Tumor specimens were collected with consent from patients at Memorial Sloan Kettering Cancer Center. Specimens were frozen in Tissue-Tek optimal cutting temperature compound (OCT) (Sakura, Torrance, CA, USA) and stored at -80°C. The MSKCC institutional review board (IRB) approved all aspects of the project.

OCT-embedded tumors were sectioned at 5 µm thickness using a cryostat and mounted to a glass slide. Slides were stored at -80°C. On the day of the experiment, the immunohistochemical procedure was performed as follows. All steps were performed at room temperature unless otherwise specified. Slides were incubated in a humidified chamber during steps 5-12. For all washing steps, slides were dipped 15x in three sequential baths of fresh PBS.

Stained tissue sections were examined and reviewed by a trained pathologist (Dr. Satish K. Tickoo, MSKCC). Scoring was performed using a modified H-score with a scoring range of 0 to 300 (7). A score of zero was assigned for negative or trace staining, 1 for weak staining, 2 for moderate staining, and 3 for intense staining. Each intensity score was multiplied by the proportion of the tumor cells with that intensity, and added together as such:

T-BsAbs were generated as previously described (8). Briefly, the IgG-[L]-scFv format used for the T-BsAbs consists of a huOKT3 scFv fused to the c-terminus of the light chain of human IgG1. The T-BsAbs differ in their complementarity-determining regions (CDRs) of IgG1, which confer specificity for each particular TAA. The parental antibody clones used to design CDRs of each T-BsAb are as follows: B7-H3 – 8H9; c-MET – onartuzumab; CD19 – FMC63; EGFR – C225 (cetuximab); GD2 – 3F8; HER2 – 4D5 (traztuzumab); L1CAM – CE7; mesothelin – MORAb-009; polysialic acid – HP35; PSMA – J591.

JN-DSRCT-1 cells were a gift from the late Dr. William L. Gerald (MSKCC). BER cells were a gift from Dr. Emily Slotkin (MSKCC). SK-DSRCT-2 cells were a gift from Dr. Marc Ladanyi (MSKCC). All cell lines were cultured in RPMI 1640 medium containing L-glutamine (Corning Scientific, Corning, NY, USA) with 10% fetal bovine serum (Gibco, Gaithersburg, MD, USA) and 1% penicillin/streptomycin (Gibco, Gaithersburg, MD, USA). Cells were incubated at 37°C with 5% CO₂ and passaged two or three times per week.

A six-well plate was coated with retronectin (20 μg/mL) and incubated overnight at 4°C. Retronectin was aspirated, 3 mL 2% BSA in PBS was added per well, and the plate was incubated for 30° minutes at room temperature. BSA was aspirated and wells were washed once with PBS. 3 mL supernatant from a click beetle red luciferase/tdTomato retrovirus (kindly provided by Vladimir Ponomarev [MSKCC]) was added to each well and the plate was centrifuged at 1240 x g for 90 minutes at 32°C. Viral supernatant was aspirated and 300,000 BER DSRCT cells were added per well. The plate was checked daily and cells were trypsinized and removed two days later when they reached ~80% confluency. Transduction was confirmed by assessing fluorescence of the red fluorescent protein tdTomato by flow cytometry and by incubating the transduced BER cells (BER-luc cells) with D-luciferin (GoldBio, St. Louis, MO, USA) and measuring luminescence using a Synergy H1 Hybrid multi-mode microplate reader (Biotek, Winooksi, VT, USA).

1,000,000 DSRCT cells per well were plated in a 96-well, U-bottom plate and incubated with T-BsAbs (10 µg/mL) in 100 µL cell culture media at 4°C for 60 minutes. Cells were washed twice with PBS, resuspended with PE-conjugated mouse anti-human IgG Fc antibody (BioLegend, San Diego, CA, USA) diluted 1:500 in PBS, and incubated at 4°C for 30 minutes. Cells were washed twice more in PBS, resuspended in 50 µL per well PBS, and flow cytometry was performed using an Attune NxT Flow Cytometer (Thermo Fisher Scientific, Waltham, MA, USA). Data was analyzed using FlowJo software (FlowJo LLC, Ashland, OR, USA).

Standard chromium-release assays were performed to assess cytotoxicity as previously described (9). Briefly, DSRCT cells were incubated with ⁵¹Cr chromate (Amersham Biosciences, Arlington Heights, IL, USA) for one hour, washed, and plated in 96-well, U-bottom plates. BsAbs and activated human T cells were added and plates were incubated for four hours at 37°C. Plates were centrifuged at 400 x g for five minutes. A gamma-counter (Packard Bioscience Company, Downers Grove, IL, USA) was used to measure radiation in supernatant.

Animal studies were conducted in accordance with Institutional Animal Care and Use Committee guidelines, under protocol 09-05-010. Six to eight-week-old male CIEA BRG C.Cg-Rag2^tm1FwaIL2rg^tm1Sug /JicTac mice (Taconic Inc., Hudson, NY, USA) were housed in pathogen-free rooms in MSKCC’s Research Animal Resource Center (RARC) and fed chow containing sulfamethoxazole and trimethoprim (Sulfatrim). Mice were injected intraperitoneally (IP) with 2 x 10⁶ BER-luc or JN-DSRCT-1-luc DSRCT cells in 200 µL PBS in the lower right quadrant of the abdomen near the midline. Initially, 5 mice per experimental group were injected with tumor cells based on previous experiments conducted using T-BsAbs in our lab that suggest this sample size would be sufficiently large enough to determine statistically significant differences in treatment efficacy. Tumors were detectable using an IVIS Spectrum in vivo imaging system (PerkinElmer, Waltham, MA, USA) one to two weeks after injection, at which time mice were randomized by tumor size into treatment groups. For tumor imaging, mice were anesthetized using isoflurane, injected retro-orbitally with 100 µL D-luciferin (30 mg/mL) and placed supine in the imager. Not all the mice in the experiment using JN-DSRCT-1-luc cells developed tumors. Mice that had no detectable tumor signal on day 20 were not included in the experiment. Mice were treated once weekly with 20 x 10⁶ activated human T cells injected retro-orbitally, twice weekly with BsAbs injected IP, and once weekly with recombinant human IL-15/IL-15R-alpha injected SC (10). BsAbs were prepared as previously described (11, 12). Tumor growth was monitored once weekly using the IVIS Spectrum in vivo imaging system. All mice that were randomized into treatment groups at the beginning of the experiments were included in the data analyses. All cages within an experiment were housed on the same rack in the same room of MSKCC’s RARC. The order of treatments and imaging was varied from week to week to minimize potential confounding. Because one person was responsible for the execution and data analysis of these experiments, blinding was not possible.

DuoSet enzyme-linked immunosorbent assays (ELISAs) were purchased from R&D Systems (Minneapolis, MN, USA). Blood was collected from mice 24 hours after the first administration of activated human T cells and BsAbs, centrifuged at 2000 x g for 10 minutes, and serum was removed and frozen at -80°C for later use. On the day the ELISA assays were performed, serum was thawed at room temperature and diluted 1:10 with reagent diluent. Assays were performed according to manufacturer’s instructions.

Statistical analysis was performed using Graphpad Prism software. One-way ANOVA with Dunnett’s multiple comparison tests were used to compare areas under the curve for tumor growth experiments and cytokine secretion in ELISA experiments. P values < 0.05 were considered statistically significant.

Tumor specimens were collected from DSRCT patients undergoing surgery at MSKCC. 14 were randomly chosen from a list of 63 archived samples with intact histology. Figure 1 shows demographic and tumor site information on these patients. 13 of the patients were male and one was female. The median age of all patients at the time of surgery was 16.6 years, the mean was 17.2 years, and the age range of all patients was 3-31 years. For specimens with tumor site information, the majority were found within the abdomen and pelvis, with one each in the retroperitoneum and the thorax. The characteristics of this sample set of tumors is representative of DSRCT tumors in general, which are most commonly found in the abdomen or pelvis of adolescent and young adult males (13, 14).

IHC was performed to evaluate the expression of known TAA targets for immunotherapy and radioimmunotherapy including B7 homolog 3 (B7-H3) (15, 16), c-Met (17, 18), epidermal growth factor receptor (EGFR) (19, 20), GD2 (21), human epidermal growth factor receptor 2 (HER2) (22), L1 cell adhesion molecule (L1CAM) (23, 24), mesothelin (25, 26), polysialic acid (27), and prostate-specific membrane antigen (PSMA) (28) using target-selective antibodies built using a IgG-[L]-scFv T-BsAb format as previously outlined (5, 9, 11, 12, 29, 30). IHC using commercial antibodies against human CD3 revealed very few CD3-positive cells (T cells) in the tumors which were almost uniformly restricted to the stroma and absent from the nests of tumor cells ( Figure 2 ). This is consistent with other reports describing DSRCT as an immunologically “cold” tumor (31–33). Importantly, these findings mean that the results of our staining with the IgG-[L]-scFv T-BsAbs are not confounded by the anti-CD3 scFv binding T cells in the tumor, and instead reflect expression of TAAs only. Figure 2 shows examples of weak (score 0-100), moderate (score 101-200) and strong (score 201-300) staining for each antigen. B7-H3, EGFR, GD2, HER2, mesothelin, and polysialic acid exhibited clear membrane staining with the moderate and strong staining tumors also exhibiting some cytoplasmic staining. In general, there is heterogeneity among tumors for all the TAAs studied. Mean score was lowest for L1CAM followed by c-MET, and EGFR. Four of the antigens examined (B7-H3, GD2, HER2, and mesothelin) had mean scores in the moderate range ( Figure 3A ). This staining strategy was repeated on a group of eight DSRCT patient-derived xenograft tumors with similar results ( Figure 3B ).

In order for BsAb-based immunotherapy or radioimmunotherapy to be effective, TAAs must be expressed on the cell surface. To determine whether this was the case for our TAAs, we performed flow cytometry to assess binding of T-BsAbs to DSRCT cell lines. The cells were not permeabilized and therefore BsAb binding could only occur with targets expressed on the outer surface of the cell membrane. Figure 3C shows binding of B7-H3, EGFR, GD2, HER2, L1CAM, and mesothelin in at least one of three DSRCT cell lines. B7-H3 showed high cell surface expression in all three cell lines, while expression of EGFR, HER2, and mesothelin, were detected at lower levels in all cell lines tested. GD2 and L1CAM were detectable on the surface of one or two cell lines, respectively, but not all three. There was no expression of c-MET or polysialic acid on any of the three DSRCT cell lines tested. The lack of cell surface c-MET expression was consistent with IHC results in Figure 2 . Although polysialic acid was found on the surface of DSRCT patient tumor specimens ( Figures 2 and 3A ), it were absent from all three cell lines tested. Outgrowth in vitro of clones missing polysialic acid with proliferative advantage could explain these differences.

One way to assess the ability of T-BsAbs to engage cytotoxic T cells and kill tumor cells is by using in vitro cytotoxicity assays. We performed standard chromium-release assays using T-BsAbs directed against our TAAs of interest and activated human T cells in culture with three DSRCT cell lines. Figure 4 shows that the T-BsAbs specific for B7-H3, EGFR, and HER2 were highly cytotoxic in all three cell lines. The mesothelin T-BsAb was moderately cytotoxic in the JN-DSRCT-1 and BER cell lines and highly cytotoxic in the SK-DSRCT-2 cell line. The GD2 T-BsAb showed minimal cytotoxicity at high concentrations, consistent with the low level of GD2 expression on the surface of these cell lines. The c-MET, L1CAM, and polysialic acid T-BsAbs exhibited no or very little cytotoxicity, even at high concentrations. In general, the level of TAA cell surface expression (shown in Figure 3C ) correlated with the level of in vitro cytotoxicity mediated by the corresponding T-BsAbs.

While several of our T-BsAbs were able to engage T cells to destroy tumor cells in vitro, the ability to do so in vivo should greatly enhance their clinical relevance. To determine in vivo anti-tumor efficacy we tested the EGFR (n=4), HER2 (n=4), and mesothelin (n=3) T-BsAbs in an intraperitoneal model of DSRCT (JN-DSRCT-1-luc), which more closely recapitulates human disease than a standard flank xenograft. Mice in these treatment groups were engrafted with human activated T cells (ATC) and given recombinant human IL-15/IL-15R-alpha to support T cell survival in vivo. ATC only (n=2) and irrelevant T-BsAb + ATC (CD19xCD3 BsAb, n=5) groups were included as negative controls. Figures 5A (showing comparisons between groups of 2-5 mice each) and 5C (showing tumor growth for all 18 mice individually) show that only the HER2 T-BsAb was able to control tumor growth (p = 0.02 compared to ATC only group). The EGFR T-BsAb performed no better than ATC alone (p = 0.28), and the mesothelin T-BsAb-treated group had one tumor that began to regress around day 40 but ultimately did not statistically differ from the ATC only group (p = 0.09). The HER2 T-BsAb completely ablated tumors by day 27 after tumor implantation, one week after treatment began ( Figure 5D ). The lack of in vivo anti-tumor efficacy by EGFR and mesothelin T-BsAbs despite their in vitro potency was surprising although similar negative results have been seen with other tumor targets, such as B7-H3 (data not shown). This discordance could result from position of the epitope being too distal from the cell membrane, insufficient antigen density or antibody affinity (34, 35). The mice did not lose weight during treatment ( Figure 5B ) and did not show any of sign of clinical toxicity until around day 60 when signs of graft versus host disease (GVHD - hair loss, skin reddening, weight loss) appeared, at which point the experiment was terminated and the animals were euthanized. This is a known consequence of engrafting activated human T cells into immunocompromised mice.

After observing the striking efficacy of the HER2 T-BsAb against JN-DSRCT-1-luc xenografts, we sought out to evaluate the efficacy of this T-BsAb in a faster-growing model of DSRCT (BER-luc). In this experiment, average total flux (photons/second) was around 10⁶ cells for BER-luc tumors at the beginning of treatment, versus 10⁴ cells with JN-DSRCT-1-luc in the previous experiment ( Figures 5A and 6A ). A total of 30 mice were used in this experiment. Four dose levels of HER2 T-BsAb were tested (0.3, 1, 3, and 10 μg per dose, n=5 mice per treatment level) and all tumors treated with at least 1 μg/dose began to shrink after the second injection of ATC and third dose of BsAb ( Figures 6A ). Figure 6C shows the tumor growth of each individual mouse. No treatment and irrelevant T-BsAb + ATC groups (CD19xCD3 BsAb) were included as negative controls (n=5 mice each). The lowest dose of HER2 T-BsAb had no effect on tumor growth at any point during the experiment compared to the negative control untreated group (p = 0.99). The two intermediate dose levels (1 μg and 3 μg/mouse) completely ablated tumors by day 49 after tumor implantation (p = 0.03 and 0.07, respectively, vs no treatment control), with no luciferase signal detectable by day 49 ( Figure 6D ). The highest dose level (10 μg/mouse) also shrunk established tumors (p = 0.04 vs no treatment control), however, 3/5 mice in this group still had detectable luciferase signal in their abdomens on day 49. None of the mice exhibited signs of toxicity and no significant weight loss was observed ( Figure 6B ). Around day 50 the mice treated with ATC began to exhibit signs of GVHD at which point the experiment was terminated and the animals were euthanized.

In an effort to evaluate the effects of different doses of HER2 T-BsAb on the activity of T cells in vivo, blood was taken from the mice 24 hours after the first administration of T-BsAbs and activated human T cells and ELISAs were performed on the serum. Interferon gamma (IFN-gamma) was detected in the blood of all mice treated with HER2 T-BsAb in a dose-dependent trend ( Figure 7D ). The overall levels of interleukin 2 (IL-2) were low and near the limit of detection for the assay, though IL-2 in the 10 μg HER2 T-BsAb group was significantly higher compared to the untreated group (p = 0.026) ( Figure 7A ). Only two mice, one in each of the highest dose groups, had detectable interleukin-6 (IL 6) in their serum at this timepoint ( Figure 7B ). Interleukin 10 (IL-10) was detectable in many of the mice treated with HER2 T-BsAb but was highest in the two highest dose groups ( Figure 7C ), with the 3 μg group being significantly higher compared to the no treatment control (p = 0.0003) and the 10 μg group approaching significance (p = 0.05). IFN-gamma was significantly elevated in the 1 μg, 3 μg, and 10 μg groups (p = 0.0016, 0.0006, and <0.0001, respectively). Finally, tumor necrosis factor alpha (TNFα) was significantly upregulated in the 10 μg HER2 T-BsAb group compared to the untreated control (p = 0.0039) ( Figure 7E ). It was detected at high levels in 3/5 mice in the 10 μg HER2 T-BsAb group but only detectable at very low levels in two other mice in different groups ( Figure 7E ). TNFα has been described as the “master of T cell exhaustion” and has been shown to influence T cells towards an exhausted phenotype (36). This finding may explain why 10 μg HER2 was less effective than 1 μg or 3μg. This phenomenon has been noted by other researchers studying biologic drugs and is the basis for the rational move away from determining the maximum tolerated dose (MTD) in early stage clinical trials and towards instead identifying the optimum biologic dose, which may well be lower than the MTD (37).