Prune-1 drives polarization of tumor-associated macrophages (TAMs) within the lung metastatic niche in triple-negative breast cancer

Summary M2-tumor-associated macrophages (M2-TAMs) in the tumor microenvironment represent a prognostic indicator for poor outcome in triple-negative breast cancer (TNBC). Here we show that Prune-1 overexpression in human TNBC patients has positive correlation to lung metastasis and infiltrating M2-TAMs. Thus, we demonstrate that Prune-1 promotes lung metastasis in a genetically engineered mouse model of metastatic TNBC augmenting M2-polarization of TAMs within the tumor microenvironment. Thus, this occurs through TGF-β enhancement, IL-17F secretion, and extracellular vesicle protein content modulation. We also find murine inactivating gene variants in human TNBC patient cohorts that are involved in activation of the innate immune response, cell adhesion, apoptotic pathways, and DNA repair. Altogether, we indicate that the overexpression of Prune-1, IL-10, COL4A1, ILR1, and PDGFB, together with inactivating mutations of PDE9A, CD244, Sirpb1b, SV140, Iqca1, and PIP5K1B genes, might represent a route of metastatic lung dissemination that need future prognostic validations.


Introduction
Triple-negative breast cancer (TNBC) accounts for 20% of breast cancers (BCs) (Neophytou et al., 2018), where the tumorigenic cells are negative for the estrogen receptor (ER), progesterone receptor (PgR), and human epidermal growth factor receptor 2 (HER2; i.e., ER -, PgR -, HER2 -) (Rakha and Chan, 2011). TNBC is the most aggressive subtype of BCs due to its aggressive clinicopathological features, including young age at onset, large tumor size (Rakha and Chan, 2011), and greater propensity for visceral metastasis to distant sites (Lin et al., 2012). Among the metastatic patients diagnosed with TNBC, 49.3% develop metastasis in the lung (Xiao et al., 2018). Due to the absence of recognized molecular targets for therapy, TNBC patients with lung metastasis have the poorest outcome compared with those diagnosed with other metastatic BC subtypes (Xiao et al., 2018;Cancer Genome Atlas, 2012).
Due to its molecular features, TNBC can be considered one of the most complex tumor disorders in humans. Thus, the application of ''-omics'' technologies is of importance to monitor the genomic evolution of TNBC, which shows dominant TP53, PIK3CA, and PTEN somatic mutations (Shah et al., 2012). Indeed, six distinct molecular TNBC entities have been described: two basal-like-related subgroups (basal-like 1 [BL1] and 2 [BL2]), two mesenchymal-related subgroups (mesenchymal [M], mesenchymal stem-like [MSL]), one luminal androgen receptor (LAR) group, and one immunomodulatory (IM) subgroup, with MSL and IM subtypes that are driven by tumor-associated stromal cells and tumor-infiltrating lymphocytes (TILs), respectively, in the tumor microenvironment (TME) (Lehmann et al., 2011(Lehmann et al., , 2016. Thus, the immune cells that infiltrate the TME have a dual function in tumor development and metastatic progression. In the early stages of tumorigenesis, they identify and eliminate tumorigenic cells because of the expression of tumor-specific antigens (a mechanism known as ''tumor immune surveillance''). Later, the tumors undergo immune-editing processes that allow the tumor variants with reduced immunogenicity to escape immunological detection and elimination, thus allowing tumor variants to be clonally selected (Spano and Zollo, 2012).
The greater genomic instability and the mutational burden of TNBC result in higher propensity to generate neoantigens (Spano and Zollo, 2012;Bianchini et al., 2016), thus generating selected tumorigenic clones within the TME that considerably influence the risk of response to chemotherapy and relapse in these patients (Dieci et al., 2014). Recently, new distinct TNBC subtypes were defined based on the types of immune infiltrating cells within the TME: i.e., the pro-tumorigenic M2-polarized tumor-associated macrophages (M2-TAMs) and the immunosuppressive regulatory T cells (Tregs) (Adams et al., 2017). The TNBC subtypes with the poorest prognosis are characterized by higher levels of infiltrating M2-TAMs (CD163 + ) and Tregs (FOXP3 + ) in a low TILs environment, with high levels of the immunosuppressive programmed death ligand-1 (PD-L1) and glycolytic monocarboxylate transporter 4 (MCT4) markers (Adams et al., 2017). Thus, clinical evidence indicates that M2-TAMs infiltrating tumor tissues act as a prognostic indicator for poor outcome for patients with TNBC, as they are correlated with higher risk for developing metastasis (Sousa et al., 2015;Yuan et al., 2014), as M2-TAMs exert their immunosuppressive functions through inhibition of the effector functions of TILs (Santoni et al., 2017).
It is worth noting the valuable functions of cytokines withing the TME, which are soluble factors that mediate the communication between tumorigenic and immune cells. Transforming growth factor b (TGF-b) is a master regulator of an immunosuppressive action, and it is produced by both tumorigenic and immune cells and can thus orchestrate the polarization processes of the immune system toward a pro-tumorigenic phenotype and promote cancer progression and metastasis, also in TNBC (Pickup et al., 2013). In addition, extracellular vesicles (EVs; including exosomes) generated by tumor cells can modulate the recruitment of the immune cells and their immunomodulation in the TME (Thery et al., 2002), with a crucial role in promoting organotropic pre-metastatic niche formation in TNBC (Hoshino et al., 2015;Chow et al., 2014).
Through this approach we have identified in TNBC a common mutation burden in tumorigenic cells and a related network of action on immunoregulation of the immunosystem, with a focus on macrophages.  Figure 1. Prune-1 protein is overexpressed in TNBC and promotes macrophage chemotaxis through STAT3 via soluble cytokines (A) Prune-1 overexpression in TNBC. RNA log2 expression analysis of Prune-1 levels of primary BC samples across different publicly available datasets, compared with normal epithelium (N Epithelium; Shelharmer dataset only). Data from 10 independent public-domain BC gene-expression datasets show the overexpression of Prune-1 in all of the BC samples compared with normal epithelium. Higher Prune-1 expression levels are seen for TNBC samples (i.e., Brown (Burstein et al., 2015); red dashed line) (n = 1779; p = 3.0 3 10 À169 ). (B) Overexpression of Prune-1 in TNBC cells enhances macrophage chemotaxis in vitro. Normalized Cell Index as a measure of cell migration/chemotaxis of J774A.1 (i.e., J774; upper panel) and RAW264.7 (i.e., RAW264; lower panel) macrophages as generated by the xCELLigence RTCA software. Migration kinetics were monitored in response to conditioned media from Prune-1-overexpressing (4T1-Prune-1, red), Prune-1-silenced (4T1-Sh-Prune-1, green), and empty vector 4T1 cell clones (black), used as chemoattractants. Dulbecco's modified Eagle's medium was used as the negative control (blue). (C-E) Conditioned media (CM) from the 4T1 clones were collected after 24 h. J774 or Raw264 macrophages were starved for 6 h in serum-free medium and then grown in the conditioned media for 30 min (C). Densitometer analyses of immunoblotting for the indicated proteins in J774 (D) and Raw264 (E) macrophages grown for 30 min in conditioned media from Prune-1-silenced and control 4T1 clones are shown. Empty vector (EV) 4T1 clones and untreated (UNT) macrophages Here, through applying data mining approaches in public gene expression resources available in BC datasets (R2 Genomics Analysis and Visualisation Platform; http://r2.amc.nl), we found higher Prune-1 expression in all BC samples (n = 1779, p = 3.0 3 10 À169 ; Figure 1A), thus confirming Prune-1 overexpression in BC. Interestingly, we found the highest expression of Prune-1 in the public TNBC dataset (i.e., Brown (Burstein et al., 2015), n = 198; Figures 1A and S1A, within the red dashed line), which thus suggested a role for Prune-1 in this highly metastatic BC subgroup. These data are not surprising, because of the major frequency of chromosome 1q21 gain in basal-like and TNBC (30%-40% (Silva et al., 2015), (Cancer Genome Atlas, 2012)) and also in those with recurrent BC (70% (Goh et al., 2017)).
Triple-negative breast cancer is known to correlate to negative expression of ERs and PgR status (i.e., ER -/ PgR -/HER2 -) (Bianchini et al., 2016). To provide deeper insight into the role of Prune-1 in TNBC, its association with ER, PgR, and HER2 status was also investigated using the public accessible dataset of tumor breast invasive carcinoma, with gene expression data acquired for the BC cohort from The Cancer Genome Atlas (TCGA; n = 1,097). In this dataset, BC samples were stratified according to their ER, PgR, and HER2 scores (as evaluated by immunohistochemistry [IHC]), which ranged from 0 to 3 + , with 0 indicating negativity. Higher expression levels of Prune-1 were identified only in those samples with negative scores for both ER and PgR (i.e.,0), and in those with scores ranging from 0 to 1 + for HER2 status (Figure S1B, within the red dashed lines). These data prompted us to suggest potential involvement of Prune-1 in the pathogenesis of TNBC.
We have previously defined Prune-1 as an inducer of the SMAD-mediated canonical TGF-b pathway in metastatic medulloblastoma group 3 (MB group3 ) . As the TGF-b cascade has a crucial role in tumor metastasis, we investigated the expression levels of its downstream effectors (i.e., SMAD2, SMAD4) in BCs using different public BC datasets. These analyses showed that both SMAD2 and SMAD4 expression levels were higher in the TNBC dataset (i.e., Brown (Burstein et al., 2015); as shown in Figure S1C; SMAD2: p = 5.6 3 10 À230 ; SMAD4: p = 2.1 3 10 À86 ), thus indicating correlation between Prune-1 and canonical TGFb signaling effectors in TNBC (Drabsch and Ten Dijke, 2012). Further correlation analyses were performed between Prune-1, SMAD2, and SMAD4 in BC samples stratified according to their ERs and PgR status, using additional gene-expression data acquired from the publicly accessible cohort of BC samples in TCGA (n = 1,097; Table S1). This analysis showed that Prune-1 positively correlated with both SMAD2 and SMAD4 levels in the BC samples with PgR, ER, and HER2 negative status (R value, 0.19-0.35 ; Table S1). Overall, Prune-1 positively correlated with the TGF-b downstream effectors in TNBC. This is of great interest for the immunoregulatory action of TGF-b exerted in the TME. Using a murine model of metastatic TNBC cells (i.e., 4T1 cells (Yoneda et al., 2000)), independent stable clones were generated: Prune-1-silenced (0.6-fold, p = 0.02; Figure S2A) and h-Prune-1-FLAG-overexpressing ( Figure S2B) 4T1 clones. We asked whether any Prune-1-induced perturbation can modulate TGF-b signaling pathways. As shown in Figure S2C, decreased levels of phosphorylated-(Ser467)-SMAD2 (i.e., phospho-SMAD2, the main effector of the canonical TGF-b cascade) were shown in Prune-1-silenced Figure 1. Continued were used as the negative controls. b-Actin levels were used as the loading control. *p < 0.05 in Student's t test compared with J774 (D) or Raw264 (E) treated with conditioned media from 4T1 EV control clones. (F) Prune-1 induces the secretion of soluble proteins by TNBC cells. Densitometer analyses of the cytokines upregulated and downregulated in the conditioned media (CM) derived from Prune-1-overexpressing (4T1-Prune-1) and Prune-1silenced (4T1-Sh-Prune-1) 4T1 clones (MultiExperiment Viewer, http://www.tm4.org/mev.html). Among the 17 cytokines modulated by Prune-1 in the conditioned media collected from Prune-1-overexpressing (4T1-Prune-1), one was upregulated (CD30) and five were downregulated (Rantes, Galectin-1, IL-17F,  in the conditioned media from 4T1-Sh-Prune-1 cell clones, thus following an opposite trend. *p < 0.05 in Student's t test comparing cytokines levels in conditioned media of 4T1-Sh-Prune-1 cells with those in 4T1 Empty Vector control clones.

Prune-1 at the interplay of communication between TNBC cells and macrophages
Tumorigenic and immune cells within the TME communicate through extracellular mediators (e.g., cytokines, EVs, exosomes), which are also sensors in the modulation of immune cells (Spano and Zollo, 2012). We previously reported that Prune-1 has an extracellular role in paracrine communication via Wnt3a cytokine secretion (Carotenuto et al., 2014). As clinical evidence indicates that M2-TAMs positively correlate with metastasis and poor outcome in TNBC (Sousa et al., 2015;Yuan et al., 2014), we here determined the potential involvement of Prune-1 in the intratumoral recruitment of M2-TAMs in TNBC.
In this regard, we evaluated the recruitment/migration of immune cells using murine J774A.1 (#ATCC-TIB-67 (Lam et al., 2009)) and Raw264.7 (ATCC-TIB-71) macrophages, by performing a real-time cell motility assays (Cell Index) in which conditioned media from 4T1-Prune-1 cell clones (as previously described) were used as chemoattractants. As shown in Figure 1B, the conditioned media from Prune-1-overexpressing 4T1 cell clones increased the migration rates of both J774A.1 and Raw264.7 macrophages ( Figure 1B, red lines), whereas the media from Prune-1-silenced 4T1 cell clones reduced their migration rate ( Figure 1B, green lines). These results were compared with those from EV control clones ( Figure 1B, black lines).
Furthermore, the conditioned media from 4T1 cell clones were also used to grow macrophages in vitro, to determine the activation status of macrophages by measuring the phosphorylation status of the STAT3 protein (i.e., pY705-STAT3 is required for dimerization and nuclear translocation (Levy and Darnell, 2002), whereas pS727-STAT3 is linked to increased STAT3 transactivation (Decker and Kovarik, 2000)), due to its prominent role in macrophage activation and during M1-M2 switch in the polarization processe . As shown in Figures 1C-1E, S2E, and S2F, pY705-STAT3 and pS727-STAT3 were significantly decreased in the J774A.1 (pY705-STAT3: 0.6-fold, p = 0.05; pS727-STAT3: 0.6-fold, p = 0.04; Figures  1D and S2E) and Raw264.7 (pY705-STAT3: 0.27-fold, p = 0.01; pS727-STAT3: 0.56-fold, p = 0.006; Figures 1E and S2F) macrophages cultured in media from Prune-1-silenced 4T1 clones, compared with those cultured in media from the control clones (EV). In contrast, we did not find any significant downregulation of total STAT3 in both J774A.1 p = 0.17,p = 0.18,Figures 1E and S2F) macrophages. However, the conditioned media collected from the Prune-1-silenced 4T1 cell clone do not completely abolish phosphorylated STAT3 protein levels in the recipient macrophages S2E,and S2F). We believe that this is because of the incomplete knockdown of m-Prune-1 in the 4T1 cell clone (0.6-fold downregulation, p = 0.02, see Figure S2A). However, it is not possible to exclude the presence of activating factors of STAT3 that are independent of Prune-1 actions.
For the above reasons, we believe that this finding might represent one of the mechanisms in the downregulation of the STAT3 pathway in these macrophages grown in the conditioned media collected from Prune-1-silenced tumorigenic cells.

Secretion of inflammatory cytokines from TNBC cells are modulated by Prune-1
We have here investigated whether Prune-1 can influence the recruitment/activation of macrophages by modulation of the secretion of extracellular mediators (i.e., cytokines, chemokines). For this purpose, we compared the levels of 144 cytokines secreted in the conditioned media collected from three different ll OPEN ACCESS iScience 24, 101938, January 22, 2021 5 iScience Article independently generated Prune-1-overexpressing 4T1 cell clones (pooled together) and compared them to those secreted by the EV control clones ( Figure S3A, upper panel). These analyses showed 17 cytokines that were differentially secreted (14 upregulated >2-fold; three downregulated <0.5-fold; see fold intensity in Figure S3B). In particular, the levels of CD30, lungkine/CXCL15, and E-cadherin were significantly decreased in the pooled conditioned media from Prune-1-overexpressing 4T1 clones, compared with those from the control clones. In contrast, the levels of LIX/CXCL5, thymus CK-1/CXCL7, thymic stromal lymphopoietin (TSLP), 6Ckine, ALK1, amphiregulin (Areg), CD36, CD40 ligand/CD154, galectin 1, IL-17F, IL-28, IL-20, JAM-A/F11R, and ''regulated upon activation, normal T cell expressed and secreted'' (RANTES or CCL5) were increased in the pooled conditioned media derived from Prune-1-overexpressing 4T1 clones, compared with conditioned media from the control clones ( Figure S3B). To further confirm these data, we determined the levels of the selected Prune-1-modulated cytokines (i.e., n = 17) in the conditioned media from Prune-1-silenced 4T1 cell clones ( Figure S3A, bottom panel). These data showed that among these 17 cytokines analyzed, E-cadherin, RANTES, galectin 1, IL-17F, IL-20, and IL-28 (also known as type III interferon-l) levels were significantly decreased ( Figures 1F and S3C). In contrast, CD30 levels increased in the conditioned media collected from Prune-1-silenced 4T1 cells, thus following an opposite trend compared with those in conditioned media from Prune-1-overexpressing 4T1 cell clones ( Figure 1F). These findings indicated that secretion of the RANTES, galectin 1, IL-17F, IL-20, and IL-28 cytokines might be positively modulated by Prune-1 expression in TNBC cells. Gene Ontology analyses indicated that these cytokines are involved in immune system process (GO:0002376, fdr:0.0046), immune response (GO:0002376, fdr: 0.0138), immune effector processes (GO:0002252, fdr: 0.0302), and innate immune responses (GO:0045087, fdr: 0.0473), thus suggesting involvement in the regulation of immune cells within the TME. Of importance, among these cytokines, IL-20 and IL-28 were reported to take part to JAK-STAT signaling pathway activation (KEGG: mmu04630, fdr: 0.004), which is involved in M1-or M2-macrophage polarization . Of interest, through its binding to IL-17RA on recipient endothelial cells, the IL17A/IL17F heterodimer was previously reported to activate the STAT3 pathway (mainly via Y705 phosphorylation) and consequently to recruit immune cells (Yuan et al., 2015).
Altogether, we have described here that overexpression of Prune-1 in 4T1 murine TNBC cells might have a role in modulation of macrophage activation through the extracellular secretion of soluble cytokines (i.e., RANTES, galectin1, IL17-F, IL-20, IL-28). We measured the ''activation'' of macrophages by evaluating their migration rate and STAT3 pathway , because its activation in TAMs has been shown to induce immunosuppression, angiogenesis, cell growth, and metastasis (Yu et al., 2007;Dang et al., 2015). Furthermore, STAT3 inhibition was also reported to ''re-educate'' TAMs via reversing their phenotype from M2 to M1 (Fujiwara et al., 2014;Zhang et al., 2013). How Prune-1 influences the expression of these cytokines in tumorigenic cells and the activation of STAT3 in macrophages will be the focus of other studies.
To determine whether the overexpression of h-Prune-1 has similar immune responses (in terms of cytokines expression) to m-Prune1, we transiently transfected h-Prune-1 or m-Prune-1 cDNA plasmid constructs into Raw264 macrophages, because of their lower Prune-1 endogenous expression levels compared with J774 macrophages ( Figure S4B). Using real-time PCR analysis, we showed the same levels of upregulation of inflammatory cytokines (i.e., IL-10, Arg1, MMP9, IL-1b) in both h-Prune-1-and m-Prune-1-overexpressing Raw264 macrophages ( Figures S4C and S4D). These data show the potential involvement of the Prune-1 protein in the modulation of cytokines expression also in immune cells (i.e., macrophages), and they also show no significant differences between the functional regulation of the human versus murine Prune-1 proteins. For these reasons and to characterize the effects of Prune-1 overexpression in the TME of BC, we generated a GEMM-overexpressing human Prune-1 in the mammary gland.
This GEMM was generated using a vector construct that contained h-Prune-1 cDNAs under the control of the MMTV promoter (Callahan and Smith, 2000) ( Figure S4E). Female mice harboring h-PRUNE-1 cDNA overexpressed in mammary glands (i.e., MMTV-Prune-1) developed mammary hyperplasia early in life (by 80 days of age), compared with the control FVB mice, as shown in Figure S4F. These MMTV-Prune-1 mice were monitored for 12 months. Although benign mammary lesions that were usually hyperplasia ll OPEN ACCESS  (Li et al., 2000;Yu et al., 2016;Herschkowitz et al., 2007), which rarely develops spontaneous metastasis to the lungs . The resulting double transgenic MMTV-Prune-1/Wnt1 mouse model (n = 31) developed mammary tumors (with 100% penetrance), as did the MMTV-Wnt1 mice (n = 44). Interestingly, the overexpression of both Prune-1 and Wnt1 in the mammary glands did not alter mammary tumor onset, compared with the MMTV-Wnt1 mice ( Figure S4H). Furthermore, through IHC analysis, we confirmed that the mammary tumors that were generated in the double transgenic MMTV-Prune-1/Wnt1 mice resemble the TNBC subgroup with undetectable levels of both ERs and PgRs ( Figure S5A), thus confirming these models as GEMM of TNBC. Nevertheless, although MMTV-Wnt1 mice did not develop lung metastasis at 2 months from tumor onset, MMTV-Prune-1/Wnt1 mice showed macro-metastasis in the lungs at 97% penetrance ( Figure 2B, Table S2).
In addition, we investigated the presence of immune infiltrating cells in the mammary TME. Among the immune infiltrating cells, we focused on M2-TAMs (CD163 + ) and Tregs (FOXP3 + ), which have been reported as markers for the subtype of TNBC with poorest prognosis (Adams et al., 2017). As shown in Figure 2C, our quantitative IHC and immunofluorescence (IF) analyses (using a quantitative pathology approach; see Methods) revealed increased levels of both M2-TAMs (i.e., CD68 + CD163 + cells) and Tregs (i.e., FOXP3 + CD4 + cells) in the TME of MMTV-Prune-1/Wnt1 mice compared with MMTV-Wnt1 (CD68: p = 2.05 3 10 À9 ; CD162: p = 4.36 3 10 À7 ; CD4: p = 1.0 3 10 À3 ; FOXP3: p = 7.0 3 10 À3 ). Of note, M2-TAMs and Tregs were also found in the metastatic niche into the lungs of the MMTV-Prune-1/Wnt1 mice ( Figure 2C). Of interest, we performed the same IHC and IF analyses on the contralateral nontumoral mammary glands of MMTV-Prune-1/Wnt1 and MMTV-Wnt1 mice, and on the lungs from MMTV-Wnt1 mice ( Figure S5B), where immune infiltrating cells were not found. Altogether these data indicate a role for Prune-1 in combination with other factors in the recruitment/activation of immunosuppressive cells in the TME of both the primary tumor and lung metastatic microenvironments in TNBC.
To better underpin the role of Prune-1 in metastatic TNBC, we also investigated the function of Prune-1 in the activation of TGF-b signaling and induction of epithelial-mesenchymal transition (EMT) .
Here, higher SMAD2/3 levels were detected in primary mammary tumors derived from the double transgenic mice (i.e., MMTV-Prune-1/Wnt1), compared with those from MMTV-Wnt1 models ( Figures S6A and S6B). Furthermore, undetectable E-cadherin and higher N-cadherin levels were found in the tumors from MMTV-Prune-1/Wnt1 mice compared with those from MMTV-Wnt1 mice, which in contrast showed higher E-cadherin and lower N-cadherin levels ( Figures S6A and S6B). Of interest, the microenvironment of the nontumoral contralateral mammary gland also shows higher SMAD2/3 and N-cadherin and lower E-cadherin levels in MMTV-Prune-1/Wnt1 compared with those from MMTV-Wnt1 models ( Figures S6C and S6D). These data thus confirm that Prune-1 enhances TGF-b and EMT also in this GEMM of metastatic TNBC.
(D) Prune-1 induces secretion of soluble cytokines in vivo. Densitometer analyses of the cytokines upregulated and downregulated in the sera collected and pooled from three MMTV-Prune-1/Wnt1 mice and MMTV-Wnt1 mice. Among the cytokines modulated by Prune-1, significant upregulation of IL-17F, IL-28, and IL-20 was seen, with an opposite trend compared with the cytokines in the conditioned media from Prune-  Figures 2D, S7A, and S7B. These cytokines were previously shown to be overexpressed in conditioned media collected from Prune-1-overexpressing 4T1 cells and downregulated in those derived from Prune-1-silenced 4T1 clones ( Figure 1F), thus translating the in-vitro findings into the in-vivo results.
Prune-1 activates metastatic pathways and enhances the migratory phenotype in murine TNBC primary cells To dissect out the function of Prune-1 in TNBC, primary murine tumorigenic cells were obtained from the tumors generated from MMTV-Prune-1/Wnt1 and MMTV-Wnt1 mice (at 2 months from the tumor onset; Figures S8A and S8B). Here we show the activation of both canonical TGF-b and Wnt signaling in MMTV-Prune-1/Wnt1 cells compared with MMTV-Wnt1 cells, as shown by increased levels of phospho-Ser467-SMAD2, phospho-Ser9/21-GSK-3b, and Wnt3a ( Figure S8C). Importantly, the same analysis also showed increased levels of EMT markers in MMTV-Prune-1/Wnt1 cells (i.e., undetectable E-cadherin, higher N-cadherin levels), increased phosphorylation levels of phospho-(Ser-473)-AKT, and decreased expression of its repressor PTEN, compared with MMTV-Wnt1 cells ( Figure S8C). These data also showed increased levels of phosphorylated Ser120-122-125-NME1 (or NDPK-A; sign of complex formation with the Prune-1 protein) in these primary cells  ( Figure S8C). Altogether, these results indicate the activation of Prune-1-metastatic pathway (as Prune-1 in complex formation with NME1; as previously described for medulloblastoma ) also in these murine TNBC primary cells (i.e., MMTV-Prune-1/Wnt1 cells), thus overall suggesting increased migratory properties of these primary Prune-1-overexpressing TNBC cells.
Overall these data indicate that Prune-1 can enhance the proliferative and migratory properties also in these primary tumorigenic cells derived from mammary tumors generated from our GEMM.

Macrophage polarization is enhanced by Prune-1 overexpression in TNBC
Whether Prune-1 has a role in the polarization process of macrophages in these primary murine TNBC cells was also investigated. For this purpose, conditioned media from MMTV-Prune-1/Wnt1 and MMTV-Wnt1 primary cells (collected over 24 h) were used to grow J774A.1 and Raw264 macrophages in vitro ( Figure 3A), to evaluate their ''polarization status'' by measuring the expression levels of ''M2-associated genes'' (Orecchioni et al., 2019) through whole-genome RNA sequencing approaches (i.e., RNAseq; Figures 3B and 3C; see Additional Data). Untreated macrophages were used as the negative control. These data showed upregulation (i.e., fold change on untreated macrophages >2, p < 0.01) of 24 genes ( Figure 3B, Table S3) and 27 genes ( Figure 3C, Table S3) belonging to ''M2-associated genes'' in the J774A.1 and Raw264 macrophages, respectively, grown in the conditioned media collected from MMTV-Prune-1/Wnt1 cells, as compared with those grown in MMTV-Wnt1-derived media. In contrast, only one gene (i.e., Emp1) was upregulated (i.e., fold change over untreated macrophages >2; p < 0.01) in J774A.1 cells treated with conditioned media from MMTV-Wnt1 cells ( Figure 3B). This thus indicated that Prune-1 takes part in the polarization processes. Of interest, among those upregulated genes in the macrophages grown in the conditioned media from MMTV-Prune-1/Wnt1 cells, 11 were common between J774A.1 and Raw264 macrophages ( Figures 3B and 3C, black boxes). These included the most representative M2-associated genes (i.e., MMP-9, arginase-1 [Arg1], IL-10 (Mantovani et al., 2002)) and IL-1b, the intracellular accumulation of which was reported during the switch from M1-status toward M2-status of macrophages (Martinez et al., 2006;Pelegrin and Surprenant, 2009). Moreover, our data also showed higher phospho-STAT3 levels in J774A.1 macrophages grown in the conditioned media from the MMTV-Prune-1/Wnt1 cells, compared with those treated with the conditioned media from the MMTV-Wnt1 cells ( Figure S9A). These data suggest that in a ''Prune-1-overexpression status'' in tumorigenic cells, modulation of the M1-M2 switch of macrophages occurs.
In contrast, J774A.1 macrophages were grown in the conditioned media derived from MMTV-Wnt1 cells or MMTV-Wnt1 cells overexpressing IL-17F ( Figures 3G and S9C). These data showed that J774A.1 macrophages expressed higher levels of M2-associated genes (i.e., Arg1, MMP9, IL-10, IL-1b) only when they were grown in conditioned media from MMTV-Wnt1 cells overexpressing IL17F, as compared with those grown in conditioned media from MMTV-Wnt1 cells or untreated macrophages ( Figure 3H). These data indicated a potential role for this inflammatory cytokine (i.e., IL17F) in the polarization process of immune cells.
Altogether, these data suggested that M2-polarization of TAMs (i.e., IL-10 High , Arg1 High , MMP9 High , IL-1b High ) in metastatic TNBC is evident in a Prune-1-overexpression status. Indeed, as shown here, Prune-1 takes part in the extracellular communication between tumorigenic and immune cells through secretion of IL-17F. Details on how this occurs will be the focus of future studies.

Prune-1 induces macrophage polarization via extracellular vesicles
Tumor-derived EVs (including exosomes) have begun to emerge as new factors in tumor progression and organotropic metastatic dissemination. These have been reported to act via several mechanisms, including modulation of the antitumor immune response in the TME (Thery et al., 2009).
As Prune-1 induces M2-macrophage recruitment in the TME and promotes lung metastasis in the present GEMM of metastatic TNBC, we investigated whether EVs derived from MMTV-Prune-1/Wnt1 cells also contribute to these processes. To define a global picture of Prune-1-driven EV-proteins, a proteomic analysis was performed for those EVs secreted by both MMTV-Prune-1/Wnt1 and MMTV-Wnt1 cells. For this purpose, EVs were isolated (as previously described ) from media derived from both MMTV-Prune-1/Wnt1 and MMTV-Wnt1 cells. Subsequently, to define potential changes in EV-protein content in Prune-1-overexpressing cells, the extracellular proteins were analyzed using label-free quantitative mass spectrometry technology (as described in Supplemental Information (Cox et al., 2014)). These analyses showed that the EVs isolated from the media from both of these murine primary TNBC cells shared 31 proteins in common, whereas those derived from the MMTV-Prune-1/Wnt1 cells and the MMTV-Wnt1 cells had 21 and 10 mutually exclusive proteins, respectively (as summarized in Figures 4A and S9D). Among the extracellular proteins secreted by the MMTV-Prune-1/Wnt1 cells, we identified some extracellular proteins linked to EMT, motility, and metastasis in BC, including vimentin (Vim; (Yamashita et al., 2013)), interferon-induced transmembrane protein 3 (Ifitm3; (Yang et al., 2013a)), and syndecan-binding protein (syntenin-1/Sdcbp (Yang et al., 2013b;Qian et al., 2013;Koo et al., 2002)) ( Table 1). These data suggested that Prune-1 promotes distant metastasis in TNBC also by modulation of EV-protein content.
Then, to investigate the potential role for these extracellular proteins secreted via EVs from MMTV-Prune-1/Wnt1 cells in the modulation of M2-associated genes (i.e., IL-10, MMP9, Arg1, IL-1b), J774A.1 macrophages were grown in conditioned media from MMTV-Prune-1/Wnt1 cells depleted of EVs (Figures 4B and S9E). These data showed reduction of IL-10, IL-1b, and Arg1 expression levels in the macrophages grown in the conditioned media without EVs, compared with those treated with complete supernatant (Figure 4C), whereas the expression levels of MMP9 were increased. This thus suggested different mechanisms of extracellular regulation of MMP9 in our model system. In contrast, the depletion of EVs from the media from MMTV-Wnt1 cells did not change the expression levels of M2-associated genes (i.e., MMP9, Arg1, IL-1b) in J774A.1 macrophages, with the exception of IL-10 ( Figures 4D and 4E). The in-vitro data further To confirm these findings, we evaluated the polarization genes in J774A.1 macrophages grown in conditioned media derived from MMTV-Wnt1 cells that had been previously pre-treated with the supernatant of MMTV-Prune-1/Wnt1 cells containing or depleted in EVs ( Figure S9F). These data showed increased  . Prune-1 overexpressed in TNBC cells promotes M2-polarization of macrophages in vitro through modulation of EV-protein content (A) Representative scheme for proteomic analyses performed on extracellular vesicles (EVs). EVs were isolated from media from murine primary MMTV-Prune-1/Wnt1 and MMTV-Wnt1 cells. Proteomic analyses were performed on isolated EVs using label-free quantitative mass spectrometry. Data show 31 extracellular proteins in common between EVs from media of MMTV-Prune-1/Wnt1 and MMTV-Wnt cells, and 21 and 10 mutually exclusive extracellular proteins from MMTV-Prune-1-Wnt cells (red) and MMTV-Wnt1 cells (green). Data are representative of two independent experiments. (B and C) J774 macrophages were grown for 48 h in conditioned media collected (after 24 h) from MMTV-Prune-1/Wnt1 cells depleted or not in EVs from the culture supernatant (B). Real-time PCR analysis of some M2-associated genes, including IL-10, Arg-1, MMP-9, and IL-1b, in J774 macrophages grown for 48 h in conditioned media from MMTV-Prune-1/ Wnt1 depleted (red) or not (light red) in EVs (C). *p < 0.05 in Student's t test compared with macrophages treated with conditioned media from MMTV-Prune-1/Wnt1 cells not depleted in EVs. (D and E) J774 macrophages were grown for 48 h in conditioned media collected (after 24 h) from MMTV-Wnt1 cells depleted or not in EVs from the culture supernatant (D). Real-time PCR analysis of some M2-associated genes, including IL-10, Arg-1, MMP-9, and IL-1b, in J774 macrophages grown for 48 h in conditioned media from MMTV-Wnt1 depleted (light green) or not (dark green) in EVs (E). *p < 0.05 in Student's t test compared with macrophages treated with conditioned media from MMTV-Wnt1 cells not depleted in EVs. iScience Article levels of IL-10, Arg1, and IL-1b in J774A.1 macrophages when the MMTV-Wnt1 cells were grown in EV-containing media derived from MMTV-Prune-1/Wnt1 cells ( Figure S9G).

ll
In summary, Prune-1 is involved in extracellular mechanisms of communication between TNBC cells and immune cells (i.e., macrophages) not only through modulation of soluble mediators (e.g., IL-17F) but also through modulation of the EV-protein content.

Genetics, RNA expression, and mutational rate in the TNBC microenvironment
To define a more global picture of the mechanism of communication between Prune-1-overexpressing TNBC cells and macrophages, we analyzed the mutational rates (through whole-exome sequencing [WES] analyses) in MMTV-Prune-1/Wnt1 cells and the global trascriptome (via RNAseq) of macrophages that received conditioned media from these tumorigenic cells ( Figure 5A).
For this purpose, we took into account all of the common genes there were overexpressed in both J774A.1 and Raw264.7 macrophages that had been grown in the media obtained from MMTV-Prune-1/Wnt1 cells (as compared with MMTV-Wnt1 cells). Then, to identify a subset of ''core genes'' upregulated in macrophages in response to Prune-1 overexpression in TNBC cells, we compared the lists of genes from the leading edge of enriched gene-sets from each canonical pathway sub-collection (i.e., Biocarta, Kegg, Pid, Reactome, Naba) and selected those shared by at least four out of five of them (Table 2). A protein interaction network was generated using the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database (confidence: 0.4; https://string-db.org/cgi/network.pl?taskId=41jaTsLzWNqT). This network identified proteins involved in immune system processes ( . Of interest, among these protein interaction networks derived from the ''core genes'' identified here, the expression levels of COL4A1, IL-10, IL1R1, and PDGFB were positively correlated with poor prognosis in BC patients in terms of 5-year survival data analyzed from the publicity available dataset of Breast Invasive Carcinoma (n = 1,075) from The Cancer Genome Atlas (TCGA) ( Figure S10).
Then, we profiled the mutation spectra in our primary murine TNBC cells that overexpressed Prune-1. WES analyses were applied to both MMTV-Prune-1/Wnt1 cells and MMTV-Wnt1 cells. We found mutually exclusive coding variants with predicted higher/moderate impact on MMTV-Prune-1-Wnt1 cells (177 variants in  Table 3; Figure S11A). Among these, 39 gene variants were also found in the public data in the Catalog Of Somatic Mutations In Cancer (COSMIC, v91; released April 7, 2020) of human basal TNBC (426 samples collected) (see Table 3). Gene Ontology analyses performed on these human genes showed some deleterious variants involved in the activation of innate immune responses (leukocyte and macrophage activation; ANKHD1, FER1L5), cell adhesion (NEXN), apoptotic pathways (BID), and DNA repair (ERCC5) ( Figure 5C). Then, among these 39 genes that were mutated in the human basal TNBC dataset available in COSMIC, we focused on those with unfavorable prognosis in BC patients in terms of their expression levels and 5-year survival. For this purpose, we took into account the analyses of survival data obtained from the publicly available dataset of breast invasive carcinoma (n = 1,075) from TCGA. Of interest, low expression levels of the PDE9A, ERCC5, Iqca, Sirpb1b, CD244, SP140, and PIP5k1b genes were associated with decreased 5-year survival rate in BC patients, thus suggesting their potential association with unfavorable prognosis ( Figure S11B). The Rrs1 and A2M genes were excluded from this analysis because only one patient out of 426 had mutations in this gene in the public data of human basal TNBC in COSMIC (v91).
Of interest, all of the genes listed earlier (with the exception for ERCC5) were also mutated in the publicly available dataset of metastatic BC (n = 216 (Lefebvre et al., 2016)), with a total frequency of 3.7% (Cbioportal for cancer genomics; https://www.cbioportal.org) ( Figure S11C), thus indicating a potential role in metastatic processes in BC.

Dataset validation in a human cohort of TNBC patients positively correlated with M2-TAMs and distant metastasis
To further underpin these data, Prune-1 protein expression was analyzed in human TNBC by IHC on tissue micro-arrays using a collection cohort of primary TNBC specimens. In this cohort analysis, 138 TNBC samples were included (113 ductal, 25 not ductal). The patients age ranged from 24 to 93 years, with mean age 57 years. At surgery, tumors >2 cm were seen for 53% of these patients (72/136; for two patients this information was not available), and metastatic lymph nodes were seen for 40.2% of patients (55/137; for one patient this information was not available). The tumor gradings were as follows: grade 3, 87.7% (121/138) and grade 1 or 2, 12.3% (17/138). The expression of the proliferation marker Ki67 was high (>20%) in 80.6% of patients (108/134) and low (%20%) in 19.4% of patients (26/134; for four patients this information was not available; see Table 4).
Prune-1 protein expression was detected in 89.9% (124/138) of the samples from this tissue cohort. In 50.7% of the samples (68/138) there was low Prune-1 expression, and in 49.3% of the samples (70/138) there was high Prune-1 expression ( Figure 5D, a, b). These findings indicated that about 50% (i.e., 49.3%) of the TNBC samples in this tissue cohort showed overexpression of Prune-1 protein.
Statistical analysis of Prune-1 protein expression (i.e., based on intensity, percent expression) in terms of the other clinicopathological parameters in this TNBC tissue cohort indicated that Prune-1 was positively correlated with Ki67 proliferative index (p = 0.011; R = 0.219), high-grade (p = 0.005; R = 0.237), disease progression (p = 0.031; R = 0.212) and presence of lung metastasis (p = 0.027; R = 0.254), as shown in Table   Figure 5. Mutational spectrum in TNBC cells overexpressing Prune-1 regulating M2-macrophages polarization (A) Representative scheme for the experimental design. DNA from MMTV-Prune-1/Wnt1 and MMTV-Wnt1 cells was used for next-generation sequencing analyses through a whole-exome sequencing approach. J774A.1 and Raw264.7 macrophages were grown in conditioned media collected from MMTV-Prune-1/Wnt1 and MMTV-Wnt1 cells for 48 h. Total RNA was extracted from these macrophages, and RNAseq analyses were performed. (B) The inflammatory protein network generated via Search Tool for Retrieval of Interacting Genes/Proteins (STRING) database using the ''core genes'' defined as the common genes that are overexpressed in both the J774A.1 and RAW264.7 macrophages grown in media obtained from MMTV-Prune-1/ Wnt1 cells (as compared with MMTV-Wnt1 cells) shared by at least four of five enriched gene sets from each canonical pathway sub-collection (i.e., Biocarta, Kegg, Pid, Reactome, Naba). The protein interaction network was generated using the STRING database (confidence: 0.4; https://string-db.org/cgi/ network.pl?taskId=41jaTsLzWNqT).

CSF2
Colony-stimulating factor 2 (granulocyte-macrophage) Granulocyte-macrophage colony-stimulating factor; cytokine that stimulates the growth and differentiation of hematopoietic precursor cells from various lineages, including granulocytes, macrophages, eosinophils, and erythrocytes; belongs to the GM-CSF family.

CSF3
Colony-stimulating factor 3 (granulocyte) Granulocyte colony-stimulating factor; granulocyte/macrophage colony-stimulating factors are cytokines that act in hematopoiesis by controlling the production, differentiation, and function of 2 related white cell populations of the blood: the granulocytes and the monocytes-macrophages. This CSF induces granulocytes; belongs to the IL-6 superfamily.

FN1
Fibronectin 1 Fibronectin type III domain containing; endogenous ligands IL-10 Interleukin-10 Interleukin-10; inhibits the synthesis of a number of cytokines, including IFN-gamma, IL-2, IL-3, TNF, and GM-CSF produced by activated macrophages and by helper T cells; belongs to the IL-10 family

IL12B
Interleukin-12b Interleukin-12 subunit beta; cytokine that can act as a growth factor for activated T and NK cells, enhance the lytic activity of NK/lymphokine-activated killer cells, and stimulate the production of IFN-gamma by resting PBMC; belongs to the type I cytokine receptor family. Type 3 subfamily.

IL12RB1
Interleukin-12 receptor, beta 1 Interleukin-12 receptor subunit beta-1; functions as an interleukin receptor that binds interleukin-12 with low affinity and is involved in IL-12 transduction. Associated with IL12RB2 it forms a functional, high-affinity receptor for IL-12. Associates also with IL23R to form the interleukin-23 receptor, which functions in IL-23 signal transduction probably through activation of the Jak-Stat signaling cascade; CD molecules IL1A Interleukin-1 alpha Interleukin-1 alpha; produced by activated macrophages, IL-1 stimulates thymocyte proliferation by inducing IL-2 release, B-cell maturation and proliferation, and fibroblast growth factor activity. IL-1 proteins are involved in the inflammatory response, being identified as endogenous pyrogens, and are reported to stimulate the release of prostaglandin and collagenase from synovial cells.

IL1B
Interleukin-1 beta Interleukin-1 beta; potent proinflammatory cytokine. Initially discovered as the major endogenous pyrogen, induces prostaglandin synthesis, neutrophil influx and activation, T cell activation and cytokine production, B-cell activation and antibody production, and fibroblast proliferation and collagen production. Promotes Th17 differentiation of T cells.
(Continued on next page) ll OPEN ACCESS iScience 24, 101938, January 22, 2021 iScience Article 4. No significant correlations were found between Prune-1 levels and patients age, lymph node metastases, and tumor histotype and size (Table 4). Altogether, these data further supported the concept that overexpression of Prune-1 can be used to predict lung metastases in TNBC.

IL1RN
Interleukin-1 receptor antagonist Interleukin-1 receptor antagonist protein; inhibits the activity of interleukin-1 by binding to receptor IL1R1 and preventing its association with the coreceptor IL1RAP for signaling. Has no interleukin-1-like activity.
Binds functional interleukin-1 receptor IL1R1 with greater affinity than decoy receptor IL1R2; however, the physiological relevance of the latter association is unsure; endogenous ligands. Upregulates the expression of protein phosphatase 1 (PP1).
The common genes there are found overexpressed in both J774A.1 and RAW264.7 macrophages grown in the media supernatant from MMTV-Prune-1/Wnt1 cells (as compared with those from MMTV-Wnt-1 cells) shared by at least 4 out of 5 enriched gene-sets coming from canonical pathway sub-collection (i.e., Biocarta, Kegg, Pid, Reactome and Naba) are listed. Gene, common gene name (HGNC); Description, gene description; Annotation, annotated using Sequence Ontology terms.
Furthermore, we investigated the potential relationships between Prune-1 and the tumorigenic signaling pathways that are responsible for the aggressive behavior of TNBC. For this purpose, we took into account the nuclear localization of the MAPK and NF-kB effectors, due to their previously reported expression and correlations with poor prognosis in TNBC (  The clinicopathological parameters of the patients from our TNBC cohort grouped according to their Prune-1 protein expression levels are shown (n. 138 patients). Prune-1 overexpression was found positively correlated with Ki67 proliferative index (p = 0.011; R = 0.219), high grade (p = 0.005; R = 0.237), disease progression (p = 0.031; R = 0.212), and the presence of lung metastasis (p = 0.027; R = 0.254). Any statistically correlations were found between Prune-1 levels and patients age, lymph node metastases, and tumor histotype and size. In bold the statistically significant p-values. LMN, lymph node metastases; G1, Grade 1 or low grade (sometimes also called well differentiated); G2, Grade 2 or intermediate/moderate grade (moderately differentiated); G3, Grade 3 or high grade (poorly differentiated); NA, not available. a Prune-1 protein expression is based on the intensity and percentage of expression. b Statistical significant association (i.e., p < 0.05).
Following these statistical associations, we investigated additional correlations between Prune-1 expression and immune-cell infiltration. We used immunosuppressive pro-tumorigenic M2-TAMs (i.e., CD68 + CD163 + cells) due to their already identified prognostic role in TNBC in the tumors with higher risk of metastatic dissemination (Sousa et al., 2015;Yuan et al., 2014). These correlation analyses of our dataset showed that patients with TNBC (n = 32) overexpressing Prune-1 were also characterized by higher numbers of tumor-infiltrating TAMs (i.e., CD68 + cells; Figure 5D, g-h; p = 0.014, R = 0.433; in Table 6) and with a trend to significant association with CD163+ cells (p = 0.07, R = 0.32; see Figure 5D, i-l), a marker of pro-tumorigenic M2-polarized TAMs (Spano and Zollo, 2012), and with FOXP3, a marker of Tregs (p = 0.08; see Figure 5D, m-n). Of interest, we did not find any strong positivity of inflammatory pathways (i.e., p65-NF-kB), p-ERK, or the presence of immune infiltrating cells (i.e., CD68, CD163, FOXP3) in the sections near tumors with high expression of Prune-1 ( Figure 5D, o-t). Altogether, these results indicate that Prune-1 is a potential biomarker related to TNBC subtypes with the poorest outcomes, as characterized by higher infiltrating M2-TAMs (CD163 + ) and Tregs (FOXP3 + ) (Adams et al., 2017). However, to define an immunologically ''cold'' or ''hot'' TME, this analysis is also supported by CD4, CD8, and PDL1 to evaluate the TILs environment and the levels of immunosuppressive markers (Adams et al., 2017). Here, we focused mainly on TAMs within the TME of TNBC due to their prognostic value (Sousa et al., 2015;Yuan et al., 2014).
In conclusion, the clinico-immunopathological correlation studies performed on our TNBC patient cohort suggest that overexpression of Prune-1 is significantly associated with the activation of proliferative and inflammatory pathways (i.e., MAPK, NF-kB), and recruitment of immunosuppressive cells with pro-tumorigenic functions (i.e., M2-TAMs) in the TME, in those patients with TNBC with high-grade tumorigenesis progression and lung metastases. At this time, we focused on these immune cell populations (i.e., TAMs) because of their negative prognostic significance, as previously reported in TNBC patients (Adams et al., 2017).

Pharmacological inhibition of Prune-1 through AA7.1 reduces metastatic foci in vivo via inhibition of M2 polarization of macrophages
To further address the role of Prune-1 in M2 polarization of macrophages, we used an anti-Prune-1 molecule (AA7.1) that has been previously shown to enhance Prune-1 degradation and to impair TGF-b signaling in metastatic medulloblastoma (Ferrucci et al., 2018). Here, the potential immunomodulatory activity exerted by AA7.1 was tested in primary murine TNBC cells. AA7.1-treated MMTV-Prune-1/Wnt1 cells showed decreased levels of Prune-1 protein ( Figure 6A, 50% reduction) and mRNA ( Figure 6B) levels, with a major action on degradation of h-Prune-1. Reduced levels of phospho-SMAD2 ( Figure 6A) and IL-17F ( Figure 6B) were also shown in the same AA7.1-treated cells.
These data indicate that AA7.1 can reduce the activation of TGF-b signaling and the secretion of this inflammatory cytokine that has previously been shown to be positively modulated by Prune-1.
This immunomodulatory action of AA7.1 was then studied by evaluating changes in the expression levels of M2-associated genes in macrophages (J774A.1) grown in conditioned media from AA7.1-treated MMTV-Prune-1/Wnt1 cells ( Figure 6C). Of importance, there were decreased levels of IL-10, Arg1, IL-1b, and MMP-9 in these J774A.1 macrophages grown in culture media derived from the AA7.1-treated cells compared with the vehicle control ( Figure 6D). To exclude any potential carry-over of AA7.1 in the tumor-cell-conditioned medium to the macrophages, we tested for AA7.1 downregulation of Prune-1 mRNA expression in J774A.1 macrophages. These macrophages treated with AA7.1 do not show downregulation of endogenous murine Prune-1 ( Figure 6E), thus indicating that the changes in the cytokine levels in the recipient J774A.1 macrophages was not due to any direct effects of AA7.1 on these immune cells, thus excluding potential carry-over effects (at least at the AA7.1 concentration used in the assay).
Of interest, there was also modulation of the extracellular protein content released from AA7.1-treated MMTV-Prune-1/Wnt1 cells. Here, there were reduced Vimentin protein levels for both the AA7.1-treated MMTV-Prune-1/Wnt1 cells and in the extracellular vesicles secreted from these treated cells ( Figure 6F).
Overall here, we have shown that AA7.1 can negatively modulate the cross-talk between TNBC and macrophages, thus affecting the activation of M2-associated genes via downregulation of Prune-1, inhibition of the TGF-b pathway, reduction of IL-17F in TNBC cells, and modulation of the extracellular protein content.
Next, to investigate whether M2-polarized macrophages can affect the migration rates of TNBC cells, we performed co-culture transwell migration assays using (1) J774A.1 macrophages previously polarized using conditioned media from MMTV-Prune-1/Wnt1 cells untreated or treated with AA7.1 and (2) MMTV-Wnt1 cells ( Figure 6G). Untreated macrophages were used as the negative control. This system allowed the migration rates of MMTV-Prune-1/Wnt1 cells (using only 2% FBS as chemoattractant) to be monitored in the presence of M2-polarized macrophages ( Figure 6H). These data showed increased migration rates only for MMTV-Prune-1/Wnt1 cells co-cultured with J774A.1 macrophages previously grown in conditioned media from untreated MMTV-Prune-1/Wnt1 cells ( Figure 6H). To this end, we show that the M2polarized macrophages can increase the motility of the TNBC cells, thus indicating that a bidirectional communication is in place between macrophages and tumorigenic cells.  iScience Article Furthermore, we also confirmed in vivo that both silencing and pharmacological inhibition (via AA7.1) of Prune-1 inhibited TNBC growth and M2-TAM polarization. For this, we used syngeneic orthotopic models with murine metastatic 4T1 TNBC cells. In the first trials, immunocompetent BALB/c mice were implanted in the mammary gland with Prune-1-silenced or EV control 4T1 clones that stably expressed the firefly luciferase gene (4T1-LUC) ( Figure S12A). These syngeneic orthotopic mice were imaged weekly for tumor growth, as evaluated using bioluminescence acquisition with an imaging system (IVIS 3D Illumina; Xenogen/Caliper). At 14 and 21 days from tumor implantation, there was significant decrease in tumor growth ( Figure S12B). Of importance, in the tumor tissues derived from the Sh-Prune-1-4T1 implanted mice there was a reduction of the number of infiltrating M2-TAMs (i.e., CD68+. CD163+ cells) in the TME (Figures S12C and S12D).
Similar results were obtained in vivo through pharmacological inhibition of Prune-1 using AA7.1. Female immunocompetent BALB/c mice were injected with 4T1-LUC cells. At 14 days from tumor implantation (i.e., once the tumors were established), the mice were grouped according to their bioluminescence values and injected intraperitoneally with 60 mg/kg AA7.1 daily or with phosphate-buffered saline (PBS) as the vehicle control ( Figure S12E). Tumor growth was monitored weekly by bioluminescence acquisition. These in-vivo data showed significant reduction of tumor growth in the AA7.1-treated mice at 35 and 42 days from tumor implantation ( Figure S12F). At the end of the experiments (i.e., 42 days from tumor implantation), the primary tumors were dissected out and embedded in paraffin for IHC analysis. These data showed significant reduction of CD163+ cells, but not CD68+ cells ( Figures S12G and S12H), thus suggesting inhibition of the M2-polarization switch of macrophages in the TME.
Finally, we also investigated the AA7.1 to reduction of metastatic foci in vivo. For this purpose, MMTV-Prune-1/Wnt1 cells were injected (via the tail vein) into immunocompetent syngeneic (strain FVB) mice. At 14 days from cell injection, the mice were grouped according to their weight and AA7.1 (60 mg/kg/ day, intraperitoneally) or PBS (as the vehicle negative control) were administered daily ( Figure 7A). At 14 days from the start of the treatment (i.e., 28 days from cell injection), for ex-vivo targeting of tumorigenic cells, the mice were injected with a fluorescent imaging probe (XenoLight RediJect 2-DG-750; PerkinElmer) that targets cells with high metabolic activity in terms of glucose uptake ( Figure 7A). Here, these ex-vivo data showed positive fluorescence signals in the lungs derived from all of the control mice (i.e., treated with PBS as vehicle). In contrast, tumorigenic cells were detected in the lungs using the ex-vivo imaging analysis following the AA7.1 treatment in only one of four mice (i.e., 25%), thus showing that AA7.1 can inhibit tumor cell homing to the lungs in vivo ( Figure 7B). Hematoxylin/eosin staining of sections of lung tissue also confirmed the reductions in the metastatic foci for the AA7.1-treated mice compared with the controls (Figures 7C and 7D).
Altogether, these data indicate that pharmacological inhibition of Prune-1 through AA7.1 can reduce M2polarization of macrophages and inhibit in vivo the TNBC cell homing to the lungs.

Discussion
Despite optimal systemic chemotherapy, metastatic TNBC remains with unmet medical needs due to a lack of targeted therapies (Rakha and Chan, 2011). Metastatic progression is the result of a complex network of communication between tumorigenic and immune cells in the TME (Spano and Zollo, 2012). TAMs, dendritic cells, T and B lymphocytes, and partially differentiated myeloid progenitors (i.e., myeloid-derived suppressor cells [MDSCs]) represent the major components of the TME in TNBC (Yu and Di, 2017), mainly due to the higher propensity of TNBC cells to generate ''neoantigens'' due to their genomic instability and mutational burden (Bianchini et al., 2016). Once recruited into the TME by tumor-secreted soluble mediators and exosomes, the immune cells contribute to promotion and maintenance of an immunosuppressive environment, which allows immune escape and as a consequence, enhances tumor and metastatic progression (Spano and Zollo, 2012). Due to the variegated nature of the TME in TNBC, new TNBC subtypes with prognostic significance have been defined (Adams et al., 2017). Among these newly identified Here, we focused on M2-TAMs; of importance, these cells were found in tumors that overexpressed Prune-1 ( Figure 5D).
Genetically engineered mouse models are a useful resource in the study of the metastatic behavior of TNBC cells under immunocompetent conditions. However, GEMMs that resemble the features of metastatic TNBC have not been described to date (Pfefferle et al., 2013). Here, we developed a GEMM of metastatic TNBC driven by Prune-1 (MMTV-Prune-1/Wnt1) as a useful resource for preclinical studies to determine the efficacy of immunotherapeutic agents for treatment of established metastatic disease. We show that Prune-1 contributes to the generation of an immunosuppressive environment in both the primary tumor and the lung metastatic niches, through recruitment and polarization of M2-TAMs ( Figure 2C).
Recently, in metastatic MB group3 , we identified Prune-1-driven intracellular signaling that is involved in the metastatic behavior linked to poor prognosis . This ''metastatic axis'' is guided by a protein complex formed by Prune-1 and NDPK-A (NME-1). In metastatic medulloblastoma, both of these genes were overexpressed. Here, in specific TNBC-based gene expression analyses we show higher expression levels of Prune-1 and SMAD-2/4 (downstream effectors of the TGF-b cascade). For these reasons, we confirmed the activation of TGF-b in vivo in mammary tumors generated in the GEMM of metastatic Prune-1-driven TNBC (i.e., MMTV-Prune-1/Wnt1 cells; Figure S6A).
Furthermore, we showed increased levels of NDPK-A (NME-1) protein phosphorylated on serine residues (i.e., Ser120, Ser122, Ser125) in primary murine Prune-1-overexpressing TNBC cells (i.e., MMTV-Prune-1/ Wnt1 cells; Figure S8C). Taken altogether, these data suggest that the Prune-1/NDPK-A protein complex might act as ''driver'' to promote metastatic dissemination also in TNBC, through enhancement of the TGFb cascade and downregulation of PTEN. Literature data based on NDPK-A supports the concept of its role as a suppressor of metastasis in BC (Yokdang et al., 2015;Steeg, 2003). Of interest, overexpression of Prune-1 was also shown in these tumors, which were also characterized by higher NDPK-A expression levels (Forus et al., 2001). We hypothesized that Prune-1 acts through the ''metastatic axis'' in TNBC, as already described for MB group3 . An additional observation comes from the detection of extracellular NDPK-A in sera from patients with BC with metastases, and positive correlation was shown between the extracellular levels of NDPK-A and tumor growth in xenograft mice models (Yokdang et al., 2015). Of importance, extracellular Prune-1 was also detected in sera from patients with early stages of non-small-cell lung cancer (Carotenuto et al., 2014). Future efforts will be aimed at investigations into the mechanisms of actions of both of these extracellular proteins (i.e., Prune-1, NDPK-A) in metastatic TNBC, with a view to their potential prognostic value.
One additional feature is linked to a common PTEN loss observed in TNBC. This was described as its overexpression in MMTV-Wnt1 mice was reported to occur prior to tumor onset but not before metastatic behavior (Li et al., 2001). However, these findings indicated that PTEN loss and activation of Wnt signaling were not sufficient to induce metastatic spread in TNBC. These data are in agreement with PTEN downregulation in tumors that overexpress Prune-1 that was recently described for MB group3 . This is also of importance due to the role of PI3K/AKT activation and PTEN loss in TNBC, as previously reported (Bianchini et al., 2016) and as observed here using gene-expression data from different public BC datasets, which showed that PTEN expression levels were lower in the TNBC dataset (i.e., Brown (Burstein et al., 2015)) than in other BC datasets (n = 1779, p = 7.3 3 10 À116 ; data not shown). Our findings show that the Prune-1-induced metastatic axis is maintained also in TNBC, and they also suggest that Prune-1 overexpression together with PTEN loss has prognostic value to identify those patients with metastatic potential in TNBC. . Continued (i.e., MMTV-Prune-1/Wnt1 cells) showed predicted deleterious variants in genes involved in activation of the innate immune response, apoptotic pathways, DNA repair, and cell adhesion. These gene variants were also found in patients with TNBC. In detail, in human, we identified deleterious variants for six genes that are mainly involved in the activation of the innate immune response. We also found upregulation of IL-10, COL4A1, ILR1, and PDGFB, the expression levels of which are negatively correlated with prognosis in patients with TNBC. Altogether, these actions induce recruitment of TAMs in the TNBC microenvironment and promote their polarization toward anti-inflammatory/ pro-tumorigenic M2-status, thus preparing the system for lung metastasis within the premetastatic niche.

OPEN ACCESS
iScience 24, 101938, January 22, 2021 37 iScience Article Transforming growth factor b has a crucial role in the TME through modulation of the polarization status of different immune cells, including TAMs (Pickup et al., 2013). Our in-vitro and in-vivo data show that Prune-1 enhances TAM recruitment and polarization toward a pro-tumorigenic M2 status, which in turn, and together with other factors, promotes STAT3 activation and increased expression levels of Arg1, MMP-9, IL-10, and IL-1b in recipient macrophages. As summarized in Figure 7E, we hypothesize here that Prune-1 acts in metastatic TNBC through activation of intracellular tumorigenic pathways (e.g., TGF-b), release of extracellular inflammatory soluble molecules (with a preferential action of IL-17F), and changes in EV protein content, from which of the importance of Vim and syntenin-1 arise, with a role in EMT and in the homing of tumorigenic cells to lung tissues .
However, TGF-b has also been reported to modulate the activation and status of the other components of the TME (including MDSCs, lymphocytes, neutrophils, dendritic cells, cancer-associated fibroblasts) (Pickup et al., 2013). For the above reasons, we cannot exclude the possibility that Prune-1 also affects other immune-infiltrating cells and stromal components in the TME of TNBC. This issue will be addressed in future studies.
In the present study, we also provide evidence that Prune-1 in TNBC activates ERK1/2-MAPK, through increased phosphorylation of ERK1/2 and NF-kB and through increased phosphorylation of p65 and its increased nuclear localization ( Figure 5D). Of importance, the question on how Prune-1 activates the ERK1/2-MAPK and NF-kB signaling pathways should be further addressed in the future. A possible mechanism might be related to Prune-1 regulation of the Wnt signaling pathway, through its interaction with GSK-3b, and to the complex network of interactions between these signaling pathways. Cross-talk with a positive-feedback loop between the Wnt and ERK pathways has been described in tumor cells (Kim et al., 2016). Moreover, the Wnt signaling pathway induces expression of S100A4, an important player in tumor progression and metastasis, which in turn positively activates the NF-kB signaling pathway (Stein et al., 2009;Boye et al., 2008). Therefore, it might act in an autocrine manner to promote NF-kB pathway activation in Prune-1-overexpressing TNBC cells.
Studies to impair TNBC have been recently reported. We have here demonstrated in vitro the pharmacological inhibition of Prune-1 in BC cells using a small molecule (AA7.1) , which affects the crosstalk between tumorigenic cells and macrophages, thus reducing the expression of M2-associated genes ( Figure 6D). We also show the activity of AA7.1 in vivo, with the demonstration of its potential to reduce M2-TAM recruitment in the TME and to inhibit lung metastatic processes ( Figures 7B and 7C) via modulation of the TGF-b pathway ( Figure 6A), IL17F expression levels ( Figure 6B), and exosomal protein content (i.e., Vim; Figure 6F). How the expression of Prune1 in macrophages influences their direct modulation and its inhibition by AA7.1 in time-dependent and dose-dependent manners will be an issue for future studies.
Altogether we show that AA7.1 has an important immunomodulatory property. Of note, we have previously shown that AA7.1 did not induce toxicity, through evaluation of the hematological, hepatic, and renal parameters in treated mice . Thus, future studies will be aimed at the testing of the pharmacological inhibition of Prune-1 using AA7.1, alone or in combination with current chemotherapy regimens and/or immunotherapeutics (e.g., checkpoint inhibitors) in TNBC.
Furthermore, we have shown increased levels of IL-1b in macrophages treated with conditioned media collected from Prune-1-overexpressing TNBC cells (Figures 3B and 3C). Indeed, high levels of intracellular IL-1b have been reported for macrophages polarized toward, but not at, the M2 phenotype (Pelegrin and Surprenant, 2009). Of interest, the intracellular accumulation of IL-1b in M2-polarizing macrophages is reported to be caused by extracellular ATP-derived pyrophosphates (Pelegrin and Surprenant, 2009). For the above reasons, Prune-1 enzymatic exopolyphosphatase/pyrophosphatase activities might also be involved in the mechanisms of macrophage polarization. By taking advantage of the use of specific inhibitors of Prune-1 enzymatic activities, future studies will address these hypotheses.
Moreover, and most importantly, our in-vitro and in-vivo data here also show that Prune-1 induces an immunosuppressive TME in TNBC by releasing extracellular soluble mediators (i.e., cytokines) and vesicle-containing proteins. Among the representative cytokines, IL-17F, IL-20, and IL-28 are targets of the Prune-1activated NF-kB signaling pathways (Otkjaer et al., 2007) (Shen et al., 2006;Osterlund et al., 2007). Once ll OPEN ACCESS 38 iScience 24, 101938, January 22, 2021 iScience Article secreted, these Prune-1-induced cytokines might act on the tumor cells in an autocrine fashion or on immune cells (e.g., macrophages) in a paracrine manner within the TME. Indeed, similar to IL-17A, IL-17F activates the NF-kB and ERK1/2-MAPK signaling pathways, to promote angiogenesis and lead to upregulation of several chemokines and cytokines, thus exacerbating the inflammatory TME (Lai et al., 2011).
In the context of EV-containing proteins, among those secreted by Prune-1-overexpressing cells, we found Vim, Ifitm3, and syntenin-1 that have been reported to have roles in TNBC. Of note, Vim contributes to the aggressive phenotype and poor prognosis in TNBC (Yamashita et al., 2013). Instead, Ifitm3 was shown to be overexpressed in invasive BC, with a function related to progression and motility of TNBC cells (i.e., MDA-MB-231 cells) (Yang et al., 2013a). Most importantly, syntenin-1 is an adaptor molecule that is involved in a variety of cellular processes, including metastasis. Indeed, high expression of syntenin-1 in BC primary tumors has been significantly related to patient overall survival and progression-free survival (Yang et al., 2013b), and it is known to be negatively correlated to ER expression (Qian et al., 2013). Furthermore, syntenin-1 was shown to be overexpressed in TNBC cells with an invasive/metastatic phenotype (i.e., MDA-MB-231 cells) (Koo et al., 2002) and to have a role in promoting cell migration and invasion both in vitro and in vivo, through activation of AKT (Hwangbo et al., 2011), integrin-a1 (Yang et al., 2013a(Yang et al., , 2013b, MAPK (Yang et al., 2013a(Yang et al., , 2013b, and TGF-b signaling and EMT (Menezes et al., 2016), definitively promoting tumor growth and lung metastasis. Interestingly, syntenin-1 enhances the canonical TGF-b pathway (i.e., mediated by SMAD activation) and EMT, which thus leads to metastatic spread (Hwangbo et al., 2016). Lastly, the expression of syntenin-1 was shown to promote lung metastasis by influencing the inflammatory network, with the induction of inflammatory cytokines (i.e., IL-17A, IL-6) and both inflammatory and immunosuppressive cells (i.e., Th17 cells and MDSCs, respectively) in the TME of metastatic melanoma . Overall, these EV-containing proteins represent an important finding that confirms the mechanism of communication between these TNBC cells and the immune cells within the TME.
Through WES analyses, we found mutually exclusive coding variants in the TNBC cells overexpressing Prune-1, with predicted higher/moderate impact. Among these variants, 39 gene variants were also found in the public database of human basal TNBC (COSMIC, v91) ( Figure 5C). These deleterious variants are involved in activation of the innate immune response (leukocyte and macrophage activation; ANKHD1, FER1L5), cell adhesion (NEXN), apoptotic pathways (BID), and DNA repair (ERCC5) ( Figure 5C). Of importance, lower expression levels of the PDE9A, Iqca1, Sirpb1b, CD244, SP140, and PIP5k1b genes was found in the BC patients with unfavorable prognosis in terms of 5-year survival analysis (dataset of Breast Invasive Carcinoma [n = 1075] from TCGA) ( Figure S11). The same genes are also mutated in metastatic BC patients (dataset of metastatic BC (n = 216 (Lefebvre et al., 2016)) ( Figure S11). Among these genes, CD244, Sirpb1b, and SV140 are involved in immune cell processes. CD244 has a crucial role in the activation of natural killer T cells (Mathew et al., 2009). It was defined as one of the markers of cytolytic activity in the TME of TNBC, together with the mutation status of TP53 mut and PIK3CA wt (Cheng et al., 2020), thus acting as a marker for response to immunotherapy. On the other hand, Sirpb1b (or CD172B) shows strong correlations with immune system pathways that positively modulate the production of pro-inflammatory cytokines, including interleukin family cytokines, TNF-a, and GM-CSF (Huang et al., 2013), and participates in neutrophil trans-epithelial migration (Ribeiro et al., 2019). Of interest, the transcriptional factor SV140 is a master regulator that acts as a modulator of the adaptive immune response in BC. It was shown to be downregulated in BC cells, related to a lower infiltration of immune cells into tumor tissues and inversely correlated to relapse-free survival in BC of the basal subtype (Da Silveira et al., 2017). Importantly, SP140 is inversely correlated to NF-kB and regulates genes involved in cytokine production, inflammatory response, and cell-cell adhesion (Karaky et al., 2018). Then, Iqca1 was shown to be among the significantly downregulated genes in TNBC as compared with normal ductal cells, by genome-wide gene expression profiling analysis (Komatsu et al., 2013). Interestingly, a higher number of mutations in the PIP5K1B gene was found in metastatic BC as compared with both invasive and noninvasive BC (Durand et al., 2018). Regarding PDE9A, which is a regulator of cGMP signaling also in BC biology, it is less expressed in BC cells (including TNBC) compared with normal human mammary epithelial cells (Tinsley et al., 2009). Of importance, PDE9A was identified as a germline-related prognostic gene for ER-negative BC, involved in controlling cell growth and angiogenesis (Escala-Garcia et al., 2020).
Altogether, deleterious mutations in these genes through alterations to the innate immune response, enhancement of migratory properties, inhibition of apoptotic pathways, and impairment of the DNA repair system are potential routes to immune evasion mechanisms and metastasis formation in TNBC. Overall,
In conclusion, these data (as summarized in Figures 7E and 7F) show that Prune-1 drives metastatic spread in TNBC through two mechanisms of action. The first is related to its induction of migration and EMT in TNBC through activation of different intracellular signaling pathways, including TGF-b. The second mechanism of action is related to its contribution to the generation of an immunosuppressive TME that is permissive to tumor growth and metastatic progression, by taking part to the communication with TAMs and inducing their recruitment and polarization toward a tumor-promoting M2 phenotype. The interplay between Prune-1 and TAMs within the TME is mediated by modulation of the release of inflammatory cytokines and extracellular vesicles driven by Prune-1. Among the cytokines, Prune-1 enhances secretion of IL-17F, IL-28, and IL-20 from TNBC cells. We also showed that EVs derived from Prune-1-overexpressing primary metastatic TNBC cells contain proteins that have roles in EMT and metastasis (e.g., syntenin-1, Vim, Ifitm3). In human, we identified deleterious variants in six genes that are mainly involved in the activation of innate immune responses. We also used RNAseq analyses to show upregulation of IL-10, COL4A1, ILR1, and PDGFB, the expression levels of which are negatively correlated with prognosis in patients with TNBC ( Figure S10). Altogether, these findings support the notion of using these genes, in the near future, as potential indicators of prognosis poor outcome.
Finally, and most importantly, these data provide hope for the use of the Prune-1-targeting drug AA7.1 and any further new small molecule derivatives as immunomodulatory agents to ameliorate metastatic dissemination also via inhibition of M2-TAM polarization. These proof-of-concept studies presented here are now ready for the definition of a route for therapeutic application in TNBC.

Limitations of the study
The study provided preclinical data in a GEMM of metastatic TNBC, demonstrating the anti-tumorigenic action of AA7.1 small molecule. The predictive value of AA7.1 for therapeutic application in human TNBC patients deserves further investigations. At this time, the mutations here reported in the animal model generating lung metastasis with homologous mutated genes in human BC specimens need to be validated for prognostic use in human BC.

Resource availability
The conditions of submission and the BioMed Central Copyright and License Agreement are accepted. All the information essential to interpreting the data presented are available in the Figure legends and Supplemental Information. All resources used (antibodies, cell lines, animals, and software tools) are included in the Supplemental Information.

Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Prof. Massimo Zollo (massimo.zollo@unina.it).

Materials availability
All the data and materials will be available from the

Supplemental information
Untreated macrophages were used as negative controls. Cell migration was driven by a 2% FBS gradient. Measurements were taken at 5-min intervals, as impedance changes across the electrodes at the bottoms of the wells, for 8 h.
For cell proliferation assays, MMTV-Prune-1/Wnt1 and MMTV-Wnt1 cells were harvested, washed with PBS, resuspended in high-glucose Dulbecco's modified Eagle's medium (Euroclone) with 10% FBS and aliquoted (1 ×10 4 cells) into single wells of plates (xCELLigence E-plate 16; Acea Biosciences). Cell proliferation was recorded at 2-min intervals, as impedance changes across the electrodes at the bottoms of the wells, for up to 24 h.
For doubling time measurements, MMTV-Prune-1/Wnt1 and MMTV-Wnt1 cells were harvested, washed with PBS, resuspended (5 ×10 3 cells) in high-glucose Dulbecco's modified Eagle's medium (Euroclone) with 10% FBS, and seeded into single wells of plates (xCELLigence E-plate 16; Acea Biosciences). Cell proliferation was recorded at 2-min intervals, as impedance changes across the electrodes at the bottoms of the wells, for up to 120 h. The doubling time was calculated using the RTCA software, from the logarithmic phase of the growth curves as a readout of cell proliferation and behavior that integrates changes in cell number, attachment, and morphology.

Coculture experiments
The effects of conditioned media from Prune-1-silenced, empty-vector 4T1 cell clones, and MMTV-Prune-1/Wnt1 and MMTV-Wnt1 cells were analyzed for activation of J774A.1 macrophages. Briefly 1 ×10 6 4T1 cells of each clone were resuspended in 10 mL complete RPMI 1640 or DMEM medium, plated into 10-cm-diameter plates, and grown at 37 °C for 24 h. Then the conditioned media were collected. One day before culture in the conditioned media, J774A.1 and Raw264.7 macrophages were plated into 10-cm-diameter plates. At the time of the culturing in conditioned media, macrophages were at 50% confluence, and after 6 h of starvation in serum-free medium, the macrophages were grown in conditioned media for 30 min (for Western blotting) or 48 h (for realtime PCR). Each experimental point was carried out in duplicate. After the culturing in the conditioned media, the macrophages were washed with PBS, collected, and used for protein and/or RNA extraction.

Cytokine antibody array
The conditioned media from three different Prune-1-overexpressing, Prune-1-silenced, and emptyvector 4T1 cell clones were pooled. In the same manner, the sera from three different MMTV-Prune-1/Wnt1 and MMTV-Wnt1 cells were also collected and pooled. The relative levels of cytokines in the pooled conditioned media and murine sera were measured (RayBio Mouse Cytokine Antibody Array C, series 2000; Prodotti Gianni), according to the manufacturer protocol. Densitometric analysis was carried out using the Quantity One software (BioRad). Expression levels were normalized to the levels of the positive control spots (contained within the membrane). For each cytokine, the mean ±standard error was determined.

Extracellular vesicles isolation
Extracellular vesicles were purified from media culture supernatants of MMTV-Prune-1/Wnt1 and MMTV-Wnt1 cells through methodology previously described , with modifications. Briefly, MMTV-Prune-1/Wnt1 and MMTV-Wnt1 cells were grown in 'exosomesdepleted medium' (obtained via overnight centrifugation at 100,000 ×g) until they reached 80% confluence. After 48 h, the conditioned medium was collected by centrifugation at 300× g for 10 min.
The supernatant was then pre-cleared to remove cells, dead cells, and cellular debris by centrigufation at 2,000× g for 20 min. The extracellular vesicles were then obtained via ultracentrifugation at 100,000× g for 70 min. After washing with PBS, the extracellular vesicles were further purified (ExoQuick Exosome Precipitation Solution; Cat.#EXOQ5A-1; System Biosciences) by incubation for 12 h and centrifugation at 1,500× g for 60 min.

Mutation analyses
Whole exome sequencing was performed on tumorigenic cells obtained from primary tumors For the MMTV-Prune-1-Wnt1 sample, for the post-alignment statistics, the initial mappable reads (i.e., number of mapped reads to mouse genome) was 74,311,106, the non-redundant reads (i.e., trimming tasks: cutting of adapter sequences, cutting of bases off the ends of reads if below a quality score of 3, cutting of windows of four bases if the average quality fell below a quality score of 15, and, finally, dropping reads shorter than 36 bp. HISAT2 version 2.1.0 3 , a spliced alignment program, was used to align the trimmed reads to the reference genome (Mouse GRCm38/mm10). The expression levels of known genes were determined using StringTie version 1.3.4d 4 , with the mm10 reference annotation as the assembly guide. Raw read counts where normalized using the median ratio normalization method 5 , and the resulting normalized count tables were fed to GFOLD version 1.1.4 6 , to perform gene expression analysis, setting the significant cut-off for fold-change to the default 0.01. Accession to RNAseq data: http://www.ebi.ac.uk/arrayexpress/experiments/E-MTAB-

9231.
Gene-set enrichment analysis: Gene-set enrichment analysis was performed using GSEA Pre-ranked with gfold=0, which GFOLD classifies as nondifferentially expressed, were removed from the list before running GSEA. The number of permutations to assess the statistical significance of the enrichment score, done by gene set, was set to 1000. To identify a subset of 'core genes' potentially responsible for most of the features associated with different phenotypes, we compared the lists of genes from the leading edge of enriched gene-sets coming from each CP sub-collection (Biocarta, Kegg, Pid, Reactome, Naba), and selected those shared by at least four out of five of these subcollections.