Human Side Population Keratinocytes Exhibit Long-Term Proliferative Potential and a Specific Gene Expression Profile and Can Form a Pluristratified Epidermis



The aim of the present study was to characterize human side population (SP) epidermal keratinocytes isolated from primary cell cultures. For that purpose, keratinocytes were isolated from normal adult breast skin samples and the Hoechst 33342 exclusion assay described for hematopoietic cells was adapted to keratinocytes. Three types of keratinocytes were studied: the SP, the main population (MP), and the unsorted initial population. SP keratinocytes represented 0.16% of the total population. In short-term cultures, they exhibited an increased colony-forming efficiency and produced more actively growing colonies than did unsorted and MP keratinocytes. In long-term cultures, SP cells exhibited an extensive expansion potential, performing a mean of 44 population doublings for up to 12 successive passages after cell sorting. Moreover, even in long-term cultures, SP keratinocytes were able to form a pluristratified epidermis when seeded on a dermal substrate. Unsorted and MP keratinocytes promoted a reduced expansion: mean values of 14 population doublings for five passages and 12 population doublings for four successive passages, respectively. To further characterize SP cells, cDNA microarrays were used to identify their molecular signature. Transcriptome profiling showed that 41 genes were differentially expressed in SP (vs. MP) cells, with 37 upregulated genes and only four downregulated genes in SP cells. The majority of these genes were functionally related to the regulation of transcription and cell signaling. In conclusion, SP human keratinocytes isolated from primary cultures exhibited both short- and long-term high proliferative potential, formed a pluristratified epidermis, and were characterized by a specific gene expression profile.


The capacity of enriching keratinocyte populations in progenitors is a critical issue in the field of human skin research for both fundamental and applied studies. Adult stem cells can be directly derived from normal human tissues, particularly tissues that are frequently discarded after surgery, like skin. However, the cell number obtained from such tissues is generally limited, and contamination with other cell types, including hematopoietic stem cells, may occur. It is thus important to venture into other strategies. One of these may be stem cell expansion under tissue culture conditions. We have thus investigated the possibility of isolating a cell population enriched in progenitors from primary cultures of human epidermal keratinocytes.

In an effort to purify stem cells from human adult tissues, different techniques are currently used. The Hoechst 33342 exclusion assay, which was first described by Goodell et al. [1] for hematopoietic cells, has been extended to cells isolated from a variety of tissues. Recently, side population (SP) cells have also been isolated from human epithelial tissues, including the mammary gland [2, 3], ocular epithelium [4, 5], and skin [6]. These epithelial cells with SP phenotype are described as primitive undifferentiated cell populations enriched in progenitors. Another widely used technique to isolate epidermal stem cells is labeling with antibodies specific to cell surface markers, including integrins. However, this technique is unsuitable for studying cultured cells because the expression of most cell adhesion proteins is altered in culture. For example, irrespective of the original cell surface phenotype, upregulation of both the α-6 integrin and transferrin receptor occurs soon after placing keratinocytes in culture [7]. Consequently, we chose the Hoechst labeling method to sort epidermal cell populations. The aim of the present study was to isolate SP cells from human primary cultured keratinocytes, to characterize their growth potential and their genomic profile.

Materials and Methods

Isolation of Primary Keratinocytes and Cell Culture

Samples of normal adult skin were obtained from reductive breast plastic surgery of healthy subjects and were provided by AFM-BTR (Banque de Tissus pour la Recherche de l'AFM, Hôpital La Pitié-Salpétrière, Paris, Skin samples from four different patients were used in the present study. Keratinocytes were isolated by incubating in trypsin 0.5% (Gibco, Grand Island, NY, overnight at 4°C and then scraping with a scalpel. Cells were then firmly resuspended with a 5-ml Gilson pipette and filtered on a 70-micron filter (BD Falcon; BD Biosciences, to remove the remaining aggregates before counting and seeding.

Primary keratinocytes were cultured using the Rheinwald and Green method [8] on a lethally irradiated feeder layer of human fibroblasts. To reduce the calcium concentration in the keratinocyte growth medium, we used calcium-free Dulbecco's modified Eagle's medium (DMEM; Cambrex, Walkersville, MD, that permitted us to obtain a final calcium concentration of 0.07 mM. The complete growth medium had the following constituents: DMEM at 1 g/l glucose (Cambrex) and Ham's F-12 media (3:1 mixture), 10% fetal calf serum (HyClone, Logan, UT,, 5 μg/ml insulin, 0.18 mM adenine, 0.4 μg/ml hydrocortisone, 1 nM cholera toxin, 2 nM triiodothyronine, 10 ng/ml epidermal growth factor, 4 mM glutamine (Sigma-Aldrich, St. Louis,, and 50 IU/ml penicillin/streptomycin (Gibco).

Hoechst 33342 Exclusion Assay

The Hoechst dye exclusion assay was performed according to the protocol of Goodell et al. [1] with the modification that staining was performed while the cells were still attached to the culture dish and in the normal growth medium. A similar labeling method has been described for pancreatic cells [9]. All experiments were performed on second-passage cultures at 50% confluence. This stage was obtained 21 days after the initiation of the culture and eight population doublings on average. The Hoechst dye was optimized by titration between 10 and 100 μg/106 keratinocytes according to the fetal calf serum used. For all experiments, Hoechst 33342 (Sigma-Aldrich) was added to the growth medium at a final concentration of 40 μg/106 keratinocytes (ranging from 1.6 to 4.8 μg/ml as a function of cell density), and the cells were incubated for 2 hours in normal culture conditions (37°C and 5% CO2). For each experiment, verapamil (Sigma-Aldrich) was added in one flask at a final concentration of 400 μg/106 cells (ranging from 16 to 48 μg/ml as a function of cell density). For analysis, cells were first treated for 5 minutes at 20°C with trypsin to eliminate feeder layer cells. Then keratinocytes were trypsinized for 10 minutes at 37°C and resuspended in PBS containing propidium iodide (Sigma-Aldrich) at a final concentration of 2 μg/ml to allow the discrimination of dead cells. Flow cytometry analysis and sorting were performed on a dual-laser MoFlo cell sorter (DakoCytomation, Inc., Fort Collins, CO, Hoechst fluorescences were excited with a UV laser tuned to 350 to 360 nm (150 mW) and collected through a 450/65-nm band-pass filter for the blue Hoechst fluorescence and 670/30-nm band-pass filter for the red Hoechst fluorescence. Propidium iodide fluorescence was excited with a second laser tuned to 488 nm (150 mW) and collected through a 630/30-nm band-pass filter. Two cell populations were sorted: SP keratinocytes that excluded the Hoechst and main population (MP) keratinocytes that showed high levels of Hoechst red and blue fluorescence. Ten different cell sortings were performed using the four different cell preparations. Seven sortings were used for proliferation studies and three for RNA preparations.

Immunofluorescence Studies

SP and MP keratinocytes were directly sorted onto glass slides, air-dried, and fixed with cold methanol. Cells were stained by indirect immunofluorescence for keratin 5 and involucrin. Briefly, cells were incubated overnight at 4°C with mouse monoclonal anti-cytokeratin 5 or anti-involucrin (Novocastra, Newcastle upon Tyne, U.K., followed by 1 hour at room temperature with Cy3-conjugated anti-mouse (Jackson ImmunoResearch Laboratories, West Grove, PA, Work dilution for all antibodies used was 1:100. At least 200 cells were counted for each antibody.

Clonogenic Assay: Colony-Forming Efficiency and Colony Characterization

For the analysis of cell adhesion, cell morphology, and colony-forming efficiency (CFE), SP and MP cells were plated at 10 cells per cm2 and grown for 14 days. Keratinocytes were then fixed with cold methanol and stained with 1% rhodamine in order to count the number of colonies by microscopy [10]. All colonies were counted regardless of their size. Analysis on at least 100 colonies was performed at each passage until senescence for each skin sample studied (n = 7 cell sortings).

Colonies were analyzed according to their size. Three classes were defined: large colonies greater than 5 mm in diameter showing a high growth potential, intermediate colonies between 1 and 5 mm in diameter containing cells in a transitional stage between growing, and abortive colonies. Finally, colonies smaller than 1 mm in diameter, in which all of the cells lacked proliferative capacity, generated only abortive colonies corresponding to approximately 10 cells per colony.

Expansion Assays

Sorted SP and MP keratinocytes were cultured as previously described. After reaching 70 to 80% confluence, SP and MP keratinocytes were replated at 1000 per cm2. Cultures were serially passaged until the growth capacity of the cells was exhausted. The number of population doublings was calculated using the following formula: g = (log N/N 0)/log 2, where N 0 represents the number of plated cells and N the total number of cells obtained at each passage.

Reconstructed Epidermis

De-epidermized human dermis (DED) was prepared according to the technique described by Régnier et al. [11]. Briefly, split-thickness human skin obtained at plastic surgery was floated in magnesium/calcium-free phosphate-buffered saline at 37°C for 10 days. Thereafter, the epidermis was separated from the dermis. Dermal cells were killed by serial freezing and thawing, and the cell-free dermis was stored at −20°C until use. SP keratinocytes at passage 13 were seeded onto the DED and cultured for 6 days in the DMEM/Ham's F-12 medium (Invitrogen, Carlsbad, CA, containing 10% fetal calf serum (Invitrogen), 10 ng/ml epidermal growth factor (EGF; BD Biosciences, San Diego,, 0.4 μg/ml hydrocortisone (Sigma-Aldrich), 10−6 M isoproterenol (Sigma-Aldrich), 5 μg/ml transferrin (Sigma-Al-drich), 2 × 10−9 M triiodothyronine (Sigma-Aldrich), 1.8 × 10−4 M adenine (Sigma-Aldrich), and 5 μg/ml insulin (Sigma-Aldrich). Thereafter, the cultures were raised to the air-liquid interface and were continued in the absence of isoproterenol, transferrin, triiothyronine, and adenine. Histologic examination of the reconstructed epidermis was performed after 7 days of culture at the air-liquid interface.

RNA Isolation and Amplification

RNA was extracted from three independent cell sortings performed on three different skin samples. Total RNA was isolated directly after cell sorting using the RNA Microprep kit (Stratagene, La Jolla, CA, following the manufacturer's protocol with the optional DNase treatment step. Extracted RNA was amplified using the MessageAmp RNA Kit (Ambion, Austin, TX, following the instructions provided by the manufacturer. One amplification round was performed. Before hybridization, the three RNA samples of each cell type were pooled.

Microarray Experiments

Microarrays were spotted by our platform with a total of 10,752 features (probes and controls). Three types of DNA collections were used for preparing the cDNA arrays: a collection of 5,760 cDNA clones from the NIB (normalized infant brain) library (kindly provided by Genethon [Evry cedex, France,] [12]), a collection of 1,536 cDNA clones from two human leukocyte libraries (ResGen Invitrogen Corporation, Paris,; Sanofi Aventis, Paris, [13]), and a collection of 2,304 probes corresponding to specific human genes (involved in apoptosis, differentiation, adhesion, and DNA repair) selected by direct interrogation of the Unigene database. Moreover, a set of 1,152 control reporters and pseudofeatures was spotted. These microarrays have already been described in previous reports [14].

Amplified RNA (1 μg) was labeled by incorporation of aminoallyl-modified dUTP (Sigma-Aldrich) followed by a subsequent incorporation of monofunctionnal Cy3 or Cy5-N-suc-cinimide esters (Amersham Biosciences, Piscataway, NJ, A complete description of the microarrays used in this study, including the protocols of production and postprocessing of slides, has been deposited into the Gene Expression Omnibus (GEO) database ( This information is available under GEO accession no. GSE2467.

Microarray Data Analysis

Four independent dye-swap hybridizations (eight microarrays) were performed. Fluorescence intensities of Cy3 and Cy5 were measured separately at 532 and 635 nm with a laser scanner (Axon Genepix 4000A; Axon Instruments, Union City, CA, The resulting data files were imported into an image analysis program (Genepix 4.0; Axon Instruments). Feature ratios were calculated using the GenePix 4.0 software, which flags spots as absent based on spot characteristics. Additionally, bad spots were manually flagged and excluded from subsequent analyses. The fluorescence values calculated by the GenePix program were saved and imported into Genespring 6.1 (Silicon Genetics, Redwood City, CA, for further analysis. We applied two normalization steps: dye-swap data transformation and intensity-dependent LOWESS (locally weighted scatterplot smoothing) in order to eliminate dye-related bias in our two-color experiments.

Selected probes had to meet the following criteria: measurements should be flagged as present in all hybridizations, mean differential expression should be above the cut-off value of 1.5 (or under 0.67), and probes defined as differentially expressed should have a Student's t test p value lower than .001. Genes were classified according to their biological functions on the basis of the Gene Ontology Consortium.

Quantitative Reverse Transcription-Polymerase Chain Reaction Analysis

Quantitative reverse transcription-polymerase chain reaction (RT-PCR) was carried out using 20 ng of amplified RNA with the SYBR Green Jumpstart Taq ReadyMix (Sigma-Aldrich) according to the protocol supplied by the manufacturer using an ABI Prism 7700 sequence detection system (Applied BioSystems, Foster City, CA, All reactions were carried out with an annealing temperature of 60°C. The following gene-specific primers were used: abcg2 sense: cagcagtgtttcagccgtgg, abcg2 antisense: ggcatctgcctttggcttca; id1 sense: accctcaacggcgagatcag, id1 antisense: gggagacccacagagcacgt; id2 sense: gcccagcatcccccagaa, id2 antisense: ggtggtcagcggcgtcct; integrin β-1 sense: caagcagggccaaattgtgg, integrin β-1-antisense: tgtcatctggagggcaaccc; dual-specificity phosphatase 6 (dusp6) sense: gtgtgtggccccaggtgtca, dusp6 anti-sense: cctcgaatcccagtccctgc; gapdh sense: TTCACCACCATG-GAGAAGGC; gapdh antisense: GGCATGGACTCTGGT-CATGA. GAPDH was used as an internal control. Fold differences were calculated by the mathematical model described by Pfaffl [15].


SP Cells Can Be Isolated from Primary Cultures of Human Keratinocytes

Human keratinocytes in primary cultures contained a distinct population of Hoechst-effluxing cells, similarly to hematopoietic cells. Cells that excluded the Hoechst were defined as SP keratinocytes, whereas MP keratinocytes showed high levels of Hoechst red and blue fluorescence (Fig. 1A). SP keratinocytes represented 0.16 ± 0.06% (n = 10 independent cell sortings) of the total cell population. Treatment with verapamil, a multidrug transporter inhibitor known to eliminate the SP population in bone marrow, also reduced the keratinocyte SP population (Fig. 1B). The use of cultured keratinocytes allowed us to isolate more SP cells than we could have directly from the tissue. On average, 4,000 SP cells could be obtained from 2.5 × 106 keratinocytes isolated from a skin sample, whereas 1 million SP cells could be isolated from the same initial number of keratinocytes after eight population doublings. Thus, tissue culture allowed a 250-fold expansion of SP cells (n = 10). We also compared the number of SP cells at passage 2 with α-6briCD71dim keratinocyte progenitors directly isolated from the primary tissue, which are a well studied stem cell candidate population. If we consider a proportion of 0.8% of these progenitors within the total initial population [16], 20,000 progenitors can be isolated from 2.5 × 106 keratinocytes. Thus, SP isolation from cultured cells at p2 allows a 50-fold expansion as compared with α-6bri CD71dim cells isolation from primary tissue.

SP Keratinocytes Are Undifferentiated Keratinocytes

In three independent experiments, SP keratinocytes were directly sorted onto glass slides and stained by indirect immunofluorescence using monoclonal antibodies directed against cytokeratin 5 (K5), which is a marker predominantly expressed in undifferentiated keratinocytes of the basal layer. K5 was expressed by both cell populations (94 ± 2% for SP and 81 ± 18% for MP keratinocytes). Moreover, cells were labeled with a monoclonal antibody directed against involucrin, which is a marker of differentiated keratinocytes of the suprabasal layers of the epidermis. Only 2 ± 2% of SP cells expressed involucrin, whereas 20 ± 5% of MP epidermal cells expressed this differentiation marker. These results show that SP keratinocytes were mainly undifferentiated cells, whereas MP keratinocytes were a heterogeneous population composed of both differentiated and undifferentiated cells.

In Vitro Expansion of the SP, MP, and Total Keratinocyte Populations

The potential of global cell expansion was characterized for the three cell types. For that purpose, keratinocytes were isolated from three different skin samples and continually passaged until their growth capacity was exhausted. For one sample, two independent cultures and cell sortings were performed. Unsorted keratinocytes promoted an expansion up to mean passage 5, with a cell output of 1.6 × 104, corresponding to 14 ± 2.6 population doublings. MP keratinocytes also promoted a low expansion, as the mean output from one plated MP cell was 4 × 103, corresponding to 12 ± 2.7 population doublings over four successive passages. By contrast, SP keratinocytes could be cultured up to an average of 12 passages after cell sorting, with a maximum of 17 passages. The mean output from one plated SP cell was 2 × 1013, corresponding to 44 ± 2.1 population doublings over a mean of 12 successive passages. One representative growth curve for the three cell types isolated from one skin sample is shown in Figure 2.

In addition, when MP cells were submitted to Hoechst 33342 labeling, they did not exhibit a typical exclusion profile. The SP cells, on the other hand, exhibited a typical exclusion profile over 12 passages after cell sorting (data not shown).

SP Keratinocytes Exhibit High Growth Potential in Short-Term Cultures

We then evaluated the CFE and colony types for each population of keratinocytes at passages 1 to 4. The mean CFE values were evaluated by counting all colonies, regardless of their size (n = 7 cell sorting experiments). We calculated the following CFE values: 10% for unsorted population, 14.5% for SP cells, and 7% for MP keratinocytes (Fig. 3A). Together, these results show that SP keratinocytes had a greater capacity to attach to the culture dish than did non-SP cells in short-term cultures.

To evaluate the growth potential of keratinocytes generated by each population, the colonies were separated in three classes according to their cell number. Unsorted keratinocytes exhibited the three types of colonies (Fig. 4A). The MP keratinocyte population was composed essentially of large keratinocytes that formed abortive colonies (Fig. 4A). SP keratinocytes produced rapidly growing colonies that were at least twice the size of those from unsorted and MP cells (Fig. 4A). These large colonies were composed of small and tight keratinocytes showing a high growth potential. These results show that SP keratinocytes had a greater growth potential than did non-SP cells in short-term cultures.

SP Keratinocytes Exhibit High Growth Potential in Medium and Long-Term Cultures

Comparison of the medium- and long-term growth potential of SP and unsorted populations confirmed that SP keratinocytes were more proliferative than unsorted cells. CFE was studied during these medium-and long-term cultures. After passage 4, unsorted keratinocytes exhibited a low CFE (0.8%) (Fig. 3B) and became unable to produce proliferative colonies (Fig. 4B). Interestingly, although SP keratinocytes had a low CFE of 1% in long-term cultures (Fig. 3C), they were still able to form a mean of 71% of growing colonies that were larger than 5 mm (Fig. 4C).

SP Keratinocytes Generate Pluristratified Epidermis

Another feature of SP keratinocytes that we investigated was their organogenic potential. Cultures initiated with SP keratin-ocytes were passaged successively up to 13 times and then seeded onto a dermal substrate. At this late stage of long-term cultures, cells were still able to form a pluristratified epidermis (Fig. 5C). The following pattern of characteristic epidermal three-dimensional differentiation was observed: a basal layer containing polygonal cells oriented perpendicularly to the underlying dermis; three to four layers of spinous cells; four to six layers of granular cells characterized by the presence of keratohyalin granules; and anucleated, flattened cornified cells forming a compact stratum corneum. In addition, we studied the organogenic potential of MP keratinocytes. As expected, these cells at a low passage number (p3) were able to form a pluristratified epidermis in the organotypic model (Fig. 5A). However, they rapidly lost this potential throughout successive passages (p6; Fig. 5B). MP cells could be thus clearly distinguished from the SP cells on the basis of this functional test.

Defining the Molecular Signature of SP Keratinocytes

To identify what signals maintain the SP phenotype, a microarray study comparing large-scale gene expression patterns between SP and MP keratinocytes was performed. Forty-one genes were found differentially expressed in SP (vs. MP) cells (Table 1). For five genes, two different probes were spotted on the arrays. Reproducibility of the array results was illustrated by the observation that similar ratios were obtained for these different probes. Part of these array results was confirmed by quantitative RT-PCR (Fig. 6).

Of the 41 regulated genes in SP keratinocytes, 37 genes were found to be induced, whereas only four genes were repressed. A high proportion of these upregulated genes were functionally related to the regulation of RNA transcription (29%) and cell signaling (11%). Interestingly, two members of the Id gene family of transcription regulators, which are involved in epithelial cell proliferation and differentiation, were modified in an opposite way: the Id2 gene expression was repressed (−1.79; Table 1) in SP keratinocytes, whereas that of Id1 was overexpressed (−1.52; Table 1). These results were confirmed by real-time quantitative RT-PCR (Fig. 6).

Because the dye efflux exhibited by bone marrow SP has been attributed to members of the ABC transporter family, we determined whether they were also expressed by SP keratinocytes. Using quantitative RT-PCR, the ABCG2 mRNA level was found expressed in these cells but not upregulated.

Finally, we found by microarray analysis that cell surface markers of epidermal stem cells (integrin α-6 and β-1, CD71, EGF receptor, and connexin 43) were not differentially expressed in SP keratinocytes. Part of these results was confirmed by real-time quantitative RT-PCR (Fig. 6).


This work is the first report on the isolation and long-term characterization of SP cells isolated from human keratinocytes in primary cultures. We show that SP keratinocytes had a dye-effluxing phenotype similar to that of bone marrow SP cells. They represented an average of 0.16% of the initial population and were undifferentiated epidermal cells. SP-derived cells exhibited both short- and long-term high proliferative potential during serial subcultivation, with a mean output of 2 × 1013 cells over 12 passages (24 weeks). Moreover, they could re-form a pluristratified epidermis in an organotypic model even at late passages.

Isolation of SP cells from human epidermis using the dye exclusion assay has already been reported, but cells were isolated directly from newborn foreskin epidermis [6]. In this study, 0.3% of SP keratinocytes was isolated from the initial population, and SP cells were enriched for quiescent cells at the G0/G1 phase of the cell cycle. However, these SP keratinocytes have not been functionally characterized.

Other studies have reported on the isolation and functional characterization of human epidermal progenitors, which can thus be compared with the present data. These studies focused on progenitor isolation directly from primary tissue using cell surface protein labeling. Keratinocytes with a low expression of EGF receptor exhibited an in vitro growth potential similar to that of the present SP keratinocytes [17]. EGF-Rlow keratinocytes showed a mean output of 3.2 × 1012, corresponding to 42 population doublings for 13 passages, and they could form a pluristratified epidermis at passage 7. An increased growth potential has also been described for human epidermal progenitors isolated from epidermis by flow cytometry on a low desmoglein 3 expression [18]. Although the long-term growth potential was not precisely reported, the authors described that the Dsg3-dull population was capable of sustained generation for 24 weeks, which is similar to the SP keratinocytes of the present study (24 weeks) and to the EGF-Rlow keratinocytes (22 weeks). In addition, P. Kaur and collaborators isolated keratin-ocytes with a phenotype of α-6bri CD71dim, which is a minor population corresponding to 8% of basal layer keratinocytes [16]. A single progenitor from this population can generate 6 × 108 cells in 14 weeks [19]. They also showed that in vivo these progenitors are able to form a reconstructed epidermis [20]. The comparison of the phenotype of this α-6bri CD71dim keratinocyte population with that of SP population allowed these authors to conclude that SP keratinocytes can be considered as a subset of the α-6bri CD71dim population. In conclusion, the Hoechst labeling method that we have developed on cultured keratinocytes allowed isolation of SP cells that exhibited a long-term growth potential similar to keratinocyte progenitors isolated on the basis of EGF-Rlow, Dsg3-dull, or α-6bri CD71dim labeling.

Concerning the mechanism of the Hoechst efflux, ABC transporters have been proposed as major actors for bone marrow SP cell phenotype [21]. We found that ABCG2 transporter was not overexpressed in SP keratinocytes isolated from the cultured cells; a similar finding was described for SP keratinocytes isolated from the tissue by Terunuma et al. [6]. The dye efflux may thus be unrelated to ABCG2 overexpression in SP keratinocytes. In general, the role of these transporters is still unclear, even in blood cells. For example, the ABCG2 transporter was recently found to be neither required nor responsible for the SP phenotype in many human blood cells [22]. Similar to ABCG2, well known cell surface markers of progenitor keratinocytes (i.e., β-1 and α-6 integrins) were not differentially expressed in SP keratinocytes. It is well known that in vitro keratinocyte expansion modifies the expression of cell surface proteins [7], as shown for nonadherent cells [23]. It is interesting to note that SP keratinocytes exhibited functional properties of progenitors, whereas they did not overexpress known cell surface markers for epidermal stem cells.

Last, we studied the molecular phenotype of SP keratinocytes via expression profiling. We found that 41 genes were differentially expressed in SP (vs. MP) cells. Thirty-seven genes were upregulated, whereas only four genes were downregulated, and part of these array results has been confirmed by quantitative PCR. Moreover, a high proportion of these genes were related to two cellular functions: cell signaling (11% of differentially regulated genes) and the activation or repression of transcription (29%). Concerning cell signaling, DUSP6 was the most induced gene in SP cells; three members of this gene family were differentially expressed in SP (vs. MP) keratinocytes. DUSPs are classified as protein tyrosine phosphatases. It has already been shown that DUSP1 is expressed in human hematopoietic stem cells and downregulated with differentiation [24]; this is similar to what we observed in human epidermal SP cells. Moreover, another member of this phosphatase family, the PTEN phosphatase (phosphatase and tensin homolog deleted on chromosome 10), has been found to regulate stem cell self-renewal and proliferation [25].

Concerning the regulation of transcription, we found that Id2 was the most downregulated gene in SP cells, whereas Id1 was upregulated. It is noteworthy that a similar pattern was already described in the mammary gland [23]. The Id gene family was recently found differentially expressed in the bulge keratinocyte stem cells of mouse hair follicle [26], and Id1 has been shown to be involved in the inhibition of the epidermal differentiation [27, 28]. Because we recently described the role that Id2 plays in keratinocyte proliferation and stress response [29], we will further assess whether Id2 is involved in the phenotype of human epidermal progenitors.

Members of translational regulators like PUM1 are also differentially expressed in SP keratinocytes. It has been found that the Pumilio family is essential for the maintenance of the stem cell pool in the germinal cell line of Drosophila and Caenorhabditis elegans [28, 29]. More recently, it has been shown that Pum1 and Pum2 are more abundantly transcribed in murine hematopoietic stem cells [30]. The role of PUM1 in keratinocyte stem cells remains to be further investigated.

Finally, we compared our data with those obtained with microarrays in stem cells from various mouse and human tissues (bone marrow, neural, embryonic, and retinal tissues) [3133]. It should be noted that the microarrays used had different numbers of spotted genes, ranging from 9,120 (present study) to 36,000 probes [31]. On a single-gene basis, we found that 30% of the upregulated genes found in epidermal SP cells were also upregulated in at least one of these stem cells.

In summary, we found that SP cells can be isolated from primary cultures of adult human epidermal keratinocytes. These cells exhibited both short- and long-term high proliferative potential, could form a pluristratified epidermis, and were characterized by a specific molecular profile. We conclude that the Hoechst exclusion assay on keratinocytes formerly expanded for two passages in primary cultures allows isolation of cell populations enriched for human epidermis progenitors. One main advantage of the cell culture strategy is that it allows more SP cells to be obtained than directly from the tissue; an average increase factor of 250 is obtained in our conditions when comparing the primary tissue with cell cultures at passage 2. If we compare the number of SP cells at passage 2 with α-6briCD71dim keratinocyte progenitors directly isolated from the primary tissue, which are a well studied stem cell candidate population, the expansion factor can be evaluated to be 50 (see Results). Another advantage is that, due to the culture step before sorting, the possibility of contamination with other types of stem cells, such as hematopoietic cells, is greatly reduced. However, further in vivo functional studies will be required to determine the full potential of these human epidermal SP keratinocytes. The common use of cultured epithelial sheets after skin burning and the long-term reconstitution of the skin demonstrated that keratinocyte progenitors remain present in cell cultures. They allow long-term regeneration of human skin in burn patients transplanted with autologous cultured epithelium [34]. However, the reconstruction does not provide a fully functional skin; notably, skin appendages such as hair follicles, sebaceous, and sweat glands are not regenerated after transplantation [35]. The recent identification of multipotent epithelial stem cells in adult skin raises new hopes for improving morphogenesis after grafting. The use of enriched populations for transplantation may accelerate wound healing and result in a better functionality of the skin after grafting.

Table Table 1.. Microarray molecular signature of SP keratinocytes
original image
Figure Figure 1..

SP profile of human keratinocytes in primary cultures after Hoechst 33342 staining. (A): The SP keratinocytes represented 0.16 ± 0.06% (n = 10) of the total cell population. (B): Co-incubation with verapamil, a multidrug transporter inhibitor, reduced the SP population. Abbreviations: SP, side population; MP, main population.

Figure Figure 2..

Long-term expansion of unsorted, MP, and SP keratinocytes. This figure represents the complete expansion data from one representative experiment from the initial plating (day 0) to the cell sorting (day 29) and the growth potential of the three studied populations (unsorted, MP, and SP keratinocytes). After cell sorting, keratinocytes were continually passaged until their growth capacity was exhausted. Abbreviations: MP, main population; SP, side population.

Figure Figure 3..

Colony-forming efficiency (CFE) of unsorted, main population (MP), and SP keratinocytes. (A): CFE in short-term cultures (until mean passage 4 after cell sorting). For each population of keratinocytes (unsorted, MP, and SP cells), the CFE was evaluated at passages 1 to 4; a mean CFE is given. Asterisks denote a significant difference (p < .05) between MP and SP colony numbers. (B): CFE in medium-term cultures (from passage 4–5). Asterisks denote a significant difference (p < .05) between unsorted and SP colony numbers. (C): CFE in long-term cultures (from passage 5–12). The CFE values of SP keratinocytes were valuated at each passage in long term-culture; a mean CFE is given. Abbreviation: SP, side population.

Figure Figure 4..

Cell morphology and growth potential of unsorted, main population (MP), and SP keratinocytes in the colony assay. (A): Colony types in short-term cultures (until mean passage 4 after cell sorting). For each population of keratinocytes (unsorted, MP, and SP cells), the colony types were evaluated at passages 1 to 4; a mean value is given. The image of a colony representative of the major type of each population of keratinocytes (unsorted, MP, and SP cells) is presented. (B): Colony types in medium-term cultures (from passage 4–5). For each population of keratinocytes (unsorted and SP cells), the image of a representative colony is presented. (C): Colony types in long-term cultures (from passage 5–12). The colony types of SP keratinocytes were evaluated at each passage in long-term culture; a mean value is given. Abbreviation: SP, side population.

Figure Figure 5..

Capacity of SP and MP keratinocytes to regenerate a pluristratified reconstructed epidermis. (A): The MP keratinocytes at a low passage number (p3) were able to form a stratified epidermis in the organotypic model. (B): But they rapidly lost this potential throughout successive passages (p6). (C): SP cells, on the other hand, maintained long-term organotypic potential (p13). The following pattern of characteristic epidermal differentiation was observed for SP cells: a basal layer containing polygonal cells oriented perpendicularly to the underlying dermis; three to four layers of spinous cells; four to six layers of granular cells characterized by the presence of keratohyalin granules; and anucleated, flattened cornified cells forming a compact stratum corneum. Abbreviations: MP, main population; SP, side population.

Figure Figure 6..

Comparison of microarray and quantitative polymerase chain reaction (PCR) results for four genes found regulated by microarrays. ABCG2, which was not spotted on the arrays, was studied by PCR only.


We thank the Banque de Tissus pour la Recherche de l'AFM (AFM-BTR, Hôpital La Pitié-Salpétrière, Paris) for collecting human skin biopsies. We also thank Cyrille Petat and Olivier Alibert for help in bioinformatics analysis, Pascale Le Minter for technical assistance, and Corinne Ferraris for helpful discussions. This work was supported by a European grant (Naimori, FP5) and an EDF grant (Comité de Radioprotection d'Electricité de France).


The authors indicate no potential conflicts of interest.