Letters https://doi.org/10.1038/s41587-019-0135-x A coiled-coil masking domain for selective activation of therapeutic antibodies Vivian H. Trang1, Xinqun Zhang1, Roma C. Yumul2, Weiping Zeng2, Ivan J. Stone2, Serena W. Wo1, Melissa M. Dominguez1, Julia H. Cochran1, Jessica K. Simmons2, Maureen C. Ryan3, Robert P. Lyon1, Peter D. Senter4 and Matthew R. Levengood 1* The use of monoclonal antibodies in cancer therapy is limited by their cross-reactivity to healthy tissue. Tumor targeting has been improved by generating masked antibodies that are selectively activated in the tumor microenvironment, but each such antibody necessitates a custom design. Here, we present a generalizable approach for masking the binding domains of antibodies with a heterodimeric coiled-coil domain that sterically occludes the complementarity-determining regions. On exposure to tumor-associated proteases, such as matrix metalloproteinases 2 and 9, the coiled-coil peptides are cleaved and antigen binding is restored. We test multiple coiled-coil formats and show that the optimized masking domain is broadly applicable to antibodies of interest. Our approach prevents anti-CD3-associated cytokine release in mice and substantially improves circulation half-life by protecting the antibody from an antigen sink. When applied to antibody–drug conjugates, our masked antibodies are preferentially unmasked at the tumor site and have increased antitumor efficacy compared with unmasked antibodies in mouse models of cancer. Antibody therapeutics can be associated with substantial side effects when target antigens are not exclusively expressed on the tissue of interest1,2. Targets with abundant expression in diseased tissues but low normal tissue expression have generally proven difficult to identify, prompting the development of strategies that harness other differences between tissues. One recent example has been the development of antibody prodrugs, in which a masking group is fused to the antibody to restrict binding in healthy tissues while leveraging disease-associated proteases to remove the mask in diseased environments3. This approach requires a custom mask for each antibody4–11, which limits rapid application and leads to uncertainties regarding expression, pharmacokinetics, immunogenicity, activation and stability. Here, we set out to design a more generalizable approach to antibody masking that could potentially address these limitations. The N-terminal residues of the light and heavy chains of a typical IgG are separated by approximately 30–35 Å and flank the complementarity-determining regions (CDRs) that form the paratope responsible for imparting antigen specificity (Fig. 1a). We reasoned that fusion of a peptide domain with propensity to form robust secondary structure could sterically block the CDRs from binding antigen. We focused on leucine zipper coiled-coils because they are highly tunable peptide domains that have been used extensively for antibody-engineering approaches, and some of these domains are derived from native human proteins (Supplementary Table 1)12. As an initial test of this strategy, different classes of coiled-coils were evaluated, including parallel heterodimeric13,14 coils with low (CC1) and high (CC2) inter-coil affinities and disulfide-linked covalent coils (CC3)15, as well as antiparallel heterodimeric (CC4)16 and helixturn-helix homodimeric (CC5)17 pairs (Fig. 1b, Supplementary Table 1 and Supplementary Fig. 1). These peptide domains were genetically fused to the N termini of the CD19-binding antibody hBU12 (ref. 18) through protease-cleavable linkers. The linker sequences chosen (PLG-LAG19 and IPVS-LRSG20) are substrate sequences for multiple matrix metalloproteinases (MMPs) such as MMP-2 and MMP-9 that are upregulated in a variety of cancers21,22. Binding of the coiled-coil hBU12 fusions was assessed on CD19expressing Raji cells by flow cytometry via competition with Alexa Fluor 488-labeled hBU12, and it was found that the constructs had decreased affinity compared with unlabeled hBU12 (Fig. 1c). The effects for the antiparallel and micromolar-affinity parallel coils were relatively modest, decreasing the affinity by 38- and 83-fold, respectively, whereas the nanomolar-affinity parallel (CC2A), covalent and helix-turn-helix domains reduced binding by over 300-fold. These results suggest that inter-coil affinity is an important factor for blocking binding. An intriguing application for the masking concept is for antibody–drug conjugates (ADCs) that can deliver potent cytotoxic small molecules to antigen-expressing cells. While the covalent parallel heterodimeric coiled-coil CC3 demonstrated excellent blocking activity, the N-terminal disulfide is incompatible for masked ADCs that employ endogenous disulfides for drug attachment. For this reason, a high-affinity parallel heterodimeric coiled-coil was designed that was devoid of cysteine residues and the resulting fusion demonstrated marked loss in binding (750-fold) compared with the parental monoclonal antibody (CC2B; see Fig. 1d). This confirmed that the N-terminal disulfide was not required for efficient blocking. To test whether cleavage of the CC2B mask could restore antibody binding, CC2BhBU12 was incubated with purified human MMP-2, resulting in efficient site-specific removal of the mask, which was confirmed by reverse-phase liquid chromatography–mass spectrometry (LC–MS) (Supplementary Fig. 2). Binding of the precleaved antibody was restored to that of unmasked hBU12, as assessed by saturation binding on Raji cells (Fig. 1d). To better understand how the CC2B mask impacts antibody binding, the kinetics of binding for CC2B-hBU12 were assessed by biolayer interferometry using recombinant CD19. As has been reported with other masking modalities5–7, the decreased affinity of Department of Protein Sciences, Seattle Genetics Inc., Bothell, WA, USA. 2Department of Nonclinical Science, Seattle Genetics Inc., Bothell, WA, USA. Department of Antibody Discovery, Seattle Genetics Inc., Bothell, WA, USA. 4Department of Chemistry, Seattle Genetics Inc., Bothell, WA, USA. *e-mail: mlevengood@seagen.com 1 3 Nature Biotechnology | VOL 37 | JULY 2019 | 761–765 | www.nature.com/naturebiotechnology 761 Letters NaTure BiOTeCHnOlOgy a b Light chain N terminus Coil masking domain hBU12 Micromolar parallel (CC1) Nanomolar parallel (CC2A) Covalent parallel (CC3) Antiparallel (CC4) Helix-turn-helix (CC5) 50 100 10,000 Geometric mean (MFI) Maximum binding (%) 100 1 s-s Antiparallel Helix-turn-helix heterodimeric (CC4) homodimeric (CC5) d Antibody binding to CD19+ Raji cells 0 0.01 Covalent (CC3) s-s Antibody c High affinity (CC2) Cleavable linker 32 Å Heavy chain N terminus Parallel heterodimeric Low affinity (CC1) 250,000 Antibody binding to CD19+ Raji cells hBU12 200,000 CC2B-IPVSLRSG -hBU12 Cleaved 150,000 100,000 50,000 0 0.01 1 100 10,000 Concentration (nM) Competitor (nM) Fig. 1 | Steric blockade of antibody CDRs as a modular antibody masking system. a, Structure of an antibody Fab domain depicting the CDRs (blue) spanning approximately 32 Å between the N termini of the heavy and light chain (rituximab, PDB 4KAQ). b, Schematic depicting a coiled-coil masking domain fused to an antibody via cleavable linkers (left) and coiled-coils with various orientations and inter-coil interactions (right). c, Binding of coiled-coil masked anti-CD19 hBU12 antibodies was assessed using flow cytometry via competition against Alexa Fluor 488-labeled hBU12 on CD19+ Raji cells. Data represent the mean. The sequences for each coiled-coil domain are shown in Supplementary Fig. 1. d, The binding of CC2B-hBU12 compared with hBU12 was assessed via saturation binding on CD19+ Raji cells. For the cleaved antibody, the CC domain was removed using recombinant MMP-2 as described in the Methods. For c and d, the data shown are mean values of independent duplicate cell samples, and two independent experiments were conducted with similar results. CC2B-hBU12 to CD19 is driven largely by a slower rate of association of the antibody, whereas the dissociation rate is only modestly affected (Supplementary Fig. 3). One distinguishing aspect of this technology is that the CC2B peptides are fused to all four antibody N termini (two light chain, two heavy chain). To understand whether this could impact efficient protease activation, we designed ‘half-cleavable’ monoclonal antibodies that contained an MMP-2cleavable sequence on only one chain (either light or heavy) and a site that is not cleaved by MMP-2 (LALGPG19) on the opposing chain. These monoclonal antibodies were treated with purified MMP-2 until either the light chain or heavy chain had been cleaved, and the resulting products were evaluated for binding by flow cytometry on CD19+ Raji cells (Supplementary Fig. 4). Two cleavage events (out of a possible four) were sufficient to restore hBU12 binding to within approximately 10-fold of the parent antibody, whereas the intact masked monoclonal antibodies had over 900fold reduced affinity. Therefore, fusion of a coiled-coil peptide to a single antibody chain provides approximately 1% of the blocking ability of a coiled-coil mask that bridges the heavy and light chains. Further, these results demonstrate that limited mask proteolysis can have a dramatic impact on CC2B-mAb binding. The CC2B domain was then applied to several other monoclonal antibodies, including rituximab (anti-CD20)23, trastuzumab (antiHER2)24, h15H3 (anti-αVβ6)25 and 145-2C11 (anti-mouse CD3)26, and binding for each antibody was tested on antigen-positive cells by flow cytometry. For each of the CC2B-masked antibodies, minimal binding was observed at concentrations as high as 2 μM, representing decreases in affinity of at least 80-fold for rituximab, 470-fold for trastuzumab, 290-fold for 145-2C11 and 1,000-fold for h15H3 (Fig. 2a and Supplementary Table 2). For each of these antibodies, the mask could be efficiently removed through incubation with purified MMP-2, and cleavage restored binding to within 762 1.7-fold of the parent antibody. These results demonstrate that, in contrast to previously described methods4–10, the CC2B domain is modular and is a general solution for antibody masking. With this approach, there is no need for extensive screening to identify suitable masking domains, and the extent of masking achieved with the CC2B domain meets or exceeds those obtained using individualized antigen-blocking technologies4–10. The CC2B mask was applied to antibodies with a wide range of affinities (0.4 nM for h15H3 compared with 25 nM for rituximab) and receptor densities (100,000 copies per cell for CD19 on Raji cells compared with 1.3 million copies per cell for HER2 on SK-BR-3 cells). An additional benefit is that the masking domain has no intrinsic affinity for the antibody paratope, minimizing the possibility that the mask will remain bound once the linker has been cleaved27. We next examined the effects of the coiled-coil mask on monoclonal antibody internalization, a key factor that drives ADC activity. A quenched fluorescence assay was utilized to track hBU12 internalization into CD19+ Ramos cells over the course of 14 h, showing that CC2B-hBU12 had decreased internalization compared with hBU12, both in terms of the rate and extent of internalization (Fig. 2b). The impact of masking on in vitro cytotoxicity was then evaluated utilizing four-load ADCs bearing an auristatinbased drug-linker, mc-MMAF28. The unmasked hBU12 and MMP-activated masked ADCs both had half-maximum inhibitory concentration (IC50) values of 2.5 ng ml−1 on Ramos cells after 96 h continuous incubation, whereas the masked ADC was much less active, with an IC50 of approximately 1.5 μg ml−1 (Fig. 2c). To further examine the biological consequences of masking, we assessed the complement-dependent cytotoxicity (CDC) activity of rituximab and CC2B-rituximab. Rituximab (4 μg ml−1) elicited approximately 60% maximal cell lysis on CD20+ Raji cells in the presence of human serum, whereas no lysis was observed with Nature Biotechnology | VOL 37 | JULY 2019 | 761–765 | www.nature.com/naturebiotechnology Letters 50,000 CC2B-hBU12-mcMMAF(4) Cleaved 4 6 8 Time (h) 10 12 14 1, 00 0 10 ,0 00 2 10 0 0 0 0. 1 0 hBU12-mcMMAF(4) 50 Rituximab CDC on CD20+ cells 80 60 40 20 0 C C 2B Concentration (ng ml–1) 1,000 co nt ro l 100,000 d Cytotoxicity of ADCs on CD19+ Ramos cells 100 10 le av ed 150,000 0.1 Concentration (nM) yp e CC2B-IPVSLRSGhBU12 Survival (%) Signal intensity per cell per well c hBU12 200,000 0 0.001 Concentration (nM) Antibody internalization on CD19+ Raji cells 250,000 10,000 ot b 100 10,000 ab Concentration (nM) 1 Cleaved Is 0 0.0001 0.01 10,000 CC2B-IPVSLRSGanti-mouse CD3 20,000 itu xi m 100 10,000 Anti-mouse CD3 R 1 Cleaved Maximum lysis (%) 0 0.0001 0.01 20,000 30,000 SG 50,000 CC2B-IPVSLRSGtrastuzumab C 100,000 Cleaved Trastuzumab LR CC2B-IPVSLRSGrituximab VS 150,000 30,000 Antibody binding to CD3+ HT-2 cells -IP Rituximab 1 200,000 Antibody binding to HER2+ SK-BR-3 cells Geometric mean (MFI) Antibody binding to CD20+ Raji cells 10 Geometric mean (MFI) a Geometric mean (MFI) NaTure BiOTeCHnOlOgy Fig. 2 | Masking and reactivation of CC2B-masked antibodies. a, The binding of CC2B-masked rituximab (anti-CD20), trastuzumab (anti-HER2) and 145-2C11 (anti-mouse CD3) was assessed on CD20-, HER2- and mouse CD3-expressing cells, respectively. Cleaved comparators were generated using recombinant MMP-2 before the binding study. Data represent the mean values of two independent cell samples and are representative of two independent experiments. b, Internalization of anti-CD19 antibodies CC2B-hBU12 and hBU12 on CD19+ Raji cells was compared using a quenched-fluorophore pair (Alexa Fluor 488 and QSY9). On catabolism of the internalized antibody, a fluorescent signal is released and measured using a high-content fluorescence imager. Data represent the mean of two independent cell samples and are representative of two independent experiments. c, hBU12-mc-MMAF(4) ADCs were assessed for cell killing after 96-h incubation at 37 °C on CD19+ Ramos cells. Cytotoxicity was assessed using CellTiter-Glo (Promega). Data represent the mean ± s.d. of four independent cell samples and are representative of two independent experiments. d, CDC was assessed for masked and unmasked rituximab using CD20+ Raji cells. Cells were incubated with 4 μg ml−1 antibody for 2 h at 37 °C in the presence of human serum and assessed for cell lysis using Sytox Green (ThermoFisher). Data represent the mean of two individual cell samples and are representative of two independent experiments. MFI, mean fluorescence intensity. CC2B-rituximab or an isotype control antibody (Fig. 2d). Re-activated CC2B-rituximab had activity comparable to the unmasked antibody. These results demonstrate that the mask is able to conditionally prevent binding and biological function, and that removal of the mask restores activity. Anti-CD3 antibodies have been shown to induce rapid and profound on-target toxicity in both humans and mice29–31. We utilized an anti-mouse CD3 mAb (145-2C11) to test whether the coiledcoil domain could prevent systemic CD3 binding on lymphocytes and the subsequent production of pro-inflammatory cytokines in vivo. As shown in Fig. 2a, the anti-mouse CD3 antibody is effectively masked with the CC2B domain, as no binding was observed at concentrations up to 2 μM on CD3+ HT2 cells. Additionally, the unmasked antibody is able to induce mouse T-cell proliferation when immobilized on a culture dish at a concentration of 500 ng ml−1, whereas no proliferative effect was observed when cells were treated with the same concentration of a masked antibody (Supplementary Fig. 5). After confirming that masking impaired in vitro function, we compared the pharmacokinetics and cytokine production of masked and unmasked antibodies in immunocompetent BALB/c mice. Following intravenous administration at a dose of 0.5 mg kg−1, the parental anti-mouse CD3 antibody was undetectable in plasma 2 d after dosing, suggesting the presence of an antigen sink due to abundant CD3 expression (Fig. 3a). In contrast, the masked antibody displayed the same pharmacokinetic profile as a non-binding isotype control antibody, consistent with the mitigation of systemic CD3 binding. We next probed the consequences of masking on the production of inflammatory cytokines such as interferon-γ (IFN-γ) and interleukin-2 (IL-2). As expected, the parental antibody elicited a large spike in inflammatory cytokines 4 h post-dose, in contrast to the masked version (Fig. 3b). These results demonstrate that the mask protects the antibody from an antigen sink and unwanted systemic CD3 engagement, which both enhances circulation half-life and mitigates on-target effects. To evaluate whether mask activation could occur in tumors, we tested the activity of anti-human CD19-targeted ADCs in a human CD19+ Ramos non-Hodgkin lymphoma xenograft tumor model. Before initiating these studies, we confirmed the presence of protease activity in the Ramos model using in situ zymography with fluorescently labeled antibodies (Supplementary Fig. 6)4. We then dosed anti-CD19 or control IgG ADCs at 6 mg kg−1 in severe combined immunodeficient (SCID) mice bearing 200 mm3 subcutaneous tumors. Both CC2B-hBU12-mc-MMAF(4) and an unmasked ADC had similar activities while a CC2B-hBU12 ADC containing the inefficiently cleaved sequence (LALGPG) was less active (Fig. 3c). As both ADCs had approximately 600-fold reductions in cytotoxicity experiments in vitro, the results demonstrate cleavage sequence-dependent anti-tumor activity. We tested the consequences of masking on the pharmacokinetics and efficacy of an ADC that cross-reacts with mouse antigen. h15H3 is an antibody directed against the αVβ6 integrin complex that binds to both human and mouse antigens with similar affinities25. This integrin complex is expressed in a variety of tumors but is also expressed on skin in mice32. Binding of the masked antibody to human integrin αVβ6-transfected human embryonic kidney (HEK) cells was impaired by greater than 1,000-fold (Supplementary Fig. 7), as was the cytotoxic activity of glucuronide-cleavable MMAE33 ADCs on αVβ6+ BxPC3 cells (Supplementary Fig. 8). Binding and Nature Biotechnology | VOL 37 | JULY 2019 | 761–765 | www.nature.com/naturebiotechnology 763 Letters NaTure BiOTeCHnOlOgy b 4,000 CC3-IPVSLRSGanti-mouse CD3 3,000 Isotype control 2,000 10 15 Time (d) CC3-IPVSLRSGanti-mouse CD3 3,000 0 Isotype control 2,000 Single dose 6 mg kg–1 at day 10 300 45 5 4,000 Isotype control 35 40 0 Anti-mouse CD3 Untreated 600 25 30 0.01 IFN-ϒ 5,000 CC2B-LALGPG-hBU12 900 0 Animals dosed at 0.5 mg kg–1 CC2B-PLGLAG-hBU12 1,200 15 20 0 4 8 12 16 20 24 Time (h) Isotype control LLQ (0.08 µg ml–1) hBU12 1,500 0 CC3-IPVSLRSG -anti-mouse CD3 0.1 ADC activity in Ramos xenograft 1,000 Anti-mouse CD3 1 Anti-mouse CD3 5 10 Concentration (pg ml–1) 10 Concentration (pg ml–1) Total Ab (µg ml–1) Pharmacokinetics of antibodies in BALB/c mice c IL-2 5,000 Median tumor volume (mm3) a Time (d) 1,000 0 4 8 12 16 20 24 0 Time (h) 1,000 h15H3 10 CC2B-PLGLAG-h15H3 CC2B-LALGPG-h15H3 1 Animals dosed at 1 mg kg–1 0.1 –1 LLQ (0.015 µg ml ) 0.01 0 5 10 15 h15H3 800 CC2B-PLGLAG-h15H3 600 Untreated 400 Single dose 3 mg kg–1 at day 6 200 0 0 10 20 30 40 50 60 70 Time (d) Tumor 20 Plasma Liver 15 10 5 0 PL Time (d) Activation of CC2B-h15H3 ADCs in HPAF-II xenograft L LA AG LG PG PL G LA LAG LG P PL G G L LA AG LG PG 100 f ADC activity in BxPC3 xenografts G Pharmacokinetics of ADCs in nude mice Cleaved heavy chain (%) e Mean tumor volume (mm3) Total Ab (µg ml–1) d Fig. 3 | In vivo assessment of masked antibodies and ADCs. a, The pharmacokinetics of 3H-labeled CC3-masked and unmasked anti-mouse CD3 antibodies were compared in BALB/c mice following 1 intravenous (i.v.) dose of 0.5 mg kg−1 (n = 3 mice per time point; data represent the mean ± s.d.). b, The levels of IL-2 and IFN-γ were quantified over the course of 24 h from mice treated with 1 mg kg−1 CC3-masked or unmasked anti-mouse CD3 antibodies (n = 3 mice per time point; data represent the mean ± s.d.). c, The efficacy of anti-human CD19 hBU12-mc-MMAF(4) ADCs was assessed in a human CD19+ Ramos xenograft in SCID mice. Mice were administered 1 intraperitoneal (i.p.) dose of 6 mg kg−1 once tumors reached 200 mm3. Data represent the median value of n = 5 mice. d, The pharmacokinetics of 3H-labeled anti-human/mouse αVβ6 CC2B-h15H3-gluc-MMAE(8) and h15H3-glucMMAE(8) ADCs were compared in nude mice after a single i.v. dose of 1 mg kg−1 (n = 3 mice per time point; data represent the mean ± s.d.). e, The efficacy of anti-human/mouse αVβ6 h15H3-gluc-MMAE(8) ADCs was assessed in a human αVβ6+ BxPC3 xenograft. Mice were administered 1 3 mg kg−1 i.p. dose once tumors reached 100 mm3. Data represent the mean value of n = 5 mice. f, CC2B-mask cleavage was assessed in tumor, plasma and liver samples of HPAF-II tumor-bearing mice at 4 d post-dose. Shown is the percentage of cleaved heavy chain, as determined by western blot analysis of antibodies purified from each tissue. Data represent the mean ± s.d. (n = 3 mice for PLGLAG and n = 2 mice for LALGPG). The difference in cleaved antibody for PLGLAG and LALGPG linkers in the tumor was statistically significant (P = 0.045, two-tailed unpaired t-test). For a–f, each experiment was conducted once but included control test articles that provided similar response in at least one additional independent experiment. LLQ, lower limit of quantitation. activity could be restored on mask cleavage. As with the anti-CD3 mAb, the unmasked αVβ6-binding ADC cleared quickly from circulation in mice due to the presence of an antigen sink, whereas the masked forms of the ADC had prolonged circulation half-lives (Fig. 3d). To assess the impacts of improved pharmacokinetics on efficacy, we tested ADC activity in three different cell line xenograft tumors that express αVβ6 integrin. In two of these models (BxPC3 and Detroit 562), increased circulation of the masked ADC led to improved efficacy in comparison with the unmasked h15H3 ADC (Fig. 3e and Supplementary Fig. 9). The masked ADC bearing a cleavable PLGLAG linker and the parent h15H3 ADC had comparable efficacy in an HPAF-II pancreatic carcinoma xenograft model, but a masked ADC bearing the scrambled LALGPG linker had only minimal effects on tumor growth (Supplementary Fig. 10). We used western blot analysis to establish ratios of masked and unmasked ADCs in the tumor 4 d after administration (Fig. 3f and Supplementary Fig. 11). There was a statistically significant (P = 0.045) 7-fold increase in cleaved ADC for the PLGLAG linker compared with the scrambled sequence, suggesting that in vivo efficacy (Supplementary Fig. 10) correlates with intratumoral activation. A similar activation trend was observed using in situ zymography (Supplementary Fig. 12). While 764 the in vivo conversion does not take into account the portion of activated ADC that bound antigen and internalized inside target cells, even low-level conversion can lead to pronounced increases in tumor antigen binding (Supplementary Fig. 13). Finally, we showed (Fig. 3f and Supplementary Fig. 11) that there was no circulating unmasked ADC resulting from cleavage of the linkers from the protease-sensitive or scrambled ADC constructs. To exclude the possibility that the unmasked ADC might be taken up by antigen-positive cells, we injected both tumor-bearing and naïve mice with a non-binding CC3-masked control antibody. In both cases, the cleaved antibodies in circulation were barely detectable (Supplementary Fig. 14). These results are consistent with systemic stability and selective intratumoral activation. We have described a modular approach for conditional antibody masking, in which parallel heterodimeric coiled-coils with high inter-coil affinities reversibly impair antibody binding and biological function. Unlike designer antibody masks that can require extensive screening to identify an antibody that is efficiently masked, this approach is general and the same mask can be readily applied to an array of antibodies of interest. The pharmacological benefits of antibody masking via coiled-coils include improved antibody pharmacokinetics and mitigation of unwanted side effects, such as Nature Biotechnology | VOL 37 | JULY 2019 | 761–765 | www.nature.com/naturebiotechnology Letters NaTure BiOTeCHnOlOgy cytokine release, that arise from systemic target engagement. With ADCs, the inclusion of disease-specific protease-cleavable linkers between the mask and antibody can impart improved efficacy of masked antibody therapeutics compared with unmasked controls due to avoidance of an antigen sink. This masking approach may facilitate the targeting of antigens that are inaccessible to conventional approaches due to undesirable normal tissue expression. The system described here should enable the rapid identification and optimization of antibodies for therapeutic applications. Online content Any methods, additional references, Nature Research reporting summaries, source data, statements of code and data availability and associated accession codes are available at https://doi.org/10.1038/ s41587-019-0135-x. Received: 21 August 2018; Accepted: 16 April 2019; Published online: 27 May 2019 References 1. Bugelski, P. J. & Martin, P. L. Concordance of preclinical and clinical pharmacology and toxicology of therapeutic monoclonal antibodies and fusion proteins: cell surface targets. Brit. J. Pharmacol. 166, 823–846 (2012). 2. Hansel, T. T., Kropshofer, H., Singer, T., Mitchell, J. A. & George, A. J. The safety and side effects of monoclonal antibodies. Nat. Rev. Drug Discov. 9, 325–338 (2010). 3. Polu, K. R. & Lowman, H. B. Probody therapeutics for targeting antibodies to diseased tissue. Expert Opin. Biol. Ther. 14, 1049–1053 (2014). 4. Desnoyers, L. R. et al. Tumor-specific activation of an EGFR-targeting probody enhances therapeutic index. Sci. Transl. Med. 5, 207ra144 (2013). 5. Donaldson, J., Kari, C., Fragoso, R., Rodeck, U. & Williams, J. C. Design and development of masked therapeutic antibodies to limit off-target effects: application to anti-EGFR antibodies. Cancer Biol. Ther. 8, 2147–2152 (2009). 6. Thomas, J. & Daugherty, P. Proligands with protease-regulated binding activity identified from cell-displayed prodomain libraries. Protein Sci. 18, 2052–2059 (2009). 7. Erster, O. et al. Site-specific targeting of antibody activity in vivo mediated by disease-associated proteases. J. Control. Release 161, 804–812 (2012). 8. Yang, Y. et al. Preclinical studies of a pro-antibody-drug conjugate designed to selectively target EGFR-overexpressing tumors with improved therapeutic efficacy. mAbs 8, 405–413 (2015). 9. Yang, Y. et al. Generation and characterization of a target-selectively activated antibody against epidermal growth factor receptor with enhanced anti-tumor potency. mAbs 7, 440–450 (2015). 10. Sandersjoo, L., Jonsson, A. & Lofblom, J. A new prodrug form of Affibody molecules (pro-Affibody) is selectively activated by cancer-associated proteases. Cell. Mol. Life Sci. 72, 1405–1415 (2015). 11. Chen, I. et al. Selective antibody activation through protease-activated pro-antibodies that mask binding sites with inhibitory domains. Sci. Rep. 7, 1–12 (2017). 12. Burkhard, P., Stetefeld, J. & Strelkov, S. V. Coiled coils: a highly versatile protein folding motif. Trends Cell Biol. 11, 82–88 (2001). 13. Thomas, F., Boyle, A. L., Burton, A. J. & Woolfson, D. N. A set of de novo designed parallel heterodimeric coiled coils with quantified dissociation constants in the micromolar to sub-nanomolar regime. J. Am. Chem. Soc. 135, 5161–5166 (2013). 14. Arndt, K., Pelletier, J., Müller, K., Plückthun, A. & Alber, T. Comparison of in vivo selection and rational design of heterodimeric coiled coils. Structure 10, 1235–1248 (2002). 15. Schmidt, M. M. Engineering antibodies for improved targeting of solid tumors. PhD thesis, Massachusetts Institute of Technology, Ch. 5 (2010); http://hdl.handle.net/1721.1/61239 16. McClain, D., Woods, H. & Oakley, M. Design and characterization of a heterodimeric coiled coil that forms exclusively with an antiparallel relative helix orientation. J. Am. Chem. Soc. 123, 3151–3152 (2001). 17. Plückthun, A. & Pack, P. New protein engineering approaches to multivalent and bispecific antibody fragments. Immunotechnology 3, 83–105 (1997). 18. Gerber, H. P. et al. Potent antitumor activity of the anti-CD19 auristatin antibody drug conjugate hBU12-vcMMAE against rituximab-sensitive and -resistant lymphomas. Blood 113, 4352–4361 (2009). 19. Jiang, T. et al. Tumor imaging by means of proteolytic activation of cell-penetrating peptides. Proc. Natl Acad. Sci. USA 101, 17867–17872 (2004). 20. Turk, B. E., Huang, L. L., Piro, E. T. & Cantley, L. C. Determination of protease cleavage site motifs using mixture-based oriented peptide libraries. Nat. Biotechnol. 19, 661–667 (2001). 21. Shay, G., Lynch, C. C. & Fingleton, B. Moving targets: emerging roles for MMPs in cancer progression and metastasis. Matrix Biol. 44–46, 200–206 (2015). 22. Kessenbrock, K., Plaks, V. & Werb, Z. Matrix metalloproteinases: regulators of the tumor microenvironment. Cell 141, 52–67 (2010). 23. King, K. M. & Younes, A. Rituximab: review and clinical applications focusing on non-Hodgkin’s lymphoma. Expert Rev. Anticancer Ther. 1, 177–186 (2001). 24. Maximiano, S., Magalhães, P., Guerreiro, M. P. & Morgado, M. Trastuzumab in the treatment of breast cancer. BioDrugs 30, 75–86 (2016). 25. Ryan, M. C. et al. Integrin αVα6 is expressed on multiple solid tumors and is a potential therapeutic target for auristatin-based antibody-drug conjugates. Canc. Res. 72, abstr. 4630 (2012). 26. Leo, O., Foo, M., Sachs, D. H., Samelson, L. E. & Bluestone, J. A. Identification of a monoclonal antibody specific for a murine T3 polypeptide. Proc. Natl Acad. Sci. USA 84, 1374–1378 (1987). 27. Lin, J. & Sagert, J. in Innovations for Next-Generation Antibody-Drug Conjugates (ed. Damelin, M.) 281–298 (Springer Int. Pub., 2018). 28. Doronina, S. O. et al. Enhanced activity of monomethylauristatin F through monoclonal antibody delivery: effects of linker technology on efficacy and toxicity. Bioconjug. Chem. 17, 114–124 (2006). 29. Stanková, J., Hoskin, D. W. & Roder, J. C. Murine anti-CD3 monoclonal antibody induces potent cytolytic activity in both T and NK cell populations. Cell. Immunol. 121, 13–29 (1989). 30. Alegre, M. et al. Hypothermia and hypoglycemia induced by anti-CD3 monoclonal antibody in mice: role of tumor necrosis factor. Eur. J. Immunol. 20, 707–710 (1990). 31. Ferran, C. et al. Cytokine-related syndrome following injection of anti-CD3 monoclonal antibody: further evidence for transient in vivo T cell activation. Eur. J. Immunol. 20, 509–515 (1990). 32. Huang, X. et al. Inactivation of the integrin B6 subunit gene reveals a role of epithelial integrins in regulating inflammation in the lungs and skin. J. Cell Biol. 133, 921–928 (1996). 33. Lyon, R. P. et al. Reducing hydrophobicity of homogeneous antibody-drug conjugates improves pharmacokinetics and therapeutic index. Nat. Biotechnol. 33, 733–735 (2015). Acknowledgements We thank our Seattle Genetics colleagues L. Benoit, D. Meyer and J. Mitchell for help with antibody–drug conjugation and C. Yu for help with internalization experiments. Author contributions V.H.T., M.R.L. and P.D.S. participated in writing, reviewing and editing of the manuscript. V.H.T. and M.R.L. participated in the planning, initiation, data generation and analysis of biological experiments. V.H.T., X.Z. and M.R.L. performed cell-based assays. V.H.T. performed zymography experiments. S.W.W. and M.M.D performed biophysical characterization. J.H.C. conducted analysis of pharmacokinetics experiments. R.C.Y., W.Z., J.K.S. and I.J.S. performed in vivo anti-tumor activity and pharmacodynamics experiments. M.C.R., R.P.L. and P.D.S. provided project oversight and review. Competing interests All of the authors were employees and shareholders of Seattle Genetics at the time of these studies. Additional information Supplementary information is available for this paper at https://doi.org/10.1038/ s41587-019-0135-x. Reprints and permissions information is available at www.nature.com/reprints. Correspondence and requests for materials should be addressed to M.R.L. Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. © The Author(s), under exclusive licence to Springer Nature America, Inc. 2019 Nature Biotechnology | VOL 37 | JULY 2019 | 761–765 | www.nature.com/naturebiotechnology 765 Letters Methods All chemicals were purchased from Sigma Aldrich unless otherwise noted. All proteases were purchased from R&D Systems with the exception of human MMP-2 (Sino Biological). Cancer cell lines were obtained from the American Type Culture Collection. Cell culture media and media components were purchased from Gibco unless otherwise stated. Auristatin-based drug-linkers, including maleimidocaproyl-MMAF maleimide (mc-MMAF) and MDpr-glucuronideMMAE-maleimide (gluc-MMAE), were prepared as previously described28,33. Antibody production. Antibodies were expressed via transient transfection of either HEK (Gibco) or Chinese hamster ovary (Gibco) cells and were purified using MabSelect SuRe Protein A resin (GE Healthcare Life Sciences). Further purification using size-exclusion chromatography was performed when necessary, using a Superdex S200 column (GE Healthcare Life Sciences). General procedure for mass spectrometric analysis of antibodies or ADCs. Reverse-phase LC–MS data were acquired on a Waters Xevo GS-S QTOF coupled to a Waters Acquity H-Class UPLC system. Samples were reduced with 10 mM dithiothreitol for 10 min at 37 °C and then separated over an analytical reversedphase column (Agilent Technologies, PLRP-S, 300 Å, 2.1 mm × 50 mm, 3 μm) at 80 °C. Samples were eluted using a linear gradient of 0.01% trifluoroacetic acid (TFA) in acetonitrile from 25% to 65% in 0.05% aqueous TFA over 5 min, followed by isocratic 65% 0.01% TFA in acetonitrile for 0.5 min at a flow rate of 1.0 ml min−1. Mass spectrometry data were acquired in ESI+ mode using a mass range of 500–4,000 m/z and were deconvoluted using MaxEnt1 to determine masses of the resulting antibodies or conjugates. Fluorescent labeling of antibodies. Antibodies were fluorescently labeled using N-hydroxysuccinimide ester-activated Alexa Fluor 488 or Alexa Fluor 647 (Life Technologies) following the vendor protocol. Briefly, antibody (1 molar equivalent) was incubated with fluorophore (4–6 molar equivalents) at pH 8 at room temperature for 1 h. Fluorescent antibody was purified using a NAP-5 column (GE Healthcare Life Sciences) and fluorophore loading was quantified using a UV-Vis spectrophotometer (Agilent). Typical fluorophore loadings were 2–4 fluorophores per antibody. For quenched fluorescent antibodies used in internalization experiments, antibodies containing an engineered cysteine at S239C of the heavy chain were used. To enable preferential conjugation to the S239C site, the entire antibody was first reduced using an excess of Tris(2-carboxyethyl)-phosphine (TCEP, 20 molar equivalents, pH 8, 37 °C, 90 min) in the presence of 1 mM diethylenetriaminepentaacetic acid (DTPA). Excess TCEP and DTPA were removed via buffer exchange using Amicon 30-kDa molecular weight cutoff centrifugal filters (EMD Millipore) into PBS containing 2 mM EDTA. The disulfides were then re-oxidized using dehydroascorbic acid (20 molar equivalents, pH 7.4, room temperature, 45 min, 2 additions). Excess dehydroascorbic acid was removed by buffer exchange in Amicon 30-kDa molecular weight cutoff centrifugal filters. QSY9-maleimide (2.5 molar equivalents, ThermoFisher) was added to the re-oxidized antibody (1 molar equivalent) containing free thiols at S239C. The reaction was incubated for 15 min at room temperature, then purified using a NAP-5 column. Drug loading was determined by LC–MS. To conjugate Alexa Fluor 488 at the interchain disulfides, the QSY9-labeled antibody was again reduced using TCEP (10 molar equivalents, 37 °C, 90 min) in the presence of 1 mM DTPA, then buffer exchanged to remove excess TCEP and DTPA. The fully reduced antibody containing 8 free Cys residues was then incubated with Alexa Fluor 488-maleimide (12 molar equivalents, pH 7.4, Life Technologies). The fully conjugated quenched fluorescent antibody was buffer exchanged and concentrated into PBS as previously described. Preparation of masked ADCs. For mixed 4-load mc-MMAF ADCs, antibody interchain disulfides were partially reduced using TCEP (~2.5 molar equivalents, pH 8, 37 °C, 90 min) in the presence of 1 mM DTPA until a mean of 4 reduced thiols were present. Excess TCEP and DTPA were removed via buffer exchange with Amicon centrifugal filters into PBS containing 2 mM EDTA. Reduced antibodies were conjugated for 15 min at room temperature with a 1.5-fold molar excess per thiol of mc-MMAF from dimethylsulfoxide stock solutions. The conjugation reaction was purified using a NAP-5 column and buffer exchanged into PBS as described above. Drug loading was determined by LC–MS. For 8-loaded MDpr-glucuronide-MMAE (gluc-MMAE) ADCs, conjugates were prepared in a similar fashion, except 12 molar equivalents of TCEP were used for the reduction step to generate an antibody containing 8 free thiols. Purification and quantification of drug loading were performed using the same methods. Masked antibody proteolysis. MMPs were activated via incubation with 1.25 mM 4-aminophenyl mercuric acetate at 37 °C for 1 h (MMP-2 and MMP-7), 2 h (MMP-13) or overnight (MMP-9). For all cleaved antibodies used in binding or cytotoxicity assays, masked antibody (~50 µg) was incubated at 37 °C for 2–16 h with activated human MMP-2 (100 enzyme units, where 1 unit is the amount of MMP-2 that hydrolyzes 1 μg of fluorogenic substrate peptide per minute at 37 °C, pH 7.5). The antibody was then NaTure BiOTeCHnOlOgy purified using MabSelect SuRe Protein A resin (GE Healthcare Life Sciences). Briefly, antibody was bound to the resin at room temperature for at least 2 h with constant mixing. The resin was then washed with PBS (5 column volumes) and antibody was eluted with 50 mM glycine pH 3.0. The eluent was then neutralized to pH 7.4 and buffer exchanged into PBS using dilution and concentration with Amicon 30-kDa molecular weight cutoff filters. Competition binding experiments. To evaluate cell binding of masked antibodies, 2 × 105 Raji cells were mixed with fluorescently labeled parent antibody (2 nM) mixed with serial dilutions of competitor (masked antibody) in staining buffer (PBS, 2% fetal bovine serum (FBS), 0.2% NaN3). Samples were incubated on ice for 1 h and washed twice with ice-cold staining buffer. Labeled cells were examined by flow cytometry on a BD LSRII gated to exclude non-viable cells and analyzed using FloJo software. The IC50 was calculated using GraphPad Prism 6. Saturation binding experiments. Binding of antibody or ADC to cell-surface antigen was assessed by flow cytometry on CD19+ Ramos, CD20+ Raji, HER2+ SKBR3, mCD3+ HT2 or αVβ6+ HPAF-II and HEK cells. Cells (2 × 105) were combined with a serial dilution of indicated antibody in staining buffer (PBS, 2% FBS, 0.2% NaN3) in a total volume of 100 µl. The cells were incubated on ice for 1 h and washed twice with ice-cold staining buffer. Cells were resuspended with anti-human IgG-AF647 (Jackson ImmunoResearch, 200-fold dilution in staining buffer) on ice for 1 h. Cells were washed twice with ice-cold staining buffer and resuspended in staining buffer. Labeled cells were examined by flow cytometry on an Attune NxT cytometer (Life Technologies) gated to exclude non-viable cells and analyzed using FloJo software. The dissociation constant (Kd) was calculated using nonlinear regression analysis in GraphPad Prism 6. Biolayer interferometry. Recombinant human CD19-human Fc fusion protein was produced in Chinese hamster ovary cells and labeled with biotin following the SureLINK chromophoric biotin labeling kit (Sera Care). Briefly, protein (1 molar equivalent) was incubated with biotin-N-hydroxysuccinimide ester (4 molar equivalents) at pH 8 at room temperature for 1 h. Biotinylated protein was purified using a NAP-5 column and biotin loading was measured using a UV-Vis (Agilent). A typical extent of biotinylation was 1–2 biotins per CD19. All kinetic measurements were assessed using the OctetRed384 (Pall Forte Bio) at room temperature. The biotinylated CD19 was bound to SAX streptavidin biosensors (Pall Forte Bio) at 7 µg ml−1 in kinetics buffer (0.1% BSA, 0.02% Tween-20, PBS) for 300 s. After equilibration, hBU12 or CC2B-IPVSLRSG-hBU12 was allowed to associate for 600 s and dissociation was measured for 2,000–6,000 s. Kinetic parameters were determined using the Octet analysis software using a 1:1 global fit binding model. Viability assay. Cell viability assays were performed using CellTiter-Glo (Promega). Ramos and BxPC3 cells (2,000 cells per well) were seeded into 96-well clear-bottom culture plates. ADC dilutions were added to each well (1,000 ng ml−1 to 0 ng ml−1) and the samples were incubated for 96 h at 37 °C. Luminescence was measured using an EnVision Multilabel Plate Reader (Perkin Elmer). The IC50 value was determined in quadruplicate and is defined as the concentration that results in half-maximal growth inhibition over the course of the titration curve. The data were fit using GraphPad Prism 6 and an IC50 was used to compare the change in cytotoxicity of parent ADC against masked ADC. CDC assay. In a 96-well plate (Costar), 100,000 Raji cells were plated in 200 μl RPMI 1640 medium containing 5 µM Sytox Green (Invitrogen), 10% normal human serum (Complement Technology) and titrated rituximab, CC2BIPVSLRSG-rituximab, cleaved CC2B-IPVSLRSG-rituximab or control antibody (hIgG1k). A solution of 1% Triton-100 was used as a positive lysis control. Samples were incubated for 2 h at 37 °C and read with the EnVision Multilabel Plate Reader (Perkin Elmer) using 504/523-nm excitation and emission filters. Each experiment was performed in triplicate. T-cell proliferation assay. T cells isolated from BALB/c mouse spleens were labeled with carboxyfluorescein succinimidyl ester (BD Horizon) by incubating cells in prewarmed PBS and carboxyfluorescein succinimidyl ester (1 µM) at 37 °C for 15 min. The labeling reaction was quenched by addition of FBS (Gibco) and cells were washed and resuspended in murine lymphocyte medium (RPMI 1640, 10% FBS, 10 mM HEPES, 4 mM GlutaMAX, 50 µM beta mercapto-ethanol and 1X penicillin/streptomycin). T-cell stimulation was assessed using a CSFE dilution assay. Either 145-2C11 or CC2B-145-2C11 antibody was coated on a U-bottom clear 96-well plate (50 μl per well, 500 ng ml−1) along with 8 µg of anti-mouse CD28 antibody. The plate was incubated overnight at 37 °C. After 16–24 h, the supernatant was removed through aspiration. In each well, 5 × 106 cells were added in a volume of 200 µl and allowed to incubate at 37 °C. On day 4, the T cells were monitored via flow cytometry using an Attune NxT cytometer (Life Technologies). Data were analyzed using FloJo software. Quantification of cytokines using multiplex immunoassay. (Protocol was adapted from MCYTOMAG-70K manufacturer protocol from Luminex Corp.) Nature Biotechnology | www.nature.com/naturebiotechnology Letters NaTure BiOTeCHnOlOgy Plasma samples obtained from mice dosed with antibodies were analyzed using a mouse cytokine/chemokine magnetic bead panel (EMD Millipore). In a 96-well plate (Costar), 25 μl plasma was combined with antibody-immobilized beads and allowed to incubate with agitation overnight at 4 °C. The plate was washed twice and 25 μl detection antibody was added into each well. The plate was sealed, agitated at room temperature for 1 h and then 25 μl Streptavidin-Phycoerythrin was added to each well. The plate was agitated at room temperature for 30 min and washed twice. The luminal signal was detected using the MAGPIX Luminex plate reader (Luminex Corp.). Quantification of cleaved antibody from xenograft studies. Frozen tissues obtained from mice dosed with antibodies were homogenized into powder on dry ice by physical disruption. The homogenized tissue was then resuspended into T-PER homogenization solution (Thermo Scientific) containing protease inhibitor cocktail using a ratio of tissue homogenate to protease inhibitor based on T-PER product instructions. Human IgG-specific affinity resin, IgSelect (GE Healthcare Life Sciences), was used to capture antibodies from the tissue homogenates by incubation at 4 °C overnight. The antibodies were eluted from the resin using 50 mM glycine buffer, pH 3, and concentrated using an Amicon Ultra 30-kDa spin filter to a volume of ~30 μl. Purified antibodies were separated on a 12% Tris glycine gel (ThermoFisher) under reduced denaturing conditions, then transferred onto nitrocellulose membranes. The masked and cleaved antibody signals were detected using an HRP-conjugated anti-human IgG (H + L) antibody (Jackson ImmunoResearch). The blots were scanned using an Amersham Imager 600 (GE Healthcare Life Sciences) and band intensities were quantified with ImageJ software (NIH). The percentage of cleaved antibody signal was calculated as cleaved signal divided by total signal of cleaved and uncleaved antibodies. Assessment of antibody internalization. Ninety-six-well plates (as described earlier) were coated with poly-l-lysine by adding 150 μl poly-l-lysine solution (450 µl poly-l-lysine with 3 ml water) into each well. Plates were incubated at room temperature for 5 min. The coating solution was removed via aspiration and the plate rinsed with PBS two times. The plates were allowed to dry overnight in a sterile environment. Raji cells were labeled with Hoescht 33342 at 0.1–0.5 µg ml−1 final concentrations at 37 °C for 20 min. The cells were washed twice with RPMI 1640 medium and plated at a density of 100,000 cells per well in a total volume of 200 µl. Quenched hBU12 or CC2B-IPVSLRSG-hBU12 was added to each well to a final concentration of 2 µg ml−1 in a total volume of 300 μl. The plate was analyzed using an IN Cell Analyzer 2200 (GE Healthcare) at 37 °C over the course of 14 h. The data were analyzed using IN Cell Analyzer Workstation v.3.7.3(x64), where the fluorescence intensity of each cell was quantified over the course of the experiment. Assessment of protease activity via in situ zymography. For hBU12 antibodies directly labeled with Alexa Fluor 647, fresh-frozen Ramos tumors were cryosectioned (5 μm) and mounted onto positively charged slides (Leica). The tumor slices were incubated in PBS at room temperature for 5 min and blocked with 2% human serum in PBS at room temperature for 30 min. A hydrophobic square was drawn around the tumor using a hydrophobic pen (ThermoFisher). The tumor slice was incubated with antibody (1 µg ml−1 in staining buffer: 25 mM Tris, 150 mM NaCl, 0.05% Tween-20, 100 μM ZnCl, 5 mM CaCl2, pH 7.4) at room temperature for 2 h. Antibody was aspirated off and slides were washed three times with TBS-T buffer (25 mM Tris, 150 mM NaCl, 0.05% Tween-20, pH 7.4) with gentle agitation for 5 min. Slides were washed one time with PBS and mounted with ProLong Diamond Antifade Mountant (ThermoFisher). For protease inhibitor control experiments, slides were pretreated with protease inhibitor cocktail III (1:100, EMD Millipore) and 50 µM Galardin (EMD Millipore) before addition of labeled antibodies. For unlabeled h15H3 antibody staining, fresh-frozen HPAF-II tumors were cryosectioned (5 μm) and mounted onto positively charged slides (Leica). The tumor slices were incubated with PBS as described above and blocked with 5% goat Nature Biotechnology | www.nature.com/naturebiotechnology serum in PBS at room temperature for 30 min. A hydrophobic square was drawn around the tumor and antibody (1 µg ml−1 in staining buffer with 1 mM MnCl2) was added. The slide was incubated with antibody at room temperature for 1 h before the slide was washed as described above. Anti-human AF594 secondary antibody (1:1,000, Invitrogen) was added and allowed to incubate at room temperature for 1 h. Antibody was removed and slides were washed with TBS-T buffer with gentle agitation at room temperature for 5 min. This process was repeated three times and then slides were washed with PBS and mounted as described above. All slides were analyzed using an Olympus IX83 inverted fluorescence microscope with ×10 or ×20 objective and then analyzed using Olympus CellSens Software. In vivo experiments. All animal handling and experimentation were performed under Seattle Genetics Institutional Animal Care and Use Committee guidelines. Seattle Genetics is fully accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International. Pharmacokinetics of masked antibodies. Pharmacokinetics experiments were performed using radiolabeled antibodies. Antibodies (1 mg) were incubated with 55 μCi N-succinimidyl propionate [propionate-2,3-3H] (Moravek Biochemicals, 80 Ci mmol−1, 1 mCi ml−1, 9:1 hexane/ ethyl acetate solution) at room temperature for 2 h at pH 8.0. The mixture was centrifuged at 4,000g for 5 min and the lower aqueous layer was removed. The aqueous layer was buffer exchanged four times using Amicon Ultra-15 Centrifugal Filter Units (Millipore, catalog No. UFC903024, 30 kDa molecular weight cutoff) to remove excess radioactive material. The radiolabeled antibodies were filtered through sterile 0.22-μm Ultrafree-MC Centrifugal Filter Units (Millipore) and the final antibody or ADC concentration was measured spectrophotometrically. The specific activity (μCi per mg) of each product was determined by liquid scintillation counting. The radiolabeled antibodies were injected at 0.5 mg kg−1 in BALB/c mice via tail vein (3 animals per dose group, randomly assigned) or at 1 mg kg−1 in nude mice via tail vein (3 animals per dose group, randomly assigned). The blood was drawn into K2EDTA tubes via the saphenous vein at various time points and processed to plasma. Plasma samples were added to Ecoscint-A liquid scintillation cocktail (National Diagnostics), and the total radioactivity was measured via liquid scintillation counting. The specific activity of the radiolabeled samples was used to calculate the antibody concentration at each time point. In vivo efficacy studies. For therapy experiments, 5 × 106 cells were injected subcutaneously into 5–8 female SCID mice (Harlan) for Ramos study or nude mice (Envigo) for the BxPC3, Detroit 562 and HPAF-II studies. Mice were randomly divided into study groups and dosed with test article via intraperitoneal injection once the tumors reached approximately 100–200 mm3. Animals were euthanized when tumor volumes reached 500–1,000 mm3. Tumor volume was calculated with the formula: volume = ½ × length × width × width. Mice showing durable regressions were terminated around day 40–66 after implant. In all xenograft studies, no weight loss or treatment-related toxicities were observed for mice treated with any of the test articles. Statistics. Statistical significance was analyzed using two-tailed t-test, as indicated in the figure legends. Further information can be found in the Nature Research Reporting Summary. Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article. Data availability The data that support the findings of this study are available within the paper and its Supplementary Information files. Reporting Summary Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Research policies, see Authors & Referees and the Editorial Policy Checklist. Statistical parameters nature research | reporting summary Corresponding author(s): Matthew R. Levengood When statistical analyses are reported, confirm that the following items are present in the relevant location (e.g. figure legend, table legend, main text, or Methods section). n/a Confirmed The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement An indication of whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one- or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section. A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistics including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted Give P values as exact values whenever suitable. For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated Clearly defined error bars State explicitly what error bars represent (e.g. SD, SE, CI) Our web collection on statistics for biologists may be useful. Software and code Policy information about availability of computer code Data collection Mass spectrometry data was acquired using UNIFI software. Flow cytometry data was acquired using Attune NxT Software (version v2.7.0). Octet data acquisition software (version 11.0.0.64) was used to collect biolayer interferometry data. Cytotoxicity data were acquired using Envision Workstation software (version 1.13.3009.1401) Multiplex immunoassay data was collected using Luminex Xponent Software Solutions (version 4.2). Western Blot data was collected using Amersham 600 software. Microscopy data was collected using Olympus CellSense Software (version 1.8.1). Data analysis Mass spectromentry data was analyzed using UNIFI software. Flow cytometry data was analyzed using FloJo software (version vx0.7). Octet data was analyzed using Octet Data Analysis software (11.0.0.4). Cytotoxicity assays and multiplex immunoassay data were analyzed using Graph Pad Prism 7. Western Blot data was analyzed using Image J (version 1.48v). Microscopy data was analyzed using Olympus CellSense Software (version 1.8.1). April 2018 For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors/reviewers upon request. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information. 1 Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: - Accession codes, unique identifiers, or web links for publicly available datasets - A list of figures that have associated raw data - A description of any restrictions on data availability The data that support the findings of this study are available within the paper and its supplementary information files. Field-specific reporting Please select the best fit for your research. If you are not sure, read the appropriate sections before making your selection. Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences nature research | reporting summary Data For a reference copy of the document with all sections, see nature.com/authors/policies/ReportingSummary-flat.pdf Life sciences study design All studies must disclose on these points even when the disclosure is negative. Sample size No sample-size calculations were performed. For in vivo experiments, in an effort to minimize animal use the minimal number of animals required to detect differences between test articles was used. Based upon historical experience 6 mice per group for PK studies (yielding 3 mice per timepoint) and 5 mice per group for tumor studies are the minimum number of animals required. For in in vitro cell-based assays, minimal variability is observed for independent cell samples so an n = 2-4 was used. Data exclusions No data were excluded from the analysis Replication The replication number is indicated in the legend of corresponding figures where applicable. Replication of experiments generated similar results and is indicated in the figure legends. In vivo pharmacokinetic and xenograft experiments were not replicated, but included control test articles that have been tested in each model in at least one additional independent experiment that yielded similar results. All models used have previously been shown to provide robust biological effect across experiments. Randomization In most cases no randomization was required. Tumor-bearing mice were randomized into groups based upon similar tumor size prior to administration of test articles. Blinding No blinding was performed because no studies were deemed to be influenced by human interpretation. Reporting for specific materials, systems and methods Materials & experimental systems Methods n/a Involved in the study n/a Involved in the study Unique biological materials ChIP-seq Antibodies Flow cytometry Eukaryotic cell lines MRI-based neuroimaging Palaeontology Animals and other organisms Human research participants Policy information about availability of materials Obtaining unique materials April 2018 Unique biological materials All unique materials used are available from the authors of the study. 2 Antibodies used The preparation of all therapeutic and surrogate antibodies is described in the Materials and Methods. Saturation binding experiments with human antibodies used secondary antibodies from Jackson Immunoresearch (109-135-098, Allophycocyanin Goat Anti-human IgG Fcg, 109-096-008, FITC F(ab')2 Fragment Goat Anti-Human IgG, or 109-116-170, R-Phycoerythrin F(Ab')2 Fragment Goat Anti-human IgG Fcg at a 1:200 dilution). For saturation binding studies involving mouse antibodies, an RPhycoerythrin F(ab')2 Fragment Goat Anti-mouse IgG Fcg antibody was used (Jackson Immunoresearch, 115-116-071, 1:200 dilution). For the T cell activation experiment, anti-mouse CD3e (eBioscience, 16-0031-85, lot# E06294-1646, 1 mg/mL) and anti-mouse CD28 (Tonbo Biosciences, clone 37.51, 40-0281-U500, 2 mg/mL) was used. For the in situ zymography experiments, a Goat Anti-human IgG (H+L), Alexa Fluor 594 conjugated antibody was used (Life Technologies, A11014 at a 1:1000 dilution). for Western blot experiments, Peroxidase F(ab')2 Fragment Goat Anti-human IgG (H+L) (Jackson Immunoresearch 109-036-088, 1:10,000 dilution ) Validation All primary aAll primary antibodies tested in the manuscript have been published or described previously and are referenced accordingly. All commercially available antibodies were validated by the vendor. Links for each antibody used are: Allophycocyanin Goat Anti-human IgG Fcg : https://www.jacksonimmuno.com/catalog/products/109-135-098 FITC F(ab')2 Fragment Goat Anti-Human IgG: https://www.jacksonimmuno.com/catalog/products/109-096-008 R-Phycoerythrin F(Ab')2 Fragment Goat Anti-human IgG Fcg: https://www.jacksonimmuno.com/catalog/products/109-116-170 R-Phycoerythrin F(ab')2 Fragment Goat Anti-mouse IgG Fcg: https://www.jacksonimmuno.com/catalog/products/115-116-071 anti-mouse CD3e: https://www.thermofisher.com/antibody/product/CD3e-Antibody-clone-145-2C11-Monoclonal/16-0031-85 anti-mouse CD28: https://www.tonbobio.com/antibodies-and-reagents/flow-cytometry-reagents/in-vivo-readytm-mousecd28-37-51.html Goat Anti-human IgG (H+L), Alexa Fluor 594: https://www.thermofisher.com/antibody/product/Goat-anti-Human-IgG-H-L-CrossAdsorbed-Secondary-Antibody-Polyclonal/A-11014 All primary antibodies tested in the manuscript have been published or described previously and are referenced accordingly. All commercially available antibodies were validated by the vendor. Links for each antibody used are: Allophycocyanin Goat Anti-human IgG Fcg : https://www.jacksonimmuno.com/catalog/products/109-135-098 FITC F(ab')2 Fragment Goat Anti-Human IgG: https://www.jacksonimmuno.com/catalog/products/109-096-008 R-Phycoerythrin F(Ab')2 Fragment Goat Anti-human IgG Fcg: https://www.jacksonimmuno.com/catalog/products/109-116-170 R-Phycoerythrin F(ab')2 Fragment Goat Anti-mouse IgG Fcg: https://www.jacksonimmuno.com/catalog/products/115-116-071 anti-mouse CD3e: https://www.thermofisher.com/antibody/product/CD3e-Antibody-clone-145-2C11-Monoclonal/16-0031-85 anti-mouse CD28: https://www.tonbobio.com/antibodies-and-reagents/flow-cytometry-reagents/in-vivo-readytm-mousecd28-37-51.html Goat Anti-human IgG (H+L), Alexa Fluor 594: https://www.thermofisher.com/antibody/product/Goat-anti-Human-IgG-H-L-CrossAdsorbed-Secondary-Antibody-Polyclonal/A-11014 Peroxidase F(ab')2 Fragment Goat Anti-human IgG (H+L): https://www.jacksonimmuno.com/catalog/ products/109-036-088accordingly. nature research | reporting summary Antibodies Eukaryotic cell lines Policy information about cell lines Cell line source(s) Raji (ATCC) SKBR-3 (ATCC) HT-2 (ATCC) Ramos (ATCC) HPAF-II (ATCC) Detroit 562 (ATCC) BxPC3 (ATCC) Authentication All cell lines were authenticated using hort tandem repeat (STR) profiling and interspecies contamination testing Mycoplasma contamination All cell lines tested negative for mycoplasma. Commonly misidentified lines No commonly misidentified cell lines were used in these studies. (See ICLAC register) Animals and other organisms Policy information about studies involving animals; ARRIVE guidelines recommended for reporting animal research Female BALB/c, nude, and SCID mice were purchased from Harlan laboratories. Wild animals The study did not use wild animals. Field-collected samples The study did not use field-collected samples. April 2018 Laboratory animals 3 Plots Confirm that: The axis labels state the marker and fluorochrome used (e.g. CD4-FITC). The axis scales are clearly visible. Include numbers along axes only for bottom left plot of group (a 'group' is an analysis of identical markers). All plots are contour plots with outliers or pseudocolor plots. A numerical value for number of cells or percentage (with statistics) is provided. Methodology Sample preparation All experiments were performed using single-color flow cytometry with authenticated human cancer cell lines or transfected cells Instrument Life Technologies Attune NxT Software Data collected using Attune NxT software and processed using FloJo (FloJo, Inc.) Cell population abundance N/A Gating strategy Nonviable cells were excluded based upon FSC/SSC gating nature research | reporting summary Flow Cytometry Tick this box to confirm that a figure exemplifying the gating strategy is provided in the Supplementary Information. April 2018 4