Rapid tumor DNA analysis of cerebrospinal fluid accelerates treatment of central nervous system lymphoma Free

The sensitivity of traditional cerebrospinal fluid (CSF) liquid-based assays for diagnosing central nervous system (CNS) neoplasms, such as CNS lymphoma, ranges from 13.3% for cytology to 23.3% for flow cytometry and 26.9% for IgH gene rearrangement (supplemental Table 1, available on the Blood website).¹ CNS tumor diagnosis thus frequently necessitates invasive neurosurgical biopsy.^2-4 Although histomolecular tissue analysis remains the gold standard, diagnostic yields and risks depend on lesion characteristics, patient comorbidities, and neurosurgeon experience.^5-7 

CSF-based approaches have been proposed to enable less invasive CNS tumor diagnosis. Digital droplet polymerase chain reaction (PCR) offers increased sensitivity, but examines a limited number of variants and requires manual threshold determination.⁸ Tumor-specific DNA alterations have been detected by next-generation sequencing (NGS) of CSF-derived DNA.^9-11 Studies of CSF from patients with known CNS lymphoma showed enhanced sensitivity by incorporating machine learning–assisted circulating tumor DNA sequencing,¹² clonotypic immunoglobulin gene rearrangement detection,¹³ and characterization of multiple biomarkers.¹⁴ Residual circulating tumor DNA in the peripheral blood after induction treatment for CNS lymphoma has also shown prognostic value that may guide personalized therapy.¹⁵ However, current turnaround times hinder the utility of these approaches because most new CNS neoplasms are diagnosed via emergency admissions, necessitating urgent diagnostic information.¹⁶ 

We previously described a rapid CSF genomic analysis technique, targeted rapid sequencing (TetRS),^2,17,18 consisting of quantitative PCR–based parallel detection of highly recurrent somatic mutations defining CNS neoplasms in the TERT promoter, MYD88, IDH1/2, H3F3A, and BRAF genes within 80 minutes of specimen acquisition, with detection limits of 0.15% variant allele fraction in tissue.² Here, we prospectively implemented rapid genotyping of CSF-derived DNA in suspected CNS neoplasms and analyzed the impact of this approach on diagnosis and treatment.

The prospective cohort in this study comprised patients at least 18 years old admitted to Massachusetts General Hospital with new CNS lesions of unknown cause for which neoplasm was suspected. Excess CSF that would otherwise be discarded was collected from clinically indicated procedures between 1 December 2020 and 10 December 2021; a total of 70 patients meeting inclusion criteria were designated the rapid genotyping cohort.

Informed consent was obtained under an institutional review board–approved protocol. Consistent with US Food and Drug Administration guidance for returning potentially actionable research findings,¹⁹ results were reported directly to treating physicians, who conducted all communication of findings via shared decision-making with patients. We evaluated real-time TetRS result integration by analyzing documented decision-making and clinical trajectories, and comparing these with historical cohorts.

To assess somatic alteration prevalence, genotyping was performed retrospectively from archived specimens of patients with primary CNS lymphoma (PCNSL) and other brain tumor specimens (supplemental Methods and supplemental Figure 1).

TetRS was used to detect hallmark mutations in the TERT promoter, MYD88, IDH1/2, H3F3A, and BRAF genes. Multiplexed GAPDH detection was an internal control for reaction integrity. Assays were performed separately on supernatant and pelleted CSF fractions.^2,17,18 Assay modifications, protocols, and validation are described in the supplemental Methods.

Written informed consent was obtained, and specimens were prospectively collected and analyzed under Massachusetts General Hospital (MGH) Institutional Review Board (IRB) protocol 2017P001581. The IRBs at MGH, Brigham and Women's Hospital, and Dana-Farber Cancer Institute approved analyses of electronic health records under MGH IRB protocol 2019P002400. The Dana-Farber Cancer Institute protocol 10-454 covered analysis of archived clinical specimens and genomic DNA extracts.

We first detailed the diverse diagnoses and historical variability in diagnostic trajectories among patients with newly suspected CNS neoplasms at our multidisciplinary referral center (historical cohort, Figure 1A). Although diagnostic evaluation was ongoing, empiric treatment regimens were initiated in 40 of 62 patients (64.5%). Median time to diagnosis was 9 days, with lymphoma diagnosed in 10 of 62 patients (16.1%), nonhematologic malignancies in 11 of 62 patients (17.7%), and no diagnosis achieved in 23 of 62 patients (37.1%) (supplemental Figure 2).

We then identified patients who established new diagnoses of CNS neoplasms before TetRS assay introduction (cohorts PCNSL and Center for Integrated Diagnostics [CID], Figure 1A), and analyzed genotyping findings alongside clinical trajectories to determine whether detection of variants in the TetRS panel from CSF could have facilitated diagnosis or treatment. Accounting for variant frequency and lumbar puncture eligibility, we found potential benefit of CSF genotyping for facilitating diagnosis or treatment in 38 of 72 patients (52.8%) (25/51 [49.0%] with PCNSL and 13/21 [61.9%] with CID, Figure 1A).

We then performed a prospective, single-center study involving hospitalized patients with new CNS lesions of unknown cause for which neoplasm was in the differential diagnosis. Excess CSF from 88 patients undergoing CSF sampling was submitted by providers for rapid genotyping in clinical real time; 70 patients meeting inclusion criteria comprised the rapid genotyping cohort (Figure 1B-C).

In the rapid genotyping cohort, detection of a mutation by TetRS shortened turnaround time of diagnostic information compared with conventional assays (median [interquartile range], 0.5 [0.5-not estimable (NE)] vs 25 [11-34] days, respectively; P < .001) (supplemental Figure 4). TetRS identified a mutation in 14 of 33 patients (42.4%) who were ultimately diagnosed with a neoplasm, including primary and secondary CNS lymphoma, glioblastoma, IDH-mutant brainstem glioma, and H3K27M-mutant diffuse midline glioma. No false-positive detection in 37 patients with nonneoplastic diagnoses was encountered (Figure 1C). In 6 of 33 patients (18.2%) diagnosed with a neoplasm, diagnoses were secured by TetRS results alone (Figure 2A-B). A number needed to diagnose (inverse of the Youden index)²⁰ analysis revealed that the number of patients needed to undergo a diagnostic assay to accurately detect a hematologic CNS malignancy was 1.18 by neurosurgical biopsy, 1.74 by TetRS, and 4.74 to 15.9 by conventional diagnostics (supplemental Table 1).

Although maximal safe neurosurgical resection may confer benefit for glioma and metastases, biopsy alone is indicated for hematologic malignancies, which may alternatively be diagnosed by CSF or vitreous studies and are initially treated with chemotherapy.^2-4,21 TetRS influenced adjuvant treatment initiation in 9 of 14 cases (64.3%) where a mutation was prospectively detected in the rapid genotyping cohort. This included eliminating the need for biopsies in 7 of 14 cases (50%) of neoplasms following tumor board discussion and shared decision-making (Figure 2C-D). In cases where a variant was detected by TetRS, treatment initiation was significantly accelerated by obviating surgical biopsies (median [interquartile range], 3 [3-NE] vs 12 [6-NE] days, respectively; P = .027) (supplemental Figure 4). No consistent differences were observed in lesional anatomy, clinical presentation, or conventional CSF-based assay performance between patients offered biopsy compared with patients in whom biopsy was not pursued (supplemental Figure 4; supplemental Results).

Recognizing this potential to impact clinical workflows in CNS lymphoma, MYD88 L265P detection from CSF was validated in a Clinical Laboratory Improvement Amendments environment, and 66 CSF specimens were then analyzed over 210 days from 4 clinical sites. MYD88 L265P was detected in 5 of 66 cases (7.6%) (Figure 3A), consistent with published estimates of 7.7% MYD88 L265P-mutant CNS lymphoma when lymphoma is suspected (77 MYD88-L265P mutated PCNSL cases of 1007 radiographically suspected cases²; P = .960), and 8 of 70 cases (11.4%) in the rapid genotyping cohort (P = .445) (Figure 3B). Median result reporting time was 2 days (range, 0-6 days).

Rapid genotyping of CSF can potentially accelerate CNS neoplasm workup by characterizing hallmark molecular features²² within 80 minutes, using universally available infrastructure in clinical molecular pathology laboratories.^2,23 Although rapid, targeted information cannot subtype lymphoma or glioma, it can focus differential diagnoses, guide surgical decisions, and inform treatment in selected cases. Detecting MYD88 L265P raises suspicion for lymphoproliferative or plasmacytic processes,^24,25 whereas negative results are counterbalanced by parallel detection of mutations indicative of common alternate diagnoses; for instance, identifying TERT promoter alterations in the absence of IDH mutations suggests glioblastoma.²⁶ 

TetRS guides surgical planning for biopsy vs resection, particularly when the differential diagnosis includes lymphoma vs glioma (case RG026; supplemental Results) or if tissue analysis is required for glioma classification,²² lymphoma subtyping, or confirmation of a neoplastic diagnosis. Our findings that 37 of 70 patients (52.9%) in the rapid genotyping cohort with suspected neoplasms ultimately had nonneoplastic or indeterminate diagnoses also highlights that resource-intensive sequencing methods may not be cost-effective upfront diagnostic approaches¹²; importantly, however, DNA isolated in the TetRS workflow is suitable for NGS analysis (case RG041; supplemental Figure 10), permitting a step-wise workflow.

Clinical integration of this approach enables critical subsequent studies. First, TetRS may improve cost-effectiveness by uniquely offering rapid turnaround from a Clinical Laboratory Improvement Amendments–certified assay using commercially available reagents and publicly available methods at <$9 per reaction²; comparably scalable quantitative PCR–based diagnostics have been deployed in resource-limited settings.²⁷ We observed that costs of incorporating rapid genotyping were offset by reductions in other CSF-based assays for suspected nonneoplastic causes (supplemental Figure 9). Second, clonal variants detected by TetRS may constitute treatment response, minimal residual disease, and recurrence biomarkers, such as MYD88 L265P correlating with disease burden (Figure 2D).

Larger studies are needed to determine the clinical impact of earlier therapy initiation, and whether our findings in a limited number of patients with CNS lymphoma are broadly applicable. We observed high specificity of this approach (supplemental Table 1), with 42% sensitivity for CNS tumor detection. Although adding more tumor-defining variants to this panel may increase sensitivity and ability to subclassify selected tumors, there may be inherent limitations of CSF genotyping by lumbar puncture. For instance, NGS-based analysis of CSF has 74% to 82.5% circulating tumor DNA (ctDNA) detection rates when CSF is sampled from intracranial compartments, but 49% to 63% detection from lumbar puncture. Tumor burden, grade, and location also correlate with ctDNA detectability.^9-11,28 

Although cases in which brain biopsies were obviated by CSF genotyping were associated with shorter times to treatment initiation (supplemental Figure 4), multiple factors influence treatment timing, because of the heterogeneity and complexity of patients with CNS tumors. Because conventional CSF flow cytometry and cytomorphologic assessment are standards accepted in hematologic malignancy diagnosis and monitoring, the impact of our approach compared with conventional assays is most apparent in patients with CNS lymphoma.

We prospectively demonstrate that targeted hallmark genomic alteration detection from CSF influences oncologic management of patients with suspected CNS tumors. Unlike prior studies demonstrating the theoretical promise of NGS-based detection of somatic alterations from CSF ctDNA among patients with known neoplasms,^2,8,9,11,28 we examined patients without established diagnoses. This cohort is more representative of the challenges unique to securing a diagnosis of a hematologic malignancy affecting the CNS to guide timely treatment. To our knowledge, our data provide the first comparisons between sequencing-based and conventional methods, prospectively calculated performance benchmarks, and multicenter results after Clinical Laboratory Improvement Amendments certification.

The authors deeply appreciate the collaborative support for these studies by many parties at the Massachusetts General Hospital (MGH) and Brigham and Women’s Hospital, including the following: Departments of Neuro-Oncology and Neurology clinical providers, nurses, and staff; Department of Pathology tissue banks; and clinical core laboratories. This work would not have been possible without the research community at the MGH Brain Tumor Research Center. The authors additionally thank Donald Glazer for critical review of the manuscript.

This work was supported by grants from the National Institutes of Health (NIH), National Institute of Neurological Disorders and Stroke 1K08NS107634-01A1 (to G.M.S.), The Darwin Project (to G.M.S.), the Brian D. Silber Memorial Fund (to G.M.S.), and the NIH, National Cancer Institute 5U01CA230697 (to B.S.C.).

Contribution: G.M.S., B.S.C., D.P.C., T.K., J.D., M.M., and J.K.L. designed and supervised the study; M.G., J.D.B., E.M., E.J.B., N.Z.G., G.E.M., A. Gallagher, and G.M.S. acquired, analyzed, and interpreted data; M.G., E.M., and G.M.S. performed statistical analysis; G.M.S., B.S.C., D.P.C., and J.K.L. provided administrative, technical, and material support; N.N., J.T., P.K.B., and K.T. contributed vital reagents; M.G., J.D.B., E.M., E.J.B., F.I., D.P.C., J.K.L., B.S.C., and G.M.S. wrote the manuscript; J.M.B., A. Gordon, S.S.J., M.P., Y.S., P.S.J., F.G.B., W.T.C., R.G., J.M.R., N.W., M.M.-L., D.A.F., B.V.N., T.T.B., L.L.R., J.T.J., D.P.C., J.K.L., B.S.C., and G.M.S. identified patients for study enrollment and analyzed and interpreted clinical data for inclusion in the manuscript; G.M.S. had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis; and all authors participated in critically revising the manuscript for important intellectual content.

Conflict-of-interest disclosure: D.P.C. and G.M.S. hold a patent pertaining to the use of peptide and locked nucleic acids as used in this study during the amplification of target DNA sequences (patent number US20170369939A1). Outside the scope of this work, the authors report the following. R.G. serves as a paid consultant for Idorsia, Inc, and Mosaic Research Management; is an advisory board member of Braintale, Inc, and the University of Wisconsin’s computed tomography board; provided expert testimony to Mary Hitchcock Memorial Hospital; has received speaker honoraria from Siemens; provided legal consulting to the Harvard Medical Institutions Risk Management Foundation; and received research funding from Samsung and the US National Institutes of Health. N.W. has received royalties from Wolters Kluwer and research funding from Merck, and was previously on the advisory board of Seattle Genetics (ended 2021). P.K.B. has consulted for Merck, Genentech-Roche, Eli Lilly, Tesaro, ElevateBio, Dantari, Voyager Therapeutics, SK Life Sciences, Pfizer, ACI, Atavistik, Sintetica, Kazia, MPM, CraniUS, Axiom, and InCephalo; serves on the Scientific Advisory Board for Kazia and CraniUS; has received grant/research support (to institution) from Merck, GlaxoSmithKline, AstraZeneca, Kazia, Genentech-Roche, Eli Lilly, Mirati, Bristol Myers Squibb, and Kinnate; and has received speaker’s honoraria from Merck, Genentech-Roche, Eli Lilly, and Medscape. D.A.F. is a shareholder for Eli Lilly. B.V.N. has consulted for Robeaute, BK Medical, Merck, BrainLab, Zeta, and Enclear. He has received grant funding from the US National Institutes of Health. T.T.B. has received publishing royalties from UpToDate, Inc, and Oxford University Press; and has received research institutional grant/research support related to a clinical trial (to Brigham and Women’s Hospital) from ONO Pharmaceuticals. L.L.R. has received honoraria from PeerView, Medscape, and Clinical Care Options; and has received consulting honoraria from AbbVie, Personal Genome Diagnostics, Bristol Myers Squibb, Loxo Oncology at Lilly, Amgen, Merck, AstraZeneca, Sanofi-Genzyme, and EMD Serono. J.T.J. holds equity in Navio Theragnostics, Akeila Bio, and The Doctor Lounge; and has participated in paid consulting from Navio Theragnostics, Recursion Pharmaceuticals, Alexion Pharmaceuticals, Springworks Pharmaceuticals, Merck Pharmaceuticals, Children’s Tumor Foundation, CEC Oncology, and Elsevier. J.D. is a consultant for Amgen, Blue Earth Diagnostics, and Unum Therapeutics; and has received research support from Novartis and publishing royalties from Wolters for UpToDate, Inc. M.M. consults for Bayer, DelveBio, Interline, and Isabl; holds equity in DelveBio, Interline, and Isabl; is an inventor of patents licensed to LabCorp and Bayer; and receives research funding from Bayer and Janssen. D.P.C. has consulted for the Massachusetts Institute of Technology, Advise Connect Inspire, Lilly, GlaxoSmithKline, Iconovir, Boston Pharmaceuticals, and Boston Scientific; serves on the advisory board of Pyramid Biosciences, which includes an equity interest; and has received honoraria and travel reimbursement from Merck for invited lectures, and from the US National Institutes of Health and Department of Defense for clinical trial and grant review. B.S.C. is a consultant for Koh Young. The remaining authors declare no competing financial interests.

Correspondence: Ganesh M. Shankar, Department of Neurosurgery, Massachusetts General Hospital, 15 Parkman St, WAC 745, Boston, MA 02114; email: gshankar@mgh.harvard.edu.