Decreased NK-cell tumour immunosurveillance consequent to JAK inhibition enhances metastasis in breast cancer models

Abstract

The JAK/STAT pathway is an attractive target for breast cancer therapy due to its frequent activation, and clinical trials evaluating JAK inhibitors (JAKi) in advanced breast cancer are ongoing. Using patient biopsies and preclinical models of breast cancer, we demonstrate that the JAK/STAT pathway is active in metastasis. Unexpectedly, blocking the pathway with JAKi enhances the metastatic burden in experimental and orthotopic models of breast cancer metastasis. We demonstrate that this prometastatic effect is due to the immunosuppressive activity of JAKi with ensuing impairment of NK-cell-mediated anti-tumour immunity. Furthermore, we show that immunostimulation with IL-15 overcomes the enhancing effect of JAKi on metastasis formation. Our findings highlight the importance of evaluating the effect of targeted therapy on the tumour environment. The impact of JAKi on NK cells and the potential value of immunostimulators to overcome the weakened tumour immunosurveillance, are worthwhile considering in the clinical setting of breast cancer.

Figure 1. The JAK/STAT pathway is active in breast cancer bone metastasis.

(a,b) pSTAT3 and pSTAT5 analysis by immunohistochemistry (IHC) in primary tumours and paired bone marrow metastases from 14 breast cancer patients. (a) Images of pSTAT3 and pSTAT5 (patient number 13) in the primary tumour and the paired bone metastasis are shown. Scale bar, 20 μm. (b) Quantification of pSTAT3 and pSTAT5 immunoreactive score (IRS) in primary tumours and paired bone marrow metastases. The boxplots show the median, the 25th and 75th percentiles and the minimum and maximum values after outlier removal. n=14 patients. pSTAT3 and pSTAT5 IRS score of primary tumours and bone metastases did not show significant differences (NS) using the exact Wilcoxon's rank-sum test. (c,d) JAK/STAT pathway activation in EO771 and 1833 breast cancer models. (c) Representative images of pSTAT3 staining in EO771 and 1833 primary mammary tumours and bone metastases from mice treated with vehicle or ruxolitinib (90 mg kg⁻¹ BID). T, tumour area. Scale bar, 100 μm. (d) Quantification of percentage of positive pSTAT3 tumour cells in primary tumours and bone metastases from EO771 and 1833 tumour models from mice treated with vehicle or ruxolitinib (90 mg kg⁻¹ BID). The percentage of positive pSTAT3 tumour cells from individual mice and the mean±s.e.m. are shown. For EO771 primary tumours vehicle=4, ruxolitinib=5; bone metastases vehicle=6, ruxolitinib=9 from two independent experiments. For 1833 primary tumours vehicle=4, ruxolitinib=5; bone metastases vehicle=4, ruxolitinib=5. *P<0.05, ***P<0.005, NS, not significant with two-tailed Mann–Whitney test.

Figure 2. JAKi enhances breast cancer bone metastasis.

(a) Left: scheme of the protocols I and II. Right:quantification of bone marrow metastases in mice intracardially (IC) injected with 0.5 × 10⁶ EO771 cells and treated with vehicle or ruxolitinib (90 mg kg⁻¹ BID). Treatment was started 3 days after tumour cell injection (rux I) or 2 days before tumour cell injection (rux II) and continued for 10 days. Bars show mean±s.e.m. of bioluminescence (total flux (p s⁻¹)) quantified in long bones positive for metastases. Representative images of mice treated for 10 days with vehicle or ruxolitinib are shown. Bone number analysed at day 10: vehicle n=10, ruxolitinib I n=14, ruxolitinib II n=15. (b) 1.3 × 10⁵ 1833 cells were injected into athymic nude mice pretreated for 2 days with ruxolitinib (90 mg kg⁻¹ BID), treatment continued for 14 days and metastases were quantified in the long bones. The graph shows mean±s.e.m. of bioluminescence (total flux (p s⁻¹)) in long bones positive for metastasis at different time points. Bone number analysed: vehicle n=10–12, ruxolitinib n=12–14. (c) 1.3 × 10⁵ 1833 cells were injected into athymic nude mice. After 6 days, treatment with vehicle or ruxolitinib (90 mg kg⁻¹ BID) started and continued for 2 weeks. The graph shows mean±s.e.m. of bioluminescence (Total Flux (p s⁻¹)) in long bones positive for metastasis at different time points. Bone number analysed: vehicle n=13–14, ruxolitinib n=10-11. *P<0.05, **P<0.01 with two-tailed Mann–Whitney test.

Figure 3. JAKi enhances metastasis dissemination by impairing anti-tumour immunity.

(a) Scheme of the experimental procedure: 1x10⁵ 4T1.2 tumour cells were injected intramammary (IM), into the mammary fat pad of Balb/c mice and animals were treated for 3 weeks with vehicle or ruxolitinib (90 mg kg⁻¹ BID). Representative images of H&E used for lung metastasis quantification are shown. (b) Quantification of primary mammary tumour volume (mm³). Data are expressed as the mean±s.e.m. Vehicle n=10, ruxolitinib n=9. (c) Quantification of spontaneous bone metastases. Bars represent the mean±s.e.m of percentage of tumour cells (CD45⁻ TR119⁻ EPCAM⁺) in the whole bone quantified by FACS. Vehicle n=7, ruxolitinib n=6. (d) Quantification of spontaneous lung metastases. Bars represent the mean±s.e.m of the metastatic index (area of metastasis/area of lung). Vehicle n=7, ruxolitinib n=6. (e) Quantification of circulating tumour cells. Bars represent mean±s.e.m of the colony number per ml of blood. Vehicle n=7, ruxolitinib n=5. (f–h) 1x10⁵ 4T1.2 tumour cells were injected IM in NSG mice and animals were treated for 16 days with ruxolitinib (90 mg kg⁻¹ BID). (f) Quantification of primary mammary tumour volume (mm³). The tumour volume in individual mice and the mean are shown. Vehicle n=7, ruxolitinib n=5. (g) Quantification of spontaneous bone metastases. The bone metastatic index (number of colonies per primary tumour volume) from individual mice and the mean are shown. Vehicle n=5, ruxolitinib n=5. (h) Quantification of spontaneous lung metastases. The metastatic index (number of nodules in the big lobe/primary tumour volume) from individual mice and the mean are shown. Vehicle n=7, ruxolitinib n=5. *P<0.05, **P<0.01 with two tailed Mann–Whitney test.

Figure 4. JAKi enhance metastasis by decreasing the NK-cell population.

(a) Quantification of total spleen cell number in EO771 and 4T1.2 tumour-bearing mice treated with vehicle or ruxolitinib (90 mg kg⁻¹ BID). Bars show mean±s.e.m. For EO771: vehicle n=7, ruxolitinib n=8; for 4T1.2: vehicle=7, ruxolitinib=6. (b) Left: representative images of NK-cell-gating for FACS analysis of EO771 spleens. Right: quantification of NK cells (CD3⁻ CD49b⁺) in spleens from EO771 tumour-bearing mice and of NK cells (CD3⁻ CD49b⁺NKp46⁺) in spleens from 4T1.2 tumour-bearing mice, after treatment with vehicle or ruxolitinib (90 mg kg⁻¹ BID). Bars show the mean±s.e.m. For EO771: vehicle n=7, ruxolitinib n=8; for 4T1.2: vehicle=7, ruxolitinib=6. (c) Quantification of NK cells (CD3⁻ CD49b⁺) in the bone marrow of EO771 tumour-bearing mice treated with vehicle or ruxolitinib (90 mg kg⁻¹ BID). Bars show the mean±s.e.m. Vehicle n=7, ruxolitinib n=8. (d) Quantification of NK cells (CD3⁻ CD19⁻ CD49b⁺ NKp46⁺) in the peripheral blood of EO771 tumour-bearing mice treated with vehicle or ruxolitinib (90 mg kg⁻¹ BID). Bars show the mean±s.e.m. Vehicle n=6, ruxolitinib n=7. FACS analyses were performed on tissue samples collected from mice treated as described in Fig. 2a (protocol II) for the EO771 model and as in Fig. 3a for the 4T1.2 model. (e) Quantification of lung metastasis in Rag2^−/−γc^−/− mice that were intravenous (i.v.)-injected with 2x10⁵ EO771 cells. Mice were subjected, or not (control), to adoptive transfer of NK cells. The number of metastases in the big lobe from individual mice, and the mean±s.e.m. are shown. Control n=4, with NK cells n=5 mice.

Figure 5. JAKi decrease NK-cell maturation and activation.

(a) Splenic NK-cell maturation analysis from tumour-free mice after treatment with vehicle or ruxolitinib (90 mg kg⁻¹ BID) for 6 days. CD3⁻ NK1.1⁺ NK cells were analysed for CD27 and CD11b expression. Bars show the mean±s.e.m. Representative images of FACS gating are shown. Vehicle n=5, ruxolitinib n=5. **P<0.01. NS, not significant with two-tailed Mann–Whitney test. (b,c) Splenic NK cells (CD3⁻ NK1.1⁺) from C57BL/6 mice treated for 6 days with ruxolitinib (90 mg kg⁻¹ BID) or vehicle, were stained for DNAM1 and KLRG1 receptors. Bars show the mean±s.d. For DNAM1: vehicle n=5, ruxolitinib n=4; for KLGR1: vehicle n=5, ruxolitinib n=5. **P<0.01, ***P<0.005, with two tailed t-test. (d) MACS-purified and IL-2 expanded NK cells were pretreated 3 h with vehicle or 0.5 μM ruxolitinib and stimulated for 1 h with IL-12 (5 ng ml⁻¹) and IL-15 (50 ng ml⁻¹). Granzyme B (GZMB) expression was detected by FACS and representative images are shown. Data are expressed as percentage of positive NK cells, and the mean±s.d. of a technical triplicate is shown. **P<0.01, ***P<0.005 with two tailed t-test. (e) Splenic NK-cell maturation analysis from EO771 tumour-bearing mice treated with vehicle or ruxolitinib (90 mg kg⁻¹ BID) as described in Fig. 2a (protocol II). CD3⁻ CD49b⁺ NK cells were analysed for CD27 and CD11b expression and representative images of FACS gating are shown. Bars show the mean±s.e.m. Vehicle n=7, ruxolitinib n=8. *P<0.05, ***P<0.005 with two tailed Mann–Whitney test.

Figure 6. Inhibition of the JAK/STAT pathway blocks proliferation and cytotoxicity of NK cells.

(a) Western blot analysis of pSTAT5, pSTAT3 and pSTAT4 levels in the human NK cell line, NK-92, after 6 h of treatment with different concentrations of ruxolitinib or BSK805. Total (T) STAT5, STAT3 and STAT4 levels are also shown. (b) Proliferation of NK-92 cells after treatment with different concentrations of ruxolitinib for 4 days or BSK805 for 3 days. Data, expressed as mean±s.d. of technical replicates (ruxolitinib n=5, BSK805 n=4), are given as the ratio of proliferating cells compared with vehicle. (c) Cytotoxicity of NK-92 cells against HeLa cells, after 6 h of culture, was analysed by flow cytometry. NK-92 cells were pretreated for 3 h with 0.5 μM ruxolitinib, or 0.5 μM BSK805 or vehicle, then counted and mixed with HeLa cells at the indicated effector–target ratio (E:T). Data are expressed as percentage of NK-92 killing ability; the mean±s.d. of technical replicates is shown. *P<0.05, **P<0.01, ***P<0.005 with two tailed t-test.

Figure 7. JAK2 activation is essential for NK-cell-mediated anti-tumour immunity.

(a) Quantification of NK cells (CD3⁻ CD49b⁺) in spleens from JAK2^fl/fl and JAK2^fl/fl Mx1-Cre mice. Bars show the mean±s.e.m. JAK2^fl/fl n=6, JAK2^fl/fl Mx1-Cre n=5. (b) NK-cell maturation stages in spleens from JAK2^fl/fl and JAK2^fl/fl Mx1-Cre mice; CD3⁻ CD49b⁺ NK cells were analysed for CD27 and CD11b expression. Bars show the mean±s.e.m. of the splenic NK-cell percentage for each gate. JAK2^fl/fl n=6, JAK2^fl/fl Mx1-Cre n=5. (c) Quantification of DNAM1⁺ NK cells (CD3⁻ CD49b⁺) in spleens from JAK2^fl/fl and JAK2^fl/fl Mx1-Cre mice. Bars show the percentage±s.e.m. JAK2^fl/fl n=6, JAK2^fl/fl Mx1-Cre n=5. (d) Top: schematic representation of the treatment schedule. Bottom: quantification of lung metastasis after i.v. injection of 2x10⁵ EO771 tumour cells and treatment with vehicle or BSK805 (90 mg kg⁻¹ daily) from day −2 to day +9. The number of metastases in the whole lung from individual mice and the mean±s.e.m are shown. Vehicle n=7, BSK805 n=6. (e) Top: schematic representation of the treatment schedule. Bottom: Kaplan–Meier curves show the incidence of metastasis detected by bioluminescent imaging performed every 3 days. Mice were treated with NK1.1 Ab (100 μg per mouse) at day −12, −7, −2, +2 and +7 from i.v. injection of EO771 tumour cells on day 0. BSK805 (90 mg kg⁻¹) was administered daily either alone, or in combination with the NK1.1 antibody, from day −2 until the end of the experiment. n=8 mice per group. Log-rank (Mantel-Cox) Test: vehicle versus BSK805 **P=0.002, vehicle versus NK1.1 Ab ***P=0.0003, BSK805 versus NK1.1 Ab NS. P=0.2952, BSK805 versus BSK805/NK1.1 Ab **P=0.0085, NK1.1 Ab versus BSK805/NK1.1 Ab NS. P=0.2405.

Figure 8. Immunostimulation with IL-15 prevents the JAKi-mediated increase of metastasis.

(a) Top: schematic representation of the treatment schedule. 4x10⁵ EO771 tumour cells were injected i.v. and mice were treated with IL-15 (10 μg per mouse) or the combination of IL-15 and BSK805 (90 mg kg⁻¹ daily) from day −2 to day +6. Bottom: quantification of NK cells (CD3⁻ NK1.1⁺) in the peripheral blood collected by tail vein puncture from randomly chosen mice at day +5 after i.v. injection of tumour cells. Bars show the percentage of circulating NK cells in individual mice and the mean±s.e.m for each group. Vehicle n=5, IL-15 n=4 and IL5 plus BSK805 n=4. *P<0.05 with Mann–Whitney test. (b) Quantification of lung metastases at the end of the experiment. The number of metastasis in the whole lung from individual mice and the mean±s.e.m are shown. n=6 mice per group. *P<0.01, **P<0.001, NS, not significant with Dunn's Multiple Comparison Test. (c) STATs are active in metastatic tumour cells, and in NK cells in the tumour environment. Left: JAK/STAT pathway inhibition in NK cells weakened anti-tumour immunosurveillance by reducing proliferation, maturation and NK-cell activation. The immunomodulatory properties of JAKi prevail over potential anti-tumour effects of blocking STAT activation in cancer cells, thus leading to enhanced metastasis. Right: the combinatorial treatment of a JAKi with the immunostimulator IL-15 blocks metastasis formation, potentially by boosting NK-cell function through multiple JAKs.