Abstract Upon penetrating the basement membrane, breast cancer cells directly interact with their surrounding adipose tissue, which forms a unique tumor-associated adipose microenvironment (TAME). However, the underlying mechanism of lipid metabolic remodeling in the TAME remains elusive. Herein, we report a Zeb1-orchestrated bidirectional communication between breast cancer cells and their adjacent cancer-associated adipocytes (CAAs). At the molecular level, breast cancer cells, through the secretion of adrenomedullin (AM), induce downregulation of Zeb1 expression to activate the Atgl/Hsl/Scd-dependent lipolysis in CAAs, resulting in the release of palmitoleic acid (POA) into the TAME. In turn, the increased POA in breast cancer competes with arachidonic acid (ARA) for the phospholipid synthesis, leaving more ARA is utilized for PDG[2] production to trigger the malignant progression of breast cancer and AM production. Importantly, disruption of Zeb1-dependent lipolytic activity and/or membrane phospholipid remodeling within the TAME dramatically diminishes the aggressiveness of breast cancer in vitro and in vivo. Subject terms: Breast cancer, Cancer metabolism, Cancer microenvironment __________________________________________________________________ Breast cancer cells interact with cancer-associated adipocytes in the microenvironment. Here, the authors report that the cross-talk between these two cell types is dependent on Zeb1, which remodels the tumor-associated adipose microenvironment to confer breast cancer progression. Introduction Growing evidence has suggested that breast cancer undergoes metabolic reprogramming during its malignant progression, with substantial alterations in a series of metabolic pathways, such as glycolysis, tricarboxylic acid cycle, and lipid metabolism^[72]1–[73]5. As the stroma of breast cancer is enriched in adipose tissue, the interaction between cancer cells and their surrounding adipocytes establishes a unique tumor-associated adipose microenvironment (TAME). This gives breast cancer cells a great advantage to exploit TAME to obtain the energy, biomembrane components, and signaling molecules necessary for cancer cell survival, invasion, metastasis, as well as responsiveness to antitumor therapies^[74]6–[75]11. Simultaneously, breast cancer cells produce various factors that reciprocally act on their adjacent adipocytes to induce lipolysis, which, in turn, facilitates lipid metabolic reprogramming to shape the unique TAME^[76]12–[77]15. Therefore, it is crucial to further elucidate the specific regulatory mechanism of lipid metabolic reprogramming in breast cancer, which will open up new avenues for targeting the TAME. During the malignant progression, breast cancer cells penetrate the basement membrane and are directly exposed to surrounding adipose tissue. Thus, adipocytes adjacent to the invasive frontier of breast cancer are referred to as cancer-associated adipocytes (CAAs)^[78]16–[79]18. It has been noted that CAAs are characterized by smaller cell size, reduced lipid droplets, and a fibroblast-like morphology^[80]19–[81]21. Given that CAAs undergo dedifferentiation, there is a notable downregulation of adipogenic differentiation markers such as peroxisome proliferator-activated receptor γ (Ppar-γ) and CCAAT/enhancer binding protein α (C/ebp-α). Conversely, the expression levels of pro-inflammatory factors, such as cytokines leukemia inhibitory factor (LIF) and C-X-C subfamily chemokines (CXCLs), were upregulated^[82]22–[83]25. CAAs also display enhanced metabolic activity and secretory function to facilitate a series of malignant phenotypes in breast cancer. For example, the lipolytic activity of CAA mobilizes the breakdown of triglycerides into free fatty acids (FFAs). Further, upon delivery into adjacent breast cancer cells, FFAs undergo mitochondrial β-oxidation to produce metabolites such as ketone bodies, which serve as an energy source for tumor growth in the nutrient-deprived microenvironment^[84]26–[85]28. On the other hand, FFAs also participate in the dynamic remodeling of cell membrane phospholipids, which consequently contributes to key processes of breast cancer metastasis and invasion, resistance to oxidative stress, as well as protection against drug-induced cytotoxicity^[86]29,[87]30. Therefore, the lipid metabolic remodeling associated with CAAs has emerged as a critical driver of breast cancer progression. However, the key regulatory factors triggering this process and their underlying mechanisms remain largely unexplored. Previous studies have established that Zeb1, as a transcriptional factor, modulates various biological behaviors of human cancers through genetic and epigenetic mechanisms^[88]31–[89]34. Especially, Zeb1 is demonstrated to substantially impact the plasticity and heterogeneity of breast cancer cells, which are crucial for tumor evolution and adaptation to environmental cues^[90]35,[91]36. Of note, recent investigations have also indicated a potential role of Zeb1 in the regulation of lipogenesis^[92]37,[93]38. For example, Gubelmann et al. systematically identified Zeb1 as a central transcriptional component determining the fate of adipocyte differentiation, which is capable of directly regulating a series of adipogenic factors, including Pref1, Ppar-γ, and C/ebpα^[94]37. Accordingly, aberrant expression of Zeb1 can substantially impact adipogenic differentiation both in vitro and in vivo. Recent investigation by Shi et al. has confirmed that Zeb1 confers adipogenic differentiation and lipogenesis in progenitor cells through targeting the mTORC1 and TGF-β signaling pathways^[95]38. Therefore, it is of paramount importance to comprehensively explore the potential mechanism by which Zeb1 regulates remodeling of the TAME in breast cancer. In this study, we delineate a bidirectional communication between breast cancer cells and their surrounding CAAs in the TAME. Mechanistically, Adipocyte-specific Zeb1 downregulation triggers the lipid metabolic reprogramming in CAAs, which leads to the release of a series of monounsaturated fatty acids (MUFAs), such as palmitoleic acid (POA), to subsequently facilitate breast cancer cell growth, invasion, and metastasis. Of note, delivery of POA into cancer cells also promotes adrenomedullin (AM) production via an ARA-PDG[2]-mediated membrane phospholipid replacement mechanism and reciprocally downregulates Zeb1 expression in adjacent CAAs to create a positive feedback loop. Collectively, our data suggest that elucidating the microenvironmental signals (i.e., interventions involved in remodeling of the TAME) influencing metabolic signatures of breast cancer could provide effective options for the treatment of advanced human cancers. Results Zeb1 plays a pivotal role in mammary adipogenesis To investigate Zeb1-mediated adipose tissue remodeling in mammary development, we crossed conditioned Zeb1 knockout mice (Zeb1^fl/fl) with adipocyte-specific Fabp4-Cre mice (Fabp4-Cre). The homozygous female mice with an adipocyte-specific deletion of Zeb1 were identified (Fabp4-Cre^+; Zeb1^fl/fl) and named as Zeb1^adiKO (Fig. [96]1a). The results of RT-PCR and immunohistochemical staining confirmed that the expression of Zeb1 was significantly downregulated in the adipose tissues of Zeb1^adiKO mice (Fig. [97]1b, c). Moreover, the analysis of mammary gland developmental phenotypes showed that the weight of mammary fat pads was significantly reduced in Zeb1^adiKO mice (Fig. [98]1d). The HE staining and Perilipin-1 immunostaining further revealed smaller adipocytes with multiple lipid droplets in their cytoplasm (Fig. [99]1e, f), accompanied by the upregulation of preadipocyte factor Pref1 and downregulation of adipogenic factors including Ppar-γ and C/ebpα (Fig. [100]1g). In contrast, the length of invasive ducts was shortened in Zeb1^adiKO mice, with a marked reduction in the number of terminal end buds (TEBs), which is indicative of terminal duct differentiation (Supplementary Fig. [101]1a). Notably, Zeb1^adiKO mice showed decreased levels of Ki67 expression in ductal epithelial cells and less collagen deposition around the ducts (Supplementary Fig. [102]1b, c). Fig. 1. Adipocyte-specific depletion of Zeb1 impairs adipocyte differentiation in vivo. [103]Fig. 1 [104]Open in a new tab a Genotypic identification of the genetic mouse model with adipocyte-specific Zeb1 knock-out (n = 5 Zeb1^fl/fl, 5 Zeb1^adiKO). b Relative mRNA levels of Zeb1 (n = 5 Zeb1^fl/fl, 5 Zeb1^adiKO). c Immunohistochemistry staining for Zeb1 (n = 5 Zeb1^fl/fl, 5 Zeb1^adiKO; scale bars, 50 μm). d Fat pad weight (n = 5 Zeb1^fl/fl, 5 Zeb1^adiKO). e HE staining for adipocyte size analysis (n = 5 Zeb1^fl/fl, 5 Zeb1^adiKO; Scale Bars, 50 μm). f Immunofluorescence staining for perilipin-1 (n = 5 Zeb1^fl/fl, 5 Zeb1^adiKO; Scale Bars, 50 μm). g Relative mRNA levels of Pref1, Ppar-γ and C/ebpα (n = 5 Zeb1^fl/fl, 5 Zeb1^adiKO). h TSNE visualization showing seven major clusters of mammary cells by single nucleus RNA-sequencing (n = 3 Zeb1^fl/fl, 3316 cells, 3 Zeb1^adiKO, 6937 cells). i Relative ratio of the indicated cell cluster by single nucleus RNA-sequencing (n = 3 Zeb1^fl/fl, 3 Zeb1^adiKO). j TSNE visualization showing five adipocyte subtypes by single nucleus RNA-sequencing (n = 3 Zeb1^fl/fl, 1548 cells, 3 Zeb1^adiKO, 4048 cells). k Pseudo-time trajectory for the adipocyte subpopulations by single nucleus RNA-sequencing (n = 3 Zeb1^fl/fl, 3 Zeb1^adiKO). l Relative ratio of the indicated adipocyte subpopulation by single nucleus RNA-sequencing (n = 3 Zeb1^fl/fl, 3 Zeb1^adiKO). m Dot plot showing the pathway enrichment of the indicated adipocyte subpopulation by single nucleus RNA-sequencing (n = 3 Zeb1^fl/fl, 3 Zeb1^adiKO). n Violin plots of the lipid metabolism-related genes for the indicated adipocyte clusters by single nucleus RNA-sequencing (n = 3 Zeb1^fl/fl, 3 Zeb1^adiKO). Data were expressed as means ± SEM. b–e, g were assessed via two-tailed unpaired Student’s t-test. m were assessed via hypergeometric test. Source data are provided as a Source Data file. Next, we used the pre-adipocytes 3T3-L1 to induce adipogenic differentiation in vitro (Supplementary Fig. [105]1d), which demonstrated elevated Zeb1 expression at both the mRNA and protein levels (Supplementary Fig. [106]1e, f). Concurrently, Pref1 was downregulated, while Ppar-γ and C/ebp-α were upregulated upon adipogenic induction (Supplementary Fig. [107]1g). Furthermore, stable cell lines with knockdown of Zeb1 were respectively established in 3T3-L1 cells (Supplementary Fig. [108]1h, i), followed by adipogenic induction. The analysis of Oil Red O staining revealed that Zeb1 knockdown significantly inhibited adipogenic differentiation in shZeb1/3T3-L1 cells (Supplementary Fig. [109]1j), accompanied by the upregulation of Pref1 expression and downregulation of Ppar-γ and C/ebpα expression (Supplementary Fig. [110]1k). On the contrary, Zeb1 overexpression performed the opposite effects to promote adipogenic phenotypes in Zeb1/3T3-L1 cells (Supplementary Fig. [111]1l–n), accompanied by the downregulation of Pref1 expression and upregulation of Ppar-γ and C/ebpα expression (Supplementary Fig. [112]1o). Collectively, these findings highlighted a crucial role of Zeb1 in the regulation of mammary adipogenesis in vitro and in vivo. Adipose-derived Zeb1 affects phenotypic variability of mammary cells To gain a comprehensive understanding of the role of Zeb1 in the development of mammary adipose tissue, we mapped the single nucleus RNA-sequencing profiles of mammary tissues from Zeb1^fl/fl and Zeb1^adiKO mice. As shown in Fig. [113]1h, cell populations were clustered based on typical lineage marker genes and visualized using t-distributed stochastic neighbor embedding (t-SNE) plots. The analysis of marker genes for cluster annotation revealed distinct clusters comprised of adipose-derived stem cells (ADSCs) marked by DPP4, adipocytes marked by Adiponectin, fibroblasts marked by Col1a1, myoepithelial cells marked by Myh11, epithelial cells marked by Epcam, endothelial cells marked by Cdh5, and immune cells marked by Ptprc (Supplementary Fig. [114]2a). Of note, Zeb1^adiKO mice displayed increased proportion of ADSCs and fibroblasts within their mammary tissue, along with a reduction in immune cell populations, compared with Zeb1^fl/fl mice (Fig. [115]1i). To further verify the phenotypic variability of adipocytes, we subdivided and visualized the adipocyte populations using tSNE diagrams. The analysis of marker genes for cluster annotation revealed four distinct cell populations, which were brown adipocytes marked by Ucp1, pre-adipocytes (ProA) and two populations of white adipocytes, including White1 and White2 (Fig. [116]1j). The annotation of adipogenic genes indicated high expression of Cfd, Ppar-γ and Adiponectin in the White1 and White2 clusters, whereas those in the ProA cluster was not evident (Supplementary Fig. [117]2b). Furthermore, by examining the potential trajectory of adipocyte subsets, we found that ADSCs were significantly enriched in the early stage of adipogenesis, which progressively differentiated into the ProA population and ultimately formed the White1, White2 and brown adipocytes (Fig. [118]1k). Notably, compared with the control Zeb1^fl/fl mice, Zeb1^adiKO mice showed a remarked increase in the proportion of ADSCs and ProA in their mammary tissues, along with a substantial decrease in the White1 cluster (Fig. [119]1l). Next, pathway enrichment analysis of the top 100 differentially expressed genes (DEGs) revealed striking differences among the ADSC, ProA, White1 and brown cell clusters (Fig. [120]1m). Specifically, the ADSC cluster was mainly enriched in pathway related to extracellular matrix (ECM) remodeling, the ProA cluster was primarily enriched in fatty acid synthesis pathway, the White1 cluster was enriched in pathway associated with fatty acid metabolism, and the brown cluster was enriched in triglyceride metabolic pathway. Moreover, we annotated the lipid metabolism-related genes in the ProA, White1, and White2 clusters, which encompassed both lipid anabolism- and catabolism-associated genes as indicated in Fig. [121]1n. The results demonstrated that the ProA cluster exhibited an overall decrease in lipid metabolism. In the White 1 cluster, the expression of lipid synthesis-associated genes such as Dgat2, Fasn, and Thrsp were upregulated, whereas those of catabolism-associated genes including Atgl, Mgl, Acox1, and Apol6 were downregulated. In contrast, the White2 cluster showed enhanced lipid catabolism but reduced lipid synthesis. Importantly, compared to the control Zeb1^fl/fl mice, the two White clusters from Zeb1^adiKO mice exhibited remarkedly decreased lipid synthesis as indicated by downregulation of Dgat2, Fasn, and Thrsp expression, but increased lipolysis as indicated by upregulation of Atgl, Scd, Acox1, and Apol6 expression. These results indicated that the loss of adipocyte specific Zeb1 expression could change the expression of lipid metabolism genes. Further analysis of intercellular interactions revealed that both Zeb1^fl/fl and Zeb1^adiKO mice exhibited specific Nrg4 expression in mammary adipocytes. However, members of the Erbb family (including Erbb3 and Erbb4), as receptors for Nrg4, were specifically expressed in the ductal, glandular and myoepithelial cells of Zeb1^fl/fl mice, but not in those epithelial cells of Zeb1^adiKO mice (Supplementary Fig. [122]2c). We also observed Jag1 expression in duct epithelial cells of Zeb1^adiKO mice. On the other hand, Notch3, as receptor of Jag ligands, was expressed in ductal cells, with Zeb1^adiKO mice displaying lower expression than the control mice (Supplementary Fig. [123]2d). These findings suggested that the Nrg-Erbb pathway predominantly mediate interactions between adipocytes and their surrounding mammary epithelial cells, whereas the Jag-Notch pathway facilitates communication among mammary epithelium themselves. Importantly, the Zeb1 might play a potential role in orchestrating these phenotypic variabilities of mammary cells, thereby affecting the maturation and development of mammary gland. Aberrant Zeb1 triggers lipid metabolic reprogramming in CAAs Considering that cancer cells interact with their adjacent adipocytes to educate cancer-associated adipocytes (CAAs) in the tumor-associated microenvironment (TME), we thus performed immunohistochemical staining to analyze the expression of Zeb1 in CAAs at the invasive front of breast tumors from MMTV-PyMT mice. The results showed that CAAs, characterized by the reduced volume and loss of lipid droplets present at the invasive front of PyMT tumors, had significantly decreased expression of nuclear Zeb1 compared with the normal mammary adipocytes (Fig. [124]2a). Further, we collected tissue specimens from 40 patients with breast cancer (Table [125]S1) and examined the contents of CAA-derived Zeb1 by immunohistochemistry. Consistently, CAAs showed downregulated Zeb1 at the invasive front of human breast cancer tissues (Fig. [126]2b), confirming that aberrant expression of adipocyte-specific Zeb1 is potentially related to the TAME remodeling in breast cancer. Fig. 2. Elevated expression of adipocyte-specific Zeb1 inhibits breast cancer development in vitro. [127]Fig. 2 [128]Open in a new tab a Immunohistochemical staining for Zeb1 in normal mammary adipocytes and CAAs in breast tumors from MMTV-PyMT mice (n = 12 PyMT mice) (A: adipose, T: tumor; scale bars, 50 μm). b Immunohistochemical staining for ZEB1 in normal breast adipocytes and CAAs in human breast cancer samples (n = 40 patients; scale bars, 50 μm). c, d Relative mRNA (c) and protein (d) levels of Zeb1 in 3T3^Ctrl-Tet and 3T3^Zeb1-Tet cells in response to different concentrations of doxycycline (DOX) (n = 3 independent experiments). e Scheme of cancer cell coculture with 3T3^Ctrl-Tet (CAA^Zeb1-off) and 3T3^Zeb1-Tet (CAA^Zeb1-on) upon DOX-induced Tet-on expression of Zeb1 for 6 days. f, g Oil Red O (f) and Bodipy-C16 fluorescence (g) staining in CAA^Zeb1-off and CAA^Zeb1-on co-culture with EO771 cells (n = 3 independent experiments; scale bars, 50 μm). h–j EdU proliferation (h), transwell invasion (i), and high-content migration (j) assays in EO771 cells pretreated with supernatant from CAA^Zeb1-off and CAA^Zeb1-on cells (n = 3 independent experiments; scale bars, 50 μm). k Protein levels of Snail and Vimentin in EO771 cells pretreated with supernatant from CAA^Zeb1-off and CAA^Zeb1-on cells (n = 3 independent experiments). Data were expressed as means ± SEM. Indicated P-values were calculated using two-tailed unpaired Student’s t-test. Source data are provided as a Source Data file. We also cocultured 3T3-L1-diffrentiated adipocytes 3T3^Adi with EO771 cells to obtain 3T3^CAA cells (Supplementary Fig. [129]3a). We found that, compared with the control 3T3^Adi cells, 3T3^CAA cells displayed a remarkable reduction of Zeb1 expression at both the mRNA and protein levels, along with upregulated Pref1 expression and downregulated Ppar-γ and C/ebpα expression (Supplementary Fig. [130]3b, c). Subsequently, we used the Tet-on system to construct 3T3-L1 stable cells with induction of Zeb1 expression upon doxycycline (DOX) treatment in a dose-dependent manner (Fig. [131]2c, d). Furthermore, 3T3^Ctrl-Tet and 3T3^Zeb1-Tet cells were respectively subjected to lipogenic differentiation, followed by co-culture with EO771 cells to establish CAAs termed as CAA^Zeb1-off and CAA^Zeb1-on cells (Fig. [132]2e). The Oil Red O and Bodipy-C16 fluorescence staining assays further demonstrated that Zeb1 induction significantly weakened the lipolytic phenotypes in CAA^Zeb1-on cells (Fig. [133]2f, g). Similarly, we repeated these experiments in MDA-MB-231 human breast cancer cells and obtained consistent results (Supplementary Fig. [134]3d, e), indicating that alteration of Zeb1 expression played a pivotal role in the lipid metabolic reprogramming in CAAs. Further, we collected the conditioned media (CM) from CAA^Zeb1-off and CAA^Zeb1-on cells to incubate with EO771 cells, respectively. The results of EdU cell proliferation, transwell invasion, and high-content migration assays showed that treatment with CM from CAA^Zeb1-off cells significantly promoted the proliferation, invasion, and migration of EO771 cells, while induction of Zeb1 expression in CAA^Zeb1-on cells reduced their pro-tumorigenic capacity (Fig. [135]2h–j). Consistently, Western blot results showed that compared with CAA^Zeb1-off-derived CM, CAA^Zeb1-on-derived CM lost the activity to enhance the expression of EMT-related genes such as Snail and Vimentin in EO771 cells (Fig. [136]2k). We also performed these experiments in MDA-MB-231 cells and obtained similar results (Supplementary Fig. [137]3f–i), On the contrary, we utilized the Tet-On system to construct 3T3-L1 stable cells with induction of Zeb1 knockdown upon DOX treatment (Supplementary Fig. [138]4a, b), which confirmed the opposite effects showing that specific downregulation of adipose-derived Zeb1 increased the lipolytic activities in CAAs (Supplementary Fig. [139]4c–e) and consequently promoted malignant phenotypes of EO771 cells (Supplementary Fig. [140]4f–i) in a paracrine action. Similarly, we also performed these experiments using MDA-MB-231 cells and obtained consistent results (Supplementary Fig. [141]5). CAA-derived Zeb1 plays a key role in breast cancer progression To further investigate the function of adipocyte-specific Zeb1 in regulating breast cancer progression in vivo, as shown in Fig. [142]3a, we generated conditional Zeb1 knock-in mice (B6-CAG-LSL-Zeb1), which were crossed with Fabp4-Cre mice to establish adipocyte-specific Zeb1 knock-in mice (Zeb1^adiTG). The results of RT-PCR and immunohistochemical staining confirmed that Zeb1 expression levels were significantly upregulated in mammary adipocytes of Zeb1^adiTG mice (Fig. [143]3b, c). Further analysis of mammary gland development phenotypes revealed that the weight of mammary fat pads from Zeb1^adiTG mice was increased compared with those from the littermate control Zeb1^TG mice (Supplementary Fig. [144]6a). Moreover, the HE staining and Perilipin-1 immunofluorescence assays indicated enlarged adipocyte size in the mammary glands of Zeb1^adiTG mice (Supplementary Fig. [145]6b, c), which was accompanied by the upregulation of C/ebp-α expression (Supplementary Fig. [146]6d). Interestingly, Zeb1^adiTG mice did not exhibit obvious abnormalities in mammary ductal development, TEB morphology or epithelial cell proliferation (Supplementary Fig. [147]6e, f). However, there was a slight increase in collagen deposition around the ducts in Zeb1^adiTG mice (Supplementary Fig. [148]6g). Fig. 3. Elevated expression of adipocyte-specific Zeb1 inhibits breast cancer development in vivo. [149]Fig. 3 [150]Open in a new tab a Genotypic identification of the genetic mouse model with adipocyte specific-Zeb1 knock-in (n = 5 Zeb1^TG, 5 Zeb1^adiTG). b Relative mRNA levels of Zeb1 (n = 5 Zeb1^TG, 5 Zeb1^adiTG). c Immunohistochemical staining for Zeb1 (n = 5 Zeb1^TG, 5 Zeb1^adiTG; scale bars, 50 μm). d EO771 breast cancer allograft in Zeb1^TG and Zeb1^adiTG mice. e Tumor volume (n = 6 Zeb1^TG, 6 Zeb1^adiTG). f, g HE (f) and immunofluorescence (g) staining for Perilipin-1 (A: adipose, T: tumor; n = 6 Zeb1^TG, 6 Zeb1^adiTG; scale bars, 50 μm). h, i, Immunohistochemical staining for Ki-67 (h) and Vimentin (i) (n = 6 Zeb1^TG, 6 Zeb1^adiTG; scale bars, 50 μm). j Scheme of PyMT breast cancer allograft tumor model. k PyMT breast cancer allograft in Zeb1^TG and Zeb1^adiTG mice. l Tumor volume (n = 6 Zeb1^TG, 6 Zeb1^adiTG). m, n HE (m) and immunofluorescence (n) staining for Perilipin-1 (A adipose, T tumor; n = 6 Zeb1^TG, 6 Zeb1^adiTG; scale bars, 50 μm). o, p Immunohistochemical staining for Ki-67 (o) and Vimentin (p) (n = 6 Zeb1^TG, 6 Zeb1^adiTG; scale bars, 50 μm). Data were expressed as means ± SEM. e, l was analyzed via two-way ANOVA with Sidak correction for multiple comparisons; b, c, f, h, i, m, o, p were assessed via two-tailed unpaired Student’s t-test. Source data are provided as a Source Data file. Subsequently, single-cell suspensions of EO771 cells were injected into the fourth mammary fat pads of female Zeb1^adiTG mice to establish an allograft model. As shown in Fig. [151]3d, EO771 allografts in Zeb1^adiTG mice displayed dramatic reduction of tumor growth (Fig. [152]3d, e). Further examination with HE and Perilipin-1 immunofluorescence staining revealed that, in the control allograft tumors, CAAs at the invasive front exhibited a marked reduction in size, with uneven and disordered lipid droplets. Conversely, the invasive front of tumor allografts from Zeb1^adiTG mice displayed clearer boundaries, with CAAs maintaining a unilocular lipid droplet morphology (Fig. [153]3f, g). Immunohistochemical staining further indicated that Ki67 expression level was decreased in Zeb1^adiTG tumor tissues, and so was the expression of Vimentin at the invasion front (Fig. [154]3h, i). In line with these, we also established the allograft model with primary PyMT tumor cells in Zeb1^TG and Zeb1^adiTG mice, revealing that overexpression of adipose-specific Zeb1 led to reduced tumorigenicity in Zeb1^adiTG mice, with alleviated lipolytic activities in CAAs (Fig. [155]3j–p). Based on these findings, we further established an experimental lung metastasis model with EO771 cells preincubated with the conditioned media from CAA^Zeb1-off and CAA^Zeb1-on cells in C57BL/6 mice (Fig. [156]4a). The results showed that compared with EO771 cells treated with the supernatant from CAA^Zeb1-off cells, the lung metastases of mice treated with the supernatant from CAA^Zeb1-on cells were significantly reduced (Fig. [157]4b–e). In consistent, we also performed these experiments with primary PyMT breast cancer cells and obtained consistent results (Fig. [158]4f–j), highlighting that aberrant expression of adipocyte-specific Zeb1 architects the interplay between CAAs and their surrounding tumor cells to eventually promote the malignant progression of breast cancer. Fig. 4. Elevated expression of adipocyte-specific Zeb1 inhibits lung metastasis in breast cancer. [159]Fig. 4 [160]Open in a new tab a Scheme of establishing the experimental lung metastasis of EO771 cells (Some elements by figdraw.com). b Representative images of lung metastases at 30 days (n = 5 mice; scale bars, 50 μm). c–e Lung weight (c), number of metastatic lesions (d) and metastatic area (e) (n = 5 mice). f Scheme of establishing the experimental lung metastasis of PyMT cells (Some elements by figdraw.com). g Representative images of lung metastases at 30 days (n = 5 mice; scale bars, 50 μm). h–j Lung weight (h), number of metastatic lesions (i), and metastatic area (j) (n = 5 mice). Data were expressed as means ± SEM. c–e, h–j were assessed via two-tailed unpaired Student’s t-test. Source data are provided as a Source Data file. CAA-derived Zeb1/Scd axis exerts pro-tumorigenic effects by producing MUFAs Next, to further elucidate the molecular mechanism by which adipocyte-specific Zeb1 modulates lipid metabolic reprogramming in CAAs, we conducted untargeted lipidomics analysis of CAA^Zeb1-off and CAA^Zeb1-on cells. Our results totally identified 3137 lipids that belonged to 40 classes (Supplementary Fig. [161]7a), which displayed a marked distinction in the lipidomic profile between the two groups of CAA^Zeb1-on and CAA^Zeb1-off cells (Supplementary Fig. [162]7b). Variable importance for projection (VIP) scores were used to further distinguish differential metabolites, revealing significant alterations (VIP > 2 and P < 0.05) in various lipid species within CAA^Zeb1-on cells, such as triacylglycerol lipids (TG), phosphatidylethanolamines (PE), phosphatidylcholines (PC), etc., as indicated in Supplementary Fig. [163]7c. We noticed that triglyceride enrichment was more pronounced in CAA^Zeb1-on cells compared with the control CAA^Zeb1-off cells (Fig. [164]5a), especially the contents of MUFA incorporation into triglyceride was significantly increased (Fig. [165]5b). We next performed medium- and long-chain fatty acid (MLCFA) metabolomic analysis and confirmed significantly reduced production of MUFAs, including palmitoleic acid (POA, C16:1N7) and oleic acid (OA, C18:1N9) in the supernatants from CAA^Zeb1-on cells (Fig. [166]5c, d). These findings demonstrated that aberrant Zeb1 expression triggered lipid metabolic reprogramming in CAAs, which led to the production of specific MUFAs. Fig. 5. Ectopic Zeb1 triggers lipid remodeling by regulating MUFA production in CAAs. [167]Fig. 5 [168]Open in a new tab a Triglyceride content by non-targeted lipidomics analysis in CAA^Zeb1-off and CAA^Zeb1-on cells (n = 5 independent experiments). b Content of monounsaturated fatty acid (MUFA) in Triglyceride (n = 5 independent experiments). c, d Targeted metabolomics analysis of medium- and long-chain fatty acids (c) and their differential contents (d) in the supernatant from CAA^Zeb1-off and CAA^Zeb1-on cells (n = 6 independent experiments). e, f RNA-sequencing combined with GSEA for gene sets related to triglyceride and fatty acid metabolisms in CAA^Zeb1-off and CAA^Zeb1-on cells (NES, normalized enrichment score) (n = 3 independent experiments). g, h Relative mRNA (g) and protein (h) levels of Atgl, Hsl, Mgl, and Scd in CAA^Zeb1-off and CAA^Zeb1-on cells (n = 3 independent experiments). i Luciferase assay for the wild-type promoters of Atgl (−2960/ + 99), Hsl (−3000/ + 8) and Mgl (−2997/ + 103) in Ctrl/3T3-L1 and Zeb1/3T3-L1 cells (n = 3 independent experiments). j Luciferase assay for the wild-type (−2746/ + 300) and E[2]-box-mutated promoters of Scd in Ctrl/3T3-L1 and Zeb1/3T3-L1 cells (n = 3 independent experiments). k, l ChIP assay for recruitment of Zeb1 to the endogenous Scd promoter in Ctrl/3T3-L1 and Zeb1/3T3-L1 cells (n = 3 independent experiments). m Scheme of CAA^Zeb1-off and CAA^Zeb1-on co-culture with EO771 cells in the presence or absence of A939572 (100 nM) for 6 days. n, o Targeted metabolomics analysis of medium- and long-chain fatty acids (n) and their differential contents (o) in the supernatant from CAA^Zeb1-off and CAA^Zeb1-on cells in the presence or absence of A939572 (n = 3 independent experiments). Data were expressed as means ± SEM. Indicated P-values were calculated using two-tailed unpaired Student’s t-test. Source data are provided as a Source Data file. To identify the key regulators of lipid metabolism downstream of Zeb1, we performed RNA-sequencing in CAA^Zeb1-off and CAA^Zeb1-on cells. The gene set enrichment analysis (GSEA) revealed the downregulation of triglyceride- and fatty acid-metabolism-associated gene sets in CAA^Zeb1-on cells (Fig. [169]5e). Specifically, the expression of a series of key enzymes involved in triglyceride catabolism (Fig. [170]5f), including adipose triglyceride lipase (Atgl), hormone-sensitive triglyceride lipase (Hsl), and monoglyceride lipase (Mgl), were significantly reduced at both the mRNA (Fig. [171]5g) and protein (Fig. [172]5h) levels. In line, their enzyme activities were verified to be decreased in CAA^Zeb1-on cells (Supplementary Fig. [173]7d–f). We also noticed a remarkable downregulation of stearoyl-CoA desaturase (Scd) expression in CAA^Zeb1-on cells (Supplementary Fig. [174]7g), which is critical for the synthesis of MUFAs such as POA and OA. To further investigate the transcriptionally regulatory effects of Zeb1 on Atgl, Hsl, Mgl, and Scd expression, we constructed luciferase reporters driven by the promoters of these genes respectively, and demonstrated that overexpression of Zeb1 significantly suppressed the promoter activities of Atgl, Hsl, Mgl, and Scd in 3T3-L1 cells (Fig. [175]5i, j). In addition, we used the JASPA motif database ([176]https://jaspar.elixir.no/) to identify two E[2]-box elements (caggta) in the promoter of Scd gene, which could be the potential binding sites for Zeb1 (Fig. [177]5j). We then performed site-directed mutations at the E[2]-1 and E[2]-2 sites, either alone or in combination. The results of luciferase activity assay showed that mutagenesis of either E[2]-box element did not affect Zeb1-induced transcriptional repression of the Scd promoter, whereas simultaneous mutation of both E[2]-boxs completely eliminated this effect. The ChIP-qPCR assay also confirmed the specific recruitment of Zeb1 to both E[2]-1 and E[2]-2 loci of the Scd promoter, which could be further enriched upon overexpression of Zeb1 (Fig. [178]5k, l). These data suggested that Zeb1 directly bound to the Scd promoter in an E[2]-box-dependent manner and thus suppressed its transcription. On note, we constructed shScd/3T3-L1 cells with specific knockdown of Scd (Supplementary Fig. [179]8a, b), followed by induction of lipogenic differentiation and co-culture with EO771 cells (Supplementary Fig. [180]8c). The results showed that adipose-specific knockdown of Scd inhibited proliferation, invasion, and migration of EO771 cells (Supplementary Fig. [181]8d–f). We also verified these observations in MDA-MB-231 cells and obtained consistent results (Supplementary Fig. [182]8g–i), demonstrating that a Scd-dependent paracrine action is involved in the pro-tumorigenic effects of CAAs. Subsequently, to identify the specific MUFAs secreted by CAAs via a Zeb1/Scd-dependent mechanism, we obtained the conditioned media from CAA^Zeb1-off and CAA^Zeb1-on cells in the presence of a Scd inhibitor, A939572, for MLCFA-targeted metabolomics analysis (Fig. [183]5m). The results indicated that, compared with CAA^Zeb1-off cells, the production of POA and OA were significantly reduced upon Tet-on induction of Zeb1 expression in CAA^Zeb1-on cells (Fig. [184]5n, o). Moreover, specific blocking of MUFA generation with A939572 resulted in a significant reduction in the release of POA and OA from CAA^Zeb1-off cells; On this basis, the A939572 treatment slightly reduced the release of POA and OA. To further elucidate the paracrine mechanism by which MUFAs mediate the pro-tumorigenic effects of adipocyte-specific Zeb1, EO771 cells were respectively incubated with the conditioned media from CAA^Zeb1-off and CAA^Zeb1-on cells in the presence of A939572. The results of EdU cell proliferation, transwell invasion, and high-content migration assays confirmed that the conditioned media from CAA^Zeb1-off cells significantly promoted the malignant phenotypes of EO771 cells (Supplementary Fig. [185]8j–l); however, blocking MUFA production using A939572 effectively attenuated the pro-tumorigenic activity of CAA^Zeb1-off cells. Of note, upregulation of Zeb1 expression led to reduced pro-tumorigenic activity of CAA^Zeb1-on cells, while these effects were not altered by A939572 treatment. We also performed these experiments in MDA-MB-231 cells and obtained similar results (Supplementary Fig. [186]8m–o). Collectively, these findings revealed that Zeb1/Scd signaling-regulated MUFAs production in CAAs influenced the malignant progression of breast cancer in a paracrine-dependent action. POA confers the dynamic remodeling of breast cancer cell membrane Next, we treated EO771 cells with POA and OA respectively to compare their impacts on the proliferation and migration of breast cancer cells. The results demonstrated that treatment with POA strongly enhanced the proliferation, invasion, and migration of EO771 cells (Fig. [187]6a–c), whereas the effects of OA were not as evident. These results were also confirmed in MDA-MB-231 cells (Supplementary Fig. [188]9a–c). Upon delivery into cancer cells, FFAs such as MUFAs primarily exert biological functions through the following pathways, including conversion into triglycerides, β-oxidation in mitochondria, and participation in phospholipid remodeling of cellular membranes^[189]29,[190]39. Thus, we performed Oil Red O staining and observed increased lipid accumulation in EO771 cells by addition of POA (Supplementary Fig. [191]9d). Moreover, pre-treatment with etomoxir (ETX), an inhibitor of carnitine palmitoyl transferase-1 (Cpt1) which is the rate-limiting enzyme in mitochondrial fatty acid β-oxidation, only partially reversed the pro-tumorigenic effects of POA on EO771 cells (Supplementary Fig. [192]9e–g). We also obtained similar results in MDA-MB-231 cells (Supplementary Fig. [193]9h–k), suggesting that, beyond serving as an energy source for fatty acid β-oxidation, POA may also promote breast cancer progression through an alternative mechanism. Fig. 6. POA-mediated ARA/PGD[2] replacement regulates the dynamic remodeling of membrane phospholipids in breast cancer cells. [194]Fig. 6 [195]Open in a new tab a–c CCK8 cell viability (a), transwell invasion (b) and high-content migration (c) assays in EO771 cells treated with POA (5 μM) and OA (15 μM) for 48 h, respectively. POA: palmitoleic acid; OA: oleic acid (n = 3 independent experiments; scale bars, 50 μm). d Structure of membrane phospholipids. e Heat map of membrane phospholipids containing either 16:1 or 20:4 fatty acids in EO771 cells in the presence or absence of POA (n = 3 independent experiments). f Contents of the indicated membrane phospholipids containing either 16:1 or 20:4 fatty acids in EO771 cells in the presence or absence of POA (n = 3 independent experiments). g RNA-sequencing combined with GSEA for gene sets related to arachidonic acid metabolism in EO771 cells in the presence or absence of POA (n = 3 independent experiments). h Relative mRNA levels of Pla2g4a and Ptgds in EO771 cells in the presence or absence of POA (n = 3 independent experiments). i, j Targeted metabolomics analysis of arachidonic acids (i) and their differential contents (j) in EO771 cells in the presence or absence of POA. PGD[2]: prostaglandin D[2]; 15S-HETE: 15-hydroxyeicosatetraenoic acid; PGF[2α]: prostaglandin F[2α]; LTB4: leukotriene B4; ARA: arachidonic acid; 12S-HETE: 12-hydroxyeicosatetraenoic acid; 13S-HETE: 13-hydroxyeicosatetraenoic acid; DHA docosahexaenoic Acid (n = 3 independent experiments). k–m, CCK8 cell viability (k), transwell invasion (l), and high-content migration (m) assays in EO771 cells treated with AT-56 (50 μM) and/or POA (5 μM) for 48 h (n = 3 independent experiments; scale bars, 50 μm). n Protein levels of Snail and Vimentin in EO771 cells treated with AT-56 and/or POA (n = 3 independent experiments). Data were expressed as means ± SEM. a, k was analyzed via two-way ANOVA with Sidak correction for multiple comparisons; b, c, f, h, j, l, m were assessed via two-tailed unpaired Student’s t-test. Source data are provided as a Source Data file. We then performed a targeted lipidomic analysis using EO771 cells treated with POA and identified 940 lipids among 18 categories (Supplementary Fig. [196]10a, b). Differential enrichment analysis showed significant increases in TG, monoacylglycerol (MG), PE and PC upon POA treatment (Supplementary Fig. [197]10c). Previous reports have shown that MUFAs are preferentially bound to the sn-2 position of membrane phospholipid^[198]29, which is usually occupied by arachidonic acid (ARA, C20:4N6) (Fig. [199]6d). We thus analyzed the overall phospholipid subclasses that bound with either C16:1 or C20:4 at the sn-2 position (Fig. [200]6e). The results showed that the enrichment of C16:1 at the sn-2 position among five types of membrane phospholipids including PC, PE, PS, PG, and PI were significantly increased upon POA addition, while the contents of C20:4 at this position were decreased (Fig. [201]6e, f). To verify the subsequent outcomes, we performed RNA-sequencing analysis and found that POA treatment upregulated the gene set related to arachidonic acid metabolism in EO771 cells (Fig. [202]6g), with a notable increase in the mRNA levels of sn-2-specific phospholipase Pla2g4a and PGD[2] synthesizing enzyme Ptgds (Fig. [203]6h). Similar results were also obtained in MDA-MB-231 cells (Supplementary Fig. [204]10d). Further, the arachidonic acid metabolomics profiling revealed that POA treatment reduced the ARA levels but increased the contents of ARA-derived metabolites such as PGD[2], PGF[2α] and 15S-HETE in EO771 cells (Fig. [205]6i, j). Considering that PGD[2] was the specific product of the Ptgds enzyme and showed the highest expression level among ARA-derived metabolites, we then focused on the impact of PGD[2] on the malignant phenotype of tumor cells. First of all, we treated EO771 cells with PGD[2] or POA. The results showed that the addition of PGD[2] could promote the proliferation, invasion and migration of EO771 cells (Supplementary Fig. [206]11a–c), which were also phenocopied the effects of POA. These results were verified in MDA-MB-231 cells (Supplementary Fig. [207]11d–f). We then incubated POA-treated EO771 cells with a Ptgds inhibitor AT-56, which effectively inhibited the pro-tumorigenic effects of POA in EO771 cells (Fig. [208]6k–m). In addition, western blot analysis confirmed that POA-promoted increase in EMT markers, including Snail and Vimentin, was able to be blocked by AT-56 treatment in EO771 cells (Fig. [209]6n). We also confirmed these results in MDA-MB-231 cells (Supplementary Fig. [210]11g–j). These results revealed that POA preferentially inserts into the sn-2 position of membrane phospholipids to trigger ARA displacement from the cell membrane, thus eventually promoting the malignant phenotypes of breast cancer cells. CAA-derived Zeb1/Scd expression correlates with Ptgds clinically To further confirm the correlation between Zeb1/Scd-mediated lipid metabolic reprogramming with the malignant progression of breast cancer, we analyzed the expression of Scd and Ptgds in tumor tissues from MMTV-PyMT mice using immunohistochemistry. We found a significant upregulation of Scd expression in CAAs at the invasive front of PyMT tumors compared to the normal mammary adipocytes (Fig. [211]7a). On the other hand, the expression of Ptgds was increased, especially in CAA-adjacent breast cancer cells compared to that in distant cancer cells (Fig. [212]7b). Importantly, the expression level of CAA-derived Scd was positively correlated with Ptgds contents in adjacent breast cancer cells. Fig. 7. Relationship of Zeb1, Scd and Ptgds in PyMT tumors and human breast cancer samples. [213]Fig. 7 [214]Open in a new tab a, b Immunohistochemical staining for Scd and Ptgds in normal mammary adipocytes and CAAs in tumors from MMTV-PyMT spontaneous breast cancer mice (n = 12 PyMT mice; A: adipose; T: tumor; I: intra-tumor; B: tumor boundary; scale bars, 50 μm). c, d Immunohistochemical staining for SCD (c) and PTGDS (d) in normal breast adipocytes and CAAs in human breast cancer tissue (n = 40 patients; scale bars, 50 μm). e–g A direct association between the expression of ZEB1, SCD, and PTGDS in 40 human breast cancer samples (n = 40 patients). h–j A direct association between the expression of ZEB1, SCD, and PTGDS with the molecular typing of breast cancer (n = 40 patients). k–m A direct association between the expression of ZEB1, SCD, and PTGDS with the advanced TNM stages (n = 40 patients). Data were expressed as means ± SEM. a–d were assessed via two-tailed unpaired Student’s t-test. e–m were assessed via Spearman’s rank correction test. Source data are provided as a Source Data file. Furthermore, these results were verified in human breast cancer tissues, in which SCD expression in CAAs and PTGDS expression in adjacent cancer cells at the invasive front were significantly upregulated to perform a positive correlation (Fig. [215]7c, d). We also observed a negative correlation between the expression level of CAA-derived ZEB1 with those of SCD and PTGDS, whereas a positive correlation was indicated between SCD and PTGDS expression (Fig. [216]7e–g). Of note, the expression of CAA-derived ZEB1 is relatively decreased in the triple-negative breast cancer (TNBC), whereas those of SCD and PTGDS are remarkedly increased in these patients (Fig. [217]7h–j). Additionally, the expression level of CAA-derived Zeb1 was negatively correlated with advanced TNM stages; however, those of SCD and PTGDS exhibited a positive correlation (Fig. [218]7k–m). Together, these data indicated that the aberrant adipocyte-specific Zeb1/Scd signaling leads to the dysregulation of Ptgds-mediated ARA/PGD[2] metabolism, thus promoting the malignant progression of breast cancer clinically. POA-induced AM production reciprocally promotes Zeb1-depdent lipolysis To further elucidate the regulatory mechanism governing the crosstalk between CAAs and breast cancer cells in feedback, we stimulated EO771 cells with POA in the presence of AT-56 and collected their conditioned media to treat with 3T3^Adi cells. The Oil Red O staining and western blot analyses revealed that the conditioned media from POA-stimulated EO771 cells promoted lipolysis in 3T3^Adi cells, which was accompanied by a significant reduction in Zeb1 expression. However, pre-treatment with AT-56 attenuated these effects (Supplementary Fig. [219]12a, b). Similar results were observed in the experiments with MDA-MB-231 cells (Supplementary Fig. [220]12c, d), suggesting that POA downregulated Zeb1 expression in adipocytes via a Ptgds/PGD[2]-dependent paracrine mechanism to induce lipolysis. To further identify the underlying paracrine factors involved in this process, we performed RNA-sequencing using EO771 cells treated with POA. The results of GO database analysis combined with the gene identification related to extracellular components showed that the mRNA level of AM was strongly upregulated upon POA treatment (Supplementary Data [221]1). In line with this, the RT-PCR and ELISA assays confirmed that the addition of POA significantly promoted AM expression in EO771 cells, while AT-56 treatment abolished these outcomes (Supplementary Fig. [222]12e, f). We also performed the experiments in MDA-MB-231 cells and obtained similar results (Supplementary Fig. [223]12g, h). To explore whether AM is regulated by PGD[2], EO771 cells were treated with different concentrations of PGD[2]. Indeed, the results of RT-PCR and ELISA revealed that PGD[2] enhanced AM expression and secretion (Supplementary Fig. [224]12i, j), with similar findings from the experiments conducted in MDA-MB-231 cells (Supplementary Fig. [225]12k, l). PGD[2] has been reported to exert its biological functions by activating the downstream cAMP/PKA pathway^[226]40,[227]41. Taken together with the ChIP-sequencing analysis from the UCSC Genome Browser ([228]https://genome.ucsc.edu/) showing that Creb1 is a key upstream transcriptional regulator of AM, we hypothesized that PGD[2] might promote AM transcription by targeting the cAMP/PKA/Creb1 signaling pathway in breast cancer cells. Therefore, we pre-treated EO771 cells with an adenylate cyclase inhibitor SQ22536 to inhibit PGD[2]-induced production of cAMP. The results of RT-PCR and ELISA showed that pre-treatment with SQ22536 effectively reduced PGD[2]-induced AM expression in EO771 cells (Supplementary Fig. [229]12m, n). Similar outcomes were also obtained in MDA-MB-231 cells (Supplementary Fig. [230]12o, p). Moreover, we established a stable EO771 cell line with specific knockdown of Creb1 and confirmed that PGD[2] upregulated AM expression in a cAMP/PKA/Creb1-dependent manner (Supplementary Fig. [231]12q–t), which were further demonstrated in MDA-MB-231 cells (Supplementary Fig. [232]12u–x). Consequently, 3T3^Adi cells were treated with different concentrations of AM. The results of Oil red O staining and western blot showed that treatment with AM promoted the lipolysis in 3T3^Adi cells in a dose-dependent manner (Supplementary Fig. [233]13a), concurrent with a gradual decrease in Zeb1 expression (Supplementary Fig. [234]13b). Previous studies have shown that AM acts on the AMR receptors to induce lipolysis in adipocytes^[235]42. Therefore, we pre-treated 3T3^Adi cells with an AMR antagonist AMA, followed by AM stimulation (Supplementary Fig. [236]13c). The results of Oil Red O staining and western blot revealed that AMA pre-treatment significantly weakened AM-induced lipolysis (Supplementary Fig. [237]13d) and restored Zeb1 expression (Supplementary Fig. [238]13e). Since AMR is a complex of calcitonin receptor-like receptor (CRLR) and receptor activity-modifying protein (RAMP)^[239]43,[240]44, we established a stable 3T3-L1 cell line with the knockdown of CRLR (Supplementary Fig. [241]13f, g) to investigate the essential role of AM/CRLR signaling in Zeb1-mediated lipolysis in adipocytes. We found that the specific knockdown of CRLR significantly blocked AM-induced lipolysis in shCRLR/3T3^Adi cells, which was accompanied by the rescue of Zeb1 expression (Supplementary Fig. [242]13h–j). Next, we used a co-culture model of cancer cells with adipocytes to validate these findings (Supplementary Fig. [243]13k). Our results demonstrated that AMA treatment significantly blocked the cancer cell-adipocyte communication that triggered the lipolysis and decreased Zeb1 expression in adipocytes (Supplementary Fig. [244]13l–o). Similarly, co-culture of cancer cells with shCtrl/3T3^Adi could no longer reduce Zeb1 expression and trigger lipolysis in adipocytes (Supplementary Fig. [245]13p–t). Altogether, these findings indicated that, through binding to the AMR receptor on cell surface of adipocytes, AM specifically downregulated Zeb1 expression and thereby triggered lipolysis. Targeting Zeb1-deployed TAME inhibits breast tumorigenesis Based on the pivotal role of adipocyte-specific Zeb1/Scd signaling in fostering TAME of breast cancer in an AM-dependent mechanism, we hypothesized that the combinational treatment with the Scd inhibitor A939572 and the AMR antagonist AMA might disrupt the interaction between cancer cells and CAAs to eventually inhibit tumorigenesis. To test this, we constructed an allograft breast cancer model using EO771 cells, followed by intraperitoneal administration of A939572 (30 mg/kg) and AMA (5 μg/day) alone or in combination (Fig. [246]8a, b). We noticed that treatment with either A939572 alone or AMA alone effectively inhibited tumor growth and reduced tumor volume and weight by ~50–60%. Remarkably, the combination of A939572 and AMA had a pronounced synergistic antitumor effect, decreasing tumor volume and weight by ~80–90% (Fig. [247]8c, d). The HE staining revealed that CAAs at the invasive front of allograft tumors were significantly smaller and irregular in shape in the saline group, while CAAs in tumors from either A939572 or AMA monotherapy group showed alleviation of lipolysis. Strikingly, the combinational treatment group maintained unilocular lipid droplet phenotype in CAAs, with a clearer boundary at the invasive front of allograft tumors (Fig. [248]8e), accompanied with elevated expression of Zeb1 (Fig. [249]8f). The immunohistochemistry analysis further demonstrated the notable decrease in Ki67 (Fig. [250]8g) and Vimentin (Fig. [251]8h) expression levels in tumors by the combinational treatment. Importantly, no adverse effect, such as body weight loss, was observed in either treatment group (Fig. [252]8i). Overall, these findings highlighted that targeting the interaction between CAAs and their surrounding cancer cells with the combinational treatment of A939572 and AMA effectively impeded Zeb1-deployed lipid metabolic reprogramming within the TAME, which eventually blocked the malignant progression of breast cancer. Fig. 8. AMA combined with A939572 targets TAME to inhibit breast cancer progression. [253]Fig. 8 [254]Open in a new tab a Scheme showing combinational treatment of A939572 with AMA in mice. b EO771 allograft tumors in the mammary fat pads. c, d Tumor volume (c) and weight (d) (n = 6 mice). e HE staining (n = 6 mice; A adipose, T tumor; scale bars, 50 μm). f–h Immunohistochemical staining for Zeb1 (f), Ki-67 (g) and Vimentin (h) (n = 6 mice; scale bars, 50 μm). i Body weight of mice (n = 6 mice). Data were expressed as means ± SEM. c, i were analyzed via two-way ANOVA with Sidak correction for multiple comparisons. d–h were assessed via two-tailed unpaired Student’s t-test. Source data are provided as a Source Data file. Simultaneously, to elucidate the feedback mechanism underlying ARA/PGD[2]/AM-dependent membrane phospholipid remodeling in breast cancer cells, we therapeutically targeted the EO771 allograft breast cancer model using AMA (5 μg/day) and AT-56 (50 mg/kg) alone or in combination (Supplementary Fig. [255]14a, b). We observed that AMA monotherapy effectively suppressed tumor growth and reduced tumor volume and weight by ~40–50%, while AT-56 alone caused a reduction of about 20%-30%. Significantly, the combination of AMA and AT-56 exhibited a pronounced synergistic effect, decreasing tumor volume and weight by around 70%-80% (Supplementary Fig. [256]14c, d). The HE staining further illustrated that, in the saline control group, CAAs at the invasive front of allograft tumors were significantly reduced in size and presented irregular morphology. Conversely, CAAs in tumors from either AMA or AT-56 monotherapy group experienced decreased lipolysis. Notably, the combinational treatment group almost maintained CAAs with unilocular lipid droplets (Supplementary Fig. [257]14e), accompanied by an upregulation of Zeb1 expression (Supplementary Fig. [258]14f). The analysis of immunohistochemistry confirmed the marked reduction in Ki67 (Supplementary Fig. [259]14g) and Vimentin (Supplementary Fig. [260]14h) expression levels in tumors by combinational therapy. Additionally, no obvious adverse effect, such as body weight loss, was observed in either treatment group (Supplementary Fig. [261]14i). In all, these findings implied the efficacy of simultaneously targeting the Zeb1/Scd/POA signaling and its downstream ARA/PGD[2]/AM axis in suppressing breast cancer progression, which is possibly achieved by inhibition of either Zeb1-deployed lipolytic activity in CAAs or membrane phospholipid metabolism in breast cancer cells. Discussion The malignant progression of breast cancer predominantly depends on the interaction between cancer cells and their surrounding adipose tissue, emerging lipid metabolic reprogramming as a crucial pathological mechanism in orchestrating the TAME. Therefore, identification of the microenvironmental signals that contribute to this interplay in the TAME may translate into improved antineoplastic therapies. Based on our findings, we uncovered a Zeb1-deployed bidirectional communication between breast cancer cells and their adjacent CAAs (Fig. [262]9). At the molecular level, breast cancer cells, through the secretion of AM, induced downregulation of Zeb1 expression to activate the Atgl/Hsl/Scd-depedent lipolysis in CAAs, which subsequently resulted in the release of POA into the TAME. In turn, accumulation of POA triggered the malignant phenotypes of breast cancer cells and their production of AM via an ARA-PDG[2]-mediated membrane phospholipid replacement mechanism. Importantly, disruption of this microenvironmental paracrine action at any level, such as inhibition in Scd and/or AM activity, severely diminishes the aggressiveness of breast cancer in vitro and in vivo. Fig. 9. Schematic of Zeb1-mediated remodeling of the TAME. Fig. 9 [263]Open in a new tab Through Zeb1-deployed lipid metabolic reprogramming, CAAs at the invasive front might prime a maladaptive TAME that reciprocally confers breast cancer with aggressive and lethal properties, providing promising therapeutic approaches to limit local tumor growth and aggressive progression by eliminating this pro-tumorigenic interplay in the TAME (Some elements by figdraw.com). According to a growing body of studies, Zeb1, as a cancer cell-intrinsic transcriptional factor, modulates various biological behaviors of breast cancer through genetic and epigenetic mechanisms. However, the potential role of Zeb1 in specific types of tumor stromal cells has been poorly reported. Recent evidence have suggested that Zeb1 acts as the core transcriptional component of adipocyte differentiation, and its aberrant expression can markedly impact adipogenesis both in vitro and in vivo^[264]37,[265]38. In line with these, we established Zeb1^adiKO homozygous female mice with adipocyte-specific knockout of the Zeb1 gene, and phenotypic analyses highlighted that Zeb1 was essential for the maturation and differentiation of mammary gland adipocytes. Moreover, single nucleus RNA-sequencing profiling of Zeb1^adiKO mice confirmed that loss of adipogenic Zeb1 expression resulted in an increased proportion of progenitor adipocytes, including ADSCs and ProA, in the mammary tissues, along with a substantial decrease in the mature White1 cluster associated with lipid anabolic characteristics. These observations suggested that the absence of adipocyte-specific Zeb1 expression conferred the metabolic reprogramming of adipocytes with a lipolytic characteristics, which might be conducive to breast tumorigenesis. Indeed, at the molecular level, our approach combining transcriptomics with metabolomics unveiled that aberrant Zeb1 specifically targeted the transcriptional activity of a series of key enzymes, including Atgl, Hsl, and Scd, involved in lipid catabolic processes. This, in turn, resulted in paracrine delivery of the lipolytic products, such as POA, into adjacent cancer cells, ultimately driving the malignant progression of breast cancer in vitro and in vivo. Consistently, in a subset of breast cancer patients with aggressive phenotypes, we observed that significantly reduced Zeb1 expression was restricted to a subpopulation of CAAs at the invasive front, where breast cancer cells interact with their TAME. We also observed a negative correlation between the contents of CAA-derived Zeb1 and Scd, revealing that dysfunction of the adipose-specific Zeb1/Scd signaling might induce lipid metabolic reprogramming and thus confer the malignant progression of breast cancer. Of note, the association between dysregulation of fatty acid metabolism and breast cancer development exhibits strong biological plausibility. Especially, alterations in FFAs can impact numerous cellular processes, including cancer cell proliferation, differentiation, migration, as well as responsiveness to anticancer therapies^[266]7,[267]45–[268]48. For example, a case-control study within the European Prospective Investigation into Cancer and Nutrition (EPIC) has suggested that increased de novo lipogenesis, acting through increased synthesis of POA, could be a relevant metabolic pathway for breast tumorigenesis^[269]45. Consistently, breast cancer cells could promote the conversion of adjacent adipocytes into CAAs, which reciprocally reduce ferroptosis in cancer cells by secreting MUFAs^[270]7. Our study further provided an alternative mechanism demonstrating that CAAs also produced elevated amounts of POA via the Zeb1/Scd pathway. Upon entering breast cancer cells, POA specifically displaced ARA at the sn-2 position of phospholipids in the cell membrane. This subsequently led to the intracellular metabolism of ARA into PGD[2], which ultimately facilitated the malignant growth and metastasis of breast cancer cells. In line with our observations, extensive evidence have indicated that tumor progression in vivo requires the Scd activity, further highlighting the significance of MUFAs in promoting cancer progression^[271]49–[272]52. On the other hand, POA-induced intracellular accumulation of PGD[2] subsequently promoted the secretion of AM by activating the cAMP/PKA/Creb1 signaling in breast cancer cells. It has been well documented that AM, as a pleiotropic hormone structurally similar to the calcitonin gene-related peptide, is abundantly present in the normal adrenal medulla, serving as a crucial endocrine factor with various physiological and pathological functions^[273]53–[274]57. In human cancers, AM is predominantly produced by cancer cells and exerts multiple tumorigenic effects within the tumor microenvironment^[275]58–[276]61. For example, Greillier et al. reported that AM induces cancer cell proliferation, migration, and invasion by activating the CRAF/MEK/ERK/MAPK pathway in pleural mesothelioma under hypoxic conditions^[277]58. Moreover, the aberrant PI3K/AKT/GSK3β signaling upregulates AM expression in renal cell carcinoma cells, which subsequently recruits mast cells in the TME to promote tumor angiogenesis^[278]59. In the present study, our results provided further evidence that breast cancer cells secrete substantial amounts of AM to bind to its membrane receptor AMR on adjacent CAAs in a paracrine action, thus performing a pro-tumorigenic function through Zeb1-deplyed lipid metabolic remodeling. Of note, AM has been shown to exert differential effects on lipolysis depending on the cellular context^[279]62–[280]64. For example, studies in murine 3T3-F442A cells have revealed that AM acts in an autocrine/paracrine action to inhibit lipolysis by extracellular inactivation of isoproterenol, which is a β-adrenergic agonist^[281]62. However, cancer cell-secreted exosomal AM activates the p38 and ERK1/2 kinases to promote lipolysis in subcutaneous adipose tissue, leading to the development of diabetes in pancreatic cancer^[282]63. This is consistent with our notion showing that breast cancer cell-derived AM contributed to lipolysis in CAAs at the invasive front, which in effect was mediated by switching Zeb1-dependent transregulation on a cascade of lipolytic genes. Despite the fact that further investigation is required to fully elucidate the underlying mechanism of Zeb1 downregulation by AM, our results have positioned Zeb1 as a key modulator involved in triggering lipid metabolic remodeling in breast cancer. Taken together with our observation showing that the combinational treatment with the Scd inhibitor A939572 and AM antagonist AMA synergistically reduces the development of breast cancer in allograft models, our data have together indicate that disruption of the interaction between cancer cells and their surrounding adipose tissue would effectively diminish the aggressiveness of breast cancer in vitro and in vivo. These could be achieved by suppressing Zeb1-dependent lipolytic activity and/or membrane phospholipid metabolism within the TAME. In summary, our findings have demonstrated that through a mechanism of Zeb1-deployed lipid metabolic reprograming, CAAs at the invasive front might prime a maladaptive TAME that reciprocally confers breast cancer cells with aggressive and lethal properties. Importantly, we have also unraveled the functional complexity and heterogeneity of Zeb1 within breast cancer cells and their surrounding adipocytes, which brings new strategies for targeted intervention. Taken together with our findings concerning the differential role of Zeb1 in major components of the TAME, targeting the adipo-specific Zeb1/Scd/AM axis might provide new insights into developing effective and precise treatment strategies for breast cancer. Therefore, our study have introduced promising therapeutic approaches to improve clinical outcomes for patients with advanced breast cancer by eliminating this pro-tumorigenic interplay in the TAME to limit local tumor growth and aggressive progression. Methods Ethics statement All of the experimental procedures involving animals were performed in accordance with a protocol that was approved by the Ethics Committee for Animal Use at the Medical College of Nankai University (2022-SYDWLL-000096). This paper describes studies involving human tissue samples that were approved by the Ethics Committee of Tianjin Medical University Cancer Hospital (approval number: NKUIRB2024059). Mice and mice housing To generate the conditional Zeb1 knockout allele (genotype Zeb1^fl/+), Loxp sites were added to both sides of exon 6 to remove the sequence that encodes a large portion of the protein and induce premature translation halt. We crossed conditioned Zeb1 knockout mice (genotype Zeb1^fl/fl) with adipocyte-specific Fabp4-Cre mice (genotype Fabp4-Cre). The homozygous female mice with an adipocyte-specific deletion of Zeb1 were identified (genotype Fabp4-Cre^+; Zeb1^fl/fl), which was named as Zeb1^adiKO. We generated conditional Zeb1 knock-in mice (genotype B6-CAG-LSL-Zeb1), which were crossed with Fabp4-Cre mice to generate Zeb1^adiTG mice with adipocyte-specific Zeb1 knock-in (genotype Fabp4-Cre^+; Zeb1^+/+). The Zeb1^fl/+ mouse, Zeb1^TG mouse, FABP4-Cre mouse (C57BL/6 background) were all purchased from GemPharmatech. The MMTV-PyMT spontaneous breast cancer mouse (C57BL/6 background) was purchased from Cyagen. All mice were placed in a temperature-controlled room (22 ± 2 °C) with a humidity of 40–60% and a light/dark period of 12 h/12 h. The mice were fed with transgenic mouse-specific feed (spfbiotech, SPF-F04-001). The mice were treated according to the protocols approved by the Animal Care and Use Committees of Medical College of Nankai University and Institute of Radiation Medicine of Chinese Academy of Medical Science. Cell culture EO771 murine breast cancer cells (ATCC, Cat# CRL-3461) were cultured in DMEM (high glucose. Servicebio, Cat# G4515) supplemented with 10% FBS (VivaCell, Cat# C04001-500) and 1% Pen/Strep (Servicebio, Cat# G4003), and MDA-MB-231 human breast cancer cells (ATCC, Cat# CRM-HTB-26) were cultured in RPMI-1640 supplemented with 10% FBS and 1% Pen/Strep. HEK293T cells (ATCC, Cat# CRL-3216) were cultured in DMEM supplemented with 10% FBS, 1% sodium pyruvate, 1% NEAA, and 2% glutamine. Mature adipocyte generation in vitro and Oil Red O staining Preadipocytes 3T3-L1 (ATCC, Cat# CL-173) were cultured in DMEM containing 10% calf serum (Gibco, Cat# 2500219), and when the fusion degree of cells in 6-well or 24-well plate was >90%, differentiation was induced by an adipogenic cocktail containing 5 μg/mL Insulin (Sigma-Aldrich, Cat# I2643), 1 μM Dexamethasone (Sigma-Aldrich, Cat# D4902), 1 mM Rosiglitazone (Sigma-Aldrich, Cat# R2408), and 0.5 mM 3-Isobutyl-l-methylxanthine (Macklin, Cat# I811775) for 2 days. The cells were then maintained in DMEM containing 5 μg/mL insulin and switched to fresh medium every 2 days. Adipocytes morphology was observed by phase-contrast microscopy and Oil Red O staining (Beyotime, Cat# C0158s). For Oil Red O staining, the adipocytes generated by the in vitro differentiation system were fixed with 4% paraoxymethylene for 20 min, washed with phosphate-buffered saline (PBS), then stained with Oil Red O solution for 20 min and covered with PBS. The staining is examined under a microscope. Cell co-culture and conditioned medium (CM) production Cancer cells and adipocytes were co-cultured using transwell system (pore size 0.4 μm; Falcon). Briefly, adipocytes were seeded on the bottom layer and induced to form lipids. Then, cancer cells were seeded on the upper layer of transwell and cultured for 6 days. Adipocytes or cancer cells cultured alone under similar conditions were served as controls. Cancer cells or adipocytes cultured in 6-well plates were replaced with DMEM containing 0.2% FBS for 24 h. Then, CM were collected, filtered by 0.22 μm filter, and stored at −80 °C. To produce CM by CAAs, mature adipocytes were co-cultured with cancer cells for 6 days in the transwell system. The upper layer of cancer cells was then removed, and the bottom layer of activated CAAs were cultured in fresh DMEM containing 0.2% FBS for another 24 h followed by CM collection. Human sample Forty patients with invasive breast cancer were obtained adipose tissue adjacent to or distant from the tumor (>5 cm) during surgery. The pathological verification, molecular classification, and clinical stage of the participants were collected (Table [283]S1). The obtained tissue was fixed with 4% paraformaldehyde. Genotyping PCR was performed with DNA from the tail biopsy. All mice were genotyped to assess FABP4-Cre and Zeb1 gene expression. Genotyping PCR primers are shown in Table [284]S2. Single nucleus RNA-sequencing 1. Nuclei isolation sorting from mouse mammary adipose tissue We first stripped the fourth pair of mouse mammary tissue, removed the mammary lymph nodes, and then washed in pre-cooled PBSE (PBS buffer containing 2 mM EGTA). Nuclei isolation was carried out using GEXSCOPE® Nucleus Separation Solution (Singleron Biotechnologies, Nanjing, China), refer to the manufacturer’s product manual. Isolated nuclei were resuspended in PBSE to 10^6 nuclei per 400 μl, filtered through a 40 μm cell strainer, and counted with Trypan blue. Nuclei enriched in PBSE were stained with DAPI (1:1,000) (TermoFisher Scientifc, D1306). Nuclei were defined as DAPI-positive singlets. 2. Single nucleus RNA-sequencing library preparation The concentration of single nucleus suspension was adjusted to 3–4 × 10^5 nuclei/mL in PBS. Single nucleus suspension was then loaded onto a microfluidic chip (GEXSCOPE® Single Nucleus RNA-seq Kit, Singleron Biotechnologies), and single nucleus RNA-seq libraries were constructed according to the manufacturer’s instructions (Singleron Biotechnologies). The resulting single nucleus RNA-seq libraries were sequenced on an Illumina Novaseq 6000 instrument with 150 bp paired end reads. 3. Primary analysis of raw read data Raw reads from single nucleus RNA-sequencing were processed to generate gene expression matrixes using CeleScope ([285]https://github.com/singleron-RD/CeleScope) v1.9.0 pipeline. Briefly, raw reads were first processed with CeleScope to remove low quality reads with Cutadapt v1.17 to trim poly-A tail and adapter sequences. Cell barcode and UMI were extracted. After that, we used STAR v2.6.1a to map reads to the reference genome GRCm38 (ensembl version 92 annotation). UMI counts and gene counts of each cell were acquired with feature Counts v2.0.1 and used to generate expression matrix files for subsequent analysis. Plasmid construction cDNA fragment encoding the full-length Zeb1 sequence was prepared by PCR and cloned into pLV-EF1-MCS-IRES-Bsd (Biosettia, Cat# cDNA-pLV03) and pCW57.1 (Addgene, Cat# 41393). The lentivirus-based vector pLV-H1-EF1α-puro (Biosettia, Cat# SORT-B19) was used to express shRNA in cells. The lentivirus-based vector Tet-pLKO-puro (addgene, Cat# 21915) was used for inducible expression of shRNA. Murine Atgl promoter (−2960/ + 99), Hsl promoter (−3000/ + 8), Mgl promoter (−2997/ + 103), and Scd promoter (−2746/ + 300) were obtained from genomic DNA by PCR. Mutagenesis of Mut-I and Mut-II in the murine Scd promoter was performed using a QuikChange® Lightning Site-Directed Mutagenesis kit (Stratagene, Cat# 210518). Primer sequences are listed in Table [286]S2–[287]3. Lentivirus generation Lentiviruses were produced by transfecting subconfluent HEK293T cells with lentiviral vectors and packaging plasmids with calcium phosphate transfection. Virus supernatants were collected 48 h after transfection, centrifuged at 75,000 × g for 90 min, suspended, and filtered through a 0.45 μm filter (Millipore, Cat# SLHV033RB). Whole-mount mammary gland staining The fourth mammary gland was surgically removed^[288]65, extended onto a slide, and immobilized overnight in Carnoy fixative (ethanol: chloroform: acetic acid, 6:3:1) at room temperature. The glands were then rehydrated in descending alcohol (70%, 35%, and 15%) for 10 min per step, washed with TBST for 5 min, and then stained overnight in carminate alum solution (Sigma-Aldrich, Cat# 1390-65-4) at 4 °C. In addition, the mammary glands were dehydrated for 10 min by a series of graded ethanol solutions (50%, 70%, 85%, 95%, and 100% alcohol), degreased overnight in a clear solution, and sealed with neutral resin. Uptake of BODIPY-C16 Cells were incubated with 100 nM BODIPY-C16 (Thermofisher, Cat# D3821) for 30 min, the medium was removed, then washed with PBS for 3 times, and finally covered with PBS for observation and imaging under a fluorescence microscope at 515 nm. RNA extraction and RT-PCR For the collection of mouse mammary adipose tissue, we first stripped the fourth pair of mouse mammary tissue, removed the mammary lymph nodes, separated the mammary adipose cells, added TRIzol reagent (YEASEN, Cat# 10606ES60) at a ratio of 1 ml trizol per 50-100 mg tissue. For the collection of cells, we digested the adherent cells with pancreatic enzymes, centrifuge them, gently wash them with normal saline or 1 × PBS twice, discard the liquid, and add TRIzol at a ratio of 1 ml Trizol per 1 × 10^6 cells. Total RNA was extracted using TRIzol. cDNA was then synthesized using reverse transcriptase (Yeasen, Cat# 11300ES92). Quantitative PCR amplification was performed using the SYBR Green Q-PCR SuperMix Kit (Yeasen, Cat# 11201ES03). β-actin was used as the normalization control. Primer sequences are listed in Table [289]S4. The experiment was run using LightCycler 480 (Roche). RNA sequencing and data processing RNA sequencing was carried out according to the assembly line of BGI-tech. Briefly, total RNA was isolated with Trizol reagent. Total RNA was enriched for mRNA, and poly-a-tailed mRNA was enriched with OligodT magnetic beads. The resulting RNA was segmented with interrupt buffer, and the randomLy obtained N6 primers were reverse-transcribed, and cDNA double-stranded DNA was synthesized to form double-stranded DNA. The synthesized double-stranded DNA was flattened and phosphorylated at the 5′-end, forming a sticky end that protrudes “A” at the 3′ end, followed by a bubble-like linker that protrudes “T” at the 3′ end. PCR amplification with specific primers. Preliminary sequencing data was generated by MGISEQ-2000. After quality control, the raw readings were filtered into clean readings, which were then processed through a top cap and cufflink algorithm. GSEA was performed by R (Cluster Profiler package) with the following parameters: nPerm = 1000, minGSSize = 10, maxGSSize = 500, P-valueCutoff = 0.05, pAdjustMethod = “BH”. Western blot analysis Cell lysates were prepared with protease inhibitors and phosphatase inhibitors in RIPA buffer (Yeasen, Cat# 20115ES60). Protein concentration was determined with BCA (Yeasen, Cat# 20201ES76). Protein Marker (Epizyme, Cat# WJ103; Aladdin, Cat# rp192041) was used to monitor the migration of proteins in SDS-polyacrylamide gel and monitor the membrane transfer efficiency of proteins. Equal amounts of total protein were separated in SDS-polyacrylamide gel electrophoresis and then transferred to PDVF membranes (Millipore, Cat# IPVH00010) and blocked. PVDF membrane was incubated at 4 °C overnight with primary antibody and then at room temperature with secondary antibody for 1 h. Labeled proteins were visualized by ECL Chemiluminescence Kit (Yeasen, Cat# 36208ES76). Densitometric analysis of western blot imaging was performed by G:Box Chemi XRQ gel doc system (Syngene). Reagents and antibodies Human recombinant AM (Cat# GC35258) and AMA (Cat# GC32615) proteins were purchased from Glpbio. Palmitoleic acid (Cat# GC31253), Oleic acid (Cat# GC30110) were purchased from Glpbio. Etomoxir (Cat# T4535), A939572 (Cat# T4515), FATP1-IN-1 (Cat# [290]T38905) and SQ22536 (Cat# 17318-31-9) were purchased from [291]TargetMol (USA). AT-56 (Cat# A872100), Atglistatin (Cat# A860897) and PGD[2] (Cat# P923874) were purchased from Macklin. Antibodies used for western blot (WB), immunohistochemistry (IHC) and Immunofluorescence (IF) were as follows: anti-Zeb1 (WB, Proteintech, Cat# 21544-1-AP, dil: 1/1000) (IHC, abcam, Cat# ab87280, dil: 1/200), anti-Crlr (WB, SAB, Cat# 37509, dil: 1/1000), anti-Creb1 (WB, Proteintech, Cat# 12208-1-AP, dil: 1/1000), anti-Scd (IHC and WB, SAB, Cat# 37896, WB dil: 1/1000, IHC dil: 1/50), anti-Ptgds (IHC and WB, SAB, Cat# 31107, WB dil: 1/1000, IHC dil: 1/50), anti-Perilipin-1 (IF, Genetex, GTX634406, dil: 1/5000), anti-Vimentin (IHC and WB, Cell Signaling Technology, Cat# 5741, IHC dil: 1/500, WB dil: 1/1000), anti-snail (WB, Proteintech, Cat# 13099-1-AP, dil: 1/1000), anti-Ki67 (IHC, Servicebio, Cat# [292]GB151142, dil: 1/500), and anti-actin (WB, Santa Cruz, Cat# sc-47778, dil: 1/1000), anti-Atgl (WB, Santa Cruz, Cat# sc-365278, dil: 1/1000), anti-Hsl (WB, SAB, Cat# 41039, dil: 1/1000) and anti-Mgl (WB, Santa Cruz, Cat# sc-398942, dil: 1/1000). Chromatin immunoprecipitation (ChIP) According to the manufacturer’s instructions, the EZ-ChIP kit (Millipore, Cat# 17-371FR) was used for ChIP testing. Cells were crosslinked with 1% formaldehyde at room temperature for 10 min, and then 125 mM glycine was added to inactivate the formaldehyde. Immunoprecipitation of chromatin extracts containing DNA fragments were performed using specific antibodies. The ChIP-enriched DNA was then uncrosslinked, and quantitative PCR was performed. The primers and antibodies used in these experiments are shown in Table [293]S2. Luciferase assay The cells were transfected with wild-type murine Atgl, Hsl, Mgl, Scd promoter, or mutant murine Scd promoter, followed by treatment with full-length Zeb1. The lysate was prepared 36 h after transfection and the luciferase activity was measured using the dual luciferase Reporting Analysis System (Promega, Cat# J3082) according to the manufacturer’s protocol. The luciferase activity was normalized to renilla luciferase value. EdU incorporation assay Cells were cultured with 20 μM EdU ([294]RiboBio, Cat# C10310-1) for 2 h, fixed with 4% paraformaldehyde at room temperature for 30 min, washed with 2 mg/mL glycine for 5 min, and then washed twice with PBS containing 0.5% Triton X-100. 100 μL Apollo 567 staining reaction buffer was added to cells and incubated in a cassette for 30 min. Cells were then washed three times with PBS containing 0.5% Triton X-100 and stained at room temperature with 100 μL Hoechst 33342 (5 mg/mL) for 30 min. Photographs were taken under a fluorescence microscope. Cell viability assay 2 × 10^3 cells/well were inoculated in 96-well plates and cultured for the indicated time points. Cell viability was then measured using the cell counting kit 8 (Yeasen, Cat# 40203ES88) assay according to the manufacturer’s instructions. Absorbance is measured at 450 nm. Each group is assigned three parallel wells. Transwell assay Matrigel (Yeasen, Cat# 40183ES08) was placed in the upper layer of the transwell chamber (pore size 8 μm; Falcon, Cat# 3464) and incubated for 0.5 h, then cells were placed in the upper chamber. and allowed to invade. After 16 h, the non-invaded cells were removed with a cotton swab, and the invaded cells were fixed with 4% paraformaldehyde and stained with crystal violet dye. The cells were counted and imaged under an optical microscope (Olympus). High-content imaging Single-cell migration was measured using high-content imaging and Harmony analysis system (PerkinElmer). 1 × 10^5 cells were incubated in 24-well plates at 37 °C for 24 h with or without different treatments. We used multidimensional automatized microscopes with an environmental chamber to keep temperature, humidity, and CO[2] constant. Prewarmed medium was added before the start of imaging. The time of image acquisition was 15–17 h. Extraction of primary cells from MMTV-PyMT tumors Euthanasia was performed on 10-week-old MMTV-PyMT female mice (C57BL/6 background), and the tumor tissues were dissected and removed in a super-clean workbench, while the surrounding normal tissues were removed and the tumor were retained. Tumor was placed in a sterile container containing pre-cooled PBS. Use surgical scissors and tweezers to cut the tumor into the smallest possible lumps (about 1 mm³), and then use a knife to chop. The chopped tumor tissue was transferred to a sterile tube containing collagenase, gently mixed, and incubated in a 37 °C water bath for 30 min, shaking every 10 min. After digestion is complete, an appropriate amount of complete medium was added to stop the digestive reaction. The suspension was filtered using a 70 μm cell screen to collect the single-cell suspension and remove large undigested tissue. The filtered cell suspension was centrifuged (1000 rpm, 5 min) and the supernatant was discarded. The cells were inoculated into the petri dish at the appropriate density, adding the appropriate amount of complete medium. The inoculated cells were cultured in an incubator at 37 °C and 5% CO[2]. Change the medium periodically and observe cell growth until the desired degree of confluence is reached. Tumor cell implantation 8-weeks-old female mice (C57BL/6 background) were used. For mammary fat pad injection, mice were anesthetized with 2.5% isoflurane. Following anesthesia, a small incision was made with scissors between the fourth nipple and the midline, and then the mammary fat pad was extracted using forceps. 50 µL of E0771/PyMT cell suspension (10 million cells/mL) mixed with 25 µL cooled matrigel was injected into the mammary fat pad by holding the needle horizontally. Skin was sutured with absorbable 6.0 silk suture. The maximal tumour size/burden permitted was 1000 mm^3 or 10% body weight, and this burden was not exceeded. After surgery, animals were monitored until recovery in a chamber on a heating pad. Tumors were allowed to develop for 7 days. The mice were then given intraperitoneal injections of 30 mg/kg A939572 (once every 2 days) and/or 5 μg/day AMA; 50 mg/kg AT-56 (once every 2 days) and/or 5 μg/day AMA for 3 weeks. The mice were then euthanized, and tumors were analyzed for volume and weight. Experimental lung metastasis model of breast cancer 8-weeks-old female mice (C57BL/6 background) were used. Firstly, EO771/PyMT cells were treated with CM-CAA^Zeb1-off and CM-CAA^Zeb1-on conditioned media for 4 days, then digested with trypsin, resuspended in PBS, and prepared for tail vein injection at a concentration of 10 million cells/mL, with 50 μL injected per mouse. For the tail vein injection, the tail was first immersed in warm water (approximately 37 °C) for a few seconds to dilate the blood vessels. 1 mL sterile syringe fitted with a 27–30 G needle was used to inject into the vein at the end of the tail. The predetermined volume of cell suspension was slowly injected, ensuring smooth administration without extravasation. After injection, the injection site was gently massaged to promote distribution. Mice were monitored until they fully regained consciousness and could move normally before being returned to their cages. 30 days later, the animals were euthanized, and lung tissues were harvested for histopathological section analysis to evaluate the number and size of metastatic nodules. India ink assay After euthanizing the mice, 5 mL of India ink (Solarbio, Cat# I8060) was injected into the lungs of the mice (15% v/v PBS dilution). The lungs were then removed and fixed overnight at 4 °C with Fekete solution (1 L, containing 880 mL 70% ethanol, 80 mL 37% formaldehyde and 40 mL glacial acetic acid). Pulmonary metastatic nodules were counted manually. Immunohistochemistry (IHC) analysis Forty pairs of normal breast tissues and breast cancer tissues were collected from Tianjin Medical University Cancer Institute and Hospital (Tianjin, China). The IHC analysis of paraffin-embedded sections were conducted following the manufacturer’s protocol. Slices were boiled in the extract to expose the antigen. Specific antibodies were applied to the sections. The slides were stained with hematoxylin, dehydrated, and mounted. Immunostaining was independently evaluated by two pathologists. The IHC score was calculated by combining the quantity score (the percentage of positive-stained areas) with staining intensity score. The quantity score ranges from 0 to 4: 0, no immunostaining; 1, 1–25% of the areas are positive; 2, 26–50% of the areas are positive; 3, 51–75% of the areas are positive; and 4, >75% of the areas are positive. The staining intensity was scored as follows: 0 (no color), 1 (light yellow), 2 (light brown), 3 (brown), and 4 (dark brown). The score for each tissue was calculated by multiplying the intensity and quantity scores (the range of this calculation was therefore 0–12). Samples with an IHC score >6 were classified as high expression, and those with an IHC score ≤6 were classified as low expression. Immunofluorescence analysis The tissue sections inoculated on a cover slide were permeated with 0.4% Triton X-100 at room temperature for 15 min, sealed with 5% goat serum at room temperature for 1 h, and incubated with primary antibody at 4 °C overnight. The sample was then washed by PBS and incubated at room temperature for 1 h with the appropriate Alexa Fluor 488 (1:200, zSBG-BIO, Cat# ZF-0511) or Alexa Fluor 594 coupled secondary antibody (1:200, zSBG-BIO, Cat# ZF-0516). The nuclei were stained with DAPI (Beyotime, Cat# C1002) at room temperature for 10 min. The images were obtained with an BX53 microscpoe (Olympus) and analyzed with ImageJ (1.42q). Enzyme-linked immunosorbent assay The cells were seeded into a six-well plate. As the cells grown to 80% confluence, they were cultured with fresh serum-free medium, and the supernatant was collected for 24 h. The Mouse AM ELISA kit (Cat# SBJ-M0203), Mouse TG ELISA kit (Cat# SBJ-0617), Mouse Atgl (Cat# SBJ-M0766), Hsl (Cat# SBJ-M1350), Mgl (Cat# SBJ-M1351), Scd (Cat# SBJ-M1177) enzyme activity kit were purchased from Senbeijia. Samples were diluted with a dilution buffer, then 50 μL of biotin-labeled trapping antibodies were added to each well and incubated at 37 °C for 1 h. Wash the orifice plate with washing buffer 3–5 times for 1 min each time. A streptavidin solution of 100 μL was then added to each well, and the plates were incubated at 37 °C for 30 min. Wash the orifice plate with washing buffer 3–5 times for 1 min each time. Finally, 100 μL TMB One-Step Substrate Reagent was added to each well and incubated at 37 °C for 15 min. The reaction was terminated by adding 100 μL Stop Solution to each hole and OD values were immediately measured at 450 nm. Untargeted lipidomics 1. Sample preparation and lipid extraction 3T3^Ctrl and 3T3^Tet cells were seeded in 6-well plates, and respectively subjected to lipogenic differentiation, followed by co-culture with EO771 cells to establish CAAs termed as CAA^Zeb1-off and CAA^Zeb1-on cells (n = 5 independent experiments). CAA^Zeb1-off and CAA^Zeb1-on cells were washed three times with pre-chilled PBS, followed by washing with 0.9% sodium chloride solution. After aspirating the sodium chloride solution, cells were scraped off in methanol, rapidly frozen in liquid nitrogen, and stored at −80 °C, and sent to Shanghai Applied Protein Technology Co., Ltd. (Shanghai, China) for testing. Lipids were extracted according to MTBE (Methyl Tertiary Butyl Ether) method. Briefly, samples were first spiked with appropriate amount of internal lipid standards and then homogenized with 200 μL water and 240 μL methanol. After that, 800 μL MTBE was added and the mixture was ultrasound 20 min at 4°C followed by sitting still for 30 min at room temperature. The solution was centrifuged at 14,000 × g for 15 min at 10 °C and the upper organic solvent layer was obtained and dried under nitrogen. 2. LC-MS/MS method for lipid analysis Reverse phase chromatography was selected for LC separation using CSH C18 column (1.7 µm, 2.1 mm × 100 mm, Waters). The lipid extracts were re-dissolved in 200 μL 90% isopropanol/ acetonitrile, centrifuged at 14,000 × g for 15 min, finally 3 μL of sample was injected. Solvent A was acetonitrile–water (6:4, v/v) with 0.1% formic acid and 0.1 mM ammonium formate, and solvent B was acetonitrile–isopropanol (1:9, v/v) with 0.1% formic acid and 0.1 mM ammonium formate. The initial mobile phase was 30% solvent B at a flow rate of 300 μL/min. It was held for 2 min, and then linearly increased to 100% solvent B in 23 min, followed by equilibrating at 5% solvent B for 10 min. Mass spectra was acquired by Q-Exactive Plus in positive and negative mode, respectively. ESI parameters were optimized and preset for all measurements as follows: Source temperature, 300 °C; Capillary Temp, 350 °C, the ion spray voltage was set at 3000 V, S-Lens RF Level was set at 50%; and the scan range of the instruments was set at m/z 200-1800. 3. Identification by lipid search “Lipid Search” is a search engine for the identification of lipid species based on MS/MS math. Lipid Search contains more than 30 lipid classes and more than 1,500,000 fragment ions in the database. Both mass tolerance for precursor and fragment were set to 5 ppm. Targeted lipidomics 1. Sample preparation and lipid extraction EO771 cells were seeded in 10-cm dishes, treated with POA (5 μM) for 2 days (n = 3 independent experiments), then washed three times with pre-chilled PBS, followed by washing with 0.9% sodium chloride solution. After aspirating the sodium chloride solution, cells were scraped off in methanol, rapidly frozen in liquid nitrogen, and stored at −80 °C, and sent to Shanghai Applied Protein Technology Co., Ltd.,(Shanghai, China) for testing. Lipids were extracted according to MTBE method. Briefly, samples were first mixed with 200 μL methanol, and then 20 μL internal lipid standards and 800 μL MTBE were added. The mixture was adequately vortexed, sonicated for 20 min at 4 °C, and then kept for 30 min at room temperature. After that, 200 μL of MS-grade water was added, and the mixture was vortexed and centrifuged at 14,000 rpm for 15 min at 4 °C. The upper organic solvent layer was obtained and dried under nitrogen. For LC-MS analysis, the samples were re-dissolved in 200 μL of IPA/ACN (9:1, v/v) solvent and centrifuged at 14,000 rpm at 4 °C for 15 min, then the supernatant was injected. 2. LC-MS/MS method for lipid analysis The analysis was performed on a UHPLC system (Nexera LC-30A, Shimadzu) coupled with QTRAP MS (6500 + , Sciex). The analytes were separated on HILIC (Phenomenex, Luna NH2, 2.0 mm × 100 mm, 3 µm) and C18 column (Phenomenex, Kinetex C18, 2.1 × 100 mm, 2.6 μm). For RPLC separation, the column temperature was set at 45 °C. Mobile phase A: 70% acetonitrile + 30% H[2]O + 5 mM ammonium acetate, mobile phase B: IPA solution. A gradient (20% B at 0 min, 60% B at 5 min, 100% B at 13 min, 20% B at 13–17 min) was then initiated at a flow rate of 0.35 mL/min. The sample was placed at 10 °C during the whole analysis process. For HILIC separation, the column temperature was set at 40 °C. Mobile phase A: 2 mM ammonium acetate +50% methanol + 50% acetonitrile, mobile phase B: 2 mM ammonium acetate +50% acetonitrile + 50% water. A gradient (3% B at 0–3 min, from 3% to 100% B at 3–13 min, 100% B at 13–17 min, 3% B at 17–22 min) was then initiated at a flow rate of 400 μL/min. 6500 + QTRAP (AB SCIEX) was performed in positive and negative switch mode. The ESI positive source conditions were as follows: Source temperature: 400 °C; Ion Source Gas1 (GS1): 50 Ion Source Gas2 (GS2): 55; Curtain Gas (CUR): 35; IonSpray Voltage (IS): +5500 V； The ESI negative source conditions were as follows: Source temperature: 400 °C; Ion Source Gas1 (GS1): 50; Ion Source Gas2 (GS2): 55; Curtain gas (CUR): 35; IonSpray Voltage (IS): −4500 V. MRM method was used for mass spectrometry quantitative data acquisition. The MRM ion pairs are shown in the attached file. A polled quality control (QC) samples were set in the sample queue to evaluate the stability and repeatability of the system. 3. Data processing MultiQuant or Sciex OS was used for quantitative data processing. The QCs were processed together with the biological samples. Metabolites in QCs with coefficient of variation (CV) less than 30% were denoted as reproducible measurements. Targeted metabolomics—medium and long-chain fatty acids 1. Sample preparation and lipid extraction 3T3^Ctrl and 3T3^Tet cells were seeded in 6-well plates and respectively subjected to lipogenic differentiation, followed by co-culture with EO771 cells to establish CAAs termed as CAA^Zeb1-off and CAA^Zeb1-on cells. Subsequently, the upper layer of EO771 cells was removed, and the adipocytes were incubated with DMEM low glucose medium containing 2% FBS for another 2 days (n = 6 independent experiments). The supernatants from each group were then collected, rapidly frozen in liquid nitrogen, and stored at −80 °C, and sent to Shanghai Applied Protein Technology Co., Ltd. (Shanghai, China) for testing. Upon slow thawing at 4 °C, an appropriate volume of each sample was mixed with 5 mL of dichloromethane-methanol solution (2:1 v/v), vortexed, then ultrasounded for 30 min at a low temperature, and washed with 2 mL of gold-labeled water. The lower layer was collected and evaporated to dryness under a stream of nitrogen. Following this, 2 mL of n-hexane was added along with an internal standard, and methylation was carried out for 0.5 h. Next, 2 mL of gold-labeled water was added, and 2000 μL of the resulting supernatant was aspirated and dried under nitrogen. The residues were resuspended in n-hexane, and the supernatants were transferred into injection vials for GC-MS analysis. An injection volume of 1 μL was used with a split ratio of 10:1 for split injection into the GC-MS system. 2. Chromatography-Mass Spectrometry (GC-MS) detection conditions Chromatographic conditions Samples were separated with a capillary column (Agilent 19091S—433UI: HP-5ms, dimensions 30 m x 250 μm x 0.25 μm) in a gas chromatography system. The temperature program was as follows: initial temperature at 80 °C; ramped at 20 °C/min to 180 °C and held for 8 min; then the temperature was increased at 5 °C/min to 280 °C and maintained for 3 min. Helium served as the carrier gas at a flow rate of 1.0 mL/min. To monitor the stability and reproducibility of the system, the quality control (QC) samples were inserted into the sample queue at regular intervals among experimental samples. Mass Spectrometric Conditions: Mass spectrometry analysis was conducted using a 5977B MSD mass spectrometer (Agilent). The operational parameters for the 5977B MSD were set as follows: inlet temperature at 280 °C; ion source temperature at 230 °C; and transfer line temperature at 250 °C. The energy was 70 eV in electron impact mode. The mass spectrometry data were acquired in Scan /SIM mode. 3. Data analysis The chromatographic peak areas and retention times were extracted using the MSD ChemStation software. Standard curves were plotted, and the content of medium and long-chain fatty acids in the samples was calculated. Targeted metabolomics—arachidonic acid 1. Sample preparation and lipid extraction EO771 cells were seeded in 10-cm dishes, treated with POA (5 μM) for 2 days (n = 3 independent experiments), then washed three times with pre-chilled PBS and subsequently washed with 0.9% sodium chloride solution. After aspirating the sodium chloride solution, cells were scraped off in the presence of methanol/acetonitrile/water (2:2:1, v/v/v), rapidly frozen in liquid nitrogen, and stored at −80 °C, and sent to Shanghai Applied Protein Technology Co., Ltd. (Shanghai, China) for testing. Take the samples at −80 °C, freeze-dry them, add 500 μL of BHT protein precipitant respectively, add 10 μL of internal standard with a concentration of 1 μg/mL, vortex for 20 s, 14000 RCF, centrifuge at 4 °C for 10 min, take 500 μL of supernatant, add 1000 μL of pure water, Purify the samples using the Oasis HLB 96-well Plate. Add 2 mL of methanol to activate the well Plate, adding 1 mL each time. Add 2 mL of pure water to balance the well plate, adding 1 mL each time. Load the samples in several portions. Add 2 mL of Washing solution A in 2 portions, adding 1 mL each time. Add 2 mL of Washing solution B in two portions, 1 mL each time. Collect the samples, add 1 mL of pure methanol to elute the samples, blow dry with nitrogen, and store at −80 °C. 2. Chromatography-mass spectrometry detection conditions Chromatographic conditions: The samples were separated using the Agilent 1290 Infinity LC ultra-performance liquid chromatography system. We used a ACQUITY UPLC BEH C18 (1.7 µm, 2.1 mm × 50 mm; Waters) Phase column. The sample was placed in an automatic sampler at 4 °C, with a column temperature of 35°C. The mobile phase A was 0.1% formic acid aqueous solution, and the mobile phase B was 0.1% formic acid acetonitrile solution. The flow rate was 400 μL/min, and the injection volume was 2 μL. The relevant liquid phase gradients are as follows: 0—1 min, phase B is maintained at 30%; From 1 to 9 min, phase B changed linearly from 30% to 90%. From 9 to 11 min, phase B remained at 90%. From 11 to 11.1 min, phase B changed linearly from 90% to 20%. From 11.1 to 14 min, phase B remained at 20%. A QC sample is set up with a certain number of experimental samples at intervals in the sample queue to detect and evaluate the stability and repeatability of the system. The sample queue is set up with a mixture of standard substances of the target substances for the correction of chromatographic retention time. A mixture of standard substances for the correction of chromatographic retention time. Mass spectrometry conditions: Mass spectrometry analysis was conducted using a 5500 QTRAP mass spectrometer (SCIEX) in negative ion mode. The conditions of the 5500 QTRAP ESI source negative ion are as follows: source temperature: 500°C; ion Source Gas1(Gas1): 50; Ion Source Gas2 (Gas2): 50; Curtain gas (CUR): 30; ionSapary Voltage Floating (ISVF): −4500 V. The ion pairs to be tested were detected by the MRM mode. 3. Data analysis The chromatographic peak area and retention time were extracted by using Multiquant 3.0.2 software. The retention time was corrected using the standard substances of the target substances for metabolite identification. Quantification and statistical analysis Experimental data were analyzed using the software GraphPad Prism 8.0 (GraphPad Software) and SPSS 17.0 (IBM). All experimental data are presented as the mean ± Standard error of mean (SEM). The correlation among the gene expression profiles of the samples was assessed by Spearman’s rank correlation test. One-way analysis of variance (ANOVA) was employed for comparisons between groups. Unpaired data were evaluated using Student’s t-test where appropriate. A significance threshold of P < 0.05 was adopted. Reporting summary Further information on research design is available in the [295]Nature Portfolio Reporting Summary linked to this article. Supplementary information [296]Supplementary Information^ (16.3MB, pdf) [297]41467_2025_61088_MOESM2_ESM.pdf^ (54.3KB, pdf) Description of Additional Supplementary Files [298]Supplementary Data 1^ (16.9KB, xlsx) [299]Supplementary Data 2^ (865.8KB, xlsx) [300]Supplementary Data 3^ (133.2KB, xlsx) [301]Supplementary Data 4^ (19.9KB, xlsx) [302]Supplementary Data 5^ (18.2KB, xlsx) [303]Supplementary Data 6^ (15.5KB, xlsx) [304]Supplementary Data 7^ (12.5KB, xlsx) [305]Reporting Summary^ (211.8KB, pdf) [306]Transparent Peer Review file^ (17.9MB, pdf) Source data [307]Source Data^ (3.2MB, xlsx) Acknowledgements