`
`
`pharmaceutics 0 Cite This: Moi. Pharmaceutics 2019, 16, 2153 2161
`
`—W
`pubs.acs.org/molecula(pharmaceutics
`
`Peptide Oligomerization Memory Effects and Their Impact on the
`Physical Stability of the GLP-1 Agonist Liraglutide
`Jameson R. Bothe,*'TG Alexandra Andrews,§ Katelyn J. Smith,§ Leo A. Joyce,Te Yogita Krishnamachari,§
`and Sandhya Kashif
`
`1‘Process Research and Development and §Pharmaceutical Sciences, Merck 8: Co., Inc, Rahway, New Jersey 07065, United States
`
`. Supporting Information
`
`Fibril Formation
`
`® Stress—> Flue-um
`
`Kinetic
`Oligomer
`
`tau)
`
`
`
`I "“3""
`
`a
`
`ABSTRACT: Peptides and proteins commonly have complex
`structural landscapes allowing for transformation into a wide
`array of species including oligomers, aggregates, and fibrils.
`The formation of undesirable forms including agregates and
`fibrils poses serious risks from the perspective of drug
`development and d‘sease. Liraglutide, a GLPl agon'st for
`the treatment of diabetes, is a conjugated peptide that forms
`oligomers that can be stabilized by pH and organic solvents.
`We have developed an analytical
`toolkit
`to overcome
`challenges inherent to Liraghrtide’s conjugated acyl chain
`and probed the impact its oligomers have on its physical
`stability. Our studies show that Liraglutide’s Oligomer states
`have significant and potentially detrimental impacts on its
`propensity to agregate and form fibrils as well as its potency. Liraghrtide delivered as a synthetic peptide is able to maintain its
`Oligomerization state in dried lyophilized powders, acting as a memory effect from its synthetic process and purification.
`Through Liraglutide’s oligomer memory efl'ect, we demonstrate the importance and impact the process for synthetic peptides
`can have on drug development spanning from discovery to formulation development.
`
`..
`Lyophilized
`Peptide
`
`Stress E
`'
`’ g
`w
`
`Equilibrium
`Oligomer
`
`KEYWORDS: fibrillation, aggregation, secondary structure, size exdusion chromatography, CD spectroscopy, bioassay
`
`Peptides are important and complex modalities for
`pharmaceutical development, and approvals of new
`molecular entities for therapeutic treatments have rapidly
`risen over the past decade."3 Continued breakthroughs in
`chemistry for peptides and proteins will enable greater
`exploration of an expanded chemical space for developing
`new therapeutics.4 A key structural element common to
`peptide based therapeutics is the addition of a conjugate to the
`peptide core to enhance its pharmaceutical properties. Some
`benefits of conjugation can include tuning the peptide’s in vivo
`halflife or improving solubility.5 In addition to enhancing
`pharmaceutical properties,
`incorporation of conjugates can
`significantly impact
`the physical state and stability of the
`peptide in the final formulated drug product ultimately dosed
`in patients. Over the course of a formulated peptide’s shelf life,
`there is potential risk that the peptide may have a propensity to
`aggregate or form fibrils.‘3 These pathways of physical
`instability are undesirable due to their potential
`risk of
`immunoégenic responses and impacting bioavailability and
`eficacy. ‘7 The complex structural
`landscapes available to
`peptides and their associated risks necessitate that robust
`analytical methods are developed to characterize and under
`stand their stability to ensure stabile formulated drug products
`are delivered to patients.
`A major and currently expanding class of therapeutic
`peptides are those that target the GLP 1
`receptor for the
`
`treatment of diabetes.8_ll Several peptides that are marketed
`or in development are conjugated to allow dosing for different
`durations ranging from once daily to once weekly.11 For
`example,
`the GLP 1 agonist Liraglutide’s primary structure
`mimics the natural GLP 1 hormone with the exception of a
`single amino acid mutation at its N terminus and the inclusion
`of a C16 acyl conjugation (Figure 81). The natural GLP l
`peptide hormone,
`that which several
`therapeutic peptides
`resemble, is known to have a complex structural
`landscape
`including a high propensity to aggregate and form fibrils.”-l4
`Recently, it was shown that the GLP l peptide is one of a few
`known peptides reported in the literature to have unique pH
`dependent fibrillation properties.l4 Generally, peptides that
`undergo fibrillation do so in a concentration dependent
`manner where increasing pe tide concentration results in
`enhanced fibrillation kinetics.l '16 In stark contrast, the GLP l
`peptide shows an inverse concentration dependence at pH < 7,
`where it has a tendency to form fibrils more rapidly at low
`concentrations. Beyond the unique aspects of the GLP l
`peptide’s pH dependent fibrillation properties, Liraglutide has
`been shown to form soluble oligomers that vary in size as a
`
`January 23, 2019
`Received:
`April 13, 2019
`Revised:
`Accepted: April 16, 2019
`Published: April 16, 2019
`
`6 ACS Publications
`
`0 2019 American Chemical Society
`
`2153
`
`00:10.1021/acsmobhamracem9b00106
`Moi Phannoeeutic 2019, 16, 2153—2161
`
`MYLAN INST. EXHIBIT 1066 PAGE 1
`
`MYLAN INST. EXHIBIT 1066 PAGE 1
`
`MYLAN INST. EXHIBIT 1066 PAGE 1
`
`

`

`Molecular Pharmaceutics
`
`function of pH.17 Similar to the GLP 1 fibrillation concen
`tration dependence, the oligomerization state of Liraglutide
`differs depending on if the solution pH is acidic or basic.
`Here, we sought to probe whether Liraglutide, with the
`addition of its conjugation and pH dependent oligomerization,
`maintains the unique fibrillation properties that have recently
`been identified for GLP 1. To characterize Liraglutide, we have
`deployed an analytical
`toolkit
`including size exclusion
`chromatography (SEC) coupled with light scattering detection,
`kinetic fibrillation/aggregation experiments, transmission elec
`tron microscopy, circular dichroism spectroscopy, and bio
`assay. Through our characterization studies, we show that
`similar to GLP 1, Liraglutide exhibits unique concentration
`dependent fibrillation and aggregation kinetics as a function of
`pH.14 During development of our characterization techniques,
`it became apparent that control of Liraglutide’s pH dependent
`oligomerization state was crucial for its physical stability. Our
`studies show that Liraglutide is able to maintain its oligomer
`state in lyophilized powders, the form in which most synthetic
`peptides are delivered. Liraglutide’s oligomer states can
`significantly impact its propensity to form fibrils, aggregate,
`and its biological potency. Previous work on recombinant
`GLP 1 peptides has shown the synthetic process can have
`impacts on solubility and generation of undesirable aggre
`gates.18,19 Given Liraglutide’s oligomer memory, our studies
`highlight the importance of controlling and understanding the
`process used in synthetic peptide synthesis and purification
`and its potential downstream impacts that span from discovery
`to preclinical development and formulation activities.
`
`■ EXPERIMENTAL SECTION
`
`Materials. Liraglutide was purchased from Bachem
`(Torrance, CA) with a potency of 97.5% and used without
`further purification. The Bachem sourced Liraglutide was used
`for all studies with the exception of those specifically noted.
`Liraglutide was purchased from Achemblock (Burlingame,
`CA) with a purity of 99.4% and used without
`further
`purification. Liraglutide peptide powder was reconstituted
`with 50 mM sodium phosphate adjusted to the desired pH
`with sodium hydroxide. All other chemicals used in this study
`were reagent grade or better.
`Size-Exclusion Chromatography. Experiments were
`carried out on a variety of HPLC/UHPLC systems including
`Waters Acquity UPLCs, Aglient 1290s, and Agilent 1200s
`controlled using Empower 3. The columns used for SEC
`experiments included the Sepax Unix C 300 Å pore size 4.6 ×
`150 mm (UHPLC) and Sepax SRT C 300 Å pore size 7.8 ×
`300 mm (HPLC). The mobile phase was 10 mM sodium
`phosphate at a pH of 6.4 or 8.1. Note, because of the slow
`kinetics of the oligomer transition, the populations of States A
`and B were not significantly impacted by the mobile phase pH.
`Flow rates were 0.3 and 1.0 mL/min for UHPLC and HPLC,
`respectively. The LC sample tray was set to 5 °C for all sample
`analysis with the exception of 25 °C for incubation studies. UV
`detection for UHPLC/HPLC was carried out at 280 nm, and
`corresponding peak areas were used for oligomer population
`determination. HPLC SEC experiments included online
`detection of multiangle light scattering (MALS), quasielastic
`light scattering (QELS/DLS), and refractive index using Wyatt
`Dawn Heleos II, Wyatt Nanostar, and Wyatt Optilab Trex
`detectors. We measured and used the refractive index
`increment of dn/dc = 0.186 mL/g for Liragluitde (matches
`within 2% of a recent study) with a Wyatt Optilab Trex.17
`
`Article
`
`Analysis of light scattering data was carried out using Wyatt
`ASTRA 6.1.7.15.
`Lyophilization. Liraglutide samples were prepared by
`reconstituting the Liraglutide lyophilized powder with 50
`mM sodium phosphate at the desired pH (6.4, 6.7, 8.1).
`Samples were allowed to incubate at 25 °C for 43 h to achieve
`their equilibrium oligomer distribution. The equilibrated
`samples were then lyophilized using an SP Scientific VP 60X
`lyophilizer. The lyophilized samples were reconstituted with
`water and immediately assayed by SEC.
`Fluorescence and Kinetic ThioT Experiments. Liraglu
`tide samples were prepared by reconstituting the lyophilized
`peptide with 50 mM sodium phosphate at the desired pH
`(6.4−8.1) at a concentration of 8 mg/mL with gentle stirring
`until
`the peptide went
`into solution. The additional
`concentrations of 1−4 mg/mL were prepared by diluting the
`8 mg/mL Liraglutide stock with 50 mM sodium phosphate at
`the desired pH. Samples were then transferred to a 96 well
`plate with a total sample volume of 200 μL with 5 μM
`thioflavin T (ThioT). Physical
`stress and fluorescence
`detection (excitation = 440 nm and detection = 480 nm)
`were carried out using a Spectramax M2. Samples were
`constantly stressed by shaking at 25 °C, and ThioT
`fluorescence was recorded every 5 min.
`Transmission Electron Microscopy. Images of peptide
`fibrils were obtained on an FEI Tecnai Spirit Biotwin
`transmission electron microscope at a voltage of 120 kV. All
`peptide solutions were diluted to 1 mg/mL as necessary with
`water. A 5 μL aliquot of the 1 mg/mL peptide solution was
`deposited on the surface of a 200 mesh carbon coated copper
`grid. After 1 min, excess liquid was blotted away with filter
`paper, and the grid was rinsed briefly with 5 μL of water. The
`rinsewater was wicked away with filter paper, and 5 μL of 1%
`uranyl acetate was then added to the grid as a negative stain to
`enhance contrast. After 1 min, excess stain was blotted away,
`and the grids were imaged immediately.
`Circular Dichroism Spectroscopy. Circular dichroism
`spectra were acquired on a Chirascan qCD Spectrophotometer
`(Applied Photophysics, Surrey, UK). The temperature was
`maintained at 25 °C during the course of the measurement,
`consisting of two repetitions scanning from 200 to 280 at 1 nm
`bandwidth. Solid Liraglutide was dissolved to a concentration
`of 1 mg/mL in 50 mM sodium phosphate at the desired pH
`value (6.4, 6.7, and 7.2). These solutions were allowed to stand
`at room temperature over the course of the experiments.
`Aliquots (75 μL) were removed from the solution and further
`diluted 4× with buffer to a final concentration of 0.25 mg/mL
`before acquisition of CD spectra. This dilution was necessary
`to bring the total absorbance below 1 AU for the entire
`spectral range. The final time course spectra were normalized
`to the initial
`time point before comparison. For
`the
`experiments pertaining to addition of organic solvent, the
`same procedure was followed with the exception of the initial
`Liraglutide solutions, which were made with 10% of either
`EtOH or TFE. For analysis of kinetic data, we assumed
`Liraglutide CD spectra to be a linear combination of States A
`and B
`θ
`exp
`
`β θ
`α θ
`= · + ·
`A
`B
`
`(1)
`
`is the experimental CD spectrum, α is the
`where θ
`exp
`percentage of State A, θ
`A is the State A reference CD
`spectrum, β is the percentage of State B, and θ
`B is the State B
`
`2154
`
`DOI: 10.1021/acs.molpharmaceut.9b00106
`Mol. Pharmaceutics 2019, 16, 2153−2161
`
`MYLAN INST. EXHIBIT 1066 PAGE 2
`
`MYLAN INST. EXHIBIT 1066 PAGE 2
`
`

`

`Molecular Pharmaceutics
`
`Article
`
`Figure 1. Characterization of Liraglutide oligomerization by SEC. (A) Liraglutide exists as two different oligomerization states that are stabilized by
`acidic or basic conditions. (B) SEC chromatograms of 4 mg/mL Liraglutide solutions at pH 6.4−8.1 recorded immediately after preparation. (C)
`SEC chromatograms of 4 mg/mL Liraglutide solutions at pH 6.4−8.1 recorded after incubation at 25 °C for 1.5 days. (D) Population of State B
`determined by SEC peak areas immediately after preparation (filled circles) and after incubation at 25 °C for 1.5 days (open circles).
`
`reference CD spectrum. Eq 1 was fit using the “Non
`LinearModelFit” function built into Wolfram Mathematica 8.
`Cell Based Potency Assay. The bioassay method used
`was based on Cisbio’s HTRF technology (Homogeneous
`Time Resolved Fluorescence) for quantitative measurement of
`cyclic AMP using Cisbio’s cAMP Dynamic 2 reagents with
`Chinese hamster ovary (CHO) cells stably expressing the
`glucagon like peptide 1 receptor (CHO GLP1R) cells. As a G
`protein coupled receptor, GLP 1 receptor mediated signaling
`involves activation of the adenylate cyclase component of
`GLP 1 receptor, which increases the level of cAMP (adenosine
`3′,5′ cyclic mono phosphate).20
`The bioactivity of samples was tested in a bioassay
`immediately after preparation of Liraglutide samples at pH
`6.4 and 8.1 that primarily populate State B. The pH 6.4 sample
`was in a kinetic oligomer distribution at pH 6.4, primarily
`populating State B (Figure 1D). Bioactivity was also tested for
`samples that had been allowed to incubate at 25 °C for 3 days.
`Here, the pH 6.4 sample had achieved equilibrium primarily
`populating State A. A freshly prepared sample at pH 8.1 was
`used as a reference to compare the potency values of the
`samples at pH 6.4 (immediate and 3 day, 25 °C) and 8.1 (3
`days, 25 °C). Percent relative potency values (reference IC50/
`× 100) were determined from the comparison of
`sample IC50
`the reference and test sample using a 4 PL curve fit. The IC50
`value obtained from the curve represents the concentration of
`Liraglutide that inhibits 50% of the maximum response.
`
`■ RESULTS AND DISCUSSION
`
`Characterization of Liraglutide Oligomerization. It
`was recently reported that Liraglutide can populate two
`oligomerization states that are stabilized by varying pH (Figure
`1A).17 At pH < 7, Liraglutide was shown to exist as a 12 mer
`
`by dynamic light scattering (DLS) and static light scattering
`(SLS). While under basic conditions, Liraglutide has been
`shown to exist as either an 8 mer17 or 7 mer21 depending on
`the technique used (light or small angle X ray scattering
`(SAXS)). While DLS, SLS, and SAXS were able to provide
`deep insights into the oligomerization of Liraglutide,
`the
`interpretation and deconvolution of light/X ray scattering data
`has significant limitations for cases such as Liraglutide, where it
`is possible that multiple states with differing size can
`simultaneously exist in solution at near neutral pH.22−25
`To overcome challenges
`inherent
`to light
`scattering
`measurements of bulk solutions, we sought to develop a size
`exclusion chromatography method able to separate the
`oligomerization states of Liraglutide. Achieving separation of
`the oligomerization states would allow for accurate determi
`nation of their populations and enable additional on or offline
`analysis by techniques such as light scattering or mass
`spectrometry. Separation of conjugated peptides is notoriously
`challenging due to their conjugation chemistries having a
`tendency to cause undesirable secondary interactions with
`column stationary phases. This is particularly the case for
`acylated peptides where a greasy carbon chain (C16 for
`Liraglutide) is covalently attached. Addition of organic solvents
`in some cases can help to minimize secondary interactions but
`are not
`ideal, because they could disrupt or generate
`aggregates/oligomers on column. Initial SEC method screen
`ing with a pure aqueous mobile phase showed a challenge in
`achieving conditions that minimized secondary interactions
`between Liraglutide and the stationary phase (Figure S2).
`Through our screen, we identified that by utilizing Sepax “C
`series” columns (UHPLC/HPLC) we could achieve con
`ditions with minimal secondary interactions, good peak shape,
`and most importantly separation of the oligomerization states
`of Liraglutide. The stationary phase of these columns has a
`
`2155
`
`DOI: 10.1021/acs.molpharmaceut.9b00106
`Mol. Pharmaceutics 2019, 16, 2153−2161
`
`MYLAN INST. EXHIBIT 1066 PAGE 3
`
`MYLAN INST. EXHIBIT 1066 PAGE 3
`
`

`

`Molecular Pharmaceutics
`
`proprietary lay down nanometer film chemically bonded to the
`silica particles likely minimizing Liraglutide’s undesirable
`secondary interactions.
`With the Sepax SEC method in hand, we were able to
`demonstrate the separation of two distinct oligomerization
`states (Figure 1B) that are stabilized as a function of pH
`consistent with a recent report by Wang et al.17 State A is
`stable at pH < 7, and State B is stable at pH > 7 (Figure 1A).
`Separation of the individual states allowed us to probe their
`structure with online DLS and MALS detection (Figure 1B).
`The hydrodynamic radius for each state (Rh,A = 3.4 ± 0.3 nm
`and Rh,B = 2.6 ± 0.2 nm) was in agreement with the radii
`recently reported for each species by bulk DLS,17 confirming
`that the oligomerization states of States A and B observed on
`column are the same as in bulk solution. Molecular weight
`analysis of the individual species using MALS showed that
`State A exists as a 13 mer (13.0 ± 0.5), and State B exists as a
`7 mer (6.7 ± 0.6), consistent with the sizes recently reported
`as 12 mer (State A) and 8 mer or 7 mer (State B).17,21 The
`differences observed in reported molecular weights are likely
`attributable to error and analysis differences between SEC
`MALS, bulk SLS, and SAXS. The combined light scattering
`data supports
`that
`the species observed by SEC are
`representative of the oligomers in bulk solution. Further, it
`was previously reported that the kinetics of the transformation
`between States A and B is slow, occurring on the time scale of
`days.17 Our SEC results were consistent with pH driven
`transitions being slow on the SEC time scale, where the mobile
`phase pH did not significantly impact the populations of States
`A and B calculated from the oligomer peak areas.
`Surprisingly, upon preparation of Liraglutide samples
`expected to favor State A at pH 6.4 by reconstitution with
`sodium phosphate buffer, we observed State B to be the initial
`primarily populated state (Figure 1B). After the samples were
`allowed to equilibrate at room temperature for several days, the
`system achieved equilibrium with Liraglutide transitioning to
`fully populate the thermodynamically favored State A (Figure
`1C). We tested synthetic Liraglutide sourced from another
`vendor and saw a similar effect (Figure S3). The observation of
`an initial kinetic distribution of Liraglutide upon reconstitution
`of the lyophilized peptide powder led to the hypothesis that
`either the thermodynamically unfavored State B could be
`kinetically favored upon hydration of the dry powder or that
`the Liraglutide oligomer immediately populated in solution is a
`memory effect
`from its prelyophilized state. In order to
`distinguish between the two scenarios, we prepared Liraglutide
`solutions that populated different distributions of States A and
`B (pH 6.4, 6.7, 8.1) and lyophilized the solutions after they had
`achieved their oligomer thermodynamic equilibrium. Similar to
`our initial experiment, we assayed the oligomer distribution
`immediately after reconstitution of the lyophilized samples
`with water. Interestingly, samples that had favored State B
`the “as is” purchased lyophilized
`upon reconstitution of
`powder now populated their
`thermodynamically favored
`oligomer distributions observed prior to our lyophilization
`(Figure 2). Through the lyophilization process, the oligomer
`distribution was not impacted by any sodium phosphate acidic
`pH swing that occurred during freezing26 because of the slow
`oligomer transformation kinetics that are likely even slower at
`subzero temperatures. This demonstrates that Liraglutide has
`an oligomer “memory effect”, where the oligomerization states
`in solution prior to lyophilization are maintained in the
`
`Article
`
`Figure 2. Populations of Liraglutide State B before and after
`lyophilization at varying pH. Fresh (black): Liraglutide population
`after reconstitution of the peptide powder with 50 mM sodium
`phosphate; Pre Lyo (red): Liraglutide State B population after
`incubation at 25 °C for 43 h prior to lyophilization; Post Lyo
`(green): Liraglutide State B population immediately after recon
`stitution of the Lyo cake with water.
`
`lyophilized powder and populated upon hydration into
`solution.
`Oligomerization Impact on Physical Stability. Given
`our observation of the oligomer “memory effect”, we sought to
`probe the impact oligomer states could have on the physical
`stability of Liraglutide. To examine the physical stability and
`tendency to form fibrils and aggregate, we carried out studies
`catalyzing fibrillation and aggregation with physical stress using
`ThioT fluorescence as a reporter of fibril formation. Liraglutide
`stress studies were done using concentrations of 1−8 mg/mL
`at pH 6.4−8.1 employing physical stress by shaking at 25 °C.
`Shown in Figure 3 are the resulting fluorescence profiles for
`stressed Liraglutide at 25 °C for pH 6.4 and 8.1 where
`fluorescence increases upon ThioT binding fibrils or
`aggregates. Liraglutide exhibits an interesting concentration
`dependence at pH 6.4 (Figure 3A) where fibrillation occurs
`more rapidly as concentration decreases. This is similar to a
`recent report on the natural GLP 1 hormone, for which a
`similar concentration dependence is observed at pH < 7.14 It is
`interesting that Liraglutide is able to maintain this unique
`property despite the addition of a bulky conjugate, mutation at
`the N terminus, and formation of oligomers in solution. This
`rare inverse concentration dependence has only been reported
`for a few peptides.14,27,28 At a basic pH of 8.1 (Figure 3B),
`Liraglutide exhibits more rapid fibrillation with increasing
`concentration, which is the classical concentration dependence
`typically observed for fibrillation.15,16,29
`In addition to the unique concentration dependence
`observed at pH 6.4, the oligomerization state of Liraglutide
`plays a significant role in its tendency to form fibrils and
`aggregate. Utilizing our knowledge of Liraglutide oligomeriza
`tion from SEC and our ability to control the populations of
`each oligomer, we carried out studies to probe the impact of
`the initial oligomer state of the system (kinetic or equilibrium)
`on aggregation and fibrillation. First, we compared the ThioT
`profile for Liraglutide samples immediately prepared at pH 6.4
`that exist in a kinetic oligomer distribution where State B is
`primarily populated (populated at ∼80%) and samples that
`had been allowed to achieve equilibrium fully populating State
`A prior to stress. As shown in Figure 3A, the ThioT profiles are
`
`2156
`
`DOI: 10.1021/acs.molpharmaceut.9b00106
`Mol. Pharmaceutics 2019, 16, 2153−2161
`
`MYLAN INST. EXHIBIT 1066 PAGE 4
`
`MYLAN INST. EXHIBIT 1066 PAGE 4
`
`

`

`Molecular Pharmaceutics
`
`Article
`
`Figure 3. Fibrillation/aggregation of 1−8 mg/mL Liraglutide solutions upon immediate sample preparation and after incubation at 25 °C for 3
`days. Six replicates were recorded, and each replicate is a different color. (A) ThioT fluorescence profile recorded during physical stress of
`Liraglutide solutions at pH 6.4 immediately after preparation (top) and after incubation at 25 °C for 3 days (bottom). (B) ThioT fluorescence
`profile recorded during physical stress of Liraglutide solutions at pH 8.1 immediately after preparation (top) and after incubation at 25 °C for 3
`days (bottom).
`
`in kinetic or
`drastically different between the samples
`equilibrium distributions particularly at low concentrations.
`Samples that were stressed while in a kinetic oligomer
`distribution at 1 mg/mL underwent more rapid changes in
`fluorescence than those that were allowed to achieve their
`equilibrium state prior to stress (Figure 1). We also observed a
`similar effect at pH 6.7 where samples stressed beginning with
`a kinetic distribution had much more rapid increases in
`fluorescence compared to those with an equilibrium oligomer
`distribution (Figure S4). In contrast
`to the samples that
`contain State A (pH 6.4 and 6.7), samples at a pH > 7 that
`exclusively populate State B upon reconstitution and at
`equilibrium became less stable as the samples aged. At pH
`8.1, samples stressed immediately after preparation underwent
`ThioT fluorescence changes slower than samples allowed to
`incubate at 25 °C for 3 days (Figure 3B). SEC MALS analysis
`of pH 8.1 samples aged at 25 °C for 3 days showed that high
`molecular weight aggregates were present at the initiation of
`the fibrillation experiments, having formed during the
`incubation period. It is possible that these aggregates could
`give rise to the higher initial fluorescence and more rapid
`fibrillation. Liraglutide samples at pH 7.2, only populating
`State B, also showed decreasing stability after the samples had
`aged for 3 days (Figure S4).
`
`These combined data from pH 6.4−8.1 show that
`Liraglutide’s fibrillation and aggregation characteristics are
`modulated both by the populations of each oligomer state and
`whether the system is in thermodynamic equilibrium prior to
`initiating physical stress. Systems (pH 6.4 and 6.7) with a
`kinetic oligomer distribution were significantly less stable
`where they were undergoing oligomer conformational trans
`formations during physical stress. The ThioT profiles are
`complex particularly at low pH having variable lag times,
`multiple phases in fluorescence increase, and a lack of signal
`plateauing. These complexities preclude global fitting of the
`data to a specific fibrillation or aggregation mechanism.30
`However,
`it
`is clear at
`low pH that when the oligomer
`distribution is not at equilibrium, Liraglutide has enhanced
`physical instability with an inverse concentration dependence.
`When under kinetic oligomer conditions, transformation from
`State B to State A may enable faster fibrillation and aggregation
`because of the presence of a catalyzing intermediate state. The
`inverse concentration presence for fibril formation could be
`from high concentrations of State B that play an inhibitory role
`in the fibrillation/aggregation process favored at acidic pH’s.
`The unique ThioT profiles observed for Liraglutide
`necessitated a deeper look with orthogonal methods to better
`understand the species that had formed upon stress. Increases
`in ThioT fluorescence are indicative of conformational
`
`2157
`
`DOI: 10.1021/acs.molpharmaceut.9b00106
`Mol. Pharmaceutics 2019, 16, 2153−2161
`
`MYLAN INST. EXHIBIT 1066 PAGE 5
`
`MYLAN INST. EXHIBIT 1066 PAGE 5
`
`

`

`Molecular Pharmaceutics
`
`Article
`
`Figure 4. TEM micrographs of Liraglutide fibrils after stress.
`
`changes; however, the fluorescence increase could be due to
`fibrillation, formation of aggregates that bind ThioT, or both.31
`To probe aggregate formation, we captured and assayed
`samples by SEC at pH 6.4 and 8.1 over the course of the
`ThioT fluorescence profiles (Figure S5). Interestingly, pH 6.4,
`samples stressed in a kinetic oligomer distribution had a higher
`tendency for aggregate formation at
`lower Liraglutide
`concentrations. The aggregates detected were ∼20−30 MDa
`’s of ∼50 and ∼100−150 nm,
`
`in size with Rh’s and Rg
`respectively. The ratio of Rg and Rh of ∼2−3 suggests that
`these aggregates are rod like. Further, formation of significant
`high molecular weight aggregates at 1 mg/mL correlated with
`the timing (∼20 h) of
`significant
`increases
`in ThioT
`fluorescence (Figure 3A). Samples at pH 6.4 incubated at 25
`°C prior to stress had different aggregation properties where
`significant aggregation was delayed and increased with
`increasing concentration (Figure S5). In contrast, pH 8.1
`aggregates had a significantly broader SEC peak shape than
`those formed at pH 6.4 (Figure S5). pH 8.1 aggregates formed
`more readily with increasing concentration with differing size
`’s of ∼60 and ∼150 nm,
`of ∼15 MDa and Rh
`’s and Rg
`respectively. Overall, SEC characterization of soluble species
`demonstrated stress induced formation of high molecular
`weight aggregates that have pH/oligomerization dependent
`properties and become significantly populated as ThioT
`fluorescence increases.
`Next, we characterized the samples poststress by TEM
`imaging to probe if fibril morphology was dependent on pH
`and the equilibrium state of Liraglutide oligomers when
`stressed. Single Liraglutide fibrils are approximately 7 nm in
`width with distinct pH and concentration dependent
`morphological differences (Figure 4). At pH 6.4, 1 mg/mL
`Liraglutide solutions developed fibrils that exist in higher order
`bundles among unstructured peptide aggregates. Increasing the
`peptide concentration to 8 mg/mL led to a more defined
`
`network where bundled fibers are composed of two to three
`individual fibrils. At pH 8.1, 1 mg/mL Liraglutide solutions
`formed well defined single fibrils that occasionally intertwine
`together. This
`intertwined morphology becomes more
`prevalent in samples at elevated concentrations. The fibril
`morphologies observed at a given pH and peptide concen
`tration do not differ substantially when comparing samples
`stressed in a kinetic or equilibrium oligomer distribution,
`despite the significant differences observed in the ThioT assay.
`This suggests that the final fibular state is less governed by the
`kinetic assembly but rather the molecular interactions between
`peptides that are linked to pH and peptide concentration.
`These unique peptide interactions at
`the molecular level
`extend to the macroscale morphology of the peptide fibril as
`observed in both synthetic32 and natural33 fibril
`forming
`peptides.
`By characterizing and controlling Liraglutide’s oligomers in
`either a kinetic or equilibrium distribution, we were able to
`probe their impact on Liraglutide’s physical stability. Overall,
`the aggregation/fibrillation data shows that both the oligomer
`identity and status of thermal equilibrium from the oligomer
`“memory effect” can play major roles in fibrillation/aggregation
`kinetics under physical stress.
`Oligomer Structure and Driving Forces for Stabiliza-
`tion. Because Liraglutide’s oligomers play such an important
`role in its tendency to aggregate and form fibrils, we probed
`their secondary structure using CD spectroscopy. A recent
`study showed that the CD spectrum of State A is a mixture of
`secondary structural elements, and State B is primarily α
`helical.17 As shown in Figure 5A, CD spectra of Liraglutide
`samples fully populated in State A (pH 6.4) or State B (pH
`7.2) were consistent with the previous work by Wang et al.17
`We have shown by SEC that Liraglutide’s oligomer distribution
`is maintained in its dried lyophilized powder form. This led us
`to inquire if Liraglutide’s
`secondary structure was also
`
`2158
`
`DOI: 10.1021/acs.molpharmaceut.9b00106
`Mol. Pharmaceutics 2019, 16, 2153−2161
`
`MYLAN INST. EXHIBIT 1066 PAGE 6
`
`MYLAN INST. EXHIBIT 1066 PAGE 6
`
`

`

`Molecular Pharmaceutics
`
`Article
`
`detection limits of our CD and SEC experiments, we cannot
`exclude any proposed mechanism.
`Our observations confirmed that secondary structure was
`distinctly linked to each oligomer state and that there are likely
`key structure driven contacts that stabilize each oligomer state.
`Next, we tested if it was possible to modulate the oligomer
`distribution of Liraglutide by factors other than pH. Previous
`structural studies of the natural GLP 1 hormone have used
`trifluoroethanol34 (TFE) to stabilize an α helical conformation
`allowing for
`structural determination and analysis by
`NMR.35,36 Similarly, a recent report showed that TFE could
`enhance Liraglutide’s α helical structural content by CD.37
`Addition of 10% TFE to Liraglutide at pH 6.4 and 6.7 drove its
`CD spectrum to be primarily composed of α helical structure,
`and SEC confirmed that State B had been fully populated at
`each pH (Figure S6). To further probe the impact of organic
`solvents on Liraglutide’s conformation, we carried out a similar
`study adding ethanol from 10 to 35% at pH 6.4 and 6.7.
`Although the change in oligomer populations was not as
`drastic as TFE, ethanol had a similar effect in stabilizing State
`B.
`Increasing the ethanol concentration to 35% pushed
`Liraglutide to fully favor State B at pH 6.4 and 6.7 (Figure
`S6). Addition of either TFE or ethanol at varying levels further
`showed that Liraglutide readily populates