Clinical Risk Stratification and Modifiable Risk Factors for Hepatitis B Virus-Related Follicular Lymphoma

Introduction

Follicular lymphoma (FL) is the most frequent indolent lymphoma, showing discordant incidence of all newly diagnosed non-Hodgkin lymphomas (NHLs) in the Western countries with 20–25%¹ and in China with 10–12%.² The experienced strategy of watchful waiting is for asymptomatic patients with advanced-stage and low-tumor-burden,³ and CD20-directed immunochemotherapy is served as standard first-line therapy of advanced-stage FL.⁴ Numerous evidences have revealed the 2–3-fold high risk of chronic hepatitis B virus (HBV) surface antigen (HBsAg)-positive infections to develop B-cell NHLs including FL, which exhibits poor overall survival (OS) and a high incidence of disease progression within 24 months (POD24).⁵ China is featured with a high prevalence of the HBV infection. HBV confers chemoresistance,⁶ and the selection of time to start the treatment and the proper therapeutic regimens including immunochemotherapy and anti-viral prophylaxis remains the great challenges to improve survival.^7,8

There are multiple clinical prognostic models to stratify the risk categories in FL patients with useful practices, including the classical Follicular Lymphoma International Prognostic Index (FLIPI) and FLIPI-2.^9,10 But they are not accurate enough to guide treatment strategies.⁴ Recently, the clinicogenomic models such as m7-FLIPI scoring with seven genes, a developed 23-gene signature model which captures various aspects in tumor biology, and BioFLIPI which integrates the microenvironment biomarker, could bring to precision medicine.¹¹ However, due to the markedly distinct outcomes in HBV-related FL differing from general FL populations,^5,12 these models ignored key factors derived from virus infection such as virus loads.¹³ Clinically, although some HBV-infected patients with low or intermediate FLIPI scoring, they also suffered early progression.

Studies have disclosed distinct genetic/epigenetic features in FLs. For example, EZH2 as histone methyltransferase (H3K27me3) plays crucial role in differentiation and function of germinal center B cells (GCB) and immune remolding. The EZH2 inhibitor (eg tazemetostat) has acquired US food and Drug Administration (FDA) approval for relapsed /refractory (R/R) FL.¹⁴ Other epigenetic modifiers, such as BCL7A, ARID1A/B and KMT2D, all commonly shared in multiple disease stages (initial, progression, transformation) and were potentially insufficient to be well compared.¹⁵ It is worth noting that HBV infection has various impacts on the natural immune response.¹⁶ Considering the influence from viral infection on immune regulations, the more aspects of epigenetic modification in lymphoma B cells beyond EZH2-dependent mechanism remains uncertain.

This study aimed to explore the risk factors derived from the HBV infection for integrative prognostic stratification based on FLIPI scoring in FL. The distinct methylation signatures of immune-regulation genes between pre-POD24-HBsAg+FL and paired POD24-HBsAg+FL were explored for better auxiliary understanding of the impact from HBV infections and provided the novel choice of epigenetic treatment strategies in future.

Materials and Methods

Patients

Totally 276 patients were retrospectively reviewed and confirmed the diagnosis of follicular lymphoma by an experienced haematopathologist between January 2010 and January 2022 from Harbin Medical University Cancer Hospital (n = 104), the Tianjin Medical University Cancer Institute and Hospital (n = 73) and the Fourth Hospital of Hebei Medical University (n = 99). All patients had previously given informed consent. Patients who were asymptomatic and the low tumor burden retained watchful waiting as the initial strategy until disease progression. The advanced-stage patients underwent rituximab-containing immunochemotherapy. Patients with a high tumor burden and responding well to initial treatment would receive rituximab-maintenance therapy. Patients with high HBV DNA load (>10⁴ copies/mL)¹⁷ received chemotherapy without rituximab. All the HBsAg+ patients were given antiviral prophylaxis. POD24 referred to progression or relapse of the disease within the initial 24 months after diagnosis following the first-line therapy. The FLIPI score was calculated based on international criterion as previous.⁴ All experiments involving human patients were approved by the Research Ethics Committee of Harbin Medical University Cancer Hospital (Approval no. KY2023-79) in accordance with the Declaration of Helsinki. Personal data were anonymized in the study. The cohort study was conducted complying with the STROBE guideline.

Construction and Validation of Risk Stratification

Risk score was calculated for each patient as a weighted sum of HBV infection-related factors determined by the univariable and multivariable Cox proportional hazard regression and FLIPI scoring using linear regression (eigenvalue, Supplementary Table 1). The integrative score for each patient was calculated as total sum of each epigenvalue for positive factors. We estimated the optimal cutoff value to maximize the Wald statistic by the Receiver Operating Characteristic (ROC) curve and dichotomized patients into high-risk and low-risk subgroups. Kaplan-Meier Log rank test was used to estimate the prognosis. Patients who were alive or experienced PD were censored at the last date known alive or at their last disease assessment. Patients were not evaluable for POD24 if they were censored (eg, lost to follow-up) or died within 24 months without progression of disease. The internal validation (n = 35) from Harbin Medical University Cancer Hospital and the external validation (n = 153) from the Shandong University Cancer Center, the First Affiliated Hospital of Harbin Medical University, and the Henan Cancer Hospital, were performed after acquiring informed consents and receiving the approval by the Research Ethics Committee of these centers. The same inclusion and exclusion criteria of these cohorts as abovementioned aligned with study cohorts.

Illumina DNAm EPIC Analysis

DNA Methylation Experiment

The Formalin Fixed Paraffin Embedded samples of paired pre- and post-lymph node biopsies from HBsAg+ patients with POD24 after cyclophosphamide, doxorubicin, vincristine and prednosine plus rituximab (R-CHOP) or bendamustine plus rituximab (BR) treatment (excluding the histological transformation) were obtained. The DNA was separated by application of the DNeasy Blood and Tissue Kit (Qiagen). The Nanodrop 2000 (Thermo) was used to assess the purity and concentration of DNA. EZ DNA Methylation Kits (Zymo Research, USA) and the 850K BeadChips with 820,000 probes were applied to detect the converted products as per the manufacturer’s instruction (Illumina).

Data Analysis

The data were primarily analyzed by a ChAMP package in R with the normalized β-value matrix. The EPICanno.ilm10b4.hg19 was used to annotate all CpG sites. Differential Methylated CpGs Positions (probes) were calculated through the champ with adjusted p values by the Benjamini–Hochberg method. CpGs with |Δβ|≥0.20 and DMPs with adjusted p ≤ 0.05 were defined. Functional Epigenetic Modules reflect protein to protein interaction network (PPI) using the FEM package. A Conumee package was applied to analyze the copy number variation. The proportions of the immunocytes were estimated by the EpiDISH16 and HEpiDISH17 methods as indicated previously.¹⁸ The two-factor ANOVA was conducted for statistical analysis with the P-value and F-value.

Myc-Driven B-Cell Lymphoma Chimeric Mice Model

The experimental procedures of mice were performed in stringent accordance with the institutional guidelines delineated by Harbin Medical University, as per the Guide for the Care and Use of Laboratory Animals and standards established by the Association for Assessment and Accreditation of Laboratory Animal Care International. The Southern Model Biology Research Center (Shanghai) and the Institutional Animal Care and Use Committee of Harbin Medical University, having vetted all procedures, duly approved the entirety of the study involving mice under protocols #HMUIRB2025004, and #SCXK (shanghai) 2019-0002. The experimental mice were age- and sex-matched, ranging from 8 to 10 weeks. No evidence of sex-based influences or biases was observed. The following strains were obtained from The Southern Model Biology Research Center (Shanghai): H11-CAG-LSL-Myc (strain NM-KI-00039), Cd19-Cre (NMX) (strain NMX-KI-190023). Using CRISPR/Cas9 technology, heterozygous mice with targeted conditional overexpression of Myc in the H11 gene were obtained. The CD19-Cre knock-in/knock-out allele has a Cre recombinase gene inserted into the first coding exon of the CD19 antigen gene as protocol (JAX stock #006785). They were back-crossed to C57BL/6J genetic background through at least six generations before used for experiments.

NP (4-hydroxy-3-nitrophenyl acetyl) was conjugated to CGG (Chicken γ-globulin) at a ratio of NP18-CGG (BiosearchTM technologies, IMMB1-003). Myc-Cd19-Cre or wild-type mice were immunized intraperitoneally with 100ug alum (Biodragon, KX0210054) precipitated 50 μg NP18-CGG. Superficial lymph nodes were analyzed. All the animal studies designs were performed complying with the ARRIVE guidelines.

Bulk RNA Sequencing

RNA was extracted from patients with POD24 or mice lymph node tissues by column purification using FFPE RNA extraction kit (Meji, IVD4144) and total RNA extraction kit (Tiangen, DP419). The RNA samples were strictly quality controlled reaching at least 100ng. Library construction was performed according to the Hieff NGS MaxUp Human rRNA Depletion Kit (rRNA and ITS/ETS) item number 12257ES96 or Poly (A) (Hieff NGS mRNA Isolation) Master Kit V2, No. 12629ES96). The purified double-stranded cDNA was subjected to end repair, A-tail addition, sequencing adapter, PCR amplification, and DNA purification beads to purify the PCR products. The library connector used a MGI connector with a double-ended index tag. The different libraries were pooled according to the requirements of effective concentration and target amount of data, and then DNBSEQ-T7 PE100 sequencing was performed. PE100 (Pair End 100 bp) refers to high-throughput double-end sequencing, with 100 bp at each end. Combined analysis was performed based on 935K data and Gene expression data of RNA chip of pre-POD24-HBsAg+FL and paired POD24-HBsAg+FL samples.

Statistical Analysis

Statistical analyses were performed by SPSS 25.0 version software (SPSS Inc. Chicago, USA), R version 4.1.3 or GraphPad Prism 9 with p value. The χ²-test was performed for categorical variables. Univariable and multivariable Cox proportional hazards regression was applied for the comparisons of PFS and OS by factor categories. Survival data using Kaplan–Meier curves were assessed differences between groups through the Log rank test. Groups were compared using an unpaired t test, Mann Witney U-test or ANOVA with Tukey’s post hoc test depending on the number of groups and distribution. The p value was <0.05 for significance (two-sided).

Results

Clinicopathological Features

Of the total 276 patients, 131 patients were seropositive for HBsAg at baseline across the analysis population (FL, n = 276). Shown by serum HBsAg arm, 42 (32.1%) patients with HBsAg+ had a high HBV DNA load (>10⁵ copies/mL). HBsAg+ patients had more nodal sites with larger sizes (32.8% vs 17.9%, P = 0.004) and higher incidence of POD24 (31.3% vs 22.1%, P = 0.031) than HBsAg− patients. There was no difference in the incidence of an absolute lymphocyte count (ALC) <1000/μL (P = 0.630) between the two arms (Table 1).

Table 1 Patient Characteristics Summary in Study and External Validation Groups

Totally 131 patients with HBsAg+ were summarized in Table 2. The older age (>60 years, 35.4% vs 24.5%, P = 0.034), high HBV DNA at baseline (52.4% vs 21.3%, P < 0.01), HBeAg-positive at baseline (42.1% vs 28.2%, P = 0.022), and low ALC at baseline (50.0% vs 30.7%, P = 0.045), demonstrated a higher incidence of POD24 events. But rapidly decreasing HBV DNA (<10³ IU/mL) within 1 month after prophylactic nucleoside analogues treatment (NAT) exhibited a lower incidence of POD24 (22.9% vs 41.0%, P = 0.026). The FLIPI scoring was not significantly associated with the incidence of POD24 (P = 0.226) (Table 2). Seventy-three patients occurred POD24 and those with HBsAg+ had a higher rate of histological transformation than HBsAg− patients (7.3% vs 3.1%, P = 0.046) (Supplementary Table 2).

Table 2 Correlation Analysis Between the POD24 Event and Clinical Parameters in HBsAg+ FL Subgroup

Risk Analysis of HBV Infection on Prognosis

An overall median follow-up in HBsAg+ patients was 38.7 months (IQR, 28.9–51.5).

Patients with HBsAg+ showed a shorter progression-free survival (PFS, log-rank P = 0.004, Figure 1A) and OS (P = 0.039, Figure 1B) than those with HBsAg−. In non-POD24 arm, HBsAg+ patients also had shorter PFS (P = 0.036, Figure 1C) and OS (P = 0.007, Figure 1D) compared with HBsAg− patients. In cases of POD24, HBsAg+ indicated no obvious correlation with OS (P = 0.896, Figure 1E). The similar association between HBsAg+ and POD24 risk was observed in an external validation cohort (Table 1 and Supplementary Figure 1). There was an association between shorter survival and high HBV-DNA load, but not HBeAg status (Supplementary Figure 2). The HBsAg+ was not an independent prognostic factor for PFS and OS (Tables 3 and 4). Among the HBsAg+ arm, decreased ALC, high HBV-DNA load, persistent viral load after prophylactic NAT, and treatment with BR regimen were associated with shorter PFS (Table 5).

Table 3 Univariable and Multivariable Cox Regression Analysis with Respect to Progression-Free Survival of Clinicopathological Variables

Table 4 Univariable and Multivariable Cox Regression Analysis with Respect to Overall Survival in All Baseline Parameters

Table 5 Univariable Cox Regression Analysis with Respect to Progression-Free Survival and Overall Survival in HBsAg+ Subgroup

Figure 1 Kaplan–Meier survival curves in study cohort (n = 276), stratified by the presence of hepatitis B surface antigen (HBsAg) at diagnosis. Patients in the HBsAg+ group had worse (A) progression-free survival (PFS) and (B) overall survival (OS) compared with those in the HBsAg− group. In non-POD24 subgroup, patients in the HBsAg+ group had poorer PFS (C) and OS (D) compared with the HBsAg− group. In POD24 subgroup, there was no significant difference of OS (E) between HBsAg+ group and HBsAg− group.

HBV-Related Factors Combined with FLIPI Improves Risk Stratification

Cox proportional regression analysis identified five risk factors for prognostic stratification involving in HBsAg, viral load decline after NAT, HBV DNA load, ALC and treatment strategies. We calculated another risk model using linear regression analysis in which HBV DNA load had the highest scores with eigenvalue of 1.029 (Supplementary Table 1), which could specially predict POD24 in HBV-associated FL patients with cutoff values as 3.236 for risk stratification by dichotomy (Area Under Curve (AUC) = 0.616, P = 0.002) (Figure 2A). Notably, in the study cohort, the model had a higher sensitivity to predict POD24 compared to the FLIPI score (35.8% versus 29.6%) (Figure 2B). The high-risk model was associated with significantly shorter OS (median OS of 67.7, P < 0.01), outperformed that in high-risk FLIPI (OS of 69.8; P < 0.01, Figure 2C). The internal validation cohort and external validation cohort confirmed that the model also had a higher sensitivity for POD24 prediction than FLIPI scoring (40.0% versus 33.3%; 39.1% versus 29.7%, respectively, Figure 2B). Meanwhile, the high-risk model in internal validation cohort (Figure 2D) and external validation cohort (Figure 2E) was related to shorter OS (62.1, P = 0.009; 58.8, P < 0.01, respectively), outperformed that in high-risk FLIPI (70.5, P = 0.038; 69.4, P < 0.01, respectively), albeit at the cost of lower specificity (Figure 2A).

Figure 2 HBV-related risk factors slightly improved the risk stratification based on FLIPI scoring. (A) The constitutions of risk stratification are showed by colorful graphs with canonical FLIPI and five HBV-associated parameters. Each classification uses a set of decreasing colors. The best cutoff value of 3.236 for the new risk stratification with AUC of 0.616 were evaluated by ROC curve. (B) The proportion of POD24 events in risk stratification of model and FLIPI, in study cohort, in internal validation cohort, and in external validation cohort, respectively. Risk groups for overall survival (OS, month) were assessed by Kaplan–Meier survival curves based on the new risk stratifications (upper) and FLIPI scoring (below) in study cohort (C), in internal validation cohort (D), and in external validation cohort (E). Abbreviations: AUC, Area under Curve; ROC, Receiver Operating Characteristic.

Methylation Profiling Distinguish in Sequential Pre-POD24-FL and POD24-FL with HBsAg+

As previous studies of susceptible methylation in FL had been limited to the initial stage of disease,¹⁹ we investigated paired pre-POD24-HBsAg+ and POD24-HBsAg+ biopsies in FL to determine whether there was a dynamic alteration in methylation status over time with disease progression. Significant high methylation in 2029 Differentially Methylated Positions (DMPs) in POD24-HBsAg+ FL relative to pre-POD24 sample (P value < 0.01 and Δβ ≥ 0.2; Figure 3A–C) and showed predominantly discrepant profiles in POD24 stage (Figure 3D). 6.3% (40/632) Cytosine-Phosphate-Guanine (CpG) loci were aggregated near to the downstream 2.5 kb from transcription start sites (TSS) and 9.3% (59/632) were near to the upstream 2.5kb sites in pre-POD24-HBsAg+ FL, but almost none located in TSS. Notably, 6.2% (39/632) CpG loci were located within the TSS in POD24-HBsAg+ FL (P ≤ 0.01, Figure 3E and F).

Figure 3 Multiple differential methylation positions (DMPs) with distinct distributions among pre-POD24-HBsAg+FL and paired POD24-HBsAg+FL. (A) The schematic map shows the distribution of DMPs on each chromosome succinctly. Hypermethylated and hypomethylated CpGs are represented in Orange and blue respectively. (B) The scatter plots show the differences of CpGs between groups. The orange points represent DMPs, and the blue points represent undifferentiated points. (C) The volcano plots show the Δβ and significance of CpGs between groups. The upper right region and the upper left region represent hypermethylated and hypomethylated points with statistical differences respectively. (D) Heat maps were used to view the similarity of samples between groups after z-Score scaling. (E) The methylation differences of 5kb upstream and downstream of TSS between groups. (F) The pie chart shows the distribution of CpG islands (left) and gene regions (right). (G) The Heatmap showed the different methylation of MHC class I genes.

Correlation with Gene Expression of Tumor Immunity

Presented based on algorithm of methylation signatures, we found a higher infiltration of CD4+ T-cells (P = 0.013), monocytes (P = 0.008), and natural killer cells (P = 0.004), but lower proportions of nonmalignant B cells (P = 0.005) and a lower lymphocyte-to-monocyte ratio (0.85 vs 6.03) in POD24-HBsAg+ than in pre-POD24-HBsAg+ sample. No significant changes of intratumoral CD8+ T cells (P = 0.341) and neutrophils (P = 0.175) between two groups (Supplementary Figure 3 and Supplementary Table 3). 682 DMPs corresponding to immune function-related genes (eg IL32, ITK, NCR2, MHC I class molecular) were identified in POD24-HBsAg+ compared to pre-POD24-HBsAg+ (Figure 3G and Supplementary Table 4).

One hundred and ninety-six DMPs showed a significant correlation with gene expression involved in the tumor immunity process (Supplementary Figures 4 and 5). For example, CXCR5, TNFRSF10B, HLA-DQB1, RUNX1 and CTH were observed incremental methylation in POD24-HBsAg+ FL while ITGB2, TBX21 and NCF4 were over-represented in pre-POD24-HBsAg+ FL. KMT2A, EP300-AS1, and ARID1B related to malignant transformation effect²⁰ were methylated, while MYC related to aggressive transformation was defective methylation (Supplementary Table 4). Our CpG datasets were extensively overlapped in immune-regulated genes profiling²¹ noted in GSE14582.²² Seed cluster analysis showed the interaction network center on TNFRSF1A (GSE23002), ACTG1, IQCE and LTA (GSE131559) in POD24-HBsAg+ FL. The FOXP1, MLH3, UBQLN1, and MTA1 were significantly over-represented in pre-POD24-HBsAg+ FL (Figure 4 and Supplementary Tables 4, 5).

Figure 4 Functional Epigenetic Modules showed seed gene (A) TNFRSF1A-, ACTG1-, IQCE-, LTA-interaction network in POD24-HBsAg+FL tumors and (B) FOXP1-, MLH3-, UBQLN1- and MTA1-interaction network in pre-POD24-HBsAg+FL. Each node is a gene, and the color of the node presents statistical significance of their contribution to the disease regulation. Blue for positive significance, yellow for negative significance, and gray for no difference. The line between nodes is the degree of connectivity between genes, and the thickness of the line represents the level of correlation.

High-density DNA methylation arrays allow for determining copy number alterations, and frequently encountered alteration include segmental chromosomal amplifications of ITK, IL7R, CD14, and TNFAIP8 in POD24-HBsAg+ FL. Whereas, segmental chromosomal deletions of IL21R, TGFB1I1, and TNFRSF12A were also related to POD24 occurrence (Supplementary Figure 6 and Supplementary Tables 6, 7).

Validation of the Correlation Between CpG Methylation and mRNA Expression

We performed whole transcriptome sequencing (RNA-seq) of 6 biopsies samples from another patient cohorts with HBsAg+ FL (n = 3, pre- and POD24-matched samples). We searched for associations between DEGs expressions (log2FC) and mCpG dela beta values located 200kb upstream of TSS. In POD24-HBsAg+ FL, out of 2877 low-methylated (Δβ ≤ −0.2) sequences, 454 DEGs were upregulated involved in metabolism-enhanced genes (SLC2A2,²³ GLS2²⁴) and immunosuppressive molecules (IL1R2, IL22RA1²⁵). Notably, 969 promoter methylations contained 115 associated down-regulated mRNA expression matched to the abovementioned (ITGB2, IL21R, C1QB²⁶) (Figure 5A). Generally, the gene with up-regulated expression primarily existed in low-methylation status (quantile) at TSS200 in POD24-HBsAg+ FL compared to pre-POD24-HBsAg+ FL (Figure 5B). Top DEGs expression (FPKM values) correlated to distinct promoter methylation sites within top specific chromosome regions (60%), also known as imprinted regions, which were expected as modifiers for expression of origin (Figure 5C and D). We next captured the methylation sites highly correlated to the CXCR5 mRNA expression were simultaneously less correlated to the mRNA expression of MYC gene (Figure 5E). These two genes were also known as mover for malignant progression of B cell,^27,28 which was representative acting as the “trans” correlations existing universally between different genes. Interestingly, “cis” correlations between same gene were showed that methylation sites were simultaneously highly or low correlated with in mRNA expression, especially existing in CXCR5 (body, spearman 0.94; top 1) which was calculated around three methylation sites on average presenting strong correlation with enhanced mRNA expression in POD24-HBsAg+ FL (Figure 5F).

Figure 5 Correlation of methylation sites and gene expression. (A) Scatterplot corresponds to the associations found between methylation (delt beta at TSS200, x axis) and gene expression (logFC, y axis). (B) The splitviolins plots depicting the distribution of CpG site methylations in DEGs from two samples. Quantile 25%, 50%, 75% lines are shown. The distributions from pre-POD24-HBsAg+FL and paired POD24-HBsAg+FL samples are colored in red and blue respectively. The heatmap (C) and the Circos plot (D) depicting the distribution of associations between CpG site methylation and the level of expression of the corresponding genes. Spearman correlation is used as measure. Euclidean distance and complete method are used in bi-clustering. The outer circle represents different chromosome numbers, and the inner circle may be refined to specific positions in the genome. The correlation was showed by connecting lines. (E) The heatmap specifically depicting the distribution of associations between same CpG site methylation and the level of expression of the distant genes involving in CXCR5 and MYC. (F) Spearman correlation coefficient including Beta value of relevant methylation site location information and FPKM value of CXCR5 gene expression in a specific gene region.

The Impact of CXCR5 on Aggressive MYC-B Cells

Given about 10% to 15% of patients with aggressive B-cell lymphoma have MYC gene rearrangement, we confirmed that MYC mRNA expression was strongly correlated to low-methylation level in POD24-HBsAg+ FL compared with pre- POD24-HBsAg+ FL (TSS200, Pearson 0.059, P < 0.0001; Figure 5A). We further considered aggressive immunogenic model from Myc-Cd19-Cre mice to infer the B-cell progressive mechanism by other methylated genes at transcriptional activation in MYC variants context. Compared with wild type C57BL/6J mice histologically, MYC^+/+ CD19-B cell lymphoma presented lymphoma cells, with medium size, with giant cells and atypia, irregular nuclei and presence of mitotic sign (Figure 6A). Furthermore, RNA-seq from MYC^+/+ CD19-B cell lymph node displayed increased CXCR5 mRNA level in line with the global changes from MYC^+/+ malignant B cells compared to MYC^−/− mice. The chemokine activity of the receptor itself did appear to be aberrantly activated in MYC^+/+ malignant B cells (Figure 6B). Although observed distinct transcriptional enhance of CXCR5 among POD24-HBsAg+ FL (P = 0.00064) involved in silencing chemokine receptors binding, interferon and immunoregulatory interaction between lymphoid cells signaling, there was defective in MYC transcription with functional categories (Figure 6C). This finding suggests that the CXCR5 pathways depicted in Figure 6B is driven by the overall transcription shift of MYC in malignant B cells towards a more paused, less proliferative state which the MYC-pathway is often associated (eg p53, mTOR signaling pathway).

Figure 6 The impact of CXCR5 expression on aggressive MYC-B cells development. (A) Experimental scheme of mice model and timeline. H&E of spleen sections from mice. Scale, 100 μm (left) 50 μm (center) or 10 μm (right). (B) RNA-seq with DEGs profiling in chimera mice, involving in Heatmap, volcano and KEGG enrichment graphic. (C) RNA-seq of DEGs profile between pre-POD24-HBsAg+FL and paired POD24-HBsAg+FL samples, presented with Heatmap, volcano and KEGG enrichment graphic. Red highlighted CXCR5 gene and related critical pathway.

Discussion

FL is known to be a disease of immune dysfunction and is susceptible to infectious pathogens, mainly viruses.²⁹ Risk-adapted treatment approaches including immunochemotherapy (eg R-CHOP, BR) and inevitable prophylactic NAT should be prioritized at the high-risk FL populations. Although anti-HBV therapy was strongly recommended in HBsAg+ patients regardless of HBV DNA levels,³⁰ HBV infection associated with inferior survival⁵ is not really on the risk list. The risk evaluation results were not upgraded by patients with heavy HBV infection and we may misjudge a high-risk patient. We consequently designed this study to investigate that not only HBsAg-positive factor but also HBV-induced immune activities should be dependent prognostic indicators. Three aspects are important to be cautious about POD24 event that firstly, accompanying time with HBV infection,³¹ which may be ascribed to the persistent antigen stimulation for exhausted immune states.³² Second, the active status of HBV replication, which could be interpreted by that a small fraction of antigen-specific CD8+ T cells survived and formed long-lasting memory T cells in an acute inflammatory context, which remained the effective functions upon encountering antigens.³³ Third, the immunosuppressive effect from cytotoxic chemotherapy with decreased ALC, which may owe to the B cell lymphopenia caused by drugs like bendamustine³⁴ and indirectly affect the T-cell infiltrations due to the loss of B-cell helper especially within intra-tumoral tertiary lymphatic structure.³³

Indeed, our result assigned a novel risk stratification that integrates HBV-related¹⁶ and immune-regulated³⁵ clinical factors into FLIPI system, could contribute to enhancing identification of high-risk patients than FLIPI scoring alone, and importantly, validated by a series of patients cohorts from domestic multi-center clinical institution.

While clinically promising and practical in the abovementioned indicators, multiple studies demonstrated that a complex pattern of frequent mutations in epigenetic modifiers that are co-founding events arising during FL development and progression.³⁶ While epigenetic therapies such as histone deacetylase inhibitors (HDACi) have previously been evaluated in FL, the results were overall disappointing and pushed our further knowledge of the epigenomic landscape of this malignancy. DNA methylation with a significant overlap between hypermethylated CpGs and target genes in FL cells, especially the relapsed with POD24 or more aggressive tumors, followed by chemoimmunotherapy.³⁷ The initial conservation of methylation profiles was introduced with more measurable difference between paired pre- and post-POD24 HBsAg+ FL samples. The restricted number of genes with modifiable functional noncoding elements, including promoter and enhancers located proximal or distal to protein-coding genes, respectively, have for the most part been captured, including IL32, ITK, NCR2 and most methylated MHC class I genes,³⁸ which contributed to immunoinactivation and failed to response to immunotherapy.³⁹ Notably, we matched methylation site and correlated gene expression as RNA-seq and ranked the candidates by their correlation. We reasoned that CXCR5, which was dependent lymphoma cell dissemination and tumor-stroma interaction,²⁷ had strong methylation regions. Additionally, CXCR5 mRNA expression can be anti-correlated with the expression of MYC mRNA at the same methylation sites which acts as proto-oncogene²⁸ as “trans” candidate. We therefore posit that the biological connection of enhanced CXCR5 expression and MYC expression with developmental pausing may exist timely, as validated in MYC^+/+ ES cells of FL models. The roles in developmental contexts in which large-scale shifts in the transcriptional programme in these two crucial genes have to rapidly occur, which may constitute a more robust mechanism of POD24 process and new methyltransferase inhibitors to be recapitulated.

However, several challenges still remain to be addressed before they could be acted as critical markers: validation using molecular trials and extension in a larger cohort. The individual molecular predictor should be adjusted to be universally applicable. Some large and collaborative efforts are strived to address these problems. Interestingly, another study showed that anti-CXCR5 chimeric antigen receptor (CAR)-T cells eradicate B-NHL cells and lymphoma-supportive T follicular help cells,²⁷ showing the promising therapeutic values of CXCR5 target aligned with our findings.

We anticipate that the regulatory relationship between methylation and cellular dormancy will have implication that extend beyond identification. The insight here provides exciting new opportunities to developmental therapies of interest to diapause post-translational modifications, strongly wishing to delay POD24 transition in the context of HBV interference.

Conclusions

Integrated HBV-related clinicopathological features into FLIPI scoring system may provide precise pre-treatment risk stratification for HBV-positive FL. Modifiable DNA methylation acts as the potential targets for the combined treatment strategy to delay POD24 occurrence in high-risk HBV-positive FL.