Association of Cerebrospinal Fluid Tumor DNA Genotyping With Survival Among Patients With Lung Adenocarcinoma and Central Nervous System Metastases

Key Points Question What is the association of molecular alterations in cerebrospinal fluid with clinical outcomes of patients with a diagnosis of lung adenocarcinoma and central nervous system metastases? Findings In this cohort study of 94 patients, next-generation sequencing of cerebrospinal fluid showed that 5 molecular subtypes were associated with survival; the molecular subtype with abundant genetic alterations was associated with the shortest survival and increased risk of death. Specifically, EGFR variants coaltered with CDK4, CDK6, MYC, and MET were associated with poor outcomes. Meaning Based on genetic profiling in cerebrospinal fluid, this population experienced heterogenous survival outcomes, and some specific molecular alterations were associated with prognosis; therefore, as liquid biopsy is performed of central nervous system metastases, cerebrospinal fluid may facilitate risk stratifying central nervous system metastases into appropriate outcomes.


Introduction
Owing to the improvement of systemic therapies for lung cancer, patients live longer, but the incidence of central nervous system (CNS) metastases also increases. Cerebrospinal fluid (CSF) has been proven better than plasma to reveal unique genetic profiling of intracranial metastases. Recent research has indicated that CSF could be used as a liquid biopsy of CNS metastases because CSF reveals unique genetic profiles of patients with non-small cell lung cancer and CNS metastases. [1][2][3] However, the clinical effect of the abundant genetic alterations in CSF remains elusive. Thus, the present study aims to explore the molecular alterations revealed in CSF samples and their associations with survival of patients with lung adenocarcinoma and CNS metastases.

Methods
This retrospective cohort study was conducted at Guangdong Lung Cancer Institute, Guangzhou, China, from July 1, 2016, to July 31, 2018, among 94 patients with late-stage lung adenocarcinoma and a diagnosis of CNS metastases (leptomeningeal metastases [LM] and brain metastases [BM]).
Leptomeningeal metastases were diagnosed based on tumor cells detected in CSF samples or detected by leptomeningeal enhancement in brain magnetic resonance imaging. Brain metastases were diagnosed based on metastatic lesions detected by brain magnetic resonance imaging. All patients provided written informed consent, and the study protocol was approved by the Research Ethics Committee of Guangdong Provincial People's Hospital. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.
All patients underwent lumbar puncture, and circulating tumor DNA was extracted from CSF samples and profiled by a 168-gene targeted next-generation sequencing panel; the preparation of CSF circulating tumor DNA and a next-generation sequencing library and sequencing data analysis were fully described in a previous study. 3 Survival was calculated from the day of diagnosis of CNS metastases until death or the last follow-up date. Survival status was classified as censored if a patient was unavailable for follow-up or survived beyond the last follow-up. Follow-up was administratively censored on December 15, 2018.

Statistical Analysis
Patients were divided into several groups using unsupervised hierarchical cluster analysis using R, version 3.3.1 (R Foundation for Statistical Computing). 4 The variant status of each gene was valued as 0 or 1, and genes with a frequency greater than 10% were included. Kaplan-Meier survival curves were used to evaluate survival. The association between survival and coaltered genes (detected frequency >5%) with EGFR (GenBank NG_007726) variants was determined by logistic regression analysis. The coaltered genes and the clinical characteristics of the patients were included in the Cox proportional hazards regression models, and hazard ratios (HRs) and 95% CIs were calculated to determine the survival difference. All P values were 2-sided and were considered statistically significant at P < .05. SPSS for Windows, version 22.0 (SPSS Inc) was used for all statistical analysis; Prism, version 8.0 (GraphPad Software) was used for making graphics.

Molecular Subtypes of Patients With Lung Adenocarcinoma With LM or BM Distinguished by CSF
The median survival of the entire cohort was 19.3 months (95% CI, 15.4-23.2 months). Unsupervised hierarchical cluster analysis was used to identify 5 major molecular subtypes associated with CSF samples obtained from patients with CNS metastases (subtype cluster 1, 9 cases; cluster II, 19 cases; cluster III, 29 cases; cluster IV, 11 cases; cluster V, 26 cases) (Figure 1).
The demographic and clinical characteristics of these molecular subtypes are shown in Table 1.

Specific Genetic Signatures of 5 Molecular Subtypes
Patients in clusters I, III, IV, and V all harbored EGFR variants, while most patients in cluster II carried ALK (GenBank NG_009445) fusion ( Figure 2B). The genetic profiles of patients in cluster I, who presented with the shortest median survival (7.5 months), were characterized by a high detection rate The vertical red lines indicate the 5 clusters.  (Figure 3).  Figure 2B).

Discussion
To our knowledge, this study of 94 patients with advanced lung adenocarcinoma with CSF samples profiled by next-generation sequencing represents the largest reported series to date, and for the first time, we defined 5 subgroups of CNS metastases characterized by different multiple alterations   that were associated with diverse outcomes. We also challenged the view that the coexistence of other variants with EGFR variants had the same association with the survival of patients with CNS metastases.
Extracranially, TP53 and somatic copy number variations were associated with lung cancer. 5,6 Concordant with previous findings, we first found that patients with CNS metastases characterized by TP53 and somatic copy number variations in CSF had poor survival. Cell cycle pathway alterations were frequently detected and may be one of the mechanisms in BM or LM. 7 Our study further verified that cell cycle-regulated genes were also identifiers of CNS metastases in patients with poorer survival. Likewise, MET copy number gain has been associated with resistance to gefitinib in LM, 8 but we did not see a statistically significant association between MET alterations and inferior survival in our study. Further investigation is needed in this regard.
We identified that a subgroup (cluster I) of patients with lung adenocarcinoma and CNS metastases experienced significantly poor survival associated with the molecular characterization of their CSF samples. This finding supports the heterogeneous response to targeted therapies that is possibly due to different genetic backgrounds. The optimal strategies for this subgroup of patients with a poor prognosis are challenging, and more radical therapies-such as a tyrosine kinase inhibitor plus chemotherapy, antiangiogenic drugs, or preventive local therapy-may be taken into consideration, 9,10 but they need further verification.

Limitations
Our study has some limitations, including its retrospective design. In addition, treatment options based on each molecular subtype were not prospectively explored, so we could not identify the optimal treatments for patients in different clusters.

Conclusions
Our results suggest that genetic alterations in CSF were associated with the survival of patients with advanced lung adenocarcinoma and CNS metastases. In those patients with EGFR variants, coalterations with CDK4, CDK6, and MYC were associated with poor outcomes. Thus, the use of CSF samples may facilitate risk stratifying CNS metastases into appropriate outcomes and provide a reference for further clinical study.