Research Article

Functional Hierarchy and Reversibility Within the Murine Spermatogenic Stem Cell Compartment

See allHide authors and affiliations

Science  02 Apr 2010:
Vol. 328, Issue 5974, pp. 62-67
DOI: 10.1126/science.1182868


Stem cells support tissue maintenance by balancing self-renewal and differentiation. In mice, it is believed that a homogeneous stem cell population of single spermatogonia supports spermatogenesis, and that differentiation, which is accompanied by the formation of connected cells (cysts) of increasing length, is linear and nonreversible. We evaluated this model with the use of lineage analysis and live imaging, and found that this putative stem cell population is not homogeneous. Instead, the stem cell pool that supports steady-state spermatogenesis is contained within a subpopulation of single spermatogonia. We also found that cysts are not committed to differentiation and appear to recover stem cell potential by fragmentation, and that the fate of individual spermatogonial populations was markedly altered during regeneration after damage. Thus, there are multiple and reversible paths from stem cells to differentiation, and these may also occur in other systems.

Maintenance of adult tissues is supported by a small number of undifferentiated stem cells that self-renew to maintain their population and produce differentiating progeny for normal tissue function. It has generally been accepted that differentiating daughter cells progress unidirectionally toward terminal differentiation. This view has recently been challenged by data suggesting that under some circumstances, differentiating cells can revert to the self-renewing stem cell pool (18). This apparent plasticity may add robustness to maintenance of the stem cell population during normal tissue maintenance and may play a crucial role in tissue regeneration after injury. However, the nature of the self-renewing stem cells and the plasticity of differentiating cells in the maintenance of tissue homeostasis and regeneration are mostly unknown, particularly in mammals.

Germ cells share a characteristic feature across all animal species. Although the most primitive cells in adult gonads are singly isolated, their differentiating progeny remain connected by intercellular bridges to form syncytial cysts of 2n cells (9, 10). Thus, the length of the cysts reflects their cell division history or lineage. This unique feature has made the germ line one of the most tractable systems to study adult stem cell self-renewal and differentiation (2, 3). The study of the spermatogenic stem cell compartment in mammals also relies on heterogeneity in the cyst length (9, 11, 12). In the mouse testis, the most primitive subset of diploid germ cells (spermatogonia) includes Asingle (As, single isolated spermatogonia), Apaired (Apr, interconnected spermatogonial pairs), and Aaligned (Aal, consisting of 4, 8, or 16 interconnected spermatogonia, specifically termed Aal-4, Aal-8, and Aal-16, respectively). A vast majority of stem cell function, if not all, resides in this population. These cells transform without cell division into more differentiating A1 spermatogonia, which subsequently undergo six mitotic and two meiotic divisions to form haploid spermatids (10, 13) (fig. S1).

The prevailing rodent stem cell model (14, 15) (Fig. 1A) assumes that the stem cell population resides in the As population and that cyst length reflects the extent of differentiation in a linear manner (9, 11). A corollary of this “As model” is that As spermatogonia are functionally homogeneous, that all As cells are stem cells, and that all cells are equivalent in each morphological category (9, 10). This simple and attractive model, proposed in 1971, has provided the framework for years of germline stem cell research in mice and other animals. However, the lack of appropriate molecular markers and experimental tools has hindered its critical evaluation.

Fig. 1

The “As model” and hierarchical gene expression in cysts of As, Apr, and Aal spermatogonia. (A) Schematic representation of the “As model.” Arrows indicate the proposed flows of the cells in each cyst type. The width of the open vertical arrows reflects the relative probability of each cyst type to differentiate. (B) Whole-mount immunofluorescence for GFP (green) and GFRα1 (magenta) in Ngn3-EGFP mouse seminiferous tubules. Arrows and arrowheads indicate NGN3+ (GFP+) and GFRα1+ cells, respectively. Scale bar, 100 μm. (C and D) Summarized cell counts from immunostaining shown in (B), shown as total frequency of single- and double-positive spermatogonia regardless of cyst length (C) and frequency of single- and double-positive cysts harboring the indicated number of cells (D). Numbers of counted cells (C) and cysts (D) are shown above each bar. In (D), cysts of other than 2n cells are omitted. (E) Whole-mount immunofluorescence using a mixture of antibodies to GFP and GFRα1 (green) and antibody to E-CAD (magenta) in Ngn3-EGFP mouse seminiferous tubules. Summarized counts are shown under the panel. Scale bar, 100 μm.

In recent years, substantial progress has been made in identifying genes that are expressed in As cells and cysts of Apr and Aal [e.g., GFRα1, PLZF, E-CAD (E-cadherin), and NGN3] (1623). Heterogeneity in gene expression among cysts of the same length has suggested possible functional heterogeneity within cells of the same cyst length (2123). Here, we used gene expression, cyst length, lineage analysis (6), and live imaging (24) to revisit the long-held assumptions of the functionality of the spermatogonial population in mice.

Stratification of spermatogonia by morphology and gene expression. Comparison of expression patterns of genes that mark the As, Apr, and/or Aal population (1623) by whole-mount double staining of seminiferous tubules, the spermatogenic center of the testis, revealed that the two genes PLZF (17, 18) and E-CAD (21) have essentially identical expression patterns and are found in essentially all the As, Apr, and Aal spermatogonia (fig. S2 and supporting online text). In contrast, two other genes, GFRα1 and NGN3, were expressed in minor and major subpopulations of the E-CAD+ total As, Apr, and Aal population, respectively, with the same gene expression observed in all the cells within an individual cyst (Fig. 1, B and C). Intriguingly, all the E-CAD+ cysts expressed either or both of these genes (Fig. 1E). Thus, spermatogonial cysts were heterogeneous in the expression of GFRα1 and NGN3 even in the same morphological fraction, except for Aal-16, which was essentially all NGN3+ (Fig. 1D).

Thus, the As, Apr, and Aal populations can be stratified by both morphology (cyst length) and gene expression (GFRα1 single-positive, GFRα1-NGN3 double-positive, and NGN3 single-positive). These two parameters are mutually correlated: Shorter cysts have a greater probability of being GFRα1 single-positive, whereas longer cysts tend to be NGN3 single-positive.

A functional hierarchy between the GFRα1+ and NGN3+ subpopulations. The observation that GFRα1+ cells are largely As or Apr, whereas NGN3+ cells are mainly Aal (Fig. 1D), suggests that cells make a transition from GFRα1+ → NGN3+ → A1 spermatogonia. To investigate this possibility, we used pulse-chase and live-imaging experiments.

First, we analyzed the fate of NGN3+ cells that were irreversibly labeled with green fluorescent protein (GFP) by a single administration of 4-OH-tamoxifen (Fig. 2A) (6). Two days after the pulse, a majority of the GFP-tagged cells were E-CAD+ GFRα1 KIT (KIT is a marker for A1 spermatogonia through early spermatocytes) (25) (Fig. 2, B to D). The labeling efficiency of the NGN3+ cells was ~30% for all morphological fractions (fig. S3 and supporting online text). Ten days after the pulse, most of the GFP-tagged cells were present in large cysts (>16 cells) that were E-CAD GFRα1 KIT+ (Fig. 2, E to H), indicating that most NGN3+ cells left the E-CAD+ compartment and went on to differentiate. Eventually, all KIT+ spermatogonia were derived from NGN3+ cells, as they retained the GFP in the transgenic mice that express GFP under the control of the Ngn3 regulatory sequence (Ngn3-GFP) (16) (fig. S4). Second, live imaging of the testis of the same Ngn3-GFP mice (16, 24) was used to directly show that a gain of GFP signal, which reflects an induction of NGN3, was frequently observed in GFP (i.e., GFRα1 single-positive) As, Apr, and Aal spermatogonia (Fig. 2I, fig. S5, and movie S1). The minor GFRα1-NGN3 double-positive cells are likely to be in transition from GFRα1 single-positive to NGN3 single-positive cells, because both signals were weaker in double-positive cells than in either of the single-positive populations.

Fig. 2

Behavior of the NGN3+ population in steady-state spermatogenesis. (A) Schedule for (B) to (G). Ngn3/CreERTM;CAG-CAT-EGFP mice were administrated 4-OH-tamoxifen; their testes were analyzed 2, 10, and 20 days after pulse. (B to G) Whole-mount staining of the seminiferous tubules for EGFP (green) and E-CAD [(B) and (E)], GFRα1 [(C) and (F)], or KIT [(D) and (G)] (magenta) 2 days or 10 days after pulse. Scale bars, 100 μm. (H) Contribution of pulse-labeled cells to the E-CAD+ total population. Averages ± SEM from 5, 3, and 5 testes for days 2, 10, and 20, respectively, are shown; *P = 0.006, **P < 0.001, respectively, against the values of day 2 (Student’s t test). (I) Selected frames from live imaging (movie S1) in Ngn3-GFP mouse testes. Orange, yellow, and green arrowheads indicate the gain of GFP fluorescence in As, Aal-4, and probable Aal-8, respectively. The indicated As cell divided into an Apr; asterisk denotes a blood vessel.

Functional hierarchy within the As spermatogonia. We then asked whether As spermatogonia are functionally heterogeneous. After a single pulse of 4-OH-tamoxifen to permanently label NGN3+ cells with GFP, the contribution of the GFP-tagged cells to the entire As population rapidly decreased (Fig. 3A), indicating that the majority of the NGN3+ As cells left the As compartment. Live imaging showed that NGN3+ As cells mostly became NGN3+ Apr, while a few of them divided into two As cells (Fig. 2I, fig. S5, and movie S1). In addition, 2 days after the pulse, a small number of labeled single cells were KIT+ (Fig. 3C), which suggested direct conversion of NGN3+ As into A1 spermatogonia, although they might represent a novel and unique population. We conclude that As cells are not equivalent, and that a vast proportion of NGN3+ As cells are essentially transit-amplifying cells rather than self-renewing stem cells.

Fig. 3

Behavior of NGN3+ As cells in steady-state spermatogenesis. (A) Percentage of labeled As in E-CAD+ As in testes of Ngn3/CreERTM;CAG-CAT-EGFP mice 2, 10, and 20 days after pulse labeling. Averages ± SEM from 5, 3, and 5 testes for days 2, 10, and 20, respectively, are shown; *P = 0.006, **P < 0.001 (Student’s t test). (B) Percentage of GFRα1+ cells in labeled As spermatogonia. Total numbers of GFP-tagged As cells for days 2, 10, and 20 are 171, 8, and 6, respectively; †P < 0.02 (χ2 test). (C) Whole-mount seminiferous tubules stained for GFP (green) and KIT (magenta) 2 days after pulse labeling, indicating rare examples of GFP-KIT double-positive single (arrow) and paired (arrowhead) cells, observed at a frequency of 1 out of ~10 to 20 GFP-tagged E-CAD+ As cells. Scale bar, 100 μm.

However, by day 20, a small number of GFP-tagged cells were still present in the As compartment (Fig. 3A). A majority of these persisting GFP-tagged As cells were also GFRα1+ (Fig. 3B). These data suggest that a small fraction of the NGN3+ cells can revert to GFRα1+ As cells. Theoretically, such “reverted” GFRα1+ As cells could be derived from NGN3+ As or from NGN3+ Apr and Aal, although the latter events would require cyst fragmentation.

Fragmentation of spermatogonial clones observed by live imaging. Live imaging of Ngn3-GFP transgenic mouse testes, in which NGN3+ cells were labeled with GFP, provided direct evidence of spermatogonial cyst fragmentation (Fig. 4 and fig. S5). The frequency of this fragmentation was much lower than that of the normal divisions that double the cyst length (fig. S5, B and C). In one case out of the three observed (Fig. 4 and movie S2), a GFP+ (i.e., NGN3+) Aal-8 cyst divided synchronously and then fragmented into two pairs of interconnected cells, which later divided into Aal-4, whereas the remainder of the 12-cell cyst underwent synchronized death. The other two cysts showed different patterns of fragmentation, excluding the possibility of a stereotypic fragmentation pattern (fig. S5C and movies S3 and S4). In these instances, the GFP signal decreased in many of the resultant shorter cysts and As, prompting us to postulate that they might have become GFRα1 single-positive. Cysts with lengths other than 2n cells were also generated as a result of cyst fragmentation.

Fig. 4

Clone fragmentation of NGN3+ Aal spermatogonia. (A and B) Selected frames from live imaging of an Ngn3-GFP transgenic mouse testis (movie S2) (A) and their schematic representation (B), indicating fragmentation of NGN3+ Aal spermatogonia. After synchronous division of an Aal-8 cyst, two pairs of connected cells, indicated by magenta and orange arrowheads in (A) and dots in (B), survive; the remaining cells in the large chain subsequently died [small, light blue arrowheads in lower left panel of (A)]. Elapsed time is indicated.

Capture of latent stem cell potential during tissue regeneration. During steady-state spermatogenesis, the As, Apr, and Aal subpopulations are constant by definition. However, during regeneration after damage, self-renewal is favored over differentiation (11). We investigated the fate of the As, Apr, and Aal subpopulations during regeneration after administration of busulfan, a drug preferentially toxic to spermatogonia including stem cells (Fig. 5, A to D). About 8 days after busulfan treatment, a majority of E-CAD+ cells had died, although some had formed small regenerating clusters. By day 18, such clusters became prominent and the local density of E-CAD+ spermatogonia had recovered to a level comparable to that in untreated testes, although their average density was still low (Fig. 5B). During the recovery period, we found that both cyst length and gene expression were altered; the average cyst length became shorter and a greater percentage of cells were GFRα1+ (Fig. 5, C and D).

Fig. 5

Behavior of spermatogonial subpopulations during regeneration. (A to D) Analysis of the As, Apr, and Aal subpopulations after administration of a submarginal dose of busulfan into Ngn3-EGFP adult mice. According to the schedule in (A), testes were processed for cell counting by whole-mount immunofluorescence. Shown in (B), (C), and (D) are numbers of total (E-CAD+) As, Apr, and Aal spermatogonia per 1000 Sertoli cells, the average length of E-CAD+ cysts, and the percentage of GFRα1+ (magenta) and NGN3+ (green) As, Apr, and Aal spermatogonia within the total E-CAD+ population, respectively. Values are shown as average ± SEM [n = 6, 6, and 3, for all the data points in (B), (C), and (D), respectively]; *P < 0.006, **P < 0.001, respectively, against the values of day 0 (Student’s t test). (E to I) Pulse-label analyses of NGN3+ spermatogonia during regeneration (solid circles) and steady-state spermatogenesis (open circles). Two days after the pulse, Ngn3/CreERTM;CAG-CAT-EGFP doubly transgenic mice were injected with busulfan to induce regeneration. In the control steady-state group, busulfan treatment was omitted. (F) and (G) Percentage of labeled cells in the E-CAD+ total population (As, Apr, and Aal) and the E-CAD+ As population, respectively, following the schedule in (E). (H) Percentage of GFRα1+ As cells in the total GFP-labeled As population. Total numbers of GFP-tagged As cells for 0, 8, and 18 days were 171, 24, and 108, respectively; †P < 0.006 (χ2 test). (I) Percentage of GFP-labeled cells within the GFRα1+ As population. Averages ± SEM from 5, 3, and 5 testes for days 0, 8, and 18, respectively, are shown in (F), (G), and (I); *P < 0.03, **P < 0.006, ***P < 0.001 (Student’s t test). Data for steady-state spermatogenesis in (F) and (G) are reproduced from Figs. 2 and 3, respectively.

A change in these two parameters during recovery could reflect preferential elimination of longer NGN3+ cysts, a decrease in their formation, an increase in cyst fragmentation, and/or reversion from NGN3+ to GFRα1+. Although the first possibility is not excluded, fate analysis of the pulse-labeled NGN3+ cells is compatible with the other three scenarios (Fig. 5, E to I). A higher percentage of labeled GFP+ cells (which had been NGN3+ at the time of labeling) were retained in the E-CAD+ population during regeneration than during steady-state spermatogenesis (Fig. 5F). Moreover, a significantly higher percentage of the labeled cells contributed to the E-CAD+ As fraction (Fig. 5G) and were positive for GFRα1 (Fig. 5H). We also found that the percentage of labeled cells in the GFRα1+ As population increased during regeneration (Fig. 5I). These findings indicate that either the labeled NGN3+ As cells remained as As and “reverted” back to being GFRα1+, or alternatively, that NGN3+ Apr or Aal cells gave rise to GFRα1+ As cells through cyst fragmentation.

In addition, the frequency of GFRα1+ Aal-8 and Aal-16 cells, which were observed in steady-state spermatogenesis only rarely, was elevated during regeneration (fig. S7). This indicates a delay of the GFRα1+ → NGN3+ transition, a NGN3+ → GFRα1+ reversion in long cysts, or both. Finally, as observed previously (6), the contribution of the labeled cells to the long-term stem cell pool in regeneration was much higher than in steady-state spermatogenesis.

We conclude that the As, Apr, and Aal spermatogonial subpopulations markedly change their behavior during regeneration, thus enabling a quicker recovery of the stem cell pool.

Extending the “As model.” Our results demonstrate a variety of different properties of the As, Apr, and Aal subpopulations that constitute the stem cell compartment (Fig. 6). Differentiation does not follow a strictly linear process where gene expression is coupled to lineage (cyst length); rather, it includes multiple pathways along the two parameters. For example, NGN3+ Aal-4 can be generated either by division of NGN3+ Apr or by gain of NGN3 expression in GFRα1+ Aal-4.

Fig. 6

Proposed spermatogonial subpopulations and their behavior with respect to their morphology (cyst length) and gene expression (GFRα1+ and NGN3+). Black arrows indicate the proposed flow of the majority of cells in each morphological group. Dashed lines show the observed modes of “reversion.” Arrows without asterisks were actually observed; those with asterisks were not observed but are proposed to occur with high probability.

This model proposes a number of important extensions to the As model (Fig. 1A). First, regardless of cyst length, the NGN3+ subpopulations including As are destined for differentiation. Thus, not all As cells act equivalently as stem cells. Second, Apr and Aal spermatogonia are not committed unidirectionally to differentiation but are capable of reverting to As or shorter cysts by clone fragmentation. In some cases this may also be accompanied by reversion in gene expression. Clone fragmentation has been previously proposed in mice (11, 26) and primates (27, 28) and demonstrated in Drosophila (4, 5), and we have directly observed this event by live imaging in vivo. Third, of the two parameters, gene expression appears to be the better indicator of the fate of individual cells over the cyst length. For example, GFRα1+ cells transform directly into KIT+ spermatogonia only occasionally (fig. S4), and eventually all the KIT+ spermatogonia are generated from the NGN3+ population, regardless of their cyst length. On the basis of increasing transformation frequency into A1 spermatogonia along the cyst length, it has been proposed that the differentiation potential increases gradually (Fig. 1A) (29). Our data suggest that this is a reflection of the increasing content of NGN3+ cells.

Implications for the definition of a stem cell. It has been generally assumed that As cells represent the entire spermatogenic stem cell population and that they support both steady-state spermatogenesis and regeneration after tissue damage or transplantation (9, 12, 30). However, our findings challenge this assumption.

We have demonstrated that NGN3+ cells can revert to being GFRα1+ As cells and can even act as long-lasting stem cells. The frequency of these events increases when tissue is damaged and regeneration takes place. This suggests that these “differentiating” cells can act as “potential stem cells,” defined as cells that do not normally self-renew during steady-state spermatogenesis but nonetheless retain latent self-renewing potential (6, 31). Live imaging of clone fragmentation suggests that the potential stem cells include not only the NGN3+ As cells, but also Apr and Aal (either NGN3+ or GFRα1+) cells. Indeed, the NGN3+ cell population, most of which are present in connected cysts, exhibits colony formation and contribution to regeneration (6).

On the other hand, the population of GFRα1+ As cells, which is positioned at the top of both hierarchies, is best related to the “actual stem cells” that support normal steady-state spermatogenesis (6, 31). In the pulse-labeling experiment in undisturbed testes, however (Figs. 2 and 3), the absolute number of GFP-tagged GFRα1+ As cells (which were originally NGN3+ at the time of labeling) decreased by day 20. This was still larger than the number of persistent patches observed 3 months after the pulse, which represent the long-lasting stem cells (fig. S6). Similarly, during regeneration after damage, the number of persistent patches, which was significantly larger than that found in steady-state spermatogenesis, was smaller than the count of GFP-tagged GFRα1+ As cells (fig. S6). Thus, only a part of the GFRα1+ As population persists for a long period to serve as functional stem cells, and a majority of this population do not self-renew continually and may act as “potential stem cells” as well. The recent report of colony-forming activity in a small fraction of KIT+ spermatogonia (8) may also represent “potential stem cells,” which could reside in the minor populations of GFRα1+ KIT+ spermatogonia (fig. S4) or in the populations of single and paired KIT+ spermatogonia that have been converted directly from NGN3+ As and Apr (Fig. 3C).

The concept of a single homogeneous cell population serving both steady-state spermatogenesis and regeneration is unnecessarily constraining. Rather, cells seem to possess a variable level of potential to act as stem cells. How their potential is manifested can be greatly influenced by the state of the tissue. Steady-state spermatogenesis favors the GFRα1+ As population, which may have the greatest potential, whereas regeneration after transplantation or damage relies on the NGN3+ and cyst (Apr and Aal) spermatogonia as well, whose potential seems to be lower.

Regulation of the stem cell compartment. The molecular mechanisms governing the transition between the GFRα1+ and NGN3+ populations have yet to be defined. In this regard, the finding that GDNF (glial cell line–derived neurotrophic factor), the ligand for GFRα1, regulates the spermatogonial expression of GFRα1 and NGN3 in a reciprocal manner (i.e., positively and negatively, respectively) (32) suggests that GDNF may be an important determinant of stem cell behavior. Indeed, Sertoli cells, an essential player for the maintenance of the spermatogenic stem cell population, express GDNF, which is essential for the long-term maintenance of spermatogenic stem cell activity in vivo (19) and crucial for spermatogonial cultures to maintain the ability to form colonies after transplantation (33, 34).

Comparison with other stem cell systems. In Drosophila, male and female germ cells appear to differentiate along a linear pathway with respect to lineage and gene expression. Their differentiation state is geographically recapitulated in the polarized gonad as a consequence of localized specialized supporting cells and extracellular factors that control self-renewal and differentiation (3). In contrast, in mouse testis, the stem cell compartment (Fig. 6) does not appear to be spatially constrained. Rather, while showing biased localization to the blood vessels and interstitium (24), the As, Apr, and Aal spermatogonia are intermingled among more differentiating germ cells and seemingly uniform supporting Sertoli cells. Given these anatomical differences, it is not surprising that flies and mice exhibit distinct controlling mechanisms.

Within mammals, the primate stem cell compartment appears to differ from that in mice. In primates, primitive spermatogonia, referred to as Adark and Apale according to nuclear morphology, are generally assumed to represent the stem cell pool (11, 28). The Adark cells are the presumptive reserve stem cells that rarely proliferate; the Apale cells represent the active stem cell pool and are continuously cycling. Whereas NGN3+ spermatogonia proliferate actively, the cell cycle status of GFRα1+ population is yet to be elucidated. It is an intriguing question whether there exists a reserve population of GFRα1+ cells in mice, equivalent to the Adark cells in primates, or whether a reserve population of stem cells is unique to primates.

The biological importance of the syncytial nature of spermatogonial proliferation across animal species remains a mystery. Nonetheless, it provides a powerful tool to monitor gene expression in the context of cell lineage. In other stem cell systems, especially in mammals, stem and progenitor cell compartments are often classified on the basis of gene expression and location; correlation of lineage and gene expression has generally not been feasible. Our study demonstrates that lineage is not strictly and linearly correlated with gene expression and that there may be multiple and reversible paths from stem cells to differentiation in other systems.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S7

Movies S1 to S4


  • * Present address: Department of Immunobiology and Hematology, Institute for Frontier Medical Sciences, Kyoto University, Kyoto 606-8507, Japan.

References and Notes

  1. We thank J. Miyazaki, T. Noce, and A. Imura for materials; T. Ogawa, R. Sugimoto, Y. Kitadate, K. Hara, and H. Mizuguchi-Takase for comments; T. Fujimori for discussion and technical advice; and M. Sukeno for technical assistance. Supported by a Grant-in-Aid for Scientific Research (KAKENHI) on Innovative Areas, “Regulatory Mechanism of Gamete Stem Cells” (S.Y.), NICHD/NIH Contraceptive Development Research Centers Program grant U54 HD4254 (R.E.B.), the Uehara Memorial Foundation, and the Naito Foundation.
View Abstract

Navigate This Article