Technical Note pubs.acs.org/ac
Method To Purify and Analyze Heterogeneous Senescent Cell Populations Using a Microfluidic Filter with Uniform Fluidic Profile Minseok S. Kim,*,†,‡ Seonghyeon Jo,† Joon Tae Park,† Hyun Young Shin,† Sun Soo Kim,§ Ogan Gurel,∥,⊥,# and Sang Chul Park† †
Well Aging Research Center, Samsung Advanced Institute of Technology, Samsung Electronics, Ltd., 130, Samsung-ro, Yeongtong-gu, Suwon-si, Gyenggi-do, Korea ‡ Department of Biomedical Engineering, Konyang University, 158 Gwanjeodong-ro, Seo-gu, Daejeon, Korea § R&D Solution Laboratory, Samsung Electronics, Ltd., Maetan3-dong, Youngtong-gu, Suwon-si, Gyeonggi-do, Korea ∥ Open Innovation Group, Samsung Advanced Institute of Technology, Samsung Electronics, Ltd., 130, Samsung-ro, Yeongtong-gu, Suwon-si, Gyeonggi-do, Korea ⊥ Campus D, 20, Yangpyeong-ro 21-gil, Yeongdeungpo-gu, Seoul, Korea # Samsung Advanced Institute of Health Sciences and Technology, Irwon-ro 81, Gangnam-gu, Seoul, Korea S Supporting Information *
ABSTRACT: To precisely purify and study aged (senescent) cells, we have designed, fabricated, and demonstrated a novel diamondstructure (DS) microfluidic filter. Nonuniform flow velocities within the microfilter channel can compromise microfluidic filter performance, but with this new diamond structure, further optimized via simulation, we achieve a uniform microfilter flow field, improving the throughput of size-based separation of senescent cells, as obtained by 39-passaged human dermal fibroblasts. After separating these aged cells into two groups, consisting of large- and small-sized cells, we assessed senescence by measuring lipofuscin accumulation and βgalactosidase activity. Our results reveal that even though these senescent cells had been equivalently passaged in culture, a high degree of size distribution and senescent phenotype heterogeneity was observed. In particular, the smaller-sized cells tended to express a younger phenotype while the larger aged cells demonstrated an older phenotype. We suggest that size-based separation of senescent cells, subtyped into small- and large-sized cohorts, offers an alternative method to purify such aged cells, thereby enabling more precise study of the mechanisms of aging, autophagy impairment, and rejuvenation.
C
clonal heterogeneity is understood to substantially hinder the study of cellular aging.8 To fully study replicative senescence and rigorously elucidate mechanisms of cellular aging, isolating the true senescent cells from these heterogeneous clonal populations represents a critical, first step in the study of aging. One representative characteristic of cell aging in this context is an increase in cell surface area.9 We hypothesized, therefore, that larger cells, as captured and isolated via the novel microfluidic technique described here, might exhibit a more senescent phenotype than the smaller cells, even though both size groups derive from the same clonal culture. Our hypothesis suggests two important implications: first, that conventional methods of isolating senescent cells, based on sequential subcloning, may not result in the homogeneous cell populations desired, and second, that cell size separation techniques can be used to more precisely isolate pure senescent
ellular senescence, as originally described by Hayflick and colleagues,1 involves a process by which the proliferation of cultured human cells decelerates and ultimately ceases. On the basis of this definition, senescent cells can be created in vitro via prolonged division in culture since replicative senescence is defined as a state in which normal somatic cells lose their proliferative capacity, causing irreversible growth arrest.2 These senescent cells have served as a model for studying aging and age-related disease, in which age-related changes have been observed in terms of cell morphology,3 metabolism,4 and reduced proliferative potential5 as well as the loss of responsiveness to growth factors.6 The replicative life span of human fibroblasts, even in single clonal populations, has shown, however, a high degree of phenotypic heterogeneity. With such clonally derived senescent cells, a certain fraction still maintains the capacity to proliferate, while other cells have entered a state of irreversibly arrested growth. This heterogeneity results in morphological and growth-rate variability, as well as differing responsiveness to signals and stochastic discrepancies in gene expression.7 In particular, this © XXXX American Chemical Society
Received: February 2, 2015 Accepted: August 31, 2015
A
DOI: 10.1021/acs.analchem.5b00445 Anal. Chem. XXXX, XXX, XXX−XXX
Technical Note
Analytical Chemistry
Figure 1. Schematic of the diamond structure (DS) microfilter and size-based separation of aged cells. (A) Fluidic velocity distribution in a wide microchannel (top) as compared to uniform flow profile in the optimized diamond structure (bottom) (B) Preparation of aged cells. Normal human diploid fibroblasts (HDFs) were replicatively subcultured (passaged 39 times) to prepare aged cells. (C) Separation of aged cells. These senescent cells were detached from the Petri dish and the suspended cells were injected into the DS microfilter. Cells passing through (escaping) the microfilter, constituting the small-sized cohort, were collected. The microfilter was then inverted and medium was allowed to flow in the reverse direction. The captured cells were then isolated, thereby harvesting the large-sized group.
After isolating these differentially sized aged cells, we then analyzed and compared their senescence phenotypes. It is wellknown that autophagy impairment induces premature senescence, which can be measured via autofluorescence as a marker for accumulated lipofuscin21,22 and by senescenceassociated β-galactosidase activity23 as a characteristic of growth-arrested cells. We therefore used these analytical techniques to measure the extent of aging in the size-separated aged cells isolated by our DS microfilter. Our results show that even with cells equivalently passaged in culture (sequential subculture), a high degree of phenotypic heterogeneity still exists. We found, in particular, the presence of two subtypes of senescent cells: a smaller, younger cohort, which we term YPACs (young phenotype aged cells) and the OPACs (old phenotype aged cells) which tend to be about 70% larger in size. These results have important implications for cellular senescence, both with respect to understanding mechanisms underlying aging and, more generally, offering a new technique for studying cellular aging.
cell types. More precise purification of senescent cells as well as a practical technique to efficiently achieve that would represent a significant advance in aging research. Several cell separation technologies are available, including acoustic separation,10 deterministic lateral displacement,11 hydrodynamic separation,11 dielectrophoresis,12 and hydrophoresis,13 although we have adopted the microfluidic filter method because it is one of the simplest and most direct methods for the separation and isolation of particles on the basis of their different sizes. Although microfluidics allows the precise manipulation and characterization of biological samples,14−17 throughput with large volumes has been limited. Several different approaches have tried to overcome this challenge. 18,19 Simply enlarging the channel width to accommodate greater microfluidic flow is one such approach, but this results in a nonuniform velocity distribution within the microfilter channel. Flow uniformity has not been closely studied in microfluidic filter applications, but when this technology is applied to size discrimination among biological cells, velocity variability becomes significant. Because cells are deformable, a nonuniform flow field causes the captured cells to experience different fluidic forces as a function of their position within the filter. This phenomenon seriously compromises the ability to discriminate between different cell sizes, thus limiting filter performance. Some innovative designs have reported uniform flow profiles, but their highly complex structures and narrow channels may limit practical use for larger particle sizes, especially in the size range necessary for biological (cell) separation.20 In this paper, we present a novel and simple diamondstructure (DS) geometry, further optimized via simulation, by which a microfluidic filter with uniform fluid profile can be implemented in practical use (Figure 1A). After preparing monocultures of aged cells (Figure 1B), with this technique we have been able to efficiently isolate these senescent cells into cohorts of different sizes (Figure 1C). This allows us to prepare a homogeneous population of senescent cells, at least on the basis of cell volume.
■
EXPERIMENTAL SECTION Cultivation of Senescent Cells and Separation of Aged Cells. Normal human diploid fibroblasts (HDFs) were obtained from the foreskin of an 11 year-old donor (M11 strain) as previously described.24 HDFs were cultured in Dulbecco’s modified Eagle’s medium (DMEM) containing 25 mM glucose supplemented with 10% fetal bovine serum (FBS), 100 units/mL penicillin, and 100 μg/mL streptomycin. When confluent, the cells were split 1:4 during early passages and 1:2 during late passages. The culture medium was changed every 4 days. The number of population doublings (PDs) (n) was calculated using the equation: n = log2 F/I, where F and I are the number of cells at the end and the beginning of one passage, respectively. Once the population doubling time of the cells exceeded 14 days, the cells were considered to be senescent. The 39-passaged human dermal fibroblasts were detached with trypsinization, after which the suspended cells were injected into the microfilter inlet reservoir (Supporting B
DOI: 10.1021/acs.analchem.5b00445 Anal. Chem. XXXX, XXX, XXX−XXX
Technical Note
Analytical Chemistry Information Figure S1). The syringe pump connected to the microfilter outlet was operated in withdrawal mode. The flow rate was 100 μL per min. Senescent Measurements on Aged Fibroblasts. The extent of senescence phenotype was assessed via two methods: senescence-associated β-galactosidase (SA-β-Gal) assay and lipofuscin autofluorescence visualization. SA-β-Gal activity was assayed using a senescence-associated β-galactosidase staining kit (Cell Signaling Technology, Inc., MA) according to the manufacturer’s protocol. When the separation process was completed, the separated cells were incubated in a CO2 incubator for 2 days to ensure sufficient cell adhesion and stable conditions for these separated cells, after which SA-β-Gal staining was conducted. Briefly, the cells were fixed with 1 mL of 1× fixative solution for 15 min after washing with 1× PBS and then incubated overnight in 1 mL of β-galactosidase staining solution prior to rinsing with 1× PBS. The cells were photographed under an inverted microscope for detection of SA-β-Gal activity. Lipofuscins are regarded as a hallmark of aging and brownyellow, autofluorescent granules which have been shown to accumulate during normal aging.25 Lipofuscin autofluorescence was observed at 488 nm using an inverted fluorescence microscope (Carl Zeiss, Germany).
Figure 2. Fluidic velocity at different positions within the filter. (A) Calculated velocity profile for the optimized DS design. The fluidic velocity was uniform for most of the filter area. (B) Measured fluidic velocity at different filter positions by flowing microbeads. The flow velocity was significantly nonuniform in the filter without DS. For the non-DS device, cells experience entirely different fluidic forces, depending on their position along the filter. In contrast, the optimized DS filter showed a mostly flat (highly uniform) fluidic velocity profile throughout the entire filter area (CV: 8.4%).
■
RESULTS AND DISCUSSION Optimization of Diamond Shape Dimension for Uniform Flow Velocity. To achieve not only sufficient flow rate but also reduced fluidic velocity, one simple solution is to use a wider channel; however, when this concept is applied to a microfluidic filter, significant variation in the fluidic velocity is observed at the filter area. The flow profile is significantly nonuniform where we observed only 6% of the entire filter area had velocities that varied by less than 5% from the average (Supporting Information Figure S2). This highly nonuniform velocity phenomenon could clearly be confirmed by analyzing the flow of microbeads within the filter chip. At the filter center, these particles experienced high flow velocities (Supporting Information Figure S3A), but toward the filter edge, the microbeads moved slowly and accumulated in the boundary area (Supporting Information Figure S3B). To overcome the nonuniformity, we placed a diamondshaped structure in front of the filtration area (e.g., between the inlet port and the filter) to equalize the central and peripheral streamlines. The diamond geometry was optimized via CFD simulations, in combination with the DOE method for computational efficiency. We performed a two-factor, threelevel full factorial DOE and found that after two orthogonal geometric variables were optimized, 98% of the entire filter area experienced fluid velocities that were within 5% of the average flow velocity. Moreover, this DS design avoids excessive pressures that can damage cells passing through the filter, an important consideration if one wishes to maintain the viability of these cells after filtration. More details regarding the procedure optimizing the DS geometry are described in Supporting Information Figure S4. Characterization of the DS filter and size-based separation of cells. We performed fluidic simulations on the diamond-structured filter as optimized by the combination of CFD simulation and DOE calculations (Figure 2A). In comparing the filter with and without DS (Supporting Information Figure S2), the DS device exhibited an extraordinarily uniform fluidic velocity profile at the filter.
The microfluidic filter channel described in this study is 28 mm in width and five equidistant points (labeled I, II, III, IV, and V in Figure 2) were selected along the filter dimensions. We then measured the velocity of microbeads with a high speed camera at each of these five points. For the non-DS filter, the maximal fluidic velocity occurred at the center of the filter (position III) progressively decreasing toward the filter channel edges with the standard deviation (SD) and coefficient of variation (CV) being 3.39 mm/s and 100% respectively. In contrast, for the DS filter, the microfilter area exhibited a highly uniform, nearly flat flow profile, where the SD and CV were 0.34 mm/s and 8.09% (Figure 2B). In certain cases (for example, with the high particle concentrations found in blood samples or when particle sizes are larger than the interstructure gaps), it can be desirable to have no structure in front of the filter area. For our case, we changed the inlet and outlet of the filter and verified the fluidic velocity at the filter area. Our results showed that the average fluidic velocity was similar (velocity difference: within 5%) and the uniformity of fluidic velocity was maintained (CV: 6.56%), although the flow direction was reversed (Supporting Information Figure S5). Heterogeneity of Senescence Phenotype As a Function of Cell Size. After confirming uniform fluidic velocity throughout the filter area, we prepared our cultured cells for size filtration. We obtained aged cells via serially passaging human diploid dermal fibroblast cells 39 times in a Petri dish, detached these cells from each other, and then injected the suspended cells through the DS filter inlet port. Relatively small-sized cells, which passed through the filter (“escaped”), were harvested from the filter outlet, and largeC
DOI: 10.1021/acs.analchem.5b00445 Anal. Chem. XXXX, XXX, XXX−XXX
Technical Note
Analytical Chemistry
Figure 3. Size-based separation of 39-passaged old human diploid fibroblasts via the DS filter. (A) An image of cells escaping the filter (small-sized group). Scale bar, 50 μm. (B) An image of cells captured by the filter, then subsequently retrieved from the filter in the second, reverse flow stage (large-sized group). Scale bar, 50 μm. (C) Size distribution of the two groups. Although successively subcultured within the same Petri dish condition, the captured cells were 70% larger on average than the escaped cells. The asterisk denotes statistically significant difference (*p < 0.001, one-way analysis of variance analysis) between the groups.
sized cells were retained (“captured”). We were able to harvest the larger, captured cells from the inlet by reversing the flow of pure medium (e.g., injecting medium into the outlet port). Figure 3A,B shows the two groups of cell morphologies and their sizes as measured by image analysis. The results showed that the cells captured by the filters and harvested in the second, reversed flow step were, on average, 70% larger than the smaller-sized cells that had initially escaped the filter during the initial flow step. The average cell sizes of escaped and captured groups were 28.3 and 48.3 μm, respectively (Figure 3C). This significant size difference was clearly observed in the unsorted population of the old cells (Supporting Information Figure S6), and the heterogeneity of cell size seemed to increase as cell passages accelerate (Supporting Information Figure S7), corresponding to previous studies.26,27 These size-separated cells were then cultivated for 2 days, studied morphologically under microscopy, and subsequently analyzed via senescence-associated β-galactosidase staining and lipofuscin accumulation to ascertain the level of senescence in each of these two groups. Figure 4A,B illustrates the cell morphologies for the escaped (small-sized) and captured (large-sized) cells, respectively. As illustrated, the captured cells (larger cells) had been relatively stretched, occupying a significantly large surface area per cell as compared with the escaped cells (smaller cells). Aging mammalian cells exhibit several characteristic changes: they lose the ability to divide, which, as mentioned before, is a defining feature of senescence, and furthermore demonstrate striking transformations, including changes in their morphology, mass, and the dynamics of subcellular organelles among other structural and functional alterations. Other, more specific, changes are also observed during senescence. These include an increase in intracellular oxygen free radicals which, being highly reactive, cause an increase in certain damaged products, such as lipofuscin. Another result, well documented in the literature, is an increase in senescence-associated β-galactosidase activity (SA- β-Gal).2 For this reason, analysis of senescence-associated βgalactosidase staining and lipofuscin accumulation represent standard methods to identify the extent of cellular aging. We performed such experiments on these two size-separated groups, with our main result being that the larger (captured) cells clearly demonstrated a more senescent phenotype than the smaller (escaped) cells. In particular, the escaped cells tended to show lower autofluorescence (Figure 4C), whereas the captured cells demonstrated higher fluorescence intensities
Figure 4. Comparison of the senescence phenotype of small-sized versus large-sized aged human fibroblasts. (A, B) Images of escaped and captured cells that were cultured for 2 days after separation. The two groups showed different morphologies with the captured cells showing a broader, more spread-out morphology. (C, D) Comparison of lipofuscin accumulation in escaped versus captured cells as measured by autofluorescence images. The large-sized group showed higher fluorescence intensities, corresponding to greater lipofuscin accumulation, than the small-sized group. (E, F) Senescence associated β-gal staining images for escaped and captured cells. The captured group exhibited over 80% of the population being stained, confirming that the more senescent population was, indeed, being preferentially trapped within the filter. All scale bars: 100 μm.
(Figure 4D). Moreover, when analyzed by the senescenceassociated β-galactosidase assay, the smaller-sized cell group stained less than 55% of its population (Figure 4E), but the larger-sized, captured cohort was clearly stained in over 80% of that cell population (Figure 4F). Even though it is assumed that sequential subculture should result in a homogeneous collection of cells, these experiments suggest otherwise. Namely, senescent cells, as prepared by D
DOI: 10.1021/acs.analchem.5b00445 Anal. Chem. XXXX, XXX, XXX−XXX
Technical Note
Analytical Chemistry
(3) Smith, J. R.; Pereira-Smith, O. M.; Schneider, E. L. Proc. Natl. Acad. Sci. U. S. A. 1978, 75, 1353−1356. (4) Hayflick, L. J. Invest. Dermatol. 1979, 73, 8−14. (5) Schneider, E. L. J. Invest. Dermatol. 1979, 73, 15−18. (6) Reenstra, W. R.; Yaar, M.; Gilchrest, B. A. Exp. Cell Res. 1993, 209, 118−122. (7) Yaar, M.; Eller, M. S. Arch. Dermatol. 2002, 138, 1429−1432. (8) Driskell, R. R.; Watt, F. M. Trends Cell Biol. 2015, 25, 92−99. (9) Rodier, F.; Campisi, J. J. Cell Biol. 2011, 192, 547−556. (10) Li, P.; Mao, Z.; Peng, Z.; Zhou, L.; Chen, Y.; Huang, P. H.; Truica, C. I.; Drabick, J. J.; El-Deiry, W. S.; Dao, M.; Suresh, S.; Huang, T. J. Proc. Natl. Acad. Sci. U. S. A. 2015, 112, 4970−4975. (11) Plouffe, B. D.; Murthy, S. K. Anal. Chem. 2014, 86, 11481− 11488. (12) Su, Y. H.; Warren, C. A.; Guerrant, R. L.; Swami, N. S. Anal. Chem. 2014, 86, 10855−10863. (13) Choi, S.; Song, S.; Choi, C.; Park, J. K. Anal. Chem. 2009, 81, 1964−1968. (14) Squires, T. M.; Quake, S. R. Rev. Mod. Phys. 2005, 77, 977− 1026. (15) Kim, M. S.; Kim, T.; Kong, S. Y.; Kwon, S.; Bae, C. Y.; Choi, J.; Kim, C. H.; Lee, E. S.; Park, J. K. PLoS One 2010, 5, e10441. (16) Kim, M. S.; Kim, J.; Lee, W.; Cho, S. J.; Oh, J. M.; Lee, J. Y.; Baek, S.; Kim, Y. J.; Sim, T. S.; Lee, H. J.; Jung, G. E.; Kim, S. I.; Park, J. M.; Oh, J. H.; Gurel, O.; Lee, S. S.; Lee, J. G. Small 2013, 9, 3103− 3110. (17) Kim, M. S.; Sim, T. S.; Kim, Y. J.; Kim, S. S.; Jeong, H.; Park, J. M.; Moon, H. S.; Kim, S. I.; Gurel, O.; Lee, S. S.; Lee, J. G.; Park, J. C. Lab Chip 2012, 12, 2874−2880. (18) Park, J. M.; Kim, M. S.; Moon, H. S.; Yoo, C. E.; Park, D.; Kim, Y. J.; Han, K. Y.; Lee, J. Y.; Oh, J. H.; Kim, S. S.; Park, W. Y.; Lee, W. Y.; Huh, N. Anal. Chem. 2014, 86, 3735−3742. (19) Oyobiki, R.; Kato, T.; Katayama, M.; Sugitani, A.; Watanabe, T.; Einaga, Y.; Matsumoto, Y.; Horisawa, K.; Doi, N. Anal. Chem. 2014, 86, 9570−9575. (20) Saias, L.; Autebert, J.; Malaquin, L.; Viovy, J. L. Lab Chip 2011, 11, 822−832. (21) Kang, H. T.; Lee, K. B.; Kim, S. Y.; Choi, H. R.; Park, S. C. PLoS One 2011, 6, e23367. (22) Terman, A.; Dalen, H.; Eaton, J. W.; Neuzil, J.; Brunk, U. T. Ann. N. Y. Acad. Sci. 2004, 1019, 70−77. (23) Dimri, G. P.; Lee, X.; Basile, G.; Acosta, M.; Scott, G.; Roskelley, C.; Medrano, E. E.; Linskens, M.; Rubelj, I.; Pereira-Smith, O.; et al. Proc. Natl. Acad. Sci. U. S. A. 1995, 92, 9363−9367. (24) Park, C. H.; Lee, M. J.; Ahn, J.; Kim, S.; Kim, H. H.; Kim, K. H.; Eun, H. C.; Chung, J. H. J. Invest. Dermatol. 2004, 123, 1012−1019. (25) Terman, A.; Brunk, U. T. APMIS 1998, 106, 265−276. (26) Chen, Q. M.; Tu, V. C.; Catania, J.; Burton, M.; Toussaint, O.; Dilley, T. J. Cell Sci. 2000, 113 (Pt 22), 4087−4097. (27) Wang, E.; Gundersen, D. Exp. Cell Res. 1984, 154, 191−202.
multipassaged subculture techniques, are actually heterogeneous in aging phenotype. We have moreover shown that this phenotypic senescence heterogeneity seems to be related to cell size, in which larger aged cells, as isolated by our novel uniform flow velocity microfluidic filter, express more senescent characteristics as determined by conventional cell aging assays. With these results, we suggest the existence of two varieties of senescent cells: the smaller-sized cells, which express a younger phenotype and larger-sized cells, which appear to have a more aged phenotype. We propose to call these smaller cells young phenotypic aged cells (YPACs) and the larger cells old phenotypic aged cells (OPACs).
■
CONCLUSIONS We have designed and fabricated a novel microfilter with a uniform flow velocity profile enabling the efficient size-based separation of senescent cells. Our results further suggest that experimental studies on senescent cell cultures may benefit from adding a purification step to isolate these two YPAC and OPAC subtypes. Homogeneity of the object of study is critical for obtaining accurate and robust results, and we would suggest, then, that the large-sized OPACs would represent a suitable cell preparation for the study of cellular aging. In addition, comparative studies of YPACs and OPACs may also yield fruitful insights into the biology of aging. In this way, we believe that applying these microfiltration techniques to the preparation of aged cells may offer a significant advantage toward the study of the cellular mechanisms of aging and rejuvenation.
■
ASSOCIATED CONTENT
S Supporting Information *
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.5b00445. Additional fabrication and analytical details; figures showing cartridge setup, simulation results, and cell images (PDF)
■
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Author Contributions
M.S.K. conceived and designed the experiments. M.S.K., S.J., J.T.P., and H.Y.S. performed the experiments. M.S.K. and S.S.K. conducted and analyzed the simulations. M.S.K. and S.J. analyzed the data. M.S.K., O.G., and S.C.P. wrote and edited the paper. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS We acknowledge Dr. Hyun Tae Kang and Dr. Young Jun Koh for helpful discussions. This research was supported by the Well Aging Research Center at the Samsung Advanced Institute of Technology and the Next-Generation BioGreen 21 Program (PJ01116401) from Rural Development Administration, Republic of Korea.
■
REFERENCES
(1) Hayflick, L. Exp. Cell Res. 1965, 37, 614−636. (2) Hwang, E. S.; Yoon, G.; Kang, H. T. Cell. Mol. Life Sci. 2009, 66, 2503−2524. E
DOI: 10.1021/acs.analchem.5b00445 Anal. Chem. XXXX, XXX, XXX−XXX
