Obesity alters the lung myeloid cell landscape to enhance breast cancer metastasis through IL5 and GM-CSF

Abstract

Obesity is associated with chronic, low-grade inflammation, which can disrupt homeostasis within tissue microenvironments. Given the correlation between obesity and relative risk of death from cancer, we investigated whether obesity-associated inflammation promotes metastatic progression. We demonstrate that obesity causes lung neutrophilia in otherwise normal mice, which is further exacerbated by the presence of a primary tumour. The increase in lung neutrophils translates to increased breast cancer metastasis to this site, in a GM-CSF- and IL5-dependent manner. Importantly, weight loss is sufficient to reverse this effect, and reduce serum levels of GM-CSF and IL5 in both mouse models and humans. Our data indicate that special consideration of the obese patient population is critical for effective management of cancer progression.

Figure 1

Obesity is associated with lung neutrophilia driven by adiposity. (a) Weight curves for the diet-induced obesity (DIO) model. 5-week-old female BL6 mice were fed a low-fat (LF) or high-fat (HF) diet for 15 weeks. LF, n = 8 mice; HF, n = 10 mice; mean ± s.e.m. (b) Left, flow cytometry of lung myeloid cells in the DIO model at 15 weeks. LF, n = 8 mice; HF, n = 10 mice; minimum–maximum boxplots, all data points shown. Right, CD11b⁺Gr1⁺ populations are shown as a red overlay on total CD11b⁺ cells, graphed on Ly6C (x axis) by Ly6G (y axis) dot plots. (c) Weight curves for the leptin-deficient genetic model of obesity (ob/ob). Female ob/ob or wild-type (WT) mice were fed a normal diet until the pre-defined weight endpoint of >40 g. n = 10 mice per group; mean ± s.e.m. (d) Left, flow cytometry of lung myeloid cells in the ob/ob model at 6 weeks. Representative plots (right) are displayed as in b. n = 10 mice per group; minimum–maximum boxplots, all data points shown. (e) Weight curves for the obesity-resistant Balb/c model. 5-week-old female Balb/c mice were fed a LF or HF diet for 15 weeks. n = 10 mice per group; mean ± s.e.m. (f) Left, flow cytometry of lung myeloid cells in the Balb/c model at 15 weeks. Representative plots (right) are displayed as in b. n = 10 mice per group, minimum–maximum boxplots, all data points shown. (g) Weight curves for the diet-switch model. 5-week-old female BL6 mice were fed a HF diet over 15 weeks, and then switched to LF diet for an additional 7 weeks (HF–LF). HF, n = 6 mice; HF–LF, n = 11 mice; mean ± s.e.m. (h) Left, flow cytometry of lung myeloid cell subsets in the diet-switch model. Representative plots (right) are displayed as in b. HF, n = 6 mice; HF–LF, n = 11 mice, minimum–maximum boxplots, all data points shown. Significance was calculated via two-tailed unpaired Student’s t-test throughout. NS, not significant. Box plots represent median and interquartile range while whiskers represent maximum and minimum values excluding outliers. Each symbol represents one mouse.

Figure 2

Obesity-associated lung neutrophilia is accompanied by prometastatic gene expression changes and enhanced metastatic progression. (a) qRT-PCR on whole lung, bone marrow (BM) or FACS-purified lung neutrophils from HF non-tumour-bearing mice. Data are displayed as log₁₀(fold change) relative to LF (centre normalized line). BM and whole lung, n = 5 mice; lung neutrophils, n = 4 mice. *P < 0.05, **P < 0.01, ***P < 0.001. (b) Primary tumour volume of 99LN cells injected via mammary fat pad (1.5 × 10⁶ cells per mouse) in the DIO model. Key time points of primary tumour progression are presented: palpable (14 d, n = 10 mice), early phases of primary tumour growth (28 d, n = 10 mice; pre-metastatic niche), late phases of primary tumour growth (56 d, n = 5 mice; micro-metastatic disease). Arrows: n = 5 mice per group were euthanized at 28 d and at 56 d for analysis of lung neutrophilia, as in d. (c) Flow cytometry of blood neutrophils in tumour-bearing LF or HF mice over time. 0 d represents baseline circulating neutrophils prior to tumour cell injection. n = 5 mice per time point (matched from 0 d-56 d). (d) Flow cytometry of lung neutrophils in tumour-bearing LF or HF mice after 28 d or 56 d. Quantification of lung neutrophils in non-tumour-bearing LF mice is included as a control (black line). The reported P values indicate significant differences between tumour-bearing LF versus HF mice, n = 5 mice per group. (e) Quantification of spontaneous micro-metastases in lung at 56 d, n = 5 mice per group. Metastases were counted manually from scanned images of lung tissues, representing ~10 mm². All data are displayed as mean ± s.e.m.; significance calculated via two-tailed, unpaired Student’s t-test throughout.

Figure 3

Obesity enhances experimental breast cancer metastasis to lung in a neutrophil-dependent manner. (a) BLI of the 5 week metastasis assay with 99LN tumour cells injected via the tail vein into the DIO model. LF, n = 8 mice; HF, n = 9 mice; Tukey boxplot, Mann–Whitney test. (b) Weekly BLI over the 5 week metastasis experiment as shown in a. Mean ± s.e.m. (one-sided); dotted lines of best fit with corresponding slopes are displayed. (c) Flow cytometry of GFP⁺ 99LN cells in DIO lung after the 5 week metastasis assay. LF, n = 8 mice; HF, n = 9 mice; minimum–maximum boxplots, all data points shown, Mann–Whitney test. (d) Left, flow cytometry of lung myeloid cells in the 5 week metastasis assay. Right, representative dot plot shows total CD11b⁺ cells from LF (grey) and HF (red) lungs, and double-positive Ly6C⁺ Ly6G⁺ cells (blue gate). LF, n = 8 mice; HF, n = 9 mice; minimum–maximum boxplots, all data points shown, two-tailed unpaired Student’s t-test. (e) Left, BLI of the 48 h metastasis assay with 99LN cells injected via the tail vein into the DIO model. Right, representative images are shown. n = 10 mice per group; Tukey boxplot, Mann–Whitney test. (f) Flow cytometry of GFP⁺ 99LN cells in DIO lung after 48 h metastasis assay. n = 10 mice per group; minimum–maximum boxplots, all data points shown, Mann–Whitney test. (g) Left, flow cytometry of lung myeloid cells in the 48 h metastasis assay. Right, representative plots are displayed as in d. n = 10 mice per group; minimum–maximum boxplots, all data points shown, Mann–Whitney test. (h) Left, BLI of the 48 h metastasis assay with 99LN cells injected via the tail vein into the DIO model, +/− a neutralization antibody against Gr1. Right, representative images are shown. LF, n = 5 mice; HF + IgG, n = 9 mice; HF + anti-Gr1, n = 8 mice; Tukey boxplot, Kruskal–Wallis and Dunn’s multiple comparisons test. (i) Flow cytometry of GFP⁺ 99LN cells in lung, corroborating results in h. LF, n = 5 mice; HF + IgG, n = 9 mice; HF + anti-Gr1, n = 8 mice; minimum–maximum boxplots, all data points shown, Kruskal–Wallis and Dunn’s multiple comparisons test. Box plots represent median and interquartile range while whiskers represent maximum and minimum values excluding outliers. Each symbol represents one mouse.

Figure 4

Serum GM-CSF is elevated in obesity in association with CD11b⁺Gr1⁺ cells. (a) Left, flow cytometry of circulating CD11b⁺Gr1⁺ cells in the DIO model. Right, CD11b⁺Gr1⁺ populations are shown as a red overlay upon total CD11b⁺ cells, graphed on Ly6C (x axis) by Ly6G (y axis) dot plots. LF, n = 8 mice; HF, n = 10 mice; minimum–maximum boxplots, all data points shown, two-tailed unpaired Student’s t-test. (b) Left, in vitro myelopoiesis assay, demonstrating increased differentiation of mouse BM cells towards a CD11b⁺Gr1⁺ phenotype after treatment with HF serum compared with LF serum from the DIO model. n = 3 independent BM isolations; mean ± s.e.m., two-tailed unpaired Student’s t-test. Right, representative flow plots are shown. (c) Left, in vitro myelopoiesis assay, demonstrating increased differentiation of mouse BM cells towards a CD11b⁺Gr1⁺ phenotype after treatment with obese serum compared with lean serum from human donors. n = 3 independent BM isolations; mean ± s.e.m., two-tailed unpaired Student’s t-test. Right, representative flow plots are shown. (d) Venn diagram of results from cross-species cytokine array (Supplementary Table 2). Out of 103 factors, 30 were elevated in HF versus LF mouse serum, and 16 out of 103 factors were elevated in obese versus lean human serum. Eight overlapping factors were identified, including CCL25, CD40L, GM-CSF, IGFBP2, IL5, IL6, MMP3 and MMP9. (e) In vitro myelopoiesis assay, testing the capacity of the 8 factors identified in d to regulate BM differentiation towards a CD11b⁺Gr1⁺ phenotype. n = 3 independent mouse BM isolations; mean ± s.e.m., one-way ANOVA and Dunnett’s multiple comparisons test. (f) In vitro myelopoiesis assay, demonstrating that GM-CSF neutralization reverses the effects of HF serum on CD11b⁺Gr1⁺ differentiation. n = 6 independent mouse BM isolations; mean ± s.e.m., one-way ANOVA and Dunnett’s multiple comparisons test. (g) qRT-PCR of Csf2 (GM-CSF) across different tissues in HF-fed animals. n = 4 mice per tissue; mean ± s.e.m. (h) qRT-PCR of Csf2 in different FACS-purified cell types from HF lung tissues. n = 5 mice per cell type; mean ± s.e.m. Box plots represent median and interquartile range while whiskers represent maximum and minimum values excluding outliers. Each symbol represents one mouse.

Figure 5

GM-CSF underlies obesity-associated lung neutrophilia and breast cancer metastasis. (a) Left, flow cytometry of CD11b⁺Gr1⁺ cells in blood and lung from WT BL6 animals treated for 5 d with rGM-CSF versus PBS. n = 5 mice per group; minimum–maximum boxplot, all data points shown, Mann–Whitney test. Right, lung CD11b⁺Gr1⁺ populations are shown as a red overlay upon total CD11b⁺ cells, graphed on Ly6C (x axis) by Ly6G (y axis) dot plots. (b) Trial schematic for c: daily rGM-CSF treatment (from day −5) followed by 99LN experimental metastasis assay (48 h; continued GM-CSF treatment). (c) Left, BLI quantification of the trial depicted in b. Right, representative images are shown. n = 5 mice per group; Tukey boxplot, Mann–Whitney test. (d) Trial schematic for e–g: antibody-based GM-CSF neutralization (day −3 and −1) followed by 99LN experimental metastasis assay (48 h) in the DIO model. (e) Left, BLI quantification of the trial depicted in d. Right, representative images are shown. LF/HF + IgG, n = 5 mice; LF + anti-GM-CSF, n = 4 mice; HF + anti-GM-CSF, n = 5 mice; Tukey boxplot, one-way ANOVA and Bonferroni’s multiple comparisons test. (f) Flow cytometry of GFP⁺ 99LN cells in lung after the trial depicted in d. LF/HF + IgG, n = 5 mice; LF + anti-GM-CSF, n = 4 mice; HF + anti-GM-CSF, n = 5 mice; minimum–maximum, all data points shown, one-way ANOVA and Bonferroni’s multiple comparisons test. (g) Flow cytometry of CD11b⁺Gr1⁺ cells in lung after the trial depicted in d. LF/HF + IgG, n = 5 mice; LF + anti-GM-CSF, n = 4 mice; HF + anti-GM-CSF, n = 5 mice; minimum–maximum, all data points shown, one-way ANOVA and Bonferroni’s multiple comparisons test. (h) ELISA of serum GM-CSF in DIO mice with orthotopic breast tumours, corresponding to the trial from Fig. 2b. Serum was isolated from tumour-bearing (TB) LF or HF animals after 28 d (early; pre-metastatic disease) and 56 d (late; early micro-metastatic disease). Non-tumour-bearing (NTB) LF animals are included for comparison. n = 5 mice per group; mean ± s.e.m., one-way ANOVA and Dunnett’s multiple comparisons test. NS, not significant. Box plots represent median and interquartile range while whiskers represent maximum and minimum values excluding outliers. Each symbol represents one mouse.

Figure 6

IL5 signalling supports lung neutrophilia. (a) qRT–PCR of Ccl25, Csf2, Igfbp2, Il5, Mmp3, Mmp9, Cd40lg and Il6 in visceral and subcutaneous fat from mice on HF diet. n = 4 mice per tissue; mean ± s.e.m. (b) Trial schematic for c: WT BL6 mice were treated daily with rIL5 or PBS (5 d), and immune cells in lung were quantified by flow cytometry. (c) Representative flow cytometry plots (left) and quantification (right) of lung neutrophils following the trial depicted in b. n = 5 mice per group; mean ± s.e.m., two-tailed unpaired Student’s t-test. (d) Representative flow plots showing gating strategy and population distribution for IL5Rα⁺ cells. (e) qRT-PCR of Csf2ra, Csf2rb and Il5ra in FACS-purified lung neutrophils, monocytes and eosinophils from the DIO model. n = 5 mice per group; mean ± s.e.m., two-tailed unpaired Student’s t-test. (f) Representative flow plots showing IL5Rα⁺ populations in human blood. Eosinophils (blue) were used as a positive gating control, n = 7 healthy donors. (g) Quantification of cell proliferation in vitro in response to rIL5 treatment via flow cytometry for Ki67⁺ cells. Cells isolated from n = 5 mice per group; Tukey boxplot, two-tailed unpaired Student’s t-test. (h) qRT–PCR of Csf2 expression in FACS-purified IL5Rα⁺ cells after treatment with rIL5 in vitro (100 ng ml⁻¹, 4 h). Cells isolated from n = 5 mice per group; mean ± s.e.m., two-tailed unpaired Student’s t-test. (i) Flow cytometry analysis of IL5Rα⁺ monocytes from blood and lung in the DIO model. n = 5 mice per group; Tukey boxplot, two-tailed unpaired Student’s t-test. (j) qRT–PCR of Csf2 expression in FACS-purified IL5Rα⁺ cell types from blood and lung in the DIO model. n = 5 mice per group; Tukey boxplot, two-tailed unpaired Student’s t-test. NS, not significant. Box plots represent median and interquartile range while whiskers represent maximum and minimum values excluding outliers. Each symbol represents one mouse.

Figure 7

Obesity enhances lung homing of neutrophils in an IL5-dependent manner. (a) Schematic representation of the adoptive cell transfer experiment. Neutrophils were isolated from BM of WT or ob/ob mice, labelled with fluorescent CellTrace dye (green and red, respectively), mixed 1:1, and then injected via the tail vein (3 × 10⁶ cells) into WT or ob/ob mice ± IL5 neutralizing antibody. Lungs were isolated for flow cytometry analysis after 4 h and 8 h to assess kinetics of neutrophil trafficking. (b) Flow cytometry validation of an equal 1:1 mix of WT (green; 49.3%) and ob/ob (red; 49.7%) donor neutrophils immediately prior to adoptive cell transfer injections. (c) Flow cytometry analysis of lung at 4 h post-adoptive transfer. The vast majority of labelled neutrophils at this time point were from ob/ob donors. n = 5 mice per recipient group, mean ± s.e.m., one-way ANOVA and Dunnett’s multiple comparisons test. (d) Representative flow cytometry plots for the data presented in c. (e) Flow cytometry analysis of fluorescently labelled circulating neutrophils 4 h post-adoptive transfer, demonstrating balanced representation of both red (ob/ob donor) and green (WT donor) cells. n = 5 mice per recipient group; mean ± s.e.m. (f) Flow cytometry analysis of lung at 8 h post-adoptive transfer. Equivalent representation of red and green donor neutrophils was observed at this time point compared with 4 h as in c. n = 5 mice per recipient group, mean ± s.e.m., one-way ANOVA and Dunnett’s multiple comparisons test. (g) Representative flow cytometry plots for the data presented in f. NS, not significant.

Figure 8

Weight loss reduces obesity-associated lung neutrophilia and metastasis in mice and humans. (a–c) Flow cytometry quantification of IL5Rα⁺ neutrophils (a), IL5Rα⁺ eosinophils (b) and IL5Rα⁺ monocytes (c) from lung in LF, HF and diet-switch (HF–LF) mice. LF, n = 5 mice; HF, n = 5 mice; HF–LF, n = 4 mice; mean ± s.e.m., one-way ANOVA and Dunnett’s multiple comparisons test. (d) Left, BLI of the 48 h metastasis assay with 99LN cells injected via the tail vein in the diet-switch trial. Right, representative images are shown. LF, n = 9 mice; HF, n = 10 mice; HF–LF, n = 10 mice; mean ± s.e.m., Kruskal–Wallis and Dunn’s multiple comparisons test. (e) qRT-PCR analysis of Cxcr2, Cxcr4, S100a8 and S100a9 expression in FACS-purified lung neutrophils. LF, n = 5 mice; HF, n = 5 mice; HF–LF, n = 4 mice; mean ± s.e.m., one-way ANOVA and Dunnett’s multiple comparisons test. (f) ELISA analysis of serum IL5 (left) and GM-CSF (right). LF, n = 5 mice; HF, n = 5 mice; HF–LF, n = 4 mice; mean ± s.e.m., one-way ANOVA and Dunnett’s multiple comparisons test. (g) ELISA analysis of serum IL5 (left) and GM-CSF (right) from human weight loss study. n = 10 donors per group; matched pre- and post-weight loss concentrations within a given individual are connected with a line; two-tailed paired Student’s t-test. NS, not significant. (h) Schematic representation of the proposed mechanism underlying obesity-associated lung neutrophilia. Adipose tissue-derived IL5 signals to IL5rα⁺ cells, leading to their expansion and upregulation of Csf2. This contributes to an environment that supports neutrophilia in the circulation and in the lungs. Lung neutrophils are reprogrammed by obesity to adopt pro-tumorigenic transcriptional signatures, and ultimately support metastatic progression.