Letter pubs.acs.org/ac
Mapping Monoclonal Antibody Structure by 2D Abundance
13
C NMR at Natural
Luke W. Arbogast, Robert G. Brinson, and John P. Marino* Institute for Bioscience and Biotechnology Research, National Institute of Standards and Technology and the University of Maryland, 9600 Gudelsky Dr., Rockville, Maryland 20850, United States S Supporting Information *
ABSTRACT: Monoclonal antibodies (mAbs) represent an important and rapidly growing class of biotherapeutics. Correct folding of a mAb is critical for drug efficacy, while misfolding can impact safety by eliciting unwanted immune or other off-target responses. Robust methods are therefore needed for the precise measurement of mAb structure for drug quality assessment and comparability. To date, the perception in the field has been that NMR could not be applied practically to mAbs due to the size (∼150 kDa) and complexity of these molecules, as well as the insensitivity of the method. The feasibility of applying NMR methods to stable isotope-labeled, protease-cleaved, mAb domains (Fab and Fc) has been demonstrated from both E. coli and Chinese hamster ovaries (CHO) cell expression platforms; however, isotopic labeling is not typically available when analyzing drug products. Here, we address the issue of feasibility of NMR-based mapping of mAb structure by demonstrating for the first time the application of a 2D 13C NMR methyl fingerprint method for structural mapping of an intact mAb at natural isotopic abundance. Further, we show that 2D 13C NMR spectra of proteasecleaved Fc and Fab fragments can provide accurate reporters on the domain structures that can be mapped directly to the intact mAb. Through combined use of rapid acquisition and nonuniform sampling techniques, we show that these Fab and Fc fingerprint spectra can be rapidly acquired in as short as approximately 30 min.
H
igh resolution structural analysis of intact monoclonal antibodies (mAbs) has to date been largely inaccessible by methods such as X-ray crystallography and solution state nuclear magnetic resonance (NMR) spectroscopy. While mAbs can be produced in adequate amounts and purity for crystal growth or NMR, the size (∼150 kDa) and inherent flexibility of these proteins can be problematic for both crystallization and solution NMR methods. Indeed, there are only three structures of intact antibodies in the protein data bank (PDB), all at relatively low resolution (1HZH1 from human; 1IGT2 and 1IGY3 from mouse). These structures show evidence of a variety of hinge conformations that support the notion of mAb structures as dynamic ensembles of conformational states in solution. In contrast, Fab and Fc fragments, derived from the defined protease cleavage of mAbs, as well as complexes formed by these fragments, routinely yield to crystallization and are represented by over a thousand structures in the PDB. On the other hand, application of NMR methods to intact mAbs is a significant challenge due to significant line broadening of signals from short transverse relaxation times that arise from slow molecular tumbling. Perdeuteration of nonlabile 1H sites and use of transverse relaxation optimized spectroscopy (TROSY)4 are commonly employed to overcome such rapid transverse relaxation processes in proteins greater than approximately 50 kDa. Indeed, using TROSY-based methods, assessment of molecular systems up to 670 kDa has been successful.5 This article not subject to U.S. Copyright. Published XXXX by the American Chemical Society
mAbs, as well as Fab and Fc domains, can be heterologously expressed in E. coli culture where perdeuteration can be achieved simply through fermentation in D2O reconstituted media. This approach, however, does not yield mammalian-type post translational glycosylation. Using an E. coli-based system to express an unglycosylated triply labeled 2H,13C,15N human IgGFc fragment, application of conventional NMR methodologies afforded nearly complete resonance assignments and assessment of structural stability in a forced degradation study.6 Using mammalian cell lines, like Chinese hamster ovaries (CHO) cells, appropriately glycosylated mAbs can be produced which can then be subject to specific proteolytic cleavage of the hinge region to yield Fab and Fc fragments. In a seminal study, mAb expression in CHO cells using 2H-, 13C-, or 15N-labeled amino acids afforded a glycosylated and selectively perdeuterated mouse IgG2β-Fc fragment.7 This labeling scheme allowed the collection of conventional triple resonance 3D NMR data and subsequent near-complete resonance assignment of the protein backbone. Similarly, expression of a 13 15 C, N-labeled human IgG1-Fc could be achieved without deuteration. Using this labeled material, assignment of 66% of the resonance of the backbone could be achieved that was Received: December 23, 2014 Accepted: March 1, 2015
A
DOI: 10.1021/ac504804m Anal. Chem. XXXX, XXX, XXX−XXX
Analytical Chemistry
Letter
procedure. Such identification does not imply recommendation or endorsement by the National Institute of Standards and Technology, nor does it imply that the material or equipment identified is necessarily the best available for the purpose. Preparation of mAb Samples. Fc and Fab fragments were prepared from the full length NIST Candidate RM8670 (NISTmAb), an IgG1κ antibody,19 by papain digest using the Pierce Fab Preparation Kit (Thermo Scientific). Briefly, 32 mg of NISTmAb was prepared in the Pierce Fab Digest Buffer at a concentration of 8 mg/mL. The sample was mixed with 37.5 μg of immobilized papain resin and incubated at 37 °C for 4 h. Progress of the reaction was monitored by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), and digestion was quenched by removal of the papain resin by centrifugation. The digested mAb was loaded onto a Protein A affinity column and washed with phosphate-buffered saline (PBS) buffer, with the flow-through collected as the Fab fraction; afterward, the column was eluted with the Pierce Elution Buffer, and the eluent was collected as the Fc fraction. The Fab fraction was concentrated to 500 μM and exchanged into 25 mM L-histidine-d3 (α-d1; imidazole-2,5-d2) using a 30 kDa cutoff centrifugal filter. The Fc fraction was passed through a 100 kDa centrifugal filter to remove any residual undigested/ partially digested mAb and then concentrated to 430 μM and exchanged into 25 mM L-histidine-d3 using a 30 kDa cutoff centrifugal filter. Total yields were approximately 17.2 mg of Fab and 7.2 mg of Fc. The combined Fab/Fc digest mixture was obtained in a similar manner from 16 mg of NISTmAb, except that following quenching of the digest reaction, the crude digest mixture was immediately passed through a 100 kDa centrifugal filter, then concentrated to approximately 500 μM Fab and 250 μM Fc, and exchanged into 25 mM Lhistidine-d3 using a 30 kDa centrifugal filter. The full length NISTmAb sample was prepared from 200 μL of 100 mg/mL stock solution in 25 mM L-histidine, pH 6.0 exchanged into 20 mM bis-tris-d19, pH 6.0, and concentrated to 0.5 mL using a 100 kDa centrifugal filter. NMR Data Acquisition and Processing. All NMR data were recorded on either a 600 or 900 MHz Bruker Avance III spectrometer equipped with triple resonance cryogenically cooled TCI probes with a triple axis gradient system. 13Cmethyl fingerprint data were collected at 45 °C (600 MHz) or 50 °C (900 MHz), while rapid acquisition experiments were collected in duplicate at 37 °C. Unless otherwise noted, 900 (600) MHz 1H−13C correlation data sets were recorded with 64 (128) scans per transient and 116 (72) × 1640 (1344) complex points corresponding to spectral widths of 32 (30) ppm × 14 (14) ppm with acquisition times of 8 (8) ms and 65 (80) ms in the t1 (13C) and t2 (1H) domains, respectively. The 1 H carrier was placed on the water resonance, and the 13C carrier was set to 21 ppm. A recycle delay of 2 s was employed for gradient-selected, sensitivity-enhanced heteronuclear single quantum coherence experiments (gsHSQC)20 and 0.5 s for selective optimized flip angle short transient (SOFAST)heteronuclear multiple quantum coherence (HMQC) experiments. Selective excitation and refocusing of methyl resonances in SOFAST experiments were accomplished using PC-921 and Reburp22 selective pulses, respectively, applied at 1.5 ppm with bandwidths of 5 ppm. NUS schedule generation and data reconstruction were performed as described in the text. Data were apodized with a shifted sine-square bell, zero-filled and doubled by forward−backward linear prediction in both dimensions prior to Fourier Transform to give a final 512 ×
sufficient for monitoring structural changes in the Fc which arose from successive trimming of the N-glycans.8 While acquisition of 2D NMR spectral maps of biomolecules typically requires 13C or 15N isotopic labeling due to the low natural abundance of these nuclei, 1.1% and 0.37%, respectively, the implementation of cryogenic probe technology9 and increases in accessible spectrometer field strengths over the past decade has opened the door to practical acquisition of 2D NMR spectra at natural isotopic abundance of protein therapeutics.10−14 These NMR methods have been successfully applied to smaller protein drug products (
Mapping Monoclonal Antibody Structure by 2D Abundance
13
C NMR at Natural
Luke W. Arbogast, Robert G. Brinson, and John P. Marino* Institute for Bioscience and Biotechnology Research, National Institute of Standards and Technology and the University of Maryland, 9600 Gudelsky Dr., Rockville, Maryland 20850, United States S Supporting Information *
ABSTRACT: Monoclonal antibodies (mAbs) represent an important and rapidly growing class of biotherapeutics. Correct folding of a mAb is critical for drug efficacy, while misfolding can impact safety by eliciting unwanted immune or other off-target responses. Robust methods are therefore needed for the precise measurement of mAb structure for drug quality assessment and comparability. To date, the perception in the field has been that NMR could not be applied practically to mAbs due to the size (∼150 kDa) and complexity of these molecules, as well as the insensitivity of the method. The feasibility of applying NMR methods to stable isotope-labeled, protease-cleaved, mAb domains (Fab and Fc) has been demonstrated from both E. coli and Chinese hamster ovaries (CHO) cell expression platforms; however, isotopic labeling is not typically available when analyzing drug products. Here, we address the issue of feasibility of NMR-based mapping of mAb structure by demonstrating for the first time the application of a 2D 13C NMR methyl fingerprint method for structural mapping of an intact mAb at natural isotopic abundance. Further, we show that 2D 13C NMR spectra of proteasecleaved Fc and Fab fragments can provide accurate reporters on the domain structures that can be mapped directly to the intact mAb. Through combined use of rapid acquisition and nonuniform sampling techniques, we show that these Fab and Fc fingerprint spectra can be rapidly acquired in as short as approximately 30 min.
H
igh resolution structural analysis of intact monoclonal antibodies (mAbs) has to date been largely inaccessible by methods such as X-ray crystallography and solution state nuclear magnetic resonance (NMR) spectroscopy. While mAbs can be produced in adequate amounts and purity for crystal growth or NMR, the size (∼150 kDa) and inherent flexibility of these proteins can be problematic for both crystallization and solution NMR methods. Indeed, there are only three structures of intact antibodies in the protein data bank (PDB), all at relatively low resolution (1HZH1 from human; 1IGT2 and 1IGY3 from mouse). These structures show evidence of a variety of hinge conformations that support the notion of mAb structures as dynamic ensembles of conformational states in solution. In contrast, Fab and Fc fragments, derived from the defined protease cleavage of mAbs, as well as complexes formed by these fragments, routinely yield to crystallization and are represented by over a thousand structures in the PDB. On the other hand, application of NMR methods to intact mAbs is a significant challenge due to significant line broadening of signals from short transverse relaxation times that arise from slow molecular tumbling. Perdeuteration of nonlabile 1H sites and use of transverse relaxation optimized spectroscopy (TROSY)4 are commonly employed to overcome such rapid transverse relaxation processes in proteins greater than approximately 50 kDa. Indeed, using TROSY-based methods, assessment of molecular systems up to 670 kDa has been successful.5 This article not subject to U.S. Copyright. Published XXXX by the American Chemical Society
mAbs, as well as Fab and Fc domains, can be heterologously expressed in E. coli culture where perdeuteration can be achieved simply through fermentation in D2O reconstituted media. This approach, however, does not yield mammalian-type post translational glycosylation. Using an E. coli-based system to express an unglycosylated triply labeled 2H,13C,15N human IgGFc fragment, application of conventional NMR methodologies afforded nearly complete resonance assignments and assessment of structural stability in a forced degradation study.6 Using mammalian cell lines, like Chinese hamster ovaries (CHO) cells, appropriately glycosylated mAbs can be produced which can then be subject to specific proteolytic cleavage of the hinge region to yield Fab and Fc fragments. In a seminal study, mAb expression in CHO cells using 2H-, 13C-, or 15N-labeled amino acids afforded a glycosylated and selectively perdeuterated mouse IgG2β-Fc fragment.7 This labeling scheme allowed the collection of conventional triple resonance 3D NMR data and subsequent near-complete resonance assignment of the protein backbone. Similarly, expression of a 13 15 C, N-labeled human IgG1-Fc could be achieved without deuteration. Using this labeled material, assignment of 66% of the resonance of the backbone could be achieved that was Received: December 23, 2014 Accepted: March 1, 2015
A
DOI: 10.1021/ac504804m Anal. Chem. XXXX, XXX, XXX−XXX
Analytical Chemistry
Letter
procedure. Such identification does not imply recommendation or endorsement by the National Institute of Standards and Technology, nor does it imply that the material or equipment identified is necessarily the best available for the purpose. Preparation of mAb Samples. Fc and Fab fragments were prepared from the full length NIST Candidate RM8670 (NISTmAb), an IgG1κ antibody,19 by papain digest using the Pierce Fab Preparation Kit (Thermo Scientific). Briefly, 32 mg of NISTmAb was prepared in the Pierce Fab Digest Buffer at a concentration of 8 mg/mL. The sample was mixed with 37.5 μg of immobilized papain resin and incubated at 37 °C for 4 h. Progress of the reaction was monitored by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), and digestion was quenched by removal of the papain resin by centrifugation. The digested mAb was loaded onto a Protein A affinity column and washed with phosphate-buffered saline (PBS) buffer, with the flow-through collected as the Fab fraction; afterward, the column was eluted with the Pierce Elution Buffer, and the eluent was collected as the Fc fraction. The Fab fraction was concentrated to 500 μM and exchanged into 25 mM L-histidine-d3 (α-d1; imidazole-2,5-d2) using a 30 kDa cutoff centrifugal filter. The Fc fraction was passed through a 100 kDa centrifugal filter to remove any residual undigested/ partially digested mAb and then concentrated to 430 μM and exchanged into 25 mM L-histidine-d3 using a 30 kDa cutoff centrifugal filter. Total yields were approximately 17.2 mg of Fab and 7.2 mg of Fc. The combined Fab/Fc digest mixture was obtained in a similar manner from 16 mg of NISTmAb, except that following quenching of the digest reaction, the crude digest mixture was immediately passed through a 100 kDa centrifugal filter, then concentrated to approximately 500 μM Fab and 250 μM Fc, and exchanged into 25 mM Lhistidine-d3 using a 30 kDa centrifugal filter. The full length NISTmAb sample was prepared from 200 μL of 100 mg/mL stock solution in 25 mM L-histidine, pH 6.0 exchanged into 20 mM bis-tris-d19, pH 6.0, and concentrated to 0.5 mL using a 100 kDa centrifugal filter. NMR Data Acquisition and Processing. All NMR data were recorded on either a 600 or 900 MHz Bruker Avance III spectrometer equipped with triple resonance cryogenically cooled TCI probes with a triple axis gradient system. 13Cmethyl fingerprint data were collected at 45 °C (600 MHz) or 50 °C (900 MHz), while rapid acquisition experiments were collected in duplicate at 37 °C. Unless otherwise noted, 900 (600) MHz 1H−13C correlation data sets were recorded with 64 (128) scans per transient and 116 (72) × 1640 (1344) complex points corresponding to spectral widths of 32 (30) ppm × 14 (14) ppm with acquisition times of 8 (8) ms and 65 (80) ms in the t1 (13C) and t2 (1H) domains, respectively. The 1 H carrier was placed on the water resonance, and the 13C carrier was set to 21 ppm. A recycle delay of 2 s was employed for gradient-selected, sensitivity-enhanced heteronuclear single quantum coherence experiments (gsHSQC)20 and 0.5 s for selective optimized flip angle short transient (SOFAST)heteronuclear multiple quantum coherence (HMQC) experiments. Selective excitation and refocusing of methyl resonances in SOFAST experiments were accomplished using PC-921 and Reburp22 selective pulses, respectively, applied at 1.5 ppm with bandwidths of 5 ppm. NUS schedule generation and data reconstruction were performed as described in the text. Data were apodized with a shifted sine-square bell, zero-filled and doubled by forward−backward linear prediction in both dimensions prior to Fourier Transform to give a final 512 ×
sufficient for monitoring structural changes in the Fc which arose from successive trimming of the N-glycans.8 While acquisition of 2D NMR spectral maps of biomolecules typically requires 13C or 15N isotopic labeling due to the low natural abundance of these nuclei, 1.1% and 0.37%, respectively, the implementation of cryogenic probe technology9 and increases in accessible spectrometer field strengths over the past decade has opened the door to practical acquisition of 2D NMR spectra at natural isotopic abundance of protein therapeutics.10−14 These NMR methods have been successfully applied to smaller protein drug products (
