Introduction

The Phosphatidylinositol-3 Kinase (PI3K) pathway is evolutionarily conserved and plays crucial roles in survival, growth, cell cycle, and metabolism (Luo et al. 2003; Manning and Cantley 2007; Nakae et al. 2001; Okkenhaug and Vanhaesebroeck 2003). PI3K phosphorylates phosphatidylinositol 4,5-bisphosphate (PIP2) producing phosphatidylinositol 3,4,5-trisphosphate (PIP3), which binds to and activates targets such as serine threonine kinase AKT (Bellacosa et al. 1998; Manning and Cantley 2007). AKT has over 20 identified substrates including the transcription factors Forkhead Box subfamily O members (FOXO-1, -3, and -4) (Brunet et al. 1999; Luo et al. 2003; Manning and Cantley 2007). Phosphorylation of FOXO factors by AKT leads to their cytoplasmic sequestration/inactivation (Brunet et al. 1999).

The PI3K pathway is almost universally activated in cancer to promote growth and survival, commonly by gain-of-function PIK3CA mutants (encoding PI3K catalytic subunit) or loss-of-function PTEN (Phosphatase and Tensin homolog deleted on chromosome ten) mutations (Li et al. 1997; Saal et al. 2005, 2007, 2008). Aberrant activation of the PI3K Pathway leads to inactivation of FOXO transcription factors (Brunet et al. 1999). However, several studies have shown exceptions to this canonical PI3K Pathway circuitry in certain advanced poor prognosis cancers, human embryonic stem (ES) cells, and naïve T cells (Bigarella et al. 2017; Keniry et al. 2013; Oh et al. 2012; Trinh et al. 2013). In these settings, PI3K expression is high, but a portion of the FOXO proteins is found in the nucleus (Keniry et al. 2013; Liang et al. 2016; Oh et al. 2012; Zhang et al. 2011). FOXO1 and FOXO3 are required for the maintenance of hematopoietic, embryonic, and neural stem cells (Miyamoto et al. 2007; Renault et al. 2009; Tothova and Gilliland 2007; Tothova et al. 2007; Yu et al. 2018; Zhang et al. 2011). In ES cells, FOXO1 directly associated with the promoters and regulated Octamer-binding Transcription factor 4 (OCT4) and Sex determining region Y-box 2 (SOX2), thereby promoting pluripotency and preventing differentiation (Zhang et al. 2011). Given that FOXO factors regulate OCT4 in stem cells, we examined whether these factors had a similar function in certain cancers (Ben-Porath et al. 2008; Ghaffari et al. 2010).

Forkhead Box O (FOXO-1, -3, and -4) transcription factors regulate cellular processes in a context-dependent manner and are partially redundant with each other (Paik et al. 2007; Tothova et al. 2007). FOXO6 is mainly expressed in the brain and regulated by distinct mechanisms (Jacobs et al. 2003; van der Heide et al. 2005). FOXO-1, -3, and -4 are excluded from the nucleus in settings with high PI3K output (via an AKT-mediated mechanism) (Brunet et al. 1999). There are a number of settings in which FOXO factors at least partially bypass AKT regulation, leading to nuclear localization (Keniry et al. 2013; Liang et al. 2016). First, FOXO1 was found mutated in 9% of diffuse large B-cell lymphoma (DLBCL) leading to constitutive nuclear localization; these mutations were associated with poor prognosis (Trinh et al. 2013). Nuclear FOXO factors were also found in basal breast cancer (BBC) cell lines such as BT549 as well as primary samples that harbored active PI3K Pathway output (Hagenbuchner et al. 2016; Keniry et al. 2013; Zhang et al. 2011). However, the function of nuclear FOXO factors in these aggressive cancers with active PI3K pathway output remained elusive.

To gain insight into novel roles for FOXO factors in aggressive poor prognosis cancers, we built genetic models using CRISPR Cas9 genome editing technology (Vazquez et al. 2018). We specifically disrupted the FOXO3 gene with a neomycin resistance cassette (NPTII) producing a truncation mutant in glioblastoma (GBM) U87MG cells (Vazquez et al. 2018). Using this genetic model, we examined known FOXO target genes (identified in ES cells) for differential expression. We found significantly decreased expression of stem genes such as OCT4 and SOX2 in the foxo3 mutant U87MG cells compared to parental U87MG control cells (Figs. 1, 2). Subsequent experiments revealed that FOXO3 more broadly promoted stem gene expression and signal transduction (Figs. 2, 3). Inhibition of the PI3K pathway with NVP-BEZ235 led to dramatically increased expression of stem markers OCT4 and ALPP (encoding alkaline phosphatase), at least in part, via an FOXO-dependent mechanism. These results suggest a potential hazard for clinical use of PI3K inhibition as a therapeutic (especially in cancers that harbor stem signatures), (Figs. 3, 4). Taken together, our findings highlight a novel role for FOXO factors in cancer, which is to promote stem programs that likely contribute to aggressiveness.

Fig. 1
figure 1

FOXO3 impacts OCT4 Gene Expression. a Schematic of foxo3 disruption mutant protein (DNABD = DNA Binding Domain) and AD = transcriptional activation domain). b Total protein lysates prepared from foxo3 mutant containing U87MG cells and control cells were examined by western blot analysis; antibodies used for western blotting are indicated. Wild-type FOXO3 was approximately 80 kDa, whereas mutant foxo3 protein was approximately 45 kDa. c Gene expression (determined by qRT-PCR) of OCT4 in foxo3 disruption mutants with or without exogenous FOXO3. d Western blot of U87MG cells with exogenous FOXO3. A significant difference was indicated by Student’s t test compared to control U87MG cells (*) or cognate foxo3 mutant cell line (**)

Full size image
Fig. 2
figure 2

FOXO3 disruption mutants had reduced stem characteristics in U87MG cells. a Gene expression (for SOX2 and TUBB3, encoding a neuronal marker) was determined by qRT-PCR in foxo3 disruption mutants. Mutants had reduced SOX2 and increased TUBB3 expression. b Lysates from foxo3 mutants and control U87MG cells were investigated by western blot analysis. Mutants had reduced STAT3 Y705 phosphorylation. c Indicated cancer cell lines were plated at a density of 180 cells per ml and grown for 2 weeks. Colonies were stained with crystal violet and counted. *Significant difference indicated by Students T test

Full size image
Fig. 3
figure 3

Exogenous FOXO3 and Dual PI3K Inhibitor NVP-BEZ235 Induce OCT4. aOCT4 gene expression (determined by qRT-PCR) in four glioblastoma cell lines: U87MG, U118MG, DBTRG, and A172. b Expression of stem genes from samples with exogenous FOXO3 measured by qRT-PCR. The fold induction was relative to the control samples (CMV5 vector alone). FOXO3 and RICTOR were positive controls, whereas ACTIN (ACTB) was a negative control. The Benjamini–Hochberg Procedure was employed to correct for multiple hypothesis testing using an FDR < 0.05.c Indicated glioblastoma cell lines were treated for 5 days with NVP-BEZ235 and analyzed by qRT-PCR for OCT4 gene expression. All cell lines were treated with 50 nM NVP-BEZ235, except LN229, which was treated with 1 µM NVP-BEZ235. d U87MG and foxo3-disrupted U87MG cells were treated with 50 nM NVP-BEZ235 for 48 h. The Tukey method was utilized for multiple comparison testing with P < 0.05 denoted with *. The ** denotes P < 0.05 based on Tukey method between U87MG NVP control cells and foxo3 mutant treated with NVP. *Significant difference indicated by Student’s T test in panels a–c

Full size image
Fig. 4
figure 4

FOXO1 contributes to OCT4 gene expression in U87MG Cells. a, b Gene expression of OCT4 or indicated FOXO factor was determined by qRT-PCR in cells with indicated esiRNA (48 h post-transfection). FOXO1 esiRNA-treated cells had decreased OCT4 gene expression. c Cells were treated with DMSO, FOXO1 inhibitor (F1i, AS1842856 200 nM), and/or NVP-BEZ235 (50 nM) for 48 h and examined by q-RT-PCR. FOXO1 inhibition led to reduced OCT4 gene expression based on Tukey Test (P < 0.05), denoted by * in this panel. d ChIP was performed using extracts prepared from NVP-BEZ235 treated U87MG cells to investigate association of FOXO1 with previously mapped OCT4 Promoter Region O2 relative to a negative control conserved upstream region of human OCT4 (NEG Seq O). FOXO1 associated with the OCT4 O2 promoter sequence. e A model is depicted in which FOXO factors directly regulate OCT4 transcription in certain cancer cells such as U87MG GBM. *Significant difference indicated by Student’s T test in a, b, d

Full size image

Results

CRISPR Cas9-mediated FOXO3 disruption reduced OCT4 gene expression in U87MG cells

Poor-prognosis cancers such as GBM and BBC harbor stem cell signatures, including expression of genes OCT4 and SOX2, but the mechanisms that induce these programs remain to be fully elucidated (Ben-Porath et al. 2008). Given that FOXO factors induce OCT4 and SOX2 genes in human and mouse ES cells, we examined the expression of these putative targets in GBM cells that harbor nuclear FOXO (Zhang et al. 2011). In our previous work, we built genetic models using CRISPR Cas9 genome editing to disrupt the FOXO3 gene in U87MG GBM cells with a neomycin resistance cassette (Vazquez et al. 2018). The foxo3 truncation mutant protein retained the DNA-binding domain, but lacked the trans-activation domain and was approximately 45 kDa compared to the 80 kDa full-length protein (Figs. 1a, b) (Vazquez et al. 2018). To assess transcriptional consequences of FOXO3 disruption in U87MG cells, we performed qRT-PCR analyses with log-phase cells grown in rich media and found significantly less OCT4 gene expression in the mutant cells compared to control cells (Fig. 1c). Importantly, exogenous expression of FOXO3 restored OCT4 gene expression in the disruption mutants (Fig. 1c). Exogenous FOXO3 also induced OCT4 protein expression in U87MG cells (Fig. 1d). These results show for the first time that FOXO3 promotes the expression of the stem master regulator OCT4 in aggressive U87MG cells.

Loss-of-function foxo3 mutants have reduced stem signal transduction and evidence of differentiation

To further assess the impact of FOXO3 disruption on stem characteristics in U87MG cells, we examined expression of SOX2, which was shown to be impacted by FOXO factors in ES cells (Zhang et al. 2011). We found that SOX2 gene expression was reduced in foxo3 mutants (Fig. 2a). Interestingly, we observed an increase in TUBB3 (Tubulin Beta 3 class III) expression in the disruption mutants, suggesting that these cells may have adopted a neuronal-like fate (Fig. 2a) (Poirier et al. 2010).

To investigate the impact of FOXO3 disruption on stem signaling pathways in cancer, we examined activation of the STAT3 (signal transducer and activator of transcription 3) transcription factor. STAT3 promotes stem cell fate in part by inhibiting cellular differentiation (Raz et al. 1999). Upon IL6 or LIF ligand binding to cognate receptors, the associated Janus Kinase 2 (JAK2) phosphorylates STAT3 on tyrosine 705 thereby promoting stem gene expression (Galoczova et al. 2018; Marotta et al. 2011). We found that STAT3 Y705 phosphorylation was reduced in foxo3 disruption mutants compared to control cells, while there was no change in total STAT3 in foxo3 mutant cells (Fig. 2b). There was no change in AKT activation as evidenced by phosphorylation of AKT on serine 473 in the foxo3 disruption mutants suggesting a specific impact on stem-related signal transduction by FOXO3 in this context. Therefore, FOXO3 disruption specifically impacted STAT3 activation in U87MG cells, consistent with the idea that FOXO factors impact stem characteristics in this setting.

Clonogenicity was reduced in foxo3 disruption mutant U87MG cells

Disruption of the FOXO3 gene in the U87MG background gave rise to cells that appeared to have a growth defect. Clonogenicity assays were performed with foxo3 mutants and U87MG parental controls to assess the impact of FOXO3 disruption on colony formation. We found that disruption of FOXO3 led to a significant decline in the number of colonies obtained (Fig. 2c), highlighting a positive role for this factor in the growth and/or survival of U87MG cells.

Exogenous FOXO3 induced OCT4 in a set of glioblastoma cell lines

To gain insight into the impact of FOXO3 on stem gene expression in a set of glioblastoma cell lines, we exogenously expressed this factor (by transfection) and examined OCT4 gene expression by qRT-PCR. We found that exogenous FOXO3 led to increased OCT4 gene expression in four glioblastoma cell lines: U87MG, U118MG, DBTRG, and A172 as well as human embryonic kidney cells (HEK 293) and basal breast cancer cells (BT549), Figs. 3a and S1A. Exogenous FOXO1 also induced OCT4 in these settings (data not shown). The ability of FOXO factors to induce the stem master regulator OCT4 suggested that they should broadly impact stem gene expression and signal transduction. Therefore, we examined the ability of exogenous FOXO3 to regulate additional genes encoding transcription factors ascribed to promote stem signatures as well as ligands that induce stem programs (Fig. 3b) (Galoczova et al. 2018; Loh et al. 2006; Molyneaux et al. 2003; O'Connor et al. 2008). The positive control RICTOR was induced by exogenous FOXO3 whereas the negative control ACTIN (ACTB) was not. We found that stem transcription factors were strongly induced by exogenous FOXO3: SOX2 and NANOG (Nanog homeobox) (Fig. 3b). Furthermore, exogenous FOXO3 significantly induced the expression of stem pathway ligands LIF (Leukemia Inhibitory Factor), IL6 (Interleukin 6), and TGFB1 (Transforming Growth Factor beta1) whereas EGF (Epidermal Growth Factor) was not induced (Fig. 3b). These data indicate that FOXO3 broadly promoted a stem program in U87MG cancer cells potentially via the induction of the master regulator OCT4 (Zhang et al. 2011).

Treatment with PI3K pathway inhibitor NVP-BEZ235 induced OCT4 in glioblastoma cells

PI3K pathway activation is a hallmark of cancer and is required for the growth and survival of cancer cells (Keniry and Parsons 2008; Luo et al. 2003). In line with this, PI3K pathway inhibitors are actively explored as potential chemotherapeutics for cancer (Lin et al. 2012; Matsushima et al. 2015). To examine the impact of PI3K inhibition in cancer cells that harbor stem signatures, we treated six glioblastoma cell lines with the dual PI3K inhibitor NVP-BEZ235 (which inhibits both PI3K and mTOR) for 5 days and then examined changes in gene expression by qRT-PCR. Strikingly, we found that OCT4 gene expression was robustly induced by NVP-BEZ235 treatment (Fig. 3c). Of note, LN229 cells required a higher dose of NVP-BEZ235 to observe this induction (1μM compared to 50 nM). NVP-BEZ235 treatment also increased OCT4 expression in HEK 293 and BT549 cells, demonstrating that the impact of FOXO factors on stem genes is not limited to glioblastoma cells (Fig. S1B). We also examined the gene expression of the stem marker alkaline phosphatase (encoded by ALPP) by qRT-PCR in NVP-BEZ235-treated samples (O'Connor et al. 2008; Yu et al. 2015, 2018). We found that PI3K inhibition led to induction of ALPP gene expression in U87MG, LN229, A172, and BT549 cells (Fig. S1C). Therefore, these data suggest that PI3K inhibition leads to increased stem gene expression in a set of glioblastoma cell lines. To test whether NVP-BEZ235-mediated OCT4 induction was FOXO-dependent, we treated control and foxo3 disruption mutants with the drug for 48 h. We found that NVP-BEZ235 only induced OCT4 in the control U87MG cells. Cells that harbored a foxo3 disruption mutant lacked induction of OCT4 (Fig. 3d).

Given that the PI3K pathway is known to regulate FOXO nuclear localization, we checked whether treatment with the dual PI3K inhibitor NVP-BEZ235 impacted the localization of wild-type and/or mutant FOXO3 protein. We found that PI3K inhibition with NVP-BEZ235 alone led to less overall protein in extracts, but had no impact on FOXO3 localization in U87MG cells, consistent with findings by Zhang et al., in stem cell contexts such as ES and HSCs (Fig. S2) (Ghaffari et al. 2010; Liang et al. 2016). This is also in agreement with Sunayama et al. who found that the regulation of FOXO factor localization is complex in the glioblastoma setting of A172 cells and that inhibition with NVP-BEZ235 alone was not sufficient to direct FOXO3 to the nucleus (Sunayama et al. 2011).

Sunayama et al. found that dual inhibition of the Ras and PI3K pathways with MEK inhibitor UO126 and NVP-BEZ235, respectively, led to differentiation of glioblastoma cells A172 as evidenced by neuronal marker TUBB3 expression. In agreement with Sunayama et al., we found induction of the neuronal differentiation marker TUBB3 upon inhibition of MEK and PI3K (Figs. S3A–B) (Sunayama et al. 2011). We also found induction of TUBB3 with NVP-BEZ235 treatment alone in U87MG cells and concomitant induction of OCT4 (Figs. S3A-B). Of note, the dual treatment of U87MG cells (with UO126 and NVP-BEZ235) led to significantly higher inductions of OCT4 and TUBB3 compared to NVP-BEZ235 treatment alone (Fig. S3A), in support of the notion that Ras and PI3K coordinately impact differentiation. All of these data highlight the dual contributions of FOXO factors and PI3K to stem cell characteristics. Loss of either FOXO (Fig. 1c) or PI3K (Figs. S3A–B) can lead to stem cell differentiation, consistent with prior reports (Daniele et al. 2015; Ghaffari et al. 2010; Ikeda and Toyoshima 2017; Jones et al. 2016; Kumazoe et al. 2017; Liang et al. 2016; Renault et al. 2009; Rivas et al. 2018).

FOXO1 contributes to OCT4 expression in U87MG cells

We sought to determine which of the FOXO factors contributed to OCT4 expression in U87MG cells. RNAi to FOXO1 or FOXO3 led to reduced expression of OCT4 in ES cells; both of these factors directly associated with the OCT4 promoter in ES cells based on chromatin immunoprecipitation and EMSA experiments in prior studies (Zhang et al. 2011). To examine the contributions of FOXO1 or FOXO3 on OCT4 gene expression in U87MG cells, we transfected cells with esiRNA to target each factor independently. We found that esiRNA-mediated reduction of FOXO1 was associated with less basal OCT4 gene expression (Fig. 4a); NVP-BEZ235-induced OCT4 gene expression was not significantly changed by FOXO1 esiRNA (data not shown). Reduced FOXO3 (via esiRNA) did not have a significant impact on OCT4 gene expression (Fig. 4b). We also examined the ability of FOXO1 to impact OCT4 gene expression by utilizing an FOXO1 inhibitor. Treatment of U87MG cells with FOXO1 inhibitor AS1842856 led to decreased basal and NVP-BEZ235-induced OCT4 gene expression (Fig. 4c). Given that both FOXO1 and FOXO3 were shown to directly associate with the OCT4 promoter in ES cells, we examined their ability to bind this promoter in extracts prepared from U87MG cells. We detected FOXO1 association with the OCT4 promoter by quantitative chromatin immunoprecipitation analyses (Fig. 4d). Therefore, our results indicate that FOXO1 directly binds to the OCT4 promoter and contributes to OCT4 gene expression in U87MG cells.

Discussion

FOXO transcription factors are best known as partially redundant tumor suppressors that can induce apoptosis and halt the cell cycle (Brunet et al. 1999; Calnan and Brunet 2008). However, recent studies challenge this paradigm and point to roles for these factors in promoting cancer. FOXO1 was found mutated to produce a constitutively nuclear protein in DLBCL, which was associated with poor prognosis (Trinh et al. 2013). Hints of pro-oncogenic roles for FOXO factors were found in additional cancers such as AML (where 40% of AML patient samples had active FOXO regardless of genetic subtype) (Sykes et al. 2011). FOXO factors hindered differentiation and apoptosis in AML cell lines and mouse models (Sykes et al. 2011). FOXO3 was shown to promote TMZ resistance in glioblastoma cells (Xu et al. 2017). Other work indicated that FOXO transcription factors were at least in part localized to the nucleus in certain PI3K-activated cancers such as BBC BT549 cells, but the function of FOXO factors in these cancers was not identified (Keniry et al. 2013). Here, we took a genetic approach to investigate the role of FOXO factors in GBM cells and found that these factors contribute to the expression of stem genes such as OCT4, likely contributing to the aggressiveness of these cancers (Figs. 1, 2, 3, 4).

FOXO1 and at least in part FOXO3 directly bind to and induce stem cell genes such as OCT4 in ES cells (Zhang et al. 2011). Intriguingly, BBC and GBM commonly harbor stem cell gene expression signatures (Ben-Porath et al. 2008). We hypothesized that FOXO factors may drive stem cell signatures aggressive BBC and GBM cancer cells. In line with this notion, our results indicated that FOXO3 disruption led to significantly reduced gene expression of OCT4 and SOX2 in U87MG cells, (Figs. 1, 2) and exogenous expression induced these genes in glioblastoma cell lines (Fig. 3a). Therefore, FOXO transcription factors regulate OCT4 gene expression in glioblastoma cells (Figs. 1c–d, 2a, 3a).

OCT4 is a potent transcriptional master regulator of pluripotency. Exogenous expression of this factor is enough to instill stem-like properties to fibroblasts (Li et al. 2011). The broad impacts that FOXO3 had on stem gene expression and signal transduction underscored a novel and exciting role for these factors in aggressive GBM (Figs. 1, 2, 3, 4). Given that FOXO factors (-1, -3, and -4) are commonly redundant, it is likely that a combination of these factors contributes to stem gene expression in U87MG cells and similar cancer settings.

Insights into FOXO3 subcellular localization and function in the setting of glioblastoma were previously described by Sunayama et al. (2011). In Sunayama et al., FOXO3 contributed to differentiation upon PI3K and Ras inhibition in glioblastoma cells. We performed similar experiments and found increased expression of TUBB3 in PI3K/Ras inhibited U87MG cells, in agreement with Sunayama et al. (Fig. S3A–B). Of note, we found increased TUBB3 gene expression in U87MG cells that were only treated with NVP-BEZ235 (Fig. S3A–B). One possibility is that Ras might not have been as activated in U87MG cells; therefore, inhibition of NVP-BEZ235 alone may have been enough to induce differentiation. We detected induction of both differentiation and stem genes under NVP-BEZ235 treatment in U87MG cells (Figs. S3A–B). Intriguingly, we do not know yet whether the same cells express stem and differentiation genes, or distinct populations express these genes.

There was one slight difference between our work and Sunayama et al.; we found that FOXO3 disruption led to increased TUBB3 (Fig. 2a), whereas Sunayama et al. found that cells with decreased FOXO3 by siRNA had reduced TUBB3 gene expression. One possible explanation for this discrepancy is that FOXO3 disruption led to a loss in stem characteristics and differentiation (including TUBB3 induction) over a period of weeks in our experiments compared to three days for Sunayama et al. (2011). It is noteworthy that although FOXO3 protein was nearly undetectable after RNAi targeting in work by Sunayama et al., TUBB3 was still induced, albeit at a reduced level. This suggests that other FOXO factors may have contributed to TUBB3 expression in these experiments. Similarly, in our experiments, other factors may have promoted TUBB3 induction (such as FOXO1) when FOXO3 was disrupted.

Our results from subcellular localization experiments for FOXO factors matched Sunayama et al. (2011). Treatment with NVP-BEZ235 alone did not appreciably change FOXO3 (wild-type or disruption mutant) subcellular localization, refining conventional paradigms for this pathway (Fig. S2). FOXO1 and FOXO4 localization was not impacted by NVP-BEZ235 treatment of U87MG cells (data not shown). Even though the localization of FOXO factors were unchanged by NVP-BEZ235 treatment, it appeared that target genes were induced (Figs. 3c and S2A).

Other studies have highlighted the ability of NVP-BEZ235 to induce FOXO output in cancer cells. Lin et al. examined feedback responses under low-dose NVP-BEZ235 conditions (Lin et al. 2014). One key point from Lin et al. was that the FOXO-dependent resurgence of PI3K activity was only observed at a relatively low dose of NVP-BEZ235 (Lin et al. 2014). Lin et al. specifically chose to use a low dose of NVP-BEZ235, because higher dosages would likely be toxic to cancer patients. Our work built on the work by Lin et al. In addition to deleterious effects mediated by FOXO that led to PI3K pathway reactivation (RICTOR induction, Fig. 3b), NVP-BEZ235 treatment had additional deleterious effects that could greatly impact drug efficacy as a chemotherapeutic. We found that stem genes were induced upon NVP-BEZ235 treatment (Figs. 3c and S1B-C). Our experiments mostly employed low dosage NVP-BEZ235 (similar to experiments performed by Lin et al.). Two sets of our experiments utilized higher doses of NVP-BEZ235. We needed to use a higher dose of NVP-BEZ235 to observe stem gene induction in the glioblastoma cell line LN229 (1 µM compared to 50 nM for other cell lines, Fig. 3c). Perhaps, the PI3K pathway is more active in LN229 cells. We also utilized a higher NVP-BEZ235 dose (1 µM) when examining subcellular localization to compare our work to Sunayama et al. (who utilized 1 µM) as well as to examine changes in gene expression with combined MEK and PI3K inhibition (Lin et al. 2014; Sunayama et al. 2011).

Targeting PI3K as a chemotherapeutic for cancer requires consideration of homeostatic feedback mechanisms and possible unwanted activation of pathways that may promote cancer aggressiveness. The efficacy of PI3K targeted therapy for advanced solid tumors was limited and did not significantly increase overall survival or objective response rate across 46 randomized-controlled clinical trials (Li et al. 2018). Work by Caino et al., demonstrated that PI3K pathway inhibition led to increased cellular motility in prostate cancer cells as well as reactivation of PI3K signal transduction (Caino et al. 2015). Hopkins et al. mitigated the homeostatic feedback mechanisms induced by PI3K inhibition by utilizing a ketogenic diet in a murine pancreatic cancer model, highlighting novel avenues that show promise for future therapeutic protocols (Hopkins et al. 2018). Our work has novel implications for transcriptional changes that arise upon PI3K inhibition. We found PI3K inhibition by NVP-BEZ235 led to the robust induction of stem genes such as OCT4 in a set of glioblastoma cell lines and BT549 cells. It remains unclear if PI3K inhibition will broadly induce stem programs across varied subtypes of cancer or if it will only be pertinent to the most aggressive cancers that already harbor stem signatures such as glioblastoma cells and BT549.

Conclusions

FOXO transcription factors function in a context-dependent manner (Paik et al. 2007). In some cancers, FOXO factors cause apoptosis (Calnan and Brunet 2008). We found that FOXO factors promote stem gene expression in GBM cells (Figs. 1, 2, 3, 4). Treatment with the dual PI3K inhibitor NVP-BEZ235 induced stem gene expression in glioblastoma cell lines and BT549 cells (Fig. 2c, d), having important implications for the efficacy of PI3K inhibition as a chemotherapeutic option. Importantly, expression of stem programs is strongly associated with poor prognosis, suggesting that PI3K inhibition may make some susceptible cancers more aggressive (Iwadate et al. 2017). Perhaps combined PI3K pathway and FOXO inhibition would be more efficacious as a cancer therapy, because it would target proliferation and residual stem cells.

Methods

Cell culture and drug treatments

Cell lines were obtained from ATCC (American Type Culture Collection, Manassas, VA) and grown under standard conditions (5% CO2, 10% FBS (fetal bovine serum), with 5% antifungal/antibacterial). Cell lines were tested for Mycoplasma using the MycoAlert Mycoplasma Detection Kit (Lonza, Basel Switzerland, cat: LT07-218); all experiments were done with mycoplasma negative cells. U87MG cells were propagated in MEM (Minimal Essential Medium). BT549 and DBTRG cells were propagated in RPMI (Roswell Park Memorial Institute 1640 Medium). HEK 293, LN18, U118MG, A172, and LN229 cells were propagated in DMEM (Dulbecco's Modified Eagle Medium). NVP-BEZ235 was purchased from Sigma (Saint Louis, MO), and utilized at a final concentration of 50 nM in indicated experiments. Two sets of experiments utilized a higher dose of NVP-BEZ235. LN229 cells were treated with 1 µM NVP-BEZ235 and data from Figs. S3A–B were generated using 1 µM NVP-BEZ235 treatment. UO126 was purchased from Selleck Chemicals (Houston, TX) and used at 10 µM final concentration. Cells were plated at a density of 2700 cells per ml and were treated for 5 days with NVP-BEZ235 (unless otherwise stated). AS1842856 was purchased from Calbiochem (Danvers, MA) and utilized at 200 nM final concentration.

Transfection

Cells were grown under standard propagating conditions (37 °C with 5% CO2 in media containing 10% FBS and 5% pen/strep) and harvested by trypsin treatment, while the cells were in the log growth phase. One million cells were transfected using Lonza Nucleofection kit V (Lonza, Basel Switzerland), program P-20 (U87MG, A172, LN229, LN18, U118, DBTRG, and BT549) or X-001 (HEK 293) and allowed to recover for 24 h in 10 cm culture dishes. The foxo3 disruption mutants and controls were electroporated using program T-020 with Lonza Nucleofection kit V and 1 million cells. One microgram of indicated plasmid was utilized in each transfection. Plasmids were described previously (Keniry et al. 2013). FOXO3-CMV5 was obtained from D. Accili (Nakae et al. 2001). Vectors prepared for CRISPR Cas9 mutagenesis were previously described (Vazquez et al. 2018). The CRISPR donor vector was prepared using a pCDNA3 backbone with chromosomal FOXO3 sequences to enable integration into FOXO3 loci to disrupt FOXO3 with a NPTII neomycin resistance cassette. Isolates were previously confirmed by sequencing and western blot (Vazquez et al. 2018). RNAi experiments utilized Lonza Nucleofection kit V and 1 million cells with 600 ng of indicated esiRNAs from Sigma per transfection (eGFP: EHUEGFP), (FOXO1: EHU156591) and (FOXO3: EHU113611). RNAi lysates were collected 48 h post-transfection.

Western blot

Total protein was obtained from indicated cells by rinsing cells with 1XPBS (phosphate-buffered saline) followed by directed lysis in 2 × sample buffer (125 mM Tris–HCL at pH 6.8, 2% sodium dodecyl sulfate (SDS), 10% 2-mercaptoethanol, 20% glycerol, 0.05% bromophenol blue, 8 M urea); 2 × sample buffer was added to each well and cells scraped with a cell scraper. The lysate was collected from each well, placed into 1.5 ml microcentrifuge tubes, and heated for 10 min at 95 °C in a dry-bath heat block. Protein lysates were separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) at 100 V for 1 h. Resolved proteins were then transferred onto a polyvinylidene fluoride (PVDF) membrane for an hour and 30 min then blocked in a 5% milk solution (Carnation powdered milk, 1X Tris-buffered saline with Tween 20 (TBST) for an hour. Membranes were incubated with indicated primary antibody overnight at 4 °C then washed for 20 min with TBST in 5-min intervals. The blot was then incubated with secondary antibody for 1.5 h. Membranes were washed for 20 min in 5-min intervals and allowed to develop using SuperSignal West Dura Extended Duration Substrate luminol solution (Pierce Biotechnology, Waltham, MA) for 5 min. A Bio Rad ChemDoc XRS + Molecular Imager was utilized for protein detection (Bio Rad Hercules, CA). Data were analyzed with NIH Image J. Antibodies were obtained from Cell Signaling Technologies (Danvers, MA): FOXO3 (75D8), FOXO1 (C29H4), FOXO4 (9472S), OCT4 (2750S), total STAT3 (9139 T), phospho STAT3 tyrosine 705 (9145 T), and Phospho-p44/42 MAPK (Erk1/2) (Thr202/Tyr204) (D13.14.4E) XP® Rabbit mAb catalog: 4370), Histone H3 (96C10, catalog: 3638), and phospho AKT serine 473 (9271S). Antibodies obtained from Cell Signaling Technologies were diluted 1:1000 in TBST containing 5% BSA and were incubated with blots overnight at 4 °C. GAPDH (G-9) was obtained from Santa Cruz Biotechnology Inc. (Dallas, TX) and utilized at a 1:2000 dilution in TBST with 5% non-fat dried milk. Beta-Actin antibody (clone AC-74, cat: A2228) was obtained from Sigma and utilized at a 1:2000 dilution in TBST with 5% non-fat dried milk.

Quantitative real-time PCR

Total RNA was prepared using the Qiagen RNeasy kit (Hilden, Germany), which was then used to generate cDNA using Superscript Reverse Transcriptase II (Invitrogen, Carlsbad, CA). Samples (cDNAs) were analyzed using (Power SYBR Green Master Mix, Applied Biosystems, Foster City, CA) and the Illumina Eco Real-time system (San Diego, CA). Expression levels were normalized to GAPDH in gene expression experiments and calculated using 2−∆∆CT method (Livak and Schmittgen 2001). Primer sequences are detailed in supplemental Table S1.

Clonogenicity assays

Indicated cells were treated with trypsin and counted using a Millipore Scepter automated cell counter (Millipore, Burlington, MA). Cells were diluted into media at a final concentration of 180 cells per ml. 2 ml of cells were plated per well of a 6-well plate and were grown for 2 weeks. Colonies were stained with 0.5% crystal violet in buffered formalin and washed with PBS. All visible colonies were counted for each sample irrespective of the cell number. Experiments were performed several times in quadruplicate.

ChIP analyses

Data presented were from U87MG cells treated with 50 nM NVP-BEZ235 for 5 days. Chromatin immunoprecipitation (ChIP) analyses were performed as previously described (Niu et al. 2003), except an additional micrococcal nuclease step was added to shear chromatin. 10 million U87MG cells grown in 15 cm dishes were cross-linked by adding formaldehyde to final concentration of 1% and incubated in room temperature for 10 min; after, 125 mM glycine was added and samples were incubated for an additional 5 min at room temperature. Cells were washed twice with PBS and scraped into ice-cold PBS, and then collected by centrifugation at 200×g. Cells were washed with PBS and resuspended in ChIP lysis buffer (1% SDS, 1 mM EDTA, 50 mM Tris–HCl pH 7.5.0) Cells were lysed for 10 min on ice, sonicated on a Branson 250 Sonicator at 20% power for 30 s 4 times. After this, 18 μl of micrococcal nuclease solution (Thermo-Fisher Waltham MA, catalog: 88216) and 430 μl of 10 mM CaCl2 were added to 10 ml chromatin mixture and incubated for 10 min. The reaction was stopped by adding 100 μl of 0.5 M EDTA. Next, chromatin preps were diluted 1–10 with dilution buffer (20 mM Tris–HCl pH 8.0, 1 mM EDTA, 0.1% sodium deoxycholate, 140 mM NaCl, 0.01% SDS, and 1% NP40). Chromatin was pre-cleared with 80 μl of magnetic protein A/G beads (Pierce Biotechnology, Waltham, MA, catalog: 78609) for 1 h at 4 °C. Pre-cleared chromatin preparations (prepared from 1 million cells) were incubated overnight with 7 µg of normal rabbit IgG (Santa Cruz Biotechnology Inc., Dallas, TX, catalog: SC2017) or FOXO1 antibody (C29H4) at 4 °C. Samples were next incubated with 80 µl of Protein A/G agarose for 1 h and then washed for 5 min with each buffer: Low Salt Buffer (20 mM Tris pH 8.0, 2 mM EDTA, 0.1% SDS, 1% Triton X100, and 150 mM NaCl), High Salt Buffer (20 mM Tris pH 8.0, 2 mM EDTA, 0.1% SDS, 1% Triton X100, and 500 mM NaCl), and LiCl Buffer (20mMTris pH 8.0, 2 mM EDTA, 0.5% sodium deoxycholate, 0.1% NP40, and 250 mM LiCl). After washes, DNA was eluted by incubating beads with Elution buffer (1% SDS, 0.1 M NaHCO3) at 37 °C for 15 min (step performed twice). Crosslinks were reversed by adding 40 µl 2.5 M NaCl to the combined eluates (500 μl) and heating at 65 °C overnight. DNA was purified using phenol:chloroform extraction and sodium acetate/ethanol precipitation. The purified DNA was subjected to quantitative real-time PCR. OCT4 promoter sequence O2 association levels (mapped by Zhang et al.) were normalized to OCT4 control sequence (utilized by Zhang et al.) in ChIP experiments and calculated using 2−∆∆CT method (Livak and Schmittgen 2001; Zhang et al. 2011). Primer sequences utilized for ChIP quantitative PCR are detailed in supplemental Table S2.

Subcellular fractionations

Subcellular fractionations were prepared as described previously (Keniry et al. 2013). 10 million log-phase cells were scraped from 15 cm plates into ice-cold PBS. Cells were washed once with ice-cold PBS (containing 1 × PI). Cells were incubated in Buffer A (10 mM Hepes pH 7.4, 10 mM KCL, 100 µM EDTA, 1 X PI) for 15 min on ice. After this, 10% NP-40 was added to the Buffer A mixture at a 1:20 dilution. Samples were vortexed and incubated on ice for 2 min with mixing by inversion every 30 s. Next, samples were centrifuged at 1000×g for 5 min. The cytoplasmic fractions were collected from the supernatant and mixed one to one with 2 × sample buffer (125 mM Tris–HCL at pH 6.8, 2% sodium dodecyl sulfate (SDS), 10% 2-mercaptoethanol, 20% glycerol, 0.05% bromophenol blue, and 8 M urea). Cytoplasmic fractions (mixed with 2 × sample buffer) were heated for 10 min at 95 °C in a dry-bath heat block. The nuclear pellets (from the 1000×g centrifugation step) were washed with Buffer A and then washed with Buffer C (20 mM Hepes pH 7.4, 400 mM NaCl, 1 mM EDTA, 1 X PI). The washed nuclear pellets were lysed in 2 × sample buffer and heated for 10 min at 95 °C in a dry-bath heat block. Samples were analyzed by western blot analyses. 25% of each nuclear fraction was resolved in each well for western blot analyses. 1% of each cytoplasmic fraction was resolved in each well for western blot analyses. Experiments were repeated several times with similar results.