Advancement in our understanding of the biology of adult stem cells and their therapeutic potential relies heavily on meaningful functional assays that can identify and measure stem cell activity in vivo and in vitro. In the mammalian nervous system, neural stem cells (NSCs) are often studied using a culture system referred to as the neurosphere assay. We previously challenged a central tenet of this assay, that all neurospheres are derived from a NSC, and provided evidence that it overestimates NSC frequency, rendering it inappropriate for quantitation of NSC frequency in relation to NSC regulation. Here we report the development and validation of the neural colony-forming cell assay (NCFCA), which discriminates stem from progenitor cells on the basis of their proliferative potential. We anticipate that the NCFCA will provide additional clarity in discerning the regulation of NSCs, thereby facilitating further advances in the promising application of NSCs for therapeutic use.
Disclosure of potential conflicts of interest is found at the end of this article.
Much of our understanding of adult stem cells, and their currently accepted definition, comes from the hematopoietic system and the intestinal epithelium [1, 2]. The discovery of considerable and lifelong cell genesis in the blood and the subsequent identification of a self-renewing cell with multilineage differentiation potential set the stage for the identification, isolation, and therapeutic use of blood stem cells . Because of the general lack of unique cell surface markers and the absence of a distinct and discernable morphological phenotype, stem cells are typically defined and studied on the basis of functional criteria. Implicit in identifying a stem cell in vitro is the discovery of culture conditions that permit competent cells to exhibit the cardinal stem cell properties of (a) self-renewal over an extended period of time, (b) generation of a large number of progeny, and (c) multilineage differentiation potential.
The discovery of culture conditions for the expansion of neural stem and progenitor cells from the mammalian central nervous system (CNS)  provided for the first time a relatively simple and robust means to investigate the activity and regulation of these precursor cells . Although other methods are available to culture neural stem cells (NSCs) [5, 6], the neurosphere assay (NSA) remains the most frequently adopted method to enrich, expand, and even calculate the frequency of NSCs . In addition, the serum-free growth conditions of the NSA are also being used to promote sphere formation by stem cells and measure their frequency from a variety of tumors (e.g., breast  and brain [9, 10]) and normal tissues (e.g., breast , skin , cardiac , pancreatic , and embryonic [6, 15, 16]).
We have recently challenged one of the central tenets of the NSA (that all neurospheres are derived from an NSC) and concluded that an exclusively one-to-one relationship between neurosphere formation and NSCs does not exist, thereby suggesting that this tenet is incorrect . Here we begin by providing direct evidence that the NSA, as currently applied, overestimates the frequency of NSCs derived from both the embryonic and adult mammalian brain. We then go on to describe a novel collagen-based semisolid assay, the neural colony-forming cell assay (NCFCA), which has the ability to discriminate stem from progenitor cells and thus provides a method to enumerate NSC frequency. Finally, we demonstrate the value of using the NCFCA to discern changes in stem versus progenitor cells by re-evaluating several highly cited studies that relied on the NSA as a measure of NSC frequency.
Materials and Methods
Striata from embryonic day 14 CD1 albino mice (Charles River Laboratories, Wilmington, MA, http://www.criver.com) were processed and cultured to generate neurospheres in NeuroCult NSC Proliferation Medium and 20 ng/ml human epidermal growth factor (hEGF) (StemCell Technologies, Vancouver, BC, Canada, http://www.stemcell.com) (complete medium). The NeuroCult NSC Proliferation Medium consists of NeuroCult NSC Basal Medium (Dulbecco's modified Eagle's medium [DMEM]/Ham's F-12 medium [F12] [1:1], glucose [0.6%], sodium bicarbonate [0.1%], and HEPES [5 mM]) mixed (9:1) with the 10 × NeuroCult NSC Proliferation Supplement as previously described [18, –20]. The NeuroCult NSC Proliferation Supplement consists of DMEM/F12 (1:1), glucose (0.6%), sodium bicarbonate (0.1%), glutamine (20 mM), HEPES (5 mM), insulin (230 μg/ml), transferrin (1,000 μg/ml), progesterone (200 nM), putrescine (90 μg/ml), and sodium selenite (300 nM). To compare neurosphere numbers in the NSA and colony numbers in the NCFCA (described below), 2,500 embryonic day 14 (E14) cells at passage 2 (p2) were plated in six-well tissue culture dishes for the NSA or in 35-mm plates for the NCFCA.
The periventricular regions (PVR) (or subependymal) of brains of adult C57/BL (Charles River Laboratories) mice infused with saline, epidermal growth factor (EGF), or arabinosylcytosine (AraC) were microdissected, and tissue was enzymatically processed with the NeuroCult Enzymatic Dissociation Kit (StemCell Technologies) or as previously described . Adult cells were plated in either the NSA or the NCFCA described in the supplemental online data. All cultures for adult cells contained the complete medium supplemented with 20 ng/ml hEGF, 10 ng/ml basic fibroblast growth factor, and 2 μg/ml heparin (StemCell Technologies). For comparing neurosphere numbers in the NSA and colony numbers in the NCFCA, 7,500 cells were plated in six-well tissue culture dishes for the NSA or in 35-mm plates for the NCFCA.
To induce differentiation, cells derived from neurospheres generated from the cells within NCFCA colonies were plated in complete NeuroCult differentiation medium (NeuroCult Differentiation Kit; StemCell Technologies) and processed further for immunocytochemistry.
Antibodies and Immunocytochemistry
Primary lineage-specific antibodies were mouse monoclonal antibody to class III β-tubulin (clone TUJ1; 1:1,000), rabbit polyclonal antibody to glial fibrillary acidic protein (1:200), mouse monoclonal IgM antibody to O4 (1:50), and rabbit polyclonal antibody to myelin basic protein (1:100). The secondary antibodies were fluorescein isothiocyanate (FITC)-conjugated goat antibody to mouse IgG and IgM (1:100), Texas Red-conjugated goat antibody to rabbit IgG (1:100), and 7-amino-4-methylcoumarin-3-acetic acid-conjugated goat antibody to rabbit IgG (1:100). Indirect immunocytochemistry was performed according to the manufacturer's protocols.
In Vivo Infusion of EGF and AraC
Twelve hours prior to surgery, osmotic mini-pumps (Alzet, Cupertino, CA, http://www.alzet.com, 1007D; 7-day infusion at flow rate of 0.5 μl/hour) were loaded with EGF (10 ng/ml) or vehicle solution (0.9% sterile physiological saline) attached to the infusion cannula, and the entire apparatus was incubated at 37°C. Adult C57BL/6 mice were anesthetized via intramuscular injection of ketamine hydrochloride (50 mg/kg) and xylazine hydrochloride (8 mg/kg) and mounted onto a stereotaxic frame, and a 1.5-cm incision was made to expose the skull. A single hole was drilled in the skull (1.25 mm width) directly above the lateral ventricle (anterior/posterior, −0.5; lateral, +1.0; depth, −3.7), and the 30-gauge cannula was lowered and fixed into place on the skull by the addition of cyanoacrylate adhesive, so as to enable unilateral infusion directly into the lumen of the lateral ventricle. Mice implanted with pumps were infused for 7 days at a flow rate of 0.5 μl/hour with an initial pump concentration of 33 μg/ml, resulting in 400 ng/ml EGF being delivered per day. For AraC infusion, pumps were loaded with 2% AraC, which was infused at a rate of 1 μl/hour for 1 or 3 days. Mice were placed on a warming plate (37°C) during recovery and monitored continuously until they regained consciousness and began taking water.
Adult mice (8–10 weeks old) were sacrificed by cervical dislocation, and their brains were removed and sectioned coronally at the level of the optic chiasm. PVR tissue was harvested, pooled, diced with a scalpel for 1 minute, and then processed according to Rietze et al. . Following centrifugation at 100 relative centrifugal force (rcf) for 7 minutes, the resulting pellet was resuspended in 1 ml of 0.1 M Dulbecco's phosphate-buffered saline (PBS) (calcium- and magnesium-free; Gibco, Grand Island, NY, http://www.invitrogen.com) and mechanically triturated until it was homogeneous. The resulting suspension was filtered through a 70-μm sieve (Falcon/BD Biosciences, North Ryde, NSW, Australia, http://www.bd.com), cells were collected via centrifugation (7 minutes at 100 rcf) and resuspended in PBS, and the viable cells were counted. For immunostaining, the suspension was incubated for 20 minutes at 4°C with FITC-conjugated peanut agglutinin (1:200; Vecta, Burlingame, CA, http://www.vectorlabs.com) and phycoerythrin-conjugated mCD24a (1:200; clone M1/69; BD Pharmingen, San Diego, http://www.bdbiosciences.com/index_us.shtml). Following incubation, the cell suspension was rinsed twice with PBS via centrifugation and finally in PBS + 1% fetal calf serum + propidium iodide (100 μg/ml; Molecular Probes, Eugene, OR, http://probes.invitrogen.com) to label dead cells before fluorescence- activated cell sorting analysis. Cells were sorted using a FACSVantage SE DiVa (Becton, Dickinson and Company, Franklin Lakes, NJ, http://www.bd.com).
A single-cell suspension of p2 E14 striatal cells or adult PVR was first filtered through a 40-μm cell strainer (StemCell Technologies), and cell counts were performed. For p2 E14 striatal cells, cells were diluted to 220,000 cells per milliliter with complete medium as described above. Primary adult cells were diluted to 650,000 cells per milliliter, also in complete medium. To perform the NCFCA, the neural colony-forming cell (NCFC) serum-free medium without cytokines (Iscove's modified Dulbecco's medium containing bovine serum albumin, insulin, transferrin, and l-glutamine), NeuroCult NSC Proliferation Supplement (as described above and previously [18, 20]), hEGF, and type 1 bovine collagen solution (3 mg/ml) were required. Briefly, the following reagents were added in the order given: 1.7 ml of NCFC serum-free medium without cytokines, 0.33 ml of NeuroCult NSC Proliferation Supplement (as described above), and 7 μl of a 10 μg/ml stock solution of EGF (to give a final concentration of 20 ng/ml). Twenty-five microliters of the cells previously diluted at 220,000 cells per milliliter (p2 E14, or 650,000 cells per milliliter for primary adult cells) was next added to the medium mixture. Last, 1.3 ml of bovine collagen was added to the cell-medium suspension, and the entire suspension was mixed. In each 35-mm dish, in duplicate, 1.5 ml of cell-medium suspension was plated, to give 2,500 cells plated per 35-mm dish (p2 E14). The cultures were incubated for 21 days with regular checks of the cultures. Because of the prolonged culture period, the medium was replenished by depositing 60 μl of complete liquid medium supplemented with concentrated EGF (0.5 μg/ml; p2 E14 cells) plus fibroblast growth factor (0.25 μg/ml) + heparin (0.01%) for adult cells in the center of the dish once every week for the total of 3 weeks.
Using mathematical modeling , we have previously shown that the NSA overestimates NSC frequency by an order of magnitude. Here, we tested this assumption directly by attempting to serially passage the cells from individual neurospheres that were harvested from cultures of embryonic day 14 (E14) striatal cells at p2 or from primary cultures of cells from adult PVR. Whereas 11 of 34 (32%) of the E14-derived neurospheres contained cells that stopped proliferating after a single passage, 7 of 34 (21%), 8 of 34 (24%), and 5 of 34 (15%) of the selected neurospheres contained cells that could be passaged 2, 3, and 4 times, respectively (Fig. 1A). Indeed, only 2 of the 34 neurospheres tested (5.9%) contained cells that demonstrated extensive self-renewal, that is, the ability to passage >6 times. When adult cell-derived neurospheres were tested, once again only 4 of 70 (5.7%) contained cells that exhibited extensive self-renewal (Fig. 1B). These results provide direct evidence that not all neurospheres are derived from stem cells and support our previously stated premise that estimation of stem cell frequency based on enumerating neurospheres in the NSA significantly overestimates stem cell number .
Accordingly, we developed the NCFCA, which enables single CNS precursor cells to exhibit their full proliferative potential and also facilitates the discrimination of stem from progenitor cells based on this criterion. The gold standard for clonal analysis traditionally entails the deposition of a single cell in a single well [22, 23] (supplemental online Fig. 3b). However, this approach is technically challenging, time-consuming, and labor-intensive and therefore is not practical for day-to-day experimentation. To overcome of these problems, we adapted the NSA serum-free medium formulation  from a suspension culture system to a collagen-containing semisolid matrix. Adherent cultures have been used previously to derive clonal colonies via single-cell deposition , low plating density , and genetic marking [24, 25] and in methycellulose [26, 27]. The advantage of the NCFCA over adherent culture systems (e.g., neural precursor [6, 24, 25, 28] or mesenchymal  stem cell cultures) or suspension culture systems  (e.g., neurosphere cultures) is that cells can be plated at a much higher density than a single cell per well, and the derivation of clonal colonies is significantly enhanced as cell migration and aggregation is inhibited (supplemental online Fig. 5).
We initially plated single-cell suspensions at low cell density in a serum-free (adapted from the NSA formulation; supplemental online Fig. 1) medium containing collagen to ensure that distinct colonies were derived from single cells and therefore clonal in origin (supplemental online Fig. 5). Next, to determine whether the collagen-based serum-free medium formulation of the NCFCA inhibited the proliferation of neural precursors, we plated p2 E14 or primary adult PVR cells in the NCFCA and NSA and compared the number of colonies and neurospheres generated after 21 and 7 days in vitro, respectively. Overall, both p2 E14-derived (Fig. 1C) and adult-derived (Fig. 1D) cells generated equivalent numbers of colonies and neurospheres per total cells plated. For p2 E14 cultures, 2.4% ± 0.9% of cells generated neurospheres, whereas 2.9% ± 1.2% formed colonies (mean ± SE; n = 8 independent experiments; p = .23). Similarly, in the adult, 402 ± 67 neurospheres versus 506 ± 61 colonies were generated per brain (mean ± SE; n = 7 independent experiments; p = .28). These results led us to conclude that the NCFCA growth conditions did not inhibit the proliferation of embryonic or adult precursors. Of note, the culture of cells in the NCFCA for 21 days enabled the maximal proliferative capacity of the cells to be manifested over time and therefore allowed for the generation of colonies with a diverse size distribution, presumably because of differences in the proliferative potential of the original colony-forming cell (Fig. 2A–2D; supplemental online Fig. 2). On closer inspection, we observed that the majority of p2 E14 (76.5% ± 1.9%; Fig. 2E) and adult (59.3% ± 1.4%; Fig. 2F) colonies were <0.5 mm in diameter (Fig. 2A), whereas only 2.8% ± 0.7% of p2 E14 cells and 2.0% ± 0.6% of adult cells formed colonies >2.0 mm. On the basis of the premise that progenitor cells would exhibit limited proliferative capacity in relation to stem cells, these data suggested that we could use the size (diameter) of the colony to distinguish its founder cell type.
Accordingly, we first categorized the colonies into four groups (<0.5, 0.5–1, 1–2, and >2 mm) and then plated all the cells from representative individual colonies into a single well of a 96- or 24-well plate and cultured the cells in NSA conditions. We found that cells within p2 E14-derived colonies with a diameter of <0.5, 0.5–1, 1–2, or >2 mm produced secondary neurospheres 25%, 43%, 71%, or 100% of the time, respectively, whereas tertiary neurosphere formation was restricted to cells within colonies 1–2 mm (50% exhibiting tertiary neurosphere formation) and >2 mm (100% exhibiting tertiary neurosphere formation; Fig. 3A). Similarly, cells within adult-derived colonies <0.5, 0.5–1, 1–2, and >2 mm generated secondary neurospheres 28%, 53%, 80%, and 100% of the time, respectively (Fig. 3B), with tertiary neurosphere formation occurring from cells in 50% of 1–2-mm colonies and from cells in 100% of >2.0-mm colonies. To preclude the possibility that secondary and tertiary neurosphere formation was related to the number of cells (or numbers of stem cells) plated per well and therefore indirectly related to colony size, we serially diluted cells harvested from p2 E14-derived colonies >2.0 mm in diameter over a range of 10,000 to 1,000 cells per cm2. Given that colonies <0.5 mm in diameter contain fewer than 1,000 cells, we pooled cells isolated from 10 individual colonies <0.5 mm in diameter, which would yield numbers of cells approximately equivalent to those harvested from colonies >2 mm in diameter, and plated all the cells into a single well. In both cases, the number of cells that were obtained from the different-sized colonies and then plated in individual wells did not affect the (in)ability of cells within colonies to form secondary and tertiary neurospheres (n = 3; data not shown). Hence, the ability of different-sized colonies to exhibit self-renewal was not related to cell number but rather to the size of the colony.
As progenitor cells are known to exhibit limited proliferative potential [17, 31], we continued to serially passage the dissociated cells from tertiary neurospheres derived from colonies >2 mm in diameter. We found that for both p2 E14-derived (n = 6) and adult-derived (n = 9) neurospheres, 100% of colonies >2 mm contained cells that could be serially passaged at least seven times, generating a characteristic geometric increase in total cell numbers at each passage (Fig. 3C) and demonstrating the stem cell attributes of extensive self-renewal and generation of a large number of progeny [17, 32]. Seven was used as the minimal number of passages to demonstrate self-renewal and proliferation over an extended period of time. Similarly, upon replating in the NCFCA, only cells from colonies >2 mm in diameter produced secondary colonies of all four sizes, including secondary colonies >2 mm in diameter, whereas cells from colonies <2 mm generated secondary colonies <1 mm in diameter or no colonies, indicating a lack of self-renewal ability (supplemental online Fig. 4). Finally, to test for multipotency, colonies from each size category were harvested and dissociated, and the resulting cells were either plated directly into serum-containing cultures to promote differentiation and test multilineage potential or initially passaged in serum-free liquid cultures before being transferred into differentiation conditions. Consistent with a putative stem cell-derived colony, cells harvested directly from embryonic or adult colonies >2 mm in diameter generated neurons, astrocytes, and oligodendrocytes, and this multilineage potential was maintained after 10 passages in the NSA conditions (Fig. 3D–3H). In contrast, >90% of cells isolated from 1–2-mm colonies were bipotent (producing astrocytes and oligodendrocytes), whereas the remaining 10% were unipotent (producing astrocytes or oligodendrocytes), indicating that the original colony-forming cell of these colonies was not multipotent.
The clonal nature of the colonies was directly addressed by coculturing experiments (using EGFP-labeled and non-EGFP-labeled cells), revealing that approximately 99% of colonies in the NCFCA were derived from single cells, in contrast to the neurosphere assay, where we found that 80%–96% (depending on plating density) of the spheres where clonally derived (supplemental online Fig. 5). Serial dilution of primary E14-derived cells found a strong positive correlation (r2 = 0.9673) between numbers of large (>2 mm) stem cell-derived colonies and total cells plated, indicating a stem cell frequency of 1:5,250 (supplemental online Fig. 3a). Single-cell deposition using a flow cytometer or manual limiting serial cell dilutions demonstrated that a single cell could form a large stem cell (>2 mm) colony, with a frequency of 1:2,688 (supplemental online Fig. 3b). Together, these results suggest that proliferation of NSCs in the NCFCA can occur cell autonomously and that the frequency of the stem cells in primary E14 ganglionic eminence lies in the range of 1:2,688–1:5,250.
We next sought to demonstrate the utility of the assay by re-evaluating several research reports that have relied on the NSA as a measure of stem cell frequency. We previously reported that infusion of EGF for 6 days into the lateral ventricle of adult mice results in a 370% increase in the number of neurospheres derived from PVR cells, and we concluded that EGF stimulates the relatively quiescent PVR stem cells to expand in vivo . When we repeated this study and compared the number of neurospheres (NSA) and total colonies (NCFCA) produced, we observed a similar increase in both the number of neurospheres (1,546 ± 352 neurospheres per brain; mean ± SE; n = 4) and total colonies (1,615 ± 130 colonies per brain; mean ± SE; n = 9) from EGF-treated animals compared with saline-infused controls (432 ± 71 neurospheres per brain; mean ± SE; n = 3; Fig. 4A). However, when the numbers of colonies in the different size categories were analyzed (Fig. 4B), it was evident that EGF infusion had virtually no effect (10.8 ± 2.7 [n = 9] in the EGF-treated group vs. 12 ± 2.3 [n = 3] in the control) on the number of stem cell-derived colonies (>2 mm in diameter). To confirm that the EGF infusion did not alter the growth characteristics of the in vivo-stimulated precursors, such that the increased numbers of small colonies now also harbored stem cells, we individually isolated and attempted to serially passage cells from 40 small (<0.5 mm) colonies in addition to pooling cells from 10 such small colonies and were unable to demonstrate that these cells exhibited stem cell characteristics in vitro (data not shown). Hence, whereas EGF infusions increased the number of progenitor cells, these results clearly demonstrate no such increase in the number of NSCs as previously suggested [33, 34].
We further questioned the conclusions drawn from in vivo studies, which relied on the NSA to monitor neural stem cell frequency. It is known that the infusion of the antimitotic agent AraC almost completely eliminates all mitotically active cells in the PVR. A similar decrease was seen in neurosphere formation, and on the basis of the assumption that every neurosphere is formed by an NSC, the authors concluded that stem cells are mitotically active [34, –36]. Consistent with earlier reports, following a 3-day infusion of AraC, we observed an approximately 80% reduction in the number of neurospheres generated (74 ± 28 neurospheres per brain; mean ± SE; n = 7) relative to the saline-infused controls (402 ± 67 neurospheres per brain; mean ± SE; n = 7; Fig. 4C). Although this reduction was mirrored in a decline in total number of colonies in the NCFCA (108 ± 14 colonies per brain; mean ± SE; n = 8; Fig. 4C), the major loss occurred in colonies <0.5 mm in diameter (402 ± 31 vs. 44.9 ± 7, control vs. AraC; n = 6 and 8, respectively; Student's t test, p = .0005); in addition, there was a smaller but still significant reduction in the number of large (>2 mm) stem cell-derived colonies (Fig. 4D, inset; 13.7 ± 1.3 saline vs. 8.2 ± 1.7 AraC infusion; n = 6 and 8, respectively; mean ± SE; Student's t test, p = .02). Akin to the 3-day AraC infusion, a 1-day infusion (Fig. 4E) produced a large (65%) reduction in the total number of colonies; however, there was no significant change in the number of large (>2 mm) stem cell-derived colonies (Fig. 4E, inset; 16.5 ± 2.9 [n = 4] saline vs. 13.5 ± 1.1 [n = 8] AraC infusion; mean ± SE; Student's t test, p = .14). These results imply that the stem cell population is relatively quiescent and, in agreement with previous studies, suggest that upon injury to the constitutively proliferating cells of the subventricular zone (SVZ), a significant portion of the stem cell population is induced to divide within the first few days after injury in an attempt to repair the damaged tissue .
Finally, as we and others have used the NSA as a readout for NSC activity in experiments aimed at purifying or enriching an NSC population [21, 38, –40], we re-evaluated the stem cell content of what we previously considered an essentially pure (1:1.28) population of adult NSCs, which we obtained by collecting PNAloHSAlo cells >12 μm in diameter . Although a significant enrichment of stem cell-derived colonies was realized, as reflected in the number of large (>2 mm) colonies (15% ± 2.0% [n = 3] vs. 1.0% ± 0.5% [n = 3]; Student's t test, p = .003), our data here indicate that the stem cell content in this precursor pool was far below what we previously reported using the NSA (Fig. 4F).
Since its introduction in 1992, the NSA has been used not only to isolate, expand, and identify the presence of an NSC population but also to enumerate NSC frequency. It is generally well accepted that the assay measures what it is purported to measure, which is apparent in its broad use  to understand the effects of exogenous signaling molecules [41, –43] and genetic alterations [41, 42, 44, 45] and as an assay to evaluate positive [38, 40] and negative  selection strategies. Recently, we challenged the cogency of the NSA by demonstrating experimentally and by mathematical modeling  that the central premise of the NSA—that all neurospheres are derived from an NSC—is not true, thereby precluding the use of the NSA to meaningfully detect changes in NSC frequency. These results implied that although the NSA provides a simple means to isolate and expand NSCs harvested from the embryonic and adult mammalian CNS, its application as a quantitative in vitro assay for measuring NSC frequency is limited. To address the need for an assay that can reliably detect alterations in NSC frequency, we have developed a new single-step assay, the NCFCA, which allows discrimination between NSCs and progenitors by the size of colonies they produce (i.e., their proliferative potential).
If enumeration of neurosphere numbers in the NSA were an accurate measure of stem cell frequency, this would imply that every neurosphere is derived from a stem cell, and therefore, all neurospheres should exhibit stem cell properties. We directly tested this assumption by attempting to serially passage the cells dissociated from individual neurospheres and thereby demonstrate that they exhibit the cardinal stem cell characteristic of extensive self-renewal. We found that out of a total of 104 spheres (34 embryonic and 70 adult), only 6 contained a cell or cells that could be passaged more than seven times, revealing that less than 6% of neurospheres were derived from cells that exhibit extended self-renewal. It has become widely accepted that demonstration of secondary or tertiary neurosphere formation is sufficient to satisfy the criteria of self-renewal [4, 46, –48]. However, on the basis of our results and the argument that the spirit of self-renewal is related to long-term and not short-term self-renewal, the demonstration of secondary or even tertiary spheres would seem inadequate to satisfy this key stem cell defining attribute.
A comparison of the number of colonies or neurospheres generated after 21 or 7 days in vitro, respectively, revealed that both p2 E14-derived (Fig. 1C) and primary adult-derived (Fig. 1D) cells generated equivalent numbers of colonies and neurospheres per total cells plated. These results argue that the growth conditions of the NCFCA are not inhibitory to the proliferation of neural precursors and suggest that the same overall population of EGF-responsive neural precursor cells is being detected. Although similar numbers of neurospheres and colonies were produced, the more informative NCFCA demonstrated that only a minor population (<3%) of these cells exhibited the cardinal in vitro properties of a stem cell (Fig. 3). These results imply that less than 3% of the neurospheres generated in the NSA are truly stem cell-derived, whereas the vast majority arise from progenitor cells. The variance between the serially passaged and the NCFCA calculated stem cell frequency (6% and 3%, respectively) may be accounted for by the relatively small number of neurospheres that were selected for subculturing (104 neurospheres compared with more than 1,000 colonies analyzed in the NCFCA) or by our stringent application of the size categories to the colonies where small differences in diameter were not further subdivided. Regardless, it is clear that the majority (>90%) of the cells that proliferate in the NSA and NCFCA do not exhibit stem cell features in vitro. We have previously estimated, on the basis of mathematical modeling of serially passaged E14-derived CNS cells, that NSC frequency is approximately 0.16% of the total cells , which is an order of magnitude lower than that predicted by counting total neurospheres in the NSA [49, –51]. Using cells from p2 neurospheres generated from E14 striatal cells, the NCFCA estimates stem cell frequency (frequency of colonies >2 mm in diameter) to be 0.07% of total cells, a value more in line with that suggested by our mathematical model .
As the above experiments provide support for the cogency of the NCFCA and its capacity to distinguish between neural stem and progenitor cells from embryonic- and adult-derived neural precursor cells, we next sought to address the utility of the assay by re-examining the conclusions of studies that have relied on the NSA as a readout of NSC frequency. Although our data support the findings that infusions of EGF or AraC increase or decrease the number of neurospheres, respectively, they do not support the conclusions drawn from those results. We had validated the NCFCA for the control and EGF-infused adult mouse CNS cells that colonies >2 mm in diameter are derived from stem cells because cells within these colonies show high proliferative potential and multilineage differentiation over time, whereas cells within colonies <2 mm in diameter are derived from progenitors that have limited proliferation and differentiation potential. Based on our validation studies, in our analysis with the NCFCA, we saw an increase mainly in the numbers of colonies <0.5 mm in diameter (progenitor-derived) and not colonies >2 mm in diameter (derived from stem cells) after in vivo infusion of EGF, suggesting that the vast majority of in vivo or in vitro EGF-responsive cells of the SVZ are not stem cells but rather are restricted progenitors with a more limited proliferative and differentiation potential. It would follow, then, that the in vivo transient amplifying cells are not converted into NSCs in vitro  and that our results support the long-standing [1, 2] and recently challenged [52, 53] notion that NSCs are relatively quiescent in vivo.
The NSA has been used as the readout for strategies aimed at purifying or enriching an NSC population [21, 38, –40]. In these cases, the reliability and accuracy of the NSA is critical for evaluating a particular enrichment strategy, with errors in the readout ultimately leading to the pursuit of an erroneous or irrelevant NSC population. We choose to re-address an earlier report of ours demonstrating an approximately 80% purification of a putative adult NSCs. The NCFCA revealed that only 15% of the colony-forming cells exhibited NSC characteristics, implying that the 80% NSC purity that we reported previously was more on the order of 12%. However, although our sorting strategy did not give the purity that we had previously thought, there was 7–15-fold enrichment relative to controls.
We believe we have demonstrated both by mathematical modeling  and direct experimental evidence that the central premise of the NSA, which is that all neurospheres are derived from a neural stem cell, is not true. The culture conditions of the NCFCA allow for a stem or progenitor cells to exhibit their full proliferative capacity (high, limited, or low), allowing NSC and neural progenitor cell frequency to be estimated on the basis of colony size categories. We have used the NCFCA to demonstrate that in vivo transit-amplifying cells are not converted in vitro into NSCs, that the in vivo NSC is relatively quiescent, and that a reported highly purified putative NSC population is composed primarily of progenitor cells. Our results suggest that some of the previous studies, where neurosphere numbers were used to measure stem cells, need to be re-addressed.
In addition to providing a more meaningful readout of NSC activity in vitro, the NCFCA offers further advantages over the NSA. For instance, it has recently been demonstrated that under certain NSA culture conditions, the vast majority of putative clonally derived spheres are chimeras . The use of a collagen-based three-dimensional semisolid medium prevents migration of single cells and aggregation of colonies, thereby virtually eliminating chimera clones. Although it is still possible for small clusters (doublets or triplets) of cells to be plated, consequently giving rise to nonclonal colonies, the NCFCA provides a reasonable and more practical alternative to the gold standard of single-cell deposition.
The meaningful estimation of stem cell number is important not only in developmental and regenerative neurobiology but also in other areas of tissue-specific stem cell research. Furthermore, regardless of tissue type, particular caution should be taken in concluding that in vitro proliferation and/or sphere formation equates to stem cell function [6, 8, , , , , –14] in vivo.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.
This work was supported in part by National Health and Medical Research Council of Australia Project Grant 301134 (to R.L.R.) and a Pfizer Australia Senior Research Fellowship (to R.L.R.). We acknowledge Carmen Mak for technical assistance.