- Email: [email protected]
Evicting hitchhiker antigens from purified antibodies
Journal of Chromatography B, 877 (2009) 1543–1552
Contents lists available at ScienceDirect
Journal of Chromatography B journal homepage: www.elsevier.com/locate/chromb
Evicting hitchhiker antigens from purified antibodies Keith A. Luhrs a , Debra A. Harris a , Scott Summers b , Missag H. Parseghian a,∗ a b
Research & Development, Peregrine Pharmaceuticals Inc., 14272 Franklin Avenue, Tustin, CA, USA Process Development, Avid Bioservices Inc., Tustin, CA, USA
a r t i c l e
i n f o
Article history: Received 7 January 2009 Accepted 26 March 2009 Available online 1 April 2009 Keywords: Antibody purification Commercial-scale purification Affinity chromatography Host cell proteins Hitchhiker antigens
a b s t r a c t Antibodies that target common cellular structures may have a propensity to bind those very same antigens as they become exposed in dead or dying cells during production in a bioreactor. Those tendencies can be accentuated if the targeted epitope is highly conserved across species. While attention to contaminants such as endotoxin, viral particles, cellular DNA and even prions has grown coincident with the emergence of the monoclonal antibody industry, it is surprising how little attention has been focused on hitchhiker antigens that may co-elute while bound to the supposedly pure antibody. In this case study, we will focus on anti-histone antibodies and the measures we have taken to eliminate stowaways, such as histone–DNA complexes. These simple measures include the addition of a quartenary amine guard column to the protein A, adjusting the ionic strength of the cell culture supernatant to 400 mM sodium chloride, and establishing a mobile phase gradient from 400 mM to 2 M during protein A chromatography. Initially adjusting the cell culture to 600 mM can compromise the quartenary amine guard column. Also, we demonstrate the applicability of these techniques in both the R&D lab and the manufacturing plant, particularly in improving the apparent potency of antibodies destined for the clinic. Given the prominence of anti-histone antibodies in chromatin immunoprecipitation (ChIP), the implications of hitchhiker antigens interferring with the results of an experiment are far-reaching, indeed, we detect them in some popularly used antibodies. Moreover, a wide variety of monoclonals that may target antigens expressed by the producer cell line may face similar problems, resulting in a decreased production yield, as well as a diminished apparent binding potency. © 2009 Elsevier B.V. All rights reserved.
1. Introduction In the past 25 years, antibody reagents have become an important tool in the research lab and the clinic. In this same period, protein purification technologies have seen great advances in the drive to make antibody production commercially viable [1]. Concomitant with improvements in purification techniques has been the advancement of antibody engineering at the structural level. In the research and pharmaceutical development laboratories, engineering improvements have focused on better production of antibodies or their derivatives in a variety of cellular systems [2,3]. For the clinic, engineering improvements have resulted in better pharmaco-kinetic profiles [4,5]. With the proliferation of antibodies against an ever increasing number of intracellular proteins and structures, one must consider the possibility that all of the engineering improvements are for naught if the binding domains are contaminated with antigens that happen to hitchhike along with the antibody during the production
∗ Corresponding author. Tel.: +1 714 508 6052; fax: +1 714 838 6940. E-mail address: [email protected] (M.H. Parseghian). 1570-0232/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.jchromb.2009.03.042
process. Unlike production at the R&D scale in a T-flask, production in a commercial bioreactor can be characterized by increasing numbers of dead cells and debris as the culture matures. In the latter case, two separate, but related, problems can arise. First is the sink effect. Necrotic or late-stage apoptotic cells have ruptured plasma membranes allowing secreted antibodies accessibility to intra-cellular structures, hence, these structures may begin acting as sinks, absorbing antibody product [6]. Second, during necrosis, some cells may release much of their contents into the media as debris providing an opportunity for antigen–antibody interactions in the supernatant [7]. In order to avoid antibody binding to cellular debris, one might shorten the length of the culture run; however, productivity may be dampened if the culture must be harvested early, say prior to reaching a level of cell viability that is less than 70%. Given the expense of setting up and running a bioreactor, premature harvest may provide a diminishing rate of return. One might suggest bioreactors that use a perfusion process can help minimize such problems. Perfusion allows for the regular exchange of media in a bioreactor such that antibody secreted into the supernatant is carried away from the cells from which they originated. However, perfusion techniques also result in the close confinement of dead cells with viable ones, resulting in antibody loss due to a
1544
K.A. Luhrs et al. / J. Chromatogr. B 877 (2009) 1543–1552
sink effect. Furthermore, a perfusion process does not prevent the accumulation of hitchhikers during an antibody’s sojourn in the bioreactor. Our focus has been the development of therapeutic monoclonal antibodies that target histones and DNA; antigens that are extracellularly exposed in the necrotic core of the tumor [6,8]. Our strategy capitalizes on the exposure of nuclear material in necrotic cells at the center of every solid tumor. In our case, the hitchhikers are chromatin components, particularly, histones, which are highly conserved among species. Here, we discuss measures we have taken to minimize the accumulation of these hitchhikers during the purification process. We evaluated the procedures at an R&D scale using antibody samples polluted with histones and DNA in excess. The location of the tainted materials is tracked throughout the purification process using ELISAs and gel electrophoresis. We then verified the effectiveness of the methodology on process scale cell cultures, demonstrating apparent improvements in antibody binding with the removal of the hitchhiker contaminants. Finally, we use the same ELISA strategy to search for hitchhikers in some commercially available antibodies from other sources. 2. Materials and methods 2.1. Reagents NHS76, a fully human antibody targeting histone H1 and DNA, was developed by Peregrine Pharmaceuticals (Tustin, CA) and Cambridge Antibody Technology (Cambridge, UK). Both, NHS76 and a separate mouse–human chimeric antibody (chTNT-1) targeting the same antigens, were manufactured at Peregrine in mammalian cell cultures incubated in a humidified 5% carbon dioxide atmosphere set at 37 ◦ C. 2.2. Antibody purification Starting material for the purification studies was prepared as a 2 mg/mL cocktail of NHS76 antibody spiked with calf histone H1 and herring sperm DNA to a final concentration of 0.1 mg/mL (both from Roche) in phosphate buffered saline (PBS; 10 mM sodium phosphate, 0.15 M sodium chloride, pH 7.2). Samples were adjusted to 400 mM or 600 mM NaCl with PBS containing 2 M NaCl. Purifications were conducted on a BioLogic DuoFlow chromatographic system (BioRad) using a 5 mL Pharmacia HiTrap Q FF column and a 1 mL Pharmacia HiTrap rProtein A FF column placed in tandem (both, GE Healthcare). Each sample was injected into a 1 mL loop and loaded onto the columns with a flow rate of 2 mL/min for 1 min, in fact, the flow rate was 2 mL/min throughout the procedure. Prior to sample loading, the system was equilibrated to the salt concentration of the adjusted sample for 10 min (20 protein A column volumes (CV)). After sample loading, the columns were further equilibrated at the same salt concentration for 10 min (20 CV). The Q column was then removed from the line and the protein A was subjected to an increase to 2 M NaCl PBS in 1 min (2 CV), rinsed with 2 M NaCl PBS for 5 min (10 CV), before being adjusted to 150 mM NaCl PBS in 1 min (2 CV). After equilibration at 150 mM NaCl PBS for 7 min (14 CV), the antibody was eluted into a fraction collector with 20 mM sodium acetate, 150 mM NaCl, pH 3.2 over 8 min (16 CV). The protein A was then re-equilibrated with 150 mM NaCl PBS for another 7 min. 2.3. Q column elution When taken off-line from the purification process, the quartenary amine column was still equilibrated in the NaCl concentration of the adjusted sample. To elute any bound materials, the column
was reattached to the system (sans any protein A column) and further rinsed with the adjusted salt concentration at 4 mL/min for 5 min (4 Q column volumes). Then it was subjected to an increase to 2 M NaCl PBS over 5 min (4 CV), rinsed with 2 M NaCl PBS for 5 min (4 CV), before being equilibrated at 150 mM NaCl PBS for 5 min (4 CV). Fractions were collected throughout the process. 2.4. Sample analysis calculations To load equal amounts of material from collected fractions onto gels or 96-well plates, an assumption was made of 100% antibody recovery during elution from the protein A. Therefore, 2 mg of NHS76 loaded on a protein A column was eluted in a 5 mL fraction of elutant, bringing the final concentration of the purified antibody to 0.4 mg/mL. For the western blot analyses and ELISAs, taking 2.5% of the total collected fraction from the protein A elutant amounts to 50 g of antibody; and, taking 5% for the DNA gel analysis amounts to 100 g of antibody. Comparable signals obtained in the ELISA analysis of NHS76 between the protein A elutants and the 50 g of pure NHS76 placed in the wells as a control, suggests that the assumption of 100% antibody recovery from the protein A column was correct. Similarly, 50 g of starting material antibody was prepared for western or ELISA analysis and 100 g for DNA analysis. In the starting material cocktail, 50 g of antibody should be accompanied by 2.5 g of H1 and DNA, each; therefore, 2.5 g controls of both antigens, in combination and separately, were also applied to the microtiter plate wells. 2.5. Western blot analysis Samples were first dried down, resuspended in an equal volume of 20 mM NaCl and Tris–glycine SDS sample buffer (2×), and then resolved on a 14% T acrylamide gel in a Tris–glycine running buffer supplied with the Novex® electrophoresis system (Invitrogen). To preserve the intact structure of the antibodies (∼150 kD), non-reduced samples were run on the gel at 200 Vfor 1.5 h. H1 histones are highly basic and run anomalously on poly-acrylamide gels, therefore, 0.5 g of H1S -3 was used as a marker for histone migration. Protein bands were transferred to nitrocellulose (Schleicher and Schuell) at 25 V for 1.5 h in a 12 mM Tris–96 mM glycine–20% methanol buffer using the Xcell SureLock® Mini-Cell and XCell IITM Blot Module Kit (Invitrogen). Detection of NHS76 required goat anti-human Fc conjugated to horseradish peroxidase (1:5000 dilution; Jackson Immunoresearch), which was incubated with the blot for 1.5 h at 25 ◦ C, then the blot was rinsed three times for 5 min (3 × 5 min) with TBST (20 mM Tris–HCl, 150 mM NaCl, 0.1% Tween 20, pH 7.4), once with TBS alone (20 mM Tris–HCl, 150 mM NaCl, pH 7.4) and developed using a 3,3 -diaminobenzidine (DAB) Tablet Set (Sigma) to produce brown bands. Subsequently, H1 detection was done on the same blot by using NHS76 (5 g/mL) as the primary antibody for 1.5 h at 25 ◦ C, washing 3 × 5 min with TBST, then using goat anti-human IgG (H + L chain) conjugated to alkaline phosphatase (1:5000 dilution; Jackson Immunoresearch) as a secondary antibody for 30 min at 25 ◦ C, washing 3 × 5 min with TBST, once with TBS alone and then developing with NBT/BCIP Tablets (Roche) to produce blue bands. 2.6. DNA agarose gel analysis Samples were dialyzed against TE (10 mM Tris, 1 mM EDTA, pH 7.5) buffer, adjusted to 0.2% SDS, and digested with 40 g/mL proteinase K (Sigma) overnight at 55 ◦ C before resolution on a 1% agarose TBE gel (89 mM Tris, 89 mM boric acid, 2 mM EDTA) using a RunOne electrophoresis system (EmbiTec). The gel was stained with ethidium bromide (BioRad) and visualized on an Alpha Innotech image analysis system (FluorChem® ).
K.A. Luhrs et al. / J. Chromatogr. B 877 (2009) 1543–1552
1545
Fig. 1. Protein A purification of an antibody spiked with antigen. (a) Chromatogram of 2 mg NHS76, spiked with 0.1 mg each of histone H1 and DNA, being purified on a 1 mL protein A column (mobile phase: PBS [150 mM NaCl]; elution buffer: 20 mM sodium acetate, 150 mM NaCl, pH 3.2; flow rate: 2 mL/min). 2.5% of the collected fraction from the sample flow through (1), and protein A elutant (2), along with an estimated equivalent amount of 50 g starting material (SM) were resolved by a non-reducing SDS-PAGE and analyzed by western blotting (b). (For starting material estimation, see Section 2.) (c) Schematic presentation of proposed purification for removal of H1–DNA antigens
Chromatin (histone and DNA) contamination. Antibody with H1 hitchhiker. (d) Sample chromatogram generated from the from the antibody preparation. modified purification scheme. The three traces are described in the main text. (e) 2.5% of the collected fraction from the protein A elutant (2), using the modified purification scheme, alongside 50 g starting material were resolved by a non-reducing SDS-PAGE and analyzed by western blotting. (b and e) The blotted NHS76 was detected using goat anti-Human Fc conjugated to horseradish peroxidase and developed with DAB substrate to produce brown bands. H1 was detected using NHS76 as a primary antibody, goat anti-human IgG (H + L chain) conjugated to alkaline phosphatase as a secondary antibody, and then developed with NBT/BCIP substrate to produce blue bands. Pure H1S -3, a member of the H1 family of histones, was used as a marker for H1 migration.
2.7. H1 and NHS76 ELISA analysis
2.8. Purification from Cell Culture
Samples were placed into 96-well microtiter plates (Costar, Cat. # 3361), in triplicate, and allowed to adhere to the wells overnight at 4 ◦ C. Any exposed areas in the wells were coated with blocking buffer (2% non-fat dry milk, NFDM, in TBS). Samples were applied at 2.5% of each collected fraction while controls were applied to the wells based on the calculation rationale described above (see Section 2.4). H1 was detected in the wells with rabbit anti-H1 polyclonals described elsewhere [9,10] at a dilution of 1:500 in blocking buffer. The same plates were used for the detection of NHS76 using goat anti-human conjugated to horseradish peroxidase at 1:40,000 in blocking buffer. Further details are elaborated in Fig. 2.
One liter of cell culture was removed from a manufacturing run for a chimeric antibody targeting H1–DNA (chTNT-1) and placed in a spinner flask that was incubated in a humidified 5% carbon dioxide atmosphere set at 37 ◦ C. A third of this sample was removed when the cells were at 91.5% viability and the remaining twothirds allowed to incubate in the spinner flask until cell viability dropped to 27.9%. Cell harvests were centrifuged to remove cellular debris, the supernatant decanted, filtered with a 0.2 m cellulose acetate filter (Corning) and the antibody purified as described in Section 3. Purification required a Pharmacia 6 cm XK16 rProtein A Sepharose FF column (GE Healthcare). Samples were loaded
1546
K.A. Luhrs et al. / J. Chromatogr. B 877 (2009) 1543–1552
Fig. 2. Locating the fractions with H1 and NHS76. 2.5% of each collected fraction was placed into ELISA wells, including the flow thru fraction during sample loading (yellow arrows and bars), material captured by the Q column (blue arrows and bars), the 2 M NaCl wash (red arrows and bars), 50 g protein A elutant (green arrows and bars), 50 g starting material (dark gray bars), 50 g pure NHS76 (orange bars), 2.5 g H1 complexed with DNA (light gray bars), 2.5 g DNA (olive bars), and 2.5 g H1 (purple bars). All samples were applied to wells in triplicate. (a) Purification of 2 mg NHS76 antibody, spiked with 0.1 mg each of histone H1 and DNA, as described in Fig. 1a and including a 2 M NaCl wash of NHS76 during protein A capture. (b) Purification including Q column. Purification with NaCl adjustment of starting material to (c) 400 mM and (d) 600 mM. An artifact peak (light blue arrows and bar in (d)) also appears in the chromatograms during the gradient drop from 2 M NaCl down to 150 mM NaCl, both, during the purification process (a) or during the offline Q column wash (d). No significant quantities of protein or DNA were found in those fractions (see (d); also see lane “Q Column 2nd Peak”, Fig. 3a). H1 was detected in the wells with rabbit anti-H1 polyclonals at a dilution of 1:500 in blocking buffer. After a 1.5-h incubation at 37 ◦ C, the wells were rinsed 3 × 5 min with TBST, incubated with goat anti-rabbit conjugated to alkaline phosphatase (1:5000 dilution in blocking buffer; Jackson Immunoresearch) for
K.A. Luhrs et al. / J. Chromatogr. B 877 (2009) 1543–1552
at 250 cm/h (8.37 mL/min) and the column was equilibrated and washed at 200 cm/h (6.7 mL/min). The elution of antibody occurred at 100 cm/h (3.4 mL/min) with 20 mM sodium acetate, 150 mM NaCl, pH 3.2 and, once eluted, it was neutralized with sodium hydroxide to pH 6.5. This elutant was then loaded onto a Pharmacia XK26/60 Superdex 200 (prep grade) size exclusion column (GE Healthcare) and fractionated into antibody monomer and aggregate using a PBS running buffer at a flow rate of 20 cm/h (1.77 mL/min). Both, protein A and size exclusion chromatography were conducted on a Pharmacia ÄKTATM FPLC purification system (GE Healthcare). Potency of the monomer fraction was tested as described below. 2.9. Binding potency of a chimeric antibody targeting the H1–DNA antigen The antigenic target for our antibodies was coated on to the wells of microtiter plates with a solution of 10 g/mL calf histone H1 and 50 g/mL herring sperm DNA (both from Roche). Any exposed areas in the wells were coated with blocking buffer (2% non-fat dry milk, NFDM, in TBS). The antibody samples were serially diluted and binding potency was assessed by incubation of the resulting concentrations in the antigen coated wells, in triplicate, for 1.5 h at 37 ◦ C. Excess antibody was washed out with TBST as described for the western analysis, and the presence of the chimeric antibody was detected with goat anti-human IgG (H + L) conjugated to alkaline phosphatase (1:8000 dilution; Jackson Immunoresearch). After a 30-min incubation at 37 ◦ C, excess secondary antibody was washed out as described for the western analysis and the ELISA developed using the SIGMA FAST p-Nitrophenyl Phosphate Tablet Set (pNPP; Sigma). The signal generated at each concentration was plotted as a sigmoidal dose–response curve and the linear portion of the curve from each sample compared to the same generated with a reference standard. An analysis of variance was used to determine the relative potency of the samples to the reference standard and this value was presented as a percentage of potency. 2.10. Screening for hitchhikers Samples of commercially available antibodies were applied to a 96-well microtiter plate, in triplicate, and allowed to adhere to the wells overnight at 4 ◦ C. Any exposed areas in the wells were coated with blocking buffer (2% NFDM in TBS). The wells were incubated as described in Fig. 5 first with streptavidin alone, to demonstrate lack of cross-reactivity with the commercial antibodies being tested, then the plate was reprobed with a biotinylated antibody targeting H1–DNA followed by streptavidin. Further details are elaborated in Fig. 5. 3. Results 3.1. Is purification with protein A sufficient? Affinity purification of antibodies using protein A or G is commonplace in the research laboratory and, in the minds of many researchers, has come to signify a level of antibody purity unmatched by other chromatographic techniques (e.g. ref. [11]). To illustrate the possibility that this may not be the case, we used NHS76, a human antibody developed at our company, as a test case. NHS76 binds DNA and a class of histone proteins,
1547
known as H1. The antibody was spiked with antigen and tested for purity after protein A purification (Fig. 1a and b). Indeed, the effectiveness of this traditional affinity purification method and our proposed modified process (Fig. 1c) were evaluated by the exposure of antibody either to defined amounts of chromatin (data not shown) or defined amounts of H1 and DNA (discussed here). The starting material for these evaluations was a cocktail of 2 mg/mL NHS76 antibody (∼13 M) “contaminated” with 0.1 mg/mL histone H1 (∼4.5 M) and 0.1 mg/mL DNA (∼0.1 M). Most cell culture media have a sodium chloride concentration between 120 and 150 mM [12]; therefore, to simulate a comparable solution, the starting material cocktail was prepared in PBS with a final sodium chloride concentration of 150 mM. These conditions were sufficient to cause some turbidity in the starting material. Data for each purification run discussed in Figs. 1 and 2 were generated by the injection of 1 mL starting material (which corresponded to 2 mg of antibody and 100 g of H1). Collected fraction peaks from each of the runs, with the exception of the protein A elutant, were dialyzed against water, lyophilized and then resuspended to a volume of 1 mL water. Regardless of analysis by SDS-PAGE or ELISA, analysis of equal volumes from each collected fraction peak would, thus, correspond to an equal percentage of the total collected fraction being compared across the board with other collected fractions. Starting material and the protein A elutant were also loaded on SDSPAGE or ELISA at an equal percentage of the total fraction; however, the protein A elutant was neither dialyzed nor lyophilized upon collection in order to preserve antibody integrity prior to testing for potency. Fig. 1a illustrates purification of the starting material on a 1 mL HiTrap Protein A column that was equilibrated with PBS (150 mM NaCl) for 20 CV prior to the injection of sample at 10 min. Elution of a peak (1) during the sample loading was expected to contain those molecules that did not interact with the protein A. After rinsing with PBS for 48 CV to ensure complete removal of any unbound material, the NHS76 was eluted with 20 mM sodium acetate, 150 mM NaCl, pH 3.2 (peak 2). Preparation and analysis of 2.5% of each collected fraction by SDS-PAGE and western blotting reveals NHS76 in peak 2, as expected. However, H1 is present in peak 2 as well, in sufficient quantities to be detected. Histones were not detected in peak 1, suggesting that most of the contaminating antigen co-eluted with antibody (Fig. 1b). To address the problem of hitchhiker antigens on this antibody, we relied on earlier chromatin research that identified a differential dissociation of each histone protein class from the nucleosome complex at increasing sodium chloride concentrations. The H1 histones dissociate from chromatin between 400 mM and 500 mM NaCl, with the other histone classes dissociating above 800 mM (e.g. see refs. [13–15]). Fig. 1c summarizes our modified approach to protein A purification which entails two salt adjustments for the removal of contaminants. The first adjustment occurs by increasing the NaCl concentration of the starting material from 150 mM to 400 mM, helping dislodge any H1 proteins and their attached NHS76 antibodies from the chromatin matrix. We prefer to adjust with a solution of 2 M NaCl in PBS. This 400 mM NaCl adjustment is designed to recover antibody that could be lost from a cell culture supernatant by inclusion in chromatin debris large enough to act as an insoluble sink. To separate the soluble antibody and histone H1 from the remaining DNA-bound chromatin debris, a 5 mL quarternary amine column (HiTrap Q) is attached in tandem as a
30 min at 37 ◦ C, rinsed 3 × 5 min with TBST, once with TBS alone and then developed with the SIGMA FAST p-Nitrophenyl Phosphate Tablet Set (pNPP; Sigma) for 30 min prior to reading on a SpectraMax M5 plate reader at 405 nm (Molecular Devices). The same plates were used for detection of NHS76 by decanting the pNPP substrate, washing the plates two times with TBS, and then incubating the wells with goat anti-human conjugated to horseradish peroxidase (1:40,000 in blocking buffer; Bethyl Laboratories). The wells were incubated for 30 min at 37 ◦ C and washed as before. This time, the substrate was 3,3 ,5,5 -Tetramethylbenzidine (TMB; KPL) which was allowed to develop for 2.5 min before stopping the reaction with an equal volume of 2 M sulfuric acid and then reading the wells on the plate reader at 450 nm.
1548
K.A. Luhrs et al. / J. Chromatogr. B 877 (2009) 1543–1552
guard column in front of the protein A. The strong anion exchanger captures the negatively charged DNA and its accompanying binding proteins (Fig. 1c). The later adjustment is expected to disrupt the antibody–antigen complex in order to rid the antibody of hitchhikers. It occurs once the antibody has been captured on the protein A column. The mobile phase is adjusted from a 400 mM NaCl PBS buffer to one that contains 2 M NaCl PBS. The captured antibody is rinsed for at least 10 CV so as to flush out any H1 protein that may have been bound to the antibody. The Q column must be removed from the flow prior to this 2 M NaCl rinse step, in order to avoid solubilizing the captured chromatin debris and transferring it onto the protein A column. Upon completion of the rinse step, the mobile phase can be adjusted to a 150 mM NaCl PBS and the protein A equilibrated prior to elution of the antibody with a low pH buffer (Fig. 1c). Such a procedure is illustrated graphically (Fig. 1c) and in a sample chromatogram encompassing three traces (Fig. 1d). The black trace tracks the NaCl adjustments in the mobile phase buffer, while the red trace tracks the actual shift in conductivity in the flow cell as the mobile phase changes, and the blue trace tracks the elution of the starting material constituents at an absorbance wavelength of 280 nm. The first absorbance peak in Fig. 1d corresponds to the elution of those molecules that do not interact with either the Q or protein A columns. The second peak corresponds to those molecules eluted from the protein A during a 2 M NaCl wash, and the final peak represents the antibody eluted with low pH buffer. Preparation and analysis of 2.5% of the collected fraction from the antibody elution, after running the procedure just described, demonstrates that the intact H1 protein is no longer co-eluting with the antibody after modifying the purification procedure (lane 2, Fig. 1e). 3.2. Why not simply adjust the starting material to 2 M NaCl prior to protein A purification? A series of purification strategies were run, each with 1 mL of the starting material. The fractions collected included the sample flow thru peak, the 2 M NaCl wash peak and the protein A elutant peak. In those purification runs incorporating a Q column, it was removed from the line prior to the 2 M NaCl wash. In order to have a complete analysis, the Q column was subsequently washed with 2 M NaCl separately and this elutant peak collected as well. Collected fraction peaks from each of the runs, with the exception of the protein A elutant, were dialyzed against water, lyophilized and then resuspended to a volume of 1 mL water. The protein composition of each peak was analyzed by ELISA (Fig. 2), while the DNA was tracked by gel electrophoresis (Fig. 3). As a starting point in our investigation, the procedure described for Fig. 1a was modified to include a 2 M NaCl washing step (Fig. 2a); otherwise, the remaining steps were unchanged, including loading the starting material in a mobile phase of PBS (150 mM NaCl) and eluting the antibody from the column as described. Equal amounts (2.5%) of each collected fraction were placed into ELISA wells alongside an equivalent amount of 50 g protein A elutant, 50 g starting material and a number of controls, all in triplicate (see Section 2 and Fig. 2 legend). The presence of histone H1 in each fraction was tracked with a cocktail of rabbit polyclonals developed against the major H1 subtypes [9,10]. The presence of the NHS76 antibody was detected in the same wells using a goat anti-human polyclonal specific for the lambda light chain. Without a Q column present, H1 was detected in each of the fractions, including a small amount in the protein A elutant (green arrow and bars, Fig. 2a). Enough hitchhikers were attached to the antibody that a 2 M NaCl wash elicits a significant peak (red arrow and bars, Fig. 2a), and yet, a 10 CV wash of the protein A column is not sufficient for the removal of all the
DNA as evidenced by its presence in the protein A elutant (150 mM NaCl, No Q, Fig. 3a). Expectedly, the NHS76 is largely restricted to the protein A elution (Fig. 2a). With the incorporation of a Q column in front of the protein A during sample loading, the chromatographic pattern changes and the distribution of H1 and DNA changes as well (Figs. 2b and 3a). No longer are H1–DNA hitchhikers eluting during the 2 M NaCl wash, rather, there is a significant elution of H1 in the sample flow thru possibly due to mechanical disruption of some H1–DNA complexes during binding of DNA to the Q column (yellow arrow and bars, Fig. 2b). The remaining balance of the H1s remain with the DNA on the Q column, and a small amount appears to be in the protein A elutant. To verify the presence of H1 on the Q column, it was re-installed after completion of the antibody purification process and subjected to its own 2 M NaCl wash (blue arrow and bars, Fig. 2b). As for the NHS76, it is present in the protein A elutant. Adjusting the starting material to a 400 mM NaCl concentration refines the purification and addresses concerns that antibody could be lost to the insoluble sink of the chromatin matrix. Due to the dissociation of the H1 histones from DNA, there is an increased elution of H1 in the sample flow thru (Fig. 2c). Based on the ELISA results, there is very little H1 present in either the Q column or, surprisingly, the 2 M NaCl wash fractions. Upon completion of the wash, the protein A was equilibrated down to 150 mM NaCl and then antibody recovered with the elution buffer at pH 3.2. This elutant has an H1 content comparable to the NHS76 control (orange bar, Fig. 2c). Tracking the DNA, one finds a greater amount of DNA captured by the Q column with the 400 mM NaCl adjustment than no adjustment at all (compare 150 mM with 400 mM NaCl, Fig. 3a). Such an observation is reasonable when one considers the removal of H1 at 400 mM NaCl results in a greater interaction of the negatively charged phosphate backbone of DNA with the positively charged amines on the Q column. Some might conjecture that a further increase in the NaCl concentration of the starting material should result in the removal of all hitchhikers in the flow thru fraction. Adjustment of the antigen spiked NHS76 to a 600 mM NaCl concentration does provide a chromatogram with significantly larger peaks in the flow thru; however, the 2 M NaCl fraction is also much larger than the comparable washes when the starting material was adjusted to 150 or 400 mM NaCl (Fig. 3b). While a greater amount of H1 is found in the sample flow thru fraction with the 600 mM NaCl adjustment, neither H1 nor the NHS76 antibody can account for the material in the 2 M NaCl wash fraction (Fig. 2d). The answer becomes clear when tracking the DNA distribution amongst the fractions (Fig. 3a). Increasing the concentration of NaCl in the starting material to 600 mM results in a chloride ion content high enough to compete with DNA binding to the Q column resulting in a fouling of the protein A column with excess DNA as was observed when no Q column was employed (150 mM NaCl, No Q, Fig. 3a). This is evidenced by the presence of DNA in the 2 M NaCl wash step and a very dim presence in the protein A elutant (600 mM NaCl, Fig. 3a). Superimposition of the peaks resulting from a 2 M NaCl wash of the Q column shows an increased presence of bound material (DNA) upon removal of H1 at 400 mM NaCl and a dramatic decrease in bound material at 600 mM NaCl (Fig. 3c). 3.3. Are hitchhikers a problem in bioreactors? Granted, the starting material we have used to illustrate our procedure is artificial in its preparation; therefore, we took 1 L of cell culture from a manufacturing run for a chimeric antibody targeting H1–DNA and pulled a third of the sample for antibody purification when 91.5% of the cells were viable. Using the conventional protein A purification in Fig. 1a, antibody was recovered
K.A. Luhrs et al. / J. Chromatogr. B 877 (2009) 1543–1552
1549
Fig. 3. Locating the fractions with DNA. (a) 5% of each collected fraction was dialyzed against TE buffer, adjusted to 0.2% SDS, and digested with 40 g/mL proteinase K overnight at 55 ◦ C before resolution on a 1% agarose TBE gel. Staining with ethidium bromide reveals that the signal from the Q column fraction intensifies as the salt is increased from 150 mM to 400 mM during the initial adjustment of the starting material. At 400 mM, fewer H1s bind the DNA making the negatively charged phosphate backbone available for greater binding to the Q column. Increasing the NaCl to 600 mM causes DNA to foul the protein A column as evidenced by its reduced presence on the Q column and increased presence in the 2 M NaCl wash. Purification without a Q column and only a salt wash does not effectively remove all of the DNA from the protein A elutant (150 mM NaCl, No Q). (b) Overlay of purification process chromatograms from Fig. 2b to d. There is an increase in the materials eluted during sample loading (flow thru) and during the 2 M NaCl wash as the salt used in the initial adjustment of the starting material is increased from 150 mM to 600 mM. (c) Materials eluted from the Q column after its separate 2 M NaCl wash. Overlay of the offline Q column chromatograms from Fig. 2b to d reveals optimal capture of hitchhikers occurs when the starting material is adjusted to 400 mM NaCl.
with a relative binding potency of 106.3% (Fig. 4a). The remainder of the culture was allowed to incubate until 27.9% of the cells were viable in the bioreactor, ensuring the production of antibody in the presence of necrotic debris. Half of this later-stage material was purified by the conventional method and the other half, using the modified protocol, incorporating a 2 M NaCl wash once the antibody was bound to protein A. Despite being purified from the same culture supernatant, the antibody purified by the conventional method had a relative binding potency of 55.5%, which was improved to 89.5% using the modified method. Removal of antigen hitchhikers with the salt wash provided a greater proportion of functional binding antibodies (Fig. 4a). As verification that this increased potency is not due to a salt induced change in the proportion of monomer to aggregate in the sample, both samples were
further purified by size exclusion chromatography before potency analysis was done on the monomer fraction. The monomer to aggregate ratio was found to be nearly identical regardless of the 2 M NaCl wash (Fig. 4b). To determine if this is a problem in other bioreactors, a number of commercially available anti-histone antibodies were screened for chromatin hitchhikers using one of our salt-washed antibodies as the detection reagent (Fig. 5a). Hitchhikers were detectable in some popularly used antibodies, while no histone–DNA contaminants were detected in a control antibody targeting -2 glycoprotein-1 (Fig. 5b and c). The implication of having hitchhikers on an antibody, particularly one used for chromatin immunoprecipitation (ChIP) analysis, is the generation of results not representative of a nuclear protein’s actual distribution in the nucleus (Fig. 5d).
1550
K.A. Luhrs et al. / J. Chromatogr. B 877 (2009) 1543–1552
Fig. 4. Removal of chromatin hitchhikers using the modified protein A purification method. (a) Summary data from three purifications of a chimeric antibody targeting H1–DNA obtained from the same cell culture. The relative potency of the antibody was high when harvested from a cell culture with few dead cells. A drop in binding potency (55.5%) seen in the low cell viability harvest is restored when a 2 M NaCl wash is incorporated into the purification scheme (89.5%). Potency curves were generated by comparing the binding of the purified antibody to a reference standard, at several concentrations, in triplicate, in ELISA wells coated with H1–DNA antigens. The linear regions of the curves were used to calculate the relative amount of functional antibodies in each sample as a percentage of those found in the reference standard. (b) Resolution of antibody monomer and aggregates by size exclusion chromatography using Superdex 200 at a linear flow rate of 20 cm/h. The ratio of monomer to aggregate did not shift all that much for this chimeric antibody with the 2 M NaCl wash. Loss of UV absorbance from H1 and DNA hitchhikers may account for the slight decrease in the peak height for the monomer from the salt washed antibody.
4. Discussion The possibility that antigen hitchhikers are a real problem in antibody purity first manifested itself in our laboratory when we noticed that some of our benchtop production lots of NHS76 would remain soluble in acidic conditions upon elution from a protein A column, but, upon neutralization to pH 7, the same samples would result in a slightly cloudy antibody solution. It was surmised immunoprecipitation of histone–DNA contaminants was occurring
as the sample approached neutral pH. A similar turbidity was seen when we prepared the starting material of 2 mg/mL NHS76 spiked with 0.1 mg/mL each of H1 and DNA. Some might suggest that ion-exchange chromatography should be sufficient for the removal of hitchhikers. Cation and anion exchangers work well in removing free-floating contaminants, however, hitchhikers bound to their respective antibodies can be shielded from charge dependent interactions. Furthermore, for antibodies targeting antigens in a complex matrix, such as
K.A. Luhrs et al. / J. Chromatogr. B 877 (2009) 1543–1552
1551
Fig. 5. Surveying commercially available antibodies for histone–DNA hitchhikers. (a) A schematic presentation for detecting chromatin hitchhikers by applying 5 g of sample (green antibodies) to ELISA wells. Chromatin hitchhikers were detected in the sample wells with a mouse–human chimeric antibody targeting H1–DNA (chTNT-1) that had been purified with a 2 M NaCl wash, in order to ensure that the detection antibody did not bring along its own hitchhikers (yellow antibody). The chTNT-1 was biotinylated with Sulfo-NHS-LC-Biotin (Sigma) at a ratio of 2 moles biotin per mole antibody (chTNT-1/B). The sample wells were incubated with 5 g/mL chTNT-1/B at 37 ◦ C for 1.5 h, rinsed 3 × 5 min with TBST, incubated with streptavidin conjugated to alkaline phosphatase (2 g/mL in blocking buffer; Jackson Immunoresearch) for 30 min at 37 ◦ C, rinsed 3 × 5 min with TBST, once with TBS alone and then developed with pNPP (Sigma) for 1 min prior to reading on a SpectraMax M5 plate reader at 405 nm (Molecular Devices). Using streptavidin, absorbance readings could not be attributed to secondary antibody cross-reactivity. (b and c) Absorbance values (405 nm) for the detection of H1–DNA were obtained by placing 5 g of some popular commercial antibodies in wells in triplicate, including an anti-H4 from supplier U (yellow) and anti-H1s from supplier U (purple) and supplier A (orange). Five micrograms of the biotinylated antibody used as the detection reagent was placed in ELISA wells, in triplicate, as a control (red; signal = 2 OD at 1 min). Other controls included 2.5 g H1 complexed with DNA (light gray), 2.5 g DNA (blue), and 2.5 g H1 (green), all applied in triplicate. (c) As a control, before incubation with chTNT-1/B for detection of chromatin, the same plate was incubated only with streptavidin conjugated to alkaline phosphatase in order to demonstrate lack of cross-reactivity with any of the samples. The plate was read after a 5-min incubation with substrate. (b) Same plate incubated with the biotinylated detection antibody and then streptavidin. The plate was read after a 1-min incubation with substrate. (d) Artifactual amplification of DNA due to hitchhikers on “ChIP-grade” antibodies can lead to data misinterpretation when the PCR products are resolved on a gel.
chromatin or the plasma membrane, such a strategy would not always work, particularly for commercial scale production. In the case of chromatin, highly basic proteins (histones) bind the highly acidic backbone of DNA. Such a complex of positively charged proteins bound to negatively charged nucleic acids may not interact with ion exchange columns avidly. Conversely, significant losses in antibody recovery can occur if hitchhiker antigens are eliminated during purification by ion exchange while antigen and antibody still remain bound to each other. Although we investigated several methods for the dissociation of histone H1 from chromatin and our antibody, including the use of excess arginine, excess lysine, or low pH, we found NaCl to work
best. We exploited the selective removal of H1 from chromatin at 400 mM NaCl. Further increases in the starting salt concentration for the antibody harvest proved counter-productive in the removal of chromatin. At 600 mM NaCl, the increased chloride concentration competed with DNA for binding to the Q column. The use of magnesium and manganese salts was avoided because they lead to chromatin compaction, preventing histone removal [16]. While this particular case study focuses on the removal of hitchhiker antigens from an antibody targeting histone H1, similar problems could arise for other research and therapeutic antibodies that may target an antigen expressed by the producer cell
1552
K.A. Luhrs et al. / J. Chromatogr. B 877 (2009) 1543–1552
line. For these other antigens, alternate means of further purification may be necessary, but the propensity of antibody to remain tightly bound to protein A under conditions of very high ionic strength may allow for similar high-salt washes to be employed in other cases, as well. Of course, while the problem may be similar, each antibody-antigen pair will likely require individualized methods to ensure the purity of the final antibody product. Concerns about conformational changes to the antibodies can be addressed by using any of a number of techniques to monitor protein structure including intrinsic tryptophan fluorescence spectroscopy [17], extrinsic bis-ANS fluorescence spectroscopy [18], or attenuated total reflectance Fourier transformed infrared spectroscopy [19]. Histone hitchhikers have already had a significant impact on the misinterpretation of scientific results, as is evidenced by inconsistencies in lupus research where scientists have struggled to understand whether auto-immune disease patients have anti-DNA antibodies, or if they really are making anti-histone antibodies that happen to bind DNA when they are contaminated with histone hitchhikers [20–22]. As demonstrated by Guth et al. [21], a popular anti-double stranded DNA antibody lost its ability to bind DNA in vitro after purification with protein G incorporating a NaCl wash. The antibody’s DNA binding ability was restored with the introduction of histones. Given the current popularity of ChIP analyses, the need for vigilance in demonstrating the purity of the antibodies being used cannot be over-stated. While our focus here has been antibodies that target histones and DNA, concerns about hitchhiking antigens should be shared by those who use antibodies against any intracellular or membrane proteins. And while hitchhikers can play a significant role in confounding the results of basic research, they can also have a negative affect on the apparent potency of an antibody. In the clinic, antigen contaminants may be benign intruders; however, their affect on the cost of production or the success of clinical trials may be anything but.
Acknowledgements We thank Pete Gagnon, Kelly Felton, Aparna Roy, Van Nguyen and Lilly Mechetner for helpful discussions during the course of the experimentation. References [1] P. Gagnon, Purification tools for monoclonal antibodies, Validated Biosystems, Tucson, AZ, 1996. [2] R. Verma, E. Boleti, A.J.T. George, J. Immunol. Methods 216 (1998) 165. [3] J.V. Gavilondo, J.W. Larrick, BioTechniques 29 (2000) 128. [4] S.D. Gillies, K.-M. Lo, C. Burger, Y. Lan, T. Dahl, W.-K. Wong, Clin. Cancer Res. 8 (2002) 210. [5] L.K. Roskos, C.G. Davis, G.M. Schwab, Drug Dev. Res. 61 (2004) 108. [6] M. Radic, T. Marion, M. Monestier, J. Immunol. 172 (2004) 6692. [7] S. Holdenrieder, P. Stieber, H. Bodenmüller, M. Busch, G. Fertig, H. Fürst, A. Schalhorn, N. Schmeller, M. Untch, D. Seidel, Int. J. Cancer (Pred. Oncol.) 95 (2001) 114. [8] F.-M. Chen, A.L. Epstein, Z. Li, C.R. Taylor, J. Nucl. Med. 31 (1990) 1059. [9] M.H. Parseghian, R.F. Clark, L.J. Hauser, N. Dvorkin, D.A. Harris, B.A. Hamkalo, Chromosome Res. 1 (1993) 127. [10] M.H. Parseghian, D.A. Harris, D.R. Rishwain, B.A. Hamkalo, Chromosoma 103 (1994) 198. [11] E. Harlow, D. Lane, Antibodies: A Laboratory Manual, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, 1988. [12] R.I. Freshney, Culture of Animal Cells: A Manual of Basic Techniques, Wiley-Liss, Inc., Wilmington, DE, 2000. [13] E. Rocha, J.R. Davie, K.E. van Holde, H. Weintraub, J. Biol. Chem. 259 (1984) 8558. [14] W. Li, S. Nagaraja, G.P. Delcuve, M.J. Hendzel, J.R. Davie, Biochem. J. 296 (1993) 737. [15] N.N. Kostova, L. Srebreva, D.V. Markov, I. Rundquist, Cytometry 58A (2004) 132. [16] P.M. Schwarz, A. Felthauser, T.M. Fletcher, J.C. Hansen, Biochemistry 35 (1996) 4009. [17] P. Garidel, M. Hegyi, S. Bassarab, M. Weichel, Biotechnol. J. 3 (2008) 1201. [18] A. Hawe, W. Friess, M. Sutter, W. Jiskoot, Anal. Biochem. 378 (2008) 115. [19] R. Ueberbacher, E. Haimer, R. Hahn, A. Jungbauer, J. Chromatogr. A 1198–1199 (2008) 154. [20] C. Kramers, M.N. Hylkema, M.C.J. van Bruggen, R. van de Lagemaat, H.B.P.M. Dijkman, K.J.M. Assmann, R.J.T. Smeenk, J.H.M. Berden, J. Clin. Invest. 94 (1994) 568. [21] A.M. Guth, X. Zhang, D. Smith, T. Detanico, L.J. Wysocki, J. Immunol. 171 (2003) 6260. [22] L.J. Mason, A. Lambrianides, J.D. Haley, J.J. Manson, D.S. Latchman, D.A. Isenberg, A. Rahman, Arthritis Res. Ther. 7 (2005) R971.
Contents lists available at ScienceDirect
Journal of Chromatography B journal homepage: www.elsevier.com/locate/chromb
Evicting hitchhiker antigens from purified antibodies Keith A. Luhrs a , Debra A. Harris a , Scott Summers b , Missag H. Parseghian a,∗ a b
Research & Development, Peregrine Pharmaceuticals Inc., 14272 Franklin Avenue, Tustin, CA, USA Process Development, Avid Bioservices Inc., Tustin, CA, USA
a r t i c l e
i n f o
Article history: Received 7 January 2009 Accepted 26 March 2009 Available online 1 April 2009 Keywords: Antibody purification Commercial-scale purification Affinity chromatography Host cell proteins Hitchhiker antigens
a b s t r a c t Antibodies that target common cellular structures may have a propensity to bind those very same antigens as they become exposed in dead or dying cells during production in a bioreactor. Those tendencies can be accentuated if the targeted epitope is highly conserved across species. While attention to contaminants such as endotoxin, viral particles, cellular DNA and even prions has grown coincident with the emergence of the monoclonal antibody industry, it is surprising how little attention has been focused on hitchhiker antigens that may co-elute while bound to the supposedly pure antibody. In this case study, we will focus on anti-histone antibodies and the measures we have taken to eliminate stowaways, such as histone–DNA complexes. These simple measures include the addition of a quartenary amine guard column to the protein A, adjusting the ionic strength of the cell culture supernatant to 400 mM sodium chloride, and establishing a mobile phase gradient from 400 mM to 2 M during protein A chromatography. Initially adjusting the cell culture to 600 mM can compromise the quartenary amine guard column. Also, we demonstrate the applicability of these techniques in both the R&D lab and the manufacturing plant, particularly in improving the apparent potency of antibodies destined for the clinic. Given the prominence of anti-histone antibodies in chromatin immunoprecipitation (ChIP), the implications of hitchhiker antigens interferring with the results of an experiment are far-reaching, indeed, we detect them in some popularly used antibodies. Moreover, a wide variety of monoclonals that may target antigens expressed by the producer cell line may face similar problems, resulting in a decreased production yield, as well as a diminished apparent binding potency. © 2009 Elsevier B.V. All rights reserved.
1. Introduction In the past 25 years, antibody reagents have become an important tool in the research lab and the clinic. In this same period, protein purification technologies have seen great advances in the drive to make antibody production commercially viable [1]. Concomitant with improvements in purification techniques has been the advancement of antibody engineering at the structural level. In the research and pharmaceutical development laboratories, engineering improvements have focused on better production of antibodies or their derivatives in a variety of cellular systems [2,3]. For the clinic, engineering improvements have resulted in better pharmaco-kinetic profiles [4,5]. With the proliferation of antibodies against an ever increasing number of intracellular proteins and structures, one must consider the possibility that all of the engineering improvements are for naught if the binding domains are contaminated with antigens that happen to hitchhike along with the antibody during the production
∗ Corresponding author. Tel.: +1 714 508 6052; fax: +1 714 838 6940. E-mail address: [email protected] (M.H. Parseghian). 1570-0232/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.jchromb.2009.03.042
process. Unlike production at the R&D scale in a T-flask, production in a commercial bioreactor can be characterized by increasing numbers of dead cells and debris as the culture matures. In the latter case, two separate, but related, problems can arise. First is the sink effect. Necrotic or late-stage apoptotic cells have ruptured plasma membranes allowing secreted antibodies accessibility to intra-cellular structures, hence, these structures may begin acting as sinks, absorbing antibody product [6]. Second, during necrosis, some cells may release much of their contents into the media as debris providing an opportunity for antigen–antibody interactions in the supernatant [7]. In order to avoid antibody binding to cellular debris, one might shorten the length of the culture run; however, productivity may be dampened if the culture must be harvested early, say prior to reaching a level of cell viability that is less than 70%. Given the expense of setting up and running a bioreactor, premature harvest may provide a diminishing rate of return. One might suggest bioreactors that use a perfusion process can help minimize such problems. Perfusion allows for the regular exchange of media in a bioreactor such that antibody secreted into the supernatant is carried away from the cells from which they originated. However, perfusion techniques also result in the close confinement of dead cells with viable ones, resulting in antibody loss due to a
1544
K.A. Luhrs et al. / J. Chromatogr. B 877 (2009) 1543–1552
sink effect. Furthermore, a perfusion process does not prevent the accumulation of hitchhikers during an antibody’s sojourn in the bioreactor. Our focus has been the development of therapeutic monoclonal antibodies that target histones and DNA; antigens that are extracellularly exposed in the necrotic core of the tumor [6,8]. Our strategy capitalizes on the exposure of nuclear material in necrotic cells at the center of every solid tumor. In our case, the hitchhikers are chromatin components, particularly, histones, which are highly conserved among species. Here, we discuss measures we have taken to minimize the accumulation of these hitchhikers during the purification process. We evaluated the procedures at an R&D scale using antibody samples polluted with histones and DNA in excess. The location of the tainted materials is tracked throughout the purification process using ELISAs and gel electrophoresis. We then verified the effectiveness of the methodology on process scale cell cultures, demonstrating apparent improvements in antibody binding with the removal of the hitchhiker contaminants. Finally, we use the same ELISA strategy to search for hitchhikers in some commercially available antibodies from other sources. 2. Materials and methods 2.1. Reagents NHS76, a fully human antibody targeting histone H1 and DNA, was developed by Peregrine Pharmaceuticals (Tustin, CA) and Cambridge Antibody Technology (Cambridge, UK). Both, NHS76 and a separate mouse–human chimeric antibody (chTNT-1) targeting the same antigens, were manufactured at Peregrine in mammalian cell cultures incubated in a humidified 5% carbon dioxide atmosphere set at 37 ◦ C. 2.2. Antibody purification Starting material for the purification studies was prepared as a 2 mg/mL cocktail of NHS76 antibody spiked with calf histone H1 and herring sperm DNA to a final concentration of 0.1 mg/mL (both from Roche) in phosphate buffered saline (PBS; 10 mM sodium phosphate, 0.15 M sodium chloride, pH 7.2). Samples were adjusted to 400 mM or 600 mM NaCl with PBS containing 2 M NaCl. Purifications were conducted on a BioLogic DuoFlow chromatographic system (BioRad) using a 5 mL Pharmacia HiTrap Q FF column and a 1 mL Pharmacia HiTrap rProtein A FF column placed in tandem (both, GE Healthcare). Each sample was injected into a 1 mL loop and loaded onto the columns with a flow rate of 2 mL/min for 1 min, in fact, the flow rate was 2 mL/min throughout the procedure. Prior to sample loading, the system was equilibrated to the salt concentration of the adjusted sample for 10 min (20 protein A column volumes (CV)). After sample loading, the columns were further equilibrated at the same salt concentration for 10 min (20 CV). The Q column was then removed from the line and the protein A was subjected to an increase to 2 M NaCl PBS in 1 min (2 CV), rinsed with 2 M NaCl PBS for 5 min (10 CV), before being adjusted to 150 mM NaCl PBS in 1 min (2 CV). After equilibration at 150 mM NaCl PBS for 7 min (14 CV), the antibody was eluted into a fraction collector with 20 mM sodium acetate, 150 mM NaCl, pH 3.2 over 8 min (16 CV). The protein A was then re-equilibrated with 150 mM NaCl PBS for another 7 min. 2.3. Q column elution When taken off-line from the purification process, the quartenary amine column was still equilibrated in the NaCl concentration of the adjusted sample. To elute any bound materials, the column
was reattached to the system (sans any protein A column) and further rinsed with the adjusted salt concentration at 4 mL/min for 5 min (4 Q column volumes). Then it was subjected to an increase to 2 M NaCl PBS over 5 min (4 CV), rinsed with 2 M NaCl PBS for 5 min (4 CV), before being equilibrated at 150 mM NaCl PBS for 5 min (4 CV). Fractions were collected throughout the process. 2.4. Sample analysis calculations To load equal amounts of material from collected fractions onto gels or 96-well plates, an assumption was made of 100% antibody recovery during elution from the protein A. Therefore, 2 mg of NHS76 loaded on a protein A column was eluted in a 5 mL fraction of elutant, bringing the final concentration of the purified antibody to 0.4 mg/mL. For the western blot analyses and ELISAs, taking 2.5% of the total collected fraction from the protein A elutant amounts to 50 g of antibody; and, taking 5% for the DNA gel analysis amounts to 100 g of antibody. Comparable signals obtained in the ELISA analysis of NHS76 between the protein A elutants and the 50 g of pure NHS76 placed in the wells as a control, suggests that the assumption of 100% antibody recovery from the protein A column was correct. Similarly, 50 g of starting material antibody was prepared for western or ELISA analysis and 100 g for DNA analysis. In the starting material cocktail, 50 g of antibody should be accompanied by 2.5 g of H1 and DNA, each; therefore, 2.5 g controls of both antigens, in combination and separately, were also applied to the microtiter plate wells. 2.5. Western blot analysis Samples were first dried down, resuspended in an equal volume of 20 mM NaCl and Tris–glycine SDS sample buffer (2×), and then resolved on a 14% T acrylamide gel in a Tris–glycine running buffer supplied with the Novex® electrophoresis system (Invitrogen). To preserve the intact structure of the antibodies (∼150 kD), non-reduced samples were run on the gel at 200 Vfor 1.5 h. H1 histones are highly basic and run anomalously on poly-acrylamide gels, therefore, 0.5 g of H1S -3 was used as a marker for histone migration. Protein bands were transferred to nitrocellulose (Schleicher and Schuell) at 25 V for 1.5 h in a 12 mM Tris–96 mM glycine–20% methanol buffer using the Xcell SureLock® Mini-Cell and XCell IITM Blot Module Kit (Invitrogen). Detection of NHS76 required goat anti-human Fc conjugated to horseradish peroxidase (1:5000 dilution; Jackson Immunoresearch), which was incubated with the blot for 1.5 h at 25 ◦ C, then the blot was rinsed three times for 5 min (3 × 5 min) with TBST (20 mM Tris–HCl, 150 mM NaCl, 0.1% Tween 20, pH 7.4), once with TBS alone (20 mM Tris–HCl, 150 mM NaCl, pH 7.4) and developed using a 3,3 -diaminobenzidine (DAB) Tablet Set (Sigma) to produce brown bands. Subsequently, H1 detection was done on the same blot by using NHS76 (5 g/mL) as the primary antibody for 1.5 h at 25 ◦ C, washing 3 × 5 min with TBST, then using goat anti-human IgG (H + L chain) conjugated to alkaline phosphatase (1:5000 dilution; Jackson Immunoresearch) as a secondary antibody for 30 min at 25 ◦ C, washing 3 × 5 min with TBST, once with TBS alone and then developing with NBT/BCIP Tablets (Roche) to produce blue bands. 2.6. DNA agarose gel analysis Samples were dialyzed against TE (10 mM Tris, 1 mM EDTA, pH 7.5) buffer, adjusted to 0.2% SDS, and digested with 40 g/mL proteinase K (Sigma) overnight at 55 ◦ C before resolution on a 1% agarose TBE gel (89 mM Tris, 89 mM boric acid, 2 mM EDTA) using a RunOne electrophoresis system (EmbiTec). The gel was stained with ethidium bromide (BioRad) and visualized on an Alpha Innotech image analysis system (FluorChem® ).
K.A. Luhrs et al. / J. Chromatogr. B 877 (2009) 1543–1552
1545
Fig. 1. Protein A purification of an antibody spiked with antigen. (a) Chromatogram of 2 mg NHS76, spiked with 0.1 mg each of histone H1 and DNA, being purified on a 1 mL protein A column (mobile phase: PBS [150 mM NaCl]; elution buffer: 20 mM sodium acetate, 150 mM NaCl, pH 3.2; flow rate: 2 mL/min). 2.5% of the collected fraction from the sample flow through (1), and protein A elutant (2), along with an estimated equivalent amount of 50 g starting material (SM) were resolved by a non-reducing SDS-PAGE and analyzed by western blotting (b). (For starting material estimation, see Section 2.) (c) Schematic presentation of proposed purification for removal of H1–DNA antigens
Chromatin (histone and DNA) contamination. Antibody with H1 hitchhiker. (d) Sample chromatogram generated from the from the antibody preparation. modified purification scheme. The three traces are described in the main text. (e) 2.5% of the collected fraction from the protein A elutant (2), using the modified purification scheme, alongside 50 g starting material were resolved by a non-reducing SDS-PAGE and analyzed by western blotting. (b and e) The blotted NHS76 was detected using goat anti-Human Fc conjugated to horseradish peroxidase and developed with DAB substrate to produce brown bands. H1 was detected using NHS76 as a primary antibody, goat anti-human IgG (H + L chain) conjugated to alkaline phosphatase as a secondary antibody, and then developed with NBT/BCIP substrate to produce blue bands. Pure H1S -3, a member of the H1 family of histones, was used as a marker for H1 migration.
2.7. H1 and NHS76 ELISA analysis
2.8. Purification from Cell Culture
Samples were placed into 96-well microtiter plates (Costar, Cat. # 3361), in triplicate, and allowed to adhere to the wells overnight at 4 ◦ C. Any exposed areas in the wells were coated with blocking buffer (2% non-fat dry milk, NFDM, in TBS). Samples were applied at 2.5% of each collected fraction while controls were applied to the wells based on the calculation rationale described above (see Section 2.4). H1 was detected in the wells with rabbit anti-H1 polyclonals described elsewhere [9,10] at a dilution of 1:500 in blocking buffer. The same plates were used for the detection of NHS76 using goat anti-human conjugated to horseradish peroxidase at 1:40,000 in blocking buffer. Further details are elaborated in Fig. 2.
One liter of cell culture was removed from a manufacturing run for a chimeric antibody targeting H1–DNA (chTNT-1) and placed in a spinner flask that was incubated in a humidified 5% carbon dioxide atmosphere set at 37 ◦ C. A third of this sample was removed when the cells were at 91.5% viability and the remaining twothirds allowed to incubate in the spinner flask until cell viability dropped to 27.9%. Cell harvests were centrifuged to remove cellular debris, the supernatant decanted, filtered with a 0.2 m cellulose acetate filter (Corning) and the antibody purified as described in Section 3. Purification required a Pharmacia 6 cm XK16 rProtein A Sepharose FF column (GE Healthcare). Samples were loaded
1546
K.A. Luhrs et al. / J. Chromatogr. B 877 (2009) 1543–1552
Fig. 2. Locating the fractions with H1 and NHS76. 2.5% of each collected fraction was placed into ELISA wells, including the flow thru fraction during sample loading (yellow arrows and bars), material captured by the Q column (blue arrows and bars), the 2 M NaCl wash (red arrows and bars), 50 g protein A elutant (green arrows and bars), 50 g starting material (dark gray bars), 50 g pure NHS76 (orange bars), 2.5 g H1 complexed with DNA (light gray bars), 2.5 g DNA (olive bars), and 2.5 g H1 (purple bars). All samples were applied to wells in triplicate. (a) Purification of 2 mg NHS76 antibody, spiked with 0.1 mg each of histone H1 and DNA, as described in Fig. 1a and including a 2 M NaCl wash of NHS76 during protein A capture. (b) Purification including Q column. Purification with NaCl adjustment of starting material to (c) 400 mM and (d) 600 mM. An artifact peak (light blue arrows and bar in (d)) also appears in the chromatograms during the gradient drop from 2 M NaCl down to 150 mM NaCl, both, during the purification process (a) or during the offline Q column wash (d). No significant quantities of protein or DNA were found in those fractions (see (d); also see lane “Q Column 2nd Peak”, Fig. 3a). H1 was detected in the wells with rabbit anti-H1 polyclonals at a dilution of 1:500 in blocking buffer. After a 1.5-h incubation at 37 ◦ C, the wells were rinsed 3 × 5 min with TBST, incubated with goat anti-rabbit conjugated to alkaline phosphatase (1:5000 dilution in blocking buffer; Jackson Immunoresearch) for
K.A. Luhrs et al. / J. Chromatogr. B 877 (2009) 1543–1552
at 250 cm/h (8.37 mL/min) and the column was equilibrated and washed at 200 cm/h (6.7 mL/min). The elution of antibody occurred at 100 cm/h (3.4 mL/min) with 20 mM sodium acetate, 150 mM NaCl, pH 3.2 and, once eluted, it was neutralized with sodium hydroxide to pH 6.5. This elutant was then loaded onto a Pharmacia XK26/60 Superdex 200 (prep grade) size exclusion column (GE Healthcare) and fractionated into antibody monomer and aggregate using a PBS running buffer at a flow rate of 20 cm/h (1.77 mL/min). Both, protein A and size exclusion chromatography were conducted on a Pharmacia ÄKTATM FPLC purification system (GE Healthcare). Potency of the monomer fraction was tested as described below. 2.9. Binding potency of a chimeric antibody targeting the H1–DNA antigen The antigenic target for our antibodies was coated on to the wells of microtiter plates with a solution of 10 g/mL calf histone H1 and 50 g/mL herring sperm DNA (both from Roche). Any exposed areas in the wells were coated with blocking buffer (2% non-fat dry milk, NFDM, in TBS). The antibody samples were serially diluted and binding potency was assessed by incubation of the resulting concentrations in the antigen coated wells, in triplicate, for 1.5 h at 37 ◦ C. Excess antibody was washed out with TBST as described for the western analysis, and the presence of the chimeric antibody was detected with goat anti-human IgG (H + L) conjugated to alkaline phosphatase (1:8000 dilution; Jackson Immunoresearch). After a 30-min incubation at 37 ◦ C, excess secondary antibody was washed out as described for the western analysis and the ELISA developed using the SIGMA FAST p-Nitrophenyl Phosphate Tablet Set (pNPP; Sigma). The signal generated at each concentration was plotted as a sigmoidal dose–response curve and the linear portion of the curve from each sample compared to the same generated with a reference standard. An analysis of variance was used to determine the relative potency of the samples to the reference standard and this value was presented as a percentage of potency. 2.10. Screening for hitchhikers Samples of commercially available antibodies were applied to a 96-well microtiter plate, in triplicate, and allowed to adhere to the wells overnight at 4 ◦ C. Any exposed areas in the wells were coated with blocking buffer (2% NFDM in TBS). The wells were incubated as described in Fig. 5 first with streptavidin alone, to demonstrate lack of cross-reactivity with the commercial antibodies being tested, then the plate was reprobed with a biotinylated antibody targeting H1–DNA followed by streptavidin. Further details are elaborated in Fig. 5. 3. Results 3.1. Is purification with protein A sufficient? Affinity purification of antibodies using protein A or G is commonplace in the research laboratory and, in the minds of many researchers, has come to signify a level of antibody purity unmatched by other chromatographic techniques (e.g. ref. [11]). To illustrate the possibility that this may not be the case, we used NHS76, a human antibody developed at our company, as a test case. NHS76 binds DNA and a class of histone proteins,
1547
known as H1. The antibody was spiked with antigen and tested for purity after protein A purification (Fig. 1a and b). Indeed, the effectiveness of this traditional affinity purification method and our proposed modified process (Fig. 1c) were evaluated by the exposure of antibody either to defined amounts of chromatin (data not shown) or defined amounts of H1 and DNA (discussed here). The starting material for these evaluations was a cocktail of 2 mg/mL NHS76 antibody (∼13 M) “contaminated” with 0.1 mg/mL histone H1 (∼4.5 M) and 0.1 mg/mL DNA (∼0.1 M). Most cell culture media have a sodium chloride concentration between 120 and 150 mM [12]; therefore, to simulate a comparable solution, the starting material cocktail was prepared in PBS with a final sodium chloride concentration of 150 mM. These conditions were sufficient to cause some turbidity in the starting material. Data for each purification run discussed in Figs. 1 and 2 were generated by the injection of 1 mL starting material (which corresponded to 2 mg of antibody and 100 g of H1). Collected fraction peaks from each of the runs, with the exception of the protein A elutant, were dialyzed against water, lyophilized and then resuspended to a volume of 1 mL water. Regardless of analysis by SDS-PAGE or ELISA, analysis of equal volumes from each collected fraction peak would, thus, correspond to an equal percentage of the total collected fraction being compared across the board with other collected fractions. Starting material and the protein A elutant were also loaded on SDSPAGE or ELISA at an equal percentage of the total fraction; however, the protein A elutant was neither dialyzed nor lyophilized upon collection in order to preserve antibody integrity prior to testing for potency. Fig. 1a illustrates purification of the starting material on a 1 mL HiTrap Protein A column that was equilibrated with PBS (150 mM NaCl) for 20 CV prior to the injection of sample at 10 min. Elution of a peak (1) during the sample loading was expected to contain those molecules that did not interact with the protein A. After rinsing with PBS for 48 CV to ensure complete removal of any unbound material, the NHS76 was eluted with 20 mM sodium acetate, 150 mM NaCl, pH 3.2 (peak 2). Preparation and analysis of 2.5% of each collected fraction by SDS-PAGE and western blotting reveals NHS76 in peak 2, as expected. However, H1 is present in peak 2 as well, in sufficient quantities to be detected. Histones were not detected in peak 1, suggesting that most of the contaminating antigen co-eluted with antibody (Fig. 1b). To address the problem of hitchhiker antigens on this antibody, we relied on earlier chromatin research that identified a differential dissociation of each histone protein class from the nucleosome complex at increasing sodium chloride concentrations. The H1 histones dissociate from chromatin between 400 mM and 500 mM NaCl, with the other histone classes dissociating above 800 mM (e.g. see refs. [13–15]). Fig. 1c summarizes our modified approach to protein A purification which entails two salt adjustments for the removal of contaminants. The first adjustment occurs by increasing the NaCl concentration of the starting material from 150 mM to 400 mM, helping dislodge any H1 proteins and their attached NHS76 antibodies from the chromatin matrix. We prefer to adjust with a solution of 2 M NaCl in PBS. This 400 mM NaCl adjustment is designed to recover antibody that could be lost from a cell culture supernatant by inclusion in chromatin debris large enough to act as an insoluble sink. To separate the soluble antibody and histone H1 from the remaining DNA-bound chromatin debris, a 5 mL quarternary amine column (HiTrap Q) is attached in tandem as a
30 min at 37 ◦ C, rinsed 3 × 5 min with TBST, once with TBS alone and then developed with the SIGMA FAST p-Nitrophenyl Phosphate Tablet Set (pNPP; Sigma) for 30 min prior to reading on a SpectraMax M5 plate reader at 405 nm (Molecular Devices). The same plates were used for detection of NHS76 by decanting the pNPP substrate, washing the plates two times with TBS, and then incubating the wells with goat anti-human conjugated to horseradish peroxidase (1:40,000 in blocking buffer; Bethyl Laboratories). The wells were incubated for 30 min at 37 ◦ C and washed as before. This time, the substrate was 3,3 ,5,5 -Tetramethylbenzidine (TMB; KPL) which was allowed to develop for 2.5 min before stopping the reaction with an equal volume of 2 M sulfuric acid and then reading the wells on the plate reader at 450 nm.
1548
K.A. Luhrs et al. / J. Chromatogr. B 877 (2009) 1543–1552
guard column in front of the protein A. The strong anion exchanger captures the negatively charged DNA and its accompanying binding proteins (Fig. 1c). The later adjustment is expected to disrupt the antibody–antigen complex in order to rid the antibody of hitchhikers. It occurs once the antibody has been captured on the protein A column. The mobile phase is adjusted from a 400 mM NaCl PBS buffer to one that contains 2 M NaCl PBS. The captured antibody is rinsed for at least 10 CV so as to flush out any H1 protein that may have been bound to the antibody. The Q column must be removed from the flow prior to this 2 M NaCl rinse step, in order to avoid solubilizing the captured chromatin debris and transferring it onto the protein A column. Upon completion of the rinse step, the mobile phase can be adjusted to a 150 mM NaCl PBS and the protein A equilibrated prior to elution of the antibody with a low pH buffer (Fig. 1c). Such a procedure is illustrated graphically (Fig. 1c) and in a sample chromatogram encompassing three traces (Fig. 1d). The black trace tracks the NaCl adjustments in the mobile phase buffer, while the red trace tracks the actual shift in conductivity in the flow cell as the mobile phase changes, and the blue trace tracks the elution of the starting material constituents at an absorbance wavelength of 280 nm. The first absorbance peak in Fig. 1d corresponds to the elution of those molecules that do not interact with either the Q or protein A columns. The second peak corresponds to those molecules eluted from the protein A during a 2 M NaCl wash, and the final peak represents the antibody eluted with low pH buffer. Preparation and analysis of 2.5% of the collected fraction from the antibody elution, after running the procedure just described, demonstrates that the intact H1 protein is no longer co-eluting with the antibody after modifying the purification procedure (lane 2, Fig. 1e). 3.2. Why not simply adjust the starting material to 2 M NaCl prior to protein A purification? A series of purification strategies were run, each with 1 mL of the starting material. The fractions collected included the sample flow thru peak, the 2 M NaCl wash peak and the protein A elutant peak. In those purification runs incorporating a Q column, it was removed from the line prior to the 2 M NaCl wash. In order to have a complete analysis, the Q column was subsequently washed with 2 M NaCl separately and this elutant peak collected as well. Collected fraction peaks from each of the runs, with the exception of the protein A elutant, were dialyzed against water, lyophilized and then resuspended to a volume of 1 mL water. The protein composition of each peak was analyzed by ELISA (Fig. 2), while the DNA was tracked by gel electrophoresis (Fig. 3). As a starting point in our investigation, the procedure described for Fig. 1a was modified to include a 2 M NaCl washing step (Fig. 2a); otherwise, the remaining steps were unchanged, including loading the starting material in a mobile phase of PBS (150 mM NaCl) and eluting the antibody from the column as described. Equal amounts (2.5%) of each collected fraction were placed into ELISA wells alongside an equivalent amount of 50 g protein A elutant, 50 g starting material and a number of controls, all in triplicate (see Section 2 and Fig. 2 legend). The presence of histone H1 in each fraction was tracked with a cocktail of rabbit polyclonals developed against the major H1 subtypes [9,10]. The presence of the NHS76 antibody was detected in the same wells using a goat anti-human polyclonal specific for the lambda light chain. Without a Q column present, H1 was detected in each of the fractions, including a small amount in the protein A elutant (green arrow and bars, Fig. 2a). Enough hitchhikers were attached to the antibody that a 2 M NaCl wash elicits a significant peak (red arrow and bars, Fig. 2a), and yet, a 10 CV wash of the protein A column is not sufficient for the removal of all the
DNA as evidenced by its presence in the protein A elutant (150 mM NaCl, No Q, Fig. 3a). Expectedly, the NHS76 is largely restricted to the protein A elution (Fig. 2a). With the incorporation of a Q column in front of the protein A during sample loading, the chromatographic pattern changes and the distribution of H1 and DNA changes as well (Figs. 2b and 3a). No longer are H1–DNA hitchhikers eluting during the 2 M NaCl wash, rather, there is a significant elution of H1 in the sample flow thru possibly due to mechanical disruption of some H1–DNA complexes during binding of DNA to the Q column (yellow arrow and bars, Fig. 2b). The remaining balance of the H1s remain with the DNA on the Q column, and a small amount appears to be in the protein A elutant. To verify the presence of H1 on the Q column, it was re-installed after completion of the antibody purification process and subjected to its own 2 M NaCl wash (blue arrow and bars, Fig. 2b). As for the NHS76, it is present in the protein A elutant. Adjusting the starting material to a 400 mM NaCl concentration refines the purification and addresses concerns that antibody could be lost to the insoluble sink of the chromatin matrix. Due to the dissociation of the H1 histones from DNA, there is an increased elution of H1 in the sample flow thru (Fig. 2c). Based on the ELISA results, there is very little H1 present in either the Q column or, surprisingly, the 2 M NaCl wash fractions. Upon completion of the wash, the protein A was equilibrated down to 150 mM NaCl and then antibody recovered with the elution buffer at pH 3.2. This elutant has an H1 content comparable to the NHS76 control (orange bar, Fig. 2c). Tracking the DNA, one finds a greater amount of DNA captured by the Q column with the 400 mM NaCl adjustment than no adjustment at all (compare 150 mM with 400 mM NaCl, Fig. 3a). Such an observation is reasonable when one considers the removal of H1 at 400 mM NaCl results in a greater interaction of the negatively charged phosphate backbone of DNA with the positively charged amines on the Q column. Some might conjecture that a further increase in the NaCl concentration of the starting material should result in the removal of all hitchhikers in the flow thru fraction. Adjustment of the antigen spiked NHS76 to a 600 mM NaCl concentration does provide a chromatogram with significantly larger peaks in the flow thru; however, the 2 M NaCl fraction is also much larger than the comparable washes when the starting material was adjusted to 150 or 400 mM NaCl (Fig. 3b). While a greater amount of H1 is found in the sample flow thru fraction with the 600 mM NaCl adjustment, neither H1 nor the NHS76 antibody can account for the material in the 2 M NaCl wash fraction (Fig. 2d). The answer becomes clear when tracking the DNA distribution amongst the fractions (Fig. 3a). Increasing the concentration of NaCl in the starting material to 600 mM results in a chloride ion content high enough to compete with DNA binding to the Q column resulting in a fouling of the protein A column with excess DNA as was observed when no Q column was employed (150 mM NaCl, No Q, Fig. 3a). This is evidenced by the presence of DNA in the 2 M NaCl wash step and a very dim presence in the protein A elutant (600 mM NaCl, Fig. 3a). Superimposition of the peaks resulting from a 2 M NaCl wash of the Q column shows an increased presence of bound material (DNA) upon removal of H1 at 400 mM NaCl and a dramatic decrease in bound material at 600 mM NaCl (Fig. 3c). 3.3. Are hitchhikers a problem in bioreactors? Granted, the starting material we have used to illustrate our procedure is artificial in its preparation; therefore, we took 1 L of cell culture from a manufacturing run for a chimeric antibody targeting H1–DNA and pulled a third of the sample for antibody purification when 91.5% of the cells were viable. Using the conventional protein A purification in Fig. 1a, antibody was recovered
K.A. Luhrs et al. / J. Chromatogr. B 877 (2009) 1543–1552
1549
Fig. 3. Locating the fractions with DNA. (a) 5% of each collected fraction was dialyzed against TE buffer, adjusted to 0.2% SDS, and digested with 40 g/mL proteinase K overnight at 55 ◦ C before resolution on a 1% agarose TBE gel. Staining with ethidium bromide reveals that the signal from the Q column fraction intensifies as the salt is increased from 150 mM to 400 mM during the initial adjustment of the starting material. At 400 mM, fewer H1s bind the DNA making the negatively charged phosphate backbone available for greater binding to the Q column. Increasing the NaCl to 600 mM causes DNA to foul the protein A column as evidenced by its reduced presence on the Q column and increased presence in the 2 M NaCl wash. Purification without a Q column and only a salt wash does not effectively remove all of the DNA from the protein A elutant (150 mM NaCl, No Q). (b) Overlay of purification process chromatograms from Fig. 2b to d. There is an increase in the materials eluted during sample loading (flow thru) and during the 2 M NaCl wash as the salt used in the initial adjustment of the starting material is increased from 150 mM to 600 mM. (c) Materials eluted from the Q column after its separate 2 M NaCl wash. Overlay of the offline Q column chromatograms from Fig. 2b to d reveals optimal capture of hitchhikers occurs when the starting material is adjusted to 400 mM NaCl.
with a relative binding potency of 106.3% (Fig. 4a). The remainder of the culture was allowed to incubate until 27.9% of the cells were viable in the bioreactor, ensuring the production of antibody in the presence of necrotic debris. Half of this later-stage material was purified by the conventional method and the other half, using the modified protocol, incorporating a 2 M NaCl wash once the antibody was bound to protein A. Despite being purified from the same culture supernatant, the antibody purified by the conventional method had a relative binding potency of 55.5%, which was improved to 89.5% using the modified method. Removal of antigen hitchhikers with the salt wash provided a greater proportion of functional binding antibodies (Fig. 4a). As verification that this increased potency is not due to a salt induced change in the proportion of monomer to aggregate in the sample, both samples were
further purified by size exclusion chromatography before potency analysis was done on the monomer fraction. The monomer to aggregate ratio was found to be nearly identical regardless of the 2 M NaCl wash (Fig. 4b). To determine if this is a problem in other bioreactors, a number of commercially available anti-histone antibodies were screened for chromatin hitchhikers using one of our salt-washed antibodies as the detection reagent (Fig. 5a). Hitchhikers were detectable in some popularly used antibodies, while no histone–DNA contaminants were detected in a control antibody targeting -2 glycoprotein-1 (Fig. 5b and c). The implication of having hitchhikers on an antibody, particularly one used for chromatin immunoprecipitation (ChIP) analysis, is the generation of results not representative of a nuclear protein’s actual distribution in the nucleus (Fig. 5d).
1550
K.A. Luhrs et al. / J. Chromatogr. B 877 (2009) 1543–1552
Fig. 4. Removal of chromatin hitchhikers using the modified protein A purification method. (a) Summary data from three purifications of a chimeric antibody targeting H1–DNA obtained from the same cell culture. The relative potency of the antibody was high when harvested from a cell culture with few dead cells. A drop in binding potency (55.5%) seen in the low cell viability harvest is restored when a 2 M NaCl wash is incorporated into the purification scheme (89.5%). Potency curves were generated by comparing the binding of the purified antibody to a reference standard, at several concentrations, in triplicate, in ELISA wells coated with H1–DNA antigens. The linear regions of the curves were used to calculate the relative amount of functional antibodies in each sample as a percentage of those found in the reference standard. (b) Resolution of antibody monomer and aggregates by size exclusion chromatography using Superdex 200 at a linear flow rate of 20 cm/h. The ratio of monomer to aggregate did not shift all that much for this chimeric antibody with the 2 M NaCl wash. Loss of UV absorbance from H1 and DNA hitchhikers may account for the slight decrease in the peak height for the monomer from the salt washed antibody.
4. Discussion The possibility that antigen hitchhikers are a real problem in antibody purity first manifested itself in our laboratory when we noticed that some of our benchtop production lots of NHS76 would remain soluble in acidic conditions upon elution from a protein A column, but, upon neutralization to pH 7, the same samples would result in a slightly cloudy antibody solution. It was surmised immunoprecipitation of histone–DNA contaminants was occurring
as the sample approached neutral pH. A similar turbidity was seen when we prepared the starting material of 2 mg/mL NHS76 spiked with 0.1 mg/mL each of H1 and DNA. Some might suggest that ion-exchange chromatography should be sufficient for the removal of hitchhikers. Cation and anion exchangers work well in removing free-floating contaminants, however, hitchhikers bound to their respective antibodies can be shielded from charge dependent interactions. Furthermore, for antibodies targeting antigens in a complex matrix, such as
K.A. Luhrs et al. / J. Chromatogr. B 877 (2009) 1543–1552
1551
Fig. 5. Surveying commercially available antibodies for histone–DNA hitchhikers. (a) A schematic presentation for detecting chromatin hitchhikers by applying 5 g of sample (green antibodies) to ELISA wells. Chromatin hitchhikers were detected in the sample wells with a mouse–human chimeric antibody targeting H1–DNA (chTNT-1) that had been purified with a 2 M NaCl wash, in order to ensure that the detection antibody did not bring along its own hitchhikers (yellow antibody). The chTNT-1 was biotinylated with Sulfo-NHS-LC-Biotin (Sigma) at a ratio of 2 moles biotin per mole antibody (chTNT-1/B). The sample wells were incubated with 5 g/mL chTNT-1/B at 37 ◦ C for 1.5 h, rinsed 3 × 5 min with TBST, incubated with streptavidin conjugated to alkaline phosphatase (2 g/mL in blocking buffer; Jackson Immunoresearch) for 30 min at 37 ◦ C, rinsed 3 × 5 min with TBST, once with TBS alone and then developed with pNPP (Sigma) for 1 min prior to reading on a SpectraMax M5 plate reader at 405 nm (Molecular Devices). Using streptavidin, absorbance readings could not be attributed to secondary antibody cross-reactivity. (b and c) Absorbance values (405 nm) for the detection of H1–DNA were obtained by placing 5 g of some popular commercial antibodies in wells in triplicate, including an anti-H4 from supplier U (yellow) and anti-H1s from supplier U (purple) and supplier A (orange). Five micrograms of the biotinylated antibody used as the detection reagent was placed in ELISA wells, in triplicate, as a control (red; signal = 2 OD at 1 min). Other controls included 2.5 g H1 complexed with DNA (light gray), 2.5 g DNA (blue), and 2.5 g H1 (green), all applied in triplicate. (c) As a control, before incubation with chTNT-1/B for detection of chromatin, the same plate was incubated only with streptavidin conjugated to alkaline phosphatase in order to demonstrate lack of cross-reactivity with any of the samples. The plate was read after a 5-min incubation with substrate. (b) Same plate incubated with the biotinylated detection antibody and then streptavidin. The plate was read after a 1-min incubation with substrate. (d) Artifactual amplification of DNA due to hitchhikers on “ChIP-grade” antibodies can lead to data misinterpretation when the PCR products are resolved on a gel.
chromatin or the plasma membrane, such a strategy would not always work, particularly for commercial scale production. In the case of chromatin, highly basic proteins (histones) bind the highly acidic backbone of DNA. Such a complex of positively charged proteins bound to negatively charged nucleic acids may not interact with ion exchange columns avidly. Conversely, significant losses in antibody recovery can occur if hitchhiker antigens are eliminated during purification by ion exchange while antigen and antibody still remain bound to each other. Although we investigated several methods for the dissociation of histone H1 from chromatin and our antibody, including the use of excess arginine, excess lysine, or low pH, we found NaCl to work
best. We exploited the selective removal of H1 from chromatin at 400 mM NaCl. Further increases in the starting salt concentration for the antibody harvest proved counter-productive in the removal of chromatin. At 600 mM NaCl, the increased chloride concentration competed with DNA for binding to the Q column. The use of magnesium and manganese salts was avoided because they lead to chromatin compaction, preventing histone removal [16]. While this particular case study focuses on the removal of hitchhiker antigens from an antibody targeting histone H1, similar problems could arise for other research and therapeutic antibodies that may target an antigen expressed by the producer cell
1552
K.A. Luhrs et al. / J. Chromatogr. B 877 (2009) 1543–1552
line. For these other antigens, alternate means of further purification may be necessary, but the propensity of antibody to remain tightly bound to protein A under conditions of very high ionic strength may allow for similar high-salt washes to be employed in other cases, as well. Of course, while the problem may be similar, each antibody-antigen pair will likely require individualized methods to ensure the purity of the final antibody product. Concerns about conformational changes to the antibodies can be addressed by using any of a number of techniques to monitor protein structure including intrinsic tryptophan fluorescence spectroscopy [17], extrinsic bis-ANS fluorescence spectroscopy [18], or attenuated total reflectance Fourier transformed infrared spectroscopy [19]. Histone hitchhikers have already had a significant impact on the misinterpretation of scientific results, as is evidenced by inconsistencies in lupus research where scientists have struggled to understand whether auto-immune disease patients have anti-DNA antibodies, or if they really are making anti-histone antibodies that happen to bind DNA when they are contaminated with histone hitchhikers [20–22]. As demonstrated by Guth et al. [21], a popular anti-double stranded DNA antibody lost its ability to bind DNA in vitro after purification with protein G incorporating a NaCl wash. The antibody’s DNA binding ability was restored with the introduction of histones. Given the current popularity of ChIP analyses, the need for vigilance in demonstrating the purity of the antibodies being used cannot be over-stated. While our focus here has been antibodies that target histones and DNA, concerns about hitchhiking antigens should be shared by those who use antibodies against any intracellular or membrane proteins. And while hitchhikers can play a significant role in confounding the results of basic research, they can also have a negative affect on the apparent potency of an antibody. In the clinic, antigen contaminants may be benign intruders; however, their affect on the cost of production or the success of clinical trials may be anything but.
Acknowledgements We thank Pete Gagnon, Kelly Felton, Aparna Roy, Van Nguyen and Lilly Mechetner for helpful discussions during the course of the experimentation. References [1] P. Gagnon, Purification tools for monoclonal antibodies, Validated Biosystems, Tucson, AZ, 1996. [2] R. Verma, E. Boleti, A.J.T. George, J. Immunol. Methods 216 (1998) 165. [3] J.V. Gavilondo, J.W. Larrick, BioTechniques 29 (2000) 128. [4] S.D. Gillies, K.-M. Lo, C. Burger, Y. Lan, T. Dahl, W.-K. Wong, Clin. Cancer Res. 8 (2002) 210. [5] L.K. Roskos, C.G. Davis, G.M. Schwab, Drug Dev. Res. 61 (2004) 108. [6] M. Radic, T. Marion, M. Monestier, J. Immunol. 172 (2004) 6692. [7] S. Holdenrieder, P. Stieber, H. Bodenmüller, M. Busch, G. Fertig, H. Fürst, A. Schalhorn, N. Schmeller, M. Untch, D. Seidel, Int. J. Cancer (Pred. Oncol.) 95 (2001) 114. [8] F.-M. Chen, A.L. Epstein, Z. Li, C.R. Taylor, J. Nucl. Med. 31 (1990) 1059. [9] M.H. Parseghian, R.F. Clark, L.J. Hauser, N. Dvorkin, D.A. Harris, B.A. Hamkalo, Chromosome Res. 1 (1993) 127. [10] M.H. Parseghian, D.A. Harris, D.R. Rishwain, B.A. Hamkalo, Chromosoma 103 (1994) 198. [11] E. Harlow, D. Lane, Antibodies: A Laboratory Manual, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, 1988. [12] R.I. Freshney, Culture of Animal Cells: A Manual of Basic Techniques, Wiley-Liss, Inc., Wilmington, DE, 2000. [13] E. Rocha, J.R. Davie, K.E. van Holde, H. Weintraub, J. Biol. Chem. 259 (1984) 8558. [14] W. Li, S. Nagaraja, G.P. Delcuve, M.J. Hendzel, J.R. Davie, Biochem. J. 296 (1993) 737. [15] N.N. Kostova, L. Srebreva, D.V. Markov, I. Rundquist, Cytometry 58A (2004) 132. [16] P.M. Schwarz, A. Felthauser, T.M. Fletcher, J.C. Hansen, Biochemistry 35 (1996) 4009. [17] P. Garidel, M. Hegyi, S. Bassarab, M. Weichel, Biotechnol. J. 3 (2008) 1201. [18] A. Hawe, W. Friess, M. Sutter, W. Jiskoot, Anal. Biochem. 378 (2008) 115. [19] R. Ueberbacher, E. Haimer, R. Hahn, A. Jungbauer, J. Chromatogr. A 1198–1199 (2008) 154. [20] C. Kramers, M.N. Hylkema, M.C.J. van Bruggen, R. van de Lagemaat, H.B.P.M. Dijkman, K.J.M. Assmann, R.J.T. Smeenk, J.H.M. Berden, J. Clin. Invest. 94 (1994) 568. [21] A.M. Guth, X. Zhang, D. Smith, T. Detanico, L.J. Wysocki, J. Immunol. 171 (2003) 6260. [22] L.J. Mason, A. Lambrianides, J.D. Haley, J.J. Manson, D.S. Latchman, D.A. Isenberg, A. Rahman, Arthritis Res. Ther. 7 (2005) R971.