Background
Metastatic prostate cancer is the 2nd-leading cause of cancer mortality among American men [
1] and is associated with significant morbidity in addition to loss of life. Androgen deprivation and other methods to disrupt androgen receptor (AR) signaling were the first targeted cancer therapy and have been the mainstay of prostate cancer therapy for more than seven decades [
2]. Two recently approved therapies for advanced prostate cancer directly target the androgen receptor signaling pathway: abiraterone by preventing androgen synthesis, and enzalutamide by blocking interaction of the receptor with its ligands. There is even evidence that docetaxel, the first approved systemic therapy with a demonstrated survival benefit, acts by disrupting AR function [
3].
Though most men with prostate cancer initially respond to these AR-targeted therapies, resistance is inevitable. As with other targeted therapies, alteration or modification of the target is a likely resistance mechanism and a number of such mechanisms have already been described including upregulation of AR expression, as well as mutation and deletion of the AR ligand binding domain [
4]-[
9]. Nevertheless, translational studies, including assessment of clinical relevance of various AR alterations, have been hampered by a lack of access to tissue. Prostate cancer typically metastasizes to bone, and biopsy of these lesions is invasive, painful, and low yield [
10]. A few programs have instituted warm autopsy programs, collecting samples of metastatic disease shortly after willing patients die from their disease [
11]. These programs are rare, however, and offer only a final snapshot of the biology of resistant disease.
Circulating tumor cells (CTCs) are emerging as a source of therapy-resistant prostate cancer material, a so-called "liquid biopsy" that could serve as a therapeutic biomarker [
12],[
13]. CTCs can be obtained through relatively non-invasive means, opening the door to serial assessments of the disease state and examine tumor cell response to therapy. In 2008 the FDA approved the use of the CELLSEARCH system (Janssen Diagnostics) to collect and enumerate CTCs in patients with prostate cancer [
14]. With this system, there is a clear relationship between the number of CTC’s and disease prognosis. Additional methods for CTC isolation have been developed, including microfluidic techniques [
15], RT-PCR detection [
16], fluorescence in situ hybridization (FISH) [
17], and enrichment based approaches using other magnetic devices [
18]-[
21], lipid content [
22], size [
23],[
24], or charge. Many of these technologies are confounded by white blood cell (WBC) contamination, are often dependent on fixed cells, have focused on CTC enumeration, and have not necessarily been conducive to molecular interrogation of isolated cells.
We sought to develop and utilize techniques that allow flexibility in translational application. The eventual goal of our research is to provide support for CTC protein interrogation as a predictive therapeutic biomarker for the personalization of mCRPC care. As AR is the pivotal molecular driver of prostate cancer evolution and progression and is the foremost therapeutic target for mCRPC, our current study focused on AR interrogation. To that end we used fluorescence activated cell sorting (FACS)-based methods allowing for protein analysis and collection of live cells. We focused our approach using ImageStreamX, which combines microscopy techniques with flow cytometry. ImageStreamX captures each cellular event in a digital photograph and calculates multiple parameters, including fluorescence intensity and location, within each captured image [
25]-[
27]. This flexibility and specificity allowed an integrated evaluation of AR within CTCs in patients with metastatic castration-resistant prostate cancer (mCRPC).
Methods
Patients
All patients were treated at the University of Chicago for mCRPC and provided informed consent per an Institutional Review Board-approved prospective clinical protocol. All patients were progressing on their current therapy by PSA or radiologic criteria. Blood (15 mL) was collected and processed within 2 hours, including isolation of mononuclear cell fraction (buffy coat). Two cohorts were evaluated: our initial cohort of FACS-sorted CTCs evaluated by immunofluorescent microscopy, and a second cohort of 20 patients with CTCs evaluated by ImageStreamX.
Isolation of CTCs by FACS
Patient blood was drawn into blood collection tubes (BD: 362753). Isolated mononuclear cells were stained with either an Alexa-488 conjugated EpCAM antibody (Biolegend, 1:100) or a PE-conjugated EpCAM antibody (Biolegend, 1:100) and a QDot800 conjugated CD45 antibody (Invitrogen, 1:100). CTCs were isolated using a MoFlo XDP flow-sorting machine to sort for EpCAM positive/CD45 negative cells.
Protein expression by immunofluoresence (IF)
Isolated CTCs that were utilized for protein expression analysis via immunofluoresence (IF) were sorted into 8-chambered slides (LAB-TEK: 154941). CTC cells were stained with combinations of the following primary antibodies overnight and secondary antibodies: Androgen Receptor (Santa Cruz, N-20 antibody, 1:100) or an Alexa-647 conjugated AR antibody (Cell Signaling, AR-XP, 1:100), PSA (Santa Cruz 7638 1:50), PE conjugated pan-cytokeratin (Santa Cruz 8018, 1:50), DyLight 680 anti-rabbit secondary (KPL, 1:100), Dylight 594 anti-goat (KPL, 1:100). DAPI containing mounting media was used to coverslip slides (Vector laboratories: H1200). Slides were visualized with the Leica SP2 laser scanning confocal microscope and images were analyzed using ImageJ software.
Targeted DNA sequencing
To extract and amplify DNA, sorted cells were collected into PCR tubes containing 3 μl PBS and DNA extracted using REPLI-g (Qiagen) following the manufacturer’s instructions for amplification from single cells. To sequence a portion of the AR gene, primers were used to amplify exons 7 and 8, and the forward primer used for capillary sequencing. (AR intron7 F GAGGCCACCTCCTTGTCAACCCTG and AR 3UTR R2 GGCACTGCAGAGGAGTAGTGCAGA).
ImageStreamX analysis
Analysis of CTCs from patients with mCRPC
Buffy coat obtained from patients with mCRPC for ImageStreamX analysis was centrifuged at 450 G for 10 minutes and washed with PBS. Cells were blocked using FcR Blocking Reagent (Miltenyi Biotec, 1:10) for 10 minutes. One μl of Biotin anti-human CD45 (Biolegend, 1:100) per 5 × 10
6 cells was added, incubated for 20 minutes, washed with PBS, and centrifuged for 10 minutes at 450 G. This was repeated twice. Anti-biotin microbeads (Miltenyi, 1:4) were added to cells at a 1:4 dilution per 10
7 cells for 15 min, washed with PBS, and centrifuged at 450 G. Cells were re-suspended in 500 μl PBS and CD45 depleted using AutoMACS Pro (Miltenyi) per the manufacturer’s instructions. Following depletion, cells were centrifuged for 10 minutes at 450 G, washed and fixed for 15 minutes with 3.2% Ultra Pure EM Grade Formaldehyde (Polysciences, Inc., 1:6), and stored at 4C for up to three months (average storage time was 4 weeks). Staining was performed using EpCAM (Biolegend;1:40), and CD45 (Life Technologies; 1:27) antibodies in the dark for 30 minutes. For intracellular staining, cells were then washed with PBS. Fix/perm buffer from FoxP3 Buffer Set (eBioscience, 1:4) was added to cells for 30 minutes then washed off with PBS. Next, cells were washed with Perm Buffer from FoxP3 Buffer Set (eBioscience, 1:10) and centrifuged at 450 G for 5 minutes. Cells were stained intracellularly with AR (Cell Signaling, 1:11), and Ki-67 (Biolegend, 1:11) in the dark for one hour. Cells were washed with Perm Buffer and then PBS. Finally, cells were stained with FxCycle Violet (Invitrogen, 1:1000) and acquired on ImageStreamX (Amnis). Single stain control with gaiting strategy for each antibody is shown in Additional file
1: Figure S1. Gating strategies and multi-marker compensation were maintained. Analysis was conducted using IDEAS software (Amnis).
Comparison of staining fixed versus unfixed cells
One million LAPC-4 cells were fixed for 15 minutes with 3.2% Ultra Pure EM Grade Formaldehyde (Polysciences, Inc.). Fixed cells and an equal number of unfixed LAPC-4 cells were washed with 1× PBS and centrifuged for 5 minutes at 450 G. Following centrifugation, cells were stained with EpCAM (Biolegend, 1:40) for 30 minutes at 4C then washed with PBS. Both fixed and unfixed cells were analyzed via flow cytometry. FACSDiVa Software was used for data acquisition and FlowJo software was used for analysis.
Determination of Nuclear versus Cytoplasmic AR localization
LAPC-4 cells were cultured in serum starved media. On day 1, either 1 μM R1881 or 10 μM of enzalutamide was added to the cells. On day 2 cells were intracellularly stained for AR as described above.
Single stain compensation controls
VCAP or CWR22RV1 cells were fixed for 15 minutes with 3.2% Ultra Pure EM Grade Formaldehyde (Polysciences, Inc.). Cells were washed with PBS, centrifuged for 5 minutes, and stained for EpCAM (Biolegend, 1:40), CD45 (Life Technologies, 1:27), AR (Cell Signaling, XP conjugated antibody, 1:11), Ki-67 (Biolegend, 1:11) and FxCycle Violet (Invitrogen, 1:1000). Cells were acquired using ImageStreamX (Amnis; Seattle, WA) and analyzed using IDEAS software (Amnis; Seattle, WA).
Evaluation of cell proliferation
A mitomycin dose response curve was performed on CWR22RV1 between 1 ug/ml to 8 ug/ml, and cells were incubated for 4 hours. The media was replaced, and the cells were incubated overnight. Cells were stained for Ki-67 and analyzed by ImageStreamX.
Statistical analyses
For our initial feasibility cohort, a detection rate of 35% was postulated as the minimum threshold of feasibility given the reported detection rate for CTCs between 50 and 70%. Thus, a minimum of six of the first fifteen analyzed patients needed to have isolated and visualized CTCs to satisfy our feasibility threshold and allow the study to continue.
A student’s t test was used to analyze the difference in Ki-67 expression following exposure to Mitomycin. A Wilcoxon signed-rank test was used to analyze the association between Ki-67 and AR expression, and between Ki-67 and similarity index. A Wilcoxon rank sum test was used to analyze the association between AR expression and prior exposure to abiraterone, and between similarity index and prior exposure to abiraterone. A mixed model with a random patient effect was used to analyze the difference in area between EpCAM+ and EpCAM- cells.
Discussion
A major unmet need in personalization of prostate cancer care is the identification and validation of therapeutic biomarkers. One major limitation in our ability to develop biomarkers for mCRPC therapies is the difficulty in obtaining patient tumor specimens that are reflective of their current disease biology. CTC interrogation offers a potential source of such biomaterial; however, evaluation of CTCs is logistically difficult, and to date, there are no established methods for the study of CTCs at the molecular level. The principal goal of this study was to develop a novel, logistically feasible and robust method for interrogating CTCs from patients with progressive CRPC. Given the central importance of AR biology in prostate cancer therapeutics, our focus was on utilizing these techniques to study AR within CTCs.
We demonstrate the feasibility of two different flow cytometry-based techniques for the evaluation of CTCs from blood of men with progressive mCRPC. Our approach is streamlined and uses equipment that is widely available. Furthermore, fixation of cells for ImageStreamX allowed for short-and-long term storage of cells, facilitating greater experimental flexibility, and templates used during sample acquisition and data analysis were able to be standardized. The specificity of the cells isolated and evaluated was confirmed subsequent to FACS CTC isolation. Direct visualization and staining of these events confirmed that these were cells that expressed EpCAM, PSA and AR, but not CD45, and contained a nucleus.
The focus of the majority of CTC evaluation to date has been enumeration. While this is informative and prognostic [
36], CTC isolation with molecular profiling of advanced disease has the potential to advance the personalization of CRPC therapy. Clinical trials are beginning to tailor therapy to specific molecular subtypes, and the analysis of CTCs will be a vital tool in identifying which therapies are most appropriate for each patient.
To that end, our goal was to show feasibility of isolation and molecular profiling of CTCs from patients with mCRPC. Given the pivotal role of AR, we focused our analysis on the expression and subcellular localization of AR. We found that AR expression and localization is heterogeneous between patients, and even within a patient. The finding that patients have CTCs with pleiotropic AR expression and localization is novel and may suggest variable or incomplete efficacy of AR-targeted therapies for such patients. Whether specific clones are selected under the therapeutic pressure of specific agents is an obvious avenue of further investigation.
With our quantitative analysis of AR protein expression using the ImageStreamX platform, our results could have implications for future clinical research with embedded CTC biomarker interrogation. We initially hypothesized that after failure of abiraterone, patients would have less active AR (demonstrated by less AR expression or nuclear localization) supporting clinical progression through a non-AR directed mechanism. In fact, our data support that patients with prior exposure to abiraterone have higher AR protein expression and statistically unchanged AR nuclear localization compared to patients who were not exposed to abiraterone. This parallels previously reported xenograft models showing increased AR expression in prostate cancer after abiraterone [
37],[
38]. Our data suggest that combining abiraterone with an AR antagonist such as enzalutamide may be effective in those patients, and especially in patients who still have nuclear, highly expressed AR. A large national phase III clinical trial of enzalutamide and abiraterone in combination is underway testing this hypothesis (NCT01949337). We also show, for the first time, that CTCs with nuclear AR have more Ki-67 staining compared to CTCs with cytoplasmic AR, implying that cells with nuclear AR are more proliferative. These results confirm that AR remains important in mCRPC patients despite clinical attempts to disrupt AR signaling. Thus, use of AR expression and localization in CTCs to guide management decisions warrants further investigation.
The parameters used to identify cells of interest are of concern when identifying rare events. We chose to focus on EpCAM + cells due to the wealth of experience supporting these selection criteria. It may well be, however, that there are additional markers which could allow identification/isolation of CTCs [
39]-[
41], which we are not collecting/isolating with our current techniques. One major benefit of the ImageStreamX multiplex protein interrogation approach is its flexibility; as our understanding of CTC biology becomes more refined, one could easily replace EpCAM-specific antibodies with other targeted markers. Of note, we did identify EpCAM-/AR + cells with ImageStreamX. These were noted to be smaller than the EpCAM + cells as a whole, leading to the presumption that the majority of these events are white blood cells that escaped negative selection. However, it is possible that they are EpCAM-low CTCs, and further evaluation of this population in particular is needed.
The relatively small sample size of this feasibility study is a limitation. A larger prospective study in a homogeneous patient population is needed to validate quantitative AR protein characterization within CTCs as a therapeutic biomarker. Fixed cells analyzed by ImageStreamX are not suitable for further interrogation of DNA or RNA expression. This is a limitation of our protein-based analysis platform. Of note, our initial FACS-based approach does allow for DNA interrogation with specificity, although this was not the focus of this current study.
There are many considerations when developing novel techniques for isolating and interrogating CTCs as potential biomarkers [
13]. One primary question that remains largely unanswered is to what extent CTCs are representative of the metastatic disease. The fact that CTC enumeration is prognostic for CRPC suggests a link between CTC and metastatic disease burden, but to what extent the CTCs are in fact part of the metastatic process is not clear. One small study of prostate cancer patients showed a lymph node metastasis was phylogenetically more closely related to the CTCs and to one particular focus of the primary prostate tumor than other prostate tumor foci [
42]. The field of CTC research is still in its relative infancy and new molecular techniques will enable further characterization of the relationship biologically between CTC and metastasis.
Other groups have performed highly elegant experiments examining CTC genomics and expression for other diseases [
43],[
42]. In CRPC, reports of CTC molecular interrogation are limited, but have included proteins associated with epithelial to mesenchymal transition and genomic changes such as loss of PTEN or TMPRSS-ERG translocations [
39],[
41]. IHC methods have also been described to ascertain AR expression in CTCs [
44]. Furthermore, in a recent study, single-cell immunofluorescence was used to suggest whether CTCs had active or suppressed AR signaling, determined by PSA and PSMA, respectively, in untreated and CRPC patients [
12]. Evaluation of AR isoform expression may have clinical value. A recent study showed that expression of the AR variant AR-v7 in CTCs may be associated with primary resistance to abiraterone and enzalutamide [
45].
ImageStreamX analysis could easily be adapted to evaluate for expression of AR-v7 using anti-AR-v7 antibodies [
8]. Of note, the AR antibody used in the present study is raised against a relatively N-terminal peptide, which is expected to bind to common truncated splice variants, including AR-V7. Mutation in the AR has also been demonstrated in mCRPC [
46],[
47]. These mutation or ligand independent splice variations have been shown to lead to enhanced activation of the AR. Although our protein-centric methods do not allow for the detection of these specific genetic or mRNA events, ImageStreamX is able to quantify nuclear localization of the AR, which is a surrogate for active nuclear hormone receptor. Other groups have also published reports suggesting AR subcellular localization in CTCs has clinical value. In their study, Darshan et. al found that cytoplasmic AR status in CTCs correlated with favorable clinical response to taxane chemotherapy [
48]. Our methods complement and expand on these studies utilizing quantitative methodologies to interrogate AR protein expression in CTCs.
In summary, our focus on AR protein analysis in CTCs has demonstrated that interrogation of AR expression and subcellular localization may have clinical relevance. Alterations in these metrics are associated with progression on abiraterone and increased expression of a marker of cellular proliferation. Given the continued focus on AR as a therapeutic target in mCRPC, quantitative analysis of AR expression and nuclear localization has the potential to serve as a predictive therapeutic biomarker. Further validation of our techniques, in the context of prospective clinical trials is warranted. As our ability to personalize prostate cancer management improves, streamlined CTC isolation and molecular profiling will become an integral piece in the management of advanced prostate cancer. Although we focus on AR expression for the purpose of this study involving patients with mCRPC, our quantitative multiplexing methods are adaptable for other pathways of interest within mCRPC and potentially broadly applicable to other metastatic malignancies.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Competing interests
The authors declare that they have no competing interests.