A re-engineered immunotoxin shows promising preclinical activity in ovarian cancer

Abstract

RG7787 is a re-engineered mesothelin-targeted immunotoxin with reduced immunogenicity composed of a humanized anti-mesothelin Fab fragment and a B-cell epitope silenced 24 kD fragment of Pseudomonas exotoxin A. High prevalence of mesothelin-positive cases and a large unmet medical need make ovarian cancer a promising indication for the clinical development of RG7787. However, ovarian cancer patients also frequently have elevated serum levels of the cancer antigen 125 (CA-125). In principle this could pose a problem, since the binding sites for CA-125 and RG7787 on mesothelin were reported to overlap. However, we show here that RG7787 can readily displace even excess amounts of CA-125 in different cellular assays. Moreover when tested in-vitro on a panel of 12 ovarian cancer cell lines, RG7787 had high cytotoxic activity on COV644, Caov-4, and SNU-119 cells and fully inhibited growth of EFO-21, KURAMOCHI, OVSAHO, and Caov-3 cells with potency values ranging from 1 to 86 pM. Finally, we evaluated the in-vivo efficacy of RG7787 in OvCa6668, a patient-derived ovarian cancer model with high levels of CA-125 expression. RG7787 had moderate monotherapy efficacy but in combination with standard chemotherapies (cisplatin, paclitaxel) achieved pronounced tumor regressions. In summary our data support clinical testing of RG7787 in ovarian cancer.

Figure 1

Expression of CA-125 and MSLN in ovarian cancer. The ovarian serous cancer cohort of The Cancer Genome Atlas (TCGA) was analysed for mesothelin (MSLN) and CA-125 expression. Log2 normalized read counts for CA-125 are plotted on the x-axis and for MSLN on the y-axis.

Figure 2

CA-125 competes with, but cannot displace anti-MSLN FAB from its epitope. (a) SPR assay schematic illustrating capture of the human MSLN extracellular domain as Fc fusion protein via an anti-huFc antibody coupled to the sensorchip. (b) No significant dissociation was observed (dark blue line, buffer injection). Both CA-125-HisAvi (purple line) and anti-MSLN Fab (green line) can bind to the captured MSLN-ECD but due to steric hindrance CA-125 reaches only partial saturation of binding sites, whereas the Fab reaches ~100% saturation as confirmed by no further signal increase upon a 2nd injection of a higher Fab concentration (light blue line). When CA-125 is offered in a second injection after saturation with the FAB (red line), no further binding occurs demonstrating that binding of both molecules to mesothelin is mutually exclusive.

Figure 3

Potency of RG7787 in the presence of excess of CA-125. (a) Prior to the assay NCI-H226 cells were preincubated overnight with or without pleural fluid, which contained 29,500 U/ml of CA-125 as determined by Elecsys CA-125 measurement. Serial dilutions of RG7787 or a non-MSLN targeting control Fab-PE molecule with cytotoxic potency for NCI-H226 cells were added for 72 h. Cell viability was measured with CellTiter-Glo. Calculated IC50s with confidence intervals are as follows: RG7787 no CA-125: 0.042 nM (95% CI [0.0374, 0.0467]) versus RG7787 with CA-125: 0.058 nM (95% CI [0.0522, 0.0636]) and Fab-PE no CA-125: 0.013 nM (95% CI [0.0127, 0.0141]) versus Fab-PE with CA-125: 0.0176 nM (95% CI [0.0171, 0.0181]). (b) NCI-H226 cells were preincubated overnight with or without 28,000 U/ml of recombinant CA-125 as determined by Elecsys CA-125 measurement. The time that cells were exposed to RG7787 or a non-MSLN targeting control Fab-PE molecule was varied. For the 72 h time point the PE molecules were incubated with the cells for the duration of the experiment. For shorter exposure times the cells were incubated with the PE molecules for 40, 80, or 160 min and then the medium was removed and fresh medium was added for the remainder of the 72 h assay period. Cell viability was measured with CellTiter-Glo. The IC50s for RG7787 without and with CA-125, respectively, in ascending order of incubation length are: 40 min: 0.145 nM (95% CI [0.1185, 0.1522]) and 0.227 nM (95% CI [0.153, 0.2486]); 80 min: 0.064 nM (95% CI [0.0502, 0.0748]) and 0.127 nM (95% CI [0.1089, 0.1388]); 160 min: 0.044 nM (95% CI [0.0386, 0.0494]) and 0.095 nM (95% CI [0.0824, 0.1047]); and 72 h: 0.017 nM (95% CI [0.01482, 0.019]) and 0.044 nM (95% CI [0.0372, 0.0505]).

Figure 4

Heterotypic adhesion assay. (a) FACS quantification of mesothelin surface levels was done using humanized SS1P IgG and an anti-human Alexa Fluor 647 secondary antibody. While the FACS graph for parental H358 cells overlaps with that of the isotype control, increasing mesothelin surface staining was observed for the stably transfected clones 4D8 < 4H3 < 2C5 < IG2 (b) On day 1, 1 × 10⁶ OVCAR3 cells were seeded per well of a 12-well dish. 1 × 10⁷ of the parental and NCI-H358-MSLN clones were stained with 2.5 µM CFSE for 10 min at 37 °C and allowed to recover overnight. On day 2, the NCI-H358 cells were collected from the plates and 1 × 10⁶ cells were pre-incubated with either anti-MSLN Fab or a control Fab for 10 min. The NCI-H358 cells were then added to the OVCAR3 cells and interaction was accelerated by a brief centrifugation. After stringent washing, all of the remaining cells in the well were harvested and analyzed by FACS for CFSE positivity. (c) Bar graph depicting the relative percentages of CFSE labeled NCI-H358 cells vs unlabeled OVCAR3 cells.

Figure 5

IHC staining of OvCa6668 tumors. Formalin fixed OvCa6668 tumors were stained for MSLN (a) or CA-125 (b). For both proteins predominantly an apical staining pattern is observed. For MSLN 80% of tumor cells are positive (staining intensity +++). For CA-125 apical membranes stain positive (staining intensity ++−+++) also in 80% of tumor cells, while in 30% weaker cytoplasmic staining is observed.

Figure 6

Efficacy of RG7787 as monotherapy and in combination with standard chemotherapies in the OVCa6668 patient-derived ovarian cancer model. OvCa6668 xenograft tumors were inoculated in NMRI nu/nu female mice. One day after randomization, treatment was started with either paclitaxel (6 mg/kg iv; q1dx5) or RG7787 (2 mg/kg iv; q2dx3 2 days off, q2dx3 2 days off, plus 7th treatment) or cisplatin (2 mg/kg ip; q7dx2) or combinations of various agents. The mean tumor volumes were plotted over time. Treatment time points are indicated by arrows and triangles. (a) Tumor growth curves are shown for combination therapy with RG7787 plus paclitaxel and the respective monotherapies and vehicle control group. From day 18 onwards the mean tumor volumes of the combination therapy group were substantially reduced as compared to the effect calculated using the bliss additivity model (dashed line). (b) Tumor growth curves are shown for combination therapy with RG7787 plus cisplatin and the respective monotherapies and vehicle control group. The mean tumor volumes of the combination therapy group begin to slightly separate from the calculated tumor growth curve based on the bliss additivity model (dashed line) only after day 23. (c) Box and whisker plots for the tumor volumes observed for the different treatment groups on day 30 (end of study date). For both combination therapy groups, asterisks mark where the mean tumor end volume based on the bliss additivity calculation lies relative to the 95% confidence interval of the actual measurements (represented by the box).