Anal. Chem. 2008, 80, 1290-1296
Absolute Quantification of Monoclonal Antibodies in Biofluids by Liquid Chromatography-Tandem Mass Spectrometry Charlotte Hagman,† Darrell Ricke,‡ Stefan Ewert,§ Stephan Bek,| Rocco Falchetto,† and Francis Bitsch*,†
BioAnalytical Sciences, Discovery Technologies, Novartis Institutes for BioMedical Research, Novartis, Basel, CH-4002 Basel, Switzerland, Bioinformatics, BioMedical Informatics, Novartis Institutes for BioMedicalResearch, Inc., Cambridge, Massachusetts 02139, Biomolecules Production, Antibody Center, Novartis Institutes for BioMedical Research, Novartis, Basel, CH-4002 Basel, Switzerland, and Soluble Biomarkers, Novartis Institutes for BioMedical Research, Novartis, Basel, CH-4002 Basel
The development of a quantification method for monoclonal antibodies in serum has been accomplished by high-performance liquid chromatography multiple reactions monitoring mass spectrometry. A human monoclonal antibody (HmAb) was used as the model protein for method development and validation. A peptide from the CDR3-region of its heavy chain was selected and used for quantifying the entire mAb. This signature peptide served as a template for the internal standard. Prior to mass spectrometric analysis approximately 50% of the total serum protein content was removed by albumin depletion. The accuracy of the method ranged between 99 and 112% in cynomolgus monkey serum. The intraassay coefficient of variation (CV) was lower than 4% at 4 µg/mL and 200 µg/mL HmAb (n ) 3). The CV at 400 µg/mL corresponded to 9% (n ) 3). In addition, the interassay variation was investigated in a male cynomolgus serum pool and in a female cynomolgus serum pool. The CV for the male cynomolgus pool at 4 µg/mL HmAb was 7% (n ) 3). The CV obtained from the female pool was 8% (n ) 3), at 4 µg/mL. The dynamic range of the method was 3 orders of magnitude. After albumin depletion of 25 µL of serum, a lowest limit of quantification of 2 µg/mL HmAb was reached in both human and cynomolgus monkey samples. Compared to small molecule drugs, therapeutic monoclonal antibodies (mAbs) offer numerous advantages. Their primary benefit is the high target binding specificity.1 The possibility to be able to engineer and optimize features such as affinity, avidity, biodistribution, effector functions, and the half-life also makes mAbs highly attractive as therapeutic agents.2,3 In order to develop * To whom correspondence should be addressed. E-mail: francis.bitsch@ novartis.com. Phone: 41 (0) 61 324 2056. Fax: 41 (0) 61 324 2218. † BioAnalytical Sciences, Discovery Technologies. ‡ Bioinformatics, BioMedical Informatics. § Biomolecules Production, Antibody Center. | Soluble Biomarkers. (1) Jones, P. T.; Dear, P. H. Nature 1986, 321, 522-525. (2) Weiner, L. M. J. Immunother. 2006, 29, 1-9. (3) Chowdhury, P. S.; Wu, H. Methods 2005, 36, 11-24.
1290 Analytical Chemistry, Vol. 80, No. 4, February 15, 2008
effective dosing strategies, the pharmacokinetics of a drug, i.e., its time-dependent concentration and distribution in biofluids and tissues, has to be understood. Protein and peptide quantification is most often performed using immunoassays such as enzymelinked immunosorbent assays (ELISA).4 This approach features high sensitivity (nano- to picomolar concentration range), but also requires a specific antibody for every new protein therapeutic agent. The antibody production and selection process is itself resource- and time-consuming, contributing to long assay development times, which may take up to 5 months for a typical validated ELISA assay used for toxicological or clinical studies. Furthermore, despite the antibody’s specificity, the presence of endogenous analytes can still severely affect the selectivity of the assay, yielding unreliable results. For example, endogenous antibodies directed against the therapeutic agent could interfere with the ELISA assay. Even humanized mAbs are potentially immunogenic and could stimulate the production of interfering endogenous antibodies.5 Soluble receptors or binding proteins may also compromise the assay.6 Finally, the lack of standardization of immunometric assays makes it difficult to compare results between different laboratories.7 In search for an alternative, the attention has turned to mass spectrometry (MS) as a tool for absolute protein and peptide quantification that is faster, easier to standardize, and more specific. MS is already widely used as the quantification tool of choice in conventional bioanalytical investigations of low-molecular-weight drug candidates. The vast experience accumulated over the years in this field can be exploited for the method’s translation to the quantitative assessment of biological molecules. Small molecules are quantified using stable isotopic-labeled molecules as internal standards, and the same principle can be used for biotherapeutic agents. The concept of absolute protein and peptide mass spectrometric quantification with stable isotopes (4) Lobo, E. D.; Hansen, R. J.; Balthasar, J. P. J. Pharm. Sci. 2004, 93, 26452668. (5) Bjerner, J. Scand. J. Clin. Lab. Invest. 2005, 65, 349-64. (6) Wadhwa, M.; Bird, C.; Dilger, P.; Gaines-Das, R.; Thorpe, R. J. Immunol. Methods 2003, 278, 1-17. (7) Fierens, C.; Sto ¨ckl, D.; Baetens, D.; De Leenheer, A. P.; Thienpont, L. M. J. Chromatogr., B: Anal. Technol. Biomed. Life Sci. 2003, 792 (2), 249-59. 10.1021/ac702115b CCC: $40.75
© 2008 American Chemical Society Published on Web 01/25/2008
was first demonstrated in biological tissue.8 Endogenous leucine enkephalin in thalamus tissue extracts was quantified using opioid peptides as internal standards. The first example describing the application of proteolysis coupled with MS for absolute protein quantification used apolipoprotein A-I as the model protein.9 The basic idea was to use a peptide proteolytically generated from the intact apolipoprotein as a representation of the protein’s concentration. Further similar efforts included work on the quantification of the G-protein-coupled receptor Rhodopsin from trypsinized membrane preparations10 and the absolute quantification of yeast proteins involved in gene silencing.11 The quantification of selected proteins in serum represents a greater challenge since the dynamic range of protein and peptide concentrations in this matrix can vary by more than 10 orders of magnitude.12 The first example for the direct, absolute quantification of a protein in serum was reported by Barnidge et al.,13 who used a prostate-specific antigen biomarker as the model protein. A limit of detection of 4.5 µg/ mL could be achieved. Anderson et al. used another strategy combining stable isotope standards and capture by antipeptide antibodies (SISCAPA approach), further improving the limit of detection.14 The drawback of this approach is that, similarly to ELISA, specific capture antibodies need to be developed for each protein under investigation. Another more generally applicable approach involves sample pretreatment by depletion of the most abundant serum proteins, such as albumin and immunoglobulins. This approach has been shown to improve MS detection and identification of lowerabundant peptides and proteins.15-17 Depletion kits using chemicalaffinity or immuno-affinity for removal of both albumin and immunoglobulins typically reduce protein content by more than 70%, while with depletion kits specific only for human serum albumin (HSA) a protein reduction of about 50% can be expected.12,18,19 HSA-specific depletion kits are either based on albumin affinity for the Cibachrome blue dye, on specific antiHSA antibodies or on immobilized peptides with affinity for albumin.20,21 These kits are ideally suited to mAb quantification in serum or plasma since the IgGs are retained after albumin depletion. In this report, we describe the development of a method for quantifying monoclonal antibodies in human serum and its validation in cynomolgus monkey serum. The method is based (8) Desiderio, D. M.; Kai, M. Biomed. Mass Spectrom. 1983, 10, 471-9. (9) Barr, J. R. Maggio, V. L.; Patterson, D. G., Jr.; Cooper, G. R.; Henderson, L. O.; Turner, W. E.; Smith, S. J.; Hannon, W. H.; Needham, L. L.; Sampson, E. J. Clin. Chem. 1996, 42, 1676-1682. (10) Barnidge, D. R.; Dratz, E. A.; Martin, T.; Bonilla, L. E.; Moran, L. B.; Lindall, A. Anal. Chem. 2003, 75, 445-51. (11) Gerber, S. A.; Rush, J.; Stemman, O.; Kirschner, M. W.; Gygi, S. P. Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 6940-5. (12) Anderson, N. L.; Anderson, N. G. Mol. Cell Proteomics 2002, 1, 845-67. (13) Barnidge, D. R.; Goodmanson, M. K.; Klee, G. G.; Muddiman, D. C. J. Proteome Res. 2004, 3, 644-652. (14) Anderson, N. L.; Anderson, N. G.; Haines, L. R.; Hardie, D. B.; Olafson, R. W.; Pearson, T. W. J. Proteome Res. 2004, 3, 235-44. (15) Anderson, L.; Hunter, C. L. Mol. Cell Proteomics 2006, 5, 573-88. (16) Taylor, P. J. Clin. Biochem. 2005, 38, 328-34. (17) Bjorkhall, K.; Miliotis, T.; Davidsson, P. Proteomics 2005, 5, 307-17. (18) Wang, Y. Y.; Cheng, P.; Chan, D. W. Proteomics 2003, 3, 243-248. (19) Greenough, C.; Jenkins, R. E.; Kitteringham, N. R.; Pirmohamed, M.; Park, B. K.; Pennington, S. R. Proteomics 2004, 4, 3107-3111. (20) Steel, L. F.; Trotter, M. G.; Nakajima, P. B.; Mattu, T.; Gonye, G. Block T. Mol. Cell Proteomics 2003, 2, 262-270. (21) Travis, J.; Bowen, J.; Tewksbury, D.; Johnson, D.; Pannell, R. Biochem. J. 1976, 157, 301-306.
on a combination of albumin depletion of serum and LC-MS/ MS of unique peptides produced by trypsin digestion. A human monoclonal antibody (HmAb) with high sequence homology to human IgG was used as the model protein for method development. Only sequences that are unique for the mAb can be utilized for its quantification. For this purpose, a sequence similarity search was performed against GenBank for the two different proteomes in this study.22 The selected signature peptide was used for the quantification of the entire mAb and served as a template for the internal standard. To reduce sample complexity and enhance the sensitivity of the method, the serum was depleted of albumin prior to analysis. Five commercially available albumin depletion kits were evaluated in human serum with regard to albumin depletion efficiency and specificity. MATERIALS AND METHODS Serum. The human serum pool was obtained from Rockland Immunochemicals (Gilbertsville, PA). The spiked human serum was centrifuged for 5 min at 3000g to separate out lipids (Centrifuge 5415R, Vaudaux-Eppendorf AG, Basel, Switzerland). The cynomolgus serum was obtained from an in-house source of animals (Novartis Pharma, Basel, Switzerland). The cynomolgus serum was centrifuged at 100 000g for 4 h to remove cellular membrane debris and other unsoluble particles (Evolution RC Sorvall, Thermo Electron Corporation, Waltham, MA). Depletion of Albumin by Immuno-Affinity Depletion Kits for Evaluation on Sodium Dodecyl Sulfate Polyacrylamide Gel Electrophoresis (SDS-PAGE). The albumin depletion kits were used according to the manufacturer’s procedure. A volume of 20-35 µL serum was typically depleted of albumin using the Vivapure anti-HSA kit (Vivascience AG, Hannover, Germany) and the ProteoExtract kit (Calbiochem, Merck KGaA, Darmstadt, Germany). Depletion of Albumin by Dye-Based Kits for Evaluation on SDS-PAGE. The recommended serum volume was 20 µL for the Montage albumin depletion kit from Millipore (Millipore Corporation, Bedford, MA). The Aurum Affi-gel from BIORAD (Biorad, Hercules, CA) can effectively deplete 125 µL of serum. For the Enchant albumin depletion kit from PALL (PALL Corporation, Ann Arbor, MI), a volume of 30 µL was recommended. SDS Polyacrylamide Gel Electrophoresis. The evaluation of the albumin depletion kits was performed by SDS-PAGE. The SDS-PAGE was performed on a XCell II (Novex, San Diego, CA) using an Tris-Gly 4-20% gel (ANAMED, Darmstadt, Germany) at 200 V for 1 h. The samples were reduced before SDS-PAGE analysis at 90 °C for 6 min to ensure complete unfolding. In each lane an equivalent of 7 µg of serum protein was loaded. Both the albumin-depleted serum and the flow-through fraction were separated by SDS-PAGE. Bioinformatics Tool for Identification of Unique Peptides in the mAb Sequence. To identify unique residues within the antibody heavy and light chains, a bioinformatics tool identified all related antibody sequences from human GenBank proteins. Antibody sequences were aligned, and the percent identity for each antibody residue position against the public antibody sequences was calculated. The tool provides an in silico (22) Benson, D. A.; Karsch-Mizrachi, I.; Lipman, D. J.; Ostell, J.; Wheeler, D. L. Nucleic Acids Res. 2005, 33, D34-38.
Analytical Chemistry, Vol. 80, No. 4, February 15, 2008
1291
tryptic digest, and for the identification of candidate signature peptides residues with low identity scores were singled out. Sample Preparation. Spiked Serum Samples. The total protein content in human serum was determined with a Life Science UVvis DU530 spectrophotometer (Beckman Coulter, Fullerton, CA) to 68.35 mg/mL using a bicinchoninic assay (Sigma-Aldrich Chemie, Steinheim, Germany) with BSA (Sigma-Aldrich Chemie, Steinheim, Germany) as a protein standard. The copper(II) sulfate solution was obtained from Sigma-Aldrich Chemie (Steinheim, Germany). Serum was spiked with HmAb at various concentrations between 1 and 1000 µg/mL. The stable isotope-labeled internal standard peptide was added to the serum at a final concentration of 0.44 µg/mL. The reference peptide, corresponding to a peptide (21 amino acids; Mw 2206 Da) from the CDR3 region of the heavy chain and containing a deuterated(D4) alanine residue, was obtained from NeoMPS (Strasbourg, France). Albumin Depletion. Human serum was spiked with HmAb and the internal standard prior to albumin depletion. A volume of 50 µL of spiked serum samples was depleted according to the protocol for the ProteoExtract albumin removal kit. The albumindepleted serum was lyophilized overnight on a Speedvac SPD 111V evaporator (Thermo Savant, Du¨sseldorf, Germany). Reduction and Alkylation. To each lyophilized pellet of albumindepleted serum, 50 µL of 8 M urea (Fluka, BioChemika, Buchs, Switzerland) and 0.4 M NH4HCO3 (pH 8.0) (Sigma, St. Louis, MO) were added. The pellets were vortexed until they completely dissolved; subsequently, the samples were centrifuged. To each sample a 100-fold molar excess of dithiolthreitol DTT (Calbiochem, La Jolla, CA) was added. The samples were then incubated at 50 °C for 30 min. After cooling the samples at room temperature, 5.8 µL of 1 M iodoacetamide (SigmaUltra, St. Louis, MO) was added to each sample. The samples were incubated in the dark for 15 min at room temperature. Digestion. Portions of 55 µL of 1 M Tris buffer (Fluka, Buchs, Switzerland) and 495 µL of H2O (Riedel-de Haen, Seelze, Germany) were added to 55 µL of the reduced and alkylated serum samples, which contained 1.7 mg of serum protein. Trypsin X-IS (Sigma, St. Louis, MO) was dissolved in 0.01% trifluoroacetic acid (TFA) (Pierce, Rockford, IL) to a final concentration of 1 µg/µL. A volume of 82 µL of the 1 µg/µL trypsin solution was added to the reduced and alkylated samples. The final protein/trypsin ratio was estimated to 20:1. After overnight incubation at 37 °C the digestion was quenched by adding TFA to a final concentration of 0.1%. All samples were evaporated to dryness overnight (Speedvac SPD 111V evaporator from Thermo Savant, Du¨sseldorf, Germany) and stored at -80 °C. High-Performance Liquid Chromatography-Tandem MS (HPLC-MS/MS). Mass Spectrometric Sample Preparation. A volume of 40 µL of H2O/AcN/formic acid (85:15:0.1%) (Riedel-de Haen, Seelze, Germany) was added to each sample. The samples were vortex until the pellet was dissolved and were then sonicated for 5 min. Finally, the samples were diluted with H2O/AcN/formic acid (85:15:0.1%) to a volume of 70 µL, which corresponds to a peptide concentration of approximately 23 µg/µL. HPLC-Multiple Reaction Monitoring (MRM) MS. The mass spectrometric analysis was performed with a 4000 QTRAP instrument (Applied Biosystems, Foster City, CA).23 Its operating software, Analyst 1.4.2, has an automated quantification optimiza1292
Analytical Chemistry, Vol. 80, No. 4, February 15, 2008
tion tool. This was used for generating the MRM method. The triply charged ion of the signature peptide (CDR3-HC; [M + 3H] 3+ ) 734.7 m/z) was selected as the precursor ion. Five MRM transitions were used for detection of the peptide; 734.7 f 493.3(y5+), 734.7 f 580.3(y6+), 734.7 f 679.4(y7+), 734.7 f 780.4(y8+), and 734.7 f 879.5(y9+). For all five transitions, the entrance potential was set to 10 V and the declustering potential to 60 V. The ion source parameters are flow-dependent parameters and were optimized at a flow rate of 200 µL/min. A 2.1 × 150 mm C-8 reverse phase column with 3.5 µm particle size was used for the separation (SymmetryShield, Waters, Milford, MA). The flow rate was 200 µL/min. Each injection contained approximately 850 µg of digested serum protein, which corresponds to 25 µL of albumin-depleted serum. Eluent A was H2O containing 0.1% formic acid, and eluent B was acetonitrile containing 0.1% formic acid (Riedel-de Haen, Seelze, Germany). The gradient started with 85% eluent A and 15% eluent B. After 10 min of initial conditions, the organic content increased to 29% in 20 min. To reduce ion source contamination, a switch valve was coupled to the Agilent 1100 HPLC system (Agilent Technologies, Waldbronn, Germany), and eluates were sent to waste during the first portion of each HPLC run. Just before elution of the signature peptide and internal standard, the valve was switched to position “open”, directing the eluent into the ion source of the mass spectrometer. After the elution of the signature peptide and the internal standard the valve was switched to position “close”, directing the eluent back to waste. To prevent source overheating, a separate pump was also connected to the switch valve, delivering a flow of 200 µL/min (70:30:0.1% H2O/AcN/formic acid) into the ion source when the valve was in the closed position. RESULTS AND DISCUSSION Comparison of Albumin Depletion Kits. It is estimated that total serum protein is comprised of more than 50% albumin.12 Its removal from serum would therefore offer obvious advantages for sensitive protein quantification.15-18 Two types of disposable albumin depletion kits were evaluated. The first is based on the Cibachrome blue dye and the second on affinity selection.19-21 The Cibachrome Blue-based kits were the Montage albumin depletion kit from Millipore, the Aurum Affi-Gel Blue mini kits from BIO-RAD, and the Enchant albumin depletion kit from PALL Life Sciences. The non-dye-based affinity selection kits were the Vivapure Anti-HSA depletion kit from VivaScience and the ProteoExtract albumin removal kit from Calbiochem. The Anti-HSA depletion kit from VivaScience should have no cross-reactivity with other human proteins. The high affinity for albumin is achieved by coupling a unique antibody fragment to a low binding crosslinked agarose gel. The ProteoExtract Albumin Removal kit from Calbiochem is based on a new (non-dye) affinity resin. These kits were evaluated with regards to their albumin depletion efficiency, specificity, and reproducibility. Serum albumin depletion was assessed by SDS-PAGE gel electrophoresis (Figure 1a). Undepleted serum was separated in lane 2, and a thick band of albumin is observed. A thick band corresponding to HSA (66 kDa) was observed with the kit from Millipore, showing that a significant amount of albumin still (23) Hager, J. W.; Yves Le Blanc, J. C. Rapid Commun. Mass Spectrom. 2003, 17, 1056-64.
Figure 1. (a) SDS-PAGE of albumin-depleted human serum. In lane 1, mixtures of reference proteins were separated. Crude serum was separated in lane 2. In lane 3, albumin-depleted serum using the Montage albumin depletion kit was separated. Serum depleted with the Vivapure Anti-HSA depletion kit was separated in lane 4. In lane 5, serum depleted using the ProteoExtract albumin removal kit was separated. Serum depleted with the Aurum Affi-Gel Blue mini kits was separated in lane 6. Finally, serum depleted using the Enchant albumin depletion kit was separated in lane 7. (b) SDS-PAGE of eluted proteins from albumin depletion column. In lane 1, mixtures of reference proteins were separated. Crude serum was separated in lane 2. In lane 3, the proteins with affinity for the Montage albumin depletion kit column were separated. Proteins eluting from the Vivapure Anti-HSA depletion kit were separated in lane 4. In lane 5, the proteins with affinity for the ProteoExtract albumin removal kit columns were separated. Proteins eluting from the Aurum Affi-Gel Blue mini kits columns were separated in lane 6. In lane 7, proteins with affinity for the Enchant albumin depletion kit columns were separated.
Figure 2. MS/MS of signature peptide. Five MRM transitions were used for detection of the signature peptide. [M + 3H]3+ at m/z 734.7 f 493.3(y5+), 734.7 f 580.3(y6+), 734.7 f 679.4(y7+), 734.7 f 780.4(y8+), and 734.7 f 879.5(y9+).
remained in serum after HSA depletion (lane 3). On the other hand, the two other Cibachrome blue-based kits, the Aurum AffiGel Blue mini kits and the Enchant albumin depletion kit, removed albumin very efficiently (lanes 6 and 7). The affinity-based kits from VivaScience and Calbiochem removed the greater part of albumin from the serum (lanes 4 and 5). The affinity depletion kit from Calbiochem was the most efficient at depleting serum of albumin compared to the VivaScience kit. The specificity of the albumin depletion kits was explored by separating the proteins that eluted from the albumin depletion columns (Figure 1b). The eluate from the VivaScience albumin depletion columns was separated on lane 4. The only prominent band corresponded to human serum albumin. The ProteoExtract albumin removal kit from Calbiochem also
showed very high specificity, as the only prominent band in lane 5 also corresponded to albumin (Figure 1b). The proteins that were eluted from the Cibachrome blue-based depletion columns were separated in lanes 3, 6, and 7. The presence of bands at 49 and 23 kDa, corresponding to the heavy and light chains of IgGs from serum, respectively, indicated a lower specificity of the depletion kits. This was most prominent for the Aurum Affi-Gel Blue mini kits from BIO-RAD and the Enchant albumin depletion kit from PALL Life Sciences. These findings were consistent with other studies where dye-based affinity kits have shown low specificity, whereas kits based on non-dye affinity selection exhibited higher specificy.17,24 The single-use columns also prevented potential carryover. Analytical Chemistry, Vol. 80, No. 4, February 15, 2008
1293
Figure 3. Transition of the triply charged signature peptide to its y8+ ion monitored as a function of time. (a) LC-MS/MS chromatogram of the blank human serum sample. (b) LC-MS/MS chromatogram of the 2 µg/mL HmAb human serum sample. Table 1. Validation in Cynomolgus Serum, with the Intra-assay Variability, n ) 3 concentration (µg/mL)
mean (µg/mL)
accuracy (%)
standard deviation
CV (%)
4 200 400
4.05 198.82 448.92
101 99 112
(0.16 (6.62 (40.34
(3.95 (3.33 (8.97
In conclusion, the albumin depletion kit that showed best depletion ability and specificity was the ProteoExtract albumin depletion kit from Calbiochem. The results shown in Figure 1a and 1b are representative for three independent experiments. By spectrophotometric analysis it was estimated that after albumin depletion of 50 µL of serum, 49% of the total protein content was removed. As a result, each albumin-depleted sample contained approximately 1.7 mg of serum proteins. Identifying Signature Peptides. In silico analysis of the mAb sequence allowed us to identify unique tryptic signature peptides (24) Ramstro ¨m, M.; Hagman, C.; Mitchell, J. K.; Derrick, P. J.; Hakansson, P.; Bergquist, J. J. Proteome Res. 2005, 4 (2), 410-6
1294
Analytical Chemistry, Vol. 80, No. 4, February 15, 2008
that could be potentially used for the quantification of the monoclonal antibody. Multiple sequence alignment analysis determined identity counts for all amino acid residues with related GenBank antibody sequences. Residues that have a low identity score were regarded as unique. The [20-38], [44-65], and [102122] peptides from the heavy chain of the mAb contained several unique residues, reflecting the variable-region heavy chain. Finally, these peptide sequences were searched against UniProt/Swissprot to exclude peptides present in the plasma proteome. A tryptic digest of the monoclonal antibody was separated on a reversed phase column, and tandem MS was performed on the [20-38], [44-65], and [102-122] peptides. Since the [102-122] peptide dissociated most efficiently, this peptide was selected for all subsequent MRM-MS experiments. The transitions for the triply charged [102-122] peptide (m/z 734.7) to five of its singly charged y ions were used for the identification of the signature peptide. The quantification of the monoclonal antibody was achieved by monitoring the transition of the triply charged peptide at m/z 734.7 to the y8+ ion at 780.4 (Figure 2). By monitoring the fragment ion that has a greater m/z ratio than the parent ion, the noise is
Figure 4. Transition of the triply charged signature peptide to its y8+ ion is monitored as a function of time. (a) LC-MS/MS chromatogram of the blank cynomolgus serum sample. (b) LC-MS/MS chromatogram of the 2 µg/mL HmAb cynomolgus serum sample. Table 2. Validation in Cynomolgus Serum, with the Interassay Variability, n ) 3 QC pools
concentration (µg/mL)
mean (µg/mL)
accuracy (%)
standard deviation
coefficient of variation (%)
cynomolgus male cynomolgus female
4 4
4.43 4.13
111 103
(0.32 (0.33
(7.24 (8.00
reduced and the sensitivity is improved.25 The [102-122] peptide was used as a template for the internal standard. If the in silico search returns multiple peptides, some additional selection criteria should be considered. For example, peptides that contain KK or RR sequences or have residues that are prone to chemical modification, such as Met, Trp, or Cys, should be avoided.26 Unstable sequences, such as Asp-Gly, and N-terminal Gln or Asn could also be problematic. Finally, the most relevant positive selection criterion concerns the peptide’s ionization efficiency and its MS/MS fragmentation properties. The fragmentation path of a peptide depends on its amino acid sequence. The presence of basic amino acids at or close to either end of the peptide, such as Arg, Lys, His, or Pro, induces the formation of
fragment ions containing both the N- and the C terminus (b and y ions, respectively). The N-terminal end of the Pro residue is also a dominant MS/MS cleavage site. If this residue is present in a peptide, extensive and predicable fragmentation paths are expected.27 LC-MS/MS in Human Serum. The selectivity of the method was good since no response above background signals of the selected transitions was detected in the blank serum sample (Figure 3a). The signature peptide and the internal standard coeluted at 26 min. The linear range and lowest limit of quantification (LLOQ) of the monoclonal antibody were determined using pooled human serum as the matrix. The concentration range of the standard curve spanned 3 orders of magnitude (2, 10, 100,
(25) Kirkpatrick, D. S.; Gerber, S. A.; Gygi, S. P. Methods 2005, 35, 265-273. (26) Bro ¨nstrup, M. Expert Rev. Proteomics 2004, 1, 503-512.
(27) de Hoffman, E.; Charette, J.; Stroobant, V. Mass SpectrometrysPrinciples and Applications, 2nd ed.; John Wiley & Sons, Inc.: New York, 2001.
Analytical Chemistry, Vol. 80, No. 4, February 15, 2008
1295
and 1000 µg/mL). All calibration points were analyzed in triplicates. The numerical analysis was performed using a linear area calibration (y ) ax + b) with a weighting factor of 1/x. The method was linear from 2 to 1000 µg/mL with R2 g 0.998. The accuracy at 2 and 1000 µg/mL was determined to 105.8 and 99.6%, respectively. The precision at 2 µg/mL HmAb was (5.4%. The lowest concentration in the calibration curve was 2 µg/mL. This corresponded to approximately 333 fmol of the selected peptide on column. According to the FDA validation guidelines, the LLOQ value requires a signal-to-noise (S/N) response at least 5 times higher compared with that of the blank sample and an accuracy and precision within (20% (Figure 3b).28 Following these criteria, in our study a LLOQ of 2 µg/mL for HmAb in human serum was achieved. Method Validation in Cynomolgus Serum. Serum from cynomolgus monkeys, which are frequently used for preclinical toxicity assessments, was selected for a limited method validation. The matrix was a pool of male and female cynomolgus monkey serum. The standard curve was comprised of six different HmAb concentrations (2, 4, 100, 300, 800, and 1000 µg/mL). To assess assay specificity, blank cynomolgus serum samples were subjected to the HPLC-MS/MS method. No response above background was detected, therefore excluding the presence of any interfering endogenous substances (Figure 4a). The accuracy for the 2 and 1000 µg/mL samples was 82.5 and 93.5%, respectively. The coefficient of variation (CV) at 2 µg/mL was 12.2 and 2.3% for the 1000 µg/mL calibration point. The S/N response for the transition of the triply charged peptide to the y8+ ion was greater than 5 (Figure 4b). Therefore, the LLOQ for this method is set at 2 µg/ mL HmAb. At the highest calibration point the detector started to reach the saturation level. Nevertheless, the same calibration model was used as for the human serum samples. All calibration points were analyzed in triplicate. The calibration equation was linear with a R2 value of 0.98. The achieved LLOQ for the HPLC-MS/MS method is higher than that for a standard ELISA-based method, which typically yields values of 0.1 µg/mL and lower. However, the HPLC-MS/ MS method has several additional advantages. First, the MS-based method is insensitive to issues related to the interference from endogenous binding proteins or antibodies specific for the biotherapeutic agent, which may have deleterious effects on an (28) Guidance for Industry. Bioanalytical Method Validation; U.S. Department of Health and Human Services, Food and Drug Administration, Centers for Drug Evaluation and Research: Rockville, MD, 2001.
1296
Analytical Chemistry, Vol. 80, No. 4, February 15, 2008
immunometric assay.5,6 Another advantage with the absolute quantification approach is the power of direct interstudy comparison. The variation of the method itself, the intra-assay variation, can be investigated using quality controls spiked with different mAb concentrations into the same matrix that was used for generating the calibration curve. The previously generated calibration equation is then used for calculating the mAb concentrations for the different quality control samples. In this study, the investigation of the intra-assay variability was performed at 4, 200, and 400 µg/mL. The CV was less than 4.0% for the two lowest QC concentrations and not more than 9.0% at 400 µg/mL HmAb (Table 1). The interassay variation was evaluated using two serum pools from male and female cynomolgus monkeys. The accuracy at 4 µg/mL HmAb for the male and female cynomolgus serum pool was determined to 103 and 111%, and the CVs were determined to be (7.2 and (8.0%, respectively (Table 2). CONCLUSIONS In conclusion, this report has shown the potential of HPLCMS/MS as a generic method to assess absolute monoclonal antibody amounts in biofluids and paves the way for further developments aimed at increasing its sensitivity and applicability. Its current sensitivity already allows its application to pharmacokinetic studies during early drug development phases, for which a LLOQ of 3-5 µg/mL is typically required. Further advantages offered by the MS-based method include improved assay accuracy, direct interassay comparison, the absence of unspecific matrix binding effects, and reduced assay development and validation time, which can be an estimated 6-8 weeks as opposed to the 4-5 months required for an ELISA-based assay. ACKNOWLEDGMENT The authors thank Dr. Rene´ Amstutz, Dr. Hermann Gram, and Dr. Reinhold Janocha for supporting this work.
Received for review November 27, 2007. AC702115B
October
15,
2007.
Accepted