Introduction
Diffuse large B-cell lymphoma (DLBCL) is the most common aggressive variant of B-cell non-Hodgkin lymphomas and is characterized by pronounced clinical and molecular heterogeneity among its subtypes. In particular, systemic DLBCL is distinguished by the potential involvement of non-lymphoid organs and tissues in the tumor process, in addition to the lymphatic system, including lesions of the central nervous system (CNS).1,2 Primary CNS lymphoma (PCNSL) is described as a separate diagnostic form. In the vast majority of cases, it is characterized histologically as DLBCL, is localized in the craniospinal region, and involves the following organs (in descending order of frequency of involvement): the brain, eyes, and membranes of the brain and spinal cord, without signs of systemic disease.3 Accumulated evidence indicates that primary CNS lymphoma is a distinct biological entity and is distinguished from systemic DLBCL, and chronic activation of BCR and NF-κB signaling pathways, in combination with a pronounced ability to evade immune surveillance, plays a crucial role in its development.4
PCNSL accounts for 4–6% of extranodal lymphoma variants, up to 1% of all lymphomas, and about 2% of CNS tumors.4,5 Largely due to improved diagnosis of the disease, the incidence of PCNSL increased threefold between 1973 and 1984. Nevertheless, according to data from the Surveillance, Epidemiology, and End Results database, the incidence rate had stabilized by 2013.6 At the same time, a continuing gradual increase in the incidence of PCNSL among the elderly has been reported.7
Notably, the prognosis of patients with PCNSL is significantly worse than that of patients with secondary CNS involvement in systemic DLBCL. Despite the fact that treatment recommendations are largely based on phase II clinical trials, there is still no consensus regarding the optimal induction regimen in the international medical community. Although induction therapy achieves remission in approximately 50% of patients overall, the risk of relapse remains unacceptably high without subsequent consolidation treatment.4,8
Although the traditional diagnosis of PCNSL, based on histomorphological analysis and immunophenotyping of malignant lymphocytes, has been in use for a considerable time, the molecular-level changes due to variations in the genomes, transcriptomes, epigenomes, proteomes, and metabolomes of tumor cells remain poorly understood due to the rarity of this disease. Further advances in understanding the molecular pathogenesis of primary CNS DLBCL are needed, including the identification of biological features and oncogenic pathways that could serve as potential therapeutic targets.1,9
In recent years, the diagnostic and prognostic role of microRNAs (miRNAs) in non-Hodgkin lymphomas, and particularly in DLBCL, has been actively studied,10,11 as well as the molecular mechanisms of their deregulation in these diseases.12,13
miRNAs are small regulatory RNA molecules with a length of 18–25 ribonucleotides that regulate gene expression at the post-transcriptional level and are functionally involved in a wide range of biological processes. These endogenous molecules are capable of binding complementarily to the 3′-untranslated regions of target messenger RNA (mRNA), which results in either translational suppression or mRNA cleavage, ultimately leading to decreased levels of the encoded protein and often to changes in intracellular signaling pathways.14
miR-142 is considered one of the crucial tumor-suppressive miRNAs in human and animal lymphoid neoplasms. The targets of miR-142-5p and miR-142-3p include a large number of key regulatory genes involved in B-cell lymphopoiesis. According to published studies, this miRNA has a wide range of tumor-suppressive functions by targeting a number of important proto-oncogenes, the dysregulation of which contributes to increased proliferation, inhibition of apoptosis, activation of B-lymphocyte survival signaling pathways, metabolic reprogramming, creation of an immunosuppressive microenvironment and immune evasion, as well as tumor cell dissemination.15
There are a number of important observations regarding the features of miR-142 expression in systemic DLBCL. Firstly, it was found that the level of this miRNA in primary samples of patients with DLBCL is significantly higher than in Burkitt lymphoma samples and in healthy lymphoid tissue.16,17 Secondly, miRNA profiling in DLBCL cell lines U2932 and SUDHL5 showed high levels of both miR-142-3p and miR-142-5p in the total miRNA pool, as well as their significant enrichment in association with the Ago-2 protein as part of the RISC complex. These findings allow both mature strands to be considered biologically highly active in the analyzed cells.18,19 Thirdly, the expression level of miR-142 in DLBCL, determined by next-generation sequencing methods, is characterized by pronounced heterogeneity, with low expression of miR-142 associated with higher survival rates in DLBCL patients receiving therapy according to the R-CHOP protocol.20,21 Finally, it was found that the mutational status of the MIR142 gene (OMIM, https://www.omim.org/entry/615657 ) was observed in systemic DLBCL at a notably high frequency, ranging from 12% to 20%.22–25 No other variant of hematological malignancies or solid tumors has revealed equally frequent mutations in MIR142 or any other miRNA genes.19 Despite this, the pathogenic role of molecular genetic alterations in miR-142 and the MIR142 gene in the development of primary CNS DLBCL remains largely unstudied.
The objective of this study was to investigate the frequency, spectrum, and functional significance of mutations in the MIR142 gene in primary CNS DLBCL.
Discussion
The descriptions of mutations in the MIR142 gene identified in chronic lymphoid leukemia,35 acute myeloid leukemia, and myelodysplastic syndrome,36,37 and in isolated cases of follicular lymphoma and myeloproliferative diseases are contained in modern literature.23,27,38 Taking into account all hematological malignancies, mutations in MIR142 are most often detected in systemic DLBCL.19
The frequency and spectrum of SNVs in the MIR142 gene in tumor tissue of 35 cases of primary CNS DLBCL were characterized in detail for the first time, and an in silico analysis of their functional effect was performed in this study. The frequency of SNVs in the study group was 37.1%, which exceeds the frequency described in the literature for systemic DLBCL, ranging from 12% to 20%.22–25
It is important to note that there were multiple SNVs in the gene sequence in one third of the cases in our study. The study, performed using NGS methods, showed an extremely high frequency of kataegis in the region of the MIR142 gene in systemic DLBCL.39 Kataegis is described as an occurrence of somatic hypermutations in a number of malignant neoplasms characterized by clusters (six or more within a thousand bp) of one type of replacement, about 70% of which are represented by transitions of type C>T.40 All of them are localized on one of the DNA strands and appear as a result of deamination of cytidine during the repair of double-stranded breaks when DNA is in a single-stranded state.41 Since we analyzed a relatively small portion of the nucleotide sequence using the Sanger direct capillary sequencing method, this does not allow us to assess the allelic location of the identified substitutions and confirm the high incidence of kataegis in our study. Supplementary NGS analysis of a subset of samples to assess mutation density within the MIR142 locus (e.g., mutations per 1000 bp) would strengthen this claim. However, kataegis can be inferred from the fact that cases with multiple mutations were characterized by transitions, including repeated substitutions of G>A, which are typical for this mutagenesis mechanism. There is also evidence of sustained activation-induced cytidine deaminase expression in a fraction of PCNSL in published data.4
A total of 18 SNVs scattered across the gene sequence were detected in the tumor tissue of primary CNS DLBCL in our study. These findings are a hallmark of this disease, distinguishing it from other types of hematological malignancies. Thus, according to current published data, all currently known SNVs in this gene in acute myeloid leukemia and myelodysplastic syndrome have been described only in the ‘seed’ region of the mature miR-142-3p chain, which is responsible for binding to targets,36,37,42,33 as well as the only currently described mutation in MIR-142 in Burkitt’s lymphoma.43 The verified SNVs affect the miR-142-3p sequence in follicular lymphoma,23,38 and in the vast majority of cases of chronic lymphocytic leukemia they are located in miR-142-5p.27,35,33 However, the pattern of SNV distribution in primary DLBCL of the CNS is not identical to systemic DLBCL (Fig. 5c). Unlike in systemic DLBCL, in which SNVs are scattered across the gene sequence and are often detected in miR-142-3p, including the seed sequence, in primary DLBCL of the CNS they overwhelmingly affect areas involved in the formation of the pri-miR-142 hairpin base. Only 16.7% changed the sequence of the mature miR-142-5p, and only one (5.6%) affected the seed sequence of miR-142-5p.
Our results are consistent with published data indicating an increased frequency of SNVs in miRNA genes in malignant neoplasms, with a distribution pattern shifting from the ‘seed’ region and mature strand toward the precursor miRNA sequences. This is believed to be associated with a selection process in the tumor favoring functionally significant substitutions.44
It is known that SNVs located in the sequence of the miRNA precursor have a profound effect on miRNA functioning, from the nuclear and cytoplasmic stages of their maturation to thermodynamic stability, the formation of secondary structure, and the selection of chains included in RISC, which ultimately affects their expression and functional activity.45,46
Using the UNAFold web server, we determined the most stable secondary conformation of the “hairpin” miRNA in thermodynamic equilibrium and its MFE, which depend on factors such as the length of the molecule, the composition of nucleotides, and the sequence of their location in the chain. MFE can be calculated by adding the free energies of complementary interacting base pairs, bulges, and loops. It is noteworthy that the most structured and stable in a series of miRNAs of the same size is considered to be the one with the highest negative value of MFE.47
The analysis showed that all SNVs identified in the study group resulted in changes in MFE in the range from 0.8 to 8.4 kcal/mol in absolute terms. In most cases (76.9%), the SNVs increased the MFE, making the predicted hairpin structure less negative and, consequently, less thermodynamically stable, while the rest (23.1%), according to the prediction of the UNAFold web server, stabilized the secondary structure of the molecule. It is worth noting that Gong et al.48 summarized the existing studies in this area and found that changes in MFE by more than 2.0 kcal/mol significantly affect the production of mature miRNA, but even changes below this threshold can alter their biogenesis.49
More recently, the thermodynamic properties of the hairpin have been shown to determine the functionality of the molecule, affecting the rate and selectivity of inclusion of a particular mature chain into RISC, as well as its dissociation from the target after silencing.50 In addition, mutations outside the seed sequence can alter the efficiency of miRNA inclusion in the RISC complex, as well as potentially disrupt interactions with RNA sponges that adsorb miRNAs and thus suppress their activity.50–54
Thus, most of the SNVs identified in this study are located outside the seed sequences and, at first glance, do not appear to have a direct impact on the choice of miRNA target gene. However, beyond thermal stability, great attention should be drawn to the secondary conformation of the miRNA hairpin. The obtained data may indicate that the pri-miR-142 hairpin has a fragile structure. As can be seen from Supplementary Figure 1, in most cases the detected SNVs, according to in silico prediction, cause a significant change in the shape and length of the pri-miRNA hairpin stem. This is significant for the first stage of miRNA biogenesis, namely, the processing of pri-miRNA by the Drosha–DGCR8 complex.
Currently, there is no complete understanding of how the microprocessor recognizes pri-miRNAs and orients itself asymmetrically on the stem-loop structure to select the correct cleavage site. The following model has been proposed: interaction of DGCR8 at the apical loop provides effective cleavage by orienting Drosha toward the basal stem. Thus, both the basal and apical junctions cooperatively coordinate cleavage position and processing efficiency. Structural features are of prime importance in defining how efficiently a pri-miRNA is recognized and processed by the microprocessor.55
Recent research has revealed two important findings. Firstly, the distances from both the lower and upper junctions of the pri-miRNA stem to the single-stranded regions of the molecule in human cells play an important role in determining the cleavage site. If these distances are suboptimal, Drosha–DGCR8 may cleave pri-miRNA at non-canonical sites.34
Secondly, while the 3′ ends of most miRNAs tend to exhibit variability, the 5′ ends of mature miRNAs are typically highly conserved. This is because nucleotides 2 through 8 comprise the seed sequence, which is critical for target recognition.56 Consequently, even a single-nucleotide shift in the pri-miRNA stem length in either direction can disrupt the coordination of Drosha–DGCR8 processing, leading to alternative cleavage and the emergence of different miRNA-5p variants with distinct seed sequences and different target gene pools. It is therefore hypothesized that SNVs located within regions forming the basal part of the pri-miR-142 hairpin may impair Drosha–DGCR8 processing accuracy, resulting in the generation of non-canonical miR-142-5p isoforms and potentially altering its target gene repertoire in CNS-type DLBCL.
The predicted normal pri-miR-142 hairpin structure in our study has key functional elements necessary for processing by the Drosha–DGCR8 complex: a stem of about 33–35 nucleotides long, an apical loop of 10–15 nucleotides with a GUG motif, single-stranded segments at the base of the stem, a CNNC motif downstream of the basal junction, a mismatched GHG motif on the 3′ strand 7–9 nucleotides from the basal junction (in which H is any nucleotide but G), a base-paired stable platform in the basal stem, and a 5′ cleavage site at nucleotide positions 14–15 from the basal junction (Fig. 5a). The primary sequence of all of these motifs has an additive effect on efficient processing. It is known that changes in the length of the stem and the number of unstructured elements (such as mismatches and bubbles on the stem), as well as disruption of main functional motifs, affect the efficiency and accuracy of processing.54
Thus, it is assumed that SNVs located in the sequence regions involved in the formation of the pri-miR-142 hairpin base may impair the cleavage process by the Drosha–DGCR8 complex, resulting in the formation of non-canonical variants of miR-142-5p and changing the pool of target genes of mature miR-142-5p in DLBCL. According to Fukunaga et al.,57 5′-isomers can also be generated by non-canonical binding of the Dicer protein to pre-miRNA.
It was found that the canonical miR-142-5p variant potentially targets 1,133 mRNAs.30 We performed a bioinformatic analysis of the effect of n.29A/G in the MIR142 gene on the switching of miR-142-5p target genes, which showed that this SNV leads to a significant change in the set of regulated genes, namely a halving of the number of potential targets (to 599) and a significant change in the spectrum of interaction. According to in silico analysis, control was retained over only 6.8% of canonical mRNA targets, while 93.2% of them were lost. Notably, 86.8% of the predicted interactions of the analyzed mutant variant of the miR-142-5p seed sequence were novel. These changes were directly reflected in the results of GO term enrichment analysis for the obtained gene lists. The GO enrichment results indicate loss of control over tumor-related pathways: regulation of gene expression, cell death, cell differentiation, cell migration, chromatin organization, chromatin remodeling, intracellular signal transduction, maintenance of cell number, regulation of canonical NF-κB signal transduction, regulation of cell adhesion, regulation of cell cycle, regulation of cellular response to growth factor stimulus, vesicle-mediated transport, and many others.
Thus, targeting of common anti-apoptotic proteins (MCL1 and MDM4) and cell cycle promoters (PKN2 and CCNH) is violated in tumors. In particular, MCL1 and MDM4 are deregulated in a significant fraction of ABC DLBCLs (the predominant subtype in PCNSL) and contribute to therapy resistance.58,59
The following is a discussion of some critical deregulated targets as a result of the mutation n.29A/G in the MIR142 gene, which are not only predicted in our work, but also have experimental evidence of interaction with miRNAs and are presented in the TarBase database.60
The origin of tumor cells in PCNS remains unclear. It is most likely that this tumor arises from transformed peripheral B lymphocytes that have escaped immune surveillance and hidden in the CNS. To do this, they overcome the blood-brain barrier, the densest blood-tissue barrier, and then survive and proliferate in an extremely poor growth-stimulus environment of the CNS. The currently accumulated data indicate that the following mechanisms play a central role in the development of a tumor: adhesion and extravasation, cell signaling, autocrine stimulation, induction and maintenance of inflammation, activation of neoangiogenesis, and escape of immune surveillance.5
The following immune escape molecules are noteworthy. Recently, several studies have reported increased CD47 expression on lymphoma cells of different types. Binding of CD47 on tumor lymphocytes to SIRPα suppresses the phagocytic function of macrophages and enables avoidance of control by innate immunity.61 Treatments that block the interaction of CD47 and SIRPα significantly suppress tumor growth and metastasis through diverse mechanisms, such as phagocytosis, antibody-dependent cellular cytotoxicity, and apoptosis.62,63 CD47 expression is frequent in ABC and a subset of GCB DLBCL with mutations in TP53 and/or CCND1, and is related to tumor-associated macrophage infiltration and disease progression. For these reasons, CD47 is a potential target for the treatment of DLBCL, especially for relapsed and refractory cases.64 The results of a recent study confirmed the critical role of macrophages and CD47 in controlling PCNSL cell growth.65,66
In turn, CD109 orchestrates reprogramming of tumor-associated macrophages to dampen immune response. Moreover, expression of CD109 is a mechanism of down-regulation of TGF-β receptor on lymphoma cells. TGF-β suppresses lymphoma growth by apoptosis and inhibiting proliferation.67,68
Another mechanism of evading the immune response by tumor cells is the loss or decrease in the expression of molecules of the major histocompatibility complex class II (MHC-II) as a result of overexpression of NFX1.69 NFX1 is a nuclear transcription factor, which downregulates MHC II antigen expression.70 MHC-II presents antigens to CD4+ T helper cells, and their suppression reduces antitumor immunity, correlating with a worse prognosis and low effectiveness of immunotherapy in many cancers.71 Unfortunately, interaction of miR-142-5p with NFX1 is currently not recorded in the TarBase and requires experimental confirmation.
PCNSL refers to lymphomas of immune-privileged sites. The development of this tumor requires that neoplastic cells overcome the blood-brain barrier. This may be facilitated by the loss of control over the following targets: ITGAV, ITGAM, ITGA9, and ALCAM. These proteins are key cell adhesion molecules involved in immune response, tumor progression, and cell signaling.
ALCAM (CD166) plays a key role in the pathogenesis of PCNSL, namely in the interaction with the vascular endothelium and migration of B-lymphocytes across the blood-brain barrier. Also, ALCAM expression is considered one of the key factors that ensures the ability of DLBCL cells to survive and proliferate in the immune-privileged environment of the CNS.39,72 ITGAV acts as a key mediator of CNS tumor neoangiogenesis and progression.73,74 It was found that abnormal expression of ITGA9 in a variety of tumors was closely related to the proliferation, metastasis, adhesion of tumor cells, and angiogenesis.75 ITGAV and ITGAM are associated with immune checkpoint blockade.76,77
Other alleged mechanisms responsible for cell proliferation and survival of lymphoma cells with n.29A/G in the MIR142 gene in the CNS include loss of control over the following targets.
It has been proven that during the initiation and progression of DLBCL, a number of intracellular signaling pathways occur, including NF-kB, PI3K-AKT-mTOR, JAK-STAT, and MAPK. Genomic investigations indicate that several signaling pathways participate in the cell proliferation and survival of lymphoma cells in PCNSL, particularly constitutive NF-κB activation and deregulated TLR, BCR, PI3K-AKT-mTOR, and JAK-STAT pathways.78 As a result of the mutation n.29A/G in the MIR142 gene, there is a disruption in the targeting of miR-142-5p on a number of MAPK (MAP3K2, MAPK9, MAPK10) and PI3K-AKT-mTOR (PIK3CA, RICTOR) pathways, contributing to downstream NF-kB activation (TNFSF13B). In addition, there is a disruption in the targeting of miR-142-5p on main signaling molecules of the JAK-STAT signaling pathway (IL22, IL6ST), which allows tumor lymphocytes to survive in a poor growth-stimulus environment of the CNS.
In particular, TNFSF13B is a cytokine that activates the NF-κB transcription pathway and supports the survival of B cells. The interaction of TNFSF13B with its receptors contributes to both canonical and non-canonical signaling of NF-κB. TNFSF13B gene expression is also regulated by NF-κB through a positive feedback mechanism, promoting autocrine regulation of growth and survival. TNFRSF13B was significantly upregulated in ABC-DLBCL, and its high expression was associated with poor prognosis in DLBCL.79,80