Senescence chips for ultrahigh‐throughput isolation and removal of senescent cells

Summary Cellular senescence plays an important role in organismal aging and age‐related diseases. However, it is challenging to isolate low numbers of senescent cells from small volumes of biofluids for downstream analysis. Furthermore, there is no technology that could selectively remove senescent cells in a high‐throughput manner. In this work, we developed a novel microfluidic chip platform, termed senescence chip, for ultrahigh‐throughput isolation and removal of senescent cells. The core component of our senescence chip is a slanted and tunable 3D micropillar array with a variety of shutters in the vertical direction for rapid cell sieving, taking advantage of the characteristic cell size increase during cellular senescence. The 3D configuration achieves high throughput, high recovery rate, and device robustness with minimum clogging. We demonstrated proof‐of‐principle applications in isolation and enumeration of senescent mesenchymal stem cells (MSCs) from undiluted human whole blood, and senescent cells from mouse bone marrow after total body irradiation, with the single‐cell resolution. After scale‐up to a multilayer and multichannel structure, our senescence chip achieved ultrahigh‐throughput removal of senescent cells from human whole blood with an efficiency of over 70% at a flow rate of 300 ml/hr. Sensitivity and specificity of our senescence chips could be augmented with implementation of multiscale size separation, and identification of background white blood cells using their cell surface markers such as CD45. With the advantages of high throughput, robustness, and simplicity, our senescence chips may find wide applications and contribute to diagnosis and therapeutic targeting of cellular senescence.

UV exposure to transfer the pattern from a mask to the photoresist layer, the silicon wafer was developed to generate a pattern on photoresist. After hard-bake, the wafer was etched by DRIE to produce channels with the desired depth. For the senescence chip, the channel depth was controlled between 30 and 35 µm. Finally, a Teflon layer was deposited on all surfaces of the silicon wafer to ensure a smooth PDMS peeling-off process. The senescence-chip was prepared by bonding the PDMS-replica channel onto a glass slide after treated with plasma (PDC-001, Harrick Plasma, USA). For parallel-processing chip, five layers of identical PDMS channels were stacked up with the inlets and outlets aligned along the vertical direction. Before use, the device was incubated at 65 °C overnight to prevent the fluid leakage and confirmed by the flow-through of 1x PBS buffer.
Microtubing was connected to the PDMS channels for fluid delivery. The devices were disposed after each run of biological samples.

Experimental
Setup. An epifluorescence microscope (IX83, Olympus, Japan) connected with a CCD camera (QIClick, QImaging, Canada) was used to observe and record the cell separation Supporting Information 3 process inside the microfluidic channel. The blood sample and 1x PBS buffer with 0.05% BSA were stored in syringes (BD Biosciences, USA). A 0.45 µm syringe filter (Acrodisc, Pall Life Sciences, USA) was connected to the syringe storing the PBS buffer to prevent contaminant from flowing into the microchannel and clogging the pillar array. Two infusion syringe pumps (NE-1600, New Era Pump Systems, USA; and KDS 100, KD Scientific, USA) were used to control the flow rates. When the senescence chip was tested with cells spiked in the undiluted whole blood, the buffer flow rate was usually 3 times of that of the blood sample. This ratio was decreased to 2 when the chip was tested with low concentration of cells or beads solution alone. Due to the sedimentation of cells, the syringe containing blood sample was vertically positioned to ensure that most of the cells flowed into the microtubing. When the senescence chip contains only one outlet with capture arrays for cell trapping, a longer microtubing was connected to the other outlet to balance the hydrodynamic resistance through both outlets. Before separation, 1x PBS buffer with 0.05% BSA flowed through the microchannel and microtubing for ~15 min to remove any remaining air bubbles and reduce nonspecific bonding to the channels. were purchased from Lonza (Lonza, Swiss). The log number used in this study was 0000471980 Supporting Information 4 (derived from a 20-year-old male). MSCs were maintained in humidified incubators at 37 °C with 5% CO2, and cultured with MSCs basal medium (Lonza) supplanted with 5% FBS. MSCs at passage 6 were cultured on 12-well plates with proper densities to avoid over confluency over a 6day period. The initial number of cells for each condition is shown in Table 1 and Table 2. For hydrogen peroxide (H2O2) treatment, 30% H2O2 solution (Sigma, USA) were diluted with MSCs basal medium into desired concentrations. Media containing 100 µM and 200 µM H2O2 as well as the basal medium control were used to incubate MSCs at 37 °C for 2 h. After that, the MSCs were washed with 1x PBS solution 3 times and cultured in the fresh media for another 3 days before analysis. For X-ray treatment, MSCs were placed on a rotating table and exposed to 1 Gy, 4 Gy or sham (0 Gy), using a RAD320 320 kVp X-ray machine (Precision X-ray Inc., North Branford, CT), operated at 300 kV, 10 mA (dose rate of 1.3 Gy/min). Cells were cultured for another 3 days and 6 days, respectively, before analysis.  For mouse bone marrow samples, 10 weeks old, male wild-type mice (stain C57BL/6) were exposed to the total body X-ray irradiation at 0 Gy (sham), 1 Gy, 4 Gy, and 6.5 Gy, with 4 mice at

Device Operation
Senescence chip for analysis of senescent cells in biofluids. The senescence chip with a 4 µm 3D filter array and a cell trapping array was used to isolate MSCs from whole blood, capture MSCs on chip, and conduct single cell analysis in situ after capture. 2 mL of fresh undiluted human whole blood spiked with ~500 fixed senescent MSCs induced by either H2O2-or X-ray was injected into the senescence-chip at a flow rate of 3 mL/h. For mouse bone marrow samples, we aliquoted ~1 x 10 6 bone marrow mononuclear cells (BM-MNCs) from each sample. The aliquots were diluted Supporting Information 6 into 2 mL with 1x PBS before loading directly on our chips for cell separation. 1x PBS buffer with 0.05% BSA was injected from another inlet at a flow rate of 9 mL/h. After MSCs were captured on the cell trapping array, the flows of cell sample and buffer were stopped, and followed by a gentle injection of staining solution to fill the whole channel and tubing. The inlet tubing was kept during incubation to generate a balance pressure and prevent backflow of the trapped cells. During the separation and staining processes, care was taken to avoid air bubbles inside the channel. After incubation, color images of the captured MSCs were recorded with the microscope for analysis.
All the experiments were repeated at least 3 times.

Senescence chip for removal of senescent cells from whole blood.
A senescence chip with a 13 µm 3D filter array was used to remove senescent MSCs from blood. Before spiked into human whole blood, the fixed senescent MSCs induced by either H2O2 or X-ray were stained overnight respectively. Therefore, the percentage of senescent MSCs in each sample was determined by the ratio of senescent MSC number to total MSC number.

Quantification of senescent cells on senescence chips. After staining the MSCs on chip
overnight, the color images of MSCs were recorded with a CCD camera in RGB mode. Microscope lamp intensity was consistent at 5 V. The images were then imported into ImageJ software to isolate their red channels, which were used to identify senescent MSCs. The grayscale of the dark region for each cell was measured with ImageJ, which define the senescent MSCs with a value smaller than 40.

RESULTS AND DISCUSSION
Design and working mechanism of senescence chips. We first developed the senescence chip which monolithically integrate two rows of tilted 3D filter array for size-based cell separation with all necessary inlets and outlets for samples and buffers (Figure 1). Two types of senescence chips were designed for different purposes. For analysis of senescent cells in small volumes of whole blood or bone marrow, the senescence chip contains a 3D-filter array to isolate MSCs, followed with a cell trap array to capture MSCs after separation for enumeration and single cell analysis of Supporting Information 8 senescent cells (Figure 1a-i). For rapid removal of senescent cells from whole blood, the senescence chip does not contain cell traps but the chip outlet is connected directly to a tubing to remove senescent cells from whole blood (Figure 1a-ii). The other end of the tubing goes to a waste or a collection tube for further analysis if needed.
We performed modeling to optimize the design of our chips (Figure 1b). A 3D filter array contained PDMS micropillars inside a channel to achieve cell separation on the x-y plane as well as in the z direction. On the x-y plane, two key parameters were taken into consideration, which were the inclination angle (θ) of micropillars relative to the main fluidic flow, and the inter-pillar spacing (d) as shown in Figure 1b-i. The pillar shape was also optimized to minimize clogging and maximize cell separation. Two types of quadrangle pillars were designed as shown in the zoom-in of Figure 1a. When moving down on the filter arrays, the particle tends to be trapped by the sharp edge of the Type-A pillars. In contrast, the particle contacts a tilted surface on Type-B pillars, which is easier to move on. Therefore, the Type-B pillars show a better performance in particle separation. For the rigid particles with diameters smaller than the pillar spacing (d), they could directly pass through the filter. When a particle has a diameter larger than the pillar spacing, we divided the hydrodynamic drag force (F) into two portions, parallel to the filter (inclination plane, F1) and perpendicular to the filter (F2). To ensure the particle could roll down on the filter, the should be established, in which L1 and L2 are the arms of forces F1 and F2. In Eq. (1), F1 and F2 could be expressed as F•cosθ and F•sinθ, while L1 and L2 could be expressed as (R2-1/4d2) 1/2 and 1/2d, in which R is the radius of the particle. Therefore, Eq.
(1) could be expressed as d < 2R/(tan2θ + 1) 1/2 (2) Supporting Information 9 From Eq. 2, a smaller pillar spacing (d) and filter angle (θ) would help particles to roll down on the pillars as shown in Figure 1a. In our design, we set the angle of filter array (θ) at 5 o . The pillar spacing (d) (also called "filter size" hereafter) is varied based on the size of cells to be isolated.
In the z direction, the PDMS pillars do not bond to the glass substrate because of their small topsurface area. Therefore, depending on the operational flow rate in the channel, an opening with varying size is created between the pillars and the glass substrate, which works like a shutter and allows smaller cells to pass through (Figure 1 b-ii and iii). For example, during the separation of MSCs from whole blood, RBCs and WBCs can easily pass through the filter from both the zdirection and x-y plane, while the MSCs with a larger size will not cross the filter but instead roll down. Our simulation shows that, compared to the 2D filter array ( Figure S1, a and c), 3D filter array ( Figure S1, b and d) can generate much more uniform flow velocity across the channel.
Therefore, the design of our 3D filter array could better reduce the system backpressure, reduce clogging of the filter, and improve the throughput.