Chromatin Organization and Transcriptional Programming of Breast Cancer Cell Identity

Abstract

The advent of sequencing technologies for assessing chromosome conformations has provided a wealth of information on the organization of the 3-dimensional genome and its role in cancer progression. It is now known that changes in chromatin folding and accessibility can promote aberrant activation or repression of transcriptional programs that can drive tumorigenesis and progression in diverse cancers. This includes breast cancer, which comprises several distinct subtypes defined by their unique transcriptomes that dictate treatment response and patient outcomes. Of these, basal-like breast cancer is an aggressive subtype controlled by a pluripotency-enforcing transcriptome. Meanwhile, the more differentiated luminal subtype of breast cancer is driven by an estrogen receptor-dominated transcriptome that underlies its responsiveness to antihormone therapies and conveys improved patient outcomes. Despite the clear differences in molecular signatures, the genesis of each subtype from normal mammary epithelial cells remains unclear. Recent technical advances have revealed key distinctions in chromatin folding and organization between subtypes that could underlie their transcriptomic and, hence, phenotypic differences. These studies also suggest that proteins controlling particular chromatin states may be useful targets for treating aggressive disease. In this review, we explore the current state of understanding of chromatin architecture in breast cancer subtypes and its potential role in defining their phenotypic characteristics.

Whole-genome sequencing studies have identified numerous genetic alterations, both germline and somatic, that increase breast cancer risk and are prognostic of patient outcomes. The most studied mutations alter protein-coding sequences that can inactivate tumor suppressor genes, such as BRCA1/2 and TP53, or amplify/overexpress oncogenes, such as MYC, HER2, and PIK3CA, among others (1). However, many genetic alterations are in non-protein-coding regions, positioned far from any annotated gene, making identification of driver mutations challenging (1, 2). Recent technological advances have begun to illuminate the complexity of the highly dynamic, 3-dimensional (3D) organization of DNA within the nucleus, revealing a higher order of gene regulation that complements the linear DNA code (3-5). The advent of chromosome conformation capture assays nearly 20 years ago (6) and the derivative technology, high-throughput chromosome conformation capture (Hi-C) (7), enabled the mapping of genome-wide, 3D chromatin contacts, leading to the detection of long-range intrachromosomal interactions that were previously undetectable (8). Additionally, the development of Assay for Transposase Accessible Chromatin using sequencing (ATAC-seq) permits capture of the interplay between chromatin condensation, DNA binding proteins, and transcriptional activity (9). The 3D organization of the genome has been directly linked to cell identity with orchestrated chromatin changes occurring throughout development, cell differentiation, and normal cell functioning (4). These new technologies have been leveraged further to demonstrate that alterations in DNA structure and organization play key roles in driving breast cancer heterogeneity, progression, and treatment response. These new findings may ultimately lead to the discovery of novel therapeutic targets, expanding treatment options (10, 11).

Breast cancer is a highly heterogeneous disease clinically divided into 4 subtypes based on the expression of HER2 and receptors for estrogen (ER) and progesterone (PR). The status of these receptors is prognostic of patient outcomes and informs treatment decisions. Breast cancer can also be divided into intrinsic molecular subtypes based on transcriptome analyses (12-14). Each subtype has a unique gene expression profile that dictates its biological features, such as metastatic potential and response to therapy, and is primarily used in the research setting to discover molecular drivers of disease. Luminal breast cancers (BCs) are the most differentiated cancers and they express ER and PR and convey relatively favorable patient outcomes. In contrast, basal-like breast cancers (BLBCs) lack ER, PR, and high levels of HER2 expression, are less differentiated, and are widely considered the most aggressive subtype. Patients with BLBC are at increased risk of recurrence and metastatic disease, and thus have worse outcomes than patients with luminal cancer. BLBCs are enriched in highly plastic stem- and progenitor-like cells that readily adapt to environmental cues and contribute to both the inter- and intratumoral heterogeneity characteristic of this subtype (15). Although subtype-specific transcriptomes have been well-characterized, events that establish these gene expression programs remain vague. Developing a deeper understanding of the mechanisms controlling the development of subtype expression profiles should provide new avenues for treating aggressive disease, overcoming therapeutic resistance, and improving patient outcomes. Several recent studies provide emerging evidence that chromatin organization drives subtype-specific transcriptomes of breast cancer (11, 16-18). Here, we provide an overview of the 3D genomic architecture in normal breast epithelium, discuss the processes that establish breast cancer subtypes and confer therapeutic resistance, and outline investigational therapies that target the epigenome and their potential impact on breast cancer outcomes.

Overview of Chromatin Structure and Normal Cell Development

Three-dimensional Organization of the Genome

From the whole chromosome level to individual DNA loops, chromatin dynamics are tightly controlled and promote transcriptional programs that define cell identity (Fig. 1). On a broad scale, chromatin is divided into A and B compartments. A compartments contain open, transcriptionally active euchromatin, and B compartments are composed of condensed, or closed, and inactive heterochromatin (7). Several other structural elements are often coincident with, but distinct from, these compartmental designations. First, each chromosome in the nucleus resides in a unique and nonrandom position, termed “chromosome territories.” Generally, gene-rich, transcriptionally active regions of chromosomes reside near the nuclear interior, whereas gene-poor, transcriptionally inactive chromosome segments reside near the nuclear periphery (19, 20). Chromatin tethering points, including those at the transcriptionally repressive nuclear lamina, can influence gene expression (21). By residing in a defined position, chromosome territories promote cis-interactions, or interactions within a chromosome, which occur much more often than trans-interactions, or those between 2 chromosomes (7, 22, 23). Cis-interactions are further governed by formation of topologically associated domains (TADs). These are self-interacting genomic regions that define points of contact between regulatory elements. TAD boundaries are demarcated by occupancy of the insulator protein, CCCTC-binding factor (CTCF), as well as cohesin (10, 24-27). Current studies support the hypothesis that cohesin establishes chromatin loops, allowing distant enhancer and promoter regions to come into close proximity and regulate gene expression. Meanwhile, CTCF specifies genomic sites of cohesin function. Consequently, CTCF prevents regulatory elements from interacting with distal DNA that resides outside of the TAD (28-35). These levels of chromatin organization are established during embryogenesis in pluripotent cells, with changes being tightly choreographed throughout cell differentiation (36).

Figure 1.

Three-dimensional organization of the genome. Chromosomes reside in distinct regions in the nucleus, termed “chromosome territories.” Chromatin is also divided into A and B compartments. Transcriptionally inactive B chromatin typically resides near the nuclear lamina at regions termed “lamin-associated domains.” Transcriptionally active A compartments often cluster in the nuclear interior. Chromatin is further organized into topologically associated domains (TADS), which are demarcated by the insulator protein CTCF and the cohesin complex. Created with BioRender.com.

Chromatin Structure is Dynamic and Controls Cell Fates

Chromatin accessibility controls cell identity by enabling the expression of transcriptional programs critical for various stages of development (37, 38). In general, pluripotent cells have widespread open chromatin accessible to transcriptional machinery and other regulatory proteins (37, 39, 40). These regions are typically characterized by increased histone acetylation, an epigenetic mark commonly associated with enhanced transcription (41, 42). Open/decondensed chromatin is highly dynamic and contributes to the markedly plastic nature of pluripotent stem cells. By beginning from an open chromatin state, multiple avenues are available for stem cells to differentiate along various lineages in response to developmental cues that are independent of genetic changes (37). As differentiation pathways ensue, lineage-specific genes can acquire both trimethylation of lysine 27 on histone H3, a mark of repressed transcription, and trimethylation of lysine 4 on histone H3, an indicator of active transcription. These bivalent domains are thought to keep genes silent during pluripotency but poised for rapid induction on receiving differentiation signals (43). In contrast, differentiated cells characteristically have more abundant regions of condensed chromatin and reduced bivalent domains, thus restricting expression to the specific genes required for a specialized cell type (41, 43, 44).

The maintenance of pluripotency and prevention of terminal differentiation involves an extensive network of transcription factors and chromatin remodelers (41, 45, 46). Several factors, including the Yamanaka pluripotency factors OCT4, SOX2, and KLF4, have pioneering functions, meaning they bind closed chromatin, initiate its relaxation, and recruit additional transcription factors (47-52). In embryonic stem cells, OCT4 depletion decreases chromatin accessibility at its resident enhancers in as little as 30 minutes, highlighting the high dynamicity of chromatin during pluripotency. Moreover, decreased chromatin accessibility and suppression of active enhancer landscapes result in rapid repression of pluripotency gene expression (53). In contrast, at other genomic regions, OCT4 depletion increases chromatin accessibility and promotes gene expression, indicating that pioneer factors also repress expression of genes that promote differentiation (53, 54). Ultimately, declining expression of pluripotency-related pioneer factors provides an avenue toward cell differentiation, wherein a different set of transcription factors and cofactors become expressed. These, in turn, repress pluripotency-related genes and activate cell-specific genes (45).

Changes in chromatin accessibility during differentiation are also accompanied by alterations in structural elements including nuclear territories, TADs, and DNA loops. During differentiation, cell-specific, actively transcribed genes coalesce toward the nuclear interior. Meanwhile, pluripotency-related genes move toward the nuclear periphery and interact with the transcriptionally repressive nuclear lamina (55). It is not yet clear whether interactions with the nuclear lamina are the cause or consequence of transcriptional changes. Moreover, a positive feedback loop may be at play, wherein reduced transcription promotes movement to the lamina, and this interaction in turn strengthens transcriptional repression (55). Differentiating cells also gain long-range, CTCF-anchored chromatin loops, suggesting greater segregation between neighboring genomic regions (56). Further, cis-interactions within TADs change dramatically during differentiation. This affords a high degree of plasticity between A and B compartments within certain TADs and likely underlies transcriptional changes in response to developmental signals (57). Together, these studies demonstrate that TAD organization, CTCF expression, and cohesin activity are all necessary for accurate differentiation and establishing cell identity (58-61).

Although several details have been elucidated regarding the stepwise manner by which transcriptional modulators dictate cell identities via chromatin relaxation and condensation, many questions remain regarding the role of the broader 3D genome in these processes. Early studies suggest that pioneer factors may establish TAD boundaries by preparing chromatin for insulator proteins, such as CTCF, to bind DNA (36, 62), or by forming phase-separated condensates that alter TAD organization (63). Although differentiation is associated with a gain of CTCF-anchored loops, overexpression of OCT4 or KLF4 erases these loops, weakening contact domain boundaries and decreasing regulatory element insulation (56). These studies suggest that pioneer factors play a central role in TAD organization during differentiation. They also suggest that dysregulated chromatin organization may endow malignant cells with features of stem/progenitor cells that promote disease aggressiveness.

Chromatin Organization Controls Cell Identity in the Normal Mammary Gland and in Breast Cancer

Chromatin Remodeling Influences Cell Identity in the Normal Mammary Gland

Although many studies aimed at discerning chromatin organization and pluripotency have used embryonic stem cells, studies in mammary epithelial cells mirror these findings. Through transplantation and lineage tracing experiments, the cell lineage of the normal mammary gland has largely been elucidated. A fetal mammary stem cell (fMaSC) population initiates gland morphogenesis and generates cells of the mature gland (64). fMaSCs initially give rise to bipotent progenitor cells, which then produce luminal and myoepithelial progenitors. These further differentiate into mature luminal and myoepithelial cells, respectively (64). Each of these cell types has unique chromatin signatures that drives their lineage-specific transcriptomes (65, 66). fMaSCs largely have an open chromatin landscape reflective of embryonic stem cells and express embryo-associated transcripts (66, 67). However, integration of RNA-seq and ATAC-seq data revealed that distal regulatory regions of genes that define mature mammary epithelial cells also reside in open chromatin in fMaSCs, with these genes being expressed at intermediate levels (66). Single-nucleus ATAC-seq further revealed distinct clusters of fMaSCs, with chromatin features evident of each adult mammary epithelial cell type (67). These data suggest that these cellular subsets are poised toward a luminal or basal fate, again promoting rapid differentiation in response to developmental cues (67). Furthermore, mammary progenitor-like cells display epigenetic features of open chromatin, including bivalent methylation domains, whereas mature epithelial cells have marks of differentiated cells (68). Together, these findings indicate that mammary stem cells harbor an open chromatin landscape and exist in a plastic, multilineage state reflective of precursor and differentiated lineages (66). Moreover, as the lineage progresses, cells establish distinct structural and epigenetic features of their particular developmental stage.

Differences in 3D Genome Organization are Associated With Distinct Breast Cancer Subtypes

Studies aimed at understanding genome organization in normal cells have improved our comprehension of its role in disease. The spatial organization of chromatin is altered in cancer in general (69-71), and mutations in chromatin organizing proteins such as CTCF are associated with multiple cancer types (72-74). TADs are also disorganized in cancer (75, 76). A 2019 study that created a consensus TAD map for the human genome revealed that a subset of TADs were prognostic for patient outcomes in cancer, further supporting the impact of chromatin topology on cancer progression (77). Although it is currently unclear how alterations in chromatin structure stimulate tumorigenesis, many studies have pointed to aberrant activation of transcriptional programs that promote dedifferentiation and cell plasticity (78, 79).

There is considerable overlap in gene expression signatures of specific breast cancer subtypes and stages of mammary gland development, suggesting that breast cancer subtypes may be closely coupled to their developmental cell of origin or that transformation events induce some level of mimicry of differentiation stages (12, 80). Luminal BCs have gene expression profiles comparable to mature epithelial cells that line the mammary lumen. In contrast, BLBCs are enriched in cells that display stem- and progenitor-like properties (12, 15, 81, 82). The relative abundance of more primitive cells in this subtype is thought to underlie its aggressive behaviors, including migration, invasion, and therapeutic resistance (83). Despite vastly different transcriptional profiles and phenotypes of breast cancer subtypes, it is still unclear how these form. Elucidating core pathway(s) involved should provide new therapeutic avenues by shifting basal tumor cell identity to a more treatable state.

Several studies have demonstrated that breast cancer subtype differentiation can occur independently of a cell of origin and, instead, can be controlled through genetic and epigenetic events, including alterations in chromatin structure (11, 66, 84-86). For example, mutations in BRCA1 are associated with BLBC, and it was originally thought that tumors driven by BRCA1 mutations originated from myoepithelial cells or myoepithelial progenitor cells (80, 85). However, inducing BRCA1 mutations in luminal progenitor cells phenocopies the human BLBC phenotype, whereas similar changes in myoepithelial progenitor cells do not (85). These data indicate that luminal progenitor cells are the cells of origin for BLBC (85, 87). It was later shown that wild-type BRCA1 drives mammary epithelial cell fate by stimulating enhancer-promoter looping and establishing superenhancers that drive expression of GATA family transcription factors (88). GATA3 is a pioneer transcription factor that promotes luminal BC differentiation and prevents BLBC cell proliferation (88). These findings exemplify how chromatin organization influences the expression of lineage-specific transcription factors and dictates subtype differentiation. Beyond BRCA1, additional enhancer and superenhancer landscapes stimulate basal-like breast cancer phenotypes (89). The transcription factor FOXC1 stimulates expression of BLBC-specific genes. It, in turn, is controlled by triple-negative breast cancer (TNBC)-specific superenhancers (90, 91). Although BLBC cells sustain FOXC1, its expression is blocked in luminal BC cells. This is likely due to a strong CTCF peak upstream of the FOXC1 promoter in luminal BC cells, which may insulate this region and suppress its activity. Consequently, expression of BLBC genes is repressed as a result of chromatin changes at other genomic sites (90). This serves as just one example of how TADs could promote different subtypes or cell identities in breast cancer.

On a broader scale, widespread differences in TAD organization and chromatin looping have been reported for breast cancer subtypes and their normal counterparts (11) (Fig. 2). Compared with normal mammary epithelial cells, BLBC cells have the most altered 3D chromatin organization compared with other subtypes as assessed by Hi-C (11). BLBC cells have fewer TAD boundaries than other subtypes, indicative of disorganized chromatin that is associated with transcriptional dysregulation. Importantly, luminal BC cells also have altered 3D genomes, but they are more similar to normal cells than BLBC cells (11). Although it is likely that these alterations contribute to subtype differences, more studies are needed to parse their precise impact as well as the mechanisms underlying differences in chromatin structure.

Figure 2.

Chromatin organization in breast cancer subtypes. Clinically, breast cancer is defined by receptor status. Triple-negative breast cancer (TNBC) lacks the amplification or overexpression of human epidermal growth factor receptor 2 (HER2) and the hormone receptors, estrogen receptor and progesterone receptor. Breast cancer is also divided into subtypes based on gene expression profiling. Basal-like breast cancer (BLBC), the major TNBC subtype, has a gene expression profile that is reflective of progenitor cells. In contrast, luminal breast cancer coincides with hormone receptor-positive disease and has a gene expression profile that is reminiscent of differentiated mammary epithelial cells. Compared with luminal breast cancer cells, basal-like breast cancer cells tend to have less organized chromatin with decreased expression of the insulator protein CTCF, disrupted topologically associating domains, and increased intra-chromosomal interactions. Created with BioRender.com.

Estrogen Receptor Signaling and Chromatin Structure

Luminal BC cells rely on an intimate connection between estrogen signaling and chromatin organization. Two estrogen receptors, ERα and ERβ, function as transcription factors following hormone ligand binding. ERα has more of an impact on breast cancer biology and hence is better understood. Several drugs targeting ERα are used to treat ER⁺ breast cancer, with tamoxifen being the oldest and most commonly prescribed. These therapies have vastly extended the lives of patients and frequently have led to durable cures. However, resistance is common (92). To assess genome-wide chromatin interactions in vehicle versus estradiol-stimulated luminal BC cells, Zhou et al performed tethered chromatin capture, a variation of Hi-C (93). This approach revealed that short-term estradiol exposure drives chromatin recompartmentalization with shifts between A and B compartments, indicative of changes in actively transcribed chromatin. Furthermore, Hi-C studies confirmed that estradiol drives widespread chromosome conformational changes including a gain in intra- and interchromosomal interactions (94). Differentially interacting loci were enriched for ERα binding and correlated with transcriptional changes, suggesting that estrogen signaling extensively alters chromosomal interactions to direct a specific transcriptional program (94-96). To further investigate the role of ER in human disease, Yang et al performed whole-genome sequencing and ERα chromatin immunoprecipitation sequencing in primary breast tumors (97). They found high rates of noncoding somatic mutations at ERα-binding sites compared with other genomic regions. Hi-C and chromatin interaction analysis with paired-end tag sequencing (98) data from luminal BC cell lines revealed increased frequency of chromatin loops at these mutated regions, as well as increased expression of closely associated target genes (97). Together, these data indicate that ERα binding sites regulate genes through long-range chromatin interactions and, most strikingly, mutations at these sites can affect phenotypes in ER⁺ breast cancer by modulating its transcriptional targets (97).

Estrogen receptor action depends on chromatin accessibility. The pioneering factors FOXA1 and GATA3 groom the chromatin landscape to enable binding of ligand-bound ERα and activation of ER-transcriptional programs (99, 100). Approximately 85% of ERα binding sites in the luminal breast cancer genome bind FOXA1, GATA3, or the cofactor NR2F2, and 25% bind all 3 (101). Depletion of these factors drives a redistribution of ERα binding by condensing chromatin at those regions, reducing ER accessibility (100, 101) and shifting the expressed genome. The insulator protein CTCF further dictates the ability of ERα to bind chromatin and alter target gene expression. Depending on genomic location, CTCF binding can be either activating or inhibitory (102-105). A CRISPR-Cas9 screen identified ERα-target genes that depend on CTCF chromatin binding and its enforcement of enhancer-promoter interactions. This revealed genes required for ERα-driven proliferation that were also dependent on CTCF (104). In contrast, CTCF can also reduce ERα chromatin association by either blocking FOXA1 binding or inhibiting DNA looping between ERα-bound enhancers and promoters at specific loci (103, 106). Fiorito et al investigated a well-characterized ERα loop at the P2RY2 locus and found that a chromatin site flanking this locus was bound by CTCF and interacted with the transcriptionally repressive nuclear lamina. Following CTCF depletion, this locus lost contact with the nuclear lamina, resulting in ERα binding and formation of an ERα-mediated chromatin loop between these sites (103). Although this one example should be expanded to genome-wide analyses, these findings point to a suppressive mechanism wherein CTCF blocks ER-induced looping via nuclear lamina contacts. Following depletion of CTCF, there was also increased expression of a set of ERα target genes, with an enrichment of genes involved in cell cycle and an increase in cell proliferation. Together, these studies indicate that ER function and chromatin organization are dynamic and interconnected, with estradiol-bound ERα leading to changes in the 3D chromatin landscape that ultimately defines the ERα-regulated transcriptome.

Breast Cancer Cell Plasticity

The updated Hallmarks of Cancer implicates “unlocking phenotypic plasticity” as an emerging characteristic of malignancy (107). Cancer cells can block differentiation, dedifferentiate to a progenitor-like state, and/or transdifferentiate to a new cell identity to promote growth and survival (107-110). Both normal and malignant progenitor cells can adapt to extracellular signaling, shift between cell identities, and acquire more migratory and invasive characteristics (111, 112). This plasticity has been linked to epigenetic reprogramming, histone modifications, and shifts in chromatin topology that maintain open-chromatin states. Decondensed regions then promote activation of transcriptional programs that establish normal tissue architecture during development or intratumoral heterogeneity in the context of cancer (107).

Several factors have been identified that govern plasticity and cell identity in normal and breast cancer cells. More than a decade ago, Weinberg and colleagues reported that transcription factors SOX9 and SLUG regulate mammary stem cell fate and that another transcription factor, SOX10, is a key downstream effector of SOX9 (113). More recently, Dravis et al coupled epigenomic and transcriptomic analyses to discover factors important for mammary epithelial stemness that could be coopted in breast cancer and again identified SOX10 (66). SOX10 is required for progenitor-like activity in normal mammary epithelial cells, including fMaSCs and luminal progenitor cells (66). SOX10 is also more highly expressed in BLBC than other subtypes (114). Moreover, it is sufficient to induce stem-like and mesenchymal phenotypes in normal fetal mammary epithelial cells (115) and has recently been shown to control plasticity in breast cancers with low expression of ER (116). These data indicate that SOX10 controls cell fates both during mammary morphogenesis as well as the initiation and progression of BLBC (66). Although SOX10 has been extensively studied in mammary cells, several other pluripotency factors have been identified that play similar roles (117-119). Collectively, these results indicate that factors driving pluripotency in normal epithelium may be exploited by breast cancer cells to stimulate plasticity and heterogeneity.

Transcriptional modulators have also been identified that ensure luminal BC cell identity and prevent acquisition of less-differentiated BLBC cell characteristics. Depletion of ARID1A, a member of the SWI/SNF chromatin remodeling complex, causes a switch from luminal identify to a more basal-like phenotype through changes in chromatin accessibility and transcriptional reprogramming (16). As indicated previously, luminal BC cells rely heavily on ERα function to sustain their identity, and ARID1A maintains chromatin accessibility to promote ERα binding and target gene expression. Hence, ARID1A disruption causes cells to lose their luminal gene expression patterns and enriches for BLBC and stemness gene signatures. Likewise, JARID1B, a histone modifying enzyme, regulates breast cancer cell identity by controlling the expression of subtype-specific breast cancer genes (17). JARID1B is commonly overexpressed in luminal BC, and its loss reduces the expression of genes specific for luminal BC with the simultaneous induction of BLBC and stemness genes (17). More recently, we found that TLE3, a transcriptional corepressor, also sustains luminal gene expression signatures while preventing the expression of basal genes (120). Together, these results and many others have revealed the highly dynamic nature of the breast epithelial cell lineage that is controlled through chromatin modulation and subsequent transcriptional changes. This plasticity may also underlie the heterogeneity in breast cancers that has been observed with single-cell sequencing, revealing that most tumors include intermixed cell populations of multiple subtypes, with one subtype dominating the overall transcriptome of the tumor (121).

The adaptability that cellular plasticity affords is characteristic of aggressive cancers (107, 122). Supporting this, the highly plastic nature of BLBCs is associated with therapy resistance, relapse, and metastatic disease (111, 123, 124). ATAC-seq analyses of TNBC cell-line subpopulations capable of metastasizing to the lung and brain revealed that these cells gain an open chromatin signature that is associated with an increase in active promoter and enhancer regions compared with the parental TNBC cells. This ultimately provides cells with metastatic functionality because it leads to the increased expression of metastasis-related genes (125). BLBC cells also express markers of epithelial-mesenchymal transition (EMT) (126), a developmental process commonly used during tissue organization, regeneration, and wound repair whereby epithelial cells reversibly shift toward a migratory, more mesenchymal phenotype. Cancer cells undergoing EMT also acquire mesenchymal properties and become more prone to metastatic colonization (127). The process of EMT is reversible and largely epigenetically controlled. The pleomorphic cytokine, TGF-β, is one of the most potent inducers of EMT. Johnson et al quantified the epigenetic and transcriptomic changes in immortalized breast epithelial cells during extended treatment and subsequent withdrawal of TGF-β using integrated ATAC-seq and RNA-seq. These studies revealed that EMT was associated with increased chromatin accessibility and expression of mesenchymal genes that is largely reversible following TGF-β withdrawal, allowing restoration of epithelial characteristics through the process of mesenchymal-epithelial transition (128). These data indicate that chromatin condensation is a dynamic process that exposes or conceals genomic regulatory elements to elicit transient phenotypic changes (128). Transient, TGF-β-driven increases in chromatin accessibility were also associated with decreased CTCF protein expression, suggesting that TAD organization may be essential to maintain epithelial differentiation. Supporting this postulate, mesenchymal-like breast cancer cell lines have lower CTCF expression than epithelial-like cell lines (128, 129). The loss of CTCF promotes cell invasion and enforced overexpression in mesenchymal breast cancer cell lines suppresses migration (129, 130). Moreover, we conducted an analysis of publicly available data from The Cancer Genome Atlas and found that BLBCs and TNBCs have lower CTCF expression than other breast cancer subtypes and that low CTCF expression is associated with shorter overall survival (Fig. 3). As a whole, these data suggest that CTCF expression promotes epithelial differentiation and that its loss disrupts TAD organization to enable transcriptomic shifts that define aggressive breast cancer phenotypes. Hence, the open chromatin signature of BLBC likely underlies its plasticity and metastatic ability.

Figure 3.

CTCF expression in basal breast cancer. (A) RNA-seq data from The Cancer Genome Atlas (TCGA) of 623 patients with basal breast cancer was queried for CTCF mRNA expression. Violin plot of CTCF expression in non-basal-like and non-TNBC samples compared with basal-like and TNBC samples. P = .0063. (B) Kaplan-Meier survival curve from KM Plotter showing probability of overall survival based on high or low CTCF expression for 431 patients with basal-like breast cancer.

Chromatin Topology and Response to Hormone Therapy

It is increasingly evident that chromatin organization is important in ER signaling, leading to the likely scenario that chromatin dynamics also plays a role in the response to endocrine therapy. Several mechanisms of endocrine resistance have been identified, including loss of ERα, mutations of ESR1, or altered expression of ERα coregulatory proteins that lead to upregulated growth and survival pathways (95, 131-134). More recently, evidence also suggests a role for chromatin topology in endocrine resistance (93, 97, 106).

Tamoxifen is a selective ER modulator (SERM) that functions through a dual mechanism. It binds to ERα and competes for estradiol binding. Tamoxifen-bound ERα also interacts with DNA and represses the transcription of many, but not all, ERα target genes (106). Identifying mechanisms of tamoxifen resistance could ultimately have a significant impact on outcomes of patients with luminal BC. Notably, tamoxifen-resistant cells have altered chromatin topology compared with treatment-naïve cells, with highly dynamic chromatin regions being more altered in tamoxifen-resistant cells compared with their parental counterparts (93). Further, these regions are more likely to bind ERα and less likely to bind CTCF at nearby chromatin boundaries. Within these dynamic chromatin regions, ERα binds to promoter-enhancer loops that are particularly enriched at genes related to cancer invasion, aggressiveness, metabolism, and glycolysis (93). These data confirm that ERα associates with dynamic chromatin regions, promoting their reorganization and altered chromatin looping. These interactions are then shifted when cells develop resistance to tamoxifen. It has been speculated that tamoxifen-bound ERα loosens CTCF binding, resulting in a relaxation of chromatin boundaries (93). These findings illustrate the complex relationship between ERα function and chromatin reorganization and how these factors together affect responsiveness to endocrine-based therapies. Future analyses should focus on whether restoring CTCF binding can reverse tamoxifen resistance, or if this is a unidirectional program, requiring targeting the downstream consequences of chromatin reorganization.

Additional studies have revealed other factors beyond ERα that regulate chromatin topology and promote endocrine resistance. For example, FOXA1 overexpression is associated with endocrine resistance (106). Formaldehyde-assisted isolation of regulatory elements sequencing (FAIRE-seq) revealed that FOXA1 is essential for maintaining an open chromatin landscape that allows ERα to bind its target sites in tamoxifen-resistant breast cancer cells (106). When FOXA1 is overexpressed, ERα chromatin binding occurs independently of estrogen ligands (106). Likewise, mutations in ARID1A are associated with endocrine resistance (16). ARID1A inactivation in ER⁺ breast cancer cells decreases chromatin accessibility at ER-binding sites that are typically necessary to maintain luminal cell fate. As mentioned previously, this further facilitates switching from luminal to basal cell identity, which is associated with endocrine resistance (16).

ERα coregulatory proteins, which alter chromatin accessibility at ERα-target gene promoters, also play important roles in modulating responsiveness to endocrine therapies. High expression of the p160 family of steroid receptor coactivators is associated with endocrine resistance (135-139). These proteins contain functional domains that interact with histone acetyltransferases (HATs) and protein arginine methyltransferases (PRMTs). Histone acetyltransferases and protein arginine methyltransferases promote epigenetic modifications that open chromatin, allowing transcriptional machinery to bind ERα target genes and conveying an ERα-driven phenotype (140-142). Although tamoxifen blocks ERα signaling, increased expression of p160 proteins independently promotes ERα-target gene transcription, resulting in drug resistance. Notably, suppressing the expression p160 family members restores sensitivity to tamoxifen treatment, suggesting that these proteins may serve as viable therapeutic targets to overcome resistance (139). Although increased expression of ERα coactivators can promote tamoxifen resistance, decreased expression of ERα corepressors can have the same effect. Normally, the nuclear hormone receptor corepressor (NCOR) family of proteins recruit histone deacetylases to promote chromatin condensation, and it has been hypothesized that this impedes ERα from binding promoter regulatory regions (143, 144). Loss of NCOR expression promotes ERα-target gene transcription, tamoxifen resistance, and is associated with worse patient outcomes (145, 146). Several other corepressors have similar properties, suggesting that this is a widespread mechanism of resistance (144). Together, these studies demonstrate that ERα requires chromatin modifying enzymes to mediate its transcriptional program. In response to endocrine therapy, a subpopulation of cells can activate signaling pathways that result in an open chromatin landscape. These events ultimately allow ERα to function independently of ligand and promote drug resistance.

Conclusions

Breast cancer subtypes are associated with distinct phenotypes that drive differences in patient outcomes. Despite their importance, mechanisms underlying subtype differentiation are only recently coming into light. The studies reviewed here support the longstanding hypothesis that it should be possible to shift BLBC cells to a less aggressive and more treatable luminal-like cell identity. Several transcriptional modulators have been shown to drive luminal and suppress basal-signature genes and phenotypes. As a result, identifying therapeutic agents that can enforce differentiated cancer cell phenotypes by targeting chromatin architecture should provide a major advance in the treatment of this disease. Of note, epigenome-targeting drugs have shown efficacy in preclinical models of TNBC (147-150), providing a rationale for further investigating their mechanisms of action and the extent to which they drive differentiation away from the BLBC state.

Although luminal BC is typically associated with better patient outcomes, resistance to endocrine therapies can also occur. In this case, the chromatin landscape also appears to be a major driver, and chromatin-modifying proteins have been suggested as therapeutic targets in this disease as well. In this regard, histone deacetylase inhibitors have been US Food and Drug Administration-approved for hematological malignancies, and several trials are currently assessing the efficacy of these drugs in solid tumors, including breast. Additionally, histone acetyltransferase inhibitors or activators are also potential targets for breast cancer treatment (151), as are pioneer transcription factors (152). Importantly, the impact of chromatin biology on therapeutic response is not limited to endocrine therapies where the major mediator is ER, a transcription factor. Indeed, a few early studies have suggested that this is also the case for resistance to taxanes (153) and trastuzumab (154), drugs that are commonly used to treat TNBC and HER2⁺ breast cancers, respectively.

In summary, the chromatin landscape and its modulation are major drivers of tumor cell fate, plasticity, and drug response in breast cancer. This can dictate breast cancer subtype as well as the extent of cellular heterogeneity within tumors. Future studies must focus on whether it is feasible to restore chromatin configurations to those associated with terminally differentiated cells. If so, this would provide a major area of focus for developing novel drugs that can enforce specific landscapes to ultimately improve patient outcomes.

Abbreviations

3D

3-dimensional

ATAC-seq

Assay for Transposase Accessible Chromatin using sequencing

BC

breast cancer

BLBC

basal-like breast cancer

CTCF

CCCTC-binding factor

EMT

epithelial-mesenchymal transition

ER

estrogen receptor

fMaSC

fetal mammary stem cell

Hi-C

high-throughput chromosome conformation capture

NCOR

nuclear hormone receptor corepressor

PR

progesterone receptor

TAD

topologically associated domain

TNBC

triple-negative breast cancer

Funding

This work was supported by National Institutes of Health grants R01CA257502, R01CA213843, and R01CA206505 (R.A.K.).

Disclosures

The authors have nothing to disclose.

Data Availability

Original data generated and analyzed during this study are included in this published article or in the data repositories listed in References.