Antagonistic TNF Receptor One-Specific Antibody (ATROSAB): Receptor Binding and In Vitro Bioactivity

Background Selective inhibition of TNFR1 signaling holds the potential to greatly reduce the pro-inflammatory activity of TNF, while leaving TNFR2 untouched, thus allowing for cell survival and tissue homeostasis. ATROSAB is a humanized antagonistic anti-TNFR1 antibody developed for the treatment of inflammatory diseases. Methodology/Principal Findings The epitope of ATROSAB resides in the N-terminal region of TNFR1 covering parts of CRD1 and CRD2. By site-directed mutagenesis, we identified Arg68 and His69 of TNFR1 as important residues for ATROSAB binding. ATROSAB inhibited binding of 125I-labeled TNF to HT1080 in the subnanomolar range. Furthermore, ATROSAB inhibited release of IL-6 and IL-8 from HeLa and HT1080 cells, respectively, induced by TNF or lymphotoxin alpha (LTα). Different from an agonistic antibody (Htr-9), which binds to a region close to the ATROSAB epitope but elicits strong TNFR1 activation, ATROSAB showed a negligible induction of IL-6 and IL-8 production over a broad concentration range. We further verified that ATROSAB, comprising mutations within the Fc region known to abrogate complement fixation and antibody-mediated cellular effector functions, indeed lacks binding activity for C1q, FcγRI (CD64), FcγRIIB (CD32b), and FcγRIII (CD16) disabling ADCC and CDC. Conlusions/Significance The data corroborate ATROSAB’s unique function as a TNFR1-selective antagonist efficiently blocking both TNF and LTα action. In agreement with recent studies of TNFR1 complex formation and activation, we suggest a model of the underlying mechanism of TNFR1 inhibition by ATROSAB.


Introduction
Tumor necrosis factor (TNF) plays an important role in the development of inflammatory diseases like rheumatoid arthritis, Crohn's disease and the relapsing phases of multiple sclerosis. TNF is a pleiotropic cytokine that is expressed as type-II transmembrane protein (mTNF) on the surface of macrophages, natural killer (NK) cells, B-and T-cells. It is processed into its soluble form (sTNF) by enzymatic cleavage. TNF activates two cell surface receptors, TNFR1 (CD120a) and TNFR2 (CD120b) [1,2,3,4]. While TNFR1 is constitutively expressed on a broad variety of cell types, TNFR2 expression is cell type-restricted, context and stimulus-dependent and found mainly on immune cells, endothelial cells and neurons [5]. In general, stimulation of TNFR1 by sTNF or mTNF leads to pro-inflammatory and proapoptotic signals [6]. In contrast, effective signaling through TNFR2 is only mediated by mTNF [7], resulting in cell proliferation, tissue homeostasis and regeneration [8,9].
Current clinical intervention in the field of inflammatory diseases is focused on the blockade of TNF, employing a soluble TNF receptor-2 fusion protein (etanercept) and anti-TNF antibodies, including infliximab, adalimumab, golimumab, and certolizumab pegol [10,11]. Regardless of their successful clinical use, long-term treatment with TNF blockers is accompanied by a higher risk of tuberculosis (TB) reactivation and serious infections, whereas the effect of TNF blockers on incidence and/or manifestation of malignancies is discussed controversially [12,13,14,15,16]. Counterintuitive were observations that TNF blockade can be associated with development of inflammatory and autoimmune diseases [17,18,19,20], indicating a highly complex regulation of TNF action in vivo.
Selective inhibition of signaling through TNFR1 holds the potential to greatly reduce the pro-inflammatory activity of TNF, while leaving TNFR2 untouched, thus allowing for cell survival, tissue homeostasis and, for the CNS, myelin regeneration [21,22]. This change of concept in the treatment of TNF-mediated inflammatory diseases, from global ligand inhibition to selective receptor blockade, has gained increasing attention [23] and has led to the development of a number of TNFR1-selective inhibitors. For instance, the TNFR1-selective mutein R1antTNF and its PEGylated form (PEG-R1antTNF) were effectively used to treat acute hepatitis, collagen-induced arthritis (CIA), experimental autoimmune encephalomyelitis (EAE), and hyperplasia in different mouse models [24,25,26,27]. A dominant-negative mutein (XENP1595) inhibits TNFR1 selectively by forming inactive complexes with sTNF and was used for the treatment of experimental colitis [28,29,30]. TNFR1 knockdown in mouse models by short hairpin RNA [31] and antisense oligonucleotides [32] led to the amelioration of CIA and reduced liver toxicity caused by radiation-induced TNF production. Furthermore, antibodies directed against TNFR1, such as H398 [33,34,35], represent another promising approach for selective TNFR1 blockage.
In previous studies we transformed a humanized Fab fragment (IZI-06.1) of H398 [36], selectively recognizing human TNFR1, into a whole IgG format [37]. This antagonistic TNF receptor one-specific antibody (ATROSAB) was shown to retain TNFR1 selectivity and to inhibit TNFR1-mediated cell responses such as cell death induction, IL-6 and IL-8 release. In addition, the kinetic constants of the binding to TNFR1 were determined using a quartz crystal microbalance (QCM) system and the epitope targeted by ATROSAB was located to the cysteine-rich domains (CRD) one and two of TNFR1 [37].
Here, we identified critical amino acids within the ATROSAB epitope of TNFR1 and studied in detail kinetic binding constants by QCM as well as functional activities in comparison to an agonistic TNFR1-specific antibody that binds an epitope in close proximity to the ATROSAB epitope. Based on these data and previous results on TNFR1-TNF signal complex formation, we suggest a model of the underlying mechanism of antibodymediated TNFR1 inhibition and activation, respectively.

ATROSAB Lacks Fc-mediated Effector Functions
The Fc region of ATROSAB is modified in order to inactivate effector functions (ADCP, ADCC and CDC) [38]. To verify that binding of Fcc receptors and C1q is blocked in ATROSAB, binding studies with human C1q as well as human soluble Fcc receptor IA (CD64), FccRIIB (CD32b), and IIIA (CD16a) were performed by ELISA. ATROSAB did not bind human sFccRIA (Fig. 1a), and sFccRIIIA (Fig. 1c). Binding to sFccRIIB (Fig. 1b) and C1q (Fig. 1d) was strongly reduced. In contrast, trastuzumab (possessing a human wild-type c1 Fc) used as positive control showed binding with EC 50 values in the described range [39,40]. Immobilization of both antibodies was confirmed with an antihuman Fc-specific antibody (not shown).
In order to demonstrate the depletion of the ADCC effector function, the potential of ATROSAB to elicit killing of TNFR1expressing target cells was tested with calcein-labeled Kym-1 cells. These cells express approximately 3.000 TNFR1 on the cell surface [41]. Flow cytometry studies confirmed binding of ATROSAB to Kym-1 (data not shown). Rituximab, a chimeric anti-CD20 IgG1, and DOHH-2 cells were included as positive control. As expected, the control antibody induced lysis of target cells by ADCC depending on the E:T cell ratio (Fig. 1e). In contrast, ATROSAB induced-cell lysis of Kym-1 cells was significantly reduced when added at 600 mg/ml even at a high E:T ratio (Fig. 1e). Similar results were obtained with lower ATROSAB concentrations (120 mg/ml and further 1:5 dilutions), whereas lower concentrations of rituximab were still able to elicit ADCC in DOHH-2 cells, particularly at the highest E:T ratio (not shown).
The potential of ATROSAB to elicit complement-dependent cytolysis of target cells (CDC) was also tested employing calceinlabeled Kym-1 as target cells. Again, Rituxumab and CD20expressing DOHH-2 cells served as positive control. As expected, the positive control antibody led to cytolysis of the target cells using serum from two different donors (Fig. 1f). In contrast, ATROSAB elicited significantly reduced CDC in antigenexpressing Kym-1 cells (Fig. 1f). Similar results were obtained with lower ATROSAB concentrations (60 mg/ml and further dilutions), whereas lower concentrations of rituximab were still able to elicit CDC in DOHH-2 cells (not shown). Taken together, these data confirm that ATROSAB does not induce a substantial CDC and ADCC activity against TNFR-1 expressing target cells.

Receptor Binding and TNF Blocking of ATROSAB
Binding of ATROSAB to human TNFR1 was analyzed by quartz crystal microbalance (QCM) measurements at high and low densities of immobilized receptor (Fig. 2a, c). At room temperature and a high receptor density, ATROSAB bound with high avidity revealing an apparent dissociation constant (K d value) of 0.12 nM. At lower receptor density, a reduced binding (K d = 1.7 nM) was observed ( Table 1). In addition, binding studies were performed at 37uC to determine binding at physiological temperature (Fig. 2b, d). At high receptor density only minor differences of the kinetic constants were observed compared to 25uC (K d value of 0.19 nM) ( Table 1). However, at lower receptor density the curves obtained at 37uC could only be fitted assuming a biphasic, i.e. mono-and bivalent binding of ATROSAB to immobilized TNFR1. The subtraction of the slowly dissociating part (obtained at high receptor density) yielded an apparent K d value of 14.2 nM for the monovalent interaction, which resulted from a faster off-rate (Fig. 2d, Table 1).
Next, we investigated the inhibitory activity of ATROSAB on TNF binding to HT1080 cells using 125 I-labeled TNF. Approximately 1500 binding sites for 125 I-labeled TNF were determined on HT1080 cells from a saturation binding curve (Fig. 3a). 125 Ilabeled TNF bound to HT1080 cells with a K d value of 0.11 nM (Fig. 3b). Binding of 125 I-labeled TNF at a concentration of 0.1 nM was inhibited by ATROSAB in a concentration-dependent manner. Analysis of the inhibition curves obtained at 4uC and 37uC revealed a two-site inhibition mode, with IC 50 values of 0.11 nM and 9.58 nM at 4uC. Similar values were determined at 37uC (IC 50 values of 0.15 nM and 6.01 nM) (Fig. 3c). These results confirm that ATROSAB exhibits its inhibitory activity directly by blocking the binding of TNF to its receptor.

Inhibition of IL-8 and IL-6 Release Induced by TNF and Lymphotoxin Alpha (LTa)
We analyzed the neutralizing activity of ATROSAB on IL-8 and IL-6 secretion induced by lymphotoxin alpha (LTa, LTa 3 ) in comparison to TNF-induced cytokine release. With the reagents used in the HT1080 cell model, LTa-induced IL-8 reached approximately 40% of the maximum value induced by TNF (Fig. 4a). Similarly, LTa-induced release of IL-6 from HeLa cells reached approximately 65% of the maximum value induced by TNF (Fig. 4b). ATROSAB inhibited IL-8 and IL-6 secretion induced by 0.1 nM LTa (5.7 ng/ ml) with an IC 50 value of 7.6 nM, which proved to be more efficient than blocking TNF (0.1 nM) induced IL6 and IL8 secretion in these in vitro models ( Fig. 4c and d).
Furthermore, we analyzed the effects of ATROSAB on TNFmediated signal transduction as determined by phosphorylation of IkBa. HT1080 cells incubated with TNF (0.1 nM) induced rapid phosphorylation of IkBa and subsequent degradation as shown by immunoblotting experiments with anti-IkBa and anti-pIkBa antibodies (Fig. 4e). In the presence of excess amounts of ATROSAB, phosphorylation and degradation of IkBa was strongly reduced, demonstrating that ATROSAB inhibits TNFinduced signal transduction. ATROSAB alone did not induce any IkBa phosphorylation.
Comparison of ATROSAB with the Agonistic Antibody Htr-9 Monoclonal antibody Htr-9 is described as a human TNFR1 agonist [42,43]. Purified Htr-9 showed in size exclusion chromatography (SEC) a major peak corresponding to intact IgG (Fig. 5a). Htr-9 bound human TNFR1 in ELISA while no binding was seen for human TNFR2 and mouse TNFR1 and TNFR2. The same results were obtained for ATROSAB (Fig. 5b) demonstrating that both antibodies are specific for human TNFR1. By QCM, a K d value of 14 nM was determined for human TNFR1-Fc at a receptor density of 48 Hz (Fig. 5c). In ELISA, Htr-9 showed strong binding to immobilized TNFR1-Fc, although binding was weaker than that seen for ATROSAB (EC 50 values: 0.08 nM for ATROSAB and 0.7 nM for Htr-9) (Fig. 5d). ATROSAB competed with Htr-9 for binding to human TNFR1-Fc (IC 50 = 3.8 nM), indicating that the two antibodies share overlapping epitopes (Fig. 5e). Htr-9 inhibited binding of TNF to TNFR1-Fc in ELISA with an IC 50 value of 58 nM. ATROSAB, included as control, showed inhibition of TNF binding with an IC 50 value of 17 nM (Fig. 5f).
Next, we analyzed ATROSAB and Htr-9 for their potential to induce cytokine secretion in vitro. For comparison, we included human TNF, which induced a strong release of IL-6 from HeLa cells and IL-8 from HT1080 cells, respectively (Fig. 6). Maximum cytokine release was observed at a TNF concentration of approximately 10 nM, which resulted in IL-8 concentrations of 6-19 ng/ml and IL-6 concentrations of 300-700 pg/ml in the cell culture supernatant. Htr-9 showed a clear agonistic activity leading to strong IL-6 and IL-8 secretion (with a maximum of 1-5 ng/ml for IL-8 and 200-300 pg/ml for IL-6 at around 30 nM of Htr-9), corresponding to approximately 26% and 50% of the maximum TNF activity, respectively (Fig. 6). In contrast, ATROSAB led only to a marginal cytokine release reaching a maximum of 120-220 pg/ml IL-8 and of 40-80 pg/ml IL-6 (background values of cells alone: 50-70 pg/ml IL-8 and 25-40 pg/ml IL-6). This weak activity corresponds to approximately 1% and 5% of the in vitro TNF activity detected in IL-8 release (HT1080) and IL-6 release (HeLa) assays, respectively. Control IgG had no discernible effect on cytokine secretion (Fig. 6).

Epitope Mapping
In order to map the epitope of ATROSAB and Htr-9, we generated a panel of mutant human TNFR1-Fc fusion proteins carrying mutations of one or two residues substituting the residue(s) either with those from mouse TNFR1 (Fig. 7a, Table 2) or with alanine. All proteins were produced in stably or transiently transfected HEK293 cells and purified by protein A chromatography. Purity of all fusion proteins was confirmed by SDS-PAGE (data not shown). Binding of ATROSAB as well as human TNF to immobilized receptor-Fc fusion proteins was analyzed by ELISA using 1 nM ATROSAB or TNF. Data were standardized towards coating control (anti-human IgG Fc-specific antibody). Under these conditions, a strongly reduced binding of ATROSAB was seen with mutants P23S, R68A, H69Q and H69A, while all mutants showed binding of TNF (Table 2). In a previous study, we found that the double mutation P23S/Q24K in a chimeric human/mouse TNFR1 completely abrogated ATRO-SAB binding [37]. This was narrowed down to a single amino acid using the same mutations in the fully human TNFR1 background. Single mutations of P23 or Q24 to either the mouse residue or an alanine revealed that P23 is critical for ATROSAB binding, while mutation of Q24 to lysine or alanine did not affect binding. Furthermore, residue H69 (mutated to Q or A) contributes also strongly to binding of ATROSAB. Binding of ATROSAB and TNF to selected mutants was further analyzed by titration experiments ( Table 3). The wild-type receptor bound ATROSAB and TNF in the subnanomolar range. Mutation of H69 to alanine reduced binding of ATROSAB approximately 440-fold (Table 3). An 18-fold reduction was observed for the mutation of the conserved residue R68 to alanine. All mutants bound TNF with similar EC 50 values as wild-type TNFR1 (Table 3). Epitope fine mapping was also performed for the described TNFR1-agonistic antibody Htr-9. Using human-mouse chimeric TNFR1 molecules [37], the epitope of Htr-9 was located between residues 29 and 137 (B2 of CRD1, CRD2 and CRD3) (not shown). Using TNFR1 mutants, two residues (L71 and S74) were found to be part of the Htr-9 epitope, discriminating the epitope of ATROSAB from that of the agonistic antibody Htr-9 (Fig. 7b, Table 2). A structural visualization in the LT-TNFR1 complex (pdb entry 1TNR; [44]) showed that the epitopes of both antibodies are located at the ligand-binding site of the receptor ( Fig. 7b and 7c).

Discussion
ATROSAB is a humanized IgG1 antibody with a mutated heavy chain described to prevent the induction of CDC and ADCC [38]. Here, we could show that ATROSAB indeed lacks binding to C1q and FccRIA and does not induce ADCC or CDC in TNFR1-expressing cells. Of note, ATROSAB shows strongly reduced binding to FccRIIIA and the inhibitory FccRIIB. Notably, co-engagement with FccRIIB was shown to be required for in vivo activity of agonistic antibodies, such as antibodies directed against members of the TNF receptor superfamily [45]. The lack of binding of ATROSAB to FccRIIB should, therefore, prevent co-engagement of this receptor with ATROSAB avoiding undesired agonistic activities.
ATROSAB binds to human TNFR1 and is thereby capable of inhibiting binding of the TNFR1 ligands, TNF and LTa. Inhibition of binding of 125 I-labeled TNF to HT1080 cells revealed a two-site competition curve, indicating mono-and bivalent interactions of ATROSAB with membrane-expressed TNFR1. This was confirmed by QCM measurements using low and high densities of surface-immobilized TNFR1-Fc. Here, the calculated apparent K d values of 14 nM for binding at low receptor density and 0.2 nM at high receptor density point to avidity effects due to mono-and divalent binding of ATROSAB, which is in accordance with similar effects described for other IgG and IgM molecules [46,47]. The reduced binding of ATROSAB at a low receptor density was mainly caused by a faster off-rate, which was particularly apparent at physiological temperature (37uC).
ATROSAB inhibited TNF-and LTa-induced IL-6 and IL-8 release from HeLa and HT1080 cells, respectively, which could be fitted by an one-site competition curve. Compared to TNF, activity of LTa was inhibited more effectively by ATROSAB, most likely due to the lower affinity of LTa for TNFR1 [48]. Interestingly, IL-6 and IL-8 release was not affected by ATROSAB at concentrations where 125 I-TNF binding was already strongly reduced (0.01-1 nM; compare Fig. 3 and 4). This finding indicates that activation of only a few TNFR1 is sufficient to induce a cellular response. Inhibition of TNFR1 signaling complexes by ATROSAB may be affected by different populations of monomeric and preassembled multimeric TNFR1 [49] and/or the inhomogeneous distribution and motility of TNFR1 in membrane microdomains [50,51,52]. Moreover, the different affinities of TNF for monomeric and dimeric TNFR1 molecules (P.S., unpublished data) may influence binding of ATROSAB and inhibition of the formation of TNFR1 signaling complexes.Contrary to a long-standing dogma, recent data support the assumption that pre-ligand-binding assembly domain (PLAD)-associated receptor dimers represent the smallest signaling unit, rather than a trimeric receptor assembly [53]. As shown by computational simulation of TNF-TNFR1 interactions, binding of TNF to a TNFR1 homodimer may abrogate the PLAD-mediated receptor-receptor interaction leading to initial complexes where a single TNF homotrimer is bound to two TNFR1 molecules [54]. These initial complexes can then form larger clusters [55] via newly formed PLAD-PLAD interactions. Notably, in this model, binding of two PLAD-linked receptors to two distinct TNF homotrimers engaged in these clusters is sufficient for receptor activation. This is in excellent agreement with a recent modeling study [56], proposing that active TNFR1 signaling units are receptor dimers conformationally switched ''ON'' by ligand binding (see also Fig. 8), while unligated TNFR1 homodimers seem to be mainly in an ''OFF'' conformation. Key differences between the ON and the OFF status might be the distances between the intracellular death domains [56,57,58], allowing efficient TRADD association only in the ON status. Interestingly, the data by Lewis et al. suggest that TNF interaction per se does not induce a conformational change in the receptor domains. TNF rather stabilizes the PLAD-linked extracellular domains in ON conformation when bound to two receptors simultaneously, such that the increased distance of the two adjacent intracellular domains allows TRADD binding. Accordingly, we propose that only TNF molecules linked to at least two receptors, i.e. integrated in larger clusters, activate their bound receptors. In line with this model, antibodies might be capable of stabilizing either the ON or the OFF conformation, depending on the particular location or geometry of the epitope. Agonistic anti-TNFR1 antibodies, such as Htr-9, mimic binding of TNF to TNFR1 shifting the equilibrium towards the active conformation. In contrast, antagonistic antibodies, such as ATROSAB, inhibit ligand binding and keep the bound receptors mainly in the inactive conformation (Fig. 8). Interestingly, agonistic Htr-9 and antagonistic ATROSAB share an overlapping epitope as shown by competition ELISA and mutational analysis, both epitopes located within the ligandbinding site of TNFR1. These findings indicate that subtle differences in binding of a bivalent IgG molecule can be responsible for converting TNFR1 into an active signaling complex and that sole receptor dimerization by antibodies is insufficient for activation.
In summary, we provide further evidence and a rational for the unique properties of ATROSAB, a promising new antibody for

Production of TNFR1 Mutants
TNFR1 mutants were generated by site directed mutagenesis using standard PCR techniques. HEK293 cells were transfected by lipofection (Oligofectamine TM Reagent, Invitrogen, Carlsbad, USA) and proteins were purified from the supernatant by protein A affinity chromatography (TOYOPEARLH AF-rProtein A-650F, Tosoh, Stuttgart, Germany).

Affinity Measurements
Affinities of ATROSAB and Htr-9 for human TNFR1-Fc were determined by quartz crystal microbalance measurements (Attana A-100 C-Fast system). Human TNFR1-Fc was chemically immobilized on a carboxyl sensor chip according to the manufacturer's protocol at a high and low density, respectively. Binding experiments were performed in PBST (PBS, 0.1% Tween 20) pH 7.4 with a flow rate of 25 ml/min at 25uC and 37uC. The chip was regenerated with 25 ml 5 mM NaOH. Before each measurement, a baseline was measured which was subtracted from To analyze the biphasic kinetics detected at 37uC and the low receptor density (27 Hz), we subtracted a set of curves, generated by using the kinetic constants measured at 37uC and the high receptor density (130 Hz) representing bivalent binding, and aligned them to the last time point of the original dataset individually for each concentration. The residual curves showed reasonable fitting to a one site binding model. The superimposition of both fits described the original data almost exactly.

Competition of 125 I-labeled TNF
TNF was labeled with Na 125 I (Hartmann Analytic GmbH) using iodination beads (IODO-BEADS Iodination Reagent, Thermo Scientific) to specific activities of 1-2610 4 cpm/ng. The retained biologic activity (40-60%) was determined in cytotoxicity assays with Kym-1 cells [64]. The iodination beads were washed with 1 ml PBS and two of the beads were pre-incubated with 10 ml Na 125 I and 50 ml PBA for 5 min at RT. 10 mg TNF was incubated for 4-10 min with iodination beads and Na 125 I. The supernatant was applied to a PD-10 column (Sephadex TM G-25 M, GE Healthcare) and eluted with 1061 ml PBS. Eluted fractions were analyzed in a gamma counter (Berthold, Wildbad, Germany) and the 2-3 main fractions were pooled. The final concentration of 125 I-TNF was adjusted to 2 ng/ml. For competition assays, HT1080 cells were detached, resuspended in a PBA suspension (PBS, 1% BSA, 0.1% NaN 3 ) and 2610 5 cells were seeded per well. 200 nM ATROSAB was titrated in dilution steps of 1:3 to a constant concentration of 0.1 nM 125 I-TNF. Non-specific binding was determined using a 200-fold molar excess of unlabeled TNF. 0.1 nM 125 I-TNF alone served as 100% binding control. The samples were incubated for 3 hours at 4uC or 37uC. Cells and supernatant were separated by centrifugation at 13,200 rpm in micro-tubes (Sarstedt) containing 150 ml of an oil mixture (dibutylphtalate/dioctylphtalate) with a density of 1.014 g/cm 3 (adjusted to let the cells settle to the ground, separated from the aqueous phase on top of the oil). The binding of 125 I-TNF was analyzed in a gamma counter (Berthold, Wildbad, Germany). The inhibition of TNF binding by ATROSAB was analyzed using a two-site competition model.  Binding of TNF to two or more TNF receptors leads to a conformational change and activation of signal transduction [56]. b) ATROSAB binding to TNFR1 results in formation of receptor-antibody complexes, however, does not lead to a conformational change in the receptor. c) Binding of agonistic antibody Htr-9 mimics activation by TNF leading to a conformational change of the receptor. doi:10.1371/journal.pone.0072156.g008 as controls). After 16 hours, the plates were centrifuged at 1,500 g for 5 minutes. The cell supernatants were analyzed directly by ELISA using the high sensitivity human ELISA Set for IL-6 and IL-8 (ImmunoTools).

IL-6 and IL-8 ELISA
100 ml anti-human IL-8 or IL-6 antibody diluted 1:100 in PBS were coated onto microtiter plates and incubated at 4uC overnight. The residual binding sites were blocked with 3% bovine serum albumin (BSA) in PBS at room temperature for 1 hour. Subsequently, the plates were incubated with 100 ml of each, the cell supernatants, diluted 1:2 to 1:75 in RPMI 1640, 5% FCS or the IL-8 or IL-6 standard (100 to 300 pg/ml titrated by stepwise dilution of 1:3) and biotinylated anti-human IL-8 or IL-6 antibody and polystreptavidin-HRP conjugate (ImmunoTools, diluted 1:15,000 in PBS, 1% BSA, 0.05% Tween20). For detection, 100 ml TMB substrate solution (3,5,39,59-tetramethylbenzidine) were administered to each well, the reaction was stopped by the addition of 50 ml 1 M H 2 SO 4 and the absorption at the wavelength of 450 nm was measured using the Infinite microtiter plate reader (Tecan I control). Prior to each step, the plates were washed three times with PBS, containing 0.005% Tween20 and twice with PBS.

IkBa Immunoblot
One million HT1080 cells in 2 ml RPMI 1640, 5% FCS were seeded in a 6-well plate one day prior to the assay. Cells were then stimulated with TNF (0.1 nM), ATROSAB (20 nM), or TNF (0.1 nM) together with ATROSAB (500 nM), respecctively, for 0, 5, 10, 25, 30 and 60 min at 37uC, 5% CO 2 . Subsequently, the supernatants were replaced by 1 ml ice cold PBS and the cells were detached mechanically. After centrifugation (500 g, 4uC, 5 min), the pellets were resuspended in ''Solubi Shu'' lysis buffer (150 mM NaCl, 1 mM EDTA, 20 mM Tris, 1% Triton-X-100, pH 7.6) and incubated on ice for 30 minutes. Cell debris was separated by centrifugation (16.000 g, 4uC, 10 min) and the total protein content was determined by Bradford assay. 40 mg of the total protein was analyzed by SDS-PAGE and transferred to a nitrocellulose membrane using a semidry-blotter. Residual binding sites were blocked with PBS, 5% skimmed milk. Phospho-specific mouse monoclonal antibody (mAb) for pIkBa, mouse mAb for total IkBa, mouse mAb for tubulin-a (loading control), and HRPconjugated rabbit anti-mouse IgG (Fc-specific) antibody were used for detecting the respective protein species. Between each detection step, the membranes were stripped (5 min incubated with ddH 2 O, 5 min 0.2 M NaOH, 5 min dd H 2 O) and blocked again with PBS, 5% skimmed milk. Signals were detected with ECL substrate solution (incubated for 2 min).

ELISA
Microtiter plates were coated with 100 ml of human TNFR1-Fc fusion protein [37] at 1 mg/ml in PBS and incubated at 4uC overnight. The residual binding sites were blocked with MPBS (2% skim milk in PBS, 200 ml per well) at room temperature for 2 hours. 100 ml ATROSAB or Htr-9 (100 nM-5.7 pM, diluted in 2% MPBS) and MPBS alone (as coating control) were incubated at room temperature for 1 hour. For detection, 100 ml of the HRPconjugated detection antibodies in MPBS (ATROSAB: goat antihuman IgG, Fab specific [Sigma Aldrich] diluted 1:5,000; Htr-9: goat anti-murine IgG, Fc specific [Sigma Aldrich], diluted 1:1,000; coating control: goat anti-human IgG, Fc specific [Sigma Aldrich], diluted 1:5,000) were used, prior to detection with 100 ml TMB substrate solution (3,5,39,59-tetramethylbenzidine). The reaction was stopped by the addition of 50 ml 1 M H 2 SO 4 and the absorption at the wavelength of 450 nm was measured using the Infinite microtiter plate reader (Tecan I control). Between each step, the plates were washed 3 times with PBS, 0.005% Tween20 and twice with PBS.

Competition ELISA
Coating, blocking, detection and wash steps were performed as described for ELISA. The samples for the competition were 100 ml 7 nM Htr-9 mixed with ATROSAB (1 mM-5.7 pM, diluted in MPBS) and MPBS alone (as coating control), applied to the plates at room temperature for 1 hour.

Complement-dependent Cytolysis (CDC)
Kym-1 target cells (JCRB, art. # JCRB0627) bearing TNFR-1 were labeled with cell-permeant calcein (Invitrogen, Germany) for 30 minutes at 1 mM final calcein concentration in assay medium (RPMI, Thermo Fisher Scientific, supplemented with 2% FBS, Moregate). Cells were washed twice in assay medium and then seeded onto 96-well microtiter plates (1610 4 cells/well). ATRO-SAB was added at a maximal concentration of 600 mg/ml (and 1:10 dilutions thereof). Dilutions were carried out in assay medium diluted with formulation buffer of ATROSAB (25 mM histidine, 102 mM NaCl, 26 mM trehalose, 0.04% Tween 20, pH 6.2) equivalent to the highest ATROSAB concentration, thus, each well contained the same amount of ATROSAB formulation buffer, compensating for potential buffer effects on cell viability per se. Then, human serum (Blood Donation Center SRK Basel; www. blutspende-basel.ch/) was added to the cell-antibody mixture and release of calcein was measured after 2 h. Three different dilutions (1:2, 1:4, and 1:8) of each of the two serum batches were carried out in medium supplemented with 10% FBS prior to addition of serum to the cell-antibody mixture. A commercially available anti-CD20 antibody (Rituximab), which is known to induce cell lysis by CDC in CD20-expressing cells, was used at 12-fold lower concentration (50 mg/ml and 1:10 dilutions thereof) as compared to ATROSAB, and served as positive control. The target cell line for this antibody was DOHH-2 (human B-cell lymphoma cell line, DSMZ, ACC-47). As controls served cells treated with 0.5% final concentration of Triton X-100 (maximal release of calcein = 100% relative cell lysis/positive control) or the medium-buffer mixture (containing buffer equivalent to the highest antibody concentra-tion), representing the condition of spontaneous calcein release (i.e. 0% relative cell lysis/negative control). Statistical analysis was performed with Student's t-test.

Antibody-dependent Cellular Cytotoxicity (ADCC)
PBMCs were purified from fresh human whole blood (blood donation centre, Basel) as source of NK cells. Kym-1 target cells bearing the ATROSAB target receptor TNFR-1 were labeled for 30 minutes with calcein (1 mM final calcein concentration in assay medium, see CDC) and subsequently washed prior to seeding onto 96-well microtiter plates (1610 4 cells/well). ATROSAB was added at a maximal concentration of 600 mg/ml (and 1:5 dilutions thereof). Dilutions were carried out in medium diluted with reconstitution buffer of ATROSAB equivalent to the highest ATROSAB concentration (600 mg/ml). Then, the PBMC effector cell fraction was added to the cell-antibody mixture at the indicated effector-to-target (E:T) cell ratios and calcein release was measured after 4 h of incubation. A commercially available anti-CD20 antibody known to induce cell lysis by ADCC in CD20expressing cells was used at 12-fold lower concentration (50 mg/ml and 1:5 dilutions thereof) as compared to ATROSAB and served as positive control (target cell line DOHH-2). As controls, cells treated with 0.5% final concentration of Triton X-100 (maximal release of calcein = 100% relative cell lysis/positive control) or the medium-buffer mixture (containing buffer equivalent to the highest antibody concentration) was added to cells instead of antibody (spontaneous release of calcein = 0% relative cell lysis/ negative control). Statistical analysis was performed with Student's t-test.