Mechanisms of Acquired Resistance to Trastuzumab Emtansine in Breast Cancer Cells

This article is featured in Highlights of This Issue, p. 1353

The HER2/erbB2 oncogene encodes a 185-kDa transmembrane receptor tyrosine kinase (RTK) that belongs to the EGFR family and regulates proliferation, differentiation, apoptosis, and metastasis (1). HER2 gene amplification occurs in approximately 20% of breast cancer and is associated with increased disease recurrence and poor prognosis (2). Trastuzumab (Herceptin), a humanized antibody directed against the HER2 extracellular domain (3), is approved for treating HER2-positive breast cancer in the metastatic (4) and adjuvant settings (5). Mechanisms attributed to trastuzumab activity include inhibition of HER2/HER3/PI3K signaling (6, 7), prevention of HER2 ectodomain shedding (8), initiation of G1 arrest via induction of the cyclin-dependent kinase inhibitor p27^KIP1 (9), and inhibition of angiogenesis (10). In addition, trastuzumab engages Fc receptor-expressing immune effector cells to induce antibody-dependent, cell-mediated cytotoxicity (ADCC; ref. 9).

Despite clinical activity, a proportion of patients do not respond to trastuzumab because of de novo or acquired resistance (11). Potential mechanisms of resistance include PTEN loss and enhanced AKT signaling (6), altered receptor–antibody interaction (12), and activation of other RTKs (13). Therefore, identification of new therapeutic agents that are effective in trastuzumab-refractory tumors is key for improving survival of breast cancer patients whose tumors overexpress HER2.

Trastuzumab emtansine (T-DM1, Kadcyla) is an antibody–drug conjugate (ADC) comprised of trastuzumab covalently bound to DM1 via a nonreducible thioether linker (4-[N-maleimidomethyl]cyclohexane-1-carboxylate, MCC). DM1, a maytansine derivative, is a potent antimitotic agent that binds microtubules similar to vinca alkaloids. T-DM1 allows intracellular delivery of DM1 selectively to HER2-overexpressing cancer cells, thereby minimizing exposure of normal tissue to the toxicities of the DM1 component and improving the therapeutic index. In preclinical models, T-DM1 potently inhibits growth of trastuzumab-sensitive and -insensitive HER2-amplified cancer cells (14). In addition, T-DM1 retains the antitumor properties of trastuzumab (15). T-DM1 was first approved by the FDA based on the phase III EMILIA trial (NCT00829166), which demonstrated that T-DM1 significantly prolonged progression-free and overall survival compared with the control arm, lapatinib plus capecitabine, in HER2-positive metastatic breast cancer (mBC) patients previously treated with trastuzumab and a taxane (16). Moreover, treatment with T-DM1 is better tolerated than chemotherapy-containing regimens.

Despite improved clinical outcome, some patients initially respond to T-DM1 treatment but develop disease progression. Primary resistance can also occur (16). Thus, resistance to T-DM1 poses a challenge in therapy for HER2-positive breast cancer. A better understanding of the molecular mechanisms of primary and acquired resistance to T-DM1 is particularly important for the development of new therapeutic strategies. To explore mechanisms of acquired resistance to T-DM1, we established resistant cells with a stepwise escalation method and identified different features of T-DM1–resistant cells. Our findings show the complexities of T-DM1 resistance in that our two models showed little overlap in identified resistance mechanisms. The data also suggest possible therapeutic strategies using combinations with inhibitors that target signal transduction resistance pathways for overcoming T-DM1 resistance.

KPL-4 breast cancer cells were a gift from J. Kurebayashi (17). BT-474M1 is an in vivo--passaged subline of BT-474 (7). Cells were cultured as previously described (14). Tetracycline (Tet)-free FBS (Hyclone) was used for maintaining BT-474M1 cells infected with inducible shRNA constructs. T-DM1–resistant lines were generated by continuous exposure to increasing concentrations of T-DM1 up to 2 μg/mL. Parental cells are designated KPL-4 P and BT-474M1 P. Established T-DM1–resistant KPL-4 and BT-474M1 pools were maintained in culture with 2 μg/mL T-DM1, hereafter designated as KPL-4 TR and BT-474M1 TR, respectively. TR-1, used for xenograft studies, are resistant cells grown without T-DM1 for 6 months. Cell line authentication was performed at Genetica DNA Laboratories using short tandem repeat loci analysis. The profiles of BT-474M1 parent and TR cells were a 100% match to reference BT-474 (ATCC HTB-20). KPL-4 parent and TR cells were 92.6% and 100% identical to the reference KPL-4 profile. The results support authentication of all lines compared with the reference profiles.

T-DM1, trastuzumab, mu4D5 (murine parent of trastuzumab), anti–IGF-1R 10H5 (18), GDC-0941 (19), and lapatinib were produced at Genentech. Monomethyl auristatin E (MMAE) was from Seattle Genetics; docetaxel from Sigma Aldrich, and cetuximab from Merck. XR9051 (20), Ko143 (21), and PHA665725 (22) were from Tocris; AMD3100 (23) from Sigma, and IGFBP5 was from R&D Systems. Alexa Fluor 488 anti-human IgG was obtained from Molecular Probes. Free DM1 (S-methyl-DM1) was obtained from ImmunoGen. As DM1 has a free sulfhydryl and can oxidize to produce DM1 disulfide dimers or undergo disulfide exchange with cysteine in tissue culture medium (24), the stable S-methyl-DM1 was used.

Cells were seeded in 96-well plates and allowed to adhere overnight. Media were removed and replaced with fresh media containing various concentrations of each drug. Cell Titer-Glo (Promega) was added after 5 days, and the luminescent signal measured using an EnVision Multilabel Plate Reader (PerkinElmer). For experiments with different inhibitors, T-DM1 and the inhibitors were added simultaneously. Data are represented as mean ± SEM, with n = 4 per treatment group.

Western immunoblot procedures were described previously (14). Antibodies were against EGFR (MBL), HER2 (Neomarkers), PTEN (Santa Cruz Biotechnology), HER3, AKT, phospho-AKT, ERK, phospho-ERK, IGF-1Rβ, MDR1, and β-actin (all from Cell Signaling Technology), DARPP-32 (Epitomics), Met (Upstate Biotechnology), or BCRP (Kamiya Biomedical).

Cells were harvested, embedded in paraffin and formalin-fixed. The PathVysion probe set, HER2 in SpectrumOrange, CEP 17 in SpectrumGreen (Abbott Molecular) was used in the dual-color FISH assay performed according to the manufacturer's protocol. Analysis was performed on an epifluorescence microscope using single interference filter sets for green (FITC), red (Texas red), and blue (DAPI), as well as dual (red/green) and triple (blue, red, green) band pass filters. At least 20 metaphase spreads and 100 interphase nuclei were analyzed in each cell line. Chromosome designation followed ISCN guidelines. Images were captured using the CytoVision software (Genetix). Gene amplification was defined as a HER2/CEP17 FISH ratio signal > 2.2.

Total RNA was extracted using the RNeasy Mini Kit (QIAGEN) according to manufacturer's instructions. Individual primer/probe sets were from Applied Biosystems or designed with Primer Express software (Applied Biosystems). Total RNA (100 ng) was used as template, and Taqman One-Step Universal Master Mix (Applied Biosystems) was used for all reactions. Reactions were performed in a 96-well plate using ABI 7500 Real-Time qPCR System. Gene expression was normalized using HP1BP3 (heterochromatin protein 1, binding protein 3) as the house-keeping gene.

Total RNA (3 μg) was converted into double-stranded cDNA using a SuperScript Choice kit (Invitrogen) and a T7-(dT) primer (Biosearch Technologies, Inc.). cDNA was purified using a Sample Cleanup Module kit (Affymetrix) and used to generate biotin-labeled cRNA using an in vitro transcription kit (Enzo Diagnostics, Inc.). Labeled cRNA was purified using a Sample Cleanup Module kit. Labeled cRNA (15 μg) was fragmented and hybridized to Human Genome U133 Plus 2.0 Arrays following the manufacturer's protocol. Arrays were washed and stained in the Affymetrix Fluidics station and scanned on GeneChip scanner 3000. Data analysis was performed using the Affymetrix GeneChip operating system and analysis software. Triplicates of each cell line (parental vs. resistant) were analyzed.

SMARTpools, individual siRNA oligonucleotides, and nontargeting siRNAs were purchased from Dharmacon/Thermo Scientific. For reverse transfection, Opti-MEM medium was mixed with siRNA to give a final concentration of 25 nmol/L; this was then combined with diluted DharmaFECT 4 (Thermo Scientific). After 20-minute incubation at room temperature, the transfection mixture was aliquoted into 96-well plates. Cells were added to each well containing siRNA and DharmaFECT complex. Forty-eight hours after transfection, cells were treated with T-DM1, and cell viability was measured after 96 hours.

Doxycycline-inducible PTEN shRNA in BT-474M1 cells was produced via lentiviral transfection with an eGFP-tagged PTEN as described (25). Fluorescence-activated cell sorting was used to select eGFP-positive cells that were collected and pooled 3 days after infection. Expanded pools were treated with 25 ng/mL doxycycline (BD Clontech) for 3 days, and endogenous PTEN knockdown assessed.

The experiments were carried out in accordance with and approved by an Institutional Animal Care and Use Committee. BT-474M1 (5 million cells, plus estrogen supplementation) or KPL-4 (3 million cells) were implanted, in Matrigel, into the number 2/3 mammary fat pad of NCR nude (Taconic Biosciences) or C.B.-17 SCID.bg (Charles River Laboratories) mice, respectively (7, 26). Mice were randomly assigned to groups when tumor volumes reached approximately 150 to 300 mm³. Tumor volumes were measured twice a week after a single i.v. injection of 5 mg/kg T-DM1 or vehicle (10 mmol/L sodium succinate, 0.02% polysorbate 20, 6% w/v trehalose dihydrate, pH 5.0).

For in vitro drug combination analysis, combination index (CI) values were derived from the Chou and Talalay method (27), using CalcuSyn software (Biosoft). CI < 1 denotes synergy; CI > 1 denotes antagonism; CI ≅ 1 denotes additivity. For in vivo xenograft studies, two-sample Student t tests, assuming equal variances and two-tailed distribution, were performed to derive P values for end-of-study tumor volumes in treated versus control groups. Statistical significance is reached with P values < 0.05.

To uncover potential resistance mechanisms to T-DM1, we selected trastuzumab-insensitive KPL-4 and trastuzumab-sensitive BT-474M1 breast cancer cells. Both cell lines are HER2 gene–amplified and are 3+ for HER2 expression by immunohistochemistry (IHC), the diagnostic criteria for clinical use of T-DM1. Cells were chronically exposed to increasing concentrations of T-DM1, over a 10-month period, to a final concentration of 2 μg/mL. The doubling time for KPL-4 TR cells was 2-fold slower than that for parental cells. Doubling times for BT-474M1 parental and resistant cells were similar. Figure 1A (KPL-4) and C (BT-474M1) shows representative dose-response curves for T-DM1, S-methyl-DM1, trastuzumab, and other anticancer agents. IC₅₀ values for parental KPL-4 and BT-474M1 cells were 0.0043 and 0.056 μg/mL, respectively, whereas KPL-4 TR and BT-474M1 TR were resistant to T-DM1 at concentrations up to 3 μg/mL. Cross-resistance to S-methyl-DM1 and other antimicrotubule agents, MMAE and docetaxel, was demonstrated in KPL-4 TR (Fig. 1A). In contrast, BT-474M1 TR cells retained sensitivity to all 3 antimitotic agents (Fig. 1C). KPL-4 parental cells are insensitive to trastuzumab, which was maintained in KPL-4 TR cells (Fig. 1A). Interestingly, we observed that T-DM1–resistant BT-474M1 cells manifested resistance to trastuzumab, compared with parental cells (Fig. 1C). We also examined the effects of lapatinib, a dual HER2/EGFR kinase inhibitor. KPL-4 TR cells were less sensitive to treatment with lapatinib (Fig. 1A), whereas BT-474M1 parental and resistant cells showed similar sensitivity (Fig. 1C).

We next performed studies to verify resistance in mouse xenograft models. Initial in vivo efforts demonstrated poor growth of BT-474M1 TR cells, and lack of T-DM1 resistance in KPL-4 TR cells (due to loss of MDR1, Supplementary Fig. S1; see transporter section below). We therefore used TR-1 cells for in vivo studies. We first established that KPL-4 TR-1 and BT-474M1 TR-1 maintained resistance in vitro (Fig. 1A and C, first plots). In mouse xenograft studies, KPL-4 P tumors are highly sensitive to T-DM1 (P = 6.26 × 10⁻⁸), whereas KPL-4 TR-1 tumors demonstrated complete resistance (Fig. 1B). BT-474M1 P tumors were also sensitive to T-DM1 (P = 0.0029, Fig. 1D). BT-474 TR-1 tumors were partially resistant to T-DM1, with complete resistance at the first measurement (day 3) compared with parental cells, where clear separation between vehicle and T-DM1 was observed. Despite slow regrowth of TR-1 tumors, statistical significance (vehicle vs. T-DM1) was not reached (P = 0.091). Our findings are similar to a recent report demonstrating partial resistance in vivo in 2 of 3 BT-474 T-DM1–resistant clones (28).

To gain broad understanding on mechanisms of T-DM1 resistance, we employed Affymetrix HG-U133 plus 2.0 arrays comparing gene expression in parental versus resistant lines. We performed cross-comparison of genes that were differentially expressed (>2-fold and P < 0.05) between KPL-4 TR and BT-474M1 TR cells (Table 1). Forty-nine genes were upregulated and 10 genes downregulated in common in resistant cells compared with parental cells. Overall, gene expression fold changes in common were modest (approximately 2- to 4-fold). BHLHE41 (basic helix-loop-helix family, member 41 transcription factor) was the most highly upregulated gene in common between KPL-4 TR and BT-474M1 TR cells (up 12.24- and 8.34-fold, respectively). However, siRNA silencing of BHLHE41 in both TR lines did not reverse T-DM1 resistance (Supplementary Fig. S2). BHLHE41 is reported to regulate sleep cycles (29), B- and T-cell development (30), and suppress metastasis (31). However, no role has yet been described in drug resistance. With the exception of BHLHE41, the more highly regulated genes (>5-fold compared with parental cells) were not found in common in both cell lines. We thus decided to focus on each resistant pair separately. Expression changes > 5-fold in resistant versus parental cells are shown in Supplementary Tables S1 and S2. Supplementary Tables S3 and S4 show gene expression changes > 2-fold. The most notable changes were observed in transporters, adhesion molecules, cytokines/chemokines, proteases and phosphatases, and signal transduction pathways.

As resistant cells were generated by prolonged exposure to T-DM1, a potential mechanism of T-DM1 resistance is loss of the target, HER2. Reduced HER2 levels in KPL-4 TR cells were demonstrated by microarray (Supplementary Table S2), qRT-PCR (Supplementary Fig. S3), and immunoblot (Fig. 2A). To further characterize expression changes, HER2 amplification was assessed by FISH. The ratio of the average HER2 gene copy number to CEP17 (centromeric protein on chromosome 17) gene copy number was 5.8 in KPL-4 parental and 2.9 in KPL-4 TR cells, demonstrating that KPL-4 TR cells had reduced HER2 gene copy number (Fig. 2B). Decreased HER2 cell surface expression was confirmed by flow cytometry analysis (Supplementary Fig. S2). To understand gene copy number at the protein level, immunoblot analysis was performed and compared with a panel of breast cancer cells that express known HER2 levels. The results demonstrate that KPL-4 TR cells did not lose HER2 expression, but rather express HER2 at the 1+ to 2+ level (Fig. 2C). These findings suggest that, during chronic exposure, T-DM1 eliminates cells expressing the highest HER2 levels. We were unsuccessful reintroducing HER2 overexpression into KPL-4 TR cells, despite using a number of different transfection methods. Given preclinical and clinical data demonstrating the requirement for HER2 overexpression for T-DM1 activity (14, 32), we did not pursue this further. In contrast, there was no decrease in HER2 protein level or HER2:CEP17 ratio in BT-474M1 TR compared with parental cells (Fig. 2A and B, respectively), or in HER2 cell surface expression or trastuzumab binding as determined by flow cytometry (Supplementary Fig. S4) and immunofluorescence microscopy (Supplementary Fig. S5). Therefore, the mechanisms of acquired resistance in this setting cannot be attributed to decreased HER2.

To investigate differences in uptake and/or processing of T-DM1, we performed studies with two radiolabeled probes: ¹²⁵I-trastuzumab for assessing antibody internalization, and trastuzumab-[³H]-DM1 for tracking DM1, as described by Erickson and colleagues (33). Although the kinetics of ¹²⁵I-trastuzumab uptake in BT-474M1 TR cells were delayed compared with parental cells (Supplementary Fig. S6A, left), by 24 hours, the total radioactivity in both parent and TR cells was equal. In cells exposed to trastuzumab-[³H]-DM1, the amount of total and fractionated (soluble and precipitable) DM1 was not different in BT-474M1 parent versus TR (Supplementary Fig. S6A, right), indicating no differences in processing in the resistant cells. ¹²⁵I-trastuzumab uptake in KPL-4 TR was lower compared with parental cells (Supplementary Fig. S6B, left), likely a result of decreased HER2. Similarly, decreased intracellular DM1 was observed in KPL-4 TR (Supplementary Fig. S6B, right), with KPL-4 parental cells showing similar processing as reported for other HER2-positive breast cancer lines (33).

Activation of alternative signaling pathways confers resistance to targeted therapeutics (34). Therefore, we evaluated expression of additional RTKs (Fig. 2A; Supplementary Fig. S3) and ligands (Supplementary Fig. S3) by Western blotting and qRT-PCR. Expression of EGFR, and to a lesser extent, IGF-1Rβ and c-Met, was elevated in KPL-4 TR relative to parental cells, whereas HER3 expression was decreased (Fig. 2A; Supplementary Table S2). Both IGF-1Rβ and c-Met are reported to mediate trastuzumab resistance in preclinical models (35, 36). Although KPL-4 are innately resistant to trastuzumab in vitro, we nevertheless tested inhibitors of IGF-1Rβ and c-Met in the context of T-DM1 resistance and found that neither RTK mediated resistance in KPL-4 TR cells (Supplementary Fig. S7). BT-474M1 TR cells did not show differences in RTK expression, with the exception of slightly decreased IGF-1Rβ.

As both EGFR and TGFα were elevated in KPL-4 TR cells (Fig. 2; Supplementary Fig. S3), we investigated the role of EGFR in T-DM1 resistance. Studies using siRNA to deplete EGFR showed that EGFR expression did not mediate resistance to T-DM1 (Supplementary Fig. S8A). Moreover, addition of exogenous TGFα to KPL-4 parental cells did not cause T-DM1 resistance (Supplementary Fig. S9B). Additional studies investigated effects of cetuximab, an EGFR-directed antibody, alone or in combination with T-DM1. Despite upregulated receptor and ligand, addition of cetuximab alone did not inhibit cell growth and did not sensitize KPL-4 TR cells when added with T-DM1 (Supplementary Fig. S8B).

Our qRT-PCR studies (Supplementary Fig. S3) also demonstrated increased expression of HER4 and one of its ligands, neuregulin (as a pan-NRG probe was used, the specific isoform was not identified). Because NRG is a reported resistance factor for multiple anticancer agents (26, 37, 38), we investigated the potential role of this receptor–ligand interaction. Using a ligand-blocking HER4 antibody (38), we were unable to resensitize KPL-4 TR cells to T-DM1 (Supplementary Fig. S9D).

Part of our investigation into RTKs included immunoblot analysis for signal transduction pathways downstream of HER2 (Fig. 3A), revealing decreased PTEN expression in BT-474M1 TR cells. As PTEN is a negative regulator of the PI3K pathway, an important survival pathway, and PTEN deficiency is a resistance mechanism to trastuzumab-based therapy (6), we hypothesized that PTEN might play a role in resistance to T-DM1. We first examined if re-expression of PTEN restored sensitivity of BT-474M1 TR cells to T-DM1. We were unable to reintroduce PTEN into BT-474M1 TR cells as the PTEN-transfected cells eventually died. We then used lentivirus-mediated delivery of PTEN shRNA to deplete PTEN in BT-474M1 parental cells. Immunoblot analysis showed depletion of PTEN protein using PTEN-directed shRNA clone 1 (Fig. 3B; Supplementary Fig. S10). We next examined response of PTEN knockdown cells to T-DM1. Ablation of PTEN expression in parental cells led to reduced activity of T-DM1 (Fig. 3C, top). As expected, PTEN knockdown also conferred resistance to trastuzumab (Fig. 3C, middle), which is in agreement with previous reports (39). However, sensitivity to S-methyl-DM1 was similar in shPTEN cells compared with control cells (Fig. 5C, bottom). These data provide evidence that resistance to T-DM1 in BT-474M1 TR cells was associated with reduced PTEN levels.

Loss of PTEN results in constitutive activation of PI3K/AKT (6). In BT-474M1 TR cells, reduced PTEN expression led to increased levels of phosphorylated AKT (Fig. 3A). Therefore, we reasoned that inactivation of PI3K may rescue T-DM1 resistance from PTEN deficiency. Combining T-DM1 with the pan-PI3K inhibitor GDC-0941 (19) was more synergistic in BT-474M1 TR cells (Fig. 3D, top) than in parental cells (Fig. 3D, bottom), with average CI values of 0.49 and 0.56, respectively. These data demonstrate that PI3K inhibition enhances sensitivity to T-DM1 to a greater degree in PTEN low T-DM1–resistant cells.

The most highly upregulated gene in BT-474M1 TR cells was dopamine and cyclic AMP-regulated phosphoprotein, DARPP-32, also known as protein phosphatase 1 regulatory subunit 1B or PPP1R1B (ref. 40; 56-fold increase; Supplementary Table S1). Both DARPP-32 and its truncated variant, t-DARPP, are recognized by the same probe sets on microarray chips. Further analysis by qRT-PCR and immunoblot showed that expression of both forms was increased relative to parental cells, with t-DARPP the predominant form (Fig. 4A and B). DARPP-32 and t-DARPP are reported to mediate trastuzumab resistance in breast cancer cells (40). To examine if DARPP-32 upregulation was sufficient to confer T-DM1 resistance, BT-474M1 TR cells were transfected with DARPP-32 pooled siRNA that targeted both the full-length and truncated forms. Cells transfected with nontargeting siRNA were used as control. mRNA and protein analysis (Fig. 4C and D) confirmed reduction in DARPP-32 levels in TR cells transfected with DARPP-32 siRNA. However, DARPP-32 depletion did not restore T-DM1 sensitivity (Fig. 4E), indicating that DARPP-32 did not mediate T-DM1 resistance in BT-474M1 TR cells.

In addition to DARPP-32, we investigated the role of two other differentially regulated genes in BT-474M1 TR, IGFBP5 and CXCR4 (5.7-fold decrease and 8.2-fold increase, respectively, average of two probes; Supplementary Table S1), which could be functionally implicated in resistance. Exposing cells to exogenous IGFBP5, or to a selective CXCR4 inhibitor, AMD3100 (23), did not reverse resistance of BT-474M1 TR cells to T-DM1 (Supplementary Fig. S11A and S11B).

A number of transporters of the ATP-binding cassette (ABC) and solute carrier (SLC) families were upregulated in both TR cell lines (Supplementary Tables S3 and S4). In KPL-4 TR cells, microarray data showed highly increased expression of ABCB1 (MDR1), with moderate increases in ABCC1 (MRP1), ABCC4 (MRP4), ABCC10 (MRP7), and ABCG2 (BCRP/breast cancer resistance protein). In contrast, modest upregulation of ABCC4, ABCC11 (MRP8), ABCD3, and ABCG1 was observed in BT-474M1 TR cells. To verify expression, qRT-PCR was performed for transporters with reported functions in cancer drug resistance (41). Increased expression of MDR1 and BCRP was confirmed in KPL-4 TR cells by qRT-PCR and immunoblot analysis, and increased MRP4 was confirmed in KPL-4 TR and BT-474M1 TR cells (Fig. 5A). We assessed additional MRP transporters (MRP1, MRP2, and MRP3) by qRT-PCR and demonstrated modest to no increase in KPL-4 TR and BT-474M1 TR cells (Supplementary Fig. S12).

To investigate the role of increased MDR1 or BCRP in T-DM1 resistance in KPL-4 TR cells, we assessed whether pharmacologic inhibition would restore T-DM1 sensitivity. Inhibition of MDR1 with XR9051, a potent and selective MDR1 inhibitor (20), resulted in substantial reversal of resistance to both T-DM1 and S-methyl-DM1 (Fig. 5B), whereas XR9051 had no effect on response of KPL-4 parental cells to either agent. Given the increase in both MDR1 and EGFR in KPL-4 TR cells, we then investigated whether combined inhibition would further reverse resistance. Addition of cetuximab with XR9051 did not enhance sensitivity to T-DM1 compared with XR9051 alone (Supplementary Fig. S8B). Despite elevated BCRP, the BCRP inhibitor Ko143 (21) did not resensitize KPL-4 TR cells to T-DM1 (Fig. 5B, right). XR9051 and Ko143 alone had no effect on cell viability (Supplementary Fig. S13).

As there is no selective MRP4 inhibitor available, we used siRNA knockdown to investigate the role of MRP4 in T-DM1 resistance. Although both TR cell lines showed increased MRP4, depletion by individual or pooled siRNA oligonucleotides did not reduce resistance to T-DM1 (Supplementary Figs. 14 and 15). Although we observed variability in knockdown efficiency, MRP4 siRNA #1 showed the greatest depletion (80% in BT-474M1 TR and 70% in KPL-4 TR) but with no reversal of resistance in either cell line. Taken together, these data support the mechanisms of T-DM1 resistance in KPL-4 TR cells as decreased HER2 and increased MDR1 expression.

SLC transporter gene expression changes were abundant in TR cells. BT-474M1 TR cells showed changes (up or down) in 16 SLC family members, whereas 31 SLC genes were differentially expressed in KPL-4 TR cells. As fold changes were modest, most notably in BT-474M1 TR cells, and because the function of SLCs in cancer drug resistance is not established (42), we initially did not follow up on these observations. Recently, Hamblett and colleagues (43) reported that loss of the lysosomal transporter SLC46A3 mediates resistance to maytansinoid-containing ADCs with noncleavable linkers. As the resistance mechanisms in BT-474M1 cells were not completely defined, we performed qRT-PCR for SLC46A3 in our resistant cells, despite no evidence for SLC46A3 loss from microarray data. Interestingly, SLC46A3 expression was lost in BT-474M1 TR cells (Fig. 6, left). Modestly reduced expression was observed in KPL-4 TR cells (–2.79 fold by microarray, Supplementary Table S4, but < 2-fold decrease by qRT-PCR, Supplementary Fig. S16). A role for SLC46A3 in T-DM1 resistance in BT-474M1 parental cells was then verified by using individual and pooled oligonucleotides for siRNA knockdown (Fig. 6, middle; Supplementary Fig. S17), which resulted in resistance to T-DM1 (Fig. 6, right), to a level similar to BT-474M1 TR cells.

Understanding mechanisms of drug resistance in preclinical models poses challenges due to the complex nature of resistance and compensatory pathways, as well as the use of different tumor cell models with diverse genetic backgrounds. Moreover, identifying resistance mechanisms for ADCs is complicated by the nature of the drug itself, in that there are multiple components (antibody, linker, cytotoxic agent) to consider. One of the mechanisms by which tumors acquire drug resistance is increased expression of ABC transporters, which actively efflux anticancer drugs out of cells. Expression of MDR1, MRP4, and BCRP was increased in KPL-4 TR, and MRP4 elevated in BT-474M1 TR, compared with parental cells. Inhibition of MDR1 with a selective inhibitor restored sensitivity to T-DM1 and DM1, whereas BCRP or MRP4 inhibition did not. Despite increased expression of multiple drug transporters in our resistant cells, collective data support a role only for MDR1 and MRP1 (44). Trock and colleagues found that breast cancer patients with tumors expressing MDR1 were 3 times more likely to fail to respond to chemotherapy than patients whose tumors were MDR1 negative (26). The clinical significance of MDR1 or MRP1 as mechanisms of drug resistance in patients receiving T-DM1 treatment has not been established. It is unclear if patients with de novo resistance to T-DM1 have higher MDR1 expression or whether acquisition of T-DM1 resistance after therapy parallels increased expression of MDR1.

In addition to MDR1 upregulation, a second predominant resistance mechanism in KPL-4 TR cells was decreased HER2 expression. These results are consistent with reported resistance mechanisms for MMAE-containing ADCs targeting CD22 and CD79b (45), as well as CD30 (46), and suggest common resistance alterations among some models. A previous report described in vitro–acquired resistance to a trastuzumab–maytansinoid ADC (44) in MDA-MB-361 and JIMT1 cells, both of which are HER2 2+ by IHC, and thus less clinically relevant for T-DM1. JIMT1 cells also express MUC4, a glycoprotein that interferes with trastuzumab binding to HER2 (47), which complicates interpretation of trastuzumab and T-DM1 activity. Increased MRP1 and decreased HER2 were the primary resistance mechanisms for JIMT1 and MDA-MB-361 cells. Global alterations in proteins involved in posttranslational modification, vesicle transport, and trafficking were also described.

Following ADC catabolism and/or linker cleavage in the lysosomal compartment, transport of free drug or catabolites across the lysosomal membrane is required for ADC activity. A unique lysosomal transporter, SLC46A3, was recently shown to transport catabolites of maytansinoid-containing ADCs with noncleavable linkers (43). Although expression changes were not observed in our Affymetrix study, we demonstrated loss of SLC46A3 expression by qRT-PCR in BT-474M1 TR cells. Furthermore, silencing of SLC46A3 expression conferred partial resistance to T-DM1 in parental cells, supporting a role for SLC46A3 loss in T-DM1 resistance. As the sole catabolite of T-DM1 is lysine-MCC-DM1 (33), these data also provided an explanation for BT-474M1 TR cells retaining sensitivity to free DM1. Acquired T-DM1 resistance in BT-474 cells was recently reported to result from T-DM1 accumulation in lysosomes, mediated by decreased lysosomal proteolytic activity (28). Alterations in specific lysosomal proteins were not described.

Approximately 50% of patients with breast cancer have a mutation in or loss of at least one copy of the PTEN gene, which results in activation of PI3K signaling (48). Nagata and colleagues reported that decreased PTEN expression resulted in activation of the PI3K/AKT pathway and inhibition of trastuzumab-mediated growth arrest in HER2-overexpressing breast cancer cells (6). Furthermore, they demonstrated that PI3K inhibitors rescued trastuzumab resistance in PTEN-deficient cells in vitro and in vivo. Importantly, patients with PTEN-deficient HER2-overexpressing breast tumors had significantly worse responses to trastuzumab-based therapy than those with tumors expressing normal PTEN (48). Berns and colleagues used large-scale RNA interference screens in BT-474 cells and identified PTEN as the only gene whose knockdown resulted in trastuzumab resistance (39). Our data revealed that BT-474M1 TR cells expressed reduced levels of PTEN compared with parental cells. Moreover, decreased PTEN reduced sensitivity to T-DM1. Interestingly, BT-474M1 TR cells were cross-resistant to trastuzumab, but maintained sensitivity to free DM1. Thus, it is likely that T-DM1 resistance is partially due to resistance to trastuzumab. In addition, we found that T-DM1 combined with the PI3K inhibitor GDC-0941 synergistically inhibited BT-474M1 TR cell growth. The combination also demonstrated enhanced growth inhibition in parental cells. These data indicate that PI3K inhibition can sensitize T-DM1–resistant BT-474M1 cells. Combining T-DM1 with inhibitors that target signaling transduction pathways might be a promising strategy to improve T-DM1 efficacy and circumvent resistance. The combination of T-DM1 with PI3K inhibitors is currently under clinical investigation.

In addition to the molecular alterations discussed above, we observed increased expression of a number of RTKs—IGF-1Rβ, c-Met, and EGFR—that are implicated in trastuzumab resistance (35, 36, 49). Functional studies, however, failed to demonstrate roles in T-DM1 resistance. We performed similar studies investigating potential roles in T-DM1 resistance for additional genes that were differentially regulated. Upregulation of DARPP-32, CXCR4, as well as decreased IGFBP5 did not mediate T-DM1 resistance. Recently, defects in cyclin B1 induction were described to play a role in acquired T-DM1 resistance in HCC1954, HCC1419, and SK-BR-3 breast cancer cells (50). These findings highlight the complex nature of molecular alterations resulting from chronic ADC exposure as well as the importance of assessing function of each molecule in the context of resistance.

In summary, our models provide a valuable tool for investigating molecular mechanisms of acquired resistance to T-DM1. It will be important to determine if the features of T-DM1 resistance observed in our studies are present in breast cancer patients who progress during T-DM1 treatment. Collective data from different T-DM1–resistant models indicate that mechanisms of acquired resistance are model dependent (28, 43, 44, 50). Thus, investigating markers of T-DM1 resistance in patients will be complex. Moreover, progression biopsies are rarely acquired from mBC patients, making posttreatment tumor analysis difficult. Given the value of understanding prognostic and predictive biomarkers, this paradigm is changing. Importantly, biomarker analysis of pre-and posttreatment specimens from several neoadjuvant trials, now in progress or completed, will enable us to assess the clinical implications of the preclinical markers identified for T-DM1 resistance.

M.X. Sliwkowski has an ownership interest (including patents) in, and is a consultant/advisory board member for, Genentech, Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: G. Li, M.X. Sliwkowski, G.D.L. Phillips

Development of methodology: G. Li, J. Guo, B.-Q. Shen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G. Li, J. Guo, D.B. Yadav, L.M. Crocker, J.A. Lacap

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G. Li, J. Guo, B.-Q. Shen, D.B. Yadav, G.D.L. Phillips

Writing, review, and/or revision of the manuscript: G. Li, J. Guo, B.-Q. Shen, D.B. Yadav, M.X. Sliwkowski, G.D.L. Phillips

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G. Li, L.M. Crocker

Study supervision: G. Li, M.X. Sliwkowski

We thank Suzie J. Scales for running HER2 immunofluorescence assays and Suchit Jhunjhunwala for microarray data analysis.

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.