Combination of mTORC1/2 inhibitor vistusertib plus fulvestrant in vitro and in vivo targets oestrogen receptor-positive endocrine-resistant breast cancer

The following primary antibodies were used in this study for immunoblotting: pRB^ser780 (CST-3590), pRB^ser807 (CST-8516), total-RB (CST-9309), cyclin D1 (CST-2922), cyclin D3 (CST-2936), pAKT^ser473 (CST-9271), pAKT^Thr308 (CST-9275), total-AKT (CST-9272), pEGFR^Tyr1068 (CST-3777), total-EGFR (CST-2232), pERBB2^Tyr1248 (CST-2243), total-ERBB2 (CST-4290), pERBB3^Tyr1222 (CST-4784), pIGF1R^Tyr1135 (CST-3918), pS6K^Ser235/236 (CST-2211), total-S6K (CST-2217), Raptor (CST-2280), RheB (CST-13879), p4EBP1^Thr37/46 (CST-2855), 4EBP1 (CST-9452), pSIN1^Thr86 (CST-14716), SIN1 (CST-12860), pER^ser167 (CST-5587), Rictor (CST-2114) and Deptor (SCT-11816) were purchased from Cell Signaling Technology. p107 (sc-318), p130 (sc-317), total-ER (sc-8002, F-10), ERBB3 (sc-415) and IGF1R (sc-713) were purchased from Santa Cruz Biotechnology. β-tubulin (T-9026) were from Sigma-Aldrich and Ki67 from Clinisciences. The following antibodies were used for immunohistochemistry: pERK1/2^Thr202/4 (CST-4370), pAKT^ser473 (CST-4060), pS6K^Ser235/6 (CST-4858), pmTOR^Ser2448 (CST-2976) and p4EBP1^Thr37/46 (CST-2855) were purchased from Cell Signaling Technology. Ki67 was purchased from Clinisciences. Reagents were obtained from the following sources: 17-β-oestradiol (E2) and 4-hydroxytamoxifen (4-OHT) from Sigma-Aldrich, fulvestrant from Tocris, and neratinib and vistusertib from SelleckChem.

Human BC cell lines MCF7, SUM44, HCC1428 and T47D were obtained from the American Type Culture Collection, USA, and Asterand. All cell lines were banked in multiple aliquots to reduce the risk of phenotypic drift and identity confirmed using short tandem repeat (STR) analysis. Cells were routinely screened for mycoplasma contamination. They were maintained in phenol red-free RPMI1640 containing 10% fetal bovine serum (FBS) and 1 nM oestradiol (E2). Long-term oestrogen-derived (LTED) equivalents modelling relapse on an AI were generated, as reported previously [8], and were maintained in phenol red-free RPMI1640 containing 10% charcoal-dextran-stripped FBS (DCC). Tamoxifen-resistant (TAMR) MCF7 cells were generated by growing wild-type MCF7 long-term in the presence of RPMI1640 containing 10% DCC + 0.01 nM E2 + 100 nM 4-OHT. Fulvestrant-resistant (ICIR) MCF7 and MCF7 LTED cell lines were generated by growing parental cells long-term in the presence of RPMI1640 containing 10% DCC + 1 nM E2 + 100 nM fulvestrant or RPMI1640 containing 10% DCC + 100 nM fulvestrant, respectively. Palbociclib-resistant (PalboR) cell lines were generated and maintained, as previously described [9, 10]. All cell lines were stripped of steroids for 48–72 h prior to the start of experiments.

Cells were seeded into 96-well tissue culture plates and allowed to attach overnight. Monolayers were then treated with increasing concentrations of the drugs, and after 72 h, cell viability was determined using the CellTiter-Glo® Luminescent Cell Viability Assay (Promega), according to the manufacturer’s protocol. Values were expressed as relative luminescence compared to the vehicle-treated control. Non-linear regression analysis was used to fit the curves, and IC₅₀ values were calculated using PRISM 7 software (GraphPad). To determine the nature of the interaction between vistusertib and fulvestrant, combination studies were performed by using Chou and Talalay’ s constant ratio combination design and quantified using CalcuSyn software (BIOSOFT, Cambridge, UK) [11]. The combination indices (CI) were obtained by using mutually non-exclusive Monte Carlo simulations. In this analysis, CI scores significantly lower than 1 were defined as synergistic, CI > 1 as antagonistic, and CI = 1 as additive.

All cells were grown in the presence of RPMI1640 containing 10% DCC for 3 days prior to seeding. They were seeded into dishes, allowed to attach overnight and treated with the appropriate drugs the following day. After 24-h treatment, total protein was extracted and immunoblotting carried out, as previously described [8].

mRNA from treated cells and from HBCx34 OvaR PDX models (n = 30 [12]) was extracted using RNeasy Mini Kit (Qiagen), quantified and reverse-transcribed with SuperScript III First-Strand Synthesis System (Invitrogen). TaqMan gene expression assays (Applied Biosystems) were used to quantify TFF1 (Hs00907239_m1), PGR (Hs01556702_m1), GREB1 (Hs00536409_m1), PDZK1 (Hs00275727-m1) and ESR1 (Hs01046818_m1), EGFR (Hs01076090_m1), ERBB2 (Hs01001580_m1), ERBB3 (Hs00176538_m1), IFG1R (Hs00609566_m1) and/or IRS1 (Hs00178563_m1) together with FKBP15 (Hs00391480_m1) as a housekeeping gene to normalise the data. The relative quantity was determined using ΔΔCt, according to the manufacturer’s instructions (Applied Biosystems).

HBCx22 OvaR and HBCx34 OvaR PDX models resistant to endocrine therapy were established as stated previously [12], in accordance with the French Ethical Committee. Efficacy studies were carried out to determine the anti-tumour activity of vistusertib alone and combined with fulvestrant administered over 90 days. The treatment groups (10–12 mice per arm) received either vistusertib (15 mg/kg daily by oral gavage) or fulvestrant (5 mg/mouse suspended in corn oil by weekly subcutaneous injection into the flank). These concentrations are in keeping with previous studies [6] and clinical achievable doses [13] for vistusertib. For the combination group, fulvestrant was dosed 2 h before administration of vistusertib. The control groups received both vehicles. To assess whether treatment with vistusertib alone or in combination with fulvestrant could further delay tumour progression, five mice from each group were followed for an additional 40 days after drug withdrawal.

Tumour diameters were measured using calipers, and volumes were calculated as V = a × b²/2, where ‘a’ is the largest diameter and ‘b’ is the smallest. Percent change in tumour volume was calculated for each tumour as (Vf − V0/V0) × 100, where V0 is the initial volume (at the beginning of treatment) and Vf is the final volume (at the end of treatment). Tumour regression (R) was defined as a decrease in tumour volume of at least 50%, taking as reference the baseline tumour volume [14].

In order to assess biomarker changes, a pharmacodynamic study was performed for 4 days of treatment with vistusertib, fulvestrant or a combination of the two drugs in the HBCx22 OvaR PDX model. Mice were sacrificed at 4 h after the final treatment and tumours resected. Excised tumours were fixed in 10% neutral buffered formalin and paraffin-embedded, and tissue microarrays (TMA) were built from the blocks. Three xenografts from each treatment group and two tissue cores per tumour were included in the TMA. Sections from the TMA were cut and stained for the expression of biomarkers, as previously described [12]. The immunohistochemically stained TMA sections were digitally scanned at ×20 with a Hamamatsu NanoZoomer-XR whole-slide scanner (Hamamatsu Photonics K.K., Hamamatsu, Japan). The quality of the images was checked manually, and the images were analysed with Visiopharm integrator system (VIS) version 2018.9.3.5303 (Visiopharm A/S) using VIS ready to use automated image analysis algorithms (APPs).

Excised tumours from HBCx34 OvaR PDX-sacrificed mice were used for a gene expression study (n = 12; 3 mice by group). Libraries were created after using TruSeq Stranded mRNA Library Prep Kit (Illumina) and sequenced using the NextSeq 500 (Illumina). RNA-seq data was aligned to the human GRCh38 reference genome using STAR Aligner (star v2.6.1a) [15]; read count for each gene was calculated with HTSeq (v0.6.1) [16]. Genes were compared for differential expression between the different treatments using edgeR [17] and were considered to be statistically expressed when the absolute fold change ≥2 and false discovery rate (FDR) <5%. These significantly expressed gene lists were subject to further functional annotation using Ingenuity Pathway Analysis (IPA) to identify altered pathways due to the corresponding treatments. For individual pathways, the Benjamini–Hochberg procedure was used to the calculate FDR in order to adjust for multiple testing. RNA-seq data supporting the findings was deposited in the NCBI (http://​ncbi.​nlm.​nih.​gov/​geo/​) with reference PRJNA564917.

We tested the anti-proliferative effect of vistusertib in a panel of isogenic cell lines modelling sensitivity or resistance to endocrine therapy (MCF7, SUM44, HCC1428 and T47D) for which the PIK3CA, PTEN and ESR1 mutation status was previously established [18, 19]. Assays were conducted in the presence of E2, to model the effects of vistusertib as a monotherapy, or in the absence of E2, to model the combination with an AI in the primary setting. MCF7 cells showed a concentration-dependent decrease in proliferation in the presence of E2 with an IC₅₀ of 20 nM. In the absence of E2, minimal further anti-proliferative effect was evident from the addition of vistusertib and the IC₅₀ was increased (Fig. 1a, Additional file 1: Table S1a). In an extended panel of ER+ cell lines, in the presence of E2, vistusertib sensitivity varied with IC₅₀ values between 30 and 500 nM (Additional file 2: Figure S1a and Additional file 1: Table S1a). Removal of E2 caused a drop in proliferation in all cell lines, as expected. Addition of vistusertib further reduced cell viability in a dose-dependent manner (IC₅₀ values between 40 and 700 nM; Additional file 2: Figure S1a and Additional file 1: Table S1a). In order to assess the effect of vistusertib in cell lines modelling resistance to an AI, escalating concentrations were tested in two MCF7 LTED models in the presence or absence of E2. Of note, the MCF7 LTED^Y537C, which harbour a hotspot ESR1 mutation in the ligand-binding domain, showed sensitivity with an IC₅₀ of 50 nM in the presence or absence of E2, in keeping with their ligand-independent phenotype (Fig. 1b). Contrastingly, MCF7 LTED^wt showed a slightly higher IC₅₀ (75 nM) (Fig. 1c). Three further LTED cell lines were assessed. HCC1428 LTED expressing wild-type (wt) ESR1, SUM44 LTED harbouring ESR1^Y537S and T47D LTED which lose ER expression showed varying IC₅₀ values between 65 and 350 nM (Additional file 2: Figure S1b and Additional file 1: Table S1a).

We further assessed sensitivity to vistusertib in cell lines modelling resistance to tamoxifen (TAMR) or fulvestrant (ICIR). In keeping with the previous data, both models showed a concentration-dependent decrease in proliferation with IC₅₀ values of 85 and 50 nM, respectively (Fig. 1d, e and Additional file 1: Table S1b). Finally, we assessed the effect of escalating doses of fulvestrant in both the presence and the absence of a fixed concentration of vistusertib in MCF7 LTED^wt and MCF7 LTED^Y537C cell lines (Fig. 1f, g and Additional file 1: Table S1c). In both cell line models, the combination with vistusertib appeared synergistic with a combination index below 1.

Previous studies have shown that blockade of mTORC1 can lead to feedback loops via IGFR and ERBB signalling networks [20, 21] (Fig. 2a). In order to test the effect of targeting both mTORC1 and mTORC2, we examined the effect of vistusertib upon key protein targets within the mTOR pathway. Immunoblot analysis of the MCF7 and LTED derivatives was assessed (Fig. 2b). Vistusertib caused a decrease in the expression of pS6RP^Ser235/6, p4EBP1^Thr37/46 and pAKT^Ser473 and an increase in Deptor and pSin1 together with a decrease in abundance of Cyclin D1, D3 and pRB indicative of cell cycle arrest. Treatment with fulvestrant alone or in combination with vistusertib reduced abundance of both phosphorylated and total ER. Despite the dual blockade of mTORC1/2, feedback loops via IGF1R and ERBB family members were evident but appeared cell line-specific. For instance, MCF7 LTED^wt showed marked increases in pIGF1R and pAKT^Thr308 in response to vistusertib. To test if the effect of vistusertib was persistent beyond a 24-h period, we performed a time course experiment and showed a gradual increase in abundance of pEGRF, pIGF1R and pSin1 markers up to 96 h of treatment (Additional file 3: Figure S2).

Evidence suggests that cross-talk between PI3K/AKT/mTOR impacts on ER function as a transcription factor. Indeed, mTORC1 via S6RP has been shown to phosphorylate ER at serine 167 [22]. We therefore assessed the effects of vistusertib on ER-mediated transcription. The relative expression of a panel of oestrogen-regulated genes (ERGs: TFF1, PGR, GREB1 and PDZK1) was evaluated in the presence or absence of E2. In MCF7 and in both MCF7 LTED derivatives, treatment with vistusertib under DCC conditions caused subtle or no changes in the expression of ERGs that was gene- and cell-specific (Additional file 4: Figure S3). Similarly, in the presence of 0.01 nM of E2, vistusertib caused small changes in the expression of the ERGs for all the three cell lines tested, but fulvestrant alone or in combination with vistusertib consistently reduced the expression of all the ERGs when compared with the vehicle control (Fig. 3). These data suggest that vistusertib does not impact ER-mediated transcription.

In order to assess the effect of vistusertib alone or in combination with fulvestrant in vivo, we adopted two PDX models of acquired endocrine-resistant BC. HBCx34 OvaR is an ER+ PDX which is resistant to E-deprivation and tamoxifen but sensitive to the anti-proliferative effects of fulvestrant [12] (Fig. 4). After a period of 64 days, all treatments showed over a 95% reduction in tumour volume (fulvestrant: 97.6%, p = 0.004; vistusertib: 96.2%, p < 0.0001; combination: 99.7%, p < 0.0001) compared to vehicle control (Fig. 4a and Additional file 5: Figure S4). Vistusertib showed greater efficacy than fulvestrant as a monotherapy over the first 50 days (adjusted p value = 0.005) and appeared similar to the combination over this time period. At the end of treatments, all xenografts were in regression or complete response in the combination arm (percentage of tumour volume change ≤50%), against four xenografts in the fulvestrant-treated group (Fig. 4a).

Analysis of the combination of vistusertib and fulvestrant appeared the most effective, showing a significant increase in efficacy compared to fulvestrant alone (p = 0.0001, Mann–Whitney test, Fig. 4a).

In order to further explore the impact of vistusertib alone or in combination with fulvestrant, tumours were resected at the end of the study and subjected to RNA-seq. Fulvestrant showed the greatest impact on gene expression (1456 upregulated and 1077 downregulated genes) versus vistusertib (291 upregulated and 174 downregulated genes) when compared with vehicle control (Fig. 4b). Noteworthy, the number of gene changes as a result of the combination largely reflected that seen for fulvestrant (1717 upregulated and 1412 downregulated genes) indicating the mitogenic driver within this PDX remains ER. In order to identify canonical pathways affected by these treatments, we conducted ingenuity pathway analysis (IPA; FDR <5%) using differentially expressed genes (FDR <5% and fold change ≥2; Additional file 6: File S1). Fulvestrant showed a dominant effect on cell cycle and oestrogen-mediated S-phase entry both as a monotherapy or in combination with vistusertib. Contrastingly, single-agent vistusertib showed no impact on ER-mediated S-phase entry. Treatment with vistusertib showed minimal although significant enrichment of EGF, ERBB and ERK/MAPK signalling compared with vehicle control (Additional file 6: File S1). In order to explore this further, we carried out targeted qRT-PCR (Fig. 4c). Treatment with fulvestrant significantly reduced the expression of TFF1, PGR, GREB1 and IRS1 but increased the expression of EGFR, ERBB2 and ERBB3. Contrastingly, vistusertib had minimal effect on the expression of ESR1, GREB1 and PGR; however, it significantly reduced TFF1 but not to the degree seen with fulvestrant or the combination. Noteworthy, vistusertib significantly increased the expression of EGFR but not ERBB2, ERBB3 or IGF1R.

In order to further explore the efficacy of the combination of vistusertib with fulvestrant, a second PDX model, HBCx22 OvaR, was assessed. HBCx22 OvaR is an ER+ model showing partial resistance to fulvestrant and harbours a 24-base-pair in-frame deletion in exome 13 in PIK3R1 [12] (Fig. 5). As expected, single-agent fulvestrant had no significant impact on tumour progression compared to vehicle control, confirming the resistant phenotype. Vistusertib as a monotherapy delayed tumour progression by 54.5% (p = 0.04) compared to vehicle control. The combination of vistusertib plus fulvestrant was the most effective treatment with tumour volumes 84.7% lower than vehicle control (p = 0.0002) (Fig. 5a). After 93 days of treatment, the therapies were withdrawn and the tumour volumes assessed for a further 40 days in order to establish the efficacy of the drugs in delaying tumour progression (Fig. 5b). Removal of therapies showed sustained anti-tumour effect in the combination group, whilst tumours treated with vistusertib alone showed significant progression.

In order to assess dynamic changes, three mice per arm were sacrificed after 4 days of therapy and tissue sections were subjected to immunohistochemical analysis. Treatment with vistusertib or vistusertib in combination with fulvestrant revealed suppression of pAKT^Ser473, p4EBP1^Thr37/46 and pS6RP^Ser235/6, as well as a slight but noticeable decrease in pmTOR^Ser2448 (Fig. 5c and Additional file 7: Figure S5a). Furthermore, fulvestrant reduced the expression of pERK1/2^Thr202/4 both alone and in combination with vistusertib. In contrast to our in vitro analysis, no alteration in the abundance of pEGFR and pIGFR was evident in response to vistusertib alone, whilst pEGFR was significantly suppressed by the combination with fulvestrant (Additional file 7: Figure S5b). Noteworthy, assessment of Ki67 showed the greatest reduction when the combination of vistusertib and fulvestrant was used (Additional file 7: Figure S5a).

Taken together, these data suggest the combination may provide greater efficacy than fulvestrant alone in ER+ acquired endocrine-resistant disease.

As increased feedback loops via ERBB and IGF1R family members were evident in vitro and from our gene expression analysis, we assessed sensitivity of MCF7-LTED^wt cell lines to the anti-proliferative effect of vistusertib, or fulvestrant combined with the pan-ERBB inhibitor neratinib, or the combination of all three agents (Fig. 6a). Fulvestrant and neratinib enhanced the anti-proliferative effect of vistusertib; however, the triple combination was the most effective. These data further support previous observations in which the triple combination targeting three cellular nodes: ERBB, ER and mTORC1, showed the greatest anti-proliferative effect [20].

More recently, CDK4/6 inhibitors have become the standard of care in the treatment of endocrine-resistant ER+ BC. Despite their efficacy, not all patients benefit and many will eventually relapse with acquired resistance. Studies suggest that cross-talk exists between CDK4 and the mTOR pathway via pTSC2 [23] and that blockade of mTORC1/2 may delay the onset of resistance to CDK4/6 inhibition [24]. To assess this, we treated three palbociclib-resistant cell line models (MCF7-^PalboR, MCF7 LTED-^PalboR and T47D-^PalboR) (Fig. 6b) with escalating concentrations of vistusertib with or without fulvestrant. All three cell lines showed sensitivity to mTORC1/2 blockade. The addition of fulvestrant further enhanced the anti-proliferative effect. Taken together, these data suggest mTORC1/2 blockade remains effective after acquisition of resistance to palbociclib.