Circulating tumor DNA (ctDNA) is readily detectable in the plasma and cerebrospinal fluid of patients with central nervous system (CNS) lymphoma and is a strong prognostic biomarker for outcomes, a team of researchers from Germany and Stanford University reported at the 2021 American Society of Hematology (ASH) Annual Meeting & Exposition.¹

“We found that ctDNA seems to robustly facilitate the noninvasive identification of CNS lymphoma in a large fraction of these patients,” said Florian Scherer, MD, of University Medical Center Freiburg. The study, whose first authors were Jurik A. Mutter, MD, of University Medical Center Freiburg, and Stefan Alig, MD, of Stanford University, was a Plenary Session presentation and a “Best of ASH” abstract.

Florian Scherer, MD

Jurik A. Mutter, MD

Stefan Alig, MD

In patients with CNS lymphoma, ctDNA testing had a sensitivity of 78% in plasma and 100% in cerebrospinal fluid. In the plasma, it accurately mirrored tumor burden, whereas in the cerebrospinal fluid, it reflected lymphoma localization. Both before and during treatment, ctDNA concentration proved to be a strong prognostic biomarker, Dr. Scherer reported.

Challenges in Diagnosing CNS Lymphoma

As Dr. Scherer pointed out, the diagnosis and management of CNS lymphoma are made difficult by the heterogeneous outcomes after methotrexate-based therapy and the poor prognosis of patients who relapse. Tools for accurately risk-stratifying patients and predicting these outcomes are largely lacking, he said.

Moreover, diagnosis still relies on invasive procedures that are often inconclusive (due to inaccurate tissue targeting) and are associated with complications, especially in elderly or frail persons. Biomarkers that facilitate the identification of patients at high risk for disease progression and allow for biopsy-free detection of CNS lymphoma are greatly needed.

Although ctDNA in the plasma and cerebrospinal fluid has become established in diagnosing and monitoring many cancers, its role in CNS lymphoma has been limited for two main reasons: concentrations of ctDNA in the plasma are extremely low, and this leads to insufficient detection (0%–32%) by previous next-generation sequencing technologies; and single-gene methods can capture the hot spot mutation MYD88 L265P, but up to 30% of patients lack this mutation and go undetected.

Use of ctDNA

To overcome these limitations, the investigators used cancer personalized profiling by deep sequencing (CAPP-Seq) and phased variant enrichment and detection sequencing (PhasED-Seq) to evaluate 270 samples from tumor biopsies, plasma, and cerebrospinal fluid.² These samples came from 67 patients, mostly with diffuse large B-cell lymphoma being treated with curative intent, as well as 44 patients with other brain tumors and 24 healthy controls.

CAPP-Seq was developed at Stanford and optimized to identify 580 distinct genetic regions frequently mutated in CNS lymphoma. Investigators applied CAPP-Seq for genotyping of both the genomic DNA from the tumor tissue and ctDNA from the cerebrospinal fluid and plasma; they used PhasED-Seq for ctDNA monitoring during treatment, capitalizing on its ability to detect one affected DNA molecule out of 2 million. Concentrations of ctDNA were correlated with radiologic measures of tumor burden and tested for associations with clinical outcomes at distinct clinical time points.

Furthermore, the investigators developed a novel classifier to noninvasively distinguish CNS lymphomas from other CNS tumors based on their mutational landscapes in plasma and cerebrospinal fluid. To that end, they used supervised training of a machine learning approach from tumor whole-genome sequencing data and their own genotyping analyses, followed by independent validation.

Study Findings

“We started by profiling the mutational landscape of tumor biopsies using CAPP-Seq and were able to detect somatic mutations in 100% of tumor specimens, with a median of 288 mutations per patient,” Dr. Scherer said. “As expected, MYD88 and CD79B were the most frequently affected genes, and 73% of patients had hot spot MYD88 L265P mutations.”

Dr. Scherer continued: “We next monitored tumor-derived mutations in both pretreatment plasma and cerebrospinal fluid obtained at diagnosis or disease progression using PhasED-Seq and identified ctDNA in 78% of pretreatment plasma and 100% of cerebrospinal fluid samples, while maintaining high specificities (96%–97%),” he said. The investigators observed highly variable ctDNA concentrations in both compartments, with 80-fold higher levels (P < .0001) in cerebrospinal fluid vs plasma. 

The researchers then found an association between ctDNA concentrations in plasma (but not cerebrospinal fluid) and total radiographic tumor volume (P < .0001). Also, detection of ctDNA in the cerebrospinal fluid was largely influenced by periventricular localization of the tumor: an 80-fold higher concentration was found in patients with periventricular involvement.  

“This indicates that the proximity of the CNS lymphoma to the ventricular system is the crucial factor for ctDNA shedding into the cerebrospinal fluid, whereas tumor burden seems to be the essential variable for ctDNA detection in the blood,” Dr. Scherer explained.

Pretreatment ctDNA as a Prognostic Biomarker

Pretreatment ctDNA proved to be a reliable prognostic marker, associated with progression-free and overall survival. For pretreatment ctDNA-positive vs ctDNA-negative patients, outcomes and hazard ratios (HRs) were as follows:

• Disease progression within 1 year: 80% vs 31% (HR = 4.6; P < .001)
• Death within 2 years: 71% vs 8% (HR = 9.6; P = .0001).

 This difference was reflected in multivariable analysis, which found pretreatment ctDNA concentration to be the only variable besides Karnofsky index to be a statistically significant prognostic factor.

By combining pretreatment ctDNA concentrations with radiographic tumor volume, even finer stratification was possible. Patients with detectable ctDNA and high tumor volumes had particularly poor outcomes: 16% were progression-free at 1 year and 14% were alive at 2 years. In contrast, among patients with undetectable ctDNA and low tumor volume, 70% were progression-free and 100% were alive at 2 years.

Moreover, in an analysis of 49 samples whose plasma was analyzed during curative-intent treatment, ctDNA positivity during induction therapy was significantly associated with outcomes, both progression-free survival (HR = 6.2; P = .0002) and overall survival (HR = 7.9; P = .004). More than 75% of patients destined for disease progression had detectable ctDNA during treatment, vs 9% of patients without detectable ctDNA. “Consequently, the ctDNA status during treatment turned out to be highly
predictive,” he said.

We found that ctDNA seems to robustly facilitate the noninvasive identification of CNS lymphoma in a large fraction of these patients.

— Florian Scherer, MD

Tweet this quote

Diagnosing CNS Lymphoma Without Surgery

The final aspect of the team’s work was to determine whether CNS lymphomas could be diagnosed using the ctDNA mutational profiles, without the need for surgical intervention throughout the course of disease. Their analysis of 183 validation samples profiled by CAPP-Seq correctly classified CNS lymphomas in 59% of the cerebrospinal fluid samples and 25% of the plasma samples. No false-positive results were observed in the non–CNS-lymphoma cohort, leading to 100% specificity and 100% positive predictive value.

“Importantly, our classifier improved the detection rate substantially, compared with MYD88 detection alone,” Dr. Scherer said. “We therefore propose a potential clinical path in which noninvasive CNS lymphoma classification [based on ctDNA in plasma or cerebrospinal fluid] directly leads to further staging and systemic therapy. If the classifier predicts non–CNS-lymphoma, the patient follows conventional diagnostic procedures, including brain biopsy.”

Dr. Scherer concluded: “We foresee an important potential future role of ctDNA as a decision-making tool to guide treatment in patients with CNS lymphoma.” n

DISCLOSURE: Dr. Scherer has received research funding from Roche Sequencing Solutions.

REFERENCE

1. Mutter JA, Alig S, Lauer EM, et al: Profiling of circulating tumor DNA for noninvasive disease detection, risk stratification, and MRD monitoring in patients with CNS lymphoma. 2021 ASH Annual Meeting & Exhibition. Abstract 6. Presented December 12, 2021.

2. Kurtz DM, Soo J, Co Ting Keh L, et al: Enhanced detection of minimal residual disease by targeted sequencing of phased variants in circulating tumor DNA. Nat Biotechnol 39(12):1537-1547, 2021.