Introduction
Growth hormone-secreting pituitary adenoma (GHPA) is characterized by increased growth hormone (GH) and insulin-like growth factor I (IGF-1) secretion [
1]. It can cause disproportionate skeletal, tissue, and organ growth, which are associated with significant morbidity and higher mortality [
2]. Notably, in the last decade, although effective treatments have been largely improved for GHPA, including advanced surgical procedure, development of new radiotherapy techniques and availability of new medical therapies, ~40% proportion of patients still cannot completely achieve normal levels of GH and IGF-1 [
3].
It is undeniable that alterations in mountainous genes have been involved in the onset and development of GHPA. Mutations in the α-subunit of the heterotrimeric Gs (
Gsα), the stimulatory G protein (
GNAS), α-subunit of protein kinase C (
PKC) and aryl hydrocarbon receptor-interacting protein (
AIP) have been discovered to be associated with GHPA [
4‐
7]. Furthermore, multiplatform analysis has been performed to identify the epigenomic underpinnings of GHPA, which reveals that hypomethylation of promoter regions would increase the expression of
GH1, somatostatin receptor 5 (
SSTR5) and programmed death ligand 1 (
PD-L1) in GHPA [
8]. On the contrary, genes like high-mobility group A1 (
HMGA1),
HMGA2, and E2F Transcription Factor 1 (
E2F1) have been reported to be down-regulated in GHPA [
9]. The complex genetic network in GHPA allows for better understanding of tumor pathology. However, information on expression levels of somatotropinoma genes and how these are specifically regulated are still limited. Thus, exploring the regulatory elements of genes and their corresponding transcription factors (TFs) is critically important for elucidating the process of GHPA.
As a dynamic central regulator of transcription, chromatin is compacted in the nucleus and restricts the access of TFs to cis-regulatory elements [
10]. In order to successfully interact with cis-regulatory elements, such as promoters and enhancers, TFs must induce chromatin remodeling of nucleosomal structures to result in different levels of chromatin accessibility [
10]. Specifically, open chromatin regions allow transcription machinery components to access cis-regulatory elements and activate gene transcription, while closed chromatin regions impair the accessibility of promoters and enhancers to transcription factors and induce gene silencing [
11]. Up to now, chromatin accessibility has been investigated in various diseases, which largely facilitates the acquisition of gene expression [
12]. However, the role of chromatin remodeling in GHPA has not been investigated yet.
Assay for transposase-accessible chromatin with high-throughput sequencing (ATAC-seq) has been regarded as a robust and sensitive method for mapping genome-wide chromatin accessibility and TFs binding [
13]. This is because during the analysis of ATAC-seq, DNA accessibility is probed with a hyperactive Tn5 transposase, which allows to simultaneously fragment and integrate the sequencing adapters preferentially in functional regions with accessible chromatin [
13]. ATAC-seq has been widely adopted to investigate the pathophysiologic mechanisms of various tumors [
14,
15].
In this study, we firstly utilized both ATAC-seq and mRNA-seq to investigate the pattern of genome-wide chromatin accessibility and characterize the chromatin accessibility landscape in GHPA.
Material and methods
Collection and storage of tissue specimens
Patients with GHPA were recruited from the Department of Neurosurgery at Huashan Hospital, affiliated with Shanghai Medical School, Fudan University. The diagnosis of GHPA was based on clinical manifestations, endocrine laboratory tests, and imaging data. Primary pituitary adenoma tissues and adjacent normal pituitary specimens with no morphologic abnormalities were obtained at surgery. It was histologically confirmed with blinding to sample identity by at least two senior pathologists after surgical resection. Previous studies have demonstrated that the use of normal pituitary tissue obtained at surgery is the preferred control in order to avoid inconsistencies in the treatment of tissue in autopsy specimens [
16,
17]. In addition, the influence of heterogeneous factors such as age, gender and other diseases can also be avoided.
Totally, six GHPA tumor tissues and six corresponding para-adenoma normal tissues were obtained and immediately stored in liquid nitrogen. Owing to the small sample size, number of cells available for testing, especially for normal tissues, was limited. Thus, in ATAC-seq library preparation, we set the tumor as two aliquots, each one containing three samples, as well as for para-adenoma normal tissues. This study was approved by the Ethics and Research Committees of Huashan Hospital, Fudan University. Tissue samples and corresponding clinical materials were obtained with the written consent from all patients.
The frozen tissues were ground separately using a liquid nitrogen-cooled mortar and pestle. The pulverized powder was then transferred into 1.5 ml RNase free falcon tubes containing 1 ml TRIzol reagent (Invitrogen, USA). The homogenates were further purified using RNeasy Mini Kit (QIAGEN, Germany) and run on Agilent Bioanalyzer 2100 to asses RNA integrity. The mRNAseq library preparation from 1 μg of total RNA was performed with TruSeq RNA Sample Prep Kit (Illumina, USA) according to the manufacturer’s instructions. 2 × 150 paired-end sequencing was performed on Illumina HiSeq X Ten platform (Novogene Biotech, China).
Assay for transposase-accessible chromatin with high-throughput sequencing (ATAC-seq)
The ATAC-seq libraries were prepared as previously described [
18,
19]. Briefly, the frozen tissues were ground separately using a liquid nitrogen-cooled mortar and pestle. The pulverized powder was then transferred into 1.5 ml falcon tubes containing lysis buffer (10 mM Tris-HCl at pH7.4, 10 mM NaCl, 3 mM MgCl
2 and 0.1% IGEPAL CA-630 [Sigma-Aldrich]) for 10 min on ice to prepare the nuclei. Immediately after lysis, nuclei were spun at 500 ×
g for 5 min to remove the supernatant. Nuclei were then incubated with the Tn5 transposome (Vazyme Biotech, China) and tagmentation buffer at 37 °C for 30 min. After the tagmentation, the stop buffer was added directly into the reaction to end the tagmentation. PCR was performed to amplify the library for 10–12 cycles using the following PCR conditions: 72 °C for 3 min; 98 °C for 30 s; and thermocycling at 98 °C for 15 s, 60 °C for 30 s and 72 °C for 3 min; following by 72 °C 5 min. After the PCR reaction, libraries were purified with Agencourt Ampure XP beads (Beckman, USA) to remove contaminating primer dimers and quantitated using Qubit 2.0 Fluorimeter (Invitrogen, USA). All libraries were sequenced on the Illumina HiSeq X Ten with 150 bp paired-end reads (Novogene Biotech, China).
mRNA-seq data processing
The quality control of mRNA sequencing data was performed via FastQC (
http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Reads from RNA-seq experiments were aligned to human genome build hg19 using STAR [
20]. The mRNA expression of a gene was quantified by fragments per kilobase million reads (FPKM) based on RefSeq gene annotation using Cuffdiff [
21]. Differential expression genes (DEGs) were identified by a
p value < 0.05, fold change (FC) > 1.5 and the average FPKM > 5 for at least one of the two groups. The aligned BAM files were converted to bigWig format and normalized by reads per kilobase million (RPKM) with the bamCoverage script in deepTools 2.0 [
22]. The deepTools 2.0 and ggplot2 [
23] from R package were used for data normalization and visualization.
ATAC-seq data processing
The quality of sequenced ATAC-seq data was evaluated with FastQC (
http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). All sequencing data were mapped onto the hg19 human genome assembly using Bowtie2 [
24] with an insert size of less than or equal to 2000 bp (command line parameter -X 2000). Duplicate reads were removed with SAMtools [
25], only non-duplicate reads kept in the BAM format were used for the subsequent analysis. MACS2 [
26] was used to identify regions of ATAC-seq peaks (with the parameters --shift -100 --extsize 200). Picard Tools (v.2.2.4,
https://broadinstitute.github.io/picard/) and ggplot2 [
23] were used to plot the distribution of paired-end sequencing fragment sizes. For visualization, the BAM files were converted to the bigWig format using the bamCoverage [
22]. The heatmaps and average profiles were also generated using the plotHeatmap and plotProfile functions in deepTools.
GO and KEGG pathway enrichment analysis
The Gene Ontology (GO) enrichment and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis were performed using clusterProfiler [
27], which is an R package for functional classification and enrichment of gene clusters using hypergeometric distribution. The enriched terms with
p value< 0.05 were considered significant.
Annotation of genomic regions
To annotate the location of ATAC-seq signal in terms of important genomic features, the annotatePeaks.pl script in HOMER [
28] was used to assign the peak files in BED format to promoter-TSS (by default defined from −1 Kb to +100 bp of transcription start site), TTS (by default defined from −100 bp to +1 Kb of transcription termination site), Intron, Intergenic, Exon, etc. with default parameters.
Differential analysis of chromatin accessibility and mRNA expression
For purposes of assessing the alteration of chromatin accessibility genome-wide, deepTools was used to compute the average scores (RPKM normalized values) for transposase hypersensitive sites in each sample. Hyper-accessible regions (hyper) associated with increased ATAC-seq signal were defined if the regions showed fold change (FC) > 1.5 in the tumor group as compared to the normal group. On the contrary, the hypo-accessible regions associated with decreased ATAC-seq signal were defined if the regions showed fold change >1.5 in normal group compared to tumor group. The heatmaps and average profiles of regions expanded to ±1.5 Kb surrounding differentially accessible regions (DARs) center were generated using the plotHeatmap and plotProfile functions in deepTools [
22]. The annotation of DARs to genomic features was performed using the HOMER suite tool annotatePeaks [
28], then the DARs neighboring genes were overlapped with DEGs from mRNA-seq data, so as to address the global relationship between the ATAC-seq signal at differentially accessible regions and the transcription levels of the corresponding DEGs.
Transcription factor motif enrichment and occurrences analysis
The findMotifsGenome function in HOMER v4.9 [
28] were used to analyze transcription factor (TF) motif enrichment of ATAC-seq peak summits (with option -size -200,200), which was also used to predict binding sites and target genes of top 50 enriched TFs. The annotatePeaks.plscript was used to identify the nearest genes of predicted TF binding sites.
Statistical analysis and data visualization
If not specified, R platform (
https://www.R-project.org/) was used to compute statistics and generate plots. Fisher’s exact test, Student’s
t test, Pearson’s correlation coefficient, Wilcoxon signed rank test were used to assess the significance. Fold change,
p values and false discovery rate were calculated in analysis. Significant differences for all quantitative data were considered when
*p < 0.05,
**p < 0.01,
***p < 0.001, and
****p < 0.0001. Genome browser images were made using the Integrative Genomics Viewer (IGV) [
29] with bigWig files processed as described above.
Discussion
The genome-wide landscape of GHPA has been investigated in a few studies, but the variation of chromatin state in GHPA has never been investigated [
8,
36]. As chromatin structure change exerts significant influence on gene transcription, it has become the “hot spot” in the study of pathological mechanism of various diseases [
37,
38]. The commonly genome-wide chromatin accessibility profiling methodologies include MNase-seq, DNase-seq, FAIRE-seq and ATAC-seq, which can further analyze the alterations in the accessible chromatin state and understand the molecular mechanisms of several diseases [
11,
39]. Due to its speed, simplicity, and low input cell number requirement, ATAC-seq has been implicated as an accurate and sensitive measure of genome-wide chromatin accessibility [
40]. Moreover, ATAC-seq has also been widely adopted to predict TF binding motifs in these regulatory regions [
41].
Here, based on the RNA-seq, 64 significantly DEGs were obtained, including 10 upregulated DEGs and 54 down-regulated DEGs, in GHPA compared with normal pituitary tissues. It is worth noting that most up-regulated DEGs in tumors participate in various hormone-related signaling pathways (such as insulin, prolactin, and growth hormone) and pathways in GH synthesis, secretion and action. GH secretion by the somatotroph cells is mediated by ghrelin and GH-releasing hormone (GHRH) and inhibited by somatostatin and IGF-1 feedback inhibition [
42]. The effects of GH-releasing hormone are mediated by GHRH-receptor (GHRHR), a seven transmembrane G-protein coupled receptor which is expressed primarily on the cell surface of somatotrophs in the anterior pituitary [
43]. Under physical conditions, GHRH specifically binding to GHRHR facilitates somatotroph cell growth, growth hormone gene transcription and growth hormone secretion [
44]. Consistent with the previous result [
45], our data show that the expression of
GHRHR is significantly up-regulated in GHPA compared with normal pituitary tissues. Somatostatin binds to SSTR1-5 to inhibit GH secretion [
46]. Previous studies have reported that SSTRs expression is increased in tumor tissue of GHPA patients [
47]. Elevated expression of SSTRs has been considered as a valuable indicator in predicting biochemical response to somatostatin analogs treatment [
48]. Here, our data show the expression of SSTR1 is up-regulated in GHPA compared with normal pituitary tissue. There are few previous studies in investigating the level of SSTR1 in GHPA. However, our results suggest that additional research is needed to uncover this issue, which may provide potential target for GHPA treatment.
Additionally, our data reveal that the down-regulated DGEs are enriched in signaling pathways that included focal adhesion pathway. Focal adhesion, a subcellular structure between the cell and the extracellular matrix, plays an important role in cell migration through tissues [
49]. Previous studies have demonstrated that the invasion and metastasis of several malignant tumors, such as breast cancer and gastric cancer, are closely related to focal adhesion [
50,
51]. From RNA-seq data, we find genes which encode collagen family are significantly down-regulated. Collagen is one of the essential components which participate in the formation of focal adhesion [
52]. Suppression of collagen might breach focal adhesion and dramatically enhance the invasion of tumor cells [
53,
54]. Therefore, we speculate that decreased collagen in GHPA may interfere with the formation of focal adhesion, leading to potential invasion of tumor.
Further, ATAC-seq were undertook to reveal the noticeable alterations in transcriptomes and chromatin accessibility of GHPA. Comparing with normal pituitary tissues, there are 64 significantly DEGs identified in GHPA, the proportion of down-regulated DEGs is higher than that of up-regulated DEGs, which is corresponding to previous studies [
55,
56]. Combined with our ATAC-seq date, most of the DEGs near hypo-accessible regions are significantly down-regulated, which clearly suggests that decreased chromatin accessibility contributes to the down-regulation of genes. Furthermore, these hypo-accessible chromatin regions were rich in the motifs of critical regulatory factors, such as CTCF, Rfx2, and SF1. According to these data, we speculate that, by affecting the DNA-binding ability of TFs and transcriptional regulators in promoter and/or enhancer, the global decreases in chromatin accessibility could inhibit the expression of critical genes associated with GHPA. For example, it is probably that the significant decrease in chromatin openness leads to loss of chromatin insulator CTCF binding, which impairs the long-range interactions between promoters and enhancers of
FSHB.
In conclusion, our study is the first to depicture global changes in chromatin accessibility and transcription in GHPA. Combination of these data can provide relevant insight into the tumor biology, especially in regulatory network that contributes to tumor formation and growth. Our results demonstrated that the decreased chromatin-accessible regions are depleted for TF binding consensus motifs, which will lead the expression decrease in GHPA terminally. However, we are aware of the limitations of this research. Firstly, the sample size of GHPA and control group is small. Secondly, more informative is needed to validate the present findings in vitro and in vivo. The mechanism and functional consequence behind the interactions between TFs and chromatin accessibility change regions is worth further investigation.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.