Fig 2: Isolation and characterization of scFvs targeting membrane proximal epitopes of mesothelin. (a) Amino acid alignment of mature, shed, and stump regions of human MSLN proteins. Homology of murine (Genbank AAH23753.1) and rabbit (NCBI Reference Seq XM_070064180.1) MSLN proteins to the human stump region is illustrated. (b) Canine B cell-derived scFv naïve/non-immune phage display library screening strategy. Positive/negative panning on full-length MSLN and shed-length MSLN proteins, respectively, enriched for scFvs targeting stump region of MSLN. Twenty-nine unique stump binders identified by nucleic acid sequencing from an initial cohort of 39 were further characterized as scFvs and a subset as T cell chimeric antigen receptors. (c) Competition ELISA assay used to rank ability of mAb15B6 reference MSLN stump-directed scFv to block binding of each of 29 stump-directed canine scFvs. Ratios of canine scFv binding to MSLN in the absence and presence of 15B6 are tabulated. The top 10 blocked scFvs (dark green shading) were chosen to be further screened as chimeric antigen receptors in order to narrow down the number of candidates for comprehensive evaluation to a total of 6. For 4 of the 29 scFvs, ELISA signals were too low to confidently determine binding ratios (“N.C.”). ScFv SH4-07, a scFv from library panning determined to be specific for the shed region of MSLN, served as a negative control for assay displaying equal binding to full-length MSLN whether 15B6 was present or not (ratio = 1). (d) Schematic of CAR construct inserted downstream of an EF1α-derived promoter within a 3rd-generation lentiviral vector plasmid comprising a CD28 transmembrane domain and 4-1BB and CD3ζ signaling domains. (e) Amino acid alignment of 15B6 reference MSLN stump-directed scFv with 6 canine stump-directed scFvs evaluated CARs in this study. All scFvs were used in the variable light chain/variable heavy chain (VL-VH) orientation. The 15B6 scFv sequence and linker were used as described by Liu et al. (22). The linker for canine scFvs was based on the pComb3X phage display vector system (30). Light chain and heavy chain CDRs were assigned according to IMGT nomenclature and identified using the Domain-GapAlign tool (55).

Fig 3: Stump CAR targeting and efficacy screen in healthy donor ND608 T cells. The cytotoxic in vitro activity of stump binding CAR T cells was assessed by titration at E:T ratios of 10:1, 3:1, 1:1, 0.3:1 and 0.1:1 as single values in 96-well plate co-cultures of tumor cell lines and evaluated by Incucyte® live cell imaging. (a) The cytotoxic activity of each CAR T cell was titrated against AsPC1 and H838 wild-type cells. Only the 1:1 E:T ratio is shown to highlight differences. A qualitative evaluation was used to select best killing based on a diminished tumor area. Another donor showed similar results. (b) To eliminate stump binders with non-MSLN specific killing, CAR T cells were co-cultured with AsPC1 and H838 cell lines from which MSLN was knocked out (KO). Only the 10:1 ratio is shown where maximum off-target activity can be seen. For both cell lines, binders ST3-09, ST4–27 and ST3–14 revealed non-MSLN specific killing and were eliminated from further evaluation. (c) To assess stump specificity, CAR T cells were co-cultured with AsPC1 and H838 MSLN-KO cells expressing lentivirus-transduced GPI-linked stump peptide. To highlight differences, only 1:1 E:T ratio results are shown and reveal ST4-11, ST4-22, ST4-39, and 15B6 show specificity for MSLN stump peptide. The non-MSLN specific binders have similar activity at 1:1 as they did with wild type cells. Another T cell donor gave similar results. (d) Stump binders were assessed in vivo to determine rate of tumor clearance and the persistence of response in an AsPC1 flank solid tumor model in NSG mice with 5 mice per group. When 2e6 AsPC1 cells were injected subcutaneously into the right flank with Matrigel and reached ~150mm3 in volume (~14 days after implantation), 1e6 CAR+ T cells were injected IV. Tumor volumes were plotted as mean with SEM and survival as Kaplan-Meier curve. (e) Stump binders were assessed in vivo to determine rate of tumor clearance and the persistence of response in an OVCAR8 peritoneal metastasis model in NSG mice with 5 mice per group. Fourteen days after intraperitoneal injection of 3e6 OVCAR8 tumor cells expressing click beetle red (CBR) luciferase for in vivo monitoring, 1e6 CAR+ T cells were injected IV. Weekly tumor BLI is plotted as mean with SEM and survival plotted as Kaplan-Meier curve. (f) Summary of initial in vitro and in vivo functional screens of MSLN stump-directed CAR T cells.

Fig 4: Schematic of mesothelin. Cartoon showing mesothelin and its GPI linkage, and binding sites for SS1, M5, and 15B6 antibodies. SS1 and M5 bind at sites distant from the MSLN proteolytic cleavage sites and can also bind to shed MSLN. The 15B6 scFv binds to cell-associated MSLN stump near the plasma membrane which remains after MSLN cleavage. Modified from Ma et al. (25).