Cutting Edge: mTORC1 Inhibition in Metastatic Breast Cancer Patients Negatively Affects Peripheral NK Cell Maturation and Number

Natural killer cells are innate lymphoid cells with cytotoxic and cytokine secretion capacities. These functions and the fact that they are circulating cells endow them with antimetastatic activity (1–3). NK cells discriminate normal and tumor cells thanks to activating and inhibitory receptors. Healthy cells express an excess of ligands for inhibitory receptors that keep NK cell effector functions at bay. In contrast, tumor cells are defective in inhibitory ligands and express activating ligands stimulating NK cell functions. The NK cell number is also a key factor for efficient NK cell responses; reduced circulating NK cell count results in poor clinical outcome in human lymphoma (4) and increases metastatic spread (1, 3, 5). Spontaneous cytotoxic activity toward target cells is restricted to mature NK cells both in mouse (6, 7) and in human (8). Their absence impairs metastatic clearance in mouse (9, 10).

The kinase mechanistic target of rapamycin (mTOR) is the catalytic subunit of two complexes: mTORC1 and 2 (11). mTORC1 is targeted by rapamycin and its analogues, the rapalogs, whereas mTORC2 is not in most situations. mTORC1 integrates extracellular cues to control in return cell metabolism (11). This property explains why rapalogs are approved as antiproliferative drugs in the course of chemotherapies. Indeed, cancer cells frequently activate the PI3K/AKT/mTOR pathway to sustain their metabolism. In particular, mTOR expression is associated with worse prognosis and endocrine resistance in breast cancer (BC) (12). Therefore, everolimus, a rapalog, is used for the treatment of hormone receptor (HR)–positive, HER2-negative metastatic BC in the case of endocrine resistance (13). At the same time, preclinical and clinical evidence suggest that BC is under immunosurveillance (14). In human, the size of the immune infiltrate containing NK cells can be positively correlated to prognosis (15), and the progressive loss of NK cell function is associated with BC progression (16). In mouse models, mTOR activity is required for NK cell development and activation (17, 18). Furthermore, specific ablation of mTORC1 in murine NK cells results in crippled antimetastatic activity (19). Finally, in vitro treatment of human NK cells with everolimus decreases effector functions (L. Besson, Y. Rocca, T. Walzer, and A. Marçais, unpublished data). Based on these data, everolimus may have an immunosuppressant effect susceptible to dampen the NK cell antitumoral response. To formally test this hypothesis in human in vivo and evaluate the impact of mTORC1 inhibition on NK cell phenotype and function, we set-up a prospective study among women with HR⁺ HER2/Neu–negative metastatic BC treated with everolimus. Indeed, prospective studies in patients are invaluable to provide information on cell physiology in a natural environment. In addition, the research question, readouts, and calculation of statistical power are developed prior to data collection, assuring accurate and standardized recording of relevant data.

Patients were recruited in the RAPANK clinical trial (NCT02536625), a prospective, multicenter cohort study undertaken at two centers in France. All patients provided written informed consent in accordance with the Declaration of Helsinki. The protocol was reviewed by a central ethics committee.

Eligible patients were women >18 y of age with histologically confirmed HR⁺ HER2/neu–negative recurrent metastatic BC that were scheduled for everolimus treatment in combination with hormonotherapy. Patients required an adequate performance status (Eastern Cooperative Oncology Group Scale of Performance Status of 0–2) and a body mass index less than 30kg/m². The disease had to be measurable according to Response Evaluation Criteria In Solid Tumors (version 1.1). Adverse events were graded according to the National Cancer Institute Common Terminology Criteria of Adverse Events version 4.0. Patients with autoimmune and inflammatory diseases as well as immune dysfunctions were not enrolled. Radiation therapy to bone or vaccination could not be performed throughout the study period.

Ten milligrams of everolimus were administered orally in combination with hormonotherapy, both once daily continuously. In case of severe toxicity, dose reductions were allowed. Patients were evaluated at baseline and then every 3 mo of the trial until the end of the study by physical examination.

Blood samples collected on EDTA were stained for phenotypic markers, and PBMCs were isolated using ficoll centrifugation for NK cell stimulation. Cytofix/Cytoperm was used for cytokine staining, Lyse/Fix Buffer and Perm Buffer III were used for phosphoprotein staining (BD Biosciences). Sample analysis was performed on a Navios 3L (Beckman Coulter), and data were analyzed using FlowJo 10.5.0 (BD Biosciences).

PBMC were cultured for 4 h at a 1:1 ratio with K562 cells or Granta B cell lymphoma cells (American Type Culture Collection) with BD GolgiStop (BD Biosciences). Granta cells were incubated with 10µg/ml anti-CD20 Ab (rituximab; Hoffman-Laroche) for 30 min at 4°C before stimulation. Cells were cultured with or without recombinant human IL-15 (100ng/ml; PeproTech) for 1 h at 37°C for phosphorylation analysis.

Upon thawing, PBMCs were rested overnight prior a 4-h coculture at different E:T ratios with K562 NanoLuc⁺ cells (20). Bioluminescence was measured in the supernatant using a Tecan Instrument luminometer after addition of Hikarazin (21).

Inclusion of 60 patients was sufficient to obtain a precision of at least 12% around our 30% estimates for the variation of granzyme B (GZMB) at 3 mo (M3), with a 95% confidence interval and an early retirement of 30% of patients. Qualitative data were described by their frequency and percentage; continuous variables were described as median (minimum–maximum). A Wilcoxon signed-rank test was used to test the primary objective of the study (i.e., the change of GZMB level in peripheral NK cells between baseline and the third month of treatment). Linear mixed effect models with subject random intercept and fixed effect of time were used to analyze change over time. Box plots showing median, 10th, 25th, 75^th, and 90th percentiles, as well as outliers, have been used to graphically describe the data. Exact p values obtained in the linear mixed models are given in the figures. Statistical analyses were performed using SAS software version 9.4 (SAS Institute, Cary, NC).

Sixty patients (Table I) withHR⁺ metastatic BC were included and treated with everolimus following written informed consent (NCT02536625). All patients had previously received endocrine therapy, and 38 patients had received at least one line of chemotherapy before inclusion. The median lymphocyte count at inclusion was 1.54 G/l. The median follow-up was 32.5 mo. Over this period, 17 patients had at least one treatment interruption. Thirty-nine patients (70%) discontinued study treatment because of progression, whereas 16 patients (29%) discontinued everolimus because of toxicity. Seven patients (13.5%) achieved an objective response. Twenty-three patients (40.4%) presented infectious events, with bacterial infection for 19 patients (33.3%) and viral infection for 4 patients (7%); specific treatment was required for 16 patients (28%). Five patients (8.3%) experienced at least a grade ≥3adverse events according to the National Cancer Institute Common Terminology Criteria of Adverse Events V4.0, notably with stomatitis (two patients, 3%) and skin rash (one patient). The primary objective of the study was the change of GZMB level in peripheral NK cells within the first 3 mo of everolimus treatment. The GZMB level was not significantly different between baseline and the third month of treatment (p = 0.48).

To test the impact of mTOR inhibition on peripheral NK cells, we analyzed blood samples from the metastatic BC patients included in this prospective trial before (baseline), 1 mo after the beginning of treatment (M1), and 3 mo after the beginning of treatment (M3) (51, 46, and 36 patients, respectively). Only samples of patients still treated with everolimus were analyzed. We first assessed the impact of everolimus on the activity of mTORC1, its primary target, in peripheral NK cells. Despite not being directly targeted by rapalogs, mTORC2 activity can be secondarily decreased by these inhibitors in some cell types (11). To measure the activity of both complexes, we quantified the phosphorylation level of the ribosomal phospho-S6 (pS6) for mTORC1 as well as the phosphorylation level of the kinase AKT on Ser⁴⁷³ (pAKT) for mTORC2. We also measured the phosphorylation of STAT5 (pSTAT5), induced by IL-15 stimulation independently of mTOR activation (17). NK cells were defined as CD56⁺CD7⁺CD3⁻CD8⁻CD14⁻CD19⁻; CD8⁺ and CD7⁻ NK cells were therefore excluded from our analysis (Supplemental Fig. 1). Phosphorylations were measured in NK cells after a 1-h culture with or without IL-15 stimulation. In both conditions, the pS6 level was significantly decreased in NK cells after the onset of the treatment (Fig. 1A). In contrast, mTORC2 activity was not altered (Fig. 1B). Regarding pSTAT5, we observed a significant decrease from the baseline level in the IL-15 condition (Fig. 1C). This could be related to mTORC1-dependent regulation of the IL-2/15Rβ expression level (17, 19). As expected, pS6 in T cells was also decreased at M3, underlying the fact that everolimus affects all cells indistinctively (Supplemental Fig. 2). Thus, everolimus negatively impacts mTORC1 activity in NK cells. This effect is not limited to NK cells and probably affects the immune system broadly.

Because everolimus negatively impacts mTORC1 activity in NK cells, we analyzed the effect of prolonged mTORC1 inhibition on NK cell biology. We first measured peripheral NK cell number and maturation. We observed a gradual decrease from baseline to M3 in the number of circulating NK cells (Fig. 2A). This decrease was restricted to the CD56^Dim compartment and reached 40% at M3, suggesting an impact of the treatment on NK cell maturation. To identify the affected maturation stages, we measured the expression of markers allowing identification of the final stages of NK cell differentiation. The expression of NKG2A and CD94 was preserved (Fig. 2B), but, the levels of CD57 and KLRG1, which mark the most mature NK cell population, significantly decreased in percentage and mean fluorescence intensity of positive cells after treatment (Fig. 2C and data not shown). These results demonstrate that mTOR activity controls the size of the NK cell pool, similar to its role in the mouse. Moreover, the fact that everolimus decreases NK cell maturation with a negative impact on the most mature CD56^Dim/CD57⁺/KLRG1⁺ NK cell population again translates in human to the fact that mTOR controls NK cell maturation in mouse models, its deficiency primarily affecting the most mature CD27^low NK cell subset (17, 19, 22). Given the antiproliferative effect of everolimus, this could be due to reduced homeostatic proliferation of the CD56^Bright subset coupled with a decreased differentiation rate. A decrease in mature NK cell representation results in increased metastatic spreading in mouse models (9, 10); the specific loss of mature NK cells described in this study could thus result in a significantly defective antimetastatic control.

We next measured the impact of the treatment on the expression of an array of activating/inhibitory receptors and metabolic readouts. Regarding activating receptors, the expression of NKp46 and NKG2D was upregulated, whereas that of CD16 was downregulated (Fig. 3A). The expression of the dual activating/inhibitory receptors 2B4 and CD161 went down over time (Fig. 3B). No change was observed for KIR2DL1, 2DL2, or 3DL1 expression (Supplemental Fig. 3A). The fact that activating receptors, such as NKp46 and NKG2D, the expression of which decreases along with maturation, were upregulated in the overall NK cell population upon everolimus treatment, whereas receptors associated with maturation, such as CD16, 2B4, and CD161, were decreased, further confirming the maturation bias toward immaturity. mTORC1 being a metabolic regulator, we next tested whether everolimus treatment led to detectable changes in metabolism. No change was observed in the expression of the nutrient transporters CD71 or CD98 nor in morphologic parameters such as size and granularity (Supplemental Fig. 3B). Several explanations can be proposed to understand this result. First, rapalogs only partially inhibit mTORC1 (23). Everolimus treatment could thus have spared metabolic functions or at least the ones we assessed. Another possibility is that at the time everolimus treatment started, metabolic parameters were already lowered in NK cells consequentlyto tumor growth, which could prevent detection of any further decrease. In particular, immunomodulatory cytokines present in BC patients, such as TGF-β, (16) dampen NK cell metabolism in tumor contexts (24).

Reducing mTOR activity results in decreased effector capacities of NK cells in mouse models in vivo and in human in vitro culture (17, 22, 25, and L. Besson, Y. Rocca, T. Walzer, and A. Marçais, unpublished data). Whether in vivo mTORC1 inhibition in human results in a similar outcome in circulating NK cells is unknown. Everolimus treatment did not affect the expression of perforin-1 or GZMB (Fig. 4A). Production of IFN-γ, MIP1-β, TNF-α, and degranulation were then measured upon coculture of PBMC with K562 cells or Granta B cells coated with anti-CD20 Ab. As shown in Fig. 4B–D, cytokine production was not affected by the everolimus treatment. In contrast, degranulation was increased in response to K562 stimulation, whereas it was conserved in response to Granta cells (Fig. 4E, Supplemental Fig. 4A). We then measured NK cell cytotoxicity using a new technique adapted to low cell numbers (20). Everolimus treatment increased target cell killing (Supplemental Fig. 4B). Hence, despite the known negative effect of mTORC1 deficiency on NK cell functions in various models, the results presented above show that in the environment prevailing in metastatic BC patients, mTORC1 inhibition is not detrimental to NK cell functions. The fact that effector functions are preserved or even increased upon everolimus treatment could result from intrinsic differences between mice and human, from indirect effect of everolimus in vivo, or from physiological alterations due to metastatic BC. In this context, a beneficial effect of partial inhibition of the PI3K–AKT–mTOR pathway on T cell function has been described in murine tumor models (26). To our knowledge, this is the first study reporting such an effect in vivo in human.

Overall, despite the moderate biological effect of the everolimus treatment, we translate to human the role of mTOR in NK cell homeostasis. Yet in contrast to in vitro findings, mTOR inhibition is, rather, beneficial to effector functions of the remaining cells, demonstrating the importance of such prospective analysis in patients.

The authors have no financial conflicts of interest.