NEW YORK – Circulating tumor DNA testing may be a promising, less invasive method of diagnosing central nervous system lymphomas and for identifying patients who will fare worse before and during treatment, researchers reported at the American Society of Hematology's annual meeting Sunday.
In a study presented during a plenary session at the meeting, researchers led by Florian Scherer, a clinical scientist and primary investigator at the University Medical Center Freiburg's hematology and oncology division, explored the potential utility of ctDNA testing in several aspects of CNS lymphoma diagnosis and prognosis. Using samples from 67 CNS lymphoma patients, Scherer and colleagues showed that ctDNA is readily detectable in plasma and cerebral spinal fluid samples, and that it accurately mirrors tumor burden in plasma samples and reflects lymphoma localization in cerebral spinal fluid. Additionally, Scherer reported that ctDNA detectable in cerebral spinal and plasma samples before and during methotrexate-based treatment is a strong negative prognostic biomarker for patients.
Importantly, the researchers showcased the ability of ctDNA to diagnose patients with CNS lymphoma. They also proposed a diagnostic workflow involving ctDNA testing that, if adopted, may allow a large number of patients to avoid invasive biopsy procedures.
CNS lymphoma is a rare and aggressive type of non-Hodgkin lymphoma, where cancer cells grow in the lymph tissues of the brain or spinal cord. Around 30 percent of patients survive for five years after diagnosis. Standard treatments for CNS lymphoma include multimodal regimens that cross the blood-brain barrier such as the MATRix chemo-immunotherapy regimen, comprising methotrexate, cytarabine, thiotepa and rituximab (Genentech/Biogen Idec's Rituxan).
This regimen allows more than half of the patients treated to "gain long-term control," said Davide Rossi, deputy head of the hematology division at the Oncology Institute of Southern Switzerland, at the meeting. However, it is not possible to accurately predict which patients are likely to be cured with the MATRix regimen and which patients will relapse using MRI. "There are some complete responders [according to MRI] who will recur and some partial responders who will not," Rossi noted, citing data from the IELSG32 trial. "Therefore, new tools and metrics for residual disease assessment are needed in CNS lymphoma."
Currently, CNS lymphoma patients' performance status and age are the most important prognostic factors. Tumor load can also serve as a prognostic marker, though as Rossi noted, "quantifiable metrics" are lacking for CNS lymphoma.
Moreover, CNS lymphomas are diagnosed using MRI-guided stereotactic needle biopsies, but cancer cells are hard to access when they're deep in the brain. "The rapid vanishing of CNS lymphoma cells following steroid therapy [in imaging] … may lead to false-negative biopsies," Rossi said. "Such points may delay diagnosis, for months even, contributing to the decline of the performance status of the patient."
Diagnosis of these tumors are a major challenge, Scherer agreed, as it "still relies on invasive neuro-surgical procedures that are often inconclusive and associated with complications, especially in elderly and frail patients."
Since CNS lymphoma cells shed DNA fragments in the blood and cerebrospinal fluid, testing for ctDNA in these sample types can serve as a less invasive biomarker. However, the use of such testing has been limited in CNS lymphomas to date, Scherer said, because ctDNA concentrations tend to be extremely low in plasma.
As such, next-generation sequencing may entirely miss detection of this biomarker or pick it up in only a third of cancer at best. Moreover, while testing for the hotspot mutation MYD88 L265P can help with CNS lymphoma detection, Scherer characterized this type of single-gene testing to be insufficient since around 30 percent of patients do not harbor this mutation.
In the study presented at the meeting, Scherer's group tried to overcome current diagnostic challenges in CNS lymphoma by using CAPP-seq and PhasED-seq developed at Ash Alizadeh's lab at Stanford University. The 67 patients in this study had primary or isolated secondary CNS lymphomas, and the majority received treatment with curative-intent methotrexate-based therapies. Researchers tested for ctDNA in tumor, plasma and cerebrospinal fluid collected at diagnosis, during treatment, and at disease progression.
Using a CAPP-seq targeted sequencing panel that covers 199 kilobases and captures 580 genetic regions frequently mutated in CNS lymphomas, researchers evaluated genomic DNA in patients' tissue samples and ctDNA in cerebrospinal fluid and plasma samples. They were able to detect somatic mutations in 100 percent of patients' tumor samples with a median of 288 mutations per patient. Mutations in MYD88, PIM1 and CD79B occurred most frequently, with MYD88 L265P mutations occurring in 73 percent of patients.
Using PhasED-seq, which allowed ultrasensitive detection of ctDNA, researchers monitored tumor-derived mutations in pretreatment plasma and cerebrospinal fluid obtained at diagnosis or at progression. They detected ctDNA in 78 percent of pretreatment plasma and in 100 percent of cerebrospinal fluid, while maintaining specificity of 96 percent and 97 percent in these samples, respectively.
Scherer's group then explored if pretreatment ctDNA concentrations are associated with radiographic tumor volume assessed by MRI. They found that while ctDNA levels in plasma were significantly correlated with total radiographic tumor volumes, the same correlation wasn't there in cerebrospinal fluid. However, ctDNA concentrations were more than 80 times higher in cerebrospinal fluid when CNS lymphoma patients had periventricular involvement.
"These results indicate that the proximity of CNS lymphomas to the ventricular system is the crucial factor for ctDNA shedding into the cerebrospinal fluid, while tumor burden on the other hand seems to be the essential variable for ctDNA detection in the blood," Scherer observed.
The researchers then turned to the question of whether pretreatment plasma ctDNA is a prognostic biomarker. They found that 80 percent of patients with detectable ctDNA progressed within one year compared to 31 percent with negative ctDNA. Similarly, 71 percent with positive ctDNA prior to treatment died within two years. In comparison, 92 percent of patients with undetectable ctDNA pretreatment were alive two years after blood draw. These data show a "significantly inferior" progression-free and overall survival for ctDNA-positive patients compared to those with undetectable ctDNA, Scherer said.
Scherer and colleagues combined pretreatment ctDNA and radiographic tumor volume measurements to try to improve their ability to risk stratify patients. They found that patients with detectable pretreatment ctDNA and high tumor volumes by MRI tended to have particularly poor outcomes, while those with undetectable ctDNA and low tumor volumes had an exceptionally favorable outcome. In fact, all patients in the latter group were alive at two years.
Experiments to assess the prognostic value of ctDNA during or after treatment yielded similar results. Around three-fourths of patients who progressed on treatment had ctDNA detectable while they were receiving therapy compared to on-treatment ctDNA-positivity in 9 percent of patients who achieved complete responses. As such, ctDNA status during treatment is also "highly predictive" for progression-free and overall survival, Scherer noted.
Finally, researchers looked at whether CNS lymphomas could be diagnosed without the need for surgical procedures. Using mutational data from 30 CNS lymphoma tumors in this study and 2,647 non-CNS lymphomas from public databases, Scherer's group developed a classifier score that could be used as a noninvasive diagnostic for CNS lymphoma. They validated the classifier in an independent cohort of 183 cerebrospinal fluid and plasma samples using CAPP-seq.
The classifier correctly diagnosed CNS lymphomas in 59 percent of CSF samples and 25 percent of plasma samples. "We did not observe any false-positive results in our non-CNS lymphoma cohort, leading to 100 percent specificity and 100 percent positive predictive value," Scherer said, adding that the classifier also improved CNS lymphoma detection rates compared to just MYD88 L265P mutation testing.
Based on this, Scherer's team proposed a clinical diagnostic path where patients presenting clinically with neurological symptoms can undergo MRI evaluation of their suspected cerebral tumor lesions, but they would also undergo ctDNA sequencing of plasma and cerebrospinal fluid for brain cancer classification and CNS lymphoma staging for systemic therapy. Only patients that are not found to have CNS lymphoma by ctDNA testing would then have to undergo biopsy and standard diagnostic procedures.
"In the future, ctDNA could help stratify patients into risk groups that might either benefit from treatment reduction or treatment intensification," Scherer concluded. "Furthermore, we envision a potential role of ctDNA for surgery-free CNS lymphoma detection or even identification of occult CNS lymphoma involvement."