METHOD FOR AFFINITY PURIFICATION

Abstract

The invention relates to a method of immunoaffinity purification which comprises the use of a binding agent which binds to an epitope that it is present at least twice on the target molecule. In another embodiment the method uses at least two different binding agents, each binding to different epitopes on the target molecule.

Claims

1. A method for purification of a target molecule, comprising: binding the target molecule to an immunoadsorbent material comprising one or more binding agents wherein: a. the binding agent is an isolated, single domain camelid antibody or fragment thereof; b. the binding agent has affinity for at least two, spatially separated epitopes on a target; and, c. the two, spatially separated epitopes on the target are immunologically identical.

2. The method according to claim 1, wherein the two, spatially separated epitopes on the target molecule are such that binding of a first epitope on the target molecule does not substantially block the binding of a second epitope on the target molecule, to the binding agent.

3. The method according to claim 1, wherein the avidity of the immunoadsorbent material for the target molecule is at least 50 fold higher than the lowest affinity of the binding agent for an individual epitope.

4-5. (canceled)

6. The method according to claim 1, wherein the binding agent comprises an immunoglobulin-derived variable domain that comprises a complete antigen binding site for an epitope on the target molecule in a single polypeptide chain and whereby the framework amino acid sequences of the variable domain have at least 50% amino acid identity with the frame work amino acid sequence of any one of SEQ ID No's 1-33.

7. (canceled)

8. The method according to claim 6, wherein the target molecule is an immunoglobulin or a fragment thereof.

9. The method according to claim 8, wherein the at least two epitopes are outside the 5 CDR's of the immunoglobulin or a fragment thereof.

10. The method according to claim 8, wherein at least one epitope is present on the light chain of the immunoglobulin.

11. The method according to claim 8, wherein the at least two epitopes are epitopes of a human immunoglobulin.

12. The method according to claim 11, wherein the epitopes are epitopes of a human immunoglobulin light chain of the kappa or lambda isotype.

13. The method according to claim 12, wherein the binding agent is selected from the group of kappa light chain binding VHH molecules selected from SEQ ID No's 1-15, or a binding agent comprising an immunoglobulin[e]-derived variable domain comprising a Complementarity Determining Region (CDR) 1, 2, and/or 3 exhibiting at 20 least 80, 85, 90, 95, 98% amino acid identity with the CDRs of the VHH molecules of SEQ ID No's 1-15.

14. The method according to claim 12, wherein the binding agent is selected from the group of lambda light chain binding VHH molecules selected from SEQ ID No's 16 to 33, or a binding agent comprising an immunoglobulin-derived variable domain comprising a Complementarity Determining Region (CDR) 1, 2, and/or 3 exhibiting at least 80, 85, 90, 95, 98% amino acid identity with the CDRs of the VHH molecules of SEQ ID No's 16-33.

15. The method according to claim 11, wherein the at least two epitopes are at least two immunologically distinct epitopes of a human IgG Fc domain.

16. An immunoadsorbent material comprising one or more binding agents wherein: a. the binding agent is an isolated, single domain camelid antibody or fragment thereof; b. the binding agent has affinity for at least two, spatially separated epitopes on a target; and, c. the two, spatially separated epitopes on the target are immunologically identical.

Description

DESCRIPTION OF THE FIGURES

[0088] FIG. 1 Alignments of human kappa (SEQ ID NO: 1-15) and human lambda (SEQ ID NO: 33) light chain binding VHHs and consensus for the CDR' regions.

[0089] FIG. 2 Sensorgrams of BIACORE affinity measurements; Fab- and IgG binding curves in BIACORE on a VHH Hu-kappa-1 coated sensor chip (A; binding and dissociation, B; dissociation only, indicating the effect on the dissociation rate (kdiss) induced by avidity.

[0090] FIG. 3 CDR amino acid sequences of human kappa light chain binding VHH molecules (SEQ ID NO: 1-15)

[0091] FIG. 4 CDR amino acid sequences of human lambda light chain binding VHH molecules (SEQ ID NO: 16-33).

EXAMPLES

Example 1

Generation of Llama VHH Ligands Against Kappa Light Chains of Human Antibodies

[0092] The following protocol is taken as an example of how specific VHH fragments can be isolated, cloned, expressed, then coupled onto the desired matrix.

[0093] Although the particular VHH fragments described in this example were derived from an immune repertoire, they also could have been selected from a non-immunized VHH library (see EP1051493, Unilever) or a synthetic/semi-synthetic non-immunized VHH library (see WO00/43507, Unilever).

[0094] A llama was immunized with polyclonal IgM prepared from human serum by precipitation and gel filtration techniques and diluted in phosphate buffered saline pH 7.4 (PBS). To increase the specificity of the immune response the llama was boosted several times following the initial immunization (days 0, 28 and 49) with 250 μg of the antigen mentioned above in specol (ID-DLO, Lelystad, The Netherlands) (Boersma et al., 1992). A heparin blood sample of about 150 ml was taken 6 days after the last immunization. Peripheral blood cells were obtained via a Ficoll-Paque centrifugation. From about 2×10.sup.8 lymphocytes the total RNA was isolated essentially as described by Chomczynnski and Sacchi (1987). DNA encoding specific VHH fragments can then be isolated using similar methods to those described in WO 94/04678 (Casterman et al) and ligated into e.g. an episornal Saccharomyces cerevisiae expression vector as previously described by Frenken et. al. (2000). From these VHH expression libraries, VHH fragments with the desired antigen binding specificity (kappa or lambda light chains of human antibodies) can be selected by direct screening of culture supernatants from individual S. cerevisiae colonies (Frenken et. al. 2000).

[0095] Alternatively selection methods based on display techniques like phage-display and yeast display can be used to isolate anti kappa light chain VHH producing clones from immune repertoires.

[0096] For screening Nunc Maxisorp binding plates were coated with human antibody antigens and subsequently blocked with 4% (w/v) milkpowder in PBS. Bound VHH fragments were detected by either a mouse anti-His mAb in combination with a polyclonal goat-anti-mouse-HRP conjugate (Bio-Rad, 172-1011) or a polyclonal rabbit anti-llama-VHH serum in combination with a polyclonal swine-anti-rabbit IgG-HPO conjugate (Dako, P217). Initial screening was performed on Maxisorp plates either coated with a human IgG1 monoclonal antibody that possessed a kappa light chain or a human IgG1 mab that possessed a lambda light chain. Llama VHH fragments that showed binding to IgG1 kappa only were further tested in ELISA on different human monoclonal antibodies (e.g. IgG, IgA, IgM) to confirm binding specificity towards the human kappa light chain. During the screening process, additional VHH fragments could be identified that showed binding specificity towards the human lambda light chain. The binding specificity of one of the anti human kappa VHH fragments (VHH Hu-kappa-1) as determined by ELISA is given in Table 1 and herein compared with two other VHH fragments that specifically bind to the Fe portion of human IgG antibodies (VHHs Hu-Fc-1 and Hu-Fc-2, respectively). These VHH fragments were obtained from an immune VHH library originating from a llama immunized with human Fc fragments prepared from polyclonal IgG from human serum. As demonstrated in Table 1, VHH Hu-kappa-1 recognizes an epitope that is present on kappa light chains of human antibodies and hence binds to an epitope that is present twice in the target molecule in the case of e.g. IgG, IgA and IgE antibodies or ten times in the case of IgM antibodies. Both VHH fragments Hu-Fc-1 and Hu-Fc-2 show to be specific for an epitope that is present on the Fc domain of all 4 human IgG subclasses, said epitope being present only once in the IgG target molecule.

[0097] The sequence of these antibodies is presented in sequence ID no's 1 to 15 for the kappa light chain binding VHHs, no's 16 to 31 for the lambda binding VHHs. The consensus for both kappa and lambda binders is displayed in FIG. 1.

TABLE-US-00002 TABLE 1 Binding specificity of anti human kappa light chain and anti human IgG Fc VHH fragments as determined in ELISA. Antibody: VHH fragment: Species Isotype/Subclass Hu-kappa Hu-Fc-1 Hu-Fc-2 Human IgG.sub.1 lambda − + + IgG.sub.2 lambda − + + IgG.sub.3 lambda − + + IgG.sub.4 lambda − + + IgG.sub.1 kappa + + + IgG.sub.2 kappa + + + IgG.sub.3 kappa + + + IgG.sub.4 kappa + + + IgG, Fab fragments + − − IgG, Fc fragments − + + IgM (polyclonal) + − − IgA (polyclonal) + − − Bovine IgG (polyclonal) − − − Mouse IgG (polyclonal) − − −

Antibody Production and Immobilized Metal Affinity Chromotography (IMAC) Purification

[0098] The VHH antibody fragments were produced at 1-Litre or 10-Litre (shake flasks) fermentation scale using a genetically modified strain of Saccharomyces cerevisiae and purified using ion exchange chromatography on SP sepharose fast flow (Amersham Biosciences). The purified antibodies were dialysed against three changes of PBS buffer, pH 7.4 or the required coupling buffer, 48 hours at 4° C., using 3.5 kDa cut-off tubing (Spectra/Por 3; Spectrum Medical Industries). The concentrations of the purified samples were determined by OD.sub.280. All purified samples were stored at −20° C. when not in use.

Example 2

Affinity Measurements

[0099] Binding affinity constants of VHH fragments Hu-kappa-1, Hu-Fc-1 and Hu-Fc-2 were determined using surface plasmon resonance analysis (SPR) on a BIACORE 3000. For this purpose, purified VHH fragments were immobilised onto the surface of a CM5 sensor chip and subsequently incubated with different concentrations of human Fab and/or human IgG antibodies in HBS-EP buffer (0.01 M HEPES, pH7.4; 0.15 M NaCl; 3 mM EDTA; 0.005% Surfactant P20). Binding was allowed for 3 minutes at 30 μl/min followed by a dissociation step of 15 minutes at 30 μl/min. Binding curves were fitted according to a 1:1 Langmuir binding model using BIACORE software. An overview of the calculated affinity data is given in Table 2. With regard to the anti Hu-kappa-1 fragment the effect of avidity is clearly demonstrated by major differences in dissociation rates between Fab- and IgG molecules. Since the anti Hu-kappa-1 fragment is immobilised onto the surface of a sensor chip, said surface can interact with two identical epitopes present on one IgG molecule simultaneously. Compared to the monovalent interaction with human Fab fragments, the dissociation rate (kdiss) for IgG is significantly lower (>1000 fold) resulting in a KD value for IgG of about 120 fM compared to 6.3 nM for Fab fragments. The latter value is in the same range as the KD values for the two anti Hu-Fc VHHs (6.4 and 2.2 nM for IgG, respectively) indicating monovalent interactions.

TABLE-US-00003 TABLE 2 B1ACORE affinity data of anti Hu-kappa-1 and anti Hu-Fc VHHs (VHHs immobilised on sensor chip). k ass k diss KA KD VHH Antigen (1/Ms) (1/s) (1/M) (M) Avidity Hu-kappa-1 Hu-Fab 1.24 × 10.sup.5 7.78 × 10.sup.−4 1.60 × 10.sup.8 6.27 × 10.sup.−9 No Hu-kappa-1 Hu-IgG 4.18 × 10.sup.5 5.29 × 10.sup.−8 .sup. 7.90 × 10.sup.12 .sup. 1.27 × 10.sup.−13 Yes Hu-Fc-1 Hu-IgG 2.15 × 10.sup.5 1.39 × 10.sup.−3 1.55 × 10.sup.8 6.44 × 10.sup.−9 No Hu-Fc-2 Hu-IgG 8.30 × 10.sup.5 1.78 × 10.sup.−3 4.66 × 10.sup.8 2.15 × 10.sup.−9 No

[0100] This effect on the dissociation rate (kdiss) induced by avidity is further illustrated by FIG. 2, which compares Fab- and IgG BIACORE binding curves sensorgrams).

Example 3

General Materials and Methods for Coupling and Chromatography Testing

Coupling to N-Hydroxysuccinimide (NHS) Activated Sepharose 4 Fast Flow

[0101] After purification the antibodies were dialysed to NHS coupling buffer. This buffer contains 0.1 M HEPES pH 8.3. For an optimal coupling efficiency the recommended ratio of volumes coupling solution/NHS sepharose is 0.5:1. The antibodies had different concentrations, 0.5-15 mg/ml, the ratio of antibody/NHS sepharose of the antibodies varied between 1:1 and 10:1. When a mixture of 2 ligands is immobilised onto the matrix a 1:1 ratio of ligands was used. The following procedure was used for coupling the antibodies to NHS activated sepharose 4 Fast Flow (General Electric Healthcare). Subsequently the matrix was washed twice with NHS coupling buffer. The NHS sepharose was mixed with the antibody solution and left overnight at 4° C. head over head or 1 hour at room temperature. After incubation the gel material was filtered over a sintered glass filter and the non-reacted groups of the gel material were blocked with Tris (0.1 M pH 8.0) 1 hour at room temperature. The coupled medium was washed using alternate low and high pH (3×10 cv PBS pH 2 and 3×10 cv PBS pH 7.4). Using the non-bound fraction the coupling efficiency was determined. This is determined by measuring the OD.sub.280 value of the coupling solution before and after coupling. The coupling efficiency was also determined by looking at the protein pattern on SDS-PAGE of coupling solution before and after coupling.

Chromatography Experiments

[0102] Columns were made of the coupled antibody matrix using HR 5/5 columns (GE health). A column volume of 400 μl was used. All the chromatography experiments were performed on an Akta explorer 100. A two buffer system was used, buffer A1 the loading buffer was PBS pH 7.4, buffer B the elution buffer e.g. PBS with addition of 8 M HCl to yield pH 2.1 or 0.1 M Glycine-HCl at pH 2 or 3. Different programs were used. Protein detection was performed on line by monitoring the signal of OD.sub.214 and OD.sub.280. Fractions were collected with a volume of 1 ml. The fractions were neutralized immediately with 20 μl Tris (2 M).

Determination of the Dynamic Binding Capacity of the Antibody Coupled Matrix

[0103] To determine the dynamic binding capacity the column was loaded with the target molecule. A flowrate of 150 cm/h was used. The amount of target molecule in the flowthrough and the elution were calculated by integration of the flowthrough and elution peak area. The dynamic capacity is the elution peak area devised by the total peak area (flowthrough elution) and this multiplied with the amount of target molecule.

Example 4

Dynamic Binding Capacities of Different VHH Matrices for Human IgG

[0104] VHH fragments (ligands) Hu-kappa-1, Hu-Fc-1 and Hu-Fc-2 were immobilised onto NHS sepharose as described earlier. Ligand densities were 20 mg of ligand per ml matrix. The dynamic binding capacity was determined using procedures as described above. The column was loaded with an amount of Human IgG higher than the expected dynamic binding capacity. Elution was performed using 0.1M Glycine buffers with pH values between 2 and 4.

[0105] As can be seen from Table 3, the highest dynamic binding capacity (DBC) for human IgG is found for the Hu-kappa-1 matrix and is almost a 2-fold higher compared to both Hu-Fc-1- and Hu-Fc-2 matrices. Affinity measurements showed that the binding affinity (KD) of this antibody for human IgG molecules (containing 2 identical kappa light chains) is about 120 fM. This is significantly higher than the actual binding affinity of Hu-kappa-1 for its epitope present on the kappa light chains as determined with monovalent binding Fab fragments (6.3 nM). The latter figure is more in the range of both tested Human Fc specific VHH fragments and generally known compositions in the prior art. This shows that the use of immuno-adsorbent materials according to the invention leads to increased binding capacity compared to systems that do not include the feature of multivalent binding.

[0106] Another important feature of the current invention is that although multivalent binding ligands leads to high binding affinities for the target molecule, the release of said molecules can still be obtained at mild elution conditions. As demonstrated in Table 3 the optimal pH of elution for the Hu-Fc-2 matrix is lower (pH 2) compared to the Hu-kappa-1- and the Hu-Fc-1 matrix (both pH 3). The binding strength of the Hu-Fc-2 fragment for its epitope on human IgG is almost 3-fold higher compared to Hu-Fc-1 and Hu-kappa-1 (2.2 nM versus 6.4 nM and 6.3 nM, respectively). Although multivalent binding of the Hu-kappa-1 matrix increases the dynamic binding capacity for human IgG compared to e.g. the Hu-Fc-1 matrix, this avidity feature clearly does not affect conditions to obtain efficient release of bound IgG. Said conditions being comparable to those for systems that do not include the feature of multivalent binding (e.g. Hu-Fc-1 matrix).

[0107] The results as mentioned in this example are measured with a low bed height and short residence time. Higher bed heights will increase residence time and the dynamic capacity. The dynamic capacity of the described Hu-kappa-1 column at 15 cm bed height and 150 cm/hr linear flow rate is 30 to 40 mg Human IgG/ml matrix. This is about a two-fold higher than the capacity of known systems that do not include the feature of multivalent binding which is specific for the current invention.

[0108] The example furthermore demonstrates that the multivalent principle of the current invention enables a unique combination of high binding capacity on one-hand and mild elution conditions on the other hand.

TABLE-US-00004 TABLE 3 Comparison of the dynamic binding capacity (DBC) of anti Human IgG matrices. pH of elution Ligand DBC giving >90% KD Density (mg release of bound for HuIgG Matrix (mg/ml) HuIgG/ml) IgG (M−1) Hu-kappa-1 20 24 3 .sup. 1.27 × 10.sup.−13 Hu-Fc-1 20 14 3 6.44 × 10.sup.−9 Hu-Fc-2 20 15 2 2.15 × 10.sup.−9

Example 5

Dynamic Binding Capacities of Matrices for Human IgG Comprising VHHs that Bind to Different Epitopes Present in the Target Molecule

[0109] This example demonstrates the feature of increased binding capacity induced by multivalent binding using at least two different VHH ligands, each ligand recognizing a different epitope on the target molecule, in this case human IgG antibodies.

[0110] For this purpose two anti human IgG Fc ligands (Hu-Fc-1 and Hu-Fc-2) were used, each ligand binding to a different epitope present on the Fc domain of human IgG antibodies as demonstrated by BIACORE binding analysis (see Table 4). In this BIACORE experiment, purified anti human IgG Fc ligands were immobilised onto the surface of a CM5 sensor chip and subsequently incubated with human Fc fragments. This human Fc capturing step was followed by incubation with either VHH ligand Hu-Fc-1 or Hu-Fc- 2. Table 4 shows that VHH ligand Hu-Fc-2 can bind to human Fc fragments when captured by immobilised VHH ligand Hu-Fc-1 and vice versa (51 RU and 181 RU, respectively) which demonstrates that each ligand binds to a different epitope present on the Fc domain of human IgG antibodies. No significant binding signals were obtained in this set-up using identical ligand pairs, indicating that each individual ligand binds to an epitope on the Fc domain of human IgG antibodies that is present or available only once.

TABLE-US-00005 TABLE 4 Binding signals of anti Hu-Fc VHHs on captured human Fc fragments in BIACORE. BIACORE set-up: Binding Capture target anti Hu-Fc VHH Signal Immobilized VHH (100 μg/ml) (100 μg/ml) (RU) Hu-Fc-1 Human Fc Hu-Fc-1 6.0 Hu-Fc-1 Human Fc Hu-Fc-2 51.0 Hu-Fc-2 Human Fc Hu-Fc-2 9.6 Hu-Fc-2 Human Fc Hu-Fc-1 181.9

[0111] This further illustrates that although IgG antibodies consist of two identical heavy- and light chains, ligands directed against antibodies can recognise epitopes that are formed by a combination of identical chains such as present in the Fc domain of said antibodies. Another feature that can occur is that the ligand binds to one part of the heavy chain of the Fc domain in such a way that it prevents binding of another identical ligand due to steric hindrance.

[0112] In this respect, light chain domains of antibodies (e.g. anti human kappa light chains) as demonstrated in. Example 4, have proven to be ideal epitopes to generate specific ligands against to allow multivalent binding to an epitope that is present twice in the case of antibodies. Said ligands are clearly not affected by this feature of steric hindrance.

[0113] Three different batches of anti human IgG Fc matrices were constructed (Hu-Fc-1, Hu-Fc-2 and Hu-Fc-1/2) onto NHS sepharose using methods as previously described. For each matrix the final ligand density was 2 mg of ligand per ml matrix. The dynamic binding capacity was determined using the procedure as previously described. The column was loaded with an amount of human IgG higher than the expected dynamic binding capacity. As can be seen from Table 5, the dynamic binding capacity for the mixed matrix Hu-Fc-1/2 is higher than for each individual matrix. The dynamic binding capacity for the mixed matrix (2.45 mg human IgG/ml matrix) is 1.5 times higher than the average value of the individual matrices (1.65 mg human IgG ml matrix) that at maximum could be expected when no multivalent binding would have occurred on the mixed matrix. Based on the multivalent binding principle of the current invention these results demonstrate that high binding capacity matrices can be obtained using combinations of different ligands, each ligand binding to a different epitope present on the target molecule.

TABLE-US-00006 TABLE 5 Comparison of the dynamic binding capacity (DBC) of mixed anti human IgG matrices. Ligand Density DBC Mixed Matrix (mg/ml) (mg HuIgG/ml) Hu-Fc-1 2 1.95 Hu-Fc-2 2 1.35 Hu-Fc-1/Hu-Fc-2 2 2.45

Example 6

Dynamic Binding Capacities of Matrices for Human Serum Albumin Comprising VHHs that Bind to Different Epitopes Present in the Target Molecule

[0114] This example demonstrates the feature of increased binding capacity induced by multivalent binding using at least two different VHH ligands, each ligand recognizing a different epitope on the target molecule, in this case human serum albumin.

[0115] For this purpose three different anti human serum albumin (HSA) specific VHH ligands were used (HSA-1, HSA-2 and HSA-3). These VHH ligands were obtained from an immune VHH library originating from a llama immunized with human serum albumin. BIACORE binding analysis according to methods as previously described revealed that ligands HSA-2 and 3 bind to the same epitope present on HSA, whereas ligand HSA-1 binds to a different epitope allowing multivalent binding of HSA when VHH ligand HSA-1 is used in combination with either VHH ligand HSA-2 or 3.

[0116] Five different batches of anti HSA matrices were constructed (HSA-1, HSA-2, HSA-3, HSA-1/2 and HSA-2/3) onto NHS sepharose using methods as previously described. For each matrix the final ligand density was 2 mg of ligand per ml matrix. The dynamic binding capacity was determined using the procedure as previously described. The column was loaded with an amount of HSA higher than the expected dynamic binding capacity. As can be seen from Table 6, the dynamic binding capacity for the mixed matrix HSA-1/2 is higher than for each individual matrix and the mixed matrix HSA-1/3. The dynamic binding capacity of mixed HSA-1/2 matrix is 2.14 mg HSA/ml matrix, which is about 1.5 times higher compared to the expected average value of 1.48 mg HSA/ml matrix when no multivalent binding would have occurred on the mixed matrix ((1.72+1.23)/2=1.48). The dynamic binding capacity of HSA for the mixed matrix HSA-2/3 is 0.83 mg HSA/ml matrix. This is in line with the expected average value of 0.85 mg HSA/ml matrix, since both ligands bind to the same epitope present on HSA and therefore can not induce multivalent binding leading to an increased binding capacity as demonstrated by the current invention.

TABLE-US-00007 TABLE 6 Comparison of the dynamic binding capacity (DBC) of mixed anti HSA matrices. Ligand Density DBC Mixed Matrix (mg/ml) (mg HSA/ml) HSA-1 2 1.72 HSA-2 2 1.23 HSA-3 2 0.46 HSA-1/2 2 2.14 HSA-2/3 2 0.83

REFERENCES

[0117] Boersma, W. J. A., Bogaerts, W. J. C., Bianchi, A. T. J., Claassen, E., 1992. Adjuvant properties of stable water-in-oil emulsions: evaluation of the experience with specol. Res. Immunol. 143, 503-512.

[0118] Chomczynnski, P., Sacchi, N., 1987. Single step method of RNA isolation by acid guanidium thiocyanate-phenolchloroform extraction. Anal. Biochem. 162, 156-159.

[0119] Frenken G. J., Richard H. J. van der Linden, Pian W. J. J. Hermans, J. Wil Bos a, Robin C. Ruuls, Bernard de Geus, C. Theo Verrips, 2000, Journal of Biotechnology 78, 11-21.