Maintenance of retinal stem cells by Abcg2 is regulated by notch signaling

One of the emerging universal characteristics of stem cells or progenitors is the preferential expression of ABCG2, a half transporter belonging to ATP-binding cassette (ABC) superfamily of transmembrane proteins (Doyle and Ross, 2003). These transporters mediate efflux of a broad spectrum of substrates including cytotoxic drugs, peptides, steroids, ions and phospholipids (Krishnamurthy and Schuetz, 2006). Their ability to preferentially exclude the DNA-intercalating dye, Hoechst 33342, is regarded to be the molecular basis for the enrichment of stem cells as side population (SP) cells by the Hoechst dye efflux assay (Goodell et al., 1996; Abbott, 2003; Doyle and Ross, 2003; Bunting, 2002). The assay involves incubation of cells with the Hoechst dye followed by dual-wavelength fluorescence-activated cell sorting (FACS), which identifies a small side population of cells with low dye accumulation, possessing stem cell properties and potential (Goodell et al., 1996). The notion that SP cell phenotype is mediated by ABCG2 was supported by observations that the perturbation of Abcg2 expression by retrovirus-mediated transduction or gene deletion affected the number of SP cells present in the hematopoietic compartment (Kim et al., 2002; Scharenberg et al., 2002; Zhou et al., 2002).

The conservation of Abcg2 expression in SP cells, enriched from a wide range of tissues including blood (Zhou et al., 2003), muscle (Meeson et al., 2004), heart (Martin et al., 2004), gonad (Lassalle et al., 2004), lung (Summer et al., 2003), intestine (Staud and Pavek, 2005) and cornea (Budak et al., 2005) suggests that ABCG2 has an important role in stem cells. Molecular studies carried out on the hematopoietic compartment have begun to shed light on the involvement of Abcg2 in stem cells. For example, the highly regulated expression of Abcg2 during hematopoiesis, i.e. high levels of Abcg2 transcripts in stem cells and their sharp decline during lineage commitment, and the blocking of lineage commitment in response to ectopic expression of Abcg2 indicates a role for ABCG2 in the maintenance of hematopoietic stem cells (Zhou, S. et al., 2001; Scharenberg et al., 2002).

The expression of Abcg2 and SP cell phenotype are also characteristics of neural stem cells, derived from different brain regions, including the retina (Hulspas and Quesenbury, 2000; Bhattacharya et al., 2003; Ahmad et al., 2004; Mouthon et al., 2006). However, little is known about the role of Abcg2 in neural stem cells and their lineage commitment. In addition, we do not know how Abcg2 expression is linked with the maintenance of stem cells. We have addressed these issues in developing retina, a suitable and accessible model of the central nervous system, where seven different types of cells are generated in an evolutionarily conserved temporal sequence by multipotential retinal stem cells (Rapaport et al., 2004). Our results demonstrate a regulated expression of Abcg2 during retinal histogenesis; levels of Abcg2 transcripts decrease with the onset of lineage commitment. This decrease in Abcg2 expression correlates with the progressive depletion of SP cell population as retinal histogenesis ensues. When Abcg2 is overexpressed in retinal progenitors, using retrovirus-mediated transduction, the differentiation is blocked, accompanied by an increase in the expression of stem cells markers and SP cell phenotype. By contrast, siRNA-mediated silencing of Abcg2 expression in retinal progenitors depletes SP cell population and promotes differentiation. In addition, we observed that Abcg2 expression and SP cell phenotype are influenced by Notch signaling, a key regulator of retinal stem cells. Our results demonstrate for the first time that ABCG2 is a downstream target of Notch signaling. Together, these observations suggest that Abcg2 participates in the maintenance of retinal stem cells or progenitors, under the regulation of Notch signaling.

To examine the involvement of ABCG2 in the regulation of retinal progenitors, we first studied the temporal pattern of expression during retinal development. Retinal cells are born in an evolutionarily conserved temporal sequence; retinal ganglion cells, cone photoreceptors, horizontal cells and the majority of amacrine cells are born during early histogenesis whereas rod photoreceptors, bipolar cells and Müller glia are generated during late histogenesis (Rapaport et al., 2004). We examined Abcg2 expression during late histogenesis for the following reasons: first, because rod photoreceptors constitute ∼80% of all cells in the rodent retina, the majority of retinal cells differentiate during this stage, and second, the number of progenitors is significantly higher at this stage, compared with those in early histogenesis where tissue is a limiting factor for SP cell analysis. RT-PCR analysis revealed an inverse correlation between expression of Abcg2 transcripts and the process of histogenesis; levels of Abcg2 transcripts decrease at each stage studied (Fig. 1A,B). The levels of Abcg2 transcripts were lowest at PN9, when the generation of ∼95% of all retinal cell types have been completed (Rapaport et al., 2004). Such a temporal pattern of expression suggested that Abcg2 expression is associated with proliferating progenitors and the attenuation in Abcg2 expression reflected a decrease in the number of such cells as differentiation proceeded. To test this premise, we examined the spatial and cell-specific expression of Abcg2 in postnatal day 1 (PN1) retina. Immunohistochemical examination of retinal sections revealed that ABCG2 immunoreactivities were predominantly distributed in the outer neuroblastic layer, where progenitors and precursors are preferentially localized in the developing retina (Fig. 1C,D). To determine whether or not ABCG2-expressing cells are proliferating, cells from PN1 retina were dissociated, chased with BrdU and subjected to double immunocytochemical analysis. We observed ABCG2⁺ cells, expressing BrdU immunoreactivity, demonstrating their proliferative nature (Fig. 1E,F).

Next, we examined the temporal association of Abcg2 expression with SP cell phenotype of retinal progenitors. We have previously demonstrated that both early and late retinal progenitors display SP cell phenotype (Bhattacharya et al., 2003; Ahmad et al., 2004). We have also demonstrated that SP cell phenotype was retina specific and not due to contamination with blood during the isolation of retinal cells (Bhattacharya et al., 2003). Since SP cell phenotype is dependent on Abcg2 expression (Zhou et al., 2002), we were interested to know whether a temporal decline in Abcg2 expression was correlated with a decrease in SP cells during late retinal histogenesis. We carried out Hoechst dye efflux assay on fresh retinal dissociates obtained from PN1, PN3 and PN6 rats and observed that the proportion of SP cells progressively decreased over time such that at PN6 only ∼0.002% of cells could be characterized as SP cells, compared with ∼0.1% SP cells at PN1 (Fig. 2A-D,E). To determine the phenotypes of SP and non-SP (NSP) cells during late histogenesis we examined the expression of transcripts corresponding to cell-specific markers (Fig. 2F). Embryonic day (E)18 retinal cells were examined as the representative stage. We observed that SP cells predominantly expressed progenitor-specific transcripts, i.e. Nestin, Notch1, Pax6 and Sox2, in addition to those corresponding to Abcg2. These transcripts were detected at relatively low levels in NSP cells. The NSP cells expressed retinal cell-type-specific transcripts, i.e. rhodopsin kinase (rod photoreceptors), mGluR6 (bipolar cells) and Brn3b (retinal ganglion cells). Such transcripts were not detected in SP cells. These observations suggested that SP cells are uncommitted retinal progenitors. The association of Abcg2 expression with retinal SP cells and the progressive decline in the SP cell population in parallel with the decrease in Abcg2 expression suggested that SP cell phenotype is due to Abcg2. To test this notion further we examined the effect of perturbation of Abcg2 expression on SP cell phenotype of retinal progenitors. Late retinal progenitors were enriched as neurospheres from E18 retina (E18 neurospheres) and transduced with Abcg2 retrovirus vector (G1-ABCG2) to overexpress ABCG2, followed by Hoechst dye efflux assay to enrich SP cells (Bhattacharya et al., 2003). Since the gain-of-function approach is thought to carry a risk of non-specific results (Zhou, C. et al., 2001; Zhou et al., 2002), we examined the effects on SP cell phenotypes and maintenance of progenitors in response to acute decrease in Abcg2 expression, using siRNA-mediated gene silencing. To attenuate Abcg2 expression, E18 neurospheres were transfected with pSuper expression vector containing Abcg2 siRNA sequence (pSuper-ABCG2_siRNA). Controls included neurospheres transduced with either empty retrovirus or transfected with pSuper containing scrambled Abcg2 siRNA sequence. We observed a remarkable increase (∼14-fold) in the proportion of SP cells in Abcg2 retrovirus-transduced neurospheres, compared with controls (14.5±3.7 vs 0.9±1.9; P<0.001) (Fig. 3A-C). By contrast, there was a significant decrease (∼tenfold) in the proportion of SP cells in neurospheres expressing Abcg2 siRNA, compared with controls (0.1±0.19 vs 1.1±0.46; ^***P<0.001) (Fig. 3D-F). The specificity of siRNA-mediated attenuation in gene expression was demonstrated by a decrease in the levels of Abcg2 transcripts (Fig. 3G), ABCG2 protein in western blot analysis (Fig. 3H) and immunocytochemical analysis (Fig. 3I-L) compared with controls. Together, these observations demonstrated that Abcg2 expression is developmentally regulated and is the molecular determinant of SP phenotype of retinal stem cells during late histogenesis.

The expression of Abcg2 in proliferating progenitors and its temporal decrease during histogenesis suggested that ABCG2 is involved in the maintenance of these cells. Maintenance of stem cells entails the maintenance of progenitor properties and the uncommitted state. To determine the influence of Abcg2 on the retinal progenitor properties, we examined their proliferation and the expression of pan neural (Nestin) and retinal (Pax6) stem-cell-specific markers in response to perturbation of Abcg2 expression in E18 neurospheres, as described above. Neurospheres were cultured in proliferating conditions for 4 days following transduction of G1-ABCG2 virus and transfection of pSuper-ABCG2_siRNA. Neurospheres were exposed to BrdU in the last 24 hours, to tag proliferating cells, followed by immunocytochemical and RT-PCR analyses of cell-specific markers. We observed a significant increase in the proportion of BrdU⁺ cells expressing Nestin or Pax6 immunoreactivities in Abcg2 retrovirus-transduced neurospheres, compared with controls (BrdU⁺ Nestin⁺ cells: 79.21±4.6 vs 46.9±2.8; BrdU⁺ Pax6⁺ cells: 73.45±3.7 vs 38.8±3.2; P<0.001) (Fig. 4A-H,I). By contrast, there was a significant decrease in the proportion of BrdU⁺ cells expressing Nestin or Pax6 immunoreactivities in neurospheres transfected with pSuper-ABCG2_siRNA, compared with controls (BrdU⁺ Nestin⁺ cells: 22.24±2.8 vs 55.81±4.6; BrdU⁺ Pax6⁺ cells: 19.72±1.9 vs 47.35±3.2, P<0.001) (Fig. 5A-H,I). RT-PCR analysis carried out on similarly treated and cultured neurospheres revealed an increase and decrease in levels of transcripts corresponding to Nestin, Pax6 and Sox2, in Abcg2 retrovirus-transduced and siRNA-treated neurospheres, respectively, compared with controls, thus corroborating the immunocytochemical results (Fig. 4 and Fig. 5J). Together, these results suggested that Abcg2 expression contributes towards the maintenance of retinal progenitor properties.

One of the mechanisms for the maintenance of retinal progenitors is to keep them from lineage commitment. Therefore, we were interested to determine whether Abcg2 expression has a negative influence on the differentiation of retinal progenitors. We addressed this question by examining the effects of perturbation of Abcg2 expression on pan neural differentiation of retinal progenitors. E18 retinal progenitors, enriched as neurospheres, were transduced with G1-Abcg2 retrovirus or transfected with pSuper-ABCG2_siRNA as described above. Instead of culturing them in proliferating conditions, neurospheres were shifted to differentiating conditions (1% serum, minus mitogen) and cultured for 5 days, followed by immunocytochemical and RT-PCR analyses of the expression of neuronal (Map2) and glial (GFAP) specific markers. In both experimental and control groups a subset of cells was detected expressing Map2 or GFAP immunoreactivities. However, there was a significant decrease in the proportion of Map2⁺ as well as GFAP⁺ cells in Abcg2 retrovirus-transduced neurospheres, compared with controls (Map2⁺ cells: 73.27±3.2 vs 41.23±2.8, P<0.001; GFAP⁺ cells: 41.05±2.3 vs 20.21±2.8; P<0.001) (Fig. 6A-H,M). By contrast, the proportion of Map2⁺ and GFAP⁺ cells increased significantly in siRNA vector-transfected neurospheres, compared with controls (Map2⁺ cells: 63.1±3.7 vs 81.01±4.6, P<0.001; GFAP⁺ cells: 32.05±2.8 vs 54.19±3.61; P<0.001) (Fig. 7A-H,M). Examination of levels of transcripts corresponding to Map2 and Gfap revealed that their levels decreased and increased in neurospheres transduced with Abcg2 and transfected with siRNA, respectively, compared with controls, corroborating immunocytochemical results (Fig. 6 and Fig. 7N). Together, these observations suggested that ABCG2 has a negative influence on pan neural differentiation of retinal progenitors. Abcg2 expression influenced the differentiation of retinal progenitors along specific retinal lineage as demonstrated by a decrease and increase in levels of transcripts corresponding to rod photoreceptors (Grk1, rhodopsin kinase), bipolar cells (mGluR6), amacrine cells (Stx1a, syntaxin1) and Müller glia (Glul, glutamine synthetase) in neurospheres overexpressing Abcg2 (Fig. 6O) and those treated with siRNA (Fig. 7O), respectively, compared with controls. To determine whether the effects of perturbation of ABCG2 expression on differentiation is reflected in changes in progenitor populations, we examined the proportion of cells expressing immunoreactivities corresponding to Nestin in differentiating conditions. In an inverse relationship to cells expressing pan neural differentiation markers, the proportion of cells expressing Nestin and transcripts increased and decreased in neurospheres transduced with Abcg2 retrovirus (Fig. 6I-L,M,N) and treated with Abcg2 siRNA (Fig. 7I-L,M,N), respectively, suggesting that Abcg2 influenced differentiation at the levels of progenitors.

Next, to determine the mechanism underlying the expansion of SP cell populations in response to the ectopic expression of Abcg2, we examined the expression of Mash1, a proneural gene, whose transient expression characterizes the proliferating intermediate retinal precursors during late retinal histogenesis (Ahmad et al., 1998). Levels of Mash1 transcripts along with those corresponding to cell cycle regulators, cyclin D1, P27^kip1 and Ki67, and the pan neuronal marker, Map2, were determined in SP and NSP cells in E18 neurospheres transduced with G1-Abcg2 or empty retrovirus. Mash1 transcripts were detected in SP and NSP cells in both conditions (Fig. 8). However, there was an increase in levels of Mash1 transcripts in SP cells in G1-Abcg2-transduced neurospheres, compared with those in controls, and the increase was at the expense of levels of transcripts in NSP cells. A similar change in the expression of cyclin D1, the major D type cyclin found in retinal progenitors (Sicinski et al., 1995; Fantl et al., 1995) was observed; levels of Ccnd1 transcripts increased in SP cells, enriched from G1-Abcg2-transduced neurospheres, compared with those in controls, at the expense of those in NSP cells. Transcripts producing P27^kip1, a cyclin kinase inhibitor that negatively regulates retinal progenitor proliferation (Ohnuma et al., 1999; Levine, 2000; Dyer and Cepko, 2001) were detected only in NSP cells and their levels decreased in NSP cells enriched from Abcg2-transduced neurospheres, compared with those in controls. By contrast, transcript producing Ki67, a nuclear protein characteristic of cells in S phase (Tirelli et al., 2002) was detected only in SP cells but their levels were higher in those overexpressing Abcg2 than in controls. Together, these observations suggested that Abcg2 favored the maintenance of proliferating progenitors over committed or differentiating precursors.

One of the key regulators of retinal stem cells is Notch signaling, which maintains a pool of stem cells throughout histogenesis by keeping them uncommitted. The fact that ABCG2 inhibited the differentiation of retinal stem cells, while maintaining progenitor properties, suggested that Abcg2 is one of the downstream targets of Notch signaling. We tested this premise as follows: first, we studied the effects of perturbation of Notch signaling on Abcg2 expression and retinal progenitor properties. To activate Notch signaling, E18 neurospheres were transduced with retrovirus expressing Notch intracellular domain (NICD) and cultured in proliferating conditions for 4 days. Neurospheres were exposed to BrdU in the last 24 hours of culture to tag proliferating progenitors, followed by immunocytochemical and RT-PCR analyses of Abcg2 expression. NICD, a product of activated Notch receptor obtained through the proteolytic activity of γ-secretase, mediates the CSL-dependent activation of Notch target genes (Mumm and Kopan, 2000). Therefore, overexpression of NICD leads to constitutive activation of Notch signaling. To attenuate Notch signaling, a batch of neurospheres was cultured under conditions promoting proliferation in the presence of DAPT and examined as described above. DAPT is an inhibitor of γ-secretase and thus attenuates Notch signaling by antagonizing the release of NICD (James et al., 2004). We observed a significant increase in the proportion of BrdU⁺ cells expressing ABCG2 in NICD-transduced neurospheres, compared with controls (39.52±2.3 vs 20.17±1.9; P<0.001) (Fig. 9A-D,G). By contrast, the proportion of such cells decreased in DAPT-treated neurospheres, compared with controls (20.17±1.9 vs 17.11±1.9; P<0.05) (Fig. 9E-G). To ascertain that these results were not non-specific but related to changes in progenitor properties in response to Notch signaling, we examined the expression of Nestin, Sox2 and Pax6 transcripts. We observed that the changes in levels of Abcg2 transcripts in response to perturbation in Notch signaling were accompanied by similar trends in levels of Nestin, Sox2 and Pax6 transcripts (Fig. 9H). The specificity of perturbation in Notch signaling was demonstrated by an increase and decrease in levels of Notch target gene, Hes1 in NICD retrovirus-transduced and DAPT-treated neurospheres, respectively, compared with controls (Fig. 9H).

Next, we examined the effects of perturbation of Notch signaling on SP cell phenotype of retinal progenitors. We subjected neurospheres transduced with NICD retrovirus or treated with DAPT to the Hoechst dye efflux assay. We observed that the proportion of SP cells increased ∼sixfold in neurospheres transduced with NICD retrovirus compared with those in untreated controls (Fig. 10A,B,D). By contrast, the proportion of SP cells was remarkably decreased in DAPT-treated neurospheres, compared with controls (Fig. 10A,C,D). The majority of cells in DAPT-treated neurospheres consisted of NSP cells. Together, these observations suggested that Notch signaling regulated expression of ABCG2 and the SP cell phenotype associated with it.

Next, we wanted to know whether or not Notch signaling directly influenced the expression of the Abcg2 gene. A prerequisite for direct activation by Notch signaling is the co-expression of Notch and Abcg2 in retinal progenitors. Double immunocytochemical analysis, carried out on E18 retinal progenitors revealed that Notch1 and ABCG2 immunoreactivities were co-localized (Fig. 11A-D). Notch signaling regulates its target genes through NICD-CSL mediated transcriptional activation. Some of the best-known Notch target genes are bHLH transcription factors belonging to hairy enhancer of split class, Hes1 and Hes5. These genes contain CSL-binding elements in their promoters, which mediate the action of CSL-NICD complex. Examination of Abcg2 promoter revealed the presence of multiple CSL-binding elements in its proximal promoter suggesting that Abcg2 is one of the Notch target genes (Fig. 11E). To test this possibility we first tested whether or not the Abcg2 CSL-binding element is involved in Notch-mediated transcriptional activation. Abcg2 promoter reporter constructs containing serial deletion of Abcg2 promoter sequence (Bailey-Dell et al., 2001) (Fig. 11E), were transfected in 293T cells, transduced with NICD retrovirus, followed by the analysis of luciferase activities. We observed a significant increase in luciferase activities (∼3000-fold) in cells transfected with –1285/+362 constructs that contained CSL-binding sites compared with those transfected with the rest of the constructs that did not and demonstrated low luciferase activities (Fig. 11F). Second, we carried out a mobility shift assay to examine the ability of Abcg2 promoter sequence containing the CSL-binding element to interact with nuclear factor(s) obtained from retinal progenitors. When the labeled wild-type oligonucleotide was incubated with nuclear extracts complexes were formed whose specificity could be shown by competitive inhibition of complex formation when an excess of unlabeled wild-type oligonucleotide was included in the reaction (Fig. 11G). Complex formation was not observed when the reaction was performed with mutated oligonucleotide. To test the specificity of the complex formation, a mobility shift assay was carried out with in vitro translated CSL protein, which showed a complex formation similar to that obtained with nuclear extract obtained from retinal progenitors. Finally, to gain an insight into CSL-mediated regulation of Abcg2 in retinal progenitors in vivo, we carried out chromatin immuno precipitation (ChIP) on neurospheres maintained in proliferating conditions. The CSL antibody immunoprecipitated a nucleosomal complex that contained DNA sequence, amplifiable by Abcg2 promoter-specific primers, suggesting that CSL is part of a transcriptional complex that regulates the activities of the ABCG2 promoter in vivo (Fig. 11H). Together, these observations suggest that the CSL-binding element, through which Notch activates its target genes, plays a role in the activation of Abcg2 expression.

The ubiquitous association of Abcg2 expression with stem cells and their function in the efflux of a broad spectrum of substrates ranging from cytotoxic drugs to small molecules of differentiation suggest a role in the maintenance of homoeostasis of these progenitor cells. This notion led us to a hypothesis that the expression of Abcg2 may constitute an integral part of the cellular mechanism that maintains retinal stem cell populations, and as such it is under the regulatory network of intercellular pathways that keep stem cells self renewing and uncommitted. The first part of the hypothesis, i.e. its role in the maintenance of retinal stem cells, is supported by temporal expression patterns of Abcg2 during late histogenesis. We demonstrate that, as observed during the lineage commitment of hematopoietic stem cells (Zhou, S. et al., 2001), the expression of Abcg2 is downregulated during the differentiation of retinal stem cells. This phenomenon is accompanied by a decrease in the proportion of SP cells at each successive stage of retinal histogenesis, suggesting that Abcg2 is the regulator of retinal stem cells, which is characterized by SP cell phenotype (Bhattacharya et al., 2003). This premise was further tested by perturbation of the expression, which demonstrated that the overexpression of Abcg2 expands retinal SP cells and at the same time attenuates the differentiation of retinal stem cells. By contrast, a targeted decrease in Abcg2 expression depletes retinal SP cells and promotes the differentiation of retinal stem cells. Taken together, these observations suggest that Abcg2 expression influences the maintenance of retinal stem cells. These observations including those that demonstrated the involvement of ABCG2 in lineage commitment in the hematopoietic compartment appear at odds with the lack of abnormalities in mice lacking a functional Abcg2 gene (Zhou et al., 2002). Given the large ABC transporter family and the evidence that transporters such as Abcg2 and ABCB1 are co-expressed in stem cells (Zhou et al., 2003; Lin et al., 2006), the normal phenotype of Abcg2-knockout mice may be due to efficient functional compensation. Such functional compensation by ABC transporters is known to occur in the hematopoietic compartment. For example, the lack of effects of the absence of murine ABCB1 genes on SP cell phenotype is regarded to be due to functional compensation by ABCG2 (Zhou, S. et al., 2001; Zhou et al., 2002).

The test of the second part of the hypothesis required the demonstration of interactions between ABCG2 and pathways that regulate the self-renewal and state of commitment of retinal stem cells. Prominent among these pathways are those mediated by Frizzled (Wnt signaling), Notch (Notch signaling), Patched (Shh signaling) and c-Kit (c-Kit signaling) (Ahmad et al., 2004). Notch signaling plays an important role in the maintenance of retinal stem cells. Attenuation in Notch signaling promotes differentiation of retinal ganglion cells (Austin et al., 1995; Ahmad et al., 1997) and rod photoreceptors (Jadhav et al., 2006) during early and late histogenesis, respectively. The temporal and spatial aspects of Abcg2 and Notch1 expression during retinal histogenesis betrays their possible interactions; levels of transcripts from both Abcg2 and Notch1 (Ahmad et al., 1995) decrease with lineage commitment and immunoreactivities corresponding to ABCG2 and Notch1 are co-localized in retinal stem cells. Our observations suggest that Abcg2 is one of the targets of Notch signaling; like Hes1 and Hes5, it is regulated through NICD-CSL-mediated transcriptional activation. The functional consequence of such regulation is demonstrated by changes in the levels of Abcg2 expression and the proportion of SP cells in response to perturbation of Notch signaling. These changes are accompanied by changes in the differentiation status of retinal stem cells.

An interesting question remains as to how the expression of ABCG2 expands SP cells and keeps them undifferentiated. There is no definite answer as yet but it could be suggested that part of the increase in SP cells, observed in response to overexpression of ABCG2, might be at the expense of NSP cells. The NSP, besides containing post mitotic cells, includes proliferating precursors in different stages of differentiation. The developing retina contains committed precursors characterized by their ability to divide and expression of proneural genes, such as Mash1 (Ahmad et al., 1998). These precursors are malleable and when exposed to conditions that promote the maintenance of retinal stem cells rather than their differentiation they revert to a progenitor stage (Ahmad et al., 1998). The fact that the expression of Mash1 increases in SP cells in neurospheres transduced with ABCG2 retrovirus compared with those in controls suggests that the Mash1-positive precursors are recruited as SP cells in response to overexpression of ABCG2 (Fig. 11). Indeed, the selective proliferation of ABCG2-expressing SP cells should also be taken into account for the expansion of retinal SP cells. The classic mechanism of the maintenance of stem cells involves the influence of the genome to promote cell proliferation, as in the case of Wnt signaling, and repress cell commitment, as in the case of Notch signaling (Ahmad et al., 2004). ABCG2, whose apparent function in stem cells is to protect them from toxins and maintain the homeostasis, might prevent cell commitment by actively extruding extrinsic small molecules that promote differentiation. Additionally, it might prevent differentiation by actively extruding key components of the differentiation-promoting signaling pathway, thus impairing it. Evidence has emerged that disparate signaling pathways are recruited for the maintenance of stem cells or progenitors and it is quite likely that disparate mechanisms, one involving an ABCG2 pump, are used to keep stem cells uncommitted to maintain their pool.

Embryos were harvested from timed pregnant Sprague-Dawley rats (SASCO), and the gestation day was confirmed by the morphological examination of embryos (Christie, 1964). Embryos were harvested in Hank's balanced salt solution (HBSS). Eyes were enucleated, and retinas were dissected and dissociated into single cells as previously described (Bhattacharya et al., 2003). Retinal dissociates were cultured in retinal culture medium or RCM (DMEM-F12, 1× N2 supplement, 2 mM L-glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin) containing 20 ng/ml EGF for 5 days to generate clonal neurospheres. The neurospheres were collected, and plated on poly-D-lysine-coated and laminin-coated glass coverslips and exposed to 5-bromo-2′-deoxyuridine (BrdU) (10 μM) for the final 24 hours. Neurospheres were frozen for RT-PCR or fixed in 4% ice-cold paraformaldehyde for immunocytochemical analysis. To induce differentiation, neurospheres were washed extensively to remove BrdU and mitogen, and cultured in RCM and 1% FBS for a further 5 days, followed by collection for RT-PCR and immunocytochemical analysis, as described above.

Retinal cells, from different stages of histogenesis were resuspended in Hoechst IMDM (10⁶ cells/ml) containing 2% FCS at 4°C overnight followed by staining with Hoechst 33342 (2.0 μg/ml) at 37°C for 30 minutes. Cells were sorted on a FACStar Plus (BDIS) cell sorter. The Hoechst dye was excited at 350 nm and its fluorescence was measured at two wavelengths using a 485 BP22 (485/22 nm band-pass filter) and a 675 EFLP (675nm long-pass edge filter) optical filter (Omega Optical, Brattleboro, VT). A 610 nm short-pass dichroic mirror was used to separate these emission wavelengths (Omega Optical). The SP region was defined on the cytometer on the basis of its fluorescence emission in both blue and red wavelengths. Side population (SP) sorting gates were established after collecting 5×10⁵ events within this live gate.

Detection of cell-specific markers and BrdU was performed as previously described (Bhattacharya et al., 2003). Briefly, paraformaldehyde-fixed cells were incubated in 1× PBS containing 5% NGS and 0, 0.2 or 0.4% Triton X-100 followed by an overnight incubation in antibodies against Nestin (DSHB), Pax6 (DSHB), ABCG2 (Chemicon), Map2 (Chemicon), GFAP (Sigma), Notch (Ahmad et al., 1995) and BrdU at 4°C. Cells were examined for epifluorescence after incubation in IgG conjugated to cyanin 3 (Cy3)/FITC. Images were captured using a cooled CCD camera (Princeton Instruments, Trenton, NJ) and Openlab software (Improvision, Lexington, MA). To determine the percentage of specific cell types in a particular condition, the number of BrdU⁺ and cell-specific antigen-positive cells was counted in 10-15 randomly selected fields in two to three different coverslips. Each experiment was repeated at least three times. Values are expressed as means ± s.e.m. Data were analyzed using the Student's t-test to determine the significance of the differences between various conditions.

Total RNA was isolated using a Qiagen RNA isolation kit. 2 μM cDNA was amplified using gene-specific primers by using the following step cycle program on a Robocycler (Stratagene): denaturation at 94°C for 30 seconds, annealing at specific temperature (see supplementary material Table S1) for 35 seconds, extension at 72°C for 40 seconds for 35 cycles followed by a final extension at 72°C for 5 minutes. PCR products were resolved on 2% agarose gels against 100bp DNA marker (MBI Fermentas). Gene-specific primers were used as listed in supplementary material Table S1.

Late retinal progenitors were enriched as neurospheres from E18 retina (=E18 neurospheres) and were transduced with ABCG2 retrovirus vector (G1-ABCG2) to overexpress ABCG2. After transduction, the medium was replaced, and neurospheres were cultured in proliferating conditions for 4 days and exposed to BrdU in the last 24 hours, to tag proliferating cells, followed by immunocytochemical and RT-PCR analyses of cell-specific markers.

For targeted silencing of Abcg2 expression, ABCG2 siRNA was cloned into pSuper vector (pSuper-ABCG2_siRNA) according to the manufacturer's (Oligogene) protocol. Neurospheres in proliferating conditions were co-transfected with pSuper-ABCG2_siRNA and pEGFP-C3 (to determine transfection efficiency) using Fugene according to the manufacturer's (Roche) protocol. A batch of neurospheres was similarly transfected with pSuper-ABCG2_missense, containing scrambled sequence, as controls. The efficiency of gene silencing was determined by immunocytochemical and RT-PCR analyses of ABCG2 protein and transcripts, respectively.

HEK293T cells were plated onto 100 mm Petri dishes at a density of 8-10×10⁶ cells. Cells were co-transfected with a series of Abcg2 promoter constructs (Bailey-Dell et al., 2001) and NICD expression construct (James et al., 2004) using a calcium phosphate-based protocol. Transfection efficiency was examined by co-transfecting cells with pGFP-C3 (Clonetech). For luciferase assay, cells were lysed in 1× reporter lysis buffer (Promega) and 100 μl lysate was diluted five times using assay reagent (Promega). Diluted samples (100 μl) were analyzed for luciferase activities using a luminometer (Pharmingen). Mean fold change for the respective deletion constructs in the experimental groups was calculated with respect to controls (pGL3-basic). Values are expressed as mean fold change ± s.e.m. from three different experiments. Statistical analysis was performed using the Student's t-test to determine the significance between different conditions.

EMSA was carried out as previously described (Ahmad et al., 1995). Briefly, oligonucleotides corresponding to Abcg2 promoter sequence containing the wild-type (GGTAGATGTTGGGATGGCTAC) or mutated (GGTAGATGTGAGATTGGCTAC) CSL binding site were end-labeled with [γ-³²P]ATP using T4 polynucleotide kinase (New England Biolabs). The binding reaction was performed in 20 μl buffer containing 20 mM Tris-HCl (pH 8.0), 75 mM KCl, 5% glycerol, 50 μg/ml bovine serum albumin (BSA), 0.025% nonidet NP-40, 1 mM EDTA, 5 mM DTT and 1 μg of poly(dI/dC). The end-labeled probe (100 pmol (20,000 cpm) was incubated on ice for 30 minutes with 5 μg of nuclear extract. Samples were loaded on a 5% polyacrylamide gel in 0.5 TBE buffer for 2 hours at 10 V/cm. The gel was dried and autoradiographed. For the competition assay, a molar excess of cold probe was added to the binding reactions.

ChIP assay was done using a modified procedure from Upstate Biotechnology. Briefly, E18 retinal progenitors were grown in proliferating conditions and histones were crosslinked to DNA by adding formaldehyde directly to culture medium to a final concentration of 1% and incubating for 10 minutes at room temperature on a rocking platform. Cells were washed three times with ice-cold PBS containing protease inhibitors. Cell pellets were resuspended in pre-warmed SDS lysis buffer (1% SDS, 10 mM EDTA, 50 mM Tris-HCl, pH 8.1). To reduce the non-specific background, the samples were pre-cleared using 80 μl salmon-sperm DNA/protein A agarose slurry at 4°C for 30 minutes. Samples were centrifuged at 100 r.p.m. for 1 minute at 4°C. Supernatants were transferred to a new tube, the immunoprecipitating antibody was added and incubation was carried out overnight at 4°C on a rocking platform. For a negative control we used no antibody or non-specific antibodies. The histone-antibody complex was precipitated using 60 μl salmon sperm DNA and protein-A-Sepharose (Upstate) for 1 hour at 4°C. Precipitates were washed sequentially at room temperature for 5 minutes, once with low salt immune complex wash buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl, pH 8.1, 150 mM NaCl.), high salt immune complex wash buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl, pH 8.1, 500 mM NaCl) and lithium salt immune complex wash buffer (2.25M LiCl, 1% IGEPAL-CA630, 1% deoxycholic acid sodium salt, 1 mM EDTA, 10 mM Tris-HCl, pH 8.1) and finally twice with TE buffer (10 mM Tris-HCl, 1 mM EDTA, pH 8.0). After completely removing the TE buffer, the precipitate was resuspended and extracted twice in 250 μl freshly prepared elution buffer (1% SDS, 0.1M NaHCO₃). To reverse the histone-DNA crosslinks, samples were heated at 65°C for 4 hours. 200 μl of initial sonicated sample was reverse crosslinked and used as input. After removing the antibodies by protease digestion of the samples, DNA was recovered and column purified. PCRs were performed using gene-specific primers (5′-GGTAGATGTTGGGATGGCTAC-3′, 5′-CCATCTACAACCCTACCGATG-3′).

We thank Douglas Ross for the ABCG2 promoter-reporter construct and Kenneth Cowen for ABCG2 retrovirus constructs. This work was supported by the Nebraska Tobacco Settlement Biomedical Research Development, COBRE, the Lincy Foundation and the Pearson Foundation.