OLIG2-STAT5 cooperation supports glioma stem cell maintenance and proliferation in glioblastoma

Abstract

Background

Glioblastoma (GBM) is driven, in part, by glioma stem cells (GSCs), which sustain therapeutic resistance and recurrence. OLIG2 is a key regulator of GSC programs but the signaling pathways that cooperate with Oligodendrocyte transcription factor 2 (OLIG2) to maintain GSC states remain incompletely defined. Although Signal transducer and activator of transcription (STAT) signaling has been implicated in GSC biology, work to date has focused primarily on STAT3, and the contribution of STAT5 remains poorly defined. Here, we investigated whether STAT5B functionally cooperates with OLIG2 in sustaining GSC phenotypes.

Methods

Transcriptomic analyses, proximity ligation assays and biochemical studies were utilized to define OLIG2-STAT5B interactions in patient-derived GSCs. Functional consequences of OLIG2, STAT5B, BIRC5, and NUSAP1 suppression were assessed using small interfering RNA/short hairpin RNA-knockdown, stemness, and migration assays in EGFRvIII-expressing GSCs. In vivo tumor growth studies were performed using orthotopic xenograft models.

Results

OLIG2 and STAT5B expression were positively correlated in GBM datasets and patient-derived models. OLIG2 physically associated with STAT5B in EGFRvIII-expressing GSCs, and loss of OLIG2 reduced STAT5 activation, supporting a functional link between these factors. OLIG2 or STAT5B suppression induced convergent transcriptional changes, including downregulation of genes associated with GSC maintenance and proliferation, notably stemness regulators BIRC5 and NUSAP1. Silencing BIRC5 or NUSAP1 impaired GSC self-renewal and migration. In orthotopic models, STAT5B knockdown prolonged survival and phenocopied OLIG2 suppression and stemness loss.

Conclusions

These findings identify STAT5B as a functional partner of OLIG2 in GBM and define anOLIG2-STAT5B transcriptional axis that sustains GSC maintenance and malignant behavior. Targeting this axis may represent a therapeutic strategy to disrupt treatment-refractory glioma cell states.

Key Points.

• OLIG2 and STAT5B form a novel complex that sustains GSC phenotypes.

• OLIG2-STAT5B transcriptional axis is critical for GSC maintenance and tumor progression.

Importance of the Study

GSCs are key drivers of glioblastoma progression, therapeutic resistance, and recurrence. Despite extensive characterization of individual transcriptional regulators in glioma biology, the cooperative networks governing GSC maintenance remain poorly defined. We report the discovery of a previously unrecognized OLIG2–STAT5B transcriptional axis that regulates GSC identity and malignant behavior. This regulatory complex coordinates expression of critical downstream targets, including stemness factor BIRC5 and mitotic regulator NUSAP1. Genetic disruption of this axis impairs stemness phenotypes and extends survival in xenograft models. Our findings illustrate how developmental transcription factors and cytokine signaling pathways converge to sustain treatment-resistant populations in GBM tumors to promote GBM recurrence. This work establishes the OLIG2-STAT5B axis to be active in EGFRvIII-expressing tumors and provides a mechanistic framework for developing GSC-directed therapeutic strategies to overcome resistance in aggressive gliomas.

Glioblastoma (GBM) is the most common and lethal primary brain tumor in adults, with a median survival of approximately 15 months despite aggressive treatment that includes surgical resection, radiation, and chemotherapy.^1,2 A major clinical challenge of GBM is its diffuse infiltration into adjacent brain tissue, which renders complete surgical resection impossible and contributes to inevitable tumor recurrence.³ Recurrent tumors are uniformly fatal and exhibit heightened resistance to therapy, highlighting the urgent need to identify molecular drivers of recurrence and therapeutic resistance.

GBM progression and resistance are driven in part by a subpopulation of tumor-initiating cells known as glioma stem cells (GSCs).^4,5 GSCs are capable of self-renewal, differentiation, and repopulation of the tumor bulk after treatment.^6,7 Unlike differentiated tumor cells, GSCs are inherently resistant to cytotoxic therapies and are thought to be the primary source of relapse.³ Transcriptional profiling has identified several key neurodevelopmental regulators that maintain GSC identity and function, including the basic helix-loop-helix transcription factor Oligodendrocyte transcription factor 2 (OLIG2). In normal development, OLIG2 regulates neural progenitor proliferation and cell fate, promoting oligodendrocyte or neuronal differentiation.^8,9 In GBM, OLIG2 promotes GSC maintenance, tumor initiation, and invasiveness.^10–12 OLIG2 is also one of the key factors required for reprogramming of differentiated glioma cells into tumor-forming GSCs.¹³ However, how OLIG2 integrates with other signaling pathways to regulate GSC behavior remains poorly defined.

Signal transducer and activator of transcription (STAT) proteins are key transcriptional effectors that mediate cellular responses to cytokines, growth factors, and oncogenic stimuli.¹⁴ Upon activation, STAT proteins translocate to the nucleus, where they regulate the expression of genes involved in proliferation, survival, and differentiation. Among these, STAT3 and STAT5 have emerged as key drivers of tumor progression and therapeutic resistance across multiple malignancies.¹⁵ In GBM, the role of STAT3 has been extensively characterized, with well-established functions in sustaining GSC phenotypes and promoting immune evasion and tumor growth.^16,17 The role of STAT5 signaling in GBM, particularly in GSC biology, remains less well defined.

STAT5 exists in 2 closely related paralogs, STAT5A and STAT5B, which can exhibit distinct but overlapping functions in hematopoiesis and cancer.¹⁸ Recent studies have demonstrated that STAT5 signaling is activated in the infiltrative, residual population of GBM cells that evade surgical resection and contribute to tumor recurrence.¹⁹ These residual cells are highly enriched in stem-like properties and are highly invasive.³ STAT5B has been identified as a downstream effector of oncogenic EGFRvIII, a constitutively active EGFR variant found in a significant subset of GBMs.²⁰ EGFRvIII promotes GSC proliferation, invasion, and resistance, yet the transcriptional mediators of these effects remain poorly defined.^21,22 The association of EGFRvIII and STAT5B signaling with infiltrative, stem-like GBM cells suggests that STAT5B may play an important role in the transcriptional programs that sustain GSC maintenance and invasion.

Here, we identify a previously unrecognized transcriptional complex between OLIG2 and STAT5B that regulates GSC proliferation and stem-like properties. We demonstrate that OLIG2 expression enhances STAT5B activation and that the 2 transcription factors form a physical complex in GSCs. Transcriptomic analyses of patient-derived GBM models reveal that knockdown of OLIG2 or STAT5B leads to convergent downregulation of genes essential for GSC maintenance and mitotic fidelity, including BIRC5 and NUSAP1. Functional studies demonstrate that knockdown of these effectors impairs GSC proliferation and migration. Furthermore, in vivo knockdown of STAT5B prolongs the survival of orthotopic GBM models, phenocopying the effects of OLIG2 loss and reducing the expression of canonical stemness markers. Together, these findings define a novel, targetable OLIG2-STAT5B transcriptional axis in GBM that promotes GSC-driven malignancy. Disrupting this pathway represents a potential therapeutic strategy to impair GSC persistence and mitigate tumor recurrence in GBM.

Methods

Cell Culture

Patient-derived GSC lines were established from primary and recurrent GBM surgical specimens at the Barrow Neurological Institute (Phoenix, AZ, United States) or derived from patient-derived xenograft (PDX) models from the Mayo Clinic Brain Tumor PDX Library (Rochester, MN, United States). GSCs were maintained as tumor spheres in non-tissue culture-treated flasks or as adherent monolayers on laminin-coated surfaces. Cells were cultured in DMEM/F-12 supplemented with B27, N2, 1% penicillin-streptomycin, 20 ng/mL EGF, and 20 ng/mL bFGF. All cell lines were routinely tested and confirmed negative for mycoplasma contamination prior to experimentation.

Expression Constructs and Lentivirus Production

Bacterial plasmids containing the coding sequences of human STAT5A (clone ID: HsCD00043806) and STAT5B (clone ID: CD00852071) were obtained and generated as previously described.²⁰ Lentiviruses were produced in Lenti-X 293T cells (Takara Bio) using polyethylenimine (Polysciences Inc.) for transfection along with second-generation packaging plasmids psPAX2, pMD2.G, and lentiviral transfer vectors (pGFP-C-shLenti or pCDH-STAT5A-HA/STAT5B-FLAG). Short hairpin RNA (shRNA) vectors targeting STAT5A and STAT5B were obtained from Horizon Discovery. shRNA vectors targeting OLIG2 and non-targeting controls were sourced from the Broad Institute RNAi Consortium and subcloned into the pGIPz vector (Fisher Scientific). Full shRNA sequences and construct details are listed in Supplementary Table S2.

Transfection With Small Interfering RNA

Transient knockdowns of STAT5A, STAT5B, OLIG2, BIRC5, and NUSAP1 were performed using Qiagen FlexiTube small interfering RNA (siRNA) reagents. For each gene, 4 individual siRNA sequences were tested by quantitative PCR, and the most effective 2 (or all 4 for OLIG2) were selected for experiments. Additional siRNAs for STAT5A and STAT5B were sourced from Ambion and Invitrogen, respectively, as previously described.¹⁹ For transfection experiments, GSCs were plated on laminin-coated plates to promote adherent monolayer growth. Transfections were performed using 10 nM siRNA with Lipofectamine RNAiMAX (Thermo Fisher Scientific) following the manufacturer’s protocol. All siRNA sequences are listed in Supplementary Table S4.

Immunoblotting Analysis

Immunoblotting was performed as previously described.²⁰ Cells were lysed in RIPA buffer supplemented with protease and phosphatase inhibitors, followed by sonication and protein quantification using the BCA assay. Equal amounts of protein (40 µg per lane) were separated on SDS-PAGE gels and transferred to nitrocellulose or PVDF membranes using a semi-dry transfer system. Membranes were blocked with either Odyssey Blocking Buffer (LI-COR Biosciences) or 5% non-fat milk in TBST and incubated with primary antibodies diluted in the corresponding buffer. Detection was performed using HRP- or fluorophore-conjugated secondary antibodies, developed with chemiluminescent substrate (SuperSignal West Pico PLUS) or imaged using the Odyssey CLx Near-Infrared system. Band intensities were quantified using Image Studio Lite (LI-COR). All Western blot images shown are representative of at least 3 independent biological replicates. For fresh-frozen patient tumor lysates, variable protein integrity and partial degradation were observed in a subset of samples. Consequently, immunoblot signals from these samples were evaluated qualitatively to confirm protein presence or relative enrichment, rather than used for direct quantitative comparisons across samples.

Proximity Ligation Assays

Proximity ligation assays (PLA) were performed on GSCs plated on laminin-coated coverslips and cultured in GSC media. After fixation and permeabilization, cells were incubated overnight at 4 °C with primary antibody pairs, including STAT3/STAT5B, HDAC1/OLIG2, STAT5B/OLIG2, STAT3/OLIG2, HA-tag/STAT3, HA-tag/OLIG2, FLAG-tag/STAT3, and FLAG/OLIG2. PLA was performed using the Duolink In Situ Red and Green Detection Kits (Sigma-Aldrich) according to the manufacturer’s protocol. Images were acquired using a Keyence BZ-X microscope with a 63X oil-immersion objective. PLA foci were quantified using the Keyence Hybrid Cell Counter software. Foci were normalized to DAPI-stained nuclei, and data represent the mean ± SD from at least 3 independent biological replicates, with 6 images quantified per condition.

Quantitative PCR

Total RNA was isolated from GSCs using the RNeasy Purification Kit (QIAGEN) or PureLink RNA Mini Kit (Thermo Fisher Scientific) according to the manufacturer’s protocol. RNA concentrations were quantified using the Spark Nano Quant plate reader. cDNA was synthesized from 1 µg of RNA using the SuperScript III First-Strand Synthesis Kit (Life Technologies). Quantitative PCR was performed using TaqMan Gene Expression Assays on a QuantStudio 6 Flex Real-Time PCR System (Thermo Fisher Scientific). Relative gene expression was calculated using the ΔΔCt method, normalized to ACTB or 18S rRNA. A full list of target genes is provided in Supplementary Table S4.

Neurosphere Formation and Limiting Dilution Assay

GSCs were dissociated into single cells for neurosphere proliferation and limiting dilution assays (LDA). For proliferation assays, cells were seeded in 96-well plates at 20 000 cells per well in growth factor–supplemented media, and sphere growth was monitored using the Incucyte SX5 Live-Cell Analysis System (Sartorius). For neurosphere initiation assays, cells were seeded at limiting dilutions ranging from 0.3125 to 640 cells per well and cultured for 3-5 days. Sphere formation efficiency was calculated using extreme limiting dilution analysis (ELDA).²³ A well was scored as positive if at least one neurosphere (defined as a cluster of ≥2–4 cells) was present.

Migration Assay

GSCs were transfected with siRNAs targeting genes of interest or a non-targeting control. Twenty-four hours post-transfection, cells were seeded into collagen-coated 96-well migration plates at a density of 75 000 cells per well. After an additional 24 hours of incubation, cell stoppers were removed and fresh medium supplemented with growth factors were added to each well. Migration into the detection zone was monitored between 24 and 72 hours post-seeding using the Incucyte SX5 Live-Cell Analysis System (Sartorius). Migration area was quantified using ImageJ software.²⁴

Cell Viability

Twenty-four hours post-transfection, GSCs were seeded into laminin-coated 96-well plates at a density of 10 000 cells per well in GSC-culture medium. After overnight incubation at 37 °C to allow cell adherence, cells were maintained in culture for 7 days. Cell viability was assessed at endpoint using the CTG assay following the manufacturer’s protocol, and luminescence was measured using a microplate reader.

In Vivo Orthotopic Studies

All experimental procedures involving laboratory animals were approved by the Animal Care Committee of Dignity Health (Approval number, A3519-01). Luciferized GBM39 cells (100 000 cells/mouse) transduced with short hairpin RNAs were injected intracranially into Fox1nu nude male mice. Tumor growth was monitored every 7-14 days by bioluminescence imaging following D-Luciferin injection, and survival was assessed using Kaplan–Meier analysis. Mice were monitored daily and euthanized at humane endpoints. Full procedural details are provided in Supplementary Methods.

RNA-Seq Data Processing and Analysis

To assess transcriptomic changes associated with STAT5A, STAT5B, and OLIG2 knockdown, RNA was isolated from GBM39 GSCs 48 hours post-siRNA transfection. Poly(A)+ mRNA was enriched, and cDNA libraries were generated and sequenced (150 bp paired-end reads, ∼30M reads/sample) on an Illumina NovaSeq 2000 platform. High-quality reads were aligned to the hg38 human genome, and differential gene expression was analyzed using the edgeR pipeline. Pathway enrichment, transcriptional subtype classification, and gene correlation analyses were performed using established bioinformatics tools. Detailed RNA-seq processing and analysis are provided in Supplementary Methods.

Immunohistochemistry

Brains were harvested at the time of sacrifice, fixed in 10% neutral-buffered formalin, and paraffin-embedded using standard procedures. Sections were stained with antibodies against SOX2, Vimentin, and NUSAP1 followed by fluorescent secondary detection. H&E staining was performed using standard protocols. Slides were imaged using an Axioscan 7 and analyzed in QuPath v6. Full staining protocols and antibody details are provided in Supplementary Methods.

Statistical Analyses

All statistical analyses were performed using GraphPad Prism version 10.5.0. A P value of <.05 was considered statistically significant. For comparisons among multiple groups, one-way ANOVA followed by Dunnett’s post hoc test was used to assess differences relative to a control group. Two-way ANOVA with Sidak’s multiple comparisons test was applied where appropriate. Comparisons between 2 groups were conducted using unpaired, 2-tailed t-tests. Survival analyses were performed using Kaplan–Meier survival curves and log-rank (Mantel–Cox) tests.

Results

Expression of STAT5B and OLIG2 is Positively Correlated in GBM Patient Cohorts

While previous studies have established a role for STAT3 in regulating stemness, the contribution of STAT5 remains poorly defined. To investigate the potential association between STAT5 and stemness in IDH-WT GBM, independent of subtype or oncogenic signaling, we assessed Pearson correlations between STAT5 expression and established stemness markers using transcriptomic data from The Cancer Genome Atlas (TCGA). We identified a strong positive correlation between STAT5B and multiple established GSC-associated genes, including OLIG2, SOX2, BRN2, and NES expression, while no such relationship was observed for STAT5A (Figure 1A, Supplementary Figure S1A-H). These correlations suggest coordinated activation of stemness-associated transcriptional programs but do not distinguish direct functional cooperation. Due to the lineage-restricted role of OLIG2 and its previously established integration with oncogenic RTK signaling, we prioritized the relationship between OLIG2 and STAT5B for further investigation. This relationship was further supported by a positive correlation between STAT5B and OLIG2 expression in the Mayo Clinic Brain Tumor PDX GBM datasets (Figure 1A). In primary and recurrent GBM patient specimens, STAT5B protein expression is positively correlated with OLIG2 levels (Figure 1B). EGFRvIII-expressing GSC lines generally showed higher pSTAT5 and OLIG2 levels than EGFR wild-type lines; however, some EGFR-amplified, EGFRvIII-negative lines also exhibited robust OLIG2 expression and STAT5 activation (Figure 1C), consistent with the fact that OLIG2-positive GBM cells can engage STAT5 through multiple upstream signaling inputs. In contrast, correlations between STAT5B and other stemness-associated markers (Supplementary Figure S1A-H) did not uniformly translate to protein-level association across models (Supplementary Figure 1I). Furthermore, we find that OLIG2-null murine GSCs (mGSCs) exhibited decreased activation of STAT5, with increase in STAT5 phosphorylation upon expression of wild-type OLIG2 (Figure 1D). Collectively, these data suggest that STAT5 may play a role in regulating stemness of GSCs, with STAT5B emerging as the predominant functional paralog.

Figure 1.

STAT5B and OLIG2 expression are positively correlated across GBM datasets and models. (A) Scatter plots and Pearson correlation coefficients for OLIG2 and STAT5A/STAT5B gene expression in IDH-WT TCGA (n = 163) and PDX (n = 59) tumor samples. (B) Immunoblot of primary and recurrent GBM tumor lysates probed for OLIG2, total STAT5, and α-tubulin as a loading control. Due to partial degradation in a subset of tumor specimens, α-tubulin levels vary across samples. Lane numbers correspond to independent patient tumor lysates, (C) Immunoblot analysis of patient-derived GSC lines with either EGFR wild-type or EGFR-mutant status, probed for phosphorylated STAT5 (pSTAT5), STAT5B, OLIG2, EGFR, EGFRvIII, phosphorylated STAT3 (pSTAT3), total STAT3, and β-actin (loading control). (D) OLIG2-null mGSCs were transduced with either eGFP (control) or wild-type (WT) Olig2 via retroviral transduction. Lysates were immunoblotted for pSTAT5, pSTAT3, Fn14, and α-tubulin (loading control).

STAT5B as a Dominant Driver of GSC Self-Renewal Capacity

To investigate the role of STAT5 in maintaining GSC self-renewal, we performed siRNA-mediated knockdown of STAT5A and STAT5B in patient-derived human GSC (hGSC) lines, EGFRvIII expressing GSC models—GBM39 and GB239. OLIG2, a well-established regulator of GSC stem-like phenotype, served as a positive control.^10,13 Knockdown efficiency for each target was confirmed by quantitative PCR (for STAT5A/B) and western blot (for OLIG2) (Figure 2H, Supplementary Figure S2A and B). Stem-like properties were assessed using LDA to quantify stem cell frequency and self-renewal capacity, while proliferative capacity was evaluated by monitoring neurosphere growth over time.

Figure 2.

STAT5 is critical for regulating glioma stem cell self-renewal and proliferation. (A, B) Extreme limiting dilution assay in GBM39 hGSCs transfected with siRNAs targeting OLIG2, STAT5A, or STAT5B. Knockdown of OLIG2 (P = .00336) and STAT5B (P = .042) significantly reduced stem cell frequency compared to siNT, while STAT5A knockdown had no significant effect (P = .282). (C) Quantification of neurosphere size at day 5 in GBM39 hGSCs. (D, E) Limiting dilution assay in GB239 hGSCs transfected with siRNAs targeting OLIG2, STAT5A, or STAT5B. Knockdown of OLIG2 (P = .00571), STAT5A (P = .0403), and STAT5B (P = .000442) all significantly reduced stem cell frequency. (F) Quantification of neurosphere size at day 5 in GB239 hGSCs. (G) Cell viability assessed by CellTiter-Glo after siRNA knockdown of STAT5B and OLIG2. One-way ANOVA with multiple comparisons showed significant reduction in sphere size and viability for each knockdown versus siNT (P < .001 [***], and P < .0001 [****]). (H) Immunoblot analysis of GBM39 lysates following siRNA knockdown showing total STAT5 and OLIG2 expression, α-tubulin used as loading control.

In GBM39 cells, knockdown of STAT5B significantly decreased sphere-initiating cell frequency (Figure 2A and B), while knockdown of both STAT5A and STAT5B led to reduced average neurosphere sizes (Figure 2C and Supplementary Figure S2C). In GB239 cells, knockdown of both STAT5A and STAT5B similarly decreased sphere-initiating frequency (Figure 2D and E) and reduced sphere size (Figure 2F and Supplementary S2D), with the effect of STAT5A knockdown on sphere size being less pronounced than STAT5B. The consistent reduction in sphere size across models indicates a shared role for STAT5A and STAT5B in promoting GSC proliferation or expansion, however, with a more predominant role for STAT5B across all models. The selective impact of STAT5A knockdown in GB239, but not in GBM39, suggests a potential context-dependent contribution to GSC maintenance. Consistent with this, knockdown of STAT5B and OLIG2 resulted in a significant reduction in cell viability across models (Figure 2G and Supplementary Figure S2E).

A Novel Interaction Between STAT5B and OLIG2

Given the demonstrated involvement of STAT5 in GSC regulation and its strong correlation with OLIG2, we next sought to explore the potential functional relationship between these 2 transcription factors. Considering their canonical roles in gene regulation, we investigated whether STAT5 and OLIG2 colocalize intracellularly. PLA revealed that OLIG2 formed significantly more complexes with STAT5 than with STAT3 (Figure 3A and 3C, Supplementary Figure S3A-C) in GBM39, GB239, and GB233 hGSCs. Co-immunoprecipitation further confirmed a OLIG2-STAT5B-containing complex in GBM39 and an mGSC line (Figure 3E).

Figure 3.

OLIG2 interacts with STAT5, with preferential binding to STAT5B. (A) Proximity ligation assay (PLA) demonstrating endogenous OLIG2–STAT5 interactions in GBM39 and GB239 hGSCs. Nuclei were visualized with DAPI, and F-actin staining was used to identify cell structure. White box indicates enlarged region of image. (B) PLA in GBM39 hGSCs transduced with either HA-tagged STAT5A or FLAG-tagged STAT5B to assess differential interaction with OLIG2. (C) Quantification of PLA foci per cell from (A) in GBM39 and GB239 hGSCs (n = 3 independent, biological replicates per cell line). (D) Quantification of PLA foci per cell from (B) in GBM39 hGSCs shows significantly increased OLIG2–STAT5B interactions compared to OLIG2–STAT5A (P = .0176, n = 3 biological replicates, unpaired two-tailed t-test (P < 0.05 [*])). All scale bars are 20 µm. (E) Co-immunoprecipitation of STAT5 in OLIG2-wild-type mGSCs and GBM39. Lysates were immunoprecipitated with anti-STAT5 antibody and immunoblotted (IB) with the indicated antibodies to assess protein–protein interactions.

To determine if EGFR inhibition disrupts this OLIG2-STAT5B-containing complex, we treated GBM39 with the fourth-generation EGFR inhibitor (EGFRi), BDTX-1535. Short-term EGFRi treatment suppressed STAT5 activation and downstream signaling along with a modest decrease in OLIG2-STAT5B PLA foci (Supplementary Figure S4A-D).

To test paralog-specific interactions, we generated stable GBM39 cell lines expressing HA-tagged STAT5A or FLAG-tagged STAT5B and repeated PLA (Figure 3B, Supplementary Figure S5A-C). OLIG2 preferentially interacted with STAT5B, as evidenced by increased PLA signals in the FLAG-STAT5B condition compared to HA-STAT5A (Figure 3D, Supplementary Figure S5D). These findings further establish STAT5B as a key paralog driving the interaction with OLIG2 in EGFRvIII-expressing GSCs.

STAT5B and OLIG2 Co-regulate Shared Transcriptional Networks

To explore the transcriptional relationship between STAT5B and OLIG2, we performed RNA sequencing of GBM39 hGSCs following siRNA-mediated knockdown of STAT5A, STAT5B, STAT5A/B, and OLIG2 (Supplementary Figure S6A). Principal component analysis (PCA) demonstrated clear separation of experimental conditions with replicates clustering tightly together (Supplementary Figure S6B). Hierarchical clustering revealed that the knockdown of STAT5B, STAT5A/B, and OLIG2 substantially altered the transcriptomic landscape, whereas STAT5A alone produced modest changes (Figure 4A).

Figure 4.

STAT5 and OLIG2 co-regulate overlapping transcriptional networks in glioma stem cells. (A) Hierarchical clustering heatmap of differentially expressed genes from RNA-seq analysis of GBM39 hGSCs transfected with siNT, siSTAT5A, siSTAT5B, or siSTAT5A&B (n = 3 biological replicates). (B, C) Hierarchical clustering of enriched GO terms from overlapping up- and downregulated genes in OLIG2(B) and STAT5B(C) knockouts. Top: Treeplot of the top 20 enriched GO terms by adjusted P value from upregulated (top) and downregulated (bottom) overlapping genes. Branching reflects functional similarity based on gene-term overlap and GO structure; circle size represents gene set size and color indicates adjusted P value. (D, E) Gene Ontology (GO) term–gene network derived from overlapping differentially expressed genes in OLIG2(D) and STAT5B(E) knockout (KO) conditions. The top 10 most significantly enriched GO terms (by adjusted P value) and up to 10 contributing genes per GO term were selected for network visualization per KO condition. Node color represents upregulated (red) genes and downregulated (blue) genes in KO conditions, with grey nodes representing GO terms. Edges represent genes enriched in connected GO terms.

Differential gene expression analysis further revealed that STAT5B and OLIG2 share 165 differentially expressed genes (DEGs), while STAT5A and OLIG2 shared only 12 DEGs, highlighting a unique transcriptional alignment between STAT5B and OLIG2 (Supplementary Figure S6C). Over-representation analysis (ORA) of these shared DEGs showed significant enrichment for cell cycle–related processes (Figure 4B-E). Gene ontology analysis revealed that both OLIG2 and STAT5B knockdown led to downregulation of genes involved in mitosis, DNA replication, and chromosome organization alongside upregulation of immune response and cytokine signaling pathways, supporting their convergent roles in cell cycle regulation and immune suppression (Supplementary Figure S6D and S6E).

To characterize these transcriptional changes in the context of GBM heterogeneity, we applied 2 established GBM classification frameworks: Transcriptional-based²⁵ and Pathway-based²⁶ classifications. Transcriptional-based classification stratifies GBM into 3 molecular subtypes (classical, mesenchymal, and proneural) based on gene expression signatures, although it offers limited predictive values for survival or therapeutic response. Pathway-based classification refines this approach, identifying GBM subtypes based on metabolic and pathway signatures: Proliferative/Progenitor (PPR), Neuronal (NEU), Mitochondrial (MTC), and Glycolytic/Plurimetabolic (GPM).

Using the Transcriptional-based subtypes, control and OLIG2 knockdown samples were consistently classified as mesenchymal (Supplementary Figure 7A). In contrast, STAT5B knockdown samples displayed increased variability across replicates, with individual samples exhibiting enrichment of proneural-associated transcriptional features rather than a uniform subtype classification. Given the observed variability, these findings indicate trends in transcriptional bias following STAT5B perturbation rather than definitive subtype switching. Under the Pathway-based subtypes, control samples were uniformly classified as PPR, while both OLIG2 and STAT5B knockdown partially shifted toward the Mitochondrial (MTC) subtype (Supplementary Figure 7B-E). This parallel shift further supports the functional convergence of STAT5B and OLIG2 in maintaining GSC proliferative programs. In contrast, STAT5A resulted in less robust profile switching, indicating a more heterogeneous or limited role in GSC subtype regulation.

STAT5B/OLIG2 Transcriptional Axis Regulates GSC Maintenance, Migration, and Chemoresistance

We next validated several genes co-regulated by OLIG2 and STAT5B through independent knockdown experiments in GBM39 (Supplementary Figure S8A) and confirmed that an anti-apoptotic protein, BIRC5 (Survivin) and microtubule binding protein, NUSAP1 were consistently downregulated (Figure 5A), whereas other candidate genes did not show consistent expression changes (Supplementary Figure S8B). To confirm that these effects were not cell line–specific, we performed independent knockdown experiments in additional GSC models, GB239 and GB233, and similarly observed downregulation of BIRC5 and NUSAP1 (Figure 5A; Supplementary Figure S8C-E). BIRC5 expression is typically limited in adult differentiated tissues but is significantly upregulated in various cancer types, particularly within the cancer stem cell compartment.²⁷ While NUSAP1 is frequently overexpressed in cancer,²⁸ its role in development and stem cell maintenance is less clearly defined. However, given its canonical role in maintaining mitotic fidelity and chromosomal stability during cell division, it is likely to contribute to the preservation of stem-like populations in which the cell cycle is tightly regulated.²⁹

Figure 5.

The STAT5B–OLIG2 transcriptional axis regulates glioma stem cell maintenance, migration, and chemoresistance. (A) Quantitative RT-PCR analysis showing decreased expression of BIRC5 and NUSAP1 following siRNA-mediated knockdown of STAT5B or OLIG2 in GBM39 and GB239 hGSCs (n ≥ 3 biological replicates). Expression of both genes was significantly reduced by siSTAT5B and siOLIG2 in both models (P < .0001 for all comparisons). (B and C) Limiting dilution assays in GBM39 hGSCs following knockdown of BIRC5 or NUSAP1 demonstrate reduced stem cell frequency: siBIRC5 (P < .0001), siNUSAP1 (P = .0236). (D) Quantification of neurosphere size at day 7 post-transfection in GBM39. (E and F) Limiting dilution assays in GB239 hGSCs following knockdown of BIRC5 or NUSAP1 demonstrate reduced stem cell frequency (P < .0001). (G) Quantification of neurosphere area at day 5 post-transfection in GB239. (H) Cell viability assessed by CellTiter-Glo after siRNA knockdown of BIRC5 and NUSAP1. (I) Migration assays in GBM39 and GB239 hGSCs following knockdown of STAT5B, OLIG2, BIRC5, or NUSAP1. Statistical significance was determined by one-way ANOVA with multiple comparisons or ELDA as appropriate; pairwise comparisons are indicated on the graphs (*P < .05, **P < .01, ***P < .001, ****P < 0.0001).

To assess the relevance of BIRC5 and NUSAP1 in GSC maintenance, we transiently reduced their expression using siRNAs and evaluated its impact of stemness properties (Supplementary Figure S9A and B). Silencing either gene led to a significant reduction in sphere-forming frequency and neurosphere size (Figure 5B-G), indicating that both BIRC5 and NUSAP1 are critical for sustaining growth in the GSC population. Knockdown of BIRC5 and NUSAP1 reduced cell viability in GBM39, whereas only NUSAP1 knockdown decreased viability in GB239 (Figure 5H), and BIRC5 knockdown reduced viability in GB233 (Supplementary Figure S9C).

Since BIRC5 and NUSAP1 have been implicated in regulating cell migration,^27,30,31 we next asked whether the STAT5B–OLIG2 transcriptional axis similarly mediates this phenotype in GSCs. To address this, we assessed migration following transient knockdown of STAT5B, OLIG2, BIRC5, and NUSAP1. Suppression of STAT5B or OLIG2 significantly reduced GSC migration capabilities (Figure 5I, Supplementary Figure S9D), recapitulating the effects observed with BIRC5 and NUSAP1 knockdown. These findings suggest that the STAT5B–OLIG2 axis, along with its downstream effectors BIRC5 and NUSAP1, contributes to GSC motility.

STAT5B and OLIG2 Knockdown Prolongs Survival In Vivo

To evaluate whether STAT5 signaling is important for the tumorigenic potential of GSCs, we generated GBM39 cells stably expressing shRNAs targeting STAT5A, STAT5B, or OLIG2 and implanted them orthotopically into immunodeficient mice (Figure 6A, Supplementary Figures S10A-C). Consistent with previous reports, OLIG2 suppression ­significantly prolonged survival. Notably, STAT5B knockdown also conferred a survival advantage, whereas STAT5A knockdown had no significant effect (Figure 6B, Supplementary Figure S10D).

Figure 6.

STAT5 and OLIG2 knockdown prolongs survival and alters tumor histopathology in vivo. (A) Schematic of the in vivo study design. Luciferase-expressing GBM39 hGSCs were transduced with lentiviral shRNA vectors targeting STAT5A, STAT5B, or OLIG2 and injected intracranially into immunocompromised mice. Created in BioRender. https://BioRender.com/4fbdseh. (B) Kaplan–Meier survival analysis of mice orthotopically injected with GBM39 cells stably expressing shRNAs targeting STAT5A, STAT5B, or OLIG2. Knockdown of OLIG2 (P = .0241) and STAT5B (P = .033) significantly prolonged survival compared to non-targeting control (shNT) (shNT n = 9, shSTAT5A n = 7, shSTAT5B n = 4, shOLIG2 n = 5). (C) Low-power representative scans of shNT, shOLIG2, shSTAT5A, and shSTAT5B tumors (left to right) stained with H&E. High-power insert shows cellular features. Solid black arrows indicate mitotic figures. (D) Representative images of necrotic features (dashed line) found within the tumor core. (E) Representative images of SOX2 (green) immunostaining. Solid white arrows indicate representative SOX2+ cells. (F) Representative images of Vimentin (green) immunostaining. Solid white arrows in the high-power insert indicate invasive tumor cells with a strong pattern of Vimentin expression. (G) Representative images of NUSAP1 (red) immunostaining. Solid white arrows indicate representative NUSAP1+ cells. Dashed white arrows indicate NUSAP1+ cells undergoing mitosis. Scale bars are 1mm in A, 20 um in C, and 50 μm in B, D, and E. Scale bars in high-power inserts are all 20 μm.

Histopathological analysis of tumor sections revealed reduced necrosis in the tumor core of shSTAT5B-treated mice (Figure 6C and D). Both shSTAT5B and shOLIG2 tumors exhibited decreased SOX2 expression (Figure 6E, Supplementary Figure S10E), consistent with impaired GSC maintenance. Vimentin staining indicated reduced tumor invasiveness in both knockdown conditions (Figure 6F). In addition, NUSAP1 expression was markedly decreased in shSTAT5B and shOLIG2 tumors (Figure 6G, Supplementary Figure S10E), further supporting the role of the STAT5B-OLIG2 axis in promoting aggressive GSC phenotypes in vivo.

Discussion

Despite significant advances in understanding GBM biology, the complex transcriptional network governing GSCs maintenance and growth remains incompletely characterized. In this study, we identify a novel transcriptional complex between OLIG2 and STAT5B that sustains GSC proliferation and maintenance. These findings provide new insights into how developmental transcription factors cooperate with cytokine signaling pathways to maintain treatment-refractory GSC populations.

Our findings that OLIG2 and STAT5B are positively correlated across multiple GBM datasets and form a physical complex in GSCs provide new insights into GSC transcriptional regulation. While OLIG2 has been shown to play a key role in maintenance of GSC identity,^13,32 the cooperative factors that modulate its activity have remained poorly defined. The preferential interaction between OLIG2 and STAT5B, rather than STAT5A, in EGFRvIII-expressing GSC models suggests paralog-specific functions that warrant further investigation. This finding is particularly relevant given that STAT5B but not STAT5A serves as a key downstream effector of EGFRvIII,^20,33 which is enriched in infiltrative tumor regions where GSCs are thought to persist after treatment. Although our mechanistic studies were performed in EGFRvIII-positive GSC models, OLIG2 and STAT5 signaling are not restricted to this genotype. Elevated STAT5 activation and OLIG2 expression are observed across multiple EGFR-altered and receptor tyrosine kinase (RTK)-driven GBMs, suggesting that OLIG2–STAT5B cooperation may extend beyond EGFRvIII-specific contexts. Systematic evaluation of this axis across additional EGFR-amplified and EGFR wild-type GBM models will be important to define the breadth, context-dependence, and therapeutic relevance of this interaction across diverse RTK-driven tumors.

The transcriptomic analysis revealing 165 shared differentially expressed genes between OLIG2 and STAT5B knockdown conditions indicates that these factors regulate overlapping cellular programs in GSCs. The shared transcriptional targets of OLIG2 and STAT5B provide insights into how GSCs integrate developmental and cytokine signaling. While the regulation of immune response pathways by STAT5B is well-established, the observation that OLIG2 knockdown similarly leads to upregulation of immune modulatory genes suggests a previously unrecognized role for OLIG2 in cytokine production and other related immune responses that requires further investigation. Emerging literature has begun to implicate OLIG2 in transcriptional programs beyond lineage specification, including potential roles in tumor–microenvironment interactions.³² This dual regulation may explain how GSCs might help maintain an immunologically privileged state while preserving their proliferative capacity through coordinate control of cell cycle machinery by both factors. Testing this hypothesis will require evaluation in immunocompetent or humanized GBM models, where tumor–immune interactions can be more directly interrogated.

Among the shared validated targets of this transcriptional axis, BIRC5 and NUSAP1, emerge as functionally relevant effectors. BIRC5 (survivin) is particularly interesting given its dual roles as a mitotic regulator and apoptosis inhibitor.²⁷ Its restriction to embryonic tissues in normal physiology but overexpression in cancer cells creates a favorable therapeutic window. BIRC5 is aberrantly overexpressed in GBM and contributes to therapeutic resistance and tumor cell survival.^34–37 Notably, BIRC5 expression is enriched in the GSC compartment and has been implicated in protecting these cells from cytotoxic stress. This therapeutic potential has been explored clinically through the peptide-based vaccine SurVaxM that targets survivin, which has completed Phase I and Phase IIa trials in newly diagnosed GBM patients, showing promising results with a median overall survival of 25.9 months when combined with temozolomide.^38,39 A larger, randomized Phase II clinical trial combining SurVaxM with temozolomide in newly diagnosed GBM is underway (NCT05163080). Our ­findings position BIRC5 as a downstream effector of an EGFR-associated, OLIG2-STAT5B transcriptional program, supporting the evaluation of survivin-targeted therapies in EGFR-altered, OLIG2-high tumors.

Similarly, the essential role of NUSAP1 in mitotic spindle organization and cytokinesis positions it as a critical mediator of GSC proliferative capacity.⁴⁰ It is also overexpressed in GBM and associated with poor prognosis.^28,31 While less extensively studied in glioma, emerging evidence suggests that NUSAP1 supports tumor progression by engaging oncogenic pathways, including NF-κB and Hedgehog signaling.^30,31,41 The consistent convergence of phenotypic outcomes when targeting either BIRC5 or NUSAP1 suggests their classification as downstream effectors rather than peripheral targets of the STAT5B-OLIG2 axis and highlights their importance as clinically relevant therapeutic targets within the GSC population in GBM.

Prior studies have demonstrated that inhibition of upstream RTK/SFK signaling in combination with TMZ can enhance therapeutic efficacy,²⁰ suggesting that disruption of STAT pathway activity may sensitize GBM to alkylating chemotherapy. Our findings provide a mechanistic context for these observations by identifying an OLIG2–STAT5B transcriptional program that may underlie this vulnerability. Definitive establishment of this mechanism will require direct combination studies pairing TMZ and/or radiation with STAT5B or upstream pathway inhibitors, along with functional assays examining DNA damage accumulation, apoptosis induction, and DNA repair pathway activation following axis suppression. These experiments will clarify whether OLIG2–STAT5B directly modulates DNA damage tolerance or instead maintains pro-survival signaling during therapeutic stress. Together, these findings position the STAT5B–OLIG2 complex as a driver of GSC persistence and a potential therapeutic vulnerability, particularly in EGFRvIII-positive, MGMT-methylated GBM.

While we observe a consistent reduction in migration following suppression of STAT5B, OLIG2, BIRC5, or NUSAP1, it is important to acknowledge that proliferation, survival, and motility programs are tightly interconnected in GSCs. Thus, the decreased migratory capacity may reflect both direct effects on cytoskeletal or invasion-associated pathways and indirect effects resulting from impaired proliferative fitness.

The in vivo knockdown of STAT5B significantly prolongs survival while reducing tumor invasiveness, which strengthens the translational potential of targeting the STAT5B-OLIG2 axis. Notably, shRNA-mediated knockdown of STAT5B or OLIG2 results in incomplete suppression of both factors, making complete inhibition of tumor growth unlikely as it has been previously observed for OLIG2 knockdown.^10,11 The phenocopying of OLIG2 suppression effects, including reduced SOX2 expression and decreased invasiveness, further supports the functional relevance of this transcriptional partnership in driving malignant tumor behavior. The reduced NUSAP1 expression in endpoint tumors further validates this downstream target in vivo. Collectively, these findings identified the STAT5B–OLIG2 axis as a potential therapeutic node for combination strategies aimed at impairing GSC persistence. Future studies employing pharmacologic STAT5B inhibition, either alone or in combination with OLIG2 inhibitor and expanding these analyses to additional PDX-derived GBM models will be important for defining the contexts in which this vulnerability can be most effectively leveraged, rather than positioning this pathway as a standalone curative target.

Our subtype analysis provides additional mechanistic insights into how the STAT5B-OLIG2 axis influences GSC identity. The shift from a Proliferative/Progenitor (PPR) signature toward a Mitochondrial (MTC) signature, as defined by the pathway-based GBM classification proposed by Garofano et al.,²⁶ upon STAT5B or OLIG2 disruption suggest that this complex actively suppresses metabolic differentiation programs. The PPR subtype’s association with high proliferative activity and stemness markers aligns with our functional data of STAT5B-OLIG2 activity. In contrast, the MTC subtype is associated with enhanced mitochondrial activity and clinically with increased therapeutic vulnerability and the most favorable clinical outcomes. This transition toward an MTC-like phenotype upon STAT5B or OLIG2 knockdown suggests that OLIG2 and STAT5B may actively repress mitochondrial differentiation programs to preserve stemness and enhance therapy resistance in GSCs. Longitudinal single-cell RNA sequencing and expanded clinical cohort analyses will be critical to elucidate how OLIG2 and STAT5B loss influences clonal dynamics, metabolic rewiring, and transcriptional drift which are key factors in therapy response and recurrence.

Several important questions remain for future investigation. While we demonstrate a cooperative interaction between STAT5B and OLIG2, it remains unclear whether these transcription factors co-occupy the same regulatory elements. In silico promoter motif analysis identified canonical STAT5 consensus motifs and OLIG2 E-box elements within proximal promoter regions of candidate target genes, BIRC5 and NUSAP1 (Supplementary Tables S5 and S6). However, these analyses alone cannot establish direct co-occupancy. Chromatin immunoprecipitation sequencing (ChIP-seq) will be necessary to determine whether they converge on shared promoters or enhancers and directly co-regulate targets like BIRC5 and NUSAP1. While this study establishes OLIG2–STAT5B cooperation using patient-derived GSCs, primary and recurrent GBM specimens, and orthotopic tumor models, higher-resolution approaches such as single-cell or spatial transcriptomics will be valuable for defining the cellular and spatial contexts in which this axis operates within patient tumors. In addition, genomic stratification may further refine the contexts in which the OLIG2–STAT5B program is engaged. Given prior evidence that EGFRvIII and STAT5 signaling are enriched in infiltrative and residual tumor regions,¹⁹ spatial profiling may help clarify how OLIG2–STAT5B activity is distributed across tumor compartments and how it intersects with intratumoral heterogeneity and therapy resistance. Understanding the relationship between EGFRvIII signaling, this transcriptional axis and the broader tumor microenvironment, including immune cell interactions, will be important for developing comprehensive therapeutic strategies. Although STAT3 has been implicated in GSC biology, our analyses did not identify it as a central component of the OLIG2-dependent transcriptional network examined here. These findings suggest that OLIG2-STAT interactions may exhibit context-dependent specificity with the STAT family. Notably, while STAT5B emerges as the dominant effector of this regulatory axis, STAT5A appears to contribute more modestly to GSC maintenance, raising the possibility of paralog-specific functions within GBM stem-like population. A more comprehensive assessment of OLIG2 interactions across STAT family members, including STAT3 and STAT5A, and across diverse GBM genotypes and signaling environment, will be important for clarifying the extent of STAT-family-specific cooperation with lineage-restricted transcription factors. In parallel, integrative phospho-proteomic and epigenomic profiling could further define whether STAT5A is differentially regulated, excluded from key chromatin regulatory regions, or functionally redundant under specific molecular contexts.

From a therapeutic perspective, the STAT5B-OLIG2 axis presents multiple potential intervention points. Direct inhibition of STAT5B activation, disruption of the protein-protein interaction, or targeting of downstream effectors like BIRC5 or NUSAP1 could all represent viable strategies.

In conclusion, we define the STAT5B-OLIG2 transcriptional axis as a critical regulator of GSC maintenance, proliferation, and tumor progression in GBM. This pathway integrates developmental and cytokine signaling to sustain malignant GSC properties and represents a promising target for preventing tumor recurrence. Our findings provide a mechanistic framework for understanding GSC transcriptional regulation and identify specific molecular targets for future therapeutic development.

Supplementary Material

Supplementary material is available online at Neuro-Oncology Advances (https://academic.oup.com/vdag127).

Author Contributions

Conceptualization: A.M., J.M., N.L.T., and S.M. Methodology: A.M., J.M., M.R.B., D.N., S.W., S.T., C.S., and K.O. Data Analysis: A.M., J.M., M.F., and D.N. Resources: N.L.T., S.M., J.C.L., and S.P.F.E. Visualization: A.M., J.M., M.F., and D.N. Formal Analysis: A.M., J.M., M.F., D.N., M.R.B., S.W., and K.O. Writing: A.M., J.M., N.L.T., and S.M. Review & Editing: A.M., J.M., S.P.F.E., J.C.L, N.L.T., and S.M. All authors reviewed and approved the final manuscript.

Conflict of Interest Statement

The authors declare no competing interests.

Funding

The work is funded by National Institutes of Health (NIH) R01CA271431 (NLT & SM). Support also provided by Mayo Clinic (NLT) and the Barrow Neurological Foundation and The Ben and Catherine Ivy Foundation (JM & SM). SM is also supported by NIH 1R01CA260003-01A1, 1R01CA264934-01, CA230725 and 1R01NS123038-01. This study was also supported by CTSA Grant Number TL1 TR002380 (AM) from the National Center for Advancing Translational Science (NCATS).

Ethics Statement

All animal studies were conducted in accordance with protocols approved by the Dignity Health Institutional Animal Care and Use Committee (IACUC), in compliance with the guidelines established by the National Institutes of Health for the humane treatment of animals.

Data Availability

The data supporting the findings of this study will be made available in a public repository upon acceptance of the manuscript.