Targeting autocrine HB-EGF signaling with specific ADAM12 inhibition using recombinant ADAM12 prodomain

Abstract

Dysregulation of ErbB-family signaling underlies numerous pathologies and has been therapeutically targeted through inhibiting ErbB-receptors themselves or their cognate ligands. For the latter, "decoy" antibodies have been developed to sequester ligands including heparin-binding epidermal growth factor (HB-EGF); however, demonstrating sufficient efficacy has been difficult. Here, we hypothesized that this strategy depends on properties such as ligand-receptor binding affinity, which varies widely across the known ErbB-family ligands. Guided by computational modeling, we found that high-affinity ligands such as HB-EGF are more difficult to target with decoy antibodies compared to low-affinity ligands such as amphiregulin (AREG). To address this issue, we developed an alternative method for inhibiting HB-EGF activity by targeting its cleavage from the cell surface. In a model of the invasive disease endometriosis, we identified A Disintegrin and Metalloproteinase 12 (ADAM12) as a protease implicated in HB-EGF shedding. We designed a specific inhibitor of ADAM12 based on its recombinant prodomain (PA12), which selectively inhibits ADAM12 but not ADAM10 or ADAM17. In endometriotic cells, PA12 significantly reduced HB-EGF shedding and resultant cellular migration. Overall, specific inhibition of ligand shedding represents a possible alternative to decoy antibodies, especially for ligands such as HB-EGF that exhibit high binding affinity and localized signaling.

Figure 1. Computational model of autocrine signaling accurately predicts ligand capture differences between AREG and HB-EGF.

(a) Schematic of computational model that incorporates receptor trafficking combined with the proteolytic release, spatial diffusion, and receptor-mediated capture of growth factor ligands. (b) Ligand capture depends on ligand-receptor binding affinity, computationally modeled and represented here as the cumulative fraction of proteolytically released ligand that is captured by EGFR over the course of 24 h. (c) The computational model accurately predicts that HB-EGF is captured at higher levels compared to AREG. The fraction of bulk free ligand, equivalent to (1 - [fraction ligand capture]), was experimentally measured by taking the ratio of supernatant ligand concentrations after 24 h with or without the EGFR blocking antibody, mAb225 (n = 2 ± SEM).

Figure 2. Ligand properties significantly influence decoy antibody efficacy.

(a,b) Computational modeling shows that increasing ligand binding affinity (a) or decreasing ligand diffusion (b) substantially reduce the ability of decoy antibodies to effectively sequester ligands and prevent them from binding to surface EGFR. Modeling results were obtained after simulating 24 h decoy antibody treatment. (c) HB-EGF and AREG exhibit differential sensitivity to decoy antibodies, based on computational modeling that incorporated measured ligand release and receptor levels in 12Z endometriotic cells, along with known differences in HB-EGF and AREG binding affinity to EGFR. Dashed line marks 10 μg/mL, the experimentally tested concentration of decoy antibodies. (d) Using 12Z, computational estimates of ligand-EGFR complex levels at 24 h post-treatment correlate with cellular migration observed over the course of 24 h in collagen I gels. Note neither the model nor the cell migration measurements show HB-EGF decoy antibody (α-HBEGF) to have a significant effect, in contrast to the AREG decoy antibody (α-AREG). Genetic HB-EGF knockdown confirms a role for HB-EGF in cell migration, and complete inhibition of ErbB signaling using dacomitinib serves as a positive control (n ≥ 3 ± SEM; *p < 0.05, two-tailed t-test, reduction in cell migration).

Figure 3. Cue-signal-response data set relates protease activities to ectodomain shedding in endometriosis cell culture.

(a) Overview schematic of cue-signal-response modeling approach to infer biochemical relationships between exogenously applied signaling cues, measured ADAM proteolytic activity “signals,” and corresponding ADAM-substrate shedding “responses.” (b) Serum-starved 12Z cells were treated with growth factor/cytokine “cues” for 3 h and simultaneously monitored in real-time for live-cell protease activity “signals” using FRET-based polypeptide probes. Left: Specific ADAM activities were then computationally inferred using the PrAMA algorithm (n = 4 ± SEM). Right: 24 h later, supernatants were analyzed using ELISA for endogenous ADAM-substrate accumulation “responses” (n = 3 reps ± SEM). (c) Pairwise Pearson correlations between protease activities ((b), left) and ADAM-substrate shedding ((b), right) were calculated for measurements as they varied across the panel of growth-factor/cytokine treatment conditions; results describe correlational relationships between patterns of inferred ADAM catalytic activity and downstream substrate proteolysis. EGF-ligand correlations were performed only for the n = 8 growth-factor treatment conditions in the presence of mAb225 to block ligand endocytosis, while receptor correlations were calculated with and without mAb225 (n = 16 conditions from n ≥ 3 reps; p-value from two-tailed t-tests).

Figure 4. ADAM12 inhibition using ADAM12 prodomain (PA12) reduces HB-EGF shedding in endometriosis cell culture.

(a) Coomassie-stained SDS-PAGE gel showing PA12 isolation from E. coli inclusion bodies (i); its purification from Ni-NTA affinity chromatography (ii); and its subsequent refolding, dialysis, and concentration (iii). (b) PA12 inhibits recombinant ADAM12 but not recombinant ADAM-10 or -17, measured by inferring inhibitory constants (K_i) from dose-response curves in a fluorogenic FRET-peptide based assay. (n = 2 ± SD). (c,d) 2 μM PA12 treatment for 3 h increases relative levels of full-length HB-EGF on the cell surface. 12Z-HE cells stably expressing HB-EGF with Myc-tagged ectodomain and a GFP-tagged cytoplasmic tail were stained, fixed, and analyzed by flow-cytometry. (d) Corresponding to c, 3 h PA12 treatment reduces supernatant accumulation of HB-EGF, measured by ELISA. (e,f) Genetic ADAM12 knockdown increases relative levels of full-length HB-EGF on the cell surface (e), while decreasing its accumulation in the supernatant (f). 12Z-HE cells were treated with siRNA for 72 h, supernatant was exchanged, and 3 h later cells were analyzed by flow cytometry (e) and supernatant HB-EGF was measured by ELISA. (c–f) (n ≥ 3 ± SEM); *p < 0.05; two-tailed student’s t-test.

Figure 5. Protease inhibition blocks autocrine signaling independent of ligand-receptor binding affinity, and ADAM12 inhibition reduces endometriotic cellular migration.

(a) In contrast to decoy antibody efficacy, the predicted ability of PA12 to block HB-EGF autocrine signaling does not depend on ligand-receptor binding affinity. Gray dashed line at right indicates the used PA12 concentration (2 μM). (b) PA12 treatment and genetic ADAM12 knockdown significantly reduce 12Z cell migration in collagen I gels, measured over 24 h of treatment, compared to their respective controls (*n = 3 ± SEM; p < 0.05, two-tailed student’s t-test).