Oral Vitamin D Rapidly Attenuates Inﬂammation from Sunburn: An Interventional Study

The diverse immunomodulatory effects of vitamin D are increasingly being recognized. However, the ability of oral vitamin D to modulate acute inﬂammation in vivo has not been established in humans. In a double-blinded, placebo-controlled interventional trial, 20 healthy adults were randomized to receive either placebo or a high dose of vitamin D 3 (cholecalciferol) one hour after experimental sunburn induced by an erythemogenic dose of UVR. Compared with placebo, participants receiving vitamin D 3 (200,000 international units) demonstrated reduced expression of proinﬂammatory mediators tumor necrosis factor- a ( P ¼ 0.04) and inducible nitric oxide synthase ( P ¼ 0.02) in skin biopsy specimens 48 hours after experimental sunburn. A blinded, unsupervised hierarchical clustering of participants based on global gene expression proﬁles revealed that participants with signiﬁcantly higher serum vitamin D 3 levels after treatment ( P ¼ 0.007) demonstrated increased skin expression of the anti-inﬂammatory mediator arginase-1 ( P ¼ 0.005), and a sustained reduction in skin redness ( P ¼ 0.02), correlating with signiﬁcant expression of genes related to skin barrier repair. In contrast, participants with lower serum vitamin D 3 levels had signiﬁcant expression of proinﬂammatory genes. Together the data may have broad implications for the immunotherapeutic properties of vitamin D in skin homeostasis, and implicate arginase-1 upregulation as a previously unreported mechanism by which vitamin D exerts anti-inﬂammatory effects in humans.


INTRODUCTION
Vitamin D is a ubiquitous fat-soluble hormone important in calcium homeostasis and bone metabolism (Jackson et al., 2006). The majority of vitamin D arises from de novo synthesis in the skin triggered by UVR, with smaller contributions from dietary sources (Bikle, 2011). Although considerable attention has been placed on vitamin D deficiency and optimizing supplementation strategies, appreciation for the diverse biological effects and long-term outcomes of vitamin D now include the modulation of immune responses, inflammatory disease, cardiovascular health, and carcinogenesis (Giovannucci et al., 2006;Martins et al., 2007;Rosen, 2011;Sanders et al., 2010;Wobke et al., 2014). However, there is a lack of evidence demonstrating that intervention with vitamin D is capable of resolving acute inflammation in target tissues and organs in humans.
Keratinocytes and macrophages produce active vitamin D within the skin (Baeke et al., 2010;Bikle et al., 1986). Vitamin D has pleiotropic effects on the immune system, including the enhancement of antimicrobial responses, induction of autophagy, and suppression of proinflammatory mediators, including tumor necrosis factor-a (TNF-a) (Di Rosa et al., 2012;Liu et al., 2006;Zhang et al., 2012). We recently demonstrated that intervention with a single dose of vitamin D is capable of rapidly attenuating an inflammatory response in a mouse model of chemical induced skin injury through inhibition of inducible nitric oxide synthase (iNOS; or NOS2 gene) and TNF-a (or TNFA gene) by activated macrophages (Au et al., 2015). Therefore, we designed a pilot, proof-of-principle interventional study in humans, modeled after a randomized, double-blinded, placebo-controlled clinical trial, to test the hypothesis that a single high dose of oral vitamin D 3 (cholecalciferol) would be capable of rapidly attenuating experimental sunburn induced by simulated solar radiation (SSR).

Dose-dependent response of high-dose oral vitamin D 3 and UV irradiation
Subjects randomized to receive various doses of vitamin D were divided into 4 categories (Figure 1a), whereas the overall study design constituting the control phase and an investigative phase (Figure 1b) required monitoring of subjects for a total of 6 weeks. The randomized treatment groups did not differ in their baseline characteristics (Table 1). No participant was taking supplemental vitamin D 3 before study initiation. Serum 25-hydroxyvitamin D 3 (25(OH)D 3 ), a marker of vitamin D 3 stores, increased after treatment in a vitamin D 3 dose-dependent fashion (Supplementary Figure S1a online). Similar trends were observed for the active form of vitamin D 3 , 1,25(OH) 2 D 3 , as well as an inactive breakdown product, 24,25(OH) 2 D 3 (Supplementary Figure S1b). No measured vitamin D 3 metabolite increased into a toxic range in any of the treatment groups throughout the study period. Furthermore, there were no instances of clinically significant hypercalcemia occurring in any of the treatment groups throughout the study period (Supplementary Figure S2 online).
Sunburn is a stereotypical inflammatory response induced by exposure to an erythemogenic dose of UVR. Sunburn is characterized clinically by redness, mediated by dermal vasodilatation, and edema, mediated by increased vascular permeability and inflammatory cell infiltration (Clydesdale et al., 2001;Cooper et al., 1993;Ouhtit et al., 2000). Although skin redness peaks early after UVR exposure, skin thickness increases steadily for up to 2 weeks after irradiation (Clydesdale et al., 2001;Ouhtit et al., 2000). Compared with one minimal erythema dose (MED), exposure to 2MED increased skin erythema 24 hours and 48 hours after irradiation, and exposure to 3MED increased skin thickness 72 hours and 1 week after irradiation (P < 0.05 for all) (Supplementary Figure S3a, S3b online). We observed saturation of skin redness after exposure to 3MED, limiting the ability to discern subtle differences among treatment groups with this high dose of UVR.

High-dose oral vitamin D 3 attenuates skin inflammation
Clinically, irradiated skin appeared red and swollen 48 hours after UVR exposure (Figure 2a). Irradiated skin also displayed histologic evidence of structural damage as compared with nonirradiated skin, including epidermal vesiculation and edema formation, which improved in a vitamin D 3 dosedependent fashion (Figure 2b). Skin expression of TNF-a and iNOS was lower in participants receiving 200,000 international units (IU) D 3 as compared with placebo 48 hours after irradiation (P ¼ 0.04 for TNF-a; P¼ 0.02 for iNOS) (Figure 2c). With higher doses of vitamin D 3 , there was a trend for decreased skin thickness after irradiation, which reached significance in the 100,000 IU D 3 group at both 72 hours (P ¼ 0.03) and 1 week (P ¼ 0.02) after irradiation (Supplementary Table S1 Figure 1. Study design and baseline characteristics. (a) The enrollment, allocation, follow-up, and analysis of participants. (b) A schematic of the parallel phase study design. Skin erythema and thickness were measured 24 hours, 48 hours, 72 hours, and 1 week after experimental sunburn induced by an erythemogenic dose of simulated solar radiation. Participants returned 2 weeks after conclusion of the control phase of the study for the investigative phase, and were subsequently randomized to receive the study drug as a postexposure treatment 1 hour after experimental sunburn on the contralateral arm. Punch biopsies for tissue TNF-a, iNOS, and microarray analyses were obtained 48 hours after experimental sunburn in both phases of the study. iNOS, inducible nitric oxide synthase; MED, minimal erythema dose; TNF-a, tumor necrosis factor-a. the 200,000 IU D 3 group had a distinct gene expression profile that was very different from the placebo group, and most closely related to the 100,000 IU D 3 group ( Figure 2d).

Elevated serum levels of 25(OH)D 3 correlate with decreased skin redness
To further investigate a potential link between vitamin D 3 and gene expression, we analyzed the global gene expression profiles of all participants, blinded to their allocated treatment groups. The dendrogram resulting from this analysis produced two clusters of participants (Figure 3a), representing an unbiased, unsupervised hierarchical clustering of all individuals based on similarities in gene expression profiles.
Of note, ARG (arginase)-1, known to enhance tissue repair and inhibit inflammation through the utilization of iNOS precursors, was significantly downregulated in cluster 1 (P ¼ 0.016) and upregulated in cluster 2 (P ¼ 0.046) ( Figure 3a) (Bronte and Zanovello, 2005). Increased ARG1 gene expression observed in cluster 2 compared with cluster 1 was subsequently validated with quantitative real-time reverse transcriptase-PCR (P ¼ 0.005) ( Figure 3b). Additionally, confocal microscopy analysis of a representative participant from cluster 2 revealed increased expression of the arginase-1 protein localized to CD163þ macrophages after vitamin D 3 treatment ( Figure 3c).
Unblinding of the participants' demographics and treatment group allocation revealed that the two clusters of participants did not differ in their baseline characteristics (Supplementary Table S2 online). However, cluster 1 contained all participants randomized to receive placebo, as well as a mixture of participants from the various vitamin D 3 treatment groups ( Figure 3a). Cluster 2 predominately contained participants who had received higher doses of vitamin D 3 , and notably none of the participants who had received placebo. Although the two clusters had similar baseline serum 25(OH)D 3 levels, participants in cluster 2 had significantly higher 25(OH)D 3 levels after treatment as compared with participants in clusters 1 (P < 0.05 for all time points) (Figure 4a). When subjects receiving placebo were excluded from this analysis, serum 25(OH)D 3 levels after treatment remained lower for participants in cluster 1 as compared with participants in cluster 2 (P < 0.05 for 24 hours, 48 hours, and 1 week) ( Figure 4b). We will now refer to participants from cluster 2 as vitamin D 3 responders and participants from cluster 1 as vitamin D 3 nonresponders.
As indicated above, body mass index, age, gender, and baseline 25(OH)D 3 stores had no effect on the serum response to oral vitamin D 3 . Furthermore, along with higher serum 25(OH)D 3 levels, vitamin D 3 responders demonstrated a statistically significant sustained reduction in skin redness at all time points after irradiation as compared with vitamin D 3 nonresponders (P < 0.05 for all) (Figure 4c), and a trend for reduced skin thickness 1 week after irradiation (P ¼ 0.09; data not shown).

DISCUSSION
In this pilot, proof-of-principle human interventional study modeled after a randomized, double-blinded, placebocontrolled trial, we provide in vivo evidence that a single high dose of oral vitamin D 3 is capable of rapidly attenuating a local inflammatory response to UVR. Participants responding to high doses of vitamin D 3 demonstrated a sustained reduction in skin redness after experimental sunburn, as well as less epidermal structural damage, reduced Abbreviations: BMI, body mass index (the weight in kilograms divided by the square of the height in meters); CI, confidence interval; D, vitamin D 3 (cholecalciferol); FST, Fitzpatrick skin type; IU, international units. expression of proinflammatory markers in the skin, and a gene expression profile characterized by upregulation of skin barrier repair genes. This study also demonstrates that regardless of baseline serum vitamin D 3 levels, a single high dose of oral vitamin D 3 is safe, with serum vitamin D 3 and calcium concentrations remaining within a normal reference range (Rosen, 2011). The simplicity and safety of high-dose oral vitamin D 3 treatment, combined with its rapid and sustained therapeutic efficacy, suggest that these proof-ofconcept findings may ultimately be translated to routine clinical use once larger studies are performed on diverse populations of subjects (Ilahi et al., 2008;Sanders et al., 2010). Moreover, upregulation of arginase-1 is associated with the anti-inflammatory effects of vitamin D 3 in humans ( Figure 5). Although arginase has been identified to be present at physiological levels in inflammatory skin diseases, tumors, and chronic wounds, to our knowledge the induction of arginase-1 expression by vitamin D 3 in human skin in vivo is previously unreported (Abd-El-Aleem et al., Bruch-Gerharz et al., 2003;Gokmen et al., 2001). These findings suggest that arginase-1 may also be a clinically useful tissue biomarker for monitoring the immunomodulatory effects of vitamin D 3 in humans. Given the presence of a putative vitamin D 3 response element upstream of the ARG1 promoter, future studies should be aimed at defining the mechanism by which vitamin D 3 treatment activates the arginase-1 pathway (Andrukhova et al., 2014).
Exploratory analyses suggest that the host's response to vitamin D 3 intervention plays a critical role in the modulation of inflammation. Participants segregated into two clusters based on similarities in global gene expression profiles, and these two clusters differed significantly in their serum vitamin D 3 levels after treatment. The pharmacokinetic properties of oral vitamin D 3 are complex, however, and an individual's serum response to oral vitamin D 3 depends on the dose of vitamin D 3 , age, body mass index, baseline vitamin D 3 stores, and genetic polymorphisms (Didriksen et al., 2013;Ilahi et al., 2008). Large randomized, double-blinded, placebo-controlled trials will be required to elucidate factors responsible for the therapeutic variability of vitamin D 3 treatment within populations and to determine whether individuals with lower baseline vitamin D 3 levels require higher treatment doses to achieve immunomodulation. Exposure to erythemogenic doses of UVR initiates an influx of inflammatory cells into the skin, generating a microenvironment rich in inflammatory mediators (Cooper et al., 1993;Ouhtit et al., 2000). Specifically, release of TNF-a by damaged keratinocytes and other inflammatory cells plays a crucial role in initiating and sustaining UVR-induced inflammation (Clydesdale et al., 2001;Di Rosa et al., 2012). Following an inflammatory insult, classically activated M1-polarized macrophages infiltrate the skin and produce iNOS as part of an oxidative burst in an evolutionarily conserved attempt to prevent infection (Bronte and Zanovello, 2005;Mills, 2012). However, excessive production of iNOS perpetuates tissue damage, retards the resolution of inflammation, and prevents tissue repair (Bronte and Zanovello, 2005;Mills, 2012). We and others have shown using murine models that vitamin D 3 inhibits the production of TNF-a, and is capable of attenuating skin inflammation by reducing macrophage-specific iNOS production (Au et al., 2015;Zhang et al., 2012).
In the presence of retinoic acid, vitamin D 3 induces the in vitro differentiation of monocytes into alternatively activated, M2-polarized CD163þ macrophages expressing arginase-1 (Di Rosa et al., 2012;Takahashi et al., 2014). Furthermore, it has been shown that exposure to acute UVR increases endogenous retinoids in the skin of mice (Gressel et al., 2015). Taken together, our results combined with these data suggest that a potential mechanism by which vitamin D 3 mediates resolution of experimental sunburn is via the upregulation of arginase-1 by endogenous repair molecules, leading to the selective induction of antiinflammatory, M2-polarized CD163þ macrophages. Additionally, vitamin D 3 may have other protective mechanisms in skin, including reducing DNA damage and keratinocyte apoptosis after experimental sunburn, as was shown in mice treated topically with the active form of vitamin D 3 immediately after exposure to UVR (Dixon et al., 2007).
It is likely that the generation of vitamin D 3 from cholesterol precursors in the skin after UVR evolved to perform vital homeostatic functions (Bikle, 2011). Moreover, skin-resident cells are capable of locally converting vitamin D 3 into its active form, which can then signal in an intracrine, autocrine, and paracrine fashion to exert diverse biological effects (Bikle, 2011;Di Rosa et al., 2012). It is worthwhile to conjecture that vitamin D 3 may provide an "endocrine barrier" within the skin, utilizing energy derived from sunlight to reduce inflammation, and promote wound healing, tissue repair, and an enhanced epidermal barrier. This would provide the host with additional protection against environmental insults by complementing the classically described brick and mortar mechanical, melanin pigment, and Langerhans cell immunologic barriers.

Screening, randomization, and study design
The Institutional Review Board at University Hospitals Cleveland Medical Center approved this pilot study, which was conducted between March 2013 and February 2015. The study was modeled after a randomized, double-blinded, placebo-controlled trial. The trial is registered with clinicaltrials.gov (NCT02920502). Twentyseven healthy adults 18 years and older were screened for eligibility and provided written informed consent (Figure 1a). A total of 25 participants were randomized to receive, in a double-blinded fashion, either placebo or a single oral dose of vitamin D 3 (cholecalciferol) at 50,000, 100,000, or 200,000 IU one hour after SSR exposure. UVR was administrated as SSR emitted from a 1,000 W Xenon arc lamp (Newport, Stratford, CT), a full spectrum light source that closely resembles natural sunlight (Clydesdale et al., 2001).
This study was designed in a parallel fashion with exposure to SSR occurring on one arm without study drug administration (control phase), followed 2 weeks later by exposure to SSR on the contralateral arm with study drug administration (investigative phase) (Figure 1b). The MED to experimentally induce sunburn was  Figure 4. Serum vitamin D 3 and skin erythema after experimental sunburn in cluster 1 and cluster 2. (a) The serum 25(OH)D 3 levels over time after study drug administration for cluster 1 (n ¼ 11) and cluster 2 (n ¼ 7). (b) The serum 25(OH)D 3 levels over time after study drug administration for cluster 1 (n ¼ 7) and cluster 2 (n ¼ 7), excluding the participants from cluster 1 who received placebo (n ¼ 4). Error bars represent the standard error of the mean for each group at each time point. (c) The change in erythema over time after experimental sunburn with 2MED for each cluster. Data points represent erythema values for individual participants at each time point. Horizontal lines represent the mean for cluster 1 (n ¼ 11) and cluster 2 (n ¼ 7) at each time point. Statistical comparisons are between cluster 1 and cluster 2 at each time point. *P < 0.05; **P < 0.01. MED, minimal erythema dose. determined for each participant during the initial screening visit as previously described (Heckman et al., 2013). Participants were irradiated with one, two, and three times the MED on the sunshielded, upper arm using plastic holed templates to ensure that adjacent skin was not exposed. A total of 20 participants completed both phases of the study and were included in the per-protocol analysis.

Quantification of vitamin D 3 metabolites and calcium
The concentrations of the vitamin D 3 metabolites 25(OH)D 3 , 1,25(OH) 2 D 3 , and 24,25(OH) 2 D 3 , as well as total serum calcium, were measured from freshly frozen serum obtained during the screening visit, as well as 24 hours, 48 hours, 72 hours, and 1 week after receiving the study drug. Serum levels of 25(OH)D 3 (ng/ml) and 1,25(OH) 2 D 3 (pg/ml) were measured by Liaison assay, and serum levels of 24,25(OH) 2 D 3 (ng/ml) were measured by liquid chromatography-mass spectrometry (Heartland Assays, Ames, IA). Toxic serum 25(OH)D 3 levels were defined as those greater than 150 ng/ml (Holick, 2003). Total serum calcium was measured by the University Hospitals Cleveland Medical Center core laboratory (Cleveland, OH), and the normal reference range was considered 8.8e10.7 mg/dl.

Primary outcomes of randomized participants
Primary outcomes included noninvasive measurements of skin erythema and thickness 24 hours, 48 hours, 72 hours, and 1 week after irradiation, as well as tissue expression of TNF-a and iNOS 48 hours after irradiation. Skin erythema (redness) was quantified using a CR300 chromameter (Minolta, Ramsey, NJ). The difference in erythema (Da*) between irradiated and nonirradiated skin (a* irrad e a* nonirrad ) was calculated for each time point after SSR exposure (Da* time ). The difference in Da* time between the investigative and control phases of the study was calculated to determine the effect of the study drug on skin redness after SSR exposure [(Da* time ) invest e (Da* time ) control ]. In this context, exposure to erythemogenic doses of UVR results in sunburn and the release of proinflammatory cytokines and chemokines in the skin, including TNF-a and iNOS, which further propagate tissue inflammation. Increased skin redness and thickness are mediated by vasodilation, an influx of inflammatory cells, and vascular congestion within the skin. The gene expression profile of skin at this time is characterized by increased expression of various proinflammatory genes. (b) Levels of vitamin D 3 rapidly rise within the serum after high-dose oral vitamin D 3 intervention. Arginase-1 is upregulated within the skin, and production of the proinflammatory mediators TNF-a and iNOS is attenuated after sunburn. Reduced skin erythema and thickness are observed clinically. The gene expression profile of skin at this time is characterized by increased expression of skin barrier genes, which help to repair the epidermal barrier and attenuate the inflammatory insult. iNOS, inducible nitric oxide synthase; TNF-a, tumor necrosis factor-a.
Skin thickness, an acute measure of edema, was quantified using a Mitutoyo 9 mm dial caliper (Northamptonshire, UK). Thickness measurements were repeated in triplicate and the mean was used for all calculations. The difference in thickness (Dth) between irradiated and nonirradiated skin (th irrad e th nonirrad ) was calculated for each time point after SSR exposure (Dth time ). The difference in Dth time between the investigative and control phases of the study was calculated to determine the effect of the study drug on skin thickness after UVR exposure [(Dth time ) invest e (Dth time ) control ].
A 6-mm punch biopsy specimen was obtained from the 3MED site 48 hours after irradiation in both the control and investigative phases of the study. RNA was extracted from fresh frozen tissue using the RNeasy Lipid Mini Kit (Qiagen, Redwood City, CA), and tissue mRNA expression of TNF-a and iNOS was quantified as previously described (Au et al., 2015). A fold change representing the difference in RNA expression between the investigative and control phases of the study was calculated for TNF-a and iNOS [(RNA 48hr ) invest / (RNA 48hr ) control ].

Unsupervised hierarchical clustering of participants
An exploratory analysis was performed utilizing clusters obtained from the unsupervised hierarchical clustering of individual participants based on similarities in their gene expression profiles, regardless of the allocated treatment group. A minimum of 3 mg of total RNA was submitted to the Gene Expression and Genotyping facility at Case Western Reserve University for microarray analysis (Cleveland, OH). Human Genome 2.0 ST arrays (Affymetrix, Santa Clara, CA) were the chosen platform to analyze transcriptomic changes incurred by treatment. For microarray analyses, normalized linear data post-(T interv ) and pre-(T control ) study drug intervention were averaged amongst the 18 participants. The microarray raw data for two participants was not interpretable given poor sample quality. A fold change representing the difference in RNA log expression between the investigative and control phases of the study was determined for each gene for each participant [(RNA 48hr ) invest À (RNA 48hr ) control ]. Data analysis was performed using Rv3.2.2/Bioconductor (R Studio, Boston, MA). The oligo package was used to read, background correct, and normalize the raw sample data using the Robust Multiarray Average algorithm (Bolstad et al., 2003). The limma package was used to create a paired design matrix and conduct differential gene expression analysis based on a linear model fit and empirical Bayes methodology (Ritchie et al., 2015).
Heatmaps and dendrograms were generated utilizing 26,599 transcripts with unique gene names using the GENE-E software (http://www.broadinstitute.org/cancer/software/GENE-E/index.html, Accessed June 14, 2016). Hierarchical clustering was used to recursively merge samples based on pairwise distance, determined using the 1-Pearson correlation coefficient and average linkage methods. Ingenuity Pathway Analysis (Qiagen, www.qiagen.com/ ingenuity) was used to determine statistically significant canonical pathways, predicted upstream regulators, and biological networks most likely affected by the set of genes differentially expressed for each group of participants.
Tissue ARG1 mRNA expression was quantified using quantitative real-time reverse transcriptase-PCR as described above. To analyze the expression of the arginase-1 protein in skin, freshly frozen biopsy samples embedded in optimal cutting temperature compound were cut into 8-mm sections and stained immunofluorescently as previously described (Au et al., 2015). CD163 was used as a marker for skin macrophages, and DAPI was used to stain nuclear DNA.

Statistical analysis
Given the pilot nature of the study design, power calculations were not performed. To control for type I error rate, hierarchical closedtesting procedures were utilized for analysis of primary outcomes. Unpaired t-tests were used to compare the means of groups with respect to changes in the primary outcomes at each time point. Fisher's exact test was used for intergroup comparisons of categorical data and for determining significance of the canonical pathways. The Benjamini-Hochberg correction was used to determine differentially expressed genes in the microarray analysis, adjusting for multiple comparisons (Benjamini and Hochberg, 1995). A two-sided P-value of 0.05 or less was considered to indicate statistical significance.