Preclinical characterization of AZD1163, a first-in-class anti-PAD2/4 bispecific antibody for the treatment of rheumatoid arthritis Authors: Gary P. Sims, Lacie Scaletta, John Andrews, Dorothy A. Sims, Martin Strain, Anna Sigurdardottir, Teneema Kuriakose, Elizabeth England, Lichchavi Rajasinghe, Fanyi Jiang, Ximing Xiong, Frances Neal, Lisa Vinall, Philip Newton, Elin Boger, Chong Kim, Helena Dahlbäck, Susanne Prothon, Jacob Leander, Erika Darrah, Mia Collins, Garry Douglas, Nicholas White, Katie Day, Katherine A. Vousden, Catherine E. Huntington, David Close Published: 2026 ABSTRACT ABSTRACTAnti-citrullinated protein autoantibodies promote inflammation and joint tissue injury and define a poor prognostic group of patients with rheumatoid arthritis (RA). Citrullinated autoantigens that drive this autoimmune response are generated by peptidylarginine deiminase (PAD) enzymes, which are predominately expressed and released by neutrophils and monocytes. Accordingly, blocking the enzymatic activity of PADs to curb the generation of citrullinated autoantigens that drive autoimmunity and tissue injury may provide therapeutic benefit. Herein, we developed a high affinity, bispecific, anti-PAD2/4, effector-null antibody, AZD1163, which potently inhibits recombinant PAD2 and PAD4 activity in both histone H3 and fibrinogen citrullination assays. AZD1163 inhibited all endogenous PAD activity in the serum of patients with RA irrespective of the presence of anti-PAD4 autoantibodies, and neutralized PAD activity in synovial fluid. AZD1163 also bound and internalized PADs expressed on cell membranes into low pH endosomes for degradation, reducing the surface expression and catalytic potential. Binding of AZD1163 to neutrophils and monocytes did not trigger complement-dependent cytotoxicity, antibody-dependent cellular cytotoxicity, or the production of proinflammatory cytokines, or otherwise impact neutrophil phagocytosis, production of reactive oxygen species, or NETosis. In a non-human primate study of pharmacokinetics (PK) and pharmacodynamics, a single dose of AZD1163 exhibited PK consistent with a half-life extended antibody and a rapid and durable suppression of endogenous PAD activity. AZD1163 has a favorable preclinical safety profile. Collectively, these in vitro and in vivo pharmacology and safety data support the clinical development of AZD1163 as a novel therapeutic strategy for RA by reducing autoantigen load. Introduction Rheumatoid arthritis (RA) is a multi-system inflammatory condition characterized by inflammatory arthritis of peripheral synovial joints. When left untreated, patients suffer progressive deforming destructive polyarthropathy, loss of function, chronic pain and fatigue, and severe extra-articular manifestations. Despite the early introduction of conventional and advanced targeted therapies, there remains a ceiling of response to therapies that has not changed since the advent of the biologics era.1 Many patients still struggle to reach clinical remission or low disease activity, and, therefore, there remains an unmet need for new therapies and better management of patients with RA. A hallmark of RA disease is the breakdown in tolerance against a post-translational protein modification known as citrullination or deimination, which results in an autoimmune response and the generation of autoantibodies in susceptible individuals.2 Anticitrullinated protein autoantibodies (ACPA), which are present in ~70% of patients with RA, are believed to drive inflammation and trigger tissue injury, although a subset of ACPA specificities may be protective.3 A clinical test that measures antibodies against cyclic citrullinated peptides (CCP) is used to detect ACPA and is a diagnostic test for RA. ACPA can be detected prior to diagnosis, and the onset of symptoms may be associated with a spike in titers and an expansion in antigen specificities.4–6 RA patients seropositive for ACPA are associated with progression to bone erosions and tissue damage.7 RA is also strongly associated with the HLA-DRB1 alleles that encode the ‘shared epitope’ sequence motif that is believed to promote antigen presentation, a breakdown in tolerance and antigen-specific T cell responses to citrullinated antigens.8,9 Citrullinated proteins are generated by peptidylarginine deiminases (PADs), a family of five calcium-dependent enzymes that catalyze the conversion of arginine residues into citrulline in protein substrates, in a process known as deimination. PAD1 is highly expressed in male and female reproductive organs and embryonic tissue.10 PAD1 and PAD3 play a critical role in the epidermis, hair follicles, and keratinocytes. PAD2 is found in the brain, skeletal muscle, female reproductive tissue, skin, spleen, and leukocytes, and PAD4 is expressed in spleen, neutrophils, monocytes, and eosinophils.11 Within these cells and tissues, PAD activity regulates embryonic development, gene regulation, cell growth, epigenetics, tissue remodeling, and inflammation.12 PAD6 expression is important in oocytes, embryos, and testis, but is catalytically inactive.13,14 PAD2 and PAD4 are strongly implicated in the generation of citrullinated autoantigens, the autoimmune response, and the disease pathology in RA.15 Genetic evidence links genes encoding PAD2 and PAD4 (PADI2 and PADI4, respectively) with RA.16,17 Increased PAD2 and PAD4 protein, PAD activity, and citrullinated proteins are present in the inflamed joints of RA patients.18–22 Neutrophils and monocytes are the likely source of PAD2 and PAD4 at inflammatory sites. Besides being released from activated immune cells, active PAD expressed on the surface of neutrophils and monocytes may also generate citrullinated antigens.23,24 Citrullination of proteins in the RA joint may directly contribute to tissue pathology, but the formation of immune complexes with ACPA are believed to be the critical driver of joint pathology. Autoantibodies against individual citrullinated antigens do not show a dominant association with disease activity or severity, instead immune complexes with a broad repertoire of citrullinated antigen specificities appear to be important.25 Interestingly, antibodies targeting PAD4 are associated with erosive arthritis and interstitial lung disease,26,27 and anti-PAD4 autoantibodies with cross-reactivity to PAD3 may promote PAD activity and fuel the autoantigen burden.26 Conversely, anti-PAD2 antibodies (Abs) negatively correlated with joint damage and extraarticular disease manifestations and are not associated with ACPA or traditional shared epitope alleles that promote ACPA responses.28 Given the prominent role for PADs in RA pathogenesis, inhibition of enzymic activity of extracellular PAD2 and PAD4 should reduce the production of citrullinated protein autoantigens, the progressive autoimmune response, and downstream inflammatory pathways. Herein, we describe the development and characterization of AZD1163, an anti-PAD2/4 bispecific antibody designed to inhibit the generation of citrullinated protein autoantigens for the treatment of RA. We show AZD1163 potently inhibits the extracellular PAD2 and PAD4 activity in serum and synovial fluid samples from patients with RA, and pharmacokinetic (PK) and safety studies support early clinical development. Results Development of anti-PAD2/4 bis-specific antibodies The five human PAD enzymes exhibit only modest sequence identities (Table 1).Table 1.Sequence identity of human PAD enzymes.% IdentityPAD1PAD2PAD3PAD4PAD6PAD110052.8158.0557.9945.36PAD2 10051.8150.3044.56PAD3 10056.2344.48PAD4 10046.44PAD6 100The sequences for alignment were the reference sequences from Uniprot with the respective accession numbers PAD1 (Q9ULC6); PAD2 (Q9Y2J8); PAD3 (Q9ULW8); PAD4 (Q9UM07); PAD6 (Q6TGC4). Table 1.Sequence identity of human PAD enzymes.% IdentityPAD1PAD2PAD3PAD4PAD6PAD110052.8158.0557.9945.36PAD2 10051.8150.3044.56PAD3 10056.2344.48PAD4 10046.44PAD6 100The sequences for alignment were the reference sequences from Uniprot with the respective accession numbers PAD1 (Q9ULC6); PAD2 (Q9Y2J8); PAD3 (Q9ULW8); PAD4 (Q9UM07); PAD6 (Q6TGC4). Sequence identity of human PAD enzymes. The sequences for alignment were the reference sequences from Uniprot with the respective accession numbers PAD1 (Q9ULC6); PAD2 (Q9Y2J8); PAD3 (Q9ULW8); PAD4 (Q9UM07); PAD6 (Q6TGC4). Since it was unlikely cross-reactive antibodies inhibitory for both PAD2 and PAD4 could be generated, two independent screening campaigns were initiated to identify anti-PAD2 and anti-PAD4 antibodies using a large single-chain variable fragment (scFv) human antibody phage display. PAD2- and PAD4-specific scFv were isolated from the phage display library in a series of repeated selection cycles on recombinant human PAD2 or PAD4 with hybrid strands, including selection cycles on recombinant cynomolgus (cyno) or mouse antigen. From the selection outputs, scFv were expressed in the bacterial periplasm and screened for their capacity to inhibit PAD2 or PAD4 activity in a fluorescence resonance energy transfer (FRET)-based trypsin cleavage assay using a short arginine-containing trypsin-cleavage peptide flanked by AlexaFluor 488 (FRET donor) and QSY7 quencher (FRET acceptor). In this assay, trypsin cleavage separates the dye and quencher, resulting in a fluorescent signal, but PAD activity deiminates (citrullinates) the arginine residue preventing trypsin cleavage, resulting in quenching of the AlexaFluor 488 fluorescence. Inhibitory scFvs blocked PAD-mediated citrullination, rendering the peptide susceptible to trypsin cleavage and allowing detectable emission from the donor. Inhibitory anti-PAD2 and anti-PAD4 scFvs that bound their respective human and cyno PADs were sequenced to identify unique clones. Unique scFv clones were initially re-expressed in bacteria and affinity purified and retested, and a subset of selection outputs were subcloned for high throughput expression in IgG format. The relative expression, binding, and inhibitory characteristics were used to identify lead anti-PAD2 and anti-PAD4 Abs for optimization. Targeted mutations were made to revert framework residues to germline, ahead of affinity maturation of the scFv sequence by block mutagenesis of the variable heavy and light chains of the complementarity-determining regions. Further mutations were made to address sequence liabilities to generate the parental anti-PAD2 and anti-PAD4 IgG1 antibodies. The parental antibodies only bound to their respective targets. Neither antibody bound to PAD3 (Figure 1(A)). The optimization process improved the affinity of the anti-PAD2 antigen-binding fragment (Fab) over > 1,000-fold for human PAD2, and >100-fold for the anti-PAD4 Fab binding to human PAD4 (Figure 1(B)). To evaluate the inhibition of PAD enzymatic activity in the presence of the antibodies, an enzyme linked immunosorbent assay (ELISA)-based histone H3 PAD activity assay was used. Optimization of both lead Fab antibodies resulted in potent suppression of PAD2 and PAD4 activity, respectively (Figure 1(C, D)). Figure 1.Development of anti-PAD2/4 bispecific antibodies.A) A bar graph labeled A with y-axis label RFU and range 0 to 100000 and x-axis categories PAD2, PAD3, PAD4. Legend: Control Ab, alphaPAD2 Ab, alphaPAD4 Ab. PAD2 bars: Control Ab about 2000 RFU, alphaPAD2 Ab about 78000 RFU, alphaPAD4 Ab about 5000 RFU. PAD3 bars: Control Ab about 2000 RFU, alphaPAD2 Ab about 6000 RFU, alphaPAD4 Ab about 5000 RFU. PAD4 bars: Control Ab about 2000 RFU, alphaPAD2 Ab about 8000 RFU, alphaPAD4 Ab about 78000 RFU. B) A bar graph labeled B with y-axis label Affinity K subscript D (nM) and range 0.01 to 1000 and x-axis categories PAD2 and PAD4. Legend: Lead alphaPAD2 Fab, Optimized alphaPAD2 Fab, Lead alphaPAD4 Fab, Optimized alphaPAD4 Fab. PAD2: Lead alphaPAD2 Fab about 200 nM, Optimized alphaPAD2 Fab about 0.2 nM. PAD4: Lead alphaPAD4 Fab about 15 nM, Optimized alphaPAD4 Fab about 0.08 nM. C) A line graph labeled C titled PAD2 with x-axis label Log [Fab] (M) range negative 12 to negative 5 and y-axis label percent PAD activity range 0 to 150. Curves: Control Fab stays about 95 to 120 percent across the range; Lead Fab stays near 100 percent until about negative 8 then drops to about 0 by about negative 6.5; Optimized Fab drops from about 100 percent near negative 11 to about 0 by about negative 8. D) A line graph labeled D titled PAD4 with x-axis label Log [Fab] (M) range negative 12 to negative 5 and y-axis label percent PAD activity range 0 to 150. Curves: Control Fab stays about 100 to 125 percent until about negative 6.5 then drops to about 50 near negative 6; Lead Fab stays near 105 to 120 percent until about negative 7.5 then drops to about 0 by about negative 6.5; Optimized Fab drops from about 110 percent near negative 11 to about 0 by about negative 9. E) A line graph labeled E titled PAD2 with x-axis label Log [Ab] (M) range negative 12 to negative 6 and y-axis label percent PAD activity range 0 to 150. Multiple curves: Bis3 Con Ab stays about 105 to 125 percent; Duet Con Ab decreases from about 110 percent near negative 11 to about 0 by about negative 8; alphaPAD2 Ab decreases from about 110 percent near negative 11 to about 0 by about negative 8.5; alphaPAD4 Ab stays about 100 to 120 percent; alphaPAD2 to PAD4 Bis3-TM decreases from about 110 percent near negative 11 to about 0 by about negative 8.5; alphaPAD4 to PAD2 Bis3-TM decreases from about 110 percent near negative 11 to about 0 by about negative 8.5; alphaPAD2 to PAD4 Duet-TM decreases from about 110 percent near negative 11 to about 0 by about negative 7.5; alphaPAD4 to PAD2 Duet-TM decreases from about 110 percent near negative 11 to about 0 by about negative 7.5. F) A line graph labeled F titled PAD4 with x-axis label Log [Ab] (M) range negative 13 to negative 7 and y-axis label percent PAD activity range 0 to 150. Bis3 Con Ab stays about 110 to 125 percent; Duet Con Ab stays about 110 to 125 percent; alphaPAD2 Ab stays about 110 to 125 percent; alphaPAD4 Ab decreases from about 110 percent near negative 12 to about 0 by about negative 10.5; alphaPAD2 to PAD4 Bis3-TM decreases from about 110 percent near negative 12 to about 0 by about negative 10.5; alphaPAD4 to PAD2 Bis3-TM decreases from about 110 percent near negative 12 to about 0 by about negative 10.5; alphaPAD2 to PAD4 Duet-TM decreases from about 110 percent near negative 12 to about 0 by about negative 9; alphaPAD4 to PAD2 Duet-TM decreases from about 110 percent near negative 12 to about 0 by about negative 9. G) A line graph labeled G titled PAD2 with x-axis label Log [Ab] (M) range negative 12 to negative 7 and y-axis label percent PAD activity range 0 to 150. Bis3 Con Ab stays about 105 to 125 percent; alphaPAD2 to PAD4 Bis3-TM decreases from about 100 percent near negative 11.5 to about 0 by about negative 8; alphaPAD2 to PAD4 Bis3-TM-YTE follows a similar decrease to about 0 by about negative 8; alphaPAD2 to PAD4 Bis3-FQQ-YTE follows a similar decrease to about 0 by about negative 8. H) A line graph labeled H titled PAD4 with x-axis label Log [Ab] (M) range negative 13 to negative 8 and y-axis label percent PAD activity range 0 to 150. Bis3 Con Ab stays about 110 to 125 percent; alphaPAD2 to PAD4 Bis3-TM decreases from about 110 percent near negative 12 to about 0 by about negative 9.5; alphaPAD2 to PAD4 Bis3-TM-YTE decreases similarly to about 0 by about negative 9.5; alphaPAD2 to PAD4 Bis3-FQQ-YTE decreases similarly to about 0 by about negative 9.5.Multiple bar and line graphs showing antibody binding, affinity and PAD activity inhibition for PAD2 to PAD4.(A) Binding specificity of anti-PAD2 and anti-PAD4 lead antibodies. (B) Affinity of lead and optimized anti-PAD2 and anti-PAD4 Fabs. (C) Inhibition of recombinant PAD2 activity with lead and optimized anti-PAD2 Fabs in Histone H3 PAD citrullination assay. (D) Inhibition of recombinant PAD4 activity with lead and optimized anti-PAD4 Fabs in Histone H3 citrullination assay. Inhibition of PAD2 (E) and PAD4 activity (F) with optimized monoclonal and bispecific antibodies in bis-3 and duet formats and both orientations. Orientation of bis-3 formats are αPAD2(Fab)-PAD4(scFv) and αPAD4(Fab)-PAD2(scFv), and for the duet mabs αPAD2K(κ-light chain)-PAD4L(λ-light chain) and αPAD4K(κ-light chain)-PAD2L(λ-light chain). Inhibition of PAD2 (G) and PAD4 activity (H) of bispecific antibodies with TM, TM-YTE, and FQQ-YTE Fc mutations. Error bars represent standard deviation. Development of anti-PAD2/4 bispecific antibodies. The optimized lead sequences were incorporated and tested in two bispecific formats: a monovalent duet key-in-hole format (one Fab on each arm of the IgG) and bivalent Bis3 format (full-length IgG with C-terminal scFv extending from the Fc). Both bispecific formats were evaluated in two orientations for inhibition of recombinant PAD2 and PAD4 activities. For the Bis3 antibodies, the arm in the Fab orientation had similar potency to the optimized parental antibody, whereas a minor loss of activity was evident for the scFv arm (Figure 1). No difference in PAD2 or PAD4 inhibition was evident between the two duet antibody orientations, but owing to the monovalency of the format, they were less potent than the bispecific antibodies (Figure 1(E, F)). To mitigate against potential effector functions driven through FcγR and complement engagement, and to enhance FcRn recycling to promote half-life, mutations may be introduced into the Fc region of the antibody.29,30 TM mutations (L234F/L235E/P331S) or FQQ mutations (L234F/L235Q/K322Q) reduce binding for FcγR and C1q, and YTE mutations (M252Y/S254T/T256E) promote FcRn recycling. Variants of the bispecific antibody containing the TM mutations or combinations of TM-YTE or FQQ–YTE mutations were evaluated in the histone H3 citrullination assay. All variants similarly exhibited potent inhibition of PAD2 (Figure 1(G)) and PAD4 activity (Figure 1(H)). Characterization of AZD1163 AZD1163 is a bivalent bispecific antibody based on a human immunoglobulin G1 antibody with Fc domain mutations to reduce effector functions (FQQ) and extend half-life (YTE). The top Fab targets PAD2, while the bottom scFvs target PAD4 (Figure 2(A)). AZD1163 in this format was chosen for further development. As expected, optimization and configuration in the final bispecific format did not alter the binding specificity, AZD1163 bound to human PAD2 and PAD4, but not PAD1 or PAD3 (Figure 2(B)). Since PAD6 is not catalytically active, the binding of AZD1163 to PAD6 was not assessed. AZD1163 bound to cynomolgus monkey and murine PAD2 with only a minor loss of affinity, whereas for PAD4 there was approximately a 10-fold loss of binding to cynomolgus PAD4 (Figure 2(C)). The lead anti-PAD4 and its optimized variants never showed cross-reactivity to murine PAD4. AZD1163 potently inhibited human, cynomolgus monkey, and murine PAD2 activity (Figure 2(D)), and human, cynomolgus monkey, but not murine PAD4 activity in the Histone H3 citrullination assay, as expected (Figure 2(E)). AZD1163 also inhibited citrullination by recombinant human PAD2 and PAD4 in a fibrinogen PAD activity assay (Supplemental Material 1). Figure 2.Characterization of AZD1163.The image A showing a schematic labeled PAD2 Fab and PAD4 scFv. The image B showing a bar graph with x-axis labels PAD1, PAD2, PAD3, PAD4 and y-axis label RFU with range 0 to 250000. Legend: Bis3 Con Ab and AZD1163. Bars: PAD1 near 0 for both; PAD2 AZD1163 about 200000 and Bis3 Con Ab near 0; PAD3 near 0 for both; PAD4 AZD1163 about 200000 and Bis3 Con Ab near 0. The image C showing a bar graph with x-axis labels PAD2 and PAD4 and y-axis label Fold-change loss of binding (relative to human) with ticks 0.1, 1, 10, 100. Legend: Human, Cyno, Murine. PAD2: Human about 1, Cyno about 2, Murine about 2. PAD4: Human about 1, Cyno about 10. The image D showing a line graph titled PAD2 with x-axis label Log [Ab] (M) from negative 12 to negative 6 and y-axis label percent PAD Activity from 0 to 140. Curves listed: HuPAD2 plus Con Ab, CyPAD2 plus Con Ab, MuPAD2 plus Con Ab stay near 90 to 130 across the range. HuPAD2 plus AZD1163 drops from about 90 at negative 11 to near 0 by about negative 9.5. CyPAD2 plus AZD1163 drops from about 100 at negative 10.5 to near 0 by about negative 8.8. MuPAD2 plus AZD1163 drops from about 100 at negative 10.2 to near 0 by about negative 8.5. The image E showing a line graph titled PAD4 with x-axis label Log [Ab] (M) from negative 13 to negative 7 and y-axis label percent PAD Activity from 0 to 140. Con Ab curves for HuPAD4, CyPAD4, MuPAD4 stay near 90 to 135. HuPAD4 plus AZD1163 drops from about 95 at negative 11 to near 0 by about negative 9.5. CyPAD4 plus AZD1163 drops from about 95 at negative 10.5 to near 0 by about negative 9.0. MuPAD4 plus AZD1163 stays near 90 to 110. The image F showing a line graph with x-axis label Time (hrs) from 0 to 20 and y-axis label Integrated Intensity (OCU times micrometer superscript 2 per well) from 0 to 1.2 times 10 superscript 7. Legend: No Ab, Control Ab, AZD1163. AZD1163 rises from near 0 to about 1.0 times 10 superscript 7 by 20. Control Ab rises to about 3 times 10 superscript 6. No Ab stays near 0. The image G showing two bar graphs. Left: y-axis label Antibody Internalization (percent of Total Cells) from 0 to 80; x-axis Control Ab and AZD1163; bars about 25 and about 60 with four asterisks above. Right: y-axis label Integrated Intensity (OCU times micrometer superscript 2 per well) from 0 to 1.5 times 10 superscript 7; x-axis Control Ab and AZD1163; bars about 1 times 10 superscript 6 and about 6 times 10 superscript 6 with two asterisks above. The image H showing a line graph with x-axis label Time (h) from 0 to 24 and y-axis label Integrated Intensity (OCU times micrometer superscript 2 per well) from 0 to 8 times 10 superscript 6. Legend: Control Ab, AZD1163, Control Ab dash pulse slash chase, AZD1163 dash pulse slash chase. AZD1163 rises from near 0 to about 6 times 10 superscript 6 by 24. Control Ab rises to about 1 times 10 superscript 6. Pulse slash chase curves track similarly, with a marked point near time 6.A mixed figure showing AZD1163 binding, inhibition results across PAD targets and species, and internalization characteristics.(A) Depiction of AZD1163 bispecific antibody. (B) Binding of 10 µg/mL AZD1163 and control bispecific antibody to recombinant human PAD1, 2, 3, and 4 was measured by ELISA and reported as relative fluorescence units (RFU). (C) Relative binding affinity of AZD1163 to cynomolgus monkey and murine PAD2 and PAD4 compared to human PAD2 and PAD4 is reported as fold-change loss. Binding to murine PAD4 was not detected. (D) Inhibition of recombinant human (Hu) PAD2 (0.05 ng/mL), cynomolgus monkey (Cy) PAD2 (0.15 ng/mL) and murine (Mu) PAD2 (0.1 ng/mL) with AZD1163 and control bispecific antibody (Con Ab) using Histone H3 PAD citrullination assay. (E) Inhibition of recombinant human PAD4 (0.3 ng/mL), cynomolgus monkey PAD4 (0.8 ng/mL), and murine PAD4 (5 ng/mL) activity with AZD1163 and control bispecific antibody using Histone H3 citrullination assay. (F) Internalization kinetic analysis of FabFluor-pH Orange labeled AZD1163 and control bispecific antibody in neutrophils was assessed using IncuCyte live cell imaging. Representative data from one of eight donors. (G) The mean percentage of neutrophils with internalized antibody and the mean integrated intensity were calculated after 22–24 hours for eight donors. (H) Internalization showing kinetics of internalization of FabFluor-pH Orange labeled AZD1163 and control bispecific antibody (filled circles), and pulse-chase internalization kinetics (unfilled circles) with 6-hour unlabeled antibody pulse, wash out, and 18-hour chase with FabFluor-pH Orange labeled antibodies (red arrow indicates the start of chase). Representative of one of four donors. Error bars represent standard deviation. Characterization of AZD1163. Since PAD2 and PAD4 are detected on the surface of some neutrophils and monocytes (Supplemental Material 2), we evaluated the consequences of AZD1163 binding to surface PADs on human neutrophils. After incubation with Fabfluor-pH Orange-labeled AZD1163, there was an initial lag of approximately 2 hours before the signal intensity accumulated over the next 20 hours, which is consistent with the internalization of the antibody-PAD complex trafficking to low pH endosomes for degradation (Figure 2(F)). After 21–24 hours, internalization was detected on approximately 60% of neutrophils, which is consistent with the proportion of neutrophils expressing PAD2 or PAD4, and the intensity of the signal was significantly greater compared to a control bispecific antibody, which did not target PAD2 or PAD4 (Figure 2(G)). When the neutrophils were preincubated with unlabeled AZD1163 for 6 hours (pulse), washed out and incubated with Fabfluor-pH Orange-labeled AZD1163 (chase), there was a small additional accumulation of signal, which indicates that AZD1163 continued to internalize PADs that may be either recycled or de novo generated over the remaining 18-hour monitoring period (Figure 2(H)). In summary, these data demonstrate that AZD1163 specifically binds and potently inhibits the enzymatic activity of human PAD2 and PAD4. The binding and inhibition profile of AZD1163 for cynomolgus monkey PAD2 and PAD4 support the use of this species for pharmacology and toxicology studies. AZD1163 also internalizes membrane associated PADs, which in combination with the effects on enzymatic activity, may contribute to the overall inhibitory function. Increased PAD activity in serum from patients with RA is inhibited by AZD1163 PAD2 and PAD4 may citrullinate arginine resides in substrates within overlapping consensus target sequences, but there is limited information on their substrate specificities, particularly in the context of RA autoantigens.31,32 Consequently, we initially compared the ability of PAD2 and PAD4 to generate citrullinated autoantigens, using six known RA autoantigens and sera from patients with RA. ACPA+ RA serum could specifically detect citrullinated variants of the six citrullinated autoantigens, and both PAD2 and PAD4 similarly generated these autoantigens in vitro (Supplemental Material 3). To verify that PAD2 and PAD4 protein levels are elevated in RA, we measured concentrations of both proteins in the serum of 87 patients and 30 healthy donors. The concentrations of PAD4 were higher than PAD2, and both PAD2 and PAD4 were significantly increased in patients with RA compared with healthy donors (Figure 3(A)). To determine if the increased circulating PAD enzymes corresponded to endogenous PAD activity, we evaluated the PAD activity in the serum using an ELISA-based histone H3 citrullination assay. Consistent with the increase in PAD concentrations, PAD activity was also increased in the serum of RA patients compared to healthy controls (Figure 3(B)). The PAD activity correlated with the PAD4, but not PAD2, concentrations in the RA serum samples (Figure 3(C)). Figure 3.Increased PAD activity in serum of RA patients is inhibited by AZD1163.The image contains eight graphs. Graph B shows PAD activity with optical density at 450 nanometers on the y axis from 0 to 2.5 for healthy donors and RA patients. Graph C shows PAD in nanograms per milliliter on the y axis from 0 to 300 and PAD activity with optical density at 450 nanometers on the x axis from 0 to 2.5 for PAD4 and PAD2. Graph D shows the number of healthy donors and RA patients on the y axis, with the number of healthy donors ranging from 0 to 30 and the number of RA patients ranging from 0 to 80, with anti-PAD4 and anti-PAD3. Graph E shows PAD2 in nanograms per milliliter on the y axis from 0 to 100, PAD4 in nanograms per milliliter on the y axis from 0 to 300 and PAD activity with optical density at 450 nanometers on the y axis from 0 to 2.5 for healthy donors, alpha PAD4 dash RA and alpha PAD4 plus RA. Graph F shows PAD activity with optical density at 450 nanometers on the y axis from 0 to 2.0 and log of AZD1163 in molar on the x axis from negative 13 to negative 7. Graph G shows percent inhibition of PAD activity on the y axis from 0 to 100 and log of monoclonal antibody in molar on the x axis from negative 13 to negative 7 for AZD1163 and control bis. Graph H shows PAD activity with optical density at 450 nanometers on the y axis from 0 to 2.0 and log of AZD1163 in molar on the x axis from negative 11 to negative 7 for RA 1, RA 2, RA 3 and EDTA control.The image contains multiple graphs showing PAD2 and PAD4 protein levels and PAD activity in serum of RA patients, and inhibition of PAD activity by AZD1163.(A) PAD2 and PAD4 concentrations in serum samples from healthy donors (HD, n = 24) and RA patients (n = 86) was measured by ELISA. (B) PAD activity in serum samples from HD (n = 30) and patients with RA (n = 87) was measured using the Histone H3 citrullination assay, and correlations between PAD activity and PAD4 (gray circles) and PAD2 (back squares). Bars indicate median values. **** denotes p < 0.001 with two-tailed non-parametric Mann Whitney test. (C) Simple linear regression between serum PAD2 and PAD4 concentrations and PAD activity. Dotted line depicts 95% confidence intervals. Pearsons’s correlation coefficient squared (r2) and P values are indicated. (D) Frequency of anti-PAD4 and anti-PAD3 autoantibodies detected in HD and patients with RA. (E) Serum PAD2 and PAD4 concentrations and PAD activity in HD, and patients with RA seropositive and seronegative for anti-PAD4 autoantibodies. Bars indicate median values. *** denotes p < 0.001 and **** denotes p < 0.0001 with non-parametric one-way ANOVA Kruskal–Wallis test. (F) Individual and grouped dose responses for AZD1163 on PAD activity in anti-PAD4 seronegative RA patients. (G) Individual and grouped dose responses for AZD1163 on PAD activity in anti-PAD4 seropositive RA patients. Error bars indicated SEM. (H) Inhibition of endogenous PAD activity with AZD1163 in whole blood from three patients with RA. Error bars indicate SD. Increased PAD activity in serum of RA patients is inhibited by AZD1163. It has been previously proposed that the activity of PAD4 may be influenced by the presence of autoantibodies targeting PAD4 or a cross-reactive PAD3/PAD4 epitope.26 Consequently, we evaluated healthy control and RA serum samples for reactivity to PAD4 and PAD3. All healthy donors (n = 30) were seronegative for both anti-PAD4 and anti-PAD3 antibodies. For the RA cohort (n = 87), 71 were seronegative, 15 were anti-PAD4 seropositive (7 were PAD4 single-positive, and 8 were PAD4/PAD3 double positive) (Figure 3(D)). To ascertain if the presence of anti-PAD4 autoantibodies may influence PAD concentrations and PAD activity, the RA serum samples were grouped into PAD4 seropositive (n = 15), PAD4 seronegative (n = 72) and compared alongside the healthy donors. As expected, the PAD2 concentrations were similar in the PAD4 seropositive and seronegative groups, whereas PAD4 and PAD activity were significantly lower in the presence of anti-PAD4 autoantibodies (Figure 3(E)). These data indicate that the anti-PAD4 antibodies may mask or reduce PAD4 levels and subsequent PAD4 activity. We next determined if the presence or absence of endogenous autoantibodies could alter the potency of AZD1163. Initially, we examined the impact of AZD1163 on 20 PAD4 seronegative RA serum that exhibited the highest PAD activity. AZD1163 concentration-dependently inhibited the PAD activity in all PAD4 seronegative RA serum samples, whereas the control bispecific antibody had no effect (Figure 3(F)). Similarly, the capacity of AZD1163 to inhibit endogenous PAD activity was also evaluated in 15 anti-PAD4 seropositive serum from RA patients, including the 8 RA samples seropositive for both PAD3 and PAD4 autoantibodies (Figure 3(G)). One of the anti-PAD3/4 double positive samples did exhibit high PAD activity, and this was completely inhibited by AZD1163 with comparable potency to the seronegative RA samples. Despite the presence of anti-PAD4 antibodies and the much lower baseline PAD activity exhibited in the other anti-PAD4 positive samples, AZD1163 concentration-dependently inhibited the PAD activity to background levels. Next, we verified that AZD1163 could also inhibit PAD activity in whole blood. AZD1163 was incubated with a 1-in-5 dilution of RA blood for 15–20 mins, plasma was harvested, and the PAD activity was measured using the histone H3 citrullination assay. As expected, AZD1163 concentration-dependently inhibited the PAD activity in the three RA blood samples tested (Figure 3(H)). Collectively, these data indicate that PAD concentrations and activity are increased in patients with RA. The presence of anti-PAD4 antibodies may reduce the overall systemic concentrations of PAD4 and/or PAD activity. Importantly, irrespective of the anti-PAD4 status, AZD1163 potently inhibits all PAD activity in the serum and blood samples of patients with RA. AZD1163 inhibits PAD activity in synovial fluid from RA patients Next, we sought to determine if AZD1163 could inhibit the endogenous PAD activity in RA synovial fluid. We initially profiled the PAD concentrations in 12 synovial fluid samples from patients with RA (Table 2). The concentrations and proportions of PAD2 and PAD4 in the synovial fluid samples were highly variable.Table 2.Concentrations of PAD2 and PAD4 in synovial fluid from RA patients.Synovial fluid IDPAD2 (ng/mL)PAD4 (ng/mL)RF statusRA14362600–RA238101+RA3892800–RA41563+RA536+RA64773140+RA7677–RA819432294+RA921298UnknownRA101231719+RA11463+RA1241208–PAD2 and PAD4 levels of RA synovial fluid were determined by ELISA (Cayman Chemicals). Rheumatoid Factor Status was provided from the clinical profile. ELISA = enzyme linked immunosorbent assay; ID = identifier; PAD = peptidyl arginine deiminase; RF = rheumatoid factor; SF = synovial fluid. Table 2.Concentrations of PAD2 and PAD4 in synovial fluid from RA patients.Synovial fluid IDPAD2 (ng/mL)PAD4 (ng/mL)RF statusRA14362600–RA238101+RA3892800–RA41563+RA536+RA64773140+RA7677–RA819432294+RA921298UnknownRA101231719+RA11463+RA1241208–PAD2 and PAD4 levels of RA synovial fluid were determined by ELISA (Cayman Chemicals). Rheumatoid Factor Status was provided from the clinical profile. ELISA = enzyme linked immunosorbent assay; ID = identifier; PAD = peptidyl arginine deiminase; RF = rheumatoid factor; SF = synovial fluid. Concentrations of PAD2 and PAD4 in synovial fluid from RA patients. PAD2 and PAD4 levels of RA synovial fluid were determined by ELISA (Cayman Chemicals). Rheumatoid Factor Status was provided from the clinical profile. ELISA = enzyme linked immunosorbent assay; ID = identifier; PAD = peptidyl arginine deiminase; RF = rheumatoid factor; SF = synovial fluid. It was evident that synovial fluid concentrations of both PAD2 and PAD4 were significantly higher than that of matched serum samples (Figure 4(A)), but there was no correlation between the PAD concentrations in the synovial fluid and the serum for either PAD2 (Figure 4(B)) or PAD4 (Figure 4(C)). PAD activity in the synovial fluid samples was evaluated with the Histone H3 citrullination assay. AZD1163 inhibited PAD activity in all synovial fluid samples with more AZD1163 required to inhibit activity in RA1 and RA3 compared to RA2 and RA4, consistent with the higher PAD concentrations in these samples (Figure 4(D)). Figure 4.AZD1163 inhibits PAD activity in synovial fluid from patients with RA.The image A showing a scatter plot of PAD with y-axis label PAD (nanogram per milliliter) on a logarithmic scale with ticks 0.1, 1, 10, 100, 1000, 10000 and x-axis categories PAD2 Serum, PAD2 SF, PAD4 Serum, PAD4 SF, with significance marks asterisk asterisk asterisk above PAD2 Serum versus PAD2 SF and above PAD4 Serum versus PAD4 SF. The image B showing a scatter plot with x-axis label Log [Serum PAD2] (nanogram per milliliter) from 0 to 3 and y-axis label Log [SF PAD2] (nanogram per milliliter) from 0 to 4, with a fitted line and text p equals 0.1009 and r superscript 2 equals 0.3005. The image C showing a scatter plot with x-axis label Log [Serum PAD4] (nanogram per milliliter) from 0 to 3 and y-axis label Log [SF PAD4] (nanogram per milliliter) from 0 to 4, with a fitted line and text p equals 0.4024 and r superscript 2 equals 0.07901. The image D showing a dose response line graph with x-axis label Log [AZD1163] (molar) from negative 13 to negative 7 and y-axis label percent PAD activity from 0 to 100, with four curves labeled RA1, RA2, RA3, RA4. The image E showing four dose response line graphs labeled RA8, RA9, RA10, RA12, each with x-axis label Log [Ab] (molar) from negative 13 to negative 7 and y-axis label PAD activity from 0 to 2.5 and a legend listing AZD1163, Anti-PAD4, Control Bis, Control Ab, EDTA CNTL.A mixed set of scatterplots and line graphs showing PAD levels in synovial fluid samples, correlations with serum levels and inhibition curves.(A) PAD2 and PAD4 concentrations in matched serum and synovial fluid samples from 12 patients with RA. Bar indicates median. Values below the lower limit of detection are not shown. *** defines p < 0.001 with non-parametric paired Wilcoxon test. Correlations between serum and synovial fluid samples for PAD2 (B) and PAD4 (C). Pearson’s correlations were performed on log transformed PAD concentrations. Pearsons’s correlation coefficient squared (r2) and P values are indicated. (D) Inhibition of PAD activity with AZD1163 in 1/10th synovial fluid from four RA donors using histone H3 citrullination assay. (E) Comparison of anti-PAD4 Ab and AZD1163 inhibition of PAD activity in synovial fluid samples. RA8 was evaluated at a 1/1000 dilution, and RA9, 10, and 12 were evaluated at a 1/200 dilution. AZD1163 inhibits PAD activity in synovial fluid from patients with RA. To verify the importance of targeting both PAD2 and PAD4 with AZD1163, inhibition of synovial fluid PAD activity with AZD1163 was compared with anti-PAD4 antibody alone. Consistent with the higher concentrations of PAD2 in the synovial fluid, the anti-PAD4 antibody only partially inhibited the PAD activity in RA8, RA10, and RA12, whereas AZD1163 fully inhibited the PAD activity to background levels in all synovial fluid samples tested (Figure 4(E)). Overall, these data collectively demonstrate that AZD1163 blocks all PAD activity in serum, blood, and synovial fluid of patients with RA, highlighting the importance of inhibiting the activity of both PAD2 and PAD4. AZD1163 does not trigger Fc-mediated effector functions or impact neutrophil functions Both PAD2 and PAD4 can be detected on the surface of human neutrophils and monocytes 23,24 (Supplemental Material 2). To mitigate against the bispecific antibody driving FcγR and complement-mediated effector functions upon binding to cells, AZD1163 carries the FQQ Fc mutations. To verify that AZD1163 does not trigger any inflammatory responses or impact key neutrophil functions, a series of secondary pharmacology assays were performed. In complement-dependent cytotoxicity (CDC) assays, AZD1163 did not induce any complement-mediated changes in cell viability compared to the control bispecific antibody, whereas complement-killing was evident with rituximab on Raji B cells in the presence of normal human serum (Figure 5(A)). In antibody-dependent cellular cytotoxicity (ADCC) assays using the KC1333 natural killer cell line, cell viability for neutrophils and monocytes treated with AZD1163 was comparable with bispecific control antibody and untreated cells, whereas rituximab decreased viability of Raji cells (Figure 5(B)). AZD1163 did not induce the production of proinflammatory cytokines (TNF-α, IFN-γ, or IL-6) when added to whole blood from healthy donors or patients with RA (Figure 5(C)) or when immobilized on a plate and incubated with PBMCs (data not shown). Treatment of TNF-activated neutrophils with AZD1163 or the control antibody did not impair phagocytosis of pHrodo green-labeled Staphylococcus aureus bioparticles (Figure 5(D)), or impact reactive oxygen species (ROS)-induced oxidation of 2,’7'dichlorofluorescein diacetate (DCFDA) into 2,’7’-dichlorofluorescein (DCF) following phorbol 12-myristate 13-acetate (PMA) stimulation (Figure 5(E)). AZD1163 neither induced NETosis (Figure 5(F)) nor affected PMA-induced NETosis (Figure 5(G)). Collectively, these in vitro data demonstrate that AZD1163 treatment does not trigger antibody- or complement-mediated killing, induce cytokines or impair effector functions in primary human neutrophils. Figure 5.AZD1163 does not trigger Fc-mediated effector functions or impact neutrophil functions.Text and plotted values are not fully legible at the provided resolution, preventing exact extraction of all axis labels, units and data values for each sub-image.Image with multiple bar charts and graphs showing effects of antibodies on ADCC, CDC, cytokine induction, and neutrophil ROS production, phagocytosis and NETosis.(A) CDC was assessed for human neutrophils, monocytes, and Raji cells 6 hours after incubation with AZD1163 or R347 Bis3-FQQ-YTE control antibody (R347) in 10% normal human sera or heat inactivated (HI) sera. Cell viability (%) was normalized to untreated cells with HI serum. Graphs show mean ± SD (n = 3). *** denotes p < 0.001 (unpaired t-test with Welch correction). (B) ADCC was assessed after a 6-hour incubation with AZD1163 or R347 Bis3-FQQ-YTE control antibody of human neutrophils, monocytes, and Raji cells seeded with KC1333 cells at 1:1 ratio. Cell viability (%) was normalized to untreated cells. Graphs show mean ± SD n = 3). **** = denotes p < 0.0001 with one-way ANOVA with Dunnett’s multiple comparison test. (C) Cytokine release was assessed in blood from healthy donors (red circles) or RA patients (black squares) incubated for 24 hours at 37°C with 10 μg/mL of antibodies as indicated. TNF-α, IFN-γ, and IL-6 were measured by Luminex. Alemtuzumab (anti-CD52 Ab) was used as a positive control. * denotes p < 0.05 and ** represents p < 0.01 using one-sided Wilcoxon matched pairs test. (D) Phagocytosis of S. aureus bioparticles (% pHrodo green positive) by neutrophils was assessed after 4 hours incubation with AZD1163 or R347 Bis3-FQQ-YTE control antibody or treatment with phagocytosis inhibitor Cytochalasin D. Graph shows mean ± SD n = 3). **** = denotes p < 0.0001 with one-way ANOVA with Dunnett’s multiple comparison test. (E) Ros production was assessed by measuring oxidation of DCFDA into DCF in neutrophils following 4 hours PMA stimulation with either no treatment, AZD1163 or R347 Bis3-FQQ-YTE control antibody or treatment with the ROS inhibitor DPI. Graphs show mean ± SEM values (n = 3). One-way ANOVA with Dunnett’s multiple comparison test, *** denotes p < 0.001, and **** denotes p < 0.0001. (F and G) Generation of neutrophil extracellular traps was measured by the quantification of extracellular DNA detected with cytotox green dye using IncuCyte live cell analysis. The effects of AZD1163 or control antibody were measured over a 6-hour period in the absence of NET induction (F) or after 6-hours incubation with 100 nM PMA to induce NETs (G). The graphs display the Integrated Intensity of NETs (GCU × µm2 /Image) represented as mean ± SEM (n = 3). AZD1163 does not trigger Fc-mediated effector functions or impact neutrophil functions. AZD1163 has an acceptable PK profile, protracted pharmacodynamics, and a favorable safety profile in non-human primates Since AZD1163 lacked rodent cross-reactivity to PAD4, to evaluate the PK and pharmacodynamics (PD), a single dose non-good laboratory practice (GLP) PK/PD study was conducted in cynomolgus monkeys (Macaca fascicularis). Groups of three female animals were dosed once by intravenous infusion (~30 minutes) with vehicle (Group A), or AZD1163 at 5 mg/kg plus weekly dosing of 10 mg/kg abatacept (Group B), 5 mg/kg AZD1163 (Group C), and 100 mg/kg of AZD1163 (Group D). Administration of abatacept to Group B was included to reduce the likelihood of immunogenicity against the test article in the low dose group. AZD1163 exhibited approximately dose-proportional PK over 5–100 mg/kg following a single intravenous dose, with average clearance of 2.2–3.6 mL/day/kg and terminal half-life of 14–20 d across the different groups (Figure 6(A)). Interpretation of dose proportionality was constrained by potential abatacept interference in the serum bioanalytical assay for Group B (5 mg/kg AZD1163 + abatacept) and by early loss of exposure consistent with suspected anti-drug antibodies in 2/3 animals in Group C (5 mg/kg AZD1163) and 1/3 animals in Group D (100 mg/kg AZD1163). Among Group D animals that maintained exposure throughout the study (n = 2/3), clearance and terminal half-life (~2.2 mL/day/kg and 20 d, respectively) were consistent with expectations for half-life-extended monoclonal antibodies in cynomolgus monkeys. Figure 6.AZD1163 inhibits pad activity in non-GLP PK/PD study in non-human primates.The x-axis label is Time (days). The x-axis range is 0 to 100. The y-axis label is Concentration (microgram per milliliter). The y-axis range is 10 superscript negative 1 to 10 superscript 4. Legend entries: Group B, 5 milligram per kilogram plus abatacept; Group C, 5 milligram per kilogram; Group D, 100 milligram per kilogram. Plotted points: Group B starts near 10 squared at day 0 and declines to about 10 to the power of 0 by day 100; Group C starts near 10 squared at day 0 and declines to about 10 to the power of 0 by day 100; Group D starts near 10 cubed at day 0 and declines to about 10 squared by day 100. The image B showing a line graph titled Group A. The x-axis label is Time (Days). The x-axis range is PRE Week 1 to Day 106. The y-axis label is PAD activity (Abs 450 nanometer). The y-axis range is 0.0 to 3.0. Legend entries: 1501 PLASMA, 1502 PLASMA, 1603 PLASMA, 1501 plus SPIKE, 1502 plus SPIKE, 1603 plus SPIKE, 1501 plus EDTA, 1502 plus EDTA, 1603 plus EDTA. Plasma lines fluctuate mostly between about 0.3 and 1.6 with one peak near 2.5. Plus spike lines stay around 2.0 to 2.5 across time. Plus EDTA lines stay near 0.0 to 0.2 across time. The image C showing a line graph titled Groups B, C, D. The x-axis label is Time (Days). The x-axis range is PRE Week 1 to Day 106. The y-axis label is PAD activity (Abs 450 nanometer). The y-axis range is 0.0 to 3.0. Legend entries: 5 milligram per kilogram AZD1163, 5 milligram per kilogram AZD1163 plus spike, 100 milligram per kilogram AZD1163, 100 milligram per kilogram AZD1163 plus spike, EDTA negative controls. 5 milligram per kilogram AZD1163 rises from about 2.2 at PRE VEH to near 0.0 by about day 4 and stays near 0.0 to about 0.2 through day 106. 5 milligram per kilogram AZD1163 plus spike increases from about 2.2 at PRE Week 1 to about 2.6 to 2.8 by about days 70 to 90 and remains above about 2.0 near day 106. 100 milligram per kilogram AZD1163 drops from about 2.2 at PRE Week 1 to near 0.0 by about day 4 and stays near 0.0 through day 106. 100 milligram per kilogram AZD1163 plus spike drops from about 2.2 at PRE Week 1 to near 0.0 by about day 4 and stays near 0.0 through day 106. EDTA negative controls remain near 0.0 across time.Three line graphs showing concentration and PAD activity over time for groups A to D.Non-GLP single IV dose PK/PD assessment of AZD1163 in cynomolgus monkeys. Groups of three cynomolgus monkeys were administered vehicle control (Group A), 5 mg/kg AZD1163 with weekly dosing of 10 mg/kg abatacept (Group B), 5 mg/kg AZD1163 (Group C), or 100 mg/kg AZD1163 (Group D). Two animals in Group C and one animal in Group D were excluded from the PK analysis and PD analysis because of suspected anti-drug antibodies (ADA) (premature loss of PK and PD). (A) PK assessment for AZD1163 at indicated timepoints for Groups B, C, and D. Samples with findings consistent with ADA-related interference are not shown (two animals in group C at t ≥ 21 and 28 d; one animal in group D at t ≥ 21 d) (B) PAD activity in plasma from individual animals in Group A vehicle controls (n = 3). (C) Average PAD activity for animals with 5 mg/kg AZD1163 administration (Groups B and C, n = 4) and 100 mg/kg AZD1163 administration (Group D, n = 2). The PAD activity was measured for endogenous plasma levels, plasma with equivalent of 400 ng/mL cyno PAD4 spike-in, and plasma with EDTA (background) as indicated. AZD1163 inhibits pad activity in non-GLP PK/PD study in non-human primates. To assess the PD of AZD1163, the histone H3 citrullination assay was used to measure the endogenous PAD activity in plasma samples at different time-points. To mimic high PAD levels that may be present in some patients with RA, the PAD activity was also measured in the presence of a 400 ng/mL cynomolgus PAD4 spike. In untreated animals, variable levels of endogenous PAD activity were observed in the plasma throughout the study. Consistently higher PAD activity was detected with the PAD4 spike-in, and all PAD activity was negated by the presence of ethylenediaminetetraacetic acid (EDTA) (Figure 6(B)). Following administration of AZD1163, all PAD activity was inhibited at the first post-treatment time-point (40 mins). Consistent with the three animals that exhibited an early loss of PK (two in Group C and one in Group D), there was also an early loss in suppression of PAD activity that was attributed to anti-drug antibody response. These three animals were excluded from the PD analysis, and the remaining 5 mg/kg animals from Group B and C were pooled for PD assessment. For the remaining 5 mg/kg administered animals (Groups B and C), there was no notable recovery of endogenous PAD activity on Day 64, and a gradual recovery in PAD activity was observed from Day 29 with the spiked samples (Figure 6(C)). For the 100 mg/kg group, there was no recovery of PAD activity in the endogenous or spiked plasma samples at the final sampling time-point at Day 106 (Figure 6(C)), which is consistent with the AZD1163 serum concentrations at the end of the study. The nonclinical safety of AZD1163 was evaluated in repeat dose toxicity studies in cynomolgus monkeys that included safety pharmacology endpoints. AZD1163 was well tolerated with no adverse toxicity findings, and no AZD1163-related changes in respiratory rate, heart rate, blood pressure, ECG intervals or ECG waveform morphology, or CNS function. These non-human primate studies demonstrate that AZD1163 exhibits favorable PK properties and a clearance consistent with an antibody engineered for improved FcRn recycling, a clear relationship between PK and PD, and a safety profile that would support evaluation in human clinical trials. Discussion Herein, we describe the preclinical development of AZD1163, a first-in-class bispecific antibody targeting PAD2 and PAD4, the enzymes responsible for the generation of citrullinated antigens that drive autoimmune responses in patients with RA. Current marketed therapies for RA target proinflammatory cytokines or nonspecifically target adaptive immune responses. We have taken a novel approach that specifically targets the generation of autoantigens by inhibiting the enzymatic activity of PAD2 and PAD4 that are prevalent in the circulation and joints of patients with RA. We demonstrate that AZD1163 specifically binds PAD2 and PAD4 and potently inhibits endogenous PAD activity and the generation of citrullinated autoantigens in serum, whole blood, and synovial fluid samples from patients with RA. Besides blocking the enzymatic activity of soluble PAD2 and PAD4, AZD1163 also binds and internalizes cell surface PADs to the low pH endosomes for degradation, thereby preventing them from contributing to the citrullinated autoantigen burden. Encouraging safety and in vivo pharmacology studies in cynomolgus monkeys have permitted its clinical development. To the best of our knowledge, AZD1163 is the first biopharmaceutical targeting the generation of autoantigens to be evaluated in clinical trials. Targeting PAD2 and PAD4 with a bispecific antibody provided the opportunity to specifically suppress cell surface and extracellular PAD activity, without impacting the potentially detrimental consequences of inhibiting intracellular PAD activity. The use of small molecule inhibitors with various degrees of specificity for PAD proteins have demonstrated efficacy in murine models of inflammatory arthritis,33–37 colitis,38 cancer,39 atherosclerosis,40 multiple sclerosis,41 and fibrosis,42 among others. The proposed mechanisms of action for these intracellular PAD inhibitors are diverse and include the regulation of NETosis, epigenetics, transcription factors and cancer suppressors, autophagy, and myelin organization. Unsurprisingly, generating small molecule inhibitors that target intracellular PAD functions and have appropriate selectivity, pharmacokinetic, and safety profiles is challenging. Since AZD1163 is a bispecific antibody developed to inhibit extracellular generation of citrullinated antigens, many of the challenges attributed to small molecules have been circumvented, with the caveat that immunomodulatory activities attributed to intracellular functions of PAD will not be modulated. Knowledge of the relative contributions of PAD2 and PAD4 in the generation of citrullinated antigens is evolving. Early reports indicated that PAD2, PAD3, and PAD4 had very specific protein specificities,31 but with improving analytical techniques our perception of the citrullinome has broadened.12,43 Analysis of the sequence in the proximity of the targeted arginine residue indicates that PAD2 and PAD4 have overlapping consensus sequences.32 Our pilot experiment demonstrates that both PAD2 and PAD4 enzymes can generate citrullinated antigens detectable by ACPA+ RA serum. Within the circulation, it was apparent that both PAD2, PAD4, and overall PAD activity were elevated in patients with RA compared to healthy controls. There was a statistically strong but modest correlation between the concentration of PAD4 and PAD activity in the RA sera (r2 = 0.4447 p < 0.0001), and there was no significant correlation between the PAD2 and PAD activity. In part, this may reflect the higher circulating PAD4 concentrations than PAD2, and it remains plausible that other factors may be involved. The PAD activity assay conditions would have normalized the calcium concentration, pH, and oxidation, but bicarbonate levels, auto-citrullination, antibodies including anti-PAD2 Abs, and potentially other serum factors could modulate activity of the enzymes.28,44,45 Measurement of PAD4 in blood samples would be a reasonable biomarker for assessing circulating PAD activity, although it remains uncertain how useful this would be for predicting PAD activity in the joint, considering a correlation between the PAD concentrations in the blood and synovial compartments was not established with the limited number of synovial samples evaluated in this study Despite the limited number of synovial fluid samples evaluated, it was evident that the amount of PAD2, PAD4, and overall PAD activity can be substantially higher than those found in the blood. The increased levels in the joint tissue likely reflect the localized inflammatory environment, and the superior blockade of PAD activity with AZD1163 compared with anti-PAD4 antibody alone highlights the importance of blocking both PAD2 and PAD4 in the synovium to minimize PAD activity and citrullination. Data linking synovial PAD activity, protein citrullination, and disease activity is limited. In RA joint tissue, Western blot and immunohistochemistry data have shown increased PAD2 and PAD4 proteins,46 and increased citrullinated proteins have been detected by immunohistochemistry and mass spectrometry.11,20,22 The presence of polyreactive anti-citrullinated protein antibody producing cells in the active RA joint and higher ACPA levels in the synovial compared to the blood compartment are indicative of a local autoimmune response,11,47,48 and increased PAD2 in the RA synovium correlated with disease activity in ACPA+ RA patients.19 Although our data indicate both PAD2 and PAD4 may contribute to local citrullination in synovial fluid, only a limited number of samples have been evaluated. A well-validated clinical cohort is needed to establish direct links between PADs and PAD activity in relation to disease activity and progression. This will be evaluated in subsequent studies. It is unlikely that other PADs significantly contribute to PAD activity in the blood or joints of patients with RA. PAD6 expression has not been reported in the joint and is not catalytically active.14 PAD1 and PAD3 protein expression is predominately limited to the skin. PAD3 mRNA has been detected in the joint,49 and PAD3 protein expression was detected in neutrophils and macrophage-like synoviocytes, although its significance remains uncertain.31,50 PAD3 citrullinates histone H3 in our assay system, but since AZD1163, which only inhibits PAD2 and PAD4, blocks all measurable PAD activity, PAD3 does not appear to contribute to extracellular PAD activity in RA patients. Besides the mammalian PADs, the Porphyromonas gingivalis PAD enzyme (pPAD) has been implicated in the etiology of RA by generating citrullinated antigens in the buccal cavity, resulting in a subsequent breakdown in tolerance and the development of the anti-citrullinated autoimmune response.51 Although pPAD may citrullinate human antigens,52 this hypothesis warrants further validation since pPAD only citrullinates C-terminal arginines,53 mainly in bacterial proteins,54 and serum from drug-naïve early arthritis patients failed to cross-react with the P. gingivalis citrullinome.55 Overall, our data highlight the importance and efficacy of targeting both PAD2 and PAD4 with AZD1163 to suppress all ex vivo extracellular PAD activity in the synovial fluid and blood samples from patients with RA. The presence of anti-PAD4 antibodies and anti-PAD3/4 cross-reactive antibodies has been linked to more progressive joint damage and interstitial lung disease in patients with RA,26,27 whereas anti-PAD2 antibodies negatively correlate with these features.28 Anti-PAD4 antibodies may promote PAD activity by binding to a cross-reactive PAD3/PAD4 epitope through binding a calcium binding site that promotes activity at low calcium concentrations.26 In our cohort of 87 patients with RA, we identified 15 patients (17%) that were seropositive for anti-PAD4 antibodies, which included 8 patients (9%) that were also seropositive for anti-PAD3 antibodies. These frequencies are somewhat lower than those previously published for anti-PAD4 and anti-PAD3/4 autoantibodies (37% and 12%).26 Contrary to our expectation, serum concentrations of PAD4 and overall PAD activity among the patients with anti-PAD4 antibodies were significantly lower than those seronegative for anti-PAD4 antibodies. Less PAD4 could suggest these autoantibodies impair detection of PAD4 in serum; however, as the PAD activity was also significantly lower, these data indicate anti-PAD4 antibodies may reduce circulating PAD4 concentrations or suppress PAD4 activity. The latter is consistent with a previous study that showed the majority of anti-PAD4 autoantibodies derived from patients with RA modestly suppressed citrullination of fibrinogen.56 Verification of low PAD activity associated with the anti-PAD4 and anti-PAD3/4 antibodies requires assessment in a larger cohort of patients with RA. The association of these anti-PAD4 antibodies with progressive joint damage and interstitial lung disease may reflect activation of complement and Fc-mediated effector responses or activation of cells expressing surface PADs.57,58 Since AZD1163 binds and internalizes PADs, reducing their cell surface availability without inducing cytokines or altering cell function in vitro, AZD1163 may perturb the pathogenic potential of anti-PAD4 antibodies. Importantly, AZD1163 can inhibit essentially all serum PAD activity irrespective of the status of PAD3 or PAD4 antibodies. AZD1163 was developed to block the generation of citrullinated autoantigens. Reducing the citrullinated autoantigen burden will limit the targets recognized by autoreactive cells and autoantibodies that mediate tissue injury and limit the availability of antigen to maintain a chronic autoimmune response. Moreover, besides generating neoepitopes, the PAD2 and PAD4 catalytic conversion of positively charged arginine to a neutral citrulline residue also trigger changes in the structure and function of proteins that may contribute to the pathology in the joint. For example, citrullination of fibrinogen promotes macrophage production of PDGF-BB and TGF-β, fibroblast activation, and extracellular matrix deposition,59,60 and citrullinated fibronectin can promote angiogenesis and a more invasive fibroblast phenotype.61,62 Consequently, besides suppressing the autoimmune response to citrullinated antigens, AZD1163 may impact disease processes in RA independent of the autoimmunity. In summary, we have generated a first-in-class bispecific antibody that potently binds and inhibits endogenous PAD2 and PAD4 activity in the blood and synovial fluid from patients with RA. The safety profile and in vivo pharmacology of AZD1163 in cynomolgus monkeys justified early clinical development (NCT06103877). Inhibition of the enzymes responsible for the generation of citrullinated antigens provides a potentially novel approach for the treatment of RA and other diseases driven by extracellular PAD activity. Materials and methods Human biological samples Blood, serum, plasma, and samples from RA patients were procured by Sanguine Biosciences. Twelve synovial fluid and matched serum samples from RA patients were obtained from Cureline (n = 7) and a small observational study (CD-IA-INDP-1056, n = 5). Serum and plasma from age-matched healthy individuals were collected from participants in the AstraZeneca internal blood collection program. Informed consent was obtained in all instances, and the studies were approved by the appropriate institutional review boards. Serum, plasma, and synovial fluid samples from patients with RA were collected, processed, aliquoted, and stored at −80°C prior to transfer to AstraZeneca. Fresh blood samples were collected and transported to the test facility within 4 hours of being drawn. PAD reagents Recombinant human PAD1, PAD2, PAD3, and PAD4, and cynomolgus monkey PAD2 and PAD4, murine PAD2 and PAD4, were produced in a high-yielding proprietary protein expression cell line based on Expi293F. All were tagged at the N-terminus with a TwinStrep-TEV-Avi tag. Expression constructs were transfected using PEI (Polysciences) and cells harvested for lysis prior to two steps of sequential purification (StrepTrap XT 5 ml and Superdex 200 26/600). Final protein concentrations were determined by UV280 and stock solutions stored at −80°C in buffer containing 20 mM HEPES, 500 mM NaCl, 0.5 mM TCEP, 5% Glycerol, 1 mM EDTA, pH 7.6. PAD binding assay The PAD enzymes were biotinylated specifically at the Avi Tag site using the BirA biotin–protein ligase reaction according to manufacturer’s protocol (Avidity # BirA500-RT). The excess biotin was removed by buffer exchange and size separation on a pre-equilibrated S200 16/600 column (50 mM HEPES, 300 mM NaCl, 5% Glycerol, 500 µM TCEP, 1 mM EDTA, pH7.6). 100% biotinylation was confirmed by liquid chromatography-mass spectrometry. Black high-binding plates (Greiner #781077) were coated with 20 µL of 2 µg/mL streptavidin (Invitrogen #S888) in Hank’s Balanced Salt Solution (HBSS) buffer for 16 hours at 4°C. Plates were washed with phosphate-buffered saline (PBS) containing 0.1% Tween20 three times and blocked with 80 µL 1% bovine serum albumin (BSA) in HBSS. After 1-hour plates were washed as described above and 20 µL of 1 µg/mL biotinylated PAD1, PAD2, PAD3, or PAD4 in 50 mM HEPES pH 7.3, 10 mM CaCl2, 120 mM NaCl, 5 mM dithiothreitol (DTT), and 0.1 mM CHAPS. After 1-hour, plates were washed and incubated 20 µL of 500 nM test or control antibodies in 50 mM HEPES pH 7.3, 10 mM CaCl2, 120 mM NaCl, 5 mM DTT, and 0.1 mM CHAPS were added. After 1.5 hours, plates were washed and 20 µL of 50 ng/mL anti-human IgG Fc horseradish peroxidase (HRP) (Southern Biotech #9040–05) in HBSS containing 0.5% BSA. After 1-hour, plates were washed and incubated with 20 µL of QuantaBlu for 30 mins at room temperature, and then 20 µL QuantaBlu stop solution was added (QuantaBlu kit ThermoFisher #15169). Fluorescence was measured using a 330 nm excitation filter and a 420 nm emission filter on a PHERAStar FSX plate reader (BMG Labtech). Binding affinity The affinities of the anti-PAD2 or anti-PAD4 Fabs to PAD proteins were measured using Biacore T100 or Biacore 8K (Cytiva) at 25°C. PAD2 and PAD4 proteins were generated in-house and biotinylated on an Avi-tag. Streptavidin was covalently immobilized to a CM5 (Cytiva #BR100530) or C1 (Cytiva #BR100535) chip surface using standard amine coupling techniques at a concentration of 16 µg/mL in 10 mM sodium acetate pH 4.5. Recombinant biotinylated PAD2 and PAD4 species were titrated onto the streptavidin chip surface in 10 mM HEPES pH 7.4, 150 mM NaCl, 0.05% Surfactant P20, 1 mM CaCl2, 1 mM DTT buffer to enable Fab binding. The anti-PAD2 and anti-PAD4 Fabs were serially diluted (0.0781–10 nM, 0.391–50 nM, or 9.38–500 nM) in 10 mM HEPES pH 7.4, 150 mM NaCl, 0.05% Surfactant P20, 1 mM CaCl2, 1 mM DTT and flowed over the chip at 50 µL/min, with 3 minutes association and 10 minutes dissociation. Multiple buffer-only injections were made under the same conditions to allow for double reference subtraction of the final sensorgram sets. Fully corrected binding data were globally fitted to a 1:1 binding model using the Biacore Insight Evaluation Software (Cytiva). The chip surface was fully regenerated with pulses of 3.0 M MgCl2. The affinities of recombinant PAD species to AZD1163 were measured using Biacore 8K (Cytiva) at 25°C. AZD1163 was minimally amine biotinylated using EZ link Sulfo-NHS-LC-biotin (ThermoFisher Scientific #A39257). The Sulfo-NHS-LC biotin was added to AZD1163 in 1:2 protein:biotin ratio, after 20 minutes of incubation at room temperature the sample was injected at 0.5 ml/minute onto a Superdex 200 Increase 10/300 GL column (Cytiva #28990944) equilibrated in D-PBS and the biotinylated AZD1163 isolated. The biotinylated AZD1163 was captured on a C1-Streptavidin surface (prepared as before). The PAD species were serially diluted (0.313–5 nM or 0.625–10 nM) in 10 mM HEPES pH 7.4, 150 mM NaCl, 0.05% Surfactant P20, 1 mM CaCl2, 1 mM DTT and injected in sequentially increasing concentrations (single cycle kinetics) over the chip at 50 µL/min, with 2 minutes association and 10 minutes dissociation at the end of the complete binding cycle. A buffer-only injection was made under the same conditions to allow for double reference subtraction of the final sensorgram sets, which were globally fitted to a 1:1- binding model using the Biacore Insight Evaluation Software (Cytiva). The streptavidin surface was fully regenerated with pulses of 3.0 M MgCl2. PAD2 and PAD4 protein ELISAs PAD2 and PAD4 levels were measured using PAD2 and PAD4 ELISA kits as described by the manufacturer (Cayman Chemicals #501450, #501460). Anti-PAD4 ELISA Meso Scale Discovery (MSD) plates were coated with 25 μL of 2 μg/mL PAD4 in PBS overnight at 2–8°C, washed and blocked with I-Block buffer (Fischer Scientific) for 1–4 hours at room temperature. A standard curve and quality control samples were prepared from the anti-PAD4 EIA Standard (Cayman Chemical #500930). Plates were washed and 25 μL of Standard, QC samples and diluted human serum samples were added and incubated for 1 hours at room temperature. Washed plates were incubated with 25 μL of 2 μg/mL ruthenylated PAD4 for 1 hours at room temperature. Plates were washed, 1x READ buffer was added, and electrochemiluminescence was measured on a Sector Imager plate reader (Meso Scale Discovery). Anti-PAD4 Ab titers >5000 U/mL were considered positive. Anti-PAD3 ELISA MSD plates were coated with 25 μL of 2 μg/mL GST-PAD3 fusion protein in PBS overnight at 2–8°C, washed and blocked with I-Block buffer (Fisher Scientific #T2015) for 1–4 hours at room temperature. Plates were washed and incubated with 25 μL of diluted serum samples for 1 hour at room temperature. Plates were washed and incubated with 25 μL of 2 μg/mL biotinylated human His10x_Avitag_Human PAD3 for 1 hour at room temperature. Plates were washed and 15 μL of 0.5 μg/mL Streptavidin SulfoTag was incubated for 30 mins. Plates were washed, 1X Read buffer was added electrochemiluminescence was measured on the MSD Sector Imager plate reader. Within consideration of the background from control serum samples, a signal above background at a 1:1024 dilution was considered positive. Histone H3 PAD activity assay Two similar Histone H3 citrullination assay protocols were used: one for the measurement of recombinant PAD activity, and another for the measurement of endogenous PAD activity in biological fluids. For the measurement of recombinant PAD activity, 50 µL of a solution of Histone H3 (Cayman, #10263) at 2 µg/mL in HBSS buffer was coated on to Black High Binding plates (Greiner, #781077) for 16 hours at 4°C. Plates were equilibrated to room temperature and washed with 100 µL PBS containing 0.1% Tween20 three times and were blocked with 80 µL 1% BSA (Sigma, #A7979) in PBS for 1 hour, and washed three times as previously described. Test and control antibodies were titrated 3-fold in deimination buffer (40 mM Tris-HCl pH 7.5, 5 mM CaCl2, 1 mM DTT final concentration) and mixed with equal volumes of either 0.05 ng/mL human PAD2, 0.15 ng/mL cyno PAD2, 0.1 ng/mL mouse PAD2, 0.3 ng/mL human PAD4, 0.8 ng/mL cyno PAD4, or 5 ng/mL mouse PAD4 prepared in deimination buffer and incubated for 30 minutes and then transferred to the histone H3 coated plate. EDTA (0.1 M), which inhibits calcium-dependent PAD activity, was added to samples to determine background signal. After 3 hours, plates were washed as described previously and 50 µL anti-Histone H3 (citrulline R2) (Abcam, ab176843) was added at a concentration of 250 ng/mL prepared in PBS containing 0.5% BSA. After 1 hour, plates were washed as described previously and 50 µL F(ab’)2 anti-Rabbit IgG HRP (ThermoFisher, A24537) was added at a concentration of 100 ng/mL prepared in PBS containing 0.5% BSA. After 1 hour, plates were washed as described previously and 50 µL of QuantaBlu working solution at room temperature was added. After 30 minutes, 50 µL QuantaBlu stop solution was added (QuantaBlu kit ThermoFisher, 15,169). Fluorescence was measured using a 330 nm excitation filter and a 420 nm emission filter on a PHERAStar FSX plate reader (BMG Labtech). For evaluation of the whole blood experiments, blood collected by venipuncture into sodium heparin tubes and transferred into deep 96-well tissue culture plates and incubated with a dilution series of test and control antibody for 20 mins. The 96-well plate was then centrifuged at 1800 ×g for 5 minutes, and the plasma was harvested and used immediately or stored at −80°C. To evaluate the PAD activity in biological fluids, 50 µL of a solution of Histone H3 (Cayman, #10263) at 1 μg/mL in HBSS buffer was and coated on half-area 96-well high-binding plates (Corning, #3690) and incubated overnight at 4°C. Plates were equilibrated to room temperature and washed with 1x MSD TBS-T wash buffer (MesoScaleDiagnostics #R61TX-1) three times, and blocked in 1% BSA (MesoScaleDiagnostics #R93BA-4) in TBS-T wash buffer for 1 hour at room temperature then washed three times as previously described. Plasma, serum, and synovial fluid samples were diluted in freshly prepared deimination buffer (40 mM Tris-HCl pH7.5, 15 mM CaCl2, 1 mM DTT final concentration). In some experiments, test and control antibodies were diluted in deimination buffer and preincubated with the biological samples for 30 minutes before 50 μL/well were transferred to the Histone H3 coated plates and incubated at 37°C for 1.5 hour or at room temperature for 2.5 hour. Plates were washed 3x in 1x MSD TBS-Tween wash buffer. Rabbit anti-human histone H3 (citrulline R8) antibody (Abcam #ab219406) was diluted in 1% BSA/TBS-T buffer at 1:2000 and added at 50 μL/well and incubated for 1 hour at room temperature. Plates were then washed 3x in 1x MSD TBS-T wash buffer. HRP-labeled F(ab’)2 goat anti-rabbit IgG cross adsorbed antibody (Invitrogen #A24537) was diluted in 1% BSA/TBS-T buffer at 1:2000 and added at 50 μL/well and incubated at room temperature for 1 hour. Plates were then washed 3x in 1x MSD TBS-T wash buffer. Room temperature equilibrated ultra-sensitive TMB (Sigma #ES022) was added at 50 μL/well and placed in the dark for 4–8 minutes. Reaction was quenched with 0.3 M sulfuric acid at 50 μL/well. Absorbance was read at 450 nm on an Envision multi-label plate reader. Internalization assays Whole blood was collected by venepuncture into sodium heparin tubes, and neutrophils were purified (>97%) with the EasySep™ Direct Human Neutrophil Isolation Kit (StemCell Technologies #19666) according to the manufacturer’s instructions. Fifty microliters of human neutrophils in X-VIVO 15 media containing 10% fetal bovine serum (FBS) were seeded on to 0.01% of P-L-Ornithine coated 96-well culture plates at the density of 8 × 104 per well to achieve 40–50% confluence. The cells were allowed to settle for 20 minutes at room temperature then 50 µL of 20 µg/mL of AZD1163 or control bispecific (R347 Bis3-FQQ-YTE) antibodies labeled with IncuCyte® FabFluor-pH Orange Antibody Labelling Reagent (Satorius #4812) in X-VIVO 15 media containing 10% FBS. The plates were placed in the IncuCyte® S×5 Live-Cell Analysis System (Sartorius) for 24 hours at 37°C with repeat scanning every 30 minutes to acquire images. For pulse-chase experiments, cells were incubated with unlabeled AZD1163 or control bispecific antibody for 6 hours (pulse), carefully washed twice with media, before 10 µg/mL of labeled AZD1163 or bispecific control antibody was added, and the plate was returned to the IncuCyte analysis system and monitored for an additional 18 hours (chase). Data was acquired and analyzed using New Analysis Definition and Cell Classification Analysis software IncuCyte SX5) to obtain the percentage of cells expressing the orange fluorescence and the internalized integrated intensity of the positive orange objects (Orange Calibrated Units × µm2/well). Data was exported and plotted with GraphPad Prism 9.4. CDC assay Neutrophils, monocytes, or Raji cells were resuspended in RPMI media containing 2 ng/mL human TNF (Sigma #H8916-10UG) and seeded at 100K per well in a 96-well plate in RPMI media. The cells were incubated for 6 hours without treatment or with either AZD1163 or control antibody (1 μg/mL or 10 mg/mL) or 5 mg/mL Anti-hCD20-hIgG1- rituximab biosimilar (InVivogen, #hcd20-mab1) in the presence of 10% human serum or 10% heat-inactivated (56°C for 30 min) serum. Cells were then lysed by adding Cell titer Glo reagent (Promega #G7570) and luminescence was measured using Envision microplate reader. The values were normalized to the untreated condition in heat-inactivated serum, which was considered as 100% cell viability. ADCC assay Neutrophils, monocytes, Raji nulls, and KC1333 cells were resuspended in RPMI media containing 2 ng/mL human TNF at 2x106 cells/mL concentration. Target cells (neutrophils, monocytes, or Raji cells) and KC1333 cells (effector cells) were seeded at 1:1 ratio either untreated or with AZD1163 or control antibody (1 mg/mL or 10 mg/mL) or 5 mg/mL rituximab. Cells were incubated for 6 hours and lysed with Cell titer Glo reagent (Promega #G7570) and luminescence was measured using Envision microplate reader. The values were normalized to the untreated condition, which was considered as 100% cell viability. Cytokine release assay Blood collected by venepuncture into sodium heparin tubes, transferred to 96-well round-bottom polystyrene culture plates and incubated at 37°C with 5% CO2 for 24 hours with antibodies to a final concentration of 10 µg/mL with PBS. The plate was centrifuged at 2000 × g for 15 minutes, and the cell-free supernatant was collected and stored at −80°C. Supernatants were thawed and IFN-γ, IL-2, IL-6, and TNF-α levels were measured using Milliplex human cytokine panel as described by the manufacturer. Phagocytosis assay Neutrophils were seeded in 96-well plates (100k/50 µL), preincubated with or without antibodies in 50 µL RPMI media for 4 hours, and incubated with 20 µg/100 µL pHrodo green Staphylococcus aureus BioParticles (ThermoFisher #P35367) in RPMI media for 1 hour at 37°C. The phagocytosis inhibitor control Cytochalasin D (Life Technologies #PHZ1063) was used at 10 µM. Neutrophils with phagocytosed pHrodo Green bioparticles were identified by flow cytometry. ROS assay Neutrophils were seeded in 96-well plates (100k/50 µL), preincubated with or without antibodies in 50 µL RPMI media for 4 hours. Neutrophils were stimulated with 100 nM PMA (Millipore #P1585-1 MG) and ROS production was assessed using 20 µM 2,’7’–dichlorofluorescin diacetate (Abcam #ab113851). DCFDA is deacetylated by cellular esterases to a non-fluorescent compound and oxidized by ROS into DCF. After a 30 minutes incubation at 37°C, the highly fluorescent DCF is detected by flow cytometry. The NADPH oxidase inhibitor diphenyleneiodonium (DPI) inhibits ROS production and was used at 20 µM as a control. NETosis assay Neutrophils were seeded on fibronectin-coated Imagelock 96-well plates (Satorius Corp #F1141, #4379) at 20k/well) in F12K media supplemented with Cytotox Green Dye (Sartorius Corp #4633) and incubated in the presence or absence of antibodies for 6 hours at 37°C. NETosis was induced with 100 nM PMA. Cytotox Green fluorescens in the presence of extracellular DNA released and was imaged continuously using the Incucyte® Live-Cell Analysis System (Sartorius). Acquired images were analyzed using the Incucyte analysis software. Non-human primate studies All animal studies were conducted in the United States of America in AAALAC accredited facilities. They were conducted in accordance with the Animal Welfare Act, current guidelines for the nonclinical safety evaluation of on biotechnology derived pharmaceuticals,63 and AstraZeneca’s Bioethics Standard. To evaluate the PK, PD, and to provide an early safety assessment, a single dose non-GLP PK/PD study was conducted in cynomolgus monkeys (Macaca fascicularis). Groups of three purpose-bred females (Orient BioResource, Alice, TX, USA) were dosed once by intravenous infusion (~30 minutes) with vehicle (Group A), or AZD1163 at 5 mg/kg plus weekly dosing of 10 mg/kg abatacept (Group B), 5 mg/kg AZD1163 (Group C), and 100 mg/kg (Group D). Administration of abatacept to Group B was included to reduce the likelihood of immunogenicity against the test article in the low-dose group. After conclusion of the study, the animals were returned to the colony. For PK assessments, blood was collected from each animal pre-dose and after treatment at the following timepoints 40 minutes, 4 hours, 8 hours, and 12 hours on Day 1; at 24 hours (Day 2), 48 hours (Day 3), 72 hours (Day 4), and 120 hours (Day 6); and subsequently on Days 8, 11, 15, 22, 29, 36, 43, 50, 57, 64, 71, 78, 85, 92, 99, and 106. PK samples were processed to serum and analyzed for AZD1163 concentrations using a qualified electrochemiluminescence immunoassay which employed MSD technology and utilized an anti-human-Fc specific monoclonal antibody to capture AZD1163 that was detected via an in-house monoclonal antibody specific for the YTE modification introduced into the Fc region of AZD1163. For PD sampling, blood was collected by venepuncture into sodium heparin tubes from the animals prior to treatment with vehicle or test articles, and after treatment at the following timepoints 40 minutes (Day 1), 24 hours (Day 2), 72 hours (Day 4), and then on Days 8, 15, 22, 29, 36, 43, 50, 57, 64, 71, 78, 85, 92, 99, and 106. The endogenous PAD activity in plasma samples was evaluated using the Histone H3 activity assay. To mimic the potentially high levels of PADs present in the blood of RA patients, the PAD activity was also evaluated in cynomolgus monkey plasma samples spiked with the equivalent of 400 ng/mL of recombinant cynomolgus monkey PAD4. The potential toxicity and toxicokinetics of AZD1163 was evaluated in GLP-compliant repeat dose toxicity studies that included safety pharmacology endpoints (including assessments of cardiovascular, respiratory, and neurological parameters). Supplementary Material Supplementary Materials.docx