Long-term treatment with tamoxifen and palbociclib leads to the downregulation of hormone receptors and alterations in the cell cycle.
Tamoxifen (TR), palbociclib (PR), and double-resistant (TPR) cell variants were generated from T47D and MCF-7 wild-type (WT) cell lines, as described in Materials and Methods. Resistance to these agents was confirmed by proliferation assays. Treatment with the hormone inhibitors tamoxifen and fulvestrant had no significant impact on the proliferation of tamoxifen- and double-resistant cells, whereas palbociclib-resistant cells showed reduced sensitivity to these agents compared to parental T47D-WT cells (Fig. 1.A). These results are consistent with a substantial decrease in ER and progesterone receptor (PGR) expression in T47D and MCF-7 resistant models (Fig. 1.B). Additionally, all resistant variants presented decreased sensitivity to palbociclib compared to T47D-WT, including the tamoxifen-resistant cells (Fig. 1.C). Furthermore, the resistant variants also showed cross-resistance to other CDK4/6 inhibitors such as ribociclib and abemaciclib. Remarkably, T47D-TR and T47D-PR xenografts maintained their resistance to tamoxifen and palbociclib, respectively (Supplementary Figure S1.A).
All T47D-resistant cells showed lower proliferation rates in cell culture as well as decreased tumor growth in vivo compared to parental cells (Fig. 1.D). In particular, the double-resistant cells formed small tumors and were unable to grow further. Notably, estradiol supplementation was necessary for tumor growth in all cell lines (Supplementary Figure S1.B).
Next, we performed cell-cycle phase analysis to determine whether the cells were arrested in any particular phase. Surprisingly, only the palbociclib-resistant variant presented a higher percentage of cells in the G2/M phase, with a simultaneous decrease in the proportion of cells in G0/G1 (Fig. 1.E). Given that palbociclib triggers cell cycle arrest in G0/G125, we speculate that T47D-PR cells were able to bypass this arrest and resume cell cycle progression, accumulating in G2/M. Overall, no significant number of aberrant mitoses was observed in this resistant variant (Supplementary Figure S1.C).
To decipher the mechanisms by which cells acquired resistance, we analyzed the expression of several cell cycle proteins (Fig. 1.F). A significant increase in cyclin E2 (T47D-TR and T47D-TPR) as well as in CDK6 and CDK1 levels (T47D-PR) was observed, along with a decrease in Rb (T47D-TPR), compared to the parental cells. When analyzing MCF-7 variants, we found higher levels of cyclin D1, CDK4, and CDK6 in MCF-7-PR cells. This upregulation of cyclins and CDKs may compensate for the inhibitory effect of the drugs on the cell cycle.
Tamoxifen and palbociclib-resistant cells exhibit an altered phenotype
Some phenotypic changes occurred during the development of resistance, as evidenced by the increased heterogeneity in cell size and morphology compared to the parental cells (Supplementary Figure S1.D). When grown on the basement membrane matrix Geltrex®, the parental cells formed large compact spheroids, some of which even displayed a lumen. However, the resistant variants formed smaller (T47D-TR) or disorganized structures with grape-like morphology (T47D-PR and T47D-TPR) (Fig. 2.A), which has been associated with decreased cell-cell junctions and a more invasive phenotype26. Reduced expression or aberrant (nuclear and cytoplasmic) localization of E-cadherin was observed in both T47D- and MCF-7-resistant cells (Fig. 2.B), in correlation with the increased migratory capacity of MCF-7-TR and MCF-7-PR variants, as evidenced by transwell assays (Fig. 2.C).
Previous studies have suggested that resistance to endocrine and chemotherapy might be associated with enrichment in stem cells, as they exhibit lower sensitivity to treatments27,28. To assess whether this was accurate in our system, we performed mammosphere formation assays to propagate stem cells in vitro29. We observed that mammospheres formed by the parental cells were regular and compact, whereas the resistant cells formed more disorganized structures (Fig. 2.D), which may be associated with their lower E-cadherin expression. Furthermore, T47D-TR cells exhibited a higher mammosphere formation frequency, as shown by ELDA analysis (Fig. 2.D), along with increased expression of stemness markers NANOG and OCT4. Additionally, T47D-PR cells showed higher levels of BCRP, a membrane transporter involved in multidrug resistance (Fig. 2.E).
To further characterize the resistant phenotype and identify associated acquired mutations, we performed whole exome sequencing (WES) of the T47D variants (Supplementary Table S1). Mutations that were not present in the parental cells were filtered by potential pathogenicity, and mutated genes were then grouped according to their function or involvement in cellular processes based on the Reactome pathway database and bibliographic research. Interestingly, in all resistant variants, the altered genes were primarily associated with epithelial-mesenchymal transition and cell migration (23–26% of the total mutated genes), stemness (13–15%), and regulation of metabolism (10–17%), followed by apoptosis/autophagy (11–15%), and cell cycle regulation (8–10%) (Fig. 2.F). Some of the mutated genes that could potentially be associated with resistance to either tamoxifen or palbociclib are shown in the Venn diagram (Fig. 2.F, Supplementary Figure S1.E). Further studies are required to validate the role of these genes in the development of resistance. Interestingly, we did not find any new mutations in CDKs, cyclin D1, cyclin E, or hormone receptor genes. It remains to be determined whether copy number variation and epigenetic or post-transcriptional regulation of these genes are involved.
mTOR inhibition reduces proliferation, migration and mammosphere-forming capacity in the resistant cells.
Previous studies have reported that PI3K/AKT/mTOR and MAPK pathways are involved in the development of resistance to tamoxifen and palbociclib. Therefore, we assessed the expression and activation of proteins associated with these pathways using western blotting. Although we did not observe differences in ERK and PTEN phosphorylation or PI3K expression in the resistant variants compared to the parental cells (Fig. 3.A), we observed enhanced AKT phosphorylation in the tamoxifen-resistant cells and S6 phosphorylation in the palbociclib-resistant cells (Fig. 3.A). Surprisingly, AKT and S6 were not simultaneously overactivated, as expected. Additionally, we found upregulation of PKCα in the tamoxifen-resistant variant MCF-7-TR and the double-resistant variant T47D-TPR.
We next examined whether the decreased ER expression and AKT/S6 overactivation found in the resistant cells were permanent changes or whether they were induced by the continuous presence of the inhibitors in the culture media. After more than 13 days of drug-free growth, resistant cells still exhibited lower ER expression and increased AKT/S6 phosphorylation, confirming that these changes were not reversible upon drug removal (Supplementary Figure S2.A, B).
Given the increased activation of PI3K/AKT/mTOR signaling observed in the resistant cells, we hypothesized that they would be especially sensitive to the inhibition of this pathway. Therefore, we evaluated the effect of specific inhibitors on cell proliferation. We tested increasing concentrations of alpelisib (PI3Kα inhibitor), MK-2206 (pan-AKT inhibitor), and rapamycin (mTORC1 inhibitor) on each resistant variant and compared their responses to those of parental cells. Interestingly, alpelisib had a lower antiproliferative effect in the palbociclib-resistant cells, showing up to a 73% reduction in the response. In contrast, MK-2206 was more effective in the tamoxifen-resistant cells, in agreement with their higher AKT phosphorylation levels, with up to a 74% increase in the response. Finally, rapamycin was more effective in both T47D-TR and T47D-PR, with up to 39% increased response in the palbociclib-resistant cells, which exhibited higher phosphorylation of S6, a downstream target of mTOR (Fig. 3.B). Remarkably, ten times lower concentrations of rapamycin than alpelisib were already effective in reducing cell proliferation.
Next, we assessed the influence of PI3K/AKT/mTOR inhibition on cell migration and stemness, two processes that were altered in the resistant cells. Treatment with everolimus (rapamycin analog) and alpelisib reduced the migration of the palbociclib-resistant cells (Fig. 3.C). However, only everolimus successfully diminished the mammosphere-forming capacity in the tamoxifen-resistant cells (Fig. 3.D). We performed these two analyses in the resistant cell variants where the processes were more evident, considering that wild-type cells have no migratory capacity (not shown) and lower mammosphere-forming capacity (Fig. 2D).
Combined inhibition of mTOR and CDK4/6 shows an enhanced antitumor effect in the resistant cells.
We next evaluated the benefits of incorporating specific PI3K/AKT/mTOR inhibitors into palbociclib therapy for the resistant variants. First, the effect of the inhibitors on target proteins was verified by western blotting (Fig. 4.A). As expected, alpelisib and MK-2206 decreased AKT phosphorylation and, to a lesser extent, S6 phosphorylation in the tamoxifen-resistant cells. However, alpelisib failed to inhibit S6 phosphorylation in the palbociclib-resistant cells, likely because of its higher basal pS6 levels, which could account for the observed lower antiproliferative effect of alpelisib in these cells (Fig. 3.B). As expected, all mTOR inhibitors (including AZD2014, a dual mTORC1/2 inhibitor) efficiently inhibited S6 phosphorylation in all the resistant variants (Fig. 4.A, Supplementary Figure S2.C). Furthermore, palbociclib inhibited Rb phosphorylation only in tamoxifen-resistant cells but increased the expression of cyclin D1 in both T47D-TR and T47D-PR cells. The combination of mTOR inhibitors with palbociclib deepen the decrease of cyclin D1, cyclin E2, and cyclin A expression as well as pRb and pS6 inhibition (Fig. 4.A, Supplementary Figure S2.C). We subsequently found that, in all resistant variants, combined treatment with rapamycin and palbociclib reduced cell proliferation to a greater extent than treatment with each inhibitor alone (Fig. 4.B, Supplementary Figure S2.D).
Regarding tumor growth, the combined treatment was more effective than each drug alone, promoting tumor regression in the tamoxifen-resistant variant. In the palbociclib-resistant variant, rapamycin as monotherapy was able to induce tumor regression, and the combination with palbociclib did not induce a greater effect (Fig. 4.C). It is relevant to note that the resistant cells/xenografts were able to grow even in presence of drugs, although at a lower rate than the parental lines. Tumor immunostaining revealed that both palbociclib and rapamycin alone diminished pRb and pS6 expression in tamoxifen-resistant tumors, whereas the combined treatment improved the inhibitory effect (Fig. 4.D). In palbociclib-resistant tumors, treatment with either palbociclib or rapamycin was insufficient to inhibit pRb and pS6, whereas their combination decreased the phosphorylation of both proteins.
Some studies have shown that the cyclin D1/CDK4/6 complex can activate the PI3K/AKT/mTOR pathway through inhibition of TSC2, a negative regulator of mTORC121,22. Since we found that inhibition of the cyclin D1/CDK4/6/Rb axis with palbociclib did not cross-regulate AKT or S6 phosphorylation in the palbociclib- and double-resistant models (Fig. 4.A, Supplementary Figure S2.C), we also evaluated the cross-effect of CDK4/6 inhibitors in sensitive T47D and MCF-7 wild-type cells. In this context, abemaciclib was most effective in reducing both Rb and S6 phosphorylation, although AKT remained active (Fig. 4.E, Supplementary Figure S2.E).
Effect of PI3K/AKT/mTOR inhibition on proliferation of tamoxifen- and palbociclib-resistant PDCs
To assess whether the results found in cell lines and summarized in the Graphical Abstract (see Supplementary Material) could be validated in models that better reflect the complexity and heterogeneity of patient tumors, we used nine PDXs established from ER + breast tumors. Genomic alterations in the PDXs have been previously analyzed using the MSK-IMPACT panel30. Some of the relevant alterations associated with the PI3K/AKT/mTOR and cyclin D1/CDK4/6/Rb pathways are shown in Fig. 5.A.
Cells were isolated from PDXs (PDCs), as described in Materials and Methods (Fig. 5.B). We performed ex vivo assays using PDCs to evaluate their response to ER, PI3K/AKT/mTOR, and CDK4/6 inhibitors as well as the expression and activation of proteins associated with these pathways. The sensitivity of each PDC to tamoxifen and palbociclib is shown in Fig. 5.C.
The expression of PTEN remained mostly constant among the different PDCs, regardless of their sensitivity to tamoxifen and palbociclib (Fig. 5.D). When analyzing PI3K/AKT/mTOR activation, we observed that alterations in the PIK3CA gene did not always correlate with greater AKT or S6 phosphorylation (Fig. 5.D, red box). Furthermore, pS6 did not always correlate with pAKT levels. Regarding the cell cycle, we did not find an association between palbociclib resistance and changes in Rb expression. However, tamoxifen-resistant PDCs exhibited an upward trend in cyclin E2 expression (Fig. 5.D, blue box), similar to that observed in the tamoxifen- and double-resistant cells (Fig. 1.F).
Further assessment of the PDCs response to PI3K and mTOR inhibition revealed that only two of them, PDC474 and PDC313, were resistant to alpelisib, while all of them were sensitive to everolimus (Fig. 5.E). Interestingly, both alpelisib-resistant PDCs lacked PIK3CA mutations, although PDC313 showed alterations in AKT1, AKT3 and MTOR. Among the seven alpelisib-sensitive PDCs, four (PDC479, PDC446, PDC343, and PDC447) presented either mutations or amplifications in PIK3CA, whereas two (PDC39 and PDC476) exhibited other pathway-associated alterations. Moreover, all PDCs were sensitive to everolimus, regardless of the presence of alterations in PIK3CA or downstream genes (Fig. 5.F).
Finally, we evaluated the PDCs response to combined inhibition of PI3K/AKT/mTOR and CDK4/6. Additionally, since these inhibitors are typically used in the clinic in combination with endocrine therapy, we tested the effect of incorporating the ER degrader fulvestrant. We analyzed three PDCs with different genomic alterations and responses to tamoxifen and palbociclib ex vivo. Treatment with PI3K or mTOR inhibitors as monotherapy significantly reduced cell proliferation, although the incorporation of palbociclib and fulvestrant did not improve the response (Fig. 5.G). When we analyzed the effect of these inhibitors on PI3K/AKT/mTOR activation and cell cycle proteins, we found that palbociclib did not affect AKT/S6 phosphorylation but increased cyclin D1 levels in all models. Alpelisib decreased pRb, pAKT, and pS6 levels only in PDC446, which harbors PIK3CA mutations. In contrast, everolimus alone or in combination with palbociclib successfully reduced pS6, pRb, cyclin D1, cyclin E2, and cyclin B1 levels in all the models, regardless of their sensitivity to tamoxifen/palbociclib and the presence of genomic alterations in the PI3K/AKT/mTOR pathway. Finally, the addition of fulvestrant resulted in more effective inhibition of cyclin D1 expression in all cases (Fig. 5.G).