Normal human neonatal dermal fibroblasts (PromoCell GmbH, Heidelberg, Germany) were cultured in Dulbecco’s modified Eagle’s Medium (DMEM), 1 g/L glucose (Lonza, Rockland, ME, USA) supplemented with 10% decomplemented fetal calf serum (Lonza), 2 mM of L-glutamine (Lonza) and 0.1 mg・ml⁻¹ Primocin™ (Invivogen, San Diego, CA, USA). Fibroblasts were cultured at 37˚C in a humidified atmosphere containing 5% CO₂. The tested ingredient (Ashland Specialty Ingredients, Wilmington, USA) was used in our experiments at the dose of 1% (1:100 dilution v:v).

The tested ingredient is a complex of hydrolyzed natural yeast proteins (Saccharomyces cerevisiae) and hydrolyzed natural soya proteins with a total dry matter weight of 7 g/kg. After screening several botanical and yeast extracts and their combination, this ingredient complex was selected for its ability to modulate the expression of telomere-binding proteins such as TRF2 [9] - [11] in ex vivo human skin, human primary keratinocytes and human neonatal fibroblasts in culture. This ingredient is described in patent FR12 00814.

The list of microRNAs with a described implication in human dermal homeostasis, fibroblast senescence and regulation of dermal extracellular matrix content was established according to the literature (Table 1). Twenty- one microRNAs were hence selected by compiling the microRNAs described in the literature: miR-10b-5p, miR-1185-5p, miR-1204, miR-143-3p, miR-145-5p, miR-146a-5p, miR-155-5p, miR-199a-3p, miR-21-5p, miR-23a-3p, miR-26a-5p, miR-29a-3p, miR-30a-5p, miR-335-3p, miR-34a-5p, miR-519c-3p, miR-542-5p, miR-638, miR-658, miR-663 and miR-7-5p.

Bioinformatics was used to obtain a list of predicted and experimentally verified target genes for the selected microRNAs in this study. Two types of analyses were conducted.

i) The first approach used miRWalk, a comprehensive database providing the largest available collection of predicted and experimentally verified (validated) microRNA-target interactions [13] . Each microRNA that was modulated in our experiments was searched in the “validated targets module” of miRWalk in order to retrieve a list of validated targets that was further processed to retain only the genes involved in extracellular matrix homeostasis and fibroblast senescence.

ii) The second approach used Ingenuity® Pathway Analysis (IPA) (Qiagen, Redwood City, CA, USA) to possibly identify new interactions that could be complementary to the ones identified through the miRWalk database. IPA predicts microRNA interactions with target genes, based on the aggregation of data from TarBase (experimentally validated data), miRecords, and TargetScan (predicted microRNA-mRNA interactions) [14] - [16] . The resulting lists of genes established by these two approaches are presented in Table 2.

Cells were grown in 100-mm culture dishes and treated or not with 1% tested ingredient in the culture medium for 48 h. The culture medium was then removed and the cells were washed with cold PBS. Total RNA including microRNAs was extracted from each sample using the mirVana miRNA Isolation Kit (Ambion, Austin, TX, USA) and collected in 100 µl of RNase-free water. RNA quantity, quality and purity were determined using a NanoDrop™ 2000 spectrophotometer (Thermo Scientific, Waltham, MA, USA) by measuring the ratio of absorbance at 260 and 280 nm (A260/A280) and 260 and 230 nm (A260/A230). For each sample (untreated and treated), 10 ng of total RNA were reverse-transcribed by using the Universal cDNA Synthesis Kit and miRCURY LNA™ (Exiqon, Vedbaek, Denmark) on an MJ MiniCycler™ thermal cycler (MJ Research, Waltham, MA, USA). Real-time qPCR to amplify the 21 selected microRNAs was performed on a StepOnePlus™ thermal cycler (Applied Biosystems, Foster City, CA, USA) using the Pick-&-Mix microRNA PCR Panel with custom selection of LNA™ microRNA primer sets (Exiqon) and ExiLENT SYBR® Green Master Mix (Exiqon), according to the manufacturer’s recommendations.

The references of the microRNA LNA™ PCR primer sets (Exiqon) used in this study are: miR-10b-5p (ref. 205637), miR-1185-5p (ref. 205926), miR-1204 (ref. 204123), miR-143-3p (ref. 205992), miR-145-5p (ref. 204483), miR-146a-5p (ref. 204688), miR-155-5p (ref. 204308), miR-199a-3p (ref. 204536), miR-21-5p (ref. 204230), miR-23a-3p (ref. 204772), miR-26a-5p (ref. 204772), miR-29a-3p (ref. 204698), miR-30a-5p (ref. 205695), miR-335-3p (ref. 205613), miR-34a-5p (ref. 205613), miR-519c-3p (ref. 204016), miR-542-5p (ref. 204198), miR-638 (ref. 204234), miR-658 (ref. 204234), miR-663 (ref. 205877), and miR-7-5p (ref. 205727). Let-7a-5p (ref. 205727) was used as the reference for analysis. The comparative Ct method was used for relative quantification of microRNA expression.

An interaction network was built using Ingenuity® Pathway Analysis (Qiagen). This network included the 18 microRNAs that showed a modulation in expression after treatment with the ingredient, and the 25 target genes described in the miRWalk database (Table 1). Around these microRNAs and genes, an interaction network was developed: the network components were grouped according to their biological activity (e.g., collagen genes, metalloproteinases, transcription factors such as SOXs, FOXO3 and sirtuins, which are related to senescence regulation, BIRK5 and AURKB components of the chromosomal passenger complex). MicroRNAs were placed at the periphery of the interaction network and predicted relationships with protein-coding genes were established. New relationships (in blue) could be established by the introduction of 8 new components in the network (Figure 2(b), in yellow): nidogen 1 (NID1), integrin beta 6 (ITGB6), integrin beta 8 (ITGB8), collagen IV alpha 5 (COL4A5), ADAM metallopeptidase with thrombospondin type 1 motif, 6 (ADAMTS6), reversion-inducing cysteine-rich protein with Kazal motifs (RECK), chymase 1 (CMA1) and fibroblast growth factor 13 (FGF13).

Twenty-one microRNAs of interest and relevant to fibroblast physiology were selected according to the literature (Table 1). Among the selected microRNAs involved in fibroblast senescence, some have been described with decreased expression levels during senescence: miR-143-3p, miR-145-5p, miR-155-5p, miR-21- 5p, miR-23a-3p [17] and miR-7-5p [18] . Other microRNAs are known to have increased expression levels during the senescence process: miR-10b-5p, miR-146a-5p, miR-199a-3p, miR-26a-5p, miR-34a-5p, miR-542-5p [17] , miR-1185-5p, miR-1204, miR-519c-3p, miR-658 [18] , miR-29a-3p, miR-30a-5p [19] , and miR-638, miR- 663 [20] . MiR-29-3p is also involved in fibroblast extracellular matrix synthesis and senescence [21] . MiR-335-3p was described as having a role in proliferation, migration and differentiation [22] . These microRNAs are listed in Table 1.

A microRNA qPCR array including the 21 microRNAs of interest was used to test the modulation of these microRNAs after application of the ingredient to cultured normal human fibroblasts for 48 h. Three microRNAs (miR-1204, miR-519c-3p, miR-658) were not detected by qPCR and were thus considered as non-expressed by fibroblasts under our experimental conditions, reducing the list of microRNAs from 21 to 18. Amplification of these 18 microRNAs (Table 2, column 1) showed RQ values between 0.573 (for miR-335-3p) and 0.919 (for miR-155-5p) (Table 2, column 2). Out of these 18 microRNAs, we selected 14 microRNAs with RQ values lower than or equal to 0.85 (Table 2, column 2, in bold): miR-10b-5p, miR-1185-5p, miR-143-3p, miR-145-5p, miR-199a-3p, miR-23a-3p, miR-26a-5p, miR-29a-3p, miR-30a-5p, miR-335-3p, miR-34a-5p, miR-638, miR-663 and miR-7-5p.

The miRWalk database allowed the identification of validated target genes involved in senescence and extracellular matrix homeostasis, in correlation with the microRNAs of interest (Table 2, column 3). Ingenuity® Pathway Analysis was used to identify new potentially modulated target genes (Table 2, column 5). The genes identified in common with the two methods are also indicated in Table 2, column 4.

Interestingly, among the target genes, family members of collagens (COL1A1, COL2A1, COL2A2, COL3A1, COL4A1, COL 4A2, COL5A3, COL11A2, COL15A1), growth factors and their receptors (EGF, EGFR, VEGFA), transcription factors (SOX2, SOX6, SOX9; FOXO3) or metalloproteinases and their inhibitors (MMP3, MMP9, MMP14; TIMP2) were identified. These genes are well known to be closely related to extracellular matrix homeostasis. Other target genes, such as SIRT1 (Sirtuin 1), SIRT2, BIRC5 (Survivin), AURKB, SHH and LIN28A may also be involved in the fibroblast aging process.

However, miR-1185-5p, miR-23a-3p and miR-335-3p showed no validated target genes in miRWalk, reducing to 11 the list of microRNAs of interest (miR-10b-5p, miR-143-3p, miR-145-5p, miR-199a-3p, miR- 26a-5p, miR-29a-3p, miR-30a-5p, miR-34a-5p, miR-638, miR-663, miR-7-5p) (Table 2, column 1, underlined). The hypothetical effect of the down-regulation of these specific microRNAs on the expression of the potential target gene identified with the miRWalk database is represented in Figure 1. The tendency of a decreased microRNA level observed in our experiments may, as a consequence, induce the up-regulation of the potential target genes listed for each microRNA.

Among the microRNAs, miR-145-5p and miR29a-3p target a large number of genes of interest (respectively, 9 and 10 genes). Some microRNAs (miR-145-5p, miR-26a-5p, miR-29a-3p, miR-30a-5p, miR-34-5p and miR-638) can target collagens (COL1A1, COL2A1, COL2A2, COL3A1, COL4A1, COL 4A2, COL5A3, COL11A2, COL15A1). Interestingly, miR-26a-5p and miR-29a-3p seem to have a high potential for modulating the extracellular matrix composition since they target 3 and 6 collagen family members, respectively (Figure 1). Among the 11 microRNAs studied, miR-26a-5p and miR-335-3p showed a noticeably higher down-regulation following ingredient treatment, with RQ values of 0.577 and 0.573, respectively.

SIRT1, a gene involved in the regulation of cellular senescence, is a miRWalk-validated potential target for miR-34a-5p. Additionally, the Ingenuity® Pathway Analysis predicted that miR-29a-3p and miR-30a-5p may also target SIRT1 and up-regulate its expression, in response to down-regulation (Figure 1).

Genes other than collagens and SIRT1 were targeted by several microRNAs (e.g., SOX2 and VEGFA by 6 microRNAs; SOX9, EGF and EGFR by 3 microRNAs), suggesting the potential importance of these genes in the control of fibroblast homeostasis, extracellular matrix secretion and dermal aging. On the other hand, miR199a-3p and miR-663 are predicted to potentially target only one gene (SOX9 and VEGF1, respectively). The down-regulation of all these microRNAs by the tested ingredient may have a positive impact on extracellular matrix secretion by fibroblasts.

Bioinformatics was used to build an overview of the potential interactions between the 18 modulated microRNAs and their potential target genes identified using the miRWalk database. By using Ingenuity® Pathway Analysis, we developed an interaction network (Figure 2(a)). This approach allowed us to further identify 8 new potential target genes closely related to the molecules already included in the network (ADAMTS6, CMA1, COL4A5, FGF13, ITGB6, ITGB8, NID1, RECK), with a potential role in extracellular matrix homeostasis. This interaction network offers the possibility of visualizing all the partners and their interconnections at once. However, 3 microRNAs (miR-10b-5p, miR-335-3p, miR-542-5p) could not be connected to the network. The resulting network without the relationships with microRNAs is depicted in Figure 2(b) to facilitate visualization. This simplified network focuses on the interactions between the 35 miRWalk potential target genes and the 8 new targets identified by using Ingenuity® Pathway Analysis. The addition of new molecules to the network widens the range of targets that may be studied and validated in further in vitro studies.