EYA3 regulation of NF-κB and CCL2 suppresses cytotoxic NK cells in the premetastatic niche to promote TNBC metastasis

Abstract

Triple-negative breast cancer cells must evade immune surveillance to metastasize to distant sites, yet this process is not well understood. The Eyes absent (EYA) family of proteins, which are crucial for embryonic development, become dysregulated in cancer, where they have been shown to mediate proliferation, migration, and invasion. Our study reveals an unusual mechanism by which EYA3 reduces the presence of cytotoxic natural killer (NK) cells in the premetastatic niche (PMN) to enhance metastasis, independent of its effects on the primary tumor. We find that EYA3 up-regulates nuclear factor κB signaling to enhance CCL2 expression, which, in contrast to previous findings, suppresses cytotoxic NK cell activation in vitro and their infiltration into the PMN in vivo. These findings uncover an unexpected role for CCL2 in inhibiting NK cell responses at the PMN and suggest that targeting EYA3 could be an effective strategy to reactivate antitumor immune responses to inhibit metastasis.

Fig. 1.. EYA3 correlates with poor prognosis in human TNBC and promotes primary tumor growth and metastasis in immune-competent models of TNBC.

(A) Gene expression for EYA3 was used to stratify the overall cohort into quartiles, and the top and bottom quartiles were compared for recurrence-free survival using the survival and survminer R packages. Recurrence-free survival was analyzed with patients living the past 10 years right censored; the log-rank test was used to calculate statistical significance. (B) Eya3 mRNA expression, as measured using real-time qRT-PCR. Eya3 expression was normalized to Gapdh levels. (C) Western blot analysis of EYA3 in 66cl4 SCR and Eya3 KD cells. (D) Experimental design for the in vivo experiment with data in (E) and (F) (largest tumor reaches 1 cm³) and (G) (size-matched tumors, 1 cm³). (E) Primary tumor growth via caliper measurement after orthotopic injection of luciferase-tagged 66cl4 SCR (n = 5), shEya3 KD1 (n = 4), and shEya3 KD2 (n = 5) cells into BALB/c mice. (F) Bioluminescence intensity in the lung window over time after primary tumor removal for mice injected with 66cl4 SCR (n = 4), shEya3 KD1 (n = 4), and shEya3 KD2 (n = 3). Two mice were euthanized and excluded due to surgical complications, and one (KD2) mouse was excluded from analysis after calculated to be a significant outlier via Grubbs’ test. (G) Bioluminescence intensity in the lung window of mice injected with 66cl4 SCR (n = 4), shEya3 KD1 (n = 3), or shEya3 KD2 (n = 4) cells after orthotopic injection into the mammary fat pad followed by primary tumor removal when all tumors reached the size of 1 cm³. Statistical analysis for (E) to (G) was performed using longitudinal mixed-model analysis in R. Means ± SEM are shown.

Fig. 2.. EYA3 promotes NF-κB signaling in TNBC cells.

(A) RNA-seq data were analyzed for top pathways enriched in the SCR group compared to the combined Eya3 KD groups. Shown are the top 10 enriched Hallmark gene sets by GSEA, plotted in order of normalized enrichment score (NES). (B) GSEA plot of Hallmark TNFα signaling via the NF-κB gene set between 66cl4 SCR and combined Eya3 KD cells. (C) Heatmap of RELA transcription factor target genes between 66cl4 SCR and Eya3 KD cells. (D to F) Western blot analyses of cytoplasmic and nuclear protein fractions of 66cl4, Met1, and E0771 SCR and Eya3 KD cells for markers of NF-κB activation.

Fig. 3.. EYA3-mediated NF-κB signaling promotes metastasis and worsens survival in immune-competent mouse models of TNBC.

(A) Western blot analysis of cytoplasmic and nuclear protein fractions of 66cl4 SCR + EV, shEya3 KD2 + EV and shEya3 KD2 + IKK2-SE for Flag-IKK2-SE and markers of NF-κB activation. (B) GSEA plot of Hallmark TNFα signaling via NF-κB gene set from RNA-seq analysis comparing gene expression in 66cl4 shEya3 KD2 + IKK2-SE versus shEya3 KD2 + EV cells. (C) GSEA plot of Hallmark inflammatory response gene set from RNA-seq analyzed in (B). (D) Primary tumor growth, as measured by caliper measurement, of 66cl4 SCR + EV (n = 9), shEya3 KD2 + EV (n = 7), and shEya3 KD2 + IKK-SE (n = 7) cells after orthotopic injection into the mammary fat pad of BALB/c mice. (E) Bioluminescence intensity in lung window of BALB/c mice after primary tumor removal for BALB/c mice injected with 66cl4 SCR + EV (n = 9), shEya3 KD2 + EV (n = 6), and shEya3 KD2 + IKK2-SE (n = 6) cells. Statistical analysis for (D) and (E) performed using a longitudinal mixed model in R. (F) Overall survival to an ethical endpoint for mice from (E). Statistical analysis was performed using log-rank test to compare survival.

Fig. 4.. Eya3 KD in TNBC cells alters immune infiltration to the premetastatic lung.

(A) Schematic representation of the experimental setup. 66cl4 SCR, shEya3 KD1, and shEya3 KD2 cells were orthotopically injected into the fourth mammary fat pad of syngeneic BALB/c mice. When tumor size exceeded 50 mm³, mice were euthanized. Lung single-cell suspensions were analyzed for immune cells by flow cytometry. (B) Growth of premetastatic primary tumors as measured by caliper measurement (n = 10 per group). Mice that did not reach 50 mm³ after ~30 days were excluded from subsequent flow analyses. (C) Kaplan-Meier curves of time for tumors to reach 50 mm³. (D to G). Flow cytometry analysis was used to examine the presence of inflammatory monocytes (D), interstitial macrophages (E), dendritic cells (F), and cytotoxic NK cells (G) in the lungs of premetastatic mice. Data were collected from 7 to 10 mice per group (SCR, n = 10; KD1, n = 8; KD2, n = 7), from two independent experiments (represented by open or closed circles). Mice that did not form tumors were removed from analysis. Statistical analysis was performed using ANOVA with sum contrasts in R.

Fig. 5.. Cancer cell–secreted factors alter cytotoxic NK cell infiltration to the lung and reduce cytotoxic NK cell activation in vitro.

(A) Experimental setup: phenol red-free, serum-free media was incubated with cancer cells ± Eya3 KD for 48 hours, then collected and filtered. Three hundred microliters of media was injected intraperitoneally into mice for 7 days, then mice were euthanized and lungs were made into single-cell suspensions. (B to E) Lungs were analyzed for immune cell subsets by flow cytometry. Data were collected from five mice per group. (F) Schematic representation of the experimental setup. Briefly, NK cells were enriched from syngeneic mouse spleens using negative selection beads and incubated for 48 hours with IL-15 on a 0.4 μM transwell placed on top of cancer cells with or without Eya3 KD. PMA + ionomycin + Golgi plug was added for the last 4 hours of incubation. (G and H) NK cell-enriched splenocytes cocultured with various 66cl4 cells were analyzed for the production of TNFα and IFN-γ by flow cytometry. Shown is a representative histogram of TNFα and IFN-γ flow cytometry staining from control conditions where NK cells have been cultured in media alone, and either not stimulated with PMA/ionomycin (Alone no stim) or stimulated with PMA/ionomycin (Alone + stim). Also shown are representative histograms of NK cells cultured with 66cl4 with or without Eya3 KD (+SCR or +KD1 or +KD2) and stimulated with PMA/ionomycin as detailed in (F). Gates were drawn based on unstained control, and % of cells expressing TNFα and IFN-γ were quantified and normalized to SCR. Data were collected from four independent experiments and graphed in (H), and statistical analysis was performed using Kruskal-Wallis, adjusted for multiple comparisons. (I) As in (H), for E0771. Data were collected from five independent experiments, and statistical analysis was performed using Kruskal-Wallis, adjusted for multiple comparisons.

Fig. 6.. EYA3 promotes expression of the premetastatic niche regulator CCL2, downstream of NF-κB.

(A) Heatmap of DEGs in the GO MF (molecular function) cytokine activity gene set from RNA-seq of 66cl4 SCR + EV, shEya3 KD2 + EV, and shEya3 KD2 + IKK2-SE. (B) 66cl4 cells with or without Eya3 KD were probed for cytokine expression by qRT-PCR, normalized to Actin, and fold change was calculated compared to SCR cells. (C) 66cl4 or E0771 cells with or without Eya3 KD were probed for Ccl2 expression by qRT-PCR and normalized to Actin. Fold change was calculated compared to SCR cells. (D) Media was incubated with 66cl4 or E0771 cells for 48 hours. Conditioned media was then probed for CCL2 by Western blot. Shown is a representative image of three independent experiments. (E) Z-transformed mRNA expression for Ccl2 was plotted between shSCR + EV, shEya3 KD2 + EV, and shEya3 KD2 + IKK2-SE from RNA-seq, related to (A). (F) Media was incubated with 66cl4 cells for 48 hours, after which CCL2 levels were determined by ELISA. (G) As in (C), for HCC70 cells. (H) As in (F), for HCC70 cells. Statistical analyses for all qRT-PCR and ELISA data were performed using one-way ANOVA assuming unequal variances; when the overall F test was found to be P < 0.05, unpaired two-tailed Welch’s t test was performed to assess differences between experimental predefined comparison groups. All experiments are representative of three independent biological experiments, which were each conducted with technical triplicates.

Fig. 7.. Expression of Ccl2 in Eya3 KD cells rescues effects of CM on cytotoxic NK cell infiltration to the lung and activation in vitro.

(A and B) Ccl2 was stably expressed, and media was analyzed for CCL2 levels by (A) Western blot and (B) ELISA. (C) NK cell–enriched splenocytes cocultured with 66cl4 cells were analyzed for the production of TNFα and IFN-γ by flow cytometry. Data were collected from four independent experiments and analyzed by unpaired two-tailed Welch’s t test. (D) Cytotoxic NK cells were analyzed by flow cytometry in the lungs of mice treated with CM. Data were collected from 10 mice per group, from two independent experiments (represented by open or closed circles), and analyzed by Kruskal-Wallis, adjusted for multiple comparisons. (E) Experimental setup: CM was collected from cancer cells with or without Eya3 KD after 48 hours of incubation. Three hundred microliters of CM was injected intraperitoneally into mice for 7 days before the injection of 66cl4 SCR cells via tail vein into all groups. Anti-asialo GM1 depleting antibody or IgG control was administered before cancer cell injection. CM was injected daily throughout the course of the experiment. (F and G) Bioluminescence intensity in the lung window was detected by IVIS imaging and mean intensity was plotted over time. Three mice were excluded from the SCR + IgG group due to poor tumor injection or immediate death upon tumor injection. (SCR media + IgG, n = 7; KD1 media + IgG, n = 10; KD1 media + anti-asialo GM1, n = 10). (H) A second experiment was conducted as in (E), with the added group of SCR media + anti-asialo GM1. Mice that died prematurely from non–cancer-related deaths were excluded (SCR media + IgG, n = 8; SCR media + anti-asialo GM1, n = 8; KD1 media + IgG, n = 10; KD1 media + anti-asialo GM1, n = 9). Statistical analysis for (F) was performed using longitudinal mixed-model analysis, and that for (G) and (H) was performed using Kruskal-Wallis, adjusted for multiple comparisons.

Fig. 8.. EYA3 expression correlates with NF-κB signaling and CCL2 expression in human BC.

(A) Schematic of pairwise correlation analysis of METABRIC patient data (–82) ranking correlated genes with EYA3 and looking at enrichment of correlated genes within Hallmark gene sets. (B) Gene expression for the METABRIC dataset (–82) was downloaded from cBioPortal, and CCL2 mRNA expression was compared against EYA3 mRNA expression for each sample. (C) Using the GSVA R package, ssGSEA was used to calculate the NF-κB score using the Hallmark gene set collection, HALLMARK_TNFA_SIGNALING_VIA_NFKB (116, 117), for each sample in the METABRIC dataset and compared to CCL2 mRNA expression. (D) CCL2 and EYA3 mRNA expression was evaluated in varying genetic subsets in the METABRIC dataset.