Abstract
IgG is a glycoprotein with an N-linked carbohydrate structure attached to the CH2 domain of each of its heavy chains. In addition, the variable regions of IgG often contain potential N-linked carbohydrate addition sequences that frequently result in the attachment of V region carbohydrate. Nonetheless, the precise role of this V region glycan remains unclear. Studies from our laboratory have shown that a naturally occurring somatic mutant of an anti-dextran Ab that results in a carbohydrate addition site at Asn58 of the VH has carbohydrate in the complementarity-determining region 2 (CDR2) of the VH, and the presence of carbohydrate leads to an increase in affinity. However, carbohydrate attached to nearby positions within CDR2 had variable affects on affinity. In the present work we have extended these studies by adding carbohydrate addition sites close to or within all the CDRs of the same anti-dextran Ab. We find that carbohydrate is attached to all the novel addition sites, but the extent of glycosylation varies with the position of the site. In addition, we find that the position of the variable region carbohydrate influences some functional properties of the Ab, including those usually associated with the V region such as affinity for Ag as well as other characteristics typically attributed to the Fc such as half-life and organ targeting. These studies suggest that modification of variable region glycosylation provides an alternate strategy for manipulating the functional attributes of the Ab molecule and may shed light on how changes in carbohydrate structure affect protein conformation.
Antibodies are glycoproteins that have at least one N-linked carbohydrate added to the constant regions of the heavy chains. Many Abs also have V region-associated carbohydrate. In fact, it has been calculated that human serum IgG has, on the average, 2.8 N-glycoside-type sugar chains/protein molecule 1 . Two of these carbohydrate moieties represent the conserved N-linked carbohydrate in the Fc region, while the remainder reflect V region glycosylation. Analysis of human myeloma proteins and Abs of other species has indeed verified the frequent presence of V region carbohydrate 2, 3, 4, 5, 6, 7 . About 18% of the VH3 sequences in Kabat’s database 8 contain a potential N-linked glycosylation site, Asn-X-Ser/Thr 9 . The presence of this site, however, does not guarantee that it is used for carbohydrate attachment 10, 11 . Although asymmetric utilization of the V region glycosylation sites has been reported 6, 12, 13 , both V regions can be glycosylated 7 .
The role of the V region carbohydrate remains unclear, but several studies suggest that its presence may influence affinity and/or specificity 4, 9, 14 . Previous studies of anti-(1→6) dextran Abs showed that carbohydrate addition sites present within the CDR2 of the VH gene 19.22.1 were actually used, and N-linked carbohydrate attached 7, 15 . This was true both for a naturally occurring somatic mutation at position 60 (Asn→Thr) resulting in glycosylation of Asn58 in the VH CDR2 (hybridoma 14.6b.1) as well as for glycosylation sites introduced at residues 54 and 60 within the CDR2 using site-directed mutagenesis. Interestingly, glycosylation of Asn58 of the VH increased the affinity of the Ab for Ag approximately 10-fold, while carbohydrate at Asn60 of the VH only increased the affinity 3-fold, and carbohydrate at VH Asn54 actually blocked the binding of Ag 15 . In the anti-dextran Ab one of the surprising observations was that the carbohydrate added at position 60 remains in the high mannose form, while position 58, only two amino acids apart in sequence, contains carbohydrate that is processed to complex glycans. While differential accessibility has been proposed as the cause of differential processing, it seems likely that the protein itself plays an important role in determining the specificity of processing.
The structure of the V region carbohydrate is frequently different from that of the constant region carbohydrate, and the structure of the V region carbohydrate can be associated with differences in binding specificity and affinity 16, 17 . For the anti-dextrans, carbohydrate attached at Asn54 and Asn58 was a biantennary complex structure containing Galα1→3Gal as a nonreducing terminus, while the carbohydrate attached to the Fc of the same Ab lacked Galα1→3Gal at its terminus 18 . Recent evidence suggests that carbohydrate can influence characteristics such as organ localization, clearance rates, and receptor binding 19, 20 . Variation in glycoform structure has been associated with some disease states 21, 22 .
In the present work we have extended our studies of V region carbohydrate by adding carbohydrate addition sites close to or within CDR1 and CDR3 of the VH as well as the CDR1, CDR2, and CDR3 of the VL of the same anti-dextran Ab. The sites chosen were predicted to be on the surface of the native protein based on the molecular model of the binding site of the 19.22.1 and were introduced as single amino acid changes to minimize the effect of amino acid substitution on the binding site 23 . We joined the mutant VL and VH to human IgG1 and κ constant regions and expressed different combinations of the mutations as chimeric Abs in the cell line Sp2/0.
We found that the position of the V region carbohydrate can influence many of the functional properties of the Ab. These include not only properties usually associated with the V region, such as affinity for Ag, but other characteristics as well, such as half-life and organ targeting, and may shed light on how changes in carbohydrate structure affect protein conformation and create potentially pathogenic molecules. These studies suggest that modification of V region glycosylation provides an alternate strategy for manipulating the functional attributes of the Ab molecule.
Materials and Methods
Mutagenesis
The heavy chain sugar addition sites were introduced by site-directed mutagenesis with a modified version of the Zoller and Smith method using as template the cDNA from the 19.22.1 VH 7 . The VHCDR1 oligonucleotide (ACA→AAC (Thr-Asn)) was 5′-ACTGGCTACAACTTCAGTAGC, while the VHCDR3 oligonucleotide (TAC→AAC (Tyr-Asn)) was 5′-GGCATTACAACGGTAGTAGC (mutagenic bases are in bold and the Asn is underlined). Carbohydrate addition sites were introduced into the VLCDRs by overlapping PCR using standard or touch-down PCR 24 using the cDNA from the 19.22.1 VL as template 7 . To amplify the V region from the plasmid DNA and introduce a complete leader region and an EcoRV (bold) site for cloning, a 5′ primer that includes a ribosome recognition site (underlined) 25, 26 and a start codon and spans nine amino acids of the leader sequence was used 27 : 5′-GGGATATCCACCATGGATTTTCAAGTGCAGATTTTCAG. The 3′ antisense external primer hybridizes to the Jκ4 region in framework region 4 and contains a splicing signal (underlined) and a SalI site (bold) for cloning: 5′-AGCGTCGACTTACGTTTTATTTCCAGCTGGCCC. The mutagenic primers used were the sense and antisense oligomers of the following sequences: VL CDR1 Ser28→Asn (AGT-AAT), 5′-GCCAGCTCAAATGTAAGTTAC; VL CDR2 Asp50→Asn (GAC-AAC), 5′-GATGGATTTATACCACATCCAAAC; and VLCDR3 Trp91→Asn (TGG-AAC), 5′-ACTGCCAGCAGAACAGTAGTAACCCG (the mutagenic bases are in bold and the Asn is underlined). The PCR products were cloned into a TA plasmid (Invitrogen, Carlsbad, CA). All V region sequences were confirmed using the Sequenase 2.0 kit (U.S. Biochemical, Cleveland, OH) with the protocol described by the manufacturer.
Cloning in expression vectors and transfection
The VH genes containing the CDR1 and CDR3 glycosylation sites were excised from M13 RF with EcoRI and were used to replace the V region in the human γ1 pSV2ΔHgptHuG1 expression vector 28 . The expression vector with the VH glycosylated in the CDR2 at Asn58 has been previously described 15 . The wild-type and the CDR mutant VL V regions were cloned as EcoRV-SalI fragments into the PCR expression vector pAG4622 27 .
Different combinations of the wild-type and mutant light and heavy chains were cotransfected into Sp2/0-Ag-14 cells by electroporation. Ten micrograms of expression vector prepared using Qiagen Maxi-Plasmid Prep (Qiagen, Valencia, CA) was linearized with PvuI and transfected into 107 cells using a Gene Pulser (Bio-Rad, Hercules, CA) and was plated in Falcon 96-well tissue culture plates (Becton Dickinson, Mountain View, CA) at 125 μl/well (2 × 106 cells/plate) with Iscove’s modified Dulbecco’s medium (IMDM) supplemented with 10% calf serum, 1% Nystatin suspension (Life Technologies, Grand Island, NY), and 100 μg/ml gentamicin sulfate (Sigma, St. Louis, MO). Selection was performed as previously described 27 with medium containing 3 μg/ml mycophenolic acid (plus hypoxanthine at 125 μg/ml and xanthine at 7.5 μg/ml). Surviving clones were screened for Ab production by ELISA with anti-human γ-coated plates and detected with anti-human κ alkaline phosphatase-conjugated Ab. Positive clones were subcloned by limiting dilution, then screened again, and the ones selected were maintained in IMDM containing 5% calf serum.
Ab analysis
The Abs were analyzed by metabolic labeling and immunoprecipitation. Between 3 and 5 × 106 cells were washed, resuspended in 1 ml of labeling medium (high glucose DMEM deficient in methionine: Life Technologies, Grand Island, NY) containing 25 μCi of [35S]methionine (Amersham, Arlington Heights, IL), and allowed to incorporate label for 3–4 h or overnight with the addition of 1% FCS or α-calf serum at 37°C under tissue culture conditions. The Abs were immunoprecipitated with 2.5 μl of a rabbit anti-human IgG Fc and anti-human Fab polyclonal antiserum and a 10% suspension of IgGSorb (Enzyme Center, Woburn, MA). The precipitated labeled Ab was then analyzed by SDS-PAGE on 5% sodium phosphate-buffered polyacrylamide gels. To examine heavy and light chains separately, the labeled sample was reduced by treatment with 0.15 M 2-ME at 37°C for 30 min and analyzed on 12.5% Tris-glycine buffered polyacrylamide gels.
Tunicamycin treatment of transfectomas
To prevent N-linked glycosylation of Abs during synthesis, transfectants were grown in the presence of tunicamycin before biosynthetic labeling 29 . Between 3 and 5 × 106 cells were washed twice with PBS and preincubated in 1 ml of tissue culture medium (IMDM with 5% calf serum) in the presence of 8 μg/ml of tunicamycin (Boehringer Mannheim, Indianapolis, IN) for 3–5 h under tissue culture conditions. Then the cells were metabolically labeled as described above in 1 ml of labeling medium containing 25 μCi of [35S]methionine and 8 μg/ml of tunicamycin. Immune precipitations were performed, and the Abs were analyzed as before.
Treatment with endoglycosidase H (Endo-H)
To determine the degree of carbohydrate processing, we digested proteins with Endo-H glycosidase, which cleaves oligomannose-type oligosaccharides. Labeled immunoprecipitated Ab bound to IgGsorb was resuspended in 100 μl of reaction buffer (0.1 M citrate containing 1 μl of 2-ME and 1 μl of 250 mM PMSF), and 10 μl of Endo-H was added (Boehringer Mannheim). The digestion was allowed to proceed for 18–24 h at 37°C. After boiling for 2 min the Abs were analyzed by PAGE under reducing conditions in a 12.5% Tris-glycine gel. A control Ab similarly treated but lacking Endo-H was included for each protein.
Purification
Chimeric Abs were purified from 2–4 l of culture supernatant containing 1% calf serum by protein A-Sepharose affinity chromatography following the protocol suggested by the manufacturer (Sigma). They were then dialyzed against 2–4 l of dialysis buffer (50 mM Tris and 150 mM NaCl, pH 7.8) and concentrated using Millipore Ultra free-15 filters (Millipore Corp., Bedford, MA), and the concentration of the Ab was determined by a bicinchoninic acid protein assay (Pierce, Rockford, IL). Purified Abs were then visualized by SDS-PAGE. The yield of the Abs varied from 1–20 mg depending on the transfectant cell line.
Quantitation of affinity
Tissue culture supernatants were harvested from overnight cultures of 3 × 106 cells. To normalize the concentrations of all Abs tested, the supernatants were diluted to yield the same signal as that determined by an anti-IgG ELISA. The tissue culture supernatants were used to determine the apparent association constant of the mutant Abs to dextran B512 using the method of Nieto et al. 30 . Flat-bottom Immulon I microtiter plates (Dynatech, Chantilly, VA) were coated with 100 μl of a 20 μg/ml solution of the Ag dextran B512 in borate-buffered saline. The plates were allowed to dry overnight at 37°C, washed with PBS, and blocked with 3% BSA. Dilutions ranging from 0–20 μg/ml of soluble Ag (dextran B512) in borate-buffered saline containing 1% BSA were mixed with Ab supernatant, and 50 μl of the mix was added to the plate in triplicate and allowed to reach equilibrium overnight at 4°C. Binding was detected with an anti-human κ Ab conjugated to alkaline phosphatase or horseradish peroxidase and the appropriate substrate. Dextran B512 (200 μg/ml) was also added to one set of triplicates to determine background. Color intensity was read at 410 or 480 nm using a Dynatech MR700 plate reader. The apparent binding constant (aka; grams per milliliter) was calculated from the amount of ligand (free dextran) necessary for 50% inhibition of Ab binding. The assay was performed three or four times, and averages and SDs were calculated to obtain the apparent binding affinity constant.
The Abs that had no detectable binding were tested in the aglycosylated forms after treating the cell cultures with tunicamycin followed by a Con A-Sepharose incubation to remove all possible glycosylated proteins. One milliliter of Con A-Sepharose (Sigma) was washed twice with PBS, and 100 μl of a 50% slurry was added to 10 ml of supernatant obtained from cells cultured overnight in the presence of tunicamycin. After an overnight incubation at 4°C the Con A-Sepharose was removed by centrifugation, and the supernatant was recovered. After a second treatment, the supernatants were used for the affinity assay.
An IAsys optical biosensor (Fisons Applied Sensor Technology, Cambridge, U.K.) was also used to determine the on and off binding rates (ka, kd) and to calculate the affinity (Kd, molar concentrations−1) of the anti-dextrans for the carboxymethylated dextran matrix of commercially available cuvettes (Fisons Applied Sensor Technology). A cuvette was activated and blocked as recommended by the manufacturer. After washing with PBS-Tween (0.05%) different concentrations of purified Abs were added to the cuvette at room temperature in 200 μl of PBS-Tween, and association was measured for 10 min. After washing the cuvette, dissociation was followed for 7 min. Complete dissociation and regeneration of the surface were obtained by a quick wash with 10 mM HCl. Using the Fastfit program (Fisons Applied Sensor Technology), the apparent association and dissociation rates were calculated for each concentration of ligand, and the Kd was calculated as kd/ka.
Half-life
Four- to eight-week-old female BALB/c mice (Taconic Farms, Germantown, NY) were given 0.1 mg/ml potassium iodide in their drinking water at least 4 days before and throughout the experiment to saturate the thyroid iodine receptors. Ten micrograms of Abs purified by affinity chromatography were radioiodinated with 125I using Pierce Iodobeads. Separation of free 125I from the labeled protein was achieved by size exclusion chromatography on a 10-ml G-50 column in PBS, pH 7.2–7.4. Two hundred microliters of the labeled protein (5 × 105 cpm) was injected i.p. into five mice for each Ab. Whole body radioactivity was then determined at various time points for up to 500 h using a sodium iodide crystal with a well sufficiently large to accommodate a mouse (W. B. Johnson & Associates, Montville, NJ).
For each protein analyzed, values obtained from at least three mice were averaged and used for subsequent analysis. Half-life (β-phase; βt1/2) was calculated using an exponential regression analysis of the data obtained after 24 h.
Biodistribution and immune response
Blood was obtained 48 h following injection by subaxillary bleeding of mice rendered unconscious by exposure to ether. The radioactive Ab was immunoprecipitated from the serum using protein A-Sepharose and was analyzed by SDS-PAGE and autoradiography. The biodistribution of radiolabeled protein was determined by dissecting out the relevant organs and then determining the radioactivity left in the organs using the whole body counter. An average weight was calculated for each organ from the values obtained from 22 mice. Data were corrected for the weight and blood volume of the excised organ and were plotted. The nature of the immune response mounted against the injected recombinant Abs by some mice was determined by an ELISA. Immulon II plates were coated overnight with 1 μg/ml of goat anti-human IgG (Zymed, South San Francisco, CA) followed by 3-h incubation at room temperature with 100 μl of 1 μg/ml dilutions of the different anti-dextran mutant Abs or with an anti-dansyl IgG1 chimeric. For each protein tested sera from two mice were pooled and diluted 1/500 in PBS-Tween; 100 μl was added to each well and allowed to bind overnight at 4°C. After washing, bound Ab was detected with a rat anti-mouse IgG-alkaline phosphatase Ab (Sigma).
Results
The hybridoma 19.22.1 was originally isolated as a clone producing IgM-κ antibodies specific for anti-(1→6) dextran 31 . The hybridoma 14.6b.1, isolated at the same time, exhibits a 10-fold higher affinity for anti-(1→6) dextran and differs at only a single amino acid in VH, where Asn at position 60 was changed to Thr, presumably as a result of somatic mutation and selection. Asn58 within 14.6b.1 then constitutes a potential N-glycosylation site, and studies showed that the increased affinity of 14.6b.1 for dextran resulted from the presence of carbohydrate at this position 31 . Subsequent studies with these V regions showed that novel glycosylation sites introduced by site-directed mutagenesis at positions 54 and 60 within the CDR2 of the VH gene had N-linked carbohydrate attached when the heavy chains were expressed by transfection into a cell line producing the murine anti-dextran light chain 15 . Despite the close proximity of the addition sites, their presence had different affects on the affinity of the Ab for Ag.
Production of the mutant Abs
To extend our investigation of the role of V region carbohydrate, site-directed mutagenesis has now been used to generate carbohydrate addition sites within or near the other two CDRs of the heavy chain and all three CDRs of the light chain. We selected the positions at which the amino acid changes were introduced using the molecular model of the 19.22.1 V region 23 as a guide (Fig. 1⇓). Residues predicted to be on the surface with a serine two positions downstream in the amino acid sequence were chosen to be mutated to an Asn, thus creating a canonical N-linked carbohydrate addition sequence Asn-X-Ser/Thr (Table I⇓). While it is generally accepted that the X amino acid cannot be a proline, there are no other known constraints on this sequence. To minimize the possible deleterious effects of an amino acid change on binding, only one amino acid was mutated in each case.
Positions of the asparagines introduced to create glycosylation addition sites. Shown is a molecular model of 19.1.2 Fv, which is similar to the combining site of 14.6b.1, the Ab used for the present studies (23). The positions at which the amino acid mutations were introduced are labeled in black. The VL is dark gray, and the VH is light gray. The graphic was produced using the program MacImdad (Molecular applications group, Stanford University, Stanford, CA).
CDR sequences and position of the mutations introduced to create carbohydrate addition sites
The anti-dextran V regions were expressed with human γ1, κ constant regions in the non-Ig-expressing mouse myeloma cell line Sp2/0-Ag14. Various combinations of the wild-type and mutant V regions were transfected to generate Abs potentially containing carbohydrate at different positions in the hypervariable loops of the VH, the VL, or both. All transfections resulted in secretion of proteins that could be detected by ELISA. To determine whether the new glycan addition sites in the other two VH CDRs and the VL CDRs were indeed used and to assess whether the presence of carbohydrate interferes with either the assembly or the secretion of the Ab, transfectants were biosynthetically labeled with [35S]methionine in the presence or the absence of tunicamycin, and the Ab was immunoprecipitated from the secretions and analyzed by SDS-PAGE. All the mutant Abs were secreted as fully assembled H2L2 molecules (Fig. 2⇓A), and all showed increased electrophoretic mobility after the transfectants were grown in tunicamycin to impede attachment of carbohydrate.
PAGE of biosynthetically labeled Abs. Transfectants were grown overnight in the presence of [35S]methionine, and the radiolabeled Ig was immunoprecipitated with rabbit anti-Fc and anti-Fab and protein A-Sepharose. A, Five percent phosphate-buffered gel showing assembly of Abs. B, Twelve percent Tris glycine-buffered gel of reduced Abs showing the light and heavy chains. +/−, treatment of transfectants with tunicamycin to prevent carbohydrate attachment.
Analysis of glycosylation
To further characterize the Abs and to determine whether sugar was present at the introduced glycosylation positions in VH or VL, the immunoprecipitated Ig was analyzed by reducing SDS-PAGE, thus separating the heavy from the light chain. The shift in size of the control Ab, which has no glycosylation sites in the V regions (VH0VL0), following tunicamycin treatment reflects the presence of the carbohydrate within CH2 (Fig. 2⇑B). All the heavy and light chains that contain an engineered glycosylation addition site in the variable region showed a further shift in size upon treatment with tunicamycin, suggesting that carbohydrate is actually attached to these sites and that the sugar can be added to both chains when the sites are available. Glycosylation sites with Asn at VH58 and VL91 appeared to always be fully glycosylated. However, there seemed to be incomplete utilization of the other sites. For VH28VL0 and VH97VL0, a portion of the secreted heavy chains seemed to contain only one carbohydrate. VH0VL50, with only V region carbohydrate in the L chain, showed incomplete VL glycosylation. All Abs with a glycosylation site at VL28 showed incomplete utilization of this site when it was present as the sole Fv carbohydrate in VH0VL28 or when it was expressed in combination with a carbohydrate containing VH in VH28VL28, VH58VL28, and VH97VL28. The glycosylated VL28 in these Abs resolves as a compact doublet that may represent two glycoforms. VH58VL28 is particularly noteworthy, in that while the heavy chain appears to be completely glycosylated, the light chain migrates as three bands, with the unusual slowly migrating band presumably reflecting the attachment of different carbohydrate structures. In all cases the aglycosylated heavy and light chains recovered after treatment of the cells with tunicamycin show identical mobility confirming that carbohydrate is responsible for the observed heterogeneity (Fig. 2⇑B).
To determine whether any of the recombinant Abs have partially processed sugars attached, immunoprecipitated labeled Ab was treated with Endo-H glycosidase, which is specific for terminal oligomannose glycans. Abs that were subjected to the same reagents and incubation treatments but without Endo-H were included for comparison. An anti-dextran Ab, glycosylated at position 60 with high mannose sugars and sensitive to Endo-H, was used as a control. Analysis of all Abs by reducing SDS-PAGE revealed no shift in size in any of the heavy or light chains, suggesting that none of the Abs contain high mannose oligosaccharides (data not shown). However, the structure of the attached carbohydrate remains to be determined.
Determination of the apparent binding constants and affinity
To assess the impact of carbohydrate present at different positions within the combining site on the affinity of the anti-dextran Abs, we determined their apparent association constants using an inhibition ELISA assay as previously described 15, 30 and compared them to that of an anti-dextran lacking carbohydrate in the V region produced in the same system. Supernatant harvested from overnight cultures of 106 cells was tested for binding to a dextran-coated plate in the presence of different concentrations of free dextran B512. The reciprocal value of the concentration of free dextran required to prevent 50% of binding is the apparent binding constant, which is expressed in grams per milliliter. The Abs that showed no binding were also tested in the aglycosylated form after production in tunicamycin-treated cells and incubation with Con A-Sepharose to remove any remaining glycosylated species.
The Ab with an engineered Asn at VH58, VH58VL0, equivalent to mAb 14.6b.1, had a 6.5-fold increase in the apparent association constant compared with that of the Ab lacking V region carbohydrate (VH0VL0; Table II⇓). The presence of carbohydrate at VL28 always had a negative influence on binding. When Asn28 VL was the only carbohydrate in the binding site, VH0VL28 bound Ag, albeit with a 2.5-fold reduced apparent affinity compared with Ab lacking V region carbohydrate VH0VL0. When the heavy chain with carbohydrate at Asn58 was paired with the light chain Asn28, the resulting Ab VH58VL28 showed a decrease in affinity compared with that of VH58VL0, while VL28 in combination with carbohydrate in the CDR1 VH28 or CDR3 VH97 resulted in Abs for which no detectable Ag binding was seen. Interestingly, VH28VL28 and VH97VL28 did bind Ag when produced in the presence of tunicamycin; however, their association constants were 2.5- and 1.6-fold lower, respectively, than that of wild-type VH0VL0, suggesting that the amino acid changes introduced influence the interaction with Ag. In contrast, VH0VL50 and VH0VL91 showed no binding even after they were produced in the presence of tunicamycin to avoid glycosylation, suggesting that the change in amino acid to generate the Asn is responsible for the lack of binding.
Apparent binding constants (aka) and affinity constants (KD) of the anti-dextran Abs
Glycosylation at Asn97 in the VH CDR3 appeared to have no influence on binding, as VH97VL0 had the same apparent binding constant as VH0VL0. This also confirms that it is the carbohydrate in the VL that has a negative contribution leading to lack of binding by VH97VL28. Carbohydrate at Asn28 of the CDR1 of the VH was not well tolerated, since VH28VL0 had a 2.7-fold decrease in affinity, and VH28VL28 showed no binding.
An IAsys biosensor (Fisons Applied Sensor Technology, Cambridge, U.K.) was used to determine the Kd (molar concentration−1) of the anti-dextran Abs through a direct measurement. A cuvette with a carboxymethylated dextran surface was used as the Ag, and different concentrations of Abs were tested for binding. Only the two Abs glycosylated in the VH CDR2, VH58VL0 and VH58VL28, had a high enough affinity to be detected by the biosensor (Fig. 3⇓ and Table II⇑). At the same concentration, VH58VL0 showed more extensive binding to the dextran cuvette than VH58VL28, consistent with its higher aka. The remaining Abs did not show detectable binding to the dextran cuvette.
IAsys sensogram of the high affinity anti-dextran Abs. Different concentrations of Ab were allowed to bind to a carboxymethylated dextran cuvette surface, and dissociation was observed after washing. Only representative concentrations are shown to demonstrate the differences in binding between the two mutant Abs.
In vivo behavior
The serum half-life in BALB/c mice was obtained by injecting radioiodinated Ab into the peritoneum of 5- to 7-wk-old female mice and counting the mice using a whole body counter to determine the residual radioactivity. For all Abs except VH58VL28, 40–50% rapidly cleared during the first 24 h; during the same time 65% of the injected dose of VH58VL28 cleared. After 24 h the remaining radioactivity was cleared at a slower rate, with 15% of the injected dose of VH58VL28 and about 30–40% of the other proteins remaining in the mice after 200 h. The calculated βt1/2 for the majority of anti-dextran Abs was 7–8 days; VH58VL28 cleared more rapidly, with a βt1/2 of 4 days, and VH0VL50 and VH0VL91 cleared more slowly, with a βt1/2 of 10 days (Fig. 4⇓). Abs recovered from the serum of two mice sacrificed at 48 h postinjection and analyzed by SDS-PAGE appeared intact.
In vivo clearance of the glycosylated Abs. Radioiodinated Ab was injected i.p. in BALB/c mice, and the disappearance of whole body radioactivity was monitored over time. The βt1/2 was calculated starting 24 h postinjection.
In some mice injected with VH28VL0, VH28VL28, VH58VL28, VH58VL0, and VH97VL28, all Abs with glycosylation in VH, an immune response was observed after 200 h as indicated by a sudden increase in the rate of clearance of the Ab (not shown). None of the mice injected with Abs glycosylated only in the light chain V region or the Ab devoid of V region carbohydrate showed evidence of an immune response even 400 h post injection. To determine the specificity of this immune response, an ELISA was performed using anti-dextrans with V region glycosylation at different positions and an anti dansyl-IgG1 chimeric Ab bound to a plate through an anti-human Fc region Ab. Sera from mice injected with VH58VL0, VH58VL28, VH28VL28 or VH0VL0 collected 400 h postinjection diluted 1/500 were added to the Ag-coated plates. Bound murine Abs were detected by an alkaline phosphatase-linked rat anti-mouse IgG. The immune response to VH58VL0 appeared to be mostly specific for the constant regions, since equivalent reactivity was seen with the chimeric anti-dansyl and anti-dextran Abs. Much of the immune response to VH58VL28 was directed at the murine variable regions and did not appear to be directed against epitopes that were strongly influenced by glycosylation. In contrast, much of the immune response against VH28VL28 appeared directed against an epitope dependent on the site of V region glycosylation. Sera from mice injected with anti-dextran lacking V region carbohydrate showed no reactivity with the constant or V region, consistent with the failure to see an acceleration in clearance (Fig. 5⇓).
Immune response ELISA. Sera recovered after 400 h from mice injected with VH58VL0, VH58VL28, VH28VL28, and VH0VL0 were incubated with plates coated with anti-dextrans with different sites of V region glycosylation or an anti-dansyl chimeric IgG1. Bound murine Ab was detected using AP-labeled rat anti-mouse IgG.
The biodistribution in mice of the injected Abs was determined by removing relevant organs from two mice for each protein 48 h after injection and determining the residual radioactivity. The radioactivity present in each organ was adjusted for the weight and blood volume, and the distribution of the counts in each mouse was compared for each protein. Only the organs that showed a significant accumulation of counts or targeting are shown in Fig. 6⇓. The position of the V region carbohydrate was found to influence the in vivo distribution of the Ab. For VH58VL28, which showed the most rapid clearance, there was less in the blood and more in the liver and kidney. Less VH97VL0 was also present in the blood, but this Ab was not rapidly cleared, and it did not accumulate in the liver. Instead, an increase was seen in the lungs and skin, suggesting that it was sequestered at these sites. VH97VL28 and to a lesser extent VH0VL28 were noteworthy in that very little was present in the skin but both persisted in the blood. Finally, VH0VL97 and VH0VL50, which had the most extended half-lives, had at least 3 times more radioactivity present in the spleen than the other Abs.
Biodistribution of the glycosylated Abs. Radioiodinated Ab was injected i.p. into BALB/c mice, and the presence of radioactivity in organs harvested at 48 h was determined. All measurements were corrected for the blood volume of the organ.
Discussion
In studies performed several years ago we demonstrated that the presence of N-linked carbohydrate at Asn58 in the VH of the Ag binding site of an Ab specific for α(1→6) dextran (TKC3.2.2 or 14.6b.1) increases its affinity for dextran approximately 10-fold 15 . Since carbohydrate plays a demonstrable role in their binding behavior, anti-dextrans provide a useful system for studying the role of V region glycosylation in Ab function. To investigate the effects of carbohydrate attached at Asn residues close to the Ag contact loops on the affinity and the in vivo behavior of Abs, site-directed mutagenesis and gene transfection techniques were used to place carbohydrate addition sequences at novel sites in all the CDRs of the anti-dextran Ab, choosing positions expected to be on the surface of the combining site and thus predicted to be accessible to glycosyltransferases. Carbohydrate addition to the V region of Abs can be a mechanism used by the immune system to modulate Ab specificity, affinity, and stability. Although sequence analysis shows that about one-third of Abs contain V region carbohydrate addition sites, examination of the germline V sequences (V base, Medical Research Council Center for Protein Engineering, Cambridge, U.K.) 32, 33, 34 shows an absence of potential N-linked carbohydrate sites, suggesting that they mainly arise from somatic mutations. However, the presence of V region glycosylation does not always have a direct impact on binding. There are Abs with glycosylation addition sites in which the attached oligosaccharide does not affect the binding site 35 . These oligosaccharides are usually present in the framework regions and could potentially be used for the attachment of drugs or markers in diagnostic or therapeutic molecules without affecting the specificity of binding 36 . In the present study we now find that only carbohydrate added at position 58 within VH significantly increases Ab affinity compared with Abs without V region carbohydrate. This increase is observed when the glycosylated heavy chain is paired with a nonglycosylated light chain or a light chain with carbohydrate at position 28. Anti-carbohydrate Abs are intrinsically of low affinity, and glycosylation at Asn VH 58 is present as a naturally occurring somatic mutant. The combining site of the anti-dextran is a shallow groove, and all of the engineered glycosylation sites would be expected to surround this groove. The Asn at VH 58 is at the very edge of the binding site, and the increased affinity of the glycosylated Ab is thought to result from the hydrophilic interactions of carbohydrate with carbohydrate (Ag).
Glycosylation at Asn97 in the VH CDR3 appears to have no influence on binding, suggesting that the attached carbohydrate is positioned so that it is not available to interact with the carbohydrate Ag. In contrast, we find that carbohydrate introduced at other sites has an adverse affect on the affinity of the Ab. Glycosylation at Asn28 of the heavy chain decreases the affinity of the anti-dextran by 2.5-fold. Glycosylation at Asn28 of the VL always resulted in a decrease in binding affinity, reducing the affinity by half when present by itself or in combination with VH58. Ag binding was abolished when VL28 was paired with either VH28 or VH97. Since Ag binding was observed for these Abs when they were produced in the presence of tunicamycin, the presence of the two bulky carbohydrate groups must impair access to the Ab binding site. In contrast, the glycosylation sites introduced at VL50 and VL91 lead to Abs that do not bind Ag even if glycosylation is prevented by growth in the presence of tunicamycin. Analysis of contact loops predicts that VL50 and VL91 make important contacts with the Ag, and it is therefore probable that the amino acids introduced at these sites to form the carbohydrate addition sites destroy the ability to bind Ag.
Although glycosylation takes place at all the V region carbohydrate addition sites introduced in the current studies, we found that some sites were only partially used. It is unclear what causes this incomplete glycosylation. It is possible that the rate of folding of the region might limit oligosaccharide attachment and processing. Carbohydrate addition takes place on the nascent polypeptide chain, and once folding of this region has occurred, sites can become inaccessible 37, 38 . We found no evidence of terminal high mannose residues in any of our Abs, since all were Endo-H glycosidase resistant. Therefore, the attached carbohydrate is available to the processing enzymes as the protein transits the Golgi apparatus. However, the altered mobility of a portion of the light chains in VH58VL28 suggests that different glycoforms are present on some of the light chains and that the presence of carbohydrate at VH58 influences the processing of carbohydrate attached at VL28.
The rules governing the use of a potential N-linked glycosylation site are not completely understood, although it is accepted that proline cannot be the middle amino acid in the canonical N-linked carbohydrate addition site. However, analysis of V region sequences that actually have N-linked carbohydrate suggests that the presence of hydrophobic amino acids such as Ala at the middle residue can inhibit site utilization. In previous studies with an anti-dansyl Ab, mutation of Ala in the sequence Asn-Ala-Thr to Gly resulted in the use of this previously unused site (our unpublished observation). It also has been observed that when proline is present in the fourth position after the Asn, the site is not used. In fact, although the majority of murine heavy chain V regions belonging to subgroup IIIb have Asn at position 58, they differ from the germline VH used in these studies in that they have a proline instead of a glutamic acid at position 61.
The presence of carbohydrate in one V region of one chain does not impede the attachment of carbohydrate to the other chain. There is no evidence of asymmetric glycosylation in any of the mutant Abs containing engineered sites in both VH and VL. The 1→6 linkages of the N-linked oligosaccharides are a major source of intramolecular flexibility, and conformational heterogeneity may contribute to structural heterogeneity, since the accessibility of enzymes to the site can be altered by changing the orientation of the flexible linkages 39 . The protein itself can also play an important role in determining the specificity of processing, and even the quaternary structure of a protein has been found to influence the structure of the carbohydrate added to different sites 40 .
Different glycoforms on the surface of immune-related proteins can give them altered or distinguishable functions 19, 20 and have been related to disease states 21, 22, 41 . Although the in vivo half-lives of the majority of mutant anti-dextrans were very similar to the expected half-life of human IgG1 in mice, three Abs showed a marked difference in the rate of clearance; VH0VL50 and VH0VL91 had a longer half-life of 10 days, while VH58VL28 had a faster clearance rate and a shorter half-life of 4 days. These data are consistent with the radioactivity measured in the different organs at 48 h postinjection, in that VH58VL28 showed a markedly lower blood retention and larger accumulation in the liver and clearance through the kidneys. It is noteworthy that both VH0VL50 and VH0VL91 were present in the spleen in increased amounts.
While mannose receptors abundant in the liver have been implicated in the rapid clearance of Abs and other proteins containing unprocessed sugars with high mannose content 42 , this does not appear to be the case for VH58VL28, since its variable region carbohydrate is not sensitive to digestion with Endo-H glycosidase. Alternatively, it is possible that this Ab is clearing through the asialoglycoprotein receptor present in the hepatocytes. This would imply that VH58VL28 is heavily galactosylated and could explain the presence of the third glycosylated species in the light chain 43, 44 . The spleen uptake of VH0VL50 and VH0VL91 suggests that they are being processed by the reticuloendothelial system 45 , but it is unclear how this would translate into a longer half-life.
The presence of carbohydrate in the V region of the heavy chain seems to increase the immunogenicity of the anti-dextrans. It is interesting that carbohydrate at VH58 enhanced the response to the constant regions when paired with a light chain lacking carbohydrate, but when paired with a light chain with carbohydrate at position 28 it induced both an anti-V and an anti-constant region response. A significant amount of the immune response to VH28VL28 was anti-idiotypic. It is noteworthy that chimeric Abs lacking V region carbohydrate did not elicit a detectable immune response during the 10–15 days of the half-life studies despite the fact that they possessed human constant regions.
Traditionally, changes in amino acid sequence have been used to alter the functional properties of Ab molecules. Our current studies demonstrate that addition of V region carbohydrate can be used to influence not only the ability of the Ab to interact with Ag but also in vivo properties such as half-life. Glycosylation at positions that result in the reduction of the serum retention of an Ab could increase its therapeutic value when rapid clearance is needed. Although the production of Ab fragments 46, 47, 48 or isotype switching 49 can be used to decrease half-life, when the effector functions of an Ab, such as Fc receptor binding and cytolysis, are important for the specific application of an Ab, conserving the constant region of an active isotype and altering the half-life by other means, such as carbohydrate engineering, are attractive options. Additionally, the affinity of some Abs can be increased or decreased by the addition of carbohydrate at defined positions in the variable region. Carbohydrate accessible on the surface of the Ab that does not alter the functional properties can provide sites for the attachment of drugs, radioactive tags, or even purification using lectins.
Footnotes
↵1 This work was supported by National Institutes of Health Grants CA16858, AI39187, and AI29470.
↵2 Address correspondence and reprint requests to Dr. Sherie L. Morrison, Department of Microbiology, Immunology, and Molecular Genetics, University of California, 1602 Molecular Sciences Building, 609 Circle Drive East, Los Angeles, CA 90095-1489. E-mail address: sheriem{at}microbio.ucla.edu
↵3 Abbreviations used in this paper: VH and VL, heavy and light chain variable regions; CDR, complementarity-determining region; IMDM, Iscove’s modified Dulbecco’s medium; Endo-H, endoglycosidase H; aka, apparent affinity constant; βt1/2, β phase half-life.
- Received June 11, 1998.
- Accepted November 6, 1998.
- Copyright © 1999 by The American Association of Immunologists
References
In this issue
