Primary central nervous system lymphoma in the molecular era: genomic drivers, liquid biopsy, and targeted therapeutic strategies

Abstract

Primary central nervous system lymphoma (PCNSL) is an uncommon and aggressive extranodal non-Hodgkin lymphoma confined to the central nervous system, including the brain, leptomeninges, spinal cord, and eyes. Although high-dose methotrexate-based induction remains the backbone of first-line therapy, long-term disease control remains challenging, particularly in older or frail patients and in those with relapsed or refractory disease. Recent genomic and multi-omics studies have reshaped the biological understanding of PCNSL by identifying recurrent alterations in MYD88, CD79B, PIM1, PRDM1, CDKN2A/B, HLA-related genes, B2M, CIITA, and immune checkpoint loci, which converge on B-cell receptor/Toll-like receptor signaling, NF-κB activation, immune evasion, and microenvironmental adaptation. These molecular insights have accelerated minimally invasive diagnostics, including targeted mutation assays, multigene circulating tumor DNA profiling, methylation-based approaches, and cytokine analysis in cerebrospinal fluid, vitreous humor, and aqueous humor. Targeted and immune-based therapies, including Bruton tyrosine kinase inhibitors, immunomodulatory agents, immune checkpoint inhibitors, mTOR inhibitors, and CD19-directed chimeric antigen receptor T-cell therapy, have shown variable degrees of clinical activity, particularly in relapsed or refractory disease. However, most evidence remains derived from early-phase trials, single-arm studies, retrospective cohorts, or small series, and treatment selection is not yet routinely biomarker-defined. This review summarizes current advances in molecular taxonomy, pathogenesis, diagnostic work-up, liquid biopsy, prognostic stratification, and therapeutic management of PCNSL, emphasizing how molecular and longitudinal biomarker data may inform individualized treatment and future clinical trial design.

Introduction

Primary central nervous system lymphoma (PCNSL) is an uncommon and aggressive extranodal non-Hodgkin lymphoma that arises within the central nervous system (CNS), including the brain, leptomeninges, spinal cord, and eyes, without evidence of systemic lymphoma at diagnosis [1, 2]. Most cases are diffuse large B-cell lymphoma (DLBCL), whereas indolent B-cell lymphomas, T-cell lymphomas, and other rare histological subtypes represent uncommon but clinically important entities. Because these rare subtypes may differ in biological behavior, diagnostic work-up, and therapeutic sensitivity, they are summarized separately in Table 1 [3, 4].

Table 1.

Histological and biological subtypes relevant to primary CNS lymphoma

Histological or biological entity	Typical setting and pathology	Characteristic molecular or immune features	Diagnostic priorities	Management considerations
Large B-cell lymphomas of immune-privileged sites: primary CNS/vitreoretinal large B-cell lymphoma	Most common PCNSL in immunocompetent adults; large CD20 +, CD79a +, PAX5 +, MUM1/IRF4 + B cells; parenchymal and/or vitreoretinal disease	MCD/C5 enrichment; frequent MYD88 L265P, CD79B, PIM1, PRDM1, CDKN2A/B loss, HLA/B2M/CIITA disruption, and immune-checkpoint alterations	Stereotactic biopsy or compartment-specific ocular/CSF sampling; integrated morphology, flow cytometry, MYD88/CD79B testing, cytokines, and systemic staging	HD-MTX-based induction is the backbone. Routine cytoreductive resection is not standard; consolidation, maintenance, and salvage are fitness- and response-adapted
Primary CNS low-grade B-cell lymphoma, including dural marginal zone lymphoma	Rare, often dural, localized, or leptomeningeal; small B-cell morphology and subtype-specific immunophenotype; generally more indolent than PCNS-DLBCL	Heterogeneous and subtype-dependent; dural marginal zone lymphoma is biologically distinct from intraparenchymal PCNS-DLBCL	Distinguish from meningioma, inflammatory disease, systemic indolent lymphoma, and PCNS-DLBCL; biopsy is usually required	Biopsy for diagnosis; involved-site radiotherapy, rituximab-based therapy, or observation in selected cases. Resection only for exceptional diagnostic or decompressive indications
Primary CNS T-cell lymphoma	Very rare; parenchymal, leptomeningeal, or multifocal; diverse T-cell immunophenotypes	Variable EBV status and molecular profile; no established recurrent PCNSL-specific driver set	Adequate tissue for morphology, immunohistochemistry, T-cell receptor clonality, EBV testing, and exclusion of systemic T-cell lymphoma	Usually adapted from CNS-penetrant PCNSL regimens, commonly HD-MTX-based when feasible; evidence for targeted therapy, radiotherapy, or transplantation is limited
Immunodeficiency-associated PCNSL	HIV, transplantation, autoimmune-disease immunosuppression, or congenital immunodeficiency; DLBCL morphology; EBV detected in approximately 79%−80%	Often lacks canonical MYD88, CD79B, PIM1, and CDKN2A alterations and HLA-related lesions; SOCS1 mutations, NOTCH-pathway alterations, IL-10/JAK2 signaling, reduced promoter methylation, and tolerogenic T-regulatory/M2-like macrophage microenvironment	Document immune status; EBER in situ hybridization; exclude opportunistic infection and systemic/post-transplant lymphoproliferative disease	Rituximab and CNS-penetrant methotrexate-based therapy when feasible plus immune reconstitution or carefully supervised reduction of immunosuppression
CNS-predominant intravascular large B-cell lymphoma	Important differential diagnosis; systemic aggressive lymphoma with large B cells confined to small-vessel lumina; CNS symptoms may dominate	Molecular overlap with immune-privileged-site and other activated B-cell lymphomas; not strict PCNSL in many cases	Random skin biopsy, marrow evaluation, FDG-PET/CT, and systemic work-up may be higher yield than brain biopsy	Treat as systemic aggressive lymphoma with CNS-directed components; do not apply a PCNSL-only algorithm without systemic evaluation

This table is organized by histological or biological entity rather than by neuroanatomical compartment. Vitreoretinal, leptomeningeal, and spinal presentations are summarized in Table 2. Recommendations for rare entities are based mainly on retrospective series, case reports, systemic lymphoma principles, and expert consensus; multidisciplinary interpretation is essential

Abbreviations: CNS central nervous system, CSF cerebrospinal fluid, DLBCL diffuse large B-cell lymphoma, EBER Epstein-Barr virus-encoded RNA, EBV Epstein-Barr virus, FDG-PET/CT fluorodeoxyglucose positron emission tomography/computed tomography, HD-MTX high-dose methotrexate, PCNSL primary central nervous system lymphoma

PCNSL differs substantially from systemic DLBCL because it develops within an anatomically and immunologically specialized compartment and is influenced by the blood–brain barrier, CNS immune surveillance, and the perivascular microenvironment. These features affect clinical presentation, tissue acquisition, drug delivery, treatment-related neurotoxicity, and patterns of relapse [5, 6].

The management of PCNSL has improved with the adoption of high-dose methotrexate (HD-MTX)-based induction therapy and consolidation approaches such as autologous stem cell transplantation or reduced-dose whole-brain radiotherapy in selected patients [7–9]. In fit patients treated with contemporary intensive regimens at experienced centers, median overall survival can exceed 5 years, and long-term survival is achievable in a substantial proportion [7–11]. Nevertheless, durable disease control remains challenging, particularly in older patients, frail patients, and those with relapsed or refractory (R/R) disease. Treatment-related neurotoxicity, limited CNS penetration of many systemic agents, and the absence of universally accepted maintenance strategies remain major unmet needs [12, 13].

Recent genomic and multi-omics studies have reshaped the biological understanding of PCNSL by identifying recurrent alterations in MYD88, CD79B, PIM1, PRDM1, CDKN2A/B, HLA-related genes, B2M, CIITA, and immune checkpoint-related regions [14, 15]. These alterations converge on B-cell receptor/Toll-like receptor signaling, NF-κB activation, immune escape, and CNS microenvironmental adaptation, thereby providing a rationale for molecularly targeted and immune-based therapeutic strategies. In parallel, liquid biopsy has emerged as a clinically relevant diagnostic and monitoring tool, particularly through detection of MYD88 L265P and other tumor-derived molecular alterations in cerebrospinal fluid, vitreous humor, or aqueous humor [16].

This review summarizes recent advances in the molecular pathogenesis, diagnostic work-up, liquid biopsy, prognostic stratification, and therapeutic management of PCNSL. Particular emphasis is placed on the translational relevance of genomic alterations, the evolving role of minimally invasive molecular diagnostics, and the current evidence supporting targeted and immune-based therapies.

Epidemiology

PCNSL is rare, with a population incidence of approximately 0.4–0.5 cases per 100,000 person-years in Western registry studies, although estimates vary by period, region, age, and immune status [17–20]. It accounts for approximately 2%−3% of primary CNS tumors and less than 1% of non-Hodgkin lymphomas [1, 5, 21]. Incidence rises steeply with age and is highest in adults older than 65 years, making PCNSL predominantly a disease of the seventh and eighth decades of life[5, 17, 18]. Advanced age is clinically important because older patients often have impaired renal reserve, comorbidities, frailty, and increased vulnerability to treatment-related neurotoxicity, all of which influence the feasibility of HD-MTX-based chemotherapy, whole-brain radiotherapy, and high-dose chemotherapy followed by autologous stem cell transplantation [12, 22].

Historically, the incidence of PCNSL increased during the HIV/AIDS epidemic, reflecting the strong association between severe immunodeficiency and Epstein-Barr virus-associated CNS lymphomagenesis [23, 24]. After the widespread introduction of effective antiretroviral therapy, the incidence of AIDS-related PCNSL declined substantially [25, 26]. In contrast, population-based analyses have documented a continuing age-specific rise among immunocompetent older adults; in the United States, incidence among adults aged 65 years or older increased by approximately 1.6%−1.7% annually in the modern registry era [18].

Immunodeficiency remains the most firmly established risk factor for PCNSL. Relevant clinical settings include HIV/AIDS, post-transplant immunosuppression, congenital immunodeficiency, and iatrogenic immune suppression [27, 28]. However, most contemporary PCNSL cases occur in immunocompetent older adults. The biological basis of this age-associated increase remains incompletely understood, but proposed contributors include immunosenescence, chronic antigenic stimulation, impaired immune surveillance, and age-related changes in the CNS microenvironment [29].

This review centers on PCNSL in immunocompetent adults and addresses immunodeficiency-associated disease where its distinct biology, epidemiology, diagnostic work-up, prognostic assessment, or management informs clinical practice.

Molecular pathogenesis

The molecular pathogenesis of PCNSL is characterized by constitutive B-cell receptor and Toll-like receptor signaling, activation of the NF-κB pathway, impaired antigen presentation, and adaptation to the immune-specialized CNS microenvironment. Most immunocompetent PCNSL is histologically classified as diffuse large B-cell lymphoma. In WHO-HAEM5, primary CNS, vitreoretinal, and testicular large B-cell lymphomas are grouped under the umbrella of “large B-cell lymphomas of immune-privileged sites,” whereas the International Consensus Classification retains primary diffuse large B-cell lymphoma of the CNS and other site-defined entities while recognizing their close biological overlap [3, 30]. The major pathogenic mechanisms are summarized in Fig. 1 [15, 31–33].

Fig. 1.

Molecular pathogenesis of primary central nervous system lymphoma. This schematic summarizes the major biological programs that converge on PCNSL B-cell survival. BCR and TLR/IL-1R signaling are driven by recurrent alterations such as CD79B mutation and MYD88 L265P, with downstream convergence through BTK, IRAK4, and CARD11, together with loss-of-function alterations in TNFAIP3. Cooperative genetic and epigenetic lesions, including PIM1, PRDM1, CDKN2A/B loss, BCL2/MCL1 dysregulation, and epigenetic remodeling, promote proliferation, apoptosis resistance, transcriptional reprogramming, and treatment resistance. Immune escape is mediated by impaired antigen presentation involving HLA class I/II, B2M, and CIITA alterations, together with checkpoint-mediated suppression through CD274/PDCD1LG2 and PD-L1/PD-L2 expression. The CNS perivascular niche provides survival and immune-modulatory cues, reflected by CXCL13, IL-10, and the IL-10/IL-6 ratio. These mechanisms support liquid biopsy using CSF, vitreous humor, or aqueous humor for MYD88 L265P, multigene tumor DNA, clonality, cytokine profiling, and molecular residual disease monitoring. Representative therapeutic strategies include BTK inhibitors, investigational IRAK4 inhibitors, PD-1 blockade, immunomodulatory drugs, mTOR inhibitors, CD19-directed CAR-T-cell therapy, and other emerging approaches. Abbreviations: B2M, beta-2 microglobulin; BCR, B-cell receptor; BTK, Bruton tyrosine kinase; CAR-T, chimeric antigen receptor T-cell therapy; CARD11, caspase recruitment domain family member 11; CIITA, class II major histocompatibility complex transactivator; CNS, central nervous system; CSF, cerebrospinal fluid; CXCL13, C-X-C motif chemokine ligand 13; HLA, human leukocyte antigen; IL, interleukin; IL-1R, interleukin-1 receptor; IRAK4, interleukin-1 receptor-associated kinase 4; MRD, molecular residual disease; mTOR, mammalian target of rapamycin; PCNSL, primary central nervous system lymphoma; PD-1, programmed cell death protein 1; PD-L1/PD-L2, programmed death-ligand 1/2; TLR, Toll-like receptor; TNFAIP3, tumor necrosis factor alpha-induced protein 3; NF-κB, nuclear factor kappa B; IMiDs, immunomodulatory drugs

Most PCNSL cases display a non-germinal center or activated B-cell-like phenotype, consistent with chronic activation of signaling pathways that support B-cell survival and proliferation [34, 35]. Genetically, PCNSL is enriched for the MCD/C5 program, characterized by co-occurring MYD88 and CD79B lesions, chronic active B-cell receptor/Toll-like receptor signaling, immune-evasion alterations, and tropism for immune-privileged extranodal sites [15, 36, 37]. Among recurrent alterations, MYD88 L265P is the most characteristic and clinically relevant mutation. MYD88 functions as an adaptor molecule downstream of Toll-like and interleukin-1 receptors, and the L265P mutation promotes assembly of the MYD88 signaling complex, leading to IRAK-mediated signaling, NF-κB activation, and enhanced lymphoma-cell survival [15, 38, 39]. CD79B mutations frequently coexist with MYD88 alterations and enhance B-cell receptor signaling by disrupting normal negative regulation of the B-cell receptor complex [15, 40]. Together, these alterations provide a biological rationale for Bruton tyrosine kinase inhibition, although mutation status alone is not yet sufficient for routine treatment selection [41].

Additional recurrent alterations affect PIM1, PRDM1, CARD11, TNFAIP3, CDKN2A/B, and genes involved in chromatin regulation or cell-cycle control [15, 32]. PIM1 mutations may contribute to survival signaling and genomic instability, whereas loss of TNFAIP3, a negative regulator of NF-κB signaling, may further reinforce pathway activation. Deletion or inactivation of CDKN2A/B can promote cell-cycle progression and may be associated with aggressive biological behavior [32, 37, 42, 43]. These findings indicate that PCNSL pathogenesis is driven by cooperative lesions converging on survival, proliferation, and treatment resistance rather than a single oncogenic driver.

Immune evasion is another central feature of PCNSL biology. Genetic or transcriptional disruption of antigen presentation pathways, including alterations involving HLA class I and class II loci, B2M, and CIITA, can reduce tumor antigen presentation and impair recognition by cytotoxic or helper T cells [15, 44]. Alterations involving CD274 and PDCD1LG2 have been reported in some molecular series, although other cohorts suggest that HLA loss is more prevalent than 9p24.1/PD-L1/PD-L2 alterations [37, 45]. These findings support the biological rationale for immune checkpoint blockade in selected patients, although clinical responses remain heterogeneous and predictive biomarkers require further validation [46, 47].

The CNS microenvironment further shapes lymphoma biology. PCNSL cells often grow in close association with perivascular spaces, reactive astrocytes, microglia, macrophages, and infiltrating lymphocytes. Chemokine and cytokine signaling, including CXCL13 and IL-10, may facilitate lymphoma-cell recruitment, survival, and immune modulation within the CNS compartment [48]. These soluble mediators are also clinically relevant because elevated IL-10, altered IL-10/IL-6 ratios, and increased CXCL13 levels in cerebrospinal fluid or ocular fluid can support diagnosis and disease monitoring when interpreted together with cytology, flow cytometry, imaging, and molecular testing [48, 49].

Epigenetic dysregulation and transcriptional reprogramming represent additional layers of PCNSL pathogenesis. A large integrative analysis identified four reproducible multi-omic consensus states. CS1 and CS2 were immune-cold and hypermethylated; CS1 showed high Polycomb repressive complex 2 activity, CDKN2A/B loss, and increased proliferation, whereas CS2 was enriched for epigenetic programs potentially relevant to immunomodulatory or demethylating approaches. CS3 displayed a heterogeneous immune phenotype, HIST1H1E enrichment, meningeal infiltration, and the least favorable prognosis. CS4 was immune-hot, enriched for JAK-STAT and NF-κB-activating alterations, and had the most favorable outcome [32]. These states remain investigational but provide a framework for biological stratification and subtype-directed trials.

Immunodeficiency-associated Epstein-Barr virus-positive PCNSL constitutes a distinct biological group rather than a simple extension of immunocompetent PCNSL. EBV-positive tumors often lack the canonical MYD88, CD79B, PIM1, CDKN2A, and HLA alterations of EBV-negative disease and instead show SOCS1 mutations, gain-of-function NOTCH-pathway lesions, IL-10 and JAK2-related signaling, reduced promoter methylation, and a tolerogenic microenvironment enriched for regulatory T cells and M2-like macrophages [50, 51].

From a translational perspective, MYD88 L265P is currently the most clinically mature diagnostic molecular marker in PCNSL, primarily because it supports diagnosis, can be detected in cerebrospinal fluid or ocular fluid, and reflects the central role of Toll-like receptor and NF-κB signaling in disease biology [16, 34, 39, 52]. CD79B, HLA pathway alterations, and PD-L1/PD-L2-related changes are biologically important and may eventually refine patient selection for BTK inhibitors or immune checkpoint inhibitors, but prospective validation is still required. Therefore, molecular pathogenesis should be viewed as a framework for improving diagnosis, risk stratification, and clinical trial design rather than as an established standalone algorithm for routine therapeutic selection.

Clinical manifestations and imaging features

The clinical presentation of PCNSL is heterogeneous and depends mainly on the anatomical compartment involved. Because the disease most commonly affects the brain parenchyma, focal neurological deficits, cognitive or behavioral changes, headache, seizures, and signs of increased intracranial pressure are frequent initial manifestations [5]. Symptoms often evolve over weeks rather than months, which may help distinguish PCNSL from slowly progressive inflammatory, infectious, or degenerative disorders. However, the clinical features are nonspecific, and diagnostic delay remains common, particularly in older patients or in those with atypical neuropsychiatric presentations [5, 12].

Anatomical involvement provides a useful framework for clinical assessment. Parenchymal PCNSL typically presents with focal deficits, headache, or mental status changes and is usually evaluated by contrast-enhanced brain MRI. Leptomeningeal involvement may present with cranial neuropathy, encephalopathy, radicular symptoms, or may be clinically silent; cerebrospinal fluid analysis and contrast-enhanced MRI can support diagnosis, although cytology alone has limited sensitivity. Primary vitreoretinal lymphoma usually presents with blurred vision, floaters, or visual loss and may mimic chronic uveitis. Spinal cord involvement is rare and may cause weakness, sensory deficits, or radicular pain; tissue confirmation can be challenging because biopsy carries substantial neurological risk. These clinical patterns and diagnostic clues are summarized in Table 2 [5, 53, 54].

Table 2.

Compartment-specific clinical presentations and diagnostic priorities in primary CNS lymphoma

Anatomical compartment	Typical presentation	Key diagnostic tests and adjuncts	Key practical caveat
Brain parenchyma	Focal deficit, cognitive or behavioral change, headache, seizure, or altered mental status	Contrast MRI with DWI/ADC, FLAIR, perfusion and SWI; stereotactic biopsy; systemic FDG-PET/CT	Avoid corticosteroids before biopsy when clinically feasible; routine extensive resection is not standard
Leptomeninges/CSF	Cranial neuropathy, encephalopathy, radiculopathy, headache, or occult CSF involvement	CSF cytology and flow cytometry; MYD88/CD79B or multigene ctDNA; IL-10, IL-10/IL-6 and CXCL13; contrast brain/spine MRI	Cytology alone has limited sensitivity; repeat and molecular testing may increase yield
Vitreous/retina	Floaters, blurred vision, reduced acuity, or chronic uveitis-like symptoms	Slit-lamp examination; vitreous biopsy; ocular-fluid cytology/flow, MYD88, clonality, IL-10 and IL-10/IL-6	Samples are often paucicellular and fragile; integrate pathology, cytokines, and molecular results
Spinal cord/nerve roots	Myelopathy, weakness, sensory loss, radicular pain, or sphincter dysfunction	Contrast spine MRI; CSF studies; systemic staging; biopsy only when diagnostic benefit outweighs neurological risk	Exclude secondary lymphoma, inflammatory myelitis, infection, and other intramedullary tumors

The table is intentionally concise and highlights compartment-specific diagnostic pathways. Anatomical distribution affects presentation, sampling strategy, and local morbidity but does not necessarily define a separate histological subtype or systemic treatment approach

Abbreviations: ADC apparent diffusion coefficient, CNS central nervous system, CSF cerebrospinal fluid, ctDNA circulating tumor DNA, DWI diffusion-weighted imaging, FDG-PET/CT fluorodeoxyglucose positron emission tomography/computed tomography, FLAIR fluid-attenuated inversion recovery, IL interleukin, MRI magnetic resonance imaging, SWI susceptibility-weighted imaging

Neuroimaging is central to the diagnostic work-up. Contrast-enhanced MRI of the brain is the preferred imaging modality for suspected PCNSL [12]. In immunocompetent patients, typical lesions are solitary or multifocal, frequently located in deep periventricular white matter, corpus callosum, basal ganglia, thalamus, or other deep structures. On non-contrast CT, lesions are often hyperdense because of high tumor cellularity. On contrast-enhanced T1-weighted MRI, PCNSL classically demonstrates avid and relatively homogeneous enhancement. On T2-weighted and FLAIR sequences, lesions may appear iso- to hypointense or moderately hyperintense relative to gray matter, often with less vasogenic edema than expected for the degree of enhancement [12, 55].

Diffusion-weighted imaging is highly informative because PCNSL commonly shows restricted diffusion with high signal on DWI and low apparent diffusion coefficient values, reflecting densely packed lymphoma cells and a high nuclear-to-cytoplasmic ratio[56, 57]. These features can support differentiation from high-grade glioma, metastasis, tumefactive demyelination, and abscess, although overlap exists and imaging alone is insufficient for definitive diagnosis. Perfusion MRI commonly shows lower relative cerebral blood volume than glioblastoma. Susceptibility-weighted imaging may demonstrate a relative paucity of intratumoral hemorrhage in treatment-naive immunocompetent PCNSL, while FDG-PET typically demonstrates intense glucose uptake and may assist in lesion characterization and systemic staging [55, 58].

Although homogeneous enhancement is the classic imaging pattern, atypical presentations occur. Ring enhancement can be observed, particularly in immunocompromised patients, in lesions with necrosis, or in R/R disease; therefore, its presence should not exclude PCNSL from the differential diagnosis [55, 59, 60]. Hemorrhage, necrosis, or marked calcification are less typical in treatment-naive immunocompetent PCNSL and should prompt consideration of alternative diagnoses, while still being interpreted in the broader clinical context.

Representative multimodal imaging findings from the same patient are shown in Fig. 2. Non-contrast CT demonstrates a hyperdense lesion in the deep posterior periventricular region with involvement of the corpus callosum (Fig. 2A). On MRI, the lesion is relatively hypointense on unenhanced T1-weighted imaging (Fig. 2B), moderately hyperintense on T2-weighted imaging (Fig. 2C), and demonstrates avid, predominantly homogeneous enhancement on axial and sagittal contrast-enhanced T1-weighted images (Fig. 2D and E). The lesion shows high signal on DWI with corresponding low signal on the ADC map, confirming restricted diffusion (Fig. 2F and G). FLAIR demonstrates hyperintensity of the lesion and adjacent periventricular signal abnormality (Fig. 2H). SWI from the same case was not available and is not shown; SWI nevertheless remains useful for assessing intratumoral hemorrhage in suspected PCNSL [12, 58, 61].

Fig. 2.

Representative multimodal neuroimaging findings in primary central nervous system lymphoma. All images are from the same patient. A Non-contrast CT shows a hyperdense lesion in the deep posterior periventricular region with involvement of the corpus callosum. B Axial unenhanced T1-weighted MRI shows relative hypointensity. C Axial T2-weighted MRI shows moderate hyperintensity with limited surrounding edema. D Axial contrast-enhanced T1-weighted MRI demonstrates avid, predominantly homogeneous enhancement. E Sagittal contrast-enhanced T1-weighted MRI demonstrates the periventricular and callosal distribution of the lesion. F DWI shows high signal intensity, with corresponding low signal on the ADC map G, indicating restricted diffusion. H FLAIR shows hyperintensity of the lesion and adjacent periventricular signal abnormality. SWI from the same patient was not available and is not shown. Abbreviations: ADC, apparent diffusion coefficient; CT, computed tomography; DWI, diffusion-weighted imaging; FLAIR, fluid-attenuated inversion recovery; MRI, magnetic resonance imaging; PCNSL, primary central nervous system lymphoma; SWI, susceptibility-weighted imaging

Diagnostic work-up and liquid biopsy

The diagnosis of PCNSL requires integration of neuroimaging, tissue pathology, cerebrospinal fluid evaluation, ocular assessment, and systemic staging. Because clinical and radiographic features are suggestive but not specific, histopathological confirmation remains the diagnostic standard whenever safely feasible [12, 58]. In patients with suspected parenchymal PCNSL, stereotactic biopsy is generally preferred over extensive surgical resection, as cytoreductive surgery has not been established as a routine therapeutic strategy and may increase morbidity in deep or multifocal lesions [12, 62].

Corticosteroids should be avoided before diagnostic biopsy whenever clinically feasible because PCNSL is highly steroid-sensitive and corticosteroid exposure can cause rapid radiographic regression, cytolysis, and nondiagnostic tissue sampling [63]. When corticosteroids are necessary because of life-threatening mass effect or severe neurological deterioration, the timing, dose, and duration should be documented, and biopsy should be pursued as soon as safely possible after multidisciplinary discussion.

Baseline diagnostic evaluation should include contrast-enhanced MRI of the brain, ophthalmological examination including slit-lamp evaluation, cerebrospinal fluid analysis when not contraindicated, and systemic staging to exclude secondary CNS involvement by systemic lymphoma [12, 58]. CSF studies should include cytology, flow cytometry, protein and glucose levels, and, when available, molecular and cytokine assays. Contrast-enhanced spine MRI is indicated when spinal cord, cauda equina, nerve-root, or leptomeningeal symptoms are present. Whole-body FDG-PET/CT is increasingly favored for systemic staging because it can detect occult systemic lymphoma and guide biopsy of extracranial disease when present[58]. Bone marrow biopsy may be considered when systemic involvement is suspected or when PET/CT findings are equivocal, but its routine diagnostic yield in otherwise isolated PCNSL appears limited [64].

Pathologically, most PCNSL cases are DLBCL. Tumor cells typically grow in sheets and frequently show angiocentric or perivascular infiltration. Immunohistochemistry usually demonstrates B-cell markers such as CD20, CD79a, PAX5, and CD19, with frequent expression of MUM1/IRF4 and variable expression of BCL6, BCL2, and MYC. CD10 expression is usually absent or uncommon, consistent with the predominance of a non-germinal center or activated B-cell-like phenotype [32, 34]. EBV-encoded RNA in situ hybridization is particularly important in immunodeficiency-associated cases, where EBV positivity is more frequent and may support the diagnosis in the appropriate clinical setting [23, 51].

For suspected primary vitreoretinal lymphoma, diagnostic evaluation should be coordinated with an experienced ophthalmologist. Vitreous biopsy remains a key diagnostic procedure, but its sensitivity may be limited by low cellularity, prior corticosteroid exposure, and cell fragility. Ancillary testing of vitreous or aqueous humor, including IL-10, IL-6, IL-10/IL-6 ratio, flow cytometry, clonality assessment, and MYD88 L265P detection, can improve diagnostic confidence when interpreted together with clinical and imaging findings [53].

Liquid biopsy has become an important adjunct in the diagnosis and monitoring of PCNSL, particularly when tissue biopsy is high-risk, nondiagnostic, or anatomically difficult. Targeted detection of MYD88 L265P remains the most clinically mature approach and can be performed in cerebrospinal fluid, vitreous humor, or aqueous humor using allele-specific PCR, digital PCR, or next-generation sequencing [16]. However, contemporary assays increasingly extend beyond a single hotspot to include CD79B and other recurrent mutations, immunoglobulin clonality, multigene circulating tumor DNA profiles, phased-variant sequencing, and methylation-based classification [39, 65–68]. Sensitivity varies with tumor burden, proximity to CSF spaces, leptomeningeal involvement, sample volume, processing time, and assay platform; therefore, a negative result does not exclude PCNSL.

Prospective and translational studies illustrate the complementary value of these platforms. Rapid CSF genotyping detected canonical tumor mutations in 42% of patients with CNS neoplasms and shortened the median time to disease-directed treatment from 12 to 3 days when biopsy could be avoided[65]. Ultrasensitive phased-variant sequencing detected pretreatment ctDNA in 78% of plasma and 100% of CSF samples in a brain-lymphoma cohort and enabled molecular classification and longitudinal monitoring [66]. Methylation markers such as cg0504 and SCG3 distinguished PCNSL tissue from other CNS tumors with an area under the curve of 1.00, although plasma sensitivity in the initial pilot series was only 20%, underscoring the developmental status of methylation-based liquid biopsy [68].

Cytokine analysis can further strengthen the diagnostic value of liquid biopsy. Elevated CSF IL-10, increased CXCL13, and a high IL-10/IL-6 ratio have been associated with PCNSL and may help distinguish lymphoma from inflammatory or infectious mimics [48, 49]. In ocular disease, elevated IL-10 or an increased IL-10/IL-6 ratio in vitreous or aqueous humor supports the diagnosis of primary vitreoretinal lymphoma, particularly when cytology is equivocal [53, 69]. However, cytokine markers should not be used in isolation because inflammatory disorders and technical variability may affect interpretation.

Beyond diagnosis, serial liquid biopsy may provide a minimally invasive approach for response assessment and molecular residual disease monitoring. Clearance of tumor-derived variants from CSF, ocular fluid, or plasma may correlate with response, whereas persistent or re-emergent ctDNA can precede clinical or radiographic relapse in selected patients [16, 66, 67, 69]. Heger and colleagues integrated baseline clinical factors, radiographic response, and peripheral residual disease into the molecular prognostic index for CNS lymphoma (MOP-C), which was strongly associated with failure-free survival across risk groups [67]. These approaches are not yet substitutes for tissue diagnosis or standardized response assessment; prospective harmonization of sampling, sequencing depth, analytical thresholds, and clinically actionable decision rules remains necessary.

Standard therapeutic framework

The treatment of PCNSL should be individualized according to age, performance status, renal function, frailty, comorbidities, neurocognitive reserve, disease extent, and patient preference. In contemporary practice, treatment is usually conceptualized as induction therapy, response-adapted consolidation, and, in selected patients, maintenance therapy. A proposed diagnostic and therapeutic decision framework is shown in Fig. 3 [1, 12, 16, 70].

Fig. 3.

Diagnostic and therapeutic decision framework for primary central nervous system lymphoma. This framework outlines an integrated, response-adapted, and fitness-adapted approach to PCNSL management. Evaluation begins with standardized MRI, including contrast-enhanced MRI, DWI/ADC, FLAIR, and assessment of mass effect. Corticosteroids should be avoided before biopsy when feasible, and stereotactic biopsy is preferred for tissue confirmation with early multidisciplinary coordination. Diagnostic confirmation integrates DLBCL morphology, B-cell markers, MUM1/IRF4, EBV testing when indicated, molecular adjuncts such as MYD88 L265P, CD79B, clonality or NGS, CSF and ocular assessment, and systemic staging to exclude secondary CNS lymphoma. First-line treatment is guided by age, renal function, frailty, comorbidity, performance status, and neurocognitive reserve. Fit patients proceed to HD-MTX-based induction, response assessment, and consolidation with thiotepa-based HDC/ASCT, reduced-dose WBRT, chemotherapy consolidation, or selected maintenance. Older, frail, or HD-MTX-limited patients may receive dose-adjusted HD-MTX, lower-intensity regimens, targeted agents, palliative RT, clinical trials, or supportive care. R/R disease requires reassessment of prior response, transplant or cellular-therapy eligibility, prior WBRT, organ function, and patient goals. Abbreviations: ADC, apparent diffusion coefficient; ASCT, autologous stem cell transplantation; BTKi, Bruton tyrosine kinase inhibitor; CAR-T, chimeric antigen receptor T-cell therapy; CNS, central nervous system; CSF, cerebrospinal fluid; ctDNA, circulating tumor DNA; CXCL13, C-X-C motif chemokine ligand 13; DLBCL, diffuse large B-cell lymphoma; DWI, diffusion-weighted imaging; EBV, Epstein-Barr virus; ECOG, Eastern Cooperative Oncology Group; FDG-PET/CT, fluorodeoxyglucose positron emission tomography/computed tomography; FLAIR, fluid-attenuated inversion recovery; HDC, high-dose chemotherapy; HD-MTX, high-dose methotrexate; IL, interleukin; IMiDs, immunomodulatory drugs; KPS, Karnofsky performance status; MRD, molecular residual disease; MRI, magnetic resonance imaging; mTOR, mammalian target of rapamycin; NGS, next-generation sequencing; PCNSL, primary central nervous system lymphoma; PD-1, programmed cell death protein 1; R/R, relapsed or refractory; RT, radiotherapy; WBRT, whole-brain radiotherapy; MATRix, methotrexate, cytarabine, thiotepa, and rituximab; R-MPV, rituximab, methotrexate, procarbazine, and vincristine; MUM1/IRF4, multiple myeloma oncogene 1/interferon regulatory factor 4

Surgery has a limited therapeutic role in PCNSL. Because most lesions are deep-seated, multifocal, and highly sensitive to chemotherapy and radiotherapy, extensive resection is not routinely recommended as definitive therapy. Surgical intervention is primarily used to obtain diagnostic tissue, relieve life-threatening mass effect in selected cases, or address diagnostic uncertainty when stereotactic biopsy is not feasible [12, 62]. Therefore, systemic CNS-penetrant therapy remains the foundation of disease control.

High-dose methotrexate (HD-MTX)-based chemotherapy is the backbone of first-line induction therapy for newly diagnosed PCNSL in patients who can tolerate it. HD-MTX doses of 3 g/m² or higher are commonly used to achieve therapeutic CNS concentrations, and treatment requires hydration, urine alkalinization, leucovorin rescue, renal function monitoring, and avoidance of interacting medications [12, 71]. HD-MTX is commonly combined with other CNS-penetrant agents. MATRix comprises methotrexate, cytarabine, thiotepa, and rituximab; R-MPV comprises rituximab, methotrexate, procarbazine, and vincristine [7, 72, 73]. In IELSG32, MATRix achieved a complete remission rate of 49% [73]; among patients who received MATRix followed by consolidation, the 7-year overall survival rate was 70% [10]. The optimal regimen depends on age, renal function, performance status, comorbidity profile, and treatment goals.

Consolidation therapy is considered after response to induction because relapse is common after induction alone. For fit patients, thiotepa-based high-dose chemotherapy followed by autologous stem cell transplantation is an important consolidation strategy and may reduce long-term neurotoxicity compared with conventional-dose whole-brain radiotherapy in appropriately selected patients [7, 72]. In the long-term PRECIS analysis, 8-year event-free survival was 67% after ASCT versus 39% after WBRT, whereas 8-year overall survival was 69% and 65%, respectively [8]. Whole-brain radiotherapy remains effective but is limited by delayed neurocognitive toxicity, particularly in older patients. Reduced-dose whole-brain radiotherapy or non-myeloablative chemotherapy-based consolidation may be considered in selected patients, depending on response depth, age, neurocognitive risk, and transplant eligibility [70, 74, 75]. In Alliance 51101, the estimated 2-year progression-free survival from randomization was 73% with myeloablative consolidation and 51% with nonmyeloablative consolidation. However, in the landmark analysis restricted to patients who completed consolidation, the corresponding 2-year progression-free survival estimates were 86% and 71%, respectively, and the difference was not statistically significant [74].

Older or frail patients require a separate therapeutic strategy. Chronological age alone should not determine eligibility for intensive therapy, but renal impairment, poor performance status, frailty, and comorbidities often limit the use of HD-MTX, cytarabine, thiotepa-based transplantation, or radiotherapy [1, 12, 76]. In patients who remain eligible for HD-MTX, dose-adjusted or reduced-intensity HD-MTX-based regimens may be used with close monitoring. For patients who are ineligible for HD-MTX or intensive consolidation, treatment options may include temozolomide, rituximab-containing regimens, lenalidomide, Bruton tyrosine kinase inhibitors, reduced-dose or palliative radiotherapy, clinical trial enrollment, or best supportive care, depending on goals of care and expected tolerance [77].

Maintenance therapy remains an evolving rather than universally standardized component of care. Lenalidomide, temozolomide, and Bruton tyrosine kinase inhibitors have been explored as maintenance strategies, particularly in older patients or patients unable to undergo intensive consolidation [78–83]. Available studies differ substantially in induction regimen, eligibility, agent, dose, duration, and outcome definition; maintenance should therefore be individualized according to response depth, treatment tolerance, relapse risk, and clinical trial availability. The potential application of maintenance therapy in older patients who are ineligible for intensive consolidation is discussed further in Section "Management of R/R disease and older patients with PCNSL".

Overall, standard therapy remains anchored by HD-MTX-based induction and response-adapted consolidation, but the field is moving toward more individualized treatment strategies. Molecular profiling, liquid biopsy, neurocognitive risk assessment, and frailty evaluation may increasingly inform treatment selection, although prospective evidence is still needed before these tools can define a fully molecularly guided treatment algorithm.

Targeted and immune-based therapies

The identification of recurrent B-cell receptor/Toll-like receptor pathway activation, NF-κB signaling, immune evasion, and CNS microenvironmental adaptation has provided a biological rationale for targeted and immune-based therapies in PCNSL. Most agents were initially evaluated in R/R disease, and evidence remains dominated by early-phase trials, single-arm studies, retrospective cohorts, or small series. Table 3 provides per-study quantitative efficacy outcomes and references to facilitate interpretation; cross-study comparisons should remain cautious because patient selection, prior therapy, response criteria, and follow-up differ substantially [1, 12].

Table 3.

Quantitative outcomes of selected targeted, immune-based, and cellular therapies in primary CNS lymphoma

Therapy	Class and setting	Study/design	Quantitative efficacy	Key safety or interpretive limitation	References
Ibrutinib	BTK inhibitor; R/R PCNSL	Phase I; 13 PCNSL	ORR 77% (10/13); CR 38% (5/13); median PFS approximately 4.6 months	Responses often short; invasive aspergillosis risk increases with intensive chemotherapy and corticosteroids	[41, 84]
Tirabrutinib	Selective BTK inhibitor; R/R PCNSL	Multicenter phase I/II; 44 patients	ORR 63.6%; CR 20.5% and CRu 15.9%; median PFS 2.9 months; median DOR 9.2 months	Acquired resistance; pneumonitis/interstitial lung disease and infection require monitoring	[85–87]
Zanubrutinib plus cytarabine	BTK inhibitor combination; R/R PCNSL	Prospective phase II; 34 patients	ORR 64.7%; CR/CRu 47.1%; median PFS 4.5 months; median OS 18 months; median DOR 9 months	Single-center study; cytopenias, especially grade 3/4 thrombocytopenia; regimen-specific rather than single-agent effect	[88]
Emavusertib (CA-4948)	IRAK4 inhibitor; investigational targeting of the MYD88 pathway	Preclinical CNS lymphoma models and early clinical investigation	Blood–brain barrier penetration and single-agent preclinical activity; mature PCNSL-specific ORR/PFS data unavailable	Should be described as investigational; clinical efficacy and optimal combinations are unestablished	[89]
Lenalidomide plus rituximab	Immunomodulatory therapy; R/R PCNSL or PVRL	REVRI phase II; 45 evaluable patients	End-of-induction ORR 35.6%; CR/CRu 29%; median PFS 7.8 months; median OS 17.7 months	Myelosuppression, infection, and thrombosis; only 5 patients completed planned maintenance	[90]
Pomalidomide plus dexamethasone	Immunomodulatory therapy; R/R PCNSL or PVRL	Phase I; 25 evaluable patients	ORR 48%; CR/CRu 32%; median PFS 5.3 months (9 months among responders)	Small study; hematologic and infectious toxicity; pseudoprogression reported	[91]
Nivolumab	PD-1 inhibitor; R/R PCNSL	Multicenter retrospective series; 22 patients	ORR 41%; CR 27%; median PFS 2.1 months; median OS 18.9 months; median DOR 20.9 months among responders	Retrospective design and potential patient-selection bias; results were not reproduced in the larger prospective study	[92]
Nivolumab	PD-1 inhibitor; R/R PCNSL	Phase II CheckMate 647; 47 patients	BICR-assessed ORR 6.4%; median PFS 1.41 months; median OS 6.77 months	Limited single-agent activity in an unselected cohort; registry-reported results; predictive biomarkers remain unresolved	[93]
Pembrolizumab	PD-1 inhibitor; R/R PCNSL/PVRL	AcSe phase II conference cohort; 50 patients	ORR 26%; CR 16%; median PFS 2.6 months; 6-month OS 60.4%; median DOR 10 months	Conference-level report; short median PFS; biomarkers and optimal sequencing unresolved	[94]
Temsirolimus	mTOR inhibitor; R/R PCNSL	Phase II; 37 patients	ORR 54%; CR/CRu 21.5%; median PFS 2.1 months; median OS approximately 3.7 months	Short-lived benefit; metabolic toxicity, cytopenias, infection, and treatment-related mortality	[95]
Tisagenlecleucel	CD19 CAR-T; R/R PCNSL	Prospective phase I/II; 12 patients	ORR 58.3%; CR 50%; 3 ongoing CRs at data cutoff; 3/7 initial responders had sustained remission	Grade 1 CRS 58.3%; ICANS 41.6%, including one grade 3 event; small cohort	[96]
Axicabtagene ciloleucel	CD19 CAR-T; R/R primary or secondary CNS lymphoma	Prospective pilot study reported as an ASCO meeting abstract; 18 patients (13 PCNSL)	ORR 94%; CR 67%; median DOR 13.4 months; median PFS 14.3 months; median OS 26.4 months	CRS 89% but no grade 3 + CRS; ICANS 44%, grade 3 + 28%; Ommaya-related infection in two patients	[97, 98]

Quantitative outcomes are presented from representative pivotal or informative studies. Values should not be compared indirectly across rows because patient selection, disease compartment, prior therapy, response criteria, sample size, and follow-up differ. Emavusertib remains investigational because mature PCNSL-specific clinical efficacy data are unavailable. In February 2026, the US prescribing-information limitation excluding PCNSL from axicabtagene ciloleucel use was removed

Abbreviations: BTK Bruton tyrosine kinase, CAR-T chimeric antigen receptor T-cell therapy, CNS central nervous system, CR complete response, CRS cytokine release syndrome, CRu unconfirmed complete response, DOR duration of response, ICANS immune effector cell-associated neurotoxicity syndrome, IRAK4 interleukin-1 receptor-associated kinase 4, mTOR mammalian target of rapamycin, ORR overall response rate, OS overall survival, PCNSL primary central nervous system lymphoma, PD-1 programmed cell death protein 1, PFS progression-free survival, PVRL primary vitreoretinal lymphoma, R/R relapsed or refractory, BICR blinded independent central review

Bruton tyrosine kinase inhibitors are among the most extensively studied targeted agents in PCNSL. In a phase I study of 13 patients with R/R PCNSL, ibrutinib monotherapy produced an overall response rate of 77%, including complete responses in 38%, with a median progression-free survival of approximately 4.6 months [41]. Responses were often not durable, and invasive aspergillosis emerged as an important safety concern when ibrutinib was combined with intensive chemotherapy or corticosteroid-containing regimens [84]. These data support BTK inhibition as a biologically rational salvage strategy while emphasizing antifungal risk assessment, drug interactions, and optimization of combination schedules.

Tirabrutinib is a more selective irreversible BTK inhibitor approved in Japan for R/R PCNSL. In the multicenter phase I/II study, the overall response rate was 63.6%, including complete or unconfirmed complete responses in 36.4%; median progression-free survival was 2.9 months and median duration of response was 9.2 months [85–87]. In a prospective phase II study, zanubrutinib plus cytarabine achieved an overall response rate of 64.7%, a complete or unconfirmed complete response rate of 47.1%, median progression-free survival of 4.5 months, and median overall survival of 18 months [88]. These results are encouraging but heterogeneous, and prospective studies should determine how efficacy relates to MYD88/CD79B status, CNS pharmacokinetics, prior methotrexate exposure, and acquired resistance.

Targeting downstream MYD88 signaling represents another emerging strategy. IRAK4 is a key signaling mediator downstream of mutant MYD88, and IRAK4 inhibition may theoretically suppress NF-κB-driven survival signaling in PCNSL harboring MYD88 mutations. CA-4948, also known as emavusertib, has shown preclinical rationale and is under clinical investigation in B-cell malignancies, including CNS lymphoma settings [89]. However, because mature PCNSL-specific efficacy data are not yet available, IRAK4 inhibitors should be presented as investigational rather than established therapeutic agents.

Immunomodulatory drugs have shown activity in R/R PCNSL. In the prospective REVRI study, lenalidomide plus rituximab achieved an end-of-induction overall response rate of 35.6%, including complete or unconfirmed complete responses in 29%; median progression-free and overall survival were 7.8 and 17.7 months, respectively [90]. Pomalidomide plus dexamethasone produced an overall response rate of 48%, including complete or unconfirmed complete responses in 32%, with a median progression-free survival of 5.3 months [91]. Myelosuppression, infection, thrombosis, and corticosteroid-related toxicity require consideration, and optimal maintenance dose and duration remain uncertain.

Immune checkpoint blockade is biologically attractive because PCNSL can harbor immune-presentation defects and PD-L1/PD-L2-related alterations. Clinical results with nivolumab, however, have been inconsistent. A multicenter retrospective series of 22 patients reported an overall response rate of 41%, including complete responses in 27%, with a median progression-free survival of 2.1 months, a median overall survival of 18.9 months, and a median response duration of 20.9 months among responders [92]. In contrast, registry-reported results from the prospective phase II CheckMate 647 PCNSL cohort, which included 47 treated patients, showed a blinded independent central review-assessed overall response rate of 6.4%, a median progression-free survival of 1.41 months, and a median overall survival of 6.77 months [93]. These divergent findings indicate heterogeneous and generally limited single-agent activity in unselected R/R PCNSL, although a minority of selected patients may derive meaningful benefit. Patient selection, disease context, response assessment, and predictive biomarkers require further clarification. In the phase II AcSé pembrolizumab cohort, reported as a conference abstract, the overall response rate was 26%, the complete response rate was 16%, the median progression-free survival was 2.6 months, and the median duration of response was 10 months [94].

mTOR pathway inhibition has been evaluated with temsirolimus in R/R PCNSL. In a phase II study, the overall response rate was 54%, including complete or unconfirmed complete responses in 21.5%, but median progression-free survival was only 2.1 months and median overall survival was approximately 3.7 months [95]. The high initial response rate coupled with short durability, metabolic toxicity, cytopenias, and infection risk supports temsirolimus as a context-dependent salvage option rather than a durable standalone strategy.

CD19-directed chimeric antigen receptor T-cell therapy is increasingly incorporated into the management of selected patients with R/R CNS lymphoma at experienced centers. In a prospective phase I/II trial of tisagenlecleucel in 12 patients with relapsed PCNSL, the overall response rate was 58.3% and complete response rate 50%; three patients remained in ongoing complete remission at data cutoff. Grade 1 cytokine release syndrome occurred in 58.3%, ICANS in 41.6%, and grade 3 ICANS in one patient, with no treatment-related deaths [96]. In a prospective pilot study reported in an ASCO meeting abstract, axicabtagene ciloleucel was evaluated in 18 patients with primary or secondary CNS lymphoma. The overall response rate was 94%, the complete response rate was 67%, the median duration of response was 13.4 months, the median progression-free survival was 14.3 months, and the median overall survival was 26.4 months; grade 3 or higher ICANS occurred in 28%, whereas no grade 3 or higher cytokine release syndrome was observed [97]. On February 5, 2026, the FDA approved a revision to the US prescribing information for axicabtagene ciloleucel that removed the Limitation of Use for primary central nervous system lymphoma [98, 99].

Neurotoxicity requires distinction between systemic immune effector cell-associated neurotoxicity syndrome (ICANS) and tumor inflammation-associated neurotoxicity (TIAN). ICANS typically presents within the first week with diffuse encephalopathy, inattention, language dysfunction, tremor, seizures, or reduced consciousness and is often temporally associated with cytokine release syndrome. TIAN is a focal, lesion-centered inflammatory syndrome that may cause new or worsened neurological deficits and radiographic edema at pre-existing tumor sites. In a 56-patient CNS lymphoma cohort, TIAN occurred in 17.9% of patients, with a median onset of 3.5 days after CAR-T-cell infusion, and larger baseline CNS tumor volume predicted its occurrence. Compared with ICANS, TIAN was less frequently associated with cytokine release syndrome, occurring with CRS in 60% versus 100% of cases, respectively. Symptoms were usually transient and fully reversible, although one fatal outcome was reported [100]. Pseudoprogression can accompany local immune activation and should be distinguished from true progression through serial imaging and clinical correlation. Retrospective data suggest that greater dexamethasone exposure at the time of lymphodepletion may be associated with earlier CNS progression. Among patients requiring bridging therapy, ibrutinib-containing regimens were associated with longer CNS progression-free survival than non-ibrutinib bridging regimens, although initial response patterns were similar and residual confounding cannot be excluded [101]. Peripherally enhancing lesions and leptomeningeal involvement have also been associated with shorter CNS progression-free survival after CAR-T therapy [102].

Overall, targeted and immune-based therapies have expanded the therapeutic landscape of PCNSL, especially for R/R disease and patients unable to receive intensive chemotherapy or transplantation. BTK inhibitors and immunomodulatory agents have clinically meaningful activity and are used as salvage options in selected patients; checkpoint inhibitors and mTOR inhibition provide context-dependent benefit; and CD19-directed CAR-T therapy is increasingly incorporated as a cellular-therapy option for eligible patients at experienced centers. Future trials should incorporate molecular profiling, liquid-biopsy monitoring, standardized neurotoxicity assessment, CNS pharmacokinetics, and rational sequencing or combination strategies.

Management of R/R disease and older patients with PCNSL

R/R PCNSL remains a major therapeutic challenge. A substantial proportion of patients relapse after HD-MTX-based induction, and outcomes are generally poor without effective salvage therapy [6]. Treatment selection should be guided by prior therapy, duration of first remission, performance status, age, renal function, neurocognitive reserve, prior whole-brain radiotherapy, transplant or cellular-therapy eligibility, and patient goals. Whenever feasible, patients with R/R disease should be considered for clinical trials or referral to centers with access to contemporary targeted and cellular therapies [12].

For patients who previously responded to HD-MTX and experienced a durable remission, HD-MTX rechallenge is a reasonable salvage option, particularly when relapse occurs after a prolonged treatment-free interval and renal function remains adequate [103]. In contrast, patients with primary refractory disease or early relapse after HD-MTX-based therapy are less likely to benefit from simple MTX rechallenge and should be considered for alternative salvage strategies. These may include cytarabine- or thiotepa-containing chemotherapy, ifosfamide-based regimens, targeted agents, immunomodulatory therapy, radiotherapy if not previously delivered, cellular therapy in selected centers, or clinical trial enrollment [12, 104–107].

For transplant-eligible patients with chemosensitive relapse, high-dose chemotherapy followed by autologous stem cell transplantation can provide meaningful disease control and should be considered, especially in patients who did not receive transplant as part of first-line consolidation [12, 104, 108]. Thiotepa-based conditioning regimens are generally preferred because of CNS penetration and established activity in PCNSL [104, 108, 109]. However, eligibility requires careful assessment of performance status, organ function, infection risk, prior treatment toxicity, and stem-cell reserve.

Whole-brain radiotherapy remains an effective salvage modality for patients who have not previously received cranial irradiation, particularly when rapid disease control is required or systemic therapy is not feasible [12, 110]. Its use must be balanced against the risk of delayed neurocognitive toxicity, which is particularly relevant in older patients and in those with pre-existing cognitive impairment. Reduced-dose or focal approaches may be considered in selected clinical contexts, but focal radiotherapy alone may be insufficient for multifocal or diffuse CNS disease [75, 110].

Targeted, immune-based, and cellular therapies are increasingly important in R/R PCNSL, especially when conventional intensive salvage therapy is unsuitable. BTK inhibitors, lenalidomide-based regimens, PD-1 inhibitors, mTOR inhibition, and CD19-directed CAR-T therapy have demonstrated activity, but the strength and durability of evidence vary across approaches. Quantitative study-level outcomes are summarized in Table 3 and should be interpreted according to line of therapy, disease status, prior treatment, CNS compartment, response criteria, and toxicity profile [94, 96–98, 101, 102, 106, 107].

Allogeneic hematopoietic stem cell transplantation is not a standard treatment for PCNSL and should not be presented as routine consolidation. Its potential rationale lies in a graft-versus-lymphoma effect, but available evidence is limited to highly selected patients, small retrospective series, and case reports [111]. Because of substantial risks, including graft-versus-host disease, infection, treatment-related mortality, and neurotoxicity, allo-HSCT should be considered only in exceptional R/R cases within experienced centers or clinical trials.

Elderly patients represent a distinct and increasingly common PCNSL population. Chronological age alone should not preclude active therapy, but older patients frequently have reduced renal reserve, frailty, comorbidities, impaired performance status, baseline cognitive vulnerability, and higher risk of treatment-related toxicity [76]. A geriatric-oriented assessment, including renal function, comorbidity burden, functional status, fall risk, cognition, social support, and medication interactions, is useful when determining treatment intensity [112].

For fit older patients with adequate renal function, HD-MTX-based therapy remains the preferred foundation of treatment, with dose adjustment and supportive care tailored to renal clearance and frailty. Intensive cytarabine-containing regimens, thiotepa-based transplantation, or whole-brain radiotherapy should be used selectively because of hematologic toxicity, infectious risk, and neurocognitive consequences [6, 11, 12, 76, 113]. For older or frail patients who are not candidates for intensive treatment, options may include reduced-intensity HD-MTX-based regimens, temozolomide, rituximab-containing approaches, lenalidomide, BTK inhibitors, palliative radiotherapy, or best supportive care, depending on expected benefit, toxicity, and patient preference [77].

Maintenance approaches are summarized in Section "Standard therapeutic framework". In older responders who are not candidates for intensive consolidation, maintenance may be considered after individualized discussion of expected benefit, treatment burden, hematologic and infectious toxicity, drug interactions, and patient preference; no universal maintenance standard has been established [78, 79, 81–83].

Overall, management of R/R and older patients requires a response-adapted and fitness-adapted strategy rather than a uniform algorithm. A durable prior response favors HD-MTX rechallenge; chemosensitive relapse may justify salvage chemotherapy followed by ASCT in eligible patients; and patients with early relapse, limited tolerance, or multiple prior therapies may benefit from targeted therapy, immunomodulatory therapy, CAR-T therapy, radiotherapy, clinical trials, or symptom-directed care. The therapeutic framework shown in Fig. 3 emphasizes this individualized approach.

Rare histological and biological subtypes

Although most PCNSL in immunocompetent adults is diffuse large B-cell lymphoma within the immune-privileged-site large B-cell lymphoma spectrum, uncommon histological and biological subtypes require separate consideration. This section focuses on entities defined primarily by lineage, histology, or immune context; compartment-specific presentations such as vitreoretinal, leptomeningeal, and spinal disease are summarized separately in Table 2 because anatomical site chiefly modifies presentation and diagnostic sampling rather than defining a distinct systemic treatment paradigm.

Immunocompetent primary CNS large B-cell lymphoma is typically positive for CD20, CD79a, PAX5, and MUM1/IRF4 and is enriched for the MCD/C5 genetic program, MYD88 L265P, CD79B, PIM1, PRDM1, CDKN2A/B loss, and antigen-presentation defects [15, 32, 37]. Primary vitreoretinal large B-cell lymphoma shares this biological spectrum and may precede, accompany, or follow parenchymal disease. Diagnosis is generally established by stereotactic biopsy or compartment-specific cytology, flow cytometry, molecular testing, and cytokine analysis; routine cytoreductive resection is not standard treatment [12, 53].

Primary CNS low-grade B-cell lymphomas, including dural marginal zone lymphoma and rare follicular or lymphoplasmacytic entities, are biologically distinct from parenchymal PCNS-DLBCL. They often show dural, localized, or leptomeningeal involvement and may follow an indolent course [114–116]. Management may include biopsy for diagnosis, involved-site radiotherapy for localized disease, rituximab-based systemic therapy, or observation in carefully selected patients. Surgical resection is not routinely required beyond diagnostic or exceptional decompressive indications.

Primary CNS T-cell lymphoma is rare and heterogeneous. Diagnosis requires adequate tissue for morphology, T-cell immunophenotyping, T-cell receptor clonality, EBV assessment when relevant, and exclusion of systemic T-cell lymphoma with secondary CNS involvement [117, 118]. Because prospective trials are lacking, treatment is generally adapted from PCNSL principles using CNS-penetrant HD-MTX-based therapy when feasible; the roles of subtype-specific targeted agents, radiotherapy, and transplantation remain poorly defined.

Immunodeficiency-associated PCNSL is clinicopathologically distinct and is EBV-positive in approximately 80% of contemporary international cases [23]. EBV-associated tumors often lack canonical alterations in MYD88 and CD79B and instead show JAK/STAT- and NOTCH-associated alterations, IL-10-related signaling, and a tolerogenic microenvironment [50, 51]. Management requires lymphoma-directed therapy together with immune reconstitution or reduction of immunosuppression when feasible, such as effective antiretroviral therapy in HIV-associated disease or carefully supervised modification of post-transplant immunosuppression.

CNS-predominant intravascular large B-cell lymphoma is an important differential diagnosis rather than strict PCNSL in many cases. Large neoplastic B cells are confined predominantly to small-vessel lumina, and systemic disease may be occult. Random skin biopsy, bone marrow evaluation, FDG-PET/CT, and systemic staging can be diagnostically valuable. Treatment is generally based on systemic aggressive-lymphoma regimens with CNS-directed components rather than on a PCNSL-only algorithm.

Anatomical presentations should be reported precisely because they affect diagnostic yield and local morbidity. Vitreoretinal disease may require ocular fluid analysis or vitreous biopsy; leptomeningeal disease may require repeated CSF cytology, flow cytometry, and molecular testing; and spinal cord or nerve-root disease may make tissue acquisition hazardous. Nevertheless, these locations do not by themselves establish a separate histological subtype, and therapy should be guided by lineage, biological classification, systemic staging, compartment involvement, and patient fitness.

Future studies should report rare histological and biological subtypes separately and avoid uncritical extrapolation from classic immunocompetent PCNS-DLBCL or systemic lymphoma. Table 1 summarizes the principal entities, characteristic biology, diagnostic priorities, and management considerations.

Prognosis and future directions

Despite therapeutic progress, PCNSL remains associated with a high risk of relapse, treatment-related toxicity, and long-term neurocognitive impairment. Prognosis is influenced by both disease biology and host factors, including age, performance status, renal function, tumor burden, deep brain involvement, cerebrospinal fluid abnormalities, immune status, and treatment tolerance [70, 119]. Conventional prognostic models remain useful for clinical risk stratification, but they do not fully capture molecular heterogeneity, frailty, neurocognitive vulnerability, or response kinetics.

The International Extranodal Lymphoma Study Group (IELSG) prognostic score incorporates five adverse factors: age older than 60 years, poor performance status, elevated serum lactate dehydrogenase, elevated cerebrospinal fluid protein concentration, and involvement of deep brain structures [120, 121]. The Memorial Sloan Kettering Cancer Center model stratifies patients primarily by age and Karnofsky performance status [119, 122]. Both models predate the molecular era and incorporate neither molecular profile nor liquid-biopsy status; they also do not capture frailty, neurocognitive vulnerability, or contemporary targeted and cellular therapies.

For immunodeficiency-associated PCNSL, a recent International Primary CNS Lymphoma Collaborative Group model identified age older than 60 years, Karnofsky performance status below 70, and EBV positivity as adverse variables. Patients with up to one, two, or all three adverse factors had median overall survival of 135, 29, and 3 months, respectively [23]. This model should be applied specifically to immunodeficiency-associated disease rather than extrapolated to immunocompetent PCNSL.

Emerging biomarkers may refine prognostication. Molecular alterations involving MYD88, CD79B, CDKN2A/B, HLA-related genes, B2M, CIITA, and immune checkpoint loci may provide information about tumor biology and therapeutic vulnerability, but their independent predictive value requires prospective validation [15, 32, 37, 123]. Liquid-biopsy dynamics may add temporal risk information: persistent or recurrent tumor-derived variants after treatment can indicate residual disease and increased relapse risk [16, 66, 67, 69]. The MOP-C model integrates clinical risk, radiographic response, and plasma peripheral residual disease and was strongly associated with failure-free survival, but external prospective validation and standardized assays are needed before routine treatment decisions [67].

Future clinical trials should move beyond uniform treatment algorithms and incorporate biologically informed stratification. Key priorities include defining biomarkers that predict response to Bruton tyrosine kinase inhibitors, immunomodulatory agents, immune checkpoint inhibitors, and cellular therapy; optimizing treatment intensity for older and frail patients; identifying safe consolidation and maintenance strategies that preserve neurocognitive function; and validating liquid biopsy for molecular residual disease monitoring [86, 124]. Given the rarity of PCNSL, international collaboration, harmonized response criteria, standardized neurocognitive assessment, and prospective biospecimen collection will be essential.

Another important future direction is the integration of patient-centered outcomes. Long-term survival alone is insufficient in PCNSL because treatment may cause substantial cognitive decline, functional impairment, ocular morbidity, infectious complications, and reduced quality of life. Future studies should therefore evaluate not only progression-free and overall survival, but also neurocognitive trajectories, frailty transitions, steroid exposure, treatment-related infections, visual outcomes in primary vitreoretinal lymphoma, and patient-reported quality of life [10, 125–127].

Conclusions

PCNSL is a biologically distinct and clinically challenging CNS lymphoma characterized by recurrent activation of B-cell receptor and Toll-like receptor signaling, NF-κB pathway dependence, immune evasion, and adaptation to the CNS microenvironment. Immunocompetent PCNS-DLBCL is closely related to the MCD/C5 genetic program and is classified within the WHO-HAEM5 immune-privileged-site large B-cell lymphoma spectrum. HD-MTX-based induction remains the foundation of first-line therapy for eligible patients, but relapse, treatment-related neurotoxicity, and limited therapeutic options for older or frail patients remain major obstacles.

Recent advances in genomic profiling, multigene and methylation-based liquid biopsy, and targeted therapy have reshaped the diagnostic and therapeutic landscape of PCNSL. MYD88 L265P remains an important minimally invasive marker, but broader ctDNA platforms can capture CD79B and additional recurrent alterations, support molecular classification, and enable longitudinal residual-disease assessment. BTK inhibitors, immunomodulatory agents, immune checkpoint inhibitors, mTOR inhibition, and CD19-directed CAR-T therapy have expanded treatment possibilities for R/R disease. CAR-T therapy is increasingly integrated at experienced centers, although optimal sequencing, durability, and management of ICANS and TIAN require further study.

Future management of PCNSL will likely depend on integrating clinical fitness, neurocognitive risk, molecular features, liquid biopsy dynamics, and treatment accessibility. A molecularly informed but clinically pragmatic approach is needed to improve survival while preserving neurological function and quality of life.

Abbreviations

ABC

Activated B-cell-like

ADC

Apparent diffusion coefficient

ADCC

Antibody-dependent cellular cytotoxicity

ASCT

Autologous stem cell transplantation

BBB

Blood-brain barrier

BCR

B-cell receptor

BTK

Bruton tyrosine kinase

CAR-T

Chimeric antigen receptor T-cell therapy

cfDNA

Cell-free DNA

CNS

Central nervous system

CR

Complete response

CRS

Cytokine release syndrome

CSF

Cerebrospinal fluid

CT

Computed tomography

ctDNA

Circulating tumor DNA

DLBCL

Diffuse large B-cell lymphoma

DOR

Duration of response

DWI

Diffusion-weighted imaging

EBV

Epstein-Barr virus

ECOG-PS

Eastern Cooperative Oncology Group performance status

FDG-PET/CT

Fluorodeoxyglucose positron emission tomography/computed tomography

FLAIR

Fluid-attenuated inversion recovery

GVHD

Graft-versus-host disease

HDC

High-dose chemotherapy

HD-MTX

High-dose methotrexate

HIV/AIDS

Human immunodeficiency virus/acquired immunodeficiency syndrome

ICANS

Immune effector cell-associated neurotoxicity syndrome

ICC

International Consensus Classification

IELSG

International Extranodal Lymphoma Study Group

IL

Interleukin

IPCG

International Primary CNS Lymphoma Collaborative Group

IP-LBCL

Primary large B-cell lymphoma of immune-privileged sites

IRAK4

Interleukin-1 receptor-associated kinase 4

KPS

Karnofsky performance status

MRI

Magnetic resonance imaging

MRD

Molecular residual disease

mTOR

Mammalian target of rapamycin

NF-κB

Nuclear factor kappa B

NHL

Non-Hodgkin lymphoma

ORR

Overall response rate

OS

Overall survival

PCNSL

Primary central nervous system lymphoma

PD-1

Programmed cell death protein 1

PD-L1

Programmed death-ligand 1

PFS

Progression-free survival

PR

Partial response

PVRL

Primary vitreoretinal lymphoma

R/R

Relapsed or refractory

SWI

Susceptibility-weighted imaging

TIAN

Tumor inflammation-associated neurotoxicity

TLR

Toll-like receptor

WBRT

Whole-brain radiotherapy

Author contributions

FY, PW, and XF were involved in writing, reading literature, and designing figures. YW and YS were involved in the overall supervision of the manuscript review and editing. YZ and YG were involved in literature review and critical revision of the manuscript. All authors gave approval before submission.

Funding

This research was supported by the Capital Medical University-Yingkang Life Science and Innovation Elite Plan project (2024KCJY0102); Beijing Natural Science Foundation-Haidian Original Innovation Joint Fund Project (L222022).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Competing interests

The authors declare no conflict of interest, financial or otherwise.