Site-Specific Characterization and Absolute Quantification of

Aug 26, 2015 - The characterization and absolute quantification of protein biopharmaceuticals and their product-related impurities, e.g., oxidation va...
0 downloads 11 Views 1MB Size
Download PDF
Article pubs.acs.org/ac

Site-Specific Characterization and Absolute Quantification of Pegfilgrastim Oxidation by Top-Down High-Performance Liquid Chromatography−Mass Spectrometry Ines C. Forstenlehner,†,‡ Johann Holzmann,‡,§ Hansjörg Toll,‡,§ and Christian G. Huber*,†,‡ †

Christian Doppler Laboratory for Innovative Tools for Biosimilar Characterization, University of Salzburg, Hellbrunnerstrasse 34, 5020 Salzburg, Austria ‡ Department of Molecular Biology, Division of Chemistry and Bioanalytics, University of Salzburg, Hellbrunnerstrasse 34, 5020 Salzburg, Austria § Analytical Characterization Biopharmaceuticals, Sandoz GmbH, Biochemiestrasse 10, 6250 Kundl, Austria S Supporting Information *

ABSTRACT: The characterization and absolute quantification of protein biopharmaceuticals and their product-related impurities, e.g., oxidation variants, is essential due to their potential impact on biological activity and immunogenicity. Here, we present site assignment and absolute quantification of oxidation variants of pegf ilgrastim, a poly(ethylene glycol) modified recombinant human granulocyte-colony stimulating factor. Pegf ilgrastim stressed with 1.0% hydrogen peroxide served as a model protein for developing a top-down high-performance liquid chromatographymass spectrometry (HPLC-MS) platform that allowed direct site assignment of Met122, Met127, and Met138 oxidation within a total analysis time of 30 min. Three different absolute quantification methods, namely, UV absorption spectroscopy, full-scan MS, and all-ion fragmentation (AIF) MS were compared. Additionally, the monitoring of all generated fragment ions or selected sets of fragment ions were evaluated for the AIF method. Linearity of calibration curves from 5.0 to 25 ng μL−1, 25 to 250 ng μL−1, and 100 to 1000 ng μL−1 was confirmed. The AIF method achieved a lower limit of detection of 0.85 ng μL−1 and a lower limit of quantification of 2.54 ng μL−1. On the basis of the comparison of relative standard deviations of interday measurements, AIF was concluded to be the method of choice for concentrations up to 50 ng μL−1, and UV measurements should be carried out above this concentration. Finally, an expired pegf ilgrastim batch was analyzed as a a real biopharmaceutical sample to confirm the feasibility of our approach for monitoring low levels of oxidation variants.

T

Met138, leading, under mildly oxidative conditions, predominantly to methionine sulfoxide derivatives.11 Met1, which represents also the site of PEGylation,12 is described to be most prone to oxidation, whereas Met122 is located within the hydrophobic core of the protein and therefore features the lowest oxidation susceptibility. Both Met127 and Met138 are found in the loop region between the third and fourth helices. With the exception of Met1, all oxidized variants are described to possess reduced biological activity.13 Although a number of reports exist about the characterization of f ilgrastim and its oxidation variants,11,13−15 studies on the PEGylated version can scarcely be found.16 Holzmann et al. employed top-down mass spectrometry (MS) for site specific assignment of oxidation in f ilgrastim and could differentiate eight different oxidation variants by exploiting diagnostic

he biopharmaceutical pegf ilgrastim is a poly(ethylene glycol) modified (PEGylated) recombinant human granulocyte-colony stimulating factor (rhG-CSF) that is primarily administered to counteract neutropenia induced by chemotherapy.1 PEGylation of proteins results in a number of advantages such as lower immunogenicity and a longer serum half-life due to reduced renal clearance.2,3 Treatment with the non-PEGylated G-CSF (f ilgrastim) needs daily injections for around 2 weeks whereas one injection of pegf ilgrastim is sufficient for each chemotherapy cycle.4 However, the Nterminally, covalently attached poly(ethylene glycol) (PEG) makes mass spectrometric analysis of the protein and its impurities significantly more challenging because of PEG heterogeneity.5 Oxidation represents one of the major product related impurities of protein biopharmaceuticals that requires strict monitoring due to its potential impact on immunogenicity and biological activity.6−10 The main oxidation sites of rhG-CSF are its four methionines, namely, Met1, Met122, Met127, and © 2015 American Chemical Society

Received: May 31, 2015 Accepted: August 26, 2015 Published: August 26, 2015 9336

DOI: 10.1021/acs.analchem.5b02029 Anal. Chem. 2015, 87, 9336−9343

Article

Analytical Chemistry fragments.11 Similar studies have not been reported for pegf ilgrastim. Bagal et al. described the intact measurement of pegf ilgrastim and oxidized variants which yielded a mass difference of 16 Da that was concluded to be oxidation. However, oxidation assignment was done by peptide mapping and only one oxidation site, the Met122, was found.16 The routine strategy for the quantitative analysis of methionine oxidation in proteins consists of proteolytic digestion, subsequent separation of oxidized and nonoxidized peptides, their detection by UV spectroscopy and/or MS, and the comparison of the peak areas of nonoxidized and oxidized peptide pairs.14,17−20 Absolute quantification has been achieved by the addition of labeled21,22 or analogue proteins23 before proteolytic digestion or labeled peptides after digestion.24 Houde et al. have demonstrated relative quantification of peptides by mass spectrometry postulating that nonoxidized and oxidized peptides would have similar ionization efficiencies and therefore ratios of peak areas. This assumption was validated by comparison of UV and MS data.17 A similar approach was applied for the quantification of nonmodified PEGylated molecules. However, only non-PEGylated peptides were exploited as surrogates for the quantitative analysis.25,26 The quantification based on peptides gathered by enzymatic digestion is, nonetheless, not capable of distinguishing specific, multiply oxidized protein variants and their abundances, but merely gives overall information on the abundance of specific modification sites. Moreover, enzymatic reactions and extensive sample preparation could lead to artifacts such as deamidation27 or oxidation,28 especially when changing the optimal storage conditions of the protein to the optimal buffer conditions for the protease. Li et al. reported a quantification method for a 20 kDa PEGylated peptide that applied in-source fragmentation prior to MS analysis in order to simplify the spectrum and quantify on the basis of a fragment,29 but the application of fragmentation to perform quantitative analysis of oxidation in a PEGylated protein has not been explored so far. The chromatographic separation of pegf ilgrastim oxidation variants and their intact mass spectrometric analysis was previously shown by our group.30 Therefore, we aim here at developing and evaluating a site-specific and absolute quantification method capable of handling intact PEGylated proteins and their oxidized variants without the need for digestion and extensive sample preparation.31 We focus on elucidating the oxidation sites of stressed pegf ilgrastim by topdown MS and intend to perform mass spectrometric quantification based on the calibration-response curve method of peak areas for intact masses or fragments.

expired 09/2011) was produced by Amgen (Breda, The Netherlands). Sample Preparation. The PEGylated rhG-CSF calibration standards were prepared in ultrapure water in appropriate concentrations over a concentration range of 0.50−1000 ng μL−1. A stock solution of carbonic anhydrase was prepared in 20% acetonitrile and 0.050% TFA and added to each calibration standard to a final concentration of 100 ng μL−1. The carbonic anhydrase was used as carrier protein for diminishing loss of calibrant because of adsorption to surfaces especially of the low concentrated solutions. Oxidation was induced in a 1.0 mg mL−1 pegf ilgrastim sample by addition of H2O2 to a final concentration of 1.0% for 15 min at room temperature. The oxidation process was stopped through buffer exchange to ultrapure water by using ultrafiltration (Amicon Ultra 0.5 mL centrifugal filters with a 3 kDa cutoff, EMD Millipore, Darmstadt, Germany). Subsequently, carbonic anhydrase was also added to the stressed samples. For the measurement of intact proteins, triethylamine was prepared in a concentration of 400 mmol L−1 in 50% methanol. The high mass calibration of the mass spectrometer was executed with 1.0 mg mL−1 ammonium hexafluorophosphate in 50% methanol and 0.10% FA. The determination of extinction coefficients for nonstressed and stressed pegf ilgrastim samples was carried out on a nanophotometer (Model P330, Implen GmbH, Munich, Germany) at 214 nm. A molar extinction coefficient of 260 348 L mol−1 cm−1 at 214 nm was determined both with an oxidized or nonoxidized 1.0 mg mL−1 standard solution of pegf ilgrastim. Buffer exchange to ultrapure water by ultrafiltration was carried out for the stressed samples because H2O2 was interfering with the UV-absorbance of the analyte. High-Performance Liquid Chromatography. The chromatographic separation of oxidation variants was carried out in a capillary/nano high-performance liquid chromatography (HPLC) system (Model Ultimate 3000, Thermo Fisher Scientific, Germering, Germany) at a flow rate of 200 μL min−1. The Discovery BIO wide pore C18 column (150 × 2.1 mm i.d., 3.0 μm particle size, 300 Å pore size, Supelco, Bellefonte, PE, USA) was operated at a column temperature of 50 °C. A gradient of solutions A (H2O + 0.10% TFA) and B (acetonitrile + 0.10% TFA) was formed as follows: 53.2% B for 0.5 min, 53.2−53.6% B for 15 min, 53.6−56.6% B for 0.50 min, 56.6−57.0% B for 15 min, 57.0−95.0% B for 0.50 min, 95.0% B for 2.5 min, 95.0−53.2% B for 0.10 min, and 53.2% B for 7.0 min. In spite of its tendency for ion suppression, TFA was chosen as an additive for ion-pair reversed-phase HPLC of proteins, because it facilitates sufficient (baseline) separation of the investigated pegfilgrastim oxidation variants.32 Moreover, TFA as additive allowed serial UV-detection at 214 nm and mass spectrometric detection of the separated proteins, which was essential for the comparison of both detection modes. For minimizing quantitative uncertainty, a full loop injection of 1.0 μL was chosen. UV detection was carried out at a wavelength of 214 nm in a 45 nL Z-shaped capillary cell. For the intact measurements, triethylamine was added via a T-piece and a syringe pump (Model Fusion 100 T, Chemyx Inc., Stafford, TX, USA) equipped with a 5000 μL syringe (Hamilton, Reno, NV, USA) before the ESI spray as described previously.30 Mass Spectrometry. All measurements were conducted on a quadrupole-Orbitrap mass spectrometer (Q Exactive) from Thermo Fisher Scientific (Bremen, Germany) by using the

■

EXPERIMENTAL SECTION Materials. Acetonitrile (LC−MS grade) was purchased from VWR (Leuven, Belgium). Hydrogen peroxide (H2O2, 35.0−36.5%) was obtained from Merck (Darmstadt, Germany). Methanol (LC−MS Chromasolv grade), trifluoroacetic acid (TFA, ≥ 99.5%), formic acid (FA, 98−100%), triethylamine (TEA, ≥ 99.5%), ammonium hexafluorophosphate (99.99%), and carbonic anhydrase (from bovine erythrocytes, 100%) were purchased from Sigma-Aldrich Chemie GmbH (Steinheim, Germany). Ultrapure water was produced in house by a Milli-Q System (Millipore Corporation, Billerica, USA). The PEGylated rhG-CSF (drug substance, 16.2 mg mL−1 solution in 10 mmol L−1 acetate, 50 mg mL−1 sorbitol, pH 3.8−4.2) was provided by Sandoz GmbH (Kundl, Austria). The expired pegf ilgrastim batch (Neulasta, 6 mg, 0.6 mL injection solution, 9337

DOI: 10.1021/acs.analchem.5b02029 Anal. Chem. 2015, 87, 9336−9343

Article

Analytical Chemistry

Figure 1. Top-down fragment ion spectrum of nonoxidized pegf ilgrastim. The diagnostic fragments, y47 (blue), y50 (red), and y78 (green), are indicated: (a) raw spectrum; (b) sequence coverage map generated with ProSight PTM. Sample: 1000 ng μL−1 pegfilgrastim, R = 140 000 at m/z 200; SID 70, NCE 12. Other measurement conditions are given in the Experimental Section.

heated ESI source and applying a heater temperature of 70 °C. The following instrumental parameters were used: sheath gas flow of 35 arbitrary units, auxiliary gas flow of 15 arbitrary units, spray voltage of 3.5 kV, capillary temperature of 300 °C, S-lens RF level of 80.0, in-source CID of 70.0 eV, AGC target of 1e6, and a maximum injection time of 100 ms. The measurements of intact proteins with postcolumn addition of triethylamine (TEA) were performed with full scan data acquisition, a scan range of m/z 2500−6000, and a resolution of 17 500 at m/z 200. The fragmentation experiments utilizing all-ion fragmentation (AIF) were carried out in the higher-energy collisional dissociation (HCD) cell upon applying a normalized collision energy (NCE) of 12 (corresponding to 48 eV collision energy at m/z 2000), a scan range of m/z 1000−3000, and a resolution of 140 000 at m/z 200. Source collision-induced dissociation (SID) and normalized collision energy (NCE) values were optimized to obtain maximum signal intensity of the diagnostic y-fragments. Individual parameters are given in the figure legends. Mass calibration was executed by using ammonium hexafluorophosphate. For the fragmentation experiments, the instrument was calibrated with a lower m/z of 670 and an upper m/z of 2626. High mass calibration was done within the mass range of m/z 2137−5559. Data Evaluation. All peaks were manually integrated using Chromeleon 7.1 (Thermo Fisher Scientific, Germering, Germany) for UV Data and Xcalibur 2.2 for MS data (Thermo Fisher Scientific, San Jose, CA, USA). The extent of oxidation in the standard substance was determined by relative quantification of the UV areas by applying the SmartPeaks integration assistant option and choosing the integration type: exponentially skimmed riders. The deconvolution of the fragment ion spectra was carried out by applying the Xtract algorithm in the software Xcalibur 2.2 (Thermo Fisher Scientific). Extracted ion current chromatograms (EICCs) were taken from the five most intense isotopic peaks and before integration, whereas reconstructed total ion current chromatograms (RTICC) were used to represent the sum of the intensities of all detected (fragment) ions. Gaussian smoothing of 7 points was generally performed.

Quantification was accomplished on the basis of external calibration using nonoxidized pegf ilgrastim as standard protein throughout. All determined peak areas were manually transferred to an Excel sheet (Microsoft Excel 2010, vers. 14.0.7116.5000, Microsoft Corporation, Redmond, WA, USA), which is available as a Supplementary Table (see Supporting Information Excel file). Calibration curves of five interday injections were constructed for three different concentration ranges: 5−25, 25−250, and 100−1000 ng μL−1. Data treatment included the calculation of the regression equation and the correlation coefficient (R2), the calculation of the confidence intervals (C.I.) based on f = n − 2 and a confidence level of 95%, and the limit of detection (LOD) and limit of quantification (LOQ) (see Supporting Information Excel file). Sequence coverage of fragmentation experiments was determined with the ProSightPTM33 online tool (https:// prosightptm2.northwestern.edu/) provided by the Kelleher Research Group (Northwestern University, Evanston, IL, USA).

■

RESULTS AND DISCUSSION Oxidation Site Assignment. PEGylated rhG-CSF contains 4 methionines, namely, Met1, Met122, Met127, and Met138, which can be oxidized through mild treatment with hydrogen peroxide to obtain a suitable reference sample.13 The chromatographic separation of the four main oxidation variants of stressed PEGylated rhG-CSF and the determination of the degree (but not the site) of oxidation by intact protein mass spectrometry have previously been shown by our group.30 Here, oxidation site assignment to the respective methionines was carried out on the basis of characteristic fragments.11 In a first step, postcolumn addition of triethylamine was used for charge-stripping to shift signals for PEGylated, multiply charged pegf ilgrastim species into a detectable m/z range. However, this was accompanied by a significant loss of signal intensity. On the other hand, when carrying out all-ion fragmentation (AIF) on the intact protein in the HCD cell, which is a mode of fragmentation that decomposes all charge states of the protein transmitted into the collision cell without prior mass selection in the quadrupole, several y-ions were 9338

DOI: 10.1021/acs.analchem.5b02029 Anal. Chem. 2015, 87, 9336−9343

Article

Analytical Chemistry readily detectable even in the absence of triethylamine. For example, y47, y50, and y78 (see fragments labeled in Figure 1) were useful as diagnostic fragments for Met122, Met127, and Met138 oxidation.11 Although the C-terminus featured good sequence coverage in our measurements, no b-ions containing the N-terminus were retrieved (see Figure 1), most probably because of the attachment of PEG to the amino group of the N-terminal methionine.12 The b-ions would therefore have an additional mass of approximately 20 kDa, which shifts them out of the upper m/z limit of the Q Exactive of m/z 6000. Another possible reason for missing b-ions is the partial fragmentation of PEG, resulting in a large number of different b-ions below the detection limit. Due to that, Met1 is not covered in the fragmentation experiments and needs to be elucidated by combining the data on the oxidation degree gathered by intact measurements with the information gained by oxidation site mapping of Met122, Met127, and Met138. Figure 2 depicts the reconstructed total ion current chromatogram (RTICC) as well as extracted ion current chromatograms (EICCs) of the diagnostic fragments y47, y50, and y78 that are suitable for differentiating the nonmodified pegf ilgrastim and its oxidized variants. In the employed gradient, the carrier protein carbonic anhydrase eluted in the column hold-up volume and did not interfere with the pegf ilgrastim variants (see Figure S-1). In total, eight different chromatographic peaks were distinguishable, each representing different oxidation variants. The first four peaks at lower retention times all contain oxidized Met122, as evidenced by the absence of the nonoxidized y78 (trace g in Figure 2) and the presence of y78 containing one, two, or three additional oxygen atoms (traces h, i, and j in Figure 2). Peak 1 contains triply oxidized y78 (trace j), which allows direct identification of oxidized Met122, Met127, and Met138. Peak 2 features a nonoxidized y47 (trace b); therefore, Met138 is not oxidized, but a singly oxidized y50 (trace e), which is directed to a Met127 oxidation, and a doubly oxidized y78 (trace i) corroborates that also Met122 is oxidized. Moreover, it can be deduced that peak 3 shows Met122 and Met138 oxidation (traces c, e, and i), while peak 4 only points toward Met122 oxidation (trace h). Met1 is not covered in this data and could only be elucidated by comparison to the data on the overall oxidation degree gained by intact measurements. However, the intact measurements were not conclusive in this respect for the first four chromatographic peaks due to insufficient signal intensity of the minor oxidation variants (see Figure S-2a,b). Nevertheless, data published on the nonPEGylated rhG-CSF11 give clear evidence that Met1 is oxidized in all those peaks as it is most prone to oxidation. In a similar manner as detailed above, we can conclude that peak 5 shows Met1, Met127, and Met138 oxidation; peak 6 features Met1 and Met127 oxidation, peak 7, oxidized Met1 and Met138, and peak 8, Met1 oxidation (see Figure 2). The peak labeled as “S” is a system peak, resulting from the steep increase of the acetonitrile gradient after an almost isocratic elution step (see Figure S-3) yielding a dense cloud of signals in the mass spectra. It is potentially composed of coeluting low level oxidation variants, e.g., oxidation on amino acid residues other than methionine. However, due to the plethora of low abundant signals, a specific variant could not be unambiguously assigned. Quantification Strategies. Since oxidation is one of the major shelf life limiting factors for protein biopharmaceuticals, its absolute quantification is of utmost relevance, especially in

Figure 2. Oxidation site assignment based on EICCs of diagnostic fragments y47 (blue), y50 (red), and y78 (green), being non-, singly (+O), doubly (+2O), or triply (+3O) oxidized (the 5 most intense isotope peaks of the charge state +4 were extracted). The zoomed chromatographic traces on the left side visualize the minor oxidation variants present in peaks 1−4. Peak identification: 1, Met122, Met127, Met138 oxidized; 2, Met122, Met127 oxidized; 3, Met122, Met138 oxidized; 4, Met122 oxidized; S, system peak; 5, Met127, Met138 oxidized; 6, Met127 oxidized; 7, Met138 oxidized; 8, nonoxidized. Met1 oxidation is not covered in this evaluation. Sample, 1.0 μL of 1000 ng mL−1 PEGylated rhG-CSF stressed with 1.0% H2O2 for 15 min at room temperature; R = 140 000 at m/z 200, SID 70, NCE 12. Further experimental details are provided in the Experimental Section.

the lower concentration range. Nonetheless, quantification methods based on MS have not been employed so far, neither for in-process control nor for product release analysis because of challenging implementation requirements for a regulated GMP environment. Still, several MS-based strategies have been described in the literature for the absolute quantification of intact proteins, including selected-reaction monitoring (SRM)34 and intact full-scan MS measurements.31 Quantification relying on full-scan MS offers the benefit of monitoring several charge states and collecting information on the whole protein, including possible other post-translational modifications such as disulfide bond formation or sequence variation with the disadvantage of higher spectral background and potential interferences due to the presence of several multiply charged species.31 In order to tackle the high complexity of intact protein mass spectra, physical separation of the protein variants before full-scan MS is mandatory for successful quantification. SRM methods, quantifying analytes on the basis of fragments generated from preselected precursor ions, on the other hand, hold the advantages of higher specificity and 9339

DOI: 10.1021/acs.analchem.5b02029 Anal. Chem. 2015, 87, 9336−9343

Article

Analytical Chemistry

Table 1. Quantification of the Oxidation Variants of the 1.0% H2O2 Stressed pegf ilgrastim Sample Based on the Different Data Evaluation Approaches Concentrationa Peak 1b [ng μL−1] Peak 2 [ng μL−1] Peak 3 [ng μL−1] Peak 4 [ng μL−1] Peak 5c [ng μL−1] Peak 6 [ng μL−1] Peak 7 [ng μL−1] Peak 8 [ng μL−1] R2 (5−25 ng μL−1) R2 (25−250 or 50−250 ng μL−1)h R2 (100−1000 ng μL−1)

UV

MS, AIF-RTICC

MS, AIF-EICC (y47)

n.d.d n.d. n.d. n.d. 102.40 ± 114.28 ± 324.77 ± 436.54 ± n.d. 0.9921h 0.9961

3.35 ± 0.19f 3.44 ± 0.19f 3.93 ± 0.18f 4.38 ± 0.18f 103.31 ± 8.17 119.48 ± 8.20 326.63 ± 47.73 495.93 ± 46.90 0.9988 0.9738 0.9506

0.85 ± 0.32g 1.55 ± 0.32f 1.24 ± 0.32g 3.00 ± 0.31f 43.11 ± 8.32 91.51 ± 8.05 162.76 ± 46.75 505.42 ± 43.41 0.9968 0.9751 0.9583

4.45e 4.44 13.16 12.96

a The concentrations are mean values of 3 consecutive injections of the 1.0% H2O2 stressed sample depicted in Figure 2; the initial total protein concentration was 1000 ng μL−1. bPeak numbering corresponds to Figure 2. cPeak 5 is described in detail in the text. dn.d. = not detectable. e95% confidence intervals were calculated using a two sided t test. fNote that the lowest calibration point used for quantification was 5.0 ng μL−1. Even though all concentrations are above the lower limit of quantification, their concentrations need to be considered estimates. gPeaks 1 to 4 are below the limit of quantification for the AIF-EICC approaches (see Table 2). Therefore, these concentrations have to be considered as estimates. hUV detection was calibrated within the range of 50−250 ng μL−1.

200 ng μL−1; therefore, this quantification strategy was no longer followed. For the other quantification approaches involving AIF, peak areas were deduced either from RTICCs of all fragment ions detected upon AIF (further called MS, AIFRTICC) or from EICCs of selected y-ions (+4 charge state and 5 most intense isotope peaks, further called MS, AIF-EICC). Calibration curves were established for three different concentration ranges, namely, 5−25, 25−250, and 100−1000 ng μL−1. For UV measurements, no calibration curve was generated for the concentration range of 5−25 ng μL−1 because of low signal-to-noise ratios, and the second calibration curve was constructed from 50 to 250 ng μL−1. Individual calibration curves and statistical parameters are provided as Supporting Information (see also the section figures of merit below). Table 1 summarizes the quantitative results of the three technical replicates of the 1.0% H2O2 stressed pegf ilgrastim analysis. The method based on UV absorbance could not detect the low abundant oxidized Met122 variants (peaks 1 to 4 in Figure 2). Peak 5, the triply oxidized variant containing Met1, Met127, and Met138 oxidation, results in comparable concentrations between the UV- and the AIF-RTICC method with a deviation of only 0.88%. However, AIF-EICC quantification based on the y47 fragment yielded a deviation of 57.9% between the two quantification methods. Close inspection of the fragment ion spectra of oxidized and nonoxidized pegf ilgrastim variants revealed that the relative signal intensities of the observed y-fragments differed considerably, which indicates an influence of methionine oxidation on fragmentation behavior (see Figure S-5). Consequently, utilization of nonoxidized protein to quantify the oxidized variants on the basis of selected y-fragments is inappropriate. Nonetheless, the decent congruence of results obtained with the UV- and AIF-RTICC methods (Table 1) prove the suitability of top-down mass spectrometric detection for intact protein quantification. These results suggest that the total abundance of ions does not change upon methionine oxidation, although the relative intensities of fragment ions may change, which is in agreement with a report that the oxidation of

lower limits of detection at the cost of higher methodological complexity. Utilizing data generated by UV-absorbance and mass spectrometric detection, we evaluated the concentration− response curves obtained either by monitoring the signals of multiply charged intact protein species (intact full scan MS) or by observing fragments generated upon AIF by means of full scan data acquisition (AIF-MS) in a quadrupole-Orbitrap mass spectrometer. UV-absorbance detection, employing solutions of the nonoxidized protein as calibration standards, served as a reference method yielding a conventional true value to estimate bias and to assess statistical parameters of the calibration, such as limits of detection/quantitation or confidence intervals. To ensure that the oxidation state does not influence the UV absorption coefficient, a nonstressed and a stressed pegf ilgrastim sample of the same concentration were measured on a nanophotometer and no significant difference could be observed in terms of absorptivity (see Figure S-4). Absolute Quantification of Oxidation Variants. A 1.0% H2O2 stressed pegf ilgrastim sample served as a model mixture as it features both very low- and high-abundant variants (see peaks 1−4 and 5−8 in Figure 2). Since pure standards for each oxidation variant were not available, we deduced the concentrations of oxidized variants from a calibration curve created with the peak areas of the nonoxidized standards. Because also the “nonoxidized” standard featured a low level of oxidation, the actual standard concentrations were adjusted after comparing the relative peak areas in UV traces of the 1000 ng μL−1 calibration standard, which revealed an overall oxidation degree of 1.1%. Measurement series included the analysis of 14 different calibration standards (0.50−1000 ng μL−1) followed by two blank measurements and one blank injection of pure water. Measuring from low to high concentration, calibration series were repeated three times; then, each sample solution was analyzed in triplicate, and then, each calibration series was run two additional times. Data evaluated from reconstructed total ion current chromatograms (RTICC) of the multiply charged intact protein species revealed that the signal-to-noise ratios were insufficient to calibrate concentrations below approximately 9340

DOI: 10.1021/acs.analchem.5b02029 Anal. Chem. 2015, 87, 9336−9343

Article

Analytical Chemistry

(Table 2). Relative process standard deviation for the different calibration ranges and detection methods are displayed in Table 2 and range from 2.4% to 15%. Comparison of the Quantification Methods. A comparison of RSDs of UV and MS measurements suggests that the method of choice for pegf ilgrastim concentrations below 50 ng μL−1 is MS-based AIF-RTICC as it features lower RSDs and lower limits of detection/quantification. Protein concentrations below 35 ng μL−1 can only be quantified by means of MS-based AIF-RTICC. For concentrations higher than approximately 100 ng μL−1, UV quantification yields more reliable results. The comparatively high LODs with intact fullscan MS with the postcolumn addition of TEA makes the qualitative or quantitative monitoring of low-concentrated product-related impurities unfeasible (see Figure S-2). Finally, the data presented in Table 2 clearly show that a serial implementation of UV absorbance and mass spectrometric detection as readily feasible with common HPLC-MS platforms such as the one utilized in this study facilitated the confident qualitative and quantitative analysis of proteins in a concentration range of 5−1000 ng μL−1, which corresponds to a molar concentration of 124 nmol L−1 to 25 μmol L−1. The relative process standard deviations in this combined setup were equal or below 6.2%, representing a sound basis for implementation in a GMP regulated environment in a not too distant future. On the other hand, a combination of all-ion fragmentation with detection of all fragment ions clearly represents a viable alternative to intact full-scan MS enabling MS-based detection to enhance the detectability of low-concentrated impurities. Our investigations also clearly show that selected-reaction monitoring (similar to AIF-EICC), which is the gold standard for peptide quantification, is only applicable when authentic reference standards (in our case well-defined oxidized protein variants) are available. Quantification of Oxidation in a Real Biopharmaceutical Sample. A pegf ilgrastim sample stored for 3 years beyond its expiry date was utilized to check for oxidation variants in a real sample. The quantification range was chosen on the basis of the sample concentration. The RTICC of the AIF-MS measurement of the expired pegf ilgrastim sample is depicted in Figure 4. The main component (peak 3) was quantified with the UV-method, which was found to be optimal for concentrations above 100 ng μL−1, and gave a concentration of 679.68 ± 13.10 ng μL−1. The oxidation variants, peaks 1 and 2, were quantified with the MS, AIF-RTICC method due to

Met127 as well as Met138 induces local conformational changes.13 Figures of Merit. The linearity of all calibration curves, except the 5−25 ng μL−1 MS, AIF-EICC (y47), was confirmed by applying the Mandel’s fitting test (see Supporting Information Excel file). Relative standard deviations (RSDs) of quintuplicate injections of each calibration standard are plotted in Figure 3 and ranged between 1.6% and 47% for the

Figure 3. Relative standard deviations of the peak areas (RSDs) at the respective concentrations of the calibration curves. RSDs represent 5 different measurements carried out within 3 days corresponding to the interday precision.

different methods. As expected, RSDs decrease with increasing sample concentration with UV- and full-scan intact MS detection. However, for the MS, AIF-RTICC method, RSDs start increasing again after a minimum at 25 ng μL−1, which is most probably a consequence of the interplay between the maximum number of ions admissible for injection into the Orbitrap35,36 (automatic gain control AGC target value) and the maximum ion collection time. At concentrations higher than 100 ng μL−1, we observed a continuous reduction in ion collection time in order to limit the number of injected ions, which may be the reason for a larger fluctuation in mass spectrometric signals (see Figure S-6). Additionally, the relative standard deviations of peak areas were evaluated for interday and intraday quintuplicate injections of the 500 ng μL−1 standard. Intraday measurements show RSDs of 0.6−1.5% whereas interday measurements are in the range of 5.8−12%

Table 2. Limits of Detection (LOD) and Limits of Quantification (LOQ) for Each Quantification Methoda LOD [ng μL−1] (n = 5) LOQ [ng μL−1] (n = 5) RSD (%) (intraday, n = 5) RSD (%) (interday, n = 5) Vx0% (5−25 ng μL−1)d Vx0% (25−250 ng μL−1)d Vx0% (100−1000 ng μL−1)d

UV

MS, intact full-scan

MS, AIF-RTICC

MS, AIF-EICC (y47)

21.8b 65.4b 0.55c 5.75c n.d. 6.18 4.11

65.3b 195.9b n.d. 11.72c n.d. n.d. 13.57

0.85b 2.54b 1.49c 7.06c 2.40 13.52 14.92

1.39b 4.18b 1.41c 6.39c 3.91 13.16 13.66

a All values are mean values from five injections carried out within 3 days. bLODs calculated according to the regression line method using the slope b and the residual standard deviation sy of the 5−25 ng μL−1 calibration curves for AIF-RTICC and AIF-EICC, 25−250 ng μL−1 for the intact measurements, and 50−250 ng μL−1 for UV measurements. cInterday and intraday RSDs were evaluated for the 500 ng μL−1 calibration standard. d Relative process standard deviation Vx0 = ((sy/b)/x)·100% with sy, residual standard deviation of the calibration, b, slope of the calibration line, and ̅ x,̅ mean of standard concentrations.

9341

DOI: 10.1021/acs.analchem.5b02029 Anal. Chem. 2015, 87, 9336−9343

Analytical Chemistry

■

Article

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.5b02029. Elution of the carrier protein carbonic anhydrase, as well as the measurement of an expired pegfilgrastim sample, the UV chromatogram of a H2O2 stressed pegfilgrastim sample, an overlay of UV measurements of oxidized and nonoxidized pegfilgrastim on a nanophotometer, the comparison of fragment ion spectra of oxidized and nonoxidized variants, the reverse correlation of actual injection times and the relative standard deviations dependent on different standard concentrations (PDF) Table of relative standard deviations, calibration curves, and Mandel's fitting test (XLSX)

■

AUTHOR INFORMATION

Corresponding Author

*Phone: (+)43 662 8044 5738. Fax: (+)43 662 8044 5751. Email: [email protected].

Figure 4. MS, AIF-RTICC of an expired pegf ilgrastim batch. (a) RTICC, (b) zoomed RTICC (3% relative abundance), (c) mass spectrum of peak 1 with the oxidation sites assigned. Sample, 1.0 μL of 1.0 mg mL−1 pegf ilgrastim (Neulasta, exp. 09/2011), measurements done 08/2014; R = 140 000 at m/z 200, SID 70, NCE 12.

Notes

The authors declare the following competing financial interest(s): Hansjörg Toll, Johann Holzmann, and Ines C. Forstenlehner are employees of Sandoz GmbH, which provides financial support for the Christian Doppler Laboratory for Innovative Tools for Biosimilar Characterization. Christian G. Huber’s salary is partly funded by the Christian Doppler Laboratory for Biosimilar Characterization. The authors declare no other competing financial interest.

their concentration below 50 ng μL−1. Peak 1, consisting of the oxidized Met127 variant, yielded a concentration of 5.23 ± 0.18 ng μL−1 (0.76% of the total detected protein mass of 688 ng, corresponding to 7.6 mg·g−1) and peak 2, oxidized Met138, yielded 3.63 ± 0.18 ng μL−1 (0.53%, corresponding to 5.3 mg· g−1). Both concentrations were above the lower limit of quantification. Although Met 1 oxidation is likely in both impurities, it was not confirmed in this analysis due to insufficient abundance of the proteins for intact mass measurement. Here, we could show that our method developed for monitoring and quantifying oxidation variants of pegf ilgrastim is feasible for monitoring low levels of impurities of a relevant biopharmaceutical sample.

■

ACKNOWLEDGMENTS The financial support by the Austrian Federal Ministry of Economy, Family, and Youth, by the National Foundation of Research, Technology, and Development, and by a Start-up Grant of the State of Salzburg is gratefully acknowledged. We thank Kai Scheffler, Martin Samonig, and Jonathan Josephs from Thermo Fisher Scientific for technical support and scientific discussions. We also acknowledge Therese Wohlschlager for her valuable comments.

■

CONCLUSIONS All-ion fragmentation combined with full scan MS detection allows both oxidation site assignment and absolute quantification of oxidized methionine and nonoxidized variants of pegf ilgrastim. In-depth data analysis reveals that the monitoring of selected ions or a set of selected ions is not appropriate to reliably quantify oxidized variants using nonoxidized standards. A comparison of UV-absorbance and MS-based absolute quantification strategies reveals that each approach has an optimal concentration range. While the high-abundant main components are better quantified by UV detection, MS detection is the method of choice for the determination of low-concentrated impurities. Since both detection modes are not providing selectivity for certain analytes, adequate chromatographic separation of all protein variants prior to their detection is indispensable. The application of both methods in series allows a reliable quantification in a concentration range of 2.5−1000 ng μL−1 with relative process standard deviations below 6.5%. The feasibility of our approach for monitoring and absolutely quantifying low level impurities in a protein biopharmaceutical was successfully exemplified on an expired pegf ilgrastim sample.

■

REFERENCES

(1) Holmes, F. A.; Jones, S. E.; O’Shaughnessy, J.; Vukelja, S.; George, T.; Savin, M.; Richards, D.; Glaspy, J.; Meza, L.; Cohen, G.; Dhami, M.; Budman, D. R.; Hackett, J.; Brassard, M.; Yang, B. B.; Liang, B. C. Ann. Oncol. 2002, 13, 903−909. (2) Piedmonte, D. M.; Treuheit, M. J. Adv. Drug Delivery Rev. 2008, 60, 50−58. (3) Jevsevar, S.; Kunstelj, M.; Porekar, V. G. Biotechnol. J. 2010, 5, 113−128. (4) Harris, J. M.; Chess, R. B. Nat. Rev. Drug Discovery 2003, 2, 214− 221. (5) Huang, L. H.; Gough, P. C.; DeFelippis, M. R. Anal. Chem. 2009, 81, 567−577. (6) Schellekens, H. Nat. Rev. Drug Discovery 2002, 1, 457−462. (7) Sharma, B. Biotechnol. Adv. 2007, 25, 310−317. (8) Silva, M. M.; Lamarre, B.; Cerasoli, E.; Rakowska, P.; Hills, A.; Bailey, M. J.; Wheeler, J. X.; Burns, C. J.; Gaines-Das, R. E.; Jones, C.; Robinson, C. J. Biologicals 2008, 36, 383−392. (9) Skrlin, A.; Kosor Krnic, E.; Gosak, D.; Prester, B.; Mrsa, V.; Vuletic, M.; Runac, D. J. Pharm. Biomed. Anal. 2010, 53, 262−268. (10) Gucinski, A. C.; Boyne, M. T., 2nd Anal. Chem. 2012, 84, 8045− 8051.

9342

DOI: 10.1021/acs.analchem.5b02029 Anal. Chem. 2015, 87, 9336−9343

Article

Analytical Chemistry (11) Holzmann, J.; Hausberger, A.; Rupprechter, A.; Toll, H. Anal. Bioanal. Chem. 2013, 405, 6667−6674. (12) Molineux, G. Curr. Pharm. Des. 2004, 10, 1235−1244. (13) Lu, H. S.; Fausset, P. R.; Narhi, L. O.; Horan, T.; Shinagawa, K.; Shimamoto, G.; Boone, T. C. Arch. Biochem. Biophys. 1999, 362, 1−11. (14) Yin, J.; Chu, J. W.; Ricci, M. S.; Brems, D. N.; Wang, D. I.; Trout, B. L. Pharm. Res. 2005, 22, 141−147. (15) Yin, J.; Chu, J. W.; Ricci, M. S.; Brems, D. N.; Wang, D. I.; Trout, B. L. Pharm. Res. 2004, 21, 2377−2383. (16) Bagal, D.; Zhang, H.; Schnier, P. D. Anal. Chem. 2008, 80, 2408−2418. (17) Houde, D.; Kauppinen, P.; Mhatre, R.; Lyubarskaya, Y. J. Chromatogr. A 2006, 1123, 189−198. (18) Griffiths, S. W.; Cooney, C. L. J. Chromatogr. A 2002, 942, 133− 143. (19) Zang, L.; Carlage, T.; Murphy, D.; Frenkel, R.; Bryngelson, P.; Madsen, M.; Lyubarskaya, Y. J. Chromatogr. B: Anal. Technol. Biomed. Life Sci. 2012, 895−896, 71−76. (20) Sandra, K.; Vandenheede, I.; Sandra, P. J. Chromatogr. A 2014, 1335, 81−103. (21) Heudi, O.; Barteau, S.; Zimmer, D.; Schmidt, J.; Bill, K.; Lehmann, N.; Bauer, C.; Kretz, O. Anal. Chem. 2008, 80, 4200−4207. (22) Ji, C.; Sadagopan, N.; Zhang, Y.; Lepsy, C. Anal. Chem. 2009, 81, 9321−9328. (23) Mayr, B. M.; Kohlbacher, O.; Reinert, K.; Sturm, M.; Gröpl, C.; Lange, E.; Klein, C.; Huber, C. G. J. Proteome Res. 2006, 5, 414−421. (24) Kirkpatrick, D. S.; Gerber, S. A.; Gygi, S. P. Methods 2005, 35, 265−273. (25) Yang, Z.; Ke, J.; Hayes, M.; Bryant, M.; Tse, F. L. J. Chromatogr. B: Anal. Technol. Biomed. Life Sci. 2009, 877, 1737−1742. (26) Wu, S. T.; Ouyang, Z.; Olah, T. V.; Jemal, M. Rapid Commun. Mass Spectrom. 2011, 25, 281−290. (27) Du, Y.; Wang, F.; May, K.; Xu, W.; Liu, H. Anal. Chem. 2012, 84, 6355−6360. (28) Perdivara, I.; Deterding, L. J.; Przybylski, M.; Tomer, K. B. J. Am. Soc. Mass Spectrom. 2010, 21, 1114−1117. (29) Li, H.; Rose, M. J.; Holder, J. R.; Wright, M.; Miranda, L. P.; James, C. A. J. Am. Soc. Mass Spectrom. 2011, 22, 1660−1667. (30) Forstenlehner, I. C.; Holzmann, J.; Scheffler, K.; Wieder, W.; Toll, H.; Huber, C. G. Anal. Chem. 2014, 86, 826−834. (31) Ruan, Q.; Ji, Q. C.; Arnold, M. E.; Humphreys, W. G.; Zhu, M. Anal. Chem. 2011, 83, 8937−8944. (32) Huber, C. G.; Premstaller, A. J. Chromatogr. A 1999, 849, 161− 173. (33) LeDuc, R. D.; Taylor, G. K.; Kim, Y. B.; Januszyk, T. E.; Bynum, L. H.; Sola, J. V.; Garavelli, J. S.; Kelleher, N. L. Nucleic Acids Res. 2004, 32, W340−345. (34) Ji, Q. C.; Rodila, R.; Gage, E. M.; El-Shourbagy, T. A. Anal. Chem. 2003, 75, 7008−7014. (35) Kalli, A.; Smith, G. T.; Sweredoski, M. J.; Hess, S. J. Proteome Res. 2013, 12, 3071−3086. (36) Bateman, K. P.; Kellmann, M.; Muenster, H.; Papp, R.; Taylor, L. J. Am. Soc. Mass Spectrom. 2009, 20, 1441−1450.

9343

DOI: 10.1021/acs.analchem.5b02029 Anal. Chem. 2015, 87, 9336−9343