Gene expression analysis of bone metastasis and circulating tumor cells from metastatic castrate-resistant prostate cancer patients

PC3 and LNCaP PCa cells were obtained from ATCC (Manassas, VA) and provided by Dr. Leland Chung (Cedars-Sinai Medical Center, Los Angeles, CA), respectively. Both cell lines were grown in RPMI-1640 with 10 % heat-inactivated fetal bovine serum (FBS). Authentication of the human cell lines used here was verified through short tandem repeat profiling by the Research Technology Support Facility of Michigan State University.

For validation studies, 80 % confluent PC3 and LNCaP monolayers were trypsinized, and diluted in complete culture medium to adjust their cell density to 50 cells/mL. Cell suspensions (100 μL) were plated into the wells of 96-well cell culture plates. Twenty-four hours later, cells in all wells were counted under the microscope, and those with exactly five cells were trypsinized and added to 7.5 mL of blood from healthy volunteer donors drawn into CellSave^® preservative tubes (Janssen Diagnostics, LLC). CellSave^® tubes containing blood spiked with cultured PCa cells, as well as matched non-spiked blood samples serving as negative controls, were processed in triplicate using the semiautomated CellSearch^® Circulating Epithelial Cell Kit (Janssen) in the CellTracks^® Autoprep^® system (Janssen) for cell enumeration at the Biobanking and Correlative Sciences Core at Karmanos Cancer Institute. Briefly, epithelial cells present in peripheral blood were magnetically captured with a ferrofluid-coupled antibody targeting the epithelial cell adhesion molecule (EpCAM), then immunostained with allophycocyanin (APC)-labeled antibodies to the leukocyte marker CD45, phycoerythrin (PE)-labeled antibodies to epithelial markers cytokeratins (CKs) 8, 18, and 19, and stained with the nuclear stain 4,2-diamidino-2-phenylindole-dihydrochloride (DAPI). The sample was transferred automatically to a cartridge in a MagNest, and finally scanned with the semi-automated fluorescence optical system Cell-Tracks Analyzer II^® (Janssen). Objects preselected and displayed by the system in a gallery were defined as CTCs and counted by a trained operator if they were round to oval in shape, 4 µm in size or larger, positive for the epithelial marker (CK-PE) and nuclear stain (DAPI) with at least 50 % overlap between the CK-PE-positive cytoplasm and the nucleus, and negative for the leukocyte marker (CD45-APC). In addition, 5 PC3 or LNCaP cells were spiked into 7.5 mL of blood from healthy volunteer donors drawn into fixative-free K3EDTA tubes (BD Vacutainer^®, Becton–Dickinson) (three experiments performed by three different investigators), to validate the sensitivity of the RNA amplification method used by us (see description below). To this end, the spiked blood was immediately processed using the CellSearch^® Profile Kit (Janssen) and the CellTracks^® Autoprep^® system, and then the tube employed containing a final volume of 900 µL CTC enriched sample was placed in a DynaMag™-15 magnet (Invitrogen) for 10 min. After carefully aspirating the liquid without disturbing the ferrofluid bead pellet concentrated on the tube wall, 1 mL of TRIzol^® (Invitrogen) was added, and the tube vortexed to lyse cells and inactivate nucleases. The lysate was transferred to a 1.5 mL RNase-free Eppendorf tube, and stored at −80 °C until RNA isolation and antisense RNA (aRNA) amplification was performed.

Written informed consent was obtained from 24 mCRPC patients enrolled to participate in the human protocol # 2011-060 (Principal Investigator: Dr. Michael L. Cher), and approved by Karmanos Cancer Institute and Wayne State University Institutional Review Board. Study inclusion criteria were: (a) prior diagnosis of mCRPC characterized by rising PSA level or clinical disease progression despite a castrate level of serum testosterone; (b) clinical decision to start a new anti-cancer systemic therapy, and no treatment with any investigational drug within 2 weeks prior to blood draw/tissue biopsy proposed herein; (c) metastatic deposit visible on an imaging study obtained within 8 weeks prior to initiation of new systemic therapy; (d) radiological evidence of accessibility to a metastatic deposit in bone by computed tomography (CT); (e) performance status of 0–2 by ECOG/Zubrod criteria; (f) Absolute neutrophil count ≥1500 mm³, hemoglobin ≥9.0 g/dL, platelets ≥100,000/mm³; (g) Prothrombin Time and Partial Thromboplastin Time should be < institutional upper limit of normal; (h) presence of one or more CTCs per 7.5 mL, assessed using the Veridex CellSearch^® Profile Kit assay, as explained above. The characteristics of the patients enrolled in this study are described in Table 1.

Blood from each patient was initially collected in CellSave^® preservative tubes and processed for CTC enumeration as described above. Patients who were positive for CTCs were subjected to a new blood draw into K3EDTA anticoagulant tubes used in the CellSearch^® Profile Kit, for CTC enrichment and RNA extraction as described above. The TRIzol^® lysate obtained was transferred to a 1.5 mL RNase-free Eppendorf tube, and stored at -80 °C until needed.

On the same day blood was drawn for CTC enrichment and RNA extraction, patients then underwent CT-guided BMBx using a battery-powered drill and biopsy needle set. Four to six BMBxs were obtained from the iliac bone of each patient. One of the cores was placed in an RNase-free Eppendorf tube and flash frozen and transported in liquid nitrogen for storage at −80 °C until processing. The remaining BMBxs were fixed for 24 h in 4 % paraformaldehyde in diethylpyrocarbonate (DEPC)-treated PBS, and decalcified in 10 % EDTA in autoclaved DEPC-treated water, pH 7.0, with agitation at room temperature for 3 days. After decalcification, each BMBx was progressively dehydrated with increasing concentrations of ethanol, and immediately paraffin-embedded using Precision Cut Paraffin (Thermo Scientific). All aqueous solutions were prepared using DEPC-treated water.

Two adjacent 5-µm sections were obtained from the formalin-fixed paraffin-embedded (FFPE) BMBxs, using at all times RNase-free technique. Immunohistochemistry (IHC) for cytokeratin was performed on one of the slides for identification of PCa cells, as previously described [14, 15], as PCa cells are the only cytokeratin-positive cells expected to be found within the BMBx. The adjacent tissue section was mounted onto an RNase-free polyethylene naphthalate (PEN) membrane glass slide (Arcturus), which was previously sprayed with RNase AWAY™ (Thermo Scientific), washed twice with DEPC-treated, nuclease-free water (Fisher Scientific), and then exposed to UV for 30 min under a laminar flow hood to render the PEN membrane more hydrophilic and improve adherence of the specimen. After air drying for about 2 h under a laminar flow hood, the section mounted on the slide was deparaffinized, hydrated with decreasing graded alcohols made with DEPC-treated water, and rapidly stained with Harris hematoxylin and alcoholic Eosin Y solution (H&E), washed twice with 100 % ethanol and xylene, then held in xylene until initiation of the laser capture microdissection (LCM) session. The H&E-stained slide was air dried and loaded onto the ArcturusXT™ LCM System. Metastatic lesions identified in the FFPE sample were excised using both the infrared (IR) and ultraviolet (UV) microdissection lasers, and collected on individual CapSure^® Macro LCM Caps (Arcturus). When needed, the adjacent section immunostained for cytokeratin was used for guidance during the LCM session to identify epithelial (PCa) cells within the bone marrow. Tumor material was deemed to be inadequate for LCM and RNA collection if the area identified as tumor was smaller than 9000 µm², with 150 PCa cells on average. To collect total RNA from the microdissected tissue, the CapSure^® Macro LCM Cap containing the captured material was immediately placed into a 0.5 mL RNase-free microcentrifuge tube filled with 50 µL of TRIzol^® reagent, vortexed for 5 min to lyse the cells, and the lysate obtained stored at −80 °C until needed.

Cultured PCa cells and CTCs were lysed with TRIzol^® reagent, as described above. Flash frozen BMBxs were placed in pre-chilled microtubes containing 2.8 mm ceramic beads and TRIzol^® reagent using a Precellys^®24 homogenizer (Peqlab LLC), 1 cycle × 30 s at 6000 rpm at 4 °C, similarly as described by others [16]. High-quality total RNA (DNA-free) was purified from cultured cells, CTCs, LCM, and frozen BMBxs lysed with TRIzol^® using Direct-zol™ RNA MiniPrep (Zymo Research), as per manufacturer’s instructions. For gene analysis of CTCs enriched from patients’ blood, spiked-in cultured cells, or tumor cells metastatic to bone recovered by LCM, RNAs were amplified using an antisense mRNA (aRNA) amplification system based on the Eberwine’s procedure [17], using the MessageAmp™ II aRNA Amplification Kit (Invitrogen). The protocol was performed as recommended by the manufacturer, except for the replacement of ArrayScript™ RT by Superscript^® VILO™ Master Mix (Invitrogen) and the replacement of 10 × First Strand buffer from MessageAmp™ II aRNA Amplification Kit (Invitrogen) by 5 ×  First Strand buffer from Superscript^® II Reverse Transcriptase (Invitrogen). This was done to increase the reverse transcription (RT) of total RNA into a first strand of complimentary DNA (cDNA). SuperScript^® VILO™ Master Mix contains a recombinant ribonuclease inhibitor and SuperScript^® III RT, which is among the best performing reverse transcriptases in terms of reproducibility and sensitivity for low copy RNA levels [18]. After RNase treatment, a second strand cDNA is generated by DNA polymerase. The resulting double-stranded cDNA was then used as a template for T7-RNA polymerase for in vitro transcription (IVT) into aRNA (also known as complimentary RNA, cRNA) and amplification. The procedure was repeated in a second round of amplification when additional aRNA yield was needed. After IVT, double-stranded cDNAs were removed by treatment with DNase I, and the amplified RNA was purified.

Concentrations of total RNA and amplified aRNA obtained from cells spiked in blood, as well as from amplified aRNA obtained from CTCs and flash frozen or FFPE BMBxs, were quantified by absorbance measurements using an Epoch™ Microplate Spectrophotometer for micro-volume analysis (BioTek). Purity of total RNA and amplified aRNA was assessed through the ratio of the absorbance of the samples at 260 and 280 nm (A₂₆₀/A₂₈₀).

Integrity of mRNA is usually assessed by the ratio of the 28S:18S rRNA species shown in denaturing agarose gel electrophoresis, based on the assumption that the quality of rRNA (more than 80 % of total RNA) reflects that of underlying mRNA. However, this approach cannot be used to assess the integrity of amplified aRNA, which does not contain rRNA and derives from minute amounts of total RNA. Therefore, we assessed the integrity of aRNA with RT-PCR (see below) using two sets of primers to probe different positions (5′ end, middle, and 3′ end regions) of EpCAM and GAPDH transcripts, by verifying the presence of each respective amplicon on agarose gels (Fig. 2a), following a strategy similar to that described by Nolan [19].

Total RNA or amplified aRNA was reverse transcribed into cDNA using iScript™ cDNA synthesis kit (Bio-Rad) according to manufacturer’s protocol. For RT-PCR, resultant cDNAs were used as a template in a PCR reaction using DreamTaq DNA polymerase (Life Technologies). Forward and reverse primers used are listed in Table 2. The following amplification conditions were used: an initial denaturation at 95 °C for 3 min, followed by 35–40 cycles (except for GAPDH, 25 cycles) of 95 °C for 30 s, 52–55 °C for 30 s and 72 °C for 1 min, followed by a final extension 72 °C for 5 min. PCR products were resolved on a 2 % agarose gel and visualized by ethidium bromide staining. DNA bands were visualized using a ChemiDoc XRS gel documentation system (Bio-Rad).

For reverse transcriptase quantitative real-time quantitative PCR (RT-qPCR), the Mastercycler RealPlex2 (Eppendorf) real-time PCR system and GoTaq qPCR Master Mix (Promega) were used. Thermal cycle parameters were as follows: initial activation at 95 °C for 2 min, 40 cycles of denaturation at 95 °C for 15 s, annealing 55 °C for 15 s, and extension at 72 °C for 30 s. The mean cycle threshold (Ct) for each gene was normalized to levels of the housekeeping gene GAPDH in the same sample. Relative fold changes in expression for each gene were calculated by the delta–delta-CT method [20].

As a proof of concept, we selected eight genes for analysis with RT-PCR in CTCs and LCM BMBxs based on their relevance to PCa bone metastasis: (a) EpCAM, which codes for a transmembrane epithelial glycoprotein [21] overexpressed in adenocarcinomas [22] and used in the CellSearch^® system to enrich CTCs via immunomagnetic separation; (b) PSA (prostate-specific antigen), a well know biomarker for PCa screening found to be positive in mCRPC patients [23]; (c) BMP-7 (bone morphogenetic protein-7), a member of the transforming growth factor-beta (TGF-β) family that is usually expressed in osteoblastic bone metastases of PCa [24], and increased in CRPC patients [25]; (d) MMP-14 (a.k.a. MT1-MMP), a membrane-tethered matrix metalloproteinase that we found to be expressed by PCa cells in skeletal metastases, and contribute to bone remodeling and intraosseous tumor growth [26]; (e) TMPRSS2-ERG gene rearrangement due to chromosomal translocations that fuse the androgen-regulated TMPRSS2 promoter with the ETS family transcription factor ERG, which is expressed in about half of advanced PCas [27‐29] and has been associated with an increased risk death in PCa patients [30‐32]; (f) IL-6 (interleukin-6), which is highly expressed by PCa cells with aggressive phenotype [33], and has been associated with resistance to chemotherapy in CRPC [34] and bone remodeling in PCa bone metastases [35, 36]; (g) Vimentin, an intermediate filament protein expressed in mesenchymal cells frequently used as a marker of epithelial to mesenchymal transition (EMT) [37]; and (h) GAPDH, a housekeeping gene used as a control.

Data obtained using RT-qPCR are presented as mean values ± SD and analyzed using the ANOVA test, with p < 0.05 considered as statistically significant.

To mimic the CTC number threshold currently used to predict overall survival (OS) in men with mCRPC [11, 21, 38‐40] and establish the method for RNA amplification from limited quantities of total RNA for gene expression, we spiked blood from healthy male volunteers with five PC3 or 5 LNCaP cells. Using the CellSearch^® Circulating Epithelial Cell Kit in the CellTracks^® Autoprep^® system, we found that the recovery rate of PC3 and LNCaP cells spiked into normal blood was 76.7 ± 6.2 % (n = 6) and 80.2 ± 3.0 % (n = 6) at the five and 50 cell level, respectively, confirming high efficiency of PCa cell recovery from blood. No epithelial cells were found in non-spiked matched blood used as negative controls. After we confirmed the cell recovery rate using the Circulating Epithelial Cell enrichment system, we then spiked five PC3 or LNCaP cells in new blood of healthy men drawn in K3EDTA tubes and processed using the CellSearch^® Profile Kit, used for RNA analysis. Minute amounts of total RNA extracted from the PC3 and LNCaP cells recovered through that CellSearch^® Profile Kit and the CellTracks^® Autoprep^® system were subjected to one and two round amplification using the MessageAmp™ II aRNA Amplification Kit. As shown in Table 3, this methodology was efficient enough to generate sufficient quantities of aRNA of good quality (A₂₆₀/A₂₈₀ ratio of 1.8–2.0) from a few PCa cells for gene expression analyses, such as RT-PCR, RT-qPCR, and cDNA microarrays if desired. To establish the nature of linear amplification, we compared relative expression levels in a set of genes with high, medium, or low expression levels, in unamplified RNA extracted from 2 × 10⁵ PC3 cells (~6–9 µg of total RNA, equivalent to ~180–270 pg of mRNA) and one- or two-round amplified aRNA derived from five PC3 cells (Table 3). The relative abundance of transcripts for the different genes assessed by RT-qPCR was maintained, confirming the linear amplification of RNA (Fig. 1).

To check the purity and quality of the aRNA amplified from RNA extracted from CTCs and PCa cells microdissected from BMBxs, we designed primers to target different regions of EpCAM and GAPDH transcripts, epithelium-specific and housekeeping genes, respectively (Fig. 2a). The primers targeting the 5′ end and the middle regions of the transcripts produced the expected size of the resultant PCR products as effectively as those targeting the 3′ end (Fig. 2b), indicating that the integrity of the aRNA amplified from both CTCs and microdissected FFPE PCa bone metastases was good. Having confirmed that the quality of the amplified aRNA was adequate, we analyzed all the CTC and BMBx samples included in the study using RT-PCR.

Of 24 patients with bone scan/CT scan evidence of one or more skeletal metastases enrolled in the study, nine had no CTCs and therefore did not undergo bone biopsy. Table 1 summarizes the demographic, clinical, and tumor characteristics of the 15 patients that had one or more EpCAM-positive CTCs in their blood. One patient was eliminated from the study due to undetectable aRNA in CTCs even after two-rounds of amplification. In addition, another patient was excluded because his BMBx showed sarcomatoid tumor cells. Of the 13 patients remaining, a second blood sample was obtained and subjected to CTC enrichment using the CellSearch^® Profile kit for RNA extraction, and subsequent aRNA amplification. RT-PCR analysis of aRNA amplified from CTCs revealed that the gene expression varied from one patient to another. As expected, EpCAM, which codes for the protein used for CTC enrichment, was expressed in all cases. PSA and IL-6 were also expressed by CTCs from all patients (Fig. 3a). Unexpectedly, the transcript for Vimentin was expressed in many of patients’ CTCs (Fig. 3a). Importantly, aRNA amplification from as little as one CTC (e.g., patient 04 M–S) was sufficient to perform gene analysis, consistent with our results with amplified aRNA obtained from five cultured PCa cells.

Of 13 patients with detectable CTCs, 7 (54 %) had cytokeratin-positive cells on immunohistochemical analysis of their BMBx. This yield is consistent with that reported in the literature [41, 42]. However, only four had a large enough area occupied by cancer cells for LCM and RNA collection. In those cases, we compared aRNA amplified from RNA extracted from ground flash frozen BMBx with that obtained from LCM FFPE sections of matching BMBx of the same patient. In the latter, areas exclusively containing PCa cells, as identified by H&E and confirmed by IHC for pan-CK, were UV laser microdissected from the bone specimens mounted on membrane glass slides, and captured with the help of a gentle IR laser, using the ArcturusXT™ LCM System (Fig. 3b). We observed that, although RNA yield was better in ground frozen bone core biopsies, the expression level of most of the genes specific to PCa was considerably diminished compared to amplified aRNA derived from PCa cells microdissected from FFPE BMBxs after equal number of PCR cycles and normalization to GAPDH (not shown). Conversely, genes commonly expressed by bone cells, such as RANK (receptor activator of nuclear factor κ-B), were highly expressed in aRNA derived from ground frozen BMBx, but minimally or not expressed in pure PCa cell populations acquired through LCM in FFPE BMBx (not shown). These results demonstrate that analysis of frozen bone samples is limited by contamination with normal bone cells.

Using at all times RNase-free instruments and technique, and aqueous reagents prepared in nuclease-free water, we were able to amplify good quality aRNA from the total RNA collected from the FFPE BMBxs (Fig. 2b). Since the aRNA yield obtained was good, we analyzed in BMBxs the same genes studied in CTCs (Fig. 3c). In order to explore whether or not there was concordance in presence or absence of gene expression as detected by conventional RT-PCR in CTCs and matched BMBxs, we lined up digitally the bands obtained for all the genes amplified in CTCs and BMBxs (same number of cycles) side by side (Additional file 1: Figure S1). We found that in 21 out of 28 comparisons, the presence or absence of detectable gene expression in CTCs and PCa cells microdissected from single bone metastasis of the same patients was concordant.