The MES-like type is divided into two subgroups: MES1-like and MES2-like (11). Both subgroups are characterized by the expression of the mesenchymal gene VIM, which encodes vimentins, a class III intermediate filament protein. Both subgroups also exhibit activity of ANXA2, LGALS3, MT1X, and NAMPT. The MES1-like subgroup shows activity in genes involved in immune and inflammatory responses (ANXA1, CHI3L1, IFITM3, C1S, C1R, C3, SERPING1, TNFRSF1A, SERPINA3, MGST1), cellular adhesion, interactions with the extracellular matrix (CD44, FN1, SPP1, LGALS1, LGALS3), antioxidant activity (SOD2, PRDX6, MT2A), and cell proliferation (TIMP1, IGFBP7, WWTR1, S100A10, S100A11). The MES2-like subgroup, on the other hand, is associated with the expression of genes linked to hypoxia adaptation, stress responses, and apoptosis (HILPDA, DDIT3, NDRG1, HERPUD1, TRIB3, EPAS1, EGLN3, BNIP3, BNIP3L, GDF15, NAMPT, ADM), glucose metabolism (ENO2, SLC2A1, SLC2A3, PFKP, PGK1, LDHA), and cell proliferation (CDKN1A, IGFBP3). Thus, the MES2-like state, in some tumors, may be linked to hypoxia and high glucose consumption. MES1-like has been identified as a hypoxia-independent subtype, while MES2-like is considered hypoxia-dependent.

The AC-like subgroup expresses markers typical of astrocytic glia (GFAP, MLC1, AQP4, S100B, GPM6B) (11). This group is also characterized by the expression of genes associated with apoptosis, cell growth, and differentiation (TSPAN7, PTPRZ1, HEPN1, NDRG2, CLU, HOPX), signaling molecules (RAMP1, EDNRB, AGT), antioxidant activity and cellular metabolism (PON2, S100A16, CST3, PLTP, PPAP2B, RAB31, DBI, PMP2, GATM, SPARC, SPARCL1, TTYH1, ATP1A2, ATP1B2, PMP22), coagulation (F3, ANXA5), and metal-binding protein (MT3). Additionally, METTL7B, a gene associated with glioblastoma cell growth and survival, is expressed in this group (12).

The OPC-like group is characterized by broad expression of two groups of genes: the first group is involved in myelination processes and is actively expressed in oligodendrocytes (PLLP, PLP1, CNP, OMG, BCAS1, SIRT2, PMP2, GPM6B, RTKN, P2RX7, CADM2, PSAT1), while the second group is predominantly expressed in oligodendrocyte progenitors (LHFPL3, GPR17, OLIG1, NKAIN4, THY1, FGF12, DBI, LPPR1, RAB31, NLGN3, NEU4, HRASLS, SERINC5, SCRG1, TNS3, GPR37L1, BCAN, VCAN, PTPRZ1, CNTN1, FIBIN, APOD) (11). Additionally, the OPC-like group may express genes related to ion channels (TTYH1, TMEM206, FXYD6), cell proliferation and division (RNF13, EPB41L2), and adhesion and intercellular interactions (ALCAM, PCDHGC3, PGRMC1).

Like the MES-like group, the NPC-like group is also divided into two subgroups: NPC1-like and NPC2-like (11). Both subgroups are characterized by the expression of genes involved in neurogenesis processes (DBN1, DCX, ELAVL4, SOX11, STMN1, TAGLN3, TUBB3), as well as cell differentiation and proliferation (CD24, HN1, MLLT11, SOX4). In addition to the activity of common genes for both NPC1 and NPC2 that are active in neural development, each group also expresses a specific set of genes. For the NPC1-like subgroup, the characteristic expression includes genes such as ASCL1, HES6, OLIG1, MYT1, TCF12, BEX1, CHD7, GPR56, DLL1, DLL3, ETV1, TSPAN13, PCBP4, SEZ6L, SEZ6, MAP2, NEU4, HIP1, MARCKSL1, BEX1. Notably, there is also expression of the glutamate kainate receptor and the enzyme involved in the utilization of gamma-aminobutyric acid (GRIK2, ABAT), as well as neural-specific adhesion proteins (TNR, NXPH1), which are typical of neurons in the brain. In the case of NPC2-like, the expression is characterized by a broad spectrum of genes crucial for neural development, encoding cytoskeletal proteins and motor proteins: MAP1B, TUBB2A, DPYSL3, DPYSL5, STMN2, STMN4, FNBP1L, PAK3, RND3, MIAT, KIF5C, KIF5A, DYNC1I1. Genes specific to this subgroup, which are involved in neural system development, include TCF4, RBFOX2, NFIB, as well as genes active in mature neurons such as HMP19, NREP, NSG1, SNAP25, ENO2, SEPT3, and CD200.

Opposite to the MES-like subtype, the other subtypes are characterized by the expression of genes related either to neurons/neuronal glia or their precursor cells. These subtypes can coexist within a single tumor and, importantly, can transit into one another during biochemical reprogramming (stress, hypoxia) (13). Thus, the previously discussed bulk classification shows strong consistency with the new single-cell classification. The classical and mesenchymal subtypes align well with the AC-like and MES-like subtypes, respectively. The proneural subtype corresponds to a combination of two cellular states: NPC-like and OPC-like, reflecting their typical coexistence within the same tumor. According to observations, the AC-like phenotype in half of the cases undergoes transition to MES-like during disease progression. In the MES-like state, there is a high infiltration of stromal and myeloid cells, which may contribute to the more malignant course of the disease in individuals with this phenotype (14). The previously described neural type likely consists largely of oligodendrocytes and neurons rather than malignant cells. As with the classical type, according to the bulk classification, the AC-like phenotype is characterized by the overexpression of EGFR. This is likely due to EGFR’s involvement in astrocyte differentiation, and thus, the overexpression of this gene may secondarily contribute to the manifestation of the AC-like phenotype (14). Additionally, PDGFRA and CDK4 are associated with OPC-like and NPC-like phenotypes, respectively, which is not surprising given the important roles of these genes in oligodendrocyte and neuron development (15, 16). NF1, expressed in the mesenchymal type according to the bulk classification, is also characteristic of the MES-like phenotype.

Another characteristic of the tumor under consideration is its heterogeneity in terms of the various subtypes within a single tumor and the differing copy numbers of genes such as EGFR, PTEN, and PDGFR. An increase in the number of copies of these genes is negatively correlated with patients’ survival rate (17, 18). This heterogeneity partly explains the disappointing results of the Rindopepimut trial, a cancer vaccine targeting EGFR-mutant glioblastoma cells (19).

In several malignant neoplasms, subpopulations of cells with stem cell properties have been identified, such as self-renewal capacity, oncogenic potential, and expression of embryonic or tissue-specific stem cell genes. These cells are referred to as cancer stem cells (CSC). In glioblastoma, such subpopulations are called glioblastoma stem cells (GSCs). GSCs exhibit significant resistance to radio- and chemotherapy, and their interaction with the tumor microenvironment profoundly impacts genetic reprogramming and immune resistance (20, 21). Markers of GSCs include CD133, MYC, A2B5, CD24, CD44, L1CAM, EGFR, PDGFRA, SOX2, OCT4, BRN2, OLIG2, ID1, FUT4, PROM1, and NES (21–26). It has been shown that each of the four types (OPC, NPC, AC, and MES) exhibits significant deviations in the expression of specific GSC markers. Thus, CD24 is characteristic of NPC-like GSCs, CD133 of OPC-like, EGFR and NES of AC-like, and CD44 of MES-like. The expression of other markers was either characteristic of two types or lacked specificity altogether. Since GSCs represent a small subpopulation of the entire tumor, their identification is challenging (27, 28). The role and functions of GSCs in glioblastoma are poorly understood and remain an area of particular interest for researchers.

Among macrophages, it is important to distinguish between resident elements – microglia, and macrophages derived from blood monocytes. Microglia is a key stationary component of the central nervous system (CNS) and constitute a significant portion of its cells, originating from the embryonic yolk sac ( Figure 3 ). Pathological conditions involving microglia are associated with severe neurodegenerative diseases (33). In contrast to normal physiology, some microglial cells acquire a pro-inflammatory phenotype in tumor contexts, producing signaling molecules such as IL-1α, IL-1β, TNF, and CXCL10 (34). This transformation can be initiated by tumor cells through TGFβ signaling. Specifically, pro-inflammatory microglia secrete IL-1β via the assembly of the NLRP1 inflammasome, activated by apolipoprotein E, thereby contributing to disease progression. Disruption of TGFβ signaling reduces the population of pro-inflammatory microglia and slows tumor growth (35). In murine models, cancer-mediated activation of mTORC1 in microglia has been shown to induce secretion of the anti-inflammatory cytokines IL10 and IL6. This anti-inflammatory signaling inhibits T cell infiltration into the tumor, thereby promoting tumor growth (36). Another study highlighted the role of selectins, which are involved in the anti-cancer immune response modulation. P-selectin mediates enhanced proliferation and invasion of glioblastoma by altering the activation state of microglia and macrophages. Pharmacological inhibition of P-selectin results in reduced tumor growth and increased survival in rodent models of glioblastoma (37). Murine glioblastoma models have shown tumor-induced reprogramming of microglia, with a reduction in the expression of genes promoting tumor cell destruction and an increase in the expression of genes supporting tumor growth (38). Microglial cells are predominant in newly diagnosed glioblastoma, while macrophages prevail in recurrent glioblastoma (39). Compared to the GBM IDH-mutant phenotype, microglia in GBM IDH-wildtype are characterized by the expression of CD14 and CD64 (40). The advent of single-cell RNA sequencing has expanded the understanding of microglial heterogeneity due to its ability to detect rare cell populations. One such population of microglia expressed genes encoding MHC I and MHC II, while another population expressed genes related to cell proliferation (CDK1, STMN1, TUBA1b, TUBB5, and TOP2A). Microglia with high MHC II expression have more genes coding for chemokines (CCL3, CCL4, CCL12), suggesting that this subset of microglia may contribute to the recruitment of other immune cells (41). A new population of immunosuppressive microglia, CD163⁺HMOX1⁺, was also identified, exhibiting anti-T cell activity through the secretion of IL-10, and this population is found to be limited in mesenchymal glioblastomas (42).

Macrophages are tissue monocytes originating from myeloid progenitors in the red bone marrow. Traditionally, macrophages are categorized into M1, the pro-inflammatory subset, and M2, the anti-inflammatory subset. However, it is challenging to clearly distinguish between these two populations, as tumor-associated macrophages (TAMs) often express markers from both groups. It is believed that in the early stages of disease, M1 macrophages predominate in tumors and limit their growth. However, tumors can recruit macrophages through a process known as M2 polarization. The M1/M2 ratio shifts towards M2 macrophages as the disease progresses, negatively impacting survival (43, 44). M1 macrophages are typically activated by interferon-gamma (IFN-γ), lipopolysaccharides from Gram-negative bacteria, tumor necrosis factor-alpha (TNF-α), and Toll-like receptors 2 and 4 (TLR2/4). M1 macrophages can induce differentiation of T cells into type 1 helper phenotype (Th1) by secreting IL-12 and activate NK cells by producing pro-inflammatory cytokines such as TNF-α, IL-1β, IL-6, IL-8, and IL-23, thereby stimulating cytotoxic responses (45–48). M2 macrophages are activated via the peroxisome proliferator-activated receptor gamma (PPARγ) and STAT6 (49, 50). The activity of the M2 macrophage subset is associated with the stimulation of Th2 T cells and regulatory T cells. M2 macrophages produce IL-1RA, IL-10, vascular endothelial growth factor (VEGF), and transforming growth factor-beta (TGF-β), which promote tumor growth (46, 48, 49). The M1/M2 polarization process correlates with the activation of several major signaling pathways, including JNK, PI3K/Akt, Notch, and JAK/STAT (50, 51). In glioma macrophages, CCL3 is a marker of the anti-tumor fraction, while CD68 and CD163 are markers of the pro-tumor fraction (52). The number of CD204⁺ macrophages increases with the malignancy grade of glioma and may contribute to the pro-tumor transformation of the glioma microenvironment. CD204⁺ macrophages co-express MMP14 and HIF-1α, factors potentially linked to the aggressiveness of glioma. In Grade III-IV gliomas, CD204 expression is associated with poor prognosis (53). Glioblastoma cells can produce Wnt-induced signaling protein 1 (WISP-1), thus promoting a pro-tumor microenvironment by enhancing the survival of glioblastoma cells on one side and tumor-associated macrophages on the other (54). By synthesizing colony-stimulating factor (CSF2), glioma cells promote M2 polarization of macrophages and, consequently, tumor growth (55). Another protein supporting M2 macrophages is osteopontin (SPP1). The level SPP1 expression correlates with glioma malignancy grade and the level of macrophage infiltration. In contrast, low osteopontin levels are associated with low levels of M2 macrophages and high levels of CD8⁺ cytotoxic cells, which have anti-tumor activity (56). CD11b⁺/CD163⁺ macrophages can stimulate the growth and survival of glioblastoma cells through PTN-PTPRZ1 signaling (57). CECR1 is a potential protein regulating M2 polarization of macrophages and is widely synthesized by these macrophages. It is suggested that this protein may stimulate migration and proliferation processes in glioblastoma cells through MAPK signaling (58).

As mentioned earlier, distinguishing between M1 and M2 macrophage subpopulations is a challenging. M1 macrophages are typically characterized by the expression of CD64 (FCGR1A), SOCS1, IL1R1, TLR2, TLR4, CSF2 (GM-CSF), CD169 (SIGLEC1), CD80, CD86, HLA-DR, and CD197 (CCR7), while M2 macrophages are marked by the hyperexpression of MRC1 (CD206), TGM2, CCL22, TLR1, TLR8, FCGR3A (CD16), CD200R1, MSR1 (CD204), CD163, CD209, and CCL2. There are also significant differences in the secreted biomolecules: M1 macrophages typically produce ROS, iNOS, TNF-α, IL-1β, high levels of IL-6, IL-12, IL-23, and low levels of IL-10; whereas M2 macrophages are characterized by the secretion of IL-8, high levels of IL-10, Arg-1, CCL2, CCL5, CCL17, CCL22, PDGF, VEGF, and MMPs (48, 50, 59). Distinguishing microglia from macrophages is also problematic due to the expression of classical markers for these cell populations, which may be significantly altered during oncopathophysiological reprogramming. For example, classical markers used to identify macrophages, such as CD45 and CCR2, can start being expressed by microglial cells in brain gliomas. In order to discriminate them, microglia can be identified by the expression of typical genes such as P2RY13, P2RY12, GPR34, SLC2A5, OLFML3, TMEM119, FCRLA, SALL1, and TREM2. In contrast, macrophages can be identified using markers such as F10, EMILIN2, F5, C3, GDA, MKI67, SELL, HP, ICAM1, CCR2, C1QA, CD207, and CD209 ( Table 1 ) (39, 41, 60–65).

B lymphocytes are a distinct class of immune cells, the main characteristic of which is the presence of specific receptors for antigens on their surface. These receptors are based on immunoglobulins, complex molecules that are specific to a single antigen. After interacting with the antigen, the B cell exits the bloodstream and differentiates into a plasma cell that produces immunoglobulins (antibodies) against the recognized antigen. Thus, B lymphocytes are the primary cellular substrate for humoral immunity. The role of these cells in oncology is controversial: according to some data, they have a pro-tumor effect, while other data suggest they have an anti-tumor effect (66). According to other data, glioma cells can induce the differentiation of naive B lymphocytes into TGF-β ⁺ B regulatory cells through the synthesis of PlGF. In turn, B regulatory cells, similarly to the mechanism described above, can inhibit anti-cancer T cell immunity (67). Lee-Chang C et al. demonstrated the dual nature of B cells as local elimination of tumor-associated B lymphocytes in implanted human glioblastomas in mice via anti-CD20 immunotherapy (intracranial injection) nearly doubled the survival of the mice. In contrast, its systemic administration via intraperitoneal injection showed no benefit. This observation helps explain the dual role of these cells depending on their location. After implantation of glioblastoma tissue in mice, there was an increase in B lymphocyte infiltration (mainly in perivascular areas) of the tumor stroma, which correlated with overall B lymphopenia. Within the glioblastoma, B cells appear to exert an immunosuppressive effect, while systemically, they possess anti-tumor properties. This pro-tumoral effect is mediated through the synthesis of inhibitory molecules such as PD-L1 and CD155 and immune-regulatory cytokines like TGFβ and IL-10. The result, as seen in other cases, is the differentiation of naive B cells into mature B regulatory cells and subsequent suppression of CD8⁺ cytotoxic T cells (by inhibiting the expression of GZMB and IFNγ by these cells) (68, 69). Another study demonstrated the pro-inflammatory role of so-called 4-1BBL⁺ B lymphocytes, which exhibit high intracellular levels of TNF and IFNγ, as well as increased expression of genes CD86 and CD69, indicating their activated state. The level of these B lymphocytes positively correlates with the level of CD69⁺CD8⁺ T lymphocytes. The ability of B cells to stimulate CD8⁺ T cells largely depends on 4-1BBL expression. The increase in 4-1BBL levels in B cells occurs after stimulation of the B-cell receptor (BCR) and co-stimulation of CD40 (70, 71). Upon local (intracranial) injection of 4-1BBL⁺ B lymphocytes in mice with implanted glioblastoma, an increase in tumor infiltration by CD8⁺ T lymphocytes producing GzmB and IFNγ was observed. These mice lived longer than the control group (71). Thus, the role of B lymphocytes in glioblastoma pathogenesis may be of great importance. Tumor cells appear to recruit B cells, promoting their differentiation into immunosuppressive B regulatory lymphocytes, thereby supporting tumor survival and immune evasion through suppression of the T cell response. As demonstrated by the experiments of Lee-Chang C et al., activated B lymphocytes could be a potential target for glioblastoma immunotherapy. B lymphocytes are identified as CD45⁺CD11b⁻CD19⁺CD20⁺ cells.

T cells are a subpopulation of lymphocytes derived from hematopoietic stem cells in the bone marrow. The distinguishing feature of T lymphocytes is their unique mechanism of antigen recognition. They recognize peptide fragments of foreign proteins that are embedded in autologous major histocompatibility complex (MHC) molecules. This molecular complex is presented to them by antigen-presenting cells (such as macrophages and dendritic cells) through interaction with the T-cell receptor (TCR). There are three main groups of T lymphocytes: T-helper cells, CD8⁺ T-killer (cytotoxic T) cells, and T-regulatory cells. All T lymphocytes share common markers, including CD3 and TCR receptors ( Table 1 ) (72). While the heterogeneity of T cells extends beyond these subclusters, they represent the most significant subpopulations of interest.

Natural killer (NK) cells are a crucial component of the innate immune system and act as the first line of defense against oncological diseases. A key distinguishing feature of NK cells is their ability to induce target cell lysis without prior antigen contact or activation (90). NK cells constitute approximately 10-15% of the lymphocytes in peripheral blood. Like CD8⁺ T cells, NK cells contain intracellular granules with perforin – a protein that forms pores in target cell membranes – and granzymes, which induce apoptosis. As mentioned above, tumor cells may downregulate the expression of MHC class I molecules to evade cytotoxic effects from CD8⁺ T cells. However, this mechanism can lead to NK cell activation, as NK cells express inhibitory receptors that interact with MHC I proteins to suppress their activation (91, 92). To facilitate such interactions, NK cells possess immunoglobulin-like receptors known as Killer Immunoglobulin-like Receptors (KIRs) on their surface. These receptors can engage with various forms of MHC class I molecules, such as HLA-A, HLA-B, and HLA-C. Another receptor that inhibits the cytotoxic activity of NK cells is the CD94/NKG2A complex, which specifically interacts with the HLA-E variant of MHC class I (93). Thus, the recognition of MHC class I proteins inhibits the activity of natural killer (NK) cells.

Additionally, virus-infected or tumor cells may express non-classical MHC class I molecules, such as MICA and MICB. Interaction of these molecules with the activating receptor NKG2D on NK cells can also trigger NK cell activation and subsequent cytolysis (94). Traditionally, natural killer (NK) cells are divided into two major subsets: CD56-dim/CD16⁺ and CD56-bright/CD16⁻. The first subset (referred to as NK1) has low expression of CD56 but high expression of CD16. NK1 cells constitute the majority of the NK cell population and exhibit pronounced cytotoxic activity. Due to their cytotoxic nature, NK1 cells express genes essential for this function, including GZMB, PRF, EMP3, ITGB2, and EB2. In contrast, the second subset (NK2) makes up a smaller proportion of NK cells and is characterized by high cytokine production and regulatory functions. NK2 cells are distinguished by their expression of XCL1 (95, 96). A reliable marker for identifying NK cells is NCR1 (NKp46) (66). Other potential markers include KLRD1 (CD94), FCGR3A, KLRC1, KLRC3, KLRB1, KLRC2, NCR3 (NKp30), and NCR2 (NKp44) ( Table 1 ). Chemokine receptors (CCR2, CCR5, CXCR3, and CX3CR1) on NK cells contribute to the anti-inflammatory response and the migration of these cells to sites of inflammation in response to CCL2, CCL3, CCL5, CCL8, CCL9, CCL11, CCL13, CXCL9, CXCL10, CXCL11, and CX3CL1. High infiltration of NK cells in renal cancer correlates with a positive prognosis (97). Glioblastoma cells can downregulate the expression of the activating NK cell receptor NKG2D through a TGFβ-dependent pathway, confirming the significant role of TGFβ as a key molecular mechanism for inhibiting the immune response by glial cells (98, 99). On the other hand, glioma cells exhibit increased expression of MHC I molecules, which also inhibits NK cell activity via the aforementioned mechanism (100).

MSCs are a type of non-malignant stromal stem cells found in many tissues. One of their important functions is to migrate to sites of injury and support regeneration. In addition to tissue-specific MSCs, there are also bone marrow-derived MSCs. MSCs can differentiate into almost any type of stromal cell, including osteocytes, chondrocytes, adipocytes, fibroblasts, etc. MSCs are typically positive for CD90, CD105, CD73, and negative for CD34, CD45, CD14, HLA-DR, CD19, CD79α, and CD11b. Tumor cells can initiate genetic remodeling of MSCs, converting them into cancer-associated MSCs (CA-MSCs). Glioblastoma cells can recruit MSCs by secreting IL-8, TGF-ss1, NT-3 and VEGF. Evidence suggests that CA-MSCs play a key role in the formation of the TME in various cancers. CA-MSCs and their derivatives are capable of initiating angiogenesis, remodeling the TME, enhancing glioblastoma cell invasiveness, supporting tumor growth, protecting tumor cells from apoptosis, attracting immune cells and reprogramming them toward immunosuppression, maintaining and promoting GSCs differentiation, and initiating resistance to chemotherapy (101–108).

Fibroblasts are an essential component of connective tissue cells. They synthesize the extracellular matrix (glycoproteins, proteoglycans, and hyaluronic acid), elastin, and collagen. Cancer-associated fibroblasts (CAFs) have been identified in the stroma of many tumors and are critical to disease progression (109) ( Figure 4 ). For a long time, CAFs were thought to be absent in glioblastomas due to the lack of fibroblasts in brain tissue. However, cell populations expressing fibroblast markers have been described in gliomas (110–112). CA-MSCs are considered precursor cells for CAFs. Unlike CA-MSCs, CAFs lose the ability for self-renewal but acquire genetic markers of fibroblasts (102, 112). The expression level of CAF markers in gliomas positively correlates with tumor grade and serves as a poor prognostic indicator. CAFs appear to promote the transformation of the glioblastoma phenotype into a mesenchymal subtype (113).

The role of CAFs in M2 polarization of macrophages in gliomas has been demonstrated through IGFBP2 expression (114). M2 macrophage polarization occurs through fibronectin (FN1) and its interaction with the TLR4 receptor on macrophage surfaces. CAFs can enrich the GSCs population through the synthesis of HGF and osteopontin. In turn, GSCs affect CAFs via PDGF and TGF-β. Furthermore, adding CAFs to a GSC population in vitro promoted tumor growth. The expression of LRP10 in GBM correlates with the extent of tumor stroma infiltration by CAFs, and the knockdown of LRP10 significantly restricted the invasive behavior of GBM (115). Saket Jain et al. identified 18 markers of CAFs: ACTA2, FAP, PDGFRA, PDGFRB, PDPN, S100A4, TNC, VIM, COL1A1, CCDC80, BGN, COL4A2, COL5A1, NR2F2, COL3A1, INHBA, STC2, and LOXL2 (116). However, according to Wu Lili et al. the most reliable CAF markers in GBM are ACTA2, COL1A1, and TNC ( Table 1 ) (115).