Glioblastoma (GBM) establishes a profoundly immunosuppressive tumor microenvironment (TME) that blunts cytotoxic immunity, primarily through T-cell dysfunction. Tumor-infiltrating lymphocytes (TILs) are scarce, undergo limited expansion, produce few cytokines, and are frequently confined to perivascular niches rather than infiltrating tumor cores 18,19. Single-cell RNA sequencing and T-cell receptor (TCR) repertoire analyses confirm that the majority of CD8⁺ TILs acquire a terminally exhausted phenotype, characterized by high expression of PD-1 (PDCD1), TIM-3 (HAVCR2), and LAG-3 (LAG3) 20. These co-inhibitory receptors are sustained by chronic antigen stimulation within the antigen-rich yet immunologically inert GBM milieu 21. Exhaustion is further stabilized by transcriptional regulators such as TOX and the NR4A family (NR4A1/2/3). TOX functions as a master epigenetic organizer, remodeling chromatin to lock dysfunctional transcriptional programs while repressing effector signatures 22,23. NR4A factors likewise impose tolerogenic states and directly suppress IL-2 and IFN-γ production 24. In concert with checkpoint signaling, these factors constrain metabolic plasticity. The hypoxic and nutrient-poor GBM TME exacerbates dysfunction by depleting ATP, elevating reactive oxygen species (ROS) and limiting mitochondrial biogenesis 25. Spatial profiling indicates that exhaustion is maximal at the tumor margin, whereas the core is almost devoid of TILs 26, underscoring both focal immune suppression and an epigenetically fixed exhausted state refractory to reversal.

A major barrier to effective GBM immunotherapy is the epigenetic fixation of T-cell exhaustion, whereby dysfunctional programs become durably imprinted and refractory to re-invigoration. Exhausted T cells acquire not only altered transcriptional states but also a distinct epigenomic identity that irrevocably separates them from memory and effector lineages 27. DNA methylation is central to this process. Whole-genome bisulfite sequencing and chromatin accessibility profiling reveal hypermethylation of loci governing IL-2 signaling, cytolytic effectors (e.g., GZMB, PRF1), and co-stimulatory receptors such as CD28 28. DNMT3A serves as a critical mediator; conditional deletion of Dnmt3a in CD8⁺ T cells partially restores effector function and increases responsiveness to PD-1 blockade in pre-clinical GBM models 29. Raman micro-spectroscopy further demonstrates methylation-driven chromatin compaction within exhausted nuclei.

Histone modifications consolidate repression. Trimethylation of histone H3 lysine 27 (H3K27me3), catalyzed by EZH2 within the polycomb repressive complex 2 (PRC2), is enriched at effector cytokine and receptor loci 30. Chromatin immunoprecipitation sequencing (ChIP-seq) confirms pervasive H3K27me3-marked silencing in exhausted TILs from human GBM 31. Pharmacological EZH2 inhibition synergizes with PD-1 blockade, restoring IFN-γ production and proliferation in glioma models 32. Single-cell ATAC-seq and CUT&Tag provide high-resolution maps of this chromatin landscape. Patient-derived exhausted T cells exhibit loss of accessible enhancers governing effector and memory programs 33. Exhaustion-specific super-enhancers regulated by TOX display H3K4me1 enrichment coupled with H3K27ac depletion, indicative of functional silencing 34. These chromatin alterations persist ex vivo, emphasizing durable epigenetic entrapment. Collectively, DNA methylation, histone modifications, and chromatin compaction entrench a fate-committed exhaustion program. Even under checkpoint blockade, exhausted T cells rarely reacquire effector competence. Consequently, of checkpoint inhibitors (e.g., HDAC, BET, or EZH2 inhibitors) are under investigation to restore T-cell plasticity and overcome GBM-mediated immune resistance 35,36.

Raman spectroscopy relies on the inelastic scattering of monochromatic light (typically a laser) to generate vibrational energy profiles that are characteristic of molecular bonds. Its capacity to differentiate nucleic-acid structures, lipid content and protein conformations without exogenous stains makes it invaluable for monitoring epigenetic states 39. Raman signals are particularly sensitive to histone modifications and DNA methylation, associated with spectral shifts in phosphate and acetyl functional groups 40. Although conventional Raman microscopy is diffraction-limited to sub-micron resolution, advanced approaches such as tip-enhanced Raman spectroscopy (TERS) push sensitivity into the nanoscale. In routine practice, conventional Raman requires approximately 10³–10⁴ leukocytes for reproducible spectra, with detection limits in the 10^-6–10^-8 M range. However, signal quality is markedly compromised in FFPE brain tissue because of strong paraffin autofluorescence.

Fourier-transform infrared (FTIR) spectroscopy probing molecular vibrations in the mid-infrared region is highly sensitive to chemical modifications in chromatin, including acetylation and methylation, as well as to global metabolic shifts in lipid and carbohydrate pools 41. FTIR also enables mapping of epigenetic compartmentalization within the nucleus through differential absorbance 42. Typically, about 10⁵ cells are required to achieve an adequate signal-to-noise ratio; nevertheless, the method is fully compatible with FFPE sections, enhancing its translational and retrospective value.

Ultraviolet–visible (UV–Vis) absorption spectroscopy affords rapid quantification of nucleic acids and proteins based on absorbance maxima at 260 nm and 280 nm, respectively. Although less specific than vibrational techniques, it is a useful preliminary or validation tool for assessing cell-state transitions and chromatin compaction 43.

NMR spectroscopy provides atomic-level structural and dynamic information on biomolecules, particularly for elucidating allosteric states of transcription factors, post-translational histone modifications and TCR–pMHC interactions 44. The advent of hyperpolarized NMR has increased sensitivity, permitting real-time monitoring of metabolic intermediates and epigenetic enzyme activities 45.

Hyperspectral imaging acquires a complete spectrum at every pixel, enabling spatially resolved biochemical characterization of tissues and cellular microenvironments. In immunology, HSI has been instrumental in profiling lymphoid-organ architecture, T-cell trafficking and immune-synapse formation with high spatial fidelity 46. Despite its excellent resolution, HSI generally requires fresh or cryopreserved specimens; compatibility with FFPE material remains limited.

Collectively, these spectroscopic platforms provide a holistic view of immune–epigenetic interfaces, permitting detection of static and dynamic biomolecular features in real time and, in many cases, at single-cell resolution.

T-cell exhaustion in the glioblastoma microenvironment is not merely a transient functional state but reflects a profound and stable epigenetic remodelling program. Chromatin accessibility assays in exhausted T cells have revealed global hypoacetylation of enhancers and loss of accessibility at effector loci such as IFNG, PRF1 and TNF, alongside increased accessibility at loci encoding inhibitory receptors such as PDCD1, CTLA4, LAG3 and HAVCR2 (TIM-3) 61. This shift is tightly regulated by a cohort of epigenetic modifiers, including DNMT3A, EZH2 and HDACs, that enforce a repressive chromatin landscape, locking T cells into a dysfunctional state (Table 1) 62. Histone acetylation levels—particularly at H3K27ac and H3K9ac—are decreased in exhausted T cells, leading to chromatin condensation and reduced transcriptional activity. Conversely, increased DNA methylation at key immune-effector genes contributes to their persistent silencing despite antigen stimulation 63. These chromatin alterations are not readily reversible by immune-checkpoint blockade alone, highlighting the need for integrated therapies capable of remodelling the epigenetic architecture of T cells in GBM.

Spectroscopic modalities such as Raman and FTIR are exquisitely sensitive to the biochemical composition and structural state of chromatin. Raman spectroscopy detects vibrational modes of molecular bonds and can differentiate between open (euchromatic) and condensed (heterochromatic) states on the basis of peak shifts and intensities in phosphate-backbone (PO₂^-) and nucleic-acid ring-breathing modes 64. For instance, chromatin decompaction induced by histone acetylation is associated with a relative increase in nucleic-acid-associated Raman peaks (785 cm^-1, 1,090 cm^-1), whereas condensed chromatin states exhibit a dominance of protein and lipid signatures 65. FTIR spectroscopy complements Raman analysis by measuring absorbance in the fingerprint region (1,000–1,800 cm^-1) corresponding to nucleic acids, amide bonds and histone modifications. Notably, acetylation of lysine residues results in shifts within the Amide I and II bands (1,650 and 1,550 cm^-1, respectively), which can be tracked as surrogate markers of histone acetylation levels 66. These spectral alterations support a hypothetical model in which T-cell exhaustion in GBM correlates with a decreased nucleic-acid-to-protein ratio and altered lipid ordering, reflecting underlying chromatin condensation and reduced transcriptional activity (Table 2).

Exhausted T cells exhibit distinct biochemical phenotypes characterized by altered metabolic activity, increased reactive oxygen species (ROS), lipid peroxidation and changes in membrane composition, all of which can be detected spectroscopically 67. Raman mapping of GBM-infiltrating T cells has revealed decreased lipid unsaturation (C=C stretching at ~1,655 cm^-1), altered protein secondary structure (amide III) and reduced nucleic-acid content, findings consistent with a senescent or metabolically suppressed phenotype 68. Spectral markers such as the 1,445 cm^-1 CH₂ scissoring band (indicative of membrane saturation) and the 1,080 cm^-1 phosphate stretch (DNA/RNA) provide quantitative indicators of the exhaustion state. These observations suggest that Raman and FTIR spectroscopy can detect functionally relevant biochemical changes in T cells that are not apparent from surface-marker profiling alone.

Cerebrospinal fluid (CSF) and GBM tumour biopsies represent minimally invasive sources for immune profiling. Recent advances in hyperspectral Raman microscopy and confocal FTIR mapping have enabled single-cell-resolution analysis of immune populations directly from clinical samples 69. For instance, T cells isolated from GBM CSF exhibit significant down-regulation of nucleic-acid spectral regions compared with peripheral T cells, reflecting reduced transcriptional activity and exhaustion 70. Moreover, biopsy-derived T cells from the GBM core and peritumoral regions display distinct FTIR and Raman fingerprints. These spatially resolved signatures correspond to differential epigenetic and metabolic states sculpted by the local tumour microenvironment, underscoring the potential of spectroscopic fingerprinting as both a diagnostic and stratification tool 71.

Combining label-free spectroscopy with single-cell omics approaches such as scRNA-seq and scATAC-seq yields a multidimensional view of T-cell exhaustion. For example, single-cell Raman spectroscopy coupled with transcriptomic profiling (Raman-seq) permits correlative mapping of spectral features with gene expression and chromatin accessibility at the individual-cell level 72. Integration with scATAC-seq further reveals spectral correlates of open versus closed chromatin, particularly at exhaustion-specific loci. This integrative framework has already been applied in pre-clinical glioma models to identify subpopulations of T cells that retain effector potential despite chronic stimulation 73. By training machine-learning algorithms on multimodal datasets (Raman + scRNA-seq), investigators can classify T cells into functional states with high accuracy and predict responsiveness to immune-checkpoint inhibitors or epigenetic modulators.

T-cell exhaustion in GBM is stabilized by epigenetic programs that repress effector gene loci and sustain inhibitory receptor expression. Reversible exhaustion retains partial chromatin accessibility and remains partially responsive to ICIs, whereas fixed exhaustion is characterized by near-complete transcriptional silencing and resistance to checkpoint blockade 84. Histone deacetylase (HDAC) inhibitors such as vorinostat and panobinostat can restore IFN-γ, granzyme B and IL-2 expression, thereby synergizing with PD-1 blockade 85. Bromodomain and extra-terminal (BET) inhibitors (e.g., JQ1) interrupt TOX/NR4A-driven dysfunction 86, whereas enhancer of zeste homolog 2 (EZH2) inhibitors (e.g., tazemetostat) de-repress Tcf7 and Runx3, promoting memory-like states 87. Nonetheless, excessive transcriptional de-repression may precipitate T-cell apoptosis 88. Single-cell ATAC-seq analyses confirm that fixed exhaustion is associated with rigid chromatin landscapes, emphasizing the relevance of epigenetic targeting 89.

Traditional immunomonitoring modalities such as flow cytometry and bulk transcriptomics require cellular labeling and sample manipulation, whereas label-free spectroscopic platforms—including Raman, Fourier-transform infrared (FTIR) and hyperspectral imaging—provide real-time, in situ molecular readouts. In GBM-infiltrating T-cells, these modalities can detect chromatin condensation, lipid peroxidation and protein conformational changes associated with exhaustion or reinvigoration 90. Diagnostic spectral signatures include the 785 cm^-1 DNA ring vibration, the 1,080 cm^-1 phosphodiester stretch and the 1,650 cm^-1 amide-I band, all reflective of chromatin and protein remodeling. Raman-derived heatmaps illustrate intratumoral heterogeneity, with reinvigorated T-cells localizing to tumor margins and exhausted subsets enriched within necrotic cores 91,92. Ongoing clinical trials are incorporating spectroscopy coupled with artificial intelligence to predict therapeutic outcomes 93.

Given the multifactorial drivers of T-cell exhaustion in GBM, monotherapy with either ICIs or epigenetic agents is frequently inadequate. Current rational strategies therefore prioritize combinatorial regimens that integrate ICIs, epigenetic modulators and metabolic adjuvants, guided by spectroscopic profiling. In preclinical glioma models, anti-PD-1 therapy combined with HDAC or EZH2 inhibitors augments intratumoral T-cell proliferation, cytokine secretion and memory-like differentiation (Figure 4). This synergy results from transcriptional reactivation of effector genes together with relief of chromatin-mediated immunosuppression 94. Incorporation of metabolic adjuvants—such as mTOR modulators, NAD⁺ precursors (e.g., nicotinamide riboside) or fatty-acid-oxidation inducers—further mitigates metabolic paralysis and restores mitochondrial fitness 95. Spectroscopy can detect these functional shifts through alterations in lipid saturation indices, NADH/FAD fluorescence and redox-sensitive spectral bands, thereby enabling real-time monitoring 96. Early-phase clinical trials evaluating triplet combinations (anti-PD-1 + HDAC inhibitor + metabolic modulator) have reported increased tumor-infiltrating lymphocyte density and partial restoration of TCR clonality, with spectroscopy embedded as a non-invasive endpoint 97 (Figure 4). Spectroscopy-derived biosignatures can also stratify exhaustion subtypes—for example, selecting BET inhibition for TOX-driven states or mTOR modulation for glycolytic dysfunction. Ultimately, spectroscopy-guided adaptive therapy, whereby spectral dynamics dictate treatment adjustments, may minimize toxicity while maintaining immune surveillance. Collectively, epigenetic, metabolic and checkpoint-directed combinations that are dynamically informed by spectral biomarkers represent a transformative frontier in GBM immunotherapy (Figure 4).

As a sole author, I confirm that I have conceived the study, designed the methodology, conducted the research, created all the illustrations independently, analysed the data, and wrote the manuscript. I also affirm full responsibility for all aspects of the work, ensuring the integrity and accuracy of the study.

None.

Not applicable.

Not applicable.

Not applicable.

AI-assisted tools were used solely for language refinement of select sections (Abstract, Introduction, Discussion). All scientific content, conceptual framing, data interpretation, and figure design remain the authors’ own. Full responsibility for the manuscript content is retained by the authors.

The authors declare that they have no competing interests.