Drak/STK17A Drives Neoplastic Glial Proliferation through Modulation of MRLC Signaling

Glioblastomas (GBM), the most common primary malignant brain tumors, infiltrate the brain, grow rapidly, and are resistant to current therapies (1). Low-grade gliomas (LGG), which are related infiltrative malignant neural neoplasms, have slower tumor growth rates, longer patient survival, and display more variable responses to therapeutics (2). To understand their genesis, these tumors have been subject to extensive genomic analyses, which show that signature genetic lesions in LGG and GBM include copy gain and amplification of chromosome 7, amplification, mutation, and/or overexpression of receptor tyrosine kinases (RTK), such as EGFR, and activating mutations in components of the PI3K pathway (1, 3, 4). Nearly 60% of GBMs show focal EGFR copy gain or amplification, which are often accompanied by gain-of-function EGFR mutations (4). The most prevalent EGFR mutant variant in GBM is EGFR^VIII (5), in which exon 2 to 7 deletion confers constitutive kinase activity, which potently drives tumorigenesis (1, 6). The most frequent PI3K pathway mutation in gliomas is loss of PTEN lipid phosphatase (4), which results in unopposed PI3K signaling.

Recent mouse models demonstrate that co-activation of EGFR and PI3K pathways in glial cells or neuro-glial stems cells induces GBM-like tumors, although these tumors do not show the full range of GBM phenotypes (7–9). Furthermore, to date, pharmacologic inhibitors of EGFR and PI3K pathway components are ineffective in improving LGG and GBM outcomes (10). Genomic studies indicate that LGGs and GBMs have other genomic alterations (3–5); however, it is unknown how these changes contribute to gliomagenesis. Taken collectively, these data suggest that there are still undiscovered biological factors that drive tumorigenesis. Given the aggressive nature of these tumors, there is a pressing need to better understand their biology and to identify additional factors that could serve as new drug therapy targets.

To investigate the biology of malignant gliomas, we developed models in Drosophila melanogaster (11). Drosophila offers unique advantages for modeling gliomas: flies have orthologs for 70% of human genes, including most known gliomagenic genes (12); Drosophila neural and glial cell types are homologous to their human counterparts (13); and versatile genetic tools are available for in vivo cell-type specific gene manipulation including RNAi (14). Although Drosophila models cannot address all aspects of glioma biology, our model demonstrates that constitutive activation of EGFR and PI3K signaling in glial progenitor cells gives rise to malignant glial tumors that recapitulate key biological features of human gliomas (11).

To discover new pathways that contribute to EGFR–PI3K–mediated glioma, we performed a kinome-wide RNAi-based genetic screen in our EGFR–PI3K Drosophila GBM model (15). Kinases were screened because they are highly conserved in terms of protein function between Drosophila and mammalian systems. One of the top candidates from this screen was death-associated protein kinase related (Drak; ref. 15). Drak and its human ortholog, STK17A, are cytoplasmic serine/threonine kinases in the Drak subfamily of cytoplasmic death-associated protein (DAP) kinases, which regulate cytoskeletal dynamics, cytokinesis, and cell adhesion and mobility (16–18). In Drosophila development, Drak promotes epithelial morphogenesis, acting downstream of Rho-GTPase signaling to control the actin cytoskeleton through phosphorylation of Spaghetti Squash (Sqh; refs. 18, 19). MRLC, the human Sqh ortholog, is a regulator of nonmuscle myosin type II (NMII) motor proteins, and phosphorylated MRLC binds to NMII and stimulates NMII-dependent contractile activity to promote cytoskeletal re-organization, morphogenesis, and cytokinesis (20, 21). MRLC phosphorylation is dynamic and tightly regulated, which allows for precise temporal and spatial changes to the cytoskeleton (20). Like other Drak family kinases, STK17A, which is expressed in GBM (22), can directly phosphorylate purified MRLC protein in vitro (18, 23). However, little is known about the mechanisms by which Drak/STK17A promotes tumorigenesis.

Here we characterize genetic and functional requirements for Drak in EGFR- and PI3K-driven neoplastic glia. We report that Drak operates downstream and in concert with EGFR signaling to phosphorylate and activate Sqh to drive proliferation of neoplastic glia. Moreover, our data show that, in mitotic neoplastic glial cells undergoing cytokinesis, phosphorylated Sqh colocalizes with Anillin, a cytoskeletal protein and known Sqh binding partner (24), and that Anillin is required for proliferation of neoplastic glia. We show that these interactions are conserved in human gliomas. STK17A is overexpressed in primary human GBM and LGG tumors and patient-derived GBM cell cultures, and elevated STK17A expression correlates with elevated EGFR, MRLC phosphorylation, and ANLN (human Anillin ortholog) levels. STK17A activity is required for tumor cell proliferation and for elevated levels of MRLC phosphorylation and ANLN. Moreover, we found that STK17A, phosphorylated MRLC, and ANLN are all colocalized to the cleavage furrow in tumor cells undergoing cytokinesis. Taken together, our data suggest that Drak/STK17A potentiates EGFR signaling to drive activation of Sqh/MRLC, which in turn regulates mitosis in gliomas through effects on Anillin/ANLN.

Drosophila stocks were obtained from the Bloomington stock center and VDRC and the specific genotypes and stocks used are listed in the Supplementary Tables S1 and S2. Drak^null(Drak^del) was a gift of David Hipfner (18). UAS-sqh^D20D21 was a gift of Guang-Chao Chen (25). UAS- dEGFR^λ was a gift of Trudi Schupbach. All stocks were cultured on standard corn meal molasses food at 25°C. Prior to publication, the UAS-Drak^dsRNA stocks were validated by PCR amplification of the dsRNA element followed by sequence validation against the published sequence available at the VDRC. To create UAS-Drak and UAS-Drak^KD (kinase dead) constructs, the RE12147 Drak cDNA was cloned into pUAS-T, site directed mutagenesis was used to convert Lys-66 to Ala, and the resulting DNA was injected into embryos and stocks for each construct were established and sequence verified. All genotypes were established by standard genetic crossing. To normalize for UAS transgene dosage in repo>dEGFR^λ;dp110^CAAX animals, a UAS-lacZ transgene was included in control genotypes.

Larval brains were dissected with forceps, fixed in 4% paraformaldehyde, processed, stained, and imaged as previously described (11). The following antibodies were used: 8D12 mouse anti-Repo (1:10, Developmental Studies Hybridoma Bank), anti-phospho-S21-Sqh (1:500, gift of Robert Ward; ref. 26), anti-Anillin (1:500, gift of Maria Giansanti; ref. 27). Secondary antibodies were conjugated to Cy3 (1:150), Alexa Fluor 488, or Alexa Fluor 647 (1:100; Jackson Laboratories). Brains were mounted on glass slides ventral side down in vectashield and whole mount imaged on a Zeiss LSM 700 confocal system. For experiments where protein levels were compared between genotypes, all samples were prepared, subjected to IHC, imaged, and image processed in a parallel manner side by side. Six or more brains were stained with each Ab combination, and representative images are shown for each result. All brain phenotypes shown were highly penetrant, with approximately 75% to 100% of animals showing the growth phenotypes described. Images were analyzed in Zeiss Zen Software and processed in Photoshop. Larval Drosophila brain hemisphere volumes were analyzed using Imaris software. Larval glial cells were counted manually in representative optical sections of age-matched brain hemispheres, matched for section plane. Statistical analyses were done using Prism.

GBM39, shared by C. David James, was created from human GBMs serially xenografted. GBM157, GBM281, GBM301, and GBM309 gliomasphere cultures were derived at UCLA and maintained in culture as described (28). HNPCs were obtained from Lonza. U87 and U87-EGFR^vIII cells were gifts of Frank Furnari. Cell culture was performed as previously described (15). Lentiviral shRNAs were prepared and used as previously described on serum cultured cells and adherent serum-free cultured gliomasphere lines (15). WST-1 assays on shRNA-treated cells were performed as previously described (15). For immunofluorescence, U87-EGFR^vIII cells were plated on glass coverslips, fixed with 4% paraformaldehyde, and stained on the coverslips with anti-β-tubulin (1:25, Developmental Studies Hybridoma Bank, AA4.3), anti-STK17A (1:1,000, Sigma, HPA037979), anti-ANLN (1:100, Sigma, HPA005680), anti-phospho-S19-MRLC (1:200, Abcam, ab2480; note this antibody detects S19 according to our sequence alignment and not S20), and/or DRAQ7 (Cell Signaling Technology, 1:200) to stain nuclear DNA and chromosomes.

Cultured cells were collected and washed with 1× PBS and lysed in RIPA buffer containing protease and phosphatase inhibitors. Whole third instar larval brains were dissected were washed with 1× PBS and lysed in RIPA buffer containing protease and phosphatase inhibitors. The following antibodies were used for immunoblotting following the manufacturer's recommendations: anti-STK17A (1:1,000, Sigma, HPA037979), anti-phospho-S19-MRLC (1:500, Cell Signaling Technology, 3671), anti-phospho-S19-MRLC (1:200, Abcam, ab2480), anti-MRLC (1:500, Cell Signaling, 3672), anti-EGFR (1:5000, BD), anti-ANLN (1:500, Sigma, HPA005680), and anti-actin (1:200, Developmental Studies Hybridoma Bank, JLA20). Bands were quantified using ImageJ.

STK17A mRNA expression data were obtained from cBioportal (www.cbioportal.org) and exacted and analyzed using RStudio. All graphs and statistics were generated from this data using Prism (29, 30). The in silico analysis results shown here are based solely upon data generated by the The Cancer Genome Atlas (TCGA) Research Network (http://cancergenome.nih.gov; refs. 2–4).

All human tumor specimens were collected from surgical specimens donated for research with written informed consent of patients and were collected and used according to recognized ethical guidelines (Declaration of Helsinki, CIOMS, Belmont Report, GCP, Nuremberg Code) in a protocol (IRB00045732) approved by the Institutional Review Board at Emory University. Paraffin-embedded human brain tumor specimens and tumor tissue microarrays (TMA) with matched control tissue were prepared and sectioned using the Winship Core Pathology Laboratory. Antigen retrieval and IHC staining was performed as specified by manufacturer's guidelines for each specific antibody (15). The following antibodies were used: anti-STK17A (1:1,000, Sigma HPA037979), anti-ANLN (1:50, Sigma, HPA005680), anti-phospho-S19-MRLC (1:200, Abcam, ab2480), anti-EGFR (1:50, Cell Signaling Technology, 4267). Results were scored by neuropathologists according to standard clinical criteria on a scale of 1 and 2 (low staining), 3 or 4 (high staining, with 4 being more uniform), and images of immunoreactivity were taken on an Olympus DP72 CCD camera.

Comparisons between two groups were performed by Mann–Whitney U test (nonparametric) for TCGA RNA-seq data analyses. Comparisons between of three or more groups were performed by one-way ANOVA with multiple comparisons to experimental controls. Comparisons between two groups were done using either paired or unpaired parametric t tests. Fisher exact test was used to analyze correlations in IHC data.

To understand GBM pathology, we developed a Drosophila GBM model. The model uses the glial-specific repo-Gal4 transcriptional driver to co-overexpress constitutively active versions of dEGFR (dEGFR^λ) and dp110 (dp110^CAAX), the catalytic subunit of PI3K, that together drive malignant transformation of postembryonic larval glia (11). The resulting glial tumors exhibit phenotypic and molecular characteristics similar to human GBM (11). To identify novel modifiers of EGFR–PI3K–driven glial neoplasia, we used our Drosophila GBM model in a genetic screen, and identified Drak, which encodes the sole Drosophila ortholog of STK17A and STK17B serine/threonine kinases (15).

Glial-specific Drak RNAi reduced neoplastic glial proliferation and altered glial morphogenesis, with Drak^dsRNA#1 yielding significantly reduced brain sizes and glial cell numbers compared with dEGFR^λ-dp110^CAAX (Fig. 1A–I). The transforming effects of EGFR–PI3K signaling were also reduced in Drak null mutants (Drak^null; ref. 18) or by co-overexpression of kinase-dead Drak (Drak^KD): such mutants showed near wild-type brain sizes and reduced glial cell numbers compared with dEGFR^λ-dp110^CAAX controls (Fig. 1A–I), indicating that Drak catalytic activity is essential for proliferation of dEGFR^λ-dp110^CAAX-mutant glia.

Growth inhibition of neoplastic glia induced by Drak knockdown or loss-of-function was not due to nonspecific glial lethality: Drak is a nonessential gene and homozygous null mutants are viable and show normal brain morphology and glial development, and Drak RNAi in wild-type larval glia caused no obvious defects (Supplementary Fig. S1A and S1B; refs. 15, 18). Thus, although Drak is not required for normal glial proliferation and development, Drak kinase activity is essential for neoplastic glial proliferation.

Because reduced Drak function had a dramatic effect on dEGFR^λ-dp110^CAAX mutant neoplastic glia, and because Drak functions downstream of EGFR in epithelia (19), we predicted that Drak may function downstream of EGFR in neoplastic glia. Overexpression of constitutively active dEGFR^λ alone elicits hyperplasia in glia (11), which we found was suppressed by loss of Drak function (Supplementary Fig. S1C and S1D), consistent with our prediction.

Previous studies show that, in developing epithelia, Drak acts downstream of EGFR and RhoGTPase (RhoA) signaling in parallel with Rho kinase (Rok; refs. 18, 19). To investigate whether RhoGTPases are also required in neoplastic glia, we tested RhoA RNAi and dominant negative constructs in our GBM Drosophila model. We found that RhoA loss-of-function caused a significant reduction in dEGFR^λ-dp110^CAAX mutant glial proliferation, but did not obviously affect wild-type glia proliferation (Fig. 1A–I; Supplementary Table S1). Glial-specific inhibition of Rho family GTPases Cdc42 and RhoL and glial-specific Rok RNAi did not as strongly suppress neoplastic glial proliferation (Supplementary Table S1; ref. 15). Thus, Drak may phosphorylate substrates to drive neoplastic glial proliferation downstream of RhoA GTPase signaling, independent of Rok. Together, our data demonstrate that Drak pathways are necessary for EGFR–PI3K–dependent glial neoplasia.

Because Drak reduction suppressed glial neoplasia in the context of constitutive EGFR–PI3K and EGFR signaling, we tested whether Drak overexpression cooperates with constitutive EGFR signaling. We found that glial-specific Drak overexpression had no obvious effect on normal larval glia morphology or number compared with wild-type (Fig. 2A–C). Glial-specific dEGFR^λ overexpression induced a significant increase in glial cell numbers compared with wild-type controls (Fig. 2A–D; ref. 11). Co-overexpression of Drak and dEGFR^λ increased glial cell numbers, and these glia lost their normal stellate shape and formed abnormal cellular aggregates that disrupted normal larval brain architecture, indicative of neoplastic transformation (Fig. 2A–E). We previously showed that overexpression of human EGFR^vIII (hEGFR^vIII), an oncogenic constitutively active mutant variant of EGFR found in GBM, cooperates with dp110^CAAX to drive neoplastic glial transformation, which was suppressed by Drak RNAi (15). Similar to dEGFR^λ, hEGFR^vIII overexpression caused glial hyperplasia (15), and Drak and hEGFR^vIII co-overexpression cooperated to drive increased proliferation and alterations in larval glial morphology consistent with neoplastic transformation (Supplementary Fig. S2A–S2D). Thus, although Drak overexpression alone has little effect in glia, Drak co-overexpression augments the ability of constitutively active EGFR to promote tumorigenesis. Thus, Drak acts as a bona fide genetic modifier in that Drak loss or gain exerts the observed effects only in the context of other oncogenic mutations.

We next sought to understand the mechanism whereby Drak contributes to glial transformation. Given that Drak catalytic activity was essential for EGFR–PI3K glial neoplasia, we predicted that Sqh, which is the Drosophila ortholog of human NMII regulatory light chain (MRLC) and only known Drak substrate (18), would be essential for EGFR–PI3K glial neoplasia. Prior studies have shown that DAP family kinases, including STK17A, the human ortholog of Drak, phosphorylate MRLC at serine and threonine residues (Thr-18 and Ser-19; refs. 18, 21, 23). Similarly, Drak phosphorylates Sqh at the conserved serine residue (Ser-21 in Sqh is equivalent to Ser-19 in MRLC; refs. 18, 21, 23). Phosphorylated Sqh binds to and stimulates the ATPase-dependent motor activity of Zipper, the sole NMII ortholog, which functions in cellular processes that require cytoskeletal contractility, such as cell migration, cytokinesis, and morphogenesis; all processes that play pivotal roles in tumorigenesis (21, 31, 32).

We found that, similar to Drak knockdown, glial-specific sqh knockdown significantly reduced glial cell numbers and rescued brain size in dEGFR^λ-dp110^CAAX mutants (Fig. 3A–F), which suggests that Drak activates Sqh by phosphorylation in transformed glia. To test this hypothesis, we used a validated phospho-specific antibody to examine levels of Ser-21-phosphorylated Sqh (Sqh-S21-P) protein in dEGFR^λ-dp110^CAAX mutant glia in the presence or absence of Drak (26). dEGFR^λ-dp110^CAAX mutant glia showed increased Sqh-S21-P compared with wild-type glia, with mitotic cells showing cortical enrichment of Sqh-S21-P (Fig. 3G), consistent with published observations regarding Sqh phosphorylation during cytokinesis (27). Drak RNAi reduced Sqh-S21-P levels (Fig. 3G), indicative of loss of Sqh activation.

We next asked whether concurrent activation of EGFR and Drak signaling influences levels of activated Sqh. Consistent with Drak activity to phosphorylate Sqh, we observed increased levels of Sqh-S21-P in dEGFR^λ;Drak^OE glia compared with Drak-over expressing glia (Fig. 3G). By Western blot analysis, we observed increased levels of Sqh-S21-P with Drak^OE;hEGFR^vIII glia compared with hEGFR^vIII-overexpressing glia (Supplementary Fig. S2E). These data are consistent with a model in which Drak increases the amount of activated, phosphorylated Sqh, which in turn supports EGFR-dependent tumorigenesis.

If Sqh phosphorylation promotes EGFR-dependent neoplasia, then overexpression of a phospho-mimetic version of Sqh should enhance oncogenic EGFR. A prior study engineered a Sqh transgene with serine-21 and threonine-20 converted to aspartic acid (Sqh^D20D21) to mimic the phosphorylated state (33). As predicted, Sqh^D20D21 and dEGFR^λ co-overexpression increased numbers of small proliferative glia that disturbed normal larval brain architecture; these changes are indicative of enhanced hyperplasia and/or neoplastic transformation (Fig. 4A–E). Thus, Sqh is a functionally relevant Drak substrate in EGFR–PI3K–mediated glial neoplasia. Taken together, our data provide strong support for a model in which Drak cooperates with oncogenic EGFR signaling to create and sustain a pool of activated, phosphorylated Sqh necessary for glial cell transformation.

Following determination that glial neoplasia in our GBM model requires Sqh, we next sought to determine which processes downstream of Sqh are essential in neoplastic glia. We used our Drosophila GBM model to test RNAi constructs against eight published Sqh binding partners (Supplementary Table S2). We found that glial-specific RNAi of anillin, which encodes a well-established Sqh binding partner and multifunctional actin-binding scaffolding protein important for cytoskeletal reorganization during cytokinesis (24), significantly reduced dEGFR^λ-dp110^CAAX mutant glia proliferation (Fig. 5A–F). Thus, anillin RNAi in glial cells may interfere with cytokinesis, thereby inhibiting tumor cell proliferation.

To determine if Anillin is an effector of dEGFR–Drak–Sqh signaling, we tested glial-specific anillin knockdown in either dEGFR^λ;Drak^OE or dEGFR^λ;sqh^D20D21 mutant glia, and observed a significant reduction in glial proliferation (Supplementary Fig. S3A–S3G), indicating that Anillin operates downstream of dEGFR–Drak–Sqh signaling. In contrast, anillin knockdown in wild-type glia had no impact on glial cell proliferation compared with wild-type controls (Supplementary Fig. S3A–S3G), showing a differential requirement for the Drak–Sqh–Anillin pathway in neoplastic glia, but not wild-type glia.

MRLC phosphorylation mediates cellular processes that require NMII-dependent cytoskeletal contractility, including mitosis and cytokinesis (21, 31, 32). Previous studies demonstrate that phosphorylated Sqh recruits Anillin to the cortex during mitosis, where it coordinates cytokinesis by linking actin and NMII/Zipper proteins in the contractile ring (24). During embryonic cellularization, Drak phosphorylation of Sqh is responsible for proper organization of contractile rings (34), most likely because Sqh phosphorylation was necessary for binding of Anillin to Zipper. To determine if Drak-dependent phosphorylation of Sqh is necessary for Anillin binding and/or localization in mitosis, we examined Anillin localization in relation to Sqh-S21-P in neoplastic dEGFR^λ-dp110^CAAX glia. Consistent with their binding in mitotic glia, Sqh-S21-P and Anillin were colocalized and enriched at the cortex and cleavage furrow (observed in mitotic cells in all dEGFR^λ;dp110^CAAX brains imaged, n = 8), and this enrichment was lost upon Drak depletion (in all dEGFR^λ;dp110^CAAX;Drak^dsRNA#1 brains imaged, n = 6; Fig. 5G). Moreover, overall Anillin levels were reduced by Drak depletion (Fig. 5G). Thus, our data support a model wherein Drak-dependent phosphorylation of Sqh promotes Anillin binding to coordinately drive cytokinesis and proliferation in neoplastic glia.

To determine whether the Drak–Sqh–Anillin pathway operates in human GBM and/or LGGs, we examined expression and function of the human orthologs of Drak, STK17A, and STK17B, and found that STK17A is overexpressed in GBMs (15). We next examined STK17A levels in a panel of patient-derived human GBM stem-cell containing gliomasphere (GSC) cultures. Compared with cultured normal human neural progenitor cells (HNPC), EGFR^vIII-positive and EGFR-mutant GSC cultures express higher levels of STK17A and ANLN (Fig. 6A). Moreover, in GSC cultures with high STK17A levels, we also saw a modest increase in Ser-19-phosphorylated MRLC (MRLC-S19-P) relative to HNPCs (Fig. 6A). Thus, at the protein level, we observed that the relationships between EGFR, STK17A, ANLN (Anillin), and MRLC phosphorylation in GBM cells recapitulate our observations from Drosophila.

To further explore pathway conservation, we examined STK17A function and localization in serum-cultured GBM cell lines and GSC lines using RNAi and immunofluorescence. Consistent with prior reports (22), we found that, in serum cultured lines, STK17A is required for proliferation, with STK17A knockdown inducing slower proliferation, apoptosis, reduced MRLC-S19-P levels, and altered cell shape and adhesion (Fig. 6B; Supplementary Fig. S4A and S4B), which is consistent with alterations in MRLC regulation (35–37). We examined the effects of STK17A loss in GSCs treated with or without ZVAD to control for the effects of apoptosis, and we found that STK17A knockdown caused GSC adhesion defects, slower proliferation, apoptosis, and reduced levels of MRLC-S19-P and ANLN levels relative to control GSCs (Fig. 6C and D; Supplementary Fig. S4C and S4D). We also observed that total MRLC levels were reduced upon STK17A knockdown, suggesting that phosphorylation may regulate total MRLC protein in GBM cells. This is consistent with published studies showing MRLC levels are regulated by proteosomal turn-over (38). Furthermore, we examined STK17A localization in EGFR^vIII-positive GBM cells and observed that STK17A protein, in conjunction with MRLC-S19-P and ANLN, was upregulated in mitotic cells and localized to the cleavage furrow in tumor cells undergoing cytokinesis (Fig. 6E and F). Thus, STK17A expression is required to promote MRLC phosphorylation and ANLN upregulation in GBM cells to coordinately regulate cytokinesis and proliferation.

We used IHC to examine protein expression of STK17A, EGFR, MRLC-S19-P, and ANLN in a collection of graded human tumor specimens. In our LGG TMA, specimens with high STK17A expression showed a statistically significant correlation with high EGFR, MRLC-S19-P, and ANLN expression (Fig. 7A). In our GBM TMA, we observed high expression of EGFR, STK17A, MRLC-S19-P, and ANLN in the majority of specimens (Fig. 7B). However, we observed some GBM specimens with high STK17A but low MRLC-S19-P expression (Fig. 7B): in these surgical GBM specimens, it is possible that the MRLC-S19-P phospho-epitope was not properly fixed and preserved. Thus, in human tumors, elevated STK17A levels co-occur with elevated EGFR, MRLC-S19-P, and ANLN levels, which recapitulates the relationship we observed between EGFR, Drak, Sqh, and Anillin in Drosophila.

We used cBioportal to process genomic data catalogued by TCGA to assess prevalence of STK17A mRNA expression and copy gain alterations and to examine relationships between STK17A alterations and well-characterized genetic lesions in gliomas (2–4, 29, 30). Well-characterized lesions include full or partial amplification of chromosome 7, which includes regions encoding both EGFR and STK17A (39), and focal EGFR amplification and mutation. In TCGA cohorts of LGGs and GBMs, STK17A mRNA expression was significantly correlated with copy number gain in 13p on chromosome 7 (7p13; Fig. 7C and D). To understand whether STK17A mRNA overexpression is specific or is passively driven by copy gain, we examined mRNA expression of neighboring genes on chromosome 7. We found that genes in close proximity to STK17A (i.e., NACAD) showed no statistically significant difference in mRNA expression between LGG specimens with chromosome 7 copy gain compared with LGG specimens with no observable copy number alterations in the same region of chromosome 7 (Fig. 7E), indicating that STK17A mRNA expression may be selectively upregulated. Thus, increased STK17A expression may have a specific role in glioma pathology.

To further assess whether STK17A is a driver of gliomagenesis, we used cBioportal to examine STK17A mRNA expression relative to IDH1 status in patients with LGG. IDH1 mutation is a common genetic alteration in LGG, and patients who harbor IDH1 mutations have a better prognosis than those with wild-type IDH1 (40–42). LGG patients with wild-type IDH1 have more aggressive tumors that behave much like primary GBMs (2, 40). Furthermore, wild-type IDH1, not mutant IDH1, is typically found in gliomas with chromosome 7 alterations (2, 40). In LGG, we found elevated STK17A mRNA expression in tumors with wild-type IDH1 compared with tumors with mutant IDH1 (Fig. 7F). In IDH1 mutant LGGs, there was not a statistically significant difference between STK17A mRNA expression between tumors with or without chromosome 7 gain (Fig. 7G), suggesting that STK17A levels are not as relevant to progression in IDH1 mutant tumors. Our observations are consistent with a previous study showing STK17A mRNA overexpression in LGGs is correlated with disease severity and worse prognosis (22).

To investigate associations between STK17A expression and patient survival, we analyzed cBioportal TCGA data to find that LGG and GBM patients with at least two-fold STK17A copy gain showed worse overall survival compared with patients with no STK17A copy gain (Supplementary Fig. S5A and S5B). Thus, LGG and GBM patients with STK17A copy gain have a worse overall prognosis.

Together, our results suggest that elevated STK17A expression drives MRLC- and ANLN-dependent cytoskeletal changes during mitosis and cytokinesis to facilitate disease progression (Fig. 7H), validating our identification of the STK17A ortholog Drak as a driver of tumorigenesis in our Drosophila GBM model.

Although EGFR and PI3K pathways play important roles in glioma progression and maintenance, effective therapies targeting these pathways remain elusive (10). We developed a Drosophila melanogaster GBM model based on co-activation of EGFR and PI3K in glia in order to gain insight into genetic and cellular mechanisms underlying gliomas (11, 15). Using our system, we identified a pathway through which the cytoplasmic serine/threonine kinase Drak specifically drives neoplastic proliferation. Consistent with published results (22), we show that the orthologous kinase STK17A drives proliferation human GBM cells, through a conserved pathway that regulates cytokinesis.

During Drosophila development, Drak acts downstream of EGFR and Rho-GTPase signaling to regulate epithelial tissue morphogenesis through Sqh phosphorylation (18, 19). Activated Sqh, like human MRLC, modulates cytoskeletal reorganization in cellular processes such as cytokinesis (32, 33, 43), which is fundamental to cancer progression (44). Drak harbors latent oncogenic activity, as it can cooperate with constitutively active EGFR to stimulate glial transformation. Similarly, in human LGG and GBM, STK17A is frequently subject to copy gain and overexpression in association with EGFR and chromosome 7 alterations. We show that, in Drosophila and human tumor cells, Sqh/MRLC is a key mediator of Drak/STK17A: Sqh/MRLC is phosphorylated at equivalent conserved sites in EGFR–PI3K mutant tumor cells in a Drak/STK17A-dependent manner, and is necessary for neoplastic growth. Thus, Sqh/MRLC is a functionally relevant and evolutionarily conserved substrate of Drak/STK17A in the context of EGFR–PI3K–driven glial tumorigenesis. This corroborates studies that have shown MRLC hyper-phosphorylation occurs in GBM tumors and that targeted inhibition of MRLC activity inhibits GBM growth and invasion (35–37, 45, 46). Together, our results establish STK17A as a disease-relevant MRLC kinase in gliomas.

MRLC phosphorylation regulates many cellular processes that require NMII-dependent contractility (21, 31, 32). To distinguish which of these processes are most relevant to Drak/STK17A function, we used a genetic approach and found that, among known Sqh binding partners, Anillin is essential in neoplastic glia. Anillin is a scaffolding protein that acts downstream of RhoGTPase to organize the cytoskeleton and contractile ring in cytokinesis (24, 47–50). As part of this functionality, Anillin binds to Zipper (NMII) when Sqh is phosphorylated and activated (24). Drak-dependent activation of Sqh promotes Anillin binding to Zipper and organization into contractile rings during cytokinesis (32, 34). We observed that phosphorylated Sqh and Anillin colocalize in a Drak-dependent manner at the cleavage furrow in EGFR–PI3K mutant glia undergoing cytokinesis. Similarly, we observed that elevated levels of phosphorylated MRLC, ANLN, and STK17A co-occur in human gliomas, and that these proteins colocalize at the cleavage furrow in EGFR-mutant GBM cells undergoing cytokinesis. Our results suggest that the primary role of Drak/STK17A in neoplastic glia is to promote cytokinesis to drive proliferation.

Two other MRLC kinases, MLCK and ROCK, and two different NMII isoforms, including NMIIA and NMIIB, regulate GBM cell migration (35–37, 45, 46). Previous studies also show that STK17A promotes GBM cell migration in vitro (22). Therefore, we did not study STK17A function in tumor cell invasion; instead, given the effects of Drak loss, we focused on Drak/STK17A function in GBM proliferation and cytokinesis. While regulation of cytokinesis in glioma cells is not well understood, our results imply that cytokinesis in glioma cells is differentially regulated relative to normal developing glia or neural stem cells. In other tumor cell types, cytokinesis is preferentially controlled by NMIIC, which is also expressed in GBM cells (45), and may therefore mediate STK17A function. Further work is needed to determine mechanisms of Drak/STK17A-dependent regulation of cytokinesis, and whether defective cytokinesis actively provokes growth arrest and apoptosis in GBM cells.

Although Drak overexpression alone causes no phenotype, Drak overexpression intensified the proliferative output of EGFR and PI3K signaling pathways through Sqh. Given that Drak modifies glial EGFR–PI3K–driven neoplasia but does not affect normal glial development, Drak and STK17A may require other signaling outputs downstream of EGFR or PI3K to drive proliferation. This is consistent with known requirements for RTK and PI3K activity in cytokinesis and other NMII-dependent processes (32). For example, previous reports indicate that increased phosphorylation of NMII occurs in GBM cells in response to EGFR signaling (37). Thus, perhaps EGFR-dependent differential phosphorylation and activation of NMII underlies the mechanism by which Drak-Sqh and EGFR cooperate to drive tumorigenesis, and the lack of NMII phosphorylation underlies the inability of Drak to promote glial proliferation when overexpressed alone. Further studies are required to explore mechanisms by which Drak/STK17A cooperates with EGFR activation to promote tumorigenesis.

In summary, our data validate use of invertebrate model organisms as a means to elucidate new aspects of glioma biology. Our research reveals that Drak/STK17A dependency may provide a molecular vulnerability and therapeutically relevant target for GBM and LGG.

No potential conflicts of interest were disclosed.

Conception and design: A.S. Chen, J. Wardwell-Ozgo, R.D. Read

Development of methodology: A.S. Chen, J. Wardwell-Ozgo, H.I. Kornblum, R.D. Read

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.S. Chen, J. Wardwell-Ozgo, N.N. Shah, D. Wright, K. Vigneswaran, D.J. Brat, R.D. Read

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.S. Chen, J. Wardwell-Ozgo, C.L. Appin, D.J. Brat, R.D. Read

Writing, review, and/or revision of the manuscript: A.S. Chen, J. Wardwell-Ozgo, N.N. Shah, K. Vigneswaran, D.J. Brat, H.I. Kornblum, R.D. Read

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.S. Chen, J. Wardwell-Ozgo, R.D. Read

Study supervision: J. Wardwell-Ozgo, R.D. Read

We thank Tim Fenton, Clay Coston Rowe, Colleen Mosley, and Hye Rim Kim for technical assistance, and Ken Moberg for critical reading of the manuscript. This work was supported by grants from the NIH/NINDS (NS065974), the Southeastern Brain Tumor Foundation, and Emory University Research Committee to R.D. Read, and a K12-IRACDA career development award from the NIH/NIGMS (GM000680) to J. Wardwell-Ozgo.

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.