Skin Transcriptomic Analysis Reveals Altered Fatty Acid Metabolism Pathways in Mice Subjected to Sleep Fragmentation

Sleep is essential for the maintenance of normal body function. Sleep loss or inadequate sleep can result in memory and cognitive impairment, immune dysregulation, metabolic dysfunction, and an increased risk of cancer. As skin health has become increasingly important in the modern world, the relationship between sleep and skin has received considerable attention. Several human and animal studies have consistently shown that inadequate sleep can lead to dysregulated skin functions, including reduced skin hydration and elasticity, increased intrinsic aging, and diminished skin barrier integrity [1‒3]. In humans, sleep quantity and quality both affect skin health. For instance, poor sleepers (those with a Pittsburg Sleep Quality Index >5 and sleep duration ≤5 h) exhibit significantly higher intrinsic-aging scores and greater skin dehydration than those of good sleepers [4]. In addition, an experimental study involving humans revealed that two consecutive nights of sleep restriction (SR; 3 h of sleep per night) resulted in lower hydration, increased transepidermal water loss (TEWL), reduced extensibility and viscosity, and increased skin pH [2]. Furthermore, a mouse model showed that 4 weeks of SR (4 h per day) disrupted skin barrier function by increasing oxidative stress [3]. In rats, total sleep deprivation (SD) resulted in ulcerative and hyperkeratotic lesions, specifically on the feet and tails [5].

The potential mechanisms underlying the dysregulation of skin function due to lack of sleep include disruption of the release of hormones (such as cortisol, melatonin, and growth hormone) and inflammatory cytokines that participate in the circadian rhythm or are elevated during sleep [6]. Moreover, under SD, changes in the skin microbiome may affect its immune defense system and barrier function [7]. Although various mechanisms have been proposed to explain the effects of sleep loss on skin function, the exact molecular mechanisms remain to be clarified. No prior studies have used untargeted approaches to elucidate the mechanisms linking sleep and skin function.

To address this, we aimed to identify the molecular pathways affecting skin function in a mouse model of sleep fragmentation (SF) by leveraging total RNA sequencing (RNA-Seq). Additionally, we investigated the effects of chronic SF on skin physiology and histology. SF, rather than SD, was used for several reasons. First, prolonged SD is rare in the real world. Second, traditional SD models can cause unnecessary stress in animals. Third, several sleep disorders associated with skin diseases are related to arousal during sleep (i.e., SF); these include obstructive sleep apnea and insomnia. However, the causal relationships remain poorly understood.

Twenty-four 6-week-old male BALB/c mice (DBL, Eumseong, South Korea) were randomly divided into two groups: home cage control (HC; n = 12) and SF (n = 12). The experiment was conducted for 8 weeks, and TEWL was measured after 4 and 8 weeks. A schematic of the experimental protocol is shown in Figure 1a. The study protocol was approved by the Animal Research Committee of Jungwon University (JWU-IACUC-2022-5).

SF was induced in a commercially available SF chamber (Model 80391; Lafayette Instruments, Lafayette, IN, USA). Inside the chamber, a sweeping bar moves from one end to the other; when the sweeping bar touches the mouse, the mouse wakes up. We set the sweeping bar to move once every 2 min to mimic the frequency of awakening (30 times per h) of people with severe sleep apnea. Although we could not perform concurrent electroencephalography monitoring to directly confirm SF induction, several studies using the same model with an SF chamber have validated its effectiveness [8, 9]. We routinely observed the behavior of the mice subjected to SF to confirm sleep disruption and maintained strict consistency in procedures throughout the experimental period.

To assess whether SF affects skin barrier function, we measured TEWL, the most direct indicator of skin barrier function in the stratum corneum. A VapoMeter (Delfin Technologies, Kuopio, Finland) was used to measure the TEWL, an indicator of skin dehydration. At 4 and 8 weeks, the mice were shaved, and the TEWL of the dorsal skin was measured 1 h later. The TEWL measurements were taken three times per mouse, and the average of these values was calculated. Because TEWL can vary with external conditions such as temperature and humidity, all measurements were performed under normal conditions (temperature 22 ± 2°C, ambient humidity 40 ± 5%).

At the end of the experiment, the skin was excised from the dorsal and flank areas and washed with cold phosphate-buffered saline (PBS). After removing the vessels, the skin tissue was fixed in 10% neutral-buffered formalin and embedded in paraffin. Thereafter, the tissue was cut into 5-μm thick sections and stained using a hematoxylin and eosin (H&E) staining kit (Abcam, Cambridge, UK). After H&E staining, images of five randomly selected areas (×100 magnification) of skin on each slide were obtained using a microscope (Eclipse Ci; Nikon, Tokyo, Japan) equipped with a digital camera to assess epidermal thickness. The images were imported into ImageJ (National Institutes of Health, Bethesda, MD, USA), and a straight line was drawn vertically from the top to the bottom of the epidermis to measure its thickness.

Masson trichrome (MT) staining was performed to assess collagen content in the dermis. Briefly, paraffin sections were dewaxed in xylene, fixed in Bouin’s solution, washed with water, stained with Weigert’s Iron Hematoxylin for 5–10 min, rinsed with distilled water, stained with Biebrich scarlet/acid fuchsin solution for 5–15 min, and then rinsed with distilled water. Afterward, the sections were treated with a phosphomolybdic/phosphotungstic acid solution for approximately 10–15 min. The solution at the top was decanted, and the sections were directly stained with an aniline blue solution for 5–10 min without washing. Thereafter, the sections were treated with 1% acetic acid for 3–5 min, dehydrated, sealed, and observed under a microscope. After MT staining, collagen density was quantitatively analyzed from five randomly selected low-power field images (×100 magnification) per slide based on a previous study [10]. The selected areas excluded regions with artifacts or uneven staining. Images were captured under identical lighting and magnification settings using a high-resolution microscope (ECLIPSE Ci-L, Nikon, Tokyo, Japan). The blue-stained collagen area in each image was quantified as a percentage of the total tissue area using ImageJ software (version 1.54i, NIH, USA). To ensure reliability, the analysis was conducted independently by two blinded investigators (D.B.L. and S.L.Y.) using consistent threshold parameters for image segmentation. The final collagen density was calculated as the mean percentage from the five images per slide.

Three animals from each group were randomly selected for total RNA-Seq. Total RNA was extracted using TRIzol reagent (Invitrogen, Carlsbad, CA, USA). RNA integrity was evaluated using a TapeStation 4000 System (Agilent Technologies, Santa Clara, CA, USA). RNA quantity was determined using an ND-2000 Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA).

Libraries were generated from total RNA using a NEBNext Ultra II Directional RNA-Seq Kit (New England BioLabs, Ipswich, MA, USA). After rRNA removal using the RIBO COP rRNA Depletion Kit (Lexogen, Vienna, Austria), rRNA-depleted RNA was subjected to cDNA synthesis and shearing, following the manufacturer’s instructions. Indexing was performed using Illumina index sequences 1–12 (Illumina, San Diego, CA, USA), and PCR was conducted during the sequence enrichment process. Library quality was assessed using an Agilent 2100 Bioanalyzer (using a DNA High Sensitivity Kit; Agilent Technology, Santa Clara, CA, USA) to determine the mean fragment size. Next, the library was quantified using a Library Quantification Kit and Step One Real-Time PCR System (Life Technologies, Carlsbad, CA, USA). Finally, high-throughput sequencing was performed using paired-end 100 bp sequences on a NovaSeq 6000 Sequencing System (Illumina).

Quality control of the raw sequencing data was conducted using FastQC. Adapters and low-quality reads (<Q20) were removed using FASTXTrimmer and BBMap. Subsequently, the trimmed reads were mapped to the reference genome using TopHat. The expression of genes, isoforms, and total RNA was estimated as fragments per kilobase per million read (FPKM) values using Cufflinks. FPKM values were normalized based on the TMM + CPM method using EdgeR in R (R Foundation for Statistical Computing, Vienna, Austria). Data mining and graphical visualization were performed using Excel-based Differentially Expressed Gene Analysis (ExDEGA v. 5.0.1; E-biogen, Seoul, South Korea).

Differential gene expression analysis was conducted using ExDEGA, with a cutoff at a normalized gene expression (log₂) of 3 and a p < 0.05. Identification of differentially expressed genes (DEGs) was based on a >1.5-fold change in transcript levels. ExDEGA was used for data mining and graphic visualization. DEG expression was visualized using a heatmap generated in the open-source software MeV v. 4.9.0 (https://sourceforge.net/projects/mev-tm4/; accessed on October 31, 2023). Distinct colors representing the log₂ ratio-adjusted z-scores were utilized to illustrate gene expression for each sample. Hierarchical clustering of genes was performed, and phylogenetic trees were generated using Euclidean distances. Following DEG filtering, Gene Ontology (GO) annotation analysis was performed using the DAVID bioinformatics platform (https://david.ncifcrf.gov; accessed on September 28, 2023). RNA-Seq data were further analyzed using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (www.genome.jp, accessed on January 2, 2024). STRING, an online search tool for identifying interacting genes (http://string-db.org), was used to identify interactions among the shared DEGs. A confidence score ≥0.4 was considered statistically significant. Cytoscape (http://www.cytoscape.org/) was used to construct and visualize the protein-protein interaction (PPI) network involving the shared DEGs.

Fluorescence staining for reactive oxygen species (ROS) was performed on mouse skin tissue as described previously [11]. Briefly, excised skin samples were immediately placed on dry ice and embedded in an O.C.T. compound (Tissue-Tek; Sakura Finetek, Torrance, CA, USA). Thereafter, 30-µm sections were obtained using a cryostat microtome (CM1860; Leica Biosystems, Nuβloch, Germany). The sections were air-dried at room temperature for 10 min and then incubated with 10 μm dihydroethidium (Sigma-Aldrich, St. Louis, MO, USA) for 30 min at 37°C. After washing in PBS, the sections were incubated with 1 μg/mL Hoechst 33342 (Thermo Fisher Scientific) for 10 min at room temperature, then mounted in Fluoroshield (Sigma-Aldrich). Subsequently, the red fluorescence channel of the image was obtained by excitation at 610 nm; absorbance was measured at 600–780 nm using a confocal microscope (LSM 700; Carl Zeiss, Oberkochen, Germany).

Total RNA was extracted from skin tissue using a Tissue RNA Extraction Kit (Cat# NAE1008, microMag^®; Next&Bio, Seoul, South Korea). For cDNA synthesis, 1 μg of total RNA was used, followed by reverse transcription using a Tetro cDNA Synthesis Kit (Bioline, London, UK). Quantitative real-time PCR (qPCR) was performed using a PowerUp SYBR™ Green Master Mix (Applied Biosystems, Waltham, MA, USA). All reactions were performed in duplicate, and specificity was examined via melting-curve analysis. The following target-specific primers were purchased from Bioneer (Daejeon, South Korea): NDUFS5, F: 5′-GGG​ACC​CGG​GCG​AAA-3′ and R: 5′-CAT​TCG​CCT​CAT​CGT​TTT​GTA​C-3′; COX6C, F: 5′-C′ACA​GAT​GCG​TGG​TCT​TCT-3′ and R: 5′-GAA​AGA​TAC​CAG​CCT​TCC​TC-3′; HADHB, F: 5′-CAG​CGC​CTG​TCC​TTA​CTC​AG-3′ and R: 5′-CAG​AGT​GGC​CCA​TGG​TCT​C-3′; UCP3, F: 5′-TTT​CTG​CGT​CTG​GGA​GCT​T-3′ and R: 5′-GGC​CCT​CTT​CAG​TTG​CTC​AT-3′; CPT1B, F: 5′-GCA​CTT​CTC​AGC​ATG​GTC​ATC​T-3′ and R: 5′-GGG​TTT​GTC​GGA​AGA​AGA​AAA​TG-3′. The 2^ΔΔ method was used to compare gene expression between the HC and SF groups. β-actin was used as the housekeeping gene for normalization.

The data are expressed as the mean ± standard error of the mean. Differences in the means were evaluated using an independent sample t test. All statistical analyses were performed using GraphPad Prism 9.0 (GraphPad Software, San Diego, CA, USA). Differences were considered significant at p < 0.05.

We examined the effects of SF on TEWL at weeks 4 and 8 of the experiment. TEWL did not differ between the HC and SF groups at 4 weeks but was significantly higher in the SF group at 8 weeks (Fig. 1b, c).

Epidermal thickness contributes to the structural integrity of the skin barrier, which indirectly affects skin hydration by regulating TEWL [12]. Therefore,` we assessed whether SF affects epidermal thickness. At 8 weeks, the SF group exhibited significantly lower epidermal thickness than that in the control (Fig. 2a, b). As dermal collagen levels affect skin integrity, we evaluated differences in collagen levels using MT staining and found that collagen density was significantly lower in the SF group than in the control (Fig. 2c, d).

Differential gene expression is illustrated in Figure 3a. In the principal component analysis plot, principal component 1 (PC1) explains 38% of the variance and PC2 24%, and the HC and SF groups are distinctly separated (Fig. 3b). Total RNA-Seq analysis identified 133 DEGs, 14 upregulated and 119 downregulated, in the SF group. The heatmap with hierarchical clustering reveals two clusters of samples with distinct gene expression patterns (Fig. 3c).

To comprehensively understand the biological functions associated with the shared DEGs, GO and KEGG pathway analyses were performed using DAVID. GO analysis of biological processes revealed that the DEGs were most significantly enriched in the following processes: fatty acid metabolism (FAM), cellular response to hypoxia, response to hypoxia, negative regulation of ROS metabolic process, protein ubiquitination, positive regulation of macroautophagy, cellular response to sterol, regulation of the neuronal apoptotic process, glycogen metabolism, and regulation of vascular endothelial growth factor production (Fig. 3d). Based on KEGG analysis, the most significantly enriched pathways were those associated with glucagon signaling, AMP-activated protein kinase signaling, thermogenesis, insulin signaling, non-alcoholic fatty liver disease, carbon metabolism, fatty acid degradation, FAM, as well as the pentose phosphate pathway and metabolic pathways (Fig. 3e).

We constructed a PPI network to further examine the interactions among the shared DEGs using the STRING database and Cytoscape. The PPI network, which identifies interactions among the proteins associated with SF, includes 35 nodes and 50 edges from the 133 DEGs (Fig. 4a). Among these, 5 were upregulated and 29 downregulated (Table 1). Interactions within the PPI network were examined via functional enrichment analysis, which revealed the involvement of fatty acid β-oxidation, response to hypoxia, FAM, and cellular lipid metabolism.

To explore the interactions in FAM processes, we constructed a PPI network using the STRING database and Cytoscape. This analysis identified 10 nodes and 20 edges in the network (Fig. 4b). Of the 10 nodes, 1 DEG was upregulated and 9 were downregulated. The top 10 hub proteins involved in FAM were identified via GO and the KEGG analysis. These included NADH dehydrogenase (ubiquinone) Fe-S protein 5 (NDUFS5), ankyrin repeat domain 23 (ANKRD23), hydroxyacyl-CoA dehydrogenase trifunctional multienzyme complex subunit beta (HADHB), acetyl-CoA carboxylase alpha (ACACA), enoyl-CoA delta isomerase 1 (ECI1), mitogen-activated protein kinase 12 (MAPK12), uncoupling protein 3 (UCP3), protein kinase AMP-activated catalytic subunit alpha 2 (PRKAA2), cytochrome c oxidase subunit VI C (COX6C), and carnitine palmitoyltransferase 1B (CPT1B). These findings suggest that SF may alter FAM.

Two main pathways (FAM and the response to hypoxia) were altered in the SF group. To confirm these results, we directly measured the ROS levels in the skin. Confocal microscopy image analysis revealed no significant differences in ROS fluorescence intensity between the groups (Fig. 5a, b). qPCR was used to assess the expression of FAM-associated genes. The expression of five FAM-associated genes was significantly lower in the SF group than in the control, confirming that FAM was altered in the SF group (Fig. 5c).

We aimed to identify the molecular pathways affecting skin function in a mouse model of SF by examining the effects of chronic SF on skin physiology and histology. Our results showed that FAM was altered in the skin of SF-treated mice, and chronic SF increased TEWL and reduced epidermal thickness and collagen content.

FAM plays a crucial role in maintaining the structural integrity and functional capacity of the skin barrier. This applies to both mouse models and humans as fatty acids contribute to the formation of key lipid components, including ceramides, free fatty acids, phospholipids, and triglycerides, which are essential for the skin barrier function [13, 14]. These lipids are essential for forming the stratum corneum’s lamellar structure, which prevents TEWL and protects against external environmental stressors [14]. The changes in FAM-related genes observed here may significantly alter the lipid composition of the stratum corneum, potentially disrupting its structure and increasing skin permeability and TEWL. An increase in TEWL, a key indicator of impaired barrier function, may render the skin more susceptible to inflammation and infection. These findings are also relevant in terms of human skin physiology. Similar to that in mice, the stratum corneum in humans relies on a delicate balance of lipids for optimal barrier function. Disruptions in FAM can lead to ceramide deficiency, which has been linked to increased permeability and inflammation, as observed in various skin disorders [15].

The inability of the skin to synthesize long-chain fatty acid metabolites, owing to the lack of an enzymatic system that adds double bonds to fatty acid chains [16, 17], may further exacerbate the changes in FAM in SF-treated mice. This may have contributed to the observed barrier dysfunction, skin dehydration, and histological changes in this group. Further research is required to clarify how these SF-induced changes in the FAM pathway are linked to these inherent metabolic limitations.

Experimental studies in both humans and animals have revealed that a lack of sleep results in dysfunction of the skin barrier and mucous membranes. For example, in healthy humans, total SD for 42 h inhibits the recovery of skin barrier function and generates elevated levels of pro-inflammatory cytokines [18]. In rats, total or paradoxical SD causes ulcerative skin lesions on the legs and tail [5]. Although elevated glucocorticoid levels, caused by dysregulation of the hypothalamic-pituitary-adrenal axis, may mediate the link between sleep loss and skin damage [19], we did not examine this possibility here. Nevertheless, a previous study using the same SF mouse model reported no significant increase in glucocorticoid levels in mice subjected to SF [9].

In our study, epidermal thickness was significantly reduced at 8 weeks of SF. A thinner epidermis and reduced amounts of collagen fibers are features of chronological skin aging [20, 21], suggesting that SF may exacerbate skin aging. However, this finding is contrary to that of another study that reported increased epidermal thickness with SR [3]. This discrepancy may be due to differences in the model system (SF vs. SR), duration of exposure (8 weeks here vs. 4 weeks in the SR study), and the mouse strain used (BALB/c here vs. ICR in the SR study). Other skin-aging markers should be evaluated to determine whether SF induces aging-like changes in the skin.

We detected no evidence of increased oxidative stress, as indicated by the similar ROS levels in the HC and SF groups, despite the hypoxia response pathways being altered in the SF group. This discrepancy may stem from the differences in sensitivity and resolution between molecular analyses (e.g., gene expression profiling) and histological techniques. Gene expression data reflect transcriptional activity and may reflect potential changes in ROS production or FAM that are not yet apparent at the macroscopic level of ROS staining. Furthermore, differences in ROS-related gene expression might reflect compensatory mechanisms activated in the SF group to mitigate oxidative damage, resulting in normalized ROS levels in tissues despite altered gene activity. However, the relationship between sleep loss and skin oxidative stress remains controversial. For instance, in healthy women, 2 days of SR significantly increased malondialdehyde (MDA) levels in forehead skin [2], while 4 weeks of SR increased MDA levels and reduced total antioxidant capacity in 8-week-old ICR mice [3]. However, in aged hairless mice, 72 h of paradoxical SD and 15 days of SR did not induce significant oxidative DNA damage in skin tissue [22].

SF frequently occurs in patients with sleep disorders such as obstructive sleep apnea, insomnia, and restless legs syndrome. In obstructive sleep apnea, sympathetic activation may trigger an imbalance in the immune neuroendocrine network, precipitating inflammatory skin disorders [23, 24]. SF may, therefore, play a causal role in the development of skin disorders. In some instances, the association between sleep disturbances and skin diseases seems to be bidirectional. For example, adult patients with atopic dermatitis are more likely to develop insomnia [25]. For children with the condition, up to 60% experience sleep disturbances, with the frequency further increasing with the exacerbation of atopic dermatitis [26]. The mechanisms underlying the bidirectional relationship involve multiple factors, including cytokine and melatonin dysregulation, chronic stress, the circadian rhythm of the skin, and acquired sleep habits [27].

Oxidative stress and inflammation are known to be interconnected, with oxidative stress frequently initiating inflammatory pathways. However, interactions between these two processes can be complex. In our RNA-Seq data, the diminished response to oxidative stress in the SF group may indicate an adaptive or compensatory mechanism. This suggests that, while oxidative stress pathways were downregulated under SF, inflammatory gene expression could have stabilized at baseline levels, with no significant changes under SF.

Additionally, the chronic nature of SF could have resulted in long-term adaptation in the skin, where inflammatory signaling might have been maintained at a constant level. This would explain the lack of a marked increase in inflammatory gene expression despite the reduced oxidative stress response. Another possible explanation could involve post-transcriptional or regulatory processes affecting protein expression or activity, which would not be fully captured by RNA-Seq analysis alone. These adaptation mechanisms may primarily affect specific pathways, such as inflammation or oxidative stress, without entirely preventing structural changes. The significant increase in TEWL observed in week 8 likely reflects cumulative structural and functional changes in the skin barrier caused by prolonged SF, progressively impairing the barrier’s ability to retain water and resulting in the observed TEWL increase.

To address the possibility that the observed differences in TEWL and skin changes might result from factors other than SF, such as behavioral responses leading to skin injury, we implemented several measures to minimize such confounding effects. First, during the experimental period, the mice in the SF group were monitored closely to ensure that behavioral factors, such as excessive scratching or self-inflicted injuries, did not compromise skin integrity. Visual inspection of all mice revealed no signs of skin abrasions, wounds, or localized inflammation, which would indicate behavioral impacts as a cause of TEWL differences. Second, the environmental conditions, including temperature, humidity, and light cycles, were carefully standardized across all experimental groups. Furthermore, all handling procedures were identical to ensure that no additional stressors could disproportionately affect the SF group. These controls helped eliminate the possibility of environmental or external influences on the observed skin barrier changes.

In the current study, we used haired BALB/c mice to investigate the effects of SF on the skin because the presence of hair in BALB/c mice allowed us to evaluate the impact of SF under conditions that more closely resemble natural physiological states. Hairless mice, on the other hand, may exhibit altered skin properties that could introduce confounding factors, such as changes in barrier function or hydration dynamics.

This study has several limitations. First, our analysis was focused on specific molecular pathways and hence may not fully capture the broad range of mechanisms involved. Second, although we employed a widely recognized animal model, the results may not be entirely generalizable to humans owing to inherent differences in skin physiology. Third, as we did not examine hormone levels or the skin microbiome, we could not examine how these factors influence skin response to SF. Finally, the relatively short duration of the experiment may not fully reflect the longer term effects of chronic SF on skin health. Addressing these limitations would help elucidate the effects of chronic SF on the skin. In conclusion, we observed that chronic SF altered FAM pathways in a mouse model, resulting in physiological and histological changes in the skin. Future studies should aim to elucidate the role of FAM pathways in SF-induced skin damage.

The study protocol was approved by the Animal Research Committee of Jungwon University (JWU-IACUC-2022-5).

The authors have no conflicts of interest to declare.

This research was supported by the National Research Foundation of Korea Grant funded by the Korean Government (NRF-2021R1I1A3060351) and Regional Innovation Strategy (RIS) through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (MOE) (2021RIS-001(1345370811).

D.-B.L., S.-W.K., and D.-W.Y.: study conception and design; D.-B.L., S.-L.Y., J.S.H., J.-Y.L., S.S.P., and J.S.M.: acquisition of data; D.-B.L., M.-R.L., J.K., and D.-W.Y.: analysis and interpretation of data; D.-W.Y.: funding acquisition; D.B.L., D.-W.Y., D.-B.L., S.-W.K., and D.-W.Y.: writing – original draft; and J.S.H., J.-Y.L., S.S.P., J.S.M., J.K., and M.-R.L.: critical review. All of the authors have approved the final version of the manuscript.

All data generated or analyzed during this study are included in this article. Further inquiries can be directed to the corresponding authors.