High-Dose Methotrexate, Ibrutinib, and Temozolomide in the Treatment of Newly Diagnosed Primary CNS Lymphoma: A Multicenter, Prospective Phase II Study Open Access

Primary diffuse large B-cell lymphoma (DLBCL) of the central nervous system (PCNSL) is a rare and aggressive lymphoma and has been recognized as a distinct entity by the World Health Organization classification and International Consensus Classification (1, 2). PCNSL accounts for 1% of non-Hodgkin lymphomas (3, 4). High-dose methotrexate (HD-MTX)–based chemotherapy is considered the backbone induction therapy for newly diagnosed (ND) PCNSL but elicits considerable resistance and a high chance of disease recurrence (5).

Temozolomide (TMZ) is an alkylating agent. It can penetrate the blood–brain barrier (6). TMZ monotherapy is efficacious in PCNSL (7). In the CALGB 50202 study, HD-MTX and TMZ combined with rituximab resulted in a high response rate and long-term survival (8). Although the therapeutic effects of rituximab on extracranial mature B-cell lymphomas are dramatic, the effect on PCNSL treatment is controversial (9). In the IELSG32 study, compared with MTX–cytarabine, the combined rituximab group had a better response rate, and rituximab brought overall survival (OS) benefits after a long follow-up (10, 11). However, rituximab did not improve the survival benefits of the MTX, etoposide, carmustine, and prednisone regimen in the HOVON 105/ALLG NHL 24 study (9).

Despite important advances in the induction chemotherapy of PCNSL in the past years, the efficacy of induction regimens has potential for improvement. Mutations of B-cell receptor (BCR) signaling molecules are common in PCNSL, including gain-of-function mutations in myeloid differentiation primary response 88 (MYD88) and cluster of differentiation 79b (CD79B), and can lead to malignant progression (12, 13). Bruton tyrosine kinase (BTK) is a tyrosine kinase found downstream of the BCR signaling pathway and a therapeutic target for PCNSL (14). To date, the results of two prospective clinical studies showed the response rate of ibrutinib monotherapy in relapsed or refractory (R/R) PCNSL was 42% to 74%, and the complete response rate (CRR) was 19% to 39%. The median progression-free survival (PFS) was 4.5 to 4.8 months (15–17). High response rates to ibrutinib were also observed in patients with ND PCNSL after treatment with the ibrutinib-based TMZ, etoposide, liposomal doxorubicin, dexamethasone, and rituximab (DA-TEDDi-R) regimen (17). Initiation of BTK inhibitor prior to R/R might further optimize patient outcomes. However, infection events such as aspergillosis associated with BTK inhibitors are of concern (17). To minimize the risk of adverse events (AE), we did not choose to combine with rituximab as an induction regimen in this study.

Combining chemotherapy with reduced-dose whole-brain radiotherapy (WBRT) can improve PFS, and a median PFS of 18 to 24 months has been reported (8, 18, 19). However, WBRT use can end in irreversible neurotoxicity (e.g., cognitive impairment) and, thus, a serious decline in quality of life (18, 19). The IELSG32 and PRECIS studies have confirmed that both autologous stem cell transplantation (ASCT) and WBRT are effective consolidation treatment options for PCNSL. However, WBRT leads to higher neurotoxicity (11, 18). In addition, intensive chemotherapy may be an effective alternative for consolidation treatment (8, 20).

The International Extranodal Lymphoma Study Group (IELSG) and Memorial Sloan Kettering Cancer Center (MSKCC) are widely used risk stratification models (21, 22). In recent years, ctDNA has become an emerging biomarker. ctDNA can provide information on genomic alterations and dynamic evaluation of the efficacy of therapy and help predict recurrence (23). Because the tumor cells are situated near the brain chamber and/or brain membrane and the tumor burden is relatively small, cerebrospinal fluid (CSF) could reflect the genetic mutations of tumors within the central nervous system (CNS) better than plasma (24). Furthermore, collecting CSF is more convenient and safer than obtaining brain tumor samples.

We designed a prospective study to elucidate the efficacy and toxicity of HD-MTX and ibrutinib in combination with TMZ (MIT) in patients with ND PCNSL. Also, we undertook targeted next-generation sequencing (NGS) on tissue samples at baseline and sequential CSF/plasma samples to compare genetic mutations between paired samples. Dynamic monitoring with CSF/plasma was done to analyze the response to treatment and tumor progression.

Between July 2021 and October 2022, 35 eligible patients were recruited from three centers. One patient was excluded due to secondary CNS lymphoma, and one patient withdrew consent for participation 10 days after treatment. Hence, 33 patients were included in the efficacy and toxicity assessment (Supplementary Fig. S1).

All patients had a pathologic diagnosis of DLBCL, and systemic lesions were excluded by imaging examinations. The median age was 56 (IQR, 45–66) years. Seventeen patients (51.5%) were men. Twenty-two patients (66.7%) had a pathologic subtype of non–germinal center B-cell–like (non-GCB) DLBCL. Twenty-five patients (75.8%) had multiple cerebral lesions and an increased level of protein in CSF. Eye involvement was documented in five patients (15.2%). Epstein–Barr virus (EBV)–encoded small RNA in situ hybridization was performed in 28 of 33 patients. All patients (n = 28) had EBV–PCNSL. The clinical characteristics of patients are summarized in Table 1.

Thirty-three patients received a median of six (range, 2–6) cycles of MIT regimen. Two patients received two cycles of treatment: one due to disease progression and one due to pulmonary tuberculosis. Five and two patients discontinued the study after four and six cycles, respectively, due to disease progression. Twenty-three patients completed six courses of induction therapy and received ibrutinib maintenance therapy. Subsequently, disease progression occurred in four patients during maintenance. One patient died due to unknown causes while in complete response (CR) during maintenance. The best overall response rate (ORR) and CRR after induction therapy were 93.9% [95% confidence interval (CI), 80.4–98.3] and 72.7% (95% CI, 55.8–84.9), respectively. The ORR and CRR at the end of induction therapy were 72.7% (95% CI, 55.8–84.9) and 69.7% (95% CI, 52.7–82.6), respectively. The median duration of follow-up was 26.1 (range, 14.9–30.4) months for survivors. The 2-year PFS was 57.6% (95% CI, 49.0–66.2; Fig. 1A). The OS was 84.8% (95% CI, 78.6–91.0; Fig. 1B). The swimmer plot for the best overall response and treatment duration are shown in Fig. 2. Cox regression analysis indicated that there are no clinical characteristics related to PFS (Supplementary Table S1). Patients with Karnofsky performance status (KPS) ≤ 60 had shorter OS (P = 0.014; Supplementary Table S2). Twelve patients age ≤60 years and who had a KPS score ≥60 chose to receive ibrutinib maintenance; no patient accepted ASCT as consolidation. To date, 18 of 23 patients receiving ibrutinib maintenance remained disease-free, and the median PFS was 26.5 (range, 16.9–30.4) months.

The detection of at least one pathogenic mutation or copy number variation (CNV) in tumor tissue, CSF, and plasma at baseline was 100% (24 of 24), 83.3% (25 of 30), and 53.6% (15 of 28), respectively (Supplementary Fig. S2A). Moreover, the number of genetic mutations and CNVs in tumor tissue was significantly higher than that in CSF and plasma (Supplementary Fig. S2B). We found that 47.0% (241 of 513) and 13.6% (50 of 369) of mutated genes in tumor tissue were also found in CSF and plasma, respectively (Supplementary Fig. S2C–S2E). The most common gene mutations in tumor tissue and CSF were Pim-1 proto-oncogene (PIM1; 83.3% and 52.0%, respectively), MYD88 (70.8% and 48.0%), B-cell translocation gene 2 (BTG2; 62.5% and 36.0%), and CD79B (58.3% and 28.0%; Fig. 3A and B). In plasma, however, they were MYD88 (40.0%), lysine methyltransferase 2D (KMT2D; 26.7%), PIM1 (26.7%), B-cell lymphoma 6 (BCL6; 20.0%), and BTG1 (20.0%; Fig. 3C). Compared with plasma, CSF had more similar genetic aberrations to tumor tissue with respect to the prevalence of positive detection, number of mutated genes, and overall “genomic landscape.” Based on comprehensive analyses of different samples according to the LymphGen classification, 63.3% (19 of 30) of patients were determined to have the MYD88 and CD79B double mutations (MCD) type, 6.7% (2 of 30) to have the BN2 type, 3.3% (1 of 30) to have the EZB type, and 26.7% (8 of 30) to have other types (Supplementary Fig. S3A–S3D; Supplementary Table S3).

Patients with a higher IELSG score (4, 5) had a higher allele fraction variant (%) and ctDNA concentration in CSF, but the latter did not reach statistical significance (Supplementary Fig. S4A and S4B). Similar trends were observed with the MSKCC scores but without statistical significance (Supplementary Fig. S4C and S4D). Further analyses revealed patients with a mutation in ATRX chromatin remodeler (ATRX) or tumor necrosis factor α–induced protein 3 (TNFAIP3) in tumor tissues had shorter PFS (Supplementary Fig. S5A and S5B). Patients with mutated genes in CSF at baseline were more prone to disease progression and had shorter PFS (Supplementary Fig. S5C). Patients who cleared ctDNA from CSF after two cycles of the MIT regimen achieved longer PFS (Supplementary Fig. S5D). The dynamic characteristics of CSF and plasma, imaging evaluation, and survival of patients are shown in Fig. 4. The ctDNA level in CSF was in accordance with the imaging evaluation and could be used to predict the therapeutic effect and risk of disease recurrence. ctDNA could not be detected in patients who achieved a CR, but it could be detected in patients with disease progression (Supplementary Fig. S6A–S6D). NGS of CSF from five patients with disease relapse revealed new gene mutations, mainly PIM1 (n = 2), caspase recruitment domain family member 11 (CARD11; n = 2), and KMT2D (n = 1).

Most of the AEs experienced by patients during induction treatment and maintenance therapy using ibrutinib were grade I/II (Table 2). The main AEs of grade III/IV were thrombocytopenia (9.1%), neutropenia (9.1%), increased level of alanine transaminase (6.1%), subdural hematoma (6.1%), and increased level of creatinine (6.1%). Two patients with grade III creatinine elevation discontinued HD-MTX and only received ibrutinib and TMZ. One patient experienced atrial fibrillation during ibrutinib maintenance.

Four patients developed subdural hematoma during treatment that necessitated temporary discontinuation of ibrutinib administration. Two patients withdrew from this study due to pulmonary infections. One patient contracted tuberculosis after two cycles of treatment. One patient was infected with coronavirus disease 2019 and subsequently died of the disease after five cycles of treatment.

HD-MTX–based polychemotherapy remains the standard treatment for ND PCNSL. ASCT is now preferred as consolidation therapy in fit patients after HD-MTX–based chemotherapy. WBRT remains an alternative approach for consolidation in fit patients with insufficient autologous stem cell harvest.

DA-TEDDi-R demonstrated high efficacy in R/R PCNSL. Nevertheless, this promising clinical activity was associated with significant toxicity, including invasive aspergillosis occurring in 39% of the patients (14). HD-MTX was excluded from the DA-TEDDi-R regimen based on antagonism with ibrutinib in preclinical assays. BTK inhibitor and other novel agents may pave the way to MTX-free regimens, but further clinical trials are needed.

The future study trend is to optimize induction treatments to reduce toxicity, so most patients receive effective consolidation strategies. The role of novel targeted agents as first-line treatment should be explored in prospective trials. The aim of our study was to assess the feasibility of response-adapted ibrutinib after HD-MTX–based chemotherapy in ND PCNSL in China. This prospective multicenter phase II study involving 33 patients and evaluating ibrutinib combined with MTX and TMZ as first-line therapy focused on patients with ND PCNSL showed an ORR of 93.3%, with 72.7% CRR achieved as the best response during treatment. Due to various objective factors, such as economic constraints, enrolled patients did not undergo ASCT consolidation following MIT induction. This trial showed that the MIT regimen followed by ibrutinib maintenance was effective in this population, meeting the predetermined threshold for response rate, without major safety concerns. The results compare favorably with historic controls, the prospective upfront treatment trials in PCNSL that are summarized in Table 3. The high response rate of the MIT regimen and the durable remissions followed by ibrutinib maintenance suggested that this regimen may improve outcomes in PCNSL. Certainly, larger samples are still needed to confirm.

We assessed the prognostic impact of several parameters at inclusion, such as molecular characteristics of the disease. Historic data indicated that the proportion of the non-GCB subtype in PCNSL ranges from 67% to 96.3% (25–28), which is higher than in this study. We found no significant statistical differences between non-GCB and GCB subtypes in ORR, CR, PFS, and OS. NGS revealed that in our cohort of patients with PCNSL, the MCD subtype accounts for 63.3% of patients, which is more prevalent than reported in extracranial DLBCL overall (8%; ref. 29). Patients with PCNSL are likely to benefit from the use of the BTK inhibitor (30). Patients with higher IELSG and MSKCC scores tended to have higher allele fraction variants and ctDNA concentrations, but due to the small sample size, statistical significance was not reached.

Based on the mutation status of baseline tumor tissue, we identified TNFAIP3 (n = 3) and ATRX mutations (n = 3) associated with poor PFS in patients with PCNSL. TNFAIP3, as a distal member of the BCR signaling pathway, may be expected to restore BCR pathway activity in the presence of ibrutinib, and its mutation status is often associated with poor efficacy of ibrutinib monotherapy (31). ATRX is a chromatin remodeling factor and transcriptional regulator involved in chromatin regulation and stability maintenance (32). It has been found to regulate multiple molecular and cellular functions, including gene expression, telomere integrity maintenance, chromatin state regulation, and DNA damage repair (33, 34). Although the association of ATRX and TNFAIP3 mutations with outcomes is statistically significant, a larger sample size is needed to validate their prognostic significance.

The prognostic value of ctDNA in CSF still needs to be confirmed. Previous studies have also confirmed that plasma-based minimal residual disease (MRD) has higher accuracy in predicting survival compared with PET/CT scans (35–37). Plasma-based MRD can also improve early detection of relapse (38). In addition, in DLBCL with CNS involvement, CSF–ctDNA had higher sensitivity in detecting central progression than conventional methods, including MRI, PET/CT, and routine inspection analysis of CSF. It can also reflect the tumor burden in the CNS and the response to treatment (39). We revealed that CSF can better reflect the genomic landscape of PCNSL than plasma. The “mutation profile” in CSF was more consistent with that of intracranial tumors (especially for the most commonly mutated genes). CSF is a promising alternative tissue for the diagnosis if sampling of intracranial tumor tissue is not possible. The dynamic CSF assessment suggests that the ctDNA detection method may become feasible for detecting MRD in PCNSL. In the present study, an absence of ctDNA in CSF at baseline and early clearance from CSF were associated with a better prognosis. In brain tumors (including cerebral lymphomas), the ctDNA load in CSF is associated with patient outcomes (24). NGS plays an important part in detecting genetic abnormalities to predict the therapeutic effect and disease recurrence.

Additionally, dynamic monitoring of CSF led to the detection of new gene mutations for patients who relapsed after treatment with the MIT regimen, such as CARD11 (n = 2), PIM1 (n = 2), and KMT2D (n = 1). CARD11 is downstream of BTK and activates NF-κB. CARD11 activation is an important factor responsible for resistance to BTK inhibitors (40, 41). Mutated PIM1 has been reported to reduce the sensitivity to BTK inhibitors. PIM1 expression has been shown to increase in ibrutinib-resistant cell lines with prolonged exposure to ibrutinib (42). The role of these genes in predicting relapse deserves attention. Whether the mutations emerged during disease recurrence as a result of clonal evolution or were undetectable at diagnosis due to the limited sensitivity of the NGS analysis cannot be ascertained.

The common adverse reaction of the MIT regimen is myelosuppression, and most AEs were grades I and II. We observed a relatively low prevalence of hematologic toxicities and infection with the MIT regimen. The reason for the lower toxicity may be related to the young age of the enrolled patients and the fewer chemotherapy agents in the MIT regimen. We also observed ibrutinib-related AEs of special interest. One patient developed atrial fibrillation during ibrutinib maintenance. It is worth mentioning that subdural hematomas were found in four patients, but they were all asymptomatic, and a surgical procedure was undertaken in one patient. No patient died or suffered sequelae related to subdural hematomas. Overall, the AEs elicited from the use of the MIT regimen were controllable. Frail/elderly patients using this regimen need regular electrocardiogram monitoring due to the potential risk of atrial fibrillation and should also be vigilant for the risk of subdural hematoma.

This trial has significant limitations. The main limitations include trial design. As an early-phase clinical trial, the sample size is small and potentially heterogeneous, and patients were not randomized. The relatively low median follow-up and small numbers of patients who completed ibrutinib maintenance introduce a potential confounding factor that needs careful consideration when interpreting our results. Furthermore, the scarce availability of brain tumor tissue limits the scope of NGS analysis. Finally, none of the patients received ASCT as consolidation for two reasons: During part of the study period, the drug thiotepa as transplantation conditioning regimen was unavailable, and even when having this choice, most patients preferred ibrutinib maintenance therapy. The benefit of ibrutinib as a maintenance therapy for PCNSL remains to be confirmed through randomized clinical trials.

This prospective phase II trial met its primary endpoint and suggested that the MIT regimen has clinical, radiologic, and biological effects in patients with ND PCNSL. This study demonstrated that the highly responsive MIT regimen followed by single-agent maintenance therapy using ibrutinib is expected to result in long-term survival and was associated with acceptable toxicities. The early clearance of ctDNA from CSF may have been related to the favorable survival outcomes.

The study protocol was approved (B2020-175) by the Ethics Committee of Sun Yat-sen University Cancer Center (Guangzhou, China). This investigation has been conducted in accordance with the ethical standards and according to the Declaration of Helsinki and national and international guidelines. All participants provided written informed consent. This study is registered (NCT04514393) at www.clinicaltrials.gov.

Patients with a pathologic diagnosis of PCNSL were eligible for this clinical trial. The inclusion criteria were as follows: (i) ages 18 to 75 years; (ii) at least one measurable characteristic lesion on the brain or spinal cord MRI with gadolinium enhancement, or PET/CT; (iii) adequate bone marrow (BM; neutrophil count >1.5 × 10⁹/L, platelet count >75 × 10⁹/L); (iv) sufficient heart, liver, and kidney functions; and (v) KPS from 40 to 100.

The main exclusion criteria were as follows: (i) congenital or acquired immunodeficiency, (ii) receipt of antitumor therapy before study enrollment (except the administration of low-dose corticosteroids), (iii) evidence of extra-CNS involvement, and (iv) other incurable malignant tumors.

Pretreatment evaluation included PET/CT and/or contrast-enhanced MRI of the brain and spinal cord. Contrast-enhanced CT scans of the neck, chest, abdomen, and pelvis, as well as BM biopsy and aspiration, were undertaken to exclude extracranial DLBCL. Slit-lamp examination and aspiration of the aqueous humor were carried out in patients with ocular symptoms or suspected eye involvement. For CSF, cytologic and biochemical parameters were examined, and genetic profiles were evaluated by NGS. Samples of blood from the peripheral circulation were used to assess liver and kidney functions. Cardiac function was tested using ultrasound and electrocardiography. These tests were done to ensure that patients had adequate BM and organ functions.

This was an open-label, multicenter, phase II trial. Patients were treated with MTX (3.5 g/m² intravenously guttae) for 6 hours on day 1 and with calcium folinate (15 mg/m², every 6 hours) for 12 hours after MTX infusion for 8 to 10 times until HD-MTX clearance. TMZ (150 mg/m², orally) was administered on days 1 to 5, and ibrutinib (560 mg, orally) was taken after HD-MTX clearance until day 21. This therapeutic regimen was repeated every 3 weeks for a maximum of six courses. Patients aged ≤60 years with a KPS score ≥60 who achieved a CR or partial response were eligible for ASCT (ASCT conditioning regimen consisted of thiotepa 250 mg/m², days −6 and −5; busulfan 0.67 mg/kg every 6 hours, days −6, −5, and −4; and cyclophosphamide 60 mg/kg, days −3 and −2) as consolidation treatment, whereas patients aged >60 years or who were unfit who achieved a CR or partial response received ibrutinib maintenance up to 2 years.

Brain/spinal cord imaging and ctDNA level in CSF were evaluated every two cycles during induction therapy, every 3 months during maintenance therapy using ibrutinib, every 6 months during follow-up, and once a year after 5 years. The treatment response was evaluated using criteria set by the International PCNSL Collaborative Group (43). Patients who have undergone at least two cycles of induction therapy are considered evaluable.

Patients with progressive disease or stable disease without improvements in symptoms withdrew from this study and received salvage therapy. AEs were graded according to the Common Terminology Criteria for Adverse Events version 5.0.

NGS was done on samples from tumor tissue, CSF, and plasma. Samples were evaluated in a central testing laboratory at Nanjing Geneseeq Technology Inc. (Nanjing, China) accredited by clinical laboratory improvement amendments (Columbia, SC, USA) and the College of American Pathologists (Cook County, IL, USA). Subsequently, the study delved into investigating the correlation among clinical characteristics, tumor mutation burden, and mutation types with treatment efficacy. As described previously (44), QIAamp DNA FFPE Tissue Kit (QIAGEN) and DNeasy Blood and Tissue Kit (QIAGEN) were used to extract DNA from formalin-fixed, paraffin-embedded (FFPE) sections and white blood cells, respectively. The DNA from white blood cells was used as the germline control. ctDNA was extracted from CSF and plasma using QIAamp Circulating Nucleic Acid Kit (QIAGEN) and stored at −80°C until use. The DNA quality was assessed using NanoDrop 2000 (Thermo Fisher Scientific), and the quantity was measured using dsDNA HS Assay Kit (Life Technologies) on Qubit 2.0. Tumor genomic DNA was extracted and fragmented into ~300 to 350 bp to prepare sequencing libraries. These libraries underwent PCR amplification and purification before targeted enrichment using a leukemia- and lymphoma-related panel to target 475 predefined genes (summarized in Supplementary Table S4). The enriched libraries were then sequenced on the HiSeq 4000 NGS platform (Illumina) to achieve mean coverage depths of at least 1,000× for FFPE samples, 5,000× for plasma and CSF, and 200× for oral swabs. Reads from each sample were mapped to the reference sequence Human Genome version 19. Somatic mutations were detected using VarScan2, whereas genomic fusions were identified using FACTERA with default parameters. CNVs were detected using ADTEx (http://adtex.sourceforge.net) with default parameters. Consistent with our previous reports (45), the positive cutoff value for allele frequency of mutations in tissue samples is 1%, whereas for ctDNA, it is 0.3%.

Genomic DNA was extracted from FFPE tumor tissue samples using the QIAamp DNA FFPE Tissue Kit (QIAGEN) following the manufacturer’s instructions. Peripheral blood samples (8–10 mL) and CSF samples (3–5 mL) were promptly centrifuged within 2 hours of collection. The plasma and CSF samples were then used for ctDNA extraction with the QIAamp Circulating Nucleic Acid Kit (QIAGEN) as per the manufacturer’s guidelines. Germline mutations were identified through sequencing of oral swabs. The quality of the DNA was assessed using the NanoDrop 2000 (Thermo Fisher Scientific), and the quantity was measured using the dsDNA HS Assay Kit (Life Technologies) on Qubit 2.0. The DNA samples were stored at −80°C for subsequent targeted sequencing. The tumor genomic DNA was fragmented into 300 to 350 bp using the Covaris M220 instrument (Covaris). Sequencing libraries were prepared using the KAPA HyperPrep Kit (KAPA Biosystems) with optimized protocols. Specifically, ctDNA underwent end-repairing, A-tailing, adapter ligation, and size selection using Agencourt AMPure XP beads (Beckman Coulter). The libraries were then subjected to PCR amplification and purification before targeted enrichment.

DNA libraries from various samples were tagged with unique indices and combined for targeted enrichment. To prevent nonspecific binding, human Cot-1 DNA and xGen Universal blocking oligos were included. A custom-designed xGen Lockdown Probes panel was utilized to enrich 475 lymphoma-related genes (Supplementary Table S4). The hybridization reaction was carried out using NimbleGen SeqCap EZ Hybridization and Wash Kit (Roche). Probe-bound fragments were captured using Dynabeads M-270 (Life Technologies) and amplified using KAPA HiFi HotStart ReadyMix. Library quantification was performed using KAPA Library Quantification Kit, and size distribution was assessed using an Agilent Technologies 2100 Bioanalyzer. For cfDNA samples, an optimized library preparation method called Automated Triple Groom Sequencing was used to detect low-abundance mutations with high sensitivity and specificity (46). The enriched libraries were sequenced on NovaSeq 6000 NGS platforms (Illumina) to achieve coverage depths of up to 5,000× in ctDNA and 1,500× in tumor DNA. All targeted NGS procedures were conducted at a testing laboratory accredited by the Clinical Laboratory Improvement Amendments and the College of American Pathologists (Nanjing Geneseeq Technology Inc.).

The FASTQ files were subjected to Trimmomatic to eliminate low-quality or unknown nucleotide (N) bases. The reads were then aligned to the Human Genome version 19 reference genome using Burrows-Wheeler Aligner–MEM. Local realignment and base quality recalibration were performed using GATK. Germline mutations were identified from control samples of oral epithelial cells with GATK3.4.0, whereas VarScan2 was used for the analysis of somatic mutations. Mutations from the Single Nucleotide Polymorphism Database and 1000 Genomes Project were filtered out, with retention of COSMIC mutations. Annotation was conducted using ANNOVAR, and genomic fusions were identified with FACTERA. CNVs were detected using ADTEx (http://adtex.sourceforge.net). Somatic variants were identified using paired normal/tumor samples for each exon. PyClone was used to infer the clonal architecture of ctDNA samples (47). The levels of ctDNA were quantified in haploid genome equivalents per milliliter of plasma or CSF, calculated by multiplying the total concentration of cell-free DNA (48). Each mutation underwent thorough manual verification to ensure its accuracy. Furthermore, the molecular typing of lymphoma was analyzed using the LymphGen tool (https://llmpp.nih.gov/lymphgen/lymphgendataportal.php) based on the confirmed somatic mutations identified (49).

The primary endpoints were the ORR, CRR, and PFS. The secondary endpoints were OS and safety parameters. The response rate was calculated with a two-sided 95% CI. The sample size was calculated using Simon’s two-stage design. Assuming a response rate of 65% for existing treatment, 33 patients (10 of 14, 25 of 33) were sufficient to demonstrate the superiority of a new therapy with an 85% response rate (two-sided test; type I error α = 0.05, type II error β = 0.2).

OS was defined as the time from treatment initiation to death by any cause. PFS was defined as the time from treatment initiation to disease progression, last clinical assessment, or death, whichever came first.

Statistical analyses were undertaken using SPSS 24.0 (IBM), R 4.0.3 (R Institute for Statistical Computing), and Prism 8.0 (GraphPad). PFS and OS were analyzed using the Kaplan–Meier curve and compared using the log-rank method. Cox regression analysis was used to explore prognostic factors for survival. The median duration of follow-up was calculated for all survivors. Heatmaps were generated using the ComplexHeatmap package, and cancer cell fraction was calculated using PyClone. The overlap of mutated genes among tumor tissue, CSF, and plasma was determined through Venn diagram analysis. Group comparisons were conducted using the Fisher exact probability method. A significance level of P < 0.05 was considered statistically significant.

The raw sequencing data for this study have been uploaded to the China National Center for Bioinformation [https://ngdc.cncb.ac.cn/gsa-human; Genome Sequence Archive for Human (GSA-Human): HRA008975]. These data are governed by human privacy regulations and are intended solely for research purposes. Access to the data must comply with the user request guidelines established by the Data Access Committee of the GSA-Human database. For inquiries about the research protocol, informed consent, and statistical analysis, please contact [email protected].

X. Hong reports other support from Nanjing Geneseeq Technology Inc. outside the submitted work. No disclosures were reported by the other authors.

Y. Gao: Conceptualization, resources, formal analysis, supervision, funding acquisition, methodology, writing–review and editing. L. Ping: Software, investigation, visualization, methodology, writing–original draft, project administration. C. Shan: Conceptualization, resources, data curation, supervision, project administration. H. Huang: Conceptualization, resources, data curation, supervision, methodology. Z. Li: Conceptualization, data curation, project administration. H. Zhou: Conceptualization, resources, data curation, supervision, project administration. M. Lai: Resources, supervision, project administration. L. Cai: Resources, supervision, project administration. B. Bai: Resources, data curation, project administration. C. Huang: Data curation, project administration. H. Chen: Data curation, methodology, project administration. X. Hong: Conceptualization, resources, software, formal analysis, supervision, methodology. X. Wang: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, methodology, project administration, writing–review and editing. H. Huang: Conceptualization, resources, supervision, funding acquisition, investigation, methodology, project administration, writing–review and editing.

We would like to express our gratitude to the patients and families who participated in this clinical study. This work was supported by grants from the National Natural Science Foundation of China (Grant No. 82170188). Xi’an Janssen Pharmaceutical Ltd. provided free ibrutinib. The sponsors had no role in the study design; collection, analysis, and interpretation of data; writing of the report; and decision to submit the article for publication.