Genomic profiling defines subtypes of prostate cancer with the potential for therapeutic stratification

Abstract

The remarkable variation in prostate cancer clinical behavior represents an opportunity to identify and understand molecular features that can be used to stratify patients into clinical subgroups for more precise outcome prediction and treatment selection. Significant progress has been made in recent years in establishing the composition of genomic and epigenetic alterations in localized and advanced prostate cancers using array-based technologies and next generation sequencing approaches. The results of these efforts shed new light on our understanding of this disease and point to subclasses of prostate cancer that exhibit distinct vulnerabilities to therapeutics. The goal of this review is to categorize the genomic data and, where available, corresponding expression, functional, or related therapeutic information, from recent large-scale and in-depth studies that demonstrate a new appreciation for the molecular complexity of this disease. We focus on how these results inform our growing understanding of the mechanisms that promote genetic instability, as well as routes by which specific genes and biological pathways may serve as biomarkers or potential targets for new therapies. We summarize data that indicate the presence of genetic subgroups of prostate cancers and demonstrate the high level of intra- and intertumoral heterogeneity, as well as updated information on disseminated and circulating tumor cells. The integrated analysis of all types of genetic alterations that culminate in altering critical biological pathways may serve as the impetus for developing new therapeutics, repurposing agents used currently for treating other malignancies, and stratifying early and advanced prostate cancers for appropriate interventions.

INTRODUCTION

Prostate cancer is the second most commonly diagnosed cancer in United States men with more than 240,000 cases reported annually. These carcinomas exhibit a remarkable diversity in behavior ranging from decades of indolence to rapid growth, dissemination and lethality. Though pathological grading provides a powerful indicator of disease behavior, clinical outcomes of tumors with the same histological patterns can vary substantially. While significant morbidity results from the overtreatment of indolent tumors, delayed diagnosis and under-treatment of aggressive malignancies contributes to an excess of 30,000 deaths per year from metastatic prostate cancers. A better understanding of the genetic and molecular characteristics defining indolent and lethal prostate cancers is key for improved patient stratification and selection of optimal therapies.

This review will focus on the field of prostate cancer genomics, highlighting chromosomal alterations that may drive cancer behavior and serve as biomarkers to guide future therapeutic directions. Genomic studies have recently strengthened our understanding of prostate cancer by clarifying: 1) the frequency, types, and mutation characteristics in prostate cancer relative to other cancers, 2) the progression of genomic alterations during disease evolution, and 3) tumor heterogeneity and clonality. Collectively, these studies indicate that integrated analyses of genetic aberrations, changes in gene expression and resulting contributions to biological functions are necessary to understand the key features underlying prostate cancer behavior.

The mutational landscape of prostate cancer

Prostate cancer is characterized by extraordinary genomic complexity^{1, 2}, including somatic copy number alterations, point mutations, and structural rearrangements. Advanced prostate cancer may be aneuploid or have large regions of copy neutral loss-of-heterozygosity (cnLOH)³. Recent advances that collectively involve detailed analyses of hundreds of primary and metastatic prostate cancers now provide a clearer picture of genomic aberrations that accompany indolent and aggressive disease.

Somatic copy number alterations (SCNA)

SCNAs are genetic gains or losses that arise during cancer development. They are evident in nearly 90% of primary prostate tumors, with deletions typically outnumbering amplifications. These SCNAs tend to be focal (≤1–5 Mb), affecting only a small portion of the genome^{4, 5}. Metastatic prostate tumors, however, display dozens to hundreds of aberrations, which can affect a large portion of the genome. This difference suggests increased genomic instability as the disease progresses. A recent detailed comparison of SCNAs among cancer types determined that prostate cancer displayed more SCNAs (averaging 46 per sample) than most of the other 26 cancer types⁴. Frequent deletions are seen on chromosomes 6q, 8p, 10q, and 13q and include genes such as NKX3-1, PTEN, BRCA2 and RB1. Castration-resistant metastatic tumors (CRPC) show frequent amplification of chromosomes X, 7, 8q, and 9q, which include the androgen receptor (AR) and MYC oncogenes. Table 1 summarizes the most frequent SCNAs in different stages of prostate cancer development.

Table 1.

Most common somatic copy number aberrations (SCNAs) in human prostate cancer

Cytoband	Event	Size (Mb)	Genes of interest	Reported frequency in prostate cancer
				primary cancer	advanced cancer	DTC or CTC
Xp11.22-q13.1	Gain	18–67.8	AR		50% CRPC 15, 24, 25, 68	45% AdvDTC6, 7, 69
1p12-q43	Gain	117			45–65% CRPC24	50%69
1q32.1-q32.3	Gain	12.50	ELK4, PTPRC, ELF3, PTPN7, MDM4, RAB7L1, RASSF5, IL24, IL10, CAMK1G	5	24, 70	45% AdvDTC
3q26.1	Gain	43.80	GMPS, PIK3CA, MLF1, SKIL, CCNL1, ECT2	13–39%5, 71	24, 70	20%7
6q14.3-15	Loss	13.67	CYB5R4, NT5E, SNX14, SYNCRIP, HTR1E, CGA, GJB7	40%5, 18, 64, 72, 73	55%64, 70, 72, 74	25%6, 7
7p22.3-q36.3	Gain	158.40		5, 18	25–55% CRPC68, 70, 72, 74	40%6, 7
8p12-q24.3	Gain	97.64	MYC, MAF, EYA1, MSC, TRPA1, KCNB2	20–30%5, 18, 64, 71–73, 75	64–82% CRPC15, 62, 68, 70, 72	50–65%6, 7, 69
8p23-p11	Loss	19.58	NKX3-1	53–67%5, 18, 71–73, 75	67–74% of CRPC15, 68, 72	36–90% of AdvDTCs; 20-23% LocDTCs6, 7, 69
9q31.3	Gain	22.79	PTPN3, AKAP2, DAPK1, SYK	5	30%70, 72	25–45% AdvDTC6, 7, 69
10p13	Loss	1.12	ITGA8, PTER, C1QL3, RSU1	18%64, 70, 73	25	6
10q11.21	Loss	0.58	RET, RasGEF1A, HNRNPF, ZNF239, ZNF485, ZNF32	64, 73		55% AdvDTC6
10q22-q24	Loss	24.91	CFLP1, KILLIN, PTEN, RNLS, LIPJ, LIPF, LIPK, LIPN, LIPM, ANKRD22, STAMBPL1, ACTA2	12–25%5, 18, 62, 64, 72–75	36–80%15, 62, 68, 70, 74	36%69
11p13-p12	Loss	4.72		4, 6, 64, 70, 73	25	45% AdvDTC6
12p13	Loss	1.46	BCL2L14, LRP6, MANSC1, LOH12CR1, DUSP16, CREBL2, GPR19, CDKN1B, ETV6	30%5, 18	30–50%4, 64, 70, 72
13q12.3- q14.2	Loss	2.63	HSPH1, B3GALTL, RXFP2, EEF1DP3, FRY, ZAR1L, BRCA2, N4BP2L1, CG030, PDS5B, KL, STARD13, EXOSC8, FAM48A, CSNK1A1L, POSTN, TRPC4, UFM1, FREM2, KBTBD6, KBTBD7, MTRF1, NAA16, OR7E37P, C13ORF15, SPERT, SIAH2, RB1, FOXO1	11–40%5, 18, 62, 64, 71–73, 75	35–95% mets15, 62, 64, 68, 70, 72, 74	21–44% of LocDTC;36–55% AdvDTC6, 69
15q25.1-q26.3	Loss	21.30		64	40% CRPC	20–25%7
16q11.2-q24.3	Loss	33.56	WWOX	33–38%5, 18, 64, 71, 73, 75	57–82%72	33%7
17p13.1	Loss	4.28	RPAIN, AIPL1, XAF, DLG4, PER1, TP53	20–30%5, 18, 64
17p13.3-p11.2	Loss	19.50		30%64, 73	51–61% CRPC
17q21.31	Loss	0.15	DHX8, ETV4	20%18	24
17q24.2-q25.3	Loss	8.90			12–41% CRPC24, 70
18q22.3	Loss	0.29	CBLN2, NETO1	20–25%5, 18, 71, 73	40%24, 72	50%7
21q22.3	Loss	0.25	ERG, NCRNA00114, ETS2, PSMG1, BRWD1, HMGN1, WRB, LCA5L, SH3BGR, C21orf88, B3GALT5, IGSF5, PCP4, DSCAM, C21orf130, BACE2, PLAC4, FAM3B, MX2, MX1, TMPRSS2	33–50%5, 8, 18, 73, 75	33%74

Note: SCNA regions are listed in chromosomal order. Well-characterized cancer genes are in bold. References are indicated for reported frequencies of SCNAs. In general, only SCNAs with a frequency >40% in at least one cancer category are listed.Size is based on reported results, and indicates the broader region of overlap across studies. Actual size reported in individual samples may vary, especially for studies using recently developed technologies such as high-density SNP CGH arrays and next-generation sequencing that permit a greater limit of resolution. In general SCNAs are smaller in primary tumors than those observed in metastases, and may only cover a portion of the region listed.

Clinically, detection of prostate SCNA from alternative tissue sources is of great current interest, as the success rate for prostate biopsy is only 60–70% even with CT guidance. Circulating and disseminated tumor cells (CTC and DTCs) in the blood and bone marrow present an opportunity for repeated testing. The difficulties lie in their rare numbers and complicated techniques for isolation. Nevertheless, new methods promise new results. Genomic profiling of DTCs from patients with advanced disease showed a large number of SCNAs, mostly concordant with corresponding metastases and previous tumors (Table 1),^{6, 7} although DTCs from men with localized disease generally have fewer SCNAs, which may not correspond well with the primary tumor SCNAs.

Structural rearrangements

Double-stranded breaks can occur when DNA unwinds during replication or transcription. Improper repair of these breaks can result in intra- and inter-chromosome rearrangement. Almost 50% of all primary prostate tumors have TMPRSS2:ERG rearrangement, which places the growth-promoting activity of the ERG oncogene under the control of the regulatory elements of androgen-responsive TMPRSS2 ⁸. Rearrangements can also result in new fusion proteins that are constitutively active or have altered function or cellular localization, as in the example of ESRP1:CRAF rearrangement ³. Several other rearrangements have been described for prostate cancer, including other ETS family rearrangements ^{9, 10}, and RAF kinase gene fusions ¹¹ as reviewed previously ¹².

Although ERG rearrangement does not affect the overall frequency of SCNAs, it is associated with deletions of 10q, 17p and 3p14 ⁵. These tumors have a distinct expression signature^{8, 13}. Tumors without ERG rearrangement are significantly enriched for 6q deletion, 7q gain, and 16q deletion⁵.

Paired-end whole genome sequencing suggests that rearrangements are much more common and complex than previously appreciated, and implies the importance of surrounding chromatin structure^{12, 14}. Sequencing of primary tumors from ‘high-risk’ prostate cancer patients showed a median of 90 rearrangements, often complex, per tumor genome. Moreover, breakpoints in TMPRSS2:ERG rearranged tumors were precise and located in accessible chromatin that was enriched in transcription factors associated with androgen-regulated transcription¹⁴. In contrast, in tumors without TMPRSS2:ERG rearrangement, breakpoints were located in transcriptionally-repressed chromatin.

Point mutations

Primary prostate cancer has a somatic mutation rate of 1~2×10⁻⁶, similar to breast, renal and ovarian cancers ^15–17. Although several thousand mutations may exist in each prostate tumor genome, only ~20 per genome are likely to impact protein stability or function. However, mutation of the DNA mismatch repair enzyme MSH6 is associated with a hypermutator phenotype ^{5, 17–19}, resulting in 25-fold more mutations than normally seen in prostate cancer. Mutations of common tumor suppressor genes, including TP53, PTEN, RB1 and PIK3CA, have also been defined in prostate cancer ^{15, 18, 20, 21}, as have activating mutations in the oncogenes KRAS and BRAF. Additional recurrent mutations are detected in factors that mediate AR function, chromatin modification and transcription. These are detailed below.

A new molecular subtype of prostate cancer has been suggested as defined by SPOP mutations ^{15, 18} (Figure 1). Point mutations at evolutionarily conserved residues of the substrate-binding cleft of this E3-ubiquitin ligase subunit were identified in up to 13% of primary tumors. SPOP mutations were enriched in tumors with somatic deletions of 5q21 and 6q21, which encode genes including the chromatin-modifying enzyme CHD1 and the tumor suppressor PRDM1 and FOXO3. But these tumors did not display ETS-rearrangement or mutations in TP53, PTEN, and PIK3CA. SPOP mutations have recently been shown to influence the stability of the SRC3/NCOA3 protein and alter AR signaling in prostate cancer cells²².

Figure 1. Relationships of common genomic alterations identified in localized and advanced in prostate cancer.

Approximately 50% of both primary and advanced prostate cancer harbor ERG gene fusion or other ETS-family gene rearrangement. Primary prostate cancers without ERG rearrangement can be subclassified based on SPINK1 overexpression¹⁰, SPOP mutations, and select somatic copy number aberrations (SCNAs) which are often mutually exclusive. The size of the pie chart pieces on the left represents approximate frequencies of each subgroup. Nearly all advanced stage prostate cancers have amplification or mutation of AR, or abnormalities of other AR pathway components. However, the genetic complexity associated with most advanced stage prostate cancers precludes their classification into distinct subgroups based on genetic profiles. Among this complexity exists a number of commonly observed genetic aberrations as shown. Metastatic tumors may display some or all of these aberrations. Therefore, the pie chart pieces on the right simply help define functional groups of various genes with mutations observed in advanced prostate cancer, without indication for frequencies or mutual exclusivity. See text for detail.

(Abbreviations: Onc/TSG refers to oncogenes and tumor suppressor genes)

Integrating genetic information to identify novel therapeutic targets

As the spectrum of genetic aberrations becomes increasingly more complex in prostate cancer, integrated analysis of genetic aberrations, epigenetics, transcriptional regulation and expression profiles is necessary to understand the molecular pathways that contribute to tumorigenesis. Results from such integrated approaches are now poised to define key targets for future prostate cancer therapeutics.

Androgen signaling pathway

Because the growth of prostate cancer is largely dependent on androgens, therapies blocking the AR signaling pathway are effective for most patients. However, several mechanisms can restore AR signaling and promote the development of castration-resistant metastatic disease (CRPC). These mechanisms include AR amplification, gain-of-function AR mutations, splice variants, and overexpression of AR or its coactivators. AR amplification is observed in metastases from ~50% of patients, and can occur through focal amplification ²³ or through gain of the entire X-chromosome, on which AR resides ^{5, 24}. AR is also frequently mutated in advanced disease ^{5, 15, 25}. The oncogenic H874Y AR mutation increases the binding affinity of AR for testosterone ²⁶. Additional mutations in the ligand-binding domain (K580R, T877A, L701H and V715M) permit inappropriate AR activation by other steroid hormones such as estrogens, progestin, and glucocorticoids ²⁷. A new AR mutation, F876L, confers resistance to the potent AR antagonist, MDV3100²⁸, attesting to the plasticity of the prostate cancer genome in responding to selective therapeutic pressures.

Beyond AR itself, other components of the AR signaling pathway are altered in up to half of primaries and nearly all metastases, indicating the critical nature of this pathway to prostate cancer at all developmental stages ⁵. The oncogenic transcriptional coactivator NCOA2, on 8q13.3, is amplified in 24% of metastases and 1.9% of primary tumors, and correlates with elevated NCOA2 transcripts. Overexpression of NCOA2 primes AR to respond to reduced androgen levels and boosts the total magnitude of AR transcriptional response. Mutations in the Ser/Thr-rich regulatory domain and the transcriptional activation domain of NCOA2 are also frequent. Clinically, NCOA2 may be an important AR-pathway biomarker in primary prostate cancer as noncastrate patients who had NCOA2 alterations showed significantly more recurrences ²⁹.

Androgen signaling can also promote co-recruitment of AR with topoisomerase II beta (TOP2B) to DNA, resulting in TOP2B-mediated double-stranded breaks and rearrangements, including TMPRSS2-ERG ³⁰. Furthermore, in response to genotoxic stress - as may be experienced by cells during radiation or other anti-cancer therapeutics - AR recruits the enzymes AID (activation-induced cytidine deaminase) and the LINE-1 encoded ORF2 endonuclease ³¹, which may also contribute to formation of rearrangements. Given that chromosomes form three-dimensional ‘transcriptional hubs’ that simultaneously coordinate the chromatin structure and transcriptional activity of multiple genes ³², ‘transcriptional hubs’ may facilitate rearrangements of genes that spatially co-localize when errors in DNA processing cause DNA breaks in response to androgen signaling ³³. These new factors warrant further investigation.

Transcription factors and chromatin modifiers

Transcription factors and chromatin modifiers work cohesively to mediate sequence-specific chromatin modifications that regulate gene expression. They play important roles in embryonic stem cells, cellular differentiation, and are altered in many cancer types, including prostate cancer. Given the potentially broad effects of alterations in chromatin structure and transcriptional regulation, disruptions in these genes may have extensive effects. These global changes may contribute to the high degree of genetic instability that is characteristic of metastatic prostate cancer, perhaps through increasing accessibility of DNA to factors that induce DNA breaks or by altering three-dimensional chromatin structures and interactions. Most notable are recurrent aberrations in the chromodomain-helicase-DNA-binding (CHD) proteins and FOXA and FOXO transcription factor families. Nevertheless, their contribution to prostate cancer oncogenesis is not fully understood to date.

Alterations in CHD genes have been commonly detected in primary and metastatic prostate cancer, and may distinguish a new subgroup of prostate cancers with increased aggressiveness. CHD proteins function in ATP-dependent chromatin remodeling. CHD1 on 5q21 is disrupted by focal deletions or mutations in up to 17% of all prostate tumors^{14, 15, 34}. Intragenic breakpoints in CHD1 yield truncated proteins¹⁴. Knockdown of CHD1 in prostate cancer cell lines has been associated with morphological changes and increased cell invasiveness³⁴. CHD1-deficiency is strongly correlated with lack of ETS-family gene rearrangements, and may represent a novel subclass of aggressive prostate cancer ^{15, 18}. Another CHD protein, CHD5, is a tumor suppressor whose expression is altered in several solid tumor types by focal deletions, mutations or DNA methylation. CHD5 mutations have been detected in multiple prostate tumors ²⁵. Loss of CHD5 correlates with increased proliferation and decreased apoptosis via the p53 pathway^35–37. Finally, the H3/K4-specific methyltransferase gene MLL2 is also frequently mutated in prostate cancer, as has previously been seen in other cancer types^{15, 38, 39}.

The Fork-head box protein A (FOXA) and O (FOXO) families belong to the larger group of highly conserved forkhead proteins, which are deregulated in several tumor types ⁴⁰. FOXO and FOXA members are transcription factors that bind to AR and regulate its association with androgen response elements ^{15, 41}. However, their roles in prostate cancer appear to be diverse based on frequent amplification or activation of FOXA1 but loss of FOXO in tumors. FOXA1 is required for prostate epithelial cell differentiation and promotes proliferation⁴². Focal amplifications of FOXA1 and mutations in the transactivation and DNA-binding forkhead domains have been reported in ~10% of prostate cancer cases^{5, 15, 18, 25}. Increased expression of FOXA1 correlates with PSA and Gleason score, and is associated with biochemical recurrence and poor prognosis ⁴³. FOXA1 likely functions by repressing AR signaling, therefore leading to dedifferentiated tumors that are more aggressive and have a higher risk for metastatic relapse. In prostate cancer cell lines, increased FOXA1 activity promotes proliferation, tumorigenesis and xenograft growth¹⁵.

FOXO proteins have tumor suppressor activity, and control the transcription of genes involved in metabolism, stress response, cell cycle arrest, cell death and DNA damage repair. In prostate cancer, FOXO1 and FOXO3A are inhibited by AKT-mediated phosphorylation, resulting in their nuclear exclusion and ubiquitin-mediated degradation⁴⁴. Deletion of FOXO1 on 13q14 has been observed in approximately one-third of prostate cancer cell lines, primary tumors and xenografts⁴⁵. Loss of FOXO1 increases the basal activity of AR and sensitizes it to lower androgen levels or other nonandrogenic ligands⁴⁶. FOXO1 also inhibits the transcriptional activity of Runx2, a transcription factor that contributes to prostate cancer cell migration, invasion and metastasis⁴⁷. Restoration of FOXO3A activity in cancer cell lines sensitizes them to radiation⁴⁸, suggesting that combination of radiation with therapies that increase FOXO3A activity might be beneficial.

Phosphatidylinositol 3-kinase (PI3K) and AKTsignaling pathways

PI3K pathway is a critical regulator of cell survival and proliferation. In prostate cancer, aberrant activation of the PI3K pathway is associated with higher Gleason grades, earlier recurrence, and a higher risk of cancer-specific mortality⁴⁹. Up to 50% primary prostate tumors and 100% metastases have aberrant expression, SCNAs or mutations in PI3K pathway members.

The major route for PI3K pathway activation is via the loss of PTEN’s tumor suppressor function as a result of PTEN copy number loss, inversions or mutation^{5, 14, 25}. Focal deletion of PTEN or aneuploidy of chromosome 10 is present in nearly half of primary prostate cancers and all metastatic tumors^{5, 50}, while inactivating mutations of PTEN account for 5–10% of primary cancers¹⁸. In addition, loss-of-function mutations and rearrangements in a PTEN-associating protein, MAGI (membrane-associated guanylate kinase inverted) have been detected in prostate cancer ¹⁴. MAGI enhances PTEN’s ability to suppress AKT activation. Besides PTEN, several other phosphatases can regulate AKT activity, including PHLPP and INPP4B. Deletion or loss of their expression is correlated with greater Gleason score and shorter time to biological recurrence ^{51, 52}.

Activation of the PI3K pathway can also occur following oncogenic activation or amplification of PIK3CA, AKT1 and MTOR²⁵. PIK3CA on 3q26.32 encodes the catalytic subunit of PI3K. Amplification of PIK3CA has been detected in 13–39% of primary tumors and up to 50% of CRPCs^{53, 54}. Activating mutations in PIK3CA and MTOR have also been detected in ~4% of primary prostate tumors^{18, 25, 54}.

Clonal heterogeneity

The high level of genetic heterogeneity within and across prostate tumor foci likely contributes to a tumor’s ability to develop therapeutic resistance. Pathologically distinct tumor foci have few commonly shared mutations, supporting the largely independent clonal origin of each neoplastic region ⁵⁵. Even among different sites within a tumor focus, unique mutation profiles are observed. These results are concordant with the high degree of intratumoral heterogeneity that has been characterized in breast, kidney and myeloproliferative cancers ^56–59. Meanwhile, the shared pattern of aberrations in patients’ metastatic tumors supports a monoclonal origin of metastasis, and may indicate aberrations that are important to the progression of prostate cancer ^{23, 25}. However, metastases also accumulate SCNAs that are unique, particularly in lymph node, brain and bone metastases ²³. These may either be passenger events that do not influence disease progression, or may result from tissue-specific selection pressures.

Given the high levels of heterogeneity, integrated analysis of biological pathways that are altered in prostate cancer will be critical. Integrated analysis of genetic and transcription data has revealed new pathways in glioblastoma ⁶⁰, and is promising to do so in prostate cancer. Recently, Taylor et al found that three well-known pathways, PI3K, RAS/RAF, and RB1 are altered in 34–43% primary tumors and 74–100% metastases ⁵. While RNA expression could not predict recurrence, they found that DNA copy-number profiling did significantly associate with outcomes. They also identified a separate group of patients whose tumors did not carry any major SCNA or aneuploidy, and who remained largely recurrence-free at five years. Thus, genomic analyses may also identify patients who are good candidates for active surveillance.

Conclusions and future studies

The use of genomic profiling to identify robust subtypes of prostate cancer lags far behind efforts in other cancer types, where the molecular subclassification has improved our clinical ability to predict patients’ overall risk and response to treatments. Examples include: HER2 in breast cancer, BRAF in melanoma, and KRAS and EGFR in lung cancer. Thus, it is imperative that future studies correlate clinical data with molecular and genetic classification of cancer samples. With the high prevalence of aberrations in the AR and PI3K/AKT signaling pathways, additional treatments targeting these signaling programs are in order. Factors that regulate chromatin, epigenetics and transcription have also emerged as highly significant and deserve further investigation⁶¹. Ideally, targeted therapy based on key perturbed pathways, as illustrated in Figure 2, can be tested in prospective clinical trials.

Figure 2. Considerations for targeted therapy based on key pathways perturbed in prostate cancer.

Current standard-of-care involves active surveillance for low-risk localized prostate cancers; hormonal therapy, radical prostatectomy, or radiation therapy for higher-risk localized disease; and androgen pathway suppression for metastatic disease with chemotherapy and immunotherapy at the time of disease progression. This figure shows the potential for targeted therapy in molecularly-defined subtypes of prostate cancer. Genomic alterations are classified based on the class of molecular pathways affected (inner circle). Therapeutic agents (outer circle) targeting respective pathways are grouped with the genes (middle circle) commonly altered in these pathways, coordinated by color wherever possible. Selected agents in various phases of clinical trials are superscripted: ¹FDA approved, ²Phase III clinical trials, ³Phase I/II clinical trials; preclinical development not marked. While anti-androgen therapy abiraterone, microtubule inhibitor cabazitaxel, and immunotherapy sipuleacel-T are already in clinical use, aberrations of NCOA2 and FOXA1 genes (white) are recent findings, the functional significance and therapeutic implications of which await further investigation.

An underlying question is how SCNAs, particularly heterozygous deletions and low-copy-number amplifications, affect expression of the genes located within the affected region and cause indirect effects on other genes. Kim et al show a modest correlation of copy number with gene expression; ~38% of amplified genes had concordant increases in expression ⁶². The area of copy-neutral LOH also warrants further attention, which can only be detected through next-generation sequencing approaches or by genomic arrays incorporating SNP markers⁶³. Large cnLOH is typically associated with homozygous mutations of gene(s) residing in the respective sequence.

Additional meta-analysis of existing genetic information may help identify aberrations that work synergistically to promote tumorigenesis. In a limited example involving five metastatic tumors, all 19q13.32 losses occur in the presence of 1p22.1 loss, whereas 17q21.31 loss concurs with 18q22.3 loss, and 21q22.3 loss with 16q23.1 loss ⁶⁴. Results such as these point to common regulation, such as through colocalization in three-dimensional space.

An important question that must be addressed centers on the molecular heterogeneity within and between primary prostate cancer foci and discrete metastasis. Developing approaches to assess distinct clones will have important implications for anticipating response and resistance to targeted therapeutics. Further, sampling multiple metastatic sites for genomic analyses poses technical and safety challenges. Enumeration of CTCs and DTCs has been shown to predict risk of relapse and quantifies patients’ treatment responses ^{65, 66}. Building on these assessments of CTC numbers, technological advances now allow for the direct molecular profiling of these populations on a single-cell basis. Results such as these could provide a view of the heterogeneity of a patient’s tumor burden, and has the advantage of resampling over the course of disease. Direct sequencing of circulating cell-free DNA offers another avenue for identifying and monitoring genomic alterations that could influence therapy selection⁶⁷.

In closing, rapidly expanding technologies and declining costs for genomic analysis are providing insights into the genetic underpinnings of prostate cancer at a rate faster than ever before. As additional studies are undertaken and new gene candidates emerge, putative driver events will be evaluated as therapeutic targets. With more novel therapies tested and approved, determining the best approach to handle genetic heterogeneity among patients will be a top research priority.