BCL-XL blockage in TNBC models confers vulnerability to inhibition of specific cell cycle regulators

Cell cycle regulators are frequently altered in Triple-Negative Breast Cancer (TNBC). Emerging agents targeting these signals offer the possibility to design new combinatorial therapies. However, preclinical models that recapitulate TNBC primary resistance and heterogeneity are essential to evaluate the potency of these combined treatments. Methods: Bioinformatic processing of human breast cancer datasets was used to analyse correlations between expression levels of cell cycle regulators and patient survival outcome. The MMTV-R26Met mouse model of TNBC resistance and heterogeneity was employed to analyse expression and targeting vulnerability of cell cycle regulators in the presence of BCL-XL blockage. Robustness of outcomes and selectivity was further explored using a panel of human breast cancer cells. Orthotopic studies in nude mice were applied for preclinical evaluation of efficacy and toxicity. Alterations of protein expression, phosphorylation, and/or cellular localisation were analysed by western blots, reverse phase protein array, and immunocytochemistry. Bioinformatics was performed to highlight drug's mechanisms of action. Results: We report that high expression levels of the BCL2L1 gene encoding BCL-XL and of specific cell cycle regulators correlate with poor survival outcomes of TNBC patients. Blockage of BCL-XL confers vulnerability to drugs targeting CDK1/2/4, but not FOXM1, CDK4/6, Aurora A and Aurora B, to all MMTV-R26Met and human TNBC cell lines tested. Combined blockage of BCL-XL and CDK1/2/4 interfered with tumour growth in vivo. Mechanistically, we show that, co-targeting of BCL-XL and CDK1/2/4 synergistically inhibited cell viability by combinatorial depletion of survival and RTK/AKT signals, and concomitantly restoring FOXO3a tumour suppression actions. This was accompanied by an accumulation of DNA damage and consequently apoptosis. Conclusions: Our studies illustrate the possibility to exploit the vulnerability of TNBC cells to CDK1/2/4 inhibition by targeting BCL-XL. Moreover, they underline that specificity matters in targeting cell cycle regulators for combinatorial anticancer therapies.


Introduction
Targeting of the anti-apoptotic protein BCL-XL together with anti-mitotic agents has been proposed as an efficient therapeutic strategy for different human cancers, including triple-negative breast cancer (TNBC), a particular aggressive subtype of breast cancer [1]. TNBC is defined by the lack of oestrogen (ER) and progesterone (PR) receptors and the absence of amplification/overexpression of the HER2 receptor tyrosine kinase (RTK) [2]. The disease accounts for ~15% of all breast cancer types and is characterised by an extraordinary molecular heterogeneity. Current therapeutic options for targeted treatment are limited due to primary or acquired resistance, and major efforts are devoted to search for molecular alterations and predict vulnerabilities for effective targeted therapy [3]. However, the high heterogeneity of TNBC challenges the identification of generic targets, and highlights the requirement for precision therapy and preclinical models that recapitulate TNBC heterogeneity.
We have recently reported the generation of a rather unique mouse model in which a subtle increase in the wild-type MET RTK expression levels in the mammary gland (MMTV-R26 Met mice) leads to spontaneous tumour formation faithfully recapitulating TNBC features [4]. These include histological, molecular and signalling heterogeneity, as well as primary resistance to chemotherapy and approved targeted therapies [5,6]. We further exploited the MMTV-R26 Met TNBC model to identify a highly effective therapeutic protocol based on the combined inhibition of the anti-apoptotic molecule BCL-XL and of the cell cycle checkpoint regulator WEE1 [4]. Cell cycle proteins are frequently overexpressed and/or overactivated in several cancer types including TNBC. For example, loss of RB or p16 INK4 is frequent across TNBC subtypes [7,8]. Alterations in cyclin D and E, CDK4/6, and CDK2 in TNBC have also been reported [9]. Furthermore, mutations in p53, a master cell cycle regulator of the G1/S checkpoint, are present in a large proportion of TNBC, thus leaving cells to mainly rely on the G2/M checkpoint to maintain DNA integrity [10,11]. Therefore, targeting cell cycle regulators provide a clear rationale for designing anti-cancer therapy [10]. The vulnerability of TNBC to inhibitors of cell cycle regulators has been demonstrated by several studies using different drugs targeting distinct signalling components. For example, the CDK4/6 inhibitor Palbociclib, effectively used in ER-positive breast cancer, has recently been explored in combination with chemotherapy in RB-positive TNBC cell models [12][13][14]. Additionally, it has been reported that a combinatorial treatment based on CDK4/6 plus BET inhibitors leads to cell division errors and death, although leading to the emergence of heterogeneous mechanisms of resistance [15].
In the present study, we employed the MMTV-R26 Met model system that recapitulates TNBC heterogeneity and primary resistance to explore the effect of inhibiting specific cell cycle regulators while blocking BCL-XL function. Our results show that targeting specific cell cycle components is important and uncover a detrimental effect on TNBC cells of CDK1/2/4 plus BCL-XL inhibition. We provide evidence that this combinatorial targeting significantly reduces the levels of RTK and AKT signalling besides perturbing cell cycle and DNA repair.
Cell viability assay. Cell viability assay was performed on MMTV-R26 Met MGT and human breast cancer cells as previously described [4]. Briefly, cells were plated in 96-well plates, and treated after 24hrs with either single or combined drugs at the indicated concentrations. Cell viability was determined 48hrs later using the Cell Titer Glo Luminescent Assay (Promega). Data are mean values of at least three independent experiments done in triplicates.
Immunocytochemistry. Protocols used were as described in [4]. Percentage of TritonX-100 and antibodies used for immunofluorescence staining are detailed in Table S8. Quantification of nucleus versus cytoplasmic staining intensity was determined using Image J Intensity Ratio Nuclei Cytoplasm Tool, RRID:SCR_018573.
Western blotting. Protein extracts were prepared, and western blot analysis was performed as previously described [4]. The antibodies used are reported in Table S8. ACTIN and Ponceau staining were used as loading controls. Ponceau stainings are shown in supplementary information (non-edited gels). For densitometric analysis, the intensity of each protein band was measured using the ImageJ software (National Institutes of Health, USA).

Reverse phase protein array (RPPA). Protein lysates of MMTV-R26 Met MGT cells treated or not with
A1155463 (1µM), R547 (3µM), or the drug combination A1155463 + R547 (1µM, 3µM) for 12hrs, were prepared according to the MD Anderson Cancer Center platform instructions. RPPA of cells treated with Adavosertib (3µM), or A1155463 + Adavosertib (1µM, 3µM) was previously reported [4], and used in this study to compare signalling changes occurring with the different drug combinations. Samples were screened with 426 antibodies to identify signalling changes in protein expression and phosphorylation levels. For data shown in Figure 3C and S2A, a cut-off p-value < 0.05 was applied. The Genseset enrichment analysis (GSEA) software [21,22]  For multiple comparisons, we used ANOVA test followed by Tukey test. All statistical analyses were performed using the GraphPad Prism and software.

High expression levels of BCL-XL and of selective cell cycle regulators correlate with poor survival outcomes of TNBC patients.
Recent studies using breast cancer tissue-specific microarray databases (GSE42568, GSE45827, and GSE54002) have reported a strong enrichment in cell cycle pathway genes [23], which we also illustrate in Figure 1A. Common up-and down-regulated genes from these three microarray datasets were highly enriched in cell cycle regulation pathways from KEGG and WikiPathways ( Figure 1B, Table S1).
Analysis of the Pan-Cancer Atlas cohort of the TCGA database revealed that about 50% of breast cancer patients have altered cell cycle genes, with breast cancer ranking second among all cancer types analysed for the percentage of amplification events ( Figure 1C). Among the four breast cancer subtypes, HER2 + and TNBC patients show the highest frequency of alteration in cell cycle regulators ( Figure   1D). Gene amplification is the major type of alterations found in breast cancer patients ( Figure 1D).
We next used a TNBC microarray database (GSE31519; 580 patients) to analyse the relevance of cell cycle gene levels in patient outcomes. We found that TNBC patients with high levels of BCL-XL or specific cell cycle regulators such as CDK1, CDK6 and WEE1, exhibit a shorter overall survival rate ( Figure 2A and B, Table S2). Interestingly, patients that concurrently highly express the anti-apoptotic factor BCL-XL and the cell cycle modulators CDK1, CDK6, or WEE1, have poor clinical outcome, thus highlighting the detrimental effect of their concomitant high expression levels ( Figure 2C, Table S2).
Intriguingly, high levels of other cell cycle modulators such as CDK2, CDK4, FOXM1, Aurora A, or Aurora B, either alone or together with BCL-XL levels, is not associated with altered patient survival ( Figure 2B and C, Table S2). The apparent significant difference between the survival curves in the combination sets is mainly due to BCL-XL levels ( Figure 2C, Table S2). Furthermore, patients with high levels of BCL-XL, CDK1, and WEE1 exhibit more aggressive TNBC tumours (classified as grade 3; Figure 2D). This was also observed taking into consideration co-expression of high versus low BCL-XL with CDK1 or WEE1 ( Figure 2E). This significant increase in aggressiveness was not observed in patients with high levels of CDK6 ( Figure 2D and E). These results may suggest a greater implication of specific cell cycle regulators, particularly when overexpressed with BCL-XL, on survival of TNBC patients.

TNBC cells are vulnerable to inhibition of selective cell cycle regulators following BCL-XL blockage.
The above findings, together with our recent studies uncovering the vulnerability of TNBC cells to combined WEE1 and BCL-XL targeting [4], drove us to explore the sensitivity of TNBC cells to inhibition of specific cell cycle regulators in combination with BCL-XL targeting. We addressed this issue using the MMTV-R26 Met TNBC model system, employing RNA-seq analysis, proteomic profiling, and cell viability assay as illustrated in Figure 3A. By Gene Set Enrichment Analysis (GSEA) of RNA-seq outcomes comparing MMTV-R26 Met tumours (n=4) to control mammary gland tissues (MMTV-Cre; n=3), we found a striking enrichment in genes related to cell cycle regulation ( Figure 3B), and to DNA replication and DNA damage pathways ( Figure S2A, B, and Table S3). Specifically, we found significantly high (which control progression through the cell cycle in concert with their cyclin regulatory subunits).
In light of these results, we evaluated cell viability of mammary gland tumour (MGT) cell lines (MGT4, 9, 11, and 13) derived from the MMTV-R26 Met model [4] when treated with inhibitors of various cell cycle regulators, acting during different phases, as illustrated in Figure 3D. Inhibition of either FOXM1 (with FDI-6), CDK4/6 (with Palbociclib), Aurora A or B (with Alisertib or Barasertib, respectively) together with BCL-XL blockage (with A1155463) did not alter the viability of the MMTV-R26 Met TNBC cells we tested ( Figure 3D-F). We observed only a partial response in MGT11 cells following combined inhibition of BCL-XL with CDK4/6, Aurora A or Aurora B ( Figure 3F). In contrast, combined targeting of CDK1/2/4 (with R547) and BCL-XL (with A1155463) was highly deleterious for all MMTV-R26 Met TNBC cells we tested ( Figure 3D and F). Concomitant inhibition of CDK1/2/4 and BCL-XL was synergistic for 3 out of 4 MMTV-R26 Met MGT cell lines, as shown by the Bliss score and the Chou-Talalay combination index score calculation ( Figure 3G and S2C). These results were corroborated by the deleterious effect of CDK1/2/4 and BCL-XL co-targeting we observed in all six human TNBC cell lines tested ( Figure 3H).
In contrast, we found only a modest effect on human non-TNBC cells (with the exception of the BT-474 cells), similar to the non-tumorigenic MMTV-R26 Met MGT2 cells ( Figure 3I and S2D). These findings illustrate the vulnerability of TNBC cells to the inhibition of specific cell cycle regulators when BCL-XL is targeted.

Combined BCL-XL and CDK1/2/4 inhibition interferes with cell cycle and survival signals, and triggers apoptosis.
We next examined, in the context of BCL-XL blockage (by A1155463), the molecular and biological consequences of CDK1/2/4 targeting (by R547) [24], and assessed specificity in alterations compared to WEE1 inhibition (by Adavosertib). We first assessed the status of cell cycle regulators and observed a drastic downregulation of both RB protein expression and phosphorylation levels following CDK1/2/4 targeting (R547), in contrast to unchanged levels following WEE1 inhibition (Adavosertib) ( Figure 4A, B, and S3). This is consistent with RB being a direct target of CDK2 and CDK4/6 [25]. FOXM1 phosphorylation levels were also severely reduced following CDK1/2/4 inhibition ( Figure 4A, B, and S3). This is again consistent with FOXM1 being a direct target of CDK1 and CDK2 [26]. In contrast, WEE1 led to a slight increase in phospho-FOXM1 concurrent with an increase in CDK1 activity, as illustrated by loss of CDK1 phosphorylation ( Figure 4A, B, and S3). Changes in p53 protein and phosphorylation levels were more pronounced following WEE1 inhibition than with CDK1/2/4 inhibition ( Figure 4A and S3), although varying in relation to the MGT p53 status that we reported previously [4]. These findings were corroborated by a semi-quantitative reverse phase protein array (RPPA) proteomic profiling, a high-throughput antibody-based technique to analyse protein activities in signalling networks ( Figure   S2E). Interestingly, we found a specific downregulation of Cyclin D3 levels following CDK1/2/4 inhibition ( Figure 4C), correlating with downregulation of PP1 ( Figure 4C), a phosphatase that stabilizes Cyclin D3 by keeping it in a dephosphorylated state [27]. No significant changes were observed in Cyclin B1, Cyclin D1, and Cyclin E1 levels ( Figure S2F). Collectively, these findings illustrate that CDK1/2/4 inhibition has drastic consequences on cell cycle regulators we tested, not significantly exacerbated by BCL-XL targeting.
We then analysed the cell cycle effects of R547 alone or in combination with BCL-XL targeting (with We next explored the consequences of BCL-XL+CDK1/2/4 inhibition on regulators of cell survival and DNA damage and compared them with those linked to the BCL-XL+WEE1 targeting that we previously reported [4]. We found a drastic decrease in anti-apoptotic XIAP and MCL1 protein levels in cells treated with BCL-XL+CDK1/2/4 targeting, accompanied by increased cleavage of Caspase3, Caspase 7, and PARP, as shown by western blot ( Figure 5A, B, and S3) and RPPA ( Figure 5C) analyses. Moreover, the combined treatments led to a high extent of DNA damage in cells, as revealed by the levels of γH2AX (a histone variant considered as a double-strand break sensor), by western blot ( Figure 5A) and immuno-cytochemistry ( Figure 5D). The high levels of DNA damage were accompanied by a downregulation of protein and/or phosphorylation levels of both ATM and ATR in cells treated with BCL-XL+CDK1/2/4 inhibitors, in contrast to unchanged, or a slight upregulation following BCL-XL+WEE1 blockage ( Figure 5A and B), indicating a deficiency in the DNA damage detection mechanism induced by the A1155463+R547 drugs. Furthermore, this combined BCL-XL+CDK1/2/4 treatment did not affect the percentage of phospho-S 10 -Histone H3 (pH3)-positive cells, and did not trigger any mitotic catastrophe events in comparison to those observed in BCL-XL+WEE1 treated cells ( Figure 5E and F).
This agrees with the blockage of CDK1 by R547, therefore promoting cell cycle exit, in contrast to premature entry into mitosis observed with BCL-XL+WEE1 blockage ( Figure 5F), as we showed in [4].
Together, these results illustrate that CDK1/2/4 inhibition together with BCL-XL targeting is as detrimental as BCL-XL+WEE1 blockage for TNBC cells, with similar perturbations of cell survival regulators, while having distinct effects on cell cycle and DNA repair components.

Combined BCL-XL and CDK1/2/4 inhibition leads to downregulation of RTK and AKT signalling.
To obtain further insights on signalling changes occurring in cells treated with BCL-XL+CDK1/2/4 inhibition and associated with their death, we bioinformatically explored RPPA outcomes on expression and/or phosphorylation levels of 426 proteins ( Figure S2E and Table S5). First, we compared alterations occurring with single (BCL-XL or CDK1/2/4) versus combined BCL-XL+CDK1/2/4 targeting.
To further examine the mechanisms underlying the effects of these drug combinations, we performed a GSEA using all signals from the RPPA outcomes from cells treated with BCL-XL+CDK1/2/4 versus BCL-XL+WEE1 inhibitors. Using the Reactome database, we found that gene-sets related to RTK signalling and second messengers, as well as PI3K/AKT signalling, were among the most significant deregulated signals in MMTV-R26 Met MGT cells treated with BCL-XL+CDK1/2/4 inhibitors ( Figure 6E). This approach delineated a possible mechanism underlying combined BCL-XL plus CDK1/2/4 targeting, showing a pronounced alteration of RTK and AKT signalling.
Interestingly, it has recently been shown that CDK1, together with Aurora kinase, ensures RTK storage The correlation between reduced phosphorylation of AKT and FOXO3a was particularly interesting, considering that AKT phosphorylates FOXO3a leading to its translocation from nucleus to cytoplasm, thus preventing its tumour suppressor function [31]. We therefore explored this aspect by performing immuno-cytochemistry to follow FOXO3a localisation in untreated and treated cells. We found a striking FOXO3a nuclear localization in cells treated with BCL-XL+CDK1/2/4 inhibition compared with cells either untreated or treated with single drugs ( Figure 7D and E). This FOXO3a nuclear retention was accompanied by a transcriptional expression of PUMA and FasL, two FOXO3a target genes involved in apoptosis ( Figure 7F). Together, these findings show that combined BCL-XL and CDK1/2/4 targeting, while perturbing cell cycle regulators, leads to a combinatorial depletion of survival and RTK/AKT signals, restoring FOXO3a tumour suppression actions.

Discussion
In this study, we provide evidence that combinatorial targeting of BCL-XL with CDK1/2/4 could be an efficient therapeutic approach for TNBC. This drug combination was potent across different TNBC subtypes, as demonstrated by its high efficacy in different human TNBC cell lines and in the heterogenous cell lines generated from distinct MMTV-R26 Met tumours. Such combination appears to be less effective on non-TNBC cells, based on the cell lines used in this study.
The alteration of cell cycle regulators in several types of cancer [10], including breast cancer, and the possibility to modulate their function, has fostered the interest to design treatment options targeting them, with some of them already being exploited in clinical trials [10].    Table S1). See Figure S1A   indicated genes. Tumour grades were defined according to standard pathological scores (1-3; 3 corresponds to the most aggressive grade). ns: not significant; * P<0.05; ** P<0.01; ***p<0.001.    Clustering was based on protein interaction. The cut-off value was predefined as p-value < 0.05 and fold change <-0.5 or >0.5. See Figure S4           Relative expression (Log2FC)