Defective Cyclin B1 Induction in Trastuzumab-emtansine (T-DM1) Acquired Resistance in HER2-positive Breast Cancer

Trastuzumab-emtansine (T-DM1) is an antibody–drug conjugate (ADC) consisting of the anti-HER2 antibody trastuzumab covalently linked to the antimitotic agent DM1 through a stable linker that potently inhibits growth of both trastuzumab-sensitive and -resistant HER2-amplified cancer cells (1). T-DM1 has mechanisms of action containing of the antitumor effects related to trastuzumab and those associated with intracellular DM1 catabolites (2, 3). DM1 is a derivative of maytansine, a highly potent antimitotic drug (4). Once bound to HER2, T-DM1 enters in the cell by receptor-mediated endocytosis and the HER2–T-DM1 complex is processed via degradation of trastuzumab in lysosomes giving rise to the intracellular release of the active catabolite Lys-MCC-DM1. In the cytoplasm, DM1 exerts its functions through binding to the beta subunit of tubulin and modifying its assembly properties. By doing so, DM1 is able to disrupt the formation of the mitotic spindle necessary for accurate chromosome segregation along mitotic process. Overall, DM1 as other antimitotic drugs elicits a mitotic arrest and cells therefore fail to complete a normal mitosis. This prolonged cell-cycle delay eventually culminates in cell death by mitotic catastrophe, necrosis, or apoptosis (5). In addition to these DM1-related actions, T-DM1 maintains properties of trastuzumab such as inhibition of HER2-directed signal transduction and activation of antibody-dependent cell-mediated cytotoxicity (2, 3).

T-DM1 is a standard second-line treatment for HER2-positive metastatic breast cancer patients based on the results of a phase III clinical trial (EMILIA trial), that compared the efficacy and toxicity of T-DM1 versus the combination of lapatinib and capecitabine in patients previously treated with trastuzumab and chemotherapy. T-DM1 offered superior activity, tolerability, and survival than lapatinib and capecitabine (6). However, resistance to T-DM1 occurs (3, 7, 8). To date, the main T-DM1 clinical resistance mechanisms studies included an exposure–response and exploratory biomarker analysis, both based on patients from the EMILIA and TH3RESA trial (9–11). In the exposure–response analysis, the data showed that higher T-DM1 exposure was associated with improved efficacy. This analysis suggested that there is an opportunity to optimize T-DM1 dose in the patient subgroup with low exposure (10). With regards to the exploratory biomarker study of the EMILIA trial, focusing only in the T-DM1 group, univariate analysis of the data suggested that there were no evident differences in progression-free survival (PFS) related to EGFR, and HER3 median mRNA concentration ratios, PIK3CA mutations status, or PTEN expression; patients with HER2 mRNA above the median had better outcomes on T-DM1 than those equal or below the median (11). In the TH3RESA phase III trial, performed in heavily pretreated HER2-positive advanced breast cancer, patients were randomized to T-DM1 or treatment of physician's choice. An exploratory biomarker analysis included HER2 and HER3 mRNA expression, PIK3CA mutation status, and PTEN protein expression. T-DM1 prolonged median PFS in all the subgroups analyzed. A numerically greater benefit was reported in patients with tumors expressing HER2 mRNA above the median (9). More recently, in the ZEPHIR trial, the use of molecular imaging of HER2 by HER2-PET/CT with ⁽⁸⁹⁾Zr-trastuzumab combined with early metabolic response assessment by FDG-PET/CT, discriminated patients with short versus long time to T-DM1 treatment failure (12).

The biomarker analyses discussed above were focused on the HER2 signaling pathway. However, given the dual mechanism of action of T-DM1, we focused on resistance potentially related to the cell-cycle–modulating effects of DM1. We identified that the G₂–M arrest induced by T-DM1 in sensitive HER2 breast cancer cells did not occur in their resistant counterparts in a CDK1/cyclin B1–dependent manner. In fresh HER2-positive breast cancer explants, lack of induction of cyclin B1 correlated with T-DM1 failure to induce apoptosis.

Breast cancer cell lines BT474, SKBR3, AU565, EFM-192A, HCC1954, and HCC1419 were obtained from the ATCC. Authenticity of the cells was tested by STR DNA Profiling analysis at the ATCC (June 2013 and December 2014) before starting the generation of resistant cells. The number of passages between thawing and use in the described experiments was five or less. T-DM1 (trastuzumab emtansine, Kadcyla) was provided by Genentech under MTA agreement (Sliwkowski MX, Lewis Phillips GD) and trastuzumab by Hospital del Mar pharmacy (Barcelona, Spain).

Cells were plated in duplicate into 12-well culture plates at a density of 8–12 × 10³ cells per well and left overnight and then on day zero, treatment was initiated. At this time T-DM1 (0.1–1 μg/mL) was added. On days 3, 7, and 10, cells were washed with PBS, trypsinized, resuspended in media, and counted with Scepter Automated Cell Counter (Millipore).

T-DM1–resistant cell lines were derived from original parental cell lines by exposure to stepwise increasing concentrations of T-DM1 in a pulse fashion (13). The protocol is summarized in Fig. 1A. Three T-DM1 acquired resistance sublines were collected named SKBR3/TDR, HCC1954/TDR, and HCC1419/TDR. In addition, vehicle-treated parental cell lines were kept in culture during this period as control cell lines. We established an exposure to 0.1 μg/mL of T-DM1 for 3 days as a reference schedule to define resistance in our model. This schedule was selected on the basis of experiments performed in MCF7 cells that do not overexpress HER2, whose growth was unaffected until higher concentrations and longer exposure to T-DM1 were used.

Formalin-fixed and paraffin-embedded cell pellets were prepared from parental and T-DM1–resistant cells to assess the status of HER2 gene by scoring its amplification following ASCP/CAP guidelines (14). The commercial PathVysion HER2 DNA probe was used.

T-DM1 binding to the cell surface was evaluated by indirect immunofluorescence staining and flow cytometry. Cells (2–5 × 10⁵) were incubated with T-DM1 (7.5 μg/mL) for 30 minutes followed by R-Phycoerythrin AffiniPure F(ab′)₂ Fragment Goat Anti-Human IgG, Fcγ Fragment Specific (Jackson ImmunoResearch; 1:500 dilution) for 30 minutes at 4°C. Trastuzumab (7.5 μg/mL) and rituximab (MabTera, Roche; 7.5 μg/mL) were used as positive and negative controls, respectively. Live/dead gate was set with DAPI counterstaining. Samples were acquired on LSR Fortessa flow cytometer (BD Biosciences), and data analyzed with FlowJo software (TreeStar).

T-DM1 internalization was evaluated by immunofluorescence staining. Cells (1.5 × 10⁵) were seeded on coverslips and treated with 1.5 μg/mL T-DM1 for 15 minutes. After washing out the drug, cells were cultured for 24 hours with or without 5 μmol/L chloroquine (a drug that induced changes in lysosomal pH) to accumulate intracellular T-DM1. Anti-human Cy3-conjugated antibody was used to detect T-DM1, phalloidin-FITC (P5282 Sigma) was used for actin staining, and nuclei were counterstained with DAPI. Sample processing was performed as reported previously (15).

Cells were seeded on 6-well plates and treated with compounds for 24 hours by T-DM1 (0.1 μg/mL). Thereafter, cells were incubated with 30 μmol/L bromodeoxyuridine (BrdUrd, B9285, Sigma) for 30 minutes at 37°C, and then subsequently trypsinized, washed with PBS, and fixed in ice-cold ethanol for at least one hour at −20°C. Cells were digested in prewarmed 1 mg/mL pepsin in 30 mmol/L HCl (pH 1.5) for 30 minutes at 37°C with gentle shaking, and then incubated in 2 mol/L HCl for 20 minutes at room temperature. Next, cells were washed with PBS and antibody buffer (0.5% w/v BSA, 0.5% v/v Tween-20 in PBS) and incubated with mouse primary antibody against BrdUrd (555627, BD Pharmingen) in antibody buffer for one hour at room temperature. After washing with PBS, samples were incubated with secondary goat anti-mouse IgG FITC-conjugated antibody for 30 minutes at room temperature in the dark. Finally, cells were washed one more time with PBS and then stained with 25 μg/mL propidium iodide. Flow cytometry was performed using a Becton Dickinson FACScan operated by the CELLQuest software.

Cdc2/CDK1 kinase activity was measured with the nonradioactive MESACUP Cdc2/CDK1 Kinase Assay Kit (MBL, International Corporation) following the manufacturer's instructions. The method is based on an ELISA assay that utilizes a biotinylated synthetic peptide as a substrate for the Cdc2/CDK1 kinases present in the samples and a mAb conjugated to horseradish peroxidase (HRP) recognizing the phosphorylated form of the peptide. The HRP substrate is then added and the intensity of the color was measured at 492 nm. The results are reported as fold induction of OD 492 nm in arbitrary units (RLA) of treated sample using untreated condition as reference set at 1.

Cells were seeded on glass coverslips and cultured as indicated in the figure legends. Cells washed by PBS, and then fixed in 4% paraformaldehyde for 10 minutes at room temperature. Then the samples permeabilized by PBS containing 0.25% Triton X-100, and then, incubated with 1% BSA in PBST for 30 minutes for blocking unspecific binding of the antibodies. In the next step, samples incubated for 1 hour with anti-α-tubulin FITC-conjugated antibody (F2168, Sigma). Cell nuclei were stained with DAPI (1:1,000; Sigma) for 15 minutes at room temperature. After washing, coverslips were mounted with a drop of mounting medium and viewed using a Leica SP5 upright confocal microscope.

Cellular protein lysates were prepared in lysis buffer [50 mmol/L Tris-HCl, pH 8.0, 100 mmol/L NaCl, 1% (v/v) Nonidet P-40, 0.1% (w/v) SDS] containing protease inhibitors and quantified by use of the Bradford assay (Bio-Rad Laboratories). Equivalent protein amounts of each sample were analyzed. Protein detection on Western blots was performed according to standard protocols. The antibodies used were: HER2 (Biogenex), cyclin B1 (sc-245), and CDC2 p34 (CDK1; sc-54) purchased from Santa Cruz Biotechnology and β-actin (A-5316) was purchased from Sigma.

For measuring apoptosis, the Annexin V and Dead Cell Assay Kit (Millipore) was used according to the manufacturer's instructions. Briefly, after treatment, the cells were incubated with Annexin V and Dead Cell Reagent (7-AAD) for 20 minutes at room temperature in the dark, and the events for dead, late apoptotic, early apoptotic, and live cells were counted with the Muse Cell Analyzer (Millipore) and analyzed with MuseSoft 1.4.0.0 (Millipore).

For transient transfection experiments, cells were transfected by use of the Amaxa 4D-Nucleofector device, according to manufacturer's instruction. For each electroporation reaction, 100 μL of complete Nucleofector solution combined with 300 nmol/L siRNA against Cyclin B1, cdc20, and scrambled siRNA as a control (duplex siRNAs were purchased from GE Dharmacon) under a specific optimized program for siRNA delivery with the Nucleofector (SKBR3 E-009).

The study was approved by the ethics committee of the Hospital del Mar and conducted following institutional guidelines. Fresh tumor specimens from women with HER2-positive breast cancer undergoing routine cancer surgery, which were not needed for diagnostic purposes, were collected to add ex vivo T-DM1 and assess its molecular effects according to our experience (16). Samples were sliced and cultured in RPMI1640 medium supplemented with 10% FBS, 2 mmol/L l-glutamine, and 100 U/mL penicillin–streptomycin for 120 hours in the absence (control) or presence of T-DM1 (0.1 μg/mL). Specimens were fixed in 10% neutral-buffered formalin for 16 hours and embedded in paraffin then assayed by IHC.

Three-micrometer–thick paraffin sections from tissue blocks of the tumors were stained for HER2 (Herceptest P980018/S010, Dako), cyclin B1(sc-245, Santa Cruz Biotechnology), Ki-67 (GA62661-2, MIB1 clone, Dako), phosphorylated (Ser10) Histone 3 (9701, Cell Signaling Technology), and active caspase-3 (9664, Cell Signaling Technology) followed by incubation with an anti-rabbit Ig dextran polymer (Flex+, Dako) and 3,3′-diaminobenzidine as chromogen in a Dako Link platform. HER2 staining was scored following ASCP/CAP guidelines (14). For the other markers, the percentage of positive tumor cells was scored.

Data are presented as mean ± SD. The GraphPad Prism software was used to construct graphs and statistical analysis. Statistical significance was determined using Student t test or one-way ANOVA. P ≤ 0.05 was considered as significant.

We assessed the half maximal inhibitory concentration (EC₅₀) values of T-DM1 in a panel of HER2-positive breast cancer cells. The median EC₅₀ values for BT474, SKBR3, AU565, EFM-192A, HCC1954, and HCC1419 cell lines at 72 hours were 0.025, 0.002, 0.005, 0.009, 0.020, and 0.018 μg/mL, respectively. Four cell lines, two sensitive to trastuzumab (SKBR3 and BT474) and two with primary resistance to trastuzumab (HCC1954 and HCC1419) were selected to generate T-DM1 resistance (TDR) by applying a pulsed administration strategy of the drug, often used for the development of chemotherapy drug resistance (Fig. 1A; ref. 13). This protocol encompasses short pulses of drug treatment followed by rest periods in drug-free media to allow the cells to recover from toxicity between treatments until a stable resistant phenotype is observed. Specifically, the procedure consisted of three consecutive cycles of 3 days on treatment followed by 3 days off treatment for each T-DM1 concentration of 1, 2, and 4 μg/mL. The entire 54-day protocol resulted in the generation of cells with varying levels of T-DM1 resistance.

We established an exposure to 0.1 μg/mL of T-DM1 for 3 days as a reference schedule to define resistance in our model. This schedule was selected on the basis of experiments performed in MCF7 cells that do not overexpress HER2, whose growth was unaffected until higher concentrations and longer exposure to T-DM1 were used. After approximately 2 months, three cell lines developed resistance to T-DM1 (called HCC1954/TDR, HCC1419/TDR and SKBR3/TDR; Fig. 1B). We were unsuccessful to generate BT474-resistant cells even after repeated attempts. HCC1954/TDR cells were completely resistant after 10 days of exposure to T-DM1 at 0.1 μg/mL. HCC1419/TDR and SKBR3/TDR cells decreased the sensitivity to the antiproliferative effects of T-DM1 by 20%–30% at 0.1 μg/mL after 3 days. This level of resistance, albeit modest, has been helpful to identify mechanisms of resistance to several agents (13). T-DM1–resistant cells had a growth rate similar to parental cells in vitro, as assessed by cell counting after 7 days of culture under the same conditions. The relative cell numbers of resistant versus parental cells were 110% ± 8.3% in HCC1954/TDR, 93% ± 4% in HCC1419/TDR and 103% ± 7% in SKBR3 (no statistical differences). The morphology of parental versus resistant cells was similar by optic microscopic observation. All resistant cell lines exhibited resistance features for at least one year after their generation. Of note, SKBR3/TDR cells retained a similar antiproliferative response to trastuzumab alone than parental cells (50% growth inhibition), suggesting that the resistance was associated to DM1 (Fig. 1C). Both parental and resistant cells showed similar EC₅₀ values for paclitaxel [HCC1954 (4.5 nmol/L), HCC1954/TDR (6.5 nmol/L), HCC1419 (6.0 nmol/L), HCC1419/TDR (5.6 nmol/L), SKBR3 (3.2 nmol/L), SKBR3/TDR (3.1 nmol/L)].

HCC1419/TDR and SKBR3/TDR cells retained the same level of HER2 amplification and protein expression than parental cells (Fig. 2A and B). However, HCC1954/TDR cells displayed decreased amplification of HER2 and reduced HER2 protein (Fig. 2A and B), and mRNA levels (6% of relative mRNA expression) compared with parental cells. Parental HCC1954 cells had a mean of 40 HER2 copies but they contained two additional subpopulations; 17% of the cells had less than 6 copies (some scored as amplified for having only one copy of CEP17) and 11.4% had 6–10 HER2 copies. During the generation of the resistant cells, T-DM1 eradicated the cells with strong HER2 amplification after a short exposure, allowing the emergence of subdominant clones with low HER2 amplification (average number of 8 HER2 copies, HER2 1+ by IHC). The emergence of HER2 low resistant cells in HCC1954 was confirmed in three independent experiments. Despite this important change in the population, the resistant cells still met the criteria for scoring them as HER2 amplified as they have more than 6 HER2 copies (14). In addition, other genes such as ORMDL3, STARD3, PPP1R1B, MIEN1 located in the minimal common region of 17q12 amplification were also downmodulated as assayed by expression arrays. SKBR3/TDR and HCC1419/TDR did not have any significant differences in the levels of HER2, ORMDL3, STARD3, PPP1R1B and MIEN1 mRNA in comparison with parental cells, as assayed by qRT-PCR. We next assayed the ability of T-DM1 to bind to cell surface HER2 by flow cytometric analysis. Although surface HER2 levels might vary in HCC1954/TDR compared with parental cells, T-DM1-HER2 binding was preserved in all TDR cells (Fig. 2C).

We assayed whether T-DM1 was capable of inducing internalization upon binding to HER2 in paired sensitive and resistant cells (15). Cells were treated with 10 nmol/L T-DM1 for 15 minutes at 37°C. After washing out the drug, cells were cultured for 24 hours with or without 50 μmol/L chloroquine to accumulate T-DM1 intracellularly. As shown in Fig. 2D, T-DM1 (red dots) was internalized into both parental and resistant cells. The magnitude of T-DM1 internalization and intracellular pattern were similar in parental and resistant SKBR3 and HCC1419 cells. In HCC1954/TDR cells we also detected T-DM1 intracellularly, but to a lesser extent than in their parental HER2-positive counterpart, in agreement with their lower HER2 amplification level and the surface expression.

T-DM1 increased significantly the percentage of cells in G₂–M phase while it induced a decrease in the S and G₀–G₁ phases in the three parental cell lines (Fig. 3A). In the resistant cells, the effects of T-DM1 on G₂–M cell-cycle arrest were less pronounced than in parental cells. In HCC1954/TDR and HCC1419/TDR, there was no increase in G₂–M. In SKBR3 there was a significant increase following T-DM1 but to a much lesser degree than in parental cells (Fig. 3A). Next, we evaluated the effects of T-DM1 on microtubule arrangement. In parental cells, T-DM1 caused the formation of multinucleated giant cells with severely defective microtubule arrangements (Fig. 3B). Previously, it has been reported that T-DM1 causes tumor growth inhibition by mitotic catastrophe (17). These morphologic alterations suggestive of mitotic catastrophe were undetected in T-DM1–resistant cells. As most cells undergoing mitotic catastrophe are destined to die by apoptosis, we evaluated the effects of T-DM1 on apoptosis induction. T-DM1 induced apoptosis in HCC1954 and SKBR3 parental cells at 48 hours but the effect on the resistant counterparts was much less pronounced under the same conditions (Fig. 3C).

We hypothesized that resistant cells may have a defective cell-cycle–regulatory machinery that does not allow T-DM1 to induce G₂–M arrest and consequently, mitotic catastrophe. We focused on the potential involvement of cyclin dependent kinase 1 (CDK1) and cyclin B1 in T-DM1 resistance. Activation of the CDK1–cyclin B1 complexes are essential for progression into M-phase, but prolonged mitotic arrest (for example because of the inability of forming the mitotic spindle in cells treated with microtubule-affecting agents such as DM1) leads to mitotic catastrophe (18). The activity of CDK1–cyclin B1 complex is regulated mainly by the expression of cyclin B1 and by the phosphorylation status of the catalytic subunit CDK1. Degradation of cyclin B1 by the proteasome after ubiquitination by the multi-subunit ubiquitin E3-ligase APC/C^cdc20 is essential for exiting mitosis. T-DM1 exposure augmented cyclin B1 expression in a panel of HER2-positive parental breast cancer cells (except BT-474; Fig. 4A).

We analyzed the effects of T-DM1 on the activity of CDK1. We first tested this effect in four parental HER2 positive cell lines and in their trastuzumab resistant (TR) counterparts that we recently reported (19). These TR cell lines were included in this analysis to confirm whether acquired trastuzumab resistance (that corresponds to the clinical setting in which T-DM1 is approved) affects T-DM1 ability to induce CDK1 activity. The results showed a significant increase in CDK1 Kinase activity 24 hours posttreatment with T-DM1 (0.1 μg/mL) in both sensitive and trastuzumab-resistant cells by using the MESACUP Cdc2/Cdk1 Kinase Assay Kit with exception of the BT-474 cell pairs (Fig. 4B). In all the cell lines with acquired trastuzumab resistance, T-DM1 exposure decreased viable cells (cell number relative to control at 3 days of 20% ± 3.2% in AU565/TR, 40% ± 6% in EFM-192A, 17% ± 3% in SKBR3/TR and 72% ± 3% in BT474/TR) and cyclin B1 was induced following drug exposure (Fig. 4B). Trastuzumab alone did not induce cyclin B1 in these cell lines (Fig. 4C).

We then assessed T-DM1 effects on the CDK1/cyclin B1 complex kinase activity and cyclin B1 levels in the paired cell lines parental and resistant to T-DM1. We performed a time–course study the HCC1954 and HCC1954/TDR cells. The activation of the mitotic kinase CDK1/cyclin B1 was detected as early as 12 hours post-T-DM1 in parental HCC1954 cells (2-fold) but not in HCC1954/TDR cells (P < 0.05). Protein levels of total CDK1 were unchanged in parental and resistant cells. Cyclin B1 protein levels were moderately increased (2-fold) at 12 hours following T-DM1 treatment in parental cells in parallel with CDK1 activity and they were maintained elevated up to 24 hours afterwards. This increase in the cyclin B1 levels was undetected in resistant cells (Fig. 4D). A similar pattern of cyclin B1 expression and appearance of CDK1/cyclin B1 kinase activity were observed with the other T-DM1–sensitive/resistant cell pairs after 24 hours of T-DM1 exposure (Fig. 4C). The analysis of the levels of cyclin B1 mRNA by qRT-PCR showed a nonsignificant increase (Supplementary Fig. S1), suggesting postranscriptional regulatory mechanisms.

We tested whether cyclin B1 knockdown could mediate resistance to T-DM1 in parental cells. Cell lines were transfected with siRNA directed against cyclin B1 mRNA or with a suitable control. Cyclin B1 levels were measured 24 and 48 hours after the transfection by Western blot confirming that they were significantly lower in the siRNA-transfected cells than in the siControl (Fig. 5A). The paired cell lines transfected with the cyclin B1 siRNA or the siControl were exposed to T-DM1 and cell viability was assayed after 48 hours. Silencing of cyclin B1 induced a significant resistance to T-DM1 in the three parental cell lines (Fig. 5A).

We next tested whether increasing the levels of cyclin B1 in resistant cells might sensitize them to T-DM1. Cdc20 is a regulatory subunit of the multi-subunit ubiquitin E3-ligase APC/C that is responsible of cyclin B1 degradation at the end of mitosis. We silenced cdc20 to promote cyclin B1 accumulation (20). As shown in Fig. 5B, cyclin B1 protein level were increased following cdc20 silencing in two of the three resistant cells. In HCC1954/TDR cells, no differences in T-DM1 sensitivity between control (scrambled siRNA) and cdc20-silenced cells were observed. However, T-DM1 resistance was partially reverted in cdc20-silenced SKBR3/TDR and HCC1419/TDR cells (Fig. 5B).

To overcome, at least in part, the issue of the limited number of cell lines, we tested the association between cyclin B1 induction and T-DM1 antiproliferative effect in a panel of 18 fresh human HER2-positive breast cancer explants. The results confirmed such association. We assayed the basal levels of cyclin B1 in 18 HER2-positive breast cancers according to our experience in fresh explants (16). In 7 cases, cyclin B1 was undetected and in the remaining 11, the percentage of tumor cells with detected cyclin B1 staining ranged from 2% to 10% (Fig. 6A). A fraction of these tissues were cultured ex vivo for 5 days in the presence of T-DM1 or control and assayed for modulation in cyclin B1 expression. There were two main observations; first, cyclin B1 was induced in the majority of explants exposed to T-DM1 (12 of 18, 66.6% of the tumors), including those with undetected baseline levels; second, the percentage of tumor cells that stained positive for cyclin B1 was dramatically increased in many specimens (61.1% of cases), a finding consistent with the induction of a G₂–M arrest by T-DM1 (Fig. 6A and B). These findings support the effect of T-DM1 on inducing a persistent accumulation of cyclin B1 in tumor, which was a hallmark of T-DM1 effects in our panel of sensitive breast cancer cells. Of note, in 4 explants with cyclin B1 detected at baseline, T-DM1 exposure did not further increase the percentage of expressing cells. We also assayed induction of apoptosis by expression of active caspase-3, proliferation (Ki67), and activation of the M-phase checkpoint kinases (phospho-Histone H3, p-H3; Fig. 6A). With regards to proliferation, Ki67 staining was not modulated by T-DM1 after 5 days while reduction in p-H3 staining paralleled the induction of apoptosis by T-DM1. The lack of an effect on Ki67 is consistent with the fact that cells arrested in G₂–M also stain positive to Ki67 (Fig. 6A; ref. 21). We found two main patterns of cyclin B1 and apoptosis changes following T-DM1 exposure: no/very low cyclin B1 accumulation and very low upregulation of apoptosis versus strong cyclin B1 (accumulation) and high induction of apoptosis. Trastuzumab alone did not induce cyclin B1 in ex vivo explants (Fig. 6B). The tumor cell areas with cyclin B1 staining and the ones with caspase-3 active staining did not frankly overlapped in the tissue sections, suggesting different stages of T-DM1 effects.

Fifteen explants were from diagnostic specimens derived from patients that received neoadjuvant treatment without T-DM1. The other three were from metastatic patients. Two received T-DM1 and had cyclin B1 and apoptosis induction ex vivo. One had de novo metastatic disease, the explant was from the diagnostic breast cancer biopsy, and received T-DM1 as second line achieving a partial response. The second patient had bone and liver disease and after several lines of treatment received T-DM1. The explant was obtained from a liver metastasis just before T-DM1 (Supplementary Fig. S2) and subsequently had a partial response.

We have generated three different HER2-positive breast cancer cell lines with various levels of acquired T-DM1 resistance by applying a pulsed administration strategy of the drug, an approach commonly used for the development of chemotherapy drug-resistant cancer cells (13). The resistant phenotype was achieved within the first two months of drug exposure, suggesting the emergence of a chemotherapy-driven mechanism. We chose initially four cells lines to develop resistance, but in one of them (BT474) we were unsuccessful. The three remaining cell lines, albeit not sufficient to establish the generalizability of the findings, were sufficient to test mechanistically the role of cyclin B1. Similarly, the degree of acquired resistance varied in the 3 cell lines but a role of cyclin B1 in T-DM1 resistance was observed in all of them. The modest resistance observed in two of the cell lines clearly suggested that cyclin B1-independent mechanisms of action of T-DM1 remain active.

The generation of resistant cell lines in a short timeframe may be caused by several mechanisms and may vary between cell lines. In HCC1954, a specific finding was a marked reduction of HER2 gene amplification after the first round of exposure to T-DM1. In parental HCC1954 cells, there was a predominant subpopulation (∼93% of cells) with high Her2 amplification and a minority subpopulation (∼7%) with low, but amplified HER2 gene. An early clonal selection of the subpopulation with lower HER2 amplification following T-DM1 exposure appears to contribute to resistance. Regardless of this, the rapid emergence of resistance, also in cell lines that retain the same level of HER2 amplification, suggests a mechanistic link with the cytotoxic DM1 component rather than to trastuzumab, by as yet unknown mechanisms such as epigenetic and/or cellular pathway rewiring. SKBR3 T-DM1–resistant cells retained sensitivity to trastuzumab, whereas the other two remained trastuzumab resistant. This suggests that multiple mechanisms of resistance may coexist in the same cell line. Additional studies are needed to understand if these mechanisms coexist in a single cell or whether this represents tumor cell heterogeneity (as in HCC1954 relative to HER2 amplification). Nevertheless, the antimitotic essence of DM1 should be considered as a source of heterogeneity generation.

The cytotoxic effect of T-DM1 might be impaired by inefficient internalization or enhanced percentage of HER2–T-DM1 complex that is recycled back to the cell surface (3). It is believed that the HER2/T-DM1 conjugate enters cancer cells via the clathrin-dependent endocytosis pathway. However, a clathrin-independent mechanism, such as caveolae membranes composed mainly by caveolin-1 has also been demonstrated (22). Furthermore, it has been shown that the high endocytic activity of stem cell–like breast cancer cells make them particularly sensitive to T-DM1 (23). We have shown that T-DM1 was internalized and exhibited a similar intracellular pattern in parental and resistant cells, albeit HCC1954/TDR cells had less T-DM1 detected intracellularly. Overall, it appeared that the pathway mediating HER2-T-DM1 endocytosis was intact in resistant cell lines.

In parental cells, T-DM1 induced G₂–M arrest and mitotic catastrophe. The phenomenon of “mitotic catastrophe” considered a mode of cell death per se resulting from prolonged mitotic arrest is now believed to represent a prestage of apoptosis or even necrosis or senescence (5). In two of the three cell lines sensitive to T-DM1, the mitotic arrest induced by T-DM1 preceded apoptosis. Mitotic catastrophe did not occur in resistant cells suggesting that the machinery for inducing cell-cycle arrest at the G₂–M phase was disrupted.

This led us to hypothesize a potential involvement of CDK1 and cyclin B1, the members of mitotic promoting factor complex, in T-DM1 resistance (18). Entry into mitosis is initiated by CDK1 and binding of CDK1 to cyclin B1 is essential for its activation. CDK1/cyclin B1 complex and the kinase activity of CDK1 is controlled by cyclin B1 accumulation. During mitosis, chromosome segregation is facilitated by the kinetochore, an assembly of proteins built on centromeric DNA that attach chromosomes to spindle microtubules. After the last unattached kinetochore is attached to microtubules and the chromosomes are properly aligned, the mitotic checkpoint is switched off. CDK1 is then inactivated as cyclin B is rapidly degraded, and cells progress through anaphase, undergo cytokinesis, and exit mitosis (24). However, in the presence of an antimitotic agents (such as DM1), the impossibility of assembling the mitotic spindle activates the mitotic checkpoint arresting the cells in mitosis for a prolonged period and they may undergo mitotic catastrophe followed by cell death (25).

In the three parental HER2-positive breast cancer cells, cyclin B1 induction and CDK1/cyclin B1 activation were observed following T-DM1 treatment and were maintained elevated for up to 24 hours. However, T-DM1 failed to raise cyclin B1 levels and, consequently, CDK1/Cyclin B1 activity in T-DM1–resistant cells. Furthermore, silencing of cyclin B1 induced resistance to T-DM1 in parental cell lines while increasing the levels of cyclin B1 in resistant cells sensitized them to T-DM1 in two out of the three cell lines. The failure to generate T-DM1 BT474–resistant cells under our experimental protocol may be related, at least in part, to the inability of T-DM1 to upregulate cyclin B1.

Other CDK–cyclin complexes have been previously reported as implicated in resistance to anti-HER2 therapies. For instance, cyclin E has a role in trastuzumab resistance and treatment with CDK2 inhibitors has been proposed for tumors displaying cyclin E amplification/overexpression (26, 27). A relationship of CDK4/cyclin D1 activity in resistance to trastuzumab and lapatinib has been also reported and high levels of cyclin D1 predicted poor response to trastuzumab (28). Preclinical studies have also shown that residual cells surviving within 48 hours following T-DM1 treatment began to reenter the cell cycle and the sequential treatment with CDK4/6 inhibitors suppressed the proliferation of these residual/resistant clones (29). A clinical trial with the CDK4/6 inhibitor palbociclib in combination with T-DM1 (NCT01976169) in HER2-positive patients is underway. On the other hand, our results showing that CDK1/Cyclin B1 activity is needed for T-DM1 action suggests a note of caution regarding possible combinations of T-DM1 with pan-CDKs inhibitors or selective CDK1 inhibitors (30).

In human breast cancer, cyclin B1 expression has been associated with poor survival (31) and is a prognostic proliferation marker in lymph node–negative breast cancer cohorts (32). In our series of fresh HER2-positive breast tumor explants, the basal levels of cyclin B1 were not significantly related to T-DM1 apoptotic or antiproliferative effects. Instead, we found that the induction of cyclin B1 following T-DM1 exposure ex vivo was highly associated to induction of apoptosis and reduced tumor cell proliferation.

Interestingly, two patients with metastastic breast cancer had a response to T-DM1 and in both of them cyclin B1 was upregulated by T-DM1 ex vivo. These anecdotal data suggest that cyclin B1 induction in explants may be associated to clinical response. On the other hand, the lack of a predictive effect of baseline cyclin B1 expression is in line with an observation in 7 HER2-positive breast cancer patients that had been treated at Hospital del Mar (Barcelona, Spain) with T-DM1 as part of a neoadjuvant clinical trial. Four of them had detectable levels of cyclin B1, of whom 3 achieved a pathologic complete response (pCR). In the 3 patients with undetected baseline cyclin B1, one pCR was achieved. Albeit patients received additional anti-HER2 agents or systemic chemotherapy as part of their neoadjuvant treatment, these few cases suggest that baseline cyclin B1 expression is not a prerequisite to achieve a pCR with a neoadjuvant regimen including T-DM1. We plan to prospectively test cyclin B1 induction as a pharmacodynamics assay (i.e., serial biopsies comparing baseline cyclin B1 expression with expression at an early time point after T-DM1 treatment) to predict T-DM1 clinical benefit in a GEICAM (Spanish Breast Cancer Research Group) study in the near future.

A few additional mechanisms of T-DM1 resistance have been proposed. For instance, those factors that reduce the intracellular DM1 load per cell, namely the overexpression of multidrug resistance (MDR) proteins or an impaired lysosomal degradation of trastuzumab that might limit the subsequent release of DM1 into the cytoplasm (3, 33). Although we cannot rule out a potential role of these mechanisms in our models, we believe that MDR did not play a significant role since paclitaxel, a microtubule-stabilizing agents and a substrate of MDR, exhibited the same cytotoxicity in parental and in resistant cells.

With regards to alterations in the endosome pathways, we believe that the detection of intracellular T-DM1 in the resistant cells, as well as the effects of cdc20 silencing on restoring T-DM1 response in resistant cells, limits the potential role of this putative mechanism of resistance. Nonetheless, a biparatopic HER2-targeting ADC containing the tubulysin variant AZ13599185 demonstrated superior antitumor activity over T-DM1 in various tumor models including T-DM1 refractory. The new ADC by targeting two nonoverlapping epitopes on HER2 can induce HER2 receptor clustering, which in turn promotes robust internalization, lysosomal trafficking, and degradation (34). Also, it has been suggested a combination of inhibitors of the chaperone HSP90 that promote HER2 targeting to lysosomes and its degradation suggested as a new strategy to improve therapy with T-DM1 (35). In addition to this, other novel anti-HER2 antibody–drug conjugate offer promising activity in T-DM1–pretreated breast cancer (36, 37). Other potential mechanisms of resistance that come from the trastuzumab part of the T-DM1 molecule have not proven its clinical utility when assayed in patient samples from clinical trials (9, 11). A mechanism of T-DM1 resistance that has been reported preclinically is the presence of the HER3 ligand, heregulin, that reduced the activity of T-DM1 in breast cancer cells by causing HER2/HER3 dimerization thus strongly activating the PI3K pathway; and this effect was reversed by the addition of pertuzumab, a HER2–HER3 dimerization inhibitor. However, the combination of T-DM1 and pertuzumab in the clinic has not proven sufficiently superior to T-DM1 alone, suggesting this mechanism operates at most in a small proportion of patients (38). Finally, T-DM1 is able to alter genes related with immune response and this can influence the response in patients (39).

In short, the studies presented here show that cyclin B1 induction allowed to differentiate T-DM1 sensitive parental cells from their counterparts with acquired resistance. The induction of cyclin B1 in T-DM1–sensitive cells was independent of their prior trastuzumab sensitivity. The mitotic arrest consequence of the failure in assembling the mitotic spindle leads to sustained CDK1/Cyclin B1 kinase activity, a hallmark of mitotic catastrophe that is resolved by apoptosis. In fresh human HER2-positive breast cancer explants, the induction of cyclin B1 by T-DM1 correlates with apoptosis, suggesting that a pharmacodynamic assay to test cyclin B1 induction in breast cancer patients treated with T-DM1 may help to identify early the patients more likely to benefit from this drug.

J. Albanell reports receiving speakers bureau honoraria from and is a consultant/advisory board member for Roche. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M. Sabbaghi, G. Gil-Gómez, P. Eroles, J. Arribas, F. Rojo, A. Rovira, J. Albanell

Development of methodology: M. Sabbaghi, G. Gil-Gómez, C. Guardia, O. Arpi, S. Menendez, S. Zazo, C. Chamizo, P. González-Alonso, F. Rojo, A. Rovira, J. Albanell

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Sabbaghi, G. Gil-Gómez, S. Servitja, O. Arpi, S. García-Alonso, S. Menendez, M. Arumi-Uria, L. Serrano, A. Muntasell, M. Martínez-García, A. Lluch, F. Rojo

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Sabbaghi, G. Gil-Gómez, C. Guardia, S. Servitja, S. García-Alonso, S. Menendez, M. Salido, A. Muntasell, P. González-Alonso, J. Madoz-Gúrpide, I. Tusquets, A. Pandiella, F. Rojo, J. Albanell

Writing, review, and/or revision of the manuscript: M. Sabbaghi, G. Gil-Gómez, S. Servitja, S. García-Alonso, L. Serrano, A. Muntasell, M. Martínez-García, P. Eroles, J. Arribas, I. Tusquets, A. Lluch, A. Pandiella, F. Rojo, A. Rovira, J. Albanell

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): O. Arpi, J. Madoz-Gúrpide, F. Rojo, J. Albanell

Study supervision: I. Tusquets, F. Rojo, A. Rovira, J. Albanell

The authors would like to thank Marc Bataller and Raul Peña for technical assistance and comments. Flow cytometry analysis was technically supported by Oscar Fornas, head of Universitat Pompeu Fabra (UPF) flow cytometry core facility. We thank Fundació Cellex (Barcelona) for a generous donation to the Hospital del Mar Medical Oncology Service. We also thank the patients for their generous participation in the study.

This work was supported by ISCiii (CIBERONC CB16/12/00481, RD12/0036/0051, RD12/0036/0070, RD12/0036/0003, PIE15/00008, PI13/00864, PI15/00146, PI15/00934, PI15/01617, PT13/0010/0005), Generalitat de Catalunya (2014 SGR 740), and the "Xarxa de Bancs de tumors sponsored by Pla Director d'Oncologia de Catalunya (XBTC). MINECO through BFU2015-71371-R grant supported work in A. Pandiella's laboratory. Our work was supported by the EU through the regional funding development program (FEDER). P. González-Alonso was supported by Fundación Conchita Rábago de Jiménez Díaz grant.

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.