Neuro-Oncology Advances 1

4(1), 1–10, 2022 | https://doi.org/10.1093/noajnl/vdac096 | Advance Access date 21 June 2022

Molecular alterations associated with improved

outcome in patients with glioblastoma treated with
Tumor-Treating Fields
  

Manjari Pandey, Joanne Xiu, Sandeep Mittal, Jia Zeng, Michelle Saul, Santosh Kesari , Amir Azadi,

Herbert Newton, Karina Deniz, Katherine Ladner, Ashley Sumrall, W. Michael Korn, and Emil Lou

Downloaded from https://academic.oup.com/noa/article/4/1/vdac096/6612779 by guest on 26 December 2022

West Cancer Center and Research Institute, Memphis, Tennessee, USA (M.P.); Caris Life Sciences, Phoenix, Arizona,
USA (J.X., J.Z., M.S., W.M.K.); Virginia Tech Carilion School of Medicine, Roanoke, Virginia, USA (S.M.); Pacific
Neuroscience Institute, Saint John’s Cancer Institute, Santa Monica, California, USA (S.K.); Arizona Oncology
Biltmore, Phoenix, Arizona, USA (A.A); Neuro-Oncology Center, Advent Health Cancer Institute, Orlando, Florida,
USA (H.N.); Division of Hematology, Oncology and Transplantation, Masonic Cancer Center, University of Minnesota,
Minneapolis, Minnesota, USA (K.D., K.L., E.L.); Levine Cancer Institute, Charlotte, North Carolina, USA (A.S.)

Corresponding Author: Emil Lou, MD, PhD, FACP, Associate Professor of Medicine, Division of Hematology, Oncology and Transplantation,
University of Minnesota, Mayo Mail Code 480, 420 Delaware Street SE, Minneapolis, MN 55455, USA (emil-lou@umn.edu).

Abstract
Background. The genomic and overall biologic landscape of glioblastoma (GB) has become clearer over the past 2
decades, as predictive and prognostic biomarkers of both de novo and transformed forms of GB have been iden-
tified. The oral chemotherapeutic agent temozolomide (TMZ) has been integral to standard-of-care treatment for
nearly 2 decades. More recently, the use of non-pharmacologic interventions, such as application of alternating
electric fields, called Tumor-Treating Fields (TTFields), has emerged as a complementary treatment option that in-
creases overall survival (OS) in patients with newly diagnosed GB. The genomic factors associated with improved
or lack of response to TTFields are unknown.
Methods. We performed comprehensive genomic analysis of GB tumors resected from 55 patients who went on
to receive treatment using TTFields, and compared results to 57 patients who received standard treatment without
TTFields.
Results. We found that molecular driver alterations in NF1, and wild-type PIK3CA and epidermal growth factor
receptor (EGFR), were associated with increased benefit from TTFields as measured by progression-free survival
(PFS) and OS. There were no differences when stratified by TP53 status. When NF1, PIK3CA, and EGFR status were
combined as a Molecular Survival Score, the combination of the 3 factors significantly correlated with improved
OS and PFS in TTFields-treated patients compared to patients not treated with TTFields.
Conclusions. These results shed light on potential driver and passenger mutations in GB that can be validated as
predictive biomarkers of response to TTFields treatment, and provide an objective and testable genomic-based
approach to assessing response.

Key Points
• Alterations in NF1 were associated with increased benefit from TTFields.
• Wild-type PIK3CA and EGFR also aligned with increased benefit from this approach.
• These results provide insight into molecular differences that can be validated to tailor
treatment.

© The Author(s) 2022. Published by Oxford University Press, the Society for Neuro-Oncology and the European Association of Neuro-Oncology.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/),
which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.
2 Pandey et al. Genomic factors affecting Tumor-Treating Fields

Importance of the Study

The application of Tumor-Treating Fields increased benefit from TTFields. This signa-
(TTFields) is a part of the standard-of-care ture opens the door to a personalized treat-
approach to treating patients with glioblas- ment approach for patients with GB. The
toma (GB). To date, the genomic factors as- value of this study is that it provides insight
sociated with improved or lack of response into the role of comprehensive genomic pro-
to TTFields have not been identified. In this filing in uncovering potential predictive and
study, we provide the first identification prognostic biomarkers associated with re-
of a molecular signature associated with sponse to TTFields.

Downloaded from https://academic.oup.com/noa/article/4/1/vdac096/6612779 by guest on 26 December 2022

Glioblastoma (GB) is the most common primary malig- prognosis of patients with GB. However, despite knowl-
nant brain tumor, accounting for approximately 50% of edge of this and several other driving mutations in pri-
brain cancers.1 Most patients die within 1-2  years of diag- mary central nervous system malignancies, published
nosis with a median progression-free survival (PFS) from data on genomic correlation to response to TTFields in
diagnosis of 6.2-7.5  months and median overall survival GB have been limited. For this study, we postulated that
(OS) of 14.6-16.7 months.2 Estimated 2- and 5-year survival next-generation sequencing (NGS) of treatment-naïve
rates of 18.5% and 6.8%, respectively.1 For over a decade, GB would produce a molecular signature indicating
the standard strategy for combination therapy consisted TTFields efficacy in the first-line setting, prior to first pro-
of maximal safe surgical resection, followed by concur- gression. Thus, we attempted to identify whether there
rent radiotherapy with daily temozolomide (TMZ) chemo- is a molecular subset of GB with differential response to
therapy, followed by maintenance treatment with TMZ for TTFields treatment. Here, we performed retrospective
6-12 months.3 evaluation of a large set of resected GB tumor samples
Over the past decade, Tumor-Treating Fields (TTFields) and performed deep sequencing to uncover differences
have emerged as a complementary treatment strategy in molecular alterations in patients whose treatment had
for newly diagnosed and recurrent cases of GB following incorporated TTFields in the first-line setting.
trials that demonstrated clinical activity. In the first-line
setting specifically, the use of TTFields was associated
with significantly higher OS compared to standard-of-care
combination therapy alone.4 TTFields comprise a form of
Methods
low-intensity alternating electric fields with intermediate Demographics and Clinical Data Collection
frequency that interfere with and prolong cell division, re-
sulting in apoptosis. Optimal electrical frequency for the This study was a retrospective, multi-institutional evalua-
most effective cell kill varies by tumor type. For GB, an tion of patients with GB treated with TTFields in the first-
intensity of 1-3 V/cm and frequency of 200  kHz were in- line setting. Data were collated from genomic profiles
vestigated and ultimately established as the standard set following biopsy or surgical resection of GB tumor speci-
of parameters for use in patients with this tumor type. mens from 6 institutions (Barrow Neurological Institute,
Analysis of the EF-14 trial showed that TTFields + TMZ was Arizona; Levine Cancer Institute, Charlotte, North Carolina;
associated with improved PFS and OS in all subgroups West Cancer Center, Memphis, Tennessee; John Wayne
regardless of age, sex, Karnofsky Performance Status Cancer Center, San Diego, California; Karmanos Cancer
(KPS), methylation status of O6-methylguanine-DNA Institute, Detroit, Michigan; Florida Hospital, Florida) be-
methyltransferase (MGMT), geographic region, or extent tween December 2014 and November 2017. Patients with
of upfront surgical resection of the tumor.4 high-grade gliomas who had undergone molecular pro-
Some preclinical studies have suggested that TTFields filing at Caris Life Sciences were identified and their med-
exposure induces apoptosis by both p53-dependent and ical records were reviewed at each participating site from
-independent mechanisms.5,6 What remains unknown is which their clinicopathological features, the treatment,
what genomic factors within tumors affect response to and outcome information were extracted, de-identified,
TTFields at the molecular as well as the cellular levels. and submitted for central analysis. For final analysis, we
To date, there is no validated predictive biomarker of only included patients with GB, WHO grade 4, and ex-
response to TTFields therapy other than compliance. cluded patients who had any treatment initiated prior to
Testable biological markers that are predictive of ef- tumor profiling. Molecular profiling was performed by
ficacy of TTFields in vivo have not yet been identified. Caris Life Sciences, as described in detail below. At each
The emergence of genomic markers predictive of treat- participating institution, clinical records of patients who
ment response to cancer-directed treatments—whether received TTFields treatment as part of their treatment plan
they take the form of cytotoxic chemotherapies, biologic were reviewed and pre-specified data points were re-
targeted agents, or immunotherapeutic agents—has corded by study co-investigators; a control cohort of sim-
formed a new landscape for the field of precision on- ilar size as the TTFields-treated cohort was also reviewed
cology. Methylation of the MGMT promoter is a well-es- and recorded. The manufacturer of the device, Novocure,
tablished predictor of response to TMZ and thus had no role in the identification of patients, compliance,
 Pandey et al. Genomic factors affecting Tumor-Treating Fields 3

Advances
Neuro-Oncology
data analysis, or any other part of this research study. No positively stained cells was >5%. The antibody for PD-1
funding was acquired for this study. was NAT105 (Cell Marque) and >=1 TIL/HPF was considered
positive.

Molecular Profiling
Statistical Analysis
All tumor samples were tested with comprehensive
molecular profiling which included NGS on DNA and Patient PFS was calculated from the date of patients’
RNA as well as MGMT promoter methylation testing by glioma histological diagnosis to the first progression after
pyrosequencing. NGS was performed on genomic DNA TTFields treatment start in the experimental arm; and to
isolated from formalin-fixed paraffin-embedded (FFPE) the first progression after first-line treatment start in the
tumor samples using the NextSeq platform (Illumina, control arm. OS was calculated from the patients’ histo-
Inc., San Diego, CA). A  custom-designed SureSelect XT logical diagnosis till patient death or last date of contact.

Downloaded from https://academic.oup.com/noa/article/4/1/vdac096/6612779 by guest on 26 December 2022

assay was used to enrich 592 whole-gene targets (Agilent Kaplan-Meier estimates of the PFS and OS were performed
Technologies, Santa Clara, CA). All variants were detected on censored data using Cox proportional hazards (PH)
with >99% confidence based on allele frequency and model. Hazard ratio and P values were calculated for inter-
amplicon coverage, with an average sequencing depth of group comparisons and P < .05 was considered significant.
coverage of >500 and an analytic sensitivity of 5%. Prior Biomarker and clinicopathological features in the TTFields
to molecular testing, tumor enrichment was achieved by and control group were compared using Fisher’s exact test.
harvesting targeted tissue using manual microdissection
techniques. Genetic variants identified were interpreted
by board-certified molecular geneticists and categorized Ethics
as “pathogenic,” “likely pathogenic,” “variant of unknown
This study was conducted in accordance with guidelines of
significance,” “likely benign,” or “benign,” according to
the Declaration of Helsinki, Belmont report, Good Clinical
the American College of Medical Genetics and Genomics
Practice, REMARK, and U.S. Common Rule. In keeping
(ACMG) standards. When assessing mutation frequencies
compliance with policy 45 CFR 46.101(b)(4), the part of this
of individual genes, “pathogenic” and “likely pathogenic”
study utilizing the Caris dataset was performed using ret-
were counted as mutations while “benign”, “likely benign”
rospective, de-identified clinical data. Therefore, this part
variants, and “variants of unknown significance” were
was considered Institutional Review Board (IRB) exempt
excluded.
and no patient consent was necessary.
For gene fusion detection, anchored multiplex PCR
was performed for targeted RNA sequencing using the
ArcherDx fusion assay (Archer FusionPlex Solid Tumor
panel). The FFPE tumor samples were microdissected to
Results
enrich the sample to ≥20% tumor nuclei, and mRNA was
isolated and reverse transcribed into complementary DNA Patient Characteristics and Survival
(cDNA). Unidirectional gene-specific primers were used
to enrich for target regions, followed by NGS (Illumina Data were collected from a total of 148 patients treated at 6
MiSeq platform). Targets included 52 genes, and the full participating institutions; patients with grade III tumor at di-
list can be found at http://archerdx.com/fusionplex-assays/ agnosis, or with GB treated in the recurrent setting (a total
solid-tumor. of 36 patients) were excluded. Fifty-five patients treated
MGMT promoter methylation was evaluated by with TTFields, and 57 treated with standard-of-care treat-
pyrosequencing. DNA extraction from paraffin- ment without TTFields, were included for final analysis.
embedded tumor samples was performed for subse- Demographic characteristics were well balanced in the 2
quent pyrosequencer-based analysis of 5 CpG sites (CpGs cohorts (Table 1). Treatment regimens for both the TTFields-
74-78). All DNA samples underwent a bisulfite treatment treated and control group patients were dominated by the
and were PCR amplified with primers specific for exon 1 use of standard-of-care concurrent chemoradiation using
of MGMT (GRCh37/hgl9 − chr10: 131 265 448-131 265 560). daily TMZ chemotherapy, followed by 5-day TMZ used in
Methylation status of PCR-amplified products is deter- 28-day cycles for 6-12  months; some patients also or in-
mined using the PyroMark system. Samples with ≥7% and stead received different chemotherapeutic or biologic
<9% methylation are considered to be equivocal or gray agents on or off clinical trials (Table 2). All patients not re-
zone results. ceiving TTFields were included in the control cohort for the
Immunohistochemistry (IHC) was performed on full FFPE purpose of this analysis. In patients treated with TTFields,
sections of glass slides. Slides were stained using auto- the average duration of use of TTFields was 198 days (IQR:
mated staining techniques according to the manufacturer’s 52-149 days); average compliance of use of the device was
instructions and were optimized and validated per CLIA/ 57%, with median use of 60%.
CAO and ISO requirements. The primary antibody used In TTFields-treated patients, PFS was 15.8  months as
against PD-L1 was SP142 (Spring Biosciences). The staining compared to 6.9  months in control patients (HR = 0.55;
was regarded as positive if its intensity on the membrane 95% CI: 0.35-0.86; P  =  .01); OS was 25.5  months in
of the tumor cells was >=2+ (on a semiquantitative scale of TTFields-treated patients vs 18.8  months in the control
0-3: 0 for no staining, 1+ for weak staining, 2+ for moderate group (HR = 0.54; 95% CI: 0.31-0.94; P = .03) (Figure 1). As
staining, or 3+ for strong staining) and the percentage of expected, among TTFields-treated patients, those with
4 Pandey et al. Genomic factors affecting Tumor-Treating Fields

  
Table 1.  Baseline Patient Characteristics

TTFields-treated Patients, N (%) Control Patients, N (%)

Patient, N 55 57
Gender
 Female 17 (31%) 23 (40%)
 Male 38 (69%) 34 (60%)
Age
  Median age 59 58
  Age range 26-79 17-75

Downloaded from https://academic.oup.com/noa/article/4/1/vdac096/6612779 by guest on 26 December 2022

Histology All glioblastoma All glioblastoma
Tumor grade All IV All IV
Primary tumor location
  Temporal lobe 16 (29%) 10 (18%)
  Frontal lobe 14 (25%) 17 (30%)
  Parietal lobe 9 (16%) 12 (21%)
  Brain, NOS 12 (22%) 17 (30%)
 Cerebellum 1 (2%) 0
  Occipital lobe 2 (4%) 0
 Thalamus 1 (2%) 1 (2%)
IDH1/2 mutation % (N) 5 (9%) 3 (5%)
MGMT methylation % (N) 24/53 (45%) 26/56 (46%)

Abbreviations: IDH, isocitrate dehydrogenase; MGMT, O6-methylguanine-DNA methyltransferase; NOS, not otherwise specified; TTFields, Tumor-
Treating Fields
  
>50% use of the device per 24-hour period had better PFS Activating mutations in the PIK3CA gene regulate the
(12.6 vs 6.9  months, P = .0274) and also OS (not reached phosphatidylinositol 3-kinase signaling cascade in many
vs 18.8 months, P = .0041) than patients in the control co- cancer types, including in infiltrative gliomas; these mu-
hort (Supplementary Figure 1). Corresponding data for tations have been associated with more disseminated
TTFields-treated patients using the device at the usage rate versions of the disease at the time of diagnosis, as well as
of >75% are shown in Supplementary Figure 2. earlier recurrence.7 Specifically, our data suggested that
mutation in the PIK3CA gene predicts a poor response
of the tumor in general, possibly associated with de-
In Search of Genomic Biomarkers Predictive of creased or lack of response to TTFields: among TTFields-
Survival in Patients Receiving TTFields treated patients, those carrying PIK3CA mutations had
a significantly shorter PFS (6.7 vs 16.8  months, Cox PH
A wide distribution of molecular alterations was detected P = .0008) and OS (10.0 vs 26.6 months; Cox PH P = .0158)
using IHC for detection of PD-1 and PD-L1, NGS (592 compared to those that were wild type for PIK3CA mu-
genes), and pyrosequencing for detection of MGMT pro- tation. This difference was not seen in the control group
moter methylation (Figure 2). The most commonly de- (11.2 vs 6.0  months PFS, Cox PH P = .6541; OS 19.2 vs
tected molecular characteristics were PD-1 expression, 19.5  months; P = .6695). As is shown in Figure 3, for the
CDKN2A mutation, MGMT promoter methylation, epi- first 0-20 months, there was a notable improvement in PFS
dermal growth factor receptor (EGFR) amplification, and for patients with wild-type PIK3CA treated with TTFields
TP53 mutation. However, for these alterations, there were compared to control treatment; however, by 30 months,
no significant differences between the TTFields vs control there was no significant difference in any of the 4 groups
group. MGMT promoter methylation was detected in 45% (Figure 3; Supplementary Figure 3A; Supplementary Table
(24/53) of patients in the TTFields group and in 46% (25/56) 1). A similar comparison was seen for OS in comparison
in the control cases. of these groups (Figure 3; Supplementary Figure 3B;
Molecular alterations were surveyed for individual as- Supplementary Table 1).
sociation with PFS and OS in TTFields-treated and control Neurofibromatosis type 1 (NF1) is a tumor suppressor
patients. We found that TP53 mutations did not have any that encodes neurofibromin, a protein prevalent in neurons
effect on outcome upon TTFields therapy (PFS: HR = 1.11, and astrocytes. Alterations in NF1 predispose to increased
P = .7685; OS: HR = 1.14, P = .8000). However, alterations risk of central nervous system tumors, of which gliomas
in PIK3CA, NF1, or EGFR, in particular, did demonstrate a are the most common subtype in this patient population.
trend toward different effects in TTFields-treated tumors NF1 mutations are prevalent in both de novo and trans-
(Figure 3; Supplementary Figure 3; SupplementaryTable 1). formed GB, in the form of both mutations and deletions.8
 Pandey et al. Genomic factors affecting Tumor-Treating Fields 5

Advances
Neuro-Oncology
  
Table 2.  Treatments Administered for the TTFields and Control Groups

TTFields-treated arm (n = 55) Treatment administered prior to TTFields therapy

  Radiation/temozolomide combination 54
 Bevacizumab 5
  Vitamin C 2
 CPT-11 2
 Nivolumab 1
Concurrent treatment with TTFields
 Temozolomide 41

Downloaded from https://academic.oup.com/noa/article/4/1/vdac096/6612779 by guest on 26 December 2022

 Bevacizumab 13
 Carboplatin 4
 Nivolumab 2
 CPT-11 1
 None 3
Duration of Optune use
 Average 198 days
 IQR 52-249 days
  Compliance % 57% (11%-95%)
Control arm (n = 57) Temozolomide 56
Bevacizumab 23
Carboplatin 17
Nivolumab/pembrolizumab 10
CPT-11 4
CCNU 5

Abbreviations: CCNU, CCNU, 1-(2-Chloroethyl)-3-cyclohexyl-1-nitrosourea (Lomustine); IQR, interquartile range; TTFields, Tumor-Treating Fields
  

  
Progression free survival Overall survival
Control
1.0 TTFields 1.0 Control
TTFields

0.8 0.8
Survival probability

Survival probability

0.6 0.6

0.4 0.4

0.2 0.2

0.0 0.0
0 10 20 30 40 50 0 10 20 30 40 50
Time (months)
Time (months)

p value Survival HR Lower 95% Upper 95% TTFields n Control n TTFields censored Control censored
PFS 0.01 15.8 m vs. 6.9 m 0.55 0.35 0.86 55 49 20 5
OS 0.03 25.5 m vs. 18.8 m 0.54 0.31 0.94 55 57 37 14

Figure 1.  Progression-free and overall survivals of patient cohorts receiving TTFields vs control cohorts. Abbreviation: TTFields, Tumor-Treating
Fields
  

Here, we found that in the TTFields-treated group, NF1 18.2 vs 14.4  months; P = .07 and OS NR vs 24.7  months;
alterations were associated with a response to TTFields P = .0415). In the control group, no difference was observed
therapy, as compared to tumors with wild-type NF1(PFS in patients whose tumors harbored the NF1 alterations vs
6 Pandey et al. Genomic factors affecting Tumor-Treating Fields

  
Molecular alteration prevalence 70%
60%
Control % Optune %
50%
40%
30%
20%
10%
0%
-1

2A

FR

53

EN

F1

FR

K4

B1

II

A
42

tu
LX

vI

R
M

C
D

TP

N
KN

K3
EG

EG

ta

FR
PT

K3
G

P1
.P

l.T

S.

S.
_S
M

S.

PI
C

PI
S.

(S

p.

EG
V.

S.
G
de

G
S.
IH

l.C

F1

S.
Am

S.
G

N
N

G
1
G

V.
N

G
C

-L

G
de

N
N

Downloaded from https://academic.oup.com/noa/article/4/1/vdac096/6612779 by guest on 26 December 2022

N

V.

N
D
_
V.

N
ro

.P
N

C
Py

C
C

IH
Control TTFields Treated

Biomarker Positive Negative Total Percent Positive Negative Total Percent

IHC.PD-1 32 20 52 62% 12 7 19 63%
CNV.del.CDKN2A 27 20 47 57% 20 34 54 37%
MGMTp methylation 26 30 56 46% 24 29 53 45%
CNV.EGFR 23 26 49 47% 21 33 54 39%
NGS.TP53 21 33 54 39% 19 36 55 35%
NGS.PTEN 19 34 53 36% 13 39 52 25%
CNV.del.TLX1 10 37 47 21% 8 46 54 15%
IHC.PD-L1 12 45 57 21% 7 48 55 13%
NGS.NF1 7 43 50 14% 9 41 50 18%
NGS.EGFR 9 46 55 16% 8 47 55 15%
CNV.Amp.CDK4 5 43 48 10% 10 44 54 19%
NGS.RB1 6 46 52 12% 6 45 51 12%
EGFRvIII 5 19 24 21% 1 21 22 5%
NGS.PIK3R1 5 46 51 10% 6 47 53 11%
NGS.PIK3CA 7 48 55 13% 4 51 55 7%

Figure 2.  Distribution of biomarker alterations detected in GB treated with TTFields vs control treatment. Abbreviations: GB, glioblastoma;
TTFields, Tumor-Treating Fields
  

wild-type NF1 (PFS 6.9 vs 6.9  months, P = .8933 and OS model (Supplementary Figure 4). Cox PH model was used
19.3 vs 19.2 months; P = .8902) (Figure 3). for survival regression. Linear models were built on the 4
EGFRvIII is a truncated form of the EGFR seen in over 50% variables, while interaction terms were applied for 2 vari-
of cases of GB; other less common variants of EGFR have ables (therapy and one biomarker). The models were as-
also been reported, including amplifications and gene fu- sessed and selected based on their log-likelihood on the
sions, in addition to the vIII form.9 EGFR wild-type tumors fitted data, and the concordance index, which is a general-
showed a trend for higher PFS with TTFields therapy 17.2 ization of AUC. The models with higher log-likelihood and
vs 12.6  months for tumors that harbored any alterations concordance index were considered to have better pre-
even though this finding was not statistically significant dictive power and fit. Based on the model’s log-likelihood
(HR = 0.517, P = .3628). When comparing patients with and concordance index, the one with all 4 variables had
EGFR alterations receiving TTFields vs no TTFields, EGFR- the best predictive power and fit (Supplementary Figure
intact patients had a PFS 4.6  months longer than EGFR- 5). Therefore, we further built a Molecular Survival
altered patients (P = .36) while the PFS was practically Score (MSS) based on the combination of PIK3CA, NF1,
identical in the control patients (Figure 3; Supplementary and EGFR.
Figure 3; Supplementary Table 1). We also evaluated TP53 The combined MSS for each patient was calculated as
alterations and found no significant differences in either follows: score of +1 was assigned for unaltered PIK3CA
PFS or OS in patients treated with TTFields in this subset. (wild type), EGFR (intact, wild-type with no fusion), and
Other markers considered important in cell cycle check- NF1 alteration, respectively; the reverse-scored as 0 for
point control, including CDKN2A, TP53, RB1, and CDK2 each factor. A  sum of the scores assigned to the 3 bio-
amplification, were not found to be associated with benefit markers was considered the Survival Score of each
from TTFields treatment. patient, with a final range of 0-3. Score of 0-2 was con-
Based on the above findings, we investigated the corre- sidered Score Low, and score of 3 was considered Score
lation and interdependency of TTFields treatment and the High. Analysis of OS showed that in patients with a high
statuses of PIK3CA (mutations), NF1 (mutation and copy MSS, the OS was not reached (95% CI: 11.6-23.6 months)
number alterations), and EGFR (EGFRvIII, fusions, and for those treated with TTFields and 19.2 months (95% CI:
amplifications) biomarkers using Spearman’s rank correla- 7.6-35.4  months) in those not treated with TTFields. For
tion. The correlation coefficient between TTFields, PIK3CA, patients with Score Low, OS was 25.5  months (95% CI:
NF1, and EGFR was close to 0, suggesting that these vari- 16.8-26.6  months) for those treated with TTFields and
ables were not dependent and can be used in the same 15.6  months (95% CI: 11.6-23.6  months) for those not
model and their interactions could be examined in the treated with TTFields. There is a statistically significant
 Pandey et al. Genomic factors affecting Tumor-Treating Fields 7

Advances
Neuro-Oncology
  
A Progression free survival PIK3CA B
Overall survival PIK3CA
1.0 Control PIK3CA WT
1.0 Control PIK3CA WT
Control PIK3CA MT
TTFields PIK3CA WT Control PIK3CA MT
0.8 TTFields PIK3CA WT
TTFields PIK3CA MT 0.8
TTFields PIK3CA MT
Survival probability

Survival probability
0.6
0.6

0.4 0.4

Downloaded from https://academic.oup.com/noa/article/4/1/vdac096/6612779 by guest on 26 December 2022

0.2 0.2

0.0 0.0
0 10 20 30 40 50 0 10 20 30 40 50
Time (months) Time (months)

Progression free survival NF1 Overall survival NF1

1.0 Control NF1 WT 1.0

Control NF1 MT

0.8 TTFields NF1 WT 0.8

TTFields NF1 MT
Survival probability
Survival probability

0.6 0.6

0.4 0.4

Control NF1 WT
0.2 0.2 Control NF1 MT
TTFields NF1 WT
TTFields NF1 MT
0.0 0.0
0 10 20 30 40 50 0 10 20 30 40 50
Time (months) Time (months)

Progression free survival EGFR Overall survival EGFR

1.0 Control EGFR WT 1.0 Control EGFR WT
Control EGFR MT Control EGFR MT
TTFields EGFR WT TTFields EGFR WT
0.8 0.8
TTFields EGFR MT TTFields EGFR MT
Survival probability
Survival probability

0.6 0.6

0.4 0.4

0.2 0.2

0.0 0.0
0 10 20 30 40 50 0 10 20 30 40 50
Time (months) Time (months)

Figure 3.  PFS (A) and OS (B) of patients treated with TTFields vs control stratified for alterations of PIK3CA (top panels), NF1 (middle panels), and
EGFR alterations (lower panels). Abbreviations: EGFR, epidermal growth factor receptor; NF1, neurofibromatosis type 1; OS, overall survival; PFS,
progression-free survival; TTFields, Tumor-Treating Fields
  

difference in PFS (P = .019) and OS (P = .0252) comparing 24-hour period), then the score remained predictive of
MSS-high vs low patients when treated with TTFields PFS (P = .034) in TTFields-treated patients with high vs
while the effect is not seen in control arms; the interaction low scores in this scenario but not for OS (Supplementary
P values were trending for both PFS and OS (Figure 4). Figure 6). Among TTFields-treated patients, those with tu-
When further examining differences in OS and PFS strati- mors with MSS of 3 tend to have higher PFS and OS than
fied by compliance (minimum use >50% within an average patients in the control cohort.
8 Pandey et al. Genomic factors affecting Tumor-Treating Fields

Progression free survival score in TTFields

  
Progression free survival score in control Overall survival score in TTFields
1.0 TTFields Score high 1.0 1.0
Control score high
TTFields Score low Control score low
0.8 0.8 0.8

Survival probability
Survival probability

Survival probability
0.6 0.6 0.6

0.4 0.4 0.4

0.2 0.2 0.2

TTFields Score high
TTFields Score low
0.0 0.0 0.0
0 10 20 30 40 0 10 20 30 40 50 0 10 20 30 40
Time (months) Time (months) Time (months)

Downloaded from https://academic.oup.com/noa/article/4/1/vdac096/6612779 by guest on 26 December 2022

Overall survival score in control
1.0 Control score high
Control score low Progression free survival Overall survival
0.8 TTFields treated Control TTFields treated Control
Score high 25.9m 7.0m NR 19.2m
Survival probability

Interaction Interaction
0.6 Score low 12.9m 4.5m p-value (score 24.5 15.5m p-value (score
P values (Cox PH) 0.0190 0.6485 and therapy) 0.0252 0.9847 and therapy)
0.1485 0.1114
HR 0.30 1.25 0.09 0.99
0.4
Lower 95% 0.11 0.48 0.01 0.35
Upper 95% 0.82 3.22 0.74 2.79
0.2

0.0
0 10 20 30 40 50
Time (months)

Figure 4.  Stratification of PFS and OS based on a combined Molecular Survival Score (MSS) for comparison of PFS and OS between TTFields vs
control groups. The OS, PFS, and statistical comparisons in all patients are shown here. Abbreviations: OS, overall survival; PFS, progression-free
survival; TTFields, Tumor-Treating Fields
  

As isocitrate dehydrogenase (IDH) mutations profoundly For more than a decade, the presence of MGMT promoter
affect patients’ survival, we excluded IDH mutants in a sub- methylation has been considered the prototype predictive bi-
group analysis and confirmed all observations in IDH-WT omarker of response to chemotherapy (TMZ) in GB, and it is
group (Supplementary Figure 7). a standard biomarker routinely tested in all patients with GB.
MGMT promoter is methylated in about 35%-50% of newly
diagnosed GBs10; in this dataset, 45% (24/53) of patients in
the TTFields group and 46% (25/56) in the control had tumors
Discussion that were MGMT methylated, thus consistent with wider re-
ports in other case series. There is increased interest and un-
The data presented in this study show that the presence derstanding of the impact of genetic alteration in IDH1 and
of NF1 mutation, EGFR wild type, and PIK3CA wild type 2 genes, which in many cases is a better determinant of out-
are suggestive of improved response to TTFields, as com- comes than histologic grades, this has led to changes in the
pared to patients whose tumors do not manifest this pro- 2016 and now the 2021 World Health Organization (WHO)
file. Furthermore, a combination of these factors, which classification of gliomas.11,12 IDH-mutated GB are typically
we designate as the TTFields MSS, provides a first attempt associated with better prognosis, about 5% of GB are IDH-
to create a robust tool to predict the impact of TTFields on mutated.13 In our dataset, 5 patients in theTTFields arm and 3
survival in patients with GB. In this study, we uncovered in the control arm had IDH mutations.
initial evidence of a molecular signature associated with Analysis of gene expression can be utilized to reveal
tumor response to therapy with TTFields. how varying cancers may be affected by TTFields and
TTFields are approved for the treatment of GB in the could signify treatment responsiveness. TP53 mutation
newly diagnosed and also in the recurrent setting. This status can influence the response to certain treatments and
therapeutic modality has increasingly been adopted as is linked to a worse prognosis. In a GB study, 4 cell lines
complementary to standard-of-care following maximal were used with varying TP53 status to influence TTFields
safe surgical resection and concurrent chemoradiation, treatment.14 Genes associated with cell cycle, cell death,
and concurrent with adjuvant chemotherapy; at the same and immune response were analyzed after TTFields appli-
time, a number of patients go on clinical trials foregoing cation and were altered despite TP53 status.14 In our study,
this treatment and there have been concerns raised about there was no difference detected in TTFields response be-
the quality of life, burden of care, and costs. These fac- tween tumors that were TP53 wild-type vs mutant. TP53
tors can be barriers in the use of TTFields therapy which is a ubiquitous tumor suppressor marker so it may not be
has significantly improved OS when incorporated into surprising that no differences were found. However, mu-
first-line treatment algorithms following GB diagnosis. tant NF1, which is more glioma-associated than TP53, was
Identification of a molecular signature associated with associated with differential response to TTFields, and this
the tumor response to this treatment is a promising signal pointed toward a composite score that, if validated,
decision-making tool. could help tailor therapeutic decision-making in the future.
 Pandey et al. Genomic factors affecting Tumor-Treating Fields 9

Advances
Neuro-Oncology
Until the past few years, compliance in device usage represents those patients who were both treated with or
over time had been the only factor associated with im- without TTFields and also had tumors sent for genomic
proved survival with TTFields. However, in the era of better profiling to Caris Life Sciences. During that time period
access, affordability, and knowledge regarding utility of (2014-2017), comprehensive genomic profiling was not
comprehensive genomic profiling in oncology, in general, as offered as often to patients with GBs. Since that time,
including in neuro-oncology, emerging targets are being comprehensive genomic profiling has become more
identified and are providing further insight into TTFields re- prevalent for patients with GB. In addition, as pointed out
sponse mechanisms. The mechanism of action of TTFields by a reviewer, patients treated with TTFields, in 2014-2017
seems to be cell cycle-dependent. Dono et al recently re- as well as now, tend to have better performance status as
ported results from a retrospective analysis that detected compared to patients who may not receive this treatment
higher post-progression survival of 14 patients with PTEN- based on physician assessment. While this study did not
mutated recurrent GBs who received treatment with comprehensively evaluate additional factors associated

Downloaded from https://academic.oup.com/noa/article/4/1/vdac096/6612779 by guest on 26 December 2022

TTFields, compared to 15 patients with PTEN wild-type tu- with outcome in this disease, such as extent of resection,
mors (22.2 vs 11.6 months, respectively). Studies like that use of steroids at baseline or throughout the treatment
one and the one we present here provide initial glimpses in course, etc., these and other factors can be taken into
the molecular biology of tumors that can either be affected consideration of design of future larger-scale studies will
by TTFields therapeutically, or otherwise provide a tumor further delve into validation of the potential biomarkers
microenvironment that is most susceptible to disruption we identify here. Such studies will be likely to uncover
via the alternating electric fields treatment strategy. new ones as accuracy and sensitivity of testing improves
To date, most of the focus on mechanism(s) of action of even further through whole transcriptome analysis and
TTFields has focused on basic cell biology, most notably other methods.
on effects on microtubules and cell division. The next fron- In summary, this retrospective study provides indication
tier of basic science research in this field is the interplay that there are potential genomic predictors of response to
of tumor genomics, most especially of driver mutations, TTFields treatment of patients with GB. As GB is the most
in laying the groundwork for a tumor microenvironment common tumor type for which this technology is in wide-
that lends itself to increased or decreased susceptibility spread use at the current time, our finding that the combi-
to cell disruption at the biophysical level. In a study with nation of NF1 mutation, EGFR wild type, and PIK3CA wild
non-small cell lung cancer (NSCLC) cells, there was a divi- type, formulated as a MSS, may be predictive of height-
sion of more responsive cell lines and less responsive cell ened response to TTFields warrants further investigation,
lines based on the treatment of TTFields. Previous studies including inclusion as a correlative biomarker in large-
used gene expression analysis to ascertain molecular scale prospective clinical trials that will be even more ro-
changes after treatment of TTFields and suggested that bust in the era of more widespread use of TTFields and also
certain genes were altered in both the less responsive cell comprehensive genomic profiling.
lines and more responsive cell lines.15 Ingenuity pathway
analysis (IPA) was performed to determine the canonical
pathways involved in the altered genes. Results suggested
alterations happened in cell cycle and mitotic pathways.15 Supplementary Material
Manifested by the IPA results, downregulation of BRCA1
Supplementary material is available at Neuro-Oncology
DNA damage response pathway was significant.15 These
Advances online.
proteins involved in the pathway are important for double-
stranded breaks within DNA, which could indicate reduced
DNA repair in TTFields-treated cells.15 Untapped aspects
and unknown factors that may merit further investigation Keywords
relating to TTFields including BRCA and other damage
response pathway effectors, and other genetic abnor- biomarkers | genomic profiling | glioblastoma | gliomas |
malities including mismatch repair deficits and tumor mu- Tumor-Treating Fields
tation burden, which have to date been associated with
response to immune checkpoint inhibitors.
Limitations of this study include potential for selection
bias inherent in any retrospective study, and relatively Funding
small numbers as compared to standard prospective
studies. The years that the patients examined in this study This research study was performed independently of and did not
were diagnosed and treated (2014-2017) represent the be- receive funding from the device manufacturer, Novocure.
ginning of the era in which TTFields began to be adopted
more widely internationally in this patient population;
nonetheless, more widespread acceptance of TTFields
therapy in the first-line setting for GB took hold more Acknowledgments
widely following that time period. Multiple institutions
were involved to maximize the number of cases analyzed; The authors would like to thank all the participating centers and
however, a number of patients from those participating support staff for contributing to this research project. We spe-
institutions were alternately enrolled in clinical trials cifically note at the journal reviewers’ request that no medical
and thus did not receive treatment with TTFields at that writer was used to produce this manuscript in any form.
time. The patient population we were able to examine
10 Pandey et al. Genomic factors affecting Tumor-Treating Fields

in the United States in 2012-2016. Neuro Oncol. 2019;21(Suppl

5):v1–v100.
Conflict of interest statement. S.K.  reports honoraria from Gerson
2. Adamson C, Kanu OO, Mehta AI, et al. Glioblastoma multiforme: a re-
Lehrman Group; Institutional Principal Investigator for clinical trials
view of where we have been and where we are going. Expert Opin
sponsored by Novocure, xCures, JLF foundation, Chimerix, Boehringer
Investig Drugs. 2009;18(8):1061–1083.
Ingelheim, Lilly, Daiichi Sankyo, Oblato, Spectrum. Institutional
3. Stupp  R, Mason  WP, van  den  Bent  MJ, et  al. Radiotherapy plus con-
Principal Investigator for clinical trials sponsored by Caris Life
comitant and adjuvant temozolomide for glioblastoma. N Engl J Med.
Sciences Precision Oncology Alliance (unpaid). A.S.  reports honor-
2005;352(10):987–996.
aria from Gerson Lehrman Group, Cardinal Health, and Curio Science;
4. Stupp R, Taillibert S, Kanner A, et al. Effect of tumor-treating fields plus
and honoraria from speakers’ bureau from Bristol Myers Squibb,
maintenance temozolomide vs maintenance temozolomide alone on sur-
Bayer, and Novocure. Travel expenses were previously provided by
vival in patients with glioblastoma: a randomized clinical trial. JAMA.
Novocure. Institutional Principal Investigator for clinical trials spon-
2017;318(23):2306–2316.
sored by Novocure, Bristol Myers Squibb, Exelixis, Kura Oncology,

Downloaded from https://academic.oup.com/noa/article/4/1/vdac096/6612779 by guest on 26 December 2022

5. Schneiderman RS, Voloshin T, Giladi M, et al. ATPS-25: p53 status de-
Oncoceutics, Chimerix. Institutional Principal Investigator for clinical
pendence of tumor treating fields (TTFields) efficacy against glioma
trials sponsored by Caris Life Sciences Precision Oncology Alliance
cancer cells. Neuro Oncol. 2015; 17:v23.
(unpaid). E.L. reports research grants from the American Association
6. Alizadeh AA, Aranda V, Bardelli A, et al. Toward understanding and ex-
for Cancer Research (AACR-Novocure Tumor-Treating Fields
ploiting tumor heterogeneity. Nat Med. 2015;21(8):846–853.
Research Award, Grant Number 19-60-62-LOU) and the Minnesota
7. Tanaka S, Batchelor TT, Iafrate AJ, et al. PIK3CA activating mutations
Ovarian Cancer Alliance; honorarium and travel expenses for a re-
are associated with more disseminated disease at presentation and
search talk at GlaxoSmithKline in 2016; honoraria and travel expenses
earlier recurrence in glioblastoma. Acta Neuropathol Commun. 2019;
for laboratory-based research talks and equipment for laboratory-
7(1):66.
based research, Novocure, LLC, 2018-present; honorarium for panel
8. Lobbous M, Bernstock JD, Coffee E, et al. An update on neurofibroma-
discussion organized by Antidote Education for a CME module on
tosis type 1-associated gliomas. Cancers. 2020; 12(1):1–15. https://
diagnostics and treatment of HER2+ gastric and colorectal cancers,
www.mdpi.com/2072-6694/12/1/114.
funded by Daiichi Sankyo, 2021 (honorarium donated to laboratory);
9. Felsberg J, Hentschel B, Kaulich K, et al. Epidermal growth factor re-
consultant, NomoCan Pharmaceuticals (unpaid); Scientific Advisory
ceptor variant III (EGFRvIII) positivity in EGFR-amplified glioblastomas:
Board Member, Minnetronix, LLC, 2018-present (unpaid); consultant
prognostic role and comparison between primary and recurrent tumors.
and speaker honorarium, Boston Scientific, USA, 2019. Institutional
Clin Cancer Res. 2017;23(22):6846–6855.
Principal Investigator for clinical trials sponsored by Celgene,
10. Mellai M, Monzeglio O, Piazzi A, et al. MGMT promoter hypermethylation
Novocure, Intima Biosciences, and the National Cancer Institute,
and its associations with genetic alterations in a series of 350 brain tu-
and University of Minnesota membership in the Caris Life Sciences
mors. J Neurooncol. 2012;107(3):617–631.
Precision Oncology Alliance (unpaid). The other authors report no
11. Brat  DJ, Aldape  K, Colman  H, et  al. cIMPACT-NOW update 3: recom-
conflicts of interest.
mended diagnostic criteria for “Diffuse astrocytic glioma, IDH-wildtype,
with molecular features of glioblastoma, WHO grade IV”. Acta
Neuropathol. 2018;136(5):805–810.
Authorship statement. Conception of study: M.P.  and J.X. 12. Louis  DN, Perry  A, Wesseling  P, et  al. The 2021 WHO classification
Identification of patients for tumor analysis: M.P., S.M., M.S., S.K., of tumors of the central nervous system: a summary. Neuro Oncol.
A.A., H.N., and A.S. Data analysis and writing the initial drafts of 2021;23(8):1231–1251.
the manuscript: M.P., J.X., and EL. Review, editing, and approval of 13. Yan H, Parsons DW, Jin G, et al. IDH1 and IDH2 mutations in gliomas. N
the final form of the manuscript for publication: all authors. Engl J Med. 2009;360(8):765–773.
14. Lee YJ, Seo HW, Baek JH, Lim SH, Hwang SG, Kim EH. Gene expres-
sion profiling of glioblastoma cell lines depending on TP53 status
after tumor-treating fields (TTFields) treatment. Sci Rep. 2020;10(1):
12272.
References 15. Karanam NK, Srinivasan K, Ding L, Sishc B, Saha D, Story MD. Tumor-
treating fields elicit a conditional vulnerability to ionizing radiation via
the downregulation of BRCA1 signaling and reduced DNA double-strand
1. Ostrom QT, Cioffi G, Gittleman H, et al. CBTRUS statistical report: break repair capacity in non-small cell lung cancer cell lines. Cell Death
primary brain and other central nervous system tumors diagnosed Dis. 2017;8(3):e2711.