Different Expression Levels of DLK2 Inhibit NOTCH Signaling and Inversely Modulate MDA-MB-231 Breast Cancer Tumor Growth In Vivo

NOTCH signaling is implicated in the development of breast cancer tumors. DLK2, a non-canonical inhibitor of NOTCH signaling, was previously shown to be involved in skin and breast cancer. In this work, we studied whether different levels of DLK2 expression influenced the breast cancer characteristics of MDA-MB-231 cells. We found that DLK2 overexpression inhibited NOTCH activation in a dose-dependent manner. Moreover, depending on the level of inhibition of NOTCH1 activation generated by different levels of DLK2 expression, cell proliferation, cell cycle dynamics, cell apoptosis, cell migration, and tumor growth in vivo were affected in opposite directions. Low levels of DLK2 expression produced a slight inhibition of NOTCH1 activation, and enhanced MDA-MB-231 cell invasion in vitro and cell proliferation both in vitro and in vivo. In contrast, MDA-MB-231 cells expressing elevated levels of DLK2 showed a strong inhibition of NOTCH1 activation, decreased cell proliferation, increased cell apoptosis, and were unable to generate tumors in vivo. In addition, DLK2 expression levels also affected some members of other cell signaling pathways implicated in cancer, such as ERK1/2 MAPK, AKT, and rpS6 kinases. Our data support an important role of DLK2 as a protein that can finely regulate NOTCH signaling and affect the tumor properties and growth dynamics of MDA-MB-231 breast cancer cells.


Introduction
In addition to its pivotal role during embryonic development and tissue homeostasis in adult life, NOTCH signaling is implicated in the development of numerous types of tumors. It is now accepted that NOTCH receptors may act both as oncogenes and as tumor suppressor proteins, depending on the cellular context [1][2][3]. The important role of NOTCH signaling in breast cancer development was established more than two decades ago with the discovery that the mouse Int3 oncogene, a truncated NOTCH4 protein corresponding to its active intracellular region (NICD4), was implicated in mammary tumors [4]. It is known that NOTCH signaling is aberrantly activated in breast cancer cell lines and primary cell samples. The overexpression of NOTCH receptors and ligands was reported in breast tumors, correlating with poorer patient prognosis. The inhibition of NOTCH signaling was shown to consistently suppress cell proliferation, invasiveness and migration, as well as the development and progression of breast tumors, which is of great importance for therapeutic targeting considerations [5].
Numerous studies established DLK1 (Delta-like homolog 1), a non-canonical ligand of NOTCH receptors, as an inhibitor of NOTCH signaling, including various works published by our group [6][7][8][9][10][11][12][13]. In a previous work, we studied the role of DLK1 in the tumor properties of MDA-MB-231 breast cancer cells [10]. We confirmed that human DLK1 inhibited NOTCH1 activation in these cells, and we revealed that their proliferation rate and invasion capabilities depended upon the level of NOTCH1 activation as negatively regulated by the level of DLK1 expression. A significant decrease in MDA-MB-231 cell proliferation and invasion was associated with high levels of DLK1 expression, which produced a significant inhibition of NOTCH1 activation. However, low levels of DLK1 expression produced low levels of inhibition of NOTCH1 activation, which unexpectedly caused an enhanced in vitro MDA-MB-231 cell invasion and increased cell proliferation both in vitro and in vivo. Therefore, NOTCH1 activation levels, as negatively regulated by DLK1, appeared critical for the function of this receptor as an oncogenic or as a tumor suppressor protein in MDA-MB-231 cells.
DLK2 (Delta-like homolog 2) is a protein highly homologous to DLK1 [14]. As with DLK1, DLK2 was shown to interact with NOTCH receptors and function as another non-canonical inhibitor of NOTCH signaling [8,11,13]. It was reported that, similarly to DLK1, DLK2 affects the growth of SK-MEL-2 melanoma cells, and the nature of its effects depends upon DLK2 expression levels and the degree of inhibition of NOTCH1 activation [9]. Recently, DLK2 was described as overexpressed in lethal prostate cancers, uveal melanoma, breast cancer cell lines and in patient tissues spanning three breast cancer subtypes (Luminal A, Luminal B, and Triple Negative) [15][16][17][18].
In this work, we explored the effects of different DLK2 expression levels on the characteristics of MDA-MB-231 breast cancer cells. We found that increased DLK2 expression levels led to the increased inhibition of the NOTCH1 receptor. However, only elevated levels of the inhibition of NOTCH1 activation, achieved by elevated levels of DLK2 expression, led to decreased cell growth. Cell cycle dynamics and apoptosis were also affected by DLK2 expression levels in opposite directions. The invasive properties of these cells were also affected. Thus, low levels of DLK2 expression led to the increased invasive properties of MDA-MB-231 cells, whereas high levels of DLK2 expression did not significantly modify them. These invasive alterations were associated with changes in the expression levels of cell adhesion proteins, such as E-cadherin and N-cadherin, involved in epidermal-tomesenchymal transition (EMT) and carcinogenesis [19]. In addition, DLK2 expression levels also affected members of other cell signaling pathways, including ERK1/2 MAPK (mitogen-activated protein kinase/extracellular signal-regulated kinase), AKT (protein kinase B), and rpS6 (ribosomal protein s6) kinases. Finally, MDA-MB-231 cells expressing high levels of DLK2 were unable to generate tumors in vivo in a nude mouse model, an effect that was not observed in control cells or cells expressing low levels of DLK2. Overall, these data support an important and complex role of DLK2 in the control of NOTCH signaling and in the tumorigenic properties and growth dynamics of MDA-MB-231 cells.

Generation of Transfected MDA-MB-231 Cells Stably Expressing Different Levels of Human DLK2 Protein
MDA-MB-231 cells, a triple-negative breast cancer cell line, display representative epithelial-to-mesenchymal transition (EMT) associated with breast cancer metastasis [20]. We selected this cell line for our studies because it represents an aggressive and fast-growing type of breast cancer, capable of generating metastasis in vivo [21]. In addition, our previous analyses confirmed that MDA-MB-231 cells do not express DLK1 [10], although they do express DLK2. We estimate that this property is suitable for studying the potential effects of the forced overexpression of DLK2 on triple-negative breast cancer cell growth without interfering with the expression of its homolog, DLK1.
MDA-MB-231 cells with different levels of DLK2 expression were obtained as described in Materials and Methods by transfecting several plates of cells with the HDLK2S plasmid or the empty vector. After each transfection, cells able to grow in a selective medium were pooled, and their DLK2 mRNA and DLK2 protein expression levels were analyzed. By these procedures, we ended up selecting two different pools of MDA-MB-231 cells that expressed low or high levels of DLK2 mRNA, compared with empty-vector cells, used as a control, as determined by RT-qPCR. These pools of cells were named HDLK2SL and HDLK2SH, respectively ( Figure 1A). they do express DLK2. We estimate that this property is suitable for studying the potential effects of the forced overexpression of DLK2 on triple-negative breast cancer cell growth without interfering with the expression of its homolog, DLK1.
MDA-MB-231 cells with different levels of DLK2 expression were obtained as described in Materials and Methods by transfecting several plates of cells with the HDLK2S plasmid or the empty vector. After each transfection, cells able to grow in a selective medium were pooled, and their DLK2 mRNA and DLK2 protein expression levels were analyzed. By these procedures, we ended up selecting two different pools of MDA-MB-231 cells that expressed low or high levels of DLK2 mRNA, compared with empty-vector cells, used as a control, as determined by RT-qPCR. These pools of cells were named HDLK2SL and HDLK2SH, respectively ( Figure 1A).
We also performed Western blot assays to determine DLK2 protein expression levels in the HDLK2SL and HDLK2SH cell pools compared with empty-vector cells. Images were quantified by densitometry. As expected, DLK2 levels were lower in HDLK2SL cells than in HDLK2SH cells. The differences between both pools, as well as between each pool and the pool of cells transfected with the empty vector, were statistically significant (Figure 1B). significantly different DLK2 mRNA and DLK2 protein expression levels, termed HDLK2SL (low expression level) and HDLK2SH (high expression level), were selected and used in subsequent assays. * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001 (Student's ttest results relative to empty vector control cells).

Different Levels of DLK2 Expression Inversely Modulate the Growth of MDA-MB-231 Cells
To analyze how different DLK2 expression levels affected the growth kinetics of MDA-MB-231 cells in vitro, the two pools of DLK2-expressing cells were grown in a culture medium for several days and their cell numbers were estimated at various time points by an MTT assay (see Materials and Methods). The results showed that cell growth was strongly dependent upon DLK2 expression levels. HDLK2SL cells showed a higher proliferation rate compared with that of cells transfected with an empty vector. However, HDLK2SH cells grew at a lower rate compared to cells transfected with an empty vector ( Figure 2A). Overall, after eight days in culture, HDLK2SL and HDLK2H cells had grown at a very significantly different rate, the cells with higher levels of DLK2 expression showing a slower growth rate.
To try to uncover the reasons of these remarkable differences between the two cell pools, we first studied the cell distribution in the different phases of the cell cycle by flow cytometry. We observed that a higher percentage of HDLK2SH cells were in the G0 phase We also performed Western blot assays to determine DLK2 protein expression levels in the HDLK2SL and HDLK2SH cell pools compared with empty-vector cells. Images were quantified by densitometry. As expected, DLK2 levels were lower in HDLK2SL cells than in HDLK2SH cells. The differences between both pools, as well as between each pool and the pool of cells transfected with the empty vector, were statistically significant ( Figure 1B).

Different Levels of DLK2 Expression Inversely Modulate the Growth of MDA-MB-231 Cells
To analyze how different DLK2 expression levels affected the growth kinetics of MDA-MB-231 cells in vitro, the two pools of DLK2-expressing cells were grown in a culture medium for several days and their cell numbers were estimated at various time points by an MTT assay (see Materials and Methods). The results showed that cell growth was strongly dependent upon DLK2 expression levels. HDLK2SL cells showed a higher proliferation rate compared with that of cells transfected with an empty vector. However, HDLK2SH cells grew at a lower rate compared to cells transfected with an empty vector ( Figure 2A). Overall, after eight days in culture, HDLK2SL and HDLK2H cells had grown at a very significantly different rate, the cells with higher levels of DLK2 expression showing a slower growth rate.
Increased apoptosis could also be a factor affecting the different overall cell growth rates observed in cells with high and low DLK2 protein levels. To study this aspect, we determined annexin V/ propidium iodide (PI) double staining by flow cytometry. Annexin V and propidium iodide staining was more intense in HDLK2SH cells and less intense in HDLK2SL cells compared to empty-vector cells ( Figure 2D). Thus, apoptosis also appeared to be affected by DLK2 expression levels consistent with the observed effects on cell growth.

Invasion Features of DLK2-Overexpressing MDA-MB-231 Cells
The increased malignancy of tumor cells is related to their ability to establish metastasis to distant organs or tissues. To achieve this, the cells must leave the original tumor and migrate. To determine whether DLK2 overexpression could affect the migration ability of MDA-MB-231 cells, in vitro invasion assays were performed as described in Materials and Methods section. Figure 3A shows that DLK2 overexpression levels altered the To try to uncover the reasons of these remarkable differences between the two cell pools, we first studied the cell distribution in the different phases of the cell cycle by flow cytometry. We observed that a higher percentage of HDLK2SH cells were in the G0 phase of the cell cycle, compared to HDLK2SL or empty-vector cells ( Figure 2B), suggesting that a decreased cell cycle rate was associated with higher levels of DLK2 expression. Consistent with this, a lower percentage of HDLK2SH cells was detected in the S phase, with a concomitant increase in the percentage of cells in the G2/M phase [22].
Using RT-qPCR, we also determined the relative mRNA levels of CCND1 (cyclinD1) and CDKN1A (p21, Cip1), two proteins involved in cell cycle progression [23]. No significant differences were observed between HDLK2SL cells and empty-vector cells in the relative expression levels of CCND1 or CDKN1A mRNAs. However, HDLK2SH cells showed significantly higher levels of CDKN1A and lower levels of CCND1 ( Figure 2C), which may help to explain the slow cell growth rate of these cells compared to that of HDLK2SL or empty-vector cells.
Increased apoptosis could also be a factor affecting the different overall cell growth rates observed in cells with high and low DLK2 protein levels. To study this aspect, we determined annexin V/ propidium iodide (PI) double staining by flow cytometry. Annexin V and propidium iodide staining was more intense in HDLK2SH cells and less intense in HDLK2SL cells compared to empty-vector cells ( Figure 2D). Thus, apoptosis also appeared to be affected by DLK2 expression levels consistent with the observed effects on cell growth.

Invasion Features of DLK2-Overexpressing MDA-MB-231 Cells
The increased malignancy of tumor cells is related to their ability to establish metastasis to distant organs or tissues. To achieve this, the cells must leave the original tumor and migrate. To determine whether DLK2 overexpression could affect the migration ability of MDA-MB-231 cells, in vitro invasion assays were performed as described in Materials and Methods section. Figure 3A shows that DLK2 overexpression levels altered the invasion potential of these cells. HDLK2SL cells showed a significant increase in invasiveness, whereas HDLK2SH cells showed no significant changes in this property compared to empty-vector cells. chymal transition. The zinc finger protein Slug, encoded by the SNAI2 gene [24,25], is one of these transcription factors. This gene is upregulated by NOTCH activation in colorectal cancer [26] and in MDA-MB-231 cells [27]. Our transfected cells showed that SNAI2 expression levels were affected by DLK2 expression levels. HDLK2SL cells showed a significant increase, whereas HDLK2SH cells showed a significant decrease, in SNAI2 expression levels ( Figure 3B).
As expected, the expression levels of CDH1 and CDH2 were affected by the changes in SNAI2 expression levels, which are caused by the different levels of DLK2 expression. Both HDLK2SL and HDLK2SH cells showed an increase in CDH1 and CDH2 mRNA expression levels. However, CDH1 expression levels were higher for HDLK2SH cells ( Figure  3C).

DLK2 Inhibited NOTCH Activation in MDA-MB-231 Cells in a Dose-Dependent Manner
In previous works, we showed that DLK2 functions as an inhibitor of NOTCH signaling in mouse preadipose cells [8,11,13] and in human melanoma cells [9]. The level of The invasiveness of breast cancer cells depends on the activity of transcription factors that modulate the expression of cell adhesion proteins, such as E-cadherin, encoded by the CDH1 gene and N-cadherin, encoded by the CDH2 gene, and the epithelial-to-mesenchymal transition. The zinc finger protein Slug, encoded by the SNAI2 gene [24,25], is one of these transcription factors. This gene is upregulated by NOTCH activation in colorectal cancer [26] and in MDA-MB-231 cells [27]. Our transfected cells showed that SNAI2 expression levels were affected by DLK2 expression levels. HDLK2SL cells showed a significant increase, whereas HDLK2SH cells showed a significant decrease, in SNAI2 expression levels ( Figure 3B).
As expected, the expression levels of CDH1 and CDH2 were affected by the changes in SNAI2 expression levels, which are caused by the different levels of DLK2 expression. Both HDLK2SL and HDLK2SH cells showed an increase in CDH1 and CDH2 mRNA expression levels. However, CDH1 expression levels were higher for HDLK2SH cells ( Figure 3C).

DLK2 Inhibited NOTCH Activation in MDA-MB-231 Cells in a Dose-Dependent Manner
In previous works, we showed that DLK2 functions as an inhibitor of NOTCH signaling in mouse preadipose cells [8,11,13] and in human melanoma cells [9]. The level of NOTCH1 signaling greatly affected the proliferation rate of melanoma cells in a non-linear fashion. For these reasons, we hypothesized that DLK2 might be acting by a similar mechanism on MDA-MB-231 cells. The overall state of NOTCH signaling depends not only on NOTCH1 but also on the activation of the other three NOTCH receptors (NOTCH 2-4) expressed by MDA-MB-231 cells [10]. To assess this, we first estimated the global levels of NOTCH signaling in MDA-MB-231 cells transiently transfected with the HDLK2S plasmid by determining the activity of a luciferase reporter gene under the control of a NOTCHdependent promoter that possesses binding sites for the CSL/RBP-Jk factor. We observed a significant decrease in luciferase reporter gene activity in MDA-MB-231 cells transiently transfected with the HDLK2S plasmid compared with cells transiently transfected with the empty vector ( Figure 4A). The degree of inhibition was comparable to that obtained when cells transiently transfected with an empty vector were treated with the NOTCH inhibitor DAPT at a concentration of 10 µM for 24 h. Thus, this result confirmed a global NOTCH inhibitory role of DLK2 on NOTCH activation and signaling.

DLK2 Modulates Phosphorylation of ERK1/2 MAPK, AKT and rpS6 Kinases in MDA-MB-231 Cells
The ERK/MAPK pathway appears to be constitutively activated in MDA-MB-231 cells, which likely is important in preventing anoikis and maintaining the anchorage-independent growth of these cells [29]. In a previous work, and in agreement with other reports [30][31][32], we showed that, in addition to NOTCH signaling, DLK1 inhibited the ERK1/2 MAPK signaling pathway in MDA-MB-231 cells [10]. Therefore, it was important Next, we analyzed NOTCH activation levels in our two pools of DLK2-transfected cells. First, NOTCH1 activation levels were determined by Western blot by using an antibody that specifically detects the active intracellular domain of NOTCH1 (NICD1). Both HDLK2SL and HDLK2SH cell pools showed a significant decrease in NICD1 levels, indicating lower levels of NOTCH1 activation. The degree of inhibition was proportional to the level of DLK2 expression, and HDLK2SH cells showed a significant and stronger decrease in NOTCH1 activation levels than HDLK2SL cells, which also showed a significant inhibition of NOTCH1 activation compared to empty-vector cells ( Figure 4B). As a control, we treated empty-vector cells with the gamma-secretase inhibitor DAPT (10 µM) for 24 h, a treatment known to suppress NOTCH receptors activation ( Figure 4B).
We also studied the expression levels of HEY1, HEY2 and HES1, three NOTCH receptors target genes ( Figure 4C) [28]. As expected, the expression levels of HEY1 and HEY2 correlated with the levels of NOTCH1 activation shown by cells with different levels of DLK2 expression. Thus, both HDLK2SL and HDLK2SH cells showed lower HEY1 and HEY2 expression levels than empty-vector cells, although this decrease was not significant for HEY2 in the case of HDLK2SL cells. However, HES1 expression levels showed an increase in both pools of transfected cells, rather than the expected decrease, and the increase was greater in HDLK2SH cells. We do not have an explanation for these data; perhaps it is possible that changes in the stoichiometry of NOTCH 1-4 receptor activation and canonical ligand expression caused by DLK2 expression levels may lead to increased expression of some NOTCH target genes but to the inhibition of others.
To further study this possibility, we decided to analyze the expression levels of the JAG1, which encodes for Jagged1, one of the most important canonical NOTCH receptor ligands in breast cancer [27], as well as the expression levels of the four NOTCH receptor genes in our DLK2-overexpressing cells. We also analyzed the expression of the JAG2 gene, but it was not expressed by these cells. We hypothesized that if DLK2 preferentially inhibited one or a few, but not all, NOTCH receptors, this inhibition might somehow be compensated by the increased expression levels of the Jagged1 ligand, or by the increased expression levels of at least one of the NOTCH receptors, which could lead to the observed increase in HES1 expression, despite lower overall levels of NOTCH activation. However, this was not what we observed. The expression level of JAG1 decreased as DLK2 expression levels increased ( Figure 4D), indicating that the direct inhibition of NOTCH activation caused by DLK2 could also be associated with an indirect inhibition caused by decreased expression levels of JAG1 ligand. Moreover, all NOTCH receptor genes, except NOTCH2, showed a significant decrease in their expression levels in both HDLK2SL and HDLK2SH cells ( Figure 4E). The most significant reduction occurred in the case of NOTCH3.

DLK2 Modulates Phosphorylation of ERK1/2 MAPK, AKT and rpS6 Kinases in MDA-MB-231 Cells
The ERK/MAPK pathway appears to be constitutively activated in MDA-MB-231 cells, which likely is important in preventing anoikis and maintaining the anchorage-independent growth of these cells [29]. In a previous work, and in agreement with other reports [30][31][32], we showed that, in addition to NOTCH signaling, DLK1 inhibited the ERK1/2 MAPK signaling pathway in MDA-MB-231 cells [10]. Therefore, it was important to study whether DLK2 overexpression would also modify the steady state of ERK1/2 MAPK activation in these cells. To this end, we analyzed the phosphorylation status of ERK1/2 MAPK in our pools of DLK2-transfected cells by Western blot analysis. For HDLK2SL cells, no changes were observed relative to empty-vector cells, whereas HDLK2SH cells showed a significant decrease in ERK1/2 MAPK phosphorylation. As a control, we treated cells transfected with an empty vector with 10 µM of the MEK inhibitor U0126, or with 30 µM of the MAPKK inhibitor PD098059 for 24 h to inhibit ERK1/2 MAPK phosphorylation ( Figure 5A). LY294002, or with 1 μg/mL of the mTOR kinase inhibitor, Rapamycin.
Finally, the PI3K/AKT pathway, an important signaling pathway related to cell quiescence and proliferation, which is activated in 30-40% of breast cancer cases [33], was not modified in MDA-MB-231 cells transfected with DLK1 [10]. However, HDLK2SL cells showed a significant increase in the activation of this pathway, whereas HDLK2SH cells showed a decrease ( Figure 5C). As a control, we treated empty vector-transfected cells with 10 μM of the PI-3 kinase inhibitor LY294002 to inhibit PI3K/AKT activation.  Consistent with the previous results, we demonstrated that the phosphorylation levels of rpS6, a substrate for ERK1/2 MAPK, were significantly decreased when DLK1 was overexpressed [10]. Accordingly, we studied the phosphorylation levels of rpS6 in our DLK2-transfected cells. In this case, however, we observed different results. HDLK2SL cells showed a significant increase in rps6 phosphorylation levels, whereas no changes were observed in HDLK2SH cells with respect to empty-vector cells ( Figure 5B). Here, as a control, we treated empty-vector-transfected cells with 10 µM of the PI-3 kinase inhibitor LY294002, or with 1 µg/mL of the mTOR kinase inhibitor, Rapamycin.
Finally, the PI3K/AKT pathway, an important signaling pathway related to cell quiescence and proliferation, which is activated in 30-40% of breast cancer cases [33], was not modified in MDA-MB-231 cells transfected with DLK1 [10]. However, HDLK2SL cells showed a significant increase in the activation of this pathway, whereas HDLK2SH cells showed a decrease ( Figure 5C). As a control, we treated empty vector-transfected cells with 10 µM of the PI-3 kinase inhibitor LY294002 to inhibit PI3K/AKT activation.

DLK2 Overexpression Affects MDA-MB-231 Tumor Development In Vivo
In our previous work [10], we observed that the significant effects exerted by DLK1 on the cellular and molecular events of MDA-MB-231 cells, which impacted their growth properties in vitro, also impacted their ability to form tumors in vivo. This ability was dependent upon DLK1 expression levels. Low levels of DLK1 expression accelerated tumor growth in vivo, whereas high levels of DLK1 expression caused tumors that grew more slowly than normal. To study the effects of different DLK2 expression levels on tumor growth in vivo, we subcutaneously injected MDA-MB-231 cells transfected with empty-vector cells, or HDLK2SL or HDLK2SH cell pools into nude mice and measured their growth rate for six weeks. The results were striking. While HDLK2SL cells showed a non-statistically significant increase in tumor growth compared to empty-vector cells, HDLK2SH cells completely failed to establish tumors in vivo ( Figure 6A,B). Furthermore, in mice injected with HDLK2SL cells, a second smaller tumor appeared near the injection site ( Figure 6C). These results indicate that different levels of DLK2 expression could have different consequences on the establishment and growth of MDA-MB-231 tumors.

DLK2 Overexpression Affects MDA-MB-231 Tumor Development In Vivo
In our previous work [10], we observed that the significant effects exerted by DLK1 on the cellular and molecular events of MDA-MB-231 cells, which impacted their growth properties in vitro, also impacted their ability to form tumors in vivo. This ability was dependent upon DLK1 expression levels. Low levels of DLK1 expression accelerated tumor growth in vivo, whereas high levels of DLK1 expression caused tumors that grew more slowly than normal. To study the effects of different DLK2 expression levels on tumor growth in vivo, we subcutaneously injected MDA-MB-231 cells transfected with empty-vector cells, or HDLK2SL or HDLK2SH cell pools into nude mice and measured their growth rate for six weeks. The results were striking. While HDLK2SL cells showed a non-statistically significant increase in tumor growth compared to empty-vector cells, HDLK2SH cells completely failed to establish tumors in vivo ( Figure 6A,B). Furthermore, in mice injected with HDLK2SL cells, a second smaller tumor appeared near the injection site ( Figure 6C). These results indicate that different levels of DLK2 expression could have different consequences on the establishment and growth of MDA-MB-231 tumors.

Discussion
In a previous work, we reported that different levels of DLK1 expression exerted a significant modulation on the growth and tumorigenic properties of MDA-MB-231 triplenegative breast cancer cell line [10]. The present work aimed to study the effects exerted by DLK2, a protein homologous to DLK1, in the tumorigenic properties of this cell line [14]. As with DLK1, DLK2 was shown to inhibit NOTCH receptor signaling [8,11,13] and

Discussion
In a previous work, we reported that different levels of DLK1 expression exerted a significant modulation on the growth and tumorigenic properties of MDA-MB-231 triplenegative breast cancer cell line [10]. The present work aimed to study the effects exerted by DLK2, a protein homologous to DLK1, in the tumorigenic properties of this cell line [14]. As with DLK1, DLK2 was shown to inhibit NOTCH receptor signaling [8,11,13] and affect the outcome of several differentiation processes, in particular adipogenesis [8,10,13,34] and chondrogenic differentiation [35]. Therefore, it was important to study the effects of different levels of DLK2 expression on the growth of MDA-MB-231 breast cancer cells and their tumorigenic properties, and then compare them with the effects exerted by DLK1 in these cells, especially since NOTCH signaling was found to be a hallmark of triple-negative breast cancer tumors [36].
The effects on the growth and invasion capabilities of MDA-MB-231 cells observed upon the reconstitution of DLK1 [10] or upon DLK2 overexpression, are consistent with the role of NOTCH signaling in the growth of triple-negative breast cancer cells. Our data indicate that, as was the case with DLK1, the in vitro and in vivo growth properties of MDA-MB-231 cells are negatively affected by elevated levels of DLK2 expression (Figures 2 and 6). High levels of DLK2 expression affected cell distribution throughout the cell cycle phases ( Figure 2B). These changes were related to a significant increase in the expression of CDKN1A (p21, Cip1), a cyclin-dependent kinase inhibitor capable of inhibiting all cyclin/CDK complexes, and a downregulation of CCND1 (cyclin D1) expression, which could result in the decreased cell growth that we observed in vitro and in vivo. The effects we detected on proliferation could also be related to the fact that high levels of DLK2 expression also increased apoptosis, as determined by annexin V/ propidium iodide assays. On the other hand, low levels of DLK2 overexpression led to increased cell proliferation, and changes in the cellular distribution throughout the cell cycle phases and the expression levels of p21, Cip1 and cyclin D1. In addition, cells with low levels of DLK2 expression showed a decreased apoptosis.
One possibility that may help to explain the similar effects caused by the overexpression of DLK1 and DLK2, despite the observed differences, is that both proteins were shown to act as inhibitors of NOTCH signaling in several cellular systems [6][7][8][9][10][11][12][13]. The data presented here confirm that DLK2 is also an inhibitor of NOTCH activation in MDA-MB-231 cells. The behavior of DLK2 as an inhibitor of NOTCH signaling in MDA-MB-231 cells closely mimics the behavior of DLK1 [10]. Both proteins inhibit NOTCH1 activation in a dose-dependent manner, as elevated levels of DLK1 or DLK2 expression correlate with the high levels of inhibition of NOTCH1 activation. In both cases, the rate of inhibition of NOTCH1 activation obtained by a high expression of DLK1 or DLK2 is similar to that obtained by treatment with 10 µM DAPT, an inhibitor of the γ−secretase complex that processes NOTCH receptors ( Figure 4B). The similar behavior of both DLK proteins as inhibitors of NOTCH activation is also revealed by analyzing their effects on the expression levels of known target genes used as markers of NOTCH signaling levels. These genes include HEY1, HEY2 and HES1. The overexpression of DLK2 generally led to decreased expression levels of HEY1 and HEY2, but significantly increased expression levels of HES1 ( Figure 4C) [10]. While the decrease in HEY1 and HEY2 expression levels was expected, the increase in HES1 expression level was contrary to expectations. We do not have an explanation for this unexpected result. Although it is known that HES1 expression levels can fluctuate widely [28], it is unlikely that the consistent effect of DLK2 overexpression on HES1 expression levels in MDA-MB-231 cells is due to these fluctuations. These and previous data reveal that, in some cells or some contexts, the inhibition of NOTCH signaling can lead to the increased expression of some target genes, suggesting the existence of feedback mechanisms that may be altered by abrupt changes in NOTCH activation levels [8,13].
Despite all of the similar effects of DLK1 and DLK2 overexpression on MDA-MB-231 cells, there are some important differences, especially in the effects caused by both proteins on NOTCH mRNAs expression levels. In this regard, low or high levels of DLK2 expression caused a decrease in the expression levels of all NOTCH genes, except NOTCH2 ( Figure 4E).
Despite being able to inhibit overall NOTCH signaling levels to similar degrees, it is likely that DLK1 and DLK2 do not inhibit this signaling pathway with the same specificity. Most likely one of the proteins inhibits one of the NOTCH receptors preferentially to the others. This different specificity is reflected in different NOTCH-dependent genes being transcribed or silenced or in the same genes being transcribed to different degrees. Thus, each DLK protein influences the differentiation outcome and growth properties of the cells in a particular way. In this regard, it is worth mentioning that NOTCH4 appears to be the most important NOTCH gene for inhibiting breast cancer growth, since the overexpression of NOTCH4 increases, whereas the inhibition of NOTCH4 reduces, the proliferation and invasiveness of triple-negative breast cancer cells [37]. Our data indicate that NOTCH4 is one of the genes whose expression is more strongly reduced by high levels of DLK2.
On the other hand, NOTCH3 increases the establishment of tumors and metastasis [38,39]. However, the opposite effects are observed in vivo, since HDLK2SH cells were totally unable to establish tumors in nude mice ( Figure 6).
DLK2 also affects the growth of MDA-MB-231 cells. The effects depend on DLK2 expression levels and the degree of inhibition of NOTCH signaling, as elevated DLK2 expression levels, which resulted in high levels of inhibition of NOTCH activation, produced a decreased cell growth and increased apoptosis. Meanwhile, low DLK2 expression levels, which resulted in low levels of the inhibition of NOTCH signaling, produced increased cell growth and decreased apoptosis. This dose-dependent inhibition of NOTCH activation by DLK proteins was also described in SK-MEL-2 melanoma cells, where it is also associated with an opposite effect on cell proliferation [9]. In this regard, previously published data showed that different levels of NOTCH signaling led to different phenotypic responses, including stimulatory or suppressive growth effects in mammary epithelial cells [40]. The effect we observed in MDA-MB-231 cells expressing high DLK2 levels is similar to that reported in other studies performed in the same cells with monoclonal antibodies that inhibit NOTCH1 activation. The treatment of cells with these antibodies resulted in the decreased cell proliferation and increased induction of apoptosis [41]. Nevertheless, it was recently reported that the overexpression of DLK2 (EGFL9) protein does not affect the proliferation of mammary epithelial cell, and that the downregulation of its expression does not affect the proliferation of mouse 4T1 and human SUM159 metastatic breast cancer cell lines [16].
We also observed that the overexpression of DLK2 in MDA-MB-231 cells resulted in gene expression changes related to invasiveness. Thus, the expression levels of the SNAI2 gene were affected by DLK2 overexpression. The Slug (SNAI2) factor enhances the epithelial-to-mesenchymal transition by modulating E-cadherin and N-cadherin expression and shows antiapoptotic activity [42]. However, our results ( Figure 3) show a somewhat contradictory situation. Low levels of DLK2 expression (HDLK2SL) produced increased levels of SNAI2 expression, which is associated with increased levels of E-and N-cadherin expression. However, high levels of DLK2 expression (HDLK2SH) resulted in decreased levels of SNAI2 expression, which is associated with even higher levels of E-cadherin expression and similar levels of N-cadherin expression than HDLK2SL cells ( Figure 3B,C). Therefore, regardless of the changes in SNAI2 expression caused by the different levels of DLK2 expression, E-cadherin and N-cadherin expression is increased. The increase in the level of E-cadherin expression is, in fact, correlated with the DLK2 expression levels of the transfected MDA-MB-231 cells. However, changes in the invasiveness properties of the cells are not consistent with this, since despite the increased levels of E-cadherin and N-cadherin expression in HDLK2SL cells relative to empty-vector cells, an increased invasiveness is observed. This effect is different from what occurs in HDLK2SH cells that do not show significant changes in invasiveness ( Figure 3A,C) compared to empty-vector cells. In this regard, previously published data show that the inhibition of NOTCH signaling led to an increase in E-cadherin expression and a decrease in the invasive capacity of MDA-MB-231 cells [43], which agrees with our findings showing that HDLK2SH cells exhibit a high rate of inhibition of NOTCH1 activation and prevent tumor development in nude mice.
Strikingly, despite the observed differences in the expression levels of slug and Ncadherin and E-cadherin caused by DLK2 overexpression in MDA-MB-231 cells, similar effects on invasiveness features were observed for MDA-MB-231 cells overexpressing different levels of DLK1 [10]. This suggests that, in addition to their effect on SNAI2 and CDH gene expression, both DLK1 and DLK2 may affect other shared mechanisms that modulate the invasive properties of MDA-MB-231 and other cancer cells. DLK1 is known to affect MMP9 expression levels through NOTCH signaling to promote lung cancer cell invasion [44]. Recently, it was also described that the ectopic expression of (DLK2) EGFL9 significantly promotes cell migration and invasion in mammary epithelial cell lines and cancer metastasis in vivo, while a knockdown of DLK2 in highly metastatic breast cancer cell lines decreased cell migration and invasion and inhibited distant tumor metastasis in vivo [16]. These effects appear to be the result of the physical interaction between DLK2 and cMET receptor. This interaction activates the cMET receptor and downstream signaling pathways that are involved in cell migration and invasion. It is possible that different levels of DLK2 expression may led to different effects on MDA-MB-231 cells depending on the degree of DLK2 interaction with cMET or NOTCH, as it was reported that NOTCH1 and cMET may crosstalk with each other and are involved in a complex feedback loop [45,46].
Moreover, Jagged1, a canonical ligand of NOTCH receptors, was reported to modulate epithelial-to-mesenchymal transition in breast cancer cells through NOTCH signaling, increase the expression levels of SNAI2 and decrease those of CDH1. Jagged1 also modulates cell proliferation, as the inhibition of Jagged1 expression in MDA-MB-231 cells is sufficient to reduce cell cycle progression [5,47]. We observed that increasing DLK2 expression levels in MDA-MB-231 cells led to a proportional and significant decrease in JAG1 expression ( Figure 4D). These and previous data argue strongly in favor of DLK1 and DLK2 acting upon common molecular mechanisms to modulate NOTCH activation, which in turn may be affected by the decreased expression of at least one of the most important NOTCH canonical ligand gene, such as JAG1. However, our data indicate that JAG1 expression levels do not directly correlate with the invasiveness of DLK2-overexpressing cells. Only MDA-MB-231 cells, with high levels of DLK2 expression that have decreased levels of JAG1 expression, correlate with reduced cell cycle progression and invasiveness in vivo.
We also analyzed the effects of different levels of DLK2 expression on other signaling pathways already known to be affected by DLK1 expression [10] and NOTCH signaling [5], which ultimately affect gene expression and cell behavior in breast cancer cells [48]. The ERK1/2 MAPK pathway, involved in cell proliferation and migration, as well as epithelialto-mesenchymal transition, is reported to crosstalk with the NOTCH1 signaling pathway in MDA-MB-231 cells [49]. DLK2 behaved similarly to DLK1 in the regulation of ERK1/2 MAPK phosphorylation. The overexpression of DLK2 led to the decreased phosphorylation of ERK1/2 MAPK, although only high DLK2 expression led to a significant inhibition of ERK1/2 MAPK phosphorylation ( Figure 5A) [10]. The activation of NOTCH1 signaling was described as leading to increased ERK1/2 MAPK activity, which in turn led to increased expression levels of JAG1 in MDA-MB-231 cells [50]. The inhibition of ERK1/2 MAPK caused by the increased expression levels of DLK2 may help explain the inhibition of JAG1 in our transfectant cells.
The mTOR pathway plays a role in breast cancer cell proliferation and anti-cancer drug resistance and was also reported to crosstalk with the NOTCH signaling pathway in several malignant cells, including breast cancer cells [51]. The phosphorylation status of rpS6 kinase differed between MDA-MB-231 cells transfected with DLK1 or DLK2. In this case, low levels of DLK2 expression caused an increase in rpS6 phosphorylation levels, and high levels of DLK2 expression did not significantly affect rpS6 phosphorylation ( Figure 5B) [10]. Additionally, NOTCH can activate AKT kinase signaling, which is involved in the growth, proliferation, motility, and survival of breast and other cancer cells [5,51]. The overexpression of DLK2 in MDA-MB-231 cells also results in changes in the level of AKT phosphorylation, which again appears to be dependent on DLK2 expression levels. In this case, low DLK2 expression levels cause an increase in AKT phosphorylation levels, whereas high DLK2 expression levels cause a decreased in AKT phosphorylation levels. These data suggest that the growth and invasion capabilities of MDA-MB-231 cells may depend on the level of activation of NOTCH signaling and/or the level of phosphorylation of ERK1/2 MAPK, rpS6 and PI3K/AKT kinases.
The data presented here, and those already published on the effect of DLK1 in MDA-MB-231 cells [10], reveal that both proteins are able to inhibit NOTCH signaling and modulate other signaling pathways affecting the growth and invasive properties of MDA-MB-231 cells in vitro and in vivo. Furthermore, our data show that the level of activation of NOTCH and the degree of phosphorylation of the kinases studied in this work, which appear to be modulated by the expression levels of DLK2, may lead to opposite effects on the features of MDA-MB-231 breast cancer cells. In addition, we revealed that the insufficient inhibition of NOTCH signaling may enhance tumor cell growth, which should be considered for future treatment strategies. To our knowledge, studies on the possible stronger effects of the combined activity of DLK1 and DLK2 on breast cancer cell growth and/or their invasive properties have not yet been performed. Thus, further studies are warranted to explore this possibility and could open a new potential therapeutic window using both proteins as therapeutic tools against this aggressive triple-negative form of breast cancer.

Plasmids
The HDLK2S plasmid contains the full-length human DLK2 cDNA (MGC Full-Length clone IMAGE ID 54954558) cloned into the HindIII-NotI restriction sites of the pLNCX2 expression vector [9].

Cell Culture and Transfection Conditions
The human breast cancer cell line MDA-MB-231 was purchased from the American Type Culture Collection (HTB-26). Cells were cultured at 37 • C in a humidified atmosphere at 5% CO 2

Cell Growth Assays
Cell proliferation was measured by MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) assays for 8 days, following the protocol provided by the manufacturer (Sigma). A total of 1500 cells per well were seeded in a 96-well plate, and the culture medium was replaced every two days. Proliferation assays were carried out for eight days. These assays were repeated at least three times.

Invasion Assays
In vitro cell invasion assays were performed in Matrigel BD BioCoat Invasion Chambers (BD Biosciences), following the manufacturer s recommendations. Briefly, 1 × 10 5 transfected cells, resuspended in 500 µL of serum-free medium, were plated in the upper chamber, which was introduced into the lower chamber containing the complete medium. Chambers were incubated for 24 h at 37 • C, in a humidified atmosphere at 5% CO 2 . Noninvasive cells were removed by wiping the top of the membrane, and invasive cells were fixed and stained. Cell invasion through the membranes was determined by microscopically counting cells present in six random fields at 100× magnification. These assays were repeated at least three times.

RNA Extraction and RT-qPCR
For gene expression analysis, cell monolayers were washed twice with PBS and detached with Trypsin/Versene (Lonza). Cells were then collected by centrifugation (180× g, 5 min, at 4 • C) and washed twice with PBS. Total RNA was isolated by using the RNeasy Kit (Qiagen, Valencia, Spain). After this, DNase treatment (Qiagen), RNA (1 µg) was reverse transcribed by using the cDNA RevertAidH Minus First Strand kit (Fermentas, Madrid, Spain), according to the manufacturer's recommendations. To perform RT-qPCR, cDNA was amplified by PCR using the SYBR-GREEN Master Mix and StepOnePlus Real-Time PCR System (Applied Biosystems, Madrid, Spain). The PCR conditions used were an initial denaturation step at 95 • C, followed by 30 s at 60 • C. GAPDH expression was used as a control to compare the CT of the different samples. The primers used to determine the expression level of GAPDH, DLK2, HEY1, HEY2, HES1, NOTCH1, NOTCH2, NOTCH3, NOTCH4, JAG1, CCND1, CDKN1A, SNAI2, CDH2 and CDH1 were previously described [10].

Luciferase Assays
To estimate NOTCH-dependent transcriptional activity, MDA-MB-231 cells were cotransfected with the pGLucWT (CSL/RBP-Jk-LUC) plasmid and the pLNCX2 (empty vector) or DLK2 expression (HDLK2S) plasmids. The pGLucWT plasmid contains the luciferase reporter gene under the control of a promoter with four copies of a CSL/RBP-Jk (recombining binding protein suppressor of hairless) binding site [6]. To measure the inhibition of NOTCH-dependent transcriptional activity caused by DAPT, cells were co-transfected with pLNCX2 and pGLucWT plasmids and incubated in the presence of 10 µM DAPT for 24 h. To measure luciferase activities, cells were lysed and processed by using the dual luciferase kit (Dual-Luciferase Reporter Assay System, Promega, Madrid, Spain) 24-48 h after transfection, following the manufacturer's recommendations. To normalize the data obtained, cells were also transfected with pRLTK (renilla expression plasmid). Luciferase and renilla activities were measured on a Monolight 3096 Microplate Luminometer (Becton Dickinson). These assays were repeated at least three times.

Animal Studies
Female nude athymic-Foxn1nu mice (5-week-old) were supplied by Harlan Laboratories (Barcelona, Spain). Stably transfected MDA-MB-231 cells (2.5 × 10 6 cells resuspended in 200 µL of PBS) were injected subcutaneously into the dorsal flanks of the mice for a total of 5 mice per group. Mice were sacrificed 41 days after injection. Tumor size was calculated by the following formula: length (mm) × width (mm). In vivo experiments were performed in accordance with Spanish and European regulations and were approved by the Animal Care and Use Committee of the University of Castilla-La Mancha.

Statistical Analysis
Data are presented as the mean ± S.D of at least three independent assays. Data were also analyzed with the GraphPad and/or SPSS software packages for two-tailed Student's t-test or Mann-Whitney U test to determine statistical significance in relation to cells transfected with pLNCX2 (empty vector). A p value ≤ 0.05 was considered statistically significant (*); a p value ≤ 0.01 was considered highly statistically significant (**); and a p value of ≤0.001 was considered extremely statistically significant (***).

Conclusions
Our data revealed for the first time that different levels of DLK2 expression in MDA-MB-231 breast cancer cells can produce opposite effects on tumor invasiveness and growth rate both in vitro and in vivo. These opposite effects are most likely due to the induction of different levels of NOTCH signaling and ERK1/2, AKT and mTOR kinases phosphorylation. Low levels of DLK2 expression resulted in a slight inhibition of NOTCH signaling activation, the increased phosphorylation of rpS6 and AKT kinases, and increased invasiveness and cell growth in vivo and in vitro. Conversely, high levels of DLK2 expression produced a strong inhibition of NOTCH signaling, decreased the phosphorylation of ERK1/2 MAPK and AKT kinases, decreased cell growth rate in vitro and failed to establish tumors in vivo.
The data presented in this work and in our previously published work reveal that both DLK proteins are capable of inhibiting NOTCH signaling and modulating kinase signaling pathways involved in cancer. Most likely, the level of NOTCH activation, modulated by DLK1 and DLK2, may led to opposite effects on the invasive properties of MDA-MB-231 breast cancer cells and tumor cell growth in vitro and in vivo. Furthermore, we revealed that the insufficient inhibition of NOTCH signaling may enhance tumor cell growth. This aspect should be considered for patient treatment strategies against this aggressive form of breast cancer. Figure 7 summarizes the effects of the levels of DLK2 expression on MDA-MB-231 cell growth and invasion properties revealed in this work.
Our data revealed for the first time that different levels of DLK2 expression in MDA-MB-231 breast cancer cells can produce opposite effects on tumor invasiveness and growth rate both in vitro and in vivo. These opposite effects are most likely due to the induction of different levels of NOTCH signaling and ERK1/2, AKT and mTOR kinases phosphorylation. Low levels of DLK2 expression resulted in a slight inhibition of NOTCH signaling activation, the increased phosphorylation of rpS6 and AKT kinases, and increased invasiveness and cell growth in vivo and in vitro. Conversely, high levels of DLK2 expression produced a strong inhibition of NOTCH signaling, decreased the phosphorylation of ERK1/2 MAPK and AKT kinases, decreased cell growth rate in vitro and failed to establish tumors in vivo.
The data presented in this work and in our previously published work reveal that both DLK proteins are capable of inhibiting NOTCH signaling and modulating kinase signaling pathways involved in cancer. Most likely, the level of NOTCH activation, modulated by DLK1 and DLK2, may led to opposite effects on the invasive properties of MDA-MB-231 breast cancer cells and tumor cell growth in vitro and in vivo. Furthermore, we revealed that the insufficient inhibition of NOTCH signaling may enhance tumor cell growth. This aspect should be considered for patient treatment strategies against this aggressive form of breast cancer. Figure 7 summarizes the effects of the levels of DLK2 expression on MDA-MB-231 cell growth and invasion properties revealed in this work.