`
`Medium-Dependence of the Secondary Structure of
`Exendin-4 and Glucagon-like-peptide-1
`
`Niels H. Andersen,a,* Yan Brodsky,a Jonathan W. Neidigha and Kathryn S. Prickettb
`aDepartment of Chemistry, University of Washington, Seattle, WA 98195, USA
`bAmylin Pharmaceuticals Inc., 9373 Towne Centre Drive, San Diego, CA 92121, USA
`
`Received 25 April 2001; accepted 2 July 2001
`
`Abstract—Exendin-4 is a natural, 39-residue peptide first isolated from the salivary secretions of a Gila Monster (Heloderma sus-
`pectum) that has some pharmacological properties similar to glucagon-like-peptide-1 (GLP-1). This paper reports differences in the
`structural preferences of these two peptides. For GLP-1 in aqueous buffer (pH 3.5 or 5.9), the concentration dependence of circular
`dichroism spectra suggests that substantial helicity results only as a consequence of helix bundle formation. In contrast, exendin-4 is
`significantly helical in aqueous buffer even at the lowest concentration examined (2.3 mM). The pH dependence of the helical signal
`for exendin-4 indicates that helicity is enhanced by a more favorable sequence alignment of oppositely charged sidechains. Both
`peptides become more helical upon addition of either lipid micelles or fluoroalcohols. The stabilities of the helices were assessed
`from the thermal gradient of ellipticity (@[y]221/@T values); on this basis, the exendin helix does not melt appreciably until tem-
`peratures significantly above ambient. The extent of helix formation for exendin-4 in aqueous buffer (and the thermal stability of
`the resulting helix) suggests the presence of a stable helix-capping interaction which was localized to the C-terminal segment by
`NMR studies of NH exchange protection. Solvent effects on the thermal stability of the helix indicate that the C-terminal capping
`interaction is hydrophobic in nature. The absence of this C-capping interaction and the presence of a flexible, helix-destabilizing
`glycine at residue 16 in GLP-1 are the likely causes of the greater fragility of the monomeric helical state of GLP-1. The intramo-
`lecular hydrophobic clustering in exendin-4 also appears to decrease the extent of helical aggregate formation. # 2001 Elsevier
`Science Ltd. All rights reserved.
`
`Introduction
`
`Diabetes mellitus is a chronic disease characterized by
`multiple metabolic abnormalities arising primarily from
`an inadequate insulin effect. Glucagon-like peptide-1
`(GLP-1),1 a mammalian hormone, has been investigated
`as a potential therapeutic agent for the treatment of this
`disease because of its favorable spectrum of antidiabetic
`actions which include a glucose dependent insulino-
`tropic action,2,3 an effect to modulate gastric emptying4
`and a possible role in appetite control.5 GLP-1 mimics
`also have potential as an obesity treatment.6 Due to its
`short duration of action, however, GLP-1 may have
`limitations as a therapeutic agent.
`
`Exendin-4 is a natural, 39-residue peptide7 that displays
`greater than 50% sequence identity to GLP-1. Exendin-
`4 has been shown to bind and act as an agonist at the
`
`*Corresponding author. Tel.: +1-206-543-7099; e-mail: andersen@
`chem.washington.edu
`
`GLP-1 receptor on prepared RINm5f cell membranes8
`and to have activities known to be important
`for
`improvement of glucose control, including stimulation
`of insulin secretion and modulation of gastric emptying.
`AC2993 (synthetic exendin-4) has been shown to be
`markedly more potent and/or long-lived in vivo than
`GLP-1 for these activities.9 AC2993 is currently in
`phase 2 clinical trials as a treatment for type 2 diabetes.
`GLP-1 and exendin-4 also have significant homology to
`glucagon. The sequences are shown below.
`
`The literature contains numerous structural studies of
`glucagon, a few of GLP-1, but none for exendin-4. Only
`the helical states of glucagon have been well character-
`ized by, for example, X-ray crystallography10 and NMR
`in the micelle-associated state.11 We and others12 have
`examined glucagon by circular dichroism (CD). The CD
`studies reveal a solvent and concentration dependent
`
`0968-0896/02/$ - see front matter # 2001 Elsevier Science Ltd. All rights reserved.
`P I I : S 0 9 6 8 - 0 8 9 6 ( 0 1 ) 0 0 2 6 3 - 2
`
`MYLAN INST. EXHIBIT 1092 PAGE 1
`
`MYLAN INST. EXHIBIT 1092 PAGE 1
`
`
`
`80
`
`N. H. Andersen et al. / Bioorg. Med. Chem. 10 (2002) 79–85
`
`and is thus expected to be fully reflected in this data but
`the data should be free of contributions from slowly
`formed aggregates. The concentration dependence of
`[y]221 in aqueous buffer (Table 1A) suggests that the
`modest helicity displayed by GLP-1 is due to coiled coil
`(or other helix bundle) formation. At the lowest con-
`centration examined, the pH 5.9 CD spectrum of GLP-1
`suggests a very small helicity for the monomeric state,
`even though coiled coil formation occurs readily at
`higher concentrations at this pH. For exendin-4, there is
`only a modest increase in [y]221 with concentration at
`pH 5.9. If significant oligomer formation occurs at
`higher concentrations, the CD spectrum of the resulting
`bundle state must be very similar to that of the mono-
`meric species. At pH 3.5, some helicity enhancement by
`oligomerization is evident (Table 1A). Exendin-4 dis-
`plays substantial helicity in aqueous pH 5.9 buffer even
`at 2.3 mM peptide concentrations. At pH 4.5–5.9, exen-
`din-4 appears to favor a monomeric helical state
`2–230 mM concentration
`throughout
`the
`range,
`although some degree of helical aggregate formation
`may occur at the highest concentration.
`
`in
`GLP-1 and exendin-4 both become more helical
`media containing lipid micelles (Table 1B) and the con-
`centration dependence of [y]221 disappears indicating
`that the peptides are monomeric under these conditions.
`A further increase in helicity occurs (particularly for
`GLP-1) in aqueous fluoroalcohol media (Table 1C). We
`assumed a common residue 7–28 helical span for both
`peptides (the largest possible one based on helix predic-
`tion algorithms,15,16 vide infra) to convert the CD data
`to fractional helix populations (whelix). These values
`appear in the last column of Table 1. In aqueous fluor-
`oalcohol media, other portions of both sequences must
`either become helical or structured in some other way
`that contributes to the negative [y]221 value (by type I/
`III turn formation, for example). This is most notable
`for exendin-4, which has a higher proportion of its total
`sequence outside of the predicted helical span.
`
`conformational equilibrium with significant helicity in
`the micelle-associated state; however, glucagon is only
`ca. 15% helical in water and forms ‘soluble’ b sheet
`aggregates with time at concentrations 100 mM. GLP-
`1 has been examined by NMR in aqueous media con-
`taining dodecylphosphocholine (DPC) micelles;13 the
`resulting NMR structure ensemble is helical from resi-
`due 7 to 29 with distortions in backbone helix geometry
`near Gly16. The authors suggested that this helix dis-
`tortion could be important for both membrane and
`receptor binding. Glucagon, which otherwise displays
`significant homology to GLP-1, lacks the flexible linker
`induced by Gly16 and displays diminished binding to
`GLP-1 receptors.3 However, exendin-4 also lacks the
`flexible linker, having MEEE in place of the central
`LEGQ segment of GLP-1, and yet has potent anti-
`diabetic actions. Parker et al.14 have presented NMR
`structures and potency data for covalently cross-linked
`GLP-1 analogues, suggesting that the C-terminal helix
`of GLP-1 is essential for the display of pharmacophore
`components. Based on this, we have compared the
`structuring preferences of GLP-1 and exendin-4; CD
`studies in both the solution state and in the presence
`of lipid micelles are presented here as well as a pre-
`liminary NMR determination of the helix stability of
`exendin-4.
`
`Results
`At peptide concentrations > 40 mM, the CD spectra of
`both GLP-1 and exendin-4 in aqueous buffer were time
`dependent. However, we did not observe, at any time
`point or concentration, a b sheet CD signature as we
`had previously observed for glucagon at high con-
`centrations in the absence of fluoroalcohol or lipid
`additives. The aggregation observed with GLP-1 (and to
`a lesser extent with exendin-4) does not appear to
`represent the formation of b sheet species.
`
`At pH 5.9, the helix signature of exendin-4 diminishes
`slightly over a 2–4 day period. At pH 3.5, there is a 33%
`drop in the [y]221 value over 24 h. Even after these
`losses in net [y]221 signal, the CD traces show no sig-
`nificant change in the characteristic intensity ratios of
`the three extrema (l=191, 208, 221 nm). These obser-
`vations suggest the formation of non-soluble aggregates
`that adhere to the walls of the glass vials used to store
`the CD samples during the stability studies. Irreversible
`aggregation is less of a problem with GLP-1 at pH 3.5.
`With GLP-1, irreversible aggregation and adsorption is
`observed at pH 5.9 but at a slower rate than observed
`for exendin-4; a six-day period is required for a com-
`parable reduction of the CD signal. The CD spectra of
`GLP-1 and exendin-4 in aqueous buffer at concentra-
`tions where aggregation is not present and the spectrum
`of the helical state of exendin-4 formed upon fluor-
`oalcohol addition are shown in Figure 1.
`
`Turning to CD data recorded immediately after peptide
`dissolution, Table 1 summarizes the [y]221 values for
`exendin-4 and GLP-1 in all media examined. Coiled coil
`formation should occur on a minutes-or-faster timescale
`
`Figure 1. Representative far UV CD spectra at 21 C: listed in order of
`increasingly negative values of [y]221: a) least helical, dotted line (. . . .),
`3.15 mM GLP-1 in pH 5.9 buffer; b) dashed line (- - - - -), 2.3 mM
`exendin-4 in pH 5.9 buffer; and c) most helical, solid line (——–),
`28.5 mM exendin-4 in 25% HFIP.
`
`MYLAN INST. EXHIBIT 1092 PAGE 2
`
`MYLAN INST. EXHIBIT 1092 PAGE 2
`
`
`
`N. H. Andersen et al. / Bioorg. Med. Chem. 10 (2002) 79–85
`
`81
`
`Curve shape analysis, as expressed in ellipticity ratios
`(R1 and R2), has emerged as an alternative measure of
`fractional helicity.17 These measures (appearing in Table
`2) confirm the %-helicities obtained from the absolute
`[y]221 values. At 0 C, the ellipticity ratios observed for
`exendin-4 in 25% HFIP or 40% TFE are within the
`expectation range for 100% helicity suggesting that the
`non-structured portions of the sequence are making lit-
`tle net contribution to the far UV CD signal. This pro-
`vides additional support for using [y]221 values as a
`measure of the fractional helicity over the helical span.
`
`Although there is still some controversy regarding the
`[y]221 value that corresponds to 100% helicity at high
`temperatures, particularly when fluoroalcohols are pre-
`sent,18 the thermal gradient of [y]221 can still serve as a
`measure of helix stability. Two representative melting
`studies appear in Figure 2. We adopted the %-loss of
` [y]221 signal on warming from 0 to 20 C as the
`measure of the thermal stability of the GLP-1 and
`exendin helices in different media, Table 3. Exendin-4
`displays less thermal helicity loss than GLP-1 under all
`conditions except in 25% HFIP. This is particularly
`notable in aqueous buffer (8.8% loss at pH 3.5) and in
`8% HFIP (6% helicity loss).
`
`values of 5%19 to 11%18 represent no net loss of heli-
`city. By any of these criteria, the exendin helix appears
`to be unusually stable. In contrast, the limited helicity of
`GLP-1 melts out rapidly on warming. At pH 5.9, the
`helix loss at 50 mM is largely attributable to disaggrega-
`tion. The ready thermal
`loss of the limited helicity
`observed at pH 3.5 is attributed to GLP-1 monomers.
`
`Additional evidence for the stability of the exendin-4
`helix comes from an NMR study that revealed surpris-
`ingly large exchange protection factors for the backbone
`NHs of Ile23, Glu24, Trp25 and Leu26. The NOESY
`spectrum panel in Figure 3 illustrates this. The exendin-
`4 sample examined had spent more than 16 h in 15 mM
`phosphate buffered D2O (pH* 5.3) at 21 C before the
`final lyophilization prior to NMR sample preparation.
`The NH resonances of the listed residues displayed 0.4–
`0.8+ fractional proton occupancy while all of
`the
`remaining NHs were essentially absent from the spec-
`trum. This degree of exchange protection requires frac-
`tional helicities in excess of 0.97 in the two C-terminal
`turns of the helix.
`
`Discussion
`
`To place these signal losses in context, we note that
`helical proteins display [y]221
`signal
`losses
`for
` T=20 C as small as 2.8% below their Tm values.
`Literature values for the intrinsic gradient for 100%
`helicity for peptides in water imply that signal
`loss
`
`Based on the time-dependent CD data neither GLP-1
`nor exendin-4 form b aggregates as is routinely observed
`for glucagon in aqueous buffer at high concentrations.
`Significant helix formation for GLP-1 in water is largely
`the result of coiled coil (or other helix bundle) forma-
`
`Table 1. Circular dichroism data for exendin-4 and GLP-1
`
`Conditions
`
`Conc. (mM)
`
` [y]221
`
`whelix
`
`A
`
`B
`
`Aqueous buffer, pH 3.5
`Aqueous Buffer, pH 3.5
`Aqueous Buffer, pH 3.5
`Aqueous Buffer, pH 5.9
`Aqueous Buffer, pH 5.9
`Aqueous, pH 4.5
`
`13 mM SDS, pH 3.5
`13 mM SDS, pH 5.9
`30 mM octyl-Glu, pH 3.5
`30 mM octyl-Glu, pH 5.9
`
`40 mM C12-PC, pH 3.5
`40 mM C12-PC, pH 3.5
`40 mM C12-PC, pH 5.9
`40 mM C12-PC, pH 5.9
`
`Exendin-4
`
`2.3
`28.5
`
`2.3
`28.5
`228
`
`28.5
`28.5
`Not soln
`Not soln
`
`28.5
`200
`28.5
`200
`
`GLP-1
`
`3.15
`50
`315
`3.15
`50
`
`50
`50
`50
`50
`
`50
`270
`50
`270
`
`Exendin-4
`
`GLP-1
`
`Exendin-4
`
`GLP-1
`
`11,370
`14,900
`
`14,500
`16,100
`17,600
`
`17,000a
`16,200a
`
`18,100
`18,000
`17,900
`19,200
`
`5700
`6560
`8900
`3670
`9980
`
`23,700
`19,700
`24,850
`20,670
`
`20,200
`19,800
`17,050
`18,050
`
`0.55
`0.70
`
`0.69
`0.76
`0.83
`
`0.83
`0.79
`
`0.85
`0.84
`0.84
`0.90
`
`0.23
`0.26
`0.34
`0.15
`0.38
`
`0.85
`0.71
`0.89
`0.75
`
`0.73
`0.72
`0.62
`0.65
`
`C
`
`D
`
`28.5
`28.5
`28.5
`
`50
`50
`50
`
`20,000
`23,700
`25,500
`
`25,650
`28,880
`32,560
`
`0.93
`1.10b
`> 1.1b,c
`
`0.92
`1.03b
`> 1.1b,c
`
`8% HFIP, pH 3.5
`40% TFE, pH 3.5
`25% HFIP, pH 3.5
`Calcd for whelix=1.00
`28,000
`21,500
` 830
` 800
`Calcd for whelix=0.00
`The [y]221 values for exendin-4 and GLP-1 in the various types of media (A, B, C) employed in this study are presented (at room temperature,
`21 2 C, except as noted). The calculation of expectation values (D) for the helical and coil states of exendin-4 and GLP-1 employed formula and
`corrections for aromatic chromophores16 in calculating yC and yH values at 0 C. A common helical segment, residues 7–28, was assumed along with
`helix and coil temperature gradients of 65/C and +65/C at l=221 nm, respectively, for calculating the fractional helicity (whelix) over that
`span. Instances of large whelix differences between exendin-4 and GLP-1 are highlighted in bold.
`aExperiments performed at 37 C to improve solubility.
`bSince the helix population calculations assumed a common 7–28 helical segment, calculated values greater than 1.00 suggests that helicity can be
`induced in the C-terminal proline-rich segment of exendin-4 at high fluoroalcohol levels.
`cIn 25% HFIP, both peptides display [y]221 values that are significantly greater than those expected for a residue 7–28 helical span (given in D).
`Apparently HFIP induces additional helicity (or helix-like turns) in the N-terminal portions of both sequences.
`
`MYLAN INST. EXHIBIT 1092 PAGE 3
`
`MYLAN INST. EXHIBIT 1092 PAGE 3
`
`
`
`82
`
`N. H. Andersen et al. / Bioorg. Med. Chem. 10 (2002) 79–85
`
`Table 2. Ellipticity ratios for exendin-4 and GLP-1 in different media
`
`Conditions
`
`Conc. (mM)
`
`R1a
`
`R2a
`
`Exendin-4
`
`GLP-1
`
`Aqueous buffer, pH 3.5
`Aqueous buffer, pH 5.9
`
`40 mM C12-PC, pH 3.5
`40 mM C12-PC, pH 5.9
`
`28.5
`28.5
`
`28.5
`28.5
`
`50
`50
`
`50
`50
`
`50
`50
`
`Exendin-4
` 1.73
` 1.82
` 1.60
` 1.67
` 2.21
`+0.97
`40% TFE or 25% HFIP
`28.5
`+1.05 0.02
` 2.34 0.04
`same media at 0 C
`28.5
`+1.06 0.05
` 2.45 0.25
`whelix=1.00 Expectation values
` 0.03 0.05
`+0.65 0.15
`whelix=0.00 Expectation values
`The ratios are at 21 C, except as noted. The last entries are the expectation values for 0 and 100% helicity. The peptide and conditions are listed for
`the other entries.
`aThe ellipticity ratios employed are defined here. R1=[y]max/[y]min and R2=[y]221/[y]min, where [y]min refers to the minimum at 197–210 nm and
`[y]max is the most positive ellipticity observed in the 190–195 nm span..17
`bThe dramatic increase toward helical values at pH 5.9 is rationalized as the result of coiled coil formation which is enhanced with increasing pH.
`
`GLP-1
` 0.10
` 1.01b
` 2.07
` 2.08
` 1.92
` 2.07
`
`Exendin-4
`
`+0.79
`+0.86
`
`+0.72
`+0.73
`
`GLP-1
`
`+0.43
`+0.68b
`
`+0.87
`+0.85
`
`+0.85
`+0.90
`
`monomeric state. Since both helix disruption by glycine
`and helix stabilization by coulombic and hydrophobic
`interactions should be modeled by current helicity pre-
`diction algorithms, we employed AGADIR20 and
`Helix1.516,21 to derive predicted helicity profiles for
`these peptides (Fig. 4).
`
`Both algorithms predict that the N-terminal residues of
`exendin-4 and GLP-1 are unstructured in solution even
`though they are required for pharmacological activity.22
`The lesser net helicity of GLP-1 is clearly evident in the
`predicted helicity profiles from both algorithms. The
`two algorithms predict quite different N-capping effects,
`but in each case the decreased helix propensity of GLP-
`1 can be attributed to Gly.16 These parameterization
`differences are also reflected in the predicted histograms
`of helicity along the sequence of exendin-4; the AGA-
`DIR algorithm appears
`to weight coulombic and
`hydrophobic side-chain interactions more heavily and
`predicts greater helicity than Helix1.5, particularly for
`the C-terminal segment. The AGADIR prediction is
`more nearly in accord with the experimental determina-
`tion (0.61 over residues 7–28 vs 0.69 as observed by
`CD). However, AGADIR fails to rationalize the locali-
`zation of NH exchange protection to the extreme C-
`terminus (residues 23–26, as evidenced by Figure 3). The
`AGADIR algorithm predicts residue fractional heli-
`cities between 0.825 and 0.872 from Glu17 through
`Leu26 at the temperature of CD data reported in Table
`1. The same temperature was employed for the D2O
`exchange study. Based on the AGADIR helicity profile,
`comparable NH protection would be expected from
`Val19 to Leu26; this is not observed. Furthermore, the
`degree of exchange protection observed for Ile23–Leu26
`at 293 K requires fractional helicities of 0.97 or greater
`at residues 21–24.
`
`The folding cooperativity and thermal stability of the
`exendin-4 helix in aqueous buffer suggests the presence
`of a stable helix-capping interaction. The thermal stabi-
`lity is even more pronounced in 8 vol% HFIP suggesting
`that the helix stabilization is a hydrophobic effect.23
`Since the only significant polar/apolar sequence differ-
`ences between exendin-4 and GLP-1 are the hydrophobic
`
`Figure 2. Thermal melting curves for exendin-4 and GLP-1 in aqueous
`buffer: circular dichroism ([y]221) versus temperature. Freshly prepared
`stock solutions were diluted to 28.5 mM (exendin-4) and 50 mM (GLP-
`1) with phosphate/acetate buffer of the pH indicated in the legend in
`the figure inset. The room temperature [y]221 values were recovered
`upon cooling from the highest temperature examined.
`
`tion. However, all of the evidence presented suggests
`that exendin-4, unlike GLP-1 or glucagon, forms a
`monomeric helical state in aqueous media. While exen-
`din-4 may still have some tendency to aggregate at
`higher concentrations, aggregation is not required for
`helix formation. The exendin sequence results in a better
`alignment of alternating negative and positive side-
`chains on one face, which can stabilize the helix via salt-
`bridging or polar H-bonds. For exendin-4, monomeric
`helix stabilization is clearly enhanced by glutamate
`ionization as evidenced by the pH effect on [y]221 (see
`Table 1A). This stabilizing feature continues to con-
`tribute in the DPC micelle-associated state, but to a
`lesser degree. In fluoroalcohol-containing media, the pH
`differences are virtually absent (data not shown). The
`diminished helicity of GLP-1 in the DPC micelle system
`relative to that of exendin-4 may be attributable to helix
`disruption at the Gly16 locus of GLP-1. However glu-
`cagon, which lacks a helix disrupting residue and has a
`better sequence alignment for salt bridging (see Fig. 5,
`vide infra), is also less helical than exendin-4 in the
`
`MYLAN INST. EXHIBIT 1092 PAGE 4
`
`MYLAN INST. EXHIBIT 1092 PAGE 4
`
`
`
`N. H. Andersen et al. / Bioorg. Med. Chem. 10 (2002) 79–85
`Table 3. Helix melting as measured by %-loss of [y]221 signal on
`warming from 0 to 20 C
`
`83
`
`Medium
`
`Aqueous buffer, pH 3.5
`Aqueous buffer, pH 5.9
`
`8% HFIP
`25% HFIP
`
`40% TFE
`
`GLP-1
`
`%-loss
`
`30
`22
`
`12
`10.3
`
`12.5
`
`Exendin-4
`
`%-loss
`
`8.8
`12.7
`
`6.0
`10.6
`
`11.0
`
`side-chain at position 21 and the non-helical proline-
`rich C-terminal extension present only in exendin-4, the
`source of the unexpected helix stabilization should be
`sought in this region. NMR studies of single site muta-
`tions in this sequence and of the structuring preferences
`associated with the introduction of the C-terminal seg-
`ment of exendin are in progress.
`
`Does this study increase our understanding of the fea-
`tures required for in vitro GLP-1 receptor binding and
`activation? For peptides with conformational versati-
`lity, there is always reason to doubt that any particular
`structural feature or structuring preference observed in
`
`any state other than the receptor-bound state can be
`used to derive the requisites for receptor binding and
`activation. In the present case, however, the C-terminal
`portion of a helix has now been observed for the DPC
`micelle-associated states of GLP-1 and exendin-4, in the
`solution-state for exendin-4, and for active covalently
`constrained GLP-1 analogues.14 DPC micelles may
`mimic the membrane context of the peptides just prior
`to binding to the receptor. The greater intrinsic stability
`of the exendin helix should reduce the entropic cost of
`binding at a receptor that requires the C-terminal helix
`state,14 thus increasing the affinity for the receptor. The
`appropriately displayed residues of the C-terminal helix
`likely represent receptor binding requirements. The N-
`terminal residues including His1 and Phe6 are known to
`be essential for agonist activity and the spatial relation-
`ship between these signaling requirements and the C-
`terminal binding unit is more difficult to assess.
`
`Amphiphilic character and the alignment of the aryl
`sidechains—F6, Y13 (not present in exendin-4), F22,
`and W25—on a single helix face has been proposed13 as
`a key feature for membrane binding. The absence of this
`feature (in GLP-1, glucagon and exendin-4) in a helical
`structure spanning this entire segment is evident in helix
`wheel depictions of these peptides (Fig. 5). The present
`study reveals that the helix disrupting effect of Gly16 is
`even greater than previously suggested;13 more amphi-
`philic conformations are certainly accessible for GLP-1.
`Helix distortions would be less likely in exendin-4 (D),
`the Gly16 to Ala mutant of GLP-1 (A), and upon dele-
`tion (B) of Gly16. The potency effects of modifications at
`Gly16 have been determined. The Gly16 to Ala mutant
`of GLP-1 binds to RINm5F cell receptor equally as well
`as GLP-124 and show only slightly diminished activity.22
`Deletion of Gly16, however, produces a 40-fold loss in
`binding affinity14 even though this change does create a
`more amphiphilic helix (B) with all of the aryl groups
`clustered together. In some support for the amphiphilic
`
`Figure 3. Panels showing the NH region of the tm=150 ms NOESY
`spectrum of exendin-4 in 30% d3-TFE/70% D2O buffer (pH*=5.9) at
`300K after extensive prior exchange in buffered 99.9+% D2O (pH*
`adjusted to 5.35). Subsequent experiments in protic media establish
`that the NHs of Val19, Leu21 and Phe22 appear at 8.06, 8.42, and
`8.32 ppm; these are entirely absent in this D2O exchanged sample.
`NOESY cross peaks that serve to confirm the assignment and establish
`that this is a helical segment are labeled and color coded by residue.
`The Phe22-Ha (4.292 ppm) and Trp25-Ha (4.281 ppm) signals are
`nearly shift coincident; Arg20-Ha and Glu24-Ha are shift coincident
`(d=4.014 ppm). A weak 22Ha/23HN NOESY peak appears when the
`spectrum is plotted with a lower contour level cut-off. No additional
`cross-peaks to NH frequencies were observed, even at lower contour
`levels.
`
`Figure 4. Comparison of the residue histograms of predicted helicity
`at pH 6 (293K) for exendin-4 and GLP-1 using the AGADIR and
`Helix1.5 algorithms. The more completely parameterized AGADIR
`algorithm predicts greater helicity for the amphiphilic exendin system.
`For comparison with Table 1, the predicted helicities for residues 7–28
`at 294 K are 0.097 for GLP-1 and 0.60 for exendin-4 using the AGA-
`DIR algorithm.
`
`MYLAN INST. EXHIBIT 1092 PAGE 5
`
`MYLAN INST. EXHIBIT 1092 PAGE 5
`
`
`
`84
`
`N. H. Andersen et al. / Bioorg. Med. Chem. 10 (2002) 79–85
`
`acetate (pH 4.5) to produce stock solutions with nom-
`inal concentrations of ca. 600 mM. The concentrations
`of the stock solutions were determined by UV using the
`Trp absorbance (e278 nm=5580 cm2/mmol) and correct-
`ing for a non-zero baseline. For the preparation of
`samples in DPC micelle solutions, concentrated peptide
`stock solutions were prepared in 30 vol% TFE. The
`dilution factor
`into the aqueous micelle mixtures
`(40 mM DPC) was large, producing final TFE con-
`centrations < 2 vol%. CD spectra were recorded using a
`JASCO model J720 spectropolarimeter with a nitrogen
`flow rate of 5 L/min. The wavelength and degree ellipti-
`city scales were calibrated using a reference d–10-cam-
`phorsulfonic acid (CSA) sample assuming that the CSA
`minimum corresponds to [y]192.5= 15,600.26 Typical
`spectral accumulation parameters were a time constant
`of 0.25 s and a scan rate of 100 nm/min with a 0.2 nm
`step resolution over the range 185–270 nm with 16 scans
`averaged for each spectrum. The accumulated average
`spectra were trimmed at a dynode voltage of 650 prior
`to baseline subtraction and smoothing using the reverse
`Fourier transform procedure in the JASCO software.
`CD spectral values for peptides are expressed in units of
`residue molar ellipticity (deg cm2/residue-dmol). Final
`peptide concentrations of 2-315mM in the CD cells
`(typically of 5 to 0.1mm path length) were obtained by
`quantitative serial dilution of the freshly prepared stock
`solutions to the required levels of fluoroalcohol and
`aqueous buffer (10–20 mM phosphate).
`
`Acknowledgements
`
`Studies at the University of Washington were supported
`by a grant from Amylin Pharmaceuticals, Inc. We wish
`to thank Ved P. Srivastava (Amylin Pharmaceuticals)
`and R. Matthew Fesinmeyer (University of Washing-
`ton) for revision suggestions following critical readings
`of drafts of this manuscript.
`
`References and Notes
`
`1. Abbreviations used: the standard one and three letter sym-
`bols for amino acid residues are employed; [y]221, the residue
`molar ellipticity at 221 nm (a measure of helicity); whelix, the
`helix mole fraction averaged over the residue span that dis-
`plays some helicity based on NMR parameters or helix/coil
`transition simulations; pH*, pH meter reading for D2O-con-
`taining solutions when H2O reference buffers are employed for
`calibration; CD, circular dichroism; CSA, d-10-camphorsulfonic
`acid; DPC, dodecylphosphocholine; GLP-1, glucagon-like-pep-
`tide-1 (7- 36)-amide; HFIP, hexafluoroisopropanol; TFE, tri-
`fluoroethanol; SDS, sodium dodecylsulfate.
`2. Fehmann, H. C.; Haebener, J. F. Trends Endocrinol.
`Metab. 1992, 3, 158.
`3. Thorens, B. Proc. Natl. Acad. Sci. U.S.A. 1992, 89, 8641.
`4. (a) Wettergren, A.; Schjoldager, B.; Mortensen, P. E.;
`Myhre, J.; Christiansen, J.; Holst, J. J. Dig. Dis. Sci. 1993, 38,
`665. (b) Dupre, J.; Behme, M. T.; Hramiak, I. M.; McFarlane,
`P.; Williamson, M. P.; Zabel, P.; McDonald, T. J. Diabetes
`1995, 44, 626.
`5. Turton, M. D.; O’Shea, D.; Gunn, I.; Beak, S. A.;
`Edwards, C. M.; Meeran, K.; Choi, S. J.; Taylor, G. M.;
`
`(G16A)-GLP-1 (A),
`representations of
`Figure 5. Helix wheel
`(desG16)-GLP-1 (B), glucagon (C) and exendin-4 (D) retaining the
`same orientation relative to F22 and W25. Non-polar residues are
`underlined; aromatic residue symbols are enboxed. This representa-
`tion, which assume a perfect a-helix conformation, is unlikely to be
`appropriate for unmodified GLP-1 even in a membrane-mimicking
`micelle system.
`
`helix hypothesis, the E/D21L change versus GLP-1 and
`glucagon (and the R18A change vs glucagon) does
`result in a more uniformly hydrophobic face in exendin-
`4. However, at this point the only feature of the receptor
`bound state that can be confidently predicted is the C-
`terminal helix. Hypotheses concerning the spatial rela-
`tionship of the C-terminal helix to signaling requisites
`near the N-terminus would be speculation at this point.
`
`Experimental
`
`Exendin-4 and GLP-1 were prepared by standard solid-
`phase
`peptide
`synthesis protocols
`(Amylin lots
`#CS171297 and #P180597, respectively). Amino acid
`compositions and purity were confirmed by amino acid
`analyses (average error in residue count < 6.1 5.9% >
`for Glx, Asx and all residues other than Trp, < 1.7% >
`for Asx/Glx/Pro/Gly/Met/Leu/Lys/His) and HPLC
`( > 98% on Vydac C18). The average molecular weights
`observed by MS were within 0.6 amu of the expectation
`based on the sequence. NMR analyses25 have confirmed
`the sequence and purity of the peptides. The results of
`one NMR experiment, an attempt to examine exendin-4
`in D2O/TFE medium after complete exchange of back-
`bone NH signals, is pertinent to the present study. A
`sample of exendin-4 was dissolved in 10:1 phosphate
`buffered D2O/TFE (peptide concentration 700 mM,
`15 mM phosphate) with pH* adjusted to 5.35. Repeated
`lyophilizations (with addition of 99.9+% D2O buffer
`each time) over ca. 3 days were performed prior to pre-
`paration of the NMR sample (3:7 d3-TFE/D2O-buffer,
`pD 5.9).
`
`CD samples were prepared by dissolving weighed
`amounts (0.5–2 mg) of freshly vacuum-dried peptides
`directly in 15 mM aqueous phosphate (pH 3.5 or 5.9) or
`
`MYLAN INST. EXHIBIT 1092 PAGE 6
`
`MYLAN INST. EXHIBIT 1092 PAGE 6
`
`
`
`N. H. Andersen et al. / Bioorg. Med. Chem. 10 (2002) 79–85
`
`85
`
`Heath, M. M.; Lambert, P. D.; Wilding, J. P.; Smith, D. M.;
`Ghatei, M. A.; Herbert, J.; Bloom, S. R. Nature 1996, 379, 69.
`6. Bray, G. A.; Tartaglia, L. A. Nature 2000, 404, 672.
`7. Eng, J.; Kleinman, W. A.; Singh, L.; Singh, G.; Raufman,
`J. P. J. Biol. Chem. 1992, 267, 7402.
`8. Go¨ ke, R.; Fehmann, H. C.; Linn, T.; Schmidt, H.; Krause,
`M.; Eng, J.; Go¨ ke, B. J. Biol. Chem. 1993, 268, 19650.
`9. Young, A. A.; Gedulin, B. R.; Bhavsar, S.; Bodkin, N.;
`Jodka, C.; Hansen, B.; Denaro, M. Diabetes 1999, 48, 1026.
`10. Sasaki, K.; Dockerill, S.; Adamiak, D. A.; Tickle, I. J.;
`Blundell, T. Nature 1975, 257, 751.
`11. Braun, W.; Wider, G.; Lee, K. H.; Wu¨ thrich, K. J. Mol.
`Bio. 1983, 169, 921.
`12. Epand, R. M.; Jones, A. J.; Sayer, B. Biochemistry 1977,
`16, 4360.
`13. Thornton, K.; Gorenstein, D. G. Biochemistry 1994, 33,
`3532.
`14. Parker, J. C.; Andrews, K. M.; Rescek, D. M.; Massefski,
`W., Jr.; Andrews, G. C.; Contillo, L. G.; Stevenson, R. W.;
`Singleton, D. H.; Suleske, R. T. J. Pept. Res. 1998, 52, 398.
`15. (a) Mun˜ oz, V.; Serrano, L. Nat. Struct. Bio. 1994, 1, 399.
`(a) Doig, A. J.; Baldwin, R. L. Prot. Sci. 1995, 4, 1325.
`16. Andersen, N. H.; Tong, H. Prot. Sci. 1997, 6, 1920.
`
`17. (a) Bruch, M. D.; Dhingra, M. M.; Gierasch, L. M. Pro-
`teins 1991, 10, 130. (b) Mun˜ oz, V.; Serrano, L.; Jime´ nez, M. A.;
`Rico, M. J. Mol. Bio. 1995, 247, 648.
`18. Luo, P.; Baldwin, R. L. Biochemistry 1997, 36, 8413.
`19. Shalongo, W.; Dugad, L.; Stellwagen, E. J. Am. Chem.
`Soc. 1994, 116, 2500.
`20. Mun˜ oz, V.; Serrano, L. Biopolymers 1997, 41, 495 Avail-
`able at http://www.embl-heidelberg.de/Services/serrano/aga-
`dir/agadir-start.html
`21. Available at http://faculty.washington.edu/nielshan/
`22. Adelhorst, K.; Hedegaard, B. B.; Knudsen, L. B.; Kirk, O.
`J. Biol. Chem. 1994, 269, 6275.
`23. (a) Andersen, N. H.; Cort, J. R.; Liu, Z.; Sjoberg, S. J.;
`Tong, H. J. Am. Chem. Soc. 1996, 118, 10309. (b) Andersen,
`N. H.; Dyer, R. B.; Fesinmeyer, R. M.; Gai, F.; Liu, Z. H.;
`Neidigh, J. W.; Tong, H. J. Am. Chem. Soc. 1999, 121, 9879.
`24. Gallwitz, B.; Witt, M.; Paetzold, G.; Morys-Wortmann,
`C.; Zimmermann, B.; Eckart, K.; Folsch, U. R.; Schmidt,
`W. E. Eur. J. Biochem. 1994, 225, 1151.
`25. Neidigh, J. W. Ph.D.
`in Chemistry; University of
`Washington: Seattle, WA, 1999, pp 226.
`26. Yang, J. T.; Wu, C. S.; Martinez, H. M. Methods Enzy-
`mol. 1986, 130, 208.
`
`MYLAN INST. EXHIBIT 1092 PAGE 7
`
`MYLAN INST. EXHIBIT 1092 PAGE 7
`
`