lncRNA BORG:TRIM28 Complexes Drive Metastatic Progression by Inducing α6 Integrin/CD49f Expression in Breast Cancer Stem Cells Open Access

Despite recent advancements in the prevention, diagnosis, and treatment of breast cancers, this disease remains a momentous health burden that accounts for a substantial proportion of cancer-associated morbidity and mortality in the United States (1). Moreover, the efficacy of conventional clinical paradigms continues to be hampered by the complex molecular heterogeneity exhibited by breast cancers, a feature reflecting the existence of multiple genetically distinct breast cancer subtypes that possess unique clinical prognoses and treatment strategies (2). Of these genetically distinct breast cancer subtypes, those known as “triple-negative” (TNBC) represent the most aggressive and lethal form of the disease; they also lack effective FDA-approved targeted therapies largely due to their failure to express hormone receptors (estrogen receptor-α and progesterone) and ErbB2/HER2 (3, 4). Despite the pathophysiologic diversity amongst breast cancers, their underlying tumorigenicity is nonetheless linked by their reliance upon breast cancer stem cells (BCSC), which are defined by their requisite capacity to: (i) initiate tumor formation; (ii) undergo self-renewal and expansion; and (iii) give rise to the heterogeneous bulk populations within developing and progressing tumors (5). Moreover, BCSCs can undergo metastatic dissemination and lethal disease recurrence, and are unique in their ability to overcome and survive a variety of cellular stressors, including exposure to chemotherapies, foreign microenvironments, and immunosurveillance (6–8). As such, enhancing our understanding of the molecular underpinnings of BCSCs represents a fundamental step in the development of targeted therapeutics capable of eradicating this resilient population of malignant cells.

Long noncoding RNAs (lncRNA) are a class of heterogeneous RNA transcripts that are >200 nucleotides in length and carry out a myriad of normal cellular and organismal functions despite their inability to be translated into functional proteins (9, 10). More recently, lncRNAs have also emerged as powerful regulators of the initiation and progression of nearly all cancers, including those of the breast (11–13). Analogous to their protein-coding counterparts, alterations in lncRNA sequence or expression readily promote aberrant carcinoma cell proliferation and dissemination (14, 15); they also enhance CSCs self-renewal, survival, and resistance to chemotherapy (9, 16, 17). Along these lines, we identified the lncRNA BMP/OP-Responsive Gene (BORG) as a novel driver underlying the aggressiveness and lethality of TNBCs, doing so through the ability of BORG to bind: (i) the epigenetic remodeler TRIM28, resulting in locus-specific chromatin binding that initiates metastatic relapse (18), and (ii) the single-strand DNA-binding protein RPA1, leading to the acquisition of prosurvival and chemoresistant phenotypes (19). Mechanistically, BORG:TRIM28 complexes promote TNBC proliferation by repressing the expression of cell-cycle inhibitors Cdkn1a and Gadd45a, an event requiring nucleotides 1042–2040 of BORG to bind TRIM28 (18). Independent of the actions of TRIM28, BORG:RPA1 complexes elicit chemoresistance and survival signaling through activation of NFκB (19). It is noteworthy that the functional consequences of aberrant BORG expression are reminiscent of the “Hallmarks” associated with BCSCs (6, 20); however, whether BORG underlies the acquisition of BCSC phenotypes in TNBCs remains unknown. Thus, the objectives of this study were to: (i) determine whether aberrant BORG expression drives the self-renewal and expansion of BCSCs; and (ii) establish the relative importance of BORG:TRIM28 complexes in promoting BCSC phenotypes.

Murine D2.HAN (D2.OR, RRID:CVCL_0I88; and D2.A1, RRID:CVCL_0I90) cells were obtained from Fred Miller (Wayne State University, Detroit, MI), while human MDA-MB-231 cells were purchased from ATCC (RRID:CVCL_0062). Normal human mammary epithelial cells (HMEC) were provided by the laboratory of Mark W. Jackson (Case Western Reserve University, Cleveland, OH). All cell lines were propagated in DMEM (Sigma-Aldrich) supplemented with 10% FBS and 1% Pen/Strep. D2.HAN and MDA-MB-231 cells were engineered to stably express firefly luciferase by transfection with pNifty-CMV-luciferase, followed by zeocin selection (500 μg/mL; Invitrogen) as described previously (18). D2.OR cells engineered to express either full-length BORG and its internal deletion mutants, Del1 and Del5, were created and characterized as described previously (18, 21), as were TRIM28-deficient D2.OR cells and BORG-deficient D2.A1 cells (18). Similarly, rendering human MDA-MB-231 cells deficient in BORG expression or D2.OR and D2.A1 derivatives deficient in ITGA6 expression was achieved by transduction with control short hairpin RNA (shRNA; i.e., nontargeting pLKO.1 vector) lentiviral particles or those specific against BORG or ITGA6, followed by selection with puromycin (5 μg/mL). shRNA vector sequences are provided in Supplementary Table S1.

D2.OR cells were authenticated using short tandem repeat analysis (ATCC), and all cell lines were regularly subjected to Mycoplasma testing using MycoAlert Mycoplasma Detection Kit (Lonza, catalog no. LT07-218), with the most recent test performed in August, 2021 with negative results. All cell lines used in the described experiments were passaged fewer than 50 times, with the experiments listed using cells passaged fewer than 20 times before data acquisition.

Aldehyde dehydrogenase (ALDH) activity was quantified in D2.OR derivatives using the ALDEFLUOR Kit (STEMCELL Technologies) per manufacturer's instructions. Briefly, ALDH substrate was resuspended in DMSO and activated by 2N HCl treatment prior to its dilution in ALDEFLUOR assay buffer. Single-cell suspensions of D2.OR derivatives (10⁵ cells/mL) were incubated for 30 minutes at 37°C in ALDEFLUOR assay buffer supplemented with ALDH substrate in the absence or presence of the ALDH inhibitor reagent, N,N-diethylaminobenzaldehyde (DEAB). The reactions were centrifuged and the resulting cell pellets were resuspended in ALDEFLUOR assay buffer immediately prior to flow cytometry analyses on a CellSimple Cell Analyzer (Cell Signaling Technology). ALDH-positive cells were identified by: (i) gating forward and side scatter to eliminate cellular debris and doublets, and (ii) comparing ALDH profiles of D2.OR derivatives obtained in the absence and presence of DEAB. Data were analyzed using FlowJo software (FlowJo LLC, BD Biosciences, v10.7; RRID:SCR_008520).

Publicly available bulk RNA sequencing (RNA-seq) data on 28 human TNBC cell lines was obtained through Sequence Read Archive (SRA; accession number SRP042620; refs. 22, 23). Correlation analysis was assessed through both Spearman rank-order correlation and Pearson correlation to compare BORG and NANOG transcript expression (ENST00000229307.8) across the 28 human TNBC cell lines. For single-cell RNA-seq studies, two publicly available SMART-seq datasets of human TNBC samples (accession numbers SRP066982 and SRP157044; refs. 24, 25) were aligned to hg38 using Kallisto-BUS (26), followed by analysis using Scanpy (27). Linear correlation and coexpression networks were defined using pyScenic (28) and scLINK (29) and statistical significance of coexpression was confirmed using hypergeometric test.

Single-cell suspensions of parental (i.e., empty vector), BORG-, TRIM28-, or ITGA6-manipulated human or murine breast cancer cells were sorted by flow cytometry on a BD FACSAria (BD Biosciences) into 96-well ultralow attachment plates (Costar; 3471) at densities ranging from 1–100 cells/well, and subsequently were propagated for 10 days in DMEM/F12 media (Gibco-Thermo Fisher Scientific) supplemented with bFGF (20 ng/mL; Invitrogen), EGF (20 ng/mL; Invitrogen), B27 (Gibco, 17504), Pen/Strep (2.5 mL), and Heparin (4 μg/mL; Sigma-Aldrich). Sphere forming frequency in the primary extreme limiting dilution analysis (ELDA) was determined on the basis of the binary ability of sorted cells to form mammospheres, which were analyzed using the ELDA software (30). Afterward, the resulting mammospheres were collected, trypsinized, and resuspended as single-cell suspensions for secondary mammosphere analyses in 96-well ultralow attachment plates as above. The resulting secondary mammospheres were enumerated on day 10 (i.e., total mammospheres and total cell number) and subjected to an additional round of tertiary mammosphere formation as above.

MDA-MB-231 or D2.OR derivatives were washed thoroughly with PBS and immediately harvested with Accutase Enzyme Cell Detachment Media (Invitrogen). The resulting cell pellets (10⁷ cells/mL) were resuspended in 100 μL ice-cold FACS Staining Buffer (PBS, 10% FBS, 0.1% sodium azide) supplemented with fluorophore-conjugated antibodies against either CD24 (BioLegend; catalog no. 311105, RRID:AB_314854; 1:20 dilution), CD44 (BioLegend; catalog no. 338806, RRID:AB_1501195; 1:20 dilution), EpCAM (BioLegend; catalog no. 118205, RRID:AB_1134176; 1:20 dilution), or CD49f (α6 integrin; Thermo Fisher Scientific; 17-0495, RRID:AB_2016626; and 12-0495, RRID:AB_891478; 1:100 dilution). The cells were stained in the dark for 40 minutes at 4°C, at which point they were washed 3× in FACS Staining Buffer prior to being analyzed on either a FACSAria or FACSAria SORP cell sorter (BD Biosciences). Specific cell staining was defined by gating against: (i) isotype control and/or unstained cell populations as negative controls, and (ii) forward and side scatter to eliminate cellular debris and doublets. Data were analyzed using FlowJo software as above. Quantifying cell surface expression of α6 integrin across D2.OR derivatives was accomplished using the aforementioned staining procedures, followed by analysis on a CellSimple Cell Analyzer (Cell Signaling Technology) and FlowJo software as above. Sorted cells were also collected via microcentrifugation, resuspended in RIPA buffer, and subjected to immunoblotting as described below.

Parental (i.e., empty vector), BORG-, TRIM28-, or ITGA6-manipulated human and murine breast cancer cells were propagated in two dimensional (2D) or three-dimensional (3D) culture (Cultrex; Trevigen) for 2 and 6 days, respectively, as described previously (18, 19). Organoids produced in 3D cultures were isolated using Cultrex 3D-Culture Cell Harvesting Kit (Trevigen), at which point total RNA was extracted using TRIzol reagent (Thermo Fisher Scientific) according to the manufacturer's instructions. Total RNA (1 μg/reaction) was reverse transcribed with iScript cDNA Synthesis Kit (Bio-Rad) and subjected to semiquantitative real-time PCR using iQ SYBR Green Supermix (Bio-Rad) as described previously (18). The primer pairs used are provided in Supplementary Table S2.

RNA-seq was performed on parental (i.e., empty vector) and BORG-expressing D2.OR derivatives that were propagated in 3D culture for 7 days, at which point the resulting organoids were harvested and lysed to isolate total RNA as described above. The RNA-seq sample preparation and analysis were performed as described previously (19). Briefly, RNA-seq reactions were performed on a HiSeq 2500 Platform (Illumina) set to rapid run mode on a total of 4 samples, with two replicate RNA-seq experiments being performed for each condition. In doing so, an average of approximately 66 million paired end, 100 nucleotide long reads were obtained for each sample. The quality of the resulting sequencing reads was assessed using FastQC (RRID:SCR_014583), followed by processing on Trim Galore (Babraham Bioinformatics, RRID:SCR_011847) to remove adapter sequences and low-quality nucleotide reads. The remaining reads were aligned to the mouse genome (Mus musculus, GRCm38/mm10) through two parallel pipelines, TopHat (RRID:SCR_013035; ref. 31) and HISAT2 (32). Afterward, differential expression analyses were accomplished using Cufflinks (RRID:SCR_014597) or HTSeq-count (RRID:SCR_011867; ref. 33), followed by edgeR (QL; ref. 34). The differential expression results of the two pipelines were used to identify positively and negatively enriched pathways and gene sets using the C2 module of the MSigDB database of gene sets and GSEA tool (35). Sequencing data were deposited in the National Center for Biotechnology Information Gene Expression Omnibus with accession number GSE116656.

CUT&RUN assays were performed using the CUT&RUN kit from Cell Signaling Technology (#86652) according to the manufacturer's instructions. Briefly, MDA-MB-231 cells were plated in Cultrex-coated 6-well plates at a density of 300,000 cells/well, and were harvested 5 days later using Cultrex 3D-Culture Cell Harvesting Kit (Trevigen). The resulting single-cell suspensions were collected (10⁵ cells/reaction), washed, and rotated with Concavalin A magnetic beads for 2 hours at 4°C with the following antibodies as indicated: (i) anti-trimethyl-Histone H3 (2 μL/reaction; Cell Signaling Technology #9751); (ii) anti-rabbit mAb IgG Isotype control (5 μL/reaction; Cell Signaling Technology #66362); or (iii) anti-TRIM28 (5 μL/reaction; Thermo Fisher Scientific #MA1-2023, RRID:AB_2209892). Subsequently, the reactions were rotated with pAG-MNase enzyme, which was activated by addition of CaCl₂ and incubated sequentially for 30 minutes at 4°C and 37°C. The resulting DNA fragments were purified with Qiagen QIAquick PCR purification kit (#21804). Input samples underwent DNA extraction and sonication (5 Watts, 30-second intervals for 25 cycles) before purifying DNA and performing qRT-PCR as described previously.

Probing publicly available chromatin immunoprecipitation sequencing (ChIP-seq) data were accomplished using the Cistrome Data Browser and Toolkit (cistrome.org/db; doi.org/10.1007/s40484-020-0204-7). Briefly, 16 TRIM28 ChIP-seq datasets derived from human progenitor and stem cell populations were identified in the Cistrome Data Browser and Toolkit. The putative target search function within the Cistrome Data Browser facilitated the mapping of TRIM28 binding peaks within 2,000 bp of the transcriptional start sites of genes that comprise the LIM_MAMMARY_STEM_CELL_UP gene set (36), which is significantly upregulated in BORG-expressing cells (see above).

D2.OR, D2.A1, or MDA-MB-231 breast cancer cells and their derivatives were harvested in RIPA lysis buffer [50 mmol/L Tris-HCl, pH 7.4, 150 mmol/L sodium chloride, 6 mmol/L sodium deoxycholate, 1% IGEPAL-630, 0.1% SDS, 1 mmol/L protease inhibitor cocktail (CalBiochem), 10 mmol/L sodium orthovanadate, 40 mmol/L β-glycerophosphate, and 20 mmol/L sodium fluoride] on ice. The resulting cell lysates were clarified by microcentrifugation and immunoblotted as described previously (18) using antibodies against ITGA6 (1:2,000 dilution; Abcam #ab181551), TRIM28 (1:1,000 dilution; Thermo Fisher Scientific #PA5-27648, RRID:AB_2545124), or β-actin (1:10,000 dilution; Sigma-Aldrich #A5441, RRID:AB_476744). Quantification of band density for proteins of interest and β-actin loading control was carried out using ImageJ software (RRID:SCR_003070).

The ability of BORG to impact the tumor-initiating capacity of D2.OR cells was assessed by implanting serial dilutions of firefly luciferase-expressing parental and BORG-expressing D2.OR cells (e.g., 50,000, 5,000, or 500 cells/mouse) into the mammary fat pads of 8-week-old BALB/c mice (The Jackson Laboratory). Each experimental arm was comprised of 10 mice (i.e., 60 mice randomly and blindly divided into six groups), whose formation of mammary tumors was monitored over a span of 14 weeks. Alterations in primary tumor size were measured twice weekly using digital calipers and all animals were subjected to bioluminescent imaging immediately prior to necropsy to identify nonpalpable tumor cells residing in the mammary fat pads. Tumor initiating capacity was calculated using the ELDA software based on the binary assessment of the absence or presence of palpable mammary tumors in each animal.

The formation of pulmonary metastases was assessed by injecting firefly luciferase- and BORG- or ITGA6-manipulated D2.OR and D2.A1 derivatives into the lateral tail veins of 8-week-old BALB/c mice (10⁶ cells/mouse; 5 mice per arm) as indicated. Pulmonary colonization and eventual metastatic burden was quantified via bioluminescent imaging. Briefly, at the annotated timepoints, the mice were injected with d-luciferin (Gold Biotechnology), anesthetized with isoflurane, and imaged for 5 minutes on an IVIS Spectrum (PerkinElmer) in vivo imaging system. Quantification was performed with Living Image software (Xenogen, RRID:SCR_014247) to determine luminescent flux through the lungs at varying timepoints after initial pulmonary inoculation. Values were normalized to total luminescence at time of injection.

All animal studies were performed according to procedures approved by the Institutional Animal Care and Use Committee at Case Western Reserve University (Cleveland, OH). This study was conducted under the protocol 2015-0103 titled “Role of Long and Short Noncoding RNAs during TGFbeta Driven Breast Cancer Development and Metastasis”.

Unless otherwise stated, all experiments were performed in biological triplicate where significance was determined by unpaired two-tailed Student t tests. P ≤ 0.05 were considered to be statistically significant. GSEAs were performed using a gene set–based permutation test to yield FDR and nominal P values that indicated the significance of enrichment of the corresponding collection of genes. As denoted in figure legends, all results were conveyed as mean ± SE unless otherwise stated.

We recently observed aberrant BORG expression to correlate positively with the tumorigenicity of TNBCs, doing so by promoting enhanced survival signaling and chemoresistance (19), and by driving metastatic dissemination and recurrence (18). Importantly, these phenotypes are reminiscent of those attributed to BCSCs (6, 20, 37), which prompted us to determine whether aberrant BORG expression also plays a role in promoting the self-renewal and expansion of BCSCs. In doing so, murine D2.OR cells were co-stained with fluorescently-labeled antibodies against α6 integrin (CD49f) and EpCAM, followed by flow-cytometry sorting to isolate populations of either non-clonogenic, differentiated cells (e.g., α6 integrin^Low) or stem-like/bipotent progenitor cells (e.g., α6 integrin^HighEpCAM^High; Fig. 1A; ref. 38). As shown in Fig. 1B, “stem-like” α6 integrin^HighEpCAM^High D2.OR cells were significantly more adept at forming mammospheres than were their “differentiated” α6 integrin^Low counterparts. Moreover, to further enrich for a phenotypically stem-like population, D2.OR cells were grown in sphere-forming assays (i.e., mammospheres), which demonstrated that mammospheres produced by D2.OR cells also harbored significantly greater quantities of BORG as compared with their parental and “differentiated” (α6 integrin^Low) counterparts (Fig. 1C). Similar associations of augmented BORG expression with BCSCs were also observed in human MDA-MB-231 cells, which were sorted into “stem-like” (CD44^HighCD24^Low) and “non-stem” (CD24^High) populations by flow cytometry (Fig. 1D; ref. 39). In doing so, BORG expression was significantly higher in “stem-like” (CD44^HighCD24^Low) MDA-MB-231 cells as compared with their “non-stem” (CD24^High) counterparts, and to parental MDA-MB-231 and normal HMECs (Fig. 1E). Taken together, these findings link robust BORG expression to the extent of “stem-like” phenotypes adopted by human and murine BCSCs.

The notion that BORG expression governs the stemness of BCSCs was reinforced by examining our RNA-seq dataset obtained from parental and BORG-expressing D2.OR organoids propagated in 3D cultures (19). Figure 1F shows that BORG significantly enriched the transcriptional signature of genes whose expression are consistently augmented in human and murine mammary stem cells (36). Included among these genes is Aldh1a3 (Fig. 1G), which encodes for the detoxifying enzyme ALDH and is critical for mammary stem cell self-renewal and chemoresistance (40–42). Accordingly, ALDEFLUOR assays (43) demonstrated that ALDH1 activity was notably higher in BORG-expressing D2.OR cells as compared with their parental counterparts (Fig. 1H). Furthermore, probing publicly available RNA-seq datasets derived from a multitude of human breast cancer cell lines revealed that BORG expression correlated positively with that of NANOG, a master transcription factor critical to maintaining stem cell pluripotency (Fig. 1I; refs. 44, 45). This correlation was further confirmed by analysis of two publicly available single-cell RNA-seq datasets derived from a total of 11 primary human TNBC tumors (24, 25). Although lncRNAs can be poorly represented in single-cell RNA-seq due to their low expression levels, BORG was detected in approximately 5% of TNBC cells. Among BORG-expressing TNBC cells, NANOG was ranked as the second most abundant gene relative to those exhibiting the strongest coexpression pattern with BORG in individual TNBC tumor cells (P < 0.0002; Fig 1J). Consistent with this correlation, we found NANOG expression to be significantly elevated in D2.OR cells engineered to overexpress BORG (Fig. 1K). Further analysis of publicly available single-cell RNA-seq datasets in all BORG-expressing cells showed that the expression of BORG correlated with that of additional genes involved in stemness and aggressive cancer phenotypes (Supplementary Fig. S1). Taken together, these findings associate BORG expression with the molecular features exhibited by human and murine BCSCs.

A hallmark of BCSCs is their propensity to survive and proliferate by forming mammospheres when propagated under nonadherent culture conditions (39, 46). To directly quantify the ability of BORG to promote the self-renewal and expansion of BCSCs, we subjected parental and BORG-expressing D2.OR cells to ELDAs (30, 47). Indeed, engineering D2.OR cells to overexpress BORG significantly increased the frequency of BCSCs present within D2.OR cells (Fig. 2A), as well as dramatically augmented the overall size of mammospheres formed by these same BCSCs (Fig. 2B). Along these lines, Fig. 2C shows that BORG significantly enhanced the ability of D2.OR cells to form secondary and tertiary mammospheres in response to serial culture conditions. In contrast, depleting endogenous BORG expression in metastatic D2.A1 cells (Fig. 2D), which harbor high quantities of BORG (18), and human MDA-MB-231 cells significantly impaired their capacity to form primary, secondary, and tertiary mammospheres in vitro (Fig. 2E,–H). Finally, we monitored the tumor-initiating capacity of parental and BORG-expressing D2.OR cells by implanting serially diluted concentrations of these D2.OR populations into the mammary fat pads of syngeneic female BALB/c mice. As shown in Fig. 2I, BORG-expressing D2.OR cells were markedly more proficient in their tumor initiating capacity as compared with their parental counterparts. Collectively, these findings indicate that BORG augments the self-renewing capacity and tumorigenicity of human and murine TNBC cells.

Recently, we established the ability of BORG to interact physically with TRIM28, an event that directs and enhances TRIM28-mediated transcriptional repression at unique genomic loci (18). Interestingly, among the many functions attributed to TRIM28 is the ability to promote the self-renewal of mammary and induced pluripotent stem cells (48–50), suggesting that the formation of BORG:TRIM28 complexes may be essential to the acquisition of BCSC phenotypes. In testing this supposition, we used CRISPR/Cas9 genome editing to inactivate TRIM28 expression in BORG-expressing D2.OR cells (Supplementary Fig. S2A; ref. 18), and subsequently monitored their ability to form mammospheres in relation to their BORG- and TRIM28-expressing counterparts. As shown in Fig. 3A,–C, inactivating TRIM28 expression and activity significantly impaired the acquisition of BCSC phenotypes afforded to D2.OR cells by BORG. Moreover, structure-function analyses that monitored the impact of six distinct internal BORG deletions (i.e., Del1–6) to alter the formation of BORG:TRIM28 complexes identified the sequences that encompassed Del4 and Del5 of BORG (i.e., nucleotides 1042–2040) as those operant in binding murine TRIM28 (18). Accordingly, Fig. 3D and E show that BORG constructs capable of binding TRIM28 (full-length and Del1-BORG) dramatically enhanced the mammosphere forming activity of D2.OR cells as compared with those engineered to express Del5-BORG mutants unable to bind TRIM28. These findings highlight the functional importance of BORG:TRIM28 complexes to elicit BCSC phenotypes in response to BORG. Along these lines, genetic inactivation of TRIM28 also prevented BORG from inducing ALDH1 activity (Fig. 3F), as well as upregulating expression of stem-associated transcripts, Aldh1a3 and Nanog (Fig. 3G). Interestingly, sorting MDA-MB-231 cells for stem-like (CD44^hiCD24^lo) and non-stem (CD24^hi) populations did not yield a change in TRIM28 protein expression (Supplementary Fig. S2B). Finally, we recently described the ability of BORG to upregulate the expression of NEMO/IKKγ and promote its activation of NFκB, events that transpire independent of the formation of BORG:TRIM28 complexes (19). Because NFκB signaling has been linked to the expansion of mammary stem cell populations (51, 52), we expressed a dominant-negative form of IκBα (i.e., “superrepressor”; refs. 19, 53) to inhibit NFκB activation in parental and BORG-expressing D2.OR cells to determine the necessity of NFκB signaling to the acquisition of BCSC phenotypes elicited by BORG. Although the NFκB pathway plays a major role in promoting chemoresistant phenotypes in response to BORG (19), ablating NFκB activity failed to impact the mammosphere-forming capacity of BORG-expressing D2.OR cells (Supplementary Fig. S2C–S2E). Collectively, these findings indicate that the stem-like traits induced by BORG are primarily driven via its manipulation of the molecular functions of TRIM28.

Although the aforementioned findings demonstrate the importance of BORG:TRIM28 complexes in promoting BCSC phenotypes, the molecular targets of these events remain to be fully elucidated. To address this question, we probed publicly available ChIP-seq datasets generated from various human stem cell populations to identify genes within the BORG-dependent mammary stem cell transcriptional signature (Fig. 1F) whose promoters are enriched for TRIM28 occupancy. In this analysis, TRIM28 is enriched at the promoter regions of eight genes found in the mammary stem cell transcriptional signature (e.g., ITGA6, TCF4, MYLK, PCDH19, ZC3H12B, FAM184A, TM7SF3, EFCAB1; data not shown). Interestingly, we observed strong binding of TRIM28 within 2 kB of the transcriptional start site of ITGA6 in both human embryonic and hematopoietic stem cells (Fig. 4A; refs. 54, 55). Along these lines, CUT&RUN analyses identified enrichment of TRIM28 at the ITGA6 promoter region in MDA-MB-231 cells (Fig. 4B). Importantly, Itga6 encodes for α6 integrin/CD49f, which plays a notable role in regulating the self-renewal and multipotency of BCSCs (56–61). Accordingly, we found that the abundance of Itga6 transcripts, as well as the quantity of cell surface α6 integrin proteins were both elevated in BORG-expressing D2.OR cells as compared with their parental and TRIM28-deficient counterparts (Fig. 4C and D). Likewise, depleting BORG expression in metastatic D2.A1 cells significantly reduced their expression of Itga6 transcripts and protein (Fig. 4E and F). Similarly, flow cytometry-based isolation of D2.OR cells that expressed robust α6 integrin (i.e., α6 integrin^High; Fig. 4G) yielded a subpopulation of D2.OR cells that: (i) possessed enhanced sphere-forming capacity (Fig. 4H), and (ii) harbored significantly elevated levels of BORG as compared with parental D2.OR cells or those that exhibited low quantities of α6 integrin (i.e., α6 integrin^Low; Fig. 4I). Collectively, these findings identify α6 integrin as a novel gene target of BORG:TRIM28 complexes; they also demonstrate the functional importance of α6 integrin to the acquisition of BCSC phenotypes elicited by BORG.

To further explore the functional importance of α6 integrin to tumorigenic activities of BORG in TNBCs, we depleted α6 integrin expression in BORG-expressing D2.OR cells (Fig. 5A and B) and subsequently measured their sphere forming capacity via ELDAs. As shown in Fig. 5C and D, depleting α6 integrin expression and activity significantly impaired the ability of BORG-expressing D2.OR cells to form both primary and secondary mammospheres. More importantly, loss of α6 integrin expression abrogated the ability of BORG to drive the metastatic outgrowth of D2.OR cells following their inoculation into the lateral tail veins of female BALB/c mice (Fig. 5E). Along these lines, we also depleted α6 integrin expression in D2.A1 cells (Fig. 5F and G) and subsequently monitored their ability to form mammospheres via ELDAs. Accordingly, Fig. 5H and I show the dose dependency of α6 integrin expression on the ability of D2.A1 cells to form primary and secondary mammospheres. Moreover, monitoring α6 integrin expression in secondary mammospheres formed by D2.A1 shα6.1 cells showed that α6 integrin expression approximated that of their parental counterparts (Supplementary Fig. S3A), suggesting that D2.A1 cells that possessed low levels of α6 integrin were negatively selected during serial mammosphere cultures. Interestingly, we also found that depleting α6 integrin in D2.A1 cells failed to impact their growth in 3D cultures (Supplementary Fig. S3B), indicating the specificity of α6 integrin in regulating stem-like phenotypes by BORG. Finally, we monitored the metastatic competency of parental, α6 integrin-deficient, and BORG-deficient D2.A1 cells following their inoculation into the lateral tail veins of BALB/c mice. As shown in Fig. 5J, genetic inactivation of α6 integrin reduced the pulmonary outgrowth of D2.A1 cells in a manner comparable with that produced by genetic inactivation of BORG. Indeed, mice harboring parental D2.A1 cells exhibited significantly shorter metastatic latency (Supplementary Fig. S3C) and larger metastatic burden (Fig. 5K and L; Supplementary Fig. S3D) as compared with those depleted of α6 integrin expression. Taken together, these findings indicate that BORG:TRIM28 complexes elicit BCSC phenotypes coupled to the tumorigenicity of TNBCs in part through their ability to induce α6 integrin expression. Moreover, these findings also suggest that targeted inactivation of α6 integrin in BORG-expressing cells dramatically reduces BORG-induced BCSC phenotypes and metastatic outgrowth.

Integrins are type I transmembrane glycoproteins that are classified into α and β subfamilies capable of interacting noncovalently to form least 24 different distinct α:β heterodimeric complexes that function in relaying bidirectional signals between cells and their microenvironments (62). Through their large extracellular domains, integrins interact with a diverse assortment of extracellular matrix (ECM) components to transduce signals from the stromal microenvironment and neighboring cells into a variety of cellular functions, including those critical for tumor development and metastatic progression (63–65). Most notably, accumulating evidence has revealed that the self-renewal and expansion of CSC populations relies heavily upon the direct interactions between integrins and their neighboring stromal components (66). Despite the vast diversity of integrin heterodimers known to exist in normal and malignant cells, the chief integrin most frequently associated with the acquisition of stem-like phenotypes is α6 integrin. Indeed, α6 integrin expression is elevated significantly in over 30 unique stem cells populations; it is also: (i) the only gene commonly observed to be upregulated in both embryonic and adult stem cells (63, 67), and (ii) essential to the function and phenotypes of normal and malignant mammary stem cells (57, 58, 63, 68–70). Herein we establish α6 integrin as a novel gene target of BORG:TRIM28 complexes and their ability to drive the self-renewal, expansion, and tumorigenicity of TNBC BCSCs.

Previously, we demonstrated that the pathophysiology BORG exerts on human and murine TNBCs transpires in a highly context-dependent manner. For instance, the formation of BORG:TRIM28 complexes reinstates cell-cycle progression and emergence from dormancy in disseminated TNBCs, an event that manifests only in 3D-culture systems or in vivo settings, not traditional 2D-culture systems (18). Similarly, maintenance of BCSC populations is contingent upon the niche in which they reside, as an optimized compendium of paracrine factors, ECM organization, and vascular and immune effector cells coalesce in permitting BCSC self-renewal and preventing their terminal differentiation (71). Interestingly, BORG expression is regulated by these same stromal components, including paracrine signals (e.g., TGFβ, BMP2, and BMP7), hypoxia, nutrient deprivation, and mechanotransduction by the ECM (18, 19, 72). It therefore stands to reason that BORG, despite being solely localized to the nucleus (21), may function as an essential effector molecule that underlies the plasticity of BCSCs in response to evolving niche-derived signals, thereby enhancing the stemness and metastatic ability of BCSCs via its ability to induce α6 integrin and enhance ECM communication. This conclusion is bolstered by the fact that BORG-induced regulation of Itga6 failed to affect the 3D outgrowth of bulk tumor cell populations (Supplementary Fig. S3B), but instead specifically suppressed that of BCSCs both in vitro and in vivo (Figs. 4 and 5).

As noted above, we show for the first time that TRIM28 binds to the promoter region of Itga6 in embryonic and hematopoietic stem cells (Fig. 4A and B), implying that Itga6 expression should be mitigated in accordance with the canonical role of TRIM28 acting as a transcriptional corepressor. In contrast, we found Itga6 to be upregulated dramatically in a BORG- and TRIM28-dependent manner (Fig. 4C and D), a paradoxical finding that likely reflects the pleiotropic roles played by TRIM28 in human cancers (73). More specifically, the sumoylation status of TRIM28 dictates its interaction with histone methyltransferases and histone deacetylases, thereby shaping epigenetic marks on histones and altering the transcriptional activity at loci bound by TRIM28 (74). Furthermore, TRIM28 enhances transcriptional activation of a number of target genes when interacting with Enhancer of Zeste Homolog 2 (EZH2) in breast cancer cells (49), indicating that TRIM28 is capable of both transcriptional repression and transcriptional enhancement. Accordingly, we recently observed BORG to be essential in coordinating the formation and function of TRIM28:EZH2 complexes in TNBC cells (K.A. Parker, S. Valadkhan, and W.P. Schiemann, in preparation). Because the plasticity exhibited by CSCs depends upon signaling cascades orchestrated by epigenetic regulators (75, 76), it is tempting to speculate that modulation of Itga6 expression reflects remodeling of the epigenetic landscape of TNBCs at the hands of BORG:TRIM28 complexes.

Finally, our discovery of the BCSC-enhancing features of BORG are profoundly significant, as they further elucidate the pro-metastatic activities of BORG (17). Indeed, disseminated TNBC cells harboring robust levels of BORG are significantly more likely to exhibit increased proliferative capacity (18), enhanced resistance to environmental and chemotherapeutic stressors (19), and elevated tumor-initiating activity (Fig. 5). Taken together, these findings cement BORG as a powerful dictator of malignant and metastatic progression; they also support the future development of BORG-directed therapeutics as a means to eradicate TNBCs, particularly their capacity to exhibit aggressive metastatic dissemination and rapid disease recurrence.

K.A. Parker reports grants from NIH during the conduct of the study. A.J. Gooding reports grants from NCI during the conduct of the study. S. Valadkhan reports grants from NIH during the conduct of the study; in addition, S. Valadkhan has a patent for BORG as a therapeutic target, patent number 9308218, issued. W.P. Schiemann reports grants from NCI and Case Comprehensive Cancer Center during the conduct of the study.

K.A. Parker: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. A.J. Gooding: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. S. Valadkhan: Conceptualization, resources, data curation, formal analysis, supervision, investigation, writing–review and editing. W.P. Schiemann: Conceptualization, resources, formal analysis, supervision, funding acquisition, project administration, writing–review and editing.

Members of the Schiemann Laboratory are thanked for critical comments and reading of the article. We also acknowledge the expertise provided by members of the Case Comprehensive Cancer Center's Shared Resource Cores, particularly the Imaging Research Core, the Cytometry and Microscopy Core, and the High Performance Computing Resources in the Core Facility for Advanced Research Computing.

Research support was provided in part by the NIH (CA236273 to W.P. Schiemann and S. Valadkhan; F30CA203233 to A.J. Gooding; T32GM008803 and F31CA257637 to K.A. Parker), and by pilot funding from the Case Comprehensive Cancer Center's Accelerator Award, which is supported by the Case Council and Friends of the Case Comprehensive Cancer Center (to W.P. Schiemann and S. Valadkhan).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.