The ELEANOR noncoding RNA expression contributes to cancer dormancy and predicts late recurrence of estrogen receptor‐positive breast cancer

Abstract The recurrence risk of estrogen receptor (ER)‐positive breast cancer remains high for a long period of time, unlike other types of cancer. Late recurrence reflects the ability of cancer cells to remain dormant through various events, including cancer stemness acquisition, but the detailed mechanism is unknown. ESR1 locus enhancing and activating noncoding RNAs (ELEANORS) are a cluster of nuclear noncoding RNAs originally identified in a recurrent breast cancer cell model. Although their functions as chromatin regulators in vitro are well characterized, their roles in vivo remain elusive. In this study, we evaluated the clinicopathologic features of ELEANORS, using primary and corresponding metastatic breast cancer tissues. The ELEANOR expression was restricted to ER‐positive cases and well‐correlated with the ER and progesterone receptor expression levels, especially at the metastatic sites. ELEANORS were detected in both primary and metastatic tumors (32% and 29%, respectively), and frequently in postmenopausal cases. Interestingly, after surgery, patients with ELEANOR‐positive primary tumors showed increased relapse rates after, but not within, 5 years. Multivariate analysis showed that ELEANORS are an independent recurrence risk factor. Consistently, analyses with cell lines, mouse xenografts, and patient tissues revealed that ELEANORS upregulate a breast cancer stemness gene, CD44, and maintain the cancer stem cell population, which could facilitate tumor dormancy. Our findings highlight a new role of nuclear long noncoding RNAs and their clinical potential as predictive biomarkers and therapeutic targets for late recurrence of ER‐positive breast cancer.


| INTRODUC TI ON
Despite advances in diagnosis and treatment, recurrence remains a fatal problem for breast cancer patients. Breast cancer is heterogeneous and classified into subtypes based on the expression of ER, PgR, and HER2. 1,2 Although the underlying mechanism is unknown, ER-negative subtypes typically relapse within 3 years of diagnosis, whereas ER-positive subtypes have persistent risk of recurrence beyond 5 years and up to 20 years. 3 Remarkably, the 15-year recurrence and mortality rates are comparable in patients who were diagnosed with ER-positive or ER-negative subtypes at early stages. 4 Late recurrence reflects the capacity of circulating tumor cells or DTCs to be in a dormant state for a long period, which eventually progress to activate metastatic growth and cause a fatal condition. 5,6 Dormant metastasis has been proposed to be linked to BCSCs, characterized as having high tumorigenicity and self-renewal ability, which express the adhesion molecule CD44 but not CD24. 7 However, the mechanisms underlying the biology of dormancy remain unclear, and new diagnostic markers are required to identify patients at high risk of late recurrence.
Deep sequencing of mammalian transcripts has revealed the presence of numerous noncoding RNAs that are not translated into proteins. Many of them are dysregulated in cancers, and act as either oncogenic or tumor-suppressing factors. Nuclear noncoding RNAs often regulate gene expression through various mechanisms, including epigenetics and 3D genome architectures. [8][9][10] Therefore, noncoding RNAs are potential diagnostic markers and therapeutic targets.
ELEANORS are a cluster of noncoding RNAs originally found in a model of aromatase inhibitor therapy-resistant breast cancer, the LTED cells. These cells were established from ER-positive breast cancer cell lines, including MCF7 and HCC1428. 11 ELEANORS are transcribed from a TAD of approximately 1 Mb (chr6: 151,750,000-152,750,000), which contains four genes including the ER-coding ESR1 gene. 12 In the LTED nucleus, ELEANORS accumulate around the ESR1 gene to form an ELEANOR cloud and activate the entire ELEANOR TAD. Among the ELEANORS, characterizations of u-ELEANOR, 11 pa-ELEANOR, 12 and ELEANOR2 13 have revealed their importance for ELEANOR cloud formation and transcriptional activation of the genes in the ELEANOR TAD. 14  In the present study, we evaluated the clinicopathologic features of ELEANORS in a wide range of clinical breast cancer samples, and identified a novel correlation between the noncoding RNAs in primary tumors and late recurrence. We also obtained mechanistic insight into cancer dormancy, including stemness gene activation by nuclear noncoding RNAs.

| Patients
We used clinical breast cancer samples from two independent cohorts: (1)

| Immunohistochemical analysis and DISH
Immunohistochemistry and DISH were carried out and evaluated as described in Appendix S1. The Abs used in this study are listed in Table S1.
We used RNA-FISH on 4 µm sections using a ZytoLight FISH

| Cell culture and siRNA and LNA transfections
HCC1428 cells (ATCC) and HCC1428-LTED cells were cultured as previously described. 12 siRNA and LNA transfections were carried out with RNAiMAX (Invitrogen). The siRNA and LNA sequences are listed in Table S2. The cells were analyzed 48 h after transfection by qRT-PCR, immunoblotting, and Transwell assay.  Detailed methods for selection of clinical tissue sections, tissue microarray preparation, classification of breast cancer subtypes, microscopic analysis, qRT-PCR analysis, immunoblotting, tumorsphere assay, Transwell assay, and CCK-8 assay are described in Appendix S1. Primers used for qRT-PCR analysis are listed in Table S3.

| Statistical analysis
Pearson's χ 2 -test was used for two-group comparisons and the ttest was used to compare the means among groups. Spearman's rank correlation coefficient (rs) was used to investigate correlations between two ranks and interpreted as follows: very weak

| ELEANOR cloud is correlated with ER expression in human breast cancer tissues
To investigate the clinical significance of ELEANORS, we assessed ELEANOR expression in 185 primary breast cancers with pathological sizes from 2 to 5 cm (pT2), including all the main subtypes. Intense ELEANOR FISH signals were often detected in the nuclei of carcinoma cells and classified as follows: score 0, score 1 (small ELEANOR), and score 2 (large ELEANOR; Figure 1A).
Furthermore, score 0 was considered as ELEANOR−, and scores 1 and 2 were considered as ELEANOR+. After excluding seven indeterminate cases with samples that were unsuitable for FISH procedures, we identified 58 cases (33%) with ELEANOR+ tumors ( Figure 1A).

The ELEANOR expression was observed in luminal A-like,
luminal B-like, and luminal-HER2 type breast cancers, which were all ER-positive ( Figure 1B) and was absent from ER-negative subtypes, HER2, and triple-negative breast cancer. We then focused only on the ER-positive group (n = 141) and found that the ELEANOR expression was significantly frequent in tumors from patients who were postmenopausal or over 50 years old ( Table 1). The presence of an ELEANOR cloud was often correlated with ER expression ( Figure 1C). The common Allred scoring system for ER using IHC showed that higher ELEANOR scores were correlated with higher proportions of ER-positive cells (ER proportion score) and higher ER staining intensities (ER intensity score) (Spearman's rank correlation test, rs = 0.30, rs = 0.25, respectively; Figure 1D). These results are consistent with previous in vitro observations 11 that ELEANORS are prevalent in LTED cells and overexpressing ER through the chromatin-regulating function of ELEANORS.

| ELEANOR expression could facilitate late recurrence in ER-positive breast cancer
Assessments of the same ER-positive group (n = 141) showed that patients with ELEANOR+ primary tumors have worse RFS and OS than ELEANOR− patients ( Figure 1E). Among the ER-positive subtypes, the difference between ELEANOR+ and − was more significant in luminal B-like subtype, but not luminal A-like subtype ( Figure S1A). We then divided the Kaplan-Meier survival curves of Figure 1E into early (0-5 and 0-8 years after surgery for RFS and OS, respectively) and late (>5 and >8 years after surgery for RFS and OS, respectively) time windows ( Figure S1B). We found that ELEANOR+ was prognostic for late (log-rank test, p = 0.006) but not early (log-rank test, p = 0.67) recurrence risk ( Figure S1B, top).
The same was true for OS ( Figure S1B, bottom panel). Furthermore, the patients with ELEANOR scores of 1 and 2 had similar worse RFS beyond 5 years after surgery, compared to those with an ELEANOR score of 0 ( Figure S2).
These results demonstrate that ELEANORS are involved in the late recurrence of ER-positive breast cancer. A multivariate analysis adjusting for grade and HER2 status showed that ELEANOR expression could serve as an independent recurrence risk factor of ER-positive breast cancer patients (hazard ratio = 2.40, p = 0.014; Table 2). Furthermore, subgroup analyses of patients who received adjuvant endocrine therapies also showed that ELEANOR+ patients had higher recurrence rates beyond 5 years after surgery ( Figure S1C). These results suggest that ELEANORS play a pivotal role in the acquisition of resistance to adjuvant endocrine therapy .

| ELEANORS maintain ER and PGR expression after metastasis
To investigate the involvement of ELEANORS in the recurrence in more detail, we next used 167 paired samples from ER-positive primary tumors and their corresponding metastases. We undertook an ELEANOR FISH analysis, excluded 28 cases that were unsuitable for the FISH procedures, and analyzed the remaining 139 cases.
We found that ELEANORS were expressed in 45 cases (32%) of primary tumors and 40 cases (29%) of metastatic sites. Consistent with the results in Figure 1E, a set of these clinical samples showed that the patients with ELEANOR+ primary tumors relapsed later than ELEANOR− patients ( Figure 2A). This again supports the idea that ELEANORS are involved in late recurrence. The analysis of this dataset also indicated that ELEANORS are overrepresented in lung metastases (Tables S4 and S5).
We then followed up the changes in ELEANOR expression for each person ( Figure 2B), as all patients in this group eventually expe-  Figure 2C).

| ELEANORS activate the CD44 gene and breast cancer stemness
Late recurrence is supported by several factors, including tumor dormancy, cancer stemness, and resistance to endocrine therapy. cells had higher sphere-forming ability as compared to the original HCC1428 cells, suggesting the presence of BCSCs ( Figure 4A).
We knocked down the ELEANOR2 RNA, the most abundant member of the ELEANORS, with LNA and then analyzed the effect on CD44/CD24 expression with a FACS. As shown in Figure 4B, the ELEANOR2 knockdown significantly decreased the BCSC population (CD44 + /CD24 −/low ).
The qRT-PCR analysis revealed that the CD44 mRNA, as well as two specific ELEANORS, ELEANOR2 and pa-ELEANOR, 11,13 are overexpressed in HCC1428-LTED cells, as compared to HCC1428 cells ( Figure 4C). Expression of CD44 mRNA decreased with knockdowns of those ELEANORS ( Figure 4D). The immunoblot analysis revealed that CD44 protein expression also correlated with ELEANOR expression ( Figure 4E). Furthermore, the FACS analysis showed that the ELEANOR knockdown decreased the CD44+ population in HCC1428-LTED cells ( Figure 4F).

MCF7-LTED cells, another ELEANOR-expressing cell line, also
showed higher sphere-forming ability ( Figure S3A), and higher expression levels of CD44 mRNA ( Figure S3B) and protein ( Figure S3D), than the parental MCF7 cells. The effect of the ELEANOR knockdown on CD44 expression in MCF7-LTED cells was significant at the mRNA level ( Figure S3C), but had minor to no effects at the protein level ( Figure Figure 5A).

TA B L E 4 Association between ELEANOR expression and pathological parameters in 139 pairs of primary and metastatic breast cancers
These findings implied that ELEANORS play a role in cancer stemness through upregulation of CD44.

| ELEANORS upregulate the CD44 gene in mouse xenografts and clinical breast cancer
To further explore the characteristics of ELEANORS in vivo,  Figure 5E).
To test whether the elevated CD44 expression occurs near the ELEANOR cloud, we simultaneously visualized ELEANORS and the sites of CD44 transcription by RNA-FISH, using tumors derived from HCC1428-LTED cells that were xenografted into ovariectomized mice. We found that CD44 more frequently overlapped with ELEANORS, as compared to the control centromere 6 ( Figure 5F).
This finding suggested that ELEANORS could upregulate CD44 gene expression by providing a transcriptionally active nuclear environment. 11 However, the possibility of indirect effect is not excluded.
To investigate whether the CD44 expression state is correlated with ELEANOR expression in clinical samples, we next analyzed CD44 expression in pT2 primary ER-positive breast cancer tissues (n = 141), as shown in Figure 1E. A representative IHC analysis of CD44 revealed more CD44-positive cells in the ELEANOR+ group (scores 1 and 2; Figure 6A). The Allred scoring system showed very weak to weak correlations between the ELEANOR scores and the proportion and intensity scores of the CD44 expression ( Figure 6B;

| DISCUSS ION
In the present study, we have evaluated clinicopathologic features of ELEANORS in primary and metastatic breast cancer. We found that 33% of primary tumors from all subtypes were ELEANOR+, and exclusively ER-positive. Approximately 41% of ER-positive primary tumors were ELEANOR+, which has higher recurrence risk beyond 5 years after surgery. We found a strong correlation between the presence of ELEANORS and active ESR1 and CD44, which might both contribute to a long dormant status ( Figure 6C). Our study provides the long-anticipated molecular mechanisms and a predictive biomarker of late recurrence.
Late recurrence reflects the capacity of DTCs remaining in a dormant state for protection from cancer therapeutics or unfavorable microenvironments, such as immune surveillance, while also maintaining their tumorigenic features for final outgrowth to macrometastasis. 5,22,23 Indeed, several studies implied that BCSCs, which are the majority of DTCs, play critical roles in the acquisition of resistance to chemotherapy 24 and endocrine therapy. 25,26 The resistance in endocrine therapy is regulated by a complex network 27 that includes modulation of ER signaling and activation of growth factor signaling factors, 28,29 such as NOTCH, 30 Wnt, 31 and Hedgehog, 32 through enrichment of BCSCs. For ER signaling, the F I G U R E 4 ELEANORS activate CD44 gene expression for breast cancer stemness. (A) Higher tumorsphere-forming ability is shown in HCC1428-long-term estrogen deprivation (LTED) cells than in HCC1428 cells. Images of tumorspheres were captured (left) and quantified were proposed. In this study, the multivariate analysis showed that ELEANORS are an independent recurrence risk factor of ER-positive breast cancer. This shed light on the potential of ELEANORS for predicting the recurrence risk beyond 5 years.
Among ER-positive breast cancer patients, we found that ELEANORS were present at similar rates in both primary tumors and metastatic sites (29%-32%). These results were somewhat surprising, because ELEANORS were originally detected as being enhanced in a recurrent ER-positive breast cancer model, LTED cells, which were estrogen-deprived for a long period to recapitulate the AI therapy. 11 In this study, we found a slight correlation between ELEANOR expression and the administration of adjuvant AI therapy, which was not statistically significant (Table S7). We also detected a significant correlation between ELEANOR expression and postmenopausal sta-  (Figures 4A and 5A) and tumorigenic features in ovariectomized mice ( Figure 5B), suggesting the correlation with BCSCs. Taken together, we propose a working model in which ELEANORS could be involved in late recurrence ( Figure 6C). In addition to maintaining an equilibrium between cell proliferation and apoptosis ( Figure 6C, left and middle 12 ), ELEANORS could activate the CD44 gene. This would promote the endowment of cancer cells with dormant BCSC properties to survive and eventually induce outgrowth to active metastasis, accompanied by the proliferation abilities resulting from ER activation ( Figure 6C, right). Nuclei with ELEANOR signals overlapping with the CEN6 (blue bar) or CD44 (red bar) signals were quantified (right). p values were calculated using the two-tailed Fisher's exact test. (A-C) Each value represents the mean ± SE. p values were determined by t-tests: *p < 0.05, ***p < 0.001 In summary, ELEANORS are proposed as a new prediction marker and therapeutic target against late recurrence in ER-positive breast cancer. Our results also provide new insights into research regarding cancer initiation, progression, and metastasis regulated by noncoding RNAs.

ACK N OWLED G M ENTS
We thank the members of Saitoh's laboratory and Dr. Tomoyuki Kitagawa (JFCR) for discussions. We greatly appreciate Drs. Rika

R E FE R E N C E S
F I G U R E 6 ELEANORS could play a role in late recurrence through CD44 gene activation. (A) Representative H&E (top), immunohistochemistry (IHC) for CD44 (middle), and ELEANOR FISH (bottom) for tissues with ELEANOR scores of 0-2. Scale bars, 100, 100, and 10 μm (top to bottom, respectively). (B) Mosaic plots showing the association between ELEANOR scores and CD44 expression levels, determined with IHC. The CD44 proportion score (left) and intensity score (right) were determined by the Allred scoring system. Actual numbers detected are indicated in each box. rs, Spearman's rank correlation coefficient. (C) A proposed model. The equilibrium mechanism between cell proliferation (estrogen receptor [ER] signaling activation, middle) and apoptosis (FOXO3 gene activation, left) through ELEANORS was previously reported. 12 This study showed a correlation between ELEANORS and active ESR1 (middle) and CD44 (right) in clinical tissues. All of these properties contribute to late recurrence through tumor dormancy. TAD, topologically associating domain