Androgen receptor DNA binding and chromatin accessibility profiling in prostate cancer

Abstract

Prostate cancer (PCa) is the second most common cancer in men. The Androgen Receptor (AR) is the major driver of PCa and the main target of therapy in the advanced setting. AR is a nuclear receptor that binds the chromatin and regulates transcription of genes involved in cancer cell proliferation and survival. In a study by Stelloo et al. (1) we explored prostate cancer on the level of transcriptional regulation by means of Formaldehyde-Assisted Isolation of Regulatory Elements and Chromatin Immunoprecipitation coupled with massive parallel sequencing (FAIRE-seq and ChIP-seq, respectively). We employed these data for the assessment of differences in transcriptional regulation at distinct stages of PCa progression and to construct a prognostic gene expression classifier. Genomics data includes FAIRE-seq data from normal prostate tissue as well as primary, hormone therapy resistant and metastatic PCa. Furthermore, ChIP-seq data from primary and resistant PCa were generated, along with multiple input controls. The data are publicly available through NCBI GEO database with accession number GSE65478. Here we describe the genomics and clinical data in detail and provide comparative analysis of FAIRE-seq and ChIP-seq data.

Specifications
Organism/cell line/tissue	Homo Sapiens
Sex	Male
Sequencer or array type	Illumina Hiseq 2000 genome analyzer
Data format	Raw: SRA study; processed: BED
Experimental factors	Normal, primary and therapy resistant tumors, lymph node metastases
Experimental features	FAIRE-seq and Androgen Receptor ChIP-seq
Consent	Leftover anonymized tissue (not traceable back to the patient and not interfering with care and/or prognosis) used for research purposes.
Sample source location	Samples were from prostate cancer patients, treated at the Erasmus University Medical Center (EMC; Rotterdam, The Netherlands), The Netherlands Cancer Institute/Antoni van Leeuwenhoek hospital (Amsterdam, The Netherlands)

Direct link to deposited data https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE65478

1. Experimental design, materials and methods

1.1. Clinical samples and experimental design

Fresh frozen tissue samples were obtained through postoperative needle biopsies targeting both tumor and normal areas of prostatectomy specimens at The Netherlands Cancer Institute (Amsterdam, The Netherlands). Tissue samples from androgen deprivation resistant tumors (from transurethral resection of the prostate (TURP)) and lymph node metastases were obtained from the Erasmus University Medical Center (Rotterdam, The Netherlands). Slides stained with hematoxylin and eosin (H&E) of the cases were reviewed by our pathologists. Clinical and pathological parameters of the selected patients are provided in Table 1. Leftover anonymized tissue, which cannot be traced back to the patient and does not interfere with care and/or prognosis, and would have been discarded otherwise, has been used in accordance with the Code of Conduct of the Federation of Medical Scientific Societies in The Netherlands. NKI and Erasmus MC institutional medical ethics committees have approved the study.

Table 1.

Patient and tumor characteristics of the selected samples.

Characteristic	Number of patients
Normal	Primary	Resistant	Metastasis
4	4	4	3
Treatment type
Untreated	3	4	0	2
Bicalutamide/cyproteron acetate	1	0	0	0
Bicalutamide/LHRH analogue	0	0	1	0
Cyproteron acetate + LHRH analogue	0	0	1	0
LHRH analogue	0	0	2	0
LHRH analogue/Cyproteron Acetate	0	0	0	1
Gleason score
6	1	0	0	0
7	2	2	0	0
8	0	0	1	1
9	1	2	0	0
10	0	0	3	2
Initial PSA (ng/ml)
Mean	8.7	19.6	149.5	135.5
Range	5.3–13.0	8.5–38.0	6.5–511.0	17.0–254.0

FAIRE-seq was performed on four normal samples, four primary, three therapy resistant tumors and three lymph node metastases (Fig. 1). Androgen Receptor ChIP-seq was carried out on four primary and three resistant tumors (Fig. 1).

Fig. 1.

FAIRE-seq and ChIP-seq analyses were performed on normal prostate tissue and prostate cancer samples from different stages of the disease.

1.2. Formaldehyde-assisted isolation of regulatory elements (FAIRE)

FAIRE was performed as previously described [2]. Briefly, fresh frozen tissues were cross-linked with 1% formaldehyde for 20 min. After washing, nuclei were isolated as described before [3]. Afterwards chromatin was sonicated, cleared by centrifugation and subjected to three consecutive phenol–chloroform–isoamyl alcohol (25:24:1) extractions. Reverse cross-linking was performed at 65 °C overnight. Subsequently, samples were treated with RNase A and proteinase K and purified by using a PCR purification kit (Roche).

1.3. Chromatin immunoprecipitation (ChIP)

Chromatin immunoprecipitation was carried out as described before [3], [4]. 10 μg of AR-N20 (sc-618; Santa Cruz) antibody was used for immunoprecipitation, with 100 μl of Protein A magnetic beads (Invitrogen).

1.4. DNA sequencing

Libraries were prepared according to Illumina DNA Sample Kit instructions. Sequencing was performed on the Illumina HiSeq 2000 Genome Analyzer using 51-bp reads. Reads were aligned to the Human Reference Genome (assembly hg19, February 2009) using bwa 0.5.9.

1.5. Data analysis

Reads that map uniquely to the genome, with MAPQ quality score above 20, were used for the analysis. FAIRE-seq and ChIP-seq peaks were called with two algorithms, MACS 1.4 [5] and DFilter 1.0 [6], against mixed input controls corresponding to each group. MACS was run with default parameters, except for p = 10^− 7 for ChIP-seq data. DFilter was run with bs = 100, ks = 50 for FAIRE-seq data and bs = 50, ks = 30, refine, nonzero for ChIP-seq data. Peaks detected by both algorithms were used for further analysis. Sequencing read depths and number of called peaks can be found in Table 2.

Table 2.

Sequencing and peak calling details.

GEO accession	Experiment	Tissue	Total number of reads	Mapped reads	% mapped reads	No. peaks
GSM1598204	FAIRE-seq	Normal	19,147,127	17,986,187	93.94	50
GSM1598205	FAIRE-seq	Normal	21,599,945	19,883,501	92.05	472
GSM1598206	FAIRE-seq	Normal	26,080,719	25,043,481	96.02	61
GSM1598207	FAIRE-seq	Normal	23,167,347	22,177,458	95.73	2837
GSM1598208	FAIRE-seq	Primary	36,827,373	34,441,896	93.52	6450
GSM1598209	FAIRE-seq	Primary	18,306,926	17,002,416	92.87	1579
GSM1598210	FAIRE-seq	Primary	32,197,589	30,568,523	94.94	13,348
GSM1598211	FAIRE-seq	Primary	28,992,853	27,590,961	95.16	2243
GSM1598212	FAIRE-seq	Resistant	37,452,682	35,655,681	95.2	80
GSM1598213	FAIRE-seq	Resistant	28,372,546	26,836,918	94.59	3497
GSM1598214	FAIRE-seq	Resistant	27,545,618	26,061,843	94.61	5754
GSM1598215	FAIRE-seq	Metastasis	39,562,972	37,594,752	95.03	2043
GSM1598216	FAIRE-seq	Metastasis	29,130,845	27,291,106	93.68	281
GSM1598217	FAIRE-seq	Metastasis	27,253,810	25,789,354	94.63	1313
GSM1598218	AR ChIP-seq	Primary	13,782,549	12,232,556	88.75	754
GSM1598219	AR ChIP-seq	Primary	18,146,927	16,009,388	88.22	402
GSM1598220	AR ChIP-seq	Primary	13,040,014	11,254,994	86.31	17,511
GSM1598221	AR ChIP-seq	Primary	9,928,626	7,080,840	71.32	3278
GSM1598222	AR ChIP-seq	Primary	12,243,485	11,160,623	91.16	7932
GSM1598223	AR ChIP-seq	Resistant	16,518,987	14,727,645	89.16	739
GSM1598224	AR ChIP-seq	Resistant	16,382,421	14,441,817	88.15	238
GSM1598225	AR ChIP-seq	Resistant	15,621,538	13,967,477	89.41	1779
GSM1598226	Input	Resistant	28,171,838	26,825,849	95.22
GSM1598227	Input	Metastasis	24,117,145	22,902,755	94.96
GSM1598228	Input	Primary	23,982,305	22,739,491	94.82
GSM1598229	Input	Primary	27,642,177	26,387,234	95.46

FAIRE-seq, ChIP-seq data and clinical annotation of the samples that are deposited in NCBI GEO under accession number GSE65478.

For further analysis, a merged list of peaks present in all samples from each technique was generated. The number of peaks detected by FAIRE-seq was 25,797, while 20,703 peaks were detected by ChIP-seq. The AR binding sites had a median width of 350 bp and peak size did not vary strongly with the largest peak size of 1202 bp (Fig. 2A). In contrast, FAIRE-seq peaks had a larger spread in size with a median size of 255 bp. The largest peak size of FAIRE-seq data was 2300 bp and a higher proportion of both small and large peaks was present (Fig. 2A). The distance to the nearest transcription start site (TSS) was determined by the GREAT tool (http://great.stanford.edu/) [7]. The number of peaks within 5 kb from the nearest TSS was significantly higher in FAIRE-seq data as compared to ChIP-seq data and the number of peaks further than 50 kb from a TSS was higher in ChIP-seq data than in FAIRE-seq (p < 10^− 15 Fisher's exact; Fig. 2B-C). This is in accordance with AR binding mainly distant enhancer elements [8], while accessible regions detected by FAIRE-seq include not only enhancers, but also promoters [9].

Fig. 2.

Comparative analysis of FAIRE-seq and Androgen Receptor ChIP-seq data. (A) Size distribution of peaks detected by FAIRE-seq and ChIP-seq in prostate cancer specimens. Pie charts showing the percentage of peaks in categories based on the distance to the nearest transcription start site (TSS) in FAIRE-seq (B) and ChIP-seq (C) data.

2. Conclusions

In conclusion, we provide a unique dataset of genome-wide epigenetic profiling of prostate cancer tissue from different stages of the disease. The dataset consists of two parts: accessible chromatin profiling by FAIRE-seq and genome-wide androgen receptor binding to DNA by ChIP-seq. We previously used this dataset to identify changes in transcriptional regulation in prostate cancer upon acquisition of resistance to hormonal therapy, as well as to derive a prognostic gene expression signature for prostate cancer [1].