Human skin keratinocytes, melanocytes, and fibroblasts contain distinct circadian clock machineries

Abstract

Skin acts as a barrier between the environment and internal organs and performs functions that are critical for the preservation of body homeostasis. In mammals, a complex network of circadian clocks and oscillators adapts physiology and behavior to environmental changes by generating circadian rhythms. These rhythms are induced in the central pacemaker and peripheral tissues by similar transcriptional–translational feedback loops involving clock genes. In this work, we investigated the presence of functional oscillators in the human skin by studying kinetics of clock gene expression in epidermal and dermal cells originating from the same donor and compared their characteristics. Primary cultures of fibroblasts, keratinocytes, and melanocytes were established from an abdominal biopsy and expression of clock genes following dexamethasone synchronization was assessed by qPCR. An original mathematical method was developed to analyze simultaneously up to nine clock genes. By fitting the oscillations to a common period, the phase relationships of the genes could be determined accurately. We thereby show the presence of functional circadian machinery in each cell type. These clockworks display specific periods and phase relationships between clock genes, suggesting regulatory mechanisms that are particular to each cell type. Taken together, our data demonstrate that skin has a complex circadian organization. Oscillators are present not only in fibroblasts but also in epidermal keratinocytes and melanocytes and are likely to act in coordination to drive rhythmic functions within the skin.

Introduction

Physiology and behavior of organisms are adapted to environmental changes by a multi-oscillatory network able to generate circadian rhythms. In this network, a central clock, localized in the suprachiasmatic nuclei of the brain and synchronized to astronomical time, sets the period and phase coherence between and within the clocks and oscillators present in the brain and peripheral organs. At the molecular level, circadian rhythms in central and in peripheral oscillators are cell-autonomous and generated by similar interconnecting transcriptional–translational feedback loops involving several transcription factors encoded by clock genes [1, 2]. Transcription of Period1–3 (Per1–3) genes and Cryptochrome1–2 (Cry1–2) genes is activated by a hetero-dimer containing the products of Bmal1 and Clock genes. In turn, PER and CRY proteins repress their own transcription by inhibiting the activity of CLOCK/BMAL1 dimer. In addition, CLOCK/BMAL1 activates expression of retinoic acid receptor-related genes Rorα-β and RevErbα, whose products in turn respectively stimulate and repress the transcription of the Bmal1 gene, further stabilizing the circadian circuitry. In addition, post-translational mechanisms regulated by distinct signaling pathways play an important role in the turnover of clock factors and ultimately lead to the generation of rhythmic processes.

Given its localization at the interface between the external environment and the body, skin performs functions that are critical for the preservation of body homeostasis, in accordance with environmental changes (e.g., temperature, humidity, light, UV irradiation). In humans, daily rhythmic variations occur in different skin functions such as barrier recovery rate [3–5], sebum production, and trans-epidermal water loss with a maximum during the day, skin temperature with a maximum at night and skin pH with a minimum at night [6]. Also, a peak of cell proliferation has been identified at night [7], when the capillary blood flow is at its maximum [4, 8].

Expression of Clock and Per1 was reported in primary human keratinocytes, melanocytes, and dermal fibroblasts as well as in skin cell lines by Zanello and coauthors [9]. Also, circadian oscillations of Bmal1, Per1, and Cry1 were shown in vivo in human skin biopsies [10]. Recent in vitro bioluminescence studies showed that primary human fibroblasts [11] and keratinocytes from the HaCat cell line [12] contain functional autonomous circadian oscillators with strong similarities to the central clock machinery. These studies also highlight the fact that these cell types represent a good model to study the role of circadian rhythms in humans. Molecular oscillators have also been described in mouse skin [13, 14].

The structural complexity of skin and the multitude of rhythmic processes taking place in its different layers lead to the hypothesis that skin may contain more than one oscillator. The aim of this study was to identify the skin cell types involved in time tracking and to investigate their molecular circadian machinery at the functional level. Our study was conducted on primary keratinocyte, melanocyte, and fibroblast cultures derived from a human skin biopsy. Clock genes were found to be expressed in the studied cell types. For the evaluation of their expression patterns, we developed an original mathematical method to analyze the set of clock genes in each cell type as a unique functional oscillator. Thus, our data confirm the presence of functional oscillators in primary fibroblasts and keratinocytes and show for the first time the presence of the molecular circadian machinery in primary human melanocytes. These autonomous oscillators might act locally and interact with signals from the central pacemaker for driving rhythmic skin functions.

Materials and methods

Skin sample and cell culture

The primary skin cell cultures were established from an abdominal skin biopsy obtained from a 36-year-old Caucasian healthy female undergoing plastic surgery, with the informed consent of the donor whose identity was kept strictly anonymous. After removal of hypodermis, the skin biopsy was rinsed twice with PBS, followed by ethanol/PBS (2/3:1/3) and a final wash in PBS.

Epidermal cell cultures

Epidermis was mechanically separated from dermis using a safety blade. The obtained epidermal strips were incubated with 0.05 % trypsin (Invitrogen, Cergy Pontoise, France) at 37 °C. After 1–1.5 h of digestion, trypsin was inactivated with M199 complete medium (E-199 medium (Invitrogen) supplemented with 10 % FBS (Biowest, Nuaillé, France), 2 mM glutamine (Invitrogen), 100 U/ml penicillin and 100 μg/ml streptomycin (Invitrogen)). The epidermal layer was gently removed from the strips by scraping the remaining dermis, collected into M199 complete medium and disaggregated into single cells by repeated pipetting. The cell suspension (containing keratinocytes and melanocytes) was filtered through a 70-μm nylon filter (BD Falcon, Le Pont de Claix, France) and centrifuged at 1,500 rpm for 5 min. Pelleted epidermal cells were resuspended with M199 complete medium supplemented with 400 ng/ml hydrocortisone (Sigma-Aldrich, Steinheim, Germany), 10 ng/ml EGF, 100 U/ml penicillin and 100 μg/ml streptomycin, seeded at 12 × 10⁶ cells in 75-cm² flasks (BD Falcon) and incubated at 37 °C in a humidified 5 % CO₂ atmosphere. At 80 % confluence, the mixed P0 epidermal cell cultures underwent differential trypsinization for separation of melanocytes from keratinocytes [15]. The P0 cultures were incubated with trypsin-EDTA 1× at 37 °C/5 % CO₂. Melanocytes preferentially detached after 30 s and were transferred to M199 complete medium. The keratinocytes were incubated at 37 °C/5 % CO₂ for an additional 4.5 min in the remaining trypsin and recovered with M199 complete medium. The two separated cell populations were centrifuged at 1,500 rpm for 5 min and the supernatant was discarded.

To obtain pure normal human keratinocyte (NHK) cultures, the isolated keratinocytes were resuspended and seeded at 1 × 10⁶ cells in 75 cm² with keratinocyte serum free medium (KSFM) (Invitrogen) supplemented with 50 μg/ml bovine pituitary extract (BPE) (Invitrogen), 5 ng/ml EGF (Invitrogen) and 100 U/ml penicillin, and 100 μg/ml streptomycin. Medium was changed every other day. Cells were split at 80 % confluence (adapted from [16]).

To obtain pure normal human melanocyte (NHM) cultures, the isolated melanocytes were resuspended and seeded at 2 × 10⁶ cells in 75 cm² flasks in melanocyte medium containing 90 % KSFM, 10 % M199 complete, 10 ng/ml βFGF (Invitrogen) and 250 ng/ml phorbol 12,13-dibutyrate (Sigma). Medium was changed every other day and the cells were split at 80 % confluence (adapted from [17], [16]).

Normal human fibroblast (NHF) culture

After removal of epidermis from the skin biopsy, dermal tissue was cut into 4 to 6-mm² pieces. Dermal explants were plated on 10-cm culture dishes and allowed 5 min to attach on the bottom of the plate. Subsequently, they were incubated with DMEM medium containing 20 % FBS, 2 mM glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin at 37 °C in a humidified 5 % CO₂ atmosphere to allow fibroblast proliferation. After 2 weeks, the dermal explants were removed and the fibroblasts were incubated with 10 % FBS/DMEM medium until confluence and then split and seeded at 1 × 10⁶ cells in 75-cm² flasks (adapted from [18], [19]).

Experimental design

Normal human keratinocyte (1.3 × 10⁴ cells/cm²), NHM (2.5 × 10⁴ cells/cm²), and NHF (1 × 10⁴ cells/cm²) cultures at P3 were prepared in triplicate. At 80 % confluence, cultures were treated with 100 nM water-soluble dexamethasone (Sigma-Aldrich) to synchronize the cells (time = 0 h) and after 20 min of incubation the medium was replaced with the cell-type-specific culture medium. All cultures were incubated at 37 °C in a humidified 5 % CO₂ atmosphere. At indicated times, the wells (n = 3 per time point) were washed twice with ice-cold PBS and the cells were immediately lysed for RNA extraction. Lysates were immediately frozen on dry ice and stored at −80 °C.

Extraction of total RNA

Total RNA was extracted from cultured cells by using the RNeasy Mini kit (Qiagen Gmbh, Hilden, Germany) according to the manufacturer’s instructions. On-column digestion with DNase was performed to ensure removal of possible genomic DNA contamination. Total RNA was eluted with 30 μl of nuclease-free water. Total RNA concentration and purity were measured using a NanoDrop ND-1,000 V 3.5 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA) (A₂₆₀/A₂₈₀ and A₂₆₀/A₂₃₀ values were between 1.8 and 2.0). Integrity of the RNA was assessed using the 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). The RNA integrity numbers (RIN) were between 9 and 10.

Reverse transcription

Total RNA (1 μg) was reverse transcribed into first-strand cDNA using the high-capacity-RNA-to-cDNA kit (Applied Biosystems, Foster City, USA) according to the manufacturer’s protocol. All cDNA samples were stored at −80 °C.

Evaluation of cell culture purity and glucocorticoid receptor expression

cDNAs from NHK, NHM, and NHF cultures were submitted to PCR (30 cycles) with primers (Sigma-Aldrich) for tyrosinase (RefSeq NM_000372.4) (forward 5′TACGGCGTAATCCTGGAAAC3′, reverse 5′ATTGTGCATGCTGCTTTGAG3′), Keratin 5 (RefSeq NM_000424.3) (forward 5′GCAGCCGGACAGAAGCCGAG3′, reverse 5′TCGGCATCCGCAATGGCGTT3′) Collagen type III, alpha 1 (Col3A1) (RefSeq NM_000090.3) (forward 5′CTGGTGAACGTGGACCTCCTGGATTGG3′, reverse 5′TGGTTCACCCTTGTCACCCTTTGGACC3′) and glucocorticoid receptor (GR) (RefSeq NM_000176.2) (forward 5′GACCAGATGTAAGCTCTCCT3′, reverse 5′GTGGTAACGTTGCAGGAACT3′).

Real-time quantitative PCR

Expression of Clock, Bmal1, Per1, Per2, Per3, Cry1, Cry2, Rorα, Rorβ, and RevErbα genes was analyzed by real-time PCR in NHK, NHM, and NHF cultures during 52 h, starting 11 h after dexamethasone synchronization.

The real-time quantitative PCR was performed using the 7300 real-time PCR system (Applied Biosystems) and the hydrolyzed probe-based TaqMan chemistry. We used optimized TaqMan gene expression assays designed to specifically amplify mRNA (Applied Biosystems; Table 1). The PCR conditions were: 1× TaqMan universal PCR master mix, no. AMPErase UNG (Applied Biosystems), 1× gene expression assay mix (containing forward and reverse primers and cognate probe; Applied Biosystems) and 1 μl of cDNA in a total volume of 20 μl. The PCR program was as follows: 10 min at 95 °C, followed by 40 cycles of denaturation at 95 °C for 15 s and annealing–elongation at 60 °C for 1 min. The acquisition of fluorescence data was performed at the end of the elongation step using the 7300 system sequence detection software V 1.3.1 (Applied Biosystems). Each PCR reaction was done in duplicate. A dilution curve of the pool of all cDNA samples was used to calculate the amplification efficiency for each assay (values were between 1.8 and 2 for all assays). No-template control (NTC) reactions were performed as negative controls for each assay. One 96-well plate corresponded to the analysis of one gene within one cell type. Transcript levels were normalized using β-actin, B2 M, and GAPDH.

Table 1.

TaqMan gene expression assays used in the study

Gene	TaqMan assay reference	RefSeq	Exon boundary	Assay location	Amplicon length (bp)
Clock	Hs00231857_m1	NM_004898.2	18–19	1,879	88
Bmal1	Hs00154147_m1	NM_001030272.1	8–9	815	112
Per1	Hs00242988_m1	NM_002616.2	22–23	3,837	66
Per2	Hs00256143_m1	NM_022817.2	8–9	1,208	121
Per3	Hs00997925_m1	NM_016831.1	20–21	3,695	104
Cry1	Hs00172734_m1	NM_004075.3	2–3	1,124	84
Cry2	Hs00323654_m1	NM_021117.3	5–6	832	75
Rorα	Hs00536545_m1	NM_134260.2	7–8	1,197	102
Rorβ	Hs00199445_m1	NM_006914.3	4–5	1,278	66
RevErbα	Hs00253876_m1	NM_021724.3	1–2	660	60
β-actin	Hs99999903_m1	NM_001101.3	1–1	53	171
GAPDH	Hs99999905_m1	NM_002046.3	3–3	157	122
B2M	Hs00984230_m1	NM_004048.2	3–4	431	81

Reference (Applied Biosystems), GenBank accession number (RefSeq), location of the assay in the gene (nucleotide number indicates position of the probe in the RefSeq sequence), size (bp base pairs) of the amplicons are given for all assays

Gene expression assays were designed across the exon–exon junction indicated in the table

qPCR data analysis was done using the qBase software (free v1.3.5) [20] for the management and automated analysis of qPCR data. Expression of target genes was quantified based on the ΔCq method modified to take into account gene-specific amplification efficiencies and multiple reference genes [20]. Transcript levels were calculated relative to the sample showing lowest expression, which was rescaled to one.

Statistical analysis

Results are presented as mean (or parameter value) ± SE. Data were treated as independent plots. All the fittings were performed by non-linear least squares regression, with the help of the mathematical software R (http://www.r-project.org) in double precision.

The following procedure was applied successively to each cell type. In the first step, the data for each gene were fitted to a cosinor-derived sine wave function added to a linear drift term y = y0 + b × t + c × cos (2π(t-tm-φ)/τ), where b is the drift, c is the amplitude (half of peak-to-trough amplitude), t is time (h), tm is the 31 h time point of the experiment, τ is the endogenous period (h) and φ is the phase (h). For some genes, the drift term was non-significant and was cancelled. The 31-h time point corresponded to approximately half the maximum time (63 h) and was selected to minimize the spurious effects of phase heterogeneity. Calculations were performed on raw data, except for fibroblasts: for these, the non-Gaussian distribution of raw data was corrected by taking the logarithms (log normal distribution). The algorithm minimized the residual sum of squares by an iterative optimization method. At the end of the process, the software provided an asymptotic estimate (by excess) of the standard error of each parameter, together with the corresponding p value. It also provided a global ANOVA for the regression, which was tested with F test. Shapiro–Wilk test was used to test the homogeneity of residuals, and Bartlett’s test for the homogeneity of variances. The absence of excessive colinearity was determined with the correlation matrix of the parameters. The results were considered acceptable only if all the following conditions were simultaneously fulfilled: (1) non-significant Shapiro–Wilk test (homogeneity of residuals) and Bartlett’s test (homogeneity of variances); (2) partial correlation coefficients between parameters less than 0.6; (3) significant p value (F test) for the global regression; (4) significant p values (t test) for each of the parameters.

The presence of damping was tested systematically by adding an exponential decrement, which never reached statistical significance. Colinearity between the parameters was nearly always present. The value of y0 was used as a constant, and the calculation was repeated with only the four remaining parameters. Under these conditions, no colinearity was found.

The second step consisted of taking all the equations for individual genes that had met the acceptance criteria in the first step and combining them into a set of simultaneous equations. Solving this by non-linear least-squares regression required introducing initial values of the parameters, which were provided by the individual regressions performed in the first step. A common period was imposed by replacing in each equation the individual value of τ by a common term. For nine genes, this resulted in a total of 28 parameters to be fitted (common τ, plus individual values of b, c and φ). No weighting was introduced. Initial parameters were adjusted if necessary to reach final convergence, and the tests of validity were then performed as above. If for any gene all parameters did not reach significance, the gene was removed and the calculation repeated with only the remaining genes, until all the criteria of validity had been met.

The resulting common period was accepted only if it fell within the circadian range (20–28 h) and if the corresponding t test was significant. The standard error of the period resulted from the global calculation as for any other parameter. In addition, an ANCOVA was used to test whether a better fit was attained with nine independent regressions than with a common period. The corresponding F test did not reach significance in any of the three cell types. This confirmed the validity of the model with a common period.

Results

Primary cultures of fibroblasts, keratinocytes and melanocytes were established from a single human skin biopsy in order to perform a comparative study of cell-type-specific clock properties by analyzing the expression of ten clock genes (Clock, Bmal1, Per1, Per2, Per3, Cry1, Cry2, RevErbα, Rorα, and Rorβ). Cell cultures were first tested for purity by RT-PCR technique with primers specific for skin cell-type differentiation markers such as epidermal Keratin 5 and Tyrosinase and dermal Collagen type III, alpha 1 (Col3A1). Figure 1a displays representative results showing the purity of NHK, NHM, and NHF. Then, expression of glucocorticoid receptor gene was tested likewise; GR transcript could be amplified in all the studied cell types (Fig. 1b), which indicated that they could be synchronized with dexamethasone, a glucocorticoid hormone analogue. Finally, clock gene expression was analyzed over 52 h starting at 11 h after synchronization with dexamethasone.

Fig. 1.

Evaluation of cell culture purity (a) and glucocorticoid receptor expression (b) by RT-PCR. Agarose gel electrophoresis of the transcript amplification products obtained for the keratinocyte marker Keratin 5 (199 bp), melanocyte marker Tyrosinase (203 bp) and fibroblast marker Collagen type III, alpha1 (Col3A1) (200 bp) and GR (384 bp) in NHK (A1, 4, 7; B1), NHM (A2, 5, 8; B2), and NHF (A3, 6, 9; B3)

Clock gene expression in human epidermal cells

Keratinocytes

Transcripts of all clock genes, except Rorβ, were detected in primary NHK. The analysis of gene expression by non-linear regression showed oscillations with a period in the circadian range for all expressed genes: Clock (21.40 ± 1.64 h, p < 0.0001), Bmal1 (23.57 ± 0.68 h, p < 0.0001), Per1 (23.52 ± 0.59 h, p < 0.0001), Per2 (22.22 ± 0.43 h, p < 0.0001), Per3 (23.05 ± 0.36 h, p < 0.0001), Cry1 (22.90 ± 0.44 h, p < 0.0001), Cry2 (22.51 ± 0.62 h, p < 0.0001), RevErbα (23.88 ± 0.50 h, p < 0.0001), and Rorα (22.83 ± 0.78 h, p < 0.0001).

A second analysis was performed in order to fit the set of individual sine wave equations with a common period. This was done with the idea that all the clock genes contribute to a common clockwork mechanism. We thus aimed at developing a more realistic model, from which reliable phase relationships between genes could be determined. We found that the rhythmically expressed genes, Bmal1, Per1, Per2, Per3, Cry1, Cry2, RevErbα, and Rorα have a common period of 23.06 ± 0.22 h (p < 0.0001) (Fig. 2). Clock gene failed in showing significant parameters. This analysis included also the phase term chosen as time point = 31 h for the study of the specific phase relationships between genes. The phase of Bmal1 gene was used as a reference. By adjusting the common period to 24 h (circadian time), phases of Per genes and Cry2 were opposed approximately 12 h with respect to Bmal1 while the phase of Cry1 was 8.5 h phase advanced and the phases of RevErbα and Rorα were respectively 8 h- and 11 h-delayed (Fig. 5a).

Fig. 2.

Clock gene expression in NHK culture. mRNA amounts measured over a 52-h time interval are expressed relative to the sample showing lowest expression. Data from all time points (n = 3) are presented, together with the superimposed sine wave curve fitted for the calculated common period of 23.06 ± 0.22 h

Fig. 5.

Phase relationships between the clock genes in NHK (a), NHM (b), and NHF (c). Phase differences were calculated by taking Bmal1 as a reference. Angular coordinates are expressed in circadian hours, with a period adjusted to 24 h. Radial coordinates are arbitrary. Error bars: standard errors of the phases, as provided by the fitting algorithm (asymptotic SE, approximated by excess)

Amplitudes were variable, the highest value was observed for Per2 (2.24 ± 0.13, p < 0.0001), followed by RevErbα (1.82 ± 0.13, p < 0.0001), Per1 (1.24 ± 0.13, p < 0.0001), Per3 (1.18 ± 0.13, p = 0.0004), Cry2 (0.89 ± 0.13, p < 0.0001) and lower amplitudes for Bmal1 (0.48 ± 0.13, p < 0.0001), Cry1 (0.57 ± 0.13, p < 0.0001), and Rorα (0.32 ± 0.13, p < 0.02).

Melanocytes

All tested genes were expressed in primary NHM. By non-linear regression analysis we found circadian oscillations for Bmal1 (24.38 ± 0.52 h, p < 0.0001), Per1 (21.17 ± 1.67 h, p < 0.0001), Per2 (26.67 ± 0.87 h, p < 0.0001), Cry1 (26.46 ± 1.09 h, p < 0.0001), Cry2 (25.40 ± 1.49 h, p < 0.0001), and RevErbα (20.20 ± 0.56 h, p < 0.0001). Rorβ showed faster oscillations (19.71 ± 1.08 h, p < 0.0001) and was dropped from further analysis.

The second analysis showed a common period of 25.22 ± 0.48 h (p < 0.0001) for Bmal1, Per1, Per2, Cry1, Cry2, and RevErbα (Fig. 3). Phase analysis by adjusting the period to 24 circadian hours showed a phase advance of approximately 7–8 h for Per1, Per2, and Cry2 and 3 h for Cry1 as compared to Bmal1 and a phase delay of approximately 8 h for RevErbα (Fig. 5b).

Fig. 3.

Clock gene expression in NHM culture. mRNA amounts measured over a 52-h interval are expressed relative to the sample showing lowest expression. Data from all time points (n = 3) are presented, together with the superimposed sine wave curve fitted for the calculated common period of 25.22 ± 0.48 h

The highest amplitude was calculated for Per2 (0.49 ± 0.04, p < 0.0001). Bmal1 and Cry1 oscillations showed medium amplitudes of 0.27 ± 0.04 (p < 0.0001) and 0.21 ± 0.05 (p < 0.0001) respectively, while oscillations for Per1 (0.11 ± 0.04, p = 0.01), Cry2 (0.14 ± 0.04, p = 0.0008), and RevErbα (0.12 ± 0.05, p = 0.0074) showed low amplitudes.

Clock gene expression in dermal human fibroblasts

All tested genes were expressed in primary NHF. Analysis of gene expression by non-linear regression was performed on the log of the relative mRNA amounts to correct for the non-Gaussian distribution of raw data. We found circadian oscillations for Bmal1 (21.79 ± 0.90 h, p < 0.0001), Per1 (22.72 ± 0.99 h, p < 0.0001), Per2 (21.26 ± 0.56 h, p < 0.0001), Per3 (28.00 ± 0.98 h, p < 0.0001), Cry1 (20.17 ± 0.66 h, p < 0.0001), and Cry2 (24.10 ± 1.14 h, p < 0.0001). For RevErbα, the period of oscillations was inferior to 20 h (19.64 ± 0.52 h, p < 0.0001) and therefore the gene was dropped from further analysis.

The second analysis showed a common period of 22.24 ± 0.39 h (p < 0.0001) for Bmal1, Per1, Per2, Per3, Cry1, and Cry2 (Fig. 4). Phase analysis showed an 11-h delay of Per1 transcripts as compared to Bmal1, while Per2, Per3 and Cry2 transcripts showed an approximately 10-h phase advance and Cry1 a 6.8-h phase advance (Fig. 5c). Amplitudes were higher for Per2 (0.49 ± 0.07, p < 0.0001), Per3 (0.43 ± 0.07, p < 0.0001), and Bmal1 (0.42 ± 0.07, p < 0.0001) and lower for Cry1 (0.29 ± 0.07, p < 0.0001) Cry2 (0.29 ± 0.07, p = 0.03), and Per1 (0.23 ± 0.07, p = 0.0009).

Fig. 4.

Clock gene expression in NHF culture. mRNA amounts measured over a 52-h interval are expressed relative to the sample showing lowest expression. Data from all time points (n = 3) are presented, together with the superimposed sine wave curve fitted for the calculated common period of 22.24 ± 0.39 h

Discussion

The aim of our study was to investigate the presence of circadian oscillators in the human skin. Starting with a human skin biopsy, we established primary cultures of keratinocytes, melanocytes, and fibroblasts and analyzed rhythmic expression of clock gene transcripts following synchronization by dexamethasone. Bioluminescence recordings are well suited for the precise determination of period, amplitude, and phase of circadian oscillations [11, 21] but give no cues about phase relationships between different genes. In order to assemble all information concerning period and phase relationships, we: (1) analyzed gene expression kinetics over a 52-h interval for all the clock genes; and (2) developed a special non-linear least squares regression method to account for the fact that all the genes involved in the same clockwork should have the same period. Sine wave functions were simultaneously fitted to the data for all genes with a single common period, while other parameters (phase, amplitude, drift) were specific to each gene. This permitted exact phase relationships between genes to be determined. Taken together, our results show that NHK, NHM, and NHF derived from a common skin sample express the canonical molecular clock components with cell-type-specific oscillating periods. The analysis also revealed opposite phases between Bmal1 and Per and Cry transcripts, as described previously for central and peripheral oscillators ([22], [23], [24], [25], [26]).

Rhythmic clock gene expression in human skin cells

Skin recently became a target tissue for the study of circadian rhythms. Clock and Per1 gene expression was identified in cultured human skin cells [9] together with Bmal1 expression in human skin biopsies [10]. To analyze circadian rhythms in specific skin cell types, we first synchronized the primary cell cultures with dexamethasone, a glucocorticoid hormone analogue shown to be able to synchronize non-neuronal cells [27] and whose receptor is ubiquitously expressed in peripheral tissues, including NHK, NHM, and NHF primary cultures (Fig. 1b).

Keratinocytes

Rhythmic activity of Bmal1 promoter was reported in hair root keratinocytes infected with a lentiviral luciferase expression vector [11]. Using a similar strategy, a circadian clock was identified recently in immortalized keratinocytes (HaCaT cell line) [12] and the expression of clock gene transcripts were described. Moreover, temperature cycles enhanced rhythmic expression of clock genes, an effect observed also in primary neonatal keratinocytes on Per2, Per3, and RevErbα transcripts. Our study describes the molecular components of the circadian machinery in human primary keratinocytes. All tested clock genes except Rorβ were found to be expressed. Mathematical analysis showed that clock gene transcripts, except Clock, had a common circadian period of 23.1 h, which is in agreement with the 23.3-h period reported for the HaCaT clock [12]. Additionally, our data allowed analyzing phase relationships between genes. Bmal1 transcripts in NHK oscillated in anti-phase with Per and Cry transcripts, indicating a functional clockwork, conforming to the current model of circadian oscillators [2].

Melanocytes

Only the transcripts and translation products of Clock and Per1 genes have been investigated in melanocytes and melanoma cell lines [9]. To our knowledge, this work is the first to investigate the presence of a functional oscillator in NHM. We report expression of all canonical clock genes and rhythmic expression of Bmal1, Per1, Per2, Cry1, Cry2, and RevErbα with a common period of 25.2 h. In addition, Per1 and Per2 transcript profiles showed an approximately 8-h phase advance relative to Bmal1. Thus, our results indicate the presence of a functional oscillator in NHM. Clock gene transcription kinetics, period and phase relationships found in MHN suggest distinct, likely cell-specific, regulatory mechanisms occurring in these cells.

Fibroblasts

Fibroblasts were the first non-neuronal cells described to show rhythmic clock gene expression in vitro [27–29]. Fibroblasts as oscillators have been intensively studied in recent years and became a conventional cell model for the study of circadian rhythms in rodents [30] and humans [11, 31–34]. Our study brings new insight into the molecular clock mechanisms in human primary fibroblasts. We show that all clock genes involved in the main transcriptional/translational feed-back loop (Clock, Bmal1, Per1, Per2, Per3, Cry1, and Cry2) and in the regulatory loop (RevErbα, Rorα, and Rorβ) are expressed in NHF. Transcripts of the main loop oscillate with a circadian period of 22.2 h (in agreement with [35] in a similar in vitro study) and have a phase relationship in accordance with that described for known central and peripheral oscillators.

Cell-specific autonomous oscillators: periods, amplitudes, and phase relationships

By dissecting the molecular clock machinery in the human skin we showed that different cell types originating from the same biopsy contain specific oscillators. It was already shown that multilayered tissues, such as mammalian retina, contain several cell-specific oscillators [36–39]. In the present study, fitting a common period to sine wave equations was successful for at least two-thirds of the expressed clock genes in all skin cell types. We observed that the calculated periods were in the circadian range and specific to the different cell types. NHF showed faster oscillations than NHK while NHM showed slower oscillations. The common periods in the different cell types were remarkably accurate, with a very low dispersion (less than half an hour), indicating a good prediction of the periods by using our mathematical model. There were also five rhythmically expressed clock genes (Bmal1, Per1, Per2, Cry1, and Cry2) found to be common to the three cell types. These genes are involved in the main translational/transcriptional feedback loop at the core of the classical molecular clockwork [2], indicating the presence of a functional oscillator to control cell physiology. Supporting evidence also came from the phase relationships between transcript oscillations. The anti-phase expression observed in NHK and NHM for Bmal1 and Per1-2 on the one hand and Bmal1 and Cry2 transcripts on the other, is a common feature of master clock and peripheral oscillators. It was also described in human skin biopsies [10] for Bmal1 and Per1 transcripts. We observed a slightly different phase relationship between these genes in NHM, with 7–8-h phase advance for Per1, 2 and Cry2 as compared to Bmal1, suggesting a melanocyte-specific clock mechanism. Likewise, cell-specific phase relationships were described in healthy human peripheral blood mononuclear cells, albeit with high inter-individual variability, where the phase difference between Per2 and Bmal1 did not exceed 4 h [40]. Although we cannot exclude that differences in phase relationships observed in our study may also reflect a donor particularity, they clearly show that cell types with a specific clock molecular composition co-exist within the same tissue. Another common feature of all studied cell types is the arrhythmic expression of Clock gene, as previously described in the SCN [41]. However, rhythmic expression of Clock was reported in the liver [42].

Probably the most important question raised by our results concerns the differences in period observed between cell types originally present within the same skin sample. Although all of them are in the circadian range, we cannot exclude different cell-specific regulatory factors that are involved in the circadian clockwork. We used dexamethasone to synchronize the cells because it has been described as very efficient for primary fibroblasts [11, 27] and HaCaT cells [12], although it is not known if it acts on melanocytes in a similar manner. We observed different amplitudes of oscillation in the three cell types, the strongest being in NHK followed by NHF and NHM. We do not know the physiological significance of these differences, but they suggest the existence of cell specificity in responses to synchronizing stimuli. Further understanding of the underlying mechanisms will require better knowledge of the circadian functions of these cell types.

We can conclude that skin reveals a complex circadian organization, with the presence of keratinocyte and melanocyte autonomous oscillators at the level of epidermis and of fibroblast oscillators at the level of dermis, that function autonomously for driving rhythmic skin functions. It is possible that such a multi-oscillatory system is particularly suited to fine tuning skin physiology as a response to varying signals coming from both the environment and the body, and ultimately assist skin to play its role as a barrier. The potential internal and external synchronizers that act on each cell type remain to be identified, as do the mechanisms regulating synchronization between the oscillators. Most intriguingly, the role of these clocks in skin physiology remains to be determined.