Evaluation of Intact Mass Spectrometry for the Quantitative Analysis of

Aug 23, 2012 - Peter Nemes , William J. Hoover , and David A. Keire. Analytical ... Stanley Weihua Zhang , Jonathan Crowther , Wenying Jian. 2017,129-...
0 downloads 0 Views 573KB Size
Download PDF
Article pubs.acs.org/ac

Evaluation of Intact Mass Spectrometry for the Quantitative Analysis of Protein Therapeutics Ashley C. Gucinski and Michael T. Boyne, II* Division of Pharmaceutical Research, Office of Testing and Research, Center for Drug Evaluation and Research, U.S. Food and Drug Administration, 1114 Market Street, Room 1002, St. Louis, Missouri 63101, United States S Supporting Information *

ABSTRACT: Implementation of modern analytical techniques, such as intact mass spectrometry, may allow for more detailed quality assessments of protein therapeutics. The complexity of the protein therapeutic manufacturing process as well as the sensitivity of these drugs to different storage conditions can lead to the presence of several undesired products, including truncations, degradation products, byproducts, and differentially modified protein variants that are difficult to detect by peptide mapping. Intact mass spectrometry can be used to identify the intact protein composition, inclusive of post-translational modifications (PTMs) but can also generate a chemical fingerprint of the different protein species present in a given sample. In this work, we systematically evaluated the influence of multiple charge states, multiple isotopes per charge state, and operating resolution on the suitability of intact mass spectrometry for quantitative analysis using insulin and somatotropin as model systems. Standard curves could be generated using absolute intensity data or using the relative ratio between the analyte and internal standard. These methods demonstrate the validity of quantitative intact mass spectrometry for the analysis of protein therapeutic drugs, thus providing a foundation for future comparative methods.

P

absence of differential post-translational modifications elsewhere in the protein. Similarly, truncations of the protein at different regions could all generate the same resulting peptide fragment ion, preventing protein variants with small structural changes yet potentially large functional changes from being distinguished. Intact mass spectrometry is a potentially advantageous tool for the quantitative analysis of protein therapeutics. In contrast to indirect and MS/MS-based approaches, intact mass spectrometry is not only able to directly measure the protein of interest, including distinguishing closely related forms, but is also selective to the presence and/or absence of post-translational modifications.9−11 Two types of intact MS experiments have been performed on large protein systems: selected-ion or reaction monitoring (SIM/SRM)12,13 and full scan intact MS experiments.14,15 While SIM and SRM experiments directly measure the intact protein, quantification is based on the intensity of a selected fragment ion or fragmentation pathway. While label free, SIM and SRM experiments are vulnerable to the same pitfalls as peptide mass mapping experiments: if variations, adulterants and/or degradation products are present that only reflect minor changes in the protein structure, the reporting fragment ion may not be affected, and the quantification would not accurately reflect the populations of different protein species that may be present. In contrast to peptide mass mapping and SIM/SRM approaches, full scan intact MS-based quantification has also been explored. The Muddiman laboratory used bovine heart

rotein and peptide-based drugs represent a large and growing market in the United States. These protein therapeutics pose a unique analytical challenge for their structural characterization and evaluation of their quality.1 Unlike small molecule drugs, protein therapeutics are made via DNA expression techniques, which often form unintended protein products in addition to the desired protein product. These unintended products can include degradants, byproducts, and differentially modified variations of the intended protein. The location and type of a post-translational modification (PTM) can be crucial for the proper structure and function of a protein.2,3 Additionally, slight changes in the manufacturing or protein synthesis processes may also result in changes in bioactivity, stability, or composition that can also compromise the safety of the drug product.4 Moreover, increasing concerns about the potential for economically motivated adulteration of protein therapeutics require nonproprietary approaches to assess protein product quality. Clearly, sensitive and specific analytical methods are necessary to identify and quantify the presence of any impurities or variant forms present in order to assess the overall purity and quality of a protein therapeutic drug product. Previous efforts to use mass spectrometry for the qualitative and quantitative analysis of proteins have relied on the use of peptide mass mapping.5−8 In this approach, isotopically labeled or derivatized peptide sequences are spiked into mixtures of peptides generated from an enzymatic digest of the protein of interest, and the relative intensity ratio between the labeled and nonlabeled peptide ions are used to quantify the amount of protein present. However, any given peptide may not be truly indicative of the full protein species. For example, the peptide selected may not be sensitive to the presence or This article not subject to U.S. Copyright. Published 2012 by the American Chemical Society

Received: July 11, 2012 Accepted: August 23, 2012 Published: August 23, 2012 8045

dx.doi.org/10.1021/ac301949j | Anal. Chem. 2012, 84, 8045−8051

Analytical Chemistry

Article

cytochrome c as an internal standard for the quantification of equine heart cytochrome c.14 Similarly, Ruan and co-workers evaluated the suitability of high-resolution intact mass spectrometry for the quantitative analysis of lysozyme spiked in monkey plasma.15 An intact LC−MS method developed by Roth and co-workers using an LTQ-Orbitrap mass spectrometer was able to obtain subfemtomole detection limits for a set of standard proteins.16 However, all of these approaches sought to identify the amount of a particular protein. From a regulatory perspective, a clear need exists to not only identify and characterize the active pharmaceutical ingredient(s) but also any unintended protein species that may be present. This is necessary not only to examine the stability and homogeneity of the drug product, which would be indicated by the presence of differentially modified variants and truncated or degraded products but also to identify the amount of any potentially dangerous adulterant that is present. Variation between all of these possible protein species may be very small, so it is important to establish a foundation for identifying and quantifying changes between samples that include closely related species. In contrast to small molecule drugs, the mass spectra for protein therapeutics feature many additional levels of complexity. While small molecule drugs are generally singly charged with only small carbon-13 isotope contributions, protein therapeutics feature multiple charge states all corresponding to the same protein species, with multiple isotopes observed for each charge state. The influence of these complexities not only on the quantitative behavior of intact MS but also on the precision and accuracy of the experiment is unknown and must be explored. Furthermore, different types of mass spectrometers afford different levels of resolving power, which may or may not allow for the isotopic distributions of each charge state to be isotopically resolved. While lower resolution instruments may not be able to distinguish between two very similar protein variants, ultrahigh-resolution instruments require much longer detection times that may also affect the quality of the quantitative measurement; the combined influence of and need for ultrahigh operating resolution for monoisotopic resolution should also be considered. Lastly, the availability and suitability of appropriate internal standards for protein therapeutics must also be taken into consideration. USP reference materials and/or isotopically labeled synthetic analogues are widely available for small molecule drugs but typically not for protein therapeutics; the complexity of producing these compounds for large protein therapeutics generally limits these products to proprietary use, making the selection of an appropriate internal standard even more challenging. All of these factors need to be taken into consideration when evaluating the ability of a method to not only differentially quantify multiple different protein therapeutic species present in a given sample but also to determine relative amounts of each of the different species present. Figure 1 summarizes the different features of the protein mass spectra that were considered in the development of quantitative intact MS methods.

Figure 1. Schematic of variables explored for evaluation of intact mass spectrometry for quantitative analysis.

in the Supporting Information, Figure S1). Small molecule formulation components were removed by using Amicon 3 kDa MWCOs (Millipore, Billerica, MA) for both the insulin and somatotropin samples. Samples were dried by speed-vacuum before dilution and reconstitution in ESI buffer (49:49:2 water/ acetonitrile/formic acid) for mass spectrometric analysis. These and all other solvents and reagents were obtained from Sigma Aldrich (St. Louis, MO) and used without further purification. Protein standard solutions were prepared over the concentration range from 1 nM to 10 μM; for insulin samples, standards were prepared with or without the presence of a second insulin variant to act as an internal standard, model contaminant, and/or adulterant. Orbitrap Mass Spectrometry Experiments. Intact MS experiments were performed using a LTQ-Orbitrap XL (Thermo Electron, San Jose, CA) with an IonMax ESI source. Samples were introduced into the mass spectrometer using a syringe pump at a flow rate of 5 μL/min with a capillary voltage of ∼4 kV. Triplicate spectra were acquired for each data point, with each spectrum being the composite of 100 averaged scans. Data processing was performed using Thermo XCalibur QualBrowser (version 2.0.7). Insulin MS spectra were collected at an operating resolution of 30 000 at m/z 400, whereas hGH spectra were collected at operating resolutions of either 7 500, 30 000, or 100 000 at m/z 400. Method Validation Parameters. In order to evaluate the goodness of fit of the intact MS standard curves, the criteria delineated in the FDA Bioanalytical Method Validation Guide were used as a guideline.17 At least 75% of all data points in the linear dynamic range must be within ±15% of the calculated fit value, with the lower limit of quantitation (LLOQ) data point as the only exception, allowed to be within ±20%. Limit of detection (LOD) and LLOQ values were calculated by using 3.3 or 10 times the signal-to-noise ratio observed at the lowest concentration in the linear dynamic range (LDR). In all cases, the signal intensity of the lowest data point measured in the standard curve was much greater than 10 times the S/N ratio; as such, the lowest point in the linear range is reported as both the LOD and LLOQ as it is the lowest concentration data point that satisfies the precision, accuracy, and linearity requirements.

■

EXPERIMENTAL SECTION Two commercially available insulin variants were obtained: insulin regular and insulin detemir (both from Novo Nordisk, Seattle, WA). Human growth hormone (hGH, somatotropin) was purchased from Sandoz, Inc. (Princeton, NJ). The two insulin variants differ by the addition of a myristoic acid on lysine at the B29 position and the loss of the most C-terminal threonine residue for insulin detemir (sequences can be found

■

RESULTS AND DISCUSSION Two sets of insulin standard curves were generated using infusion ESI-MS on the LTQ Orbitrap XL mass spectrometer: both an insulin detemir standard curve in the presence of a 25 nM insulin regular internal standard and an insulin regular 8046

dx.doi.org/10.1021/ac301949j | Anal. Chem. 2012, 84, 8045−8051

Analytical Chemistry

Article

Figure 2. (left) Intact mass spectrum of an equimolar insulin regular (R)/insulin detemir (D) mixture. Each insulin species has a concentration of 2 μM. At this concentration, charge states from +6 to +3 are observed and (right) isotopic distribution of the +5 charge state from 2 μM insulin regular.

Figure 3. Absolute (left) and relative (right) quantitative standard curves for insulin detemir in the presence of a 25 nM insulin regular internal standard. Absolute intensity data show curves using one, two, three, or five summed isotopes, whereas the relative ratio plot uses the most intense isotope of the most intense charge state (+5, m/z 1184).

standard curve in the presence of a 2 μM insulin detemir internal standard were generated in order to examine the ability to quantify different protein therapeutic species present in excess or trace amounts relative to the targeted protein species. The intact mass spectrum of an equimolar insulin regular: detemir mixture is shown in Figure 2 along with the isotopic distribution of the insulin regular +5 charge state (m/z 1162). Between one and five of the most abundant isotopes from each charge state were summed to explore the influence of the number summed on the goodness of fit of the linear curve. The availability of differentially modified variations of insulin affords their use as both model internal standards and as model impurities. Both absolute and relative quantitation standard curves for the insulin detemir analyte system using the most intense charge state are shown in Figure 3. The quantitative parameters of the standard curves are summarized in Tables 1 and 2 for the insulin detemir and insulin regular standard curves, respectively. Standard Curve No. 1: Variable Insulin Detemir with 25 nM Insulin Regular Internal Standard. In the presence of the 25 nM insulin regular internal standard, the largest LDR using a single charge state was achieved by constructing a standard curve using the intensity of the second most intense charge state (+4, m/z 1480) of insulin detemir. This curve was linear from 50 to 5000 nM (i.e., a concentration range from 2 to 200-fold greater than that of the internal standard) with R2 values of 0.9994, using one, two, three, or five most intense summed isotopes from the +4 charge state. The % RSD values ranged from 0.58 to 3.17%, showing that the standard curve

using the second most intense charge state also features high precision. This result is important because in a potentially complex biological product sample, the isotopic distribution of one charge state may overlap with that of a degradant or degradant with a salt adduct, etc., skewing the result. Absolute quantification by intact MS does not require that the most intense charge state be used but is also possible for multiple charge states for a given protein system. For this example, the standard curve generated using the most intense charge state (with any number of summed isotopes) afforded two separate linear regions, with small LDRs of 1 order of magnitude each (25−250 nM or 250 nM− 2500 nM). Despite the small linear ranges, the R2 and % RSD values for this standard curve were very high and low, respectively (R2 = 0.9991 or 0.9929, % RSDs ranged from 0.16− 1.17% or 0.16−4.28%), demonstrating that a good linear fit was still possible. As this method is meant for the quantitative analysis of drug product, the sample of interest can be diluted into the appropriate linear range for analysis after construction of the standard curve because of the likely excess of protein therapeutic available in relation to the sensitivity of the MS experiment. While quantitative information can be obtained using a single charge state independently, the largest LDRs were observed using the summed intensity of the two or three most abundant charge states. Both the sum of the two and the sum of the three most intense charge states featured LDRs from 5 to 750 nM, covering over 2 orders of magnitude, with R2 values greater than 0.999. In addition, the standard curve using the absolute 8047

dx.doi.org/10.1021/ac301949j | Anal. Chem. 2012, 84, 8045−8051

Analytical Chemistry

Article

Table 1. Summary of Linearity and Precision of Insulin Detemir Standard Curves in the Presence of 25 nM Insulin Regular Internal Standard charge states (CS) used most intense charge state (+5)

no. of summed isotopes per CS one two three five

second most intense charge state (+4)

one two three five

third most intense charge state (+6)

one two three five

+5 and +4

one two three five

+5, +4, and +6

one two three five

LDR absolute (LDR relative)

R2 absolute (R2 relative)

% RSD range absolute (% RSD range relative)

25−250 nM; 250−2500 nM (25−500 nM; 100−5000 nM) 25−250 nM; 500−2500 nM (25−500 nM; 250−5000 nM) 25−250 nM; 250−2500 nM (25−500 nM; 250−5000 nM) 25−250 nM; 250−2500 nM (25−500 nM; 250−5000 nM) 50−5000 nM (100−2500 nM) 50−5000 nM (250−2500 nM) 50−5000 nM (250−2500 nM) 50−5000 nM (250−2500 nM) − (500−5000 nM) − (500−5000 nM) − (500−5000 nM) − (500−5000 nM) 5−750 nM; 250−5000 nM (50−500 nM; 250−5000 nM) 5−750 nM; 250−5000 nM (50−500 nM; 250−2500 nM) 5- 750 nM; 250−5000 nM (50−500 nM; 250−2500 nM) 5−750 nM; 250−5000 nM (50−500 nM; 250−2500 nM) 5−750 nM; 250−5000 nM (50−500 nM; 250−5000 nM) 5−750 nM; 250−2500 nM (50−500 nM; 250−2500 nM) 5−750 nM; 250−2500 nM (50−500 nM; 250−2500 nM) 5−750 nM; 250−2500 nM (50−500 nM; 250−2500 nM)

0.9991; 0.9929 (0.9994; 0.9993) 0.9991; 0.9915 (0.9995; 0.9994) 0.9991; 0.9928 (0.9998; 0.9994) 0.9991; 0.9928 (0.9997; 0.9993) 0.9994 (0.9992) 0.9994 (0.9986) 0.9994 (0.9983) 0.9994 (0.9990) − (0.9995) − (0.9993) − (0.9990) − (0.9986) 0.9994; 0.9949 (0.9999; 0.9998) 0.9994; 0.9948 (0.9999; 0.9997) 0.9994; 0.9948 (0.9999; 0.9983) 0.9994; 0.9948 (0.9998; 0.9994) 0.9993; 0.9944 (0.9998; 0.9995) 0.9993; 0.9944 (0.9998; 0.9980) 0.9993; 0.9944 (0.9998; 0.9980) 0.9993; 0.9944 (0.9998; 0.9980)

0.16−1.17%; 0.16−4.28% (0.56−1.43%; 0.56−2.99%) 0.29−0.85%; 0.21−4.14% (0.76−1.70%; 0.58−2.39%) 0.12−1.04%; 0.12−4.01% (0.27−1.75%; 0.19−2.27%) 0.22−1.23%; 0.16−3.90% (0.20−1.36%; 0.54−1.95%) 0.58−3.17% (0.58−4.10%) 0.55−3.26% (0.48−3.46%) 0.35−3.14% (1.18−2.61%) 0.45−3.21% (1.38−4.13%) − (0.49−4.78%) − (0.66−5.61%) − (1.29−6.45%) − (1.77−5.43%) 0.31−13.98%; 0.25−3.80% (0.28−1.19%; 0.55−3.24%) 0.38−13.96%; 0.21−3.76% (0.45−1.51%; 0.28−3.18%) 0.20−14.75%; 0.20−3.64% (0.41−1.65%; 0.64−2.27%) 0.28−14.25%; 0.21−3.61% (0.32−1.58%; 0.36−1.97%) 0.29−13.98%; 0.26−3.91% (0.28−1.13%; 0.31−2.95%) 0.35−13.98%; 0.23−3.86% (0.25−1.50%; 0.25−2.27%) 0.19−14.84%; 0.19−3.75% (0.32−1.48%; 0.46−2.49%) 0.27−14.40%; 0.20−3.73% (0.18−1.46%; 0.19−2.58%)

quantitative results, standard curves using fewer (or even a single) charge states and isotope per charge states are adequate for the absolute quantification of proteins using intact MS. Standard curves based on the relative intensity and concentration ratios between insulin detemir and the 25 nM insulin regular internal standards were also generated. The largest LDR was observed using the most intense isotope of the most intense charge state (100−5000 nM, R2 = 0.9994), demonstrating that linear behavior was observed when the analyte was between 4 and 200 times more concentrated than the internal standard protein signal. Incorporation of additional isotopes from the +5 charge state narrowed the LDR to 250− 5000 nM, similar to the effect of the incorporation of additional isotopes per charge state observed in the absolute quantification data. Adequate relative standard curves could also be generated using the +4 charge state or the summed two or three most intense charge states, further demonstrating the suitability of this technique for protein therapeutic quantification.

intensity of the two most intense charge states featured a second linear range from 250 to 5000 nM (R2 > than 0.994), while the sum of the three most intense charge states featured a second linear range from 250 to 2500 nM. The LDRs observed were consistent independent of the number of isotopes per charge state (one, two, three, or five) used to construct the standard curve. As the number of summed isotopes per charge state did not alter the LDR, this suggests that the incorporation of additional isotopes from a given charge state (or charge states) is unnecessary and does not improve the quality of the standard curve. In light of this, only standard curves constructed using the most abundant isotope will be discussed for the sake of simplicity. In related work by Silva and co-workers, the authors stated that the averaged signal responses of three different peptide signals was necessary for absolute quantification of proteins using LCMSE.18 Our results show that while standard curves using the average intensity of three charge states yield 8048

dx.doi.org/10.1021/ac301949j | Anal. Chem. 2012, 84, 8045−8051

Analytical Chemistry

Article

Table 2. Summary of Linearity and Precision of Insulin Regular Standard Curves in the Presence of 2 μM Insulin Detemir Internal Standarda no. of summed % RSD range charge isotopes per LDR absolute R2 absolute absolute (% RSD 2 CS (LDR relative) (R relative) range relative) states used

no. of summed % RSD range charge isotopes per LDR absolute R2 absolute absolute (% RSD 2 states used CS (LDR relative) (R relative) range relative) one two +5 three five one two +6 three five one two +4 three five one +3

two three

a

20−2500 nM (20−2500 nM) 20−2500 nM (20−2500 nM) 20−2500 nM (20−2500 nM) 20−2500 nM (20−2500 nM) 20−2000 nM (20−2000 nM) 20−2000 nM (20−2000 nM) 20−2000 nM (20−2000 nM) 20−2000 nM (20−2000 nM) 250−4000 nM (100−4000 nM) 250−4000 nM (100−4000 nM) 250−4000 nM (250−4000 nM) 250−4000 nM (250−4000 nM) − (250−4000 nM) − (250−4000 nM) − (250−4000 nM)

0.9977 (0.9965) 0.9976 (0.9964) 0.9977 (0.9964) 0.9977 (0.9964) 0.9975 (0.9986) 0.9974 (0.9986) 0.9974 (0.9986) 0.9974 (0.9986) 0.9981 (0.9976) 0.9980 (0.9971) 0.9980 (0.9974) 0.9980 (0.9980) − (0.9983) − (0.9983) − (0.9984)

−

0.06−2.45% (0.15−3.03%) 0.01−2.14% (0.08−2.82%) 0.19−1.81% (0.28−2.86%) 0.03−1.65% (0.26−2.82%) 0.62−3.39% (0.57−3.82%) 0.54−2.35% (0.59−3.48%) 0.44−3.32% (0.67−3.79%) 0.38−3.22% (0.47−4.40%) 0.27−2.47% (0.48−19.41%) 0.05−2.14% (0.39−24.94%) 0.36−2.20% (0.36−17.10%) 0.17−2.31% (0.75−5.23%) − (0.57−6.62%) − (0.63−6.77%) − (1.02−6.17%)

five

+5 and +4

+5, +4, and +6

+5, +4, +6, and +3

−

(250− (0.9983) 4000 nM) 20−2500 nM one (10−2500 nM) 20−2500 nM two (20−2500 nM) 20−2500 nM three (20−2500 nM) 20−2500 nM five (20−2500 nM) 40−2500 nM one (10−2000 nM) 40−2500 nM two (40−2500 nM) 40−2500 nM three (40−2500 nM) 40−2500 nM five (40−2500 nM) 10−4000 nM one (40−2500 nM) 10−4000 nM two (40−2500 nM) 10−4000 nM three (40−2500 nM) 10−4000 nM (40−2500 nM) five

− (0.72− 6.20%) 0.9979 (0.9970) 0.9978 (0.9967) 0.9979 (0.9968) 0.9979 (0.9971) 0.9977 (0.9967) 0.9976 (0.9965) 0.9977 (0.9966) 0.9976 (0.9969) 0.9983 (0.9968) 0.9983 (0.9967) 0.9983 (0.9968) 0.9983 (0.9971)

0.17−2.73% (0.25−7.50%) 0.17−2.17% (0.13−9.09%) 0.21−2.31% (0.17−6.66%) 0.26−2.09% (0.04−3.19%) 0.07−1.57% (0.42−7.26%) 0.14−1.47% (0.09−8.60%) 0.20−1.55% (0.28−6.43%) 0.33−1.59% (0.28−3.00%) 0.03−3.69% (0.41−7.03%) 0.14−3.69% (0.13−8.16%) 0.22−2.91% (0.31−6.08%) 0.33−2.69% (0.28−2.98%)

No distinct linear range observed for +3 charge state using absolute intensity alone.

Standard Curve No. 2: Variable Insulin Regular with 2 μM Insulin Detemir Internal Standard. Both absolute and relative quantification standard curves were generated for insulin regular in the presence of a 2 μM insulin detemir internal standard. While the previous example explored the relative range of linearity when an impurity is present in great excess relative to the control (internal standard) protein, the use of a concentrated internal standard mimics the scenario of having degradant or other unintended products in a smaller relative abundance in comparison to the intended product. While a LDR from 20 to 2500 nM was obtained when using one isotope from the most intense charge state to generate the standard curve (1−125% of the internal standard concentration), the largest insulin regular LDR (10−4000 nM, R2 = 0.9983) was obtained from the standard curve constructed using the sum of the most intense isotope from all four observed charge states (+3, +4, +5, and +6). Using one charge state, the % RSD ranged from 0.06 to 2.45%, while using the sum of all four charge states yielded a % RSD range from 0.03 to 3.69%. Therefore, using absolute intensity, the concentration of a second insulin species can be determined when the second insulin species is between 2-fold more concentrated than the analyte and 200-fold less concentrated than the analyte. The standard curves generated using the 2 μM internal standard present larger LDRs than those obtained in the presence of a 25 nM insulin internal standard. Adjusting the

automatic gain control on the Orbitrap and increasing the concentration of the internal standard eliminated the switch in data acquisition from maximum injection times to a maximum number of ions, which was the presumed cause of the dual linear ranges in the earlier example. In terms of the suitability for this method to be used in a regulatory setting, the latter scenario demonstrates promise, as the relative amount of degradant and impurity products present in a drug product are likely to exist at low levels relative to the amount of the drug product. In order to address the accuracy of the technique, an additional insulin regular sample with a concentration of 1.75 μM was analyzed. Using the one isotope−one charge state standard curve, the 1.75 μM sample featured a percent recovery of 107.0% with a % RSD precision value of 1.24%, while the one isotope−four charge state standard curve yielded a percent recovery value of 109.1% with a % RSD of 1.16%. Both accuracy values are within the generally accepted range of 80− 120%, in accord with the FDA Bioanalytical Method Validation Guidance for Industry.17 In order to determine at what level of precision one insulin variant can be quantified in the presence of another, a twotailed t test was performed on the two lowest data points of the LDR (i.e., 10 and 20 nM for the standard curve using one isotope from four charge states, 20 and 40 nM, for the standard curve using one isotope and one charge state). For the standard curve using one isotopes from four charge states, the t test P value 8049

dx.doi.org/10.1021/ac301949j | Anal. Chem. 2012, 84, 8045−8051

Analytical Chemistry

Article

for 10 and 20 nM was 0.0013. As a P value less than 0.05 is considered to be statistically significant, a protein at a concentration of 0.5% of the internal standard concentration can be differentiated from a protein present at a concentration of 1% of the internal standard concentration. Similarly, the t test for the one isotope−one charge state standard curve yielded a P value of 0.0017, demonstrating that the 20 nM (1%) and 40 nM (2%) samples can be distinguished from one another. Similar to the absolute quantification data, relative quantification of insulin regular in the presence of 2 μM insulin detemir features the best LDR when constructing the standard curve using the most intense isotope of the most intense charge state (LDR 20−2500 nM, R2 = 0.9965, % RSD range from 0.15 to 3.03%). Again, adequate standard curves for relative quantification could be generated using any of the four observed charge states independently or in summation, with one, two, three, or five summed isotopes per charge state (see Table 2). Of note is the ability to generate a relative quantification curve for the +3 charge state, the least intense of the four observed charge states for the insulin variants. Using absolute intensity alone, no clear linear trend can be elucidated. However, as the signal fluctuation of the +3 charge state of the similar insulin variant internal standard is similar, the relative ratio is linear over the range from 250 to 4000 nM. Influence of Operating Resolution. In order to demonstrate the applicability of this technique for larger protein therapeutic species and explore the influence of resolving power on the quantification of such larger protein species, calibration curves were generated for somatotropin using infusion ESI-MS on the LTQ Orbitrap XL mass spectrometer at three different operating resolutions: 7 500, 30 000, and 100 000 at m/z 400. The charge state distribution of somatotropin, along with a magnification of the most abundant +15 charge state at each operating resolution, is shown in Figure 4. While baseline isotopic resolution is achieved using Rs = 100 000, the entire charge state envelope is observed as a single peak using operating resolutions of either 30 000 or 7 500. (Note that this is in contrast to insulin, MW 5808, where isotopic resolution was easily obtained when using an operating resolution of 30 000.) Standard curves were generated using the absolute intensity of the +15 charge state at each operating resolution, and the figures of merit of each fit are summarized in Table 3. In light of the previously discussed results for insulin, only the most abundant isotope of the +15 charge state was used for construction of the calibration curve at Rs = 100 000. Excellent linear agreement between the intensity of the somatotropin +15 charge state and its concentration were observed at all three operating resolutions, with each LDR spanning greater than 1 order of magnitude. While a LDR of 2 orders of magnitude was observed for Rs = 100 000, adequate standard curves were also obtained using operating resolutions of 7 500 or 30 000. Additionally, the average % RSD values for each of the three different resolution standard curves are less than 8%; the majority of % RSD values determined are less than 4%, which comments on the precision of the method. These results suggest that while ultra-high mass resolution capable instruments may improve the quality of the calibration by increasing specificity, such instruments are not essential in order to use intact mass spectrometry as a quantitative tool. However, the use of low-resolution instruments for this type of quantitative analysis relies on the absence of other spectral interferences in the same m/z range; a preliminary analysis using a highresolution instrument to screen for the presence of any spectral

Figure 4. Intact MS spectrum of 1 μM somatotropin (hGH). Inset: Comparison of isotopic distribution of +15 charge state at operating resolutions of 7 500, 30 000, or 100 000 at m/z 400.

Table 3. Comparison of Quality of Linear Fit at Different Operating Resolutions for Somatotropin LDR R2 % RSD range % RSD average

Rs = 7 500

Rs = 30 000

Rs = 100 000

100−1500 nM 0.9981 1.26−6.86% 3.70%

50−1000 nM 0.9993 2.30−28.34% 7.95%

25−2500 nM 0.9986 0.40−22.50% 5.46%

interferences within a given m/z range would provide additional confidence that any m/z signal observed would correspond to the protein of interest, facilitating the use of lowresolution instruments for intact quantitative analysis. Influence of Charge States and Number of Isotopes per Charge State for Somatotropin. While the standard plots for insulin detemir shown in Figure 3 and the somatotropin results for the Rs = 100 000 experiments were constructed using only the most intense isotope per charge state, it is unknown how many isotopes per charge state or how many charge states are required in order to accurately describe the quantitative behavior of larger protein therapeutic systems using intact mass spectrometry. To this end, the top four charge states of somatotropin were explored both independently and in sum, using the top 1, 2, 3, 5, 7, or 10 isotopes from each charge state for the somatotropin model system (MW 22 125 Da). The LDR, R2, % RSD ranges, and averages are reported in the Supporting Information, Table S1. Similar to the results observed for insulin, adequate standard curves could be generated using any of the top four most abundant charge states independently or in summation. The largest LDR was obtained when using the most intense isotope from the most intense charge state (25−2500 nM, R2 = 0.9986, average % RSD 5.35%), with the next largest LDR observed when using the summed intensity of the most intense isotope from the four most intense charge states (50−2000 nM, R2 = 0.9991, average % RSD 5.70%). No enhancement of the LDR was observed with the incorporation of additional isotopes from a given charge state; for the one charge state scenario, the LDR was unchanged when 1, 2, or 3 isotopes were summed together, but the LDR narrowed when using 5, 7, or 10 charge states. No distinct pattern relating the % RSD and the number of charge states or isotopes per charge state used to construct the standard plots was observed. While the LDR of the calibration varied depending on the number of summed charge states and 8050

dx.doi.org/10.1021/ac301949j | Anal. Chem. 2012, 84, 8045−8051

Analytical Chemistry

■

isotopes, all combinations of charge states and isotopes per charge state explored were able to demonstrate in some capacity linear behavior with respect to concentration.

REFERENCES

(1) Baker, K. N.; Rendall, M. H.; Patel, A.; Boyd, P.; Hoare, M.; Freedman, R. B.; James, D. C. Trends Biotechnol. 2009, 20, 149. (2) Terzoglou, A. G.; Routsias, J. G.; Moutsopoulos, H. M.; Tzioufas, A. G. Clin. Exp. Immunol. 2006, 146, 60. (3) Cloos, P. A. C.; Christgau, S. Biogerontology 2004, 5, 139. (4) Srebalus Barnes, C. A.; Lim, A. Mass Spectrom. Rev. 2007, 26, 370. (5) Ji, C.; Sadagopan, N.; Zhang, Y.; Lepsy, C. Anal. Chem. 2009, 81, 9321. (6) Heudi, O.; Barteau, S.; Zimmer, D.; Schmidt, J.; Bill, K.; Lehmann, N.; Bauer, C.; Kretz, O. Anal. Chem. 2008, 80, 4200. (7) Ye, H.; Hill, J.; Kauffman, J.; Gryniewicz, C.; Han, X. Anal. Biochem. 2008, 379, 182. (8) Ye, H.; Hill, J.; Kauffman, J.; Han, X. Anal. Biochem. 2010, 400, 46. (9) Kelleher, N. L. Anal. Chem. 2004, 196A. (10) Siuti, N.; Kelleher, N. L. Nat. Methods 2007, 4, 817. (11) Capriotti, A. L.; Cavaliere, C.; Foglia, P.; Samperi, R.; Lagana, A. J. Chromatogr., A 2011, 1218 (49), 8760. (12) Ji, Q. C.; Gage, E. M.; Rodila, R.; Chang, M. S.; El-Shourbagy, T. A. Rapid Comm. Mass Spec. 2003, 17, 794. (13) Mao, Y.; Moore, R. J.; Wagnon, K. B.; Pierce, J. T.; Debban, K. H.; Smith, C. S.; Dill, J. A.; Fuciarelli, A. F. Chem. Res. Toxicol. 1998, 11, 954. (14) Gordon, E. F.; Mansoori, B. A.; Carroll, C. F.; Muddiman, D. C. J. Mass Spectrom. 1999, 34, 1055. (15) Ruan, Q.; Ji, Q. C.; Arnold, M. E.; Humphreys, W. G.; Zhu, M. Anal. Chem. 2011, 83, 8937. (16) Roth, M. J.; Plymire, D. A.; Chang, A. N.; Kim, J.; Maresh, E. M.; Larson, S. E.; Patrie, S. M. Anal. Chem. 2011, 83, 9586. (17) United States Food and Drug Administration. Guidance for Industry: Bioanalytical Method Validation; U.S. Department of Health and Human Services, Center for Drug Evaluation and Research (CDER), Center for Veterinary Medicine (CVM), May 2000; http:// www.fda.gov/cder/guidance/index.htm. (18) Silva, J. C.; Gorenstein, M. V.; Li, G.-Z.; Vissers, J. P. C.; Geromanos, S. J. Mol. Cell. Proteomics 2006, 5, 144.

■

CONCLUSIONS Herein we have demonstrated the suitability of intact MS methods for the quantitative analysis of different insulin variants. The individual amounts of each mass-different insulin variant could be determined using intact MS. Results for somatotropin demonstrated that while ultrahigh-resolution mass spectrometers are capable of quantitative analysis, lower resolution instruments can also be used. Here we have developed an understanding of how both internal and external calibration can be used in order to build a foundation for making relative quantitative comparisons between protein species. While the manual approach for data analysis taken here is laborious, it does not preclude the use of vendor deconvolution or isotopic matching software to aid in automation of the quantitative analysis. However, additional understanding of the fidelity of the data intensity output to the raw data input is needed before further automation could be considered. By identifying and then working in the linear dynamic range of the available mass spectrometer, the precision and accuracy of a change from the expected composition of a protein therapeutic could be measured in order to identify deviations or differences between drug products that might be indicative of a safety or purity issue of the drug product. As these analyses are intended for the analysis of drug product, the concentration of the drug product samples can easily be adjusted to fit within the linear range of the mass spectrometer. An additional benefit of using mass spectrometry quantitatively is the ability to also simultaneously obtain a quantitative chemical fingerprint of the drug product. The applicability of similar methods for larger proteins and alternative types of post-translational modifications will continue to be explored.

■

Article

ASSOCIATED CONTENT

* Supporting Information S

Protein sequence information and a table of hGH charge state and isotope precision and linearity data. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Phone: (301) 796-0113. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS Disclaimer: the findings and conclusions of this article have not been formally disseminated by the Food and Drug Administration and should not be construed to represent any agency determination or policy. This project was supported in part by an appointment to the Research Participation Program at the Center for Drug Evaluation and Research administered by the Oak Ridge Institute for Science and Engineering and the U.S. Food and Drug Administration for A.C.G. This work was funded by the FDA Center for Drug Evaluation and Research Critical Path Program and the Regulatory Science and Review Enhancement (RSR) Program. 8051

dx.doi.org/10.1021/ac301949j | Anal. Chem. 2012, 84, 8045−8051