Direct neuron-to-glioma synapses exist and result in excitatory post-synaptic currents (EPSCs), glioma cell depolarization, and tumor proliferation (17, 18). This pathological integration was first shown with glutamatergic synapses (17, 18), but recent work shows that glioma cells can receive GABAergic (19), cholinergic (20), serotonergic, adrenergic, and dopaminergic synaptic inputs (6, 7), allowing them to infiltrate a variety of brain-wide circuits. By communicating with a diversity of neuronal subpopulations, gliomas are highly adaptable, driving tumor progression and invasion throughout the brain while simultaneously complicating LFP interpretations, impeding therapeutic targeting and contributing to a myriad of neurological symptoms that affect patients’ quality of life.

Glioma cells themselves produce non-synaptic, activity-dependent potassium currents with low input resistance, similar to those made by normal glial cells, as well as autonomous rhythmic calcium-based oscillations generated from a small subset of “pacemaker-like” glioma cells (17, 21). These electrochemical signals then spread through a synchronized, gap-junction coupled tumor cell network (18, 22, 23). Many studies have also implicated activity-dependent secretion of synaptic regulator proteins in the ability of glioma cells to form synapses, as well as in proliferation and invasion (5, 24–28). In addition, other groups have highlighted epigenetic mechanisms that facilitate neuron-glioma interactions including hypomethylation with an upregulation of synaptic (29) and neuroactive ligand-receptor interaction genes (30). Extensive chromatin remodeling has also been shown to sustain neuron-glioma synapses through enhancer rewiring and transcription factors that regulate synaptic organization and axon guidance genes (31). Furthermore, glioma cells receive and respond to afferent neuronal inputs from a variety of local and far-ranging sources. They also contain intrinsic mechanisms and regulatory networks that bolster communication with neurons. All of which not only drives tumor progression but also dynamically influence extracellular current measurement at any given moment.

In the opposite direction, glioma cells induce neuronal hyperexcitability through mechanisms such as aberrant glutamate transport and impaired inhibitory signaling (32–37). Inhibitory neurons are not lost with gliomas but are instead recruited to participate in excitatory activity (38). Further, GABAergic signaling, which is typically thought of as inhibitory through cell hyperpolarization, can instead become depolarizing (19). One explanation for this is that many gliomas are thought to originate from oligodendrocyte precursor cells (OPCs) (39, 40), and normal GABAergic neuron-to-OPC synapses during development can be excitatory due to high intracellular chloride (41). In addition, OPC-like glioma cells are important drivers of epileptic activity at the leading edge via dysregulation of voltage-gated potassium channels leading to increased extracellular potassium and neuron depolarization (42, 43), as well as upregulation of synaptic gene expression programs and increased colocalization with neurons (30). Lastly, there are subtype-specific propagators of hyperexcitability such as increased levels of D-2-hydroxyglutarate produced by IDH mutated gliomas (44), a mechanism that IDH wildtype tumors may be largely insensitive to (45), and certain PIK3CA driver variants (46). These subtype-specific alterations may allude to why different types of gliomas, such as IDH mutant tumors, have higher rates of glioma-related epilepsy (44, 47, 48).

This imbalance between excitatory and inhibitory synaptic activity (E/I) promotes seizure-like events (SLEs) at the tumor margin and peritumoral regions, whereas the tumor core remains electrically silent (17, 34). SLEs can consequently develop into glioma-related epilepsy (GRE) - a common and often refractory symptom of gliomas (49). Ultimately, this reprogramming of canonical inhibitory signaling may represent an escape mechanism for tumor cells seeking to exploit neuronal activity-regulated factors. Glioma-induced neuronal excitability can further drive activity-dependent mechanisms of tumor growth and invasion, feeding a malignant cycle. Although the distorted E/I dynamics challenge treatment strategies and traditional electrophysiological analyses, tumor-associated E/I imbalances may be informative at a greater scale using intraoperative recordings (50–53), as discussed in the next section.

Neuron-to-glioma communication, synchronous glioma-cell networks and glioma-induced neuronal activity drive tumor progression and likely alter field potentials simultaneously. LFP interpretations are further confounded by the heterogeneous structural, spatial, and temporal relationships between the underlying current sources and sinks, especially as infiltrative gliomas such as GBM must be thought of as whole brain diseases. A less ambiguous approach may be to explore the fundamental changes in excitability at the level of the single neuron using single-unit recordings, which have become more feasible in recent years.

The modalities for measuring LFP in clinical settings include scalp EEG, cortical surface electrocorticography (ECoG, Figure 1B), and intracortical stereoencephalography (sEEG) (54). Subdural recording devices can be used in combination with direct electrical stimulation (DES) for brain mapping (use of electrical stimulation to disrupt function) during intraoperative resections of tumors situated in or near eloquent regions. Although LFP can be measured using any number of intracranial electrodes, clinical macroelectrodes have the advantage of low impedance, wider spatial coverage, handling ease, and FDA approval. While standard brain mapping techniques using clinical macroelectrode grids currently help avoid surgical damage to functional areas, they have limits.

Most conventional clinical electrodes have a diameter of 4mm and a pitch of 1cm, which averages electrophysiological signal over a relatively large area of cortex while still leaving a significant area uncovered. As a result, they lack the spatial resolution for localizing coordinated hubs of neural activity (55). Also, increasing distance between the recording electrode and current sources directly reduces signal quality, thus limiting the utility for tumors situated anywhere but directly under the grid (12). This effect is attributable to signal attenuation and dispersion, as well as further amplification of context-dependent confounders (e.g. spatial orientation, neuronal morphology) when recording from larger populations of neurons, all of which result in lower signal-to-noise ratio of recordings. Since electrodes are not useful for recording signals from white matter, neural recordings cannot help in many scenarios of deep invasion, which is common given that gliomas often invade along axon tracts (56). These limitations highlight that existing methods for detecting neural activity come with a tradeoff between precision and spatial coverage. Although FDA-approved ECoG grids are the standard-of-care, novel research electrodes are currently being developed to provide finer resolution while maintaining the ability to record from clinically significant spatial areas. For example, both Neuropixels probes and high-density thin-film electrodes have already demonstrated superior resolution for functional localization during brain mapping (Figure 2) (57–60). With the ability to characterize specific waveforms and finely track regional spike propagations, these higher resolution electrodes could contribute significantly to our understanding of glioma electrophysiology.

Even with traditional ECoG, patterns of electrical activity produced by physiological and pathological states can be detected through oscillatory and aperiodic components of the power spectrum (Figure 1C) (61). Oscillatory components are typically thought of in relation to canonical frequency bands. In the normal brain, high frequency signals like in the high gamma range (typically >70 Hz and <200 Hz) exhibit discrete spatial localizations and have been attributed to fast spiking sub-populations, such as those activated in task-dependent circuits (62–64). In contrast, lower frequencies tend to be more spatially diffuse and have been interpreted as slower fluctuations of synchronous inputs such as those modulating attention, memory, or perception (55, 65–67). These frequency-specific oscillations are represented as peaks of power above the aperiodic component (68). The aperiodic component, in contrast, represents non-oscillatory activity and can be assessed with parameters such as the offset and slope (68). The offset represents a uniform shift in power across frequencies and is correlated with increased neuronal spiking activity, whereas the slope of the aperiodic component has been shown to reflect the E/I balance, with a steeper slope indicating a lower E/I ratio (increased inhibitory tone) and a flatter slope indicating a higher E:I ratio (increased excitatory tone) (69).

In some cases, glioma-related electrophysiological trends are diffuse and non-specific to tumorigenesis, mirroring those seen in other brain pathologies such as increased slow-wave activity observed in stroke patients and increased high frequency oscillations in glioma-related epilepsy (33, 70, 71). However, recent work demonstrates the possibility of more specific tumor-associated signals such as altered E/I balance reflected in the aperiodic slope (15, 72, 73), frequency-specific changes that delineate peritumoral versus intratumoral regions (73), and task-dependent high gamma alterations within well-defined functional circuits (4, 5, 74–76). Further, given that brain mapping is commonly employed intraoperatively for tumor resections, these signatures represent promising biomarkers to more precisely identify and characterize pathological tissue.

Recent studies have begun to explore the relationship between the aperiodic slope and different factors of glioma pathology (15, 72, 73, 77). Using intraoperative ECoG recordings, one group found that in higher frequency bands (gamma: 30–50 Hz and high gamma: 70–150 Hz), glioma-infiltrated cortex had a flatter aperiodic slope compared to normal-appearing cortex, indicating an excitation-dominant state (77). They also found a positive correlation between these flatter aperiodic slopes and increased tumor cell infiltration, glutamatergic gene expression, and tumor grade, with IDH-wildtype GBMs, the most aggressive glioma subtype, displaying the flattest aperiodic slope (77). Other studies have explored the aperiodic slope over frequencies below 50 Hz, with one showing that IDH-wildtype GBMs preferentially occur in brain regions with intrinsically steeper aperiodic slopes (lower E/I ratio, inhibitory-dominant state) as measured from healthy participants (15). Peritumoral regions have also shown steeper slopes than intratumoral regions in the 20–40 Hz range (73).

These findings highlight that the interpretation of aperiodic slope is highly dependent on the frequency range considered (78), emphasizing the need for systematic characterization across bands. However, across studies, a consistent observation remains an increase in offset, and frequency-specific power as discussed below, in glioma-infiltrated tissue outside of the tumor core, despite molecular subtype, suggesting an overall elevation in neuronal activity (15, 72, 73, 79). Taken together, these data indicate that while general increases in neuronal activity are a common feature of gliomas, variation in E/I balance reflected by the slope may help stratify molecular subtypes, distinguish tumor boundaries, and provide further insight into evolving tumor dynamics.

Most studies so far focused on task-dependent alterations of canonical language circuits in patients with gliomas infiltrating dominant, left frontal cortical language areas (4, 5, 74, 75). Using intraoperative ECoG, differences in high gamma power (HGp) between normal-appearing and glioma-infiltrated cortex have been investigated during audiovisual naming tasks (4, 5). HGp is closely associated with increased spiking activity and neuronal synchrony (64). At sites with normal-appearing brain, ECoG recorded a task-dependent increase in HGp that was spatially confined to known areas of speech planning and initiation. At glioma-infiltrated sites, activation was also task-dependent, but maximal HGp increased more diffusely, including recruitment of non-canonical language areas. These differences are echoed by task-based fMRI (tb-fMRI) studies of syntax networks where healthy participants had task-dependent activation of three separate syntax networks with little cross-talk between networks, while glioma patients showed a lack of network boundaries with increased inter-network activity (74, 75). For motor tasks, ECoG has been used to record task-dependent differences of finger or wrist exercises in patients with motor cortex glioma (MCG) (76). In non-MCG patients, individual fingers had distinct locations within the motor cortex and individualized responses to direct electrical stimulation, but in MCG patients, these distinct spatial representations were lost and replaced instead by diffuse arrangements and remodeled activation patterns upon stimulation. Moreover, neural activity in glioma-infiltrated regions is clearly task-relevant and physiologically organized. However, gliomas also induce pathologic circuit remodeling with more diffuse high gamma activity and synchrony with other nearby areas.

Language studies have also been used to examine the effects of glioma infiltration on cortical activity during increasingly difficult tasks, such as monosyllable versus polysyllable conditions (4, 5). In normal-appearing cortex, the more difficult tasks elicited an upregulation of HGp which was not detectable when recording from glioma-infiltrated regions. Recordings from glioma regions produced below-chance decoding of the easier versus harder word trials compared to normal regions. Comparably, harder syntactic load elicited significantly decreased activation of syntax networks on tb-fMRI in glioma patients compared to healthy controls (74, 75). Taken together, this suggests that hyperconnected glioma-to-glioma and neuron-to-glioma networks may limit the ability to perform selective and specific processes at the local level, thus reducing the ability to coordinate more dynamic behavioral responses. Similar pathophysiology is seen in Alzheimer’s Disease (AD) patients, for example, where AD-associated cognitive impairment correlates with neuronal hyperexcitability and hyper-synchronization of brain regions (80–82). Indeed, gliomas are known to cause cognitive impairment (83, 84), and some studies have linked this with higher functional connectivity on imaging and scalp EEG (85, 86). Furthermore, conditions of increased task complexity may be crucial to revealing abnormal patterns intraoperatively (87).

In summary, recent work demonstrates that gliomas produce altered patterns of neural activity that can be detected intraoperatively using ECoG. However, current limitations include the difficulty of rigorous intraoperative studies, incomplete comparisons between glioma subtypes, anatomical regions, and between primary versus recurrent tumors, and importantly, the lack of spatial recording methods that match cortical physiology. Leveraging novel detection methods will allow more precise characterization of electrophysiological biomarkers to improve understanding and management of this disease.