doi:10.1006/jmbi.2001.4608 available online at http://www.idealibrary.com on
`
`J. Mol. Biol. (2001) 308, 783794
`
`Identification of Long-range
`Islet Amyloid Polypeptide:
`Contacts and Local Order on the
`Fibrillogenesis Pathway
`Shae B. Padric k and Andre w D. Miranker *
`
`Departmentof Molecular
`Biophysicsand Biochemistry
`YaleUniversity, 266 Whitney
`Avenue,PO Box208114,New
`Haven,CT 06520,USA
`
`*Correspondingauthor
`
`Introduction
`
`The pathology of type II diabetes includes deposition of amyloid in the
`extra cellular space surrounding the b-cells of
`the endocrine pancreas.
`The principle component of thesedeposits is an insoluble fibrillar form of
`a normally soluble 37 residue peptide hormone, islet amyloid polypep-
`tide. Multiple sequenceanalysis and peptide synthesis have identified a
`core set of residues (20 to 29) as intrinsically amyloidogenic. As the fibril-
`logenesisof the 20-29 peptide often requires conditions that deviate con-
`siderably from physiological, residues 20 to 29 may be necessary,but not
`sufficient, for amyloidosis. We aim to determine the structural role of
`residues outside this core in the context of in vitro fibrillogenesis of the
`wild-type peptide at physiological pH and ionic strength. Specifically, we
`make use of an intrinsic fluorescent probe, tyrosine 37 (Y37), to explore
`the role of the C terminus in fibrillogenesis. Our protocol permits steady
`state measurement of the lag phase and fiber conformational statesof the
`protein under identical conditions. Theseare compared to a non-amyloi-
`dogenic variant of islet amyloid polypeptide from rat and N-acetyl-tyrosi-
`namide as models of
`the unfolded state under matched conditions.
`Spectral, quenching and anisotropic properties of Y37 in the fiber state
`indicate that
`the C terminus is packed in a well-defined environment
`with near frozen rigidity. The presence of a fluorescence resonance
`energy transfer pathway shows Y37 is near F15 and F23. The lag-phase
`conformation, while considerably less ordered than the fiber,
`is more
`ordered than unfolded models. Differences in anisotropy between the lag
`and fiber state were used to monitor fibrillogenesis in real time. Parallel
`assessmentof fiber formation using the histological dye, ThT, indicate
`that ordering at the C terminus of islet amyloid polypeptide is coincident
`with, and thus indicative of, fiber formation.
`
`# 2001Academic Press
`Keywords:IAPP; tyrosine fluorescence;type II diabetes;amyloid; amylin
`
`The deposition of normally soluble protein as
`amyloid fiber is central to a number of diseases
`including,
`for example, Alzheimer s and dialysis
`
`Abbreviations used: FRET,fluorescenceresonance
`energy transfer; IAPP, islet amyloid polypeptide;
`hIAPP, human IAPP; hIAPPlag, the lag phase
`conformation of hIAPP; hIAPPfib , the fibrillar
`conformation of hIAPP; rIAPP, rat IAPP;
`HFIP, 1,1,1,3,3,3-hexafluoroisopropanol;NAYA,
`N-acetyl tyrosinamide; nMAC, N-methyl acetamide;
`ThT, Thioflavin T.
`E-mail address of the corresponding author:
`Andrew.Miranker@yale.edu
`
`related amyloidosis.1 This problem is particularly
`fascinating as proteins that are apparently unre-
`lated in sequenceand in their native conformations
`nevertheless aggregate into fibrils with common
`structural and histological features.2 These features
`include a crossed-b sheetorganization in which the
`b-strands are arranged perpendicular to the fibril
`axis, and the display of green birefringence upon
`binding of the dye Congo Red. Amyloid fiber for-
`mation kinetics also share common features.3
`Fibrillogenesis reactions are characterized by a lag
`phase during which no fibers are observed. This is
`then followed by a period of fiber formation on a
`timescale that can be shorter than the lag phase.
`Furthermore, the lag phase can often be bypassed
`by nucleating a reaction with preformed fibers.
`
`0022-2836/01/040783 12$35.00/0
`
`# 2001Academic Press
`
`Novo Nordisk Ex. 2038, P. 1
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`784
`
`Structure Formation during Fibrillogenesis of IAPP
`
`type II diabetics have
`Greater than 90 % of
`amyloid deposits surrounding the islet cells of the
`pancreas.4 These deposits are predominantl y made
`of
`fibers composed of
`islet amyloid polypeptide
`(IAPP). IAPP is a 37-residue peptide which is co-
`secreted with insulin by the b cells. A number of
`possible metabolic roles have been suggested for
`IAPP, e.g. slowing of pyloric emptying into the
`duodenum5 and attenuation of insulin secretion by
`b cells.6 Increased insulin secretion (and therefore
`IAPP secretion) is the normal response to insulin
`resistancein type II diabetics. As IAPP fibers have
`been shown to be toxic to cultur ed b cells,7 the
`development of IAPP fibers is thought to amplify
`the diabetic state through severe depletion in the
`number of b cells available to secreteinsulin.
`The covalent structure of
`IAPP in diabetics is
`identical to that found in healthy individuals. Fur-
`thermore, in vitro fiber formation occurs at IAPP
`concentrations below the concentration found in
`the secretory granule (ca 400 mM).8 10 This implies
`that a change in the environment of
`IAPP is
`responsible for the conformational change to inso-
`luble amyloid fiber
`in diabetics. Subsets of
`the
`IAPP sequence, e.g. 20-2911 and 30-3712 have been
`characterized and shown to form fibers. Indepen-
`dent
`fiber formation by these peptides suggests
`that they may be in a similar conformation to that
`seenfor the same stretch of amino acid residues in
`the full-length fiber. We note, however, that
`the
`concentrations required for
`fiber
`formation by
`these sequences are often orders of magnitude
`higher than that required for the full-length pep-
`tide. This suggeststhat additional
`interactions are
`necessaryfor the formation and stability of wild-
`type IAPP fibers.
`Here, we have used the intrinsic aromatic resi-
`dues of
`IAPP to study conformational change
`induced by fibrillogenesis of
`the full-length pep-
`tide. As wild-type IAPP possessesa single tyrosine
`and no tryptop hans (Figure 1), we are able to
`make several specific and complementary studies
`of
`this residue. Excitation and emission profiles,
`steady state anisotropy, quenching behavior and
`resonant energy transfer between internal aro-
`matics are all used to develop a structural descrip-
`tion of IAPP prior to and after fibrillogenesis.
`
`Results
`Polymerization time for human IAPP (hIAPP) is
`strongly dependent on reaction conditions and
`procedures for preparation of stock solutions.13 For
`this work, we chosea single reaction condition that
`
`reproducibly generatesa lag phase in excessof ten
`minutes with fibrillogenesis complete in less than
`1 hr (Figure 2). We refer to the conformational
`ensemble of hIAPP before fibrillogenesis as hIAP-
`Plag and after fibrillogenesis as hIAPPfib . Our time-
`scale is determined from the midpoint,
`t1/2 ,
`for
`maximal
`fluorescence enhancement of
`the histo-
`logical dye, Thioflavin T (ThT)14 (Figure 2). The
`reaction (initiated by a 1:40 dilution of a 1 mM
`stock solution of hIAPP in HFIP, with 100 mM
`KCl, 50 mM potassium phosphate (pH 7.4) at
`25 C) yields a final protein concentration of
`25 mM. This readil y permits steady state intrinsic
`fluorescencemeasurementsto be conducted before
`and after the transition. There are three fluorescent
`residues in hIAPP: F15, F23 and Y37 (Figure 1).
`The maximum absorbance of
`tyrosine
`is
`e275 ˆ 1.4 f f i103 M 1 cm 1 whil e that of phenyl-
`alanine is e255 ˆ 0.22 f f i103 M 1 cm 1.15 Further-
`more,
`the quantum yield of
`fluorescence for
`tyrosine is approximately fourfold higher than for
`phenylalanine. We can assume,therefore, that light
`detected upon excitation at 278 nm is entirely
`derived from Y37.
`Two reference molecules are used to evaluate
`hIAPPlag and hIAPPfib under matched solution con-
`ditions. As a baseline for interpretation of fluor-
`escence,we use a tyrosine analogue. Our choice,
`N-acetyl-tyrosinamide (NAYA),
`reflects the pre-
`senceof C-terminal amidation in IAPP. As a model
`of the unfolded state, we use a sequencevariant of
`IAPP from rat (rIAPP) (Figure 1). The relevance of
`rIAPP to these studies is derived from two obser-
`vations. First, the far UV CD spectrum of rIAPP
`under our
`reaction conditions is characteristic
`of a random coil
`(data not shown).13 Second,
`although rIAPP is 84 % identical
`to hIAPP, it
`is
`non-amyloidogenic. This has been attributed to the
`presence of
`three proline residues, which are
`known to be strong b-sheet breakers.16 These two
`referencemolecules permit changesin the spectral
`properties of rIAPP versus hIAPPlag and hIAPPlag
`to be interpreted in terms of
`versus hIAPPfib
`unfolded to lag-phase and lag-phase to fiber
`structural transitions, respectively.
`
`Electronic environment of Y37
`The excitation maximum for NAYA under our
`reaction condition s is 275 nm (Figure 3(a)), exci-
`tation of which yields an emission profile with a
`maximum at 300 nm (Figure 3(b)). At wavelengths
`that are longer than their peak maxima, the exci-
`tation profile of NAYA, rIAPP, hIAPPlag are nearly
`
`hIAPP
`rIAPP
`NAYA
`
`KCNTATCATQRLAN FLVHSSNNFGAILSSTNVGSNT Y-NH2
`KCNTATCATQRLAN FLVRSSNNLGPVLPPTNVGSNT Y-NH2
` Ac- Y-NH2
`........10........20........30.......
`Figure 1. Amino acid sequence of hIAPP, rIAPP, and NAYA. NAYA is shown here schematically. Changes
`between hIAPP and rIAPP sequencesare in boxes.Intrinsic fluorescent aromatic residues are highlighted in bold.
`
`Novo Nordisk Ex. 2038, P. 2
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`Structure Formation during Fibrillogenesis of IAPP
`
`785
`
`readily seen at 250 nm. As this is 24.5nm below
`the excitation maximum for NAYA and as the
`slope of the excitation of NAYA at 250 nm can be
`seenapproaching 0, it seemsunlikely that intensity
`at this wavelength is the result of direct excitation
`of tyrosine. One possibility is that this trend is the
`result of excitation of phenylalanine. As we collect
`emission 25 nm above the emission l max phenyl-
`alanine and, since the quantum yield for phenyl-
`alanine is 1/4 that of
`tyrosine, our observations
`would reflect fluorescenceresonanceenergy trans-
`fer (FRET), not direct emission of phenylalanine.
`This can occur with measurable efficiency,
`for
`example in histone-like protein from Thermoplasma
`acidophilum,19 provi ded the distance between
`donor
`(F) and acceptor (Y)
`is less than 18 A
`(efficiency ˆ 10 %).
`To confirm that our observation is due to FRET,
`we first compared the difference spectrum of
`hIAPPfib and NAYA with the excitation spectrum
`of
`free phenylalanine (Figure 3(c)). A correspon-
`dence can clearly be seen at wavelengths below
`270 nm. Second, we were able to monitor
`the
`quenching of the phenylalanine fluorescenceby the
`presence of the acceptor, tyrosine (Figure 3(d)). By
`monitoring
`emission at 278 nm rather
`than
`303 nm, we ensured that fluorescence of phenyl-
`alanine would predominate over that of tyrosine.
`The excitation spectrum of hIAPPlag clearly shows
`a maximum at 258 nm indicating that some direct
`emission of phenylalanine fluorescencecan still be
`measured from this conformational state. Upon
`fiber formation, a distinct peak for phenylalanine
`fluorescence is no longer observed. This indicates
`complete quenching of
`the fluorescence of both
`phenylalanine residues.
`Emission profiles of
`rIAPP and hIAPPlag are
`indistinguishable from NAY A (Figure 3(b)),
`although intensities are diminished by
`80%
`(Figure 3(b), inset). This loss of intensity between
`free tyrosine and tyrosine incorporated in peptides
`is well known and largely attributed to quenching
`by the carbonyl group of peptide bonds.17,20 By
`contrast, the emission profile of hIAPPfib differs
`markedly from NAYA. A substantially greater pro-
`portion of hIAPPfib emission is at wavelengths that
`are shorter than the maximum (300 nm). While
`this is due in part to a small increasein the direct
`scatter intensity brought about by the presenceof
`fibers, scatter cannot account for the diminishment
`of
`intensity above 300 nm. Blue shifts in trypto-
`phan fluorescenceare commonly attributed to bur-
`ial of the indole moiety in the hydrophobic core of
`a protein. This does not follow for tyrosine as its
`emission is far lesssensitive to environment. Emis-
`sion l max of phenol derivatives such as NAYA,
`for
`example, is unchanged in solvents as diverse as
`water and 1,4-dioxane.17
`
`Solvent accessibility of the C terminus
`The excited state of tyrosine, as any fluorophore,
`returns to the ground state via a number of path-
`
`a)
`
`1
`
`Relative Intensity
`
`b)
`
`Relative Intensity
`
`hIAPP Fiber
`Soluble hIAPP
`
`0
`300
`
`420
`380
`340
`Wavelength (nm)
`
`460
`
`Lag-phase
`10
`
`0
`
`30
`
`40
`
`20
`Time /min
`Figure 2. Fiber formation kinetics followed by ThT
`fluorescence.(a) Fluorescenceof ThT is enhanced upon
`binding fiber. hIAPP as preformed fibers (thick line), or
`monomer stock solution in HFIP (thin line) was diluted
`to identical concentration (in monomer units) with ThT
`containing buffer. For fibers, the fluorescence excitation
`spectrum between 430 and 460 nm is dramatically
`increased. Emission is detected at 482 nm. (b) Kinetics
`followed by ThT fluorescence. Aliquots of a standard
`fiber formatio n reaction are removed at successive time
`points and quenched by dilution with ThT assay buffer.
`Excitation intensities are integrated from 430-460nm.
`
`the relative
`(Figure 3(a)). By contrast,
`identical
`intensity of excitation transitions at these longer
`wavelengths is greater for hIAPPfib . An increasein
`relative intensity at wavelengths near 287 nm has
`also been observed in absorbanceand fluorescence
`excitation measurements of the interaction of free
`tyrosine with peptide backbone mimics, such as
`N-methyl acetamid e (nMAC).17,18 These investi-
`gations have suggested that
`these lower energy
`transitions are enhanced when the tyrosine
`hydroxyl oxygen accepts a hydrogen bond. We
`made an analogous study under our reaction con-
`ditions by determining the difference spectrum of
`NAYA in the presenceand absenceof 1 M nMAC
`(Figure 3(c)). Comparing this to the difference spec-
`trum of NAYA and hIAPPfib clearly shows simi-
`larities at wavelengths above 275 nm.
`Excitation profiles at wavelengths that are short-
`er than 265 nm show a clear progression in the flu-
`orescent intensity of rIAPP, hIAPPlag and hIAPPfib
`compared to NAY A (Figure 3(a)). This is most
`
`Novo Nordisk Ex. 2038, P. 3
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`Structure Formation during Fibrillogenesis of IAPP
`
`1
`.5
`
`00
`
`hIAPPfib
`hIAPPlag
`rIAPP
`
`NAYA
`
`Relative Signal
`
`290
`
`300
`320
`310
`Emission Wavelength (nm)
`
`330
`
`hIAPPfib
`
`hIAPPlag
`
`1
`
`0.8
`
`0.6
`
`0.4
`
`0.2
`
`0
`
`b)
`
`Normalized Signal
`
`d)
`)
`
`Relative Intensityd
`
`1
`
`0.8
`
`0.6
`
`0.4
`
`0.2
`
`0
`250
`
`NAYA
`rIAPP
`hIAPPlag
`hIAPPfib
`260
`270
`280
`Excitation Wavelength (nm)
`
`290
`
`hIAPPfib - NAYA
`
`free phe
`
`NAYA in 1M nMAC - NAYA
`
`786
`
`a)
`
`Normalized Signal
`
`c)
`
`Relative Intensity
`
`250
`
`280
`270
`260
`Excitation Wavelength (nm)
`
`290
`
`230
`
`240
`250
`260
`Excitation Wavelength (nm)
`
`270
`
`Figure 3. Intrinsic fluorescenceof IAPP. (a) Normalized fluorescenceexcitation spectra for NAYA, rIAPP, hIAPPlag
`and hIAPPfib . All spectra acquired under identical conditions. Emission is observed at 303 nm. (b) Normalized fluor-
`escenceemission spectra for NAYA,
`rIAPP, hIAPPlag and hIAPPfib . All spectra acquired under identical conditions.
`Excitation wavelength is 278 nm. Curve identities are shown in (a). The spectra for NAYA, rIAPP and hIAPPlag over-
`lay. (Inset) Relative fluorescent intensity for samples in (a) and (b). Intensities measured at excitation wavelength
`275 nm and emission wavelength 300 nm. IAPP concentrations were 25 mM in all samples. NAYA concentration
`5 mM and scaled by 5. (c) Difference spectrum of hIAPPfib and NAYA shown in (a). Two distinct peaks are observed,
`one with a l max above 278 nm, the other below. An excitation spectrum of 500 mM free phenylalanine is shown (emis-
`sion observed at 278 nm), and corresponds to the blue shifted peak of hIAPPfib -NAYA. The difference spectrum of
`[NAYA in 1 M nMAC]-[NAYA in H 2O] and corresponds to the red shifted peak of hIAPPfib-NAYA. Curves have
`been offset for clarity. Latter curve was smoothed using a 2 nm rolling average. (d) Fluorescenceexcitation spectrum
`of hIAPP, detecting at 278 nm. A peak for phenylalanine is readily apparent in hIAPPlag and absent in hIAPPfib .
`
`ways, of which, only one is emission of a photon
`detectable as fluorescence. The proportion
`of
`molecules that emit a photon can be greatly dimin-
`ished by addition of a soluble quenching agent
`to the solution. Quenching is dependent, in part,
`on contact of
`the quenching agent with the
`fluorophore. This provides an effective means of
`characterizing the local environment of a fluoro-
`phore in terms of solvent accessibility, e.g. Garzon-
`Rodriguez et al.21 We have made use of
`this to
`study the conformation of
`IAPP in terms of
`the
`solvent accessibility of Y37.
`The effect of the quenching agent, acrylamide,
`was determined for NAYA,
`rIAPP, hIAPPlag and
`hIAPPfib over a quenching agent concentration
`range of 0 to 100 mM (Figure 4(a)). As expected,
`NAYA is most easily quenched with 60 % of the
`fluorescence intensity lost at 100 mM acrylamide.
`By comparison,
`the extent
`to which rIAPP is
`
`quenched by 100 mM acrylamide is smaller. The
`difference between rIAPP and NAYA most likely
`stems from the slower diffusion of rIAPP. Quench-
`ing of Y37 in hIAPPlag shows a further reduction
`over rIAPP. While the difference is small, it is con-
`sistent and is also qualitatively observed when I
`is used as the quenching agent. The difference
`is most
`readily
`between rIAPP and hIAPPlag
`explained by a conformational difference between
`the two that either affords a greater degree of sol-
`vent protection for Y37 in hIAPPlag or a signifi-
`cantly larger hydrodynamic radius for hIAPPlag.
`Conversion of hIAPPlag to a fibrillar state results
`in a dramatic reduction in the sensitivity of Y37 to
`quenching. At 100 mM acrylamide, fluorescencein
`hIAPPfib is reduced by only 39 %compared to 51 %
`in hIAPPlag. A reduction in quenching may result
`from either solvent exclusion, or a reduced
`diffusion rate. Collisional quenching is, however,
`
`Novo Nordisk Ex. 2038, P. 4
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`Structure Formation during Fibrillogenesis of IAPP
`
`787
`
`Furthermore, the fits consistently extrapolate to an
`intercept of 1( 0.03) (Figure 4(a)). This strongly
`suggeststhat only single conformers are contribut-
`ing to our observations. The slopes of the fits, KSV,
`describe the ease of quenching in a physically
`interpretable way. Our measurements were
`repeated using I
`as the quenching agent
`(Figure 4(b)). The trends observed for acrylamide
`and I are similar, although the absolute magni-
`tudes are not. The reduced degree by which I
`quenchescompared to acrylamide likely reflects its
`reduced quenching efficiency. The similarity of the
`behavior of
`the two quenching agents suggests
`that the quenching of Y37 in IAPP is the result of
`diffusion-controlled collisions of quenching agent
`with phenol groups that are in structurally homo-
`geneousenvironments.
`An independent approach to measuring solvent
`accessibility is the measurement of the pKa of the
`tyrosine. If buried or partially buried, tyrosine will
`ionize at an increased pKa.22 Tyrosinate, tyr osine s
`conjugate base, has dramatically less fluorescent
`intensity at 303 nm. Therefore, we can measurethe
`pKa of
`tyrosine by monitoring the loss of
`fluor-
`escent intensity as a function of pH (Figure 5). We
`presumed that hIAPPlag would be affected by pH
`and therefore measured the pKa of
`tyrosine in
`NAYA, rIAPP and hIAPPfib only. SubsequentThT
`analysis of hIAPPfib samplesincubated at thesepH
`values showed a <10 % variation in fiber content.
`NAYA exhibits a pKa of 9.9, which is a commonly
`observed value for tyrosine pKa. Titration of rIAPP
`yields data within error to that of NAYA (data
`now shown). hIAPPfib however shows a pKa which
`is increased by nearly two full pKa units, to a pKa
`of 11.8. We attribute this shift
`to partial hydro-
`phobic burial of Y37.
`
`Fiber
`NAYA
`
`1
`
`0.8
`
`0.6
`
`0.4
`
`0.2
`
`8
`
`Fraction of Protonated Tyrosine
`
`9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`pH
`Figure 5. Titration of Y37 in hIAPPfib . Titration of
`5 mM of hIAPPfib is observed by quenching of Y37 fluor-
`escenceintensity. Lower baseline for both NAYA and
`hIAPPfib correspond to zero intensity. hIAPPfib did not
`dissolve at high pH, as measured by ThT. 5 mM samples
`of NAYA show 50 % titration at pH 9.9, consistent with
`literature values. Curves are plots of
`the Henderson-
`Hasselbachequation for a single ionizing species,using
`pKas of 9.9 and 11.8,respectively.
`
`Figure 4. Measurement of solvent accessibility by
`quenching of
`intrinsic fluorescence. (a) Stern-Volmer
`analysis of quenching of
`IAPP fluorescence by acryl-
`amide. Data was fit
`to the equation F0/ F ˆ KSV[Q] ‡ b,
`where F is the fluorescence intensity at concentration
`[Q] of quencher. For clarity, points shown are averages
`of
`three measurements, although statistics were deter-
`mined using all points. All
`lines extrapolate back to
`b ˆ 1( 0.03) at [Q] ˆ 0 M. Pearsons R for fits of acryl-
`amide quenching are 0.997, 0.994, 0.985, and 0.944 for
`NAYA,
`rIAPP, hIAPPlag and hIAPPfib , respectively. (b)
`Values of KSV for quenching of tyrosine fluorescenceby
`acrylamide (filled bars) or iodide (open bars). Heavy
`line at KSV ˆ 14.2M 1 represents a theoretical estimate
`of acrylamide quenching (see Materials and Methods)
`for a 100 % solvent exposed tyrosine in monomeric
`IAPP being quenched by acrylamide. Error bars are the
`standard error in the slope of the fit.
`
`the fluorophore and
`to the sum of
`proportional
`quencher diffusio n rates.15 As the diffusio n rate for
`both hIAPPlag and hIAPPfib is small in comparison
`to that of acrylamide, diminished sensitivity of
`hIAPPfib to quenching reflects burial of Y37.
`Quenching can arise from either binding of
`quenching agent, or diffusion controlled collisions.
`Furthermore, multiple protein conformations may
`contribute to the observed quenching behavior. To
`address theseissues,fluorescencein the absenceof
`quenching agent (Fo) and as a function of quench-
`ing agent concentration (F) was measured. Plots of
`Fo/ F as a function of acrylamide concentration
`(Stern-Volmer analysis) show linear behavior.
`
`Novo Nordisk Ex. 2038, P. 5
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`788
`
`Structure Formation during Fibrillogenesis of IAPP
`
`The increase in fluorescence anisotropy from
`rIAPP to hIAPPlag can be explained by an increase
`in local order and/or by an oligomeric complex.
`The latter can be tested for using static light scatter.
`Detecting light scatter at 340 nm (Figure 7) we
`observe that scatter by rIAPP is marginal. By con-
`trast, hIAPPfib produces considerable scatter, pre-
`sumably as fiber length is well in excessof 340 nm.
`It is surprising that hIAPPlag also produces scatter
`intensity within a factor of
`two of that seen for
`hIAPPfib . This is consistent with hIAPPlag contain-
`ing large oligomeric complexes.
`Simultaneous collection of perpendicular and
`parallel emission permits measurement of the ani-
`sotropy in real time (Figure 8). This approach gen-
`erates a lag phase with t1/2 ˆ 16( 3) minutes and
`elongation time constant of 180( 50) seconds.This
`is consistent with our kinetics measured by mass
`spectrometry under similar conditions.24 Further-
`more, fiber formation kinetics of hIAPP25 can be
`measured indirectly using the histological dye,
`ThT.14 Changes in the fluorescence intensity of the
`dye are directly proportional
`to the quantity of
`fiber in the sample. We conducted this measure-
`ment in parallel with the measurement of anisotro-
`py. As aliquots of the reaction are quenched by
`dilution into assaybuffer containing ThT, they can
`be assayedafter anisotropy data collection is com-
`plete. The profile of
`the increase in ThT fluor-
`escenceclearly parallels the transition measured
`by anisotropy (Figure 8). This demonstrates that
`the formation of a rigid environment around
`tyrosine 37 is coincident with ThT detectable fiber
`formation.
`The standard deviations of our measurementsof
`anisotropy for NAYA,
`rIAPP and hIAPPlag are
`small (<0.01). This is also true for hIAPPfib
`(SD
`0.001),however, in reactions under similar but not
`identical solution conditions (data not shown),
`hIAPPfib anisotropy values range from 0.22to 0.26.
`
`1
`
`Relative intensity
`
`0
`
`hIAPPfib
`rIAPP
`hIAPPlag
`Figure 7. Static light scatter of rIAPP, hIAPPlag, and
`hIAPPfib. 90 static light scatter is observed from objects
`with a dimension on the order of or greater than the
`wavelength of
`light used for measurement (340 nm).
`Reported values are the average of three independent
`measurementswith buffer contribution subtracted. Error
`bars are 1SD.
`
`Rotational
`
`freedom of Y37
`
`Anisotropy of fluorescenceemission, r, is depen-
`dent on the ability of a fluorophore to rotate
`during the lifetime of its excited state. Rotational
`freedom can therefore be related to the size and
`rigidity of the atoms surrounding the fluorophore.
`As conversion of precursor to amyloid fiber
`is
`necessarilyaccompanied by an increasein molecu-
`lar weight, steady state anisotropy was measured
`(Figure 6). A clear and consistent progression of
`increasing anisotropy was observed with NAYA <
`rIAPP < hIAPPlag5 hIAPPfib .
`We assume that the development of anisotropy
`is the result of rigid, spherically shaped packing of
`protein atoms around the tyrosine side-chain. We
`further assumethat the fluorescent lifetime of tyro-
`sine, t 0, does not change in the three states of
`IAPP. The fundamental, or maximal, anisotropy
`for NAYA, ro ˆ 0.278 (l ex ˆ 280 nm), is determined
`from the anisotropy of NAY A in frozen solution.23
`Within these caveats, the rotational correlation
`time, y, can be extracted using the Perrin equation,
`(r 1 ˆ
`r0
`1 ‡ t 0/( r0y)), and the relative molecular
`volumes, which are proportional
`to y are 1:2:4:100
`for NAYA,
`rIAPP, hIAPPlag and hIAPPfib respect-
`ively (Figure 6). As many assumption s are required
`for this assessment,it
`is impossible to conclude
`more than that there is a progressive increase of
`conformational order near the phenol group in
`NAYA,
`rIAPP, and hIAPPlag. The transition of
`hIAPPlag to hIAPPfib however results in a dramatic
`is
`change. Indeed,
`the anisotropy for hIAPPfib
`nearly the same as NAYA in frozen solution. This
`strongly suggeststhat the increaseof anisotropy is
`the result of oligomeric assembly.
`
`Relative volume
`
`100
`
`10
`
`frozen NAYA
`
`0.3
`
`0.2
`
`0.1
`
`Anisotropy
`
`0
`
`1
`
`NAYA
`rIAPP
`hIAPP lag
`hIAPP fib
`Figure 6. Steady state anisotropy of Y37. Anisotropy
`of NAYA, rIAPP, hIAPPlag and hIAPPfib are shown with
`open bars. Dashed line is the value observed for the ani-
`sotropy of NAY A in frozen solution.23 Relative volume
`(filled bars) is calculated for a spherical
`fluorophore
`whose rotational diffusion describesthe observed aniso-
`tropy. Averages of three measurementsare shown, error
`bars are
`1SD. Note, error shown for hIAPPfib
`is
`0.001. However, we observe anisotropy to be depen-
`dent on reaction conditions with values ranging from
`0.22to 0.26.
`
`Novo Nordisk Ex. 2038, P. 6
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`Structure Formation during Fibrillogenesis of IAPP
`
`789
`
`ter value to represent 100 % exposure to solvent.
`The KSV measured for hIAPPfib
`(KSV ˆ 6.4 M 1)
`therefore reflects a solvent exposure of 55 %. This
`may reflect partial burial of all
`fluorophores, or
`complete burial of a fraction of the fluorophores.
`The latter
`is improbable given the apparent
`absence of curvature in the Stern-Volmer plots
`(Figure 4(a)). As a complementary approach to
`assessthe existence of
`two or more fiber popu-
`lations with distinct solvent accessibilities, we
`monitored quenching of tyrosine fluorescence by
`deprotonation. Our
`titration curve for NAYA
`yields a pKa of 9.9 while that of hIAPPfib is 11.8
`(Figure 5). A shift of
`2 units is consistent with
`partial burial and is comparable to pKa shifts seen
`in folded proteins, e.g. thermolysin.22 Furthermore,
`as the pKa could be determined using a single site
`model and since deprotonation results in a com-
`plete loss of
`fluorescence intensity,
`it
`is unlikely
`that our measurement of solvent exposure reflects
`a weighted average of exposed and occluded
`populations. This is particularly important, as the
`fiber preparation contains a mixture of
`fiber
`morphologies.26 Our observation of linear quench-
`ing behavior and single component ionization indi-
`cates either that a single local conformation is
`present under our conditions or that all confor-
`mations have similar solvent exposure for their C
`termini.
`fluorescence
`The maximum possible value for
`anisotropy is 0.4,27 which can occur provided the
`fluorophore is rigidly oriented during the excited
`state lifetime, it does not interact with its environ-
`ment and the absorption and emission moments
`are parallel. For real samples,a limiting anisotropy,
`r0, can be determined by making measurementsin
`a vitreous environment (e.g. frozen to
`62 C, in
`70% (v/v ) propylene glycol). For our model com-
`pound, NAYA, this yields r0 ˆ 0.278,while for pro-
`teins and peptides, values range from 0.21(histone
`H1) to 0.28 (leu5-enkephalin).23 At 25 C in an aqu-
`eous environment, we measure the steady state
`anisotropy of hIAPPfib
`to be 0.235. This implies
`that the conformational flexibility of Y37 in hIAPP-
`is comparable to proteins in frozen solution.
`fib
`Specifically, Y37 does not rotate on a timescale
`shorter than the lifetime of the excited state (1-3ns).
`For heterogeneousfibers, this is not necessarily
`the expected result. Fibers formed from SH3
`domains, for example, have helical repeat lengths
`that vary withi n a given fiber.28 One interpretation
`of this is that fibers are fluid at a local level. Given
`the loss of entropy upon fiber formation, this could
`be an effective means of recapturing some of this
`loss. By contrast, well-defined internuclear dis-
`tancesand unusually narrow linewidths have been
`determined for fibers formed from peptide frag-
`ments of Ab using solid state NM R methods.29,30
`These measurements are consistent with homo-
`geneous, crystal-like rigidity. Thus, the side-chain
`of Y37 in hIAPPfib behavesmuch like the residues
`of fibrillar Ab. This is surprising given that Y37 is
`
`ThT intensity
`Anisotropy
`
`0.3
`
`0.2
`
`0.1
`
`Anisotropy
`
`0
`
`0
`
`10
`
`30
`
`40
`
`5.5
`
`4.5
`
`3.5
`
`ThT intensity (CPS/105)
`
`2.5
`
`20
`Time /min
`Figure 8. Fiber formation kinetics of hIAPP followed
`by Y37 fluorescenceanisotropy. A single fiber formation
`reaction is initiated and split
`into two aliquots. Half
`is
`used for real-time anisotropy measurements,the remain-
`der for ThT assays.Overlay of ThT dye-binding assay
`intensity and tyrosine anisotropy show close agreement
`of the midpoints of the fiber formation transition. Real
`time anisotropy measurements made with one second
`averaging, nine second delay between points to mini-
`mize photobleaching.
`
`Sensitivity of apparent anisotropy to the exact
`composition of the fibril mixture is likely to con-
`tribute to this variability. Using negative stain
`transmission electron microscopy, our fiber prep-
`arations appear to be formed from 5 nm-wide pro-
`tofibers which have assembled into a range of
`higher order fiber morphologies (data not shown).
`This is consistent with previously reported data in
`which 5 nm protofibers are also seen to assemble
`into several well-characterized morphologies, e.g.
`8 nm and 11 nm forms.26 A second source of varia-
`bility may arise from the contribution of 1-2 mM of
`precursor which remains after
`fibrillogenesis is
`apparently complete.24
`
`Discussion
`The structural determinants of amyloid for-
`mation are challenging to characterize since fiber
`preparations are intrinsically
`heterogeneous in
`both length and width. The intermediates of fiber
`formation exist transiently allowing only limited
`time for study. The experiments performed here
`have nevertheless allowed us to elucidate several
`aspectsof the structure of both the lag-phase and
`fiber conformations of IAPP.
`The observation that
`intrinsic fluorescence in
`NAYA,
`rIAPP and hIAPPlag
`is more readily
`quenched than in hIAPPfib
`(Figure 4) can be
`explained by a significant change in exposed sur-
`face area. We estimate this change by comparing
`hIAPPfib with rIAPP. The predicted KSV for tyro-
`sine in rIAPP in an extended conformation is
`14.2M 1
`(see Materials and Methods). As our
`measurement,11.7M 1, is similar, we take this lat-
`
`Novo Nordisk Ex. 2038, P. 7
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`790
`
`Structure Formation during Fibrillogenesis of IAPP
`
`eight residues outside the proposed amyloidogenic
`core.
`Upon conversion of hIAPPlag
`the
`to hIAPPfib ,
`emission l max
`is unchanged (Figure 3(b)). How-
`ever, a changein the profile shifts the first moment
`(intensity weighted wavelength average) by 2 nm
`to shorter wavelengths. This is