Research ArticleCancer therapy

Skp2-dependent reactivation of AKT drives resistance to PI3K inhibitors

See allHide authors and affiliations

Sci. Signal.  13 Mar 2018:
Vol. 11, Issue 521, eaao3810
DOI: 10.1126/scisignal.aao3810

Skp’ing down a path of resistance

Predicting how a tumor can adapt to a drug may help identify the patients most likely to respond as well as strategies to avoid resistance. Activation of the kinase PI3K drives the growth of many cancers, but resistance to targeted PI3K pathway inhibitors is common. Clement et al. analyzed triple-negative breast cancer cell lines and found that high abundance of the ubiquitin ligase Skp2 is a predictive biomarker of PI3K inhibitor resistance. Using cultured cells and in tumors in mice, they found that Skp2 ubiquitylated the PI3K pathway kinase AKT, which promoted its activation independently of PI3K, such that PI3K inhibitors were ineffective at suppressing growth. Thus, a Skp2-targeted combination therapy may prevent PI3K inhibitor resistance in some patients with this aggressive type of breast cancer.

Abstract

The PI3K-AKT kinase signaling pathway is frequently deregulated in human cancers, particularly breast cancer, where amplification and somatic mutations of PIK3CA occur with high frequency in patients. Numerous small-molecule inhibitors targeting both PI3K and AKT are under clinical evaluation, but dose-limiting toxicities and the emergence of resistance limit therapeutic efficacy. Various resistance mechanisms to PI3K inhibitors have been identified, including de novo mutations, feedback activation of AKT, or cross-talk pathways. We found a previously unknown resistance mechanism to PI3K pathway inhibition that results in AKT rebound activation. In a subset of triple-negative breast cancer cell lines, treatment with a PI3K inhibitor or depletion of PIK3CA expression ultimately promoted AKT reactivation in a manner dependent on the E3 ubiquitin ligase Skp2, the kinases IGF-1R (insulin-like growth factor 1 receptor) and PDK-1 (phosphoinositide-dependent kinase-1), and the cell growth and metabolism-regulating complex mTORC2 (mechanistic target of rapamycin complex 2), but was independent of PI3K activity or PIP3 production. Resistance to PI3K inhibitors correlated with the increased abundance of Skp2, ubiquitylation of AKT, cell proliferation in culture, and xenograft tumor growth in mice. These findings reveal a ubiquitin signaling feedback mechanism by which PI3K inhibitor resistance may emerge in aggressive breast cancer cells.

INTRODUCTION

The phosphoinositide 3-kinase (PI3K) signaling pathway is frequently deregulated in virtually all human solid tumors including breast cancer as well as hematological malignancies (1). Amplifications, somatic mutations, and other genetic lesions in genes encoding proteins in PI3K signal relay play critical roles in breast cancer by regulating phenotypes such as cell proliferation, survival, metastasis, and metabolic reprogramming (2). The three class IA PI3K catalytic subunits p110α, p110β, and p110δ are encoded by the PIK3CA, PIK3CB, and PIK3CD genes, respectively, and all are activated downstream of receptor tyrosine kinases (RTKs), although PIK3CB can also be activated by G protein (heterotrimeric guanine nucleotide–binding protein)–coupled receptor signaling (3). Activated PI3K phosphorylates phosphatidylinositol 4,5-bisphosphate [PI(4,5)P2] to generate the second messenger phosphatidylinositol 3,4,5-trisphosphate (PIP3). Consequently, synthesis of PIP3 leads to activation of several downstream effector proteins, including the serine and threonine protein kinase AKT, also known as protein kinase B (PKB) (2).

AKT is activated by interaction of the pleckstrin homology (PH) domain with either phosphatidylinositol 3,4-bisphosphate [PI(3,4)P2] or PIP3 (47). Maximal AKT activation is achieved by phosphorylation at both Thr308 and Ser473 mediated by the phosphoinositide-dependent kinase-1 (PDK-1) and mammalian or mechanistic target of rapamycin complex 2 (mTORC2), respectively (810). Signal termination of PI3K and AKT signaling, on the other hand, is mediated primarily by the tumor suppressors phosphatase and tensin homolog (PTEN), which dephosphorylates PIP3 back to PI(4,5)P2 (11, 12), and inositol polyphosphate-4-phosphatase types I and II (INPP4A and INPP4B) that dephosphorylate PI(3,4)P2, resulting in PI3P (13). Moreover, protein phosphatase 2A (PP2A) and PH domain leucine-rich repeat-containing protein phosphatase (PHLPP) dephosphorylate AKT at Thr308 and Ser473, respectively, and thereby also participate in signal termination of AKT signaling (14, 15). Once fully activated, AKT mediates downstream signal relay by phosphorylating a myriad of substrates whose phosphorylation is causally implicated in multiple phenotypes associated with malignancies (2).

Amplification and gain-of-function somatic mutations of the PIK3CA gene occur in many human carcinomas but are most prevalent in breast tumors, particularly in estrogen receptor (ER)–positive breast cancer patients, where about 40% of cases harbor one of the two most frequent activating PIK3CA hotspot mutations, H1047R and E545K (16, 17). Loss-of-function mutations, deletions, and loss of heterozygosity (LOH) in the PTEN gene are also frequently observed in cancers and result in PIP3 accumulation and hyperactivation of AKT signaling (18). Additional genetic lesions in the PI3K and AKT pathway include mutations of PIK3R1 that lead to misregulation of p110α, LOH of INPP4B, activating mutations or copy number gain of the AKT1, AKT2, and AKT3 genes, amplification of PDPK1, as well as a plethora of genetic alterations of upstream RTKs such as HER2 (2).

The frequency of genetic lesions in the PI3K and AKT pathway has made it an attractive target for the development of small-molecule inhibitors for therapeutic benefit (19). Pan-specific, isoform-specific (PIK3CA, PIK3CB, and PIK3CD), and dual PI3K inhibitors, mTOR inhibitors, and catalytic and allosteric AKT inhibitors are being evaluated in preclinical as well as phase 1 and 2 clinical trials; however, aside from the PIK3CD inhibitor idelalisib, to date, they have shown only relatively modest responses in patients (19). The development of drug resistance, including incomplete inhibition of PI3K, reactivation of PI3K and AKT, and activation of compensatory prosurvival pathways, is often cited as a major obstacle in achieving complete clinical responses (20, 21). For example, in HER2-amplified tumors, AKT inhibition induces the expression of HER3 such that the combination of HER-targeted kinase inhibitors with AKT inhibitors improves antitumor efficacy (22). An adaptive resistance mechanism to PI3K inhibition in PIK3CA-mutant breast cancer cells, increased phosphorylation of the tumor suppressor retinoblastoma (RB) has been observed and is reversed using a combination of PI3K and cyclin-dependent kinase 4 and 6 (CDK4/6) inhibitors (23). In addition, prolonged treatment with the p110α inhibitor BYL-719 in patients harboring activating PIK3CA mutations leads to loss of PTEN, providing an alternative mechanism of PI3K activation (24). Additional mechanisms of PI3K inhibitor resistance have previously been uncovered, including enhanced ER function mediated by the epigenetic regulator lysine methyltransferase 2D (KMT2D) (25, 26). A systematic analysis of genes that promote PI3K inhibitor resistance also identified proto-oncogene proviral integration site for Moloney murine leukemia virus (PIM) kinase overexpression as an AKT-independent mechanism of resistance (27). Other mechanisms of PI3K inhibitor resistance including signaling through ribosomal S6 kinases 3 and 4 (RSK3/4) and AXL have also been uncovered (28, 29).

In the context of additional potential mechanisms of PI3K inhibitor resistance, noncanonical AKT reactivation mechanisms have been described. Nonproteolytic ubiquitination by the S-phase kinase-associated protein 2 (Skp2) promotes the activation of AKT (30), and Skp2 overexpression correlates with AKT activation in cancer and is associated with poor prognosis (3133). Here, we further investigated this pathway of AKT activation in the context of PI3K inhibitor–treated triple-negative breast cancer (TNBC) cells.

RESULTS

PI3K inhibition induces reactivation of AKT signaling

Numerous studies have shown that AKT signaling is reactivated in various cancer cell lines after treatment with PI3K inhibitors. To examine the effect of PI3K inhibitors on AKT signaling, MDA-MB-468 breast cancer cells were treated with two different concentrations of BKM-120 (a pan–class I PI3K inhibitor, trade name buparlisib) and the AKT-specific inhibitor MK-2206. As expected, AKT phosphorylation at Ser473 and Thr308 was blocked after a 3-hour treatment with both PI3K and AKT inhibitors (Fig. 1A). However, when cells were treated for 48 hours with BKM-120, inhibition of AKT phosphorylation was no longer observed, and reactivation of AKT signaling was evidenced by enhanced AKT phosphorylation at Ser473 and Thr308 as well as phosphorylation of the AKT substrates PRAS40 and glycogen synthase kinase 3β (GSK3β).

Fig. 1 PI3K inhibition and depletion lead to AKT reactivation.

(A) MDA-MB-468 cells were treated with BKM-120 (1 or 5 μM) and MK-2206 (10 μM) for 3 or 48 hours, alone or with 1 μM of BKM-120, BYL-719, or MK-2206 for an additional 3 hours, as indicated. Cells were then harvested, and lysates were immunoblotted for the indicated total and phosphorylated proteins. (B) MDA-MB-468 cells were infected with shRNA lentivirus targeting PIK3CA or control vector (pLKO). After selection, cells were harvested, and lysates were immunoblotted for the indicated total and phosphorylated proteins. (C) MDA-MB-231 cells were infected with shRNA lentivirus targeting PIK3CA or control pLKO. Cells were serum-starved overnight (−) or stimulated with IGF-1 for 20 min (+) and harvested, and lysates were immunoblotted with the indicated antibodies. (D) MDA-MB-231 cells were infected with shRNA lentivirus targeting PIK3CA, PIK3CB, PIK3CD, PIK3CG, or control pLKO. Cells were serum-starved overnight (−) or stimulated with IGF-1 for 20 min (+) and harvested, and lysates were immunoblotted with the indicated antibodies. Blots are representative of at least three independent experiments. See also fig. S1 (A to C) for additional experiments pertaining to BKM-120 time course, depletion of PIK3CA with multiple shRNA hairpins, and analysis of pAKT1 pSer473 and pSer474 with PIK3CA depletion.

AKT reactivation in response to PI3K inhibition was observed in MDA-MB-468 cells as long as 120 hours after treatment with BKM-120 (fig. S1A). AKT reactivation was not affected by the addition of fresh BKM-120 or BYL-719 (a selective p110α inhibitor, also known as alpelisib), for an additional 3 or 48 hours after the initial treatment (Fig. 1A). Re-addition of PI3K inhibitors did not reduce AKT phosphorylation at Ser473 and Thr308, nor did the treatment affect the phosphorylation of PRAS40 and GSK3β, suggesting that AKT reactivation was independent of PI3K activity. However, AKT phosphorylation was detectably eliminated by the addition of MK-2206 alone or in addition to BKM-120 (Fig. 1A).

Depletion of PIK3CA with two distinct short hairpin RNA (shRNA) hairpins in MDA-MB-468 cells also induced AKT reactivation evidenced by increased AKT phosphorylation at Ser473, Thr308, and phosphorylation of PRAS40 and GSK3β (Fig. 1B). Similarly, AKT reactivation was induced by PIK3CA depletion in MDA-MB-231 breast cancer cells in response to insulin-like growth factor 1 (IGF-1) stimulation (Fig. 1C). Moreover, AKT reactivation was also observed with four additional and distinct shRNA hairpin sequences, indicating that AKT reactivation is unlikely resulting from shRNA off-target effects (fig. S1B). Furthermore, both AKT1 and AKT2 showed reactivation in PIK3CA-depleted cells (fig. S1C), and AKT reactivation was not due to compensatory expression of PIK3CB or PIK3CD (Fig. 1C). Finally, depletion of PIK3CA, PIK3CB, PIK3CD, and PIK3CG in MDA-MB-231 cells induced AKT reactivation in response to IGF-1 (Fig. 1D), indicating that AKT reactivation is a common event upon PI3K pathway inhibition.

AKT reactivation upon PI3K pathway inhibition is restricted to a subset of TNBC cell lines

We next tested whether AKT reactivation upon PI3K pathway inhibition is a common event in breast cancer cells. Surprisingly to us, whereas AKT reactivation was observed in SUM149PT, HCC1937, or MDA-MB-468 cells, in the ER-positive breast cancer cell line MCF7, PIK3CA depletion showed a marked inhibition of AKT phosphorylation and activity (Fig. 2A). We then tested a larger panel of breast cancer cell lines using the same PIK3CA shRNA approach and evaluated AKT Ser473 and Thr308 phosphorylation and activity (pPRAS40 and pGSK3β). The results (Fig. 2B) revealed that of the total 17 breast cancer cell lines tested here, 6 showed AKT reactivation and other cell lines in this panel showed inhibition of AKT. All six cell lines that showed AKT reactivation upon PIK3CA depletion are TNBC, although this was not a defining characteristic of the reactivation phenotype, because several TNBC lines showed AKT inhibition and not reactivation upon PI3K pathway inhibition (Fig. 2B).

Fig. 2 PI3K depletion does not lead to AKT reactivation in all breast cancer cell lines.

(A) SUM149PT, HCC1937, MDA-MB-468, and MCF7 cells were infected with shRNA lentivirus targeting PIK3CA, control pLKO, or shGFP as additional negative control. Cells were serum-starved overnight (−) or stimulated with IGF-1 for 20 min (+) and harvested, and lysates were immunoblotted with the indicated antibodies. Blots are representative of at least three independent experiments. (B) Summary table showing the different breast cancer cell lines tested in this study for AKT reactivation. TNBC classification according to Lehmann et al. (55): BL1, basal-like 1; BL2, basal-like 2; M, mesenchymal; MSL, mesenchymal stem–like; LAR, luminal androgen receptor. Note that BT20 and MDA-MB-435 are only classified as TNBC.

AKT reactivation is dependent on mTORC2 and PDK-1

Given that AKT reactivation was observed upon IGF-1 stimulation (Fig. 1C), we next treated PIK3CA-depleted MDA-MB-231 cells with the IGF-1R inhibitor NVP-AEW541. AKT reactivation in PIK3CA-depleted cells was completely blocked upon IGF-1R inhibition (Fig. 3A). Similarly, AKT reactivation in PIK3CA-depleted cells was attenuated in Rictor-depleted cells, indicative of an mTORC2 requirement for Ser473 phosphorylation as part of the AKT reactivation mechanism (Fig. 3B). By contrast, depletion of the mTORC1 component Raptor (RPTOR) actually further promoted AKT reactivation compared to PIK3CA depletion alone (Fig. 3B), consistent with previous studies demonstrating that Raptor depletion induces AKT activation (9). Therefore, AKT reactivation upon PI3K pathway inhibition is mTORC2-dependent.

Fig. 3 AKT reactivation is dependent on upstream regulators.

(A) MDA-MB-231 cells were infected with shRNA lentivirus targeting PIK3CA or control pLKO. Cells were serum-starved overnight (−) and treated with the IGF-1R inhibitor NVP-AEW541 for 15 min before stimulation with IGF-1 for 20 min (+). Cells were harvested, and lysates were immunoblotted with the indicated antibodies. (B) MDA-MB-231 cells were infected with shRNA lentivirus targeting PIK3CA or control pLKO alone or with the indicated combinations of shRNA specific for Raptor or Rictor, respectively. Cells were serum-starved overnight (−) or stimulated with IGF-1 for 20 min (+). Cells were harvested, and lysates were immunoblotted with the indicated antibodies. (C) MDA-MB-231 cells were infected with shRNA lentivirus targeting PIK3CA or control pLKO. Cells were serum-starved overnight (−) and treated with the PDK-1 inhibitor GSK2334470 for 15 min before stimulation with IGF-1 for 20 min (+). Cells were harvested, and lysates were immunoblotted with the indicated antibodies. Data are representative of at least three independent experiments. (D) MDA-MB-231 cells were infected with shRNA lentivirus targeting PIK3CA or control pLKO in the absence or presence of complementary DNA (cDNA) directing expression of p110α. Cells were serum-starved overnight (−) or stimulated with IGF-1 for 20 min (+). Cells were harvested, and lysates were immunoblotted with the indicated antibodies. Data are representative of at least three independent experiments. WT, wild type.

AKT reactivation in PIK3CA-depleted cells was also blocked by pretreatment with the PDK-1 inhibitor GSK2334470 (Fig. 3C) (34). Finally, wild-type p110α was expressed in control cells, and as expected, this promoted AKT phosphorylation at Ser473 and Thr308 and PRAS40 and GSK3β phosphorylation (Fig. 3D). Notably, enhanced AKT activity was also observed in PIK3CA-depleted cells. Together, these data show that AKT reactivation in PI3K pathway–inhibited cells is dependent on the canonical IGF-1R/PDK-1/mTORC2 pathway.

AKT reactivation is PI3K-independent

We next determined whether AKT reactivation in PI3K pathway–inhibited cells required any form of PI3K activity. Control and PIK3CA-depleted cells were treated with BKM-120, BYL-719, A66, TGX-221, or IC-87114, inhibitors that are specific to p110α (PIK3CA), p110β (PIK3CB), and p110δ (PIK3CD), respectively. Control cells showed AKT phosphorylation and activity in response to IGF-1, and this was robustly attenuated by all PI3K inhibitors. However, AKT reactivation in PIK3CA-depleted cells was unaffected by either pan or specific PI3K inhibitors, indicating that AKT reactivation is largely independent of PI3K activity.

To provide further evidence for PI3K-independent AKT reactivation, PTEN was expressed in control and PIK3CA-depleted cells. Whereas expression of PTEN robustly attenuated AKT and PRAS40 phosphorylation in control cells, as expected, there was no effect on AKT reactivation in PIK3CA-depleted cells (Fig. 4B). We also measured the production of PIP3 and PI(3,4)P2 in control and PIK3CA-depleted cells in response to IGF-1. Depletion of PIK3CA in MDA-MB-231 cells promoted AKT and PRAS40 phosphorylation (Fig. 4C). In control cells, we detected a significant increase in PIP3 also in response to IGF-1. However, depletion of PIK3CA profoundly reduced PIP3 production (Fig. 4C), yet under these conditions, robust AKT activity is observed. Similar responses were detected with PI(3,4)P2 (Fig. 4D). Thus, class I PI3K and PIP3/PI(3,4)P2 do not prominently contribute to AKT reactivation in MDA-MB-231 breast cancer cells in which the PI3K pathway is inhibited.

Fig. 4 AKT reactivation is PI3K-independent.

(A) MDA-MB-231 cells were infected with shRNA lentivirus targeting PIK3CA or control pLKO. Cells were serum-starved overnight (−) and treated with indicated inhibitors for 15 min before stimulation with IGF-1 for 20 min (+). Cells were harvested, and lysates were immunoblotted with the indicated antibodies. DMSO, dimethyl sulfoxide. (B) MDA-MB-231 cells were infected with shRNA lentivirus targeting PIK3CA or control pLKO and then transfected with PTEN WT or control pcDNA vector. Cells were serum-starved overnight (−) or stimulated with IGF-1 for 20 min (+). Cells were harvested, and lysates were immunoblotted with the indicated antibodies. (C and D) MDA-MB-231 cells were infected with shRNA lentivirus targeting PIK3CA or control pLKO. Cells were serum-starved overnight (−) or stimulated with IGF-1 for 20 min (+). Phospholipids were isolated, PIP3 (C) or PI(3,4)P2 (D) was quantified by enzyme-linked immunosorbent assay (ELISA), or cell lysates were immunoblotted with the indicated antibodies. Data are means ± SEM from three independent experiments, each carried out in triplicate. **P < 0.01 by a two-sided Student’s t test.

AKT reactivation is Skp2-dependent

To identify the molecular mechanism by which AKT is reactivated in a PI3K-independent manner, we investigated the role of the ubiquitin E3 ligase Skp2, which has been shown to contribute to AKT activation through ubiquitination (30, 35). Strikingly, we detected increased abundance of total Skp2 protein in breast cancer cell lines depleted with PIK3CA and that show AKT reactivation (Fig. 5A). This was also evident at the mRNA level with three distinct PIK3CA shRNA hairpins (Fig. 5B).

Fig. 5 AKT reactivation is Skp2-dependent.

(A) The indicated cell lines were infected with shRNA lentivirus targeting PIK3CA or control pLKO. Cells were serum-starved overnight (−) or stimulated with IGF-1 for 20 min (+) and harvested, and lysates were immunoblotted with the indicated antibodies. (B) MDA-MB-231 cells infected with PIK3CA and/or SKP2 shRNA lentiviral vectors or control pLKO. SKP2 mRNA expression was measured by real-time reverse transcription polymerase chain reaction (RT-PCR), and corresponding lysates were immunoblotted with indicated antibodies. Data are means ± SEM from three independent experiments. **P < 0.01 by a two-sided Student’s t test. (C) Immunoblotting in lysates from MDA-MB-468 cells treated with BKM-120 (1 μM) for up to 144 hours and then treated again with BKM-120 (1 μM) for an additional 30 min. (D) Immunoblotting in lysates from MDA-MB-231 cells infected with PIK3CA, SKP2, or PIK3CA in combination with SKP2 shRNA lentiviral vectors or control vector (pLKO) and serum-starved overnight (−) or stimulated with IGF-1 for 20 min (+). (E) Immunoblotting in lysates from MDA-MB-468 cells infected with SKP2 shRNA lentiviral vector or control pLKO and then treated with BKM-120 (1 μM) for 3 or 48 hours. (F) Immunoblotting in lysates from MDA-MB-231 cells, parental and resistant to 1 μM BKM-120 (BKM-120), that were infected with SKP2 shRNA lentiviral vector or control pLKO vector and serum-starved overnight (−) or stimulated with IGF-1 for 20 min (+). (G) MDA-MB-231 cells, parental and BKM-120–resistant, were infected with SKP2 shRNA lentiviral vector or control pLKO. After transfection with pcDNA3 or His-ubiquitin (His-Ub), His-ubiquitin complexes were isolated, followed by immunoblotting. Blots are representative of at least three independent experiments. See also fig. S1D for additional time course with BKM-120 in cells expressing cDNA for SKP2.

Similar to PIK3CA depletion, PI3K inhibition (by BKM-120) also promoted Skp2 expression (Fig. 5C). In this experiment, to confirm that AKT reactivation was PI3K-independent, we treated cells with BKM-120 for an additional 30 min after the initial drug administration, and this had no substantially apparent effect on AKT phosphorylation. AKT reactivation required Skp2 in MDA-MB-231 cells given that SKP2 depletion with each of four distinct shRNA hairpins in the context of PIK3CA depletion attenuated AKT reactivation when compared to PIK3CA depletion alone (Fig. 5D). Depletion of SKP2 in control-treated MDA-MB-468 cells attenuated AKT phosphorylation, as previously documented (30). AKT reactivation in cells treated with BKM-120 for 48 hours was also attenuated with SKP2 depletion (Fig. 5E). We also generated MDA-MB-231 cells resistant to 1 μM BKM-120 using a dose escalation protocol. In agreement with the above data, BKM-120–resistant cells show AKT reactivation and Skp2 expression when compared to parental cells (Fig. 5F) and SKP2 depletion attenuated AKT reactivation in IGF-1–stimulated cells. Furthermore, expression of Skp2 promoted AKT phosphorylation and activation, and this was further promoted upon chronic treatment with BKM-120 (fig. S1D). Finally, we used an established cell-based ubiquitination assay to evaluate AKT ubiquitination in the reactivation mechanism. Notably, AKT ubiquitination was induced in parental cells in response to IGF-1 stimulation (Fig. 5G). As expected, depletion of SKP2 suppressed AKT ubiquitination as well as phosphorylation and activity. Robust ubiquitination of AKT in response to IGF-1 was also detected in MDA-MB-231 cells that were resistant to BKM-120, and this was prominently attenuated by SKP2 depletion (Fig. 5G). Together, these data demonstrate that AKT reactivation upon PI3K pathway inhibition depends, at least in part, on Skp2 activity.

AKT reactivation promotes cell proliferation

We next explored the functional significance of AKT reactivation by Skp2. The spheroid growth of MDA-MB-231 parental cells in three-dimensional (3D) Matrigel was significantly reduced by treatment with BKM-120 (Fig. 6, A and B). However, spheroid growth in BKM-120–resistant MDA-MB-231 cells was not significantly affected by BKM-120 (Fig. 6B), although spheroid growth was significantly inhibited upon shRNA-mediated depletion of SKP2 or by treatment with the AKT inhibitor MK-2206. Similar results were obtained with anchorage independence of growth assays; MDA-MB-231 parental cells formed colonies in soft agar, but treatment with BKM-120 significantly inhibited the number of colonies growing in this anchorage-independent manner (Fig. 6, C and D). Moreover, additional treatment with MK-2206 or depletion of SKP2 did not rescue the defect in anchorage-independent growth. On the other hand, in BKM-120–resistant cells, which also formed a substantial number of colonies, treatment with BKM-120 did not reduce colony formation, but additional treatment with MK-2206 or depletion of SKP2 in BKM-120–resistant cells did abrogate colony formation (Fig. 6, C and D). These results show that AKT reactivation in the context of PI3K inhibitor resistance is primarily driven by Skp2.

Fig. 6 Skp2 promotes 3D culture and anchorage independence of growth in BKM-120–resistant cells.

(A) Representative images of MDA-MB-231 spheroids from parental or BKM-120–resistant cells that were infected with SKP2 shRNA lentiviral vector or control pLKO and grown for 14 days in a 3D Matrigel/growth medium mixture and treated with BKM-120 and/or MK-2206 (each 1 μM). Scale bar, 500 μm (magnification, ×4). (B) Quantitation of spheroid sizes from (A) was performed using ImageJ software. Data are means ± SEM. ***P < 0.001 by a two-sided Student’s t test. (C) Representative images of MDA-MB-231 colony growth from parental or BKM-120–resistant cells that were infected with SKP2 shRNA lentiviral vector or control pLKO and grown in an agar/growth medium mixture for 28 days in the presence of DMSO, BKM-120, and/or MK-2206 (each 1 μM). (D) Number of colonies in soft agar obtained in (C) was quantitated using MATLAB software. Data are means ± SEM. **P < 0.01 by a two-sided Student’s t test. All data and blots are representative of at least three independent experiments.

Depletion of SKP2 inhibits the progression of BKM-120–resistant breast cancer xenografts

Because SKP2 depletion and inhibition of AKT significantly retarded both anchorage-dependent and anchorage-independent cell growth, we performed xenograft experiments using parental and BKM-120–resistant cells with SKP2 depletion. We first confirmed that BKM-120–resistant cells display AKT reactivation that is reversed with SKP2 depletion (Fig. 7A). Cells were then subcutaneously injected into mice, and tumorigenicity was assessed in xenograft growth. Tumors derived from xenografted BKM-120–resistant cells grew larger than those derived from parental MDA-MB-231 cells (Fig. 7, B to E). Moreover, consistent with in vitro assays (Fig. 6), depletion of SKP2 resulted in significantly retarded tumorigenesis in both parental and BKM-120–resistant cells. Finally, BKM-120–resistant cells displayed more pronounced effects upon SKP2 depletion with respect to delay in tumor growth (Fig. 7, D and E). Together, these results indicate that Skp2 is a major mechanism for tumor growth of BKM-120–resistant tumors.

Fig. 7 Skp2 depletion attenuates tumor growth in vivo.

(A) Immunoblotting of lysates from MDA-MB-231, parental and BKM-120–resistant, cells that were infected with SKP2 shRNA lentiviral vector or control pLKO. Cells were serum-starved overnight (−) or stimulated with IGF-1 for 20 min (+). (B and C) Tumor weight (B) and volume (C) of xenografts formed from cells described in (A) inoculated subcutaneously into nude mice (n = 10). Cells (1 × 106) were injected in each flank. Tumor measurements were taken 22 days after injection. Data are means ± SEM from 10 mice for each condition. **P < 0.01, ***P < 0.001 by a two-sided Student’s t test. (C) Tumor xenograft volumes taken 22 days after injection. (D) Growth curves of tumors described in (B). Tumor volume was measured starting day 1 after injection, then on days 5 and 8, and every 2 days up to day 22. (E) Representative pictures of tumors described in (B), surgically removed at day 22.

DISCUSSION

The aberrant activation of the PI3K/AKT/mTOR signaling pathway is frequently observed in virtually all cancer types, including breast cancer. Somatic mutations and amplification of PIK3CA, the gene that encodes the p110α catalytic subunit of class I PI3K, occur with high frequency in both hormone receptor–positive breast cancer and TNBC (1, 36). Although a large number of potent and highly selective PI3K inhibitors have been developed and are under clinical evaluation, the emergence of resistance mechanisms that result in pathway bypass or downstream effector reactivation limits therapeutic efficacy in patients (19, 20). Therefore, elucidating the molecular mechanisms underlying acquired resistance to PI3K-targeted therapies may afford a critical diagnostic assessment of which subtype of breast cancers will develop drug resistance and inform the development of combination therapies to limit resistance. Here, we identified a previously unidentified mechanism of resistance to PI3K inhibition in breast cancer cells. We show that inhibition of PI3K (with either shRNA-mediated depletion of p110α, or administration of PI3K inhibitors) promotes AKT reactivation by the E3 ubiquitin ligase Skp2 and the canonical upstream AKT kinases PDK-1 and mTORC2, resulting in enhanced growth of breast cancer cells both in vitro and in mouse xenografts. We further identified that AKT reactivation was independent of PI3K and PIP3. Skp2 is implicated in the development and progression of various cancers, including breast cancer (37, 38). We propose that this Skp2-AKT mechanism may represent a central node for a subset of breast cancers that develop resistance to PI3K inhibitors.

The precise molecular mechanism(s) leading to increased Skp2 expression in PI3K inhibitor–treated (and resistant) cells is not yet clear. A number of pathways are likely to contribute. Activation of Notch1, a transcriptional activator of SKP2 (39), is frequently observed in PI3K inhibitor–resistant cells (40); potentially, mutations that enhance Notch signaling could be a source of the resistance and, therefore, a genetic biomarker for patients likely to relapse after BKM-120 treatment. Alternatively, inactivation of FOXO3-mediated suppression of SKP2 may also contribute to the increase in Skp2 in PI3K inhibitor–treated cells (41). Notably, Skp2 expression and AKT kinase activation positively correlate in breast cancer cells (42, 43). AKT itself is a positive regulator of Skp2 function through transcriptional, translational, and posttranslational mechanisms. Specifically, AKT1 induces SKP2 transcription in an E2F1-dependent manner (44) and antagonizes the transcriptional repressor function of FOXO3 that binds the SKP2 promoter (45); AKT1 and AKT2 reportedly enhance the translation of SKP2 mRNA through activation of the translation initiation protein eIF4E (46); and AKT1 promotes Skp2 protein stability in part through direct phosphorylation of Skp2 at Ser72 (43, 47). Collectively, these findings indicate that AKT and Skp2 reciprocally enhance their functional activation through a positive feedback loop.

Whether this mechanism of AKT reactivation mediated by Skp2 in the absence of productive PI3K activity occurs in patients treated with inhibitors, such as BKM-120 or BYL-719, remains to be determined. Likewise, it is presently unknown what genetic or epigenetic characteristics define the ability of breast cancer cells to co-opt Skp2-mediated AKT activation in response to PI3K inhibition. This is further compounded by the current lack of any predictive biomarkers that would allow a classification of cells that display AKT rebound activation as part of PI3K inhibitor resistance. Our observation that AKT reactivation upon depletion of any PI3K isoform was surprising, given that, under conditions of depletion of an individual p110 isoform, other isoforms should compensate and retain intact PI3K signaling. That depletion of PIK3CA alone was sufficient to quantitatively eliminate PIP3 and PI(3,4)P2 in IGF-1–stimulated cells suggests that inhibition of individual p110 isoforms cannot be efficiently compensated for, at least under these conditions. It is also worth speculating that other PI3K effectors (such as SGK isoforms) could also display rebound activation in the context of resistance to PI3K inhibitors. SGK3 may be particularly relevant, given that SGK3, and not AKT, mediates the PI3K oncogenic signal in certain PIK3CA-mutant cells (48), and in a manner that also requires the PI(3,4)P2 phosphatase INPP4B (49).

Further work with mouse models of PIK3CA-driven breast cancer treated with combination regimens of PI3K and Skp2 inhibitors will be required to support the model proposed by these data. However, this concept is supported by studies that have revealed therapeutic efficacy of Skp2 inhibitors in mouse models (35, 50, 51). Nonetheless, the findings uncover a mechanism of PI3K inhibitor resistance in breast cancer cells in which the Skp2 ubiquitin ligase enables PI3K-independent AKT signaling.

MATERIALS AND METHODS

Cell culture and transfection

BT20, Hs578T, MCF7, MDA-MB-231, MDA-MB-435, and MDA-MB-468 cells were obtained from the American Type Culture Collection (ATCC) and cultured in Dulbecco’s modified Eagle’s medium (DMEM; Cellgro) supplemented with 10% fetal bovine serum (FBS; Gibco). BT549, HCC1143, HCC1806, HCC1937, HCC38, HCC70, and ZR-75.1 cells obtained from the ATCC were grown in RPMI 1640 medium (Cellgro) supplemented with 10% FBS. MDA-MB-453 and SKBR3 cells obtained from the ATCC were grown in McCoy’s 5A medium (Cellgro) supplemented with 10% FBS. SUM149PT and SUM159PT were grown in Ham’s F12 medium (Cellgro) with 5% FBS and insulin (5 μg/ml; Sigma-Aldrich) and hydrocortisone (500 ng/ml; Sigma-Aldrich). Cells were passaged for no more than 6 months and routinely assayed for mycoplasma contamination. Transient transfections were performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s protocol. To generate MDA-MB-231–resistant cells, the parental cells were treated with increasing concentrations of BKM-120 until a target concentration of 1 μM BKM-120 was achieved.

Growth factors and inhibitors

Cells were stimulated for 20 min with recombinant human IGF-1 (R&D Systems) at a final concentration of 100 ng/ml. PI3K inhibitors BKM-120 (Selleck Chemicals), BYL-719 (Selleck Chemicals), A66 (Selleck Chemicals), TGX-221, and IC-87114 [gifts from P. Shepherd, University of Auckland; (52)] were added to cells 15 min before IGF-1 stimulation at a final concentration of 1 μM. The AKT inhibitor MK-2206 (Active Biochem) was added to cells 15 min before IGF-1 stimulation at a final concentration of 1 μM. NVP-AEW541 (Cayman Chemical) and GSK2334470 (Tocris) were added to cells 15 min before growth factor stimulation at a final concentration of 1 μM.

Antibodies

Antibodies recognizing AKT (4685), phospho-AKT Ser473 (4060), phospho-AKT Thr308 (2965), PRAS40 (2691), phospho-PRAS40 (2997), GSK3β (9315), phospho-GSK3β (9336), p110α (4249), phospho-S6K T389 (9205), PTEN (9188), actin (4970), and p110γ (5405) were purchased from Cell Signaling Technology. Antibodies recognizing p110β (sc-602), p110δ (sc-7176), and Raptor (sc-81537) were from Santa Cruz Biotechnology. Antibody recognizing Rictor (A300-459A) was from Bethyl Laboratories. Antibody recognizing Skp2 (32-3300) was from Invitrogen. Antibody recognizing p85 was generated in house and has been described (53). Horseradish peroxidase (HRP)–conjugated anti-rabbit and anti-mouse immunoglobulin G antibodies were from Millipore.

Plasmids

(His)6-ubiquitin plasmid was a gift from H.-K. Lin (University of Texas MD Anderson Cancer Center, Houston). PTEN–wild-type plasmid was obtained from Addgene.

RNA interference

For shRNA-mediated silencing of PIK3CA, PIK3CB, PIK3CD, and PIK3CG, a set of oligonucleotides composed of a target shRNA sequence and its complement was synthesized. The hairpin sequences were generated as follows: PIK3CAshRNA#2, 5′-CCGGGCACAATCCATGAACAGCATTCTCGAGAATGCTGTTCATGGATTGTGCTTTTTTG-3′ (sense) and 5′-AATTCAAAAACACAATCCATGAACAGCATTCTCGAGAATGCTGTTCATGGATTGTG-3′ (antisense); PIK3CAshRNA#3, 5′-CCGGGCATTAGAATTTACAGCAAGACTCGAGTCTTGCTGTAAATTCTAATGCTTTTTTG-3′ (sense) and 5′-AATTCAAAAAGCATTAGAATTTACAGCAAGACTCGAGTCTTGCTGTAAATTCTAATGC-3′ (antisense); PIK3CBshRNA#1, 5′-CCGGGCAACAGCTTTGCATGTTAAACTCGAGTTTAACATGCAAAGCTGTTGCTTTTTG-3′ (sense) and 5′-AATTCAAAAAGCAACAGCTTTGCATGTTAAACTCGAGTTTAACATGCAAAGCTGTTGC-3′ (antisense); PIK3CDshRNA#2, 5′-CCGGGATCTTTCTCTCTGACTATACCTCGAGGTATAGTCAGAGAGAAAGATCTTTTTG-3′ (sense) and 5′-AATTCAAAAAGATCTTTCTCTCTGACTATACCTCGAGGTATAGTCAGAGAGAAAGATC-3′ (antisense); SKP2shRNA#2, 5′-CCGGGAAGGCCTGCGGCTTTCGGATCCTCGAGGATCCGAAAGCCGCAGGCCTTCTTTTTTG-3′ (sense) and 5′-AATTCAAAAAAAGGCCTGCGGCTTTCGGATCCTCGAGGATCCGAAAGCCGCAGGCCTT-3′ (antisense).

The oligonucleotide pairs were each annealed and inserted into pLKO.1. For shRNA-mediated knockdown of SKP2 and PIK3CG, simple hairpin shRNAs in the pLKO.1 lentiviral vector designed by the RNAi Consortium were ordered from Thermo Fisher Scientific: 5′-GCCTAAGCTAAATCGAGAGAA-3′ (SKP2shRNA#4; clone TRCN0000007530), 5′-CCACGATCATTTATGGACCAA-3′ (SKP2shRNA#5; clone TRCN0000007531), 5′-CCATTGTCAATACTCTCGCAA-3′ (SKP2shRNA#6; clone TRCN0000007532), and 5′-TAATTCCCAAGAATGTGCCCG-3′ (PIK3CGshRNA#1; clone TRCN0000033282). To produce lentiviral supernatants, 293T cells were cotransfected with VSVG (vesicular stomatitis virus glycoprotein), psPAX2, and control-pLKO, PIK3CA, PIK3CB, PIK3CD, PIK3CG, or SKP2 shRNA-containing vectors for 48 hours.

Immunoblotting

Cells were washed with ice-cold phosphate-buffered saline (PBS) and lysed in radioimmunoprecipitation assay buffer [150 mM tris-HCl (pH 7.4), 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, 20 mM sodium fluoride, 1 mM sodium pyrophosphate, 1 mM sodium orthovanadate, 50 nM calyculin A, proteinase inhibitor cocktail] for 15 min at 4°C. Cell extracts were centrifuged at 13,000 rpm for 10 min at 4°C, and protein concentration was measured with the Bio-Rad protein assay reagent using a Beckman Coulter DU-800 machine. Lysates were then resolved on 8% acrylamide gels by SDS–polyacrylamide gel electrophoresis and transferred electrophoretically to nitrocellulose membrane (Bio-Rad) at 100 V for 60 min. The blots were blocked in TBST buffer [10 mM tris-HCl (pH 8), 150 mM NaCl, 0.2% Tween 20] containing 5% nonfat dry milk for 30 min and then incubated with the specific primary antibody at 4°C overnight. Membranes were washed three times in TBST and incubated with HRP-conjugated secondary antibody for 1 hour at room temperature. Membranes were washed three times and developed using chemiluminescent HRP substrate (Millipore).

Quantitative RT-PCR analysis

RNA was extracted using the Qiagen RNeasy Mini Kit, and the reverse transcription reaction was carried out with Applied Biosystems TaqMan Reverse Transcriptional Reagent. After mixing the resulting template with Skp2 (Mm00449925_m1) or glyceraldehyde-3-phosphate dehydrogenase (GAPDH; Mm99999915_g1) primers and Applied Biosystems TaqMan Fast Universal PCR Master Mix, the RT-PCR reaction was performed with the Applied Biosystems 7500 Fast Real-Time PCR System.

PIP3 and PI(3,4)P2 assay

PIP3 and PI(3,4)P2 were extracted as described in the manufacturer’s protocol (K-2500s and K-3800, Echelon). Briefly, after IGF-1 stimulation, the medium was removed from the plates and 10 ml of ice-cold 0.5 M trichloroacetic acid (TCA) was immediately added. Cells were scraped and centrifuged at 3000 rpm for 5 min. The pellet was resuspended in 3 ml of 5% TCA/1 mM EDTA, vortexed, and centrifuged at 1500 rpm for 5 min. To extract neutral lipids, 3 ml of MeOH/CHCl3 (2:1) was added to the pellet, vortexed three times over 10 min, and centrifuged at 1500 rpm for 5 min. Acidic lipids were extracted by adding 2.25 ml of MeOH/CHCl3/12 M HCl (80:40:1), vortexed four times over 15 min, and centrifuged at 1500 rpm for 5 min. The supernatant was transferred to a new 15-ml tube, and 0.75 ml of CHCl3 and 1.35 ml of 0.1 M HCl were added to the supernatant, vortexed, and centrifuged at 1500 rpm for 5 min to separate organic and aqueous phases. The organic (lower) phase was collected and dried in vacuum dryer for 1 hour. Once isolated from cells, PIP3 and PI(3,4)P2 were measured using ELISA kits (K-2500s and K-3800, Echelon) according to the manufacturer’s instructions.

3D culture assay

3D cultures were performed as described (54). Chamber slides were coated with growth factor–reduced Matrigel (BD Biosciences) and allowed to solidify for 30 min. Subsequently, 5000 MDA-MB-231 cells were seeded to each chamber slide in assay medium (DMEM supplemented with 2% FBS and 5% Matrigel). Assay medium was replaced 4 days after seeding. ImageJ was used to quantify 2D surface area on day 14.

Soft agar colony formation assay

Soft agar assays were performed by coating six-well plates with 2 ml of 0.8% noble agar/growth medium (10% FBS/DMEM) with indicated inhibitors and allowed to solidify at 20°C. Cells (2 × 104) were plated in 1 ml of top layer 0.4% noble agar/growth medium with indicated inhibitors. Growth medium with inhibitors was replaced every 4 days, and cells were counted and measured 28 days after seeding. Quantitation was performed using MATLAB software (MathWorks).

Ubiquitination assay

Cell-based in vivo ubiquitination assays were performed as described (30). MDA-MB-231 parental and BKM-120–resistant cells were transfected with His-ubiquitin or pcDNA3 as control. Subsequently, cells were infected with Skp2 shRNA lentiviral vector or control pLKO. Cells were serum-starved overnight or stimulated with IGF-1 for 20 min and lysed in denatured buffer (6 M guanidine-HCl, 0.1 M Na2HPO4/NaH2PO4, 10 mM imidazole). The cell lysates were incubated with nickel beads for 3 hours, washed, and subjected to immunoblotting analysis.

Tumor xenografts

Six- to 8-week-old female nude mice were purchased from Taconic and maintained in a pathogen-free environment. All procedures were carried out under the approval of the Institutional Animal Care and Use Committee at the Beth Israel Deaconess Medical Center and comply with the federal guidelines for maintenance and care of laboratory animals. Ten mice for each experimental condition were injected subcutaneously in the flank with 1.0 × 106 MDA-MB-231 cells in PBS with 50% Matrigel. Tumor volume was measured by volume = 0.52 × length × width2.

Statistics and reproducibility

Sample sizes and reproducibility for each figure are denoted in the figure legends. Unless otherwise noted, all immunoblots are representative of at least three biologically independent experiments. Statistical significance between conditions was assessed by two-tailed Student’s t tests. All error bars represent SEM, and significance between conditions is denoted as *P < 0.05, **P < 0.01, and ***P < 0.001.

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/11/521/eaao3810/DC1

Fig. S1. Reactivation of AKT in response to PIK3CA depletion and inhibition.

REFERENCES AND NOTES

Acknowledgments: We thank the members of the Toker and Wei laboratories for productive discussions, P. Shepherd (University of Auckland) and H.-K. Lin (The University of Texas MD Anderson Cancer Center) for providing reagents, and R. Arora (Beth Israel Deaconess Medical Center) for evaluation of statistical analyses. Funding: This study was supported in part by Susan G. Komen for the Cure Postdoctoral Fellowship (KG111139 to E.C.), the NIH (CA177910 and CA200671 to A.T.; GM089763 and GM094777 to W.W.), and the Ludwig Center at Harvard (to A.T.). Author contributions: E.C., A.T., and W.W. designed the study and interpreted the results. E.C. performed all experiments unless otherwise noted. H.I. performed the ubiquitination experiments and mouse xenograft studies. N.T.N. performed a subset of the shRNA and Skp2 expression assays. E.C., A.T., H.I., and W.W. wrote the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.
View Abstract

Navigate This Article