Consideration of Conformational Transitions and Racemization during
`Process Development of Recombinant Glucagon-like Peptide-1
`
`RICHARD I. SENDEROFF*, KATHLEEN M. KONTOR, LOTTE KREILGAARD, JIN JYI CHANG, SUNIL PATEL,
`JONATHON KRAKOVER, JANE K. HEFFERNAN, LEO B. SNELL†, AND GARY B. ROSENBERG
`
`Contribution from ZymoGenetics Corporation, 1201 Eastlake Ave. East, Seattle, Washington 98102, and Novo Nordisk A/S,
`Novo Alle, DK-2880, Bagsvaerd, DK.
`
`Received July 15, 1997.
`
`Accepted for publication November 18, 1997.
`
`Abstract 0 Physicochemical characterization of dry, excipient-free
`recombinant glucagon-like peptide-1 (rGLP-1) indicates the conforma-
`tion and purity of the bulk peptide is dependent on the purification
`scheme and the in-process storage and handling. The recombinant
`peptide preparations were highly pure and consistent with the expected
`primary structure and bioactivity. However, variations in solubility were
`observed for preparations processed by different methods. The
`differences in solubility were shown to be due to conformational
`differences induced during purification. A processing scheme was
`identified to produce rGLP-1 in its native, soluble form, which exhibits
`FT-IR spectra, consistent with glucagon-like peptide-1 synthesized by
`solid-state peptide synthesis.
`rGLP-1 was also found to undergo base-
`catalyzed amino acid racemization. Racemization can impact the yield
`and impurity profile of bulk rGLP-1, since the peptide is exposed to
`alkali during its purification. A combination of enzymatic digestion
`using leucine aminopeptidase (which cleaves N-terminal L-amino acids
`. D-amino acids) and matrix-assisted laser desorption ionization mass
`spectrometry was used to identify racemization as a degradation
`pathway. The racemization rate increased with increasing temperature
`and base concentration, but decreased with increasing peptide
`concentration. The racemized peptides were shown to be less
`bioactive than rGLP-1.
`
`Advances in biotechnology are leading to the discovery
`of an ever increasing number of proteins and peptides with
`pharmaceutical applications. Biopharmaceutical develop-
`ment and realization of the therapeutic potential of these
`proteins and peptides is dependent on their production and
`purification in commercially viable quantities. Unfortu-
`nately, proteins possess multiple functional groups in
`addition to three-dimensional structure, making them
`inherently unstable molecules. This complicates the re-
`combinant production and purification of homogeneous
`protein preparations having the desired biological and
`physicochemical characteristics.1,2 Our studies involving
`glucagon-like peptide-1 (GLP-1) address these issues and
`illustrate the importance of physicochemical characteriza-
`tion in support of process development (in addition to
`formulation and analytical development) of recombinant
`proteins and peptides.
`Glucagon-like peptide-1 (Figure 1) is a 31 amino acid
`peptide (3356 Da, pI 5.4) which promotes the secretion of
`insulin and has been identified as a mediator of satiety.3-5
`Much of its biological evaluation has been conducted with
`peptide synthesized by solid-state peptide synthesis (sGLP-
`1). Clinical evaluation and commercialization of GLP-1
`requires a recombinant production process, making a
`
`* To whom correspondence should be addressed (tel, 206-442-6766;
`fax, 206-442-6608; e-mail, senderor@zgi.com).
`† Novo Nordisk A/S.
`
`Figure 1sAmino acid sequence of GLP-1.
`
`pharmaceutical product economically feasible. Both the
`physicochemical properties and the biological properties of
`the recombinant peptide must be equivalent to the syn-
`thetic peptide, so that the formulation, administration
`protocols, and clinical data established for sGLP-1 can be
`applied toward the development of a recombinant GLP-1
`(rGLP-1) product. A production process which yields a dry,
`excipient-free bulk peptide intermediate simplifies char-
`acterization studies (i.e. stability, conformation), dosage
`form development (e.g. immediate release vs sustained
`release formulations), in vivo considerations (e.g. dosing
`protocols), and storage issues (e.g. temperature, capacity).6
`We conducted studies to characterize the physicochemical
`behavior of rGLP-1 in support of process development of
`dry, excipient-free bulk peptide. Our results indicate that
`the conformation and purity of the bulk peptide is depend-
`ent on the purification scheme and the in-process storage
`and handling.
`
`Experimental Section
`
`GLP-1 PreparationsssGLP-1 was synthesized by solid-state
`peptide synthesis and was obtained from Bachem California (lot
`#FGLP9201BR). s(D-Ser8)GLP-1 was synthesized by solid-state
`peptide synthesis at ZymoGenetics. rGLP-1 was overexpressed
`in yeast (Yarrowia lipolytica), purified from culture media, and
`subsequently lyophilized from either an ammonium hydroxide
`solution or a suspension in water. rGLP-1(A) was the original
`batch and was purified by a combination of hydroxylapatite
`chromatography, reverse phase chromatography, and precipitation
`(Figure 2A). Prior to lyophilization, rGLP-1(A) was precipitated
`(by pH adjustment) to remove organic solvent from the reverse
`phase eluate, resolubilized ((cid:25)10 mg/mL) in 0.05 M ammonium
`hydroxide, and subsequently lyophilized. Portions of rGLP-1(A)
`were reprocessed by three different schemes (Figure 2B), yielding
`rGLP-1(B), rGLP-1(C), and rGLP-1(D), to determine the effects
`of processing on the resultant peptide structure and solubility
`behavior. Bulk lyophilization was performed using Lyoguard
`freeze-drying bags (W. L. Gore & Associates) in a Dura-Stop MP
`Tray Dryer/20 Liter Dura-Dry MP (-85 °C) freeze-drying system
`(FTS Systems, Stone Ridge, NY) equipped with a liquid nitrogen
`trap.
`
`© 1998, American Chemical Society and
`American Pharmaceutical Association
`
`S0022-3549(97)00272-4 CCC: $15.00
`Published on Web 01/07/1998
`
`Journal of Pharmaceutical Sciences / 183
`Vol. 87, No. 2, February 1998
`
`Novo Nordisk Ex. 2042, P. 1
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`luciferase gene controlled by a cyclic AMP response element.
`Exogenous GLP-1 binds to the receptor and activates adenylate
`cyclase resulting in an increase in intracellular cyclic AMP. Cyclic
`AMP then activates transcription of the luciferase gene. The
`expressed luciferase is capable of catalyzing oxidation of luciferin,
`generating photons which are quantitated in a scintillation counter
`(LucLite Luciferase Reporter Gene Assay Kit, TopCount Micro-
`plate Scintillation Counter; Packard). Luciferase expression is
`directly proportional to the concentration of GLP-1 binding to the
`receptor and a standard four-parameter curve can be constructed
`from known concentrations of sGLP-1 for quantitation of unknown
`samples. The assay is carried out in a deep-well titer plate
`(Beckman #267007) format.
`Matrix-Assisted Laser Desorption Ionization Mass Spec-
`trometry (MALDI-MS)sA combination of N-terminal enzymatic
`digestion using leucine aminopeptidase (LAP) and MALDI-MS of
`GLP-1 variants and isolated aqueous degradation products was
`used to identify the base-catalyzed aqueous degradation pathway-
`(s). Samples were prepared at concentrations of (cid:25)0.1 mg/mL. The
`N-terminal enzymatic digestion was initiated by adding 20 (cid:237)L of
`sample solution to 80 (cid:237)L of Tris (50 mM)/MgCl2 (50 mM) buffer
`(pH 8.5), 4 (cid:237)L ofn-octyl glucopyranoside (5 mg/mL), and 10 (cid:237)L of
`leucine aminopeptidase suspension (Sigma, L-5658, 3 mg/mL; 750
`units/mg; 1 unit will hydrolyze 1 (cid:237)mol of substrate per minute at
`pH 8.5 at 25 °C). The reaction mixture was incubated for 48 h at
`40 °C, and was then stopped by adding 10 (cid:237)L of acetonitrile/water/
`trifluoroacetic acid (50/50/0.1). MALDI-MS was performed using
`a Micromass TofSpec SE in linear mode. A 1.5-(cid:237)L portion of the
`digested sample was mixed with 1.5 (cid:237)L of R-cyano-4-hydroxycin-
`namic acid solution (Hewlett-Packard) containing 2% trifluoro-
`acetic acid, and the resulting mixture was spotted onto the target
`and allowed to air-dry.9,10 Spectra were accumulated several times
`from random places on the spot to verify consistent relative peak
`intensity thereby ensuring a homogeneous sample spot.
`
`Results and Discussion
`The process by which rGLP-1(A) was produced resulted
`in a highly pure (1.5% peptide contaminants by RP-HPLC,
`3.5% residual acetate, 5.9% residual moisture) white
`powder which was consistent with sGLP-1 in terms of
`amino acid sequence, molecular weight, and in vitro
`bioactivity. However, rGLP-1(A) did not exibit the expected
`solubility in neutral phosphate buffer based on solubility
`studies conducted with sGLP-1. Equilibrium solubility
`determinations by pH titration of a 1.7 mg/mL sGLP-1
`solution (23 °C) resulted in a pH-solubility profile consis-
`tent with expectations for a peptide having a pI 5.4. A
`minimum solubility of 0.2 mg/mL at approximately pH 6.0
`and a precipitation zone (solubility < 1.7 mg/mL) between
`approximately pH 4.5 and 6.5 was observed. The solubility
`of sGLP-1 allowed for preparation of solutions at concen-
`trations >10 mg/mL in neutral phosphate buffers, which
`was required for in vivo biological evaluations. However,
`equivalent preparations could not be prepared with rGLP-
`1(A) due to its limited solubility in neutral phosphate
`buffers (<1 mg/mL). Both sGLP-1 and rGLP-1(A) are
`readily soluble (>10 mg/mL) in acidic (50 mM phosphoric
`acid) and basic (50 mM ammonium hydroxide) solvents,
`which allowed for in vitro bioactivity comparisons. Even
`though the in vitro bioactivities of sGLP-1 and acid-
`solubilized rGLP-1(A) were equivalent, the differences in
`solubility behavior in neutral phosphate buffers revealed
`inconsistencies between the sGLP-1 and rGLP-1(A) bulk
`peptides which impacted formulation and subsequent in
`vivo biological comparisons.
`It has been previously shown that chemical and shear
`pertubations can induce precipitation of sGLP-1 which is
`related to conformational transitions.11 Secondary struc-
`ture transitions in proteins have also been associated with
`salt-induced precipitation and dehydration.12,13 FTIR spec-
`tra of rGLP-1(A) revealed significant differences (area of
`overlap ) 73%) in the conformationally sensitive amide I
`
`Figure 2sSummary of processing schemes for rGLP-1 variants. Key:
`(A)
`purification of rGLP-1(A) and (B) reprocessing of rGLP-1(A) yielding rGLP-
`1(B), rGLP-1(C), and rGLP-1(D).
`
`Fourier Transform Infrared Spectroscopy (FTIR)sFTIR
`analysis was used to characterize conformation differences be-
`tween GLP-1 variants. FTIR spectra were collected at 25 °C using
`a Magna 550 IR spectrometer (Nicolet Instrument Corp.) equipped
`with a KBr beam splitter and a DTGS detector. Samples were
`prepared by mixing 0.5-1 mg of GLP-1 with 300 mg of KBr, which
`was then compressed into a pellet. Each spectra was obtained by
`collecting a 256-scan interferogram using a single-beam mode with
`a 4-cm-1 resolution. Background and water vapor spectra were
`subtracted from each sample spectra. The resultant difference
`spectra were smoothed using a seven-point Savitsky-Golay func-
`tion to remove the possible white noise. The inverted second-
`derivative GLP-1 spectra (OMNIC software; Nicolet Instrument
`Corp.) were curve-fit using Guassian band profiles (GRAMS
`software; Galactic Industries). The individual secondary struc-
`tural elements (amide I band region) were assigned to wavenum-
`bers on the basis of carbonyl stretching in the protein backbone
`structure according to the method of Dong and Caughey.7 Semi-
`quantitative comparison of spectra from rGLP-1 preparations to
`sGLP-1 (reference) spectra were made be estimating the area of
`overlap.8
`HPLC AnalysissThe analytical HPLC method was used to
`estimate the purity of bulk peptide, assay samples from kinetic
`studies, isolate degradation products, and estimate the concentra-
`tion and purity of isolated peak fractions. The HPLC system was
`composed of a Hewlett-Packard HP1090 with diode array detec-
`tion. Samples in the autosampler were maintained at (cid:25)5 °C
`during analyses. Hewlett-Packard Chemstation software was used
`for data aquisition and integration. [Column: PLRP-S (Polymer
`Labs), 250 (cid:2) 4.6 mm i.d., 8-(cid:237)m particle size, 300 Å pore size.
`Conditions:
`load at 40% B (1 min) and then linear AB ramp to
`65% B (25% B/min) and elution using a linear AB gradient from
`65% to 80% (0.5% B/min), where eluent A was 0.1% TFA in water
`and eluent B was 0.1% TFA in acetonitrile/water (50:50, v/v); flow
`rate, 1 mL/min; detection wavelength, 215 nm.]
`Luciferase BioassaysThe luciferase bioassay was used to
`evaluate in vitro biologic activity of GLP-1 and its aqueous
`degradation products. The assay uses baby hamster kidney (BHK)
`cells that express the human GLP-1 receptor and contain a
`
`184 / Journal of Pharmaceutical Sciences
`Vol. 87, No. 2, February 1998
`
`Novo Nordisk Ex. 2042, P. 2
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`Figure 3sSecond-derivative infrared spectra in the amide I region, comparing each rGLP-1 variant to sGLP-1. Corresponding secondary structural compositions
`are shown in Table 1.
`
`Table 1sSecondary Structural Composition of sGLP-1 and rGLP-1
`Variants
`
`GLP-1 preparation
`
`structure
`R-helix
`(cid:226)-sheet
`turn
`intermolecular (cid:226)-sheet
`
`sGLP-1 rGLP-1(A)
`
`rGLP-1(B)
`
`rGLP-1(C)
`
`rGLP-1(D)
`
`46.2
`25.6
`20.9
`7.3
`
`35.0
`8.0
`18.0
`39.0
`
`26.5
`20.0
`17.4
`36.1
`
`64.3
`15.5
`13.3
`6.9
`
`50.2
`20.9
`22.9
`6.0
`
`region compared to sGLP-1 (Figure 3A; Table 1). Signifi-
`cant bands were found at 1620 and 1696 cm-1 for rGLP-
`1(A), which were not present in sGLP-1. These bands are
`associated with the formation of intermolecular (cid:226)-sheet
`structures.14,15 A corresponding decrease in R-helical
`structures compared to sGLP-1 is reflected by a reduction
`in the peak at 1657 cm-1. These differences suggest that
`in-process treatment of rGLP-1(A) induced loss of disor-
`dered elements and formation of nonnative protein ag-
`gregates in the dried solid. Interestingly, Kim et al. did
`not observe bands in the 1610-1620 cm-1 region for GLP-1
`precipitates induced by shaking or exposure to phenol,
`suggesting the transitions induced during rGLP-1(A) pro-
`cessing are different.11 Although in both cases, variants
`were produced with decreased solubility. The fact that the
`in vitro bioactivities of sGLP-1 and acid solubilized rGLP-
`1(A) were equivalent suggests that the structural transi-
`tions do not alter the GLP-1 receptor binding interactions
`or the peptide refolds upon acid solubilization and subse-
`quent dilution into the working concentration range of the
`bioassay (0.01-100 ng/mL).
`Portions of rGLP-1(A) were reprocessed to determine if
`the structural changes (along with the resultant insolubil-
`ity in neutral phosphate buffers) were reversible and
`whether they were due to chemical pertubations (presence
`
`of acetate contaminant), in-process pH-induced precipita-
`tion, or dehydration. The resultant FTIR spectra are
`shown in Figure 3B-D, and the corresponding estimates
`of secondary structural compositions are shown in Table
`1. Preparation of rGLP-1(B), by washing (to remove the
`acetate contaminant) and lyophilizing as a slurry, did not
`dramatically alter the major secondary structural elements
`or the resultant solubility. In fact, rGLP-1(B) exhibited
`less R-helix (reduced 1657 cm-1 band) and increased
`(cid:226)-sheet (increased 1647 cm-1 band) compared to rGLP-1(A).
`Preparation of rGLP-1(C) by washing, resolubilizing in 0.05
`M ammonium hydroxide, and lyophilizing as a solution
`induced structural changes resulting in a spectra more
`consistent with sGLP-1 (area of overlap ) 91.6%). The
`bands at 1620 and 1696 cm-1 were reduced or eliminated
`and the band at 1657 cm-1 was increased. These transi-
`tions are associated with formation of additional R-helix
`and less (cid:226)-sheet compared with sGLP-1. However, rGLP-
`1(C) still exhibited reduced solubility in neutral phosphate
`buffer (<1 mg/mL), suggesting that R-helial content alone
`is not a determinant of or a prerequisite for solubility.
`rGLP-1(D) was prepared by resolubilizing (unfolding) in 6
`M urea/1% acetic acid, followed by reverse phase chroma-
`tography and cation exchange chromatography (eluting
`directly into 0.05 M ammonium hydroxide, thereby elimi-
`nating the pH-induced precipitation step), and lyophilizing
`as a solution. The resultant FTIR spectra for rGLP-1(D)
`was almost identical with those of sGLP-1 (area of overlap
`) 97.5%), and rGLP-1(D) was soluble in neutral phosphate
`buffer (>10 mg/mL). This suggests that the in-process
`structural alterations were caused by pH-induced precipi-
`tation. Elimination of this step allowed for the preparation
`of excipient-free rGLP-1 powder having conformational and
`solubility properties equivalent to those of sGLP-1.
`
`Journal of Pharmaceutical Sciences / 185
`Vol. 87, No. 2, February 1998
`
`Novo Nordisk Ex. 2042, P. 3
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`Figure 4sAnalytical HPLC chromatograms indicating base-catalyzed degradation. Key:
`mg/mL; 0.10 M NH4OH; 37 (cid:176)C for 1 day).
`
`(A) rGLP-1 standard and (B) partially degraded rGLP-1 sample (5
`
`Purification of rGLP-1 by this scheme requires the
`peptide to be exposed to 0.05 M ammonium hydroxide (pH
`10.7). Although the resulting preparation was highly pure,
`preliminary HPLC studies indicated that GLP-1 can un-
`dergo base-catalyzed degradation to three primary (and
`several minor) degradation peak fractions (Figure 4). This
`can potentially lead to peptide impurities in the final
`product. Therefore, we conducted studies to identify the
`base-catalyzed degradation products, determine their ki-
`netic dependencies, and evaluate their in vitro bioactivity.
`The base-catalyzed degradation peak fractions, identified
`in Figure 4 as DPF-1, DPF-2, and DPF-3, were isolated by
`HPLC. Electrospray ionization mass spectrometry (ESI-
`MS) indicated that the peptides present in each degrada-
`tion peak fraction had equivalent molecular weight to
`rGLP-1. These findings ruled out hydrolysis, deamidation,
`oxidation, diketopiperazine formation, and succinimide
`formation as possible degradation pathways. However,
`amino acid racemization can occur in proteins and peptides
`following exposure to heat and alkali.16 Peptide variants
`containing racemized amino acids could not be identified
`by ESI-MS since there are no mass differences in the intact
`peptides. A combination of enzymatic digestion using LAP
`and MALDI-MS was used to determine whether rGLP-1
`variants containing racemized amino acids were present
`in the degradation peak fractions. LAP cleaves N-terminal
`L-amino acids at a higher kinetic rate than D-amino acids.
`Although the kinetic rate of cleavage varies for the L-amino
`acids, a peptide variant containing a D-amino acid can be
`identified by comparing its MALDI-MS spectra following
`LAP digestion with that of the control peptide. For
`instance, peaks corresponding to peptide fragments con-
`taining an N-terminal amino acid which was inefficiently
`cleaved would be evident in the MALDI-MS spectra.
`Therefore, MALDI-MS of a peptide containing a single
`D-amino acid following LAP digestion would indicate a peak
`corresponding to the mass of the peptide fragment contin-
`ing the N-terminal D-amino acid. Theoretically, the diges-
`tion would terminate at this point in the peptide sequence.
`
`186 / Journal of Pharmaceutical Sciences
`Vol. 87, No. 2, February 1998
`
`In our case (Figure 5), this is evident by comparing the
`spectra following LAP digestion of s(D-Ser8)GLP-1 with
`those of sGLP-1 and rGLP-1. The spectra of sGLP-1 and
`rGLP-1 are equivalent and indicate intense peaks corre-
`sponding to inefficient cleavage of glutamic acid (Glu3,
`Glu15) and glycine (Gly4, Gly16) N-terminal amino acids.
`Minor peaks are also evident corresponding to peptide
`fragments containing N-terminal threonine, aspartic acid,
`serine, and alanine. However, the resultant spectra for s(D-
`Ser8)GLP-1 indicates that the LAP digestion essentially
`terminates at the N-terminal D-Ser8 peptide fragment.
`Similar analysis of rGLP-1 degradation peak fractions
`(DPF-1, DPF-2, DPF-3) indicate that racemized product(s)
`is present in each isolate, since in each case the LAP
`digestion does not proceed to the extent of the sGLP-1 and
`rGLP-1 controls. Analysis of DPF-1 indicates significant
`peaks corresponding to peptide fragments containing N-
`terminal Thr7, Ser8, Val10, Ser11, and Ser12 amino acids.
`Likewise, analysis of DPF-3 indicates significant peaks
`corresponding to peptide fragments containing N-terminal
`Thr7, Ser8, Val10, and Ser11 amino acids. This suggests that
`DPF-1 and DPF-3 each contain several racemized GLP-1
`variants. The method is not capable of identifying more
`than one racemized amino acid in a given peptide, since
`the LAP digestion will terminate at the first N-terminal
`D-amino acid. Therefore, spectra showing multiple peptide
`fragments which are not consistent with the controls are
`indicative of several different racemized variants in the
`preparation. Analysis of DPF-2 suggests that this isolate
`is composed of a GLP-1 racemized variant(s) containing
`D-Ser11, since the LAP digestion essentially terminates at
`a peptide fragment containing an N-terminal Ser11. Our
`results indicate serine residues in GLP-1 are most sensitive
`to base-catalyzed racemization.
`As shown in Figure 6, the degradation peak fractions
`containing racemized GLP-1 variants were less bioactive
`than GLP-1 in the luciferase bioassay, and bioanalytical
`stability data were consistent with reverse phase HPLC.
`This indicates that the analytical methods were stability
`
`Novo Nordisk Ex. 2042, P. 4
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`Figure 5sMALDI-MS spectra of GLP-1 variants following LAP digestion. Each peak corresponds to a peptide fragment containing an N-terminal amino acid
`(indicated) which was inefficiently cleaved.
`
`indicating. Kinetic studies were carried out as a function
`of temperature, peptide concentration, and ammonium
`hydroxide concentration. No differences in racemization
`behavior were observed between sGLP-1 and rGLP-1
`variants. As shown in Figure 7, the degradation rate
`increases with increasing temperature and decreasing
`peptide concentration. The mechanism of amino acid
`racemization involves the removal of the R-proton of an
`amino acid, resulting in the formation of a planar carban-
`ion.17 Factors which determine the relative racemization
`rates of amino acids in proteins and peptides are the
`electron-withdrawing capacities of the various R-substit-
`uents, neighboring group effects, inherent structure-related
`
`effects, and matrix effects.17,18 GLP-1 is known to undergo
`self-association, leading to an oligomer with a higher helical
`content than the monomer.11,18
`Increases in solution
`viscosity were observed as peptide concentrations were
`increased from 5 mg/mL (nonviscous solution) to 25
`mg/mL (gel-like solution), suggesting that self-association
`increased as a function of peptide concentration, as ex-
`pected. The decrease in the racemization rate with in-
`creasing peptide concentration may be related to changes
`in intramolecular interactions of neighboring residues
`and/or steric constraints resulting from self-association.
`The degradation rate also increases as a function of
`ammonium hydroxide concentration (Figure 8). This was
`
`Journal of Pharmaceutical Sciences / 187
`Vol. 87, No. 2, February 1998
`
`Novo Nordisk Ex. 2042, P. 5
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`Figure 6sAnalysis of rGLP-1 degradation by HPLC and the luciferase
`bioactivity assay. Key:
`(A) comparison of pseudo-first-order plots of base-
`catalyzed rGLP degradation (5 mg/mL, 37 (cid:176)C, 0.05 M ammonium hydroxide)
`determined by HPLC and the luciferase bioassy and (B) response curves
`comparing the in vitro bioactivity of racemized rGLP-1 variants to rGLP-1.
`
`expected, since racemization rates in proteins are expected
`to be first-order in [OH-] at pH >9.19 Racemization in
`proteins at high temperature and pH extremes is not
`usually considered consequential in terms of protein stabil-
`ity in formulations.20 Chemical degradation pathways
`which compromise shelf life (T90 < 2 years at 25 °C) are
`those with high relative reactivity at pH 5-7, such as
`deamidation, aspartic acid isomerization, succinimide for-
`mation, oxidation, and disulfide formation.21 However,
`recombinant proteins and peptides are often exposed to pH
`extremes during their purification. Therefore, base-
`catalyzed racemization can impact the impurity profile of
`bulk recombinant peptide and protein products.
`These studies demonstrate the importance of conducting
`physicochemical characterization in parallel with process
`development of protein therapeutics. An understanding of
`the key elements of the process which impact the properties
`and purity of the bulk protein product is necessary for
`efficient and appropriate process validation and selection
`of suitable analytical methods, as well as identification of
`bulk product specifications. For instance, understanding
`the mechanisms and kinetics of rGLP-1 base-catalyzed
`degradation allows in-process storage time limits to be
`established and validated. Likewise, identifying the de-
`pendence of GLP-1 solubility properties on peptide confor-
`mation using FTIR analysis demonstrates that confirming
`the FTIR spectra of dry, excipient-free bulk rGLP-1 to be
`consistent with a reference is useful to ensure product
`equivalence and quality. Incorporation of physicochemical
`
`188 / Journal of Pharmaceutical Sciences
`Vol. 87, No. 2, February 1998
`
`Figure 7sPseudo-first-order plots compaing the effects of rGLP-1 concentra-
`(2) 5
`tion and temperature on degradation rates in 0.05 M NH4OH. Key:
`mg/mL, (b) 10 mg/mL, and (9) 25 mg/mL.
`
`Figure 8sPseudo-first-order plots comparing the effects of ammonium
`hydroxide concentration on rGLP-1 degradation rates (5 mg/mL; 30 (cid:176)C). Key:
`(2) 0.05 M NH4OH (pH 10.7); (b) 0.10 M NH4OH (pH 10.9); (9) 0.20 M
`NH4OH (pH 11.3).
`
`characterization studies in parallel with process develop-
`ment represents an expanded role for preformulation
`activities in the development of protein and peptide thera-
`peutics compared to conventional drugs. This distinction
`stems from the fact that the production and purification
`of recombinant proteins is carried out under aqueous
`conditions, whereas conventional organic drugs are typi-
`cally synthesized and purified in organic solvents. There-
`fore, as our studies with GLP-1 show, aqueous degradation
`and conformational transitions can effect the yields, purity,
`and properties of bulk protein and peptide products.
`
`Novo Nordisk Ex. 2042, P. 6
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`Acknowledgments
`We would like to thank John F. Carpenter (University of
`Colorado, College of Pharmacy) for use of his laboratory and
`assistance with FTIR analyses. We would also like to recognize
`Niels C. Kaarsholm, Svend Havelund, and Svend Ludvigsen (Novo
`Nordisk) for their efforts in characterizing the physical, biochemi-
`cal, and structural properties of sGLP-1 in solution, which included
`solubility, self-association, and solution structure analyses.
`
`References and Notes
`1. Senderoff, R. I.; Kontor, K. M.; Heffernan, J. K.; Clarke, J.
`K.; Garrison, J. K.; Kreilgaard, L.; Lasser, G. W.; Rosenberg,
`G. B. J. Pharm Sci. 1996, 85, 749-752.
`2. Welfle, K.; Misselwitz, R.; Schaup, A.; Geerlach, D.; Welfle,
`H. Proteins: Struct., Funct., and Genet. 1997, 27, 26-35.
`3. Thorens, B.; Waeber, G. Diabetes 1993, 42, 1219-1225.
`4. Adelhorst, K.; Hedegaard, B. B.; Knudsen, L. B.; Kirk, O. J.
`Biol. Chem. 1994, 269, 6275-6278.
`5. Turton, M..D.; O’shea, D.; Gunn, I.; Beak, S. A.; Edwards,
`C. M. B.; Meeran, K.; Choi, S. J.; Taylor, G. M.; Heath, M.
`M.; Lambert, P. D.; Wilding, J. P. H.; Smith, D. M.; Ghatei,
`M. A.; Herbert, J.; Bloom, S. R. Nature 1996, 379, 69-72.
`6. Overcashier, D. E.; Brooks, D. A.; Costantino, H. R.; Hsu, C.
`C. J. Pharm. Sci. 1997, 86, 455-459.
`7. Dong, A.; Caughey, W. S. Methods Enzymol. 1994, 232, 139-
`175.
`8. Kendrick, B. S.; Dong, A.; Allison, S. D.; Manning, M. C.;
`Carpenter, J. F. J. Pharm. Sci. 1996, 85, 155-158.
`
`9. Beavis, R. C.; Chaudhary, T.; Chait, B. T. Org. Mass
`Spectrom. 1992, 27, 156-158.
`10. Anderson, J. S.; Svensson, B.; Roepstorff, P. Nat. Biotechnol.
`1996, 14, 449-457.
`11. Kim, Y.; Rose, C. A.; Liu, Y.; Ozaki, Y.; Datta, G.; Tu, A. T.
`J. Pharm. Sci. 1994, 83, 1175-1180.
`12. Przybycien, T. M.; Bailey, J. E. Biochim. Biophys. Acta 1991,
`1076, 103-111.
`13. Prestrelski, S. J.; Tedeschi, N.; Arakawa, T.; Carpenter, J.
`F. Biophys. J. 1993, 65, 661-671.
`14. Ismail, A. A.; Mantsch, H. H.; Wong, P. T. Biochim. Biophys.
`Acta 1992, 1121, 183-188.
`15. Allison, S. D.; Dong, A.; Carpenter, J. F. Biophys. J. 1996,
`71, 2022-2032.
`16. Schwass, D. E.; Finley, J. W. J. Agric. Food Chem. 1984, 32,
`1377-1382.
`17. Neuberger, A. Adv. Protein Chem. 1948, 4, 298.
`18. Lambert, W. J.; Grucza, R. A.; Stamper, G. F.; Chrunyk, B.
`Pharm. Res. 1994, 11, S-83.
`19. Bada, J. L. Interdis. Sci. Rev. 1982, 7, 30-46.
`20. Powell, M. F. In Formulation, Characterization, and Stability
`of Protein Drugs: Case Histories, Pearlman, R.; Wang, Y. J.
`(Eds.), Plenum Press: New York, 1996; pp 1-12.
`21. Powell, M.F. In Formulation and Delivery of Proteins and
`Peptides; Cleland, J. L., Langer, R., Eds.; ACS Division of
`Biochemical Technology. IV. Series; American Chemical
`Society: Washington DC, 1994; p 112.
`JS9702729
`
`Journal of Pharmaceutical Sciences / 189
`Vol. 87, No. 2, February 1998
`
`Novo Nordisk Ex. 2042, P. 7
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`