Preferential Antibody and Drug Conjugate Targeting of the ADAM10 Metalloprotease in Tumours

Simple Summary ADAM10 is a cell surface protein that releases other proteins from cells, thereby controlling a range of functions during normal development. Its activity is normally tightly regulated, but it can become deregulated in cancer calls. We previously showed that an active form of ADAM10 is elevated in tumours in both mice and humans, and that we could detect this active form using an antibody which we developed that binds to a specific region of ADAM10, which is apparently hidden in the inactive form. We now provide insight explaining the specificity of our antibody (8C7) for active ADAM10, and show that it preferentially targets tumours when injected into mice. We thus conjugated cytotoxic drugs to 8C7 in order to preferentially bind and kill tumour cells that contain activated ADAM10. Our experiments show that our ‘8C7 antibody–drug conjugates’ specifically kill cells expressing 8C7-reactive ADAM10, and can inhibit the growth of tumours in mice, without significant side effects, suggesting their potential as a novel approach for targeted cancer therapy. Abstract ADAM10 is a transmembrane metalloprotease that sheds a variety of cell surface proteins, including receptors and ligands that regulate a range of developmental processes which re-emerge during tumour development. While ADAM10 is ubiquitously expressed, its activity is normally tightly regulated, but becomes deregulated in tumours. We previously reported the generation of a monoclonal antibody, 8C7, which preferentially recognises an active form of ADAM10 in human and mouse tumours. We now report our investigation of the mechanism of this specificity, and the preferential targeting of 8C7 to human tumour cell xenografts in mice. We also report the development of novel 8C7 antibody–drug conjugates that preferentially kill cells displaying the 8C7 epitope, and that can inhibit tumour growth in mice. This study provides the first demonstration that antibody–drug conjugates targeting an active conformer of ADAM10, a widely expressed transmembrane metalloprotease, enable tumour-selective targeting and inhibition.


Introduction
A disintegrin and metalloprotease (ADAM) 10 is a transmembrane protein with extracellular metalloprotease (MP), disintegrin (D) and cysteine-rich (C) domains. It is an
Effects of ADC treatment (at the indicated concentration ranges) on cell viability were measured using an MTS assay (CellTiter 96, Promega) after 3-4 days.

Animal Studies
Immuno-deficient athymic mice (BALB/c nude) or NSG mice (Nod/SCID/gamma) (5-6 weeks old) were obtained from the Animal Resources Centre (Canning Vale, Western Australia, Australia). All animals were handled in strict accordance with good animal practice, as defined by the National Health and Medical Research Council (Australia) Code of Practice for the Care and Use of Animals for Experimental Purposes, and approved by the Austin Hospital Animal Ethics Committee. Patient-derived xenografts were generated by subcutaneous insertion of 2-3 mm square samples of fresh colorectal tumours into the flank of NSG mice, and serial passage in NSG mice.

Biodistribution Studies
8C7 was coupled to 111 In with the bifunctional metal ion chelator CHX-A diethylenetriamine-penta-acetic acid (DTPA), and to 125 I by direct coupling using chloramine-T, as described previously [22], following which binding to immobilised antigen was confirmed. For the biodistribution experiments, BALB/c nude mice were injected subcutaneously on the flank with 10 7 LIM1215 cells in 30% growth factor-reduced Matrigel (BD), and after 30 days (at a tumour size of approximately 250 mm 3 ), the mice received a tail vein injection of 111 In-8C7 (5 ug/8 µCi) and 125 I (5 ug/7.05 µCi). Five mice/time point were sacrificed after 2 h, 2 d and 5 d, and tissue, blood and tumour samples were taken and weighed. Sample radioactivity was determined using a gamma counter, calculated as a fraction (%) of injected dose per gram of tissue, and graphed for each time point. For imaging, only 111 In-8C7 [47.4 µg/75.9 µCi] was injected, and gamma camera/MRI scans (Mediso, Budapest, Hungary) were taken of anaesthetised mice.

Therapeutic Studies
Mice were injected subcutaneously on the flank with 10 7 LIM1215 colon cancer cells in 30% growth factor-reduced BD Matrigel (BALB/c nude mice), or 5 × 10 6 U251MG glioma cells in PBS (NSG mice). When tumours reached 100-200 mm 3 , randomised groups of n ≥ 5 mice were treated twice weekly with I.P. injection of 8C7-PBD control IgG-PBD (0.3-0.5 mg/kg), naked 8C7 (10 mg/kg) or vehicle control (PBS). Tumour volumes were determined by caliper measurement (volume = length × width 2 /2) and mouse weights were measured over the treatment course. Mice were humanely euthanized at or before the ethical endpoint, and tumours were recovered for analysis. Statistical significance was determined using the student's T test. Results are representative of at least 2 replicate experiments.

Flow Cytometry, Histology and Western Blot Analysis
Flow cytometry was performed on live cells using a BD LSR II flow cytometer and a BD FACS Canto II (BD Biosciences). Tumour immunohistochemistry (IHC) and fluorescence (IF) was performed on sections (6 mm) of OCT-embedded fresh-frozen resected tumours. For IHC, sections were stained (Vector Labs ABC Kit), hematoxylin counterstained and imaged (Aperio AT2). IF staining was with Alexa-labelled secondary antibodies (Abcam) and nuclear stain (DAPI), and imaging was carried out using a Zeiss LSM 980 confocal microscope. Images were analysed with Aperio (Imagescope version 12.4.2.5010) and HALO software (version 3.1.1076.429). Western blot analysis of tumour lysates was performed as previously described [20], after equalising protein content, as determined by BSA assay (Pierce).

Results
We previously showed that our ADAM10 monoclonal antibody 8C7 is specific for ADAM10, and that it recognises both human and mouse forms, reflecting their high homology; with 95% sequence conservation in the C domain and 100% sequence conservation of residues contacting the 8C7 complementarity-determining regions (CDRs) [19]. This cross-reactivity enables assessment of the relative binding of 8C7 to ADAM10 in tumour and normal tissue in mice with human tumour xenografts. We thus evaluated the biodistribution of mAb 8C7 in mice bearing human LIM1215 colon cancer xenografts, which we have previously shown to possess high levels of active ADAM10 [20]. 8C7 was conjugated to 111 In ( Figure 1A,C) or 125 I ( Figure 1B,C) and intravenously injected, and tumour and tissue samples were collected over 6 days and analysed for radioactivity. Both conjugates showed similar distribution profiles, with gradual clearance from the blood, and clearance from most tissues consistent with blood pool clearance, with no specific tissue uptake and retention. In comparison, antibody retention in the tumour was more prolonged, indicating preferential binding to tumour cells compared to normal organs ( Figure 1C,D). The enhanced tumour uptake of 111 In-8C7 compared to 125 I-8C7 may reflect the known rapid internalisation of 8C7 [19], as the Indium label is better retained under these conditions, while 125 I is dehalogenated and diffuses out of the tumour cell [22]. SPECT/MRI imaging of 111 In-8C7 2 days post injection also indicated preferential targeting of the tumour. The 111 In-8C7 signal detected in liver and spleen was consistent with metabolism of the 111 In-chelate in these organs, rather than with higher antibody retention ( Figure 1D,E).
( Figure 1C,D). The enhanced tumour uptake of 111 In-8C7 compared to 125 I-8C7 may reflect the known rapid internalisation of 8C7 [19], as the Indium label is better retained under these conditions, while 125 I is dehalogenated and diffuses out of the tumour cell [22]. SPECT/MRI imaging of 111 In-8C7 2 days post injection also indicated preferential targeting of the tumour. The 111 In-8C7 signal detected in liver and spleen was consistent with metabolism of the 111 In-chelate in these organs, rather than with higher antibody retention ( Figure 1D,E). We next analysed the binding of 8C7 to epithelial cells isolated from fresh human colorectal tumour and adjacent normal colon tissue. We used flow cytometry to detect the binding of fluorescent Alexa-647-labelled 8C7, or of 4A11, an anti-human ADAM10 antibody that we developed which binds to both active and inactive ADAM10 [20]. While the non-selective ADAM10 mAb 4A11 bound to both normal colon and tumour epithelial (CD45 negative, EpCam positive) cells, 8C7 preferentially bound to tumour cells, suggesting selective targeting of active ADAM10 in the tumour (Figure 2A). Quantitative PCR of ADAM10 mRNA expression confirmed comparable ADAM10 expression to that of a larger set of normal and tumour tissues from a cohort of 19 colorectal cancer patients, in which expression was independent of tumour stage ( Figure S1A,B). We also generated patient-derived tumour xenograft (PDX) lines grown in immunodeficient mice, and compared 8C7 binding to tumours versus normal tissues ( Figure 2B,C). 8C7 selectively bound to tumours in multiple colon PDX lines, irrespective of the initial tumour stage, compared to its low binding to normal colon, and to spleen, which was also recently reported to express ADAM10 [23]. In this case, the control antibody 4A11 did not show significant We next analysed the binding of 8C7 to epithelial cells isolated from fresh human colorectal tumour and adjacent normal colon tissue. We used flow cytometry to detect the binding of fluorescent Alexa-647-labelled 8C7, or of 4A11, an anti-human ADAM10 antibody that we developed which binds to both active and inactive ADAM10 [20]. While the non-selective ADAM10 mAb 4A11 bound to both normal colon and tumour epithelial (CD45 negative, EpCam positive) cells, 8C7 preferentially bound to tumour cells, suggesting selective targeting of active ADAM10 in the tumour (Figure 2A). Quantitative PCR of ADAM10 mRNA expression confirmed comparable ADAM10 expression to that of a larger set of normal and tumour tissues from a cohort of 19 colorectal cancer patients, in which expression was independent of tumour stage ( Figure S1A,B). We also generated patientderived tumour xenograft (PDX) lines grown in immunodeficient mice, and compared 8C7 binding to tumours versus normal tissues ( Figure 2B,C). 8C7 selectively bound to tumours in multiple colon PDX lines, irrespective of the initial tumour stage, compared to its low binding to normal colon, and to spleen, which was also recently reported to express ADAM10 [23]. In this case, the control antibody 4A11 did not show significant binding to normal mouse tissues (not shown), consistent with its specificity for human ADAM10 [20]. binding to normal mouse tissues (not shown), consistent with its specificity for human ADAM10 [20]. The crystal structure of the ADAM10 extracellular domain suggests a bent, 'closed' conformation, where the MP and C domains are in contact, partially obscuring both the catalytic cleft and the C domain [13] ( Figure 3A,C). Comparison with the previously determined crystal structure of 8C7 in complex with the isolated D + C domains of ADAM10 (18) showed a steric clash between the MP domain in the closed conformation and 8C7 binding [13] (Figure 3B,C). In addition, 8C7 binding correlates with ADAM activity [13,20], indicating that the selectivity of 8C7 for active ADAM10 may be due to the MP domain inhibiting 8C7 binding in the inactive, closed conformation. To test this hypothesis, we generated a membrane-expressed ADAM10 construct lacking the MP domain (ADAM10D + C), which we transfected into HEK293 cells, which also express endogenous wild type (wt) ADAM10. Flow cytometry showed that 8C7 clearly bound to cells expressing ADAM10D + C on the cell surface, but not parental cells with only the wtADAM10, whereas 4A11 was able to bind to both forms ( Figure 3D). This data indicates that 8C7 is selective for an ADAM10 conformation where the C domain is exposed via displacement of the MP domain. ADAM10D + C-expressing cells thus provide a convenient model for testing 8C7 in vitro as a potential means to target cytotoxic payloads in cells with active, 'open' ADAM10. (B) Tumours were recovered from mice bearing two human colorectal PDX lines, and single cell isolates were stained with 8C7, 4A11 or control IgG, before analysis by flow cytometry. Normal colon and spleen were also analysed for 8C7 binding (human-specific 4A11 was not used). Histograms show overlaid binding of ADAM10 antibodies with an isotype control. (C) Quantitation of 8C7 binding to tumour cell isolates, from multiple passages of six distinct colorectal PDX tumour lines (n = 14 tumours), compared to normal tissue (% positive cells). Binding of the anti-human ADAM10 mAb 4A11 to PDX tumour cells is also shown for comparison.
The crystal structure of the ADAM10 extracellular domain suggests a bent, 'closed' conformation, where the MP and C domains are in contact, partially obscuring both the catalytic cleft and the C domain [13] (Figure 3A,C). Comparison with the previously determined crystal structure of 8C7 in complex with the isolated D + C domains of ADAM10 (18) showed a steric clash between the MP domain in the closed conformation and 8C7 binding [13] (Figure 3B,C). In addition, 8C7 binding correlates with ADAM activity [13,20], indicating that the selectivity of 8C7 for active ADAM10 may be due to the MP domain inhibiting 8C7 binding in the inactive, closed conformation. To test this hypothesis, we generated a membrane-expressed ADAM10 construct lacking the MP domain (ADAM10D + C), which we transfected into HEK293 cells, which also express endogenous wild type (wt) ADAM10. Flow cytometry showed that 8C7 clearly bound to cells expressing ADAM10D + C on the cell surface, but not parental cells with only the wtADAM10, whereas 4A11 was able to bind to both forms ( Figure 3D). This data indicates that 8C7 is selective for an ADAM10 conformation where the C domain is exposed via displacement of the MP domain. ADAM10D + C-expressing cells thus provide a convenient model for testing 8C7 in vitro as a potential means to target cytotoxic payloads in cells with active, 'open' ADAM10.
Next, we generated 8C7 antibody-drug conjugates, by coupling 8C7 with the DNA binding cytotoxin PBD, the microtubule disruptor MMAE, or the Topoisomerase I inhibitor DXd (DX8591 derivative), all of which have been clinically tested with other antibodies [24,25]. The drugs were attached using a cleavable linker, enabling their release upon internalisation into cells. Linker-ADC structures are shown in Figure 4A. Flow cytometry analysis confirmed 8C7-ADCs retained binding to ADAM10D + C cells ( Figure S2). We then tested the effects of the ADCs on the viability of ADAM10D + C cells and parental HEK293 cells expressing only endogenous ADAM10. All 8C7-ADCs were able to efficiently kill cells expressing ADAM10D + C containing the exposed 8C7 epitope, compared to the corresponding control ADCs ( Figure 4B). In contrast, the parental cells were equally resistant to the control and 8C7-ADCs, confirming the selectivity of 8C7-ADCs. 8C7-PBD had the highest activity, consistent with the known potency characteristics of PBD dimers [24]. Similar to parental HEK293 cells, 8C7-ADCs were ineffective against LIM1215 colon and U251 glioblastoma (GBM) tumour cells expressing ADAM10 ( Figure S3), consistent with the low binding of 8C7 to live cells in vitro [19] compared to that in tumours, likely reflecting a predominantly inactive conformation on cells cultured in vitro [20].  Next, we generated 8C7 antibody-drug conjugates, by coupling 8C7 with the DNA binding cytotoxin PBD, the microtubule disruptor MMAE, or the Topoisomerase I inhibitor DXd (DX8591 derivative), all of which have been clinically tested with other antibodies [24,25]. The drugs were attached using a cleavable linker, enabling their release upon internalisation into cells. Linker-ADC structures are shown in Figure 4A. Flow cytometry and U251 glioblastoma (GBM) tumour cells expressing ADAM10 ( Figure S3), consistent with the low binding of 8C7 to live cells in vitro [19] compared to that in tumours, likely reflecting a predominantly inactive conformation on cells cultured in vitro [20]. Finally, we evaluated 8C7-ADCs in mice bearing tumours. We concentrated on 8C7-PBD, which showed the highest potency in vitro. In addition to LIM1215 colon xenografts, we also used U251 glioblastoma (GBM) xenografts, since ADAM10 expression is associated with poor prognosis in GBM [26], where it mediates Notch and other oncogenic signalling [27,28]. Mice treated with 8C7-PBD exhibited significantly inhibited tumour growth compared to those treated with vehicle or control IgG, and also compared to mice subjected to treatment with a control IgG-PBD conjugate ( Figure 5A,B). 8C7-PBD was also more effective than naked 8C7 at >20-fold higher dose (p < 0.01 in GBM model, Figure 5B). Tumours from 8C7-PBD-treated mice showed decreased staining of the endothelial marker CD31, and increased staining for phospho-histone γH2AX, a marker for G2-M cell cycle arrest ( Figure 5C,D, Figure S4A), known to result from PBD-induced DNA cross-linking [29]. These results are consistent with 8C7-PBD targeting a subpopulation of tumour cells with high Notch activity, that support angiogenesis [20,30]. Indeed, Western blot analysis of tumour lysates indicated decreased Notch 1 levels in 8C7-PBDtreated tumours, with the one outlier being the largest, most resistant tumour ( Figure  5E). Mouse weights and overall health were not significantly affected, and were equivalent to control ADC, consistent with minimal binding of 8C7 to ADAM10 in normal tissues ( Figure S4B). 8C7-PBD was also effective against a human colon cancer PDX model ( Figure S5A). We also tested 8C7-DXd in mice bearing U251 tumours; this also effectively Finally, we evaluated 8C7-ADCs in mice bearing tumours. We concentrated on 8C7-PBD, which showed the highest potency in vitro. In addition to LIM1215 colon xenografts, we also used U251 glioblastoma (GBM) xenografts, since ADAM10 expression is associated with poor prognosis in GBM [26], where it mediates Notch and other oncogenic signalling [27,28]. Mice treated with 8C7-PBD exhibited significantly inhibited tumour growth compared to those treated with vehicle or control IgG, and also compared to mice subjected to treatment with a control IgG-PBD conjugate ( Figure 5A,B). 8C7-PBD was also more effective than naked 8C7 at >20-fold higher dose (p < 0.01 in GBM model, Figure 5B). Tumours from 8C7-PBD-treated mice showed decreased staining of the endothelial marker CD31, and increased staining for phospho-histone γH2AX, a marker for G2-M cell cycle arrest ( Figure 5C,D, Figure S4A), known to result from PBD-induced DNA cross-linking [29]. These results are consistent with 8C7-PBD targeting a subpopulation of tumour cells with high Notch activity, that support angiogenesis [20,30]. Indeed, Western blot analysis of tumour lysates indicated decreased Notch 1 levels in 8C7-PBD-treated tumours, with the one outlier being the largest, most resistant tumour ( Figure 5E). Mouse weights and overall health were not significantly affected, and were equivalent to control ADC, consistent with minimal binding of 8C7 to ADAM10 in normal tissues ( Figure S4B). 8C7-PBD was also effective against a human colon cancer PDX model ( Figure S5A). We also tested 8C7-DXd in mice bearing U251 tumours; this also effectively inhibited tumour growth while being well tolerated, with mouse weights being unaffected compared to control mice ( Figure S5B,C).
Cancers 2022, 13, x FOR PEER REVIEW 9 of 13 inhibited tumour growth while being well tolerated, with mouse weights being unaffected compared to control mice ( Figure S5B,C).

Discussion
The metalloprotease ADAM10 is highly expressed in a range of tumour types, where it mediates activation of signalling pathways associated with tumour 'stemness', proliferation and angiogenesis. Accordingly, its higher expression in tumours correlates with poor patient prognosis for a range of tumour types, suggesting ADAM10 is a potential therapeutic target [3]. Previous attempts to target ADAM10 have focused on small molecule inhibitors targeting the metalloprotease active site, based on hydroxamate analogues originally used for MMPs. While the efficacy and specificity of such inhibitors have improved, none have yet been approved for therapy [3,31,32].
An alternative approach is to employ monoclonal antibodies, which have the advantages of high affinity, specificity and stability in vivo [33]. They can either be inhibitory (e.g., by competing with ligand or substrate binding), or induce cell death (by immune-mediated cell-or complement-dependent cytotoxicity, or by direct attachment of cytotoxic drugs) [34]. Antibody-drug conjugates (ADCs) are prominent amongst newly approved cancer treatments [35,36]. Clearly, these must preferentially target tumour antigens to avoid toxicity to normal tissues. While the wide expression of ADAM10 in normal tissues could undermine such an approach [37], we previously showed preferential binding of our antibody, 8C7, to active ADAM10 which was upregulated in tumours compared to normal tissues, both in mice and in human samples [20]. We now support this with data from biodistribution studies of radiolabelled 8C7 in mice bearing human colorectal cancer xenografts, in which the mouse/human cross-reactive antibody preferentially accumulated in tumours compared to normal tissues. Furthermore, we show that the activity-dependent selectivity of 8C7 is likely due to conformational exposure of the 8C7 epitope in the C-domain of active ADAM10, which is obscured by the MP domain in the auto-inhibited conformation [13], since removal of the MP domain facilitates robust binding of 8C7. While previous studies have indicated conformational changes upon activation of ADAM proteases [14][15][16], and the ADAM10 crystal structure shows a 'closed' inactive conformation, with interaction of the MP and C domains likely causing steric hindrance of substrate binding [13], this is the first direct evidence that movement of the metalloprotease domain allows access to a hidden epitope in the C domain recognized by 8C7. The selectivity of 8C7 for active ADAM10 in tumours [20] is consistent with active ADAM10 having an open structure, allowing substrate and antibody access ( Figure 6).

Discussion
The metalloprotease ADAM10 is highly expressed in a range of tumour types, where it mediates activation of signalling pathways associated with tumour 'stemness', proliferation and angiogenesis. Accordingly, its higher expression in tumours correlates with poor patient prognosis for a range of tumour types, suggesting ADAM10 is a potential therapeutic target [3]. Previous attempts to target ADAM10 have focused on small molecule inhibitors targeting the metalloprotease active site, based on hydroxamate analogues originally used for MMPs. While the efficacy and specificity of such inhibitors have improved, none have yet been approved for therapy [3,31,32].
An alternative approach is to employ monoclonal antibodies, which have the advantages of high affinity, specificity and stability in vivo [33]. They can either be inhibitory (eg., by competing with ligand or substrate binding), or induce cell death (by immunemediated cell-or complement-dependent cytotoxicity, or by direct attachment of cytotoxic drugs) [34]. Antibody-drug conjugates (ADCs) are prominent amongst newly approved cancer treatments [35,36]. Clearly, these must preferentially target tumour antigens to avoid toxicity to normal tissues. While the wide expression of ADAM10 in normal tissues could undermine such an approach [37], we previously showed preferential binding of our antibody, 8C7, to active ADAM10 which was upregulated in tumours compared to normal tissues, both in mice and in human samples [20]. We now support this with data from biodistribution studies of radiolabelled 8C7 in mice bearing human colorectal cancer xenografts, in which the mouse/human cross-reactive antibody preferentially accumulated in tumours compared to normal tissues. Furthermore, we show that the activitydependent selectivity of 8C7 is likely due to conformational exposure of the 8C7 epitope in the C-domain of active ADAM10, which is obscured by the MP domain in the autoinhibited conformation [13], since removal of the MP domain facilitates robust binding of 8C7. While previous studies have indicated conformational changes upon activation of ADAM proteases [14][15][16], and the ADAM10 crystal structure shows a 'closed' inactive conformation, with interaction of the MP and C domains likely causing steric hindrance of substrate binding [13], this is the first direct evidence that movement of the metalloprotease domain allows access to a hidden epitope in the C domain recognized by 8C7. The selectivity of 8C7 for active ADAM10 in tumours [20] is consistent with active ADAM10 having an open structure, allowing substrate and antibody access ( Figure 6). Figure 6. Specificity of mAb 8C7 for an active conformer of the metalloprotease ADAM10 lies in its binding to an epitope in the C domain (green), which is obscured by the metalloprotease domain (MP) in the inactive conformer. Furthermore, 8C7 selectively targets ADAM10 in tumours, and shows improved anti-tumour activity when conjugated to cytotoxic payloads (red).
We previously showed partial inhibition of tumour growth in mice using naked 8C7 alone, likely due to it inhibiting ADAM10-mediated Notch activation [20]. Based on its demonstrated tumour selectivity, we sought to improve its anti-tumour efficacy by developing 8C7-ADCs coupled with cytotoxic drugs to inhibit cell replication. We tested 8C7 coupled to a microtubule inhibitor, MMAE, a DNA cross-linker, PBD dimer SGN3199, Figure 6. Specificity of mAb 8C7 for an active conformer of the metalloprotease ADAM10 lies in its binding to an epitope in the C domain (green), which is obscured by the metalloprotease domain (MP) in the inactive conformer. Furthermore, 8C7 selectively targets ADAM10 in tumours, and shows improved anti-tumour activity when conjugated to cytotoxic payloads (red).
We previously showed partial inhibition of tumour growth in mice using naked 8C7 alone, likely due to it inhibiting ADAM10-mediated Notch activation [20]. Based on its demonstrated tumour selectivity, we sought to improve its anti-tumour efficacy by developing 8C7-ADCs coupled with cytotoxic drugs to inhibit cell replication. We tested 8C7 coupled to a microtubule inhibitor, MMAE, a DNA cross-linker, PBD dimer SGN3199, and a Topoisomerase I inhibitor (DXd). 8C7-ADCs inhibited cell proliferation and viability in vitro compared to matched control IgG-ADCs, with selectivity for cells expressing ADAM10 with the 8C7 epitope exposed. Importantly, experiments in mice confirmed 8C7-ADCs can inhibit growth of colorectal and glioblastoma xenografts, compared to control IgG-ADCs, or 8C7 alone. The effects on tumour growth indicated 8C7-ADCs targeted a subpopulation of tumour cells which exhibited cell cycle arrest, and this was associated with decreased blood vessel density and Notch1 expression, consistent with previous work showing selectivity for tumour stem-like cells with high Notch activity [20]. 8C7-ADCs also appeared well-tolerated at the concentrations used in our experiments, with only a slight trend of weight loss in mice treated with PBD ADCs (both 8C7 and control IgG), indicating minimal non-antigen-related uptake in normal tissues. However, since other studies in animals and humans have shown toxicity of ADCs using PBD dimers [38], we also tested 8C7 coupled to the new generation topoisomerase I inhibitor DXd, used in the anti-HER2 ADC Trastuzumab deruxtecan (DS-8201a, or T-DXd) [25]. 8C7-DXd also effectively inhibited the growth of U251 GBM xenografts in mice with no apparent adverse effects.

Conclusions
In summary, we show evidence for conformation-specific binding of antibody 8C7 to ADAM10, and its tumour selectivity in mice, supporting development of 8C7-based ADCs. We furthermore show promising activity of our novel 8C7-ADCs in colorectal and glioblastoma xenograft models. This is the first demonstration of the use of ADCs against ADAM proteases as a basis for tumour-selective targeting. Our results indicate a potential for clinical development of ADCs utilising humanised 8C7, which we are currently progressing.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/cancers14133171/s1, Figure S1. Quantitative reverse transcriptase PCR of ADAM10 mRNA from human colorectal tumours compared to nearby normal tissue. Figure S2. Flow cytometry analysis comparing 8C7 and 8C7-ADC binding to HEK293 cells with the exposed 8C7 epitope (ADAM10D + C). Figure S3. Viability assays of tumour cells treated with 8C7-ADCs and matched control ADCs. Figure S4. A. Analysis of U251 tumours from 8C7-PBD-treated mice, and B. Monitoring safety of PBD ADCs. Figure S5. A. Tumour volumes following 8C7-PBD treatment of NSG mice bearing human colon cancer patient-derived xenografts. B, C. 8C7-DXd treatment of NSG mice bearing U251 cell xenografts.