CD133 is a marker for long-term repopulating murine epidermal stem cells.

Maintenance, repair, and renewal of the epidermis are thought to depend on a pool of dedicated epidermal stem cells (EpiSCs). Like for many somatic tissues, isolation of a nearly pure population of stem cells is a primary goal in cutaneous biology. We used a quantitative transplantation assay, using injection of keratinocytes into subcutis combined with limiting dilution analysis, to assess the long-term repopulating ability of putative murine EpiSC populations. Putative EpiSC populations were isolated by FACS sorting. The CD133(+) population and the subpopulation of CD133(+) cells that exhibits high mitochondrial membrane potential (DΨm(hi)) were enriched for long-term repopulating EpiSCs versus unfractionated cells (3.9- and 5.2-fold, respectively). Evidence for self-renewal capacity was obtained by serial transplantation of long-term epidermal repopulating units derived from CD133(+) and CD133(+)ΔΨm(hi) keratinocytes. CD133(+) keratinocytes were multipotent and produced significantly more hair follicles than CD133(-) cells. CD133(+) cells were a subset of the previously described integrin α6(+)CD34(+) bulge cell population, and 28.9±8.6% were label-retaining cells. Thus, murine keratinocytes within the CD133(+) and CD133(+)ΔΨm(hi) populations contain EpiSCs that regenerate the epidermis for the long term, are self-renewing, multipotent, and label-retaining cells.

INTRODUCTION

Murine epidermis is maintained by tissue stem cells that can be defined by their long-term repopulating and self-renewal abilities. These defining features are essential to the identification and characterization of epidermal stem cells (EpiSCs) and their progeny.

Keratinocytes spontaneously form cysts with a differentiated keratinizing epidermis, following traumatic or surgical implantation into the human subcutis (Hall et al., 2006; Silver and Ho, 2003; Ohnishi and Watanabe, 1999). Production of such cysts has been used to study conjunctival epithelium (Doran et al., 1980; Wei et al., 1997), lung alveolar cells (Yu et al., 2007) and epidermal keratinocytes (Grimwood et al., 1988; Zheng et al., 2005). Epidermal keratinocytes produced stratified squamous structures with keratohyalin granules, stratum corneum, basement membrane, and protein expression indicating a fully differentiated epidermis (Doran et al., 1980; Grimwood et al., 1988). Thus, subcutaneous injection of keratinocytes into mice results in the phenotype and differentiation pattern of the original epidermis.

CD133+ marked human progenitors in kidney (Bussolati et al., 2005), nervous system (Uchida et al., 2000) and in epithelial tissues including prostate (Richardson et al., 2004), foreskin (Yu et al., 2002; Mizrak et al., 2008; Guo and Jahoda, 2009) and colorectal adenocarcinoma (Corbeil et al., 2000). CD133 marked murine progenitors in neural cells (Corti et al., 2007; Coskun et al., 2008), liver (Rountree et al., 2007), kidney (Weigmann et al., 1997) and intestine (Snippert et al., 2009; Zhu et al., 2009). We selected CD133 as a possible marker of murine EpiSCs.

Murine embryonic stem cells with high mitochondrial membrane potential (ΔΨm) show decreased differentiation and increased teratoma formation (Schieke et al., 2008). TMRM, a fluorescent derivative of R123, was used to isolate putative EpiSCs based on high ΔΨm, as it provides more accurate quantification than the parent compound (Scaduto et al., 1999).

In both human and murine epidermis, integrin α6hiCD71lo keratinocytes showed features of stem cells (Kaur and Li, 2000; Li et al., 1998; Tani et al., 2000).

In this study, epidermal allografts were produced by injection of putative EpiSCs into murine subcutis. The frequency of long-term repopulating EpiSCs was determined by limiting dilution analysis (Schneider et al., 2003; Charruyer et al., 2009; Strachan et al., 2008). Enrichment for murine EpiSCs was 3.9- fold over unfractionated (UNF) keratinocytes in CD133+ keratinocytes and 5.2-fold in CD133+ΨΔmhi keratinocytes. CD133+ and CD133+ΨΔmhi keratinocytes displayed superior long-term repopulating and self-renewal ability, multipotency, and label retention.

RESULTS

Generation of murine epidermal repopulating units in an allograft model

Epidermal structures generated from injection of keratinocytes were termed epidermal repopulating units (ERUs) pursuant to hematologic terminology (Szilvassy et al., 1990). After injection of GFP+ keratinocytes, 100% of 100 ERUs analyzed were GFP+, confirming derivation from GFP+ keratinocytes injected (Figure 1a).

Figure 1

Generation of ERUs following subcutaneous injection of murine keratinocytes in an allograft model

(a) ERUs formed following injection of GFP+ murine keratinocytes into a GFP− NOD/SCID mouse are GFP+ confirming derivation from the GFP+ donor keratinocytes. Fluorescence microscopy (488 nm) (left panel). H&E of an adjacent section (right panel). 100 ERUs were analyzed (3 experiments). (b–d) Epidermal differentiation is seen in ERUs. (b) Keratin 14 is expressed in the basal layers. (c) Involucrin is expressed in the suprabasal epidermis. (d) Laminin is present at the basement membrane of the ERU. Positive controls were intact murine epidermis. Negative controls were performed with omission of the primary antibody. (e–g) ERUs are formed from single cells. Keratinocytes were labeled with Vybrant™DiI or DiO and mixed in a 1:1 ratio before injection into NOD/SCID mice. Resultant ERUs (arrow heads) were green or red, but not mixed. (e) Four of the ERUs produced by injection of 20,000 keratinocytes. Fluorescence microscopy (left panel), DCI of the same section (middle panel) and H&E of the adjacent section (right panel). (f) A DiO positive ERU and (g) a DiI positive ERU. Scale bars = 10 µm in a–d; 25 µm in e; and 10 µm in f and g.

Immunohistochemistry showed keratin 14 in the basal layers (Figure 1b) and involucrin in the suprabasal layers (Figure 1c). Linear fluorescence at the ERU periphery of the ERU with anti-laminin antibody (Figure 1d) indicated a basement membrane.

To determine whether ERUs originated from single cells, keratinocytes incubated with Vybrant™ DiI (565nm, red) or DiO (501nm, green) were mixed 1:1. One week after injection of 100,000, 40,000, 20,000 or 6,250 keratinocytes, 17/79 (21%), 5/46 (10%), 0 (0%) and 0 (0%) bi-colored ERUs, respectively (Figure 1e, f and g), indicating that at lower doses ERUs originated from a single cell.

In this in vivo repopulation assay only EpiSCs originally injected (and their progeny) persist after 9 weeks, while transit amplifying cells (TACs) (and their progeny) are no longer present

Long-term repopulation combined with limiting dilution analysis has been used to quantify EpiSCs (Schneider et al., 2003; Charruyer et al., 2009; Strachan et al., 2008). As short-term repopulating cells exhaust their proliferative ability over time, the frequency of ERUs decreases. When only ERUs from long-term repopulating keratinocytes remain, ERU frequency does not change at subsequent time points. Here we used a transplantation assay modified from previous studies (Schneider et al., 2003; Strachan et al., 2008). UNF keratinocytes were injected at a range of doses (1–100,000 cells) and the frequency of ERUs determined at different repopulation times by limiting dilution analysis (Table1). The frequency of ERUs decreased between 1 and 6 weeks (P≤0.001; n=5) from 1 in 48 (SE 1 in 35–66) to 1 in 790 (SE 1 in 576–1,084). After 6 weeks no significant change in the frequency of ERUs was detected. The overall likelihood ratio test for differences in EpiSC frequencies between weeks 1 and 24 yielded a significant result (P<0.001), but between weeks 6–24 (P = 0.25), 9–24 (P = 0.63), 12–24 (P=0.34), and 18–24 (P=0.88) yielded non-significant results. In previous transplantation studies (Schneider et al., 2003; Strachan et al., 2008) no ERUs were lost after 9 weeks and no significant change was found after 6 weeks in the present model. Therefore, we selected 9 weeks as the time at which we are studying ERUs derived only from EpiSCs.

Table 1

Frequency of epidermal repopulating units at different durations of repopulation

Duration ofrepopulation		1 week	6 weeks	9 weeks	12 weeks	18 weeks	24 weeks
Celldose
		Response ratio: positive samples/total number of injections
100,000		1/1	5/5	3/4	5/5	4/4	5/6
25,000		4/4	4/4	3/3	7/7	3/3	4/4
6,250		4/4	9/9	6/7	9/10	2/2	3/3
1,562		4/4	8/11	10/13	7/11	3/7	4/11
391		4/4	3/6	3/7	3/7	3/5	6/12
98		7/9	2/4	0/4	0/5	1/3	2/7
24		6/11	1/4	0/1	0/1		2/3
5		0/1
1		0/2
ERU FrequencySE1, 2		1 in 48**35–66	1 in 790576–1,084	1 in 1,4911,109–2,002	1 in 1,8021,342–2,420	1 in 1,373917–2,054	1 in 1,242921–1,674

p ≤ 0.001 between 1 week and all the other time points

SE expressed as 1 in…

Tests for inconsistency (CHI-Square Pearson, Deviance) were NS

CD133+ and CD133+ΔΨmhi keratinocytes are located in the bulge and are enriched for long-term repopulating EpiSCs

CD133 immunostaining was located in the bulge of neonatal and adult hair follicles, as seen in humans (Jiang et al., 2010) (Figure 2a and b).

Figure 2

CD133+ and CD133+ΔΨmhi keratinocytes are located in the murine bulge and are enriched for long-term repopulating epidermal stem cells

. (c) Isolation of CD133+ and (d) isolation of CD133+ΔΨmhi murine keratinocytes based on isotype control. (e) Isolation of integrin α6hiCD71lo murine keratinocytes. The selected population was based on the 7–10% of cells most α6 integrinhiCD71lo (putative EpiSCs). A second population (right upper quadrant) was also selected based on the 7–10% of cells most α6 integrinhiCD71hi (putative TACs) (Kaur and Li, 2000; Li et al., 2004; Li et al., 1998; Li, 2005) (based on isotype controls). (f) ERUs formed following injection of CD133+ keratinocytes into a NOD/SCID mouse. (g) EpiSC frequency, based on in vivo transplantation combined with limiting dilution analysis, in isolated keratinocyte populations at 9 weeks. (h) Summary of the data in Figure 2g. Bar graph showing the fold enrichment over UNF keratinocytes for the selected populations. Error bars = mean±SE (* P≤0.05; **P≤0.001; n=3). Scale bars= 10 µm for a, b, f. bg= bulge

For FACS isolation of CD133+ keratinocytes, an isotype control was used to set a gate resulting in < 1% of total cells in the positive gate. 7AAD was used to exclude dead cells. In day 4 neonates, the CD133+ keratinocytes constituted 2 – 7.6% of total (mean=4.2±3.1%, n=7) (Figure 2c). In 10–12 week adults, the CD133+ keratinocyte population was not significantly different (2–4.2% of total cells, mean=2.8±1%, P=0.4, n=4).

High membrane potential (OΨmhi) and low membrane potential (OΨmlo) were defined as 5% highest and lowest TMRM fluorescence, as previously (Schieke et al., 2008; Schieke et al., 2006). Approximately 2% of total cells were CD133+ OΨmhi. Therefore, we selected the 2% CD133+ OΨmhi and 2% CD133− OΨmlo populations. (Figure 2d).

Integrin α6hiCD71lo/hi keratinocytes (7–10% total) were selected based on appropriate isotype controls (Figure 2e), as previously described (Tani et al., 2000; Li et al., 2004; Youn et al., 2004).

When injected into the subcutis of NOD/SCID mice, CD133+ cells produced ERUs with keratinized epidermis, indistinguishable from those from UNF cells (Figure 2f). 1 in 379 (SE 1 in 274–526) CD133+ keratinocytes was an EpiSC capable of long-term repopulation, vs. only 1 in 9,487 (SE 1 in 6,644–13,547) CD133− cells. 1 in 285 (SE 1 in 220–371) CD133+ΔΨmhi cells was capable of long-term repopulation vs. only 1 in 5,323 (SE 1 in 3,410–8,308) CD133−ΔΨmlo cells. The enrichment provided by CD133+ΔΨmhi vs. CD133+ cells was not significantly different (P=0.49). One in 1,488 (SE 1 in 1,079–2,052) integrin α6hiCD71lo cells was capable of long-term repopulation, similar to UNF cells [1 in 1,491 (SE 1 in 1,109–2,002)] (Figure 2g). Thus, the CD133+ population was enriched for long-term repopulating EpiSCs 3.9 fold over the UNF population and the CD133+ΔΨmhi population was enriched for long-term repopulating EpiSCs 5.2 fold (Figures 2g and h), while the CD133− and CD133−ΔΨmlo populations were depleted. The integrin α6hiCD71lo population of keratinocytes was not enriched for long-term repopulating cells in this assay.

To exclude the possibility of contamination of CD133+ keratinocytes by CD133+ dermal papilla cells, we studied versican GFP-tagged mice that express GFP in dermal papilla cells (Ehama et al., 2007; Kishimoto et al., 1999). CD133+ keratinocytes isolated from versican-GFP epidermis were GFP− (Supplementary Figure S1a).

In pilot studies, SSEA1+, CD44+, and CD133+CD44+ populations were not enriched in long-term repopulating EpiSCs (data not shown).

CD133+ and CD133+ΔΨmhi murine keratinocytes express stem cell markers and display functional stem cell characteristics, including self-renewal, multipotency and label retention

Overlap between CD133+ and previously described EpiSC markers was analyzed. Using histology, cells co-expressing CD133 and integrin α6 (Li et al., 1998) were located in the hair follicle bulge (Figure 3a). Cells co-expressing CD133 and Delta 1 (Estrach et al., 2008) were located in the bulge (Figure 3b). Integrin α6+CD34+ keratinocytes exhibited EpiSC properties (Trempus et al., 2003;Trempus et al., 2007). Using flow cytometry, 97.1±1.8% of CD133+ keratinocytes were integrin α6+ (not shown, n=4), 99±1% were CD34+ (Figure 3c, upper panel, n=3), and 78.7±14% were integrin α6+CD34+ (Figure 3c, middle panel, n=3). Only 0.8 ± 0.2% of CD133+ keratinocytes were integrin α6hiCD71lo (Figure 3c, lower panel, n=4). Using direct counting of FACS isolated cells, 94.3±4.2% of CD133+ cells (Figure 3d) and 95.3±1.1% of CD34+ cells (not shown) were keratin 14+ (n=3), consistent with basal keratinocytes. Keratin 15 co-localized with label retaining cells (Lyle et al., 1998; Lyle et al., 1999). Using direct counting 15.5±4% of CD133+ keratinocytes were keratin 15+ (Figure 3e, n=4). CD133+ keratinocytes were Lgr5− (data not shown) (Jaks et al., 2008). (4.5±2.1% of CD34+ cells are CD133+, 9.7±7.8% of integrin α6+ are CD133+ and 12.2±2.2% of integrin α6+CD34+ cells are CD133+). Thus, CD133+ keratinocytes are a subset of the α6+CD34+ bulge population.

Figure 3

CD133+ keratinocytes co-localize with integrin α6+, Delta1+ and CD34+ keratinocytes

(a) CD133 and α6 integrin co-expressing keratinocytes in the bulge (arrows). (b) CD133 and Delta-1 co-expressing keratinocytes in the bulge (arrows). (c) Overlap between CD133+, CD34+ and integrin α6 populations. CD133+ cells are CD34+ (upper panel), integrin α6+CD34+ (middle panel) and not integrin α6hiCD71lo (lower panel). (d) 94.3±4.2% of CD133+ keratinocytes were keratin 14+ (direct counting, 200–500 cells/experiment). (e) 15.5±4% of CD133+ cells were keratin 15+ (direct counting, 200–500 cells/experiment). Scale bars= 10 µm. bg= bulge

Bmi-1 is associated with stem cell self-renewal (Lee et al., 2008; Reinisch et al., 2007; Lacroix et al., 2010). More CD133+, ΔΨmhi and CD133+ΔΨmhi cells expressed nuclear Bmi-1 (19.8±7.6%; P<0.05, 28.7±5.3%; P<0.001 and 45.9±10%; P<0.001 respectively) than UNF, CD133−, ΔΨmlo and CD133−ΔΨmlo cells (5±2.3%, 2.8±2.6%, 6.4±4.1% and 4.8±4.5% respectively) (Figures 4a and b). To further study self-renewal ability, 4,000 CD133+ or CD133+ΔΨmhi and 4,000–25,000 CD133− or CD133−ΔΨmlo keratinocytes were injected into NOD/SCID mice. Primary injection sites were harvested at 9 weeks and cell suspensions obtained, followed by re-injection into secondary recipient mice. Secondary sites (harvested at 9 weeks) were analyzed microscopically. ERUs were detected in both CD133+ and CD133+ΔΨmhi secondary transplants, indicating self-renewal ability, while no ERUs were detected in CD133− or CD133−ΔΨmlo secondary transplants (Figure 4c, n=3). These results indicate greater self-renewal ability in the CD133+ and CD133+ΔΨmhi populations.

Figure 4

CD133+ and CD133+ΔΨmhi murine keratinocytes display stem cell characteristics

(a) Nuclear expression of Bmi-1 in CD133+, ΔΨmhi and CD133+ΔΨmhi keratinocytes. Cells with nuclear expression (Nuc), cytoplasmic expression (C) or no expression (x) were counted (ε100 cells per population). (b) Bar graph of the percentage of cells with nuclear Bmi-1 expression. Error bars = mean±SD (* P≤0.05; **P≤0.001; n=3). (c) ERUs were produced after secondary transplantation of ERUs from 4,000 CD133+ and 4,000 CD133+ΔΨmhi keratinocytes but not after secondary transplantation of ERUs from 4,000 CD133− and 4,000 CD133−ΔΨmlo keratinocytes. (d) Hair follicles derived from CD133+ keratinocytes that were co-injected with day 2 murine dermal papilla cells (H&E). Bar graph: error bars = mean±SD, P=0.001, n=3). (e) CD133 immunostaining of hair follicle derived from CD133+ keratinocytes co-injected with day 2 murine dermal papilla cells and H&E of the adjacent section. (f) 28.9±8.6% of CD133+ and 37.3±5.9% of CD133+ΔΨmhi keratinocytes are label retaining cells (arrows indicate BrdU− cells; mean ± SD, n=3). Scale bars= 10 µm for a, c, f; 50 µm for d (low power images); 10 µm for d (high power images); 20 µm for e. bg=bulge; hf=hair follicle

Regeneration of hair follicles in vivo, using injection of mixtures of epidermal and dermal cells into immunodeficient mice is well-described (Zheng et al., 2005; Morris et al., 2004; Yang and Cotsarelis, 2010). Multipotency was studied using co-injection of 30,000 to 90,000 keratinocytes and 100,000 neonatal (day 2) GFP-tagged dermal papilla cells. Eighteen days after injection, CD133+ keratinocytes formed greater numbers of hair follicles than CD133− keratinocytes (22.3±2.8 vs. 2.7±2.6 hair follicles per 30,000 cells injected, respectively, P=0.01, n=3) (Figure 4d). These follicles also expressed CD133 in the bulge (Figure 4e). Injection of 100,000 dermal papilla cells did not produce hair follicles (n=4).

BrdU incorporation in neonatal mice was used to study label retaining ability as previously described (Blanpain et al., 2004). After a 30 day chase, 28.9±8.6% of CD133+ cells (n=3, 1000 cells counted) and 37.3±5.9% of CD133+ΔΨmhi cells were label retaining cells (n=3, 1000 cells counted) (Figure 4f and Supplementary Figure S2).

CD133+ and CD133+ΔΨmhi murine keratinocytes have less colony forming ability in vitro than CD133− and CD133−ΔΨmlo keratinocytes

It has been assumed that colony forming efficiency (colonies/100 cells plated) reflects EpiSC number. However, greatest in vitro short-term proliferative ability is not associated with greatest long-term repopulating ability in vivo (Strachan et al., 2008). 4,000 cells of each population (CD133+, CD133+ΔΨmhi, CD133−, CD133−ΔΨmlo and UNF) were plated in 35mm dishes. The CD133− and CD133−ΔΨmlo populations showed significantly greater relative clonogenic ability (1.11±0.1 and 0.47±0.06 fold) vs. CD133+ and CD133+ΔΨmhi populations (0.23±0.07 and 0.07±0.12 fold, respectively) (Figure 5a). Given this in vitro result, we examined short-term repopulation in vivo. UNF and CD133− populations also had greater short-term repopulating ability in vivo at one week vs. CD133+ and CD133+ΔΨmhi populations, [1 in 48 (SE 1 in 35–66) and 1 in 77 (SE 1 in 52–144) vs. 1 in 712 (SE 1 in 492–1032) and 1 in 495 (SE 1 in 364–671) respectively] (Figure 5b). Thus, the CD133+ population was enriched for keratinocytes with long-term (in vivo), but not short-term (in vivo or in vitro) repopulating ability. Conversely, the CD133− population showed minimal long-term repopulating ability (in vivo), but contained keratinocytes with short-term repopulation ability (in vivo and in vitro).

Figure 5

In both in vitro and in vivo studies the short-term repopulating cells reside in the CD133− population, rather than the CD133+

(a) Murine keratinocyte populations were sorted and 4,000 cells plated in each well. After 3–4 weeks wells were stained with toluidine blue. Relative colony forming efficiency of CD133+, CD133+ΔΨmhi, CD133− and CD133−ΔΨmlo keratinocytes is shown in the bar graph. Error bars = mean±SD (*P≤ 0.05; n=3) (b) Assessment of ERU frequency using in vivo transplantation and limiting dilution analysis at 1 week (short-term repopulation) in vivo.

DISCUSSION

These studies show that murine CD133+ keratinocytes (a subset of integrin α6+CD34+ keratinocytes) and CD133+ΔΨmhi keratinocytes, contain long-term repopulating, self-renewing, multipotent EpiSCs containing increased proportions of cells with nuclear Bmi-1 expression and label retaining ability. The CD133− population contains the clonogenic cells in vitro and the short-term repopulating cells in vivo. The CD133+ and CD133+ΔΨmhi populations, while containing long-term repopulating cells, are not clonogenic in vitro nor short-term repopulating cells in vivo.

While it was believed 10% of basal cells were EpiSCs, many studies have found a frequency on a lower order (Schneider et al., 2003; Charruyer et al., 2009; Bickenbach, 1981; Bickenbach and Chism, 1998; Triel et al, 2004; Redvers et al, 2006; Winter and Bickenbach, 2009). Here the frequency of EpiSCs was 1 in 1,491 (SE 1 in 1,109–2,002). This is expected to be an underestimate, as it is most probable that not all EpiSCs achieve proliferation in this model. In this study, keratinocytes were isolated from freshly-obtained day 2–4 neonatal murine epidermis and analyzed by FACS for integrin α6 expression. Integrin α6+ (CD49f) expressing basal cells constituted 44 ± 7 % of total keratinocytes (n= 5), in keeping with previous findings (Schneider et al, 2003). Therefore, in this study, on the order of 0.01% of murine basal cells are EpiSCs. While our studies were performed using neonatal murine dorsal epidermis, our result is on the same order as that found using unperturbed adult murine epidermis in vivo (Clayton et al., 2007).

CD133+ΔΨmhi cells were studied for long-term repopulating ability, nuclear Bmi-1 expression, and label retention. Although nuclear Bmi-1 expression was increased in the CD133+ΔΨmhi vs. CD133+ population (p=0.02), the EpiSC enrichment (5.2 fold vs. 3.9 fold, P=0.49) and the number of label retaining cells (37.3±5.9% vs. 28.9±8.6%, P=0.23) in CD133+ΔΨmhi keratinocytes was not significantly different. Also, while there is strong evidence for integrin α6hiCD71lo as a marker of human EpiSCs (Kaur and Li, 2000; Li et al., 1998; Li et al., 2004; Li, 2005) and in vitro studies showed that integrin α6hiCD71lo murine keratinocytes are quiescent and small, with high nuclear/cytoplasmic ratio (Tani et al., 2000; Yano et al., 2005), our studies indicated that murine integrin α6hiCD71lo keratinocytes are not enriched for long-term repopulating cells in vivo, in this assay. Other functional characteristics of integrin α6hiCD71lo cells were not tested.

CD133+ cells were bulge cells and were CD34+ [AC133 antibody was produced by inoculation of CD34+ human cells (Yin et al., 1997; Bidlingmaier et al., 2008)]. While 94.3±4.2% of CD133+ keratinocytes were keratin 14+, there may exist a more primitive K14− stem cell. While side population cells were integrin α6hiCD71lo (Redvers et al, 2006), CD34− and distinct from the bulge population (Redvers et al, 2006), 99±1% of CD133+ keratinocytes were distinct from integrin α6hiCD71lo cells, CD34+ and localized in the hair follicle bulge. CD133+ cells thus appear to be distinct from both integrin α6hiCD71lo cells and side population cells.

CD133+ cells are keratinocytes because CD133+ (CD34+) cells are keratin 14+ and integrin α6+, consistent with basal epidermal cells. Possible contamination of CD133+ keratinocytes by CD133+ dermal papilla cells was excluded using versican GFP-tagged mice (CD133+ dermal papilla cells are GFP+). Isolated CD133+ keratinocytes were GFP−. Furthermore, the GFP+/CD133+ dermal papilla cells were less than 2% of pure dermal papilla cell preparations, so that contaminating CD133+ cells could not explain the 4.2±3.1% CD133+ cells consistently found in the total keratinocyte preparation. Finally, isolated CD133+ cells grow a stratified epithelium in vivo, not dermal tissue, when injected into murine subcutis.

The colony forming efficiency results reflected short-term in vivo results. CD133− keratinocytes were clonogenic in vitro and were short-term repopulating cells in vivo. CD133+ cells didn’t grow in vitro and were not short-term repopulating cells in vivo, but were long-term repopulating cells in vivo. Culture conditions may favor proliferation of TACs and/or reprogram proliferative ability [for review see (Cotsarelis, 2006)] or in vitro analysis may reflect a wounding response, not homeostasis (Kaur, 2006). These results are consistent with the belief that stem cells have minimal clonogenic ability in vitro and that not all colony forming cells constitute stem cells (Pavlovitch et al., 1991; Budak et al., 2005; Li et al., 1998; Selver et al., 2011; Strachan et al., 2008; Schofield, 1978; Loutit et al., 1981; Louis et al., 2008).

Lineage tracing indicates that follicular cells do not contribute to interfollicular epidermis during homeostasis (Gazizadeh et al, 2001; Morris et al, 2004; Tumbar et al, 2004; Levy et al, 2005; Ito et al., 2005). Keratin 15+ bulge cells formed interfollicular epidermis in transplantation assays (at 4 weeks) (Morris et al, 2004), but lineage analysis after wounding demonstrated that keratin 15+ bulge cells and their progeny contribute to the interfollicular epidermis only transiently, while unlabeled cells persisted (Ito et al, 2005). In another study, lineage tracing after wounding demonstrated that Shh-expressing cells (present throughout the pilosebaceous unit and comprising both keratin 15+ and keratin 15− cells) and their progeny contributed to interfollicular epidermis for at least 4 months (Levy et al, 2007). By contrasting and synthesizing the results of these two sets of experiments Levy et al suggested the possibility of a distinct cell in the follicle (derived from Shh+ but not keratin 15+ cells) with the ability to become a long-term repopulating stem cell of interfollicular epidermis (Levy et al, 2007). Our studies show that bulge-derived CD133+ cells (85% keratin 15−) form both interfollicular epidermis and hair follicles in transplantation assays and form interfollicular epidermis for the long-term (9 weeks). Future lineage tracing is indicated to determine how CD133+ cells and their progeny contribute to follicular and interfollicular epidermis in vivo, during homeostasis as well as wounding.

Finally, our results characterize CD133+ EpiSCs using functional properties and provide a basis for future studies aimed at quantitative comparison of the enrichment in long-term repopulating stem cells and short-term repopulating progenitors provided by different EpiSC isolation strategies.

MATERIALS AND METHODS

Mice

IACUC approval (VAMC San Francisco, CA) was obtained and work performed in adherence to institutional guidelines. C57BL/6TgN(ACTbEGFP1Osb), NOD/SCID (Jackson Laboratory, http://www.jax.org) and Versican GFP-tagged transgenic mice (kindly provided by Jiro Kishimoto as a gift to Daniel Bikle) (Ehama et al., 2007; Kishimoto et al., 1999) were used.

Keratinocyte and Fibroblast isolation

Excised skin was incubated in dispase and then trypsin (Schneider et al., 2003) and a cell suspension obtained. Fibroblasts were isolated by incubating the dermis in 0.25% collagenase IA (Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com) at 37°C for 1h.

Flow Cytometry

Keratinocytes were sorted using a FACSAria (BD Biosciences) and analyzed with CellQuest™ software. Antibodies included APC-CD133 [AC133 (Corti et al., 2007; Rountree et al., 2007; Snippert et al., 2009)], APC-IgG1K isotype control, Alexa Fluor® 488-SSEA-1, PE-CD44 (all from eBioscience, San Diego, CA, http://www.ebioscience.com) and FITC-integrin α6, FITC-IgG2a isotype control, PE-CD71, PE-IgG1K isotype control and FITC-CD34 (all from BD Pharmingen, San Diego, CA). PE-integrin α6 antibody was from Abcam (Cambridge, MA, http://www.abcam.com).

To sort for keratinocytes with high ΔΨm, cells were treated with TMRM 25nM for 15min (Schieke et al., 2008; Schieke et al., 2006). Cells with high ΔΨm accumulate more of this potentiometric dye. (Invitrogen).

Epidermal regeneration in vivo

Keratinocytes in Progenitor Cell Technology Epidermal Keratinocyte Medium Complete (CNT07, Chemicon, Temecula, CA, http://www.chemicon.com) were injected with Matrigel™ (0.5 mg/ml; BD Biosciences) 1:1 (vol:vol) into the subcutis of NOD/SCID mice. CNT-07 is a protein-free fully-defined, calcium 0.07 mM formulation with no antibiotics/antimycotics, containing amino acids, minerals, vitamins and organic compounds. Grafts were harvested and examined for the presence of ERUs histologically.

Limiting Dilution Analysis of ERU Frequency

For each time point injection sites were scored as positive if at least one ERU was observed microscopically (Schneider et al., 2003) and the ratio of positive/total sites determined for each dose. Statistical software for limiting dilution analysis (L-CALC, Stemsoft, Vancouver, BC, Canada, http://www.stemsoft.com) was used. The Chi square statistic was used and 5% or less type I error considered significant.

Immunohistochemistry

Involucrin and keratin 14 primary antibodies (Abcam) and VECTASTAIN Elite ABC reagent followed by ImmPACT NovaRED Peroxidase Substrate (Vector, Burlingame, CA) were used on paraffin-embedded sections.

Immunofluorescence

Antibodies included CD133 (AC133) (eBioscience), laminin, Delta1, Lgr5, Keratin15, Keratin14 and Bmi-1 (all from Abcam) and integrin α6 (Santa Cruz Biotechnology, Inc, Santa Cruz, CA, http://www.scbt.com/). AlexaFluor 488-IgF1 (H+L), AlexaFluor 594-IgG and AlexaFluor 488-IgG (Invitrogen) secondaries were used on cell cytospins and paraffin-embedded sections. 300–500 cells were analyzed per sample.

Label retaining cells

Three-day mice were injected twice daily for 3 days with BrdU (75 µg/dose). Skin was collected 30 days later. FITC conjugated BrdU antibody (Abcam) was used to localize label retaining cells using microscopy.

Multipotency

CD133+ or CD133− keratinocytes and Versican GFP-tagged dermal papilla cells (day 2) mixed with Matrigel™ 1:1 (vol:vol) were injected into NOD/SCID subcutis. Grafts were harvested 18 days after injection and examined histologically.

Colony Forming Efficiency

Keratinocytes were seeded at clonal density (100–500 cells per cm2) (Strachan et al., 2008; Morris et al., 1988) onto six-well plates with CNT07 medium. After 3–4 weeks cells were fixed, stained with toluidine blue (Sigma-Aldrich), and colony forming efficiency (number of colonies per 100 cells seeded) was expressed relative to UNF.

Statistical analysis

For comparisons of colony forming efficiency, percentage of cells exhibiting nuclear Bmi-1 and percentage of ERUs labeled with both DiI and DiO, a two-tailed Student’s t test was used. For comparisons of the frequency of repopulating units in each subpopulation at each time point, a two-tailed t test within the L-CALC program (Stemsoft Software Inc) was used.