IL-21 Stimulates the expression and activation of cell cycle regulators and promotes cell proliferation in EBV-positive diffuse large B cell lymphoma

The clinical features of EBV-positive diffuse large B cell lymphoma (DLBCL) indicate a poorer prognosis than EBV-negative DLBCL. Currently, there is no efficacious drug for EBV-positive DLBCL. The cytokine interleukin-21 (IL-21) has been reported to be pro-apoptotic in DLBCL cell lines and is being explored as a new therapeutic strategy for this type of lymphomas. However, our previous studies showed that IL-21 stimulation of EBV-positive DLBCL cell lines leads to increased proliferation. Here, analysis of a rare clinical sample of EBV-positive DLBCL, in combination with a NOD/SCID mouse xenograft model, confirmed the effect of IL-21 on the proliferation of EBV-positive DLBCL cells. Using RNA-sequencing, we identified the pattern of differentially-expressed genes following IL-21 treatment and verified the expression of key genes at the protein level using western blotting. We found that IL-21 upregulates expression of the host MYC and AP-1 (composed of related Jun and Fos family proteins) and STAT3 phosphorylation, as well as expression of the viral LMP-1 protein. These proteins are known to promote the G1/S phase transition to accelerate cell cycle progression. Furthermore, in NOD/SCID mouse xenograft model experiments, we found that IL-21 treatment increases glucose uptake and angiogenesis in EBV-positive DLBCL tumours. Although more samples are needed to validate these observations, our study reconfirms the adverse effects of IL-21 on EBV-positive DLBCL, which has implications for the drug development of DLBCL.


IL-21 promotes cell viability and survival of primary cells derived from an EBV-positive DLBCL clinical sample.
To confirm our previous finding on the EBV-positive DLBCL cell line Farage that IL-21 induced cell proliferation rather than apoptosis, we collected primary cells (named 'Patient-1') from a clinical sample of EBV-positive DLBCL. After 48 h of IL-21 treatment, we observed a significant apoptosis reduction in these cells (Fig. 1a) compared to the significantly increased apoptosis in EBV-negative DLBCL primary tumours under similar conditions as previously reported 16 . In addition, IL-21 promoted the viability of the primary cells and of the EBV-positive DLBCL cell line Farage, but reduced viability in the EBV-negative DLBCL cell line MC116 (Fig. 1b). The total cell number of EBV-positive DLBCL cells increased significantly after 48 h with/ without IL-21 treatment, indicating cell proliferation in both cases. Using RNA-seq analysis of EBV latency gene transcripts, we found that the EBV-positive DLBCL primary cells expressed the full set of EBV latency genes (indicating a type III latency), which is similar to Farage cells (Fig. 1d). C1orf43 and CHMP2A served as the house-keeping genes 2 . After IL-21 treatment, the expression of Blimp-1 that orchestrates plasma cell differentiation and the viral protein LMP-1 was upregulated in the patient-derived cells as shown by RNA-seq analysis and western blot (Fig. 1c,e), and phosphorylation of STAT3 was also upregulated (Fig. 1e). These results are the same as our previously described in the EBV-positive DLBCL cell line Farage after IL-21 treatment 13,18 . The RNA-seq analysis, combined with our previously reported Western blot results 13,18 suggests that the expression and regulation of these key genes are similar in Farage cells and the primary cells (Fig. 1c-e), which confirmed our finding in cell lines using a primary sample. Although there was no significant difference in the body weights of tumour-bearing mice (Fig. 2a), tumour proliferation in mice treated with IL-21 was increased (Fig. 2b,c). This was further confirmed by comparison of the size and weights of tumours obtained 28 days post the start of measurement (Fig. 2d,e). To measure tumour areas and evaluate tumour glucose uptake, representative mice were evaluated with 18 F-FDG micro-PET/CT three weeks after continuous treatment with IL-21 ( Fig. 2f,g). A larger signal area was observed in the group treated with IL-21 confirming better tumour growth in these mice. In addition, tumour standard uptake in the IL-21 injection group was increased by 17% compared to that in the control group indicating a higher rate of tumour glucose uptake. Next, we analyzed the effect of IL-21 treatment on angiogenesis. Consistent with the in vitro gene expression data for VEGFA and VEGFB showing induction after IL-21 treatment of Farage cells, in vivo we observed promotion of tumour angiogenesis (Fig. 3a,b). After sectioning each of three representative tumours, anti-CD31 (endothelial marker) antibody was used for staining. Sections from the IL-21 group exhibited significantly  -1) were treated with IL-21 (100 ng/mL for 48 h) or left untreated. Samples were stained with anti-Annexin V antibodies to measure cell apoptosis by flow cytometry. The experiment was done in triplicate and one representative sample is shown. Statistical analysis of the flow cytometry data was collected from 3 sample replicates. (b) The viability of EBV-positive and EBV-negative DLBCL cells was assessed by erythrosine B after 48 h of treatment with IL-21. Data in a and b are expressed as the means ± SEMs (n = 3); two-tailed t test, **P < 0.01, *P < 0.05. (c) The indicated DLBCL cell lysates were probed with the indicated antibodies. β-Actin served as the loading control. (d) Z-score normalized log2-(FPKM + 1) of EBV latency genes in gene expression are shown in the heatmap. C1orf43 and CHMP2A were served as the house-keeping gene control. RNA-seq data was collected from 3 sample replicates. (e) RNA-seq analysis of selected genes in Farage cells and in the primary cells. Blue indicates decreased expression and red indicates increased expression compared to the IL-21 untreated control. Log 2 -(FPKM + 1) data from the RNA-seq analysis displayed as the means ± SEMs (n = 3). Statistical analysis in d and e was performed using FDR-based q values; *q < 0.01. Scientific RepoRtS | (2020) 10:12326 | https://doi.org/10.1038/s41598-020-69227-0 www.nature.com/scientificreports/ increased CD31 microvascular staining densities (Fig. 3c,d). These findings may explain the increase in subcutaneous blood vessels around the tumours.

Gene expression induced by IL-21 exposure is similar in EBV-positive DLBCL primary cells and the Farage cell line.
The previous results suggested that IL-21 treatment of EBV-positive DLBCL primary cells and cells from the Farage cell line had similar effects. To study whether this reflected similar gene expression regulation in the two types of cells following IL-21 exposure, we analysed the global gene expression of the two types of cells before and after 48 h of IL-21 exposure. The MA plot shows that IL-21 treatment upregulated the expression of 1512 and 1566 genes in Farage and primary cells, respectively, and downregulated the expression of 1,830 and 1,573 genes ( Fig. 4a,b). To identify the lymphoma associated genes, we performed a real-time literature analysis using PALM-IST (Pathway Assembly from Literature Mining-an Information Search Tool) 24 . We scanned 61,835 article abstracts in the PubMed databases using lymphoma as the main keyword and generated a disease-associated gene (DAG) library (Supplementary Table S1). Based on the DAG library, we performed hierarchical clustering analysis of the fold change of each differential expressed gene (DEG) for each independent sample in the Farage and primary cells groups. The results indicated that IL-21 showed similar regulation of DAGs in Farage cells and the primary cells. Specifically, 80.2% of DAGs in the Farage and primary cells had similar regulatory trends after IL-21 treatment (Fig. 4c).
To identify the most significant signalling pathways regulated by IL-21 in EBV-positive DLBCL, phyper was used to perform KEGG pathway functional enrichment analysis for all DEGs. After IL-21 treatment, DEGs in the Farage group and the primary cells group were highly enriched in the cytokine-cytokine receptor interaction pathway, chemokine signalling pathway and JAK-STAT signalling pathway (Fig. 4d,e). In addition, DEGs enriched in the JAK-STAT signalling pathway showed overall upregulation ( Supplementary Fig. S2).
The master regulator of cell proliferation MYC is upregulated in EBV-positive DLBCL primary and Farage cells upon IL-21 exposure. To identify the main effector of IL-21-induced EBV-positive DLBCL cell proliferation, we applied an RNA-sequencing-based ABC/GCB calling method combining corresponding CRISPR screening data to seek more hints 25 . This method classifies DLBCL tumours based on RNAseq gene expression data into activated B cell-like (ABC) and germinal centre B cell-like (GCB) types. Patients with ABC-type DLBCL have a lower overall survival rate than patients with GCB-type DLBCL. In our study, based on the RNA-seq data characteristics, Farage cells were classified as GCB DLBCL, and the primary cells were classified as ABC DLBCL (Fig. 5a). The calling results remained unchanged in both groups after IL-21 treatment. Besides, six DLBCL driver genes 25 , namely CD97B, MYC, BCL6, IRF4, TGFBR2 and CD22, were dif-

Molecular interaction network analysis identifies a cell cycle module in which MYC plays a central role.
To identify the molecular network involved in IL-21-induced cell proliferation, we conducted STRING network analysis using functional differential expression data from Farage cells and the primary cells.
We obtained a molecular interaction network containing 350 genes and 683 edges ( Supplementary Fig. S3e To obtain the core molecular interaction module for IL-21-induced cell proliferation, we extracted a subnetwork consisting of 25 genes and 73 edges from cluster 3 by disabling the Fluff function in MCODE (Fig. 5c).
The genes that form this core module are significantly enriched in the biological processes of cell cycle regulation and mitotic cell cycle G1/S transition and significantly enriched in the KEGG cell cycle pathway (Table 1, Sup-   Table S7). This subnetwork module describes the regulatory effect of IL-21 on cell cycle progression in EBV-positive DLBCL from a molecular interaction perspective. We found from this subnetwork module that MYC occupies a central position in the module with regard to cell cycle regulation.
IL-21 promotes cyclins in G1/S phase transition in EBV-positive DLBCL but not in EBV-negative DLBCL. To confirm the results from RNA-seq and the network analysis, we extracted the differential  www.nature.com/scientificreports/ expression results from the RNA-seq data and used western blotting to evaluate changes in gene expression at the protein level (Fig. 6a,b). We found that consistent with its effect in Farage cells, IL-21 induced upregulation of c-Myc, c-Jun, c-Fos, cyclin D2 and cyclin E1 protein expression in the primary cells (Fig. 6b). Subtype-specific DEG analysis of cyclin D and E confirmed that only cyclin D2 and cyclin E1 are significantly regulated by IL-21 in both primary cells and Farage ( Supplementary Fig. S4). Since these cyclins are highly correlated with the restriction (R) point transition in the cell cycle, we further examined the phosphorylation status of Rb, which is a negative regulator of cell division whose phosphorylation is associated with cell cycle entry. After IL-21 treatment, we observed increased Rb phosphorylation in both Farage cells and the primary cells (Fig. 6b). The relative protein levels were similar to the differential expression results from RNA-seq analysis (Fig. 6a,b). By contrast, in the EBV-negative DLBCL cell line MC116, IL-21 did not induce upregulation of c-Jun, cyclin D2, and cyclin E1 (Fig. 6b) and no clear increase in the levels of phosphorylated Rb protein was detected in MC116 cells (Fig. 6b). Therefore, we propose a hypothetical model for growth regulation by IL-21 in EBV-positive DLBCL in which IL-21 signalling leads to STAT3 phosphorylation and induces LMP-1 expression to form a positive autoregulatory loop, which amplifies the signals for Myc and AP-1 (composed of homoor heterodimers formed by related Jun and Fos family proteins) transcription (Fig. 7a). The activation of c-Myc and AP-1 results in increased cyclin D2 and cyclin E1 expression (Fig. 7b), promotes R point transition, and ultimately leads to hyperphosphorylation of Rb which accelerates cell cycle progression (Fig. 7c).

Discussion
In the present work, we confirmed previous observations of IL-21-induced proliferation in EBV-positive DLBCL 18 using an EBV-positive primary clinical sample and an in vivo NOD/SCID mouse xenograft model. These results differ from the tumour regression observed in a previous mouse study using MC116, an EBV-negative DLBCL 16 . By RNA-seq analysis using hierarchical clustering, we found that the direction of the regulation of DEGs in the primary cells and in Farage cells after IL-21 treatment was similar. This similarity suggests that Farage cells are an appropriate research tool for the study of type III latency EBV-positive DLBCL. We also conducted a joint analysis of the primary cells and Farage cells with reference to the DLBCL subtype calling model 25 and corresponding CRISPR screening data proposed by the Dave laboratory 25,26 . Here, genes associated with ABC/ GCB classification were used to classify samples into subgroups, and an RNAseqSubtypeScore 25 was applied accordingly. This score indicates the difference of calculated characteristic value between ABC and GCB. Our results showed that the unsupervised hierarchical clustering did not distinguish the IL-21 treated cells and IL-21 untreated cells (Fig. 5a). Interestingly, according to the ABC/GCB classification based CRISPR oncogene targets 25 , our results showed that MYC is the only DLBCL essential gene that was upregulated by IL-21 in both the primary cells and Farage cells (Fig. 5a,b). Based on the role of MYC in cell proliferation, we hypothesised that MYC plays a key role in IL-21-induced proliferation of EBV-positive DLBCL cells. Using bioinformatic methods, we extracted the core subnetworks of co-DEGs in Farage cells and the primary cells. Functional enrichment analysis revealed a subnetwork involved in the regulation of the mitotic cell cycle (Table 1), with MYC at the centre of the core module (Fig. 5c).
Myc acts as the primary transcriptional centre for multiple networks in B cells 27,28 . It is a known target of the STAT3 transcription factor 29-31 and previous publications have identified several target genes of Myc [32][33][34] . Myc directly and indirectly activates the expression of cyclin D2 and cyclin E to promote R point transition 32,[35][36][37][38][39] . The level of D-type cyclins is primarily controlled by extracellular signals 40 , which, together with CDK4/6, are responsible for inducing Rb phosphorylation. Once cells move through the R point, the remaining cyclins, including E-type cyclins, behave in a pre-programmed manner until the end of M phase and no longer respond Given the evidence that LMP-1, an oncoprotein encoded by EBV, mimics the constitutively activated CD40 signal in B lymphocyte responses 42 , it is plausible that IL-21 induces EBV-positive DLBCL cell proliferation by a similar mechanism. It has also been reported that cells with high LMP-1 expression are responsible for the poor drug www.nature.com/scientificreports/ response of EBV-positive DLBCL 43 . Combining these results with our previous observations, we hypothesise that the LMP-1 induced in EBV-positive DLBCL cells may be a key factor for proliferation. LMP-1 leads to activation of the JAK/STAT signalling pathway 44 and is critical for the proliferation and survival of EBV-positive B cells 13,18 . Evidence suggests that activated STAT3 acts as an activator of the LMP-1 L1-TR promoter 45 , which creates a positive autoregulatory loop since STAT3 itself can be activated by LMP-1 44 . Our results showed a significant increase in STAT3 phosphorylation as well as LMP-1 expression after IL-21 treatment in both Farage cells and the patient-derived primary cells. LMP-1 signalling has been shown to have an important effect on the G1/S transition in human B cells 46 with LMP-1 specifically inducing mitogenic B cell activation via Myc upregulation 47 .
Importantly, LMP-1 also induces mitogenic B cell activation via members of the Jun/AP-1 family 47 . After IL-21 treatment, the mRNA and protein levels of c-Jun and c-Fos were significantly increased in the EBVpositive DLBCL cell line and the primary cells. It is known that the transcription of Jun and other AP-1 family members is rapidly and transiently stimulated by many external signals that promote mitogen-induced cell cycle progression 38,48 . Additionally, there is evidence that AP-1 blockade inhibits Rb phosphorylation, decreases the expression of major cyclins in the G1 phase of the cell cycle (D and E cyclins), and decreases the activity of CDK2 and CDK4 37 . Collectively, these indications suggest that the different outcomes of IL-21 signalling in EBV-negative DLBCL and EBV-positive DLBCL cells, both showing an upregulation of MYC, are most likely due to the presence/absence of LMP-1.
Moreover, we demonstrated that the expression of CCL22 was significantly decreased and that the expression of CXCL10 was considerably increased in both Farage cells and the primary cells after IL-21 treatment (Supplementary Fig. S5). Low CCL22 expression and the presence of EBV are associated with decreased survival in patients with DLBCL 49 . High expression of CXCL10 is closely associated with poor survival in patients with DLBCL and secondary central nervous system involvement was more frequent in the high CXCL10 group 50,51 . Therefore, in the EBV-positive DLBCL cell line and the primary cells, upregulation of CXCL10 and downregulation of CCL22 by IL-21 might be detrimental to DLBCL prognosis.
The lack of sufficient cell numbers of clinical samples that can be used for functional studies is one of the major reasons hindering the investigation of EBV-positive DLBCL. This problem can be effectively solved by NGS technology, which produces large amounts of bioinformatic data from a limited amount of material. In our literature analysis in the PubMed database, we used lymphoma as the keyword to extract lists of hotspot genes and build disease-related gene sets to successfully expand our knowledge of disease-related transcriptional regulatory signals. After identifying the core transcription subnetwork through network analysis and enrichment analysis, we validated the prediction results using wet laboratory methods. Our comprehensive research findings show that NGS can satisfactorily match these research protocols and help to predict core molecular mechanisms.
Our study shows that the inappropriate use of IL-21 in patients with EBV-positive DLBCL not only may fail to achieve therapeutic effects but also may lead to accelerated proliferation of DLBCL. In addition, this work suggests that IL-21, as a component of the tumour microenvironment, may also contribute to the proliferation of EBV-positive DLBCL cells. In drug development, key regulators such as the IL-21 receptor and LMP-1 may be valuable as therapeutic targets for EBV-positive DLBCL, and monoclonal antibodies or small molecule antagonists targeting these regulators may be beneficial for the survival of patients. Further functional studies of the key regulatory nodes identified in the current study are needed to confirm our computational data. Although it is currently difficult to integrate large sample resources into systematic research, we hope that this analysis on a clinical sample can further increase researchers' attention to EBV status in DLBCL treatment and supplement biological information in this field. Isolation of primary EBV-positive DLBCL B lymphocytes. All procedures followed were in accordance with the ethical standards and approved by the responsible committee of the First Affiliated Hospital of Nanjing Medical University, Jiangsu Province Hospital (2017-SRFA-167). Primary cells were derived from ascites of the patient and informed consent was obtained from all subjects. Density gradient separation of primary cells was performed with lymphocyte separation medium (Sinopharm Chemical Reagent Co., Ltd.). B lymphocytes were further isolated via negative selection using a BD IMag Human B Lymphocyte Enrichment Set-DM according to the manufacturer's instructions. In brief, the biotinylated Human B Lymphocyte Enrichment Cocktail (component  consists of the following biotinylated antibodies: anti-human CD3 (clone UCHT1), anti-human CD41a (clone HIP8), anti-human CD43 (clone L60), and anti-human CD235a (clone GA-R2). These monoclonal antibodies recognise antigens expressed on erythrocytes, platelets, and peripheral leukocytes, but not on B lymphocytes. The BD IMag Streptavidin Particles Plus-DM (component 51-9000810) magnetic particles, which contain streptavidin covalently conjugated to their surfaces, were used for the B lymphocyte isolation. Immunohistochemistry. Tumour tissues were fixed in 4% paraformaldehyde, embedded in paraffin, and sectioned before staining with haematoxylin-eosin (H&E) and antibodies against CD31 (GB11063-2, Servicebio). The areas of positive staining were quantified using ImageJ software and expressed as a percentage.

Reagents
RNA sequencing analysis. The RNA-seq process was performed on the BGISEQ-500 platform (The Beijing Genomics Institute). Sequencing reads containing low-quality reads, adaptor-polluted reads and reads with a high content of unknown bases were removed by SOAPnuke (v1.5.2) before downstream analyses. Clean reads were mapped to reference transcripts using Bowtie2 (v2.2.5) 54 . Gene expression was calculated for each sample using RSEM (v1.2.12) 55 . The DEGseq algorithm was used to identify differentially expressed genes (DEGs) 56 . Gene Ontology (GO) classification, functional enrichment analysis, KEGG pathway classification 57-59 and functional enrichment analysis were performed for the DEGs.
Network construction and analysis. The molecular interaction information was obtained from the STRING 11.0 online database (https ://strin g-db.org) 60 . Interactions with a confidence score of greater than 0.400 were selected for network construction. The Molecular Complex Detection (MCODE) plugin 61 in Cytoscape software (v3.5.1) was used to further divide the whole network into modules using a cut-off for the node connectivity degree of greater than 3 (Haircut: enabled; Fluff: enabled; Node Score Cut-off: 0.2; K-Core: 2; Max. Depth: 100), generating Clusters 1-3 ( Supplementary Fig. S3). By disabling the Fluff function, which expands the core cluster by one neighbour shell, we further extracted the core module through Cluster 3 based on the above parameters.
Statistical analysis. All statistical analyses, excluding those for RNA-seq, were performed using GraphPad Prism 7. Differences between individual groups were tested for statistical significance using unpaired two-tailed Student's t tests (*P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.). The false discovery rate (FDR) was used in functional enrichment analyses; the q value was utilised as an FDR-based measure of significance for differential expression analysis 63 .

Data availability
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.