Human limbal stem cells produce transit amplifying progenitors that migrate centripetally to regenerate the corneal epithelium. Coexpression of CCAAT enhancer binding protein δ (C/EBPδ), Bmi1, and ΔNp63α identifies mitotically quiescent limbal stem cells, which generate holoclones in culture. Upon corneal injury, a fraction of these cells switches off C/EBPδ and Bmi1, proliferates, and differentiates into mature corneal cells. Forced expression of C/EBPδ inhibits the growth of limbal colonies and increases the cell cycle length of primary limbal cells through the activity of p27Kip1 and p57Kip2. These effects are reversible; do not alter the limbal cell proliferative capacity; and are not due to apoptosis, senescence, or differentiation. C/EBPδ, but not ΔNp63α, indefinitely promotes holoclone self-renewal and prevents clonal evolution, suggesting that self-renewal and proliferation are distinct, albeit related, processes in limbal stem cells. C/EBPδ is recruited to the chromatin of positively (p27Kip1 and p57Kip2) and negatively (p16INK4A and involucrin) regulated gene loci, suggesting a direct role of this transcription factor in determining limbal stem cell identity.
Stem cells have the unique capacity to self-renew and generate committed, transit amplifying (TA) progenitors that differentiate into the cell lineages of the tissue of origin (Niemann and Watt, 2002; Fuchs et al., 2004; Cotsarelis, 2006; Blanpain et al., 2007). The most important function of TA cells is to increase the number of differentiated progeny produced by each stem cell division, thus enabling stem cells to divide infrequently, at least under normal tissue homeostasis. The cornea provides an ideal experimental system for studying stem cells of human stratified epithelia (Lavker and Sun, 2003). Human corneal stem cells are segregated in the basal layer of the limbus, which is the vascularized zone encircling the cornea and separating it from the bulbar conjunctiva. The corneal epithelium lies on the avascular Bowman's membrane and is formed by TA keratinocytes that migrate millimeters away from their parental limbal stem cells (Schermer et al., 1986; Cotsarelis et al., 1989; Lehrer et al., 1998; Pellegrini et al., 1999a).
Clonal analysis of squamous human epithelia, including the cornea, has identified three types of clonogenic keratinocytes, giving rise to holoclones, meroclones, and paraclones in culture (Barrandon and Green, 1987; Pellegrini et al., 1999a). Holoclone-forming cells have all the hallmarks of stem cells, including self-renewing capacity (Rochat et al., 1994; Claudinot et al., 2005), telomerase activity (Dellambra et al., 2000), and an impressive proliferative potential—a single holoclone can generate the entire epidermis of a human being (Rochat et al., 1994). Holoclone-forming cells generate all the epithelial lineages of the tissue of origin (Pellegrini et al., 1999a; Oshima et al., 2001; Blanpain et al., 2004; Claudinot et al., 2005), permanently restore massive epithelial defects (Gallico et al., 1984; Romagnoli et al., 1990; Pellegrini et al., 1997, 1999b; Ronfard et al., 2000), and can be retrieved from human epidermis regenerated from cultured keratinocytes years after grafting (De Luca et al., 2006). We have recently shown that a defined number of genetically corrected stem cells regenerate a normal epidermis in patients with genetic skin adhesion disorders (Mavilio et al., 2006). The paraclone is generated by a TA cell, whereas the meroclone has an intermediate clonal capacity and is a reservoir of TA cells (Barrandon and Green, 1987; Pellegrini et al., 1999a).
The p63 gene produces full-length (TAp63) and N-terminally truncated (ΔNp63) transcripts initiated by different promoters. Each transcript is alternatively spliced to encode three different p63 isoforms, designated α, β, and γ (Yang et al., 1998). The p63 gene products are essential for the morphogenesis and the regenerative proliferation of stratified epithelia (Mills et al., 1999; Yang et al., 1999). In particular, ΔNp63α sustains the proliferative potential of basal epidermal keratinocytes (Parsa et al., 1999; Koster et al., 2004; McKeon, 2004; Nguyen et al., 2006). In the human corneal epithelium, high levels of ΔNp63α identify limbal stem cells both in vivo and in vitro, whereas ΔNp63β and ΔNp63γ correlate with corneal regeneration and differentiation (Pellegrini et al., 2001; Di Iorio et al., 2005).
In mammary gland epithelial cells, the CCAAT enhancer binding protein δ (C/EBPδ) transcription factor regulates cell cycle by inducing a G0/G1arrest. This effect is specific for epithelial cells and for the G0/G1 phase, as C/EBPδ expression does not increase in other types of G0/G1-arrested cells or in mammary cells arrested at other stages of the cell cycle (O'Rourke et al., 1999; Hutt et al., 2000). C/EBPδ is a member of a highly conserved family of leucine zipper transcription factors expressed in a variety of tissues and cell types and involved in the control of cellular proliferation and differentiation, metabolism, and inflammation (Ramji and Foka, 2002; Johnson, 2005). At least six members of the family have been isolated and characterized (C/EBPα–C/EBPζ), with further diversity produced by the generation of different polypeptides by differential use of translational initiation sites, and extensive protein–protein interactions within the family and with other types of transcription factors (Ramji and Foka, 2002; Johnson, 2005).
In this paper, we show that C/EBPδ and ΔNp63α are coexpressed by human limbal stem cells in vivo and in vitro and that the expression of C/EBPδ is restricted to a subset of mitotically quiescent ΔNp63α+/Bmi1+ cells. Forced expression of a constitutive C/EBPδ or of a tamoxifen-inducible estrogen receptor (ER)–C/EBPδ fusion protein in human primary limbal keratinocytes shows that C/EBPδ is instrumental in regulating self-renewal and cell cycle length of limbal stem cells.
Coexpression of C/EBPδ and ΔNp63α in quiescent human limbal cells
Experiments were performed on four uninjured and five wounded corneas, referred to as resting and activated cornea, respectively (Di Iorio et al., 2005). We have previously shown that ΔNp63α is expressed by 10% of resting limbal basal cells endowed with stem cell properties and that activated ΔNp63α+ limbal cells contain ΔNp63β and ΔNp63γ, proliferate, and migrate to the central cornea to restore a wounded epithelium (Di Iorio et al., 2005).
Immunofluorescence analysis on resting limbal sections revealed that C/EBPδ and ΔNp63α were coexpressed in the same patches of basal cells (Fig. 1 A, left). Both transcription factors were undetectable in suprabasal cell layers (Fig. 1 A) and in the entire corneal epithelium (not depicted). Limbal cell nuclei were stained with DAPI to estimate the proportion of C/EBPδ+/ΔNp63α+ cells in the basal layer. 1 mm of resting limbal epithelium contained a mean of 15 C/EBPδ+/ΔNp63α+ cells, equivalent to ∼10% of the basal layer. Upon corneal wounding and limbal activation, ΔNp63α appeared in many basal and some suprabasal limbal cells, whereas C/EBPδ remained confined to ∼10% of the basal layer (Fig. 1 A, middle). Of note, C/EBPδ+ limbal cells invariably coexpressed ΔNp63α (Fig. 1 A). In activated limbus, patches of C/EBPδ+/ΔNp63α+ basal cells flanked by C/EBPδ−/ΔNp63α+ cells were commonly observed (Fig. 1 A, right), whereas neither resting nor activated central corneal epithelium expressed C/EBPδ (not depicted). C/EBPδ+/ΔNp63α+ resting limbal cells did not express Ki-67, a proliferation-associated nuclear antigen present throughout the cell cycle but absent in G0/G1-arrested cells (not depicted). In activated limbus, proliferating Ki-67+ limbal cells expressed ΔNp63α, but not C/EBPδ, whereas C/EBPδ+ cells contained ΔNp63α but not Ki-67 (Fig. 1 B). Thus, C/EBPδ and ΔNp63α are coexpressed by quiescent limbal basal cells, whereas ΔNp63α, but not C/EBPδ, is expressed in proliferating limbal cells.
The cyclin/Cdk inhibitors p27Kip1 and p57Kip2 negatively regulate G1 progression. Nuclear levels of p27Kip1 are high in quiescent cells (Sherr and Roberts, 1999). Mitogenic and/or oncogenic signals activate different kinases that phosphorylate p27Kip1 on serine and tyrosine residues, promoting its export from the nucleus and cytoplasmic proteolysis, thereby leading to cell proliferation (Rodier et al., 2001; Chu et al., 2007; Grimmler et al., 2007; Kaldis, 2007). Of note, p57Kip2, which inhibits cyclin D–Cdk4/6 complexes (Yamazaki et al., 2006), is highly expressed in mouse epidermal stem, but not TA, cells (Dunnwald et al., 2003). Immunofluorescence analysis on limbal sections revealed that C/EBPδ, p27Kip1, and p57Kip2 were coexpressed in the nucleus of the same patches of basal cells (Fig. 1 C, arrowheads). Such cells also expressed ΔNp63α (Fig. 1 C, arrowheads). C/EBPδ+ cells were flanked by C/EBPδ−/ΔNp63α+ cells containing cytoplasmic, but not nuclear, p27Kip1 and p57Kip2 (Fig. 1 C, brackets). Finally, p27Kip1 and p57Kip2 were never detected in a fully activated limbus or in the corneal epithelium (not depicted). These data are consistent with the notion that p27Kip1 and p57Kip2 are localized in the nucleus of quiescent cells, appear in the cytoplasm at the G1–S transition, and are not expressed by actively proliferating cells (Rodier et al., 2001), and confirm that C/EBPδ is expressed only by quiescent limbal basal cells.
Expression of C/EBPδ, ΔNp63α, and Bmi1 in limbal stem cells
Immunofluorescence analysis showed that ΔNp63α is abundantly and uniformly expressed in holoclones (Fig. 2 A), is expressed in a subset of meroclone cells, and is not expressed in paraclones (Di Iorio et al., 2005). Western analysis showed that clonal evolution, i.e., the transition from holoclones to paraclones, is accompained by a progressive disappearance of ΔNp63α and a relative enrichment in ΔNp63β and ΔNp63γ (Fig. 2 B). Strikingly, C/EBPδ expression was detected exclusively in holoclones (Fig. 2 B) and confined to a subpopulation of ΔNp63α+ cells (Fig. 2 A). C/EBPδ+/ΔNp63α+ cells were not proliferating, as shown by the mutually exclusive expression of C/EBPδ and Ki-67 (Fig. 2 C). Of note, although Ki-67 and C/EBPδ were never expressed in the same cell (Fig. 2 C), large areas of the colony were formed by nonproliferating yet C/EBPδ-negative cells (Fig. 2 C, dots), suggesting that the expression of C/EBPδ was not merely related to the proliferative status of the limbal cell.
C/EBPα and -β are the most commonly expressed and thoroughly studied isoforms of the C/EBP family (Ramji and Foka, 2002). In particular, C/EBPα and -β are known to positively regulate the program of squamous differentiation in the epidermis (Oh and Smart, 1998; Zhu et al., 1999). Accordingly, we found that C/EBPα and -β were contained in the suprabasal layers of both human limbal and corneal epithelium (unpublished data). Of note, however, although C/EBPβ was expressed in all limbal clonal types (Fig. 2 B), we could not detect C/EBPα in cultured limbal colonies.
Gene profiling experiments have led to the identification of genes that are commonly expressed in adult stem cells. Among these genes, Bmi1, a member of the polycomb group of transcription factors, plays a crucial role in the renewal of hematopoietic and neural stem cells (Lessard and Sauvageau, 2003; Molofsky et al., 2003, 2005; Park et al., 2003) and is expressed in clonogenic, multipotent, and self-renewing murine hair follicle stem cells (Claudinot et al., 2005). Immunofluorescence performed on resting limbal sections revealed that C/EBPδ and Bmi1 were coexpressed by the same limbal basal cells (Fig. 2 D). In particular, the basal layer of palisades of Vogt, where limbal stem cells are thought to be concentrated, was formed by C/EBPδ+/Bmi1+ cells (Fig. 2 D). Accordingly, Western blot analysis showed that Bmi1 was expressed in holoclones but not in meroclones and paraclones (Fig. 2 B).
Collectively, these data indicate that C/EBPδ, ΔNp63α, and Bmi1 colocalize in limbal stem cells of the resting corneal epithelium in vivo and in limbal holoclone-forming cells in vitro and that expression of C/EBPδ is restricted to a subset of ΔNp63α+ cells that are mitotically quiescent both in vivo and in vitro.
C/EBPδ regulates the cell cycle of human limbal stem cells
Primary limbal cultures were infected with a lentiviral vector expressing either an epitope-tagged human C/EBPδ or a control protein (a truncated form of the p75 low-affinity NGF receptor [ΔNGFr]) under the control of a constitutive phosphoglycerokinase (PGK) promoter. Both vectors expressed GFP under the control of an internal ribosomal entry site (IRES) element (Fig. 3 A, RRL-δ-G and RRL-N-G). Transduction efficiency on clonogenic cells was ∼90%, as calculated by GFP expression. After 2 d of cultivation, the size of untransduced colonies increased nearly threefold (Fig. 3 B, red circles), whereas the size of C/EBPδ+/GFP+ colonies increased only slightly (Fig. 3 B, yellow circles). Control cells reached confluency 5 d after plating (Fig. 3 B). In contrast, a 5-d culture of C/EBPδ+/GFP+ cells showed well-defined colonies composed of small, tightly packed cells (Fig. 3 B).
Replicative senescence and differentiation of keratinocytes are associated with increased levels of p16INK4A and involucrin, which indicate irreversible exit from the cell cycle and onset of terminal differentiation, respectively (Dellambra et al., 2000). C/EBPδ-transduced cells contained threefold less p16INK4A and involucrin than ΔNGFr-transduced cells (Fig. 3 D). C/EBPδ-transduced cells contained four- and threefold more p27Kip1 and p57Kip2 than control cells, respectively (Fig. 3 D). A cell cycle profile revealed that ∼55, 35, and 10% of the control cells were in the G1, S, and G2–M phases, respectively (Fig. 3 C). In sharp contrast, most of the C/EBPδ-transduced cells were in the G1 phase of the cell cycle (Fig. 3 C). The amount of apoptotic cells was negligible in both C/EBPδ-transduced and control cells (Fig. 3 C). Finally, C/EBPδ-dependent growth inhibition was associated with neither increase of p21Waf1/Cip1 or pRb expression (Fig. 3 D) nor activation of the p53 checkpoint pathway (not depicted). These data indicate that the growth inhibitory effect of C/EBPδ was not due to replicative senescence, terminal differentiation, or apoptosis.
To investigate whether the growth inhibitory effect of C/EBPδ was reversible, we transduced primary limbal cells with a lentiviral vector expressing an N-terminal fusion between C/EBPδ and a modified, 4-hydroxytamoxifen (4OHT)–inducible ligand binding domain of the human ER (Littlewood et al., 1995; Fig. 4 A, RRL-ERδ-G). In mock-transduced (RRL-ER-G) cells, C/EBPδ was found predominantly in the nucleus (Fig. 4 B, ER, middle). In the absence of 4OHT, the ER-C/EBPδ chimeric protein was found in the cytoplasm of transduced limbal cells, and nuclear translocation was observed within 12 h from the addition of 1 μM 4OHT to the culture medium (Fig. 4, B [top], C, and D). In contrast, C/EBPβ was present only in the nucleus, irrespective of the presence of 4OHT (Fig. 4 B [bottom] and Fig. D). Members of the C/EBP family are known to form homo- and heterodimers. In the absence of 4OHT, ER-C/EBPδ sequestered also the endogeneous C/EBPδ in the cytoplasm, as indicated by the absence of nuclear immunofluorescent staining (Fig. 4 D, −4OHT), the absence of endogeneous C/EBPδ in nuclear extracts, and the presence of C/EBPδ in the corresponding cytoplasmic extracts (Fig. 4 B, middle, −4OHT) of RRL-ERδ-G–transduced cells. Colonies of cells transduced with the control vector showed a progressive and linear increase in their size, irrespective of the presence of 4OHT (Fig. 4, E and F). In sharp contrast, the growth of ER-C/EBPδ+/GFP+ colonies was strictly dependent on the localization of the ER-C/EBPδ chimera (Fig. 4, E and F): (1) addition of 4OHT at day 1 considerably slowed the growth of transduced colonies; (2) removal of 4OHT at day 4 was promptly followed by a linear increase of the size of GFP+ colonies; and (3) readdition of 4OHT at day 6 again induced a growth arrest. Of note, untransduced, ΔNGFr- and ER-transduced primary limbal cells duplicated every 17–19 h, whereas C/EBPδ-transduced cells showed a doubling time of 41 h (Fig. 5 E, green). These data show that C/EBPδ lengthened the limbal cell cycle by forcing cells into the G1 phase without altering their capacity for multiplication.
C/EBPδ-dependent mitotic quiescence is mediated by p27Kip1 and p57Kip2
Semiquantitative RT-PCR was performed on control and C/EBPδ-transduced cells using p27Kip1- and p57Kip2-specific primers. As shown in Fig. 5 (A and B), we observed a 5–10-fold increase of both p27Kip1 and p57Kip2 transcripts in RRL-δ-G (C/EBPδ)–transduced limbal cells and RRL-ERδ-G–transduced cells treated with 4OHT (+4OHT), as compared with RRL-N-G (ΔNGFr)–, RRL-ER-G (ER)–, and RRL-ERδ- G–transduced cells not treated with 4OHT (−4OHT). To prove the role of p27Kip1 and p57Kip2 in mediating the effect of C/EBPδ on keratinocyte cell cycle, C/EBPδ-transduced limbal cells were transfected with siRNAs specifically targeted to the p27Kip1 and p57Kip2 mRNAs. Transfection efficiency was 84 ± 2 and 77 ± 3%, respectively (Fig. 5 C). Western blot analysis showed that siRNA-p27Kip1 and siRNA-p57Kip2 caused a strong decrease of the expression of p27Kip1 and p57Kip2, but not of C/EBPδ and a control protein (Fig. 5 D).
Untransduced and C/EBPδ-transduced cells showed a doubling time of 18 and 41 h, respectively (Fig. 5 E). C/EBPδ-transduced cells transfected with either siRNA-p27Kip1 or siRNA-p57Kip2 showed a doubling time of 23.5 and 23 h, respectively. Of note, C/EBPδ-transduced cells transfected with both siRNAs simultaneously, showed a doubling time of 18.5 h, a value undistinguishable from that of untransduced control cells (Fig. 5 E). These data show that p27Kip1 and p57Kip2 mediate C/EBPδ-induced mitotic quiescence.
C/EBPδ binds to the p63, involucrin, p27Kip1, p57Kip2, and p16INK4A loci in vivo
To provide evidence for a direct contribution of C/EBPδ in regulating the expression of ΔNp63α, involucrin, p27Kip1, p57Kip2, and p16INK4A, we analyzed recruitment of C/EBPδ to these loci by a chromatin immunoprecipitation (ChIP) assay on cultured limbal keratinocytes three and five passages after transduction with either RRL-δ-G or the control, RRL-N-G vector. Protein–DNA complexes were immunoprecipitated with antibodies specific for C/EBPδ or the Flag epitope and with control IgGs. Immunoprecipitated chromatin DNA was analyzed by PCR with primers specific for different regions of the p63, involucrin, p27Kip1, p57Kip2, and p16INK4A loci (Fig. 6 A, red arrowheads), containing evolutionarily conserved and/or putative C/EBP binding elements.
In C/EBPδ-transduced cells, vector-derived (Flag-tagged) C/EBPδ was found associated to the p63 locus in intron 3 (at position +147873 to +148041) and in an evolutionarily conserved, keratinocyte-specific enhancer in intron 5 (+202579 to +202761; Antonini et al., 2006). Primers designed to amplify other sequences from the p63 locus detected the correct fragment only in the input samples (Fig. 6, A and B). Binding of Flag-tagged C/EBPδ was also observed to a region upstream of the involucrin promoter (−421 to −119) containing a C/EBP responsive element previously characterized in keratinocytes (Agarwal et al., 1999; Balasubramanian and Eckert, 2004) and upstream of the p27Kip1 (−227 to +14), p57Kip2 (−622 to −398), and p16INK4A (−1020 to −871) loci (Fig. 6 B). Binding to all these sites was observed specifically in C/EBPδ-transduced cells and was more pronounced at the fifth than at the third passage (Fig. 6 B). The signals obtained with the anti-Flag antibody were always weaker than those obtained with the anti-C/EBPδ antibody, probably reflecting a lower immunoprecipitation efficiency. In control, ΔNFGr-transduced cells, a weak but specific signal was observed at the p63, involucrin, and p27Kip1 loci in chromatin immunoprecipitated with the anti- C/EBPδ but not the anti-Flag antibody. Binding was observed at the third but not at the fifth passage (Fig. 6 B), most likely as a result of the presence of endogenous C/EBPδ activity in a subset of early passage cells, which is lost in later passages.
Chromatin from the same cells was also immunoprecipitated with antibodies specific for all isoforms or only the α isoforms of p63. Binding of ΔNp63α was observed in the intron 5 enhancer of the p63 locus (Antonini et al., 2006) in both C/EBPδ-transduced and control cells. Binding was more pronounced in C/EBPδ+ than in control cells, reflecting either an increased recruitment of ΔNp63α to the enhancer or simply the increased proportion of cells expressing ΔNp63α in these cultures. Interestingly, ΔNp63α and C/EBPδ appear to bind the same regions in the p63 and p27Kip1 loci (Fig. 6 B). These results suggest that the p63, involucrin, p27Kip1, p57Kip2, and p16INK4A loci might be direct targets of C/EBPδ activity, in some cases in combination with ΔNp63α.
C/EBPδ promotes self-renewal of holoclone-forming cells
Clonogenic ability and proliferative potential are distinct properties of epithelial cells. Keratinocyte stem cells are endowed with high clonogenic and high proliferative capacity, and self-renewal occurs when both properties are maintained. Conversely, TA cells are clonogenic but have a limited capacity for multiplication.
Serially cultivated, untransduced, or ΔNFGr-transduced limbal cells showed a progressive decrease of their clonogenic capacity (Fig. 7, A and C) and ceased to proliferate after 60–75 d (or 9–11 passages) in culture (Fig. 7 B). Replicative senescence occurs because of clonal evolution, as indicated by the progressive increase of aborted, paraclone-type colonies (Fig. 7 D) and by the replacement of ΔNp63α with ΔNp63β and ΔNp63γ expression (Fig. 2 B and Fig. 7, E and F). In sharp contrast, both clonogenic ability (Fig. 7, A and C) and proliferative capacity (Fig. 7 B) of C/EBPδ-transduced cells were maintained indefinitely. This effect was due to the capacity of enforced C/EBPδ expression to promote self-renewal and halt clonal evolution in holoclones, as indicated by the following evidence: (1) serially cultivated C/EBPδ-transduced cells showed no increase in the number of paraclones (Fig. 7 D) or replacement of ΔNp63α with ΔNp63β and ΔNp63γ (Fig. 7, E and F); (2) statistical analysis of cell size (Di Iorio et al., 2006), a major marker of clonogenic stem cells (Barrandon and Green, 1985), showed that C/EBPδ-transduced cells were nearly 10-fold smaller than control cells (325.93 vs. 3,035.25 μm3); (3) clonal analysis revealed that the percentage of holoclone-forming cells decreased and eventually set to zero in serially cultivated control cells but remained constant in C/EBPδ-transduced cells (10–15% of inoculated cells); and (4) ER- C/EBPδ was able to fully sequester also endogeneous C/EBPδ in the cytoplasm of limbal cells in the absence of 4OHT (Fig. 4 B). Such cells ceased to express ΔNp63α (not depicted) and underwent replicative senescence in only two passages as compared with 9–11 passages of control untransduced cells (Fig. 7 B).
To investigate whether C/EBPδ was able to rescue TA cells from their terminal fate, we transduced different single cell–derived clones. As expected, holoclone, meroclone, and paraclone type clones displayed a progressive decrease in clonogenicity and ΔNp63α content (Fig. 8). Forced expression of C/EBPδ was able to sustain the self-renewal of holoclones and meroclones and, hence, of clones still containing ΔNp63α+ cells, but not that of ΔNp63α− paraclones (Fig. 8). All cells in transduced holoclones and meroclones expressed ΔNp63α (not depicted), further suggesting that C/EBPδ is able to foster self-renewal only of ΔNp63α+ cells and to maintain expression of ΔNp63α in such cells.
These data prompted us to investigate whether forced expression of ΔNp63α was sufficient to sustain limbal cell self-renewal. Primary limbal cultures and single cell–derived clones were infected with a lentiviral vector expressing the ΔNp63α isoform (Fig. 8, RRL-ΔNα-G). ΔNp63α-transduced holoclones underwent regular clonal evolution and ceased to proliferate after 11 passages, a value identical to control untransduced cells (Figs. 7 and 8). ΔNp63α was therefore unable to sustain limbal stem cell self-renewal both in primary cultures (unpublished data) and in clones. Finally, simultaneous infection with lentiviral vectors expressing C/EBPδ and ΔNp63α was unable to rescue clonogenic ability and self-renewal in paraclones, suggesting that loss of self-renewal is an irreversible process, at least in limbal keratinocytes.
Exceptional progress has been made in understanding the molecular mechanisms regulating keratinocyte stem cells. The role of transcription factors, such as p63, tcf3, CCAAT displacement protein, and GATA-3, and of adhesion and signaling molecules, such as integrins, Wnt/β-catenin, c-Myc, Notch, hedgehog, Sgk3, and bone morphogenic proteins, in controlling hair follicle and epidermal development and stem cell fate has been highlighted (Niemann and Watt, 2002; Fuchs et al., 2004; Cotsarelis, 2006; Blanpain et al., 2007). Molecular phenotyping of some of the keratinocyte stem cell niches helped explain how stem cells interact with the microenvironment to maintain their properties (Morris et al., 2004; Tumbar et al., 2004). Little is known, however, on the regulation of perhaps the most important property of epithelial stem cells, that is, their capacity to self-renew. It has been shown that the Rho guanosine triphosphatase Rac1 sustains murine epidermal stem cell renewal and human epidermal stem cell clonogenicity by negatively regulating MYC (Benitah et al., 2005). However, differences exist between different lining epithelia and among animal species. For instance, Rac1 stimulates differentiation and not self-renewal in the intestinal epithelium (Stappenbeck and Gordon, 2000), whereas the CD34 antigen identifies murine but not human hair follicle stem cells (Cotsarelis, 2006).
We took advantage of the availability of human corneas to carry out genetic manipulation experiments on primary, clonogenic limbal stem cells and show that C/EBPδ plays a key role in regulating their cell cycle and self-renewal properties. Our findings are graphically summarized in Fig. 9. According to this model, a defined number of mitotically quiescent limbal stem cells coexpress Bmi1, ΔNp63α, and C/EBPδ under normal homeostasis. Coexpression of Bmi1, ΔNp63α, and C/EBPδ therefore identifies limbal holoclones and is part of the genetic program maintaining stem cell identity. Bmi1 fosters self-renewal of haematopoetic and neural stem cells through regulation of the p16INK4A and p19ARF pathways (Lessard and Sauvageau, 2003; Molofsky et al., 2003, 2005; Park et al., 2003; Walkley et al., 2005) and might play a similar role also in limbal stem cells. ΔNp63α sustains the proliferative potential of stem cells in several stratified epithelia, including the cornea (Parsa et al., 1999; Pellegrini et al., 2001; Koster et al., 2004; McKeon, 2004; Di Iorio et al., 2005; Nguyen et al., 2006). We show here that C/EBPδ regulates mitotic quiescence of limbal keratinocytes by forcing cells in the G0/G1 phase of the cell cycle. Even under culture conditions specifically designed to promote keratinocyte proliferation, forced C/EBPδ expression greatly increases the cell cycle length through activation of the cell cycle inhibitors p27Kip1 and p57Kip2. The growth inhibitory effect of C/EBPδ is not due to replicative senescence or terminal differentiation, as confirmed by the down-regulation of p16INK4A and involucrin. Perhaps more important, C/EBPδ promotes the self-renewal of ΔNp63α+ limbal stem cells, as suggested by the block of clonal evolution and the indefinite maintenance of the number of holoclones during serial cultivation of C/EBPδ-transduced limbal keratinocytes.
Stem cells are capable of shifting from a homeostatic state of relative quiescence to rapid proliferation under specific conditions (activation). In the ocular surface, this shift occurs upon central corneal wounding (Lehrer et al., 1998; Di Iorio et al., 2005). This explains the apparently opposing actions of C/EBPδ and ΔNp63α. On one hand, C/EBPδ induces mitotic quiescence (through a positive regulation of p27Kip1 and p57Kip2) and self-renewal of limbal stem cells; on the other, it preserves their proliferative potential (essential for stem cell–dependent tissue regeneration) through a positive regulation of ΔNp63α. In this way, when some limbal stem cells are released from C/EBPδ-dependent mitotic constraints, as in a corneal damage, they can unchain their remarkable p63-dependent proliferative capacity, multiply, and migrate to repair a corneal wound. This process is, however, irreversible and leads to limbal stem cell terminal differentiation (Fig. 9). Our data therefore strengthen the notion that proliferation and self-renewal capabilities are two related, albeit distinct, processes. At least in human limbal stem cells, proliferation potential relies on the expression of ΔNp63α, whereas self-renewal requires also C/EBPδ. Similarly, Bmi1 is essential for the self-renewal of neural stem cells but does not influence the proliferative capacity of their committed progeny (Molofsky et al., 2003). The notion that ΔNp63α induces the expression of growth factor receptors and adhesion molecules regulating survival and motility of epithelial cells (Carroll et al., 2006) is consistent with our proposed model.
Our data establish an interesting parallel with the hematopoietic system, where quiescence and self-renewal of stem cells have been recently shown to be linked and regulated by p27Kip1, p57Kip2, and Mad1 (Scandura et al., 2004; Walkley et al., 2005; Yamazaki et al., 2006). Indeed, loss of p27Kip1 allows relatively quiescent hematopoietic stem cells to rapidly enter the cell cycle to restore haematopoiesis (Walkley et al., 2005). Finally, we show that C/EBPδ is directly associated in vivo, alone or in combination with ΔNp63α, to chromatin-surrounding promoters or regulatory elements of the p63, p27Kip1, p57Kip2, and p16INK4A loci, suggesting a direct role of this transcription factor in determining the genetic program of self-renewing stem cells.
The role of C/EBPδ described here is intriguing. Indeed, C/EBPs have been mainly related to cellular differentiation. C/EBPα, -β, and -δ are instrumental in regulating adipogenesis, whereas C/EBPα, -ε, and -β orchestrate myeloid differentiation into mature neutrophils, atypical neutrophils, and macrophages (Rosen et al., 2000; Ramji and Foka, 2002), and C/EBPδ regulates learning and long-term memory in the central nervous system (Sterneck et al., 1998; Taubenfeld et al., 2001). The importance of the C/EBP family in cellular differentiation also extends to other cell types, including hepatocytes, ovarian luteal cells, intestinal epithelial cells, and epidermal keratinocytes. For instance, it has been shown that C/EBPα and -β induces cell cycle exit in normal keratinocytes and positively regulates the program of squamous differentiation in the epidermis (Oh and Smart, 1998; Zhu et al., 1999). However, C/EBPβ promotes keratinocyte proliferation and skin tumor formation in the presence of oncogenic Ras or in response to carcinogens (Zhu et al., 2002; Sterneck et al., 2006) and fosters hepatocyte proliferation during liver regeneration after partial hepatectomy (Greenbaum et al., 1998). Mammary epithelial cells from C/EBPβ-deficient mice have a proliferation defect that leads to impaired ductal morphogenesis and a failure to lactate (Robinson et al., 1998; Seagroves et al., 1998), and ectopic C/EBPβ expression in human mammary epithelial cells induces hyperproliferation and a partially transformed phenotype (Bundy and Sealy, 2003). Finally, C/EBPδ induces late differentiation events in epidermal keratinocytes (Smith et al., 2004) and is indeed detected in the subrabasal layers of the human epidermis (unpublished data). Therefore, the biological effects of C/EBPs appear to be highly species and cell context specific, suggesting that role that C/EBPδ exerts in the human corneal epithelium might not necessarily be observed in other squamous epthelia.
The mechanisms controlling C/EBPδ expression and function in the limbus, as well as the downstream mediators of C/EBPδ activity in controlling stem cell quiescence and self-renewal, remain to be determined. The expression of the C/EBPs has been found to change markedly during several physiological and pathophysiological conditions through the action of extracellular signals. C/EBPs are subject to extensive species- and tissue-specific posttranscriptional regulation and phosphorylation-mediated changes in DNA binding activity and nuclear localization (Ramji and Foka, 2002). Furthermore, the different C/EBP proteins are able to form heterodimers in all intrafamilial combinations and to associate with other factors (Ramji and Foka, 2002). A combination of biochemical, cellular, and genetic experiments is necessary to acquire a more comprehensive description of upstream regulators and downstream targets of C/EBPδ and to elucidate the networks of protein interactions and regulatory pathways that control its activity in human limbal stem cells.
Materials And Methods
Human specimens, cell culture, and cell cycle analysis
Corneas taken from organ donors and considered unsuitable for transplantation (solely because of hepatitis seropositivity of the donor) were examined with a slit lamp immediately before retrieval and classified as resting corneas, which did not show epithelial defect, dehydration, edema, or inflammation, or activated corneas, which had central corneal epithelial defects and/or abrasions, usually as a result of incomplete closure of the eyelids after death. Resting and activated corneas were taken 3.93 ± 0.69 and 6.79 ± 2.9 h from death, respectively. Corneas were provided by D. Ponzin and A. Ruzza (The Veneto Eye Bank Foundation, Venice, Italy).
Swiss mouse 3T3-J2 cells (a gift from H. Green, Harvard Medical School, Boston, MA) were grown in DME supplemented with 10% calf serum. Keratinocytes were cultivated on a feeder layer of lethally irradiated 3T3-J2 cells, and colony forming efficiency (CFE) assays and calculation of the number of cell generations and population doublings were performed as described previously (Pellegrini et al., 1999a; Dellambra et al., 2000). Clonal analysis was performed from subconfluent primary cultures as described previously (Pellegrini et al., 1999a, 2001). In brief, single cells were inoculated onto multiwell plates containing a feeder layer of 3T3 cells. Clones were identified after 7 d of culture under an inverted microscope and transferred to replicate dishes. One dish (1/4 of the clone) was fixed 9–12 d later and stained with rhodamine B for clonal type classification (Barrandon and Green, 1987; Pellegrini et al., 1999a). The second dish was used for further experiments and analyses. In selected experiments, 100 limbal cells were plated in 100-cm dishes and cultured for 1 wk. Colonies were then examined under a microscope (Axiovert 200 M; Carl Zeiss MicroImaging, Inc.): large round colonies with smooth and regular borders and formed entirely by small cells with scarce cytoplasm were classified as holoclones (Di Iorio et al., 2005) and were subjected to immunofluorescence.
For cell cycle analysis, subconfluent keratinocyte cultures were trypsinized and fixed in 70% ethanol at 4°C. Samples (106 cells) were rehydrated in PBS/1% FCS at room temperature for 10 min and stained with 20 μg/ml propidium iodine for 30 min at 4°C. Flow cytometry was performed using a LSR II FACScan (Becton Dickinson).
Immunofluorescence and Western analysis
The following antibodies were used: rabbit anti-C/EBPδ, anti-RasGAP, anti-Rb, and p57Kip2 purified IgG (Santa Cruz Biotechnology, Inc.); 4A4 pan-p63 mAb (BD Biosciences); p16INK4A, p21Waf1/Cip1, and p27Kip1 mAbs (Exalpha Biologicals, Inc.); involucrin and Ki67 mAbs (Novocastra); Bmi1 mAb (Upstate Biotechnology); rabbit anti-p63α unconjugated and FITC-conjugated purified IgG raised against a synthetic peptide (NH2-DFNFDMDARRNKQQRIKEEGE-COOH) comprising the C terminus post-SAM domain of p63α (Primm; Di Iorio et al., 2005). Secondary rhodamine- or FITC-labeled antibodies were obtained from Santa Cruz Biotechnology, Inc. For immunofluorescence analysis, keratinocyte colonies were fixed (3% paraformaldehyde/2% sucrose in PBS, pH 7.6), permeabilized (0.5% Triton X-100 in PBS), and coated with 0.5% BSA/PBS for 1 h at RT. Paraformaldehyde-fixed corneal samples were embedded in OCT, frozen, and sectioned. Immunofluorescence was performed on fixed colonies and 5–7-μm corneal sections as described previously (Di Iorio et al., 2005). Confocal analyses were done with a confocal analyzer (LSM510-META; Carl Zeiss MicroImaging, Inc.). Multitrack analysis was used for image acquisition. For immunoblots, mass or clonal cultures were extracted on ice with RIPA buffer (0.15 mM NaCl/0.05 mM Tris/HCl, pH 7.5/1% Triton X-100/1% sodium deoxycholate/0.1% SDS). Nuclear and cytoplasmic protein extraction was performed using the NE-PER Nuclear and Cytoplasmic Extraction kit (Pierce Chemical Co.) following conditions supplied by the manufacturer. Equal amounts of samples were electrophoresed on 7.5% SDS-polyacrylamide gels and transferred to polyvinylidene difluoride filters (Immobilon-P; Millipore). Immunoreactions were performed as described previously (Pellegrini et al., 2001) using antibodies at a 1:500 dilution. Immobilon bound antibodies were detected by chemiluminescence with ECL (GE Healthcare).
Total RNA was extracted from keratinocyte cultures, purified with RNase Micro kit (QIAGEN), and quantified by spectrophotometry. RT-PCR was performed using the One Step RT-PCR kit (QIAGEN). cDNAs were synthesized from 0.5–2 μg of total RNA, and PCR reactions were performed using 20, 24, 28, 32, 36, and 40 cycles. β-Actin was used for normalization. Ethidium bromide–stained agarose gels were visualized with an Image Station 440 CF (Kodak). Quantification was performed using 1D 3.5 software (Kodak). Primer sequences for p63 isoforms and annealing temperatures were as described previously (Di Iorio et al., 2005). The following primers and annealing temperatures were used for p27, p57, and β-actin RT-PCR: p27Kip1, 5′-AGTGTCTAACGGGAGCCCTA-3′ and 3′-GTCCATTCCATGAAGTCAGC-5′ (annealing temperature 60°C, 829 bp); p57Kip2, 5′-CACGATGGAGCGTCTTGTC-3′ and 3′-CTTCTCAGGCGCTGATCTCT-5′ (annealing temperature 60°C, 699 bp); and β-actin, 5′-GAGCGCAAGTACTCCGTGT-3′ and 3′-ACGAAGGCTCATCATTCAAA-5′ (annealing temperature 58°C, 548 bp).
The human C/EBPδ cDNA was cloned by RT-PCR from total RNA extracted from the THP1 cell line using specifically designed primers containing EcoR1 recognition sequences. To generate a N-terminal Flag epitope–tagged C/EBPδ sequence, the EcoR1 C/EBPδ cDNA fragment was subcloned into the pFlagCMV-2 plasmid (Sigma-Aldrich), after inserting an EcoRV restriction site at position 913 by PCR (for all primers, see Table S1, available at http://www.jcb.org/cgi/content/full/jcb.200703003/DC1). To generate a Flag-tagged ER-C/EBPδ fusion sequence, a MfeI–EcoRI PCR–amplified fragment containing the modified ligand binding domain of the human ER (Littlewood et al., 1995) was fused to the C terminus of the Flag epitope before inserting the C/EBPδ cDNA. The Flag-C/EBPδ, Flag-ER- C/EBPδ, and Flag-ER cassettes were extracted as EcoRV fragments and cloned downstream the human PGK promoter and upstream an IRES-EGFP cassette into the blunted SmaI–BamHI sites of the pRRL.ppt.PGK.IRES.GFP.WPRE lentiviral vector (Urbinati et al., 2005), to obtain the RRL-δ-G, RRL-ERδ-G, and RRL-ER-G vectors. The control RRL-N-G vector was generated by cloning an NcoI–EcoRV fragment encoding a truncated form of the low-affinity, p75 NGFr (ΔNGFr; Bonini et al., 1997) into the same vector backbone. The RRL-ΔNα-G vector was obtained by inserting the cDNA of ΔNp63α isoform (Yang et al., 1998) into the same vector backbone.
Lentiviral stocks pseudotyped with the vesicular stomatitis G protein (VSV-G) were prepared by transient cotransfection of 293T cells using a three-plasmid system (the transfer vector and the helper plasmids pCMVΔR8.74, encoding Gag, Pol, Tat, and Rev, and pMD.G, encoding VSV-G), as previously described (Dull et al., 1998). Viral titers were determined by transduction of HeLa cells with serial dilution of the vector stocks and ranged from 107 to 108 TU/ml. Transduction efficiency was evaluated by scoring GFP and/or ΔNGFr transgene expression by flow cytometry.
Transduction of limbal keratinocytes
Subconfluent primary or clonal limbal cultures were trypsinized. 2 × 104 cells were resuspended in 1 ml of culture medium containing 8 mg/ml polybrene and transduced with lentiviral vector stocks at a MOI of 25, overnight at 37°C. Gene transfer efficiency was assessed 4 d after transduction by scoring GFP+ cells by confocal fluorescence microscopy (LSM510-META; Carl Zeiss MicroImaging, Inc.). In selected experiments (Fig. 4), 1 μM 4OHT was added every 12 h for 3 d to RRL-ERδ-G–transduced cultures. 4OHT was then removed from the culture medium for 2 d and readded until control cultures reached confluency.
Transfection of siRNA-p27 and siRNA-p57
These experiments were performed using the siRNA-p27 and siRNA-p57 RNAi Human/Mouse Starter kit (QIAGEN). The siRNA duplexes were designed using the HiPerformance Design Algorithm licensed from Novartis AG, integrated with a stringent in-house homology analysis tool. Double-stranded RNAs were synthesized by QIAGEN. Primary cultured human limbal epithelial cells previously transduced with a lentiviral vector carrying the C/EBPδ cDNA (RRL-δ-G) were plated into 24-well plates at 4 × 104 cells/cm2. 48 h later, cells were transfected using fluorescently labeled p27Kip1 (Alexa Fluor) and p57Kip2 (Cy3) siRNAs at a final concentration of 67 nM (siRNA to Hiperfect reagent ratio 1:6) and incubated under normal growth conditions (37°C and 5% CO2). Transfection efficiency was determined 24 h after siRNA addition through laser-scanning confocal microscope (LSM510-META) analysis. Nonsilencing siRNAs (Alexa Fluor) were used as negative controls at the same conditions of transfection (67 nM; 1:6 ratio) and cell density.
ChIP assays were performed essentially as described previously (Testa et al., 2005). Chromatin was prepared from 107 limbal keratinocytes at the third and the fifth passage after transduction with either the RRL-δ-G or the RRL-N-G vector. Nuclear extracts were sonicated to obtain DNA fragments ranging from 400 to 800 bp in length. The equivalent of ∼5 × 105 cells was immunoprecipitated with rabbit anti-CEBPδ (Santa Cruz Biotechnology, Inc.), mouse anti-Flag (Sigma-Aldrich), mouse anti-p63 (4A4 pan63; BD Biosciences), and rabbit anti-p63α antibodies. Immunoprecipitations with mouse and rabbit IgGs (BD Biosciences) were included as controls. Immunoprecipitated DNA was analyzed by PCR with primers spanning regions containing known or putative CEBPδ and p53/p63 binding motifs within the genomic loci of p63 (from position +70720 to +70951, +147873 to +148041, +151191 to +151408, +202579 to +202761, and +234830 to +235057 from the transcription start site), involucrin (−421 to −119), p27Kip1 (−227 to +14), p57Kip2 (−622 to −398), and p16INK4A (−1020 to −871). Specific primers are listed in Table S1.
Online supplemental material
Table S1 gives the primers used in this study.
This work was supported by grants from the Ministry of Health, the Istituto Superiore di Sanità, and the European Commission, VI Framework Program, Cornea Engineering (grant NMP2-CT-2003-504017) and Skintherapy (grant LSHB-CT-2005-512073).
V. Barbaro, A. Testa, and E. Di Iorio contributed equally to this paper.
Abbreviations used in this paper: 4OHT, 4-hydroxytamoxifen; C/EBP, CCAAT enhancer binding protein; CFE, colony forming efficiency; ChIP, chromatin immunoprecipitation; ER, estrogen receptor; IRES, internal ribosomal entry site; NGFr, NGF receptor; PGK, phosphoglycerokinase; TA, transit amplifying.