TO THE EDITOR
Sebaceous glands (SGs) are essential skin appendages which synthesize and secrete sebum, a lipid-rich substance that supports the skin’s antimicrobial defenses and promotes hydration (Zouboulis 2004). Aberrant SG activity has been implicated in dermatological disorders, including acne vulgaris and atopic dermatitis. Factors demonstrated to regulate SG activity include androgens, retinoids and the skin immune environment (Choa et al. 2021; Kobayashi et al. 2019), though their mechanisms are incompletely understood.
Because SGs constitute a distinct minority of skin tissue, studying molecular mechanisms of SG regulation in vivo is difficult. In whole skin bulk RNA sequencing (RNAseq) approaches, SG transcriptional changes may be masked by more abundant cell types. Single-cell RNAseq (scRNAseq) approaches may be less sensitive for detection of differentially expressed genes (DEGs), (Chung et al. 2022) especially in low abundance cell types (Joost et al. 2016; Takahashi et al. 2020). While immortalized human sebocyte cell lines have clear utility (Thiboutot et al. 2003), they lack the complex multicellular cutaneous environment and its cues. Here, we applied laser-capture microdissection (LCM) followed by RNAseq (LCM-RNAseq) to define SG gene transcriptional profiles in vivo with high resolution (Supplementary Figure S1). We analyzed mouse and human SG transcriptomes and compared SG gene expression to whole skin bulk RNAseq and scRNAseq datasets.
Gene expression in murine and human LCM-derived SGs was first compared to whole mouse and human skin (Meisel et al. 2018; Swindell et al. 2017). By principal component analysis, the two tissue types segregated into distinct clusters (Figure 1a, g). We observed 3491 genes enriched in mouse SGs and 3315 genes enriched in human SGs (Figure 1b, c & h, i). The DGAT, AWAT, and ELOVL gene families were enriched in human and murine SGs (Figure 1d, j), and are critical for lipid metabolism and sebum generation (Holmes 2010). Gene ontology (GO) analysis of the SG enriched genes and gene set enrichment analysis (GSEA) revealed SG-enriched pathways related to lipid, steroid, and retinol metabolism in both mouse and human datasets (Figure 1e, f & k, l), major biological processes known to occur in SGs (Makrantonaki et al. 2011). Genes, GO terms and GSEA pathways enriched in whole skin related to skin development and structure (Supplementary Figure S3). When comparing results between species, >800 SG-enriched homologous genes were shared (Figure 1m, Supplementary Table S3). GO analysis of shared and species-specific genes revealed that of the top 100 GO biological processes, ~50% were related to lipid or steroid metabolism in the shared set (Figure 1n). This suggests that many cross-species similarities in SGs are related to processing lipids and hormones.
Figure 1: LCM allows for SG gene enrichment in mouse and human tissue.
Analysis of LCM-derived SG compared to whole skin transcriptome of murine (a-f) and human (g-l) tissue. (a, g) PCA depicting PC1 and PC2 of RNA sequencing results comparing SGs (n = 3 in mouse, n = 2 in human), to whole skin (n = 7 in mouse, n = 4 in human) in mouse (a) and human (g). (b, h) Volcano plot representing genes enriched in mouse (b) or human (h) SGs (3491 in mouse (yellow), 3315 in human(orange)), log2-fold change > 1, Benjamini-Hochberg adj-p-val > 0.05. (c, i) Heatmap of expression levels of differentially expressed genes (DEGs) in mouse (c) and human (i). (d, j) Table of selected DEGs enriched in mouse (d) and human (j) SGs. (e, k) Selected GO terms enriched in mouse (e) and human (k) SGs identified from the GO knowledgebase with FDR adj-p-val > 0.05, and number of genes in the dataset within each term listed beside the bar. (f, l) GSEA enrichment plots showing pathways overrepresented in mouse (f) and human (l) SGs, identified via Reactome and KEGG databases. Genes shown in ranked order according to running enrichment scores. (m, n) Comparison of SG enriched genes from mouse and human (m) with lipid associated GO pathways from shared and species-specific lists quantified (n). LCM, laser capture microdissection; SG, sebaceous gland; PCA, principal component analysis; GO, gene ontology; FDR, false discovery rate; GSEA, gene set enrichment analysis.
To compare LCM-RNAseq to scRNAseq, we selected publicly available scRNAseq data of mouse (Joost et al. 2016) and human skin (Takahashi et al. 2020) (Figure 2a). First, we compared SG gene enrichment in the LCM-RNAseq dataset to SG clusters in the single cell dataset. LCM-RNAseq yielded >10- or >170-fold greater enrichment for SG genes than scRNAseq in murine and human skin, respectively (Figure 2b, e). To examine enrichment of functional pathways, we compared GO biological processes enriched in LCM-RNAseq-specific, scRNAseq-specific, and shared gene lists. Over 50% of biological process GO terms were detected only by LCM-RNAseq (Figure 2c, f). For example, in mice, “autophagy” related GO terms appeared 14 times in the LCM-RNAseq specific set, but were absent in the shared or single cell sets (Figure 2c). Autophagy has been shown to contribute to murine SG morphology and function (Rossiter et al. 2018). To further illustrate this point, in human SG, only 2 “lipid metabolism” related GO terms were found in scRNAseq data compared to 53 in LCM-RNAseq data (Figure 2f).
Figure 2: LCM-derived SG transcriptome has increased sensitivity for SG-specific and overall gene detection compared to sebaceous gland single cell sequencing.
(a) Workflow for comparison of the LCM and single cell methods of SG sequencing in mouse (b-d) and human (e-g). (b, e) Number of SG enriched genes detected by LCM or scRNAseq, defined by log2-fold change > 1 and adj-p-value < 0.05 compared to whole skin (LCM) or remaining clusters (scRNAseq) in mouse (b) or human (e). (c, f) Shared and method-specific GO terms enriched from SG genes in mouse (c) and human (f) with important biological processes highlighted. (d, g) Relative expression compared to reference gene (Scd1 in mice, DCD or FADS2 in humans) of known SG genes or keratinocyte/dermis genes using LCM or scRNAseq in mice (d) and human (g). NA: gene not present in dataset post-processing. (h) Relative expression (compared to above reference genes) of autophagy genes discovered in the LCM SG dataset in mouse and human compared to relative expression from scRNAseq. (i) Representative immunofluorescence images of TEX264 (green) counterstained with Cytokeratin 14 (red) and DAPI (blue) from mouse or human skin sections. Scale bars: 100 μm in mouse tissue, 200 μm in human tissue. Mean fluorescence intensity of TEX264 was quantified and normalized to epidermis to determine fold-change. LCM, laser capture microdissection; SG, sebaceous gland, scRNAseq, single cell RNAseq; GO, gene ontology. **p < 0.01, *****p < 0.00001 by Student t test (mouse). Data are shown as mean ± SD. qPCR experiment performed twice.
To compare expression levels between techniques, we normalized average expression of all genes to a scRNAseq cluster-defining reference gene. We find variable results between the two methods, though LCM overall yielded higher expression levels of lipid metabolic genes such as ELOVL4 and DGAT2 (Figure 2d, g). A recent scRNAseq study parsed SG cells to high levels and defined SG cluster genes including Pparg, Fasn, and Cidea (Veniaminova et al. 2023), which were expressed at variable levels in these comparative datasets as well (Figure 2d, g). Lastly, expression levels of canonical keratinocyte and dermal/fibroblast genes were negligible in both datasets (Figure 2d, g). Given the higher expression levels of SG genes from LCM, we examined the LCM-RNAseq dataset for previously under-investigated pathways expressed in SGs. Autophagy GO terms were enriched in mouse SGs, and many overlapped with human SGs (Supplementary Figure S4a). Thus we focused on two genes in the autophagy pathway: TEX264 and MAP1LC3B, which are essential in the formation of autophagosomal vacuoles (An et al. 2019). Expression of TEX264 and MAP1LC3B was significantly increased in LCM-RNAseq in both mouse and human (Figure 2h). Immunofluorescence staining of the LCM-RNAseq samples for these autophagy genes showed clear expression in the SGs in both species, with TEX264 having a significant or trending increase in SGs compared to epidermis, and MAP1LC3B with clear expression in both SGs and epidermis tissue in both species (Figure 2i, Supplementary Figure S4b). Additionally, in SEB-1 immortalized human sebocytes, relative expression of TEX264 and MAP1LC3B were similar to or greater than DGAT1, a lipid metabolism gene essential for SG function (Chen et al. 2002) (Supplementary Figure S4c).
Here, we used LCM to isolate SGs from in vivo murine and human skin to characterize the transcriptomic landscape at homeostasis in the complex cutaneous environment. We define murine and human SG enriched gene sets and SG biological processes. We demonstrate altered SG gene expression levels when compared to scRNAseq studies, perhaps due to limited cell number and sequencing depth in these scRNAseq clusters. A limitation of LCM-RNAseq is the inability to separate sebocyte lineages by LCM-RNAseq. Specialized approaches to enrich for sebocytes could be employed together with scRNAseq to circumvent this issue. Additionally, the small sample size of LCM datasets is a limitation, which reflects the labor-intensive nature of the method. As with any rapidly evolving technology, the advantages and disadvantages of each approach should be carefully weighed with respect to the experimental goals.
Supplementary Material
| Sample | Sample origin | Tissue type | Tissue location | Tissue extraction type | Sebaceous glands isolated (#) | Area isolated (um^2) | Cells in sebaceous gland cluster |
|---|---|---|---|---|---|---|---|
| Mouse_SG_1 | Mouse | Sebaceous gland | Back | Laser capture microdissection | 1187 | 2,490,766 | NA |
| Mouse_SG_2 | Mouse | Sebaceous gland | Back | Laser capture microdissection | 974 | 1,936,628 | NA |
| Mouse_SG_3 | Mouse | Sebaceous gland | Back | Laser capture microdissection | 1103 | 2,860,728 | NA |
| Mouse_wholeskin_1 | Mouse | Whole skin | Back | Bulk skin extraction | NA | NA | NA |
| Mouse_wholeskin_2 | Mouse | Whole skin | Back | Bulk skin extraction | NA | NA | NA |
| Mouse_wholeskin_3 | Mouse | Whole skin | Back | Bulk skin extraction | NA | NA | NA |
| Mouse_wholeskin_4 | Mouse | Whole skin | Back | Bulk skin extraction | NA | NA | NA |
| Mouse_wholeskin_5 | Mouse | Whole skin | Back | Bulk skin extraction | NA | NA | NA |
| Mouse_wholeskin_6 | Mouse | Whole skin | Back | Bulk skin extraction | NA | NA | NA |
| Mouse_wholeskin_7 | Mouse | Whole skin | Back | Bulk skin extraction | NA | NA | NA |
| Human_SG_1 | Human | Sebaceous gland | Nose | Laser capture microdissection | 383 | 45,438,467 | NA |
| Human_SG_2 | Human | Sebaceous gland | Nose | Laser capture microdissection | 426 | 29,743,045 | NA |
| Human_wholeskin_1 | Human | Whole skin | Buttock or upper thigh | Bulk skin extraction | NA | NA | NA |
| Human_wholeskin_2 | Human | Whole skin | Buttock or upper thigh | Bulk skin extraction | NA | NA | NA |
| Human_wholeskin_3 | Human | Whole skin | Buttock or upper thigh | Bulk skin extraction | NA | NA | NA |
| Human_wholeskin_4 | Human | Whole skin | Buttock or upper thigh | Bulk skin extraction | NA | NA | NA |
| Mouse_single_cell | Mouse | Epidermis | Back | Epidermis tissue extraction | NA | NA | 16 |
| Human_single_cell | Human | Epidermis | Back | Scalp micrograft extraction | NA | NA | 30 |
HUMAN AND ANIMAL MATERIALS STATEMENT.
Animal materials used in this study were collected with approval by the University of Pennsylvania Institutional Animal Care and Use Committee (IACUC). Human materials used in this study were collected with approval by University of Pennsylvania under IRB protocol 808225 which pertains to the use of de-identified archived tissue specimens which do not meet the criteria for human subjects research. General consent for use of these specimens for educational and research purposes was obtained at the time of biopsy. We isolated sebaceous gland tissue from three skin samples and performed bulk RNA sequencing on this tissue, no identifying information is included in publication and all samples remained de-identified throughout analysis, therefore, the study does not meet the criteria for human subjects research.
ACKNOWLEDGEMENTS
The authors thank Dr. Daniel Beiting, PhD for developing an open-source RNA sequencing analysis toolkit, formed into an instructional course at the University of Pennsylvania. This work was supported by grants to EAG (R01NR015639), the Resource Cores and pilot and feasibility funding (to TK) from the Penn Skin Biology and Disease Research Center (supported by NIAMS P30AR069589). JCH was supported by the Penn Dermatology Research Training Grant (NIAMS T32AR007465) and NRSA Fellowship (NIAMS F31AR079845). Models within figures were created using BioRender.com.
Abbreviations:
- SG
sebaceous gland
- RNAseq
RNAsequencing
- DEG
differentially expressed gene
- LCM
laser capture microdissection
- PCA
principal component analysis
- GO
gene ontology
- GSEA
gene set enrichment analysis
Footnotes
CONFLICT OF INTEREST STATEMENT
The authors state no conflict of interest.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
DATA AVAILABILITY STATEMENT
FASTQ files generated from mouse and human SG RNA isolation and sequencing have been uploaded to the Gene Expression Omnibus (GEO) database for public use under the accession number GSE230682. Files will also be stored in the Skin Genes Query database hosted by the University of Michigan SBDRC, located at https://medicine.umich.edu/dept/sbdrc/core-services/functional-analytics-core.
REFERENCES:
- An H, Ordureau A, Paulo JA, Shoemaker CJ, Denic V, Harper JW. TEX264 is an ER-resident ATG8-interacting protein critical for endoplasmic reticulum remodeling during nutrient stress. Mol. Cell 2019;74(5):891–908.e10 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Chen HC, Smith SJ, Tow B, Elias PM, Farese RV. Leptin modulates the effects of acyl CoA:diacylglycerol acyltransferase deficiency on murine fur and sebaceous glands. J. Clin. Invest 2002;109(2):175–81 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Choa R, Tohyama J, Wada S, Meng H, Hu J, Okumura M, et al. Thymic stromal lymphopoietin induces adipose loss through sebum hypersecretion. Science. American Association for the Advancement of Science; 2021;373(6554) Available from: https://science.sciencemag.org/content/373/6554/eabd2893 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Chung KB, Oh J, Roh WS, Kim T-G, Kim D-Y. Core Gene Signatures of Atopic Dermatitis Using Public RNA-Sequencing Resources: Comparison of Bulk Approach with Single-Cell Approach. J. Invest. Dermatol 2022;142(3 Pt A):717–721.e5 [DOI] [PubMed] [Google Scholar]
- Holmes RS. Comparative genomics and proteomics of vertebrate diacylglycerol acyltransferase (DGAT), acyl CoA wax alcohol acyltransferase (AWAT) and monoacylglycerol acyltransferase (MGAT). Comp. Biochem. Physiol. Part D Genomics Proteomics. 2010;5(1):45–54 [DOI] [PubMed] [Google Scholar]
- Joost S, Zeisel A, Jacob T, Sun X, La Manno G, Lönnerberg P, et al. Single-Cell Transcriptomics Reveals that Differentiation and Spatial Signatures Shape Epidermal and Hair Follicle Heterogeneity. Cell Syst. Elsevier; 2016;3(3):221–237.e9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kobayashi T, Voisin B, Kim DY, Kennedy EA, Jo J-H, Shih H-Y, et al. Homeostatic Control of Sebaceous Glands by Innate Lymphoid Cells Regulates Commensal Bacteria Equilibrium. Cell. 2019;176(5):982–997.e16 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Makrantonaki E, Ganceviciene R, Zouboulis C. An update on the role of the sebaceous gland in the pathogenesis of acne. Dermatoendocrinol. 2011;3(1):41–9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Meisel JS, Sfyroera G, Bartow-McKenney C, Gimblet C, Bugayev J, Horwinski J, et al. Commensal microbiota modulate gene expression in the skin. Microbiome. 2018;6(1):20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Rossiter H, Stübiger G, Gröger M, König U, Gruber F, Sukseree S, et al. Inactivation of autophagy leads to changes in sebaceous gland morphology and function. Exp. Dermatol 2018;27(10):1142–51 [DOI] [PubMed] [Google Scholar]
- Swindell WR, Sarkar MK, Liang Y, Xing X, Baliwag J, Elder JT, et al. RNA-seq identifies a diminished differentiation gene signature in primary monolayer keratinocytes grown from lesional and uninvolved psoriatic skin. Sci. Rep 2017;7:18045. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Takahashi R, Grzenda A, Allison TF, Rawnsley J, Balin SJ, Sabri S, et al. Defining Transcriptional Signatures of Human Hair Follicle Cell States. J. Invest. Dermatol. Elsevier; 2020;140(4):764–773.e4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Thiboutot D, Gilliland K, Cong Z, Jabara S, McAllister JM, Sivarajah A, et al. Human Skin is a Steroidogenic Tissue: Steroidogenic Enzymes and Cofactors Are Expressed in Epidermis, Normal Sebocytes, and an Immortalized Sebocyte Cell Line (SEB-1). J. Invest. Dermatol. Elsevier; 2003;120(6):905–14 [DOI] [PubMed] [Google Scholar]
- Veniaminova NA, Jia Y, Hartigan AM, Huyge TJ, Tsai S-Y, Grachtchouk M, et al. Distinct mechanisms for sebaceous gland self-renewal and regeneration provide durability in response to injury. bioRxiv; 2023. p. 2023.05.05.539454 [DOI] [PMC free article] [PubMed]
- Zouboulis CC. Acne and sebaceous gland function. Clin. Dermatol. 2004;22(5):360–6 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
FASTQ files generated from mouse and human SG RNA isolation and sequencing have been uploaded to the Gene Expression Omnibus (GEO) database for public use under the accession number GSE230682. Files will also be stored in the Skin Genes Query database hosted by the University of Michigan SBDRC, located at https://medicine.umich.edu/dept/sbdrc/core-services/functional-analytics-core.