In silico analysis of imprinted gene expression in the mouse skin

Imprinted genes help mediate embryonic cell proliferation and differentiation, but their roles after birth are far less well understood. A subset of 16 imprinted gene network (IGN) genes is expressed at higher levels in stem cell progenitor cells of adult skeletal muscle and epidermis compared to their differentiated counterparts. While these genes function in muscle regeneration, their role in the skin is poorly understood. We assessed the expression profiles of these 16 IGN genes in publicly available datasets and revealed elevated expression of IGN genes in the telogen and early anagen phases in mouse skin. We also identified IGN genes among a list of previously identified hair cycle-associated genes. Furthermore, our results suggest that IGN genes form part of a larger network and function predominantly as upstream regulators of hair cycle-regulated genes. Based on these in silico data, we propose a potential novel role of these 16 IGN genes as upstream regulators of hair cycle-associated genes. We speculate that IGN gene dysregulation participates in syndromes characterized by an impaired hair cycle. Thus, IGN gene expression might serve as a point of therapeutic intervention for patients suffering from cutaneous pathologies such as common hair-loss disorders.


14
Imprinted genes help mediate embryonic cell proliferation and differentiation, but their roles 15 after birth are far less well understood. A subset of 16 imprinted gene network (IGN) genes is 16 expressed at higher levels in stem cell progenitor cells of adult skeletal muscle and epidermis 17 compared to their differentiated counterparts. While these genes function in muscle regeneration, 18 their role in the skin is poorly understood. We assessed the expression profiles of these 16 IGN 19 genes in publicly available datasets and revealed elevated expression of IGN genes in the telogen 20 and early anagen phases in mouse skin. We also identified IGN genes among a list of previously 21 identified hair cycle-associated genes. Furthermore, our results suggest that IGN genes form part 22 of a larger network and function predominantly as upstream regulators of hair cycle-regulated 23 genes. Based on these in silico data, we propose a potential novel role of these 16 IGN genes as 24 upstream regulators of hair cycle-associated genes. We speculate that IGN gene dysregulation

29
Genomic imprinting is an epigenetic regulatory mechanism that confers expression of selected 30 genes from one parental allele, and thus, is independent of classical Mendelian inheritance 1 . 31 Genomic imprinting is established in parental germline cells, and is maintained throughout  Imprinted genes are functionally distinct, but most are involved in controlling the transition of 41 cells between their quiescent, proliferative and/or differentiated states during fibroblast cell cycle 42 withdrawal, adipogenesis in vitro, and muscle regeneration in vivo 1 . These genes function 43 cooperatively in the regulation of specific biological pathways by forming co-expressed 44 networks 1 . One of these subnetworks comprises 16 imprinted genes (hereafter referred to as the expressed at high levels during embryonic and early postnatal life, but are silenced in the adult, 58 except in muscle satellite cells, hematopoietic stem cells and skin stem cells 5,7 . In mouse skin 59 tissue, it has been reported that Cdkn1c, Dlk1, Grb10, H19, Igf2, Mest, Ndn, Peg3 and Plagl1 are 60 expressed at higher levels in epidermal stem cells compared to those in non-stem cells 61 (keratinocytes) 7 . While imprinted genes have been shown to play a role in muscle regeneration 62 and hematopoiesis, their functions in skin tissue are poorly understood 8,9 ; thus, this was the 63 focus of our study. 64 Skin stem cells are multipotent adult stem cells that can self-renew and differentiate into multiple 65 cell lineages to form the different layers of the skin as well as the hair follicle. The cyclic activity 66 of hair follicles organizes the growth and renewal of hair. During its life span, hair undergoes 67 growth, degeneration and regeneration in concert with the activation and quiescence of 68 epidermal stem cells located in the bulge of the hair follicle 10,11 . The cyclic activity of hair 69 growth is divided into the anagen (growth), catagen (regression), and telogen (resting) phases 12 .

70
Follicular stem cells are maintained in a quiescent state during the telogen phase. Once activating 71 signals are received from upstream regulatory systems, a new cycle of hair growth is initiated 72 (anagen phase) 11,13,14 . After the active growth phase, proliferating matrix cells in the hair 73 follicles are induced to undergo coordinated apoptosis (catagen phase) 12 . Following the catagen 74 phase, the hair follicles eventually undergo transition to the telogen phase, during which hairs are 75 no longer produced due to inactivation of the follicular stem cells 12 .

76
In this study, we hypothesized that IGN genes play important roles in skin/hair biology 77 throughout life (after birth) in addition to their well-known activity during embryonic/fetal 78 growth. Thus, we aimed to identify the potential function of the 16 IGN genes in mouse skin  To study the expression of IGN genes in mouse skin, we searched the NCBI GEO databank for a 94 dataset that includes untreated, unaffected mouse skin samples. Using the search term 'skin AND 95 C3H/HeJ', we identified dataset GSE45513 which contains three samples of skin transcription 96 profiles from 10-week-old C3H/HeJ mice. Analysis of GSE45513 with the webtool GEO2R 97 revealed expression of all 16 IGN genes in the mouse skin samples (Fig. 1).

98
To identify publicly available datasets in which members of the IGN could be examined in the 99 skin, we searched for one IGN gene H19 using the search term 'H19[gene symbol], AND skin' 100 in the NCBI GEO Profiles database 17 . In this search, 156 datasets were identified. These datasets 101 were manually curated for differential expression of H19 across all samples within a dataset   For comparison, we also examined the absolute expression of six known telogen-activated genes 128 (Ar, Esr1, Lhx2, Nr1d1, Sox18, and Stat3), and six known telogen-repressed genes (Elf5, Foxn1, 129 Grhl1, Lef1,Msx2,and Vdr) 18 . We observed that the median IGN gene expression in the telogen 130 phase was elevated compared to that in the mid-anagen phase. This trend was less marked for 131 Dcn and Igf2r. Some of the IGN genes (i.e., Gnas, H19, Meg3 and Plagl1) were expressed at 132 even higher levels than the known telogen-activated genes (Fig. 3A). Igf2r) did not follow this trend (Fig. 3B).

141
In summary, our analysis shows that like H19, IGN genes are, in general, expressed at higher 142 levels in the telogen phase compared to those in the anagen phase. cycle-associated genes (hereafter referred to as the Lin1-dataset; Table S1) 21 . The P-value cut-158 off for the F-test was set previously to 0.05, as it was found that >80% of known genes 159 exhibiting hair cycle-dependent expression had a P-value of <0.05 determined using this 160 computational approach 21 . As the pool of these 2,289 hair cycle-associated genes was restricted Gatm, Ndn, and Slc38a4) identified in our study were grouped in the 'anti-hair growth' category 172 and showed a decline in expression levels during the anagen phase. Igf2r was categorized as a 173 'hair growth' gene, with peak expression early in the anagen phase. In contrast, Sgce was 174 categorized as a 'catagen-related' gene, with a decrease in expression during the catagen phase.

175
Only one IGN gene (Dcn) belonged to a gene cluster that could not be categorized according to 176 the three main profile patterns (Table 1). 177 Next, we analyzed an independent dataset (previously reported by Lin et al. and hereafter 178 referred to as the Lin2-dataset; Table S2), which comprises a set of 6,393 mRNA probe sets and   Table S1 as input to identify potential upstream regulators of these genes. We   Fig. 5). In the IPA, this network of genes was predicted to be associated with 232 'Dermatological Diseases and Conditions' in the 'Top Diseases and Functions' category. To 233 explore potential functions of this network further, we sought to determine whether these 35 234 genes were among the hair cycle-associated genes listed in Table S1. Our analysis revealed that 235 10 genes from this network had previously been identified as hair cycle-associated genes (Table   236   S4). Further analysis revealed that 27 of the 35 genes in the larger IGN network were among the 237 potential upstream regulators of the hair cycle-associated genes listed in Table S3 (Table S4 and (Table 3).  Table S3 and Fig. 5). Exogenous administration of its gene product, decorin, 276 was shown to accelerate the anagen phase and delay catagen phase transition, and was 277 categorized as a positive regulator of the hair growth cycle 34 . It has also been shown that 278 physiologic concentrations of Igf2, which is among our predicted upstream regulators of hair 279 cycle-associated genes (Table S3) (Table S3, S4 and highlighted in red in Fig. 5). Among the genes predicted to be directly 285 regulated by the IGN, we identified several known to be associated with regulation of the hair 286 cycle, including Akt, Erk, Pi3k, and Vegf (Fig. 5). Indeed, the PI3K-Akt signaling pathway plays including short anagen hair syndrome 54 , which is associated with a synchronized pattern of scalp 312 hair growth, 55 and androgenic alopecia, which is a very common type of hair-loss 32 . As H19 313 overexpression was shown to activate Wnt signaling to maintain the hair follicle regeneration 314 potential, it has already been suggested that H19 could be a target for treatment of androgenetic 315 alopecia 27 .

316
The limitations of our study should be noted. Our study was designed to examine the potential 317 function of the 16 IGN genes in mouse skin tissue after birth by reanalyzing publicly available 318 transcription profiles. Thus, this study consists of a in silico analysis of these 16 IGN genes, but 319 does not include wet-lab characterization and functional analysis. Not only is the hair follicle a 320 complex mini-organ that presents some challenges to wet-lab investigations, but the   performed and dataset quality was assessed by the generated graphical plots provided in GEO2R.

379
In brief, these comprised: 1) a volcano plot, generated using limma, displaying statistical reference set and the 'Upstream Regulator' analytics. Similarly, the network analysis was 437 generated using the 16 IGN genes as input data, performing a 'Core analysis' and using the 438 'Networks' analysis tool option in IPA. Using the same Core analysis of the 16 IGN genes, we 439 also performed a canonical pathway analysis using the 'Canonical Pathways' tab in the IPA 440 application. The resulting -log(P-values) were calculated by the IPA software using Fisher's 441 exact test to determine the probability that the association between the IGN genes and the 442 identified canonical pathways is due to chance alone. A -log(P-value) of ≥1.3 was considered 11. Data availability 619 The datasets underlying the results are available in the NCBI GEO DataSets (GSE11186) at 620 ncbi.nlm.nih.gov/gds/ as well as in the supplemental material of references 18 and of 21 .     Insulin-like growth factor 2 Expression peaks at early anagen phase Mest Mesoderm-specific transcript Expression peaks at early anagen phase Peg3** Paternally expressed gene 3 Expression peaks at early anagen phase Plagl1 PLAG1-like zinc finger 1 Expression peaks at early anagen phase * Defined by Lin et al. 18 , ** Gene was also identified as a potential hair cycle-associated gene in 673 the Lin1-dataset (see Table 1)