Background
Diffuse large B-cell lymphoma (DLBCL) is a highly aggressive form of non-Hodgkin’s lymphoma (NHL), originating from B-lineage lymphocytes, accounting for 30–40% of cases [
1‐
3]. Moreover, this lymphoid neoplasm has highly variable gene expression profiles and genetic alterations [
4,
5]. The most common up-front treatment is R-CHOP [
6]. However, due to its heterogeneity, 60% of patients are curable with combination therapy and the remainders still succumb to the disease [
7]. Unfortunately, those who develop the disease refractory to up-front treatment or relapse after remission have a poor prognosis [
8‐
10]. To understand the underlying mechanisms of DLBCL progression, more studies are necessary. As a result, stratifying DLBCL patients and developing predictive models can provide more precise molecular subtyping and, therefore, more customized treatment.
Epigenetic alterations are directly linked to lymphoma pathogenesis. As one of these post-translational modifications, histone acetylation has been extensively studied [
11]. Post-translational modifications of histones regulate transcription and DNA repair and are linked to the stable maintenance of repressive chromatin [
12]. The N-terminal tail of histones contains lysine residues that are tightly regulated by acetylation regulators. Acetylation reactions are catalyzed by histone acetyltransferases (HATs), while deacetylation reactions are catalyzed by histone deacetylases (HDACs) [
13]. Besides, other regulators recognize modified histones and play a role in acetylation. The three types of regulators are regarded as “writers,“ “readers,“ and “erasers.“ Increasing evidence suggests that the tumor microenvironment (TME) plays a critical role in malignancy development and maintenance. The process is accomplished through sustained proliferation and immune evasion [
14,
15]. As a result of several signal molecules being activated and inhibited, histone acetylation has recently been linked to cancer and TME. In various types of human malignancies, aberrant expression of HDACs has been reported [
16,
17]. It has also been reported that many HDAC inhibitors are effective against several hematologic and solid malignancies [
18,
19]. The HAT paralogs p300 and CBP are involved in many vital cellular processes and have critical roles in several pathological conditions, including cancer [
20‐
22]. Current studies indicated that somatic mutations affecting CREBBP and EP300 are a hallmark of DLBCL [
23]. Multiple studies have also shown that histone deacetylase inhibitors (HDACi) can improve the abilities of the immune system to eradicate tumor cells by changing the TME through various mechanisms [
24,
25].
Collectively, histone acetylation is critical in regulating TME. Targeting histone acetylation modulators can disrupt the resistance to cancer immunotherapy. Recent research has concentrated on individual histone acetylation modulators and their impact on cancer treatment and prognosis. We retrospectively collected transcription information from public databases to better understand how histone acetylation regulators impact the immune system. We investigated histone acetylation regulatory variables and infiltration of immune cells, as well as the value of HAscore in targeted DLBCL immunotherapy.
Materials and methods
Dataset acquisition and clinical samples
The Gene Expression Omnibus (GEO) and the Cancer Genome Atlas (TCGA) databases provided gene expression data and comprehensive clinical annotations. Transcriptome data was derived from fragments per kilobase million and converted into transcripts per million (TPM) and then transformed into log2 (TPM + 1) values for further analysis. The UCSC Xena database also provided genomic mutations (including somatic mutations and copy number variations (CNV). This study focused on the DLBCL cohorts (GSE10846 and GSE31312) as well as the TCGA-DLBC. The R package limma “normalizeBetweenArrays” package was used for normalization after gene symbol conversion (R version: 4.1.2; Bioconductor version: 3.13). A method in “sva” package called “ComBat” was utilized to adjust for batch effects caused by non-biotechnical bias [
26,
27].
Cell lines
Human DLBCL cell lines OCI-LY1, OCI-LY3, OCI-LY8, OCI-LY10, and U2932 were cultured in IMDM (Gibco, MD, USA), supplemented with 10% FBS (Gibco). The cells were maintained under optimal conditions, with a temperature of 37 °C and a humid atmosphere containing 5% CO2. Peripheral blood mononuclear cells and serum were isolated from healthy volunteers.
Lentiviral generation and cell transfection
The stable knockdown of KAT2A was encoded by cloning the shRNAs into lentiviral vectors (Beijing Syngentech Co., Ltd., China). The manufacturer’s instructions were followed for lentivirus infection. To select the stably transfected cells, the medium was supplemented with puromycin (2.0 g/ml, Sigma-Aldrich, USA). A 72-hour timeframe was used for the collection and analysis of cells.
qRT-PCR
Total RNA was extracted using RNAiso Plus (TaKaRa, Dalian, China) following the manufacturer’s protocol. Reverse transcription reactions were undertaken using reverse transcription reagents (Vazyme). Expression levels of specific genes were then measured using qRT-PCR on LightCycler 480II system (Roche, Basel, Switzerland). Normalized results were determined based on GAPDH expression.
The KAT2A primer sequence was as follows:
KAT2A: forward 5-CCCGCTACGAAACCACTCAT-3, reverse 5-GCATGGACAGGAATTTGGGGA-3.
Cell proliferation assay
The cells were added to 96-well plates at a density of 1 × 104 per well. The proliferation of cells was determined after 24, 48, and 72 h of exposure to CCK-8 (Dojindo Laboratories, Kumamoto, Japan). After incubation for four hours, the OD was measured at 450 nm.
Western blot
The Western blot analysis was performed as described previously [
10]. The primary antibody used in this assay was KAT2A (66575-1-Ig, Proteintech). GAPDH was used as an internal reference.
Flow cytometry
Cell cycle and apoptosis were determined by Navios flow cytometer. Cells were stained for 15 min and monitored for cell cycle. Annexin V-PE/7-aminoactinomycin (7AAD) apoptosis detection assays (BD Biosciences) were used to detect apoptosis.
Consensus clustering analysis
38 recognized histone acetylation regulators were isolated from the published literature [
28‐
30]. An unsupervised clustering algorithm of 884 DLBCL samples was performed based on their expression levels for 38 histone acetylation regulators. The patients were clustered using the R package “consensusClusterplus”, and the classification was performed 1,000 times to ensure stability [
31]. Principal Component Analysis (PCA) was employed to illustrate the clustering conditions of given samples with tumors by dimension reduction.
Gene set variation analysis (GSVA)
We used the “GSVA” R package to analyze differences in histone acetylation modification patterns between biological processes. The gene sets “c2.cp.kegg.v7.5.1.symbols” were accessed from MSigDB. Genes related to histone acetylation modifications were functionally annotated with a threshold of FDR < 0.05.
Identification of DEGs in histone acetylation modification characteristics
Patients were divided into three groups by their expression of histone acetylation regulators. An analysis of differentially expressed genes (DEGs) exhibiting various characteristics among histone acetylation patterns was conducted using the limma R package [
32]. Accordingly, the DEGs were selected based on |logFC|> 1 and adjusted
P values < 0.001. Kyoto Encyclopedia of Genes and Genomes (KEGG) [
33‐
35] pathway analysis and Gene Ontology (GO) biological were processed using the R package “clusterProfiler” and “org.Hs.eg.db”.
Estimation of tumor environment cell infiltration
Adapted from Charoentong’s study [
36], the gene set was used to define immune cell types. A fraction of the immune cells infiltrating each sample was determined based on their relative abundance. Based on the single-sample gene set enrichment analysis (ssGSEA) algorithm, ESTIMATE was calculated by the R package “ESTIMATE” based on immune cell and stromal cell-specific gene expression levels to calculate a score reflecting the level of immune cell and stromal cell infiltration. Boxplots of ESTIMATE score, immune score, and stromal score for different risk groups were demonstrated using the R package “ggpubr”.
Evaluation and generation of HAscore
There has been an effort to develop a method for quantifying the pattern of histone acetylation modifications by developing a scoring approach (HAscore). The overlapping DEGs from different HA clusters were first accessed to divide patients into several clusters via unsupervised clustering. Following a univariate Cox regression model, prognostic-related DEGs were selected to construct HA gene signatures. Following determining the prognostic value of gene signature scores, a method similar to the Genome Grading Index [
37] was applied to define the HAscore for each patient: HAscore = ∑(PC1i + PC2i), where i indicates the expression value of each histone acetylation regulator.
Statistical analysis
Patients’ survival was analyzed by the Kaplan-Meier analysis, and overall survival (OS) between subgroups by the log-rank test was determined. Using the “surv-cutpoint” function in the “survminer” R software package, the optimal cut-off point was determined, resulting in the classification of patients into high- and low-HAscore subgroups, as well as high- and low-KAT2A expression subgroups. The difference between three or more groups was examined using one-way ANOVA and Kruskal-Wallis tests [
38].
p values were analyzed using two-sided statistical tests in all statistical analyses, and
p < 0.05 was considered statistically significant. At least three independent experiments were conducted, and the mean and standard deviation (SD) were used in describing the experimental data. R and GraphPad Prism (version 8.0) were utilized to conduct all the statistical analyses.
Discussion
Histone acetylation is an epigenetic modification necessary for cancer biology. KAT7, for example, inhibits tumor cell proliferation and invasion by acetylating H4K5 at promoters of FOXO1 and FOXO3a genes [
40], and HDAC5 loss increased H3K27-ac acetylation and circumvented oncogene-related cell cycle repression [
41]. Considerable research has been conducted on individual histone acetylation regulators and their role in cancer; however, a comprehensive assessment of these regulators and their interactions remains lacking. Our research aims to develop a more complete understanding of the different histone acetylation patterns and their associated biological properties.
Based on a preliminary analysis of the expression of histone acetylation regulator genes, most genes were significantly different between normal and tumor samples with an association with prognosis. According to unsupervised clustering of 36 modulators, patients were categorized into three histone acetylation phenotypes. Survival analysis revealed three distinct patterns of histone acetylation modification associated with prognosis. To further characterize these different histone acetylation phenotypes, we determined DEGs among them. The HAscore model was developed based on these genes to assess histone acetylation phenotypes in individual patients quantitatively. All patients in HAcluster A have a low HAscore, associated with the poorest survival outcome. Moreover, we found a close relationship between most acetylation-related genes. One of these risk factors, KAT2A, has been found to be closely related to histone acetylation regulators.
In order to shed light on the mechanisms underlying the different prognoses of patients with different phenotypes, the biological characteristics of each pattern were examined. The results of GSVA analysis confirmed that HAcluster A showed enrichment in several metabolically relevant pathways. Tumorigenesis and tumor progression are supported by abnormal metabolic activity, which allows cells to obtain essential nutrients from the environment [
42]. Moreover, we found that PARP, JAK-STAT, MAPK, NOTCH, and apoptosis were significantly activated in HAcluster A. Enrichment of these malignancy signaling pathways may be indicative of malignant progression, which may result in a poor prognosis [
43‐
45].
A growing number of studies have suggested that TME components may contribute to cancer development [
46‐
48]. An investigation into the relationship between histone acetylation modifications and TME cell infiltration was undertaken to to understand the antitumor immune response to DLBCL better. HAcluster A, with the lowest OS, exhibited a significantly higher level of MDSCs, which play a crucial role in immunosuppression [
49]. MDSC is characterized by its ability to suppress immune cell function, including inhibition of T cell proliferation [
50]. Each pattern is characterized by a different degree of TME infiltration, with immunosuppression characteristic of HAcluster A group. By analyzing histone acetylation patterns, we evaluated the potential therapeutic effects of HAscore based on differences in signaling pathways and tumor microenvironments between the different patterns. In the high-score group, higher levels of immune-activating cells were observed, including activated CD4 T cells and CD56 bright natural killer cells. CD4 + T cells have antitumor activity through the production of effector cytokines that activate CD8 + T cells [
51]. By modulating DC and T cell responses, CD56 bright natural killer cells can positively influence the anticancer response [
52]. This indicated that the HAscore contributed to further defining the microenvironment and thereby guiding more effective treatment.
Furthermore, our research revealed that KAT2A correlates with DLBCL prognosis and functions independently. An enzyme in the HAT family, lysine acetyltransferase 2 A (KAT2A), is involved in transcriptional activation through histone acetylation, histone succinylation, and recruitment of transcriptional coactivators [
53]. Several studies have demonstrated that KAT2A functions as an epigenetic oncogene in several cancers [
54,
55]. In previous studies, KAT2A has been shown to be a viable target for reducing the growth of acute myeloid leukemia by significantly promoting myeloid differentiation and apoptosis [
56]. However, little was investigated between KAT2A and DLBCL. In this study, we utilized the bioinformatic methods to find that histone regulator KAT2A was a risk factor in DLBCL. The stable KAT2A KD cell lines were constructed for further study. Moreover, our research found that KAT2A deficiency could prohibit cell proliferation, promote cell apoptosis, and arrest cells in the G2/M phase. These findings suggest that inhibiting KAT2A may inhibit tumor growth.
In conclusion, our findings suggest that genetic signatures derived from histone acetylation regulators can be used to provide personalized survival assessments for patients newly diagnosed with DLBCL. It might provide clinicians with valuable information regarding treatment decisions, follow-up, and prognosis for their patients. There are some limitations to the study, despite its strengths. All the data we used were from public databases. It would be beneficial to collect more prospective real-world data to confirm its clinical utility. On the other hand, the specific mechanism of the effect of KAT2A in DLBCL still needs to be further explored.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.