Pairing Microwell Arrays with an Affordable,Semiautomated Single-Cell Aspirator for the Interrogation of Circulating Tumor Cell Heterogeneity

Abstract

Comprehensive analysis of tumor heterogeneity requires robust methods for the isolation and analysis of single cells from patient samples. An ideal approach would be fully compatible with downstream analytic methods, such as advanced genomic testing. These endpoints necessitate the use of live cells at high purity. A multitude of microfluidic circulating tumor cell (CTC) enrichment technologies exist, but many of those perform bulk sample enrichment and are not, on their own, capable of single-cell interrogation. To address this, we developed an affordable semiautomated single-cell aspirator (SASCA) to further enrich rare-cell populations from a specialized microwell array, per their phenotypic markers. Immobilization of cells within microwells, integrated with a real-time image processing software, facilitates the detection and precise isolation of targeted cells that have been optimally seeded into the microwells. Here, we demonstrate the platform capabilities through the aspiration of target cells from an impure background population, where we obtain purity levels of 90%–100% and demonstrate the enrichment of the target population with high-quality RNA extraction. A range of low cell numbers were aspirated using SASCA before undergoing whole transcriptome and genome analysis, exhibiting the ability to obtain endpoints from low-template inputs. Lastly, CTCs from patients with castration-resistant prostate cancer were isolated with this platform and the utility of this method was confirmed for rare-cell isolation. SASCA satisfies a need for an affordable option to isolate single cells or highly purified subpopulations of cells to probe complex mechanisms driving disease progression and resistance in patients with cancer.

Keywords

circulating tumor cells automation microwell array single cell transcriptomics

Introduction

Rare-cell populations can have a tremendous influence in disease development, progression, and resistance. Isolation and analysis of rare cells offers a method to study the pertinent cell-to-cell heterogeneity that may harbor drivers of disease progression or treatment resistance.^1,2 However, rare cells (e.g., circulating tumor cells [CTCs]) are generally difficult to isolate at high purity.³ The lab-on-a-chip (LOC) field has produced many iterations of rare-cell enrichment devices, leveraging both physical and biological phenomena.⁴ However, the majority of these platforms are bulk enrichment techniques, and significant levels of background or contaminating cells remain after isolation. While these methods can provide an averaged measure of cells found in the entire sample, the presence of contaminating cells can directly alter results and interpretation.^5,6 Moreover, devices that excel in cell yield/recovery often perform poorly in cell purity,⁷ establishing a need for methods that provide high rare-cell purity to avoid the bias of contaminating cells and enable meaningful clinical analysis. Furthermore, we envision that if widely accessible methods or technologies were available that could maintain cell viability and integrate with a range of existing LOC technologies, then a multitude of clinically relevant endpoints could be studied.

As mentioned previously, CTCs are an important example of a rare-cell population of interest, as demonstrated by the wealth of CTC enrichment microdevices developed in recent years. Moreover, CTCs have been shown to contain clinically relevant information with regard to cancer progression and therapeutic resistance.^8–10 CTCs are found in peripheral blood at a frequency of ~1 in 10⁸ to 10¹⁰ peripheral blood cells,^11,12 and those isolated from patients with prostate cancer have been shown to exhibit high degrees of heterogeneity.¹³ This heterogeneity extends to a range of molecular alterations including genomic, transcriptomic, and proteomic differences in both solid tumor biopsies and CTCs from liquid biopsies.^14,15 For example, single-cell RNASeq of prostate cancer CTCs identified intra- and interpatient heterogeneity in the expression of androgen receptor (AR) splice variants, which is known to confer resistance to AR-targeted therapies.¹⁶ Accordingly, there is a critical need to understand the molecular alterations that account for, and may drive, cancer progression to develop improved biomarkers and therapeutic modalities for patients with advanced disease. To do so, there is a need for a system that works in tandem with bulk LOC enrichment devices that can retrieve purified, live CTC populations while maintaining requisite cell viability that enables robust downstream single-cell interrogation.

For many years single-cell studies have been accomplished using cell micromanipulators (CMMs), which integrate directly with microscopes to interrogate cells of interest from a larger population.^17–19 These studies—and many that followed—used manually operated CMMs. Manual micromanipulators have been essential for advances in neuroscience,²⁰ microbial science,²¹ and in vitro fertilization.²² Unfortunately, such manual models can be cumbersome, requiring a practiced hand to operate accurately. The inefficiency of these models results in low sample throughput. To mitigate the issues with slow throughput, companies have recently developed fully automated CMM technologies. Automated platforms that have already been used to isolate CTCs include the AVISO CellCelector and the RareCyte CyteFinder/CytePicker.^23,24 Being fully automated, these technologies present major advantages in sample processing and throughput, but these benefits come at a steep price—they can be as much as 100 times more expensive than manual CMMs. In order to make advanced molecular analysis of CTCs more accessible to researchers, a less expensive, high-throughput, single-cell manipulation technique is needed to couple with upstream bulk enrichment technologies.

To address the need for an affordable yet robust CMM platform, we have developed a semiautomated system for purifying rare-cell populations. This system is designed to work in tandem with various upstream LOC bulk enrichment technologies with a fabrication cost of less than $3000 ( Suppl. Table S1 ). Furthermore, this system—referred to as the semiautomated single-cell aspirator (SASCA)—integrates modularly with standard inverted microscopes that are already present in many modern laboratories, especially those investigating CTCs ( Fig. 1A ). This integration reduces cost by obviating the need to purchase expensive equipment. In addition to microscope integration, the SASCA couples with established LOC technologies to immobilize cells and increase throughput ( Fig. 1B ).

Figure 1.

Technology Integration—the SASCA platform is engineered to integrate with inverted microscopes. (A) The aspiration apparatus itself is made up of several components, including the multiaxis micromanipulator (i), the mounting cantilever (ii,v), and the 10 µL syringe pump (iii). The syringe is connected via microtubing (iv) to the glass pipette (50 µm inner diameter) (vi). (B) The platform attaches easily to the microscope (i) and integrates well with the processing device (MAs, silicone isolators, and PCR caps) (ii). (C) To entrap the cells, the samples are seeded into MAs. (D) Samples can be imaged and processed efficiently, using the microscope’s own x, y, z mechanical movements to navigate to cells expressing target markers (EpCAM, green) (i). When identified, target cells can be aspirated by lowering the z plane of the microscope itself, which drops the pipette tip into the target well for aspiration (ii). Afterwards the z plane is raised back to the focus plane, bringing the pipette tip out of the well (iii).

For years microwell arrays (MAs) have been used in microfluidics to isolate individual cells for various single-cell assays. Previous applications have included single-cell PCR, cytokine secretion, and natural killer cell interactions.^25–27 For the SASCA, sample cells are evenly distributed in a polydimethylsiloxane (PDMS) MA to immobilize and contain the cells, thereby enabling robust single-cell selection and negating the worry of accidental contaminant aspiration as seen with some glass slide setups ( Fig. 1C ). The technique of pairing a CMM with a MA was previously demonstrated in 2009 by Ozawa et al. where a manual CMM was used to select rare targets from a heterogeneous population. Specifically, they isolated rare B-cell populations expressing type A influenza nucleoprotein from the blood of vaccinated patients for downstream antibody generation, illustrating the enrichment capabilities of the method.²⁸

We further improved our final purity by matching experimental results with a Pólya-Aeppli probability distribution to optimize our microwell seeding method and developed a custom software that incorporates real-time detection and localization of target cell subpopulations. To isolate a cell in a MA that has been identified using the software, the SASCA leverages the microscope’s own mechanisms to navigate the MA in an automated fashion. These components contribute to improved objectivity of target cell selection, throughput, and sample purity. In this study we tested the efficacy of the SASCA platform in rare-cell purification through the collection of CTCs from the blood of patients with prostate cancer, showing that high-throughput phenotypic analysis of such samples can be accomplished at an affordable price.

Materials and Methods

Semiautomated Single-Cell Aspirator

The single-cell aspirator presented in this study is an assemblage of commercially available items and custom engineered parts. A four-axis mechanical CMM (MX130R; Siskiyou, Grants Pass, OR) was used to hold a glass micropipette (50 µm tip inner diameter, MGM 1C-50; FIVEphoton Biochemicals, San Diego, CA) and perform initial pipette alignments. The CMM was bolted to an aluminum mounting cantilever that was custom machined (dimensions listed in Suppl. Fig. S1 ) out of aluminum square bar stock (⅝ in., 2EZP2; Grainger, Lake Forest, IL). The cantilever base was machined out of aluminum flat bar stock (¼ × 3 in., 2EYZ6; Grainger) and embedded with three neodymium magnets (one R822CS-N52 and two R622CS-N52; K&J Magnetics, Pipersville, PA), which held the base in place when it was slid into the mounting plate ( Suppl. Fig. S2A–C ). The mounting plate was made of a 1/64 in. thick steel sheet (10 × 4 in., 16NH24; Grainger) that was bent into the shape of the cantilever base. The steel mounting plate was attached to the top of the microscope’s motorized nosepiece housing ( Suppl. Fig. S2B ) using double-sided tape (ARCare, 90106, Glen Rock, PA). Fluid movements for aspirating and dispensing cells were controlled by a manual microinjector syringe pump (10 µL syringe model; Sutter Instruments, Novato, CA), which was connected to the CMM via a 3 ft length of plastic microtubing (054397; United States Plastic Corp., Lima, OH).

Microwell Arrays

MAs were fabricated out of PDMS (Dow SYLGARD 184, Midland, MI) using soft lithography replica molding techniques.^29,30 A mold for the MA was fabricated using established SU-8 photolithography techniques³¹ and placed in a plastic petri dish. Using a standard 1:10 ratio of cross-linking agent to prepolymer, we mixed and de-gassed a PDMS solution before pouring it over the molds. The PDMS was cured at 70 °C for 4 h and allowed to cool for at least 1 h before removal from the mold. Finally, before they were ready for use, the MAs were cleaned using a Soxhlet extractor filled with 70% EtOH.

SASCA Media

During the single-cell aspiration process, two specialized solutions were used interchangeably to prevent cell adhesion and maintain cell viability. In initial experiments, a specific SASCA media was developed using Dulbecco’s modified Eagle’s medium (DMEM) with 4.5 mg/mL glucose, 0.1% bovine serum albumin (BSA; A9056; Sigma-Aldrich, St. Louis, MO), 0.1% Pluronic F-127 (P2443; Sigma-Aldrich), and 10 mM HEPES buffer (15630-080; Gibco, Grand Island, New York). Pluronic F-127 was specifically used because it has been shown to inhibit cell adhesion in biomedical research.^32–34 We also found a commercially available phosphate-buffered saline (PBS) blocking buffer (SuperBlock; Thermo Fisher, Waltham, MA) to be a suitable SASCA media replacement when performing cell line nucleic acid extraction experiments.

Microwell Loading

MAs were placed on precleaned microscope slides (12-550-15; Fisherbrand, Hampton, NH) before being loaded. Due to the hydrophobicity of PDMS, the microwells do not spontaneously fill with media. To promote microwell filling, the empty bull’s-eye MAs were placed in a vacuum chamber for 30 min as previously described.³⁵ Immediately following PDMS gas evacuation, 60 µL of either SASCA media or SuperBlock buffer was pipetted onto each bull’s-eye—within the inner edge of the “moat”—and allowed to fill for 15 min. After the microwells had filled, the samples were added.

The volume of the patient samples following initial enrichment could vary from 50 to 80 µL, which was near the maximum volume a bull’s-eye could contain within the inner rim of its moat,³⁵ so before adding the sample to its respective bull’s-eye(s), 50 µL of the initial 60 µL of liquid used to fill the device was removed. Leaving behind 10 µL ensures that the wells remain filled and covered with a thin film of media; if this film is allowed to evaporate, the wells will not refill without requiring an additional vacuum chamber treatment. Next, the sample is pipetted onto its respective bull’s-eye MA, being careful not to spill into the moat and risk losing cells.

Once the sample is seeded, the MA device is placed in a covered omni-tray and left in a darkened chamber for the cells to settle into the microwells. The cells were allowed to settle for approximately 15 min, but no less than 11 min because the theoretical settling velocity of cells (2.8 cm/h) suggests that a cell at the top of a 5 mm tall liquid droplet requires almost 11 min to reach the bottom.³⁶ After the cells are fully settled, the moat is filled with 10 µL of media or buffer, merging with the sample droplet and pinning it within the outer bounds of the bull’s-eye moat. The moat-filling step is critical for subsequent bright-field imaging.

Cell Seeding Simulations

To quantify how different protocols for cell collection might influence final sample purity, we simulated the process of cell seeding and collection. After performing a virtual cell seeding to randomly distribute 250 target cells with different levels of background cells, microwells with target cells were collected in two ways. Using a “naive” approach, we collect each microwell that is encountered containing target cells. Using an optimized or “ranked” approach, we visit microwells in order of decreasing purity. In both cases, collection proceeds until 50 target cells are collected. The simulation was rerun 60 times for each level of initial sample purity using R statistical software. Collected/final sample purities (each containing 50 target cells and a variable number of background cells) were then calculated. Final purities using the naive and targeted (i.e. ranked) approach were then compared.

Automated Microscopy and Target Cell Identification

The SASCA system was mounted to a Nikon Eclipse Ti-E inverted fluorescence microscope (Nikon USA, Melville, NY), and images were acquired using a 10× objective (CFI Plan Fluor series; Nikon) in bright-field and three fluorescence channels: 390 nm (Hoechst), 560 nm (EpCAM), and 648 nm (Exclusion). To identify target cells, a montage of tile images was acquired across the entire bull’s-eye MA and analyzed using custom software written in MATLAB (The MathWorks, Natick, MA). This process was fully automated and performed in real time during image acquisition in order to minimize the time spent prior to the aspiration of cells. When a new tile of the montage is acquired, the software recognizes it as it appears in the monitored folder. Once recognized, the tile is processed and scanned for cells. It should be noted that while the real-time image processing program created for this platform was written in MATLAB, it runs independently of MATLAB software, so users do not need to purchase and install additional software to run the program.

The analysis algorithm involves distinct processes specific to the different image channels. All cells were detected based on fluorescence intensity using the Hoechst channel. The mask of detected nuclei was combined with that of microwells detected in the bright-field channel in order to restrict interest to cells lying within microwells. Fluorescence images for EpCAM and Exclusion were background subtracted, and then the mean fluorescence intensity (MFI) within detected cells was calculated. After processing all MA images, a plot of EpCAM versus Exclusion MFI for all cells was displayed in a graphical user interface (GUI) and thresholds were manually set on this plot to define the target and exclusion cell staining profiles. Cells that were EpCAM⁺/Exclusion⁻ were considered to be CTCs, while peripheral blood mononuclear cells (PBMCs) were defined as EpCAM⁻/Exclusion⁺. For the subsequent image-guided aspiration, a list of CTC locations was generated and imported into the imaging software as a multipoint list. By clicking through each point on this list, the microscope stage automatically locates to the specific cell coordinates, allowing the user to quickly navigate a sample.

In some experiments where cell recovery was being characterized, the bull’s-eye MA was again imaged following aspiration to confirm removal of all target cells, and the collected cells were separately imaged to evaluate target cell purity. However, this follow-up imaging step was not performed during experiments where nucleic acid endpoints were the focus, as maintaining cell viability was prioritized. In transcriptomic experiments the cells were aspirated and immediately dispensed into lysis buffer to protect molecular integrity.

Cell Aspiration from Microwell

The following steps were used to ensure successful cell aspiration from microwells. With the tip positioned above the target microwell ( Fig. 1D-i ), the glass pipette was lowered into the well by lowering the z plane of the microscope ( Fig. 1D-ii ). This process brings the cells slightly out of focus, yet the blurred image of the target cell can still be tracked visually when using 10× magnification. Slight deformation of the microwell’s elastic wall was used as an indicator for when the pipette tip had been properly lowered into the well ( Fig. 1D-ii ). Next, the syringe pump is used to generate a suction force within the pipette tip, aspirating the contents of the well. Finally, the pipette tip is raised out of the microwell as the level of the z plane is returned to its initial position, bringing the cells back into full focus ( Fig. 1D-iii ).

Dispensing Cells

Aspirated cells can be dispensed into a droplet of buffer in various fluid containers. Specifically, in this study cells were dispensed into situation-specific buffers held within either a silicone isolator well array (JTR10S-A-2.0; Grace Bio-Labs, Bend, OR) or a clear, flat-bottom PCR strip cap. The buffers used in this study included the SASCA media, SuperBlock buffer, and lysis buffer (RLT Plus + βME). The volumes of the buffers used also depended on the specific endpoint (i.e., cell imaging and quantification, and RNA or DNA extraction). The silicone isolators were placed adjacent to MAs on the glass slide, and PCR strip caps were placed proximal to the slide on the sample processing device. This limits the time spent between aspirating and dispensing cells, promoting maintenance of cell viability. Silicone isolators were used for imaging and quantifying aspirated cells to characterize target cell recovery and final purity, with each well containing 35 µL of SASCA media. For the purposes of mRNA extraction, isolators were filled with lysis buffer. Alternatively, for whole genome amplification (WGA), cells were dispensed into inverted PCR caps containing 1–4 µL of PBS. Cell transfer to the caps was visually confirmed by keeping the micropipette tip in focus while imaging in real time as cells were dispensed using the microsyringe pump. Tubes were then placed on the caps, reoriented to the upright position, and briefly centrifuged before continuing with the WGA protocol.

Cell Culture

The prostate cancer cell line 22Rv1 (ATCC CRL-2505, Old Town Manassas, VA) was a gift from Scott Dehm at the University of Minnesota. Monocytic cell line THP-1 (ATCC TIB-202) was a gift from Fotios Asimakopoulos at the University of Wisconsin. Prostate cancer line LAPC4 was a gift from Doug McNeel at the University of Wisconsin. Prostate cancer cell line C4-2B was a gift from Felix Feng at the University of California, San Francisco. LAPC4 cells were cultured in DMEM media (Corning Cellgro, Corning, NY) supplemented with 20% fetal bovine serum (Gibco). Remaining cell lines were cultured in RPMI 1640 medium (Corning Cellgro) supplemented with 10% fetal bovine serum. All media included 2% penicillin-streptomycin (HyClone, South Logan, UT). THP-1 medium was also supplemented with 1% Gibco GlutaMAX (LifeTech Carlsbad, CA).

Cell Preparation

Cells were isolated by trypsin digestion and washed with 1× PBS + 0.1% BSA. 22Rv1 cells were stained with anti-EpCAM (phycoerythrin; VU-1D9; Abcam, Cambridge, UK) and THP-1 cells were stained with anti-human CD45 (AlexaFluor 647; BioLegend, San Diego, CA). Both cell lines were also incubated with Hoechst 33342 (Life Technologies) to visualize nuclei.

Blood Processing and Cell Isolation

Whole blood was collected from patients with metastatic prostate cancer. The study was approved by the University of Wisconsin Institutional Review Board and patients supplied written informed consent. PBMCs were isolated using a Ficoll-Pacque Plus (GE Healthcare, Chicago, IL) gradient before undergoing CD45 depletion (MACS; Miltenyi Biotec, Bergisch Gladbach, Germany). The VERSA^13,37,38 platform was used to capture CTCs with an anti-EpCAM antibody (R&D, Minneapolis, MN) and stain for extracellular markers EpCAM (phycoerythrin; VU-1D9; Abcam), CD45, CD34, and CD11b (AlexaFluor 647; BioLegend).

Reverse Transcription, Preamplification, and Quantitative RT-PCR

Preliminary mRNA extraction (see Fig. 4B ) to demonstrate the enrichment ability of the aspirator was completed as described by Strotman et al.³⁹ Purification of total RNA at low cell numbers (see Fig. 5A ) was performed using the RNEasy Plus Micro Kit (Qiagen, Hilden, Germany). cDNA was generated using the High Capacity cDNA Reverse Transcriptase Kit (LifeTech) and Bio-Rad C1000 Thermo Cycler (Bio-Rad, Hercules, CA), as per manufacturer guidelines. The generated cDNA was then amplified for 14 cycles using a Taqman PreAmp Master Mix (LifeTech) before being diluted 1:20 in 1× TE (10 mM Tris-HCL, pH 8, 1 mM EDTA). For TaqMan assays, 5 µL of diluted cDNA template was mixed with 10 µL iTaq master mix (Bio-Rad), 1 µL TaqMan Gene Expression Assay (Life Technologies), and 4 µL nuclease free (NF) water. Each reaction was amplified for 45 cycles (denatured at 95 °C for 15 s followed by annealing at 60 °C for 1 min) using a CFX Connect Real-Time PCR System (Bio-Rad). The primers used are listed in Supplemental Table S2 .

Whole Genome Amplification

Cells isolated using the SASCA were dispensed directly into a droplet of PBS placed at the bottom of a PCR tube before undergoing lysis and WGA using a modified version of the Repli-G Single Cell Kit (Qiagen). Recommended modifications to the kit as listed in Sho et al. included a 30 min lysis period (as opposed to the 10 min recommended in the kit’s protocol) and division of the amplification step into 16 individual reactions. We adjusted this methodology to include the extended lysis period but reduced the number of parallel reactions to eight. Amplified genomic DNA underwent a preamplification step before undergoing qPCR as described above. The primers used are listed in Supplemental Table S2 .

Results

SASCA Platform

Microscope Integration

The SASCA consists of three main parts: (1) the mechanical micromanipulator ( Fig. 1A-i ), (2) the magnetic mounting cantilever ( Fig. 1A-ii ), and (3) the manual syringe pump ( Fig. 1A-iii ). The syringe pump and micromanipulator are connected by a length of plastic microtubing ( Fig. 1A-iv ), to allow precise fluid movements. This manual syringe pump was chosen over more advanced models because it was less expensive, but with a 4.2 nL resolution it still provides adequate control when aspirating cells. The mounting cantilever ( Fig. 1A-vi ) attaches the micromanipulator to the z axis of the microscope via a magnetic base ( Suppl. Fig. S2A ) rather than the microscope stage ( Fig. 1A-v ). By mounting to the z axis, the micropipette tip ( Fig. 1A-vi ) can be easily moved in and out of wells using the focus knob on the microscope. A shaped steel plate provides a ferromagnetic mounting pad for the cantilever and is a permanent fixture on the microscope, though it does not affect the operation of the microscope ( Suppl. Fig. S2B ). Moreover, it allows for quick and precise attachment of the aspirator to the microscope by guiding the cantilever base into position where it is held in place by the magnets ( Suppl. Fig. S2A–C ).

The unique mounting method implemented by the SASCA platform provides two significant advantages over traditional mounting techniques. First, this approach holds the aspirator stationary in the x-y directions relative to the microscope stage ( Fig. 1B ). To navigate the microwell platform, the stage is moved using the microscope’s built-in x-y joystick. This allows the pipette tip to remain fixed in the field of view while the user scans the sample ( Fig. 1C ). This concept of camera following the end effector is similar to what is used on specific manufacturing robots that implement object-related machine vision.⁴⁰ The technique uses a camera (i.e., microscope objective) that is fixed in such a way that it continuously follows the motion of the end effector (i.e., micropipette tip). By keeping the aspirator tip in the field of view at all times when processing a sample, the processing time is greatly reduced because there is no need to adjust CMM alignment knobs as with mechanical CMMs that use traditional mounting techniques.

The second significant advantage of the SASCA’s unique mounting approach involves the movement of the aspirator tip in and out of target microwells. When a cell is being aspirated, the z-axis mounted CMM is moved up and down the z plane using the microscope’s objective focus knob ( Fig. 1D ). This major advantage provides quick yet precise movements in and out of microwells, providing the user with confidence that they are selecting their target of interest and not perturbing other cells. By exploiting the microscope’s own maneuvering mechanisms, the SASCA platform is able to achieve the precision and accuracy of automated platforms while remaining affordably integratable with existing microscopy systems.

Microwell Integration

MAs can be used to significantly aid single CTC isolation.^23,41,42 Viscous forces at the microscale help to slow fluid flow within the microwells, thereby preventing cells from being dislodged or removed by exterior flows. In the context of single-cell aspiration, this allows the contents of one microwell to be aspirated without aspirating contaminants from adjacent wells ( Fig. 1D ). We have enhanced the interface between users, aspirators, and microwells by integrating a pipette-friendly MA design that we recently developed.^35,43 The original bull’s-eye MA design was modified slightly for this rare-cell application by removing the recessed region in the center to maximize the number of microwells, reduce dead volume, and minimize cell loss. The resulting microwells had dimensions of ~50 × 50 × 50 µm (well volume of 125 pL), with the well-to-well spacing in the array also being 50 µm in either the x or y direction ( Fig. 1C ).

To contain samples into finite arrays, a circular moat (250 µm deep) was fabricated around an array of ~2800 microwells, giving the MA a diameter of 6 mm. The moat provides a surface tension-driven barrier at the exterior edge of the sample. Unfortunately, refraction and interference near the periphery of the sample droplet resulted in dark regions in the bright-field channel, obfuscating the microwells ( Suppl. Fig. S2D ). This was problematic for our software-based identification, because it only considered cells within software-detected microwells. To address this, a moat-filling step was added after the cells had settled. To eliminate any bright-field shadow near the edge of the MA, the moat is filled with 10 µL of extra media so it merges with the sample ( Suppl. Fig. S2E ). It was important to only fill the moat after the cells had settled into the microwells to prevent the loss of cells. Moving the droplet edge shadow beyond the perimeter of the MA greatly improved the success of identification and aspiration of cells near the periphery of the bull’s-eye, which ensured that no rare cells were excluded from software detection.

Optimizing Microwell Seeding and Collection

When the MA is seeded with a heterogeneous suspension of cells, it is possible for multiple cells to settle into a single microwell, which can result in a mixed population of cells occupying the same well. Because sample purity is of the utmost importance, the goal of microwell seeding is to maximize the ability to isolate and collect the rare target cells while minimizing collection of contaminating background cells. To optimize this process, one must be able to predict/estimate how many cells will occupy each microwell. Traditionally, a Poisson distribution has been used to model microwell seeding;^44,45 however, this model does not take into account the fact that cell suspensions are generally not monodisperse (i.e., some cells cluster with one another). To account for the presence of clusters, we used the closely related Pólya-Aeppli distribution instead.

The Pólya-Aeppli distribution is typically used to model the “Poisson delivery” of clusters of items where the number of items per cluster follows a geometric distribution. Two parameters define the Pólya-Aeppli distribution. The first parameter is identical to the standard Poisson parameter, λ (i.e., average number of clusters delivered to a microwell). The second parameter, p, defines the geometric distribution of cluster sizes (i.e., the number of cells observed in the different clusters). In this case, p = 1/(1 + average number of cells per cluster). Figure 2A illustrates how the probability distribution of cells per well shifts when settling objects are considered to be cell clusters (one to six cells per cluster) rather than a monodisperse suspension.

Figure 2.

Optimizing cell seeding with probability distributions. (A) The probability distribution of cells per microwell shifts if objects settling into wells are considered to be clusters of cells rather than individual cells. The occurrence of single-cell clusters remains the most probable, but accounting for multicell clusters increases the probability of wells containing two to five cells. (B) Traditional cell seeding distribution is modeled using a Poisson distribution compared with experimental cell line seeding conditions with a range of four different cell–well ratios (1:1, 1:2, 1:4, and 1:8). (C) However, a Pólya-Aeppli distribution accounts for multicell clusters and proves to be a much better predictor of cell distribution in MAs using the same data as in B. The shaded ribbons shadowing each model line represents the 95% confidence interval.

To confirm that a Pólya-Aeppli distribution improves upon established microwell seeding models, experimental cell-count distributions were measured from MAs seeded at different densities. These experimental distributions were fitted with either a Poisson distribution ( Fig. 2B ) or a Pólya-Aeppli distribution ( Fig. 2C ) using maximum likelihood estimation (R, stats4 and polyaAeppli packages^46,47). The Pólya-Aeppli distribution followed the data much more closely and was able to model changes in suspension density via changes in λ while maintaining a consistent estimate of the average cluster size. To our knowledge, this represents the first proposal and demonstration of using the Pólya-Aeppli distribution to better model cell seeding distributions in MAs.

Given that some microwells will contain 100% target cells (i.e., 100% purity) while others will contain differing numbers of contaminating cells (i.e., < 100% purity), we proposed to optimize the collection of seeded target cells by visiting microwells in order of decreasing purity. This ranked approach can be compared with a naive approach, which is defined as indiscriminately collecting cells from any microwell containing at least one target cell (i.e., in random order of purity). The value of a ranked approach can be seen in the simulations summarized in Supplemental Figure S3 . If a naive approach is used, the simulated results suggest that a sample purity ≥50% can be achieved if the cell-to-well ratio is kept at or below 1:1 ( Suppl. Fig. S3A ). If a ranked approach is used, the final purity increases to 100%. Given this result, we developed a semiautomated collection method that enables the ranking of microwells by purity to increase the final purity.

Semiautomated Processing

After cells are seeded into microwells, target cells can be identified in two separate ways. With the manual image and capture (MIC) technique used above, the user scans through the microwell by eye, cycling through the fluorescent channels along the way to identify target cells. Once a target cell has been discovered, it is subsequently aspirated. This approach becomes tedious, even with the high maneuverability of the SASCA. In comparison, the automated image and manual capture (AIMC) method incorporates real-time image processing software into the workflow. The software prescans a sample for cells exhibiting phenotypes of interest. In less than 5 min, the program identifies potential targets, records and processes data on these targets, and outputs a list of discovered cells (and their locations) to assist the user in their decision-making process ( Fig. 3A ). All of these functions are performed simultaneously while the sample is scanned by a microscope macro. With AIMC, the user is directed toward potential targets, enhancing both the output purity achieved and the speed at which targets are isolated.

Figure 3.

Semiautomated detection of CTCs. (A) Cell detection algorithm using multichannel images. Cells that were EpCAM⁺/Exclusion⁻ were considered targets (red box). The locations of these cells were used to guide subsequent aspiration. (B) Scatterplots of fluorescence intensity of two mixed cell line populations in three distinct bull’s-eye samples. The bull’s-eye MAs were imaged before and after aspiration, demonstrating the successful identification and collection of a rare target cell subpopulation (yellow ellipse) within heterogeneous samples. The cells were seeded onto MAs at an approximate initial purity of 1%.

The coordinate list generated is loaded into the imaging software and used to quickly navigate from one point of interest to another. Once the cell of interest is identified, the user validates whether or not the cell should be targeted for aspiration. To confirm that this semiautomated identification and navigation can reliably collect rare populations of cells within a sample, heterogeneous mixtures of background and target cells were prepared at an initial concentration of approximately 1% target cells. This heterogeneous sample was seeded into three separate bull’s-eye MAs and processed using the AIMC technique. By collecting only cells that were identified as targets by the algorithm, it was shown that every target cell in each of the three samples was identified and collected ( Fig. 3B ).

The semiautomated method of identification and navigation increases processing speed without compromising cell recovery and final purity, allowing users to quickly scan through the samples in a directed fashion. The workflow for the SASCA is divided into three parts: upstream LOC bulk enrichment, SASCA cell purification, and downstream rare-cell interrogation ( Fig. 4 ). LOC bulk enrichment is vital because it aims to decrease the number of cells to a manageable quantity for seeding into MAs. Platforms available for upstream CTC isolation rely on a combination of immunoaffinity mechanisms (positive/negative selection), size filtration, electrophoresis, and inertial focusing.⁴⁸ However, as described above these platforms do not provide single-cell resolution, necessitating further enrichment using the SASCA. Once pure populations of viable cells have been isolated, they can be interrogated for molecular heterogeneity with deep genome sequencing, used for organoid culture, or transciptomically analyzed. These downstream analytic methods can offer more comprehensive insight into the genomic heterogeneity and CTC diversity found in patients.

Figure 4.

SASCA sample throughput. Top: The general SASCA workflow begins with the initial bulk enrichment of CTCs from whole blood using any variety of LOC enrichment techniques. Once the bulk sample is enriched, it is processed using the SASCA platform. The sample is seeded into a MA (20 min), the sample is imaged, cells are identified in real time (optional; 4 min), and targets are gently aspirated for downstream assays. Bottom: When comparing MIC and AIMC protocols, it was found that the AIMC protocol increased aspiration rate (*p = 0.03) while not sacrificing final purity or CTC transfer efficiency (n = 3).

To illustrate the improvement of aspiration with AIMC, heterogeneous cell mixtures were made with an initial target cell purity of approximately 1% and 10%. These ratios were chosen because they should be representative of sample purity after an initial bulk enrichment step is completed. Target cells were identified using either the MIC or AIMC method, before being aspirated, dispensed, and counted to quantify output purity and transfer efficiency. As shown in Figure 4 , both methods provided highly pure samples (90%–100%), while very few cells were lost between aspirating and dispensing (85%–95% transfer efficiency). We observed increased output purity and speed of target isolation when using the AIMC method relative to the MIC. In fact, as impurity increased, aspiration speed increased as well (p = 0.03; Fig. 4 ), demonstrating the value of semiautomated processing for providing viable, highly pure CTC samples for advanced cell interrogation.

Aspiration Enriches qPCR Endpoint

To test the isolation capabilities of the SASCA, we focused on the enrichment of prostate cancer CTCs because they offer a heterogeneous system with a rare subpopulation of cells that have been shown to provide insight into disease progression and drug resistance. To evaluate the accuracy of the SASCA, we tested our ability to separate individual prostate cancer cell lines from a mixed cell sample based on phenotypic differences. To replicate a VERSA-enriched CTC sample, a human prostate cancer cell line (22Rv1) was spiked into samples of a human monocyte cell line, THP-1. We began by labeling the 22Rv1 cell line using anti-EpCAM-PE antibodies and the THP-1 cell line with anti-CD45-AlexaFluor 647 and mixing the cells at a ratio of 1:100. Pure samples of both 22Rv1 and THP-1 cells were also used as starting populations for aspiration to provide a pure cell benchmark. Each cell mixture was added to one or multiple bull’s-eye MAs to ensure at least 50 cells were available for aspirations at a distribution of 720 cells per MA. The 22Rv1 cells were aspirated using the SASCA with the MIC method, prior to implementation of the real-time imaging software. To quantify the accuracy of this process, the SASCA purified populations were imaged and manually identified as either a 22Rv1 target cell or THP-1 contaminant cell, as per their staining phenotype. Using this method, we achieved an output purity of 83.8% with the SASCA platform ( Fig. 5A ).

Figure 5.

Cell line enrichment. (A) Fluorescently labeled 22Rv1 and THP-1 cells were combined at a ratio of 1:100. The cell mixture was added to MAs and 22Rv1 cells, identified using the MIC method, and then isolated using the SASCA platform. Cells were imaged to quantify percent purity achieved (upper corner of scatterplot). Pure cell populations were also isolated. (B) mRNA was extracted from approximately 50 cells from the initial cell populations indicated, as well as cells aspirated and described in A. qRT-PCR was performed using primers specific for each gene listed. Data describe the relative expression of each gene normalized to the expression of POLR2A and P0 housekeeping genes. Expression was represented by a color scale from lowest (green) to middle (black) to highest (red). White indicates that no expression was detected. (C) C4-2B cells were seeded onto MAs, and the SASCA platform was used to isolate low cell numbers. Cells were dispensed into lysis buffer before undergoing mRNA extraction, reverse transcription, and qRT-PCR for primers specific to each listed gene. (D) LAPC4 cells were dispensed into PCR tubes before undergoing WGA and qRT-PCR.

RNA was isolated from the aspirated cells to evaluate enrichment of prostate specific transcripts. The chosen genes are all involved in AR signaling (AR splice variant AR_4/5, TMPRSS2, and the prostate specific antigen KLK3), the expression of which occurs in 22Rv1 cells but not THP-1 cells. Figure 5B represents the qPCR results from this experiment. As expected, pure 22Rv1 cells expressed prostate specific markers and THP-1 cells expressed high levels of leukocyte marker PTPRC (CD45) in both the prepurified and aspirated samples. Compared with the pure THP-1 population, we observed reduced expression of PTPRC in the aspirated prostate cancer sample. Additionally, all prostate specific markers showed increased expression in the aspirated sample, indicative of a highly enriched 22Rv1 population. Here, it is helpful to consider the composition of the initial cell mixture populations. Assuming an estimated 50 cells in each mixture prior to purification, only 0.5 cells would be found in a 1:100 mixture. This is reflected by the lack of expression of prostate specific markers. However, after purification with the aspirator, the expression of these markers is significantly improved. This corresponds to a greater concentration of 22Rv1 cells in the final population of cells, indicating that the aspirator successfully enriched the target 22Rv1 population. Though we demonstrate our ability to precisely identify and isolate rare-cell subpopulations, we further optimized the aspiration by automating the image processing software and applying the AIMC technique to isolate target cells from heterogeneous cell mixtures, with the aim of increasing purity and decreasing sample acquisition time.

Molecular Analysis of Low Cell Numbers

To determine the minimum number of cells necessary for obtaining transcriptomic and genomic endpoints, we captured defined numbers of cells with the SASCA platform before subjecting them to downstream analysis. First, the prostate cancer cell line C4-2B was used to assess mRNA extraction and whole transcriptome amplification at low cell numbers. Starting with three different cell inputs (10, 5, and 1), we were able to successfully amplify all targeted prostate specific genes (AR splice variants AR 4/5 and AR V7, and androgen-signaling NKX3.1) as well as housekeeping genes POLR2A and RPLOP0 for all starting inputs ( Fig. 5C ). All Ct (threshold cycle) values reported were under 35. A clear trend is visible between increased cell number and decreased Ct value, demonstrating our ability to isolate precise numbers of cells and assess their gene expression.

Next, low numbers of aspirated LAPC4 cells (15, 12, 5, 3, and 1) were isolated with SASCA, followed by WGA. Amplification of genomic DNA was assessed using a qPCR assay composed of genes known to play a role in cancer.⁴⁹ Using a threshold for detection set at 36 Ct values, we chose a binary rating system to assess whether or not we had attained high-quality WGA DNA ( Fig. 5D ). A starting input of ≥5 cells appears to provide a lower threshold for cell input, above which we are able to retain consistent quantities of DNA that should be reliable enough to perform downstream applications. Inputs of ≥5 cells gave expression readouts on all target genes, nearly half of genes for 3 cells, and only 2 genes for 1 cell. We hypothesize that due to a limited DNA template from which to amplify, stochastic amplification occurs, thus providing inexact WGA products isolated from less than five cells. As seen in Figure 5D , an increase in cellular input lessens this variability and is likely necessary for performing optimal downstream sequencing analysis. These findings are concordant with those from Sho et al., which showed that a cellular input of 5–10 cells is needed to reliably obtain amplified WGA products suitable for precise next-generation sequencing and aCGH (array comparative genomic hybridization) analysis. Although we were able to obtain a low degree of single-cell resolution for downstream genomics, we believe that a higher starting cellular input (5–10 cells) is imperative for generating amplified DNA with high fidelity. With regard to CTCs, it is not unreasonable to surmise that at least five CTCs would be found in a peripheral blood draw sample and accordingly be seeded into a MA for discovery and aspiration. Consequently, so long as the minimum threshold of cellular input is reached, the integration of the SASCA platform with a modified WGA protocol should yield high-quality WGA products from CTCs satisfactory for downstream sequencing analysis. Successful sequencing results enable a method to characterize the genomic landscape of a patient’s disease, thereby increasing the clinical utility of a liquid biopsy.

Aspiration Selection of CTCs in CRPC

We applied the SASCA technology to interrogate CTCs from patients with castration-resistant prostate cancer (CRPC). CTCs were first enriched using the VERSA, an EpCAM antibody-based positive-selection system. Cells were stained in the VERSA with EpCAM and Exclusion markers (CD45/CD34/CD11b), as well as a nuclear stain. Captured cells were transferred to MAs and then imaged to identify target cells. EpCAM⁺/Exclusion⁻ events were identified as CTCs, while EpCAM⁻/Exclusion⁺ events were identified as PBMCs. As seen in Figure 6A , CTCs and PBMCs showed clear phenotypic distinction upon fluorescent microscopy analysis. Cells from both populations were isolated from the same patient using SASCA and separately underwent RNA extraction ( Fig. 6B ). Gene expression analysis of CTC and PBMC populations showed expression of prostate specific markers exclusively in the CTC population, including the AR splice variant AR_V7, the presence of which has been suggested to act as a predictor of therapeutic resistance to both abiraterone acetate and enzalutamide. These results are concordant with the clinical narrative of the patient, who, after undergoing disease progression on abiraterone, did not show a clinical response to enzalutamide. While only demonstrated in a single patient, these results are indicative of the clinically relevant information that can be attained when using the SASCA platform to isolate and interrogate rare-cell populations. It is worth noting that CD45 (PTPRC) expression is observed in the CTC population. This could be explained by either a contaminant PBMC being collected in the CTC population or the presence of CD45^low CTCs, which have been previously reported to exist.^50,51

Figure 6.

Enrichment of patient sample. (A) Images of representative CTCs and PBMCs isolated from whole blood of a patient with CRPC using the SASCA platform, after initial enrichment with the VERSA. Cells are stained for Hoechst, EpCAM, and an exclusion channel (CD45, CD34, CD11b). Scale bar = 25 µm. (B) Matched qPCR gene expression profile of isolated CTC and PBMC populations from the same patient.

Discussion

The SASCA platform leverages multiple technologies to improve purification of live rare-cell populations. Here, we describe its application in CTC isolation from liquid biopsies. Our initial CTC enrichment is performed using the VERSA device, which employs EpCAM-labeled paramagnetic beads to bind and extract EpCAM-expressing cells. Although this method provides a high degree of enrichment, the purity is often insufficient for precise molecular analysis of CTC populations, which have been shown to hold clinically relevant information. Acquiring viable, purified cells is imperative for reliable molecular characterization, including transcriptomic and genomic assays. The presence of contaminating background cells can mask changes occurring in rare cells and obscure expression of potentially clinically relevant biomarkers. Additionally, the quality of genomic sequencing is enhanced with a higher-purity cellular input, allowing visualization of copy number changes, insertions/deletions, gene rearrangements, and mutations. To overcome this obstacle, the SASCA platform detects and isolates rare cells on an individual basis, according to their phenotype. Live cells are collected and can be used for a wide variety of downstream analysis applications.

We sought to develop a strategy that integrates LOC bulk enrichment approaches with single-cell selection in order to improve rare-cell purities. As we developed each component of this tandem approach, achieving high purity of isolated cell populations was our primary objective, which in the past could potentially hinder cell yield and throughput. However, in this platform we sought to improve purity while minimally diminishing cell yield and throughput. While there is potential for cell loss when transferring from the upstream LOC device to the SASCA, this step is still necessary to achieve the level of purity required for advanced downstream analytics. To this end, the SASCA was designed to maximize output purity by integrating with a standard inverted microscope and a MA device. The platform itself is mounted to the z axis of the microscope, improving the ease of sample navigation, which in turn improves sample throughput. Guided by the semiautomated software, the MA device facilitates confident rare-cell collection. Combining these features greatly increases output purity without sacrificing speed and recovery. Moreover, all of the components used to fabricate the SASCA platform are easily accessible and affordable, which could benefit other labs performing rare-cell studies, where fluorescent microscopes with motorized stages are already available and requisite to perform such studies.

Commercially available technologies exist for manipulating individual cells within a bulk population. Previous rare-cell studies have used CMMs to isolate and interrogate individual cells. CMMs currently are available as manual, motorized, or automated models. The more motorized and automated a model is, the more expensive it is; automated models can cost up to $300,000, while the simplest mechanical models start at $750. However, cost is inversely proportional with regard to maneuverability and sample throughput. The SASCA was designed to harness advantages from both systems. To circumvent these drawbacks, the SASCA was designed to be mounted to the z axis of the microscope. Similar to the way robotic manipulators function through keeping their end effector in view, the SASCA keeps the micropipette tip in the field of view of the user at all times by virtue of its z-axis integration. The advantage gained from this is ease of handling by the user, as well as reliable precision and increased throughput. We assess how the SASCA compares to other CMMs in Supplemental Figure S4 , where we consider design parameters we found to be important.

Similar to the SASCA, the CellRaft System is a commercially available platform that integrates microscopes with MAs to isolate subpopulations of cells, and it also is mounted to the z axis. CellRaft illustrates how the integration of microtechnology with microscopy can be leveraged to enable microscale applications and capabilities. While our study and CellRaft show the benefits of attaching a microscope modification to the z axis, users should verify with their microscope manufacturer to ensure it is able to withstand the addition of the SASCA platform.

In addition to our unique CMM mounting method and our semiautomated collection approach (AIMC), we have provided an optimized method of seeding heterogeneous samples into microwells. Previous attempts in the literature to optimize microwell seeding density have used a Poisson distribution to model how cells distribute to wells. Indeed, the Poisson distribution is very effective at modeling truly monodisperse suspensions of cells; however, in reality, cell suspensions typically also contain small clusters of cells as well. Therefore, we have chosen to use the Pólya-Aeppli distribution. The Pólya-Aeppli distribution models Poisson delivery of clusters of items (i.e., cells) where the number of items in the clusters follows a geometric distribution. This distribution modeled the experimental seeding data significantly better than the Poisson distribution. To our knowledge, we are the first to suggest and demonstrate application of the Pólya-Aeppli distribution to microwell seeding. Together, the tools described above were designed to be flexible and enable translation to alternative microscopy platforms, representing a significant contribution to rare-cell enrichment technologies.

Although the single-cell isolation platform introduced here was used for CTC interrogation, the capabilities of the SASCA are not limited to CTCs, as advanced single-cell isolation presents a variety of applications for rare-cell studies.⁵² Traditionally, molecular biology has studied tissue slices (tumor, brain, embryo, etc.), blood, or populations of cultured cells, which have varying levels of heterogeneity that researchers aim to better understand. For example, the analysis of neural progenitor cells could improve our understanding of developmental neural disease progression but requires the ability to isolate subpopulations of neural cells. Advances such as microdissection and single-cell isolation have allowed analysis of the heterogeneity of complex systems, which could provide new insights into disease progression.⁵³ As studies into the heterogeneity of diseased and developing tissue continue to increase, the SASCA provides an affordable, highly controllable platform to facilitate this research.

Conclusion

Improving the molecular evaluation of tumoral heterogeneity requires robust technology able to separate individual live cells for interrogation. To address these challenges, we have developed a platform that integrates low-cost technologies, such as MAs and manual micromanipulators, with real-time imaging software to isolate viable single cells based on phenotypic markers. Our platform successfully demonstrated enhanced purification of cell line subpopulations, resulting in the enrichment of gene expression endpoints. When applied to patient samples, we were also able to enrich both prostate cancer CTC and PBMC populations. The SASCA platform presents a highly controlled, affordable method for the isolation of subpopulations of cells that could provide the specificity and reliability needed for rare-cell investigation.

Supplemental Material

supplemental_material_for_SASCA_by_Tokar_and_Stahlfeld_et_al – Supplemental material for Pairing Microwell Arrays with an Affordable, Semiautomated Single-Cell Aspirator for the Interrogation of Circulating Tumor Cell Heterogeneity

Supplemental material, supplemental_material_for_SASCA_by_Tokar_and_Stahlfeld_et_al for Pairing Microwell Arrays with an Affordable, Semiautomated Single-Cell Aspirator for the Interrogation of Circulating Tumor Cell Heterogeneity by Jacob J. Tokar, Charlotte N. Stahlfeld, Jamie M. Sperger, David J. Niles, David J. Beebe, Joshua M. Lang and Jay W. Warrick in SLAS Technology

Footnotes

Acknowledgements

We would like to thank all patients who participated in this study. We are also grateful for the help of the UWCCC clinical research group, especially Jamie Wiepz, Kelly Bush, Amy Forsyth, Dorothea Horvath, Jane Straus, Mary Jane Staab, Dr. Glenn Liu, Dr. Douglas McNeel, Dr. Christos Kyriakopoulos, and Dr. Hamid Emamekhoo. Finally, we would like to acknowledge David J. Guckenberger for his engineering expertise, Croix Kimmel for his help in manufacturing devices, and Yasmin Alvarez-Garcia and Loren Stallcop for their help with contributions to the Pluronic F-127 protocol.

Supplemental material is available online with this article.

Declaration of Conflicting Interests

The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: David J. Beebe holds equity in Bellbrook Labs LLC, Tasso Inc., Stacks to the Future LLC, Lynx Biosciences LLC, Onexio Biosystems LLC, and Salus Discovery LLC. Joshua M. Lang holds equity in Salus Discovery, LLC.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by a Movember-Prostate Cancer Foundation Challenge Award (to J. M. Lang and D. J. Beebe), NIH grant no. 1R01CA181648 (to J. M. Lang), Department of Defense Synergistic Idea Development Award PC140746 (to J. M. Lang), NIH grant no. 5R33CA137673 (to D. J. Beebe), and University of Wisconsin Carbone Cancer Center support grant NIH P30 (to D. J. Beebe and J. M. Lang).

ORCID iDs

Jacob J. Tokar

Charlotte N. Stahlfeld

References

Ebinger

Ozdemir

Tiedt

; et al. Characterization of a Novel Dormant, Drug Resistant, Stem Cell Subpopulation in Acute Lymphoblastic Leukemia. Eur. J. Cancer 2016, 61, S207.

Avery

S. V.

Microbial Cell Individuality and the Underlying Sources of Heterogeneity. Nat. Rev. Microbiol. 2006, 4, 577–587.

Shen

Chen

Current Detection Technologies for Circulating Tumor Cells. Chem. Soc. Rev. 2017, 46, 2038–2056.

S. R.

; et al. Advances of Lab-on-a-Chip in Isolation, Detection and Post-Processing of Circulating Tumour Cells. Lab Chip 2013, 13, 3163–3182.

Schehr

J. L.

Schultz

Z. D.

Warrick

; et al. High Specificity in Circulating Tumor Cell Identification Is Required for Accurate Evaluation of Programmed Death-Ligand 1. PLoS One 2016, 11, e0159397.

Sieuwerts

A. M.

Kraan

Bolt-de Vries

; et al. Molecular Characterization of Circulating Tumor Cells in Large Quantities of Contaminating Leukocytes by a Multiplex Real-Time PCR. Breast Cancer Res. Treat. 2009, 118, 455–468.

Lin

Rivera-Báez

Fouladdel

; et al. High-Throughput Microfluidic Labyrinth for the Label-Free Isolation of Circulating Tumor Cells. Cell Syst 2017, 5, 295–304.e4.

Goldkorn

Ely

Quinn

D. I.

; et al. Circulating Tumor Cell Counts Are Prognostic of Overall Survival in SWOG S0421: A Phase III Trial of Docetaxel with or without Atrasentan for Metastatic Castration-Resistant Prostate Cancer. J. Clin. Oncol. 2014, 32, 1136–1142.

de Bono

J. S.

Scher

H. I.

Montgomery

R. B.

; et al. Circulating Tumor Cells Predict Survival Benefit from Treatment in Metastatic Castration-Resistant Prostate Cancer. Clin. Cancer Res. 2008, 14, 6302–6309.

10.

Beltran

Jendrisak

Landers

; et al. The Initial Detection and Partial Characterization of Circulating Tumor Cells in Neuroendocrine Prostate Cancer. Clin. Cancer Res. 2016, 22, 1510–1519.

11.

Nelson

N. J.

Circulating Tumor Cells: Will They Be Clinically Useful?

J. Natl. Cancer Inst. 2010, 102, 146–148.

12.

Stott

Toner

; et al. Circulating Tumor Cells: Approaches to Isolation and Characterization. J. Cell Biol. 2011, 192, 373–382.

13.

Sperger

J. M.

Strotman

L. N.

Welsh

; et al. Integrated Analysis of Multiple Biomarkers from Circulating Tumor Cells Enabled by Exclusion-Based Analyte Isolation. Clin. Cancer Res. 2017, 23, 746–756.

14.

Ross

A. A.

Cooper

B. W.

Lazarus

H. M.

; et al. Detection and Viability of Tumor Cells in Peripheral Blood Stem Cell Collections from Breast Cancer Patients Using Immunocytochemical and Clonogenic Assay Techniques. Blood 1993, 82, 2605–2610.

15.

Miyamoto

D. T.

Lee

R. J.

Stott

S. L.

; et al. Androgen Receptor Signaling in Circulating Tumor Cells as a Marker of Hormonally Responsive Prostate Cancer. Cancer Discov. 2012, 2, 995–1003.

16.

Miyamoto

D. T.

Zheng

Wittner

B. S.

; et al. RNA-Seq of Single Prostate CTCs Implicates Noncanonical Wnt Signaling in Antiandrogen Resistance. Science 2015, 349, 1351–1356.

17.

Chambers

A Micromanipulator for the Isolation of Bacteria and the Dissection of Cells. J. Bacteriol. 1923, 8, 1–5.

18.

Chambers

New Micromanipulator and Methods for the Isolation of a Single Bacterium and the Manipulation of Living Cells. J. Infect. Dis. 1922, 31, 334–343.

19.

Fraser

R. A

Simple Method of Micro-Pipette Isolation of Single Cells, Canad. Pub. Health J. 1933, 24, 145–148.

20.

Stoney

S. D.

Jr. Thompson

W. D.

Asanuma

Excitation of Pyramidal Tract Cells by Intracortical Microstimulation: Effective Extent of Stimulating Current. J. Neurophysiol. 1968, 31, 659–669.

21.

Brehm-Stecher

B. F.

Johnson

E. A.

Single-Cell Microbiology: Tools, Technologies, and Applications. Microbiol. Mol. Biol. Rev. 2004, 68, 538–559.

22.

S. C.

Bongso

Sathananthan

; et al. Micromanipulation: Its Relevance to Human In Vitro Fertilization. Fertil. Steril. 1990, 53, 203–219.

23.

Lohr

J. G.

Adalsteinsson

V. A.

Cibulskis

; et al. Whole-Exome Sequencing of Circulating Tumor Cells Provides a Window into Metastatic Prostate Cancer. Nat. Biotechnol. 2014, 32, 479–484.

24.

Campton

D. E.

Ramirez

A. B.

Nordberg

J. J.

; et al. High-Recovery Visual Identification and Single-Cell Retrieval of Circulating Tumor Cells for Genomic Analysis Using a Dual-Technology Platform Integrated with Automated Immunofluorescence Staining. BMC Cancer 2015, 15, 360.

25.

Gong

Ogunniyi

A. O.

Love

J. C.

Massively Parallel Detection of Gene Expression in Single Cells Using Subnanolitre Wells. Lab Chip 2010, 10, 2334–2337.

26.

Choi

Love

K. R.

Gong

; et al. Immuno-Hybridization Chain Reaction for Enhancing Detection of Individual Cytokine-Secreting Human Peripheral Mononuclear Cells. Anal. Chem. 2011, 83, 6890–6895.

27.

Yamanaka

Y. J.

Berger

C. T.

Sips

; et al. Single-Cell Analysis of the Dynamics and Functional Outcomes of Interactions between Human Natural Killer Cells and Target Cells. Integr. Biol. 2012, 4, 1175–1184.

28.

Ozawa

Kinoshita

Kadowaki

; et al. MAC-CCD System: A Novel Lymphocyte Microwell-Array Chip System Equipped with CCD Scanner to Generate Human Monoclonal Antibodies against Influenza Virus. Lab Chip 2009, 9, 158–163.

29.

Xia

Whitesides

G. M.

Soft Lithography. Angew. Chem. Int. Ed Engl. 1998, 37, 550–575.

30.

Qin

Xia

Whitesides

G. M.

Soft Lithography for Micro- and Nanoscale Patterning. Nat. Protoc. 2010, 5, 491–502.

31.

Rettig

J. R.

Folch

Large-Scale Single-Cell Trapping and Imaging Using Microwell Arrays. Anal. Chem. 2005, 77, 5628–5634.

32.

Lange

J. R.

Steinwachs

Kolb

; et al. Microconstriction Arrays for High-Throughput Quantitative Measurements of Cell Mechanical Properties. Biophys. J. 2015, 109, 26–34.

33.

Tharmalingam

Ghebeh

Wuerz

; et al. Pluronic Enhances the Robustness and Reduces the Cell Attachment of Mammalian Cells. Mol. Biotechnol. 2008, 39, 167–177.

34.

Luk

V. N.

G. C.

Wheeler

A. R.

Pluronic Additives: A Solution to Sticky Problems in Digital Microfluidics. Langmuir 2008, 24, 6382–6389.

35.

Warrick

J. W.

Timm

Swick

; et al. Tools for Single-Cell Kinetic Analysis of Virus-Host Interactions. PLoS One 2016, 11, e0145081.

36.

Wang

Belovich

J. M.

A Simple Apparatus for Measuring Cell Settling Velocity. Biotechnol. Prog. 2010, 26, 1361–1366.

37.

Casavant

B. P.

Guckenberger

D. J.

Berry

S. M.

; et al. The VerIFAST: An Integrated Method for Cell Isolation and Extracellular/Intracellular Staining. Lab Chip 2013, 13, 391–396.

38.

Casavant

B. P.

Strotman

L. N.

Tokar

J. J.

; et al. Paired Diagnostic and Pharmacodynamic Analysis of Rare Non-Small Cell Lung Cancer Cells Enabled by the VerIFAST Platform. Lab Chip 2014, 14, 99–105.

39.

Strotman

O’Connell

Casavant

B. P.

; et al. Selective Nucleic Acid Removal via Exclusion (SNARE): Capturing mRNA and DNA from a Single Sample. Anal. Chem. 2013, 85, 9764–9770.

40.

Pérez

Rodríguez

Í.

Rodríguez

; et al. Robot Guidance Using Machine Vision Techniques in Industrial Environments: A Comparative Review. Sensors 2016, 16, 335.

41.

Park

E. S.

Yan

J. P.

Ang

R. A.

; et al. Isolation and Genome Sequencing of Individual Circulating Tumor Cells Using Hydrogel Encapsulation and Laser Capture Microdissection. Lab Chip 2018, 18, 1736–1749.

42.

Stevens

Oomens

Broekmaat

; et al. VyCAP’s Puncher Technology for Single Cell Identification, Isolation, and Analysis. Cytometry A 2018, 93, 1255–1259.

43.

Timm

A. C.

Warrick

J. W.

Yin

Quantitative Profiling of Innate Immune Activation by Viral Infection in Single Cells. Integr. Biol. 2017, 9, 782–791.

44.

Jones

M. C.

Kobie

J. J.

Delouise

L. A.

Characterization of Cell Seeding and Specific Capture of B Cells in Microbubble Well Arrays. Biomed. Microdevices 2013, 15, 453–463.

45.

Ahrberg

C. D.

Lee

J. M.

Chung

B. G.

Poisson Statistics-Mediated Particle/Cell Counting in Microwell Arrays. Sci. Rep. 2018, 8, 2438.

46.

CRAN—Package polyaAeppli. https://cran.r-project.org/web/packages/polyaAeppli/index.html (accessed Dec 20, 2018).

47.

R: The R Project for Statistical Computing. https://www.r-project.org/ (accessed Dec 20, 2018).

48.

Ferreira

M. M.

Ramani

V. C.

Jeffrey

S. S.

Circulating Tumor Cell Technologies. Mol. Oncol. 2016, 10, 374–394.

49.

Sho

Court

C. M.

Winograd

; et al. Precision Oncology Using a Limited Number of Cells: Optimization of Whole Genome Amplification Products for Sequencing Applications. BMC Cancer 2017, 17, 457.

50.

Nel

Jehn

Gauler

; et al. Individual Profiling of Circulating Tumor Cell Composition in Patients with Non-Small Cell Lung Cancer Receiving Platinum Based Treatment. Transl. Lung Cancer Res. 2014, 3, 100–106.

51.

Nel

Gauler

T. C.

Bublitz

; et al. Circulating Tumor Cell Composition in Renal Cell Carcinoma. PLoS One 2016, 11, e0153018.

52.

Tokar

J. J.

Warrick

J. W.

Guckenberger

D. J.

; et al. Interrogating Bronchoalveolar Lavage Samples via Exclusion-Based Analyte Extraction. SLAS Technol. 2017, 22, 348–357.

53.

Junker

J. P.

van Oudenaarden

Every Cell Is Special: Genome-Wide Studies Add a New Dimension to Single-Cell Biology. Cell 2014, 157, 8–11.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

1.13 MB