Visual Abstract

Bispecific T cell engagers have demonstrated clinical efficacy; however, their use can be accompanied by severe toxicity. Mechanistic understanding of these toxicities is limited by a lack of suitable immunocompetent preclinical models. In this study, we describe an immunocompetent mouse tumor model that exhibits bispecific T cell engager–induced toxicity and recapitulates key features similar to those in human cytokine release syndrome. In this study, toxicity occurred between the second and fourth injections of an NK Group 2D bispecific T cell engager protein. Symptoms were transient, peaking 3–4 h after treatment and resolving by 8 h. Mice developed weight loss, elevated plasma cytokines, a significant reduction in spleen white pulp, and lymphocyte infiltration in the liver. Systemic cellular immune changes also occurred; notably, an increase in CD8+ T cell activation, an increase in myeloid cells in the blood, and a population of Ly-6Cint monocytes (CD11b+Ly-6GF4/80) emerged in the liver and spleens of bispecific protein–treated mice. IFN-γ was primarily produced by CD8+ T cells in the spleen and was required for the observed changes in both T cell and myeloid populations. Rag deficiency, IFN-γ deficiency, or depletion of either CD4+ or CD8+ T cells prevented toxicity, whereas perforin deficiency, GM-CSF deficiency, or modulation of the myeloid population through clodronate-mediated depletion showed a partial abrogation of toxicity. Together, these findings reveal that T cell activation by a bispecific T cell engager leads to changes in the host myeloid cell population, both of which contribute to treatment induced toxicity in immunocompetent mice.

This article is featured in In This Issue, p.2865

T cells are key players in antitumor immunity (1, 2). One approach to enhance the function of T cells is to redirect them to recognize a tumor Ag using a bispecific Ab. These are called bispecific T cell engagers (bsTCEs). A successful strategy has been to connect two single-chain variable fragments (scFv) with a flexible serine-glycine linker (3). The bsTCE is then able to engage T cells through binding to CD3 with one arm and redirect them to a tumor Ag with the other arm. The bridging of CD3 with the target-expressing cell results in cytolytic synapse formation, perforin- and granzyme-mediated, MHC-independent killing of target Ag-expressing cells, polyclonal T cell activation and proliferation, and cytokine production (4). There are now more than 40 bsTCEs in clinical development, with the majority of these being evaluated for hematologic indications, although several for solid tumor indications are now entering clinical trials (5). The most advanced is the U.S. Food and Drug Administration–approved, CD19 × CD3–targeting bsTCE, blinatumomab, which is approved for the treatment of adults and children with relapsed/refractory or minimal residual disease–positive B cell precursor acute lymphoblastic leukemia (ALL) (6).

However, the clinical success of blinatumomab is accompanied by frequent and sometimes severe cytokine release syndrome (CRS) and neurotoxicity (7, 8). This toxicity profile has similarities to patients treated with chimeric AgR (CAR) T cells, which redirect T cells to tumors through genetic engineering (7). Clinical manifestations of CRS can range from mild flu-like symptoms to more severe sepsis-like symptoms (9). In blinatumomab-treated patients, CRS occurred in ∼15% of relapsed/refractory ALL and 7% in minimal residual disease–positive ALL patients in clinical trials (10). CRS typically occurs within the first few days of treatment and is characterized by high levels of circulating serum cytokines for both bsTCE and CAR T cell–treated patients (1116). The high levels of cytokines are thought to drive a macrophage dysfunction seen in some patients (7). Neurological symptoms may include headache, delirium, seizure, cerebral edema, and intercranial hemorrhage and occur in ∼65% of patients receiving blinatumomab (9, 10, 17), although only ∼13% develop severe neurotoxicity (10).

For blinatumomab, these toxicities are primarily managed by suspending treatment and through administration of dexamethasone (7). An IL-6R–blocking Ab, tocilizumab, has been U.S. Food and Drug Administration approved for treatment of CAR T cell–induced CRS. This approval was based on a retrospective analysis demonstrating that 50–70% of CAR T cell–treated patients developing CRS responded to tocilizumab treatment (18). There have also been 30 cases identified in a retrospective analysis in which tocilizumab was used for the management of blinatumomab-induced CRS (19). Of the 22 cases with reported outcomes, 21 reported resolution of symptoms (19). However, it is difficult to interpret the importance of tocilizumab in resolution of symptoms in these cases, as patients also discontinued blinatumomab treatment and were treated with corticosteroids (19). Together, these data suggest that IL-6 can be an important mediator of CRS; however, the fact that it is not effective in all cases suggests that other mechanisms are at play and that additional methods to treat these symptoms are still needed. A major hindrance is that the underlying mechanisms remain poorly understood, in part because these toxicities are not recapitulated in the immunodeficient xenograft models that are used to study most human T cell–redirecting therapies. An increased understanding of mechanisms driving toxicities may lead to improved strategies to manage such events and improve patient outcomes.

We have developed both human and murine versions of a bsTCE to investigate immune interactions in syngeneic immunocompetent mouse tumor models. These bsTCEs are constructed from the extracellular portion of either mouse or human NK Group 2D (NKG2D) receptor linked to an anti-CD3 scFv and do not contain an Fc (20, 21). NKG2D is an activation receptor found on NK cells and some T cell subsets and normally functions in the recognition of stressed, damaged, or transformed cells (22). NKG2D ligands (NKG2D-Ls) are often expressed on tumor cells but are absent on most normal tissues, making them excellent immunotherapeutic targets (22). NKG2D-Ls are species specific. The human bsTCE, hNKG2D-LxCD3 (hNKG2D-OKT3), recognizes two families of ligands: MHC class I chain–related proteins A (MICA) and B (MICB) as well as UL16-binding proteins (ULBP1–6) (20). This bispecific protein has demonstrated the ability to specifically activate T cells against ligand-positive human tumor cell lines and primary tumor samples (20). The murine bispecific, mNKG2D-LxCD3 (mNKG2D-2C11), recognizes the ligands Rae1, Mult-1, and H60 (21). In vitro, mNKG2D-2C11 engages both T cells and tumor cells, resulting in T cells producing IFN-γ and cytotoxicity against ligand-positive tumor cells (21). mNKG2D-2C11 has also demonstrated in vivo antitumor efficacy (21). In this study, we describe the development of an immunocompetent mouse tumor model that exhibits mNKG2D-2C11–associated toxicity, with features similar to those observed in human CRS, and explore mechanisms underlying the observed symptoms.

Murine colon cancer MC38 cells were obtained from Dr. R. J. Barth (Geisel School of Medicine). Cell lines are tested routinely for mycoplasma by PCR at the time of experiments and retested periodically. After thawing, cells were passaged two to three times before use. MC38 cells were cultured in complete DMEM with a high glucose concentration (SH30022.01; GE HyClone Laboratories) supplemented with 10% heat-inactivated FBS (SH30910.03; GE HyClone Laboratories), 100 U/ml penicillin, 100 μg/ml streptomycin (SV30010; GE HyClone Laboratories), 1 mM sodium pyruvate (25-000-Cl; Corning Cellgro), 10 mM HEPES (25-060-Cl; Corning Cellgro), 0.1 mM MEM nonessential amino acids (25-025-Cl; Corning Cellgro), and 50 μM 2-ME (M6250-100ML; Sigma-Aldrich).

mNKG2D-2C11 and TZ47-2C11 bsTCEs were constructed as described (21, 23). The bsTCEs were produced using the Expi293 Expression System (A14635; Thermo Fisher Scientific). Manufacturer’s instructions were followed for transfection of Expi293 cells. Cultures were maintained at 37°C, 8% CO2, with shaking at 100 rpm for 5 d after transfection. Anti-CD4 and anti-CD8 depletion Abs were purified from hybridoma supernatants of GK1.4 and 2.43 adapted with low serum obtained from Dr. M.J. Turk (Geisel School of Medicine). Hybridomas were grown in Hybridoma-SFM media supplemented with 0.05% ultra-low IgG FBS (12045076, A3381901; Life Technologies). Cell-free supernatant was then harvested and filtered using 0.45-μm filter units and glass 1.0-μm fiber prefilters (SCHVU05RE, AP1507500; MilliporeSigma).

The bsTCEs were purified by nickel chromatography using an AKTA start purification system (29-0220-94; GE Healthcare). Briefly, supernatants were loaded onto a HisTrap HP column (17524801; GE Healthcare) diluted in nickel column binding buffer (300 mM NaCl, 50 mM NaH2PO4, and 20 mM imidazole [pH = 7.4]) and eluted using a stepwise gradient of imidazole with the elution buffer (300 mM NaCl, 50 mM NaH2PO4, and 480 mM imidazole [pH = 7.4]). The anti-CD4 and anti-CD8 Abs were purified using HiTrap Protein A HP columns (29048576; GE Healthcare). The Abs were eluted with 0.1 M glycine (pH 2.2) (BP381-500; Thermo Fisher Scientific). One hundred microliters of neutralization buffer (Tris-HCl [pH 9]) was added to fraction collection tubes prior to elution. Eluted fractions containing desired protein as determined by SDS-PAGE gel (S6650; Thermo Fisher Scientific) were combined and buffer exchanged into PBS (21-040-CV; Corning Cellgro) using HiTrap Desalting columns (29-0486-84; GE Healthcare). Protein concentration was quantified by NanoDrop.

All experiments were conducted according to protocols approved by Dartmouth College’s Institutional Animal Care and Use Committee. C57BL/6 (B6, wild type [WT]) were purchased from Charles River National Cancer Institute (Frederick, MD). Rag1−/− mice (B6 background) were obtained from Dr. Y. Huang (Geisel School of Medicine). IFN-γ reporter with endogenous poly(A) transcript (GREAT) mice (B6.129S4-Ifngtm3.1Lky/J) were obtained from Dr. M.J. Turk (Geisel School of Medicine) (stock 017581; The Jackson Laboratory). Perforin-deficient mice (C57BL/6Prf1tm1Sdz/J) and IFN-γ–deficient mice (B6.129S7-Ifngtm1Ts/J) were obtained from The Jackson Laboratory (stock 002407 and 002287). GM-CSF–deficient mice (C57BL/6 background) were bred in our facility.

MC38 tumor cells (1 × 106) in 0.4 ml HBSS (SH30031.02; GE HyClone Laboratories) were injected s.c. into mice on day 0. When tumors reached ∼40 mm2 (days 10–14), treatment with the indicated bispecific protein was initiated (unless otherwise noted). Dose and interval of treatment are indicated in each experimental setup. bsTCEs were diluted in 0.4 ml HBSS and injected i.v.

The health status of mice was graded on a scale from 1 to 4 by an experimenter blinded to the group treatment status as follows: 1, normal and healthy; 1.5, walking somewhat slowly; 2, moving slowly; 2.5, moving slowly and hunched posture; 3, little movement, hunched posture, and respiratory distress; 3.5, little movement upon touch, hunched posture, and respiratory distress; and 4, death. Mice were also weighed at the start of each experiment, daily during the course of treatment with bsTCEs, and 2–3 times/wk after treatment.

Anti-CD4 (clone GK1.5), anti-CD8α (clone 2.43), anti–murine IL-6R (clone 15A7, BE0047; Bio X Cell), and control rat γ-globulin (012-000-002; Jackson ImmunoResearch Laboratories) were given i.p. at a dose of 200 μg diluted in 0.4 ml HBSS. In T cell depletion experiments, Abs were given 1 d prior to the first injection of bsTCE. Anti–IL-6R was given 3 h prior to the third injection of bsTCE. The 1400W (214358-33-5; Cayman Chemical) was administered i.p. daily for 7 d at a dose of 100 μg diluted in 0.2 ml HBSS starting 5 h before the first injection of bsTCE or before the second and third injections of bsTCE. Clodronate or control liposomes (CP-005-005; LIPOSOMA research liposomes) were given at a dose of 0.2 ml i.v. 1 d prior to the initiation of bsTCE treatment.

Mouse blood was collected via tail bleed or cardiac puncture into 300-μl Microvette 100 EDTA Tubes (101093-992; VWR International) centrifuged at 1300 rcf for 10 min. Plasma was then collected and frozen at −20°C. Samples were then analyzed by multiplex cytokine analysis by DartLab.

Blood, spleen, liver, tumor, and tumor-draining inguinal lymph nodes were removed from mice and processed as follows for flow cytometry. Blood was collected as described above. Plasma was removed and frozen. The remaining cell layer was subjected to two rounds of RBC lysis by incubating for 5 min with 0.5 ml of RBC lysis buffer (0.16 M NH4Cl, 0.01 M KHCO3, and 0.1 mM EDTA). Cells were washed twice with PBS after each round of RBC lysis. The spleen, lymph node, liver, and tumor were manually dissociated. Spleens were washed with HBSS and then incubated for 3 min in 3 ml RBC buffer. After 3 min, 10 ml of HBSS was added to neutralize the RBC lysis buffer, and the suspension was filtered through a 70-μm cell strainer (21008-952; VWR International). Cells were pelleted as before, resuspended in PBS. Lymph nodes were pelleted at 4.6 × 1000 × g for 1 min and resuspended in PBS. Livers and tumors were dissociated in 1.5 mg/ml collagenase and 0.4 mg/ml DNase (10103586001, DN25-100MG; Sigma-Aldrich) in HBSS and then incubated for 1 h at 37°C with shaking. The disaggregated tissue was strained through 70-μm cell strainer and washed with 10 ml HBSS. RBCs were then lysed in these tissues, as described for the spleen. Cells were washed with HBSS and resuspended in PBS. Cells from spleens, lymph nodes, livers, and tumors were counted by hemocytometer, and 5 × 105 cells were stained for each sample. Samples were stained with 1:100 dilution of Zombie UV (423108; BioLegend), followed by blocking with anti-CD16/32 (101302; BioLegend) and 2% mouse serum (015-000-120; Jackson ImmunoResearch Laboratories). Samples were then incubated with the following staining mixture: CD45.2 PE at a 1:50 dilution (560695, clone 104; BD Biosciences), CD3 PE-Cy7 at a 1:200 dilution (560591, clone 17A2; BD Biosciences), CD4 APC-Cy7 at a 1:200 dilution (552051, clone GK1.5; BD Biosciences), CD8 BV711 at a 1:200 dilution (563046, clone 53-6.7; BD Biosciences), CD44 BV605 at a 1:200 dilution (563058, clone IM7; BD Biosciences), CD62L BV510 at a 1:50 dilution (563117, clone MEL-14; BD Biosciences), NK1.1 BV421 at a 1:50 dilution (108741, clone PK136; BioLegend), CD49b FITC at a 1:50 dilution (553857, clone DX5; BD Biosciences), CD11b PE-Cy5 at a 1:50 dilution (101210, clone M1/70; BioLegend), F4/80 BV650 at a 1:200 dilution (123149, clone BM8; BioLegend), Ly-6C APC at a 1:200 dilution (128016, clone HK1.4; BioLegend), Ly-6G BUV395 at a 1:200 dilution (563978, clone 1A8; BD Biosciences), and CD19 BUV737 at a 1:200 dilution (564296, clone 1D3; BD Biosciences). Samples were fixed with methanol-free 0.5% formaldehyde (04018-1; Polysciences) and run on a Yeti/ZE5 (Propel Labs, Fort Collins, CO). Flow cytometry data were analyzed using FlowJo V10. (FlowJo).

Dissected tissues were fixed in 10% formalin (10662-328; VWR International). Paraffin sections were generated and stained with H&E by the Pathology Research Resource at Dartmouth–Hitchcock Medical Center. Histological analysis was performed in an independent and blinded manner by pathologist Dr. Z. Grada (Geisel School of Medicine). Tissues were taken for analysis from each mouse: brain, heart, kidney, liver, lungs, and spleen. Tissues were rated as normal, or pathology was noted.

Paraffin-embedded, 2.5% buffered, formalin-fixed tissue specimens (brain, lung, heart, pancreas, spleen, liver, kidney, inguinal lymph node, and tumor) from adult C57BL/6 mice underwent deparaffinization and heat-induced epitope retrieval using sodium citrate buffer (pH 6) prior to Ab staining with polyclonal goat IgG anti-mouse pan-specific Rae1 primary Abs (AF1136; R&D Systems) in a 1:20 dilution overnight at 4°C in a humidity chamber. Contiguous sections from each sample were incubated with primary Abs preincubated for 4 h with mouse rRae1 Fc chimeric protein (1998-RA; R&D Systems) at a 1:10 M ratio as specificity controls. Blocking reagents, secondary Abs, and diaminobenzidine chromogen were added using a high-sensitivity, streptavidin-conjugated anti-goat Ab kit (CTS008; R&D Systems). Slides were counterstained with methylene blue. B16F10-Rae1 melanoma cells were stained as positive controls. Specimens were also incubated with no primary Abs as negative controls.

To study toxicity of the T cell–redirecting bispecific proteins, MC38 mouse colon carcinoma cells, which endogenously express Rae-1, were injected s.c. into B6 mice. Once tumors reached ∼40 mm2 (days 10–14 posttumor injection), treatment with either mNKG2D-2C11 or a control bsTCE (TZ47-2C11) was initiated. Both bsTCEs are similar in size and have the same anti-CD3 scFv domain through which they can bind murine T cells and lack Fc domains. However, the control bsTCE is specific for human B7H6, which has no expression in mouse tissues, whereas mNKG2D-2C11 is specific for murine NKG2D-Ls (Fig. 1A). Mice were injected i.v. with 10 μg of bsTCE every other day for a total of four doses. Mice were monitored for symptoms of toxicity in a blinded manner and given a health score between 1 and 4, in which a score of 1 was a normal healthy mouse, and higher scores indicate progressively worse symptoms (see 2Materials and Methods for scoring system). Mice experienced decreased activity, hunched posture, piloerection, respiratory distress, and weight loss. Mouse distress was observed between the second and fourth treatments but never occurred after a single injection or when more than four injections were given (Fig. 1B, Supplemental Fig. 1A). The symptoms were transient, peaking 3–4 h after treatment, with complete resolution by 8 h. Symptoms were also dose dependent. When mice were treated with 2-, 5-, or 10-μg doses of mNKG2D-2C11, only mice treated at the 10-μg dose experienced toxicity (Fig. 1C).

FIGURE 1.

Kinetics and dose response of mNKG2D-2C11–induced toxicity. (A) Diagram showing construction of treatment (mNKG2D-2C11) and control (TZ47-2C11) bsTCEs. (BD) MC38 (1 × 106) cells were injected s.c. into WT B6 mice. Treatment was initiated when tumors reached ∼40 mm2. (B) Mice received a total of four 10-μg i.v. injections of bsTCE every other day. (Left) Health scoring was blinded and evaluated 3 and 24 h after each treatment. (Right) Weight, normalized to day 0, at the indicated time points. Data pooled from seven experiments (n = 30). (C) Mice received a total of four i.v. injections of either 2, 5, or 10 μg bsTCE every other day. Health scoring was blinded and evaluated every hour for the first 8 h and then 24 h after each treatment. Data pooled from two experiments (n = 8). (D) Tumor- or nontumor-bearing WT B6 mice received a total of four 10-μg i.v. injections of bsTCE every other day. Health scoring was blinded and evaluated 3 and 24 h after each treatment. Data are pooled from two experiments (n = 8). (E) Representative images of Rae1-positive staining in the liver of B6 mice or contiguous tissue sections blocked with recombinant Rae1. Slides counterstained with methylene blue. Health scores in (B) and (D) are shown ±SD; weight and (C) are shown ±SEM. Statistical significance determined by repeated measures two-way ANOVA with Bonferroni multiple comparisons test (B and C) or with Tukey multiple comparisons test (D). *p < 0.05, ***p < 0.001. NS, not statistically significant.

FIGURE 1.

Kinetics and dose response of mNKG2D-2C11–induced toxicity. (A) Diagram showing construction of treatment (mNKG2D-2C11) and control (TZ47-2C11) bsTCEs. (BD) MC38 (1 × 106) cells were injected s.c. into WT B6 mice. Treatment was initiated when tumors reached ∼40 mm2. (B) Mice received a total of four 10-μg i.v. injections of bsTCE every other day. (Left) Health scoring was blinded and evaluated 3 and 24 h after each treatment. (Right) Weight, normalized to day 0, at the indicated time points. Data pooled from seven experiments (n = 30). (C) Mice received a total of four i.v. injections of either 2, 5, or 10 μg bsTCE every other day. Health scoring was blinded and evaluated every hour for the first 8 h and then 24 h after each treatment. Data pooled from two experiments (n = 8). (D) Tumor- or nontumor-bearing WT B6 mice received a total of four 10-μg i.v. injections of bsTCE every other day. Health scoring was blinded and evaluated 3 and 24 h after each treatment. Data are pooled from two experiments (n = 8). (E) Representative images of Rae1-positive staining in the liver of B6 mice or contiguous tissue sections blocked with recombinant Rae1. Slides counterstained with methylene blue. Health scores in (B) and (D) are shown ±SD; weight and (C) are shown ±SEM. Statistical significance determined by repeated measures two-way ANOVA with Bonferroni multiple comparisons test (B and C) or with Tukey multiple comparisons test (D). *p < 0.05, ***p < 0.001. NS, not statistically significant.

Close modal

Toxicity was also evaluated using two additional tumor models, a B16F10-Rae1 metastatic lung tumor model and an RMA-Rae1 lymphoma model. Similar toxicity was also observed in these tumor models (Supplemental Fig. 1B, 1C). These observations prompted us to evaluate the dependence of toxicity development on presence of a tumor. Evaluation of toxicity in nontumor-bearing mice revealed that toxicity occurred regardless of the presence of the tumor (Fig. 1D). As NKG2D-L expression has been reported on some normal healthy mouse tissues (24, 25), this likely represents an on-target, off-tumor response, recapitulating a common clinical issue. Indeed, immunohistochemical analysis in mice revealed positive Rae1 expression in the liver. The Rae1-specific Ab was blocked by preincubation with a rRae-1–fusion protein in contiguous sections from the same tissue sample, thus demonstrating Ag-binding specificity (Fig. 1E). Other tissues examined (heart, lymph node, kidney, and spleen) showed no positive staining compared with Rae1 protein–blocked slides (Supplemental Fig. 1D, 1E). Together, these findings establish an immunocompetent mouse tumor model that demonstrates dose-dependent, transient, on-target, off-tumor symptoms of toxicity when treated with mNKG2D-2C11.

At the point of maximal symptoms of toxicity, 3 h after the third bsTCE injection, plasma cytokines, pathology and immune populations were evaluated. The cytokine profile in these mice was elevated for IL-6, IL-10, IFN-γ, TNF-α, and IL-2 (Fig. 2A, 2B), which was highly similar to those reported in patients receiving blinatumomab treatment (11, 12). In addition to the cytokines that have been reported from patients, we examined a wider panel of cytokines, many of which were elevated (Fig. 2A, Supplemental Fig. 2). When fold change between control versus mNKG2D-2C11–treated mice was examined, IFN-γ (145-fold) was identified as the cytokine with the greatest increase (Fig. 2A).

FIGURE 2.

mNKG2D-2C11–mediated toxicity is associated with elevated plasma cytokines and changes in T cell and myeloid populations. MC38 (1 × 106) cells were injected s.c. into WT B6 mice. Subsequently, mice received a total of three 10-μg i.v. injections of bsTCE every other day. Three hours after the third bsTCE injection, (A and B) plasma was collected and analyzed by multiplex cytokine array. (A) All cytokines evaluated are shown ranked by greatest fold change between mNKG2D-2C11– and control-treated mice. Dashed line indicates a fold change >2. All cytokines above this threshold were significantly different from the control group by unpaired two-tailed Mann–Whitney U test. (B) Plasma concentration of select cytokines shown as mean ± SD. Data pooled from two experiments (n = 8). (C and D) Tissues were paraffin embedded, sectioned, and stained with H&E. (C) Representative images of spleens (above, original magnification ×4) or livers (below, original magnification ×40 [left] and ×200 [right]) are shown. (D) Percentage of mice displaying each phenotype are shown. Statistical significance determined by two-tailed Fisher exact test performed on the number of mice with each phenotype. Data pooled from six experiments (n = 17–21). (EH) Flow cytometry was performed to analyze cell populations in the indicated tissues. (E) Representative flow plots showing changes in T cell and myeloid populations in the spleen. The indicated cell populations are shown as a percentage of CD3+ (%CD3+) cells in (F), percentage of CD8+ (%CD8+) T cells in (G), and percentage of CD45+ (%CD45+) cells in (H) and (I) show Ly-6C+ populations (CD11b+Ly-6GF4/80) as percentage of CD11b+ (%CD11b+) cells. Data pooled from four experiments (n = 9–12). Statistical significance determined by unpaired t test without assuming a consistent SD. Error bars show mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001. NS, not statistically significant.

FIGURE 2.

mNKG2D-2C11–mediated toxicity is associated with elevated plasma cytokines and changes in T cell and myeloid populations. MC38 (1 × 106) cells were injected s.c. into WT B6 mice. Subsequently, mice received a total of three 10-μg i.v. injections of bsTCE every other day. Three hours after the third bsTCE injection, (A and B) plasma was collected and analyzed by multiplex cytokine array. (A) All cytokines evaluated are shown ranked by greatest fold change between mNKG2D-2C11– and control-treated mice. Dashed line indicates a fold change >2. All cytokines above this threshold were significantly different from the control group by unpaired two-tailed Mann–Whitney U test. (B) Plasma concentration of select cytokines shown as mean ± SD. Data pooled from two experiments (n = 8). (C and D) Tissues were paraffin embedded, sectioned, and stained with H&E. (C) Representative images of spleens (above, original magnification ×4) or livers (below, original magnification ×40 [left] and ×200 [right]) are shown. (D) Percentage of mice displaying each phenotype are shown. Statistical significance determined by two-tailed Fisher exact test performed on the number of mice with each phenotype. Data pooled from six experiments (n = 17–21). (EH) Flow cytometry was performed to analyze cell populations in the indicated tissues. (E) Representative flow plots showing changes in T cell and myeloid populations in the spleen. The indicated cell populations are shown as a percentage of CD3+ (%CD3+) cells in (F), percentage of CD8+ (%CD8+) T cells in (G), and percentage of CD45+ (%CD45+) cells in (H) and (I) show Ly-6C+ populations (CD11b+Ly-6GF4/80) as percentage of CD11b+ (%CD11b+) cells. Data pooled from four experiments (n = 9–12). Statistical significance determined by unpaired t test without assuming a consistent SD. Error bars show mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001. NS, not statistically significant.

Close modal

Tissue sections of the brain, heart, lung, liver, spleen, and kidneys were stained with H&E and evaluated by a pathologist blinded to treatment conditions. This analysis showed a significant reduction in white pulp and marginal zone in mouse spleens (17/18 mice, 94%) and increased lymphocyte infiltration in the liver of animals treated with mNKG2D-2C11 (18/21 mice, 86%) (Fig. 2C, 2D). However, the other organs evaluated (brain, heart, lung, and kidneys) were within normal limits, including no evidence of liver hepatocyte death (data not shown).

Additionally, systemic changes in immune populations were found in the evaluated organs (blood, liver, spleen, tumor draining lymph node, and tumor). The primary changes found were in the T cell and myeloid compartments (Fig. 2F–I, Supplemental Fig. 3). Within the T cell compartment, the percentage of CD4+ T cells was decreased and percentage of CD8+ T cells was increased significantly in the lymph node, spleen, and liver (p < 0.01) and showed a similar trend in the tumor (p = 0.06). However, in the blood, the opposite occurred, with CD4+ T cells increased (p < 0.001) and CD8+ T cells decreased (p < 0.01) (Fig. 2F). In the blood, lymph node, spleen, and liver, the percentage of CD8+CD44+ T cells increased (p < 0.05) (Fig. 2G), likely indicative of CD8+ T cell activation. Indeed, the overall percentage of T cells (CD3+) was decreased in the blood (p < 0.001), which may reflect a transient lymphopenia (margination) effect, which has been observed with other CD3-targeting Abs (26) (Fig. 2H). In the myeloid compartment, the total percentage of CD11b+ cells was increased in the blood, and the percentage of Ly-6Cint monocytes (CD11b+Ly-6GF4/80) increased in the spleen and liver, with concomitant decrease in the percentage of Ly-6Chi monocyte population (all p < 0.001) (Fig. 2H, 2I). Overall, this model was similar to the cytokine profile observed in patients receiving blinatumomab and was associated with systemic immune cell changes, including histological changes in the spleen and liver, and increased percentages and activation of CD8+ T cells as well as changes in the myeloid population.

To confirm the requirement for T cells in the development of this toxicity, the same experimental setup was examined in WT and Rag−/− mice. WT mice experienced toxicity, but no toxicity was observed at any point in the Rag−/− mice (Fig. 3A). To investigate the contribution of T cell subsets, depletion of CD4+ or CD8+ T cells was achieved by treatment with depleting Abs 1 d prior to the first treatment with bsTCEs. Although CD8+ T cells showed the most prominent activation after mNKG2D-2C11 treatment (Supplemental Fig. 3D–F), depletion of either CD4+ T cells or CD8+ T cells was sufficient to inhibit the development of symptoms in mice (Fig. 3B). IFN-γ was the cytokine that showed the greatest fold increase and is important for immune activation, so we investigated whether IFN-γ was required for the development of toxicity. Indeed, symptoms did not occur in IFN-γ−/− mice compared with WT mice (Fig. 3C). In addition to IFN-γ production, activation of T cells results in cytotoxic killing of target cells through perforin and granzyme. To determine whether perforin-mediated killing might be involved in this acute toxicity, we compared the response of perforin-deficient mice with WT mice and found that toxicity was reduced but not completely prevented in perforin−/− mice (Fig. 3D). Thus, the data have shown that both CD4+ and CD8+ T cells, as well as IFN-γ, are required for mNKG2D-2C11–mediated toxicity, whereas perforin deficiency reduced, but did not completely prevent, toxicity development.

FIGURE 3.

T cell activity and function are required for mNKG2D-2C11–induced toxicity. MC38 (1 × 106) cells were injected s.c. into WT B6 mice. Subsequently, mice received a total of four 10-μg i.v. injections of bsTCE every other day. Health scoring was blinded and evaluated 3 h after each treatment and additionally after 24 h where indicated. (A) Experiment was conducted in WT or Rag−/− mice. (B) Mice received 200 μg i.p. of CD4 or CD8 depleting mAb or Ig control 1 d prior to the first injection of bsTCE. (C) Experiment was conducted in WT or IFN-γ−/− mice. (D) Experiment was conducted in WT or perforin−/− mice. Figure legend notation indicates number of mice developing toxicity/total treated. In (A)–(C), data pooled from two experiments (n = 7–8). In (D), data pooled from five experiments (n = 20–22). All data are shown ±SD. Statistical significance determined by repeated measures two-way ANOVA with Tukey multiple comparisons correction. ***p < 0.001. NS, not statistically significant.

FIGURE 3.

T cell activity and function are required for mNKG2D-2C11–induced toxicity. MC38 (1 × 106) cells were injected s.c. into WT B6 mice. Subsequently, mice received a total of four 10-μg i.v. injections of bsTCE every other day. Health scoring was blinded and evaluated 3 h after each treatment and additionally after 24 h where indicated. (A) Experiment was conducted in WT or Rag−/− mice. (B) Mice received 200 μg i.p. of CD4 or CD8 depleting mAb or Ig control 1 d prior to the first injection of bsTCE. (C) Experiment was conducted in WT or IFN-γ−/− mice. (D) Experiment was conducted in WT or perforin−/− mice. Figure legend notation indicates number of mice developing toxicity/total treated. In (A)–(C), data pooled from two experiments (n = 7–8). In (D), data pooled from five experiments (n = 20–22). All data are shown ±SD. Statistical significance determined by repeated measures two-way ANOVA with Tukey multiple comparisons correction. ***p < 0.001. NS, not statistically significant.

Close modal

We next examined how IFN-γ affected the systemic changes in immune cell populations observed. We found that changes in both T cell and myeloid populations were inhibited in IFN-γ−/− mice (Fig. 4A–D). Neither the changes in percentage of CD4+ T cells and CD8+ T cells (Fig. 4A) nor the changes in CD11b+ (Fig. 4C) or Ly-6Cint monocyte populations (Fig. 4D) were seen in nearly all tissues examined in IFN-γ−/− mice. In these experiments, the increased percentage of CD8+CD44+ T cells only reached statistical significance in WT mice in the lymph node and spleen, and IFN-γ deficiency inhibited the increase in this population in the lymph node, but not in the spleens, of treated mice (Fig. 4B).

FIGURE 4.

IFN-γ is produced mainly by splenic CD8+ T cells and is required for changes in immune cell populations and plasma cytokines. MC38 (1 × 106) cells were injected s.c. into WT or IFN-γ−/− mice. Subsequently, mice received a total of three 10-μg i.v. injections of bsTCE every other day. Three hours after the third bsTCE injection, flow cytometry was performed to analyze cell populations in the indicated tissues. The specified cell populations are shown as a percentage of CD3+ cells (A), percentage of CD8+ (%CD8+) T cells (B), percentage of CD45+ (%CD45+) cells (C), and percentage of CD11b+ (%CD11b+) cells in (D). Blood was also collected and analyzed for plasma cytokines by multiplex bead array (E). Data pooled from two experiments (n = 6). Error bars show mean ± SD. (F and G) Tissues were paraffin embedded, sectioned, and stained with H&E. (F) Percentage of mice with spleen pathology. Data from one experiment (n = 3). (G) Percentage of mice with liver pathology. Data from two experiments (n = 5–6). (HK) Tissues from IFN-γ reporter mice treated as in (A)–(D) were analyzed by flow cytometry for YFP expression (indicating IFN-γ production). (H) Quantification of the YFP+ cells as a percentage of CD45+ cells for each population. (I–K) Representative flow plots are shown with splenic YFP+ NK cells shown in (I), splenic T cells shown in (J), and liver T cells shown in (K). Data pooled from two experiments (n = 7). For (A)–(D), statistical significance determined by two-way ANOVA with Tukey multiple comparison correction. For (E), significance determined by Kruskal–Wallis test with Dunn posttest. In (I), significance determined by unpaired t test without assuming a consistent SD. *p < 0.05, **p < 0.01, ***p < 0.001.

FIGURE 4.

IFN-γ is produced mainly by splenic CD8+ T cells and is required for changes in immune cell populations and plasma cytokines. MC38 (1 × 106) cells were injected s.c. into WT or IFN-γ−/− mice. Subsequently, mice received a total of three 10-μg i.v. injections of bsTCE every other day. Three hours after the third bsTCE injection, flow cytometry was performed to analyze cell populations in the indicated tissues. The specified cell populations are shown as a percentage of CD3+ cells (A), percentage of CD8+ (%CD8+) T cells (B), percentage of CD45+ (%CD45+) cells (C), and percentage of CD11b+ (%CD11b+) cells in (D). Blood was also collected and analyzed for plasma cytokines by multiplex bead array (E). Data pooled from two experiments (n = 6). Error bars show mean ± SD. (F and G) Tissues were paraffin embedded, sectioned, and stained with H&E. (F) Percentage of mice with spleen pathology. Data from one experiment (n = 3). (G) Percentage of mice with liver pathology. Data from two experiments (n = 5–6). (HK) Tissues from IFN-γ reporter mice treated as in (A)–(D) were analyzed by flow cytometry for YFP expression (indicating IFN-γ production). (H) Quantification of the YFP+ cells as a percentage of CD45+ cells for each population. (I–K) Representative flow plots are shown with splenic YFP+ NK cells shown in (I), splenic T cells shown in (J), and liver T cells shown in (K). Data pooled from two experiments (n = 7). For (A)–(D), statistical significance determined by two-way ANOVA with Tukey multiple comparison correction. For (E), significance determined by Kruskal–Wallis test with Dunn posttest. In (I), significance determined by unpaired t test without assuming a consistent SD. *p < 0.05, **p < 0.01, ***p < 0.001.

Close modal

Plasma cytokines were also evaluated in WT or IFN-γ−/− mice. Although IFN-γ−/− did not completely inhibit the elevation of any cytokines evaluated, the mean concentration of IL-2, IL-6, and TNF-α were all decreased in mNKG2D-2C11–treated IFN-γ−/− mice compared with WT-treated mice (Fig. 4E).

Additionally, the development of pathology in mouse tissues was evaluated. Reduction in white pulp and marginal zone in mouse spleens occurred in 100% of mNKG2D-2C11–treated mice regardless of whether they were WT or IFN-γ−/− (Fig. 4F). However, IFN-γ−/− mice had increased baseline lymphocyte infiltration in the liver compared with WT mice. In the control-treated WT group, 0/6 mice were scored positive for lymphocyte infiltration in the liver, and 6/6 were scored as having severe infiltration in the WT mNKG2D-2C11–treated group. In contrast, control-treated IFN-γ−/− mice had 2/5 (40%), scored as having mild lymphocyte infiltration. Despite this increased baseline infiltration, differences were observed when compared with the mNKG2D-2C11–treated IFN-γ−/− group. In this group, 1/6 (17%) of mice had no infiltration, 2/6 (33%) had mild infiltration, and 3/6 (50%) had severe infiltration (Fig. 4G). These results indicated that there was an increase in liver lymphocyte infiltration in mNKG2D-2C11–treated IFN-γ−/− mice. However, this increase in lymphocyte infiltration was blunted compared with mNKG2D-2C11–treated WT mice, despite higher baseline infiltration, suggesting that this phenotype is associated with toxicity development in contrast to the loss in splenic white pulp, which occurred regardless of toxicity.

To determine the cellular source of IFN-γ, we employed reporter mice that express yellow fluorescent protein (YFP) under the control of the IFN-γ promoter. YFP expression was evaluated in NK and T cells at the peak of toxicity in the spleen, liver, blood, lymph node, and tumor. This analysis indicated that CD8+ T cells in the spleen were the main source of IFN-γ production, although NK cells also contributed. In the liver, there was also a trend for increased YFP expression in T cells (p = 0.08) (Fig. 4H–K). These data are consistent with the observation that pathology occurred in the spleen and liver and likely reflects T cell activation in these organs, possibly in response to endogenous expression of NKG2D-Ls.

In contrast to IFN-γ–deficient mice, perforin-deficient mice had very similar changes in T cell and myeloid populations compared with mice in the WT groups (Fig. 5A–D). The only differences between perforin−/− mice in comparison with WT mice were in the blood and lymph node. Within these tissues, the trends in the T cell and myeloid populations were similar but failed to reach statistical significance in perforin−/− mice, whereas changes were significant in WT mice when comparing mNKG2D-2C11–treated mice to the control groups (Fig. 5C). However, all other cellular changes in both the T cell and myeloid compartment were similarly affected in WT and perforin−/− mice when treated with mNKG2D-2C11 (Fig. 5A, 5B, 5D). Analysis of plasma cytokines in the control versus treatment group of both WT and perforin−/− mice showed elevation of IFN-γ, TNF-α, and IL-10. However, the increase in mean plasma concentration of IL-2 and IL-6 in control- versus the mNKG2D-2C11–treated group was reduced in perforin−/− mice compared with WT mice (Fig. 5E). Additionally, pathology analysis in perforin−/− mice was similar to WT mice, with reduced white pulp and marginal zone in the spleen (6/6 mice, 100%) and severe lymphocyte infiltration in the liver portal tract (6/6 mice, 100%) (Fig. 5F, 5G). These data have shown that IFN-γ, but not perforin, is primarily responsible for driving the cellular and cytokine changes in immune populations. However, the effect of perforin deficiency on cytokine levels may explain the decreased toxicity observed in these mice.

FIGURE 5.

Perforin minimally impacts changes in cellular immune populations. MC38 (1 × 106) cells were injected s.c. into WT or perforin−/− mice. Subsequently, mice received a total of three 10-μg i.v. injections of bsTCE every other day. Three hours after the third bsTCE injection, flow cytometry was performed to analyze cell populations in the indicated tissues. The indicated cell populations are shown as a percentage of CD3+ cells (A), percentage of CD8+ (%CD8+) T cells (B), percentage of CD45+ (%CD45+) cells (C), and percentage of CD11b+ (%CD11b+) cells (D). Blood was also collected and analyzed for plasma cytokines by multiplex bead array (E). Data pooled from two experiments (n = 6). Error bars show mean ± SD. (F and G) Tissues were paraffin embedded, sectioned, stained with H&E, and evaluated by a pathologist blinded to treatment. (F) Percentage of mice with spleen pathology. (G) Percentage of mice with liver pathology. Statistical significance determined by two-tailed Fisher exact test performed on the number of mice with each phenotype. Data from two experiments (n = 6). For (A)–(D), statistical significance determined by one-way ANOVA with Tukey multiple comparison correction. For (E), significance determined by Kruskal–Wallis test with Dunn posttest. *p < 0.05, **p < 0.01, ***p < 0.001.

FIGURE 5.

Perforin minimally impacts changes in cellular immune populations. MC38 (1 × 106) cells were injected s.c. into WT or perforin−/− mice. Subsequently, mice received a total of three 10-μg i.v. injections of bsTCE every other day. Three hours after the third bsTCE injection, flow cytometry was performed to analyze cell populations in the indicated tissues. The indicated cell populations are shown as a percentage of CD3+ cells (A), percentage of CD8+ (%CD8+) T cells (B), percentage of CD45+ (%CD45+) cells (C), and percentage of CD11b+ (%CD11b+) cells (D). Blood was also collected and analyzed for plasma cytokines by multiplex bead array (E). Data pooled from two experiments (n = 6). Error bars show mean ± SD. (F and G) Tissues were paraffin embedded, sectioned, stained with H&E, and evaluated by a pathologist blinded to treatment. (F) Percentage of mice with spleen pathology. (G) Percentage of mice with liver pathology. Statistical significance determined by two-tailed Fisher exact test performed on the number of mice with each phenotype. Data from two experiments (n = 6). For (A)–(D), statistical significance determined by one-way ANOVA with Tukey multiple comparison correction. For (E), significance determined by Kruskal–Wallis test with Dunn posttest. *p < 0.05, **p < 0.01, ***p < 0.001.

Close modal

In addition to changes in the T cell compartment, flow cytometry analysis revealed changes in the myeloid cell compartment. We sought to determine the impact of this population through clodronate depletion experiments. When compared with sham or control liposome treatment, mice injected with clodronate liposomes had reduced development of mNKG2D-2C11–induced toxicity. The observed toxicity in each group (at any point during evaluation) was 11/12 (92%) of mice in the sham treated group, 9/11 mice (82%) treated with control liposomes, and only 3/11 mice (27%) in the clodronate-treatment group (Fig. 6A). Therefore, several functional properties of macrophages were investigated. GM-CSF, IL-6, and inducible NO synthase (iNOS) all have important functional roles for myeloid cells and have also been implicated in development of toxicity in clinical and preclinical studies in subjects treated with T cell–redirecting therapeutics (8, 16, 18, 2729). Indeed, toxicity development was reduced in this model when GM-CSF–deficient mice were tested (Fig. 6B), but was unaltered in the presence of IL-6R blocking Ab (Fig. 6C) or by pharmacological inhibition of iNOS with 1400W (Fig. 6D). These data demonstrate a role for myeloid cells in the development of mNKG2D-2C11–mediated toxicity with partial contribution from GM-CSF, but not from IL-6 or iNOS.

FIGURE 6.

Phagocytic cell depletion or GM-CSF deficiency reduce development of mNKG2D-2C11–mediated toxicity. MC38 (1 × 106) cells were injected s.c. into WT B6 mice. Subsequently, mice received 10-μg i.v. injections of bsTCE every other day. Health scoring was blinded and evaluated 3 h after each treatment and additionally after 24 h where indicated. (A) Mice received 200 μl i.v. of clodronate, control liposomes, or HBSS 1 d prior to the first injection of bsTCE. Data pooled from three experiments (n = 8–12). (B) Experiment was conducted in WT or GM-CSF−/− mice. Data pooled from two experiments (n = 8–9). (C) Mice received 200 μg i.p. of IL-6R blocking mAb or isotype 3 h prior to the third injection of bsTCE. Data pooled from two experiments (n = 8). (D) Mice received 100 μg i.p. of the iNOS inhibitor 1400W or HBSS either daily or 5 h prior to second and third injections of bsTCE. Data pooled from two experiments (n = 8). All data are shown ±SD. Statistical significance determined by repeated measures two-way ANOVA with Tukey multiple comparisons test (A, B, and D) or with Bonferroni multiple comparisons test (C). *p < 0.05, **p < 0.01, ***p < 0.001. NS, not statistically significant.

FIGURE 6.

Phagocytic cell depletion or GM-CSF deficiency reduce development of mNKG2D-2C11–mediated toxicity. MC38 (1 × 106) cells were injected s.c. into WT B6 mice. Subsequently, mice received 10-μg i.v. injections of bsTCE every other day. Health scoring was blinded and evaluated 3 h after each treatment and additionally after 24 h where indicated. (A) Mice received 200 μl i.v. of clodronate, control liposomes, or HBSS 1 d prior to the first injection of bsTCE. Data pooled from three experiments (n = 8–12). (B) Experiment was conducted in WT or GM-CSF−/− mice. Data pooled from two experiments (n = 8–9). (C) Mice received 200 μg i.p. of IL-6R blocking mAb or isotype 3 h prior to the third injection of bsTCE. Data pooled from two experiments (n = 8). (D) Mice received 100 μg i.p. of the iNOS inhibitor 1400W or HBSS either daily or 5 h prior to second and third injections of bsTCE. Data pooled from two experiments (n = 8). All data are shown ±SD. Statistical significance determined by repeated measures two-way ANOVA with Tukey multiple comparisons test (A, B, and D) or with Bonferroni multiple comparisons test (C). *p < 0.05, **p < 0.01, ***p < 0.001. NS, not statistically significant.

Close modal

In this study, we have used an immunocompetent, syngeneic mouse tumor model to study the mechanisms contributing to toxicity induced by the T cell–engaging bispecific mNKG2D-2C11. We established that mNKG2D-2C11 induces a transient, dose-dependent toxicity that coincides with high levels of plasma cytokines. Additionally, we demonstrated that IFN-γ, produced primarily by CD8+ T cells in the spleen, is required for development of these CRS-like symptoms and changes in immune cells. We also establish that both CD4+ and CD8+ T cells as well as phagocytes contribute to the observed toxicity. Furthermore, data suggest that T cell cytotoxicity via perforin contributes to severity but is not required for development of this toxicity.

The toxicity observed was independent of a tumor and is likely the result of on-target, off-tumor toxicity, although it is difficult to pinpoint the source of the target. We did observe Rae1 expression in the liver as well as some indications of T cell activation in this tissue. Mouse NKG2D recognizes several ligands, including Rae1, H60, and Mult1, and there are several isoforms (22). The mRNA expression of NKG2D-Ls is insufficient to predict surface expression, as NKG2D-Ls are regulated at both the transcriptional and posttranslational levels (22). A possible explanation for the systemic responses observed in this mouse model is that NKG2D-Ls may be present at very low levels in additional mouse tissues, such as the spleen. Alternatively, a T cell response against hepatocytes may act as a triggering event that induces myeloid cells to upregulate NKG2D-L, which may occur under stress conditions (25, 3037).

One of the main strategies to manage toxicity of bsTCEs is to use a stepwise dosing schedule to reduce the frequency and severity of toxicities in patients treated with blinatumomab (38). Although the mechanism is not understood, this phenomenon is observed clinically after blinatumomab treatment (39) and has been observed in multiple other preclinical models (26, 28, 40). One of the main hypotheses is that many target-expressing cells are depleted early in treatment, leading to reduced T cell activation and, therefore, cytokine release in subsequent cycles. However, it was recently shown ex vivo that a reduced cytokine response could still occur when fresh target cells were supplemented during repeated treatment cycles (26). Another possible explanation is that the initial T cell activation and cytokine release activate negative feedback mechanisms, such as PD-1 expression (41), that restrain T cell cytokine release during subsequent dosing. Considering the dose responsiveness of mNKG2D-2C11–induced toxicity and that toxicity tapered off after the third dose, it is possible that a stepwise dosing schedule would minimize toxicity. Other feedback mechanisms may limit unchecked T cell activation, such as IFN-γ and IL-10, both of which were elevated and have been reported to downregulate the expression of NKG2D-Ls on mouse and human tumor cells (25, 42, 43).

Although the pathophysiology of CRS is incompletely understood (9), some recent studies have identified several components that may be involved in the development of CRS. T cell–derived IFN-γ (9, 14, 44), TNF-α (26, 45), and GM-CSF (28, 29) have all been identified as cytokines that may be involved in initiating the cascade leading to cytokine storm. The initial T cell cytokine release is thought to result in the activation of bystander cells, especially monocyte (27) and macrophage (16) populations. These myeloid cells have been shown to be the main producers of IL-1 (16, 26, 27), IL-6 (16, 26, 27), and NO (16), which can contribute to the development of CRS. Several of the studies implicating these cytokines in CRS development have been CAR T cell studies, including preclinical studies regarding GM-CSF (28, 29), IL-1 (16, 27), IL-6 (16, 27), and NO (16). Although the clinical presentation of CRS in patients treated with bsTCEs is similar to CRS observed in CAR T cell–treated patients, caution should be used when extrapolating these results (26). These treatment modalities have some major differences, such as the kinetics of CRS onset and severity correlating with the magnitude of initial T cell activation (26). In this work, GM-CSF was not as critical a component in the development of CRS symptoms as has been observed for CAR T cells (28).

A recent study reported a mechanism of toxicity for a bsTCE in which T cell derived TNF-α–mediated activation of myeloid cells, which drove the production of myeloid-derived IL-1 and IL-6 (26). However, this study did not show that the modulation of cytokines had a phenotypic manifestation on symptoms in the mice, making it difficult to determine if this these cytokine changes were sufficient to prevent CRS symptoms (26). Additionally, although it has been suggested that IFN-γ plays an important role in initiating CRS (9), neither CAR T cell studies nor bsTCE studies have directly shown a role of IFN-γ in CRS. A murine model of CRS induced by treatment with an anti-CD3 mAb (145-2C11) has shown that prophylactic treatment with an IFN-γ blocking Ab protected mice against pathological changes. These symptoms correlated with intermouse strain differences in the ability of the strain to produce IFN-γ (44). In this study, we show that IFN-γ was required for the development of mNKG2D-2C11 bsTCE–induced toxicity. Given that IFN-γ is a driving factor in the development of this CRS-like toxicity in mice and the correlation of IFN-γ with the severity of clinical CRS (14), it is possible that IFN-γ blockade would be effective in inhibiting CRS. Recently, the IFN-γ blocking Ab Gamifant (emapalumab-lzsg) was approved for the treatment of macrophage activation syndrome (46), which shares many of the characteristics associated with severe CRS and CRS-induced macrophage activation syndrome. Although inhibition of IFN-γ may mitigate toxicity, IFN-γ is also a critical factor in mediating the antitumor immune response. However, in severe cases that are refractory to other treatment interventions, this might be a viable option.

One of the most consistent factors between preclinical models of CRS is the involvement of the myeloid population. This likely represents the ideal breakpoint to intervene in the development of toxicity, either upstream of monocyte/macrophage activation or through its downstream mediators. However, depletion of this population may be a suboptimal strategy, as one CAR T cell study using clodronate to mitigate CRS symptoms also showed that the antitumor response was reduced by this treatment (27). Clinical data indicate that IL-6 blockade is not universally effective in reversing symptoms and preventing cytokine release–related death (14, 15, 18, 26). Cases of CRS that are refractory to both tocilizumab and glucocorticoids support the need to identify additional therapeutic interventions to treat CRS (9, 47).

In summary, this work has investigated systemic immune changes occurring during toxicity induced by the mNKG2D bsTCE. We have found activation of both T cell and myeloid populations, and we show a critical role for T cell–derived IFN-γ in initiating this toxicity. The lack of impact of IL-6R blockade may make this model ideal to identify additional points that could be therapeutically targeted. Together, these findings shed light on the molecular mediators of mNKG2D-2C11–induced toxicity and set the stage to elucidate additional druggable targets for the inhibition of bsTCE-induced toxicity.

We thank DartLab; the immune monitoring and flow cytometry shared resource at the Norris Cotton Cancer Center for assistance with flow cytometry and Luminex assays; and the staff at the Center for Comparative Medicine and Research for assistance with animal care. The authors also thank the National Cancer Institute Biological Resource Branch for human rIL-2.

This work was supported in part by grants from the National Cancer Institute (NCI), National Institutes of Health (CA164178 to C.L.S.), National Institute of General Medical Sciences (GM008704 to C.G.-P.), and National Institute of Allergy and Infectious Diseases (AI007363 to T.A.C.). DartLab at the Norris Cotton Cancer Center is supported in part by NCI Cancer Center Grant 5P30 CA023108-37.

The online version of this article contains supplemental material.

Abbreviations used in this article:

ALL

acute lymphoblastic leukemia

bsTCE

bispecific T cell engager

CAR

chimeric AgR

CRS

cytokine release syndrome

iNOS

inducible NO synthase

NKG2D

NK Group 2D

NKG2D-L

NKG2D ligand

scFv

single-chain variable fragment

WT

wild type; YFP, yellow fluorescent protein.

1
Smyth
,
M. J.
,
G. P.
Dunn
,
R. D.
Schreiber
.
2006
.
Cancer immunosurveillance and immunoediting: the roles of immunity in suppressing tumor development and shaping tumor immunogenicity.
Adv. Immunol.
90
:
1
50
.
2
Maher
,
J.
,
E. T.
Davies
.
2004
.
Targeting cytotoxic T lymphocytes for cancer immunotherapy.
Br. J. Cancer
91
:
817
821
.
3
Brinkmann
,
U.
,
R. E.
Kontermann
.
2017
.
The making of bispecific antibodies.
MAbs
9
:
182
212
.
4
Wolf
,
E.
,
R.
Hofmeister
,
P.
Kufer
,
B.
Schlereth
,
P. A.
Baeuerle
.
2005
.
BiTEs: bispecific antibody constructs with unique anti-tumor activity.
Drug Discov. Today
10
:
1237
1244
.
5
Labrijn
,
A. F.
,
M. L.
Janmaat
,
J. M.
Reichert
,
P. W. H. I.
Parren
.
2019
.
Bispecific antibodies: a mechanistic review of the pipeline.
Nat. Rev. Drug Discov.
18
:
585
608
.
6
Stein
,
A.
,
J. L.
Franklin
,
V. M.
Chia
,
D.
Arrindell
,
W.
Kormany
,
J.
Wright
,
M.
Parson
,
H. R.
Amouzadeh
,
J.
Choudhry
,
G.
Joseph
.
2019
.
Benefit-risk assessment of blinatumomab in the treatment of relapsed/refractory B-cell precursor acute lymphoblastic leukemia.
Drug Saf.
42
:
587
601
.
7
Jain
,
T.
,
M. R.
Litzow
.
2018
.
No free rides: management of toxicities of novel immunotherapies in ALL, including financial.
Blood Adv.
2
:
3393
3403
.
8
Teachey
,
D. T.
,
S. R.
Rheingold
,
S. L.
Maude
,
G.
Zugmaier
,
D. M.
Barrett
,
A. E.
Seif
,
K. E.
Nichols
,
E. K.
Suppa
,
M.
Kalos
,
R. A.
Berg
, et al
.
2013
.
Cytokine release syndrome after blinatumomab treatment related to abnormal macrophage activation and ameliorated with cytokine-directed therapy.
Blood
121
:
5154
5157
.
9
Shimabukuro-Vornhagen
,
A.
,
P.
Gödel
,
M.
Subklewe
,
H. J.
Stemmler
,
H. A.
Schlößer
,
M.
Schlaak
,
M.
Kochanek
,
B.
Böll
,
M. S.
von Bergwelt-Baildon
.
2018
.
Cytokine release syndrome.
J. Immunother. Cancer
6
:
56
.
10
Amgen. BLINCYTO: Highlights of prescribing information. Available at: https://www.pi.amgen.com/~/media/amgen/repositorysites/pi-amgen-com/blincyto/blincyto_pi_hcp_english.pdf. Accessed: October 3, 2019.
11
Zhu
,
M.
,
A.
Kratzer
,
J.
Johnson
,
C.
Holland
,
C.
Brandl
,
I.
Singh
,
A.
Wolf
,
S.
Doshi
.
2018
.
Blinatumomab pharmacodynamics and exposure-response relationships in relapsed/refractory acute lymphoblastic leukemia.
J. Clin. Pharmacol.
58
:
168
179
.
12
Zhu
,
M.
,
B.
Wu
,
C.
Brandl
,
J.
Johnson
,
A.
Wolf
,
A.
Chow
,
S.
Doshi
.
2016
.
Blinatumomab, a bispecific T-cell engager (BiTE(®)) for CD-19 targeted cancer immunotherapy: clinical pharmacology and its implications.
Clin. Pharmacokinet.
55
:
1271
1288
.
13
Davila
,
M. L.
,
I.
Riviere
,
X.
Wang
,
S.
Bartido
,
J.
Park
,
K.
Curran
,
S. S.
Chung
,
J.
Stefanski
,
O.
Borquez-Ojeda
,
M.
Olszewska
, et al
.
2014
.
Efficacy and toxicity management of 19-28z CAR T cell therapy in B cell acute lymphoblastic leukemia.
Sci. Transl. Med.
6
: 224ra25.
14
Teachey
,
D. T.
,
S. F.
Lacey
,
P. A.
Shaw
,
J. J.
Melenhorst
,
S. L.
Maude
,
N.
Frey
,
E.
Pequignot
,
V. E.
Gonzalez
,
F.
Chen
,
J.
Finklestein
, et al
.
2016
.
Identification of predictive biomarkers for cytokine release syndrome after chimeric antigen receptor T-cell therapy for acute lymphoblastic leukemia.
Cancer Discov.
6
:
664
679
.
15
Hay
,
K. A.
,
L. A.
Hanafi
,
D.
Li
,
J.
Gust
,
W. C.
Liles
,
M. M.
Wurfel
,
J. A.
López
,
J.
Chen
,
D.
Chung
,
S.
Harju-Baker
, et al
.
2017
.
Kinetics and biomarkers of severe cytokine release syndrome after CD19 chimeric antigen receptor-modified T-cell therapy.
Blood
130
:
2295
2306
.
16
Giavridis
,
T.
,
S. J. C.
van der Stegen
,
J.
Eyquem
,
M.
Hamieh
,
A.
Piersigilli
,
M.
Sadelain
.
2018
.
CAR T cell-induced cytokine release syndrome is mediated by macrophages and abated by IL-1 blockade.
Nat. Med.
24
:
731
738
.
17
Garcia Borrega
,
J.
,
P.
Gödel
,
M. A.
Rüger
,
Ö. A.
Onur
,
A.
Shimabukuro-Vornhagen
,
M.
Kochanek
,
B.
Böll
.
2019
.
In the eye of the storm: immune-mediated toxicities associated with CAR-T cell therapy.
HemaSphere
3
: e191.
18
Le
,
R. Q.
,
L.
Li
,
W.
Yuan
,
S. S.
Shord
,
L.
Nie
,
B. A.
Habtemariam
,
D.
Przepiorka
,
A. T.
Farrell
,
R.
Pazdur
.
2018
.
FDA approval summary: tocilizumab for treatment of chimeric antigen receptor T cell-induced severe or life-threatening cytokine release syndrome.
Oncologist
23
:
943
947
.
19
Choudhry
,
J.
,
M.
Parson
,
J.
Wright
.
2018
.
A retrospective review of tocilizumab for the management of blinatumomab (a bispecific T cell engager)-induced cytokine release syndrome (CRS).
Blood
132
(
Suppl. 1
):
5211
.
20
Godbersen
,
C.
,
T. A.
Coupet
,
A. M.
Huehls
,
T.
Zhang
,
M. B.
Battles
,
J. L.
Fisher
,
M. S.
Ernstoff
,
C. L.
Sentman
.
2017
.
NKG2D ligand-targeted bispecific T-cell engagers lead to robust antitumor activity against diverse human tumors.
Mol. Cancer Ther.
16
:
1335
1346
.
21
Zhang
,
T.
,
C. L.
Sentman
.
2011
.
Cancer immunotherapy using a bispecific NK receptor fusion protein that engages both T cells and tumor cells.
Cancer Res.
71
:
2066
2076
.
22
Spear
,
P.
,
M. R.
Wu
,
M. L.
Sentman
,
C. L.
Sentman
.
2013
.
NKG2D ligands as therapeutic targets.
Cancer Immun.
13
:
8
.
23
Wu
,
M.-R.
,
T.
Zhang
,
A. T.
Gacerez
,
T. A.
Coupet
,
L. R.
DeMars
,
C. L.
Sentman
.
2015
.
B7H6-Specific bispecific T cell engagers lead to tumor elimination and host antitumor immunity.
J. Immunol.
194
:
5305
5311
.
24
Vilarinho
,
S.
,
K.
Ogasawara
,
S.
Nishimura
,
L. L.
Lanier
,
J. L.
Baron
.
2007
.
Blockade of NKG2D on NKT cells prevents hepatitis and the acute immune response to hepatitis B virus.
Proc. Natl. Acad. Sci. USA
104
:
18187
18192
.
25
Nausch
,
N.
,
A.
Cerwenka
.
2008
.
NKG2D ligands in tumor immunity.
Oncogene
27
:
5944
5958
.
26
Li
,
J.
,
R.
Piskol
,
R.
Ybarra
,
Y. J.
Chen
,
J.
Li
,
D.
Slaga
,
M.
Hristopoulos
,
R.
Clark
,
Z.
Modrusan
,
K.
Totpal
, et al
.
2019
.
CD3 bispecific antibody-induced cytokine release is dispensable for cytotoxic T cell activity.
Sci. Transl. Med.
11
:
eaax8861
.
27
Norelli
,
M.
,
B.
Camisa
,
G.
Barbiera
,
L.
Falcone
,
A.
Purevdorj
,
M.
Genua
,
F.
Sanvito
,
M.
Ponzoni
,
C.
Doglioni
,
P.
Cristofori
, et al
.
2018
.
Monocyte-derived IL-1 and IL-6 are differentially required for cytokine-release syndrome and neurotoxicity due to CAR T cells.
Nat. Med.
24
:
739
748
.
28
Sentman
,
M.-L.
,
J. M.
Murad
,
W. J.
Cook
,
M.-R.
Wu
,
J.
Reder
,
S. H.
Baumeister
,
G.
Dranoff
,
M. W.
Fanger
,
C. L.
Sentman
.
2016
.
Mechanisms of acute toxicity in NKG2D chimeric antigen receptor T cell-treated mice.
J. Immunol.
197
:
4674
4685
.
29
Sterner
,
R. M.
,
R.
Sakemura
,
M. J.
Cox
,
N.
Yang
,
R. H.
Khadka
,
C. L.
Forsman
,
M. J.
Hansen
,
F.
Jin
,
K.
Ayasoufi
,
M.
Hefazi
, et al
.
2019
.
GM-CSF inhibition reduces cytokine release syndrome and neuroinflammation but enhances CAR-T cell function in xenografts.
Blood
133
:
697
709
.
30
Hamerman
,
J. A.
,
K.
Ogasawara
,
L. L.
Lanier
.
2004
.
Cutting edge: toll-like receptor signaling in macrophages induces ligands for the NKG2D receptor.
J. Immunol.
172
:
2001
2005
.
31
Nedvetzki
,
S.
,
S.
Sowinski
,
R. A.
Eagle
,
J.
Harris
,
F.
Vély
,
D.
Pende
,
J.
Trowsdale
,
E.
Vivier
,
S.
Gordon
,
D. M.
Davis
.
2007
.
Reciprocal regulation of human natural killer cells and macrophages associated with distinct immune synapses.
Blood
109
:
3776
3785
.
32
Sirén
,
J.
,
T.
Sareneva
,
J.
Pirhonen
,
M.
Strengell
,
V.
Veckman
,
I.
Julkunen
,
S.
Matikainen
.
2004
.
Cytokine and contact-dependent activation of natural killer cells by influenza A or Sendai virus-infected macrophages.
J. Gen. Virol.
85
:
2357
2364
.
33
Molinero
,
L. L.
,
M. B.
Fuertes
,
G. A.
Rabinovich
,
L.
Fainboim
,
N. W.
Zwirner
.
2002
.
Activation-induced expression of MICA on T lymphocytes involves engagement of CD3 and CD28.
J. Leukoc. Biol.
71
:
791
797
.
34
Diefenbach
,
A.
,
A. M.
Jamieson
,
S. D.
Liu
,
N.
Shastri
,
D. H.
Raulet
.
2000
.
Ligands for the murine NKG2D receptor: expression by tumor cells and activation of NK cells and macrophages.
Nat. Immunol.
1
:
119
126
.
35
Rabinovich
,
B. A.
,
J.
Li
,
J.
Shannon
,
R.
Hurren
,
J.
Chalupny
,
D.
Cosman
,
R. G.
Miller
.
2003
.
Activated, but not resting, T cells can be recognized and killed by syngeneic NK cells.
J. Immunol.
170
:
3572
3576
.
36
Cerboni
,
C.
,
A.
Zingoni
,
M.
Cippitelli
,
M.
Piccoli
,
L.
Frati
,
A.
Santoni
.
2007
.
Antigen-activated human T lymphocytes express cell-surface NKG2D ligands via an ATM/ATR-dependent mechanism and become susceptible to autologous NK- cell lysis.
Blood
110
:
606
615
.
37
Ward
,
J.
,
M.
Bonaparte
,
J.
Sacks
,
J.
Guterman
,
M.
Fogli
,
D.
Mavilio
,
E.
Barker
.
2007
.
HIV modulates the expression of ligands important in triggering natural killer cell cytotoxic responses on infected primary T-cell blasts.
Blood
110
:
1207
1214
.
38
Chen
,
X.
,
C.
Kamperschroer
,
G.
Wong
,
D.
Xuan
.
2019
.
A modeling framework to characterize cytokine release upon T-cell-engaging bispecific antibody treatment: methodology and opportunities.
Clin. Transl. Sci.
12
:
600
608
.
39
Lee
,
K. J.
,
V.
Chow
,
A.
Weissman
,
S.
Tulpule
,
I.
Aldoss
,
M.
Akhtari
.
2016
.
Clinical use of blinatumomab for B-cell acute lymphoblastic leukemia in adults.
Ther. Clin. Risk Manag.
12
:
1301
1310
.
40
Amann
,
M.
,
M.
Friedrich
,
P.
Lutterbuese
,
E.
Vieser
,
G.
Lorenczewski
,
L.
Petersen
,
K.
Brischwein
,
P.
Kufer
,
R.
Kischel
,
P. A.
Baeuerle
,
B.
Schlereth
.
2009
.
Therapeutic window of an EpCAM/CD3-specific BiTE antibody in mice is determined by a subpopulation of EpCAM-expressing lymphocytes that is absent in humans.
Cancer Immunol. Immunother.
58
:
95
109
.
41
Kinter
,
A. L.
,
E. J.
Godbout
,
J. P.
McNally
,
I.
Sereti
,
G. A.
Roby
,
M. A.
O’Shea
,
A. S.
Fauci
.
2008
.
The common γ-chain cytokines IL-2, IL-7, IL-15, and IL-21 induce the expression of programmed death-1 and its ligands.
J. Immunol.
181
:
6738
6746
.
42
Bui
,
J. D.
,
L. N.
Carayannopoulos
,
L. L.
Lanier
,
W. M.
Yokoyama
,
R. D.
Schreiber
.
2006
.
IFN-dependent down-regulation of the NKG2D ligand H60 on tumors.
J. Immunol.
176
:
905
913
.
43
Serrano
,
A. E.
,
E.
Menares-Castillo
,
M.
Garrido-Tapia
,
C. H.
Ribeiro
,
C. J.
Hernández
,
A.
Mendoza-Naranjo
,
M.
Gatica-Andrades
,
R.
Valenzuela-Diaz
,
R.
Zúñiga
,
M. N.
López
, et al
.
2011
.
Interleukin 10 decreases MICA expression on melanoma cell surface.
Immunol. Cell Biol.
89
:
447
457
.
44
Matthys
,
P.
,
C.
Dillen
,
P.
Proost
,
H.
Heremans
,
J.
Van Damme
,
A.
Billiau
.
1993
.
Modification of the anti-CD3-induced cytokine release syndrome by anti-interferon-γ or anti-interleukin-6 antibody treatment: protective effects and biphasic changes in blood cytokine levels.
Eur. J. Immunol.
23
:
2209
2216
.
45
Suntharalingam
,
G.
,
M. R.
Perry
,
S.
Ward
,
S. J.
Brett
,
A.
Castello-Cortes
,
M. D.
Brunner
,
N.
Panoskaltsis
.
2006
.
Cytokine storm in a phase 1 trial of the anti-CD28 monoclonal antibody TGN1412.
N. Engl. J. Med.
355
:
1018
1028
.
46
U.S. Food and Drug Administration
.
2018
.
FDA news release: FDA approves first treatment specifically for patients with rare and life-threatening type of immune disease.
Available at: https://www.fda.gov/news-events/press-announcements/fda-approves-first-treatment-specifically-patients-rare-and-life-threatening-type-immune-disease.
47
Turtle
,
C. J.
,
L. A.
Hanafi
,
C.
Berger
,
T. A.
Gooley
,
S.
Cherian
,
M.
Hudecek
,
D.
Sommermeyer
,
K.
Melville
,
B.
Pender
,
T. M.
Budiarto
, et al
.
2016
.
CD19 CAR-T cells of defined CD4+:CD8+ composition in adult B cell ALL patients.
J. Clin. Invest.
126
:
2123
2138
.

C.L.S. has patents and patent filings on bsTCE proteins. These conflicts are managed by the policies of Dartmouth College. The other authors have no financial conflicts of interest.

Supplementary data