NFκB ACTIVATION BY HYPOXIC SMALL EXTRACELLULAR VESICLES DRIVES ONCOGENIC REPROGRAMMING IN A BREAST CANCER MICROENVIRONMENT

Abstract

Small extracellular vesicles (sEV) contribute to the crosstalk between tumor cells and stroma, but the underlying signals are elusive. Here, we show that sEV generated by breast cancer cells in hypoxic (sEV^HYP), but not normoxic (sEV^NORM) conditions activate NFκB in recipient normal mammary epithelial cells. This increases the production and release of inflammatory cytokines, promotes mitochondrial dynamics leading to heightened cell motility and disrupts 3D mammary acini architecture with aberrant cell proliferation, reduced apoptosis and EMT. Mechanistically, Integrin-Linked Kinase packaged in sEV^HYP via HIF1α is sufficient to activate NFκB in the normal mammary epithelium, in vivo. Therefore, sEV^HYP activation of NFκB drives multiple oncogenic steps of inflammation, mitochondrial dynamics, and mammary gland morphogenesis in a breast cancer microenvironment.

INTRODUCTION

Small (40–150 nm) extracellular vesicles (sEV) originating from the late endosomal trafficking machinery or via shedding of the plasma membrane [1] are important mediators of tumor-stroma crosstalk [2]. Released by virtually all cells, including breast cancer cells [3], sEV can package a diverse array of bioactive proteins and RNAs that affect multiple traits of a tumor microenvironment, including stemness [4], exit from dormancy [5] and inflammatory responses [6]. In addition, sEV have been linked to enhanced cell motility of both tumor and stromal components [7]. Mechanistically, this involves modulation of directional cell movements [8], integrin-dependent adhesion [9], and, more recently, deregulated mitochondrial dynamics with increased mitochondrial accumulation at the cortical cytoskeleton to fuel membrane requirements of cell motility [10].

Despite these advances and their clear disease-relevance [11], the requirements of sEV signaling, especially in the regulation of a tumor microenvironment, are still mostly unknown [12]. In particular, a role of sEV in pleiotropic inflammatory mechanisms, capable of influencing local and distant recurrences, as well as response to therapy, has not been widely investigated [13].

In this study, we examined the effect of breast cancer-derived sEV generated in hypoxic conditions [10], a common occurrence in a tumor microenvironment, on inflammatory reprogramming of normal recipient mammary epithelial cells.

RESULTS AND DISCUSSION

Hypoxic sEV-associated transcriptome.

We isolated sEV from HER2⁻/ER⁺/PR⁺ human breast adenocarcinoma MCF7 cells maintained under conditions of normoxia (sEV^NORM) or hypoxia (sEV^HYP). sEV quantified after each collection had comparable morphology (Supplementary Fig. S1A), yield (Supplementary Fig. S1B) and size distribution (Supplementary Fig. S1C), in agreement with previous findings [10]. sEV^NORM and sEV^HYP isolated from HER2⁻/ER⁻/PR⁻ human breast adenocarcinoma MDA-231 or normal mammary epithelial MCF10A cells were also comparable (Supplementary Figure S1B). sEV production in these settings was associated with negligible (<5%) cell death (Supplementary Figure S1D). By RNA-Seq profiling [10], sEV^HYP, but not sEV^NORM produced by MCF-7 cells caused the upregulation of several NFκB target genes in recipient MCF10A cells (Fig. 1A), including TNFα, IFNγ, IL1β, IL6, IL17, RelA, and PRKCD (Supplementary Fig. S1E).

Fig. 1: Breast cancer-derived sEV^HYP activate NFκB in MCF10A recipient cells.

A Bioinformatics analysis of NFκB target genes upregulated by sEV^HYP in recipient MCF10A cells by RNA-Seq. B MCF10A cells treated with sEV^NORM or sEV^HYP for 24 h were analyzed for nuclear accumulation of p50 NFκB by fluorescence microscopy. Representative images. Scale bar, 25 μm. Ctrl, control. Cyan, nuclei; yellow, p50 NFκB. C The conditions are as in B and nuclear accumulation of p50 NFκB was quantified at the indicated increasing concentrations of sEV^NORM or sEV^HYP. Mean±SD (two independent experiments). D MCF10A cells treated with sEV^NORM or sEV^HYP were incubated with pXSC or transfected with p65 NFκB -directed siRNA (si-p65) and analyzed for expression of the indicated NFκB target genes, by qRT-PCR. Mean±SD (N=3). E Conditioned media from MCF10A cells treated as in D were analyzed for cytokine release by LEGENDplex. Mean±SD (N=4). F sEV-induced nuclear accumulation of p50 NFκB was quantified in the presence or absence of pXSC. For panels C-F, numbers correspond to p values by 1-way Anova with Tukey’s posttest.

In validation experiments, sEV^HYP treatment increased NFκB transcriptional activity using a luminescence reporter gene (Supplementary Fig. S1F), promoted nuclear accumulation of the p50 subunit of NFκB in a dose-dependent fashion (Fig. 1B, C), and upregulated the mRNA levels of NFκB target cytokines, IL6, IL8, TNFα and MCP1 (Fig. 1D) in recipient MCF10A cells. This was associated with increased release of IL6 and TNFα in the supernatant of sEV^HYP-treated MCF10A cells (Fig. 1E). In terms of specificity, a small molecule inhibitor of NFκB DNA binding, 1,4-phenylenebis(methylene)selenocyanate (pXSC) or siRNA silencing of the p65 subunit of NFκB prevented sEV^HYP-induced NFκB transcriptional activity (Supplementary Fig. S1F), blocked nuclear accumulation of p50 NFκB (Fig. 1F), and suppressed inflammatory cytokine production (Fig. 1D) and release (Fig. 1E) to levels of control cultures. sEV^NORM had no effect on NFκB signaling in the presence or absence of pXSC or p65 silencing (Fig. 1D–F, Supplementary Fig. S1F).

Requirements of ILK and HIFIα for sEV^HYP induced NFκB activity.

We next looked at the composition of NFκB -inducing sEV^HYP. Consistent with previous observations [10], sEV^HYP produced by MCF7 or MDA-231 breast cancer cell types contained higher levels of Integrin-Linked Kinase (ILK), compared to sEV^NORM, whereas no differences were observed in sEV produced by MCF10A cells, by flow cytometry (Supplementary Fig. S2A). Similarly, 3D reconstruction of confocal microscopy images from MCF7 cells showed increased co-localization of ILK with the sEV marker, CD63 in hypoxia, but not normoxia (Supplementary Fig. S2B, C). siRNA silencing of HIF1α in sEV^HYP-producing MCF7 cells suppressed the co- localization of ILK with CD63, by confocal microscopy (Supplementary Fig. S2D), and reduced the levels of ILK in sEV, by Western blotting (Supplementary Fig. S2E, F). Similar results were obtained with an independent breast cancer cell type, T47D (Supplementary Fig. S2D). Conversely, HIF1α silencing did not affect ILK levels in sEV^NORM, by confocal microscopy (Supplementary Fig. S2B–D) or Western blotting (Supplementary Fig. S2E, F). Other cellular (calnexin) or sEV (Flotillin 1 and TSG101) markers were also unaffected (Supplementary Fig. S2E).

Based on these data, we next asked if ILK and/or HIF1α were important in sEV^HYP induction of NFκB. In these experiments, siRNA silencing of ILK or HIF1α in sEV-producing breast cancer cells suppressed nuclear accumulation of p50 NFκB (Supplementary Fig. S3A) and cytokine mRNA upregulation in recipient cells (Supplementary Fig. S3B). Small molecule inhibitors of ILK (Cpd22) or its downstream target, Akt (MK2206) gave similar results, preventing nuclear localization of p50 NFκB (Supplementary Fig. S3C) and upregulation of cytokine mRNA levels (Supplementary Fig. S3D), compared to controls. As an independent approach, transfection of MCF10A cells with WT ILK, but not loss of function L207W ILK mutant was sufficient to induce nuclear accumulation of p50 NFκB (Supplementary Fig. S4A, B), increase cytokine mRNA levels (Supplementary Fig. S4C) and upregulate NFκB target proteins, MMP9 and BclXL, by Western blotting (Supplementary Fig. S4D).

NFκB regulation of sEV^HYP-induced mitochondrial-fueled cell motility.

Consistent with recent findings [10], sEV^HYP stimulated mitochondrial dynamics in MCF10A cells (Fig. 2A) resulting in increased rates of both mitochondrial fusion and fission (Fig. 2B). This was accompanied by redistribution of mitochondria to the cortical cytoskeleton of sEV^HYP recipient cells (Fig. 2C, Supplementary Fig. S5B). NFκB inhibition with pXSC normalized mitochondrial dynamics in the presence of sEV^HYP (Fig. 2A), restored the rate of mitochondrial fusion and fission events to levels of control cultures (Fig. 2B, Supplementary Fig. S5A) and prevented the redistribution of mitochondria to the cortical cytoskeleton (Fig. 2C, Supplementary Fig. S5B). Treatment of recipient cells with sEV^NORM had no effect on mitochondrial dynamics or mitochondrial localization to the cortical cytoskeleton in the presence or absence of pXSC (Fig. 2B, C and Supplementary Fig. S5A, B).

Fig. 2: NFκB regulation of sEV^HYP-dependent mitochondrial dynamics and cell motility.

A, B MCF10A cells treated with sEV^HYP were incubated with or without pXSC and analyzed for changes in mitochondrial volume indicative of organelle fusion (>1.3-fold, positive y-scale) or fission (<0.7-fold, negative y-scale) over 40-sec intervals (A) and the number of mitochondrial fusion and fission events was quantified (B). Each line corresponds to a single cell. Mean±SD (N=3). Numbers correspond to p values by unpaired t-test. C MCF10A cells treated with sEV^NORM or sEV^HYP were imaged after 24 h by confocal microscopy and mitochondrial accumulation at the cortical cytoskeleton was quantified. Mean±SD (N=3). D MCF10A cells as in C were analyzed for single cell motility in 2D contour plots with quantification of speed of cell movements (Velocity, Vel, μm/min) and distance traveled by individual cells (Dis, μm). The cutoff velocities for slow (black, <0.53 μm/min)- or fast (red, >0.53 μm/min)-moving cells are indicated. E MCF10A cells transfected with siCtrl or si-p65 NFκB were treated with sEV^NORM or sEV^HYP and analyzed for directional cell migration in a wound closure assay at the indicated time intervals. Mean±SD (N=3). F MCF10A cells incubated with sEV^NORM or sEV^HYP were treated with pXSC or transfected with si-p65 NFκB and analyzed for migration on PET inserts.Mean±SD (N=3). For panels C and F, numbers correspond to p values by 1-way Anova with Tukey’s posttest.

Increased mitochondrial dynamics mediated by sEV^HYP fuels directional cell motility [10, 14] and a role of NFκB in this response was next investigated. Treatment with pXSC or siRNA silencing of p65 NFκB suppressed single cell motility mediated by sEV^HYP (Fig. 2D, Supplementary Fig. S5C), inhibited directional cell motility in a wound closure assay (Fig. 2E, Supplementary Fig. S5D) and blocked cell migration in a Transwell assay (Fig. 2F). In rescue experiments, reconstitution of recipient MCF10A cells with WT p65 NFκB (Supplementary Fig. S5E) normalized single cell motility and restored the speed of cell movements and the distance traveled by individual cells to levels of control transfectants (Supplementary Fig. S5F). Conversely, reconstitution with a loss-of-function T305D p65 NFκB mutant had no effect (Supplementary Fig. S5E, F).

Deregulation of epithelial morphogenesis by sEV^HYP-NFκB signaling.

When tested in a model of 3D mammary gland formation [10], sEV^HYP treatment increased MCF10A acini surface area and reduced acini circularity (Supplementary Fig. S6A, B). This was associated with extensive oncogenic changes, including nuclear accumulation of p50 NFκB, sustained upregulation of ILK, increased cell proliferation by Ki-67 staining and direct cell counting (Supplementary Fig. S6C) and impaired apoptosis with reduced caspase-3 expression and upregulation of survivin (Fig. 3A, B). sEV^NORM had no effect on MCF10A acini size and circularity (Supplementary Fig. S6A, B) or oncogenic changes (Fig. 3A, B, Supplementary Fig. S6C). In addition, sEV^HYP treatment induced hallmarks of EMT in MCF10A acini with upregulation of vimentin, N-cadherin and nuclear accumulation of SNAIL by fluorescence microscopy (Fig. 3A, B) as well as Western blotting (Supplementary Fig. S6E). Mechanistically, NFκB inhibition with pXSC restored MCF10A acini morphogenesis (Supplementary Fig. S6B), suppressed oncogenic changes (Fig. 3A, B), including cell proliferation (Supplementary Fig. S6D), and prevented EMT (Fig. 3A, B, Supplementary Fig. S6E) induced by sEV^HYP. As an independent approach, small molecule inhibition of ILK or Akt with Cpd22 or MK2206, respectively, also prevented sEV^HYP upregulation of N-cadherin and vimentin by Western blotting (Supplementary Fig. S6E).

Fig. 3: NFκB regulation of 3D mammary acini morphogenesis.

A MCF10A acini in 3D culture seeded in the presence of Ctrl, sEV^NORM or sEV^HYP with or without pXSC (added once at t=0) were analyzed by fluorescence microscopy 21 d after seeding (representative images; scale bar, 15 μm). B The conditions are as in A and the expression of p50 NFκB, ILK, Ki-67, cleaved (Cl) caspase 3, survivin, SNAIL and vimentin was quantified. Magenta, nuclei; yellow, individual protein of interest. MFI, mean fluorescence intensity. Mean±SD (N=3). Numbers correspond to p values by 1-way Anova with Tukey’s posttest.

Finally, we tested the potential relevance of this pathway, in vivo, and we isolated sEV from mouse breast adenocarcinoma AT3 cells under hypoxic or normoxic conditions. AT3-derived sEV^NORM or sEV^HYP had comparable size particle distribution (Supplementary Fig. S7A) and expressed sEV (flotillin, TSG101), but not cellular (calnexin, GM130) markers (Supplementary Fig. S7B). Injection of AT3-derived sEV^HYP in the abdominal mammary gland of immunocompetent C57BL/6 mice was associated with increased ILK expression in the recipient epithelium after 6 weeks, by immunohistochemistry (IHC) (Fig. 4A, B). Consistent with the data above, sEV^HYP injection in the normal mammary epithelium resulted in nuclear accumulation of p50 NFκB, in vivo, by immunofluorescence staining (Fig. 4C, D), as well as EMT, with upregulation of vimentin (Fig. 4A, B), decreased level of E-cadherin and nuclear localization of SNAIL (Fig. 4C, D).

Fig. 4: sEV-induced EMT in the mammary gland, in vivo.

A, B AT3-derived sEV^HYP were injected in the abdominal mammary gland of C57BL/6 mice and tissue samples harvested after 6 weeks were analyzed by immunohistochemistry (A) and reactivity with antibodies to ILK or vimentin was quantified (B). Ctrl, control. Representative images. Scale bar, 100 μm. C, D The conditions are as in A and mammary gland tissues harvested after 6 weeks were analyzed by immunofluorescence (C, representative images) and changes in expression of p50 NFκB, ILK, E-cadherin and SNAIL were quantified (D). Scale bar, 50 μm. MFI, mean fluorescence intensity. Mean±SD. Numbers correspond to p values by unpaired t-test.

In sum, we have shown that sEV packaged by breast cancer cells in hypoxia [10] activate NFκB in a normal mammary epithelium, promoting inflammatory cytokine release, mitochondrial dynamics with heightened cell migration, disruption of 3D acini morphogenesis and EMT, in vivo. A role of sEV, let alone sEV^HYP in NFκB activation has not been previously described. This pathway may be uniquely suited to globally alter a hypoxic breast cancer microenvironment, which occurs commonly in patients, favoring field cancerization [15] and oncogenic inflammatory changes [13]. Key effectors of this response included HIF1α, which has been linked to sEV-induced immunologic and metabolic changes in breast cancer [16] as well as ILK, leading to downstream activation of its pleiotropic oncogenic target, Akt [17]. Although its mechanism(s) of action continue to be debated [18], these data identify ILK as an important therapeutic target to disrupt sEV-induced oncogenic and inflammatory signaling during breast cancer development.