Introduction
Primary central nervous system lymphoma (PCNSL) is a rare form of non-Hodgkin lymphoma (NHL) comprising 2.2 % of all central nervous system (CNS) tumours [
1]. It encompasses lymphoma exclusively involving the brain, spinal cord, eyes, meninges, and cranial nerves, with 90–95 % classified histologically as diffuse large B-cell lymphoma (DLBCL). The majority of PCNSL are sporadic and the incidence increases with age. A minority are attributable to immunosuppressed states, including HIV infection or iatrogenic immunosuppression following organ transplantation. In the era of effective combined antiretroviral therapy (cART), the frequency of HIV-associated PCNSL has diminished [
2]. The involvement of critical sites within the CNS presents both diagnostic and therapeutic challenges, with outcomes consistently inferior to systemic DLBCL. Neurocognitive dysfunction and impaired performance status are frequent at clinical presentation, whilst histological confirmation is inherently risky and often yields small tissue biopsies. Moreover, choice of cytotoxic therapy is limited by the inability of many drugs employed for systemic NHL treatment to penetrate the blood–brain barrier (BBB) efficiently. Since the initial description of PCNSL in 1975 [
3], treatment algorithms have evolved from whole-brain radiotherapy (WBRT) as a single-modality treatment towards a multi-agent, high-dose methotrexate (MTX)-based, chemotherapy approach where WBRT is reserved for consolidation or for relapsed disease. Given the rarity of PCNSL, together with challenges conducting clinical trials in this patient group, data from randomised studies are scarce and the level of evidence to guide therapeutic decisions is often low. This review covers recent advances in our understanding of biological and clinical aspects of PCNSL, chiefly primary cerebral DLBCL, and potential implications for clinical practice.
Diagnosis
The diagnosis of CNS lymphoma can be a particular challenge because of lesional response to corticosteroids and MRI features that are shared with other pathologies. The majority of PCNSL are diagnosed via stereotactic biopsy or, less commonly, by flow cytometric analysis of cerebrospinal fluid (CSF) lymphocytes. The conventional approach has been to avoid surgical resection given the risk of neurological sequelae and lack of therapeutic benefit [
4]. However, a recent unplanned secondary analysis of the G-PCNSL-SG-1 trial has challenged this view, describing an apparently superior progression-free survival (PFS) for those undergoing complete or subtotal resection [
5]. However, this study had a number of limitations, and independent verification in a well-designed and controlled study would be required to change practice.
Rubenstein et al. recently evaluated the utility of CXCL13 (a mediator of B-cell migration) and IL-10 as diagnostic biomarkers with the ability to discriminate CNS lymphoma from other CNS [
6•]. The mean concentration of CXCL13 protein in CSF from newly diagnosed PCNSL and SCNSL was >50-fold higher than in CSF from patients without CNS lymphoma (p < 1 × 10
−7). The concentration of IL-10 in CSF from PCNSL and SCNSL patients was also markedly elevated compared with non-lymphoma comparators (p < 2.3 × 10
−5). Notably, for patients with PCNSL, both CXCL13 and IL10 levels below the median were associated with significantly longer PFS, although statistical independence from pre-existing clinical risk scores was not shown. The positive predictive value of CXCL13 and IL-10 elevation in CSF was 95 % in the identification of newly-diagnosed HIV-negative PCNSL, with an 88 % negative predictive value [
6•]. These interesting findings potentially offer the opportunity for CNS lymphoma diagnosis without brain biopsy, particularly where tissue biopsy is deemed high-risk or of low diagnostic yield. The precision and reproducibility of the diagnostic cut-offs, however, will need to be prospectively evaluated.
Magnetic resonance imaging (MRI) is the principal modality for the detection and monitoring of CNS lesions and recent publications have focussed on the diagnostic and prognostic role of advanced MRI techniques. Cellular density is higher and vascularity is reduced in PCNSL compared to other CNS malignancies, which is reflected in lower apparent diffusion coefficient (ADC) and relative cerebral blood flow (rCBV) values on diffusion-weighted and perfusion MRI, respectively. In support of a prior report suggesting that ADC values are predictive of outcomes in PCNSL [
7], a recent study of 23 patients showed that those with baseline ADC
min <384 × 10
−6 mm/s had inferior PFS and overall survival (OS) [
8]. A study by the same group reported that low baseline rCBV predicted inferior OS in a small cohort of 25 patients. Patients with both low ADC
min and low rCBV had the worst outcomes with 0 % OS at five years compared to 100 % for those with high values for both [
9]. A multi-centre phase II study failed to identify ADC as an independent prognostic factor, but numbers were small (n = 28) and the two-year PFS was lower in those with an ADC below the median (57 % vs. 86 % p = 0.27) [
10•].
Whole-body
18FDG PET-CT has an increased sensitivity for the detection of systemic DLBCL over conventional CT staging [
11], and has an important role in the exclusion of systemic lymphoma at presentation. PCNSL lesions characteristically exhibit homogeneous, high-avidity
18FDG uptake [
12], and one small study has suggested that this may assist in differentiating PCNSL from other intracranial malignancies where MRI findings are equivocal [
13]. Pre-imaging corticosteroid therapy is a potential confounding factor, however, and the additional diagnostic value of
18FDG PET over modern MRI brain imaging remains poorly defined [
12]. The prognostic impact of pre-treatment
18FDG PET was evaluated in a retrospective study of 42 patients by Kasenda et al., demonstrating inferior OS on multivariate analysis (p = 0.018) for patients with increased
18FDG activity relative to cerebellar uptake at diagnosis [
14], consistent with data from an earlier small (n = 17) study [
15].
Molecular
Improved characterisation of PCNSL genotype and phenotype, albeit from small studies with restricted availability of diagnostic material, has the potential to provide prognostic information and identify key molecular pathways that may serve as potential targets for novel therapeutics [
16]. An activated B-cell like phenotype is typical (95 % MUM-1+, 50–80 % BCL6+, 10 % CD10+), but evidence of ongoing somatic hypermutation and the preservation of an open reading frame suggests ongoing germinal centre exposure. Therefore, PCNSL does not neatly conform to either of the principal molecular profiles identified in systemic DLBCL, namely germinal centre and activated B-cell subtypes, and appears to exhibit unique transcriptional features by gene expression profiling [
17]. In contrast to systemic DLBCL, high expression of BCL-2, BCL-6, and MYC by immunohistochemistry is frequent (70 % of cases studied [
18]) and it has been speculated that this may contribute to the adverse prognosis of PCNSL. Recently, the only multi-centre trial to prospectively evaluate PCNSL biomarkers demonstrated that BCL-6 expression, but not MYC, correlated with inferior survival [
19••]. Whilst some studies support this finding [
20,
21], other retrospective analyses found that BCL-6 overexpression correlated with superior outcomes [
22,
23]. Heterogeneous treatment approaches, sample size, and variable methodologies or cut-offs may explain these discrepant findings.
The most frequent genomic aberration identified in PCNSL tissue is deletion of 6p21 involving the HLA locus (56–79 %) [
24], a lesion found commonly in DLBCL arising in immune-privileged sites [
25], and represents a potential mechanism for immune escape. Deletions within the 6q22-23 region (34–50 %), which contains numerous tumour suppression genes [
20,
24,
26] and homozygous silencing/deletion of the cell cycle regulator CDKN2A (45–64 %) [
24,
27] have apparent adverse prognostic significance. The MyD-88 L265P activating mutation appears to be a common molecular aberration identified (38–50 %), but no impact on clinical outcome has been demonstrated [
21,
24]. The resultant activation of the NFκB pathway, which is also upregulated by less frequent mutations or amplification of MALT1 [
28], CARD11 [
29], PRDM1 [
21,
30], and TBL1XR1 [
24], highlights a key survival pathway and potential therapeutic target. Other dysregulated signalling pathways of potential significance include B-cell receptor signalling, with CD79B mutations in 20 % [
31,
32], and the JAK/STAT pathway [
6•,
33]. Most of the available data, however, has emerged from relatively small studies, thus the frequency and prognostic significance of most individual genomic aberrations requires further validation, preferably in the context of prospective clinical trials.
Compliance with Ethics Guidelines
Open Access This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.