Background
Brain metastases occur in nearly 30–40% of lung cancer patients and always result in poor clinical outcomes [
1]. The treatment strategies for lung cancer patients have become increasingly precise due to evaluations of molecular tumor characteristics. Thus, identifying actionable alterations and choosing optimal treatments have become increasingly important [
2]. Accumulating evidence has revealed the genetic heterogeneity and different clinical responses between primary lung tumors and metastatic brain tumors [
3‐
5]. As obtaining intracranial tumor samples is invasive and risky, there is an urgent need to identify alternative biopsy types to evaluate the genetic landscape and tumor evolution of central nervous system (CNS) tumors.
Circulating tumor DNA (ctDNA) is a promising biomarker to characterize the mutational profiles of tumors. Through real-time dynamic monitoring, ctDNA can also be used to trace tumor evolution and treatment responses, and identify resistance mechanisms [
5,
6]. However, due to the blood-brain barrier, plasma is not an ideal source to evaluate the genetic characteristics of CNS tumors [
7]. Thus, cerebrospinal fluid (CSF), which can be collected through a minimally invasive lumbar puncture procedure, is becoming an appealing surrogate as it is contained throughout the CNS and is in contact with intracranial lesions [
8]. Recent studies have shown that CSF-ctDNA represents the genetic alterations of brain tumors better than plasma, and also represents the changes in brain tumor burden [
9]. However, most current studies included very few samples and were retrospectively designed. No study has reported the use of CSF ctDNA to predict intracranial responses or to monitor mutational evolution during treatment.
We conducted this prospective cohort study to compare the genetic landscapes of paired CSF and plasma, and sought to trace intracranial tumor evolution and treatment responses through serial ctDNA sequencing of paired CSF and plasma samples in non-small cell lung cancer (NSCLC) patients with brain metastases.
Methods
Study design and participants
This prospective study (ClinicalTrials.gov identifier: NCT 03257735) was designed and performed at the Sun Yat-sen University Cancer Center (Guangzhou, China). Main inclusion criteria included following: (1) NSCLC newly diagnosed by histopathology; (2) brain metastases confirmed by enhanced brain MRI at primary diagnosis, with at least one intracranial lesion whose longest diameter was > 5 mm; (3) treatment-naïve (no previous systemic therapy, surgery, or radiotherapy); (4) brain metastases were asymptomatic or responding to corticosteroid treatment, followed by systemic therapy as the first-line treatment; and (5) had no contraindication for lumbar puncture. All patients meeting the inclusion criteria were enrolled consecutively between January 2017 and December 2020 at our institution. For first-line treatment, patients received systemic therapies based on their mutational characteristics. Paired CSF and blood samples with or without primary extracranial tumor samples were collected at baseline (before treatment), 8 weeks after the initiation of first-line treatment, and at the time of disease progression. At each time point, 5 ml of CSF and 8 ml of blood were collected, and radiographic evaluations were performed within one week (i.e., enhanced computerized tomography scans for extracranial lesions and enhanced magnetic resonance imaging for intracranial lesions). Radiographic evaluations were then performed at approximately 8-week intervals until tumor progression. The clinical response was assessed according to the Response Evaluation Criteria in Solid Tumors (RECIST), version 1.1 by a radiologist who was blinded to the patients’ clinical information. This study was approved by the ethics committee of the Guangdong Association Study of Thoracic Oncology (GASTO ID:1028, Approval No. A2017-003). All patients provided written informed consent to participate in the study and provide samples for tumor genetic profiling.
Next-generation sequencing and data processing
CSF and whole blood samples were collected in cell-free DNA BCT tubes (Streck Inc., La Vista, NE, USA). Within 24 h of CSF and blood collection, the cellular fraction was removed by two-step centrifugation at 4°C (1900 g for 10 min within 2 h of collection and 16,000 g for 10 min). The white blood cells were used for genomic DNA extraction (DNeasy Blood & Tissue Kit, Qiagen) as the germline controls. Samples were stored at -80°C until further processing. CtDNA was extracted using the QIAamp Circulating Nucleic Acid Kit (Qiagen), as previously reported [
10], and analyzed using comprehensive genomic profiling of 425 cancer-related genes in a central testing laboratory (Nanjing Geneseeq Technology, Jiangsu, China), as previously described [
10‐
13]. For patients with extracranial tumor tissues, genomic DNA was also extracted from tissue biopsy samples using the QIAamp DNA FFPE Tissue Kit (Qiagen) and subjected to next-generation sequencing with the same panel described above. Sequencing was performed on the Illumina HiSeq4000 platform and data analysis was performed as previously described [
10‐
12]. In brief, sequencing data were analyzed by Trimmomatic [
14] to remove low-quality (quality < 15) or N bases, and then mapped to the human reference genome hg19 using the Burrows-Wheeler Aligner (
https://github.com/lh3/bwa/tree/master/bwakit). PCR duplicates were removed by Picard (available at:
https://broadinstitute.github.io/picard/). The Genome Analysis Toolkit (GATK) (
https://software.broadinstitute.org/gatk/) was used to perform local realignments around indels and base quality reassurance. Single nucleotide polymorphisms (SNPs) and indels were analyzed by VarScan2 [
15] and Haplotype Caller/Uni edGenotyper in GATK, with the mutant allele frequency (MAF) cutoff of 0.2% for cfDNA samples, and a minimum of three unique mutant reads. Common SNPs were excluded if they were present in > 1% population frequency in the 1000 Genomes Project or the Exome Aggregation Consortium (ExAC) 65,000 exomes database. The resulting mutation list was further filtered by an in-house list of recurrent artifacts based on a normal pool of whole blood samples. Gene fusions were identified by FACTERA [
10,
16‐
18]. The medium depth of coverage after the removal of PCR duplicates was > 2000× and > 500× for liquid biopsies and tissue samples, respectively.
The ctDNA levels were calculated as previously described [
19,
20]: maxVAF = max variant allele frequency; ctDNA concentration (hGE/mL) = mean ctDNA VAF * cell-free DNA concentration (pg/mL) / 3.3, which assumes that each haploid genomic equivalent (hGE) weighed 3.3 pg.
Analysis of the consistency of genetic mutations
To evaluate the mutations that were consistently identified in the paired samples (sample A and sample B), the overlapping genes and unique genes from each sample were illustrated using a Venn diagram. Consistency was calculated using the following equation:
$$\mathrm{Consistency of sample A to sample B }\left(\%\right) = \frac{A \cap B}{A}\times 100\%$$
Consistency was analyzed for all mutant-positive samples. When considering the factors of the somatic mutations and driver gene variants, consistency was analyzed after removing the copy number variants (CNVs) or non-driver gene variants from the samples.
Mutation clonality analysis
At baseline, a mutation was considered clonal if its VAF was more than 25% of the maxVAF, and it was defined as subclonal if it was below this threshold [
21]. Newly acquired mutations in post-treatment samples were always defined as subclonal mutations.
Statistical analysis
Survival data were analyzed using Kaplan-Meier curves and Cox proportional hazard regression analyses. Comparisons of the continuous variables were made using an unpaired two-tailed t-test, while comparisons of the categorical variables between groups were performed using the chi-squared test or Fisher’s exact test. The concordance of genomic alterations between CSF and plasma, and concordance of dynamic changes in ctDNA and radiographic response was assessed using Cohen’s kappa coefficient. All statistical analyses were performed using R software (version 3.5.3). Two-tailed p-values of less than 0.05 were considered statistically significant.
Discussion
To the best of our knowledge, this is the largest prospective study investigating the genetic landscape of CSF ctDNA vs plasma ctDNA in NSCLC patients with brain metastases. This study is also the first to demonstrate that dynamic changes in CSF ctDNA could better predict intracranial tumor response than plasma, and can be used to monitor intracranial mutational evolution during treatment.
The presence of the blood-brain barrier prevents the release of ctDNA from intracranial tumors to peripheral plasma, and thus, CSF was considered a promising surrogate to represent the genetic landscape of intracranial lesions. In this study, the ctDNA-positivity rate was 63.7% in baseline CSF samples, which was also consistent with previous studies [
26,
27], and may be related to the tumor burden and the distance of brain lesion to its nearest ventricle. More importantly, we found that ctDNA-positivity in baseline CSF samples was an adverse prognostic factor for NSCLC patients with brain metastases, even after adjusting for gene type and other characteristics. Alexandra et al. also reported that ctDNA-positivity in baseline CSF samples was related to high disease burden and poor outcomes in glioma patients [
28]. Our results indicated that the presence of ctDNA in baseline CSF may be an early indicator of outcomes in NSCLC patients with brain metastases.
Patients with brain metastases harbored different genetic alterations in brain tumors vs primary tumors and exhibited different treatment responses with poor clinical outcomes [
29]. In our previous study, we revealed that driver mutations in metastatic brain tumors were highly concordant with paired primary lung tumors, while genetic alterations in cell cycle regulator genes and the PI3K pathway genes were enriched in brain metastases [
30]. Due to the invasiveness of obtaining brain tumor samples, we hypothesized that paired CSF and plasma samples could be used to reflect the genetic profiles of brain metastatic tumors and primary tumors, respectively. In this study, we found a high degree of genetic divergence between the CSF and paired plasma samples, while concordance in driver mutations was observed. A higher number of unique CNVs was also identified in CSF ctDNA than in plasma. A previous study also found that CNVs were enriched in the CSF from leptomeningeal metastases in NSCLC patients [
31], thus, indicating that CNVs may contribute to brain metastases in lung cancer patients. The identification of genetic alterations in brain metastases using CSF could facilitate precise treatments for such patients.
One of the advantages of liquid biopsies is that they can facilitate real-time tracking of tumor evolution and the prediction of treatment responses. Dynamic changes in the ctDNA in plasma have been reported to be related to tumor response [
32‐
34]. However, the genetic profiles of plasma ctDNA poorly represent those of brain tumors, and thus, the use of CSF to assess intracranial responses should be explored.
Our study was the first prospective study to reveal that the dynamic changes in CSF ctDNA after treatment could better predict intracranial tumor response than plasma. Patients with CSF ctDNA response 8 weeks after treatments had significantly longer intracranial PFS than patients without CSF ctDNA response. Dynamic changes in CSF ctDNA had better concordance with radiographic intracranial responses than plasma. Notably, when patients had opposing intracranial and extracranial tumor responses, the changes in CSF and plasma ctDNA followed the same trends as the intracranial and extracranial tumor burden, respectively. We then tracked the mutational evolution of CSF and plasma during treatment, and found that CSF harbored more subclonal mutations after treatment than plasma, thus, indicating different evolutionary patterns. The clonal mutations that remained in more than 80% in CSF after 8 weeks also predicted a shorter intracranial PFS. Collectively, our study revealed that dynamic monitoring of CSF ctDNA could be used to predict intracranial tumor response and track the evolution of brain tumors.
This study also had some limitations. First, we could not compare the genetic landscapes between CSF and brain tumor tissues due to the invasiveness of obtaining brain lesion samples. Thus, the unique somatic mutations and CNVs detected in CSF ctDNA might reflect the distinct genomic profiles of metastatic brain tumors. However, we are also aware of the possibility that alterations only detected in CSF may not be detected in plasma samples due to high background noise and varying ctDNA fractions. Second, due to the relatively low level of ctDNA-positivity in CSF, the sample size of patients with serial CSF sampling was small. Finally, though patients meeting the inclusion criteria were enrolled consecutively, we restricted our cohort to NSCLC patients with stable brain metastases. Thus, further exploration using large cohorts of patients with advanced CNS involvement should be warranted in the future.
Conclusions
In conclusion, our study supported the use of CSF as a liquid biopsy for brain tumors, and investigated the unique genetic profiles of metastatic brain tumors by sequencing CSF ctDNA. CSF ctDNA-positivity at baseline may be an early indicator of adverse outcomes in NSCLC patients with brain metastases. More importantly, dynamic changes in CSF ctDNA could better predict intracranial tumor response than plasma, and track intracranial clonal evolution during treatment.
Open AccessThis article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit
http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (
http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.