mTOR Inhibition and T-DM1 in HER2-Positive Breast Cancer

HER2 gene amplification or protein overexpression (HER2 positivity) occurs in approximately 15% to 20% of breast cancers and confers poor prognosis (1). However, the addition of anti-HER2 antibodies to standard chemotherapy significantly improved the outcome of patients with HER2-positive breast cancer (1–3). More recently, antibody–drug conjugates (ADC), combining the specificity of mAbs with the cytotoxic potential of chemotherapeutic drugs, have proven highly effective in breast cancer (4, 5). Ado-trastuzumab emtansine, also called T-DM1, consists of a potent anti-tubulin maytansinoid (DM1) bound by a noncleavable linker to the mAb trastuzumab. T-DM1 was the first approved ADC for therapy in solid tumors and is a standard of care for patients with metastatic HER2-positive breast cancer that have progressed to first-line trastuzumab-based therapy or have residual disease following neoadjuvant anti-HER2–based therapy (6, 7). Still, a considerable proportion of patients may present with de novo or acquired resistance to anti-HER2 agents. Consequently, the pursuit of novel strategies that increase response rates and/or delay the appearance of therapy resistance remains one of the primary interests of research in this field.

In an elegant preclinical study, mTOR inhibition greatly improved the antitumor effects of trastuzumab (8). The mTOR protein is a main downstream effector of PI3K signaling. mTOR signals through two different multiprotein complexes termed mTORC1 and mTORC2 (9). Two ensuing randomized clinical trials testing the combination of trastuzumab-based therapy plus everolimus (a mTORC1 inhibitor) in patients with HER2-positive breast cancer lead to statistically significant improvements in progression-free survival, but considering the modest magnitude of the benefit and the added side effects, it did not reach routine clinical practice (10, 11). The combination of mTOR inhibitors and T-DM1 remains unexplored, likely in part due to the results mentioned above.

In contrast to trastuzumab, T-DM1 requires receptor-mediated internalization and entrance into the endolysosomal pathway. Once T-DM1 reaches the lysosome, it suffers proteolytic degradation by lysosomal proteinases leading to the release of lysine-linked DM1. Linker-bound DM1 then traverses the lysosomal membrane and accesses the cytosol to exert its antimitotic effect through binding to tubulin (12). Consistent with this, altered T-DM1 intracellular trafficking and lysosomal processing has been postulated as one of the main resistance mechanisms to T-DM1 (13–17). Other mechanisms of resistance that have been described are decreased HER2 expression, upregulation of multidrug transporters, loss of cyclin B1 induction, and increased PI3K pathway activation (17, 18).

Due to the important role that lysosomal processing plays in the mechanism of action of T-DM1 and since mTOR is a master regulator of lysosome dynamics (9, 19, 20), we felt that the combination of T-DM1 with mTOR inhibitors deserved testing. Here, we examined the effects of mTOR inhibition plus T-DM1 in a wide panel of HER2-positive breast cancer cells, including two patient-derived xenograft (PDX) primary culture models. The combination was superior to T-DM1 or everolimus alone in HCC1954, EFM192A, MDA-MB-361 lines, and in two cultures from primary tumor cells derived from HER2-positive PDXs, but not in BT474 and SKBR3 cells. In BT474 and in the trastuzumab-resistant HCC1954 cell line, we characterized the biochemical and cellular effects of mTOR inhibition as well as lysosomal dynamics in different experimental conditions. Finally, we confirmed the superior antitumor activity of T-DM1 plus everolimus (TE) in vivo in subcutaneous and orthotopic HCC1954 mouse xenografts and in engrafted cells from PDX-derived primary cultures.

Breast cancer cell lines HCC1954, SKBR3, BT474, EFM192A, and MDA-MB-361 were obtained from the ATCC. Authenticity of the cells was tested by short tandem repeat (STR) DNA profiling analysis at the ATCC. PDX118 and PDX433 were obtained from de Vall d'Hebron Institute of Oncology (Barcelona, Spain). These PDX models were established from sites of distant metastases in Vall d'Hebron Institut of Oncology (21, 22) and recently used by our group. PDX-118 was established after the patient received systemic treatment with trastuzumab, lapatinib, and standard chemotherapeutics and it is from a skin biopsy. PDX-433 was a hepatic metastasis where the patient received trastuzumab and T-DM1. The HCC1954 and EFM192A were grown in RPMI and SKBR3 and BT474, PDX118 and PDX433 were grown in DMEM: F12, supplemented with 10% FBS, and containing high glucose (4500 mg/liter) and antibiotics (penicillin 100 U/mL, streptomycin 100 µg/mL). MDA-MB-361 were grown in DMEM supplemented with 20% FBS and antibiotics (penicillin 100 U/mL, streptomycin 100 µg/mL). Cell lines were cultured at 37°C in a humidified atmosphere in the presence of 5% CO₂ and 95% air. The number of passages between thawing and use in the described experiments was fifteen or less. Detection of Mycoplasma was conducted at cell culture core facility of our institution. T-DM1 (Kadcyla) was provided by the Hospital del Mar pharmacy (Barcelona, Spain). Everolimus, TAK-228, bafilomycin A1, and chloroquine phosphate were obtained from Selleckchem. DM1 (mertansine; HY-19792/CS-5804) was obtained from MedChemExpress. For in vitro studies, drugs were prepared at 10 mmol/L in DMSO and stored at −20°C. Chloroquine phosphate was prepared at 100 mmol/L in water and stored at −20°C. T-DM1 was prepared at 20 mg/mL in water and stored at −20°C. For in vivo studies, everolimus was prepared in PEG400 and T-DM1 was prepared in physiologic serum. IgG1 isotype control from Sigma-Aldrich (I5154) was prepared in physiologic serum.

Cells were plated in triplicate in 12-well plates at a density of 15,000 to 70,000 per well (based on the optimal density for each cell line) in 1 mL of culture medium. After 24 hours, the cells were treated with drugs, either alone or in combination, for 72 hours. Two different approaches (automated cell counting or crystal violet staining) were used to assess the effects of different treatments on cell viability. The cells were trypsinized and resuspended in PBS for counting using the Scepter Automated Cell Counter (Millipore). For the experiments with crystal violet, at the end of the experiments, cells were stained with crystal violet solution (10% acetic acid, 10% absolute ethanol, and 0.06% crystal violet) for 1 hour. After that, the plates were washed with PBS. Images of each plate were scanned and quantified using ImageJ and plotted as arbitrary units.

Western blots were performed according to standard protocols. Cells were plated at a density of 8 × 10⁵ to 1,2 × 10⁶ in 100 mm2 dishes (based on the optimal density for each cell line) and, after 24 hours, cells were treated with drugs as indicated in figure legends. The following antibodies were used: p-Akt XP (Ser473; 4060), Akt (9272), p-S6 (Ser235/236; 2211), S6 (2217), LC3B-I-II (2775), p62/SQSTM1 (5114), cleaved-PARP (9541), p-ULK1 (Ser757; 6888), ULK1 (8054), LAMP1 (9091), regulatory protein associated with mTOR (raptor) and mTOR from Cell Signaling Technology, cathepsin B (sc-6493) from Santa Cruz, α-tubulin (T5168) and β-actin (A-5316; Sigma-Aldrich), rapamycin insensitive companion of mTOR (rictor; Bethyl laboratories), and calnexin (Stressgen Biotechnologies Corporation). Anti-mouse (NA931) and anti-rabbit (NA934) horseradish peroxidase (HRP)-conjugated secondary antibodies purchased from GE Healthcare Life Sciences.

T-DM1 was analyzed by immunoprecipitation with protein A-sepharose, followed by Western blotting with antibodies to either human Ig (that recognize trastuzumab) or anti-DM1 antibodies. The latter were prepared in-house by injecting rabbits with keyhole limpet hemocyanin-coupled DM1, followed by purification of the specific anti-DM1 antibodies by affinity chromatography over a column prepared with BSA-copled DM1.

Lentiviral infection was performed as described in (23). The lentiviral vectors containing short hairpin RNA (shRNA) for raptor, rictor, and mTOR were obtained from Addgene.

Cells were plated at a density of 8 × 10⁵ in 100-mm² dishes and after 24 hours cells were treated with the different drugs. For cell cycle analysis, cells were fixed by 70% of ethanol overnight and then stained with the Muse Cell Cycle reagent (Merck Millipore) during 30 minutes at room temperature in the dark. Stained cells were analyzed by Muse Cell Analyzer (Millipore) according to manufacturer's protocol. Apoptosis was assayed by determining cleaved PARP by Western blot.

Lysosome number was measured by Lysotracker Red DND-99 (Molecular Probes) staining according to manufacturer's protocol. Internalization of T-DM1 and its colocalization with lysosomes (i.e., intracellular accumulation) were measured by detecting T-DM1 and LAMP1 as described in (13). Briefly, acidic organelle staining was followed with 100 nmol/L LysoTracker Red DND-99, which was added 30 minutes before fixing. Phase contrast images were obtained using conventional photomicroscopy.

All animal work was conducted following the Barcelona Biomedical Research Park (PRBB) Institutional Animal Care and Scientific Committee guidelines. Five-week-old female BALB/c nude mice were subcutaneously inoculated in their flank or in the mammary fat pad with 1.5 × 10⁵ (subcutaneous) and 1.5 × 10⁶ (orthotopic) HCC1954 cells mixed with matrigel. Five-week-old female NOD/SCID mice were subcutaneously inoculated in their flank with 5 × 10⁶ PDX118 cells mixed with matrigel. Tumor growth was measured two or three times a week depending on its growth. When tumors reached approximately 100 to 200 mm³ mice were randomized to four groups. Treatment groups are indicated in figure legends. T-DM1 was administered at 1 mg/kg by intravenous injection on days 1, 22, and 43 of treatment. Everolimus was administered at 1 mg/kg by oral gavage five times a week. At the end of the experiment, animals were sacrificed and tumor tissue was harvested frozen or formalin-fixed and paraffin-embedded (FFPE).

FFPE fixed samples blocks were cut in 3-µm tissue sections were immunostained in a Dako Link platform (pH3 and cleaved caspase-3) or HER2 IHC assays PATHWAY (Ventana/Roche; HER2). The following antibodies were used: cleavedcaspase-3 (9664) and pH3 (9701) from cell signaling and HER2 (790–2991) from Roche. The percentages of cleaved caspase-3 and pH3-positive tumor cells for each case were determined according to our own procedures (24). Xenograft tumor HER2 expression was determined by Ventana's HER2 IHC assays PATHWAY, an FDA-approved in vitro diagnostic test marketed by Roche.

Human primer pair sequences are SLC46A3 forward: 5′-TTGGATTCACCACTCTGCTG-3′ and reverse: 5′-GAGGCACTACCCAAAGCTGA-3′ (25); GAPDH forward: 5′-GGAGTCAACGGATTTGGTCGTA-3′ and reverse: 5′ GGCAACAATATCCACTTTACCAGAG-3′. Gene expression was calculated as 2 to the power of - ΔΔCt, where ΔΔCt = (Ct_Gene – Ct_{ATP5E or GAPDH}) assay.

Statistical analysis was performed by SPSS version 18.0 (SPSS, Inc.). For in vitro experiments, differences in means between conditions was assessed using one-to-one t tests, with the null hypothesis being no differences between groups. For in vivo experiments two-way ANOVA with posthoc Tukey test for pairwise differences was performed. Contrasts with P < 0.05 were considered statistically significant. Significant group-group differences are depicted in corresponding plots using different thresholds, as described in the figure legends. Barplots and Boxplots were designed using Graph Pad Prism v6.05. The boxplot in Fig. 4F was generated using the ggplot2 library on R software version 4.3.0.

Everolimus forms a complex with FKBP12, binding to the FRB (FKBP12-rapamycin–binding) domain of the mTORC1 complex and acting as a selective allosteric inhibitor of mTOR (9). We tested the effects of T-DM1 with and without everolimus on the cell viability of seven HER2-positive breast cancer cell lines with different therapeutic profiles: two trastuzumab-resistant cell lines (HCC1954 and MDA-MB-361), three trastuzumab-sensitive lines (SKBR3, EFM192A, and BT474), and in two cultures from primary tumor cells derived from HER2-positive PDXs, PDX118 and PDX433 (refs. 21, 22; see Methods; and Supplementary Fig. S1A). We used everolimus at 100 nmol/L because at this concentration its effect on cell viability plateaued (Supplementary Fig. S1B; ref. 26). In all cells but BT474, addition of everolimus to T-DM1 caused a significant decrease in cell viability compared with T-DM1 alone (Fig. 1A; Supplementary Fig. S1C). In SKBR3 cells, the activity of the combination was not significantly superior compared with everolimus alone.

Since T-DM1 strongly depends on lysosomal processing to exert its cytotoxic effect (12), we reasoned that the addition of chloroquine would decrease T-DM1 efficacy. Chloroquine is a dibasic lipophilic amine drug that passively diffuses through cell membranes. When chloroquine reaches the acidic lumen of lysosomes, it becomes protonated and consequently trapped (27, 28), impairing lysosomal function. Of note, chloroquine by itself has shown to inhibit cell proliferation and induce apoptosis in preclinical breast cancer models at concentrations of 10 µmol/L or higher (29, 30). Therefore, we chose a relatively low concentration of chloroquine (5 µmol/L) to avoid any significant impact on cell viability while still impacting lysosomal function. Overall, we observed a decrease in T-DM1 antitumor effects after the addition of chloroquine (Fig. 1A). This was observed in all cell lines except MDA-MB-361, in which notably T-DM1 alone also caused no significant decrease in cell numbers. Furthermore, chloroquine did not impact cell viability or modify the activity of everolimus, coherent with the antiproliferative effect of everolimus not depending on lysosome function. In HCC1954, EFM192A, and MDA-MB-361 cells, chloroquine also reverted the effects of the TE combination, although this effect was modest. Findings from these initial experiments were compatible with our initial hypothesis, namely that everolimus increases T-DM1 efficacy in HER2-positive breast cancer. However, our observations in BT474 suggested that certain cells may be resistant to the TE combination.

We subsequently decided to study the TE combination in a trastuzumab-resistant model. Based on the literature, HCC1954 cells are trastuzumab-resistant in vitro (31) and in vivo (32). Thus, we henceforth focused on this model to further characterize the effects of the TE combination. Furthermore, our findings in BT474 cells suggested that our results may not be universal across all HER2-positive breast cancer models. Hence, we ran parallel experiments in BT474 cells to explore potential mechanisms of resistance to the combination. We present these data in a dedicated section in the text, in Fig. 4 and Supplementary Fig. S3.

To confirm the role of lysosome inhibition in T-DM1 antitumor activity, we first replaced chloroquine with the lysosomal proton pump V-ATPase inhibitor bafilomycin A1 (Supplementary Fig. S1D). We observed a significant abrogation of both T-DM1 and TE activity, supporting the notion that impaired lysosomal function decreases T-DM1 efficacy (13, 14). Second, we substituted T-DM1 for unconjugated DM1 (uDM1), which freely penetrates cell membranes and thus does not need lysosomal processing to inhibit tubulin polymerization. Like TE, the combination of uDM1 and everolimus significantly decreased cell viability compared with each drug alone (Fig. 1B). In contrast with the experiments with T-DM1, the addition chloroquine did not revert the effect of uDM1 alone or in combination with everolimus.

Next, to confirm the involvement of mTOR as a regulator of the antitumoral efficacy of T-DM1, we repeated the previous experiments replacing everolimus with TAK-228, an ATP-competitive mTOR kinase inhibitor (33). As shown in Fig. 1C, treatment with TAK-228 mimicked the results of the experiments with everolimus in terms of cell viability. Next, we explored T-DM1 efficacy after knockdown of different components of the mTOR. We either silenced Raptor or Rictor, which are essential components of mTOR complex 1 and 2, respectively. We also knocked down the mTOR kinase, which abrogates signaling through both complexes. Western blot confirmed a significant decrease of each knocked down component (20.7% mTOR kinase vs. control, 80.2% Raptor vs. control, and 49.8% Rictor vs. control) which was preserved in conditions with T-DM1 (34% mTOR kinase vs. control, 74% Raptor vs. control, and 40.4% Rictor vs. control; Fig. 1D, left). Notably, we observed a significant decrease in cell viability by knocking down any of the mTOR components in the absence of T-DM1 (Fig. 1D, right), which was consistent with our previous results with everolimus and TAK-228. In this regard, the largest effect was observed after knockdown of the mTOR kinase. Addition of T-DM1 further decreased cell viability in all conditions. However, this decrease after the addition of T-DM1 was only significant after either mTOR kinase or Raptor knockdown. Together, the results from both pharmacologic and genetic inhibition of mTOR suggested that the effects of the TE combination were primarily due to mTORC1 inhibition.

To study the role of everolimus in T-DM1 intracellular trafficking, we first analyzed T-DM1 intracellular accumulation, a trait that was associated with sensitivity to T-DM1 in a recent study (13). Indeed, we observed an increased accumulation of T-DM1 with the addition of everolimus (Fig. 2A). We then analyzed the influence of each treatment on lysosomal biogenesis by using LAMP1 (a specific marker of lysosomal membrane) and cathepsin B (a lysosomal enzyme) levels as surrogate markers of lysosome number. As expected, the addition of everolimus increased LAMP1 and Cathepsin B levels, an effect that was sustained over time (Fig. 2B). Furthermore, everolimus also increased lysosome number as measured by LysoTracker Red immunofluorescent staining, which occurred independently of T-DM1 (Fig. 2C). We did not observe significant changes in lysosome numbers with T-DM1 alone.

Next, we studied T-DM1 internalization and its colocalization with lysosomes. Cells were treated with a 15-minute pulse of T-DM1, after which the drug was washed out (Fig. 2D). Immediately after the pulse, most T-DM1 was bound to the cell membrane. After 24 hours, most T-DM1 had been processed or externalized and consequently, T-DM1 staining decreased. With the addition of everolimus, intracellular T-DM1 concentration increased and was mainly localized to the lysosomal compartment. Interestingly, we observed a clearly higher amount of intracellular T-DM1 in the conditions with chloroquine. We speculate that this is due to the lysosomal blockage caused by chloroquine on the final steps of T-DM1 processing, and that what we observe corresponds to the high numbers of T-DM1 molecules that appear sequestered in the lysosomal compartment. Furthermore, this amount appears to be greater in the condition with everolimus, suggesting that everolimus may increase cellular uptake of T-DM1. This would be consistent with previous studies showing increased ligand-receptor internalization after mTOR inhibition (34). Together with the results from the previous experiments, these results suggested that mTOR inhibition increases intracellular T-DM1 levels, leading to increased lysosomal processing of the drug.

To study the integrity of the mTOR pathway in the different experimental conditions, we analyzed the phosphorylation of S6 ribosomal protein (pS6_S235/236) and Akt (pAkt_Ser473), which are downstream effectors of mTORC1 and mTORC2, respectively. Consistent with its mechanism of action, everolimus abrogated phosphorylation of S6 (Fig. 3A). The effect of everolimus on the pathway was not altered by T-DM1, uDM1, chloroquine, or bafilomycin A1 (Fig. 3A; Supplementary Fig. S2A and S2B) TAK-228 also decreased the levels of pS6, although to a lesser extent than everolimus (Fig. 3B) and this was not significantly altered in conditions T-DM1 or CHQ.

In contrast to everolimus, TAK-228 also decreased phosphorylation of AKT, which is consistent with its dual mTORC1 and mTORC2 inhibition. Therefore, the effects observed on the mTOR pathway were coherent with the known mechanisms of action of everolimus and TAK-228. These effects did not differ appreciably with the addition of any of the other compounds, including T-DM1.

We then studied the effect of the TE combination on apoptosis induction and cell-cycle modifications. According to the literature and our previous experience (18), T-DM1 causes mitotic arrest at G₂–M phase and ultimately cell death due to mitotic catastrophe. We did not observe a significant increase in G₂–M cells with the TE combination with respect to the other conditions. Rather, a predominant increase of cells in G₀–G₁ was observed in the presence of everolimus, which persisted in the TE condition (Fig. 3C). These results did not differ significantly when experiments were repeated at 48 hours (Supplementary Fig. S2C). Regarding apoptosis, we observed similar levels of cl-PARP between T-DM1 alone and the TE combination (Fig. 3D). Thus, the cumulative decrease in cell viability observed with T-DM1 after mTOR inhibition in HCC1954 does not appear to be caused by a clear increase in apoptotic or G₂–M cells.

Besides regulating lysosomal biogenesis, mTOR also regulates autophagy (9, 20). Autophagy is an evolutionarily conserved catabolic process that mediates the degradation of large protein aggregates or damaged organelles (35). In recent years, autophagy has gained momentum as a crucial mechanism for cancer cell survival and therapeutic resistance (36). Inhibition of mTOR is a bona fide method to induce autophagy. In this context, lysosome inhibitors such as chloroquine or bafilomycin A1 are commonly used to assess variations in autophagic flux, since they block the final stages of the autophagic process (37). To assess if the effect of the TE combination was dependent on autophagic activity, we assessed autophagic flux in the different experimental conditions and using different markers. Briefly, we analyzed LC3B-I/LC3B-II turnover as well as changes in SQSTM1/p62 and pULK1 levels (37). An increase of the LC3B-II/LC3B-I (lipidated/unlipidated LC3B) ratio is used to measure autophagosome formation and is especially apparent in the presence of late-phase autophagy inhibitors such as chloroquine. Indeed, we observed a clear increase in LC3B-II accumulation in the presence of everolimus and TAK-228, which was not evident with T-DM1 alone (Fig. 3E and F; Supplementary Fig. S2D and S2E). Degradation of p62 is another widely used marker to monitor autophagic activity because p62 directly binds to LC3B and is degraded during the autophagic process, and Atg1/pULK is involved predominantly in the induction of autophagy by mTOR. Coherent with the changes observed in LC3B-II, we observed a decrease of p62 and pULK in the conditions with everolimus and TAK-228. Chloroquine (5 µmol/L) caused the expected increase in LCB3I/II ratio and the expected changes in p62, namely a slight accumulation or stability (similar to control or T-DM1 conditions) when compared with conditions in which autophagy is inducted (everolimus and TAK-228; Supplementary Fig. S2D and S2E). Importantly, T-DM1 alone (with or without chloroquine) did not appear to induce or inhibit autophagy. These findings suggest that autophagy activation was preserved in all our experimental conditions and was primarily driven by mTOR inhibition.

Despite not observing a significant decrease in cell viability with the addition of mTOR inhibitors to T-DM1 in BT474 cells (Figs. 1A and 4A), results thus far showed that the effects of T-DM1 with and without mTOR inhibition regarding mTOR downstream signaling (Fig. 4B), autophagy (Supplementary Fig. S3A), as well as T-DM1 internalization and accumulation (Fig. 4C and D; Supplementary Fig. S3B) were overall consistent with those observed in HCC1954 cells.

Furthermore, and in contrast to what we observed with the TE combination, uDM1 plus everolimus did show a cumulative decrease in cell viability in BT474 cells (Fig. 4E). However, consistently with our previous experience (18) and that of other groups (13, 15), we had to employ a 10-fold–higher T-DM1 concentration in BT474 than in HCC1954 to achieve a similar decrease in cell viability. Similarly, higher uDM1 concentrations were needed in BT474 to achieve effects similar to the ones observed in HCC1954. Together, these findings suggest that BT474 may be less sensitive to T-DM1 both by lacking optimal lysosomal processing of T-DM1 as well as by having intrinsic resistance to the DM1 moiety. With regard to the former, several studies have highlighted the potential role of specific lysosomal proteins such as V-ATPase or SLC46A3 (16, 17, 38) in T-DM1 resistance. Such reports suggest that SLC46A3 expression is required for exporting the product of cleaved T-DM1, lysine-MCC-DM1, from the lysosome to the cytosol (16, 17), a role that appears to be specific to all noncleavable linker-based ADCs (16, 38). Interestingly, we explored the expression of SLC46A3 in our cell lines and found that BT474 expression of the transporter is among the lowest within HER2-positive cells, while HCC1954 has one of the highest (Fig. 4F and G).

We established three tumor xenograft models to test the in vivo antitumor activity of the TE combination. Cells were injected subcutaneously in the flank (HCC1954 and PDX118), as well as orthotopically in the mammary fat pads (HCC1954) of immune-deficient mice. Mice were then treated with either T-DM1, everolimus alone, or the combination of both drugs (TE). Everolimus was administered at 1 mg/kg by oral gavage five times per week, based on published data (39). As shown in Fig. 5A–C, TE was superior to T-DM1 in all models. In the PDX118 model, this difference did not reach a statistical significance. Of note, engrafted tumors from these cells grew much slower than those from HCC1954. We also noted that PDX118 cells were more sensitive to everolimus than HCC1954, which we had also confirmed in vitro (Supplementary Fig. S1B).

Tumors treated with TE showed a decrease in mitotic markers and increased apoptosis, as determined by phospho-histone H3 and cleaved caspase-3 levels, respectively. In addition, the tumors treated with the combination showed ischemic necrosis. The results of the subcutaneous xenografts are shown in Fig. 5D and the results of the orthotopic xenografts are shown in Supplementary Fig. S4A. Similar findings were observed in the PDX118 xenografts (Supplementary Fig. S4B). Finally, it has been debated that HER2 downregulation plays an important role in the efficacy of anti-HER2 treatments (17, 40–42). In HCC1954 models, we found no effect on HER2 expression neither in vitro (Supplementay Fig. 4C) nor in vivo (Supplementary Fig. S4D) in any of the experimental conditions. Altogether, the results from the in vivo models corroborated the increased efficacy of TE compared with T-DM1 alone, by decreasing cell division and increasing tumor cell apoptosis.

Here we report that T-DM1 plus everolimus in HER2-positive breast cancer is superior to T-DM1 alone in a wide range of HER2-positive cell lines. Interestingly, we observed significant activity of the combination with a relatively low concentration of T-DM1, compared with our previous experience with T-DM1 in the same cell lines (18). We observed in five out of seven cell lines that the combination of T-DM1 and everolimus had a greater effect on cell viability that either T-DM1 or everolimus alone. Our results suggest that the improved effect of TE may be a common effect in HER2-positive breast cancer, but not universal. We selected HCC1954 cells for further studies because (i) they were sensitive to the combination, (ii) they are trastuzumab-resistant, and (iii) they grow well in vivo. Indeed, we found promising antitumor activity of TE in both subcutaneous and orthotopic HCC1954 mouse models. We also validated the results of the TE combination in an in vivo model of PDX-derived primary culture cells stemming from distant metastases from a patient treated with trastuzumab and lapatinib. Together, our findings indicate that TE warrants further investigation in the clinical setting.

Most clinical experience with mTOR inhibition in HER2-positive breast cancer comes from studies in combination with trastuzumab, since activation of the PI3K/mTOR pathway was shown to drive resistance to trastuzumab in preclinical models (8). Two studies addressed this question by adding everolimus to trastuzumab and chemotherapy, failing to show a significant benefit compared with trastuzumab alone (BOLERO-1; ref. 10), or reporting a modest clinical benefit at the expense of increased toxicity (BOLERO-3; ref. 11). The results of these studies may explain the relative scarcity of clinical trials combining mTOR inhibitors and T-DM1. In contrast, promising evidence regarding the combination of T-DM1 with PI3K inhibitors is emerging. At the time of writing this manuscript, at least two clinical trials of T-DM1 with PI3K inhibitors had been published. In this context, the PI3Kα isoform-specific inhibitor alpelisib was tested in a phase I trial, with 6 out of 14 patients (43%) showing an objective response. Of note, 3 of the responding patients (out of 10 included in the study) had received previous T-DM1 (43). Similarly, a response was observed in eight out of 24 patients participating in a phase I trial with taselisib and T-DM1, and they seemed to occur independently of previous T-DM1 treatment and/or PIK3CA mutation (44). Other trials testing the combination of T-DM1 and PI3K inhibitors are ongoing (NCT02390427, NCT00928330). Interestingly, the design of these studies appears to be based on the same hypothesis as the BOLERO studies, namely that PI3K activation is a common resistance mechanism to early-line anti-HER2 therapy. To the best of our knowledge, there is only one study, reported thus far only as abstract (45), that has previously reported increased activity of T-DM1 with mTOR inhibition in a trastuzumab-resistant breast cancer model. Importantly, however, it was conducted on only two cell lines (KPL4 and MCF7 neo/HER2) different from any of those employed in our study, and the experimental drug (GDC-0980) was a dual PI3K/mTOR inhibitor.

The different effects observed in each cell line used in our study may be explained by their different molecular backgrounds. Subgroup analyses of the BOLERO-1 and BOLERO-3 trials revealed that only patients with hormone receptor (HR)-negative disease derived a significant benefit from the addition of everolimus (10, 11, 46). Similarly, a molecular analysis of patients included in both studies showed that only patients whose tumors harbored a “hyperactive” PI3K pathway (defined as the presence of PTEN loss or PI3KCA activating mutation in tumor tissue) benefitted from everolimus plus trastuzumab-chemotherapy (47). Notably, the benefit with everolimus in HER2-negative breast cancer appears to be independent of PI3K pathway status (48), suggesting that this phenomenon is restricted to HER2-positive disease. Of the cell lines employed in our study, HCC1954 and SKBR3 are HR-negative, while BT474, EFM192A, and MDA-MB-361 are HR-positive. All but SKBR3 (PTEN loss) harbor a pathogenic PIK3CA mutation, although only HCC1954 and MDA-MB-361 occur in bona fide gain-of-function hotspots (p.H1047R and p.E545K, respectively). Interestingly, the effect of adding everolimus was highest in these two cell lines. In contrast, we found that the combination failed to increase the cell viability effects compared with T-DM1 or everolimus alone in BT474 and SKBR3 cells. As they differ in HR status, we cannot suggest that this drives resistance to the combination. In our experiments, BT474 cells appeared to be less sensitive to T-DM1 both by lacking optimal lysosomal processing as well as by having intrinsic resistance to the DM1 moiety. Interestingly, we found that BT474 express significantly lower SLC46A3 than HCC1954, which has been reported to regulate the transport of insoluble linker-bound DM1 from the lysosome to the cytosol (16, 38). Importantly, no mutation has been reported in this gene in any of the cell lines used in our study. Finally, we also observed differences in sensitivity to everolimus between HCC1954 and PDX118 cells, both in vitro and in vivo. Notably, the subcutaneous HCC1954 model was less sensitive to everolimus alone than the orthotopic one. Of note, we observed that HER2 remained highly expressed in HCC1954 models, both in vitro and in vivo. Altogether, our findings highlight that disease biology, beyond HR status, must be taken into consideration to select patients for therapy with T-DM1 and mTOR inhibitors.

Two recent preclinical studies, one conduced in HER2-positive breast- and another in HER2-positive gastric cancer, found that T-DM1 strongly inhibits mTOR and induces autophagy (30, 49). Interestingly, they reported discordant results in antitumor activity with T-DM1 and the early-stage autophagy inhibitor LY294002. Zhang and colleagues found an increased antitumor activity of T-DM1 plus LY294002 versus T-DM1 alone in subcutaneous NCI-N87 gastric cancer xenografts, concluding that LY294002 abrogates cytoprotective autophagy caused by T-DM1 (49). In contrast, Liu and colleagues reported that LY294002 reverts T-DM1 antitumor activity in BT474 and SKBR3 cells in vitro, hypothesizing that the autophagy induced by T-DM1 facilitates apoptosis (30). Findings from these two studies stand in contrast with our results since we did not find any significant effect of T-DM1 on mTOR downstream signaling or autophagy, despite observing consistent antitumor activity across several models. Of note, both studies assessed the effects of T-DM1 on mTOR and autophagy using T-DM1 concentrations more than 10-fold higher than the ones we employed in our experiments. Indeed, no alterations of mTOR downstream signaling or in LC3B-II/I ratio were observed by Liu and colleagues when they used T-DM1 concentrations similar to ours (30). Taking these findings into account, we postulate that i) the effects of T-DM1 on mTOR signaling and autophagy may be concentration-dependent, and ii) biological differences between HER2-positive gastric- and breast cancer may dictate the different fate of cells treated with T-DM1 and autophagy inhibitors. Of note, and despite chloroquine being a well-known autophagy inhibitor, we used it in our experiments to highlight role of lysosomal processing in T-DM1 activity. Importantly, together with the data reported by Liu and colleagues (30), our results do not support the use of T-DM1 in combination with autophagy inhibitors in HER2-positive breast cancer.

To summarize, our results support that the combination of T-DM1 with mTOR inhibitors is active in multiple HER2-positive breast cancer models. This effect might be related, at least partially, to mTOR-dependent lysosomal processing of T-DM1. Our findings might be applicable to other ADCs such as trastuzumab deruxtecan, which has recently proven highly effective in patients with HER2-positive breast- (5) and gastric cancer (50). In fact, the superiority of trastuzumab deruxtecan compared with T-DM1 will shift the clinical use of T-DM1 to later lines in HER2-positive advanced breast cancer. Hence, T-DM1 and mTOR inhibitors combinations might be of importance in this new clinical setting.

D. Casadevall reports grants from Instituto Carlos III (ISCIII; Spanish Healthcare Ministry) during the conduct of the study; nonfinancial support from Roche; and non-financial support from Pfizer outside the submitted work. A. Lluch reports grants from Roche during the conduct of the study; grants and personal fees from Novartis, Pfizer, Roche/Genentech, Eisai, Celgene; grants from Roche Pharma AG, AstraZeneca, Merck, PharmaMar, Boehringer Ingelheim, Amgen, GlaxoSmithKline; and grants from Pierre Fabre outside the submitted work. F. Rojo reports personal fees from Roche, AstraZeneca, BMS, MSD, Novartis, Eli Lilly and Company; and personal fees from GSK outside the submitted work. J. Albanell reports personal fees from Roche, Daichii Sankyo–AstraZeneca; and personal fees from Seagen outside the submitted work; in addition, J. Albanell has a patent for EGFRmut licensed, and a patent for LCOR issued; and J. Albanell has received consulting or advisory role fees, research funding, or grant travel and accommodation support from Roche, Pfizer, Seattle Genetics, AstraZeneca, Amgen, MSD, Eli Lilly and Company, and Daiichi Sankyo. No disclosures were reported by the other authors.

D. Casadevall: Conceptualization, data curation, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. A. Hernández-Prat: Data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. S. García-Alonso: Formal analysis, investigation, visualization, methodology, writing–review and editing. O. Arpí-Llucià: Formal analysis, investigation, methodology, writing–review and editing. S. Menéndez: Resources, validation, methodology. M. Qin: Formal analysis, methodology, writing–review and editing. C. Guardia: Validation, investigation, methodology, writing–review and editing. B. Morancho: Resources. F.J. Sánchez-Martín: Formal analysis, validation, methodology, writing–review and editing. S. Zazo: Validation, methodology, writing–review and editing. E. Gavilán: Conceptualization, methodology, writing–review and editing. M.A. Sabbaghi: Supervision, investigation, writing–review and editing. P. Eroles: Methodology, writing–review and editing. J.M. Cejalvo: Writing–review and editing. A. Lluch: Resources, funding acquisition, writing–review and editing. F. Rojo: Conceptualization, formal analysis, supervision, funding acquisition, methodology, writing–review and editing. A. Pandiella: Conceptualization, resources, supervision, funding acquisition, investigation, methodology, writing–review and editing. A. Rovira: Conceptualization, formal analysis, supervision, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. J. Albanell: Conceptualization, resources, supervision, funding acquisition, project administration, writing–review and editing.

The authors acknowledge the help and collaboration of the staff at the PRBB Animal Research Facility. This work was supported by ISCIII (CIBERONC CB16/12/00481, CB16/12/00241, PI18/00382, PI18/00006, PI18/01219), Generalitat de Catalunya (2017 SGR 507). MINECO through gBFU2015-71371-R grant and the CRIS Cancer Foundation supported work in AP lab. D. Casadevall was supported by ISCIII (Rio Hortega Research Contract CM16/00023 and Juan Rodés Research Contract JR18/00003). F.J. Sánchez-Martín and S. Menéndez were supported by Department de Salut Generalitat de Catalunya (PERIS SLT002/16/00008 and PERIS SLT006/17/00040). M. Qin received financial support from the China Scholarship Council (CSC) for her doctoral fellowship. Work carried out in our laboratories receives support from the European Community through the Regional Development Funding Program (FEDER).

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.