This article is mentioned in:

Introduction

Prostate cancer (PCa) is the second most common type of cancer affecting men worldwide, with an incidence of 1.1 million new cases each year (1). The most widely accepted pathological grading of PCa is the Gleason Score (GS), which is based on architectural features of the gland and is comprised of the sum of two Gleason Patterns (GP) (2). The GS grading system is among the most effective predictors of clinical outcomes following radical prostatectomy (RP), and is universally used to guide PCa treatment (3). Within this system, GS6 PCas are low grade, whereas GS8 or higher are classified as high grade. GS7 prostate adenocarcinomas consisting of variable proportions of GP3 and GP4 are considered to be intermediate grade, and represent the most heterogeneous group of neoplasms with diverse clinical outcomes (4,5) Thus, GS7 PCa is the focus of the current study, as identifying specific prognostic factors that can provide additional information on disease progression and treatment outcomes for patients with GS7 PCa would be of marked clinical benefit.

GP4 is assigned to adenocarcinomas when they contain any of the following architectural patterns: Small/large fused glands, poorly formed glands, glomeruloid and cribriform architecture (4–6). The cribriform pattern is commonly associated with intraductal carcinoma (IDC), a distinct histopathological entity characterized by malignant cells spanning the lumen of prostate ducts and acini (7–9). Although specific morphologies of PCa, including cribriform architecture and IDC, have been recognized for decades, their independent clinical significance is only now emerging (8). Studies have demonstrated that IDC in biopsy and/or RP tissues is indicative of a possible adverse clinical course and metastatic disease, warranting further aggressive treatment (10–13). Similarly, cribriform architecture has been demonstrated to be a clinically significant independent prognosticator for PCa-specific mortality (14–17).

The unequivocal identification of these pathological features based on the morphological criteria alone has been reported as challenging (18,19). In this regard, molecular features such as DNA methylation alterations may be of value. Aberrations in DNA methylation deregulate the genome and contribute to the loss of tissue homeostasis observed in aging and in diseases such as PCa (20). They may constitute the driver and the passenger events of tumorigenesis, as is evident from the widespread hypermethylation at the promoter regions of tumor suppressor genes, which is a hallmark of PCa (20,21). Detection of tumor-specific DNA methylation alterations in biopsy tissues may, in the future, serve as molecular indicators of cribriform architecture and/or of IDC. Additionally, DNA methylation markers are emerging as promising biomarkers of prostate carcinogenesis, thus complementing, if not altogether avoiding, the requirement for the histopathological examination of biopsy tissue samples (22). The results of our previous studies demonstrated that the DNA hypermethylation of a panel of seven genes [adenomatous polyposis coli (APC), cytochrome P450 family 26 subfamily A member 1 (CYP26A1), homeobox D3 (HOXD3), HOXD8, Ras association domain family member 1 (RASSF1), T-box 15 (TBX15) and transforming growth factor-β2 (TGFβ2)] is associated with PCa disease progression and clinical outcome (23–25). These markers were initially identified through a genome-wide methylation screen of low-grade and high-grade PCas (23). Subsequently, the prognostic potential of a subset of these markers was validated in an independent cohort (24,25). However, to the best of our knowledge, their association with the tumor architectural features of aggressive PCa has not yet been investigated.

In the present study, cribriform architecture and IDC were characterized in GP4 tumors derived from GS7 RP specimens, and their association with the aforementioned panel of seven DNA methylation markers was examined.

Materials and methods

Clinical samples and information

The present study was approved by the University Health Network Research Ethics Board. Retrospective formalin-fixed paraffin-embedded (FFPE) RP specimens from a total of 91 patients diagnosed with PCa between 1998 and 2001 at the University Health Network were included in the current study. This is a subset of a larger cohort of 246 patients (consisting of 91 GS7 tumors) that was characterized in our previous study (17). As previously reported, GS ≥8 tumors were limited in the present cohort and, thus, were excluded from analysis (17). The clinical and pathological characteristics of the cohort are listed in Table I. All samples, as well as clinical and pathological follow-up data, were obtained according to protocols approved by the Research Ethics Board of Mount Sinai Hospital, Toronto, and the University Health Network (Toronto, Canada).

Table I. Clinicopathological characteristics of the study cohort.

Table I.

Clinicopathological characteristics of the study cohort.

Clinicopathological characteristics	Value (range)
Total number of patients	91
Gleason scorea, no. of patients
7 (3+4)	63
7 (4+3)	28
Pathological stage, no. of patients
pT2	45
pT3a	34
pT3b	11
pT4	1
Surgical margin status, no. of patients
Negative	65
Positive	26
Average preoperative PSAb	8.56 (2.51–32.70)
Average prostate weight, gramsb	50.3 (22.0–143.5)
Median age, years	63.4 (47.6–74.5)

a ISUP 2005 GS assigned by an expert genitourinary pathologist

b n=85. PSA, prostate-specific antigen.

Histological evaluation

The complete set of hematoxylin and eosin (H&E)-stained slides from each RP tissue sample was collected and reviewed by an expert genitourinary pathologist in order to assign a modified GS (International Society of Urological Pathology 2005), along with the pathological stage, prostate weight and surgical margin status (4). For each GS7 PCa, the most representative GP4 tumor regions were marked on the H&E-stained slides corresponding to an area of ≥80% neoplastic cellularity, and the cells isolated from these regions were used for DNA methylation analysis. Only pure GP4 tissue was analyzed. The same marked areas on the H&E-stained slides were also independently reviewed by another genitourinary pathologist for the presence or absence of cribriform architecture, which is characterized by confluent epithelial proliferation with multiple lumina without intervening stroma, and of IDC, which is defined as a lumen-spanning solid or cribriform expansile neoplastic proliferation within the prostate gland or ducts (Fig. 1). Therefore, the presence of cribriform architecture and IDC in the same GP4 area was investigated, an area that was manually macrodissected using a scalpel for subsequent DNA methylation analysis, as described previously (23–25). All GP4 lesions were determined to contain an equal tumor cell percentage within the selected regions of analysis.

Figure 1. Representative hematoxylin and eosin stains of prostate cancer Gleason pattern 4 tissues with (A) cribriform architecture (marked with a black arrow) and (B) intraductal carcinoma (marked with a blue arrow). Magnification, ×100.

DNA methylation analysis

DNA was extracted from 10 µm sections of previously prepared FFPE blocks matching the selected H&E slides using a QIAamp DNA Mini kit (Qiagen, Inc., Valencia, CA, USA), as previously described (23–25). DNA methylation was analyzed in the same GP4 area used for histological evaluation. Briefly, a total of 100–400 ng extracted DNA from the GP4 region for each GS7 case was bisulfite modified using EZ DNA Methylation Gold kit (Zymo Research Corp, Irvine, CA, USA), according to the manufacturer's protocol.

Methylated genes were detected using a MethyLight quantitative polymerase chain reaction (qPCR) assay that was conducted using the TaqMan® Gold with Buffer A pack (Thermo Fisher Scientific, Inc., Waltham, MA, USA), as described previously (23–25). The primer and probe sequences are presented in Table II. A percent methylation ratio (PMR) score was calculated to assess the methylation of each gene in GP4 separately for individual cases, as described previously, according to the formula of Eads et al (26): (Gene of Interest/ALU) sample/(Gene of Interest/ALU) CpGenome ×100%. CpGenome is a commercially available fully methylated DNA (EMD Millipore, Billerica, MA, USA). ALU was a control reaction that measured the level of input DNA and normalized the signal for each methylation reaction (27).

Table II. Primer and probe sequences used in the MethyLight assay for APC, CYP26A1, HOXD3, HOXD8, RASSF1, TBX15, TGF-β genes and ALU control reaction.

Table II.

Primer and probe sequences used in the MethyLight assay for APC, CYP26A1, HOXD3, HOXD8, RASSF1, TBX15, TGF-β genes and ALU control reaction.

Genea,b	Sequence
ALU	Forward: 5′-GGTTAGGTATAGTGGTTTATATTTGTAATTTTAGTA-3′
	Reverse: 5′-ATTAACTAAACTAATCTTAAACTCCTAACCTCA-3′
	Probe: 5′FAM-CCTACCTTAACCTCCC-3′BHQ1
APC	Forward: 5′-GAACCAAAACGCTCCCCAT-3′
ENSG000001349821	Reverse: 5′-TTATATGTCGGTTACGTGCGTTTATAT-3′
Chr5: 112737781-1127378562	Probe: 5′FAM-CCCGTCGAAAACCCGCCGATTA-3′BHQ1
CYP26A1	Forward: 5′-TTGTAGAGATTCGACGTACGCGG-3′
ENSG00000095596	Reverse: 5′-AAAACCTTCCGTCAAACATCCTCTACG-3′
Chr10: 93069103-93069230	Probe: 5′FAM-ACGCCCACGACGTACCCGCTTCCTTAC-3′BHQ1
HOXD3	Forward: 5′-TTAAAGGTTTATGGTTGCGC-3′
ENSG00000128652	Reverse: 5′-TTACGAACACTAAACTACACCCG-3′
Chr2: 176163186-176163274	Probe: 5′FAM-ACAAAACGTTCCCGACGCTTCTAAAA-3′BHQ1
HOXD8	Forward: 5′-TAGTCGGTTTTGGTTCGTTGC-3′
ENSG00000175879	Reverse: 5′-CGTTCTAAAACGAAAAAAAAAACTCGCG-3′
Chr2: 176128934-176129040	Probe: 5′FAM-TCCTCGAACAAAACGCGACTCCCGAATCTC-3′BHQ1
RASSF1	Forward: 5′-ATTGAGTTGCGGGAGTTGGT-3′
ENSG00000068028	Reverse: 5′-ACACGCTCCAACCGAATACG-3′
Chr3: 50340720-503407843	Probe: 5′FAM-CCCTTCCCAACGCGCCCA-3′BHQ1c
TBX15	Forward: 5′-GCGGTTTTGTAAGTATATTGTTGCG-3′
ENSG00000092607	Reverse: 5′-ACTCCGAATAAAACAAAAACTAAAATCCG-3′
Chr1: 118984438-118984535	Probe: 5′FAM - CAAATAACGCCGCCGAACGCCT-3′BHQ1
TGF-β	Forward: 5′-TTTTAGGAGAAGGCGAGTCG-3′
ENSG00000092969	Reverse: 5′-CTCCTTAACGTAATACTCTTCGTCG-3′
Chr1: 218346933-218347007	Probe: 5′FAM-TCTCGCGCTCGCAAACGACC-3′

a Ensembl gene accession number.

b Human genome assembly GRCh38/hg38.

c The RASSF1A TaqMan hydrolysis probe sequence contains the SNP rs4688725. APC, adenomatous polyposis coli; CYP26A1, cytochrome P450 family 26 subfamily A member 1; HOX, homeobox; RASSF1, Ras association domain family member 1; TBX15, T-box 15; TGFβ2, transforming growth factor β2; ALU, Arthrobacter luteus element.

Statistical analysis

The Mann-Whitney U test was performed to analyze the differences in the median continuous PMR across different clinicopathological categories. Pearson's Chi square test was used to analyze proportional differences in Cribriform and/or IDC status and clinicopathological categories. All statistical analyses were conducted using SPSS version 21 software (SPSS, IBM Software, Armonk, NY, USA). For all stated methods, a two-sided value of P≤0.05 was considered to indicate a statistically significant difference.

Results

Clinicopathological features associated with GP4 cribriform architecture and/or IDC

The gland morphology of 91 GP4 tumor specimens obtained from RPs with GS7 pathology composed of comparable proportion (≥80%) of tumor cells was examined in the current study. The presence of a cribriform growth pattern was identified in the GP4 areas of 61/91 (67.0%) GS7 cases, whereas IDC was present in 21/91 (23.0%) GS7 cases. All 21 RP tissue specimens with IDC were also positive for cribriform architecture. Notably, the proportion of cases positive for cribriform architecture and/or IDC did not significantly differ between GS7 cases composed predominantly of GP4 (GP4+3; where the proportion of GP4 is >50% of the total pattern seen) vs. predominantly GP3 (GP3+4); however cribriform and/or IDC positive status was significantly associated with a more advanced pathological stage (pT3 vs. pT2; P=0.007; Table III) and higher preoperative prostate-specific antigen levels (PSA; P=0.022; Table III).

Table III. Association between cribriform and/or IDC-positive (n=61) and negative (n=30) cases and clinicopathological features.

Table III.

Association between cribriform and/or IDC-positive (n=61) and negative (n=30) cases and clinicopathological features.

	Cribriform and/or IDC
Parameter	Negative	Positive	P-value
Gleason score, % cases
7 (3+4)	34.9	65.1	0.552a
7 (4+3)	28.6	71.4
Pathological stage, % cases
pT2	46.7	53.3	0.007a
pT3	20.0	80.0
Surgical margins, % cases
Negative	30.8	69.2	0.481a
Positive	38.5	61.5
Age, % cases
Below median	32.6	67.4	0.941a
Above median	33.3	66.7
Preoperative PSA, % cases
Below median	45.5	54.5	0.022a
Above median	22.0	78.0
Prostate weight, % cases
Below median	38.1	61.9	0.542a
Above median	31.8	68.2
Median PMR
APC	31.7	47.3	0.045b
HOXD3	25.8	27.1	0.888b
HOXD8	25.7	27.7	0.278b
CYP26A1	16.3	18.5	0.426b
RASSF1	69.5	99.2	0.007b
TBX15	10.0	21.6	0.013b
TGFβ2	0.0	0.0	0.287b

{ label (or @symbol) needed for fn[@id='tfn6-ol-0-0-6140'] } Cribriform architecture and/or IDC in association with clinicopathological features and median PMR values within the exact same GP4 area. P-values were obtained using the

a Pearson's χ² test

b Mann-Whitney U test. IDC, intraductal carcinoma; PMR, percent methylation ratio; PSA, prostate-specific antigen; GP, Gleason pattern; APC, adenomatous polyposis coli; CYP26A1, cytochrome P450 family 26 subfamily A member 1; HOX, homeobox; RASSF1, Ras association domain family member 1; TBX15, T-box 15; TGFβ2, transforming growth factor β2.

DNA hypermethylation in areas with GP4 cribriform architecture and/or IDC

Gene specific DNA methylation differences were investigated between GP4 tumor areas positive for cribriform and/or IDC, and cases that did not harbor these features. For all genes investigated, PMR values were increased in GP4+3 vs. GP3+4, and in pT3 vs. pT2 cases (Table III). However, a significant association was only observed between CYP26A1 PMR and GS (GP4+3 vs. GP3+4; P=0.004; Table IV), as well as between RASSF1 and TBX15, and pathological stage (pT3 vs. pT2; P=0.027 and P=0.011, respectively; Table IV). Among the seven genes investigated, GP4 with a cribriform/IDC component exhibited a significant increase in the median PMR for APC, RASSF1 and TBX15 (Table IV). The median PMR of APC in cases negative for cribriform and/or IDC features was 31.7%, whereas the median PMR was 47.3% in positive cases (P=0.045). The median PMR of RASSF1 was 69.5% in negative cases and 99.2% in positive cases (P=0.007). TBX15 methylation also exhibited a significantly increased PMR in cribriform/IDC positive cases, as compared with in negative cases, with a median PMR of 10.0% in negative cases and 21.6% in positive cases (P=0.013). An increased median PMR for APC and TBX15, but not for RASSF1, was also detected within GP3 areas of the same GS7 tumors that harbored cribriform/IDC features within GP4 areas (P=0.028, P=0.025 and P=0.280, respectively). Due to a limited number of biochemical recurrence events in the cohort, the prognostic utility of these biomarkers could not be comprehensively evaluated in the current study.

Table IV. Median PMR values of methylation markers and Mann-Whitney U P-values stratified according to Gleason score and pathological stage.

Table IV.

Median PMR values of methylation markers and Mann-Whitney U P-values stratified according to Gleason score and pathological stage.

	Median PMR values
Parameter	APC	HOXD3	HOXD8	CYP26A1	RASSF1A	TBX15	TGFb2
Gleason score
7 (3+4)	40.0	24.8	25.5	15.3	92.3	16.5	0.0
7 (4+3)	52.8	31.6	28.8	26.0	93.5	25.7	0.2
Mann-Whitney U P-value	0.120	0.084	0.492	0.004	0.523	0.220	0.779
Pathological stage
pT2	32.4	24.9	25.6	13.8	82.2	11.4	0.0
pT3	46.6	27.3	26.9	18.8	100.1	27.8	0.0
Mann-Whitney U P-value	0.082	0.272	0.738	0.227	0.027	0.011	0.397

[i] PMR, percent methylation ratio; APC, adenomatous polyposis coli; CYP26A1, cytochrome P450 family 26 subfamily A member 1; HOX, homeobox; RASSF1, Ras association domain family member 1; TBX15, T-box 15; TGFβ2, transforming growth factor β2.

Discussion

The effective management of patients with PCa is hindered by a lack of optimal, highly sensitive and specific biomarkers that are able to predict the risk of disease aggressiveness at the time of diagnosis (28). Therefore, PCa treatment decisions are made with limited, and at times conflicting, information (28). In the present study, prognostic DNA methylation biomarkers were incorporated along with histological evaluation of cribriform and IDC patterns, in order to examine their association. To the best of our knowledge, this is the first study to investigate DNA methylation aberrations in tumor areas consisting of specific tumor architectural features, including cribriform and/or IDC, which provide information on the aggressive potential of GP4 carcinoma glands in GS7 tumors (17). Aberrations to DNA methylation in certain GP morphologies require further investigation as it may improve personalized patient treatment by utilizing biomarker information. The DNA methylation of APC, CYP26A1, HOXD3, HOXD8, RASSF1, TBX15 and TGFβ2 genes were investigated in the current study, as they have previously been revealed as promising prognostic biomarkers for PCa with the potential for clinical utility in independent patient cohorts (23–25).

Of the seven genes analyzed, the median methylation levels of APC, RASSF1 and TBX15 were significantly associated with cribriform architecture and/or IDC within the same GP4 regions of GS7 RP cases, harboring a similar proportion and purity of tumor cells. Increased methylation of APC and TBX15, but not RASSF1, was also detected within GP3 areas of the same GS7 tumors, which harbored cribriform/IDC features within GP4 areas. Hypermethylation of APC, RASSF1 and TBX15 in the presence of cribriform architecture and/or IDC suggests these molecular markers may serve as indicators of these architectural features. This provides novel evidence for a link between cribriform and/or IDC features and methylation biomarkers, and warrants further investigation of additional hyper- and hypo-methylation events in association with various architectural patterns in PCa. Notably, the presence of cribriform architecture and/or IDC, rather than the precise percentage of cribriform cells, was associated with increased APC, RASSF1, and TBX15 methylation within the carcinoma area (data not presented). Given these results, it is proposed that there is an extensive increase in DNA methylation within cribriform/IDC architecture, as well as in the adjacent carcinoma tissue. This concept must be evaluated in future studies, as it may have clinically significant implications for the detection of DNA methylation-based biomarkers as surrogate indicators of cribriform and/or IDC patterns in biopsy tissue samples. This requires further investigation into the sensitivity and specificity of APC, RASSF1 and TBX15 PMR, in order to discriminate between cribriform and/or IDC positive and negative GP4 cases. The detection of increased DNA methylation of APC, RASSF1 and TBX15, indicating high-grade disease, may facilitate personalized treatment strategies, including intense follow-up or a lower threshold for initiating adjuvant radiotherapy for patients. Future studies are also necessary to explore the potential for the identification of a cribriform/IDC-associated DNA methylation signature in easily accessible biofluids, such as urine samples, as it may serve as a surrogate to examine epigenetic events heralding cribriform/IDC architecture in the prostate. In particular, due to the likely potential of IDC cells to travel into the prostatic urethra via the antecedent prostatic ductal system, it may be valuable to investigate APC, RASSF1 and TBX15 hypermethylation in association with IDC features in post-digital rectal exam urine samples obtained from patients with PCa.

Notably, not all of the seven methylation biomarkers were associated with cribriform and/or IDC architecture, suggesting unique epigenetically regulated pathways may be involved in PCa progression and its associated morphology. However, the potential biological roles of APC, RASSF1 and TBX15 methylation in the formation of these clinicopathological entities are not yet well characterized, and must be investigated in further studies. It has been demonstrated that DNA methylation at the promoters of APC and RASSF1 tumor suppressor genes can decrease gene expression levels, thus inducing signaling pathways important in cancer biology, including MYC, ERK1/2 and p21; however, the role of TBX15 in cancer has not yet been characterized (29–34). Conversely, it is possible that DNA methylation events are not biologically associated with cribriform and/or IDC architecture, but are separate hallmarks of (epi)genetic derangement in PCa.

One limitation to the current study is that it was carried out using a restricted number of RPs. Another potential limitation of the study is that it is an association study between biomarkers and clinical indices, not PCa-associated mortality that may predict clinical benefit. This was an exploratory study aimed at acquiring novel insight into DNA methylation aberrations in specific PCa tumor architectures. Therefore, only RP tissue samples with an abundant quantity of carcinoma were used. The results of the current study warrant future analysis of larger multi-site retrospective cohorts in order to elucidate the clinical utility and predictive clinical benefit of DNA methylation and cribriform/IDC architecture in GP4 PCa, especially in a biopsy setting.

In summary, the association of seven DNA methylation markers with cribriform and IDC architecture was assessed in a series of 91 GP4 carcinoma glands in RP tissues with GS7 PCa. Presence and prevalence of cribriform and/or IDC architecture was confirmed in prostate tumors, similar to previous studies on architectural patterns in PCa RPs (10–17). Furthermore, a significant increase in APC, RASSF1 and TBX15 methylation in association with cribriform/IDC features was demonstrated. Future studies are required to further validate the clinical utility and functional relevance of DNA methylation events in cribriform and IDC architecture, and whether it should be routinely included in the RP pathology reports. Due to the challenges in the unequivocal identification of these clinicopathological entities by certain pathologists, the evaluation of these morphological features must be incorporated in the rapidly developing field of digitized histopathology (18,19,34). This may provide clinically translatable imaging markers for PCa prognosis. External validation of these findings within a larger population with a longer follow-up time is, therefore, essential.

Acknowledgements

This study was supported by the National Cancer Institute of Canada/the Canadian Prostate Cancer Research Initiative (grant no. 18568); Movember Translation Acceleration Grant from Prostate Cancer Canada (grant no. 2014-01); the Ontario Graduate Scholarships; and the Campbell Family Fellowship award. The authors thank Mr Michael Nesbitt for organizing and maintaining the patient clinical information that was used in this study.

Glossary

Abbreviations

Abbreviations:

References