| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
,
,
* Institute of Molecular Biotechnology, RWTH Aachen University, Aachen, Germany;
Institute of Applied Microbiology, University of Natural Resources and Applied Life Sciences, Vienna, Austria;
Polymun Scientific Immunbiologische Forschung GmbH, Vienna, Austria; and
Fraunhofer IME, Aachen, Germany
1Correspondence: Institute of Molecular Biotechnology, RTWH Aachen University, Worringerweg 1, 52074 Aachen, Germany. E-mail: rademacher{at}molbiotech.rwth-aachen.de
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
RI for kinetic analysis. The plant-derived antibody showed similar kinetic properties and 89% of the binding capacity of its CHO-derived counterpart, but was only 33% as efficient in HIV-1 neutralization assays. Our results show that plant suspension cultures can be used to produce human antibodies efficiently and that the analysis methods used in this study, including biosensor technology, provide useful functional data about antibody performance. This highlights important issues raised by the use of plant systems to produce human biologics.—Sack, M., Paetz, A., Kunert, R., Bomble, M., Hesse, F., Stiegler, G., Fischer, R., Katinger, H., Stoeger, E., Rademacher, T. Functional analysis of the broadly neutralizing human anti-HIV-1 antibody 2F5 produced in transgenic BY-2 suspension cultures.
Key Words: FcRI ectodomain • surface plasmon resonance • antibody activity • recombinant pharmaceutical glycoprotein
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
2
3
4)
.
When choosing a nonhuman production system, several issues must be considered. Several studies have shown that the effector functions, and thus the clinical performance of antibodies, can be affected by the nature of the attached N-glycans (5
, 6)
. In particular, it has been shown that antibody-dependent cellular cytotoxicity is significantly increased for antibodies lacking core-
(1,6)-fucose and, to a lesser extent, containing a bisecting GlcNAc (7)
. A reduction in sialic acid and galactose of serum IgG has been observed in patients with inflammatory autoimmune diseases (8)
. A recent report by Kaneko et al. has demonstrated that sialylation of the N-glycans attached to Asn297 is the structural determinant for the anti-inflammatory activity of intravenous gamma globulin. A fraction enriched for sialic acid containing N-glycans showed a 10-fold increased ability to inhibit inflammatory responses in the K/N serum arthritis model (9)
. A comparative analysis of N-glycans attached to Fc and Fab has shown that Fc sialylation is a specifically regulated process (10)
.
Antibody production in a nonhuman host may result in the presence of nonhuman N-glycans that may induce an immune response, especially when repeated or long-term system applications are considered. In plants, ß(1,2)-xylose and (1,3)-fucose may be present in Fc-associated N-glycans (11
, 12)
. Glyco-engineering has been successfully used to modify the N-glycosylation pathways in different systems including CHO cells (13)
, insect cells (14)
, yeasts (15)
, and plants (16
, 17)
. Because higher plants are less amenable to genetic manipulation, different cellular targeting has been explored as an alternative strategy to control N-glycosylation. KDEL-mediated retrieval of antibodies from the cis-Golgi has been proposed for preventing addition of the plant-specific ß(1,2)-xylose and (1,3)-fucose residues. In tobacco plants, KDEL-tagged cPIPP (mouse/human chimeric IgG1/
) (18)
and 14D9 (mouse IgG1/) (19)
where efficiently retained while small amounts of complex type N-glycans were found on CB.Hep1 (mouse IgG2b/) (20)
and SO57 (human IgG1/
) (21)
. Thus, the efficiency of this approach appears to be antibody specific and may not always be sufficient to prevent formation of complex type N-glycans.
Expression of human ß (1
, 4)
-galactosyltransferase (GalT) in plants has been found to significantly reduce the presence of ß(1,2)-xylose and even more of (1,3)-fucose (22
, 23)
. However, minor amounts of hybrid N-glycans containing ß(1,2)-xylose were still present. The use of a hybrid enzyme containing the CTS domain of a plant core ß(1,2)-xylosyltransferase led to antibodies that were essentially devoid of plant-specific sugars (24)
. The (1,3)-arm and the core of the resulting hybrid type N-glycans are identical to N-glycans found on human antibodies while the (1,6) arm contains mannose residues found on ER-retained antibodies.
With many different production platforms, antibody mutants, and glycoforms being developed and encountered, sensitive methods for the analysis not only of their structures, but also their functional properties, are required. Effector functions of antibodies can be investigated in detail by surface plasmon resonance using soluble ectodomains of human Fc receptors (9
, 25
26
27
28
29)
. Recently the soluble ectodomain of FcRI (CD64) (31)
has been characterized (25)
. FcRI binds to the Fc part of monomeric human IgG1 and IgG3 with high affinity (Ka=1.2·109 M–1) (30)
. The interactions between human IgG1 and FcRI have been shown to be influenced by the N-glycans attached to Asn297 (5
, 31)
. Aglycosylated IgG1 and IgG3 exhibit significantly reduced binding to FcRI. Binding is also reduced by mutation of Asp265, which contacts the primary GlcNAc. Functional and structural analysis of different glycoforms has revealed a conformational change of the CH2 domain that depends on the structure of the oligosaccharides and affects the binding sites for Fc receptors (32
33
34)
. The structure of the N-glycans attached to the antibody constant domains and the functional consequences are therefore important properties for evaluating a production system and strategy and its derived product for its suitability for therapeutical application.
Here we investigated the production and functional characterization of the human monoclonal antibody 2F5 (35)
in BY-2 suspension cells. The antibody 2F5 recognizes the highly conserved linear epitope ELDKWA of gp41 of HIV-1; it has broad neutralization capacity and is being evaluated in animal and clinical trials for its efficiency in HIV-1 prevention and treatment (36
, 37)
. 2F5, in combination with other potent antibodies, conferred protection against intravenous, intravaginal, or oral challenge with simian-human immunodeficiency virus in adult and neonate rhesus macaques (38
39
40)
. Recent clinical trials in HIV-1-infected humans indicated efficiency in some patients in combination with the potent neutralizing antibodies 2G12 and 4E10 (41
, 42
, and unpublished results).
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Cloning and transformation of the 2F5 expression vector
cDNAs of the heavy and light chains of the 2F5 antibody were isolated together with their existent sequence for the signal peptide from the eukaryotic expression vectors p2F5IgG1 and p2F5LC, respectively (43)
. The sequence coding for a C-terminal ER retention signal (SEKDEL) (44)
was fused to the 3' end of both cDNAs by PCR. Each fragment was cloned in an expression cassette consisting of the duplicated CaMV-35S promoter, the TEV 5'-untranslated region, and the CaMV-35S transcriptional terminator. For coexpression of heavy and light chains, the two cassettes were inserted head-to-tail into the binary vector pTRAk, a derivative of pPAM (GenBank no. AY027531). The tandem expression units were separated by scaffold attachment regions (SAR) of the tobacco RB7 gene (GenBank no. U67919). An expression cassette for DsRed (45)
fused to a plastid transit peptide was cloned behind the antibody chains cassettes to generate the final construct pTRAk-2F5ER-Ds (Fig. 1
).
|
The plasmid was electroporated into Agrobacterium strain GV3101::pMP90RK (46)
. Recombinant Agrobacteria were added to 3 ml of a 3-day-old BY-2 suspension cell culture supplemented with 200 µM acetosyringone to give to a final OD600 of 0.05. After 3 days of coincubation at 26°C, cells were plated onto solid medium containing 100 mg/L kanamycin and 200 mg/L cefotaxime. After 4 wk of incubation at 26°C, kanamycin-resistant calli were screened for DsRed fluorescence and transferred to fresh plates. After 3 wk, the expanded calli were screened for 2F5 accumulation (data not shown) and four lines were used to establish suspension cultures. The suspension with the highest antibody accumulation was used for scale-up.
Antibody extraction and purification
Cells were separated from the culture medium by filtration and stored at –20°C. The frozen cells were thawed in two volumes of extraction buffer (PBS pH 7.4 with 5 mM ETDA and 5 mM ß-mercaptoethanol) and disrupted in a Microfluidizer M110-L (Microfluidics Corporation, Newton, MA, USA). Insoluble material was removed by centrifugation (30 min/4000 g/4°C). The supernatant was adjusted to pH 8, stirred for 2 h at 4°C and centrifuged at 30,000 g for 30 min. The clear supernatant was passed through a paper filter and applied to a disposable column (Bio-Rad, Hercules, CA, USA) packed with 2 ml protein A Ceramic HyperD F (BioSepra, Marlborough, MA, USA) using a flow rate of 2 ml per minute (240 cm/h). Nonspecifically bound material was removed by extensive washing with PBS containing 5 mM EDTA. The antibody was eluted with 100 mM glycine pH 2 containing 10% maltose. Elution fractions were immediately adjusted to pH 4.5 by addition of 1 M unbuffered sodium acetate and analyzed for antibody concentration by Biacore (Piscataway, NJ, USA). Antibody-containing fractions were pooled and dialyzed extensively against 10 mM sodium acetate pH 4.5 containing 10% maltose and 1 mM EDTA. The preparation was centrifuged at 30.000 g to remove residual insoluble material and the protein concentration of the supernatant was determined using the BCA assay (Pierce, Rockford, IL, USA).
SDS-PAGE, EMSA, and immunoblot analysis
Proteins were separated on 4–12% precast Bis-Tris NuPAGETM gels (Invitrogen, Carlsbad, CA, USA) under reducing and nonreducing conditions. The SeeBlue Plus2 marker (Invitrogen) was used as size standard.
The electrophoretic mobility of functional BY22F5 was shifted by adding the peptide ARP7073 (CQNQQEKNEQELLELDKWAS, CFAR, Centralized Facility for AIDS Reagents), which contains the linear epitope recognized by the antibody 2F5. CHO2F5 and CHO2G12 were used as positive and negative controls, respectively. One microgram of each antibody was incubated with or without 1.6 µg of ARP7073 in a total volume of 5 µl for 30 min at 28°C. After adding 5 µl of 2 x loading buffer (200 mM phosphate buffer pH 6.5 containing 10% sucrose), proteins were separated on a 10% continuous PAA gel at 100 V for 3 h at 4°C using 100 mM NaH2PO4 buffer pH 6.5 as gel and running buffer. Proteins were visualized by staining with Coomassie brilliant blue.
For Western blot analysis, proteins were electrotransferred to nitrocellulose membranes (Amersham Biosciences, Arlington Heights, IL, USA). Membranes were blocked for 1 h at room temperature with 5% (w/v) skim milk dissolved in PBS containing 0.05% Tween-20 (PBST). AP-conjugated goat polyclonal specific for human -chain or -chain were used for detection (Sigma-Aldrich, St. Louis, MO, USA). Antibody bands were visualized by incubation with NBT-BCIP.
ELISA
Microtitre plates (High-binding; Greiner Bio-One GmbH, Frickenhausen Germany) were coated overnight at 4°C with 400 ng/ml ARP7073 peptide (NIBSC, Centralized Facility for AIDS Reagents, EU program EVA/AVIP) or polyclonal goat anti-human Fab (Sigma-Aldrich, I-3266) diluted 1:500 in 50 mM carbonate buffer pH 9.6. Wells were blocked with 200 µl of 5% skim milk dissolved in PBST for 1 h at room temperature. Two-fold serial dilutions of CHO2F5 and BY22F5 were applied to the wells, followed by AP-conjugated polyclonal goat anti-human -chain (Sigma-Aldrich, A3187) diluted 1:5000 in PBST. Bound antibodies were detected by incubation with p-nitrophenyl phosphate and the OD405 was read. Steps plates were washed with PBST between incubations.
The binding data was evaluated using Origin 5.0 by determining the reactivity (slope) in the linear range of the dose-response curve by linear regression. The reactivity against polyclonal goat anti-human Fab was used to measure the total antibody concentration, and reactivity against the ARP 7073 peptide was used to measure the active antibody concentration. The relative antigen binding activity of BY22F5 was derived by first dividing the peptide binding reactivity by the total antibody binding reactivity, then normalizing to the ratio obtained for CHO2F5.
Surface plasmon resonance analysis
Binding assays were performed on a Biacore2000 instrument. Coupling to a CM5-rg sensorchip was done following the standard EDC/NHS protocol using 100 µg/ml protein A or protein G in 10 mM sodium acetate buffer pH 4.25 and HBS-EP as running buffer. After immobilization, the surface was conditioned with three 15 s pulses of 100 mM HCl. Immobilization levels were 2.5 kRU for protein A and 950 RU for protein G, and the binding capacity of the surfaces was verified by injection of known amounts of CHO2F5. The fraction of active binding sites was determined by capturing 1146 to 1286 RU of CHO2F5 or BY22F5, followed by injection of 50 µg/ml of peptide ARP7073 using the kinject mode. The assay was performed at 25°C using a flow rate of 30 µl/min. The surface was regenerated by a 15 s injection of 30 mM HCl. A reference cell was used for blank subtraction and the fraction of active binding sites was calculated using Eq. 1. The molecular mass of BY22F5 is slightly higher than that of CHO2F5 due to the presence of the C-terminal SEKDEL sequence. The N-glycans of BY22F5 and CHO2F5 were assumed to have the same mass. Based on published results (18)
, an average mass for the N-glycans of 3200 Da was added to the calculated masses of the polypeptides.
f = Rpeptide/(RmAb · 2 · Mrpeptide/MrmAb)
Mrpeptide = 2431 Da
MrmAb = 153.6 kDa (BY22F5) = 150.8 kDa (CHO2F5)
RmAb = amount of antibody captured to protein A surface
 | (1) |
The recombinant ectodomain of human FcRI (rsCD64) and the tobacco-derived mouse/human chimeric antibody cPIPP have been described before (18
, 25
, 47)
. For analyzing binding to rsCD64, 187–201 RU of recombinant antibody were captured to protein G, followed by injection of rsCD64 at 16 nM. A flow cell with an activated/deactivated surface was used as reference. A buffer injection was used as a second reference and injection of rsCD64 over the protein G surface was used as a third reference. The binding curves were then normalized to 200 RU of captured antibody. The resulting curves were fitted to a simple monovalent interaction model to determine the rsCD64 binding capacity and kinetic and equilibrium constants. The relative activity of BY22F5 is the ratio of the binding capacities (Rmax) determined for BY22F5 and CHO2F5.
HIV neutralization assay
HIV-1 neutralization was assessed in a syncytium inhibition assay that is routinely used for quality control of CHO-derived C2F5 batches. Briefly, ten 2-fold serial dilutions (start concentration: 100 µg/ml) of BY22F5, CHO2F5, and non-neutralizing 3D6 (control) were preincubated with HIV-1 RF at 102.4 TCID50/ml for 1 h at 37°C. AA-2 substrate cells were added at a density of 4 x 105 cells/ml and further incubated for 5 days. Experiments were performed with eight replicates per antibody dilution step. The presence of 
1 syncytium per well after 5 days was scored as positive infection. The 50% inhibiting concentrations (IC50) were calculated by the method of Reed and Muench and are concentrations present during the antibody virus preincubation step (48)
.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Purification and characterization of BY22F5
Purification of BY22F5 was followed on-line during all steps by Biacore on a protein A surface (data not shown). Thereby we confirmed efficient binding to and elution from the protein A matrix. No significant amount of antibody was detected in the flow-through or the wash fractions (data not shown). BY22F5 was enriched >500-fold by a single step protein A chromatography and 2.2 mg were retrieved from 340 g (wet weight) of BY-2 cells, yielding 6.4 mg/kg wet cell weight or 1.8 mg/L suspension culture. The purified antibody was entirely intact and contained only minor amounts of impurities (Fig. 2
A). Neither the heavy chain (Fig. 2B
) nor the light chain (Fig. 2C
) showed degradations. The electrophoretic mobility of the light chain of BY22F5 is slightly lower than that of CHO2F5 because of the C-terminal addition of the SEKDEL signal. The electrophoretic mobilities of the heavy chains do not show a significant difference probably because the size difference due to the SEKDEL signal was not resolved. Nonreducing denaturing SDS-PAGE and immunoblot were performed to investigate the state of assembly of plant- and mammalian-produced 2F5. No bands corresponding to free light or heavy chains were detected (Fig. 3
A), showing that the antibody was efficiently assembled within the ER of BY-2 cells.
|
|
Functional characterization
The antigen and rsCD64 binding properties of BY22F5 and CHO2F5 were compared in detail using different types of experiments.
The electrophoretic mobility shift assay (EMSA) was performed using phosphate buffer at pH 6.5 in order to have near-physiological conditions. Under these conditions, 2F5 migrates only a short distance into the gel. Nevertheless, a small difference in the electrophoretic mobility of free BY22F5 and free CHO2F5 was observed (Fig. 3B
, lanes 1, 3), which is due to the presence of the SEKDEL motif on all chains of the BY22F5. When ARP7073 was present in a 100-fold molar excess, a clear shift was seen and no bands remained at the position of free antibody (Fig. 3B
, lanes 2, 4). In contrast, the band for the CHO2G12 was not affected by the presence of the ARP7073 peptide, since this antibody recognizes a high-mannose carbohydrate cluster epitope in gp120 that is not present in the ARP7073 peptide. The results show that neither 2F5 antibody preparations contained detectable amounts of completely inactive molecules. The resolution and sensitivity of the assay, however, are not high enough to reveal a small fraction of inactive molecules. Moreover, partially inactive molecules (i.e., antibodies with only one active binding site) may not be revealed.
A quantitative comparison of the antigen binding activities of the plant and mammalian-derived antibody preparations was done by ELISA. The total antibody concentration was determined using goat-anti-human Fab as capture reagent. For measuring the active antibody concentrations, the peptide ARP7073 was used for coating. This enables the use of goat anti-human Fc-specific secondary antibody for both setups, minimizing possible error sources and allowing the most direct comparison of the two antibody samples. ELISA reactivities obtained for the plant- and mammalian-expressed 2F5 showed no significant differences (Fig. 4
A). The mean relative activity of BY22F5 was 105% (±16% SD) of CHO2F5, and thus identical within experimental limits (Table 1
).
|
|
Surface plasmon resonance was used to perform a highly accurate quantitative comparison of the antigen binding activities of CHO2F5 and BY22F5. This assay is standard-independent and allows determination of absolute binding activities. For both antibody preparations, five repetitive measurements were performed. The amounts of captured antibody and subsequently bound peptide were determined from the sensorgrams and the fraction of active binding sites was calculated (Fig. 4B
and Table 2
). The absolute peptide binding activity for the CHO-derived 2F5 was 97% (±1.1% SD) and for the BY-2-derived product 89% (±2.3% SD), resulting in a relative reactivity of BY22F5 for single binding sites of 91% (±2.1% SD). Thus, the activity of BY22F5 was slightly lower than that of the CHO2F5 standard.
|
Apart from antigen binding, effector functions mediated by the Fc region are important properties of antibodies intended for medical applications. Here we used the recently described recombinant ectodomain of human FcRI (25)
and compared its interaction with BY22F5 and CHO2F5. The mouse/human antibody cPIPP was used as an additional reference. This antibody also carries the KDEL signal on both antibody chains and the attached N-glycans are of the high-mannose type (18)
. The recombinant antibodies were captured to immobilized protein G, followed by injection of 16 nM of rsCD64. The binding curves were fitted to a monovalent interaction model (Fig. 5
). While NtcPIPP and CHO2F5 exhibited nearly identical binding to rsCD64, slight differences were observed for BY22F5 (Table 3
). The association rate constants were almost identical whereas the dissociation rate constants of CHO2F5 and BY22F5 were slightly different. The resulting KD differed by a factor of 2. The determined binding capacities were 51.6 RU for CHO2F5 and 45.8 RU for BY22F5. This results in a relative rsCD64 binding activity of 89% and agrees with the absolute single binding site antigen activity.
|
|
The HIV neutralization activity of BY22F5 was compared with the CHO2F5 using a syncytium inhibition assay with the laboratory adapted virus HIV-1 RF. The CHO2F5 had an IC50 of 9.6 µg/ml whereas the BY22F5 surprisingly had a 3-fold higher IC50 of 29.2 µg/ml. The control antibody 3D6 had no neutralizing activity (results not shown).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
, 50)
, a C-terminal SEKDEL tag was fused to both antibody chains and resulted in a 5- to 10-fold higher accumulation (data not shown). Even though accumulation of the KDEL-tagged 2F5 was only 1.8 mg/L suspension culture, purification by single-step protein A chromatography was quite efficient. SDS-PAGE under reducing and nonreducing conditions and subsequent immunoblot analysis showed that the purified antibody was intact and was efficiently assembled. BY22F5 displayed a lower electrophoretic mobility than CHO2F5 in native PAA gel electrophoresis at pH 6.5. This is likely due to the addition of the SEKDEL tag to the light and heavy chains, resulting in an increase of the molecular mass of the full-size antibody of 2.8 kDa. The theoretically calculated pI and the net charge at pH 6.5 are lower for BY22F5 (pI=9.15, z=+17.2) than for CHO2F5 (pI=9.48, z=+25.2). Although real values may deviate from this calculation, these differences seem to be sufficient to explain the lower electrophoretic mobility of BY22F5. The oligosaccharides attached to Asn297 are not likely to be responsible because they are mainly neutral N-glycans with a similar averaged size (Fig. 6
). Upon addition of excess amounts of the peptide ARP7073, mobilities decreased due to a combination of increased size and decreased net charge of the complexed antibodies. The shift was clearly visible and complete, showing that both preparations did not contain significant amounts of totally inactive molecules. The antigen binding activities of BY22F5 and CHO2F5 were then compared by ELISA. The relative antigen binding activity for BY22F5 was 105% (±16% SD) of the CHO2F5 standard. Thus, the antigen binding activity of BY22F5 and CHO2F5 were indistinguishable by ELISA and EMSA. However, the resolution and sensitivity of these assays are not high, and low amounts of totally inactive or partially inactive antibody would not be revealed. The antigen binding activity determined by ELISA had a relatively high SD and the assay performance depends on suitable standard and detection reagents. Moreover, full-size antibodies are bivalent molecules, and it is usually difficult to discriminate between molecules with one or two active binding sites. 2F5 binds to its antigen with high affinity and therefore it is likely that, in ELISA, antibody molecules with only one active binding site will also bind efficiently to the coated peptide ARP7073. Thus, for characterization of antibody quality, a direct and precise determination of the single binding sites activity is highly desirable. This is particularly important for molecules intended for therapeutical application.
|
Using a high-resolution biosensor assay with sufficient accuracy and reproducibility, we were able to reveal a small but significantly lower activity of BY22F5. The absolute single binding site activity determined for BY22F5 was 89.2%, with an SD of only 2.3%. The CHO2F5 had an absolute single binding site activity of 97.7% with an even smaller standard variation of just 0.94%. Except for the lower saturation signal of the BY22F5 antibody, no obvious differences have been observed. In addition to higher precision, biosensor analysis can be automated; it is easier to perform and evaluate, and is independent of suitable standard and detection reagents. The resolution may be even further increased by using analytes with a higher molecular mass to improve the signal-to-noise ratio. Taken together, these results clearly demonstrate the usefulness of Biacore analysis for the functional characterization of recombinant antibodies.
The analysis is not restricted to antigen binding but can be extended to other parts of the antibody molecule by employing suitable ligands. This was shown here using the ectodomain of human FcRI as an example, but other antibody binding molecules may be equally suited to address more functional aspects. Soluble Fc receptors are particularly interesting ligands, as they are sensitive to the N-glycan structures (7
, 31
, 51)
and are more likely to reveal differences that have a therapeutical impact (52
53
54)
. The availability of the recombinant ectodomain of human FcRI (rsCD64) enabled us to perform binding experiments to investigate the functionality of the antibody constant region comprising the effector functions. The kinetic analysis of rsCD64 binding to BY22F5 and CHO2F5 revealed minor differences: BY22F5 exhibited a slightly faster dissociation rate constant and an 
2-fold higher equilibrium dissociation constant. Moreover, BY22F5 exhibited only 89% of the binding capacity of CHO2F5 for rsCD64, which is consistent with the lower single binding site activity observed earlier. The reduced rsCD64 binding capacity and single antigen binding site activity may be explained by the residual presence of impurities, degradation products, or denatured antibody molecules. In contrast, the different dissociation kinetics, though only small, is likely due to structural differences between BY22F5 and CHO2F5.
BY22F5 and CHO2F5 harbor different types of N-glycans. Whereas CHO2F5 is produced by a mammalian cell line and is secreted into the medium, BY22F5 is retained within the ER of a plant cell by means of a KDEL tag. As a consequence, BY22F5 is expected to harbor only high-mannose type N-glycans (Fig. 6A
), similar to those described for NtcPIPP. This antibody has the same isotype (IgG1/) and also contains KDEL tags at the C-terminal ends of both the light and heavy chain and is efficiently retained in the ER of transgenic tobacco plants (18)
. Since no 2F5 antibody has been detected in the medium supernatant, we conclude that the antibody 2F5 is also efficiently retained within the ER of the tobacco BY2 suspension cells and therefore carries only high-mannose type N-glycans. In contrast, the CHO2F5 carries mammalian complex type glycans similar to those found on human serum IgG (Fig. 6B
) (see refs. 3
, 11
for a comparison of structures). Despite these differences, the binding kinetics of NtcPIPP derived from transgenic tobacco plants and 2F5 derived from CHO cells is identical, and binding of BY22F5 is only marginally different. We therefore conclude that high-mannose type and mammalian secretory type N-glycans support the same or at least a similar conformation of the Fc region. Antibodies produced in the CHO Lec-1 cell line carry only Man5 structures and exhibit a reduced binding to FcRI on U937 cells (5)
. The majority of the N-glycans of NtcPIPP are Man7 and Man8, and there is a possibility that a higher abundance Man5 or Man6 in BY22F5 may affect the binding to rsCD64.
The faster dissociation may also be due to slight interference of the SEKDEL signal at the C-terminal end of the BY22F5 light chains that may come into proximity to the bound rsCD64 and affect its dissociation behavior. The light chain of NtcPIPP contains a longer spacer sequence (VDGGGGSAAARGSEKDEL) and therefore may exert a different influence.
Despite the near identical antigen and FcRI binding activities BY22F5 exhibited a 3-fold lower HIV neutralization capacity compared with CHO2F5. This difference is much larger than expected based on antigen and rsCD64 binding activities.
Neutralization by 2F5 might be particularly sensitive to small structural differences outside the paratope since the 2F5 epitope is located in the membrane proximal external region (MPER) of gp41 (36)
. Reduction of the neutralization capacity has been observed after class switching CHO2F5-IgG to polymeric IgA and IgM (55)
, because access to the buried epitope is probably restricted by the size of the larger polymeric antibodies. CHO2F5 and BY22F5 differ in their structure by the presence of the C-terminal SEKDEL tags and their N-glycans. We cannot conclude whether these structural differences or other factors are responsible for the reduced HIV neutralization activity of BY22F5 in the syncytium inhibition assay. More detailed investigations are required to analyze whether downstream processing issues (e.g., residual plant-derived impurities, presence of antibody aggregates, or currently undetected protein modifications) might be responsible for the discrepancy noted between in vitro antigen binding activity and HIV neutralization activity. It will be essential to resolve this issue, as higher dose requirements are not acceptable for therapeutical applications either in terms of production costs or treatment efficiency. Transgenic plant cell suspensions are easy to generate, handle, and cultivate and offer a high level of experimental control. Establishing homogenous transgenic cell lines is significantly faster than generation of transgenic plant lines, which may require several generations. Compared with the production of recombinant proteins in transgenic plants, the generation and initial upscaling of transgenic suspension cell lines is considerably faster and requires fewer resources (56)
. The cost of protein production in plant cell culture may be lower than in mammalian cell culture, and suspension cultures allows sterile containment in compliance with GMP conditions (1)
. In this respect the recent approval of a veterinary vaccine developed by Dow AgroSciences LLC by the U.S. Department of Agriculture Center for Veterinary Biologics represents an important milestone (57)
.
Taken together, our results show that plant suspension cultures can be efficiently employed to produce human antibodies. Combined with powerful analytical methods, such as biosensor technology, plant suspension cultures represent an important tool for the characterization and development of plant-derived biologics.
|
ACKNOWLEDGMENTS |
|---|
Received for publication September 21, 2006. Accepted for publication January 18, 2007.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  K. Ramessar, T. Rademacher, M. Sack, J. Stadlmann, D. Platis, G. Stiegler, N. Labrou, F. Altmann, J. Ma, E. Stoger, et al. Cost-effective production of a vaginal protein microbicide to prevent HIV transmission PNAS, March 11, 2008; 105(10): 3727 - 3732. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||
