Co-expression of the metabolic GOT2 with a GPC3-targeted CAR-T overcomes the challenges of the solid tumor microenvironment, substantially improving therapeutic efficacy in solid tumor xenografts Kathleen R. Whiteman, Tapasya Pai, Eugene Choi, Taylor Hickman, Tyler Johnson, Taylor Friedman, Luke #P227 Barron, Madaline Gilbert, Binzhang Sheng, Seth Ettenberg, Kathleen McGinness, and Greg Motz Introduction Results The metabolic demands of cancer cells in the solid tumor microenvironment (TME) create an GOT2 overexpression increases critical metabolites in T cells BOXR1030 activity is superior to a control CAR in TME-like conditions BOXR1030 T cells resist TME-driven dysfunction unfavorable T cell environment through depletion of critical nutrients and amino acids and Standard conditions Standard conditions a) b) c) accumulation of waste products. This drives T cell dysfunction and inhibits the effectiveness of Standard conditions CD8+ PD1+Tim3+ CD4+ PD1+Tim3+ TIL cytokines a) b) c) 60 1.5 immunotherapies. To overcome these and other TME challenges, we developed the BOXR (bolt-on Mock 60 GPC3 CAR a) 100 8000 200000 chimeric receptor) platform in which engineered T cells co-express both a chimeric-targeting receptor GPC3 CAR p=0.007 BOXR1030 and a “bolt-on” transgene. In a screen of 100+ for enhanced T cell function when co- 75 6000 150000 BOXR1030 40 1.0 50 40 expressed with an anti-glypican-3 (GPC3) CAR, we identified the first candidate of our BOXR potential 4000 100000 (pg/mL) GSH γ platform, BOXR1030, which co-expresses the transgene glutamic-oxaloacetic 2 GSSG 25 IFN 50000 % Cytotoxicity 2000 p=0.015 (GOT2), a critical enzyme involved in mitochondrial metabolism. Here, we present preclinical CD3+ T Cell Count Glutamate Malate p<0.001 Malate NADPH NADP+ 0 Percent (%) 20 0.5 0 0 20 per T cell (pg/mL)

characterization of the mechanism of action of BOXR1030. Percent (%) HepG2 (GPC3+) NCI-H929 (GPC3-) HepG2 (GPC3+) NCI-H929 (GPC3-) HepG2 (GPC3+) NCI-H929 (GPC3-) γ GOT2 GOT2 IFN TCA Chronic stimulation in low Hypoxia Chronic stimulation in low glucose Hypoxia ⍺KG OAA Aspartate Aspartate 0.0 The BOXR platform cycle 0 0 GPC3 BOXR1030 GPC3 BOXR1030 d) e) GPC3 CAR f) 0.0020 g) 0.0010 2 7 14 0.004 GPC3 CAR 0.0015 p=0.0062 CAR CAR Critical biosynthetic pathways p=0.025 Day p<0.001 p=0.001 0.0008 BOXR T cells overcome immune suppression in the TME AAs, proteins, nucleotides 0.0015 Citrate 0.003 Mice bearing JHH7 tumors were dosed with either a control GPC3 CAR or BOXR1030 T cells. T cells were isolated and analyzed by flow Metabolites directly produced by GOT2 0.0010 0.0006 cytometry or were evaluated ex vivo for cytokine production. (a) CD8+ T cells over time. Data are mean of 5 animals. (b) CD4+ T cells on day 0.0010 0.002 14. Data are mean of 5 animals. (c) Cytokine production from ex vivo cultured T cells. Data are mean of 3 animals. BOXR components 0.0004 0.0005 0.0005 chimeric receptor ‘bolt-on’ 0.001 0.0002 (ACTR, CAR, TCR) § chimeric receptor: Tumor targeting modality (CAR, RNA transcriptional profile of BOXR1030 is consistent with reduced

transgene Proliferation (1/gMFI) Proliferation (1/gMFI) Proliferation (1/gMFI) ACTR, TCR, etc.) drives cancer cell targeting and Proliferation (1/gMFI) 0.0000 0.0000 0.000 0.0000 exhaustion and reduced metabolic stress immunosuppressive cells attack GPC3 BOXR1030 GPC3 BOXR1030 Standard Low Glc Normoxia Hypoxia (Treg, MDSC) CAR CAR a) b) § bolt-on: novel transgene re-programs T cell biology % significant GO ID Process name P adj value to improve functionality in the tumor genes Control GPC3 CAR or BOXR1030 T cells were incubated with either GPC3+ HepG2 or GPC3-negative NCI-H929 tumor cells. 24 hours later a) TOX2 microenvironment c) e) GO:0006986 response to unfolded protein 19.05 1.32E-10 BOXR T cell b) cytotoxicity and b) cytokine production was measured. c) In a separate assay with similar conditions, T cell proliferation was measured on day 800 secreted factors 4 7. Representative data from a single donor. In a separate study, control GPC3 CAR T cells were labeled with cell trace violet, and incubated GO:0042026 protein refolding 40.00 1.50E-09 metabolites Targeting key mechanisms of immunosuppression with GPC3+ Hep3B tumor cells. d) To mimic chronic stimulation under TME-like conditions, T cells were cultured in low glucose (2mM) and GO:0070370 cellular heat acclimation 75.00 2.95E-05 ) 600 restimulated with targets 3 days after initial incubation or e) co-cultured in hypoxia (1.5% O2). Similarly, control GPC3 CAR and BOXR1030 T positive regulation of interleukin-23 § Competition for metabolites 3 GO:0032747 40.00 3.23E-03 cells were compared under identical conditions f) and g). Following 7 days, CAR+CD3+ T cells were evaluated by flow cytometry for loss of production padj § Immunosuppressive cells (MDSC, Treg) GPC3 CAR BOXR1030 tumor cells cell trace violet signal as a readout of T cell proliferation. Data are presented as means of 11 unique donors. GO:0010941 regulation of cell death 22.22 6.37E-03 2 400 log10( § Exhaustion due to chronic stimulation positive regulation of cytokine production - GOT2 GO:1900017 22.22 6.37E-03 involved in inflammatory response 1 200 BOXR1030 T cells have superior antitumor activity in CAR-resistant chaperone mediated protein folding GSH:GSSG ratio GO:0051085 20.00 7.27E-03 requiring cofactor mU/mL AST at 50 mins GAPDH xenograft models chaperone-mediated protein complex 0 0 GO:0051131 15.38 1.04E-02 HepG2 b) c) assembly GPC3 BOXR1030 GPC3 BOXR1030 a) 200000 32 1500 positive regulation of tyrosine CAR CAR (HCC) GO:0042523 12.50 1.31E-02

) 5mM phosphorylationlog2(Fold Change) of Stat5 protein b) Discovery of BOXR1030 3 log2(Fold Change) 16 150000 1000 TOX2 expression over time in T cells TOX2 expression at 4 hours in CD4+ and CD8+ T cells Screening of 100+ of bolt-ons led to the discovery of BOXR1030 8 c) 4hr 24hr d) CD4+ CD8+ f) 100000 5 mM d) Alanine untreated 600 800 BOXR1030 4 GSH 500 Glucose (mM) 2 50000 600

a) b) CD3+ T Cell Count Aspartate 2 Ala 400

Asp (mm volume Tumor GPC3 CAR > 100 genes testing literature-based hypotheses Met Gln 0 1 400 0 p=0.04 Gly 0 20 40 60 80 100 0.01 0.1 1 10 100 p=0.034 1 Glu GOT2 Blood JHH7 200 Ser Glucose Concentration (mM) HepG2 Hep3B p=0.06 p=0.029 Co-express CAR + bolt-on transgene NCI-H446 Val Time (Day) 200 Thr Normalized Counts Normalized Counts Lys d) e) f) Asn Assess basic T cell function 2000 JHH7 Hep3B 1500 NCI-H446 ControlUntreated 0 0 0.5 Pro 1500 GPC3 BOXR1030 GPC3 BOXR1030 GPC3 BOXR1030 GPC3 BOXR1030 His (HCC) (HCC) (SCLC) ) GPC3 CARCAR CAR CAR CAR CAR ) )

Phe 3 Proliferation Cytokines Cytotoxicity 3 Trp 3 GPC3BOXR1030 CAR + bolt-on Control GPC3 CAR and BOXR1030 T cells from 2 unique donors were co-cultured with immobilized GPC3 protein for 4 and 24 hours, Leu 1500 Tyr followed by RNA extraction. RNA was then analyzed by RNA-Seq. In a parallel experiment, CD4 and CD8 T cells from a single donor were 1000 Assess function in simulated solid TME GPC3 CAR Arg 1000 isolated for RNA analysis. a) At 4hrs, a number of GO processes were differentially expressed with high significance. Red=immune response 0.25 GSSG Ile genes; blue=cell stress genes b) At 4hours, many genes were differentially expressed and genes associated with significantly different GO Secreted T cell Nutrient humanized anti-GPC3 BOXR1030/GPC3 CAR (fold) Chronic scFv: tumor targeting 1.0 1.1 1.2 1.3 1.4 1.5 1000 processes are indicated by color. Shown is data from an individual donor. c) Notably, a key driver of T cell exhaustion, TOX2, was stimulation inhibitors shortages extracellular Relative Amount (BOXR1030/GPC3 CAR) substantially lower in BOXR1030 from both donors at 4 and 24 hours. d) 4 hours after stimulation, TOX2 expression was substantially 500 500 lower in BOXR1030 T cells in both CD4+ and CD8+ T cell subsets. Regulatory Suppressive intracellular Hypoxia 4-1BB: co-stimulation T cells myeloid cells 0.125 500 Tumor volume (mm volume Tumor Tumor volume (mm volume Tumor CD3ζ: TCR signaling (mm volume Tumor Conclusions 0 0 0 Assess activity in stringent xenograft models 0 10 20 30 40 50 0 20 40 60 80 100 10 30 50 70 Unum’s BOXR T cell platform is a novel approach to discover transgenes that overcome the challenges faced a) Depiction of biochemical steps related to GOT2 activity. b) Western blot analysis of a control GPC3 CAR and BOXR1030 T Time (Day) Time (Day) Time (Day) by T cells in solid tumors, and led to the discovery of BOXR1030. BOXR1030 is a GPC3-targeted CAR that a) BOXR candidates were generated by cloning a library of literature-derived, hypothesis-driven bolt-on genes into vectors cells for GOT2 following activation with Hep3B GPC3+ target cells. GADPDH was used as a loading control. c) Aspartate co-expresses GOT2. The addition of GOT2 led to improved metabolic and transcriptional profiles associated containing a GPC3-targeted CAR that had either a 4-1BB or CD28 co-stimulation domain. BOXR constructs were screened aminotransferase activity of BOXR1030 compared to a control GPC3 CAR. Lysates were evaluated following activation with with superior activity in the face of diverse TME challenges in vitro and in vivo. These results demonstrate Hep3B GPC3+ target cells. d) A targeted panel of 116 metabolites involved in energy metabolism was evaluated by CE-MS through Unum’s novel solid tumor microenvironment assays, followed by assessment of activity in stringent xenograft models. that engineering of T cell immunometabolism is an effective and potent strategy to overcome the challenges Candidates were selected on their ability to overcome multiple TME-challenges, while maintaining specificity and tolerability. for both BOXR1030 and a control GPC3 CAR. Means of metabolite values (pmol/10e6 cells) from 2 unique donors were a) Mice bearing HepG2 tumors were treated with control GPC3 CAR T cells. b) Proliferation of a control GPC3 CAR was measured by flow BOXR1030 was selected from this screening. b) BOXR1030 is a GPC3-targeted CAR that contains a humanized scFv, a 4-1BB co- pooled and plotted as a ratio of BOXR1030 relative to a GPC3 CAR. Metabolites of interest are highlighted. e) The ratio of cytometry following co-culture with GPC3+ JHH7 target cells in a dose-titration of glucose. c) Interstitial free glucose was measured from a of the solid tumor microenvironment and prevent T cell dysfunction in the TME. IND-enabling studies with stimulation domain and co-expresses the transgene glutamic-oxaloacetic transaminase 2 (GOT2), a critical enzyme involved in glutathione (GSH) to glutathione disulfide (GSSG) is depicted from the metabolite analysis described. f) Amino acids detected panel of GPC3+ xenograft tumors. Mice bearing d) JHH7 e) Hep3B or f) NCI-H446 xenograft tumors were treated with either control GPC3 BOXR1030 are underway with the expectation that BOXR1030 will be evaluated clinically in the treatment of from the metabolic profile (cysteine below the limit of detection). cellular metabolism. CAR T cells, or BOXR1030 T cells. Data shown as means (n=5/group). GPC3+ malignancies.