Access the full text.
Sign up today, get DeepDyve free for 14 days.
www.nature.com/npjbcancer ARTICLE OPEN Lipid exposure activates gene expression changes associated with estrogen receptor negative breast cancer 1 2 3 1 2 4 Shivangi Yadav , Ranya Virk , Carolina H. Chung , Mariana Bustamante Eduardo , David VanDerway , Duojiao Chen , 5 4 6 1 1 4 3,7,8,9 Kirsten Burdett , Hongyu Gao , Zexian Zeng , Manish Ranjan , Gannon Cottone , Xiaoling Xuei , Sriram Chandrasekaran , 2 10 1 1 ✉ ✉ Vadim Backman , Robert Chatterton , Seema Ahsan Khan and Susan E. Clare Improved understanding of local breast biology that favors the development of estrogen receptor negative (ER−) breast cancer (BC) would foster better prevention strategies. We have previously shown that overexpression of speciﬁc lipid metabolism genes is associated with the development of ER− BC. We now report results of exposure of MCF-10A and MCF-12A cells, and mammary organoids to representative medium- and long-chain polyunsaturated fatty acids. This exposure caused a dynamic and profound change in gene expression, accompanied by changes in chromatin packing density, chromatin accessibility, and histone posttranslational modiﬁcations (PTMs). We identiﬁed 38 metabolic reactions that showed signiﬁcantly increased activity, including reactions related to one-carbon metabolism. Among these reactions are those that produce S-adenosyl-L-methionine for histone PTMs. Utilizing both an in-vitro model and samples from women at high risk for ER− BC, we show that lipid exposure engenders gene expression, signaling pathway activation, and histone marks associated with the development of ER− BC. npj Breast Cancer (2022) 8:59 ; https://doi.org/10.1038/s41523-022-00422-0 INTRODUCTION validated this signature in an independent set of 36 human samples and re-conﬁrmed the above results in fresh frozen tissues Breast cancer is a heterogeneous disease with different molecular obtained from a new set of ER+ and ER− breast cancer patients, subtypes that are characterized, at a minimum, by the expression each time using laser capture microdissection (LCM) to obtain of the estrogen receptor (ER), progesterone receptor (PR), and epithelial cells from tumor and CUB samples . Again, we found Human epidermal growth factor receptor 2 (HER2)/neu . Although signiﬁcantly higher expression of LiMe genes in CUBs from women multiple statistical tools have been developed to quantify breast 2 with ER− breast cancer, compared to both CUBS from women cancer risk , they do not predict breast cancer subtypes. Current with ER+ breast cancer, and breast epithelium from a control breast cancer prevention with selective estrogen receptor group of women undergoing reduction mammoplasty. However, modulators (SERM) and aromatase inhibitors decreases the risk the speciﬁc genes comprising this overexpressed set had no of estrogen-receptor (ER) positive breast cancer sub-types, but not speciﬁc function or group of functions in common and did not 3–5 those without ER expression . Thus, determining the etiologic/ suggest speciﬁc mechanistic explanations as to why lipid biologic factors that favor the development of ER-negative breast metabolism pathways would aid ER− breast cancer development. cancer will potentially enable the development of both strategies In the present study, we address possible mechanistic explana- to identify women at risk for ER-negative disease as well as tions for our previous observations. targeted preventive and therapeutic agents. Major reprogramming of cellular energetics is one of two Given the poor understanding of the genesis of sporadic ER- 11 emerging hallmarks of cancer . Altered lipid metabolism is negative breast cancer, we set out to study this using the posited to be a driver of carcinogenesis in various cancers, contralateral, unaffected breast of patients with unilateral breast 12 13,14 15 including ovarian , prostate , liver and triple negative breast 16,17 cancer as a model. Studies of metachronous contralateral breast cancer . Increased lipid metabolism has also been shown to cancer show a similarity in the ER status of the contralateral cancer serve as a survival signal that enables tumor recurrence and has 6–8 to the index primary . Therefore, the contralateral unaffected been suggested as an Achilles heel for combating breast cancer breast (CUB) of women undergoing surgical therapy for newly progression . Despite this recognition of the importance of fatty diagnosed unilateral breast cancer can be employed as a model to acid metabolism, its role in the transformation of a normal cell to discover potential markers of subtype-speciﬁc risk. In a previous the malignant state is largely unknown. Metabolomic studies of study, we performed Illumina expression arrays on epithelial cells the concentrations of several free fatty acids in primary breast from the CUB of breast cancer patients and identiﬁed a lipid tumors, including linoleate, palmitate, and oleate, as a function of metabolism (LiMe) gene signature which was enriched in the breast cancer subtype have revealed signiﬁcant differences across CUBs of women with ER- breast cancer . Among these are genes the subtypes, with the highest concentrations in basal-like breast that control critical steps in lipid and energy metabolism. We cancer . Conjugation of long-chain fatty acids to carnitine for 1 2 Department of Surgery, Feinberg School of Medicine, Northwestern University, Chicago, IL 60611, USA. Department of Biomedical Engineering, Northwestern 3 4 University, Evanston, IL 60208-2850, USA. Department of Biomedical Engineering, University of Michigan, Ann Arbor, MI 48109, USA. Center of for Medical Genomics, 5 6 Indiana University School of Medicine, Indianapolis, IN 46202, USA. Department of Preventive Medicine, Northwestern University, Chicago, IL 60611, USA. Department of Data Sciences, Dana Farber Cancer Institute, Harvard T.H. Chan School of Public Health, Boston, MA 02215, USA. Program in Chemical Biology, University of Michigan, Ann Arbor, MI 8 9 48109, USA. Rogel Cancer Center, University of Michigan Medical School, Ann Arbor, MI 48109, USA. Center for Computational Medicine and Bioinformatics, University of Michigan, Ann Arbor, MI 48109, USA. Department of Obstetrics and Gynecology, Northwestern University Feinberg School of Medicine, Chicago, IL 60611, USA. email: email@example.com; firstname.lastname@example.org Published in partnership with the Breast Cancer Research Foundation 1234567890():,; S. Yadav et al. transport into the mitochondria and subsequent fatty acid pathways identiﬁed by GSEA analysis revealed linked clusters oxidation (FAO) was observed to be highest in basal-like breast involved with the nervous system and a second, separate group of cancers, followed by luminal B ~HER2-enriched, with luminal A linked clusters involved with growth factor stimulation, regulation tumors displaying the lowest levels . Another study, which of the MAPK cascade, and ERBB signaling (Fig. 1e). We validated utilized Raman spectroscopy to interrogate tissue, revealed that the expression of a number of genes that GSEA analysis histologically normal breast tissue centimeters removed from the determined were signiﬁcantly upregulated in MCF-10A with breast malignancy have signiﬁcantly higher polyunsaturated fatty octanoate treatment using real-time qPCR (Fig. 1f). In order to acid levels compared with normal tissue from cancer-free validate our ﬁndings in a second cell line, we chose MCF-12A cells. subjects . Sweeney et al. provide the history of the establishment of this cell The kinetic and thermodynamic properties of the chromatin line as well as a demonstration that the cells are non-responsive to modiﬁcation reactions are commensurate with the dynamic range estrogen . 1645 genes were upregulated and 330 downregulated of the physiological concentrations of the corresponding inter- 21 (FDR = 0.01) in the octanoate treated MCF-12A. Comparison of mediates in metabolism . Therefore, we sought to determine if octanoate treated MCF-10A and MCF-12A GSEA reveals consider- the LiMe signature we observed in the CUBs of ER- patients is able overlap for Gene Ontology Biological Processes (GOBP, associated with chromatin modiﬁcations and histone PTMs Supplementary Fig. 2A), Reactome gene sets (R, Supplementary secondary to changes in metabolism fostered by exposure to Fig. 2B), and KEGG gene sets (K, Supplementary Fig. 2C). Similar to medium and long-chain fatty acids. the linked clusters involved with the nervous system seen in the MCF-10As, the octanoate-treated MCF-12A are enriched for gene RESULTS sets of nerve development, synapses and neurotransmitters, and Lipid facilitates transcriptional reprogramming in non- axons. Additional overlap includes: adenylate cyclase pathways transformed mammary cells and cell fate speciﬁcation (upregulated genes, GOBP), cell cycle We established an in vitro model by exposing estrogen and and cell cycle checkpoints (downregulated genes, GOBP), cardiac progesterone receptor (PR) negative MCF-10A cells to octanoate conduction, and muscle contraction (R), and MAPK and HEGDE- (OA), a medium chain eight-carbon fatty acid. Due to its small size HOG signaling (K). Examination of individual genes in the NOTCH and lipophilic nature octanoate does not depend on fatty acid and Wnt pathways listed in Fig. 1d reveals that in MCF-12A cells transport proteins to traverse cell membranes and is readily exposure to octanoic acid increased the expression of DLL4 by 22,23 oxidized in the mitochondria to form acetyl-CoA .We 25.4-fold (p = 1.93E−21, FDR 7.00E−21), that of HEY1 2.07-fold performed RNA-seq to determine the effects of octanoate (p = 7.49E−29, FDR 3.60E−86), NOTCH3 4.75-fold (p = 1.34E−87, treatment on gene expression in the MCF-10A cells. RNA-seq FDR 3.04E−86), WNT11 4.85-fold (p = 4.56E−38, FDR 2.96E−37) analysis revealed that 24 h of octanoate treatment produces a and FZD4 2.29-fold (p = 3.36E−26, FDR 1.24E−25). Thus, treat- transcriptional proﬁle that is completely distinct from vehicle- ment with medium chain fatty acids induces signiﬁcant changes treated controls (Fig. 1a, Supplementary Fig. 1A, B). Genes with in transcription. initially low expression (negative values of lnðE =E Þ) are ctrl ctrl;avg upregulated (corresponding to positive values of lnðE =E Þ) oct ctrl Evaluating the lipid composition of the serum of ER− and while genes with initially high expression (positive values of ER+ BC patients lnðE =E Þ) are downregulated upon octanoate treatment ctrl ctrl;avg Next, we investigated whether dietary lipids, which are mainly (corresponding to negative values of lnðE =E Þ) . More oct ctrl long chain fatty acids (LCFAs), have a similar effect on the gene speciﬁcally, there is a clear trend for initially highly expressed transcriptional proﬁle to that of MCF-10A cells. In order to genes in the control condition to be downregulated upon octanoate treatment while genes with initial low expression in determine the speciﬁc lipid(s) to evaluate experimentally, we the control condition were upregulated. Differential expression sought to determine the differences in the percent composition of analysis performed using DESeq2 revealed a total of 2132 lipid species as a function of ER expression in serum from patients 9,10 upregulated and 632 downregulated genes (FDR = 0.01, |logFC| who had donated CUB samples for our original studies .A >1) in the octanoate treated cells (Supplementary Fig. 1C). comprehensive lipid proﬁle of these serum samples was Pathway enrichment analysis of the differentially expressed genes performed by the Northwest Metabolic Research Center at induced by the 5 mM octanoate treatment was performed and the University of Washington, with measurement of more than 700 top 25 upregulated and downregulated pathways are shown in lipids. For each of the measurements, the association between the Fig. 1b. Speciﬁcally, this analysis revealed that among the top measured value and ER status was evaluated using regression altered biological processes are second messenger mediated models, adjusting for BMI, age, and menopausal status. ER was a signaling, the Notch signaling pathway, adenylate cyclase- categorical variable used to describe subjects having ER+ or ER− activating adrenergic receptor signaling, cell morphogenesis, cancers, or controls undergoing reduction mammoplasty. As the and differentiation. In contrast, downregulated genes are involved purpose of this experiment was to identify a lipid for ensuing in cell cycle processes, transcriptional regulation of tumor experiments, lipid species were ranked for effect size comparing suppressor genes such as p53, and cell cycle checkpoints (Fig. serum from patients subjects with ER- disease to those with ER+ 1b). Additional gene set enrichment analysis (GSEA) investigating disease (Supplementary Table 1). There were 28 serum samples top pathways with coordinated upregulation or downregulation from donors with ER− disease and 28 from ER+ donors. Three of of genes demonstrated that the top pathways associated with the top four lipid species with the largest effect size were noted to octanoate treatment included positive regulation of cell morpho- contain linoleic acid: cholesterol ester (CE) 18:2, phosphatidyl genesis, a process involved in differentiation, as well as several choline (PC)16:0/18:2, and triacylglycerol (TAG) 54:6-FA18:2 (Table oncogenic pathways associated with breast tumorigenesis, 1). Linoleic acid as a free fatty acid ranked 11th in the analysis. including ERBB, WNT, and NOTCH signaling pathways (Fig. 1c). Linoleic acid is the most highly consumed polyunsaturated fatty Subsequent leading-edge analysis of these top upregulated acid in the human diet , its presence in serum CE has been signaling pathways- Lipid storage pathways (I), Wnt pathway (II), strongly correlated with intake , and its concentration in adult Notch signaling (III) and ERBB pathway (IV) shows clear association of core enrichment genes with octanoate treatment across adipose tissue has more than doubled in the past half century . replicates (Fig. 1d). Network analysis of octanoate-associated Therefore, linoleic acid (LA) was included in subsequent studies. npj Breast Cancer (2022) 59 Published in partnership with the Breast Cancer Research Foundation 1234567890():,; S. Yadav et al. a b (I) (II) (IV) (III) Octanoic acid and Linoleic acid inﬂuence chromatin packing transcriptionally active and inactive states. Thus, our next step behavior was to explore the changes in chromatin structure of fatty acid The state of chromatin is intimately linked with the regulation of treated MCF-10A cells by employing partial wave spectroscopic gene transcription, undergoing dynamic changes between (PWS) microscopy, which quantiﬁes chromatin packing scaling (D) Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 59 S. Yadav et al. Fig. 1 Lipid-rich environment enables transcriptional reprogramming in mammary epithelial cells. a Twenty-four hour treatment of MCF- 10A cells with 5 mM octanoate results in a completely distinct transcriptional proﬁle compared to untreated controls. E is the expression of ctrl genes in the control condition across all 3 control replicates, E is the average expression for the control condition across all genes and ctrl;avg replicates, E is the expression of genes across all 3 octanoate replicates. E =E represents the ratio of expression of a particular gene to ctrl ctrl;avg oct ctrl the average expression across all control cells. Thus, a positive value of ln corresponds to genes that are highly expressed in the control ctrl;avg ctrl conditions while a negative value of ln corresponds to genes that have an initial lower expression in the control condition. E /E oct ctrl ctrl;avg represents the ratio of expression of a particular gene for octanoate-treated versus vehicle control-treated cells. Genes with initially low expression are upregulated while genes with initially high expression are downregulated upon octanoate treatment. b Gene ontology analysis of differentially expressed genes induced by octanoate treatment. Upregulated and downregulated genes were ﬁrst identiﬁed using DESeq2 (FDR < 0.01, |logFC| > 1) for 5 mM octanoate treated cells compared to vehicle-treated control cells. Pathway enrichment analysis was performed on identiﬁed differentially expressed genes with annotations from online pathway databases (KEGG, Hallmark, Canonical Pathways, Reactome, BioCarta) and Gene Ontology Biological Processes. Pathway enrichment was ranked by p-value on a −Log scale and a selection from the top 25 pathways associated with upregulated genes (in red) and downregulated genes (in blue) are shown. c GSEA analysis of Gene Ontology Biological Processes showing top pathways associated with octanoate treatment with FDR < 0.1 related to differentiation, cell signaling, and metabolic processes. d List of core enrichment genes differentially expressed in treated replicates-T4, T5, T6 versus control replicates- C1, C2, C3: (I) Lipid storage pathways (II) Wnt pathway (III) Notch pathway (IV) ERBB pathway, each pathway as identiﬁed by GSEA leading edge analysis. Expression values are represented as colors and range from red (high expression) to dark blue (lowest expression). e Network analysis of pathways associated with the octanoate phenotype in GSEA analysis of Gene Ontology Biological Processes. f qPCR analysis of genes associated with the NOTCH pathway (mean ± s.d.). Two genes, NOTCH3 and DLL4 show remarkable upregulation upon 5 mM octanoate treatment compared to other identiﬁed genes such as NOTCH1. Statistical signiﬁcance was determined by the unpaired t-test with Welch’s correction (**P < 0.01, *P < 0.05). period. Our results showed signiﬁcant increases in chromatin Table 1. Lipid species in the serum of CUB patients ranked by effect packing scaling upon exposure to lipids suggesting that there is size, ER negative compared to ER positive. an increase in the dynamic range of gene expression and transcriptional gene network heterogeneity (Fig. 2a, b). These Order Variable ER_Negaitve_Effect_compared_to_ER_Positive signiﬁcant changes in chromatin packing behavior also indicate 1 CE(18:2) 97.76 signiﬁcant changes in chromatin accessibility, which is directly associated with chromatin structure . 2 PC(16:0/18:2) 51.96 3 SM(20:0) 42.09 ATAC sequencing reveals increased chromatin accessibility in 4 TAG54:6- 28.78 regulatory regions of genes in the MAPK and cAMP signaling FA18:2 pathways in lipid treated mammary cells 5 SM(22:0) 23.33 To acquire more detailed insight into the speciﬁc regions of open 6 PC(16:0/18:1) 23.13 chromatin that were made accessible by LA treatment, we 7 CE(16:1) 20.32 proceeded with ATAC sequencing on LA-treated MCF-10A cells. 8 TAG54:5- 16.57 We examined the genomic locations of ATAC-seq peaks, FA18:2 representing open chromatin sites, and discovered 1704 open 9 SM(22:1) 16.01 chromatin sites. Open chromatin regions were overrepresented 10 CE(18:3) 15.51 within 1 kb of transcription start sites (TSSs) by 40-fold relative to 11 FFA(18:2) 15.37 the whole genome (Fig. 2c). Further, KEGG pathway analysis revealed 326 open chromatin regions with a log fold change > = 12 PC(18:0/18:2) 13.77 1.5 and FDR < 0.05 compared to vehicle treated cells. Among the 14 PC(18:1/18:2) 12.38 top pathways that were upregulated signiﬁcantly upon LA 14 TAG52:4- 11.37 treatment are MAPK signaling pathway, PI3K-AKT signaling FA18:2 pathway, and the cAMP adenylate cyclase pathway (Fig. 2d). 15 TAG54:5- 10.83 Additionally, motif analysis conducted using ‘HOMER’ showed FA18:1 that chromatin regions made accessible/inaccessible by LA Abbreviations: CE cholesterol ester, FFA free fatty acids, PC phosphatidyl- treatment have binding motifs for a number of transcription choline, SM sphingomyelin, TAG triacylglycerol. TAGs have 3 acyl chains, factors (Fig. 2e). These data reveal that linoleic acid affects but it is only possible to measure the length and number of double bonds chromatin heterogeneity and increases/decreases the accessibility of 1 of them. For example, TAG54:6-FA18:2 has 1 chain that is an 18:2 FA of speciﬁc regions that include transcription factor binding sites. and the other two have a total of 54 carbons and 6 double bonds. The whole TAG has 72 carbons in the chain (plus 3 from the glycerol backbone) and a total of 8 double bonds. Notch pathway genes are overexpressed in patients at high risk of ER- disease Next, we sought to determine whether the genes, or sets of genes/pathways that we identiﬁed in our in vitro study were also in live cells . D represents the power-law scaling relationship differentially expressed in vivo in tissue of patients at risk for ER− between the 1D size of the chromatin polymer i.e., the number of and ER+ breast cancer. We took advantage of RNA from the CUB nucleotides and the 3D space the chromatin polymer occupies. of breast cancer cases utilized in our previous studies, which Recent evidence indicates that higher chromatin packing scaling revealed the association of LiMe genes in the CUBs of women with is associated with increased intercellular and intra-network 9,10 unilateral ER- breast cancer . We combined the data from the transcriptional heterogeneity as well as increased malignancy 24,30,31 and chemoresistance in cancer cells . PWS was used to RNA and ATAC sequencing experiments and collated a list of 44 evaluate the effect of OA and LA on chromatin packing scaling in genes of interest and 3 housekeeping genes. The list consists of live MCF-10A cells. Images were obtained every 6 h over a 24 h the genes from the HEDGEHOG, NOTCH, WNT, EMT, PPARγ, and npj Breast Cancer (2022) 59 Published in partnership with the Breast Cancer Research Foundation S. Yadav et al. Vehicle Linoleic acid Octanoate Vehicle adenylate cyclase pathways (Supplementary File 1). TaqMan low with participants comprised of 28 matched triplets of women with density arrays were utilized to measure the expression of these ER-positive breast cancer, ER-negative breast cancer, and reduc- genes in CUBs of ER− and ER+ cases compared with the tion mammoplasty controls. The three groups were matched by reduction controls. The study population included 84 women, age, race, and menopausal status as shown in Supplementary Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 59 S. Yadav et al. Fig. 2 Linoleic acid alters large-scale chromatin packing behavior in MCF-10A cells. a Representative PWS microscopy images of MCF-10A cell nuclei at 24 h after treatment with vehicle controls and lipids—octanoate and linoleic acid. Scale bars, 10 μm. Chromatin packing scaling (D) map of nuclei shows an increase in chromatin packing scaling upon lipid treatment as demonstrated by an increase in red regions. b Changes in average chromatin packing scaling among MCF-10A cells upon treatment with vehicle controls and lipids compared to untreated cells. Signiﬁcance was determined using unpaired Kolmogorov–Smirnov t-test (****P < 0.0001, *P < 0.05). Bar graphs show the mean change in intranuclear D across cell populations for N = 88 cells PBS (vehicle for octanoate), N = 110 cells Octanoate (OA), N = 103 cells BSA (vehicle for linoleic acid), and N = 94 Linoleic acid (LA). c Enrichment of genomic locations for 1704 open chromatin regions (FDR < 0.05, logFC > 1) in LA treated MCF-10A cells. The enrichment of peaks in each type of genomic region relative to the whole genome is shown on the y- axis. Two ATAC-seq libraries were used for the analysis. d Pathway analysis for the regions with increased chromatin accessibility in linoleic acid-treated cells identiﬁed using the KEGG database. e Biplot showing changes in chromatin accessibility for speciﬁc regions identiﬁed by HOMER analysis. Motifs with a signiﬁcant increase in the chromatin accessibility are shown in blue and those with a signiﬁcant decrease in accessibility are shown in yellow (FDR < 0.05 and |logFC| > 1). Fig. 3A. As noted in our original publication, ANOVA revealed a ACSL3, and CPT1B were increased by LA exposure in MCF-10A cells signiﬁcant difference in BMI across the three groups with BMI in and mammary organoids. Additionally, we observed a signiﬁcant the reduction mammoplasty control group (30.0 ± 5.8) notably increase in DLL4 expression followed by HEY1, HEY2, and NOTCH1 higher than in ER-negative cases (25.3 ± 6.3, p = 0.015), but not in the lipid-treated mammary cells (Fig. 4a). We revisited the ATAC signiﬁcantly higher than in the ER-positive group (26.7 ± 5.5, p = sequencing data to examine the effect of LA on chromatin 0.136) . There was no signiﬁcant difference in HER2 status architecture near key genes in the DLL4/NOTCH signaling pathway between ER-positive and ER-negative cases. The majority of the and observed increased accessibility around the transcription start selected genes had higher expression in high-risk CUB specimens sites of DLL4, NOTCH1, and HEY1 showing signiﬁcant lowered chromatin density with p-values of 1.62e−17, 0.017 and 0.03 than the controls, irrespective of the ER status of the index tumor respectively (Fig. 4b, c). (Supplementary Fig. 3B). The comparison between the ER− and ER+ CUBs revealed that in the ER− CUBS there is increased expression of genes that function in the Notch pathway: NOTCH1 The NOTCH signaling pathway is activated in vitro by octanoic (1.7-fold, p = 0.002, BH_adjP = 0.07), NOTCH4 (1.5-fold, p = 0.04, acid treatment BH_adjP = 0.3), DLL1 (1.2-fold, p = 0.07, BH_adjP = 0.4) and HEY 1 Intracellular Notch binds to the transcriptional repressor RBP-Jk in (1.5-fold, p = 0.05, BH_adjP = 0.3), in addition to the SMO gene the nucleus, thereby converting it into an activator and inducing (1.5-fold, p = 0.05, BH_adjP = 0.3), which is a key component of the expression of downstream target genes. Therefore, to the hedgehog signaling pathway (Fig. 3). Comparing ER− to determine if the NOTCH pathway is activated by OA, we control, increased expression was observed for GPR161 (1.7-fold, transfected a RBP-Jk reporter construct into MCF-10A cells. The p = 0.05, BH_adjP = 0.7), which plays a role in the Hedgehog LUC/REN ratio is increased 2-fold by exposure to OA (Fig. 5) pathway via cAMP signaling, and IGF2 (2.8-fold, p = 0.07, BH_ indicating that the NOTCH pathway is functionally activated by adjP = 0.7), which signals via both the MAPK and PI3K-AKT the lipid. pathways. Altogether, these data reveal upregulation in NOTCH signaling in benign breast tissue samples from women at risk for Fatty acids drive ﬂux through metabolic reactions resulting in ER− disease, suggesting that dysregulation of these pathways increased histone methylation may play a role in the early stages of ER- cancer development. While most of the experiments reported by McDonnell et al. were performed in AML 12 liver cells, these investigators also LA increases the expression of Notch pathway genes and demonstrated increased H3K9 acetylation in octanoate-exposed speciﬁc genes involved in fatty acid oxidation in vitro MCF7 and MDAMB-231 breast cancer cells . Therefore, we sought The increased expression of Notch pathway genes we discovered to determine if these same experimental conditions would lead to in the ER- CUBs, along with the similar ﬁndings in MCF-10A cells H3K9 acetylation in a non-malignant MCF-10A cells. We exposed exposed to octanoate (described above), led us to test the MCF-10A non-transformed ER - breast epithelial cell line to 5 mM hypothesis that long chain fatty acids have similar effects on gene octanoate (OA) for 24 h in medium containing both glucose expression. We, therefore, investigated whether an increased LA (1.441 g/L) and glutamine (0.292 g/L). Western blot analysis environment inﬂuences the expression of Notch pathway genes demonstrated that octanoate exposure of MCF-10As resulted in and speciﬁc genes involved with FAO in vitro. We treated MCF- increased acetylation at both H3K9 and H3K14 (Fig. 6a). To 10A cells and mammary organoids from reduction mammoplasty demonstrate that this was a fatty acid-speciﬁc effect, we treated patient samples with LA for 24 h and then quantiﬁed changes in the cells with 1,4-Cyclohexanedimethanol (1,4-CHDM), an alcohol gene expression using RT-qPCR. To begin with, we assayed the with the same formula as octanoate; no acetylation was observed genes involved in the activation of FAO. Upon entering cells, free consequent to the alcohol exposure (Supplementary Fig. 4A). To fatty acids are converted into fatty acyl-CoA molecules by the validate the speciﬁcity of the antibody against the acetylated enzymes of the acyl-CoA synthetase (ACS) family . Notably, acyl- histone lysines, we treated MCF-10A cells with sodium butyrate, a CoA synthetase long chain (ACSL3) is one of the LiME genes found histone deacetylase (HDAC) inhibitor. Sodium butyrate treatment to be upregulated in high-risk ER- CUBs samples. Generation of increased the acetylation of H3K9 and H3K14 as shown in acetyl-CoA occurs through a cyclical series of reactions in which a Supplementary Fig. 4B. fatty acid is shortened by two carbons per cycle, eventually To exhaustively explore the impact of octanoate treatment on generating acetyl co-A. Acetyl co-A is a substrate for ketogenesis, metabolic pathways, we used ﬂux balance analysis (FBA) . FBA which is initiated by the mitochondrial enzyme 3-hydroxy-3- makes use of genome-scale metabolic network models that methylglutaryl-CoA synthase 2 (HMGCS2), another of the pre- contain all known metabolic reactions in a cell or tissue based on viously identiﬁed LiMe genes. The mechanism for LCFAs oxidation evidence from the published literature . Genome-scale metabolic is slightly more complex than for MCFAs, as this is regulated models have been widely used to predict the metabolic behavior 38–42 primarily via the enzyme carnitine palmitoyltransferase 1 (CPT1), of various mammalian cell types . Here we used the Recon1 the rate-limiting enzyme of FAO which enables transport into the human network model that maps the relationship between 3744 mitochondria. As shown in Fig. 4a, the expression of HMGCS2, reactions, 2766 metabolites, 1496 metabolic genes, and 2004 npj Breast Cancer (2022) 59 Published in partnership with the Breast Cancer Research Foundation S. Yadav et al. NOTCH 4 HEY 1 Fig. 3 Notch pathway is overexpressed in CUB samples of patients at high risk of ER− disease. Expression of genes from various pathways in matching CUBs from ER-negative, ER-positive patients, and controls. The log2-transformed relative (log2RE) amounts of mRNA expression −(CtX−CtGAPDH) normalized to the housekeeping gene and expressed as log 2 = −(CtX − Ct GAPDH) where Ct is threshold cycle and X is gene of interest. IGF2 and GPR161 were signiﬁcantly higher in ER-negative versus control. Genes from the Notch pathway were signiﬁcantly higher in ER negative CUBs in comparison to ER positive patients. Mann–Whitney test was used to test the pairwise differences between the samples (ER+,ER−, Control) * P < 0.05; ** P < 0.01. Boxplots show mean and SEM with whiskers indicating 1–99th percentile. Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 59 S. Yadav et al. metabolic enzymes . This model was augmented with biochem- the ﬂux through the substrates for the histone modiﬁcations. ical reactions corresponding to histone acetylation and methyla- These models were previously used to predict bulk histone 38,44 tion , allowing us to predict the consequences of octanoate- acetylation levels in various cell lines based on the nuclear ﬂux of induced metabolic changes on histone modiﬁcations by tracking acetyl-coA directed towards histone acetylation . Similarly, bulk npj Breast Cancer (2022) 59 Published in partnership with the Breast Cancer Research Foundation S. Yadav et al. Fig. 4 Increased DLL4/Notch signaling is associated with the stimulated fatty acid oxidation. a qPCR data showing increase in lipid metabolism genes (green) and Notch pathway genes (red) after 24 h linoleate treatment in MCF-10A and mammary organoids (mean ± s.d.). Organoid I was donated by a postmenopausal 61-year-old with a BMI of 22 and Organoid II by a premenopausal 28-year-old with a BMI of 31. Statistical signiﬁcance was determined by the unpaired t-test with Welch’s correction (****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05). b Chromatin accessibility in the lipid-treated cells around the transcription start site (TSS) of NOTCH1, HEY1, and DLL4 (FDR < 0.001). c Gene tracks and increase in peaks for the Notch genes in LA treated cells with the exact location on the chromosome. d Leading edge scores for genes of interest associated with the NOTCH signaling pathway as determined by GSEA leading edge analysis. DLL4, HEY1, HEY2, NOTCH3, and NOTCH4 were identiﬁed as core enrichment genes in the NOTCH pathway. octanoic and linoleic acid treatment, we performed liquid Cignal RBP-Jk reporter assay chromatography/mass spectrometry on tryptic peptides isolated from the nuclei of treated and control MCF-10A cells. Increased methylation in the OA treated cells was observed in various Positive control histone proteins including H3K9me1/2/3, H3.1K27me2/3, Negative control H3.3K36me2/3, H3K79me1/2, and H3K4 (Fig. 6e) together with Vehicle increased acetylation of H3K14 and H4K16 (Fig. 6d). Similarly OA increased methylation was observed for H3.1: K27me1, H3.3: K23me1, H3.1: K36me3, H3.3: K36me2, and H3.3: K36me3 in LA 1 treated cells (Fig. 6e). Notably, the GSEA analysis showed a signiﬁcant correlation of H3K27 methylation (NES = 2.47, FDR q- 0 value = 0.05) and H3K4 methylation (NES = 1.24, FDR q-value = 0.1) with octanoate treatment (Supplementary Fig. 4D) suggesting this lipid-rich environment eventuates in histone methylation in Vehicle vs. OA mammary epithelial cells. p-value = 0.07 Fig. 5 Effects of OA on Notch signaling. NOTCH transcriptional DISCUSSION activity was measured using the Cignal RBP-Jk reporter assay The known determinants of risk for ER-negative breast cancer are following exposure of MCF-10A cells to 5 mM octanoic acid (OA) for genetic (either speciﬁc racial inheritance, germline mutations in 24 h. Luciferase levels were normalized to Renilla luciferase. The genes such as BRCA1) or systemic/behavioral factors (premeno- results were plotted as fold change with respect to the untreated. 47 48 (n = 3, mean ± SEM). The p-value was calculated by unpaired t-test. pausal obesity , absence of a breastfeeding ). In contrast, few if any local factors in the breast environment serve to identify women at risk for ER-negative tumors. Local in-breast factors are histone methylation levels can be predicted based on the nuclear of great interest, however, since they may be more speciﬁcally ﬂux of S-adenosyl-L-methionine (SAM) . The model predicted targetable for breast cancer prevention than systemic factors. Of octanoate treatment would result in increased histone methyla- note, the two strongest risk factors for breast cancer overall (other tion levels, with a more modest increase in histone acetylation than high penetrance germline mutations) are local: atypical levels (Fig. 6c). As a comparison, we repeated this analysis with 49 50 proliferative lesions, and extremely dense breast tissue . This immortalized hepatocyte cells used by McDonnel et al.; they reasoning motivated us to investigate the local breast biology that found a signiﬁcant increase in histone acetylation after octanoate may promote the development of ER-negative rather than ER- treatment . We calculated metabolic ﬂux in these hepatocytes positive breast cancer, using the CUB of women undergoing using the transcriptomics data from McDonnel et al and found a surgery for a unilateral primary breast cancer as a model for ER- much larger increase in histone acetylation after octanoate 7,51 speciﬁc breast cancer risk . In our initial study, we identiﬁed a treatment (Supplementary Fig. 4E). These results suggest that highly correlated lipid metabolism (LiMe) gene signature, which the impact of metabolic alterations on histone acetylation is cell- was enriched in the CUBs of women with ER- breast cancer. 45,46 type speciﬁc, as observed in prior studies . Overall, out of the To explain the biologic basis for this association, we developed 3759 reactions in the model, we identiﬁed 38 that showed an in vitro model wherein we exposed MCF-10A and MCF-12A, ER- signiﬁcant increased activity after octanoate treatment (p-value < negative, non-tumorigenic epithelial cell lines, or breast organoids 0.01; Supplementary Fig. 4C). As expected, reactions involved in derived from reduction mammoplasty samples to an extracellular lipid and fatty acid metabolism, speciﬁcally triacyl glycerol milieu rich in medium or long chain fatty acids. This model system synthesis and glycerophospholipid metabolism were upregulated. has now enabled us to demonstrate that the exposure of breast Interestingly, among the upregulated reactions were several epithelial cells to these fatty acids results in a dynamic and reactions related to the one-carbon metabolic pathway, which profound change in gene expression, accompanied by changes in links folate, SAM, methionine, glycine, and serine metabolism (Fig. chromatin packing density, chromatin accessibility, and histone 6f). The reactions catalyzed by methionine adenosyltransferase, PTMs. The histone modiﬁcations, in turn, are the result of both the methionine synthase, adenosyl homocysteinase, 5,10-methylene- lipid-engendered increased expression of the requisite enzymes tetrahydrofolatereductase, glycine N-methyltransferase, and for- and the increased production of their substrates. Our metabolic myltetrahydrofolate dehydrogenase were all predicted to have ﬂux analysis revealed the upregulation of several reactions related increased activity after treatment (p-value < 0.01). These reactions to the one-carbon metabolic pathway, which links folate, SAM, likely support increased histone methylation by providing one methionine, glycine, and serine metabolism. This insight was not carbon units. Examining the reaction ﬂuxes/activities in the OA- evident upon analysis of differential gene expression, which is not treated MCF-12A cells (Fig. 6f), we see differences in one carbon surprising as gene expression changes often do not reﬂect the ﬂux metabolism and glutathione metabolism similar to what we of metabolic reactions . The substrates for histone methylation observed in the MCF-10A cells. and acetylation reactions often have cellular concentrations that Lipid exposure eventuates in histone methylation. In order to are commensurate with enzyme Km values, and thus these proﬁle the speciﬁc histone marks signiﬁcantly changed by the reactions are sensitive and responsive to changes in metabolism . Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 59 Positive control Negative control Vehicle OA Relative Luciferase activity (fold) S. Yadav et al. Organoids MCF-10A Vehicle Acetylation LA ns ns Vehicle Methylation LA **** **** **** ns **** **** **** **** **** ns **** * ** Fig. 6 Fatty acids drive histone modiﬁcations and metabolic ﬂux. Western blot of histone acetylation at H3 lysine K9 and K14 in MCF-10A cells and organoids treated with a octanoate and b linoleic acid. c The effect of octanoate treatment on histone acetylation and methylation ﬂux in MCF-10A cells predicted using genome-scale metabolic modeling. Proteomic acetylation (d) and methylation (e) proﬁling measured by mass spectrometry of MCF-10A cells treated in triplicate with 5 mM octanoate for 24 h in a complete media compared to vehicle (left) and 0.5 mM linoleate for 24 h in complete media compared to vehicle (right) (mean ± s.d.). Two-way ANOVA was performed to determine the statistical signiﬁcance and corrected for multiple comparisons using Sidak test (****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05, ns not signiﬁcant). f Heatmap of reaction ﬂux differences predicted by metabolic modeling to be differentially active (p-value < 0.01) between control and treatment (increased ﬂux in red, decreased ﬂux in blue). The corresponding pathways (subsystem) that each reaction belongs to is listed in the legend. npj Breast Cancer (2022) 59 Published in partnership with the Breast Cancer Research Foundation H3: K14AC H4: K16AC H3: K9ME1 H3: K9ME2 H3: K9ME3 H3.1: K27ME1 H3.1: K27ME3 H3.3: K27ME1 H3.1: K36ME3 H3.3: K36ME2 H3.3: K36ME3 H3: K79ME1 H3: K79ME2 H4: K20ME2 H3K4ME1 Relative abundance % Relative abundance % Subsystem Pathway S. Yadav et al. Heatmap of metabolic reaction flux difference MCF10A Ubiquinol-6 cytochrome c reductase, Complex III Electron transfer flavoprotein Electron transfer flavoprotein-ubiquinone oxidoreductase Acyl-CoA oxidase (hexadecanoyl-CoA), peroxisomal Diacylglycerol acyltransferase Lipase Lipase.1 Monoacylglycerol acyltransferase Methenyltetrahydrofolate cyclohydrolase, mitochondrial Glycerol kinase Lipase.2 Cytochrome c oxidase, mitochondrial Complex IV Histone methylation Retinyl ester hydrolase Retinol acyltransferase Sarcosine dehydrogenase (m) Peroxidase (multiple substrates) Glycine N-methyltransferase Glycerol-3-phosphate dehydrogenase (FAD), mitochondrial Adenosylhomocysteinase Catalase A, peroxisomal Catalase Methionine adenosyltransferase Methionine synthase 5,10-methylenetetrahydrofolatereductase (NADPH) S-Adenosyl-L-methionine intracellular diffusion D-lactate dehydrogenase Hydroxyacylglutathione hydrolase D-Lactaldehyde:NAD+ oxidoreductase (glutathione-formylating) Lactaldehyde dehydrogenase 1-acylglycerol-3-phosphate O-acyltransferase 1 Subsystem Glycerol-3-phosphate acyltransferase Fatty-acid--CoA ligase (octanoate) Lecithin retinol acyltransferase (11-cis) Vitamin A Metabolism Retinyl ester hydrolase (11-cis) Urea cycle/amino group metabolism Phospholipase A2 Triacylglycerol Synthesis Retinyl ester hydrolase (9-cis) Transport, Endoplasmic Reticular Inositol oxygenase Glyceraldehyde-3-phosphate dehydrogenase ROS Detoxification Enolase Pyruvate Metabolism NADH dehydrogenase, mitochondrial Nucleoside-diphosphatase (dUDP) Pentose Phosphate Pathway Ribonucleoside-diphosphate reductase (GDP) Oxidative Phosphorylation 5-nucleotidase (dGMP) Nucleotides Purine-nucleoside phosphorylase (Deoxyguanosine) Ribonucleoside-diphosphate reductase (UDP) Miscellaneous Ribonucleoside-diphosphate reductase (CDP) Methylation DUTP diphosphatase, nuclear Methionine Metabolism Nucleoside-diphosphate kinase (ATP:dUDP), nuclear DCMP deaminase Inositol Phosphate Metabolism Deoxyuridine phosphorylase Glycolysis/Gluconeogenesis Thioredoxin reductase (NADPH) Glycine, Serine, and Threonine Metabolism Aldehyde dehydrogenase (acetaldehyde, NAD) Deoxyribokinase Glycerophospholipid Metabolism Deoxyribose-phosphate aldolase Folate Metabolism Aldehyde dehydrogenase (acetylaldehyde, NAD), mitochondrial Fatty Acid Metabolism 2-deoxy-D-ribose 1-phosphate phosphorylase Aldehyde dehydrogenase (acetaldehyde, NADP) MCF-12A H2o Exchange Retinol Dehydrogenase (All-Trans) Carboxylic Acid Dissociation Retinal Dehydrogenase (Nadph) Co2 Exchange Acyl-Coa Oxidase (Hexadecanoyl-Coa), Peroxisomal R Group Artificial Flux R Total Flux R Group To Palmitate Conversion Phosphoglucomutase D-Glucose Exchange Glutathione Oxidoreductase Glutathione Peroxidase, Mitochondria L-Lactate Dehydrogenase Glutamate Dehydrogenase (Nadp), Mitochondrial 0 Adenylate Kinase Methylenetetrahydrofolate Dehydrogenase (Nad) Methenyltetrahydrofolate Cyclohydrolase Formyltetrahydrofolate Dehydrogenase Glycine Hydroxymethyltransferase, Reversible −1 Glycine Exchange Fatty-Acid--Coa Ligase (Octanoate) Beta Oxidation Of Med/Long Chain Fatty Acid Aspartate N-Acetyltransferase, Mitochondrial N-Acetyl-L-Aspartate Amidohydrolase Isoleucine Transaminase Acetyl-Coa C-Acetyltransferase, Mitochondrial Catalase A, Peroxisomal Acetyl-Coa Synthetase Aspartate Transaminase Glutaminase (Mitochondrial) Isocitrate Dehydrogenase (Nadp+) Aconitate Hydratase Pathway Retinol Acyltransferase Retinyl Ester Hydrolase Vitamin A Metabolism R Group Coenzyme A Ligase Malate Dehydrogenase Valine, Leucine, and Isoleucine Metabolism Atp-Citrate Lyase Urea cycle/amino group metabolism Hco3 Equilibration Reaction Tryptophan metabolism Electron Transfer Flavoprotein Electron Transfer Flavoprotein-Ubiquinone Oxidoreductase R Group Synthesis Ubiquinol-6 Cytochrome C Reductase, Complex Iii Pyruvate Metabolism Biomass_objective Isoleucine Transaminase, Mitochondrial Oxidative Phosphorylation Inorganic Diphosphatase Nucleotides Atp Synthase (Four Protons For One Atp) Miscellaneous 2-Aminoacrylate Hydrolysis L-Serine Hydro-Lyase Glycolysis/Gluconeogenesis Octanoate (N-C8:0) Exchange Glycine, Serine, and Threonine Metabolism Pyruvate Dehydrogenase Ammonia Exchange Glutathione Metabolism L-Lactate Dehydrogenase.1 Glutamate metabolism O2 Exchange Galactose metabolism Hexokinase (D-Glucose:atp) Glutamate Dehydrogenase (Nad) (Mitochondrial) Folate Metabolism Utp-Glucose-1-Phosphate Uridylyltransferase Fatty Acid Metabolism Nucleoside-Diphosphate Kinase (Atp:udp) Bicarbonate Exchange Exchange Retinal Dehydrogenase Citric Acid Cycle H+ Exchange Alanine and Aspartate Metabolism Carboxylic Acid Dissociation.1 Retinol Dehydrogenase (All-Trans,Nadph) Fig. 6 Continue. Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 59 S. Yadav et al. Fig. 7 Proposed model illustrating the orchestration of lipid-induced molecular changes. Sensors: Senses the fatty acid-rich environment and perturb cellular metabolism providing the essential substrate for histone modiﬁcations and thereby turning on the Mediators- histone PTMs, which consequently activates the Effectors- Notch, adenylate cyclase, and MAPK-ERK the key protein signaling associated with ER− breast cancer. Increasing substrate concentration can increase the product of the Additional clues regarding the association of our experimental reaction even if there is no increase in the expression of the ﬁndings with ER-negative breast cancer comes from GWAS data. A study that included 21,468 ER-negative cases and 100,594 controls enzyme. Our proteomics data reveal increased methylation at identiﬁed independent associations of ten single nucleotide H3K27me1 and H3K36me2/3 in cells treated with octanoate in polymorphisms (SNPs) with the development of ER− breast both MCF-10A and MCF12 A, and H3K4me1 in MCF-10A cells; cancer . Pathway analysis was performed by mapping each SNP GSEA analysis showed that genes with ontologies related to to the nearest gene. This identiﬁed several pathways implicated in histone methylation at H3K27 and H3K4 exhibit changes in susceptibility to ER-negative, but not ER+ breast cancer. Included expression in the lipid-treated cells. among these was the adenylate cyclase (AC) activating pathway. The goal of our investigation was to develop speciﬁc mechan- One of the signiﬁcantly altered biologic processes that we istic explanations as to why lipid metabolism pathways would aid identiﬁed by RNA sequencing of the octanoic acid-treated cells ER− breast cancer development. The data have revealed a is adenylate cyclase-activating adrenergic receptor signaling. number of possibilities, all of which will have to be explored Adenylate cyclase signals via cyclic AMP. Regions of chromatin further. Mammary stem cell differentiation is a hierarchical with increased accessibility are associated with increased gene organization, and lineage tracing experiments have determined 52 expression; our ATAC-Seq results show that linoleic acid exposure that NOTCH1 expression exclusively generates ER− luminal cells . signiﬁcantly increased accessibility to genes in the cAMP signaling A subsequent study by these investigators revealed that during pathway. In their discussion of ER− GWAS results, Milne et al. mammary embryogenesis Notch signaling prevents the genera- suggest that stimulation of the beta 2 adrenergic-adenylate tion of basal precursors, and cells expressing active NOTCH1 cyclase-cAMP-β-arrestin–Src–ERK pathway may play a role in the exclusively give rise to the ER− (Sca1-/CD133-) lineage at any genesis of ER− breast cancer. MetaCore analysis of our RNA- developmental stage from mouse embryonic day 13.5 to sequencing data reveals similar pathway activation, however, it is postpartum day 3 . Even more interesting given our focus on the beta1 adrenergic receptor that demonstrates increased the origins of ER-negative breast cancer was their observation that expression in the octanoate treated cells. In addition, our ATAC- pubertal cells retain plasticity. Ectopic activation of Notch1 in basal seq data showed increased RAP1 signaling pathway accessibility. cells at puberty was able to completely switch their identity to ER- Adenylate cyclase signaling also functions via Epac-Rap1-B-raf- negative luminal cells. MEK-ERK, with this signaling shown to be responsible for npj Breast Cancer (2022) 59 Published in partnership with the Breast Cancer Research Foundation S. Yadav et al. sustained ERK activation that occurs 10–30 min after cAMP CUB samples activation . The MAPK (ERK) pathway can be stimulated by Patients diagnosed with unilateral breast cancer and undergoing means other than adrenergic receptor-ligand binding. Activation contralateral prophylactic mastectomy at Prentice Women’s Hospital of Northwestern Medicine were recruited under an approved protocol of this pathway by overexpression of EGFR + EGF, c-erbB-2, RAF1, (NU11B04), with exclusions for neoadjuvant treatment, prior endocrine or MEK in MCF7 cells leads to estrogen-independent growth and therapy, or pregnancy/lactation during the prior 2 years. A group of downregulation of ERα expression . These results suggest that reduction mammoplasty (RM) patients were also recruited as standard risk hyperactivation of the MAPK(ERK) pathway plays a role in the controls. All participants provided written informed consent. The fresh generation of the ER− phenotype in breast cancer. We observed tissues were frozen and stored in liquid nitrogen. Tissue samples from 56 MAPK activation in our analysis of differentially expressed genes, bilateral mastectomy cases (28 ER+ and 28 ER−) and 28 healthy RM controls were used in this study. The ER+ cases, ER− cases, and controls i.e., “positive regulation of the MAPK cascade,” and in the analysis were matched by age, race, and menopausal status. of regions of chromatin with signiﬁcantly increased chromatin. Using stratiﬁed LD score regression, a statistical method for identifying functional enrichment from GWAS summary statistics, Mammary organoids preparation SNPs associated with the H3K4me3 histone mark were deter- Tissues were collected from women admitted for reduction mammoplasty, mined to be contributing to the heritability of ER-negative breast who were recruited under an approved IRB protocol (NU15B07). All participants provided written informed consent. Breast tissue to be cancer, (2.4-fold, P = 0.0005) . Increased activity of the one- processed is transferred into a sterile petri dish and chopped into small carbon pathway is associated with increased H3K4 trimethylation 38,57 pieces using a scalpel. The minced tissue was transferred to a sterile 50 ml in stem cells and cancer cell lines . Restriction of methionine tube and 30 ml of Kaighn’s Modiﬁcation media (Gibco #21127022) with consequent modulation of SAM and S-Adenosyl-L- containing collagenase from Clostridium histolyticum (Sigma Aldrich, homocysteine (SAH) levels affects methylation at H3K4me3, #C0130) was added, ﬁnal collagenase concentration is 1 mg/mL. Media H3K27me3, and H3K9me3, with H3K4me3 exhibiting the largest containing collagenase is ﬁltered using a 0.22 μm ﬁlter. The Falcon tube is changes (45). Interestingly, this restriction leads to loss of sealed with paraﬁlm and tissue is gently dissociated on a shaker at 100 rpm and 37 °C, overnight (16 h). The following day, organoids are collected by H3K4me3 at the promoters of colorectal cancer (CRC)-associated the centrifugation of the suspension at 114 × g for 5 min. The supernatant genes, with resulting decreased expression (p = 0.02, Fisher’s is discarded, and the organoid pellet washed two-three times with PBS. exact test). A computational model developed to identify the Organoids with a size between 40 and 100 μm are collected and direct inﬂuences on methionine concentrations in humans resuspended in fresh media (3 mL) and added to a six-well plate (Ultra- suggests that dietary intake explains about 30% of the variation Low Attachment Surface plate, Corning # CLS3471). Organoids are allowed in methionine concentration, and fats (arachidic acid in this to stabilize for 24 h before use in the experiments. model) are among the foods contributing to higher methionine levels . Fatty acid preparation In conclusion, we have demonstrated in the present study that Sodium octanoate (OA) was dissolved in PBS. To bind linoleic acid (Sigma # exposure of breast epithelial cells in vitro to fatty acids results in L8134) to BSA, it was initially dissolved in water to yield a 50 mM ﬁnal epigenetic effects that produce dynamic and profound changes in concentration. 0.12 g of BSA was dissolved in 1.2 ml of water resulting a the expression of genes that have been associated with the 10% BSA solution. A 0.2 ml aliquot of the Na linoleate solution was development of ER- breast cancer (Fig. 7). Next steps include combined with the 10% BSA solution. After 15 min of slow stirring at 37 °C, 0.6 ml of water was added to bring the ﬁnal concentration of Na linoleate demonstrating that these same changes are observed in vivo. As to 5 mMol/L . Linoleic acid bound to BSA (Sigma # L9530) was dissolved mentioned in the introduction, polyunsaturated fatty acids are in water. present in normal breast tissue. Although we measured lipid species in the serum of the donors of the CUB specimens, fatty Lipid analysis acids can also be mobilized from adjacent adipose tissue; LC-MS grade methanol, dichloromethane, and ammonium acetate were adipocytes have been shown to be a reservoir of lipids for breast purchased from Fisher Scientiﬁc (Pittsburgh, PA) and HPLC grade cancer stem cells . We hypothesize that the expression of genes 1-propanol from Sigma-Aldrich (Saint Louis, MO). Milli-Q water was associated with the development of ER- breast cancer is obtained from an in-house Ultrapure Water System by EMD Millipore consequent to lipid stimulation of one-carbon metabolism with (Billerica, MA). The Lipidyzer isotope labeled internal standards mixture resultant changes in histone methylation. Important roles for consisting of 54 isotopes from 13 lipid classes was purchased from Sciex glycolysis, glutaminolysis, lipogenesis, and mitochondrial activity (Framingham, MA). have been demonstrated in oncogenesis; the one-carbon pathway has comparatively received less attention and the insights we Sample preparation provide here generate new questions regarding lipid metabolism Frozen plasma samples were thawed at room temperature (25 °C) for and ER-negative breast cancer, to be pursued in future 30 min, vortexed; 25 μL of plasma was transferred to a borosilicate glass investigations. culture tube (16 × 100 mm). Next, 0.475 mL of water, 1.45 mL of 1:0.45 methanol:dichloromethane, and 25 μL of the isotope labeled internal standards mixture were added to the tube. The mixture was vortexed for MATERIALS AND METHODS 5 s and incubated at room temperature for 30 min. Next, another 0.5 mL of water and 0.45 mL of dichloromethane were added to the tube, followed Cell culture by gentle vortexing for 5 s, and centrifugation at 2500 × g at 15 °C for MCF-10A and MCF-12A cell lines were obtained from American Type 10 min. The bottom organic layer was transferred to a new tube and 0.9 mL Culture Collection (ATCC) and cultured in mammary epithelial cell growth of dichloromethane was added to the original tube for a second extraction. basal medium with single quots supplements and growth factors (#Lonza The combined extracts were concentrated under nitrogen and recon- CC-4136). Cells were treated with the medium-chain fatty acid sodium stituted in 0.25 mL of the mobile phase (10 mM ammonium acetate in octanoate (OA; Sigma # C5038) dissolved in PBS; and long-chain fatty acid 50:50 methanol:dichloromethane). Linoleic acid (LA; Sigma # L8134) complexed with fatty acid-free BSA (Roche 10775835001). Alternatively, Linoleic Acid bound to BSA (LA-BSA; Mass spectrometry Sigma # L9530) was used. PBS and BSA were used as the vehicle control in experiments containing OA and LA, respectively. Cells were counted using Quantitative lipidomics was performed with the Sciex Lipidyzer platform an Invitrogen Countess automated cell counter using the Trypan blue consisting of Shimadzu Nexera X2 LC-30AD pumps, a Shimadzu Nexera X2 exclusion method and seeded at the indicated densities. All experiments SIL-30AC autosampler, and a Sciex QTRAP 5500 mass spectrometer were done in complete MEBM media with fatty acids or vehicle. equipped with SelexION for differential mobility spectrometry (DMS). Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 59 S. Yadav et al. 1-propanol was used as the chemical modiﬁer for the DMS. Samples were ATAC Seq Library preparation and sequencing 6 71 introduced to the mass spectrometer by ﬂow injection analysis at 8 μL/min. 1×10 cells were pelleted and lysed in ATAC-resuspension buffer . Each sample was injected twice, once with the DMS on (PC/PE/LPC/LPE/ Extracted nuclei were processed for TN-5 mediated tagmentation using SM), and once with the DMS off (CE/CER/DAG/DCER/FFA/HCER/LCER/TAG). the Illumina Tagment DNA Enzyme and buffer kit (Nextera Illumina # The lipid molecular species were measured using multiple reaction 20034210): Transposon reaction mix as 2X TD Buffer-25 µl, Tn5 monitoring (MRM) and positive/negative polarity switching. Positive ion Transposase-2.5 µl, 1X PBS containing nuclei-16.5 µl, 10% Tween- mode detected lipid classes SM/DAG/CE/CER/DCER/HCER/DCER/TAG and 20–0.5 µl (Sigma # P9416), 1% Digitonin-0.5 µl (Promega # G9441) and negative ion mode detected lipid classes LPE/LPC/PC/PE/FFA. A total of water at 37 °C, on a thermomixer at 1000 rpm for 30 min. Tagmented DNA 1070 lipids and fatty acids were targeted in the analysis. was isolated by Nucleospin PCR clean-up (Takara Bio USA, Inc # 740609.250). Libraries were ampliﬁed for 8 cycles and puriﬁed using AMPure XP (Agencourt # A63880). Fragment sizes were determined using Data processing 106 LabChip GXII Touch HT (PerkinElmer), and 2 × 50 paired-end Data were acquired and processed using Analyst 1.6.3 and Lipidomics sequencing performed on NovaSeq S1 6000 ﬂow cell (Illumina) to yield Workﬂow Manager 18.104.22.168. For statistical analysis, we evaluated the lipid 100M reads per sample. species enrichments in the ER+,ER−, and control groups. The different groups were compared in pair-wise and the log-fold changes of lipid enrichment were derived, along with the effect sizes and p-values inferred ATAC-seq data sequencing and peak calling from the regression models using the lipid measurement as an input Illumina adapter sequences and low-quality base calls were trimmed off variable and group information as the output variable. the paired end reads with Trim Galore v0.4.3. Sequence reads were aligned to human reference genome hg38 using bowtie2 with default settings. Duplicate reads were discarded with Picard. Reads mapped to mitochon- Library preparation and RNA sequencing drial DNA together with low mapping quality reads were excluded from MCF-10A: RNA was isolated with Qiagen RNeasy Plus Mini Kit (# 74134). further analysis. MACS2 was used to identify the peak regions with options The concentration and quality of total RNA in samples were assessed using -f BAMPE -g hs –keep-dup all -B -q 0.01 . Peaks for samples in the same the Agilent 2100 Bioanalyzer. RNA Integrity Number (RIN) of the vehicle condition were merged using the function ‘merge’ of bedtools and peaks and octanoate sample was 9.9 and 9.8, respectively. Sequencing libraries for samples in different conditions were intersected using the function of were prepared from a total of 100 ng of RNA using KAPA RNA HyperPrep ‘intersect’ of bedtools . Kit. Single-Indexed adapters were obtained from KAPA (catalog# KK8701). Library quality was assessed using the KAPA Library Assay kit. Each indexed library was quantiﬁed and its quality accessed by Qubit and Agilent Differential chromatin accessibility analysis Bioanalyzer, and 6 libraries were pooled in equal molarity. 5 μLof 4nM The number of cutting sites of each sample was counted using the script pooled libraries were denatured, neutralized and a ﬁnal concentration of 74 dnase_cut_counter.py of pyDNase (version 0.2.4) . The raw count matrix 1.5 pM of pooled libraries was loaded to Illumina NextSeq 500 for 75 bp was normalized by CPM. R package edgeR (version 3.16.5) was used to single-read sequencing. Approximately 80 M ﬁltered reads per library was conduct the differential accessibility analysis for all 66,853 common peaks. generated. A Phred quality score (Q score) was used to measure the quality Signiﬁcantly different accessible chromatin regions under different of the sequencing. More than 88% of the sequencing reads reached Q30 conditions were deﬁned as the threshold 0.05 for FDR. With the cutoff 1 (99.9% base call accuracy). Single-end FASTQ reads from RNA-seq for the absolute value of fold change, comparing the treatment group with measurements were aligned and mapped to hg38 ENSEMBL genome vehicle control group, we obtained 1704 signiﬁcantly increased peaks and using STAR alignment . Transcriptions per million (TPM) from mapped 3340 signiﬁcantly decreased peaks. reads were estimated using RSEM from the STAR aligned reads . The DESeq2 Bioconductor R package was employed to determine differen- Motif analysis tially expressed genes for the octanoate treatment group compared to the vehicle-treated controls with FDR cutoff= 0.01 and |log FC| ≥ 2 to identify Motif analysis was conducted for signiﬁcantly changed chromatin regions a reasonable number of differentially expressed genes, on the order of using ‘ﬁndMotifsGenome.pl’ script of HOMER (version 4.9) with default several thousands of genes total, for subsequent analysis settings . The principal component analysis was conducted to detect the MCF-12A: RNA isolation and library preparation as above. The Illumina important motifs using the relative enrichment of motifs calculated from NovaSeq 6000 platform was employed for 100 bp paired-end sequencing. HOMER reports. Biplot was used to visualize the principal component The sequence reads were mapped to the hg38 reference geneome using analysis results. STAR (Spliced Transcripts Alignment to a Reference) . To evaluate the quality of the RNA-seq data, the number of reads that fall into different Genomic distribution of open chromatin regions annotated regions (exonic, intronic, splicing junction, intergenic, promoter, UTR, etc.) of the reference genes was determined with bamUtils . More We calculated the overall genomic distribution of open chromatin regions, than 83% of the sequencing reads reached Q30. Low quality mapped reads comparing the treatment to the vehicle . We used the hg38 refseq genes (including reads mapped to multiple positions) were excluded and annotation from UCSC Genome Browser to deﬁne the genomic features. featureCounts was used to quantify the gene level expression. All TSSs were considered in the analysis if a gene had multiple TSSs. The Differential gene expression analysis was performed with edgeR . In this formula for reported enrichment is (a/b)/(c/d). a is the number of peaks workﬂow, the statistical methodology applied uses negative binomial overlapping a given genomic feature, b is the number of total peaks, c is generalized linear models with likelihood ratio tests. the number of regions corresponding to the feature, and d is the estimated number of discrete regions in the genome where the peaks and feature could overlap. Speciﬁcally, d is equal to (genome size)/ (mean peak size + Gene ontology analysis of differentially expressed genes mean feature size), following the implementation in the bedtools ﬁsher Gene ontology pathway analysis for biological processes was performed (version 2.26.0). on each set of differentially expressed genes using Metascape . Pathway analysis for open chromatin regions GSEA analysis For the 326 open chromatin regions with logFC ≥ 1.5 and FDR < 0.05 Raw counts were ﬁrst estimated using HTSeq from STAR-aligned reads . comparing the treatment with the vehicle, we extracted the target genes Next, replicates for control cells and treated cells were merged and of the 326 chromatin regions. The function ‘enrichKEGG’ from the R normalized using modules from the GenePattern software package . package ‘clusterProﬁle’ (version 3.6.0) was used to conduct KEGG 69,70 GSEA was performed on these DESeq-normalized reads using pathway analysis with organism = ”hsa” and adjusted.pval=0.05. annotations from online databases, including KEGG, Hallmark, Reactome, BioCarta, and Canonical Pathways. The normalized enrichment score (NES) Validation of candidate genes qRT-PCR of these top 20 pathways associated with the control and the octanoate- treated condition is shown with nominal p-value = 0.0. Metascape was Treated cells and organoids were washed with PBS and RNA was isolated employed to perform network analysis on these top 20 pathways with Qiagen RNeasy plus mini Kit (# 74134). cDNA was synthesized using associated with each treatment condition. the SuperScript VILO cDNA synthesis kit (ThermoFisher #11755250). npj Breast Cancer (2022) 59 Published in partnership with the Breast Cancer Research Foundation S. Yadav et al. Real-time qPCR was performed using Applied biosystem QuantStudio 5 cells were transfected with Transcription Factor Reporter, Negative Control real time PCR System (Thermo Scientiﬁc). Expression data of the studied reporter or Positive Control constructs using the Neon Transfection System genes was normalized to RPLP1 to control the variability in expression 10 µl kit (Invitrogen, Walthan, MA; # MPK1025). Twenty-four hours after −ΔΔCT levels and were analyzed using the 2 method described by Livak and transfection, cells were exposed to Sodium octanoate (OA, 5 mM) dissolved Schmittgen . TaqMan gene expression assays and TaqMan fast advanced in PBS or PBS for 24 h. Cells were then lysed using Passive Lysis Buffer master mix (# 4444556) were purchased from ThermoFisher Scientiﬁc and (Promega, Madison, WI) and transferred to a 96-well white ﬂat-bottom the list of the assays is provided in Supplementary File 1. plate (Corning, Tewksbury, MA). Luciferase activity was measured with the Dual-Luciferase Reporter Activity system (Promega; # E1910) using a Biotek Cytation 3 multiwell reader. Fireﬂy luciferase activity was normalized to qRT-PCR based TaqMan low density array assays Renilla luciferase activity. All transfections were performed in triplicate. Based on histological diagnosis, benign breast epithelium was identiﬁed Luciferase measurement for each biological replicate was performed in and captured by laser capture microdissection (LCM). RNA was isolated three technical replicates. Values are expressed as mean ± SEM. The p- with Qiagen RNeasy plus mini Kit (# 74134). RNA quality was checked for integrity using Bioanalyzer 2100 (Agilent). 100 ng RNA was reverse value was calculated by unpaired t-test. transcribed using High Capacity RNA-to-cDNA Master Mix (ThermosFisher #4388950) and preampliﬁed for 14 cycles using TaqMan PreAmp Master Flux based analysis (FBA) Mix 2X (ThermoFisher #4488593) and pooled assay mix for the genes in which we were interested. Pre-ampliﬁed cDNA were diluted to 1:20 with 1X We calculated the relative activity of reactions in MCF-10A and MCF-12A TE buffer and mixed with Fast advanced master mix (ThermoFisher # cells by interpreting gene expression data using the Recon1 human 44,81 4444965) Each sample was loaded in duplicate in 384-well microﬂuidic metabolic model augmented with histone modiﬁcations . We then cards customized with 47 genes of interest including three housekeeping identiﬁed a metabolic ﬂux state that is most consistent with gene genes (GAPDH, RPLP0, and RPLP1). TaqMan assays with best coverage expression data in control and octanoate treatment. This was achieved by attribution were used for the TLDA study as recommended by the maximizing the activity of reactions that are associated with upregulated manufacturer. A list of the genes and the Assay ID for the primers obtained genes and minimizing ﬂux through reactions that are downregulated in a from ThermoFisher is provided in Supplementary File 2. Real Time PCR condition, while simultaneously satisfying the stoichiometric and thermo- reactions were carried out in QuantStudio 7 Flex system for 40 cycles using 44,81 dynamic constraints embedded in the model using linear optimization . comparative Ct (ΔΔCt) method. Results were analyzed using Expression- The glucose, fatty acid, and glutamine levels in the simulations were Suite software. adjusted based on the growth media used for culturing the cells. All p- values were corrected for multiple comparisons. Statistical analysis Prior to performing the analyses, the log2-transformed relative (log2RE) amounts of mRNA expression were normalized to GAPDH and expressed LC/MS based post-translation histone modiﬁcation −(CtX−CtGAPDH) as log 2 = −(CtX − CtGAPDH), where Ct is threshold cycle. 2 quantitation The Mann–Whitney test was performed to identify genes with pairwise MCF-10A were treated with 5 mM octanoate or 500 µM LA. After 24 h, cells differences between ER+ and ER− samples. The analyses were adjusted were snap frozen for nuclei extraction. Histones were acid-extracted from for multiple testing, 34 genes, using the Benjamini–Hochberg (BH) 100% nuclei, derivatized via propionylation reaction and digested with adjustment in order to control the false discovery rate at the two-sided trypsin. Each sample was resuspended in 50 µL of 0.1% TFA/mH2O and 2 µl 0.05 level. Boxplots were used to visualize differences in log2RE by group. was injected with 3 technical replicates. Multi-reaction monitoring (MRM) The log2RE analyses were conducted using the R statistical environment technology was used for histone analysis using a triple-quad mass [R] version 3.5.1. spectrometer, which is programmed to fragment only speciﬁc precursor peptides and measure the intensity of speciﬁc product ions. Final results Live cell PWS imaging show changes of relative abundances of histone mark modiﬁcation. Error Before treatment and imaging, MCF-10A cells were seeded in 6 well black bars are +/− one standard deviation obtained from sample technical culture plates, at least 35% conﬂuency, and allowed to adhere overnight replicate intensities. before the treatment with 500 µM LA and 5 mM Octanoate. We based the concentration of LA used in the experiment on the range in human plasma: 0.2–5.0 mmol/L . To determine the chromatin packing behavior of Western blotting MCF-10A cells under varying treatment conditions, live-cell PWS images Cells and organoids were plated and allowed to stabilize overnight and were acquired at 37 °C and 5% CO conditions. Imaging was performed then treated the next day for the indicated times. At the end of the using the commercial inverted microscope (Leica DMIRB) Hamamatsu treatment, cells were collected, washed with PBS, and lysed in radio Image-EM CCD camera C9100-13 coupled to a liquid crystal tunable ﬁlter immunoprecipitation assay (RIPA) buffer (ThermoFisher Scientiﬁc # 89900) (LCTF; CRi Woburn, MA) to acquire mono-chromatic spectrally resolved including protease inhibitors (ThermoFisher Scientiﬁc # 78430). Protein is images that range from 500 to 700 nm at 1 nm intervals produced by a estimated using the BCA protein assay kit (ThermoFisher Scientiﬁc # 23227) broad band illumination provided by an Xcite-120 LED Lamp (Excelitas, 79,80 and loaded on 4–12% Bis Tris acetate gel using MES buffer, blotted on a Waltham, MA) as previously described . Brieﬂy, PWS measures the polyvinylidene ﬂuoride (PVDF) membrane (Invitrolon 0.45 μM) and blocked standard deviation of internal optical scattering originating from with blocking buffer (10% skimmed milk) for 1 h at room temperature. chromatin in the nucleus, which is related to variations in the refractive index distribution (Σ). To obtain the interference signal directly related to Primary antibodies were purchased from Cell Signaling Technologies – refractive index ﬂuctuations in the cell, we normalized measurements to an AcH3K9 (rabbit mAb Cell Signaling #9649) at 1:250 dilution, AcH3K14 independent reference measurement acquired in an area of the plate (rabbit mAb Cell Signaling #7627) at 1:250 dilution and H3 (Rabbit mAb without cells. These normalized spatial variations of refractive index are Cell Signaling #9715) at 1:1000 dilution. Membranes were incubated in linearly proportional to nuclear mass density ﬂuctuations, according to the primary antibody overnight at 4 degrees Celsius with shaking. Blots were Gladstone–Dale relation, and are characterized by chromatin packing washed with PBS+ 0.1% Tween 20, three times 5 min each, and probed scaling, D, the power-law relationship between the mass M of the with secondary antibodies (Anti-rabbit IgG, HRP-linked Antibody, Cell chromatin polymer and the three-dimensional space it occupies R, i.e., Signaling #7074) at a concentration of 1:10,000 for 1 h at room D 80 M~R . The measured change in chromatin packing scaling between temperature. Blots/lysates were derived from the same experiment and treatment conditions was quantiﬁed by ﬁrst averaging D within each cell’s were processed concurrently. Uncropped images of the original Western nucleus and then averaging nuclei from over 100 cells per condition. blots are provided in Supplementary File 3. Notch reporter assay Reporting summary Notch pathway function was analyzed by measuring the transcriptional activity of its downstream component RBP-jk using a Cignal RBP-Jk Dual Further information on research design is available in the Nature Research Luciferase Reporter assay (Qiagen, Germantown, MD; # 336841). MCF-10A Reporting Summary linked to this article. Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 59 S. Yadav et al. DATA AVAILABILITY 23. Schönfeld, P. & Wojtczak, L. Short- and medium-chain fatty acids in energy metabolism: The cellular perspective. J. Lipid Res. 57, 943–954 (2016). The datasets generated and analyzed during the current study are publicly available 24. Virk, R. K. A. et al. Disordered chromatin packing regulates phenotypic plasticity. in the Gene Expression Omnibus: accession numbers GSE126799 (RNA-seq, MCF- Sci. Adv. 6, eaax6232 (2020). 10A), GSE190572 (RNA-seq, MCF-12A), and GSE190573 (ATAC-seq). 25. Sweeney, M. F., Sonnenschein, C. & Soto, A. M. Characterization of MCF-12A cell phenotype, response to estrogens, and growth in 3D. Cancer Cell Int. 18,43 (2018). CODE AVAILABILITY 26. Whelan, J. & Fritsche, K. Linoleic acid. Adv. Nutr. 4, 311–312 (2013). No custom codes were written to analyze the data presented in this manuscript. The 27. Ahel, D. et al. Poly(ADP-ribose)-dependent regulation of DNA repair by the following software versions were used: Trim Galore v0.4.3.; pyDNase version 0.2.4; R chromatin remodeling enzyme ALC1. Science 325, 1240–1243 (2009). package edgeR version 3.16.5; HOMER version 4.9 with default settings; bedtools 28. Guyenet, S. J. & Carlson, S. E. Increase in adipose tissue linoleic acid of US adults ﬁsher version 2.26.0; ‘enrichKEGG’ from the R package ‘clusterProﬁle’ version 3.6.0; R in the last half century. Adv. Nutr. 6, 660–664 (2015). statistical environment [R] version 3.5.1. 29. Almassalha, L. M. et al. Macrogenomic engineering via modulation of the scaling of chromatin packing density. Nat. Biomed. Eng. 1, 902–913 (2017). Received: 26 December 2020; Accepted: 31 March 2022; 30. Almassalha, L. M. et al. The global relationship between chromatin physical topology, fractal structure, and gene expression. Sci. Rep. 7, 41061–41061 (2017). 31. Almassalha, L. M. et al. The greater genomic landscape: The heterogeneous evolution of cancer. Cancer Res. 76, 5605–5609 (2016). 32. Yadav, S. et al. Abstract P3-02-07: Fatty acid metabolism is associated with chromatin remodeling in mammary epithelial cells. Cancer Res. https://doi.org/ REFERENCES 10.1158/1538-7445.SABCS19-P3-02-07 (2020). 1. Weigelt, B. & Reis-Filho, J. S. Histological and molecular types of breast cancer: Is 33. Heinz, S. et al. Simple combinations of lineage-determining transcription factors there a unifying taxonomy? Nat. Rev. Clin. Oncol. 6, 718–730 (2009). prime cis-regulatory elements required for macrophage and B cell identities. Mol. 2. Gail, M. H. et al. Projecting individualized probabilities of developing breast Cell 38, 576–589 (2010). cancer for white females who are being examined annually. J. Natl Cancer Inst. 34. Steensels, S. & Ersoy, B. A. Fatty acid activation in thermogenic adipose tissue. 81, 1879–1886 (1989). Biochim. Biophys. Acta Mol. Cell Biol. Lipids 1864,79–90 (2019). 3. Fisher, B. et al. Tamoxifen for prevention of breast cancer: report of the National 35. McDonnell, E. et al. Lipids reprogram metabolism to become a major carbon Surgical Adjuvant Breast and Bowel Project P-1 Study. J. Natl Cancer Inst. 90, source for histone acetylation. Cell Rep. 17, 1463–1472 (2016). 1371–1388 (1998). 36. O’Brien, E. J., Monk, J. M. & Palsson, B. O. Using genome-scale models to predict 4. Cuzick, J. et al. Long-term results of tamoxifen prophylaxis for breast cancer–96- biological capabilities. Cell 161, 971–987 (2015). month follow-up of the randomized IBIS-I trial. J. Natl Cancer Inst. 99, 272–282 37. Orth, J. D., Thiele, I. & Palsson, B. O. What is ﬂux balance analysis? Nat. Biotechnol. (2007). 28, 245–248 (2010). 5. Cuzick, J. et al. Use of anastrozole for breast cancer prevention (IBIS-II): Long-term 38. Chandrasekaran, S. et al. Comprehensive mapping of pluripotent stem cell results of a randomised controlled trial. Lancet 395, 117–122 (2020). metabolism using dynamic genome-scale network modeling. Cell Rep. 21, 6. Bouchardy, C. et al. Risk of second breast cancer according to estrogen receptor 2965–2977 (2017). status and family history. Breast Cancer Res. Treat. 127, 233–241 (2011). 39. Frezza, C. et al. Haem oxygenase is synthetically lethal with the tumour sup- 7. Kurian, A. W. et al. Second primary breast cancer occurrence according to hor- pressor fumarate hydratase. Nature 477, 225–228 (2011). mone receptor status. J. Natl Cancer Inst. 101, 1058–1065 (2009). 40. Mardinoglu, A. et al. Personal model-assisted identiﬁcation of NAD(+) and glu- 8. Swain, S. M. et al. Estrogen receptor status of primary breast cancer is predictive tathione metabolism as intervention target in NAFLD. Mol. Syst. Biol. 13, 916 of estrogen receptor status of contralateral breast cancer. J. Natl Cancer Inst. 96, (2017). 516–523 (2004). 41. Shlomi, T., Cabili, M. N., Herrgard, M. J., Palsson, B. O. & Ruppin, E. Network-based 9. Wang, J. et al. Lipid metabolism genes in contralateral unaffected breast and prediction of human tissue-speciﬁc metabolism. Nat. Biotechnol. 26, 1003–1010 estrogen receptor status of breast cancer. Cancer Prev. Res. 6, 321–330 (2013). (2008). 10. Wang, J. et al. Overexpression of lipid metabolism genes and PBX1 in the con- 42. Yizhak, K., Chaneton, B., Gottlieb, E. & Ruppin, E. Modeling cancer metabolism on tralateral breasts of women with estrogen receptor-negative breast cancer. Int. J. a genome scale. Mol. Syst. Biol. 11, 817 (2015). Cancer 140, 2484–2497 (2017). 43. Duarte, N. C. et al. Global reconstruction of the human metabolic network based 11. Hanahan, D. & Weinberg, R. A. Hallmarks of cancer: The next generation. Cell 144, on genomic and bibliomic data. Proc. Natl Acad. Sci. USA 104, 1777–1782 (2007). 646–674 (2011). 44. Shen, F., Boccuto, L., Pauly, R., Srikanth, S. & Chandrasekaran, S. Genome-scale 12. Nieman, K. M. et al. Adipocytes promote ovarian cancer metastasis and provide network model of metabolism and histone acetylation reveals metabolic energy for rapid tumor growth. Nat. Med. 17, 1498–1503 (2011). dependencies of histone deacetylase inhibitors. Genome Biol. 20, 49 (2019). 13. Mitra, R., Le, T. T., Gorjala, P. & Goodman, O. B. Jr. Positive regulation of prostate 45. Carrer, A. et al. Impact of a high-fat diet on tissue acyl-CoA and histone acet- cancer cell growth by lipid droplet forming and processing enzymes DGAT1 and ylation levels. J. Biol. Chem. 292, 3312–3322 (2017). ABHD5. BMC Cancer 17, 631 (2017). 46. Li, J. T. et al. BCAT2-mediated BCAA catabolism is critical for development of 14. Singh, K. B. et al. Prostate cancer chemoprevention by sulforaphane in a pre- pancreatic ductal adenocarcinoma. Nat. Cell Biol. 22, 167–174 (2020). clinical mouse model is associated with inhibition of fatty acid metabolism. 47. Kerlikowske, K. et al. Risk factors that increase risk of estrogen receptor-positive Carcinogenesis 39, 826–837 (2018). and -negative breast cancer. J. Natl Cancer Inst. https://doi.org/10.1093/jnci/ 15. Huang, Li,T. et al. HIF-1-mediated suppression of acyl-CoA dehydrogenases and fatty djw276 (2017). acid oxidation is critical for cancer progression. Cell Rep. 8,1930–1942 (2014). 48. Li, K. et al. Risk prediction for estrogen receptor-speciﬁc breast cancers in two 16. Camarda, R. et al. Inhibition of fatty acid oxidation as a therapy for MYC- large prospective cohorts. Breast Cancer Res. 20, 147 (2018). overexpressing triple-negative breast cancer. Nat. Med. 22, 427–432 (2016). 49. Worsham, M. J. et al. Multiplicity of benign breast lesions is a risk factor for 17. Park, J. H. et al. Fatty acid oxidation-driven Src links mitochondrial energy progression to breast cancer. Clin. Cancer Res. 13, 5474–5479 (2007). reprogramming and oncogenic properties in triple-negative breast cancer. Cell 50. Alonzo-Proulx, O., Mainprize, J. G., Harvey, J. A. & Yaffe, M. J. Investigating the Rep. 14, 2154–2165 (2016). feasibility of stratiﬁed breast cancer screening using a masking risk predictor. 18. Havas, K. M. et al. Metabolic shifts in residual breast cancer drive tumor recur- Breast Cancer Res. 21, 91 (2019). rence. J. Clin. Invest. 127, 2091–2105 (2017). 51. Reiner, A. S. et al. Hormone receptor status of a ﬁrst primary breast cancer 19. Cappelletti, V. et al. Metabolic footprints and molecular subtypes in breast cancer. predicts contralateral breast cancer risk in the WECARE study population. Breast Dis. Markers 2017, 7687851 (2017). Cancer Res. 19, 83 (2017). 20. You, S. et al. Raman spectroscopic analysis reveals abnormal fatty acid compo- 52. Rodilla, V. et al. Luminal progenitors restrict their lineage potential during sition in tumor micro- and macroenvironments in human breast and rat mam- mammary gland development. PLoS Biol. 13, e1002069 (2015). mary cancer. Sci. Rep. 6, 32922 (2016). 53. Lilja, A. M. et al. Clonal analysis of Notch1-expressing cells reveals the existence of 21. Reid, M. A., Dai, Z. & Locasale, J. W. The impact of cellular metabolism on chro- unipotent stem cells that retain long-term plasticity in the embryonic mammary matin dynamics and epigenetics. Nat. Cell Biol. 19, 1298–1306 (2017). gland. Nat. Cell Biol. 20, 677–687 (2018). 22. Grabacka, M., Pierzchalska, M., Dean, M. & Reiss, K. Regulation of ketone body 54. Milne, R. L. et al. Identiﬁcation of ten variants associated with risk of estrogen- metabolism and the role of PPARalpha. Int. J. Mol. Sci. https://doi.org/10.3390/ receptor-negative breast cancer. Nat. Genet. 49, 1767–1778 (2017). ijms17122093 (2016). npj Breast Cancer (2022) 59 Published in partnership with the Breast Cancer Research Foundation S. Yadav et al. 55. Li, Y., Dillon, T. J., Takahashi, M., Earley, K. T. & Stork, P. J. Protein kinase 81. Shen, F., Cheek, C. & Chandrasekaran, S. Dynamic network modeling of stem cell A-independent ras protein activation cooperates with Rap1 protein to mediate metabolism. Methods Mol. Biol. 1975, 305–320 (2019). activation of the extracellular signal-regulated kinases (ERK) by cAMP. J. Biol. Chem. 291, 21584–21595 (2016). 56. Creighton, C. J. et al. Activation of mitogen-activated protein kinase in estrogen ACKNOWLEDGEMENTS receptor alpha-positive breast cancer cells in vitro induces an in vivo molecular We would like to thank Professors Matthew D Hirschey and Neil Kelleher for phenotype of estrogen receptor alpha-negative human breast tumors. Cancer advice regarding histone proteomics and, Jeannie Camarillo and the North- Res. 66, 3903–3911 (2006). western Proteomics Core for conducting the histone proteomic analysis; The 57. Mentch, S. J. et al. Histone methylation dynamics and gene regulation occur Northwest Metabolic Research Center (NW-MRC) at University of Washington for through the sensing of one-carbon metabolism. Cell Metab. 22, 861–873 (2015). performing the lipidomics analysis; The Center for Medical Genomics at the 58. Wang, T. et al. JAK/STAT3-regulated fatty acid beta-oxidation is critical for breast Indiana University School of Medicine for RNA library preparation and RNA cancer stem cell self-renewal and chemoresistance. Cell Metab. 27, 136–150 e135 sequencing, and ATAC sequencing; The NU Seq Core facility for providing The (2018). QuantStudio 7 Flex system; Natalie Pulliam for consenting patients and collecting 59. Pappas, A., Anthonavage, M. & Gordon, J. S. Metabolic fate and selective utili- tissue, and our many lab colleagues for feedback. This work was supported by the zation of major fatty acids in human sebaceous gland. J. Invest. Dermatol. 118, Breast Cancer Research Foundation (S.A.K.), Bramsen-Hamill Foundation (S.A.K.), 164–171 (2002). Otzen Family Foundation (S.E.C.), Ulla and Bertil Brunk Foundation (S.E.C.), R35 60. Dobin, A. et al. STAR: Ultrafast universal RNA-seq aligner. Bioinformatics 29,15–21 GM13779501 from The National Institute of General Medical Sciences, NIH (S.C.). (2013). 61. Li, B. & Dewey, C. N. RSEM: accurate transcript quantiﬁcation from RNA-Seq data with or without a reference genome. BMC Bioinform. 12, 323 (2011). AUTHOR CONTRIBUTIONS 62. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and Conceptualization and project design: S.Y., S.E.C., and S.A.K.; experiments: S.Y. and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014). M.B.E.; sequencing: X.X and H.G; RNA-seq analysis: R.K.A.V., Z.Z., G.C, and M.R.; 63. Breese, M. R. & Liu, Y. NGSUtils: A software suite for analyzing and manipulating ATAC Seq analysis: D.C.; PWS microscopy: D.V.D.; metabolic ﬂux analysis: C.H.C. next-generation sequencing datasets. Bioinformatics 29, 494–496 (2013). and S.C.; statistical analysis: K.L.B.; writing (original draft): S.Y., S.E.C., and S.A.K.; 64. Liao, Y., Smyth, G. K. & Shi, W. featureCounts: An efﬁcient general purpose pro- writing (review and editing): V.B., S.C., and R.C. The authors approved the gram for assigning sequence reads to genomic features. Bioinformatics 30, completed version. 923–930 (2014). 65. Robinson, M. D., McCarthy, D. J. & Smyth, G. K. edgeR: A Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics 26, COMPETING INTERESTS 139–140 (2010). 66. Zhou, Y. et al. Metascape provides a biologist-oriented resource for the analysis The authors declare no competing interests. of systems-level datasets. Nat. Commun. 10, 1523 (2019). 67. Anders, S., Pyl, P. T. & Huber, W. HTSeq−a Python framework to work with high- throughput sequencing data. Bioinformatics 31, 166–169 (2015). ADDITIONAL INFORMATION 68. Reich, M. et al. GenePattern 2.0. Nat. Genet. 38, 500–501 (2006). Supplementary information The online version contains supplementary material 69. Mootha, V. K. et al. PGC-1α-responsive genes involved in oxidative phosphor- available at https://doi.org/10.1038/s41523-022-00422-0. ylation are coordinately downregulated in human diabetes. Nat. Genet. 34, 267–273 (2003). Correspondence and requests for materials should be addressed to Seema Ahsan 70. Subramanian, A. et al. Gene set enrichment analysis: A knowledge-based Khan or Susan E. Clare. approach for interpreting genome-wide expression proﬁles. Proc. Natl Acad. Sci. USA 102, 15545 (2005). Reprints and permission information is available at http://www.nature.com/ 71. Ackermann, A. ATAC-seqProtocolOmni-KaestnerLab.pdf. (2019). reprints 72. Zhang, Y. et al. Model-based analysis of ChIP-Seq (MACS). Genome Biol. 9, R137 (2008). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims 73. Quinlan, A. R. & Hall, I. M. BEDTools: A ﬂexible suite of utilities for comparing in published maps and institutional afﬁliations. genomic features. Bioinformatics 26, 841–842 (2010). 74. Piper, J. et al. Wellington: A novel method for the accurate identiﬁcation of digital genomic footprints from DNase-seq data. Nucleic Acids Res. 41, e201 (2013). 75. Miyamoto, K. et al. Chromatin accessibility impacts transcriptional reprogram- Open Access This article is licensed under a Creative Commons ming in Oocytes. Cell Rep. 24, 304–311 (2018). Attribution 4.0 International License, which permits use, sharing, 76. Yu, G., Wang, L. G., Han, Y. & He, Q. Y. clusterProﬁler: An R package for comparing adaptation, distribution and reproduction in any medium or format, as long as you give biological themes among gene clusters. OMICS 16, 284–287 (2012). appropriate credit to the original author(s) and the source, provide a link to the Creative 77. Livak, K. J. & Schmittgen, T. D. Analysis of relative gene expression data using real- Commons license, and indicate if changes were made. The images or other third party time quantitative PCR and the 2(-Delta Delta C(T)) method. Methods 25, 402–408 material in this article are included in the article’s Creative Commons license, unless (2001). indicated otherwise in a credit line to the material. If material is not included in the 78. Abdelmagid, S. A. et al. Comprehensive proﬁling of plasma fatty acid con- article’s Creative Commons license and your intended use is not permitted by statutory centrations in young healthy Canadian adults. PLoS One 10, e0116195 (2015). regulation or exceeds the permitted use, you will need to obtain permission directly 79. Almassalha, L. M. et al. Label-free imaging of the native, living cellular from the copyright holder. To view a copy of this license, visit http://creativecommons. nanoarchitecture using partial-wave spectroscopic microscopy. Proc. Natl Acad. org/licenses/by/4.0/. Sci. USA 113, E6372–E6381 (2016). 80. Eid, A. et al. Characterizing chromatin packing scaling in whole nuclei using interferometric microscopy. Opt. Lett. 45, 4810–4813 (2020). © The Author(s) 2022 Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 59
npj Breast Cancer – Springer Journals
Published: May 4, 2022
Access the full text.
Sign up today, get DeepDyve free for 14 days.