Microfluidic Device for On-Chip Immunophenotyping and Cytogenetic Analysis of Rare Biological Cells

The role of circulating plasma cells (CPCs) and circulating leukemic cells (CLCs) as biomarkers for several blood cancers, such as multiple myeloma and leukemia, respectively, have recently been reported. These markers can be attractive due to the minimally invasive nature of their acquisition through a blood draw (i.e., liquid biopsy), negating the need for painful bone marrow biopsies. CPCs or CLCs can be used for cellular/molecular analyses as well, such as immunophenotyping or fluorescence in situ hybridization (FISH). FISH, which is typically carried out on slides involving complex workflows, becomes problematic when operating on CLCs or CPCs due to their relatively modest numbers. Here, we present a microfluidic device for characterizing CPCs and CLCs using immunofluorescence or FISH that have been enriched from peripheral blood using a different microfluidic device. The microfluidic possessed an array of cross-channels (2–4 µm in depth and width) that interconnected a series of input and output fluidic channels. Placing a cover plate over the device formed microtraps, the size of which was defined by the width and depth of the cross-channels. This microfluidic chip allowed for automation of immunofluorescence and FISH, requiring the use of small volumes of reagents, such as antibodies and probes, as compared to slide-based immunophenotyping and FISH. In addition, the device could secure FISH results in <4 h compared to 2–3 days for conventional FISH.

Fabrication of microtrap device. The single bed microtrap device consisted of 2 levels; one level with an interleaving channel network and the second level with cross channels that formed the microtraps to retain biological cells for FISH or immunophenotyping. The interleaving channels were ~30 µm deep and the second level contained the cross-channels that were either 4 or 2 µm deep. SU-8 lithography was performed using two optical masks with the appropriate alignment marks to allow for two-step photolithography. SU-8 structures were formed on a Si wafer that contained metal-patterned alignment marks. The microtraps, which were formed by the cross channels when a cover plate was bonded to the substrate, consisted of a 4 or 2 µm thick layer of SU-8 2002 spin coated at 3000 rpm for 30 s and soft baked at 100°C for 3 min. The wafer was then exposed to 160 MJ/cm 2 of UV light for 28 s through a dark field glass mask using a Quintel exposure station with the mask alignment marks aligned to the optical mask. Post exposure baking was done at 100°C for 4 min followed by SU-8 development. For the interleaving channels, which consisted of input/output channels that were interconnected using the cross-channels, SU-8 2015 was spin coated at 1300 rpm for 30 s to yield a 30 µm thick resist layer on top of the exposed cross-channel layer. Soft baking was done at 100°C for 10 min. Fiducial (alignment) marks were used for proper alignment of the second mask with respect to the exposed wafer.
UV exposure was performed at 300 mJ/cm 2 for 53 s and post baked at 100°C for 10 min. The wafer was developed first in SU-8 developer for 5 min followed by fresh SU-8 developer solution for 2 min and rinsed with IPA (isopropyl alcohol) for 30 s. After drying, the wafer was hard baked at 150°C for 20 min and 70°C for 10 min to anneal surface cracks in the resist, which were evident after development. The developed wafer was then cleaned using a plasma asher at 600 W for 2 min and used as a relief for subsequent PDMS casting.
The SU-8 relief consisted of different designs with different pore widths (4 μm, 6 μm or 8 μm; not shown here). Figure S1A and B shows an example of a bright-field image of the microtrap device transferred into PDMS where line scans are shown in Figures S1C and S1D for 4 m and 2 m microtrap devices (depth of traps), respectively. A 3-D scan of the device is shown in Figure S1E and F, where the blue area represents the ~30 m deep interleaving channels and the microtraps are shown in the light red area. Measurements of the depth of the microtraps were analyzed using the Keyence VK-X200 series laser scanning confocal microscope. According to the measurements, 4 m deep microtraps showed an average of 3.5  0.1 m from depth analysis of the PDMS device while 2 m deep microtraps showed a depth of 1.5  0.1 m, where both gave ~0.5 m depth deviations from the designed value. Figure S2 shows the fabrication steps for the 8-bed microtrap device. Similar to the single bed device, it was fabricated on a Si wafer using SU-8, but in this case it required 3 lithography steps. As with the single bed device, the Si wafer possessed alignment marks for the 3 lithography steps. Lithography Step 1 was performed to fabricate the microtrap layer by spin coating SU-8 with a controlled thickness of 2 or 4 m. Photoresist was exposed through a mask to define the microdtrap positions and following SU-8 development, the cross-channels were generated. Next, SU-8 photoresist was spin coated onto the Si wafer followed by UV exposure through another mask and resist development to fabricate the interleaving channels (lithography Step 2). The SU-8 layer thickness for the interleaving channels were ~30 m in depth. The final lithography step (Step 3) utilized an SU-8 layer (~150 m) processed as described for Steps 1 and 2, but with a different mask to fabricate the distribution channels that possessed 150 m depths. The distribution channels were used to allow for one input port that supplied sample to all 8 beds with a common output port. The fluid flow into the distribution channels was perpendicular to the fluid flow in the interleaving input channel network to assure uniform fluid supply to all 8 beds.
Flow simulation of the microtrap device. COMSOL simulations were performed to determine the proper fluidic operational parameters for containing biological cells using the microtrap device without causing damage to the cells so that they could be subjected to immunophenotyping or FISH. For the simulation, we chose the device design with microtraps of 4 m width and 2 m depth as they provided the highest containment efficiency (see Figure S3A). Figure S3B shows the linear velocity profile through a small section of the device for a 10 µL/min input volumetric flow rate (2x magnified view). The region simulated was 3.5 mm from the entrance side of the device with a simulated length of 0.8 mm (total length of the trapping region is 16 mm; see Figure   3 in the main text for the region of the device simulated here). The profile line graph associated with the linear velocity across a series of interleaving channels (see yellow line in Figure S3B) at 10 μL/min is shown in Figure S3C. The linear flow velocity in the center of the interleaving channels was higher (~2.75 m/s; average) compared to the cross-channels forming the microtraps, where the velocity was ~0.5 m/s (average). The highest velocity reached inside the device at 10 L/min volumetric flow rate was ~3 m/s Step 1 was performed to fabricate the microtraps. SU-8 was spin coated onto the Si wafer and microtrapping structures (2 or 4 m depth) were produced by exposing the resist to UV light followed by development and etching.
Step 2 was performed on the Si wafer repeating the same processes as in Step 1, but with a different mask. Here, the depth of the interleaving channels were designed to be ~30 m.
Step 3, which is the final step, was performed to fabricate the deeper distribution channels with a depth around 150 m.
in the area where the fluid enters the input interleaving channels. As can be seen from Figure S3B, the input interleaving channels seems to show a rather constant linear velocity down the length of this channel network, but does decrease further down the device as noted in Figure 3, which is a consequence of the fact that these channels have a termination point. The output interleaving channels show the reverse behavior because the input side of these channels possesses the termination point.  Figure 2B of the main text with a dashed box. Note: input and output interleaving channels are labeled.

Input Channels
Output Channels

Output Channels
The velocity at all of the input interleaving channels near the walls and in the middle area of these channels were 5 X 10 -4 m/s and 1.75 m/s, respectively, as determined by COMSOL -this is consistent with the fact that at these flow rates, the flow is laminar with the velocity higher at the center of each channel and lower at the walls due to the no-slip condition (this was seen for all channels of the device, including the cross-channels). The line graph shows the flow profile at the entrance of the device where the velocity was higher (represented by positions 1, 4 and 6) at the center of the interleaving input channels and the velocity being lower at the interleaving output channels (represented by positions 2 and 5) as well as the cross-channels (position 3). This relationship is only seen at the input side of the device as indicated by the yellow line shown in Figure S3C. Figure S3D shows the shear stress in the input and output interleaving channels as well as the crosschannels as calculated using Newton's law (in this case, the same section of the device was simulated as shown in Figure S3B and C). The line graph in Figure S3E shows the shear rates at positions of the device as designated by 1-6 listed in Figure S3D. The interleaving channels had lower shear rate in the middle of the channel. Higher shear rates were found in close proximity of the walls (i.e., 10 µm), consistent with the results shown in Figure S3B and C. In the microtraps, higher shear stress was observed compared to the input/output interleaving channels. However, it is important to note that cells do not travel through the microtraps, see Figure S3A. Similar to the velocity, the shear rates decreased in both the input interleaving channels and the cross channels down the length of the device. Table 2  Cells were re-suspended in freshly prepared Carnoy's solution before used in FISH analyses. Cells were spotted onto a glass slide and the slides were immediately placed on a hot plate at 37°C and left to dry for ~15 min increasing the temperature to 80°C. The slides were treated with 70% acetic acid for 30 min and 2X SSC (pH 7.3) at 37°C (1 o C) for 5 min followed by dehydrating successively in an ethanol series (70%, 85%, and 100%) at room temperature for 2 min each and allowed to dry completely as predenaturation steps. Ten μL from the stock FISH probe solution was heated to 37°C ( 1°C). The FISH probe mixture (10 μL) was applied to each slide and covered with a cover slip carefully followed by sealing with rubber cement and allowed to dry completely. Cells with probes were denatured at 75 o C (

Phenotypic characterization of RPMI-8226 cells via flow cytometry.
CPCs primarily express CD38, CD138, and in a few cases CD56. Additionally, expression of CD45 is very weak or not expressed at all. Figure S4 provides flow cytometry scatter plots showing the expression patterns of CD56, CD38 and CD138 for RPMI-8226 cells, which serve as a model for CPCs associated with multiple myeloma. From the plot, 98% of the cells expressed CD38, 90% expressed CD138, and 74% expressed CD56 with <0.001% expressing CD45, consistent with literature data for this cell line [3,4].
On-slide immunophenotyping of the RPMI-8226 cell line was also used to further determine protein expression. This CPC model cell line expressed CD138 (FITC channel in Figure S5A(i)), CD38 (APC channel in Figure S5A(ii)), and CD56 (APC channel in Figure S5B(i)), but did not express CD45 (FITC channel in Figure S5B(ii)). Figure S6 shows the FISH procedure carried out on-slide as a control to validate our on-chip FISH processing and imaging. The slide preparation is represented as a schematic shown in Figure S7A for a general overview of the process. Figure S7B shows FISH processed cells (RPMI-8226) that were deposited onto a glass slide and imaged by a fluorescence microscope within a cytogenetic laboratory. The images were acquired using a Nikon 60x oil immersion objective. Image  microscope stage. Low magnification was used to spot the cells with FISH signal that could be further analyzed using a higher magnification objective. These points were then marked in the xy plane with multi-point imaging. Upper and lower thresholds were set to image all of the FISH probes contained within the nucleus of the cell and images were captured using automated imaging via a Keyence BZ-X700 microscope in several z-planes. Figures S8 and S9 show composite images from DAPI, FITC (green) and Cy5 (red) color channels.

On-slide (conventional) FISH.
The objective used was a Nikon CFI Plan APO VC 60X (oil) and N/A 1.4 with a working distance of 0.17 mm. Images were taken as a z-stack in the range of 15 µm varying the distance by 1 µm in the z-direction.
The final image was acquired by adding all z-plane images into a single composite image at a certain region of interest.
For the SUP-B15 cell line, it is expected to see two distinct red and two green signals with the TEL/AML1 FISH probes. The presence of 2 red and 2 green signals ( Figure S8A) were seen in >98% cells in agreement with the literature for this cell line. For the MLL break-apart probes, we would expect to see two red/green (or yellow) fusion signals and that was seen in most of the cells as noted from Figure   S8B. Figure  Images were acquired using a Keyence microscope and a Nikon 63x oil objective with DAPI, FITC (green) and Cy5 (red) filters. Figure S9. FISH processed images of SUP-B15 cell line showing signal for Philadelphia chromosome. Philadelphia chromosome is produced by the fusion of two genes in chromosome 9 and 22. Specifically, the translocation of ABL1 gene region (9q34.11-q34.12, red) chromosome 9 and translocation BCR gene region (22q11.22-q11.23, green) of chromosome 22 will result in Philadelphia translocation (t(9;22)(q34.12;q11.23), yellow). Images were acquired from Keyence microscope using Nikon 63x oil objective with DAPI, FITC and Cy5 filters. yellow). As is evident from the image in Figure S9, there are yellow signals in >90% cells, which indicated the presence of the Philadelphia chromosome.
To process the images, z-stacks were loaded in all 3-channels and brightness and contrast were adjusted in the blue (DAPI), green, and red channels. Then, the full focus function was used to focus the 3-D projection of the z-stack to one single image. The processed images were then filtered through "haze reduction" and "black balance" filters to isolate the cell nuclei in the blue channel and to enhance the FISH signals in the green and red channels. DAPI processed images were then loaded to FIJI and the threshold was adjusted such that all the nuclei were clearly visible. The adjusted image was then used with Dilate (2X) and Watershed functions to clear out the single nuclei. After cell nuclei were separated, a mask was created and the green and red channel images were processed with the image calculator and the "AND" function to eliminate noise outside the cell nuclei. Afterwards, three channels were color shaded with the appropriate filter colors and images were overlaid to achieve the final processed images with FISH signal.