ADA-RESPONSE SPECIFICATION ASSAY

Abstract

Herein is reported a method for determining the epitope of an antibody specifically binding to a therapeutic antibody comprising the steps of a) incubating a sample, which comprises serum and the antibody specifically binding to a therapeutic antibody, separately with i) at least a Fab fragment of the therapeutic antibody, and ii) at least a Fab fragments of the therapeutic antibody in which the HVRs forming a paratope have been replaced with germline sequences, and detecting the binding or non-binding of the antibody specifically binding to a therapeutic antibody to the at least a Fab fragment in any of i) to ii), and b) determining the epitope of the antibody specifically binding to a therapeutic antibody to be in the at least one HVR that has been replaced in ii) if binding is detected in i) and non-binding is detected in ii).

Claims

1. A method for determining the epitope of an anti-drug antibody, which is specifically binding to a therapeutic antibody specifically binding to a therapeutic target, comprising the following steps: a) separately incubating aliquots of a sample, which comprises the anti-drug antibody, with i) at least a Fab fragment of the therapeutic antibody, and ii) at least a Fab fragment of the therapeutic antibody in which all HVRs of the paratope to the therapeutic target have been replaced with germline or non-binding HVRs, and detecting the binding or non-binding of the anti-drug antibody to the at least a Fab fragment in any of i) to ii), and b) determining the epitope of the anti-drug antibody to be in the paratope to the therapeutic target if binding is detected in i) and non-binding is detected in ii).

2. The method according to claim 1, wherein the non-binding or germline HVR is obtained from the same germline as the Framework-regions of the therapeutic antibody are.

3. The method according to any one of claims 1 to 2, wherein the therapeutic antibody is a mono- or bispecific Fab.

4. The method according to any one of claims 1 to 3, wherein the paratope is formed by a VH/VL-pair.

5. The method according to any one of claims 1 to 4, wherein the therapeutic antibody is a bispecific Fab, wherein the first paratope is formed by the HVRs H-1, H-3 and L-2 and the second paratope is formed by the HVRs H-2, L-1 and L-3, in step a) step ii) is the therapeutic antibody in which HVRs H-1, H-3 and L-2 have been replaced with non-binding HVRs, step a) further comprises vi) the therapeutic antibody in which HVRs H-2, L-1 and L-3 have been replaced with non-binding HVRs, step b) is determining the epitope of the anti-drug antibody to be in the first paratope of the bispecific Fab if binding is detected in i) and vi), and non-binding is detected in ii), in the second paratope of the bispecific Fab if binding is detected in i) and ii), and non-binding is detected in vi), in the HVR-Framework region-junction of the first paratope if binding is detected in i) and vi) and non-binding is detected in ii), in the HVR-Framework region-junction of the second paratope if binding is detected in i) and ii) and non-binding is detected in vi).

Description

DESCRIPTION OF THE FIGURES

[0240] FIG. 1 Gene usage frequencies observed for the VH domain of human (n=9340) Ig sequences published in the IMGT/LIGM-DB database. Each of the mapped IGHV genes usage was calculated as the percentage of the total unique population of productive and in-frame sequences according to IMGT/HighV-QUEST Statistical Analysis Report (reproduced from B. Shi, et al., Ther. Biol. Med. Model. 11 (2014) 30, supplemental FIG. 1A).

[0241] FIG. 2 Germlinization scheme heavy chain variable domain. The respective stretches of the heavy variable domain denoted with “F”, “C” and “—” as well as “J” are used as query sequence in the alignment with the human IMGT germline repertoire to identify the most identical germline sequence. “-” represent a gap used in the alignment whereas the corresponding sequence of the identified germline is used to replace the respective HVR sequence of the antibody.

[0242] FIG. 3 Germlinization scheme light chain variable domain. The respective stretches of the light variable domain denoted with “F”, “C” and “—” as well as “J” are used as query sequence in the alignment with the human IMGT germline repertoire to identify the most identical germline sequence. “-” represent a gap used in the alignment whereas the corresponding sequence of the identified germline is used to replace the respective HVR sequence of the antibody.

[0243] FIG. 4 Schematic illustration of the bispecific Fab biotherapeutic, engineered Fab variants and control molecules generated for detection of domain-specific anti-drug antibodies (ADAs). Constructs 2-4 are modified versions of original Duta-original drug molecule.

[0244] FIG. 5: Domain detection assay principle: distinct domains (described in FIG. 4) are directly coated to a Nunc MaxiSorp™ plate and used to capture domain-specific ADAs. Bound ADAs are detected by a Dig-labeled anti-Cynomolgus IgG. Signals are generated using an anti-Dig antibody coupled to horseradish peroxidase (HRP), leading to substrate/ABTS color conversion.

[0245] FIG. 6: Schematic representation of an exemplary domain detection assays according to the current invention as used in example 5. Five different Fab variants, representing different domains are immobilized on the surface. ADAs with different specificities can attach to different domains. The further recognition of the ADAs is done by an anti-Cynomolgus antibody.

[0246] FIG. 7: Expected positivity patterns of the exemplary assay setup of the method according to the current invention as used in example 5.

[0247] FIG. 8: ADA-response “type 8” determined with the assay according to the invention.

[0248] FIG. 9: ADA-response “type 9” determined with the assay according to the invention.

[0249] FIG. 10: ADA-response “type 10” determined with the assay according to the invention.

[0250] FIG. 11: ADA-response “type 11” determined with the assay according to the invention.

EXAMPLES

Chemicals, Reagents and Equipment

[0251] The therapeutic antibodies and specific assay reagents were all provided by Roche Diagnostics GmbH, Penzberg, Germany, and stored in aliquots at −80° C. until use. The bispecific therapeutic Fab (“Fab-original”, drug) consists of a monoclonal Fab directed against two different antigens, which was available from recombinant expression. It was used after purification and analytical characterization. Biotinylated (Fab-original-Bi) and digoxygenylated (Fab-original-Dig) versions of the bispecific construct “Fab-original” were used as capture and detection reagents, respectively, in the bridging ADA assay.

[0252] For domain detection assay (DDA) and domain competition assay (DCA), the following un-labeled reagents were used for capturing: “Fab-original” (bispecific therapeutic Fab), “Germ1-control” (bispecific therapeutic Fab, wherein the first paratope is intact and the second paratope had been back-mutated to a germline sequence) and “Germ2-control” (bispecific therapeutic Fab, wherein the second paratope is intact and the first paratope was back-mutated to a germline sequence), “GermGerm-control” (bispecific therapeutic Fab, wherein both paratopes were back-mutated to a respective germline sequence), Ranibizumab (commercial human Fab biotherapeutic approved for intraocular application), a non-binding human Fab (DP47). A digoxygenylated monoclonal anti-cynomolgus IgG antibody was used as detection reagent. A polyclonal sheep antibody Fab against digoxygenin conjugated to horseradish peroxidase (HRP) (pAb-Dig-S-Fab-HRP) was used as a second detection reagent in all ADA assays.

[0253] Positive control, PC=generated in vivo or via phage-display

[0254] Capture control, CC=surrogate for PC when PC is not available

[0255] The following compounds served as positive controls (PC), e.g. as (functional (cloning, generation, immobilization) capture controls (CC): total ADA screening assay: monoclonal mouse antibody directed against human IgG kappa (mAb-anti-hu-kappa, Roche Diagnostics GmbH, Mannheim, Germany); DDA: digoxygenylated recombinant antigen-1 and digoxygenylated recombinant antigen-2 were used as capture controls (detection via anti-Ckappa), pre-dose cynomolgus monkey CTAD (CTAD=citrate-theophylline-adenine-dipyridamole; anti-coagulant) plasma samples from each animal were used as negative control.

[0256] Pooled cynomolgus monkey CTAD plasma was prepared from 40 individual drug-naive female and male animals. Plasma samples were obtained from Sera Laboratory International, Ltd., Haywords Heath, UK. The washing buffer (phosphate-buffered saline (PBS)/0.05% Tween 20/0.002% Bronidox) and 2,2′-azino-bis ethylbenzthiazoline-6-sulphonic acid (ABTS) substrate were provided by Roche Diagnostics GmbH, Mannheim, Germany. Ready to use LowCross Buffer was obtained from Candor Bioscience GmbH, Wangen, Germany, and was used as dilution and assay buffer in the bridging ADA assay (ELISA). All chemicals were of analytical grade.

[0257] Three dose groups of four animals (adult monkeys; cynomolgus monkeys) each were treated with 0 (placebo), 5 and 10 mg/eye for up to 43 days. A second treatment was administered at day 29. From this study, 39 plasma samples of bispecific Fab-original construct-dosed animals were used for ADA testing. Out of these, 31 samples were found positive in the total ADA screening assay (samples tested with a dilution factor 1:20). Finally, 16 samples from 8 different animals (from each animal the pre-dose sample as a specific negative control and day-35-after-treatment-sample) were selected for further evaluation by the domain-specific approaches because of significant signal intensities (>1.0 AU) reflecting high ADA concentrations.

[0258] Streptavidin-coated microtiter plates (SA-MTP) for the bridging ADA assay were obtained from Microcoat Biotechnologie GmbH, Bemried, Germany. Un-coated Nunc MaxiSorp™ microtiter plates for the DDA were obtained from Thermo Fisher Scientific, Germany.

Example 1

Bridging ADA Assay for Detection of the Overall ADA Response (ADA Screening Assay)

[0259] A bridging ADA assay was applied to analyze plasma samples obtained from a cynomolgus monkey study (see above) for detection of ADA responses directed against the bispecific <AG1/AG2> Fab (anti-antigen-1/antigen-2 bispecific Fab). The samples were tested at a dilution factor of 1:20. The bridging ADA assay served as a reference for the domain-specific approaches.

[0260] The bridging ADA assay is a sandwich enzyme linked immunosorbent assay (ELISA). Generally, antibody preparations were made in dilution buffer (LowCross Buffer, Candor Bioscience GmbH, Wangen, Germany). Reagents and samples were incubated at room temperature with shaking at 500 rpm. Washing steps consisted of three cycles of applying 300 of washing buffer followed by a final aspiration step. Generally, 15 μL of samples were added to 285 μL of incubation mixture containing 0.5 μg/mL each of Duta-original-Bi and Duta-original-Dig in a polypropylene pre-incubation plate. The PC (mAb-anti-hu-kappa M-1.7.10) was processed in parallel. After 1 h of incubation, 100 μL of each sample mixture were transferred in duplicate to a SA-MTP for 1 h of further incubation. After washing with PBS/Tween to remove unbound material and a final aspiration, 100 μL of the polyclonal anti-Dig-S-Fab-HRP conjugate (12.5 mU/mL) were added. After 1 h of incubation, the SA-MTP was washed three times followed by final aspiration. Then the substrate solution ABTS was added and the HRP catalyzed a color reaction with readout at 405 nm wavelength (reference wavelength: 490 nm). Only samples containing bridged complexes, i.e., antibodies bound to both the Duta-original-Bi and Duta-original-Dig, were able to generate signals. The signal intensities were proportional to the amount of ADAs present. Absorbance values were determined in duplicate wells for each sample. The absorbance values were averaged and accepted if the precision of the mean value was <20% CV. The screening cut point (CP) was evaluated according to Shankar et al., using the 95% percentile applied to 40 drug-naive donor samples. By calculating a plate specific CP (blank of pooled plasma multiplied by a normalization factor), negative samples (<plate-specific CP) and positive samples (>plate-specific CP) were identified. In this context, it should be mentioned that drug tolerance was not considered a crucial parameter. The drug is administered intravitreal, and study samples in this equation are taken at time points with low drug levels, i.e. prior to subsequent dosing. Since Fab-original levels in the samples were B.L.Q., an evaluation of drug tolerance was considered unnecessary.

Example 2

Generation of Bispecific and Variant Fab Fragments

[0261] Synthetic genes encoding the bispecific construct Duta-original and the monospecific, partly germlined constructs Germ1-control and Germ2-control as well as the fully germlined, non-binding construct GermGerm-control were purchased from Geneart.

[0262] Synthetic antibody genes were cloned into a vector, which is a bicistronic vector for periplasmic expression of Fab fragments in E. coli, and bears an expression cassette comprising a LacZ promoter, LC ribosome binding site, LC signal peptide, Vk variable domain, Ck constant domain, HC ribosome binding site, HC signal peptide, VH variable domain, IgG1 CH1 constant domain and IgG1 upper hinge.

[0263] Plasmids encoding Fab fragments were transformed into TG1 E. coli cells (Zymo Research), and single colonies were pre-cultured at 37° C. in TB media, supplemented with 2% glucose to inhibit antibody expression. Upon reaching lag phase, pre-cultures were diluted in Erlenmeyer shaker flasks into TB expression media supplemented with a final concentration of 0.05% glucose, and upon reaching log phase expression cultures were induced by supplementation with IPTG to a final concentration of 1 mM. Fab fragments were expressed for 16 hours at 30° C., and culture supernatants were clarified by centrifugation.

[0264] In order to ensure only full-length Fab fragments with intact heavy and light chain constant domains were used in the ADA analysis, the Fab fragments were double affinity purified in two steps, firstly using CaptureSelect IgG-CH1 resin and secondly using CaptureSelect Kappa XL resin, both purchased from GE Healthcare. The protein concentration of purified Fab fragments was determined by spectrophotometry, using absorbance at 280 nm.

Example 3

Domain Detection Assay (DDA) for Detection of Domain-Specific ADA Responses

[0265] The domain detection assays (DDAs) represented a four-step sandwich enzyme linked immunosorbent assay (ELISA).

[0266] Each 3.33 μg/mL and 150 μL per well of Fab-original, Germ1-control, Germ2-control, GermGerm-control, Ranibizumab and a non-binding human Fab were incubated for 1 hour on a Nunc MaxiSorp™ MTP as capturing reagents. After washing the plate 3 times with PBS/Tween to remove unbound material, 150 μL assay buffer was used to block the plate for 30 minutes to avoid unspecific binding from the samples in the next step. After washing the plate, 3 μL of each sample was diluted with 297 μL assay buffer and 100 μL per sample in duplicate were incubated for one hour. The capture controls (recombinant antigen-1-Dig and recombinant antigen-2-Dig) were processed in parallel. The pre-dose sample of each animal represented the negative control. The digoxygenylated monoclonal anti-cynomolgus IgG was used as detection reagent (100 μL/well). Wells with capture controls were filled in with 100 μL assay buffer. The polyclonal sheep antibody Fab fragment against Dig conjugated to horseradish peroxidase (HRP) (pAb-Dig-S-Fab-HRP) was applied as a second detection reagent in all ADA assays (100 μL/well). ABTS was added to each well and the ensuing color reaction was monitored by photometrical readout at 405 nm (reference wavelength 490 nm).

[0267] To differentiate between positive and negative samples, a cut-point (CP) determination was performed with the set of 8 pre-dose samples of the cynomolgus monkey study for each of the 5 DDAs.

TABLE-US-00001 TABLE Capture control assays used for the 5 different DDAs with expected positivity/negativity. Germ<AG1 > Germ<AG2> GermGerm- Controls F(ab)<AG1/AG2> control control control Ranibizumab Capture control 1 (AG1-Dig) + + − − + Capture control 2 (AG2-Dig) + − + − − Capture control 3 (anti-hu-kappa-Dig) + + + + +

[0268] All samples are measured in 3 dilutions (1 to 100/1 to 1000/1 to 10000). The samples were diluted in Low Cross Buffer™ and measured in parallel in all 5 assays.

Example 4

Capture Control Assays (CCA)

[0269] In order to test the reproducibility of the coating procedure for the 5 different DDAs, three different binding control assays were established. First, Nunc MaxiSorp™ plates were coated as described in the DDAs. Subsequently, digoxygenin labeled anti-human kappa light chain antibody or digoxygenin labeled antigen-1 or antigen-2 were incubated at a concentration of 50 ng/mL. A polyclonal sheep antibody Fab fragment against digoxygenin conjugated to HRP (pAb<Dig>S-Fab-HRP) was applied as a second detection reagent. ABTS was added to each well and the ensuing color reaction was monitored by photometrical readout at 405 nm (reference wavelength 490 nm). These assays were used to check the reproducibility of the coating and the functionality of the coated proteins. In addition, during the sample measurement, the capture control 1 assay was performed on the measuring plate as quasi-positive control (“surrogate PC”) to ensure the comparability of individual measurements on different plates. The signal of this CCA was monitored and the assay was stopped at signal intensities of 1.8-2.2 OD.sub.405 nm.

Example 5

Characterization of the Domain Detection Assay (DDA) According to the Invention

[0270] The samples used were taken from a preclinical tolerability study in cynomolgus monkeys with the bispecific anti-antigen-1/antigen-2 Fab as a drug. Based on ADA positivity in the initial bridging assay for ADA screening (Wessels, U., et al., Bioanal. 10 (2018) 803-814), 16 plasma samples from 8 different animals treated with bispecific anti-antigen-1/antigen-2 Fab were selected for ADA characterization with the method according to the current invention. From each animal, the pre-dose sample as a specific negative control and the Day-35 post dose sample was tested. Pre dose cynomolgus monkey citrate-theophylline-adenosine-dipyridamole (CTAD) plasma samples from each animal were used as negative control for the DDA.

Assay Qualification

[0271] The example of the method according to the invention described in this example is for characterization of the ADA response and is based on the result of 5 individual DDAs being analyzed in relation to each other (see FIG. 6 and FIG. 7 showing the setup used in this example and the expected positivity pattern). This requires the generation of 5 distinct positive controls. This is a general challenge with DDAs and not limited to the assay described in this publication (Gorovits, B., et al., J. Immunol. Meth. 408 (2014) 1-12; Hock, M. B., et al., AAPS J. 17 (2015) 35-43). For an optimal qualification of these assays, specific positive controls would have to be generated, which can be achieved by immunization of mice with the complete compound and a good screening concept. Another possibility would be the use of already available antibodies against individual domains (Stubenrauch, K., et al., J. Pharm. Biomed. Anal. 114 (2015) 296-304). Due to the characteristics of the Fab therapeutic used in this example, the second approach was not possible, and unfortunately specific immunizations are very time consuming and laborious. Without positive controls, a standard approach for qualification of these assays was not possible. Therefore, a combination of capture control assays and an assay specific cut point determination was used as practical approach leading to reliable data.

Capture Consistency

[0272] The data from the capture control assay allows for the determination of a precision value. This precision value reflects the reproducibility of the coating. The anti-kappa IgG antibody based CCA is a measurement showing efficiency of Fab surface coating. This value was determined for all 5 individual DDAs. The AG1 and AG2 based CCAs also show whether the CDRs of the bound Fab fragments are accessible for either the antigen or potential ADAs. Precision data shows very little variation of 1 to 3% for all 3 CCAs in 3 different intra assay runs.

TABLE-US-00002 TABLE Precision data for all 3 capture control assays in the five Domain detection assays. Germ<AG1> Germ<AG2> GermGerm- Control (50 ng/mL) F(ab)<AG1/AG2> control control control Ranibizumab Capture control AG1 -Dig Intra-assay precision (% CV) 1% 1% —* —* 1% Inter-assay precision (% CV) 5% 10%  —* —* 4% Capture control AG2-Dig Intra-assay precision (% CV) 2% —* 2% —* —* Inter-assay precision (% CV) 3% —* 9% —* —* Capture control anti-human-kappa-Dig Intra-assay precision (% CV) 1% 3% 2% 1% 2% Inter-assay precision (% CV) 2% 2% 2% 2% 4% *Construct without binding competency for the respective target

[0273] It can be seen that the amount of coated antibody as well as their ability to bind the antigens is constant. This robustness in the coating is a prerequisite for the comparability of the data. This data also shows that the used constructs all have or do no longer have the desired functionalities.

DDA Cut Point Determination

[0274] A study specific cut point was calculated for all 5 DDAs to assess the ADA positivity by measuring the corresponding pre-dose samples (n=8 animals). For the calculation of the cut point, the 99% percentile was used. Multiplicative normalization factors as well as corresponding cut points for all assays are listed in the following Table.

TABLE-US-00003 Germ<AG1 > Germ<AG2> GermGerm- Parameter F(ab)<AG1/AG2> control control control Ranibizumab MRD 100 100 100 100 100 Capture control 2 (AG2-Dig) + − + − − Capture control 3 (anti-hu-kappa-Dig) + + + + +

[0275] The assay range was determined by the use of digoxygenin-labelled antigens (antigen-1 and antigen-2). At a concentration of 50 ng/ml antigen-1-Dig or antigen-2-Dig the assay was developed until a signal of 1.8-2.2 OD.sub.405 nm was reached. This signal range was used as a criterion to stop the reaction in the real sample measurement. Real samples were diluted 1:100 in assay buffer. In case signals above 2.2 OD.sub.405 nm were reached, samples were further diluted in 1:10 steps.

[0276] A cut-point determination was performed with the respective pre-dose samples for each individual construct.

TABLE-US-00004 TABLE Cut-point determination: signals of the pre-dose samples in the DDA assays were obtained. A 99% percentile was used to calculate the assay specific cut-point as shown below. Non-binding predose-samples Fab-original Germ1-contol Germ-2 control GermGerm Ranibizumab human Fab Animal1 0.113 0.141 0.153 0.051 0.032 0.046 Animal2 0.080 0.192 0.072 0.050 0.031 0.054 Animal3 0.070 0.142 0.090 0.082 0.034 0.055 Animal4 0.069 0.151 0.069 0.069 0.037 0.058 Animal5 0.076 0.150 0.062 0.048 0.041 0.092 Animal6 0.078 0.110 0.067 0.045 0.035 0.050 Animal7 0.066 0.155 0.063 0.074 0.041 0.067 Animal8 0.081 0.133 0.064 0.084 0.045 0.081 mean (0D405nm) 0.079 0.146 0.080 0.063 0.037 0.063 SD (0D405nm) 0.015 0.023 0.031 0.016 0.005 0.019 CP = mean*2,576 SD 0.117 0.206 0.159 0.104 0.049 0.104 (0D405nm)

Drug Tolerance of DDAs

[0277] The samples that were used are from late time points (day 35 after drug administration). Therefore, the measured levels of residual drug are very low and no impact of residual drug on these assays is to be expected. To discuss this theoretically, it has to be said that in this assay the labelled drug is coated with a very high concentration, which is in direct competition with the residual drug in the samples. The theoretical drug tolerance should therefore be at least as good as with a standard bridging assay. If necessary, this can be improved in a similar way to bridging ADA assays. For example, the samples can be pre-treated with acid to dissolve existing drug-ADA complexes (see, e.g., Kavita, U., et al., J. Immunol. Meth. 448 (2017) 91-104).

[0278] The assay has been shown to be highly drug tolerant, as mass concentration of the capture antibody/domain was orders of magnitude higher as compared to the residual drug levels present in the samples at day 35 post dose. This was also confirmed by titration experiments, showing that specific ADA recognition signals were still highly positive, when the drug was diluted out of the sample.

Sample Analysis with Domain Detection Assays

[0279] The numerical data for all samples with all 3 dilutions used can be found in the following table.

TABLE-US-00005 TABLE Numeric ADA domain detection assay results. Values used for graphical illustration in FIGS. 8 to 11 are depicted in bold. mean signals OD [405 nM] Animal 1 Animal 2 Predose Postdose Predose Postdose Predose Postdose Predose Postdose ADA response against (1:100) (1:100) (1:1000) (1:10000) (1:100) (1:100) (1:1000) (1:10000) Bispecific F(ab) <AG1/AG2)> 0.080 1.756 1.704 1.505 0.066 1.738 1.663 1.418 Germ<AG1>-control 0.192 1.824 1.810 1.389 0.155 1.797 1.725 1.259 Germ<AG2>-control 0.072 1.886 1.850 1.513 0.063 1.860 1.753 1.163 GermGerm-control 0.050 1.906 1.740 0.707 0.074 1.666 0.714 0.112 Ranibizumab (commercial Fab) 0.031 1.951 1.888 0.831 0.041 1.674 0.611 0.082 mean signals OD [405 nM] Animal 3 Animal 4 Predose Postdose Predose Postdose Predose Postdose Predose Postdose ADA response against (1:100) (1:100) (1:1000) (1:10000) (1:100) (1:100) (1:1000) (1:10000) Bispecific F(ab) <AG1/AG2)> 0.113 0.797 0.171 0.035 0.070 1.687 1.581 0.992 Germ<AG1>-control 0.141 0.557 0.078 0.029 0.142 1.690 1.537 0.563 Germ<AG2>-control 0.153 0.588 0.089 0.032 0.090 1.770 1.590 0.645 GermGerm-control 0.051 0.042 0.030 0.028 0.082 0.098 0.033 0.026 Ranibizumab (commercial Fab) 0.032 0.032 0.022 0.022 0.034 0.160 0.026 0.019 mean signals OD [405 nM] Animal 5 Animal 6 Predose Postdose Predose Postdose Predose Postdose Predose Postdose ADA response against (1:100) (1:100) (1:1000) (1:10000) (1:100) (1:100) (1:1000) (1:10000) Bispecific F(ab) <AG1/AG2)> 0.076 1.769 1.751 1.246 0.078 1.758 1.679 0.905 Germ<AG1>-control 0.150 1.755 1.642 0.856 0.110 1.742 1.478 0.468 Germ<AG2>-control 0.062 2.009 1.799 0.789 0.067 2.011 1.647 0.510 GermGerm-control 0.048 1.215 0.175 0.039 0.045 1.296 0.170 0.039 Ranibizumab (commercial Fab) 0.041 1.312 0.244 0.037 0.035 1.509 0.354 0.045 mean signals OD [405 nM] Animal 7 Animal 8 Predose Postdose Predose Postdose Predose Postdose Predose Postdose ADA response against (1:100) (1:100) (1:1000) (1:10000) (1:100) (1:100) (1:1000) (1:10000) Bispecific F(ab) <AG1/AG2)> 0.069 1.571 0.879 0.150 0.081 1.775 1.192 0.269 Germ<AG1>-control 0.151 1.548 0.601 0.097 0.133 1.649 0.996 0.185 Germ<AG2>-control 0.069 1.405 0.314 0.053 0.064 1.460 0.404 0.066 GermGerm-control 0.069 0.058 0.031 0.029 0.084 0.083 0.031 0.027 Ranibizumab (commercial Fab) 0.037 0.084 0.026 0.023 0.045 0.100 0.025 0.019

[0280] Different dilutions were selected for the comparison of the data with each other as shown in FIGS. 8 to 11. The linear range of the chromogenic ABTS substrate is limited and the samples must be diluted to different extents to make the signal heights clearly distinguishable. For 6 of 8 animals, the 1 to 1000 dilution was used. In animal 3, the immune response was weak and the 1 to 100 dilution was chosen and in animal 1 the 1 to 10000 dilution was chosen because the immune response was very strong.

[0281] Four ADA response patterns (in the following termed 8, 9, 10 and 11 in line with the corresponding results in FIGS. 8 to 11) can be discriminated which are representative for all eight ADA positive animals of the study.

[0282] Pattern 8: Animals 1 and 2 reveal a mixed ADA response against different parts (CDRs and backbone) of the molecule. A strong response against the fully binding competent <AG1/AG2> bispecific F(ab), the Germ<AG1>-control and the Germ<AG2>-control can be seen. A moderate response was detected against the fully germlined control construct (GermGerm-control) and Ranibizumab, indicating the presence of ADAs against the constant regions of the molecule. The majority of ADAs appears to be directed against both CDRs in similar proportions.

[0283] Pattern 9: Animals 3 and 4 show only a recognition of the CDR regions with no ADAs against the constant regions. In animal 3, the signal against both CDRs here is lower than in Pattern A, suggesting a more moderate ADA response, whereas animal 4 reveals a similar signal strength as Pattern A. In these two animal also, a similar number of ADAs against both CDRs appear to exist.

[0284] Pattern 10: Animals 5 and 6 are similar to pattern 8, indicating a mixed response against both CDRs and the constant regions. However, a much lower proportion of ADAs against the constant regions of the molecule was detected as compared to pattern 8.

[0285] Pattern 11: Animals 7 and 8 show no ADAs against the constant regions and only against the CDRs. Unlike Pattern 9 and all other animals, we see a different distribution of ADAs between the two CDRs against antigen 1 and 2. There seem to be more ADAs against the AG1 binding site.

[0286] Overall, all pre-dose samples were ADA negative in all five DDAs. Therefore, we conclude that all observed immunogenicity responses from the Day-35 post-dose cynomolgus study samples are treatment-related ADA responses.

[0287] The signals of the pre-dose samples were also very well comparable with small variations in all 5 DDAs. Thus, these values were used to calculate DDA specific in study cut points (see above: DDA cut point determination). For this proceeding, the number of baseline samples used is relevant and the recommendation (Shankar, G., et al., J. Pharm. Biomed. Anal. 48 (2008) 1267-1281; Amaravadi, L., et al., Bioanal. 7 (2015) 3107-3124) is to use at least 50 pre-dose samples. This is not possible with significantly lower animal numbers in many preclinical studies such as this. Nevertheless, given the exploratory nature of this measurement, this approach has been chosen.

[0288] Summarizing the above, with the method according to the current invention all eight animals were found to show ADAs against the CDRS and only in two of eight animals, these responses were slightly different. Due to this mixed immune response, it is not likely that one of the two CDR regions of the bispecific molecule was predominantly responsible for the observed immune responses. It was not observed that individual domains exhibit strongly different immunogenicity.

[0289] One more point that had to be addressed was the question whether the engineered bispecific F(ab) fragment used here, with its slightly different structure to a wild type Fab fragment, bears a higher immunogenicity risk. Four out of eight animals showed antibodies against the constant regions. In no sample higher signals in the DDA with the Germ-Germ variant were found compared to the DDA with Ranibizumab. This would be evidence of an immune response against a neoepitope only present on the Germ-Germ variant which does not exist on Ranibizumab, an ocular therapeutic known for its low immunogenicity, as shown in several clinical studies (Figurska, M., et al., Klin. Oczna. 112 (2010) 147-150).

[0290] Based on the above findings, all immunogenicity responses from the above cynomolgus study are treatment-related ADA responses. The dual/bispecific Fab specificity of the treatment-induced ADAs was confirmed. Overall, the immunogenicity response is directed against the HVRs and the constant Fab part. In 50% of the animals, a purely “monospecific” response against the HVRs is observed, whereas in the other half a mixed response against both HVRs and backbone is seen.

[0291] No ADA response against the constant backbone without HVR contribution is observed. Notably, anti-drug antibodies directed against the constant part of the bispecific Fab are cross-reactive to classical human Fab molecules like the non-binding (DP47) control molecule and Ranibizumab. This is an important result towards a potential human application, as there is no hint for molecule-intrinsic immunogenic properties in the bispecific Fab.