Human Epidermal Stem Cell Differentiation Is Modulated by Specific Lipid Subspecies

Human epidermal stem cell differentiation is modulated by specific lipid subspecies

Matteo Vietri Rudana, Ajay Mishraa,b, Christian Klosec, Ulrike S. Eggertd, and Fiona M. Watta,1

aCentre for Stem Cells and Regenerative Medicine, King’s College London, SE1 9RT London, United Kingdom; bEuropean Bioinformatics Institute, CB10 1SD Hinxton, United Kingdom; cLipotype GmbH, 01307 Dresden, Germany; and dRandall Centre for Cell and Molecular Biophysics, King’s College London, SE1 1UL London, United Kingdom

Contributed by Fiona M. Watt, July 16, 2020 (sent for review June 4, 2020; reviewed by Nils Joakim Faergeman and Kathleen J. Green) While the lipids of the outer layers of mammalian epidermis and such as phospholipids and sphingomyelins, the granular layers their contribution to barrier formation have been extensively exhibiting higher levels of glycosylceramides and cholesterol described, the role of individual lipid species in the onset of sulfates, and the cornified layer mostly composed of cholesterol, keratinocyte differentiation remains unknown. A lipidomic analy- fatty acids, and ceramides, particularly ω-acylceramides that are sis of primary human keratinocytes revealed accumulation of only found in the outermost epidermal layer (14). These differ- numerous lipid species during suspension-induced differentiation. ences are the cumulative result of keratinocyte lipid metabolism, A small interfering RNA screen of 258 lipid-modifying enzymes sebaceous gland secretion, and microbial production (10, 12, 15). identified two genes that on knockdown induced epidermal dif- Beyond their roles as structural components of cells and en- ferentiation: ELOVL1, encoding elongation of very long-chain fatty ergy storage molecules, lipids function as bioactive compounds in acids protein 1, and SLC27A1, encoding fatty acid transport protein a wide number of cellular processes. Posttranslational addition 1. By intersecting lipidomic datasets from suspension-induced dif- of lipid moieties to certain proteins is necessary for their function ferentiation and knockdown keratinocytes, we pinpointed candidate (16), and multiple classes of lipids are involved in endocrine bioactive lipid subspecies as differentiation regulators. Several of (e.g., steroid hormones), paracrine (e.g., eicosanoids), or intra- these—ceramides and glucosylceramides—induced differentiation cellular signaling (e.g., the phosphatidylinositol bisphosphate/ when added to primary keratinocytes in culture. Our results reveal diacylglycerol second-messenger system) (17). However, several the potential of lipid subspecies to regulate exit from the epidermal other lipid classes, such as the sphingolipids ceramide and stem cell compartment. sphingosine-1-phosphate (18, 19), have signaling properties CELL BIOLOGY across multiple tissues, including the epidermis (20). Within each lipids | epidermis | differentiation | keratinocytes | lipidomics lipid class a cell produces a plethora of individual lipid species (i.e., carbon chain length, desaturation, and hydroxylation vari- he skin is the largest organ in the human body, performing ants), whose importance is only beginning to be investigated (21, Tthe primary functions of preventing water evaporation and 22). Indeed, several studies suggest tight regulation and diverse protecting the internal organs from damaging external agents. functions for specific lipid molecular structures in a variety of The outermost part of the skin, the epidermis, is a stratified cellular processes, ranging from cell division (23) to the innate squamous epithelium formed predominantly by keratinocytes (1, 2). immune response (24). These cells migrate from the basal layer, home of the stem cell While the lipid composition of the outer layers of the epi- compartment, upward toward the surface of the skin through the dermis has been well described, the potential role of individual spinous and granular layers to finally reach the cornified layer. lipid species in the early phases of keratinocyte differentiation During this migration, keratinocytes undergo a terminal differenti- remains to be investigated. Due to the high complexity of the ation process, accumulating transglutaminase-cross-linked proteins and secreting lamellar bodies enriched in highly hydrophobic lipid Significance species. Ultimately, keratinocytes lose their nuclei and become corneocytes, held together by cross-linked sheets of apolar lipids in Cells generate a vast repertoire of lipid molecules whose a structure that has been likened to bricks and mortar (1, 3). functions are poorly understood. To investigate whether lipids Human epidermis can be reconstituted in culture, forming can regulate cell fate decisions, we carried out a systematic stratified sheets in which the stem cell compartment and key lipidomic analysis and perturbation of lipid metabolism in cul- elements of the terminal differentiation process are preserved tured human epidermal keratinocytes, determining associa- (1, 4). Clonal growth assays are used as a quantitative readout of tions with the onset of differentiation. We identified individual stem cell abundance in cultures of human epidermal keratino- lipid species that induced exit from the epidermal stem cell cytes (4–6). Keratinocytes can be induced to undergo terminal compartment. Our observations suggest that more research is differentiation in a near-synchronous manner when maintained warranted on the regulation of biological processes via lipid as a single-cell suspension in medium containing methylcellulose species, moving beyond the more conventional contribution of (7, 8). The cells undergo a differentiation commitment phase proteins and nucleic acids. after 4 h and increase expression of differentiation markers between 8 and 24 h (8, 9). While detailed transcriptomic and Author contributions: M.V.R., A.M., U.S.E., and F.M.W. designed research; M.V.R., A.M., and C.K. performed research; U.S.E. contributed new reagents/analytic tools; and M.V.R., proteomic analysis of suspension-induced differentiation was U.S.E., and F.M.W. wrote the paper. carried out previously (9), little is known about changes in the Reviewers: N.J.F., University of Southern Denmark; and K.J.G., Northwestern University. lipid composition of keratinocytes as they exit the stem cell Competing interest statement: C.K. is the head of research and development at compartment. Lipotype GmbH. Lipids are essential for the establishment of an efficient epi- This open access article is distributed under Creative Commons Attribution License 4.0 dermal barrier (10). Alterations in the lipid composition of the (CC BY). cornified layer can lead to morphological defects in the epider- 1To whom correspondence may be addressed. Email: [email protected]. mis and diseases such as ichthyoses (11–13). The predominant This article contains supporting information online at https://www.pnas.org/lookup/suppl/ lipid species change across the different epidermal layers, with doi:10.1073/pnas.2011310117/-/DCSupplemental. the basal and spinous layers enriched in more polar lipid classes First published August 25, 2020.

www.pnas.org/cgi/doi/10.1073/pnas.2011310117 PNAS | September 8, 2020 | vol. 117 | no. 36 | 22173–22182 Downloaded by guest on September 28, 2021 lipid makeup of the epidermis and the limitations of methodol- inhibition of PKC (Fig. 1B). None of the other lipid classes showed ogies available to manipulate lipids, we sought to approach this such a marked change with time in suspension, although it was question in a more tractable system, namely, cultured primary notable that several families were affected by PKC inhibition (SI human keratinocytes. We have used a combination of small in- Appendix,Fig.S1A). Unsupervised clustering based on lipid species terfering RNA (siRNA) screening and lipidomic analysis of abundance showed three main groups of samples: 1) those at the primary human keratinocytes grown in culture. We find changes 0 h time point (adherent cells), 2) untreated cells suspended for 12 in sphingolipids and phospholipids associated with differentia- and 24 h (differentiated cells), and 3) untreated cells suspended tion and show that several ceramide and glucosylceramide mol- for 4 and 8 h (committed cells) together with all of the PKCi- ecules can induce differentiation of cultured keratinocytes. treated samples (Fig. 1C). Consistent with this, principal component analysis (PCA) separated the adherent cell sample from all of the Results others along the first component and the differentiated cells along the The Lipid Composition of Keratinocytes Changes during Suspension-Induced second, while the committed and inhibited cells could not be clearly Differentiation. Single-cell suspension-induced differentiation of pri- separated (Fig. 1D). mary human keratinocytes can be blocked in the presence of a protein In order to identify the most critical lipid species that accu- kinase C inhibitor (PKCi) (Fig. 1A) (9). To examine how lipids were mulated during commitment and differentiation, the samples affected in suspension culture, we analyzed samples using a shotgun were first classified into four categories (“adherent,”“commit- lipidomics approach (25), comparing control and PKCi-treated cells ment,”“differentiated,” and “inhibited”) according to the model (Dataset S1). Analysis of lipid changes at the class level showed sig- in Fig. 1A and subsequently examined using sparse partial least nificant accumulation of ceramides and hexosylceramides as squares regression (sPLS) analysis to maximize separation be- keratinocytes underwent differentiation in suspension, mirroring tween the sample groups, coupled with discriminant analysis the increase in these lipid classes that occurs during epidermal (DA) to pinpoint the main lipid species contributing to such differentiation in vivo (14). This accumulation was blocked on separation (26). The first and second components in the sPLS

A B C Z−score −2 −1 0 1 2

PC 17:1;0−16:0;0 PC 20:1;0−20:1;0 PC 18:0;0−22:6;0 PC 18:2;0−16:1;0 PC 17:1;0−15:0;0 PC 17:0;0−16:0;0 PC 17:1;0−16:1;0 PC 20:1;0−15:0;0 PC 22:4;0−18:0;0 PC 18:2;0−18:2;0 PC O− 16:2;0−16:1;0 PI 18:2;0−18:1;0 PC 18:1;0−15:0;0 PC 18:1;0−15:1;0 PC 18:1;0−17:1;0 PC 18:0;0−18:1;0 PC 16:0;0−14:1;0 PS 18:1;0−22:5;0 PC 18:0;0−16:1;0 HexCer 42:2;3 CE 27:1;0−14:1;0 PC O− 16:0;0−20:3;0 PC O− 16:1;0−20:4;0 PC 17:0;0−14:1;0 PC 18:0;0−20:3;0 PC 18:2;0−15:0;0 PC O− 16:0;0−14:0;0 PE O− 18:1;0−22:4;0 PC 18:0;0−16:2;0 Ceramides PC 20:1;0−20:2;0 PC 18:0;0−14:1;0 PC O− 18:0;0−22:5;0 PE O− 18:0;0−22:4;0 CE 27:2;0−18:1;0 CE 27:2;0−16:0;0 PC 19:1;0−20:4;0 PC 16:1;0−14:0;0 PC 16:0;0−18:3;0 PE 18:2;0−16:1;0 SM 34:1;3 PS 20:3;0−18:0;0 PS 18:0;0−19:1;0 PS 18:3;0−18:0;0 PI 20:5;0−18:1;0 PS 18:0;0−18:0;0 0.8 PS 16:0;0−20:2;0 PG 18:2;0−18:1;0 PI 18:1;0−22:3;0 PI 18:1;0−20:3;0 PE 18:0;0−19:2;0 PE 20:1;0−20:4;0 PG 16:1;0−20:1;0 PC 14:0;0−14:0;0 Sample collection points PE O− 18:2;0−16:0;0 SM 35:1;3 PI 18:1;0−22:5;0 PE O− 16:1;0−22:6;0 PG 18:1;0−18:0;0 PG 16:0;0−20:2;0 PE 18:1;0−18:1;0 PS 18:1;0−22:1;0 PS 18:0;0−19:0;0 PS 18:1;0−21:0;0 PS 16:1;0−20:1;0 PI 20:4;0−19:1;0 PG 16:1;0−18:1;0 PI 18:1;0−21:3;0 PC 17:0;0−16:1;0 PS 18:0;0−22:2;0 SM 31:1;2 PE 20:0;0−20:4;0 PS 18:1;0−24:1;0 PA 18:1;0−18:1;0 PA 18:1;0−18:0;0 PA 14:3;0−20:0;0 PA 16:0;0−16:0;0 PS 16:1;0−24:0;0 PA 14:0;0−18:1;0 PC 16:1;0−16:1;0 PC 16:0;0−16:2;0 PC 16:1;0−14:1;0 PI 14:1;0−18:1;0 SM 34:0;2 0.6 PI 20:5;0−16:0;0 PI 17:0;0−20:2;0 PS 18:1;0−26:1;0 PC 20:1;0−19:1;0 PE 18:1;0−16:1;0 PE O− 18:2;0−20:5;0 PE 18:1;0−19:3;0 PE 18:1;0−20:5;0 PC O− 18:0;0−20:4;0 PE 20:3;0−16:0;0 PC 18:0;0−15:0;0 CE 27:1;0−19:1;0 PE 22:5;0−16:0;0 DAG 18:2;0−16:1;0

% PE 20:4;0−16:1;0 PI 16:0;0−17:1;0 PS 18:1;0−24:0;0 PS 20:4;0−18:1;0 PS 18:2;0−18:1;0 PE 21:3;0−18:0;0 PS 16:0;0−22:0;0 PS 18:1;0−24:3;0 PS 16:0;0−17:1;0 PS 17:1;0−17:0;0 PS 22:6;0−18:1;0 PS 16:1;0−26:0;0 PE 16:0;0−16:0;0 PC O− 16:0;0−16:2;0 PC 15:1;0−15:1;0 SM 33:1;3 PC 17:0;0−20:4;0 SM 35:2;3 PI 20:3;0−18:0;0 PC 18:0;0−18:2;0 PE O− 18:2;0−18:1;0 * PE O− 16:0;0−18:1;0 PE 18:0;0−16:1;0 PC 18:1;0−14:0;0 PS 16:1;0−17:0;0 0.4 PE 18:2;0−18:2;0 DAG 20:2;0−18:0;0 PC 16:1;0−16:2;0 PE 22:2;0−18:0;0 PE 22:3;0−18:0;0 PC 16:0;0−14:0;0 PC 18:0;0−18:3;0 PE O− 16:1;0−22:4;0 PG 18:2;0−18:0;0 PI 16:0;0−22:3;0 PG 16:0;0−18:1;0 0 h 4 h 8 h 12 h 24 h PE O− 16:1;0−20:1;0 PE 22:3;0−16:1;0 PE 18:1;0−16:2;0 PS 16:1;0−22:2;0 PS 17:1;0−22:0;0 PS 16:1;0−26:1;0 PC 18:0;0−17:1;0 PA 18:1;0−20:3;0 PI 16:0;0−19:2;0 PC 20:3;0−14:0;0 PC 20:1;0−17:0;0 PE 18:0;0−20:5;0 PS 16:0;0−18:2;0 PC 22:4;0−16:0;0 PC 19:2;0−18:1;0 PC 15:0;0−15:1;0 PE 22:4;0−16:0;0 PE O− 18:0;0−22:6;0 PE 22:4;0−18:0;0 SM 41:1;2 SM 36:2;2 pmol SM 36:1;2 PG 18:2;0−20:3;0 PC 16:2;0−14:0;0 PC O− 18:0;0−18:2;0 PE O− 18:1;0−22:5;0 PC 19:1;0−14:0;0 SM 39:1;2 DHERENT 0.2 PI 22:5;0−18:0;0 A PS 16:0;0−24:0;0 PS 18:0;0−21:3;0 PC 18:1;0−17:2;0 SM 43:1;2 SM 44:3;2 DIFFERENTIATED PI 18:1;0−21:2;0 PE 20:4;0−16:0;0 PI 19:3;0−18:0;0 PC O− 16:0;0−22:5;0 PE O− 18:2;0−20:4;0 SM 41:2;2 SM 43:2;2 SM 42:3;2 COMMITMENT SM 40:2;2 KERATINOCYTES SM 40:1;2 SM 38:1;2 SM 42:1;2 SM 44:2;2 SM 35:1;2 DAG 18:1;0−17:0;0 KERATINOCYTES DAG 18:0;0−14:0;0 DAG 16:0;0−16:0;0 DAG 18:0;0−18:0;0 CE 27:2;0−20:0;0 DAG 18:2;0−16:0;0 UNDIFFERENTIATED PS 16:1;0−18:0;0 ( ) PI 18:0;0−21:2;0 CE 27:1;0−14:0;0 CE 27:1;0−24:1;0 CE 27:1;0−20:2;0 CE 27:1;0−20:1;0 CE 27:1;0−22:1;0 PI 18:3;0−18:0;0 PA 16:0;0−20:2;0 PI 17:0;0−18:1;0 PI 18:2;0−20:4;0 PI 18:2;0−20:1;0 0.0 PI 17:0;0−22:5;0 PI 18:1;0−18:1;0 PI 16:2;0−18:0;0 PI 18:1;0−19:1;0 PI 16:1;0−18:2;0 Cer 41:1;2 Cer 41:2;2 Cer 36:1;2 Cer 40:2;2 Cer 40:1;2 HexCer 42:1;2 Cer 42:2;2 Cer 42:1;2 Cer 42:3;2 4 8 PG 16:0;0−18:2;0 Cer 34:2;2 PE 22:3;0−18:1;0 PE 18:1;0−22:6;0 CE 27:1;0−22:5;0 12 24 Cer 39:1;2 HexCer 40:1;2 HexCer 34:1;2 HexCer 36:1;2 PC 17:0;0−15:0;0 HexCer 40:2;2 PI 17:1;0−17:0;0 PI 18:2;0−17:0;0 PC 18:3;0−16:1;0 PC 18:1;0−16:0;0 PC 18:1;0−20:5;0 HexCer 41:2;2 HexCer 38:1;2 HexCer 42:2;2 PG 17:0;0−18:1;0 PC 16:0;0−16:3;0 PE 20:3;0−16:1;0 PE 22:4;0−17:0;0 PI 14:0;0−16:0;0 SM 36:3;3 PC 16:1;0−15:0;0 PE 20:2;0−18:1;0 Control DAG 16:0;0−16:1;0 PE O− 18:1;0−18:1;0 PC O− 15:1;0−16:1;0 PC 22:4;0−18:1;0 PS 14:0;0−18:1;0 PC 18:1;0−16:1;0 PC 20:1;0−18:2;0 Cer 30:1;2 PE 20:4;0−20:4;0 PA 17:0;0−18:1;0 SM 44:1;2 Cer 33:1;2 Cer 34:1;2 Cer 38:1;2 DAG 18:0;0−18:1;0 Cer 32:1;2 Cer 37:1;2 PC 17:0;0−22:6;0 PC 20:3;0−16:0;0 PE 20:3;0−18:2;0 PC 20:2;0−17:1;0 PC 20:2;0−18:1;0 Cer 36:2;2 HexCer 41:1;2 PC 17:0;0−17:1;0 PC 18:1;0−17:0;0 PG 18:1;0−20:1;0 PG 16:0;0−19:1;0 PI 20:4;0−19:0;0 PC 18:0;0−22:5;0 PC 22:5;0−17:0;0 PKCi PC 20:3;0−17:0;0 PC 19:1;0−18:0;0 PC 18:1;0−12:0;0 PC 20:2;0−16:1;0 PC 20:2;0−16:0;0 CE 27:1;0−20:3;0 PC 19:1;0−16:0;0 PC 19:1;0−16:1;0 Hexosylceramides PC 20:2;0−18:2;0 PC 18:2;0−16:0;0 PC 20:2;0−18:0;0 PC 19:1;0−18:1;0 PC 18:0;0−20:4;0 PC 19:0;0−18:1;0 PC 19:1;0−17:0;0 PC 20:2;0−14:0;0 PC 20:3;0−16:1;0 PC 20:3;0−18:2;0 PE 16:0;0−16:1;0 PE 22:1;0−18:0;0 PI 18:0;0−20:2;0 PE 20:0;0−16:1;0 PI 18:2;0−20:2;0 PE 20:1;0−17:1;0 PI 14:1;0−16:0;0 PI 16:0;0−18:1;0 PI 18:0;0−22:1;0 PE 18:1;0−18:2;0 PE 20:2;0−16:0;0 IFFERENTIATION PC 19:1;0−18:2;0 D 1.5 DAG 18:1;0−18:2;0 DAG 18:1;0−14:0;0 PS 16:0;0−20:1;0 Chol PI 18:2;0−18:2;0 CE 27:1;0−17:1;0 CE 27:1;0−16:0;0 CE 27:1;0−17:0;0 PI 16:1;0−20:4;0 CE 27:1;0−20:5;0 CE 27:1;0−20:4;0 CE 27:1;0−18:3;0 CE 27:1;0−18:2;0 PC 22:4;0−17:0;0 INHIBITION PC 18:1;0−18:3;0 PI 16:0;0−22:5;0 PI 16:1;0−17:0;0 CE 27:1;0−15:0;0 CE 27:1;0−18:0;0 CE 27:1;0−16:1;0 CE 27:1;0−18:1;0 PE 20:1;0−18:2;0 PE 19:1;0−16:1;0 PE 17:0;0−16:1;0 SM 33:2;2 PE 18:1;0−15:0;0 PE 18:0;0−15:0;0 PE 16:0;0−20:5;0 PE 16:0;0−16:2;0 PE 18:0;0−17:1;0 PE 20:2;0−16:1;0 PE 18:1;0−17:0;0 PG 16:0;0−20:1;0 PI 16:0;0−21:2;0 PG 17:0;0−18:0;0 PE O− 17:1;0−18:1;0 PE 18:0;0−14:1;0 PA 16:0;0−20:1;0 DAG 20:1;0−18:1;0 PI 16:0;0−18:2;0 PI 16:0;0−19:1;0 PE 20:2;0−20:4;0 PE 18:1;0−18:3;0 PE 18:0;0−18:1;0 Primary keratinocytes PA 16:0;0−18:2;0 PG 18:1;0−19:1;0 PI 16:1;0−19:1;0

% PE 18:1;0−20:4;0 DAG 18:0;0−18:2;0 DAG 18:1;0−17:1;0 DAG 18:0;0−20:3;0 DAG 16:1;0−16:1;0 1.0 DAG 17:0;0−16:0;0 PS 18:1;0−21:1;0 PC 20:0;0−20:4;0 PI 18:3;0−18:1;0 DAG 17:0;0−16:1;0 DAG 18:1;0−16:1;0 PE O− 16:0;0−22:5;0 PE 18:0;0−22:5;0 PE 18:0;0−20:3;0 in suspension PE O− 16:1;0−22:3;0 PE O− 16:1;0−20:2;0 PC 18:2;0−17:1;0 PE O− 18:1;0−20:3;0 DAG 18:0;0−20:4;0 PE 20:1;0−20:1;0 DAG 18:1;0−20:4;0 PE O− 16:1;0−20:4;0 PE O− 18:2;0−22:6;0 PE O− 16:0;0−22:6;0 PE O− 18:1;0−20:5;0 PE 18:1;0−17:1;0 PI 16:1;0−20:1;0 SM 34:2;2 SM 42:2;2 SM 33:1;2 SM 32:1;2 in methylcellulose SM 34:1;2 PC O− 16:0;0−20:5;0 PE O− 18:2;0−22:5;0 DAG 20:3;0−16:0;0 PS 16:0;0−20:3;0 DAG 18:0;0−16:1;0 DAG 18:1;0−15:0;0 PE 18:2;0−16:0;0 PE O− 18:1;0−22:6;0 PI 16:0;0−20:1;0 PI 18:2;0−20:3;0 + PKCi PI 17:0;0−20:1;0 PI 16:0;0−20:4;0 PI 18:2;0−18:0;0 PI 16:1;0−20:3;0 PI 16:1;0−16:1;0 PI 16:2;0−16:0;0 PC 20:0;0−18:1;0 PI 16:0;0−20:2;0 PA 16:0;0−20:3;0 PI 14:0;0−18:2;0 PI 14:1;0−18:0;0 PG 16:0;0−18:0;0 PI 14:0;0−16:1;0 * PG 18:0;0−20:2;0 0.5 PI 17:0;0−20:3;0 pmol PE 20:3;0−17:0;0 PI 16:0;0−22:4;0 PA 16:0;0−18:1;0 PI 16:0;0−22:1;0 PI 16:0;0−19:3;0 PE O− 16:0;0−20:5;0 PI 16:1;0−20:2;0 PS 18:0;0−20:0;0 PS 18:1;0−22:0;0 PS 18:1;0−22:3;0 PS 16:1;0−20:3;0 DAG 18:2;0−18:2;0 PI 17:0;0−17:0;0 PE 20:3;0−18:1;0 PS 16:1;0−20:2;0 PS 16:1;0−18:1;0 * PS 18:0;0−22:1;0 Time in suspension 0.0 8 h 4 h 0 h 4 8 24 h 12 h 12 24 Time (h) 8 h + PKCi4 h + PKCi 12 h + PKCi 24 h + PKCi DFE Contribution on component 2 PE 20:3;0-18:2;0 PG 18:2;0-18:1;0 PC 16:1;0-14:0;0 PC O- 18:1;0-16:1;0 PC 18:2;0-14:0;0 Cer 40:2;2 PC 18:1;0-18:2;0 PA 16:0;0-16:0;0 Principal Component Analysis sparse Partial Least Squares regression PC 20:2;0-18:1;0 PC 18:2;0-18:2;0 PC 22:5;0-18:1;0 PC 20:3;0-16:0;0 PG 18:2;0-18:2;0 PC 18:0;0-18:1;0 PC 20:4;0-16:0;0 PS 18:2;0-18:0;0 PC 18:0;0-20:5;0 Cer 40:1;2 PC 22:3;0-16:0;0 Component 1 (28%) HexCer 42:1;2 PC 18:1;0-15:0;0 PC 20:3;0-16:1;0 PC 16:1;0-16:1;0 0 10 20 30 PI 18:2;0-18:1;0 PC 20:3;0-18:1;0 PC 17:0;0-18:2;0 PC 18:1;0-16:0;0 PI 18:1;0-20:3;0 10 PS 18:2;0-22:0;0 PC 18:1;0-17:0;0 PS 18:1;0-18:0;0 PE O- 16:1;0-22:5;0 PC 16:0;0-16:2;0 HexCer 40:1;2 PI 18:1;0-18:1;0 CE 27:1;0-22:5;0 Legend PG 18:2;0-20:3;0 DAG 18:2;0-16:1;0 PS 20:3;0-18:0;0 HexCer 41:2;2 HexCer 38:1;2 00hr HexCer 42:2;3 PS 18:1;0-18:1;0 HexCer 34:1;2 04hr 5 HexCer 36:1;2 HexCer 42:2;2 PC 18:0;0-18:3;0 08hr PI 16:0;0-22:4;0 HexCer 40:2;2 PI 16:1;0-20:4;0 12hr Outcome 0 0 24hr -0.10 -0.05 0.00 0.05 0.10 PKCi 04hr 10 Adherent PKCi 08hr 5 Commitment -5 0 Differentiated

PC2: 16% expl. var PKCi 12hr Contribution on component 3 SM 35:1;3 Inhibited -5 PC 17:1;0-16:1;0

PKCi 24hr SM 32:1;2 Component 3 (8%) 3 Component DAG 17:1;0-16:0;0 HexCer 41:1;2 PC 15:0;0-15:0;0 -10 PE 18:1;0-20:4;0 DAG 18:2;0-18:2;0 PC 19:1;0-16:1;0 PC O- 16:1;0-14:0;0 PA 18:1;0-18:0;0 -15 SM 33:1;2 PI 17:0;0-20:3;0 -10 PE O- 16:1;0-20:4;0 Component 2 (16%) PC O- 16:0;0-16:2;0 CE 27:1;0-20:2;0 PI 18:0;0-21:3;0 Cer 30:1;2 PC O- 16:0;0-16:0;0 PE O- 18:1;0-20:5;0 PS 18:1;0-24:1;0 PS 16:1;0-18:0;0 PE O- 18:0;0-22:5;0 PG 18:2;0-18:0;0 DAG 20:3;0-18:1;0 CE 27:1;0-14:1;0 Outcome PS 18:1;0-20:0;0 PE O- 18:1;0-18:1;0 PE 18:0;0-18:3;0 PE 20:1;0-18:2;0 -30 -20 -10 0 10 PI 18:2;0-18:0;0 Adherent Differentiated PS 18:2;0-22:1;0 CE 27:1;0-18:2;0 PC 17:0;0-14:0;0 PC1: 30% expl. var CE 27:1;0-24:1;0 Commitment Inhibited PS 16:1;0-18:1;0 PS 18:1;0-20:2;0 PE O- 16:1;0-16:1;0 PC 16:0;0-20:5;0 PC 20:1;0-17:0;0 PE O- 18:1;0-20:3;0 PC 18:1;0-22:6;0 CE 27:1;0-17:0;0 PI 20:5;0-16:0;0 CE 27:1;0-24:0;0 PC O- 16:2;0-16:1;0 PA 18:2;0-18:1;0 PC 20:4;0-16:1;0 PE 22:5;0-16:1;0 CE 27:1;0-16:1;0

-0.2 -0.1 0.0 0.1 0.2

Fig. 1. Keratinocyte lipid composition changes during suspension-induced differentiation. (A) Schematic overview of the experimental strategy and keratinocyte response to treatments (9). (B) Class-level variation of ceramides and hexosylceramides during suspension-induced keratinocyte differentiation, expressed as a percentage of the total sample lipids. Error bars indicate SDs. (C) Heat map representation and two-dimensional (2D) clustering of samples (Euclidean distance and complete-linkage clustering) based on the Z-scores of all lipid species (y axis) identified in the lipidomic analysis of suspension-induced keratinocyte differentiation; different time points/conditions are shown along the x axis; dendrogram branches are color-coded according to the schematic in A.(D) Sample variation along the first two principal components. (E) Separation along three components by sPLS after fitting the samples to the model shown in A.(F) Contribution of the 50 most discriminant lipid species to the separation along the second (Upper panel) and third (Lower panel) components of sPLS. Error bars indicate SDs; P values are calculated using multiple t tests with Holm–Sidak adjustment for multiple comparisons (*P < 0.05).

22174 | www.pnas.org/cgi/doi/10.1073/pnas.2011310117 Vietri Rudan et al. Downloaded by guest on September 28, 2021 regression separated adherent and differentiated samples from These results reveal that perturbation of lipid metabolism in the rest, respectively, while the samples undergoing commitment primary human keratinocytes can interfere with terminal could be isolated along the third component (Fig. 1E). The differentiation. contribution of lipid species to components 2 and 3 could therefore identify which discriminating molecules were enriched Lipidomic Analysis Identifies Potential Lipid Regulators of Keratinocyte in the committed and differentiated samples, where a total of Differentiation. When a lipid-modifying enzyme is down-regulated, 145 lipids were found to accumulate (Fig. 1F and SI Appendix, it is difficult to predict how the lipid composition of the cell will be Fig. S1C and Table S1). affected. This is because cellular lipid metabolism is extremely “ ” These results show that during suspension-induced differen- complex and redundant and many of the building blocks of tiation keratinocytes extensively change their lipid composition complex lipid species are shared across multiple lipid classes (18). and accumulate a number of specific lipid species. To accurately define the lipid species involved in regulating differentiation, we performed a comprehensive lipidomic analysis of ELOVL1 SLC27A1 Knockdown of Lipid-Modifying Enzymes Can Induce Differentiation. cells transfected with siRNA targeting or or a Having identified a panel of lipids enriched during keratinocyte nontargeting siRNA to control for the potential influence of differentiation, we asked whether any of them regulated the transfection reagents (31). We collected samples at 24, 48, or 72 h onset of differentiation. Despite numerous advances, current posttransfection and performed lipidomics analysis on them methodologies do not allow direct and systematic manipulation (Dataset S3). of lipid molecular species (27). We therefore introduced per- A class-level analysis of the lipidomics data did not show any obvious divergence in accumulation/depletion trends between turbations in the lipid composition of adherent cultures of pri- the control and knocked-down samples (SI Appendix, Fig. S3B). mary human keratinocytes by transfecting them with a panel of When looking at the full panel of individual lipid species, un- 258 siRNAs against lipid-modifying enzymes (23). The trans- supervised clustering of the samples (Fig. 3A) revealed that the fected cells were incubated under conditions (feeder-free main factor contributing to sample grouping was time, with keratinocyte serum-free medium [KSFM]) that would en- cultured primary keratinocytes exhibiting a striking plasticity in rich for undifferentiated cells or were treated with medium their lipid composition, even in control conditions. Principal supplemented with fetal bovine serum, which is known to component analysis was consistent with this observation, in that stimulate accumulation of differentiated cells (28, 29). Immu- the three different time points, irrespective of treatment, sepa- nofluorescence staining for involucrin, an early marker of differ- rated neatly along the first component (Fig. 3B). However, at 48

entiation in cultured keratinocytes (9, 30), was used as a readout and 72 h the ELOVL1 knockdown cells clustered separately from CELL BIOLOGY of differentiation (Dataset S2). the other samples, indicating that the down-regulation of this Modified Z-score transformation of the data (using the sample enzyme caused a shift in the lipid makeup of the cells. Con- median and median absolute deviation) allowed pooling of both versely, the siSLC27A1 samples were never clearly distinguish- culture conditions in a single dataset. The siRNA screen yielded able from the nontargeting control samples. reproducible results, as indicated by the good correlation ob- served between each replicate and the mean of the quadrupli- Specific Lipid Molecules Can Induce Keratinocyte Differentiation in cates (Pearson’s r ∼ 0.8; SI Appendix, Fig. S2A). No correlation Culture. To identify the critical lipid species that accumulated was seen when plates containing different siRNAs were plotted in the knockdown cells, we conducted a sPLS-DA analysis on the against each other (SI Appendix, Fig. S2B), ruling out possible 48 and 72 h samples separately (Fig. 3 C and D). The analysis contributions of a sample’s position within a plate. siRNA identified 195 unique discriminant lipid species enriched in against involucrin was used to estimate transfection efficiency siELOVL1 cells at either time point. Interestingly, despite the and to validate the antibody (SI Appendix, Fig. S2C). Unsuper- similarity to the controls, the analysis was also able to separate vised two-way clustering of the screen results grouped together with good accuracy the siSLC27A1 samples and allowed us to replicates of the same treatment but not related enzymes pinpoint 148 discriminant lipid species that accumulated in these E F SI Appendix C D (Fig. 2A). Ten knockdowns elicited a response, with four inhib- cells (Fig. 3 and and , Fig. S3 and and Tables iting differentiation and six inducing it, based on statistical sig- S2 and S3). We next compared the lipids enriched in committed and differentiated cells (Fig. 1) with the ones that accumulated nificance and fold change with respect to the average of control ELOVL1 SLC27A1 wells within the same plate (Fig. 2B). in or knocked-down keratinocytes to see if The candidate enzymes were validated by confirming the effect there were any common species. Remarkably, these two inde- of knockdown on involucrin protein expression and analyzing the pendent experimental approaches, which stimulated differentiation in completely different ways, yielded an overlap of 26 and 16 transcription of a panel of additional differentiation markers in candidate bioactive lipid species in the case of the ELOVL1 and either differentiating (for the knockdowns that inhibited differ- SLC27A1 C knockdown, respectively, that could potentially function entiation, Fig. 2 ) or nondifferentiating conditions (for the “ ” A D as differentiation inducers ( intersection sets ;Fig.4 ). The over- knockdowns that induced differentiation, Fig. 2 ). None of the lapping lipid species were sphingolipids (12/42), more specifically candidate differentiation-inhibiting knockdowns significantly im- ceramides (6/42) and hexosylceramides (6/42), and glycerophospholipids pacted the ability of keratinocytes to undergo differentiation, (30/42), with the most represented types being phosphatidylcholines whereas the knockdown of three enzymes was indeed able to in- (8/42) and phosphatidylserines (7/42). The fatty acid moieties pre- duce expression of involucrin and at least one other differentiation sent had chain lengths ranging between 14 and 24 carbon atoms, marker. We subsequently performed colony formation assays to containing up to four unsaturations (Table 1). further validate these effects, and in two cases (siELOVL1 and We sought next to verify whether any of the species identified siSLC27A1) we were able to confirm a significant induction of could stimulate keratinocytes to undergo differentiation. Hex- keratinocyte differentiation on the basis of a reduction in colony osylceramides can have a glucose or a galactose as their head formation, a surrogate of stem cell activity (9)(Fig. 2F). We also group, but we assumed them to be glucosylceramides due to the performed colony formation assays on the whole panel of candi- established presence and importance of this lipid class in the date differentiation-inhibiting knockdowns but did not observe epidermis (14, 32). We tested two ceramides and two gluco- any significant effects (Fig. 2E). Deconvolution of the two vali- sylceramides from our intersection sets and included one more dated siRNA pools did not suggest off-target effects (SI Appendix, ceramide species that differed from a candidate molecule by Fig. S2E). having a chain two carbon atoms longer and still appeared in the

Vietri Rudan et al. PNAS | September 8, 2020 | vol. 117 | no. 36 | 22175 Downloaded by guest on September 28, 2021 A Effect of knockdown B

HADH AGPS Diff. Diff. PLA2G1B SMPD2 AWAT2 SLC27A1 Effect on differentiation: AWAT1 ACSM2A ELOVL1 PTGS1 inhibition induction SQLE SERINC1 ELOVL4 ALDH8A1 ALOX12 DGKE FDFT1 OXSM AGPAT5 FAR1 DEGS2 Inhibition Induction ELOVL3 AGMO FADS3 ACSM2B ELOVL6 NAGA GPAT2 ACSL4 ACAA2 −4−2 024 ORMDL3 DPAGT1 GBGT1 DHCR24 AKR1B1 ALOX15B ACSM1 ALOX12B Value HADHA ASAH2 ACSL3 MBOAT7 FADS1 AGPAT9 FA2H PIK3C2B siDHCR7 PIGK AGPAT3 MECR AGK ACSL5 DGAT2L6 Untreated SMPDL3A ASAH2C PPAP2C LYPLA1 SGMS1 SCD PGS1 PIGO PIK3C2A LSS SOAT1 FDPS IDI2 KDSR PLSCR2 ACACB LPIN1 AGPAT6 GPAM AGPAT4 PIK3CD CERS5 PLD1 AGPAT2 SCD5 CBR4 LCLAT1 LYPLA2 CERS4 PTGS2 C14orf1 ACSL6 ALOX5 EBP ELOVL2 HEXB GK B4GALT6 GLB1 PGAP2 ORMDL1 SGMS2 PTPLAD2 AGPAT1 SACM1L PTPLB GNPAT siLARGE MVD siALDH8A1 PI4K2A PEX2 CEPT1 ST8SIA5 CRLS1 DOLK (p) PTEN CHPT1 PIK3CG PTDSS2 SGPP1 SPTLC2 SAMD8 MTMR3 CERK 10 CERS1 LPGAT1 B3GALT4 SSSPTB PIGB PLSCR3 siPPAP2A siELOVL1 PLA2G15 CHKA SGPP2 A4GALT ACADVL ACOT8 CYP51A1 ST3GAL5 FADS6 siELOVL3 SC5DL LPCAT2 PPAP2B

UGCG -log PISD GBA2 GALC siSMPD4 PTPLA DPM1 GGPS1 B3GNT5 ALOX15 PIGW GLA PLCE1 ALG9 LPCAT3 PLA2G6 siELOVL4 LIAS PLA2G4C PIGN CERS6 PIGA LPIN2 PLA2G4F siALDH1A2 ABHD5 siSLC27A1 SMPD3 LCAT PEMT Scramble HMGCS2 PIK3R1 B4GALNT1 INPP5E PIK3CA PGAP3 PLCB1 AASDHPPT PPT1 SPTLC1 ACER3 PLA2G5 DEGS1 TAZ DGAT1 ORMDL2 IDI1 PSAPL1 PIGG CERS2 PLA2G4B CDS1 ACADL B3GALNT1 PIK3C2G OCRL SPHK1 PLA2G4A PSAP GBA3 CDS2 LPCAT4 MOGAT2 HMGCS1 OLAH ACSL1 SMPD1 SERINC4 SGPL1 MGLL ELOVL7 ECHS1 ACER1 ASAH1 SLC27A2 HSD17B4 HMGCR FASN FADS2 ACSM3 DGAT2 PI4KB ALDH1A2 MTM1 UGT8 ACER2 LIPT2 CDIPT SPTLC3 SMPD4 PNPLA3 GBA PPAP2A LARGE ELOVL5 MBOAT1 020123 8 PECR FAR2 MBOAT2 PTPLAD1 SSSPTA PLSCR1 MOGAT3 SRD5A3 PTDSS1 SERINC5 SPHK2 ST8SIA1 PI4KA ST8SIA3 PIKFYVE TECR FIG4 EPT1 DOLPP1 PLSCR4 HPGDS SMPDL3B PIK3CB -1 -0.5 0 0.5 1 1.5 ST6GALNAC6 GAL3ST1 CH25H PIP5K1A CERS3 MOGAT1 ALDH3A2 LIPC LPCAT1 DHCR7 Vehicle Serum log2(Fold Change) C D 72 h p.t. EF 72 h p.t. 1.0 siALDH8A1 1.0 0.8 Clonogenicity assay Clonogenicity assay siALDH1A2 0.8 siDHCR7 1.0 1.0 siLARGE 0.6 siELOVL1 0.6 50 80 0.5 0.4 0.4 siPPAP2A siELOVL3 0.5 0.2 40 siSMPD4 siELOVL4 60 0 0.2 * siSLC27A1 30 Knockdown 0

IVL 40

PPL ** efficiency Knockdown 20 EVPL TGM1 SBSN IVL PPL

ZNF750 efficiency

EVPL 20 TGM1 SBSN Colonies/well Colonies/well 10 ZNF750 siScramble 96 h p.t. 96 h p.t. 0 0 siScramble siScramble siALDH1A2 siALDH8A1 siDHCR7 siLARGE siELOVL1 siPPAP2A siELOVL3 siELOVL4 siSMPD4 siSLC27A1 *

0 1 2 3 siSMPD4 0.0 0.5 1.0 1.5 siLARGE siELOVL1 siPPAP2A siSLC27A1 siScramble siScramble siALDH8A1 % Involucrin-positive cells % Involucrin-positive cells siALDH1A2 (fold change) (fold change)

Fig. 2. siRNA-mediated knockdown of lipid-modifying enzymes can affect keratinocyte differentiation. (A) Heat map representation of the Z-scores of involucrin levels in 258 lipid-modifying enzyme knockdowns (y axis) after 2D clustering of samples (x axis) (Euclidean distance and complete-linkage clustering). (B) Volcano plot of the screening results to identify hits based on statistical significance (P < 0.05) and fold change (FC) with respect to nontargeting siRNA controls included in each individual plate (FC < 0.6 for differentiation inhibition; FC > 2.5 for differentiation induction). Validation of differentiation- inhibiting (C) and differentiation-inducing (D) knockdowns by qPCR of differentiation markers (Upper) and immunofluorescence staining of involucrin (Lower); p.t.: posttransfection. Colony formation assay validation of differentiation-inhibiting (E) and differentiation-inducing (F) knockdowns with representative images for each siRNA. Error bars indicate SDs; P values are calculated using Dunnet’s multiple comparison test (*P < 0.05, **P < 0.01).

set enriched during suspension-induced differentiation. As a the skin microbiome (15), epidermal lipids can regulate exit from negative control, we employed one ceramide and one sphingo- the epidermal stem cell compartment. By integrating lip- myelin species that did not appear in either of our enriched lipid idomic datasets with an siRNA screen of lipid modifying en- sets. Notably, the negative control ceramide had the same zymes, we identified ELOVL1 and SLC27A1 as genes that, chemical structure as the lipids in the intersection sets, with the when knocked down, caused keratinocytes to undergo dif- exception of bearing an α-hydroxylated fatty acid moiety. ferentiation. The knockdown of ELOVL1 and, to a lesser We incubated cultured primary human keratinocytes for 48 h extent, SLC27A1 caused a shift in the lipid composition of with increasing concentrations of our candidate bioactive lipids keratinocytes, and introduction of individual ceramides and or the control lipids. At the highest concentration tested (100 μM) glucosylceramides mimicked the ability of the knockdowns to all candidate lipids, but not the control lipids, produced a signif- promote differentiation. It should be noted that our analysis icant increase in involucrin-expressing cells (Fig. 4B). None of the did not cover the full gamut of lipid classes, lacking, for ex- bioactive lipids were toxic, as assessed by cleaved caspase 3 staining ample, eicosanoids and fatty acids. It is therefore possible that (SI Appendix,Fig.S4). To further validate the effect of the lipids, additional lipid species also participate in the regulation of we tested whether they could affect the ability of keratinocytes keratinocyte differentiation. to form colonies. Incubation with the lipids for 48 h reduced the ELOVL1 catalyzes the elongation of saturated and monoun- average number of colonies formed in all conditions, albeit not saturated C20 to C26 acyl-CoAs. ELOVL1 activity is linked to significantly. Addition of either candidate glucosylceramide the production of sphingolipids as it is regulated by ceramide significantly decreased colony size (Fig. 4C). synthase (CerS) enzymes (33). Elovl1 knockout mice die shortly after birth due to skin barrier deficiencies caused by the impaired Discussion formation of lipid lamellae and defective desquamation (34). Our studies show that in addition to forming the epidermal These mice are depleted of ≥C26 ceramides and, similar to our barrier, acting as intracellular signaling molecules, and modulating results, accumulate ≤C24 ceramides but do not present any

22176 | www.pnas.org/cgi/doi/10.1073/pnas.2011310117 Vietri Rudan et al. Downloaded by guest on September 28, 2021 Z−score Contribution on component 1

A C E PE O- 18:2;0-22:1;0 TAG 48:2;0 PI 18:1;0-20:1;0 −2 −1 0 1 2 Cer 38:1;2 DAG 14:1;0-18:1;0 TAG 50:2;0 PE 14:0;0−18:1;0 PE O− 16:0;0−22:1;0 PS 16:1;0−24:0;0 PS 18:1;0−24:1;0 LPG 16:1;0 TAG 44:1;0 PE O− 16:1;0−24:2;0 PS 16:1;0−24:1;0 PE O- 18:1;0-20:2;0 Cer 44:2;2 PC 14:1;0−17:0;0 PC O− 18:0;0−14:1;0 PE O− 16:1;0−16:0;0 PG 16:1;0−16:1;0 PG 14:0;0−18:1;0 PE O− 16:0;0−16:1;0 PE O− 18:1;0−14:1;0 DAG 14:0;0−18:2;0 PE 16:1;0-20:1;0 PC 16:1;0−24:1;0 PC 14:0;0−15:0;0 PE 14:0;0−18:2;0 SM 42:2;2 PE 18:1;0−24:1;0 PS 18:1;0−24:0;0 Cer 44:1;2 Cer 42:1;2 SM 44:2;2 PE O- 18:2;0-20:2;0 PE 16:1;0−24:1;0 TAG 46:0;0 PE O− 18:2;0−18:3;0 PS 16:0;0−24:1;0 PG 14:0;0−18:2;0 PS 18:2;0−24:1;0 PS 16:0;0−22:2;0 PC 12:0;0−18:0;0 PC 16:1;0−22:2;0 SM 44:2;2 PG 16:1;0−18:1;0 PC O− 16:0;0−16:0;0 PG 16:1;0−18:2;0 PC O− 16:1;0−16:0;0 PE 14:0;0−16:1;0 PE 18:2;0−20:3;0 PC O− 17:0;0−17:0;0 PC O− 18:0;0−16:0;0 PS 18:1;0−18:1;0 PG 18:1;0-20:1;0 DAG 16:0;0−20:2;0 PS 16:1;0−22:2;0 CL 68:5;0 PI 18:2;0−20:4;0 DAG 16:1;0−20:2;0 DAG 16:1;0−18:1;0 DAG 12:0;0−18:1;0 LPI 16:1;0 PI 18:1;0−18:1;0 PE O- 18:2;0-19:2;0 DAG 14:0;0−20:1;0 PE O− 16:0;0−22:2;0 PI 18:1;0−20:3;0 PS 16:1;0−20:2;0 PE O− 18:2;0−14:1;0 PC 16:1;0−17:1;0 PI 16:1;0−20:2;0 PI 18:0;0−18:3;0 PI 16:0;0−20:2;0 CL 64:3;0 Cer 32:1;2 PG 16:0;0−18:2;0 PC 16:1;0−18:1;0 PS 16:1;0−20:0;0 PE O− 18:1;0−16:0;0 PG 16:0;0−20:1;0 DAG 18:1;0−18:2;0 DAG 16:0;0−20:3;0 PI 18:1;0−19:2;0 CL 70:4;0 Cer 44:1;2 PI 16:1;0−20:3;0 PA 14:1;0−18:0;0 PI 18:1;0−18:2;0 PS 16:1;0−18:1;0 PI 14:0;0−18:2;0 PC 14:1;0−14:1;0 PS 14:1;0−18:0;0 PE 18:1;0−20:4;0 CL 66:2;0 PE O- 16:0;0-16:1;0 PI 18:0;0−18:2;0 PC 16:1;0−16:1;0 PC 16:1;0−18:0;0 PE 16:1;0−22:4;0 PG 18:1;0−18:3;0 DAG 14:0;0−16:1;0 PS 18:0;0−20:4;0 PS 18:1;0−20:3;0 PI 17:1;0−18:2;0 PE O- 18:2;0-22:3;0 PS 16:1;0−24:2;0 PI 16:2;0−18:0;0 PE O− 16:0;0−22:4;0 DAG 18:0;0−20:3;0 PA 16:0;0−18:2;0 PI 18:1;0−22:2;0 PG 16:0;0−20:2;0 CL 72:4;0 PI 16:2;0−18:1;0 HexCer 40:1;2 TAG 50:1;0 PC 18:1;0−19:2;0 TAG 56:5;0 PC 18:1;0−20:3;0 PE O− 16:2;0−18:1;0 sparse Partial Least Squares regression PE O− 16:2;0−16:0;0 PE O− 16:2;0−16:1;0 PE O− 16:1;0−24:1;0 TAG 56:6;0 PC 16:1;0-20:1;0 PE O− 18:1;0−16:2;0 PC O− 18:1;0−18:0;0 PC O− 16:1;0−18:0;0 PC 18:1;0−22:3;0 PG 16:1;0−20:3;0 PE O− 16:2;0−18:0;0 PC O− 17:1;0−16:1;0 PI 17:0;0−17:0;0 PC O− 18:1;0−22:5;0 PG 17:1;0−18:1;0 PC 14:0;0-16:0;0 PC O− 16:1;0−18:2;0 PC 14:0;0−16:0;0 PC O− 16:0;0−16:1;0 PE O− 16:1;0−16:1;0 PE O− 16:1;0−14:0;0 PS 18:1;0−24:2;0 PE O− 18:1;0−16:1;0 PC O− 18:0;0−16:1;0 PC 14:0;0−14:0;0 PC 16:0;0-16:0;0 PE O− 18:2;0−24:1;0 TAG 48:3;0 TAG 46:2;0 Cer 42:2;2 PE O− 18:1;0−14:0;0 PE O− 16:1;0−12:0;0 CL 66:5;0 PI 18:1;0−20:2;0 CL 64:4;0 PS 18:1;0-24:1;0 PI 18:0;0−20:3;0 PI 18:0;0−20:2;0 PE O− 16:1;0−22:2;0 PE O− 18:2;0−16:1;0 PE O− 16:1;0−22:1;0 CL 66:4;0 PI 18:2;0−20:2;0 CL 68:4;0 PS 18:1;0−20:2;0 TAG 48:1;0 TAG 46:1;0 TAG 48:2;0 TAG 48:1;0 PE O− 16:1;0−24:3;0 TAG 56:2;0 PC O− 16:1;0−16:1;0 TAG 60:3;0 TAG 58:2;0 TAG 60:2;0 PE O- 16:0;0-20:1;0 Cer 42:0;2 PE O− 18:2;0−24:2;0 PC O− 16:1;0−22:2;0 TAG 51:2;0 TAG 54:5;0 PS 18:0;0−24:1;0 PC 16:1;0−16:2;0 PS 18:1;0−22:3;0 LPI 20:2;0 PE O- 16:1;0-14:0;0 PC O− 16:0;0−18:1;0 PC O− 18:0;0−14:0;0 PC 14:0;0−18:2;0 PC O− 18:1;0−16:2;0 PC 18:1;0−20:2;0 PC O− 18:1;0−14:0;0 PC 14:0;0−18:1;0 PC O− 17:1;0−16:0;0 PC O− 16:2;0−20:2;0 PC O− 16:2;0−18:1;0 PE 18:1;0-24:0;0 Cer 34:0;2 PC O− 16:2;0−16:1;0 PG 19:1;0−19:1;0 PS 17:0;0−20:3;0 PC O− 18:0;0−20:4;0 PE O− 18:1;0−18:1;0 PS 18:0;0−22:3;0 10 LPS 18:0;0 PS 18:2;0−20:1;0 SM 32:1;2 PA 16:0;0−20:1;0 PS 18:2;0−22:1;0 PE O− 18:2;0−12:0;0 DAG 18:1;0−20:1;0 DAG 18:0;0−20:2;0 PE O− 18:1;0−22:2;0 PC 18:2;0−20:2;0 PE O− 18:2;0−19:2;0 PC 16:0;0−22:2;0 PE 14:0;0-16:0;0 PA 18:1;0−18:1;0 PE O− 16:1;0−21:2;0 PI 18:1;0−22:3;0 PI 15:0;0−20:2;0 LPI 20:4;0 PG 16:0;0−20:4;0 PC O− 17:2;0−18:1;0 PC O− 17:1;0−18:2;0 PC 12:0;0−16:1;0 DAG 16:1;0-16:1;0 TAG 53:2;0 PG 20:3;0−20:3;0 PC O− 16:1;0−22:3;0 PC 18:1;0−19:0;0 PE O− 18:2;0−22:0;0 PS 16:0;0−22:3;0 PG 16:1;0−22:2;0 PC 12:0;0−17:1;0 PC 20:1;0−20:1;0 PC O- 16:0;0-16:0;0 PC O− 16:1;0−20:3;0 TAG 51:3;0 PE O− 16:2;0−14:0;0 PG 18:2;0−20:3;0 PI 20:2;0−20:3;0 PE O− 17:2;0−20:1;0 CL 68:6;0 PC 18:1;0−18:2;0 PC 18:1;0−18:1;0 Cer 44:2;2 PE O− 18:2;0−19:1;0 PC O− 18:1;0−18:1;0 PC 18:0;0−18:1;0 TAG 54:2;0 PC O− 18:1;0−20:3;0 Chol PE 18:0;0−20:1;0 Cer 40:1;2 PC 16:2;0−18:1;0 PE O− 18:1;0−22:3;0 PE 18:0;0-20:2;0 PC 14:0;0−20:1;0 PG 18:1;0−20:1;0 PE O− 18:2;0−20:1;0 PG 18:2;0−20:1;0 TAG 56:3;0 PE 18:1;0−20:2;0 SM 32:0;2 PC 15:0;0−22:1;0 PE O− 16:1;0−21:3;0 TAG 60:3;0 PE 18:2;0−20:2;0 PC O− 18:0;0−18:1;0 PS 18:0;0−20:2;0 PC O− 18:2;0−20:1;0 PC 18:0;0−20:2;0 PC O− 16:1;0−20:1;0 PE O− 18:2;0−20:2;0 PE O− 18:2;0−22:3;0 PC O− 16:2;0−20:1;0 Cer 40:2;2 PC O− 18:1;0−20:1;0 PC O− 18:0;0−20:2;0 PE O− 18:2;0−22:1;0 PE O− 18:2;0−22:2;0 PE O− 18:1;0−20:2;0 PE O− 17:1;0−20:2;0 PE O− 18:2;0−18:1;0 PC O− 16:0;0−22:2;0 PE 17:1;0−19:1;0 PC 18:1;0-20:1;0 PS 18:0;0−20:0;0 PS 17:1;0−20:1;0 PE O− 18:2;0−16:2;0 PE O− 16:0;0−20:2;0 PE 16:1;0−20:2;0 PE O− 16:0;0−22:3;0 PC 16:1;0−20:3;0 PC 16:1;0−18:2;0 PC 14:0;0−20:2;0 HexCer 40:2;2 PC 16:0;0−20:2;0 PE O− 16:1;0−20:2;0 PC O− 18:1;0−16:0;0 PE O− 18:2;0−14:0;0 PE O− 18:2;0−16:0;0 PC 16:1;0−20:2;0 PE O− 16:1;0−22:3;0 PG 18:1;0−20:2;0 PC O− 16:0;0−20:2;0 PE 18:1;0-20:1;0 PG 18:1;0−18:1;0 PG 18:2;0−20:2;0 PC O− 18:1;0−14:1;0 PG 18:1;0−18:2;0 PE O− 18:1;0−22:1;0 PE 16:0;0−20:2;0 PE O− 16:2;0−20:1;0 PS 18:0;0−22:2;0 PC O− 16:0;0−22:1;0 PI 18:2;0−20:3;0 PS 18:1;0-22:1;0 PI 16:0;0−22:3;0 PG 18:2;0−18:2;0 PG 18:1;0−20:3;0 PC O− 16:1;0−20:2;0 PC O− 18:1;0−16:1;0 PG 16:1;0−20:2;0 PC O− 18:2;0−16:1;0 PC O− 16:0;0−16:2;0 PC O− 18:1;0−20:2;0 PC 16:0;0-20:1;0 PC O− 18:0;0−20:3;0 PC O− 18:0;0−20:1;0 PC O− 16:0;0−22:3;0 PE 18:0;0−20:2;0 PC 14:0;0−20:3;0 PC O− 16:0;0−20:3;0 PG 18:1;0−20:4;0 LPG 18:1;0 TAG 52:4;0 SM 38:1;2 TAG 50:3;0 TAG 50:2;0 PS 18:1;0−22:2;0 TAG 58:4;0 TAG 56:4;0 PC O− 18:2;0−20:2;0 PC O− 18:2;0−14:0;0 TAG 54:4;0 PC O− 18:2;0−18:1;0 PE O- 18:2;0-20:1;0 PC 18:0;0−20:3;0 TAG 54:3;0 TAG 52:3;0 PC O− 18:2;0−16:0;0 TAG 58:3;0 PE O− 16:1;0−18:1;0 TAG 52:2;0 PC O− 16:1;0−18:1;0 PS 18:2;0−20:0;0 SM 36:1;2 PI 17:0;0−18:2;0 PS 17:0;0−20:1;0 PE O− 18:0;0−22:3;0 PE 18:1;0−22:2;0 PC O− 16:1;0−22:5;0 PE O− 18:2;0−17:0;0 PG 18:1;0−22:5;0 PE 18:0;0−22:2;0 PS 18:1;0−19:1;0 PS 18:1;0-20:1;0 PS 17:0;0−17:1;0 PS 16:0;0−22:1;0 TAG 53:3;0 PC O− 18:2;0−20:3;0 PI 18:1;0−22:4;0 PG 18:0;0−20:3;0 Cer 42:2;1 PE 18:1;0−20:1;0 PS 20:1;0−20:2;0 PE 20:1;0−20:1;0 PE O- 16:1;0-20:1;0 PC 17:1;0−20:1;0 PC 18:1;0−20:1;0 PC 16:0;0−20:1;0 PS 18:1;0−20:1;0 PE O− 18:1;0−20:1;0 SM 32:1;2 SM 40:2;2 Cer 34:2;2 PE 16:1;0−20:1;0 Cer 38:2;2 PG 16:1;0−20:1;0 PE 20:1;0−20:2;0 PE O− 18:2;0−20:0;0 5 PC O− 16:0;0−20:1;0 Cer 32:1;2 PC 16:1;0−20:1;0 Cer 40:2;2 PE O− 16:1;0−20:1;0 PS 18:1;0−22:1;0 PE 18:1;0-24:1;0 PS 16:1;0−20:1;0 PS 16:1;0−22:1;0 DAG 16:1;0−20:1;0 PE O− 16:0;0−20:1;0 PI 18:1;0−20:1;0 PI 18:2;0−20:1;0 PE 20:2;0−20:2;0 PI 16:1;0−19:1;0 PC 14:1;0−20:1;0 SM 40:2;2 PC 14:0;0−22:1;0 PI 20:1;0−20:2;0 PS 16:0;0−20:1;0 PI 16:1;0−20:1;0 PC 16:1;0−19:1;0 HexCer 40:2;2 HexCer 36:1;2 PE 16:1;0−20:0;0 SM 36:2;2 PE O- 18:1;0-20:1;0 SM 38:1;2 SM 36:1;2 PS 18:0;0−20:1;0 PS 20:1;0−20:3;0 SM 38:2;2 PC 17:1;0−20:2;0 Cer 38:2;2 PI 20:1;0−20:4;0 SM 32:2;2 SM 36:2;2 PI 17:1;0−20:1;0 PS 20:1;0−20:1;0 PI 17:0;0−20:2;0 PS 18:1;0−21:1;0 PC 18:2;0−18:2;0 Cer 34:1;2 LPE 18:2;0 PS 17:0;0−20:2;0 Cer 38:1;2 DAG 18:1;0−20:2;0 HexCer 38:1;2 DAG 16:0;0−20:1;0 PE O− 18:0;0−18:1;0 PS 18:1;0−20:0;0 PE O− 16:2;0−18:2;0 PC 14:0;0−19:1;0 DAG 18:0;0−20:4;0 PC 15:0;0−20:1;0 DAG 18:1;0−20:3;0 PC 18:2;0−20:1;0 SM 38:2;2 DAG 18:1;0−19:1;0 PC 14:0;0−20:0;0 LPE 20:1;0 PC 20:1;0−20:2;0 PC 20:2;0−20:2;0 PS 17:1;0−20:0;0 PS 17:1;0−22:1;0 TAG 55:3;0 PE 17:1;0−20:2;0 PG 18:0;0−18:0;0 PC 16:1;0−18:3;0 PE 16:1;0−22:1;0 PC 16:0;0−20:6;0 PE 12:0;0−18:1;0 PE 17:1;0−20:1;0 PC 18:0;0−20:1;0 PE O− 16:2;0−20:3;0 PC 17:0;0−20:4;0 PC O− 18:0;0−20:5;0 PE 14:0;0−20:1;0 LPE 16:2;0 CL 78:4;0 PC 12:0;0−18:2;0 PI 16:0;0−16:2;0 PS 18:0;0−22:5;0 PE O− 18:1;0−12:0;0 PE 16:1;0−22:2;0 PE O− 18:0;0−20:2;0 PC 16:1;0−22:1;0 PI 18:1;0−18:3;0 PC O− 16:2;0−18:2;0 PE O− 16:0;0−18:1;0 PE O− 18:2;0−19:3;0 PE 16:0;0−16:2;0 Cer 32:2;2 PE O− 17:0;0−16:2;0 PC 16:0;0−22:1;0 PE 16:0;0−16:0;0 PS 16:0;0−20:2;0 PI 18:0;0−21:3;0 PI 20:1;0−20:3;0 PC 18:1;0−22:2;0 PE 18:0;0−20:5;0 PI 16:0;0−18:3;0 PI 14:1;0−16:0;0 PI 14:0;0−16:1;0 Cer 44:2;1 LPC O− 16:1;0 DAG 16:0;0−16:2;0 PA 16:1;0−16:1;0 PC 14:1;0−18:2;0 -0.10 -0.05 0.00 0.05 0.10 CL 74:3;0 Legend PA 14:0;0−16:0;0 LPE 16:1;0 LPI 18:1;0 DAG 18:0;0−18:1;0 PA 18:0;0−18:1;0 PG 18:0;0−20:2;0 LPS 18:1;0 PA 14:0;0−18:1;0 PG 16:0;0−18:0;0 DAG 14:1;0−18:0;0 DAG 16:1;0−18:0;0 LPA 18:1;0 PA 16:0;0−16:1;0 DAG 14:0;0−18:1;0 DAG 18:1;0−18:1;0 CL 80:5;0 LPI 16:0;0 DAG 18:0;0−18:2;0 PC O− 17:0;0−16:0;0 PG 16:1;0−22:3;0 PC 16:0;0−22:3;0 PC O− 16:2;0−16:0;0 PC 15:0;0−17:0;0 PC O− 17:1;0−20:1;0 PC O− 17:0;0−20:2;0 LPE 24:4;0 LPE 14:1;0 PA 18:0;0−20:3;0 PC O− 18:2;0−17:0;0 PG 14:0;0−18:0;0 PS 17:0;0−18:0;0 LPE 24:5;0 Cer 32:0;2 PA 16:0;0−20:2;0 PC O− 16:0;0−17:0;0 PC O− 16:1;0−22:1;0 PC O− 17:0;0−22:2;0 PI 16:0;0−19:2;0 PI 16:0;0−21:3;0 PI 17:1;0−19:1;0 PE 20:1;0−20:3;0 PC 14:1;0−18:1;0 PC O− 16:2;0−18:0;0 PC O− 17:2;0−17:0;0 Cer 36:1;3 PI 14:0;0−20:1;0 PG 16:1;0−19:1;0 Control 48h PE O− 16:1;0−24:4;0 PC 12:0;0−16:0;0 PC O− 17:0;0−20:1;0 SM 30:1;4 PE O− 16:1;0−16:2;0 PA 18:0;0−18:2;0 PS 17:1;0−17:1;0 PC 17:0;0−20:1;0 PI 16:1;0−20:4;0 PC 16:1;0−20:4;0 PC O− 18:2;0−20:4;0 PC O− 17:2;0−16:1;0 PA 17:0;0−17:1;0 PC 18:1;0−18:3;0 PC 16:0;0−19:2;0 PC O− 17:0;0−19:2;0 PG 16:1;0−18:0;0 PS 18:0;0−22:1;0 0 PE 18:2;0−19:1;0 PC 18:0;0−22:5;0 PE O− 17:1;0−19:1;0 PI 20:2;0−20:2;0 PC 14:1;0−20:2;0 Contribution on component 2 CL 76:2;0 PE 16:1;0−22:0;0 PI 16:0;0−22:1;0 PI 20:0;0−20:3;0 PS 18:1;0−20:4;0 PS 20:0;0−20:3;0 PC O− 18:0;0−16:2;0 Cer 38:1;3 ELOVL1 48h PE 16:0;0−22:1;0 PE O− 16:2;0−22:3;0 PS 16:1;0−20:3;0 PS 18:2;0−18:2;0 PS 18:2;0−22:0;0 PS 20:0;0−20:4;0 PE O− 16:1;0−20:0;0 PC O− 16:2;0−20:3;0 PE 18:2;0−22:1;0 PI 18:2;0−18:2;0 CL 72:0;0 PC 14:0;0−20:4;0 PI 16:1;0−18:2;0 PE O− 17:1;0−16:0;0 PE O− 17:0;0−16:1;0 PE 18:0;0−22:1;0 PE O− 18:1;0−20:0;0 CL 74:6;0 PC 16:1;0−20:0;0 PC O− 17:1;0−19:1;0 LPE 18:1;0 PG 16:1;0−22:4;0 SM 42:2;3 DAG 15:0;0−15:0;0 LPA 16:1;0 DAG 16:0;0−17:0;0 PC 16:1;0−19:2;0 PE 18:2;0−24:0;0 PI 18:0;0−22:6;0 Cer 40:1;1 CE 18:2;0 CL 66:3;0 SLC27A1 48h PG 16:0;0-18:0;0 DAG 14:0;0−17:0;0 DAG 15:0;0−16:0;0 DAG 15:0;0−17:0;0 DAG 15:0;0−18:0;0 DAG 15:0;0−18:1;0 DAG 16:0;0−17:1;0 DAG 16:1;0−17:0;0 DAG 18:0;0−20:1;0 HexCer 34:0;2 PC 18:0;0-20:3;0 HexCer 42:1;3 PC 18:2;0−20:4;0 PC O− 16:0;0−19:2;0 PC O− 16:0;0−20:5;0 PC O− 17:1;0−19:2;0 PE 18:1;0−21:1;0 PE O− 16:0;0−17:1;0 PE O− 17:1;0−22:6;0 PE O− 17:2;0−20:3;0 PE O- 16:1;0-20:2;0 PI 16:0;0−22:6;0 PC O− 17:0;0−18:3;0 PI 18:1;0−19:1;0 TAG 54:1;0 DAG 14:1;0−16:0;0 PC 14:1;0−17:1;0 PI 16:0;0−22:4;0 DAG 14:0;0−16:0;0 PE O− 18:0;0−16:1;0 SM 34:0;2 PE O− 17:1;0−17:0;0 PE 16:0;0−22:4;0 PC O− 16:1;0−16:2;0 PG 18:0;0−20:6;0 SM 42:2;3 PC 16:2;0−17:1;0 PE 15:0;0−18:1;0 PE 16:0;0−17:1;0 PE 16:1;0−17:0;0 LPC 18:1;0 PE 18:1;0−19:2;0 PA 16:0;0−18:1;0 PC 17:0;0−17:1;0 PI 18:0;0−18:1;0 CL 70:6;0 CL 72:5;0 PS 16:0;0−18:0;0 PC 16:2;0−18:0;0 PC 18:1;0−24:1;0 PA 16:0;0−18:0;0 PI 14:0;0-18:2;0 PE 18:1;0−22:0;0 PS 18:1;0−22:0;0 PI 16:0;0−20:1;0 PC 14:1;0−16:1;0 PS 18:0;0−18:1;0 DAG 16:0;0−18:2;0 DAG 16:0;0−16:1;0 PA 16:1;0−18:0;0 DAG 16:0;0−18:1;0 PI 16:1;0-18:1;0 PA 16:1;0−18:1;0 PS 16:1;0−18:0;0 PI 16:0;0−18:2;0 PI 16:1;0−18:1;0 PE 16:2;0−18:1;0 PE 16:1;0−18:2;0 PC 16:0;0−16:1;0 PC 16:0;0−16:0;0 PE 16:1;0−16:1;0 PS 16:1;0-18:1;0 PE 16:1;0−18:1;0 PI 16:0;0−20:3;0 PC 14:0;0−14:1;0 DAG 14:1;0−18:1;0 DAG 16:1;0−16:1;0 PE 14:0;0−16:0;0 PE 18:1;0−20:5;0 PC O− 16:0;0−18:3;0 PE 18:2;0−20:4;0 LPE 18:2;0 PS 18:2;0−24:0;0 PA 14:1;0−18:1;0 DAG 14:0;0−20:2;0 PC 16:0;0−21:1;0 PC 16:2;0−18:2;0 PC 18:0;0−24:1;0 PC 18:2;0−20:3;0 PC O− 17:0;0−21:1;0 PE 17:1;0−18:2;0 PI 16:0;0-18:2;0 PE O− 18:1;0−19:2;0 PI 16:1;0−20:0;0 PS 16:0;0−22:0;0 PS 16:1;0−22:3;0 TAG 46:3;0 TAG 50:4;0 CL 72:2;0 PS 16:1;0−22:0;0 PC 14:0;0−16:2;0 CL 64:4;0 PS 14:0;0−18:1;0 PS 16:0;0−16:1;0 PE 14:1;0−16:0;0 PE 14:1;0−18:0;0 PI 16:0;0−18:1;0 PI 16:0;0−16:1;0 PI 16:1;0−18:0;0 PG 16:0;0−16:1;0 PI 14:1;0−18:0;0 PE 16:0;0−16:1;0 DAG 16:0;0-18:1;0 PI 14:0;0−18:1;0 PE 14:1;0−18:1;0 PI 16:1;0−16:1;0 PI 14:1;0−18:1;0 PC 16:0;0−24:1;0 PC 18:1;0−24:0;0 PE 16:1;0−22:3;0 PC 15:0;0−16:1;0 PC 14:0;0−17:1;0 LPC 16:1;0 PE 15:0;0−22:3;0 Component 2 (14%) PC O− 16:0;0−22:6;0 PE 17:1;0−21:3;0 PE 20:0;0−20:4;0 CL 74:9;0 PE 17:1;0−22:6;0 PC 17:1;0−19:1;0 PE 20:4;0−20:4;0 PE 17:0;0−20:2;0 PC O- 18:2;0-20:3;0 PG 16:0;0−20:3;0 PI 18:0;0−19:2;0 PI 16:0;0−18:0;0 LPI 20:3;0 PE O− 17:1;0−18:2;0 PI 18:0;0−22:4;0 PI 18:1;0−20:4;0 -5 PI 16:0;0−20:4;0 PS 18:0;0−22:6;0 PI 18:0;0-18:2;0 CL 68:3;0 PI 18:0;0−22:5;0 LPI 18:0;0 PI 18:0;0−22:3;0 PE O− 16:0;0−20:5;0 CE 20:1;0 PE O− 17:2;0−18:2;0 PE O− 17:0;0−18:3;0 PE 16:0;0−18:3;0 DAG 16:1;0-18:0;0 PE O− 16:1;0−19:3;0 PC 16:1;0−22:5;0 TAG 48:0;0 PI 17:1;0−18:0;0 PE 16:0;0−20:3;0 PC O− 17:0;0−16:1;0 TAG 56:1;0 PE O− 17:1;0−16:1;0 PE O− 18:1;0−18:2;0 TAG 52:2;0 PE O− 16:1;0−17:1;0 Cer 42:0;1 PE 18:1;0−22:6;0 PE 16:0;0−22:5;0 PE O− 17:0;0−18:2;0 CE 14:0;0 CE 16:1;0 CE 18:1;0 CE 16:0;0 PC 14:1;0−18:0;0 LPE 18:0;0 PC 16:0;0−20:3;0 PE 18:0;0−22:5;0 PI 18:0;0−20:4;0 PE O− 18:1;0−22:4;0 PE O− 18:1;0−20:5;0 PE 17:0;0−20:4;0 PE 18:0;0−20:4;0 PE 18:0;0−18:3;0 HexCer 42:2;2 PC O- 18:1;0-18:0;0 PE 18:2;0−18:2;0 PE O− 18:1;0−22:6;0 PS 17:1;0−18:0;0 PS 17:0;0−18:1;0 PE 17:1;0−18:1;0 PE O− 18:1;0−22:5;0 PE 18:0;0−18:1;0 PC 15:0;0−17:1;0 PE O− 18:1;0−20:4;0 TAG 54:3;0 PE O− 16:1;0−20:4;0 PE 18:1;0−22:5;0 PE 18:0;0−18:2;0 PE O− 17:1;0−20:4;0 PE O− 17:0;0−20:4;0 PE 17:0;0−18:1;0 PE O− 16:1;0−22:5;0 CL 68:2;0 CE 24:1;0 PI 16:1;0-20:2;0 PC 18:1;0−20:5;0 PE O− 16:1;0−22:6;0 PS 18:0;0−20:3;0 PC O− 17:0;0−15:0;0 PE 18:1;0−18:3;0 PE 16:0;0−20:4;0 PE 16:0;0−18:1;0 PE 16:0;0−18:2;0 PE 16:1;0−20:4;0 PE 18:1;0-19:2;0 PE O− 16:1;0−18:2;0 PC 15:0;0−15:0;0 PE 16:1;0−22:5;0 PE 17:0;0−18:2;0 PE 17:1;0−17:1;0 PE O− 17:0;0−20:3;0 PC O− 17:0;0−16:2;0 CL 72:7;0 PE 17:0;0−20:3;0 PI 18:1;0-20:3;0 PE O− 18:0;0−22:5;0 PE 16:0;0−19:1;0 PE 20:1;0−20:4;0 HexCer 40:1;2 PC 16:0;0−18:1;0 HexCer 42:1;2 PE 18:0;0−22:6;0 PC O− 17:0;0−17:1;0 PC 15:0;0−18:1;0 LPE 18:0;0 Cer 34:0;2 PE O− 18:2;0−20:5;0 PE O− 16:1;0−20:5;0 PE O− 16:1;0−22:4;0 PI 17:1;0−18:1;0 PE O− 16:0;0−20:4;0 PE 18:1;0−18:2;0 PC 16:0;0−17:1;0 PE 16:1;0−19:1;0 HexCer 34:1;2 PI 18:0;0-20:3;0 PI 18:0;0−20:1;0 PE O− 17:0;0−17:1;0 PE O− 18:2;0−17:1;0 PC 17:1;0−17:1;0 PE O− 18:0;0−20:3;0 PE O− 16:0;0−22:5;0 PE O− 17:1;0−22:5;0 PC 18:0;0−19:1;0 PE O− 18:0;0−20:4;0 PC 16:1;0-18:2;0 PC O− 18:2;0−18:0;0 PE 18:1;0−20:0;0 PC 17:0;0−18:2;0 PE O− 16:1;0−18:3;0 PC 16:0;0−16:2;0 PC 18:0;0−18:3;0 PC 15:0;0−18:2;0 PE 18:0;0−22:4;0 PE O− 17:1;0−17:1;0 PS 16:1;0-22:0;0 PC 18:1;0−22:4;0 PI 18:0;0−19:3;0 PI 16:0;0−19:1;0 PC 16:0;0−22:4;0 PS 18:0;0−19:1;0 PI 14:0;0−18:0;0 PI 17:0;0−20:3;0 PC O− 18:1;0−20:4;0 PC O− 16:0;0−18:2;0 PI 18:0;0-20:2;0 PC O− 16:1;0−20:4;0 PC O− 16:0;0−20:4;0 PC 18:0;0−20:4;0 PC O− 17:1;0−17:1;0 PE 16:1;0−17:1;0 PC O− 18:0;0−18:2;0 PE O− 16:0;0−22:6;0 PC O− 16:0;0−22:4;0 PC O− 18:0;0−20:6;0 PI 18:0;0-18:3;0 PE O− 16:1;0−20:3;0 PE O− 18:2;0−22:6;0 PC 18:0;0−18:2;0 PE O− 18:2;0−20:3;0 PC 17:1;0−18:1;0 PE O− 18:2;0−15:0;0 PE O− 18:1;0−20:3;0 PE 18:1;0−22:4;0 PE 18:0;0−21:3;0 PE O− 17:2;0−18:0;0 DAG 18:0;0-20:2;0 PI 17:0;0−18:1;0 PE O− 16:1;0−19:2;0 PI 16:0;0−22:5;0 PS 16:0;0−20:3;0 PE O− 16:1;0−24:5;0 PS 18:0;0−22:4;0 TAG 52:1;0 PC 16:0;0−22:5;0 PE O− 16:1;0−19:1;0 PE 18:1;0-18:1;0 PC 18:1;0−20:4;0 PE O− 17:1;0−18:1;0 PC O− 17:2;0−16:0;0 PC 17:0;0−18:1;0 PC 16:0;0−20:4;0 PC 18:1;0−22:5;0 PE O− 18:1;0−17:1;0 PC 14:0;0−17:0;0 PE O− 18:2;0−20:4;0 LPE 18:1;0 PE 18:0;0−20:3;0 PE O− 18:2;0−22:5;0 PE O− 17:2;0−18:1;0 PE O− 17:2;0−16:0;0 PE O− 18:0;0−18:2;0 PC O− 18:1;0−17:1;0 PC O− 17:0;0−18:2;0 PE 18:1;0−20:3;0 LPC 16:0;0 CE 16:0;0 LPC 16:1;0 LPE O− 18:1;0 -10 PC O− 16:0;0−17:1;0 SM 34:0;2 PC 14:0;0−18:0;0 PC 16:0;0−18:3;0 PC O− 17:2;0−18:0;0 PC O− 17:1;0−18:1;0 PC 17:1;0−18:0;0 DAG 18:1;0-18:1;0 LPC 18:1;0 LPC 18:0;0 PC 18:1;0−22:1;0 LPE 20:3;0 PC O− 18:1;0−18:2;0 PC O− 16:1;0−17:0;0 PE 18:0;0−22:3;0 PE O− 18:2;0−18:2;0 PC O− 16:0;0−22:5;0 PS 18:0;0-24:1;0 PE 14:0;0−18:0;0 PG 16:1;0−22:5;0 PE O− 18:1;0−18:3;0 PC 17:1;0−22:5;0 PE O− 16:0;0−20:3;0 PE 16:2;0−18:0;0 PI 18:1;0−22:5;0 PG 18:0;0−18:2;0 PE 18:1;0−18:1;0 CE 18:1;0 PC O− 17:1;0−16:2;0 PC 20:1;0−20:4;0 PC 17:0;0−20:6;0 PE O− 18:0;0−22:4;0 Cer 34:0;1 Cer 32:0;1 LPC 18:3;0 PC 18:0;0−22:1;0 PE 15:0;0−20:1;0 PG 17:0;0−18:1;0 PI 18:1;0-20:2;0 LPE 20:4;0 CE 18:0;0 PE O− 18:0;0−20:5;0 PA 17:1;0−17:1;0 PC O− 18:1;0−15:0;0 PC O− 16:1;0−17:1;0 PE O− 16:0;0−18:2;0 SM 42:1;2 HexCer 44:2;2 PI 16:0;0-20:2;0 PE 18:1;0−24:0;0 Cer 42:1;1 PE 18:1;0−19:1;0 PE 16:1;0−24:0;0 PE O− 16:2;0−20:4;0 PC O− 18:1;0−18:3;0 PC O− 17:1;0−15:1;0 -10 0 10 PC 15:1;0−16:1;0 PE O− 17:1;0−20:5;0 PI 16:2;0-18:0;0 PG 16:0;0−18:1;0 PG 18:0;0−18:1;0 PC O− 17:1;0−15:0;0 PE 16:1;0−18:0;0 PC 16:1;0−17:0;0 PE 16:1;0−20:3;0 PC O− 18:2;0−18:2;0 PC 15:0;0−16:0;0 PS 18:1;0−18:2;0 PI 18:2;0-20:3;0 PE O− 16:1;0−14:1;0 PS 16:0;0−18:2;0 PS 16:0;0−18:1;0 PC 16:0;0−18:2;0 PS 18:1;0−19:0;0 CL 72:6;0 PS 18:0;0−18:2;0 PI 18:0;0−20:5;0 CL 70:5;0 PC 17:1;0-18:0;0 PA 14:0;0−18:0;0 PA 16:0;0−16:0;0 PE 16:0;0−22:3;0 PC O− 17:0;0−15:1;0 PC O− 17:1;0−17:0;0 PC 15:1;0−16:0;0 HexCer 38:1;2 PE O− 17:0;0−19:1;0 PE 18:0;0−19:2;0 PI 14:0;0-18:0;0 SM 34:1;3 PE O− 18:0;0−20:1;0 SM 40:1;2 SM 34:1;2 PC 16:0;0−20:5;0 PC 15:1;0−18:1;0 PC 16:0;0−19:1;0 PC O− 17:0;0−19:1;0 SM 34:2;2 PC O− 17:2;0−17:1;0 PA 18:0;0-18:1;0 PE O− 17:1;0−20:3;0 PC 18:1;0−20:0;0 LPC O− 18:1;0 PE O− 18:2;0−22:4;0 PE 18:1;0−22:3;0 PE 16:0;0−20:1;0 PE 18:2;0−20:1;0 PE 18:1;0−22:1;0 PC 18:1;0−19:1;0 PC O- 17:2;0-16:1;0 PC 17:1;0−18:2;0 Component 1 (27%) PA 16:0;0-18:1;0 LPE O- 18:1;0 PC O- 17:1;0-16:2;0 DAG 18:0;0-18:1;0 PG 16:0;0-20:3;0

-0.15 -0.10 -0.05 0.00 0.05 0.10 0.15

Contribution on component 1 siELOVL1 48h siELOVL1 72h siELOVL1 24h siScramblesiSLC27A1 48h 48hsiSLC27A1siScramble 72h 72hsiSLC27A1siScramble 24h 24h PE O- 16:1;0-24:1;0 TAG 46:1;0 PE O- 18:2;0-20:2;0 TAG 48:1;0 PE 16:1;0-20:1;0 PE O- 16:1;0-16:0;0 B D F PI 18:1;0-20:1;0 PS 18:1;0-24:0;0 PE 20:1;0-20:1;0 TAG 56:3;0 Cer 34:2;2 PC 16:1;0-20:1;0 PS 18:1;0-24:1;0 Principal Component Analysis sparse Partial Least Squares regression PS 18:1;0-20:1;0 PE O- 18:1;0-20:1;0 PE O- 18:2;0-22:3;0 Cer 38:1;2 Cer 40:2;2 PE 20:1;0-20:2;0 PE O- 16:1;0-20:1;0 20 Cer 44:1;2 HexCer 44:2;2 10 PC O- 16:0;0-16:0;0 PE O- 16:1;0-14:0;0 PG 18:1;0-20:1;0 Cer 44:2;2 PC 14:0;0-20:1;0 SM 42:1;2 Cer 32:1;2 PS 20:1;0-20:3;0 PE 18:1;0-24:1;0 HexCer 40:2;2 PE 18:1;0-24:0;0 PE 18:1;0-20:1;0 TAG 60:3;0 PC 16:0;0-20:1;0 SM 44:2;2 SM 34:1;2 SM 40:2;2 5 SM 36:2;2 PC 18:1;0-20:1;0 PE O- 18:2;0-20:1;0 PS 20:1;0-20:1;0 SM 36:1;2 10 SM 34:1;3 Cer 38:2;2 Legend HexCer 38:1;2 SM 32:1;2 SM 38:2;2 siScramble 24h Legend SM 38:1;2 siScramble 48h 0 Control 72h -0.10 -0.05 0.00 0.05 0.10 siScramble 72h ELOVL1 72h siELOVL1 24h SLC27A1 72h Contribution on component 2 CELL BIOLOGY LPE O- 18:1;0 0 siELOVL1 48h PC O- 18:1;0-18:2;0 PE O- 16:2;0-18:0;0 HexCer 42:2;2 PA 14:0;0-18:1;0 siELOVL1 72h PC O- 18:2;0-16:1;0 PE O- 18:2;0-18:3;0 PC 14:1;0-18:1;0 -5 PC O- 18:0;0-18:2;0 siSLC27A1 24h TAG 56:5;0 Component 2 (16%) TAG 52:3;0 PC O- 17:1;0-17:1;0 PC O- 16:0;0-16:2;0 siSLC27A1 48h PE O- 18:2;0-17:1;0 DAG 18:1;0-20:2;0

Component 2 (11%) PA 16:1;0-18:0;0 PE O- 16:2;0-14:0;0 siSLC27A1 72h DAG 16:0;0-16:1;0 PI 18:0;0-21:3;0 DAG 14:0;0-18:1;0 PE 16:1;0-18:2;0 PE 16:1;0-20:3;0 PC O- 17:0;0-16:2;0 PE 16:0;0-20:4;0 TAG 54:4;0 -10 -10 PA 18:0;0-18:2;0 PE O- 16:0;0-18:2;0 PC O- 16:1;0-18:2;0 PC O- 16:1;0-20:2;0 PG 16:1;0-20:2;0 PS 18:0;0-22:4;0 PA 16:0;0-18:1;0 TAG 54:3;0 PC O- 16:0;0-20:3;0 PC O- 16:0;0-22:2;0 DAG 18:1;0-18:1;0 010 DAG 18:0;0-20:3;0 PC O- 17:2;0-18:1;0 PC O- 17:1;0-18:2;0 PE O- 18:1;0-22:2;0 PC 15:0;0-16:0;0 Component 1 (26%) PA 18:0;0-18:1;0 DAG 16:0;0-18:1;0 DAG 18:0;0-18:2;0 SM 34:0;2 DAG 16:1;0-18:0;0 DAG 16:0;0-20:2;0 PG 18:1;0-20:3;0 DAG 18:0;0-18:1;0 -20 -10 0 10 PC O- 18:1;0-16:1;0 Component 1 (30.6%) -0.15 -0.10 -0.05 0.00 0.05 0.10 0.15

Fig. 3. ELOVL1 or SLC27A1 knockdowns promote accumulation of specific lipid molecules. (A) Heat map representation and 2D clustering (Euclidean distance and complete-linkage clustering) of the Z-scores of all lipid species identified in the lipidomic analysis of differentiation-inducing knockdowns and control keratinocytes. (B) Sample variation along the first two principal components. (C and D) Sample separation along the first two components by sPLS of the 48 h (C) and 72 h (D) samples. (E and F) Contribution of the 50 most discriminant lipid species to the separation along the first and second components of sPLS for the 48 h (E) and 72 h (F) samples.

major alterations in keratinocyte differentiation (34). Recent known to be important in the biology of the upper epidermal studies have described a mutation in ELOVL1 in humans that layers (32). Our results are in line with an early report that a causes similar alterations in the lipid profile and is associated mixture of glucosylceramides can induce differentiation in fetal with dry skin and ichthyotic keratoderma, pointing to abnormal rat skin keratinocytes (41) and shed light on the individual mo- differentiation and therefore supporting our findings (35, 36). lecular species responsible. A more recent study (42) detailed FATP1, encoded by the SLC27A1 gene, is a fatty acid trans- lipid profile changes and alterations in keratinocyte differentia- porter that possesses a very long chain fatty acyl-CoA synthetase tion in patients with autosomal recessive congenital ichthyosis activity (37). It is expressed in the basal layer of adult human with mutations in the ceramide synthase 3 (CerS3) gene. These epidermis (38). FATP1 can be localized in the endoplasmic re- patients showed a deficiency in longer-chain (≥C26) ceramides. ticulum or in the plasma membrane (39), which might indicate a Interestingly, CerS3 can influence the activity of ELOVL1 by role in the intracellular distribution of lipids rather than their enhancing further elongation of C24 fatty acyl-CoA to C26/C28 abundance. This would be consistent with SLC27A1 knockdown fatty acyl-CoA (34). These observations point to a key role of affecting differentiation without strongly perturbating the overall the ≤C24/≥C26 ceramide (and possibly hexosylceramide) ratio − − lipid composition of keratinocytes. Slc27a1 / mice do not ex- in regulating keratinocyte differentiation. Building on these hibit skin phenotypes (39), and a loss-of-function mutation in studies, we now define a number of specific lipid subtypes that humans linked to Melkersson–Rosenthal syndrome does not are able to influence the fate of keratinocytes. Notably, these produce an epidermal phenotype (40). This potentially indicates molecules must operate within restricted chemical boundaries, as the existence of compensatory mechanisms to maintain skin the addition of structurally similar lipids to the growth medium homeostasis in vivo. was unable to significantly affect keratinocyte differentiation. A common set of individual lipid molecules was enriched in The amplitude of the effect of the bioactive lipids we describe cells stimulated to differentiate by suspension culture or by ei- may depend on several factors, such as how efficiently the lipid ther ELOVL1 or SLC27A1 knockdown. The bioactive lipid molecules are delivered to the appropriate subcellular destina- mediators we tested, ceramides and glucosylceramides, are tion or a dependency of the biological effect on the contextual

Vietri Rudan et al. PNAS | September 8, 2020 | vol. 117 | no. 36 | 22177 Downloaded by guest on September 28, 2021 A C ELOVL1 SLC27A1 knockdown knockdown

4 165 128 40

ll 30

26 16 /we 20

103 Colonies 10 B Suspension-induceduced 0 C14:0 Ceramide differentiation C24:1 GlucosylCeramide

1.5 **** ** 1.5 *

1.0 1.0 100 * (fold change of %)

) *** (fold change of %) 2 Involucrin-positive cells

0.5 Involucrin-positive cells 0.5 0.1 1 10 100 0.1 1 10 100 10 Concentration (µM) Concentration (µM) (mm C16:0 Ceramide C24:0 Ceramide *** 1.5 1.5 1 s s **** area y f%) o 0.1 olon

1.0 nge 1.0 a C h c (fold (fold change of %) Involucrin-positive cell Involucrin-positive cell 0.01 0.5 0.5 0.1 1 10 100 0.1 1 10 100 Concentration (µM) Concentration (µM) EtOH C18:0 GlucosylCeramide C16:0 2-OH Ceramide C16:1 Sphingomyelin C14:0 Cer C16:0 Cer C24:0 Cer 1.5 * 1.5 1.5 C18:0 GlcCer C24:1 GlcCer %) f o e

ng 1.0 1.0 1.0 a (fold ch (fold change of %) Involucrin-positive cells

0.5 Involucrin-positive cells 0.5 0.5 0.1 1 10 100 0.1 1 10 100 0.1 1 10 100 Concentration (µM) Concentration (µM) Concentration (µM)

Fig. 4. Specific lipid molecules can induce keratinocyte differentiation in culture. (A) Overlap between the discriminant lipid sets enriched during ELOVL1 knockdown, SLC27A1 knockdown, or suspension-induced keratinocyte differentiation. (B) Keratinocyte response to different doses of ceramides and glucosylceramides identified in the ELOVL1 intersection enriched lipid set (teal), in the SLC27A1 intersection lipid set (yellow), in the suspension-induced differentiation enriched lipid set (orange), or in no set (white box). Error bars indicate SDs, P values are calculated using one-way ANOVA with Dunnett’s multiple comparison test (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001), comparing each lipid treatment to vehicle-treated cells (1% ethanol, represented by a dashed line in the graphs). (C) Effect of bioactive lipid treatments on the colony formation ability of keratinocytes. Shown are the total number of colonies per well (dark purple bars) as well as the number of abortive colonies (magenta bars). Representative images for each treatment are shown below the bar plot. Error bars indicate SDs, P values are calculated using one-way ANOVA with Dunnett’s multiple comparison test (colony number) or the Kruskal–Wallis test with Dunn’s multiple comparison test (colony size) (*P < 0.05, ***P < 0.001).

lipid (and perhaps protein) composition of the cell (24). This cells and neighboring basal cells in culture were recently latter concept is consistent with the idea that lipids operate by reported (47). affecting the functionality of whole assemblies of molecular Whereas keratinocytes undergoing suspension-induced dif- partners within their localized environment, which makes their ferentiation lack cell–cell contacts, the siRNA experiments and effects challenging to evaluate using the more linear approaches lipid treatments were performed on adherent cells which have that are commonly employed when studying proteins (27). intercellular junctions. In the principal component analysis of the Our data offer insights into the influence of lipid composition lipidomic data the majority of the variance (component 1) sep- on the modulation of cellular behavior and represent a starting arated cells that had been disaggregated and immediately lysed platform to study the role of lipid species in the cellular biology (time 0) from cells that had been maintained in suspension of the epidermis at the molecular level. In this respect, while the (Fig. 1D), suggesting that intercellular adhesion may affect lipid regulatory potential of individual lipid subspecies on cellular composition. Within the epidermis, basal keratinocytes delami- function is largely unexplored, there are a number of interesting nate as they move into the suprabasal layers during differentia- possibilities. The Wnt pathway is a key regulator of stem cells in tion, and lipids could potentially play a role in this process by many, if not all, tissues, and Wnts require lipid modifications for altering the properties of the cell membrane or modulating the activity (43). Therefore, their action could be affected by direct activity of junctional proteins (48, 49). For example, the locali- interactions with key lipids within membranes or by changes in zation of the desmosomal cadherin desmoglein-1 to specific their local environment due to lipid variation. Lipid composition glycosylceramide-enriched membrane domains is essential for can affect the stiffness and mechanosensitivity of plasma mem- keratinocyte delamination during the early phases of differenti- branes (44–46), and differences in the stiffness of epidermal stem ation (49). Further analysis will be required to discover how the

22178 | www.pnas.org/cgi/doi/10.1073/pnas.2011310117 Vietri Rudan et al. Downloaded by guest on September 28, 2021 Table 1. Lipid molecules enriched both upon ELOVL1 or SLC27A1 knockdown and upon suspension-induced differentiation of primary human keratinocytes Knockdown Lipid species* Max fold change† Tested lipid Differentiation induction

siELOVL1 Ceramide – 32:1:2 (C14:0 Cer) 1.7 (72 h) C14:0 ceramide Yes siELOVL1 Ceramide – 34:1:2 (C16:0 Cer) 1.7 (48 h) C16:0 ceramide Yes siELOVL1 Ceramide – 34:2:2 (C16:1 Cer) 1.5 (72 h) —— siELOVL1 Ceramide – 38:1:2 (C20:0 Cer) 8.3 (72 h) —— siELOVL1 Ceramide – 40:1:2 (C22:0 Cer) 1.3 (72 h) C24:0 ceramide (accumulated in Yes suspension-differentiated cells) siELOVL1 Ceramide – 40:2:2 (C22:1 Cer) 1.7 (72 h) —— siELOVL1 Hexosylceramide – 34:1:2 1.5 (72 h) —— (C16:0 HexCer) siELOVL1 Hexosylceramide – 36:1:2 Accumulated‡ C18:0 glucosylceramide Yes (C18:0 HexCer) (48 and 72 h) ‡ siELOVL1 Hexosylceramide – 38:1:2 Accumulated —— (C20:0 HexCer) (48 and 72 h) siELOVL1 Hexosylceramide – 40:1:2 3 (72 h) —— (C22:0 HexCer) ‡ siELOVL1 Hexosylceramide – 40:2:2 Accumulated (C22:1 HexCer) (48 and 72 h) siELOVL1 Phosphatidylcholine – 17:0:0;17:1:0 3.7 (48 h) ——

siELOVL1 Phosphatidylcholine – 18:0:0;18:1:0 1.1 (72 h) ——CELL BIOLOGY siELOVL1 Phosphatidylcholine – 18:0:0;18:2:0 2 (72 h) —— siELOVL1 Phosphatidylcholine – 18:1:0;18:1:0 1.2 (72 h) —— siELOVL1 Phosphatidylcholine – 18:1:0;18:2:0 1.1 (48 h) —— siELOVL1 Phosphatidylcholine – 18:2:0;18:2:0 1.3 (48 h) —— siELOVL1 Phosphatidylethanolamine 1.05 (48 h) ether – 18:1:0;18:1:0 siELOVL1 Phosphatidylethanolamine 1.2 (72 h) —— ether – 18:1:0;20:3:0 siELOVL1 Phosphatidylethanolamine 1.2 (72 h) —— ether – 18:1:0;20:4:0 ‡ siELOVL1 Phosphatidylinositol – 16:1:0;20:4:0 Accumulated (48 h) —— siELOVL1 Phosphatidylserine – 14:0:0;18:1:0 1.8 (48 h) —— siELOVL1 Phosphatidylserine – 16:1:0;20:0:0 3.2 (48 h) siELOVL1 Phosphatidylserine – 18:1:0;20:0:0 1.9 (48 h) —— ‡ siELOVL1 Phosphatidylserine – 18:2:0;20:0:0 Accumulated (48 h) —— siELOVL1 Phosphatidylserine – 18:2:0;22:1:0 3.7 (72 h) —— siSLC27A1 Hexosylceramide – 42:2:2 1.1 (72 h) C24:1(Δ9) glucosylceramide Yes (C24:1 HexCer) siSLC27A1 Phosphatidic acid – 16:0:0;18:2:0 3.2 (48 h) —— siSLC27A1 Phosphatidylcholine – 16:1:0;16:1:0 1.1 (48 h) —— siSLC27A1 Phosphatidylcholine – 18:0:0;20:3:0 1.3 (48 h) —— siSLC27A1 Phosphatidylcholine ether – 16:1:0;16:1:0 1.2 (72 h) —— siSLC27A1 Phosphatidylcholine ether – 16:1:0;20:4:0 4.2 (72 h) —— siSLC27A1 Phosphatidylcholine ether – 18:1:0;16:1:0 1.3 (72 h) —— siSLC27A1 Phosphatidylcholine ether – 18:1:0;20:4:0 1.1 (72 h) —— siSLC27A1 Phosphatidylethanolamine 1.1 (72 h) —— ether – 16:1:0;20:3:0 siSLC27A1 Phosphatidylglycerol – 16:1:0;18:0:0 Accumulated‡ (48 h) —— siSLC27A1 Phosphatidylglycerol – 18:2:0;18:2:0 2.4 (72 h) —— siSLC27A1 Phosphatidylglycerol – 18:2:0;20:3:0 7 (72 h) —— siSLC27A1 Phosphatidylinositol – 16:2:0;18:0:0 Accumulated‡ (48 h) —— siSLC27A1 Phosphatidylinositol – 18:2:0;20:2:0 4 (48 h) —— siSLC27A1 Phosphatidylserine – 16:1:0;18:1:0 1.3 (48 h) —— ‡ siSLC27A1 Phosphatidylserine – 18:0:0;19:1:0 Accumulated (48 h) ——

*Lipids are annotated as “Class – Carbon chain length: Unsaturations: Hydroxylations.” For sphingolipids, the numbers refer to the total in the molecule; for phospholipids, individual lengths of the two fatty acid moieties are reported. The alternative notation used for sphingolipids refers to the carbon chain length of their fatty acid moiety and its number of unsaturations. †Fold changes are calculated based on the percentage picomole relative to total sample lipids. ‡ A numeric value could not be calculated due to undetectable levels in the control.

Vietri Rudan et al. PNAS | September 8, 2020 | vol. 117 | no. 36 | 22179 Downloaded by guest on September 28, 2021 cellular- and tissue-level architecture of the epidermis is im- Operetta High-Content Imaging System. For validation experiments, a sim- pacted by the lipid changes we have identified. In addition, lipids ilar protocol was performed, but using the ONTARGET plus siRNA series are important in inflammation (24, 50) and could, potentially, (Horizon Discovery). mediate aspects of epidermal immune signaling. In conclusion, we have systematically shown that the lipid Immunofluorescence. Antibody staining of cells was performed in the same composition of human keratinocytes critically impacts their dif- way as described in siRNA Screening and Transfection. Plates were washed once with PBS and fixed with 4% paraformaldehyde incubated for 10 min at ferentiation, down to the level of individual lipid subspecies. Our room temperature. Plates were then permeabilized by incubating with 0.2% comprehensive approach has allowed us to identify a set of Triton-X-100 in PBS for 5 min at room temperature, incubated with blocking specific lipid molecules that induce differentiation when added buffer (10% FBS, 0.25% fish skin gelatin in PBS) for 1 h and stained over- to primary keratinocytes in culture. Since it is not technically night at +4 °C with the primary antibody anti-involucrin (SY3 or SY7 clones) feasible to specifically target the basal layer of the epidermis for or anti-cleaved caspase 3 (Asp175) (Cell Signaling catalog no. 9661) diluted lipid delivery, either in in vitro reconstituted skin or in vivo, the to 1 μg/mL or according to manufacturer’s instructions in blocking buffer. physiological relevance of our findings remains to be explored Plates were then washed three times with PBS, stained with the secondary fully. It also remains technically challenging to dissect the exact antibodies Alexa Fluor 555 donkey anti-mouse (Thermo Fisher Scientific) mechanisms through which the lipid subspecies identified are and/or Alexa Fluor 488 donkey anti-mouse (Thermo Fisher Scientific) at 1 μg/mL, operating. Nevertheless, the results we report corroborate a the nuclear dye DRAQ5 (abcam) at 10 μM, and Alexa Fluor 647 Phalloidin number of recent studies (23, 24) in underlining the importance (Thermo Fisher Scientific) at 12.6 nM in blocking buffer. Secondary stains were and largely unexplored regulatory potential of individual lipid incubated for 2 h at room temperature protected from light, and plates subspecies in cellular function and lay the foundation for further were finally washed three times with PBS before being imaged using the exploration of the role of lipids in epidermal differentiation. Perkin-Elmer Operetta High-Content Imaging System. Materials and Methods High-Content Imaging Analysis. Images acquired with the Perkin-Elmer Op- eretta High-Content Imaging System were analyzed using custom algo- For a summary of materials and resources used, see SI Appendix, Table S4. rithms in the Perkin-Elmer Harmony high-content analysis software package (SI Appendix, Fig. S2D). Nuclei were initially defined using the DRAQ5 Primary Human Keratinocyte Culture. Primary male human keratinocytes 2 channel; small (<2,000 μm ) and highly irregular (roundness < 0.6) nuclei (strain Km) isolated from neonatal foreskin were cultured at 37 °C on mi- were excluded from the analysis. To minimize the misattribution of the cy- totically inactivated 3T3-J2 cells in complete F12/adenine/Dulbecco’s modi- toplasm areas, the level of cytoplasmic staining was inferred from the fied eagle medium (FAD) medium, containing one part Ham’s F-12, three fluorescence intensity in a ring around the nucleus, as involucrin staining parts Dulbecco’s modified eagle medium, 100 μM adenine, 10% (vol/vol) was homogeneous throughout the cytoplasm. Terminally differentiating fetal bovine serum (FBS), 0.5 μg/mL hydrocortisone, 5 μg/mL insulin, 0.1 nM cells were identified by manual thresholding of the involucrin perinuclear cholera toxin, and 10 ng/mL epidermal growth factor (EGF), as described fluorescence intensity. The complete Harmony image analysis sequence has previously (9). Primary keratinocytes were used in experiments at passages = 4–7. For knockdown or overexpression experiments, keratinocytes were been deposited in Open Science Framework (https://osf.io/wf7v2/?view_only grown in KSFM medium (Gibco) supplemented with 0.15 ng/mL EGF and 4b1f32c0e8d14c3fa13b95d51247e503). Data from both screening conditions 30 μg/mL bovine pituitary extract. When subculturing or seeding cells for an (Dataset S2) was transformed using modified Z-scores (calculated with the experiment, we filtered the disaggregated keratinocytes through a 40 μm sample median and median absolute deviation) and compiled into a single nylon strainer to remove cell clumps and large differentiated keratinocytes. data set. Fold changes were calculated with respect to negative control samples present in the same plate as each test sample. siRNA Screening and Transfection. We transfected primary human keratinocytes with a custom siRNA library targeting 258 lipid biosynthetic enzymes RNA Isolation, Complementary DNA Synthesis, and qPCR. RNA was extracted purchased from Dharmacon (siGENOME pools) (23). In addition, every plate from all samples using the RNeasy Mini kit (Qiagen) according to the man- contained two negative nontargeting siRNA control wells, two no-siRNA ufacturer’s instructions and subsequently reverse transcribed using the wells, and two involucrin-targeting siRNA wells. Nontargeting and no- QuantiTect Reverse Transcription kit (Qiagen) according to the manufac- siRNA wells were used as negative controls, while involucrin-targeting turer’s instructions. The complementary DNA was diluted to 5 ng/μL, and wells were used as positive controls for the estimation of transfection effi- specific targets were amplified by qPCR using the Fast SYBR Green Master ciency and validation of antibody specificity. The screening was performed Mix (Thermo Fisher Scientific). Primer details can be found in SI Appendix, in quadruplicate in two different culture conditions. Reverse transfection of Table S4. The amplification program used was as follows: 3 min incubation siRNAs was performed in accordance with the transfection reagent manu- at 95 °C followed by 40 cycles of denaturation for 3 s at 95 °C and annealing/ facturer’s instructions. siRNAs were diluted in KSFM medium and arrayed in extension for 25 s at 60 °C; melting curve analysis was also performed to 96-well plates (Greiner μClear). The siRNAs were incubated for 20 min with ensure amplifications were specific. Expression of all targets was normalized μ 10 L of INTERFERin-HTS transfection reagent (PolyPlus Transfection) diluted against the expression of three reference genes (RPL13A, ATP5B, and TBP) 1/200 in water. Approximately 8,000 keratinocytes were then seeded in ev- and against control sample expression (ΔΔCq). ery well in 125 μL complete KSFM medium. Final volume was 175 μL per well, and the final siRNA concentration was 30 nM. Four hours after transfection, Suspension-Induced Keratinocyte Differentiation. Keratinocytes were differ- the medium was changed to fresh complete KSFM with added 1% penicillin/ entiated in suspension as described previously (8, 9). Preconfluent cultures streptomycin. Two days after transfection, half the plates were switched to were disaggregated in trypsin/ethylenediaminetetraacetic acid and resus- complete KSFM + penicillin/streptomycin supplemented with 10% FBS to pended at a concentration of 105 cells/mL in medium containing 1.45% induce keratinocyte differentiation; in the remaining plates, the medium methylcellulose, supplemented with either 5 μM GF109203X (Tocris) to in- was refreshed. Ninety-six hours after transfection, the plates were washed hibit PKC or dimethylsulphoxide as a control. Aliquots were plated in six-well once with phosphate-buffered saline (PBS) and fixed with 4% paraformal- plates coated with 0.4% polyhydroxyethylmethacrylate; this ensured that dehyde incubated for 10 min at room temperature. Plates were then per- – meabilized by incubating with 0.2% Triton-X-100 in PBS for 5 min at room there was no cell substratum adhesion. The suspended cells were subse- temperature and incubated with blocking buffer (10% FBS, 0.25% fish skin quently incubated at 37 °C. At each collection time point, the methylcellu- gelatin in PBS) for 1 h and stained overnight at +4 °C with the primary an- lose was diluted with PBS, and the cells were recovered by centrifugation. tibody anti-involucrin (SY3 or SY7 clones) diluted to 1 μg/mL in blocking The experiment was performed in duplicate. buffer. Plates were then washed three times with PBS, stained with the secondary antibody Alexa Fluor 555 donkey anti-mouse (Thermo Fisher Sci- Clonogenicity Assays. Five hundred keratinocytes were plated on a 3T3 feeder entific), the nuclear dye DRAQ5 (abcam), and Alexa Fluor 647 Phalloidin layer per well of a six-well dish. After 12 d, feeders were removed, and (Thermo Fisher Scientific) at a concentration of 1 μg/mL, 10 μM, and 12.6 nM keratinocyte colonies were fixed in 4% paraformaldehyde (Sigma) for in blocking buffer, respectively. Secondary stains were incubated for 2 h at 10 min then stained with 1% rhodanile blue (1:1 mixture of rhodamine B room temperature protected from light, and plates were subsequently and Nile blue A [Acros Organics]). Colony number and size were scored washed three times with PBS before being imaged using the Perkin-Elmer using ImageJ.

22180 | www.pnas.org/cgi/doi/10.1073/pnas.2011310117 Vietri Rudan et al. Downloaded by guest on September 28, 2021 Lipid Treatment of Cultured Primary Human Keratinocytes. All lipid molecules R heatmap.2 function (package Rplots) with default settings. Samples in the were purchased from Avanti Polar Lipids (see SI Appendix, Table S4)in siRNA knockdown time course were analyzed by PCA using the mixOmics powder form. The lipids were dissolved in ethanol at a stock concentration package (v6.10.9) (54) with parameter scale = T after removal of near-zero of 10 mM and stored at −20 °C. In each experiment, 15,000 keratinocytes variance predictors. To maximize separation between the samples at 48 and were seeded per well of a 96-well plate and incubated overnight before 72 h posttransfection and identify discriminating lipid species, we employed adding the lipids. Upon treatment, the stock solutions were heated to 37 °C, sPLS-DA using the mixOmics package with the following parameters: diluted to the indicated concentrations in culture medium preheated to ncomp = 3, keepX = rep(250,3), near.zero.var = T. Samples in the suspension- 37 °C, and finally added to the monolayers. Vehicle concentration was 1%, induced differentiation time course were analyzed by PCA using the equal to the level present in the highest lipid concentration. mixOmics package with parameters ncomp = 3, scale = T after removal of near-zero variance predictors. To maximize separation between the samples Lipid Extraction for Mass Spectrometry (MS) Lipidomics. MS-based lipidomics and identify discriminating lipid species, we classified the samples according was performed by Lipotype GmbH (51). A two-step chloroform/methanol to the model detailed in Fig. 2 (9) and employed sPLS-DA using the mixOmics procedure was used to extract the lipids (52), and samples were spiked with package with the following parameters: ncomp = 3, keepX = rep(250,3), an internal lipid standard mixture. This standard comprised cardiolipin 16:1/ near.zero.var = T. Discriminating lipids in the siRNA knockdown time course 15:0/15:0/15:0 (CL), ceramide 18:1;2/17:0 (Cer), diacylglycerol 17:0/17:0 (DAG), experiment were identified as lipids enriched in siELOVL1 that were contrib- hexosylceramide 18:1;2/12:0 (HexCer), lysophosphatidate 17:0 (LPA), lyso- uting to the separation from siScramble along component 1 in the sPLS-DA at phosphatidylcholine 12:0 (LPC), lysophosphatidylethanolamine 17:1 (LPE), 48 or 72 h posttransfection (loading > 0.01) and lipids enriched in siELOVL1 or lysophosphatidylglycerol 17:1 (LPG), lysophosphatidylinositol 17:1 (LPI), siSLC27A1 that were contributing to the separation from siScramble along lysophosphatidylserine 17:1 (LPS), phosphatidate 17:0/17:0 (PA), phosphati- component 2 at 48 or 72 h posttransfection (loading > 0.01). Enrichment and dylcholine 17:0/17:0 (PC), phosphatidylethanolamine 17:0/17:0 (PE), phosphatidylglycerol fold change were determined by comparing the means of the replicates of 17:0/17:0 (PG), phosphatidylinositol 16:0/16:0 (PI), phosphatidylserine 17:0/17:0 each sample. Discriminating lipids in the suspension-induced differentiation (PS), cholesterol ester 20:0 (CE), sphingomyelin 18:1;2/12:0;0 (SM), triacylglycerol time course experiment were identified as lipids enriched in the “differenti- 17:0/17:0/17:0 (TAG), and cholesterol D6 (Chol). The organic phase was dried in ated” class samples (see Results) that were contributing to the separation a speed vacuum concentrator. The dried extract was resuspended in 7.5 mM from all other sample classes along component 2 in the sPLS-DA (loading < −0.01) ammonium acetate in chloroform/methanol/propanol (1:2:4, vol:vol:vol), dried or lipids that were enriched in “commitment” class samples (see Results) again, and then suspended in a 33% ethanol solution of methylamine in that were contributing to the separation from all other sample classes along chloroform/methanol (0.003:5:1, vol:vol:vol). Liquid handling was performed component 3 in the sPLS-DA (loading < − 0.01). Enrichment was determined on the Hamilton Robotics STARlet robotic platform with the Anti Droplet by comparing the median of each sample class. Control feature.

Data Analysis. Data are presented as mean and standard deviation (SD). MS Data Acquisition. Sample analysis (53) was performed on a Q Exactive

Statistical analysis was performed using Microsoft Excel and GraphPad Prism CELL BIOLOGY mass spectrometer (Thermo Scientific) with a TriVersa NanoMate ion source 8.0. Normality was assessed based on quantile–quantile plot inspection. See (Advion Biosciences). Analysis was performed in positive and negative ion modes (resolution = 280,000 at m/z = 200 for MS and resolution = 17,500 at figure legends for details of the statistical tests used. For the siRNA screen- = m/z = 200 for tandem MS [MSMS], where m/z indicates the mass-to-charge ing, n 4 replicates (independent transfections). For validation experiments, ≥ = ratio). The trigger for MSMS was an inclusion list of MS mass ranges scanned n 2. For the knockdown time course lipidomics, n 3 replicates (inde- in 1-Da increments. MS and MSMS data were combined to monitor CE, DAG, pendent transfections). For the suspension-induced differentiation lip- = ≥ and TAG ions as ammonium adducts; PC and PC O- as acetate adducts; and idomics, n 2 replicates. For lipid-induced involucrin expression, n 6 ≥ CL, PA, PE, PE O-, PG, PI, and PS as deprotonated anions. MS alone was used independent treatments. For lipid-induced apoptosis induction, n 3 inde- ≥ to monitor LPA, LPE, LPE O-, LPI, and LPS as deprotonated anions; Cer, pendent treatments. For colony formation assays, n 2 hexaplicates of in- HexCer, SM, LPC, and LPC O- as acetate adducts; and cholesterol as the dependent transfections or n = 12 independent treatments. ammonium adduct of an acetylated derivative. Data Availability. All study data are included in the article and SI Appendix. Data Analysis and Postprocessing. Data were analyzed with in-house-developed lipid identification software based on LipidXplorer (53). Data postprocessing ACKNOWLEDGMENTS. We thank all members of the F.M.W. laboratory for and normalization were performed using an in-house-developed data man- helpful discussions. We are grateful for funding from the Department of agement system. Lipids identified with a signal-to-noise ratio of >5anda Health via the National Institute for Health Research comprehensive Bio- ’ ’ signal intensity fivefold higher than in corresponding blank samples were medical Research Centre award to Guy s and St Thomas National Health Service Foundation Trust in partnership with King’s College London and analyzed further. King’s College Hospital NHS Foundation Trust. This work was funded by grants to F.M.W. from the UK Medical Research Council (Grant G1100073) Downstream Analysis of Lipidomics Data. Lipid amounts from the lipidomics and the Wellcome Trust (Grant 096540/Z/11/Z), from a grant to U.S.E. from experiments were normalized by the total amount of lipids in each sample. the Wellcome Trust (Grant 110060/Z/15/Z), and from a grant to F.M.W. The samples were compared by unsupervised hierarchical clustering using the from Unilever.

1. F. M. Watt, Mammalian skin cell biology: At the interface between laboratory and 11. P. M. Elias, Epidermal lipids, barrier function, and desquamation. J. Invest. Dermatol. clinic. Science 346, 937–940 (2014). 80, S44–S49 (1983). 2. K. A. U. Gonzales, E. Fuchs, Skin and its regenerative powers: An alliance between 12. K. R. Feingold, The outer frontier: The importance of lipid metabolism in the skin. stem cells and their niche. Dev. Cell 43, 387–401 (2017). J. Lipid Res. 50, S417–S422 (2009). 3. Z. Nemes, P. M. Steinert, Bricks and mortar of the epidermal barrier. Exp. Mol. Med. 13. M. Akiyama, Corneocyte lipid envelope (CLE), the key structure for skin barrier 31,5–19 (1999). function and ichthyosis pathogenesis. J. Dermatol. Sci. 88,3–9 (2017). 4. T. Hirsch et al., Regeneration of the entire human epidermis using transgenic stem 14. M. A. Lampe, M. L. Williams, P. M. Elias, Human epidermal lipids: Characterization and cells. Nature 551, 327–332 (2017). modulations during differentiation. J. Lipid Res. 24, 131–140 (1983). 5. Y. Barrandon, H. Green, Three clonal types of keratinocyte with different capacities 15. Y. Jia, Y. Gan, C. He, Z. Chen, C. Zhou, The mechanism of skin lipids influencing skin for multiplication. Proc. Natl. Acad. Sci. U.S.A. 84, 2302–2306 (1987). status. J. Dermatol. Sci. 89, 112–119 (2018). 6. P. H. Jones, F. M. Watt, Separation of human epidermal stem cells from transit am- 16. M. D. Resh, Covalent lipid modifications of proteins. Curr. Biol. 23, R431–R435 (2013). plifying cells on the basis of differences in integrin function and expression. Cell 73, 17. Y. Nishizuka, Intracellular signaling by hydrolysis of phospholipids and activation of 713–724 (1993). protein kinase C. Science 258, 607–614 (1992). 7. H. Green, Terminal differentiation of cultured human epidermal cells. Cell 11, 18. Y. A. Hannun, L. M. Obeid, Principles of bioactive lipid signalling: Lessons from 405–416 (1977). sphingolipids. Nat. Rev. Mol. Cell Biol. 9, 139–150 (2008). 8. J. C. Adams, F. M. Watt, Fibronectin inhibits the terminal differentiation of human 19. K. Iwabuchi et al., Role of ceramide from glycosphingolipids and its metabolites in keratinocytes. Nature 340, 307–309 (1989). immunological and inflammatory responses in humans. Mediators Inflamm. 2015, 9. A. Mishra et al., A protein phosphatase network controls the temporal and spatial 120748 (2015). dynamics of differentiation commitment in human epidermis. eLife 6, e27356 20. Y. Uchida, Ceramide signaling in mammalian epidermis. Biochim. Biophy. Acta 1841, (2017). 453–462 (2014). 10. P. W. Wertz, B. van den Bergh, The physical, chemical and functional properties of 21. A. Shevchenko, K. Simons, Lipidomics: Coming to grips with lipid diversity. Nat. Rev. lipids in the skin and other biological barriers. Chem. Phys. Lipids 91,85–96 (1998). Mol. Cell Biol. 11, 593–598 (2010).

Vietri Rudan et al. PNAS | September 8, 2020 | vol. 117 | no. 36 | 22181 Downloaded by guest on September 28, 2021 22. A. Nguyen, S. A. Rudge, Q. Zhang, M. J. Wakelam, Using lipidomics analysis to de- 39. M.-H. Lin, J. H. Miner, Fatty acid transport protein 1 can compensate for fatty acid termine signalling and metabolic changes in cells. Curr. Opin. Biotechnol. 43,96–103 transport protein 4 in the developing mouse epidermis. J. Invest. Dermatol. 135, (2017). 462–470 (2015). 23. G. E. Atilla-Gokcumen et al., Dividing cells regulate their lipid composition and lo- 40. X. G. Xu et al., Exome sequencing identifies FATP1 mutation in Melkersson-Rosenthal – calization. Cell 156, 428–439 (2014). syndrome. J. Eur. Acad. Dermatol. Venereol. 31, e230 e232 (2017). 24. M. S. Köberlin et al., A conserved circular network of coregulated lipids modulates 41. Y. Uchida, M. Iwamori, Y. Nagai, Activation of keratinization of keratinocytes ω innate immune responses. Cell 162, 170–183 (2015). from fetal rat skin with N-(O-linoleoyl) -hydroxy fatty acyl sphingosyl glucose 25. M. A. Surma et al., An automated shotgun lipidomics platform for high throughput, (lipokeratinogenoside) as a marker of epidermis. Biochem. Biophys. Res. Commun. 170, 162–168 (1990). comprehensive, and quantitative analysis of blood plasma intact lipids. Eur. J. Lipid 42. K.-M. Eckl et al., Impaired epidermal ceramide synthesis causes autosomal recessive Sci. Technol. 117, 1540–1549 (2015). congenital ichthyosis and reveals the importance of ceramide acyl chain length. 26. K.-A. L. Cao, D. Rossouw, C. Robert-Granié, P. Besse, A sparse PLS for variable selection J. Invest. Dermatol. 133, 2202–2211 (2013). when integrating omics data. Stat. Appl. Genet. Mol. 7, 35 (2008). 43. H. Clevers, K. M. Loh, R. Nusse, Stem cell signaling. An integral program for tissue 27. E. Muro, G. E. Atilla-Gokcumen, U. S. Eggert, Lipids in cell biology: How can we un- renewal and regeneration: Wnt signaling and stem cell control. Science 346, 1248012 – derstand them better? Mol. Biol. Cell 25, 1819 1823 (2014). (2014). 28. J. T. Connelly et al., Actin and serum response factor transduce physical cues from the 44. M. A. Ayee, I. Levitan, Paradoxical impact of cholesterol on lipid packing and cell microenvironment to regulate epidermal stem cell fate decisions. Nat. Cell Biol. 12, stiffness. Front. Biosci. 21, 1245–1259 (2016). 711–718 (2010). 45. L. O. Romero et al., Dietary fatty acids fine-tune Piezo1 mechanical response. Nat. 29. K. W. Mulder et al., Diverse epigenetic strategies interact to control epidermal dif- Commun. 10, 1200 (2019). ferentiation. Nat. Cell Biol. 14, 753–763 (2012). 46. M. Saeedimasine, A. Montanino, S. Kleiven, A. Villa, Role of lipid composition on the 30. F. M. Watt, Involucrin and other markers of keratinocyte terminal differentiation. structural and mechanical features of axonal membranes: A molecular simulation J. Invest. Dermatol. 81 (suppl. 1), 100s–103s (1983). study. Sci. Rep. 9, 8000 (2019). 31. C. Özbalci, E. M. Storck, U. S. Eggert, RNAi transfection results in lipidome changes. 47. S. A. Mobasseri et al., Patterning of human epidermal stem cells on undulating Proteomics 19, e1800298 (2019). elastomer substrates reflects differences in cell stiffness. Acta Biomater. 87, 256–264 32. S. Hamanaka et al., Human epidermal glucosylceramides are major precursors of (2019). stratum corneum ceramides. J. Invest. Dermatol. 119, 416–423 (2002). 48. S. N. Stahley et al., Desmosome assembly and disassembly are membrane raft- 33. Y. Ohno et al., ELOVL1 production of C24 acyl-CoAs is linked to C24 sphingolipid dependent. PLoS One 9, e87809 (2014). synthesis. Proc. Natl. Acad. Sci. U.S.A. 107, 18439–18444 (2010). 49. O. Nekrasova et al., Desmosomal cadherin association with Tctex-1 and cortactin- 34. T. Sassa et al., Impaired epidermal permeability barrier in mice lacking elovl1, the Arp2/3 drives perijunctional actin polymerization to promote keratinocyte delamination. Nat. Commun. 9, 1053 (2018). gene responsible for very-long-chain fatty acid production. Mol. Cell. Biol. 33, 50. C. Albanesi, S. Madonna, P. Gisondi, G. Girolomoni, The interplay between kerati- 2787–2796 (2013). nocytes and immune cells in the pathogenesis of psoriasis. Front. Immunol. 9, 1549 35. A. Kutkowska-Kazmierczak et al., Dominant ELOVL1 mutation causes neurological (2018). disorder with ichthyotic keratoderma, spasticity, hypomyelination and dysmorphic 51. J. L. Sampaio et al., Membrane lipidome of an epithelial cell line. Proc. Natl. Acad. Sci. – features. J. Med. Genet. 55, 408 414 (2018). U.S.A. 108, 1903–1907 (2011). 36. N. Mueller et al., De novo mutation in ELOVL1 causes ichthyosis, acanthosis nigricans, 52. C. S. Ejsing et al., Global analysis of the yeast lipidome by quantitative shotgun mass hypomyelination, spastic paraplegia, high frequency deafness and optic atrophy. spectrometry. Proc. Natl. Acad. Sci. U.S.A. 106, 2136–2141 (2009). J. Med. Genet. 56, 164–175 (2019). 53. I. Mitroulis et al., Modulation of myelopoiesis progenitors is an integral component of 37. M.-H. Lin, D. Khnykin, Fatty acid transporters in skin development, function and trained immunity. Cell 172, 147–161.e12 (2018). disease. Biochim. Biophys. Acta 1841, 362–368 (2014). 54. F. Rohart, B. Gautier, A. Singh, K.-A. Lê Cao, mixOmics: An R package for ’omics 38. M. Schmuth et al., Differential expression of fatty acid transport proteins in epidermis feature selection and multiple data integration. PLoS Comput. Biol. 13, e1005752 and skin appendages. J. Invest. Dermatol. 125, 1174–1181 (2005). (2017).

22182 | www.pnas.org/cgi/doi/10.1073/pnas.2011310117 Vietri Rudan et al. Downloaded by guest on September 28, 2021