Distinct microRNA expression signatures of primary and secondary central nervous system lymphomas

Central nervous system (CNS) lymphoma is a rare and aggressive non-Hodgkin lymphoma that might arise in the CNS (primary CNS lymphoma, PCNSL) or disseminates from a systemic lymphoma to the CNS (secondary CNS lymphoma, SCNSL). Dysregulated expression of microRNAs (miRNAs) is associated with various pathological processes and miRNA expression patterns may have diagnostic, prognostic and therapeutic implications. However, miRNA expression is understudied in CNS lymphomas. Here, we performed expression analysis of 798 miRNAs in 73 CNS lymphoma samples using the NanoString platform, followed by a detailed statistical analysis to identify potential novel biomarkers characterizing subgroups and to examine differences based on their primary and secondary nature, molecular subtype, mutational patterns and survival. We describe the general expression patterns of miRNAs across CNS lymphoma samples and identified 31 differentially expressed miRNAs between primary and secondary groups. Additionally, we identified 7 more miRNAs associated with a molecular subtype and 25 associated with mutation status. Using unsupervised clustering methods, we defined a small but distinct primary CNS lymphoma subgroup, with characteristically different expression patterns compared to the rest of the cases. Finally, we identified differentially regulated pathways in the above comparisons and assessed the utility of miRNA expression patterns in predicting survival. Our study identifies a novel CNS lymphoma subgroup defined by distinct miRNAs, proves the importance of specific miRNAs and pathways in their pathogenesis, and provides the basis for future research.

1 0 1 efficacy of novel targeted therapies between these subgroups [7,8]. A more precise, gene 1 0 2 expression-based molecular subtype assignment can be achieved from formalin-fixed paraffin  The discovery of microRNAs (miRNAs, miRs) has opened a new field for unraveling and 1 0 8 therapeutically targeting diseases. These small non-coding RNAs regulate diverse biological 1 0 9 processes through post-transcriptional gene expression modulation. Based on their seed 1 1 0 sequence, miRNAs bind to multiple target mRNAs, thereby promoting their degradation or 1 1 1 6 myriad of pathological processes including hematological malignancies [13,14], and distinct 1 1 3 miRNA expression patterns may also have diagnostic, prognostic and therapeutic implications 1 1 4 [15][16][17][18][19]. As miRNAs remain relatively well preserved in archival FFPE tissue specimens, they 1 1 5 are readily available as a valuable source of information in cancer tissues [20][21][22]. For the 1 1 6 quantification of miRNA transcripts in FFPE samples, the NanoString nCounter technology is 1 1 7 a preferable choice over quantitative reverse transcription polymerase chain reaction (RT-1 1 8 PCR) [23,24], with a high reproducibility similar to other platforms [25,26]. The NanoString 1 1 9 assay also performs well in relative quantification studies [27]. MiRNA expression of PCNSL has been studied using different methods such as RT-PCR [28- peripheral blood [35], serum [34] or plasma [33]. It has been shown and further confirmed 1 2 5 that the combined detection of miR-21, miR-19b and miR-92a in CSF allowed a reliable 1 2 6 diagnosis of PCNSL. Moreover, these miRNAs emerge as promising tools in treatment 1 2 7 monitoring and follow-up [28,29], similarly to U2 small nuclear RNA fragments [30]. In 1 2 8 addition, plasma miR-21 may serve as a diagnostic [33], and serum miR-21 both as a 1 2 9 diagnostic and prognostic marker for PCNSL [34]. Other miRNAs with prognostic value in 1 3 0 PCNSL include miR-151a-5p and miR-151b, together with 10 additional miRs [35]. CSF 1 3 1 levels of miR 21 may have potential as a predictor of chemotherapeutic effect [33]. Measuring miR-30c in the CSF can differentiate between PCNSL and SCNSL, as elevated In this study, we performed expression profiling of 798 human miRNAs in 73 FFPE brain 1 3 8 biopsy samples of primary and secondary CNS lymphomas using the NanoString platform, followed by a bioinformatics analysis to reveal changing expression signatures. We aimed to 1 4 0 identify potential novel biomarkers characterizing subgroups among brain lymphomas, as 1 4 1 well as to examine differences based on their primary and secondary nature, molecular 1 4 2 subtype, mutational patterns and survival. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 8, 2021. Sample collection and patient information 1 4 7 FFPE brain biopsy specimens of 64 patients with PCNSL and 9 patients with SCNSL were 1 4 8 1 9 We carried out a similar analysis using the 19 miRNAs defined during the binary clustering, 4 0 2 that are not expressed in our "small cluster". However, in this case, as the miRNAs are not  Additional, more detailed results for the SCNSL vs PCNSL, PRDM1 mutated vs non-mutated 4 0 8 and "small cluster" analysis using the C2 -C7 gene sets are available in Supplementary file 6. Finally, we asked whether the expression profile of specific miRNAs is associated to patient survival for patients with "Expressed" or "Not expressed" status of hsa-miR-4488 was one (Supplementary file 7), we found no significant differences between groups after FDR 4 2 0 correction. hsa-miR-18a-5p had the lowest FDR value (0.12). Of note, the survival analysis is  . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 8, 2021. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) In this study, we performed expression profiling of 798 human miRNAs in a large number of 4 6 3 CNS lymphoma cases. We compared the miRNA expression patterns of FFPE brain biopsy 4 6 4 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) We identified 28 up-and 3 downregulated miRNAs in SCNSL compared with PCNSL ( Table   4 6 8 1). Out of these, 23 up-and all 3 downregulated miRNAs are completely novel and until now, 4 6 9 were not described in the context of CNS lymphomas. Reassuringly, the remaining miRNAs described miRNAs, we found miR-30c-5p, that is significantly increased in CSF samples of in PCNSL and DLBCL cases compared to controls [16,28,29,33,34,[62][63][64][65][66]. In our study, . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) other members of the miR-17-92 cluster (miR-17-5p and miR-20a) to be upregulated in 4 9 0 PCNSL compared with nodal DLBCL [32,37]. Moreover, high levels of miR-19b-1 and miR-4 9 1 92a-1 were detected in the CSF of PCNSL patients [28,29].

9 2
Regarding the molecular subtypes, cases in the ABC group showed significantly higher  -222-3p [18, 71, 73] has already been associated with the ABC 4 9 6 subtype in DLBCL. We found that the ABC molecular subtype also correlated with higher 4 9 7 miR-522-3p, miR-454-3p and miR-455-5p expression compared with the UC subgroup. This is the first study demonstrating differentially expressed miRNAs in association with the 4 9 9 mutational status of the PRDM1, MYC and CARD11 genes in CNS lymphomas. According to 5 0 0 the literature, the only miRNA that has already been connected to any of these genes is miR-5 0 1 30a-5p, which directly targets PRDM1 and modulates the WNT/beta-catenin pathway [77]. However, this association is not connected to the mutational status of PRDM1 itself. It is widely known, that the different miRNA profiling platforms do not perform consistently 148b-3p, miR-32-5p, miR-411-5p and miR-379-5p by ddPCR and/or RT-PCR methods. Based on our data, pathway enrichment analysis revealed several downregulated pathways circumstantial, as we did not directly measure the differential regulation of the genes 5 1 0 comprising a pathway, but only their regulators, these pathways might be attractive targets for  . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) suggesting a possible therapeutic target in PCNSL, as these lymphomas might be more include protein secretion, the p53 pathway, MYC target genes, the G2M checkpoint, E2F 5 2 1 transcription factor target genes, apoptosis genes, and genes related to androgen response.

2 2
Considering that PCNSL is a rare and aggressive disease and the prognosis is poor [84], the 5 2 3 pathways and molecular mechanisms analyzed in this study might be considered as novel 5 2 4 drug targets.

2 5
Considering the pathway level changes in PRDM1 mutated samples, some drugs might be 5 2 6 more effective in patients without PRDM1 mutations. The TGF-Beta signaling, the PI3K- might be decreased [80,82,85]. rest of the cohort. Additional sample clustering using a k-means based method with normalized expression values defined a sample cluster (SCluster1) consisting of 10 samples.

3 6
Five samples from SCluster1 also overlapped with the small cluster defined in the binary . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 8, 2021. ; https://doi.org/10.1101/2021.02.05.21249862 doi: medRxiv preprint 2 5 samples with markedly distinct miRNA expression patterns. Interestingly, the lack of 5 3 9 expression of 19 miRNAs was found to be associated with the small cluster in the binary 5 4 0 clustering analysis. Moreover, all of these 19 miRNAs were part of a miRNA k-means cluster SCNSL vs PCNSL analysis. The WNT/beta-catenin pathway activated here, was described as  Based on these results, this is a well-defined sample group within the PCNSL cases contributing significantly to the distinct expression patterns between PCNSL and SCNSL.

4 8
Survival analysis of all cases using the binary expression data showed miR-4488 to be 5 4 9 significantly associated with a worse overall survival, however, we did not find any  Our study identifies a novel CNS lymphoma subgroup defined by distinct miRNA expression 5 5 5 patterns, describes putative subtype and cell-of-origin biomarkers, and proves the importance  . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 8, 2021. Permissions to use the archived tissue have been obtained from the Local Ethical Committee Not applicable. The datasets supporting the conclusions of this article are included within the article, its additional files and at the NCBI GEO database with accession number GSE162956.

6 9
Competing interests The authors declare the following competing interest. James Storhoff and Ning Chen were 5 7 1 employees of NanoString Technologies when the original NanoString analyses were carried 5 7 2 out. James Storhoff is currently an employee of Veracyte. Ning Chen is currently an 5 7 3 employee of Adaptive Biotechnologies. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 8, 2021.
: 9 9 9 1 -9 9 9 6 . 5 9 5   . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 8, 2021.  r  a  c  t  i  c  e   6  2  9   r  e  c  o  m  m  e  n  d  a  t  i  o  n  s  t  o  s  t  a  n  d  a  r  d  i  z  e  p  r  e  -a  n  a  l  y  t  i  c  a  l  v  a  r  i  a  b  l  e  s  i  n  t  h  e  d  e  t  e  c  t  i  o  n  o  f  c  i  r  c  u  l  a  t i n g a n d . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 8, 2021. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 8, 2021.  . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 8, 2021.  . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 8, 2021. ; https://doi.org/10.1101/2021.02.05.21249862 doi: medRxiv preprint 3 2 and secondary central nervous system lymphoma samples categorized by molecular subtypes. samples. Both axes are log 10 based, and the hexagon color scale shows the number of 8 2 8 miRNAs falling into a particular median expression range. As can be seen on the plot (bright while the y-axis shows the -log 10 transformed FDR corrected p-value. b) Volcano-plot of 8 3 6 differential expression results comparing mutated and non-mutated samples for a specific a) x-axis shows the log 2 fold change of a specific miRNA, while the y-axis shows the -log 10 8 3 9