A Stable Niche Supports Long-Term Maintenance of Human Epidermal Stem Cells in Organotypic Cultures



Stem cells in human interfollicular epidermis are still difficult to identify, mainly because of a lack of definitive markers and the inability to label human beings for label-retaining cells (LRCs). Here, we report that LRCs could be identified and localized in organotypic cultures (OTCs) made with human cells. Labeling cultures for 2 weeks with iododeoxyuridine (IdU) and then chasing for 6–10 weeks left <1% of basal cells retaining IdU label. Whole mounts demonstrated that LRCs were individually dispersed in the epidermal basal layer. Some LRCs, but not all, colocalized with cells expressing melanoma chondroitin sulfate proteoglycan, a putative stem cell marker. Although we found LRCs in both collagen- and scaffold-based OTCs, only the scaffold-OTCs supported long-term survival and regeneration. LRCs' short survival in collagen-OTCs was not due to loss of appropriate growth factors from fibroblasts. Instead, it was due to expression of metalloproteinases, especially matrix metalloproteinase (MMP)-2 and MMP-14, which caused collagen fragmentation, matrix degradation, and dislocation of specific basement membrane components bound to epidermal integrins. Blocking MMP activation not only abrogated MMP-dependent matrix degradation but also increased longevity of the epidermis and the LRCs in these cultures. Such findings indicate that the stem cell niche, the microenvironment surrounding and influencing the stem cell, is essential for stem cell survival and function, including long-term tissue regeneration.

Disclosure of potential conflicts of interest is found at the end of this article.


Author contributions: S.M. and H.-J.S.: collection and/or assembly of data, data analysis and interpretation; M.A. and B.F.-H.: collection and/or assembly of data; K.B.: figure preparation and conception; H.-J.B. and A.M.: provision of study material or patients; J.R.B.: manuscript writing, final approval of manuscript; P.B.: conception and design, manuscript writing, financial support.

Human epidermis is a stratified squamous epithelium, which provides the essential protective barrier against water loss and penetration by infectious agents. To guarantee life-long tissue integrity, the epidermis must continuously regenerate. Several studies have suggested that 1% or less of the basal cells are stem cells (reviewed in [1]). It is assumed that stem cell asymmetric division forms a hierarchy, with one daughter cell remaining as a stem cell that proliferates rarely, whereas the second daughter becomes a transit-amplifying cell, which proliferates for a finite number of cell cycles before ultimately differentiating.

The fact that stem cells rarely divide has allowed them to be identified as label-retaining cells (LRCs). It was shown that the epidermis contained slowly cycling cells that could maintain a 3H-thymidine DNA label for up to 240 days [2]. Young mice were labeled when the epidermis was highly proliferative and then chased with no label as the mice grew. This allowed the label to dilute in a proliferation-associated manner, leaving only infrequently dividing cells retaining label. LRCs have been identified as clusters in the bulge region of the hair follicles, as well as individual cells in the interfollicular epidermis (IFE) [3, 4] (reviewed in [5]). Although it is argued that hair follicle stem cells are the predominant stem cells in the epidermis, the LRC method allowed demonstration of a distinct stem cell population in the IFE that is responsible for epidermal regeneration and tissue homeostasis [6, 7]; only upon wounding do the hair follicle stem cells contribute to regeneration of the IFE [8].

Since LRC-type labeling cannot be applied to humans, attempts have been made to develop alternative protocols to characterize human IFE stem cells [9, [10], [11]–12]. Rapid adhesion to a collagen type IV matrix has been used to enrich for highly proliferative human skin keratinocytes that expressed high levels of β1- or α6-integrin subunits [9, 10]. High expression of α6-integrin combined with low expression of the transferrin receptor CD71 has been proposed as a method to isolate human epidermal stem cells [10]. Melanoma chondroitin sulfate proteoglycan (MCSP) has also been proposed as a marker for human epidermal stem cells [13]. MCSP identified a cluster of putative stem cells in the upper part of the epidermal rete ridges, which correlated with cells showing high expression of β1-integrin but did not correlate with the location of the α6highCD71low cells, which were found in the lower part of the rete ridges [9, 13]. Although expression patterns of β1highMCSPhigh and α6highCD71low did not overlap, long-lived and highly clonogenic cells were present in both populations, suggesting that stem cells in the IFE may not be restricted to either of these two sites. Grafting LacZ-marked human keratinocytes onto immune-deficient mice or infecting the keratinocytes directly on the backs of nude mice supported evidence that stem cells were intermittently distributed throughout the basal layer of human IFE [12, 14, 15].

To characterize human epidermal stem cells, an approach that allowed functional analysis was needed. We devised a method that recapitulates stem cell growth in vitro in organotypic cultures (OTCs). We previously showed that human epidermal keratinocytes can initially form an intact epidermis when grown on a collagen-based dermal equivalent. However, epidermal regeneration only lasted for 3–4 weeks [16]. Later, we demonstrated that when human fibroblasts were provided a scaffold, they developed a matrix closely authentic to human dermis, and that under these conditions epidermal tissue regeneration was supported for 15 weeks or longer [17, 18].

In the current study, we present evidence that to establish a proper stem cell microenvironment in vitro, human epidermal stem cells require an authentic dermal equivalent, rather than one artificially made from collagen type I. We show that LRCs can initially be identified in both collagen-OTC and scaffold-OTC models but that long-term maintenance and regenerative capacity of LRCs is found only in the scaffold-OTC model. Comparison of our two models provides an excellent functional approach to study the stem cell niche (i.e., the microenvironment surrounding and influencing the stem cell) and its contribution to stem cell maintenance in human tissue.

Materials and Methods

Isolation and Culture of Cells

Normal human skin samples were obtained from several body sites and from several individuals of various ages. Obtaining skin samples was approved by the Heidelberg Ethics Commission. Keratinocyte cultures were established and routinely passaged as described previously [19]. First- and second-passage cultures of keratinocytes were used for all experiments. They were established and grown on lethally irradiated normal human fibroblasts [20] in keratinocyte culture medium consisting of a 1:3 mixture of Ham's F12 and Dulbeccos's modified Eagle's medium (traditionally named FAD-medium) containing 5% fetal calf serum (FCS) and antibiotics ([21], additives as in 16). For subcultures, keratinocytes were split 1:3 to 1:5. Human dermal fibroblasts were isolated from explant cultures of de-epidermized dermis. Fibroblasts were expanded up to eight passages in Dulbecco's modified Eagle's medium with 10% FCS.


Collagen-OTCs and scaffold-OTCs were produced and handled as described [16, [17]–18]. The scaffold (Hyalograft 3D) was kindly provided by Fidia Advanced Biopolymers (Abano Terme, Italy, http://www.fidiapharma.com). For dermal equivalents, 6 × 104 fibroblasts per cm2 were inoculated in collagen-OTCs (collagen I from rat tail tendon), and 1.8 × 105 were inoculated in scaffold-OTCs. After preincubation for at least 1 day in OTC medium (FAD with 10% FCS, 50 μg/ml l-ascorbic acid, and 500 units/ml of aprotinin [Bayer, Leverkusen, Germany, http://www.bayer.com]), 3 × 105 keratinocytes per cm2 were seeded on top in OTC medium, and the dermal equivalent was kept submerged for 1 day and then raised to the air medium interphase and kept air-exposed in OTC medium with 200 units/ml aprotinin until being processed for routine histology or frozen for cryostat sectioning and immunohistochemistry. When indicated, 4 μg/ml Ilomastat (Millipore, Billerica, MA, http://www.millipore.com) was added to the medium to inhibit matrix metalloproteinase (MMP) activity.


Frozen sections from OTCs and different donor skins (different sex, age, and site) were fixed in 80% methanol (5 minutes, 4°C) and acetone (2 minutes, −20°C) or in acetone only (10 minutes, 4°C). After rehydration and blocking in 3% bovine serum albumin (BSA) in phosphate-buffered saline (PBS), primary antibodies were applied for 1.5 hours at room temperature (RT) or overnight at 4°C. Sections were washed and incubated with fluorochrome-conjugated secondary antibodies for 40 minutes (RT) with 5 μg/ml Hoechst bisbenzimide to stain nuclei, washed, mounted in Permafluor (Beckman Coulter, Krefeld, Germany, http://www.beckmancoulter.com), and examined using an Olympus AX-70 microscope (Olympus, Hamburg, Germany, http://www.olympus-global.com) equipped with epifluorescence illumination. Digital images were captured with a charge-coupled device camera (F-View) using AnalySIS Pro 6.0 software (both Soft Imaging Systems, Muenster, Germany, purchased at http://www.olympus-sis.com).

To detect 5′-iodo-2′-deoxyuridine (IdU) or 5′-bromo-2′-deoxyuridine (BrdU)-labeled cells, sections were fixed in 100% ethanol at −20°C for 10 minutes, incubated with 2 N HCl for 5 minutes at RT, and blocked with a Streptavidin/Biotin Blocking Kit (Vector Laboratories, Burlingame, CA, http://www.vectorlabs.com), followed by blocking with 3% BSA or 5% normal serum from the host species of the secondary antibody. Sections were incubated with primary antibody, washed, incubated with biotin-conjugated secondary antibody for 40 minutes at RT, washed, incubated with fluorochrome-conjugated streptavidin plus 5 μg/ml Hoechst bisbenzimide for 30 minutes at RT, and mounted in Permafluor. For double staining of IdU and keratin or collagen IV, antibodies were incubated simultaneously. For double staining of IdU and MCSP or Ki67, antibodies were incubated consecutively.

Primary antibodies and their sources are listed in supplemental online Table 1. Two mouse monoclonal antibodies against MCSP with different IgG subtypes were used to allow double staining with IdU by the use of subtype-specific secondary antibodies. Secondary antibodies were obtained from Dianova (Hamburg, Germany, http://www.dianova.de) and Molecular Probes (Karlsruhe, Germany, http://probes.invitrogen.com). Hoechst 33258 bisbenzimide for nuclear staining was obtained from Sigma-Aldrich (Taufkirchen, Germany, http://www.sigmaaldrich.com).

BrdU and IdU Labeling Procedure

OTCs were labeled by adding either BrdU/5′-bromo-2′-deoxycytidine or equimolar IdU/5′-iodo-2′-deoxycytidine (IdC) (all from MP Biomedicals, Irvine CA, http://www.mpbio.com) to the culture medium. Pulse-labeling began on the day that keratinocytes were seeded onto the dermal equivalents. Several concentrations (1, 5, and 10 μM) and pulses (two pulses of 6 hours, two to six pulses of 12 hours, and one to three pulses of 24 hours) were tried. Cultures were harvested at different time points and analyzed for BrdU or IdU labeled nuclei by immunofluorescence. Optimal labeling with 1 μM IdU/IdC for two pulses of 12 hours for collagen-OTCs and 2 continuous weeks for scaffold-OTCs was determined. To determine the initial labeling efficiency, cultures were harvested 12 hours after the last pulse or after the end of a 2-week labeling period. Chase periods for cultures ranged from 1.5 to 10 weeks. Two independent experiments were performed with collagen-OTCs, and five independent experiments were performed with scaffold-OTCs, to determine the number of LRCs. Comparisons were made between different labeling regimens and different chase periods ranging from 2 to 10 weeks with at least two replicates for each regimen and chase period.

Quantification of IdU-Labeled Cells

Sections of scaffold-OTCs (two per time point) were stained for IdU and counterstained with Hoechst to detect nuclei. Digital images of the entire section were captured. Total number of nuclei in the basal layer and the number of IdU-labeled basal nuclei were counted for each sampling time in discontinuous sections for each culture. For cultures from chase day 0, only one section per culture was counted, whereas homogeneous label of all basal cells was confirmed by microscopy on five additional sections. For chase periods of 2, 4, and 6 weeks, cells were considered IdU-labeled when the label covered more than two-thirds of the nucleus. Nuclei with only a few speckles of IdU labeling were considered negative and likely had undergone several cell divisions with subsequent dilution, as described recently [22]. For the 2-week chase period, 2 sections (∼4,000 nuclei) from each culture were analyzed; for the 4-week chase, 6 or 7 sections (∼15,200 nuclei) from each culture were analyzed; and for the 6-week chase, 14–18 sections (∼28,400 nuclei) from each culture were analyzed. Percentages of IdU-positive cells were calculated for each section, and statistical analysis was performed to compare the percentage of LRCs between the 4- and 6-week chase periods. Values from both time points were normally distributed. Unpaired t test (Prism 4; GraphPad Software, Inc., San Diego, http://www.graphpad.com) was performed and showed a statistically significant difference between percentages of LRCs after the 4- and 6-week chase period (p < .0001; α = 0.05). The results were depicted as scatter plots with means ± SDs.

Whole-Mount Preparation and Staining

The method for mouse tail epidermal whole mounts [23] was modified for human epidermis in OTCs by peeling off the epidermis from the dermal equivalent without prior treatment. Epidermal sheets were fixed in 3.7% formaldehyde in PBS for 2 hours at RT, rinsed twice with PBS, and stored in PBS with 0.2% NaN3 at 4°C. For fluorescence staining of Ki67 and IdU, epidermal sheets were placed in blocking buffer (3% BSA/0.5% Triton X-100 in PBS) for 1 hour at RT. Ki67 antibody was applied as a dilution in blocking buffer and incubated overnight at RT with gentle agitation. After being washed five times for 45 minutes in washing buffer (0.2% Tween 20 in PBS), secondary antibody (anti-rabbit, Cy2), diluted in 0.5% Triton X-100 in PBS, was incubated overnight at RT. Following five 45-minute washes in washing buffer, whole mounts were incubated with 2N HCl for 25 minutes at RT for subsequent staining of IdU. Samples were washed in washing buffer four times for 10 minutes each time and then blocked with blocking buffer for 1 hour at RT. IdU antibody diluted in blocking buffer was incubated overnight. Washing steps, incubation of the appropriate secondary antibody (anti-mouse Cy3) with the addition of 5 μg/ml Hoechst bisbenzimide to stain nuclei, and final washing steps were performed according to the described protocol. After the last washing step, epidermal sheets were mounted in Permafluor and analyzed with the Olympus AX-70 microscope as described for stained sections.

In Situ Zymography for Localization of Proteolytic Activity

To visualize in situ gelatinolytic activity, we used the highly quenched, fluorescein-labeled DQ-gelatin (EnzChek; Molecular Probes). Upon local proteolytic cleavage, quenching is released, and fluorescence signals can be visualized and quantified [24]. Unfixed cryostat sections (6–8 μm) were incubated in a solution containing 40 μg/ml DQ-gelatin and 5 μg/ml 4,6-diamidino-2-phenylindole for 30 minutes at RT and then coverslipped. Without washing, the sections were examined under a fluorescence microscope, and fluorescence was documented.

RNA Isolation and Reverse Transcription-Polymerase Chain Reaction Analysis

Total RNA was isolated from Tissue-Tek OCT Compound (Sakura Finetek, Torrance, CA, http://www.sakura.com)-frozen tissue sections or from dermal equivalents using RNeasy according to the manufacturer's instructions (Qiagen, Hilden, Germany, http://www1.qiagen.com). One microgram of total RNA from each sample was reverse-transcribed to cDNA (Omniscript; Qiagen) as described in detail previously [25]. The cDNA template was used in a polymerase chain reaction (PCR) with the primers listed in supplemental online Table 2. PCR conditions were as follows: for glyceraldehyde-3-phosphate dehydrogenase, denaturation for 2 minutes at 95°C; 30 seconds at 94°C, 30 seconds at 60°C, and 1 minute at 72°C (23 cycles); and final extension for 10 minutes at 72°C; and for the growth factors, 15 seconds at 94°C, 30 seconds at 55°C, and 30 seconds at 72°C (25–30 cycles).


Putative Stem Cell Markers Did Not Define a Population of Stem Cells in Human IFE

Aiming to identify the stem cells in human IFE, we investigated expression of all previously described epidermal stem cell markers, including β1- and α6-integrins, MCSP, the transferrin receptor CD71 and p63, as well as stem cell markers assigned to the hair follicle bulge cells, namely keratins 15, 19, and CD200. Furthermore, other potential stem cell markers were analyzed, such as the embryonic transcription factor oct4, components of drug pumps (ABCG2 and ABCB1), and stem cell markers from other tissues (CD34, CD133, CD117, or the frizzled receptor Fz4). However, even with this extensive panel of markers, identification of the stem cells as a specific subpopulation of basal keratinocytes was not possible (details given in supplemental online data).

IdU LRCs Can Be Identified in OTCs of Human Skin

One important functional characteristic of epidermal stem cells is their low rate of proliferation. In mouse epidermis, interfollicular and follicular stem cells have been identified by long-term retention of a DNA label [2, 4, 26]. Since we previously demonstrated that the epidermis in scaffold-based OTCs was regenerated and maintained for >15 weeks [17], we anticipated finding stem cells in this epidermis. To label them, we reasoned that we needed to add label at a phase of intense proliferation. Since this was provided only during early tissue reconstruction [17], we added BrdU directly after seeding the keratinocytes onto the dermal equivalent. To our surprise, the epidermis formed in the presence of BrdU failed to develop properly (Fig. 1A, 1B), even when concentrations and labeling times were decreased (1 μM for one 24-hour pulse). Thus, BrdU appeared to be toxic for human skin keratinocytes in vitro.

Figure Figure 1..

Effect of BrdU and IdU labeling on tissue regeneration and organization in OTCs. (A, B): col-OTCs were labeled with BrdU (1 μM; one time for 24 hours). Histological sections of 2-w-old cultures show deficient tissue organization in BrdU-labeled (A) versus unlabeled (B) cultures. (C, D): col-OTCs were labeled with IdU (5 μM; one time for 24 hours). Histological sections of 2-w-old cultures show no difference in tissue organization between labeled (C) and unlabeled (D) cultures. (E–H): Histology of sca-OTCs labeled with IdU (1 μM; 14 days) shows no detrimental effects of labeling on tissue regeneration in short-term 2-w cultures (E) or long-term 8-w cultures (F). Lower-magnification pictures show the early hyperplastic stage (2 w) of growth with a still thin stratum corneum (G) and, in contrast, an 8-w-old epithelium at the stage of tissue homeostasis (thinner epithelium) with an extensive stratum corneum (H). Scale bars = 25 μm (A–F) and 50 μm (G, H). Abbreviations: BrdU, 5′-bromo-2′-deoxyuridine; col, collagen; IdU, 5′-iodo-2′-deoxyuridine; OTC, organotypic culture; sca, scaffold; w, week.

However, substitution of BrdU with IdU allowed us to regenerate a well-formed epidermis no matter what concentration (1–10 μM) or labeling scheme we used. The optimal labeling scheme was determined to be 1 μM IdU given either twice for 12 hours for the collagen-OTCs (Fig. 1C, 1D) or continuously for 2 weeks for the scaffold-OTCs (Fig. 1E–1H). Initially, epidermal tissue was regenerated with a well-stratified and differentiated structure in both types of OTCs. The epidermis in the collagen-OTCs remained vital for only 3–4 weeks, whereas in the scaffold-OTCs, the epidermis maintained vitality for longer than 12 weeks. This allowed us to use chase periods after labeling for >6 weeks and thus made the scaffold-OTCs a good model in which to examine LRCs.

After labeling with IdU for 2 days (two times for 12 hours each time), nearly all cells were labeled (Fig. 2A). By 2 weeks after labeling, LRCs were identified that remained in the basal layer, whereas other labeled cells could be seen in the suprabasal differentiating layers (Fig. 2B). The most obvious difference between the various labeling protocols was the intensity of the label in individual nuclei. Although the cells appeared heavily labeled directly after labeling (Fig. 2C), only small foci remained visible after chase periods of 3–5 weeks in the cultures labeled for <1 week (Fig. 2D). This finding agrees with the high proliferative rate seen during the first period of tissue development [17] and agrees with the colocalization of labeled cells with Ki67 (Fig. 3A, inset). Since stem cells must be present in the initial formation of the tissues, these combined results suggest that stem cells are proliferating during tissue regeneration, and as such are labeled at this early stage with IdU.

Figure Figure 2..

IdU label-retaining cells (LRCs) proliferate and contribute to tissue regeneration. (A): After labeling with IdU for 2 d, >95% of basal cells stained positive for IdU in col-OTCs. An example of col-OTC labeled with IdU (1 μM; two times for 12 hours each) is shown. (B): After a chase period of 2 w, rare LRCs (red) were still labeled in col-OTCs. col-OTC labeled with IdU (5 μM; two times for 12 hours each) is shown. (C): In 6-d-old sca-OTCs treated with six 12-hour pulses of 1 μM IdU, ∼100% of the epidermal basal cells were labeled 12 hours after the final pulse (red). (D): In sca-OTCs labeled with IdU <1 w (example of 1 μM, six times for 12 hours each, is shown), LRCs with small foci of IdU were detectable after a 3-w chase period (red). (E): Treatment of sca-OTCs with 1 μM IdU for 2 w labeled ∼100% of the epidermal cells (red). (F): The percentage of labeled cells (red) was reduced to ∼40% after a 2-w chase (4 w in culture). The number was further reduced to only single dispersed LRCs after 4-w (G) and 6-w (H) chase periods. (I): IdU-positive LRCs (red) and Ki67-positive proliferating cells (green arrows) in 8-w-old culture (6-w chase). (J): In rare cases Ki67-positive LRCs were found (yellow arrow), demonstrating that LRCs can still proliferate and thus participate in tissue regeneration. All sections were counterstained with Hoechst to identify nuclei (blue). White arrows mark LRCs. Dotted lines indicate basement membrane. Scale bars = 50 μm (A–H) and 25 μm (I, J). (K, L): Percentage of IdU-labeled epithelial basal cells after various chase periods. (K): Decrease in percentage of IdU-labeled cells over time (0 = initial labeling, n = 2; 2-w chase period, n = 4; 4-w chase, n = 13; 6-w chase, n = 32). Means ± SDs. n = number of sections counted. (L): Scatter plot depicting statistical analysis of percentage of IdU-LRCs at 4- and 6-w chase periods. Means ± SDs (p < .0001). Abbreviations: col, collagen; d, day; IdU, iododeoxyuridine; OTC, organotypic culture; sca, scaffold; w, week.

Figure Figure 3..

Whole mounts of epidermal sheets from organotypic cultures (OTCs) allow for detection of the number and distribution of label-retaining cells (LRCs) up to a chase period of 10 w. (A): Directly after labeling, double staining of IdU and Ki67 showed many proliferating labeled cells (inset). (B–D): Epithelium after 6 (B), 8 (C), and 10 (D) w of chase, stained for IdU and Ki67 (B, C) or IdU only (D). (E–G): Distribution of LRCs (after 8 w of chase) in a single epithelium, showing areas of multiple (E), few (F), or hardly any (G) LRCs. (H, I): Even after 10 w of chase, some LRCs costained with Ki67 (arrows), demonstrating proliferation of long-term LRCs also. (J, K): Some IdU-labeled LRCs (red) colocalized with MCSP (green) in scaffold-OTCs (2-w label) after a chase period of 4 w (J). Likewise, a similar number of LRCs showed no costaining with MCSP after a 4-w chase (K). White arrows indicate MCSP-negative LRCs; yellow arrow marks MCSP-positive LRC. Hoechst (blue) stained all nuclei. Dotted lines indicate basement membrane. Scale bars = 100 μm (A–G), 50 μm (H, I), and 25 μm (J, K). Abbreviations: d, day; IdU, iododeoxyuridine; MCSP, melanoma chondroitin sulfate proteoglycan; w, week.

LRCs Contribute to Tissue Regeneration in Scaffold-OTCs

With a constant labeling for 2 weeks, label intensity of LRCs was maintained over a 6–10-week chase period, allowing us to accurately determine the percentage of LRCs over time in the scaffold-OTCs. At the end of the labeling period, the percentage of labeled basal keratinocytes was nearly 100% (Fig. 2E, 2K). The number of labeled cells was reduced to approximately 40% after 2 weeks (Fig. 2F, 2K) and to only individual basal LRCs after 4 and 6 weeks of chase (Fig. 2G, 2H). Quantifying LRCs in discontinuous sections revealed a frequency of 2.8% after 4 weeks and 0.9% after 6 weeks (Fig. 2K, 2L). At both time points, the LRCs did not seem to be evenly distributed but were dispersed throughout the basal layer, with some sections containing more, some sections containing fewer, and some sections containing no LRCs.

This fate and this distribution of labeled cells were even more obvious when analyzing whole mount preparations of the epidermis. The initial ∼100% labeled cells (Fig. 3A) rapidly decreased to ∼1% of the basal cells by the end of the 6-week chase period (Fig. 3B). The number of LRCs was only slightly decreased after 8 and 10 weeks of chase (Fig. 3C, 3D), suggesting that a homeostasis had been established with the stem cells. The whole mount analyses further highlighted the uneven distribution of LRCs. In Figure 3, the same OTC shows areas with several (Fig. 3E), areas with few (Fig. 3F), and areas with hardly any or no LRCs (Fig. 3G). We saw a similar distribution in an OTC derived from keratinocytes obtained from a different donor. Thus, the number of LRCs decreased with time to ∼1% of the basal cell population in long-term cultures. Furthermore, distribution of LRCs showed no discernable pattern. This is similar to what was reported for mouse IFE [23] and is confirmed by double labeling with MCSP. When we investigated the distribution of LRCs in relation to the expression pattern of MCSP (as the most distinct and clearly defined stem cell marker), we found that some LRCs double-stained for MCSP, some did not, and some MCSP-positive cell clusters lacked any LRCs (Fig. 3J, 3K) for time points up to 6 weeks of chase.

Staining with the proliferation marker Ki67 demonstrated that many proliferating cells were also IdU-labeled at the beginning of the chase period (Fig. 3A, inset). By the end of the chase periods, the number of IdU-labeled cells had greatly decreased to leave LRCs. At these later times, Ki67-positive cells were clearly identified close to LRCs both in sections and in whole mounts (Figs. 2L, 3B, 3C). Importantly, some Ki67-labeled LRCs were seen at all time points, including specimens from the 10-week chase period (Figs. 2J, 3H, 3I). This suggests that although LRCs in the long-term OTCs are largely quiescent under homeostatic conditions, they are capable of proliferation. These data indicate that IdU labeling did not hinder the development of a normal epidermis and that LRCs proliferate in scaffold-OTC cultures and contribute to tissue regeneration.

The Stem Cell Microenvironment Determines Survival and Long-Term Regenerative Capacity of Epidermal Stem Cells

Both collagen-OTCs and scaffold-OTCs allowed labeling of LRCs in the epidermal basal layer. This suggests that irrespective of the dermal equivalent used, initially a stem cell hierarchy was established in vitro. However, only the scaffold-OTCs supported long-term regeneration. Since the fibroblasts in the scaffold-OTC established a dermal equivalent similar to that in skin in situ [17, 27], only the scaffold-OTC provided an authentic stem cell microenvironment to support the stem cells, allowing them to support epidermal regeneration. To test this hypothesis, we first evaluated the distribution of several putative stem cell markers in both types of OTCs. In scaffold-OTCs, expression of β1- and α6-integrins was restricted to the basal cells, with a distribution largely identical to that of intact skin (compare Fig. 4A, 4C with supplemental online Fig. 1A, 1B). In contrast, in the collagen-OTCs, a massive shedding had occurred that was visible as a distinct β1- or α6-positive layer within the dermal collagen matrix below the basement membrane, with little protein remaining with the basal keratinocytes (Fig. 4B, 4D). MCSP-positive cells were clustered in the basal layer of the epidermis in scaffold-OTCs, similar to normal skin with no rete ridges (compare Fig. 4E with Fig. 4G), whereas in collagen-OTCs, MCSP expression showed little or no clustering (Fig. 4F). Expression of keratins K19 and K15, both postulated stem cell markers of the hair follicle bulge region, showed no differences between the two OTCs. Keratin 19 stained small-size groups or single basal cells (Fig. 4H), whereas keratin 15 stained nearly all basal cells, as it does in rete ridge-poor epidermis (Fig. 4I, 4J).

Figure Figure 4..

sca-OTCs, but not col-OTCs, provide an environment that allows for maintenance of epidermal stem cells and thus long-term epidermal regeneration. (A–F): Immunofluorescence analysis of 2-week-old sca-OTCs (A, C, E) and col-OTCs (B, D, F). Distribution of β1-integrin (red) was normal in sca-OTCs (A) but showed shedding in col-OTCs (B). α6-Integrin (red) was normally localized in sca-OTCs (C) but shed in col-OTCs (D). MCSP (green) expression was clustered in basal cells of sca-OTCs (E), similar to that seen in normal, rete ridge-poor, human breast skin (G). In col-OTCs, MCSP-positive basal cells were distributed as smaller-sized groups of cells or single cells (F). Keratin 19 (green) showed a scattered distribution, with positive cells found as small groups or single cells mainly in the basal layer and only a few suprabasal cells (H) (2-week-old sca-OTC shown). Keratin 15 (green) was widely present in basal cells, with some variation in signal intensity in both normal human breast skin (I) and OTCs (J) (2-week-old sca-OTC shown). Scale bars = 50 μm (A–F, H, J) and 25 μm (G, I). Abbreviations: col, collagen; MCSP, melanoma chondroitin sulfate proteoglycan; OTC, organotypic culture; sca, scaffold.

Loss of the Epidermal Regeneration in Collagen-OTCs Is Not Due to Changes in Fibroblast-Produced Growth Factors

Since fibroblast-produced growth factors can regulate the growth and differentiation of the keratinocytes [28, 29], we reasoned that the restricted maintenance of the epidermis in the collagen-OTCs could be due to the exhaustion of fibroblasts to provide the appropriate growth factors. To test this, we performed reverse transcription PCR for amphiregulin, fibroblast growth factor-2 (FGF-2), insulin-like growth factor 2, and transforming growth factor-α. These were chosen on the basis of an Affymetrix (Santa Clara, CA, http://www.affymetrix.com) analysis of 1-week-old OTCs, which showed that they were increased only when keratinocytes were added onto the dermal equivalents as compared with keratinocyte-free dermal equivalents (data not shown). We found mRNA expression of all four growth factors in 1-week-old collagen-OTCs with induction by cocultivated keratinocytes (Fig. 5, lanes 1 and 2). Expression of nearly all growth factors continued mostly unchanged throughout the 3–4-week culture period (Fig. 5, lane 3). Only the expression of FGF-2 was slightly decreased. Thus, it is not the lack of these growth factors that contributes to the loss of epidermal vitality in collagen-OTCs.

Figure Figure 5..

mRNA expression of growth factors is unchanged in collagen-organotypic cultures (OTCs) over time, and only FGF-2 shows a slight decrease of expression. Shown is reverse transcription polymerase chain reaction analysis of amphiregulin, FGF-2, IGF-2, and TGF-α. Lane 1 for each set, 1-week-old OTCs without keratinocytes; lane 2, 1-week-old OTCs with keratinocytes; lane 3, 4-week-old OTCs with keratinocytes. Abbreviations: FGF, fibroblast growth factor; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; IGF, insulin-like growth factor; M, markers; TGF, transforming growth factor.

Destruction of the Stem Cell Niche Is Causal for Epidermal Atrophy in Collagen-OTCs

Even though the fibroblasts appeared the same in both types of OTCs, the epidermis in the collagen-OTCs was not maintained, and β1- and α6-integrins were shed (Fig. 4B, 4D). Since this suggested that components in the basement membrane and dermal matrix had been degraded, thereby altering the stem cell environment in the collagen-OTCs, we next examined the expression and localization of several basement membrane components. In scaffold-OTCs, collagen types IV (coll IV) and VII (coll VII), laminins 5 (laminin 332) and 10 (laminin 511), and nidogen showed expression patterns comparable to that of skin in situ, distinctly marking the basement membrane zone (exemplified here for nidogen and coll VII; Fig. 6A, 6C). In the collagen-OTCs, coll VII and laminin 10 were properly located, although coll VII was partially still inside the basal keratinocytes and thus was not processed properly (Fig. 6D). In contrast, nidogen and laminin 5, binding partners of β1- and α6-integrins, were co-shed from the basement membrane zone, indicating cleavage and dislodging of the entire integrin/basement membrane complex (compare Fig. 6B with Fig. 4B, 4D).

Figure Figure 6..

col-OTCs show disintegration of basement membrane and dermal matrix components. (A–F): Comparative immunofluorescence analysis of 3-week-old sca-OTCs (A, C, E) and col-OTCs (B, D, F). Costaining of α6-integrin (green) and nidogen (red) showed a normal basement membrane organization in sca-OTCs (A), whereas massive co-shedding in col-OTCs led to the appearance of a split basement membrane (B). col type VII (red) was deposited at the basement membrane in both sca-OTCs (C) and col-OTCs (D); however, basal cells in col-OTCs showed considerable cytoplasmic staining, indicative of impaired protein secretion (enlarged in inset [D]). The distribution of keratin 2e (green) was similar in both types of OTCs. β4-Integrin (green) was regularly localized in sca-OTCs (E) but displayed a more irregular diffuse subepithelial pattern in col-OTCs (F). col type I (red) showed a fibrillar organization in sca-OTCs (E) versus a sponge-like texture in col-OTCs (F). Scale bars = 50 μm (A–F) and 10 μm (inset [D]). Abbreviations: col, collagen; OTC, organotypic culture; sca, scaffold.

Loss of the Epidermal Stem Cell Niche Is Due to Increased MMP Activity

One major difference between scaffold- and collagen-OTCs is the appearance of collagen type I, which showed a higher content and a more sponge-like distribution in the collagen-OTCs as compared with the more fibrous texture in scaffold-OTCs (Fig. 6E, 6F). Since it is well known that collagen type I can trigger MMP activity, we reasoned that an active degradation by MMPs may have been responsible for the above described changes. In scaffold-OTCs, expression and distribution of MMP-1, MMP-2, and MMP-14, major players in extracellular matrix remodeling, resembled those found in normal skin, with little to no MMP-1 (data not shown) and some MMP-2 expressed in the nucleated epithelial layers. MMP-14, a pro-MMP activator, was found in the upper part of the epidermis (Fig. 7A), where it may be involved in remodeling during terminal differentiation. In contrast, MMP-1 had accumulated beneath the epithelium in collagen-OTCs (data not shown), and MMP-2 was present in the epidermis, but also in the dermal part of the collagen-OTC below the epidermis (Fig. 7B). MMP-14 was found throughout the epidermis, and most importantly in the basal keratinocytes (Fig. 7B), arguing for MMP-activation in the basement membrane zone. Indeed, we show subepithelial gelatinolytic activity by in situ zymography (Fig. 7C, 7D) and formation of specific collagen fragments (Fig. 7E, 7F) exclusively in the collagen-OTCs.

Figure Figure 7..

Dermal equivalent integrity and functionality is destroyed by metalloproteinase degradation of matrix proteins in col-OTCs. The distribution of MMP-2 and MMP-14, their proteolytic activity, and degradation products were compared in 4-week-old sca-OTCs (A, C, E) and col-OTCs (B, D, F). MMP-2 (green) was present in low amounts in the epidermis and along the basement membrane in sca-OTCs (A) and strongly expressed, particularly in the dermal equivalent, in col-OTCs (B). MMP-14 (red) was found in the most upper vital layers of the epidermis in sca-OTCs (A) but was present throughout the entire epithelium, particularly in the basal cells, in the col-OTCs (B). In situ zymography using DQ-gelatin revealed no activity in sca-OTCs (C) but intense proteolysis (green) beneath the epidermis of col-OTCs (D), as well as in the upper part of the epidermis, which correlates to proteolytic processing involved in terminal differentiation. In agreement with lack of MMP activity, the dermal matrix of sca-OTCs showed no specific col fragments (E), detectable by C1, 2C antibody (red). In contrast, col type I was specifically fragmented in col-OTCs (F). (G–J): Effects of MMP inhibition. In situ zymography using DQ-gelatin on 5-week-old col-OTCs revealed strong activity in untreated cultures (G) and no activity in cultures treated with MMP inhibitor (H). Accordingly, shedding of α6-integrin was massive in untreated cultures (I), whereas shedding was completely prevented upon inhibition of MMP activation (J). Scale bar = 50 μm. Abbreviations: col, collagen; inhib., inhibitor; MMP, matrix metalloproteinase; OTC, organotypic culture; sca, scaffold.

To test whether or not MMPs were causal for degradation of the stem cell niche and in turn causal for epidermal atrophy, we treated the collagen-OTCs with Ilomastat, an inhibitor of MMPs, including MMP-1, -2, and -14. Not only did this treatment prevent gelatinolytic activity (Fig. 7G, 7H), more importantly, it also prevented cleavage and shedding of the integrin, nidogen, and laminin 5 complexes (exemplified here for α6-integrin; Fig. 7I, 7J). Furthermore, as hypothesized, inhibiting MMP activity doubled the life extension of an actively regenerating epidermis in the collagen-OTCs (up to 8 weeks tested so far). Thus, preserving a proper microenvironment is essential for maintaining stem cell function that is, long-term epidermal regeneration.


Despite more than 30 years of research on epidermal stem cells, unequivocal identification of the human epidermal stem cell or its niche is still elusive, primarily because of the lack of good model systems in which to study the stem cell. Using the putative stem cell markers published over the years, we tried to localize the epidermal stem cells or the stem cell niche in normal human skin samples for comparison with locations in the scaffold-OTCs. However, we found that the various markers stained different clusters of cells within the human epidermal basal layer (supplemental online Fig. 1). Thus, none of them could be used to identify only the stem cells. Another approach is to place epidermal cells in culture and perform clonal analysis [30, 31]. However, when cells are dissociated from the tissue and grown as two-dimensional standard cultures, they change their polarity, as well as cell-cell and cell-matrix interactions. Such changes may be misleading when examining stem cell and niche characteristics. Furthermore, since the most prominent functional characteristic of stem cells is their distinct proliferative character that is, their capacity to divide infrequently in undamaged steady-state epidermis [4, 30], three-dimensional in vitro models may be superior for establishing a reliable stem cell niche.

The scaffold-OTC model presented here is the first to demonstrate that a microenvironment can be established in vitro suitable for maintaining human epidermal stem cells because it is the only one that recapitulates a LRC-based stem cell hierarchy with long-term epidermal regeneration. This supported our previous finding that only the dermal equivalent formed in the scaffold-OTCs was comparable to dermis in situ [17, 18, 27]. Although both types of dermal equivalents, collagen- and scaffold-based, were sufficient for establishing an epidermis, the collagen-OTCs remained vital for only 3–4 weeks—the time required for one complete epidermal regeneration cycle. This suggests that either epidermal stem cells are not present in these cultures or, as proposed earlier, that they are forced to terminally differentiate [32]. We provide evidence that epidermal stem cells are initially present because using the same keratinocytes in the scaffold-OTC enabled us to maintain the epidermis through tissue regeneration into tissue homeostasis. We also show that by preventing degradation with a MMP inhibitor, we can extend the life span of the epidermis in the collagen-OTCs to 8 weeks instead of 3–4 weeks. This represents at least two regeneration cycles. Finally, the long-term growth of the scaffold-OTCs permitted a long chase period, making it possible to identify LRCs for the first time in organotypic cultures of normal human epidermis. This now allows us to demonstrate that (a) labeling LRCs in vitro is possible provided that appropriate tissue organization is present, (b) LRCs are present at a frequency (<1%) similar to what has been reported in vivo [33], and (c) LRCs are distributed throughout the epidermal basal layer. Thus, a stem cell hierarchy was established in the scaffold-OTC model, suggesting that a proper stem cell microenvironment was established and maintained.

To identify the distribution of the stem cells, we prepared sections and whole mounts from the OTCs. Although the latter allow examination of large areas, the analysis of tissue sections was used for a more precise study of the location of LRCs in relation to putative stem cells markers. LRCs were observed as individual, nonclustered cells in a “random” distribution, with no evidence of a pattern suggestive of epidermal proliferative units (i.e., a central stem cell surrounded by a ring of transit-amplifying cells [34, [35], [36]–37]). This accounted for all parallel cultures (>4 weeks of chase), as well as for epithelia formed by keratinocytes from two independent donors. Since a comparable distribution was also described for the interfollicular epidermis of mouse back and tail skin [23], this may suggest that in the OTCs the stem cells may re-establish a pattern similar to that in the skin in situ. The random distribution was further confirmed when comparing LRC location in relation to MCSP. MCSP-expressing cells have previously been reported to mark clusters of epidermal cells at the tops of the rete ridges and in rete ridge-poor epidermis, and MCSP has been used as a surface marker to enrich for human epidermal stem cells [13, 38]. In scaffold-OTCs, we found clusters of MCSP-positive cells similar to those seen in human skin tissues. However, MCSP did not stain all LRCs, and it did not stain only LRCs.

Logically, stem cells should reside in highly protected locations. In human skin, the question is, where are these locations? When considering UV or chemicals, which expose the skin from the outside, the bottom of the epidermal rete ridges would appear to be the best place for a stem cell niche. On the other hand, inflammation or a burst of oxidative stress comes through the dermis and would more readily damage cells in the bottom of the epidermal rete ridges. In this case, the top of the rete would appear to be the best stem cell niche. Thus, either position could be problematic, depending on the type of insult. Furthermore, not every area of skin has rete ridges to provide this kind of microenvironment. Another consideration is the loss and re-establishment of the stem cell niche postinjury, which takes place independent of structure. A dispersed distribution of stem cells would assist in this type of recovery. Our finding that LRCs are dispersed individually support the last possibility and agrees with several in vivo studies in which IFE healing was facilitated by dispersed epidermal stem cells [8, 12, 33].

Stem cells establish as LRCs in collagen-OTCs; however, after 3–4 weeks the entire epithelium is atrophic. This is not due to lack of growth factors secreted by the fibroblasts and known to be required for successful epidermal growth [28, 39]. Instead, epidermal atrophy was correlated with matrix degradation. Different from what was expected, however, this degradation process seemed to be highly specific. We show that the expression pattern of MMP-14, the pro-MMP activator (reviewed in [40]) generally expressed in the upper suprabasal layers, was now found in the basal cells, thus allowing MMP-14 to activate MMP-1, made by the basal keratinocytes, and/or MMP-2, expressed by the dermal fibroblasts. This caused degradation of the collagen matrix, as well as degradation of some basement membrane components (e.g., coll IV or laminin 10). Most unexpectedly, it also caused shedding of the integrins α6 and β1 together with two other basement membrane components, laminin 5 and nidogen. Since this process resulted in hardly any integrins remaining on the cell surface of the basal keratinocytes, we propose that this was insufficient for further stable attachment and the keratinocytes entered a lack of anchorage-dependent differentiation process. That MMPs were causal for this process is demonstrated by the fact that upon inhibition of MMP activation, none of the degradation processes occurred, consequently maintaining an intact stem cell niche that allowed the keratinocytes to remain vital and proliferatively active for at least 8 weeks—a doubling of the life span of the collagen-OTCs. The specific dislodgement of the integrin/nidogen/laminin5-complexes now prompts us to further concentrate on their role as promising candidate components regulating the stem cell niche and thus regulating long-term epidermal maintenance.


We have shown that the scaffold-OTC model allows establishment and maintenance of an interfollicular human epidermis, with a microenvironment, the stem cell niche, permitting formation of a LRC stem cell hierarchy. We determined that it was not lack of epidermal stem cells that restricts epidermal growth in collagen-OTCs but the loss of the proper microenvironment that is the crucial parameter. Understanding which components are necessary to maintain the stem cell niche and how they regulate the regenerative capacity of the stem cell itself will help to elucidate their functional impact on human epidermal stem cells in development and disease.

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflicts of interest.


We thank Iris Martin, Katrin Schmidt, Christa Etzelsdorfer, and Gaby Blaser for excellent technical assistance and Angelika Lampe for help in preparing the manuscript. We extend our thanks to Fidia Advanced Biopolymers, for kindly providing Hyalograft 3D scaffold and Baxter AG, Vienna, for the generous donation of TissueCol-fibrin glue for the preparation of scaffold-based organotypic cultures. This work was supported by the Landesstiftung Baden-Wuerttemberg and in part by the Deutsche Krebshilfe (Forschungsverbund Tumorstammzellen) (both to P.B.) and the Bundesministerium für Bildung und Forschung (grant number FKZ 03 12120) (to H.-J.S.).