Safer topical treatment for inflammation using 5α-tetrahydrocorticosterone in mouse models

Graphical abstract


Introduction
Inflammatory skin disease is highly prevalent worldwide: according to estimates published in 2012, eczema affects approximately 230 million people globally, and topical anti-inflammatory glucocorticoids are the most common treatment [1]. However, the therapeutic benefits of these drugs are offset by severe side-effects [2].
Following application of glucocorticoids to the skin, adverse effects manifest locally (skin atrophy and impaired wound healing) and systemically (e.g. osteoporosis, abdominal obesity, and glaucoma) [3]. Skin atrophy is characterised by increased transparency and striae, due to suppression of cell proliferation and inhibition of collagen synthesis, whereas delayed wound healing is attributed to dysregulation of keratinocyte proliferation, fibroblastic activity, and angiogenesis causing delayed formation of granulation tissue [4,5]. These factors restrict the use of more potent glucocorticoids and also the length of treatment, particularly in vulnerable patients such as infants and the elderly. Accordingly, novel drugs or delivery modes have been sought to improve the therapeutic index of topical steroids. This may be achieved by ''dissociated glucocorticoids" which invoke trans-repression through the glucocorticoid receptor (GR) without trans-activation [6].
5a-Tetrahydrocorticosterone (5aTHB) is a naturally occurring glucocorticoid metabolite, formed by reduction of the A-ring of corticosterone, the principle endogenous glucocorticoid in rodents. Two 5a-reductase isozymes catalyse the rate-determining step in 5aTHB synthesis from corticosterone: the type I isozyme is expressed in the liver, kidney and skin; the type II isozyme is expressed principally in the liver, genital tract, prostate and skin. Recent studies have suggested that 5aTHB exhibits a dissociated profile of action, possibly acting through GR. 5aTHB displayed anti-inflammatory properties when administered subcutaneously in mice subjected to thioglycollate-induced peritonitis, but did not induce the chronic adverse metabolic changes (such as increased insulin and body mass) produced by corticosterone [7]. 5aTHB and corticosterone suppressed infiltration of neutrophils into the inflamed peritoneum to a similar extent, but 5aTHB had a lesser effect on macrophage recruitment, suggesting a different profile of action. When infused systemically, 5aTHB demonstrated rapid clearance from the systemic circulation [7], likely limiting its oral use, but the same characteristic would be beneficial in topical treatments in which systemic effects are unwanted.
This investigation addressed the hypothesis that topical 5aTHB can effectively reduce skin inflammation through a GR-dependent mechanism, but with limited effects on processes such as angiogenesis and collagen deposition, the inhibition of which underpin most common problems associated with topical glucocorticoid therapy. The efficacy of topical 5aTHB application for treating skin inflammation was compared with that of corticosterone in vivo using croton oil-induced dermatitis and the role of GR was assessed by pre-administration of the GR antagonist mifepristone (RU486). This is a steroid-responsive model of dermal inflammation, driven primarily by neutrophil accumulation, commonly used for testing anti-inflammatory treatments for irritant dermatitis [8,9]. An in vivo model of sponge implantation allowed comparison of the effects of corticosterone and 5aTHB on inflammation, angiogenesis and collagen deposition.

Animal welfare, models and ethical statements
Male mice (C57BL/6, 8-10 week old) were from Harlan Laboratories (Shardlow, UK). Animals were allowed to acclimatise for one week prior to experimentation, maintained under controlled conditions of light (lights on 0700-1900 h) and temperature (18-20°C) and allowed free access to standard chow (Special Diet Services, Witham, UK) and drinking water. All experimental procedures were performed under UK Home Office guidelines.

In vivo model of inflammatory angiogenesis
An autoclaved cubic polyurethane sponge (0.5 cm 3 ; grade XE1700V, Caligen Foam Ltd., Accrington, U.K.) was implanted sub-cutaneously on each flank of mice anaesthetised with isoflurane (Merial, Harlow, UK). For the delivery of steroids, one silastic pellet (Silastic 20 medical grade, Dow Corning Ò , Midland, USA) impregnated with vehicle (silastic only), corticosterone (3 mg/pellet) or 5aTHB (3 or 15 mg/pellet) was inserted in each sponge. Validation of the preparation and release properties of these pellets has been reported [10]. Each animal in the treatment groups had an intervention-impregnated sponge (corticosterone or 5aTHB) implanted on the right side, and a vehicle sponge on the left. Animals in the control group had vehicle sponges implanted on each side. 21 days after surgery, mice were euthanized, sponges excised and pellets removed. Sponges were bisected and one half was preserved in 10% formalin (in PBS) and the other immersed in RNAlater solution.

Histological analysis
Paraffin-embedded ear tissue and sponge slices (5 lm) were rehydrated and stained with haematoxylin and eosin (H&E) or with Picrosirius red solution (PRS) for collagen content analysis (sponges only), prior to dehydration and mounting. Images were captured using the software QCapture Pro 7 (QImaging, Canada).
On H&E stained ear sections the width of the dermis (in lm) and cell infiltration (number of nuclei) were assessed as further measurements of swelling and inflammation at both sides of the central cartilage layer. For each parameter three different areas were analysed and the mean value calculated.

Quantitation of neutrophils by myeloperoxidase activity
Myeloperoxidase (MPO) activity in ear tissue was quantified using a fluorometric detection kit (ENZO Life Sciences (Exeter, UK)).

Quantification of angiogenesis, macrophage infiltration and collagen content in sponges
For each sponge section vessels were counted in three different areas at magnification 100X. Vessels in sections stained with H&E were recognized by their round appearance and the presence of erythrocytes in the lumen. Vessels identified in sections immunostained for CD31 and aSMA marker were counted. The numbers of newly formed vessels and vessels positive for either CD31 or aSMA are represented as percentage of the vehicle group (mean value = 100%). Macrophages were counted as the number of cells positive for the marker F4/80 and expressed as a proportion of the total number of cells (blue nuclei; magnification 200X). PRS staining was quantified calculating the number of red pixels at a magnification of 100X using ImageJ software (NIH, USA).

Immunostaining for CD31, aSMA and F4/80 in sponges
Immunostaining of sponge slices was performed using a Leica Staining Robot with robotic antigen retrieval for F4/80 using trypsin solution (0.5 mg/mL in PBS, 10 min, 37°C), or for aSMA manual antibody retrieval using a pressure cooker under standard conditions (citrate NCL pH 6 buffer); antigen retrieval was not performed for CD31. For F4/80 and aSMA, samples were blocked UK)) prior to incubation with polymer (30 min, F4/80 Impress Kit as above; aSMA as above). For CD31 a Leica Refine Kit (as above) was used, which included a hydrogen peroxide block (5 min), primary antibody (120 min, 1:200 (Abcam)) and polymer incubation (15 min). Immunostaining was completed with incubation with 3,3 0 -diaminobenzidine (10 min) and counterstaining with haematoxylin (5 min), both from the Leica Refine Kit.

Total RNA extraction and PCR
Total RNA was extracted from ears and sponges using the RNeasy Fibrous Tissue Mini Kit, and first strand cDNA synthesis was performed using the QuantiTect Reverse Transcription kit (Qiagen Ltd, West Sussex, UK). Real-time quantitative PCR was performed using a LightCycler Ò 480 (Roche Diagnostics, Mannheim, Germany) with primers (6 pmol each, Table 1), corresponding 5 0 FAMlabelled probe (UPL, 2 pmol) and LightCycler Ò 480 Probes Master [11]. Data were normalized for the mean of the transcript abundance of two housekeeping genes, TATA-binding protein and glyceraldehyde 3-phosphate dehydrogenase, the abundances of which did not differ between groups, and represented as percentage of the respective control group whose mean value was set to 100%.

Data analysis
All data were analysed using GraphPad Prism6 software and presented as mean ± SEM, using statistical tests as indicated. Significant differences were reported when p 6 0.05. For histological analysis, the operator was blinded to treatment.

5aTHB reduces croton oil-induced inflammation with a different time course than corticosterone
Increasing concentrations of corticosterone reduced croton oilinduced ear swelling in a dose-dependent manner after 6 and 24 h Histological and biochemical analyses showed that croton oilinduced swelling was accompanied by enlargement of the dermal layer at both 6 and 24 h (Fig. 2a, b, e, f), cell infiltration ( Fig. 2g-i) and increased MPO activity (Fig. 2j, k). Application of corticosterone reduced dermal width at 6 and 24 h but 5aTHB only had this effect at 24 h ( Fig. 2a, b, e, f). Corticosterone and 5aTHB reduced cell accumulation to a similar extent at 24 h, although corticosterone also had this effect after only 6 h (Fig. 2h, i). Moreover, 5aTHB was more effective than corticosterone in suppressing MPO activity, an indicator of neutrophil activity, after 24 h, but did not have any effect at 6 h ( Fig. 2j, k).

5aTHB and corticosterone have discrepant effects on de novo angiogenesis in sponges
Sponges containing vehicle pellets (V) were bright red, suggesting extensive vascularization, whereas those with corticosterone pellets (3 mg) were pale (similar to pre-implantation sponges) indicating less vascularization (Fig. 5a). Sponges with 5aTHB (3 mg) had a similar appearance to vehicle-treated sponges, whereas those with 15 mg 5aTHB were somewhat paler, although the lack of vascularization was not as extensive as that shown by sponges containing corticosterone. Vessels were identified by H&E staining (Fig. 5b), and immunoreactivity for CD31 (Fig. 5c) and aSMA (Fig. 5d) associated with vascular structures. Quantification showed that corticosterone dramatically reduced vessel density whether assessed by H&E (Fig. 5e), CD31 (Fig. 5f) or aSMA (Fig. 5g) staining. In contrast, an equivalent dose (3 mg) of 5aTHB in sponges had little effect on vascular density (Fig. 5e, f, g). The higher dose (15 mg: equipotent in anti-inflammatory model) reduced vascular density assessed by H&E (Fig. 5e) and CD31 (Fig. 5f) staining but less dramatically than with corticosterone. Strikingly, 5aTHB had little effect on the density of aSMA positive vessels (Fig. 5g). Macrophage infiltration of sponges was observed (Fig. 6a, c) and was reduced following treatment with corticosterone or 5aTHB, although this did not achieve significance at the lower dose of 5aTHB (Fig. 6c). Collagen staining (Fig. 6b, d) was lower in sponges treated with corticosterone, compared with vehicle-treated controls, whereas 5aTHB had no effect at either concentration.

Discussion
Anti-inflammatory properties, but a safer systemic side effect profile, of 5aTHB, an A-ring reduced metabolite of corticosterone, have been demonstrated previously in vivo in a murine model of thioglycollate-induced peritonitis [7]. This investigation addressed the hypotheses that 5aTHB has advantages over conventional glucocorticoids as a topical treatment for irritant dermatitis, and that its actions are mediated through GR. 5aTHB was demonstrated to be an effective topical anti-inflammatory agent in an in vivo model of irritant dermatitis induced by croton oil, with effects mainly on pathways regulating vascular permeability to cell infiltration rather than suppression of pro-inflammatory markers. However, experiments using the GR antagonist RU486 suggested these effects may not be mediated by occupation of GR. Furthermore, in comparison with corticosterone, use of 5aTHB may be associated with fewer adverse local effects (less inhibition of angiogenesis and collagen deposition).
The direct suppression of inflammation by topical 5aTHB was assessed using croton oil-induced irritant dermatitis, a wellcharacterised, steroid-responsive model, used extensively in the pharmaceutical industry [6,[12][13][14]. Croton oil promotes the release of cytokines and chemokines from keratinocytes and other cells, stimulating in turn the release of pro-inflammatory cytokines such as IL1b and IL6 from epidermal and dermal cells, causing swelling and infiltration of leukocytes [15][16][17][18]. The predominant effect at the early stages of the process is tissue oedema which peaks 6 h-post treatment; this phase is followed by cell infiltration (mainly neutrophils) which is at its highest between 12 and 24 h [19][20][21][22]. This time course of changes was recapitulated here: at 6 h croton oil produced a marked swelling accompanied by cell infiltration and increased MPO activity, whereas at 24 h swelling persisted but was reduced, while cell infiltration and MPO activity increased further. This model has been used extensively to test novel antiinflammatory drugs, including glucocorticoids (e.g. dexametha-   sone, betamethasone and hydrocortisone) [21,[23][24][25][26]. Here, the effects of 5aTHB on skin inflammation were compared with those of corticosterone, the active rodent glucocorticoid. Previous reports of corticosterone efficacy were not available but our data suggest that at 6 h it is twice as potent as one of the most frequently employed topical glucocorticoids, hydrocortisone [21,27]. Notably, the dose-dependent reduction of dermal inflammation by 5aTHB exhibited a slower time course than corticosterone with no effect at 6 h. Significant suppression of inflammation was, however, achieved at 24 h, albeit requiring a 3-5 fold higher dose than corticosterone to match efficacy. This suggests that 5aTHB may predominantly target cell recruitment, which is prominent at 24 h post CO-treatment, with neutrophils believed to be the most abundant cell type present [19][20][21][22]. Analysis of MPO activity showed that both steroids reduced neutrophil infiltration; interestingly, this was suppressed more effectively by 5aTHB than by corticosterone at 24 h. The lack of response to 5aTHB to reduce swelling at 6 h was accompanied by a failure to reduce cell number.
Experiments with the GR antagonist RU486 suggested that the anti-inflammatory effect of B on swelling is mediated by GR. However, this seems not to be the case for 5aTHB. The fact that the antiinflammatory actions of 5aTHB were not antagonised by RU486 was unexpected, albeit in keeping with its different time course of action to corticosterone. 5aTHB is a glucocorticoid metabolite, previously shown capable of displacing dexamethasone from rat GR in vitro [28]. RU486 acts as a GR antagonist by binding to the ligand-binding domain, blocking the recruitment of essential GR co-activators [29,30], and if 5aTHB does operate through GR it does so by an unconventional mechanism, perhaps through allosteric means. Further genetic and pharmacological studies are now required to elucidate the mechanism of action of 5aTHB.
To start to address this, the profile of action of 5aTHB was explored further through expression profiling and compared with responses produced by corticosterone. At 24 h, the impact of 5aTHB on expression of genes involved in regulation of inflammation and vascular homeostasis in the CO model was strikingly similar to those of corticosterone. A noted difference was in the upregulation of Dusp1 by 5aTHB in contrast to corticosterone. DUSP1 is a negative regulator of the MAPK signal transduction pathway, and is increased in response to inflammatory and irritant stimuli, leading to decreased production of cytokines, chemokines and adhesion molecules [31,32]. A glucocorticoid-mediated increase in its protein abundance has been reported previously in many cell types [33][34][35][36][37]. DUSP1 is also upregulated by nonsteroidal anti-inflammatory compounds [32] and by new selective modulators of GR [38]. 5aTHB and corticosterone also increased transcripts of the anti-inflammatory gene AnxA1 (lipocortin). Lipocortin was originally described as being responsible for the antiinflammatory properties of glucocorticoids [39], again diminishing neutrophil recruitment by suppressing adhesion to endothelial cells [39]. Both corticosterone and 5aTHB strongly suppressed many factors involved with tissue permeability and cell recruitment. The effects of 5aTHB on expression of ECM proteins were less marked than those of corticosterone, whereas both lacked effects on transcription of most collagen genes, only decreasing Col4a1 transcripts. Type IV collagen is one of the minor components of the skin (less than 5%) [5], mainly forming the basement membrane; as a consequence both steroids are likely to have little impact on skin thickness of croton-oil treated ears after 24 h. However accurate measures of skin thickness cannot be made in the inflamed skin in vivo. Previous in vitro microarray studies on cultures of primary human keratinocytes showed that dexamethasone, a glucocorticoid known to cause skin thinning, decreases transcripts of Mmp9, Mmp10, Keratin6, FilaminA and Col4a1 after 24 h incubation [40], a profile similar to that of corticosterone in our model.
Two of the major side effects of topical anti-inflammatory glucocorticoids are skin thinning and delayed wound healing, requiring careful assessment of potential adverse effects for any proposed topical steroid therapy. Angiogenesis and collagen turnover are important factors in skin homeostasis, with angiogenesis particularly important for dermal healing [41]. Following the demonstration that cortisone inhibits angiogenesis in vitro, a number of ''angiostatic steroids" have been identified [42]. These include 5b-tetrahydrocortisol, one of the A-ring reduced metabolites of cortisol (the major glucocorticoid in humans). Here the impact of 5aTHB on angiogenesis, inflammation and collagen deposition was studied in response to implantation of subcutaneous sponges [43]. This model is steroid sensitive [44] and equipotent anti-inflammatory doses of corticosterone and 5aTHB (based on results from the croton-oil model) were administered in the sponges. 5aTHB reduced angiogenesis but to a lesser degree than, and with a different pattern to, corticosterone. It was notable that both compounds reduced vessel density, predominantly by reducing the number of CD31-positive vessels. However, unlike corticosterone, 5aTHB did not reduce the number of aSMA positive vessels. CD31-staining identifies endothelial cells lining vascular structures and can be used to detect small (capillaries) and larger vessels [45]. In contrast, aSMA, staining detects pericytes and smooth muscle cells, and can be used to identify larger/maturing vessels with a functional medial layer [46]. Thus, these results suggest that, although 5aTHB does reduce new vessel formation, it does not adversely affect the number of mature vessels and, hence, may be less likely to attenuate wound healing. Caveats do, however, apply in that retention of 5aTHB and corticosterone within sponges was not directly compared in this study.
In terms of local inflammation, both corticosterone and 5aTHB inhibited macrophage infiltration. Inflammatory cells, in particular monocytes/macrophages, are central players in angiogenesis as they are a source of pro-angiogenic cytokines [47]. Indeed, it is often difficult to distinguish between direct angiostatic properties and those that are a consequence of reduced inflammation. In the croton oil model there was a bigger effect of 5aTHB compared to corticosterone on neutrophil infiltration, whereas this is not true regarding the recruitment of macrophages analysed in the angiogenesis model. This suggests that neutrophils might be the preferential cellular target for 5aTHB in an inflammatory setting. Importantly, macrophages have a pivotal role in wound repair [41], and a compound with fewer inhibitory effects on their infiltration might be more desirable.
Relatively few changes in gene transcripts for factors that regulate aspects of vascular function were consistent with 5aTHB having a limited effect on the developing vasculature, with only a modest reduction in Pecam1 which support the immunostaining data for CD31. In contrast, corticosterone inhibited expression of most of the factors assessed (Icam1, Pecam1, VE-cadherin, E-selectin and Acta2), consistent with its potent anti-angiogenic properties [44], and again reinforce the data obtained with the immunostaining of CD31 and aSMA in sponge sections. These factors contribute to stabilisation of cell-cell contact between endothelial cells [45], and the dynamics of cell rearrangement central to vessel formation [48,49]. The only exceptions were the increased expression of Vegfa and Vcam1. Vegfa has a key role in stimulating angiogenesis [50][51][52], so would be expected to be reduced by corticosterone (although it is possible that the increased expression is a consequence of insufficient angiogenesis in the sponges). However, this was accompanied by reduced Vegfr2 transcript number, consistent with inhibition of angiogenesis seen in previous studies with other glucocorticoids [53]. VEGFR2 is the main receptor responsible for the majority of the effects of VEGF on endothelial cells [53] and is associated with proliferation of new vessels during wound healing [53].
Intriguingly, despite its potent anti-inflammatory properties, corticosterone only produced a small reduction in transcripts of pro-inflammatory genes. 5aTHB, despite reducing macrophage infiltration, did not reduce expression of any of the transcripts analysed and, indeed, increased abundance of transcripts of Mcp1. A similar induction of Mcp1/MCP1 promoted vessel growth in infarcted hearts in mice [54], suggesting this may be a compensatory mechanism in response to an ischemic event. In previous studies of peritonitis 5aTHB also suppressed inflammatory cell recruitment, but with a more profound effect on neutrophils compared with macrophages [7].
The extra-cellular matrix (ECM) plays a vital role in wound healing and many studies show that glucocorticoids cause skin damage by depleting collagen isoforms [55]. This again could be modelled in the implanted sponges with consistent reductions in collagen deposition and expression of the four collagen isoform transcripts analysed in corticosterone-treated sponges. Importantly, 5aTHB had little effect on collagen staining, significantly reducing only Col1a2 expression. Skin collagen fibrils are made mostly of type 1 collagen, a heterodimer composed by three alpha chains, two encoded by Col1a1 and one by Col1a2 [5]. The lack of effects on Col1a1 transcripts by 5aTHB might explain why this compound did not affect the overall collagen content in sponges. Furthermore, the absence of effect on Col4a1 transcripts suggest a reduced impact, compared to corticosterone, on angiogenesis, as type IV collagen is a major component of the basement membrane (facilitating the differentiation of endothelial cells during formation of capillary-like structures) [56]. This contrasts with the dermatitis model where both steroids had a small effect to reduce Col4a1, but over a shorter time course. Metalloproteinase enzymes (MMPs) influence wound healing, predominantly by promoting ECM degradation and cell migration [57], whereas tissue inhibitors of MMPs (TIMPs) oppose these actions. Corticosterone, but not 5aTHB, decreased expression of MMPs (Mmp10 and Mmp2) within the sponges, which may contribute to alterations in matrix composition and to inhibition of angiogenesis [58,59].
Studies of mechanism in the croton oil and sponge model were mainly limited to analysis of transcript abundance, due to paucity of tissue. While complementary analysis confirmed effects on CD31 (Pecam1), aSMA (Acta2) and collagen protein, GCs have been shown to inhibit translation rather than transcription for some molecules involved in inflammation [60,61]. As a consequence, while no significant effect of 5aTHB or B was found on transcription of cytokines, it remains possible that they may be affecting the translation of those transcripts.
In conclusion, 5aTHB is an effective topical anti-inflammatory steroid, reducing acute skin inflammation possibly binding to a different receptor than GR. Its limited effects on angiogenesis and de novo deposition of collagen (combined with previous reports of its rapid systemic elimination and, therefore, reduced metabolic side-effects), suggest it may have potential as a novel topical antiinflammatory treatment and careful dissection of its mechanism of action is now required.

Conflicts of interest
RA and BRW are inventors on a relevant patent held by the University of Edinburgh.