Navitoclax enhances the effectiveness of EGFR-targeted antibody-drug conjugates in PDX models of EGFR-expressing triple-negative breast cancer

Patient-derived xenograft models of TNBC were developed and provided by Y.D., M.F., and A.W. (Huntsman Cancer Institute). As previously described [17], tumor fragments were orthotopically transplanted into female NOD.scid mice (Charles River Labs; RRID:IMSR_ARC:NODSCID). For the PDX studies presented within Figs. 2, 3, and 4 and Supplemental Figure 7, each mouse was transplanted with one tumor fragment in order to generate one tumor per mouse. Procedures were completed in accordance with IACUC#0990 and Harvard University ARCM policies.

100 mpk ABT-263/navitoclax (AbbVie) was administered p.o. once per day. ABT-263 was formulated according to AbbVie recommendations in 60% phosal 50 PG (Lipoid), 30% polyethylene glycol 400 (DOW Chemical), and 10% ethanol. 10 mpk ABT-414 (AbbVie) or the non-tumor targeted ADC AB095-MMAF (AbbVie) was administered i.p. once per week; 0.5 mpk ABBV-321 (AbbVie) or the non-tumor targeted ADC AB095-PBD (AbbVie) was also administered i.p. once per week. ADCs were prepared in 0.9% sterile saline for injection (Hospira). The vehicle-treated mice received the vehicles corresponding to drug-treated mice. Tumor-bearing mice were randomized according to pre-treatment tumor volumes. For the studies presented within Figs. 2, 3, and 4 and Supplemental Figure 7, tumor volumes prior to treatment initiation are presented in Supplemental Figure 2, Supplemental Figure 3, Supplemental Figure 4, Supplemental Figure 6, and Supplemental Figure 7. For the studies presented within Figs. 3 and 4 and Supplemental Figure 7, the number (n) of tumor-bearing animals randomized per group were in agreement with the results from our initial in vivo studies (Fig. 2), which determined statistical significance between vehicle- and combo-treated PDX using n = 3–6 tumor-bearing animals per group. For the initial studies (Fig. 2), the 14-day treatment plan was based upon our previous work which demonstrated the efficacy of navitoclax and a HER2-targeted ADC (T-DM1) within the same time frame [15]. Pre-treatment body weights were used for individualized mpk calculations. Body weights measured at the experimental endpoints were used to calculate percent reductions in body weight.

Formalin-fixed paraffin-embedded (FFPE) tumors and Bouin’s-fixed paraffin-embedded livers were prepared, processed, sectioned, and H&E-stained by the Harvard Rodent Histopathology Core.

Serum alanine aminotransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP), and albumin (ALB) levels were measured by Charles River Laboratories, Clinical Pathology Services (Shrewsbury, MA).

FISH assays were performed using bacterial artificial chromosome clones RP11-148P17 and RP11-81B20 (Oakland Children’s Hospital) to construct probes for a 333-kb region including EGFR. Probes were biotin-labeled using the Random Prime DNA Labeling System (Invitrogen) according to the manufacturer’s protocol and detected with rhodamine-streptavidin. A FITC-labeled probe to the chromosome 7 centromeric region (CEP7) was purchased from Abbott/Vysis. The specificity of probe binding was verified using normal lymphocyte metaphase spreads. Dual-color FISH was performed on whole tissue sections. Slides were counterstained with DAPI/Antifade (Vector Labs) and evaluated using an Olympus BX51 fluorescence microscope. Hybridization signals were scored in ≥ 20 tumor cells for each case. The average EGFR signals per cell, average CEP7 signals per cell and, EGFR:CEP7 ratio were calculated. Previously described EGFR copy number classifications were applied [9, 18].

EGFR (Cell Signaling Technologies 4267), BCL-2 (Dako M0887), and BCL-X_L (Cell Signaling Technologies 2764) IHC assays and H-scores [19] were performed as previously described [15, 20].

For in vivo tumor measurements in 2-dimensions, caliper-based measurements approximated length (l) and width (w). These values were used to calculate volume via ½(l × w²). For ex vivo tumor measurements in 3-dimensions, caliper-based measurements approximated length, width, and height (h). These values were used to calculate volume via \( \pi \times \raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$6$}\right.\left(l\times w\times h\right) \). For each tumor, tumor growth was calculated as a percentage of pre-treatment tumor volume and based upon pre- and post-treatment in vivo tumor volumes. Tumor-bearing mice that died before the experimental endpoints were excluded from volume comparisons; however, tumor data for these mice is presented in the Additional file 1.

Histology slides were visualized and images were acquired via a Nikon Eclipse E200 microscope, Idea color camera, and SPOT software for MAC.

GraphPad Prism for MAC was utilized for statistical tests (Welch’s t test) and correlation analyses (Pearson’s r).

HCI-010 or HCI-025 patient-derived xenografts were dissociated using a combination of mechanical and enzymatic disruption methods. PDX were cut into small pieces using a scalpel and added to digestion medium composed of DMEM/F12 (Gibco) with DNase I recombinant grade I (130 U/ml, Roche), hyaluronidase from bovine testes (0.1 KU/ml, Sigma), and collagenase type I from rat tail (2 mg/ml, Gibco). To obtain a single-cell suspension, tumor samples were incubated in a dissociation medium at 37 °C with continuous shaking for 45–90 min. Freshly isolated PDX-derived tumor cells were cultured in advanced DMEM/F12 (Gibco) with N2 supplement (Gibco), B27 supplement (Gibco), 1% l-glutamine (Gibco), and 10% FBS (Gibco) at 37 °C under 5% CO₂. For DBP, PDX-derived tumor cells were seeded in 384 well plates (Corning 3764BC) and treated for 24 h with ABT-263 (AbbVie), ABBV-321 (AbbVie), ABT-263+ABBV-321, AB095-PBD (AbbVie), or ABT-263+AB095-PBD. DBP was performed as previously described [21, 22]. Briefly, BIM peptide dilution curves were generated using untreated tumor cells. A concentration of BIM peptide that corresponded to 10% of cytochrome c-negative cells was selected based upon the dilution curves and was used to perform BH3 profiling on the drug-treated tumor cells. Twenty-four hours post-treatment, the tumor cells were washed three times with PBS using the BioTek 406EL plate washer. Consequently, the BIM peptide was added using the HP D300e Digital Dispenser (Hewlett-Packard Development Company) at the designated concentration. Tumor cells were incubated in a BH3 profiling buffer containing 0.002% digitonin for 1 h. Tumor cells were fixed in paraformaldehyde for 15 min. The fixative was subsequently neutralized using a Tris/Glycine buffer. Tumor cells were stained overnight with Hoechst 33342 (tumor nuclei, Invitrogen), anti-cytochrome c Alexa Fluor 647-conjugated antibody (BioLegend), anti-cytokeratin antibody (tumor cells, BioLegend), and anti-vimentin antibody (tumor cells, BioLegend). Prior to imaging, the stain solution was removed via washes using the BioTek 406EL plate washer (BioTek). Fluorescence microscopy images from DBP plates were acquired using the IXM XLS high-content widefield microscope (ICCB Longwood Screening Facility, Harvard Medical School). Analysis of the images was performed using the MetaMorph Microscopy Automation and Image Analysis Software. The Multi Wavelength Cell Scoring Module was used to quantify the fraction of cytochrome c-positive tumor cells per well. All subsequent data analysis was performed in Microsoft Excel or GraphPad Prism.

Tissue microarrays (TMA) corresponding to patient triple-negative breast cancers (HTMA-240) were obtained from the Breast Biorepository at the DFCI/HCC. HTMA-240 samples were previously collected via DFCI IRB#93085 and were de-identified prior to analysis. HTMA-240 was sectioned by the Brigham & Women’s Hospital Specialized Histopathology Core, and unstained FFPE were assayed for EGFR and co-expression of BCL-2/X_L according to the IHC methods described above. H-scores were determined by breast pathologist Dr. Deborah Dillon (BWH) according to standard procedures [19] and using reference controls [EGFR, MDA-MB-468 cell-derived xenografts; BCL-2, human tonsil tissue; BCL-X_L, HCI-025 patient-derived xenografts].

Seven previously generated patient-derived xenograft models of HER2−/ER−/PR− breast cancer [16] were established and maintained via orthotopic transplantation in female NOD.scid mice [17]. Five models were derived from primary breast tumors that were either treatment naïve (HCI-002; HCI-004; HCI-016; HCI-019) or had undergone prior treatment (HCI-001), whereas two models (HCI-010; HCI-015) were derived from advanced metastatic breast cancers with multiple prior treatment exposures (Table 2).

To compare EGFR, BCL-2, and BCL-X_L expression levels within these tumor models, we performed IHC assays and H-score (H) assessment for each protein (Fig. 1 and Table 2). These models displayed a range of EGFR expression levels (Fig. 1a), with HCI-004 (H_x̅ = 76) as the lowest EGFR-expressing tumor model and HCI-010 (H_x̅ = 228) as the highest EGFR-expressing tumor model. Two models (HCI-002 and HCI-016) were distinguished by high BCL-2 expression whereas the other five models expressed low levels of BCL-2 (Fig. 1b). There was also variation in the expression of BCL-X_L, which ranged from H-scores of 10 to 220 (Fig. 1c).

To compare EGFR within these tumor models, we performed FISH assays and, as previously described [9], classified EGFR copy number alterations (Supplemental Table 1a). Based upon EGFR:CEP7 assessments (Supplemental Table 1b), none of the models examined was characterized by EGFR:CEP7 ≥ 2 and classified as EGFR-amplified; however, HCI-001, HCI-002, HCI-004, HCI-015, HCI-016, and HCI-019 were characterized by EGFR trisomy, and HCI-010 was characterized by EGFR polysomy (Supplemental Table 1b). Consistent with EGFR polysomy and EGFR expression (Fig. 1 and Table 2), the HCI-010 model was distinguished by the greatest proportion of tumor cells with ≥ 4 EGFR per cell (~ 26%). For the other PDX models evaluated, there was no obvious correlation between EGFR copy numbers and EGFR expression levels.

To obtain an initial assessment of which PDX models were most sensitive to navitoclax+ABT-414 treatments, we evaluated the efficacy of this drug combination in all seven EGFR-expressing PDX models. For these studies, tumor-bearing mice (n = 4–6 per group) were randomized into one of two groups: one to be treated with navitoclax+ABT-414 and another with the vehicles for both agents (Supplemental Figure 1a). Navitoclax (100 mpk) was administered once per day and ABT-414 (10 mpk), once per week. Using caliper-based measurements, we calculated tumor volumes pre- (Supplemental Figure 2a) and post-treatment (Supplemental Figure 2b). Fourteen days post-treatment (Fig. 2), we observed significant reductions in tumor growth in five of seven combination-treated tumor models (HCI-001; HCI-002; HCI-004; HCI-010; HCI-019). Tumor regressions were observed in HCI-002, HCI-004, and HCI-010, with the most consistent and substantial tumor regressions being observed in HCI-010 tumors (Fig. 2a). Compared to the other PDX models, HCI-010 tumors were distinguished by the highest EGFR expression levels and the highest percentage of cells with ≥ 4 EGFR copies (Table 2 and Supplemental Table 1). Non-significant reductions in tumor growth were observed in the two additional combination-treated tumor models (HCI-015; HCI-016); however, these determinations could have been limited by sample size (Fig. 2a).

Macroscopic evaluation of H&E-stained tumor sections from the vehicle-treated and combo-treated tumors (Supplemental Figure 2c) supported the tumor growth inhibition observed within HCI-001, HCI-002, HCI-004, and HCI-010 tumors. Microscopic evaluation of the H&E sections corresponding to these tumors (Fig. 2b, c) also revealed pathological responses within HCI-002 and HCI-010 tumors as well as HCI-015 tumors. Although these responses were characterized by a reduction in viable invasive tumor cell content, each model exhibited unique treatment-associated histology. For example, HCI-002-treated tumors were characterized by regions of multi-focal cellular dropout whereas HCI-010-treated tumors were characterized by a desmoplastic stroma. Interestingly, HCI-015-treated tumors displayed reduced tumor cell content associated with expanded areas of necrosis.

Based upon our initial in vivo studies, which indicated that HCI-010 tumors were the most sensitive to combined navitoclax+ABT-414 treatments, we next compared single agents and combined treatments in the HCI-010 PDX model. As a control, we also included a non-tumor [tetanus toxoid] targeted ADC (AB095-MMAF) as a single agent or in combination with navitoclax. For these studies, tumor-bearing mice (n = 5 per group) were randomized into one of six groups: navitoclax, ABT-414, navitoclax+ABT-414, AB095-MMAF, navitoclax+AB095-MMAF, and vehicles (Supplemental Figure 1b). Navitoclax (100 mpk) was administered once per day, 5 days per week, and MMAF-loaded ADCs (10 mpk), once per week. Similar to the initial studies, we calculated tumor volumes pre- (Supplemental Figure 3a) and post-treatment (Supplemental Figure 3b). Single-agent ABT-414 treatment had no effect, whereas treatment with navitoclax reduced tumor volume and caused regressions of ~ 20% on average in a subset of treated mice (Fig. 3a and Supplemental Figure 3c). However, combined navitoclax+ABT-414 treatment induced significant tumor regressions of ~ 40% on average (Fig. 3a and Supplemental Figure 3c), similar to the findings for the combined treatment of HCI-010 presented in Fig. 2. Tumor size was unaffected by single-agent AB095-MMAF treatment, whereas tumors treated with navitoclax+AB095-MMAF were comparable in size to tumors treated with single-agent navitoclax. Microscopic evaluation of the H&E slides also revealed notable pathological responses associated with navitoclax treatments, characterized by a reduction in viable invasive tumor cell content (Fig. 3b). Together, these results provided evidence that navitoclax enhances the cytotoxicity of ABT-414.

Since TNBCs are often enriched for EGFR expression in the absence of EGFR mutation or amplification, we also assessed an alternative EGFR-targeted ADC (ABBV-321) that displays an enhanced affinity for wild-type EGFR and carries the ultra-potent pyrrolobenzodiazepine (PBD) dimer [23] as an alternative cytotoxic payload. To evaluate the efficacy of ABBV-321 combined treatment in the HCI-010 model, tumor-bearing mice (n = 4–5 per group/2 studies) were randomized into six groups: navitoclax, ABBV-321, navitoclax+ABBV-321, AB095-PBD, navitoclax+AB095-PBD, and vehicles (Supplemental Figure 1c). 100 mpk of navitoclax was administered once per day, 5 days per week. Based upon ABBV-321’s enhanced EGFR affinity and potency, 0.5 mpk of PBD-loaded ADCs was administered once per week. We assessed tumor volumes pre- (Supplemental Figure 4a) and post-treatment (Supplemental Figure 4b). As observed in the previous experiments, navitoclax caused a reduction in HCI-010 tumor volume and induced partial tumor regression (Fig. 4a). However, unlike single-agent treatment with ABT-414, ABBV-321 single-agent treatment resulted in dramatic tumor regressions, on average ~ 66% (Fig. 4a). Notably, navitoclax enhanced the effectiveness of ABBV-321 as evidenced by dramatic and near-complete tumor regressions, on average ~ 88% (Fig. 4a). Macroscopic analysis of the H&E slides (Supplemental Figure 4d) indicated the tumor size reductions induced by ABBV-321 or navitoclax+ABBV-321 treatment. Consistent with these measurements, pathological responses characterized by the elimination of viable invasive carcinoma were associated with ABBV-321 and were dramatically enhanced by navitoclax (Fig. 4b).

We extended these studies to include an additional PDX (HCI-025). We selected HCI-025 after characterizing EGFR, BCL-2, and BCL-X_L expression levels in another seven PDX models of TNBC (Supplemental Table 3). HCI-025 was characterized by EGFR low polysomy (Supplemental Table 1) and comparable to HCI-010, high EGFR expression levels (H = 230) (Supplemental Figure 5). Compared to the HCI-010 model (Supplemental Figure 4), HCI-025 tumors expand at a faster rate in vivo (Supplemental Figure 6). To evaluate ABBV-321 combined treatment within the HCI-025 model, we performed a similar six-group treatment study (Supplemental Figure 1d) and calculated tumor volumes pre- (Supplemental Figure 6a) and post-treatment (Supplemental Figure 6b). HCI-025 tumors were sensitive to single-agent ABBV-321, displaying tumor regressions on average ~ 36% (Fig. 4c). As observed with HCI-010, navitoclax also enhanced the effectiveness of ABBV-321 in HCI-025 tumors, as evidenced by dramatic tumor regressions, on average ~ 68% (Fig. 4c). Consistent with these measurements, macroscopic analysis of the H&E slides highlighted ABBV-321-associated tumor size reductions (Supplemental Figure 6d). Moreover, pathological responses characterized by the elimination of viable invasive carcinoma were associated with combined navitoclax treatments (Fig. 4d).

AB095-PBD (non-tumor targeted ADC) and navitoclax+AB095-PBD also caused reductions in tumor volume and regressions in the HCI-010 and HCI-025 PDX models; however, responses to ABBV-321 alone were significantly greater than to AB095-PBD alone, and the responses to navitoclax+ABBV-321 were greater than navitoclax+AB095-PBD, supporting a contribution of EGFR-targeted effects to the efficacy of ABBV-321. These observations are consistent with some level of antigen-independent accumulation and uptake of AB095-PBD, which could be due to enhanced permeability and retention (EPR) of antibody-drug conjugates [24], tumor-associated macropinocytosis [25‐28], or extracellular cleavage of the ADC linker [29]. These effects have been previously described for ADCs and other agents in vivo [25, 29‐32].

To address combination effectiveness within the context of larger tumors, we evaluated treatment effects in the HCI-025 model after tumors had reached a volume of ~ 382.5 mm³ on average (Supplemental Figure 7). These pre-treatment tumor volumes were three times larger than the pre-treatment tumor volumes used in the experiments shown in Fig. 4 and Supplemental Figure 6a (~ 120 mm³ on average). Treatment groups and schedules were as described above (Supplemental Figure 1d). Significant ABBV-321 treatment effects, characterized by tumor growth arrest, were observed in mice carrying larger tumors (Supplemental Figure 7). Importantly, navitoclax once again enhanced the efficacy of ABBV-321 and resulted in moderate regressions characterized by ~ 21.5% on average (Supplemental Figure 7).

Adverse events were observed in HCI-010 and HCI-025 tumors treated with the PBD-loaded antibodies combined with navitoclax. Deaths were documented in two out of nine HCI-010 and three out of ten HCI-025 mice treated with navitoclax+ABBV-321 (Supplemental Figure 8a and e). Deaths were also documented in two out of nine HCI-010 and two out of ten HCI-025 mice treated with navitoclax+AB095-PBD (Supplemental Figure 8a and e). Interestingly, neither body weight reductions (Supplemental Figure 8b and f) nor tumor growth inhibition (Supplemental Figure 8c and d & g and h) distinguished HCI-010 or HCI-025 tumor-bearing mice that died. To provide additional insight into the nature of the toxicity, nontumor-bearing female NOD.scid mice (n = 5 per group) were treated with the PBD-loaded antibodies and or navitoclax according to the doses and schedules presented in Supplemental Figure 1d. Deaths were not observed in nontumor-bearing mice treated with PBD-loaded antibodies combined with navitoclax (Supplemental Figure 8i); however, body weight reductions (< 15%) were observed after day 12 (Supplemental Figure 8j-l). Seventeen days post-treatment, the serum was collected for clinical chemistry analyses, and carcasses were saved for necropsies. Microscopic evaluation of the livers revealed no obvious histopathology in any of the treatment groups except for those treated with PBD-loaded antibodies combined with navitoclax. Livers collected from animals treated with combined agents were distinguished by occasional apoptotic and or necrotic hepatocytes. To better assess liver function, we evaluated the serum samples for alanine aminotransferase (ALT; Supplemental Figure 8 m), aspartate aminotransferase (AST; Supplemental Figure 8n), alkaline phosphatase (ALP; Supplemental Figure 8o), and albumin (ALB; Supplemental Figure 8p). One of the five mice treated with either navitoclax+ABBV-321 (#16; Supplemental Figure 8 m and n) or navitoclax+AB095-PBD (#27; Supplemental Figure 8 m and n) displayed significantly elevated levels of ALT and AST (3.2-, 7.7-, 2.8-, and 4.2-fold higher than the mean of the vehicle-treated ALT and AST levels, respectively). None of the mice displayed significantly elevated levels of ALP (Supplemental Figure 8o), whereas there was a trend towards decreased levels of ALB in mice treated with combined agents (Supplemental Figure 8p). Together, these results suggest possible hepatotoxicity in some animals treated with combined agents.

Dynamic BH3 profiling (DBP) [21, 22] was performed on PDX-derived tumor cells corresponding to HCI-010 (Fig. 5a) and HCI-025 (Fig. 5b) in order to assess the extent of mitochondrial apoptotic priming induced by ABBV-321 and/or ABT-263/navitoclax treatments. It was also of interest to assess whether AB095-PBD displayed apoptotic priming in vitro as this compound caused reductions in tumor size and regressions in vivo (Fig. 4). DBP revealed that both navitoclax and ABBV-321 single-agent treatments induced mitochondrial-associated apoptotic signaling within HCI-010 tumor cells, whereas only ABBV-321 single-agent treatments induced mitochondrial-associated apoptotic signaling within HCI-025 tumor cells. Notably, tumor cells derived from either HCI-010 or HCI-025 exhibited enhanced priming under combined navitoclax+ABBV-321 treatments in vitro. These responses were remarkably consistent with the treatment responses observed in vivo (Fig. 4). Surprisingly, treatments in vitro with AB095-PBD as a single agent failed to induce mitochondrial apoptotic signaling within either HCI-010 (Fig. 5a) or HCI-025 (Fig. 5b), and treatment with navitoclax+AB095-PBD in HCI-010 did not induce additional apoptotic priming relative to navitoclax alone, suggesting that the tumor reductions associated with this non-targeting ADC are dependent on factors within the tumor microenvironment in vivo.

Our pre-clinical results provide evidence that navitoclax enhances the effectiveness of EGFR-targeted ADCs and highlight the potential clinical utility of navitoclax+ABBV-321 as an effective treatment option for EGFR-expressing TNBC. To evaluate the co-expression of EGFR and BCL-2/X_L within patient triple-negative tumors, we utilized a human tumor microarray (HTMA-240) comprising triple-negative breast cancers. Forty-six TNBC cases (TN#) arrayed on HTMA-240 were evaluable for EGFR, BCL-2, and BCL-X_L. To compare EGFR expression levels within these tumors, we performed IHC assays and H-score (H) assessment as described above (Fig. 1). EGFR expression (H > 0) was detected in 87% of tumors (n = 40), and the majority of these cases were characterized by EGFR H-scores ≥ 50 (n = 27) (Fig. 6). Co-expression of the pro-survival proteins BCL-2 (Fig. 6a) and BCL-X_L (Fig. 6b) was also characterized for these tumors. Only 28% of tumors evaluated were distinguished by BCL-2 expression greater than an H-score of 50 (n = 13); the majority either expressed low levels (0 < H < 50; n = 10) or undetectable levels (H = 0; n = 23) of BCL-2. Compared to BCL-2, BCL-X_L expression was detected in 100% of tumors (n = 46), and the majority of these cases were characterized by BCL-X_L H-scores ≥ 50 (n = 44). Although comparative assessment of BCL-2 and BCL-X_L (Fig. 6) revealed predominant BCL-X_L expression within tumors (BCL-X_L > BCL-2; n = 35), the subset of tumors distinguished by notable BCL-2 expression co-expressed BCL-X_L at levels less than BCL-2 (BCL-X_L < BCL-2; n = 7) or comparable to BCL-2 (BCL-X_L ≈ BCL-2; n = 4).