Pathology and new insights in central nervous system lymphomas

Purpose of review

Primary central nervous system lymphoma (PCNSL) is a rare central nervous system (CNS) malignancy, which represents a heterogenous group of tumors. Among PCNSL, diffuse large B-cell lymphoma of the CNS (CNS-DLBCL) represents the most common tumor type. Multiomics studies have recently revealed the complex genomic landscape of these rare diseases. These findings lead to a potential new molecular and epigenetic classification.

Recent findings

Our review is focused on CNS-DLBCL in immunocompetent patients. CNS-DLBCL are derived from self-reactive/polyreactive precursor cells. An early molecular event such as MYD88 mutation leads to escape elimination of precursor cells, which, by a dysregulated GC reaction, acquire auto-/polyreactivity of the B-cell tumoral cells for antigens physiologically expressed in the CNS. Most of CNS-DLBCL tumor cells harbor a non-GCB, ABC-like immunophenotype associated with a late GC (exit) B-cells genotype by gene expression profiling. Various mechanisms of genetic alterations are involved in the pathogenesis of PCNSL, including point mutations [nonsomatic hypermutation (SHM), aberrant SHM (aSHM)], SHM/aSHM, chromosome copy gains or losses, and DNA hypermethylation. Constitutive NFκB activation plays a key role in lymphoma cell proliferation and survival by dysregulation of toll-like receptor (mutations of CARD11 and MYD88), BCR (CD79B), JAK-STAT, and NFκB signaling pathways.

Summary

Multiomics approaches have succeeded to substantially improve the understanding of the pathogenesis, as well as the molecular and epigenetic events in PCNSL. Challenges remain due to the obvious heterogeneity of CNS-DLBCL, and improvement is needed for their classification.

INTRODUCTION

Primary central nervous system lymphoma (PCNSL), a subtype of extra-nodal non-Hodgkin's lymphoma, is a rare CNS malignancy, which represents only 2–4% of newly diagnosed CNS tumors in immunocompetent patients [1,2^▪▪]. In immunocompromised patients, PCNSLs are described as specific entities, grouped in the family of CNS lymphomas named “immunodeficiency-associated CNS lymphomas” in the 2021 CNS WHO classification [1,3]. Interestingly, these tumors have specific imaging characterized by multifocal lesions with necrosis and present specific pathogenesis. Indeed, most of immunodeficiency-associated lymphomas are EBV-related [1,3]. Molecularly, in immunocompetent patients, genetic and epigenetic alterations implicated in the PCNSL pathogenesis are involved in the late germinal center (GC) exit B-cell phenotype (CD10–BCL6+IRF4/MUM1+) [4,5]. Recurrent molecular alterations have not only been specifically described in the B-cell receptor (BCR) and toll-like receptor (TLR) pathway (Myeloid differentiation factor 88 (MYD88)), nuclear factor-kB (NF-κB) pathways (CD79B, MYD88 L265P, PIM1 mutations) but also in genes associated with chromatin structure and modification, cell-cycle regulation, and immune recognition [2^▪▪,3,4,6,7]. Among PCNSLs, diffuse large B-cell lymphomas of the CNS (CNS-DLBCLs) represent the most common tumor type (90%) with a predominantly nongerminal center B-cell-like (non-GCB) immunophenotype and share non-CNS activated B-cell (ABC)-DLBCL genetic alterations [8^▪▪]. The incidence rate of CNS-DLBCLs is 4–6% of all extra-nodal lymphomas with an overall annual incidence rate of 0.47 cases per 100 000 population [3,5]. Large-scale molecular profiling and multiomics studies highlighted that CNS-DLBCLs are molecularly different and have less favorable prognosis than its systemic counterpart, in part due to the location [2^▪▪,4]. Recent studies have underlined the molecular heterogeneity of systemic DLBCLs by deciphering different molecular clusters based on the presence of mutations (MYD88, CD79B, PIM1, BTG2 mutations), immunoglobulin heavy (IgH) locus and B-cell lymphoma 6 (IgH-BCL6) translocations, and chromosomes 6 and 9 copy gains and losses [2^▪▪,9–11]. PCNSLs seem to be more genetically related to the “MCD,” “C5,” or “MYD88-like” molecular clusters [8^▪▪]. Despite dramatic improvement in the molecular characterization of the PCNSLs, the understanding of the pathogenic mechanisms, especially noncoding and regulatory changes, is not fully understood [8^▪▪]. This review aims to provide recent advances in PCNSL, especially CNS-DLBCL found in immunocompetent patients, focusing on the clinico-pathological, molecular, and epigenetic data. 

Box 1.

no caption available

PATHOPHYSIOLOGY AND HISTOGENETIC ORIGIN OF PRIMARY CENTRAL NERVOUS SYSTEM LYMPHOMA

The understanding of PCNSL's pathogenesis is limited by its low incidence and the small size of biopsy samples [8^▪▪,12]. The exact pathogenesis of PCNSL and its restriction to the CNS compartment is not fully elucidated [8^▪▪,13]. Indeed, the histogenetic origin of the tumor cells is still debatable, as the CNS is known to not harbor lymphatic vessels or lymph nodes [13]. Interestingly, recent studies showed the presence of lymphatic vessels, parallel to dural venous sinuses, in the meninges of mammals [14]. Moreover, there is growing evidence about the existence of a “glymphatic system,” which uses periarterial spaces and constantly clears interstitial fluid into the cerebrospinal fluid through the brain parenchyma [14,15].

In 2019, Cai et al. [5] proposed the hypothesis that systemic B-cells are recruited to the brain during activation of the immune system and undergo malignant transformation in the CNS compartment or outside the CNS during a germinal center reaction in a lymphoid organ [16].

In the 2021 CNS WHO classification, the hypothesis is that the tumor cells are derived from precursor cells and harbor late germinal center exit B-cell phenotype [3,17] with expression of pan B-cell markers, that is, CD19 (100% of cases), CD20 (100%), and CD79a, germinal center -associated molecule BCL6 (60–100%) and post-germinal center associated markers, that is, MUM1/IRF4 (90–100%) with 10–20% of cases CD10+ [3,5,8^▪▪]. Absence of late post-germinal center marker CD138 expression associated with coexpression of BCL-6 and MUM-1 are correlated with the activated immunophenotype and the early post-germinal center origin of PCNSL [18].

The hypothesis proposed by Montesinos-Rongen et al.[19] (Fig. 1) is that an early molecular event such as MYD88 mutation leads to escape elimination of self-reactive/polyreactive precursor cells, which, by dysregulated germinal center reaction, acquire auto-/polyreactivity of the B-cell tumoral cells for antigens physiologically expressed in the CNS such as S100 and MPLZ1, MBP, and MOBP proteins (constituents of the myelin/oligodendrocytes mostly expressed in white matter) leading to uncontrolled proliferation. In case of normal maturation of a naive B-cell, upon germinal center, the somatic hypermutation (SHM) (which leads to modification of the IG and BCL6 genes) leads to the differentiation in plasma cells or memory B-cells [19]. The SHM plays a key role in the pathogenesis of PCNSLs. This phenomenon physiologically targets rearranged IG genes and BCL-6. In PCNSLs, other genes associated with tumorigenesis such as BCL2, MYC, PIM1, PAX5, RHOH, KLHL14, OSBPL10, and SUSD2 genes, are involved in SHM [4,9]. Through this aberrant SHM (aSHM) of proto-oncogenes and the alterations of DNA repair, methylation, and expression programs, there is an aberrant expression of genes implicated in TLR, BCR, and NF-κB pathways [12]. Following further molecular events, B-cells lose their ability to terminate SHM, decrease BCR antigen binding affinity, and increase auto-/polyreactivity of the B-cells (such as IgM+ memory B cells), including higher affinity for proteins expressed in the CNS compartment. This phenomenon gives the tumoral cells the ability to perfectly adapt to the CNS microenvironment [19]. Interestingly, upregulation of proteins that are recognized by the B-cells is observed in reaction of inflammation and tumoral infiltration in the CNS compartment such as endoglin and galectine-3 expressed by the CNS endothelial cells [12]. By contrast, in immunodeficient patient, the molecular pathogenesis is specifically different, as immunodeficiency-associated lymphomas are EBV-related [20].

FIGURE 1.

An early molecular event such as MYD88 mutation lead to escape elimination of self-reactive/polyreactive precursor cells which, by dysregulated GC reaction, acquire auto-/polyreactivity of the B-cell tumoral cells for antigens physiologically expressed in the CNS. Aberrant SHM (aSHM) followed by further molecular events give the tumoral cells the ability to perfectly adapt to the CNS microenvironment. Adapted from Montesinos-Rongen et al., Cai et al.[5,12,19]. CNS: central nervous system; GC: germinal center; aSHM: aberrant SHM; BCR: B-cell receptor.

Neurotropic viruses, that is, HHV6, HHV8, and SV40 have been excluded from having pathogenetically ability to activate B-cells in the CNS compartment [21].

RADIOLOGICAL AND PATHOLOGICAL FEATURES OF PRIMARY CENTRAL NERVOUS SYSTEM LYMPHOMA

As illustrated in Fig. 2, gold standard imaging modality is MRI including gadolinium injection. PCNSLs appear as a solitary lesion in 60–70% of cases, the remaining cases being a multifocal disease [3,22,23]. The typical locations are in the cerebral hemispheres (38%), deep brain region such as thalamus and basal ganglia (16%), corpus callosum (14%), and periventricular region (12%) or cerebellum (9%). Spinal cord involvement is exceptional [3,22,24–27]. The preferential location of PCNSLs for white matter of the cerebral hemispheres and deep brain regions could be explained by the affinity of B-cell tumor cells for white matter constituent MPLZ1, MBP, and MOBP proteins, as explained above [3,19]. PCNSLs are typically hypointense on T1-weighted MRI images and isointense to hyperintense on T2-weighted MRI images [28]. Contrast enhancement is variable and can harbor a rim enhancement corresponding to the central necrosis [29]. In immunodeficiency-associated lymphomas, multifocal lesions with large areas of necrosis are more common with almost all of these tumors being multifocal at diagnosis [3,23]. By contrast to diffuse gliomas and metastases, only a moderate amount of peritumoral edema is found in PCNSLs and these lesions are frequently restricted on diffusion-weighted images [22,28]. On the basis of a cohort of 100 PCNSLs, Wilhelm et al.[24] provided the conclusion that PCNSLs are contrast-enhancing lesions with a diameter of at least 15 mm in contact with the subarachnoid space. Recently, recommendations from the International PCNSL Collaborative Group (IPCG) have been published for standardized imaging practices, including contrast-enhanced MRI of the brain and spine (if spinal symptoms are present), ophthalmologic and cerebrospinal fluid evaluation [22,30].

FIGURE 2.

Radiological features. Left temporal (a, f), right parietal (b, g), right lenticulostriate (c, h), posterior right falcorial (d, i), corpus callosum (e, j) well defined tumoral growth hyperintense on gadolinium-enhanced T1-weighted (a, b, c, d, e), T2-weighted showing slight hyperintense and perilesional oedema (f, g, h, i, j) Δ: Primary CNS lymphoma.

Brain biopsies are considered to be the gold standard for the assessment of PCNSL diagnosis using histological and immunohistochemical analysis [3,31]. Table 1 listed the currently recognized PCNSL sub-groups in the 2021 CNS WHO classification. CNS-DLBCLs represent the most common tumor type (90%). The typical morphological aspect, as illustrated in Fig. 3, is a highly cellular, patternless tumor with commonly large areas of geographical necrosis and vasocentric growth pattern, mostly found in the periphery of the tumor. The front of the invasion can be well delineated or diffusely infiltrating with single cells infiltrating the tissue [3,29]. The tumoral cells are large, sometimes pleomorphic, cells with round to oval vesicular nuclei with fine chromatin and prominent nucleoli with usually scant cytoplasm, mimicking centroblasts or immunoblasts morphology [3]. Brisk mitotic activity and apoptotic bodies are frequent. Associated with tumoral cells, reactive astrocytes (GFAP+) and mixed reactive inflammatory population are found, including macrophages, B and cytotoxic T-cells lymphocytes (CD3+), and activated microglia (CD45, CD68+, HLA-DR+) [3]. This complex neuroimmune network represents the hallmark of PCNSL and could play a role in the prognosis of patients such as reactive perivascular T-cell infiltrates (RPVI) [32]. Patients with RPVI-positive lesions had significantly better overall survival (OS) especially for patients treated with high-dose methotrexate-based chemotherapy (3-year OS: 59 ± 14 vs. 42 ± 9%) [32].

Table 1.

2021 CNS WHO classification of central nervous system lymphomas

CNS lymphomas
	Incidence	Specific morphological and IHC features
Primary diffuse large B-cell lymphoma of the CNS	0.47 cases per 100 000	Angiocentric tumoral cells. Positivity for B-cell markers CD19, CD20, and CD79a
Immunodeficiency-associated CNS lymphomas	Unknown (rare)	Positivity of tumoral cells for B-cell markers CD19, CD20, and CD79a and EBV-associated proteins EBNA1–6, LMP1, and EBV-encoded small RNAs 1 and 2 (EBER1 and EBER2)
Lymphomatoid granulomatosis	Unknown (rare)	Polymorphous lymphoid infiltrate and EBV+, CD20+, CD30+/−, CD15− large neoplastic B cell
Intravascular large B-cell lymphoma	0.5 cases per 1 000 000	Large atypical B cells CD20+ are present, confined to the lumina of cerebral blood vessels
Miscellaneous rare lymphomas in the CNS
MALT lymphoma of the dura	Unknown (rare)	Dural-based neoplastic marginal zone B- cells (CD20+, CD79a+, CD21+, CD23+, CD35+)
Other low-grade B-cell lymphomas of the CNS	Unknown (rare)	Most cases: extranodal marginal zone lymphoma
Anaplastic large cell lymphoma (ALK+/ALK−)	Unknown (rare)	Large neoplastic lymphoid cells CD30+, negative for B-cell markers, ALK+ or ALK-
T-cell and NK/T-cell lymphomas	Unknown (rare)	T-cell: Positivity for T-cell markers (CD2, CD3, CD4, CD8,CD5, CD7). NK-cell: Positivity for CD3, CD2, CD5, CD7, CD56

CNS, central nervous system; IHC, immunohistochemistry [3].

FIGURE 3.

Histopathological features. (a) Patternless tumor with large areas of geographical necrosis (HE staining, x200). (b) Front of the invasion with single cells infiltrating the brain tissue (HE staining, x200). (c) Angiocentric large tumoral cells with round to oval vesicular nuclei, fine chromatin and prominent nucleoli (HE staining, x400). (d) Brisk mitotic activity and apoptotic bodies are frequent (HE staining, x400). HE, hematoxylin-eosin.

As illustrated in Fig. 4, the tumor cells showed B-cells markers PAX5, CD19, CD20, CD22, and CD79a immunopositivity with surface immunostaining of IgM, IgD (not IgG). Most cells are positive for BCL-6 (60–80%), IRF4 (MUM1) (90%), and BCL-2 (82%) and negative for plasma cells markers (CD68 and CD138) corresponding to the late germinal center exit B-cell phenotype [3].

FIGURE 4.

Immunohistochemical features. (a, b) Diffuse immunopositivity in tumoral B-cells for CD45 (a) and CD20 (b) (x400). (c) CD3 immunopositivity in reactive cytotoxic T-cell lymphocytes (x400). (d) CD10 expression is uncommon in CNS-DLBCLs (x400). (e) GFAP immunopositivity highlighted reactive astrocytes commonly found in the neighboring brain tissue (x400). (f) Ki67 proliferation index, higher than 90% in this case (x400).

Regarding prognosis markers, Liu et al.[18] reported that the mean KI-67 index (usually higher than 70% and even 90%) was 88% and is associated with significant poor prognosis (<90% vs. ≥90%) for patients with PCNSL [3,29]. Interestingly, some studies have performed PD-L1 immunohistochemistry, which harbored membranous expression in 30% of cases, suggesting a potential interest of immunotherapy in PCNSL [33,34].

On the basis of morphology, PCNSLs and systemic DLBCLs are indistinguishable [17]. However, CD10 expression is uncommon in CNS-DLBCLs (less than 10%) and indicates possible systemic DLBCL secondly located in the CNS compartment [3]. Therefore, PCNSL needs to be clinically distinguished from the CNS location of systemic DLBCL [17].

MOLECULAR PROFILES AND GENE EXPRESSION SIGNATURES OF PRIMARY CENTRAL NERVOUS SYSTEM LYMPHOMA

As developed in Chapter 1, various mechanisms of genetic alterations are involved in the pathogenesis of PCNSL, including point mutations (non-SHM, aSHM), SHM/aSHM (hotspot target PIM1 and BTG2 genes), chromosomal translocations, and chromosomal copy gains or losses [5,12,17].

Chromosomal abnormalities included copy number gains of 2q37, 3q12.3, 7q, 11q, 9p24.1 (affecting CD274/PD-L1(programmed death-ligand 1) and PDCD1LG2/PD-L2(programmed death-ligand 2)), 18q21 and chromosome 12, copy number deletions of 6p21–22 [human leukocyte antigen (HLA) locus], 6q21, 8q12, 9p21.3 (CDKN2A biallelic loss), 19p13 (CDKN2D), and translocations of IgH-BCL6[2^▪▪,8^▪▪,35^▪,36,37].

BCL6 is a transcriptional repressor expressed by the germinal center B-cells, required for germinal center formation and is involved in the regulation of genes associated with cell-cycle control, inflammation, and lymphocyte differentiation [36]. BCL6 gene translocations, with many partners loci including immunoglobulin, are frequently found in systemic DLBCL (more than 40%) defining high-grade B-cell lymphoma, with MYC and BCL2 and/or BCL6 rearrangements in the 2016 WHO classification of lymphoid neoplasms [8^▪▪,38,39]. The IGH-BCL6 translocation, mostly found in ABC-DLBCLs, leads to a substitution of the promoter of BCL6 by that of the partner and deregulation expression of intact BCL6 protein [36]. Few reports have described the IGH-BCL6 translocation PCNSLs. In 2006, BCL6 translocation have been described in 36% of PCNSLs with unknown impact on their prognosis [37]. Another series of 75 PCNLS reported a frequency of 45 and 17% of BCL6 and IGH rearrangements, respectively [40]. Interestingly, they found a significantly worse OS associated with BCL6 translocation [40]. Recently, Radke et al.[8^▪▪] reported that all PCNSL cases harbored IG gene rearrangements. Interestingly, MYC and BCL2 translocations seem to be rare in PCNSLs [3].

The mutated genes include members of the B-cell receptor signaling (i.e., MYD88, CD79B, CARD11), cell cycle/apoptosis regulator (i.e., TP53, CCND3, BTG2, PIM1, CDKN2A, ATM), chromatin (i.e., KMT2D), and transcriptional (i.e., C-MYC, PRDM1, TBL1XR1) regulator pathways [16]. The tumoral cell also harbor high load of rearranged, somatically mutated IG genes, reflecting the late germinal center exit B-cell phenotype [3,17,41]. Many point mutations reported in PCNLs are associated with NFκB activation. NFκB signaling pathways play a key role in lymphoma cell proliferation and survival by deregulation of TLR (mutations of CARD11 and MYD88), BCR (CD79B), and JAK-STAT [2^▪▪,8^▪▪,16,18,42–45].

PCNSLs typically harbored MYD88 (L265P) (which encodes for a B-cell signaling adaptor protein) activating mutations, CD79B (Y196X), PIM1, CARD11[8^▪▪]. They also harbored CNV including CDKN2A biallelic loss [33]. As previously described, MYD88 and CDKN2A loss represent early clonal event in the tumoral pathogenesis of PCNSL [33]. In a recent study of 64 PCNSL cases, 210 somatic mutations across 14 genes analyzed (average of 3.3 mutations per case ranging between 0 and 10) confirmed the most frequent genes mutated: MYD88 (66%), PIM1 (41%), CD79B (17%) and CARD11 (8%). Other mutations were also observed including KMT2D (31%), PRDM1 (30%), C-MYC (19%), IRF4 (19%), CD79B (17%), TP53 (11%), CCND3 (9%), PAX5 (3%), CSMD2 (3%), and CSMD3 (3%) [16]. By contrast, the most frequent molecular profile for systemic DLBCLs infiltrating the CNS were PRDM1 (50%), followed by MYD88 (42%) and PIM1 (25%) mutations [16]. In 2022, whole exome sequencing of 51 PCNSLs reported 50 driver genes such as MYD88 (67%), CD79B (63%), OSBPL10 (83%), HLA-A/B/C (40%/63%/53%), PRDM1 (40%), TOX (50%), TBL1XR1 (40%), CD58 (37%), PIM1 (70%), ETV6 (50%), BCL11A (30%), CDKN2A (83%), GRHPR (60%), FOXC1 (20%), and DAZAP1 (20%). They also described MYC mutations (17%) and no MYC translocations were evidenced [8^▪▪].

Focus to the most important gene mutations in PCNSLs: PIM1 gene mutations are described in 25--100% of cases [8^▪▪,16,35^▪,46–48]. Zhou et al.[46] have divided CNS-DLBCL into two groups CDP (PIM1 and/or CD79B mutations) and non-CDP (without PIM1 and CD79B mutations) with better outcome (long-term survival) observed for CDP group after high-dose methotrexate-based polychemotherapy.

MYD88 and CD79B (CD79A and CD79B are the two B-cell receptor subunits) are, respectively, mutated in 38–94% and 17–51.2% of PCNSL cases, respectively [8^▪▪,16,33,35^▪,45,46,49]. Both genes are involved in the NF-kB signaling, which is associated with an increase in cell division [33]. Interestingly, MYD88 L265P is more frequently mutated in PCNSLs (59.8%) than in DLBCLs from other sites (16.5%) and is not described in EBV-positive cases [8^▪▪]. Although Nayyar et al.[33] reported that CD79B mutations were associated with better OS (median OS of 105 months in CD79B nonmutated tumors vs. OS not reached in CD79B-mutated tumors) and PFS (median PFS of 35 months in CD79B nonmutated tumors vs. median PFS of 112 months in CD79B-mutated tumors), there was no prognosis impact for MYD88 mutations. The prognostic implication of MYD88 mutations is quite debated, as studies reported conflicting conclusions in the literature [33,43,50]. In 2021, Curran et al.[43] reported in a consecutive series of 57 patients with PCNSL that MYD88 mutation is associated with better survival in multivariate model analysis, while in 2018, Takano et al.[50] reported worse OS for patients among the age of 65 years (11.5 vs. 56.2 months P < 0.04).

CDKN2A is a tumor suppressor gene included in angiogenesis, cell death, invasiveness, and growth suppression [8^▪▪,51]. CDKN2A gene molecular alterations are widely described in CNS tumors and particularly in infiltrating gliomas such as IDH-wt glioblastoma with important prognostic implication of CDKN2A homozygous deletion in IDH-mutant astrocytoma [52,53]. Regarding CDKN2A gene deletion in PCNSL, WES analysis of 36 PCNSL reported the deletion in 44% of cases, while another series reported recurrent and often biallelic CDKN2A deletion in 60% (out of 19 PCNSL cases) to 83% (out of 51 PCNSL cases) [8^▪▪,33,35^▪,36,54]. The frequency of this molecular alteration suggests that CDK4/6 inhibitors could be a therapeutic option for PCNSL patients [33].

CARD11 gene mutations are found in 10–16% of PCNSLs [16,33,55]. CARD11 is a downstream member of the BCR pathway and mutations were described to be associated with Ibrutinib, a Bruton tyrosine kinase (BTK) inhibitor able to cross the blood–brain barrier, resistance [33,56].

Other genes mutations included BTG2 (12.5–92.7%), PRDM1 (19%–40%), and TBL1XR1 (14–40%) [8^▪▪,35^▪,47,57]; this latter mostly co-occurred with MYD88 mutations (both activated NF-kB signaling pathway) [8^▪▪,49].

In 2000, Alizadeh et al.[58] underlined the systemic DLBCL heterogeneity and introduced the molecular sub-classification based on gene expression, including germinal center B-cell type, ABC-type, and some “unclassified” cases. The differential gene expression signatures interested proliferation, T cells, and lymph-node biology such as the genes that distinguished germinal center B cells from other stages in B-cell ontogeny. These subgroups are associated with significant prognostic implication. Patients with germinal center B-cell type DLBCL had a significantly better OS than those with ABC-type [16,48,58]. As previously described, most of CNS-DLBCL tumor cells harbor non-GCB, ABC-like immunophenotype using Hans classification (95% in recent series) associated with late germinal center (exit) B-cells (with blocked terminal B-cell differentiation) genotype by gene expression profiling [8^▪▪]. This could explain that CNS-DLBCLs have a worse outcome than systemic DLBCLs.

The mutational pattern described in the literature suggested that PCNSL are genetically related to “MCD” (MCD enriched genes: PIM1, MYD88 and CD79B mutations), “C5,” or “MYD88-like” DLBCL subtypes [2^▪▪,8^▪▪,35^▪,59]. In 2022, whole genome and transcriptome sequencing were performed to study noncoding regulatory changes and structural variants of PCNSL samples from 42 immunocompetent patients. They succeed to classify the cases with 67% MCD, 3% BN2 and ST2, while 23% were nonclassified cases and one case was assigned to two subtypes (MCD/BN2) in the same way [8^▪▪].

Recent studies of global gene expression described specific clustering and RNA expression signatures of PCNSL as compared to secondary CNS lymphoma and meningeal CNS lymphoma, which grouped together with ABC and GCB-DLBCLs [8^▪▪]. SHM, which we previously described to have a key role in the pathogenesis of PCNSL, has been shown to be more extended in PCNSL than systemic DLBCLs [8^▪▪,42]. In contrast to systemic DLBCL, PCNSL harbored significantly more single nucleotide variations even in intronic and intergenic regions [8^▪▪]. Even if there is a molecular overlap with systemic DLBCLs, GC-like and ABC-like PCNSLs do not harbor different molecular profile. However, studies reported that CD79B, CARD11, CSMD2, and CSMD3 have been exclusively described in ABC cases [16]. Radke et al.[8^▪▪] also reported higher frequency of the BTG2 proto-oncogene mutations in PCNSLs as compared to ABC-like systemic DLBCLs [8^▪▪,35^▪,46,47].

These findings reinforce the hypothesis that PCNSL is a distinct clinical entity with a specific molecular signature independently of the cell of origin classification [8^▪▪,16,47].

EPIGENETIC CHANGES AND DNA METHYLATION SIGNATURES OF PRIMARY CENTRAL NERVOUS SYSTEM LYMPHOMA

DNA methylation is a relatively stable component of the epigenome, not only specific to the cell but also to the tissue of origin and therefore can be used to establish lineage classification. A study of the cancer methylome, which also reflects additional somatic alterations, succeeds to identify new-specific methylation clusters among CNS tumors and is now widely used in the neuropathological diagnosis [60,61]. Interestingly, as previously described in 2009, Vogt et al.[62] did not succeed to distinguish PCNSLs (n = 26) from DLBCLs, which invade the CNS (n = 78) by DNA methylome but identified 2847 CpGs differentially methylated between PCNSLs and DLBCLs, which invade the CNS [62,63]. As expected, PCNSLs harbor specific cluster from other CNS tumors by DNA methylation analysis [60,64].

Hernandes et al.[2^▪▪] proposed a molecular subclassification of PCNSL, with specific clinical implication, based on a multiomic approach on 64 immunocompetent EBV-PCNSL cases: methylation, mutations, CNVs, fusions, gene expression, TCR/BCR clonotypes, tumor microenvironment, tumor localization, and clinical data. Four PCNSL subtypes (CS1 to CS4), associated with significant different OS in multivariate models, were described: CS4 was associated with the better median OS (66.8 months) followed by CS1 (26.2 months), CS2 (18 months), and CS3 (13.8 months). RNA gene expression profiling seemed to be sufficient to identify these sub-groups. Regarding mutations, CS4 was enriched in MPEG1, PIM2, and SOCS1 mutations (a negative regulator of the JAK pathway and an activator of the STAT3 pathway). Although CS2 cluster did not exhibit enriched gene mutations, CS3 cluster showed more HIST1H1E mutations and CS1 was enriched in mutations involved in the NF-kB pathway, B-cell differentiation and proliferation, and B-cell lymphomagenesis [2^▪▪]. In their study, MCD-PCNSL was enriched in CS1 PCNSL subtypes. The correlation between DNA methylation and transcriptome analysis highlighted that 12.4% out of total of 27 111 gene were correlated with their promoter methylation including TERT, CD79A, and MGMT, as previously described by Chu et al.[65]. DNA hypermethylation leading to epigenetic silencing is one of the various mechanisms implicated in the pathogenesis of PCNLs [53]. DNA hypermethylation also targets reduced folate carrier (RFC) genes in 30% of cases [66], CDKN2A[37] or DAPK[16], which could be associated with therapeutic implications [53].

CONCLUSION

Multiomics approaches have succeeded to substantially improve the understanding of the pathogenesis, molecular, and epigenetic events in PCNSL. However, challenges remain in refining the obvious heterogeneity of CNS-DLBCL. Recently defined clinical, molecular, and epigenetic subtypes of PCNSL could help decipher new targeted therapeutic approaches.

Acknowledgements

The authors would like to thank Dr Marie-Lucie Racu for her logistical help.

Financial support and sponsorship

None.

Conflicts of interest

There are no conflicts of interest.

REFERENCES AND RECOMMENDED READING

Papers of particular interest, published within the annual period of review, have been highlighted as:

• ▪ of special interest

• ▪▪ of outstanding interest