Integrative molecular characterization of Chinese prostate cancer specimens

Abstract

Prostate cancer (PCa) exhibits epidemiological and molecular heterogeneity. Despite extensive studies of its phenotypic and genetic properties in Western populations, its molecular basis is not clear in Chinese patients. To determine critical molecular characteristics and explore correlations between genomic markers and clinical parameters in Chinese populations, we applied an integrative genetic/transcriptomic assay that combines targeted next-generation sequencing and quantitative real-time PCR (qRT-PCR) on samples from 46 Chinese patients with PCa. Lysine (K)-specific methyltransferase 2D (KMT2D), zinc finger homeobox 3 (ZFHX3), A-kinase anchoring protein 9 (AKAP9), and GLI family zinc finger 1 (GLI1) were frequently mutated in our cohort. Moreover, a clinicopathological analysis showed that RB transcriptional corepressor 1 (RB1) deletion was common in patients with a high risk of disease progression. Remarkably, four genomic events, MYC proto-oncogene (MYC) amplification, RB1 deletion, APC regulator of WNT signaling pathway (APC) mutation or deletion, and cyclin-dependent kinase 12 (CDK12) mutation, were correlated with poor disease-free survival. In addition, a close link between KMT2D expression and the androgen receptor (AR) signaling pathway was observed both in our cohort and in The Cancer Genome Atlas Prostate Adenocarcinoma (TCGA-PRAD) data. In summary, our results demonstrate the feasibility and benefits of integrative molecular characterization of PCa samples in disease pathology research and personalized medicine.

INTRODUCTION

Prostate cancer (PCa) is the second most frequent malignancy in men worldwide, accounting for more than 300 000 cancer-related deaths per year. In China, approximately 6.03% of adult males were suspected to have PCa in 2015, and the annual incidence of PCa in the Chinese population is increasing by 4.7%.1 Actionable molecular sequencing is a plausible and promising strategy for improving treatment results by targeted therapy. However, the incidence, prognosis, and treatment outcomes differ among ethnicities due to heterogeneity at the genetic and phenotypic levels.2,3 Genetic and transcriptional alterations have been studied extensively in Western populations; however, little is known about the genomic aberrations associated with the pathogenesis of PCa in Chinese populations.

Next-generation sequencing (NGS) has enabled extensive profiling of genome-wide alterations during PCa tumorigenesis. Recent sequencing studies have revealed various somatically acquired genetic, epigenetic, and transcriptional changes in PCa, such as alterations in phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA), tumor protein P53(TP53), speckle type BTB/POZ protein (SPOP), chromodomain helicase DNA-binding protein 1 (CHD1), and external transcribed spacer (ETS) transcription factor family.4 Notably, NGS approaches not only provide numerous insights into the molecular-based pathogenesis but also identify specific biomarkers for better patient stratification and targeted therapy. The hyperactivation of oncogenic signaling pathways, such as phosphatidylinositol 3-kinase (PI3K)/phosphatase and tensin homolog (PTEN)/AKT, is considered a prognostic biomarker for PCa and is closely correlated with advanced localized or metastatic disease.5 SPOP, CHD1, andETS transcription factor (ERG) statuses can be used for segregation, as they suggest future androgen receptor (AR) loss and possibly aggressive cancer development.6 DNA repair pathway defects indicate that patients would benefit from treatment with a poly(ADP)-ribose polymerase (PARP) inhibitor that has been approved by the US Food and Drug Administration for metastatic castration-resistant PCa monotherapy.7 Hence, discovering and validating genomic aberrations is urgently needed for PCa prediction, prognosis, and treatment in Chinese patients.

In this study, we performed a comprehensive molecular characterization of Chinese patients with PCa and explored correlations between identified genomic markers and clinical factors. We used quantitative real-time PCR (qRT-PCR) to detect the expression of focal PCa-related genes in 46 Chinese PCa samples. Moreover, we reanalyzed previously generated gene sequencing data8 and detected copy number variants (CNVs) by identifying clustered amplicons with significant changes in normalized log₂ read depth.9 We provide an integrative overview of genetic/transcriptomic alterations in Chinese patients with PCa; these findings will improve our understanding of disease biology and have implications for personalized medicine.

PATIENTS AND METHODS

Patients and samples

The study included a total of 46 PCa samples collected at Nanfang Hospital (Guangzhou, China). All samples were histologically confirmed as PCa by two independent pathologists. Patients who received preoperative hormone therapy, chemotherapy, and radiotherapy were excluded. Detailed information for all samples is summarized in Supplementary Table 1. This study was approved by the Medical Ethics Committee of Nanfang Hospital (NFEC-2017-145), and the specimens were used with written informed consent by the Department of Pathology, Nanfang Hospital, in accordance with the Declaration of Helsinki.

Supplementary Table 1.

Primer sequences used in our quantitative real-time polymerase chain reaction analysis

Gene	Type	Forward primer	Reverse primer
AR	AR pathway	CCAGATCAGGGTTGAAGAGAAA	ATATGCTGGGTGACAAGAAGAG
KLK2	AR pathway	GATAAGGCCGTGAGCAGAAA	GACCTGAACAAACCTCCTGTAA
KLK3	AR pathway	GCTGGGAACTGCTATCTGTTAT	AGCAGGGAGAGAGTGAGATAG
CAMKK2	AR pathway	CCCTAGACTCCACACAATAACC	AAGCCCTGGTTTCCTCATAC
FKBP5	AR pathway	GTCTCCCACGTGTGTATTATGG	AATGGGCACCCTGTAGTTATTT
ABCC4	AR pathway	GGAACTCCCACACTAAGGAATC	CTCTCCAGAGCACCATCTTTC
STK39	AR pathway	CAGTTGAGTGTCAGCTGATGTA	GGAGGAAATGGGCAGAAAGA
ZBTB10	AR pathway	GACATGGATCTCGACGTTATGG	CATCCCGAGGGAATTCTGTATC
NUSAP1	Cell cycle	CCTCTTGTGATGAGACTGAGATAC	CAGTCTTGCACCTTCTCCTT
KIF11	Cell cycle	CCGTTCTGGAGCTGTTGATAA	TGTTCTTTCTACAAGGGCAGTAA
CDC20	Cell cycle	TTTGGCCAGTGGTGGTAATG	CCTTGATGCTGGGTGAATGT
FOXM1	Cell cycle	GGGAACAACAAAGGCAATGG	GCAAAGATCAGGGAAGGTAAGA
TMPRSS2: ERG	Subtype	CTGGAGCGCGGCAGGAA	GTAGGCACACTCAAACAACGACTGG
ERG	Subtype	GAAGACTGGACTCAGGACATTT	GCTGTGTCCTTTCTCCTAACA
ETV1	Subtype	GGTGGGTGGGAGTAATCTAAAC	CCCATCCCAAGCCTAAGTAAA
ETV4	Subtype	GAAGAAAGGAACCTGGGATGAG	CCTAGGAAAGGGCAGAAGAAAG
FLI1	Subtype	CATCTCCTACATGCCTTCCTAC	TGATGCGGCTCCAAAGAA
CHD1	Subtype	TGGCAAAGGGAGAGAATATGG	CAACCTTGGCAGGAAGAGAT
SPINK1	PCa main alterations	GCTTCTGAAGAGACGTGGTAAG	CCAAGGCACTGAGAAGAAAGA
RAF1	PCa main alterations	CTCGTGGACAGAGAGATTCAAG	CTCCGTGCCATTTACCCTTAT
PTEN	PCa main alterations	CTGCCAGCTAAAGGTGAAGATA	ATCACCACACACAGGTAACG
PIK3CA	PCa main alterations	CTGGTTCAGCAGTGTGGTAA	CCTGCGTGGGAATAGCTAAA
RB1	PCa main alterations	CTGTCTGAGCACCCAGAATTAG	GTCCAAATGCCTGTCTCTCAT
MYC	PCa main alterations	CATACATCCTGTCCGTCCAAG	GAGTTCCGTAGCTGTTCAAGT
NKX3-1	PCa main alterations	CGGAGACCCAAGTGAAGATATG	CAAAGAGGAGTGCTTCTCCAA
EZH2	PCa main alterations	CTCGGTGTCAAACACCAATAAAG	AGTGCCAATGAGGACTCTAAAC
KMT2D	PCa main alterations	ACCATGTGAAGAACAGGAAGAG	TCACCCTGGCTCAGATTAGA
GAPDH	House keeping	CTCCTCACAGTTGCCATGTA	GTTGAGCACAGGGTACTTTATTG

AR: androgen receptor; PTEN: phosphatase and tensin homolog

DNA extraction, amplicon capture, next-generation sequencing, and analysis

Samples for targeted NGS were dissected from 4-mm unstained sections, and genomic DNA was extracted using a GeneRead DNA FFPE Tissue Kit (Qiagen, Hilden, Germany) according to the manufacturer's protocol. The extracted DNA was quantified using a Qubit fluorometer (Invitrogen, Carlsbad, CA, USA). Genomic DNA with a yield of >40 ng was used for library preparation. Amplicon enrichment was performed using the prostate cancer panel within the GeneRead DNAseq Targeted Panels V2 (Qiagen). Library preparation was performed using the GeneRead DNA Library Core Kit for Ion Torrent Proton (Qiagen) and barcoded with Ion Xpress™ Barcode Adapters (Thermo Fisher, Waltham, MA, USA). Sequencing of barcoded pools was performed using the Ion Torrent Proton, and the data were evaluated using Cloud-Based DNAseq Sequence Variant Analysis (Qiagen). Called variants were filtered by Ion Reporter (Thermo Fisher) to remove synonymous or noncoding variants and University of California, Santa Cruz (UCSC) common single-nucleotide polymorphisms (SNPs). High-confidence candidate somatic drivers were included in subsequent analyses.

For CNV determination, germline DNA (extracted from peripheral blood mononuclear cells) was collected from seven independent healthy males and sequenced to obtain the CNV baseline. CNV calling was performed using Cloud-Based DNAseq Sequence Variant Analysis. For targeted genes, CNVs were defined as clustered amplicons with significant changes in normalized log₂ read depth, as previously reported.9 In particular, CNVs were evaluated as follows. 1) Amplicons were grouped into clusters of neighbors at a genomic locus (amplicon clustering). 2) The log₂ read depth at each priming site was compared between sample and control datasets (filtered clustering). P-values were calculated by weighted t-tests (total read depth was used as the weight). 3) Only CNV calls with scores of >50 indicated strong evidence (CNV identification).

RNA extraction, reverse transcription, and PCR analysis

The same tissues for NGS were used to isolate RNA using the AllPrep DNA/RNA FFPE Kit (Qiagen) according to the manufacturer's protocol. The extracted RNA was quantified using a Qubit fluorometer (Thermo Fisher). Reverse transcription (RT) of 1 μg of RNA was performed using the QuantiTect Reverse Transcription Kit (Qiagen). The RT product was mixed with nuclease-free water, primers, and the SYBR Premix Ex Taq II Kit (Takara, Shiga, Japan). Quality PCR was performed using the CFX384 Touch Real-Time PCR system (Bio-Rad, Hercules, CA, USA). Data were analyzed using the 2^−ΔΔCt method, and normalized target gene expression was visualized using R statistical software (GraphPad Software, La Jolla, CA, USA).

Statistical analyses

Kaplan–Meier analysis was performed using R statistical software. Disease-free survival (DFS) was defined as the time from diagnosis to recurrence or date of last follow-up. The log-rank test was used to calculate the P- values or the Kaplan–Meier analysis.

A large prostate cancer dataset from the Cancer Genome Atlas (TCGA), with both gene expression and clinical follow-up data, was used for the correlation analysis. Level 3 gene expression data were downloaded from TCGA data portal (https://tcga-data.nci.nih.gov/tcga/). RNA-seq by Expectation-Maximization (RSEM) normalized expression data were obtained.

Correlations between the recurrently altered genes and clinical parameters were evaluated using Fisher's tests; gene expression correlations were evaluated based on Pearson's correlation coefficients, as determined using R statistical software.

RESULTS

Application of integrative genetic/transcriptomic assays for Chinese clinical formalin-fixed paraffin-embedded (FFPE) PCa samples

To identify the genetic and transcriptomic spectra of genes associated with the pathogenesis of PCa in Chinese patients, we performed integrative DNA- and RNA-based assays requiring only approximately 50 ng of FFPE DNA and 500 ng of FFPE RNA. The assays consist of a multiplex PCR-based panel for gene sequencing and a qRT-PCR array for the detection of gene expression. The gene sequencing panel was used for target enrichment of the exonic regions of the 32 most commonly mutated genes in human PCa samples, and the amplification or deletion of these genes was evaluated as described previously.9 The qRT-PCR assay included 28 genes, including robust housekeeping genes, AR transcriptional modules, cell cycle-related genes, and genes defining molecular subtypes or molecular drivers based on previous transcriptomic studies (Supplementary Table 1). The integrative assays were performed using samples obtained from a cohort of 46 Chinese patients with PCa after radical prostatectomy or transurethral resection of the prostate (Supplementary Table 2).

Supplementary Table 2.

46 Chinese prostate cancer samples assessed by target gene sequence and quantitative real-time polymerase chain reaction

Sample ID	Target NGS	RT-qPCR	Age	Site	Percentage tumor	Gleason score	TNM stage	tPSA	Disease free status	NCCN risk stratification	EAU risk classification
PR101	Yes	Yes	68	RP	30	8	T3aN0Mx	158.69	Disease free	High	Locally advanced
PR102	Yes	Yes	73	RP	50	6	T3aN0Mx	>50	Recurred	High	Locally advanced
PR103	Yes	Yes	66	RP	50	7	T2aN0M0	16.9	Disease free	Favorable intermediate	Intermediate risk
PR104	Yes	Yes	61	RP	30	8	T2aN0M0	7.027	Disease free	High	High risk
PR105	Yes	Yes	49	RP	70	7	T4N0Mx	>100	Disease free	Very high	Locally advanced
PR106	Yes	Yes	68	RP	50	7	T3aN0Mx	>100	Recurred	High	Locally advanced
PR107	Yes	Yes	71	RP	30	6	T2aN0M0	9.3	Disease free	Low	Low risk
PR108	Yes	Yes	67	RP	60	7	T2aN0M0	9.85	Disease free	Favorable intermediate	Intermediate risk
PR109	Yes	Yes	67	RP	70	4	T2aN0Mx	3.143	Disease free	Low	Low risk
PR110	Yes	No*	67	RP	80	10	T4N0Mx	21.523	Recurred	Very high	Locally advanced
PR111	Yes	Yes	68	RP	70	7	T4N0Mx	9.101	Recurred	Very high	Locally advanced
PR112	Yes	Yes	71	RP	50	9	T3bNxM1	17.95	Recurred	Metastatic	Metastatic
PR113	Yes	No*	66	RP	80	7	T4N0Mx	9.66	Recurred	Very high	Locally advanced
PR114	Yes	Yes	67	RP	70	7	T2bNxM0	43.57	Disease free	High	High risk
PR115	Yes	Yes	63	RP	90	4+3+5	T2cN0M0	20.37	Disease free	High	High risk
PR201	Yes	Yes	73	RP	70	7	T2bN0M0	18.58	Disease free	Unfavorable intermediate	Intermediate risk
PR202	Yes	Yes	54	RP	80	8	T4N0Mx	114.93	Recurred	Very high	Locally advanced
PR203	Yes	Yes	68	RP	90	7	T2cN0M0	8.76	Disease free	Unfavorable intermediate	High risk
PR205	Yes	Yes	65	RP	50	8	T2cN0M0	13.86	Disease free	High	High risk
PR206	Yes	Yes	73	RP	40	8	T3bN0M0	35.16	Recurred	Very high	Locally advanced
PR207	Yes	Yes	62	RP	40	7	T2cN0M0	6.41	Recurred	Unfavorable intermediate	High risk
PR208	Yes	Yes	49	RP	80	6	T2cN0M0	123.12	Recurred	High	High risk
PR210	Yes	Yes	73	RP	60	8	T2bN0M0	22.9	Recurred	High	High risk
PR212	Yes	Yes	60	RP	60	7	T2cN0M0	43.03	Disease free	High	High risk
PR213	Yes	Yes	69	RP	70	10	T3aN0M0	3.79	Disease free	Very high	Locally advanced
PR214	Yes	Yes	57	RP	70	7	T2bN0M0	12	Disease free	Favorable intermediate	Intermediate risk
PR215	Yes	Yes	65	RP	60	6	T2cN0M0	8.11	Disease free	Favorable intermediate	High risk
PR302	Yes	Yes	60	RP	70	6	T2aN0M0	10.88	Disease free	Favorable intermediate	Intermediate risk
PR304	Yes	Yes	75	RP	70	7	T2bN0M0	4.65	Disease free	Unfavorable intermediate	Intermediate risk
PR305	Yes	Yes	80	RP	70	6	T2bN0M0	7.01	Disease free	Favorable intermediate	Intermediate risk
PR306	Yes	Yes	66	RP	90	7	T3aN0M0	98.16	Disease free	High	Locally advanced
PR307	Yes	Yes	67	RP	50	5	T2bN0M0	5.81	Disease free	Favorable intermediate	Intermediate risk
PR308	Yes	Yes	56	RP	70	7	T2aN0M0	13.7	Disease free	Favorable intermediate	Intermediate risk
PR309	Yes	No*	59	RP	60	7	T2N1M1	7.69	Recurred	Metastatic	Metastatic
PR310	Yes	Yes	63	RP	60	9	T2cN0M0	9.22	Recurred	High	High risk
PR311	Yes	Yes	63	RP	95	8	T2cN1M0	67.61	Disease free	Regional	Locally advanced
PR312	Yes	Yes	52	RP	40	7	T2cN0M0	19.22	Recurred	Unfavorable intermediate	High risk
PR313	Yes	Yes	77	TURP	65	7	T4N1Mx	157.86	Disease free	Regional	Locally advanced
PR315	Yes	Yes	77	TURP	95	8	T2cNxM1	>500	Disease free	Metastatic	Metastatic
PR401	Yes	Yes	67	RP	60	6	T2aN0M0	9.27	Disease free	Low	Low risk
PR403	Yes	Yes	49	RP	70	6	T2bN0M0	9.21	Disease free	High	High risk
PR404	Yes	Yes	44	RP	70	7	T2bN0M0	24.4	Disease free	High	High risk
PR414	Yes	No	72	RP	40	8	T2cN0M0	91.48	Recurred	Favorable intermediate	Intermediate risk
PRY01	Yes	Yes	59	RP	30	7	T2N0M0	14.082	Disease free	Favorable intermediate	Intermediate risk
PRY02	Yes	Yes	63	RP	30	6	T2bN0M0	12.18	Disease free	Favorable intermediate	Intermediate risk
PRY04	Yes	Yes	72	RP	30	7	T2cN0M0	20.2	Recurred	High	High risk

*The samples didn’t perform gene expression analysis by qRT-PCR due to low RNA quality. Specimen type: RRP: radical retropubic prostatectomy; TURP: transurethral resection of the prostate. qRT-PCR: quantitative real-time polymerase chain reaction; NGS: next-generation sequencing; TNM: Tumor-node-metastasis; tPSA: total PSA; RP: radical prostatectomy; NCCN: National Comprehensive Cancer Network; EAU: European Association of Urology

Mutation and copy number variant detection by targeted gene sequencing

Ion Torrent NGS after multiplex PCR-based targeted enrichment was applied to 46 FFPE PCa samples, yielding an average of 6 243 762 mapped reads and an average coverage of 3527× (1156–5419×) over targeted bases per sample. The complete coverage details are provided in Supplementary Table 3. An initial 5909 variants were filtered for quality, SNPs, and functional relevance. As described previously,8 we identified 245 nonsynonymous (missense, in-frame deletion, nonsense, stop-loss, and frameshift) mutations with an allele frequency of ≥3% in all sequencing specimens (average: 5; range: 1–34) across 27 genes. All high-confidence mutations are shown in a heatmap in Figure 1 and are detailed in Supplementary Table 4 ^{(178.8KB, pdf)} . The most frequently mutated genes were lysine (K)-specific methyltransferase 2D (KMT2D, 63.04%), zinc finger homeobox 3 (ZFHX3, 50.00%), A-kinase anchoring protein 9 (AKAP9, 32.61%), GLI family zinc finger 1 (GLI1, 32.61%), thrombospondin type 1 domain containing 7B (THSD7B, 19.57%), APC regulator of WNT signaling pathway (APC, 15.22%), cyclin-dependent kinase 12 (CDK12, 15.22%), lysine demethylase 4B (KDM4B, 15.22%), mediator complex subunit 12 (MED12, 15.22%), and zinc finger protein 595 (ZNF595, 15.22%).

Supplementary Table 3.

Sequencing statistics for Chinese prostate cancer samples assessed on target gene sequence

Sample ID	Total reads	Read depth	Percentage of bases covered at ³100×	SNPs/MNPs	Insertions/deletions
PR101	5068352	2830	98	98	6
PR102	5947960	3373	98	104	4
PR103	5286208	3106	98	105	6
PR104	6744522	3889	98	95	8
PR105	6929606	4076	99	109	6
PR106	5894989	3467	98	105	6
PR107	5911356	3381	98	112	9
PR108	5918910	3321	98	117	11
PR109	6543896	3709	98	104	5
PR110	5653814	3064	96	111	7
PR111	6314382	3550	98	114	10
PR112	4459067	2496	98	111	9
PR113	6677205	3636	98	150	11
PR114	5470762	3094	98	100	7
PR115	9498650	5419	99	126	8
PR201	3145871	1584	92	176	9
PR202	2801865	1572	95	129	9
PR203	2633500	1488	98	114	7
PR205	9233792	4882	96	153	11
PR206	7980130	4619	98	121	10
PR207	7660193	4497	99	141	11
PR208	8228356	4574	98	112	9
PR210	7929129	4429	98	121	11
PR212	7932509	4595	98	123	11
PR213	7707601	4476	98	121	14
PR214	6895340	4048	98	117	7
PR215	8254890	4630	97	107	8
PR302	5687529	3129	96	113	7
PR304	7384117	4227	98	134	8
PR305	5473565	3106	98	127	9
PR306	6107461	3432	97	105	7
PR307	6090712	3486	98	139	5
PR308	6220821	3516	98	102	5
PR309	7586732	4021	95	206	8
PR310	4196760	2343	97	116	5
PR311	6496237	3710	98	119	9
PR312	6120013	3559	98	126	6
PR313	6981974	3865	97	137	8
PR315	5432214	3133	98	102	5
PR401	5855582	3217	95	121	7
PR403	5291636	2727	98	111	10
PR404	9418627	5355	98	130	10
PR414	7055318	4060	99	106	6
PRY01	8761095	5121	99	146	13
PRY02	1989740	1156	98	104	10
PRY04	2340079	1305	97	94	7

SNPs: single-nucleotide polymorphisms; MNPs: multiple-nucleotide polymorphisms

Figure 1.

Integrative molecular profiling of Chinese FFPE prostate cancer (PCa) samples. A heatmap of somatic alterations for each sample is shown. Clinicopathological information is provided in the header, including Gleason score, T stage, tumor percentage, and molecular subtype based on qRT-PCR analysis. High-level somatic CNVs and mutations are indicated according to the legend. FFPE: formalin-fixed paraffin-embedded; qRT-PCR: quantitative real-time polymerase chain reaction; CNVs: copy number variants. All abbreviated names of genes are defined in Supplementary Table 6.

Using an approach adapted from multiplexed PCR-based targeted Ion Torrent NGS data,9 we simultaneously assessed CNVs in 25 genes, yielding a total of 312 high-level CNVs across the 46 PCa specimens (average: 7 high-level CNVs per sample; range: 0–15). A copy number heatmap for all samples is shown in Figure 1.

Assessment of critical PCa transcriptomic alterations by qRT-PCR

We performed qRT-PCR using RNA isolated from the sequencing cohort of 43 FFPE PCa samples (the sequencing samples, excluding those with low RNA quality). Normalized target gene expression is shown in Figure 2. Combined with target sequencing, we performed basic PCa molecular subtyping for the 43 Chinese patients (Figure 1 and 2). TMPRSS2:ERG was detected in nine cancer samples, which showed approximately 70-fold greater TMPRSS2:ERG expression than the median of the 34 other TMPRSS2:ERG-negative samples. Two samples showed marked ETS variant 4 (ETV4) overexpression, indicating the presence of a gene fusion in this ETS family member. Overexpression of serine peptidase inhibitor, Kazal type 1 (SPINK1), was identified in four samples (the SPINK1⁺ subtype). Moreover, by targeted gene sequencing, we observed high-confidence SPOP mutations in samples PR110, PR112, and PR203, each of which was ETS-negative by qRT-PCR. Among these samples, PR203 also showed high SPINK1 expression, suggesting the overlap of these alterations.

Figure 2.

Gene expression signatures of PCa assessed by qRT-PCR. Unsupervised hierarchical clustering of gene expression signatures for PCa-related genes. Normalized target gene expression (log₂) for 28 robust target genes assessed by the qRT-PCR panel. qRT-PCR: quantitative real-time polymerase chain reaction; PCa: prostate cancer. All abbreviated names of genes are defined in Supplementary Table 6.

Supplementary Table 6.

The official full name of genes listed in the figures

Gene abbreviation	Official full name
KMT2D	Lysine methyltransferase 2D
ZFHX3	Zinc finger homeobox 3
AKAP9	A-kinase anchoring protein 9
GLI1	GLI family zinc finger 1
THSD7B	Thrombospondin type 1 domain containing 7B
APC	APC regulator of WNT signaling pathway
CDK12	Cyclin-dependent kinase 12
KDM4B	Lysine demethylase 4B
MED12	Mediator complex subunit 12
ZNF595	Zinc finger protein 595
PIK3CA	Phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha
ZNF473	Hypoxia inducible factor 1 subunit alpha
NCOA2	Nuclear receptor coactivator 2
PDZRN3	PDZ domain containing ring finger 3
SCN11A	Sodium voltage-gated channel alpha subunit 11
SYNE3	Spectrin repeat containing nuclear envelope family member 3
IKZF4	IKAROS family zinc finger 4
SPOP	Speckle type BTB/POZ protein
TP53	Tumor protein p53
AR	Androgen receptor
NKX3-1	NK3 homeobox 1
NRCAM	Neuronal cell adhesion molecule
PTEN	Phosphatase and tensin homolog
TFG	TRK-fused gene
CDKN2A	Cyclin-dependent kinase inhibitor 2A
NIPA2	NIPA magnesium transporter 2
RB1	RB transcriptional corepressor 1
OR5L1	Olfactory receptor family 5 subfamily L member 1
MYC	MYC proto-oncogene, bHLH transcription factor
TBX20	T-box 20
KLK2	Kallikrein-related peptidase 2
KLK3	Kallikrein-related peptidase 3
CAMKK2	Calcium/calmodulin-dependent protein kinase kinase 2
FKBP5	FKBP prolyl isomerase 5
ABCC4	ATP-binding cassette subfamily C member 4
STK39	Serine/threonine kinase 39
ZBTB10	Zinc finger and BTB domain containing 10
NUSAP1	Nucleolar and spindle-associated protein 1
KIF11	KIF11
CDC20	Cell division cycle 20
FOXM1	Forkhead box M1
TMPRSS2:ERG	transmembrane serine protease 2: ETS transcription factor ERG
ERG	ETS transcription factor ERG
ETV1	ETS variant 1
ETV4	ETS variant 4
FLI1	Fli-1 proto-oncogene
CHD1	Chromodomain helicase DNA-binding protein 1
SPINK1	Serine peptidase inhibitor, Kazal type 1
RAF1	Raf-1 proto-oncogene
EZH2	Enhancer of zeste 2 polycomb repressive complex 2 subunit

PTEN: phosphatase and tensin homolog; AR: androgen receptor; ETS: external transcribed spacer

AR-targeted genes were included based on previous microarray analyses of genes regulated by androgens in PCa.10 Our results showed that the AR signature was heterogeneous in all treatment-naïve PCa specimens. AR activity might predict the response to therapies targeting the AR axis. PR315 and PRY04 had markedly lower AR activity than that of other samples, suggesting different responses to anti-androgen therapy compared with those of samples with high AR activity.

In addition, four cell cycle-related genes, nucleolar and spindle-associated protein 1 (NUSAP1), kinesin family member 11 (KIF11), cell division cycle 20 (CDC20), and forkhead box M1 (FOXM1), were evaluated as indicators of cell proliferation, a fundamental aspect of tumor biology.11 Most of the samples showed relatively low expression of cell cycle progression genes, suggesting low aggressiveness in the early stages of PCa. Our results support the use of qRT-PCR as a complementary approach to target gene sequencing for assessing critical transcriptomic events in PCa.

Molecular features correlated with clinical characteristics and prognosis

We subdivided the genomic aberrations in Chinese PCa patients identified by integrative genetic/transcriptomic assays into nine specific pathways (Supplementary Table 5). These molecules are commonly altered in Western patients with PCa and have clinical implications.12 We analyzed the correlations between these genomic aberrations and clinicopathological characteristics, including prostate-specific antigen (PSA) level, Gleason score, T stage, distant metastasis, and clinical risk stratification from the National Comprehensive Cancer Network (NCCN) and the European Association of Urology (EAU). As shown in Supplementary Table 5, activation of the PI3K/AKT/PTEN pathway and deletion of RB transcriptional corepressor 1 (RB1) were dramatically correlated with older age (P = 0.035 and P = 0.041, respectively). The ETS fusion was more common in patients at an advanced T stage (P = 0.030). Notably, an increased risk of progression was found in patients with the RB1 deletion, both in NCCN risk stratification (P = 0.006) and in EAU risk classification (P = 0.022). In addition, recurrent mutated or copy number-altered genes were included in the correlation analysis. GLI mutation was notably correlated with distant metastasis (P = 0.030).

Supplementary Table 5.

Clinical features and driver alterations in Chinese prostate cancer patients

Classification	All (n)	PTEN loss function mutation and deletions PIK3CA amplification			NCOA2 amplification SPOP mutation			AR activity			MYC amplification			RB1 deletion			Cell cycle-related gene expression			ETS fusion status			APC mutation/deletion			GLI1 mutation
		Yes (n)	No (n)	P	Yes (n)	No (n)	P	High (n)	Low (n)	P	Yes (n)	No (n)	P	Yes (n)	No (n)	P	High (n)	Low (n)	P	Yes (n)	No (n)	P	Yes (n)	No (n)	P	Yes (n)	No (n)	P
Age (year)
≤60	12	1	11	0.035*	4	8	0.416	7	4	0.310	7	5	1.000	2	10	0.041*	6	5	0.736	2	10	0.701	5	7	1.000	4	8	1.000
>60	34	15	19		6	28		14	18		21	13		19	15		15	17		9	25		15	19		11	23
Total PSA levels (ng ml−1)
<10	17	7	10	0.787	6	11	0.110	8	7	0.538	9	8	0.687	6	11	0.146	8	7	0.925	5	12	0.824	7	10	0.477	8	9	0.155
10–20	10	3	7		0	10		6	4		6	4		3	7		5	5		2	8		6	4		1	9
>20	19	6	13		4	15		7	11		13	6		12	7		8	10		4	15		7	12		6	13
Gleason score
<7	11	4	7	1.000	4	7	0.427	6	5	0.335	6	5	0.860	4	7	0.555	5	6	0.632	4	7	0.505	4	7	0.865	5	6	0.207
7	21	7	14		4	17		11	8		13	8		9	12		8	11		5	16		9	12		8	13
>7	14	5	9		2	12		4	9		9	5		8	6		8	5		2	12		7	7		2	12
T stage
pT2	33	12	21	0.703	9	24	0.335	16	16	0.349	20	13	0.714	12	21	0.066	15	17	0.687	6	27	0.030*	14	19	1.000	11	22	0.699
pT3	7	3	4		0	7		2	5		5	2		6	1		3	4		1	6		3	4		3	4
pT4	6	1	5		1	5		3	1		3	3		3	3		3	1		4	2		3	3		1	5
Distance metastasis
Yes	3	0	3	0.541	1	2	0.529	1	1	1.000	2	1	1.000	2	1	0.162	2	0	0.233	0	3	1.000	3	0	0.075	3	0	0.030*
No	43	16	27		9	34		20	21		26	17		10	33		19	22		11	32		17	26		12	31
NCCN risk stratification
Low	3	1	2	0.173	0	3	0.607	3	0	0.314	1	2	0.378	2	1	0.006*	2	1	0.314	1	2	0.352	0	3	0.099	1	2	0.263
Favorable intermediate	11	3	8		3	8		6	5		4	7		0	11		4	7		3	8		5	6		4	7
Unfavorable intermediate	5	4	1		2	3		3	2		4	1		3	2		3	2		1	4		4	1		1	4
High	15	7	8		2	13		4	11		11	4		10	5		6	9		2	13		5	10		5	10
Very high	7	1	6		1	6		3	2		4	3		3	4		4	1		4	3		3	4		1	6
Regional	2	0	2		1	1		1	1		2	0		1	1		0	2		0	2		0	2		0	2
Metastatic	3	0	3		1	2		1	1		2	1		2	1		2	0		0	3		3	0		3	0
EAU risk classificiton
Low risk	3	1	2	0.376	0	3	0.862	3	0	0.078	1	2	0.242	2	1	0.022*	2	1	0.512	1	2	0.183	0	3	0.210	1	2	0.165
Intermediate risk	12	3	9		3	9		8	4		5	7		1	11		4	8		4	8		6	6		4	8
High risk	15	8	7		4	11		4	11		12	3		8	7		8	7		1	14		6	9		4	11
Locally advanced	13	4	9		2	11		5	6		8	5		8	5		5	6		5	8		5	8		3	10
Metastatic	3	0	3		1	2		1	1		2	1		2	1		2	0		0	3		3	0		3	0

*P < 0.05. EAU: European Association of Urology; AR: androgen receptor; ETS: external transcribed spacer; PSA: prostate-specific antigen; NCCN: National Comprehensive Cancer Network; APC: APC regulator of WNT signaling pathway; GLI1: GLI family zinc finger 1

The prognostic impacts of the five specific pathways and recurrently altered genes were assessed. Log-rank tests were used to analyze the outcome for each individual subtype. MYC proto-oncogene (MYC) amplification (P = 0.036), RB1 deletion (P = 0.029), APC mutation/deletion (P = 0.032), and CDK12 mutation (P = 0.029) were associated with worse progression-free survival (PFS) in the 46 Chinese patients with PCa (Figure 3). Furthermore, we explored the clinical characteristics related to prognosis in Chinese populations. As expected, PFS decreased with increasing T stage (P = 0.0024), distant metastasis (P = 0.0012), NCCN risk stratification (P = 0.0011), and EAU risk classification (P = 0.00039) (Supplementary Figure 1 ^{(1.6MB, tif)} ). These results indicated a high predictive value of these risk factors, which may contribute to clinical decision-making in Chinese patients with PCa.

Figure 3.

MYC amplification, RB1 deletion, APC mutation/deletion, and CDK12 mutation are linked to worse prognosis in our cohort. Survival analysis was performed in Chinese PCa patients treated according to the presence or absence of (a) MYC amplification, (b) RB1 deletion, (c) APC mutation/deletion, and (d) CDK12 mutation. PCa: prostate cancer; MYC: MYC proto-oncogene, bHLH transcription factor; RB1: RB transcriptional corepressor 1; APC: APC regulator of WNT signaling pathway; CDK12: Cyclin-dependent kinase 12.

KMT2D is positively correlated with the AR signaling pathway in PCa

Another advantage of the integrative genetic/transcriptomic assay is the ability to explore the correlations underlying recurrently altered genes. We performed a correlation analysis of the expression of the 27 robust target genes across all samples. We observed correlated expression of genes within the AR signaling pathway and cell cycle progression modules (Figure 4a). In particular, positive correlations were identified between KMT2D and AR (r = 0.31, P = 0.046), the AR downstream target calcium/calmodulin-dependent protein kinase kinase 2 (CAMKK2) (r = 0.45, P = 0.046), the cell cycle progression gene CDC20 (r = 0.32, P = 0.037), and SPINK1 (r = 0.40, P = 0.008) (Figure 4b). Because the AR signaling pathway is indispensable for PCa carcinogenesis, we further validated the correlation between KMT2D and AR signaling using the Cancer Genome Atlas Prostate Adenocarcinoma (TCGA-PRAD) data. Consistent with our qRT-PCR results, KMT2D expression was strongly correlated with AR mRNA levels (r = 0.70, P < 0.001; Figure 4c left) and AR protein levels (r = 0.32, P < 0.001; Figure 4c right). Patients with AR-V7 had higher KMT2D expression (P = 0.036; Figure 4d), suggesting that KMT2D contributes to the AR splice variants. These findings highlight the close linkage between KMT2D and AR signaling and suggest that KMT2D plays a critical role in the activation of the AR axis in PCa.

Figure 4.

KMT2D is positively correlated with the AR signaling pathway in PCa. (a) Heatmap of correlations (r) between normalized gene expression for the 27 robust target genes across all samples shown in Figure 2. (b) KMT2D mRNA levels were positively correlated with AR, CAMKK2, CDC20, and SPINK1 expression in PCa tissues in our cohort. (c) KMT2D mRNA levels were positively correlated with both gene and protein expression of AR in TCGA-PRAD. (d) mRNA expression of KMT2D in AR-V7-positive and AR-V7-negative cases from TCGA-PRAD. PCa: prostate cancer; TCGA-PRAD: The Cancer Genome Atlas Prostate Adenocarcinoma. All abbreviated names of genes are defined in Supplementary Table 6.

DISCUSSION

We used a rapid, robust, and high-throughput approach for the characterization of gene mutations, copy number variants, and crucial transcriptomic events in Chinese patients with PCa by combined multiple-PCR-based deep NGS and qRT-PCR. This integrative assay required only 50 ng of DNA and 500 ng of RNA, which can be co-isolated from most routine clinical samples, making it practical for academic research settings as well as routine clinical settings. In this study, we added to the current knowledge of the genetic heterogeneity of prostate tumors among ethnic populations and identified correlations between clinical and prognostic factors and recurrently altered genes, providing a basis for therapeutic decisions. Moreover, we observed a close relationship between KMT2D and AR signaling in a gene co-expression analysis, highlighting the important role of KMT2D in the activation of the AR axis in PCa.

Our findings expand our understanding of the ethnicity-specific molecular landscape of PCa. As a heterogeneous disease, previous studies have reported notable disparities in PCa epidemiology among ethnic groups.13 Emerging evidence suggests that these differences might be the result of distinct molecular characteristics.14,15 Ren et al.16 identified salient genetic characteristics of PCa in Chinese patients. Similarly, the molecular profiles of Chinese patients with PCa identified in our cohort differed from that of Western patients. The most frequently mutated genes were KMT2D, ZFHX3, AKAP9, and GLI1 in our cohort; recurrently mutated genes in Western populations are SPOP, titin (TTN), TP53, mucin 16 (MUC16), MED12, and Forkhead Box A1 (FOXA1).17 These results might be due to either de novo ethnicity-specific mutations in Chinese patients or high-depth sequencing effects. Asian and Western populations have distinct genetic polymorphisms,18,19 familial aggregation,20 and diet.21 Basic science and epidemiological studies have demonstrated that these factors have a role in the clinical heterogeneity of PCa.22 Thus, the differences in these factors might influence the tumor genome and contribute to the ethnic genetic heterogeneity of PCa. Moreover, unlike other large-scale sequencing studies in Western patients, in which the sequencing depth is about 100×, in this study, we focused on 32 PCa-correlated genes and the average sequencing depth was approximately 3527×. The higher sequencing depth can remarkably increase the sensitivity of single-nucleotide variant detection;23 hence, the high coverage might account for the high mutation frequency of some genes in our cohort.

In addition to mutation, using an algorithm developed by Grasso et al.,9 we detected CNVs in our amplicon-based sequencing data. We observed copy number gains of KMT2D, AKAP9, MYC, and GLI and copy number losses of PTEN, RB1, and APC in Chinese patients. PTEN inactivation and copy number deletions account for the majority of PTEN loss-of-function cases.5 We found a low frequency of these genomic changes, i.e., 23.9% (11/46) in our cohort, compared with approximately 40% in Western samples.15 MYC amplification and RB1 deletion are two major CNVs detected in multiple genomic analyses.12 We also found MYC gains (60.9%, 28/46) and RB1 deletions (45.7%, 21/46) in our samples. However, the frequencies were higher than those in Western populations, in which MYC amplification and RB1 were deleted in 7.4% (37/498) and 16.3% (81/498) cases from TCGA-PRAD, respectively. For critical transcriptomic events, we identified a lower frequency of the ETS fusion (23.9%, 11/46) in our patients than in Caucasian patients (with a prevalence of approximately 50%).24 The results of other large-scale studies further support the low frequency of the ETS fusion in Asian patients,14,15,25 indicating regional differences in molecular subtypes.

Furthermore, we identified clinical associations for the recurrently altered genes. Copy number deletion of RB1 was remarkably correlated with a high risk of disease progression. ETS fusion and GLI1 mutations were more common in patients with a higher T stage and distant metastasis, respectively, than in other patients. Importantly, poor prognosis was associated with various alterations, including MYC amplification, RB1 deletion, APC mutation or deletion, and CDK12 mutation. MYC and RB1 are involved in the regulation of the cell cycle; amplification of MYC prompts progression from G1 to S phase,17,26 and MYC amplification status is predictive of biochemical recurrence after prostatectomy.27 However, its mRNA or protein levels show weak prognostic ability.28 RB1 is a critical negative regulator of the G1–S checkpoint and is responsible for repressing E2F family transcription factors.29 According to a recent study, RB1 deletion occurs early in PCa, prior to subclonal diversification.30 Consistent with our results, a large PCa tissue microarray study indicated that the deletion of RB1 is linked to an adverse phenotype and poor prognosis in this disease.31 APC is a multifunctional protein and is as a negative regulator of the Wnt pathway.32 The inactivation of APC by mutations or hypermethylation is associated with tumor aggressiveness.33,34 In Chinese populations, Geng et al.35 showed that APC mutations are associated with both PCa progression and all-cause mortality. CKD12 is a cyclin-dependent kinase; however, Wu et al.36 found that inactivation of CDK12 could define a distinct subtype of PCa, leading to increased neoantigen burden and T cell infiltration. Moreover, the loss of CDK12 is associated with genomic instability and predicts a worse poor prognosis in breast cancer.37 These findings further indicated that these genomic events are associated with clinical parameters and might serve as biomarkers in Chinese patients. However, given the small sample size, an independent cohort of patients is needed to confirm these results in the future.

Our integrative genetic/transcriptomic assays might also have utility for personalized medicine. The five main pathways involved in our analysis have clinical implications, and drugs targeting these alterations have been developed.12 For instance, the PI3K/AKT/PTEN pathway was upregulated in 34.78% of samples in our cohort. Therapeutically, to target this pathway, pan-PI3K and dual PI3K–mTOR inhibitors have been developed, and the AKT-targeting drugs BKM120 and MK-2206 show clinical efficacy in PCa.6 Cancer immunotherapies have become a revolutionized approach to the treatment of cancers, including PCa.38 The inactivation of CDK12 delineates a distinct immunogenic class of PCa, and a clinical study showed that patients with metastatic PCa who harbor a biallelic CDK12 loss have a higher likelihood of response to immunotherapy.36 In our cohort, 15.22% (7/46) of patients had CDK12 mutations. Among them, PR110, PR113, PR206, and PR309 were in an advanced state and relapsed after radical prostatectomy. These patients might benefit from immune checkpoint inhibitors after progression to the final stage of this disease. In addition, we evaluated AR signaling pathway activity. Consistent with previous results, most of our treatment-naïve cases showed high AR activity, indicating a better response to anti-androgen therapy.10 Compared with other cases, PR305 and PRY04 showed decreased AR activity, suggesting a heterogeneous response. Therefore, careful attention is needed when administering hormonal therapy. Above all, our findings suggest that molecular profiling using a targeted panel could provide valuable information, i.e., personalized mutation profiles for actionable targets, thereby contributing to clinical treatment decisions.

Notably, we detected a close link between KMT2D and the AR signaling pathway in our cohort using the integrative assays. These results were further confirmed using an additional dataset, TCGA-PRAD, showing that KMT2D expression was positively correlated with AR expression at the mRNA and protein level. Moreover, AR-V7-positive patients had high KMT2D expression levels. KMT2D is a histone lysine methyltransferase involved in the monomethylation of H3K4 residues (histone H3 lysine 4-monomethylation, H3K4me1). We previously concluded that, in PCa, KMT2D epigenetically upregulates LIF receptor alpha (LIFR)and Kruppel-like factor 4 (KLF4) expression, activating PI3K/AKT and epithelial–mesenchymal transition (EMT) oncogenic pathways and promoting PCa outgrowth and metastasis.8 Recent studies have reported that KMT2D is a transcriptional coactivator that recognizes target genes via DNA-bound transcription factors.39 In breast cancer, KMT2D is required for FOXA1, PBX homeobox 1 (PBX1), and estrogen receptor (ER) recruitment and activation and contributes to resistance to PI3Kα inhibitors.40 Thus, the positive correlation between KMT2D and the AR signaling pathway raises the possibility that KMT2D serves as a transcriptional coactivator and controls AR activation through posttranslational modification of epigenetic regulators. Further research is required to comprehensively clarify the role of KMT2D in PCa.

CONCLUSION

We applied an integrative approach to identify the critical genomic and transcriptomic events in 46 routine clinical Chinese PCa samples. This approach is helpful in characterizing retrospective material and analyzing correlations between clinical parameters and molecular changes. More importantly, focused assessments of genomic and transcriptomic features are important for disease pathology research as well as for advancements in personalized medicine.

AUTHOR CONTRIBUTIONS

SDL carried out the study design, data acquisition, statistical analysis, and interpretation and drafted and revised the manuscript. HYW carried out the data acquisition and statistical analysis. XPY participated in the study design and coordination and helped draft the manuscript. QLZ helped data acquisition. YBW provided the important suggestions for manuscript revision. QW and WHH supervised the study and provided the fund support. All authors read and approved the final manuscript.

COMPETING INTERESTS

All authors declared no competing interests.

Supplementary Table 4

High confidence somatic nonsynonymous mutations identified by target gene sequence

Supplementary Figure 1

T stage, distance metastatic, NCCN risk stratification and EAU risk classification are linked to worse prognosis in our cohort. Survival analysis was performed on Chinese PCa patients treated according to the presence or absence of T stage (a), or distance metastatic (b), or NCCN risk stratification (c), and EAU risk classification (d).