In recent years, evidence has accumulated that point to the importance of proliferation control and cell cycle regulation in neurogenesis. Hypotheses have been put forth suggesting the expansion of the primate neocortex is the direct result of evolutionary changes in cell cycle parameters (Kornack and Rakic,1998; Kornack,2000), that cell number output in anatomically defined cortical areas and to specific laminae within a specific area are dependent on cell cycle parameters (Miyama et al.,1997; Caviness et al.,2003; Lukaszewicz et al.,2005), and that the decision to remain in the cell cycle or differentiate is dependent on the time a progenitor cell spends in G1 (Calegari and Huttner,2003; Calegari et al.,2005). In addition, increasing numbers of mutations are being described that alter the proliferation characteristics of CNS progenitors, resulting in profound developmental alterations (Levine and Green,2004; Mochida and Walsh,2004; Del Bene and Wittbrodt,2005; Donovan and Dyer,2005).
In many tissues, the unambiguous identification of proliferative stem and progenitor cells is not easily accomplished as proliferative cell populations are not typically compartmentalized in a concise manner and can be mixed with non-dividing cells. A good example of this is the developing vertebrate retina, where a significant proportion of postmitotic differentiating cells are interspersed with retinal progenitor cells (RPCs) in the neuroblast cell layer (NBL; Hinds and Hinds,1979). Even in developing tissues such as the cerebral cortex, which has well-defined proliferative compartments, cell heterogeneity within these compartments complicates the unambiguous identification of distinct populations of progenitors (Gal et al.,2006).
A common way to distinguish proliferative cells from postmitotic cells is by the detection of differentially expressed RNAs or proteins. In many cases, however, the markers may not be direct readouts of proliferation or exclusive to proliferative cells. For example, transcription factors are often used as markers, but expression may be limited to subsets of progenitors or they may also be detectable in postmitotic cells. Furthermore, perturbations in signaling pathways may alter transcription factor expression. These issues necessitate the use of multiple markers to identify progenitor populations and assess their proliferative status. Potentially less restrictive markers are cell cycle proteins such as cyclins, Proliferating Cell Nuclear Antigen (PCNA), and the human nuclear proliferation associated antigen, Ki-67. While cell cycle proteins are attractive because they are fairly direct readouts of proliferation status, their expression characteristics can vary among tissues or cell lines. For many cell cycle proteins including widely used PCNA and Ki67, their expression profiles in progenitors including RPCs and their kinetics of downregulation in postmitotic cells in different CNS regions have not been well documented.
Other methods for identification of proliferative populations rely on readouts of cell cycle–related activities. Examples include the incorporation and detection of nucleotide tracers such as tritiated thymidine ([3H]-TdR) or halogenated uridine derivatives (e.g., bromodeoxyuridine; BrdU) (see Hayes and Nowakowski,2000, and references therein). When cells are briefly exposed to these tracers at the termination of an experiment, the tracers are incorporated into newly synthesized DNA and mark cells in S phase. Labeled cells are then identified by autoradiography for [3H]-TdR or immunohistochemistry for BrdU and other halogenated derivatives. Another example is the immunodetection of phosphoserine histone H3 (pHH3), which marks cells in M phase (Prigent and Dimitrov,2003). As each marker identifies cells in a specific cell cycle phase, they reveal only a subset of the proliferative cell population. Even so, these markers are often used to estimate proliferation status or progenitor cell population size. This can be misleading in comparative studies as changes in the number of cells that incorporate nucleotide tracers or express pHH3 can be the result of alterations in the relative lengths of any of the cell cycle phases, transient cell cycle arrest, permanent cell cycle exit, or even survival of specific subpopulations. While cumulative tracer labeling can be used to estimate the size of proliferating cell populations, cell heterogeneity with tissue compartments and cells undergoing mitosis or cell cycle exit during the labeling interval can complicate the analysis (Nowakowski et al.,1989). Thus, identifying markers that accurately identify proliferative populations would provide another approach to accomplish similar goals.
In this study, we characterized the expression patterns of several cell cycle proteins during retinal development by immunohistochemical detection on tissue sections and by flow cytometry using a set of antibodies that detect different retinal cell populations (Table 1). The combination of these approaches provides a comprehensive view of marker expression in developing tissues such as the retina, which contain complex and dynamically changing cell populations.
In the mouse, the interval of retinal histogenesis extends from approximately embryonic day (E) 11 to postnatal day (P) 8 (Sidman,1961). During this period, RPCs perform two essential functions: they proliferate to produce sufficient numbers of precursors and they permanently exit the cell cycle as they differentiate into neurons and glia. Because the production of postmitotic, differentiated cells overlaps with cell proliferation and RPCs are not easily distinguished from postmitotic cells, we characterized the expression of several cell cycle proteins with the goal of identifying reliable markers of RPCs.
We initially analyzed the expression of PCNA, Ki67, and Mini Chromosome Maintenance Protein 6 (MCM6). PCNA and Ki67 are widely used as markers of proliferating cells and we identified MCM6 in a survey of cell cycle genes expressed in the developing retina (see Supplemental Fig. 1, which can be viewed at www.interscience.wiley.com/jpages/1058-8388/suppmat; unpublished observations). MCM proteins are essential for DNA replication (Tye,1999) and depletion of the MCM5 protein disrupts proliferation in the zebrafish retina (Ryu et al.,2005). Initial examination of these proteins in the P0 retina indicated robust expression of PCNA and MCM6 (see below) whereas Ki67 expression showed variation in staining intensity (Suppl. Fig. 2). Because of their robust detection, the expression characteristics of PCNA and MCM6 were analyzed in more detail.
We compared the expression patterns of MCM6 and PCNA over the interval of mouse retinal histogenesis by double label immunofluorescence (representative ages are shown in Fig. 1). Between E9–E13, both proteins are detected in most cells of the retina and surrounding tissues including the lens, embryonic vasculature, retinal pigment epithelium, and extraocular tissues (Fig. 1A–F). At E16, MCM6 and PCNA proteins are detected in the neuroblast layer (NBL) and excluded from cells located in the developing ganglion cell layer and inner boundary of the NBL, a region we collectively term the differentiated cell layer (DCL; Fig. 1G–I). This general pattern was observed in retinal sections through P3 (Fig. 1J–L). By P6, cells expressing both markers are largely confined to a narrow band of cells in the developing inner nuclear layer (Fig. 1M–O), a pattern consistent with the location of the last remaining RPCs. The double-labeled cells observed in the nascent outer nuclear layer are likely to be RPCs undergoing interkinetic nuclear migration, and those at the inner retinal boundary (adjacent to DCL) are likely to be astrocytes, endothelial cells, or pericytes, which are of non-retinal origin (Fig. 1J–O). By P15, all retinal cells have exited the cell cycle, RPCs are no longer present, and MCM6 and PCNA proteins are not detected (Fig. 1P–R). In sum, MCM6 and PCNA proteins are expressed similarly during retinal histogenesis in a pattern consistent with the laminar distribution of RPCs.
Cell Cycle Profiles of PCNA and MCM6 Expression
Since the DNA content within a cell is an indicator of a cell cycle phase, we used flow cytometry to simultaneously measure the extent of marker expression and DNA content on a per cell basis (Fig. 2). The flow cytometry profile for the total retinal cell population at P0 labeled with propidium iodide (PI) revealed that the majority of cells have a DNA content consistent with G1 or G0 (Fig. 2A; G0 is defined as postmitotic). PI staining was then combined with PCNA or MCM6 staining to determine their cell cycle expression profiles and the data are displayed as density plots (Fig. 2B–E; red indicates regions of highest cell density). The dashed line on each plot demarcates the boundary between marker negative cells ((−); below the line) and marker positive cells ((+); above the line) as determined by the profiles obtained for control samples (Fig. 2B,D). Compared to the controls, the PCNA- and MCM6-labeled samples show a significant upward shift in fluorescence intensity in all phases of the cell cycle (Fig. 2C,E), and as expected, a significant proportion of cells in the G1/G0 phases are not positive for either marker. While the PCNA and MCM6 flow cytometry profiles are generally similar, one notable difference is that cells in the G1/G0 phases separate into two distinct peaks in PCNA-stained samples (Fig. 2C; see also Fig. 6B) whereas MCM6+ and MCM6− cells in the G1/G0 phases are not as easily separable as indicated by the broad shoulder that traverses the boundary between positive and negative cells (Fig. 2E).
Using the flow cytometry data, we estimated the percentage of cells in each phase of the cycle at P0 (Fig. 3). In the total cell population, 79% of the cells are in the G1/G0 phases, 16% in S, and 5% in G2/M (Fig. 3A). Since postmitotic cells are included in these calculations, these numbers do not accurately reflect the cell cycle distribution of RPCs. Based on PCNA staining, we found that 65% of the PCNA+ cells are in the G1/G0 phases, 27% in S, and 8% in G2/M (Fig. 3B), whereas 95% of the PCNA− cells are in the G1/G0 phases and 5% are in S and G2/M (Fig 3C). Importantly, the presence of PCNA− cells in the S and G2/M phases has minimal effects on the cell cycle profile and proportion of RPCs (Fig. 3 D,E; RPCs are defined as the sum of the PCNA+ cells, PCNA− S phase cells, and PCNA− G2/M cells).
Our data thus far show that MCM6 and PCNA are expressed in the majority of RPCs throughout retinal development and suggest that they are downregulated in postmitotic cells. To address this further, we directly compared the MCM6 and PCNA patterns to markers of postmitotic, differentiating cells (Figs. 4–6). The acetylated form of class III beta-Tubulin (acTUBB3) is an early marker of ganglion cells, amacrine cells, and horizontal cells (Sharma and Netland,2007). At E13 and P0, the expression patterns of MCM6 and acTUBB3 are largely complementary although many acTUBB3+ cells are weakly positive for MCM6 in the inner retina (Fig. 4A–F; arrowheads in A–C, brackets in D–F). Double-positive mitotic cells were also observed (Fig. 4D–F; arrows) suggesting that TUBB3 is acetylated prior to cell cycle exit. In contrast, no overlap in expression was observed between MCM6 and Neurofilament-M (NEFM; Fig. 4G–I), which is also expressed in ganglion cells, amacrine cells, and horizontal cells but has a later onset of expression compared to acTUBB3 (Troy et al.,1990; Chien and Liem,1995). Consistent with this, we did not detect NEFM in mitotic cells. MCM6 was compared to Recoverin (RCVRN), an early marker of rod and cone photoreceptors (Sharma et al.,2003), and we did not observe strong MCM6 expression in RCVRN+ cells although weak MCM6 immunoreactivity is evident in some cells (Fig. 4J–L). Finally, we did not observe colocalization of MCM6 with Calretinin (CALB2), a marker of amacrine, displaced amacrine, and ganglion cells (Fig. 4M–O) (Haverkamp and Wassle,2000). As described earlier, the MCM6+ cells at the inner retinal boundary are likely to be of non-retinal origin. The sum of these observations indicates that MCM6 is detectable in newly postmitotic cells but is downregulated as differentiation progresses.
In contrast to MCM6, PCNA and acTUBB3 are coexpressed in relatively few cells, and the double-positive cells are confined primarily to cells in mitosis (Fig 5A–F). PCNA was not detected in acTUBB3+ cells in the newly forming amacrine cell layer (Fig. 5D–F), in NEFM+ cells (Fig. 5G–I), or in RCVRN+ cells (Fig. 5J–L). These results suggest that PCNA protein levels diminish rapidly as RPCs exit the cell cycle and with faster kinetics than MCM6.
Flow cytometry was used to directly compare the cell cycle expression profiles for PCNA and acTUBB3 in P0 retinal cells (Fig. 6). To do this, cells were triple labeled with PI, anti-PCNA, and anti-acTUBB3. As expected, the PCNA expression profile (Fig. 6B) is similar to that shown in Figure 2C, and the acTUBB3 expression profile is complementary to the PCNA profile (compare Fig. 6B and D). Next, cells were partitioned on the basis of their DNA content (PI fluorescence intensity) to determine the extent of PCNA and acTUBB3 coexpression in each phase of the cell cycle. Cell density plots show that in all phases, most PCNA+ cells are acTUBB3−, and PCNA− cells are acTUBB3+ (Fig. 6E–J; quadrants II and IV, respectively). Double-positive cells are present in the G1/G0 and G2/M phases (quadrant III), but these are minor subpopulations as they account for ∼6% of the total cell population. Double-negative cells are present in all phases (quadrant I) with the majority of them found in the G1/G0 phases (∼16% of total cells) and we predict these cells are newly postmitotic rod photoreceptor precursor cells, as these cells do not express acTUBB3 (Sharma and Netland,2007).
Expression Patterns and Cell Cycle Profiles of Cyclin D1 (CCND1), Cyclin A2 (CCNA2), Cyclin B1 (CCNB1), and pHH3
Since CCND1, BrdU, and pHH3 are frequently used to measure proliferation and identify progenitor cell populations, we examined their expression patterns and cell cycle profiles in the P0 retina (Figs. 7 and 8). We also examined CCNA2 and CCNB1, as they are critical regulators of cell cycle–related activities (Fung and Poon,2005). In general, CCND1, CCNA2, BrdU, CCNB1, and pHH3 are expressed in patterns that overlap with MCM6 and PCNA (Fig. 7), although in some pHH3+ cells, MCM6 and PCNA are expressed at low levels (Fig. 7M–R; arrowheads in P–R). We also observed that the patterns of the cyclins, BrdU, and pHH3 are distinct from each other, which suggests that they detect different subpopulations of RPCs although some overlap is expected between markers. For example, CCND1, CCNA2, CCNB1, and pHH3 are detected in mitotic cells (Fig. 7A–F, J–L; arrowheads). One notable difference among these markers is that CCNB1 is detected in the DCL and this pattern is consistent with its expression in postmitotic cells.
Next, we performed flow cytometry to characterize the cell cycle expression profiles for CCND1, CCNA2, CCNB1, and pHH3 (Fig. 8). In general, each marker was detected in the appropriate cells based on their known functions: CCND1 was detected in cells of the G1/G0 phases (Fig. 8B), CCNA2 was detected in S and G2/M cells (Fig. 8C), and CCNB1 and pHH3 were detected in G2/M cells (Fig. 8E,F). Interestingly, the profiles of the cyclins also revealed some unexpected distributions. CCND1+ cells were detected in all phases of the cell cycle and the cellular fluorescence intensity was fairly uniform regardless of cell cycle phase. In contrast, the intensity of CCNA2 detection was proportional to DNA content in each cell. With the possible exception of a small cohort of G1/G0 cells, cells in G1/G0 and early S phases had the lowest CCNA2 expression, and cells in late S and G2/M phases had the highest expression. As with PCNA and CCND1, a small cohort of G2/M cells show reduced CCNA2 expression. The cell cycle profile for CCNB1 is quite distinct from CCND1, CCNA2, PCNA, and MCM6. Most cells are below the defined threshold for immunoreactivity (dashed line) with the exception of some cells in G1/G0 and G2/M phases, and the majority of CCNB1+ cells in G1/G0 phases are likely to be those observed in the DCL (Fig. 7J–L). Interestingly, the general upward shift in fluorescence intensity observed in the cells analyzed by flow cytometry (Fig. 8F) is likely due to relatively weak detection of CCNB1 in most cells (Fig. 7K).
The goal of this study was to identify markers that accurately label RPCs and reveal their proliferative status during retinal histogenesis. We focused on proteins associated with the cell cycle because of its essential role in coordinating the activities needed for cell proliferation. Immunohistological labeling of retinal cryosections allowed us to correlate the spatial and temporal expression patterns of markers with each other and with specific cell populations, but one limitation to this approach is the difficulty in directly assessing the cell cycle status of individual cells. Incorporating flow cytometric analysis allowed us to evaluate the expression of markers with respect to cell cycle phase. The combined utilization of these approaches provides a more detailed picture of the RPC population and expression characteristics of the markers tested than could be obtained from either technique alone.
To accurately identify the RPC population, we sought markers whose expression met the following criteria: they should be expressed in all RPCs, and they should not be expressed in postmitotic cells. With respect to the first criterion, we found PCNA and MCM6 are the most comprehensive RPC markers. Both are expressed in all phases of the cell cycle and in most if not all RPCs throughout retinal histogenesis. Ki67 may also be expressed widely in RPCs, but its highly variable expression in P0 cells limits its potential as a general RPC marker. With respect to the second criterion, the expression characteristics of MCM6 make it less amenable for excluding postmitotic cells. Flow cytometry analysis revealed a graded reduction in MCM6 labeling in cells in the G1 and G0 phases and immunohistological analysis indicated weak, but detectable expression in differentiating cells. In contrast, we detected fewer postmitotic PCNA+ cells and our flow cytometry analysis of PCNA and acTUBB3 revealed a low proportion of double-stained cells in the G1 and G0 phases. Based on its overall expression characteristics, we conclude that PCNA is the best RPC marker of the ones tested.
CCND1 is also broadly expressed in RPCs. Immunohistological analysis revealed expression in cells undergoing mitosis and flow cytometric analysis revealed expression in all cell cycle phases. These findings are surprising because CCND1 function is generally thought to be specific for the G1 phase although recent reports suggest functions in the G2 phase (Stacey,2003). Importantly, our observations indicate that CCND1 should not be used in isolation to identify cells in G1 phase.
Although only analyzed at P0 in this study, CCND1 is likely to be expressed in the majority of RPCs during most stages of retinal histogenesis (unpublished observations). We also found that the proportion of CCND1+ cells (∼52%) is similar to the proportion of PCNA+ cells in the wild type P0 retina (∼54%), which is consistent with our previous findings (Green et al.,2003). In addition, it is unlikely that CCND1 protein persists in postmitotic cells as its forced expression in photoreceptors causes apoptosis (Skapek et al.,2001). While its overall expression characteristics indicate attributes of a suitable general RPC marker, its use for this purpose needs to be carefully evaluated in developmental studies as many signaling pathways are known to regulate CCND1 expression (Musgrove,2006).
In this study, we found that flow cytometry was not only efficient at revealing the cell cycle distribution of the retinal cell population, but was also informative for resolving expression patterns of proteins with respect to the cell cycle. While CCND1, CCNA2, CCNB1, and pHH3 are all expressed in cells in the expected cell cycle phase(s), we also observed several unpredicted patterns. CCND1 expression is not confined to the G1/G0 phases, but is also found in cells in the S and G2/M phases. While the cell cycle profile of CCNA2 reveals an expression pattern consistent with its functions in the S and G2 phases, the monotonic rise in expression level from early S to G2 was not expected, especially since the A-cyclins are thought to be essential for the G1 to S phase transition. It may be that CCNA2 does not need to be highly expressed to promote the G1 to S transition or that CCNA2 is in a complex in early S phase that is difficult to detect by our technique. We also observed that a small population of G2/M phase cells and most cells in G1/G0 had reduced CCNA2 expression. This is expected since the mitotic cyclins, which include CCNA2 and CCNB1, are rapidly degraded by the Anaphase Promoting Complex/Cyclosome from anaphase through early G1, and their downregulation is essential for mitotic exit and preventing precocious S phase entry (Peters,2002). CCNB1 is also expressed in postmitotic cells, a finding that suggests it has functions outside of the cell cycle, which is consistent with a previous report of CCNB1 expression in chick retinal ganglion cells as well as with recent studies indicating functions for cell cycle proteins in axonal outgrowth or activation of cell death pathways in postmitotic neurons (Lefevre et al.,2002; Becker and Bonni,2005; Herrup and Yang,2007). Importantly, our findings reveal that expression of cell cycle proteins such as the cyclins does not directly report the cell cycle phase of an expressing cell and in some cases may not accurately indicate proliferative status.
One limitation of flow cytometry is that proliferating cells in the G1 phase cannot be distinguished from postmitotic cells (G0) since cells in both populations have the same DNA content. If these populations are not compartmentalized or easily separable, markers of proliferating and/or postmitotic cells can be incorporated into the analysis. By using three fluorophores, we were able to separate proliferative and post-mitotic cells into separate phases based on PCNA, acTUBB3, and DNA labeling. This analysis was not only effective in showing that PCNA is expressed in all cell cycle phases, it supported the finding from our immunohistological analysis that PCNA is a comprehensive and definitive marker of RPCs.
Other limitations of flow cytometry are that morphological (i.e., spatial) information is lost upon tissue dissociation. Another constraint is that it is often necessary to start with at least 200,000 cells for each procedure since cell loss during the dissociation and staining procedures can be substantial. While each of these issues needs to be considered with respect to the analyses being done, the advantage of this technique is that it allows the direct identification of cells in each phase of the cell cycle. Although flow cytometry and in situ immuonohistochemistry have these inherent limitations, combining them makes it possible to more accurately and directly assess the cell cycle profiles and proliferative status of progenitor cell populations in the CNS.
Animal Use and Tissue Preparation
Black Swiss and Swiss Webster mice were obtained from Taconic laboratories. To collect embryos, mice were time-mated and the day following conception counted as embryonic day 0 (E0). Pregnant females and pups older than P5 were sacrificed by asphyxiation with carbon dioxide gas followed by cervical dislocation. Pups between the ages of P0 and P4 were anesthetized by rapid immersion in ice and sacrificed by decapitation. Embryos were sacrificed by decapitation. Tissue samples were fixed in 4% neutral buffered formalin for periods ranging from 1 hr to overnight at 4°C and subsequently stepped through a graded sucrose series for cryoprotection. Samples were mounted in OCT (Tissue Tek), frozen on dry ice, and stored at −80°C.
The antibodies used in this study are listed in Table 1. Where possible, the protein names and symbols used in this study were in accordance with those adopted by the Mouse Genome Nomenclature Committee at the time of submission (URL: http://www.informatics.jax.org). Twelve-micron cryosections were pretreated with PBS followed by 30 minutes in blocking buffer (2% goat or donkey serum [dependent on primary antibody], 0.15% Triton-X 100, 0.1% azide, and PBS, pH 7.5). Primary antibodies were diluted in blocking buffer and incubated on slides for 2 hr at room temperature or 4°C overnight and detected with Alexa-Dye™ conjugated secondary IgGs (Molecular Probes) diluted in blocking buffer and incubated on sections for 45 min at room temperature. Sections were coverslipped in Fluormount-G (Southern Biotechnology).
Image Analysis and Processing
Sections were analyzed by epi-fluorescence using a Nikon E-600 microscope and images were captured in gray scale mode with a Spot-RT Slider CCD camera (Diagnostic Instruments, Sterling Heights, MI). Confocal images were scanned with an Olympus Flouview 300 microscope at the University of Utah Health Sciences Imaging Core Facility. RGB images were constructed from individual monochrome channels using Photoshop 7.0 (Adobe Systems Inc., San Jose, CA). The levels function was used to adjust digital images to be consistent with visual observations.
Retinal cell suspensions were prepared by digesting whole retinal tissue with 0.25% activated Trypsin (Invitrogen) in Hanks Buffered Saline Solution lacking calcium and magnesium (Invitrogen). Samples were incubated at 37°C for 8–10 min with gentle rotation. Trypsin was inactivated with fetal bovine serum (FBS; final volume 10% v/v; Invitrogen). Cell suspensions were rinsed twice with PBS, pH 7.5. One milliliter of cells (between 5 × 105 and 1 × 106 cells/ml) was added dropwise to ice-cold methanol for fixation and stored at −20°C.
Samples were warmed to room temperature and rehydrated by rinsing twice in PBS, pH 7.5. Samples were incubated in blocking buffer for 15 min followed by direct addition of primary antibodies (final concentrations listed in Table 1). Samples were incubated for 30 min at RT with gentle mixing every 5 min. Samples were rinsed with PBS followed by secondary antibody incubation for 30 min at RT with gentle mixing. Samples were rinsed with PBS followed by incubation in 500 μl RNase A (250 U/ml; Sigma) for 5 min followed by addition of 100 μg/ml propidium iodide (PI; Invitrogen) for 45 min. Note that all steps requiring buffer changes were preceded by centrifugation (approximately 220g for 5 min at RT). Pelleted cells were subsequently resuspended by mild trituration.
Between 1 × 104 and 3 × 104 cells were analyzed on a Becton Dickinson FACScan Analyzer at the University of Utah Health Sciences Flow Cytometry Core Facility. The software packages used were: CellQuest (Becton Dickinson) for data acquisition; ModFit (www.vsh.com) for cell cycle modeling; and FCSPress 1.4 (http://www.fcspress.com) for data presentation.
We thank Drs. Sabine Fuhrmann, Brian Jones, Elizabeth Leibold, Robert Marc, and Todd Raleigh for critically reading the manuscript at various stages of this project, and we thank Drs. Yuan Wu, Eric Green, Wayne Green, Chris Rodesch, and Anna Clark for their contributions to this work.