Clinical Implications of CSF-Ctdna in CNS Involvement of Newly Diagnosed Diffuse Large B Cell Lymphoma: An Improvement of Diagnosis, Treatment and Evaluation

Background: CNS involvement (CNSi) is a rare and often fatal event, occurring in around 5% of patients (pts) with diffuse large B-cell lymphoma (DLBCL) during primary treatment or shortly after completion (SCNSL). Furthermore, the potential mechanisms of CNS dissemination were limited. Circulating tumor DNA from cerebrospinal fluid (CSF-ctDNA) has played an important part in the application of liquid biopsy in CNS cancers. In this study, we provided new insights into the feasibility of using CSF-ctDNA as an essential tool for CNSi diagnosis, treatment, and evaluation specifically in newly diagnosed (ND) DLBCL pts.

Methods: Between December 2021 and April 2023, 168 ND DLBCL pts and 24 ND PCNSL pts have been treated in our center. One hundred DLBCL pts were enrolled in the training cohort while the other 68 pts refused to enter in our study. Another 26 ND DLBCL pts diagnosed between May 2023 and June 2023 were entered into a validation cohort. The flow of treatment and samples collections were according to the study protocol shown in Figure 1A. CSF-ctDNA positive (CSF-ctDNA(+)) criteria: (1) CSF-ctDNA detected mutations consistent with tissue; (2) Mutations detected in CSF-ctDNA were inconsistent with tissues, but were classified as grade I and grade II mutations; (3) At least one mutation was detected with an abundance >0.5%. The criteria of CNSi positive (CNSi(+)) were CSF-ctDNA(+) or/and other evidences detected by conventional methods (CM, including MRI, PET-CT, routine inspection analysis of CSF).

Results: In the training cohort, 25 pts (25%) were diagnosed with CSF-ctDNA (+) while 21 pts were according to the first criteria and 4 pts were according to the second criteria. Among the 25 pts (25.0%) who were finally diagnosed with CNSi (+), only 3 pts (12.0%) have evidences of CNSi both in CM and CSF-ctDNA while similar results were shown in validation cohort (Figure 1B). The unique mutations detected in CSF were rare compared with plasma and tissue mainly due to early detection with no clone evolution (Figure 1C). The mean variant allele frequency was significantly higher in tumor tissue than CSF in SCNSL (Figure 1D). The pretreatment ctDNA burden was significantly higher in CSF of PCNSL while higher in plasma of SCNSL (Figure 1F). The distributions of gene subtypes for the 25 CNSi (+) DLBCL pts were shown in Figure 1E. Only 11 pts (44%) were MCD subtype while nearly 80% PCNSL pts were MCD subtype. There existed significant differences of gene mutation profile between SCNSL and PCNSL (Figure 1G). We explored the association of clinical and biological factors with CNSi in ND DLBCL pts. Of note, high plasma ctDNA burden (≥2.18) were significantly related to CNSi (+) (Figure 1H). In multivariable analyses, CSF protein level ≥0.4, high plasma ctDNA burden (≥2.18) and high risk CNSi sites involvement were independent risk factors for CNSi (Figure 1I). We use these 3 risk factors to build CNSi-IPI model which can significantly better predict CNSi than CNS-IPI (Figure 1I). According to CNSi-IPI, pts were stratified into low-risk (0-1), intermediate-risk (2) and high-risk (3). CNSi (+) were observed in 5.0%, 48.3% and 72.7% in the training cohort and 5.9%, 50.0% and 100.0% in the validation cohort in the 3 risk groups (Figure 1J). Pts with CARD11, JAK2, ID3, and PLCG2 mutation were more predominant with CNSi while FAT4 mutation seemed to promote tumor cells disseminating to CNS (Figure 1K). Furthermore, the down-regulated genes of CNSi DLBCL pts significantly enriched in PI3K-AKT signaling pathway, focal adhesion, regulation of action cytoskeleton and tight junction pathways, which might be the potential mechanisms promoting CNS dissemination in ND DLBCL (Figure 1L). The dynamic change of disease status and ctDNA burden of all the 25 CNSi (+) pts were shown in Table 1.

Conclusion: In summary, our study highlighted the clinical implications of CSF-ctDNA in CNSi detection of ND DLBCL pts (Figure 1M): (1) CSF-ctDNA exhibits higher sensitivity than CM in detecting CNSi in ND DLBCL. (2) Different genomic and clinical landscapes were observed between SCNSL and PCNSL. (3) The CNSi-IPI is a robust, highly reproducible tool that can be used to estimate the risk of CNSi in ND DLBCL. (4) Down-regulation of PI3K-AKT signaling pathway, focal adhesion, regulation of action cytoskeleton and tight junction pathways contribute to CNS dissemination in ND DLBCL. (5) Furthermore, CSF-ctDNA can be used to monitor CNS tumor burden and response to treatment.