Oncogenic cell fate decision in breast epithelial cells I: growth factors and mechanisms

NIH LINCS program contributed to the creation and preprocessing of an MCF10a perturbation dataset at Oregon Health and Science University (OHSU). MCF10a cells were treated with ligands as follows: 10 ng/ml EGF, 40 ng/ml HGF, and 10 ng/ml OSM. Samples were collected at 1, 4, 8, 24 and 48 hours post treatment for RPPA and cyclic-IF measurements, and 24 and 48 hours for RNA-seq and ATAC-Seq. RPPA protein samples were prepared and preprocessed according to the standard preprocessing methods developed at the MD Anderson Cancer Center RPPA core and passed QC checks. The normalized RPPA log2 values are joined with their experimental metadata obtained from the MD Anderson cancer center and stored as level 3 data. Replicates were median summarized in the RPPA Level-4 data and obtained from Synapse for 1, 4, 8, 24, and 48-hour time points for treatment versus PBS control. We used a log 2 fold change (FC) cutoff for downstream functional analysis. |FC| >1.3 RNA-seq reads were mapped to the GRCh38.99 human transcriptome using OMICSOFT. The resulting count matrix normalization and differential analysis was performed using the “DEseq2” package available through R/Bioconductor[ 40 ]. Genes with at least one count, were retained for further analysis within each matrix. The pairwise comparisons for each treatment at each timepoint were performed with respect to their respective PBS controls at every timepoint. DEGs were called at a Benjamini-Hochberg false adjusted p < 0.05. ATAC-seq data preprocessing was performed using TrimGalore! package, removing sequencing adaptors and selecting for all paired-end sequences. Trimmed reads were aligned to the GRCh38 human genome using BBMap v37.95 in the BBTools suite followed by sorting and indexing of bam files using SAMtools v1.3, and annotation of PCR duplicates using the Picard v2.3.0 MarkDuplicates function. All duplicates and mitochondrial, chromosome X, chromosome Y, and Epstein-Barr virus (EBV) reads were removed using SAMtools v1.3. Only seventeen different samples were suitable for further analysis. MACS2 was used to call peaks on the ATAC-seq data and determine regions of open chromatin. The specifically called open regions of chromatin were then passed to the R package Diffbind v2.8.0 to determine regions of differential accessibility between each control and treatment conditions using DESEQ2. Differentially accessible regions of DNA were annotated using the R package, ChIPseeker, defining the promoter region −1000 to 500 bp from the TSS. Genes with differential chromatin expression were defined as those called DEGs (|FC| > 1.3 and FDR P-adjusted p-value 1.3 and adjusted p-value FDR P < 0.05) occurring in a promoter. ATAC-seq enrichment analysis was conducted to determine the TF footprints associated with the gain or loss of chromatin accessibility and motifs enriched for these regions using GimmeMotifs. The gimmemotifs function was used with both the HOCOMOCOv11 pwm databases and the Homer motif-finding algorithm. Pathway enrichment analysis was identified using hypergeometric enrichment of DEGs in KEGG, Gene ontology, Hallmark and Reactome databases. Functional pathways describing temporal activation were constructed through a combination of curated pathways and manual literature search. Enrichment terms ranked by their p-value are presented for any comparative analysis. Using the TRANSFAC database (version 2020), TFs-target relationships were identified, and TFs were enriched using hypergeometric enrichment analysis using an R package. The identified transcriptional networks, including TFs and their targets, were further filtered based on whether they were significantly differentially expressed genes (DEGs) in treatment vs. control at 24 and 48 hours, respectively. Identification of ligand-specific gene expression by pairwise comparison across ligand treatments was performed for all combinations of three treatments and controls to find differentially regulated genes between any two compared conditions. To find the ‘ligand specific’ outcomes for a condition, the intersection of three pairwise comparisons, obtained between the condition and all other conditions, was taken. This cross intersection allowed us to remove the commonly regulated genes. For example, to find EGF-specific genes, we first applied the DESEQ2 between EGF and control (PBS), giving a list of genes differentially regulated in EGF versus PBS. Then the same conditions were applied between EGF and OSM; EGF and HGF respectively, each of which returned a list of genes differentially regulated in EGF versus the compared condition. p-value cutoff ≤0.05 was used for all the cases. The intersection of these lists represented the genes differentially regulated in EGF compared with all other conditions. We denoted this intersection as EGF-specific genes. Similarly, this method was repeated for HGF, and OSM, revealing the ligand specific genes for each condition.

The cell fate of epithelial cells to maintain homeostasis or transform to tumor cells results from a dynamic interplay between cellular mechanisms involving signaling and growth factors[ 41 ]. In this project, MCF10a cells was treated with growth-regulatory ligands EGF, HGF and OSM to assess mechanisms involved in the cell fate processes. Alterations in proteins were measured at 1, 4, 8, 24, and 48 hours post treatment using reverse phase protein array (RPPA), alterations in transcripts and chromatin accessibility measured at 24 and 48 hours post treatment using RNA-seq and ATAC-seq methods, and protein expression and localization were monitored at 1, 4, 8, 24, and 48 hours post treatment using cyclic-immunofluorescence (cyclic-IF) method (Fig. S1). RPPA measurements showed significant overlap in response to the three treatments (EGF, HGF, and OSM), with alteration at least at one time point for 95 proteins with EGF, 93 with HGF, and 89 with OSM treatments with respect to PBS control ( Fig. 1a ). RPPA characterization of canonically defined oncoproteins revealed changes in expression of 44 oncoproteins[ 42 – 44 ] ( Fig. 1c ). Analysis of all significant proteins at each individual time point revealed consistent functional enrichment of several modules across all treatments. The functional enrichment included an increase in hallmark oncogenic phenotypes such as cell growth and proliferation (cell cycle, mTOR signaling, PI3K signaling, MAPK signaling, cellular senescence, TGFβ signaling), inflammation (HIF1α signaling, TNFα signaling), angiogenesis (AMPK signaling, VEGF signaling, ERK signaling), and cytoskeletal changes (gap junctions, tight junctions) ( Fig. 1b (i – iii) ). Additionally, OSM treatment showed an increase in EMT processes (Hippo signaling, VEGF, angiogenesis) and pro-inflammatory processes (chemokine signaling, IL17 differentiation). All three treatments activated Receptor tyrosine kinases (RTKs) (HER2_pY1248 and HER3_pY1289) that can lead to the activation of PI3K/mTOR, MEK/ERK, and STAT3 signaling pathways ( Fig.1d ). However, EGF showed a downregulation of both EGFR and ERBB2 total proteins. Similar activation of AKT, ERK and mTOR was observed as early as 1 hour, along with an increase in activation of its targets EIF4EBP1, S6K at later time points as measured by RPPA in all three treatments. An increase in mTOR1 via PIK3CA led to the increase in the protein expression of TP53BP1, PXN, FOXM1, EIF4EBP1, RPS6, and FOXO3. PIK3CA is also known to the increase phosphorylation of AKT and S6 which play a significant role in proliferation, angiogenesis, metabolism, and loss of apoptosis. Further inactivation of CDKN1A, CDKN1B was observed in all three treatments, indicating cell cycle progression. However, although proteins such as AKT, STAT3 and GSK3B showed consistent trends in all three treatments, EGF showed a significant increase in the expression of phosphorylated AKT-pS473/pS308, ERBB2-pY1248 and GSK3A-pS21-S9 compared the other treatments, indicating treatment-specific activation of downstream pathways. Additionally, in OSM treatment, phosphorylation of STAT3-pY705 led to its nuclear localization as seen in the cyclicIF measurements and triggered enhanced phenotypic changes in cell cycle (CCND2, MYC), anti-apoptosis (BCL2), EMT (SNAIL, ZEB1, HIF1A), and Metastasis (SERPINE, MYC)[ 45 , 46 ]. HIF1A further upregulated VEGF and ZEB1, thus, promoting EMT, migration, and invasion. RPPA data showed the upregulation of several transcription factors (TF). To further characterize the role of RPPA-TFs in altering transcriptional regulation, we performed an analysis of the RPPA-TFs and their differentially expressed (DE) target genes obtained from RNA-seq analysis. Functional enrichment was performed for the targets of each TF individually. Analysis showed common modules such as cell cycle, ECM interaction, mTOR signaling pathway, metabolism, inflammation, and EMT across multiple TFs and downregulation of cellular senescence, endocytosis, and p53 mediated apoptosis ( Fig 2a ). EMT causes dissociation of cells, loss of polarity, and an increase in the plasticity of cells[ 47 – 50 ]. EMT-related processes such as HIF1α signaling, proteoglycans in cancer, RAP signaling were enriched in the above RPPA-TF-target analysis. Canonical markers of EMT[ 51 – 53 ] such as CDH2, VIM, SMN1, CTNNB1, and FN1 were significantly upregulated and epithelial markers such as CDH1, OCLN and CD24 were downregulated in all three treatments ( Fig. 2b ). All three ligands showed induction of EMT via mTOR, MAPK, and ERK pathways. These pathways regulate E2F1 which acts as an EMT regulator directly via mTOR and indirectly via ZEB1/2. E2F1 also upregulated MMPs and ADAM12 which further upregulated EMT[ 54 ]. In OSM, the transcriptomic regulation of HIF1α, TGFRβ1, JUN, and FOS via STAT3 led to the further activation of TFs (SNAI1and ZEB1) which regulate EMT gene expression[ 16 , 55 – 58 ] To evaluate overall transcriptional changes that contribute to downstream functional changes, we analyzed all DE genes in EGF, HGF, and OSM treatments at 24 and 48 hours (Fig. S2). Functional enrichment showed cell cycle, metabolism, mTOR signaling, PI3K signaling, MAPK signaling, HIF1α signaling, and regulation of cytoskeleton consistent with the proteo-transcriptomic analysis. In addition, we also observed an increase in gene expression associated with DNA repair pathways, ribosome biosynthesis, metabolism, protein biosynthesis, and ECM interactions common to all three treatments. ( Fig. 3a ). The increase in proliferation, DNA damage repair pathways, and change in the extracellular matrix proteins suggest putative mechanisms for oncogenesis and tumor migration. Association of DEGs with the KEGG enrichment showed mTOR regulated downstream protein synthesis via S6K, EIF4B, nucleotide biosynthesis via ATF4, MTHFD2, S6K, loss of autophagy via FOXO3, ULK1, and cytoskeletal rearrangement via RHO in all three treatments (Fig. S3). We further analyzed the oncogenesis gene list that was curated previously to investigate the genome-wide expression of oncogenes in all three treatments ( Fig. 3c ). Increased expression of oncogenes indicated an increase in oncogenic potential consistent with the proteomic analysis. TF enrichment was performed for all DE TFs based and their DE targets and EMT genes as targets from TRANSFAC database. Top TFs were associated with modules such as cell cycle (E2F1, SP3, EGR1), metabolism (NFYA, NFYB), growth, and proliferation (JUN, FOS) ( Fig. 3b ). EMT target genes were regulated by JUN, TCF7L2, AR, and HIF1A. To decipher the ligand-specific transcriptional changes, we performed a pairwise comparison analysis using a previously published method (see Methods )[ 59 ]. Functional enrichment of ligand-specific DE genes showed regulation of cell cycle, DNA damage repair proteins, PI3K, FOXO, p53 signaling, loss of apoptosis, ECM interaction, focal adhesion, HIPPO signaling, and TGFβ signaling in EGF treatment, and TH17, Osteoclast differentiation, NF-κB, JAK-STAT, IL-17, TNFα signaling, cytokine-receptor interacting and NOD signaling in OSM treatment. We observed TGFβ signaling in EGF and OSM treatments. In EGF treatment EMP3, TGFBR2 RHOA and MAPK1 were upregulated while in OSM treatment, we see upregulation of TGFBR1, SMAD3, ZEB1 and NREP was driven by TGFBR2 activation and its downstream effects and TGFBR1 activation in OSM ( Fig. 3f ). Treatment-specific TFs and their targets-based gene networks were constructed, and key hub genes such as E2F1, ETS1, and NRF1 were identified ( Fig. 3d and e ). E2F1 played a key role in EGF specific treatment and led to the regulation of multiple DE targets that participate in cell cycle (CCND1, CDC6, CDC25a, PTTG1, CDK2, PLK1, DFB4, BRCA1, CCNB2, CCNE2) and EMT (SMC2, FGF2, FGFR1, CDKN2C, NRF1). Other hub genes identified in the EGF treatment network were BRCA1, NRF1, THBS1, HMGA1, and FOSL1. These genes contribute to changes in cytoskeletal organization, increased tumor progression, and invasion aiding EMT[ 60 ]. OSM treatment showed a repertoire of hub TF such as JUN, FOS, SMAD3, HIF1A, STAT3, CREB1, ESR1, and CEBPB. These TFs regulate EMT, inflammation, and cell cycle. Thus, the transcriptomic analysis shows that E2F1 and its targets were regulated in all three treatments; however, EGF showed enhanced regulation of E2F1 compared to other treatments. AP1 complex (JUN and FOS) were regulated by both E2F1 and STAT3 in the OSM condition, leading to EMT. Transcriptional changes were validated using changes in chromatin accessibility obtained from the ATAC-seq data. Differentially accessible regions (DARs) were identified using Diffbind[ 61 ] in EGF and OSM at 24 and 48 hours with respect to PBS. Peaks occurring in the promoter region [−1000, +500 base pairs (bp) from TSS] were annotated using ChIPseeker[ 62 ]. These regions showed enrichment of genes associated with proteoglycans, focal adhesion, RAP signaling, and loss of apoptosis in EGF treatment, and RAP, RAS, and PI3K signaling in OSM treatment. To identify the regulator of these promoter regions, Motif enrichment analysis with GimmeMotifs[ 63 ] was performed with the Homer algorithm using the HOCOMOCOv11databbase[ 64 ]. In EGF treatment, the analysis revealed binding sites for cell cycle-related TFs and MAPK activator (ELK1),[ 65 , 66 ] and cell cycle modulators (E2F1, SP2, E2F2)[ 59 , 67 – 70 ] ( Fig 4a ). In OSM treatment, upregulation of differentially accessible regions showed enrichment of AP1 complex (FOSL1, FOSL2, JUNB, JUND, FOS, FOSB), cell cycle, and pro-inflammatory regulators (STAT3, NFYA) ( Fig. 4b ) These results further supported the role of E2F1 in EGF and AP1 in OSM which were consistent with the transcriptomic analysis. The cyclic-IF data across time following the treatments shows dynamic changes in localization of the proteins. To associate mechanisms inferred from the omics measurements with phenotypic outcomes, we categorized the cyclic-IF outcomes into EMT, RTK signaling, cell cycle, and inflammation-associated mechanisms. An increase in nuclear and plasma membrane localization of proteins associated with inflammation (STAT1, STAT3, PDL1, NFKB-P65), RTK signaling (MET, RPS6), and EMT proteins (VIM, KRT18, KI67, JUN) were observed in both EGF and OSM treatments ( Fig 4c ). The EMT progression was supported by the decrease of CDH1 in the plasma membrane and the nucleus and an increase of CTNNB1 in the nucleus in both treatments. The nuclear intensities of NDRG1-pT346 showed an increase in OSM in contrast to EGF treatment. A temporal increase in localization of cell cycle genes (CCND1, CDKN1A, HES1) was seen in EGF. In the OSM treatment, cyclic-IF data showed reduced presence of CCND1, CDKN1A consistent with RNA-seq measurements at 48 h.

Taken together, our analyses of the LINCS measurements provide an integrated multi-omics insight into initiation and progression of oncogenesis. All three growth regulators, EGF, HGF, and OSM, activated downstream pathways mTOR, ERK, and MAPK leading to increased pro-oncogenic processes, and EMT. The TF-target analysis identified the role of E2F1 mediated oncogenesis and EMT in all the ligands. In addition to the role of E2F1, in OSM treatment, STAT3 led AP1 activity led to a strong EMT signal. This suggests potential mechanisms for growth factor-driven oncogenesis. Targeting these mechanisms has the potential to interfere with pro-oncogenic mechanisms and alter pathological cell states. In vivo, several other cell-intrinsic growth factors and cytokines crosstalk participate in cell fate decisions. To explore the combinatorial effects of such stimuli, we study in a companion paper the effect of endogenous ligands namely TGFβ, BMP2, and IFNγ in combination with EGF in determining the pro- and anti-oncogenic cell fates of normal breast epithelial cells.

The transformation of normal epithelial cells into an oncogenic state is orchestrated by multiple cellular processes. The availability of multi-omics longitudinal measurements on a normal breast epithelial cell treated with distinct growth factors, in the LINCS study, offers the potential to identify key mechanisms. These mechanisms poise the epithelial cell for transition to a tumor, and through analyses of the LINCS measurements we identify the pathways that are pro-oncogenic and transformation-resistant and explore temporal causality of these mechanisms[ 71 , 72 ]. EGF is known to play a key role in promoting cell proliferation, differentiation, and survival[ 73 ], while HGF, which is largely similar in response, additionally regulates cell migration mechanisms[ 74 ]. OSM, another native epithelial cell factor, plays a key role in controlling proliferation while increasing cell motility[ 75 ]. Our analyses of LINCS MCF10a data broadly support these mechanisms while providing significant detailed insights into phenotypic changes that promote tumorigenic transformation of epithelial cells. All three treatments demonstrate changes in cell state that contribute to pro and anti-oncogenic cell fates. EGF, HGF and OSM, regulate mTOR, AKT and ERK pathways which promote oncogenic transformation in epithelial cells[ 76 – 78 ]. Notably the expression of oncogenes and oncoproteins such as BRCA1, BRCA2, CCND1, CCNE1, HER2, EGFR, c-KIT, MET, TP53, ATM, CHEK1, CHEK2, NF1, PDL1, have been reported to be elevated in breast cancer cells and associated with increased cell cycle, metabolism, DNA damage repair pathways and EMT[ 79 ]. Consistent with these findings, most breast cancer genes are upregulated in all three treatments, according to our study. It has been shown that mTOR signaling regulates proliferation via EIF4EBP1 and S6K, metabolism and immune response via AKT, NDRG1, PDK1 and PRKCA[ 80 ]. This regulation of downstream targets of mTOR is also seen with EGF and HGF and OSM treatments[ 81 – 84 ] leading to increased proliferation and EMT. In addition, OSM, as an analog of IL6, regulates cytokine signaling and shows the activation of immune response[ 85 ]. Thus, these growth regulators play a pivotal role during initial stages of cell states that lead to oncogenic transformation. Our analysis of the multi-omics measurements of response to ligand treatments provides unique mechanistic insights. We identified various pathways in which EGF regulated oncogenesis. In our analysis, E2F1 is seen to be regulated by EGF through NRF1, RB1, and mTOR signaling and leads to increase in both cell cycle and EMT. This is consistent with CDK-RB-E2F pathway that is critical for the control of cell proliferation, angiogenesis, metastasis, and tumor progression[ 86 ]. EGF is also shown to transactivate TGFβ receptor signaling that leads to EMT[ 87 ]. Further, E2F1 can lead to transcriptional activation of ZEB1 which causes cytoskeletal remodeling by binding to the proximal promoter region and recruiting HDAC1/2, while also downregulating CDH1. This relation between E2F1, ZEB1 and CDH1 is also observed in our data. In addition to E2F1 regulation through mTOR, OSM treatment also shows activation of JAK/STAT signaling leading to enhanced EMT. Increase in EMT is also supported by the temporal activation of STAT3-pY705 as seen in RPPA, increased TF activity in transcriptomic and nuclear localization in cyclic-IF measurements[ 88 – 90 ]. We identified two key downstream effects due to STAT3 signaling namely activation of AP1 complex and regulation of hypoxia via HIF1A. AP1 binds in the promoter region and leads an increase in FN1, VIM and CDH2[ 91 , 92 ] AP1 also regulates ZEB1 and SNAI1, which are known EMT TFs [ 93 , 94 ] Transcriptional regulation of HIF1A led to the regulation of EMT (PLAUR, CXCL8, FGF2) and angiogenesis (VEGFA, ITGB1, PTGS2). The relationship between HIF1A-induced hypoxia and OSM treatment can be attributed to metabolic changes and molecular changes consistent with EMT in epithelial cells.[ 95 , 96 ] EGF and OSM treatments showed distinct EMT responses. EGF treatment shows EMT is regulated by E2F1. It is accompanied by an increase in the expression of CTGF, NR2F1, CDH2, PMP22, CDKN2C, FGF2, and EMP3 showing an upregulation of the cell-cycle. Under OSM treatment, EMT is regulated by E2F1 and AP1. The downstream signaling indicates that STAT3 drives pro-inflammatory cytokine signaling. Further, extracellular matrix organization (VCAN, VIM, ADAM12, LIM, FBN1, LTBP2), inflammatory factors (TNFAIP6) TGFβ pathway (TFP1, NREP) and enzyme activity (SULF1, TGM2) are regulated in OSM treatment. Differences in the EMT signaling can be attributed to the differences in the signaling of TGFβ receptors[ 97 ]. It is shown in literature that EGF activates the TGFβ receptor and, OSM treatment activates the TGFβ via STAT3 which causes canonical TGFβ signaling via SMAD activation[ 98 ].

Breast cancer is heterogeneous in nature and is classified based on the expression of hormone receptors (HR) and HER2 including Luminal A (HER2−, HR+), luminal B (HER2+, HR+), HER2 overexpressing (HER2+, HR-), and triple-negative breast cancer (HER2-, HR-)[ 1 – 4 ]. Cellular heterogeneity arises from underlying cellular and molecular landscape, involving protein interactions, gene expression, and changes in the epigenetic landscape, leading to diverse tumor phenotypes[ 5 , 6 ], which include increased proliferation, angiogenesis, epithelial to mesenchymal transition (EMT), metastasis, cell plasticity, and loss of apoptosis. These processes help in tumor initiation, leading to changes in cell state, invasion, and proliferation. Integrative analyses of cellular signaling, transcriptomic, and epigenomic data provide the opportunity to identify mechanisms that induce and promote tumor pathology. Tumor growth and proliferation are accompanied by an increase in both autocrine and paracrine signaling of growth regulators[ 7 – 10 ]. In general, growth factors are responsible for proliferation, development, and homeostasis in normal breast cells. Aberrant expression of growth factor proteins and abnormal cellular response to these proteins lead to oncogenesis. Since the discovery of soluble growth factor EGF[ 11 ], studies have established that amplification or overexpression of endogenous growth factors such as EGF, HGF, TGFβ, VEGF, and IGF leads to cell proliferation, migration, angiogenesis, and metastasis[ 10 , 12 – 15 ]. While EGF is required for the growth and development of breast epithelial cells, overexpression of EGF leads to a sustained activation of several crucial members of the ERBB Receptor Tyrosine Kinases (RTKs) family. Sustained activation of these RTKs is known to mediate MAPK mediated tumorigenesis, c-MYC mediated anti-apoptotic signals, hyperactivation, and nuclear localization of kinases such as p-AKT and p-ERK1/2 which promote proliferation and EMT[ 16 , 17 ]. Human secreted cytokine Oncostatin-M (OSM), a growth regulator of the IL-6 superfamily, is also a potent inducer of cancer cell de-differentiation and stem cell plasticity[ 18 ]. OSM also causes the induction of transcription factors such as SNAI1, SNAI2, ZEB1, FN1, and CDH2 which are well-known markers of EMT in human breast cancer cells[ 16 ]. High throughput sequencing studies in loss/gain of function using siRNA, shRNA, and CRISPR have led to new avenues for understanding the genetic vulnerabilities and discovery of drug targets[ 17 , 19 , 20 ]. These advances have led to the identification and characterization of various oncogenes such as growth factor receptors (ERBB2/3, c-MET, OSMR), cell cycle modulators (CDKs, CCNDs, and others), kinases (AKT, PI3K), and transcription factors (c-MYC, c-FOS). For example, loss of function of ERBB2/3 blocks cell proliferation and cell cycle[ 21 , 22 ] and inhibition of cell cycle regulators such as CDKs (1–6) induces apoptosis[ 23 – 26 ]. Upregulation of PIK3CA leads to the formation of malignant tumors[ 27 ]. The oncogenic role of PIK3CA and CDK proteins are attributed to ZEB1 activation which controls EMT[ 28 ]. OSMR also signals via JAK/STAT pathway where STAT3 regulates cyclin D-1 and c-MYC to increase proliferation. STAT3 can also inhibit BAX/BCL-2-related caspase-dependent apoptosis[ 29 – 31 ]. Recent developments in high throughput technologies have aided in early diagnosis and development of novel therapeutic strategies[ 32 – 38 ]. In order to provide an integrated view of cellular mechanisms that lead to the development of tumor, the National Institutes of Health launched a multi-institutional study, called Library of Integrated Network-based Cellular Signatures (LINCS), which made multi-omics measurements on normal breast epithelial cells, MCF10a treated with tumor growth-regulating ligands namely EGF, HGF, and OSM[ 39 ]. Experimental measurements were taken longitudinally with reverse phase protein assay (RPPA), RNA-seq, ATAC-seq and cyclic-IF. These measurements provided the ability to decipher potential tumor activating and suppressing mechanisms through integrative analyses. In our analysis, integration of multi-modal data showed that EGF, HGF, and OSM treated cells activate mTOR, STAT3, and ERK pathways leading to enhanced cell cycle, inflammation, and EMT. These phenotypes are transcriptionally regulated by E2F1 in all three ligands and are further enhanced in OSM treated cells by the activation of the AP1 complex. Our study offers key insights into pathways and underlying mechanisms for oncogenesis in normal breast cells to help identify early-stage mechanisms and potential therapeutic targets. We present the results of these analyses in the following sections.

Supplementary Files This is a list of supplementary files associated with this preprint. Click to download. Supplementary.docx Supplementary.docx