The three main challenges of cancer treatment are metastases, recurrence, and acquired therapy resistance. These challenges have been closely linked to circulating cancer cell clusters. A detailed understanding of their genetic and morphological composition is essential. This will not only improve our knowledge of basic cancer biology but enable the successful development of much needed therapies preventing the three main challenges mentioned above. Extensive research effort is underway to isolate, capture, and analyze circulating tumor cells. However, few if any current efforts specifically target cancer cell clusters, and their much greater ability to initiate new tumors. Growing scientific consensus over the last five years has convincingly established the importance of targeting circulating cancer cell clusters verses individual CTCs to prevent the occurrence of metastatic disease. Based on the increased clinical importance of cancer cell clusters as the main driver of cancer metastasis, new and improved methods are much needed to access these larger multi-celled structures. Microfluidic devices offer a readily accessible platform for a customizable microenvironment for cell isolation and analysis. In this study, we show how a well-known passive micromixer design (staggered herringbone mixer - SHM) can be optimized to induce maximum chaotic advection within antibody-coated channels of dimensions appropriate for the capture of cancer cell clusters. The device’s principle design configuration is called: Single-Walled Staggered Herringbone (SWaSH). The preliminary empirical results of our work show that utilization of extensive simulation and modeling can accelerate the development of a working prototype that allows for target-specific cancer cell cluster isolation.

The fight against metastatic cancer is among the foremost challenges we face in our efforts to truly improve patient outcomes of many if not most forms of cancer. The identification of circulating cancer cell clusters as primary vehicles of the metastatic process1–6 have opened up a new therapeutic target space holding great promise. This is confirmed by the most recent findings by Aceto and colleagues6 clearly showing that eliminating circulating cancer cell clusters directly impacts metastatic potential in an orthotopic patient-derived-xenograft models of breast cancer.

Microfluidic technologies hold great promise in cancer diagnosis, monitoring of disease progression, and potentially, identification of optimized therapy.7 By extension, microfluidic devices can be utilized to acquire and analyze important biological samples from patients via liquid biopsies that are substantially less invasive and present a lower risk than traditional biopsies.7 Over recent years various microfluidic approaches have been proposed for the capture and isolation of circulating cancer cells.8,9 However, nearly all of these methods focus on individual circulating tumor cells (CTC) and often exhibit an intrinsic bias against larger multicellular structures.10 Circulating CTC clusters have a demonstrated higher ability to initiate metastases in comparison to individual circulating tumor cells, however they are difficult to obtain.1,3,11,12 Only few studies have reported reliable estimates on the size of these clusters, ranging from 2-50 cells.1 However, the authors commented on the intrinsic bias of the utilized approach against clusters of a larger size.1 Thus, the development of technologies to isolate and identify CTC clusters independent of their size is necessary to enable their thorough investigation. Furthermore, the extremely small number of circulating clusters requires an ability to process a larger sample volume and any useful technology have the capability to process a sample both rapidly and gently to preserve viable cell. Maintaining the viability of the isolated cancer cell clusters will allow more relevant analysis of their tumorigenic potential and enable cutting edge downstream analysis.

Handling of microliter volumes of liquids in small microchannels experiences extremely laminar flow with little to no advection. The observed Reynolds numbers (Re) in microfluidic systems often range from 0.2 – 5, which is substantially below a value of 2300 - 2900 commonly considered the transition point from laminar to turbulent flow.13,14 Laminar flow dramatically reduces the contact time of circulating cells with a potential capturing agent. Therefore, microfluidic channels need to be designed to introduce chaotic advective flow in order to generate more interactions between the rare cell clusters and the channel walls presenting a capturing moiety.15 The herringbone (HB) pattern is an arrangement of checkmark-like contours in an alternating array. This pattern’s use as a successful chaotic micromixer feature was first reported in 2002.16 Microfluidic micromixers either have these staggered HB features on the bottom or the top of each channel. In our work we expanded on the original design and substantially optimized important characteristics, such as periodicity and specific dimensions to facilitate efficient capture of cell clusters with a range of sizes in contrast to other efforts where the focus was on single cell capture only.

Here we show the production and functionalization of a microfluidic device capable of increasing the contact surface time to enable capture of cancer cell clusters and individual cancer cells. The design of our microfluidics device advances the work of Stroock and colleagues16,17 who showed that an HB pattern within microchannels that allows for increased chaotic flow at low Reynold’s numbers. This approach was also investigated by Stott et al.15,18 However, none of their designs was intended for or has shown an ability to process larger cell clusters, or target very rare cancer cell types.

Immuno-capture approaches to capture and isolate target cells have been successfully used by a number of microfluidic devices.19 To readily exploit the wealth of suitable antibodies available to target the wide diversity of cell surface receptors found on different cancer types, easy and rapid customization of the capture surfaces is important. For this we developed an alginate hydrogel coating with covalently attached streptavidin derivatives that can readily by functionalized with any biotinylated antibody. Furthermore, the device should have the ability to process whole blood samples with only a minimum amount of pre-processing, such as standard anti-coagulant treatments (i.e. citrate-, heparin-, or EDTA-buffered Vacutainer® collection) to minimize the loss of the rare target cells and clusters.20 Thus the coating has to exhibit the ability to prevent non-specific binding and biofouling commonly associated with the use of undiluted whole blood, while still allowing efficient capture of the target cell clusters. Incorporating our design enhancements and modifications into a single device, we termed this new approach Single-Walled Staggered Herringbone (SWaSH).

In order to generate a microfluidic chip meeting our requirements we employed a four-stage approach. Firstly, we used computer modeling and simulation analysis to determine ideal parameters for channel dimensions as well as HB characteristics. Secondly, we adapted photolithography to generate a suitable mold for the PDMS substrate. Thirdly, we optimized coating chemistry and functionalization properties for the immunocapture approach. Finally, we tested the resulting device prototype utilizing relevant biological fluids and cell-based models.

The underlying principle of fluid-dynamics have been established for quite some time.21,22 However, recent advances in computational resources and methodologies drastically enables readily available capabilities to conduct meaningful simulation analyses on non-standard geometries. Uniaxial flow is not desired for this device design since the objective is to generate more cell-to-antibody interactions, so that is where the HB configuration comes into play. We utilized extensive simulation analyses to identify ideal starting geometries for the initial prototyping.

1. Finite element analysis (FEA)

For this work’s FEA analysis, COMSOL Multiphysics® version 3.5 and a trial version of 5.1 (Palo Alto, CA) was used with the “Laminar Flow,” “Transport of Diluted Species,” and “Particle Trajectories” modules. All geometric objects, both 2D and 3D, were designed and exported using SolidWorks 2015 (Dassault Systèmes, Waltham, MA).

2. SWaSH configuration analysis

For the configuration analysis, channel depth was kept at 100 μm, HB depth was set at 50 μm, and the HB short arm to long arm ratio kept at a 1:3. The varying parameters were HB groove width, spacing between HB grooves, and the number of HB in the same orientation per pitch. The deterministic factor of success is increased “cell-to-surface” interactions. For this we placed 40 tracer particles uniformly distributed at the inlet cross-section. These tracer particles were defined as massless and followed a previously solved velocity profile. The quantification involved counting the number of inlet particles that hit any boundary within the domain.

1. Photolithography

100 mm bare Si wafers were cleaned and used as substrates. For the master mold, we used MicroChem® SU-8 100 (Westborough, MA), a negative photoresist (PR), and coated the substrate uniformly to the desired thickness (100 μm for channels and 50 μm for HBs). The photomask transparency was prepared using CoventorWare® 2010 and printed by CAD/Art Services, Inc. (Bandon, OR). Two masks were prepared, one for the channels and one for the herringbones with a dark field polarity. The first mask used was the channel mask that included the channel corridor from the inlet port, which then branched out into 32 parallel channels for 15 mm, and finally branched back into one outlet port. The parallel channels are each 200 μm wide with 100 μm spacing between any two channels. Both the inlet and outlet ports are 3 mm in diameter. The second mask was designed to be placed exactly over the 32 parallel channels, from the beginning to the end of the horizontal sections. Each HB is 80 μm wide with 120 μm spacing between subsequent HBs. Every HB has a short arm length of 50 μm and a long arm length of 150 μm (1:3 ratio). The positioning of each arm changes every 10 HBs, with a total of 4 pitch changes (supplementary material Fig. S1). A detailed photolithography protocol can be found in the supplementary material.

2. Elastomer casting over master mold—Soft lithography

The PDMS (Dow Corning Sylgard 184 Silicone Elastomer Kit, Auburn, MI) was mixed at a 10 to 1 weight to weight (w/w) ratio of base elastomer to curing agent. The PDMS was degassed in a vacuum desiccator for 30-60 min. In order to prevent unwanted bonding, a passivation layer needs to be introduced between photoresist (PR) features and crosslinked PDMS. Chemical vapor deposition of organosilane, tridecafluoro-1,1,2,2-tetrahydrooctyl trichlorosilane (Gelest SIT8174.0, Morrisville, PA) for passivation was introduced to prevent unwanted bonding.23 To cast the mold, degassed PDMS was slowly poured onto the mold and cured at 120°C for 20 minutes. Access to the inlet and outlet ports in the PDMS were cut using a 1.5mm biopsy punch (supplementary material Fig. S2A). Finally, the PDMS was irreversibly plasma bonded onto a glass slide using the Technics Series 85 Reactive Ion Etcher (Pleasanton, CA) at 150W for 30 seconds (supplementary material Fig. S2B).

1. Hydrogel coating of the device microchannels

The microfluidic channel walls were functionalized using an alginate hydrogel derivatized with a covalently bonded streptavidin.24 To generate a uniform alginate coating, a 2.5 M CaCl2 solution was pumped through the PDMS channels using a Harvard Apparatus 11+ syringe pump (Holliston, MA) leading to an enrichment of Ca2+ ions onto the channel walls. The Cy5 labeled streptavidin (ZyMAX™ Streptavidin-CyTM5, Invitrogen, Carlsbad, CA) was covalently attached to the un-crosslinked alginate (FMC Biopolymer AS, Sandvika, Norway) before pumping into the device.25 The Ca2+-mediated crosslinking was allowed to occur over 30 minutes at 4°C. After crosslinking, the microchannels were flushed with 50 mM MES buffer pH 7 at a high flow rate (∼1 mL/min) to remove any un-crosslinked alginate-streptavidin-Cy5. A uniform coating was observed on all inner channel surfaces, 5-15 μm thickness, which was optically verified using an inverted fluorescent microscope (Zeiss Axiovert 200M, Germany) and analyzed using ImageJ open-source software (NIH, Maryland). Antibody functionalization was performed by perfusing a biotinylated antibody in 50 mM MES buffer, pH 7 at 10 μL/min for 30 min through the chip. Biotinylated anti-human CD38 antibody (#303518, BioLegend, San Diego, CA) diluted 1:2000 in 50 mM MES buffer was pumped through the microchannels and allowed to incubate for 30 min. The chip was subsequently washed using 50mM MES buffer pH7.0 at a flow rate 200 μL/min for 30 minutes. A fluorophore-labeled secondary antibody (Alexa Fluor® 594 anti-mouse IgG1) was utilized to immune-stain the immobilized anti-CD38 antibody.

All cell lines were obtained from the American Type Culture Collection (ATCC) and sequence verified. The human prostate cancer cell line PC3, was stably transfected with green fluorescent protein (PC3-GFP) as described previously26 and used as a model cell line to evaluate flow characteristics. The human multiple myeloma RPMI-8226 B-cell line and the PC3-GFP cells were cultured in RPMI medium supplemented with 10% FBS and 1x antibiotic-antimycotic at 37°C with 5% CO2. The PC3 cells were gently trypsinized with 0.25% trypsin + 2mM EDTA in PBS for 10 minutes at 37°C, or until cells were detached and subsequently diluted to the desired concentration. Finally, the RPMI-8226 B-cells were stained with Hoechst (Thermo Fisher, Hoechst Stain 33342) and the PC3-GFP cells were either imaged at the intrinsic fluorescence (Ex. 488 nm/Em. 525 nm) or stained with SYTO85 (Ex. 567 nm/Em. 583 nm; Thermo Fisher, S11366) according to the manufacturer’s protocol.

The human breast cancer cell line MDA-MB-468 were cultured in DMEM medium supplemented with 10% FBS. Cells were stained with Hoechst 33342 in 500 μL of 1X TBS with 5% BSA 0.1% Tween-20.

PC3-GFP cells, as described above, were seeded as single cells in a 24-well ultra-low attachment plate (Corning 3473) at 500 cells per well in 500 μL media. Incubation was allowed for 72 hours and cluster formation confirmed by visual inspections. An average of 80 clusters with approx. 50 cells per cluster were found in each well. Cell surface presentation of STEAP-1,27,28 the prostate cancer specific epitope targeted for the immune-capture of the PC3-GFP clusters, was confirmed by immune fluorescence staining on the clusters (data not shown; #042391, US Biological, Salem, MA).

Use of the alginate hydrogel provides a biocompatible surface with low inherent non-specific binding properties. In order to evaluate the ability of our functionalized hydrogel coating to resist or minimize whole blood clogging we obtained normal donor blood collected in sodium heparin BD Vacutainer® blood collection tubes (San Diego Blood Bank, San Diego, CA). Whole blood was pumped at 100 μL per minute for 20 minutes through our chip within 4 hours of collection. Channels were continuously observed for clogging or restricted flow via microscopy video imaging.

The original design was adapted to reduce the width and increase the total number of microchannels. The previous designs used 8 channels, each of 2 mm width.15 Whereas, our design makes use of 32 channels, each of 200 μm width and 100 μm spacing, which will increase the available chip surface to cross-sectional area by approx. 1.4-fold (Fig. 1). Our goal was to target rare CTC clusters with a size ranging between 50-150 μm in diameter (∼2-50 cells per cluster) based on the cluster sizes identified in patients and described in the literature.1 Thus, the channel dimensions and HB design had to be modified to function for both clusters and individual cells. Further, the design should be compatible with relatively high flow rates thus allowing large volumes to be processed rapidly while increasing the ability to capture extremely rare cells and clusters.

FIG. 1.

A) Top and profile view 2D drawing of SWaSH configuration with bifurcation from single inlet. (all units in mm) B) 3D model extruded from 2D drawing of SWaSH configuration. Only one channel with one pitch change (10 HB grooves per pitch) was utilized in the FEA analysis.

FIG. 1.

A) Top and profile view 2D drawing of SWaSH configuration with bifurcation from single inlet. (all units in mm) B) 3D model extruded from 2D drawing of SWaSH configuration. Only one channel with one pitch change (10 HB grooves per pitch) was utilized in the FEA analysis.

Close modal

Using simulation analysis, we determined optimal channel dimensions and flow rates that result in the highest number of particle (i.e. rare cell cluster) to-wall (i.e. all hydrogel coated surfaces) interactions to allow the capture of single cells and larger cell clusters. FEA simulations were done to optimize the HB dimensions for the desired channel width (200 μm) and initial results showed that the HBs provide a three-dimensional flow as compared to a one-dimensional flow (i.e. uni-axial) normally found in smooth rectangular microchannels as seen in figure 2D. Although the flow is still laminar, since it continues to be dominated by viscous forces, the HBs “stretch” and “fold” the fluid in the transverse direction (i.e. z-axis) while travelling down the length of channel (Fig. 2C). This validates results found by Stroock and McGraw29 and increases cell-to-surface antibody interactions. It is important to note that mass transport of a diluted species in a bulk microfluidic flow remains strictly diffusion-limited. However, in the z-dimension the channel is expanding and contracting, while also changing in the x- and y-dimensions due to the staggered HB design and the plasticity of the PDMS. This causes a steady axial pressure gradient, due to the anisotropic hydraulic resistance to the viscous flown, and allows micro-vortices to occur. This improved mixing can be defined by the dimensionless Péclet number.30 This number is the ratio of the advective transport rate and diffusive transport rate, where a high Péclet number can be defined as “chaotic advection.”

FIG. 2.

A) Sliced representation of localized Reynolds number (colored scale bar). Although the numbers are too low to induce traditional turbulent flow, the heterogeneous distribution of localized maxima and minima aids chaotic advection. B) Shows the modeled interaction pattern of the top row of 40 massless particles released at the face of the inlet in a 5 by 8 grid in a liquid medium model. Note, 5 out of 8 particles experience at least one interaction in the first two pitches of the HB pattern. C) FEA results of HB’s effect on stretching and folding the fluid in the transverse direction (i.e. z-axis). Colored tracings represent simulated particle trajectories and their velocities over a 4-mm longitudinal section. D) FEA simulation of particle behavior within the channel on a cross-sectional view. The differential tracing between time=0 (upper panel) and time=200s (lower panel) shows the chaotic movement of particles in flow. The color scale represents particle velocities in mm/s.

FIG. 2.

A) Sliced representation of localized Reynolds number (colored scale bar). Although the numbers are too low to induce traditional turbulent flow, the heterogeneous distribution of localized maxima and minima aids chaotic advection. B) Shows the modeled interaction pattern of the top row of 40 massless particles released at the face of the inlet in a 5 by 8 grid in a liquid medium model. Note, 5 out of 8 particles experience at least one interaction in the first two pitches of the HB pattern. C) FEA results of HB’s effect on stretching and folding the fluid in the transverse direction (i.e. z-axis). Colored tracings represent simulated particle trajectories and their velocities over a 4-mm longitudinal section. D) FEA simulation of particle behavior within the channel on a cross-sectional view. The differential tracing between time=0 (upper panel) and time=200s (lower panel) shows the chaotic movement of particles in flow. The color scale represents particle velocities in mm/s.

Close modal

Numerous simulations were performed by varying different properties of the HB pattern such as channel configuration, and flow velocities to optimize for our deterministic factor cell-to-surface interactions (Fig. 3). The simulation results showed that the optimum HB groove width was 80 μm (Fig. 3A), channel width of 200 μm (Fig. 3B), and 10 HBs per pitch (data not shown). Based on the FEA studies, the final dimensions of the device were: channel width: 200 μm, channel height overall: 150 μm, HB extension height: 50 μm, HB groove width: 80 μm, HB spacing: 200 μm. Length of the channels was set to 30 mm providing a total capture zone volume of 12.26 μl.

FIG. 3.

Results from the FEA simulation: A) variable HB groove width showing that at 60 to 80 μm provides a sustained local maximum of the percentage of particles interacting with the boundaries over the length of the channel. B) The variable channel widths tested suggest that between 200-240 μm are the most favorable conditions for particle-boundary interactions, C) Inlet velocities between 3 to 5 mm/s exhibit the maximum particle-boundary interactions. Although particle-boundary interactions appear to be rising at velocities at and above 8 mm/s, the shear forces at these velocities would preclude efficient cell-capturing via immunocapture approaches. Reynolds number are consistently and substantially below the level needed for turbulent flow and show the expected linear correlation with flow velocity, further indicating that particle boundary-interactions are primarily driven by chaotic advection.

FIG. 3.

Results from the FEA simulation: A) variable HB groove width showing that at 60 to 80 μm provides a sustained local maximum of the percentage of particles interacting with the boundaries over the length of the channel. B) The variable channel widths tested suggest that between 200-240 μm are the most favorable conditions for particle-boundary interactions, C) Inlet velocities between 3 to 5 mm/s exhibit the maximum particle-boundary interactions. Although particle-boundary interactions appear to be rising at velocities at and above 8 mm/s, the shear forces at these velocities would preclude efficient cell-capturing via immunocapture approaches. Reynolds number are consistently and substantially below the level needed for turbulent flow and show the expected linear correlation with flow velocity, further indicating that particle boundary-interactions are primarily driven by chaotic advection.

Close modal

After settling on the channel and herringbone final dimensions, the next set of simulations were done to see the effect of particle position within the microchannel on its probability to interact with a boundary. At the channel inlet, 18 particles were homogeneously (x-axis) released at 6 different heights (z-axis; supplementary material Fig. S3). The results show that particles entering the channel closer to the HB “ceiling” experience more of the transverse effects and thus more likely to meet a wall. In fact, the particles released in the center of the channel are the least likely to interact with the wall and continue down the length of the channel. However, as shown in Figures 2C and 2D the HB pattern induced transversal folding of the flow appears to be extensive enough to allow particles to experience reach regions of high chaotic advection and thus high probabilities of wall interactions. Nevertheless, it is important to keep in mind that these results are used chiefly as a directional tool for determining the best design for fabrication and implementation. Fabrication of the prototypes and careful analysis of flow behavior of cells and clusters within the devices helped to validate the theoretical scaffold provided here.

Following the procedure described above and in more detail in the supplementary material, we generated the SU-8 mold for the first prototype SWaSH chip (Fig. 4A–C). With the photo-patterned mold made, a two-part elastomer was poured over the mold’s features to create the foundation of the final experimental microfluidic device. The specific elastomer being used is polydimethylsiloxane (PDMS), a common type of silicone polymer. PDMS is the most predominately used microfluidic device material used today due to its low-cost, biocompatibility, optical clarity, and ease of rapid fabrication.31 

FIG. 4.

A) Microscopy image of the first PR layer showing the plain microchannels and the inlet branching. B) Microscopy image of the first and second layer of the cured PR, illustrating the microchannels with the HB extensions on top (interior channel ceiling). C) Fully cured and cleaned PR mold ready for elastomer casting with PDMS. D) Cast PDMS chip mounted and plasma bonded to a microscope slide. Inlet and outlet ports were cute with a 1.5 mm biopsy punch prior to plasma bonding.

FIG. 4.

A) Microscopy image of the first PR layer showing the plain microchannels and the inlet branching. B) Microscopy image of the first and second layer of the cured PR, illustrating the microchannels with the HB extensions on top (interior channel ceiling). C) Fully cured and cleaned PR mold ready for elastomer casting with PDMS. D) Cast PDMS chip mounted and plasma bonded to a microscope slide. Inlet and outlet ports were cute with a 1.5 mm biopsy punch prior to plasma bonding.

Close modal

We developed a biocompatible coating of the channel walls that allows covalent binding of streptavidin molecules, which in turn allow flexible and easy immobilization of any biotinylated antibody. The functionalized, biocompatible coating facilitates target cell capture and reduces non-specific binding of non-targeted cells. The Cy5-labeled streptavidin was utilized to visualize the crosslinked and functionalized alginate hydrogel coating within the microchannels. Images were captured using a fluorescence Zeiss 200M microscope. The alginate coating appeared thin and uniform throughout all channels shown by the streptavidin-Cy5 fluorescence signal along all interior channel surfaces including the side walls and around the edges of the HBs (Fig. 5).

FIG. 5.

Cy5-labeled alginate hydrogel coated chip shows a consistent and uniform coating across the channels. A) 2D-fluorescence microscopy image at 5x near the inlet of the capture zone. B) Cross-sectional view (yellow, magenta and cyan lines denote the shown slices) through the confocal laser scanning microscopy reconstruction of two channels show a uniform coating along all walls. C) The full 3D-reconstruction confirms a consistent and homogeneous coating along the walls of a channel section (A full movie showing this is provided in the supplementary material).

FIG. 5.

Cy5-labeled alginate hydrogel coated chip shows a consistent and uniform coating across the channels. A) 2D-fluorescence microscopy image at 5x near the inlet of the capture zone. B) Cross-sectional view (yellow, magenta and cyan lines denote the shown slices) through the confocal laser scanning microscopy reconstruction of two channels show a uniform coating along all walls. C) The full 3D-reconstruction confirms a consistent and homogeneous coating along the walls of a channel section (A full movie showing this is provided in the supplementary material).

Close modal

Full characterization of the coating was performed on an Olympus FV3000 confocal laser scanning microscope (Fig. 5B). 3D reconstruction of the confocal stack allowed full visualization of the coating on all inner channel surfaces which showed clean and clear channels (Fig. 5C; supplementary material movie SM1).

Subsequently, the alginate-streptavidin coated SWaSH chip was functionalized with a biotinylated antibody, as described above. Immunofluorescence validation with a secondary antibody against the primary CD38-reactive antibody showed an even and uniform distribution on all surfaces (Fig. 6A).

FIG. 6.

A) A representative section of a coated and functionalized channel shows a uniform distribution of the secondary antibody anti-mouse IgG1Alexa Fluor® 594 (red) binding to the anti-humanCD38 antibodies. B) Only a very few PC3-GFP (CD38-negative) labeled cells (green) show non-specific binding to the channels of a chip that has been coated and functionalized with alginate and anti-CD38 antibody. C) RPMI-8226 Multiple Myeloma B-cells (CD38+) that have been labeled with Hoechst 33342 stain (blue) are selectively captured in the coated and functionalized channels. D) A captured RPMI-8226 Multiple Myeloma B-cell cluster (CD38-positive) remains intact and immobilized in the channel following an 8-minute flush with 20mM TRIS buffer pH 7.4 at 200μL/min., which equates to an approximate flow velocity of 5 mm/s.

FIG. 6.

A) A representative section of a coated and functionalized channel shows a uniform distribution of the secondary antibody anti-mouse IgG1Alexa Fluor® 594 (red) binding to the anti-humanCD38 antibodies. B) Only a very few PC3-GFP (CD38-negative) labeled cells (green) show non-specific binding to the channels of a chip that has been coated and functionalized with alginate and anti-CD38 antibody. C) RPMI-8226 Multiple Myeloma B-cells (CD38+) that have been labeled with Hoechst 33342 stain (blue) are selectively captured in the coated and functionalized channels. D) A captured RPMI-8226 Multiple Myeloma B-cell cluster (CD38-positive) remains intact and immobilized in the channel following an 8-minute flush with 20mM TRIS buffer pH 7.4 at 200μL/min., which equates to an approximate flow velocity of 5 mm/s.

Close modal

After coating and functionalization of the SWaSH device, we tested for non-specific cell binding. The GFP-expressing prostate cancer cells (PC3), with no CD38 surface antigen expression, were suspended in MES buffer at a concentration of 1,000,000 cells/mL, flowed through the device and visualized under a fluorescence microscope. This experiment showed that there was little to no non-specific binding of the PC3-GFP positive cells to the alginate hydrogel (Fig. 6B; supplementary material movie SM2).

To demonstrate that the device and its functionalized hydrogel coating aids in selective targeting of cells and clusters we performed the two following experiments.

Firstly, we determined the ability of the chip to capture targeted single cells, while minimizing non-specific cell-binding. For this we mixed target cells (RPMI-8226 Multiple Myeloma B-cells, CD38-positive) with a GFP-positive, CD38-negative control cancer cell line (PC3-GFP) at a 1:20 ratio. Before mixing the two cell lines, they were incubated with biotinylated anti-human CD38 for 2 hours and then washed several times to rinse off any unbound antibody. Approximately 50,000 RPMI-8226 cells and 1,000,000 PC3-GFP cells were suspended in 1mL of 20mM Tris buffer pH 7.4. The cell mixture was pumped through the device for 2.5 minutes at a flow rate of 200 μL/min, followed by a 30-minute incubation step at room temperature. After incubation, a 20mM TRIS buffer flush was run for 8 minutes at 200 μl/min, eliminating any non-specifically bound cells. Finally, captured and isolated cells and cell clusters were visualized under DAPI (Ex. 360 nm/Em. 497 nM) and GFP/FITC (Ex. 488 nm/Em. 525 nm) filters using a fluorescent microscope (Fig. 6C and 6D). On average 3439 ±1553 RPMI cells were captured in all available channels of a capture zone out of the 25,000 cells that were contained in the 500 μL that was passed through the device. In contrast to this, an average of only 487 ±135 PC3-GFP cells were found on the capture zone out of the 500,000 cells flowed through. Although the effective capture rate is only at 13.7%, the high selectivity against the PC3-GFP with a non-specific binding of less than 0.1% is very promising (Table I). Although the RPMI-8226 clusters we succeeded in capturing may not be directly comparable to the metastatically potent CTC clusters described above, their physical appearance, size, and cell number matches the described characteristics of reported CTC clusters.1,32 Furthermore, the RPMI-8226 cell clusters described in this study are likely to exhibit lower structurally integrity than the CTC clusters found in patients’ blood. This indicates that exposure to the shear forces the clusters experiences at 200 μl/min is not sufficient to dissociate them from the capturing antibody or lead to a breakdown of cluster integrity. Interestingly, the isolated cluster was able to withstand extensive washing of the channels with 20 mM TRIS buffer at flow velocities exceeding 5 mm/s.

TABLE I.

Analysis of capture efficiency in 5 capture zones. The number of RPMI and PC3-GFP cells was counted for each zone and capture efficiency back-calculated to the number of cells passed through the device in 500 μL.

Capture% from total% from total
Zone# RPMIs(25,000)# PC3-GFP(500,000)
1959 7.84 392 0.078 
4005 16.02 348 0.070 
4092 16.37 479 0.096 
5397 21.59 522 0.104 
1741 6.96 696 0.139 
Average 3439 ± 1553 13.8 ± 6.21 487 ± 135 0.097 ± 0.027 
Capture% from total% from total
Zone# RPMIs(25,000)# PC3-GFP(500,000)
1959 7.84 392 0.078 
4005 16.02 348 0.070 
4092 16.37 479 0.096 
5397 21.59 522 0.104 
1741 6.96 696 0.139 
Average 3439 ± 1553 13.8 ± 6.21 487 ± 135 0.097 ± 0.027 

Secondly, we wanted to determine the ability to specifically capture tumor cell clusters. For this we mixed 100,000 MDA-MB-468 breast cancer cells (see section II D) with approx. 1000 PC3-GFP clusters cultured as described in section II E, in 2 mL of MES buffer. 0.65 mL of the resulting cell suspension (approx. 325 clusters and 33,000 MDA-MB-468 single cells) were passed through the chip at 25 μL/min. The device was subsequently flushed with 1.5 mL of MES buffer at 25 μL/min to remove unbound cells. The entire chip was imaged by fluorescence microscopy and the number of captured PC3-GFP clusters as well as Hoechst 33342 stained MDA-MB-486 single cells were quantified using manual counting and ImageJ particle analysis. 232 captured PC3-GFP clusters (71.4%) were identified and 370 MDA-MB-468 cells (1.1%). Interestingly, tumor cell cluster capture appears to occur preferentially at the trailing edge of the HB chevrons, indicating that the chaotic advection induced by the pattern leads to effective cluster-boundary interaction (Fig. 7; supplementary material Fig. S4).

FIG. 7.

Fluorescence microscopy overlay of a section of the SWaSH device functionalized with anti-STEAP1 to capture PC3-GFP tumor cell clusters (green). The alginate hydrogel coating is shown in red (Cy5-labeled alginate), and the Hoechst 33342 stained MDA-MB-468 cells in blue. A composite image of the entire capture zone can be found in the supplementary material (Fig. S4). Interestingly, from the MDA-MB-468 cells that were counted on the device most were found attached to a cluster (>80%).

FIG. 7.

Fluorescence microscopy overlay of a section of the SWaSH device functionalized with anti-STEAP1 to capture PC3-GFP tumor cell clusters (green). The alginate hydrogel coating is shown in red (Cy5-labeled alginate), and the Hoechst 33342 stained MDA-MB-468 cells in blue. A composite image of the entire capture zone can be found in the supplementary material (Fig. S4). Interestingly, from the MDA-MB-468 cells that were counted on the device most were found attached to a cluster (>80%).

Close modal

Treatment of our chip with whole blood lead to no observable clogging of any channel for 20 minutes at the relatively high flow-rates tested (50 μL/min). No apparent RBC lysis was observed, and analysis of channels post blood flow indicated little binding of any biological matter even without any subsequent washing step (Fig. 8). There are approximately 4-11 million PBMC per mL of whole blood, and the number of residual cells that can be seen after the whole blood was passed through the device are very low, and we expect will readily be washed off. This strongly supports the extreme suitability of our design to be utilized with undiluted whole blood samples.

FIG. 8.

A) Image of channel section from a alginate hydrogel-coated chip perfused with undiluted whole blood from a normal blood donor. B) A similar section of the same chip after the blood perfusion illustrating that only a few residual cells adhere to the channel surface. No buffer washing was performed, which would have further eliminated any non-specifically bound biological matter.

FIG. 8.

A) Image of channel section from a alginate hydrogel-coated chip perfused with undiluted whole blood from a normal blood donor. B) A similar section of the same chip after the blood perfusion illustrating that only a few residual cells adhere to the channel surface. No buffer washing was performed, which would have further eliminated any non-specifically bound biological matter.

Close modal

Isolation, capture, and characterization of CTCs and CTC clusters from patients’ whole blood is a tremendously important milestone in the comprehensive understanding of their role in the pathology of most cancers. Analysis of these extremely rare cancer cells is an essential step in the development of new targeted therapies for these malignancies. A number of methods for isolating circulating cancer cells have been reported, however little to no effort has been directed towards capturing the native cell clusters that have been identified as the main drivers of metastasis, recurrence, and acquired resistance to therapy. Most reported methods require extensive pre-treatment of the patient whole blood sample, resulting in substantial loss of the already rare target cells. Furthermore, previously described approaches can only deal with small sample volumes (1.0 – 2.0 mL) and low flow rates (<20μL/min) which require relatively lengthy processing times, impacting the biological ‘naivety’ of the isolated cells.

We believe that our work provides an important advancement in the development of a technology that, ultimately, can allow the capture of these rare clusters from patient whole blood. We demonstrate that our microfluidic device with its optimized design pattern, can induce sufficient chaotic flow to allow the immune-capture of clusters and individual cancer cells simultaneously. We have shown that our chip design combined with the targeted immune-capture approach can lead to efficient specific binding of the target cells and clusters, with very little non-specific binding of unwanted cells. Furthermore, our device can process sample volumes in excess of 5 mL due to higher flow rates achievable, which is key for isolating rare cancer cells from patient blood samples in a timely and efficient manner.

Our future work will focus on utilizing the optimized chip design to isolate and capture circulating CTCs and, more importantly, their clusters directly from patient whole blood. A careful selection of the appropriate antibodies as well as identification of a suitable cancer type and disease stage will be necessary to evaluate the devices performance on whole blood samples derived from patients.

We provide a number of supplementary material as follows. Four additional figures are provided. 1) Illustrates the photomask utilized in the SU-8 based photolithography. 2) Are photography of the cast and mounted PDMS device. 3) Additional data of the FEA analysis highlighting the benefit SWaSH configuration in contrast to a rectangular channel of the same dimensions. 4) The full fluorescence microscope image (DAPI, GFP, and Cy5 channels) of the SWaSH device utilized for the PC3-GFP tumor cell cluster capture. Furthermore, we provide a link to two movies. 1) Showing the 3D reconstruction of a laser scanning confocal fluorescence imaging of the a channel section coated with the fluorophore labeled alginate hydrogel. 2) A movie of GFP-labeled PC3 prostate cancer cells flowing through the SWaSH device. The long exposure of the latter recording visualizes the chaotic movements the cell experience when passing through the channels. Finally, a detailed photolithography protocol as utilized in the production of the SWaSH device.

We gratefully acknowledge the following people for their support and guidance during this project. Nicole Bata for culturing the PC3-GFP clusters. Dr. Marcus Kaul, for generously providing access to the Zeiss Axiovert 200M. Dr. Victory Thaney, for teaching us Slidebook and answering incessant questions about fluorescence microscopy. Leslie Boyd, for help with the Olympus FV3000 laser scanning microscope. All members of the Cosford lab at SBP and Kassegne lab at SDSU, past and present for their helpful discussions. Part of this work was supported by a STRIVE award from the Sanford Burnham Prebys Medical Discovery Institute.

The authors JKA and PT have filed a patent application on the SWaSH design and uses thereof for the capture of circulating tumor cells and tumor cell clusters.

1.
N.
Aceto
,
A.
Bardia
,
D. T.
Miyamoto
,
M. C.
Donaldson
,
B. S.
Wittner
,
J. A.
Spencer
,
M.
Yu
,
A.
Pely
,
A.
Engstrom
,
H.
Zhu
,
B. W.
Brannigan
,
R.
Kapur
,
S. L.
Stott
,
T.
Shioda
,
S.
Ramaswamy
,
D. T.
Ting
,
C. P.
Lin
,
M.
Toner
,
D. A.
Haber
, and
S.
Maheswaran
,
Cell
158
(
5
),
1110
1122
(
2014
).
2.
A.
Fabisiewicz
and
E.
Grzybowska
,
Med Oncol
34
(
1
),
12
(
2017
).
3.
Y.
Suo
,
C.
Xie
,
X.
Zhu
,
Z.
Fan
,
Z.
Yang
,
H.
He
, and
X.
Wei
,
Cytometry A
91
(
3
),
250
253
(
2017
).
4.
M.
Giuliano
,
A.
Shaikh
,
H. C.
Lo
,
G.
Arpino
,
S.
De Placido
,
X. H.
Zhang
,
M.
Cristofanilli
,
R.
Schiff
, and
M. V.
Trivedi
,
Cancer Res
78
(
4
),
845
852
(
2018
).
5.
N.
Riggi
,
M.
Aguet
, and
I.
Stamenkovic
,
Annu Rev Pathol
13
,
117
140
(
2018
).
6.
S.
Gkountela
,
F.
Castro-Giner
,
B. M.
Szczerba
,
M.
Vetter
,
J.
Landin
,
R.
Scherrer
,
I.
Krol
,
M. C.
Scheidmann
,
C.
Beisel
,
C. U.
Stirnimann
,
C.
Kurzeder
,
V.
Heinzelmann-Schwarz
,
C.
Rochlitz
,
W. P.
Weber
, and
N.
Aceto
,
Cell
176
(
1-2
),
98
112.e114
(
2019
).
7.
A.
Kulasinghe
,
H.
Wu
,
C.
Punyadeera
, and
M. E.
Warkiani
,
Micromachines (Basel)
9
(
8
) (
2018
).
8.
M. M.
Ferreira
,
V. C.
Ramani
, and
S. S.
Jeffrey
,
Mol Oncol
10
(
3
),
374
394
(
2016
).
9.
T. A.
Burinaru
,
M.
Avram
,
A.
Avram
,
C.
Marculescu
,
B.
Tincu
,
V.
Tucureanu
,
A.
Matei
, and
M.
Militaru
,
ACS Comb Sci
20
(
3
),
107
126
(
2018
).
10.
A. F.
Sarioglu
,
N.
Aceto
,
N.
Kojic
,
M. C.
Donaldson
,
M.
Zeinali
,
B.
Hamza
,
A.
Engstrom
,
H.
Zhu
,
T. K.
Sundaresan
,
D. T.
Miyamoto
,
X.
Luo
,
A.
Bardia
,
B. S.
Wittner
,
S.
Ramaswamy
,
T.
Shioda
,
D. T.
Ting
,
S. L.
Stott
,
R.
Kapur
,
S.
Maheswaran
,
D. A.
Haber
, and
M.
Toner
,
Nat Methods
12
(
7
),
685
691
(
2015
).
11.
K. J.
Cheung
,
V.
Padmanaban
,
V.
Silvestri
,
K.
Schipper
,
J. D.
Cohen
,
A. N.
Fairchild
,
M. A.
Gorin
,
J. E.
Verdone
,
K. J.
Pienta
,
J. S.
Bader
, and
A. J.
Ewald
,
Proc Natl Acad Sci U S A
113
(
7
),
E854
E863
(
2016
).
12.
M.
Umer
,
R.
Vaidyanathan
,
N. T.
Nguyen
, and
M. J. A.
Shiddiky
,
Biotechnol Adv
36
(
4
),
1367
1389
(
2018
).
13.
H.
Schlichting
,
K.
Gersten
, and
K.
Mayes
(
Springer
,
Berlin
,
2016
), pp. 1 online resource (xxviii, 805 pages).
14.
A. E.
Kamholz
and
P.
Yager
,
Biophys J
80
(
1
),
155
160
(
2001
).
15.
S. L.
Stott
,
C. H.
Hsu
,
D. I.
Tsukrov
,
M.
Yu
,
D. T.
Miyamoto
,
B. A.
Waltman
,
S. M.
Rothenberg
,
A. M.
Shah
,
M. E.
Smas
,
G. K.
Korir
,
F. P.
Floyd
, Jr.
,
A. J.
Gilman
,
J. B.
Lord
,
D.
Winokur
,
S.
Springer
,
D.
Irimia
,
S.
Nagrath
,
L. V.
Sequist
,
R. J.
Lee
,
K. J.
Isselbacher
,
S.
Maheswaran
,
D. A.
Haber
, and
M.
Toner
,
Proc Natl Acad Sci U S A
107
(
43
),
18392
18397
(
2010
).
16.
A. D.
Stroock
,
S. K.
Dertinger
,
A.
Ajdari
,
I.
Mezic
,
H. A.
Stone
, and
G. M.
Whitesides
,
Science
295
(
5555
),
647
651
(
2002
).
17.
A. D.
Stroock
,
S. K.
Dertinger
,
G. M.
Whitesides
, and
A.
Ajdari
,
Anal Chem
74
(
20
),
5306
5312
(
2002
).
18.
S. L.
Stott
,
R. J.
Lee
,
S.
Nagrath
,
M.
Yu
,
D. T.
Miyamoto
,
L.
Ulkus
,
E. J.
Inserra
,
M.
Ulman
,
S.
Springer
,
Z.
Nakamura
,
A. L.
Moore
,
D. I.
Tsukrov
,
M. E.
Kempner
,
D. M.
Dahl
,
C. L.
Wu
,
A. J.
Iafrate
,
M. R.
Smith
,
R. G.
Tompkins
,
L. V.
Sequist
,
M.
Toner
,
D. A.
Haber
, and
S.
Maheswaran
,
Sci Transl Med
2
(
25
),
25ra23
(
2010
).
19.
M.
Antfolk
and
T.
Laurell
,
Anal Chim Acta
965
,
9
35
(
2017
).
20.
Y.
Chen
,
P.
Li
,
P. H.
Huang
,
Y.
Xie
,
J. D.
Mai
,
L.
Wang
,
N. T.
Nguyen
, and
T. J.
Huang
,
Lab Chip
14
(
4
),
626
645
(
2014
).
21.
Y. A.
Çengel
and
J. M.
Cimbala
,
Fluid mechanics: Fundamentals and applications
, Third edition, ed. (
McGraw Hill
,
New York
,
2014
).
22.
F. M.
White
,
Fluid mechanics
, Eighth edition, ed. (
McGraw-Hill Education
,
New York, NY
,
2015
).
23.
L. H.
Ting
,
S.
Feghhi
,
S. J.
Han
,
M. L.
Rodriguez
, and
N. J.
Sniadecki
,
J. Nanotechnol. Eng. Med
2
(
4
),
041006
(
2011
).
24.
N.
Nakajima
and
Y.
Ikada
,
Bioconjug Chem
6
(
1
),
123
130
(
1995
).
25.
A.
Hatch
,
G.
Hansmann
, and
S. K.
Murthy
,
Langmuir
27
(
7
),
4257
4264
(
2011
).
26.
M.
Degtyarev
,
A.
De Maziere
,
C.
Orr
,
J.
Lin
,
B. B.
Lee
,
J. Y.
Tien
,
W. W.
Prior
,
S.
van Dijk
,
H.
Wu
,
D. C.
Gray
,
D. P.
Davis
,
H. M.
Stern
,
L. J.
Murray
,
K. P.
Hoeflich
,
J.
Klumperman
,
L. S.
Friedman
, and
K.
Lin
,
J Cell Biol
183
(
1
),
101
116
(
2008
).
27.
I. M.
Gomes
,
P.
Arinto
,
C.
Lopes
,
C. R.
Santos
, and
C. J.
Maia
,
Urol Oncol
32
(
1
),
53.e23
5
3.e2
9
(
2014
).
28.
I. M.
Gomes
,
C. R.
Santos
, and
C. J.
Maia
,
Genes Cancer
5
(
3-4
),
142
151
(
2014
).
29.
A. D.
Stroock
and
G. J.
McGraw
,
Philos Trans A Math Phys Eng Sci
362
(
1818
),
971
986
(
2004
).
30.
H. A.
Stone
,
A. D.
Stroock
, and
A.
Ajdari
,
Annu Rev Fluid Mech
36
,
381
411
(
2004
).
31.
T.
Fujii
,
Microelectron Eng
61-62
,
907
914
(
2002
).
32.
D.
Marrinucci
,
K.
Bethel
,
A.
Kolatkar
,
M. S.
Luttgen
,
M.
Malchiodi
,
F.
Baehring
,
K.
Voigt
,
D.
Lazar
,
J.
Nieva
,
L.
Bazhenova
,
A. H.
Ko
,
W. M.
Korn
,
E.
Schram
,
M.
Coward
,
X.
Yang
,
T.
Metzner
,
R.
Lamy
,
M.
Honnatti
,
C.
Yoshioka
,
J.
Kunken
,
Y.
Petrova
,
D.
Sok
,
D.
Nelson
, and
P.
Kuhn
,
Phys Biol
9
(
1
),
016003
(
2012
).

Supplementary Material