Estradiol Protects Dermal Hyaluronan/Versican Matrix during Photoaging by Release of Epidermal Growth Factor from Keratinocytes*

Background: Skin aging involves UVB-induced degeneration of the dermal extracellular matrix. Results: Estrogen induces epidermal growth factor expression in keratinocytes thereby stimulating hyaluronan synthase 3 and versican expression in dermal fibroblasts of UVB-irradiated skin. Conclusion: Paracrine release of epidermal growth factor in response to estrogen maintains hyaluronan and versican-rich extracellular matrix. Significance: Estrogen prevents specific aging responses in the hyaluronan matrix of photoaged skin. Hyaluronan (HA) and versican are key components of the dermis and are responsive to ultraviolet (UV)B-induced remodeling. The aim of this study was to explore the molecular mechanisms mediating the effects of estrogen (E2) on HA-rich extracellular matrix during photoaging. Hairless skh-1 mice were irradiated with UVB (three times, 1 minimal erythema dose (80 mJ/cm2), weekly) for 10 weeks, and endogenous sex hormone production was abrogated by ovariectomy. Subcutaneous substitution of E2 by means of controlled-release pellets caused a strong increase in the dermal HA content in both irradiated and nonirradiated skin. The increase in dermal HA correlated with induction of HA synthase HAS3 by E2. Expression of splice variant 2 of the HA-binding proteoglycan versican was also increased by E2. In search of candidate mediators of these effects, it was found that E2 strongly induced the expression of epidermal growth factor (EGF) in UVB-irradiated epidermis in vivo and in keratinocytes in vitro. EGF in turn up-regulated the expression of HAS3 and versican V2 in dermal fibroblasts. HAS3 knockdown by shRNA caused inhibition of fibroblast proliferation. Furthermore, HAS3 and versican V2 induction by E2 correlated positively with proliferation in vivo. In addition, the accumulation of inflammatory macrophages, expression of inducible cyclooxygenase 2, as well as proinflammatory monocyte chemotactic protein 1 were decreased in response to E2 in the dermis. Collectively, these data suggest that E2 treatment increases the amount of dermal HA and versican V2 via paracrine release of EGF, which may be implicated in the pro-proliferative and anti-inflammatory effects of E2 during photoaging.

by intrinsic and extrinsic aging responses and by an overlap with photo-carcinogenesis under certain circumstances (1). The role of the ECM 2 of the skin in this aging process is well established. In particular, the cleavage of collagen by matrix metalloproteinases has been demonstrated (2). The partially degraded collagen network heals imperfectly through de novo collagen synthesis, leaving microscars in the skin. Collagen fragments that are released as a result of matrix metalloproteinase-induced collagen cleavage are bioactive and participate in the regulation of fibroblast phenotype during photoaging. Moreover, collagen fragments reduce de novo synthesis of dermal collagen (3). Thus, matrix degradation and altered matrix expression influence fibroblast phenotypes and may thereby perpetuate UVB-induced aging responses. Other dermal ECM molecules are also reportedly affected by UVB irradiation, particularly hyaluronan (HA) and proteoglycans (4 -7). HA is abundant in the dermis and is thought to contribute to water content, turgidity of the skin, and the diffusion of soluble factors and nutrients (5). Furthermore, HA can support the proliferative phenotype of fibroblasts and possibly opposes apoptosis (8,9). HA is synthesized at the plasma membrane by HA synthase isoenzymes-1, -2, and -3 (HAS1-3) (10). These enzymes extrude HA into the extracellular space after assembly of UDP-glucuronic acid and UDP-N-acetyl-D-glucosamine into a growing chain of ␤(1-3)-linked D-glucuronic acid and N-acetyl-D-glucosamine disaccharides. The repeating disaccharides are linked by hexosaminidic ␤(1-4) bonds that form high molecular weight HA of up to 10 7 Da and up to 20 m in length (11). HAS isoenzymes are expressed at a relatively low copy number per cell but can rapidly produce large amounts of HA. In addition to transcriptional regulation of HAS enzymes, it has recently been shown that regulation also takes place at the post-transcriptional level through phosphorylation, glycosylation, and mono-ubiquitination (12)(13)(14). Furthermore the formation of homo-and heterodimers has now been demonstrated (13).
With respect to skin aging, it has been shown that HA is reduced by chronic UVB irradiation, and this loss likely contributes to the aged phenotype of skin (8,(15)(16)(17). However, a limited number of studies have also demonstrated either no change or increased dermal HA in response to UVB (15,18). The latter is likely an acute response associated with heliodermatitis.
Hyaluronan is bound by versican, a large chondroitin sulfate proteoglycan, through specific binding domains termed link modules that each consist of ϳ100 amino acids and are part of the link protein domain. The link protein domain is composed of an immunoglobulin domain and two link modules and is present in the N-terminal globular G1 domain of versican (19). Because multiple versican molecules bind to one chain of HA, large networks of HA and versican are formed. These HA and versican-rich matrices are known to critically govern the proliferative and migratory phenotype of mesenchymal cells (20).
Recently, it has been demonstrated that versican accumulates in response to UVB irradiation (7). However, little is known about what the specific functions of versican might be during skin aging, how versican is regulated during skin aging, and specifically whether versican is responsive to estrogen.
Skin aging is accelerated after menopause and positive effects on the skin including increased thickness, increased moisture, decreased wrinkling, and improved wound healing responses became obvious upon estrogen treatment of postmenopausal women (21). However, hormone replacement therapy is confounded by thrombotic and malignant complications and is now restricted to short term use in selected cases (22). It is therefore of great interest to better understand the molecular mechanisms that underlie the protection of the skin matrix by E 2 . The aim of this study was to investigate the effect of E 2 on the dermal hyaluronan and versican matrix during UVB-induced skin aging by use of ovariectomized and E 2 treated hairless skh-1 mice.

EXPERIMENTAL PROCEDURES
UVB Irradiation of Mice-Female hairless mice (Skh:Hr1) (Charles River Laboratories) were housed according to standard procedures. Mice were randomly assigned to either sham procedure or bilateral ovariectomy (OVX, X) at the age of 8 weeks as described previously (23). OVX mice were subdivided into four treatment groups, which received either placebo or E 2 Ϯ UVB irradiation. E 2 treatment was performed by implantation of subcutaneous slow release hormone pellets (Innovative Research of America) prepared to dispense 1.1 g/day E 2 for the duration of the 10-week experimental period. Placebo (P) pellets served as control. After OVX and pellet implantation at 8 weeks, half of the mice were irradiated with UVB light and the other half served as nonirradiated controls. Sham-operated animals received placebo pellets Ϯ UVB irradiation. Thus, six experimental groups (Fig. 1A) were compared in total as follows: 1) sham, placebo (S,P); 2) OVX, placebo (X,P); 3) OVX, E 2 (X,E 2 ); 4) sham, placebo, UVB (S,P UVB); 5) OVX, placebo, UVB (X,P UVB); and 6) OVX, E 2 , UVB (X,E 2 UVB). Animals were exposed to UVB radiation in an irradiation chamber as described previously (8) using UV lamps with fluorescent bulbs (280 -320 nm with a peak at 313 nm TL 20W/12; Philips, Eindhoven, The Netherlands). UVB irradiation was performed three times per week at a dose of 80 mJ/cm 2 (irradiation time 1 min 36 s) equaling one minimal erythema dose over a period of 10 weeks (Fig. 1A). The light intensity was determined by means of a UV meter (Waldmann, Villingen-Schwennigen, Germany). Skin biopsies from the dorsal skin of 1 ϫ 1.5 cm 2 in size were obtained from control and UVBirradiated animals after 10 weeks. All animal experiments were approved by the local ethical committee for animal experiments.
Digital image analysis was performed using a modified approach based on Dai et al. (8). Bright field images (8-bit) of the stained sections were captured using a Leica DM2000 microscope (Leica Microsystems, Wetzlar, Germany) at ϫ100 magnification. For the quantification of the staining intensity, ImageJ software 1.41 Version (National Institutes of Health) was used according to instructions of ImageJ. The color deconvolution plug-in was employed to separate the color channels of hemalaun and DAB. Negative controls and strongly DAB-positive images were used to determine the thresholds and the analyzed range of signals. This procedure minimized the background interference and maximized the signal of DAB-positive tissue. Thresholds were set for the complete analysis of one antigen. The area fraction is defined as percentage of the area classified as DAB-positive. The measurement was performed in the papillary dermis excluding regions that contained hair follicles. Per skin section data from three randomly selected areas were averaged.
UVB irradiation of the cells was performed with a Bio-Sun irradiation system (Vilbert Lourmat, Munich, Germany) containing two 30-watt UVB sources (312 nm). During the UVB irradiation procedure (100 mJ/cm 2 ; irradiation time ϳ10 s), cells were kept in phosphate-buffered saline solution, which was replaced by DMEM containing 10% charcoal-treated FCS with or without 100 nM ␤-estradiol (Sigma) immediately after the irradiation. Cells were harvested 24 h after stimulation.
For the cell media transfer experiments, NEHKs were UVBirradiated and stimulated with E 2 as described above in keratinocyte media 2. The supernatant was collected 24 h after stimulation and was transferred to fibroblasts in the presence or absence of erlotinib (3 M) (LC Laboratories, Woburn, MA), EGF-neutralizing antibody, or IgG isotype control (both 0.35 g/ml, Abcam, Cambridge, UK). Fibroblasts were harvested 24 h after exposure to conditioned media.
RNA Isolation and Quantification of Gene Expression-Total RNA was isolated using RNeasy total RNA kits (Qiagen, Hilden, Germany). The RNA concentration was determined via photometric measurement at absorbance 260/280. Total RNA (1 g) was reverse-transcribed using the SuperscriptIII first-strand synthesis system for reverse transcriptase-PCR (RT-PCR) (Invitrogen). Primers to analyze target mRNA expression levels were designed employing Primer Express 3.0 software (Applied Biosystems, Darmstadt, Germany) based on published mRNA sequences. The sequences are given in Table 1. Real time RT-PCR was performed in triplicate using SYBR Green PCR master mix (Applied Biosystems) as described (15). The 2 Ϫ⌬⌬C(T) method was used for comparison of the relative expression between control and treated cells. A melting curve analysis was performed after every run, and a negative control containing only master mix, the primer pair, and water was included on every plate.
Knockdown of HAS3-HAS3 knockdown was achieved by usage of the MISSION TM lentiviral shRNA knockdown system (Sigma) as described previously (24). A scrambled shRNA served as control, and the anti-HAS3 hairpin sequence was 5Ј-CGGGCTCTACAACTCTCTGTGGTTCTCGAGAACCA-CAGAGAGTTGTAGAGCTTTTTG-3Ј.FuGENE 6 (Roche Applied Science) was used for the transfer into the packaging cell line HEK293T (ATCC, Wesel, Germany). For better stability of the produced lentiviral particles, the medium was changed to Iscove's modified Dulbecco's medium after 16 h. The lentiviruses were harvested 24 h later and concentrated by centrifugation with poly-L-lysine as reported previously (25). Fibroblasts were transfected with a multiplicity of infection of 5 and kept in normal growth medium for at least 5 days before stimulation with conditioned keratinocyte medium. Before using the cells in proliferation assays, the efficacy of the lentiviral HAS3 knockdown was verified by quantitative RT-PCR (data not shown) and Western blot analysis (anti-HAS3, 1:1000, Sigma) (supplemental Fig. 1).

Gene
Primer sequence Determination of HA and EGF in Cell Culture Supernatants-HA concentration in the supernatants was determined using an HA test kit based on bHABP (Corgenix, Peterborough, UK) 24 h after stimulation and was normalized to total cellular protein. The EGF concentration in the supernatants was determined by EGF human ELISA kit (Abcam, Cambridge, UK) according to the manufacturer's protocol 24 h after stimulation.
Immunoblotting-In equal aliquots of cell culture supernatants or skin extracts, versican was first enriched by DEAE ion exchange chromatography and was chondroitin ABC lyase-digested, and subsequently core proteins were separated on SDS-PAGE and Western blotted as described previously (26). LF99 (1:1000) was used as primary antibody.
Statistical Analysis-All data sets were analyzed by one-way analysis of variance and the Bonferroni post hoc test. Data pre-FIGURE 1. Induction of HA by E 2 in vivo. Hairless skh-1 mice were ovariectomized (OVX, X) at the age of 7-8 weeks or were sham-operated (S). Ovariectomized and sham-operated animals received either a subcutaneous placebo pellet (P) as control or a subcutaneous long release E 2 pellet (1.1 g E 2 /day/mouse) (E 2 ). Subsequently, mice were subjected to UVB irradiation (three times 1 minimal erythema dose, weekly) for 10 weeks. At age of 18 weeks, skin biopsies were obtained, and the amount of HA was quantified. A, experimental protocol and abbreviations. B, affinity histochemistry of the skin using biotinylated bHABP and quantitative analysis (area fraction) of HA staining in the papillary dermis. C, measurement of HA in skin extracts using a bHABP-based kit. ϫ100 magnification; n ϭ 7-12, mean Ϯ S.E.; *, p Ͻ 0.05 versus S,P; #, p Ͻ 0.05 versus S,P UVB; ࡗ, p Ͻ 0.05 versus X,P; OE, p Ͻ 0.05 versus X,P UVB. sented in Fig. 5 were analyzed by Student's t test. Data are presented as means Ϯ S.E., and statistical significance was assigned at the level of p Ͻ 0.05.

E 2 Elevates
Dermal HA in UVB-irradiated and Nonirradiated Skin-Hairless skh-1 mice were ovariectomized to deplete endogenous sex hormones. The lack of E 2 was substituted in OVX mice by implantation of subcutaneous long release pellets (1.1 g E 2 /day/mouse). This nonirradiated group allowed the characterization of the effect of E 2 on the intrinsically aged dermal HA matrix at 18 weeks of age. OVX led to slightly reduced amounts of HA as shown by affinity histochemistry and biochemical quantification in skin extracts (Fig. 1, B and C). Of note, E 2 substitution substantially elevated dermal HA above the control level (S,P) suggesting a stimulatory effect of E 2 .
To investigate the role of estrogen during photoaging, skh1 mice were irradiated with UVB (three times 1 minimal erythema dose (80 mJ/cm2), weekly) for 10 weeks. Subsequently, skin biopsies were collected from UVB-exposed areas. UVBinduced loss of dermal HA (S,P UVB) and again substitution of E 2 prevented the decline of HA (X,E 2 UVB) in a way that HA levels were even higher than those observed in nonirradiated, non-OVX controls (S,P, Fig. 1, B and C).
Dermal HAS3 Expression Is Responsive to E 2 in the Skin of UVB-irradiated and Nonirradiated Mice-To analyze the mechanisms that mediate the elevation of HA content in response to E 2 , mRNA expression of Has isoenzymes and hyaluronidases (Hyal) 1 and 2 were analyzed. Of note, Has3 mRNA expression was strongly induced by E 2 in nonirradiated and irradiated OVX mice ( Fig. 2A). HAS1 was significantly downregulated in OVX mice that received UVB irradiation (X, P UVB). This effect was reversed by E 2 , which could therefore contribute to the recovery of HA matrix in E 2 -supplemented mice after irradiation (supplemental Fig. 2A). Has2 by contrast was up-regulated after OVX (X, P), which could potentially counteract the loss of HA to some extent. The other conditions examined did not affect Has2 expression (supplemental Fig.  2B). Therefore Has3 was considered as the most important HAS isoenzyme mediating the recovery of HA matrix after E 2 substitution. Increased Has3 expression in response to E 2 was also validated at the protein level by immunocytochemistry and immunoblotting. As a result, HAS3 protein detection confirmed the regulation of Has3 mRNA (Fig. 2, B and C) by E 2 . Western blot analysis detected an additional band below the predicted size of HAS3 (63 kDa). The identity of the band is not known, but the detected protein showed the same expression pattern as the mature HAS3.
Nonirradiated skin also showed induction of Hyal2 in response to OVX (supplemental Fig. 2D), which was prevented by E 2 substitution and could therefore contribute to decreased HA content in response to OVX and the increase of HA in response to E 2 treatment. Hyal1 was not regulated in a fashion that could explain the observed changes in dermal HA (supplemental Fig. 2C). Finally, expression of the HA receptors CD44 and receptor of HA-mediated motility (Rhamm) were determined by real time RT-PCR (supplemental Fig. 2, E and F). Cd44 expression was found to be unresponsive to both OVX/E 2 and to UVB irradiation. Rhamm, however, was down-regulated by UVB irradiation in OVX animals, and E 2 substitution partially restored expression, suggesting that HA signaling via RHAMM is responsive to E 2 .
Quantitation of E 2 receptors by real time RT-PCR revealed that E2R␤ was not detectable in the dermis of female hairless skh-1 mice in any of the experimental groups (data not shown). In contrast, the E2R␣ receptor (supplemental Fig. 2G) was highly abundant, but expression was not affected by any of the experimental interventions. These data suggest that the effects of E 2 were mediated by the E2R␣ receptor.  Fig. 1. At the age of 18 weeks, skin biopsies were obtained, and the amount of HAS3 was analyzed. A, Has3 mRNA expression in skin extracts. B, HAS3 immunostaining and quantitative image analysis of the papillary dermis The 63-kDa band was quantified by densitometry and related to tubulin loading control (55 kDa). C, HAS3 immunoblotting (H64) of total skin extracts and quantitative analysis. ϫ100 magnification; n ϭ 7-12, mean Ϯ S.E.; *, p Ͻ 0.05 versus S,P; #, p Ͻ 0.05 versus S,P UVB; ࡗ, p Ͻ 0.05 versus X,P; OE, p Ͻ 0.05 versus X,P UVB. E 2 Up-regulates Versican Content in Dermal Matrix-Subsequently, the regulation of the HA-binding proteoglycan versican was characterized. Immunostaining of versican revealed that OVX did not affect versican but that substitution of E 2 caused marked dermal versican accumulation in nonirradiated skin (X,E 2 , Fig. 3A). Furthermore UVB irradiation itself raised the versican content of the dermis in the immunohistochemical staining (Fig. 3A), and this was further increased by E 2 . Four splice variants of versican (V 0 , V 1 , V 2 , and V 3 ) are known. To obtain a read-out of the overall versican mRNA expression, a primer that did not distinguish between the splicing variants was initially used. The mRNA expression of pan-Versican was not changed or even decreased in response to UVB (Fig. 3B), but a trend toward up-regulation was seen in X,E 2 samples, which reached statistical significance in UVB irradiated X,E 2 mice. Immunoblotting of versican in skin extracts revealed a major versican band at ϳ200 kDa, which likely represented the versican splice variant V2 (Fig. 3C) and closely resembled the regulation of versican as evidenced by immunostaining. Therefore, mRNA expression of Versican V2 was analyzed by real time RT-PCR (Fig. 3D) and found to parallel the protein data shown in Fig. 3, A and C.
Comparable changes in versican expression were also determined in the epidermis by quantitation of immunostainings (data not shown) suggesting a similar response to E 2 in the epidermis. Therefore, part of the data from whole skin extracts (mRNA and immunoblotting) also reflect increased epidermal versican. In contrast, regulation of HAS3 was not detected in the epidermal compartment (data not shown). Future studies may address the effect of E 2 and versican specifically in the epidermis.
Thus, dermal HA and versican were markedly increased by E 2 in both nonirradiated and irradiated skin. Furthermore, the data suggest that HAS3 and versican V2 are the molecular targets of this regulation. With respect to quantitative remodeling, the UVB-irradiated dermal matrix of OVX mice is characterized by a proportional loss of HA and an increase in versican.
Effects of E 2 on HAS Isoenzymes and Versican in Dermal Fibroblasts in Vitro-The above described in vivo results suggest a strong effect of exogenous E 2 on the dermal HA and versican matrix. In search of the underlying regulatory mechanisms, two possibilities were considered. First, direct transcriptional effects of E 2 on HAS isoenzymes and VERSICAN in fibroblasts may be responsible. Second, indirect effects of E 2 on the expression of growth factors that may in turn affect gene expression could be involved. Therefore, it was tested whether the HAS isoenzymes and VERSICAN are responsive to E 2 and UVB in vitro. In monolayer cultures of human skin fibroblasts, HAS3 mRNA was not affected by either UVB or E 2 (Fig. 4A), although HAS3 immunoblotting even suggested a slight decrease of HAS3 protein after UVB exposure (Fig. 4B). pan-VERSICAN mRNA was down-regulated in response to UVB and UVB ϩ E 2 (Fig. 4C). However, immunoblotting of versican from supernatants revealed no effect on versican V2 expression (Fig. 4D). Therefore, in contrast to the in vivo results, HAS3 and versican were clearly not directly induced by E 2 in fibroblasts in vitro. Next, concentration-dependent responses of fibroblasts exposed to E 2 and UVB were examined by real time RT-PCR. Using 0.1, 1, 10, and 100 M E 2 revealed that HAS3 mRNA was not regulated, whereas HAS1 and HAS2 mRNA were even down-regulated in fibroblasts 24 h after stimulation (data not shown). Application of 10, 50, and 100 mJ UVB/cm 2 led to down-regulation of HAS1 and HAS2 at 100 mJ/cm 2 and a weak reduction of HAS3 mRNA at 10 mJ/cm 2 (data not shown). The transcriptional down-regulation of all three HAS isoenzymes in response to direct UVB irradiation of fibroblasts could potentially be related to the loss of HA in response to UVB in vivo.
However, in vivo only Has1 mRNA was slightly reduced in UVB-irradiated skin of X,P mice (supplemental Fig. 2, A and B). With respect to pan-VERSICAN mRNA expression, dose responses revealed that only 100 nM E 2 inhibited expression as seen in Fig. 4C, whereas lower concentrations had no effect (data not shown). After UVB irradiation, a small increase of pan-VERSICAN mRNA expression (1.39 Ϯ 0.15-fold, p Ͻ 0.05 compared with control, n ϭ 3) was detected at 10 mJ/cm 2 , followed by significant inhibition by 100 mJ/cm 2 as presented in Fig. 4C. All considered, in cultured fibroblasts neither the estrogen response nor the acute UVB response mimicked the results obtained in vivo. Therefore, a direct regulation of the involved genes in fibroblasts by E 2 and UVB appeared unlikely to be responsible for the induction of HA and versican in vivo in response to E 2 .
EGF Is Induced in Keratinocytes in Response to E 2 and Induces HAS3 and Versican Expression in Fibroblasts-Next, a possible paracrine mechanism was addressed. For this purpose, skin biopsies were examined with regard to differential regulation of growth factors known to be involved in the regulation of the HA matrix. Specifically, it was searched for candidates that were diminished by OVX and increased by E 2 substitution both in nonirradiated and irradiated skin. Egf was identified as one candidate that showed reduced mRNA expression in OVX mice, which could be rescued by E 2 in both nonirradiated and UVB-irradiated mice (Fig. 5D). EGF protein expression was therefore quantified in skin extracts by ELISA (Fig. 5E) confirming the induction by E 2 in nonirradiated and irradiated OVX animals. Furthermore, expression of EGF protein specifically in the epidermis was determined by immunostaining (Fig.  5F). The results confirmed that EGF is indeed induced in the epidermis and in total skin extracts by E 2 . In contrast, Tgf␤1-3 mRNA patterns did not correlate with the changes of HA, HAS3, and versican V2 in response to E 2 . Therefore, EGF was considered as the most promising candidate for a paracrine mediator released in response to E 2 from keratinocytes that may in turn regulate HAS3 and versican V2 in dermal fibroblasts.
The effect of EGF on dermal fibroblasts was subsequently examined in vitro. Fig. 6 shows that EGF induces HAS3 mRNA, HAS3 protein, and HA synthesis in skin fibroblasts (Fig. 6,  A-C). On the immunoblots, the same lower molecular weight band as described in Fig. 2C was detected. Quantification (data not shown) revealed the same regulation of this smaller protein as the mature HAS3. Furthermore, EGF induced total VERSICAN mRNA (Fig. 6D), versican V2 protein (Fig. 6E), and VERSICAN V2 mRNA (Fig. 6F). Next, it was addressed whether UVB and E 2 affect the release of EGF from keratinocytes. Of note, E 2 induced expression and release of EGF in both nonirradiated and irradiated keratinocytes (Fig. 7, A and B) as determined by quantitative real time RT-PCR and ELISA. In dermal fibroblasts EGF expression was not stimulated by E 2 and UVB under the current experimental conditions (data not shown). Next, conditioned medium derived from keratinocytes exposed to E 2 and E 2 plus UVB was used to stimulate dermal fibroblasts as illustrated in Fig. 7C. The conditioned medium from keratinocytes exposed to E 2 and E 2 plus UVB was found to induce HAS3 and VERSICAN V2 mRNA expression (Fig. 7, D and E). The induction of HAS3 mRNA and versican V2 mRNA was blocked by both an EGF receptor kinase inhibitor, erlotinib, and by a neutralizing antibody against EGF (Fig. 7, D and E). Furthermore, the supernatants of E 2 -treated keratinocytes stimulated the proliferation of fibroblasts as determined by [ 3 H]thymidine incorporation (Fig. 7F). This pro-proliferative effect of conditioned medium from E 2 -stimulated keratinocytes was EGF-dependent because erlotinib and a neutralizing antibody against EGF were able to block the effect (Fig. 7F). Next, it was tested whether HAS3 induction might be responsible for this pro-proliferative effect. Indeed, lentiviral knockdown of HAS3 (suppression of HAS3 mRNA to 0.6 Ϯ 0.1-fold of scrambled control, p Ͻ 0.05, n ϭ 6) prohibited proliferative response in fibroblasts induced by the conditioned keratinocyte medium (Fig. 7G). Therefore, it is proposed (i) that E 2 induces expression and release of EGF from keratinocytes in vivo and (ii) that EGF induces in a paracrine manner HAS3, HA, and versican expression in fibroblasts of the papillary dermis.
Differential Effects of E 2 on Dermal Cell Proliferation and Inflammation-Next, it was attempted to link the specific changes of dermal HA and versican to functional effects in the dermis during photoaging. The proliferation of dermal fibroblasts as determined by Ki-67 immunostaining and mRNA expression was increased by E 2 both in nonirradiated and in UVB-irradiated skin (Fig. 8, A and B). This was further supported by the induction of Tpx2 (microtubule-associated homolog) mRNA expression, which is also indicative for proliferative activity (Fig. 8C). HA and versican are both thought to contribute to proliferation of dermal fibroblasts (see Fig. 7G). It is conceivable that this proliferative response to E 2 is in part due to HA synthesis via HAS3 and possibly also increased versican V2.
It has been suggested that HA plays a role in inflammation either as a result of generation of HA fragments or the formation of supramolecular complexes of HA and HA-binding proteins that form HA cables and allow monocyte/macrophage adhesion. Interestingly, in the UVB-irradiated dermis of OVX mice, which contained a reduced amount of HA and an elevated amount of versican, Mac-2 immunostaining, immunoblotting, and mRNA expression, indicative for macrophage content, were increased (Fig. 9, A-C). Furthermore, Mac-2 mRNA and protein expression were strongly reduced by E 2 suggesting an anti-inflammatory effect of E 2 during skin aging (Fig. 9, A-C). To further support this hypothesis, the expression profiles of the chemokines and cytokines Mcp-1, IL6, and Tnf␣ were analyzed and found to be suppressed in at least one of the E 2 -treated groups (Fig. 9, D--F). Finally, the expression of inducible cyclooxygenase 2 (Cox2) was analyzed as an indicator of inflammatory activity. Cox2 mRNA expression and protein expression (data not shown) were also strongly suppressed by E 2 (Fig. 9G). In summary, the in vivo data are consistent with a pro-proliferative and anti-inflammatory effect of estrogen that coincided with induction of both HAS3, HA, and versican in UVB-irradiated skin (Fig. 10).

DISCUSSION
Estrogen has been known for a long time to have beneficial effects on the skin, which is particularly obvious after the onset of menopause or the initiation of estrogen treatment of postmenopausal women (21). Estrogens partially protect from skin aging by increasing thickness, reducing the wrinkling, and augmenting the moisture of the skin (28 -30). It has been demonstrated that estrogen treatment prevents loss of collagen type 1 and increases de novo collagen synthesis (31). Furthermore, estrogens increase dermal glycosaminoglycans, which in turn may contribute to the thickness and moisture of the skin (32)(33)(34). Loss of HA from the skin appears to be part of the postmenopausal aging response. In male mice, the administration of estrogen increases HA content of the skin (35). However, systematic animal studies in females addressing the regulatory effect of estrogens on dermal HA, HAS isoenzymes, HA receptors, and hyaluronidases and importantly the underlying molecular mechanisms are lacking. Furthermore, to date nothing is known about how E 2 may regulate the content and function of HA and versican in the skin. Therefore, the aim of this study was to investigate changes of HA, HAS isoenzymes, hyaluronidases, HA receptors, and the HA-binding proteoglycan versican in the context of photoaging and E 2 .
This study clearly shows that E 2 is a strong inductor of dermal HA and versican in both nonirradiated skin and UVB-irradiated skin in hairless skh-1 mice. Moreover, HAS3 was identified as the responsible target and therefore an important E 2 -dependent regulator of dermal HA homeostasis during intrinsic and extrinsic skin aging. E 2 mediates changes in gene expression by activation of intracellular E2R␣ and -␤, which translocate into the nucleus upon ligand binding and affect the expression of genes with estrogen response elements in the promoter region. In addi-tion, nongenomic E 2 effects occur through the GPR30 pathway (36). It has been shown that both E2R␣ and -␤ are expressed in the skin and specifically in skin fibroblasts (21). However, considerable variability appears with respect to the ratio of both ER subtypes in various organs, ages, and species (37,38). Therefore, the expression of E2R␣ and -␤ was determined by real time PCR in hairless skh-1 mice. The E2R␣ receptor was strongly expressed in the skin biopsies, whereas the E2R␤ receptor was not detectable. Therefore, the above mentioned effects of E 2 on dermal HA and versican during photoaging were attributed to E2R␣. However, the contribution of nongenomic effects cannot be excluded.
Because the in vivo experiments suggested that induction of HAS3 and versican V2 expression likely played a key role in the estrogen response during photoaging, it was addressed whether E 2 directly induced these genes in fibroblasts. However, no evidence for direct transcriptional induction of HAS3 and VERSICAN V2 gene expression by E 2 was found. As an alternative explanation, the induction of paracrine factors by E 2 that subsequently induce expression of HAS3 and VERSICAN appeared plausible. This hypothesis is in line with the report that E 2 indeed affects dermal ECM via paracrine mechanisms in vivo. It has been shown that postmenopausal loss of collagen might be due to lack of E 2 -mediated expression of transforming growth factor ␤1 (TGF␤1) (39).
Here, EGF displayed an expression pattern in response to E 2 substitution that paralleled the changes of HA and HAS3 both in photoaged skin and in nonirradiated skin. Specifically, EGF mRNA and protein showed a trend to reduced expression by OVX and strong induction by E 2 . EGF is known to induce the expression of HAS2 and HAS3 in keratinocytes, thereby promoting the pro-proliferative and migratory response to EGF (40,41). Because skin extracts were used to determine Egf mRNA and protein, which reflect epidermis plus dermis, both keratinocytes and fibroblasts might be the source of EGF in response to E 2 . However, immunostaining of EGF already suggested a pronounced EGF response in the epidermis. The in vitro experiments confirmed that E 2 indeed induced EGF expression in keratinocytes but not in dermal fibroblasts (data not shown). Subsequently, EGF released in vitro in response to E 2 and UVB from keratinocytes was shown to stimulate HAS3, versican V2 expression, and proliferation in human skin fibroblasts. The pro-proliferative effect was shown to be dependent on EGF and HAS3 as shown by EGF inhibitors and lentiviral transduction with shHAS3. Therefore, it is proposed that E 2 stimulates keratinocytes to release EGF, which in turn induces HAS3 and versican V2 expression in dermal fibroblasts. The limitation of the in vitro experiments is that they reflect an acute response to single UVB irradiation, whereas in vivo repetitive chronic dosing of UVB was applied. However, the results are in agreement with the observation in other biologic systems that EGF is involved in various estrogen-induced responses (42,43) or mimics E 2 responses (44) both in vitro and in vivo. FIGURE 7. EGF release from keratinocytes in response to E 2 and UVB. A, keratinocytes were treated with 100 nM E 2 , 100 mJ/cm 2 UVB, or both. After 24 h, EGF mRNA expression was analyzed by real time RT-PCR. B, EGF protein was determined by ELISA in cell culture supernatants of keratinocytes treated as described in A. C, schematic representation of the experiments shown in D-G using keratinocyte-conditioned medium to investigate paracrine EGF effects. D and E, supernatants as shown in B were used to stimulate human skin fibroblasts. After 24 h, HAS3 mRNA expression and VERSICAN V2 mRNA expression were determined by real time RT-PCR in the presence or absence of 3 M erlotinib, EGF neutralizing antibody (EGF nAB, 0.35 g/ml), or isotype control IgG (0.35 g/ml). F, cell culture supernatants as in B were used to stimulate fibroblasts. [ 3 H]Thymidine incorporation was determined as a measure of DNA synthesis and proliferation; G, [ 3 H]thymidine incorporation in response to the supernatants as in B in fibroblasts pretreated with scrambled or HAS3-targeting lentiviral shRNA; n ϭ 3-6, mean Ϯ S.E.; *, p Ͻ 0.05 versus control.
These findings might also be relevant with respect to the aged phenotype of fibroblasts. Interestingly, age-associated resistance to phenotypic activation of fibroblasts is accompanied by reduced HA synthesis and the failure to up-regulate HA in response to growth factors (45). EGF appears to be involved in this process, because EGF signaling is critically required for TGF␤1-induced HA synthesis (46,47). Furthermore, fibroblast proliferation is stimulated by EGF, and in aging fibroblasts this proliferative response declines partly due to down-regulation of the EGF receptor (48,49). Thus, EGF is a growth factor that opposes the aging phenotype of dermal fibroblasts, and this response involves in part HA synthesis and signaling. It is there-  . Inflammatory responses to E 2 and UVB in vivo. The inflammatory response was analyzed in the dermis of skin biopsies derived from 18-week-old hairless skh-1 mice treated as described in the legend of Fig. 1. A, immunostaining of Mac-2 as marker of inflammatory macrophages; B, MAC-2 mRNA expression; C, Mac-2 immunoblotting and quantification related to tubulin as loading control (55 kDa); D, MCP-1 mRNA expression; E, IL6 mRNA expression; F, TNF␣ mRNA expression; G, COX2 mRNA expression as an inducible gene associated with inflammation and release of inflammatory chemokines and cytokines. ϫ100x magnification; n ϭ 3-12, mean Ϯ S.E.; *, p Ͻ 0.05 versus S,P; #, p Ͻ 0.05 versus S,P UVB; ࡗ, p Ͻ 0.05 versus X,P; OE, p Ͻ 0.05 versus X,P UVB.
fore likely that induction of EGF in keratinocytes by E 2 and paracrine stimulation of HAS3 and versican V2 in dermal fibroblasts is involved in the attenuation of the aged fibroblast and skin phenotype by E 2 .
In vitro studies have to date provided ample information about the function of individual single matrix components, including collagens, HA, HAS isoenzymes, and versican. However, it is obvious that during physiologic and pathophysiologic responses, changes in matrix composition occur that simultaneously involve multiple matrix molecules. This dramatically affects the cellular microenvironment. Yet cellular responses likely represent the consequences of the combined effects of these complex changes. These include alterations in micromechanic forces and matrix receptor signaling through integrins, the HA receptors CD44 and RHAMM, and activation of alternative receptors by degraded matrix molecules such as toll-like receptors that mediate "danger signals" (50). Therefore, a possibly important finding of this study was that photoaging resulted in a dermal matrix characterized by low HA content and high versican content. At present, it is unknown what the consequence of this shift in the relative abundance of HA and versican may be for the phenotype of cells and the aged phenotype of the skin. It is remarkable, however, that treatment with E 2 increases both HA and versican resulting in a matrix with high HA and high versican content. Versican has four splicing variants that are characterized by loss of glycosaminoglycanbearing domains and therefore major differences in molecular weight and chondroitin sulfate content (51). The functions of the different splice variants appear to differ, in part opposing each other, in various biologic systems (52). The presented data point toward versican V2 as the target of E 2 /EGF-mediated regulation in the skin. Although the dermal function of versican V2 has not been addressed, in the central nervous system it has been suggested to regulate matrix assembly (53). Therefore, versican V2 might be an interesting candidate for regulating matrix assembly and function in skin homeostasis and skin aging.
In this study, the proliferative capacity of fibroblasts in the dermis was positively correlated with increased HA and versican content after E 2 treatment. This is in line with the hypothesis that pericellular matrix consisting of both pericellular HA and versican promotes proliferation (20). In vitro knockdown of FIGURE 10. UVB and E 2 -mediated remodeling of dermal matrix during extrinsic aging. Schematic summary of the working hypothesis that was deduced from the presented data is shown. After chronic UVB exposure, the dermal matrix undergoes remodeling from HA-rich and versican-poor matrix with low proliferative and low inflammatory activity (left panel) to a HA-low and versican-rich ECM characterized by low proliferative but high inflammatory activity (middle panel). E 2 via paracrine release of EGF from the epidermis up-regulates HAS3 and versican that cause transition into HA-rich and versican-rich dermal matrix with higher proliferative activity and reduced inflammation. This E 2 -mediated matrix remodeling partially overwrites the UVB-induced degeneration of the dermal matrix possibly setting the stage for skin regeneration due to increased pro-proliferative activity and by inhibition of inflammatory responses.
HAS3 prevented fibroblast proliferation in response to EGF secreted into conditioned medium from E 2 -stimulated keratinocytes. The role of versican splice variant V2 in the E 2 response in the dermis should be addressed in future studies. An additional effect of estrogen was to reduce the content of inflammatory macrophages and the expression of MCP-1 and COX2 in nonirradiated and UVB irradiated skin. It is believed that the supramolecular structure of the pericellular matrix and the recruited molecules determine whether the matrix acts homeostatic (5), pro-inflammatory (54,55), or pro-migratory (20). Therefore, consideration for future studies might be that the ratio of HA and different versican splice variants plays a role in the inflammatory response to UVB in the dermis.
In conclusion, E 2 represents a critical regulator of dermal HA and versican content during photoaging, through release of EGF from keratinocytes, which in turn induces the expression of HAS3 and versican V2 in dermal fibroblasts in a paracrine manner. As a consequence, fibroblast proliferation is increased by E 2 , and inflammation is inhibited. Therefore, this study identifies novel molecular targets of E 2 that may contribute to the protective function of this hormone on dermal matrix during intrinsic and extrinsic aging responses.