Epigenomic profiling of stem cells within the pilosebaceous unit identifies PRDM16 as a regulator of sebaceous gland homeostasis

The epidermis consists of different compartments such as the hair follicle (HF), sebaceous gland (SG) and interfollicular epidermis (IFE), each containing distinct stem cell (SC) populations. However, with the exception of the SCs residing within the HF bulge, other epidermal SC populations remain less well understood. Here we used an epigenomic strategy that combines H3K27me3 ChIP-seq and RNA-seq profiling to identify major regulators of pilosebaceous unit (PSU) SC located outside the bulge. When applied to the bulk of PSU SC isolated from mouse skin our approach identified both previously known and potentially novel non-bulge PSU SC regulators. Among the latter, we found that PRDM16 was predominantly enriched within the Junctional Zone (JZ), which harbors SC that contribute to renewal of the upper HF and the SG. To investigate PRDM16 function in the PSU SC, we generated an epidermal-specific Prdm16 Knock-out mouse model (K14-Cre-Prdm16fl/fl). Notably, SG homeostasis was disturbed upon loss of PRDM16 resulting in enlarged SGs, and excessive sebum production, resembling some of the features associated with human acne and sebaceous hyperplasia. Importantly, PRDM16 is essential to shut down proliferation in differentiating sebocytes. Overall, our study provides a list of putative novel regulators of PSU SC outside the bulge and identifies PRDM16 as a major regulator of SG homeostasis.


Introduction
The JZ connects the HF with the SG, an important skin appendage that contains lipid-  We previously showed that, by combining H3K27me3 ChIP-seq and RNA-seq profiling, 126 the main regulators of the different cell types found within a heterogeneous population 127 5 can be identified by those genes that, although highly expressed, are embedded within 128 broad H3K27me3 domains (Rehimi et al., 2016;Schertel et al., 2015;Shim et al., 2020). 129 Using this simple epigenomic approach to dissect epidermal SC heterogeneity, here we 130 identified several putative regulators of non-bulge SCs. Among them, we found that

155
To identify important regulators of HFSC populations residing outside the bulge, a 156 differential functional heterogeneity score (dFH score) (   Genes predicted as non-bulge SC regulators display spatially restricted 205 expression within the pilosebaceous unit. 206 We noticed that among the top 200 dFH score genes, there were several that were not  Moreover, we also observed the expression of DLK1 in lower bulge SC. (P33, especially in the JZ (Fig 2A-B). Moreover, in the tail skin of P56 mice, both 258 proteins were also co-expressed in the HG and the SGs (Fig 2A), although HG 259 expression might be artifactual as it was not reproducibly observed among replicates 260 ( Fig 1D). Overall, these results show that PRDM16 represents a novel marker for the JZ  sebum-producing sebocytes accumulate (Fig 4B). Co-staining of SCD1 and Ki67 340 confirmed that, upon loss of PRDM16, differentiated sebocytes still proliferate (Fig 4D).

341
These results suggest that seboctye proliferation and differentiation, two processes that but also frequently observed among lipid-producing sebocytes located in the SG interior.

424
Therefore, PRDM16 seems to be necessary to switch off proliferation in differentiating  The dFH score was computed as follows: 551 • H3K27me3B g is the breadth of the H2K27me3 mark in Αlpha6 + /Sca1cells at the gene 552 g. We added 1 to the H3K27me3 breadth to avoid mathematical issues with zeros.  All mouse genes and their corresponding dFH scores are listed in Data S1.  (B) H3K27me3 and H3K4me3 levels at the indicated regions were measured by ChIP-qPCR analysis using α6 +ve /Sca1 -ve cells isolated from P56 adult mice skin. Chr2 Neg corresponds to an intergenic region in mice chromosome 2 and serves as a H3K27me3 negative control; the promoter of Sox2, which is inactive in α6 +ve /Sca1 -ve HFSC, is expected to be marked by H3K27me3 but not H3K4me3; the promoter of Eef1a, a housekeeping gene active in α6 +ve /Sca1 -ve PSU SC, is expected to be marked by H3K4me3 but not by H3K27me3.

dFH g = (H3K27me3B g +1) x log 2 (DE g _FoldChange) x -log(DE g _qValue)
(C) ChIP-seq (H3K4me3 and H3K27me3) and RNA seq profiles generated from α6 +ve /Sca1 -ve stem cells around the HOX locus. The HOXA cluster can be clearly divided in active (high RNA-seq and high H3K4me3 levels) and inactive (low RNA-seq and high H3K27me3) domains, which illustrates the quality of the ChIP-seq data generated from sorted α6 +ve /Sca1 -ve PSU SC.