Humanization of a strategic CD3 epitope enables evaluation of clinical T-cell engagers in a fully immunocompetent in vivo model

T-cell engagers (TCEs) are a growing class of biotherapeutics being investigated in the clinic for treatment of a variety of hematological and solid tumor indications. However, preclinical evaluation of TCEs in vivo has been mostly limited to xenograft tumor models in human T-cell reconstituted immunodeficient mice, which have a number of limitations. To explore the efficacy of human TCEs in fully immunocompetent hosts, we developed a knock-in mouse model (hCD3E-epi) in which a 5-residue N-terminal fragment of murine CD3-epsilon was replaced with an 11-residue stretch from the human sequence that encodes for a common epitope recognized by anti-human CD3E antibodies in the clinic. T cells from hCD3E-epi mice underwent normal thymic development and could be efficiently activated upon crosslinking of the T-cell receptor with anti-human CD3E antibodies in vitro. Furthermore, a TCE targeting human CD3E and murine CD20 induced robust T-cell redirected killing of murine CD20-positive B cells in ex vivo hCD3E-epi splenocyte cultures, and also depleted nearly 100% of peripheral B cells for up to 7 days following in vivo administration. These results highlight the utility of this novel mouse model for exploring the efficacy of human TCEs in vivo, and suggest a useful tool for evaluating TCEs in combination with immuno-oncology/non-immuno-oncology agents against heme and solid tumor targets in hosts with a fully intact immune system.

Historically, TCEs have often been evaluated using xenograft models in immunodeficient mice. In this model, the mice are transfused with donor-derived human immune cells, implanted with human tumor cells, and treated with TCE therapeutic molecules. While having the advantage of testing fully human antibodies, the reconstituted immune system in this model does not resemble an intact mouse, is known to develop alloreactivity to implanted tumors, results in donor-to-donor variations, and restricts evaluation of off-target toxicities associated with treatment.
More recently, efforts have focused on evaluating TCEs in syngeneic models. In particular, murine bispecific technologies have been developed to allow surrogate TCEs to be evaluated in immunocompetent wildtype mice 7,8 . While this addresses many of the limitations associated with xenograft models, the CD3 arm used for the molecules targets murine CD3 and may not accurately recapitulate the activity of anti-human CD3 bispecifics.
Finally, to evaluate the in vivo efficacy of anti-human CD3 TCEs in an immunocompetent mouse, transgenic models expressing human CD3 components have been developed. However, in these models, complete replacement of the entire CD3 complex or replacement with the CD3E chain alone result in deficiencies in T-cell frequencies or function 9,10 . More recent developments have focused on human/murine chimeras, replacing targeted exons within the murine Cd3e locus 11,12 . While these models are more promising, the chimeras introduce mutations in murine CD3E at residues important for interactions with CD3G and CD3D.
Thus, we sought to develop a preclinical syngeneic model that was both generalizable to a broad class of TCEs and minimalistic in modifications to the murine TCR/CD3 complex. Here, we generated a transgenic mouse where we introduced a targeted replacement of an N-terminal DAENI motif in murine CD3e with a GNEEMGGITQT motif from hCD3E. We identified this region on hCD3E as the target of many anti-human CD3 arms from TCEs currently in the clinic. Importantly, this modest replacement in the design of the hCD3E epitope (hCD3E-epi) knock-in mouse does not impact interactions within the TCR complex. Consequently, we do not observe deficiencies in T-cell frequency or function that have been noted with other models. Further, we showed the utility of the model by demonstrating in vivo efficacy with an anti-human CD3/CD20 TCE in a B-cell depletion study. www.nature.com/scientificreports/

Results
Structural rationale for hCD3E-epi knock-in mouse design. To aid in the development of a syngeneic model that would enable the evaluation of a broad class of TCEs, we characterized the binding properties of a panel of anti-CD3 arms for TCEs currently in the clinic. This panel of anti-human CD3 antibodies includes not only some of the common cynomolgus monkey (cyno) cross-reactive SP34-derived clones, but also unique clones as well (Table S1) 13,14 .
First, we used surface plasmon resonance (SPR) to confirm the binding properties of our panel of anti-CD3 Fabs to purified human, cyno, and murine Fc-tagged CD3E/D heterodimers (Table 1 and Fig. S2). We verified that the anti-human CD3 Fabs in our panel, hCD3.1-hCD3.5 as well as SP34, bound potently to cyno and hCD3E/D, but not to mCD3E/D. Additionally, we used the murine surrogate anti-CD3 clone, 2C11, to confirm binding to mCD3E/D, but not cyno or human CD3E/D (Table 1 and Fig. S3) 15 . Thus, this in vitro characterization supported the inability of selected clinical anti-human CD3 variants to bind to mCD3E.
To inform the humanization strategy of mCD3E, we then determined the epitopes for our panel of anti-CD3 binding arms using hydrogen/deuterium exchange mass spectrometry (HDX-MS) (Fig. 1b). HDX reveals changes in solvent accessibility and/or hydrogen bonding induced by complex formation on protein backbone amides. Enzymatic digestion of the Fc-tagged hCD3E/D heterodimeric protein alone with pepsin under conditions used in the HDX-MS workflow provided 98.1% and 97.6% sequence coverage of the E-and D-subunits, respectively, giving confidence to complete epitope characterization using this approach (Fig. S4). We next performed HDX on the complexes between our panel of anti-human CD3 containing bispecifics and the hCD3E/D heterodimer. The peptide-level deuterium uptakes for these complexes were evaluated relative to hCD3E/D alone to reveal regions of the protein that were protected (Fig. S5). The epitopes for all the anti-human CD3 containing bispecifics in our panel mapped to the N-terminus (Fig. 1b).
To validate our engineering strategy of incorporating the anti-human CD3 epitope into mCD3E, we purified a recombinant, chimeric human/murine CD3E protein where we replaced a 5-residue N-terminal stretch in mCD3E with a 11-residue stretch from hCD3E and expressed this as a h/mCD3E-mCD3D Fc-tagged heterodimer. We demonstrated similar binding affinities of our anti-human CD3 Fabs to this chimera relative to hCD3E by SPR (Table 1). These comparable affinities indicate that this modest change in mCD3E was sufficient to accommodate binding of the anti-human CD3 variants. Notably, 2C11, which has an epitope on mCD3E that includes both the N-terminus as well as residues adjacent to the N-terminus, still bound to this chimera, but with reduced affinity (Table 1).
We next sought to understand if there were additional surface accessible epitopes on hCD3E that could be targeted by novel anti-human CD3 antibodies not represented in our panel, or if this minimal replacement of the N-terminus in mCD3E would be sufficient to study a broad class of TCEs. We focused on residues conserved between human and cyno CD3E to enable rapid clinical development. In evaluating the structure of the hCD3E subunit alone, there appear to be multiple regions on the surface of the protein that would be tractable for binding a human/cyno cross-reactive anti-CD3 antibody (Fig. 1c). However, a recent report on the cryo-EM structure of the TCR/CD3 complex reveals that most of these regions on CD3E become buried in the complex and are not accessible to an antibody 5 (Fig. 1d).
In the context of the TCR/CD3E/D complex, only three regions appear to be human/ cyno cross reactive and surface exposed: (1) a membrane proximal patch, (2) a region adjacent to the N-terminus, and (3) the N-terminus (Fig. 1d). The first human/ cyno cross-reactive region closest to the membrane would likely be inaccessible to anti-CD3 antibodies. The second region proximal to the N-terminus is a relatively small area that would be difficult to target with a high affinity anti-CD3E binder. Additionally, this second area is adjacent to an insertion in the hCD3E protein that makes the local surface different from the cyno CD3E protein and thus, more challenging to identify antibodies that are cross-reactive that target this epitope ( Fig. 1d and Fig. S3). This leaves the N-terminus as the most tractable epitope for targeting by anti-CD3E antibodies that are human/cyno cross-reactive.
Generation of immunocompetent mice with a humanized CD3E strategic epitope. Following identification of a high value epitope for CD3 biologics, we designed a strategy to humanize the murine Cd3e locus in an immunocompetent background. The corresponding murine epitope was modified through a germline replacement of exon 4 and exon 5 by homologous recombination in C57BL/6 (B6) mouse embryonic stem (ES) cells with a targeting construct engineered to introduce the human germline sequence. This enabled www.nature.com/scientificreports/ creation of the strategic humanization by replacing the DAENI motif with a GNEEMGGITQT motif (Fig. 2a). Bioinformatic evaluation of ENCODE data for Cd3e and CD3E germline sequence did not identify any predicted regulatory motifs that would be removed or transferred through the germline targeting event (data not shown) indicating the targeting strategy should be low risk to alter physiological expression or expression pattern in the humanized mice. Following germline transmission and breeding to B6 mice the humanized Cd3e mice were born at normal Mendelian ratios and then bred to homozygosity (hCD3E-epi). The hCD3E-epi mice were first evaluated for genomic integrity following humanization. Splenocytes from the hCD3E-epi mice were sequenced by Targeted Locus Amplification (TLA) to enable complete gene sequencing following the humanization 16 . Data from TLA confirmed the correct targeting event and established that no unexpected genomic alteration occurred in or around the Cd3e locus itself or within neighboring genes, including genetically linked Cd3d and Cd3g genes (Fig. S1A).
Following genomic DNA confirmation, the hCD3E-epi mice were evaluated for transcript functionality. Endpoint RT-PCR assays were set up across the humanization region for both wild-type (WT) and hCD3E-epi mice. Data resulted in only a single, expected band for both genotypes (Fig. 2b). Amplicons were then sequenced and resulted in the designed humanized sequence across the exon boundaries (Fig. S1B). The results confirmed the strategic humanization generated the expected transcript and did not result in cryptic or impaired splicing. To determine if the humanization strategy altered expression levels or patterns, we then isolated T cells from the thymus of 7-week-old mice. T cells were isolated in similar quantities (data not shown) and quantitative PCR was performed with WT and hCD3E-epi specific assays as well as a common assay for comparison between the hCD3E-epi and WT strains. The assays distinguished between the different strains, and the results from the common assay confirmed similar relative Cd3e expression in WT and hCD3E-epi mice ( Fig. 2c; unpaired t-test, p = 0.43).
Previously generated human CD3 transgenic models have resulted in altered T-cell function and a range of phenotypes, including thymic dysplasia 9,10 . hCD3E-epi mice were designed to maintain native CD3-complex and function to support T-cell activation and function. Phenotypic evaluation of the hCD3E-epi mice began with an initial evaluation of relative thymus weight in 7-week-old mice to gain insight into T-cell function and   www.nature.com/scientificreports/ and undergoing positive selection (Fig. 4c,f) were unaffected by humanization of the Cd3e locus. Similarly, no differences were observed in the frequency of CD4 and CD8 T cells in the spleen, indicating that T cell development and thymic emigration were unaffected by Cd3e humanization (Fig. 4g,h). Consistent with SPR results, anti-human CD3 (hCD3.1-KLH) bound to the humanized CD3E on the surface of hCD3E-epi T cells and human peripheral blood T cells with similar affinity, but did not bind to WT C57BL/6T cells as expected (Fig. 5a). To ensure normal functionality of T cells from hCD3E-epi mice, T cell activation with a plate-bound anti-human CD3 antibody (hCD3.1-KLH) was monitored after 72 h of stimulation. Similar dosedependent upregulation of T cell activation markers (CD69 and CD25) was observed on CD4 (not shown) and CD8 T cells from hCD3E-epi splenocytes and human PBMC's, but not WT mice stimulated with plate-bound anti-human CD3 (Fig. 5b,c), indicating that anti-human CD3 antibodies are able to effectively bind and crosslink the T cell receptor to activate hCD3E-epi T cells in vitro.
To explore the utility of this mouse model for preclinical assessment of human CD3E-targeted biologic therapies, we tested the ability of hCD3E-epi T cells to mediate cytotoxic killing of target cells following treatment with a T cell-engaging CD3-bispecific antibody (TCE). Incubation of WT and hCD3E-epi total splenocytes with a bispecific T cell engager targeting human CD3 and murine CD20 (hCD3.1-mCD20), resulted in efficient T cell-redirected killing of B cells in hCD3E-epi but not WT spleens after 48 h of in vitro treatment (Fig. 5d). In contrast a CD3-bispecific targeting murine CD3 and murine CD20 (m2C11-mCD20) was able to deplete B cells in both hCD3E-epi and WT splenocyte cultures (Fig. 5d). T cell activation kinetics (CD69, CD25 on CD8-pos T cell) also mirrored B cell killing in this in vitro setup (Fig. 5e,f). These results indicate that the humanized CD3E epitope is distinct from the epitope recognized by a widely used commercial anti-mouse CD3E antibody, and furthermore, suggests that humanization has minimal effects on the native murine CD3 complex.
To investigate in vivo relevance of this mouse model, we treated WT and hCD3E-epi mice intravenously (IV) with the same TCE's used in the in vitro killing assay and monitored peripheral B cell depletion over 7 days. A single IV dose of hCD3.1-mCD20 was sufficient to deplete nearly 100% of CD19-positive B cells from the www.nature.com/scientificreports/ peripheral blood (day 1-7 post treatment) (Fig. 6a,b) and spleen (day 7 post treatment) (Fig. 6c) of hCD3E-epi mice, but not WT mice, as expected. B cell depletion in hCD3E-epi mice was also accompanied by rapid activation of peripheral T cells, as measured by upregulation of surface CD69 and CD25 on blood CD8 T cells 24 h post-treatment (Fig. 6d,e). Both B cell depletion and T cell activation in hCD3E-epi mice required dual engagement of CD3 and CD20 since control antibodies targeting CD3 and an irrelevant targeting arm (KLH) had no obvious effect on peripheral cellularity or immune activation state (Fig. 6a,c,d,e). Importantly, similar levels of B cell depletion and peripheral T cell activation were observed following treatment of both WT and hCD3E-epi mice with m2C11-mCD20, as well as with an anti-mouse monoclonal CD20 depleting antibody (mCD20-IgG2a) (Fig. 6a-e). Serum exposure analysis of the treatment antibodies demonstrated clearance rates consistent with native mouse IgGs (Fig. 6f). We did observe a modest reduction in peripheral exposure of m2C11-mCD20 in WT mice compared to hCD3E-epi mice (Fig. 6f black filled circles vs. purple filled triangles), which likely reflects the increased affinity of m2C11 for the native murine CD3E over chimeric hCD3E-epi. These differences in exposure did not impact the treatment response, as we saw similar levels of peripheral B cell depletion and T cell activation in WT and hCD3E-epi mice treated with m2C11-CD20 TCEs (Fig. 6a-e). These results indicate that hCD3E-epi mice are suitable for studying both murine and human CD3 targeted therapies, and also retain normal antibody-dependent-cellular-cytotoxicity function.

Discussion
In this study, we describe the design and generation of a hCD3E epitope knock-in mouse model. To inform on a strategy to humanize murine CD3E, we first characterized the binding epitope for a small panel of anti-human CD3E arms from TCEs currently in the clinic. The epitope for all of these antibodies mapped to an N-terminal region of hCD3E that is conserved in cyno CD3. With this knowledge, we engineered minimal changes in the murine Cd3e locus with replacement of just a 5-residue motif in mCD3E with an 11-residue motif from hCD3E. These transgenic mice displayed normal T cell development, function, and frequencies. Finally, we were able to demonstrate the utility of this model in a preclinical setting through in vitro T cell killing and activation assays, as well as through B cell depletion in mice treated with a hCD3/CD20 TCE. In conclusion, we show that making this www.nature.com/scientificreports/ targeted, minimal replacement in mCD3E resulted in a syngeneic model where T cell function is not impaired, which enables studying TCEs that have anti-human CD3E arms. Of note, our panel of anti-human CD3 clinical antibodies did all map to the N-terminus of human CD3E, likely influenced by a common desire to obtain a human/ cyno cross-reactive antibody that would be surface accessible in the context of the TCR/CD3 complex. However, other strategic epitopes on CD3 have been reported. For example, OKT3 does map to a region adjacent to the N-terminus of hCD3E (Fig. S2), but consistent with our structural analysis, OKT3 is reported to have a weak affinity for hCD3E 18 . Further, others have reported targeting distinct sites on the TCR/CD3 complex 19 . These studies suggest that strategic selection of alternative epitopes may influence signaling cascades that affect T cell activation and ultimately TCE activity 19 . To characterize molecules binding hCD3E epitopes not targeting the N-terminus in a preclinical setting, additional strategies Figure 6. In vivo B cell depletion and T cell activation following intravenous CD3-CD20 bispecific antibody treatment. (a,b) CD19 + B cell frequencies, as a percentage of total CD45 + leukocytes, were measured in the peripheral blood of hCD3E-epi (a) and WT C57BL/6 mice on day 0 (pre-dose), 1, 3 and 7, and in the spleen on day 7 (c), following intravenous treatment with1 mg/kg of the indicated CD3-bispecific antibodies, or 10 mg/ kg of mCD20-mG2a. Peripheral T cell activation was measured as a percentage of total blood CD8 + T cells that upregulated CD69 (d) or CD25 (e) in hCD3E-epi and WT C57BL/6 mice on day 1 post treatment with the indicated antibodies. (f) Peripheral exposure of indicated the CD3-bispecific antibodies was measured in the serum hCD3E-epi and WT C57BL/6 mice at 1, 24, 72, and 168 h following IV treatment. Each point represents an individual mouse (n = 4/group). Error bars represent ± SEM. P values were calculated using unpaired student T-test. NS (not significant p > 0.05), *p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001, ****p ≤ 0.0001. Data are representative of 2 independent experiments. www.nature.com/scientificreports/ for model development would need to be considered. However, for many TCEs, our model provides a way to assess T cell activity and therapeutic effects with native TCR characteristics. TCEs represent a growing class of bispecific molecules that are showing promising results in cancer immunotherapy. Thus far, two TCEs have been clinically approved-catumaxomab, an EPCAM/CD3 bispecific, for malignant ascites and blinatumomab, a CD19/CD3 bispecific, for acute lymphoblastic leukemia (ALL) 20 , which has sparked efforts to advance additional TCEs targeting both hematologic and solid tumor indications. In the clinic, TCEs have been shown to be particularly effective in treating hematologic tumors but have been met with significant challenges in the solid tumor setting, including limited efficacy and off-tumor on-target toxicities.
Despite signs in the preclinical setting that TCEs demonstrated promising efficacy and reasonable therapeutic windows against solid tumors, there is limited clinical success. In part, this dichotomy could be due to the restricted models available to evaluate TCEs in vivo. Typically, TCEs have been evaluated with xenograft models in immunodeficient mice. Unfortunately, these models do not recapitulate the immune system of an intact mouse and restrict the ability to evaluate on-target toxicities. While utilizing surrogate TCEs in syngeneic models does enable the assessment of molecules in a more relevant, immunocompetentt setting, there is the possibility that results using the mouse bispecifics might not translate to the human setting. More recent efforts to generate transgenic mouse models with a humanized CD3 offer the most promise to evaluate TCEs, to define mechanisms that limit activity, and to identify approaches to avoid normal tissue toxicities. However, models generated to date are typically confounded by deficiencies in T cell frequencies or function 9,10 . The hCD3E-epi knock-in mouse enables characterization of both efficacy and toxicity for TCEs that have an anti-human CD3 binding arm and a murine/human cross-reactive tumor antigen arm 21 .
This novel hCD3E-epi knock-in model offers the opportunity to further evaluate and characterize TCEs in a solid tumor setting leveraging syngeneic tumor cell lines 21 . Some of the challenges associated with the discovery of effective and safe therapies in solid tumor indications include heterogenous expression of the tumor antigen, an immune suppressive tumor microenvironment, and high expression of solid tumor targets in normal tissue leading to off-tumor, on-target toxicities 22 . In this context, TCEs will need to be evaluated in combination with other immunotherapy agents to achieve potent activity 12,23 . In the hCD3E-epi mouse model, we are able to evaluate both single agent activity and combination activities to further understand the role of each of these factors, as well as ways to generate better molecules in this setting.

Methods
Human CD3E knock-in mice. A Cd3e targeting vector was constructed from genomic C57BL/6N mouse DNA (genOway). Targeting was designed to replace mouse DAENI amino acid sequence with its human counterpart GNEEMGGITQT, spanning exons 4 and 5. A loxP-flanked neomycin cassette was inserted in mouse intron 4. Linearized targeting vector was electroporated into C57BL/6N embryonic stem cells. G-418 resistant ES cell clones carrying the recombined locus were identified by PCR while the integrity of the human epitope sequence was confirmed by sequencing.
Recombined ES cell clones were microinjected into C57BL/6N blastocysts, giving rise to male chimeras. These male mice were crossed with C57BL/6N mice expressing Cre-recombinase to produce humanized Cd3e heterozygous mice, without the neomycin cassette. Heterozygous mice were validated by PCR allowing the identification of the wild-type allele and the humanized epitope allele lacking the neomycin cassette. A subset of PCR validated animals was then further characterized by DNA sequencing to fully confirm the integrity of the humanized region (data not shown). Heterozygous humanized animals were interbred to produce homozygous humanized Cd3e mice for study. Mendelian ratios were observed.
Genotyping of hCD3E-epi mice. The hCD3E-epi mice were genotyped using the following primer pair: Fwd: 5' -TGG GTT GTT ATC TAT TCT GTC TGG AAG ACG -3'; Rev: 5' -TTA GCA TAT AGG CAG TAG ACA GGG ACT TGG -3' , which produced a 325 bp band for the wild-type allele and a 398 bp band for the humanized allele. An additional forward primer, specific to the Neo selection cassette, was used in conjunction with the reverse primer listed above, to confirm deletion of the Neo cassette: Neo Fwd: 5'-ACA ACA GCA CCA TTG TCC ACT TGT CC -3' . A 259 bp band would be produced with this primer pair only in the presence of the intact Neo cassette.

RT-PCR for evaluation of hCD3E-epi RNA splicing.
To verify proper hCd3e splicing in hCD3E-epi mice, end-point RT-PCR was performed on total RNA extracted from splenic T cells using the following primers: Fwd: 5'-CTG TCT GCG TCT GGT GCC TTCTT -3'; Rev: 5' -TGG GTC CAC AGA AGG CGA TGTCT -3' . Spleens were extracted from 20-week old WT and hCD3E-epi mice, and single cell suspensions were prepared using gentleMACS Dissociator (Miltenyi Biotec). T cell enrichment was performed using the Pan T Cell Isolation Kit II (Miltenyi Biotec), and total RNA was then extracted using the RNeasy Mini Kit (Qiagen). Quality of the extracted RNA was verified on LabChip (PerkinElmer), and RT-PCR was performed using the SuperScript III One-Step RT-PCR Kit (Invitrogen). Resulting amplicons were analyzed on LabChip (PerkinElmer) or Sanger sequenced and confirmed against the hCD3E-epi reference sequence.
qRT-PCR for evaluation of hCd3e and mouse endogenous Cd3e gene expression. qRT-PCR experiments were performed on total RNA extracted from thymic T cells using fluorescent probes designed specifically to the mutant and WT sequences within the humanization region, and a common sequence shared between the WT and humanized alleles spanning the exon 6 -7 boundary. Murine Actb was used as a reference gene. Thymi were extracted from male and female 7-week old WT and hCD3E-epi mice and dissociated into single cell suspensions using gentleMACS Dissociator (Miltenyi Biotec). T cell enrichment and total RNA extrac- www.nature.com/scientificreports/ tion were performed as described above. Quality of extracted RNA was verified on LabChip (PerkinElmer) and qRT-PCR was performed in triplicates using the iTaq Universal Probes One-Step Kit (Bio-Rad) on the CFX96 Touch Real-Time PCR Detection Systemin (Bio-Rad). Gene expression data were analyzed with CFX Maestro software (Bio-Rad).

Targeted Locus Amplification (TLA) analysis of hCD3E-epi mice.
Bone marrow from 20-week old hCD3E-epi mice was harvested and processed according to the bone marrow isolation protocol provided by Cergentis (Cergentis, The Netherlands). Frozen bone marrow samples were sent to Cergentis for amplification, sequencing, and bioinformatic analysis.
Histological Analysis of hCD3E-epi mice. After mice were sacrificed, thymi from six hCD3e epi and WT littermate mice were collected, fixed in 10% neutral buffered formalin and then embedded in paraffin. 4-µm thick sections were stained with H&E and CD3e. IHC staining was performed using an automated staining system (Leica Bond RX, Leica Biosystems) with a rabbit monoclonal CD3e antibody (clone D4V8L, Cell Signaling Technology, dilution 1:100). The antibody was detected with the Leica Bond Polymer Refine detection kit (Leica Biosystems, cat# DS9800). Positive (mouse spleen) and negative (mouse heart) controls were added in IHC staining. All stained slides were reviewed by an experienced pathologist (X.L.) under light microscope.

Assessment of T cell development in hCD3E-epi mice. Single cell suspension of spleen and thymus
from WT and CD3e KI mice were prepared in RBC lysis buffer (Biolegend) using a gentle MACS tissue dissociator (Miltenyi Biotech) and resuspended at 1 × 10 7 cells/ml in PBS (Corning) containing 2% HI-FBS (VWR), and 1 mM EDTA (Corning). 1 × 10 6 cells were pelleted in 96 well round bottom plates and resuspended in PBS + 2%FBS + 1 mM EDTA (Flow buffer) containing Fc-block (BD), and live/dead fixable dye (Invitrogen) for 20 min. Cells were re-pelleted and resuspended in surface staining cocktail containing antibodies against mouse TCR-beta, CD8, CD4, B220, CD25, CD5, CD44, CD69 at concentrations recommended by manufacturer. Cells were washed 2 times with flow buffer and fixed with cytofix (BD) followed by 2 washes according to manufacturer's instructions. Cells were acquired on an LSRFortessa X-20 (BD), and analyzed using Flowjo V10.6.2. Thymocytes were analyzed from the live B220-negative population, using surface CD8 and CD8 to define SP, DP, and DN thymocytes. Surface CD44 and CD25 on DN thymocytes was used to define the DN1 (CD44-pos CD25neg), DN2 (CD44-pos, CD25-pos), DN3 (CD44-neg, CD25-pos) and DN4 (CD44-neg, CD25-neg) transition stages. Positively selected DP and SP thymocytes were identified by surface CD5 and CD69 upregulation. All populations were quantified as a relative percentage of the parent population. To measure T cell activation, anti-hCD3.1-KLH was coated onto wells of a 96-well flat bottom plate starting at 10 ug/ml in PBS with 6 × fourfold serial dilutions. Plates were incubated at 4 °C overnight and washed 3 times with RPMI (Corning) containing 10% HI-FBS. Splenocytes from WT C57BL/6 and hCD3E-epi mice, and human PBMC's were prepared as above, and 200 ul of cells at 5 × 10 6 cells/ml were added to pre-coated wells and incubated at 37 °C and 5% CO 2 for 72 h. Following stimulation, cells were pelleted into 96-well round bottom plates, incubated with Fc-block and live/dead fixable dye, stained with fluorescently conjugated antibodies against mouse or human CD4, CD8, CD19, CD25, and CD69, and fixed prior to analysis as decribed above. Cells were acquired and analyzed as above, and T cell activation was measured as the percentage of CD8-positive T cells that upregulated surface CD25 or CD69.

Binding of anti-human and anti-mouse CD3 antibodies to hCD3E-epi T cells and T cell activa
In vitro T cell-redirected killing of CD20-positive B cells. RBC-lysed single cell suspensions of splenocytes from WT C57BL/6 and hCD3E-epi mice were prepared as described above and resuspended in RPMI + 10% HI-FBS. 100 µl of total splenocytes at 1 × 10 7 cells/ml were transferred to wells of a 96 well round bottom plate and mixed with a 2 × solution (starting at 20 µg/ml) of fourfold serially diluted CD3-bispecific antibodies (hCD3.1-KLH, hCD3.1-mCD20, m2C11-KLH, m2C11-mCD20). Cells were incubated at 37 °C and 5% CO 2 for 72 h prior to measurement of B cell depletion by flow cytometry. Cells were pelleted and pre-incubated with Fc-block and live-dead fixable dye, followed by staining with fluorescently conjugates antibodies against CD4, CD8, CD19, CD25, and CD69 and cytofix treatment as described above. Cells were acquired with a BD LSRFortessa X-20 in plate-mode, with settings adjusted so the same volume of sample would be acquired for every well. Only live cells were included in the final analysis with FlowJo V10. 6 www.nature.com/scientificreports/ In vivo B cell depletion. WT C57BL/6 (n = 4) and hCD3E-epi mice (n = 4) were pre-bled submentally (30 µl into EDTA tubes) to obtain pre-treatment peripheral B cell frequencies, and then treated the following day intravenously with 1 mg/kg of CD3-bispecific antibodies (hCD3.1-KLH, hCD3.1-mCD20, m2C11-KLH, and m2C11-mCD20) or 10 mg/kg mCD20-mG2a. 30 µl of whole blood was collected by submental method from all mice into EDTA tubes on day 1, 3, and 7 post-treatment. Blood samples were cleared of red blood cells with ACK lysing solution (Biolegend) and stained as described above with fluorescently-conjugated antibodies against mouse CD45, CD19, CD4, CD8, CD25, and CD69. The percentage of CD19-positive B cells remaining in the blood was calculated as a frequency of total live CD45-poitive cells, and T cell activation was measured as the percentage of CD8-positive T cells expressing surface CD25 or CD69. Similar analyses were performed on RBClysed splenocytes from mice on day 7 after treatment to measure B cell depletion in secondary lymphoid organs.
Antibodies for flow cytometry. Protein expression and purification. Human or murine chimeric Fabs were individually cloned into the pTT5 expression plasmids (GenScript) with an N-terminal osteonectin signal peptide and a C-terminal poly-histidine tag on the heavy chain. Supplemental Table S1 lists the VH and VL sequences. The Fabs were transiently expressed in Expi293 cells (ThermoFisher Scientific), and the clarified supernatants were purified using Ni Sepharose excel resin (GE Healthcare Life Sciences) followed by a size exclusion purification step using a Superdex 200 column (GE Healthcare Life Sciences) into 1 × PBS. Human CD3E/D, cyno CD3E/D, murine CD3E/D, and chimeric human/murine CD3E/D were generated as Fc-fused heterodimers, and similarly cloned into pTT5 expression plasmids. Each heterodimer was transiently co-expressed in Expi293 cells. Clarified supernatants were purified using MabSelect Sure LX resin (GE Healthcare), washed with PBS, eluted in 100 mM pH 3.6 sodium citrate buffer, and neutralized with 1 M pH 8.0 Tris buffer. Each CD3 heterodimer was further purified on a Superdex 200 column into 1 × PBS.
HDX-MS approach for epitope characterization. HDX-MS was conducted as described previously 24 .
In brief, non-deuterated experiments were performed to generate a list of common peptides for CD3 prior to epitope mapping experiments. For epitope mapping in CD3, samples including CD3 alone and protein complex of mAb/CD3 were prepared at 1 to 1 molar ratios at 20 µM final protein concentration in PBS at 25 °C for 1 h prior to HDX. HDX was conducted on an HDX PAL robot (LEAP Technologies, Carrboro, NC) where 5 µL of each sample was diluted into 55 µL of D 2 O buffer (10 mM phosphate buffer, D2O, pD 7.0) and the reactions were carried out for different periods of time: 20 s, 1 min, 10 min and 240 min. By the end of each labeling reaction period, the reaction was quenched by adding quenching buffer (100 mM phosphate, 4 M GdmCl, 0.4 M TCEP, pH 2.5, 1:1, v/v) and 50 µL of quenched sample was injected into Waters nanoACQUITY UPLC HDX Manager™.
Proteins were digested online using a Enzymate pepsin column, 300 Å, 5 µm, 2.1 mm × 30 mm (Waters Corp., Milford, MA, USA), at 20 °C and for 3 min. The deuterated peptides were trapped and desalted for 3 min on an ACQUITY UPLC BEH C18 VanGuard Pre-column (130 Å, 1.7 µm, 2.1 mm × 5 mm). Peptides were separated with an ACQUITY UPLC C18 BEH 1.0 × 100.0 mm column (Waters Corp.) with a 10 min gradient of 8-85% acetonitrile/water containing 0.1% formic acid operated under 40 µL/min flow rate. All the chromatographic elements were held at 0.0 ± 0.1 °C. Mass spectra were obtained with a Waters Synapt G2si Q-TOF equipped with standard ESI source (Waters Corp.). The instrument configuration was the following: capillary was 3.5 kV, sampling cone at 35 V, source temperature of 80 °C and desolvation temperature of 175 °C. Mass spectra were acquired over an m/z range of 260 to 2000. www.nature.com/scientificreports/ For HDX-MS data analysis, peptic peptides were identified through a combination of exact mass analysis and MSE using ProteinLynx Global Server 3.0.2 (Waters Corp., Milford, MA, USA). Peptide-level deuterium uptakes were calculated using Waters DynamX ™ 3.0 software. Surface plasmon resonance (SPR) spectroscopic determination of anti-CD3 variants binding parameters. SPR was used to determine binding parameters for the anti-CD3 Fabs to the Fc-tagged human CD3E/D, cyno CD3E/D, murine CD3E/D, and the chimeric human/murine CD3E/D heterodimer variants with a BIACORE® T200 SPR spectrometer (Biacore AB). The CD3 variants were captured on a C1 chip where either anti-human Fc pAb or anti-murine Fc pAb was first amine coupled to the surface. The CD3 variants were then bound by a concentration series of the Fabs in HBS pH 7.4 running buffer supplemented with 0.05% Tween-20, 1 g/L BSA at 37 °C. All data were double-referenced and fitted to a 1:1 Langmuir binding model with mass transport to determine equilibrium dissociation constants (K D ).

Research ethics.
All animal studies that were conducted are in accordance with the ARRIVE guidelines and are in compliance with relevant ethical regulations. Animal studies performed at Bristol-Myers Squibb were carried out in an AAALAC, Intl-accredited facility and approved by the Bristol-Myers Squibb Institutional Care and Animal Use Committee. www.nature.com/scientificreports/