Real-world experience with circulating tumor DNA in cerebrospinal fluid from patients with central nervous system tumors

Authors

Hickman RA

Miller AM

Holle BM

Jee J

Liu SY

Ross D

Yu H

Riely GJ

Ombres C

Gewirtz AN

Reiner AS

Nandakumar S

Price A

Kaley TJ

Graham MS

Vanderbilt C

Rana S

Hill K

Chabot K

Campos C

Nafa K

Shukla N

Karajannis M

Li B

Berger M

Ladanyi M

Pentsova E

Boire A

Brannon AR

Bale T

Mellinghoff IK

Arcila ME

Doi

10.1186/s40478-024-01846-4

PMID: 39289779 · DOI: 10.1186/s40478-024-01846-4 · Journal: Acta Neuropathologica Communications (2024)

TL;DR

Hickman et al. report the real-world clinical implementation of prospective cerebrospinal fluid (CSF) circulating tumor DNA (ctDNA) sequencing using the FDA-authorized MSK-IMPACT assay at Memorial Sloan Kettering Cancer Center. They profiled 1,007 CSF samples from 711 adult and pediatric patients with over 90 distinct primary CNS and metastatic tumor types collected between November 2018 and November 2022. Genetic alterations were detected in 53.0% of CSF samples from patients with clinically documented CNS tumors (489/922), while none of 85 CSF samples from patients without CNS tumors were ctDNA-positive (specificity 100%). ctDNA positivity was associated with a three-fold increased risk of death (HR 3.23, 95% CI 2.58–4.05, p < 0.001), and the distribution of clinically actionable alterations mirrored tumor-type specific patterns in the AACR GENIE cohort PMID:39289779.

Cohort & data

  • Sample size: 1,007 CSF samples from 711 patients (adult and pediatric); 150/711 patients (21.1%) contributed multiple samples (median 2, range 2–12) PMID:39289779.
  • Cancer types: Over 90 distinct tumor types, including lung cancer (n=188), breast cancer (n=150) and gliomas (n=148) as the largest categories, plus metastatic melanoma, GI cancers, and embryonal tumors PMID:39289779.
  • Negative controls: 85/1,007 CSF samples from patients without clinically documented CNS involvement PMID:39289779.
  • Dataset: csf_msk_2024 PMID:39289779.
  • Assays: msk-impact-panel — IMPACT468 (n=274 samples) and IMPACT505 (n=733 samples) on cfDNA from CSF; a paired plasma subset (n=31 patients) was profiled with msk-access (ACCESS129) PMID:39289779.
  • Annotation: Variants annotated against oncokb for clinical actionability; level 1–4 and resistance levels (R1, R2) per the OncoKB schema PMID:39289779.
  • Collection period: November 2018 to November 2022; median follow-up 240 days (IQR 112–483 days) following CSF collection PMID:39289779.
  • Reference genome: GRCh37 csf_msk_2024.

Key findings

  • ctDNA detection rate: 53.0% (489/922) of CSF samples from patients with clinically documented CNS tumors harbored at least one somatic genetic alteration PMID:39289779.
  • Specificity: 100% (0/85 ctDNA-positive among patients without CNS tumors) PMID:39289779.
  • Total alterations: 7,110 somatic alterations across the 489 ctDNA-positive samples (3,944 non-synonymous mutations, 2,980 somatic copy number alterations, 186 structural rearrangements); median 4 mutations/sample (IQR 2–8, range 1–415) PMID:39289779.
  • Variant allele frequencies: Median VAF 38.7% (IQR 23.2–51.1%, range 1–100%), reflecting high tumor purity of CSF cfDNA PMID:39289779.
  • Tumor-type variation: GI cancers had the highest proportion of ctDNA-positive CSF samples and highest median VAF; embryonal tumors had the fewest PMID:39289779.
  • Clinical actionability: 50.7% (248/489) of ctDNA-positive samples carried a level 1 OncoKB actionable alteration; lung carcinomas had the highest level 1 actionability PMID:39289779.
  • Survival association: ctDNA positivity was associated with a three-fold increased risk of death (HR 3.23, 95% CI 2.58–4.05, p < 0.001). Median overall survival was 854 days shorter in ctDNA-positive patients (235 days, 95% CI 177–272 vs. 1,089 days, 95% CI 796–not reached) PMID:39289779.
  • Metastatic vs. primary CNS tumors: Samples from patients with CNS metastasis were more likely to be ctDNA-positive than those from primary CNS tumors (OR 2.60, 95% CI 1.96–3.46, p < 0.001) PMID:39289779.
  • Leptomeningeal disease (lung cancer): In NSCLC, leptomeningeal involvement strongly predicted ctDNA positivity (OR 20.17, 95% CI 9.65–42.16, p < 0.0001). CSF ctDNA outperformed cytology for leptomeningeal disease detection (sensitivity 85.4% vs. 61.7%; NPV 80% vs. 66%; specificity 78.7%; PPV 84.4%) PMID:39289779.
  • CSF vs. tumor tissue: Across 56 tumor/CSF pairs (55 patients), 999 alterations were detected — 43% (434/999) shared, 27% (273/999) private to CSF, 29% (292/999) private to tumor. For OncoKB level 1–3A alterations, 41/53 (77%) were shared. CSF VAFs were significantly higher than paired tumor (median 32% vs. 24%, p < 0.01, Mann–Whitney U). TMB in tumor and CSF correlated strongly (Spearman r = 0.81, p < 0.001) PMID:39289779.
  • CSF vs. plasma: Across 31 patients, 54% (77/142) of alterations were shared; 75% (24/32) of OncoKB level 1–3A alterations were shared. CSF VAFs were markedly higher than plasma (median 36.4% vs. 2.3%, p < 0.01) PMID:39289779.
  • Pre-analytic factors: Volumes <2 mL critically reduced ctDNA positivity (8.3% for primary CNS tumors, 20.0% for metastatic CNS tumors); positivity climbed to 37.8% / 68.6% at 5 mL with no further improvement above 10 mL. 99.2% of samples were processed within the 14-day Streck-tube stability window. Median sequencing coverage 65X; coverage correlated with cfDNA yield (r = 0.79, p < 0.001) PMID:39289779.
  • Mutational signatures: Of 35 CSF samples with TMB ≥ 13.8 mutations/Mb, dominant signatures included UV (n=3) in metastatic cutaneous melanoma, APOBEC (n=7) in breast cancer, smoking (n=8) in lung cancer, and temozolomide (n=2) in gliomas PMID:39289779.

Genes & alterations

  • TP53 — most frequently altered gene across all tumor types (n=242, 49% of 489 ctDNA-positive samples) PMID:39289779.
  • EGFR — sensitizing mutations and high-level amplification in lung cancer; serial CSF sequencing captured emergence of acquired-resistance gatekeeper and other mutations including p.T790M, p.C797S, p.L792H, p.L718Q, p.L718V, and p.G724S; EGFRvIII detected in gliomas PMID:39289779.
  • MET — amplification in lung cancer; exon 14 skipping mutations; resistance mutation p.Y1230N emerged on crizotinib (patient transitioned to capmatinib) PMID:39289779.
  • ERBB2 — mutation and amplification in lung cancer; amplification common in breast cancer PMID:39289779.
  • KRAS — mutations in lung cancer; off-target alteration on EGFR-targeted therapy PMID:39289779.
  • BRAF — p.V600E in metastatic melanoma; fusions as off-target resistance in EGFR-mutant lung cancer; BRAF::KIAA1549 fusions detected in gliomas PMID:39289779.
  • PIK3CA — p.E545K common in breast cancers; off-target alteration on EGFR-targeted therapy in lung cancer PMID:39289779.
  • ALK — EML4::ALK fusions in lung carcinomas; resistance mutations p.G1202R and p.G1269A detected on targeted therapy progression PMID:39289779.
  • EML4 — fusion partner in EML4::ALK rearrangements in lung carcinomas PMID:39289779.
  • RET — rearrangements with diverse gene partners in lung carcinomas PMID:39289779.
  • ROS1 — rearrangements with diverse gene partners in lung carcinomas PMID:39289779.
  • IDH1 — p.R132H in IDH-mutant gliomas PMID:39289779.
  • STK11 — mutations in lung cancer PMID:39289779.
  • KEAP1 — mutations in lung cancer PMID:39289779.
  • H3-3A — p.K28M histone mutation supporting diagnosis of DMG (diffuse midline glioma) per the 2021 WHO CNS Classification, identified by CSF ctDNA when biopsy tissue was insufficient PMID:39289779.
  • KIAA1549 — partner in BRAF::KIAA1549 fusions in gliomas PMID:39289779.

Clinical implications

  • CSF ctDNA sequencing via msk-impact-panel is feasible and clinically informative in a routine hospital setting, providing actionable cancer genotype information in a large fraction of patients with CNS tumors PMID:39289779.
  • For NSCLC patients, CSF ctDNA had greater sensitivity than cytology for leptomeningeal disease detection (85.4% vs. 61.7%), supporting use as a complementary or preferred diagnostic modality PMID:39289779.
  • Serial CSF ctDNA profiling captured clonal evolution and emergence of resistance mechanisms — EGFR gatekeeper mutations, ALK resistance mutations, and MET p.Y1230N — directly informing treatment changes (e.g., crizotinib → capmatinib) PMID:39289779.
  • ctDNA detection in CSF was associated with significantly shortened overall survival (HR 3.23) across nearly all tumor subtypes (except embryonal and GI cancers), suggesting prognostic utility of CSF ctDNA status PMID:39289779.
  • CSF liquid biopsy outperformed plasma for CNS cancer monitoring (median VAF 36.4% vs. 2.3% for shared mutations), reflecting the blood–brain barrier limiting tumor DNA shed into peripheral circulation; CSF is preferred for genomic monitoring of CNS-confined disease PMID:39289779.
  • Mutational signatures (UV in melanoma, smoking in lung, APOBEC in breast, temozolomide in gliomas) and tumor mutation burden derived from CSF ctDNA aided in determining or confirming the primary tumor site, particularly when biopsy was inconclusive PMID:39289779.
  • Pre-analytic recommendations: collect ≥5 mL CSF in Streck tubes; volumes <2 mL substantially reduce sensitivity; processing within 14 days preserves stability PMID:39289779.

Limitations & open questions

  • Sensitivity ceiling: Only 53% of CSF samples from patients with known CNS disease were ctDNA-positive; the 47% false-negative rate limits utility as a standalone diagnostic test PMID:39289779.
  • Volume dependence: Samples below 2 mL had ctDNA positivity rates as low as 8.3% for primary CNS tumors, creating a practical barrier when CSF volume is limited PMID:39289779.
  • Heterogeneous cohort: The study encompasses over 90 tumor types with variable ctDNA shedding propensities; disease-specific conclusions are limited for rarer subtypes PMID:39289779.
  • Treatment outcome data: While actionable alterations and resistance mutations are described, the study does not systematically report treatment decisions or outcomes driven by CSF ctDNA findings PMID:39289779.
  • Single-institution study: All samples from MSK; generalizability to settings without equivalent pre-analytic infrastructure and sequencing capabilities is uncertain PMID:39289779.
  • Resistance examples are illustrative: Clonal evolution and resistance monitoring findings are presented as case studies rather than systematic longitudinal analyses across the cohort PMID:39289779.
  • CSF vs. plasma comparison limited: Only 31 patients had contemporaneous CSF and plasma samples, restricted to genomic regions overlapping msk-impact-panel and msk-access PMID:39289779.
  • Authors’ caveats: They suggest more targeted assays (e.g., digital PCR on pre-capture NGS libraries) could confirm subthreshold variants, but recovery of sufficient cell-free DNA remains the rate-limiting step PMID:39289779.

Citations from this paper used in the wiki

  • “We identified genetic alterations in 489/922 (53.0%) CSF samples with clinically documented CNS tumors. None of 85 CSF samples from patients without CNS tumors had detectable ctDNA.” (Abstract) PMID:39289779
  • “TP53 was the most frequently altered gene across all tumor types (n = 242, 49% of 489 ctDNA + samples).” (Results, p. 4) PMID:39289779
  • “Across all cancer types, detection of a genetic alteration in the CSF was associated with a three-fold increased risk of death (HR: 3.23, 95% CI: 2.58–4.05, P < 0.001).” (Results, p. 7) PMID:39289779
  • “Across the 489 ctDNA + samples, 248 (50.7%) had a level 1 OncoKB actionable alteration.” (Results, p. 6) PMID:39289779
  • “patients with metastatic disease and leptomeningeal involvement were more likely to have positive CSF ctDNA than those with parenchymal brain metastases in the absence of leptomeningeal involvement (OR: 20.17; CI: 9.65–42.16; p < 0.0001)” (Results, p. 7) PMID:39289779
  • “detection of a genetic alteration in the CSF had greater sensitivity than positive cytology for the presence of leptomeningeal disease (sensitivity: 85.4% vs. 61.7%)” (Results, p. 7) PMID:39289779
  • “Compared to the alterations identified in plasma, mutations detected in CSF cfDNA were identified at significantly higher VAFs (CSF: median VAF = 36.4%, IQR = 34.3% vs. Plasma: median VAF = 2.3%, IQR = 10.7%, P < 0.01, Mann–Whitney U test)” (Results, p. 7) PMID:39289779
  • “MSK-IMPACT™ solid tumor assay, a custom hybridization capture-based panel targeting all coding regions of 468 (n = 274) or 505 (n = 733) genes” (Methods, p. 2) PMID:39289779

This page was processed by paper-compiler on 2026-05-01.