Fully human monoclonal antibody targeting activated ADAM10 on colorectal cancer cells

Abstract

Metastasis and chemoresistance in colorectal cancer are mediated by certain poorly differentiated cancer cells, known as cancer stem cells, that are maintained by Notch downstream signaling initiated upon Notch cleavage by the metalloprotease ADAM10. It has been shown that ADAM10 overexpression correlates with aberrant signaling from Notch, erbBs, and other receptors, as well as a more aggressive metastatic phenotype, in a range of cancers including colon, gastric, prostate, breast, ovarian, uterine, and leukemia. ADAM10 inhibition, therefore, stands out as an important and new approach to deter the progression of advanced CRC. For targeting the ADAM10 substrate-binding region, which is located outside of the catalytic domain of the protease, we generated a human anti-ADAM10 monoclonal antibody named 1H5. Structural and functional characterization of 1H5 reveals that it binds to the substrate-binding cysteine-rich domain and recognizes an activated ADAM10 conformation present on tumor cells. The mAb inhibits Notch cleavage and proliferation of colon cancer cell lines in vitro and in mouse models. Consistent with its binding to activated ADAM10, the mAb augments the catalytic activity of ADAM10 towards small peptide substrates in vitro. Most importantly, in a mouse model of colon cancer, when administered in combination with the therapeutic agent Irinotecan, 1H5 causes highly effective tumor growth inhibition without any discernible toxicity effects. Our singular approach to target the ADAM10 substrate-binding region with therapeutic antibodies could overcome the shortcomings of previous intervention strategies of targeting the protease active site with small molecule inhibitors that exhibit musculoskeletal toxicity.

Keywords: ADAM10; COLO205; Chemotherapy; Colorectal cancer; Monoclonal antibody; Notch; Xenograft.

Fig. 1.

(A) SDS-PAGE profile of the purified bovine (b) ADAM10 disintegrin + cysteine-rich domain construct (D+C), used as an antigen to screen a large naïve human Fab library. (b) indicates bovine protein, (h) indicates human protein. (B) Characterization of fully human 1H5 Fab. Binding profile of individual Fab binders (left) and binding of 1H5 Fab (middle) to ADAM10 D+C. Competitive ELISA with the murine 8C7 antibody demonstrating that 1H5 Fab (15 nM fixed concentration) binds to a similar ADAM10 epitope (right). (C) Pull-down experiment showing that ADAM10 D+C (b) binds to protein A Sepharose bead-bound murine 8C7 and human 1H5. Lane 1: Low molecular weight standards (Bio-Rad). Lane 2: bead bound murine 8C7 IgG + ADAM10 D+C. Lane 3: bead bound human 1H5 IgG + ADAM10 D+C. Lane 4: bead bound ADAM10 D+C without any pre-bound mAb. Lane 5: ADAM10 D+C input. (D) Human 1H5 IgG binds specifically to immobilized human (h) and bovine (b) ADAM10 D+C, but not to human ADAM17 D+C or human ADAM19 D+C. Wells were coated with the ADAM D+C constructs and ELISA was performed as described in the Materials and Methods.

Fig. 2.

Alamar blue cell viability assays with multiple cancer cell lines. Percent growth inhibition is shown for (A) colon (COLO205, LIM1215), (B) breast (MDA-MB-231, SKBR-3), (C) ovarian (SKOV-3, OVCAR-3) and (D) glioblastoma (U87-MG) and non-small cell lung cancer (HCC-827) tumor cell lines treated with human 1H5. The data represent mean of triplicate determinations and two independent experiments, and the bar plots show the effect of treatment of mAbs on cancer cells relative to the control, mean ± SEM (described in Materials and Methods).

Fig. 3.

1H5 inhibits Notch cleavage. Sandwich ELISA was used to measure the levels of total (A) and cleaved (B) Notch1 in COLO205 cells upon treatment with 1H5. The data represent mean of triplicate experiments, and the bar plots show the effect of treatment with 1H5 relative to untreated control, mean ± SEM. Comparison of notch levels between treated and untreated groups was performed using independent t test. Total Notch1 levels did not significantly differ between the two groups (A), p = 0.162. On the other hand, the mAb-treated group showed significant decrease of the cleaved Notch1 levels when compared with the untreated control (B), p < 0.001.

Fig. 4.

Anti-tumor effect of 1H5 in a colon cancer xenograft model. 6–8 weeks old female athymic nude mice were used. 5 million COLO205 cells per mouse were implanted (subcutaneous with Matrigel). Four groups were used (n = 4) for the experiment. When tumor volumes reached ~100 mm3, the four groups were injected as follows: (Group 1) sterile PBS (as a control). (Group 2) Irinotecan (i.p.) 20 mg/kg, three doses, once a week starting day 12. (Group 3) 1H5 (i.p.) at a dose of 30 mg/kg, biweekly (total 7 doses) starting day 7. (Group 4) Irinotecan (20 mg/kg, i.p., once a week starting day 12, total three doses) + continued 1H5 treatment (30 mg/kg, i.p., biweekly, total 7 doses). (A) Mean tumor volume ± 1 SD from day 7–35. 83 % tumor growth inhibition was recorded for the Group 4, treated with Irinotecan+ 1H5, 54 % for mAb alone (Group 3) and 48 % for Irinotecan alone (Group 2). No change in mouse weight or presence of diarrhea was observed in any of the groups. (B) Independent t test analysis showed significant difference in tumor volume reduction (day 35) between PBS and all treatments (with mAb and chemotherapy alone: p = 0.002; with combined treatment: p < 0.001). Significant tumor volume reduction differences were also observed between chemotherapy alone and combined treatment (p < 0.001), and between mAb treatment alone and combined treatment (p = 0.004). ANOVA with Dunnett multiple comparison post-hoc analysis likewise showed statistically significant tumor volume reduction between all treatment regimens vs. PBS control (all are p < 0.001), as well as between combination therapy and monotherapy (vs. chemotherapy p = 0.004; vs. mAB p = 0.011). In both analyses, the difference between mAb treatment and chemotherapy was not statistically significant. In the box and whiskers plot, the black horizontal lines indicate the average value, the top and bottom of the boxes, the interquartile range, and the whiskers, the range. Color coding on Panel B is the same as on Panel A.

Fig. 5.

Crystal structure of the 1H5/ADAM10 (D+C) complex. (A). Overall structure of the ADAM10(D+C)/1H5 complex shown in ribbon view, as well as binding interface comparison (zoom-in insets) with the ADAM10(D+C)/8C7 structure showing a similar epitope region with distinct recognition strategies (see also Fig. S2 for epitope comparison). (B). Superimposition of the mAbs, 1H5 and 8C7, in complex with the ADAM10 D+C domain, illustrating the different antibody approaching angles. ADAM10 is shown as ribbons (green: 1H5 bound; grey: 8C7 bound). The antibodies are shown as molecular surfaces (1H5: nontransparent orange and blue; 8C7: semi-transparent yellow and cyan). (C) Overlay of the ADAM10 D+C (in green) / 1H5 (in orange and blue surface representation) complex and the ADAM10 ECD (in magenta) structures showing a partial overlay (indicated with the black circle) of 1H5 with the M domain in the autoinhibited ADAM10 conformation.

Fig. 6.

1H5 preferentially binds to the activated, tumor cell-specific conformation of ADAM10. Cellular ELISA was performed to gauge the binding of 1H5, relative to the binding of the control anti-ADAM10 mAb1427 (commercial, R&D systems) to ADAM10 expressed on the cell surface of colon cancer cell lines LIM1215, COLO205, as well as HEK293 cells and HEK293 cells transfected with human ADAM10. MAb1427 bind equally well to the activated (tumor-associated) and the autoinhibited conformation of ADAM10^,. The graph show the 1H5/mAb1427 signal ratio observed for the noted cell line relative to the 1H5/mAb1427 signal ratio observed for the untransfected HEK293 cells. Specifically, on the Y axes is plotted the value of: (A(1H5)/A(mAb1427))/(A(1H5-HEK)/A(mAb1427-HEK)) where A(1H5-HEK) is the signal for 1H5 using the untransfected HEK293 cells; A(mAb1427-HEK) is the signal for the mAb1427 using the untransfected HEK293 cells; A (1H5) is the signal for 1H5 using the cells that are being evaluated; A(mAb1427) is the signal for mAb1427 using the cells being evaluated. 1H5 binds to ADAM10 on tumors approximately fourteen-folds better than to ADAM10 on HEK2923 cells. The data represent triplicate determinations and two independent experiments, mean ± SEM; P < 0.001 by unpaired two-tailed independent t test (cancer cell lines vs. HEK293 cells).

Fig. 7.

(A) SDS-PAGE profile of purified human (h) and bovine (b) catalytically active ADAM10 extracellular domains (ECD). (B and C) FRET-based peptide cleavage assays. The data shows that 1H5, similar 8C7 (but even more efficiently), promotes cleavage of a short peptide substrate by ADAM10. The data represent mean of triplicate determinations and two independent experiments (mean fluorescence ± SEM). Maximum dispersion was within 10 % of the mean value. Bovine or human ADAM10(ECD)-antibody complexes were formed at 1:1 molar ratio prior to the assay, and the assay was carried out in the presence of 50 μM of a fluorogenic peptide as described in the Materials and Methods section. GM6001 is a MMP inhibitor, which was also used as a control. Statistical analysis revealed significant difference/increase in peptide cleavage (fluorescence) between the 1H5 and 8C7 treated and untreated ADAM10 samples (p < 0.001, for both 1H5 and 8C7 treatment, one way ANOVA with a Dunnett’s multiple comparison post hoc test), as well as between the 1H5-treated and 8C7-treated ADAM10 samples (p = 0.003, two-tailed independent t test). Treatment with the control mAb1427 or the GM6001 inhibitor did not significantly change the ADAM10 enzymatic activity (p > 0.05).

Fig. 8.

Schematic representation of a proposed mechanism for ADAM10 activation and interactions with substrates and the 1H5 antibody. In the autoinhibited conformation, which is the predominant conformation of the ADAM10 ectodomain observed in solution [31], the cysteine rich domain partially occludes the ADAM10 active site, hindering substrate binding. In the open ADAM10 conformation, the active site is fully accessible for substrate binding [39]. In addition to interacting with the active site, ADAM10 cell-surface substrates also interact with the D+C region of the molecule, which is required for substrate selection and/or proper substrate positioning for productive cleavage. The open ADAM10 conformation can be stabilized by binding of substrates, such as Notch, or other regulatory molecules, such as tetraspanins. The latter not only regulate the activity of ADAM10 by stabilizing the open conformation, but also selectively enhance the cleavage of certain substrate, while downregulating the cleavage of other substrates [2,3,14,31,39]. The 1H5 mAb, in this regard, acts in a similar fashion to the tetraspannins: it (i) selectively binds and stabilizes the activated, open ADAM10 conformation; and (ii) it selectively downregulates the cleavage of certain ADAM10 substates (Notch), while upregulating (or not affecting) the cleavage of other substrates (e.g., peptides, APP).