The isolation and characterization of CTC subsets related to breast cancer dormancy

Uncovering CTCs phenotypes offer the promise to dissect their heterogeneity related to metastatic competence. CTC survival rates are highly variable and this can lead to many questions as yet unexplored properties of CTCs responsible for invasion and metastasis vs dormancy. We isolated CTC subsets from peripheral blood of patients diagnosed with or without breast cancer brain metastasis. CTC subsets were selected for EpCAM negativity but positivity for CD44+/CD24− stem cell signature; along with combinatorial expression of uPAR and int β1, two markers directly implicated in breast cancer dormancy mechanisms. CTC subsets were cultured in vitro generating 3D CTC tumorspheres which were interrogated for biomarker profiling and biological characteristics. We identified proliferative and invasive properties of 3D CTC tumorspheres distinctive upon uPAR/int β1 combinatorial expression. The molecular characterization of uPAR/int β1 CTC subsets may enhance abilities to prospectively identify patients who may be at high risk of developing BCBM.

We have previously reported the discovery of CTCs that do not express the common carcinoma epithelial cell adhesion molecule (EpCAM-negative CTCs) and possess high competence to generate BCBM in xenografts 13 . We posited that specific EpCAM-negative CTCs subpopulations, shed from the primary tumor and found in the circulation, avoid organ arrest with extreme efficiency by the concomitant presence of stem cell and quiescence properties. The molecular switch to differentiate quiescence in malignant CTCs depends on the cross-talk between CTCs and the tumor microenvironment. Of note, previous studies have established the presence of two neoplastic markers, urokinase plasminogen activator receptor (uPAR) and integrin β 1 (int β 1) promoting tumor cell growth and proliferation when they interact with the extracellular brain microenvironment 14,15 . However, the loss of uPAR and int β 1 expression strikingly reduces proliferative signals causing a shift from an invasive or metastatic to a dormant state, and directly implicating these two biomarkers in mechanisms of tumor cell dormancy in vivo 1,2,14,15 .
Here, we report the isolation of subsets of EpCAM-negative breast cancer CTCs containing stem-cell properties (CD44 + /CD24 − ) by multiparametric flow cytometry with a combinatorial uPAR and int β 1 expression and their abilities to grow long-term in vitro. Second, we characterized CTC subsets possessing six cell surface expression markers (CD45 − /EpCAM-negative/CD44 + /CD24 − /uPAR +/− /int β 1 +/− ) to determine the expression profiling of candidate genes related to breast cancer and embryonic stem-cell pathways and demonstrate their tumor origin as putative CTCs. Third, we investigated CTC subsets for cell adhesion, proliferation properties, and for subset abilities to generate in vitro 3D CTC tumorspheres (3D-spheroids) and invade into extracellular matrix. Lastly, we sorted uPAR and int β 1 CTCs at single-cell level by employing the DEPArray ™ platform and performed mutation analyses to reveal unique genomic signatures of uPAR/int β 1 CTC subsets.
In summary, we provide first-time evidence for the isolation of intra/inter-patient EpCAM-negative, uPAR/int β 1 CTCs subsets with distinct capabilities for long-term in vitro growth; along with mechanistic link of these CTC subsets to cell adhesion, proliferative and invasive properties relevant to BCBM onset.
Second, we carried out comprehensive genotyping analyses on CTC subsets derived from BCBM patients either with the presence or absence of uPAR/int β 1 expression. We applied short tandom repeat (STR) DNA fingerprinting (16 loci). These CTC subsets possessed unique STR DNA fingerprinting profiles and were distinct from ones employing cancer cell lines from available databases (http://bioinformatics.istge.it/clima/) and from each other (Fig. 2b).
Third, we interrogated CTC subsets by their abilities to be viable and expand in vitro. We were able to grow CTCs as non-adherent 3D CTC tumorspheres regardless of whether they were derived from BCBM vs no BCBM patients and independent of uPAR/int β 1 expression (uPAR + /int β 1 + , uPAR − /int β 1 − , uPAR + /int β 1 − and uPAR -/int β 1 + ). We were able to grow CTC subsets under normal aerobic conditions (37 °C with 5% CO 2 ) using 1% soft agar on 6-well tissue culture plates 17 (Fig. 3). Of note, lowering O 2 levels to hypoxic conditions (37 °C with 3-4% CO 2 ) did not significantly affect CTC subsets growth. CTCs subsets were passaged using 0.25% trypsin (Gibco Life Technologies, Inc.). However, they tended to grow and expand as clusters (CTC tumorspheres) and dissociated only as singlets or paired cells. CTC-generated tumorspheres grew in vitro having two distinct cell sizes. We classified CTCs < 5 μ M diameter as small CTCs and > 5 μ M as large CTCs (Fig. 3, white arrows). We also observed 3D CTC tumorspheres to expand as an endomembrane partitioning-like system (Supplementary Fig. 1) in which the endomembrane furrow separates the daughter and mother cell during cell-division events 18 .
Biomarker profiling of CTC subsets. To validate the specific expression of cell-surface markers used for CTC enrichment, we performed RT-PCR analyses. We amplified mRNAs from 3D CTC tumorspheres obtained from breast cancer patients with or without BCBM, and analyzed them by RT-PCR to assess Scientific RepoRts | 5:17533 | DOI: 10.1038/srep17533 expression levels of neoplastic (uPAR/int β 1), tumor epithelial (EpCAM), circulating endothelial (CD31), mesenchymal stem cell (CD73, CD90 and CD103) and breast cancer stem cell (CD44 + /CD24 − ) markers. We detected the presence of neoplastic and breast cancer stem cell markers coupled with negativity for EpCAM (Fig. 4a). Next, to confirm that isolated CTCs subsets did not represent non-CTC populations, we evaluated specific transcript levels for the expression of mesenchymal stem cells (CD73, CD90 and CD105) and circulating endothelial (CD31) markers (Fig. 4a). These markers were not expressed in in vitro 3D CTC tumorspheres (Fig. 4a). The absence of circulating mesenchymal and endothelial markers suggests that these putative 3D CTCs tumorspheres had a non-hematopoietic origin and that they did not derive from non-CTC populations. Moreover, we assessed in vitro 3D CTC tumorspheres to retain original gene expression patterns irrespective of the initial selection under long-term in vitro culture conditions. Further, we assessed protein expression of CTC subsets uPAR and int β 1 markers by immunofluorescence on in vitro 3D CTC tumorspheres. We found that these CTC subsets possessed a characteristic combinatorial expression pattern on their cell-surface (Fig. 4b). Lastly, we verified the neoplastic origin and proliferating abilities of 3D CTC tumorspheres by evaluating the pan-cytokeratin and Ki67 expression and confirmed their detection in uPAR/int β 1 3D CTC tumorspheres (Fig. 4c).

CTC single-cell genotyping.
To dissect the heterogeneity of CTC subsets at a single-cell level, we captured cells positive or negative for uPAR, int β 1 and HER2 expression markers using the dielectrophoretic array platform DEPArray ™ (Silicon Biosystems, Inc.), following a pre-enrichment step of CD45 − / EpCAM-negative/CD44 + /CD24 − CTCs derived from BCBM and no BCBM patients (Fig. 5). Of note, DEPArray ™ technology enables the isolation of viable CTCs for interrogation of CTCs on a cell-per-cell basis, the smallest functional unit of cancer 19 . CTC subsets were sorted per DEPArray ™ specifications (all-or-none threshold for CTC marker expression) employing uPAR, int β 1 and HER2 selection. Next, the genomic content of DEPArray ™ -sorted CTCs containing combinatorial expression of these markers (uPAR +/− /int β 1 +/− and HER2 +/− ) was assessed at single-cell level. Single CTCs were amplified employing the Ampli1 ™ WGA method (Silicon Biosystems, Inc.) and mutation analyses of > 200 hallmark cancer genes were carried out by applying the MassARRAY ™ detection system (Sequenom, Inc.) on DEPArray ™ -sorted single CTCs (n = 7). We were able to detect the presence of HSP90AB1 C2139T, PRKCB G785T, AURKC C154G and JAK2 A2049CT cosmic mutations in BCBM-derived CTCs at the

Characterization of CTC subsets revealed distinct in vitro biological patterns.
To interrogate 3D CTC subsets for multiple in vitro properties as related to steps of the metastatic cascade, we investigated the spatial-temporal kinetics of in vitro 3D CTC tumorspheres formation by performing  in vitro 3D CTC tumorspheres using stem cell and non-adherent conditions. White arrows indicated smallvesicle-like cells. Images were taken at 40X by phase contrast microscopy (Zeiss, Inc.). Representative images are shown.
Second, we assessed the proliferative, adhesive and invasive capacities of patient-derived EpCAM-negative CTC subsets. Cell proliferation assays applying 3D non-adherent cells methodologies to 3D CTC tumorspheres revealed that these subsets possessed differential in vitro proliferation abilities that correlated with the combinatorial expression of uPAR and int β 1 markers. Further, uPAR + /int β 1 − and uPAR − /int β 1 + CTC tumorspheres showed an additive proliferative capacity between days 9 and 12 (Fig. 6b).
Third, to evaluate CTC subsets adhesion capabilities, we grew those using Trevigen ® basement membrane extract (BME) tumorsphere assays [20][21][22] . We observed high adhesion of uPAR − /int β 1 + CTC tumorspheres on BME matrix at 48 hours while the other three CTC subsets showed no attachment in adhesion assays even up to 96 hours incubation time (Fig. 6c). Cell migration and invasion are fundamental processes which regulate important cellular events such as angiogenesis, invasion and metastasis of cancer cells. Interestingly, EpCAM-negative CTC subsets aggregated and formed in vitro 3D CTC tumorspheres. Accordingly, we determined how CTC tumorspheres generate invadopodia under well-controlled in vitro conditions, capable to become motile and to invade into extracellular matrix (ECM) of the 3D-invasion assay (Fig. 7a). Invadopodia formation by invading CTCs recapitulates the early steps of brain colonization observed in vivo 23 . To this end, we assessed Trevigen ® 3D tumorsphere invasion assays 20 on in vitro 3D CTC tumorspheres and visualized invadopodia formation. We used non-invasive poorly metastatic MCF7 and highly metastatic 231Br breast cancer cells as negative and positive controls, respectively. We processed invasion matrix to monitor invadopodia formation at day 4. Non-invasive control MCF7 cell-derived spheroids did not form any protrusions whereas invadopodia formation was noted employing invasive 231Br spheroids. Of note, protrusions and tiny ruffle-like invadopodia were observed in uPAR + /int β 1 − and uPAR + /int β 1 + CTC subsets at day 11 (Fig. 7b, yellow arrows). Conversely, no invadopodia formation was observed in uPAR − /int β 1 − and uPAR − /int β 1 + 3D CTC subset spheroids plated on BME invasion matrix per assay specifications 20 . These results demonstrate that the uPAR/int β 1 biomarker axis enables invadopodia formation when subjected to the proper tumor microenvironment and factors. They are of relevance because the formation of invadopodia in CTC is required for the in vivo extravasation through blood-brain barrier as the early step toward CTC colonization of brain and BCBM development 23 .
Fourth, we confirmed the EpCAM status of in vitro 3D CTC tumorspheres by FDA-cleared CellSearch ® CTC testing which is however capable to capture only CTCs positive for EpCAM 24 . We spiked ~100 cells of EpCAM-negative in vitro 3D CTC tumorsphere cells in blood from normal healthy donors. We were able to capture only 1/100 EpCAM-positive CTCs from CellSearch ® analyses (Supplementary table   S3). These findings demonstrates that the EpCAM-negative CTC subsets retain their expression under long-term in vitro conditions. CTC gene expression profiling. CTCs containing stem cell properties undergo embryonic trans-differentiation at distant organs during metastasis. We performed real-time-PCR (RT 2 -PCR) human embryonic stem cell array (Qiagen) profiling to determine the expression of 83 candidate genes in FACS-sorted EpCAM-negative, uPAR + /int β 1 + and uPAR − /int β 1 − stem cell CTC subsets derived from clinically diagnosed breast cancer patient with or without BCBM. Real-time PCR analyses revealed > 30 fold increased expression of CDC42, CDK1, FGF2, RIF1, HSPA9 and KLF4 genes between uPAR + / int β 1 + and uPAR -/int β 1 − CTC subsets over the five internal controls of RT 2 PCR profiler array (Qiagen) and in relation to patient BCBM status (Fig. 8a). Further, CDC42 and POU5F1 gene expression level were relatively higher (> 8 fold) when uPAR + /int β 1 + compared with uPAR − /int β 1 − CTC subsets in breast cancer patient without BCBM (Fig. 8b). These findings suggest that uPAR + /int β 1 + CTC subsets possess gene profiles for increased proliferation, DNA damage repair pathway and relate closely to BCBM onset.

Discussion
CTCs are the "seeds" of uncurable metastasis and can represent a promising and effective alternative to invasive tumor biopsies to detect, monitor and combat solid tumors in patients 3-6 . However, thus far, only one platform CellSearch ® (Janssen Diagnostics, LLC.) has been cleared by the FDA for CTC clinical testing and application. While CellSearch ® provided a breakthrough in the CTC field, there are known limitations by this platform since it captures only CTCs positive for the epithelial cell adhesion molecule (EpCAM-positive CTCs) 24,25 . Furthermore, CellSearch ® involves a fixation step and CTCs captured this way cannot be interrogated further for other downstream application such as RNA-based measurements and culturing CTCs under in vitro and in vivo conditions. This can be particularly relevant towards discriminating CTC critical for the development of metastasis vs ones non metastasis-competent ("irrelevant" CTCs) 4 . These insights have an added impact in breast cancer, a disease known to have high frequency of recurrence following excision of the primary tumor 26,27 . We have previously demonstrated that EpCAM-negative CTCs isolated from breast cancer patients were competent for metastasis in xenografts 13 . Further, we have reported identifiers relevant to the breast cancer brain-metastasis-selected CTC profile suggesting their biological and functional relevance in BCBM 13 . Considering the heterogeneity of CTCs, we hypothesized that multiple and contrasting biomarkers are responsible for mechanisms leading to BCBM onset; and additive or alternative to the brain-metastasis selected CTC profile 13 . The purpose of this study was to identify, isolate and characterize CTC subsets with properties related to breast cancer dormancy. We focused on EpCAM-negative CTCs possessing alternative combinations of urokinase plasminogen activator receptor (uPAR) and integrin β 1 (int β 1), two biomarkers known to be directly implicated in breast cancer dormancy 1,2 .
We applied multiparametric flow cytometry and CD45 − /CD44 + /CD24 − as initial selection markers and specific criteria for EpCAM-positive and EpCAM-negative CTCs: EpCAM-negative PBMCs derived from breast cancer patients sorted through multiparamteric flow cytometry followed by the selection of uPAR/int β 1 combinatorial CTC subset expression (Fig. 1). First, gene expression profiling of 83 breast cancer candidates revealed that enriched CTC population disseminate from their primary neoplastic breast tumor and have their unique gene signature (Fig. 2a). Furthermore, the presence of a unique STR DNA fingerprinting of sorted cells revealed their authenticity as putative CTCs which were distinct from human breast cancer cell lines (Fig. 2b). Of note, embryonic stem-cell gene expression profiling revealed the high expression of CDK1, HSPA9, CDC42, FGF2, KLF4 and RIF1 genes in uPAR + /int β 1 + CTC subsets when compared with uPAR − /int β 1 − CTC subsets in BCBM patients (Fig. 8a). FGF2 and KLF4 genes play an important role in blood-brain barrier permeability 28,29 , RIF1 is involved in DNA repair pathways 30 whereas CDK1 and CDC42 are profoundly implicated in mechanisms regulating cell proliferation 31,32 . Accordingly, the high expression of above-indicated genes suggests the BCBM competency of uPAR + /int β 1 + CTC subsets additive to the brain metastasis-selected marker profile we have previously discovered 13 .
Second, we were able to grow FACS-sorted CTC populations and to expand them as 3D CTC tumorspheres under in vitro conditions (Fig. 3). It was recently reported that CTC clusters derived from primary breast cancer tumor have more metastatic competency compared to single CTCs 33 . Our in vitro CTC subsets population expanded as 3D tumorspheres in non-adherent stem-cell conditions; however, they did not fully dissociate when trypsinized suggesting metastatic competency. We also observed cellular protrusions stemming at the periphery of these 3D CTC tumorspheres during in vitro expansion ( Fig. 3 and Supplementary Fig. 1). Di Vizio et al. 34 found that tumor microvesicles present in the circulation of aggressive form of prostate cancer and their presence in tumor microenvironment may be functionally relevant in potentiating metastasis. Our findings using uPAR/int β 1 CTC subsets are consistent with these notions. Thus, elucidating the mechanisms for the generation of tumor-associated vesicles, termed oncosomes, and how they mediate intracellular signaling will be of significance in metastatic breast cancer.
Third, we investigated whether these CTCs subsets retain their initial selective markers uPAR and int β 1 under in vitro conditions. We observed that the combinatorial expression of uPAR and int β 1 remains constant to their selection and were not altered during in vitro expansion (Fig. 4a). The lack of mesenchymal (CD90, CD73 and CD105) 35 and circulating endothelial (CD31) 36 markers expression in 3D CTC tumorspheres indicate that these putative 3D CTC tumorspheres are non-hematopoietic, tumorigenic, and contain stem-cell properties, eg, presence of the CD44 + /CD24 − axis. Further, positivity of Ki67, cytokeratins (CK) along with EpCAM negativity in 3D CTC tumorspheres (Fig. 4b,c) suggest their hybrid or plastic state required for transition/interchange of mesenchymal to epithelial properties postulated for metastasis to occur 37 .
Disseminated EpCAM-negative CTCs undergo mesenchymal-epithelial transition (MET) at distant organs, invade the tissue and then become localized to generate metastatic tumors. Accordingly, CTC adhesion, proliferation, invasion and tumorsphere formation are of value to characterize CTCs at cellular and molecular levels. The neoplastic markers, uPAR and int β 1 interact with each other to drive tumor growth by regulating the cross-talk with the target organs of microenvironments. Interestingly, the ablation of uPAR and int β 1 switches the proliferative cell to dormant G 0 -G1 arrest state resulting in tumor suppression in vivo 1,15 . We observed uPAR + /int β 1 + 3D CTC tumorspheres to be more proliferative compared with CTC populations containing the uPAR − /int β 1 − CTC subsets having this dormancy axis (Fig. 6a,b). Additionally, the presence of invadopodia formation/cell invasiveness in uPAR + /int β 1 − and uPAR + /int β 1 + 3D CTC tumorspheres advocates for their metastatic competency (Fig. 7b). Fourth, uPAR/int β 1 CTC subsets underwent expansion in size, volume and number prior to CTC clustering and 3D CTC tumorspheres formation at variable rates via an endomembrane partitioning-like system ( Supplementary Fig. 1, yellow arrows) 18 . Further experiments with xenografts and live-cell imaging using membrane binding and nuclear dyes will be required to confirm the mechanism of CTC clustering and 3D CTC tumorsphere formation in vivo. Regardless, our findings are of significance to clinical dormancy since CTCs shed from the primary tumor exhibit various CTC circulator phenotypes via a mechanism(s) of expansion that are yet unknown. These phenotypes are dependent on the biomarker expression such as presence of uPAR and int β 1 axis. Multiple circulator CTC phenotypes must exist; they resist apoptosis, undergo evolution and clonal selection via DNA damage and active DNA repair pathways, and avoid arrest and adhesion to target organs with extreme efficiency. Selected CTC clones specific for uPAR/int β 1 biomarker axis undergo proliferation and expansion for a long-term niche pool, and CTC clustering for secondary tumorsphere formation. For example, it is known that breast cancer cells grow in a disorganized fashion on reconstituted basement membrane assays by employing int β 1 and epidermal growth factor (EGF)-dependent signaling pathways 36 . We observed int β 1-dependent adhesion capabilities of dormant tumor populations in uPAR -/int β 1 + 3D CTC tumorspheres when grown on BME matrix (Fig. 6c). This suggests that int β 1 + "dormant" CTCs might undergo some degree of differentiation but they become non-proliferative in the absence of uPAR expression 1,38 .
Lastly, heterogeneous populations of CTCs harbor genetic and epigenetic changes at single-cell level [39][40][41][42] and exhibit distinct breast cancer phenotypes 43 . We used the DEPArray ™ platform (Silicon Biosystems, Inc.) to dissect CTCs at single-cell level derived from BCBM vs no BCBM followed by MassARRAY ™ mutation analysis (Sequenom, Inc., Supplementary table S2). We detected common cosmic mutation PRKCB G785T in patient-derived CTCs, regardless of their expression markers (uPAR/int β 1/HER2) or brain metastasis clinical status. However, BCBM-derived CTC subsets contained cosmic mutation HSP90AB1 C2139T in uPAR + /int β 1 + /HER2 + CTC, and AURKC C154G and JAK2 A2049CT mutations in two different CTCs containing uPAR − /int β 1 -/HER2expression. Accordingly, while the variability of genetic mutations at the single-cell CTC level confirmed the high heterogeneity of CTCs, it can provide a better approach to evaluate the biology of CTCs by targeting these mutations and assessing their impact.
In conclusion, the detailed characterization and application of uPAR/int β 1 CTC subsets can be useful to decipher cellular and molecular mechanisms of organ-homing CTCs and to better understand breast cancer dormancy versus CTCs abilities to adhere, proliferate and invade, which are hallmark properties of tumor progression. This study represents a step forward towards early detection and treatment of breast cancer-associated brain metastasis. The extension of these investigations will be a clinically useful tool in personalized medicine applications for effective drug screening/testing method rather than cellular transplantation.  44.7%) did not have brain metastasis ( Table 1). Details of each selected patient were provided in the supplementary table S1. Only patients starting a new line of therapy were enrolled in the present study. Patients with concurrent disease(s) were excluded. Peripheral blood (25-45 mls/patient) was obtained at the middle of vein puncture after the first 5 ml of blood was discarded to avoid contamination by normal epithelial cells. All samples (25-45 mls blood) were collected using CellSave TM (Janssen Diagnostics, LLC) or EDTA tubes in sterile conditions according to CTC testing to be performed, and provided immediately to the laboratory for CTC analysis.
Peripheral blood mononuclear cells isolation. PBMCs were isolated as described elsewhere 44 .
Briefly, PBMCs from whole blood were isolated by using red blood cell lysis buffer (154 mM NH4Cl, 10 mM KHCO 3 , 0.1 mM EDTA) at a ratio of 1:25, followed by incubation at room temperature (25 °C) for 5 min, then pelleting remaining blood cells at 300 g for 10 min. Cell pellets, consisting mostly of mononucleated cells, was washed with 20 ml 1X PBS and centrifuged at 300 g for 5 mins. PMBCs were then counted by hemocytometer used for fluorescent labeling and capturing CTC using multi-parametric CTC subsets culture and growth conditions. FACS-selected CTC populations were grown as tumorsphere using Mammocult ™ media (StemCell Technologies, Inc.). Enriched CTCs were seeded on 1% agarose in 6-well tissue culture plate. Mammocult ™ media was then applied and used to grow 3D CTC tumorspheres by incubating cells at 37 °C and 5% CO 2 . 3D CTC tumorspheres were passaged with 0.25% trypsin-EDTA (Gibco Life Technologies, Inc.). CTC subsets were STR DNA fingerprinted (Fig. 2a). They were genetically analyzed by the MassARRAY ™ detection system (Sequenom, Inc.) to ensure tumor cell fidelity, and periodically assessed for pathogen-free Mycoplasma testing. They were used for experimental work only within the first 30-40 days of culture.
CellSearch ® CTC analyses. CellSearch ® CTC procedures were applied for 3D CTC tumorsphere analyses. Briefly, approximately 100 cultured cells from each FACS-selected group (uPAR + /β 1 int + , uPAR + /int β 1 − , uPAR − /int β 1 + and uPAR − /int β 1 − ) were spiked into 7.5 ml of peripheral blood from normal donors collected in CellSave TM tubes (Janssen Diagnostics, LLC.) tubes. Samples were loaded onto the CellTracks ® AutoPrep. The system added anti-epithelial cell adhesion molecule (EpCAM) ferrofluid to cells. Cells were automatically stained with anti-CK-PE to identify intracellular cytokeratins 8, 18 and 19 with anti-CD45/APC to identify leukocytes and with DAPI to identify cell nuclei 23,24  Reverse-Transcriptase PCR (RT-PCR). cDNA was isolated from in vitro 3D CTC tumorspheres and amplified by using REPLI-g WTA Single Cell Kit (Qiagen) according to manufacturer instructions protocol. Briefly, cells were lysed followed by gDNA removal. The subsequent reverse transcription reaction was performed by using oligo dT primer to amplify polyA + mRNA enrichment transcripts. The synthesized cDNA was ligated using a high-efficiency ligation mix followed by whole transcriptome amplification of cDNA with the REPLI-g SensiPhi DNA polymerase enzyme. RT-PCR were then performed by using gene specific primers (Supplementary table S4). Real-time PCR profiling. Amplified cDNA were purified by ExoSAP-IT (Affymetrix, Inc.) and were subjected to real-time PCR amplification using SYBR green method (Applied Biosystems, Inc. CTC proliferation assays. 3D CTC tumorspheres containing about 500 cells were grown in 96-well at different time points. Cell proliferation assays were performed by incubating cells with tetrazolium salt WST1 (Roche Life Technologies, Inc.) for 8 hrs and OD were measured at absorbance 450 nm and 600 nm. Student paired, type 2 t-test was applied to calculate p-value for statistical significance in between CTC subsets containing combinatorial expression of uPAR and int β 1 at different time points.
3D CTC tumorsphere growth assays. 3D CTC tumorspheres were dissociated as single CTC units or pairlets and then scored using hemocytometer and confirmed for cell viability using 1:1 Trypan Blue (Gibco Life Technologies, Inc.). Twenty-four well flat-bottom plates were coated with 1% soft agar and approximately 10-35 trypsinized CTC units/subset were suspended in 100 μ l of Mammocult ™ (StemCell Technologies, Inc.) media were added in each well in multiples. The tissue culture plate was then incubated at 37 °C to analyze the spatial-temporal kinetics of 3D CTC tumorsphere formation with a 10 week period. CTC growth rate was observed under 10X magnification and images were captured and analyzed every week under 40X magnification using phase-contrast microscopy (Zeiss, Inc.). BME adhesion assays. CTC subsets were aliquoted into 96 well flat-bottom tissue culture plates coated with Cultrex ® Basement Membrane Extract, PathClear ® (BME) (Trevigen ® , Inc.) and incubated for 96 hours at 37 °C. Plates were then analyzed every 24 hrs for in vitro 3D CTC tumorsphere BME adhesion using 10X magnification. Images were captured at 96 hrs endpoint under 40X magnification using phase-contrast microscopy (Zeiss, Inc.).
3D spheroid BME cancer cell invasion assays. CTC subsets were dissociated using 0.25% trypsin as single CTC units or pairlets. CTCs were scored using hemocytometer and confirmed for cell viability using 1:1 Trypan Blue (Gibco Life Technologies, Inc.). Dissociated CTCs were detected as per the protocol provided by the Trevigen ® assay kit (Trevigen ® , Inc.) 20 . Approximately 15-25 trypsinized CTC units/subset were suspended in 40 μ l of Mammocult ™ media and 10 μ l of 1 × 3D spheroid ECM was mixed well and a total volume of 50 μ l added to each well in triplicates in 96 well round-bottom plates provided by the kit. Non-invasive MCF7 and highly invasive 231Br breast cancer cells were used, and cell viability was confirmed after trypsinization. Approximately 10 3 cells in 40 μ l of growth media and 10 μ l of 1 × 3D spheroid ECM were mixed well and a total volume of 50 μ l was added to each well. The tissue culture plate was incubated at 37 °C for monitoring the 3D CTC tumorspheres formation under microscope with an endpoint of day 4. The images were captured and analyzed every two days under 40× magnification using phase contrast microscopy (Zeiss, Inc.). The invasion matrix was added into each well at day 4 and incubated for 1 hr to gel as per assay protocol. 100 μ l of growth media was added to each well and the plate was put in 37 °C incubator for regular monitoring of invadopodia formation from until day 11. Images were captured and analyzed every two days under 40X magnification using phase-contrast microscopy (Zeiss, Inc.).