Phenotypic variations in glioma stem cells: regulatory mechanisms and implications for therapeutic strategies

Abstract

Glioma represents the most prevalent primary tumors of the central nervous system, originating from glial cells. Cancer stem cells have the ability to extensively proliferate, self-renew, and form colonies, which contribute to tumorigenesis. Studies have found a population of cells within glioblastoma exhibiting characteristics similar to those of cancer cells, termed glioma stem cells (GSCs). GSCs have two distinct phenotypes: mesenchymal subtype (MES) and proneural subtype (PN). Despite the vital role of these subtypes in glioma biology, there is a significant lack of comprehensive reviews focusing on the regulatory mechanisms underlying each phenotype.This review integrates emerging insights into the regulatory mechanisms underlying GSCs plasticity, with dedicated analysis of novel pathways governing PN and MES phenotypes and their dynamic transition. By examining these critical elements, we aim to contribute to the development of novel therapeutic strategies in the ongoing fight against gliomas.

Background

Glioma stem cells (GSCs) are a subpopulation of cells within glioblastomas, characterized by the ability to self-renew and differentiate into various types of glial cells. GSCs are the primary contributors to the occurrence, progression, and recurrence of glioblastoma. Subsequently, investigating the molecular pathways and metabolic mechanisms governing GSCs behavior may offer novel and reliable therapeutic targets for the prevention and treatment of gliomas. The MES subtype of GSCs is primarily associated with the expression of genes such as CD44, CD109, Octamer-Binding Transcription Factor 4 (OCT4), Aldehyde Dehydrogenase 1 Family Member A3 (ALDH1A3), Epidermal growth factor receptor (EGFR), and Chitinase 3-like 1 (Chi3l1/YKL40), whereas the PN subtype is primarily characterized by the expression of genes such as CD133, Oligodendrocyte Transcription Factor 2 (Olig2), sex-determining region Y-box 2 (SOX2), and Notch2. These gene expression profiles are crucial in determining the distinct biological behaviors and phenotypes of each glioblastoma subtype [1–5].

The MES and PN GSCs exhibit distinctly different survival characteristics and traits such as growth patterns and metabolic processes [6–8]. Accordingly, investigating the regulatory mechanisms of these two subtypes of phenotypes is crucial and, therefore, guides this study. Literature review reveals that numerous studies have investigated the regulatory mechanisms on the expression and maintenance of the stemness in GSCs; however, there is limited research on the regulatory mechanisms governing the distinct phenotypes of these GSCs subtypes [7]. In this study, we aim to investigate the regulatory mechanisms of these two subtypes, thereby significantly contributing to our understanding of the phenotypic regulation of GSCs, particularly the MES subtype, which is vital for effectively combating the onset, progression, and recurrence of glioblastoma. This review extends prior foundational work through integrating emerging evidence on non-coding RNA networks, extracellular vesicle signaling dynamics, and treatment-induced proneural-to-mesenchymal transition (PMT) mechanisms, alongside metabolic reprogramming, thereby providing potential therapeutic strategies by identifying key pathways and regulatory molecules that could serve as potential treatment targets [7, 9].

Differences in the pathways regulating the GSCs phenotypes

To explore and comprehensively summarize the regulatory mechanisms of the two GSCs phenotypes, it is vital to understand the key signaling pathways involved within these two subtypes. In PN GSCs, the Notch and Wnt pathways are critically important, whereas in MES GSCs, the NF-κB, C/EBPβ, HIPPO and STAT3 pathways are activated and implicated [2, 10–14]. We list the relevant pathways and their upstream and downstream molecules in Table 1.

Table 1.

Differences in pathways regulating the phenotypes of GSCs

GSCs phenotype	Signaling pathway	Regulatory Mechanism	References
PN	Notch	Regulated by Norrin	[25]
	PI3Kα/AKT	Regulated by PIK3CA	[26]
	Wnt		[12]
MES	Wnt	Regulated by LGR5	[18]
		Regulated by YKL40/CD44	[27, 28]
		Regulated by Grb10	[29]
	Wnt5a	Targeting NFAT	[20, 21]
	NF-κB	Regulated by TNF-α; Targeting TGM2	[22, 30]
		Regulated by EGFRvⅢ; Targeting ALDH1A3	[31]
		Regulated by HDAC1	[32]
		Regulated by BACE2	[33]
		Regulated by CXCL8/AKT	[34]
		Regulated by IFI16	[35]
		Regulated by MMPs	[36]
		Regulated by HIF1/PLOD1	[37]
		Regulated by FOSL1/UBC9	[38]
		Regulated by MLK4	[23]
		Regulated by MIF	[39]
		Regulated by LINC01057	[40]
		Via TAB2/NF-κB axis	[41]
	STAT3	Regulated by TAB2; Targeting RTVP-1	[41, 42]
		Regulated by RBPJ; Targeting ANGPTL4	[11, 43]
		Regulated by ITGA2	[44]
		Regulated by BACE2	[33]
		Regulated by YKL40/CD44	[27, 28]
		Regulated by TGM2	[30]
		Regulated by HuR	[35]
		Regulated by miR-184-3p	[45]
		Regulated by H4K16 via H4K16/STAT3/MAPK axis; Targeting PDIA3P1	[46]
	MAPK	via H4K16/MAPK axis; Targeting PDIA3P1	[46, 47]
		Regulated by NFAT1/NDEL1	[48]
		Regulated by NEAT1/let-7 g-5p	[49]
		Regulated by A3AR; Targeting Snail, Twist, ZEB1	[50–53]
	PI3K/AKT
		Regulated by YKL40/CD44; Targeting ID1	[27, 28, 54]
		Regulated by TGM2	[30]
		Regulated by Rictor	[55, 56]
	C/EBPβ	Regulated by BACE2; Targeting CD109	[2, 33]
		Regulated by TGM2; Targeting NNMT	[30, 57]
	HIPPO	Regulated by YAP/TAZ	[58, 59]
	Shh	Regulated by Gli1/BMI1	[60]

Regulated by A3AR, and targeting Snail, Twist, and ZEB1. These two mechanisms are concurrently involved in both the MAPK and PI3K/AKT signaling pathways

The Notch signaling pathway, often involved in regulating brain cell development, regulates cellular de-differentiation, proliferation, and survival [15]. In GSCs, this pathway positively influences tumor cell growth, self-renewal, invasion, and migration [16]. It exhibits a strong correlation with genes and characteristics associated with the PN subtype and is it rarely implicated in recurrent tumors [10, 17]. Additionally, the Notch signaling interacts with Wnt signaling to facilitate the proliferation and self-renewal of the PN GSCs while inhibiting their differentiation to maintain their phenotype [12]. Research has indicated that the Wnt pathway may be involved in MES transition. Leucine-rich repeat-containing G protein-coupled receptor 5 (LGR5), a marker of GSCs, inhibits the phosphorylation of β-catenin, thereby activating the Wnt/β-catenin pathway and promoting the MES transition, enhancing their invasive capabilities. LGR5 can also be induced by OLIG2 to negatively regulate the Wnt signaling during cell differentiation [18, 19]. Moreover, the non-canonical Wnt5a pathway can interact with Notch1 signaling to promote the expression of MES phenotype [20]. This interaction is probably associated with Wnt5a-mediated endothelial differentiation and the activation of the downstream molecule nuclear factor of activated T cell (NFAT) [21]. The NF-κB pathway is primarily activated in response to various stimuli, such as radiation, inflammatory factors, and stress, and plays a key role in regulating cell proliferation and survival. In GSCs, this pathway is primarily activated by tumor necrosis factor-alpha (TNF-α), resulting in the upregulation of CD44, a hallmark gene of MES GSCs, while downregulating genes associated with PN GSCs, thereby increasing tumor cells resistance to radiation [22, 23]. Moreover, the platelet-derived growth factor receptor-β (PDGFR-β) mediated pathway is closely associated with the expression of PN subtype markers; it also has the capacity to activate STAT3 signaling molecules and promote the presentation of typical characteristics of MES subtypes such as invasiveness [24].

The regulatory mechanisms of phenotypes in PN GSCs subtypes

ASCL1 facilitates the acquisition of PN gene expression in GSCs

Achaete-scute homolog 1 gene (ASCL1) is a member of the basic helix-loop-helix (bHLH) family of transcription factors involved in the regulation of neuronal differentiation and phenotypes [61]. This protein requires dimerization with other bHLH proteins to effectively bind to DNA and exert its regulatory functions. ASCL1 is a specific marker for PN subtype glioblastoma and PN GSCs [62]. The expression of ASCL1 promotes the upregulation of key molecules essential for the development of mature neurons, contributing to a neuronal-like morphology in GSCs. This highlights the inherent role of in driving the PN phenotype in GSCs. In contrast, N-myc downstream-regulated gene 1(NDRG1) acts as a tumor suppressor by inhibiting cancer cell proliferation, invasion, and promoting apoptosis and differentiation [63]. While the specific mechanisms of NDRG1 in GSCs remain underexplored, some studies show that NDRG1 can induce a shift towards the MES phenotype. However, ASCL1 directly binds to the promoter region of NDRG1, repressing its expression and thereby facilitating the expression of the PN phenotype in GSCs [64]. Additionally, ASCL1 can directly interact with epidermal growth factor receptor (EGFR), a marker gene for MES GSCs, within its intragenic region, thereby inhibiting the transition of GSCs to the MES phenotype [64]. Furthermore, research has shown that Norrin, a product of the Norrie disease protein (NDP) gene, acts as an atypical Wnt ligand in ASCL1^low GSCs (likely MES subtype) to inhibit cell growth via Frizzled class receptor 4 (FZD4) through the classical Wnt pathway. Conversely, in ASCL1^high GSCs, NDP functions as an upstream regulator of the Notch signaling pathway, promoting GSCs growth (Fig. 1) [25].

Fig. 1.

ASCL1 facilitates GSCs to acquire PN expression. In GSCs with high ASCL1 expression, NDRG1 and EGFR expression is inhibited, promoting the progression of GSCs towards a neural phenotype. Additionally, it can activate the Notch signaling pathway through Norrin to promote GSCs growth. Conversely, in GSCs with low ASCL1 expression, cell growth is inhibited through the FZD4 receptor-mediated interaction with the Wnt pathway

OLIG2 and SOX2 are key proneural transcription factors

OLIG2 and ASCL1 both belong to the bHLH transcription factor family and are transcriptionally activated by ASCL1 [65]. Activated OLIG2 can form a positive feedback loop with the Epidermal Growth Factor Receptor (EGFR), jointly promoting the maintenance of the PN glioma subtype characteristics [66]. Furthermore, OLIG2 directly inhibits the cell cycle inhibitor gene p21, thereby maintaining the stemness of GSCs [67]. Notably, when OLIG2 expression is downregulated, platelet-derived growth factor α (PDGFα) levels increase, accompanied by downregulation of SOX2 expression, collectively promoting the PMT [66, 68].

SOX2, as a key factor in maintaining the stemness of PN GSCs, works alongside stem cell markers like Nestin to maintain the stemness and subtype characteristics of PN GSCs [69]. Its expression is positively regulated by multiple factors, including Tripartite Motif Containing 26 (TRIM26), TRIM24, myeloid Elf-1 like factor (MEF), the long non-coding RNA (lncRNA) MALAT1, EphA2, and CDK8 [69–74]. The Notch signaling pathway drives SOX2 expression by activating SOX9 transcription; in turn, SOX2 upregulates NOTCH1 expression, forming a positive feedback loop that jointly maintains PN subtype characteristics [75]. Additionally, transforming growth factor β (TGFβ) can directly upregulate SOX2 expression via SOX4, a process that antagonizes the inhibitory effect of FHL3 on SOX4 [76, 77].

PDGF signaling serves as a critical pathway for the PN GSCs

The PDGF signaling pathway plays a central regulatory role in the progression of GSCs. This pathway directly promotes GSC proliferation and growth by inducing the expression of methyltransferase-like 3, while stabilizing the Inhibitor of DNA-binding protein (ID2) through the E2F transcription factor-mediated ubiquitin-specific peptidase 1 (USP1), thereby sustaining the survival of PN GSCs [78, 79]. At the metabolic level, PDGF significantly enhances aerobic glycolysis in GSCs by activating the phosphoinositide 3-kinase (PI3K)/AKT signaling axis [80]. Notably, sorting nexin 10 (SNX10) maintains GSCs stemness and proliferation through PDGFRβ, while LINC02283 cooperatively promotes GSCs survival and expansion through specific interaction with PDGFRA [81, 82]. This stemness regulatory network is also modulated by miRNAs – PDGF signaling utilizes specific miRNA clusters (e.g., miR-106 and miR-17) to maintain stem cell properties, whereas other miRNAs (e.g., miR-29a) exert tumor-suppressive effects by targeting PDGF signaling [83, 84].

Within the tumor microenvironment, M2-polarized microglia significantly enhance GSCs migratory and invasive capabilities through activation of PDGFRβ [85]. PDGF signaling additionally induces endothelial nitric oxide synthase (eNOS) expression, driving Notch pathway activation via the NO/cGMP/PKG cascade to establish cross-pathway coordination [86]. Critically, PDGFR can also promote mesenchymal phenotype by inducing transcription factor SNAIL expression through the NFκB signalin, thereby complementing the PMT mechanism triggered by OLIG2 downregulation in the PN subtype [87].

PATZ1 is involved in the maintenance of the PN phenotype

In their study, Guadagno et al. demonstrated that PATZ1, a zinc finger protein containing a POZ/BTB domain and an AT hook, is associated with the expression of markers for the PN subtype, such as OLIG2 and SOX2, and it is enriched in PN GSCs. Additionally, PATZ1 may also bind to the promoter region of CXCR4, potentially inhibiting the CXCR4 pathway, which is implicated in the PMT [88, 89].

Additionally, increased expression levels of p110α in PN GSCs facilitates the activation of the phosphoinositide-3-kinase (PI3K)/AKT/mTOR signaling pathway, which is important for maintaining the PN phenotype [26].

MES GSCs exhibit intense aggressiveness compared to PN GSCs. Consequently, investigating therapeutic strategies to maintain GSCs in the PN phenotype and even exploring strategies to convert MES GSCs to PN GSCs for further treatment could provide novel modalities for the treatment and prevention of glioma recurrence.

The regulatory mechanisms of phenotypes in MES GSCs subtypes

The heterogeneity of GSCs constitutes the core of their malignant nature and therapeutic resistance. Among these, the MES subtype has garnered significant attention due to its pronounced invasiveness, resistance to radiotherapy, chemotherapy, and targeted therapies as well as its role in shaping an immunosuppressive tumor microenvironment (TME). A profound understanding of the mechanisms sustaining MES phenotype homeostasis and the PMT is crucial for developing effective therapeutic strategies. These complex processes primarily revolve around core transcriptional hub networks and are profoundly regulated by the TME.

Environmental stressors and microenvironmental signaling

MES phenotype is closely related to invasiveness and treatment resistance. Its establishment and maintenance are highly dependent on several core transcription factor hubs, which integrate multiple upstream signals and regulate extensive effector gene networks. And several studies have demonstrated that in GSCs, the PN GSCs can transition to MES GSCs in response to adverse environmental conditions such as hypoxia, drug treatment, and radiation therapy. This transformation is a significant mechanism resulting in glioma recurrence. Key regulators of this process include STAT3, C/EBPβ, TAZ (Transcriptional co-activator with PDZ-binding motif), and NF-κB pathways [13, 14, 22, 90]. Conversely, there are mechanisms for the reverse transformation of the MES back to the PN subtype [41].

Hypoxic stress

Hypoxia remodels GSCs fate through multiple pathways. Under hypoxic conditions, tumor cells trigger endoplasmic reticulum (ER) stress, leading to the translocation and expression of cell surface GRP78 (csGRP78) [91–94]. This process prevents lysosomal degradation of β-site APP-cleaving enzyme 2 (BACE2), subsequently activating key pathways closely associated with the MES phenotype, including STAT3, C/EBPβ, and NF-κB (p65) [33].

The core effector of hypoxia, hypoxia-inducible factor-1α (HIF-1α), accumulates and upregulates procollagen-lysine, 2-oxoglutarate 5-dioxygenase 1 (PLOD1). PLOD1 binds to IκB, thereby releasing and activating NF-κB [37]. Concurrently, it promotes M2 polarization of tumor-associated macrophages (TAMs) and facilitates the binding of triggering receptor expressed on myeloid cells-1 (TREM1). This interaction enhances PLOD1 expression, while TREM1 activates the intracellular Smad2/3 signaling pathway by binding to TGFβR, collectively promoting PMT [95, 96]. HIF also induces the production of adenosine and activates the A3 Adenosine Receptor (A3AR). A3AR regulates the expression of CD73 and prostatic acid phosphatase (PAP) through HIF1α and HIF2α, activating PI3K/AKT and Erk signaling to promote key transcription factors involved in PMT (e.g., Snail, Twist, and ZEB1) [50–53].

Additionally, under hypoxic conditions, ZDHHC18 stabilizes BMI1 and activates Gli1, leading to the activation of genes in the Sonic Hedgehog (Shh) pathway and affecting the expression of N-cadherin, E-cadherin, and MMP2 [60, 97–99]. In contrast, ZDHHC23 degrades BMI1, promoting the PN phenotype [97]. The lncRNA HIF1α-antisense RNA 2 (HIF1A-AS2) is also induced under hypoxia (Fig. 2) [100].

Fig. 2.

Hypoxic conditions cause the PMT. Under hypoxic conditions, GRP78 can migrate from the ER to the cell membrane, preventing the lysosomal degradation of BACE2, while activating the STAT3, C/EBPβ, and NF-κB pathways to express genes associated with MES and PMT. ZDHHC18 stabilizes BMI1 by inhibiting the ubiquitination induced by RNF144A, thereby promoting the expression of the MES/PMT genes. Conversely, under normal oxygen conditions, ZDHHC23 induces the ubiquitination and degradation of BMI1, leading to the expression of the PN phenotype. HIF-1 increase the expression of PLOD1, thereby activating the NF-κB pathway. Similarly, HIF1A-AS2 interact with IGF2BP2 and DHX9 to enhance the expression of downstream mRNAs such as FOSL1, maintaining the MES phenotype. FOSL1 can activate IKK through UBC9, thereby activating the NF-κB pathway. In M2 TAMs, HIF-1α promotes the expression of TREM1, which then activates the Smad2/3 signaling pathway by releasing TGFβ2 that binds to TGFβR on GSCs cell membranes. Additionally, HIF can induce the production of adenosine and activate A₃AR, which activates the expression of CD73/PAP through HIF1α/2α, further producing A₃AR. Meanwhile, A₃AR can activate the PI3K/AKT and Erk signaling pathways, promoting the expression of TFs such as Snail and Twist, which are significantly related to mesenchymal transition

Therapeutic pressures

Therapeutic pressures such as radiotherapy, chemotherapy (TMZ), and anti-angiogenic therapy (bevacizumab) directly cause cellular damage and stress responses, potently inducing the activity of key factors such as C/EBPβ and NF-κB, and exacerbating hypoxia [2, 47, 101–104]. Radiotherapy enhances mesenchymal characteristics through the ROS-xCT axis and activates NF-κB by phosphorylating IKKα via mixed lineage kinase 4 (MLK4) [23, 105]. TMZ chemotherapy triggers p38α-MAPK and upregulates PDIA3P1 to block the degradation of C/EBPβ [47]. Bevacizumab therapy also specifically upregulates HIF1α and mesenchymal-related genes (e.g., Nicotinamide N-methyltransferase (NNMT), YKL40) [101].

Tumor microenvironment crosstalk

Tumor microenvironment cells influence the malignant phenotype of GSCs through paracrine networks. NFAT, a family of transcription factors that are activated in the cytoplasm by the calcium-dependent calcineurin [106]. NFAT1 binds to the neurodevelopment protein 1-like 1 (NDEL1) promoter to increase its expression, resulting in the activation of the Erk pathway which sustains the malignant phenotype of GSCs [48]. Concurrently, NFAT1 promotes the transcription of complement C3 in TAMs. The secreted C3a binds to C3aR, simultaneously activating a Ca²⁺-NFAT1 positive feedback loop and activating Gal-9/TIM-3 to induce M2-like polarization of TAMs (Fig. 3) [107–109]. Furthermore, GSCs secrete periostin (POSTN) to recruit TAMs to the GSCs niche. M2 TAMs secrete TGF-β, IL-10, and TNF-α, establishing the foundation for activation of STAT3, C/EBPβ, and NF-κB pathways.​[22, 30, 107, 110, 111]

Fig. 3.

NFAT facilitates the maintenance of the MES phenotype of GSCs. In MES GSCs, NFAT is activated in the cytoplasm through calcineurin, enabling it to exert its biological functions. NFAT1 induces the expression of NDEL1, which activates the Erk signaling pathway to promote the MES phenotype. Additionally, NFAT2 induces the expression of HDAC1, subsequently inhibiting the acetylation of p65 in the NF-κB protein, thereby enhancing its biological activity. This inhibition promotes the expression of MES marker genes including CD44 and YKL40. YKL40 can also enhance the expression of CD44, and these two molecules can interact to activate the β-catenin, STAT3, and AKT, thereby maintaining the MES phenotype. MES GSCs secrete POSTN to recruit M2 TAMs, and then M2 TAMs secrete TGF-β1 to promote the MES phenotype of GSCs. In TAMs, NFAT1 induces the expression of C3 and TIM-3, facilitating M2 polarization. Subsequently, C3 enhance the NFAT1/Ca²⁺ pathway through autocrine activity via C3aR, resulting in the formation of a positive feedback loop. TIM-3 receives Gal-9 secreted by MES GSCs to promote M2 polarization. Additionally, MES GSCs secrete CXCL8, which activates NF-κB via the AKT pathway to maintain their own phenotype. It can also promote M2 polarization in TAMs through the CXCL2-JAK2/STAT3 axis

Extracellular vesicles (EVs) have a crucial impact within the TME and are instrumental in the PMT (Fig. 4) [112]. M2 TAMs can activate the RelB-p50 axis and STAT3 signaling pathway by releasing small extracellular vesicles (sEVs) containing molecules such as miR-27a-3p, which target and inhibit the 3' UTR of chromatin domain helicase DNA-binding protein 7 (CHD7) [45, 113]. Additionally, EVs also reflect the unique characteristics and pathway activities of different GSCs subtypes, with PN GSCs being highly susceptible to the effects of EVs, which drive their transformation into the MES subtype. This phenomenon contributes to the plasticity and heterogeneity of GSCs subtypes [114–116]. EVs released by MES GSCs not only activate NF-κB and STAT3 signaling pathways in PN GSCs through C/EBPβ but also stimulate angiogenesis in endothelial cells and promote tumor cell survival. Furthermore, the miR-155-5p carried by these EVs promotes PMT in gliomas through its interaction with acetyl-CoA sulfotransferase 12 (ACOT12) [117–120]. Endothelial cell secrete EVs containing matrix metalloproteinases (MMPs) that are transported to PN GSCs, resulting in Notch pathway inactivation and NF-κB pathway activation [36]. Neuron-derived exosomes suppress expression of RNA binding motif protein 15 (RBM15) through miR-184-3p, reducing N6-methyladenosine (m6A) modification levels on Discs Large Homolog 3 (DLG3) mRNA and consequently activating the STAT3 signaling pathway [45].

Fig. 4.

Extracellular vesicles mediate the mechanisms of PMT. EVs derived from MES GSCs can activate STAT3 and NF-κB signaling through C/EBPβ to promote the transition of PN GSCs to MES GSCs. sEVs containing lncRNAs such as miR-27a-3p, miR-22-3p and miR-221-3p from M2 TAMs can enhance this transition by inhibiting the effects of CHD7 on STAT3 and NF-κB pathways. Additionally, EVs with MMPs from endothelial cells promote PMT via NF-κB signaling and simultaneously inhibiting Notch signaling. Furthermore, exosomes secreted by neurons containing miR-184-3p can block the expression of RBM15, thereby preventing DLG3 from suppressing the STAT3 pathway and facilitating mesenchymal transition. Moreover, MES GSCs secrete EVs containing EGFR to endothelial cells which can induce angiogenesis, aiding tumor cell survival, and release EVs containing miR-504 that inhibit M2 polarization of TAMs

Core transcriptional hubs and phenotype output

NF-κB signaling hub

The NF-κB transcription factor integrates diverse signals including TNF-α, hypoxia, and radiotherapy. TNF-α secreted by TAMs upregulates TGM2 and inducible T-cell co-stimulatory ligand (ICOSLG) through the classical NF-κB pathway, thereby inducing the expression of key transcription factors such as C/EBPβ, STAT3, and TAZ, while promoting the production of IL-10 and TGF-β to sustain the malignant phenotype [22, 30, 111]. This hub consolidates MES traits through dual mechanisms: NFAT2 upregulates histone deacetylase 1 (HDAC1) expression by binding to its promoter, enhancing p65 deacetylation to amplify NF-κB transcriptional activity. This persistently drives expression of critical MES markers including CD44 and YKL40 (Fig. 3) [32]. YKL40 interacts with CD44 to activate downstream β-catenin, STAT3, and AKT signaling, forming a positive feedback loop that stabilizes MES state and stemness [27, 28].

In PMT, actin-related protein-2/3 complex 1B (ARPC1B) inhibits the ubiquitination of DNA-binding protein IFI16 by tripartite motif-containing 21 (TRIM21) in GSCs, promoting NF-κB pathway activation and driving PMT [35]. During radiotherapy, this pathway can also be activated through macrophage migration inhibitory factor (MIF) and lncRNA LINC01057 [39, 40]. Hypoxia-induced HIF1A-AS2 (HIF1α-antisense RNA 2) interacts with insulin-like growth factor 2 mRNA-binding protein 2 (IGF2BP2) and ATP-dependent RNA helicase A (DHX9) to activate downstream messenger RNAs [100]. FOS-like antigen 1 (FOSL1) enhances transcription of the E2-conjugating enzyme UBC9, facilitating SUMOylation of the key Deubiquitinases (DUB) enzyme CYLD. This activity prevents deubiquitination of NF-κB intermediates (NEMO, TRAF2, TRAF6), ultimately promoting IKK complex phosphorylation to activate NF-κB signaling and induce PMT [38]. Salmonella pathogenicity island 1 (SPI1) activates NF-κB signaling via TNF-α and also binds the FKBP12 promoter to increase its transcription, elevating MES marker expression. This interaction also amplifies TGF-β1/PI3K/Akt signaling, further supporting the maintenance of MES phenotype [121, 122]. Additionally, NF-κB can enhance mesenchymal transformation by initiating ALDH1A3 transcription in EGFRvIII-driven contexts [31].

C/EBPβ signaling hub

C/EBPβ is robustly activated under environmental stressors including hypoxia, radio/chemotherapy, and the immunosuppressive microenvironment orchestrated by M2 TAMs. It synergizes with NF-κB and STAT3 signaling to drive PMT [13, 22, 30, 90, 111, 117].

Its stability is maintained through multiple protective mechanisms: tissue transglutaminase 2 (TGM2) induces calcium-dependent aggregation and degradation of growth arrest and DNA damage-inducible transcript 3 (GADD153), thereby alleviating its inhibitory effect on C/EBPβ [30]. Simultaneously, protein disulfide isomerase family A, member 3 pseudogene 1 (PDIA3P1) prevents the binding of MDM2 to C/EBPβ, thereby blocking its ubiquitination and subsequent degradation [47].

This master regulator governs the unique metabolic adaptations and epigenetic landscape of MES GSCs. Within the perinecrotic niche where MES GSCs reside, C/EBPβ upregulates nicotinamide N-methyltransferase (NNMT), which transfers methyl groups from S-adenosylmethionine (SAM) to nicotinamide (NAM) while suppressing DNA methyltransferases (DNMTs). This metabolic imbalance inhibits DNA methylation in MES GSCs, creating an epigenetic environment conducive to MES phenotype maintenance [57]. C/EBPβ further coordinates the ALDH1A3 core regulatory network (Fig. 5), Mao et al. demonstrated that MES GSCs predominantly utilize an aldehyde dehydrogenase (ALDH)-mediated glycolytic pathway, establishing ALDH1A3 as a specific marker essential for maintaining the mesenchymal subtype [3]. The Forkhead family FOXD1 directly activates ALDH1A3 transcription, while ubiquitin-specific proteases 9X (USP9X) and USP21 stabilize both ALDH1A3 and FOXD1 through deubiquitination [123–125]. ALDH1A3 converts retinaldehyde into retinoic acid (RA), inducing TGM2 expression [126]. TGM2 subsequently upregulates inhibitor of DNA binding 1 (ID1) by activating PI3K/Akt pathway. The co-expression of TGM2 and ID1 maintain the phenotype of MES by promoting cell proliferation and survival [54]. Moreover, C/EBPβ induces CD109 expression, modulating downstream Yes-associated protein (YAP)/TAZ and HIPPO signaling to facilitate PMT [2].

Fig. 5.

ALDH1A3 can maintain the MES phenotype. ALDH1A3 plays a significant role in supporting the MES phenotype, whereas USP9x and USP21 deubiquitinate ALDH1A3 and FOXD1, respectively, with FOXD1 possessing the ability to activate the expression of ALDH1A3. EGFRvIII also activate the expression of ALDH1A3 through NF-κB. Moreover, ALDH1A3 can oxidize retinaldehyde to retinoic acid, which in turn promotes the expression of TGM2. TNF-α enhance the expression of TGM2 by activating the NF-κB pathway through TNFR. TGM2 can then increase the expression of ID1 via the PI3K/AKT pathway and degrade GADD153 in a Ca²⁺-dependent manner, thereby increasing the expression of C/EBPβ. This interaction promotes the MES phenotype and supports the growth and radiation resistance of these GSCs

STAT3 signaling hub

STAT3 plays a central role in cytokine secretion and TME remodeling. Yuan et al. discovered that MES GSCs secrete CXCL8, which binds to their autologously expressed CXCR2 receptor, activating the JAK2/STAT3 pathway to form a self-sustaining autocrine loop that maintains STAT3 activity and mesenchymal traits. This process simultaneously promotes polarization of TAMs toward an immunosuppressive M2-like phenotype [34]. These M2 TAMs subsequently secrete TGF-β and sEVs containing miR-27a-3p, miR-22-3p, and miR-221-3p, which activates the STAT3 signaling pathway (Fig. 4) [45, 107, 113].

Integrin alpha-2 (ITGA2) and Enhancer of Zeste Homolog 2 (EZH2) directly activates STAT3 to promote PMT [44, 127]. STAT3 also interacts with the promoter of angiopoietin-like 4 (ANGPTL4) to facilitate the maintenance of the MES phenotype [11]. ARPC1B, which crucially regulates the NF-κB pathway, additionally promotes PMT by stabilizing the RNA-binding protein HuR to activate STAT3 signaling [35]. The long non-coding RNA MIR222HG is anchored to histone H4 after transcription, binding to the YWHAE/HDAC5 complex, leading to the deacetylation of H4K16. This process activates both STAT3 and MAPK signaling pathways, thereby facilitating PMT [46].

Furthermore, IL6-activated STAT3 and C/EBPβ synergistically bind to the promoter of related to testis-specific, vespid and pathogenesis protein 1 (RTVP-1), upregulating its expression to promote MES markers while suppressing PN markers. RTVP-1 additionally drives MES transformation through IL6 and CXCR4 pathways [42]. Human mesenchymal stem cells derived from glioblastoma can mediate PMT via the IL6/STAT3 pathway in conjunction with recombination signal binding protein for immunoglobulin kappa J region (RBPJ) [43, 128].

YAP/TAZ signaling hub

The YAP/TAZ hub, which is mainly regulated by the classic HIPPO pathway, locks the MES state through a transcription complex, which controls multiple downstream factors to maintain the MES phenotype of GSCs [129, 130]. Activation of the HIPPO pathway activates the Mammalian Ste20-like 1 and 2(MST1/2) and Large Tumor Suppressor 1 and 2 (LATS1/2) cascade, leading to inhibition of YAP/TAZ. Conversely, HIPPO pathway inactivation results in YAP/TAZ activation and downstream regulation [131]. In addition, YAP is also affected by the activation of the EGFR-AKT axis [132]. According to findings by Bhat et al., YAP/TAZ binds to TEAD at the promoters of PMT-associated genes in the nucleus, where this complex upregulates expression of MYC and EGFR to maintain stemness in GSCs, while simultaneously sustaining the MES phenotype and promoting PMT [14, 133]. Research indicates that YAP and OLIG2 mutually suppress each other's expression in the MES and PN subtypes, thereby sustaining their respective phenotypes [134, 135].

In a study by Yang et al., insulin-like growth factor 2 (IGF2) mRNA-binding protein 1 (IMP1) binds to methylated YAP mRNA, thereby stabilizing its translation and activating the HIPPO signaling pathway. Subsequently, YAP/TAZ increase the expression of IMP1 which forms a feed-forward loop to promote the phenotype of MES [136]. Moreover, YAP/TAZ can also enhance the radiation resistance of GSCs through its downstream (Growth Differentiation Factor 15) GDF15 [137]. YAP/TAZ further participate in the αvβ3-mediated pathway that enhances the glucose transporter GLUT3 expression, thereby promoting the survival of GSCs [138]. CD146 prevents the degradation of YAP by inhibiting the expression of LATS1, thereby promoting the HIPPO signaling pathway [58]. Additionally, the DUB Otubain-2 (OTUB2) activates the YAP/TAZ via HIPPO-independent way [59] and mutant-type p53 may regulate YAP/TAZ stability through AKT2 and WASP-interacting protein (WIP) [139].

Auxiliary regulatory networks

The Wnt5a/β-catenin signaling pathway cooperates with EGFR-expressing EVs secreted by MES GSCs to promote angiogenesis and enhance the malignant MES phenotype [118, 119, 140]. Interferons (IFNs) can drive mesenchymal expression through STAT1 signaling [141].

Long non-coding RNAs (lncRNAs), while non-protein-coding, play critical roles in determining GSCs subtype phenotypes. Guardia et al. highlighted the importance of differential lncRNA expression in determining the phenotypes of various subtypes of GSCs [142]. Nuclear enriched abundant transcript 1 (NEAT1) enhances the malignant properties of GSCs by suppressing let-7g-5p and increasing MAP3K1 expression. Let-7g-5p exert its effects by inhibiting VSIG4-mediated mesenchymal transformation [49, 143]. Conversely, certain lncRNAs suppress the mesenchymal phenotype. For instance, miR-370-3p directly binds to NEAT1 to inhibit its ability to promote PMT [144]. MiR-194-3p can inhibit MES phenotype by downregulating the NF-κB and IL6/STAT3 signaling axis through the suppression of TGF-β Activated Kinase 1/MAP3K7 Binding Protein (TAB2) [41]. During specific treatment processes, miR-128 and miR-200 can be upregulated to inhibit BMI1 and the Shh pathway, respectively, thereby suppressing PMT [98]. MiR-504 targets Grb10 in GSCs and FZD7 in GBM, thereby downregulating the Wnt pathway and suppressing mesenchymal transformation. Additionally, it can inhibit the M2 polarization of TAMs [29, 145]. MiR-34a targets Rictor, a component of the mTORC2 complex, inhibiting AKT phosphorylation, the Wnt/β-catenin pathway, and MMP9 activity, which ultimately reduces the proliferation and invasiveness of MES GSCs [55, 56]. Autophagy features are essential in maintaining the MES phenotype of GSCs, with genes such as DRAM1 and p62 mediating this effect. However, the expression of miR-93 can suppress autophagy-related genes that are highly expressed in MES GSCs, whereas being upregulated in PN GSCs to maintain the PN phenotype [146, 147].

Also, circNCAPG facilitates the nuclear translocation of Ras responsive element binding protein 1 (RREB1) to enhance the TGF-β1 signaling pathway [148]. Conversely, extracellular vesicles secreted by human mesenchymal stem cells containing miR-744-5p can inhibit the TGF-β1/MAPK signaling pathway, thereby suppressing M2-like polarization of TAMs [149]. A2B adenosine receptor (A2BAR), S100A4, a fatty acid-binding protein 7 (FABP7), high-mobility group A (HMGA2) and the FGFR1/FOXM1 pathway can regulate the expression of genes associated with PMT, such as Snail, ZEB1, and Twist [150–154]. Ubiquitin carboxyl-terminal hydrolase L3 (UCHL3), a DUB, acts on POLD4 (DNA polymerase delta 4, accessory subunit), thereby stabilizing it and promoting PMT [155]. Moreover, the Ca²⁺-activated K⁺ channel KCa3.1 along with integrin alpha-6 are highly expressed in MES GSCs, promoting their radioresistance [156–158]. Catenin alpha-like 1 modulates the inflammatory factor CCL2, promoting the mesenchymal transformation of GSCs, potentially due to the chemotactic effect of the CCL2 on mesenchymal stem cells towards GSCs enrichment [159, 160].

In addition to the numerous molecules already discussed that are important in phenotype of glioma stem cells, several others with less well-understood mechanisms may also serve as potential targets for therapy. These include specific serine/threonine kinases, the receptor tyrosine kinase AXL, MET, and MAPK-interacting kinases, which contribute to maintaining the MES GSCs phenotype [3, 161–163]. The receptor tyrosine kinase-like orphan receptor 1 is also implicated in promoting the MES phenotype of GSCs [164].

Therapeutic application

PN GSCs and MES GSCs exhibit distinct characteristics and play different roles in the occurrence and development of gliomas. Typically, PN subtypes have a better prognosis and exhibit greater sensitive to radiation compared to the MES subtypes. Consequently, precise identification of the PN and MES subtypes is critical for designing subtype-specific therapeutic regimens, which may significantly improve patient survival outcomes [165]. However, accurate diagnosis of GSCs subtypes remains a substantial clinical challenge due to intratumoral heterogeneity and dynamic phenotypic plasticity.

Potential targets according to GSCs subtypes

Currently, we identify GSCs on the basis of the expression of cell surface markers such as CD133, CD44, and CD15. Some lncRNAs, including miR-155-5p, have been proven to be significantly correlated with the diagnosis and prognosis of gliomas [120, 142, 166]. Therefore, in-depth research into the underlying mechanisms of these GSCs subtypes is vital for enhancing our understanding of their transcriptional networks, which can not only provide more avenues for glioma treatment but also reveal more biomarkers associated with the PN and MES subtypes. These findings will aid in early diagnosis and prognosis assessment and provide a basis for subsequent targeted therapies. E. Agosti et al. reported that inhibiting signaling pathways, such as the Notch, Wnt/β-catenin, STAT3, and PI3K/AKT pathways, in GSCs can suppress the progression of gliomas [167]. Similarly, L. P. Helweg et al. reported that inhibiting the NF-κB signaling pathway effectively suppressed the activity of GSCs and treated gliomas[168] H. C. Hung et al. discovered that manipulating the Shh signaling pathway could induce autophagy and death in GSCs, thereby reducing glioma formation and providing further therapeutic options [169]. We can infer that by targeting specific signaling pathways and their related upstream and downstream molecules that regulate the two subtypes of GSCs, we can develop targeted therapies specific to each subtype, enabling personalized treatment approaches.

The previously mentioned targets, such as Notch, STAT3, TNF-α, NF-κB, and tumor-associated macrophages, have demonstrated significant progress in clinical research. Notably, emerging potential therapeutic targets against GSCs, identified in recent years, have begun to enter clinical evaluation and exhibit considerable research value. These novel targets not only provide innovative therapeutic strategies but also hold promise for improving patient prognosis; a comprehensive summary of candidate agents targeting key signaling pathways in GSCs, along with their clinical status and associated challenges, is provided in Table 2.

Table 2.

GSCs-Targeted Therapy: Candidate Agents for Key Pathways and Clinical Challenges

Target/Signaling	Medicine	Clinical Status	GSCs Evidence	Challenges	Ref
HIPPO/YAP	Metformin	Phase Ⅰ	In vitro	Lack of TAZ suppression; Elevated adverse events incidence with metformin-TMZ combination	[170, 171]
YAP/TAZ-TEAD	Verteporfin	Phase 0	In vitro	Lack of self-paired controls, potential bias	[133, 172]
αvβ3	Cilengitide (halted)	Phase Ⅲ	In vitro/In vivo	Lack of significant benefit	[138, 173]
SOX2 mTOR pathway	Rapamycin	Phase Ⅰ	In vitro	Limited monotherapy efficacy	[174, 175]
NF-κB/STAT3 signaling	GSK343	Preclinical	In vitro/In vivo	GSK343 suppresses EZH2 but fails to inhibit c-Myc in GSCs	[176]
	Melatonin※ (via EZH2)	Preclinical	In vitro/In vivo	Undefined effective dose of melatonin; Significant in vitro-in vivo disparity	[177]
NOTCH signaling			In vitro		[178]
	RO4929097 (GSI)	Phase Ⅱ	In vitro, Phase Ⅱ	Suboptimal efficacy with failed GSC suppression	[179, 180]
HDAC2-c-Myc	Vorinostat (Suberanilohydroxamic Acid/SAHA)	Phase Ⅱ	In vitro/In vivo	No significant improvement in combination therapy outcomes	[181–183]
Shh signaling	Sonidegib	Phase Ⅱ	In vitro/In vivo	Efficacy restricted to medulloblastoma; Limited response in gliomas	[169, 184]
Shh/NF-κB signaling	SANT-1 Guggulsterone	No evidence	In vitro	Synergistic GSCs inhibition demonstrated; Lacks clinical validation	[185]
PI3K/AKT/mTOR signaling	JQ1	No evidence	In vitro/In vivo	No glioma-specific clinical trials conducted	[186]
	ACT001	No evidence	In vitro	Multitarget engagement with unelucidated regulatory mechanisms	[187]
	RSL3	No evidence	In vitro	Inability to cross the blood–brain barrier (BBB)	[188]
	Vistusertib	Phase I/II	In vitro/In vivo	Pediatric trial data only; Insufficient efficacy in unstudied adult population	[189, 190]
	NVP-BEZ235	Phase IIb (Unknown status)	In vitro	Unresolved toxicity profiles; Requires optimized dosing and further preclinical/clinical validation	[191]
	MTI-31	Preclinical	In vitro/In vivo	Potential immunogenic reactions (e.g., anaphylaxis); Needs safety validation	[192]
	Everolimus	phase II	In vitro/In vivo	No significant efficacy improvement with elevated toxicity	[193, 194]
PI3K/AKT signaling ERK signaling NF-κB signaling	Rosmarinic acid	Preclinical	In vitro	Poor BBB penetration; Clinically safe dose below in vitro effective concentration	[195]
NF-κB signaling	CP-673451 (via PDGFR)	Preclinical	In vitro	BBB penetration requires experimental validation; Needs subtype-specific targeting	[196]
	Nabiximols (Cannabinoid)	Phase Ⅰb	In vitro	Significant adverse events; Discordant preclinical-clinical efficacy	[197, 198]
	Disulfiram	Phase Ⅰ/Ⅱ	In vitro/In vivo	Undetectable intratumoral drug levels; Limited therapeutic efficacy	[199, 200]
FOXM1-Survivin	Bortezomib	Phase Ⅱ	In vitro/In vivo	Significant OS improvement with expected safety profile; Requires larger cohorts	[201, 202]
EGFR signaling	Gefitinib	Phase Ⅱ	In vitro/In vivo	Inadequate control of EGFR pathway activity due to signaling network complexity	[203]
Wnt/β-catenin	Trotabresib (BET inhibitor)	Phase Ⅰ	In vitro/In vivo	Limited patient enrollment; Phase Ib trial ongoing (NCT04324840))	[204, 205]
STAT3 signaling	WP1066	Phase Ⅰ	In vitro/In vivo	Unvalidated combination strategies; Needs clinical trial confirmation	[206, 207]
	STX-0119	No evidence	In vitro/In vivo	Unintended KLF-4 activation requires functional clarification	[208]
	WP1193	No evidence	In vitro/In vivo	Compensatory pathway activation; Lacks in situ model validation	[209]
	ODZ10117	No evidence	In vitro/In vivo	Unverified BBB penetration	[210]
	Siramesine	No evidence	In vitro/In vivo	Suboptimal efficacy at low doses; Requires enhanced BBB delivery	[211]
	Napabucasin	No evidence	In vitro/In vivo	Unassessed TME impact	[212]
TGFβ	RGFP966 (via HDAC3)	No evidence	In vitro/In vivo	No glioma-specific clinical trials conducted	[213]
	Galunisertib	Phase Ⅱ	In vitro	Lomustine combination failed to enhance efficacy	[214, 215]
PRMT	SGC707 (via PRMT3)	No evidence	In vitro/In vivo	No glioma-specific clinical trials conducted	[216]
CDK	TG02 (via CDK9)	Phase Ⅰb	In vitro/In vivo	Limited efficacy with exacerbated toxicities when combined with alkylating agents	[217, 218]
	SNS032 (via CDK9)	No evidence	In vitro/In vivo	Poor BBB penetration; Requires clinical validation of human efficacy	[219]
	THZ1 (via CDK7)	Preclinical	In vitro/In vivo	Requires clinical validation of human efficacy	[219, 220]
	CYC065 (via CDK2/9)	Preclinical	In vitro	Requires clinical validation of human efficacy	[220, 221]
	Palbociclib (CDK4/6)	Phase Ⅱ	In vitro/In vivo	No significant efficacy/prognostic biomarker improvement	[193]
CD146	Nanoparticles with folic acid-polyethylene glycol (FA-PEG-COL NPs)	No evidence	In vitro/In vivo	Challenging drug synthesis; No glioma clinical trials	[222]
xCT	Sulfasalazine	Phase I	In vitro/In vivo	Poor aqueous solubility limits dosing; Selective efficacy against GSCs	[223]
TGM2 C/EBPβ signaling	GK921	No evidence	In vitro	Clinical significance of GK921 requires validation	[30]
ALDH1A3	MCI-INI-3	No evidence	In virto	Ineffective monotherapy with limited therapeutic potential; Needs further investigation	[224]

Melatonin exerts inhibitory effects on NOTCH, NF-κB, and STAT3 signaling pathways in GSCs

Challenges and future directions in glioma stem cell-targeted therapy

Model limitations and barriers to clinical translation

Current therapeutic strategies targeting GSCs primarily rely on in vitro systems (e.g., neurosphere cultures) and in vivo models (e.g., primary patient-derived xenograft [PDX] models). Although indispensable, these systems exhibit critical limitations: In vitro neurosphere cultures lack the complex immune components and vascular interactions present in vivo, which critically influence GSCs plasticity and treatment responses. For in vivo models, subcutaneous xenografts in animals are often used for preliminary validation when drugs fail to penetrate the blood-brain barrier (BBB). However, this approach cannot accurately recapitulate drug efficacy in the intracranial microenvironment; further investigation using advanced delivery methods is essential to evaluate therapeutic potential [188]. Meanwhile, emerging human brain organoid models provide a highly biomimetic platform for glioma research by preserving key features of human brain cellular architecture and physiological organization. However, their low maturity and the absence of TME and vascular components still make it difficult to fully simulate the pathological progression of gliomas in the human brain. As a cutting-edge technology, this system holds considerable potential for refinement in precisely modeling intracranial pharmacodynamic responses [225, 226].

Clinical translation also faces numerous obstacles. The inadequate BBB permeability hinders many potent drugs from reaching GSCs niches, severely limiting therapeutic efficacy [195]. Additinoally, GSCs possess intricate regulatory networks, increasing the risk of off-target effects, unintended crosstalk with key signaling pathways, and subsequent failure of targeted therapies [171, 177, 178]. Furthermore, the immunosuppressive tumor microenvironment not only shields GSCs but also diminishes treatment effectiveness [181]. Notably, given the pivotal role of STAT3 in immune regulation, novel STAT3 inhibitors must carefully balance antitumor activity with the preservation of host immune homeostasis [212]. The intratumoral heterogeneity among GSCs leads to selective therapeutic resistance, such as the inhibitory effect of Cilengitide being effective only inhibits the minor subset of GSCs addicted to glucose/Glut3 metabolism [138]. Finally, adaptive feedback resistance remains a major challenge; incomplete pathway inhibition often triggers compensatory activation of alternative signaling axes, ultimately driving drug resistance [194, 203].

Emerging strategies to overcome barriers

To break through current translational bottlenecks in GSCs-targeted therapy, multidimensional innovative strategies are advancing in parallel. In drug delivery technologies, FA-PEG modified nanoparticles achieve tumor enrichment via the enhanced permeability and retention (EPR) effect and are efficiently internalized by glioma cells through folate receptor-mediated active transport, significantly enhancing intracranial targeting efficacy [222]. Engineered exosomes and artificial vesicles offer novel pathways to bypass the BBB [227, 228]. Furthermore, physical delivery technologies have demonstrated substantive progress, Specific Mode Electroacupuncture Stimulation (SMES, NCT06818331) and ultrasound-mediated BBB opening (NCT03744026) have entered clinical validation, substantially improving central nervous system delivery success rates for therapeutic agents.

Combination therapies strategies aim to systematically overcome resistance networks, dual-pathway inhibitors SANT-1 and guggulsterone suppress compensatory escape by blocking Shh/NF-κB crosstalk, while the immunomodulatory combination of SAHA with anti-PD-L1 antibodies directly eliminates GSCs while counteracting Treg-mediated immunosuppression [181, 185]; concurrently, BACE1-targeted reprogramming of tumor-promoting TAMs toward tumor-suppressive phenotypes represents a promising immunoenvironmental modulation strategy [229]. Such regimens have extended to clinical practice, exemplified by ongoing trials of dendritic cell vaccines loaded with GSC antigens combined with PD-1 blockade (NCT04888611). In addition, PMT significantly contributes to enhanced therapeutic resistance, making PMT-targeted strategies critically important. Recent studies have indicated that nefllamapimod can inhibit PDIA3P1-mediated PMT, thereby reducing GSCs resistance and enhancing the efficacy of TMZ in anti-tumor treatments. Furthermore, inhibiting the expression of POLD4 through the inhibitor of UCHL3 can suppress mesenchymal transition, thereby increasing the radiation sensitivity of MES GSCs. The peroxisome proliferator-activated receptor γ, which is exclusively expressed in MES GSCs, modulates PMT by inhibiting the STAT3 pathway, thereby suppressing tumor progression. Additionally, this receptor serves as a diagnostic marker for glioma patients [47, 155, 230]. Elucidating these mechanisms of glioma treatment resistance enables the development of comprehensive combination therapies that provide personalized therapeutic solutions.

The development of precision medicine can provide new paradigms for personalized intervention. An integrated diagnosis-treatment-prognosis scheme based on the MES GSCs specific marker PPARγ and high-throughput drug screening (such as NCT05380349 and NCT05043701) are accelerating the process of translating targeted drugs from mechanism research to clinical application [230].

GSCs-targeted clinical trials: emerging strategies

On the basis of information retrieved from the clinical trial database (https://clinicaltrials.gov/) (Table 3), our investigation revealed that current clinical trials targeting GSCs predominantly focus on mRNA-based immunotherapy utilizing GSCs-derived molecular payloads. Concurrently, research initiatives exploring precision diagnostics and intraoperative localization through GSCs-specific molecular biomarkers are demonstrating progressive advancements. At present, there are few clinical trials specifically targeting unique GSCs markers, and most targeted therapies remain in the preclinical research stage. However, emerging studies such as NCT05380349 and NCT02654964 (Swedish Medical Center), along with NCT05043701 (Oslo University Hospital), are pioneering personalized therapeutic strategies that aim to achieve precise molecular targeting of GSCs. These investigations aim to optimize molecularly tailored therapeutic approaches for improved GSCs eradication.

Table 3.

Clinical Trials Targeting Glioma Stem Cells

NCT number	Title	Target	Status	Intervention
NCT05772767	Modulation of Ciliogenesis in Glioma Stem Cells (RF2019-1236878)	Modulation of Ciliogenesis	Unknown status	Biological sample collection, dissecting ciliogenesis players
NCT04888611	Neoadjuvant PD-1 Antibody Alone or Combined With DC Vaccines for Recurrent Glioblastoma	PD-1, Immunotherapy Targets for GSCs-Associated Antigens	Unknown status	Camrelizumab, GSCs-DCV
NCT01269411	RO4929097 in Treating Patients With Recurrent Invasive Gliomas	p75^NTR, Notch receptor	Terminated	Surgery, RO4929097
NCT05328089	Vacuolar ATPase and Drug Resistance of High Grade Gliomas	V-ATPase	Recruiting	Proton Pump Inhibitors (PPI)
NCT06348693	Development of Therapeutic Approaches Modulating Molecular Targets Implicated on Cancer Stem Cell-related Aggressiveness	Sphingosine-1-phosphate (S1P) signaling pathway	Recruiting	Tumor biopsies and biomarker investigation
NCT03548571	Dendritic Cell Immunotherapy Against Cancer Stem Cells in Glioblastoma Patients Receiving Standard Therapy (DEN-STEM)	Survivin, hTERT, Immunotherapy Targets for GSCs-Associated Antigens	Recruiting	Dendritic cell immunization, Adjuvant temozolomide
NCT02010606	Phase I Study of a Dendritic Cell Vaccine for Patients With Either Newly Diagnosed or Recurrent Glioblastoma	Immunotherapy Targets for GSCs-Associated Antigens	Completed	Dendritic cell vaccination, standard temozolomide chemotherapy and radiation therapy
NCT02820584	A Phase I Study of Immunotherapy With GSC -Loaded Dendritic Cells in Patients With Recurrent Glioblastoma	Immunotherapy Targets for GSCs-Associated Antigens	Completed	GSCs-loaded autologous dendritic cells
NCT00846456	Safe Study of Dendritic Cell (DC) Based Therapy Targeting Tumor Stem Cells in Glioblastoma	Immunotherapy Targets for GSCs-Associated Antigens	Completed	Dendritic cell vaccine with mRNA from stem cells
NCT00890032	Vaccine Therapy in Treating Patients Undergoing Surgery for Recurrent Glioblastoma Multiforme	Immunotherapy Targets for GSCs-Associated Antigens	Completed	Dendritic Cells
NCT01171469	Vaccination With Dendritic Cells Loaded With Brain Tumor Stem Cells for Progressive Malignant Brain Tumor	Immunotherapy Targets for GSCs-Associated Antigens	Completed	Dendritic Cells; Imiquimod
NCT05341947	Activated Autologous T Cells Against Glioma Cancer Stem Cell Antigens for Patients With Recurrent Glioblastoma	Immunotherapy Targets for GSCs-Associated Antigens	Not yet recruiting	Activated T cells
NCT05380349	Personalized Cancer Stem Cell High-Throughput Drug Screening for Glioblastoma	GSCs	Recruiting	Chemotherapy
NCT05043701	Individualized Systems Medicine Functional Profiling for Recurrent Glioblastoma (ISM-GBM)	GSCs	Recruiting	Chemotherapy
NCT02654964	Cancer Stem Cell High-Throughput Drug Screening Study	GSCs	Completed	Chemotherapy
NCT05556486	Mapping of Tumor Stem Cells in the Resection Marigin During Extirpation of Highly Malignant Gliomas Using GlioStem (CeNo2)	The internal structure of the GSCs	Recruiting	GlioStem (stem cell marker)
NCT04868396	Patient-derived Glioma Stem Cell Organoids	GSCs	Active, not recruiting	Tumor biopsy
NCT02039778	Stem Cell Radiotherapy and Temozolomide for Newly Diagnosed High-grade Glioma (STRONG)	GSCs	Terminated	Stem Cell Radiotherapy and Temozolomide
NCT01872221	Study of the Capacity of the MRI Spectroscopy to Define the Tumor Area Enriched in Glioblastoma Stem Cells. Proof of Concept Study (STEMRI)	GSCs	Completed	Surgery, the standard radio-chemotherapy stupp protocol

Conclusions

Glioma stem cells exhibit remarkable phenotypic plasticity between the proneural and mesenchymal subtypes, driven by dynamic transcriptional networks, metabolic reprogramming, and bidirectional interactions with the tumor microenvironment. The MES phenotype, characterized by increased invasiveness, therapy resistance, and immunosuppressive TME remodeling, represents a critical therapeutic target. Key regulatory axes such as the NF-κB/STAT3/C/EBPβ, Wnt5a/β-catenin, and YAP/TAZ- HIPPO pathways converge to sustain MES traits, whereas hypoxia, radiotherapy, and extracellular vesicle-mediated crosstalk potentiate PMT. Notably, metabolic enzymes (e.g., ALDH1A3, NNMT), epigenetic modifiers (e.g., HDAC1, m6A regulators), and immune-evasive mechanisms (e.g., C3a-TAMs polarization) synergistically reinforce MES dominance.

Emerging strategies to counteract therapeutic resistance include dual targeting of PMT drivers and stemness pathways (e.g., the USP9X/FOXD1-ALDH1A3 axis), and disrupting immunosuppressive niches via M2-TAMs repolarization. Recognition of subtype-specific vulnerabilities, such as PN dependence on Notch/Wnt signaling and MES addiction to NF-κB, supports personalized combinatorial regimens. Promising clinical targets (PRMT1 and CD146) and diagnostics are transitioning from preclinical validation to early-phase trials, yet challenges persist in overcoming intratumoral heterogeneity and blood-brain barrier penetration.

Future efforts should prioritize: 1) The development of microenvironment-modulating agents (e.g., TGF-β/IL-6 inhibitors) to block PMT-inducing signals; 2) Biomarker-guided therapeutic sequencing to exploit metabolic-epigenetic interdependencies; 3) Engineering nanocarriers for dual targeting of GSCs and their vascular-immune niches. Unraveling the contextual regulation of GSCs plasticity will be paramount for achieving durable remission in glioblastoma.

Abbreviations

GSCs

Glioma stem cells

MES

Mesenchymal subtype

PN

Proneural subtype

PMT

Proneural-to-mesenchymal transition

OCT4

Octamer-binding transcription factor 4

Olig2

Oligodendrocyte transcription factor 2

SOX2

Sex-determining region Y-box 2

NFAT

Nuclear factor of activated T cell

TNF-α

Tumor necrosis factor-alpha

ASCL1

Achaete-scute homolog 1 gene

EGFR

Epidermal growth factor receptor

FZD4

Frizzled class receptor 4

PATZ1

POZ, AT-hook and zinc-finger 1

TAMs

Tumor-associated macrophages

Chi3l1/YKL40

Chitinase 3-like 1

ALDH

Aldehyde dehydrogenase

USP

Ubiquitin-specific protease

ID1

Inhibitor of DNA binding 1 protein

EVs

Extracellular vesicles

lncRNAs

Long non-coding RNAs

m6A

N6-methyladenosine

GRP78

Glucose-regulated protein 78

SPI1

Salmonella pathogenicity island 1

ITGA

Integrin alpha

MLK4

Mixed lineage kinase 4

GPR

G-protein-coupled receptor

PRMT

Protein arginine methyltransferase

CDK

Cyclin-dependent kinases

xCT

The cystine-glutamate transporter

CLIC1

Chloride intracellular channel-1

EphA2

Eph Receptor A2

DRD1

D1 dopamine receptor

CSNK1D

Encode casein kinase 1δ

LMO2

Lim domain only 2

TRF2

Telomere repeat-binding factor 2

KLF6

Kruppel-like factor 6

FMR1

Fragile X messenger ribonucleoprotein 1

HIF-1

Hypoxia-inducible factor-1

HOXC8

Homeobox C8

FOX

Forkhead box

Author contributions

Guofeng Tian: Conceptualization, formal analysis, writing – original draft, writing – review & editing, and images. Yifu Song: Writing – review & editing, Investigation, Validation. Yaochuan Zhang: Formal analysis, Investigation, Validation. Liang Kan: Writing – review & editing & images, Design, Supervision. Ana Hou: Writing – review & editing, Design, Validation, Supervision. Sheng Han: Writing – review & editing. Conceptualization, Resources, Design, Supervision, Funding acquisition.

Funding

This work was supported by grants from the National Natural Science Foundation of China (nos. 82273482 and 81772653).

Data availability

The datasets generated and/or analysed during the current study are available in the [ClinicalTrials.gov] repository, [https://clinicaltrials.gov/].

Declarations

Competing interests

The authors declare no potential conflicts of interest.