The life in a gradient: calcium, the lncRNA SPRR2C and mir542/mir196a meet in the epidermis to regulate the aging process

Introduction

The aging process is to a large part still enigmatic. This is attributed to the fact that aging is multifactorial and is dependent on such processes as senescence, loss of proteostasis, changes on the epigenetic level, altered mitochondrial functions and many more (reviewed in [1]). The aging process in the skin is even more complex, because this tissue consists essentially of two layers (epidermis and dermis) that is predominantly populated by two different cell types: keratinocytes and fibroblasts, respectively. If a hallmark of an aged dermis has to be named, it is the changed architectural organization of the extracellular matrix, finally resulting in folds and wrinkles. The dermis is also a wonderful example for the force of extrinsic aging. Increased UV-exposure (′photoaging′) results in increased ROS levels that stimulates the activation of matrix metalloproteases thus leading to an increased turnover of the extracellular matrix [2].

One of the hallmarks of an aged epidermis, however, is the breakdown of the epidermal calcium gradient [3, 4]. Calcium is a key driver for keratinocyte differentiation and thus for the formation of the epidermal barrier. On the one hand, this barrier protects us against environmental influences such as invading microorganisms, xenobiotics, pollutants and mechanical stress; on the other hand, it reduces the transepidermal water loss. The protective effect of the epidermis is built on three pillars: (1) a heavily cross-linked protein network (cornified envelope; the “bricks” of the epidermis): (2) lipids as a multilamellar barrier between the cells (mainly ceramides; the “mortar” of the epidermis); and (3) the acidic skin pH [5]. The small ion calcium promotes the expression of most cornified envelope specific proteins including involucrin, fillagrin, loricrin and the small proline rich repeat (SPRR) family [6–8]. All these proteins are crosslinked by the formation of Nɛ-(γ glutamyl) lysine (isopeptide) bonds in a Ca²⁺-dependent manner [9, 10]. In addition, the secretion of the lipid filled “lamellar bodies” and thus the formation of the “mortar” between the keratinocytes is driven by calcium [11].

Despite its undeniable essential functions in the epidermis, high concentrations of this ion can be detrimental. Ca²⁺ leads to fully differentiated keratinocytes (corneocytes) in the stratum corneum and cell death through a highly regulated cell death pathway. Accordingly, calcium would be a high risk factor for the progenitor cells that reside in the stratum basale. To cope with such different demands for calcium a gradient is established. The Ca²⁺ concentration is lowest in the stratum basale (cells attached to the basal lamina; in immediate proximity to the dermis) and constantly rises from the startum spinosum to the stratum granulosum, where a peak is reached. However in the stratum corneum, the outmost layer of the epidermis, a clear drop of calcium can be observed [12]. During aging the overall amount of calcium in the epidermis is not decreasing, but the epidermal calcium gradient is collapsing. This has two consequences for the epidermis: A lowered calcium concentration in the stratum granulosum and an increased Ca ²⁺-level in the stratum basale. This redistribution leads to a multitude of side effects: The cornified envelope gets restructured, the epidermal barrier gets rigid and the skin pH rises [3, 8, 13]. It is quite possible that another prominent phenotype of skin aging is attributed to the redistribution of calcium. The skin turnover (the shedding of dead corneocytes and its replacement) is markedly increased from ~30 days to ~60 days [14]. This phenomenon is mainly attributed to the so called Hayflick limit, but it is also likely that the increased leakage of calcium into the stratum basale is involved in the reduced renewal of keratinocytes.

In this paper we are not considering factors that contribute to the breakdown of the epidermal calcium gradient, but we demonstrate how the epidermis adapts to the altered calcium levels in each stratum by the expression of lncRNAs and miRNAs.

Results

Previously, we were able to demonstrate that the cornified envelope is rebuilt during the aging process as a direct response to the collapsing calcium gradient. These changes include such hallmarks as down-regulation of loricrin (the most prevalent protein in the CE) and an upregulation of most small proline rich repeat proteins [3, 15, 16]. It has to be stated that the so called “CE precursor family“ has its origin in a single progenitor gene and presumably arose due to a series of gene duplications [17]. Therefore, all members of this family (loricrin, involucrin, SPRR1A-1B, SPRR2A-2G, SPRR3, SPRR4, LCE1A-1F, LCE2A-2D, LCE3A-3E, LCE4A-6A and LEP7) are located in the epidermal differentiation complex on the chromosomal region 1q21 [18, 19]. This family also harbors a long non-coding RNA (lncRNA): SPRR2C. During evolution, a C at position 102 was mutated into a T. In this way the 6^th codon (glutamine) was changed into a STOP (TAG). Careful re-evaluation of a series of microarray data using Gene Chip HGU133 Plus 2.0 Arrays (Affymetrix) [15] showed a differential expression of this pseudogene. In this study five non-inflamed foreskin tissues from middle aged individuals (age: 18 years – 28 years) were compared with foreskin samples obtained from elderly people (58 years to 74 years). The analysis revealed a 13.96-fold up-regulation of SPRR2C (p-value=0.016). When the same aged foreskin samples were compared with three young non-inflamed foreskin samples (3-5 years) (Rinnerthaler et al., unpublished data) a 8.99-fold up-regulation with increasing age was observed (p-value: 0.071). A RT-PCR analysis comparing young, middle aged and old foreskin samples confirmed the strong increase in SPRR2C expression with age (Figure 1A). On average the SPRR2C levels in old people were 116-fold and 41-fold increased compared to young and middle-aged skin samples, respectively. The rise of this lncRNA in some elderly is dramatic (more than 1050-fold). The presence of this C to T transition resulting in a premature STOP codon is not specific for humans but occurred late in hominid evolution (Figure 1B). Because humans as well as chimpanzees and bonobos have this pseudogene, it is very probable that there is a certain evolutionary advantage in keeping the pseudogene.

In several publications a role for lncRNAs as ‘miRNA sponges’ was suggested [20–22]. Accordingly, SPRR2C was already identified as a miRNA-196a-5p target [23]. By using several online available tools, the most promising miRNAs were predicted that are potentially binding to the lncRNA SPRR2C. According to this analysis, the best candidates are hsa-mir542-5p, hsa-mir125a, hsa-mir135a-5p, hsa-mir196a-5p, hsa-mir491-5p and hsa-mir552-5p. The minimal free energy (mfe) for the hybridization between the miRNAs and their target SPRR2C was recalculated by using RNAhybrid [24]. In all cases the seed regions of the predicted miRNAs (nucleotides 2/3-7/8 from the 5′-end) perfectly match SPRR2C (Figure 2A). The potential binding sites of these 6 miRNAs in the SPRR2C sequence are shown in Figure 2B. The two best hits according to our analysis were hsa-mir542-5p and hsa-mir196a-5p with a mfe of -36.6 kcal/mol and -32.7 kcal/mol, respectively. Attributed to the ancestral origin most members of the “CE precursor family“ are targets of the same miRNAs (Supplementary Figure 1).

Using a RT-PCR approach, we analyzed the regulation of hsa-mir542-5p, hsa-mir125a-5p, hsa-mir135a-5p, hsa-mir196a-5p, hsa-mir491-5p and hsa-mir552-5p in young, middle-aged and old foreskin samples (N=5 of each age). The exact same foreskin samples were used as in Figure 1A and data are presented in Figure 3. It has to be stated that the levels of all miRNAs that were tested show a dramatic continuous increase with age. The expression levels of mir542-5p rise more than 10-fold on the transition from childhood to adolescence. In old age, a more than 200-fold up-regulation of mir542-5p was observed. The fold-increase for all other miRNAs is as follows: 35-fold (hsa-miR-196a-5p); 132-fold (hsa-miR-491-5p); 176-fold (hsa-miR-125a-5p); 37-fold (hsa-miR-135a-5p) and 124-fold hsa-miR-552-5p. In most of the skin samples a high SPRR2C expression level correlates very well with a high abundance of hsa-mir542-5p (Pearson Correlation Coefficient R=0.7967).

For the following experiments, we focused on two miRNAs: mir196a-5p as a bona fide SPRR2C binding miRNA [23]; miR542-5p due to its high expression and low mfe. The SPRR2C binding of mir542-5p and mir196a-5p was validated as follows: Either full length SPRR2C or a specific region of SPRR2C (nucleotides 407-628) were cloned into the luciferase reporter plasmid psiCHECK™-2. The latter region was chosen, because it harbors most miRNA binding sites (Figure 2B). In mammalian cells two RNAs are produced from this construct: (1) a firefly luciferase mRNA used for normalization; (2) a renilla luciferase (downstream)-SPRR2C (upstream) hybrid mRNA. The stability of the hybrid mRNA was tested via RT-PCR. For this assay HaCaT cells were co-transfected with either a negative control miRNA or hsa-mir542-5p or hsa-mir196a in combination with one of the three psiCHECK™-2 vectors (psiCHECK™-2; psiCHECK™-2-SPRR2C; psiCHECK™-2-SPRR2C (281-486 fragment)). The successful transfection with the miRNA was verified by RT-PCR analysis with specific primers after RNA isolation and poly-A-tailing (data not shown). In a first step the RNA levels of the renilla luciferase in relation to the firefly luciferase mRNA was analyzed via RT-PCR. Confirming the results of Maru et al. [23] we saw a decrease of renilla luciferase levels that were expressed from the psiCHECK™-2-SPRR2C (407-628) vector after addition of mir196a-5p (data not shown). However, it has to be stated that in our hands the mir196a effect on luciferase expression was dramatically less (a 10% decrease compared to a 70% decrease as seen by Maru). In case of mir542-5p transfection (40 nM) a highly significant 20% decrease of renilla luciferase abundance expressed from the psiCHECK™-2-SPRR2C (407-628) vector was detected. This reduction was even confirmed at lower mir-542-5p concentrations albeit at lower levels indicating a concentration dependent degradation of SPRR2C (data not shown). By using a commercially available assay measuring dual luciferase activity (renilla luminescence compared to firefly luminescence) we saw a nearly complete abolishment of renilla luciferase activity (~23-fold reduction in activity) after transfection with 125 nM mir-542-5p (Figure 4). In cells co-transfected with the control vector psiCHECK™-2 and mir-542-5p no downregulation of renilla luciferase activity was observed.

In a next step the effect of a SPRR2C overexpression on keratinocyte differentiation and cornified envelope formation was tested. SPRR2C was cloned into the retroviral transfer vector pLL3.7 [25] and after lentiviral production in HEK293FT cells, HaCaT cells were transduced with the lncRNA. Using qPCR the overexpression of SPRR2C was confirmed (~90-fold; Figure 5). In addition, the expression level of selected CE specific genes was tested. The overexpression of the pseudogene SPRR2C resulted in an increase of LCE2A (11.5-fold), LCE1A8 (7.7-fold), SPRR2B (6-fold), LCE3A (4.8-fold), SPPR2G (2.8 -fold) and loricrin (2.5-fold). The positive effect of the lncRNA SPRR2C was confirmed by Western blot analysis. Previously, SPRR2A was already established as a marker for keratinocyte differentiation that responds to skin aging and were thus chosen for this study [3]. In fact, SPRR2C overexpression also results in an increase in SPRR2A protein levels (Supplementary Figure 2).

The transfection of HaCaT cells with mir196a-5p had the opposite effect, although on modest levels. In case that a knock-down of SPRR2C was observed, the differentiation status of keratinocytes decreased as deduced from loricrin expression levels (in the mean a 2-fold reduction; the Pearson correlation coefficient R between SPRR2C and loricrin expression is 0.88; N=6). Surprisingly, the transfection with mir542-5p showed no correlation between SPRR2C and loricrin levels (R=-0.30). In each single experiment (N=6) mir542-5p induced a SPRR2C response ranging from a 19.7-fold up-regulation to a 5.5-fold down-regulation.

Therefore, another approach was chosen to test the involvement of mir542-5p in keratinocyte differentiation. Ca²⁺ is a key player in the transition from keratinocytes to corneocytes and thus reduces the cell division rate. In fact, addition of 0.8 mM CaCl₂ increased the doubling time of HaCaT cells from 27 hours (0.5 mM CaCl₂) on average (N=4) to 66 hours (N=6; p-value=0.0308). No significant further change in doubling time was observed by rising the calcium concentration to 1mM CaCl₂ (Figure 6). Transfection of HaCaT cells with mir542-5p completely restored the cell division rate (30 hours at 0.8 mM CaCl₂; 33 hours at 1 mM CaCl₂). Based upon the finding that mir542-5p specifically counterbalances the effects of high calcium concentrations, the role of this miRNA was tested at increased Ca²⁺ levels. At 0.8 mM CaCl₂ (compared to 0.09 mM CaCl2) a more than 6-fold rise of loricrin levels was observed (Figure 5B). This confirms previously published loricrin expression levels [26]. Additionally, we also detected 9-fold increased SPRR2C expression levels upon addition of 0.8 mM CaCl₂. After transfection with mir542-5p, SPRR2C, loricrin and filaggrin levels decreased: (1) SPRR2C: 272-fold down; (2) loricrin: 5-fold down; and (3) filaggrin: 5.6-fold-down. In accordance to results presented above, the mir542-5p effects on CE formation are less pronounced at lower CaCl₂ concentrations (data not shown). It also has to be stated that the transfection process itself had some effect on the differentiation status of keratinocytes (Figure 5B).

In the epidermis ions, RNAs as well as proteins are organized in a gradient. Therefore, it is of utmost importance that the antagonists mir542-5p and SPRR2C meet in some epidermal strata. In a RNA ISH (in situ hybridization) approach both SPRR2C and mir542-5p where localized in human skin samples by digoxygenin labelled probes. Most CE proteins are synthesized in and localize to the outermost stratum spinosum, stratum granulosum and stratum corneum. Although SPRR2C is not encoding a protein, the localization of the lncRNA is not unusual and can be found in the stratum granulosum and corneum (Figure 7B, 7C). In addition, the RNA in situ hybridization experiments confirm the RT-PCR results (Figure 1A). SPRR2C is hardly expressed in young foreskin samples and shows increased levels from middle-aged to old individuals. In Figure 7A–7C representative images are shown. The expression and localization of SPRR2C is not specific for the body region and was observed in foreskin as well as eyelid samples. The exact same phenotypes were observed for eyelid samples obtained from middle-aged and old women after blepharoplasty (data not shown). RNA ISH experiments also show an increase in mir542-5p levels in aged skin samples (Figure 7D, 7E) and thus confirm a previous RT-PCR analysis (Figure 3). The expression of mir542-5p is equally distributed across all epidermal strata with the exception of stratum corneum. In oversaturated specimen, mir542-5p was also observed in stratum corneum (data not shown). No traces of this miRNA can be localized in the dermis and it seems to be epidermal specific. In young skin samples mir542-5p is practically undetectable. The RNA ISH experiments demonstrate that both mir542-5p and SPRR2C meet in the stratum granulosum and corneum.

Discussion

The central findings of this paper are summarized in Figure 8. In healthy epidermis a calcium gradient is established. This is of absolute necessity because Ca²⁺ is essential for differentiation of keratinocytes and therefore detrimental to progenitor cells. The highest amounts of Ca²⁺ can be found in the stratum granulosum. In this epidermal layer as well as in the outermost stratum spinosum most members of the CE precursor family as well as ω-OH ceramides are formed in a calcium dependent manner. All these proteins get cross-linked and form the cornified envelope, a multi-protein mesh below the plasma membrane. Additionally, the phospholipids in the plasma membrane get replaced by ceramides and thus the keratinocytes transform into corneocytes that are finally shed. All the processes listed above are in strict dependence on high amounts of calcium. These facts also explain that calcium is a serious threat to the progenitor cells in the stratum basale, the basic layer of the epidermis, due to the induction of the differentiation process. Therefore, the calcium concentration in this stratum is kept low (Figure 8A).

During aging, a still not completely understood breakdown of the epidermal calcium gradient takes place [3]. The calcium in the epidermis is not lost, but its concentration increases in the stratum basale and decreases in the stratum granulosum. This has two consequences: (1) a less pronounced epidermal barrier with e.g. reduced levels of loricrin; (2) a reduced proliferation rate of progenitor cells in the stratum basale (Figure 8B). In the course of this work, we could demonstrate that a rise of the calcium concentrations doubles the time of the cell cycle (from 27 hours to 66 hours) in HaCaT cells.

To cope with this altered calcium distribution the epidermis reacts with an increased expression of miRNAs as hsa-mir542-5p, hsa-mir125a-5p, hsa-mir135a-5p, hsa-mir196a-5p, hsa-mir491-5p and hsa-mir552-5p. We could show here that in the skin of elderly people the amount of these miRNAs increases at least 35-fold. The highest expression level was found for mir542-5p with a 200-fold upregulation in aged skin. The localization of the latter miRNA was studied in an RNA ISH approach. In fact, the presence of mir542-5p in the stratum basale was corroborated as well as the high expression level in the skin of elderly people. In cell culture experiments after transduction of HaCaT cells we could demonstrate that mir542-5p counterbalances the calcium induced slow-down of the cell cycle (Figure 8B). These data fit quite well to results published by the Grillari group previously [27]. They showed that senescent skin fibroblasts as part of the senescence-associated secretory phenotype significantly secrete hsa-mir542-5p, hsa-mir196a-5p, hsa-mir125a-5p and mir491-5p and pack them into small extracellular vesicles [27]. It is therefore quite possible that the fibroblasts deliver these vesicle packed miRNAs to the epidermis in aged skin and thus modulate proliferation and differentiation of keratinocytes.

The increased levels of the miRNAs mentioned above have a major drawback for the outermost layers of the epidermis: For example, the small non-coding RNA mir542-5p is not specifically localized to the stratum basale but can be found in all other epidermal layers. Especially in the stratum granulosum and corneum, these RNA molecules can be harmful for the skin and could result in a breach of the epidermal barrier (Figure 8C). A transfection of HaCaT cells with mir196a-5p and mir542-5p revealed decreased mRNA levels of most CE precursor family members. By an unknown mechanism, the RNA degradation mediated by mir542-5p is strictly dependent on the presence of high amounts of calcium.

To prevent damage to the skin a compensation mechanism based on a lncRNA was established (Figure 8D). We could show that during skin aging SPRR2C levels, a lncRNA that arose during hominid evolution, are increased. Bioinformatics analysis revealed that SPRR2C is targeted by mir542-5p, mir125a-5p, mir135a-5p, mir196a-5p, mir491-5p and mir552-5p. Of all these miRNAs, mir542-5p is the best candidate and thus showed the lowest mfe upon binding to SPRR2C. It also has to be mentioned that these miRNAs are not specific for SPRR2C but target the mRNAs of most CE precursor family members. In this way, SPRR2C acts as a decoy for miRNAs and thus promotes the establishment of the CE and the epidermal barrier as a whole. Previously, binding of mir196a-5p to SPRR2C was demonstrated [23]. By applying a reporter approach based on firefly and renilla luciferases, we could confirm a binding of mir196a-5p and mir542-5p to SPRR2C. A qPCR analysis also demonstrated that after transfection of HaCaT cells with either mir196a-5p or mir542-5p levels of SPRR2C are drastically reduced.

Opposing to miRNAs SPRR2C is promoting keratinocyte differentiation. Overexpression of SPRR2C in cell culture experiments in fact resulted in increased RNA levels of most CE precursor family members. SPRR2C itself is not promoting cell differentiation in the stratum basale, because this lncRNA is strictly localized to the stratum granulosum/corneum where the transition from keratinocytes to corneocytes takes place. This result also stresses the importance of gradients and localizations of proteins, ions and RNAs in the epidermis. To promote keratinocyte differentiation calcium as well as lncRNAs are concentrated in the outermost layers of the skin. As an exception to this rule, miRNAs are evenly distributed over all epidermal layers. In the stratum basale this small non-coding RNAs facilitate cell proliferation. In the stratum granulosum their action is nullified by the lncRNA SPRR2C. Furthermore, it would be interesting to see if these miRNAs fulfil a specific action in the stratum spinosum during skin aging.

Materials and Methods

Supplementary Materials

Supplementary Figures

Supplementary Table 1