Bystander Effects, Pharmacokinetics, and Linker-Payload Stability of EGFR-Targeting Antibody-Drug Conjugates Losatuxizumab Vedotin and Depatux-M in Glioblastoma Models

Overexpression of receptor tyrosine kinases (RTK) is a common oncogenic driver event, and multiple antibody-drug conjugates (ADC) capitalize on RTK differential expression in cancer versus normal tissues (1–3). The epidermal growth factor receptor (EGFR) is a highly expressed RTK in numerous solid malignancies and is mutated and/or amplified in approximately half of patients with glioblastoma (GBM; refs. 4, 5). In this context, there is tremendous interest in evaluating EGFR-targeted ADCs in GBM, and AbbVie has designed multiple such ADCs, which are at different stages of clinical development. These include Depatuxizumab mafodotin (Depatux-M), which has a noncleavable linker and a cell-impermeant monomethyl auristatin F (MMAF) toxin (6, 7), and Losatuxizumab vedotin (ABBV-221), which uses an affinity matured antibody, a cathepsin-B cleavable linker, and a cell-permeant monomethyl auristatin E (MMAE) toxin (8). Previously, we demonstrated that Depatux-M is highly potent against EGFR-amplified GBM. However, its limited distribution across the brain tumor–blood barrier significantly limited efficacy in many orthotopic tumor models (9). Using CED to directly infuse drug into intracranial tumors significantly increased tissue exposure and efficacy of Depatux-M (10). The focus of the current study was to compare key features of efficacy, safety, and bystander effects of Depatux-M and ABBV-221 with regard to potential use for GBM treatment.

The toxin and linker used in Depatux-M and ABBV-221 are highly similar but confer very different biological behavior. MMAF is conjugated using a noncleavable maleimidocaproyl (mc) linker, which requires degradation of the antibody within the lysosome to release Cys-mc-MMAF. Moreover, MMAF contains a negatively charged carboxylic acid that prevents the released Cys-mc-MMAF toxin from crossing intact biologic membranes (7, 11). In contrast, the inclusion of a peptide cleavable mc-valine-citrulline-p-aminobenzyl-carbamate linker in ABBV-221 enables enzymatic release of MMAE within the lysosome, and the lipophilicity and neutral charge of MMAE enable diffusion across membranes (8, 12). Thus, after release, MMAE can diffuse into and kill adjacent, nontargeted cells. This bystander effect is an important strategy to kill subpopulations of tumor cells that lack high-level ADC target expression but potentially also affects normal cells (13, 14). In this study, ABBV-221 provided robust in vitro bystander cytotoxicity for nontargeted GBM cells and normal astrocytes. In comparing the CED infusion of the two ADCs, both had significant efficacy against GBM tumors; however, ABBV-221 was associated with a higher risk of neuronal toxicity. Pharmacokinetic evaluation of ABBV-221 indicated prolonged exposure of the intact ADC and released MMAE in the brain following CED delivery.

Therapeutic antibodies and clinical grade ADCs were provided by AbbVie, including isotype control (AB095), anti-EGFR antibody (ABT-806), isotype-ADC controls [AB095-MMAE, drug to antibody ratio (DAR): 2.8; AB095-MMAF, DAR: 4.8] and EGFR-targeting ADCs (ABBV-221, DAR: 2.0 and 3.1; Depatux-M, DAR: 4.1). MMAE was obtained from MedChemExpress, Monmouth Junction, NJ (#HY-15162). Commercially available antibody reagents were used for immunohistochemistry staining or Western blotting, as detailed in the Supplementary Data.

GBM patient-derived xenograft (PDX) lines were maintained as previously described (15). Short-term explant cultures of GBM6, GBM10, GBM14, GBM39, GBM43, and GBM108 PDX lines; U87 glioma (ATCC, Cat#HTB-11, RRID:CVCL_GP63); and SVG-A astrocytes (Ximbio, UK, Cat#153563, RRID:CVCL_5G13) cell lines were grown in DMEM (Fisher Scientific, Hampton, NH, Cat#MT10013CV), supplemented with 10% fetal bovine serum (Atlanta Biologicals, Flowery Branch, GA, Cat#S11150) and 1% penicillin/streptomycin (Fisher Scientific, Hampton, NH, Cat#MT30001CI) at 37°C, 5% CO₂. The identity of all the PDXs and U87 cells was verified by short tandem repeat (STR) analysis. Cell lines were tested every 6 months for mycoplasma using the MycoAlert mycoplasma detection kit (Lonza, Switzerland, Cat#LT07-418). Detailed methods for CellTiter-Glo viability assay, antibody internalization, in vitro bystander cytotoxicity, and Western blots are provided in the Supplementary Data. Doses of all treatments in cytotoxicity assay for each line are reported in µg/mL, µmol of antibody, and µmol of toxin in Supplementary Table S1.

Female athymic nude (Strain#553), FVB (Strain#559, RRID:MGI:2160605), and C57BL/6 (Strain#556, RRID:MGI:2160593) mice were purchased from Charles River Laboratories (Wilmington, MA). All animal experiments were approved by the Mayo Clinic Institutional Animal Care and Use Committee, Protocol #A00004595-19 and #A00006636-22. Prior experience with the models and treatment conditions was used to determine the sample size for in vivo studies. Power analysis was not performed. Orthotopic tumor inoculation, bioluminescence imaging, and CED infusions were conducted as previously described (9, 10). The CED parameters are detailed in the Supplementary Data.

For efficacy studies, nude mice with established intracranial tumors were randomized into treatment groups by bioluminescence imaging (BLI) signal or body weight. Doses were selected based on a pilot dose-finding study and published studies (9, 10). CED infusions (20 µL) of AB095 (120 µg/mouse), AB095-MMAE (88 µg/mouse in GBM39; 48 µg/mouse in GBM6 and GBM108), AB095-MMAF (51.2 µg/mouse), ABBV-221 (66 µg/mouse) and Depatux-M (60 µg/mouse) were given every 21 days. For systemic dosing, ABBV-221 (6 mg/kg) or Depatux-M (5 mg/kg) were given intraperitoneally once a week. Mice with orthotopic tumors were observed daily and euthanized upon reaching a moribund state.

Toxicity studies were performed in C57BL/6 mice that received a single CED infusion of AB095 (228 µg/mouse), AB095-MMAE (88 µg/mouse), ABBV-221 (82 µg/mouse), or Depatux-M (740 µg/mouse). Doses for AB095-MMAE and ABBV-221 were selected by matching toxin levels to 60 µg/mouse of Depatux-M based on prior published work (10), and AB095 and Depatux-M were infused at available stock concentrations. Brains were collected and processed by formalin-fixation paraffin-embedding over 14 days post-treatment.

Pharmacokinetics studies were performed in FVB mice with wild-type and BCRP^−/−:MDR1a/b^−/− (triple knockout; TKO) genotypes (breeder pairs from Taconic Biosciences, Germantown, NY). Single injections of MMAE or ABBV-221, at similar doses as used in efficacy studies, were administered either intraperitoneally (IP; 0.5 and 5 mg/kg, respectively) or via CED (570 ng/mouse and 60 µg/mouse, respectively). Blood and brains were collected at multiple time points from 40 minutes to 24 hours from start of CED infusion, and samples were stored at −80°C until subsequent analysis.

Doses in µg and µmol (of Ab and toxin) for the in vivo studies are detailed in Supplementary Table S2.

Tissue sectioning, hematoxylin/eosin staining, and IHC for NeuN and CD68 were performed at the Mayo Clinic Pathology Research Core, and details are provided in Supplementary Data. All IHC-stained slides were scanned at 200× magnification on the Axioscan microscope. Quantification of NeuN was performed using ImageJ (RRID:SCR_003070) Cell Counter by a technician blinded to the treatments, and an average of four high power fields (HPF) per mouse brain were plotted in the graph.

Brains were homogenized in 2× (w/v) 5% BSA. Brain concentrations were corrected for residual blood estimated at 1.4% of brain weight with blood concentrations approximated by plasma concentrations (16). MMAE was analyzed using reverse-phase chromatography tandem mass spectrometry. Released MMAE (MMAE that was not conjugated to an antibody at the time of sample collection) was separated from plasma, the internal standard (MMAF) was added to samples or spiked blank matrix for standards and quality control, and 5× volume of acetonitrile was added for protein precipitation. Samples were vortexed for 5 minutes and centrifuged (13,000 rpm, 10 minutes), and the supernatant was dried using a speedvac. Samples were reconstituted in mobile phase (95% water + 5% acetonitrile + 0.1% formic acid) for injection. For total MMAE analysis, MMAE was first liberated from the antibody using enzymatic deconjugation. Details of the LC-MS/MS assay and the deconjugation protocol are provided in Supplementary Data.

Plasma and brain samples were analyzed for concentration–time profiles of total or released MMAE using Phoenix WinNonlin version 8.4 (Certara Inc, West Windsor Township, NJ). Pharmacokinetic parameters and metrics were calculated by performing noncompartmental analysis. The area under the curve (AUC) was determined by linear trapezoidal integration, where the AUC to the last time point (AUC_Last) was calculated directly. Variance for AUC_Last was calculated using the Bailer method as reported in Phoenix WinNonlin (17).

The brain homogenate was prepared by manually homogenizing FVB brain tissue in 3× (w/v) PBS, pH 7.4. A 50 µL reaction was set up in 50 mmol/L Tris-PBS buffer (pH 7.4). ABBV-221 (2 mg/mL) was incubated in 30 µL FVB brain homogenate on a thermo-block for 24 hours at 37°C, 500 rpm. MMAE was quantitated using LC/MS-MS, and serial dilution of the reaction mix was used for cytotoxicity assays.

Cumulative survival probabilities were estimated using the Kaplan–Meier method. Survival comparisons across groups were performed using the log-rank test. Analysis of variance or two-sample t-tests were used where appropriate. P-values < 0.05 were considered statistically significant.

All data are available on request from the corresponding author.

The cytotoxicity of auristatin and auristatin-conjugated antibodies was evaluated in three EGFRviii-amplified GBM PDXs (GBM6, GBM39, and GBM108) and an EGFR non-amplified (GBM10) PDX. As anticipated, high truncated-EGFR expression was observed only in the EGFRviii-amplified lines (Supplementary Fig. S1A). In a cytotoxicity assay, all four cell lines were highly sensitive to unconjugated MMAE toxin in the subnanomolar range, with GBM39 being exceptionally sensitive to free toxin (Fig. 1A, 50% effective concentrations; EC₅₀ in Supplementary Table S3).

The cytotoxicity of two highly related EGFR-targeted antibodies conjugated to MMAE (ABBV-221) or MMAF (Depatux-M) was evaluated in the same four PDX lines (Supplementary Table S3). Treatments were compared with corresponding nontargeted isotype-matched ADC controls (AB095-MMAE or AB095-MMAF). Both targeted ADCs demonstrated comparable efficacy with EGFR-specific cytotoxicity limited to the EGFRviii-amplified tumors. Of the EGFRviii-amplified lines, GBM39 was the most sensitive and GBM108 was the least sensitive to either ABBV-221 (Fig. 1B) or Depatux-M (Fig. 1C). While GBM10 was insensitive to all ADCs at concentrations of 1 µg/mL and below, nonspecific cytotoxicity was observed for EGFR-targeted and corresponding control ADCs at higher concentrations (Supplementary Figs. S1B and S1C). Thus, compared with isotype control ADCs, only the EGFRviii-amplified PDXs demonstrated EGFR-specific cytotoxicity with ABBV-221 or Depatux-M.

ADC internalization and trafficking to the lysosome were evaluated in parallel to gain further insight into the specificity of cytotoxicity across the four PDXs. Accumulation of the EGFR-specific and control ADCs in the lysosome was evaluated using a Fab-fluor reagent, which binds the Fc region of the antibody and fluoresces at the acidic pH within the lysosomal compartment. Using this assay, ABBV-221 and Depatux-M accumulation was much higher than isotype controls in the EGFRviii-expressing PDXs (GBM6, GBM39, GBM108) but not in GBM10 (Figs. 1D and E; Supplementary Figs. S1D and S1E). Auristatins inhibit microtubule polymerization and exhibit cytotoxicity through the induction of apoptosis. Consistent with EGFR-specific cytotoxicity in GBM6, markers of apoptosis were observed only following ABBV-221 and Depatux-M exposure for 48 hours (Fig. 1F). Collectively, these data confirm the expected EGFR-specific cytotoxicity in GBM PDXs with both ABBV-221 and Depatux-M.

The bystander killing potentials of ABBV-221 and Depatux-M, which differ in cell permeability of the released toxins, were compared using a conditioned media strategy. EGFRviii-amplified (GBM6, GBM108) and non-amplified (GBM10) PDX lines were incubated with various ADCs (1 µg/mL) for 4 days and conditioned media were collected. Using LC-MS/MS to measure released toxin levels, MMAE was 1.5 ± 0.6 and 0.7 ± 0.3 nmol/L in ABBV-221 treated GBM6 and GBM108 conditioned media, respectively, whereas concentrations were below the limit of quantification in AB095-MMAE treated GBM6, GBM108, and all GBM10 conditioned media (Fig. 2A). Because Cys-mc-MMAF released from Depatux-M lacks bystander potential (18, 19), toxin levels were not measured in the corresponding conditioned media.

The cytotoxic potential of conditioned media was evaluated using two EGFR non-amplified PDXs (GBM14 and GBM43) and normal human astrocytic SVG-A cells (Supplementary Fig. S1A). All three of these models are sensitive to unconjugated MMAE (Supplementary Fig. S2A) but not to killing by direct incubation with ABBV-221 or Depatux-M (Supplementary Figs. S2B and S2C; Supplementary Table S3). Conditioned media harvested from EGFRviii-amplified cells (GBM6) treated with ABBV-221 induced significant killing of GBM43 (P < 0.0001 and P < 0.0001), GBM14 (P = 0.002 and P = 0.001) and SVG-A (P = 0.001 and P = 0.001) cells compared with conditioned media harvested after AB095-MMAE or Depatux-M incubation, respectively (Figs. 2B–D). Conditioned media from ABBV-221 treated EGFRviii-amplified GBM108 cells also suppressed proliferation of GBM43 cells (Supplementary Fig. S2D). Consistent with a dose-response for MMAE release, conditioned media collected from GBM6 cells treated with 1 (P < 0.0001) or 0.5 µg/mL (P < 0.0001) ABBV-221 induced bystander killing in GBM43 cells, while conditioned media from 0.1 µg/mL (P = 0.8) treatment had no effect compared with conditioned media from AB095-MMAE treated cells (Supplementary Fig. S2E). Conditioned media obtained from EGFR non-amplified GBM10 cells post ABBV-221 treatment did not affect the viability of GBM14 (P = 0.6), GBM43 (P = 0.6), or SVG-A (P = 0.7) cells compared with AB095-MMAE treated conditioned media (Figs. 2B–D; Supplementary Fig. S2E). These data demonstrate the potential of ABBV-221 to kill EGFR-low cells after treatment of EGFRviii-amplified cells and subsequent toxin release.

The in vivo efficacy of ABBV-221 and Depatux-M was compared after intraperitoneal (IP) injections or intracranial CED infusion in three EGFRviii-amplified orthotopic GBM PDXs: GBM6, GBM39, and GBM108. Doses were selected based on prior published work (9, 10) and survival benefit provided by a single CED infusion of ABBV-221 (2–200 µg/mouse) at four dose levels in a pilot dose-finding study using GBM6 (Supplementary Fig. S3A). GBM6 and GBM39 were transduced with a lentiviral construct encoding eGFP/fLuc2, which enabled BLI to track intracranial tumor growth. CED infusions of either ABBV-221 or Depatux-M significantly reduced BLI signal in GBM6 and GBM39 compared with AB095 or relevant AB095-ADCs. In comparison, IP dosing of either EGFR-targeted ADC resulted in reduced tumor growth of GBM39 but not in GBM6 (Figs. 3A and B; Supplementary Figs. S3B and S3C). CED infusions of AB095-MMAE and AB095-MMAF resulted in reduction of BLI signal in GBM6 compared with AB095 infusion (P = 0.001 and P = 0.0003, respectively), whereas no significant change occurred in GBM39 (P = 1 for both; Supplementary Figs. S3B and S3C). These data suggest nonspecific ADC uptake by GBM6 at the relatively high concentrations delivered to the tumors.

The mice were monitored daily until reaching a moribund state as a definitive read-out for efficacy. CED infusions of ABBV-221 (66 µg/mouse) and Depatux-M (60 µg/mouse) significantly extended the survival of all three PDX models, GBM39 (median: 195 and >300 days; P = 0.03 and P = 0.002), GBM6 (median: 125 and 140 days; P = 0.05 and P = 0.003) and GBM108 (median: >300 and 219 days; P = 0.003 and P = 0.003), compared with AB095 (120 µg/mouse) control (medians: 20, 49 and 36 days, respectively; Figs. 3C and D; Supplementary Table S4). IP injections of ABBV-221 (6 mg/kg) or Depatux-M (5 mg/kg) provided significant survival benefits compared with CED-infused AB095 control in GBM39 (P = 0.02 and P = 0.02, respectively) and GBM108 (P = 0.002 and P = 0.02, respectively), whereas a significant increase in survival in GBM39 with IP ABBV-221 was observed (median: 251 and 68 days, respectively, vs. 20 days) compared with GBM108 (median: 61 and 53 days, respectively, vs. 36 days; Figs. 3C and D), which is likely related to the relatively open BBB in GBM39 (9). The control ADCs AB095-MMAE (GBM6 and GBM108: 48 µg/mouse; GBM39: 88 µg/mouse) and AB095-MMAF (51.2 µg/mouse), infused via CED, extended survival of all three PDX models to varied extents compared with AB095 CED (GBM39: 19 and 58 days; P = 0.4 and 0.002, GBM6: 57 and 93 days; P = 0.3 and 0.003, GBM108: 94 and 17 days; P = 0.02 and 0.6, respectively; Figs. 3C and D; Supplementary Table S4), suggesting potential antitumor activity of nonspecific ADCs at the infused concentrations in the brain tumor microenvironment.

The toxicity of ABBV-221 and Depatux-M delivered by CED infusion was evaluated in immunocompetent, nontumor-bearing mice. After infusion, the mice were euthanized at prespecified time points (7 and 14 days) or earlier if moribund, and brains were processed for histologic analysis. AB095-MMAE (88 µg/animal) and ABBV-221 (82 µg/animal) induced notable pathological changes of apoptosis and neuronal loss, specifically in the infused, ipsilateral brain but not in the contralateral brain, harvested 5 to 14 days after infusion (Fig. 4A). In contrast, CED infusions of AB095 (228 µg/animal) or Depatux-M (740 µg/animal) had minimal effects, even at much higher doses. An increase in activated microglial/macrophage cells as a marker of inflammation was evaluated by staining for CD68, demonstrating an enhanced number of CD68⁺ cells in the infused hemisphere with AB095-MMAE or ABBV-221 compared with Depatux-M or AB095 (Fig. 4B). Finally, the impact on neurons was quantified by analyzing the density of NeuN-positive cells. AB095-MMAE or ABBV-221 infusion resulted in significantly decreased NeuN⁺ cells to 199.6 ± 52.3 (P < 0.0001 and P = 0.0004) or 184.3 ± 84.5 (P = 0.0005 and P = 0.004) neurons per HPF per mouse, respectively, compared with AB095 (353.6 ± 27.2 neurons per HPF per mouse) or Depatux-M (358.9 ± 54.2 neurons per HPF per mouse) infusions, respectively (Figs. 4C and D). These results demonstrate that Depatux-M causes minimal neuronal toxicity and a potentially broader therapeutic window within the brain than ABBV-221.

The concentration–time profiles following intraperitoneal and CED infusions of MMAE (unconjugated toxin) and ABBV-221 were evaluated in immunocompetent FVB mice to gain greater insight into the observed toxicity with the latter. Following bolus IV injection, MMAE (0.5 mg/kg) displayed a rapid distribution in plasma and brain in FVB mice with a terminal half-life of 5.5 hours. Consistent with partial exclusion by the BBB, the ratio of brain to plasma exposure (AUC, Kp_Brain) was 0.1 (Fig. 5A; Supplementary Table S5). To better understand the mechanism of MMAE exclusion from the brain, parallel studies were performed in TKO mice, which lack BCRP and P-gp expression. The ratio of brain to plasma MMAE exposure in TKO mice was 1.2 (Fig. 5A; Supplementary Table S5). CED infusion (570 ng/mouse) into the right hemisphere of wild-type FVB mice resulted in approximately tenfold greater MMAE concentrations in the right versus left hemisphere (P < 0.001; Fig. 5B). Interestingly, significant plasma levels were observed at the end of the CED infusion, but the levels rapidly dropped in the next 2 hours of the infusion (Fig. 5B; Supplementary Table S5). This rapid fall-off in plasma MMAE concentration after CED suggests that the clearance of MMAE from the vascular compartment is rapid.

Similar pharmacokinetic analyses of MMAE were performed following ABBV-221 delivery. By analyzing samples incubated with and without papain, which frees MMAE from ABBV-221, the levels of total and released MMAE were determined by LC-MS/MS. After IP injection of 5 mg/kg ABBV-221, the total MMAE peak plasma concentration was ∼400 ng/mL, and, consistent with a known long half-life of the ADC, levels of total MMAE did not appreciably change over 24 hours. Levels of released MMAE were similarly low in both brain and plasma, consistent with limited distribution of ABBV-221 into the central nervous system (CNS) and linker stability in plasma (Fig. 5C; Supplementary Table S5). Following CED (60 µg ABBV-221 per animal), total MMAE levels were approximately tenfold higher in the right versus left cerebral hemispheres (P = 0.002; Fig. 5D). While plasma levels were initially comparable to the non-infused left hemisphere, these levels continued to rise in the first few hours after infusion. They ultimately reached levels of the infused right hemisphere. The concentrations of total MMAE remained relatively constant over the 24-hour sampling period, whereas levels of released MMAE in the right and left hemispheres continued to rise during this timeframe (Fig. 5D; Supplementary Table S5). These data demonstrate relatively prolonged exposure of released MMAE in the brain following CED infusion of ABBV-221, which partly explains the neurotoxicity associated with MMAE ADCs.

The increase in released MMAE concentration over time in the brains of nontumor-bearing mice suggested that the peptide linker (mc-val-cit-PABC) was potentially unstable in the brain microenvironment. To test this, ABBV-221 (2 mg/mL) was incubated with buffer (50 mmol/L Tris-PBS, pH 7.4) or with FVB brain homogenates for 24 hours, and the levels of released MMAE were measured. In comparison to buffer only (5.3 ± 0.2 nmol/L), incubation with brain homogenate resulted in a nearly 40-fold increase in released MMAE, 197.7 ± 39.6 nmol/L; P = 0.001 (Fig. 6A). To validate the potential for cytotoxicity of the released MMAE, MMAE-sensitive U87 glioma cells (Fig. 6B) were incubated with brain homogenates. As seen in Fig. 6C, marked cytotoxicity was observed in U87 cells specifically incubated with solution collected from the ABBV-221/brain homogenate compared with relevant controls (P = 0.0006). Notably, FVB brain homogenate alone did not cause cytotoxicity in U87 cells (P = 0.09). These results suggest that the instability of the ABBV-221 linker in the brain microenvironment, resulting in the release of MMAE, could contribute to the observed neurotoxicity.

ADCs targeting RTKs are highly effective therapies for solid malignancies expressing sufficient levels of the targeted receptor. The antitumor effects of traditional RTK inhibitors result from disruption of receptor signaling, and the emergence of compensatory signaling pathways is a predominant mechanism of resistance (20, 21). In contrast, ADCs capitalize on differential RTK expression in tumor versus normal tissue to selectively deliver highly toxic payloads specifically to tumor cells. The therapeutic potential of this approach is highlighted by the profound survival benefits conferred by the two FDA-approved ADCs targeting EGFR 2 (HER2). Both trastuzumab emtansine (T-DM1) and trastuzumab deruxtecan (T-DXd) are highly effective in patients with HER2-overexpressing breast cancer who have progressed on HER2-targeted small molecules and therapeutic antibodies (3, 22). The survival benefit afforded by these and other approved ADCs provides a strong rationale to evaluate relevant ADC therapeutics in GBM and other treatment refractory cancers.

High-level EGFR expression is a pathogenic marker for approximately half of patients with GBM and is a rational drug target. Unfortunately, all EGFR-targeted therapies tested so far have failed to provide clinically meaningful survival benefits in patients with GBM (5, 23). While multiple mechanisms contribute to the failure of these therapies, molecular heterogeneity, the emergence of compensatory survival signaling pathways, and limited drug distribution across the BBB are key aspects that must be considered (9, 24). The current article explores parallel strategies to address these issues using CED to achieve high-level drug delivery and an ADC with a cell-permeant MMAE toxin (ABBV-221) to provide bystander killing of EGFR-low subpopulations of tumor cells. CED is a clinically viable technique and was used here as a representative strategy to bypass the BBB (25, 26). Consistent with enhanced and uniform delivery, CED administration improved brain tumor distribution of ABBV-221 and Depatux-M and significantly extended survival in intracranial EGFRviii-amplified GBM PDX models. While protein-toxin conjugates have been safely delivered via CED in clinical trials (27, 28), the elevated neuronal toxicity observed specifically with ABBV-221 suggests that the toxicity of ADCs must be critically evaluated in relation to the delivered toxin and the tissue-specific stability of the chemical linker securing the toxic warhead to the antibody.

Intratumoral heterogeneity of RTK expression is common in GBM (20, 29), and failure to kill EGFR-low subpopulations of cells can result in treatment failure. To overcome RTK heterogeneity issues, next-generation ADCs are being developed using toxins that can diffuse out of targeted tumor cells to kill adjacent tumor cells lacking the ADC target. This potential for bystander killing is one mechanism proposed to account for the greater efficacy of trastuzumab deruxtecan (bystander capable topoisomerase I inhibitor) against HER2-amplified tumors and those expressing lower HER2 levels, compared with T-DM1 (bystander incapable microtubule inhibitor; refs. 13, 30–32). Extending this concept to EGFR-expressing GBM, the MMAE payload of ABBV-221 demonstrated proficient bystander killing in our in vitro assays, which is consistent with the cell-permeable nature of MMAE, compared with the cell-impermeant MMAF toxin from Depatux-M. Unlike heterogeneous EGFR expression typical of most EGFR-amplified patient samples, presumably, the in vivo selection process associated with generating PDXs results in relatively uniform, high-level EGFR expression in EGFR-amplified GBM PDXs (33, 34). This prevented us from testing the bystander killing potential of ABBV-221 in native PDXs. Although mixed orthotopic PDXs could have been used to evaluate bystander killing, the recognition of elevated neuronal toxicity with ABBV-221 diminished enthusiasm for a comprehensive evaluation of this agent’s antitumor bystander effects.

The two EGFR-targeting ADCs used in this study, ABBV-221 and Depatux-M, have very different neurotoxicity profiles following CED infusion. In our previous study, an infusion of 200 µg Depatux-M had no impact on neuronal density (10), and escalation to 740 µg in the current study was similarly well tolerated. In contrast, infusion of ABBV-221 (82 µg) or the nontargeted AB095-MMAE (88 µg) resulted in significant neuronal loss and a corresponding increase in CD68⁺ microglial/macrophage cells associated with an inflammatory response. ABBV-221 uses an affinity matured antibody (AM-1) that differs by only three amino acid residues in the heavy chain complementarity-determining regions compared with the ABT-806 antibody used for Depatux-M (7). Although both ADCs use highly similar auristatin toxins and release significant toxin (Cys-mc-MMAF or MMAE) levels in the brain microenvironment following CED infusion in nontumor-bearing mice (10), the negative charge of MMAF prevents the released toxin from diffusing across intact membranes and entering cells. Thus, Depatux-M is incapable of causing bystander killing of either tumor or normal cells. Although low-level antibody/ADC internalization into normal brain cells has been reported (35, 36), the lack of significant neurotoxicity with Depatux-M suggests that neuronal loss with ABBV-221 is directly related to the bystander-capable cytotoxicity of the latter ADC.

The toxicity studies described here were performed in mice without orthotopic tumors. Thus, the bystander-induced neuronal toxicity observed with ABBV-221 likely resulted from the release of MMAE within the brain microenvironment. We previously reported similar neuronal toxicity following CED infusion of a related EGFR-targeted ADC (Serclutamab talirine; Ser-T), which is built on the same AM-1 as ABBV-221 but with a cell-permeant, potent DNA crosslinking toxin (10, 37). Several mechanisms could contribute to toxin release from ADCs in brain parenchyma, including catalytic activity of extracellular enzymes and uptake/processing by resident brain cells such as microglia using macropinocytosis or Fc-gamma receptor–mediated internalization (38). Both ABBV-221 and Ser-T use valine-citrulline linkers that can be readily cleaved by cathepsin-B within the lysosomal compartment of all cells. Moreover, cathepsin-B and other related family members are present in the brain microenvironment (39). Consistent with enzymatic cleavage in the normal brain, appreciable levels of released MMAE were detectable following CED infusion of ABBV-221 or incubation of this ADC with brain homogenate. Notably, MMAE had a long residence time within the brain following infusion of either unconjugated MMAE or ABBV-221, which could be an important factor contributing to the significant toxicity observed with ABBV-221.

The potential for ADC-induced neuronal toxicity is a critical consideration in the context of evolving strategies to enhance drug delivery into the CNS. Microtubule-targeted chemotherapy agents (vinca alkaloids and taxanes) commonly cause peripheral neuropathy, and neuropathy is a toxicity common to FDA-approved ADCs with microtubule-targeted toxins (MMAE, DM1, DM4; refs. 40–42). The high rate of axonal transport in neurons requires constant mobilization by tubules, which can explain their sensitivity to microtubule inhibitors (36, 43). Although the specific mechanism of toxin uptake into peripheral nerves is not clear, the lower incidence of neuropathy of ADCs with cell-impermeant DM1 or MMAF, compared with cell-permeant DM4 or MMAE toxins, is consistent with a bystander-mediated toxicity effect (44). The exclusion of ADCs and small molecule microtubule inhibitors from the CNS by the BBB can explain the relatively rare incidence of CNS toxicity with these agents. However, multiple strategies are being pursued to enhance drug delivery across the BBB in GBM and other brain tumors, such as intrathecal injection, CED, focused ultrasound, brain-targeted nanoparticles, and receptor-mediated transcytosis (45, 46). In this context, further dissection of mechanisms of neuronal toxicity related to toxin and linker chemistry is essential to maintain a favorable therapeutic window for these treatment strategies.

The dire prognosis of GBM and the potential for significant efficacy with ADC therapy continue to generate enthusiasm for this strategy. Coupled with our prior studies, the presented data highlight the potential for significant antitumor efficacy with EGFR-targeted ADCs in GBM. However, conjugation with different ADC-relevant toxins, such as other microtubule inhibitors, topoisomerase inhibitors, or DNA alkylators, needs further evaluation. Several clinical ADCs targeting cell surface markers upregulated in subsets of GBM, such as Her2, Her3, TROP-2, MET, AXL, and B7-H3, have similar potential for significant therapeutic efficacy (47, 48). Regardless of the targeted marker, improving delivery into tumor regions with an intact BBB is a critical consideration for GBM. CED is a well-established neurosurgical technique that has been used to deliver a variety of small and large-molecule therapeutics in GBM and is one of several strategies to overcome delivery issues (25, 26). Although our studies identified relatively high plasma levels of ABBV-221 and MMAE after CED infusion in nontumor-bearing mice, we hypothesize that this observation reflects catheter reflux associated with a short catheter path length in the mouse brain (49, 50). Nonetheless, the neuronal toxicity and plasma drug levels observed in mice highlight the importance of evaluating the stability, toxicity, and pharmacokinetics of ADC infusion into the brain microenvironment in larger animal models before human testing.

In conclusion, both Depatux-M and ABBV-221, when delivered intracranially via CED, had significant efficacy in EGFRviii-amplified GBM PDX models; however, ABBV-221 demonstrated profound neurotoxicity. While auristatins are highly effective toxins against GBM, improved linker stability within the brain microenvironment will be required to minimize neuronal toxicity, especially when using cell-penetrant, bystander-capable microtubules or other neuronal toxins. The key findings of this study highlight the importance of careful evaluation of ADCs for dose optimization, stability, bystander capability, and payload potency for GBM.

S. Rathi was a student at the University of Minnesota when the work was conducted. S.K. Gupta reports grants from NCI/NIH during the study, and AbbVie provided the ADCs used in the study. J.E. Eckel-Passow reports grants from NIH during the study. E.B. Reilly reports employment with AbbVie. J.N. Sarkaria reports grants from AbbVie Inc during the conduct of the study, as well as grants from Bayer, Wayshine, Black Diamond, Karyopharm, Boston Scientific, Wugen, Rain Therapeutics, Sumitomo Dainippon Pharma Oncology, SKBP, Boehringer Ingelheim, AstraZeneca, ABL Bio, ModifiBio, Inhibrx, Otomagnetics, and Reglagene outside the submitted work. No disclosures were reported by the other authors.

S. Jain: Conceptualization, resources, data curation, software, formal analysis, supervision, validation, methodology, writing–original draft, project administration, writing–review and editing. J.I. Griffith: Conceptualization, data curation, software, supervision, methodology, writing–review and editing. K.A. Porath: Conceptualization, data curation, software, supervision, methodology, writing–review and editing. S. Rathi: Data curation, software, writing–review and editing. J. Le: Data curation, writing–review and editing. T.I. Pasa: Software, writing–review and editing. P.A. Decker: Resources, software, formal analysis. S.K. Gupta: Resources, writing–review and editing. Z. Hu: Resources, data curation, writing–review and editing. B.L. Carlson: Resources, data curation, writing–review and editing. K. Bakken: Resources, data curation, writing–review and editing. D.M. Burgenske: Resources, writing–review and editing. T.M. Feldsien: Resources, writing–review and editing. D.R. Lefebvre: Resources, writing–review and editing. R.A. Vaubel: Resources, writing–review and editing. J.E. Eckel-Passow: Resources, software, formal analysis, writing–review and editing. E.B. Reilly: Resources, writing–review and editing. W.F. Elmquist: Conceptualization, resources, supervision, methodology, project administration, writing–review and editing. J.N. Sarkaria: Conceptualization, resources, supervision, funding acquisition, validation, methodology, writing–original draft, project administration, writing–review and editing.

We thank AbbVie Inc. for providing AB095, ABT-806, AB095-MMAE, AB095-MMAF, ABBV-221, and Depatux-M for this study. This study was funded by grants from the Mayo Clinic, The National Brain Tumor Society, NIH/NCI (U19-CA264362 and U54-CA 210180), and AbbVie Inc.