The characterization of a humanized rabbit anti-TNFα monoclonal antibody

Tumor necrosis factor alpha (TNFα) is now regarding as a key role in the pathogenesis of immune-mediated disease such as Rheumatoid arthritis (RA), Crohn's Disease, Psoriatic arthritis and Plaque Psoriasis. HERE we have successfully developed an anti-hTNFα monoclonal rabbit antibody(HZ3M) with high binding and neutralizing activity based on RabMAbs platform. Rabbit hybridomas, immunized subcutanrously with 0.4 mg human TNFα, were generated from the rabbit splenocytes and a total of 142 hybridoma clones with specific binding to human TNFa were obtained. The anti-TNFa RabMAbs showed better neutralizing activity and higher antigen binding affinity compared to Humira and Remicade, the elimination phase half-life 58.2h respectively. In vivo efficay studies, normal mice or human TNF-alpha transgenic mice were injected with 1.0 mg/kg Humira (positive control), HZ3M at 0.33?1.0 or 3.0 mg/kg, or solvent (negative control), showed that HZ3M is able to bind and neutralize hTNFα in transgenic and normal mice as well as normal rabbits.Clearly dose-dependent response can be determined. Compared to marketed anti-TNFa drug Humira, the efficacy of HZ3M is seems to show significant longer holding time.Our observations indicate that the HZ3M derived from RabMAb preclinical safety study, and might have a therapeutic role in RA treatment.

fused in-frame to the constant region of H chain at BspE1 site and cloned into another pcDNA 3.1 or pCEP4 vector at BamH I and Pme I sites. Recombinant RabMAbs were expressed in HEK 293 or CHO cells by cotransfection of the L and H chain plasmids. For humanized anti-human TNFa RabMAb cloning and expression, the humanized VH was fused in-frame to the constant region of human IgG1 H chain at Nhe I site and clones into pTT5 vector 4 at Hind III and Not I sites, and the humanized VK to the constant region of human IgG kappa chain at BsiWI site and into another pTT5 vector at Hind III and Not I sites. The L and H chain plasmids were co-transfected into 293-6E cells 4 and supernatants were harvested 5 days after transfection. The recombinant RabMAbs in the supernatants were purified through protein A column, dialyzed in PBS buffer and quantified at OD280 nm.

Antigen Binding ELISA
For the direct antigen binding ELISA assay, recombinant human TNFa (Shanghai Primegene Bi-Tech ) was coated onto microtiter plates (Greiner Bio-One) in 0.05M carbonate-bicarbonate buffer overnight at 4°C, and non-specific binding was blocked with 1% BSA/TBST (Tris-Buffer Saline Tween-20) for 1 hour at room temperature.
Rabbit anti-human TNFa hybridoma supernantants or recombinant antibodies were added to the plates for binding to the antigen.
Alkaline phosphatase-conjugated goat anti-rabbit (Jackson ImmunoResearch) and goat anti-human IgG (H+L) antibody (Pierce/Thermo Scientific) were added at a dilution of 1:5000 to detect bound rabbit and human antibodies, respectively. The ELISA plates were then processed for pNPP substrate development and the optical density was measured at 405 nm with an ELISA plate reader (Labsystems Multiskan Ascent Photometric plate reader).

L929 TNFa Induced Cell Cytotoxicity Assay
L929 cells (ATCC, 3.5 x 10 4 /well) were seeded into 96-well assay plates in 100 ul of Dulbeco's modified Eagle's medium (Cellgro) supplemented with 2% fetal bovine serum (Hyclone, heat inactive) and incubated overnight at 37 C, 5% CO2 in a humidified incubator. Recombinant human TNFa (R&D, final assay concentration, 1ng/ml) was preincubated with or without anti-TNFa hybridoma supernatant, parental or humanized recomboinant RabMAb or control antibody at 37 C for 2 hrs, and then 50 ul of the preincubated mixture was transferred to the assay plates after addition of 50 ul of actinomycin D at a final concentration of 1ug/ml (Sigma).
Incubation of the assay plate was continued for 18 h followed by addition of 20 µl of a MTT Reagent (R&D). After another 4 hrs incubation, 100 µl of Detergent Reagent (R&D) was added.
Next day, the absorbance of each well was determined with Wallac 1420 Victor plate reader at 570 nm. Survival was calculated as the percentage of the staining value of untreated cultures. Percent cytotoxicity is the difference between control (100%) and percent survival. To stain, the cells were permeabilized for 5 mins with 0.1% Triton-100 /TBS (Tris-Buffer Saline) at room temperature, blocked for 1 hr in 1% BSA in TBS and then incubated for 1 hr with mouse anti-NF-KB p65 antibody (BD biosciences) at 1:1000 dilution in 1% BSA. After three washes in TBST, anti-mouse-HRP antibody (ImmunoVision Technologies) was added and the cells were incubated with the secondary antibody for 2 hrs. The assay plate was then processed for DAB (ImmunoVision Technology) substrate development and cell stain was observed under a microscope.

Humanization of TNFa RabMAb
A phylogenetic tree was built by alignment of VH and VL amino acid sequences of selected RabMAbs. Sequence alignment and phylogenetic analysis were performed using the ClustalX software. The parental RabMAb and the most closely related human germline sequence were then aligned. RabMAb residues were changed to match the human germline sequence except for the residues which are known to be likely critical in antigen binding. By MLG analysis 5,6 , some of the critical residues subjected to change during the in vivo maturation process were identified and

Analysis of Glycosylation on Anti-TNFa RabMAbs
Purified recombinant antibodies in glycoprotein denaturing buffer (0.5% SDS and 1% β-mercaptoethanol) were heated at 100°C for 10 min, and then sodium phosphate, pH 7.5 and NP-40 were added to a final concentration of 50 mM and 10%, followed by addition of PNase F (New England Bioland). Deglycosylation reaction was performed at 37 C for 10 min. The glycosylated and deglycosylated proteins were separated on a 4-20% Novex Tris-Glycine gel (Invitrogen) by reducing, denatured gel electrophoresis and visualized by Gelcode (Pierce) blue stain.

Affinity Measurement of Anti-TNFa RabMAbs by competition ELISA
The optimal coated antigen and antibody concentrations used in competition ELISA were pre-determined by direct antigen binding ELISA (see Antigen Binding ELISA for the detailed method

Affinity Measurement of Anti-TNFa RabMAbs by BIAcore
Affinities measured by BIAcore were performed on BIAcore 3000 (BIAcore, Inc., Uppsala, Equilibrium dissociation constants were then calculated from the rate constants (KD = koff/kon).

Pharmacokinetics study
The HZ3M (also named as HZD-RabMAb above) and anti-human TNFα RabMAb was produced as described above. Humira control was purchased from Abbott Laboratories, lot number 41366SP40. Eight weeks old male C57BL/6 mice, weight 20 ± 2g, were purchased from Vital River Laboratories. The animals were given free diet during the study.
The serum drug concentration was measured by ELISA (enzyme-linked immunosorbent assay) 7 .
The 96-well microtiter plates were coated with human recombinant TNFa (Beijing Biosea Biotechnology Co.,LTD) as the capture reagent. After the animal blood samples or Humira were incubated with the coated plate, goat-anti-human IgG-HRP (from Zhongshan Goldenbridge Biotechnology Co.,LTD) was used as the detecting reagent.
The pharmacokinetics analysis was carried by using the single-compartment open model with first order absorption to calculate elimination phase half-life 8,9 Efficacy Study Testing samples: Humanized rabbit antiTNFa monoclonal antibody HZ3-M was produced and purified as described above. Humira was purchased from Abbott Laboratories, lot number 65204LJ40.
Method: The animals were divided into six groups as shown in table 1. The HZ3M (resting agent), Humira (positive control), or solvent (negative control) were injected i.p. at 0.1ml/10g body weight volume. The HZ3M (testing agent), Humira (positive control), or solvent (negative control) were given to the animals by i.p. injection, once every three days for six weeks, at 0.1ml/10 g body weight volume.
The RA symptoms were analyzed and scored once a week, based on the limbs swollenness, toes curvature, toes separation, toe joints and ankle joints mal-shape 10 . The mice survival rate was counted 24 and 48 hours after the treatment.
Fever study: Animal: Closed colony of New Zealand rabbits, 50% male and 50% female, were purchased from Shenyang Pharmaceutical University Experimental Animal Center.
Treatment: The animals were divided to four groups (normal, negative control, positive control, and HZ3-M) and subjected to various treatments via ear vein injection as shown in table 3. The animal Rectal temperature was measured every 30 min within 4 hours after the treatment.

Generation of Anti-human TNFa RabMAbs with Neutralizing Activity
Three rabbits were immunized with full length human TNFa expressed in E. coli. Antisera from the rabbits were tested for their activity to neutralize TNFa induced cytotoxicity in L929 cells.
Antisera from all the rabbits showed the neutralizing activity (Fig. 1). Rabbit hybridomas were generated from the rabbit splenocytes and a total of 142 hybridoma clones with specific binding to human TNFa were obtained. Forty-four RAbMAbs that neutralize TNFa induced cytotoxicity in L929 cells were identified from screening of the 142 clones in the cell-based assay (Fig 2a). The neutralizing activity was confirmed by their capability to block NFkB nuclear translocation (Fig   2b). To rank the neutralizing antibodies, cDNAs for the neutralizing antibodies were clones and expressed and dose-responsive neutralization of TNFa-induced cytotoxicity assay was performed with the recombinant antibodies. IC50 was calculated and summarized in Table 4. Antigen binding affinity for some of these antibodies was also measured (Table 4). Currently marketed anti-TNFa drugs, Humira and Remicade, were used as controls. Some of anti-TNFa RabMAbs showed better neutralizing activity and higher antigen binding affinity compared to Humira and Remicade.   NFkB in the cells was probed with mouse anti-NFkB antibodya and its location was detected by anti-mouse HRP-polymer and DAB stain.
The cells without (No TNF) or with (TNF alone) TNFa treatment or with TNFa treatment plus control antibody (Cntrl Ab) were run in parallel.

Sequence and Structure Analysis of Anti-TNFa Neutralizing RabMAbs
Phylogenetic trees based on variable region amino acids sequences of light (VK) and heavy (VH) chains of the neutralizing RabMAb are shown in Figure 3a. There are seven lineage-related groups identified. In the same lineage-related group, all the VHs are in the same cluster, so as all the VKs.
Sequence alignments of the lineage related groups are shown in Figure 3b (VH) and 3c (VK).
RabMAbs in a lineage-related group have the same length of CDRs and their sequences in the heavy chain CDR3 are very similar as well. For some lineage groups, the CDR2 sequence is more diverse than the CDR3 in the heavy chains ( Fig. 2b Group A, B and G). In addition, some significant sequence variations were observed in the CDR1, CDR2 or CDR 3 of the VKs (Fig. 3c, Group A, D and G), suggesting the light chain contributes to the RabMAb activity.

Humanization of Anti-TNFa RabMAb Clone 858
In the lineage-related group F, all antibodies have a good neutralizing activity and antigen binding affinity (Table 4). In addition, the group F contains the most RabMAbs among the lineage-related groups, and does not have uncommon sequence and structure features that were found in other lineage-related groups or RabMAbs (Table 5). One RabMAb, Clone 858, in the group F was selected as a therapeutic lead candidate for humanization and further studies.
Clone 858 was humanized into a human IgG1 subclass with kappa light chain by using the MLG (mutational lineage-guided) method 5 . First, the heavy (VH) and light (VK) variable region sequences of Clone 858 were blasted against the human germnline VH and VK database. The closest human germline sequences, VH4-b/JH3 and VKIO2/JK4, were identified as the template for Clone 858 humanization ( Fig. 4a and b). Second, antibody sequences within the Clone 858 lineage group ( Fig. 3c and 3d, group F) were aligned to the template ( Fig. 4a and b). Third, the rabbit residues in the frameworks potentially involved in CDR contacts or inter-chain contacts were identified based on the knowledge from human and mouse antibodies. These residues, marked by "#", were not changed (1_, 2Q, 7F, 9L, 37V, 40A, 43S, 67S, 71R, 78V, 82M, 91F, 105P in VH and 1A, 2Y, 22K, 43P in VK) (Fig 4a and 4b). At CDR2 position 59 of VH, F is interchangeable to Y within the group, so it was mutated to the human residue Y at the position marked by * in Figure 4a. The same rule was applied to CDR1 position 28 of VK. N was changed to S at the position marked by * in Figure 4b. From the phylogenetic analysis, more residues in the frameworks were also considered not critical to structural activity of the antibody and humanized as marked by "$" ( 46Q -E of VH and 70E -D of VK). Residues at position 3D and 7T were substituted with human residue Q and S, respectively, as marked by "^" because they are away from the CDR loops and facing solvent (based on homology modeling). The parental Clone 858 antibody was 65% identical to the human germline frameworks (VH/VK). After humanization, the sequence identity is 90%.

Comparison of Humanized Anti-TNFa RabMAb Clone 858 (HZD-RabMAb) to the Parental RabMAb and Marketed anti-TNFa drugs
To investigate whether HZD-RabMAb retains the properties of the parental antibody, we first compared N-glycosylation on the HZD and parental RabMAb and marketed anti-TNFa drugs.
The HZD RabMAb, the parental, Humira and Remicade were treated with or without N-Glycosidase F (PNGase F), and analyzed by gel electrophoresis. Like Humira and Remicade, without PNCase treatnment, HZD-RabMAb only showed one heavy chain band (Fig. 5, at 50 KD).
However, the parental RabMab showed two heavy chain bands, which is likely due to additional glycosylation site on the heavy chain CH1, predicted by sequence analysis, and partially gylcosyated product (Fig. 5, at 50 KD). Treatment with PNGase resulted in one deglycosylated heavy chain product for the HZD-RabMAb as well as for Humira, Remicade and the parental RabMAb (Fig.5, compare + PNGase to -PNGase). These results indicated HZD-RabMAb is the same as Humira and Remicade, but differentiated from the parental in glycosylation. In addition to glycosylation, RabMab kappa light chain has a Cys-Cys pair between the framework 3 and the constant region (CK) 13 . Replacement of rabbit CK with human CK and substitution of the Cys with Pro in the framework 3 (Fig. 4b) for humanization could affect the light chain folding and make antibody structurally fragile. To address this issue, thermal stability of HZD-RAbMAb was examine and compared to Humira and Remicade. The thermal stability profile is shown in Fig. 6. HZD-RabMAb started its structure melting at about 74°C as Humira and Remicade, indicating HZD-RabMAb is structurally as stable as the two marketed anti-TNFa drugs. Biological activity of HZD-RabMAb was tested for its TNFa neutralizing activity and binding activity. The relationship of L929 cell viability and antibody concentration is shown in Fig. 7.
IC50 for HZD-RAbMAb is similar to that for the parental RabMab , but 1.5 times better than that (4.0 vs 9.0 ng/ml) for Remicade, indicating a high potency of HZD-RabMAb in neutralizing TNFa-induced cell toxicity. The antigen binding affinity measured by BIAcore is summarized in Table 6. Humanization of the RabMAb didn't change its binding affinity to TNFa. Under the measurement conditions, HZD-RAbMAb showed picomolar affinity to the antigen.  The animal arthritis symptoms were examined once a week after the first treatment and scored based on the limbs swollenness, toes curvature, toes separation, toe joints and ankle joints mal-shape, as described in the Materials and Methods. Humira was used as the positive control.
The effect of HZ3M on normal and arthritis mice (Taconic 1006 mice) is shown in table 8 and figure 9. Compare with negative control mice, *p<0.05, **p<0.01 Figure 9. The effect of HZ3M on normal and arthritis mice The arthritis symptoms were significantly reduced in Humira and HZ3M treated Taconic 1006 mice after two weeks of treatment.
Mice muscle strength study: The mice were subject to the muscle strength test once a week on the Rota-Rod Instrument during the treatment. The holding time was recorded, above 60 seconds were counted as 60 second.
During the treatment period, the Taconic 1006 mice suffer from the arthritis progression.    Compare with negative control mice, **p<0.01

Mice survival study
The Normal mice did not survive from the deadly dose of hTNFα with effect enhancer D-galactosamine. With the protection of anti-TNFα monoclonal antibodies, mice death rate is decreased. The anti-TNFα monoclonal antibodies do not just bind to TNFα but also block the interaction between TNFα and its receptors, so the bioactivities of TNFα can be neutralized. In the HZ3M group, the death rates are decreased with those increasing doses. A clear dose-dependent relationship between mice survival and HZ3M administration dose is seen in the table 11.

Fever study
hTNFα is able to increase the temperature of rabbit with D-Gln. The anti-TNFα antibodies have neutralized the bioactivities of TNFα after administrated to rabbits. And like those in mice survival study, dose-dependent behaviour is found in the rabbit fever study as well (table 12 and figure11). Compare with negative control mice, *p<0.05, **p<0.01 Figure 11 the effect of HZ3M on rabbit fever

DISCUSSION AND CONCLUSIONS
In this study, we sought to develop a new anti-TNFa monoclonal drug, with high binding and neutralizing activity based on RabMAbs platform. Using MLR strategy, the humanization version of anti-TNFα RabMAb HZD-RabMAb (also named as HZ3M) is produced. The humanized RabMAb inherited most the activities of binding and neutralizing and the thermal stability. In vivo PK study, the behaviour of HZ3M is similar with its RabMAb version and two marketed anti-TNFa drugs. As expected with any formulation of HZ3M, it will be more safety to develop a humanied antibody, multiple experiments from our laboratories have primarily observed the effects of HZ3M on RA.In vivo efficay studies show HZ3M is able to bind and neutralize hTNFα in transgenic and normal mice as well as normal rabbits.
Clearly dose-dependent response can be determined, compared to marketed anti-TNFa drug Humira, the efficacy of HZ3M is seems to be better, HZ3M was easy to see the advantages of the new antibody, the hafe-life was 167 hours, longer than drugs on the market Humira, which was 105 hours as a contral drug in our Pharmacokinetics study. The survival rate reasch showed a clear dose-dependent relationship between the mice survival rate and HZ3Madministration dose, in 0.1μg/g dose, it seems better survival rate than Humira, further more, in 0.13 μg/g dose, the survival rate was up to 100%.
The survival rate further supports the administration of HZ3M to patients RA.These results encourage the further development of HZ3M derived from RabMAb to preclinical safety study.