A Large-Scale RNAi Screen Identifies SGK1 as a Key Survival Kinase for GBM Stem Cells

GBM is the most common type of primary tumor of the brain, accounting for approximately 45% of malignant gliomas. The hallmark histologic features of this tumor include high mitotic index, diffuse brain infiltration, presence of necrotic regions as well as microvascular proliferation in the tumors (1). The current standard of care for patients is maximal safe resection surgery of tumor from the brain, followed by a regimen of radio and chemotherapy with the DNA alkylating agent, temozolomide. This therapy reduces the tumor bulk, but is not curative as recurrence is very common. Even with this aggressive treatment regimen, the prognosis for GBMs remains poor with median survival of only 14–15 months postdiagnosis and a low 5.3% 5-year survival rate (2).

Solid tumors such as GBM usually outgrow the normal nutrient supply available resulting in the presence of a range of oxygen concentrations in different parts, ranging from 0.1% to 5%. This is usually in the form of a gradient, with higher amount of oxygen available to regions near vasculature, and this oxygen availability decreases as a function of distance from the blood vessel (3). Necrotic regions are formed in niches exhibiting severe hypoxia/anoxia and lack of nutrient supply (4). Hypoxic regions are a hallmark of GBM pathogenesis with the tumors exhibiting necrotic cores termed as pseudopalisades (1). The physiologic importance of tumor hypoxia is underscored by the poor prognosis associated with the increasing volume of necrotic and hypoxic niches found in the tumor. Furthermore, the presence of these niches negatively impacts the effectiveness of the radio and chemotherapy (5).

At the molecular level, GBM tumors display a significant degree of intratumoral heterogeneity. Current evidence suggests that a subpopulation of cells termed GBM stem-like cells (GBM-SC) are critical for the initiation and maintenance of the tumor. These cells possess the ability to self-renew as well as differentiate into various brain-specific lineages such as neuron and astrocyte like cells. These cells are positive for various stem cell–specific markers such as nestin, Olig2, and CD133 as well as possess similar characteristics to neural stem cells in culture. This population has been isolated from various brain tumors using both marker-based separation as well as phenotypic isolation in neural stem cell–specific media (6). GBM-SCs have been shown to possess a gene expression signature more similar to the parental tumor as compared with serum lines derived from the same tumor and traditional serum glioma lines used in the literature (7). Compared with the non-stem bulk tumor, these cells possess high tumorigenic potential and form tumors that phenocopy the characteristics of the parental tumors (8, 9). Furthermore, this GBM-SC in vitro model is clinically relevant, as this population is more chemo and radioresistant than the bulk tumor cell population and is enriched posttreatment (10, 11). Conventional therapy is able to target the bulk tumor, but the GBM-SC population preferentially survives due to its resistant nature. Subsequently, these cells can proliferate and form a tumor, leading to recurrence (12). Targeting this population specifically, in addition to the tumor bulk, should be advantageous for treatment of GBMs.

GBM-SCs are also enriched in hypoxic microenvironments, which in turn aid in the resistance properties of the cells as well as increase their tumorigenicity and their ability to self-renew (13). All cells primarily respond to hypoxic stress through the transcription factors, hypoxia-inducible factors (HIF) 1α and 2α (14). Indeed, these factors have shown to be important in GBM-SC for self-renewal, survival, and tumorigenicity, although HIF2α is thought to be the primary mediator of this response (15). Although the hypoxic response through HIF1α and HIF2α is well studied, the role of other genetic factors involved in the response, which may be independent of regulation by the HIF pathway, are not as well understood. Given the critical role of the hypoxic microenvironment in GBMs and its role in GBM-SC population maintenance, there is a need to identify genes that are critical to the hypoxic response in these cells, which may be targeted to produce a more efficacious therapeutic response. GBM-SCs under hypoxic conditions, with elevated chemotherapeutic and radiotherapeutic resistance genes, may escape treatment and give rise to recurrence. To find effective treatment alternatives, targeted therapies against GBM-SCs need to be developed, which can decrease cancer cell viability in both normoxic as well as hypoxic conditions to achieve complete remission of the tumor.

To identify genes that are important for the growth and survival of GBM-SCs, we have performed an unbiased pooled shRNA screen to identify genes that are important in GBM-SC growth and survival under both normoxic and hypoxic conditions. Using this negative selection screen, we have screened approximately 10,000 genes in two different patient-derived GBM-SC lines and have validated SGK1 as a potential target gene that is important for survival of GBM stem cells in vitro as well as in vivo.

Primary GBM stem cell (GBM-SC) lines (GS6-22, GS7-2, GS11-1, GS13-1 and MGG 8), serum glioma lines (U251, A172, SF126, SF763), and human fibroblast line, Hs27, were cultured and differentiated (where stated) as described previously (16). MGG8 lines were kindly provided by Hiroaki Wakimoto, PhD (Massachusetts General Hospital, Boston, MA; ref. 9). H04 and SW06 normal human neural progenitor cells were provided by Dr. Dennis Steindler. The other GBM-SCs were established in our laboratory. For hypoxia experiments, C-chambers with a custom gas mix (5% CO₂, balance nitrogen; Airgas) was setup and used as per manufacturer's instructions (BioSpherix) and cells were incubated for indicated times. The cells were treated with the following inhibitors for indicated times in this study: GSK650394 (SGK1; Tocris), Monastrol (KIF11). Cell lines were authenticated by STR profiling by Genetica DNA Laboratories in July 2017.

GBM-SCs were transfected with siRNA using Lipofectamine RNAiMax transfection reagent, according to the manufacturer's protocol (Invitrogen). For viral transduction of GBM-SCs and other glioma lines, cells were transduced with indicated virus particles and transduced cells selected by puromycin (0.1–1 μg/mL). Cells were allowed to grow for 5–10 days or 3 days after selection for viability assays and RNA/protein isolation, respectively. For CRISPR-based gene editing, guide sequences were designed (www.crispr.mit.edu) and cloned into LenticrisprV2 plasmid (a gift from Feng Zhang - Addgene 52961) according to Zhang laboratory guidelines.

GBM-SCs were transduced with pooled lentiviral library at MOI of 0.5 to minimize the number of cells with multiple shRNA integrants. Infections were performed at a scale to achieve >200 infected cells per shRNA to minimize intercellular variation. After puromycin selection for 2 days, genomic DNA was isolated from a portion of the cultures as a reference time point for infection efficiency/shRNA. The remaining cells were placed either under normoxic (21% O₂) or hypoxic (1% O₂) conditions for 10 and 18 days, respectively. Screening was performed in triplicate for both cell lines in both conditions. Following two-step PCR from genomic DNA, which added an index barcode sequence for each condition, next-generation sequencing (NGS) was performed in the Tufts University Core Facility using the Illumina HiSeq 2500 High Throughput Sequencer. The Galaxy software platform was used to separate indexed samples and barcode deconvolution was performed using Decipher Deconvolution software provided by the manufacturer (Cellecta Inc). Data analysis was performed by calculating log₂ fold change of each barcoded shRNA in the experimental sample (hypoxia or normoxia) compared with the cognate shRNA in the control reference sample and hits identified as described in the Supplementary Methods section.

To measure viability of GBM-SC and neural stem cell lines in response to shRNA or inhibitor treatment, cells were plated at subconfluent densities in 96-well plates. Lentiviral particles and inhibitors were added to the wells containing the cells as indicated and the cells were allowed to grow for 5–7 days. Cell viability was assessed using PrestoBlue Cell Viability reagent (Life Technologies) and fluorescence was measured in a Tecan Spectrafluor Plus plate reader (Ex 544 nm/Em 590 nm). Cell viability of adherent glioma (SF126, SF763, U251, A172) and fibroblast (Hs27) cell lines was assayed using CellTiter 96 AQ_ueous One Solution Cell Proliferation (MTS) assay (Promega) according to the manufacturer's protocol.

shRNA or drug-treated cells were stained using an Annexin V-GFP/PI staining kit (BD Pharmingen) according to the manufacturer's protocol. Briefly, 1 × 10⁴ cells were infected in 6-well plate with control or targeted shRNA and selected with puromycin. Cells were treated with accutase, washed twice with 0.1% BSA in PBS, and resuspended in PBS at 1 × 10⁶ cells/mL density. FITC-conjugated Annexin V and/or propidium iodide (PI) was added to the cells and incubated for 15 minutes at room temperature. Control lentivirus–infected cells, which were unstained or stained with only FITC Annexin V or PI, were used as controls. Cells were analyzed using a FACSCalibur flow cytometer and data were analyzed using Summit software.

To identify genes important for the proliferation and survival of GBM stem cells under normoxic and hypoxic conditions, we performed a pooled shRNA screen. Approximately 10,000 genes were targeted with an average of five independent barcoded shRNAs per gene on two patient-derived GBM stem cell lines. The GS6-22 and MGG 8 GBM stem cell lines were each transduced, in triplicate, with the lentiviral library and grown under normoxic (21%) or under hypoxic conditions (1% O₂). Following 10 days under normoxia or 18 days under hypoxic conditions, shRNA barcodes were amplified by PCR from genomic DNA, and high-throughput sequencing was performed. Individual shRNA abundances were enumerated, normalized, and the normoxic/hypoxic samples were compared with reference control (Fig. 1A). The relative hairpin abundances from the normoxic and hypoxic samples were analyzed using median absolute deviation (MAD) method as described previously (17). A cutoff of MAD ≤ −3 and fold change (condition/reference) ≤0.5 was applied to identify hairpins causing a significant growth phenotype. A gene was considered a candidate hit if at least 2 hairpins targeting the gene, matched the above criteria.

We identified 457 (∼4.5%) and 727 (∼7%) genes, which were essential for GS6-22 cells under normoxic and hypoxic conditions respectively. Of these, 247 genes were important to maintain viability under both conditions in this cell line (Fig. 1B; Supplementary Table S1). MGG 8 cells required 531 (∼5%) genes and 178 (∼2%) genes for survival or proliferation in normoxia and hypoxia, respectively, out of which 123 genes were essential in both conditions (Fig. 1C; Supplementary Table S1). Under normoxic conditions, 173 genes are essential to both cell lines, which represent 32% and 38% of the total normoxia hits in either GS6-22 and MGG 8 cells, respectively. A total of 117 genes (∼1.2%) were found to be required for the growth of both cell lines under hypoxic conditions (Supplementary Fig. S1A and S1D). Comparing the genes required for growth under normoxic and hypoxic conditions for both cell lines screened, we identified 81 genes which were essential for cellular growth and survival (Fig. 1D and E; Supplementary Table S2). These essential genes represent the best candidates for drug discovery for targeting GBM stem cells across multiple GBM tumors under both normoxia and hypoxia (Supplementary Table S2). A previous large-scale comparison of multiple RNAi screens identified a common set of genes that were required for growth or survival of a multiplicity of cell lines (18). We found that nearly half of the genes identified as hits in our screen (∼52%) are genes essential for survival of multiple cell types and thus, may not be ideal for drug development due to potential therapeutic side effects (Supplementary Table S3). However, the enrichment of these genes in the screen confirms the validity of the screen. In addition, we also identified targets that were required only under either normoxic or hypoxic conditions (Supplementary Table S5; Supplementary Fig. S1D).

The list of essential genes consists in part of various proteins involved in key housekeeping functions of the cells. Top hits of the screen include genes important in transcription (e.g., POLR2B), translation (e.g., RPS6, RPL30), as well as the proteasomal machinery (e.g., PSMA1, PSMB2; Supplementary Table S2). In addition to known essential genes, various hits were identified which have not been described as commonly essential in different cell lines (Supplementary Table S4). We independently validated the top hits of the screen including Ring Box-1 (RBX1), an E3 ubiquitin ligase and Kinesin 11 (KIF11). RBX1 and KIF11 play a critical role in cell cycle progression (19, 20). Depletion of RBX1 protein by shRNA resulted in a significant decrease in cell viability of the GBM-SC lines (Fig. 1F; Supplementary Fig. S1B). KIF11 was validated as an essential gene using depletion by siRNA (Fig. 1G). In addition, treatment of GBM-SCs with the specific small-molecule inhibitor of KIF11, monastrol, demonstrated a significant decrease in GBM-SC viability, further indicating its functional requirement in the GBM-SCs (Supplementary Fig. S1C).

Kinases regulate processes critical for survival and growth of cells and many kinase inhibitors are available and clinically approved (21). We examined the kinases, which were found to be important in cellular proliferation and survival of GBM-SCs in normoxic and hypoxic conditions. Of the approximately 500 kinases tested, 64 kinases (10.6%) were hits in GS6-22 or GBM 8 in at least one of the oxygen conditions. Only three serine/threonine kinases, Polo-like Kinase 1 (PLK1), Serine/Threonine Kinase 36 (STK36), and Serum and Glucocorticoid regulated Kinase 1 (SGK1) were identified as essential for both GS6-22 and MGG 8 cell lines under both normoxic and hypoxic conditions (Fig. 2A). As the common essential kinases were few in number, we performed pathway analysis on all the kinases scored as hits in the screen, by combining kinases which were unique or common to either condition or cell line. The top represented pathways included glioma signaling and metabolic salvage pathways (Fig. 2B). PLK1 is a mitotic kinase that has been previously implicated in GBM (22). STK36 is part of the hedgehog pathway, a pathway implicated in GBM stem cells, but has not been well studied (23, 24).

SGK1 has been extensively characterized and has been implicated in the growth of serum grown glioma cell lines (25), but has not been demonstrated to be a required kinase for GBM stem cell function. It is a member of the AGC kinase family that includes AKT, and has a known role in regulating cell fate (26). We initially validated SGK1 as a protein essential for GBM-SC proliferation and survival using three shRNAs independent from the constructs used during screening. Depletion of SGK1 mRNA and protein led to a significant decrease in cell viability of GS6-22 and MGG 8 cell lines (Fig. 2C and D; Supplementary Fig. S2A). Similar sensitivity to SGK1 knockdown was observed in three additional patient-derived GBM-SC lines indicating that it is frequently required for growth and survival of GBM-SCs (Supplementary Fig. S2B).

The requirement of SGK1 for in vivo tumor growth was assayed using a stereotactic intracranial xenograft mouse model. Luciferase-tagged MGG 8 cells were transduced with SGK1 shRNA or scrambled sequence lentivirus, selected, and injected into female NCR/nude mice. SGK1 depletion led to a significant increase in mouse survival as compared with the scrambled control, indicating the importance of SGK1 for in vivo tumor growth (Fig. 2E).

Additional validation of SGK1 essentiality was tested using CRISPR/Cas9–based gene knockout. We designed two small guide RNAs targeting SGK1 gene exons. Expression of small guide RNAs with Cas9 abolished SGK1 protein expression and resulted in a severe loss of cellular viability of GS6-22 cell line compared with the nontargeting guide sequence (Fig. 3A and B). MGG 8 and GS11-1 cell lines were also susceptible to knock out of SGK1 by the guide RNAs (Fig. 3C). Furthermore, to determine whether the kinase activity for SGK1 is essential for GBM-SCs, GS6-22 and MGG 8 cell lines were treated with the SGK1-specific inhibitor, GSK650394. Treatment with the inhibitor decreased cell viability in both GS6-22 and MGG 8 cell lines (Fig. 3D–F).

We next investigated whether other glioma cells, which do not possess stem-like properties, were sensitive to SGK1 inhibition. To this end, we tested the effect of SGK1 knockdown and inhibition on traditional serum glioma lines that are not stem-like cells and are grown in presence of 10%–15% serum in the culture media. Guide RNAs targeting SGK1 were used to knockout SGK1 in Cas9-expressing U251 and A172 glioma cell lines. Although efficient SGK1 protein knockdown was achieved, U251 cell proliferation and survival was not affected by depletion of SGK1 as compared with nontargeting guide RNA control (Fig. 4A and B). Furthermore, the A172 cell line only exhibited modest sensitivity to SGK1 knockout (Fig. 4A). Although A172 shows a trend of sensitivity to the guide RNAs, this cell line overcomes the proliferation defect after passaging and grows at a rate similar to the nontargeting control (data not shown). This difference in sensitivity between the different cell types is not due to the lack of SGK1 expression in the cell lines used (data not shown). Interestingly, using the Oncomine database to analyze SGK1 mRNA expression in various GBM-SC and serum cell lines data using the Lee study database, we observed that the serum lines, which are insensitive to SGK1 depletion, actually have the highest level of expression of SGK1 (Supplementary Fig. S3A and S3B). To test sensitivity of the cell lines to SGK1 pharmacologic inhibition, we measured change in cellular viability of serum glioma lines in response to increasing doses of SGK1 inhibitor. The glioma cell lines were refractory to the effect of SGK1 pharmacologic inhibition as compared with the GBM-SC cell lines (Fig. 4C). In addition, the human normal fibroblast cell line, HS27 was also insensitive to SGK1 knockdown by shRNA as well as pharmacologic inhibition by GSK650394 (Fig. 4D; Supplementary Fig. S3C).

Because of the lack of a growth phenotype observed in non-stem cell lines following SGK1 knockout, we hypothesized that SGK1 activity is exclusively important in stem cell cultures. To test this, we treated normal neural progenitor lines, H04 and SW06, with the SGK1 inhibitor GSK650394. Both cell lines showed decreased viability when treated with the drug (Fig. 5A). Although the effect was not as pronounced as observed in GBM-SCs, these cells display sensitivity to SGK1 inhibition. GBM-SCs may be differentiated in culture in the presence of media that is growth factor deprived and contains 2% serum when grown on laminin-coated plates. This provides a direct assay to test the essentiality of SGK1 in differentiated cell types. We inhibited SGK1 activity using inhibitor GSK650394 on the differentiated and undifferentiated GS6-22 cell line cultured on laminin. Interestingly, we observed a loss of sensitivity of GBM-SCs to SGK1 inhibition upon differentiation, indicating that SGK1 is no longer required by the cells upon differentiation (Fig. 5B). GS11-1 cells also show a similar pattern of drug sensitivity upon differentiation (Fig. 5C). This desensitization is not due to loss of SGK1 protein in differentiated GBM-SCs as there is only a modest decrease in SGK1 expression (Fig. 5D).

The reduced number of viable cells observed due to SGK1 depletion could be due to either the loss of proliferation or due to death of the GBM-SCs. SGK1 depletion did not impact the expression of known stem cell markers (Supplementary Fig. S4). As SGK1 is a known prosurvival kinase, we tested the effect of SGK1 depletion on apoptosis (27). SGK1 knockdown by shRNA and inhibition using GSK650394, led to a significant increase in the Annexin V/PI double positive cell population as compared with the control (Fig. 6A and B). This indicates an induction of apoptosis in response to SGK1 loss. The canonical pathway for apoptosis induction is via the caspase cascade (28). Cleavage of PARP, a downstream target of active caspase-3, is also a marker for induction of apoptotic signaling cascade (29). Caspase-3 and PARP cleavage was detected exclusively in the cells with SGK1 knockdown by shRNA compared with the scrambled control, further indicating the activation of apoptosis in SGK1-depleted cells (Fig. 6C). These apoptosis markers were also detected by Western blotting of GS6-22 and MGG 8 GBM-SC lines upon treatment with the SGK1 inhibitor as well as upon SGK1 knockout by CRISPR (Fig. 6D; Supplementary Fig. S5A and S5B). Further studies are required to understand the molecular pathways involved in the role of SGK1 in GBM-SC survival. These results indicate that SGK1 functions as a critical cellular survival, antiapoptotic kinase in GBM-SCs.

GBM-SCs are important targets for drug discovery as they are relatively resistant to treatment with conventional chemo- and radiotherapy and are thought to be responsible for tumor relapse (10, 11). The enrichment of these cells in hypoxic niches further enhances these properties leading to incomplete tumor clearance and disease recurrence in the patients (30). Hypoxic conditions in the microenvironment may change the dependency of the cells on specific genes as it does to the ability of these cells to resist chemical and radiologic treatment (31). To effectively target GBM-SCs, a better understanding is needed of the genes important for their proliferation and survival under normal oxygen conditions as well as hypoxic conditions. Therefore, we have conducted an RNAi screen of GBM-SC under both these conditions.

We tested the effect of knockdown of approximately 10,000 genes on the proliferation and survival under normoxic and hypoxic conditions of two separate patient derived GBM-SC lines. This study focused on the hits common to both cell lines, under both conditions. These hits include genes that are required for key processes in cells such as transcription, translation, and proteasomal degradation. The targeting of these pathways is not ideal due to their critical functions in many normal cells, but may be effective due to the dysregulation and hyperactivation of these pathways in cancer tissues (32, 33). Comparison of the hits derived from GS6-22 and MGG 8, showed that a little over one-third of the hits were common to both cell lines (32% and 38%, respectively) under normoxic conditions. Under hypoxic conditions, only 17% of hits in GS6-22 were also identified in MGG 8 cell lines. This points to significant heterogeneity between GBMs, which is not unexpected despite a relatively small number of genes and pathways that are commonly mutated in GBM (34). These results suggest that a personalized therapeutic strategy will be needed for effective therapy of GBM rather than a common therapy for all GBMs as is now the standard of care. Moreover, as most of the hits are not in genes that are frequently mutated in GBM, our results indicate that the set of potential target genes is much larger than just the mutated genes. This suggests that the search for targeted therapeutics should move beyond targeting genes with mutations. Similar results have been found in large-scale inhibitor screens of cancer cell lines (35).

Some of the top core essential genes from the screen included Ring-box 1 (RBX1), Kinesin Family 11 (KIF11), and RAN. RBX1 is an important component of the SCF E3 ubiquitin ligase complex, which targets key cell-cycle regulators and transcription factors by ubiquitination (36). RBX1 has been found to be essential for maintaining genomic integrity and has been shown to be important for proliferation and survival of cancer cells (37). Gastric cancers with high RBX1 expression have a poorer prognosis, but its importance in GBMs remains unknown (38). Kinesin family 11 (KIF11) is a motor protein which binds to microtubules during the M-phase and anaphase of cell division and is essential for proper chromosomal segregation (39). KIF11 has been targeted in cancer cells by small-molecule compounds and work is ongoing in preclinical GBM models examining the potential of KIF11 as a therapeutic target (20). Ran (ras-related nuclear protein), functions in transport of biomolecules across the nuclear pores. This protein has been previously reported as being critical for GBM-SC survival (40). The detection of these genes along with other known essential genes provides further validation that our screen successfully identified important functional genes for GBM.

Comparison of the normoxic and hypoxic hit lists revealed that the majority (∼70%) of the genes identified as hits under hypoxic conditions were also essential under normoxia (Fig. 1D). GBM-SCs have been shown to be dependent on HIF1α and HIF2α, even under normoxic conditions (41). We postulate that this may be due to the existence of a psuedohypoxic state in the GBM-SCs, which drives the hypoxic signaling network and makes the cells sensitive to the disruption of this pathway. This pseudohypoxic state may be driven in part by the constitutive expression of HIF2α in the GS6-22 and MGG 8 cell lines (data not shown).

The hypoxia-specific hits provide further understanding of the regulation of tumor cells by the microenvironment and may provide tumor-specific targets for GBM. Various hypoxia relevant genes were enriched in the screen. PFKFB4, encodes the bifunctional enzyme PFK2/FBPase-2, is a key regulator of glycolysis and has been shown to be stimulated under hypoxic conditions (42). Interestingly, PFKFB4 was identified as a gene important for GBM-SC growth in a previous kinome and phosphatome screen (43). Hypoxia Upregulated 1 (HYOU1) is hypoxia-responsive gene that was identified as a hit in our screen. HYOU1 is upregulated in invasive breast cancers and is known to protect against hypoxia-induced cell death (44, 45). The role of selected hypoxia-specific hits is currently under investigation. However, another important implication of our results is that some drugs effective under normoxic conditions, may not be effective under hypoxia. Thus, hits which are specific to normoxic conditions only, may not be the best targets for drug development due to their potential lack of efficacy against cells under hypoxic conditions. In this study, we have focused on genes that are essential under both oxygen conditions.

Kinases regulate many important cellular processes and are important drug targets. There are more than 25 kinase inhibitors already approved as oncology drugs (46). We validated SGK1 as a key prosurvival gene in GBM-SC proliferation and survival, using shRNA, CRISPR and a SGK1-specific inhibitor. Previously, SGK1, an AGC kinase family member, is upregulated by various external stimuli, and its role in cellular survival is well known (26, 47). SGK1 is a transcriptionally induced gene in the presence of a number of different environmental stimuli including oxidative and osmotic stress and UV irradiation (47, 48). The primary role of SGK1 in tumor development is as a regulator of cell death, primarily through the regulation of the balance of pro- and antiapoptotic proteins through the FOXO3 and NF-κB cascades (26). Recent studies highlight the importance of SGK1 in GBM biology. A kinome screen identified SGK1 as a key regulator of mTOR1 in an mTOR2-dependent, but AKT- and S6 Kinase-independent manner in NF2-deficient meningiomas (49). In another kinome and phosphatome screen to identify GBM-SC vulnerabilities, SGK1 was one of the hits of the screen which significantly increased cell death as measured by PI staining, but was not validated in this study (43). Interestingly, we did not observe SGK1 dependence in traditional glioma serum grown lines as well as in differentiated GBM-SC lines. A previous study implicated SGK1 as essential in some serum grown glioma lines, but this result was based on the use of a single chemical inhibitor that could have off-target effects (25). Two large-scale RNAi screens of over 20 serum grown glioma lines using 12 to 21 shRNAs to SGK1 did not find SGK1 as an essential gene (50, 51). The differential effect of SGK1 inhibition on undifferentiated versus differentiated GBM-SCs is intriguing as it suggests that the GBM stem cells preferentially depend on SGK1 for survival. This is in contrast to AKT and PDK1 inhibition, which produces a robust growth deficit in serum glioma lines, neural stem cells as well as in differentiated GBM-SCs (data not shown). Although there is significant overlap of SGK1 targets with Akt, our results suggest that Akt does not compensate for the loss of SGK1 in GBM-SCs. This indicates that an SGK1-specific downstream target may be playing a key role in survival of this cell type. Further investigation is required to elucidate the mechanistic difference in response of these cells to SGK1 depletion.

We have found that SGK1 depletion and inhibition leads to an increase in Annexin V and PI staining measured by FACS analysis. This induction of apoptosis is further evidenced by the increase in caspase-3 and PARP cleavage. Caspase 3 is a mediator of the apoptotic signaling cascade and targets and cleaves the PARP protein. PARP cleavage is a marker for caspase dependent cell death. We detected the presence of the cleaved forms of caspase-3 and its downstream target PARP upon SGK1 protein depletion by shRNA and CRISPR as well as upon inhibition using SGK1 inhibitor GSK650394. This indicates that suppression of SGK1 function in the GBM-SCs leads to apoptotic cell death. However, in contrast to results in other cell types, we observed no change in FOXO3 and NFκB signaling upon SGK1 knockdown in these cells, and further work is required to elucidate the molecular pathways regulating SGK1′s antiapoptotic role in GBM-SCs (Supplementary Fig. S6A–S6C).

In summary, to identify new genetic targets for GBM-SC, we employed an unbiased pooled shRNA screening approach. SGK1, in addition to other potential targets, was validated as key essential gene in multiple GBM-SCs. It may be effective to target SGK1 in GBM, as its inhibition seems to be well tolerated in non-stem cell types. The refractory nature of differentiated GBM-SC and normal human fibroblast lines to SGK1 depletion may be a beneficial trait in this context, due to the reduced likelihood of it affecting normal cell types of the body. The mild phenotype of SGK1 knockout mouse indicates that it is indeed not required for survival of normal cell types though it is well known for regulated sodium reabsorption in the kidney (52). SGK1 may provide an avenue for therapeutic intervention for GBMs either as single targeting agent or in combination with current standard of care drugs such as temozolomide.

B.H. Cochran is a consultant/advisory board member for GLG Group and Smart Information Flow Technologies. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Kulkarni, B.H. Cochran

Development of methodology: S. Kulkarni, S. Goel, B.H. Cochran

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Kulkarni, S. Sengupta

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Kulkarni, S. Goel, S. Sengupta, B.H. Cochran

Writing, review, and/or revision of the manuscript: S. Kulkarni, S. Goel, S. Sengupta, B.H. Cochran

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Kulkarni, B.H. Cochran

Study supervision: S. Kulkarni, B.H. Cochran

This work was supported primarily through NIH grant 1R01NS072414. The Tufts University Flow Cytometry and Neuroscience core facilities were supported by NIH grant P30 NS047243.

The authors thank Dr. Dennis Steindler for providing us with H04, SW06 neural progenitor cell lines and Dr. Haraoki Wakimoto for the MGG 8 cell line used in this article. The authors would also like to acknowledge the support and guidance from the Tufts University Flow Cytometry and Neuroscience core facilities for FACS and qPCR experiments, respectively.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.