Therapy-induced senescence promotes breast cancer cells plasticity by inducing Lipocalin-2 expression

The acquisition of novel detrimental cellular properties following exposure to cytotoxic drugs leads to aggressive and metastatic tumors that often translates into an incurable disease. While the bulk of the primary tumor is eliminated upon exposure to chemotherapeutic treatment, residual cancer cells and non-transformed cells within the host can engage a stable cell cycle exit program named senescence. Senescent cells secrete a distinct set of pro-inflammatory factors, collectively termed the senescence-associated secretory phenotype (SASP). Upon exposure to the SASP, cancer cells undergo cellular plasticity resulting in increased proliferation, migration and epithelial-to-mesenchymal transition. The molecular mechanisms by which the SASP regulates these pro-tumorigenic features are poorly understood. Here, we report that breast cancer cells exposed to the SASP strongly upregulate Lipocalin-2 (LCN2). Furthermore, we demonstrate that LCN2 is critical for SASP-induced increased migration in breast cancer cells, and its inactivation potentiates the response to chemotherapeutic treatment in mouse models of breast cancer. Finally, we show that neoadjuvant chemotherapy treatment leads to LCN2 upregulation in residual human breast tumors, and correlates with worse overall survival. These findings provide the foundation for targeting LCN2 as an adjuvant therapeutic approach to prevent the emergence of aggressive tumors following chemotherapy.


INTRODUCTION
Breast cancer affects more than one in ten women worldwide [1]. Currently, neoadjuvant chemotherapy is extensively used to treat breast cancer patients as it reduces tumor burden, thus downstaging the disease [2]. However, most non-targeted anticancer agents do not only trigger cytotoxicity in dividing cells, but also engage specific cellular response programs, including senescence, on both cancer cells and the tumor microenvironment [3].
Cellular senescence refers to the stable cell proliferation arrest caused by either telomere shortening, oncogene activation or genotoxic stress, all of which converge towards the activation of a sustained DNA damage response (DDR) [4]. Because of its engagement in preneoplastic lesions, senescence was originally hypothesized to serve as a barrier to malignant transformation, by preventing the proliferation of cells harboring an altered genetic content [5]. Additionally, senescent cells accumulate over time in mammals and contribute to the health defects associated with aging [6,7]. The beneficial impact of senescence as a tumor suppressor mechanism early in life along with its detrimental impact on aging phenotypes led to the theory of antagonistic pleiotropy of senescence [8]. One of the hallmarks of senescence that could rationalize these otherwise contradictory features is the senescence-associated secretory phenotype (SASP) [9].
The SASP consists of a discrete set of pro-inflammatory cytokines, chemokines and growth factors secreted by senescent cells, in a cellular and senescence inducer-specific manner. As such, the SASP may contribute to the "inflammaging", a sterile inflammation that develops as individuals age. In addition to its potential impact on aging phenotypes, exposure to the SASP has been reported to promote aggressive traits in tumor models, including increased cellular proliferation [10], enhanced angiogenesis [11], and activation of the epithelial-to-mesenchymal transition (EMT) [12]. However, the molecular mechanisms engaged by senescent cells to drive tumorigenesis remain poorly understood.
Here, we demonstrate that exposure to the SASP or senescenceinducing neoadjuvant chemotherapy results in the potent upregulation of LCN2 expression in breast cancer cells in vitro and in human cancer samples, respectively, which correlates with increased cellular plasticity and poor prognosis.

RESULTS
The IL-1-dependent SASP promotes cellular plasticity in breast cancer cells We have previously demonstrated that inactivation of the IL-1 pathway can be used to uncouple SASP production from senescence-associated cell cycle exit [19]. We leveraged this property of IL-1α-inactivated cells to determine the impact of SASP exposure on cancer cells' properties. MCF7 breast cancer cells were exposed to conditioned media (CM) from wild-type (WT) TERTimmortalized IMR90 (IMR90T) fibroblasts, WT IMR90Ts rendered senescent through ectopic expression of oncogenic Ras G12V , and senescent IL-1α −/− IMR90Ts (Fig. 1A). MCF7 cells exposed to CM from senescent WT cells (Ras CM) migrated significantly faster than cells exposed to CM from growing cells in a scratch assay (Fig. 1B,  C). Strikingly, CM collected from senescent IL-1α −/− cells was unable to promote cancer cell migration (Fig. 1B, C). These observations suggest that exposure to the IL-1α-dependent SASP is sufficient to stimulate breast cancer cells' migration. To determine whether SASP-induced increased migration correlates with enhanced chemotactic capacities, MCF7 cells cultured with CM for 2 days were allowed to migrate for 48 h in a transwell assay. MCF7 cells cultured with Ras CM displayed increased transwell migration compared to cells cultured with growing CM or cells cultured with CM from IL-1α −/− senescent cells (Fig. 1D). Consistent with our previous demonstration that SASP production is dependent on the IL-1α/IL-1R axis, chemotactic-based migration was significantly reduced when CM was obtained from senescent IMR90Ts depleted for IL-1R (Fig. 1E). Furthermore, exposure to senescent CM resulted in morphology changes in MCF7 cells, which then adopted a fibroblast-like appearance, a feature of EMT [20]. By contrast, MCF7 cells treated with CM from senescent IL-1α −/− cells maintained their cobblestone-like morphology and strong cell-cell adhesions (Fig.  1F). Accordingly, MCF7 cells exposed to senescent CM exhibited loss of expression of the epithelial marker E-cadherin (Fig. 1G). Finally, the proportion of MCF7 cells expressing the EMT-associated surface marker CD44 [21] increased upon exposure to SASP (Fig.  1H). Taken together, these results indicate that exposure to the IL-1dependent SASP induces cellular plasticity in breast cancer cells, as evidenced by increased migratory properties and the engagement of an at least a partial EMT program.
Exposure to the SASP induces expression of Lipocalin-2 in breast cancer cells To begin to decipher the molecular mechanisms underlying the impact of exposure to the SASP on breast cancer cells' properties, we profiled the transcriptome of MCF7 cells exposed to CM from growing or senescent IMR90Ts. Using a log 2 fold change cutoff 1 and an adjusted p value of <0.05, 1981 genes were found differentially expressed between MCF7 cells exposed to growing CM and MCF7 exposed to senescent CM. Pathways upregulated in senescent CM samples include inflammatory response and extracellular matrix organization ( Fig. 2A). Gene sets that were significantly enriched in senescent CM compared to growing CM samples included Epithelial-to-Mesenchymal Transition and Protein Secretion (Fig. 2B). These results are consistent with the previous demonstration that exposure to the SASP activates an inflammatory response and the initiation of an EMT program [22,23]. The most upregulated transcript in MCF7 cells exposed to senescent CM encodes the protein Lipocalin-2 (LCN2, or NGAL) (Fig. 2C). We confirmed the upregulation of LCN2 mRNA and protein levels via qRT-PCR (Fig. 2D) and Western Blot (Fig. 2E). LCN2 was not upregulated in MCF7 cells treated with CM from senescent IL-1α −/− cells (Fig. 2D, E). Exposure to SASP from cells rendered senescent by etoposide (Etop) treatment resulted in a significant increase in LCN2  mRNA levels, indicating that LCN2 upregulation was independent of the stimulus used to induce senescence (Fig. 2F). Of note, the LCN2 upregulation induced by exposure to genotoxic stressinduced SASP also correlated with increased migration (Fig. 2G). We extrapolated these observations to independent breast cancer cells with various ER, PR and HER2 status, and detected a consistent upregulation of LCN2 upon exposure to the SASP (Fig. 2H). In these conditions, these cells also exhibited an elongated and mesenchymal-like morphology (Fig. S1). In addition, we observed an increase in the rate of migration of MDA-MB-231 and T47D cells exposed to the SASP for 2 days ( Fig. 2I and Fig. S2). Taken all together, these results indicate that the SASP secreted from cells induced to senesce by various stimuli promotes migration and LCN2 upregulation in several breast cancer cell lines.

LCN2 upregulation is required for SASP-induced cell plasticity
High expression of LCN2 results in increased migratory properties in cancer cells [24,25]. However, the molecular basis for LCN2 upregulation remain unexplored. On account of the substantial LCN2 upregulation we detected in cells treated with senescent CM, we hypothesized that the SASP may enhance aggressive breast cancer phenotypes at least in part through upregulation of LCN2. We first successfully inactivated LCN2 in MCF7 cells by CRISPR/Cas9 induced gene deletion, as evidenced by Western Blot analysis (Fig. 3A, B). In agreement with the undetectable to low levels of LCN2 expressed in MCF7 cells grown in normal conditions, LCN2 −/− MCF7 cells did not exhibit any proliferation defects (data not shown). Strikingly, scratch assays indicated that the SASP-induced increase in migration in MCF7 was largely dependent on the presence of LCN2. Indeed, exposure to the SASP had no noticeable impact on the ability of LCN2 −/− MCF7 to close the gap left by the scratch even after 3 days (Fig. 3C, D). We also tested the migration capabilities of LCN2 −/− cells via transwell assay, and consistent with the scratch assay results, the SASPdependent increase in chemotactic migration was impaired upon LCN2 inactivation (Fig. 3E). To determine the transcriptional programs engaged by LCN2 in breast cancer cells exposed to the SASP, transcriptome analysis was performed on SASP-treated LCN2 +/+ and LCN2 −/− MCF7 cells. Using a log2 fold change cutoff 1 and an adjusted p value < 0.05, we confirmed LCN2 as one of the most differentially expressed genes (Fig. 3F). Gene Ontology analysis indicated that pathways upregulated in LCN2 +/+ samples include Epithelial-to-Mesenchymal transition and E2F targets (Fig.  3G). Similarly, differentially enriched gene sets by GSEA in wild type versus LCN2 −/− cells, included "Epithelial-to-Mesenchymal Transition", as well as "MYC targets" and "TNFα Signaling via NF-κB". Notably, these pathways have been associated with a loss of cell identity and induction of plasticity in breast cancer [26][27][28]. Collectively, these results indicate that LCN2 is required for SASPinduced plasticity of breast cancer cells.

SASP-induced LCN2 expression promotes breast cancer progression in vivo
We next sought to assess the impact of LCN2 expression on tumor progression in vivo. We injected Luciferase-expressing MDA-MB-231 cells into mammary fat pads of nude mice and followed tumor progression. Importantly, LCN2 status had no impact on MDA-MB-231 tumor growth (Fig. 4A, B). This observation is consistent with the undetectable to low LCN2 expression levels in these cells in the absence of a prosenescence stimulus (Fig. 2H). Importantly, Western Blot analysis revealed that SASP-dependent upregulation of LCN2 levels was only transient, as removal of conditioned media from senescent fibroblasts resulted in a strong downregulation of LCN2 (Fig. 4C). Therefore, to ensure continuous SASP exposure and upregulation of LCN2 in tumors, we opted for a co-injection model with senescent human fibroblasts. MDA-MB-231 cells exhibited increased LCN2-dependent proliferation and tumor progression when co-injected with senescent fibroblasts (Fig. 4D-F). IHC revealed that LCN2 +/+ tumors expressed reduced levels of E-cadherin and increased Ki67, compared to their LCN2 −/− counterparts. Apoptosis could not account for the difference in tumor size between experimental groups since LCN2 −/− tumors were negative for cleaved-caspase 3 (Fig. 4G). These results indicate that SASP-dependent LCN2 upregulation promotes tumor cell plasticity in vivo and results in the development of more aggressive tumors that are highly proliferative.

SASP-induced LCN2 protects breast cancer cells from genotoxic stress
Tumor cell plasticity and EMT are closely associated with therapy resistance in breast cancer [29][30][31]. Based on the phenotypes elicited by LCN2 −/− breast cancer cells in vivo, we hypothesized that SASP-mediated LCN2 upregulation could confer breast cancer cells a chemo-protective phenotype. Indeed, cell viability and Annexin V assays indicated that LCN2 +/+ MDA-MB-231 cells exposed to SASP were more resistant to doxorubicin than LCN2 +/+ MDA-MD-231 cells exposed to normal medium. However, the increased resistance to doxorubicin upon exposure to the SASP was abrogated in LCN2 −/− MDA-MB-231 cells (Fig. 5A, B). To validate these findings in vivo, we injected MDA-MB-231 cells into mammary fat pads of nude mice. Once tumors were established, mice were treated with doxorubicin. While LCN2 +/+ and LCN2 −/− tumors grew at a comparable rate prior to doxorubicin treatment, doxorubicin injection resulted in a dramatic sensitization of LCN2 −/− tumors (Fig. 5C, D). One week after doxorubicin treatment, LCN2 +/+ tumors were significantly larger than their LCN2 −/− counterparts, indicating that resistance to doxorubicin is enhanced by SASP-induced LCN2 upregulation. Together, these results suggest that the exposure to senescence-inducing stimuli promotes detrimental cellular plasticity in breast tumors through LCN2 expression.
LCN2 expression is induced following chemotherapy and is a poor prognostic factor in breast cancer patients To assess the clinical relevance of the previous observations, we analyzed LCN2 levels in biopsy samples from individual breast cancer patients collected prior to or following neoadjuvant chemotherapy treatment. Upon chemotherapy treatment, we detected an increased positivity for the SASP marker IL-6 ( Fig. 6A, B). Along IL-6 upregulation, samples collected after neoadjuvant chemotherapy treatment exhibited increased LCN2 positivity, while all biopsy samples collected before chemotherapy treatment displayed undetectable or low LCN2 expression (Fig. 6A, C). An opposite pattern of expression was observed for the epithelial marker, E-cadherin (Fig. 6A, D). We further analyzed the correlations between LCN2 levels and prognosis in breast cancer patients using publicly available expression databases. The analysis revealed that LCN2 was upregulated at the mRNA level in 125 of 2,507 patient samples (7%). Patients with increased levels of LCN2 have an inferior overall (data not shown) and relapse-free survival compared to those with unaltered levels of LCN2 (Fig. 6E). Additionally, 52.8% of patients with high LCN2 levels had received chemotherapy treatment prior to analysis, while only 18.5% of the patients with low levels of LCN2 had (Fig. 6F), further indicating a correlation between chemotherapy treatment and increased LCN2 levels. These data suggest that LCN2 could be a potential prognostic biomarker for breast cancer survival.

DISCUSSION
Cellular plasticity promotes tumor progression in breast cancer, through metastatic spread and resistance to conventional therapies. The molecular pathways that contribute to the plasticity of breast cancer cells, including EMT and dedifferentiation, are currently being uncovered. However, the stimuli leading to the acquisition of a cellular plastic phenotype within an established tumor remain largely unknown. We report here that exposure to the SASP induces breast cancer cell plasticity in an LCN2dependent manner.
Our results are consistent with previous findings showing that breast cancer cells exposed to the SASP exhibit epithelial cell scattering and reduced cell-cell adhesions, features of EMT, and increased expression of stemness markers [32,33]. Strikingly, we demonstrate that LCN2 upregulation is critical for the engagement of an EMT-associated expression program in breast cancer cells upon SASP exposure. Furthermore, while we cannot conclude that all breast cancer cells are responsive to SASP-induced LCN2increased migration, our results indicate that breast cancer cells of different molecular subtypes exhibit these properties. Such observation suggests that a large proportion of breast tumors could acquire more aggressive traits when surrounded by senescent cells. In the context of breast cancer, the source of the SASP can be several-fold: replicatively exhausted cells that accumulate in the organism with age secrete SASP factors that can reach tumor cells. Another source of the SASP may stem from precancerous or cancerous cells themselves with constitutively active drivers of mitogenic signals, resulting in hyper-proliferation and fork collapse, known as oncogene-induced senescence (OIS). Finally, exposure to genotoxic therapies, including radiation therapy or chemotherapy, can drive both normal and tumor cells into senescence, in a process termed Therapy-Induced Senescence (TIS), leading to general and local SASP production. TIS has emerged as a novel functional target to improve cancer therapy [34]. However, accumulating evidence indicates that senescent cancer cells are capable of reentering the cell cycle and promoting tumor relapse and metastasis. Recent findings show that a senescence-like population of chemotherapy-resilient cells is capable of initiating cancer recurrence by increasing their stemness potential [35][36][37].
Previous studies have linked LCN2 and EMT or metastasis of breast cancer cells [15,38], but the mechanisms involved remain unclear. Downregulation of the estrogen receptor ERα induces expression of the transcription factor Slug, driving EMT. Our transcriptome analyses indicate that in wild-type MCF7 cells, ERα is indeed downregulated and Slug is upregulated upon treatment with senescent CM compared to those treated with growing CM (data not shown). Studies have demonstrated that ERα signaling helps maintain the epithelial phenotype through inhibition of Snail activity [39]. However, based on our demonstration that SASP-induced LCN2 promotes migration in ER-breast cancer cells, it is unlikely that LCN2 mediates its effects on breast cancer plasticity through the ERα pathway exclusively. Accordingly, our transcriptomic analyses revealed that LCN2 levels show no correlation with ERα expression. Instead, LCN2 could drive EMT through its interaction with MMP9, reducing E-cadherin expression levels on the cell surface [40]. Further experiments will be necessary to elucidate the mechanism employed by LCN2 to promote EMT in breast cancer cells.
Lipocalin-2 (LCN2) was first characterized as an iron-binding protein, sequestering it from Gram-negative bacteria and inhibiting their proliferation. LCN2 iron-binding sequestration properties have also been implicated in the etiology of Leptomeningeal Metastasis (LM) [41]. Indeed, LCN2 upregulation allows metastatic cancer cells to thrive in an iron-limiting environment such as the cerebrospinal fluid. Accordingly, the administration of iron chelators slows cancer progression in a mouse model of LM [41]. Iron facilitates tumorigenesis by driving cell proliferation [42]. Therefore, LCN2 upregulation could result in an increased ability to sequester iron leading to an aggressive tumor growth in breast cancer cells [43].
The experiments presented and conclusions drawn here focus on the cancer cell autonomous impact of SASP-induced LCN2 upregulation. However, based on the pleiotropic impact of iron metabolism, it is also likely that LCN2 modulates the anti-tumor immune response in vivo. Indeed, LCN2 has the ability to upregulate human leukocyte antigen G (HLA-G), which can promote tumor immune escape in mouse models [44]. Furthermore, by conferring cancer cells with the ability to outcompete macrophages for iron, LCN2 may also contribute to the generation of a tumor microenvironment that promotes tumor growth, a possibility that remains to be investigated. For these reasons, therapeutic inhibition of LCN2 could provide a therapeutic relief for patients with breast tumors that are poised to undergo plasticity, for example following exposure to genotoxic therapies.

Senescence induction and condition media (CM) harvest
For Ras-induced senescence, IMR90T-Ras G12V -ERT2 were treated with 200 nM tamoxifen (Sigma) continuously for 10 days. Fresh media and tamoxifen were added every 2-3 days. Cells were treated with an equal volume of ethanol as control. On day 8 of Ras induction, media was replaced with serum-free media and harvested after 48 h. For etoposideinduced senescence, cells were treated with 50 μM etoposide (Sigma) or an equal volume of DMSO as a control for 48 h. Etoposide-containing medium was then replaced by normal culture media for 5 days. 7 days after etoposide treatment, media was replaced with serum-free media and harvested after 48 h. Conditioned media was aliquoted and flash frozen in liquid nitrogen before storing at −80°C.

Scratch assays
Cells were grown in six-well plates until confluent. A P200 tip was used to create vertical scratches. Media was then changed to media containing 1% FBS supplemented by CM from growing or senescent cells. The data are presented as relative gap width compared to the gap width of each sample at day 0. Values were subjected to two-way ANOVA followed by Tukey's multiple comparisons test. Data are presented as means ± SEM (n = 3).

Transwell migration assays
MCF7 cells were exposed to 1-3 mL CM from growing or senescent cells. Fresh CM was added every day for a total of 2 days. A total of 50,000 cells in 100 μL serum-free DMEM were seeded on top of transwells containing 8 μm pores for use in 24-well plates. One milliliter of DMEM supplemented with 20% FBS was added to the bottom of the wells and cells were allowed to migrate for 48 h. Values are expressed as fold change in the number of cells compared to the number of cells cultured in the corresponding growing CM. Values were subjected to two-way ANOVA followed by Tukey's multiple comparisons test. Data are presented as means ± SEM (n = 3).

Immunofluorescence
MCF7 cells were treated with CM from growing or senescent cells. CM was added every day for 2 days. Cells were then fixed and incubated with mouse anti-E-cadherin (Millipore, MAB1199) at 1:200 dilution in blocking solution at 37°C for 1 h. Cells were then washed and incubated in Cy3-conjugated donkey anti-mouse IgG (Jackson Immunoresearch) for 1 h at RT. Cells were mounted with mounting medium containing Dapi (Vectashield). Slides were examined on a Zeiss AxioImager A2 microscope. A total of 100 cells per cover slip were counted. Amount E-cadherin staining was quantified as number of red pixels per Dapi-positive cells using Image J to calculate pixels. Values were subjected to a paired t test. Data are presented as means ± SEM (n = 3).

Transcriptomics analysis
MCF7 cells were cultured in CM from growing or Ras-induced senescent cells for 2 days. MCF7 LCN2 +/+ and LCN2 −/− cells were cultured in CM from Ras-induced senescent cells for 2 days. RNA quality assessment, library preparation and sequencing were performed by the NYU School of Medicine Genome Technology Center or by Genewiz. Strand-specific libraries were prepared using the TruSeq RNA library Prep kit, and libraries were sequenced on an Illumina HiSeq2500 using 50-bp paired-end reads. Sequences were mapped to the hg10 genome, and analysis was done as previously described (Proudhon, 2016).

Bioluminescence
For in vivo luminescence of Luciferase, mice were injected i.p. with of 150 mg of D-Luciferin (ThermoFisher) per kg of body weight. Fifteen minutes later, the mice were anesthetized with isoflurane and luminescence was measured with a PerkinElmer IVIS Spectrum system. Analysis of the tumor size was performed in a single-blind manner.
Cell viability assay 400,000 cells were plated in triplicate in 12-well plates and allowed to adhere overnight. Cells were then treated with increasing concentrations of doxorubicin and CM for 24 h. Wells were washed with PBS and cells were fixed with 2% glutaraldehyde in PBS for 15 min. Cells were then stained with crystal violet (0.1% in 10% ethanol) for 30 minutes. After washing and drying, cells were distained in 10% acetic acid for 15 min. Optical density (OD) was measured at 595 nm absorbance. Values were subjected to two-way ANOVA followed by Tukey's multiple comparisons test. Data is presented as means ± SEM (n = 3).

Annexin V staining
Treated and untreated cells were collected, without washing to collect all floating cells, centrifuged at 1300 rpm for 3 min. After discarding supernatant, 5 μL of Annexin V (BioLegend) was added to cells resuspended in 200 μL of binding buffer (BioLegend). Cells were then incubated for 30 min at room temperature in the dark before centrifuging them again. Cells were resuspended in 200 μL of binding buffer after discarding the supernatant and analyzed via flow cytometry using a FACSCalibur Flow Cytometer (BD). Values were subjected to two-way ANOVA followed by Tukey's multiple comparisons test. Data are presented as means ± SEM (n = 3).

Animal studies
The number of animals needed to achieve statistical power was calculated based on a two-sided Wilcoxon nonparametric test with a significance level of 5%. No animal was excluded from the study. Mice were randomly allocated to experimental and control groups.

Study approval
Animal work and human subject work were performed accordingly to approved protocols from the NYU School of Medicine's IACUC and IRB. Written informed consent was obtained from each human participant before study procedure.

DATA AVAILABILITY
Raw transcriptomic data were deposited at GEO and are available under accession numbers GSE198661 and GSE198685. Additional data that support the findings of this study are available in figshare with the identifier: https://doi.org/10.6084/ m9.figshare.20017631.