Development and Application of Liquid Biopsies in Metastatic Prostate Cancer

Abstract

Purpose of Review

Metastatic prostate cancer is a lethal and highly heterogeneous malignancy, associated with a broad spectrum of potentially actionable molecular alterations. In the past decade, disease profiling has expanded to include not only traditional tumor tissue, but also liquid biopsies of cells and genetic material circulating in the blood. These liquid biopsies offer a minimally invasive, repeatable source of tumor material for longitudinal disease profiling but also raise new technical and biological challenges. Here we will summarize recent advances in liquid biopsy strategies and the role they have played in biomarker development and disease management.

Recent Findings

Technologies for analysis of circulating tumor cells (CTCs) continue to evolve rapidly, and the latest high content scanning platforms have underscored the phenotypic heterogeneity of CTC populations. Among liquid biopsies, CTC enumeration remains the most extensively validated prognostic marker to date, but other clinically relevant phenotypes like androgen receptor (AR) localization or presence of AR-V7 splice variant are important new predictors of therapy response. Serial genomic profiling of CTCs or circulating tumor DNA (ctDNA) is helping to define primary and acquired resistance mechanisms and helping to guide patient selection for targeted therapies such as poly(adenosine diphosphate [ADP] ribose) polymerase (PARP) inhibition.

Summary

The era of liquid biopsy-based biomarkers has arrived, driven by powerful new enrichment and analysis techniques. As new blood-based markers are identified, their biological significance as disease drivers must be elucidated to advance new therapeutic strategies, and their clinical impact must be translated through assay standardization, followed by analytical and clinical validation. These efforts, already ongoing on multiple fronts, constitute the critical steps toward more effective precision management of advanced prostate cancer.

Introduction

Prostate cancer is the most prevalent malignancy affecting men in the USA, with an estimated 161,360 new diagnoses and 26,730 deaths expected this year [1]. Androgen deprivation therapy (ADT) has long been the cornerstone of treatment for metastatic disease, and more recently a survival benefit was demonstrated when ADTwas combined with chemotherapy or newer androgen-directed therapies [2–4]. Irrespective of treatment, most patients ultimately progress to castrate-resistant prostate cancer (CRPC), and optimally choosing among the recently approved therapeutic options in this disease has driven the search for new biomarkers. Next-generation sequencing (NGS) of primary tumors and metastases has begun to elucidate the molecular landscape of prostate cancer and to identify candidate biomarkers and potential disease drivers [5, 6••, 7–9]. These studies have revealed that prostate cancer is characterized by significant intra- and inter-patient heterogeneity, and that somatic alterations accrue over time and with exposure to therapies.

Given this heterogeneity, it is unlikely that tissue from one disease site at one point in time can fully represent the molecular profile of the disease. Yet, obtaining sequential tumor biopsies from multiple metastatic sites at serial time points during the disease course is impracticable for reasons of patient safety, comfort, and cost. An alternative source of tumor material that may represent the full repertoire of alterations present in metastatic disease can be found in a routine blood draw. So-called liquid biopsies are repeatable and minimally invasive and allow real-time monitoring of a patient’s treatment response. Circulating tumor cells (CTCs), cell-free DNA (cfDNA), and more recently extracellular vesicles (EVs) have been utilized as sources of tumor material for study (Fig. 1). In this review, we will summarize the most recent developments for liquid biopsies’ role in advanced prostate cancer with an emphasis on clinical utility for personalized medicine.

Fig. 1.

Schematic of liquid biopsy analytes in prostate cancer and overview of profiling options. EVs, extracellular vesicles; cfDNA, cell-free DNA; CTC, circulating tumor cells; SSNVs, somatic single-nucleotide variants; CNV, copy number variation; WES, whole exome sequencing; WGS, whole genome sequencing; nAR-V7, nuclear androgen-receptor splice variant 7; LN, lymph nodes

Circulating Tumor Cells

CTCs are cancer cells shed into the bloodstream from the primary tumor or metastatic sites and have the potential to provide a non-invasive, repeatable sampling means for genomic, phenotypic, and functional analysis. Originally identified in 1869 by Ashworth [10], their potential is only now being realized due to major technical advances and investment in clinical implementation. A comprehensive survey of CTC technologies is beyond the scope of this review, but all strategies aim to surmount the same central challenge: the scarcity of CTCs relative to the vast numbers of red blood cells and leukocytes in blood. To date, only the automated CellSearch system, originally developed by Immunicon, then by Janssen/J&J and more recently by Menarini Silicon Biosystems, has been FDA cleared for enumeration of CTCs in metastatic prostate cancer [11]. CellSearch enriches and identifies putative CTCs using immunomagnetic beads targeting the cell surface epithelial cell adhesion molecule (EpCAM), followed by immunofluorescent staining for cytokeratins (CK 8, CK 18, CK 19), nuclear DAPI, and the leukocyte cell surface marker CD45. After this automated process, candidate CTCs (EpCAM+CK+DAPI+CD45−) are manually reviewed and verified.

In recent years, this canonical definition of a CTC has been shown to identify many but not all of the phenotypically heterogeneous cancer cells in blood. In particular, non-canonical CTCs can include those undergoing epithelial to mesenchymal transition (EMT) and neuroendocrine-like cells that may express low levels of epithelial markers and thus may be overlooked using the CellSearch criteria. Although the biological and clinical importance of these subpopulations is still being studied, their presence has helped to motivate the development of multiple alternative CTC enrichment platforms that sort CTCs based on their larger size and lower deformability rather than based on cell surface markers (Table 1). More recent approaches have eschewed enrichment altogether, instead using rapid, high-content imaging of all the nucleated blood cells [26]. These instruments employ a “no cell left behind” strategy—red blood cell removal followed by staining and scanning all nucleated blood cells on glass slides.

Table 1.

Examples of recent CTC enrichment and detection platforms

Enrichment type	Platform	Company	Comments	Ref.
Immuno-affinity-positive selection	CellSearch® platform	Menarini Silicon Biosystems	FDA-cleared system for CTC detection; extensive analytic and clinical validation; high reproducibility and specificity; low EpCAM-expressing cells, e.g., neuroendocrine-like or EMT, may not be captured	[11]
	LiquidBiopsy® Platform	Cynvenio	Automated workflow; multi-antibody capture cocktails including PSMA for increased CTC capture; CLIA certified downstream NGS workflow; plasma removal possible for cfDNA analysis; CTC enrichment compatible with targeted sequencing but not WES/WGS	[12]
	AdnaTest	AdnaGen/Qiagen	Simple workflow for CTC capture and gene expression analysis; adaptable downstream targets, e.g., AR-FL and AR-V7; no CTC enumeration data; false-positive signal due to low-level gene expression in contaminating cells	[13]
	VERSA	University of Wisconsin-Madison	Versatile platform; customizable capture antibody; simultaneous analysis of RNA, DNA, and protein; marker-negative cells may not be captured	[14]
Immuno-affinity-negative depletion	RosetteSep™ Depletion Cocktail	Stemcell Technologies	Cost-effective and fast with intact, viable unbound CTCs isolated; customizable; applicable to PDX cancer model; low purity	[15]
Dielectrophoretic enrichment and cell sorting	DEPArray™ NxT	Menarini Silicon Biosystems	Single-cell sorting and recovery for downstream analysis; requires pre-enrichment and offline antibody staining ofCTCs	[16]
	ApoStream®	ApoCell Inc.	Epitope-independent permitting isolation of heterogeneous populations of CTCs; intact, viable unbound CTCs isolated; blood requires preprocessing	[17]
Size and/or deformability	Parsortix™ System	Angle PLC	Epitope-independent permitting isolation of heterogeneous populations of CTCs; intact, viable unbound CTCs isolated; plasma removal possible for cfDNA analysis; no CTC detection platform (enrichment only)	[18]
	ClearCell® FX1 System	Clearbridge BioMedics	Epitope-independent inertial focusing permitting isolation of heterogeneous populations of CTCs; intact, viable, unbound CTCs isolated; high enrichment protocol; RBC lysis is required; no CTC detection platform (enrichment only)	[19]
Filtration-based methods	ISET® Technology	Rarecells	Epitope-independent, filtration-based permitting isolation of heterogeneous populations of CTCs for cytomorphological and immunofluorescent characterization; identifies distinct populations of CTCs compared to CellSearch	[20]
	Parylene-C slot microfilter	University of Southern California/California Institute of Technology	Epitope-independent, filtration-based permitting isolation of heterogeneous populations of CTCs for immunofluorescent analysis; fast, simple, and cost-effective; not amenable to single-cell genomic analysis	[21,22]
Size and immuno-affinity positive/negative	CTC-iChip	Massachusetts General Hospital	Intact, viable unbound CTCs isolated; high detection rates, enrichment of heterogeneous populations of CTCs; not commercially available	[23•]
	GEDI Chip	Cornell University	High detection rates; multi-marker capture; not commercially available	[24]
High-resolution scanning and recovery	AccuCyte® System, CyteFinder®, and CytePicker®	RareCyte	Non-enrichment-based approach, applicable to any cancer type, algorithmic CTC detection; complete workflow for CTC imaging and single-cell picking for downstream molecular analysis; plasma removal possible for cfDNA analysis; processing time (8 slides per 7.5-ml blood sample)	[25]
	Epic™ platform	Epic Sciences	Non-enrichment-based approach, applicable to any cancer type, complete workflow for CTC imaging and single-cell picking for downstream molecular analysis; processing time (16 slides per 7.5-ml blood sample); commercially available as a send out assay but not as an instrument for purchase	[26, 27]

GEDI geometrically enhanced differential immunocapture, CTC circulating tumor cell, cfDNA cell-free DNA, EMTepithelial to mesenchymal transition, PDX patient-derived xenografts, RBC red blood cell, FDA Food and Drug Administration, CLIA clinical laboratory improvement amendments, VERSA versatile exclusion-based rare sample analysis, ISET isolation by size of epithelial/throphoblastic tumor cells

CTC Enumeration

CTC counts have prognostic value in both metastatic hormone-sensitive prostate cancer (mHSPC) and metastatic castrate-resistant prostate cancer (mCRPC). In a small cohort of mHSPC patients, baseline CTC counts correlated with response to androgen deprivation therapy (ADT), with a threshold of ≥ 3 CTC prognostic for poor prostate-specific antigen (PSA) response [28]. In a phase II clinical trial of ADTalone ± cixutumumab (SWOG 0925) in mHSPC, lower baseline CTCs counts were correlated with PSA response [29]. Recently, preliminary data from a phase III trial of ADT ± orteronel, a cyp-17 inhibitor, in mHSPC (SWOG 1216) confirmed that baseline CTC counts were associated with poor prognostic factors such as elevated PSA, presence of bone metastases, and disease severity [30].

In mCRPC, baseline CTC count was prognostic of overall survival (OS), initially using a categorical cut point of favorable (< 5 CTCs per 7.5 ml) vs. unfavorable (≥ 5 CTCs per 7.5 ml) CTC counts [11]. Subsequently, CTC counts were also assessed as a continuous variable, and it was shown that any increase in counts post treatment was associated with a worse prognosis [21, 31••]. Conversely, a study evaluating CTC reduction as a response biomarker in two large clinical mCRPC cohorts (COU-AA-301 and IMMC-38) demonstrated that a 30% decline in CTC counts was associated with a better OS, evaluable as early as 4 weeks post treatment [32]. Utility of a biomarker can be gauged by determining whether it meets the so-called Prentice criteria of surrogacy, where the surrogate endpoint (measured biomarker) correlates with clinical outcome and fully reflects any effect of therapy on clinical outcome [33]. In one recent mCRPC study [31••], combination of CTC counts and lactate dehydrogenase (LDH) levels satisfied Prentice criteria, underscoring the potential clinical utility of CTC counts in this disease state.

CTC enumeration requires further clinical validation as a predictive marker that guides therapy selection. In one early attempt in breast cancer (S0500), CTC counts measured after first-line therapy served as a decision point, wherein patients with persistently elevated counts were randomized to continued therapy vs. change to another chemotherapy [34]. While CTC counts were highly prognostic in this trial, their value as a predictive marker was not shown (i.e., switching chemotherapy based on CTC counts did not improve outcome). Factors that may have contributed to this result included relatively small treatment arms that required a very large treatment effect in order to reach statistical significance. This would be difficult to achieve given the poor prognosis of the cohort overall (metastatic disease that progressed on first-line therapy). In addition, the study utilized categorical cut points (< 5 or > 5 CTCs/7.5 ml) and might have had different results had treatment arm been assigned based on CTC counts as a continuous measure. Such early studies can inform future trial designs, which will hopefully build upon these valuable insights to further explore the predictive utility of CTC enumeration.

CTC Characterization

While early studies aimed to enumerate CTCs, subsequent work has focused increasingly on characterizing the DNA, RNA, or protein profile of enriched cells, either pooled or singly. Initial feasibility studies of CTC characterization investigated prostate cancer-specific markers using techniques such as qPCR and fluorescent in situ hybridization (FISH). TMPRSS2-ERG fusion is particularly prevalent in prostate cancer and in one such study was investigated in CTCs as a biomarker for response to abiraterone in a cohort of CRPC patients [35]. While there was no significant association between CTC TMPRSS2-ERG status and clinical outcome, this study nevertheless provided an important proof of principle that CTCs might be used as apt surrogates for solid biopsies. Concordance with tumor tissue was demonstrated again when PTEN loss identified by FISH in CTCs correlated with PTEN status in matched tumor biopsies [36]. In a phase III clinical trial of first-line docetaxel ± atrasentan in mCRPC (SWOG 0421), a new live CTC telomerase activity assay was investigated in parallel to baseline CTC counts [22]. Among patients with high CTCs (CellSearch counts ≥ 5), elevated telomerase activity was associated with shorter OS [21, 37]. Other markers evaluated in advanced prostate cancer included EMT and neuroendocrine phenotype [38, 39]. Degree of phenotypic heterogeneity, in and of itself, was recently found to be important in advanced prostate cancer, marked by a multitude of non-canonical subtypes like CK-negative cells, small CTCs, and CTC clusters [40]. In an advanced prostate cancer cohort starting second-line therapy, low CTC heterogeneity was associated with better survival with second-line anti-androgen therapy, whereas high heterogeneity was associated with better survival on taxane chemotherapy [41]. The biological underpinnings of these observed clinical associations warrant careful study that may illuminate promising new therapeutic avenues.

CTC Characterization: Androgen Receptor

Persistent AR signaling despite ADT is a major driver of disease progression. Consequently, AR has been extensively studied in CTCs as a potential biomarker in late-stage disease. Genomic profiling of AR in CTCs identified many of the same genetic mutations that had previously been implicated in disease progression from autopsy studies, including the emergence of AR gene amplifications [42, 43]. These proof-of-principle studies highlighted the potential utility of CTCs for serial monitoring for newly acquired AR mutations as a mechanism of treatment resistance.

Beyond genomic profiling, the protein phenotype of AR has been investigated in CTCs as well. Serial monitoring of CTCs identified a subset of mCRPC patients in whom CTC AR was retained predominantly in the cytoplasm, as opposed to the nucleus. This reduction of nuclear translocation and signaling was associated with significantly better response to docetaxel chemotherapy [44]. In another study that used CTC PSA and PSMA protein profiling as a surrogate for AR signaling activity, persistence of high “AR-on” CTCs in the face of abiraterone therapy was associated with worse outcome [45].

More recent CTC studies have focused on AR-V7, a constitutively active AR splice variant lacking the ligand-binding domain. In mCRPC, detection of the AR-V7 transcript by qPCR from EpCAM-enriched CTCs was associated with resistance to enzalutamide and abiraterone but not docetaxel or cabazitaxel [46••, 47]. In a subsequent study, presence of AR-V7-positive CTCs was associated with worse clinical outcome compared with presence of AR-V7-negative CTCs, which was in turn worse than having no CTCs, independent of number of prior lines of hormonal therapy [48]. An alternative approach for AR-V7 detection in CTCs employed an immunofluorescent protein staining assay [49••]. This strategy allowed single-cell resolution of nuclear AR-V7 protein localization in multiple CTC subtypes, including CK-negative cells, which was not possible with a qPCR assay. Using this assay, mCRPC patients with AR-V7-positive CTCs were found to have poor response to second-generation anti-androgens and better OS with taxane chemotherapy. A subsequent re-analysis of the same cohort reinforced the importance of nuclear localization [27]: A clear distinction was observed, where only nuclear-stained AR-V7 CTCs were strongly associated with resistance to second-generation anti-androgens and better OS with taxane therapy. Notably, in these studies only 20% of patients had AR-V7-positive CTCs. Although more exploratory, additional AR splice variants have been identified in CTCs [23•], and one recent study demonstrated that the presence of any AR splice variant was associated with worse outcome [50].

CTC Next-Generation Sequencing: Genomic Analysis

The advent of whole-genome amplification methodologies has enabled high-throughput analysis from picogram quantities of nucleic acid contained within single cells. In a seminal study by Lohr and colleagues, whole exome sequencing (WES) was performed on single CTCs isolated from mCRPC patients [51]. Census-based sequencing (multiple CTC libraries combined for analysis) was used to successfully identify somatic single-nucleotide variants (SSNVs) that had been previously identified in multi-focal primary tumor or metastatic biopsies. In other studies, whole genome sequencing (WGS) was used for low coverage as well as for more comprehensive analysis of genetic aberrations across single CTCs [52, 53]. In one case study analysis, comparative WGS analysis of matched primary and metastatic tumors and CTCs was performed. Increased sequencing depth and coverage enabled detection of both SSNVs and structural variants in CTCs, which had high concordance with matched tumor tissues [53]. Using low-pass analysis, clinically actionable focal copy number alterations were identified such as AR gain and PTEN loss. Furthermore, overall genomic instability was inferred from large-scale state transition analysis (chromosome breaks per 10-Mb regions) [52]. With further clinical validation, this type of high-throughput analysis may be valuable as a predictor of response to therapies targeting DNA repair, such as poly-ADP ribose polymerase (PARP) inhibition.

CTC Next-Generation Sequencing: Transcriptome Analysis

Gene expression has been performed on single CTCs, admixtures of enriched CTCs and leukocytes, or whole unenriched blood, and each approach has inherent advantages and limitations. From a sensitivity and specificity stand point, single-cell expression analysis has obvious advantages: RNA extracted from isolated cells can be analyzed by multiplexed qPCR or RNA sequencing (RNA-Seq) without concerns of contaminating leukocyte signal [54, 55]. In a proof-of-principle study, Cann et al. successfully performed single-cell whole transcriptome analysis on EpCAM-enriched prostate CTCs, identifying prostate-specific gene expression in the majority of CTCs and overexpression of multiple functional pathways when compared to non-malignant prostate biopsy [54]. Another study utilized RNA-Seq to query the enrichment of signaling pathways involved in enzalutamide resistance. Although performed in a relatively small cohort of patients, non-canonical Wnt signaling was implicated in treatment resistance, with significant enrichment of downstream signaling targets [45]. In these studies, CTCs from the same patient clustered together, reflecting greater inter-than intra-patient heterogeneity.

Despite these important advances, single-cell transcriptomic strategies must contend with the challenges of non-uniform amplification of miniscule and often degraded starting material, PCR bias, allelic imbalance, high cost, and technical burden. Even if these challenges are surmounted, the inherent biological heterogeneity of single cells and the technical heterogeneity introduced by amplification and library preparation mandate that a considerable number of live CTCs be captured, extracted, amplified, and analyzed in order to arrive at a meaningful transcriptional phenotype representative of a patient’s underlying tumor biology and predictive of clinical course. These questions and challenges demand further study and will benefit from continued development of more robust and efficient methodologies.

At the opposite extreme, non-enriched whole blood gene expression analysis, using RNA preservative tubes such as PAXgene, has also yielded valuable biomarker signatures. A study of 97 mCRPC patients using an analytically validated five-gene prostate-specific panel demonstrated that RT-PCR detection of two out of five genes was associated with worse OS [56]. Another study, by the same group, validated these results while also highlighting the clinical applicability of this approach compared to CellSearch and AdnaTest [57]. Using a different approach, two prospective discovery studies each identified a unique expression signature in whole blood that was associated with clinical outcome in mCRPC [58, 59]. While these gene panels were prognostic and hence are of potential clinical utility, they consisted of genes expressed in leukocytes which far outnumbered the much rarer CTCs in the PAXgene samples. Consequently, it may be more challenging to parley these clinical associations toward a mechanistic understanding of cancer drivers and new therapeutic strategies.

A third approach pursued with minimal success to date has been to expand prostate CTCs in culture in order to yield a pure and plentiful population of cells for expression profiling. Successful culturing of viable CTCs directly from a mCRPC patient in three-dimensional organoid cultures was accomplished from one sample in one study [60]. However, the field must await the development of more rapid and robust culture protocols that could be used reliably to expand CTCs from most if not all patients. Even when such protocols are developed, the expression profile of cultured populations would have to be compared with that of parental cells (or matched tumor and metastases) in order to demonstrate that their transcriptional phenotypes did not change in culture and still significantly represent that phenotype of the patient’s tumor.

Circulating Cell-Free DNA

Cell-free DNA (cfDNA) refers to all DNA fragments released by cells (malignant and non-malignant) into the circulation, where circulating tumor DNA (ctDNA) refers to the subset of cfDNA that is derived specifically from tumor cells. Currently, the rate and mechanism of release is unknown, but it is generally accepted that ctDNA can originate from multiple metastatic lesions (either actively shed or through apoptotic/necrotic tumor cells), CTCs, and primary tumor [61]. Through a standard blood draw followed by plasma DNA purification, cfDNA can be isolated for genomic profiling using next-generation technologies such as low-coverage WGS or customized targeted sequencing [62•]. At its most basic, the absolute amount of ctDNA and the ratio of ctDNA to cfDNA can be used as a biomarker. In one study by Carreira et al., estimated ctDNA levels in mCRPC patients were correlated with tumor burden and response to abiraterone treatment [63••]. Likewise, in abiraterone-treated mCRPC patients, detection of increasing ctDNA quantities was shown to correlate with overall tumor burden markers (e.g., alkaline phosphatase) and significantly worse clinical outcomes [64••].

Concordance between ctDNA and tumor tissue has been investigated. In one recent study, liquid biopsies were collected contemporaneously with matched biopsies of metastatic sites (lymph node, visceral, bone) from 45 patients with mCRPC [65]. Evaluable ctDNA samples contained all DNA alterations, at similar allele frequencies, found by WES of the metastatic lesions, including potentially actionable alterations in genes such as BRCA2 and AR. Furthermore, unique alterations were identified solely in the ctDNA fraction, suggesting that liquid biopsies may in fact outperform traditional biopsies of single lesions by more fully representing clonal disease evolution and heterogeneity across all metastatic sites [65].

ctDNA offers utility beyond baseline genomic profiling, because as a liquid biopsy it is uniquely suitable for serial analysis throughout the disease course. Emergence of new genomic aberrations can serve as a sign of disease progression, at the same time pointing to new mechanisms of resistance that may be targeted therapeutically. In mCRPC, several studies have focused on genomic alterations in AR before and after treatment with abiraterone. In pretreatment samples, AR copy number gains or point mutations (T878A or L702H) have been identified in up to 45% of patients, and their presence was associated with shorter progression-free survival (PFS) and OS [63••, 64••, 66]. Similarly, in studies of mCRPC patients initiating enzalutamide therapy, patients harboring these same AR aberrations had shorter PFS and OS when compared to those with normal AR [67, 68]. At progression, emergence of AR alterations and other clinically actionable mutations were identified in ctDNA samples, suggestive of acquired resistance, and providing a rationale for novel therapies targeting this pathway [63••, 68, 69•].

Targeting androgen signaling has long been the cornerstone of therapy for mCRPC and hence a focus of ctDNA analysis. However, a more recent strategy has focused on germline and acquired DNA repair defects involving genes such as BRCA1 and 2, ATM, and PALB2 [6••, 70•]. The presence of germline mutations in these genes was associated with a significantly increased risk of metastatic prostate cancer [70•], and acquired somatic mutations also were observed at much higher frequencies in mCRPC which progressed on multiple therapies [6••, 71]. Tumors harboring DNA repair defects, specifically in homologous recombination of double-strand breaks, are susceptible to inhibition of poly(adenosine diphosphate [ADP] ribose) polymerase (PARP), which functions in single-strand break repair. Inhibition of PARP-mediated single-strand break repair, combined with mutations impacting double-strand break repair mechanisms, effectively blocks DNA repair via both single- and double-strand break repair mechanisms, resulting in “synthetic lethality” and cell death [72]. Initial evidence of clinical efficacy was observed in a phase I study of the PARP inhibitor olaparib, which demonstrated efficacy in patients harboring germline BRCA mutations [73]. In a subsequent phase II trial (TOPARP-A), 33% of subjects had defects in homologous recombination genes, and olaparib induced a response in 88% of these patients [74••]. In this trial, a ≥ 50% reduction in cfDNA was associated with improved OS and radiographic PFS at 8 weeks post olaparib treatment [75]. Response also was associated with reduction in the allele frequency of somatic homologous recombination deficiency (HRD) mutations to < 5% after 8 weeks of treatment and regression of germline/loss of heterozygosity (LOH) HRD mutations to allele frequencies ~ 50% [75]. Whole exome sequencing identified mechanisms of acquired resistance with reversal of the initial sensitizing frameshift alteration, with somatic deletions establishing the open reading frame and as such reinstating a functional protein [75]. A subsequent olaparib clinical trial, TOPARP-B, is currently under way and includes only mCRPC patients with alterations in DNA repair genes. In a separate study, mCRPC patients harboring germline BRCA2 mutations, combined with LOH in ctDNA (~ 90% of cases), had significantly worse time to progression and PFS on abiraterone or enzalutamide while PFS on docetaxel was similar, suggesting that cfDNA analysis of this type might be used to prompt an early switch to chemotherapy or PARP inhibition [76].

Extracellular Vesicles

Circulating extracellular vesicles (EVs) are beginning to emerge as a third component of the liquid biopsy profile. This heterogeneous population of cell-derived particles, consisting of exosomes, microvesicles, and large oncosomes, carry biomolecules (nucleic acids, proteins, lipids) that can be assayed for molecular tumor profiling [77, 78]. Methodologies for isolation and characterization of EVs are still evolving and undergoing standardization to allow rapid and reproducible analysis of large numbers of samples. While newer enrichment and quantitation methods are emerging (e.g., NanoSight, Malvern), gradient centrifugation and immunoblotting currently remain the gold standard. Whereas distinctions between tumor and normal host-derived exosomes are difficult to delineate, large oncosomes appear to be a class of EVs that is predominantly secreted by tumor cells and therefore holds great promise for future liquid biopsy studies [79, 80]. Clinical studies in metastatic prostate cancer are limited to date, but a few have been reported. In a small cohort of mCRPC patients, exosome-derived RNAwas isolated and assayed for the presence of AR-V7 prior to initiation of abiraterone or enzalutamide therapy [81]. Identification of AR-V7, in ~ 40% of patients, was associated with a significantly shorter PFS and OS compared with AR-V7-negative patients, consistent with findings from CTC AR-V7 studies. A second study employed global miRNA expression profiling using a previously validated RNA-Seq assay and identified exosomal miR-1290/miR-375 to be correlated with a shorter OS in mCRPC independent of therapy [82]. Interestingly, miRNAs have also been profiled from cell-free plasma samples in mCRPC and were associated with clinical outcomes. Specifically, expression levels of members of the miR-200 and miR-17 family members were significantly associated with response to docetaxel and with OS [83]. The relationship and possible overlap between these cell-free miRNA levels and miRNAs found in EVs depend on the isolation methods employed and will require clarification in future studies.

Conclusion

Over the past decade, multiple seminal studies have demonstrated that enumeration and characterization of CTCs and genomic profiling of cfDNA can provide valuable information about disease burden, prognosis, potential response to therapy, mechanisms of resistance, and alternative therapeutic targets. As these powerful new technologies and assays move toward clinical implementation, several questions and challenges remain to be addressed: (i) Assay development: At its most basic, which component of the liquid biopsy is most informative? CTCs are challenging to enrich and recover, whereas ctDNA is simpler to handle and is usually released from established tumor sites even if few CTCs are available for analysis. On the other hand, CTCs may represent the most aggressive, disseminating disease phenotype, whereas ctDNA may derive from cells that are dying in response to treatment and therefore may be less informative regarding resistant clones. Furthermore, CTCs may yield additional cellular profiles (protein, RNA, live cells assays) beyond genomic phenotypes. More recently, EVs have emerged as a “third horse in the race” and may combine some benefits of both: more plentiful and resilient-like cfDNA, but yielding multiple molecular phenotypes like CTCs. However, much remains to be clarified about EV biology, clinical significance, and analysis techniques. Ultimately, CTCs, cfDNA, and EVs will likely each yield unique and potentially complementary data, and a multi-parametric liquid biopsy approach might prove to be most beneficial. (ii) Analytical and clinical validation: Transition from a newly discovered phenotype to a reliable liquid biomarker requires analytical validation of sensitivity and reproducibility of capture, amplification, and analysis (e.g., predetermined thresholds for variant calls), ultimately in a CLIA setting. These new assays must also reliably correlate with clinical outcomes or therapy response using standard measures like the Prentice criteria of surrogacy. (iii) Clinical utility: Ultimately, even clinically validated liquid biopsy assays must prove to be broadly applicable, safe, and cost-effective, and produce a measurable improvement in real-world disease outcomes. (iv) Biological significance: Beyond their empiric association with clinical outcome, new liquid biomarkers ultimately should lead to novel therapeutic strategies based on new biological insights. Understanding which of the CTCs in a liquid biopsy represent the aggressive, metastasisinitiating cells may lead to new strategies to disrupt cancer dissemination. Likewise, elucidating the mechanism of cellular nucleic acid and vesicle release and identifying distinct subsets of ctDNA and EVs will clarify which tumor cells are represented by these liquid biopsy components. As importantly, cancer phenotypes that emerge over serial liquid biopsies reflect candidate mechanisms of disease resistance and progression, which can serve as a basis for new basic and translational studies in pursuit of these novel therapeutic targets. It is gratifying to witness the rapid advance of liquid biopsies in all of these areas, from development of enabling technologies, to discovery of new phenotypes that illuminate drugable driver mechanisms, to progression of analytically validated CTC and cfDNA assays into prospective clinical validation. These developments raise high hopes of clinical utility in advanced prostate cancer, a field where liquid biopsies may soon play a key role in patient management and therapy selection.