Abstract
Antibody–drug conjugates (ADC) utilizing noncleavable linker drugs have been approved for clinical use, and several are in development targeting solid and hematologic malignancies including multiple myeloma. Currently, there are no reliable biomarkers of activity for these ADCs other than presence of the targeted antigen. We observed that certain cell lines are innately resistant to such ADCs, and sought to uncover the underlying mechanism of resistance.
The expression of 43 lysosomal membrane target genes was evaluated in cell lines resistant to ADCs bearing the noncleavable linker, pyrrolobenzodiazepine payload SG3376, in vitro. The functional relevance of SLC46A3, a lysosomal transporter of noncleavable ADC catabolites whose expression uniquely correlated with SG3376 resistance, was assessed using EPHA2-, HER2-, and BCMA-targeted ADCs and isogenic cells overexpressing or genetically inactivated for SLC46A3. SLC46A3 expression was also examined in patient-derived xenograft and in vitro models of acquired T-DM1 resistance and multiple myeloma bone marrow samples by RT-PCR.
Loss of SLC46A3 expression was found to be a mechanism of innate and acquired resistance to ADCs bearing DM1 and SG3376. Sensitivity was restored in refractory lines upon introduction of SLC46A3, suggesting that expression of SLC46A3 may be more predictive of activity than target antigen levels alone. Interrogation of primary multiple myeloma samples indicated a range of SLC46A3 expression, including samples with undetectable levels like multiple myeloma cell lines resistant to BCMA-targeting DM1 and SG3376 ADCs.
Our findings support SLC46A3 as a potential patient selection biomarker with immediate relevance to clinical trials involving these ADCs.
Although antibody–drug conjugates (ADC) utilizing the payload DM1 have been approved for clinical use, and several are in clinical development that target both solid tumors and hematologic malignancies, predictive biomarkers beyond antigen expression have yet to be identified. Herein, we show that loss of lysosomal transporter SLC46A3 expression is a mechanism of innate and acquired resistance to ADCs bearing the payloads DM1 or SG3376 (a PBD dimer) with noncleavable linkers, making SLC46A3 a potential biomarker for patient selection in ongoing trials with ADCs bearing these payloads.
Introduction
Antibody–drug conjugates (ADC) combine a mAb with a cytotoxic drug (warhead) to preferentially eliminate antigen-positive cells for the treatment of cancer (1). Four ADCs are approved for clinical use: brentuximab vedotin for the treatment of Hodgkin lymphoma (2), ado-trastuzumab emtansine (T-DM1) for the treatment of erb-b2 receptor tyrosine kinase 2 (ERBB2, HER2)-positive metastatic breast cancer (3), inotuzumab ozogamicin for the treatment of acute lymphoblastic leukemia (4), and gemtuzumab ozogamicin for the treatment of CD33-positive acute myeloid leukemia (5). Upon binding to the targeted antigen on the cell surface, ADCs prepared with enzymatically cleavable linkers (e.g., brentuximab vedotin) or acid-labile hydrazone linkers (e.g., inotuzumab ozogamicin and gemtuzumab ozogamicin) are internalized and processed within the cell, releasing the cytotoxic warhead after linker cleavage. Warheads released in this manner are typically membrane permeable and capable of bystander killing (6–8). In contrast, ADCs with noncleavable linkers, such as T-DM1, rely upon proteolytic degradation of the antibody in the lysosome to release an amino acid linker warhead (9). These catabolites are generally not membrane permeable (10, 11) and therefore require transport from the lysosome to reach their intracellular target (12). Several ADCs utilizing noncleavable linkers are currently in clinical development and target both solid tumors and hematologic malignancies, including diffuse large B-cell lymphoma and multiple myeloma (refs. 13–16).
Although T-DM1 therapy has shown significant clinical benefit, most patients eventually relapse despite continued treatment (17–19). Currently, the only predictive biomarker for response to T-DM1 is overexpression of HER2 (20), which is detected by measurement of HER2 protein levels by IHC or of HER2 gene amplification by fluorescent or chromogenic in situ hybridization (21). Acquired resistance to T-DM1 has been evaluated in several preclinical studies, and due to the complexity of this molecule, several potential mechanisms have emerged: (i) antigen loss and/or downregulation, (ii) increased expression of drug transporters MDR1 (ABCB1) and MRP1 (ABCC1), (iii) defects in ADC trafficking, and/or (iv) changes in receptor and signaling pathways (22). In addition, aberrations in lysosomal pH and proteolytic activity (23) and loss of the lysosomal transporter solute carrier family 46 member 3 (SLC46A3; refs. 12, 24) have been observed in T-DM1–resistant cell lines, highlighting a potential role for the lysosome in T-DM1 resistance.
Until recently, the cytotoxic warheads used for most ADCs in clinical development were based on antimitotic agents, namely, the auristatins and maytansines (1). ADCs prepared with the cleavable linker dimeric pyrrolobenzodiazepine (PBD) payloads talirine (SGD-1910; refs. 25, 26) or tesirine (SG3249; ref. 27) are now entering the clinic. PBD dimers are a class of compounds that form sequence-selective DNA crosslinks in the minor groove of DNA, leading to cell death (28, 29) and thereby offering an alternative mechanism of action for tumor targeting. New classes of PBD payloads with noncleavable linkers (30), including the benzyl tether-linked PBD SG3376 (31), are also being synthesized.
Herein, we report that certain cell lines that are sensitive to ADCs prepared with the cleavable linker drug, SG3249, were innately resistant to those with the noncleavable linker drug, SG3376. Through gene expression profiling of sensitive and resistant cell lines, we discovered that expression of the lysosomal transporter SLC46A3 was required for cytotoxic activity of ADCs prepared with SG3376. Absence of SLC46A3 contributed to innate and acquired resistance to ADCs bearing noncleavable DM1 and SG3376, further suggesting that expression of SLC46A3 is a candidate patient selection biomarker for ADCs incorporating such payloads.
Materials and Methods
Antibodies and generation of ADCs
Antibody cloning, expression, and purification were carried out as described previously (32–34). IgG1-isotype control, 1C1, and trastuzumab antibodies were made at MedImmune. The anti-TNFRSF17 (BCMA) antibody was prepared using VH and VL sequences retrieved from patent application US 2014/0105915 A1. Trastuzumab drug conjugate, T-DM1 (ado-trastuzumab emtansine), was purchased from Blue Door Pharma. All antibodies used in this study, except for those conjugated to DM1 and monomethyl auristatin F (MMAF), were engineered for the site-specific conjugation of two drugs per antibody as described by Dimasi and colleagues (32).
The PBD dimer SG3249 was synthesized at Novasep. mcMMAF and DM1–succinimidyl-4-(N-maleimidomethyl)cyclohexane-1-carboxylate (SMCC) were purchased from ALB Technology. SG3541 was prepared at Spirogen as described in the Supplementary Materials. SG3376 was prepared at Spirogen according to published methods (31). Conjugation of SG3249 (27), SG3376 (31), and SG3541 was performed as described previously (32, 34).
To generate the mcMMAF ADCs with a drug to antibody ratio (DAR) of 4 at the antibodies' native cystines, the antibodies were reduced using 2.5-molar equivalent (MEq) tris(2-carboxyethyl)phosphine in PBS pH 7.2–1 mol/L EDTA for 1 hour at 37°C. The reaction mixture was then incubated with 4 equivalents of mcMMAF for 1 hour at 25°C in PBS pH 7.2–1 mmol/L EDTA–10% v/v dimethylsulfoxide. The reaction was quenched by the addition of 4 equivalents (over the mcMMAF) of N-acetyl cysteine.
To generate the DM1 ADCs with DAR 4 at the antibody lysines, the antibodies at 2 mg/mL in 1 mmol/L borate buffer, pH 8.5, were incubated at room temperature with 8 MEq DM1. The conjugation reaction was monitored with reduced reverse-phase LC/MS as described previously (32, 34), and the reaction was stopped when a DAR of 4 was reached (total reaction time, approximately 1 hour). The ADCs were purified and characterized analytically as described previously (32, 34).
Human cells
Primary multiple myeloma samples were obtained from patients after informed consent was obtained, in accordance with the Declaration of Helsinki and under the auspices of a protocol approved by the Dana-Farber Cancer Institute Review Board. Myeloma cells were purified before RNA isolation as described previously (16). RT112, HT29, and KYSE-410 cells were obtained from the European Collection of Authenticated Cell Lines, and Du145, JIMT-1, JJN-3, OPM2, and KMS-12-BM cells were obtained from Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH. N87-TR and BT-474-TR cells were made at MedImmune as described previously (35). All other cell lines were obtained from the ATCC. Cells were grown according to the manufacturer's instructions. All cell lines were authenticated by short tandem repeat DNA profiling (IDEXX BioResearch Laboratories) and were found to be negative for Mycoplasma using PCR and the MycoSEQ Mycoplasma Detection Assay Kit (4407876, Thermo Fisher Scientific).
Patient-derived xenograft models
The ST1616B HER2+ breast PDX model was established by using tissue from a patient with lung metastasis refractive to T-DM1, and ST1616B/TDR was established from the parent model by using chronic administration of T-DM1 over several passages. Both models were developed and tested at South Texas Accelerated Research Therapeutics (START) according to protocols approved by the International Animal Care and Use Committee. For in vivo studies, tumor fragments were harvested and implanted subcutaneously into the flank of 7- to 8-week old female athymic nude mice (Charles River Laboratories). Animals were matched by tumor volume and randomly assigned to control and treatment groups (n = 3–5 animals per group). Tumor volume and animal weight data were collected electronically with a digital caliper and scale; tumor dimensions were converted to volume with the formula tumor volume (mm3) = width2 (mm2) × length (mm) × 0.52. Study endpoint was a mean control tumor volume of approximately 1.0 to 2.5 cm3; change in tumor volume of each group was compared with the control.
IHC
IHC was performed on formalin-fixed, paraffin-embedded cell pellets or tissue sections mounted on glass slides. N87, N87-TR, BT-474, and BT-474-TR cell pellets were stained using the PATHWAY anti-HER-2/NEU (4B5) rabbit monoclonal primary antibody (790-2991, Ventana Medical Systems). PDX tumor tissue sections were stained using a HER2/NEU (SP3) rabbit mAb (237R-16-ASR, Cell Marque).
In vitro cytotoxicity assays
Cells were seeded into white-walled 96-well plates at a density of 3,000 cells per well. Treatments were added in triplicate the next day and cell viability was determined 3 to 6 days later, using the Cell Titer-Glo Luminescent Cell Viability Assay Kit (G7572, Promega). Luminescence was measured with an EnVision 2104 Multilabel Reader (Perkin Elmer). Cell viability was calculated as a percentage of control untreated cells, and data were analyzed with GraphPad Prism software v7.02 for Windows, using the log [inhibitor] versus response model.
RNA isolation and mRNA expression analysis
RNA was isolated from cell lines with an RNeasy Plus Mini Kit (74134, Qiagen), and cDNA was generated using SuperScript III First-Strand Supermix (18080400, Thermo Fisher Scientific). Each cDNA was preamplified using TaqMan PreAmp Master Mix (4391128, Thermo Fisher Scientific), diluted 1:10 in DNA suspension buffer (T0220, Teknova) and loaded onto a Fludigm dynamic array for analysis on the Fludigm Biomark. Relative gene expression was calculated as 2(– ΔCt) × 1,000, where ΔCt is the Ct value for each gene minus the Ct value of GAPDH for each sample. Ct values of >25 were considered out of range. All TaqMan assays were purchased from Thermo Fisher Scientific, including Hs99999905_m1 (GAPDH), Hs03045080_m1(TNFRSF17), and assays listed in Supplementary Table S1.
Flow cytometry
Cells were stained on ice with their respective antibodies at concentrations indicated in figure legends, followed by detection with goat anti-human IgG (H+L)–AF647 (A-21445, Thermo Fisher Scientific). Fluorescence of live, single cells was determined on a BD-LSRII and data were analyzed using FlowJo software (FlowJo).
Internalization
Cells were incubated on ice with ADCs at a concentration of 5 μg/mL. Cells were washed and suspended in RPMI + 10% FBS and were either placed on ice (time 0) or into an incubator set to 37°C. At desired time points, cells were removed from the incubator and placed on ice. At the end of the experiment, cells were washed and incubated with anti-human-AF647 secondary antibody (A-21445, Thermo Fisher Scientific). Fluorescence of live, single cells was determined on a BD-LSRII and data were analyzed with FlowJo software (FlowJo). Internalization is reported as 100 – [(mean fluorescence intensity (MFI) at each time point/MFI at time 0) × 100].
Lysosomal trafficking
ADCs were labeled with Fab-pHast human secondary fluorescent conjugate (PH-01, ATS Bio) according to the manufacturer's instructions. Cells were incubated with 3 μg/mL of the labeled ADCs at 37°C for desired time points. After washing, the fluorescence of live, single cells was determined on a BD-LSRII and data were analyzed with FlowJo software (FlowJo).
Western blotting
EPHA2 degradation experiments and immunoblotting were performed as described previously (36), using anti-ECK/EPHA2 antibody clone D7 (05-480) and GAPDH (G-8795) antibodies from Sigma-Aldrich. Images were captured using an ImageQuant LAS 4000 mini (GE Healthcare).
Generation of SLC46A3-expressing cells
Full-length human SLC46A3 with a C-terminus Rho1D4 tag was amplified by PCR from TrueORF Gold-validated human cDNA (SLC46A3-NM_181785; OriGene) with the forward primer [5′CCATAGAAGATTCTAGA (GCTAGC) ATGAAGATTTTATTTGTAGAACCTGCCATTTTCC3′] and reverse primer [5′GATCGCAGATCCTT (GCGGCCGC) TTAGGCAGGGGCCACCTGAGATGTCTCGGTCCTGTCTGAAGCATCTTCACTGG3′], and cloned into a pCDH1-CMV-MCS1-EF1-Puro vector (CD510B-1, System Biosciences). Lentiviruses were made with the pPACKH1 HIV Lentivector Packaging Kit (LV510A-1, System Biosciences). Cells were transduced with lentivirus and incubated for 4 days, at which time cells were split into T75 flasks and were selected in 1 μg/mL puromycin (A1113803, Thermo Fisher Scientific) before use.
Generation of SK-BR-3 KO Cells
CRISPR crRNA sequences against human SLC46A3 exon2 (NM_181785) were identified with the CRISPR design tool (crispr.mit.edu; ref. 37). Sequences of two appropriate crRNAs for use in an all-in-one dual nickase CRISPR-Cas9n plasmid containing a GFP tag (pD1401-AD; ATUM) were chosen (38). SK-BR-3 cells were transfected with Lipofectamine 3000 reagent, and GFP-positive cells were collected by cell sorting on a FACSAria Fusion Cell Sorter (BD Biosciences) 48 hours after transfection. Cell populations resulting from limited dilution cloning were screened by Sanger sequencing and qRT-PCR to confirm CRISPR knockout of SLC46A3. Relative quantitation was performed with TaqMan assays Hs01308309_cn (SLC46A3) and Hs99999901_s1 (18S; Thermo Fisher Scientific).
Analysis of microarray data
Raw gene expression data from primary multiple myeloma bone marrow samples (39) was obtained from the Gene Expression Omnibus (accession no. GSE6477) and normalized with the frozen robust multiarray analysis (40), using the R statistical software package. In cases where multiple Affymetrix probes mapped to an individual gene, probes with the highest interquartile range were selected as representative for that gene. This resulted in an expression dataset across unique genes. To identify genes that were differentially expressed among relapsed/refractory multiple myeloma, newly diagnosed multiple myeloma, and healthy donors, ANOVA was run for each gene across disease groups. Tukey Honest Significant Difference test was applied to determine pairwise statistical differences between groups. Fold changes for each gene were also calculated between pairs of disease groups to determine the magnitude of change of expression of each gene in the dataset. P values resulting from this analysis were adjusted for multiple comparisons by determining the FDR for each gene, using the Benjamini–Hochberg procedure (41).
Statistical analysis
Data are presented as mean ± SD or mean ± SEM as stated in the figure legends. Statistical analysis was performed with GraphPad Prism software v7.02 for Windows. Statistically significant differences were tested by using specific tests as indicated in the figure legends. P < 0.05 was considered statistically significant.
Results
Innate resistance to ADCs containing the noncleavable linker drug SG3376
The Eph receptor A2 (EPHA2) tyrosine kinase is expressed on the surface of tumor cells and can be specifically targeted by the human mAb 1C1 for the delivery of cytotoxic warheads (36). We used a panel of EPHA2-positive cell lines to compare the cytotoxic activity of ADCs that had been prepared by conjugating antibody 1C1 to PBD payloads containing cleavable (1C1-SG3249) or noncleavable (1C1-SG3376) linkers (Fig. 1A). Although most cell lines were sensitive to both ADCs, Hep G2 and T24 were uniquely resistant to the noncleavable ADC (1C1-SG3376; Fig. 1B). Resistance to the noncleavable ADC could not be explained by target expression, because T24 and ES-2 cells express nearly equivalent levels of EPHA2 (MFI ratio, 68.9 and 72.2 for T24 and ES-2, respectively), but exhibit strikingly different sensitivities to the noncleavable ADC (Fig. 1B and C).
Expression of SLC46A3 correlates with sensitivity to noncleavable 1C1-SG3376 ADC
Effective ADCs bind to their intended target on the tumor cell surface and are subsequently internalized and trafficked to the lysosome for efficient release of the cytotoxic warhead (1). Therefore, we next used flow cytometry to examine binding and internalization of 1C1-SG3249 and 1C1-SG3376 in T24 and ES-2 cells. 1C1-SG3376 and 1C1-SG3249 bound equally well to both cell lines (Fig. 2A) and were similarly internalized (Fig. 2B). Intracellular trafficking was assessed by labeling ADCs with a pH-sensitive dye that only fluoresces when exposed to low-pH environments and using flow cytometry to monitor internalization kinetics. MFI increased over time (Fig. 2C), indicating that both ADCs were internalized and delivered to the acidic endosomal/lysosomal compartment. Finally, EPHA2 protein degradation was observed after treatment of T24 and ES-2 cells with 1C1-SG3249 and 1C1-SG3376 (Fig. 2D), demonstrating that antibody–EPHA2 complexes were delivered to, and processed by, the lysosome in both cell lines.
Because the binding and intracellular trafficking of the 1C1-SG3249 and 1C1-SG3376 ADCs were similar for the sensitive and resistant cells, we hypothesized that catabolites of 1C1-SG3376 may rely upon active transport to leave the lysosome and that resistant cells lack the transporter(s) necessary for lysosomal escape. We examined 43 target genes focused on lysosomal membrane transporter genes (42) by multiplex RT-PCR, using TaqMan gene expression assays on a 96.96 Fluidigm Dynamic Array to identify genes with expression levels correlating with cell killing. Of the 43 genes examined, only the expression of the lysosomal transporter SLC46A3 correlated with 1C1-SG3376 sensitivity in our cell panel (Supplementary Table S1; Fig. 2E), suggesting that the transporter may play a role in sensitivity of a cell to the noncleavable ADC.
Silencing of SLC46A3 alters the potency of SG3376 and DM1 ADCs
To further define the role of SLC46A3 in SG3376 ADC activity and to determine whether the above findings were target- or payload-specific, we conjugated trastuzumab to cleavable linker (SG3249) or noncleavable linker (SG3376 and MMAF) payloads. We then tested their cytotoxic activity along with T-DM1 in SK-BR-3 cells and in SK-BR-3 knockout cells lacking SLC46A3 (SK-BR-3 KO). Sequencing and quantitative PCR confirmed complete knockout of SLC46A3 (Supplementary Fig. S1A), and flow cytometry was used to confirm that HER2 levels were similar in the SK-BR-3 and SK-BR-3 KO cells (Supplementary Fig. S1B). Knockout of SLC46A3 markedly decreased the potency of T-SG3376 and T-DM1 (Fig. 3A and B), but did not affect the potency of T-SG3249 (Fig. 3C). This finding suggests that the role of SLC46A3 is neither target- nor payload-dependent, despite the structural dissimilarity between PBDs (Fig. 1A) and maytansinoids (Fig. 3D). Loss of SLC46A3 expression had no impact on the potency of T-MMAF (Fig. 3E). This result is in agreement with previously published data (12) and suggests that not all noncleavable linker drugs require SLC46A3 for their activity. Similar to SG3249, SG3541, a protease-cleavable analogue of SG3376, was conjugated to trastuzumab (T-SG3541; Fig. 3D) and was found to be active in both parental SK-BR-3 and SK-BR-3 KO cells lacking SLC46A3 (EC50, 2.8 and 2.1 ng/mL, respectively; Fig. 3F).
Diminished SLC46A3 expression in cells with acquired T-DM1 resistance
Given the role of SLC46A3 in the transport of T-DM1 catabolites from the lysosome, we hypothesized that loss of this transporter may be a mechanism of acquired resistance. T-DM1–resistant cell lines were established by treating NCI-N87 and BT-474 cells with T-DM1 in vitro at gradually increasing concentrations until stable resistant lines emerged; these were named N87-TR and BT-474-TR (35). These cells are refractory to at least 1 μg/mL T-DM1 in vitro but are sensitive to T-MMAF (Supplementary Fig. S2), suggesting that lysosomal proteolytic function is retained in these cells. Although antigen loss is a common mechanism of acquired resistance to T-DM1 in preclinical models (43–45), we confirmed that both cell lines maintained a high level of HER2 (HER2 3+), like the parental cells as measured by IHC (Fig. 4A) and flow cytometry (Fig. 4B). Using qRT-PCR, SLC46A3 was nearly undetectable in both N87-TR and BT-474-TR cells as compared with their parental lines (Fig. 4C).
We reasoned that if loss of SLC46A3 expression is the primary mechanism of T-DM1 resistance in these cells, they would also be cross-resistant to T-SG3376. Indeed, cell killing assays showed that, although T-SG3376 ADCs were active in N87 and BT-474 parental cell lines, they were completely inactive in the T-DM1–resistant N87-TR and BT-474-TR lines (Fig. 4D and E). We next forced expression of SLC46A3 into both T-DM1–resistant (TR) and parental cells and evaluated the cytotoxicity of T-DM1 and T-SG3376. Lentiviral expression of SLC46A3 into N87-TR and BT-474-TR cells restored sensitivity to T-DM1 and T-SG3376 (Fig. 4D–G). Notably, the enhanced levels of SLC46A3 in the N87 and BT-474 transfectants augmented maximum cell killing and potency for the T-SG3376 ADCs (Fig. 4D and E), but not for the T-DM1 ADCs (Fig. 4F and G). Flow cytometry confirmed that HER2 levels were not increased by forced expression of SLC46A3 (Fig. 4B), and therefore, the increased activity observed in the SLC46A3 transfectants cannot be explained by an increase in targeted antigen. Similar data were obtained using the ADC 1C1-SG3376 in N87 cells, which are positive for EPHA2 (Supplementary Fig. S3).
Diminished SLC46A3 expression in a T-DM1–resistant PDX model
A lung biopsy collected from a patient 1 month after relapse from a 13-month response to T-DM1 therapy was used to generate a patient-derived xenograft (PDX) model in immunocompromised mice (46). T-DM1 was found to be efficacious in the parent ST1616B model, but chronic treatment of the model with T-DM1 over three passages in mice resulted in a T-DM1–insensitive model, designated ST1616B/TDR that recapitulated the drug resistance observed in the patient. Weekly administration of 5 mg/kg T-DM1 caused tumor regression in the sensitive model (ST1616B), whereas weekly dosing of T-DM1 at a higher dose of 10 mg/kg was inactive in the resistant (ST1616B/TDR) model (Fig. 5A). We next evaluated HER2 expression in the sensitive and resistant PDX tumors by IHC (Fig. 5B). Both models expressed a high level of HER2, ruling out antigen loss as a mechanism for this resistance. Finally, we isolated RNA from ST1616B and ST1616B/TDR tumor tissue and measured SLC46A3 gene expression by qRT-PCR. Average levels of SLC46A3 expression in the ST1616B/TDR tumors were reduced by 92% as compared with the sensitive ST1616B model (Fig. 5C).
SLC46A3 as a potential patient selection biomarker for DM1 and SG3376 ADCs
Although loss of SLC46A3 expression may be a mechanism of acquired resistance to T-DM1, the frequency of SLC46A3-driven innate resistance in HER2+ tumors is unclear. Comparison of mRNA expression across several tumor types in the publicly available Cancer Cell Line Encyclopedia database (ref. 47; Supplementary Fig. S4) showed that breast cancer cell lines expressed relatively high levels of SLC46A3, whereas cell lines derived from prostate, multiple myeloma, esophageal, and upper aerodigestive tract malignancies were among the lowest expressers. In agreement with these data, we measured SLC46A3 expression in a panel of 16 HER2+ cell lines by qRT-PCR and found only the esophageal squamous-cell carcinoma line, KYSE-410, to be without expression of this transporter (Supplementary Fig. S5). In contrast, five of nine (56%) multiple myeloma cell lines had undetectable SLC46A3 (Fig. 6A).
B-cell maturation antigen (BCMA) is universally expressed in multiple myeloma (16, 48) and is being targeted in clinical trials with ADCs bearing the noncleavable linker drugs, DM1 and MMAF (14, 15). Thus, the expression of BCMA and SLC46A3 in multiple myeloma cell lines and in CD138+ cells isolated from the bone marrow of a patient with multiple myeloma was measured by qRT-PCR. Expression of BCMA varied across the primary multiple myeloma samples, only 9 of 99 (9.1%) of which expressed levels as high as the multiple myeloma cell line, NCI-H929 (Fig. 6B). We detected a wide range of SLC46A3 expression in the primary multiple myeloma samples, including several samples with undetectable levels of SLC46A3 (Fig. 6C). Likewise, we analyzed gene expression data from a recently published study (39) and found that SLC46A3 expression was significantly decreased in newly diagnosed (1.8-fold decrease, FDR < 0.001) and relapsed/refractory MM (2.2-fold decrease, FDR < 0.001) as compared with bone marrow taken from healthy donors (Fig. 6D). We then prepared anti-BCMA ADCs conjugated with either cleavable linker (SG3249) or noncleavable linker drugs (SG3376, DM1, and MMAF) and measured their cytotoxic activity in SLC46A3-positive NCI-H929 or SLC46A3-negative MM.1R and JJN-3 MM cell lines. The noncleavable DM1 and SG3376 ADCs were cytotoxic to the NCI-H929 cells, but were inactive in the MM.1R and JJN-3 cells, whereas all cell lines were sensitive to ADCs prepared with SG3249 and MMAF (Fig. 6E and F). Next, we forced expression of SLC46A3 in the JJN-3 cell line (JJN-3–SLC46A3) to evaluate the impact of increased SLC46A3 expression in a cell line with BCMA levels that were more representative of primary multiple myeloma (49). Introduction of SLC46A3 into JJN-3 cells did not alter surface BCMA levels as compared with parental cells (MFI ratio, 1.3 and 1.2 in JJN-3 and JJN-3–SLC46A3 cells, respectively) but rendered them highly sensitive to BCMA-Ab–SG3376, shifting the EC50 of this ADC from >500 ng/mL in JJN-3 cells to 5.5 ng/mL in the JJN-3–SLC46A3 cells (Fig. 6F). Likewise, the cytotoxic activity of BCMA-Ab–DM1 was markedly improved in JJN-3–SLC46A3 cells (EC50, 441 ng/mL) as compared with the parental cells, where it was completely inactive. The potency of the BCMA-Ab–SG3249 and BCMA-Ab–MMAF ADCs was unchanged, having an EC50 of 8–10 ng/mL for BCMA-Ab–SG3249 and 13–21 ng/mL for BCMA-Ab–MMAF in JJN-3 and JJN-3–SLC46A3 cells, respectively (Fig. 6F).
Discussion
Lysosomes are highly acidic, intracellular organelles that function in numerous physiologic processes, including breakdown of macromolecules, autophagy, plasma membrane repair, and metabolism. ADCs bearing noncleavable linker drugs rely upon lysosomal degradation of the antibody to release an amino acid linker warhead, which must then escape the lysosome to bind its intracellular target (11). The lysosomal membrane transporter, SLC46A3, was recently identified to regulate the efflux of noncleavable DM1 catabolites from the lysosome to the cytoplasm (12) and contribute to T-DM1 resistance in vitro (24). Herein, we report that SLC46A3 is also necessary for cytotoxic activity of noncleavable ADCs constructed with a PBD payload, SG3376, which is structurally and mechanistically distinct from DM1. Furthermore, we show that loss of SLC46A3 expression leads to acquired and innate resistance to these ADCs.
Modifications to the basic components of an ADC (antibody, linker, or warhead) can influence the efficacy of the molecule, including its activity in drug-resistant settings.
For example, replacement of a noncleavable linker with a protease-cleavable linker could overcome acquired T-DM1 resistance in an in vitro–derived T-DM1–resistant cell line, 361-TM, even when warheads with a similar mechanism of action are used (43). Similarly, we found that SG3541, a protease-cleavable analogue of SG3376, exerted full activity in SLC46A3-negative cell lines that are resistant to the noncleavable linker-drug, SG3376. Likewise, the cleavable linker-drug, SG3249 (tesirine), which releases the warhead SG3199 having the same DNA cross-linking mechanism of action as SG3376 and SG3541, was equally active in SLC46A3-positive and SLC46A3-negative cell lines.
Modification of the warhead to an alternate mechanism of action has also been shown to overcome drug resistance and is routinely employed by next-generation ADCs. For example, non–Hodgkin lymphoma tumor models with acquired resistance to anti–CD22-vc-MMAE have been found to be sensitive to an ADC-delivering the DNA inhibitor–based warhead, PNU-159682, via the same cleavable linker previously used with MMAE (50). The anti-CD33 ADC SGN-CD33A, which bears the PBD payload talirine, has been shown to be cytotoxic to MDR1-positive cells that are resistant to the CD33-targeting ADC gemtuzumab ozogamicin, which bears a calicheamicin warhead (26). In this study, the T-DM1–resistant cell models, BT-474-TR and N87-TR, demonstrated that acquired resistance may span distinct warhead classes with unique mechanisms of action. A key common feature of SG3376 and DM1 catabolites appears to be an exposed terminal amino acid (i.e., lysine, cysteine), but this feature is also shared in the catabolite of MMAF ADCs, which were not reliant on SLC46A3 for activity. This latter fact suggests that further work is necessary to determine what shared component(s) of these catabolites is being recognized by SLC46A3.
We noticed that forced SLC46A3 expression in the N87, N87-TR, BT-474, and BT-474-TR cell lines enhanced maximum cell killing and potency, as compared with parental lines, for T-SG3376 but not T-DM1 ADCs. Because HER2 antigen levels were similar across these cell lines, these data may indicate that catabolites of DM1 are better substrates than SG3376 for SLC46A3 transport, thereby requiring a lower threshold of transporter expression than SG3376 to deliver a therapeutically active amount of drug to the cytosol. T-DM1 is conjugated through lysines and T-SG3376 is conjugated through cysteines, which could suggest that lysine is a better substrate than cysteine for SLC46A3. This could also be explained by the differences in the mechanisms of action of DM1 and SG3376. In contrast to DM1, whose biological target (microtubules) resides in the cytosol, the catabolites of SG3376 must be further trafficked from the cytosol and enter the nucleus through the nuclear pore complexes to bind to DNA. Increased levels of SG3376 in the cytosol, facilitated by the increase in SLC46A3 in the lysosome, would enable the diffusion and accumulation of drug within the nucleus, leading to a more potent biological effect. Small molecules with a molecular mass of less than 30–60 kDa (51, 52) can pass through the nuclear pore complex by passive diffusion. SG3376 and its predicted cysteine-adduct catabolite have a molecular mass of 1.070 and 1.191 kDa, respectively. Therefore, it is unlikely that active transport regulates the influx of SG3376 catabolites into the nucleus.
The phase III TH3RESA trial of T-DM1 in patients with HER2-positive breast cancer previously treated with two or more HER2-targeted therapies demonstrated an objective response rate of 31%, suggesting that at least in some patients, HER2 remains an attractive target for advanced-stage breast cancer (53). In some patients, this disease is refractory or innately resistant to T-DM1 therapy, and the use of archival diagnostic tumor tissue for determination of HER2 expression in the TH3RESA trial is an obstacle for determining the underlying mechanisms of T-DM1 resistance (54). Most patients who initially respond to T-DM1 treatment eventually relapse, despite continued treatment (17). A tumor from one such patient was used to derive a PDX model that offered a unique opportunity to examine SLC46A3 levels in a clinically relevant model with known levels of HER2 expression and T-DM1 sensitivity. Although this model indicated that loss of SLC46A3 expression is a mechanism of acquired resistance, the expression of HER2 and SLC46A3 in additional clinical biospecimens collected before and after disease progression on T-DM1 will need to be evaluated to confirm these findings.
Cells that lack SLC46A3 expression are resistant to noncleavable DM1 and SG3376 ADCs, regardless of the targeted antigen. Therefore, SLC46A3-driven resistance is not likely to be confined to T-DM1 but may instead be a shared mechanism of resistance among ADCs bearing these payloads. Multiple myeloma is currently being targeted in clinical phase I trials by two anti-BCMA ADCs conjugated via a noncleavable linker to either DM1 (14) or MMAF (55). In this study, SLC46A3-negative multiple myeloma cells were resistant to BCMA-targeting ADCs with DM1 and SG3376 payloads, but were sensitive to those with SG3249 or MMAF. In primary multiple myeloma bone marrow samples, a range of SLC46A3 expression was evident, including several samples with little to no detectable transcript. Our data suggest that SLC46A3-low or SLC46A3-negative multiple myeloma may be resistant or less responsive to treatment with ADCs bearing DM1 or SG3376. Conversely, high levels of SLC46A3 may indicate patients in whom DM1 or SG3376 ADCs may be more efficacious and may be more predictive of activity than target antigen expression alone. This is exemplified by the JJN-3–SLC46A3 model, in which the introduction of SLC46A3 produced a dramatic increase in sensitivity to BCMA-Ab–DM1 and BCMA-Ab–SG3376 ADCs as compared with parental JJN-3 cells, which were refractory to these ADCs. Our data suggest that SLC46A3 levels may be an important predictor of sensitivity, particularly in cells expressing low levels of surface antigen. Taken together, our findings support SLC46A3 as a potential patient selection biomarker with immediate relevance to ongoing or planned clinical trials involving noncleavable DM1 and SG3376 ADCs, regardless of their target antigens and tumor types.
Disclosure of Potential Conflicts of Interest
K. Kinneer is an employee of MedImmune and holds ownership interest (including patents) in AstraZeneca. J. Meekin is an employee of MedImmune and holds ownership interest (including patents) in AstraZeneca. S. Phipps holds ownership interest (including patents) in AstraZeneca. C.M. Kiefer holds ownership interest (including patents) in AstraZeneca. M.C. Rebelatto holds ownership interest (including patents) in MedImmune and AstraZeneca. N. Dimasi holds ownership interest (including patents) in AstraZeneca. S. Sridhar holds ownership interest (including patents) in AstraZeneca. R. Herbst is an employee of MedImmune. P.W. Howard holds ownership interest (including patents) in AstraZeneca. No potential conflicts of interest were disclosed by the other authors.
Authors' Contributions
Conception and design: K. Kinneer, S.J. Gregson, N. Dimasi, M.J. Wick, P.W. Howard, D.A. Tice
Development of methodology: K. Kinneer, J. Meekin, A.C. Tiberghien, Y. Tai, S. Phipps, C.M. Kiefer, N. Dimasi, M.J. Wick, P.W. Howard
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Kinneer, J. Meekin, Y. Tai, S. Phipps, C.M. Kiefer, N. Dimasi, A.D. Moriarty, K.P. Papadopoulos, M.J. Wick, K.C. Anderson, P.W. Howard
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Kinneer, J. Meekin, Y. Tai, C.M. Kiefer, N. Dimasi, S. Sridhar, M.J. Wick, P.W. Howard, D.A. Tice
Writing, review, and/or revision of the manuscript: K. Kinneer, A.C. Tiberghien, Y. Tai, S. Phipps, C.M. Kiefer, M.C. Rebelatto, N. Dimasi, A.D. Moriarty, K.P. Papadopoulos, S. Sridhar, S.J. Gregson, M.J. Wick, L.A. Masterson, K.C. Anderson, R. Herbst, P.W. Howard, D.A. Tice
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.D. Moriarty
Study supervision: K. Kinneer, N. Dimasi, M.J. Wick, D.A. Tice
Responsible for IHC assays and interpretation of staining: M.C. Rebelatto
Provision of synthetic compounds used in the study: L.A. Masterson
Acknowledgments
We gratefully acknowledge Binyam Bezabeh, Ben Ruddle, and Ryan Fleming (MedImmune) for the preparation and analytical characterization of the ADCs used in this study; Terrence O'Day (MedImmune) for statistical support; and Radhika Rayanki (MedImmune) for carrying out flow cytometric sorting of the SK-BR-3 KO cells. We thank Deborah Shuman (MedImmune) for careful review and editing of the manuscript and Sepi Farshadi (MedImmune) for help with the preparation of figures. We also thank Megan Groves (START) for her technical contributions to the PDX studies. This study was funded by MedImmune, the global biologics R&D arm of AstraZeneca.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.