Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Stem-like breast cancer cells in the activated state resist genetic stress via TGFBI-ZEB1

Stem-like breast cancer cells in the activated state resist genetic stress via TGFBI-ZEB1 www.nature.com/npjbcancer ARTICLE OPEN Stem-like breast cancer cells in the activated state resist genetic stress via TGFBI-ZEB1 1,2,4 1,2 2,3 1,2 1,2 2,3 1,2✉ Qi Sun , Yufen Wang , Adam Officer , Brianna Pecknold , Garrett Lee , Olivier Harismendy and Jay S. Desgrosellier Breast cancer cells with stem-like properties are critical for tumor progression, yet much about these cells remains unknown. Here, we characterize a population of stem-like breast cancer cells expressing the integrin αvβ3 as transcriptionally related to activated stem/basal cells in the normal human mammary gland. An unbiased functional screen of genes unique to these cells identified the matrix protein TGFBI (BIG-H3) and the transcription factor ZEB1 as necessary for tumorsphere formation. Surprisingly, these genes were not required for cell proliferation or survival, but instead maintained chromosomal stability. Consistent with this finding, CRISPR deletion of either gene synergized with PARP inhibition to deplete αvβ3 stem-like cells, which are normally resistant to this therapy. Our findings highlight a critical role for TGFBI-ZEB1 protection against genetic stress as a key attribute of activated stem- like cells and suggest that disrupting this ability may enhance their “BRCAness” by increasing sensitivity to PARP inhibitors. npj Breast Cancer (2022) 8:5 ; https://doi.org/10.1038/s41523-021-00375-w INTRODUCTION Despite the importance of CSCs for breast cancer progression, studies of these cells are limited by their scarcity and a lack of Tumor-initiating cancer stem cells (CSCs) bearing similarities to appropriate cell line models that reflect the heterogeneity in a adult mammary stem cells (MaSCs) are important contributors to 1–4 patient tumor. In the present study, we make use of our previously breast cancer progression and metastasis . However, adult characterized heterogeneous breast cancer cell line models to MaSCs are highly dynamic, frequently changing their cell state— overcome this limitation. These cell lines better recapitulate the a physiological condition due to altered gene expression or 2,15 intratumoral heterogeneity in patient disease , including a signaling—in response to hormonal cues. In fact, the mammary subset of αvβ3 CSCs, and allow us to directly assess a role for gland is one of the most dynamic organs in adult women, these cells compared to other neighboring tumor cell types. Based undergoing robust epithelial remodeling in response to hormones on our prior findings, we hypothesized that tumor cells bearing during the menstrual cycle and pregnancy that is driven by stem αvβ3 may similarly express genes found in stem/basal cells in cells. While normally quiescent, MaSCs respond to hormones response to hormonal signaling during the menstrual cycle or indirectly via paracrine signals to become active and contribute to pregnancy. Furthermore, since αvβ3 is a biomarker of aggressive 5–8 9 epithelial remodeling since they lack hormone receptors . cancer cells, we propose that these cells may contain unique These active stem cells exhibit enhanced proliferation and genes/pathways that could serve as potential vulnerabilities. To 5–8 migration , features that make this signaling state likely to be address these questions, we performed unbiased whole tran- hijacked by tumor cells. This raises the tantalizing question of scriptome analysis of αvβ3 CSCs. These findings represent an whether some of the most aggressive CSCs may further acquire initial step toward revealing similarities between these cells and properties associated with activated stem cells. normal mammary cell types and identifying key pathways that We previously showed that the cell surface receptor integrin may control their aggressive behavior. αvβ3 is a key switch turned-on by activated stem cells as they are mobilized for epithelial remodeling during pregnancy . Using αvβ3 as a marker, we further characterized a unique and RESULTS particularly aggressive population of stem-like breast cancer Surface αvβ3 marks stem-like cells enriched for tumor 11 + cells . Unexpectedly, we found αvβ3 CSCs in aggressive patient initiation tumors that were either estrogen receptor-positive (ER ), human Breast cancers are heterogeneous, with cells representing epidermal growth factor receptor-positive (HER2 ), or triple- different mammary lineages often found in the same tumor, negative , suggesting these cells may contribute to disease 14,16 including those with stem-like properties . We previously progression in all clinical subtypes. Notably, αvβ3 expression was showed in patient breast cancers that cells expressing the surface not synonymous with traditional CSC profiles, such as CD44 / marker integrin αvβ3 represent a stem-like cancer cell subset Low1 + Low2 12 CD24 , CD49f /EpCAM , the claudin-low intrinsic subtype associated with disease progression in a diverse array of 2,13,14 + 11 + or mesenchymal markers . Instead, αvβ3 cells represented a subtypes . Despite the potential significance of αvβ3 CSCs for distinct subset of these broader classifications . Our prior findings disease progression, few good models exist to study these cells in provided valuable insight regarding the aggressive nature of the context of other non-stem cells. Additionally, the scarcity of αvβ3 CSCs and emphasized the need to further elucidate the these cells in patient samples represents another practical unique genes and signaling pathways required for their function. limitation to studying these cells. One potential in vitro model 1 2 Department of Pathology, University of California, San Diego, La Jolla, CA 92093, USA. Moores Cancer Center, University of California, San Diego, La Jolla, CA 92093, USA. 3 4 Department of Medicine, University of California, San Diego, La Jolla, CA 92093, USA. Present address: Department of General Surgery, The First Affiliated Hospital of Xi’an Jiaotong University, Xi’an 710061, China. email: jdesgrosellier@ucsd.edu Published in partnership with the Breast Cancer Research Foundation 1234567890():,; Q. Sun et al. a b Tumor Incidence 32.6% 57.3% 10 EpCAM high EpCAM low Number of sorted cells injected v3 neg. v3 pos. v3 neg. v3 pos. 50,000 2/2 2/2 1/2 2/2 35,000 5/11 5/12 3/13 10/13 15,000 0/8 1/8 2/6 3/8 2 Frequency 1/64,755 1/59,918 1/91,076 1/24,615 6.58% 3.49% 95% Confidence Interval (1/31,119- (1/29,954- (1/40,447- (1/14,335- 2 3 4 5 10 10 10 10 134,748) 119,857) 205,083) 42,226) P = 0.0324 v3 c d Differentiation Low + Parental HCC38 EpCAM /v3 * 5 5 5 30.1% 61.3% 5.09% 73.7% 10 10 4 4 10 10 3 3 10 10 2 2 0 10 10 4.32% 4.26% 5.28% 15.9% - + - + v3 v3 v3 v3 2 3 4 5 2 3 4 5 high low EpCAM EpCAM 10 10 10 10 10 10 10 10 v3 Fig. 1 Integrin αvβ3 enriches for tumor-initiating ability and stem-like properties. a Representative FACS density plot of HCC38 breast cancer cells showing the live, CD49f cells according to their cell surface EpCAM and αvβ3 status. b Table describing the frequency of tumor formation per fat pad injected for each sorted cell type. Results pooled from four independent experiments. c Histogram showing the estimated number of tumor-initiating cells from the data in (b). b, c Statistics by Extreme Limiting Dilution Analysis (ELDA), which uses a chisquare likelihood ratio test to calculate p-values between groups. *P < 0.05. d Representative FACS density plots showing differentiation of Low + - sorted EpCAM /αvβ3 cells re analyzed after 6 weeks (10 passages). a, d n = 3 independent experiments. See also Supplementary Fig. 1. for our studies is the heterogeneous HCC38 cell line, which types, representing an ideal in vitro model to begin parsing critical + high consists of luminal-like (CD49f /EpCAM ) and stem-like cell gene expression and signaling differences. + low 13 types (CD49f EpCAM ) . Our analysis of surface αvβ3 in these Low + cells further identified a population of EpCAM /αvβ3 cells enriched for stemness properties such as tumorsphere formation αvβ3 CSCs express genes associated with activated stem and self-renewal . Thus, to more closely reflect the situation in cells patients’ tumors, we examined the HCC38 breast cancer cell line To discover critical stemness genes in αvβ3 CSCs, and determine as a potential model for our studies of αvβ3 CSCs. any similarities with normal mammary cell types, we performed To rigorously compare stemness traits in vivo, we sorted HCC38 bulk RNA-Seq analysis. The principal component analysis high- cells into four populations based on their EpCAM and αvβ3 status lighted a surprising amount of distinction between αvβ3 and (Fig. 1a and Supplementary Fig. 1a) prior to evaluating their αvβ3 cells (Fig. 2a), even greater than that due to EpCAM status tumor-initiating potential in vivo (Fig. 1b, c). Sorted cells were 2 alone, a widely-used marker to identify stem-like cells . This was injected orthotopically into the inguinal mammary gland fat pads high Low even more surprising since EpCAM and EpCAM cell types of adult female immunocompromised mice, then compared for are widely separated and distinct populations, whereas αvβ3 their ability to initiate new tumors in limiting dilution assays (Fig. expression represented a continuum of high and low expressers Low + 1b, c). We now show that EpCAM /αvβ3 cells possess about a (Fig. 1a). Meanwhile, both αvβ3 cell types exhibited a high 4-fold greater ability to initiate tumors relative to other HCC38 cell degree of similarity at the transcriptional level (Fig. 2a). This types (Fig. 1b, c). This is consistent with our prior tumorsphere suggests a potential relationship between these cell types, results and further supports their characterization as stem-like consistent with our differentiation results (Fig. 1d). To probe this cells. Another important attribute of stem-like cells is their ability Low + relationship further we compared the expression of a few select to differentiate. To determine if EpCAM /αvβ3 cells also markers of normal mammary cell types. Since αvβ3 expression has possessed this property we cultured sorted cells for exactly 10 previously been shown to occur on both stem/basal and luminal passages prior to re-analyzing by flow cytometry (Fig. 1d and progenitor cells in the normal murine and human mammary Supplementary Fig. 1b). This showed that indeed these cells were 10,17,18 gland we examined markers previously established to capable of differentiating into all three of the other cell types differentiate between these two cell types . Our analysis of these analyzed. Comparison with parental HCC38 cells showed that mammary cell markers showed that both αvβ3 cell types are there was a great deal of lineage specificity with regards to each Low + enriched for genes associated with stem/basal, but not luminal sorted cell type, with EpCAM /αvβ3 cells displaying a High + progenitor cells (Fig. 2b and Supplementary Fig. 2a), consistent preference for differentiating into EpCAM /αvβ3 cells (Fig. High with our hypothesis that αvβ3 CSCs display characteristics of 1d and Supplementary Fig. 1b). Interestingly, the EpCAM / adult MaSCs. αvβ3 cells changed the least, suggesting that they represent a more stable differentiated cell type (Supplementary Fig. 1b). Thus, To further probe any potential similarity between αvβ3 CSCs similar to patients’ tumors, we show that the HCC38 cell line and activated stem/basal cells from the normal mammary gland contains a rare subset of αvβ3 CSCs, in addition to non–stem cell we assessed the differentially expressed genes (DEG) within each npj Breast Cancer (2022) 5 Published in partnership with the Breast Cancer Research Foundation HCC38 cells Tumor Initiation 1234567890():,; Tumor-initiating cells (Estimate per 100,000) EpCAM EpCAM Follicular (Inactive) Q. Sun et al. a b High - Low - Mammary cell markers EpCAM /v3 EpCAM /v3 + - High + Low + (v3 vs v3 ) EpCAM /v3 EpCAM /v3 1.5 Low EpCAM High EpCAM 1.0 Luminal progenitor 0.5 genes 0.0 -0.5 Stem/basal genes -5 -1.0 -10 -5 0 5 PC1: 30% variance c d Activated stem/basal cell gene sets Low High EpCAM EpCAM Hypoxia Response to abiotic stimulus EpCAM High DEG v3- v3+ EMT EpCAM High Tissue development 146 468 Regulation of cell death Response to oxygen levels Response to external stimulus Cellular response to stress 104 425 Myc targets V1 Posttranscriptional regulation of gene expression EpCAM Low v3- v3+ Ribonucleoprotein complex biogenesis EpCAM Low mRNA processing Ribosome biogenesis -3 -2 -1 0 1 2 3 + - Enrichment Score (v3 vs v3 ) Fig. 2 αvβ3 CSCs are similar to activated stem/basal cells from the normal human mammary gland. a Principal component analysis (PC) performed on bulk RNA-Seq data from each of the indicated HCC38 sorted cell types. b Relative expression of select gene markers of stem/ + − basal or luminal progenitor cells in αvβ3 versus αvβ3 cells. Data represent the mean ± s.e.m. # = not significant. c Venn diagrams depicting the number of differentially expressed genes (DEG) identified in each cell type. The selection criteria was ≥1.5-fold change in gene expression + − and P < 0.05. d Comparison of αvβ3 versus αvβ3 cell GSEA results with the top gene sets enriched in stem/basal cells during luteal (Active) versus follicular (Inactive) menstrual cycle phases. b, d Statistics by Student’s t-test with Benjamini-Hochberg multiple comparisons test. a–d n = 3 independent experiments. See also Supplementary Fig. 2. cell type (Fig. 2c) and performed gene set enrichment analysis Identification of key genes unique to αvβ3 CSCs (GSEA). While αvβ3 cell types were closely related (Fig. 2a, b), we Based on our findings that αvβ3 CSCs enrich for stemness Low + identified 180 genes enriched in EpCAM /αvβ3 cells compared properties such as tumor initiation (Fig. 1b, c) we wished to High + to EpCAM /αvβ3 cells (Fig. 2c and Supplementary Fig. 2b). We determine the key genes and signaling pathways critical for their then compared gene sets enriched in both αvβ3 cell types with function. We began by selecting several gene sets associated with those from activated stem/basal cells in the normal human αvβ3 CSCs based on their relevance to breast cancer, stem cells, mammary gland. Since data from pregnancy is unavailable, we or signaling pathways (Fig. 3a and Supplementary Fig. 3a). By compared our GSEA results with published data from normal determining which of the 180 DEG’s identified in Fig. 2c were basal/stem cells during the luteal (Active) versus follicular present within each gene set, we identified 20 candidate genes 19 + (Inactive) phases of the menstrual cycle . Many of the same unique to αvβ3 CSCs (Fig. 3b and Supplementary Fig. 3b), hormone-induced changes that occur during the luteal phase also referred to as our αvβ3 CSC signature. In order to perform a happen during pregnancy. The results were striking, as gene sets functional screen of these genes, we sought to identify appro- found in activated stem/basal cells were overwhelmingly shared priate surrogate cell lines for our αvβ3 CSCs. For this analysis, we by αvβ3 cancer cells, while those in inactive cells were not (Fig. made use of published gene sets from 28 breast cancer cell lines 2d and Supplementary Fig. 2c). Interestingly, of the gene sets that were previously used to classify these cells according to their 13 + analyzed, the αvβ3 cells differed only in the genes involved in intrinsic subtype . Comparison with our αvβ3 CSC signature the cellular response to stress, with this representing a unique revealed specific enrichment in the claudin-low cell type (Fig. 3c), Low + 11 feature distinguishing EpCAM /αvβ3 cells (Fig. 2d). These consistent with prior characterization of these cells as stem-like . findings highlight an association between αvβ3 CSCs and the However, careful analysis of each of the eight claudin-low cell lines activated state in normal mammary stem/basal cells and suggest revealed that only three of them displayed any enrichment that a heightened response to stress may be a key distinguishing beyond the parental HCC38 cells (Fig. 3d), in which αvβ3 CSCs feature of these cells. are only a small fraction of the total cells. The three cell lines Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 5 TAGLN ACTA2 SNAI2 CDK1 TP63 MYLK CAV1 KIT ELF5 PC2: 19% variance Log2 fold change Luteal (Active) Log2 Fold Change Q. Sun et al. ab High - EpCAM /v3 High + EpCAM /v3 Low - EpCAM /v3 Gene sets with significantly modulated genes Low + EpCAM /v3 Low + - (EpCAM /v3 vs v3 ) WNT5A Schuetz Breast Cancer Ductal Invasive Up VGLL3 PRICKLE1 Lim Mammary Stem Cell Up TSPAN5 COBL 1 Koinuma Targets of Smad2 or Smad3 SFRP4 POSTN Kim WT1 Targets Up 0 FREM2 RSAD2 Plasari TGFB1 Targets 10hr Up RBPMS2 -1 ZEB1 Wong Adult Tissue Stem Module DACT1 Negative Regulation of Canonical Wnt -2 TGFBI IL11 01 234 DLC1 LTBP2 - Log 10 (q-value) ALCAM PLAUR PTHLH AMOTL1 c d Breast cancer cell lines Claudin-low cell lines 0.6 0.5 0.4 0.4 0.3 0.2 0.2 0.0 0.1 -0.2 0.0 -0.4 -0.1 Fig. 3 αvβ3 CSCs express unique genes associated with aggressive breast cancers and normal mammary stem cells. a Gene set Low + − enrichment analysis (GSEA) for EpCAM /αvβ3 versus αvβ3 cells. Statistics by Student’s t-test with Benjamini−Hochberg test. Dashed line Low + indicates statistical significance. b Heat map of the 20 most differentially expressed genes upregulated in EpCAM /αvβ3 cells and found in the gene sets in (a) (Log2 scale). c Box and whiskers plot showing the relative enrichment for the αvβ3 CSC gene signature in cell lines representing different molecular subtypes. Boxplots represent medians (center line) and interquartile range (IQR; box), and whiskers represent the maximum and minimum values within 1.5 times the IQR from the edge of the box. Statistics by ANOVA with Tukey’s test. *P < 0.05. Cell lines in each category: Luminal B; n = 7, HER2; n= 7, Basal-like; n = 6, Claudin-low; n = 8. d Claudin-low cells lines with significant enrichment for the αvβ3 CSC gene signature compared to parental HCC38 cells (dashed line). a–d n = 3 independent experiments. See also Supplementary Fig. 3. identified (MDA-MB-231, BT549, and Hs578T) represent some of surrogate for our αvβ3 CSCs, as characterized in Fig. 3d. Tumor- the most widely used breast cancer cell lines due to their high sphere formation in methylcellulose wasselectedasour primary tumorigenicity and metastatic potential, highlighting the potential endpoint since it is a critical stemness property. Results from these studies identified two genes as highly relevant for further study: significance of our candidate genes for aggressive disease. These TGFBI (Transforming Growth Factor Beta Induced; initially termed BIG- findings serve to distinguish αvβ3 CSCs as a unique cancer cell H3) and ZEB1 (Fig. 4b and Supplementary Fig. 4a). We rigorously type that is not synonymous with the claudin-low classification validated these targets by showing ZEB1 protein enrichment in and identify appropriate surrogate cell lines in which to screen our sorted αvβ3 CSCs from HCC38 cells (Fig. 4c). Additionally, we used candidate genes for their role in stemness. another heterogeneous cell line (SUM149) to show conserved Low + expression of both TGFBI and ZEB1 specifically in EpCAM /αvβ3 Characterization of ZEB1 and TGFBI as candidate genes cells (Fig. 4d, e and Supplementary Fig. 4b). Importantly, ZEB1 protein required for stemness levels were specifically associated with the surrogate cell lines We next wished to perform an unbiased assessment of the key genes enriched for the αvβ3 CSC gene signature (Supplementary Fig. 4c), as and signaling pathways responsible for the more aggressive nature 20 well as the LM2-4 metastatic variant of the MDA-MB-231 cell line of αvβ3 CSCs. Here, we used αvβ3 as a marker of activated stem-like (Supplementary Fig. 4d), all of which we previously showed to cells, with candidate genes selected without regard for a potential express αvβ3 . direct link to αvβ3 signaling. To identify candidate genes, we A secreted ECM protein, TGFBI (Transforming Growth Factor Beta performed QPCR analysis of the 20 DEGs from Fig. 3b to select those Induced; BIG-H3) has been shown to paradoxically enhance 21 22 that displayed the most consistent and robust expression in αvβ3 anchorage-independence , similar to our findings with αvβ3 . CSCs relative to the other three cell types (Fig. 4a). This identified six While best known for its role in epithelial-mesenchymal transforma- genes for further analysis (Fig. 4a). We then performed an siRNA tion (EMT) , the transcription factor ZEB1 also mediates non-EMT functional screen to identify which of these genes was most critical functions that may be more important for its role in tumor 24,25 for αvβ3 CSCs (Fig. 4b and Supplementary Fig. 4a). To simulta- progression . In fact, recent studies unexpectedly found ZEB1 in neously examine the function of multiple genes after transient siRNA a subset of basal/stem cells in normal human mammary glands knockdown we employed the BT549 cell line as an appropriate where it promotes oncogene-induced transformation . While the npj Breast Cancer (2022) 5 Published in partnership with the Breast Cancer Research Foundation Luminal B Her2 Basal-like Claudin-low HCC1395 MDA-436 SUM159 SUM1315 HCC38 MDA-231 BT549 Hs578T v3 CSC Score v3 CSC Score Q. Sun et al. a b QPCR Validation Functional Screen (BT549) Low - EpCAM /v3 10 35 Low + EpCAM /v3 8 High - EpCAM /v3 High + EpCAM /v3 20 * 0 0 c e SUM149 cells 91.1% 5.71% Low - EpCAM /v3 Low High EpCAM EpCAM Low + EpCAM /v3 250 4 v3: - + - + High - EpCAM /v3 3 High + ZEB1 EpCAM /v3 -actin 2.97% 0.22% 1 2 3 4 10 10 10 10 v3 ZEB1 TGFBI Fig. 4 Identification of TGFBI and ZEB1 as candidate genes unique to αvβ3 CSCs. a QPCR validation of candidate genes in sorted HCC38 cell types. b Functional screen for candidate genes necessary for methylcellulose colony formation after transient siRNA knockdown in BT549 cells. Target gene knockdown was validated by QPCR. Statistics by one-way ANOVA with Dunnett’s test. *P < 0.05. c Representative immunoblot of lysates from sorted HCC38 cells. β-actin is shown as a loading control. Molecular weight markers are indicated in kilodaltons. d Representative FACS density plot of the live, CD49f SUM149 cells according to their cell surface EpCAM and αvβ3 status. e QPCR validation of candidate genes in sorted SUM149 cells. a, e Samples were run in duplicate with GAPDH as a loading control. Expression is shown relative Low − to the EpCAM /αvβ3 cells (dashed lines). a, b, e Data represent the mean ± s.e.m. a–e n = 3 independent experiments. See also Supplementary Fig. 4. exact identity and function of these cells is still a mystery, it suggests αvβ3 CSCs identified by a rigorous, unbiased and systematic they may be similar to ZEB1 breast cancer cells. approach. Discovery of TGFBI-ZEB1 as a key stemness-related signaling TGFBI-ZEB1 promotes chromosomal stability and resistance to module PARP inhibition In order to further assess the relevance of these two genes for We next considered the cell biological basis for these effects on stemness properties, we generated TGFBI and ZEB1 knockout cells tumorsphere formation. Our GSEA results suggest that the ability using CRISPR/Cas9 in our surrogate stem-like LM2-4 and BT549 cell to respond to cellular stress was a distinguishing feature of αvβ3 lines. We further validated these cells, showing significantly CSCs. In fact, while TGFBI and ZEB1 have diverse cellular functions, reduced ZEB1 protein levels (Fig. 5a) as well as decreased they may play a common role in reducing a certain type of genetic 24,27 amounts of TGFBI mRNA expression (Supplementary Fig. 5a) and stress called chromosomal instability (CIN) . CIN is a hallmark of secreted TGFBI protein (Supplementary Fig. 5b). Surprisingly, these cancer and an important stress in cancer cells that limits validation studies showed that TGFBI deletion also resulted in transformation . In fact, a recent study showed that normal adult decreased levels of ZEB1 protein (Fig. 5a). Importantly, deletion of MaSCs were inherently more tumorigenic due to suppression of ZEB1 did not decrease levels of TGFBI mRNA (Supplementary Fig. CIN via ZEB1 . Our independent discovery of ZEB1 as one of the 5c). We also noted that deleting either gene had no effect on most DEG in αvβ3 CSCs, suggested that it may play a similar role protein levels of the β3 subunit (Supplementary Fig. 5d). These in these cells. unexpected findings suggest that these two independently Using multiple methods, we now show that CRISPR knockout of identified candidate genes are linked within the same pathway ZEB1 or TGFBI enhances CIN in stem-like cell lines. We began by To examine this possibility and validate their role in stemness, measuring staining for the DNA damage marker phospho-γH2AX we tested our TGFBI and ZEB1 CRISPR knockout cells in assays of and found that DNA strand breaks increased in our knockout cells primary tumorsphere formation and self-renewal (Fig. 5b). (Fig. 6a, b). In contrast, there was no effect on proliferation as Deletion of either TGFBI or ZEB1 resulted in an approximately assessed by incorporation of fluorescently-labeled EdU (Fig. 6a, c) 50% decrease in primary tumorspheres, while subsequent self- or apoptosis measured by PARP cleavage (Supplementary Fig. 6a). renewal assays showed an almost 75% decrease due to ZEB1 Since staining with phospho-γH2AX indicates the presence of DNA knockout in BT549 cells (Fig. 5b). These findings highlight an strand breaks that could lead to missegregation of chromosomes, important role for these genes in stemness and are consistent we quantified micronuclei as a direct measure of CIN and with their function within the same pathway. Indeed, we show observed increased levels associated with both knockout cell that adding recombinant human TGFBI protein (rhTGFBI) is lines (Fig. 6d). To robustly examine differences in CIN we also sufficient to drive ZEB1 protein expression in control and TGFBI evaluated potential copy number alterations (CNA) and found knockout cells (Fig. 5c) and specifically rescue defective tumor- higher levels in our knockout cells (Fig. 6e). Analysis of data from sphere formation caused by TGFBI deletion (Fig. 5d). Our findings The Cancer Genome Atlas further supported these findings by highlight a potential new TGFBI-ZEB1 signaling module specificto showing that high ZEB1 expression in tumors corresponded with Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 5 SUM149 cells ZEB1 WNT5A DACT1 TGFBI ALCAM TSPAN5 siControl siZEB1 siWNT5A siDACT1 siTGFBI siALCAM siTSPAN5 mRNA Levels (Fold Change) EpCAM Colonies mRNA Levels (Fold Change) Q. Sun et al. a b Primary Tumorspheres Secondary Tumorspheres LM2-4 BT549 50 * ZEB1 TGFBI ZEB1 TGFBI * * * 25 Ctrl KO KO Ctrl KO KO ZEB1 30 -actin 37 10 0 0 LM2-4 BT549 LM2-4 BT549 Ctrl TGFBI KO ZEB1 KO c d Rescue w/ rhTGFBI BT549 Ctrl + Vehicle * * Ctrl + rhTGFBI Ctrl TGFBI KO rhTGFBI TGFBI KO + Vehicle 30 * * (ng/mL) : - 20 100 500 - 20 100 500 TGFBI KO + rhTGFBI ZEB1 -actin LM2-4 BT549 Fig. 5 Discovery of a TGFBI-ZEB1 signaling module required for stemness. a Representative immunoblot of lysates from TGFBI and ZEB1 CRISPR knockout cells. b Tumorsphere assays in methylcellulose to assess the effects of TGFBI or ZEB1 CRISPR knockout on primary colonies (left) or self-renewal (right) in the indicated cell lines. c A representative immunoblot experiment, showing rescue of ZEB1 protein levels after treatment with rhTGFBI for 24 h with the indicated doses compared to vehicle control (PBS). a, c β-actin is shown as a loading control. Molecular weight markers are indicated in kilodaltons. d Rescue of primary colony formation in TGFBI knockout cells with 500 ng/mL rhTGFBI. Data represent the mean ± s.e.m. Statistics by one-way (b) or two-way (d) ANOVA with Dunnett’s(b) or Tukey’s test (d). *P < 0.05. n = 3(a–c)or n =5(d) independent experiments. See also Supplementary Fig. 5. low levels of CIN (Supplementary Fig. 6b). Thus, TGFBI-ZEB1 now show a striking correlation between αvβ3 CSCs and signaling appears to promote stemness in αvβ3 CSCs by activated stem/basal cells from the normal human mammary suppressing the endogenous genetic stress caused by CIN. gland. We further identified key genes associated with this cell Tumor cells with defective DNA repair due to BRCA mutations state including the secreted matrix protein TGFBI (BIG-H3) and the are highly sensitive to PARP inhibition due to an accumulation of transcription factor ZEB1. Taken together, our findings suggest double-strand breaks that tips the balance toward cell death . that these genes may operate as a TGFBI-ZEB1 signaling module Since we observed increased DNA strand breaks in our TGFBI and to promote stemness by protecting against genetic stress, such as ZEB1 CRISPR knockout cells (Fig. 6a, b), we hypothesized that CIN. In fact, downregulating TGFBI-ZEB1 sensitized αvβ3 CSCs to some of these may fail to be repaired, possibly leading to synergy PARP inhibition, laying the foundation for a potential new with PARP inhibitors such as Olaparib (Lynparza). While Olaparib is treatment strategy to reduce breast cancer progression. clinically-approved for BRCA-mutant breast and ovarian cancers, Unbiased analysis of critical genes and pathways in αvβ3 CSCs led to our surprising discovery of a TGFBI-ZEB1 signaling module. non-BRCA mutant cancers are completely refractory to this treatment . Indeed, we show that while deletion of either TGFBI While TGFBI is a secreted ECM protein that would normally bind to or ZEB1 had no effect on 2D cell viability (Supplementary Fig. 6c), integrins and elicit adhesion-dependent responses, we and others knockout of either gene synergized with Olaparib in two different have now shown that it can also enhance anchorage-independent stem-like cell lines (Fig. 6f and Supplementary Fig. 6d). Signifi- growth . This is similar to our surprising finding that the integrin αvβ3 also promotes anchorage-independence , and suggests the cantly, we observed sensitivity to Olaparib at similar doses that are two may function as a possible ligand-receptor pair in stem-like effective against a BRCA-mutant cell line (Supplementary Fig. 6e). These data are consistent with an important role for TGFBI-ZEB1 in cells. The transcription factor ZEB1 is perhaps best known for its reducing CIN in αvβ3 CSCs, highlighting the ability to control role in EMT; however, it is also important for functions not related 24,25 genetic stress as a critical attribute of stem-like cells. Additionally, to EMT that may be even more critical for tumor progression . our findings suggest that PARP inhibition may be an effective An unexpected result of gene atlas studies from the normal human mammary gland was the discovery of a subset of stem/ precision therapy for more than just BRCA-mutant disease, and + 26 basal cells expressing ZEB1 . Notably, these cells were specificto that a similar approach may be able to eliminate αvβ3 CSCs and reduce breast cancer progression. human glands and not observed in mice . Further corroborating this finding, a different study identified ZEB1 expression in enriched populations of human MaSCs, where it surprisingly DISCUSSION functioned to promote oncogene-induced transformation . Thus, While stem cells in the adult mammary gland are dynamic, cycling in the normal mammary gland, ZEB1 is expressed in cells that through active and inactive cell signaling states due to hormonal display stem cell properties. While there is still much to learn 10,30,31 signaling , it is unclear if stem-like tumor cells possess a about these cells, our new findings suggest they may represent similar ability. Our prior work identified the integrin αvβ3as a MaSCs in the activated state and display traits similar to ZEB1 surface marker of activated stem cells in the adult mammary breast cancer cells. Together, our results highlight a potential new gland, suggesting that tumor cells expressing this marker may TGFBI-ZEB1 pathway specificto αvβ3 CSCs that we identified feature a similar activated signaling state. By comparing our whole through a rigorous, unbiased and systematic approach. transcriptome sequencing data from sorted αvβ3 CSCs with While CIN is a hallmark of cancer cells, too much may act to limit published gene sets enriched in stem/basal cells during the luteal tumor progression . In fact, normal cell types that can better (Active) and follicular (Inactive) phases of the menstrual cycle, we tolerate CIN, such as MaSCs are much more likely to undergo npj Breast Cancer (2022) 5 Published in partnership with the Breast Cancer Research Foundation Colonies Colonies Colonies Q. Sun et al. a b LM2-4 DNA strand breaks 35 * * Ctrl Ctrl TGFBI KO ZEB1 KO * * TGFBI KO ZEB1 KO LM2-4 BT549 Proliferation Ctrl n.s. TGFBI KO n.s. ZEB1 KO LM2-4 BT549 d e Micronuclei TGFBI KO adj. p<0.01 8 16 n.s. * Ctrl * TGFBI KO 6 12 ZEB1 KO 2 4 0 0 LM2-4 BT549 -0.25 0 0.25 0.5 Log2 Ratio CNA (KO vs Ctrl) BT549 ZEB1 KO adj. p<0.01 * * 3 n.s. Ctrl TGFBI KO 1 ZEB1 KO -0.25 0 0.25 0.5 0.1 1 10 [Olaparib (M)] Log2 Ratio CNA (KO vs Ctrl) Fig. 6 Decreased TGFBI-ZEB1 signaling enhances chromosomal instability and sensitivity to PARP inhibition. a Representative immunofluorescent staining for the DNA damage marker phospho-γH2AX (red) or detection of fluorescently-labeled EdU (red) after 90 min incubation to assess cell proliferation. Nuclei are stained blue in all images. Scale bars, 40 μm. Percentage of cells positive for p-γH2AX (b), EdU (c), or micronuclei (d) relative to total nuclei. Data calculated from four random fields per condition for each experiment. e Volcano plots depicting the copy number alterations (CNA) in BT549 TGFBI or ZEB1 knockout cells relative to controls. Statistics performed by Student’s t-test corrected for multiple testing using the Benjamini-Hochberg method. Red dots represent the 112 (TGFBI KO) and 129 (ZEB1 KO) segments with statistically significant differences (adjusted p-value < 0.01) out of 459 total segments examined. Black dots are not significant (n.s.). f XTT cell viability assays comparing Olaparib sensitivity in the indicated BT549 cell types. Curves are plotted relative to vehicle controls for each group and fitted by non-linear regression. *P < 0.05 for 0.5, 0.75, 1, 2, 5, and 10 μM Olaparib (TGFBI or ZEB1 KO versus control). n = 3 (a–d, f)or n =4(e) independent experiments. Data represent the mean ± s.e.m. Statistics by one-way (b−d) or two-way (f) ANOVA with Dunnett’s test. *P < 0.05. n.s.= not significant. See also Supplementary Fig. 6. oncogenic transformation , suggesting that control of genetic consistent with the enhanced metabolic activity identified in stress is an important attribute of more aggressive tumor cells. stem-like breast cancer cells found in metastases . Thus, our While TGFBI and ZEB1 have diverse cellular functions, they may investigation of the biological effects of TGFBI-ZEB1 support a role 24,27 play a common role in reducing CIN . In fact, a recent study in limiting the effects of genetic stress and maintaining showed that MaSCs were inherently more tumorigenic due to chromosomal stability, suggesting this may be a defining attribute suppression of CIN via ZEB1 . Our independent discovery of ZEB1 and potential vulnerability of stem-like cells. as one of the most differentially-expressed genes in αvβ3 CSCs, PARP inhibitors are clinically-approved and highly effective suggested that it may play a similar role. We now show that treatments for BRCA-mutant breast and ovarian cancers . Tumor CRISPR knockout of either ZEB1 or TGFBI increased CIN in two cells with defects in DNA double-strand break repair, such as BRCA stem-like cell lines with no effect on cell proliferation or survival. mutations, are more sensitive to PARP inhibitors, such as Olaparib, This genetic instability may be caused by endogenous factors which prevent single-strand break repair and drive further genetic such as cell replication or the production of reactive oxygen instability, resulting in cell death. However, while PARP inhibition species during metabolism. The latter of which would be is an effective treatment against BRCA-mutant breast and ovarian Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 5 % Micronuclei-positive cells EdU p-H2AX Viable cells (% Control) % Micronuclei-positive cells -Log10 (Adjusted P-value) -Log10 (Adjusted P-value) % EdU-positive cells % p-H2AX-positive cells Q. Sun et al. cancers, non-BRCA mutant cancers are completely refractory to Bulk RNA-Seq this treatment . We now provide proof-of-concept that disrupting After sorting HCC38 cells, RNA was purified from an equal number of cells per experiment for each cell type (approximately 50–70,000 cells) using a TGFBI-ZEB1 signaling not only increased CIN, but functioned much RNeasy Mini Kit (Qiagen). The samples were then submitted to the IGM like a BRCA mutation by increasing sensitivity to PARP inhibition. Genomics Core at UCSD for validation of RNA quality and sequencing was We hypothesized that this may lead to synergy with Olaparib, performed on a HiSeq 4000 (Illumina, San Diego, CA, USA). which normally affects only BRCA-mutant cancers. Indeed, our new data shows that TGFBI or ZEB1 deletion enhances “BRCA- Bioinformatics analysis ness”, and synergizes with Olaparib. Our findings suggest that RNA-seq data were analyzed with a pipeline implemented in the disrupting key mediators of chromosomal stability in αvβ3 CSCs, BCBio-nextgen workflow manager https://zenodo.org/record/4686097#. such as TGFBI and ZEB1, can sensitize these normally resistant YRLzj4hKiM8. Briefly, we aligned reads to GRch37 reference genome using cells to treatment with clinically-approved PARP inhibitors. There- 33 34 STAR and quantified expression levels with Salmon 0.13.1 . We then fore, these findings represent a crucial initial step laying the annotated genes with BioMART , keeping only protein coding genes with foundation for further study of the activated stem cell state as a more than one read count for analysis. DEG was then identified using key contributor to recurrence and metastasis in patient disease DESEQ2 . Gene Set Enrichment analysis was performed using R package LIGER on the Hallmark and Reactome gene sets available in MSigDB . The and outline a potential therapeutic strategy for targeting 20 gene signature was derived by overlap analysis between the two these cells. differential gene expression analyses followed by manual curation for gene set membership. The public gene expression profile from breast cancer cell lines was obtained at NCBI GEO (GSE50470) and the corresponding MATERIALS AND METHODS intrinsic subtype information obtained from Prat et al. . The CSC signature Cell lines score was calculated according to Barbie et al. implemented in the The following breast cancer cell lines were purchased from ATCC gseapy python package (v0.9.8). (Manassas, VA, USA): HCC38, MDA-MB-436, MCF-7, T47D, BT474, MDA- MB-468, BT-20, HCC1187, Hs578T, BT549, and MDA-MB-231. LM2-4 cells, a Real-time qPCR highly metastatic variant of the MDA-MB-231 cell line was a gift from qPCR experiments on cultured cells were performed by collecting total Robert Kerbel. All cell lines were tested and shown to be free of RNA using the RNeasy Mini Kit (Qiagen) and reverse transcribing with the mycoplasma. The HCC38, BT549, and LM2-4 cells were further authenti- High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Thermo cated by short tandem repeat (STR) testing. Cells used in mice were Fisher Scientific). Relative mRNA levels from sorted cells were examined additionally tested and found to be negative for an extensive panel of using the Cells-to-CT kit (Life Technologies) according to the manufac- mouse pathogens. Cell lines were cultured in complete DMEM medium turer’s instructions. Lysates were prepared from 90,000 freshly sorted (DMEM supplemented with 10% fetal bovine serum (FBS) + 1% L- HCC38 cells. Real-time qPCR was performed using iTaq Universal SYBR glutamine, sodium pyruvate, non-essential amino acids, and antibiotic/ Green Supermix (Bio-Rad, Hercules, CA, USA) and run on a LightCycler 480 antimycotic). qPCR System (Roche, Basel, Switzerland). See Supplementary Methods for a list of primers. Cell transfection and lentiviral transduction Plasmids containing enhanced specificity Cas9 and the appropriate guide Immunoblotting RNA’s in the pLentiCRISPRv2 vector were purchased from GenScript Whole cell lysates were prepared from cell lines with RIPA lysis buffer (Piscataway, NJ, USA) for generating stable knockout with lentivirus and (100 mM Tris pH 7.5, 150 mM sodium chloride, 0.1% deoxycholate, 0.1% selected using puromycin. Transient transfections for all CRISPR/Cas9 SDS, 50 mM NaF, Protease inhibitor cocktail (Roche), 2 mM PMSF, 2 mM vectors into 293T cells were performed with Lipofectamine 3000 sodium orthovanadate) combined with scraping and the lysates cleared by (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA), while HiPerFect centrifugation. Standard Western blotting procedures were performed. The following primary antibodies were used for immunoblotting at a dilution of Transfection Reagent (Qiagen, Hilden, Germany) was used for siRNAs. All 1:1000: ZEB1 (3396, Cell Signaling Technology, Danvers, MA, USA), Full- transfections were performed according to the manufacturers’ instructions. length PARP (9532, Cell Signaling Technology), Hsp90 (sc-13119, Santa FlexiTube siRNAs (Qiagen) included AllStars negative control, ZEB1 Cruz, Dallas, TX, USA) and β-actin (MABT825, MilliporeSigma). Treatments (SI04339587), WNT5A (SI04384184), DACT1 (SI00359275), TGFBI with rhTGFBI (R&D Systems, Minneapolis, MN, USA) or vehicle control (PBS) (SI02780722), ALCAM (SI02780155), and TSPAN5 (SI04151665). were performed at the specified doses for 24 h prior to harvesting lysates. All blots or gels derive from the same experiment and were processed in Generation of CRISPR knockout cells parallel. See Supplementary Information for unedited blots (Supplemen- Stable knockout of select genes was achieved by transducing BT549 or tary Fig. 8a–c). LM2-4 cells with lentivirus vectors expressing enhanced specificity Cas9 and the appropriate guide RNA’s targeting human TGFBI or ZEB1 TGFBI ELISA (GenScript) and pooling puromycin-resistant cells. A vector lacking the Conditioned media from BT549 and LM2-4 cell lines was collected after guide RNA was used as a negative control. Successful knockout of the 48 h in phenol-free complete DMEM culture medium. Secreted TGFBI levels respective targets was verified by Western blot for ZEB1 and Real-time were then quantified with the Human TGFBI (BIGH3) ELISA kit (Invitrogen, QPCR for TGFBI. Thermo Fisher Scientific) according to the manufacturer’s instructions. Conditioned media was diluted 1:50 (BT549) or 1:200 (LM2-4) in order to fit on the standard curve. Flow cytometry Single-cell suspensions were prepared from cultured HCC38 or SUM149 cells, blocked in 0.5% BSA/PBS, and stained with the following antibodies Immunofluorescence staining prior to sorting: CD49f-PE 1:10 (555736, GoH3; BD Biosciences, San Jose, Immunofluorescence staining was performed on cultured cells in a 4-well CA, USA); EpCAM–Alexa 647 1:20 (324212, 9C4; BioLegend, San Diego, CA, chamberslide (Nunc, Thermo Fisher Scientific) fixed brieflyin2% USA); and αvβ3–biotin 1:40 (MAB1976B, LM609; MilliporeSigma, Burling- paraformaldehyde in PBS at room temperature and permeabilized. All ton, MA, USA) and Streptavidin-Brilliant Violet 421 1:80 (BioLegend). samples were blocked with 1:80 normal goat serum in 0.1% BSA/PBS Propidium iodide solution (0.5 μg/ml) was used to detect dead cells. Viable before incubation in phospho-γH2AX primary antibody diluted 1:100 cells were collected by sorting with a FACSDiva or FACSAria machine (BD (9718, Cell Signaling Technology) overnight at 4 °C, followed by incubation Biosciences). In some cases, differentiation assays were performed by with DAPI and 1:500 Alexa Fluor 647-conjugated secondary antibody culturing sorted cells for exactly 10 passages (6–7 weeks after sorting) (A32733, Invitrogen, Thermo Fisher Scientific) for 2 h at RT. Alternatively, before re-analyzing by flow cytometry. See Supplementary Information for proliferation was measured by incubating live cells with Edu for 2 h at 37 °C gating strategies (Supplementary Fig. 7a–c). prior to labeling the Edu with Alexa Fluor 647 according to manufacturer’s npj Breast Cancer (2022) 5 Published in partnership with the Breast Cancer Research Foundation Q. Sun et al. instructions (Invitrogen, Thermo Fisher Scientific). DAPI was used to Statistics visualize nuclei. Cells stained by either method were imaged with a Nikon Data presentation and statistical tests are indicated in the figure legends. A1R confocal microscope and images captured from four randomly Two-tailed Student’s t-tests were used for comparing two means while selected fields with a 60× objective. Micronuclei were then manually ANOVA was performed for 3 or more data sets. Post hoc analysis was counted for each field or the percent positive cells quantified using Image performed using an appropriate multiple comparisons test as indicated in the legends. For all analyses, P < 0.05 was considered statistically J software. significant. Statistical analysis was performed using GraphPad Prism software (San Diego, CA, USA). Cell viability See Supplementary Material for additional methods. XTT cell viability assays were performed by first seeding cells into a 96-well tissue culture plate at the following density per well: LM2-4 (4,000), BT549 Reporting summary (2,000), HCC38 (4,000), or MDA-MB-436 (7,500). After cells attached Further information on research design is available in the Nature Research overnight the indicated concentrations of Olaparib (Lynparza) (Sell- Reporting Summary linked to this article. eckchem, Houston, TX, USA) or vehicle alone (DMSO) were then added to the wells in 100 μL phenol-free complete DMEM medium. After 48 h, XTT substrate (Thermo Fisher Scientific) was added to the wells and DATA AVAILABILITY incubated for 2 h before reading the A450 nm on a plate reader. Cell RNA sequencing data that support the findings of this study have been deposited in viability for each of the indicated treatments was expressed as a percent of the NCBI Sequence Read Archive (SRA) with the primary accession code the vehicle control wells. PRJNA750073. Copy number alterations Received: 14 April 2021; Accepted: 6 December 2021; Genomic DNA was prepared from 5 × 10 cells with the DNeasy Blood and Tissue Kit (Qiagen) and submitted to the IGM Genomics Core at UCSD. CNA was determined by hybridizing 200 ng of DNA per sample onto an Infinium CoreExome-24 Array (Illumina) and analyzing with Nexus CN software (version 7.5). The log2 ratio of signal intensity were calculated for each SNP REFERENCES on the array and copy number segments identified using the Bioconductor 1. Al-Hajj, M., Wicha, M. S., Benito-Hernandez, A., Morrison, S. J. & Clarke, M. F. copynumber package , including winsorization to remove outliers, Prospective identification of tumorigenic breast cancer cells. Proc. Natl Acad. Sci. imputation of missing value, and multi-sample copy number segmentation USA 100, 3983–3988 (2003). using multipcf. The differences in log2 ratio of all larger segments (more 2. Lim, E. et al. Aberrant luminal progenitors as the candidate target population for than 10 kb) were compared between control and KO BT549 cells. basal tumor development in BRCA1 mutation carriers. Nat. Med. 15, 907–913 (2009). 3. Prat, A. & Perou, C. M. Deconstructing the molecular portraits of breast cancer. Tumorsphere assays Mol. Oncol. 5,5–23 (2011). Primary tumorsphere formation was assessed in cells grown under 4. Malanchi, I. et al. Interactions between cancer stem cells and their niche govern anchorage-independent conditions in methylcellulose. BT549 (15,000) or metastatic colonization. Nature 481,85–89 (2011). LM2-4 cells (10,000) were cultured in 0.9 mL of 1% methylcellulose/ 5. Cai, S. et al. A quiescent Bcl11b high stem cell population is required for main- complete DMEM medium in ultra-low adhesion 24-well dishes (Corning, tenance of the mammary gland. Cell Stem Cell 20, 247–260 (2017). e5. Corning, NY, USA) and cells cultured for 10−12 days. Primary tumorspheres 6. Fu, N. Y. et al. Identification of quiescent and spatially restricted mammary stem were assessed by counting colonies consisting of at least 6 cells from 4 cells that are hormone responsive. Nat. Cell Biol. 19, 164–176 (2017). fields per well with a 10× objective. We measured self-renewal by 7. Fu, N. Y. et al. Foxp1 is indispensable for ductal morphogenesis and controls the exit of mammary stem cells from quiescence. Dev. Cell 47, 629–644 (2018). e8. collecting primary tumorspheres by dilution in at least 3 volumes of PBS, 8. Prater, M. D. et al. Mammary stem cells have myoepithelial cell properties. Nat. dissociating them with trypsin for approximately 10 min, and re-seeding in Cell Biol. 16, 942–950 (2014). 1-7. 1% methylcellulose before evaluating secondary colonies after 10 days. For 9. Asselin-Labat, M. L. et al. Steroid hormone receptor status of mouse mammary treatment with rhTGFBI (R&D Systems), a single dose of 500 ng/mL was stem cells. J. Natl Cancer Inst. 98, 1011–1014 (2006). added only once, when initially embedding the cells, and compared 10. Desgrosellier, J. S. et al. Integrin alphavbeta3 drives slug activation and stemness against cells receiving the same volume of vehicle (PBS). in the pregnant and neoplastic mammary gland. Dev. Cell 30, 295–308 (2014). 11. Sun, Q. et al. Proapoptotic PUMA targets stem-like breast cancer cells to suppress Orthotopic breast cancer metastasis. J. Clin. Invest. 128, 531–544 (2018). 12. Prat, A. et al. Phenotypic and molecular characterization of the claudin-low Tumors were generated by injection of a limiting number of sorted HCC38 intrinsic subtype of breast cancer. Breast Cancer Res. 12, R68 (2010). cells diluted in 50 μl sterile PBS and injected into the inguinal fat pads of 8- 13. Prat, A. et al. Characterization of cell lines derived from breast cancers and normal to 10-week-old adult female nonobese diabetic/severe combined immu- mammary tissues for the study of the intrinsic molecular subtypes. Breast Cancer nodeficiency/interleukin-2 receptor γ chain knockout (NSG) mice (pur- Res. Treat. 142, 237–255 (2013). chased from UCSD Animal Care Program colony). All mice were monitored 14. Creighton, C. J. et al. Residual breast cancers after conventional therapy display weekly for tumor formation by gentle palpation. Most tumors formed mesenchymal as well as tumor-initiating features. Proc. Natl Acad. Sci. USA 106, within 8−10 weeks. Tumor volume was measured with calipers twice 13820–13825 (2009). weekly by a blinded observer. The experiment was concluded, and mice 15. Shipitsin, M. et al. Molecular definition of breast tumor heterogeneity. Cancer Cell were sacrificed just prior to the tumors reaching the maximum allowable 11, 259–273 (2007). size of 2 cm or 14 weeks, whichever came first. Estimated tumor-initiating 16. Lawson, D. A. et al. Single-cell analysis reveals a stem-cell program in human cell frequencies were calculated with the Extreme Limiting Dilution (ELDA) metastatic breast cancer cells. Nature 526, 131–135 (2015). web-based tool http://bioinf.wehi.edu.au/software/elda/. Primary tumor 17. Asselin-Labat, M. L. et al. Gata-3 is an essential regulator of mammary-gland mass was determined by assessing the wet weight of the resected tumors. morphogenesis and luminal-cell differentiation. Nat. Cell Biol. 9, 201–209 (2007). The remaining tumor-free mice were harvested at 14 weeks and the 18. Lim, E. et al. Transcriptome analyses of mouse and human mammary cell sub- populations reveal multiple conserved genes and pathways. Breast Cancer Res. absence of any detectable tumor was confirmed by whole-mount staining. 12, R21 (2010). 19. Murrow L. M. et al. Changes in epithelial proportions and transcriptional state Ethics underlie major premenopausal breast cancer risks. Preprint at bioRxiv https://doi. All mouse studies described were approved by the UCSD Institutional org/10.1101/430611 (2020). Animal Care and Use Committee (IACUC) and performed in accordance 20. Munoz, R. et al. Highly efficacious nontoxic preclinical treatment for advanced metastatic breast cancer using combination oral UFT-cyclophosphamide metro- with the guidelines set forth in the NIH’s Guide for the Care and Use of nomic chemotherapy. Cancer Res. 66, 3386–3391 (2006). Laboratory Animals (National Academies Press, 2011). Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 5 Q. Sun et al. 21. Nummela, P. et al. Transforming growth factor beta-induced (TGFBI) is an anti- Program (TRDRP) High Impact Research Project Award 28IR-0054 (to J. S. D.). We would adhesive protein regulating the invasive growth of melanoma cells. Am. J. Pathol. also like to acknowledge the flow cytometry shared resource facility in the UCSD Moores 180, 1663–1674 (2012). Cancer Center. Bulk RNA-Seq was conducted at the IGM Genomics Center, University of 22. Desgrosellier, J. S. et al. An integrin alpha(v)beta(3)-c-Src oncogenic unit promotes California, San Diego, La Jolla, CA (MCC grant # P30CA023100). anchorage-independence and tumor progression. Nat. Med. 15,1163–1169 (2009). 23. Ye, X. et al. Distinct EMT programs control normal mammary stem cells and tumour-initiating cells. Nature 525, 256–260 (2015). AUTHOR CONTRIBUTIONS 24. Morel, A. P. et al. A stemness-related ZEB1-MSRB3 axis governs cellular pliancy Q.S. helped design the project, performed experiments, analyzed data, and and breast cancer genome stability. Nat. Med. 23, 568–578 (2017). generated figures. Y.W., B.P., and G.L. designed and conducted experiments and 25. Maturi, V., Enroth, S., Heldin, C. H. & Moustakas, A. Genome-wide binding of analyzed data. O.H. and A.O. analyzed the RNA-Seq data and generated figures. J.S.D. transcription factor ZEB1 in triple-negative breast cancer cells. J. Cell Physiol. 233, designed the project, analyzed the data, supervised the overall project, and wrote the 7113–7127 (2018). manuscript. 26. Nguyen, Q. H. et al. Profiling human breast epithelial cells using single cell RNA sequencing identifies cell diversity. Nat. Commun. 9, 2028 (2018). 27. Zhang, Y. et al. TGFBI deficiency predisposes mice to spontaneous tumor COMPETING INTERESTS development. Cancer Res. 69,37–44 (2009). The authors declare no competing interests. 28. Cleary, J. M., Aguirre, A. J., Shapiro, G. I. & D’Andrea, A. D. Biomarker-guided development of DNA repair inhibitors. Mol. Cell 78, 1070–1085 (2020). 29. Noordermeer, S. M. & van Attikum, H. PARP inhibitor resistance: A tug-of-war in ADDITIONAL INFORMATION BRCA-mutated cells. Trends Cell Biol. 29, 820–834 (2019). 30. Joshi, P. A. et al. Progesterone induces adult mammary stem cell expansion. Supplementary information The online version contains supplementary material Nature 465, 803–807 (2010). available at https://doi.org/10.1038/s41523-021-00375-w. 31. Asselin-Labat, M. L. et al. Control of mammary stem cell function by steroid hormone signalling. Nature 465, 798–802 (2010). Correspondence and requests for materials should be addressed to Jay S. 32. Davis, R. T. et al. Transcriptional diversity and bioenergetic shift in human breast Desgrosellier. cancer metastasis revealed by single-cell RNA sequencing. Nat. Cell Biol. 22, 310–320 (2020). Reprints and permission information is available at http://www.nature.com/ 33. Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29,15–21 reprints (2013). 34. Patro, R., Duggal, G., Love, M. I., Irizarry, R. A. & Kingsford, C. Salmon provides fast and Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims bias-aware quantification of transcript expression. Nat. Methods 14, 417–419 (2017). in published maps and institutional affiliations. 35. Smedley, D. et al. BioMart−biological queries made easy. BMC Genomics 10, 22 (2009). 36. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014). 37. Liberzon, A. et al. The Molecular Signatures Database (MSigDB) hallmark gene set Open Access This article is licensed under a Creative Commons collection. Cell Syst. 1, 417–425 (2015). Attribution 4.0 International License, which permits use, sharing, 38. Barbie, D. A. et al. Systematic RNA interference reveals that oncogenic KRAS- adaptation, distribution and reproduction in any medium or format, as long as you give driven cancers require TBK1. Nature 462, 108–112 (2009). appropriate credit to the original author(s) and the source, provide a link to the Creative 39. Nilsen, G. et al. Copynumber: Efficient algorithms for single- and multi-track copy Commons license, and indicate if changes were made. The images or other third party number segmentation. BMC Genomics 13, 591 (2012). material in this article are included in the article’s Creative Commons license, unless 40. Hu, Y. & Smyth, G. K. ELDA: Extreme limiting dilution analysis for comparing indicated otherwise in a credit line to the material. If material is not included in the depleted and enriched populations in stem cell and other assays. J. Immunol. article’s Creative Commons license and your intended use is not permitted by statutory Methods 347,70–78 (2009). regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/. ACKNOWLEDGEMENTS This research was supported by funding from the National Natural Science Foundation of China Grants 82073271 and 81702362 (to Q. S.) and a Tobacco-Related Disease Research © The Author(s) 2022 npj Breast Cancer (2022) 5 Published in partnership with the Breast Cancer Research Foundation http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png npj Breast Cancer Springer Journals

Stem-like breast cancer cells in the activated state resist genetic stress via TGFBI-ZEB1

Loading next page...
 
/lp/springer-journals/stem-like-breast-cancer-cells-in-the-activated-state-resist-genetic-NI07AG0306
Publisher
Springer Journals
Copyright
Copyright © The Author(s) 2022
eISSN
2374-4677
DOI
10.1038/s41523-021-00375-w
Publisher site
See Article on Publisher Site

Abstract

www.nature.com/npjbcancer ARTICLE OPEN Stem-like breast cancer cells in the activated state resist genetic stress via TGFBI-ZEB1 1,2,4 1,2 2,3 1,2 1,2 2,3 1,2✉ Qi Sun , Yufen Wang , Adam Officer , Brianna Pecknold , Garrett Lee , Olivier Harismendy and Jay S. Desgrosellier Breast cancer cells with stem-like properties are critical for tumor progression, yet much about these cells remains unknown. Here, we characterize a population of stem-like breast cancer cells expressing the integrin αvβ3 as transcriptionally related to activated stem/basal cells in the normal human mammary gland. An unbiased functional screen of genes unique to these cells identified the matrix protein TGFBI (BIG-H3) and the transcription factor ZEB1 as necessary for tumorsphere formation. Surprisingly, these genes were not required for cell proliferation or survival, but instead maintained chromosomal stability. Consistent with this finding, CRISPR deletion of either gene synergized with PARP inhibition to deplete αvβ3 stem-like cells, which are normally resistant to this therapy. Our findings highlight a critical role for TGFBI-ZEB1 protection against genetic stress as a key attribute of activated stem- like cells and suggest that disrupting this ability may enhance their “BRCAness” by increasing sensitivity to PARP inhibitors. npj Breast Cancer (2022) 8:5 ; https://doi.org/10.1038/s41523-021-00375-w INTRODUCTION Despite the importance of CSCs for breast cancer progression, studies of these cells are limited by their scarcity and a lack of Tumor-initiating cancer stem cells (CSCs) bearing similarities to appropriate cell line models that reflect the heterogeneity in a adult mammary stem cells (MaSCs) are important contributors to 1–4 patient tumor. In the present study, we make use of our previously breast cancer progression and metastasis . However, adult characterized heterogeneous breast cancer cell line models to MaSCs are highly dynamic, frequently changing their cell state— overcome this limitation. These cell lines better recapitulate the a physiological condition due to altered gene expression or 2,15 intratumoral heterogeneity in patient disease , including a signaling—in response to hormonal cues. In fact, the mammary subset of αvβ3 CSCs, and allow us to directly assess a role for gland is one of the most dynamic organs in adult women, these cells compared to other neighboring tumor cell types. Based undergoing robust epithelial remodeling in response to hormones on our prior findings, we hypothesized that tumor cells bearing during the menstrual cycle and pregnancy that is driven by stem αvβ3 may similarly express genes found in stem/basal cells in cells. While normally quiescent, MaSCs respond to hormones response to hormonal signaling during the menstrual cycle or indirectly via paracrine signals to become active and contribute to pregnancy. Furthermore, since αvβ3 is a biomarker of aggressive 5–8 9 epithelial remodeling since they lack hormone receptors . cancer cells, we propose that these cells may contain unique These active stem cells exhibit enhanced proliferation and genes/pathways that could serve as potential vulnerabilities. To 5–8 migration , features that make this signaling state likely to be address these questions, we performed unbiased whole tran- hijacked by tumor cells. This raises the tantalizing question of scriptome analysis of αvβ3 CSCs. These findings represent an whether some of the most aggressive CSCs may further acquire initial step toward revealing similarities between these cells and properties associated with activated stem cells. normal mammary cell types and identifying key pathways that We previously showed that the cell surface receptor integrin may control their aggressive behavior. αvβ3 is a key switch turned-on by activated stem cells as they are mobilized for epithelial remodeling during pregnancy . Using αvβ3 as a marker, we further characterized a unique and RESULTS particularly aggressive population of stem-like breast cancer Surface αvβ3 marks stem-like cells enriched for tumor 11 + cells . Unexpectedly, we found αvβ3 CSCs in aggressive patient initiation tumors that were either estrogen receptor-positive (ER ), human Breast cancers are heterogeneous, with cells representing epidermal growth factor receptor-positive (HER2 ), or triple- different mammary lineages often found in the same tumor, negative , suggesting these cells may contribute to disease 14,16 including those with stem-like properties . We previously progression in all clinical subtypes. Notably, αvβ3 expression was showed in patient breast cancers that cells expressing the surface not synonymous with traditional CSC profiles, such as CD44 / marker integrin αvβ3 represent a stem-like cancer cell subset Low1 + Low2 12 CD24 , CD49f /EpCAM , the claudin-low intrinsic subtype associated with disease progression in a diverse array of 2,13,14 + 11 + or mesenchymal markers . Instead, αvβ3 cells represented a subtypes . Despite the potential significance of αvβ3 CSCs for distinct subset of these broader classifications . Our prior findings disease progression, few good models exist to study these cells in provided valuable insight regarding the aggressive nature of the context of other non-stem cells. Additionally, the scarcity of αvβ3 CSCs and emphasized the need to further elucidate the these cells in patient samples represents another practical unique genes and signaling pathways required for their function. limitation to studying these cells. One potential in vitro model 1 2 Department of Pathology, University of California, San Diego, La Jolla, CA 92093, USA. Moores Cancer Center, University of California, San Diego, La Jolla, CA 92093, USA. 3 4 Department of Medicine, University of California, San Diego, La Jolla, CA 92093, USA. Present address: Department of General Surgery, The First Affiliated Hospital of Xi’an Jiaotong University, Xi’an 710061, China. email: jdesgrosellier@ucsd.edu Published in partnership with the Breast Cancer Research Foundation 1234567890():,; Q. Sun et al. a b Tumor Incidence 32.6% 57.3% 10 EpCAM high EpCAM low Number of sorted cells injected v3 neg. v3 pos. v3 neg. v3 pos. 50,000 2/2 2/2 1/2 2/2 35,000 5/11 5/12 3/13 10/13 15,000 0/8 1/8 2/6 3/8 2 Frequency 1/64,755 1/59,918 1/91,076 1/24,615 6.58% 3.49% 95% Confidence Interval (1/31,119- (1/29,954- (1/40,447- (1/14,335- 2 3 4 5 10 10 10 10 134,748) 119,857) 205,083) 42,226) P = 0.0324 v3 c d Differentiation Low + Parental HCC38 EpCAM /v3 * 5 5 5 30.1% 61.3% 5.09% 73.7% 10 10 4 4 10 10 3 3 10 10 2 2 0 10 10 4.32% 4.26% 5.28% 15.9% - + - + v3 v3 v3 v3 2 3 4 5 2 3 4 5 high low EpCAM EpCAM 10 10 10 10 10 10 10 10 v3 Fig. 1 Integrin αvβ3 enriches for tumor-initiating ability and stem-like properties. a Representative FACS density plot of HCC38 breast cancer cells showing the live, CD49f cells according to their cell surface EpCAM and αvβ3 status. b Table describing the frequency of tumor formation per fat pad injected for each sorted cell type. Results pooled from four independent experiments. c Histogram showing the estimated number of tumor-initiating cells from the data in (b). b, c Statistics by Extreme Limiting Dilution Analysis (ELDA), which uses a chisquare likelihood ratio test to calculate p-values between groups. *P < 0.05. d Representative FACS density plots showing differentiation of Low + - sorted EpCAM /αvβ3 cells re analyzed after 6 weeks (10 passages). a, d n = 3 independent experiments. See also Supplementary Fig. 1. for our studies is the heterogeneous HCC38 cell line, which types, representing an ideal in vitro model to begin parsing critical + high consists of luminal-like (CD49f /EpCAM ) and stem-like cell gene expression and signaling differences. + low 13 types (CD49f EpCAM ) . Our analysis of surface αvβ3 in these Low + cells further identified a population of EpCAM /αvβ3 cells enriched for stemness properties such as tumorsphere formation αvβ3 CSCs express genes associated with activated stem and self-renewal . Thus, to more closely reflect the situation in cells patients’ tumors, we examined the HCC38 breast cancer cell line To discover critical stemness genes in αvβ3 CSCs, and determine as a potential model for our studies of αvβ3 CSCs. any similarities with normal mammary cell types, we performed To rigorously compare stemness traits in vivo, we sorted HCC38 bulk RNA-Seq analysis. The principal component analysis high- cells into four populations based on their EpCAM and αvβ3 status lighted a surprising amount of distinction between αvβ3 and (Fig. 1a and Supplementary Fig. 1a) prior to evaluating their αvβ3 cells (Fig. 2a), even greater than that due to EpCAM status tumor-initiating potential in vivo (Fig. 1b, c). Sorted cells were 2 alone, a widely-used marker to identify stem-like cells . This was injected orthotopically into the inguinal mammary gland fat pads high Low even more surprising since EpCAM and EpCAM cell types of adult female immunocompromised mice, then compared for are widely separated and distinct populations, whereas αvβ3 their ability to initiate new tumors in limiting dilution assays (Fig. expression represented a continuum of high and low expressers Low + 1b, c). We now show that EpCAM /αvβ3 cells possess about a (Fig. 1a). Meanwhile, both αvβ3 cell types exhibited a high 4-fold greater ability to initiate tumors relative to other HCC38 cell degree of similarity at the transcriptional level (Fig. 2a). This types (Fig. 1b, c). This is consistent with our prior tumorsphere suggests a potential relationship between these cell types, results and further supports their characterization as stem-like consistent with our differentiation results (Fig. 1d). To probe this cells. Another important attribute of stem-like cells is their ability Low + relationship further we compared the expression of a few select to differentiate. To determine if EpCAM /αvβ3 cells also markers of normal mammary cell types. Since αvβ3 expression has possessed this property we cultured sorted cells for exactly 10 previously been shown to occur on both stem/basal and luminal passages prior to re-analyzing by flow cytometry (Fig. 1d and progenitor cells in the normal murine and human mammary Supplementary Fig. 1b). This showed that indeed these cells were 10,17,18 gland we examined markers previously established to capable of differentiating into all three of the other cell types differentiate between these two cell types . Our analysis of these analyzed. Comparison with parental HCC38 cells showed that mammary cell markers showed that both αvβ3 cell types are there was a great deal of lineage specificity with regards to each Low + enriched for genes associated with stem/basal, but not luminal sorted cell type, with EpCAM /αvβ3 cells displaying a High + progenitor cells (Fig. 2b and Supplementary Fig. 2a), consistent preference for differentiating into EpCAM /αvβ3 cells (Fig. High with our hypothesis that αvβ3 CSCs display characteristics of 1d and Supplementary Fig. 1b). Interestingly, the EpCAM / adult MaSCs. αvβ3 cells changed the least, suggesting that they represent a more stable differentiated cell type (Supplementary Fig. 1b). Thus, To further probe any potential similarity between αvβ3 CSCs similar to patients’ tumors, we show that the HCC38 cell line and activated stem/basal cells from the normal mammary gland contains a rare subset of αvβ3 CSCs, in addition to non–stem cell we assessed the differentially expressed genes (DEG) within each npj Breast Cancer (2022) 5 Published in partnership with the Breast Cancer Research Foundation HCC38 cells Tumor Initiation 1234567890():,; Tumor-initiating cells (Estimate per 100,000) EpCAM EpCAM Follicular (Inactive) Q. Sun et al. a b High - Low - Mammary cell markers EpCAM /v3 EpCAM /v3 + - High + Low + (v3 vs v3 ) EpCAM /v3 EpCAM /v3 1.5 Low EpCAM High EpCAM 1.0 Luminal progenitor 0.5 genes 0.0 -0.5 Stem/basal genes -5 -1.0 -10 -5 0 5 PC1: 30% variance c d Activated stem/basal cell gene sets Low High EpCAM EpCAM Hypoxia Response to abiotic stimulus EpCAM High DEG v3- v3+ EMT EpCAM High Tissue development 146 468 Regulation of cell death Response to oxygen levels Response to external stimulus Cellular response to stress 104 425 Myc targets V1 Posttranscriptional regulation of gene expression EpCAM Low v3- v3+ Ribonucleoprotein complex biogenesis EpCAM Low mRNA processing Ribosome biogenesis -3 -2 -1 0 1 2 3 + - Enrichment Score (v3 vs v3 ) Fig. 2 αvβ3 CSCs are similar to activated stem/basal cells from the normal human mammary gland. a Principal component analysis (PC) performed on bulk RNA-Seq data from each of the indicated HCC38 sorted cell types. b Relative expression of select gene markers of stem/ + − basal or luminal progenitor cells in αvβ3 versus αvβ3 cells. Data represent the mean ± s.e.m. # = not significant. c Venn diagrams depicting the number of differentially expressed genes (DEG) identified in each cell type. The selection criteria was ≥1.5-fold change in gene expression + − and P < 0.05. d Comparison of αvβ3 versus αvβ3 cell GSEA results with the top gene sets enriched in stem/basal cells during luteal (Active) versus follicular (Inactive) menstrual cycle phases. b, d Statistics by Student’s t-test with Benjamini-Hochberg multiple comparisons test. a–d n = 3 independent experiments. See also Supplementary Fig. 2. cell type (Fig. 2c) and performed gene set enrichment analysis Identification of key genes unique to αvβ3 CSCs (GSEA). While αvβ3 cell types were closely related (Fig. 2a, b), we Based on our findings that αvβ3 CSCs enrich for stemness Low + identified 180 genes enriched in EpCAM /αvβ3 cells compared properties such as tumor initiation (Fig. 1b, c) we wished to High + to EpCAM /αvβ3 cells (Fig. 2c and Supplementary Fig. 2b). We determine the key genes and signaling pathways critical for their then compared gene sets enriched in both αvβ3 cell types with function. We began by selecting several gene sets associated with those from activated stem/basal cells in the normal human αvβ3 CSCs based on their relevance to breast cancer, stem cells, mammary gland. Since data from pregnancy is unavailable, we or signaling pathways (Fig. 3a and Supplementary Fig. 3a). By compared our GSEA results with published data from normal determining which of the 180 DEG’s identified in Fig. 2c were basal/stem cells during the luteal (Active) versus follicular present within each gene set, we identified 20 candidate genes 19 + (Inactive) phases of the menstrual cycle . Many of the same unique to αvβ3 CSCs (Fig. 3b and Supplementary Fig. 3b), hormone-induced changes that occur during the luteal phase also referred to as our αvβ3 CSC signature. In order to perform a happen during pregnancy. The results were striking, as gene sets functional screen of these genes, we sought to identify appro- found in activated stem/basal cells were overwhelmingly shared priate surrogate cell lines for our αvβ3 CSCs. For this analysis, we by αvβ3 cancer cells, while those in inactive cells were not (Fig. made use of published gene sets from 28 breast cancer cell lines 2d and Supplementary Fig. 2c). Interestingly, of the gene sets that were previously used to classify these cells according to their 13 + analyzed, the αvβ3 cells differed only in the genes involved in intrinsic subtype . Comparison with our αvβ3 CSC signature the cellular response to stress, with this representing a unique revealed specific enrichment in the claudin-low cell type (Fig. 3c), Low + 11 feature distinguishing EpCAM /αvβ3 cells (Fig. 2d). These consistent with prior characterization of these cells as stem-like . findings highlight an association between αvβ3 CSCs and the However, careful analysis of each of the eight claudin-low cell lines activated state in normal mammary stem/basal cells and suggest revealed that only three of them displayed any enrichment that a heightened response to stress may be a key distinguishing beyond the parental HCC38 cells (Fig. 3d), in which αvβ3 CSCs feature of these cells. are only a small fraction of the total cells. The three cell lines Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 5 TAGLN ACTA2 SNAI2 CDK1 TP63 MYLK CAV1 KIT ELF5 PC2: 19% variance Log2 fold change Luteal (Active) Log2 Fold Change Q. Sun et al. ab High - EpCAM /v3 High + EpCAM /v3 Low - EpCAM /v3 Gene sets with significantly modulated genes Low + EpCAM /v3 Low + - (EpCAM /v3 vs v3 ) WNT5A Schuetz Breast Cancer Ductal Invasive Up VGLL3 PRICKLE1 Lim Mammary Stem Cell Up TSPAN5 COBL 1 Koinuma Targets of Smad2 or Smad3 SFRP4 POSTN Kim WT1 Targets Up 0 FREM2 RSAD2 Plasari TGFB1 Targets 10hr Up RBPMS2 -1 ZEB1 Wong Adult Tissue Stem Module DACT1 Negative Regulation of Canonical Wnt -2 TGFBI IL11 01 234 DLC1 LTBP2 - Log 10 (q-value) ALCAM PLAUR PTHLH AMOTL1 c d Breast cancer cell lines Claudin-low cell lines 0.6 0.5 0.4 0.4 0.3 0.2 0.2 0.0 0.1 -0.2 0.0 -0.4 -0.1 Fig. 3 αvβ3 CSCs express unique genes associated with aggressive breast cancers and normal mammary stem cells. a Gene set Low + − enrichment analysis (GSEA) for EpCAM /αvβ3 versus αvβ3 cells. Statistics by Student’s t-test with Benjamini−Hochberg test. Dashed line Low + indicates statistical significance. b Heat map of the 20 most differentially expressed genes upregulated in EpCAM /αvβ3 cells and found in the gene sets in (a) (Log2 scale). c Box and whiskers plot showing the relative enrichment for the αvβ3 CSC gene signature in cell lines representing different molecular subtypes. Boxplots represent medians (center line) and interquartile range (IQR; box), and whiskers represent the maximum and minimum values within 1.5 times the IQR from the edge of the box. Statistics by ANOVA with Tukey’s test. *P < 0.05. Cell lines in each category: Luminal B; n = 7, HER2; n= 7, Basal-like; n = 6, Claudin-low; n = 8. d Claudin-low cells lines with significant enrichment for the αvβ3 CSC gene signature compared to parental HCC38 cells (dashed line). a–d n = 3 independent experiments. See also Supplementary Fig. 3. identified (MDA-MB-231, BT549, and Hs578T) represent some of surrogate for our αvβ3 CSCs, as characterized in Fig. 3d. Tumor- the most widely used breast cancer cell lines due to their high sphere formation in methylcellulose wasselectedasour primary tumorigenicity and metastatic potential, highlighting the potential endpoint since it is a critical stemness property. Results from these studies identified two genes as highly relevant for further study: significance of our candidate genes for aggressive disease. These TGFBI (Transforming Growth Factor Beta Induced; initially termed BIG- findings serve to distinguish αvβ3 CSCs as a unique cancer cell H3) and ZEB1 (Fig. 4b and Supplementary Fig. 4a). We rigorously type that is not synonymous with the claudin-low classification validated these targets by showing ZEB1 protein enrichment in and identify appropriate surrogate cell lines in which to screen our sorted αvβ3 CSCs from HCC38 cells (Fig. 4c). Additionally, we used candidate genes for their role in stemness. another heterogeneous cell line (SUM149) to show conserved Low + expression of both TGFBI and ZEB1 specifically in EpCAM /αvβ3 Characterization of ZEB1 and TGFBI as candidate genes cells (Fig. 4d, e and Supplementary Fig. 4b). Importantly, ZEB1 protein required for stemness levels were specifically associated with the surrogate cell lines We next wished to perform an unbiased assessment of the key genes enriched for the αvβ3 CSC gene signature (Supplementary Fig. 4c), as and signaling pathways responsible for the more aggressive nature 20 well as the LM2-4 metastatic variant of the MDA-MB-231 cell line of αvβ3 CSCs. Here, we used αvβ3 as a marker of activated stem-like (Supplementary Fig. 4d), all of which we previously showed to cells, with candidate genes selected without regard for a potential express αvβ3 . direct link to αvβ3 signaling. To identify candidate genes, we A secreted ECM protein, TGFBI (Transforming Growth Factor Beta performed QPCR analysis of the 20 DEGs from Fig. 3b to select those Induced; BIG-H3) has been shown to paradoxically enhance 21 22 that displayed the most consistent and robust expression in αvβ3 anchorage-independence , similar to our findings with αvβ3 . CSCs relative to the other three cell types (Fig. 4a). This identified six While best known for its role in epithelial-mesenchymal transforma- genes for further analysis (Fig. 4a). We then performed an siRNA tion (EMT) , the transcription factor ZEB1 also mediates non-EMT functional screen to identify which of these genes was most critical functions that may be more important for its role in tumor 24,25 for αvβ3 CSCs (Fig. 4b and Supplementary Fig. 4a). To simulta- progression . In fact, recent studies unexpectedly found ZEB1 in neously examine the function of multiple genes after transient siRNA a subset of basal/stem cells in normal human mammary glands knockdown we employed the BT549 cell line as an appropriate where it promotes oncogene-induced transformation . While the npj Breast Cancer (2022) 5 Published in partnership with the Breast Cancer Research Foundation Luminal B Her2 Basal-like Claudin-low HCC1395 MDA-436 SUM159 SUM1315 HCC38 MDA-231 BT549 Hs578T v3 CSC Score v3 CSC Score Q. Sun et al. a b QPCR Validation Functional Screen (BT549) Low - EpCAM /v3 10 35 Low + EpCAM /v3 8 High - EpCAM /v3 High + EpCAM /v3 20 * 0 0 c e SUM149 cells 91.1% 5.71% Low - EpCAM /v3 Low High EpCAM EpCAM Low + EpCAM /v3 250 4 v3: - + - + High - EpCAM /v3 3 High + ZEB1 EpCAM /v3 -actin 2.97% 0.22% 1 2 3 4 10 10 10 10 v3 ZEB1 TGFBI Fig. 4 Identification of TGFBI and ZEB1 as candidate genes unique to αvβ3 CSCs. a QPCR validation of candidate genes in sorted HCC38 cell types. b Functional screen for candidate genes necessary for methylcellulose colony formation after transient siRNA knockdown in BT549 cells. Target gene knockdown was validated by QPCR. Statistics by one-way ANOVA with Dunnett’s test. *P < 0.05. c Representative immunoblot of lysates from sorted HCC38 cells. β-actin is shown as a loading control. Molecular weight markers are indicated in kilodaltons. d Representative FACS density plot of the live, CD49f SUM149 cells according to their cell surface EpCAM and αvβ3 status. e QPCR validation of candidate genes in sorted SUM149 cells. a, e Samples were run in duplicate with GAPDH as a loading control. Expression is shown relative Low − to the EpCAM /αvβ3 cells (dashed lines). a, b, e Data represent the mean ± s.e.m. a–e n = 3 independent experiments. See also Supplementary Fig. 4. exact identity and function of these cells is still a mystery, it suggests αvβ3 CSCs identified by a rigorous, unbiased and systematic they may be similar to ZEB1 breast cancer cells. approach. Discovery of TGFBI-ZEB1 as a key stemness-related signaling TGFBI-ZEB1 promotes chromosomal stability and resistance to module PARP inhibition In order to further assess the relevance of these two genes for We next considered the cell biological basis for these effects on stemness properties, we generated TGFBI and ZEB1 knockout cells tumorsphere formation. Our GSEA results suggest that the ability using CRISPR/Cas9 in our surrogate stem-like LM2-4 and BT549 cell to respond to cellular stress was a distinguishing feature of αvβ3 lines. We further validated these cells, showing significantly CSCs. In fact, while TGFBI and ZEB1 have diverse cellular functions, reduced ZEB1 protein levels (Fig. 5a) as well as decreased they may play a common role in reducing a certain type of genetic 24,27 amounts of TGFBI mRNA expression (Supplementary Fig. 5a) and stress called chromosomal instability (CIN) . CIN is a hallmark of secreted TGFBI protein (Supplementary Fig. 5b). Surprisingly, these cancer and an important stress in cancer cells that limits validation studies showed that TGFBI deletion also resulted in transformation . In fact, a recent study showed that normal adult decreased levels of ZEB1 protein (Fig. 5a). Importantly, deletion of MaSCs were inherently more tumorigenic due to suppression of ZEB1 did not decrease levels of TGFBI mRNA (Supplementary Fig. CIN via ZEB1 . Our independent discovery of ZEB1 as one of the 5c). We also noted that deleting either gene had no effect on most DEG in αvβ3 CSCs, suggested that it may play a similar role protein levels of the β3 subunit (Supplementary Fig. 5d). These in these cells. unexpected findings suggest that these two independently Using multiple methods, we now show that CRISPR knockout of identified candidate genes are linked within the same pathway ZEB1 or TGFBI enhances CIN in stem-like cell lines. We began by To examine this possibility and validate their role in stemness, measuring staining for the DNA damage marker phospho-γH2AX we tested our TGFBI and ZEB1 CRISPR knockout cells in assays of and found that DNA strand breaks increased in our knockout cells primary tumorsphere formation and self-renewal (Fig. 5b). (Fig. 6a, b). In contrast, there was no effect on proliferation as Deletion of either TGFBI or ZEB1 resulted in an approximately assessed by incorporation of fluorescently-labeled EdU (Fig. 6a, c) 50% decrease in primary tumorspheres, while subsequent self- or apoptosis measured by PARP cleavage (Supplementary Fig. 6a). renewal assays showed an almost 75% decrease due to ZEB1 Since staining with phospho-γH2AX indicates the presence of DNA knockout in BT549 cells (Fig. 5b). These findings highlight an strand breaks that could lead to missegregation of chromosomes, important role for these genes in stemness and are consistent we quantified micronuclei as a direct measure of CIN and with their function within the same pathway. Indeed, we show observed increased levels associated with both knockout cell that adding recombinant human TGFBI protein (rhTGFBI) is lines (Fig. 6d). To robustly examine differences in CIN we also sufficient to drive ZEB1 protein expression in control and TGFBI evaluated potential copy number alterations (CNA) and found knockout cells (Fig. 5c) and specifically rescue defective tumor- higher levels in our knockout cells (Fig. 6e). Analysis of data from sphere formation caused by TGFBI deletion (Fig. 5d). Our findings The Cancer Genome Atlas further supported these findings by highlight a potential new TGFBI-ZEB1 signaling module specificto showing that high ZEB1 expression in tumors corresponded with Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 5 SUM149 cells ZEB1 WNT5A DACT1 TGFBI ALCAM TSPAN5 siControl siZEB1 siWNT5A siDACT1 siTGFBI siALCAM siTSPAN5 mRNA Levels (Fold Change) EpCAM Colonies mRNA Levels (Fold Change) Q. Sun et al. a b Primary Tumorspheres Secondary Tumorspheres LM2-4 BT549 50 * ZEB1 TGFBI ZEB1 TGFBI * * * 25 Ctrl KO KO Ctrl KO KO ZEB1 30 -actin 37 10 0 0 LM2-4 BT549 LM2-4 BT549 Ctrl TGFBI KO ZEB1 KO c d Rescue w/ rhTGFBI BT549 Ctrl + Vehicle * * Ctrl + rhTGFBI Ctrl TGFBI KO rhTGFBI TGFBI KO + Vehicle 30 * * (ng/mL) : - 20 100 500 - 20 100 500 TGFBI KO + rhTGFBI ZEB1 -actin LM2-4 BT549 Fig. 5 Discovery of a TGFBI-ZEB1 signaling module required for stemness. a Representative immunoblot of lysates from TGFBI and ZEB1 CRISPR knockout cells. b Tumorsphere assays in methylcellulose to assess the effects of TGFBI or ZEB1 CRISPR knockout on primary colonies (left) or self-renewal (right) in the indicated cell lines. c A representative immunoblot experiment, showing rescue of ZEB1 protein levels after treatment with rhTGFBI for 24 h with the indicated doses compared to vehicle control (PBS). a, c β-actin is shown as a loading control. Molecular weight markers are indicated in kilodaltons. d Rescue of primary colony formation in TGFBI knockout cells with 500 ng/mL rhTGFBI. Data represent the mean ± s.e.m. Statistics by one-way (b) or two-way (d) ANOVA with Dunnett’s(b) or Tukey’s test (d). *P < 0.05. n = 3(a–c)or n =5(d) independent experiments. See also Supplementary Fig. 5. low levels of CIN (Supplementary Fig. 6b). Thus, TGFBI-ZEB1 now show a striking correlation between αvβ3 CSCs and signaling appears to promote stemness in αvβ3 CSCs by activated stem/basal cells from the normal human mammary suppressing the endogenous genetic stress caused by CIN. gland. We further identified key genes associated with this cell Tumor cells with defective DNA repair due to BRCA mutations state including the secreted matrix protein TGFBI (BIG-H3) and the are highly sensitive to PARP inhibition due to an accumulation of transcription factor ZEB1. Taken together, our findings suggest double-strand breaks that tips the balance toward cell death . that these genes may operate as a TGFBI-ZEB1 signaling module Since we observed increased DNA strand breaks in our TGFBI and to promote stemness by protecting against genetic stress, such as ZEB1 CRISPR knockout cells (Fig. 6a, b), we hypothesized that CIN. In fact, downregulating TGFBI-ZEB1 sensitized αvβ3 CSCs to some of these may fail to be repaired, possibly leading to synergy PARP inhibition, laying the foundation for a potential new with PARP inhibitors such as Olaparib (Lynparza). While Olaparib is treatment strategy to reduce breast cancer progression. clinically-approved for BRCA-mutant breast and ovarian cancers, Unbiased analysis of critical genes and pathways in αvβ3 CSCs led to our surprising discovery of a TGFBI-ZEB1 signaling module. non-BRCA mutant cancers are completely refractory to this treatment . Indeed, we show that while deletion of either TGFBI While TGFBI is a secreted ECM protein that would normally bind to or ZEB1 had no effect on 2D cell viability (Supplementary Fig. 6c), integrins and elicit adhesion-dependent responses, we and others knockout of either gene synergized with Olaparib in two different have now shown that it can also enhance anchorage-independent stem-like cell lines (Fig. 6f and Supplementary Fig. 6d). Signifi- growth . This is similar to our surprising finding that the integrin αvβ3 also promotes anchorage-independence , and suggests the cantly, we observed sensitivity to Olaparib at similar doses that are two may function as a possible ligand-receptor pair in stem-like effective against a BRCA-mutant cell line (Supplementary Fig. 6e). These data are consistent with an important role for TGFBI-ZEB1 in cells. The transcription factor ZEB1 is perhaps best known for its reducing CIN in αvβ3 CSCs, highlighting the ability to control role in EMT; however, it is also important for functions not related 24,25 genetic stress as a critical attribute of stem-like cells. Additionally, to EMT that may be even more critical for tumor progression . our findings suggest that PARP inhibition may be an effective An unexpected result of gene atlas studies from the normal human mammary gland was the discovery of a subset of stem/ precision therapy for more than just BRCA-mutant disease, and + 26 basal cells expressing ZEB1 . Notably, these cells were specificto that a similar approach may be able to eliminate αvβ3 CSCs and reduce breast cancer progression. human glands and not observed in mice . Further corroborating this finding, a different study identified ZEB1 expression in enriched populations of human MaSCs, where it surprisingly DISCUSSION functioned to promote oncogene-induced transformation . Thus, While stem cells in the adult mammary gland are dynamic, cycling in the normal mammary gland, ZEB1 is expressed in cells that through active and inactive cell signaling states due to hormonal display stem cell properties. While there is still much to learn 10,30,31 signaling , it is unclear if stem-like tumor cells possess a about these cells, our new findings suggest they may represent similar ability. Our prior work identified the integrin αvβ3as a MaSCs in the activated state and display traits similar to ZEB1 surface marker of activated stem cells in the adult mammary breast cancer cells. Together, our results highlight a potential new gland, suggesting that tumor cells expressing this marker may TGFBI-ZEB1 pathway specificto αvβ3 CSCs that we identified feature a similar activated signaling state. By comparing our whole through a rigorous, unbiased and systematic approach. transcriptome sequencing data from sorted αvβ3 CSCs with While CIN is a hallmark of cancer cells, too much may act to limit published gene sets enriched in stem/basal cells during the luteal tumor progression . In fact, normal cell types that can better (Active) and follicular (Inactive) phases of the menstrual cycle, we tolerate CIN, such as MaSCs are much more likely to undergo npj Breast Cancer (2022) 5 Published in partnership with the Breast Cancer Research Foundation Colonies Colonies Colonies Q. Sun et al. a b LM2-4 DNA strand breaks 35 * * Ctrl Ctrl TGFBI KO ZEB1 KO * * TGFBI KO ZEB1 KO LM2-4 BT549 Proliferation Ctrl n.s. TGFBI KO n.s. ZEB1 KO LM2-4 BT549 d e Micronuclei TGFBI KO adj. p<0.01 8 16 n.s. * Ctrl * TGFBI KO 6 12 ZEB1 KO 2 4 0 0 LM2-4 BT549 -0.25 0 0.25 0.5 Log2 Ratio CNA (KO vs Ctrl) BT549 ZEB1 KO adj. p<0.01 * * 3 n.s. Ctrl TGFBI KO 1 ZEB1 KO -0.25 0 0.25 0.5 0.1 1 10 [Olaparib (M)] Log2 Ratio CNA (KO vs Ctrl) Fig. 6 Decreased TGFBI-ZEB1 signaling enhances chromosomal instability and sensitivity to PARP inhibition. a Representative immunofluorescent staining for the DNA damage marker phospho-γH2AX (red) or detection of fluorescently-labeled EdU (red) after 90 min incubation to assess cell proliferation. Nuclei are stained blue in all images. Scale bars, 40 μm. Percentage of cells positive for p-γH2AX (b), EdU (c), or micronuclei (d) relative to total nuclei. Data calculated from four random fields per condition for each experiment. e Volcano plots depicting the copy number alterations (CNA) in BT549 TGFBI or ZEB1 knockout cells relative to controls. Statistics performed by Student’s t-test corrected for multiple testing using the Benjamini-Hochberg method. Red dots represent the 112 (TGFBI KO) and 129 (ZEB1 KO) segments with statistically significant differences (adjusted p-value < 0.01) out of 459 total segments examined. Black dots are not significant (n.s.). f XTT cell viability assays comparing Olaparib sensitivity in the indicated BT549 cell types. Curves are plotted relative to vehicle controls for each group and fitted by non-linear regression. *P < 0.05 for 0.5, 0.75, 1, 2, 5, and 10 μM Olaparib (TGFBI or ZEB1 KO versus control). n = 3 (a–d, f)or n =4(e) independent experiments. Data represent the mean ± s.e.m. Statistics by one-way (b−d) or two-way (f) ANOVA with Dunnett’s test. *P < 0.05. n.s.= not significant. See also Supplementary Fig. 6. oncogenic transformation , suggesting that control of genetic consistent with the enhanced metabolic activity identified in stress is an important attribute of more aggressive tumor cells. stem-like breast cancer cells found in metastases . Thus, our While TGFBI and ZEB1 have diverse cellular functions, they may investigation of the biological effects of TGFBI-ZEB1 support a role 24,27 play a common role in reducing CIN . In fact, a recent study in limiting the effects of genetic stress and maintaining showed that MaSCs were inherently more tumorigenic due to chromosomal stability, suggesting this may be a defining attribute suppression of CIN via ZEB1 . Our independent discovery of ZEB1 and potential vulnerability of stem-like cells. as one of the most differentially-expressed genes in αvβ3 CSCs, PARP inhibitors are clinically-approved and highly effective suggested that it may play a similar role. We now show that treatments for BRCA-mutant breast and ovarian cancers . Tumor CRISPR knockout of either ZEB1 or TGFBI increased CIN in two cells with defects in DNA double-strand break repair, such as BRCA stem-like cell lines with no effect on cell proliferation or survival. mutations, are more sensitive to PARP inhibitors, such as Olaparib, This genetic instability may be caused by endogenous factors which prevent single-strand break repair and drive further genetic such as cell replication or the production of reactive oxygen instability, resulting in cell death. However, while PARP inhibition species during metabolism. The latter of which would be is an effective treatment against BRCA-mutant breast and ovarian Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 5 % Micronuclei-positive cells EdU p-H2AX Viable cells (% Control) % Micronuclei-positive cells -Log10 (Adjusted P-value) -Log10 (Adjusted P-value) % EdU-positive cells % p-H2AX-positive cells Q. Sun et al. cancers, non-BRCA mutant cancers are completely refractory to Bulk RNA-Seq this treatment . We now provide proof-of-concept that disrupting After sorting HCC38 cells, RNA was purified from an equal number of cells per experiment for each cell type (approximately 50–70,000 cells) using a TGFBI-ZEB1 signaling not only increased CIN, but functioned much RNeasy Mini Kit (Qiagen). The samples were then submitted to the IGM like a BRCA mutation by increasing sensitivity to PARP inhibition. Genomics Core at UCSD for validation of RNA quality and sequencing was We hypothesized that this may lead to synergy with Olaparib, performed on a HiSeq 4000 (Illumina, San Diego, CA, USA). which normally affects only BRCA-mutant cancers. Indeed, our new data shows that TGFBI or ZEB1 deletion enhances “BRCA- Bioinformatics analysis ness”, and synergizes with Olaparib. Our findings suggest that RNA-seq data were analyzed with a pipeline implemented in the disrupting key mediators of chromosomal stability in αvβ3 CSCs, BCBio-nextgen workflow manager https://zenodo.org/record/4686097#. such as TGFBI and ZEB1, can sensitize these normally resistant YRLzj4hKiM8. Briefly, we aligned reads to GRch37 reference genome using cells to treatment with clinically-approved PARP inhibitors. There- 33 34 STAR and quantified expression levels with Salmon 0.13.1 . We then fore, these findings represent a crucial initial step laying the annotated genes with BioMART , keeping only protein coding genes with foundation for further study of the activated stem cell state as a more than one read count for analysis. DEG was then identified using key contributor to recurrence and metastasis in patient disease DESEQ2 . Gene Set Enrichment analysis was performed using R package LIGER on the Hallmark and Reactome gene sets available in MSigDB . The and outline a potential therapeutic strategy for targeting 20 gene signature was derived by overlap analysis between the two these cells. differential gene expression analyses followed by manual curation for gene set membership. The public gene expression profile from breast cancer cell lines was obtained at NCBI GEO (GSE50470) and the corresponding MATERIALS AND METHODS intrinsic subtype information obtained from Prat et al. . The CSC signature Cell lines score was calculated according to Barbie et al. implemented in the The following breast cancer cell lines were purchased from ATCC gseapy python package (v0.9.8). (Manassas, VA, USA): HCC38, MDA-MB-436, MCF-7, T47D, BT474, MDA- MB-468, BT-20, HCC1187, Hs578T, BT549, and MDA-MB-231. LM2-4 cells, a Real-time qPCR highly metastatic variant of the MDA-MB-231 cell line was a gift from qPCR experiments on cultured cells were performed by collecting total Robert Kerbel. All cell lines were tested and shown to be free of RNA using the RNeasy Mini Kit (Qiagen) and reverse transcribing with the mycoplasma. The HCC38, BT549, and LM2-4 cells were further authenti- High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Thermo cated by short tandem repeat (STR) testing. Cells used in mice were Fisher Scientific). Relative mRNA levels from sorted cells were examined additionally tested and found to be negative for an extensive panel of using the Cells-to-CT kit (Life Technologies) according to the manufac- mouse pathogens. Cell lines were cultured in complete DMEM medium turer’s instructions. Lysates were prepared from 90,000 freshly sorted (DMEM supplemented with 10% fetal bovine serum (FBS) + 1% L- HCC38 cells. Real-time qPCR was performed using iTaq Universal SYBR glutamine, sodium pyruvate, non-essential amino acids, and antibiotic/ Green Supermix (Bio-Rad, Hercules, CA, USA) and run on a LightCycler 480 antimycotic). qPCR System (Roche, Basel, Switzerland). See Supplementary Methods for a list of primers. Cell transfection and lentiviral transduction Plasmids containing enhanced specificity Cas9 and the appropriate guide Immunoblotting RNA’s in the pLentiCRISPRv2 vector were purchased from GenScript Whole cell lysates were prepared from cell lines with RIPA lysis buffer (Piscataway, NJ, USA) for generating stable knockout with lentivirus and (100 mM Tris pH 7.5, 150 mM sodium chloride, 0.1% deoxycholate, 0.1% selected using puromycin. Transient transfections for all CRISPR/Cas9 SDS, 50 mM NaF, Protease inhibitor cocktail (Roche), 2 mM PMSF, 2 mM vectors into 293T cells were performed with Lipofectamine 3000 sodium orthovanadate) combined with scraping and the lysates cleared by (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA), while HiPerFect centrifugation. Standard Western blotting procedures were performed. The following primary antibodies were used for immunoblotting at a dilution of Transfection Reagent (Qiagen, Hilden, Germany) was used for siRNAs. All 1:1000: ZEB1 (3396, Cell Signaling Technology, Danvers, MA, USA), Full- transfections were performed according to the manufacturers’ instructions. length PARP (9532, Cell Signaling Technology), Hsp90 (sc-13119, Santa FlexiTube siRNAs (Qiagen) included AllStars negative control, ZEB1 Cruz, Dallas, TX, USA) and β-actin (MABT825, MilliporeSigma). Treatments (SI04339587), WNT5A (SI04384184), DACT1 (SI00359275), TGFBI with rhTGFBI (R&D Systems, Minneapolis, MN, USA) or vehicle control (PBS) (SI02780722), ALCAM (SI02780155), and TSPAN5 (SI04151665). were performed at the specified doses for 24 h prior to harvesting lysates. All blots or gels derive from the same experiment and were processed in Generation of CRISPR knockout cells parallel. See Supplementary Information for unedited blots (Supplemen- Stable knockout of select genes was achieved by transducing BT549 or tary Fig. 8a–c). LM2-4 cells with lentivirus vectors expressing enhanced specificity Cas9 and the appropriate guide RNA’s targeting human TGFBI or ZEB1 TGFBI ELISA (GenScript) and pooling puromycin-resistant cells. A vector lacking the Conditioned media from BT549 and LM2-4 cell lines was collected after guide RNA was used as a negative control. Successful knockout of the 48 h in phenol-free complete DMEM culture medium. Secreted TGFBI levels respective targets was verified by Western blot for ZEB1 and Real-time were then quantified with the Human TGFBI (BIGH3) ELISA kit (Invitrogen, QPCR for TGFBI. Thermo Fisher Scientific) according to the manufacturer’s instructions. Conditioned media was diluted 1:50 (BT549) or 1:200 (LM2-4) in order to fit on the standard curve. Flow cytometry Single-cell suspensions were prepared from cultured HCC38 or SUM149 cells, blocked in 0.5% BSA/PBS, and stained with the following antibodies Immunofluorescence staining prior to sorting: CD49f-PE 1:10 (555736, GoH3; BD Biosciences, San Jose, Immunofluorescence staining was performed on cultured cells in a 4-well CA, USA); EpCAM–Alexa 647 1:20 (324212, 9C4; BioLegend, San Diego, CA, chamberslide (Nunc, Thermo Fisher Scientific) fixed brieflyin2% USA); and αvβ3–biotin 1:40 (MAB1976B, LM609; MilliporeSigma, Burling- paraformaldehyde in PBS at room temperature and permeabilized. All ton, MA, USA) and Streptavidin-Brilliant Violet 421 1:80 (BioLegend). samples were blocked with 1:80 normal goat serum in 0.1% BSA/PBS Propidium iodide solution (0.5 μg/ml) was used to detect dead cells. Viable before incubation in phospho-γH2AX primary antibody diluted 1:100 cells were collected by sorting with a FACSDiva or FACSAria machine (BD (9718, Cell Signaling Technology) overnight at 4 °C, followed by incubation Biosciences). In some cases, differentiation assays were performed by with DAPI and 1:500 Alexa Fluor 647-conjugated secondary antibody culturing sorted cells for exactly 10 passages (6–7 weeks after sorting) (A32733, Invitrogen, Thermo Fisher Scientific) for 2 h at RT. Alternatively, before re-analyzing by flow cytometry. See Supplementary Information for proliferation was measured by incubating live cells with Edu for 2 h at 37 °C gating strategies (Supplementary Fig. 7a–c). prior to labeling the Edu with Alexa Fluor 647 according to manufacturer’s npj Breast Cancer (2022) 5 Published in partnership with the Breast Cancer Research Foundation Q. Sun et al. instructions (Invitrogen, Thermo Fisher Scientific). DAPI was used to Statistics visualize nuclei. Cells stained by either method were imaged with a Nikon Data presentation and statistical tests are indicated in the figure legends. A1R confocal microscope and images captured from four randomly Two-tailed Student’s t-tests were used for comparing two means while selected fields with a 60× objective. Micronuclei were then manually ANOVA was performed for 3 or more data sets. Post hoc analysis was counted for each field or the percent positive cells quantified using Image performed using an appropriate multiple comparisons test as indicated in the legends. For all analyses, P < 0.05 was considered statistically J software. significant. Statistical analysis was performed using GraphPad Prism software (San Diego, CA, USA). Cell viability See Supplementary Material for additional methods. XTT cell viability assays were performed by first seeding cells into a 96-well tissue culture plate at the following density per well: LM2-4 (4,000), BT549 Reporting summary (2,000), HCC38 (4,000), or MDA-MB-436 (7,500). After cells attached Further information on research design is available in the Nature Research overnight the indicated concentrations of Olaparib (Lynparza) (Sell- Reporting Summary linked to this article. eckchem, Houston, TX, USA) or vehicle alone (DMSO) were then added to the wells in 100 μL phenol-free complete DMEM medium. After 48 h, XTT substrate (Thermo Fisher Scientific) was added to the wells and DATA AVAILABILITY incubated for 2 h before reading the A450 nm on a plate reader. Cell RNA sequencing data that support the findings of this study have been deposited in viability for each of the indicated treatments was expressed as a percent of the NCBI Sequence Read Archive (SRA) with the primary accession code the vehicle control wells. PRJNA750073. Copy number alterations Received: 14 April 2021; Accepted: 6 December 2021; Genomic DNA was prepared from 5 × 10 cells with the DNeasy Blood and Tissue Kit (Qiagen) and submitted to the IGM Genomics Core at UCSD. CNA was determined by hybridizing 200 ng of DNA per sample onto an Infinium CoreExome-24 Array (Illumina) and analyzing with Nexus CN software (version 7.5). The log2 ratio of signal intensity were calculated for each SNP REFERENCES on the array and copy number segments identified using the Bioconductor 1. Al-Hajj, M., Wicha, M. S., Benito-Hernandez, A., Morrison, S. J. & Clarke, M. F. copynumber package , including winsorization to remove outliers, Prospective identification of tumorigenic breast cancer cells. Proc. Natl Acad. Sci. imputation of missing value, and multi-sample copy number segmentation USA 100, 3983–3988 (2003). using multipcf. The differences in log2 ratio of all larger segments (more 2. Lim, E. et al. Aberrant luminal progenitors as the candidate target population for than 10 kb) were compared between control and KO BT549 cells. basal tumor development in BRCA1 mutation carriers. Nat. Med. 15, 907–913 (2009). 3. Prat, A. & Perou, C. M. Deconstructing the molecular portraits of breast cancer. Tumorsphere assays Mol. Oncol. 5,5–23 (2011). Primary tumorsphere formation was assessed in cells grown under 4. Malanchi, I. et al. Interactions between cancer stem cells and their niche govern anchorage-independent conditions in methylcellulose. BT549 (15,000) or metastatic colonization. Nature 481,85–89 (2011). LM2-4 cells (10,000) were cultured in 0.9 mL of 1% methylcellulose/ 5. Cai, S. et al. A quiescent Bcl11b high stem cell population is required for main- complete DMEM medium in ultra-low adhesion 24-well dishes (Corning, tenance of the mammary gland. Cell Stem Cell 20, 247–260 (2017). e5. Corning, NY, USA) and cells cultured for 10−12 days. Primary tumorspheres 6. Fu, N. Y. et al. Identification of quiescent and spatially restricted mammary stem were assessed by counting colonies consisting of at least 6 cells from 4 cells that are hormone responsive. Nat. Cell Biol. 19, 164–176 (2017). fields per well with a 10× objective. We measured self-renewal by 7. Fu, N. Y. et al. Foxp1 is indispensable for ductal morphogenesis and controls the exit of mammary stem cells from quiescence. Dev. Cell 47, 629–644 (2018). e8. collecting primary tumorspheres by dilution in at least 3 volumes of PBS, 8. Prater, M. D. et al. Mammary stem cells have myoepithelial cell properties. Nat. dissociating them with trypsin for approximately 10 min, and re-seeding in Cell Biol. 16, 942–950 (2014). 1-7. 1% methylcellulose before evaluating secondary colonies after 10 days. For 9. Asselin-Labat, M. L. et al. Steroid hormone receptor status of mouse mammary treatment with rhTGFBI (R&D Systems), a single dose of 500 ng/mL was stem cells. J. Natl Cancer Inst. 98, 1011–1014 (2006). added only once, when initially embedding the cells, and compared 10. Desgrosellier, J. S. et al. Integrin alphavbeta3 drives slug activation and stemness against cells receiving the same volume of vehicle (PBS). in the pregnant and neoplastic mammary gland. Dev. Cell 30, 295–308 (2014). 11. Sun, Q. et al. Proapoptotic PUMA targets stem-like breast cancer cells to suppress Orthotopic breast cancer metastasis. J. Clin. Invest. 128, 531–544 (2018). 12. Prat, A. et al. Phenotypic and molecular characterization of the claudin-low Tumors were generated by injection of a limiting number of sorted HCC38 intrinsic subtype of breast cancer. Breast Cancer Res. 12, R68 (2010). cells diluted in 50 μl sterile PBS and injected into the inguinal fat pads of 8- 13. Prat, A. et al. Characterization of cell lines derived from breast cancers and normal to 10-week-old adult female nonobese diabetic/severe combined immu- mammary tissues for the study of the intrinsic molecular subtypes. Breast Cancer nodeficiency/interleukin-2 receptor γ chain knockout (NSG) mice (pur- Res. Treat. 142, 237–255 (2013). chased from UCSD Animal Care Program colony). All mice were monitored 14. Creighton, C. J. et al. Residual breast cancers after conventional therapy display weekly for tumor formation by gentle palpation. Most tumors formed mesenchymal as well as tumor-initiating features. Proc. Natl Acad. Sci. USA 106, within 8−10 weeks. Tumor volume was measured with calipers twice 13820–13825 (2009). weekly by a blinded observer. The experiment was concluded, and mice 15. Shipitsin, M. et al. Molecular definition of breast tumor heterogeneity. Cancer Cell were sacrificed just prior to the tumors reaching the maximum allowable 11, 259–273 (2007). size of 2 cm or 14 weeks, whichever came first. Estimated tumor-initiating 16. Lawson, D. A. et al. Single-cell analysis reveals a stem-cell program in human cell frequencies were calculated with the Extreme Limiting Dilution (ELDA) metastatic breast cancer cells. Nature 526, 131–135 (2015). web-based tool http://bioinf.wehi.edu.au/software/elda/. Primary tumor 17. Asselin-Labat, M. L. et al. Gata-3 is an essential regulator of mammary-gland mass was determined by assessing the wet weight of the resected tumors. morphogenesis and luminal-cell differentiation. Nat. Cell Biol. 9, 201–209 (2007). The remaining tumor-free mice were harvested at 14 weeks and the 18. Lim, E. et al. Transcriptome analyses of mouse and human mammary cell sub- populations reveal multiple conserved genes and pathways. Breast Cancer Res. absence of any detectable tumor was confirmed by whole-mount staining. 12, R21 (2010). 19. Murrow L. M. et al. Changes in epithelial proportions and transcriptional state Ethics underlie major premenopausal breast cancer risks. Preprint at bioRxiv https://doi. All mouse studies described were approved by the UCSD Institutional org/10.1101/430611 (2020). Animal Care and Use Committee (IACUC) and performed in accordance 20. Munoz, R. et al. Highly efficacious nontoxic preclinical treatment for advanced metastatic breast cancer using combination oral UFT-cyclophosphamide metro- with the guidelines set forth in the NIH’s Guide for the Care and Use of nomic chemotherapy. Cancer Res. 66, 3386–3391 (2006). Laboratory Animals (National Academies Press, 2011). Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 5 Q. Sun et al. 21. Nummela, P. et al. Transforming growth factor beta-induced (TGFBI) is an anti- Program (TRDRP) High Impact Research Project Award 28IR-0054 (to J. S. D.). We would adhesive protein regulating the invasive growth of melanoma cells. Am. J. Pathol. also like to acknowledge the flow cytometry shared resource facility in the UCSD Moores 180, 1663–1674 (2012). Cancer Center. Bulk RNA-Seq was conducted at the IGM Genomics Center, University of 22. Desgrosellier, J. S. et al. An integrin alpha(v)beta(3)-c-Src oncogenic unit promotes California, San Diego, La Jolla, CA (MCC grant # P30CA023100). anchorage-independence and tumor progression. Nat. Med. 15,1163–1169 (2009). 23. Ye, X. et al. Distinct EMT programs control normal mammary stem cells and tumour-initiating cells. Nature 525, 256–260 (2015). AUTHOR CONTRIBUTIONS 24. Morel, A. P. et al. A stemness-related ZEB1-MSRB3 axis governs cellular pliancy Q.S. helped design the project, performed experiments, analyzed data, and and breast cancer genome stability. Nat. Med. 23, 568–578 (2017). generated figures. Y.W., B.P., and G.L. designed and conducted experiments and 25. Maturi, V., Enroth, S., Heldin, C. H. & Moustakas, A. Genome-wide binding of analyzed data. O.H. and A.O. analyzed the RNA-Seq data and generated figures. J.S.D. transcription factor ZEB1 in triple-negative breast cancer cells. J. Cell Physiol. 233, designed the project, analyzed the data, supervised the overall project, and wrote the 7113–7127 (2018). manuscript. 26. Nguyen, Q. H. et al. Profiling human breast epithelial cells using single cell RNA sequencing identifies cell diversity. Nat. Commun. 9, 2028 (2018). 27. Zhang, Y. et al. TGFBI deficiency predisposes mice to spontaneous tumor COMPETING INTERESTS development. Cancer Res. 69,37–44 (2009). The authors declare no competing interests. 28. Cleary, J. M., Aguirre, A. J., Shapiro, G. I. & D’Andrea, A. D. Biomarker-guided development of DNA repair inhibitors. Mol. Cell 78, 1070–1085 (2020). 29. Noordermeer, S. M. & van Attikum, H. PARP inhibitor resistance: A tug-of-war in ADDITIONAL INFORMATION BRCA-mutated cells. Trends Cell Biol. 29, 820–834 (2019). 30. Joshi, P. A. et al. Progesterone induces adult mammary stem cell expansion. Supplementary information The online version contains supplementary material Nature 465, 803–807 (2010). available at https://doi.org/10.1038/s41523-021-00375-w. 31. Asselin-Labat, M. L. et al. Control of mammary stem cell function by steroid hormone signalling. Nature 465, 798–802 (2010). Correspondence and requests for materials should be addressed to Jay S. 32. Davis, R. T. et al. Transcriptional diversity and bioenergetic shift in human breast Desgrosellier. cancer metastasis revealed by single-cell RNA sequencing. Nat. Cell Biol. 22, 310–320 (2020). Reprints and permission information is available at http://www.nature.com/ 33. Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29,15–21 reprints (2013). 34. Patro, R., Duggal, G., Love, M. I., Irizarry, R. A. & Kingsford, C. Salmon provides fast and Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims bias-aware quantification of transcript expression. Nat. Methods 14, 417–419 (2017). in published maps and institutional affiliations. 35. Smedley, D. et al. BioMart−biological queries made easy. BMC Genomics 10, 22 (2009). 36. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014). 37. Liberzon, A. et al. The Molecular Signatures Database (MSigDB) hallmark gene set Open Access This article is licensed under a Creative Commons collection. Cell Syst. 1, 417–425 (2015). Attribution 4.0 International License, which permits use, sharing, 38. Barbie, D. A. et al. Systematic RNA interference reveals that oncogenic KRAS- adaptation, distribution and reproduction in any medium or format, as long as you give driven cancers require TBK1. Nature 462, 108–112 (2009). appropriate credit to the original author(s) and the source, provide a link to the Creative 39. Nilsen, G. et al. Copynumber: Efficient algorithms for single- and multi-track copy Commons license, and indicate if changes were made. The images or other third party number segmentation. BMC Genomics 13, 591 (2012). material in this article are included in the article’s Creative Commons license, unless 40. Hu, Y. & Smyth, G. K. ELDA: Extreme limiting dilution analysis for comparing indicated otherwise in a credit line to the material. If material is not included in the depleted and enriched populations in stem cell and other assays. J. Immunol. article’s Creative Commons license and your intended use is not permitted by statutory Methods 347,70–78 (2009). regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/. ACKNOWLEDGEMENTS This research was supported by funding from the National Natural Science Foundation of China Grants 82073271 and 81702362 (to Q. S.) and a Tobacco-Related Disease Research © The Author(s) 2022 npj Breast Cancer (2022) 5 Published in partnership with the Breast Cancer Research Foundation

Journal

npj Breast CancerSpringer Journals

Published: Jan 13, 2022

References