Annals of Oncology 30: 211–218, 2019

doi:10.1093/annonc/mdy544
Published online 21 December 2018

REVIEW

Cerebrospinal fluid cell-free tumour DNA as a liquid

biopsy for primary brain tumours and central
nervous system metastases

J. Seoane1,2,3,4*, L. De Mattos-Arruda1, E. Le Rhun5,6,7, A. Bardelli8,9 & M. Weller10

1
Vall d’Hebron Institute of Oncology, Vall d’Hebron University Hospital, Barcelona; 2Institució Catalana de Recerca i Estudis Avançats (ICREA), Barcelona; 3CIBERONC,
Barcelona; 4Universitat Autònoma de Barcelona, Cerdanyola del Vallès; 5Lille University, Inserm U1192 PRISM, Villeneuve d’Ascq; 6Neuro-oncology, Department of
Neurosurgery, University Hospital, Lille; 7Neuro-oncology, Breast Unit, Department of Medical Oncology, Oscar Lambret Center, Lille, France; 8Candiolo Cancer
Institute-FPO, IRCCS, Candiolo (TO); 9Department of Oncology, University of Torino, Candiolo (TO), Italy; 10Department of Neurology, University Hospital and
University of Zurich, Zurich, Switzerland

*Correspondence to: Prof. Joan Seoane, Vall d’Hebron Institute of Oncology, C/Natzaret, 115-117, 08035 Barcelona, Spain.
Tel: þ34-93-254-34-50; E-mail: [email protected]

Challenges in obtaining tissue specimens from patients with brain tumours limit the diagnosis and molecular characterisation
and impair the development of better therapeutic approaches. The analysis of cell-free tumour DNA in plasma (considered a
liquid biopsy) has facilitated the characterisation of extra-cranial tumours. However, cell-free tumour DNA in plasma is limited in
quantity and may not reliably capture the landscape of genomic alterations of brain tumours. Here, we review recent work
assessing the relevance of cell-free tumour DNA from cerebrospinal fluid in the characterisation of brain cancer. We focus on the
advances in the use of the cerebrospinal fluid as a source of cell-free tumour DNA to facilitate diagnosis, reveal actionable
genomic alterations, monitor responses to therapy, and capture tumour heterogeneity in patients with primary brain tumours
and brain and leptomeningeal metastases. Profiling cerebrospinal fluid cell-free tumour DNA provides the opportunity to
precisely acquire and monitor genomic information in real time and guide precision therapies.
Key words: cerebrospinal fluid, circulating cell-free tumour DNA, glioblastoma, brain metastasis,
liquid biopsy, brain cancer

Invasive surgical procedures have been the cornerstone treat-

Introduction ment and a diagnostic tool in patients with primary brain
Genomic characterisation of tumour tissue has been established tumours and in selected patients with brain metastasis.
as crucial for state-of-the-art diagnostic and therapeutic However, collecting tumour tissue from central nervous system
approaches to cancer. However, characterisation of cancer is (CNS) malignancies is complex, can be risky, and sometimes
challenged by constitutive and evolving intra-tumour and inter- unfeasible, at least with purely diagnostic intent. Surgery has a
lesion heterogeneity, which requires thorough and continuous role in improving disease control in patients with primary
analysis of genomic complexity over time. This is particularly tumours or with a single, resectable brain metastasis, whereas
relevant in brain malignancies where the genomic landscape patients with disseminated systemic disease are frequently not
changes in response to treatment or during relapse and can differ candidates for routine neurosurgical procedures [1, 2].
from the primary extra-cranial lesion in the case of brain metasta- Moreover, specimens may be small and not representative
ses. Yet, obtaining samples for characterisation and correct diag- hampering correct diagnosis or even necessitating multiple sur-
nosis can be difficult in brain cancer patients. The anatomical gical samplings to clarify final pathological diagnosis. In add-
location of the tumour limits access due to the risk and complex- ition, the surgical intervention strategy and assessment of the
ity of intracranial surgical procedures. surgical risk–benefit balance depend on the tumour prognosis.

C The Author(s) 2018. Published by Oxford University Press on behalf of the European Society for Medical Oncology.
V
All rights reserved. For permissions, please email: [email protected].
Review Annals of Oncology
This implies that an intraoperative histological diagnosis may aiding in monitoring response to therapy, and allowing deconvo-
be required possibly delaying the surgical procedure. Repeat lution of tumour heterogeneity in patients with CNS cancer
surgical interventions may be needed to differentiate tumour (Figure 1).
pseudoprogression induced by treatment from true relapse.
The challenges in obtaining tumour tissue have led physicians
to rely on primary archival tumour specimens. Thus, in some
cases, therapies for brain cancer are selected based on the mo- Primary brain tumours
lecular characteristics of the primary tumour that can differ
from the current tumour manifestation [3, 4]. Diagnostic considerations
Plasma cell-free circulating tumour DNA (ctDNA) has been Primary brain tumours encompass a large variety of lesions
used as a ‘liquid biopsy’ in the context of tumour genomic char- with diverse natural course, response to treatment, and prog-
acterisation [5–12]. ctDNA is the fraction of the total cell-free nosis. The histological grade and molecular genetic make-up
DNA that is derived from tumour cells and can be defined by the determine prognosis, with median overall survivals ranging
presence of genomic alterations. ctDNA detected in plasma has from <1 year (e.g. in glioblastoma of the elderly) to long-term
shown promise in characterising tumours and allowing patients survival including cures (e.g. pilocytic astrocytoma and other
and their cancers to be monitored over time. Analyses of muta- rare circumscribed lesions). The clinical hallmark of glioblast-
tions in plasma ctDNA have demonstrated high concordance oma is aggressive growth, local invasiveness, and inexorable re-
with genomic alterations in the tumour [10]. currence [30–32]. In recent years, the development of novel
However, in the context of primary brain tumours and brain sequencing technologies and DNA methylation profiling
metastasis, plasma ctDNA has been shown to be in low abun- coupled to bioinformatics tools has yielded an unparalleled,
dance and present in a limited number of patients [8, 13–16]. comprehensive view of the genome and epigenome of brain
Importantly, the cerebrospinal fluid (CSF) is in intimate contact tumours [33–36].
with brain malignancies and has been recently proved to contain The 2016 update of the WHO classification incorporated well
ctDNA. The CSF space involves the intracerebral ventricles, sub- established molecular parameters into the classification of brain
arachnoid spaces of the spine and brain (cisterns and sulci), and tumours, specifically gliomas. The analysis of the CSF ctDNA of
the central spinal cord canal. The CSF is renewed three to five a cohort of diffuse gliomas indicated that they could be sub-
times daily and is produced by the choroid plexus. The CSF circu- typed by analysing the IDH1 and IDH2, ATRX, TP53, TERT,
lates in a craniocaudal direction from ventricles to spinal sub- H3F3A and HIST1H3B mutational status, facilitating the classi-
arachnoid space from where it is removed via craniocaudal fication of diffuse gliomas and providing prognostic informa-
lymphatic routes and the venous system [17]. The CSF space is tion [28]. Moreover, the presence of mutations in the TERT
separated from the vascular system by the blood–CSF barrier, promoter found in CSF ctDNA correlated with outcome [37].
while the blood–brain barrier is located between the brain paren- In the case of diffuse midline gliomas, the detection of H3F3A
chyma and the vascular system [18]. and HIST1H3B mutations in the CSF could confirm diagnosis
CSF has been explored as a source of ctDNA for precisely char- [28]. This is of major relevance since the anatomical location of
acterising brain cancers. Studies reported before the era of high- this type of tumours increases the risk of obtaining surgical
throughput sequencing showed that some molecular alterations specimens.
or gene mutations can be detected in the DNA present in the CSF CSF ctDNA was detected in a large proportion of patients
of patients with brain tumours [19–24]. Importantly, massively with brain primary tumours (Table 1). However, CSF ctDNA is
parallel sequencing methods have recently been used to analyse not found in all brain tumours. For example, in some low grade
cell-free tumour DNA from CSF to comprehensively characterise gliomas, CSF ctDNA was not detected or was not informative
somatic alterations including gene mutations and copy number [28]. Technological advances may improve sequencing sensitiv-
alterations [15, 25–29] (Table 1). ity in the future, thus reducing the number of non-informative
DNA was isolated from CSF (ranging from 0.75 to 10 ml) usu- cases.
ally obtained from a lumbar puncture and DNA sequencing (i.e.
droplet digital PCR, targeted sequencing, whole-exome sequenc-
ing, or shallow whole-genome sequencing) allowed the identifi-
Therapeutic considerations
cation of ctDNA. Notably, CSF ctDNA enabled the identification ctDNA diagnostic applications with potential therapeutic impli-
of genomic alterations in patients with systemic metastatic bur- cations remain limited for adult patients with primary brain
den including brain metastasis, or disease restricted to the brain tumours. The most relevant biomarker for glioblastoma in terms
(primary tumours and brain metastasis) [15, 25–27]. Higher of choice of therapy remains promoter methylation of the
grade brain tumours were more likely to exhibit detectable CSF MGMT gene [38]. Efforts at the detection of MGMT promoter
ctDNA than lower grade ones [26] and, in some cases, the dis- methylation in CSF of glioma patients showed higher sensitivity
tance of the tumour to CSF spaces could determine the amount than in plasma [39]. Future applications with therapeutic impact
of CSF ctDNA [27]. are likely to include the monitoring of EGFRvIII and amplified
Here, we focus on the studies related to ctDNA obtained from epidermal growth factor receptor (EGFR) in patients undergoing
CSF. Nowadays, the CSF liquid biopsy is increasingly allowing EGFR-targeted therapy [40]. The evaluation of ctDNA during the
molecular diagnoses, providing information on prognosis, facili- follow-up of patients and especially at recurrence can confirm the
tating the identification of new actionable genomic alterations, molecular status and may help to deliver precision therapies.

212 | Seoane et al. Volume 30 | Issue 2 | 2019

 Table 1. Studies detecting ctDNA in CSF
Annals of Oncology

Patient population Sequencing technologies Method for CSF collection/ Tumour-derived Main findings Comparison Extra-cranial Refs
amount collected DNA found in with plasma disease

Volume 30 | Issue 2 | 2019

CSF (% of cases) ctDNA analysed

Four glioblastoma, 2 medulloblas- Targeted capture massively parallel Lumbar puncture, cerebral shunts, 58–60 CSF ctDNA is enriched in brain 冑 冑 [15]
toma, and 17 brain metastases sequencing, digital droplet PCR autopsy/ tumours and produces better results
(from breast and NSCLC) than plasma ctDNA.
Thirty-five primary brain and spinal
Targeted capture massively parallel Cerebral shunts (during surgical pro- 57–88 All adjacent tumours to the CSF reser- – – [26]
cord tumours sequencing, whole-exome sequenc- cedure)/average of 4.8 ml voir had CSF ctDNA detectable
ing in four cases (range¼0.75–10 ml)
Twelve primary brain tumours and Targeted capture massively parallel Lumbar puncture and one sample 50–63 Drug-resistance mutations in patients – 冑 [27]
41 brain metastases sequencing Ommaya reservoir/5 ml of CSF whose CNS disease progresses during
kinase inhibitor therapy is identified in
CSF ctDNA.
A vestibular schwannoma, a men- Targeted amplicon sequencing and Lumbar puncture, cerebral shunts 85 Tumour mutations were detectable in 冑 – [25]
ingioma, five brain metastases and digital PCR (during surgical procedure)/1–10 ml the CSF ctDNA of patients with differ-
three leptomeningeal metastases of CSF ent types of brain tumours
Twenty primary diffuse glioma Targeted amplicon sequencing and Lumbar puncture before surgery; two 85 A sequencing platform to simultan- – – [28]
tumours digital PCR samples cisterna magna, one sample eously test seven genes IDH1, IDH2,
cerebral shunt/2 ml of CSF TP53, ATRX, TERT, H3F3A, HIST1H3B in
the CSF ctDNA allowing the subclassifi-
cation of diffuse glioma
Thirty-eight TERT mutant glioblast- Sequenced uni-directionally on an Directly after opening the dura (durot- 78–98 CSF ctDNA identifies TERT promoter 冑 [37]
oma (34 primary and 4 recurrent Ion Torrent PGM omy), through dissection of the con- mutations
glioblastoma) NGS system, digital droplet PCR vexity subarachnoid space/2–4 ml of
CSF
Thirteen glioma tumours Untargeted, low-coverage WGS Lumbar puncture/10 ml of CSF 39 Combining analyses of SCNAs with – – [29]
(<0.4) to detect SCNAs DNA fragmentation allows detection
of ctDNA in CSF using sWGS data at
low cost

CNS, central nervous system; GBM, glioblastoma; LM, leptomeningeal metastasis; NSCLC, non-small cell lung cancer; SCNAs, somatic copy number alterations; s-WGS, shallow whole-genome sequencing.

doi:10.1093/annonc/mdy544 | 213
Review
Review Annals of Oncology
Molecular diagnosis
monitoring actionable mutations and therapy resistance using
prognosis CSF ctDNA appears to be an application of CSF-based liquid
biopsies that could be close to clinical practice.
First- (erlotinib and gefitinib) and second-generation (afati-
nib) EGFR tyrosine kinase inhibitors (TKI) have shown activity
against brain metastasis from non-small-cell lung cancers
Brain
GBM metastasis (NSCLC) that harbour EGFR mutations [56–58]. A number of
third-generation EGFR-TKI that also target mutant EGFR
Monitor Actionable T790M, which confers therapeutic resistance, are in various
tumor genomic
CSF alterations
phases of clinical investigation to target brain metastases (osimer-
burden
tinib, rociletinib, ASP-8273, HM-61713). In anaplastic lymph-
oma kinase (ALK) gene-rearranged (ALK)-NSCLC, second-
generation ALK inhibitors with increased potency such as alecti-
ctDNA
nib and ceritinib have apparently superior CNS penetration com-
pared with crizotinib and share significant therapeutic potential
[53–55]. Breast cancer studies have focussed primarily on tar-
geted therapies [e.g. lapatinib, pertuzumab, ado-trastuzumab
Evolving heterogeneity
emtansine (T-DM1)] used for HER2-positive cancers [51, 52, 59,
(tracking diversity and evolution)
60]. In patients with melanoma and brain metastases, substantial
clinical activity has been observed with BRAF and MEK inhibi-
Figure 1. Potential use of CSF ctDNA as a liquid biopsy for primary tors, e.g. dabrafenib plus trametinib [49, 61], resulting in an
brain tumours and brain metastasis. intracranial response rate of nearly 60% [61]. Ongoing clinical
trials exploit the cytotoxic T-lymphocyte-associated antigen 4
Brain metastases and programmed death 1 pathways as target for immune check-
point inhibitor therapy [50, 62]. Actionable genomic alterations
Diagnostic considerations with potential therapeutic implications have been identified in
the CSF ctDNA [15, 25–27], including EGFR, ALK, HER2,
Brain metastases from solid tumours are more frequent than pri- BRAF-targetable kinases, and others associated with DNA integ-
mary brain tumours. They may occur in 20%–40% of advanced rity such as BRCA1 and BRCA2 [63].
stage cancers, particularly in lung cancer, breast cancer and melan- Analysis of CSF ctDNA has also shown gene mutations associ-
oma [41–43]. Recent reports on the branched evolution of cancer ated with therapy resistance [15, 27, 64]. Drug-resistance muta-
at different sites including metastasis to the brain have reinforced tions in patients whose CNS disease progressed during TKI
the need of sequential molecular profiling across the disease trajec- therapy (EGFR, ALK, HER2, or BRAF) were identified in CSF
tory [4, 44]. Brain metastases exhibit different genomic alterations ctDNA in one-third of cases [27]. This included a NRAS G12R
than the primary extra-cranial tumours [4] indicating that the mutation in the CSF of a BRAF V600E-mutant (and NRAS-
brain lesion-specific genomic alterations should be identified to se- negative) melanoma; a PIK3CA H1047R mutation in the CSF of a
lect the optimal therapeutic approach [4, 45]. CSF ctDNA and not HER2-amplified breast cancer patient, potentially associated
plasma ctDNA can be a good surrogate marker in such situations with trastuzumab resistance; and an EGFR T790M mutation in
since ctDNA from brain lesions is enriched in the CSF. the CSF of a patient with EGFR-mutant NSCLC who did not re-
Trunk mutations, present in all cancer cells, as well as private gen- spond to a second-generation EGFR-TKI [27]. ESR1 mutations
omic alterations, present in just a subpopulation of cells or in specif- can confer resistance to aromatase inhibitor therapy in advanced
ic metastatic lesions, can be identified in the CSF [15, 27]. This estrogen receptor-positive breast cancers, but not to fulvestrant
allows opportunities for deconvolving tumour heterogeneity. CSF [65, 66]. A clinical trial is evaluating ESR1 mutations in plasma
and plasma ctDNA were compared in a series of samples that ctDNA to predict the efficacy of a change of the hormone therapy
included multiregional metastatic sites from postmortem speci- (aromatase inhibitor changed to fulvestrant) (ClinicalTrials.gov
mens of patients with disseminated breast cancers including brain Identifier: NCT03079011). Translation of this type of clinical trial
metastases [15]. For example, mutations found in the CSF ctDNA design to the setting of patients with brain metastasis is envi-
allowed to discern the origin of leptomeningeal and brain metasta- sioned using liquid biopsies. In the context of multiple metastases
sis implants separately in a patient with Li Fraumeni syndrome and and discordant clinico-radiological findings, analysing a single-
two concurrent tumours, a metastatic breast cancer and esthesio- lesion biopsy is inadequate in guiding the selection of targeted
neuroblastoma [15]. CSF ctDNA analysis captured trunk mutations therapy [67]. Parallel analyses of serial CSF and plasma ctDNA
and, importantly, private mutations to the brain and to the menin- samples may be warranted.
geal deposits. These observations highlight the potential applica- Although the most common initial clinical presentation of
tions of CSF ctDNA to complement diagnosis of brain metastasis. metastatic HER2-positive breast cancer is with extra-cranial
metastases, CNS progression occurs in a substantial proportion
of patients during the course of the disease [68]. It has also been
Therapeutic considerations shown that extensive extra-CNS disease control, with HER2 tar-
Several targeted therapeutic agents have demonstrated clinical ac- geting, might drive high incidence of CNS progression [69, 70].
tivity against established brain metastases [46–55] and This situation remains a major challenge where genomic analysis

214 | Seoane et al. Volume 30 | Issue 2 | 2019

Annals of Oncology Review
of ctDNA in the CSF might allow interrogation of the molecular Further work is warranted to consolidate CSF ctDNA as a com-
status of progressive CNS metastasis. For example, in a case vi- plementary tool for the diagnosis and characterization of lepto-
gnette, CSF ctDNA analysis captured CNS genomic alterations in meningeal metastasis. Accordingly, the EANO ESMO guideline
patients with absent or minimal extracranial tumour disease bur- advises caution in over-interpreting ctDNA detected in CSF as a
den, where plasma ctDNA profiling did not play a diagnostic role proof of leptomeningeal seeding.
[15]. A HER2-positive metastatic breast cancer patient with di-
vergent responses of brain metastases underwent autopsy. Copy
number alteration testing of three spatially separated brain meta-
Therapeutic considerations
stases, in addition to CSF ctDNA and plasma ctDNA sampling, Recent and ongoing studies address the role of CSF ctDNA in
showed ERBB2 amplification, a hallmark of HER2-positive breast patients with EGFR-mutant NSCLC and leptomeningeal metas-
cancer in CSF ctDNA and not in the plasma analysis [15]. tasis [27, 64]. Forty NSCLC patients with suspected leptomenin-
geal metastasis were profiled, including 35 patients with a
confirmed leptomeningeal metastases [64]. EGFR T790M and
MET amplification were detected in 21% and 39% in CSF
Leptomeningeal metastases ctDNA, respectively, suggesting a resistance profile to EGFR-TKI
associated with leptomeningeal disease. The BLOOM study
Diagnostic considerations (ClinicalTrials.gov Identifier: NCT02228369) investigates osi-
mertinib, an oral, irreversible third-generation EGFR-TKI select-
Leptomeningeal metastasis, defined by the multifocal seeding of ively active against the EGFR T790M resistance mutation [76].
the leptomeninges by malignant cells, is a rare but often rapidly Encouraging activity has been seen in patients with leptomenin-
fatal manifestation of advanced cancer [71–73]. Diagnosing lepto- geal metastasis from NSCLC and results of EGFR-mutant ctDNA
meningeal metastasis relies on clinical symptomatology, MRI, and analyses are being awaited (ClinicalTrials.gov Identifier:
on detecting malignant cells in the CSF through CSF cytology. Its NCT02228369) [77]. Thus, ctDNA analysis should be considered
incidence is increasing and prognosis remains poor despite radio- for the EGFR and T790M status in the CSF at diagnosis and in
therapy, systemic and intrathecal chemotherapy, and precision case of suspicion of progression of NSCLC leptomeningeal meta-
treatments in molecularly selected patients [71, 72]. stases to guide the therapeutic decision.
Two principal diagnostic applications of ctDNA studies in
patients with leptomeningeal metastasis emerge: first, detecting
CSF ctDNA in patients with leptomeningeal metastasis may com-
plement diagnostic profiling in patients with negative cytology, Road to clinical practice
second, identifying actionable genomic alterations in CSF ctDNA To integrate the assessment of ctDNA obtained from CSF liquid
has the potential to define an optimal targeted therapy [15]. CSF biopsies into current standards of care, several questions and con-
ctDNA revealed mutations in 50% of patients with primary brain troversies have to be addressed. For almost all primary brain
tumours despite their CSF being negative for malignant cells tumours, extent of resection is an important prognostic factor.
[27], further, among patients with brain metastases, somatic Thus, situations where a surgical intervention is not an option, but
mutations were found in 100% of patients with positive cytology a diagnosis would still be welcome, are rare. These might include
and in 25% of patients with negative cytology [27]. patients with major comorbidities thought to be at high risk of
A pivotal study compared CSF profiling with CSF cytology complications, e.g. those with high bleeding risk for various rea-
results in the same CSF extraction [15]. Analysis of three meta- sons. Furthermore, there are instances where initial stereotactic
static breast cancer patients with clinical signs and symptoms biopsies of brain lesions are not informative and where non-
suggestive of leptomeningeal metastasis showed that CSF ctDNA neoplastic lesions, e.g. neuroinflammatory or neuroinfectious dis-
analysis was more sensitive than cytology in detecting leptomen- eases are a differential diagnosis. In these cases, detection of
ingeal metastasis, and leptomeningeal infiltration was confirmed tumour-defining genomic alterations in the CSF ctDNA may
during autopsy [15]. A molecular case report compared paired greatly aid in further management. Furthermore, DNA methyla-
profiling of matched CSF ctDNA and plasma ctDNA from a pa- tion profiling may represent a novel approach that will undoubted-
tient with HER2-positive metastatic breast cancer. The patient ly also be explored for confirming tumour diagnoses from small
developed CNS progression and leptomeningeal metastasis tissue samples, including CSF [36]. Future studies will also need to
whereas the systemic extracranial metastases showed a clinical determine in how far serial assessments of ctDNA load in the CSF
and radiological response to treatment with T-DM1 [74]. CSF may aid in situations where response assessment based on MRI
ctDNA revealed an enrichment of ERBB2 amplification, MYC alone remains challenging, including brain tumours treated with
amplification and PIK3CA and TP53 driver gene mutations, pre- immunotherapy, and help clinical decision making.
sumably reflecting CNS progression whereas decreasing mutant The situation is different for patients with brain metastases
allelic fractions of selected mutations in plasma ctDNA likely from solid tumours. For brain metastases from unknown pri-
reflected a partial clinical response in the extracranial compart- mary tumours, either rapid neurosurgical intervention as clinic-
ment [74]. In metastatic melanoma spreading to the leptomenin- ally needed or initial work-up by chest abdomen CT or FDG-PET
ges, CSF examination using PCR-based techniques has been are standard procedures [78] whereas liquid biopsies have so far
successfully used for diagnosis and monitoring response to ther- not assumed a role. However, patients with new brain lesions
apy based on the detection of driver mutations, e.g. affecting detected by neuroimaging who are known to suffer from a malig-
BRAF [64, 75]. nancy are not routinely sent for neurosurgical resection unless

Volume 30 | Issue 2 | 2019 doi:10.1093/annonc/mdy544 | 215

Review Annals of Oncology
this is thought to be in the best interest of the patient, e.g. because and monitoring brain-specific characteristics through CSF
there are concerns regarding the validity of the radiological diag- ctDNA may expedite the design of targeted therapies. Yet, no li-
nosis, or because the patient is neurologically symptomatic. In quid circulating biomarkers have been validated and integrated
such circumstances, notably with tumours with targetable lesions into clinical practice for primary brain tumours or brain metasta-
or in patients pre-exposed to chemotherapy or targeted therapy, ses. CSF ctDNA is a promising instrument to evaluate CNS
it may be of major interest to ascertain whether the molecular tu- malignancies in real-time and guide therapeutic management of
mour profile has changed, to select the most appropriate treat- patients.
ment. In such scenarios molecular tumour profiling from ctDNA The treatment of human cancer has shifted towards a precision
from the CSF could become most valuable. medicine paradigm, in which the selection of a targeted therapy
Similar considerations apply to patients with known leptomen- will rely upon the genetic anomalies in individual patients. We
ingeal metastases, and targeted treatments are available for the predict that characterising brain tumours will be feasible using
major primary tumours associated with leptomeningeal metasta- CSF ctDNA in the near future. In addition, combining plasma
sis: lung cancer, and breast cancer, and melanoma. The role of ctDNA with CSF ctDNA, morphological analyses and imaging
ctDNA in the CSF remains controversial if neither MRI nor rou- methods would ideally be complementary for patients with brain
tine CSF studies suggest the presence of leptomeningeal metastasis. metastases and systemic disseminated disease. Thus, liquid bi-
In such situations, it can at present not be clarified whether ctDNA opsy approaches based on CSF are opening new avenues for the
detected in the CSF signifies leptomeningeal tumour cell seeding. better managing of brain cancer patients.
Thus, further studies are required to show that ctDNA in the CSF
alone may justify treatment directed against leptomeningeal me-
tastasis [79]. However, the identification of the EGFR mutation or
T790M in the CSF of patients with CNS metastases help guide the Funding
therapeutic decision in NSCLC patients. The authors acknowledge Asociación Espa~ nola contra el Cáncer
Regarding the current data, ctDNA should be explored for the (JS, LDMA), Fondo de Investigación Sanitaria (FIS) Instituto de
diagnosis and in case of suspicion of progression of leptomeningeal Salud Carlos III grant (PI16/01278) (JS), and the FERO -EDM
metastases. CSF and plasma ctDNA should be evaluated in parallel. support-, LaCaixa and Cellex foundations (JS, LDMA).
Fondazione Piemontese per la Ricerca sul Cancro-ONLUS 5 per
mille 2011 e 2014 Ministero della Salute (AB). H2020 grant agree-
Current limitations ment no. 635342-2 MoTriColor (AB) and AIRC IG no. 17707
(AB). AIRC Special Program 5 per mille metastases project no.
Current data on ctDNA are mainly reported in small cohorts of 21091.
patients, including sometimes different primary tumours.
Analyses of large cohorts of patients should be carried out.
Technical issues such as the potential blood contamination in the
CSF sample or the minimum time interval between surgery and Disclosure
CSF analysis for ctDNA have to be evaluated. Confirmation stud-
All authors have declared no conflicts of interest.
ies are needed to validate the role of ctDNA analysis for the diag-
nosis and follow-up of patients. Importantly, the feasibility of the
CSF analysis in patients with brain tumours have to be considered References
when the lumbar puncture is contra-indicated due to risk of her- 1. Owen S, Souhami L. The management of brain metastases in non-small
niation related to the presentation of the space-occupying CNS cell lung cancer. Front Oncol 2014; 4: 248.
tumours, or abnormal coagulation. 2. Ferguson SD, Wagner KM, Prabhu SS et al. Neurosurgical management
of brain metastases. Clin Exp Metastasis 2017; 34: 377–389.
3. Johnson BE, Mazor T, Hong C et al. Mutational analysis reveals the ori-
gin and therapy-driven evolution of recurrent glioma. Science 2014;
Discussion 343(6167): 189–193.
4. Brastianos PK, Carter SL, Santagata S et al. Genomic characterization of
Conclusions and future perspectives brain metastases reveals branched evolution and potential therapeutic
targets. Cancer Discov 2015; 5(11): 1164–1177.
Increasing understanding of the genomic and epigenomic char- 5. Best MG, Sol N, Zijl S et al. Liquid biopsies in patients with diffuse gli-
oma. Acta Neuropathol 2015; 129(6): 849–865.
acteristics of primary brain tumours and brain and leptomenin-
6. De Mattos-Arruda L, Cortes J, Santarpia L et al. Circulating tumour cells
geal metastases has uncovered the extraordinary complexity of and cell-free DNA as tools for managing breast cancer. Nat Rev Clin
these tumours [4, 48, 80, 81]. Nevertheless, identifying bio- Oncol 2013; 10(7): 377–389.
markers to assist in the diagnosis, prognosis, prediction of tar- 7. Murtaza M, Dawson SJ, Tsui DW et al. Non-invasive analysis of acquired
geted therapy responses, serial monitoring and mechanisms of resistance to cancer therapy by sequencing of plasma DNA. Nature 2013;
therapy resistance for patients with CNS malignancies [80] 497(7447): 108–112.
8. Bettegowda C, Sausen M, Leary RJ et al. Detection of circulating tumor
remains challenging, in part because of difficulties in accessing
DNA in early- and late-stage human malignancies. Sci Transl Med 2014;
CNS tumour-derived tissue. 6: 224ra224.
CNS malignancies demonstrate considerable spatial and tem- 9. Siravegna G, Mussolin B, Buscarino M et al. Clonal evolution and resist-
poral intra-tumour and inter-tumour heterogeneity. For patients ance to EGFR blockade in the blood of colorectal cancer patients. Nat
with primary brain tumours or with brain metastases, identifying Med 2015; 21(7): 827.

216 | Seoane et al. Volume 30 | Issue 2 | 2019

Annals of Oncology Review
10. Phallen J, Sausen M, Adleff V et al. Direct detection of early-stage cancers 32. Omuro A, DeAngelis LM. Glioblastoma and other malignant gliomas: a
using circulating tumor DNA. Sci Transl Med 2017; 9: 403. clinical review. JAMA 2013; 310(17): 1842–1850.
11. Alix-Panabieres C, Pantel K. Clinical applications of circulating tumor 33. Sottoriva A, Spiteri I, Piccirillo SG et al. Intratumor heterogeneity in
cells and circulating tumor DNA as liquid biopsy. Cancer Discov 2016; human glioblastoma reflects cancer evolutionary dynamics. Proc Natl
6(5): 479–491. Acad Sci USA 2013; 110(10): 4009–4014.
12. Siravegna G, Marsoni S, Siena S, Bardelli A. Integrating liquid biopsies 34. Cancer Genome Atlas Research Network, Brat DJ, Verhaak RG et al.
into the management of cancer. Nat Rev Clin Oncol 2017; 14(9): Comprehensive, integrative genomic analysis of diffuse lower-grade glio-
531–548. mas. N Engl J Med 2015; 372: 2481–2498.
13. Lavon I, Refael M, Zelikovitch B et al. Serum DNA can define tumor- 35. Ceccarelli M, Barthel FP, Malta TM et al. Molecular profiling reveals bio-
specific genetic and epigenetic markers in gliomas of various grades. logically discrete subsets and pathways of progression in diffuse glioma.
Neuro Oncol 2010; 12(2): 173–180. Cell 2016; 164(3): 550–563.
14. Chen WW, Balaj L, Liau LM et al. BEAMing and droplet digital PCR 36. Capper D, Jones DTW, Sill M et al. DNA methylation-based classifica-
analysis of mutant IDH1 mRNA in glioma patient serum tion of central nervous system tumours. Nature 2018; 555(7697):
and cerebrospinal fluid extracellular vesicles. Mol Ther Nucleic 469–474.
Acids 2013; 2: e109. 37. Juratli TA, Stasik S, Zolal A et al. TERT promoter mutation detection in
15. De Mattos-Arruda L, Mayor R, Ng CK et al. Cerebrospinal fluid-derived cell-free tumor-derived DNA in patients with IDH wild-type glioblasto-
circulating tumour DNA better represents the genomic alterations of mas: a pilot prospective study. Clin Cancer Res 2018; 24(21): 5282–5291.
brain tumours than plasma. Nat Commun 2015; 6: 8839. 38. Weller M, van den Bent M, Tonn JC et al. European Association for
16. Boisselier B, Gallego Perez-Larraya J, Rossetto M et al. Detection of Neuro-Oncology (EANO) guideline on the diagnosis and treatment of
IDH1 mutation in the plasma of patients with glioma. Neurology 2012; adult astrocytic and oligodendroglial gliomas. Lancet Oncol 2017; 18(6):
79(16): 1693–1698. e315–e329.
17. Ghersi-Egea JF, Strazielle N, Catala M et al. Molecular anatomy and 39. Wang Z, Jiang W, Wang Y et al. MGMT promoter methylation in serum
functions of the choroidal blood-cerebrospinal fluid barrier in health and cerebrospinal fluid as a tumor-specific biomarker of glioma. Biomed
and disease. Acta Neuropathol 2018; 135(3): 337–361. Rep 2015; 3(4): 543–548.
18. Redzic Z. Molecular biology of the blood-brain and the blood- 40. Figueroa JM, Skog J, Akers J et al. Detection of wild-type EGFR amplifi-
cerebrospinal fluid barriers: similarities and differences. Fluids Barriers cation and EGFRvIII mutation in CSF-derived extracellular vesicles of
CNS 2011; 8(1): 3. glioblastoma patients. Neuro Oncol 2017; 19(11): 1494–1502.
19. Schmitt-Graff A, Hummel M, Anagnostopoulos I et al. [Primary brain 41. Tabouret E, Chinot O, Metellus P et al. Recent trends in epidemiology of
lymphoma in acquired immunodeficiency syndrome. brain metastases: an overview. Anticancer Res 2012; 32(11): 4655–4662.
Immunophenotype and molecular pathologic characterization in stereo- 42. Taillibert S, Le Rhun E. [Epidemiology of brain metastases]. Cancer
tactic biopsy, autopsy and cerebrospinal fluid cytology]. Pathologe 1995; Radiother 2015; 19(1): 3–9.
16: 75–80. 43. Stelzer KJ. Epidemiology and prognosis of brain metastases. Surg Neurol
20. Rhodes CH, Honsinger C, Sorenson GD. PCR-detection of tumor- Int 2013; 4(Suppl 4): S192–S202.
derived p53 DNA in cerebrospinal fluid. Am J Clin Pathol 1995; 103(4): 44. Paik PK, Shen R, Won H et al. Next-generation sequencing of stage IV
404–408. squamous cell lung cancers reveals an association of PI3K aberrations
21. Swinkels DW, de Kok JB, Hanselaar A et al. Early detection of leptomen- and evidence of clonal heterogeneity in patients with brain metastases.
ingeal metastasis by PCR examination of tumor-derived K-ras DNA in Cancer Discov 2015; 5(6): 610–621.
cerebrospinal fluid. Clin Chem 2000; 46(1): 132–133. 45. Chen G, Chakravarti N, Aardalen K et al. Molecular profiling of patient-
22. Shi W, Lv C, Qi J et al. Prognostic value of free DNA quantification in matched brain and extracranial melanoma metastases implicates the
serum and cerebrospinal fluid in glioma patients. J Mol Neurosci 2012; PI3K pathway as a therapeutic target. Clin Cancer Res 2014; 20(21):
46(3): 470–475. 5537–5546.
23. Yang H, Cai L, Zhang Y et al. Sensitive detection of EGFR mutations in 46. Rochet NM, Kottschade LA, Markovic SN. Vemurafenib for melanoma
cerebrospinal fluid from lung adenocarcinoma patients with brain meta- metastases to the brain. N Engl J Med 2011; 365(25): 2439–2441.
stases. J Mol Diagn 2014; 16(5): 558–563. 47. Welsh JW, Komaki R, Amini A et al. Phase II trial of erlotinib plus con-
24. Touat M, Duran-Pena A, Alentorn A et al. Emerging circulating bio- current whole-brain radiation therapy for patients with brain metastases
markers in glioblastoma: promises and challenges. Expert Rev Mol Diagn from non-small-cell lung cancer. J Clin Oncol 2013; 31(7): 895–902.
2015; 15(10): 1311–1323. 48. Seoane J, De Mattos-Arruda L. Brain metastasis: new opportunities to
25. Pan W, Gu W, Nagpal S et al. Brain tumor mutations detected in cerebral tackle therapeutic resistance. Mol Oncol 2014; 8(6): 1120–1131.
spinal fluid. Clin Chem 2015; 61(3): 514–522. 49. Long GV, Trefzer U, Davies MA et al. Dabrafenib in patients with
26. Wang Y, Springer S, Zhang M et al. Detection of tumor-derived DNA in Val600Glu or Val600Lys BRAF-mutant melanoma metastatic to the
cerebrospinal fluid of patients with primary tumors of the brain and spi- brain (BREAK-MB): a multicentre, open-label, phase 2 trial. Lancet
nal cord. Proc Natl Acad Sci USA 2015; 112(31): 9704–9709. Oncol 2012; 13(11): 1087–1095.
27. Pentsova EI, Shah RH, Tang J et al. Evaluating cancer of the central ner- 50. Margolin K, Ernstoff MS, Hamid O et al. Ipilimumab in patients with
vous system through next-generation sequencing of cerebrospinal fluid. J melanoma and brain metastases: an open-label, phase 2 trial. Lancet
Clin Oncol 2016; 34(20): 2404–2415. Oncol 2012; 13(5): 459–465.
28. Martinez-Ricarte F, Mayor R, Martinez-Saez E et al. Molecular diagnosis 51. Lin NU, Dieras V, Paul D et al. Multicenter phase II study of lapatinib in
of diffuse gliomas through sequencing of cell-free circulating tumor patients with brain metastases from HER2-positive breast cancer. Clin
DNA from cerebrospinal fluid. Clin Cancer Res 2018; 24: 2812–2819. Cancer Res 2009; 15(4): 1452–1459.
29. Mouliere F, Mair R, Chandrananda D et al. Detection of cell-free DNA 52. Bachelot T, Romieu G, Campone M et al. Lapatinib plus capecitabine in
fragmentation and copy number alterations in cerebrospinal fluid from patients with previously untreated brain metastases from HER2-positive
glioma patients. EMBO Mol Med 2018; 10: 12. metastatic breast cancer (LANDSCAPE): a single-group phase 2 study.
30. Stupp R, Mason WP, van den Bent MJ et al. Radiotherapy plus concomi- Lancet Oncol 2013; 14(1): 64–71.
tant and adjuvant temozolomide for glioblastoma. N Engl J Med 2005; 53. Gadgeel SM, Gandhi L, Riely GJ et al. Safety and activity of alectinib
352(10): 987–996. against systemic disease and brain metastases in patients with crizotinib-
31. Furnari FB, Fenton T, Bachoo RM et al. Malignant astrocytic glioma: resistant ALK-rearranged non-small-cell lung cancer (AF-002JG): results
genetics, biology, and paths to treatment. Genes Dev 2007; 21(21): from the dose-finding portion of a phase 1/2 study. Lancet Oncol 2014;
2683–2710. 15(10): 1119–1128.

Volume 30 | Issue 2 | 2019 doi:10.1093/annonc/mdy544 | 217

Review Annals of Oncology
54. Costa DB, Shaw AT, Ou SH et al. Clinical experience with crizotinib in 67. Russo M, Siravegna G, Blaszkowsky LS et al. Tumor heterogeneity and
patients with advanced ALK-rearranged non-small-cell lung cancer and lesion-specific response to targeted therapy in colorectal cancer. Cancer
brain metastases. J Clin Oncol 2015; 33(17): 1881–1888. Discov 2016; 6(2): 147–153.
55. Crino L, Ahn MJ, De Marinis F et al. Multicenter phase II study of 68. Bendell JC, Domchek SM, Burstein HJ et al. Central nervous system
whole-body and intracranial activity with ceritinib in patients with ALK- metastases in women who receive trastuzumab-based therapy for meta-
rearranged non-small-cell lung cancer previously treated with chemo- static breast carcinoma. Cancer 2003; 97(12): 2972–2977.
therapy and crizotinib: results from ASCEND-2. J Clin Oncol 2016; 69. Martin AM, Cagney DN, Catalano PJ et al. Brain metastases in newly
34(24): 2866–2873. diagnosed breast cancer: a population-based study. JAMA Oncol 2017;
56. Iuchi T, Shingyoji M, Sakaida T et al. Phase II trial of gefitinib alone 3(8): 1069–1077.
without radiation therapy for Japanese patients with brain metastases 70. Priedigkeit N, Hartmaier RJ, Chen Y et al. Intrinsic subtype switching
from EGFR-mutant lung adenocarcinoma. Lung Cancer 2013; 82(2): and acquired erbb2/her2 amplifications and mutations in breast cancer
282–287. brain metastases. JAMA Oncol 2016; 3(5): 666–671.
57. Porta R, Sanchez-Torres JM, Paz-Ares L et al. Brain metastases from lung 71. Remon J, Le Rhun E, Besse B. Leptomeningeal carcinomatosis in non-
cancer responding to erlotinib: the importance of EGFR mutation. Eur small cell lung cancer patients: a continuing challenge in the personalized
Respir J 2011; 37(3): 624–631. treatment era. Cancer Treat Rev 2017; 53: 128–137.
58. Schuler M, Wu YL, Hirsh V et al. First-line afatinib versus chemotherapy 72. Dudani S, Mazzarello S, Hilton J et al. Optimal management of lepto-
in patients with non-small cell lung cancer and common epidermal meningeal carcinomatosis in breast cancer patients—a systematic review.
growth factor receptor gene mutations and brain metastases. J Thorac Clin Breast Cancer 2016; 16(6): 456–470.
Oncol 2016; 11(3): 380–390. 73. Le Rhun E, Weller M, Brandsma D et al. EANO-ESMO clinical practice
59. Swain SM, Baselga J, Miles D et al. Incidence of central nervous system guidelines for diagnosis, treatment and follow-up of patients with lepto-
metastases in patients with HER2-positive metastatic breast cancer meningeal metastasis from solid tumours. Ann Oncol 2017; 28(Suppl 4):
treated with pertuzumab, trastuzumab, and docetaxel: results from the iv84–iv99.
randomized phase III study CLEOPATRA. Ann Oncol 2014; 25(6): 74. Siravegna G, Geuna E, Mussolin B et al. Genotyping tumor DNA in cere-
1116–1121. brospinal fluid and plasma in a HER2 positive breast cancer with brain
60. Bartsch R, Berghoff AS, Preusser M. Breast cancer brain metastases metastases. ESMO Open 2017; 2(4): e000253.
responding to primary systemic therapy with T-DM1. J Neurooncol 75. Taillibert S, Chamberlain MC. Leptomeningeal metastasis. Handb Clin
2014; 116(1): 205–206. Neurol 2018; 149: 169–204.
61. Davies MA, Saiag P, Robert C et al. Dabrafenib plus trametinib in 76. Ballard P, Yates JW, Yang Z et al. Preclinical comparison of osimertinib
patients with BRAFV600-mutant melanoma brain metastases (COMBI- with other EGFR-TKIs in EGFR-mutant NSCLC brain metastases mod-
MB): a multicentre, multicohort, open-label, phase 2 trial. Lancet Oncol els, and early evidence of clinical brain metastases activity. Clin Cancer
2017; 18(7): 863–873. Res 2016; 22(20): 5130–5140.
62. Berghoff AS, Preusser M. Targeted therapies for melanoma brain meta- 77. Yang JC-H, Cho BC, Kim D-W et al. Osimertinib for patients (pts) with
stases. Curr Treat Options Neurol 2017; 19(4): 13. leptomeningeal metastases (LM) from EGFR-mutant non-small cell lung
63. Chakravarty D, Gao J, Phillips S et al. OncoKB: a precision oncology cancer (NSCLC): updated results from the BLOOM study. J Clin Oncol
knowledge base. JCO Precision Oncol 2017; 1(1): 1–16. 2017; 35: 9022.
64. Jiang B-Y, LI Y, Chuai S et al. NGS to reveal heterogeneity between cere- 78. Wolpert F, Weller M, Berghoff AS et al. Diagnostic value of
brospinal fluid and plasma ctDNA among non-small cell lung cancer (18)F-fluordesoxyglucosepositron emission tomography for patients with
patients with leptomeningeal carcinomatosis. J Clin Oncol 2017; brain metastasis from unknown primary site. Eur J Cancer2018; 96: 64–72.
35(Suppl 15): 9022. 79. Le Rhun E, Bertrand N, Dumont A et al. Identification of single nucleo-
65. Spoerke JM, Gendreau S, Walter K et al. Heterogeneity and tide polymorphisms of the PI3K-AKT-mTOR pathway as a risk factor of
clinical significance of ESR1 mutations in ER-positive metastatic central nervous system metastasis in metastatic breast cancer. Eur J
breast cancer patients receiving fulvestrant. Nat Commun 2016; 7: Cancer 2017; 87: 189–198.
11579. 80. Tanaka S, Louis DN, Curry WT et al. Diagnostic and therapeutic avenues for
66. Fribbens C, O’Leary B, Kilburn L et al. Plasma ESR1 mutations and the glioblastoma: no longer a dead end? Nat Rev Clin Oncol 2013; 10(1): 14–26.
treatment of estrogen receptor-positive advanced breast cancer. J Clin 81. Dagogo-Jack I, Gill CM, Cahill DP et al. Treatment of brain metastases
Oncol 2016; 34(25): 2961–2968. in the modern genomic era. Pharmacol Ther 2017; 170: 64–72.

218 | Seoane et al. Volume 30 | Issue 2 | 2019