Ultra‐Specific Isolation of Circulating Tumor Cells Enables Rare‐Cell RNA Profiling

The clinical potential of circulating tumor cells (CTCs) in managing cancer metastasis is significant. However, low CTC isolation purities from patient blood have hindered sensitive molecular assays of these rare cells. Described herein is the ultra‐pure isolation of CTCs from patient blood samples and how this platform has enabled highly specific molecular (mRNA and miRNA) profiling of patient CTCs.

Comparison of technologies with the reported device for CTC sorting and isolation Table S2. Parameters and devices studied during the 2nd generation devices to arrive at final spiral design Table S3. PDAC patient information and status Table S4. Data for individual patient samples either miRNA or mRNA profiled Video S1. Magnetic bead-labeled cancer cells being magnetically sorted from the waste stream into the collection outlet to achieve ultra-specific CTC sorting

Fabrication of PDMS Microfluidic Chip. Microfluidic channels were designed using
AutoCAD and master molds were fabricated using SU8-2100 negative photoresist (Microchem Corp.) following standard photolithography procedures. All molds were fabricated with a thickness of 100µm. A 10:1 ratio of PDMS polymer to curing agent (Dow-Corning) was mixed and de-bubbled prior to pouring over the SU8 molds. The PDMS was allowed to cure at 65⁰C overnight. After curing, the PDMS chips were manually cut and peeled from the mold, trimmed to size, and inlet/outlet holes were punched. PDMS chips were bonded by oxygen plasma to cleaned glass slides. Tygon tubing was inserted to connect device inlets and outlet ports to syringes. All flow was driven by Harvard Apparatus syringe pumps. Prior to running samples, tubing and chips were primed with PBS and 1% Pluronic F127 solution for several minutes to prevent cells from sticking to PDMS surfaces.
Polymer Micro Beads. Device prototypes for the spiral inertial sorter and the passive mixer were first evaluated using fluorescent polymer beads. 20µm (FITC, Polymer Sciences), 15µm (DAPI, Invitrogen), 10µm (Dragon Green, Bangs Lab), and 7µm beads (Sky Blue, Spherotech) were used for flow characterization at concentrations ranging from 10 4 -10 5 beads mL -1 . Also, magnetic fluorescent polymer beads (8µm, Bangs Lab) were used to characterize the magnetic sorter.

Inertial Sorting
Theory of Inertial Sorting 11 . The separation principle of the spiral is based on the concept that particles flowing along a rectangular, curved channel will experience a combination of inertial lift forces, F L and Dean drag forces, F D , the magnitudes and directions of which, are dependent on particle size and the relative position of the particle across the channel cross section. These two lift forces are oppositely directed and they result in a net lift force, F L that focuses different streams of particles within the microchannel. This force is given by where ρ is the fluid density, G is the fluid shear rate given by G=U max /D h , where U max is the maximum fluid velocity and D h is the hydraulic radius of the channel. C L represents the lift coefficient which is a function both of the channel's Reynolds number, Re and the particle position along the channel cross section. 10-11 Particles experience a transverse drag force from the Dean vortices, which is calculated by assuming Stokes drag such that where U Dean is the average Dean velocity given by U Dean =1.8 x 10 -4 De 1.63 where De is the Dean Number given by De=Re(D h /2R) 1/2 . R is the radius of curvature of the channel and μ is the fluid viscosity. For the spiral device studied, the estimated ratio of F D /F L for a CTC (a p ~20μm) is 2.148 while F D /F L for a WBC (a p ~10μm) is 0.468 so that the 2 particles equilibrate at different cross sectional positions. Essentially, the magnitude of the ratio of F D /F L which varies with a p 3 (a p being particle diameter), determines the equilibrium position adopted by particles of different sizes. 11 Thereupon, the spiral device can be engineered so that particles of different diameters occupy single, focused streams along the channel length. Ideally, in the present study, when a patient's blood sample is flowed through the inertial spiral separator, all of the small blood cells will be focused to the outer wall thus resulting in a suspension consisting mainly of cancer cells and some WBCs in the innermost channel outlet.
Device Design. The optimized spiral design for the integrated chip consists of a 500 µm wide channel that expands to 1000 µm at the outlet and branches into one collection outlet (innermost outlet, 200 µm wide) and three waste outlets. The largest radius on the spiral is 15 mm and the height of the microfluidic channels is 100 µm. The spiral device operates at 1700 µL min -1 as a stand-alone platform, for effective sorting of large (15-20 µm) particles from smaller particles (7-10 µm). Once the inertial sorter is fitted into the integrated device, the spiral operating flow rate is dropped to 1200 µl/min for resistance-matching and to maintain effective CTC sorting.  Table S2. Parameters and devices studied during the 2nd generation of devices to arrive at a final spiral design. Figure S1. Inertial Sorter Optimization: parameters and devices studied during the 2nd generation device optimization to arrive at final spiral design (A) Optimized device dimensions showing resistance matching channels (B) Different outlet designs tested to facilitate optimal inertial cell sorting in stand-alone spiral.

Passive Mixing in Serpentine Channel
The serpentine configuration for the passive mixer has been chosen since the turns in the channels give rise to centrifugal Dean Forces that allow particles to cross fluid stream lines and become mixed. Dean forces increase with increasing fluid flow rate to lead to rapid mixing and high mixing efficiencies. 27 Typically, for the microchannel-based passive mixer, flows through the device are primarily laminar with values of Reynolds' number, (Re), in the range of 0.01<Re<100 (Re~30 in the passive mixer studied) 13 . Because both viscous and inertial forces are appreciable in this range of Re, flow around the serpentine bend generates "secondary flows" in both the axial and radial directions thus increasing the interfacial area over which diffusion and mixing occur. The serpentine channel has a height of 100µm and a width of 250 µm and is capable of mixing particles at flow rates of 50 µL min-1 or higher. As the flow rate increased, mixing occurred sooner and was primarily dependent on chaotic advection rather than diffusion, which arises at slower flow rates where Dean forces are weak.
Preliminary evaluation of mixing was done by observing the mixing patterns of yellowand blue-dyed buffer solutions generated in the mixing module. Using flow rates of 10-100 µL min-1, the mixing quality was characterized by the distance along the mixer required for the two distinct streams of yellow and blue, to become one well-mixed green stream (see SI Figure S1).
The highly compact design of the mixer enabled fluids to be completely mixed by the third channel segment, at a length of ~4cm. Cancer cell labeling with EpCAM coated magnetic beads

was examined through static experiments (PANC-1 cells and EpCAM-labeled beads in an
Eppendorf tube) and on-chip experiments (on-chip passive mixing of PANC-1 cells and EpCAM-labeled beads subsequently flowed into reservoirs). Calculations demonstrate that for a cell to experience a magnetic force sufficient enough to pull it in the direction perpendicular to flow, and therefore allow it to be sorted, a cell of average diameter of 20 µm requires less than 1/3 bead coverage to be deflected to the desired magnetic sorter outlet (see SI Section 4).For each incubation time point (5, 10, 15 and 30 min), the quality of mixing was evaluated by immediately imaging the mixed effluent to quantify the extent to which cells were labeled with magnetic beads. Cells were characterized as having no coverage, 1/3-2/3 coverage, and greater than 2/3 coverage (see SI Figure S3). In static conditions, with no other cells present, at least 15 minutes of incubation time was required to achieve at least 1/3 surface coverage on 85% of cells (SI Figure S6). Figure S2. Mixing progression through passive mixer. Dyed PBS was used to demonstrate how increasing flow rates lead to mixing occurring closer to the beginning of the channel (a)10 µL min -1 (b)40 µL min -1 (c)70 µL min -1 (d)100 µL min -1 . Scale = 250µm.

Magnetic Bead Coverage Efficiency
Calculated Requirements. COMSOL modeling of magnets (K&J Magnetics, B224) shows a range of magnetic fields of 0.22-0.28T across the microchannel ( Figure S2). These values are sufficient to saturate the magnetization of superparamagnetic microparticles as indicated by the hysteresis curve provided by the manufacturer. Therefore velocities are proportional to the magnetic field gradient and magnetic particle movement can be measured and used for Stokes law calculations. Figure S3. Comsol calculation of magnetic field strength at channel edges. Magnets (500mT) were positioned to be 1mm from the magnetic sorter channel.
The magnetic force on a magnetic particle in the laminar flow regime can be determined by applying Stokes law: where η is the fluid viscosity, µ t is the terminal velocity perpendicular to flow (which can be approximated by tracking bead movement across the channel), and D is particle diameter.
Magnetic beads were placed introduced to a microfluidic channel and permanent magnets were placed 1mm from the channel edge. Particle movement was recorded using a high-speed camera.
This leads to a deflection velocity, , of N • 7.5µm s -1 . Next cells were tracked while traveling in laminar flow through the microchannel and linear velocity was determined to be 125mm s -1 resulting in a residence time, or magnetic exposure time, of 0.2s (magnet length, 25mm). Then each magnetic bead is capable of deflecting the cell 1.5µm toward the magnet. Based on channel dimensions and streamlines, the maximum distance a cell needs to move is 100µm, therefore a minimum of 65 beads are required to ensure cell collection. Most observations of magnetic bead coverage of PANC-1 cells were minimum of 1/3 coverage which would be roughly 400 beads based on surface area. Thus, it is reasonable to assume that sufficient magnetic bead coverage is achieved for cell collection within the magnetic sorter module. The channel length is 25mm with a height of 100µm.

Minimum Magnetic Bead Coverage of Deflected Cells.
Cells labeled with 1 μm magnetic and diluted in PBS were flowed through the magnetic sorter as previously described. The outputs of each outlet were examined via scanning microscopy to determine bead coverage required for collection. Figure S3 shows the minimum bead coverage necessary for cells to be deflected into the collection stream at a sample flow rate of 50µL min -1 . Video S1 provides further confirmation of labeled cells being pulled from the sample stream into the bottom outlet while remaining blood cells continue to upper waste outlet and are not deflected by the magnets.

Experimental Labeling Efficiency and Magnetic Cell Sorting
To determine appropriate bead dilution and incubation time to ensure sufficient labeling of PANC-1 cells (1/3 surface coverage or higher of ~85% of cell), labeling experiments were carried out in buffer and blood. To test the on-chip mixing and labeling procedure, cells and diluted beads were flowed through the passive mixer at 100µL min -1 for a total volume of 1-1.5 mL. The cells and beads were allowed to incubate for the allotted time in the reservoirs with occasional rocking to prevent cell/bead settling and to enhance bead-cell mixing. As a negative control, equivalent volumes of cell sample and beads were mixed in an Eppendorf tube. The tube was occasionally rocked to prevent cell/bead settling. After each incubation time, a drop of mixed sample was scanned on a slide at 20X and cells were counted based on their bead coverage ( Figure S4). These experiments were performed with 1:10 and 1:5 bead dilutions and cell concentration of approximately 10 5 ( Figure S5). Based on these results, 1:10 bead dilution was used and cell labeling was further tested in diluted blood to ensure bead-cell labeling efficiency was maintained even in blood. Figure S5 shows that 92% of cells have sufficient coverage after only 5 min incubation in reservoirs. were tested. For the 50 µL min -1 bias condition, contamination was 2.35% ± 2.72 and efficiency was 88.39% ± 5.79 in the collection channel. For the 75 µL min -1 bias setup, WBC contamination decreased to less than 0.5% but the recovery efficiency was also reduced ( figure   S6). Since the purity was higher using the 75 µl/min bias conditions this setup was finally implemented.

Bead/Antibody Reactivity
Assuming that magnetic sorting of CTCs may be useful for commercial applications of point-ofcare testing the stability of the 1µm streptavidin coated superparamagnetic beads were evaluated up to 100 days after the beads were functionalized with biotinylated EpCAM per manufacturer's instructions. Day 0 indicates a sample taken after 30min of incubation with the newly functionalized beads. This experiment was repeated incrementally using the same stock of beads that had been functionalized on Day 0 and kept at 4°C while not in use. For each time point, 100µl of PANC-1 cells (~10 6 /ml) were incubated for 30min at room temperature with 10µl of functionalized beads (original stock concentration). As seen in Figure S7, beads remain active through Day 100 and continue to show the same full surface coverage of PANC-1 cells. This indicates the functionalized beads remain active and capable of cell antigen interaction when stored at 4°C. Figure S8. Examination of the prolonged reactivity of antibody coated beads. PANC-1 cells were treated with anti-EpCAM coated beads for 30min and then imaged. Time points indicate the numbers of days after beads were initially conjugated to EpCAM antibody. Beads were stored at 4°C and original stock concentration for the duration. Scale bar = 25 µm

Cell viability
To ensure cells processed through the integrated magnetic device would be viable for further cellular studies, the viability of cells was assessed through the MTT assay ( Figure S7). PANC-1 cells were either seeded directly into a 96-well plate (Control), processed through the integrated device in media without magnetic beads (Test 1), or processed through the integrated device in media with the addition of magnetic beads (Test 2). Magnetic beads were exposed to UV light for 8 hours prior to use in these experiments to promote experimental sterility. After collecting the output of the integrated device, the processed cells were also plated in 96 well plates at 4000 cells/well. Every 24 hours, cells were subjected to MTT protocol and absorbance at 570nm was recorded. Control Day 1 was used as the basis to normalize the results. Media was changed on Day 3. This experiment was repeated 3 times and showed similar growth patterns. Cells that were processed through the device were viable and resulted in nearly the same growth rate as the control cells. Figure S9. MTT assay to determine cell viability. Cells were processed through the integrated device either without magnetic beads (Test 1) or with magnetic beads (Test 2). Control cells were unprocessed. Then cells were plated in 96 well plates and grown for 5 dayswith MTT assay performed on one plate per day to assess growth. Growth was normalized to control cells on Day 1. (n=5, bars indicate standard deviation)