Confocal Raman Characterization of Different Protein Desorption

Dec 30, 2013 - Confocal Raman Characterization of Different Protein Desorption Behaviors from Chromatographic Particles. Yuewu Xiao*†, Thomas Stoneâ...
0 downloads 0 Views 1MB Size
Article pubs.acs.org/ac

Confocal Raman Characterization of Different Protein Desorption Behaviors from Chromatographic Particles Yuewu Xiao,*,† Thomas Stone,† Wilson Moya,† Paul Killian,† and Thomas Herget‡ †

EMD Millipore, 80 Ashby Road, Bedford, Massachusetts 01730, United States Merck Millipore, Frankfurter Strasse 250, D 64293 Darmstadt, Germany



ABSTRACT: Confocal Raman spectroscopy is a nondestructive analytical technique that combines the chemical information from vibrational spectroscopy with the spatial resolution of confocal microscopy. It was applied, for the first time, to measure protein desorption from chromatographic particles. Monoclonal antibody was loaded onto the Fractogel EMD SO3 (M) cation exchanger at either pH 5 or pH 4. Confocal Raman measurement suggests that only the protein loaded at pH 5 is able to release from chromatographic particles in the elution buffer. Detailed comparison of highquality spectra indicates that, while proteins loaded at both pH values showed a predominant β-sheet conformation, protein loaded at pH 4 has a broader amide I band with more intensity in the >1680 cm−1 region. This small but clear and reproducible amide I bandwidth increase is not observed for protein in the solution state at pH 4. No definitive assignment of the increased Raman intensity in the >1680 cm−1 region could be made, but it might be related to structural changes involved in the association of protein molecules in the adsorbed state, which helps to explain the nearly 100% retention under elution conditions of the monoclonal antibody adsorbed at pH 4 in chromatographic particles.

protein and the medium surface could conceivably lead to preferred protein-stationary phase interactions.21−24 Protein reorientation from a space-consuming to space-saving geometry has been observed on Fractogel EMD SO3 (M), a tentacletype25 cation exchanger, to explain the higher maximum binding capacities compared to other media.23 Characterization of protein in the eluate fractions from a chromatographic column helps to understand the adsorptioninduced effects. If the unfolded species is slow to refold in the eluate or the intermediate state is preserved after protein release from chromatographic particles, it can be differentiated from the native form using ultraviolet circular dichrosim spectroscopy (UV-CD), 6, 26 fluorescence spectroscopy,2,12,23,24,26 second-order derivative UV spectroscopy26−28 and light scattering.16,17 Other approaches rely on the in situ amide hydrogen/deuterium (H/D) isotope labeling while protein is still bound to chromatographic particles. The exchanged deuterium is then characterized by nuclear magnetic resonance (NMR) spectroscopy7−9 or mass spectrometry (MS)1,5,6,10,11 to determine protein conformational stability. More specifically, the fully solvent exposed amino acids on the protein surface have a characteristic H/D exchange time of ∼6 s, up to 108 faster than those inside the protein.29 Change in the hydrogen-exchange behavior of a particular amino acid with a

Protein adsorption onto chromatographic particles may cause a number of effects including conformational changes,1−15 aggregation,14−20 preferred binding orientation,21−24 and/or different combinations of mixed mode interactions. Conformational changes have been studied most especially with hydrophobic interaction chromatography (HIC) and reversephase chromatography (RPC). In a simple scenario, native protein, particularly one with backbone flexibility, may unfold to form multiple conformational states during passage through a column.2−7 These conformers with distinct adsorption behaviors lead to broad, shouldered, tailed or multiple peaks in the elution profile.2−7 The least retained species is usually native protein; whereas those eluting later are unfolded forms due to an increased stationary-phase exposure with more interaction of their hydrophobic residues.2−7 Further studies indicate that the correlation between retention and unfolding is complicated,8−12 which can be affected by protein desorption kinetics, protein rigidity and chromatographic particle pore structures. Another mechanism to explain longer retention times is adsorption-induced protein aggregation. Elution usually requires disruption of lateral intermolecular interactions.14,15 Preferred protein orientation on the particles is based on the consideration that most protein properties, e.g., surface morphology, ionic and hydrophobic patches, are not uniformly distributed across the protein exterior. The spatial variations may also be true for chromatographic particles in terms of pore structure, ligand density and other surface chemistry features. The heterogeneous distribution of functional groups on the © 2013 American Chemical Society

Received: April 12, 2013 Accepted: December 30, 2013 Published: December 30, 2013 1007

dx.doi.org/10.1021/ac401081s | Anal. Chem. 2014, 86, 1007−1015

Analytical Chemistry

Article

deuterated solvent can therefore indicate whether the residue has moved from the interior of the molecule to the exterior, or vice versa. In practice, native and unfolded proteins adsorbed onto a chromatography column have different isotope labeling kinetics. A number of methods9,30 are available to quench the exchange reaction to preserve the specific exchange-labeled state in the eluate for NMR and/or MS analysis. With these various solution-state spectroscopy techniques, enormous information has been obtained about protein adsorptioninduced effects that depend on selection of chromatographic particles,4−11 loading and elution conditions,12−16 and type of protein.4,10−12 However, one concern is that the postexposure characterization may not always adequately reflect the protein state when it is actually interacting with the media.10,15 There are a few reports of in situ characterization of protein conformational changes in chromatographic particles. Although CD is an excellent tool to analyze protein on nanometer-sized particles that are stable in suspension or solution,31−33 its use is limited for micrometer-sized chromatographic particles due to difficulties in minimizing light-scattering effects. Fluorescence spectroscopy differentiates native and unfolded adsorbed proteins based on its ability to monitor the microenvironments in which tryptophan and tyrosine are occupied.2,12,34 A red shift of the emission maximum indicates that amino acids buried in the interior have been exposed to the hydrophilic solvent due to protein unfolding.2,12,34 As fluorescence spectroscopy generally does not provide a high spectral resolution, interpretation of the result becomes difficult when the protein contains multiple tryptophan or tyrosine residues. Recently, vibrational spectroscopy has been applied to quantify secondary structural components and thus conformational changes of protein in chromatographic particles. When Raman spectroscopy was applied to characterize lysozyme in alkyl-bonded silicate media15 and α-lactalbumin in a HIC column,6 it agrees with the solution-state NMR or UV-CD approach6,9 that two proteins in the adsorbed state are typically depleted in the helix content and enriched in the sheet content. Similar α-helix to βsheet conversion is detected by FTIR for BSA in a HIC column14 and HSA in a C6 medium.35 Part of the new conformation converted from α-helix is presumably attributed to the intermolecular β-sheet of protein aggregate in chromatographic particles.14,15 Model proteins used in vibrational spectroscopy studies have significant amount of α-helix structure that is easily differentiated from intermolecular/ intramolecular β-sheet in the amide I frequency. Raman spectroscopy enjoys several important advantages in characterizing protein. First, it can be applied to protein in varied states: aqueous solution, precipitated fibril, amorphous aggregate, solid, and crystal. The spectra can be directly compared to estimate different conformations present in these states.36 Second, the H2O bending vibration that obscures the amide I band in IR spectroscopy has a low intensity in Raman spectra, reducing the error inherent in large background solvent subtraction. Third, in addition to characterization of the amide I and other amide vibrations, additional features are present in Raman spectra that help to monitor the environment of numerous amino acid side chains, including aromatic amino acids, acidic residues, and sulfur containing residues.37,38 Furthermore, when Raman spectroscopy is carried out in the confocal mode, it provides chemical information with μm-sized spatial resolution. For example, confocal Raman spectroscopy has recently been applied in our lab to measure protein distribution in semiopaque chromatographic particles.39

In this paper, we applied confocal Raman spectroscopy to measure protein desorption from the strong cation exchanger Fractogel EMD SO3 (M). The medium used in this report was characterized to have high dynamic binding capacities of monoclonal antibodies (mAbs) under optimized conditions that include the loading pH at 5.40 Similar observations of high dynamic binding capacities were also reported when a mAb, EMD Millipore mAb04, was loaded at pH 5 onto the Fractogel EMD SO3 (M) medium.41 It was also reported that when mAb04 was loaded at pH 4, little or no protein was observed to release from the chromatographic particles under elution conditions (pH 6, [Na+] gradient from 0 to 1 M, superficial velocity 100 cm/h).41 To better understand the impact different loading pH values have on the elution profiles, we loaded mAb04 onto the cation exchanger at either pH 5 or pH 4 and then applied confocal Raman spectroscopy to measure protein conformation. We find that protein loaded at pH 4 shows small but clear and reproducible Raman features that may be related to structural changes involved in protein association. The presence of oligomers instead of monomers helps to explain the irreversible binding character between protein adsorbed at pH 4 and the chromatographic particles.



EXPERIMENTAL SECTION Reagents and Materials. The cation exchanger, Fractogel EMD SO3 (M) (EMD Millipore, Billerica, MA), was supplied in 20% ethanol with 0.15 M NaCl. The resin was washed extensively with Milli-Q water and then equilibrated with a loading buffer of either pH 5 or 4. The buffer solution, 15 mM acetic acid (pKa = 4.76) and sodium acetate, was adjusted with a saturated NaCl solution to have a conductivity of 2 mS/cm (∼20 mM [Na+]) measured using an Oakton CON 11 standard conductivity meter (Eutech Instruments). Monoclonal antibody used for this work, EMD Millipore mAb04 (isoelectric point or pI = 8.2), was expressed in a Chinese Hamster ovary cell culture and purified by Protein A affinity chromatography. Protein purity, ∼99%, was measured by Protein A HPLC using a POROS A HPLC (Applied Biosystems) column on an Agilent 1260 HPLC platform. Prior to use, the protein solution was dialyzed against the respective loading buffer and then diluted to a concentration of 2 mg/mL. Protein concentration was determined by absorbance at 280 nm with an extinction coefficient of 1.532 mL mg−1 cm−1. Protein desorption from the chromatographic particles was measured by application of the elution buffer (15 mM bis-tris, pH 6 with 1 M NaCl). Two buffers for analysis of the solution-state protein were prepared in a slightly different way from the protein loading buffer. In one case, NaClO4 replaced NaCl for conductivity adjustment; and in the other case, D2O replaced H2O as the solvent. For buffer exchange, protein solution in the loading buffer was concentrated 60-fold using an Amicon Ultra-50K centrifugal filter (EMD Millipore) and an Eppendorf centrifugal concentrator 5810R (Alert Scientific) at 4000g and 4 °C, then the original sample volume was restored with the changing buffer; this concentration/washing step was repeated a minimum of three more times to exchange the loading buffer thoroughly. Protein Loading. About 40 μL of gravity settled Fractogel EMD SO3 (M) particles was suspended in 35 mL of mAb04 solution (2 mg/mL, 2 mS/cm, pH 5 or 4) to ensure an excess of protein for binding. Protein and the chromatographic particles were gently mixed for 24 h at room temperature in a 50 mL reaction tube by end-over-end rotation. Samples were 1008

dx.doi.org/10.1021/ac401081s | Anal. Chem. 2014, 86, 1007−1015

Analytical Chemistry

Article

in the buffer mainly due to their gravity and van der Waals interaction between the particles and the substrate, with minor contribution from the optical trapping effects described by transferring the momentum of the incident radiation to the particle.43−47 As described previously,39 a line scan of 90 μm with a 0.5 μm piezoelectric step was carried out, beginning from ∼10 μm above the particle to ∼15−20 μm below the Teflon support. A total of 180 spectra were generated to cover the 60− 65 μm in diameter particle, each being collected with an integration time of 4 s and coaddition of two scans. Confocal Raman spectra suggest that protein loaded at pH 5 or pH 4 for 24 h distributes uniformly throughout the particle. The zerotime point spectrum in the protein desorption profile was obtained by averaging about 60 spectra corresponding to the top half of the particle. A similar procedure was performed to obtain Raman spectrum of the protein-free particle. The next time point in the protein desorption profile is ∼15 min after immersion in the elution buffer (pH 6, 1 M [Na+] in H2O). This 15 min delay allows for the particle to completely settle onto the Teflon membrane from the buffer solution. Line scan of 30 μm covering the top half of the particle with a 5 μm piezoelectric step generated seven spectra, each being collected with an integration time of 4 s and coaddition of two scans. The ∼1 min measurement time is expected to be negligible compared to the ∼15 min particle settlement time, and different depths were conceived to be measured at one time. Figure 1 shows confocal Raman spectra of the proteinretained and protein-free particles at different depths and loading conditions. The limited spectral range (1570−1750 cm−1) was selected to focus on the protein amide I band at ∼1667 cm−1. Assuming homogeneous pore structures throughout the particle, all spectra have been normalized on the basis of the Raman intensity at 1718 cm−1 mainly assigned to CO stretching of the polymethacrylate matrix. The contribution of protonation under acidic conditions of the carboxylate (−COO−) group(s) in mAb04 is very small and can generally be neglected. Spectral comparison at the zero-time point (Figure 1a,b, top) suggests that the particle with protein loaded at pH 5 shows an amide I intensity that is ∼30% greater than when loaded at pH 4, generally agreeing with previous results that show the static binding capacities of mAb04 in the Fractogel EMD SO3 (M) medium at pH 5 and 4 were ∼90 and ∼66 mg/mL, respectively.41 Further difference between the two adsorbed proteins is evident in their desorption behaviors. As seen in Figure 1a, for the particle with protein loaded at pH 5 and immersed in the elution buffer for 15 min, all seven spectra of different depths (0−30 μm) showed an intensity decrease at 1667 cm−1, when compared with that of the zero-time-point protein-loaded particle (Figure 1a, top). However, this relative intensity decrease at ∼1667 cm−1 was not observed in the seven spectra of different depths (0−30 μm) for the particle with protein loaded at pH 4 and immersed in the elution buffer for 15 min (Figure 1b). Spectra in Figure 1 therefore suggest that protein adsorbed at pH 4 is fully retained, whereas protein adsorbed at pH 5 can be released from particles in the elution buffer. Figure 2 shows depth profiles of protein retained in chromatographic particles after being immersed in the elution buffer for different times, calculated on the basis of Raman intensity at ∼1667 cm−1 in the normalized difference spectra (protein-loaded particle − protein-free particle). In Figure 2a, for protein loaded at pH 5, relatively flat intensity profiles could be observed throughout the particle, and the Raman intensity

then centrifuged to remove protein supernatant, diluted by the loading buffer and centrifuged again. Three cycles of centrifugation−dilution were carried out to remove free protein from the particles. The final protein-loaded Fractogel EMD SO3 (M) particles were stored in the loading buffer (pH 5 or 4, 20 mM [Na+] in H2O) or in a deuterated buffer (pD 5 or 4, 20 mM [Na+] in D2O) overnight for further studies. Raman Spectroscopy. All Raman experiments were carried out on an alpha-300R confocal Raman microscopy (WITec, Germany), an f/4 lens-based spectrometer with a focal length of 300 mm. Two gratings, 600- and 1800-lines/mm, were utilized to achieve low- and high-spectral resolutions. A thermoelectric-cooled back-illuminated CCD camera (1024 × 127 pixels, pixel size 26 × 26 μm, Andor Technology) was equipped to collect Raman signals, but only 18 out of 127 pixels in the central region were selected to read the spectra. A He− Ne laser (632.8 nm), which operated at 17 mW and passed through a 100 μm diameter single-mode optical fiber, was used to excite the sample. The confocal aperture of 100 μm in diameter was controlled by the multimode collection optical fiber. Wavelength calibration was achieved using low-pressure gas discharge lines of mercury and argon. The procedure to measure the depth profile of protein in chromatographic particles was described previously.39 In brief, prior to each confocal Raman measurement, about 10 particles were allowed to completely settle onto a hydrophobic Teflon membrane (0.2 μm, EMD Millipore Millex-FG) from ∼500 μL of a buffer solution. Two types of buffers were tested, the storing buffer (pH/pD 5 or 4, 20 mM [Na+] in H2O/D2O) to obtain protein adsorption profile and the elution buffer (pH 6, 1 M [Na+] in H2O) to obtain protein desorption profile. For acquisition of protein adsorption profiles, four particles with protein loaded at pH 5 were measured in the odd-numbered runs and four particles with protein loaded at pH 4 were measured in the even-numbered runs. A 60× water dipping objective lens (MRD07620 CFI Water Apo, Nikon) with a numerical aperture of 1.0 was used to collect Raman scattering signals. Other experimental conditions such as integration time and piezoelectric step for confocal Raman experiments were specified for each figure as discussed below. Protein in solution state was analyzed in a similar behavior to the particle. A droplet of ∼10 μL protein solution in the measuring buffer (H2O containing NaClO4 or D2O containing NaCl for conductivity adjustment) was placed on the Teflon membrane and covered with a PFA film (DuPont PFA 50LP, 12 μm in thickness, RI ∼1.35). On top of the PFA film ∼500 μL of H2O or D2O was deposited, into which the water dipping objective was dipped. The buffer solution without protein was also measured for solvent background subtraction.



RESULTS Desorption Profile of Protein from Chromatographic Particles. Fractogel EMD SO3 (M) resin has a cross-linked polymethacrylate matrix to which the sulfoisobutyl surface modification is bound through linear polymeric chains,39 the so-called “tentacles”.25 Particles are sized from 40 to 90 μm in diameter with a porosity of ∼61%, an average pore size of ∼80 nm in diameter and a ligand density of ∼380 μmol/g.42 Only particles with a diameter of 60−65 μm have been selected for spectroscopic measurements in this report. In the protein desorption profile measurement, the zero-time point is defined by examining the Raman spectra of proteinloaded particles in the loading buffer. Particles were well fixed 1009

dx.doi.org/10.1021/ac401081s | Anal. Chem. 2014, 86, 1007−1015

Analytical Chemistry

Article

Figure 2. Time series of confocal depth profiles of protein retained in the Fractogel EMD SO3 (M) particle after being immersed in the elution buffer. Protein was loaded onto the chromatographic particle at either pH 5 (a) or pH 4 (b) prior to elution. Calculation was based on the amide I intensity at 1667 cm−1 in the Raman difference spectra (protein-loaded particle − protein-free particle, after spectral normalization based on the 1718 cm−1 band). Profiles times were 15, 30, 45, 60, 75, 90, 105 and 120 min.

(protein-loaded particle − protein-free particle) representing proteins adsorbed in chromatographic particles and measured under different conditions. In Figure 3e,f, for protein adsorbed at pH 5 or 4 and measured in respective H2O buffers, the amide I band at ∼1667 cm−1 suggests that β-sheet is their predominant secondary structure. Further confirmation is available by examining their amide III band at ∼1240 cm−1, characteristic of the β-sheet conformation. In addition to the amide I and III bands, Raman spectroscopy is also beneficial in monitoring the average local environment of some amino acid side chains. For example, tryptophan shows a Fermi doublet at 1360 and 1340 cm−1.48 Decrease in the intensity ratio of two bands (I1360/I1340) suggests that the tryptophan indole ring is moved to a more hydrophilic environment or exposed to an aqueous medium. Unfortunately, the polymeric matrix of chromatographic particles has sufficient Raman signals in this region, to cause uncertainty in spectral calculation of the adsorbed protein and difficulty in determining whether or not there are small adsorption-induced environmental perturbations of tryptophan in the protein. In Figure 3g,h, for protein loaded at pH 5 or 4 and then measured in respective D2O buffers, there is a 6 cm−1 amide I downshift (from ∼1673 to ∼1667 cm−1), agreeing with the previous observation that H/D exchange causes a 5−10 cm−1 downshift in the protein amide I region.49 The more obvious isotope labeling effect is the significant amide III intensity decrease at ∼1240 cm−1, which is shifted to ∼985 cm−1 outside the coverage using the 1800 line/mm grating in one shot. Note proteins loaded at pH 5 and 4 did not show significant difference in the amide III intensity decrease, when they were measured in the D2O buffers. This seems to suggest that proteins loaded at the two pH values have a similar degree of unfolding. However, another explanation is that for mAb04 loaded at pH 4, protein unfolding and solvent exclusion due to

Figure 1. Confocal Raman spectra of the protein-retained Fractogel EMD SO3 (M) particle after being immersed in the elution buffer for 15 min. Protein was loaded onto the chromatographic particle at either pH 5 (a) or pH 4 (b) prior to elution. The top and bottom spectra in each panel correspond to zero-time point protein-loaded and proteinfree particles. All spectra were normalized on the basis of the 1718 cm−1 band.

decreases with elution time for successive profiles. Approximately 60−120 min are required to reach the desorption equilibrium, comparable with the time usually seen in practicing the cation exchanger for mAb purification. The protein recovery of ∼70% is slightly low, likely due to presence of a small amount of oligomers in the adsorbed state and/or absence of a flow rate in the protein desorption experiment. On the other hand, the current procedure provides a simple approach to investigate the intrinsic binding strength of protein with chromatographic particles, by removing other affecting factors such as superficial velocity and column packing. From Figure 2b, for particles loaded with protein at pH 4, different soaking times (0−2 h) in the elution buffer produced overlapped depth profiles, suggesting that no protein is released from the particles. Raman Characterization of Protein Adsorbed at pH 5 or 4 on Chromatographic Particles. To understand the different protein desorption profiles in Figures 1 and 2, we used the 1800 lines/mm grating to record Raman spectra (Figure 3a−d) of protein-loaded and protein-free particles in either H2O or D2O buffers at pH 5 or 4. Each spectrum is an average of four particles; and for each particle, about 60 confocal Raman spectra were measured covering the top half of the particle, each being collected with an integration time of 30 s and coaddition of 2 scans. Figure 3e−h shows difference spectra 1010

dx.doi.org/10.1021/ac401081s | Anal. Chem. 2014, 86, 1007−1015

Analytical Chemistry

Article

Figure 4. (a) Overlapped Raman spectra of protein loaded at pH 5 (in green, four odd-numbered runs) or pH 4 (in red, four even-numbered runs) in the Fractogel EMD SO3 (M) particles and measured in the H2O buffers; (b) Overlapped Raman spectra of protein loaded at pH 5 (in green, four odd-numbered runs) or pH 4 (in red, four evennumbered runs) in the Fractogel EMD SO3 (M) particles and measured in the D2O buffers.

perturbation of Trp in the adsorbed protein at pH 4.31 A similar amide I′ bandwidth increase can be observed for the protein adsorbed at pH 4 and measured in the D2O buffer (Figure 4b). Use of D2O in Raman measurement eliminates the bending vibration of H2O, either bound to protein or free in chromatographic pores, at ∼1630−1640 cm−1 in the amide I region. This vibrational mode might cause uncertainty in protein spectral calculation and complicate explanation of small amide I intensity changes.51 The small amide I/I′ bandwidth increase for protein loaded at pH 4 (Figure 4a,b) is unlikely due to baseline shift in the spectrometer, thermal fluctuation, drift in laser power and sample scattering efficiency, or other experimental errors. First, the essentially arbitrary order in acquiring spectra of proteins loaded at pH 5 and pH 4, respectively with odd-numbered and even-numbered runs, proves that the spectral difference between two pH values is not a result of artifact in any format. Second, the control spectra of protein in the solution state do not show discernible difference in amide I/I′ bandwidth between pH/pD 5 and 4 (see Figure 5b below). If the amide I/ I′ bandwidth change seen between two adsorbed proteins were due to an artifact, it is unlikely it would occur only for protein in particles but not for protein in the solution state. And third, spectra of protein-free particles measured at pH/pD 5 and 4 are nearly identical (data not shown) and thus no band shift has been introduced during background subtraction to calculate the protein spectra in Figure 4a−b. Raman Spectra of Protein in Solution at pH 5 and 4. Raman spectra of protein dissolved in the H2O and D2O buffers were recorded using the 1800 lines/mm grating for high spectral resolution. In the H2O buffer that contains NaClO4 to adjust solution conductivity, the salt is extremely Raman sensitive and often utilized as the internal standard for H2O background subtraction in calculating protein spectrum.43,52 Before and after the high spectral resolution measurement, the 600 lines/mm grating was selected to record spectra that include the 934 cm−1 Raman band of the NaClO4 salt and the

Figure 3. (a−d) Raman spectra of the protein-loaded and protein-free Fractogel EMD SO3 (M) particles. Protein was loaded at either pH 5 (a and c, in green) or pH 4 (b and d, in red). Spectra were measured in either H2O buffers (a, pH 5; b, pH 4) or D2O buffers (c, pD 5; d, pD 4). Spectra of protein-free particles were in gray. For each condition, four particles were measured to obtain the average spectra; (e−h) Raman difference spectra of two types of particles (protein-loaded particle − protein-free particle) corresponding to respective spectra in (a−d).

enhanced protein−resin and/or protein−protein contact are occurring simultaneously, resulting in a net neutral effect on the H/D exchange behavior.7,11 Use of the 1800 lines/mm grating improves the spectral resolution to ∼0.5 cm−1/pixel in the spectral range of 1600− 1700 cm−1, when other parameters (632 nm red laser for excitation and 300 mm focal length for signal collection) have been fixed. This is three times better than what could be achieved with the 600 lines/mm grating. Good separation of two close bands at 1617/1614 and 1604 cm−1, assigned to tryptophan/tyrosine and phenylalanine,37 has been achieved in this report (Figure 3) but not in the previous work.39 With this improved spectral resolution, we then compare normalized protein spectra in the expanded region of 1520−1750 cm−1, where chromatographic particles and the adsorbed protein show less overlapping features than in other regions (Figure 4). Four spectra acquired under the same protein loading and spectral measurement condition have been shown to demonstrate data reproducibility. Protein loaded at pH 4 and measured in the H2O buffer (Figure 4a) has a broader amide I band than that at pH 5, clearly seen from the fact that the higher wavenumber falling edge of the red spectra of pH 4 is above the green spectra of pH 5. The red spectra have also shown a small intensity decrease at ∼1552 cm−1 and a frequency shift, likely indicating the microenvironmental 1011

dx.doi.org/10.1021/ac401081s | Anal. Chem. 2014, 86, 1007−1015

Analytical Chemistry

Article

A column at pH ∼3−4 and then adjusted to have pH ∼5 for the subsequent cation exchange chromatography. Elimination of the buffer exchange procedure, i.e., to load the protein eluted from the Protein A column directly to the cation exchanger, would represent a potential cost-saving approach. Although it might be effective for some mAbs to be processed on some cation exchangers, we find that this approach does not work for a hydrophobic mAb, EMD Millipore mAb 04, on a strong cation exchanger, Fractogel EMD SO3 (M). Protein loaded at pH 4 cannot elute from the chromatographic particles under elution conditions.41 The research presented here is therefore an effort to understand how and why the loading pH affects mAb desorption from the cation exchanger. Implication of Amide I Intensity Increase in the >1680 cm−1 Region. The amide I vibration is a property of the peptide group involving the CO stretching vibration with minor contributions from the out-of-phase CN stretching vibration, the CCN deformation, and the NH in-plane bending.53 Its frequency is determined mainly by the hydrogen bonding of the polypeptide bond, which varies in different conformations. For example, α-helix and β-sheet have intrachain and interchain hydrogen bondings, i.e., the CO group forms a hydrogen bond with the −NH group of the peptide bond in the same chain at different sequence positions or in another adjacent chain. Other factors that contribute to the amide I frequency include the coupling effects between amino acid residues54 and presence of water in the amide hydrogenbonding.55 The amide I band is the sum of individual contributions of all amide groups, and thus its position and/ or shape are sensitive to protein backbone conformational changes. The periodic α-helix and β-sheet structures show wellrecognized amide I bands at ∼1656 and ∼1667 cm−1. This characterization is based on correlating the amide I position with the crystallographically determined fraction of each secondary structural element present in a protein.51 Unfortunately, it is difficult to make a definitive assignment of the “aperiodic” (not repeating structure) Raman band in the region 1680−1700 cm−1, which has been attributed, in different studies, to β-turn,51,56 disordered structure (non-hydrogen bonded)57 and/or intermolecular β-sheet conformation.38,55 It is thus difficult to define the exact structural changes that are related to the amide I intensity increase in the >1680 cm−1 region observed for mAb04 adsorbed at pH 4 in chromatographic particles. On the other hand, previous Raman studies on monoclonal antibodies, although very limited, may provide some clues to the increased amide I line width. In one study, Raman spectroscopy was applied to investigate the relationship between lyophilization-induced mAb structural changes and long-term storage protein stability.56 Lyophilization resulted in a broader amide I band with an intensity decrease at ∼1670 cm−1 and an intensity increase in the >1680 cm−1 region.56 The magnitude of the amide I shift correlates directly with the mAb association during storage.56 Little or no amide I line width change could be detected if an excipient was added to the formulation to stabilize protein secondary structure during lyophilization;56 and correspondingly, poststorage reconstitution had much less mAb aggregates.56 In another study, Raman spectroscopy was applied to study pH effects on mAb stability. Amide I shift from 1669 to 1689 cm−1, comparable in magnitude with the observation in Figure 4a/b, was accompanied by protein association, when the pH of the

Figure 5. (a) (From top to bottom) Raman spectra of protein in the solution state at pH 5 (in green), pH 4 (in red), pD 5 (in green), and pD 4 (in red); (b) (from top to bottom) Overlapped Raman spectra of protein in solution of pH 5 (in green, four odd-number runs) or pH 4 (in red, four even-numbered runs), Overlapped Raman spectra of protein in solution of pD 5 (in green, four odd-number runs) or pD 4 (in red, four even-numbered runs).

protein amide I band at ∼1667 cm−1 in one shot. Consistency of their frequency and intensity in two spectra of different times indicates that there is no detectable water evaporation during high spectral resolution Raman measurement and that protein does not have obvious structural changes due to laser irradiation. Figure 5a showed Raman spectra of mAb04 in the H2O and D2O buffers at pH/pD 5 and 4. Spectral comparison indicates that deuterium exchange causes a ∼6 cm−1 amide I downshift and a significant amide III intensity decrease at ∼1240 cm−1. These isotope labeling effects have been observed for the adsorbed protein in chromatographic particles measured in the D2O buffer (Figure 3). But, unlike the protein adsorbed at pH 4 and measured at pH/pD 4 (Figure 4), protein in solution of pH/pD 4 does not have amide I/I′ shift that is large enough to be detected by Raman spectroscopy (Figure 5b). This agrees with previous mAb studies that when the solution pH was adjusted from 7 to 2, obvious Raman spectral changes could be observed around pH ∼2 but not above pH 3.38



DISCUSSION Monoclonal antibody purification usually employs a template, which often includes Protein A and cation exchange chromatography. The target protein is eluted from the Protein 1012

dx.doi.org/10.1021/ac401081s | Anal. Chem. 2014, 86, 1007−1015

Analytical Chemistry

Article

mAb solution was decreased from 7 to 2.37,38 Both case studies seem to suggest that the increased Raman intensities in the >1680 cm−1 region observed for mAb04 adsorbed at pH 4 might be related to some conformational changes involved in protein association in the chromatographic particles. About 1% amide I intensity shift (Figure 4) indicates that among ∼1400 amino acids in mAb04, at least 14 residues have experienced structural perturbations. Possible Explanation for Retention of the Protein Adsorbed at pH 4 in Chromatographic Particles. Interaction between the mAb04 molecules loaded at pH 4 and the Fractogel EMD SO3 (M) cation exchanger is so strong that protein cannot release from chromatographic particles under either static (Figures 1 and 2) or dynamic41 elution conditions. One possible explanation considers the fact that mAb04 is more positively charged at pH 4, to have more electrostatic interactions with the −SO3 surface modification groups in the chromatographic particles. Only the surface charges of the mAb04 molecular are involved in the electrostatic interaction, but their number is difficult to know. For a simplified comparison of surface charges, the mAb04 molecule with known amino acid sequence is calculated to have overall positive charges of 54 at pH 5 and 106 at pH 4. However, in discussing protein adsorption onto an ionexchange column, the steric effect also play a key role that only a limited number of surface charges (