Association of CD44−/CD24− Breast Cancer Cells with Late Stage Tumor Recurrence

Tumor metastasis remains the main cause of breast cancer-related deaths, especially the later breast cancer distant metastasis. This study assessed CD44−/CD24− tumor cells in 576 tissue specimens for associations with clinicopathological features and metastasis and then investigated the underlying molecular events. The data showed that level of CD44−/CD24− cells was associated with later postoperative distant tumor metastasis. Furthermore, CD44−/CD24− triple negative cells could spontaneously convert into CD44+/CD24− cancer stem cells (CSCs) with properties similar to CD44+/CD24− CSCs from parental MDA-MB-231 cells in terms of gene expression, tumor cell xenograft formation, and lung metastasis in vitro and in vivo. Single-cell RNA sequencing identified RHBDL2 as a regulator that enhanced spontaneous CD44+/CD24− CSC conversion, whereas knockdown of RHBDL2 expression inhibited YAP/NF-κB signaling and blocked spontaneous CD44−/CD24− cell conversion to CSCs. These data suggested that the level of CD44−/CD24− tumor cells could predict breast cancer prognosis, metastasis, and response to adjuvant therapy.


Introduction
Breast cancer is the most prevalent malignancy in women, and its incidence is increasing worldwide, especially in developed countries (Bianchini et al., 2016;Siegel et al., 2020).
Different treatment strategies, such as surgical resection, hormone therapy, targeted therapy, radiation therapy, and chemotherapy, have greatly improved the survival of breast cancer patients (Bray et al., 2018;Burstein et al., 2014;Early Breast Cancer Trialists' Collaborative, 2015;Khan et al., 2012;Saini et al., 2012). However, as such advancements prolong patients' survival, they also lead to an unfortunate and considerable number of patients facing the risk in developing later breast cancer metastasis, even 20-40 years after breast cancer diagnosis (Sharma, 2018). Notably, breast cancer metastasis occurring 5-8 years after initial surgical resection has become a significant cause of treatment relapse, tumor progression, and poor survival of patients (Nishimura et al., 2013); thus, further research on the underlying molecular mechanisms and gene alterations could help us to identify novel biomarkers and treatment strategies to effectively control breast cancer metastasis and progression. To date, clinical strategies for breast cancer treatment remain suboptimal. Currently, the best option available is continuous tamoxifen treatment for 10 years, which has been shown to reduce cancer recurrence and mortality of patients; however, a significant proportion of patients may be overtreated (Bianchini et al., 2016;O'Conor et al., 2018). In this regard, we also urgently need to discover novel biomarkers to predict treatment effectiveness and to improve treatment success and prognosis among breast cancer patients.
Indeed, tumor metastasis is a multistep process, involving tumor cell escape from the primary site, migration into neighboring tissues, extravasation, survival, and colonization, all leading to the formation of new tumor lesions at a secondary site (Drabsch and ten Dijke, 2011;Klein, 2008;Scott et al., 2012;Syn et al., 2016). The rate-limiting step of cancer metastasis is tumor growth at the secondary site, because the tumor lesion initially lacks sufficient vasculature to provide nutrients for cancer cell growth. Thus, the newly arrived tumor cells may grow to a certain size in the new and harsh microenvironment and then experience growth arrest in that organ. However, once they regain their proliferative ability, late metastasis will occur (Langley and Fidler, 2007). Molecularly, CD44+/CD24-tumor cells from the breast primary tumor lesion are associated with distance metastasis (Abraham et al., 2005), and these cells display increased motility and invasiveness (Liu et al., 2010) similar to chemoresistant cancer stem cells (CSCs) (Velasco-Velazquez et al., 2011). Previous studies have shown that CD44 + /CD24breast CSCs might be a dominant factor in relapse of triple negative breast cancer (TNBC), due to their possession of potent self-renewal and differentiation capacities to differentiate into mature CD44 -/CD24 -, CD44 + /CD24 + , and CD44 -/CD24 + cancer cells (Geng et al., 2014;Wang et al., 2014). Indeed, injection of up to 1000 breast CSCs was able to generate a solid tumor mass in immunocompromised mice (Chaffer et al., 2011;Iliopoulos et al., 2011). Thus, the number of breast CSCs in the secondary site could affect the efficiency of early metastasis formation, and stem cells are more prone to be resistant to chemotherapy (De Angelis et al., 2019). Moreover, previous studies reported that non-CSCs, such as CD44 -/CD24 -TNBC cells, are able to spontaneously convert into CSCs to renew the CSC pool, resulting in chemoresistance (Gruber et al., 2016;Kim et al., 2015;Ye et al., 2018). Thus, the dormant CD44 -/CD24tumor cells that have previously been colonized in the metastatic site may be able to spontaneously convert to CSCs to regain proliferative ability and drug resistance, resulting in later tumor metastasis.
Therefore, detection of CD44 -/CD24cells may be useful in the prediction of later breast cancer metastasis.
In the present study, we aimed to identify the molecular mechanisms by which CD44 -/CD24cell to CSC conversion promotes later breast cancer metastasis. We first performed a retrospective analysis of CD44 -/CD24breast cancer cells in tissue specimens from patients 5 enrolled from three academic medical centers to identify any associations between the presence of these cells and postoperative tumor metastasis. We then performed in vitro and in vivo experiments to confirm CD44 -/CD24cell conversion to CD44 + /CD24 -CSCs and then assessed the properties of the converted CSCs in vitro and in vivo. This study provides novel insight into the role of CD44 -/CD24tumor cells in later breast cancer metastasis and into the potential use of CD44 -/CD24cells as a biomarker to predict survival and metastasis in breast cancer patients or targeting of RHBDL2 as a novel therapeutic approach in the future.

Results
Association of high CD44 -/CD24tumor cell level with late breast cancer distant metastasis CD44 + /CD24breast CSCs were previously found to not have a significant impact on the prognosis of breast cancer patients (Kaverina et al., 2017;Mylona et al., 2008). In the present study, we performed a retrospective analysis of CD44 + /CD24cell levels in 576 patients using immunofluorescence staining to examine their association with clinicopathological factors. The patients were divided into a training set (n= 355) and testing set (n= 221) ( Table   S1) for assessment of cell membrane proteins CD44 and CD24 (Fig. S1), leading to four types of breast cancer cells (CD44 -/CD24-, CD44 + /CD24 -, CD44 -/CD24 + , and CD44 + /CD24 + ), as shown in Fig. S2A. The level of CD44 -/CD24cells was significantly higher in breast cancer tissues of patients with postoperative tumor metastasis in the training set (P<0.0001; Fig. 1A). The receiver-operating characteristic curve analysis showed a decision threshold of 19.5% CD44 -/CD24cancer cells, and thus, we used this cut-off-point to perform a subgroup analysis (the discrimination criteria are shown in Fig. 1B). In the 6 training set of patients, the metastasis rate was 1.97-fold higher in patients with a high level of CD44 -/CD24breast cancer cells (>19.5%) than in those with a low level of CD44 -/CD24tumor cells (<19.5%). Similar data were obtained after analysis of the three breast cancer subtypes (i.e., luminal: 63.1% vs. 32.58%; HER-2: 49.02% vs. 15%; TNBC: 56.96% vs. 20.28% for high vs. low CD44 -/CD24tumor cells; Fig. 1C). Our univariate analysis showed that the CD44 -/CD24cancer cell subgroup, CSCs percentage, lymph node metastasis, N stage, estrogen receptor and progesterone receptor status, and molecular subtype were all predictors of DFS (Table S2). Furthermore, the DFS of patients with a high CD44 + /CD24or CD44 -/CD24tumor cell level was shorter than that of patients with a low frequency (Fig. 1E), which was also observed in the three breast cancer subtypes ( Fig. S2D and S3A). The kinetic curve pattern analysis also showed that patients with a high CSC percentage had a high risk of developing early tumor metastasis within the first 3 years after diagnosis (Fig. S3C). Some patients with high CD44 -/CD24tumor cell levels also possessed a high CSC percentage and showed a higher rate of postoperative metastasis. However, to exclude the effects of CSCs, we defined patients with <2% CD44 + /CD24and > 19.5% CD44 -/CD24tumor cells as the C1 group, and our data revealed that these C1 patients had a high risk of developing later tumor metastasis after 5-7 years (Fig. 1D).

Level of CD44 -/CD24breast cancer cells predicts success of postoperative TNBC treatment
We then confirmed our data using the testing set of patients and found that the metastasis rate was higher in patients with a high CD44 -/CD24tumor cell level among all breast cancer subtypes (luminal: 63.1% vs. 32.58%; HER-2: 49.02% vs. 15%; TNBC: 56.96% vs. 20.28% for high vs. low CD44 -/CD24cells; Fig. 2A). The DFS and OS of patients with a high CD44 -7 /CD24percentage were shorter than those of patients with a low CD44 -/CD24cell percentage (Fig. 2CD). Moreover, the DFS and OS of patients stratified by breast cancer subtype also were shorter with a high CD44 -/CD24percentage vs. low CD44 -/CD24percentage ( Figure S5A,B). The kinetic curve pattern analysis also confirmed that C1 patients had a later metastasis peak than all other patients (Fig. 2B).
Furthermore, we analyzed the predictive value of the CD44 -/CD24cell percentage for response to treatment. In the training set of patients, cases with a high CD44 -/CD24cell level after chemotherapy had a higher risk of developing tumor metastasis 4 years after initial diagnosis and treatment than did cases with a low CD44 -/CD24cell percentage (Fig. S3B).
Moreover, a low CSC percentage led to different risks for developing tumor metastasis 5 years after diagnosis and adjuvant chemotherapy between the C1 and C0 patients (Fig. S3D).
TNBC patients with high and low CD44 -/CD24cell levels also had different risks of metastasis 4 years after diagnosis and chemotherapy, indicating that chemotherapy per se did not effectively inhibit tumor progression (metastasis) (Fig. 2F). In this regard, detection of CD44 -/CD24cells in tumor tissues could be a valuable predictor of chemotherapy response.
The data from our testing group of patients confirmed these results ( Fig. 2E and S4CD). In addition, patients with luminal breast cancer treated with endocrine therapy also showed significantly different prognoses after stratification by percentage of CD44 -/CD24cells in both the training and testing groups of patients (Fig. 1FG).

Spontaneous conversion of CD44 -/CD24 -TNBC cells into CD44 + /CD24 -CSCs in vitro
Previous studies showed that non-CSCs are able to spontaneously convert into CSCs to renew the CSC pool, although it remained unclear whether the stem cells derived from the non-CSCs have the same biological behavior as the original stem cells (Najafi et al., 2019).
In the present study, we first designed the experiments illustrated in Fig. 3A to characterize 8 the percentages of different cell subtypes among MDA-MB-231 cells using cell culture and flow cytometric cell sorting of parental MDA-MB-468 cells. We found that the percentage of CD44 + /CD24 -CSCs was 18.7% and that of CD44 -/CD24 -CSCs was 73.4%. Next, we sorted the CD44 -/CD24cells and continued to culture them in the stem cell (SC) and normal complete L15 medium for 7 days. We found that 28.3% of CD44 -/CD24cells were able to spontaneously convert into CD44 + /CD24 -CSCs, which was higher than the 24.7% observed when the cells were cultured in the normal complete L15 medium (Fig. 3B). Similar results were observed in the parental MDA-MB-468 cells (Fig. S5). Moreover, the percentage of CD44 + /CD24 -CSCs was even higher after 21 days in culture than after 7 days in culture ( Fig.   3C), suggesting that CD44 -/CD24 -TNBC cells were able to spontaneously convert into CD44 + /CD24 -CSCs.

Similar properties of newly converted CSCs from CD44 -/CD24 -TNBC cells to those directly derived from TNBC cells in vitro
The clinical usefulness of CD44 -/CD24cells in breast cancer lesions is obvious; thus, we explored the underlying mechanism by assessing whether these newly converted DEGs associated with tumor cell differentiation and dedifferentiation after 24 h in cell culture (including RHBDL2, DSCC1, ZNF710, ATP8B3, and others; Fig. 6B and Fig. S6A). Our GO and KEGG analyses revealed that the reverse differentiation predominantly affected the expression levels of proteins involved in cell organelle formation, metabolism, signal transduction, and transcriptional regulation (Fig. S6BC). Furthermore, we performed Kaplan-Meier curve analysis and log-rank test to analyze associations between select DEGs and breast cancer prognosis using The Cancer Genome Atlas (TCGA) data ( Fig. 6C). This analysis showed that high expression of HIST1H4H, ZNF710, RHBDL2, DSCC1, ARL6IP1, and PPME1 mRNAs was associated with a poor prognosis in breast cancer patients, whereas low expression of SLC35F5, G2E3, ATP8B3, and MED22 mRNAs also was associated with a poor prognosis of breast cancer patients. Furthermore, we cultured breast cancer cells for 7 days and then divided them into CD44 + /CD24breast cancer stem cells, CD44 -/CD24double negative cells, and dedifferentiated stem cells for detection of the expression levels of different mRNAs. We found that RHBDL2 expression was high during tumor cell dedifferentiation, suggesting that RHBDL2 may play a key role in the process of breast cancer cell dedifferentiation.

Downregulated RHBDL2 expression inhibits YAP/nuclear factor (NF)-κB/interleukin (IL)-6 signaling and spontaneous CD44 -/CD24cell conversion to CD44 + /CD24 -CSCs
To further explore how RHBDL2 and the related gene signaling are involved in the control of spontaneous CD44 -/CD24cell conversion to CD44 + /CD24 -CSCs, we first assayed the levels of total YAP1 and phosphorylated YAP1 in MDA-MB-231 cells (Fig. 7A) and found that phosphorylated YAP1 was increased with RHBDL2 silencing. Previous studies revealed that the YAP signaling is able to suppress USP31 expression, a potent inhibitor of NF-κB signaling in cells (17,20). We thus further examined USP31 expression and NF-κB phosphorylation in CD44 -/CD24cells sorted from both the negative control (NC) and RHBDL2-silenced MDA-MB-231 cells. Our results showed that USP31 expression was significantly higher in CD44 -/CD24 -RHBDL2-silenced cells than in NC MDA-MB-231 cells, whereas the level of phosphorylated NF-κB was significantly lower in CD44 -/CD24 - We next selected the CD44 -/CD24cells using flow cytometric sorting and transfected them with an RHBDL2-specific siRNA. After 7 days, the cell content was observed by flow cytometric analysis. The results showed that the ratio of CD44 + /CD24 -CSCs dedifferentiated from NC cells was 6.5% and the ratio of siRHBDL2 cells converted to stem cells was 0.7% increase USP31 and decrease NF-κB signaling, and in turn, induce spontaneous conversion of CD44 -/CD24 -TNBC cells into CD44 + /CD24 -CSCs, whereas knockdown of RHBDL2 expression had the opposite effects in TNBC cells (Fig. 7G).

Discussion
Our current study revealed that (a) CD44 -/CD24 -TNBC MDA-MB-231 cells were able to spontaneously convert to CD44 + /CD24 -CSCs, which is consistent with previous studies showing that non-tumor and neoplastic non-stem cells and CD44 -/CD24breast cancer cells can spontaneously convert into CSCs in vitro (Italiano and Shivdasani, 2003;Zoppino et al., 2010); The dedifferentiation (conversion) of non-CSC tumor cells into CSCs also occurs in other types of human cancers, such as glioblastoma and intestinal stroma melanoma, and contributes to intra-and inter-tumor heterogeneity (Stepanova et al., 2003;Tzimas et al., 13 2006; Wei et al., 2019). Although the underlying molecular mechanisms remain to be defined, CSC plasticity and heterogeneity could induce tumor progression and resistance to therapy (Das et al., 2020;Kilmister et al., 2020;Martin-Castillo et al., 2013;Thankamony et al., 2020). For example, intratumoral heterogeneity is a major ongoing challenge for effective cancer therapy, while CSCs are responsible for intratumoral heterogeneity, therapeutic resistance, and metastasis, which may be because cancer cells exhibit a high level of plasticity and an ability to dynamically switch between CSC and non-CSC states or among different subsets of CSCs (Thankamony et al., 2020). Another previous study showed that the naturally high CD44 expression and low CD24 expression (Murohashi et al., 2010).
However, but further studies are needed to fully understand how RHBDL2 and the downstream YAP/NF-κB/IL-6 signaling mediate the regulation of this conversion.
Furthermore, our multicenter retrospective analysis of CD44 -/CD24cancer cells in breast cancer tissue specimens showed that the CD44 -/CD24cancer cell population was associated with postoperative breast cancer metastasis. Our multivariate Cox regression analysis revealed that percentages of CD44 -/CD24and CD44 + /CD24cells, tumor N stage, and molecular subtypes were all predictors of DFS for breast cancer patients. The percentage of CD44 -/CD24cancer cells could be a better predictor of later breast cancer metastasis (up to 12 years after initial breast cancer diagnosis) than of early metastasis (within 5 years). Our C1 subgroup of patients had a higher risk of later metastasis at 5-7 years after surgery. The ROC curve analysis showed that the percentage of CD44 -/CD24cells could predict later metastasis in cases with a low CSC percentage. These results further suggest that CD44 -/CD24cells could be able to spontaneously convert into CSCs and cause distant metastasis, although another theory of the clonal evolution also indicates that successive mutations accumulating in a given cell line may lead to clonal outgrowth in response to microenvironmental selection pressures (Meacham and Morrison, 2013).
In addition, our RNA-seq analysis identified several DEGs during the dynamic process of CD44 -/CD24 -TNBC cell conversion into CSCs, i.e., expression of RHBDL2, YAP1, and other genes were significantly increased. Consistently, YAP/USP31/NF-κB signaling was activated in CD44 -/CD24 -MDA-MB-231 cells, which is consistent with previous research showing that YAP1-mediated suppression of USP31 enhances NF-κB activity to promote sarcomagenesis (Kemeny and Fisher, 2018;Mehta et al., 2018). YAP1 can enhance CSC stemness in several types of human cancers, and aberrant YAP1 activation was associated with a low level of TNBC differentiation and poor survival of breast cancer patients (Bora-15 Singhal et al., 2015;Hansen et al., 2015). Our current data could indicate that upregulated RHBDL2 during CD44 -/CD24 -TNBC cell conversion into CSCs could enhance YAP expression to, in turn, inhibit USP31 and attenuate its inhibition of NF-κB signaling, leading to the enhanced conversion of CD44 -/CD24 -TNBC cells into CSCs.
In summary, postoperative breast cancer metastasis, especially later metastasis (e.g., ≥5 years after diagnosis), is an important unresolved issue. Our current results showed that patients with high CD44 -/CD24cell populations had a high risk of developing metastasis 5-7 years after diagnosis. Knockdown of RHBDL2 expression caused YAP1 reduction and enhanced USP31/NF-κB signaling, leading to inhibition of the conversion of CD44 -/CD24 -TNBC cells into CSCs. Furthermore, inhibition of the YAP1 signal could attenuate or even reverse the dedifferentiation process of TNBC cells. These findings provide novel insight into the molecular process of the dedifferentiation of non-CSCs into CSCs, which was been reported in glioblastoma, intestinal stroma melanoma, and other cancers, leading to cancer progression. Thus, our current findings suggest a novel postoperative therapeutic strategy for TNBC. Future studies will investigate targeting of RHBDL2 to control breast cancer resistance to chemotherapy.

Patients and tissue specimens
Paraffin-embedded cancer tissue samples were collected from 576 breast cancer patients who This study was approved by the ethics committee of all three hospital review boards, and each participant signed an informed consent form before being included in this study.

Immunofluorescence staining
Paraffin-embedded cancer tissue blocks from 576 breast cancer patients were collected and used to construct a tissue microarray (Edge and Compton, 2010) . The levels of CD44 and Additionally, tumor weight and sizes were measured in each group.

Statistical analysis
The data were expressed as the mean ± the standard deviation (SD), and the differences between the groups were analyzed by chi-squared or Student's t tests, as applicable. The statistics of all biological experiments were based on more than three independent replicated experiments. The DFS of each group of patients was estimated by the Kaplan-Meier curves and analyzed using the log-rank test. All statistical analyses were performed using SPSS 23.0 (SPSS, Inc., Chicago, IL, USA). A p-value less than or equal to 0.05 was considered statistically significant.

Conflicts of interest
The authors declare that there are no conflicts of interest in this work.