Author contributions: E.A.S.: conception and design, collection and/or assembly of data, data analysis and interpretation, development of mathematical model, and manuscript writing; B.A.H.: collection and/or assembly of data and data analysis and interpretation; A.M.M.: conception and design; J.L.: collection and/or assembly of data; S.C.M. and K.R.S.: development of mathematical model; R.C.R. and P.J.H.: conception and design, financial support, data analysis and interpretation, and manuscript writing.
Disclosure of potential conflicts of interest is found at the end of this article.
First published online in STEM CELLSEXPRESS September 10, 2011.
Although new neurons are produced in the subventricular zone (SVZ) of the adult mammalian brain, fewer functional neurons are produced with increasing age. The age-related decline in neurogenesis has been attributed to a decreased pool of neural progenitor cells (NPCs), an increased rate of cell death, and an inability to undergo neuronal differentiation and develop functional synapses. The time between mitotic events has also been hypothesized to increase with age, but this has not been directly investigated. Studying primary-cultured NPCs from the young adult and aged mouse forebrain, we observe that fewer aged cells are dividing at a given time; however, the mitotic cells in aged cultures divide more frequently than mitotic cells in young cultures during a 48-hour period of live-cell time-lapse imaging. Double-thymidine-analog labeling also demonstrates that fewer aged cells are dividing at a given time, but those that do divide are significantly more likely to re-enter the cell cycle within a day, both in vitro and in vivo. Meanwhile, we observed that cellular survival is impaired in aged cultures. Using our live-cell imaging data, we developed a mathematical model describing cell cycle kinetics to predict the growth curves of cells over time in vitro and the labeling index over time in vivo. Together, these data surprisingly suggest that progenitor cells remaining in the aged SVZ are highly proliferative. STEM CELLS 2011;29:2005–2017.
Neural progenitor cells (NPCs) retaining the capacity to produce new neurons in the adult mammalian brain reside primarily in the subventricular zone (SVZ) and the subgranular zone of hippocampal dentate gyrus . Neuron production declines in both areas during normal aging [2, 3]; this phenomenon is correlated with cognitive decline [4, 5]. However, the cellular mechanisms underlying age-related neurogenic decline are unclear. Neurogenesis is a complex, multistep process, and the documented age-related decline could be due to a decreased pool of neural stem cells, slower cell cycle progression, a lower survival rate, a deficit in migration capacity, or an inability to undergo neuronal differentiation and develop functional synapses.
Although it can be challenging to examine these aspects of neurogenesis separately, previous investigations into the cellular mechanism of neurogenic decline have uncovered fewer neural stem cells existing in aged SVZ . Of these remaining cells, a lower percentage are capable of differentiating into functional neurons; fewer aged NPCs could evoke action potentials after differentiation protocols, although those that did differentiate showed physiological characteristics indistinguishable by age . In addition, researchers have observed a survival deficit in aging NPCs under both growth and differentiation conditions [7, 8]. Much investigation of the proliferative capacity of aging NPCs in vitro has relied on gross analyses such as total cell or sphere counts that can be influenced by multiple confounding factors. Several in vivo studies, using single or sequential pulses of bromodeoxyuridine (BrdU), have concluded that aging leads to a loss of NPCs in the SVZ [6, 7]. Others have suggested that neurogenic decline may be due to NPCs undergoing quiescence or a lengthening of the cell cycle, perhaps due to increased tumor suppressor expression [8, 9] or decreased growth factor responsiveness [10, 11]. However, cell cycle kinetics have not been empirically determined in aging NPCs.
Thymidine-analog markers of S-phase can be used to determine the number of cells dividing at a given time (the mitotic index). The number of cells in S-phase at a given time gives no information about the rate of cell cycle transit or re-entry. However, several of these markers can be used in concert to calculate the time between successive S-phases. This measure is affected by the amount of time needed to transit through the cell cycle, as well as the latency and likelihood for NPCs to re-enter the cell cycle, all factors which could affect net proliferative activity. In this study, we quantified the time between successive S-phase labels and cytokinetic events in young adult and aged adult NPCs. Surprisingly, aged cultures appear to contain both a highly quiescent population and a highly proliferative population; in contrast, many young NPCs divide sporadically. This study demonstrates that, although a fewer number of aged cells are cycling at a given time, the actively cycling NPCs remaining in the aged mouse forebrain undergo more cell divisions in a given period of time than those in the young adult forebrain.
MATERIALS AND METHODS
All experiments were performed as approved by the University of Washington Institutional Animal Care and Use Committee. Female C57BL/6 mice were housed at 21°C with access to food and water ad libitum. Adult NPCs were isolated as previously described . Briefly, wild-type C57BL/6 mice, 3 months and 18 months of age, were overdosed with Beuthanasia and transcardially perfused with ice-cold saline. Brain tissue, not including olfactory bulbs or cerebellum, was mechanically and enzymatically dissociated with collagenase-DNase solution. To remove debris, myelin and red blood cells, the cell suspension was mixed with a percoll solution and centrifuged. The isolated progenitor cells were grown in proliferation media, consisting of Dulbecco's modified Eagle's medium/F12 supplemented with 2 mM glutamine, 1% N2 (Gibco, Carlsbad, CA, www.invitrogen.com/site/us/en/home/brands/Gibco.html), 50 μg/ml heparin (Sigma, St Louis, MO, www.sigmaaldrich.com/united-states.html), 20 ng/ml epidermal growth factor (Peprotech, Rocky Hill, NJ, www.peprotech.com), and 20 ng/ml fibroblast growth factor-2 (Peprotech). Cultures were passaged by mechanical dissociation, and used for in vitro experimentation between passages 3 and 15. All in vitro experiments were performed at least three times; replications included at least two independent cell isolates of each age.
Growth curves were extrapolated from passage rate and viable cell counts at passage. Forebrain-derived cultures initially formed neurospheres in vitro and were capable of continued passage as spheres or a monolayer. Cell number and viability was quantified at each passage using the ViCell automated cell counting system (Beckham Coulter, Indianapolis, IN, https://www.beckmancoulter.com). Cells were plated at 106/10 cm tissue culture dish. Four days after passage, cells were trypsinized and counted. To quantify the mitotic index and live-dead cell count of young and aged NPCs, live cells were collected and incubated at 37°C with Hoechst (0.5 mg/ml) for 30 minutes then incubated at 37°C with propidium iodide (1 μg/ml). Cells were then subjected to sorting and analysis with FACSdiva (BD Biosciences, Franklin Lakes, NJ, www.bdbiosciences.com /instruments/software/facsdiva/index. jsp). Cellular debris and doublets were sorted out by side scatter analysis, and propidium iodide–positive dead cells were quantified and gated out. Remaining cells were plotted by Hoechst width to assay mitotic phase ratios. Cell fractions within G1, S, and G2/M phases were binned and compared with two-tailed t tests in Excel. Time-lapse live-cell imaging was performed using a Nikon TiE inverted widefield fluorescence microscope (nikoninstruments.com/Information-Center/Perfect-Focus-System-PFS), with an environmental chamber for temperature and CO2 control, attached to an EMCCD camera. Cells were first infected with a lentiviral construct expressing green fluorescent protein (GFP) under a constitutive promoter, which was produced in accordance with NIH guidelines for recombinant DNA. Labeled cells were plated at low density with uninfected, age-matched cells (1:100) on poly-L-lysine-coated 60-mm dishes and were photomicrographed every 15 minutes for 48 hours at ×30 under phase and GFP using NIS Elements software (Nikon Instruments, Melville, NY, www.nis-elements.com). Time-lapse live-cell imaging data were analyzed using Fisher's exact test.
To characterize markers of progenitor cell phenotype, NPCs were plated in 24-well plates at a density of 10,000 cells per well on laminin- and poly-L-lysine-coated glass coverslips for 4 days in proliferation media. Cells were then fixed in 4% paraformaldehyde at room temperature for 5 minutes, rinsed three times with phosphate-buffered saline (PBS), and blocked for 1 hour in PBS with 0.08% Triton X-100 and 5% donkey serum. Cells were then labeled with anti-Nestin mouse monoclonal antibody (Chemicon MAB353, 1:1,000, www.millipore.com), anti-CD133 mouse monoclonal antibody (14-1331-82, 1:333, eBioscience, www. ebioscience.com), anti–SRY box 2 (anti-Sox2) goat polyclonal antibody (SC17320, 1:250, Santa Cruz, www.scbt.com) and anti-KI67 rabbit polyclonal antibody (NCL-Ki67p, 1:500, Novocastra, www.leica-microsystems.com/products/total-histology/novocastra-reagents). Terminal deoxynucleotidyl transferase dUTP nick end label–positive (TUNEL+) apoptotic cells were quantified using TdT Reagent Kit (Chemicon S7160). The following secondary antibodies were diluted 1:2 in 50% glycerol, then 1:250 in PBS with 0.08% Triton X-100 and 5% donkey serum: Jackson Labs (www.jacksonimmuno.com) Cy2-conjugated donkey anti-rat, RedX-conjugated donkey anti-mouse, and Cy2-conjugated donkey anti-rabbit. To quantify the number and rate of cycling cells, we used the antigenically distinct thymidine analogs chlorodeoxyuridine (CldU) (Sigma C6891-100 mg) and iododeoxyuridine (IdU) (Sigma I7125-5G). Cells were plated on coated coverslips as previously, and exposed to CldU (4.6 μg/ml) and uridine (1 mg/ml) for 30 minutes to label dividing cells . Although exposure to thymidine analogs has been reported to have cytotoxic effects on mammalian cells, no deleterious effects have been observed at these concentrations. At 12-hour, 15-hour, 18-hour, 21-hour, or 24-hour after removal of CldU, cells were treated with IdU (7.2 μg/ml) for 30 minutes then immediately fixed with 4% paraformaldehyde. Anti-CldU rat monoclonal antibody clone BU1/75 (1:250, Novus, www.novusbio.com) and anti-IdU mouse monoclonal antibody clone B44 (BD Biosciences 347-580, 1:250) were applied sequentially for 90 minutes at 37°C, followed by incubation with appropriate secondary antibodies for 60 minutes at room temperature. All cells were costained with 4′,6-diamidino-2-phenylindole (DAPI) (Sigma D9542). Fluorescence microscopy was performed using a Zeiss Axioskop two with attached Optronics camera and StereoInvestigator software (MBF Biosciences, Williston, VT, www.mbfbioscience.com/stereo-investigator). Fractions of labeled cells were compared using a two-tailed t test in Excel.
Quantification of Dividing Cells In Vivo
To quantify NPCs in the young adult and aged SVZ, mice aged 3 months (n = 8) and 20 months (n = 8) were injected with BrdU (50 mg/kg) once daily for 12 days. The animals were divided into two groups, and either euthanized immediately following the final injection or 28 days after the final injection. To quantify cell cycle re-entry in the young adult and aged SVZ, mice aged 3 months (n = 6) and 18 months (n = 6) were injected with a single pulse of CldU (50 mg/kg), then with three pulses of IdU (50 mg/kg) 16 hours, 18 hours, and 20 hours later. Animals were euthanized with 0.04 ml Beuthanasia, then transcardially perfused with ice-cold saline followed by 4% paraformaldehyde. Brains were removed and serially sectioned into 20-μm slices. Tissue sections were stained on glass slides after being subjected to antigen capture using 0.01 M sodium citrate and 2 N HCl. BrdU+ mitotic cells and BrdU-retaining cells were labeled with anti-BrdU rat monoclonal antibody (Novus MB500169, 1:200), anti-Nestin mouse monoclonal antibody (Chemicon MAB353, 1:250), anti-Sox2 goat polyclonal antibody (Santa Cruz SC17320, 1:250), anti–T-box brain 2 (anti-Tbr2) rabbit polyclonal antibody (gifted from Robert Hevner's laboratory) and anti-glial fibrillary acidic protein (anti-GFAP) rabbit polyclonal antibody (Z0334, 1:300, Dako, www.dako.com). Double-thymidine-analog labeling was performed by sequentially applying anti-CldU rat monoclonal antibody clone BU1/75 (Novus, 1:250) overnight at 4°C and anti-IdU mouse monoclonal antibody clone B44 (BD Biosciences 347-580, 1:250) for 2 hours at 37°C, followed by incubation with appropriate secondary antibodies for 2 hours at room temperature. Cells were counted in 100 × 200 μm grids on NIS Elements software in 10-20 optical sections. Three-dimensional reconstruction of z-stack images was performed with Volocity software (Perkin Elmer, Waltham, MA, www.perkinelmer.com/pages/020/cellularimaging/products/volocity.xhtml). Counts were adjusted for total SVZ area. Fractions of labeled cells were compared using a two-tailed t test in Excel.
To calculate cell cycle transit time using a cumulative BrdU labeling protocol, animals were injected with BrdU (50 mg/kg) once every 3 hours for 18 hours. A cohort of animals (n = 4 for each age group at each time point) was sacrificed 1 hour after each BrdU injection. Perfusion, BrdU labeling, and cell quantification were performed as described above. The total number of BrdU+ cells in the SVZ of each animal was plotted, and regression lines were fitted to the points [14, 15]. The x value at which BrdU labeling reaches a plateau is Tc − Ts, the time required to transit through the cell cycle subtracted by the time required for transit through S-phase. The y value at this point is referred to as GF, the total number of proliferating cells in the SVZ. The y intercept of the curve, denoting the number of cells labeled at the first time point, is equal to (Ts/Tc) × GF. Tc and Ts are calculated based on a curve fit solved for a minimum sum of squares (SS) . After calculating the SS and degrees of freedom for each data set compared to a model (the curve fit to the control group), the data sets were subjected to nonlinear regression analysis and were compared using an F test.
Mitotic Index in the Aging Brain
Previous investigators have reported a dramatic difference in the BrdU labeling index in young and aged SVZ using a single-day BrdU pulsing protocol [8, 11]. Yet the accumulated number of cells in S-phase over an extended period of time has not been determined. To calculate the number of BrdU+ cells in SVZ after an extended labeling protocol, young (3 months, Fig. 1A) and aged (20 months, Fig. 1B) mice were intraperitoneally injected with 50 mg/kg BrdU for 12 days, then sacrificed. We observed a 21% fewer BrdU+ cells in the aged SVZ, compared with the young adult SVZ (p < .05, Fig. 1C). After this extended labeling protocol, we observed no difference in the fraction of SOX2+ cells that incorporated BrdU (p > .05, Fig. 1D). BrdU+ mitotic cells (in green), colabeled with the neural stem cell marker SOX2 (in red), are observed in the dorsolateral SVZ of young (Fig. 1E) and aged adult mice (Fig. 1F), and along the lateral ventricle of young (Fig. 1G) and aged adult mice (Fig. 1H). BrdU+ cells in the young (I, K, and M) and aged adult (J, L, and N) SVZ colabel with Nestin (I-J), Tbr2 (K-L), and GFAP (M-N). We observed significantly more BrdU+ cells in the aged SVZ colabeled with Sox2 (p < .01) and significantly fewer BrdU+ cells in the aged SVZ colabeled with Nestin (p < .001); we observed no differences in fractions of cells colabeled with Tbr2 or GFAP (O).
We injected a second group of mice with 50 mg/kg BrdU intraperitoneally for 12 days, then sacrificed them 28 days afterward. We then quantified label-retaining cells, the population of slowly dividing cells , in the young (Supporting Information Fig. 1A) and aged (Supporting Information Fig. 1B) SVZ. We observed a 45% decrease in the number of label-retaining cells in the SVZ with age (p < .05, Supporting Information Fig. 1C). We also quantified GFAP+ cells in the young (Supporting Information Fig. 1A) and aged (Supporting Information Fig. 1B) SVZ. Although we observed greater numbers of GFAP+ cells with an astrocytic morphology in the aged brain (p < .001, Fig. 1N), there were no significant differences in the fraction of GFAP+-labeled cells in the population of BrdU+ cells or BrdU+ label-retaining cells (p > .05, Supporting Information Fig. 1D). The 45% decrease in BrdU+ label-retaining cells suggests a decrease in the population of slowly dividing stem-like cells in the aged brain. To more closely investigate young and aged NPCs, we carried out in vitro assays on cultured cells.
Growth of Aging NPCs in Culture
Once forebrain-derived cultures entered logarithmic growth phase and could be actively passaged, we characterized their immunophenotypes and proliferative activity. Both young (Fig. 2A-2C) and aged (Fig. 2F-2H) NPCs expressed uniformly high levels of the stem cell markers Nestin (Fig. 2A, 2F), CD133 (Fig. 2B, 2G), and SOX2 (Fig. 2C, 2H), demonstrating that young and aged NPCs are antigenically similar (p > .05, Fig. 2K-2M). However, a greater percentage of young NPCs (Fig. 2D) than aged NPCs (Fig. 2I) stained positive for KI67, a marker of active proliferation (p < .05, Fig. 2N). We further immunophenotyped the KI67+ cells, and found that a greater fraction of aged KI67+ cells were SOX2+ (p < .05, Fig. 2Q), although there were no significant age-related differences in colabeling of KI67 with GFAP or doublecortin (DCX) (p > .05, Fig. 2P, 2R). Colabeling in the KI67+ population adds up to more than 100% because there is likely to be overlap between the KI67+ populations that are SOX2+ and those that are DCX+.
Next, we quantified the viability of young (Fig. 2E) and aged (Fig. 2J) NPCs. Aged cells displayed a significant increase in TUNEL+ staining, an indication of apoptotic activity (p < .01, Fig. 2O). Using the ViCell automated cell quantification system, we also observed a significant increase in trypan blue+ nonviable cells in aged cultures (p < .01, Fig. 2S). To further investigate the mitotic index of these cells, we quantified the number of young and aged NPCs in each phase of the cell cycle using Fluorescence-activated cell sorting (FACS; Fig. 2T, 2U). Live cells were labeled with 2 μg/ml Hoechst; cellular doublets and debris were sorted out by side scatter analysis and propidium iodide exclusion. Remaining cells were binned to assay mitotic phase ratios. A significantly higher percentage of 3-month cells were observed in G2/M phase of the cell cycle, compared with 18-month cells (31.0% vs. 12.4%, p < .0001). Together, these data support the notion that fewer aged adult cells are undergoing mitosis at a given time, and that these cells are less viable than young adult cells. We hypothesized that progression through the cell cycle might be altered in aged cells as well.
Live-Cell Imaging Analysis
To empirically quantify the time between mitotic events, we took time-lapse images of live cells. To better visualize individual cells, we labeled young and aged NPCs with a lentivirus containing GFP under a constitutive promoter, and plated them at 1:100 on a layer of unlabeled cells of the same age. By taking photomicrographs of young and aged NPCs every 15 minutes for 48 hours, we could track whether a single cell divided, then whether daughter cells divided additional times (Fig. 3A, Supporting Information Videos 1-4). We categorized these cells, separating the fractions of nondividers (no mitoses in 48 hours), rare dividers (one mitosis in 48 hours), and prolific dividers (more than one mitosis in 48 hours). We observed that significantly more aged cells had no mitotic events, compared with young cells (67% vs. 22%, p < .001). However, comparing only the dividing cells, a greater proportion of aged NPCs than young NPCs were prolific dividers (56% vs. 29%, p < .05). A 79% decrease was observed in the rarely dividing population with age (p < .0001), while a nonsignificant decrease was observed in the prolifically dividing population with age (p > .05). The average number of mitotic events for the entire population was 1.07 for young cells and 0.57 for aged cells, but among the mitotic population, the average number was 1.37 events for young cells and 1.70 events for aged cells. In addition, we measured the time between mitotic events in each culture, by subtracting the timestamps on frames from consecutive cytokineses. A great deal of variability was observed in times between cell divisions; aged cells exhibited an average time between mitoses of 16.8 hours, while young cells that divided more than once exhibited an average time between mitoses of 20.7 hours (p > .05, Fig. 3B). Similar to results with TUNEL labeling and ViCell quantification methods (Fig. 2O, 2S), aged cultures had a significantly greater incidence of cell death under live-cell imaging (p < .001, Fig. 3C). Of only dividing cells, aged cultures displayed significantly higher rates of cell death (p < .05, Fig. 3C). Significantly fewer aged cells also migrated out of the field of view during the 48-hour period of live-cell imaging (p < .01, Fig. 3D). Of only dividing cells, however, aged cultures did not have significantly different rates of migration (p > .05, Fig. 3D). These data provide direct evidence for a rather surprising result: specifically, that aged NPC cultures have fewer actively cycling cells than do young NPC cultures, but aged cells that are cycling undergo a significantly greater number of divisions.
To quantify re-entry into the cell cycle, we sequentially labeled young and aged NPCs with two antigenically distinct thymidine analogs (Fig. 4A). All mitotic cells were labeled by 30-minute CldU pulse at time 0. After rinsing and further incubation, cells were labeled with a 30-minute IdU pulse and immediately fixed, at 3-hour time points throughout the following day. Cells were not presynchronized, but a population of coincident cells labeled by CldU in an initial S-phase was investigated for re-entry into cell cycle at multiple independent time points by colabeling with IdU. The fraction of IdU+ cells of total DAPI+ cells was significantly higher in young cultures (Fig. 4B) than in aged cultures (Fig. 4C) after a 30-minute pulse in vitro, agreeing with previous results that fewer cells are dividing at a given time in aged cultures (p < .0001, Fig. 4D). However, a significantly larger fraction of CldU+ aged cells colabeled with IdU at the 12-, 15-, 18-, and 21-hour time points (12-hour, p < .01; 15-hour, p < .0001; 18-hour, p < .0001; 21-hour, p < .001; 24-hour, p > .05, Fig. 4E). These data suggest that aged cells have an increased tendency to re-enter the cell cycle within a day of a previous S-phase labeling.
To test whether this effect was also observed in vivo, we treated young and aged mice with a single intraperitoneal injection of 50 mg/kg CldU, followed by intraperitoneal injections of 50 mg/kg IdU at 16 hours, 18 hours, and 20 hours post-CldU (Fig. 5A). This protocol identifies cells rapidly re-entering cell cycle, by covering the peak time between successive mitoses observed in young and aged NPCs in vitro. We observed double-labeling in 18.2% of the CldU+ cells in the young adult SVZ (Fig. 5B) and in 34.7% of the CldU+ cells in the aged adult SVZ (Fig. 5C) after this protocol (p < .001, Fig. 5D). The number of IdU+ cells yields a quantification of the number of cells dividing at a given time. In vivo, we observed significantly fewer IdU+ cells (p < .00001, Fig. 5E) and significantly fewer CldU+ cells (p < .0001, Fig. 5F) in the aged SVZ. Despite the many complex processes that govern cell genesis in the SVZ, these data confirm that aged cells are more likely to re-enter cell cycle than young adult cells over a 22-hour period in vivo.
Cumulative BrdU Labeling to Calculate Cell Cycle Transit Time
To calculate the time required for cells to transit through S-phase and the entire cell cycle, we performed cumulative BrdU labeling according to a method developed by Nowakowski and coworkers [14–16]. Young and aged mice were injected with BrdU once every 3 hours; 1-hour after each injection, several animals in each age group were sacrificed. A schematic depicting this protocol is shown (Fig. 6A). Representative pictures of the young (Fig. 6B) and aged (Fig. 6C) SVZ after 1, 4, and 7 BrdU injections are shown. BrdU+- labeled cells are plotted (Fig. 6D). Cell cycle transit times were calculated according to two equations: Tc − Ts, the x value at which labeling reached a plateau and Ts/Tc × GF, the y value at time 0 (3,360 cells in the young SVZ and 1,516 in the aged SVZ). GF is equivalent to the number of BrdU+ cells at the plateau (5,602 cells in the young SVZ and 2,214 cells in the aged SVZ). From these equations, we calculate that young cells take 9.8 hours to transit through the entire cell cycle, with 4.9 hours for S-phase, while aged cells take 7.6 hours to transit through the entire cell cycle, with 4.2 hours for S-phase. These curves are significantly different (p < .001), unless each data point is normalized to its respective GF, suggesting that the total number of cycling cells is significantly different between groups, but cell cycle length is not.
Mathematical Modeling of Cell Cycle Kinetics
To verify that the cell expansion observed in vitro is consistent with our time-lapse live-cell imaging data and double-thymidine-analog labeling data, we developed a simple mathematical model applying our empirical data regarding age-related differences in cell cycle kinetics (Supporting Information Table 1). The number of cells counted in culture (the first observation), is the starting population (N(0)); this population was subjected to an iterative equation describing cellular activity over a 48-hour period. The equations used to model the cell counts over time are described in Figure 7; N(t) equals the number of cells present in the previous time step and N(t + 2) equals the number of cells present in the current step (after 48 hours). The original cell count (N(0)) is cycled through this equation describing cell cycle kinetics empirically determined over a 48-hour period under live-cell imaging; the computed result (N(t + 2)) is then set as the next iteration's N(t). The predictions over time are plotted against actual cell counts derived from samples quantified every 4 days from young and aged cultures (Fig. 7).
At each iteration, the previous cell population (N(t)) is added to the newly divided cells, quantified by the fraction of cells entering cell cycle in a 48-hour period of live-cell imaging (N(t) × E, shown in Fig. 3A). All cells are subject to death, so this number was multiplied by the fraction of surviving cells (S) observed after 48 hours under live-cell imaging (Fig. 3C). This simple model, predicting a growth curve based on cell cycle entry and survival, is depicted in Figure 7A. The regression values describing the fit between predicted results and values observed by counting cells at passage are r2 = .964 for young cells and r2 = .767 for aged cells. Next, we adapted this simple model to take into account cell cycle re-entry: the fraction of cells entering cell cycle was multiplied by the number of new cells produced by the dividing cells (the fraction of cells dividing once create two cells each (two × D1), the fraction of cells dividing twice create three cells each (3 × D2), and so on, as quantified during 48 hours of time-lapse live-cell imaging (Fig. 3A). This adapted model is depicted in Figure 7B. The regression values describing the fit between predicted results using this model and values observed by counting cells at passage are r2 = .998 for young cells and r2 = .940 for aged cells. This model is therefore superior to the previous model, which took into account only the number of actively cycling cells and survival, so we conclude that cell cycle re-entry is an important factor in the net proliferative activity of a population of NPCs.
We then modeled the thymidine-analog labeling index over time in vivo, using the number of CldU+ cells after a single pulse (shown in Fig. 5F), as the starting index (L(0)). This population of actively dividing cells was subjected to an adapted iterative equation to generate a model-derived final index (L (12) computed), which we fit to the actual number of BrdU+ cells after 12 thymidine-analog pulses (L (12) in vivo), to obtain predicted values of N, the total number of cells in the SVZ (described in Supporting Information Methods and Supporting Information Table 2). These predicted values of N are 6,334 in the young adult brain and 4,791 in aged adult brain. The model-predicted labeling indices are compared in Supporting Information Figure 2 to actual BrdU+ cell counts in the young adult and aged adult SVZ after an extended 12-day labeling protocol.
Although molecular mechanisms regulating proliferation, including cell cycle regulatory proteins and niche-derived factors, have been manipulated to recover neurogenic activity in aged NPCs [8–11, 17, 18], an actual slowing of cell cycle progression in the normal aged brain has not been demonstrated. Decreased number of NPCs, decreased capacity for differentiation, and decreased cell survival have been shown to underlie age-related neurogenic decline in the SVZ [6–8]. Several studies additionally suggest that a proliferative deficit may be responsible for neurogenic decline in vivo [7, 10], although this phenomenon has been difficult to empirically separate from changes in cell number or survival rate . In testing these hypotheses, it has proved notably difficult to empirically separate a proliferative index (the number of cells dividing at a given time) from the proliferative activity (specifically, the time between one cell division and the next), especially considering the problem of heterogenous populations of NPCs. Although a fewer number of aged NPCs are dividing at a given time, according to FACS analysis, KI67 labeling, and BrdU labeling, we show in the present study that aged cells that are cycling undergo a greater number of cell divisions. It is not clear whether the rarely dividing population is selectively lost, or whether cell cycle progression is sped up in individual remaining cells. Because of the limitations of cell fate identification, it cannot be definitively stated whether the sizes of stem and progenitor populations are affected differentially in vivo. Cultures used in this study were derived from whole forebrain tissue, which increases the heterogeneity of stem and progenitor cells present. However, NPCs from the SVZ are thought to be a major contributor to these cultures. The short time frames of each experiment were designed to target primarily the NPC population, and the experiments performed in vivo were designed to test the activity of SVZ progenitors specifically.
Using time-lapse live-cell imaging (Fig. 3), we found that a lower fraction of the aged culture undergoes cell division, but aged cells that do divide are highly proliferative dividers. Compared with young adult cultures, aged cultures contain fewer rare-divider cells, which divide one time in 48 hours (55% vs. 15%, p < .0001), and fewer cells that exhibit more than 20 hours between cell divisions (52% vs. 19%, p = .0516). It is not clear whether this slowly dividing population is lost or whether aged cells on average speed up cell cycle progression, since the result in either case would be more cell divisions on average in the population. In support of the former view, the number of slowly dividing cells that retain BrdU+ labeling after 28 days are significantly reduced in the aged animals (Supporting Information Fig. 1). In support of the latter view, the time between successive cell divisions observed by time-lapse live-cell imaging or cumulative BrdU labeling is ∼20% lower in aged cultures, although these results are not significant, due to high variability. Two populations can be distinguished in Figure 3B: cells with an interval greater than 20 hours between cytokinetic events, and cells with less than 20 hours between cytokinetic events. Calculations of cell cycle length by cumulative BrdU labeling in vivo are shorter than those calculated by time-lapse imaging (9.8 hours in young cells and 7.6 hours in aged cells). The BrdU+ cells under investigation in the accumulation experiment are primarily quickly dividing progenitors, labeled within an 18-hour timeframe, while some of the cells observed under time-lapse imaging may have been stem-like, with more than 20 hours between cytokinetic events. Heterogeneity of progenitor populations is a source of variability that decreases the statistical power of cell cycle length comparisons.
Double-thymidine-analog-labeling experiments (Figs. 4, 5) demonstrate that the proliferative deficit observed in aged animals is due to a decreased number of dividing cells, not slower cell cycle transit or less cell cycle re-entry. Fewer aged NPCs were observed in S-phase after a 30-minute thymidine-analog pulse in vitro and in vivo, in agreement with our data showing less KI67+ labeling and a lower FACS mitotic profile in aged cells. The time between S-phase peaks was not significantly different between cultures in this experiment. Interestingly, however, higher numbers of aged cells were observed with double-labeled nuclei at time points between 12 hours and 21 hours in vitro (Fig. 4E) and between 16 hours and 20 hours in vivo (Fig. 5D). The high double-labeling index could be interpreted as aged cells being blocked in S-phase of the cell cycle, but since full cytokinetic events were observed under live-cell imaging, this possibility seems unlikely. Therefore, we conclude that aged NPCs which are actively cycling are more likely to re-enter S-phase of cell cycle within a day after completing a round of cell division.
Significant age-related changes in cell cycle re-entry are demonstrated directly in Figures 4 and 5, using double-thymidine-analog labeling, and indirectly in Figures 3A and 7A and 7B, using live-cell time-lapse imaging and mathematical modeling. In Figures 3A and 7A and 7B, changes in cell cycle length and cell cycle re-entry are indistinguishable, while cell cycle re-entry is not quantified in Figure 6 at all. Together, these results suggest that aged NPCs do not have a significantly different cell cycle length compared with young NPCs, but are more likely to re-enter into cell cycle. Perhaps, however, increased cell cycle re-entry is caused by an abbreviated G1 phase—in other words, these two phenomena (cell cycle length and cell cycle re-entry) may be effectively inseparable due to empirical limitations.
Cell cycle regulatory proteins may affect the proliferative activity of NPCs across the lifespan. We demonstrated previously that the remaining population of NPCs in the aged forebrain have lower expression and inducibility of the tumor suppressor protein p53 . Impairment in this cell cycle regulatory pathway can affect the rate of cell division . However, increased p16 expression has also been reported with age [8, 19], and p16 null mice have higher rates of new neuron production in the aged SVZ, compared with wild-type mice . However, there is evidence that p16 has differential roles in cell cycle progression depending on cellular context and age; for example, altered p16 expression in hematopoetic stem cells does not cause replicative senescence [20, 21]. The net effect of the bidirectional changes in tumor suppressor protein expression documented in aging NPCs has not been thoroughly investigated. However, intriguing correlations have been shown between a shortened G1 phase of the cell cycle and an expansion of the progenitor population at the expense of neuronal differentiation, during development [22, 23], in adulthood , or after injury of the adult brain . Multiple mechanisms could be responsible for age-related changes in NPC population size and cell cycle progression, with the balance between senescence and regeneration achieved by compensation on a molecular scale. Cellular selection within the neurogenic niche may also affect the observed progenitor phenotype. Recently, a loss of morphologically identifiable proliferative subpopulations has been reported in the aging hippocampus ; it is not known whether similar subpopulations are selectively affected in the aging SVZ.
Given single or sequential pulses of BrdU over the course of a single day, previous investigators have observed dramatically decreased numbers of mitotic cells in the aged SVZ [6–8, 10]. Our data, both in vitro (Figs. 2, 4) and in vivo (Fig. 5E, 5F), also point to a greatly decreased number of cells undergoing mitosis at a given time. However, aged cells are more likely to re-enter S-phase of the cell cycle within a day, both in vitro (Fig. 4E) and in vivo (Fig. 5D). We hypothesized that quicker cell cycle progression could partially compensate for a lower cell number in the aged brain over time. To test this hypothesis, we developed an iterative mathematical model incorporating empirically derived cell cycle kinetics to predict the regenerative activity of NPCs over time, as both the number of cycling cells and the number of cells produced in a given period of time are taken into account. Mathematical modeling based on empirically derived data has been used previously to predict cellular activity in the case of a tumor [26, 27]. These previous models used differential MRI data from individual patients to predict tumor growth by calculating values describing only proliferation and invasiveness. Similarly, we are confident that populational activity in vitro and in vivo can be described here using empirical data. The data used to build the model were derived from a single in vitro methodology, while the observed data to which the model was compared are determined by alternate methodologies. In addition, the variables incorporated into the model reflect specific cellular phenomena, while observed values are dependent upon net proliferative activity. While our model predicts the in vitro growth curves and in vivo thymidine-analog labeling quite well, some factors may not have been fully accounted for in the models. Quantifying true values for survival, migration, and total cell number in vivo may create a better fit with the observed data. It is notable that this model, based on data from live-cell imaging experiments showing more mitotic events in the actively cycling population of aged cells, predicts observations of cellular behavior both in vitro and in vivo with a high level of accuracy. Together, these models support the hypothesis that aged NPCs partially compensate for a smaller population over time, with more mitotic events per cell.
Previous authors have observed that the decrease in BrdU+ cells in the aged SVZ matches the decrease in new neurons observed in the olfactory bulb [8, 11]. These data suggest that a decreased stem cell population is primarily responsible for age-related neurogenic decline; however, survival and differentiation are also significantly impaired with age. The ability of the remaining population of dividing cells to more quickly progress into the next cell cycle may lead to an expansion of the progenitor cell pool; this population is then pared down again by impaired survival and differentiation (Supporting Information Fig. 3). While such a case may be helpful in maintaining some level of olfactory cortical repair under normal conditions, a dysregulation in cell cycle kinetics could predispose the remaining population of neural stem cells to become cancer stem cells. Alterations in cell cycle progression during normal aging may resolve a paradox wherein aging leads to both neurogenic decline [2, 6–8, 10, 11] and an increased risk for brain tumors [28–30]. Further investigation of altered cell cycle regulation in aging NPCs may yield insights into NPC dysfunction in aging and cancer.
Paradoxically, aging leads to both a decreased regenerative capability in the brain and an increased risk of brain tumors; both conditions are thought to be caused by NPC dysregulation. In this study, we found that aging NPCs have increased progression from one cell cycle to the next, allowing aged cells to undergo more divisions in a given time. This activity partly compensates for age-related deficits in population size, cell survival, and neuronal differentiation. These findings provide a cellular mechanism linking the decline in regenerative activity with an increased capacity for proliferation.
We thank Jason Barber for assistance with statistical analysis and Denise Inman for valuable discussions. E.A.S. is supported by the University of Washington Training Grant in Developmental Biology (HDO7183-28), the University of Washington Retirement Association Fellowship in Aging Biology, and the American Foundation for Aging Research. B.A.H. is supported by the Mary Gates Undergraduate Fellowship. This work was supported by AG029406 (R.C.R. and P.J.H.) and NSO46724 (P.J.H.).
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.