ViaChip for Size-based Enrichment of Viable Cells

Abstract

Live cells acquire different fates including apoptosis, necrosis, and senescence in response to stress and stimuli. Rapid and label-free enrichment of live cells from a mixture of cells adopting various cell fates remains a challenge. We developed a ViaChip for high-throughput enrichment of Viable cells via size-based separation on a multi-stage microfluidic Chip. Our chip takes advantage of the characteristic increase in cell size during cellular senescence and decreases during apoptosis and necrosis, in comparison to their viable and healthy counterparts. The core component of our ViaChip is a slanted and tunable 3D filter array in the vertical direction (z-gap) for rapid and continuous cell sieving. The shape of the 3D filter array is optimized for target cells to prevent clogging during continuous separation. We demonstrated enrichment of live human and mouse mesenchymal stem cells in culture and from live animals, as well as the removal of senescent and necrotic MSCs, respectively, achieving an enrichment efficiency of ~67% with the continuous flow at 1.5 mL/hour. With further improvements in throughput and separation efficiency, our ViaChip could find applications in cell-based drug screening for anti-cancer and anti-aging cell therapies.

1. Introduction

Separation and sorting of cells from complex, heterogeneous mixtures is important yet challenging in biomedicine [1]. For example, purification of viable cells serves as the first step for cancer drug screening and cell therapy [2–4]. At least 85% cell viability for live cells is required for reproducible cancer drug screenings [3]. Current standard cell sorting techniques, such as flow cytometry [5] and magnetic-activated cell sorting devices [6], often require labeling with single or multiple “tags” or “labels” to identify cells of interest. Further limitations in these cell sorting platforms for clinical sample processing include (a) requirement of a large number of cells; (b) bulky instrumentation that occupies large bench footprints; (c) high operating pressures that could result in a loss of cell function or cell viability; and (d) increased risks of sample contamination in the lengthy process. Emerging label-free cell sorting methods that utilize the physical characteristics of cells can potentially overcome the aforementioned limitations. Label-free circulating tumor cell sorting methods using microfluidics have been reviewed previously [7]. Depending on the existence of the external force fields, label-free cell sorting methods can be classified into two main categories: passive and active methods [7]. Previous works have demonstrated active cell sorting by mechanisms including acoustophoresis [7, 8], dielectrophoresis [9], magnetophoresis [10], and centrifugation [11]. On the other hand, passive cell sorting methods include for example filtration[12–15], geometries of microstructure [16, 17], deterministic lateral displacement (DLD) [18, 19, 20, 21], initial microfluidics including spiral microchannel [22–25], contraction-expansion structure [26], and serpentine microchannel [27].

Passive cell sorting methods do not require the use of chemical reagents and have other advantages including short sample preparation time, no external force field operation, relatively high processing throughput, and high cell viability. For example, microfluidics using the DLD mechanism can usually process at a high flow rate with high accuracy and offer the capacity for sorting cells of a similar size. Liu et al. reported a capture rate of 99% and 80% for breast cancer cell lines MCF-7 and MDA-MB-231 from 10x diluted blood at 2mL/min, respectively [20]. However, DLD devices usually require complex structures and long channels, which can lead to excessive fluid resistance and the requirement of high driving pressures [7]. Furthermore, a single DLD device has the difficulty in processing whole blood samples due to clogging and cell-cell collisions. Inertial microfluidics is another type of passive method. It can achieve relatively high throughput and is easy to be integrated with other cell sorting methods [24].

However, its cell sorting performance may be affected by cell size and flow velocity. In addition, cell separation solely based on the size difference is insufficient for achieving high-accuracy separation owing to the heterogeneity of cells. Furthermore, the purity of target cells may be impaired when they are sorted from undiluted blood due to the cell-cell collisions [7]. Filtration, on the other hand, has the advantages of a relatively simple structure amenable for mass production. However, for one-dimensional filters, clogging is inevitable while processing high-density cell samples. Crossflow filtration, another type of filtration, offers greater potential for sorting similar-sized cells than other size-dependent methods [12, 14]. However, the cell attachment to micropillar or weir can still block the channel and the deformed target cells might pass through the pores or weirs, resulting in reduced sorting efficiency [7]. Among the reported passive cell sorting methods, it remains very challenging to enrich viable cells from a complex cellular mixture containing healthy, senescent, and necrotic cell populations of similar size through a single microfluidic chip.

We had previously reported a microfluidic chip (senescence-chip) to isolate a small number of senescent cells in a high throughput manner by taking advantage of the cell size increase during cellular senescence [12]. In this work, we further developed the technology for complex cell separation by taking advantage of the increased cell size characteristic of senescence [14, 28, 29], as well as decreased cell size observed in the later stages of necrosis [22, 30]. We designed a multi-stage ViaChip for size-based cell separation to increase the separation resolution and efficiency. We optimized the filter shape and the distance between the filters to reduce cell clogging when loading cells at a high concentration (> 10⁴/mL). Our design includes a tunable z-gap to allow small cells to flow through without experiencing the high pressure that might cause cell damage. We utilized a peristaltic pump, for future in vivo mouse experiments, to reduce the risk of contamination during cell sorting. As a proof of principle, we demonstrated that our ViaChip could operate at a flow rate of 50 μL/min for label-free separation and enrichment of viable cells from a complex mixture of healthy, senescent, and necrotic mesenchymal stem cells. The novelty and breakthrough nature of this work include the following. First, this is the first demonstration of a microfluidic chip (ViaChip) that could separate three different populations of cells with different cellular fates: live, senescent, and necrotic cells. Second, there were dramatic improvements from our previous senescence-chip that enabled our current ViaChip to be used for future in vivo mouse experiments. Third, we built an automatic cycling fluidic system to sort cells in a continuous flow format.

2. Material and Methods

2.1. ViaChip design and fabrication.

The 1-stage or 2-stage ViaChip (Figure 1) consists of a slanted filter array with an inclination angle relative to the fluidic flow within the chip, the height of the chip is 37, 40, or 55 μm depending on the cell types. The chip consists of two inlets and two outlets for the 1-stage ViaChip, while two inlets and three outlets for the 2-stage ViaChip. An array of circular posts with a diameter of 80 μm and a gap of 80 μm between posts is designed before the filter array to prevent large debris from entering the functioning module of the chip. All silicon master was fabricated by the standard photolithography including lithography, oxide deposition, Teflon deposition, and dry etching as detailed in the supporting information. The process scheme to fabricate the silicon master was shown in Figure S1. The fabrication features two lithography processes to create an etching depth difference between the filter array features and the channel structures. The polydimethylsiloxane (PDMS) soft lithography techniques [31] was then applied to fabricate the working chips. Polystyrene beads with varied sizes were purchased from Polysciences, Inc. (Warminster, PA) and Bangs Laboratories, Inc (Fishers, IN) and were used to validate the separation performance of ViaChip.

Figure 1.

The design, fabrication, and separation performance of the ViaChip. (a) ViaChip features and fabrication procedures. (b) The images of 2-stage and 1-stage ViaChip, respectively. The scale bar is 1 inch. (c) The images of the cross-section of the ViaChip show filter features and chip channel heights. The scale bar is 50 μm. The z-gap and the channel height for each application are listed in the table. (d) The percentages of 20 μm, 15 μm beads, and basal mMSCs collected at the outlet (i) by the 2-stage or 1-stage ViaChip.

2.2. Experimental setup.

An epifluorescence microscope (IX83, Olympus, Japan), connected with a CCD camera (QIClick, QImaging, Canada), and controlled by Olympus cellSens Dimension™ software, was used to observe and record the motion of the beads or cell separation process, and to take images of beads or cells inside the ViaChip. A three-channel peristaltic pump (EW-78001–70, Cole-Parmer, USA) was used to control the fluidic system. In all studies in this paper, the flow rates pumping to two inlets were 25 μL/min. Several other flow rates were also investigated, and the cell size distribution under different flow rates is shown in Figure S2. We found that 25 μL/min is the optimal flow rate in our microfluidic system for viable cell separation. Dulbecco’s phosphate-buffered saline (DPBS, 21–031-CV, Corning, USA) was used as diluent and buffer inlet when analyzing beads in the ViaChip, while for cell or mouse-derived bone marrow samples, we used cell culture media in the buffer inlet. All tubes connected to inlets or outlets were incubated with cell medium for 10 minutes before sample loading to reduce non-specific binding. The number of cells or polystyrene beads was visualized under the microscope and counted by a BioRad TC20 cell counter (BioRad, USA).

2.3. Cell culture and staining.

Human mesenchymal stem cells (MSCs) were derived from a 20-year-old male and purchased from Lonza (Lonza, 0000471980). Mouse mesenchymal stem cells (MSCs) were purchased from Cyagen (Strain C57BL/6 Mouse Mesenchymal Stem Cells, MUBMX-01001, Cyagen, USA). Cells were cultured in a humidified incubator (Symphony 5.3A CO₂ incubator, VWR, USA) at 37°C with 5% CO₂. MSCs basal medium (PT-3238, Lonza) with MSCGM™ SingleQuots supplement kit (PT-4105, Lonza) was used for hMSC culture, and MesenCult™ Expansion kit (5513, STEMCELL Technologies, Canada) was used for mMSC culture. Both media were supplemented with 1% penicillin-streptomycin (15140122, ThermoFisher Scientific, USA) to prevent cell contamination during culture. Basal MSCs were stained with either Hoechst (Hoechst 33342, H3570, ThermoFisher Scientific, USA) for 10 minutes or cell tracker (C2102, CellTracker™ green BODIPY™ dye, ThermoFisher Scientific, USA) for 15 minutes. In all cases, adhesive cells were detached using Trypsin/EDTA (CC3232, Lonza) at 37°C for 10 minutes. The size of trypsinized cells was measured by the analytical tool provided in the Olympus cellSens Dimension software.

2.4. Senescent cell model and staining.

hMSCs or mMSCs senescence was induced chemically using hydrogen peroxide or using X-ray radiation. The medium containing hydrogen peroxide (H₂O₂) was prepared by diluting 30% H₂O₂ solution (H1009, Sigma, USA) with MSCs culture medium to desired concentrations. For the H₂O₂ treatment experiments, the MSCs were cultured in the medium containing H₂O₂ at 37 °C for 3 hours. After that, the MSCs were washed with 1xDPBS 2 times and cultured in the fresh media for a variable number of days before analysis. For X-ray radiation treatment, MSCs were placed on a rotating table and exposed to 1 Gy, 4 Gy, or 6 Gy radiation. After treatment, cells were cultured for a variable number of days before analysis. The senescent cells were identified by the senescence detection kit (K320, BioVision, USA), which targets senescence-associated β-galactosidase (SA-β-gal) activity, following the manufacturer’s provided protocol. The senescence progression in response to H₂O₂ concentration or X-ray radiation dosage over time is shown in Figure S3 (hMSCs) and S4 (mMSCs). For the experiments using the mixture of basal and H₂O₂-treated hMSCs, the H₂O₂-treated hMSCs were stained with cell tracker (C34551, CellTracker™ orange CMRA Dye™ dye, ThermoFisher Scientific, USA), while basal hMSCs were stained with Hoechst for 10 minutes to differentiate two cell populations.

2.5. Necrotic cell model and staining.

To induce cell necrosis, the detached cells after culture were stored at a 4°C refrigerator (GDM-23-SCI-TSL01, VWR, USA) for a variable number of days. The ratio of necrotic cells increased with the storage time. The apoptotic/necrotic cells were identified by the apoptosis kit (K201, BioVision, USA), using annexin V-FITC and SYTOX™ green dye, by following the manufacturer’s provided protocol. The annexin V-FITC targets phospholipid phosphatidylserine on the surface membrane due to apoptosis, while SYTOX™ green dye targeting nucleic acid in the necrotic cells showed a higher level of green fluorescence. The fraction of necrotic cells as a function of the storage time is shown in Figure S5.

2.6. Mouse model and staining.

Male C57BL/6 mice at ten weeks of age were exposed to 6.5 Gy X-ray (n=4) or sham control (n=4), using a Precision X-ray Inc X-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). Mice were returned to their cage and left undisturbed for 6 days, then euthanized before collection of their bone marrow. Bone marrow samples were collected from the hind-leg femurs by flushing the contents of the marrow with approximately 5 mL of 1xPBS supplemented with 0.5% FBS and 8 μM EDTA. The bone marrow cell suspension was filtered through a 70μm nylon filter and further diluted with 1x DPBS buffer to a total volume of 10 mL. For samples that could not be processed on the same day of collection, Tirofiban (30 μg/mL; SML0246–10MG, Sigma, USA) was added to the bone marrow sample before storage at 4°C to alleviate cold-induced platelet aggregation[32]. The MSCs and Hematopoietic Stem Cells (HMCs) were identified by Hoechst, Stem Cells Antigen-1 antibody (anti-sca-1, 130–116-490), Platelet-derived Growth Factor Receptor-a antibody (anti-PDGFRa or anti-CD140a, 130–102-502), and c-Kit antibody (anti-c-Kit, or anti-CD117, 130–122-948). Anti-sca-1, anti-CD140a, and anti-CD117 antibodies were purchased from Miltenyi Biotech, USA. The procedure of cell staining and the gallery of cell images are summarized in the supporting information. All bone marrow samples were filtered with Pre-Separation Filters (130–095-823, Miltenyi Biotech) to remove large debris before loading to the ViaChip. For all experiments with the mouse bone marrow, the samples were treated with RBC lysis buffer (420301, BioLegend, USA) at room temperature for 30 minutes to reduce the number of red blood cells before staining or cell counting.

3. Results and Discussion

3.1. Fabrication and Characterization of the 2-stage ViaChip.

An illustration of a 2-stage ViaChip is shown in the top left in Figure 1(a), where the enlarged illustration of filter features on the x-y plane is highlighted by a blue box. The main features in Figure 1(a) include θ₁, θ₂, and ld which are 4°, 30°, and 8 μm, respectively, and the sd, shortest distance, is ~ 4 μm. The θ₁ is the inclination angle of the pillar array relative to the main fluidic flow. The θ₂ is the smaller angle in the single unit of pillar array. The ld is the longest inter-pillar spacing, while sd is the shortest inter-pillar spacing. The diameter of the particle or cell that is larger than 8 μm (cutoff particle 2R, where R is the radius) is expected to roll along with the filter feature and will not be captured in the space between two filter features. The mechanism of the cell rolling is detailed in our previously published paper [12]. The average cell size of human mesenchymal stem cells (hMSCs) and mouse mesenchymal stem cells (mMSCs) is larger than 8 μm, so they are least likely to be captured at the gaps between the filter features. The fabrication process of the ViaChip is shown in Figure 1(a) (right panel). Note that there will be a z-gap between the filter array and the glass substrate. The images of the 2-stage ViaChip (left) and 1-stage ViaChip (right) are shown in Figure 1(b). In the 2-stage ViaChip, the larger particles/cells are expected to roll over the filter array at the first stage then flow to the top branch at the second stage for further separation.

Figure 1(c) shows the cross-section of the ViaChip. The cross-sections of the ViaChip of various channel heights and z-gaps shown in Figure 1(c) include: (1) for hMSCs separation, (2) for mMSC or beads separation, and (3) for in-vitro bone marrow-derived mice samples separation. The z-gap and the channel height are marked, and their dimensions are listed in the table. By adjusting the z-gap, the ViaChip can separate cells of different sizes. Compared to our previous work using a syringe pump [12], the peristaltic pump was used in this study to minimize contamination since only the tubing of the pump will come in contact with the biofluids.

The separation efficiency for beads of 20 μm and 15 μm, and basal mMSCs, using 2-stage and 1-stage ViaChip was summarized in Figure 1(d). Both 2-stage and 1-stage ViaChip could sort 15 μm or 20 μm beads at an efficiency of > 80% to the outlet (i). However, compared to a 1-stage ViaChip (14.3%), much fewer basal mMSCs were collected at the outlet (i) when using a 2-stage ViaChip (2.8%). The flow controlled by a peristaltic pump generated a sinusoidal flow stream within the chip because of the pressure pulsation in the fluidic system. The sinusoidal flow stream caused the leaking of some small basal mMSCs to the outlet (i). By using a 2-stage ViaChip, most small basal mMSCs leaked in the 1^st stage flowed through the top branch to the outlet (ii), not the outlet (i). Only larger senescent cells were expected to remove from the cell solution and collected at the outlet (i). We observed the > 80% recovery rates of beads and mMSCs and found only a few beads/cells clogged in the filter array area when the loading concentration of bead/cell is > 10⁴/mL.

3.2. Separation of hMSCs by the 2-stage ViaChip.

The 2-stage ViaChip was used to separate hMSCs based on their cell fates. In our previous study [12], we showed that senescence of hMSCs can be induced by H₂O₂. The percentage of senescent cells increased with the post-culture time for both untreated and H₂O₂ treated hMSCs. The percentage of senescent hMSCs after treatment with 300 μM H₂O₂ increased from 9.9%, 16.5%, to 57.7% for 1, 3, and 7 days post-treatment (Figure S3). On the other hand, the senescence percentage of basal hMSCs increased from 8%, 12.3%, to 30.7% after 1, 3, and 7 days of culture without H₂O₂ treatment (Figure S3). The percentage of cellular senescence was significantly different between untreated and H₂O₂ treated cells 7 days post-treatment (p=0.0038). On the other hand, we found that the necrotic cell percentage of H₂O₂ (300 μM) treated hMSCs increased from 25% to 56% while basal hMSCs cells from 12 to 44% when stored at a 4°C refrigerator for 2 days (Figure S5)

The stored H₂O₂ treated hMSCs were prepared at a cell density of ~10⁴/mL as the sample input. The ViaChip with a z-gap of 12.1 +/− 2 μm and a channel depth of 54.9 +/− 1 μm was used for hMSC separation. The represented images of hMSCs running on the 2-stage ViaChip in the highlighted area (1) were shown in Figure 2(b–1). Most H₂O₂ treated hMSCs (red) were rolling on the filter array (white dot line in the image), while necrotic hMSCs (green) were able to pass through the filter array. We also observed at the area (3) in Figure 2(a) that many more green cells flowed to the lower branch than the upper branch. The timelapse images of the flow trajectory in 1.1 secs for an hMSC cell in the area (2) to (3) are shown in Figure 2(b-2, b-3). The cell flowed on the filter array then flowed to the upper branch of the ViaChip. Around 50–70% hMSCs acquired from the outlet (i) were senescent cells, as confirmed by the SA-β-gal ⁺ staining after collection from the outlet (Figure 2(b–4)).

Figure 2.

The separation performance of hMSCs of the 2-stage ViaChip. (a) The illustration of the 2-stage ViaChip with highlighted blue, red, yellow, and green boxes. (b-1) The H₂O₂ treated hMSCs flowed on or through the filter array in the area highlighted by a blue box. (b-2, b-3) Time-lapse images (0, 0.4, 0.8, 1, 1.1 sec) showing the movement of an hMSC in the area highlighted by red and yellow boxes. The scale bar is 200 μm. (b-4) A represented image of hMSCs acquired from the outlet (i), highlighted with a green box, after labeling with SA-β-gal. The scale bar is 50 μm. (c) The percentage of necrotic hMSCs in the sample input and each outlet. The input hMSCs were treated with 300 μM H₂O₂ for 3 hours. (d) The percentage of necrotic, senescent, and viable hMSCs in the sample input and each outlet. The input hMSCs were a mixture of basal and H₂O₂ treated hMSCs.

We also applied H₂O₂ treated hMSCs containing about 60% necrotic cells, to the 2-stage ViaChip. The percentage of necrotic hMSCs acquired from the outlet (i), outlet (ii), and outlet (iii) was 30.2+/−6.3%, 53.8+/−8.9%, and 70.1+/−11.2%, respectively (Figure 2(c)). The necrotic cell percentage in the solution collected at the outlet (i) was the lowest while highest at the outlet (iii), and cell percentage collected at the outlet (ii) was close to that in the sample input.

We next evaluated the separation efficiency of ViaChip for separating senescent and necrotic hMSCs. The basal hMSCs were mixed with H₂O₂ treated hMSCs to generate a mixture of ~40% senescent and ~35% necrotic hMSCs as the sample input. Applying this sample to the 2-stage ViaChip, the necrotic, senescent, and viable hMSC percentage at the outlet (i) was 18.5%, 60%, and 12.5%, at the outlet (ii) was 38.5%, 44%, and 23%, and at the outlet (iii) was 68.5%, 30%, and 4.5% (Figure 2(d)). The reagent in the senescence detection kit will not label the necrotic cells because the beta-galactosidase enzymatic activity, which catalyzes the hydrolysis of β-galactosides into monosaccharides [33], is lost in the necrotic cells. We observed that senescent cells were the dominant population at the outlet (i), while necrotic cells were the dominant population at the outlet (iii). The percentage of the viable cell population was lower in both outlet (i) and (iii), compared to the sample input. We observed that senescent hMSCs tended to flow to the outlet (i), while necrotic hMSCs tended to flow to the outlet (iii). The percentages of the three cell populations collected at the outlet (ii) were similar to that of the sample input. By adjusting the z-gap of the ViaChip, only smaller necrotic cells were allowed to flow through the pillar array to the outlet (ii) or (iii) with minimum deformation. Our results suggest that the 2-stage ViaChip separates hMSCs based on their cell fates, including senescent, necrotic, and viable (normal) cell populations.

Since our 2-stage ViaChip could separate hMSC based on their fates, with enrichment of senescent cells at the outlet (i), necrotic cells at the outlet (iii), and viable cells at the outlet (ii), we further explored its potential for high-throughput, high-efficiency viable cell enrichment, through a sequential separation procedure (Figure 3(a)). We reintroduced the cells collected at the outlet (ii) to sample inlet and repeated the procedure two more times. At each run, the cell size, the percentages of necrotic, senescent, and viable cells were measured. The basal and H₂O₂ (300 μM) treated hMSCs were mixed as the sample input, where the size of the basal and H₂O₂ treated hMSCs was 23.3 +/− 2.9 μm and 29.4 +/− 3.2 μm, respectively (Figure 3(b–1)). Both basal and H₂O₂ treated hMSCs were cultured for 7 days and stored at a 4°C refrigerator for another 2 days. The percentage of necrotic, senescent, and viable cells of basal was 65.9%, 16.5%, and 17.6%, respectively, while they were 37%, 51.7%, and 11.3%, respectively for H₂O₂ treated hMSCs (Figure 3(b–2)).

Figure 3.

The separation performance of hMSCs through the 2-stage steps ViaChip with multiple cycles. (a) The illustration shows the separation sequence of hMSCs using the 2-stage ViaChip, where the 2^nd and 3^rd sample inputs were the solutions collected from the outlet (ii) in the preceding cycle, respectively. (b) The cell input was the mixture of basal (38%): H₂O₂ (62%) hMSCs. (b-1) The size distribution of basal or H₂O₂ (300 μM, 3 hours) treated hMSCs. (b-2) The necrotic, senescent, and viable cell percentage in basal or H₂O₂ hMSCs. (c) The cell size distribution in the initial cell input, in the outlet (i)-(iii) after the 1^st run of separation, in the outlet (i)-(iii) after the 2^nd run of separation, and in the outlet (i)-(iii) after the 3^rd run of separation. The necrotic (d-1), senescent (d-2), and viable hMSC percentage (d-3) in the sample input, outlet (i)-(iii) at 1^st, 2^nd, and 3^rd run, respectively.

The basal and H₂O₂ treated hMSCs were mixed to have the sample input with ~ 30% necrotic cells, ~ 40% senescent cells, and ~30% viable cells. The cell size distribution in the sample input, from the outlet (i)-(iii) after 1^st run of separation, from the outlet (i)-(iii) after 2^nd run of separation, and outlet (i)-(iii) after 3^rd run of separation was shown in Figure 3(c). The average size of the cells in the sample input is 23.8 +/− 4.3 μm. The average size of cells from the outlet (i), (ii), and (iii) after 1^st run of separation was 27.8 +/− 5.1 μm, 25.5 +/− 4.3 μm, and 20.5 +/− 3.7 μm, respectively. The size of cells collected at the outlet (iii) was ~ 5 μm and ~ 7.3 μm smaller than those collected at the outlet (ii) and outlet (i), respectively. The difference between cell size from the outlet (iii) compared to the outlet (i) or between outlet (iii) and outlet (ii) is statistically significant with a p-value <0.001 and 0.006, respectively. After the 2^nd run of separation by ViaChip, the average size of cells from the outlet (i), (ii), and (iii) was 26 +/− 3.9 μm, 22.6 +/− 4 μm, and 18.9 +/− 2.9 μm, respectively. The size of cells collected at the outlet (iii) was ~ 3.7 μm and ~ 7.1 μm smaller than those collected at the outlet (ii) and outlet (i), respectively. The difference between cell size from the outlet (iii) and outlet (i) and between outlet (iii) and outlet (ii) is statistically significant with a p-value <0.001 and 0.024, respectively. After the 3^rd run of separation, the size of cells from the outlet (i), (ii), and (iii) was 23.1 +/− 3 μm, 22.7 +/− 3.3 μm, and 20.4 +/− 2.6 μm, respectively. The size of cells collected at the outlet (iii) was ~ 2.3 μm and ~ 2.7 μm smaller than those collected at the outlet (ii) and outlet (i), respectively. The difference between cell size from the outlet (iii) and outlet (i) and between outlet (iii) and outlet (ii) was statistically significant with a p-value of 0.035 and 0.07, respectively. However, the size difference between outlet (ii) and outlet (iii) was not different (p>0.05). In summary, the total cell size distribution became smaller after each run of separation on a ViaChip (Figure 3(c)), suggesting larger (senescent) and smaller (necrotic) cells were gradually removed at each run with the enrichment of viable cells.

We then analyzed the enrichment of each cell fate at the three outlets for the three cycles (Figure 3(d–1)). For all three runs, the cells collected from the outlet (i) had the lowest necrotic cell percentage while cells from the outlet (iii) had the highest percentage. After three cycles, the percentage of necrotic cells in the sample input decreased from 29% to 19%, and from 59% to 28% in the outlet (iii), suggesting that necrotic cells were removed gradually at each run. On the other hand, the cells collected from the outlet (i) have the highest senescent percentage while outlet (iii) had the lowest. After three cycles, the percentage of senescent cells in the sample input decreased from 39% to 33%, and from 52% to 36% in the outlet (i), suggesting that senescent cells were removed at each run. On the other hand, the percentage of viable cells in the sample input increased from 32% after the first cycle to 48% after the third cycle as shown in Figure 3(d–3). The percentage of viable cells increased in each outlet after the 2^nd and 3^rd runs. Overall, through three sequential cycles of separation by the ViaChip, we observed an enrichment in viable cells (from 32% to 48%), and the size distribution range became more uniform. Therefore our 2-stage ViaChip could increase the enrichment of viable cells by increasing the number of separation cycles.

3.3. Separation of mMSCs by the 2-stage ViaChip.

For separating the mMSCs, the z-gap of the 2-stage ViaChip was adjusted to 8.2 +/− 1.5 μm with the channel depth of 40.1 +/− 1.6 μm as listed in Figure 1(c). The average size (cultured for 1 day) of mMSC was 13.3 +/− 1.1 μm. We used H₂O₂ or X-ray radiation to induce cellular senescence of mMSCs as shown in Figure S4. The mMSCs were less resistant to the H₂O₂ compared to hMSCs, and a lower concentration of 150 μM was on mMSCs for three hours. On the other hand, a dosage of 6 Gy was used for X-ray radiation. After culturing for 1 day, the size of mMSCs treated with H₂O₂ or X-ray radiation was 16.8 +/− 2.5 μm or 17.1 +/− 2.7 μm, with an increase of 3.5 or 3.8 μm compared to basal mMSCs, respectively.

The H₂O₂ treated mMSCs were separated by the ViaChip (Figure 4). The average size of mMSCs collected at the outlet (i), (ii), and (iii) was 19.3 +/− 2.5 μm, 15.1 +/− 1.9 μm, and 13.6 +/− 1.7 μm, respectively. The average size of mMSCs collected at the outlet (iii) was ~ 1.5 μm and ~ 5.7 μm smaller than those collected at the outlet (ii) and outlet (i), respectively. The difference in cell size between outlet (iii) and outlet (i) was statistically different (p-value <0.001). In the case of mMSCs treated with 6 Gy radiation, the size of cells from the outlet (i), (ii), and (iii) was 19.2 +/− 3.2 μm, 15.5 +/− 2.6 μm, and 14.4 +/− 1.9 μm, respectively. The size of cells collected at the outlet (iii) was ~ 1.1 μm and ~ 4.8 μm smaller than those collected at the outlet (ii) and outlet (i) (p-value <0.001). Considering that the average cell size in the input solution was similar for these two cell inputs, the separation performance based on the cell size was very consistent.

Figure 4.

The separation performance of mMSCs through the 2-stage ViaChip. (a) The size distribution of H₂O₂ (150 μM, a-1) treated or X-ray (6 Gy, a-2) radiated mMSCs from the sample input and each outlet. (b) The purity of mMSCs > 15 μm at input, outlet (i), outlet (ii), and outlet (iii). mMSCs were either treated with H₂O₂ (red circle) or exposed to X-ray radiation (blue diamond). (c) The morphologies of basal mMSCs collected in the sample input, outlet (i), and outlet (ii) and post-cultured for 7 days. (d-1) The percentage of basal mMSCs in the sample input, outlet (i), and outlet (ii). (d-2) The doubling time of basal mMSCs in the sample input, outlet (i), and outlet (ii). (e-1) The illustration shows the separation sequence of mMSCs cyclic fluidic system, where outlet (i) connects with buffer inlet while outlet (ii) & (iii) connect with sample input. the 2^nd and 3^rd cell inputs are the solutions collected from the outlet (ii) of the 1^st and 2^nd cycle, respectively. (e-2) The percentage mMSCs in the outlet (i), outlet (ii)+(iii), and the recovery rate of mMSCs after 990 seconds of cycling operation. (e-3) The size distribution of H₂O₂ (150 μM, a-1) treated mMSCs from the sample input, outlet (i), outlet (ii)+(iii). The mMSC in the sample input is the mixture of basal and H₂O₂ (150 μM) treated mMSCs.

We next analyzed the purity of larger mMSC (>15 μm) at each outlet as shown in Figure 4(b). The threshold of 15 μm is used to estimate the capture rate and purity since the size change of senescent cells was dynamic. The purity rate was defined as (target cells collected in outlet/total cells collected in outlet) ×100%. The purity of cells > 15 μm was ~5%, 43.5%, 32.5% for the basal, H₂O₂ treated, and X-ray radiated mMSCs, respectively. The population of >15 μm cells significantly increased when mMSCs were treated with H₂O₂ or X-ray radiation. For H₂O₂ treated mMSCs, the purity of mMSCs >15 μm at outlet (i), outlet (ii), and outlet (iii) was 100%, 50%, and 16%, respectively. For X-ray radiated mMSCs, the purity of mMSC >15 μm at outlet (i), outlet (ii), and outlet(iii) was 93.8%, 68.2%, and 17.1%, respectively. For both experiments, the purity of mMSCs > 15 μm collected in the outlet (i) >90% suggested the ability of ViaChip to enrich > 15 μm mMSCs. We then analyzed the capture rate of mMSCs >15 μm through our ViaChip. The capture rate was defined as: (target cells collected in the outlet /target cell collected in inlet) ×100%. Combining outlet (i) and (ii), the capture rate of mMSCs > 15 μm at outlet (i) and (ii) for H₂O₂ treated and X-ray radiated mMSCs was 80% and 60%, respectively. Compared to the reported cross-filtration system, Miller et al. [13] applied Parsortix™ Cell Separation System to reach ~66% or ~ 92% capture rate of BT549 or SKBR-3 cells from the whole blood. The capture rate of our ViaChip will be affected by the size increase due to the senescence progression of cells. The larger the mMSCs due to the senescence progression, the higher the capture rate could be achieved.

We also separated the basal mMSCs through the 2-stage ViaChip under the sterile condition and studied the growth rate of cells collected from the sample input, outlet (i), and outlet (ii), respectively. The cell images after culture for 7 days are shown in Figure 4(c). The cell viability collected from all outlets (over cell input in the inlet) was > 95%. The cells collected at the outlet (i) and (ii) were healthy and could grow and proliferate. The percentages of cells in the sample input, outlet (i), and outlet (ii) were shown in Figure (d-1). As expected, most of the basal mMSCs flow to the outlet (iii). The doubling time of mMSCs collected from input, outlet (i), and outlet (iii) is 20, 22.6, and 19.7 hours, respectively. The mMSCs collected at the outlet (i) grew more slowly compared to the sample input and those collected at the outlet (ii). The slower proliferation of cells suggested a higher population of senescent cells collected at the outlet (i). This result shows that the 2-stage ViaChip can separate senescent cells from the viable cell population for basal mMSCs.

We next built an automatic cyclic fluidic system including sensors, valves, and bubble trappers to circulate fluidics in two closed loops. One loop connected the buffer inlet with the outlet (i), and another loop connected the sample input with outlets (ii) and (iii), as shown in Figure 4(e–1). The pump and sensors were controlled using a custom python script (see Supporting Information). Using the closed-loop system for separation of the mixture of basal and H₂O₂ treated mMSCs on ViaChip, the average percentages of cells in the outlet (i) or outlet (ii)+(iii) were 15.1 +/− 7.6% or 84.9 +/− 7.6%, respectively (Figure 4(e–2)). The recovery rate after a 990-second operation was 82.2%. The average size of mMSCs in the sample input or collected at the outlet (i), (ii)+(iii) after the 990-second operation was 16.7 +/− 3.2 μm, 19.9 +/− 1.9 μm, and 15.1 +/− 2.9 μm, respectively (Figure 4(e–3)). The average size of mMSCs collected at the outlet (ii)+(iii) was ~ 4.8 μm smaller than that at the outlet (i). Therefore, larger mMSCs in the buffer-outlet(i) loop can be separated through our automatic cyclic fluidic system, and this system could be further developed for in vivo applications in the future.

3.4. Separation of Mouse Bone Marrow Samples using the 2-stage ViaChip.

To investigate the size distribution of MSCs and Hematopoietic stem cells (HSC) in mouse bone marrow, bone marrow cells were collected and stained by Hoechst, anti-sca-1, anti-CD140a, and anti-CD117. The Hoechst stains the nucleus of MSCs and HSCs. The anti-sca1 conjugated with Vio® Bright FITC, anti-CD140a conjugated with PE, and anti-CD117 conjugated with APC were visualized in the green, red, and white color channel, respectively (Figure 5a and Figure S8). Sca1⁺CD140a⁺ double-positive cells are considered MSC [34, 35] while sca1⁺CD117⁺ double-positive cells are considered HSC [36, 37]. The mMSC cell line is anti-sca1⁺CD140a⁺ double-positive and CD117⁻ negative (see Figure S7 for representative images of stained cells).

Figure 5.

The separation performance of mouse bone marrow samples through the 2-stage ViaChip. (a) An image showing the staining of bone marrow-derived MSCs and HSCs from sham and 6.5 Gy irradiated mice 6 days after radiation exposure. The scale bar is 50 μm. (b) The size distribution of bone marrow-derived MSCs and HSCs from sham and 6.5 Gy irradiated mice. (c) The purity of cells > 10 μm in input, outlet (i), outlet (ii), and outlet (iii) for sham (red circle) and X-ray irradiated mice (blue diamond). (d) The size distribution of bone marrow-derived cells collected from the input, outlet (i), outlet (ii), and outlet (iii) from sham and 6.5 Gy irradiated mice. (e) The SA-β-gal⁺ cell fraction (indicated with an arrow) in the bone marrow of sham or 6.5 Gy irradiated mice (left). The scale bar is 20 μm. The average size of senescent cells found in the bone marrow of sham or 6.5 Gy irradiated mice (right). (f) The senescence cell ratio in the input, outlet(i), outlet (ii), and outlet (iii) for sham (red circle) and X-ray irradiated mice (blue diamond).

The size distribution of bone marrow-derived MSC and HSC found in sham or 6.5Gy irradiated mice are shown in Figure 5(b). Bone marrow-derived MSC was larger than HSC independent of X-ray exposure. We found that the average size of bone marrow-derived MSCs in 6.5 Gy irradiated mice (10 +/− 1.6 μm) was larger than that in sham irradiated mice (8.7 +/− 1.4 μm) (p-value < 0.001). The average size of bone marrow-derived HSCs in 6.5 Gy irradiated mice (8 +/− 1 μm) was also larger than that in sham control mice (7 +/− 1 μm) (p=0.012). The size of bone marrow-derived MSC and HSC from sham irradiated mice was close to previously reported values [38].

Compared to the size of mMSC cell lines in Figure 4, which was 13.3 +/− 1.1 μm, the size of MSCs from mice samples was 3–4.5 μm smaller on average, consistent with the fact that self-renewing cells are usually smaller in diameter [38]. To separate the mouse bone marrow-derived cells, the z-gap of the 2-stage ViaChip was further reduced to 5.1 +/− 0.5 μm and the channel depth of the chip was 37.1 +/− 2.6 μm (Figure 1(c)). To prevent coagulation within the chip, 1.5 mg/mL K₂-EDTA was added to mMSC culture media. Bone marrow samples, which contained K₂-EDTA already, from either sham or 6.5 Gy irradiated mice were allowed to flow through the 2-stage ViaChip. The purity of cells > 10 μm at sample input, outlet (i), outlet (ii), and outlet (iii) for sham irradiated mice (red circle) or X-ray irradiated mice (blue diamond) is shown in Figure 5(c). The purity of cells > 10 μm in the sham and 6.5 Gy irradiated mouse bone marrow samples was 6.9 +/− 0.4% and 11.2 +/− 0.4%. Around 4.3% more > 10 μm cells were observed in samples from 6.5 Gy radiated mice. If we excluded the red blood cells in the sample, the percentage of cells > 10 μm in mice bone marrow samples from the sham and 6.5 Gy irradiated mice was 21.2 +/− 0.5% and 36.6 +/− 2% (only count cells > 6 μm), respectively, showing a significant increase of > 10 μm cells in the irradiated mice bone marrow. The purity of cells > 10 μm at outlet (i), outlet (ii), and outlet (iii) for X-ray radiated mice was 17.7 +/− 4.2%, 14.6 +/− 0.4%, and 7.6 +/− 2.5%, respectively. The difference between outlet (i) and outlet (iii), and between outlet (ii) and outlet (iii) was statistically significant (p-value < 0.001). On the other hand, the purity of cells > 10 μm at outlet (i), outlet (ii), and outlet (iii) for sham irradiated mice was 7 +/− 0.1%, 7.2 +/− 0.1%, and 7.1 +/− 0.6%, respectively. There is no significant difference in purity of cells > 10 μm between the three outlets, and all percentages were close to that in the sample input. There was a significant increase in the purity of cells > 10 μm at the outlet (i) or outlet (ii) between 6.5 Gy irradiated mouse bone marrow samples and sham irradiated control samples.

As shown in Figure 5(d), the average cell size at the input, outlet (i), outlet (ii), and outlet (iii) for 6.5 Gy irradiated mice was 10.1, 9.9, 10.1, and 8.9 μm, while the average cell size at input, outlet (i), outlet (ii), and outlet (iii) for sham irradiated mice was 9.2, 9.4, 9.2, and 9.1 μm. For irradiated mouse-derived bone marrow samples, the average size of cells collected at the outlet (iii) was smaller than those at the outlet (i) or (ii). On the other hand, there was no significant difference in the average size of cells collected at each outlet in bone marrow samples from sham irradiated mice.

We next identified the SA-β-gal⁺ cells in the 6.5 Gy or sham irradiated mouse bone marrow samples at each outlet. Representative SA-β-gal⁺ images are shown in Figure 5(e). The average bone marrow-derived senescence cell size in sham or 6.5 Gy irradiated mice was 10.5 +/− 1.4 μm or 11.1 +/− 2 μm, respectively (Figure 5(e)) (p=0.12). After exposure to 6.5 Gy X-rays, mice were returned to their cage and left undisturbed for 6 days before euthanasia. We expect the size of the senescent cells in X-ray irradiated mouse bone marrow samples to further increase by increasing the time between X-ray irradiation and sample collection. We then counted the number of senescent cells in the sample input and solution collected from the outlet (i), outlet (ii), and outlet (iii). The senescent cell concentration in 6.5 Gy irradiated mice in the input, outlet (i), outlet (ii), and outlet (iii), was 8.6, 26.2, 5.2, and 5.4 per 1×10⁴ cells (population of cell size >10 μm), respectively (Figure 5(f)). The capture rate of senescent cells at outlet (i), outlet (ii), and outlet (iii) was 58.6%, 22.4%, and 8.6%, respectively. Significantly more senescent cells were observed in the outlet (i) compared to the other two outlets. On the other hand, the senescent cell concentration from sham irradiated mouse bone marrow samples in the input, outlet (i), outlet (ii), and outlet (iii) was 4.2, 3.6, 3.2, and 3.8 per 1×10⁴ cells, respectively (Figure 5(f)). There was no significant difference in senescent cell ratio among sample input and all three outlets. We observed a higher senescence concentration from the bone marrow sample of 6.5 Gy irradiated mice compared to that from sham irradiated mice. Because of the size increase in senescent cells in the 6.5 Gy irradiated mouse bone marrow samples, the senescent cell concentration in the outlet (i) was also significantly increased. It is worth noting that the total cells (excluding red blood cells) in the 6.5 Gy irradiated mouse samples was only 8.6% of that in the sham mice sample, indicating the severe reduction in cell numbers after X-ray radiation exposure.

The above results demonstrate that our ViaChip could achieve the separation of human mesenchymal stem cells from three different cell fates (Fig. 2), through the continual cycling separation (Fig. 3). We used mouse mesenchymal stem cells to validate the separation with > 95% cell viability (Fig. 4). We further showed the separation of senescent cells in bone marrow samples collected from 6.5 Gy irradiated mice (Fig. 5). These results pave the way for the future applications of removing unhealthy cells from live mice in vivo using our ViaChip. To accommodate the flow velocity of live cells flowing in the blood circulation of mice, the flow velocity of cells within the ViaChip has to be confined to a specific range. We experimented with the average flow velocity ~12 mm/s, close to the blood velocity in mouse retina arteriole blood vessels (8 to 20 mm/s, 25.3 mm of blood vessel diameter) [39]. Higher flow velocity may increase the risk to damage healthy blood cells in vivo. Processed under the current flow rate, the mMSC remained viable after flowing through our ViaChip and could be cultured further (Fig. 4(d)). The processed flow rate could be increased in the future by several methods, such as parallel ViaChip processing, stacking microfluidics layers, as well increase in the ViaChip dimension.

4. Conclusions

In summary, we have developed a 2-stage ViaChip that incorporates 3D filter arrays with a tunable z-gap, for the enrichment of the viable human MSCs by removal of senescent and necrotic cell populations. Our 2-stage ViaChip could also isolate senescent mouse MSCs. The mouse MSC population after senescent cell removal grows faster on average than those before removal. Moreover, ~58.6% of senescent cells found in bone marrow samples from mice treated with X-ray radiation could be collected at the outlet that is for senescent cell removal. We also validated that by adjusting the z-gap, the ViaChip could separate different cell lines with different average cell sizes. The 3D filter array in the ViaChip is optimized to prevent the clogging for live-cell separation, and maintain the healthy states of cells for subsequent cell culture. Compared to the existing microfluidic technologies, our platform has distinct advantages. It works directly with bone marrow samples with very minimum clogging and coagulation. The separation can operate in a continuous flow with a flow rate of 1.5 mL/hour. The 2-stage design can minimize the cell leaking to side channels and allow the usage of the peristaltic pump to make senescence/necrosis in vivo dialysis feasible. Our ViaChip could find potential applications in in-vitro and in vivo cell separation and potentially be utilized for drug screening and cell therapies.

Highlights.

• Label-free enrichment of viable mesenchymal stem cells from senescent and necrotic populations

• Cycling separation of cells in-vitro from mouse bone marrow samples on a microfluidic chip

• Adjustable multi-stage microfluidic chip for studying different fates of various cell types

Author Biographies

Dr. Po Ying Yeh is a Senior Engineer at Newomics Inc. Dr. Yeh received his Ph. D. degree in Mechanical Engineering from the University of British Columbia, Canada in 2009. His current research interests are cellular senescence and microfluidics.

Dr. Antoine Snijders is a Staff Scientist at the Lawrence Berkeley National Laboratory. Dr. Snijders received his M.S. in Biomedical Sciences from the VU Amsterdam (The Netherlands) in 1999. In 2004 he received his Ph.D. cum laude at the University of Utrecht (The Netherlands) in Cancer genetics and Molecular Biology. Dr. Snijders’s laboratory has over 20 years of experience in using cell culture and mouse model systems to identify phenotypic outcomes associated with environmental exposures by using a strategy that integrates systems genetics, the microbiome and discovery approaches with mechanistic information to ultimately address key questions concerning the effects of environmentally relevant exposures on human health.

Dr. Daojing Wang is the Founder and CEO of Newomics Inc. Dr. Wang had been a Principal Investigator at Lawrence Berkeley National Laboratory between 2002 and 2013. Dr. Wang obtained his B.S. degree from University of Science and Technology of China (USTC) in 1994, Ph.D. in chemistry from Princeton University in 1999, and completed his postdoctoral training at the University of California, Berkeley in 2000. Dr. Wang’s current research interests are developing platforms and solutions to enable precision medicine.