Integrative molecular profiling of routine clinical prostate cancer specimens

To enable widespread personalized medicine for prostate cancer, the authors developed an integrative DNA/RNA based molecular profiling assay (MiPC) compatible with routine specimens. Applied to over 50 samples, MiPC identified known molecular subtypes, detected novel mutations and copy-number alterations, and informed on heterogeneity between primary tumors and paired lymph node metastases.

Abstract

Background

Comprehensive molecular profiling led to the recognition of multiple prostate cancer (PCa) molecular subtypes and driving alterations, but translating these findings to clinical practice is challenging.

Patients and methods

We developed a formalin-fixed paraffin-embedded (FFPE) tissue compatible integrative assay for PCa molecular subtyping and interrogation of relevant genetic/transcriptomic alterations (MiPC). We applied MiPC, which combines capture-based next generation sequencing and quantitative reverse transcription PCR (qRT-PCR), to 53 FFPE PCa specimens representing cases not well represented in frozen tissue cohorts, including 8 paired primary tumor and lymph node metastases. Results were validated using multiplexed PCR based NGS and Sanger sequencing.

Results

We identified known and novel potential driving, somatic mutations and copy number alterations, including a novel BRAF T599_V600insHT mutation and CYP11B2 amplification in a patient treated with ketoconazole (a potent CYP11B2 inhibitor). qRT-PCR integration enabled comprehensive molecular subtyping and provided complementary information, such as androgen receptor (AR) target gene module assessment in advanced cases and SPINK1 over-expression. MiPC identified highly concordant profiles for all 8 tumor/lymph node metastasis pairs, consistent with limited heterogeneity amongst driving events. MiPC and exome sequencing were performed on separately isolated conventional acinar PCa and prostatic small cell carcinoma (SCC) components from the same FFPE resection specimen to enable direct comparison of histologically distinct components. While both components showed TMPRSS2:ERG fusions, the SCC component exclusively harbored complete TP53 inactivation (frameshift variant and copy loss) and two CREBBP mutations.

Conclusions

Our results demonstrate the feasibility of integrative profiling of routine PCa specimens, which may have utility for understanding disease biology and enabling personalized medicine applications.

introduction

Comprehensive molecular profiling has led to a relatively complete portrait of the prostate cancer (PCa) genomic/transcriptomic landscape, leading to robust molecular subtypes defined by essentially mutually exclusive transcriptional and/or genetic events [1]. The most common subtype-defining lesion, ETS gene fusions, which occur in ∼50–60% of PSA screened PCa (ETS⁺), can be detected at both the genetic and transcript levels. A subset of ETS fusion negative (ETS⁻) tumors harbor mutations in SPOP (SPOP^mut), deletion/mutation of CHD1 (CHD1^del), and/or over-expression of SPINK1 (SPINK1⁺), which require both genetic and transcriptional (or protein based) approaches for identification [1]. Alterations in key potential therapeutic targets (such as PTEN or BRCA2) or genes associated with specific histologic subtypes (such as RB1 mutation or deletion in prostatic neuroendocrine/small cell prostate carcinoma [SCC)]) have also been described and may occur in both ETS⁺ and ETS⁻ cancers [1–3]. Similarly, androgen receptor (AR) amplification or mutation is commonly observed in progressive disease after anti-androgen therapy (termed castration resistant prostate carcinoma [CRPC]) [4]. Transcriptional evaluation provides a real-time readout of AR signaling activity that impacts interpretation of genomic events [1]. Likewise, assessment of cell cycle/proliferation related genes is included in many recently developed prognostic assays for localized PCa [5].

Despite major advances in next generation sequencing (NGS) technology, comprehensive, integrative genomic/transcriptomic characterization remains challenging in routine patient care, as such approaches typically require large amounts of fresh frozen tissue [6], while clinical specimens are processed using formalin fixed paraffin embedding (FFPE). Hence, FFPE compatible approaches for identifying molecular subtypes, driving and/or potentially targetable genetic lesions and assessment of key transcriptional modules may have utility both in localized and advanced PCa, and are critical for the widespread adoption of personalized medicine strategies for all patients with PCa.

materials and methods

cohort

Archived clinical formalin-fixed paraffin-embedded (FFPE) PCa tissue specimens were obtained from the University of Michigan Department of Pathology Tissue Archive and the Specialized Program of Research Excellence (SPORE) Tissue Core. Fifty three prostate cancer specimens from 41 patients were used for next-generation sequencing and/or qRT-PCR with IRB approval. DNA/RNA co-isolation was performed essentially as described [7] from 3–10 × 10 um FFPE sections per sample. See Supplementary Methods for additional information.

MiPC-D

We designed a custom Agilent Haloplex capture panel for Ion Torrent Sequencing using the SureDesign website that targeted 436 kb (See supplementary Table S1, available at Annals of Oncology online and Supplementary Methods). Library preparation was performed using the Ion HaloPlex Target Enrichment System (Agilent, Santa Clara, CA), starting with 200 ng DNA per sample according to the manufacturer's instructions. Three to five libraries were combined for template preparation and sequencing using an Ion Torrent 318 chip on the Ion Torrent Personal Genome Machine (PGM) sequencer as described [7]. Data analysis was performed essentially as described [7, 8] and detailed in the Supplementary Methods and supplementary Table S2, available at Annals of Oncology online.

MiPC-R

We designed an 8 × 48 format TaqMan low density array (TLDA) to interrogate the expression of 43 target genes and five housekeeping genes (assays given in supplementary Table S3, available at Annals of Oncology online). Reverse transcription (RT) of 200 ng RNA was performed using gene specific priming and qPCR was performed on the ABI 7900 Sequence Detection System (Applied Biosystems). Detailed methods and assay qualification is described in the Supplementary Methods.

results

design and application of an integrative FFPE compatible assay for comprehensive PCa profiling (MiPC)

To enable comprehensive profiling of routine PCa samples for precision medicine, we developed an integrative DNA and RNA-based assay (MiPC), compatible with ∼200 ng FFPE isolated DNA and RNA (Figure 1A supplementary Figure S1, available at Annals of Oncology online). MiPC consists of a custom Haloplex capture panel and a TLDA qPCR array for detecting DNA and RNA alterations, respectively. The Haloplex capture panel (MiPC-D), targeting ∼0.44 Mb, was designed to assess key recurrently mutated, amplified, deleted or rearranged genes, identified in previous PCa profiling studies, as well as pan-cancer altered therapeutic targets, when coupled with Ion Torrent sequencing (supplementary Table S1, available at Annals of Oncology online). The TLDA qPCR assay (MiPC-R) was comprised of 48 expression assays targeting robust housekeeping genes, AR⁺ and AR⁻ transcriptional modules, proliferation/cell cycle genes, and genes that define key molecular subtypes or driving alterations (supplementary Table S3, available at Annals of Oncology online). We applied MiPC to a cohort of 53 FFPE PCa specimens from 41 patients, representing a mix of cases not well studied in frozen tissue cohorts, as detailed in the Supplementary Results and supplementary Table S4, available at Annals of Oncology online.

Figure 1.

Integrative molecular profiling of routine archived prostate cancer (PCa) samples. (A) Overview of the MiPC assay, which utilizes next generation sequencing (NGS) and qRT-PCR assays to perform integrative profiling of formalin-fixed paraffin-embedded (FFPE) PCa tissue samples. (B) A heatmap of transcriptional module gene expression and somatic alterations for each assessed sample are shown. Clinicopathological information is provided in the header, including specimen type (prostatectomy or transurethral resection [RRP/TURP], RRP node, prostatic biopsy [Pros. Bx]) or metastasis), prior treatment, molecular subtype (by MiPC) and tumor content estimated by H&E staining. Paired samples are indicated by the same color. Samples excluded for CNA calling due to suboptimal profiling data and those not assessed by MiPC-D are indicated. Average transcriptional module expression (AR⁺, AR⁻ and cell cycle/proliferation [C.C./Prolif.]) is shown according to the color scale, with samples excluded due to insufficient housekeeping gene expression indicated. High-level somatic CNAs (copy number ratios >2 or <0.6) and mutations are indicated according to the legend (point = point mutation; fp = frame-preserving indel; fs = frame-shifting indel; stop = non-sense).

single nucleotide variant (SNV) and short insertion/deletion (indel) detection with MiPC

Ion Torrent NGS after MiPC-D capture was applied to 48 FFPE PCa specimens and a pooled benign matched normal FFPE tissue specimen (PRMN), yielding an average of 1 505 078 mapped reads and 242 × coverage over targeted bases per sample (supplementary Table S5, available at Annals of Oncology online). We identified a total of 106 high confidence somatic single nucleotide variants (SNV)/indels across the 48 samples (average 2, range 0 to 14), including known drivers in low tumor content samples (as estimated by hematoxylin and eosin [H&E] staining), as well as dominant and subclonal mutations in the same sample (Supplementary Results). All high-confidence somatic, non-synonymous mutations are shown in an integrative molecular heatmap in Figure 1B and given in supplementary Table S6, available at Annals of Oncology online.

copy number alteration assessment with MiPC

Using an approach adapted from multiplexed PCR-based targeted Ion Torrent NGS data [8], we simultaneously assessed copy number alterations (CNAs) in 93 genes from MiPC-D data. Four out of 48 samples (8%) had excessively noisy CNA profiles and were excluded from this analysis as described in the Supplementary Results and Figure S2, available at Annals of Oncology online. Copy number ratio heatmaps for all informative samples are shown in Figure 2A and supplementary Figure S3, available at Annals of Oncology online. At the gene level, we identified 117 high-level CNAs across the 44 informative PCa samples (average 2 high-level CNAs per sample, range 0 to 11), as shown in Figure 1B. High-level AR amplifications, which are known to develop in response to anti-androgen therapy, were identified in four samples (PR-2, PR-25, PR-49 and PR-14), all of which were from patients with CRPC. We confirmed mutations and CNAs identified by MiPC-D using multiplexed PCR-based sequencing (Ampliseq) as described in the Supplementary Results and Figure S4, available at Annals of Oncology online. DNA rearrangements were not robustly detected using the Haloplex technology (Supplementary Results, supplementary Table S7 and Figure S5, available at Annals of Oncology online).

Figure 2.

Copy number profiles and gene expression signatures of prostate cancer (PCa) assessed by MiPC. (A) Unsupervised hierarchical clustering of copy number profiles from MiPC-D. Copy number ratios (log₂) for 93 genes assessed by the MiPC-D target capture next generation sequencing (NGS) panel are shown according to the color scale. Clinicopathological information is provided as in Figure 1, and paired samples are indicated by the same color. Selected genes of interest are indicated. All genes are labeled in supplementary Figure S3A, available at Annals of Oncology online. (B) Unsupervised hierarchical clustering of gene expression signatures from MiPC-R. Normalized target gene expression (log₂) for 31 robust target genes assessed by the MiPC-R qRT-PCR panel are shown according to the color scale. Genes activated (AR⁺, black) or repressed (AR⁻, white) by androgen receptor signaling, cell cycle/proliferation related genes (orange) and genes used for assessing molecular subtyping or representing key alterations (green) are indicated. Clusters defined by T2:ERG (and ERG), AR⁺/AR⁻ and proliferation/cell cycle module expression are indicated. All genes are labeled in supplementary Figure S6A, available at Annals of Oncology online.

assessment of key PCa transcriptome alterations using MiPC

We performed RT using gene specific priming on co-isolated RNA followed by qRT-PCR using the MiPC-R TLDA assay on all 53 FFPE PCa samples and three benign matched normal FFPE prostate tissues. As described in the Supplementary Results and supplementary Tables S3 and 4, available at Annals of Oncology online), 31 target genes showed robust performance across 49 cancer samples and two benign prostate tissues (PR-41MN and PR-22MN), with unsupervised hierarchical clustering of normalized target gene expression shown in Figure 2B and supplementary Figure S6A, available at Annals of Oncology online. We observed correlated–expression of genes within the AR⁺, AR⁻ and proliferation/cell cycle modules (supplementary Figure S6B, available at Annals of Oncology online), and five clusters of samples were observed based largely on T2:ERG (and ERG), AR⁺/AR⁻ and proliferation/cell cycle module expression (Figure 2B). Average module expression and subtype-defining expression events (i.e., T2:ERG detection) are incorporated into the integrative heatmap (Figure 1B). To assess the technical reproducibility, we subjected PR-74 to duplicate RT and qPCR with MiPC-R (PR-74a and PR-74b), which showed highly correlated normalized target gene expression (r² = 0.99) and these samples cluster tightly as expected (Figure 2B). Taken together, our results support the ability of MiPC-R to assess key PCa transcriptomic events.

molecular subtyping using MiPC

In order to assess the ability of MiPC to perform basic PCa molecular subtyping, we integrated results from the DNA and RNA assays. As shown in Figures 1B and 2B, across the 49 PCa samples assessable by MiPC-R, T2:ERG (and ERG over-expression) was detected in 23 cancer samples (T2:ERG⁺), including PR-26 (SCC as described above), which showed ∼3000 fold-less T2:ERG expression than the median of the 22 other T2:ERG positive samples (supplementary Figure S7, available at Annals of Oncology online), consistent with loss of AR driven TMPRSS2 expression in this sample (T2:ERG^low). Three additional samples showed marked ERG over-expression without detectable T2:ERG (ERG⁺), consistent with alternate 5′ fusion partners or distinct T2:ERG isoforms not assessed using our qRT-PCR assay (supplementary Figure S7, available at Annals of Oncology online). Validation of these findings, including the ability of 3′ ETS expression as performed herein to serve as a surrogate for ERG⁺ status, is described in the Supplementary Results.

Detailed description of samples with other subtype defining alterations identified by MiPC—including other ETS gene rearrangements, genetic/transcriptomic alterations in SPOP, SPINK1 and/or CHD1, RAF/RAS family alterations, and hypermutator phenotype—is provided in Figure 3 and supplementary Figure S7, available at Annals of Oncology online and the Supplementary Results. Taken together, through interrogation of both DNA and RNA alterations, these results demonstrate the ability of MiPC to robustly molecularly subtype routine clinical FFPE PCa specimens.

Figure 3.

Identification of known molecular prostate cancer (PCa) subtypes and novel alterations through integrative profiling with MiPC. Copy number, somatic mutation and gene expression profiles of selected PCa of interest are shown. (A and B) Individual copy number plots and a heatmap of copy number ratios from MiPC-D. In (A) copy number ratios (log₂) for individual target regions are shown as points (colored by genes). Gene-level copy number ratios (log₂), indicated by black bars, are derived from the weighted average of individual target region copy number ratios. Genes are plotted in genome order and vertical gray lines separate individual chromosomes. (C) Somatic mutations from MiPC-D. (D) Gene expression signatures from MiPC-R. Individual cases are described in the text and Supplementary Results. For example, PR-25 is a CRPC specimen from a patient treated with ketoconazole, which demonstrates AR and CYP11B1 amplification (ketoconazole is a potent CYP11B1 inhibitor), one copy PTEN and 13q (including RB1) loss, PTEN non-sense mutation, high AR⁺ and cell cycle/proliferation module expression, and ETV5 outlier expression.

potential applications of MiPC for PCa precision medicine

Our integrative approach, which assesses both genomic and transcriptome alterations, is particularly useful for assessing alterations in AR signaling. AR activity may have utility in predicting response to therapies targeting the androgen signaling axis. For example, the integrative molecular profile of PR-25, a dural metastasis from a patient who had previously undergone anti-androgen therapy, chemotherapy and extensive second-line ketoconazole treatment is shown in Figure 3. In addition to high-level AR amplification, PR-25 also harbor-marked CYP11B1 amplification (Figure 3A and B). Despite being located on the frequently amplified chromosome 8q24 locus, CYP11B1 has not been previously reported as focally amplified in the treatment of naïve PCa or CRPC. As ketoconazole is a second-line antiandrogen and potent CYP11B1 inhibitor [9], CYP11B1 amplification in this case suggests an adaptive response to therapy. By qRT-PCR, PR-25 was ETV5⁺ (supplementary Figure S7, available at Annals of Oncology online) and had the highest AR expression in our cohort, with active AR signaling supported by high- and low-expression of the AR⁺ and AR⁻ modules, respectively (Figure 3D). Lastly, PR-25 showed high expression of the cell cycle/proliferation module (5th highest expression in our cohort).

Integrative profiles from PR-26, a biopsy of a SCC liver metastasis that harbored a RB1 homozygous deletion and decreased AR⁺ module expression, as well as PR-22, which harbored a likely activating BRAF T599_V600insHT mutation (validation in supplementary Figure S8, available at Annals of Oncology online) and AR and AKT1 activating mutations, are shown in Figure 3 and further described in the Supplementary Results.

assessment of heterogeneity in paired FFPE samples using MiPC

An advantage of using an FFPE compatible integrative assay, such as MiPC, is its ability to assess specimens that are not well represented in research cohorts, such as paired (concurrent or subsequent) samples, which can be used to assess heterogeneity in key clinical scenarios that may guide personalized medicine strategies. Hence, in our cohort, we assessed a total of 22 paired samples from 10 patients, including eight primary tumor/lymph node metastasis pairs (see Figures 1, 4 and supplementary Table S4, available at Annals of Oncology online), subsequent transurethral resections (TURPs) from a patient with CRPC (PR-27 and PR-34 as described in Figure 3 and the Supplementary Results), and separately macrodissected areas of well differentiated (PR-72), poorly differentiated (PR-73) and SCC (PR-74) from the same TURP specimen (as described below). Critically, the basic molecular subtype for all paired samples was concordant, and key genetic drivers and major transcriptional profiles were highly stable in paired primary tumors/lymph node metastases (Figure 4 and Supplementary Results).

Figure 4.

Assessment of paired primary prostate cancer (PCa) and lymph node metastases by MiPC demonstrates concordant molecular subtypes. CNAs, somatic mutation and gene expression profiles of paired primary PCa and concurrent lymph node metastases assessed by MiPC are shown. (A and B) Individual copy number plots and a heatmap of copy number ratios from MiPC-D for selected paired tumor/lymph node metastases are shown. In (A) copy number ratios (log₂) for individual target regions are shown as points (colored by genes), with gene-level copy number ratios indicated by black bars. Samples with PTEN deletions are plotted. Genes are plotted in genome order and vertical gray lines separate individual chromosomes. In (B) highly discordant genes from paired specimens are shown as bolded cells. (C) Somatic mutations from MiPC-D. Samples without mutations are not shown. Mutations exclusively present in lymph node metastases but not primary tumors are shown in bold. (D) Gene expression signatures from MiPC-R. Note, PR-07.1, PR-05.1 and PR-05.3 showed sufficient housekeeping gene expression only for assessing ETS status and full MiPC-R profiles from these tumor/lymph node metastasis pairs are not shown. Discordant AR⁺ module expression driving discordant clustering of PR-42 and PR-43 by MiPC-R (see Figure 2B), is indicated by a red box.

comprehensive molecular profiling of divergent histologic components using MiPC and exome sequencing

Lastly, we applied MiPC, as well as more comprehensive exome sequencing, to examine the utility of integrative profiling for identifying molecular underpinnings of diverse histologic components observed in the same tumor. Concurrent conventional acinar PCa and SCC have been shown to represent clonal progression through evaluation of T2:ERG fusions as clonal markers [10–12]. Hence, we used macrodissection guided by immunohistochemistry (IHC) to isolate areas of well differentiated (PR-72, morphologically consistent with Gleason score 3+4 = 7), SCC morphology but no neuroendocrine marker expression (PR-73) and frank SCC (PR-74, with chromogranin and synaptophysin expression by IHC) from the same FFPE TURP specimen from a patient with CRPC. As detailed in supplementary Figure S9, Table S8 and the Supplementary Results, available at Annals of Oncology online, MiPC and exome sequencing of these components demonstrated expected differences in transcriptional modules, and support complete TP53 inactivation and CREBBP mutations as drivers of the SCC transition in this case.

discussion

Here we report the development and application of an FFPE compatible approach for integrative, comprehensive molecular profiling of routine PCa specimens through combined Haloplex capture-based next-generation sequencing and qRT-PCR (MiPC). Importantly, this approach is compatible with ∼200 ng co-isolated DNA and RNA, which can be obtained from most routine specimens where precision medicine approaches are likely to be employed (aggressive high volume disease at biopsy, prostatectomy specimens, and core biopsies/excisions of metastases). However, as described in the Supplementary Discussion, alternative approaches will be needed to enable integrative personalized medicine strategies from more limiting specimens.

Unlike approaches that only profile DNA-based alterations, such as the FoundationOne test (Foundation Medicine) [13], our integrative approach also allows assessment of critical transcriptional modules (cell cycle/proliferation and AR regulated genes) and expression events, which may inform on therapeutic strategies in the future. For example, AR splice variant expression has recently been shown to predict response to second-line anti-androgen treatment [14]. Likewise, our transcriptional module of genes negatively regulated by AR signaling is novel. Although MET has previously been shown to be negatively regulated by active AR signaling [15], the other three genes, MIAT, ARNTL2 and S100A6, have not been associated with AR signaling previously and warrant further study, with ARNTL2 and S100A6 showing strong anti-correlation with AR and AR⁺ module expression, supplementary Figure S6B, available at Annals of Oncology online). Importantly, approaches that only profile DNA-based alterations cannot detect such splice isoforms or transcriptional modules. See Supplementary Discussion for additional comparisons between our approach and other methodologies.

Application of MiPC identified known and novel point mutations/indels and copy number alterations in PCa, including a novel somatic BRAF V600insHT mutation. As described in the Supplementary Discussion and supplementary Table S9, available at Annals of Oncology online, across more than 600 sequenced PCa (including our current results), ∼2–4% of patients with PCa harbor somatic BRAF alterations are potentially susceptible to therapies targeting the BRAF/MEK signaling axis [16, 17]. The need to develop precise medicinal strategies for small subsets of patients is similarly highlighted by the lack of driving/targetable alterations in our cohort, with 39% and 19% of untreated and treated samples, respectively, having ≤1 alteration by MiPC-D profiling).

Our cohort included advanced prostate cancer specimens, including paired tumor/lymph node metastases and CRPC/SCC. Although the relatively small cohort size and lack of paired pre-/post-treatment specimens precluded identification of alterations associated with specific treatment, we anticipate that our approach can now be applied to such specimens. See Supplementary Discussion for additional insight into the limited heterogeneity observed between integrative profiles from paired tumor/lymph node metastases, as well as insights from the MiPC/exome sequencing from separately dissected acinar and SCC components in the same sample, which support alterations in TP53 and CREBBP as drivers of the SCC transition in that case.

In summary, we have developed an integrative approach (MiPC) that assesses key PCa genomic and transcriptomic events from routine clinical FFPE specimens. Such an approach may be useful in characterizing retrospective material or enabling biomarker informed clinical trials. More specifically, focused assessment of both the genome and transcriptome from clinical specimens and distinct macrodissected tumor components may better inform on disease biology as well as facilitate personalized medicine adoption.

funding

This work was supported in part by the Evans Foundation/Prostate Cancer Foundation (to F.Y.F, K. E. K. and S.A.T (no grant number)) and the National Institutes of Health (R01 CA183857 to S.A.T. and R01 CA181605 to P.S.N. and S.A.T.). F.Y.F. and S.A.T. were supported by University of Michigan Prostate SPORE Career Development Awards. P.S.N., F.Y.F. and S.A.T. are supported by a Stand Up To Cancer–Prostate Cancer Foundation Prostate Dream Team Translational Cancer Research Grant. Stand Up To Cancer is a program of the Entertainment Industry Foundation administered by the American Association for Cancer Research (SU2C-AACR-DT0712). S.A.T. is supported by the A. Alfred Taubman Medical Research Institute (no grant number).

disclosure

The University of Michigan has been issued a patent on the detection of ETS gene fusions in prostate cancer, on which SAT is listed as a co-inventor. The University of Michigan licensed the diagnostic field of use to Gen-Probe, Inc., who has sublicensed some rights to Ventana/Roche. The University of Michigan has applied for a patent on the detection of SPINK1 in prostate cancer, on which SAT is listed as a co-inventor. The University of Michigan licensed the diagnostic field of use to Gen-Probe, Inc., who has sublicensed some rights to Ventana/Roche. SAT serves as a consultant to, and has received honoraria from, Ventana/Roche. SAT has a sponsored research agreement with Compendia Bioscience/Life Technologies/ThermoFisher that provided access to the pooled gene specific primers and supported the targeted multiplexed PCR-based confirmatory experiments. Compendia Bioscience/Life Technologies/Fisher had no other role in the data collection, interpretation, or analysis, and did not participate in the study design or the decision to submit for publication. All remaining authors have declared no conflicts of interest.