Role of Neural (N)-Cadherin in Breast Cancer Cell Stemness and Dormancy in the Bone Microenvironment

Simple Summary Breast cancer (BrCa) patients experience tumour recurrence 7–25 years after mastectomy. Dormant BrCa cells often home to the bone and, upon reactivation, lead to secondary lesions. We previously demonstrated that, in the bone marrow, dormant BrCa cells are located near the endosteal niche enriched in spindle-shaped N-cadherin-high Osteoblasts (SNOs), show a stem signature like the LT-HSCs and express high levels of Notch2. Here, we observed that Notch2High human BrCa MDA cells are also N-CadherinHigh. When injected in vivo, N-CadherinHigh MDA cells show lower aggressiveness, and higher ability to colonize the endosteal niche and to adhere to SNOs compared to N-CadherinLow MDA cells. These cells have features typical of cancer stem cells, displaying a gene signature similar to the HSCs. In contrast, N-cadherin has a negative role in mouse 4T1 cell stemness and HSC mimicry. In brief, our results identified a role of N-Cadherin in BrCa dormancy and stemness, also highlighting the differences between human and mouse BrCa cell lines. Abstract Breast cancer cells that interact with spindle-shaped N-Cadherin+ Osteoblasts (SNOs) are recognised to become dormant through a Notch2-dependent mechanism. We found that Notch2High human BrCa MDA-MB231 (MDA) cells also expressed high level of N-Cadherin. This prompted us to hypothesize that N-Cadherin could have a role in MDA-SNO interaction. Of note, the expression of N-Cadherin in MDA cells reduced tumour incidence and bone osteolysis in BrCa mouse model. Moreover, similarly to Notch2High MDA cells, the N-CadherinHigh MDA cells revealed a high expression of the canonical Haematopoietic Stem cell (HSC) markers, suggesting an HSC mimicry, associated with higher ability to form mammospheres. Interestingly, N-CadherinHigh MDA cells showed greater capacity to adhere to SNOs, while the inhibition of SNO-mediating MDA cell proliferation was unremarkable. To investigate whether these features were shared by mouse BrCa, we used the 4T1 cell line in which N-Cadherin expression was abolished and then rescued. At variance with MDA cells, 4T1 cells expressing N-Cadherin revealed that the latter was associated with a lower expression of the HSC marker, Cxcr4, along with a lower capacity to form mammospheres. Furthermore, the rescue of N-Cadherin expression increased cell-cell adhesion and reduced proliferation of 4T1 cells when they were co-plated with SNOs. In conclusion, we demonstrated that: (i) N-CadherinHigh and Notch2High MDA cells showed similar HSC mimicry and dormancy features; (ii) N-Cadherin mediated BrCa-SNO adhesion; (iii) N-Cadherin had a positive Notch2-dependent role on SNO-induced dormancy and HSC mimicry in MDA cells, and a negative role in 4T1 cell stemness and HSC mimicry.


Introduction
Breast cancer (BrCa) represents the second most common malignancy in women, with thousands of new cases diagnosed every year [1] and a five-year survival rate around 90% [1]. It is estimated that 20-45% of "disease-free" patients experience tumour recurrence

Breast Cancer Cell Culture
Human MDA-MB-231 (MDA) and mouse 4T1 breast cancer cell lines, parental or transfected with turbo-GFP (MDA GFP and 4T1 GFP ), were used for all experiments. Cells were cultured in high glucose Dulbecco's Modified Eagle Medium (DMEM) with the addition of 1% glutamine, 1% penicillin-streptomycin and 10% FBS.

Genetically Modified 4T1 Breast Cancer Cells
Parental 4T1 cells were evaluated for N-cadherin expression by cytofluorimetry revealing, as expected, the presence of a 4T1 N-cadherin high subpopulation (Supplementary Figure S1a). N-Cadherin expression was then subjected to non-homology mediated CRISPR/Cas9 knockout by plasmids carrying two gRNAs targeting the N-Cadherin gene co-transfected with a linear donor containing genes encoding for the GFP tag and puromycin resistance (Supplementary Figure S1b). After stable transfection, the integration of the linear donor was confirmed in two 4T1 N-cadherin knock-out cell populations (4T1 Cdh2KO1-GFP and 4T1 Cdh2KO2-GFP ) using a primer pair specific for puromycin, confirming the success of the transfection protocol (Supplementary Figure S1c). Non-transfected 4T1 cells were included in the analyses as control. Cytofluorimetric analyses further confirmed the absence of N-Cadherin expression in the GFP-positive cells (knock-out cells) (Supplementary Figure S1d-g). After validation of transfection, 4T1 Cdh2KO1-GFP and 4T1 Cdh2KO2-GFP were sorted by FACS using the GFP tag. Then, to generate a rescued model, the FACS-sorted 4T1 Cdh2KO1-GFP and 4T1 Cdh2KO2-GFP were stably transfected with the mouse Cdh2-turboGFP expression vector (4T1 Cdh2Res1-GFP and 4T1 Cdh2Res2-GFP ), which restored N-Cadherin expression (Supplementary Figure S1h). Cells transfected with the empty-turboGFP vector (4T1 Cdh2KO1-GFP and 4T1 Cdh2KO2-GFP ) were used as control. Finally, the 4T1 Cdh2Res2-GFP that showed a non-stable Cdh2 over-expression, and the relative control knock-out cells (4T1 Cdh2KO2-GFP ) were excluded from the analysis (Supplementary Figure S1i). On the contrary, the 4T1 Cdh2KO1-GFP and the 4T1 Cdh2Res1-GFP were used for the experiments and labelled 4T1 Cdh2KO-GFP and 4T1 Cdh2Res-GFP .

Mouse Primary Osteoblast Cell Isolation
Mouse primary osteoblasts were isolated from the calvariae of 8-day-old CD1 mice using three-step enzymatic digestion with a solution containing 25 mg/mL of porcine trypsin and 1 mg/mL of Clostridium histolyticum type IV collagenase in Hanks' Balanced Salt Solution. The supernatant from the first digestion, containing mainly fibroblasts, was discarded, while those from the second and the third digestions, enriched in primary murine osteoblasts, were centrifuged at 300× g for 8 min and the cells were then cultured in high glucose DMEM supplemented with 1% glutamine, 1% penicillin-streptomycin and 10% FBS. At confluence, cells were trypsinised and plated according to the experimental protocol.

Magnetic-Activated Cell Sorting (MACS)
MDA and 4T1 parental or genetically modified cells, and primary mouse osteoblasts were sorted using MACS. Cells were detached with sorting buffer containing DPBS, 5% BSA and 0.5 M of EDTA. Resuspended cells were incubated for 20 min at 4 • C using N-Cadherin or Notch2 biotinylated primary antibodies. Then, cells were incubated again in the same condition using streptavidin-conjugated magnetic microbeads and were eluted through the magnetic column to separate the antigen-depleted and antigen-enriched cell populations. The cells obtained from this procedure were used for RNA isolation, in vitro assays, and in vivo experiments. Supplier, product code and dilution of the primary and secondary antibodies used for the MACS are listed in Supplementary Table S2. 2.6. RNA Extraction, RNA Deep Sequencing (RNAdSeq) Analysis and Gene Expression RNA was extracted using TRIzol ® according to the manufacturer's instructions. The RNA quality was assessed using electrophoresis agarose gel and was quantified by Nanodrop ® using an absorbance of 260 nm wavelength. The RNA purity was assessed measuring the 260/280 nm ratio and 260/230 nm ratio for the protein and phenol presence respectively.
For RNA dSeq, 3 independent RNA preparations for Notch2 High and Notch2 Low MDA cells were precipitated in ethanol and sent to Omega Bioservice (Norcross, GA, USA) for the RNA dSeq analysis. The generated RNA dSeq datasets, containing the expression profile of 36,000 genes for each sample/condition, was interrogated to examine the expression of the CDH2 gene.
For conventional gene expression analyses, 1 µg of RNA was retro-transcribed using Revertaid First Strand cDNA Synthesis. Semiquantitative PCR was performed using OneTaq ® Hot Start 2X Master Mix, while real-time PCR was performed using Luna ® Universal qPCR Master. Primer sequences used to assess gene expression are listed in Supplementary Table S1. Full agarose gel pictures are shown in Figures S6-S8.

Flow Cytometry
Cells were detached with a sorting buffer containing DPBS, 5% BSA and 0.5 M of EDTA. Resuspended cells were incubated with primary antibodies against Notch2 and N-Cadherin for 1 h at 4 • C. Then, secondary incubation was performed using fluorochromeconjugated secondary antibodies; then, cells were analysed by the FACS Melody ® (BD) and FlowJO software. Unmarked cells were used to set the laser for the fluorescence threshold.
For the analysis of cells sorted from genetically modified 4T1 (4T1 Cdh2KO-GFP and 4T1 Cdh2Res-GFP ), the cell gaiting was performed using the GFP fluorescence. Supplier, product code and dilution of the primary and secondary antibodies used for the analyses are listed in Supplementary Table S1.

Animals
For the in vivo experiment, 4-week-old CD1 or Balb-C nude/nude (nu/nu) female mice were purchased from Charles River (Écully, France Human MDA and mouse 4T1 BrCa cells were injected into the left tibia of 4-week-old female Balb/c nu/nu immunocompromised (for human cells) or Balb/c immunocompetent (for mouse cells) mice (1 × 10 4 cells/0.01 mL PBS) anesthetized with intraperitoneal injection of 80 mg/kg of ketamine and 10 mg/kg of xylazine. Animals were monitored daily for body weight, food intake, behaviour, and survival.
To follow the progression of osteolytic lesions, mice were subjected to weekly X-ray analysis (X-ray parameters: peak kilovoltage [kVp] = 36 kV for 10 s) using a Cabinet X-ray system (Faxitron model no. 43855A; Faxitron X-Ray Corp., Buffalo Grove, IL, USA). At the end of the experiment, mice were subjected to final X-ray analysis and then sacrificed to perform anatomical dissection for the evaluation of bone and visceral metastases.

Limiting Dilution Assay (LDA)
Different dilutions (50,000 to 100 cells/mice) of human MDA cells MACS-sorted into N-Cadherin High and N-Cadherin Low subpopulations were subcutaneously injected in Balb/c nu/nu immunocompromised female mice anesthetized with intraperitoneal injection of 80 mg/kg of ketamine and 10 mg/kg of xylazine. Animals were monitored daily for body weight, food intake, behaviour, and survival. The tumor incidence was evaluated after 4 weeks from the injection of the tumor cells. The stem cell frequency was estimated using the Extreme Limiting Dilution Assay (ELDA) [30] available at https: //bioinf.wehi.edu.au/software/elda/ (accessed on 25 February 2022).

Micro-Computed Tomography (µCT) Analysis
Left tibias harvested from the tumour cell-injected mice were fixed in 4% formaldehyde for 48 h and then scanned by µCT SkyScan 1174. The scan was performed with a 9.80 µm resolution using the X-ray voltage of 50 kV. The Skyscan NRecon software was used to reconstruct the images using a modified Feldkamp algorithm. Three-dimensional (3D) analysis was carried out employing a marching cubes-type model with a rendered surface. The cortical bone parameters were calculated on 300 consecutive slides starting from 100 µm below the growth plate, where the osteolytic lesions were located. Pratt's algorithm was adopted to take 2D measurements. Threshold values were applied for segmenting trabecular bone. Bone cortical variables were selected according to Bouxsein et al. [31].

Histology
Left tibias were decalcified for 48 h in Osteodec and then embedded in paraffin. Livers were fixed in 4% paraformaldehyde and embedded in paraffin. Microtome sectioning was used to obtain tissue slices of 5-µm thickness. Liver sections were stained with haematoxylin and eosin while tibia sections were also processed for immunohistochemistry or immunofluorescence staining.

Immunohistochemistry and Immunofluorescence
For immunohistochemistry, mouse tibia sections and human primary BrCa tissue arrays were deparaffined and incubated with 0.07 M citrate buffer (pH 6) for 30 min at 96 • C and for 10 min at room temperature. The blocking was made with 3% H 2 O 2 and 5% BSA. Then samples were incubated overnight at 4 • C with primary antibodies against N-Cadherin or human pan-Cytokeratin AE1/AE3. The staining signals were revealed using the SignalStain ® Boost IHC Detection Reagent (HPR rabbit or mouse). Sections were For immunofluorescence, tissue sections or fixed cells (4% paraformaldehyde) were incubated with primary antibodies against human pan-Cytokeratin AE1/AE3, N-Cadherin or Ki67, either singularly or in combinations. Primary antibody incubations were carried out at room temperature for 1 h, then overnight at 4 • C. Then, incubations with secondary antibodies conjugated with AlexaFluor 488 or 594 were performed for 1 h at room temperature. Nuc-Spot ® or DAPI were used to stain the nuclei. The supplier, product code and dilution of the primary and secondary antibodies used for the analyses are listed in Supplementary Table S2.

Histomorphometry
Endosteal niche colonization analysis was performed for counting the number of cytokeratin positive cells in proximity of the endosteum (4 mm 2 in area, 50 µm away from the growth plate and 20 µm away from the endocortical surface) [12], and their distance from the endosteal surface was measured. For liver metastases, sections were evaluated for metastasis number/mm 2 and for the percentage of metastasis area over total tissue area. Primary BrCa tissue arrays were analysed counting the number of N-Cadherin + cells on total surface. All histomorphometric analyses were performed using the software Fiji ® by ImageJ (version 1.53).

Osteoblast/BrCa Cells Coculture Assay
Mouse primary calvarial osteoblasts were MACS-sorted into SNOs or NON-SNOs using anti-N-Cadherin-biotin antibody and Streptavidin-conjugated magnetic microbeads, as described above. For the knock-down experiment, MACS-sorted MDA cells were incubated with CDH2-or Scramble (SCR)-siRNA for 48 h before proceeding with the coculture. 7 × 10 4 -1 × 10 5 sorted cells were seeded in 96-well plates and incubated overnight in a humidified CO 2 incubator (5% CO 2 , 37 • C). The day after, the MDA GFP or the 4T1 GFP , MACS-sorted into N-Cadherin High and N-Cadherin Low or Notch2 High and Notch2 Low , or the non-sorted 4T1 Cdh2-KO-GFP and 4T1 Cdh2-Res-GFP were seeded on SNO or NON-SNO monolayers as above. After 1 h, cultures were extensively washed and the number of GFP + cells were counted using an Olympus IX inverted fluorescence microscope. The counting was repeated after 24, 48, 72 h. BrCa cell density in the cocultures were 1 × 10 3 for MDA GFP cells and 1 × 10 2 for 4T1 GFP cells.

Primary and Secondary Mammosphere Formation Assay
Primary mammosphere assays were performed using 8 × 10 3 suspended cells seeded in non-adhesive Petri dishes with serum-free DMEM, supplemented with 1% N2, 1% B27, 1% penicillin/streptomycin and 1% L-glutamine. They were incubated for 6 days in a humidified CO 2 incubator. For secondary mammospheres, the primary mammospheres were disaggregated using trypsin to obtain single-cell suspensions and cultured again under the same conditions. Imaging for the analysis was performed using the SXView Software (version 2.2.0.172). Mammosphere volume was calculated using the formula: (1)

TaqMan™ Array Mouse Stem Cell Pluripotency
Two micrograms of total RNA isolated from 4T1 Cdh2KO-GFP and the 4T1 Cdh2Res-GFP cells were retrotranscribed into cDNA using Revertaid First Strand cDNA Synthesis. Then 200 ng of cDNA per condition were loaded in the TaqMan™ Array Mouse Stem Cell Pluripotency plates (Applied Biosystem cat: 4414080) containing specific Fluorescein Amidite (FAM)tagged probes for mRNAs involved in mouse stemness and pluripotency. Real-time PCR was performed using the TaqMan™ Fast Advanced Master Mix (Applied Biosystem cat: Cancers 2022, 14, 1317 7 of 21 4444556). Gene expression was normalized by mouse Gapdh and expressed as fold vs. the 4T1 Cdh2KO-GFP . The real-time arrays were run in triplicate.

Statistical Analyses
Results are expressed as mean ± Standard Deviation (SD). Sample size is indicated in the figure legends. Groups' comparisons were performed carrying out independent samples Student's t-tests and non-linear regression, fitting with F-test when dealing with continuous parameters. Data from RNA dSeq were analysed using a false discovery rate (FDR)-adjusted p-value. To assess the distributional pattern of the BrCa cells in the bone marrow in relation to the endosteal surface we used a cumulative frequency distribution with Gaussian regression and the F-test. For the ELDA, the p-value was calculated using a Student's t-tests with 95% of confidence [30]. The statistical methods are indicated in the figure legends and the p values are indicated in the figures.

N-Cadherin and Notch2 Are Co-Expressed in MDA Cells
Our previous data showed that the Notch2 High MDA BrCa cell line, which behaved as dormant cells in in vitro and in vivo BrCa dormancy models, also expressed N-Cadherin [12]. These results suggested the presence of an MDA cell subpopulation co-expressing higher levels of both Notch2 and N-Cadherin that lodged in proximity of the endosteal area, prompting us to further investigate the role of N-Cadherin in the BrCa phenotype and dormancy.

Role of N-Cadherin in In Vivo Tumour Growth and Dormancy
To address the relevance of N-Cadherin in the in vivo tumour growth and dormancy, MACS-sorted N-Cadherin High and N-Cadherin Low MDA cells were injected into the tibia medullary cavity of female Balb-c nu/nu mice ( Figure 2a). After 4 weeks, X-rays and µCT analyses showed a trend of reduction in osteolytic lesion incidence (chi-square test; p = 0.06) ( Cadherin High cell-injected mice (Figure 2e,f). Interestingly, histopathological analyses revealed that the liver metastases were smaller in N-Cadherin High compared to N-Cadherin Low cell-injected mice (Figure 2g,h), with a trend of reduction in their number and incidence (Figure 2i and Supplementary Figure S2b). Lung metastases were instead undetectable.  Next, we analysed the endosteal niche colonization capability of Cadherin High and N-Cadherin Low MDA cells at 3-and 7-days post-injection, evaluating the tumour cell distribution in relation to their vicinity to the endosteal surface. Histomorphometric analysis demonstrated that the distance of cytokeratin + MDA cells from the endosteal surface was lower in the N-Cadherin High compared to the N-Cadherin Low cell-injected mice at all the time points analysed (Figure 2j,k). In line with this observation, the number of MDA cells nearby the endosteal surface was higher in the N-Cadherin High cell-injected mice after 7 days (Figure 2l,m).
Finally, we investigated if these features were shared by the mouse breast cancer cell line, 4T1. Of note, we did not observe significant differences in the in-bone tumour growth and endosteal niche colonization capability when we injected MACS-sorted N-Cadherin High or N-Cadherin Low 4T1 cells into the tibia medullary cavity of female Balb/C mice after 3 and 7 days (Supplementary Figure S2c-e). Longer term experiments were not possible due to the high aggressiveness of the 4T1 that grew fast in the tibia and quickly colonised the liver (Supplementary Figure S2f).
Altogether, these results demonstrated that the expression of N-Cadherin is associated with reduced MDA cell aggressiveness in the bone microenvironment and increased ability to home to the endosteal niche. Furthermore, these features were unique to the human MDA cells, not shared, at least in our experimental conditions, by the mouse 4T1 cells.

N-Cadherin Mediates BrCa Cell Adhesion onto SNOs In Vitro
To establish the role of N-Cadherin in the SNO-mediated tumour cell dormancy, MACS-sorted N-Cadherin High and N-Cadherin Low MDA GFP cells were seeded onto sorted SNOs or NON-SNOs as previously described [12]. N-Cadherin High MDA GFP cell adhesion, measured after 1 h from plating, was significantly higher when they were plated on SNOs compared to all other conditions tested (Figure 3a). In a time-course of 24-72 h, the number of MDA GFP cells was lower in MDA-SNO compared to MDA-NON-SNO cocultures, regardless of the N-Cadherin status (Figure 3b), suggesting that the expression of N-Cadherin does not affect the inhibition of MDA cell proliferation induced by SNOs [12]. In line with this observation, the N-Cadherin knock-down by a specific siRNA reduced the N-Cadherin High MDA GFP cell adhesion to SNOs (Figure 3c) without affecting the SNO-dependent inhibition of MDA proliferation (Supplementary Figure S3a).
Interestingly, N-Cadherin High MDA cells showed higher Notch2 transcriptional expression when compared with the N-Cadherin Low counterpart (Supplementary Figure S3b), and FACS analysis revealed that 58.4% of the N-Cadherin High MDA cells were also Notch2 High (Figure 1e). We also co-cultured MACS-sorted Notch2 High and Notch2 Low MDA GFP cells with SNOs or NON-SNOs. Although cell adhesion measured after 1 h from plating was not statistically different in Notch2 High and Notch2 Low MDA GFP co-cultured with SNOs and NON-SNOs (Figure 3d), during the subsequent time-course the number of Notch2 High MDA GFP cells increased less when plated on SNOs compared to the co-plating with NON-SNOs (Figure 3e). Interestingly, co-cultures between Notch2 High MDA GFP cells and NON-SNOs or Notch2 Low MDA GFP cells and SNOs impaired tumour cell proliferation only partially, suggesting that both players should be present to induce the maximum inhibitory effect (Figure 3e).
Next, we tested the role of N-Cadherin in SNO-induced BrCa dormancy using the mouse 4T1 cells. To this purpose, we generated genetically modified 4T1 cells using the CRISPR/Cas9 technology. Briefly, N-Cadherin expression was firstly abolished (4T1 Cdh2-KO ) and subsequentially rescued (4T1 Cdh2-Res ) (Supplementary Figure S1).
In our previous work, we demonstrated that SNOs induced cellular dormancy in parental unsorted 4T1 cells [12]. Here, we showed no difference in adhesion and number of 4T1 Cdh2-KO-GFP cells co-cultured with SNOs or NON-SNOs for 48 h (Figure 4a,b). Interestingly, the number of 4T1 Cdh2-Res-GFP cells plated onto SNOs was higher after 1h of co-plating (Figure 4c), but their number increased less in the subsequent 48 h compared to 4T1 Cdh2-KO-GFP (Figure 4d). Next, we tested the role of N-Cadherin in SNO-induced BrCa dormancy using the mouse 4T1 cells. To this purpose, we generated genetically modified 4T1 cells using the CRISPR/Cas9 technology. Briefly, N-Cadherin expression was firstly abolished (4T1 Cdh2-KO ) and subsequentially rescued (4T1 Cdh2-Res ) (Supplementary Figure S1).
In our previous work, we demonstrated that SNOs induced cellular dormancy in parental unsorted 4T1 cells [12]. Here, we showed no difference in adhesion and number of 4T1 Cdh2-KO-GFP cells co-cultured with SNOs or NON-SNOs for 48 h (Figure 4a,b). Interestingly, the number of 4T1 Cdh2-Res-GFP cells plated onto SNOs was higher after 1h of co-plating (Figure 4c), but their number increased less in the subsequent 48 h compared to 4T1 Cdh2-KO-GFP (Figure 4d Of note, unlike MDA cells, in the 4T1 cells we observed that the level of Notch2 expression was independent of N-Cadherin expression. In fact, no significant differences were found in Notch2 expression between 4T1 Cdh2-KO-GFP and 4T1 Cdh2-Res-GFP cells and between N-Cadherin High and N-Cadherin Low MACS-sorted 4T1 cells (Supplementary Figure S4a,b). In agreement with this observation, when 4T1 Cdh2-KO-GFP and 4T1 Cdh2-Res-GFP cells MACS-sorted into Notch2 High and Notch2 Low were co-plated with SNOs or NON-SNOs, we confirmed that the slower growth rate in the 4T1 cells was related to the presence of N-Cadherin and not to the Notch2 status. In fact, in all conditions tested, 4T1 Cdh2Res-GFP -Notch2 High and 4T1 Cdh2Res-GFP -Notch2 Low cultures showed the lowest number of cells per plate when compared to 4T1 Cdh2KO-GFP -Notch2 High and 4T1 Cdh2KO-GFP -Notch2 Low cells plated onto both SNOs and NON-SNOs for 48 h (Figure 4e,f). These results suggest that the axis Notch2/N-Cadherin is not involved in the SNO-mediated cell dormancy, while N-Cadherin strongly reduces the proliferation ability of 4T1 cells both on SNOs and NON-SNOs.  Taken together these data indicated that N-Cadherin plays a role in MDA and 4T1 cell adhesion especially onto SNOs, while Notch2 is pivotal in SNO-induced dormancy in human MDA cells, but not in mouse 4T1 cells.

The Role of N-Cadherin in HSC Mimicry and Cancer Stem Cell-like Phenotype
We previously demonstrated that the dormant Notch2 High MDA cells showed HSC mimicry, along with a cancer stem cell-like phenotype and reduced 2D cell proliferation, when compared with the Notch2 low counterpart [12]. To evaluate whether these features were shared by the N-Cadherin High BrCa cells, we first investigated whether MACS-sorted N-Cadherin High MDA cells showed an HSC-like gene signature. Our analysis demonstrated that they expressed higher mRNA levels of the HSC markers, CD34, TEK Receptor Tyrosine Kinase (TIE2), and C-X-C Motif Chemokine Receptor 4 (CXCR4) compared to N-Cadherin Low MDA cells (Figure 5a), along with a higher expression of the stem cell related genes SOX2 and NAGOG (Figure 5b). Moreover, we confirmed the presence of a canonical cancer stem cell signature independent from N-Cadherin expression. Indeed, both N-Cadherin High and N-Cadherin Low MDA cells expressed high CD44, medium ALDH and low CD24 mRNA (Figure 5c and Supplementary Figure S5a). On the contrary, a lower expression of the cell proliferation marker EdU was found in N-Cadherin High compared to N-Cadherin Low MDA cells (Figure 5d). Of note, N-Cadherin High cells were able to initiate more primary (Figure 5e,f) and secondary (Figure 5g,h) mammospheres. Moreover, primary mammospheres from the N-Cadherin High MDA cells were larger (Figure 5e) and more numerous (Figure 5f) than mammospheres generated by N-Cadherin Low MDA cells. Secondary mammospheres were larger in the N-Cadherin High compared to the N-Cadherin Low MDA cells (Figure 5g), while their number was very variable and showed only a trend to increase (Figure 5h). The cancer stem cell-like phenotype was further confirmed employing an in vivo Limiting Dilution Assays (LDA) that revealed a higher stem cell frequency in the N-Cadherin High compared to the N-Cadherin Low MDA cells subpopulation (Table 1).
Interestingly, the analysis of the epithelial-to-mesenchymal transition (EMT) markers revealed a higher expression of E-Cadherin in the N-Cadherin High vs. N-Cadherin Low MDA cells (Supplementary Figure S5b).
Next, the phenotypic comparison between MDA and 4T1 cells revealed a different impact of N-Cadherin on cellular stemness. In fact, unlike MDA cells, both 4T1 Cdh2-Res-GFP and 4T1 N-Cadherin High showed a lower ability to initiate both primary (Figure 5i    Taken together these results showed that, unlike 4T1, N-Cadherin High MDA cells display an HSC-like gene signature and a cancer stem cell-like phenotype similar to the Notch2 High MDA cells. Moreover, the role of N-Cadherin in mouse 4T1 cancer cell stemness differs from its role in human MDA cell stemness.

N-Cadherin Status in Human Primary BrCa and Correlation with Survival
The KMplot ® , containing protein expression data and survival information from four independent cohorts of 1193 BrCa patients, was used to test the correlation between N-Cadherin protein expression and survival. The analysis revealed the presence of a positive correlation between the N-Cadherin and the overall survival when we analysed the whole dataset (Figure 6a). Furthermore, when data were stratified for oestrogen (ER), progesteron (PR) receptor, Human Epidermal growth factor Receptor 2 (HER2) or triple negative subtypes, we confirmed a significant and positive correlation between high N-cadherin level and overall survival in patients with either ER-or PR-positive breast cancers (Figure 6b,d). On the contrary, no significant correlations were found in the survival of patients with either ER-or PR-negative, HER2 positive and negative, or triple negative breast cancers (Figure 6c,e,f-h).
Finally, we investigated the expression of N-Cadherin in a BrCa tissue array containing 64 different samples of primary tumours. We observed that N-Cadherin positive cells represented a small subpopulation (0.76 ± 0.3 cell/mm 2 ) (Figure 6i and Supplementary Table  S4). Moreover, histopathological analysis revealed that the number of N-Cadherin positive cells was higher in poorly differentiated primary BrCa (Figure 6j). In contrast, the number of N-cadherin positive cells was unchanged when we stratified the tumours according to the presence of distant metastasis (Figure 6k) and PR and HER2 status (Figure 6l,m), while, when we stratified our samples for the ER status, a trend of increase in the N-cadherinpositive cells (p = 0.08) was found in the ER-positive compared to the ER-negative primary breast cancers (Figure 6n). Moreover, no changes in the number of N-Cadherin positive cells was found when we compared the ER, PR and HER2 single negative with triple negative specimens (Figure 6o).
to the ER-negative primary breast cancers (Figure 6n). Moreover, no changes in the number of N-Cadherin positive cells was found when we compared the ER, PR and HER2 single negative with triple negative specimens (Figure 6o).

Discussion
Tumour cell dormancy is an intricate mechanism involving different molecular pathways and cell-cell interactions in accordance with the type of cancer and the microenvironmental signalling [32]. In our previous work we demonstrated that dormant BrCa cells interact with a specific osteoblast subpopulation, known as spindle-shaped N-Cadherin High osteoblasts (SNOs), remaining cell cycle arrested due to the Notch2 pathway. Moreover, we demonstrated that dormant BrCa cells compete with HSCs for bone marrow engraftment and endosteal niche colonization thanks to their HSC mimicry features [12]. In this work, we added a new piece in this complex puzzle showing a potential role of N-Cadherin in the SNO-mediated BrCa cell dormancy and cellular stemness.
The role of N-Cadherin in tumour biology is very complex, and varies according to the cellular context and the type of tumour [21]. N-cadherin has been reported to induce or suppress tumour development and spreading according to the type of cancer [22,[24][25][26].
In the BrCa context, the role of N-Cadherin is still poorly understood, probably because its function is tightly related to cellular and microenvironmental conditions. Data in the literature demonstrate that the expression of N-Cadherin increases BrCa cell adhesion to the stroma and stimulates motility, enhancing metastatic spread [15,25,26,33]. In contrast, a recent paper demonstrated that N-Cadherin drives human BrCa dormancy in the bone marrow in association with connexin 43 [26]. Moreover, in line with our findings, the author showed that N-Cadherin was expressed by the stem compartment of BrCa cells, contributing to the maintenance of the cellular dormancy [26].
We found the expression of N-Cadherin in the dormant Notch2 High MDA BrCa cells, confirming the possible association between N-Cadherin expression and BrCa cellular dormancy. In line with this, MDA cells expressing high level of N-Cadherin showed a lower aggressiveness in the bone microenvironment associated with a lower incidence of osteolytic lesions alongside an increased endosteal niche engraftment, indicating that the N-Cadherin signalling prompted the tumour cells to acquire a dormant-like phenotype. The ability to lodge in proximity of the endosteal niche led us to assume that they were able to interact with the SNOs. Our assumption was confirmed in vitro by experiments showing that MDA cells expressing N-Cadherin were able to interact with SNOs and that the presence of N-Cadherin increased the ability of tumour cells to adhere to SNOs. This observation was further confirmed by the fact that the knock-down of the N-Cadherin expression in MDA cells reduced their ability to adhere to SNOs. Of note, unlike Notch2 High cells, the proliferation of the N-Cadherin High MDA cells in the presence of SNOs was unremarkable, suggesting that N-Cadherin mediates the adhesion of BrCa cells to SNOs, while Notch2 mediates the inhibition of SNO-induced cell proliferation. Accordingly, further analyses confirmed that only about 58% of N-Cadherin High MDA expressed high levels of Notch2. Altogether these data prompted us to hypothesize that the homophilic N-Cadherin interaction between MDA cells and SNOs is used by tumour cells for engraftment to the endosteal niche, while only a small subpopulation co-expressing Notch2 also acquires the dormant phenotype.
As mentioned, the ability to colonize the endosteal niche and interact with SNOs is a typical feature of HSCs [10,19] shared also by the dormant Notch2 High MDA cells [12]. Intriguingly, although conflicting data about the role of N-cadherin in HSCs are described in literature [19,[27][28][29], Arai et al. demonstrated that N-Cadherin is expressed by HSCs and that its overexpression promotes HSC quiescence [29]. In agreement with this observation, our data showed that the N-Cadherin expression was associated with the acquisition of a bone marrow-specific cancer stem cell phenotype and HSC like-signature in MDA cells. In fact, we found a higher expression of the canonical HSC markers CXCR4, TIE-2 and CD34 and of the stem cell related markers SOX2 and NANOG in the N-Cadherin High MDA cells along with a lower cell proliferation rate and a higher ability to form primary and secondary mammospheres. Moreover, in vivo LDA showed a higher stem cell frequency in the N-Cadherin High MDA cells, further confirming their ability to initiate a new cancer. Interestingly, the analysis of canonical cancer stem cell markers revealed that both N-Cadherin High and N-Cadherin Low MDA cells display the typical cancer stem cell phenotype CD44 High /CD24 Low /ALDH + [34]. These findings suggest that N-Cadherin expression identifies a specific subpopulation, expressing HSC-and stem cell-related markers, within the canonical CD44 High /CD24 Low /ALDH + cancer stem cell population. These data were in line with the ability of the N-Cadherin High to colonize the endosteal niche resulting in lower aggressiveness in the bone/bone marrow microenvironment and in the concept that dormant cells should have stem features to initiate a new tumour after their reactivation [6,7].
We also performed our experiments using the mouse BrCa cell line 4T1, and were able to establish syngeneic interactions with mouse bone cells in vitro and in vivo. We demonstrated that N-Cadherin is required in 4T1 cells for the interaction with SNOs in vitro. Indeed, the lack of N-Cadherin reverted the impairment of the 4T1 cell proliferation induced by SNOs. On the other hand, the re-expression of the N-Cadherin enhanced the adhesion of 4T1 cells onto SNOs, slowing down their proliferation. On the contrary, in vivo analysis of 4T1 revealed no differences associated with N-Cadherin expression. This was probably because 4T1 are characterised by a strong aggressiveness and extremely high proliferation rate in the bone microenvironment. Interestingly, in 4T1 cells there was no co-expression of N-Cadherin and Notch2, indicating that different signalling pathways could be involved in mouse BrCa cellular dormancy. This was also confirmed by the observation that, unlike MDA cells, 4T1 cells expressing a high level of Notch2 showed a higher proliferation rate when co-cultured with SNOs.
Surprisingly, when we analysed the stem features of 4T1 cells expressing N-Cadherin we found an opposite effect compared to the MDA cells. Indeed, we observed that high N-Cadherin expression was associated with a lower expression of the HSC marker Cxcr4 and the mouse stem related genes Bxdc2, Dnmt3b, Gata4, Grb7, Il6st, Lamb1-1, Lamc1, Nog and Tert, along with a lower ability to form primary and secondary mammospheres. Other HSC markers including Tie-2 and Sca-1 were not expressed by 4T1 cells in our experimental conditions. These results were confirmed using both genetic manipulation of the N-Cadherin and cell sorting for N-Cadherin status.
Finally, according to the fact that there are no reliable markers to predict tumour dormancy in clinical practice [35], we investigated whether N-Cadherin could be useful to this scope. The analyses carried out using public datasets demonstrated that N-Cadherin protein expression in the primary tumour was correlated with a better prognosis in an unselected BrCa patients' cohort. In line with this, the analyses of a primary BrCa tissues array demonstrated a trend of increase in the number of N-Cadherin positive cells in ER positive tumours, known to be less malignant, suggesting that N-Cadherin expression could be associated with less aggressive human primary tumours. However, we also found a higher number of N-Cadherin-positive cells in primary tumours classified as less differentiated, which are known to be more aggressive [36]. This is in line with a recent publication showing that N-Cadherin expression in patients with ductal carcinoma in situ is predictive of synchronous invasion [37]. Another explanation of this conflicting result could be represented by the fact that less differentiated tumour cells usually show a stem-like phenotype. In confirmation, according to our and other's [12,26,38] results, N-Cadherin is likely to be expressed mainly in the stem cellular compartment. A limitation of this study is that most of the tumours analysed were ductal carcinomas, and that the absence of more differentiated tumours in our BrCa tissue array did not allow a complete correlation analysis between type and differentiation grade of primary tumours and N-Cadherin expression. Moreover, even if our data were generated using triple negative BrCa cell lines, we did not find association between N-Cadherin expression and the overall survival in patients harboring triple negative breast cancers. Similar results were found when we analysed the number of N-Cadherin positive cells in triple negative specimens present in our BrCa tissues array. This could be partially explained by the fact that the number of patients derived from the public dataset carrying a triple negative tumor was relatively low (68 N-Cadherin Low and 27 N-Cadherin High samples) and the absence of triple positive specimens in our BrCa tissue array forced us to compare the triple negative samples only with the ER, PR and HER2 single negative tumors.
For this reason, further larger studies are needed to clarify the possible role of N-Cadherin as an early dormancy marker in BrCa patients.

Conclusions
In conclusion, in this work we defined the role of N-Cadherin in BrCa dormancy and stemness, highlighting the differences between human and mouse cell lines. Overall, we can conclude that N-Cadherin could play a role in the induction and maintenance of tumour cell dormancy, in cooperation with Notch2. Moreover, we demonstrated that N-Cadherin is mainly involved in the anchorage of tumour cells to the endosteal niche, rather than in the inhibition of cell proliferation. Therefore, we can speculate that targeting N-Cadherin could be a potential co-adjuvant therapy to be administered in combination with anti-tumoral drugs to prevent the engraftment and the dormancy of BrCa cells lodged in the endosteal niche.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/cancers14051317/s1, Figure S1: Generation of genetically modified 4T1 cells, Figure S2: In vivo analysis of MDA-and 4T1-N-Cadherin High BrCa cells, Figure S3: Effect of N-Cadherin down regulation on proliferation to SNOs and NON-SNOs and Notch2/N-Cadherin co-expression analysis in MDA cells, Figure S4: Notch2 and N-Cadherin co-expression analysis in 4T1 cells, Figure S5: Effect of N-Cadherin expression on HSC mimicry, stemness and EMT of MDA and 4T1 BrCa cells, Figure S6: Full agarose gel images for Figures S1 and S3, Figure S7: Full agarose gel images for Figures S4 and S5, Figure S8: Full agarose gel images for Figure S5, Table S1: Primer sequences, Table S2: Antibody information, Table S3: TaqMan™ Array Mouse Stem Cell Pluripotency gene expression, Table S4: Breast Cancer tissue array donor information.