Genomic signaling of vitamin D

.


Introduction
The secosteroid vitamin D 3 was named some 100 years ago a vitamin [1], because it cured experimentally induced rickets in dogs and rats [2]. However, humans and many other species synthesize vitamin D 3 endogenously, when in their skin the direct cholesterol precursor 7dehydrocholesterol is exposed to UV-B radiation (290-320 nm) ( Fig. 1, top) [3,4]. In fact, all species that are able to synthesize cholesterol can also make vitamin D 3 . For example, phytoplankton produces vitamin D 3 since some 750 million years [5] but uses it for UV-B scavenging [6] and not for endocrine purposes [7]. Endocrinology of vitamin D 3 started some 550 million years ago in lamprey (Petromyzon marinus), which is a boneless fish that already had a high affinity receptor for 1,25(OH) 2 D 3 [8]. This was confirmed by crystal structure analysis of the VDR ligand-binding domain (LBD) of lamprey [9] compared with that of humans [10]. Also other vertebrate species like bony fish, amphibians, reptiles, birds and mammals developed a functional VDR [11], which, with the exception of teleost fishes, originates from only one VDR gene [12]. The production of 1,25(OH) 2 D 3 from vitamin D 3 requires two hydroxylation steps and the transport of the molecules, i.e., functional vitamin D 3 endocrinology involves also metabolic enzymes [8], such as CYP (cytochrome P450) 2R1 and CYP27B1 (Fig. 1, top), and transporters like the vitamin D transport protein (encoded by the GC gene) [13]. In addition to these canonical vitamin D 3 metabolites the enzyme CYP11A1 was demonstrated to add hydroxyl groups to carbon 17, 20 and 22 of the vitamin D 3 molecule [14]. These metabolites were suggested to contribute to the prevention and attenuation of COVID-19 [15].
innate immune cells, such as monocytes and macrophages, in their fight against pathogenic microbes like the intracellular bacterium Mycobacterium tuberculosis [18]. Furthermore, vitamin D prevents overreactions of the adaptive immune system and avoids in this way the onset of autoimmune diseases, such as multiple sclerosis [19]. In addition, there are numerous indications for the involvement of vitamin D in other diseases, such as the prevention of certain types of cancer [20], but the majority of these effects are related to the immune-regulatory function of the compound. The molecular basis of the physiological functions of vitamin D is the regulation of target genes by 1,25(OH) 2 D 3 -activated VDR in respective tissues [21].
Since genomic signaling of vitamin D involves not only the activation of VDR but also the production of mRNA from target genes and their translation into protein, it takes several hours before physiological effects are detected [22]. In contrast, there are observations about rapid (seconds to minutes) actions in response to vitamin D stimulation, such as the transport of calcium i) by enterocytes [23], ii) in an osteogenic sarcoma cell line or iii) in primary chicken myocytes. These rapid effects at membranes led to the hypothesis that there may be a membrane receptor that mediates non-genomic actions of vitamin D. This claim is analogous to steroid hormones, such as progesterone or testosterone, for which membrane receptors are suggested [24,25].
The main aim of this review is to discuss the principles of genomic signaling of vitamin D with a focus on the molecular actions of VDR. This may serve as a reference for understanding also rapid and non-genomic effects of vitamin D.

VDR as a member of the nuclear receptor superfamily
The close structural similarity of the members of the nuclear receptor superfamily suggests that they have the same evolutionary origin, which most likely was an orphan receptor involved in the control of energy metabolism [26,27]. Multiple rounds of gene duplications created the family of 48 nuclear receptors encoded by the human genome [28]. In a trial and error process some members of the superfamily learned over millions of years to bind and get activated by a hydrophobic ligand in the size range of cholesterol. The closest relatives of VDR (NR1I1) within the family are the so-called adopted orphan nuclear receptors PXR (pregnane X receptor, NR1I2), CAR (constitutive androstane receptor, NR1I3), LXR (liver X receptor) β(NR1H2), LXRα (NR1H3) and FXR (farnesoid X receptor, NR1H4) [29]. PXR and CAR control xenobiotic detoxification pathways, while LXR and FXRs regulate lipid metabolism. Interestingly, all six nuclear receptors can get activated by micromolar concentrations of cholesterol derivatives like bile acids and oxysterols [30][31][32][33].
VDR is the only member of the NR1H/NR1I subgroup, which evolved into an endocrine receptor that accommodates its specific ligand, 1,25 (OH) 2 D 3 , in sub-nanomolar concentrations [8]. This made the physiological profile of VDR more similar to the 11 other endocrine members of the superfamily [34]. These are the nuclear receptors for retinoic acid (RARα, RARβ and RARγ) and thyroid hormone (THRα and THRβ) as well as those for the steroids estrogen (ESR1 and ESR2), testosterone (AR (androgen receptor)), progesterone (PGR), aldosterone (MR (mineralocorticoid receptor)) and cortisol (GR (glucocorticoid receptor)) [35,36].
It is a general principle in biochemistry that receptors have an affinity for their ligands in the order of their physiological concentration levels [37]. Since the cellular concentration of 1,25(OH) 2 D 3 is in the order of 0.1 nM, a functional endocrine receptor like VDR needs to have an K D -value in this range. In contrast, an adopted orphan nuclear receptor like LXR functions well as a sensor for up to millimolar levels of oxysterols with a K D -value in the respective order. The same principle applies also for metabolizing enzymes, such as CYP2R1, CYP27B1 and CYP24A1 in case of vitamin D endocrinology. However, the circulating concentrations of vitamin D 3 and 25(OH)D 3 are in the order of 10-100 nM. Therefore, the enzymes do not have to fully embed the vitamin D compounds and can allow for this substrate binding site to be in partial contact to the aqueous environment [38].
Many nuclear receptors have an interesting triangular relationship, because the genes, which encode for specific metabolic enzymes and transporters controlling the concentration of the ligands, are targets of the respective receptors [39,40]. For example, the down-regulated vitamin D target gene CYP27B1 encodes for an 1α-hydroxylase that converts 25-hydroxyvitamin D 3 (25(OH)D 3 ) into 1,25(OH) 2 D 3 [41]. In contrast, the up-regulated vitamin D target gene CYP24A1 encodes for a 24-hydroxylase that inactivates 1,25(OH) 2 D 3 [42]. These and many comparable examples of other nuclear receptors suggest that there was a coevolution of metabolic enzymes, transporters and the nuclear receptors regulating them. Thus, hundreds of million years of evolution resulted in a fine-tuned system of nuclear receptors, ligands and enzymes that form the functional basis for genomic signaling not only of vitamin D but also for other endocrine hormones and nutritional regulators.

Molecular action of the VDR
The members of the nuclear receptor superfamily have a modular structure, of which the amino-terminal DNA-binding domain (DBD) and the carboxy-terminal LBD are most important. The DBD consists of 60-70 highly conserved amino acids that form two zinc fingers, each of which specifically recognize three base pairs of a hexameric DNA sequence motif. In contrast, the conservation of the LBD is rather on the basis of its 3D structure than on its amino acid sequence (Fig. 1, bottom  left). The LBD of all nuclear receptors has a 3-layer sandwich structure formed by 11-13 α-helices [43]. The ligand binding pocket (LBP) is a cavity of a volume of 300-700 Å 3 for endocrine nuclear receptors and up to 1400 Å 3 for adopted orphan receptors, which locates in the lower part of the LBD [44]. For example, for VDR the LBP is formed by 40 primarily non-polar amino acids being well-adapted to the shape of the 1,25 (OH) 2 D 3 molecule [10,45].
The volume of a potential nuclear receptor ligand is an important criterium for the ability to bind to the LBP. Ideally, the volume and shape of both the ligand and the LBP should be similar, so that the latter embeds the compound. Endocrine ligands like 1,25(OH) 2 D 3 and estrogen fit very well into the LBP of their respective receptors. Therefore, they are bound with high specificity. In contrast, the larger size of the LBP of adopted orphan receptors allows to accommodate a wider range of compounds, each of which is bound with rather low affinity. A second aspect are interactions between polar groups of the ligands, which are primarily hydroxy groups, and specific polar amino acids within the LBP. 1,25(OH) 2 D 3 has, in contrast to steroid hormones, even three hydroxy groups at carbons 1, 3 and 25, which form hydrogen bonds with Y143 (helix H1) and S278 (helix H5), S237 (helix H3) and R274 (helix H5) as well as H305 (loop between helices H6 and H7) and H397 (helix H11), respectively (Fig. 1, bottom right) [10,46]. Thus, the three polar contacts are the main reason, why VDR has with a K D -value of 0.1 nM an outstandingly high affinity for its natural ligand [47]. In addition, the 40 hydrophobic amino acid residues that form the ligand binding pocket snugly embed 1,25(OH) 2 D 3 and its synthetic analogues [45]. These hydrophobic interactions further contribute to the high affinity of VDR for its ligands and in parallel explain differences in the receptor binding affinity of each compound [48].
The binding of ligand to the LBP of a nuclear receptor induces a conformational shift within the LBD that also affects the structure of its surface. Importantly, more than 100 VDR crystal structures (of different ligands and species) confirm that there is only one agonistic conformation of the protein, i.e., the structure of the agonistic ligand has no specific effect on the receptor [49]. This conformational change results in an exchange of proteins that interact with the nuclear receptor, such as a shift from corepressors to coactivators [50,51]. For example, the binding of 1,25(OH) 2 D 3 or synthetic agonists to VDR causes an efficient dissociation of corepressors like NCOR1 (nuclear receptor corepressor 1) and allows the binding of coactivators like NCOA1 (nuclear receptor coactivator 1) and members of the Mediator complex. The latter complex forms a protein bridge to the basal transcriptional machinery, which marks together with RNA polymerase II (Pol II) transcription start sites (TSSs) of vitamin target genes [52]. Moreover, coactivator proteins interact with chromatin modifying enzymes, which leads to local changes of histone modification at enhancer and promoter regions [53] ( Fig. 2, right). Furthermore, proteins of chromatin remodeling complexes interact in a ligand-dependent way with VDR and other nuclear receptors [54]. Thus, the ligand-triggered activation of a nuclear receptor like VDR results in significant changes in the protein interaction profile of the transcription factor, which has significant consequences on gene activation and repression.

Specific activation of vitamin D target genes
The human genome encodes for approximately 1600 different transcription factors, which have the unique property to bind specifically a genomic DNA sequence. This is in contrast to histone proteins that bind as parts of nucleosomes in a regular fashion every 200 base pairs to DNA but do not show any sequence specificity [55]. This protein-DNA complex is referred to as chromatin and is the physical expression of the epigenome [56,57]. Transcription factors and nucleosomes compete for genomic DNA and in most cases nucleosomes are the "winners". Therefore, in a terminally differentiated cell only some 10% of the genome is located in so-called euchromatin, which is the active form of chromatin, where nucleosomes leave sufficient space for transcription factor binding [58]. In this way, activated transcription factors bind only to those promoter and enhancer regions that are located within accessible chromatin [52] and in addition comprise a specific binding site for the transcription factor complex.
Most transcription factors act as dimeric complexes, in order to cooperatively increase their DNA binding affinity. The best characterized partner protein of VDR is the nuclear receptor RXR (retinoid X receptor) (Fig. 2, right) [59]. Each nuclear receptor recognizes with the two zinc fingers of its DBD a hexameric sequence motif. Accordingly, the VDR-RXR heterodimer binds two hexamer motifs, which a preferentially in a direct repeat arrangement with three intervening base pairs, which are so-called DR3-type response elements [59].
The genome-wide binding pattern of a transcription factor is referred to as its cistrome and is for most transcription factors very tissue specific. The VDR cistrome has been determined in more than 10 human tissues and cell types [60][61][62]. In the absence of ligand, the cistrome comprises only some 200-2,000 sites, while after ligand stimulation the number increases in average 2.5-fold [60]. Taking all investigated tissues together, there more than 20,000 genomic VDR binding sites known in the human genome. Interestingly, only a few hundred of these sites are persistently bound by VDR, while the vast majority of loci are found transiently after ligand activation [62]. It is assumed that these persistent VDR sites represent the prime contacts of vitamin D with the genome, i.e., they are kind of hotspots of the genomic signaling of vitamin D, from where secondary responses to the nuclear hormone are coordinated.
Promoters and enhancers are stretches of accessible genomic DNA that comprise binding sites for one or several transcription factors. They differ only in their location in relation to the TSS of the respective target gene: promoters are rather close (+/-1000 base pairs of the TSS), while the distance of enhancers in only limited by the size of the respective topologically associated domain (TAD), in which the gene is located (Fig. 2, left). TADs have an average size of about 1 million base pairs and subdivide the genome into a few thousand functional domains [63], each of which contain 1-10 coregulated genes [56]. When activated VDR binds to an accessible enhancer region, it can regulate all genes within the same TAD that have accessible TSS regions. This explains why some vitamin D targets, such HLA (human leukocyte antigen) genes and those encoding for chemokines of the CXCL (C-X-C motif chemokine ligand) family are each clustered in the same genomic region [22].
The ligand-stimulated binding of VDR to thousands of enhancer regions is the first epigenome-wide effect of vitamin D stimulation [60]. The VDR binding is followed by changes of the histone modification pattern at these (and neighboring) enhancers (H3K27ac) as well as at TSS regions (H3K4me3) [64]. This results in a vitamin D-dependent change of chromatin accessibility [65]. In addition, in a liganddependent fashion pioneer transcription factors, such as PU.1 (purinerich box 1) [66] and CEBPα (CCAAT enhancer binding protein α) [67], associate with VDR bound enhancers and help the receptor in increasing the local accessibility of its binding sites (Fig. 2, right). Regulatory DNA loops form between these vitamin D-activated enhancer regions and TSS regions of vitamin D target genes within the same TAD. For up-regulated target genes this results in increased activity of Pol II and more mRNA production. In contrast, for down-regulated genes there is a multitude of specific mechanisms, the core event of which is the inhibition of an activating transcription factor or chromatin modifier.
The above described mechanism of genomic signaling of vitamin D applies to all primary target genes and takes place in all VDR expressing tissues and cell types. Since most human tissues express the VDR gene [1,16], the physiological impact of vitamin D is clearly broader than assumed decades ago [68]. Furthermore, in addition to primary vitamin D target genes, there is a large number of secondary targets, which do not require the direct binding of VDR to their regulatory region, but are under the control of transcription factors, such as BCL6 (BCL6 transcription repressor), NFE2 (nuclear factor, erythroid 2), POU4F2 (POU class 4 homeobox 2) and ELF4 (E74 like ETS transcription factor 4), which are encoded by primary vitamin D target genes [69]. Thus, genomic signaling of vitamin D results in the regulation of several hundred target genes per VDR expression tissue.

Non-genomic vitamin D signaling
Compared to the genomic signaling of vitamin D and other steroidal nuclear hormones, there is no superfamily of related membrane receptors explaining the molecular mechanisms of rapid, non-genomic effects of estrogen [70,71], testosterone [72], aldosterone [73] or cortisol [74]. Accordingly, there was no comparable evolutionary driven development of a membrane-based, non-genomic signaling pathways. Furthermore, there is no cocrystal of any of the ligands bound to the suggested receptors.
At present, the enzyme PDIA3 (protein disulfide isomerase family A member 3, also known as ERp57 or 1,25D 3 -MARRS) is the best studied candidate for an alternative binding protein of 1,25(OH) 2 D 3 . However, a direct contact of the ligand with the PDIA3 protein could not be demonstrated. PDIA3 is involved in intestinal calcium absorption and skeletal development [75][76][77][78], i.e., it contributes to core functions of genomic vitamin D signaling. At membranes PDIA3 associates with caveolin 1 [79,80], which is the main protein of caveolae (small invaginations of the plasma membrane) [81]. Furthermore, it was suggested that PDIA3 may serve as a molecular chaperone for VDR [82], i.e., rapid effects of vitamin D may be explainable through the location of VDR at cell membranes, including caveolae [83,84]. For example, rapid effects of 1,25(OH) 2 D 3 had been reported at a concentration of 0.3 nM [85]. Since only VDR posses that high affinity for 1,25(OH) 2 D 3 , it is more than likely that the observed effects are based on a mechanism that uses VDR as a sensor for the molecule.
Genomic and non-genomic effects of vitamin D are primarily studied in experimental settings that use stimulations with supra-physiological concentrations of 1,25(OH) 2 D 3 in the range of 10-100 nM [86], because under these extreme conditions the system is most responsive. However, under in vivo conditions vitamin D endocrinology aims on homeostasis [87], i.e., there is no drastic rise in 1,25(OH) 2 D 3 concentrations. However, there may be exceptional situations in 1,25(OH) 2 D 3

Fig. 2. Vitamin D signaling in the context of TADs and chromatin.
TADs are architectural loops that subdivide the human genome into a few thousand subunits (left). The interaction of VDR-bound enhancers with the TSS region of gene 2 creates a regulatory loop (right). VDR (red) binding in complex with RXR (light green) is supported by the pioneer transcription factors PU.1 (dark blue) and CEBPα (green). This leads to changes in histone marks at promoters (H3K4me3 and H3K27ac) and enhancers (H3K27ac), in order to activate Pol II. This finally leads to mRNA transcription of respective vitamin D target genes. ac, acetylated; me, methylated.
producing cells, such proximal tubules of the kidneys and CYP27B1 expressing immune cells, in which the concentrations of the hormone are significantly higher than measured in serum. Under these conditions there may be the chance that rapid, non-genomic signaling of vitamin D are effective that do involve alternative receptors than VDR.

Conclusions
Genomic signaling of vitamin D is in full accordance with the mechanism of action of other nuclear hormones, such as estrogen and cortisol. Several hundred million years of evolution of vertebrate species resulted with VDR in a protein that binds with extremely high affinity (K D = 0.1 nM) the biologically most active vitamin D metabolite 1,25 (OH) 2 D 3 . In parallel, the endocrine vitamin D system evolved, in order to regulate not only calcium homeostasis, but also innate and adaptive immunity as well as energy metabolism. The molecular mechanisms of ligand activated VDR, which binds to enhancer regions, changes its protein interaction partners and finally activates Pol II to enhance target gene transcription, are well understood based on crystal structure, epigenome and transcriptome data.
Under some experimental conditions rapid, non-genomic effects of vitamin D exist and may be mechanistically understood via the action of PDIA3 and/or membrane associated VDR. It is unclear, whether these rapid response occur in the real life in vivo condition, where vitamin D regulates homeostasis. However, in certain physiological conditions, such as calcium transport, non-genomic effects of vitamin D seem to contribute to the genomic actions and fine-tune them.

Financial support and sponsorship
This publication is part of the WELCOME2 project that has received funding from the European Union's Horizon2020 research and innovation program under grant agreement no. 952,601 and from the David and Amy Fulton Foundation, Seattle, US.

Declaration of Competing Interest
The author declares that he has no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.