Pharmaceutical Research, Vol. 17, No. 2, 2000
`
`Research Paper
`
`Metal-Catalyzed Oxidation of Brain-
`Derived Neurotrophic Factor
`(BDNF): Analytical Challenges for the
`Identification of Modified Sites
`
`Jana L. Jensen,1 Carl Kolvenbach,2 Suzanne Roy,3
`and Christian Scho¨neich1,4
`
`Received September 1, 1999; accepted December 10, 1999
`
`Purpose. We examined the metal-catalyzed oxidation of brain-derived
`neurotrophic factor (BDNF) using the Cu(II)/ascorbate/O2 model oxi-
`dative system.
`Methods. Electrospray ionization mass spectrometry, peptide mapping
`and amino acid analysis were utilized to determine the nature of the
`covalent modification induced by the metal-catalyzed oxidative system.
`Additionally, analytical ultracentrifugation, the Bradford assay, circular
`dichroism and ANSA dye-binding were used to determine the nature
`of any conformational changes induced by the oxidation.
`Results. Exposure of BDNF to the Cu(II)/ascorbate/O2 system led to
`the modification of ca. 35% of Met92 to its sulfoxide, and to subsequent
`conformational changes. The proteolytic digestion procedure was sensi-
`tive to this conformational change, and was unable to detect the modifi-
`cation. Chemical digestion with CNBr, however, was not sensitive to
`this change, and allowed for
`the identification of
`the site of
`modification.
`Conclusions. The modification of Met92 to its sulfoxide rendered the
`oxidized BDNF inaccessible to proteolytic digestion, due to conforma-
`tional changes associated with the oxidation.
`KEY WORDS: brain-derived neurotrophic factor (BDNF); metal-
`catalyzed oxidation (MCO); cyanogen bromide (CNBr); methionine
`sulfoxide; quality control; protein conformation.
`
`storage and delivery. One major pathway for covalent post-
`translational modification of proteins is oxidation; in particular,
`metal-catalyzed oxidation constitutes a significant problem for
`protein pharmaceuticals (1–4). Metals are present as contami-
`nates in water and buffers, and are used in various purification
`and refolding procedures during the production of recombinant
`proteins, e.g. metal-affinity chromatography (5–7). Transition
`metals can promote protein oxidation in the presence of prooxi-
`dants, oxygen or peroxides (2–4,8,9), which may lead to unde-
`sired loss of drug and/or enrichment of degradation products
`with unknown properties.
`Detection and characterization of oxidative modifications
`are necessary to ensure quality control of the biotechnology
`product. Additionally, characterization of the modification is
`critical for achieving an understanding of the oxidation mecha-
`nism, so that rational strategies for avoiding such reactions can
`be developed. While a variety of analytical techniques are used
`for these purposes, the predominant technique for monitoring
`the integrity of protein pharmaceuticals is peptide mapping (10).
`In this work, we examine the metal-catalyzed oxidation
`of a recombinant protein, brain-derived neurotrophic factor
`(BDNF), by the Cu(II)/ascorbate/O2 system (general mecha-
`nisms of metal-catalyzed protein oxidation are described in
`reference 8). We will demonstrate the inability of peptide map-
`ping to detect an oxidative modification of BDNF which occurs
`at significant levels (ca. 35% relative to total protein), and at
`lower prooxidant concentrations, specifically targets Met92.
`
`MATERIALS AND METHODS
`
`Materials
`
`INTRODUCTION
`
`Recent advances in biotechnology have led to the emer-
`gence of recombinant proteins as a significant class of therapeu-
`tics. The high degree of complexity of these molecules provides
`a myriad of possibilities for degradation during production,
`
`should
`
`be
`
`addressed.
`
`(e-mail:
`
`Recombinant BDNF was provided by Amgen, Inc., (Thou-
`sand Oaks, CA). The protein was exchanged into doubly dis-
`tilled deionized water (dd. H2O) via ultrafiltration using
`Microcon-3 microconcentrators (3 kD cutoff) from Amicon,
`Inc. (Beverly, MA). Typically, 500 ml of a 1 mg/ml BDNF
`solution in phosphate buffered saline (PBS), pH 7.4, was loaded
`onto a microconcentrator and centrifuged at 12,000 g for 50
`minutes, followed by three washes with 300 ml dd. H2O. The
`retentate was collected and the BDNF concentration determined
`by UV spectroscopy (Shimadzu UV-160 spectrophotometer;
`Kyoto, Japan) using an extinction coefficient of 1.76 cm2/mg
`at 280 nm. The BDNF solution was aliquotted in amounts of
`1 Department of Pharmaceutical Chemistry, University of Kansas, 2095
`270 mg/vial which were dried in a vacuum centrifuge (Labconco
`Constant Avenue, Lawrence, Kansas 66047.
`Centrivap Concentrator; Kansas City, Mo) and stored at 2708C
`2 Pharmaceutical Research and Development, Amgen Inc., Thousand
`for subsequent experiments. Sequencing grade endoproteinase
`Oaks, California 91320.
`Lys-C (Endo Lys-C) was obtained from Promega (Madison,
`3 Department of Product Development, Amgen Inc., Thousand Oaks,
`WI).
`8-Anilinonaphthalen-1-sulfonic
`acid
`(ANSA) was
`California 91320.
`obtained from Sigma (St. Louis, MO). For amino acid analysis,
`4 To whom correspondence
`schoneic@ukans.edu)
`4N methanesulfonic acid, phenyl isothiocyanate and Amino
`ABBREVIATIONS: BDNF, brain-derived neurotrophic factor; Endo Acid Standard H were from Pierce (Rockford, II) and ACS
`Lys-C, endoproteinase Lys-C; ANSA, 8-anilinonaphthalene-1-sulfonic
`grade NaOH (filtered prior to use) was from Fisher (Pittsburgh,
`acid; CNBr, cyanogen bromide; EDTA, ethylenediaminetetraacetic
`PA). Formic acid (99%) was from Fluka (Ronkonkoma, NY),
`acid; ESI-MS, electrospray ionization mass spectrometry; MALDI-
`and cyanogen bromide (CNBr), 99.995%, was from Aldrich
`TOF MS, matrix-assisted laser desorption ionization—time of flight
`(Milwaukee, WI). Experiments with CNBr were performed in
`mass spectrometry; HPLC/ESI-MS, reversed phase HPLC coupled on-
`a fume hood with proper safety attire, including gloves, goggles
`line to ESI-MS; SEC, size exclusion chromatography; SDS-PAGE,
`and a filtered mask. Waste from these experiments was disposed
`sodium dodecyl sulfate polyacrylamide gel electrophoresis; CD, circu-
`of in a manner appropriate for hazardous chemicals. All other
`lar dichroism; DSC, differential scanning calorimetry; a.m.u., atomic
`reagents were of the highest grade commercially available.
`mass units; SDS, sodium dodecyl sulfate.
`
`0724-8741/00/0200-0190$18.00/0 q 2000 Plenum Publishing Corporation
`
`190
`
`-190-
`
`
`
`
`MAIA Exhibit 1036
`MAIA V. BRACCO
`IPR PETITION
`
`

`

`Metal-Catalyzed Oxidative Modification of BDNF
`
`191
`
`Reaction Conditions
`Oxidation reactions were conducted on samples containing
`20 mM BDNF, a concentration at which BDNF exists entirely
`as a dimer (11). Unless noted otherwise, the oxidation of BDNF
`by the Cu(II)/ascorbate/O2 system was conducted under the
`following conditions: 20 mM BDNF/20 mM CuCl2/2 mM ascor-
`bate. Reactions were run in a volume of one milliliter in 20
`mM phosphate buffer, pH 7.4, and were incubated for one hour
`at 258C. The reagents were added in the following order: buffer,
`BDNF, CuCl2 and ascorbate. After addition of CuCl2, the mix-
`ture was incubated for five minutes at 258C to promote metal
`binding to the protein. The reaction was then started by the
`addition of ascorbate. For control samples, dd. H2O was added
`in place of ascorbate. In some cases, oxidation was stopped by
`the addition of a final concentration of 20 mM EDTA. All
`solutions were made with dd. H2O. The phosphate buffer stock
`solution was Chelex-treated (5 g Chelex to 100 ml buffer solu-
`tion; stirred for 1 hour, then filtered) to minimize metal contami-
`nation. Stock solutions of CuCl2 and ascorbate were freshly
`prepared prior to the reactions.
`
`Microcon-3 microconcentrators) and vacuum centrifugation.
`Approximately 270 mg BDNF was dissolved in 0.2 M Tris
`buffer, pH 8.5, containing 6 M guanidine-HCl and 1 mM EDTA.
`20 ml of 0.1 M dithiothreitol was added and the sample was
`incubated for one hour at 458C. After cooling to room tempera-
`ture, 40 ml of 0.1 M iodoacetic acid was added and the sample
`was placed in the dark for 30 minutes to allow for carboxymeth-
`ylation of the cysteine residues. The samples were then desalted
`by extensive ultrafiltration using Microcon-3 microconcentra-
`tors and dried by vacuum centrifugation. The reduced and alkyl-
`ated BDNF was reconstituted in 0.15 M ammonium bicarbonate
`buffer containing 2.2 M urea. 1.5% (w/w) Endo Lys-C was
`added and the mixture was incubated for 12 hours at 378C. At
`this point, an additional 0.75% (w/w) Endo Lys-C was added
`and the mixture incubated for an additional 4 hours at 378C.
`The primary structure of BDNF and the Endo Lys-C cleavage
`sites are shown in Fig. 1. This protease was found to give the best
`cleavage pattern for BDNF, and the most reproducible digestion.
`
`Chemical Digestion
`
`Samples were prepared for chemical digestion by acetone-
`Electrospray Ionization Mass Spectrometry (ESI-MS)
`HCl precipitation. Approximately 270 mg BDNF was dissolved
`in 900 ml of 70% formic acid. Cyanogen bromide was prepared
`Samples were prepared for ESI-MS analysis by acetone-
`HCl precipitation (12), followed by dissolution in dd. H2O as a 1 M solution in 70% formic acid, and 100 ml added to
`prior to analysis. ESI-MS experiments were performed on an
`the BDNF in formic acid. The sample (in a glass vial covered
`Autospec-Q tandem hybrid mass spectrometer (VG Analytical with aluminum foil) was incubated overnight at room tempera-
`Ltd., Manchester, UK), equipped with a Mark III ESI source
`ture, then vacuum dried. After digestion, samples were dena-
`and an OPUS data system. Samples were trapped and desalted
`tured, reduced and alkylated as in the proteolytic digestion
`prior to ESI by loading onto a trapping column (1.5 cm 3 1 mm procedure: samples were dissolved in 0.2 M Tris buffer, pH
`of polymeric beads with 400 A˚ pores, Michrom BioResources,
`8.5, containing 6 M guanidine-HCl and 1 mM EDTA. 20 ml
`Auburn, CA) with 0.1% acetic acid at 250 ml/min and eluting
`of 0.1 M dithiothreitol was added and the samples were incu-
`bated for one hour at 458C. After cooling to room temperature,
`the retained sample into the ESI source with 70% methanol/
`30% H2O with 0.1% acetic acid at 10 ml/min.
`40 ml of 0.1 M iodoacetic acid was added and the samples
`were placed in the dark for 30 minutes.
`
`Peptide Mapping by RP-HPLC
`
`Amino Acid Analysis
`Hydrolyses were performed on a Waters PICOxTAG Work
`Station (Waters, Milford, MA). Samples were prepared for
`Analysis of peptide fragments from the proteolytic diges-
`amino acid analysis by acetone-HCl precipitation (12), and
`tions of BDNF was performed on a Vydac C18 column (4.6 3
`were dissolved in formic acid prior to hydrolysis. Vapor-phase
`250 mm) (Vydac, Hesperia, CA) at 608C and 0.7 ml/min, using
`HCl hydrolysis was performed with 6 N HCl containing 0.1% UV detection at 215 nm. Typically, 1.5 nmol of protein was
`(w/v) phenol at 1108C for 24 hours. This method causes loss
`injected onto the column. A gradient from 0% to 50% acetoni-
`of Trp residues (12) and can cause conversion of methionine
`trile with 0.1% trifluoroacetic acid over 70 minutes was
`sulfoxide, Met(O), back to methionine (Met) (13–15). To quan-
`employed for the separation and elution of the peptide frag-
`titate Trp and Met (O), hydrolysis was also performed with 4 ments. Satisfactory signal-to-noise ratios for the separated frag-
`N methanesulfonic acid, which can readily determine Trp, Met ments were achieved for injections of 0.5 nmol protein and
`and Met(O) residues (16). Hydrolysis and neutralization were
`above. Samples which had been digested with CNBr and subse-
`conducted following the method of Simpson et al. (16). Briefly,
`quently denatured, reduced and alkylated were desalted on the
`protein samples were hydrolyzed with 4 N methanesulfonic Vydac C18 column. The resulting fragments were collected
`acid at 1158C for 22 hours, then neutralized with 4 N NaOH together and submitted as a mixture for analysis by ESI-MS.
`prior to derivatization. Hydrolyzed amino acids from both pro-
`cedures were derivatized with phenyl isothiocyanate and sepa-
`rated on a Waters Spherisorb S5ODS2 column (4.6 3 250 mm)
`(Waters, Milford, MA) at 458C and 0.8 ml/min, using UV
`detection at 254 nm. For complementary analysis, samples were
`also sent to Commonwealth Biotechnologies, Inc. (Richmond,
`VA) for analysis by alkaline hydrolysis with NaOH.
`
`Proteolytic Digestion
`Samples were prepared for proteolytic digestion by ultrafil-
`tration (three dd. H2O rinses at 12,000 g for 50 minutes using
`
`Fig. 1. Primary structure of BDNF. Three disulfide bonds exist: Cys13
`(S-S-)Cys80, Cys58(-S-S-)Cys109 and Cys68(-S-S-)Cys111. Endo Lys-C
`cleavage sites are indicated with arrows, and the positions of the four
`Met residues are indicated.
`
`-191-
`
`
`
`
`

`

`192
`
`Jensen, Kolvenbach, Roy, and Scho¨neich
`
`MALDI-TOF Mass Spectrometry
`
`Fluorescence Spectroscopy of 8-Anilinonaphtalene-1-
`sulfonic Acid (ANSA)
`
`Microcon-3 microconcentrators). Far-UV CD spectra (185–260
`nm) were obtained on a Jasco J-720 spectropolarimeter (Jasco,
`Matrix-assisted laser desorption ionization-time of flight
`Easton, MD), using a 0.1 cm path length quartz cell. The spectra
`(MALDI-TOF) mass spectra were obtained on a Hewlett Pack- were recorded at 258C at a concentration of approximately 100
`ard model G2025A (Hewlett Packard, Palo Alto, CA), which mg/ml. Measurements were taken at intervals of 0.5 nm with
`typically measures molecular masses below 3 kDa with an
`a speed of 20 nm/min, and were averaged over 3 accumu-
`accuracy of at least 0.1%. Peaks from the peptide map were
`lated scans.
`collected and vacuum centrifuged to dryness. Samples were
`reconstituted in acetonitrile/isopropanol/0.1% trifluoroacetic
`acid (1:3:2 (v/v/v)). The matrix used was a standard solution
`of sinapinic acid (3,5-dimethoxy-4-hydroxycinnamic acid) in
`acetonitrile/H2O (1:1 (v/v)).
`
`Fluorescence spectra were obtained using a Photon Tech-
`nology International (Monmouth Junction, NJ) spectrofluoro-
`meter (Quanta Master Luminescence Spectrometer, QM1).
`ANSA (100 mM) containing 20 mM EDTA, was incubated with
`RP-HPLC coupled on-line to ESI mass spectrometry was
`10 mM BDNF (from either an oxidized or control sample) in
`used to analyze the peptide maps. The reverse-phase assay for
`20 mM sodium phosphate buffer, pH 7.4, at 378C for 30 minutes.
`peptide mapping was optimized at a flow rate of 50 ml/min on
`The fluorescence emission spectra (excitation, 370 nm) were
`a 13 250 mm Vydac C18 column (Vydac, Hesperia, CA), monitored between 400 and 600 nm. Binding of ANSA to
`which was coupled to the Autospec-Q tandem hybrid mass
`protein was determined by subtracting the emission spectrum
`spectrometer. The column effluent was monitored with UV of ANSA from that of ANSA in the presence of BDNF.
`detection at 215 nm, and post UV detector, 8 ml/min were split
`off into the ESI source. Mass spectral data were collected and
`analyzed on the OPUS data system.
`
`HPLC/ESI-MS
`
`RESULTS
`
`Mass Spectrometry
`
`Analytical Ultracentrifugation
`
`ESI-MS experiments were performed to determine if expo-
`Sedimentation equilibrium and velocity experiments were
`sure to the Cu(II)/ascorbate/O2 system caused any covalent
`performed on a Beckman XL-l analytical ultracentrifuge (Beck- modification of BDNF. The spectrum in Fig. 2a reveals a signifi-
`man Coulter, Inc., Fullerton, CA), equipped with an eight cell
`cant amount of product (ca. 35% of total protein) with a molecu-
`rotor. Samples were dialyzed extensively against 10 mM sodium lar weight of 13652 6 1, i.e. 16 a.m.u. higher than that of the
`phosphate buffer, pH 7.4, at 48C prior to centrifugation. For
`the sedimentation equilibrium experiments, samples were
`diluted with 10 mM sodium phosphate buffer, pH 7.4 to concen-
`trations of 5, 10 and 20 mM. Samples were centrifuged at three
`different speeds (15,000, 20,000 and 27,000 rpm) for 24 hours
`at 208C. Equilibrium was confirmed by monitoring the UV
`absorbance at 229 nm, measured every 5 hours. Raw data were
`imported into the software program Kdalton (developed by
`John Philo, Amgen) for analysis via nonlinear curve fitting.
`For sedimentation velocity experiments, samples of 20 mM
`BDNF were equilibrated to 258C and centrifuged at 40,000
`rpm. Sedimentation was monitored using UV absorbance at
`229 nm. The data generated was fit to a single species model
`using the modified Fujita-MacCosham function in the software
`program SVEDBERG, version 6.10 (developed by John
`Philo, Amgen).
`
`Protein Concentration
`
`Soluble protein concentration was determined using the
`CoomassieTPlus Protein Assay Reagent from Pierce (Rockford,
`II), which is a modification of the Bradford method (17). Sample
`analysis involved the addition of the Coomassie dye to the
`protein solution, with subsequent measurement of absorbance at
`595 nm on a Shimadzu UV-160 spectrophotometer (Shimadzu,
`Kyoto, Japan).
`
`Circular Dichroism Spectroscopy
`
`Samples were prepared for CD experiments by ultrafiltra-
`tion (three dd. H2O rinses at 12,000 g for 50 minutes using
`
`Fig. 2. ESI-MS spectra for oxidized and control BDNF. (a) Oxidized;
`(b) control.
`
`-192-
`
`
`
`
`

`

`Metal-Catalyzed Oxidative Modification of BDNF
`
`193
`
`native protein (M.W. 5 13636), for the sample exposed to 20
`denatured, reduced and alkylated as described in the Experimen-
`mM CuCl2 and 2 mM ascorbate. This modified species is absent
`tal section. The sample was then prepared for ESI-MS experi-
`in the control sample (Fig. 2b). The existence of an M 1 16 ments via ultrafiltration,
`followed by an acetone-HCI
`peak in the spectrum of the oxidized sample is indicative of
`precipitation. ESI-MS spectra clearly showed oxidized carboxy-
`methylated BDNF (M.W. 5 14006) at a level of ca. 35% of
`BDNF modification via a single oxygen addition.
`We investigated the possibility that the oxidation seen in
`total protein, similar to that seen for BDNF exposed to metal-
`catalyzed oxidation without further reduction and alkylation.
`the ESI-MS data was an artifact of the procedure itself, as it
`has been reported that oxidation of analytes can occur during
`The observed molecular mass indicated that the oxidized BDNF
`ionization in the mass spectrometer (18). Although acetone- was completely reduced and alkylated at all six cysteine resi-
`HCl precipitation of samples was performed prior to ESI-MS
`dues. This control effectively eliminates the possibility that the
`analysis, it is possible that minute amounts of ascorbic acid 116 modification on the oxidized BDNF was reduced during
`and CuCl2 remained in the sample and promoted oxidation
`sample preparation for enzymatic digestion.
`A discrepancy exists between ESI-MS, which showed clear
`during ionization. In order to rule out this possibility, a control
`experiment was performed. A sample containing 20 mM BDNF
`evidence for oxidized BDNF, and proteolytic mapping, which
`and 20 mM CuCl2 in 20 mM phosphate buffer was prepared
`could not resolve any product peaks. A possible rationale for
`and incubated at room temperature for one hour, after which
`this discrepancy is that oxidized BDNF is not accessible to
`Endo Lys-C for proteolytic digestion due to aggregation and/
`an acetone-HCl precipitation was performed. Immediately prior
`to analysis by ESI-MS, a solution of 2 mM ascorbic acid was
`or conformational changes, or likewise, oxidatively modified
`proteolytic fragments may undergo aggregation or adhesion to
`added to the sample. No evidence for oxidation of the BDNF
`was seen under these conditions (data not shown). Clearly the
`surfaces. A range of analytical techniques was employed to
`116 modification seen in the oxidized BDNF sample did not
`determine the nature of any potential structural changes induced
`occur during ionization in the mass spectrometer.
`by the oxidation of BDNF, as described below.
`
`Amino Acid Analysis
`
`Analytical Ultracentrifugation
`
`Results from HCl hydrolysis of oxidized BDNF and its
`control revealed no change in the content of any of the amino
`acid residues. One amino acid particularly sensitive to metal-
`catalyzed oxidation is His (2,3,8). In order to control whether
`HCl hydrolysis would have indicated the potential loss of His
`from BDNF, we subjected BDNF to more extensive oxidation
`by 40 mM Cu(II) and 4 mM ascorbate. Subsequent to HCl
`hydrolysis, amino acid analysis indicated a loss of up to 63%
`of His (relative to the initial content). Hence, our experimental
`method would have indicated the loss of His if it had occurred
`under our original reaction conditions employing 20 mM Cu(II)
`and 2 mM ascorbate. After oxidation by 20 mM Cu(II) and
`2 mM ascorbate, methanesulfonic acid and NaOH hydrolysis
`revealed some loss of Met (ca. 5%), and the formation of Determination of Soluble Protein
`Met(O) (ca. 16% relative to the original content of Met). These
`results do not quantitatively agree with those from ESI-MS
`analysis; however, they do provide qualitative identification of
`the modified amino acid residues and indicate that Met oxida-
`tion is the only detectable chemical modification.
`
`The possibility of aggregation beyond the native dimer
`was investigated by analytical ultracentrifugation. Results from
`sedimentation equilibrium experiments revealed that BDNF
`existed as dimer under the conditions used for the oxidation
`reactions, and also showed that BDNF remained entirely in the
`dimer form after oxidative modification. Data from sedimenta-
`tion velocity experiments confirmed the existence of a single
`species in all samples, with molecular weights corresponding
`to that of the dimer. No evidence for the presence of multimers
`was seen by either ultracentrifugation technique, indicating that
`metal-catalyzed oxidation of BDNF did not induce further
`aggregation.
`
`The possible loss of soluble protein was investigated using
`a modified Bradford assay. No loss of soluble protein occurred
`upon oxidation of BDNF.
`
`Circular Dichroism
`
`Proteolytic Digest
`
`Protein secondary structure can be assessed through the
`In order to localize the chemical modification to a particu-
`use of far-UV CD spectroscopy. Figure 4 compares the CD
`lar sequence in the protein, oxidized BDNF was subjected to
`spectra of control and oxidized BDNF. Native BDNF has been
`estimated to consist of ca. 74% b-sheet and 21% b-turn struc-
`proteolytic digestion, followed by RP-HPLC separation of the
`proteolytic fragments. All fragments of the native protein,
`ture, based on deconvolution analysis of the circular dichroism
`except fragment K12 (Arg117-Arg119), were easily identified via
`spectrum (11). The broad negative signal between 190 and 220
`nm has been attributed to a combination of b-sheet and b-turn
`on-line ESI-MS or peak collection followed by MALDI-TOF
`MS. The peak assignments are given in Fig. 3. Neither procedure
`structure (11,19,20), while the distinctive peak at 232 nm has
`showed evidence for a fragment being 16 a.m.u. heavier than the
`been attributed to both tertiary structural elements (20,21) and
`a combination of secondary b-structure elements (20). Oxidized
`identified native sequences of BDNF for the oxidized sample.
`We investigated the possibility that the modified Met resi- BDNF did not show a significant change in the peak at 232
`due (Met(O)) in oxidized BDNF was reduced during sample
`nm. However, a slight increase was seen in the magnitude of
`the negative peak at 190–220 nm for the oxidized sample
`preparation procedures prior to the enzymatic digestion. A sam-
`relative to the control, suggesting an increase in b-structure.
`ple which was exposed to the Cu(II)/ascorbate/O2 system was
`
`-193-
`
`
`
`
`

`

`194
`
`Jensen, Kolvenbach, Roy, and Scho¨neich
`
`Fig. 3. Endo Lys-C maps of oxidized (upper) and control (lower) BDNF, monitored by RP-HPLC at 215 nm.
`
`ANSA Binding
`
`or a change in the local environment of the bound ANSA which
`enhances its fluorescence. Derivative analysis of the spectra
`ANSA is one of a class of fluorescent probes that has been
`revealed a blue shift of ca. 10 nm in the peak maximum upon
`used to study the surface hydrophobicity of proteins (22,23). metal-catalyzed oxidation, indicating that the ANSA is located
`It was thought that the intense fluorescence exhibited by the
`in a more apolar or hydrophobic region in the oxidized BDNF
`dye in the presence of proteins was due to hydrophobic interac-
`than in the control BDNF. Although ANSA may induce further
`tions, i.e. binding, between the naphthalene moiety and exposed
`conformational changes upon binding to the protein (induced-
`hydrophobic sites on proteins. Recent work, however, has
`fit binding model) (24,26), it is evident that metal-catalyzed
`shown that ANSA binding is actually dominated by electrostatic
`oxidation of BDNF results in conformational changes which
`forces between the sulfonate group of ANSA and exposed
`promote ANSA binding.
`cationic groups on the protein (24,25). Furthermore, ANSA
`fluorescence upon binding to a protein depends on many vari-
`ables, of which hydrophobicity is only one. Nonetheless, bind-
`ing and subsequent fluorescence of ANSA can indicate changes
`in overall structure of the protein molecule. Here, ANSA bind-
`ing was used to determine if any changes in 38 structure were
`induced upon oxidation. Figure 5 shows the fluorescence spectra
`of ANSA in the presence of non-oxidized and oxidized BDNF.
`The increase in ANSA fluorescence intensity suggests an oxida-
`tion-dependent conformational change of BDNF resulting in
`either exposure of additional sites to which ANSA can bind,
`
`Alternate Approaches Using Denaturants
`
`Clearly BDNF experienced structural modifications upon
`oxidation by the Cu(II)/ascorbate/O2 oxidative system. Oxi-
`dized BDNF may exist in a conformation which is not fully
`accessible to Endo Lys-C for proteolytic digestion. Two
`approaches were taken to overcome this problem. One approach
`involved increasing the urea level to 4 M during the digestion
`procedure, in an effort to unfold the protein to a greater extent.
`
`Fig. 4. Far-UV CD spectra of oxidized (---) and control (—) BDNF.
`(Molar ellipticity: [Q], dec cm2 dmol21).
`
`Fig. 5. Fluorescence spectra of ANSA in the presence of oxidized
`(---) and control (—) BDNF.
`
`-194-
`
`
`
`
`

`

`Metal-Catalyzed Oxidative Modification of BDNF
`
`195
`
`As expected, the integrity of the Endo Lys-C was compromised
`and only partial digestion was achieved. The second approach
`utilized sodium dodecyl sulfate (SDS) as a solubilizing agent
`in place of guanidine/HCI and urea during the denaturation,
`reduction, alkylation and subsequent digestion procedure. New
`peaks were detected in the peptide map of the oxidized sample
`relative to the control; however, these could not be identified
`by ESI-MS or MALDI-TOF mass spectrometry, likely due to
`contamination of the samples with detergent and salt.
`
`Chemical Digestion
`
`Results from amino acid analysis of oxidized BDNF indi-
`cated the formation of Met(O). Accordingly, a chemical diges-
`tion procedure which is sensitive to this oxidative modification
`was employed. Oxidized and control BDNF samples were
`chemically digested using CNBr, which cleaves on the C-termi-
`nal side of Met residues, but is incapable of cleavage at Met(O)
`residues (27). ESI-MS analysis of the CNBr digest mixture
`showed all of the expected fragments for the control and oxi-
`dized samples. Several of these fragments were observed in the
`mono or di-formylated form. Formylation of CNBr fragments
`generated in 70% formic acid has been reported (28). In addition
`to the expected CNBr fragments, the oxidized sample also
`contained a peak of M.W. 5 7052 6 1, representing a mono-
`([Gly62-
`oxidized non-formylated combination fragment
`Arg119](O); theoretical average M. W. 5 7053.08). This sug-
`gested that Met92 was oxidized to Met(O). Significantly, no
`evidence for 116 modifications of the other fragments was
`seen, indicating that all other amino acid residues remained
`unmodified in this system.
`
`DISCUSSION
`
`In the pharmaceutical industry, determination of post-
`translational protein modifications is critical for assessing the
`integrity of biotechnology products. A range of standard quality
`control analyses exists for this purpose, and chief among these
`is proteolytic peptide mapping (10). Clearly the analysis of
`metal-catalyzed oxidative modifications of BDNF by this tech-
`nique is problematic. For oxidized BDNF, the proteolytic pep-
`tide map was unable to detect the oxidative modification,
`although this modification occurred at Met92 at levels of ca.
`35% as seen by ESI-MS and was identified as a Met fi Met(O)
`modification by amino acid analysis and CNBr digest. Conse-
`quently, this inadequacy represents a potentially severe prob-
`lem: failure of the quality control system.
`As proteolytic digestion of proteins generates a large num-
`ber of fragments, peptide maps are often quite complicated,
`and complete resolution of all peaks in a reasonable time frame
`may be difficult if not impossible. Therefore, the possibility
`that the oxidized BDNF fragment was co-eluting with another
`fragment was addressed by conducting experiments with RP-
`HPLC coupled on-line to ESI mass spectrometry. No evidence
`for a BDNF proteolytic fragment bearing a 116 modification
`was seen.
`Control experiments verified that the BDNF modification
`seen by ESI-MS after exposure to the Cu(II)/ascorbate/O2 oxi-
`dative system was not an artifact of the ionization procedure.
`Additionally, it was verified that oxidized BDNF was not
`reduced during sample preparation procedures prior to proteo-
`lytic digestion and subsequent peptide mapping.
`
`Analytical ultracentrifugation showed no evidence for the
`presence of aggregates beyond the dimer. Furthermore, results
`from the modified Bradford assay revealed no loss of soluble
`protein upon oxidation.
`Clearly, oxidized BDNF was not lost via chemical reduc-
`tion or physical aggregation; neither did loss of oxidized BDNF
`occur due to decreased solubility. A remaining possibility was
`that the oxidized BDNF was not fully accessible for digestion
`by Endo Lys-C as a result of conformational changes induced
`by oxidation of the Met residue. The conformational integrity
`was assessed by CD and ANSA dye binding.
`Evaluation of BDNF secondary structure by far-UV CD
`indicated a slight increase in b-structure upon oxidation. Such
`behavior has also been observed for other peptides and proteins,
`e.g. human relaxin (2). Dado et al. (29) demonstrated the ability
`of Met to behave as a conformational “switch”: oxidation of
`the Met residue to Met(O) caused conversion of the peptide 28
`structure from a-helix to b-sheet.
`Although BDNF 28 structure increased upon oxidation,
`the 38 structure was modified to some degree as evidenced by
`results from the ANSA experiments. Metal-catalyzed oxidation
`of BDNF led to the exposure of ANSA binding sites. It is not
`known if this unfolding occurred in a localized manner at a
`specific site or sites within the protein, or if it occurred in a
`more globalized fashion as a result of loss of contact between
`b-sheets or the monomer units.
`Clearly, the conformational integrity of BDNF was com-
`promised upon metal-catalyzed oxidation. The Endo Lys-C
`digestion procedure seemed to be sensitive to this conforma-
`tional change, while the CNBr digestion procedure was appar-
`ently not. BDNF contains 11 Lys residues which are dispersed
`throughout the protein sequence. Examination of the crystal
`structure of the BDNF:Neurotrophin-3 heterodimer (30) (Pro-
`tein Data Bank file 1BND) indicates that these residues are
`primarily located on the surface of the native BDNF dimer.
`Presumably, the conformational changes induced by oxidation
`cause a positional shift in one or more of these residues, render-
`ing them inaccessible to cleavage by Endo Lys-C. It is intriguing
`that these Lys residues remained inaccessible even when the
`reduced and alkylated BDNF was exposed to 2.2 M urea during
`the digestion procedure. Increased levels of denaturant may well
`unfold BDNF completely, but these conditions compromised the
`proteolytic activity of Endo Lys-C. SDS appeared to success-
`fully “solubilize” the oxidized BDNF so that it could be properly
`digested; however, this detergent interfered with the mass spec-
`trometric analyses.
`We have demonstrated the inability of proteolytic diges-
`tion/peptide mapping to detect a covalent modification which
`occurred at levels of ca. 35% of native BDNF. It was verified
`that this modification was induced by metal-catalyzed oxidation
`via the Cu(II)/ascorbate/O2 model system. By CNBr digestion,
`it was shown that BDNF was modified by conversion of Met92
`to Met(O). Spectroscopic analyses revealed that this chemical
`modification induced definite structural changes in the protein,
`causing certain cleavage sites in the protein to become inaccessi-
`ble to proteolytic enzymes, even under denaturing conditions.
`This study stresses the need for a battery of analytical tests
`that will not only exhaustively detect modified and/or degraded
`product