ATRA sensitivity of TNBC cell-lines characterized by N1ICD expression
The ATRA-score is an experimental index which measures the anti-proliferative action of ATRA in cancer cells (16),(17). The low ATRA-score values determined in most TNBC cell-lines indicate that they are generally resistant to ATRA (Fig. 1A). Indeed, only three (HCC-1599, MB-157 and MDA-MB157) of the eighteen TNBC cell-lines considered are endowed with high ATRA-score values and respond to ATRA. Two additional TNBC cell-lines (HCC-1806 and HCC-70) present with measurable ATRA-scores, while all the other ones show undetectable values. Remarkably, HCC-1599, MB-157 and MDA-MB157 cell-lines are reported to be characterized by internal-deletions of the NOTCH1-gene (19),(20),(21),(22) which cause constitutive expression of N1ICD, the γ-secretase cleavage product and transcriptionally active form of NOTCH1 (23). Consistent with this, HCC-1599, MB-157 and MDA-MB157 are the only TNBC cell-lines expressing N1ICD (Fig. 1A).
To verify the presence of NOTCH1-gene alterations in HCC-1599, MB-157 and MDA-MB157 cells, we performed RNA-sequencing (RNA-seq) experiments. The results demonstrate an exon expression imbalance of the NOTCH1-gene, which indicates the presence of an internal deletion (Fig. 1B). In HCC-1599 cells, the deletion involves exons 3–27, it eliminates the NRR (Negative-Regulatory-Region) from the primary NOTCH1-gene product (19) and it results in a gain-of-function phenotype (19),(20),(22),(24),(25). In MB-157 and MDA-MB157 cells, we do not detect the reported SEC16A-NOTCH1 fusion-gene product (20), although the two cell-lines show deletion of NOTCH1 exons 2–27. Sequencing of the PCR products (exon-1/exon-28 primers) obtained from the HCC-1599, MB-157 and MDA-MB157 RNA confirms the presence/position of the NOTCH1-gene internal deletions (Fig. 1C-D).
NOTCH1 role in the growth of ATRA-sensitive TNBC cell-lines
To investigate whether N1ICD activation contributes to the proliferation of HCC-1599, MB-157 and MDA-MB157 cells, we conducted studies with the γ-secretase inhibitor, DAPT [N-(N-(3,5-difluorophenacetyl)-L-alanyl)-S-phenylglycine-t-butyl-ester].
First, we transfected HCC-1599, MB-157, MDA-MB157 and three TNBC/retinoid-insensitive (MDA-MB231, HCC-38 and MDA-MB436) cell-lines with a reporter construct, containing a luciferase cDNA driven by the promoter of the NOTCH1 target-gene, HES1 (26) (27). Consistent with constitutive NOTCH1 activation, HCC-1599, MB-157 and MDA-MB157 cells show high basal levels of luciferase-activity which are reduced by DAPT (Fig. 2A). In contrast, vehicle- and DAPT-treated MDA-MB231, MDA-MB436 or HCC-38 cells are devoid of significant luciferase-activity. In HCC-1599, MB-157 and MDA-MB157 cells, DAPT-dependent suppression of luciferase-activity is due to the expected inhibition of NOTCH1 cleavage into N1ICD (Fig. 2B).
Subsequently, we evaluated the anti-proliferative action of DAPT (1µM, 9 days). DAPT reduces the proliferation/survival of HCC-1599, MB-157 and MDA-MB157 cells, while it does not affect MDA-MB231, MDA-MB436 and HCC-38 cell-growth (Fig. 2C). In HCC-1599, MB-157 and MDA-MB157 cells, the action of DAPT is time- and concentration-dependent (Fig. 2D) and it is replicated with PF-03084014 (28),(29),(30), another γ-secretase inhibitor (Fig. 2E).
Cross-resistance to ATRA of DAPT-resistant cells
A comparison of the dose-response and time-dependent curves obtained in HCC-1599, MB-157 and MDA-MB157 cells with ATRA (Suppl.Fig.S1) and DAPT/PF-03084014 (Fig. 2D-E) indicates that the more a cell-line is sensitive to the retinoid, the more it responds to γ-secretase inhibitors. To support the idea, we developed MB-157 cells with induced DAPT-resistance. Long-term culturing of MB-157 cells in the presence of DAPT (1 µM) resulted in the isolation of two independent and DAPT-resistant cell-lines of clonal origin (MB-157RCL7A and MB-157RCL15A). Relative to the parental MB-157 counterpart, MB-157RCL7A and MB-157RCL15A cells show resistance not only to DAPT, but also to ATRA (Fig. 3A). Cross-resistance is specific to ATRA, as no difference in sensitivity to other chemotherapeutics, such as VP16 (etoposide-phosphate), is observed (Fig. 3B). This cross-resistance to DAPT and ATRA supports NOTCH1 involvement in the anti-proliferative responses to the retinoid.
Synergistic anti-tumour effects of ATRA and γ-secretase inhibitors
We studied the growth-inhibitory effects of ATRA, DAPT and ATRA + DAPT combinations in retinoid-sensitive HCC-1599/MB-157/MDA-MB157 cells and retinoid-resistant MDA-MB231/MDA-MB436/HCC-38 cells (Fig. 3C). ATRA causes a time-dependent reduction in the growth of HCC-1599, MB-157 and MDA-MB157 cells. The anti-proliferative effect of DAPT is obvious in HCC-1599 and MB-157, while it is less evident in MDA-MB157 cells. In MB-157 cells, ATRA + DAPT is significantly more active than ATRA or DAPT alone. A similar effect is likely to occur in HCC-1599 cells too (Fig. 3C), although the strong anti-proliferative action of ATRA and DAPT alone masks the interaction. As for MDA-MB157 cells, they are only mildly sensitive to DAPT (see also Fig. 2C) and the combination of ATRA + DAPT is no more effective than ATRA alone. The lack of constitutive NOTCH1 activation renders MDA-MB231, MDA-MB436 and HCC-38 cells equally resistant to DAPT, ATRA and ATRA + DAPT. In HCC-1599 and MB-157 cells, we evaluated the additive/synergistic nature of the potentiating effects exerted by ATRA + DAPT. Isobologram analysis (31),(32) of the dose-response data shows synergistic interactions between ATRA and DAPT in both cell-lines (Fig. 3D). In HCC-1599 cells, the synergism is confirmed with combinations of ATRA and PF-03084014 (Suppl.Fig.S2).
To establish whether the anti-tumour action of ATRA and γ-secretase inhibitors is replicated in vivo, we performed experiments with xenografts of HCC-1599 cells. Tumor-bearing animals were treated with vehicle, ATRA, PF-03084014 or ATRA + PF-03084014 for 18 days. The volume of the tumors was determined up to 15 days following treatment cessation (Fig. 4A). ATRA and PF-03084014 reduce the growth of HCC-1599 tumours (16),(33). The ATRA + PF-03084014 combination is more effective than each compound alone. The efficacy of the various treatments was further evaluated using “Percentage-Growth-Inhibition” (%GI) and “Absolute-Growth-Delay” (AGD). The %GI value is indicative of short-term anti-tumor effects, while AGD defines long-term delay of tumor regrowth. ATRA + PF-03084014 causes a more significant reduction of the %GI value than ATRA or PF-03084014 alone (Fig. 4B). At the end of all treatments, tumours invariably tend to regrow. Nevertheless and consistent with an increased long-term efficacy of the combination, the AGD value is higher following treatment with ATRA + PF-03084014 than ATRA or PF-03084014 alone (Fig. 4C).
The NOTCH1-gene internal deletion observed in HCC-1599, MDA-MB157 and MB-157 cells results in the expression of an amino-terminal deleted transmembrane precursor protein (TM-N1ICD) which is constitutively cleaved into N1ICD by γ-secretase (23). Thus, we evaluated whether ATRA, DAPT and ATRA + DAPT perturb the levels of N1ICD and TM-N1ICD. In the three cell-lines, DAPT suppresses N1ICD and increases the levels of the TM-N1ICD precursor, as a consequence of γ-secretase inhibition (Fig. 5A). ATRA reduces the amounts of N1ICD in HCC-1599 and MDA-MB-157 cells, while no effect is observed in the MB-157 counterpart. Consistent with this, the levels of TM-N1ICD are reduced by the combination of ATRA + DAPT only in HCC-1599 and MDA-MB-157 cells. Overall the data suggest that the mechanisms underlying N1ICD down-regulation by ATRA and DAPT are different. In addition, they indicate that ATRA exerts diverse effects on the NOTCH1 pathway in MB-157 cells relative to the HCC-1599 and MDA-MB-157 counterparts. Consistent with the last observation, ATRA causes a significant down-regulation of the TM-N1ICD mRNA in HCC-1599 and MDA-MB-157, but not in MB-157, cells (Fig. 5B). Exposure of MB-157RCL7A and MB-157RCL15A cells to DAPT for 24 hours does not cause the same decrease of N1ICD intracellular levels observed in parental MB-157 cells (Fig. 5C). Noticeably, ATRA does not alter the amounts of N1ICD in either parental or MB-157RCL7A and MB-157RCL15A cells.
To evaluate the specificity of the effects exerted by ATRA and DAPT we performed similar experiments in MDA-MB-231, MDA-MB-436 and HCC-38, three TNBC cell-lines of the panel which express significant amounts of the intact NOTCH1 protein (Fig. 5D) and do not respond to ATRA (Fig. 1) or DAPT (data not shown). In the three cell-lines, DAPT and ATRA alone or in combination exert no significant effects on the constitutive amounts of NOTCH1 protein and do not induce the appearance of the N1ICD cleavage product.
Transcriptomic perturbations afforded by ATRA and DAPT alone and in combination
To get insights into the early perturbations afforded by the retinoid and the γ-secretase inhibitor on the gene-transcription profiles of HCC-1599 and MB-157 cells, we performed comparative RNA-seq studies in the two cell-lines exposed to ATRA, DAPT and ATRA + DAPT for 8 hours.
In HCC-1599 cells, ATRA and DAPT up- and down-regulate a large number of genes (FDR < 0.05) (Suppl.TableS1; Fig. 6A). GSEA (Gene-Set-Enrichment-Analysis; HALLMARK annotations) of the genes down-regulated by ATRA and DAPT indicates that both compounds modulate the MYC-dependent gene-network negatively (Suppl.TableS2; Fig. 6B). This suggests that down-regulation of the MYC-pathway contributes to the anti-proliferative action of the two compounds. Furthermore, both ATRA and DAPT reduce oxidative-phosphorylation, consistent with a growth-inhibitory action involving a decrease in mitochondrial activity (34). As for the up-regulated pathways, ATRA causes a significant enrichment of the “Interferon-alpha-response” and “Interferon-gamma-response” gene-networks (18). Remarkably, “Interferon-alpha-response” is the only up-regulated gene-network equally enriched by ATRA and DAPT (Suppl.TableS2; Fig. 6B). The RNA-seq data obtained in HCC-1599 cells were further analysed to identify genes commonly up- or down-regulated by ATRA and DAPT. To this purpose, we first reduced the number of potential hits, using a threshold value for the expression fold-change caused by ATRA or DAPT (> 40%) (Suppl. Table S1; Fig. 6A). ATRA up-regulates 43% and down-regulates 28% of the genes regulated by DAPT in the same direction (Fig. 6C-D). The large fraction of common genes regulated by ATRA and DAPT is consistent with an ATRA-dependent down-regulation of the N1ICD transcription factor. The ATRA + DAPT combination causes a more sustained up- and down-regulation of 49% and 77% of these common genes, respectively (Suppl.TableS1; Fig. 6A; Fig. 6C-D). This suggests that many of the common genes are regulated by ATRA and DAPT via different regulatory mechanisms. We used the ATRA/DAPT common genes to generate a NOTCH1-oriented protein-protein interaction network. The results indicate that ATRA and DAPT induce the expression of 23 genes and reduce the expression of 11 genes whose products interact with NOTCH1 directly or indirectly (Suppl.Fig.S3).
In MB-157 cells, ATRA, DAPT and ATRA + DAPT modulate the expression of a smaller number of genes than in HCC-1599 cells (Suppl.TableS1; Fig. 7A). GSEA demonstrates that ATRA up-regulates three gene-networks significantly (Fig. 7B). As observed in HCC-1599 cells, “Interferon-alpha-response” is one of these top-enriched gene-networks. Surprisingly, ATRA up-regulates the MYC-dependent gene-network, which is the opposite of what is occurring in ATRA- or DAPT-treated HCC-1599 cells and DAPT-treated MB-157 cells. The fraction of genes commonly regulated by ATRA and DAPT (Fig. 7C) is much smaller in MB-157 than HCC-1599 cells (up-regulated = 29% vs 43%; down-regulated = 3% vs 28%), which supports the idea that the retinoid affects the NOTCH pathway by acting downstream of NOTCH1 expression and N1ICD activation. Co-treatment of MB-157 cells with ATRA + DAPT enhances the action of ATRA or DAPT in the majority (up-regulated = 77%; down-regulated = 64%) of these common genes (Suppl. Table S1; Fig. 7C-D). This is consistent with the hypothesis that the majority of the ATRA/DAPT common genes are modulated by the two compounds via different mechanisms not only in HCC-1599, but also, in MB-157 cells.
We compared the global effects of ATRA, DAPT and ATRA + DAPT on the 50 HALLMARK gene-sets in HCC-1599 and MB-157 cells (Fig. 7E). ATRA, DAPT and/or ATRA + DAPT down-regulate the “E2F-targets”, “G2M-checkpoint” and “MYC-targets” gene-networks in both cell-lines. “E2F-targets”, “G2M-checkpoint” down-regulation may simply be the consequence of the anti-proliferative effect afforded by the two compounds alone or in combination, while the effect on “MYC-targets” is likely to be of mechanistic relevance. It is also noticeable that ATRA + DAPT leads to an up-regulation of the “Interferon-alpha-response” and “Interferon-gamma-response” in both HCC-1599 and MB-157 cells.
Role of RARα and RARβ in ATRA anti-proliferative effects
ATRA is a pan-RAR agonist, activating the RARα, RARβ and RARγ retinoid-receptors with equal efficiency (35),(36). To identify the RAR receptor(s) mediating the activity of ATRA in TNBC cell-lines, we exposed HCC-1599, MB-157 and MDA-MB157 cells to AM580 (RARα agonist), UVI2003 (RARβ agonist), BMS961 (RARγ agonist) as well as ATRA for 9 days (Fig. 8A). AM580 reduces the growth of these cell-lines in a dose-dependent manner and the effects of the RARα agonist and ATRA are quantitatively similar. In contrast, UV2003 and BMS961 do not alter the growth of HCC-1599, MB-157 and MDA-MB157 cells. Thus, ligand-dependent activation of RARα seems to be the primary determinant of ATRA-dependent action in TNBC cell-lines, as previously observed in luminal breast-cancer cell-lines (16).
To establish whether ATRA modulates the expression of RARα, RARβ and RARγ, we exposed six TNBC and six retinoid-sensitive Luminal breast cancer cell-lines to ATRA for 24 hours (Fig. 8B). Consistent with binding, activation and proteasome-degradation of the receptor (37), ATRA reduces the levels of RARα in almost all the cell-lines. In contrast, ATRA has no effect on the basal expression levels of RARγ in any cell-line (data not shown). As for RARβ, the product of a direct retinoid target-gene (38), no cell-line expresses detectable amounts of the receptor in basal conditions and the action of ATRA depends on the cellular phenotype and retinoid-sensitivity. In the TNBC context, ATRA up-regulates RARβ only in HCC-1599, MB-157 and MDA-MB157 cell-lines (Fig. 8B), indicating an association between RARβ induction and ATRA-sensitivity. In the luminal context, there is no cell-line which responds to ATRA with an induction of RARβ. Thus, RARβ is likely to play a functional role in ATRA growth-inhibitory action only in the case of TNBC cell-lines. With respect to this, it is remarkable that ATRA does not induce RARβ in DAPT- and retinoid-resistant MB-157RCL7A and MB-157RCL15A cells (Fig. 8C).
To support the functional relevance of RARβ induction, we silenced the retinoid-receptor in HCC-1599 and MB-157 cells by stable infection of retroviral constructs containing two distinct shRNAs targeting RARβ (shRARβ-a and shRARβ-b) and a control shRNA (shCTRL). In both HCC-1599 and MB-157 cell-lines, shRARβ-a and shRARβ-b suppress the ATRA-dependent induction of RARβ (Fig. 8D-E,left). In contrast, shCTRL does not alter the up-regulation of RARβ caused by ATRA in parental HCC-1599 and MB-157 cells. In these experimental conditions, RARβ knock-down induces ATRA resistance. Indeed, ATRA-dependent growth-inhibition is significantly reduced in shRARβ-a/shRARβ-b infected HCC-1599 and MB-157 cells relative to the parental or shCTRL infected counterparts (Fig. 8D-E,right). The functional results obtained indicate that RARβ contributes to the growth-inhibitory action of ATRA in sensitive TNBC cell-lines.
ATRA-dependent RARβ induction is mediated by RARα activation, as the phenomenon is replicated by AM580 (Fig. 8F), while UV2003 and BMS961 are completely inactive (data not shown). The observation is supported by the results obtained with the RARα antagonist, ER-50891, which blocks the induction of RARβ triggered by ATRA or AM580 in HCC-1599 cells (Fig. 8F). RARβ may contribute to ATRA anti-proliferative action in a ligand-dependent or ligand-independent manner. However, the comparative growth-inhibitory studies performed in HCC-1599 cells with the above mentioned RAR agonists support the idea that RARβ contributes to ATRA-dependent growth-inhibition in a ligand-independent manner (Fig. 8G). In fact, the anti-proliferative effect exerted by ATRA in HCC-1599 cells is entirely recapitulated by the RARα agonist, AM580, which is incapable of binding and activating RARβ at the concentrations used in our experimental conditions.