[1]
132w
The generation of differentiated progeny in the cortex ultimately relies on a limited population of stem cells. However, the number of cell divisions required to generate such large quantities of neurons and glia cannot only be dependent on the differentiative, asymmetric divisions of these stem cells, as the mitotic burden could lead to fatal genetic errors in this long-lasting precursor population. Rather, general defining features of stem cells are their capacity to be maintained in a slowly dividing state, to preserve the stem cell pool over time, as well as their contribution to a transient population of intermediate progenitors that are committed to differentiation (Mamber et al., 2010;Takashima et al., 2013;Tiede et al., 2007). However, how the proliferative dynamics of stem and progenitor cells are regulated in the cortex has remained elusive.
[2]
166w
The germinal region of the developing cortex consists of cells with stem cell-like properties, termed radial glia cells (RGCs), and their more differentiated progeny, intermediate progenitor cells (IPCs) (Molyneaux et al., 2007). In the cortex, neural cells become committed to differentiation through the sequential expression of proneural basic-helix-loop-helix (bHLH) proteins such as Ngn2 and the IPC transcription factor Tbr2. Subsequently, differentiating cells exit the cell cycle, upregulate postmitotic neuronal subtype-specific proteins, such as Tbr1 or Cux1, and migrate into the appropriate cortical layer. Wnt/b-catenin signaling is ongoing in cortical precursor cells and is involved in regulating RGC and IPC cell-cycle kinetics (Woodhead et al., 2006). Increasing canonical Wnt signaling, through stabilization of b-catenin protein, stimulates cell-cycle progression in the developing cortex and spinal cord (Chenn and Walsh, 2002), in part by activating the expression of D-type cyclins (Megason and McMahon, 2002). Concomitantly, removal of Wnt/b-catenin signaling decreases proliferation of cortical precursors and results in premature commitment to neuronal differentiation (Machon et al., 2003;Zechner et al., 2003).
[3]
286w
High-mobility group (HMG) box transcription factors of the Sox family are intrinsic determinants that have been implicated in the regulation of stem and progenitor cell proliferation (Kormish et al., 2010). Members of the SoxB1 group of proteins (Sox1-3) are expressed in all neural precursors cells (NPCs) in the CNS (Pevny and Placzek, 2005). They share a high degree of amino acid identities, both within and outside the HMG domain, and exhibit extensive functional redundancy (Miyagi et al., 2009). Analyses conducted both in the developing and adult CNS demonstrate that SoxB1 protein function is both sufficient and a prerequisite to maintaining cells in a mitotically active precursor state (Sarkar and Hochedlinger, 2013). Despite this fact, overexpression or misexpression of SoxB1 proteins in immortalized cells in vitro has actually been shown to reduce their proliferation rate (Cox et al., 2012;Otsubo et al., 2008), while removal of one of the SoxB1 genes, Sox2, causes quiescent Muller-glia progenitors in the adult mouse retina to become mitotically active (Surzenko et al., 2013). However, the mechanisms by which differences in Sox2 expression levels can alter a cell's cell-cycle kinetics are not understood. Interestingly, in the developing mouse cortex, the level of Sox2 expression decreases among progenitor cells as they progress through neurogenesis. Cortical RGCs, with stem cell features such as multipotency and neurosphere-forming capacity, express higher levels of Sox2 compared to Tbr2 + IPCs, which are committed to differentiation (Hutton and Pevny, 2011), while both these cell populations express higher levels of Sox2 than postmitotic cells expressing the neuronal marker Tbr1 (Hutton and Pevny, 2011). However, whether differential expression of Sox2 has a function in regulating the proliferation or differentiation of stem and progenitor cells in the developing cortex is not known.
[4]
130w
In this study, we have used functional and genome-wide approaches to characterize how cell-cycle progression of RGCs and differentiating IPCs is regulated in the mouse cortex. We demonstrate that high levels of Sox2 demarcate stem cell-like RGCs, which are not yet committed to differentiation, and maintain these cells in a slowly self-renewing state by suppressing the expression of Cyclin D1. Upon differentiation, proneural bHLH factors decrease Sox2 expression, which results in derepression of Cyclin D1 expression, commitment to proliferation, and the transition to an IPC state. Together, our results imply that the dosage-dependent regulation of stem and progenitor cell proliferation by Sox2 is a general mechanism in the developing and adult CNS and that it is vital for the ability of neural cells to be maintained in an undifferentiated state.
[1]
253w
To examine how the proliferation of stem and progenitor cells is regulated in the embryonic cortex, we first defined the molecular signatures of 96 NPCs by single-cell RNA sequencing (RNA-seq) (Picelli et al., 2013;Figure S1A). Profiling of the NPCs, using principal component analysis (PCA; component 1 and 2 scores), revealed that cells separated relative to their progression through differentiation, such that cells expressing above-average levels of Sox2, or the RGC marker Pax6, grouped separately from cells expressing high levels of the early panneuronal marker Tubb3 (Figure 1A). Similarly, along a single PCA axis (component 2), cells expressing the highest levels of Sox2 grouped separately from cells expressing higher levels of differentiation genes, such as Dll1 and Tubb3 (Figure 1B). By separating cells at different positions along the PCA axis into four bins of 24 cells (B1-B4), we found that proliferation genes, including those encoding cyclins, PCNA, and the cell-cycle licensing factor Mcm4, were more highly expressed in the more differentiated compartment of the PCA (Bin2) compared to that in which the highest levels of Sox2 were found (Bin1) (Figure 1C; Figure S1B). Gene ontology (GO) term enrichment of the genes most highly expressed within each bin revealed that genes associated with DNA replication were specifically enriched in Bin2, while neuronal terms were enriched in Bin4 (Figure 1D). Thus, these analyses provide evidence that Sox2 is expressed at its highest levels in cortical NPCs uncommitted to neurogenesis and that the level of Sox2 decreases in differentiating cells as they upregulate genes promoting cell-cycle progression.
[2]
287w
In the embryonic day 13.5 (E13.5) cortex, all Sox2 + cells expressed the proliferation marker Ki67 (Figure S1C), while cells labeled by an 1 hr pulse of bromodeoxyuridine (BrdU), which incorporates into DNA during S phase, were much more likely to be found in cells expressing the lowest levels of Sox2 (Figures 1E-1H). When divided into two groups based on Sox2 expres-sion levels, the fraction of BrdU-labeled Sox2 low cells was 2.6 times greater than that of Sox2 high (Figure 1I) (for definition of Sox2 high and Sox2 low , see the Supplemental Experimental Procedures), and PH3 + M phase cells were similarly more frequent in the Sox2 low population (Figure S1D). Concomitantly, $40% of the Sox2 high cells remained BrdU negative even after a 14 hr treatment with BrdU, although this treatment was sufficient to label the vast majority (>90%) of the Sox2 low cells (Figure 1I). Moreover, Sox2 low cells were much more likely to be double labeled by a pulse of BrdU that was followed 24 hr later by a pulse of 5-ethynyl-2 0 -deoxyuridine (EdU) (Figures 1J and S1E). High and low expression of the other two SoxB1 members, Sox1 and Sox3, could similarly define slowly and rapidly dividing cortical NPCs (data not shown). In the subventricular zone (SVZ) of the adult mouse brain, the relationship between Sox2 levels and BrdU incorporation or PH3 expression followed that found in NPCs of the embryonic cortex, such that a 24 hr pulse of BrdU labeled a fraction of Sox2 low cells four times greater than of Sox2 high cells . Hence, in the developing and adult brain, the likelihood of finding a cell committed to proliferation is inversely correlated with its expression level of Sox2.
[3]
305w
To address if high levels of SoxB1 expression were refractory to cortical NPC competence for cell-cycle re-entry, we increased their expression levels in E13.5 cortices by overexpressing Sox1, Sox2, Sox3 or GFP in the ventricular zone (VZ) using in utero electroporations (Figure 1P). In comparison with GFP-electroporated cells, overexpression of Sox2 for 16 hr decreased the percentage of BrdU + cells labeled by a 1 hr pulse (15 hr after electroporation) from 25% to $8% (Figures 1Q, 1R, and 1T) and similarly decreased the fraction of PH3 + cells (Figure 1U) without forcing cell-cycle exit (Figures S1G and S1H). Overexpression of Sox1 or Sox3 had a similar effect and decreased the proportion of BrdU-incorporating cells to 12% and 11%, respectively (Figure S1I). We next examined proliferation in cortical cells after we reduced the level of Sox2 only or the level of all SoxB1s simultaneously, by electroporating small hairpin RNA (shRNA) constructs targeting specifically Sox2 or all SoxB1 transcripts (shRNA-Sox2 or shRNA-SoxB1; Figures S1J-S1L). In accordance with the results above, a decrease in Sox2 or SoxB1 levels for 16 hr increased the proportion of BrdU + -transfected cells to over 32% and 45%, respectively (Figures 1S and 1T), whereas the fraction of PH3 + cells was more than doubled by shRNA-SoxB1 misexpression (Figure 1U). Moreover, the proportion of cells double labeled by a pulse of BrdU that was followed 24 hr later by a pulse of EdU was decreased by Sox2 overexpression and increased by shRNA-SoxB1 misexpression (Figures 1V and S1M). However, it is important to point out that the rapidly proliferative state induced by shRNA-SoxB1 was only transient, and 72 hr after electroporation, most of the transfected cells had downregulated Sox2 expression completely and differentiated into postmitotic neurons (Figure S1N). Nevertheless, while high levels of Sox2 reduced BrdU incorporation, decreased SoxB1 expression promoted cell-cycle progression.
[4]
132w
RGCs divide at the apical surface, where cells going through symmetrical, proliferative divisions have division planes that are vertical to the apicobasal membrane, whereas cells going through asymmetrical, differentiative divisions have more horizontal cleavage planes (Konno et al., 2008). Consistent with the finding that Sox2 levels are decreased in differentiating progenitors, daughter cells of a symmetrical RGC division generally expressed higher and more uniform levels of Sox2 compared to daughter cells of an asymmetrical, differentiative division (Figures S2A-S2C). Interestingly, overexpression of Sox2 increased symmetrical divisions and decreased the fraction of basal IPC divisions, whereas shRNA-SoxB1 increased asymmetric divisions and the fraction of basal IPC divisions (Figures S2D-S2F). Thus, while high levels of Sox2 promote RGCs to undergo symmetrical divisions, a reduction in Sox2 levels causes cells to go through differentiative, asymmetric divisions.
[5]
143w
We next examined how the transition between a slowly proliferating Sox2 high and a rapidly proliferating Sox2 low state is regulated. Since our single-cell analysis demonstrated that the level of Sox2 expression decreases as cells become progressively more differentiated, we correlated the expression levels of markers that reflect the progression of cortical neurogenesis, Neurog2 (Ngn2), Eomes (Tbr2), and Tbr1, with those of Sox2. PCA revealed that cells separated based on their expression levels of Sox2 (black triangle), Neurog2 (red triangle), and Tbr1/2 (green triangle) (Figure 2A). Interestingly, the ordered appearance of these differentiation markers during cortical neurogenesis corresponded with their decreasing correlation coefficients with Sox2 (Figure 2B). Thus, when cells were ordered by Sox2 levels, Neurog2 expression increased simultaneously as the level of Sox2 started to decrease, while as cells upregulated Eomes and Tbr1, the level of Sox2 expression decreased further (Figure 2C).
[6]
178w
We detected a similar relationship between the expression of Sox3 and differentiation markers in cortical NPCs analyzed with immunohistochemistry. In a region of E13.5 cortices corresponding to that shown in Figure 1E, 40% of Ngn2 + cells were Sox3 high , yet only 25% of the Tbr2 + cells in the VZ were Sox3 high (Figures 2D-2J). Cells in the SVZ that continued to express Tbr2 and upregulated Tbr1 were Sox3 negative (Figures 2F-2J). Furthermore, the degree of BrdU incorporation, following a 1 hr pulse, increased with the onset of Tbr2, such that 28% of Ngn2 + was BrdU + , while 42% of the Tbr2 + cells within the VZ were BrdU + (Figures 2D-2K). Tbr2 + /Tbr1 + cells in the SVZ were generally BrdU negative. Consistent with these results, Tbr2 expression was decreased by Sox2 and increased by shRNA- Sox2 transfections, respectively (Figures S2G-S2I and S2K). In the SVZ of the adult brain, we found similar dynamics in Sox expression levels and the onset of differentiation markers, including, GFAP, Mash1, Tbr2, and PSA-NCAM (Figures S2M-S2Z).
[7]
209w
Since expression of Neurog2 increases as cells start to downregulate Sox2 (Figure 2C), we examined if Ngn2 could promote NPCs to downregulate Sox2 expression and thereby commit to a rapidly proliferating state. Electroporation of Ngn2 into E13.5 mouse cortices resulted in a rapid decline in Sox2 expression levels. After 16 hr, only a small fraction of the Ngn2-transfected cells were Sox2 high , and after 32 hr, most electroporated cells were Sox2 negative (Figures 2L and 2M). Moreover, both Ngn2 and Mash1 could act in a dose-dependent manner in P19 cells to repress a luciferase reporter containing the Sox2 enhancer SRR1 (Miyagi et al., 2004) (Figures 2N and S2L). Concomitantly, 16 hr after Ngn2 electroporation, the fraction of Tbr2 + and BrdU + transfected cortical NPCs was almost double that of the GFP-transfected cells (Figures 2O and 2Q and S2J and S2K), whereas after 32 hr, most of the Ngn2-transfected cells were both Ki67 and BrdU negative (Figures 2O and 2Q; data not shown). Importantly, the capacity of Ngn2 to promote this transient increase in NPC proliferation was blocked by cotransfected Sox2 (Figures 2P and 2Q). Thus, in cortical NPCs, proneural bHLH proteins can promote the transition into a rapidly self-renewing state via a reduction in Sox2 expression levels.
[8]
369w
That the level of Sox2 could act as a regulatory effector of NPC proliferation prompted us to screen for Sox2-regulated target genes relevant to this process. To detect direct Sox2 target genes, we performed in vivo chromatin immunoprecipitation sequencing (ChIP-seq) analysis on embryonic cortical tissue (I) A smaller fraction of Sox2 high than Sox2 low cells is BrdU + after both 1 hr and 14 hr pulses (n = 4). (J) Sox2 low cells are more likely to be double labeled by a pulse of BrdU that was followed 24 hr later by a 1 hr pulse of EdU (n = 3). (K-M) Sox2 + cells in the adult SVZ that incorporated BrdU after a 24 hr pulse. Boxed area in (K) represents regions shown in (L) and (M). (N) Number of BrdU + cells within ten Sox2 level bins in SVZ (n = 4). (O) BrdU labels a smaller fraction of Sox2 high then Sox2 low cells in the adult SVZ (n = 4). (P-U) Sox2 electroporation in E13.5 cortices decreases the fraction of BrdU + or PH3 + GFP-electroporated cells (n = 6) (P-R, T, and U), while shRNA-SoxB1 (n = 7) or shRNA-Sox2 (n = 3) increases the proportion of BrdU + or PH3 + cells (Q and S-U). (V) The proportion of cells double labeled by a pulse of BrdU that was followed 24 hr later by a 1 hr pulse of EdU was decreased by Sox2 and increased by shRNA-SoxB1 (n = 3). (legend continued on next page) (Figure 3A). In parallel, we compared global gene expression using mRNA-seq of fluorescence-activated cell sorting (FACS)purified cortical cells 16 hr after misexpressing Sox2, shRNA-SoxB1, or GFP (Figure 3A). Characterization of the two ChIP-seq replicates confirmed that they were highly overlapping (Figures S3A and S3B) and allowed us to identify substantially conserved Sox2-bound regions (peaks) with Sox binding sequences as the most centrally enriched motifs (Figures 3B and S3C), distributed within 500 kb of 4,353 genes (Table S1). The overlap between Sox2-bound genes in cortical NPCs and embryonic stem (ES) cell-derived NPCs was over 70%, whereas the overlap was less than 50% with genes reported to be bound by Sox2 in ES cells (Figures S3D and S3E).
[9]
133w
Comparing the cortical Sox2 ChIP-seq with the mRNA-seq analysis of the electroporated cortical NPCs revealed a direct correlation between peak height and Sox2-dependent expression change of the closest gene (Figure 3C), thus providing a functional link between both screens. In comparison with genes expressed over 5 RPKM (reads per kilobase per million mapped reads) in GFP-transfected cells, we found hundreds of genes up-and downregulated >1.5-fold in cells overexpressing Sox2 or shRNA-SoxB1 (Figures 3D, 3E, and S3H; Table S2). Gene ontology analysis of all Sox2-bound genes, and target genes positively or negatively regulated by increased or decreased Sox2 levels (Figures 3F and 3G), revealed that genes downregulated by Sox2 and/or upregulated by shRNA-SoxB1 misexpression (Figure 3G) were more than twice as enriched as the other groups for neural precursor cell proliferation terms (Figure 3H).
[10]
210w
Given that the GO analysis suggested that Sox2-repressed target genes are involved in regulating NPC proliferation, we next examined the nine Sox2-bound genes, which were both downregulated by Sox2 and upregulated by shRNA-SoxB1 (Figure 3G), for their effect on cell-cycle regulation (Figure 3I). In situ hybridization analysis confirmed that all of these genes are expressed in E13.5 cortices (Figure S4A). Four of the nine genes (Ccnd1, Bhlhe40, D16ertd472e, and Fam181b) significantly increased the proliferation rate of electroporated cortical NPCs compared to GFP-transfected control cells (Figure 3I). However, of these, only Ccnd1, which encodes for the G1/S phase transition licensing factor Cyclin D1, retained its full ability to promote proliferation when coelectroporated with Sox2 (Figures 3J-3O). In situ hybridization of electroporated E13.5 cortices or chick neural tubes confirmed that the expression level of Ccnd1 was downregulated by Sox2 and upregulated by shRNA-SoxB1 or a dominant-negative variant of Sox3 (Shih et al., 2010) (dnSoxB1; Figures S4B-S4E). In addition, our single-cell analysis also confirmed a strong negative correlation between the expression levels of Sox2 and Ccnd1 (correlation coefficient À0.132) in cortical NPCs under nonoverexpressing physiological conditions (Figure 3P). Thus, Sox2 regulates cell-cycle progression by acting as a direct suppressor of genes that promote NPC proliferation, of which Cyclin D1 acts epistatically to Sox2.
[11]
81w
Notably, forced expression of Cyclin D1 for a prolonged time has been demonstrated to induce cells to become postmitotic and upregulate markers of differentiation (Lange et al., 2009). We found that even though Sox2 could prevent neuronal differentiation in the presence of high Cyclin D1 levels for 72 hr, Sox2 could not prevent cells from becoming postmitotic (Figure S4F), indicating that the reduction of Cyclin D1 expression is vital to the ability of Sox2 to maintain NPCs in a proliferationcompetent state.
[12]
207w
The Ccnd1 promoter is the third most Sox2-occupied region in the entire genome (Table S1) and has been shown to be positively regulated by b-catenin and Tcf/Lef transcription factors (Tetsu and McCormick, 1999). ChIP quantitative PCR (qPCR) analysis of mouse cortical NPCs confirmed strong Sox2 binding to a region À100 to À230 bp upstream of the Ccnd1 open reading frame (region C) and less binding to a region À350 to À512 bp upstream (region B) (Figures 4A and 4B). The binding of Lef1 to the Ccnd1 promoter followed that of Sox2 and could primarily be detected at region C and, to a lesser extent, at region B (Figures 4A and 4B). We next generated a luciferase reporter construct (Ccnd1-Luc) containing the conserved Sox2-bound region of the Ccnd1 promoter (À409 to +94 bp; Figures 4A and S5A). In chick NPCs, electroporated Ccnd1-Luc was repressed 3.4-fold by cotransfected Sox2 (Figure 4C) and, to a similar extent, by Sox1 or Sox3 overexpression (Figure S5B). In contrast, misexpression of dnSoxB1 or the DNA-binding domain of Sox2 (HMG Sox2 ) acted oppositely to Sox1-3 and activated Ccnd1-Luc approximately 3-fold (Figure 4C), demonstrating that in NPCs, the activity of the Cyclin D1 reporter responds as endogenous Ccnd1 to altered levels of Sox2 activity.
[13]
128w
The isolated region of the Ccnd1 promoter contains nine conserved, potential Sox2 binding motifs (SM1-SM9; Figure 4A), though SM8 and SM9 are downstream of transcriptional start (Figure S5E). All nine sites were bound by recombinant Sox2 protein in a DNA-binding gel shift assay (Figure 4D), although the sites closest to a consensus Sox motif, SM4, SM5, and SM7 (marked in red), were bound most efficiently. Surprisingly, individual or combinatorial mutations of the high-affinity sites had little impact on Sox2-mediated repression (Figures 4E and S5C). In contrast, individual or combinatorial mutations of the low-affinity motifs had a major impact on Sox2 mediated repression and reduced it up to 100% (Figures 4E and S5C). Thus, low-affinity sites, but not high-affinity sites, are necessary for Sox2-mediated repression of the Ccnd1 promoter.
[14]
145w
To examine the concentration-dependent regulation of the Cyclin D1 promoter by Sox2, Ccnd1-luc was transfected together with various amounts of Sox2 into human embryonic kidney 293 (HEK293) cells. Similar to Ccnd1 expression in the cortex, Ccnd1-luc was repressed in a concentration-dependent fashion by Sox2 (Figure 4F). Interestingly, the concentration dependence of Sox2 repression was reduced and nearly abolished by the loss of one or more low-affinity sites (Figures 4F and S5D), whereas high-affinity sites were insignificant in this context. Moreover, at a genome-wide level, we found a significant enrichment of degenerated Sox motifs in peaks associated with genes negatively regulated by Sox2 overexpression in cortical NPCs (Figure 4G), while peaks near genes positively regulated by Sox2 instead had an enrichment of consensus Sox motifs (Figure S5F). Thus, multiple low-affinity sites appear to be a common feature of regulatory sequences facilitating graded Sox2 target gene repression.
[15]
234w
The Wnt signaling transducer b-catenin can increase Ccnd1 promoter activity and NPC proliferation (Megason and McMahon, 2002;Tetsu and McCormick, 1999). We found that a stabilized version of b-catenin (DNb-catenin) (Megason and McMahon, 2002) could activate Ccnd1-Luc nearly 8-fold in chick neural cells (Figure 5A) and nearly doubled the fraction of BrdU + -electroporated cortical NPCs (Figures 5B and 5D), functions of DNb-catenin that were completely blocked by cotransfected Sox2 (Figures 5A, 5C, and 5D). However, while the mutation of low-affinity Sox sites (SM1-SM3 and SM6) had only a minor effect on the capacity of Sox2 to block DNb-catenin-induced Ccnd1-Luc activation (Figure 5E), this Sox2 activity was significantly more dependent on the high-affinity sites (SM5 and SM7) (Figure 5E). The major Sox2 peak associated with Ccnd2 (Cyclin D2), which is expressed at similar levels in Sox2 high (Bin1) and Sox2 low (Bin2/3) cells (Figure S5J), lacks clusters of degenerated Sox motifs but contains a consensus Tcf/Lef and Sox motif in close proximity (Figure S5G). Interestingly, although reducing SoxB1 function activated a Ccnd2 reporter in vivo, Sox2 overexpression could not repress the reporter and was only able to reduce its activation by DNb-catenin (Figures S5H and S5I). Thus, while the Ccnd1 and Ccnd2 reporters contain consensus Sox motifs, which allow Sox2 to hamper b-cateninmediated activation, only the Ccnd1 promoter also contains appropriate clusters of low-affinity sites, through which Sox2 can repress Ccnd1-Luc in a concentration-dependent fashion.
[16]
107w
Lef/Tcf Proteins Sox2 possesses only weak intrinsic regulatory capacity (Kamachi et al., 1998), and control of target gene activity is often dependent on an interaction with partner transcription factors (Kamachi et al., 1998). Coelectroporated Lef1 strongly increased the capacity of Sox2 to repress the Ccnd1 promoter in NPCs (Figure 5F), while the repression of Ccnd1-Luc by Tcf7L1 was increased in the presence of Sox2 (Figure S6C). Moreover, through coimmunoprecipitation experiments in HEK293 cells, we found a strong physical interaction between Sox2 and Lef1/Tcf7L1, apparently mediated by the HMG domain (Figure 5G; Figures S6A and S6B). Thus, Tcf/ Lef proteins potentiate Sox2's ability to repress the Ccnd1 promoter.
[17]
163w
We found three potential Tcf/Lef binding sites (LM1-LM3; Tetsu and McCormick, 1999) to be important for b-catenin and Tcf7L1 protein activity (Figures S6D and S6E) in the Ccnd1 À409 bp promoter region (Figures 4A, 4B, and S5A). In line with the cooperative function found between Sox2 and Lef1, mutational analyses revealed a significant role for LM1-LM3 in Sox2-mediated repression of Ccnd1-luc (Figure 5H). Interestingly, DNb-catenin-mediated activation was mostly dependent on the consensus motif LM2 (Figures S6D and S6E;Tetsu and McCormick, 1999), and in its absence, HMG Sox2 or dnSoxB1 was unable to activate Ccnd1-Luc (Figure S6F and data not shown). Consistent with this, in the presence of a dominantnegative version of Tcf7L2 (dnTcf7L2) that cannot recruit b-catenin (Megason and McMahon, 2002), shRNA-SoxB1 was unable to increase proliferation of cortical NPCs (Figures 5I-5K). Hence, the activation of the Ccnd1 promoter and the increase in cortical proliferation caused by reduced SoxB1 function appear to be based on a relief of the suppression of b-catenin activity.
[18]
117w
Genome-wide, Tcf/Lef motifs were highly enriched and found in 24% of all Sox2 peaks (Figure S6G). Moreover, we identified closely spaced Sox and Tcf/Lef motifs (4-10 bp) specifically enriched in peaks associated with genes negatively regulated by Sox2 in cortical NPCs (Figure 5L). In line with this observation, of the nine target genes that were both downregulated by Sox2 and upregulated by shRNA-SoxB1 in cortical NPCs (Figure 3G), five genes were associated with peaks containing modules of conserved, 4 bp-to 10 bp-spaced Sox and Tcf/Lef motifs (Figure S7). Thus, a module of 4 bp-to 10 bp-spaced Sox and Tcf/Lef motifs is one significant characteristic associated with genes negatively, but not positively, regulated by Sox2 in cortical NPCs.
[19]
271w
The cooperative function between Sox2 and Lef1 proteins in the repression of the Ccnd1 promoter encouraged us to examine if this was dependent on the recruitment of known Tcf/Lef cofactors Gro/Tle family repressors (Arce et al., 2006;Liu et al., 2014) (Figure 6A). Although both Lef1 and Sox2 interacted weakly with Tle1 in transfected HEK293 cells (Figure 6B), the interaction between Lef1 and Tle1 was greatly enhanced by the presence of increasing levels of Sox2 (Figure 6B). Sox2 was found to interact with Tle1 via its C-terminal, non-DNA-binding domain (Figure 6C). Moreover, while Tle1 acted synergistically with Sox2 to repress the Ccnd1 promoter, its own repressive activity was completely blocked by mutating Sox or Lef sites or coelectroporating HMG Sox2 (Figures 6D and S6H). Moreover, the ability of Tle1 to decrease the proliferation rate of cortical NPCs (Figures 6E-6G) was dependent on SoxB1 function and was abolished in the presence of coelectroporated shRNA-SoxB1 (Figures 6E-6G). To examine if Gro/Tle corepressors were essential to the repressive activity of Sox2, we took advantage of the finding that the interaction between Lef1 and Gro/Tle is dependent on a Groucho binding sequence (GBS) in Lef1 (Lef1 GBS ) (Arce et al., 2009). Although Sox2 synergized with cotransfected Lef1 to repress Ccnd1-Luc (Figure 5F), Lef1 GBS fully blocked the capacity of Sox2 to repress Ccnd1-Luc (Figure 6H) and misexpression of Lef1 GBS increased the fraction of BrdU-incorporating electroporated cortical NPCs to over 45% even in the presence of coexpressed Sox2 (Figures 6I-6K). Thus, the capacity of Sox2 to repress the Ccnd1 promoter and NPC proliferation is dependent on its cooperative function with Lef1 in recruiting Gro/Tle corepressors.
[1]
113w
The development and maintenance of organs as diverse as the CNS, intestine, hair follicle, and hematopoietic system rely on a limited population of slowly self-renewing stem cells to divide asymmetrically and supply a pool of rapidly dividing and differentiating progenitor cells. However, despite the importance of this proliferative hierarchy, how it is regulated has remained elusive. In this study, we have focused on how competence for cell-cycle entry is regulated in coordination with the transition from stem cell-like RGCs to differentiating RGCs and IPCs. We provide mechanistic insights, with a genome-wide perspective, into how the stem cell determinant Sox2 acts in a concentrationdependent fashion to regulate cell-cycle progression in these two cellular states.
[2]
161w
Our analyses indicate that uncommitted cortical RGCs are less prone to enter the cell cycle than differentiating progenitor cells. This finding is consistent with the proliferative behavior of precursor cells in other stem cell niches, including those of the adult brain (Mamber et al., 2010;Takashima et al., 2013;Tiede et al., 2007). However, previous studies, also conducted in the developing cortex, have shown that the transition from stem cell-like RGCs to differentiating Tbr2 + IPCs is associated with a lengthening of the cell cycle (Arai et al., 2011). These discrepancies can possibly be explained by our finding that Tbr2 + IPCs proliferate with different rates depending on whether the cells are within the VZ or SVZ. Although the overall Tbr2 + population incorporates BrdU to a lesser extent than the average precursor cell within the VZ (Figure 2K), Tbr2 + IPCs within the VZ (Sox3 + ) appear to be more rapidly dividing compared with the general precursor cell within the VZ.
[3]
258w
The model proposed in this paper is partly based on the finding that the level of SoxB1 expression is inversely correlated with a cell's likelihood of incorporating BrdU or expressing the M phase marker PH3. However, can we exclude the possibility that low SoxB1 levels are not varying with the different phases of the cell cycle? A number of observations argue against this interpretation. First, we failed to detect a cell-cycle phase that is generally low in SoxB1 expression, as both SoxB1 low and SoxB1 high cells are labeled by a short pulse of BrdU and the M phase marker PH3 (even though consistently twice as many positive cells are Sox2 low in both cases). Moreover, our single-cell data indicate that markers of all phases of the cell cycle are collectively expressed in both the Sox2-high (Bin1) and Sox2-low (Bin2/3) compartment of the PCA (Figures 1C and S1B). Second, a cumulative BrdU treatment over 14 hr failed to incorporate in almost 40% of Sox2 high cells, despite labeling nearly all Sox2 low cells, suggesting Sox2 high cells are refractory to S phase entry over a long period. Finally, in contrast to Hes1, which has been shown to oscillate with an approximately 2 hr interval (Shimojo et al., 2008), Sox2 was shown in similar experiments to be expressed at relatively constant levels throughout the cell cycle (Imayoshi et al., 2013). Thus, rather than being associated with a certain phase of the cell cycle, low levels of SoxB1 expression are more likely a characteristic of cortical cells committed to differentiation.
[4]
291w
An important question addressed in this paper is how differences in the expression level of a transcription factor, such as Sox2, can be interpreted and result in a specific response in target gene expression. In this study, we have observed that the Sox2-targeted region of the Cyclin D1 promoter contains Sox motifs of different affinities and that the low-affinity Sox2 motifs are essential for the graded repression of the Ccnd1 promoter to increasing amounts of Sox2. This implies that the differences in Cyclin D1 expression between cells with high and low SoxB1 levels may be explained by the occupancy grade of Sox2 motifs in its promoter region. Our model is analogous to the situation in the Drosophila embryo, in which high and low levels of the Bicoid (Bcd) transcription factor along the anterior-posterior axis result in anterior expression of Orthodenticle (otd) and posterior expression of hunchback (hb). It is proposed that otd expression relies on off-consensus, low-affinity Bcd binding sites and thus requires higher levels of the Bcd protein, while hb expression relies both on consensus, high-affinity sites as well as low-affinity sites (Ephrussi and St Johnston, 2004). Notably, by analyzing Sox2-bound regions at a genome-wide level, we identified a significant enrichment of degenerated Sox sites in peaks associated with genes that are negatively regulated by Sox2 overexpression in NPCs. Hence, together with previous findings in the Drosophila embryo, our data indicate that the presence of binding motifs of varying affinities may reflect a common mechanism for translating the expression level of transcription factors into specific target gene expression. However, despite the similarities between our models, future experiments are needed to confirm differential binding of Sox2 to the Ccnd1 promoter in cells expressing high versus low levels of Sox2 in vivo.
[5]
281w
Transactivation experiments in vitro and studies in ES cellderived NPCs, in which the level of SoxB1 expression was either increased or decreased (Bergsland et al., 2011;Engelen et al., 2011;Kamachi et al., 1998), suggest that these proteins mainly function to activate target genes. Unexpectedly, our gain-and loss-of-function studies in cortical NPCs in vivo revealed that approximately half of the Sox2 regulated target genes were repressed by Sox2 overexpression. Furthermore, we identified a stronger correlation between negative gene regulation and Sox2 binding than with positive gene regulation (Figure S3H). Nevertheless, the capacity of Sox2 to both activate and repress target gene expression in a population of cortical NPCs argues that the regulatory activity of Sox2 is determined by contextual differences at the DNA level. Indeed, the interaction with Tcf/Lef proteins is essential for the capacity of Sox2 to repress the Ccnd1 promoter and NPC proliferation. Moreover, Tcf/Lef biding sites are among the most overrepresented motifs in Sox2 peaks, and a module with Sox and Tcf/Lef motifs spaced between 4 and 10 bp was specifically enriched in peaks associated with genes repressed by Sox2. One possibility is that this particular distance between Sox and Tcf/Lef motifs is favorable for the establishment of a bimodal surface that facilitates the recruitment of Gro/Tle corepressors. By studying the regulation of the Ccnd1 promoter, our findings propose a model for a SoxB1-dependent biphasic repressive mechanism: Tcf/Lef motifs, together with low-affinity Sox sites, allow SoxB1 proteins to act in a concentration-dependent fashion to repress Ccnd1 expression through the stabilization of a Gro/Tle corepressor complex. In addition, a module of Tcf/Lef motifs and high-affinity Sox sites facilitate SoxB1 proteins to moderate b-cateninmediated Ccnd1 activation in a dose-independent fashion (Figure 7).
[6]
127w
Apart from regulating the proliferative dynamics of cortical cells, SoxB1 proteins have previously been shown to prevent NPCs from committing to differentiation. However, the mechanisms by which SoxB1 proteins maintain undifferentiated features of neural cells have remained unresolved. Our finding that prolonged overexpression of Cyclin D1 eventually causes NPCs to become postmitotic, even in the presence of forced Sox2 expression, suggests that controlling the expression of Cyclin D1 is an essential part of this mechanism. Thus, apart from reducing the frequency of genetic errors incorporated into the long-lived stem cell population, the ability of high levels of Sox2 to slow cell-cycle entry through the repression of Cyclin D1 expression is likely a vital function by which SoxB1 proteins maintain neural cells in an undifferentiated stem cell-like state.
[7]
147w
While this study focuses on the role of SoxB1 protein in the regulation of stem and progenitor cells in the healthy brain, it is notable that reduced Sox2 expression levels are a hallmark of gastric cancers and high levels of Sox2 have tumor suppressor functions in brain cancer cells (Cox et al., 2012). In addition, it is tempting to speculate that other Sox transcription factors may employ a similar regulatory mechanism in controlling precursor self-renewal. For instance, proliferation of intestinal cells, which is largely dependent on Wnt/b-catenin signaling (Holland et al., 2013), is reduced by Sox9 overexpression (Formeister et al., 2009) and increased upon the loss of Sox9 (Bastide et al., 2007). Together with our findings, these data suggest that specific Sox proteins may be commonly employed to control the proliferative hierarchy of stem and progenitor cells of diverse origins through the reduction of Wnt/b-catenin-activated proproliferative genes.