The complex molecular epileptogenesis landscape of glioblastoma

Abstract

The cortical microenvironment surrounding malignant glioblastoma is a source of depolarizing crosstalk favoring hyperexcitability, tumor expansion, and immune evasion. Neosynaptogenesis, excess glutamate, and altered intrinsic membrane currents contribute to excitability dyshomeostasis, yet only half of the cases develop seizures, suggesting that tumor and host genomics, along with location, rather than mass effect, play a critical role. We analyzed the spatial contours and expression of 358 clinically validated human epilepsy genes in the human glioblastoma transcriptome compared to non-tumor adult and developing cortex datasets. Nearly half, including dosage-sensitive genes whose expression levels are securely linked to monogenic epilepsy, are strikingly enriched and aberrantly regulated at the leading edge, supporting a complex epistatic basis for peritumoral epileptogenesis. Surround hyperexcitability induced by complex patterns of proepileptic gene expression may explain the limited efficacy of narrowly targeted antiseizure medicines and the persistence of epilepsy following tumor resection and clarify why not all brain tumors provoke seizures.

Full text

Article
The complex molecular epileptogenesis landscape
of glioblastoma
Graphical abstract
Highlights
d157 genes linked to human monogenic epilepsy are enriched
at the tumor leading edge
dDosage-sensitive dysregulation of multiple gene pathways
validates proepileptogenicity
dLeading edge epilepsy-linked genes recapitulate early
developmental expression patterns
dEpistatic genetic remodeling defines a basis for
pharmacoresistant epilepsy therapy
Authors
Victoria Soeung, Ralph B. Puchalski,
Jeffrey L. Noebels
Correspondence
jnoebels@bcm.edu
In brief
Soeung et al. report spatial analysis of
monogenic epilepsy-linked gene
expression in human glioblastoma. The
dysregulated genes affect multiple
molecular pathways, and their dose-
dependent and epistatic interactions
explain both the presence and absence of
seizures in individual patients, as well as
the basis for pharmacoresistance to
narrowly targeted antiseizure drugs.
Soeung et al., 2024, Cell Reports Medicine 5, 101691
August 20, 2024 ª2024 The Author(s). Published by Elsevier Inc.
https://doi.org/10.1016/j.xcrm.2024.101691 ll

Article
The complex molecular epileptogenesis landscape
of glioblastoma
Victoria Soeung,
1
Ralph B. Puchalski,
3
and Jeffrey L. Noebels
1,2,4,
*
1
Developmental Neurogenetics Laboratory, Department of Neurology, Baylor College of Medicine, Houston, TX, USA
2
Center for Cancer Neuroscience, Baylor College of Medicine, Houston, TX, USA
3
Ben and Catherine Ivy Center for Advanced Brain Tumor Treatment, Swedish Neuroscience Institute, Seattle, WA, USA
4
Lead contact
*Correspondence: jnoebels@bcm.edu
https://doi.org/10.1016/j.xcrm.2024.101691
SUMMARY
The cortical microenvironment surrounding malignant glioblastoma is a source of depolarizing crosstalk fa-
voring hyperexcitability, tumor expansion, and immune evasion. Neosynaptogenesis, excess glutamate,
and altered intrinsic membrane currents contribute to excitability dyshomeostasis, yet only half of the cases
develop seizures, suggesting that tumor and host genomics, along with location, rather than mass effect, play
a critical role. We analyzed the spatial contours and expression of 358 clinically validated human epilepsy
genes in the human glioblastoma transcriptome compared to non-tumor adult and developing cortex data-
sets. Nearly half, including dosage-sensitive genes whose expression levels are securely linked to monogenic
epilepsy, are strikingly enriched and aberrantly regulated at the leading edge, supporting a complex epistatic
basis for peritumoral epileptogenesis. Surround hyperexcitability induced by complex patterns of proepilep-
tic gene expression may explain the limited efficacy of narrowly targeted antiseizure medicines and the persis-
tence of epilepsy following tumor resection and clarify why not all brain tumors provoke seizures.
INTRODUCTION
Pathological corticalhyperexcitabilityin the glioblastomamicroen-
vironment is both an early warning and a lasting clinical legacy of
this aggressive brain cancer. Seizures are the presenting sign in
nearly one-half of patients with glioblastoma, may promote tumor
growth, and typically persist as pharmacoresistant epilepsy after
tumor resection,
1
indicating a sustained restructuring of network
excitability in the surrounding host neocortex. While mechanisms
underlying acquired peritumoral epileptogenesis and its impact
on malignant cell invasion, cognitive impairment, and survival are
under active exploration,
2
correcting this deleterious outcome,
particularly in view of treatment advances prolonging survival,
3
re-
mains a foremost goal in the clinical management of brain tumors.
The early pathophysiology of tumor-related epilepsy (TRE) is
not fully explained by mass effect. While edema and increased
intracranial pressure might contribute at later stages, persis-
tence following tumor resection indicates that tissue compres-
sion alone is not required, and the large fraction of cases without
a seizure history supports the inference of genetic risk as a prin-
cipal component. Recent evidence in preclinical mouse glioblas-
toma models reveals that peritumoral synaptic imbalance and
cortical hyperexcitability are determined by the specific onco-
genes selected to drive tumor formation
4
as well as intrinsic sus-
ceptibility to epilepsy,
5
pointing to a process of gradual network
remodeling typical of acquired focal epileptogenesis that de-
pends upon tumor and host genomics and cortical location
rather than the space-occupying mass.
While little is known about the in situ electrobiology of human
brain tumors prior to their clinical detection, serial studies of tu-
mor progression in a new generation of in utero electroporation
(IUE) mouse glioblastoma models are providing key insights
into tumor epileptogenesis. In these immunocompetent, co-
isogenic tumor models, hyperexcitability onset coincides with
the emergence of tumor cell subclones expressing mRNA pro-
files enriched with synapse-related genes.
6
Alternative IUE tu-
mor driver genes significantly determine the TRE phenotype,
and epileptic tumors show a high ratio of excitatory to inhibitory
synaptic markers in the tumor margin absent in non-epileptic tu-
mors.
4,7
Peritumoral network remodeling during this period is
also accompanied by a sequential loss of fast-spiking interneu-
rons, perineuronal nets, altered glutamate exporter expression,
microglial activation,
5
and the emergence of glutamate-linked
microcircuit hyperactivity.
8
These defects confirm earlier obser-
vations in immunodeficient xenograft models,
9,10
indicating they
arise independently of changes in the adaptive immune
landscape.
11
Far more is known about the gene determinants of cortical sei-
zures in the absence of brain tumor. Over 350 genes are causally
linked to monogenic epilepsy by virtue of their de novo recur-
rence in patients and populate clinical exome screening panels
in wide use for precision epilepsy diagnosis.
12
These clinically
validated risk genes encode a broad array of ion channel sub-
units, transporters, synaptic proteins, and neuronal migration
factors, as well as oncogenic and intracellular metabolic path-
ways,
13
invoking a spectrum of candidate network excitability
Cell Reports Medicine 5, 101691, August 20, 2024 ª2024 The Author(s). Published by Elsevier Inc. 1
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
ll
OPEN ACCESS

mechanisms ranging from fast (msec) synaptic activity to slow
(minutes) paracrine signaling and ultraslow (hours) transcrip-
tional dysregulation. Both gain and loss-of-function mutations
in many of these genes are pathogenic, and, in some, haploid
dose alterations are also sufficient. To date, only a few candidate
genes have been implicated by differential expression in bulk
low-grade gliomas,
14,15
but without spatial resolution or valida-
tion of a direct role in intact human network hyperexcitability.
16,17
Given the biologically diverse pathways and molecular plasticity
within glioma microniches during tumor evolution,
18
the relative
contribution of any single one of these genes to epileptic network
synchronization is unclear, and a simultaneous large-scale
spatial mapping of their regulated expression, case concor-
dance, and fold change relative to unaffected developing and
adult neocortex has not yet been performed.
Here, we analyze the mRNA expression of 358 genes for hu-
man monogenic epilepsy (Epi358) according to their anatomical
expression in human spatial and developmental transcriptome
databases (see methods for details). The extended analysis in-
cludes epilepsy gene sets linked to channelopathies, receptors,
transporters, exocytotic release, familial migraine with seizures,
molecular targets of autoimmune epilepsies, and somatic gene
mutations leading to epileptic focal cortical dysplasia. Our anal-
ysis reveals pathogenic patterns of spatial enrichment and
extensive dysregulation of epilepsy-linked genes at the tumor
leading edge (LE) when compared to the pure tumor cell region
and the healthy cortex. Across functional clusters, the transcrip-
tion profiles of genes for ion homeostasis and macromolecular
biosynthesis most closely resemble those in developing unaf-
fected cortex. The microvascular zone of tumor cell proliferation
and migration encompassing the blood-brain barrier (BBB) is the
second most dysregulated region of proepileptogenic genes.
Most other tumor cell regions show little significant enrichment
compared to the pure cellular tumor. These complex patterns
of coordinate dysregulation may allow more precise matching
of tumor molecular excitability subtypes with antiseizure medi-
cines to direct future gene-guided clinical management of
glioblastoma.
RESULTS
Monogenic sources of epileptogenesis aggregate at the
tumor LE
To characterize the intratumoral heterogeneity of epilepsy-linked
genes, we first determined the comparative enrichment ratios
(ERs) of epilepsy gene transcripts for 6 anatomically and molec-
ularly defined tumor regions compared to the pure cellular tumor
(CT)
19
(see Methods). Our analysis reveals the distinct arealiza-
tion of the excitability transcriptome and highlights the spatial
heterogeneity across tumor regions (Figure 1A). We found a large
proportion (157/358) of epilepsy genes enriched by 1.5-fold or
higher at the LE (containing only 1%–3% tumor cells), a similar
fraction (172/358) without significant change, and an interesting
subset (25/358) enriched in pure tumor cell regions relative to the
LE (high-resolution details in Figure S1;Table S1). At the LE-
adjacent infiltrating tumor (IT) (containing 10%–20% tumor cells),
118/358 genes are enriched but to a lesser magnitude as ex-
pected by the lower neuronal content of this region. Other intra-
tumoral regions each contain less than 50 enriched genes as
compared to the CT. In contrast, both the LE and IT also contain
the fewest under-enriched genes (25/358 and 2/358, respec-
tively), while the microvascular proliferation (MVP) region con-
tains the most under-enriched genes within the tumor (113/
358). Genes enriched at the LE have the highest statistical signif-
icance and case concordance (Figure S1;Table S1). Only 19% of
Epi358 genes show low concordance at the LE (67/358), and
concordance is lower in other regions consistent with tumor
cell heterogeneity. To assess whether seizure status altered
regional RNA expression levels, we compared RNA expression
between cases presenting with seizures and those without
(FC
sz
)(Figure S2;Table S1). We found that despite a greater
than 10-fold elevation (FC
sz
)ofFOSB, an immediate-early gene
sensitive to excess neuronal depolarization, relatively few
Epi358 genes are consistently amplified over 2-fold by a history
of seizures in this cohort.
Heterogeneous proepileptic pathways are enhanced in
distinct intratumoral zones
We then conducted a Gene Ontology analysis to classify the
Epi358 genes by biological processes and identified three major
categories, (1) cell growth and division (CGD), (2) ion channel and
transporter (ICT), and (3) macromolecule biosynthesis (MBS)
(Figure 1B). Intriguingly, 81% of the genes in the ICT category
are enriched at the LE, whereas the majority of MBS genes
show equivalent expression at the LE and CT. Approximately
50% of genes in the CGD category are enriched only at the LE
while the other half are expressed equally between the LE and
CT (Figure 1C). Thus, a significant number of epilepsy-linked
genes show specific enrichment at the LE, and their dysregu-
lated expression would be masked in a bulk tumor analysis (Fig-
ure 1D). For example, both deletions and heterozygous
missense mutations of SNAP25, a key regulator of transmitter
exocytosis, are proepileptic.
20
Despite the low level originally re-
ported in bulk human glioma suggesting a role in tumor suppres-
sion,
21
we found that SNAP25 is markedly enriched and upregu-
lated at the LE consistent with a high peritumoral synaptic
density but is not enriched in deep tumor regions (Figure S1),
where, in the absence of significant synaptic connectivity, it is
unlikely to contribute to aberrant network activity. Another
instructive example is LGI1, a gene so named upon its discovery
in bulk tissue as ‘‘low in glioma.’’
22
LGI1 was initially considered a
putative glioma suppressor with no known role in excitability, yet
it is now recognized as a presynaptic protein linked to potassium
channel function
23
and autoimmune epilepsy.
24
We consistently
find LGI1 strongly enriched and upregulated (Figure S1) at both
the LE and IT but not in other tumor regions. Thus, genes may
exert congruent or opposing network excitability and growth-
promoting effects, either singly or as a composite group, in
different spatial contexts that may be obscured in bulk tissue
analyses.
Tumor lobar location does not greatly affect
dysregulation of Epi358 genes at the LE
Next, we ascertained the relative magnitude of dysregulation at
the cortical tumor margin by comparing the mean expression
of the Epi358 genes at the LE to expression levels in anatomically
2Cell Reports Medicine 5, 101691, August 20, 2024
Article
ll
OPEN ACCESS

aligned, age-matched, healthy human frontal, temporal, or pari-
etal lobe cortical samples published in the Human Protein Atlas
database.
25
The majority of genes show altered expression
levels at the LE compared to non-tumor cortex, suggesting
that although the neuronal, glial, and immune cells at the LE
are microscopically similar to healthy cortex, they are highly tran-
scriptionally remodeled. The Epi358 genes fall into multiple
distinct groups according to their intratumoral ER and fold
change (FC
Cx
) profile per cortical lobe (Figure 2A, groups 1–
15). Intriguingly, the majority of epilepsy-linked transcripts
Figure 1. Monogenic sources of epileptogenesis aggregate at the tumor LE
(A) The LE is enriched for monogenic sources of epileptogenesis. Upper: heatmaps ordered by magnitude of enrichment at the LE compared to cellular tumor
zone. Nearly 50% of monogenic sources of epilepsy are either enriched specifically at the LE (157/358) or equally expressed in the LE and CT (172/358). Lower:
percent of cases enriched in each zone. Detailed gene list given in Figure S1.
(B) Epi358 genes are involved in diverse cellular processes. Circular tree plot of the Epi358 genes based on hierarchical clustering by biological function. Inner
circle: Gene Ontology analysis defined three broad categories: (1) cell growth and division (CGD), (2) ion channel and transporter (ICT), and (3) macromolecule
biosynthesis (MBS). Middle circle: the enrichment ratio (ER) for genes in the ICT category is almost exclusively enriched at the LE. Outer circle: regional RNA-seq
results in improved concordance for all categories compared to bulk tumor analysis. Detailed gene list given in Figure S1.
(C) Epilepsy-linked ion channels show the strongest enrichment at the LE. Bar plot quantifying the enrichment status of Epi358 genes at the LE compared to the
CT. A large percentage of genes in the CGD category and greater than 80% of genes in the ICT category show strong LE enrichment.
(D) Case concordance at the LE is greater for all functional categories as compared to bulk analysis. Bar plot quantifying concordance of gene expression of
Epi358 genes at the LE as compared to concordance from bulk tumor analysis. For genes in the CGD and MBS categories, there were 1.49-fold and 1.44-fold
more genes that display high-medium concordance at the LE compared to bulk analysis, respectively. Genes in the ICT category show the most improved case
concordance as compared to bulk tumor analysis with 3.57-fold more genes with high-medium concordance at the LE.
Cell Reports Medicine 5, 101691, August 20, 2024 3
Article
ll
OPEN ACCESS

Figure 2. Complexity and degree of dysregulation support an epistatic basis of epileptogenesis at the LE
(A) The majority of genes linked to monogenic epilepsy are similarly dysregulated across the frontal, temporal, and parietal lobes. Heatmap of the ER and FC
Cx
of
Epi358 genes at the LE compared to different lobes of healthy human cortex. Lanes show samples from the frontal (FL), temporal (TL), parietal (PL) lobes, and
pooled cortex (Cx). The expression patterns were stratified according to congruence of ER and FC of tumors derived from different cortical lobes, forming 15
distinct groups (color coded). Groups 1–9 share a similar FC
Cx
and ER pattern at the LE, whereas groups 10–12 and 13–14 share only a similar FC
Cx
or ER pattern,
respectively. Group 15 genes show a variable ER and FC
Cx
pattern. Forty-one of the Epi358 are validated pathogenic dosage-sensitive genes (rCNVs) (see
Figure S3 for more details on the 41 epilepsy-linked rare copy-number variants; rCNVs).
(legend continued on next page)
4Cell Reports Medicine 5, 101691, August 20, 2024
Article
ll
OPEN ACCESS

show a high congruence between ER and FC
Cx
despite the
different locations of the cortical tumors (Figure 2A, 218/358,
groups 1–9). Another large subset (Figure 2A, 85/358, groups
10–12) shows high congruence of fold change relative to the
native cortex but differs in intratumoral enrichment profile. The
high congruence of FC
Cx
across cortical regions suggests that
for these epilepsy-linked genes, the tumor is similarly remodeling
the LE despite any intrinsic differences in the surrounding
cortical tissue. However, lobar synaptic connectivity differences
adjacent to the LE may also contribute to why some tumors pro-
voke peritumoral epilepsy.
Complexity and degree of LE dysregulation support an
epistatic basis of epileptogenesis
We further characterized the 218 genes that showed high
congruence of dysregulation at the LE across the frontal, tempo-
ral, and parietal lobes (groups 1–9). We conducted a Gene
Ontology analysis to cluster the genes according to similar bio-
logical function for groups 1–3 (enriched at the LE) and groups
4–6 (equal LE and CT expression), respectively (Figure 2B),
and then conducted a Kyoto Encyclopedia of Genes and Ge-
nomes (KEGG) pathway analysis of major clusters of these
genes. Our analysis shows striking dysregulation of multiple
genes involved in neuronal functions such as synaptic vesicle cy-
cle and GABAergic synaptic function at the LE. Although the ma-
jority of genes are downregulated compared to healthy cortex, a
subset in each pathway is aberrantly upregulated. Similar to the
genes enriched at the LE, those in group 4–6 show high dysregu-
lation, with a large proportion downregulated and a few aber-
rantly upregulated. Overall, the transcriptome profile at the LE
differs greatly from region- and age-matched healthy cortex,
indicating a complex dysregulation of multiple pathways and
an epistatic basis for peritumoral hyperexcitability and
epileptogenesis.
Dosage-sensitive genes validate the pathogenicity of
large fold changes at the LE
Although transcript-level changes only provide plausible evi-
dence of actual functional impact, many heterozygous nonsense
or missense mutations that lead to an equivalent loss or gain of
function cause seizures in patients,
26
and the experimental liter-
ature is replete with overexpression and haploinsufficient seizure
models based on gene dosage. Important human examples of
epilepsy-linked dosage-sensitive genes include SCN1A,
SCN2A,CHD2,SYNGAP1, and DEPDC5. We determined that
41 of the Epi358 genes are known dosage-sensitive genes,
where allele haploinsufficiency and/or triplosufficiency in non-tu-
mor brain lead to epilepsy in human patients (Figure 2A second
column, genes listed in Figure 3). Aberrant dosage of any of
these genes alone is sufficient evidence of pathogenicity. We
therefore adopted a level of 0.5- and 1.5-fold change in transcript
density as a benchmark for prioritizing pathogenicity. We found
that 16/25 haploinsufficient genes are downregulated, and 2/26
triplosensitive genes are upregulated at the LE (Figures S3A and
S3B). Thus, 18/41 dosage-sensitive genes that have been inde-
pendently linked to clinical seizure and neurodevelopmental dis-
orders show dysregulated expression profiles sufficient for clin-
ical expression of epileptogenesis at the LE.
Homeostatic mechanisms of ion equilibrium are
pathologically remodeled at the LE
Ion channelopathy is the best understood class of monogenic
epilepsy, and each channel is linked to multiple variant-defined
clinical syndromes. We found that of the three major functional
gene categories, only the ICT group contains a majority of genes
that are enriched and dysregulated at the LE (Figures S3C–S3E).
Detailed functional and computational analyses in both heterol-
ogous cells and in situ cortical networks have established that
pathogenic channel variants with slight to major shifts in either
biophysical activation or deactivation properties lead to a range
of epilepsy phenotypes and antiepileptic drug sensitivities. Simi-
larly, impaired interactions with their regulatory subunits repro-
duce current defects across many channel subtypes.
27
The
epistatic consequences of multiple gain or loss-of-function var-
iants are complex due to the cell type-specific roles of ion chan-
nels in sculpting excitability within synaptic microcircuits. The
combinatorial outcome of even two distinct epilepsy gene muta-
tions on a seizure phenotype depends on the specific network
affected and may either exacerbate or attenuate network hyper-
excitability.
28,29
We find extensive levels of transcriptional dysre-
gulation within all major classes of voltage-gated ion channels
(details are in the following paragraph), highlighting the tumor-
induced departure from homeostatic mechanisms maintaining
excitatory and inhibitory equilibrium in the surrounding cortical
circuitry.
Sodium channels are enriched and dysregulated at the
LE
Voltage-gated sodium channels mediate depolarization in
response to changes in membrane potential, enabling the initia-
tion and propagation of action potentials. Four pore-forming
alpha subunits linked to epilepsy, SCN1A,2A,5A, and 8A, and
the regulatory subunit SCNB1 are all significantly enriched at
the LE and decreased elsewhere in the tumor (Figure 2C). How-
ever, only SCN1A and SCN2A have elevated expression
compared to healthy cortex,whereas SCN5A,8A, and SCNB1
have reduced expression. Experimental studies of epilepsy-
linked clinical variants of these sodium channels indicate that
either gain or loss of function can contribute to network hyperex-
citability. Coordinate changes in channel genes may also reflect
local altered upstream transcription factors. For example, the
(B) Genes from groups 1–3 (LE enriched) and groups 4–6 (LE equal to CT) were hierarchically clustered according to biological function. KEGG pathway analysis
of prominent clusters identified pathways dysregulated at the LE. The dysregulation of multiple epilepsy-linked genes involved in diverse pathways supports an
epistatic basis for epileptogenesis at the LE.
(C) Simultaneous dysregulation of multiple voltage-gated ion channels disrupts ion homeostasis and supports pathological spreading depolarization. Heatmap of
the FC
Cx
and sample-wise regional ER for voltage-gated sodium channel (VGSC), voltage-gated calcium channel (VGCC), and voltage-gated potassium channel
(VGKC) subunits linked to epilepsy. Of the VGSCs, only SCN1A and SCN2A are upregulated compared to wild type (WT). Almost all VGCC and VGKC are
downregulated except for a few calcium-activated potassium channels. Asterisks (*) denotes genes implicated in clinical cases of spreading depolarization.
Cell Reports Medicine 5, 101691, August 20, 2024 5
Article
ll
OPEN ACCESS

NRSF/REST transcription corepressor silences sodium channel
transcription
30
and thus decreased REST levels at the LE favor
elevated sodium channel transcripts at this location and sup-
pression where REST is more highly elevated, particularly at
the MVP (Figure 2C).
Calcium channels are enriched and uniformly
downregulated at the LE
PQ, N, and R-type channels (CACNA1A,CACNA1B,andCAC-
NA1E, respectively) mediate excitation-coupled vesicular
release at central synapses. We consistently found peritumoral
enrichment of these calcium channels (Figure 2C), along with
enriched expression of requisite presynaptic exocytotic ma-
chinery genes (Figure S4A), consistent with the excess synap-
togenesis at the tumor margin seen in experimental glioblas-
toma models,
4,7,31
yet all presynaptic calcium channels are
uniformly downregulated compared to healthy cortex (Fig-
ure 2C). However if dysregulation of these channels is un-
equally distributed across a microcircuit of excitatory and
inhibitory neurons, the resulting excitatory:inhibitory release
imbalance can explain why both gain and loss-of-function var-
iants in these subunits may result in epilepsy.
32
L-type calcium
channels mediate excitation-transcription coupling but are un-
common causes of epilepsy; their genes, CACNA1C and CAC-
NA1D, are enriched at both the LE and MVP zone (Figure 2C).
Low-voltage-activated T-type calcium channel genes CAC-
NA1G and CACNA1H mediate rebound bursting in thalamo-
cortical circuits and, when elevated, promote
33
or, when low-
ered, prevent
34
generalized spike-wave seizures. Their
downregulation at the LE is consistent with the absence of
this specific aberrant electroencephalography (EEG) pattern
in glioblastoma tumor cases.
Potassium channels are predominantly enriched and
downregulated at the LE
Potassium channel subunits comprise the largest class of epi-
lepsy channelopathy,
35
with both gain and loss of function lead-
ing to prominent epilepsy phenotypes.
36
These channels
exhibit complex combinatorial control over compartmental
membrane excitability in dendrites and axons,
37
are modulated
by various factors (i.e., membrane voltage, calcium ions,
G-protein-coupled receptor [GPCR] signaling, and/or ATP),
and also mediate a diverse range of essential cellular functions,
including cellular homeostasis and membrane polarization. A
large subset of potassium channel genes is strongly enriched
at the peritumoral LE (KCNA1,2;KCNAB2;KCNC1;KCNH1,5;
KCNS1;KCNT1) compared to tumor regions (Figure 2C). In
contrast, KCNQ2, a gene encoding the non-inactivating Kv7
M-current does not show strong enrichment for any tumor re-
gion. Most of these genes are downregulated at the LE
compared to healthy cortex except for a subset of calcium-acti-
vated channels (Figure 2C). The calcium-activated potassium
channel subunit KCNMB1 is downregulated at the LE but en-
riched at the MVP, a region rich in actively migrating cells. In
contrast, the subunits KCNMB2,3,4show strong enrichment
and upregulated expression at the LE. In neurons, mutations
of KCNMB1 and KCNMB4 lead to epilepsy, but their role in tu-
mor cells is unknown. However, the related calcium-activated
potassium channel, KCNN4, is involved in glioma cell migra-
tion.
38
The enriched expression of potassium channel subunits
attheLEwithaknownroleinbothintrinsicexcitability
and cellular proliferation, so-called ‘‘oncochannelopathy
genes,’’
39,40
merits further exploration.
Breakdown of peritumoral ion homeostasis facilitates
pathological spreading depolarization
Overall, our results indicate that multiple genes from different ion
channel classes are simultaneously dysregulated at the LE, facil-
itating the pathological breakdown of ion homeostasis. One such
pathological event is spreading depolarization. Spreading depo-
larization (SD) is a slowly propagating pathological wave of
neuronal and glial depolarization resulting from the breakdown
of membrane ion homeostasis and is implicated in neurologic
defects associated with brain tumor progression such as hypox-
ic brain injury and migraine aura syndromes.
41
This glioblastoma
excitability biomarker was detected emanating from the tumoral
LE of glioblastoma mouse models.
5,7
Three genes are linked to
SD threshold in monogenic syndromes of familial hemiplegia
(FHM1–3) with seizures, including PQ-calcium channel, CAC-
NA1A, in FHM1, astrocytic sodium potassium ATPase,
ATP1A2, in FHM2, and sodium channel, SCN1A, in FHM3. Func-
tional studies indicate that gain-of-function mutations in CAC-
NA1A
42
and loss-of-function mutations in ATP1A2
43
and
SCN1A
44
lower SD threshold. All three genes are enriched at
the LE (Figure 2C), and downregulation of ATP1A2 is consistent
with ATP1A2 loss-of-function mutations resulting in SD.
Although SCN1A is upregulated at the LE, its functional levels
depend on its regulatory subunit, SCN1B, which is strongly
downregulated at the LE, highlighting the epistatic interactions
between ion channel subunits. While gain-of-function mutations
of CACNA1A are associated with increased synaptic glutamate
release, we found a strong downregulation of CACNA1A at the
LE (Figure 2C). However, transmitter release might be enhanced
by upregulation of other ion channels such as HCN1. Hyperpo-
larization and ATP-activated cyclic nucleotide channels (HCNs)
conduct mixed cation (predominantly K
+
and Na
+
) pacemaker
currents that regulate neuronal bursting, and a spectrum of mu-
tations in HCN1–4has been identified in patients with epilepsy.
45
We show that HCN1 transcripts are strikingly enriched and upre-
gulated at the LE (Figure 2C; Table S1), leading the list of Epi358
channel genes. Interestingly, a search in the Ivy Glioblastoma
Atlas Project (GAP) for the strongest enrichment of all ion chan-
nel genes at the LE identified KCNS1, a potassium channel
linked to brain edema and increased intracranial pressure
46
but
not yet linked to epilepsy (Figure 2C). This gene is strongly en-
riched at the LE and shows uniform downregulation compared
to healthy cortex. Overall, the complex dysregulation of ion
channel transcripts at the LE is supportive of a hyperexcitable
environment conducive to SD and other dysfunctions of ion
homeostasis.
Peritumoral remodeling of glutamatergic and
GABAergic signaling promotes hyperexcitability
Glutamate receptors show subtype-specific regional
dysregulation
Mesoscale imaging in experimental
8
and human glioblastoma
47
reveals extensive extracellular glutamate accumulation contrib-
uting to epilepsy,
48
tumor progression,
49,50
and cortical SD.
5
High levels of glutamate are likely supplied by tumor cells and
6Cell Reports Medicine 5, 101691, August 20, 2024
Article
ll
OPEN ACCESS

astrocytes
51
and maintained by defective uptake,
52
but the
involvement of receptors in this pathway has not been spatially
resolved. We detected extensive LE enrichment of ionotropic
(NMDA/GRIN,AMPA/GRIA), and metabotropic (GRM) gluta-
mate receptor subunits relative to tumor regions (Figures 3A
and S4B). NMDA (GRIN1,2A,3A,2B, and 2C) receptor subunits
concentrating at the LE suggest a mechanism for excitotoxicity;
however, the FC
Cx
values are uniformly depressed. In contrast,
AMPA receptor (GRIA1-4) expression shows extensive LE and
intratumoral enrichment consistent with a role in tumor
growth,
53,54
although only GRIA3 is so far genetically linked to
epilepsy. GRIA2 uniquely shows a major FC
Cx
increase at the
LE. Other members, including the extrasynaptic tonic glutamate
receptor dsubunit gene GRID1, and kainate receptors show an
elevated LE and tumor enrichment profile, yet a low FC
Cx
. Me-
tabotropic glutamate receptors (mGlur1–8) are strongly LE en-
riched and display subtype-specific downregulation compared
to healthy cortex (Figure S4B). Mutations in these genes have
not yet been linked to epilepsy.
Dysregulation of glutamate transport genes supports
abnormal glutamate homeostasis
Synaptically releasable glutamate is concentrated in vesicles by
proton-dependent (VGLUT1–3/SLC17A6–8) and zinc-depen-
Figure 3. Peritumoral remodeling of gluta-
matergic and GABAergic signaling promotes
hyperexcitability
(A) Dysregulation of glutamate signaling genes
supports abnormal glutamate signaling and ho-
meostasis. Ionotropic glutamate receptors show
subtype-specific regional dysregulation. Although
most NMDARs and AMPARs show strong enrich-
ment at the LE, only GRIA2 was upregulated, and
most other glutamate receptors are downregulated
in comparison to healthy cortex. Bottom: evidence
for increased transcript levels of vesicular gluta-
mate transport (SLC17A6,8) and astrocytic
(SLC7A11) glutamate transporter.
(B) Dysregulation of GABA signaling genes sup-
ports abnormal GABA signaling favoring hyperex-
citability. GABA synthesis genes are enriched and
upregulated. GABA transport to and from the
extracellular space are uniformly downregulated.
GABA receptor dysregulation and chloride mem-
brane gradient is conducive to depolarizing GABA
signaling. GABA receptors display a subtype-spe-
cific upregulation at the LE. Although SLC12A5
(KCC2) is uniformly enriched at the LE, it is
consistently downregulated as compared to
healthy cortex without a parallel increase in
SLC12A2 (NKCC1) expression, indicating a peri-
tumoral KCC2:NKCC1 ratio conducive to depola-
rizing GABA signaling.
dent (ZNT3/SLC30A3) transporters. We
found upregulation of the vesicular trans-
porters, SLC17A6 and SLC17A8 at the
LE, but not SLC17A7 or SLC30A3 (Fig-
ure 3A). Similarly, the astrocytic cysteine-
glutamate exporter (XCT/SLC7A11) is en-
riched and upregulated at the LE and may
also contribute to elevated extracellular glutamate levels
5
(Fig-
ure 3A). Three high-affinity neuronal and glial plasma membrane
transporters (SLC1A1–3,6) are responsible for greater than 90%
of cellular glutamate uptake in healthy brain and are all linked to
epilepsy. These genes are LE enriched but expressed at healthy
cortical levels (Figure 3A). The increased expression of gluta-
mate exporters coupled with the lack of increase in cellular gluta-
mate uptake may contribute to the excess glutamate accumula-
tion seen in clinical and preclinical studies.
Evidence for increased GABA synthesis without parallel
increase in transport
GABA, the primary inhibitory neurotransmitter, is synthesized
from glutamate. Deficiencies in either GAD1
55
or GAD2
56
reduce
inhibitory GABA signaling, and both are strongly LE enriched
(Figure 3B). GABA can also be synthesized by outer mitochon-
drial membrane flavoenzymes, MAOA and MAOB. Like GAD1,
MAOA is also elevated at the LE. However, contrary to GAD1
deficiency, which can lead to seizures, duplication of MAOA
has been reported in a patient with epilepsy.
57
Vesicular pack-
aging of GABA by VGAT/SLC32A1 and reuptake via plasma
membrane transporters via GAT1–3/SLC6A1,13,11 is strongly
enriched at the LE but is largely downregulated at the LE
compared to healthy cortex (Figure 3B).
Cell Reports Medicine 5, 101691, August 20, 2024 7
Article
ll
OPEN ACCESS

GABA receptors show uniform LE enrichment and
subtype-specific upregulation
Ionotropic GABA receptors are pentamers, and mutations in all
subunits are linked to human epilepsies.
58
GABAergic transmis-
sion plays a critical role in network synchronization,
59
stem cell
proliferation, migration, synaptogenesis, and immune cell func-
tion.
60,61
GABA receptors are reduced in bulk glioblastoma tu-
mors compared to lower grade gliomas, and in deep glioblas-
toma tumor regions relative to the perimeter.
62
However, we
detected uniform enrichment of all ionotropic a,b, and gsubunits
at the LE compared to tumor cells, and a mixed FC
Cx
pattern of
receptor expression compared to healthy cortex (Figure 3B). Me-
tabotropic GABA receptors (GABBR1,2) are also enriched at the
LE. However, these genes are largely downregulated. Metabo-
tropic GABA signaling is critical for epilepsy in non-malignant
focal epileptic cortical dysplasia, including tuberous sclerosis.
63
The peritumoral membrane chloride gradient is
conducive to depolarizing GABA signaling
The effect of GABA on network excitability depends upon the
postsynaptic transmembrane chloride gradient. This gradient un-
dergoes a shift from depolarizing to hyperpolarizing during early
brain development as chloride exporter, KCC2 (SLC12A5),
expression increases, reversing the balance of Cl

import medi-
ated by NKCC1 (SLC12A2). KCC2 loss-of-function mutations
lead to epilepsy,
64
and a knockin mouse of a KCC2 mutation
that prevents phosphorylation-dependent inactivation sup-
presses convulsant-induced seizures,
65
validating the importance
of this pathway and serving as a reminder of the critical role of
posttranslational modulation for this and other genes. Analysis
of chloride transport dysregulation in glioblastoma has yielded
mixed results in human tissue and murine xenograft models.
66–68
However, we consistently found KCC2 (SLC12A5) is enriched at
the LE but downregulated compared to healthy cortex (Figure 3B)
without a parallel increase in NKCC1 (SLC12A2), indicating a
KCC2:NKCC1 ratio that favors depolarizing GABA signaling,
which may contribute to peritumoral hyperexcitability.
The LE and MVP are distinct proepileptic zones
Our analysis identified the LE and MVP zone as the most proepi-
leptic tumor regions (Figure 1A). A closer examination of the
enrichment profile identified 32 epilepsy-linked genes enriched
at the MVP zone compared with the CT (Figure 4A). Approxi-
mately half of those genes are also selectively enriched at the
MVP compared to the LE and may influence neurovascular reac-
tivity and BBB integrity.
69
For example, the mechanosensitive
stretch non-selective cation channels PIEZO1 and PIEZO2 are
exclusively enriched at the MVP (Figure 4B), and upregulation
of both genes has been reported in a case of seizures with
cortical compression.
70
While genomic variants are not as yet
linked to genetic epilepsy, THSN1, the gene encoding astroglial
thrombospondin1 was recently implicated in glioblastoma syn-
aptic remodeling
31
and is also enriched at the MVP (Figure 4B).
Pathological astroglial-based BBB alterations provide vasogenic
mechanisms for local inflammation leading to hyperexcitability,
as well as defective transport of antiepileptic drugs.
71
Autoimmune epilepsy antigens are enriched at the LE
Autoimmune epilepsy is an uncommon result of pathogenic anti-
body binding to neuronal antigens leading to seizures
72
and typi-
cally diagnosed in the context of non-CNS cancers or no discov-
erable malignancy. Molecular targets of autoimmune antibodies
include NMDA, AMPA, GABA, and glycine receptors (GAD65,
GAD67,LGI1,HU/ELAV4,NEUREXIN1–3, and CASPR2
73
), and
germline mutation of each of these targets is linked to mono-
genic epilepsy. Surprisingly, we found that genes for 22 autoim-
mune antigens show striking enrichment at the LE, while none
are enriched at the MVP, pointing to an unexplored immunolog-
ical pathway underlying glioblastoma-related epilepsy (Fig-
ure 4C). Glioblastoma is not currently recognized as one of the
causative tumors underlying autoimmune epilepsy,
72,74,75
except in a single case report of seizures due to anti-glutamate
receptor cerebral spinal fluid antibodies subsequently found to
arise from an unrecognized glioblastoma,
76
suggesting this
pathway merits further exploration.
Differences in the regulation of MTOR pathways at the
LE and MVP
Mutations in genes of the MTOR pathway are one of the primary
causes of focal cortical dysplasia (FCD), the most common
cause of MRI-detectable, pharmacoresistant epilepsy. Somatic
mutations of these genes generate localized cortical cell line-
ages and thus represent a model of non-malignant peritumoral
epileptogenesis, and some (KRAS,NF1,NIPBL, and PCDH19)
have been linked to cellular growth in glioblastoma.
77–79
We
found that most MTOR pathway genes show lowered expression
at the LE compared to healthy cortex, except for NPRL2,KRAS,
and PTEN, which show increased expression (Figure 4D).
Although KRAS and PTEN are upregulated, the MTOR repres-
sors, TSC1,TSC2,DEPDC5,NPRL3,RPTOR, and STRADA,
are all downregulated at the LE compared to healthy cortex,
highlighting how different MTOR downstream pathways may
be enhanced while others are repressed in glioblastoma (Fig-
ure 4D). PTEN is upregulated and enriched in both the LE and
MVP regions, but the repressor TSC1 is highly LE enriched yet
downregulated at the LE, highlighting their different regional
roles in tumorigenesis. MTOR inhibitors are effective antiseizure
therapy for FCD and suppress glioblastoma tumor cell prolifera-
tion in vitro,
80
and their in vivo profile suggests further therapeutic
study of the effects of MTOR signaling on glioblastoma-related
epilepsy and tumorigenesis.
Epilepsy gene dysregulation at the LE partially
recapitulates early development
Finally, we sought evidence for the important hypothesis that tu-
mor-induced gene dysregulation may recapitulate patterns of
plasticity characteristic of early cortical development.
81
We
compared the degree of dysregulation of 254 Epi358 genes
aberrantly expressed at the LE to healthy cortex during early pre-
natal and postnatal development (see Methods) and found that
54% of the LE dysregulated genes resemble a FC
Cx
pattern
similar to that of at least one early developmental stage (Fig-
ure 5A). Suppressors of MTOR signaling are uniformly downre-
gulated at the LE and at all developmental stages tested (Fig-
ure 5B), suggesting that MTOR signaling itself is upregulated in
both tumor LE and healthy cortex during these early develop-
mental stages. Genes involved in neuronal migration are upregu-
lated in healthy cortex at the prenatal and infancy stage
compared to adult, as expected based on the high demand for
8Cell Reports Medicine 5, 101691, August 20, 2024
Article
ll
OPEN ACCESS

neuronal displacement (Figure 5B). Glioblastoma is more preva-
lent in older adults (>40 years of age), but many of the MTOR
signaling genes such as TSC1,DEPDC5,NF1, and NPRL2 all
showed expression levels at the LE, which more closely
resemble healthy earlier development as opposed to healthy
levels in adult cortex (Figure 5C). Overall, the data suggest that
the LE is upregulated for specific MTOR signaling pathways,
which are normally active during healthy cortical development.
Gene Ontology analysis also identified key differences in regu-
lation of membrane potential and postsynaptic membrane
signaling at the LE as compared to healthy development (Fig-
ure 5D). Most genes in these functions are similarly downregu-
lated. However, less coordinate upregulation is found among
pre- and postsynaptic membrane genes except in the vesicular
transport to synapse cluster, which shows a high similarity of
genes upregulated at the LE and in healthy development (Fig-
ure 5D). Different ion channel and transmitter subtypes modulate
different functions in the developing brain versus the adult
brain.
82
For example, NMDA-dependent glutamate currents
enhance the survival of adult neural progenitor cells but are not
required for migration.
83
On the other hand, embryonic GABA
and NMDA-dependent signaling is important for radial migration
of hippocampal neurons,
84
whereas embryonic AMPA receptors
modulate prenatal tangential migration.
85
Intriguingly, our anal-
ysis shows that all tested NMDA receptors are downregulated
similarly to prenatal stages, but none of the AMPA receptor sub-
units show expression patterns resembling early development
(Figure 5E). For example, GRIA1, the only AMPA receptor upre-
gulated in the early development of the healthy cortex (Figure 5E),
is downregulated at the LE (Figure 3A). Conversely, GRIA2, the
only AMPA receptor upregulated at the LE compared to the adult
cortex (Figure 3A), is considerably downregulated at all tested
developmental stages (Figure 5E). In vitro studies of 10–15 DIV
neuronal cultures show that both NMDA and AMPA receptors
Figure 4. The LE and MVP are distinct proepileptic zones
(A) The microvasculature of proliferation is enriched in proepileptic genes. Scatterplot of the ER of the Epi358 genes for the LE and MVP regions. Based on the ER
at the MVP, 32 Epi358 genes were enriched at the MVP compared to the CT, but half of these genes were also enriched at the LE.
(B) The microvasculature is enriched in potential epileptogenic genes. Heatmap of MVP-enriched genes implicated in seizures or synaptogenesis.
(C) Autoimmune epilepsy antigens are enriched at the LE but not the microvascular of proliferation zone. Scatterplot of the FC
Cx
plotted as a function of the ER
between the MVP and LE. Results show that no genes associated with autoimmune epilepsy are enriched at the MVP while nearly all of them are enriched at the
LE. Genes associated with dysplasia are enriched in both the LE and MVP regions.
(D) Differences in the regulation of MTOR pathways at the LE and the microvascular of proliferation zone. Heatmap of the FC
Cx
and ER of genes related to
dysplasia. Growth-related genes for epileptic cortical dysplasia are dysregulated at the LE. KRAS is enriched and upregulated only at the LE whereas PTEN is
enriched at the LE and MVP. NPRL2 is also upregulated at the LE but shows no enrichment as compared to the CT. Most genes involved in cortical dysplasia did
not show strong regional tumor enrichment.
Cell Reports Medicine 5, 101691, August 20, 2024 9
Article
ll
OPEN ACCESS

Figure 5. Peritumoral remodeling partially recapitulates early developmental programs
(A) A majority of dysregulated Epi358 genes recapitulates expression patterns in early development. Top: heatmap of the 254 dysregulated Epi358 genes (177/
354 downregulated, 77/358 upregulated) and the fold change at the LE and multiple developmental stages (EP, LP, IN, CH) compared to healthy adult cortex.
Genes where fold change at the LE resembles any of the tested developmental stages are marked by a cyan bar (left most column). Bottom: bar plot of the percent
of dysregulated genes which resemble development (shades of cyan) stratified by the percent of genes from each functional cluster, MBS, ICT, and CGD.
(B) Recapitulation of early developmental TOR signaling at the LE. Bar plot of genes significantly enriched per growth/migration-related Gene Ontology biological
process. Positive y axis represents the number of genes upregulated and negative y axis represents genes downregulated.
(C) A subset of regulators of MTOR signaling resembles early developmental patterns. Heatmap of the cortical developmental indices for MTOR signaling genes.
The genes in FL glioblastoma tumor and healthy FL per developmental stage were normal ized using a cortical developmental index (CDI, see Methods), revealing
trends in developmental expression. Asterisks (*) denote MTOR-related genes, which had LE expression similar to prenatal levels.
(legend continued on next page)
10 Cell Reports Medicine 5, 101691, August 20, 2024
Article
ll
OPEN ACCESS

can induce calcium ion influx into cells
86
and might be the source
of slow inward calcium currents seen in glioma co-culture as-
says. However, only AMPA receptor antagonists inhibit these
slow inward calcium currents in vitro.
50
Thus, our analysis sug-
gests a predominant AMPA receptor-mediated slow calcium
current in peritumoral networks. In contrast to the age-specific
pattern seen in glutamate receptors, about half of the GABA re-
ceptor subunits resemble an early developmental stage while the
other half more closely resembled adult expression levels. In
healthy neocortex, the impact of GABA receptor signaling on
proliferation is mixed; these receptors limit proliferation of radial
glial type adult neural progenitor cells
87
but promote proliferation
of a subset of type II adult neural progenitor cells with a higher
internal chloride ion concentration.
88
Our analysis shows a
marked decrease in SLC12A5 (KCC2) at the LE (Figure 3A), but
no compensatory reduction in SLC12A2 (NKCC1), suggesting
a higher internal chloride ion concentration resembling type II
adult neural precursor cells. Thus, at the LE, SLC12A5 levels
resemble early development, but SLC12A2 more closely resem-
bles adult levels, which may suggest that transcriptional dysre-
gulation at the LE suppresses SLC12A5 expression independent
of SLC12A2 (Figure 5E). Overall, our data suggest peritumoral re-
modeling of a subset of genes involved in ion homeostasis re-
sembles early development, highlighting the importance of
viewing tumor-induced transcription profiles through the lens
of developmental programs and mechanisms.
DISCUSSION
Spatial analysis of RNA enrichment and dysregulation patterns
of genes linked to monogenic epilepsy phenotypes reveals an
unexpectedly complex molecular substrate for the emergence
of epistatic gene interactions in the glioblastoma cortical micro-
environment. A majority of clinically validated, proepileptic gene
pathways are strongly dysregulated at the tumor LE, identifying a
priority gene set underlying proepileptic and antiepileptic circuit
remodeling. Both gain and loss of expression in these genes,
each known to critically control the balance of network excit-
ability, define a rich network excitability dysregulome that can
modify seizure risk in patients with glioblastoma. This epistatic
complexity replicates the one previously reported for patterns
of genomic ion channel variants in non-tumor human epilepsy
cases, where computational modeling validated their combina-
torial pathogenic effects.
29
The extent of epilepsy-linked tran-
script remodeling therefore provides a parsimonious explanation
for epilepsy risk in individuals with this brain tumor, as well as a
molecular basis for their pharmacoresistance to narrowly tar-
geted conventional antiseizure medicines. A better understand-
ing of epilepsy gene landscape patterns could identify potential
targets and help guide future antiepileptic therapy for glioblas-
toma. For example, none of the greater than 30 currently
approved antiseizure drugs target proepileptic GPCRs,
89
which
are prominently upregulated at the LE.
The spatial arrangement of molecular plasticity is also informa-
tive. Unlike the ‘‘surround inhibition’’ motif characteristic of an
acute convulsant-induced cortical seizure focus,
90
we find that tu-
mor-induced epileptogenesis, which evolves over weeks in glio-
blastoma mouse models,
5
replaces this native inhibitory barrier
with surround excitation. Heterogeneous profiles identified within
tumor subregions distant from the margin suggest additional
distinct intratumoral excitability thresholds responsible for glioma
cell depolarization as wellas microvascular BBB defects that could
guide future clinical diagnostic classification and management.
We identified striking subtype-specific changes in glutamate re-
ceptors at the tumor margin and within the tumor, for example,
the strong upregulation of AMPA receptor subunits that resemble
the complex cortical layer-specific changes in temporal lobe epi-
lepsy tissue.
91
We also identify candidate BBB defects enriched
at the MVP. These highly concordant patterns are masked in
bulk tissue samples,
92
highlighting the value of regional analysis
of malignant peritumoral networks.
The transcriptional control mechanisms driving the striking het-
erogeneity of epilepsy gene dysregulation within regional tumor
excitability niches remain to be explored. Some genes dysregu-
late only in the LE, some in both the LE and tumor, and other
compartmental patterns are evident. Whereas voltage-gated so-
dium and calcium gene dysregulation is LE predominant, enrich-
ment of potassium channels and glutamate receptors extends
into the puretumor zone, a region largely devoid ofrecurrent excit-
atory synaptic connections critical for seizure generation, where
they may nonetheless contribute to paracrine signaling affecting
glial, immune, and tumor cell biology. This source of non-synaptic
depolarization defines an alternative driver of tumor progression
independent of EEG-detectable seizure activity. Extracellular po-
tassium and glutamate excess in both compartments may also
combine to trigger SD waves identified in mouse glioblastoma
models.
5,7
These spatial excitability profiles may also refine glio-
blastoma subtype classification. Isocitrate dehydrogenase (IDH)
genes are currently the primary classifiers of diffuse glioma sub-
types, and their role in hyperexcitability could overlap with many
metabolically linked epilepsy genes.
13
IDH mutation status alters
normal oxidation of isocitrate to alpha-ketoglutarate in mitochon-
dria and affects oncogenesis. In Ivy GAP, IDH1 but not IDH2
expression is low at the LE; both genes are markedly elevated in
the tumor zone, and IDH2 is also elevated at the MVP. Interest-
ingly, IDH3a is a recent candidate risk gene for epilepsy.
93,94
Limitations of the study
A principal limitation to interpreting the pathological transcriptome
is the unclear relationship between transcript density, proteome
(D) Pre- and postsynaptic genes showed subtype-specific changes at the LE compared to normal development. Bar plot of genes significantly enriched ac-
cording to their related Gene Ontology biological process. Positive y axis represents the number of genes upregulated and negative y axis represents down-
regulated genes.
(E) NMDA receptors and a subset of GABA receptors recapitulate early developmental patterns. Heatmap of the cortical developmental indices for major
glutamate and GABA receptor genes. The genes in FL glioblastoma tumor and healthy FL samples per developmental stage were normalized using a cortical
developmental index (CDI, see Methods), revealing subtype-specific trends in developmental expression. Asterisks (*) denote genes with LE levels resembling
prenatal levels.
Cell Reports Medicine 5, 101691, August 20, 2024 11
Article
ll
OPEN ACCESS

lifespan, and the plasticity of intrinsic membrane and synaptic
network firing properties, each of which varies in the brain across
multiple timescales. Both glioblastoma
95
and healthy neurons
96
display activity-driven circadian fluctuations in mRNA transcripts,
which, once alternatively spliced, may be toxic at some synapses
and imperceptible at others, and the levels themselves offer little
insight into the extent of cell type-specific posttranslational modi-
fications that could masktheir final impact. Forexample, migrating
glioblasts remodel more extensively than stationary tumor cells,
97
and both episodic seizures
98,99
and persistent depolarization
100
alter these patterns, rendering the layers of neuron-tumor cross-
talk difficult to unravel. Further ambiguity resides in the potential
for malignantstem cell subclones to producemultiple non-uniform
lineages, as shown by discordant regional case profiles, and the
age of their progeny may extend unevenly across tumor re-
gions.
18,101,102
Therefore, even an anatomically segmented tran-
scriptome environment is challenging tosummarize across space
and time and is only a molecular protomap of network hyperexcit-
ability. Despite these limitations, as a snapshot of malignant brain
tumor biology, this spatial portrait of excitability dyshomeostasis
provides an invaluable reference point and a rich source of candi-
date gene targets for therapeutic exploration aimed at returning
the peritumoral cortical landscape to its native excitability.
STAR+METHODS
Detailed methods are provided in the online version of this paper and include
the following:
dKEY RESOURCES TABLE
dRESOURCES AVAILABILITY
BLead contact
BMaterials availability
BData and code availability
dEXPERIMENTAL MODEL AND SUBJECT DETAILS
dMETHOD DETAILS
BCellular compartmentation and case complexity
BCalculation of enrichment ratio per tumor region
BCalculation of case concordance
BHierarchal clustering of the Epi358 genes into functional groups
based on biological process
BCalculation of fold change of epilepsy genes at the peritumoral LE
vs. healthy cortex
BIdentification of pathogenic copy number variants that are epilepsy
linked
BCuration of gene set of disease-linked genes and presynaptic genes
BComparison of FC
Cx
at the LE to expression levels during healthy
early development
BOverrepresentation analysis of development related functions
BDetermination of cortical developmental index for fold changes
compared to WT
dQUANTIFICATION AND STATISTICAL ANALYSIS
SUPPLEMENTAL INFORMATION
Supplemental information can be found online at https://doi.org/10.1016/j.
xcrm.2024.101691.
ACKNOWLEDGMENTS
This work was supported by NIH/NCI R01CA223388 and the Blue Bird Circle
Foundation (J.L.N.).
AUTHOR CONTRIBUTIONS
J.L.N., V.S., and R.B.P. contributed to the project conceptualization and dis-
cussed results, analytical methods, and control databases. V.S. performed
formal data analysis and prepared data for visualization. J.L.N. and V.S. wrote
and edited the manuscript with support from R.B.P.
DECLARATION OF INTERESTS
The authors declare no competing interests.
Received: March 30, 2024
Revised: May 30, 2024
Accepted: July 25, 2024
Published: August 20, 2024
REFERENCES
1. Vecht, C., Royer-Perron, L., Houillier, C., and Huberfeld, G. (2017). Sei-
zures and Anticonvulsants in Brain Tumours: Frequency, Mechanisms
and Anti-Epileptic Management. Curr. Pharm. Des. 23, 6464–6487.
https://doi.org/10.2174/1381612823666171027130003.
2. Mastall, M., Wolpert, F., Gramatzki, D., Imbach, L., Becker, D., Schmick,
A., Hertler, C., Roth, P., Weller, M., and Wirsching, H.-G. (2021). Survival
of brain tumour patients with epilepsy. Brain 144, 3322–3327. https://doi.
org/10.1093/brain/awab188.
3. Mellinghoff, I.K., Van Den Bent, M.J., Blumentha l, D.T., Touat, M., Peters,
K.B., Clarke, J., Mendez, J., Yust-Katz, S., Welsh, L., Mason, W.P., et al.
(2023). Vorasidenib in IDH1- or IDH2-Mutant Low-Grade Glioma. N. Engl.
J. Med. 389, 589–601. https://doi.org/10.1056/NEJMoa2304194.
4. Yu, K., Lin, C.-C.J., Hatcher, A., Lozzi, B., Kong, K., Huang-Hobbs, E.,
Cheng, Y.-T., Beechar, V.B., Zhu, W., Zhang, Y., et al. (2020). PIK3CA
variants selectively initiate brain hyperactivity during gliomagenesis. Na-
ture 578, 166–171. https://doi.org/10.1038/s41586-020-1952-2.
5. Hatcher, A., Yu, K., Meyer, J., Aiba, I., Deneen, B., and Noebels, J.L.
(2020). Pathogenesis of peritumoral hyperexcitability in an immunocom-
petent CRISPR-based glioblastoma model. J. Clin. Invest. 130, 2286–
2300. https://doi.org/10.1172/JCI133316.
6. John Lin, C.-C., Yu, K., Hatcher, A., Huang, T.-W., Lee, H.K., Carlson, J.,
Weston, M.C., Chen, F., Zhang, Y., Zhu, W., et al. (2017). Identification of
diverse astrocyte populations and their malignant analogs. Nat. Neuro-
sci. 20, 396–405. https://doi.org/10.1038/nn.4493.
7. Curry, R.N., Aiba, I., Meyer, J., Lozzi, B., Ko, Y., McDonald, M.F., Rose-
nbaum, A., Cervantes, A., Huang-Hobbs, E., Cocito, C., et al. (2023). Gli-
oma epileptiform activity and progression are driven by IGSF3-mediated
potassium dysregulation. Neuron 111, 682–695.e9. https://doi.org/10.
1016/j.neuron.2023.01.013.
8. Meyer, J., Yu, K., Luna-Figueroa, E., Deneen, B., and Noebels, J. (2024).
Glioblastoma disrupts cortical network activity at multiple spatial and
temporal scales. Nat. Commun. 15, 4503. https://doi.org/10.1038/
s41467-024-48757-5.
9. Robert, S.M., Buckingham, S.C., Campbell, S.L., Robel, S., Holt, K.T.,
Ogunrinu-Babarinde, T., Warren, P.P., White, D.M., Reid, M.A., Es-
chbacher, J.M., et al. (2015). SLC7A11 expression is associated with sei-
zures and predicts poor survival in patients with malignant glioma. Sci.
Transl. Med. 7, 289ra86. https://doi.org/10.1126/scitranslmed.aaa8103.
10. Tewari, B.P., Chaunsali, L., Campbell, S.L., Patel, D.C., Goode, A.E., and
Sontheimer, H. (2018). Perineuronal nets decrease membrane capaci-
tance of peritumoral fast spiking interneurons in a model of epilepsy.
Nat. Commun. 9, 4724. https://doi.org/10.1038/s41467-018-07113-0.
11. Yeo, A.T., Rawal, S., Delcuze, B., Christofides, A., Atayde, A., Strauss, L.,
Balaj, L., Rogers, V.A., Uhlmann, E.J., Varma, H., et al. (2022). Single-cell
RNA sequencing reveals evolution of immune landscape during glioblas-
toma progression. Nat. Immunol. 23, 971–984. https://doi.org/10.1038/
s41590-022-01215-0.
12 Cell Reports Medicine 5, 101691, August 20, 2024
Article
ll
OPEN ACCESS

12. Guerrini, R., Balestrini, S., Wirrell, E.C., and Walker, M.C. (2021). Mono-
genic Epilepsies: Disease Mechanisms, Clinical Phenotypes, and Tar-
geted Therapies. Neurology 97, 817–831. https://doi.org/10.1212/WNL.
0000000000012744.
13. Tumiene, B., Ferreira, C.R., and Van Karnebeek, C.D.M. (2022). 2022
Overview of Metabolic Epilepsies. Genes 13, 508. https://doi.org/10.
3390/genes13030508.
14. Kumar, P., Lim, A., Hazirah, S.N., Chua, C.J.H., Ngoh, A., Poh, S.L., Yeo,
T.H., Lim, J., Ling, S., Sutamam, N.B., et al. (2022). Single-cell transcrip-
tomics and surface epitope detection in human brain epileptic lesions
identifies pro-inflammatory signaling. Nat. Neurosci. 25, 956–966.
https://doi.org/10.1038/s41593-022-01095-5.
15. Niesen, C.E., Xu, J., Fan, X., Li, X., Wheeler, C.J., Mamelak, A.N., and
Wang, C. (2013). Transcriptomic Profiling of Human Peritumoral
Neocortex Tissues Revealed Genes Possibly Involved in Tumor-
Induced Epilepsy. PLoS One 8, e56077. https://doi.org/10.1371/jour-
nal.pone.0056077.
16. Kalita, O., Sporikova, Z., Hajduch, M., Megova Houdov a,M., Slavkovsky ,
R., Hrabalek, L., Halaj, M., Klementova, Y., Dolezel, M., Drabek, J., et al.
(2021). The Influence of Gene Aberrations on Survival in Resected IDH
Wildtype Glioblastoma Patients: A Single-Institution Study. Curr. Oncol.
28, 1280–1293. https://doi.org/10.3390/curroncol28020122.
17. Pollak, J., Rai, K.G., Funk, C.C., Arora, S., Lee, E., Zhu, J., Price, N.D.,
Paddison, P.J., Ramirez, J.-M., and Rostomily, R.C. (2017). Ion channel
expression patterns in glioblastoma stem cells with functional and ther-
apeutic implications for malignancy. PLoS One 12, e0172884. https://
doi.org/10.1371/journal.pone.0172884.
18. Neftel, C., Laffy, J., Filbin, M.G., Hara, T., Shore, M.E., Rahme, G.J., Rich-
man, A.R., Silverbush, D., Shaw, M.L., Hebert, C.M., et al. (2019). An Inte-
grative Model of Cellular States, Plasticity, and Genetics for Glioblas-
toma. Cell 178, 835–849.e21. https://doi.org/10.1016/j.cell.2019.06.024.
19. Puchalski, R.B., Shah, N., Miller, J., Dalley, R., Nomura, S.R., Yoon, J.-G.,
Smith, K.A., Lankerovich, M., Bertagnolli, D., Bickley, K., et al. (2018). An
anatomic transcriptional atlas of human glioblastoma. Science 360,
660–663. https://doi.org/10.1126/science.aaf2666.
20. Ka
´dkova
´, A., Murach, J., Østergaard, M., Malsam, A., Malsam, J., Loli-
cato, F., Nickel, W., So
¨llner, T.H., and Sørensen, J.B. (2024). SNAP25 dis-
ease mutations change the energy landscape for synaptic exocytosis
due to aberrant SNARE interactions. Elife 12, RP88619. https://doi.org/
10.7554/eLife.88619.
21. Huang, Q., Lian, C., Dong, Y., Zeng, H., Liu, B., Xu, N., He, Z., and Guo, H.
(2021). SNAP25 Inhibits Glioma Progression by Regulating Synapse
Plasticity via GLS-Mediated Glutaminolysis. Front. Oncol. 11, 698835.
https://doi.org/10.3389/fonc.2021.698835.
22. Chernova, O.B., Somerville, R.P., and Cowell, J.K. (1998). A novel gene,
LGI1, from 10q24 is rearranged and downregulated in malignant brain tu-
mors. Oncogene 17, 2873–2881. https://doi.org/10.1038/sj.onc.
1202481.
23. Schulte, U., Thumfart, J.-O., Klo
¨cker, N., Sailer, C.A., Bildl, W., Biniossek,
M., Dehn, D., Deller, T., Eble, S., Abbass, K., et al. (2006). The Epilepsy-
Linked Lgi1 Protein Assembles into Presynaptic Kv1 Channels and In-
hibits Inactivation by Kvb1. Neuron 49, 697–706. https://doi.org/10.
1016/j.neuron.2006.01.033.
24. Nobile, C., Michelucci, R., Andreazza, S., Pasini, E., Tosatto, S.C.E., and
Striano, P. (2009). LGI1 mutations in autosomal dominant and sporadic
lateral temporal epilepsy. Hum. Mutat. 30, 530–536. https://doi.org/10.
1002/humu.20925.
25. Uhle
´n, M., Fagerberg, L., Hallstro
¨m, B.M., Lindskog, C., Oksvold, P.,
Mardinoglu, A., Sivertsson, A
˚., Kampf, C., Sjo
¨stedt, E., Asplund, A.,
et al. (2015). Tissue-based map of the human proteome. Science 347,
1260419. https://doi.org/10.1126/science.1260419.
26. Carvill, G.L., Matheny, T., Hesselb erth, J., and Demarest, S. (2021). Hap-
loinsufficiency, Dominant Negative, and Gain-of-Function Mechanisms
in Epilepsy: Matching Therapeutic Approach to the Pathophysiology.
Neurotherapeutics 18, 1500–1514. https://doi.org/10.1007/s13311-
021-01137-z.
27. Br€
unger, T., Pe
´rez-Palma, E., Montanucci, L., Nothnagel, M., Møller,
R.S., Schorge, S., Zuberi, S., Symonds, J., Lemke, J.R., Brunklaus, A.,
et al. (2023). Conserved patterns across ion channels correlate with
variant pathogenicity and clinical phenotypes. Brain 146, 923–934.
https://doi.org/10.1093/brain/awac305.
28. Glasscock, E., Qian, J., Yoo, J.W., and Noebels, J.L. (2007). Masking ep-
ilepsy by combining two epilepsy genes. Nat. Neurosci. 10, 1554–1558.
https://doi.org/10.1038/nn1999.
29. Klassen, T., Davis, C., Goldman, A., Burgess, D., Chen, T., Wheeler, D.,
McPherson, J., Bourquin, T., Lewis, L., Villasana, D., et al. (2011). Exome
Sequencing of Ion Channel Genes Reveals Complex Profiles Confound-
ing Personal Risk Assessment in Epilepsy. Cell 145, 1036–1048. https://
doi.org/10.1016/j.cell.2011.05.025.
30. Conti, L., Crisafulli, L., Caldera, V., Tortoreto, M., Brilli, E., Conforti, P.,
Zunino, F., Magrassi, L., Schiffer, D., and Cattaneo, E. (2012). REST Con-
trols Self-Renewal and Tumorigenic Competence of Human Glioblas-
toma Cells. PLoS One 7, e38486. https://doi.org/10.1371/journal.pone.
0038486.
31. Krishna, S., Choudhury, A., Keough, M.B., Seo, K., Ni, L., Kakaizada, S.,
Lee, A., Aabedi, A., Popova, G., Lipkin, B., et al. (2023). Glioblastoma re-
modelling of human neural circuits decreases survival. Nature 617,
599–607. https://doi.org/10.1038/s41586-023-06036-1.
32. Mayo, S., Go
´mez-Manjo
´n, I., Marco-Herna
´ndez, A.V., Ferna
´ndez-Martı
´-
nez, F.J., Camacho, A., and Martı
´nez, F. (2023). N-Type Ca Channel in
Epileptic Syndromes and Epilepsy: A Systematic Review of Its Genetic
Variants. Int. J. Mol. Sci. 24, 6100. https://doi.org/10.3390/
ijms24076100.
33. Ernst, W.L., Zhang, Y., Yoo, J.W., Ernst, S.J., and Noebels, J.L. (2009).
Genetic Enhancement of Thalamocortical Network Activity by Elevating
a1G-Mediated Low-Voltage-Activated Calcium Current Induces Pure
Absence Epilepsy. J. Neurosci. 29, 1615–1625. https://doi.org/10.
1523/JNEUROSCI.2081-08.2009.
34. Song, I., Kim, D., Choi, S., Sun, M., Kim, Y., and Shin, H.-S. (2004). Role of
the a1G T-Type Calcium Channel in Spontaneous Absence Seizures in
Mutant Mice. J. Neurosci. 24, 5249–5257. https://doi.org/10.1523/
JNEUROSCI.5546-03.2004.
35. D’Adamo, M.C., Catacuzzeno, L., Di Giovanni, G., Franciolini, F., and
Pessia, M. (2013). K+ channelepsy: progress in the neurobiology of po-
tassium channels and epilepsy. Front. Cell. Neurosci. 7, 134. https://
doi.org/10.3389/fncel.2013.00134.
36. Niday, Z., and Tzingounis, A.V. (2018). Potassium Channel Gain of Func-
tion in Epilepsy: An Unresolved Paradox. Neuroscientist 24, 368–380.
https://doi.org/10.1177/1073858418763752.
37. Trimmer, J.S. (2015). Subcellular Localization of K+ Channels in Mamma-
lian Brain Neurons: Remarkable Precision in the Midst of Extraordinary
Complexity. Neuron 85, 238–256. https://doi.org/10.1016/j.neuron.
2014.12.042.
38. Cuddapah, V.A., Turner, K.L., Seifert, S., and Sontheimer, H. (2013). Bra-
dykinin-Induced Chemotaxis of Human Gliomas Requires the Activation
of K
Ca
3.1 and ClC-3. J. Neurosci. 33, 1427–1440. https://doi.org/10.
1523/JNEUROSCI.3980-12.2013.
39. Prevarskaya, N., Skryma, R., and Shuba, Y. (2018). Ion Channels in Can-
cer: Are Cancer Hallmarks Oncochannelopathies? Physiol. Rev. 98,
559–621. https://doi.org/10.1152/physrev.00044.2016.
40. Petersson, S., Persson, A.-S., Johansen, J.E., Ingvar, M., Nilsson, J., Kle-
ment, G., Arhem, P., Schalling, M., and Lavebratt, C. (2003). Truncation
of the Shaker-like voltage-gated potassium channel, Kv1.1, causes me-
gencephaly. Eur. J. Neurosci. 18, 3231–3240. https://doi.org/10.1111/j.
1460-9568.2003.03044.x.
Cell Reports Medicine 5, 101691, August 20, 2024 13
Article
ll
OPEN ACCESS

41. Hills, K.E., Kostarelos, K., and Wykes, R.C. (2022). Converging Mecha-
nisms of Epileptogenesis and Their Insight in Glioblastoma. Front. Mol.
Neurosci. 15, 903115. https://doi.org/10.3389/fnmol.2022.903115.
42. Eikermann-Haerter, K., Yuzawa, I., Qin, T., Wang, Y., Baek, K., Kim, Y.R.,
Hoffmann, U., Dilekoz, E., Waeber, C., Ferrari, M.D., et al. (2011).
Enhanced Subcortical Spreading Depression in Familial Hemiplegic
Migraine Type 1 Mutant Mice. J. Neurosci. 31, 5755–5763. https://doi.
org/10.1523/JNEUROSCI.5346-10.2011.
43. Reiffurth, C., Alam, M., Zahedi-Khorasani, M., Major, S., and Dreier, J.P.
(2020). Na
+
/K
+
-ATPase aisoform deficiency results in distinct
spreading depolarization phenotypes. J. Cereb. Blood Flow Metab. 40,
622–638. https://doi.org/10.1177/0271678X19833757.
44. Aiba, I., Ning, Y., and Noebels, J.L. (2023). A hyperthermic seizure un-
leashes a surge of spreading depolarizations in Scn1a-deficient mice.
JCI Insight 8, e170399. https://doi.org/10.1172/jci.insight.170399.
45. Kessi, M., Peng, J., Duan, H., He, H., Chen, B., Xiong, J., Wang, Y., Yang,
L., Wang, G., Kiprotich, K., et al. (2022). The Contribution of HCN Chan-
nelopathies in Different Epileptic Syndromes, Mechanisms, Modulators,
and Potential Treatment Targets: A Systematic Review. Front. Mol. Neu-
rosci. 15, 807202. https://doi.org/10.3389/fnmol.2022.807202.
46. Farago
´, N., Kocsis, A
´.K., Brasko
´, C., Lovas, S., Ro
´zsa, M., Baka, J., Ko-
va
´cs, B., Mikite, K., Szemenyei, V., Molna
´r, G., et al. (2016). Human
neuronal changes in brain edema and increased intracranial pressure.
Acta Neuropathol. Commun. 4, 78. https://doi.org/10.1186/s40478-
016-0356-x.
47. Nakamura, Y., Inoue, A., Nishikawa, M., Ohnis hi, T., Yano, H., Kanemura,
Y., Ohtsuka, Y., Ozaki, S., Kusakabe, K., Suehiro, S., et al. (2022). Quan-
titative measurement of peritumoral concentrations of glutamate,
N-acetyl aspartate, and lactate on magnetic resonance spectroscopy
predicts glioblastoma-related refractory epilepsy. Acta Neurochir. 164,
3253–3266. https://doi.org/10.1007/s00701-022-05363-y.
48. Buckingham, S.C., Campbell, S.L., Haas, B.R., Montana, V., Robel, S.,
Ogunrinu, T., and Sontheimer, H. (2011). Glutamate release by primary
brain tumors induces epileptic activity. Nat. Med. 17, 1269–1274.
https://doi.org/10.1038/nm.2453.
49. Takano, T., Lin, J.H., Arcuino, G., Gao, Q., Yang, J., and Nedergaard, M.
(2001). Glutamate release promotes growth of malignant gliomas. Nat.
Med. 7, 1010–1015. https://doi.org/10.1038/nm0901-1010.
50. Venkataramani, V., Tanev, D.I., Strahle, C., Studier-Fischer, A., Fank-
hauser, L., Kessler, T., Ko
¨rber, C., Kardorff, M., Ratliff, M., Xie, R., et al.
(2019). Glutamatergic synaptic input to glioma cells drives brain tumour
progression. Nature 573, 532–538. https://doi.org/10.1038/s41586-
019-1564-x.
51. Tardito, S., Oudin, A., Ahmed, S.U., Fack, F., Keunen, O., Zheng, L.,
Miletic, H., Sakariassen, P.Ø., Weinstock, A., Wagner, A., et al. (2015).
Glutamine synthetase activity fuels nucleotide biosynthesis and supports
growth of glutamine-restricted glioblastoma. Nat. Cell Biol. 17, 1556–
1568. https://doi.org/10.1038/ncb3272.
52. Ye, Z.-C., Rothstein, J.D., and Sontheimer, H. (1999). Compromised
Glutamate Transport in Human Glioma Cells: Reduction–
Mislocalization of Sodium-Dependent Glutamate Transporters and
Enhanced Activity of Cystine–Glutamate Exchange. J. Neurosci. 19,
10767–10777. https://doi.org/10.1523/JNEUROSCI.19-24-10767.1999.
53. Ishiuchi, S., Yoshida, Y., Sugawara, K., Aihara, M., Ohtani, T., Watanabe,
T., Saito, N., Tsuzuki, K., Okado, H., Miwa, A., et al. (2007). Ca
2+
-Perme-
able AMPA Receptors Regulate Growth of Human Glioblastoma via Akt
Activation. J. Neurosci. 27, 7987–8001. https://doi.org/10.1523/JNEUR-
OSCI.2180-07.2007.
54. Taylor, K.R., Barron, T., Hui, A., Spitzer, A., Yalc¸ in, B., Ivec, A.E., Ger-
aghty, A.C., Hartmann, G.G., Arzt, M., Gillespie, S.M., et al. (2023). Gli-
oma synapses recruit mechanisms of adaptive plasticity. Nature 623,
366–374. https://doi.org/10.1038/s41586-023-06678-1.
55. Chatron, N., Becker, F., Morsy, H., Schmi dts, M., Hardies, K., Tuysuz, B.,
Roselli, S., Najafi, M., Alkaya, D.U., Ashrafzadeh, F., et al. (2020). Bi-
allelic GAD1 variants cause a neonatal onset syndromic developmental
and epileptic encephalopathy. Brain 143, 1447–1461. https://doi.org/
10.1093/brain/awaa085.
56. Kash, S.F., Johnson, R.S., Tecott, L.H., No ebels, J.L., Mayfield, R.D., Ha-
nahan, D., and Baekkeskov, S. (1997). Epilepsy in mice deficient in the
65-kDa isoform of glutamic acid decarboxylase. Proc. Natl. Acad. Sci.
USA 94, 14060–14065. https://doi.org/10.1073/pnas.94.25.14060.
57. Klitten, L.L., Møller, R.S., Ravn, K., Hjalgrim, H., and Tommerup, N.
(2011). Duplication of MAOA, MAOB, and NDP in a patient with mental
retardation and epilepsy. Eur. J. Hum. Genet. 19, 1–2. https://doi.org/
10.1038/ejhg.2010.149.
58. Maillard, P.Y., Baer, S., Schaefer, E
´., Desnous, B., Villeneuve, N., Le
´pine,
A., Fabre, A., Lacoste, C., El Chehadeh, S., Piton, A., et al. (2022). Molec-
ular and clinical descriptions of patients with GABA
A
receptor gene var-
iants ( GABRA1 , GABRB2 , GABRB3 , GABRG2 ): A cohort study, review
of literature, and genotype–phenotype correlation. Epilepsia 63, 2519–
2533. https://doi.org/10.1111/epi.17336.
59. Dossi, E., and Huberfeld, G. (2023). GABAergic circuits drive focal sei-
zures. Neurobiol. Dis. 180, 106102. https://doi.org/10.1016/j.nbd.2023.
106102.
60. Tian, J., Chau, C., Hales, T.G., and Kaufman, D.L. (1999). GABAA recep-
tors mediate inhibition of T cell responses. J. Neuroimmunol. 96, 21–28.
https://doi.org/10.1016/S0165-5728(98)00264-1.
61. Blanchart, A., Fernando, R., Ha
¨ring, M., Assaife-Lopes, N., Romanov,
R.A., Anda
¨ng, M., Harkany, T., and Ernfors, P. (2017). Endogenous
GABAA receptor activity suppresses glioma growth. Oncogene 36,
777–786. https://doi.org/10.1038/onc.2016.245.
62. Smits, A., Jin, Z., Elsir, T., Pedder, H., Niste
´r, M., Alafuzoff, I., Dimberg,
A., Edqvist, P.-H., Ponte
´n, F., Aronica, E., and Birnir, B. (2012). GABA-
A Channel Subunit Expression in Human Glioma Correlates with Tumor
Histology and Clinical Outcome. PLoS One 7, e37041. https://doi.org/
10.1371/journal.pone.0037041.
63. Levinson, S., Tran, C.H., Barry, J., Viker, B., Levine, M.S., Vinters, H.V.,
Mathern, G.W., and Cepeda, C. (2020). Paroxysmal Discharges in Tissue
Slices From Pediatric Epilepsy Surgery Patients: Critical Role of GABAB
Receptors in the Generation of Ictal Activity. Front. Cell. Neurosci. 14, 54.
https://doi.org/10.3389/fncel.2020.00054.
64. Duy, P.Q., David, W.B., and Kahle, K.T. (2019). Identification of KCC2
Mutations in Human Epilepsy Suggests Strategies for Therapeutic Trans-
porter Modulation. Front. Cell. Neurosci. 13, 515. https://doi.org/10.
3389/fncel.2019.00515.
65. Moore, Y.E., Deeb, T.Z., Chadchanka r, H., Brandon, N.J., and Moss, S.J.
(2018). Potentiating KCC2 activity is sufficient to limit the onset and
severity of seizures. Proc. Natl. Acad. Sci. USA 115, 10166–10171.
https://doi.org/10.1073/pnas.1810134115.
66. Pallud, J., Le Van Quyen, M., Bielle, F., Pellegrino, C., Varlet, P., Cresto,
N., Baulac, M., Duyckaerts, C., Kourdougli, N., Chazal, G., et al. (2014).
Cortical GABAergic excitation contributes to epileptic activities around
human glioma. Sci. Transl. Med. 6, 244ra89. https://doi.org/10.1126/sci-
translmed.3008065.
67. Campbell, S.L., Robel, S., Cuddapah, V.A., Robert, S., Buckingham,
S.C., Kahle, K.T., and Sontheimer, H. (2015). GABAergic disinhibition
and impaired KCC2 cotransporter activity underlie tumor-associated ep-
ilepsy: Reduced KCC2 Underlie Tumor-Related Epilepsy. Glia 63, 23–36.
https://doi.org/10.1002/glia.22730.
68. MacKenzie, G., O’Toole, K.K., Moss, S.J., and Maguire, J. (2016).
Compromised GABAergic inhibition contributes to tumor-associated ep-
ilepsy. Epilepsy Res. 126, 185–196. https://doi.org/10.1016/j.eplepsyres.
2016.07.010.
69. Montgomery, M.K., Kim, S.H., Dovas, A., Zhao, H.T., Goldberg, A.R., Xu,
W., Yagielski, A.J., Cambareri, M.K., Patel, K.B., Mela, A., et al. (2020).
Glioma-Induced Alterations in Neuronal Activity and Neurovascular
Coupling during Disease Progression. Cell Rep. 31, 107500. https://
doi.org/10.1016/j.celrep.2020.03.064.
14 Cell Reports Medicine 5, 101691, August 20, 2024
Article
ll
OPEN ACCESS

70. Ren, Y., Liu, Y., Wu, H., Meng, Q., Zhang, J., Li, H., Dong, S., Lian, H., Du,
C., and Zhang, H. (2023). Subdural osteoma in an adolescent patient with
epilepsy: an unusual case report and literature review. Childs Nerv. Syst.
39, 3281–3288. https://doi.org/10.1007/s00381-023-06015-x.
71. Lo
¨scher, W., and Friedman, A. (2020). Structural, Molecular, and Func-
tional Alterations of the Blood-Brain Barrier during Epileptogenesis and
Epilepsy: A Cause, Consequence, or Both? Int. J. Mol. Sci. 21, 591.
https://doi.org/10.3390/ijms21020591.
72. Geis, C., Planaguma
`, J., Carren
˜o, M., Graus, F., and Dalmau, J. (2019).
Autoimmune seizures and epilepsy. J. Clin. Invest. 129, 926–940.
https://doi.org/10.1172/JCI125178.
73. Flammer, J., Neziraj, T., R€
uegg, S., and Pro
¨bstel, A.-K. (2023). Immune
Mechanisms in Epileptogenesis: Update on Diagnosis and Treatment
of Autoimmune Epilepsy Syndromes. Drugs 83, 135–158. https://doi.
org/10.1007/s40265-022-01826-9.
74. Vogrig, A., Gigli, G.L., Segatti, S., Corazza, E., Marini, A., Bernardini, A.,
Valent, F., Fabris, M., Curcio, F., Brigo, F., et al. (2020). Epidemiology of
paraneoplastic neurological syndromes: a population-based study.
J. Neurol. 267, 26–35. https://doi.org/10.1007/s00415-019-09544-1.
75. Giammello, F., Galletta, K., Grillo, F., Brizzi, T., Cavallaro, M., Mormina,
E., Scelzo, E., Allegra, C., Stancanelli, C., Rodolico, C., et al. (2023). Para-
neoplastic neurological syndromes of the central nervous system: a sin-
gle institution 7-year case series. Acta Neurol. 123, 1355–1369. https://
doi.org/10.1007/s13760-023-02232-y.
76. Rokutanda, T., Inatomi, Y., Yonehara, T., Takahashi, Y., Hirano, T., and
Uchino, M. (2008). A case of glioblastoma misdiagnosed initially due to
positive finding of anti-glutamate receptor antibody. Rinsho Shinkeigaku
48, 497–500. https://doi.org/10.5692/clinicalneurol.48.497.
77. Lo
´pez-Rivera, J.A., Leu, C., Macnee, M., Khoury, J., Hoffmann, L., Coras,
R., Kobow, K., Bhattarai, N., Pe
´rez-Palma, E., Hamer, H., et al. (2023).
The genomic landscape across 474 surgically accessible epileptogenic
human brain lesions. Brain 146, 1342–1356. https://doi.org/10.1093/
brain/awac376.
78. Kim, J.K., Cho, J., Kim, S.H., Kang, H.-C., Kim, D.-S., Kim, V.N., and Lee,
J.H. (2019). Brain somatic mutations in MTOR reveal translational dysre-
gulations underlying intractable focal epilepsy. J. Clin. Invest. 129, 4207–
4223. https://doi.org/10.1172/JCI127032.
79. Chung, C., Yang, X., Bae, T., Vong, K.I., Mittal, S., Donkels, C., Westley
Phillips, H., Li, Z., Marsh, A.P.L., Breuss, M.W., et al. (2023). Comprehen-
sive multi-omic profiling of somatic mutations in malformations of cortical
development. Nat. Genet. 55, 209–220. https://doi.org/10.1038/s41588-
022-01276-9.
80. Schreck, K.C., Allen, A.N., Wang, J., and Pratilas, C.A. (2020). Combina-
tion MEK and mTOR inhibitor therapy is active in models of glioblastoma.
Neuro-Oncol. Adv 2, vdaa138. https://doi.org/10.1093/noajnl/vdaa138.
81. Jung, E., Alfonso, J., Monyer, H., Wick, W., and Winkler, F. (2020).
Neuronal signatures in cancer. Int. J. Cancer 147, 3281–3291. https://
doi.org/10.1002/ijc.33138.
82. Go
¨tz, M., Nakafuku, M., and Petrik, D. (2016). Neurogenesis in the Devel-
oping and Adult Brain-Similarities and Key Differences. Cold Spring
Harb. Perspect. Biol. 8, a018853. https://doi.org/10.1101/cshperspect.
a018853.
83. Platel, J.-C., Dave, K.A., Gordon, V., Lacar, B., Rubio, M.E., and Bordey,
A. (2010). NMDA receptors activated by subventricular zone astrocytic
glutamate are critical for neuroblast survival prior to entering a synaptic
network. Neuron 65, 859–872. https://doi.org/10.1016/j.neuron.2010.
03.009.
84. Manent, J.-B., Demarque, M., Jorquera, I., Pellegrino, C., Ben-Ari, Y.,
Aniksztejn, L., and Represa, A. (2005). A noncanonical release of GABA
and glutamate modulates neuronal migration. J. Neurosci. 25, 4755–
4765. https://doi.org/10.1523/JNEUROSCI.0553-05.2005.
85. Manent, J.-B., Jorquera, I., Ben-Ari, Y., Aniksztejn, L., and Represa, A.
(2006). Glutamate acting on AMPA but not NMDA receptors modulates
the migration of hippocampal interneurons. J. Neurosci. 26, 5901–
5909. https://doi.org/10.1523/JNEUROSCI.1033-06.2006.
86. Fischer, W., Franke, H., Scheibler, P., Allgaier, C., and Illes, P. (2002).
AMPA-induced Ca(2+) influx in cultured rat cortical nonpyramidal neuro-
nes: pharmacological characterization using fura-2 microfluorimetry.
Eur. J. Pharmacol. 438, 53–62. https://doi.org/10.1016/s0014-2999(02)
01296-7.
87. Liu, X., Wang, Q., Haydar, T.F., and Bordey, A. (2005). Nonsynaptic
GABA signaling in postnatal subventricular zone controls proliferation
of GFAP-expressing progenitors. Nat. Neurosci. 8, 1179–1187. https://
doi.org/10.1038/nn1522.
88. Tozuka, Y., Fukuda, S., Namba, T., Seki, T., and Hisatsune, T. (2005).
GABAergic excitation promotes neuronal differentiation in adult hippo-
campal progenitor cells. Neuron 47, 803–815. https://doi.org/10.1016/j.
neuron.2005.08.023.
89. Yu, Y., Nguyen, D.T., and Jiang, J. (2019). G protein-coupled receptors in
acquired epilepsy: Druggability and translatability. Prog. Neurobiol. 183,
101682. https://doi.org/10.1016/j.pneurobio.2019.101682.
90. Prince, D.A., and Wilder, B.J. (1967). Control Mechanisms in Cortical
Epileptogenic Foci*: ‘‘Surround’’ Inhibition. Arch. Neurol. 16, 194–202.
https://doi.org/10.1001/archneur.1967.00470200082007.
91. Pfisterer, U., Petukhov, V., Dem harter, S., Meichsner, J., Thompson, J.J.,
Batiuk, M.Y., Asenjo-Martinez, A., Vasistha, N.A., Thakur, A., Mikkelsen,
J., et al. (2020). Identification of epilepsy-associated neuronal subtypes
and gene expression underlying epileptogenesis. Nat. Commun. 11,
5038. https://doi.org/10.1038/s41467-020-18752-7.
92. Elias, A.F., Lin, B.C., and Piggott, B.J. (2023). Ion Channels in Gliomas—
From Molecular Basis to Treatment. Int. J. Mol. Sci. 24, 2530. https://doi.
org/10.3390/ijms24032530.
93. Drumm, M.R., Wang, W., Sears, T.K., Bell-Burdett, K., Javier, R., Cotton,
K.Y., Webb, B., Byrne, K., Unruh, D., Thirunavu, V., et al. (2023). Postop-
erative risk of IDH mutant glioma–associated seizures and their potential
management with IDH mutant inhibitors. J. Clin. Invest. 133, e168035.
https://doi.org/10.1172/JCI168035.
94. Van Opijnen, M.P., Tesileanu, C.M.S., Dirven, L., Van Der Meer, P.B., Wij-
nenga, M.M.J., Vincent, A.J.P.E., Broekman, M.L.D., Dubbink, H.J.,
Kros, J.M., Van Duinen, S.G., et al. (2023). IDH1/2 wildtype gliomas grade
2 and 3 with molecular glioblastoma-like profile have a distinct course of
epilepsy compared to IDH1/2 wildtype glioblastomas. Neuro Oncol. 25,
701–709. https://doi.org/10.1093/neuonc/noac197.
95. Dong, Z., Zhang, G., Qu, M., Gimple, R.C., Wu, Q., Qiu, Z., Prager, B.C.,
Wang, X., Kim, L.J.Y., Morton, A.R., et al. (2019). Targeting Glioblastoma
Stem Cells through Disruption of the Circadian Clock. Cancer Discov. 9,
1556–1573. https://doi.org/10.1158/2159-8290.CD-19-0215.
96. Gonzalez, J.C., Lee, H., Vincent, A.M., Hill, A.L., Goode, L.K., King, G.D.,
Gamble, K.L., Wadiche, J.I., and Overstreet-Wadiche, L. (2023). Circa-
dian regulation of dentate gyrus excitability mediated by G-protein
signaling. Cell Rep. 42, 112039. https://doi.org/10.1016/j.celrep.2023.
112039.
97. Cuddapah, V.A., Robel, S., Watkins, S., and Sontheimer, H. (2014). A
neurocentric perspective on glioma invasion. Nat. Rev. Neurosci. 15,
455–465. https://doi.org/10.1038/nrn3765.
98. Berger, T.C., Vigeland, M.D., Hjorthaug, H.S., Etholm, L., Nome, C.G.,
Taubøll, E., Heuser, K., and Selmer, K.K. (2019). Neuronal and glial
DNA methylation and gene expression changes in early epileptogenesis.
PLoS One 14, e0226575. https://doi.org/10.1371/journal.pone.0226575.
99. Soon, H.R., Gaunt, J.R., Bansal, V.A., Lenherr, C., Sze, S.K., and Ch’ng,
T.H. (2023). Seizure enhances SUMOylation and zinc-finger transcrip-
tional repression in neuronal nuclei. iScience 26, 107707. https://doi.
org/10.1016/j.isci.2023.107707.
100. Xu, X., Johnson, Z., and Xie, H. (2022). Neuronal Depolarization Induced
RNA m5C Methylation Changes in Mouse Cortical Neurons. Biology 11,
988. https://doi.org/10.3390/biology11070988.
Cell Reports Medicine 5, 101691, August 20, 2024 15
Article
ll
OPEN ACCESS

101. Tirosh, I., and Suva
`, M.L. (2020). Tackling the Many Facets of Glioblas-
toma Heterogeneity. Cell Stem Cell 26, 303–304. https://doi.org/10.
1016/j.stem.2020.02.005.
102. Couturier, C.P., Ayyadhury, S., Le, P.U., Nadaf, J., Monlong, J., Riva, G.,
Allache, R., Baig, S., Yan, X., Bourgey, M., et al. (2020). Single-cell RNA-
seq reveals that glioblastoma recapitulates a normal neurodevelopmen-
tal hierarchy. Nat. Commun. 11, 3406. https://doi.org/10.1038/s41467-
020-17186-5.
103. Schaff, L.R., and Mellinghoff, I.K. (2023). Glioblastoma and Other Primary
Brain Malignancies in Adults: A Review. JAMA 329, 574–587. https://doi.
org/10.1001/jama.2023.0023.
104. Yu, G., Li, F., Qin, Y., Bo, X., Wu, Y., and Wang, S. (2010). GOSemS im: an
R package for measuring semantic similarity among GO terms and gene
products. Bioinforma. Oxf. Engl. 26, 976–978. https://doi.org/10.1093/
bioinformatics/btq064.
105. Boyle, E.I., Weng, S., Gollub, J., Jin, H., Botstein, D., Cherry, J.M., and
Sherlock, G. (2004). GO::TermFinder–open source software for access-
ing Gene Ontology information and finding significantly enriched Gene
Ontology terms associated with a list of genes. Bioinforma. Oxf. Engl.
20, 3710–3715. https://doi.org/10.1093/bioinformatics/bth456.
106. Wu, T., Hu, E., Xu, S., Chen, M., Guo, P., Dai, Z., Feng, T., Zhou, L., Tang,
W., Zhan, L., et al. (2021). clusterProfiler 4.0: A universal enrichment tool
for interpreting omics data. Innovation 2, 100141. https://doi.org/10.
1016/j.xinn.2021.100141.
107. Xu, S., Dai, Z., Guo, P., Fu, X., Liu, S., Zhou, L., Tang, W., Feng, T., Chen,
M., Zhan, L., et al. (2021). ggtreeExtra: Compact Visualization of Richly
Annotated Phylogenetic Data. Mol. Biol. Evol. 38, 4039–4042. https://
doi.org/10.1093/molbev/msab166.
108. Miller, J.A., Ding, S.-L., Sunkin, S.M., Smith, K.A., Ng, L., Szafer, A., Eb-
bert, A., Riley, Z.L., Royall, J.J., Aiona, K., et al. (2014). Transcriptional
landscape of the prenatal human brain. Nature 508, 199–206. https://
doi.org/10.1038/nature13185.
109. Collins, R.L., Glessner, J.T., Porcu, E., Lepamets, M., Brandon, R., Laur-
icella, C., Han, L., Morley, T., Niestroj, L.-M., Ulirsch, J., et al. (2022). A
cross-disorder dosage sensitivity map of the human genome. Cell 185,
3041–3055.e25. https://doi.org/10.1016/j.cell.2022.06.036.
110. Gandal, M.J., Nesbitt, A.M., McCurdy, R.M., and Alter, M.D. (2012).
Measuring the maturity of the fast-spiking interneuron transcriptional
program in autism, schizophrenia, and bipolar disorder. PLoS One 7,
e41215. https://doi.org/10.1371/journal.pone.0041215.
111. Hanamsagar, R., Alter, M.D., Block, C.S., Sullivan, H., Bolton, J.L., and
Bilbo, S.D. (2017). Generation of a microglial developmental index in
mice and in humans reveals a sex difference in maturation and immune
reactivity. Glia 65, 1504–1520. https://doi.org/10.1002/glia.23176.
16 Cell Reports Medicine 5, 101691, August 20, 2024
Article
ll
OPEN ACCESS

STAR+METHODS
KEY RESOURCES TABLE
RESOURCES AVAILABILITY
Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Jeffrey
Noebels (jnoebels@bcm.edu).
Materials availability
This study did not generate any new reagents.
Data and code availability
Data used in this study are listed in the key resource table and are detailed in the Methods Details section below. No original code was
generated for this study. All heatmaps were generated using ggplot2 and Tidyverse packages in R. Any additional information
required to reanalyze the data reported in this work is available from the lead contact upon request.
EXPERIMENTAL MODEL AND SUBJECT DETAILS
This study did not generate any new experimental models. Information about cases analyzed in this study are listed in references
detailed in the Methods Details section.
METHOD DETAILS
Cellular compartmentation and case complexity
The Ivy GAP
19
database was generated by microscopically sampling 7 resected tumor regions, including the neuronal leading edge
(LE) and adjacent zones of infiltrating tumor (IT), dense tumor cells (CT), perinecrotic zone (PNZ), pseudopalisading necrosis (PAN),
REAGENT or RESOURCE SOURCE IDENTIFIER
Deposited data
Ivy GAP Allen Brain Institute https://glioblastoma.alleninstitute.org/rnaseq/search/index.html
RNA TCGA cancer sample gene data,
Human Protein Atlas version 23.0,
Ensembl version 109
Human Protein Atlas https://www.proteinatlas.org/about/download
BrainSpan Allen Brain Institute https://www.brainspan.org/
RNA HPA brain gene data, The Human
Protein Atlas version 23.0 and Ensembl
version 109
Human Protein Atlas https://www.proteinatlas.org/about/download
RNA HPA PFC brain gene data,
The Human Protein Atlas version
23.0 and Ensembl version 109.
Human Protein Atlas https://www.proteinatlas.org/about/download
Software and algorithms
ggplot2 R https://cran.r-project.org/web/packages/ggplot2/index.html
Tidyverse R https://cran.r-project.org/web/packages/tidyverse/index.html
clusterProfiler Bioconductor https://bioconductor.org/packages/release/
bioc/html/clusterProfiler.html
GOSemSims Bioconductor https://bioconductor.org/packages/release/
bioc/html/GOSemSim.html
STAT R https://cran.r-project.org/web/packages/STAT/index.html
tidytree R https://cran.r-project.org/web/packages/tidytree/index.html
ggtreeExtra Bioconductor https://bioconductor.org/packages/release/
bioc/html/ggtreeExtra.html
Cell Reports Medicine 5, 101691, August 20, 2024 e1
Article
ll
OPEN ACCESS

hyperplastic blood vessel in the cellular tumor (HBV) and microvascular proliferation (MVP). Each region was identified histologically
and validated with cell-type specific RNA-Seq population markers.
Histological definitions of tumor regions
The leading edge (LE) samples were selected to contain zones containing only 1–3 tumor cells per visual field reflecting peritumoral
neural networks, while the IT infiltrating zone composition was 10–20 tumor cells per visual field, and the CT tumor zone contained an
average of 300/1 ratio of tumor cells to normal cells. The PAN and PNZ zones were defined by aggregation of nuclei around necrosis
with or without pseudopalisading cells respectively. And the HBV and MVP zones were distinguished by blood vessels with either
hypertrophic thickened walls or sharing endothelial walls respectively.
Molecular validation of some Epi358 genes
The expression of only 9 of the Epi358 genes in the Ivy database have been validated by ISH at the LE and despite the uniform mo-
lecular genetic procedures used to build this database and evidence of clear segmentation in 6 of these genes (Figure S5), extended
analysis could reveal anomalies in compartmental values. In future studies, single cell RNAseq analysis of spatial samples will provide
validation as well as the identity of neuronal and glial subtypes for each gene.
Further genetic inclusion and exclusion criteria
Per the revised definition of glioblastoma published by the World Health Organization in 2021,
103
we omitted samples derived from
patients with IDH1/2 mutation, resulting in the exclusion of four patients. To facilitate interpretation of regional tumor mechanisms, we
also excluded tumors which were derived from multiple lobes in the brain (n= 6 patients excluded), resulting in a total of 163 samples
from 29 patients which were each probed for 25,873 genes. A summary of the number of patients and samples collected for each
region of the tumor and full methodological details are available at the Allan Brain Atlas/Ivy GAP website: https://help.brain-map.
org/display/glioblastoma/Documentation.
Calculation of enrichment ratio per tumor region
The enrichment ratio is an intra-tumoral comparison which highlights tumor heterogeneity and can be used to identify genes enriched
per tumor region. Regional RNA-Seq was downloaded from Ivy GAP
19
database, and samples were segregated by tumor lobar loca-
tion. Per frontal, temporal, and parietal lobar locations, the enrichment ratio was calculated for six anatomically defined tumor regions
LE, IT, PNZ, PAN, HBV, and MVP as compared to expression levels at the CT. We defined an ER of R1.5-fold as enriched in the
region of interest (ROI) as compared to the CT, and an ER of %0.5 as enriched in the CT compared to the ROI.
ERregion of interest =
Avg RNAregion of interest
Avg RNA pure tumor cell region
In order to compare the enrichment ratios across cortical regions, we determined that the relative expression of the Epi358 genes
did not vary greatly amongst the frontal, temporal, and parietal lobes in healthy adult brain (Figure S6, lane 1). Comparison of the
tumoral enrichment ratios per lobe demonstrated that the majority of genes also showed a similar enrichment ratio regardless of
cortical regions (Figure S6, lanes 2–6). Due to the high congruence of enrichment ratios from tumors derived from different cortical
regions, we display the average enrichment ratio across the frontal, temporal, and parietal lobes per tumor region in Figure 1A.
To display the heterogeneity within the cellular tumor, we calculated the enrichment ratio of the average RNA of all samples per
patient for each region of the tumor as a function of the lobar-matched average expression of the cellular tumor.
ERper patient =
RNApatient X
Avg RNA Lobar matched CT
Calculation of case concordance
We used the coefficient of variation (stdev/average) as a measure of case concordance and defined a low concordance as a CV R
0.50, a high concordance as a CV %0.25, and a medium concordance as intermediate values between 0.25 and 0.50. Regional RNA-
Seq was downloaded from Ivy GAP database, and TCGA bulk RNA-Seq was downloaded from the Human Protein Atlas website. The
coefficient of variation was first determined per lobar location before averaging for the final coefficient of variation displayed in Fig-
ure 1.
Coefficient of Variation ðCV Þ=AvgðCVFL;CVTL ;CVPL Þ
Calculation of fold Change between cases with and without seizures (FC
sz
):
We stratified the Ivy GAP RNA-Seq according to patient history, forming two groups, cases with seizures and cases without sei-
zures. We then determined the fold change between these two groups per tumoral region.
FCsz =
Avg RNAseizure cases
Avg RNA non seizure cases
e2 Cell Reports Medicine 5, 101691, August 20, 2024
Article
ll
OPEN ACCESS

Hierarchal clustering of the Epi358 genes into functional groups based on biological process
To hierarchically cluster the Epi358 genes based on their biological processes listed in the Gene Ontology Database, we used the
godata function of GOSemSims R Package to generate a semantics similarity database for the Biological Processes (BP GO) of hu-
man genes listed in the Gene Ontology Database. The mgeneSim function from the GOSemSim package was then used to calculate
the pairwise semantic similarities of the Epi358 genes based on their BP GO semantic annotations using the Wang method.
104
The
hclust function from the STATS package in R was then used to hierarchically cluster the Epi358 genes based on BP GO semantics
similarity matrix scores using the ward.D method. Hierarchical clustering was successful for 349/358 Epi358 genes, resulting in 3
major clusters. We conducted a GO over-representation analysis
105
of the BP for each of the clusters using the clusterProfiler pack-
age
106
to assign the major categories displayed in Figure 1: (1) Cell Growth and Division; (2) Ion Channel and Transport; (3) Macro-
molecule Biosynthesis. The hierarchal cluster was then converted into a phylo object used for phylogenetic analysis in R using the
as.phylo function in the tidytree package.
107
The phylogenic object based on the BP GO of the Epi358 genes were then visualized
using ggtreeExtra package
107
in R as a fan tree diagram with an open angle of 30. For the circular dendrogram in Figure 1, we an-
notated the circular dendrogram with the ER and CV per tumor region using the geom_fruit, and geom_tile function in ggtreeExtra.
Calculation of fold change of epilepsy genes at the peritumoral LE vs. healthy cortex
To compare LE expression to healthy human brain, we downloaded region matched healthy human cortical RNA-Seq datasets from
the Allen Brain Institute Brain Span Project
108
and the Human Protein Atlas
25
Brain and Pre-frontal Brain Gene Data. To compare LE
and cortical tissue RNA-Seq across different datasets, we normalized all genes by five genes which showed invariant expression
across multiple cortical regions in both the Allen Brain Atlas and Human Protein Atlas healthy cortical datasets (RNF10, M6PR,
AIFM1, APEX1, CNDP2). We determined the expression fold change for each lobar location of the derived LE (i.e., FL, TL, PL), as
well as the pooled cortical sample (Cx), which included gene expression at the frontal lobe, temporal lobe, and parietal lobe.
FCCx =
Avg RNALE per FL;TL;or PL
Avg RNA Healthy Fl;TL;or PL
To determine the congruence across the ER and FC across the FL, TL, and PL, we categorized the data roughly into those with ERs
or FCs >1.5, <0.5, or between those two threshold values, assigning them as enriched, under-enriched, no change for ERs and then
upregulated, downregulated, no change for FCs. The categorized variables were then sorted for genes which showed identical cat-
egorical ERs and/or FCs, forming 15 groups displayed in Figure 2.
Identification of pathogenic copy number variants that are epilepsy linked
A prior study conducted a meta-analysis of rare copy number variants on the 22 autosomes, cumulating data from 17 different sour-
ces.
109
We extracted data from this resource and determined which genes located on these rare copy number variants were clinically
validated as epilepsy linked, resulting in a panel of 41 genes. We then determined the ER and FC
Cx
for these epilepsy-linked genes
located within pathogenic copy number variants.
Curation of gene set of disease-linked genes and presynaptic genes
Relevant epilepsy linked genes were compiled from a diagnostic panel of 320 genes (https://www.invitae.com/en/providers/test-
catalog/test-03401) and recent literature, resulting in 358 epilepsy/dysplasia-linked genes (Table S1). Dysplasia/cortical malforma-
tion genes were identified from a panel of genes linked to malformations of cortical development.
79
Presynaptic genes were curated
from the literature for genes involved in the cytoskeletal matrix assembled at the assembly zone (CAZ), SNARE family proteins
including synaptotagmins (SYT) and synaptobrevins (VAMP), synaptic vesicle proteins such as clathrin and syntaxins, and adaptor
proteins which may regulate synaptic vesicle composition (AP complexes). In addition, murine gene ontology annotations for synap-
tic exocytosis processes (GO:0140029) was included for further consideration if involved in priming, docking, fusion or targeting of
synaptic vesicles.
Comparison of FC
Cx
at the LE to expression levels during healthy early development
Human RNA-Seq at the frontal, temporal, and parietal lobes for different developmental stages was downloaded from the Allen Brain
Institute BrainSpan Project (https://www.brainspan.org/).
108
To make comparisons across the BrainSpan and HPA RNA-Seq data-
sets, we first normalized the BrainSpan data to adult cortical levels within the BrainSpan dataset (Adult.A). Next, we normalized
Adult.A levels from BrainSpan and Adult.B expression levels from HPA using five genes which showed invariant expression across
multiple cortical regions in both the Allen Brain Atlas and Human Protein Atlas healthy cortical datasets (RNF10, M6PR, AIFM1,
APEX1, CNDP2). A ratio was formed using normalized Adult.B/Adult.A levels per cortical lobe. We then determined the FC
Cx
compared to healthy adult cortex for each early developmental stage per FL, TL, and PL using the equation for FC
cx
below. Due
to the high congruence of expression of Epi358 genes across the FL, TL, and PL, we displayed the average FC
Cx
for these three lobes
in Figure 5.
FCDev:norm =
Avg RNADev Stage per FL;TL;or PL
Avg RNA Adult:A per FL;TL;or PL
Cell Reports Medicine 5, 101691, August 20, 2024 e3
Article
ll
OPEN ACCESS

FCAdult:norm:ratio =
Avg RNAAdult:B per FL;TL;or PL
Avg RNA Adult:A per FL;TL;or PL
FCCx =
FCDev:norm per FL;TL;or PL
FC Adult:norm:ratio per FL;TL;or PL
Overrepresentation analysis of development related functions
The Epi358 genes were stratified into genes which were upregulated and downregulated per LE and early developmental stage,
forming 10 expression groups. We then conducted a Gene Ontology over-representation analysis
105
of the biological process for
each of the expression groups using the clusterProfiler package.
106
A significance level of less than 0.05 for the adjusted p-value
was used to filter for significant results. We then prioritized developmentally relevant Gene Ontology terms involved in growth
signaling, migration, and pre- and post-synaptic functions.
Prioritized GO terms:
Growth signaling: GO:0032007.
Migration: GO:0001764.
Presynaptic Function: GO:0042391, GO:0007416, GO: 0099003.
Postsynaptic Function: GO: 0099565, GO: 0060079.
Determination of cortical developmental index for fold changes compared to WT
The Cortical Developmental Index (CDI) is a modified strategy for comparing developmental stages which was previously tested in
fast-spiking interneurons
110
and microglia.
111
The CDI normalizes expression across different stages so that the condition with a zero
CDI score represents the minimum value of all the tested conditions and a score of 1 equals the maximum CDI value of all the con-
ditions tested. To facilitate comparison across Ivy GAP LE and BrainSpan datasets, we normalized FL expression by genes which
showed invariant expression across all developmental stages at the FL (ARFGAP2, BSDC1, DMAP1). The CDI was then determined
for the normalized expression at LE samples derived from the FL and normalized FL expression in healthy cortex from various devel-
opmental stages.
CDI =
FCxstage FCxminimum across all stages
FCxmaximum stage FCxminimum across all stages
QUANTIFICATION AND STATISTICAL ANALYSIS
An unpaired two-sample t-test and Welch’s t-test was used to determine statistical significance of intratumoral enrichment ratio for
samples with and without equal variances respectively. A threshold p-value of 0.05 was used to determine statistical significance.
e4 Cell Reports Medicine 5, 101691, August 20, 2024
Article
ll
OPEN ACCESS