I. METABOLIC ACTIVATION OF VITAMIN D

A. Introduction

An appreciation that vitamin D₃ represents only a precursor to its functionally active form, 1α,25-dihydroxyvitamin D₃ [1,25-(OH)₂D₃], is arguably one of the most important developments in vitamin research during the latter half of the 20th century. The discovery of the two activation steps involved in the metabolism of vitamin D₃ to the hormone 1,25-(OH)₂D₃ (sect. i) set the stage for the elucidation of the role of vitamin D in the physiological events involved in calcium and phosphate homeostasis (sect. ii). The realization that it was the metabolites of vitamin D that were important led to an intense focus on the molecular events surrounding the mechanism of action of 1,25-(OH)₂D₃ , which resulted in the discovery of the vitamin D receptor (VDR) and its interaction with the transcriptional machinery inside vitamin D target cells (sect. iii). Subsequently, this led to the demonstration of new biological actions of 1,25-(OH)₂D₃ , in particular, its effects on the regulation of growth and differentiation of certain specialized cell types (sect. iv), which represent involvements of vitamin D not even envisioned when 1,25-(OH)₂D₃ was first discovered. Furthermore, the knowledge of vitamin D metabolism also provided the impetus to study the regulation of cytochrome P-450-containing enzymes involved in the process as well as stimulating the chemical synthesis of a wide range of vitamin D analogs (sect. v). This review seeks to summarize our current understanding of the molecular events surrounding the physiological action of vitamin D in these many varied areas. For a more detailed account of the subjects described in this review, particularly those of a clinical nature, the reader is directed to a recently published text (96).

B. Overview of Metabolism

Vitamin D, in the form of vitamin D₃ , is made from 7-dehydrocholesterol in the skin by exposure to ultraviolet light (270–300 nm range). Alternatively, vitamin D, in the form of either vitamin D₂ or vitamin D₃ , can be derived from dietary sources (Fig. 1 A). Both vitamin D₃ and vitamin D₂ undergo the same activation process, involving first 25-hydroxylation in the liver, followed by 1α-hydroxylation in the kidney, to make the biologically active compounds 1,25-(OH)₂D₃ and 1,25-(OH)₂D₂ , respectively (Fig. 1 B). There is little evidence that these two active forms differ in their mode of action, and because most is known about the synthesis and action of 1,25-(OH)₂D₃ , this review focuses on the natural D₃ compound. The metabolic activations of vitamin D₃ are carried out by specific cytochrome P-450-containing enzymes, the vitamin D₃-25-hydroxylase (CYP27) and possibly another P-450 in the hepatocyte and the 25-hydroxyvitamin D-1α-hydroxylase (CYP1α) in the renal proximal tubular cell. Both of the known hydroxylases are located in the inner mitochondrial membrane of these cells (Fig. 2). The synthesis of 25-hydroxyvitamin D₃ (25-OH-D₃) by the liver appears to be only loosely regulated, whereas the synthesis of 1,25-(OH)₂D₃ by the renal 1α-hydroxylase is tightly regulated by the levels of plasma 1,25-(OH)₂D₃ and calcium. The renal enzyme is strongly upregulated by the hormone parathyroid hormone (PTH), a point that is discussed further in section ii. A third vitamin D-related mitochondrial cytochrome P-450-containing enzyme, the 25-hydroxyvitamin D-24-hydroxylase (CYP24), was originally believed to be exclusively located in the kidney and to be involved only in the metabolism of 25-OH-D₃ to 24,25-dihydroxyvitamin D₃ [24,25-(OH)₂D₃]. The 24-hydroxylation of 1,25-(OH)₂D₃ was first realized with the isolation of 1,24,25-(OH)₃D₃ and subsequently shown to occur in all vitamin D target tissues including enterocytes, osteoblasts, keratinocytes, and parathyroid cells. Thus it is now known that CYP24 will use 1,25-(OH)₂D₃ as substrate also. Because CYP24 is widely distributed around the body, is strongly induced in target cells by 1,25-(OH)₂D₃ , and prefers 1,25-(OH)₂D₃ as substrate to 25-OH-D₃ , its role appears to be catabolic. The enzyme CYP24 catalyzes several steps of 1,25-(OH)₂D₃ degradation, collectively known as the C-24 oxidation pathway which starts with 24-hydroxylation and culminates in the formation of the biliary excretory form, calcitroic acid (Fig. 3). Thus our current view is that both the synthesis and degradation of 1,25-(OH)₂D₃ are tightly regulated events, attesting to the fact that the concentration of this potent hormone requires fine control at the cellular level, and hence a set of highly specific and finely tuned cytochrome P-450 exist for the purpose.

C. Hepatic 25-Hydroxylation

Vitamin D does not circulate for long in the bloodstream but, instead, is immediately taken up by adipose tissue for storage or liver for further metabolism. In humans, tissue storage of vitamin D can last for months or even years. Ultimately, vitamin D₃ undergoes its first step of activation, namely, 25-hydroxylation, in the liver (28) (Fig. 1 B). Early data suggested that the liver is the only significant site of 25-hydroxylation in vivo, although there were occasional reports of intestinal and renal extracts containing this activity (346). Research, therefore, focused on purification of the major hepatic enzyme activity. Over the years, there has been some controversy over whether 25-hydroxylation is carried out by one enzyme or two and whether this cytochrome P-450-based enzyme is found in the mitochondrial or microsomal fractions of liver. Madhok and DeLuca (207) reported that a rat liver microsomal system requiring NADPH, molecular oxygen, a flavoprotein, and a cytochrome P-450 was capable of 25-hydroxylation of vitamin D₃ , but the cytochrome P-450 responsible has never been cloned. There was some speculation that the microsomal enzyme might be CYP2C11, but this cytochrome is male specific (129) and other data have also been presented that indicate that human microsomes do not possess 25-hydroxylase activity (285). Recently, Axen et al. (9) have purified a pig liver microsomal 25-hydroxylase with an NH₂-terminal sequence different from that of CYP2C11 and that is capable of the 25-hydroxylation of both vitamins D₂ and D₃ . Currently, only the mitochondrial 25-hydroxylase has been purified to homogeneity and subsequently cloned (7 44 351). The cytochrome P-450 involved is known as CYP27 or P-450c27 because it is a bifunctional cytochrome P-450 which in addition to 25-hydroxylating vitamin D₃ also carries out side-chain hydroxylation, including 27-hydroxylation of cholesterol-derived intermediates involved in bile acid biosynthesis (from which it derives its name) (252).¹ The primary amino acid sequences of three species of CYP27 are depicted in Figure 4. Even though 25-hydroxylation of a variety of vitamin D compounds, including vitamin D₃ , has been clearly demonstrated in cells transfected with CYP27 (119), there is still some skepticism in the vitamin D field that a single cytochrome P-450 can explain all the metabolic findings observed over the past two decades of research. The many unexplained observations suggesting that other cytochrome P-450 might perform 25-hydroxylation of vitamin D at nanomolar concentrations of substrate that exist in vivo include the following. 1) Perfused rat liver studies by Fukushima et al. (104) demonstrate kinetics consistent with two 25-hydroxylase enzyme activities: a high-affinity, low-capacity form (presumably microsomal) and a low-affinity, high-capacity form (presumably mitochondrial; CYP27).

2) Dietary studies show regulation, albeit weak, of the liver 25-hydroxylase in animals given normal intakes of vitamin D after a period of vitamin D deficiency (20), which is not explained by a transcriptional mechanism, since the gene promoter of CYP27 lacks a vitamin D-responsive element (VDRE) (118) or demonstrable responsiveness to 1,25-(OH)₂D₃ whereas it is regulated by bile acids (354).

3) Clinical studies show no obvious 25-OH-D₃ or 1,25-(OH)₂D₃ deficiency occurs in patients suffering from the genetically inherited disease cerebrotendinous xanthomatosis, in which CYP27 is mutated. [Although a subset of these patients suffer from osteoporosis, this is more likely because of biliary defects leading to altered enterohepatic circulation of 25-OH-D₃ (18).]

4) Substrate specificity studies using transfected recombinant human CYP27 show that the enzyme does not 25-hydroxylate vitamin D₂ (119); it 24-hydroxylates vitamin D₂ instead, evoking the question: Which cytochrome P-450 synthesizes 25-OH-D₂?

Observations that are explained by the existence of CYP27 include 1) the occasional reports of extrahepatic 25-hydroxylation of vitamin D₃ mentioned above (346), which are consistent with the detection of CYP27 mRNA in a number of extrahepatic tissues including kidney and bone (osteoblast) (10 147), and 2) the abundance of 24-hydroxylated metabolites [e.g., 24-OH-D₂ , 1,24-(OH)₂D₂ , and 24,26-(OH)₂D₂] in the blood of vitamin D₂-intoxicated animals (143 161 178 320). It is worth noting from perusal of Figure 4 that the recently cloned CYP1α is more closely related to CYP27 than it is to CYP24, a surprising fact given that CYP1α and CYP24 are both renal cytochromes P-450 and appear to be reciprocally regulated, thereby implying CYP27 might have evolved to metabolize vitamin D after all. Thus, although CYP27 remains the best-characterized cytochrome P-450 capable of 25-hydroxylation, it may not be the only 25-hydroxylase, and its full physiological importance remains to be established.

The product of the 25-hydroxylation step, 25-OH-D₃ , is the major circulating form of vitamin D₃ and in humans is present in plasma at concentrations in the range 10–40 ng/ml (25–125 nM) (140). The main reason for the stability of this metabolite is its strong affinity for the vitamin D-binding (globulin) protein of blood (DBP) (70). The metabolic fate of 25-OH-D₃ is dependent on the calcium requirements of the animal. An urgent need for calcium results in renal 1α-hydroxylation, whereas an abundance of calcium results in 24-hydroxylation (see sect. ii). These two alternative pathways are discussed in turn below.

D. Renal 1α-Hydroxylation

The enzyme 25-hydroxyvitamin D₃-1α-hydroxylase is responsible for the tightly regulated step that involves the introduction of a 1α-hydroxyl group into the A ring of 25-OH-D₃ , thereby creating the hormone 1,25-(OH)₂D₃ . The specific location of this enzyme in the kidney became apparent (100) even before the unequivocal identification of 1,25-(OH)₂D₃ (139). Experiments involving nephrectomized animals have confirmed that the kidney as the major source of the circulating pool of 1,25-(OH)₂D₃ . The renal 1α-hydroxylase enzyme comprises a cytochrome P-450, a ferredoxin, and a ferredoxin reductase (112) (see Fig. 2). The cytochrome P-450 for the 1α-hydroxylase enzyme, CYP1α, was recently cloned from rat, mouse, and human (228a 298 316 332), and the amino acid sequences of these are compared with the other known vitamin D-related cytochromes P-450 in Figure 4. Because of the close resemblance with CYP27, it has been suggested that CYP1α (P-4501α) be termed CYP27B1 (298). Both 1α-hydroxylases have short mitochondrial targeting sequences but share many regions of similarity with other members of the family including the classical heme-, ferredoxin-, and oxygen-binding sites indicated in Figure 4. The 1α-hydroxylase is induced by PTH through a cAMP/phosphatidylinositol 4,5-bisphosphate (PIP₂)-mediated signal transduction mechanism that is still to be defined at the molecular level (132). The enzyme appears to be downregulated by vitamin D status, possibly through a VDR-mediated transcriptional mechanism involving the hormonal product 1,25-(OH)₂D₃ , although there were early claims that 1,25-(OH)₂D₃ might act directly on its own synthesis through an allosteric mechanism (117). Work using the perfused vitamin D-deficient rat kidney (281) elegantly shows that the downregulation of 1α-hydroxylation takes 2–4 h after exposure to 1,25-(OH)₂D₃ and is blocked by inhibitors of protein synthesis and transcription. In the same model, the disappearance of the 1α-hydroxylase is mirrored by the reciprocal appearance of the renal 25-OH-D₃-24-hydroxylase, in effect a complete “switchover” from 1α- to 24-hydroxylating activity in the isolated organ over the 4-h period. The exact reciprocal regulation of the two enzymes, first demonstrated in vivo by Tanaka et al. (336) two decades ago, led some workers to postulate that the 1α- and 24-hydroxylases might share a single cytochrome P-450 polypeptide chain, its catalytic properties modified by NH₂-terminal truncation (111) or regulated by the phosphorylation state of the ferredoxin component of the enzyme (300). The cloning of two distinct cytochromes P-450, representing 1α-hydroxylation and 24-hydroxylation, suggests the first hypothesis to be incorrect, the switchover process probably being accomplished by de novo protein synthesis of the required cytochrome P-450. The second hypothesis involving regulation of enzyme activity through ferredoxin phosphorylation remains a possibility for fine-tuning of enzyme activity, although the details of this remain obscure at this point. The cloning of CYP1α has already led to mapping of the human gene to chromosome 12q13.1-q13.3 (316), the same locus as the gene defect for vitamin D dependency rickets type I (192), a disease cured by small doses of exogenous 1,25-(OH)₂D₃ and which had been previously postulated to be because of a mutated version of the 1α-hydroxylase (99). The next decade will see a clarification of the molecular and clinical aspects surrounding this key regulatory step of calcium homeostasis.

There have been indications that there are extrarenal versions of the 1α-hydroxylase existing in cells of monocyte/macrophage, placental, and keratinocyte lineages (4 22 83). There is strong evidence that the extrarenal enzyme located in macrophages plays a major role in certain granulomatous conditions (e.g., sarcoidosis), causing uncontrolled elevations of blood 1,25-(OH)₂D₃ levels, which subsequently result in troublesome hypercalcemia and hypercalciuria (3). The normal function of this and other extrarenal 1α-hydroxylases remains obscure at this time, although some have postulated a paracrine or autocrine role for locally produced 1,25-(OH)₂D₃ . Once again, it is safe to predict that the molecular basis for various reports of extrarenal activity in cultured cells in vitro (4 22 83) and in certain human disease states will be resolved shortly with the cloning of the renal enzyme (298 316 332), as will its mechanisms of regulation.

E. 24-Hydroxylation

The discovery of 24,25-(OH)₂D₃ (138 327) predated even the identification of 1,25-(OH)₂D₃ and the recognition of 24-hydroxylation as a metabolic step allowed for the search for the enzyme activity. The relative ease with which 24,25-(OH)₂D₃ was generated in such large amounts was a clue that the metabolic step was upregulated rather than downregulated by vitamin D administration. The earliest report of the 25-OH-D₃-24-hydroxylase was a subcellular localization study (174) using vitamin D-replete chicken kidney tissue which established that the 24-hydroxylase is a mitochondrial cytochrome P-450-containing enzyme, and this was followed by a reconstitution experiment involving partially purified enzyme components (248). Evidence was also emerging both in vivo and in vitro that the 24-hydroxylase was not confined to the kidney but could be found in classical vitamin D target tissues including the intestine and bone (185 347). In the late 1970s, it became evident that 24-hydroxylation was probably only the first step in an inactivation process. First, it was shown to be induced by 1,25-(OH)₂D₃ itself (172 187 335 337 340), and the product 1,24,25-(OH)₃D₃ was 10 times less biologically active than 1,25-(OH)₂D₃ (56 335). Second, 24,25-(OH)₂D₃ and its 1α-hydroxylated analog, 1,24,25-(OH)₃D₃ , could be converted to further metabolic products containing a 24-oxo and/or 23-hydroxy groups as well (237 246 330). The perfused rat kidney was helpful in establishing the production of these catabolites (160) but revealed two further pieces to the puzzle of the metabolic role of 24-hydroxylation. First, the perfused rat kidney allowed for a clarification of the temporal relationship of these catabolites, in effect, suggesting the existence of a pathway from 1,25-(OH)₂D₃ and/or 25-OH-D₃ (160), and second, the perfused kidney generated two side chain-cleaved molecules: a 23-alcohol (159) and a 23-acid (208 276), not observed previously in vitro. The 1α-hydroxylated 23-acid, calcitroic acid, is observed in vivo and has been shown to be the principal biliary excretory form of 1,25-(OH)₂D₃ (94). These discoveries led to the rationalization of many findings from a number of laboratories into a hypothesis that 24-hydroxylation is the first step of a target cell C-24 oxidation pathway (Fig. 3) whose major function is to convert 1,25-(OH)₂D₃ to calcitroic acid (208 276). The demonstration of vitamin D-inducible, calcitroic acid production in bone (UMR106) and kidney (208) was followed by demonstration of vitamin D-inducible C-24 oxidation pathway activity in a number of vitamin D target cells including intestine (Caco-2), keratinocyte (HPKIA-ras), and breast (T47D and MCF-7) (215 278 344).

In the early 1990s, Okuda's group succeeded in cloning the cytochrome P-450, CYP24 or P-450cc24, representing the 24-hydroxylase (62 148 247). The amino acid sequences of three species of CYP24 are shown in Figure 4. It belongs to the same subfamily as the other two known vitamin D-related cytochromes. The structure of the gene for CYP24 has been described for several species (rat, mouse, and human), and in each case shown to possess two VDRE in its proximal promoter (249 367 368). These VDRE allow for the 1,25-(OH)₂D₃ VDR-mediated upregulation of CYP24 from undetectable to strongly detectable expression at the mRNA level within 4 h (297). This is consistent with the enzyme activity pattern first observed in the kidney but subsequently reported for a variety of vitamin D target cells after initial exposure to hormone (127 202 209). Metabolic studies with the recombinant CYP24 protein produced in bacteria or insect cells have been equally enlightening. Akiyoshi-Shibata et al. (5) and Beckman et al. (15) have shown that CYP24 is a multicatalytic enzyme capable of several if not all of the steps illustrated in Figure 3. It is likely that CYP24 is able to catalyze three successive oxidations, two at C-24 and one at C-23, to give an intermediate that is subsequently cleaved by an unknown mechanism. All prevailing evidence suggests that C-24 oxidation is a highly efficient process giving rise to molecules of lower biological activity (e.g., calcitroic acid) such that if C-24 hydroxylation is blocked by a general cytochrome P-450 inhibitor such as ketoconazole, 1,25-(OH)₂D₃ hormone action is extended (265). This argues that the main role of the C-24 oxidation pathway is attenuation of the biological signal inside target cells.

Recently, this hypothesis was tested when St. Arnaud et al. (315) engineered a CYP24-null mouse. At this time, the results have been reported only in abstract form, but it is clear that the defect is not lethal during embryonic development. Instead, at least one-half the mice exhibit hypercalcemia/hypercalciuria in early neonatal life and quickly die before weaning from nephrocalcinosis. The other half of animals survive and appear healthy perhaps due to the upregulation of some alternative vitamin D-catabolic pathway (C-26 hydroxylation or 26,23-lactone formation). All CYP24-null animals exhibit abnormal bone histology, characterized by excessive unmineralized bone matrix and reminiscent of that previously observed in experiments involving exogenous 1,25-(OH)₂D₃ intoxication (136). This could be caused by an inability of target cells, in this case osteoblasts, to turn off the 1,25-(OH)₂D₃ signal in the absence of the C-24 oxidation pathway. An alternative explanation that CYP24 is needed for the synthesis of some essential 24-hydroxylated metabolite of vitamin D [e.g., 24,25-(OH)₂D₃] needed for a yet to be defined role in bone formation has been proposed (314) but seems unlikely for several reasons. Among the data that argue against such an essential role for 24-hydroxylated metabolites are the findings from the study of fluorinated analogs of vitamin D (149 234 338). These analogs are irreversibly blocked in the C-24 and C-23 positions of the side chain with one or more atoms of fluorine and therefore unable to undergo 24- or 23-hydroxylation. However, they are fully biologically active in all known in vivo functions of vitamin D. Furthermore, the biochemical machinery required for transducing the signal from a 24-hydroxylated metabolite has never been satisfactorily demonstrated in bone (or any other) cells. Thus it appears that 24-hydroxylation is not essential for vitamin D to fulfill its many biological roles in vertebrate biology. Therefore, evidence from the CYP24-null mouse seems to confirm our hypothesis that the C-24 oxidation pathway is a complex, self-induced mechanism for limiting the action of 1,25-(OH)₂D₃ in vitamin D target cells once the initial wave of gene expression has been initiated (see sect. iii). This is discussed in greater detail in section v.

II. ROLE OF VITAMIN D IN CALCIUM HOMEOSTASIS

A. Development of Calcium Homeostatic Mechanisms

Calcium is undoubtedly one of the most tightly regulated substances in plasma of higher animals (81 270). Its concentration is held constantly at 1 mmol ionized calcium or 10 mg/100 ml of total calcium. The ionized calcium concentration of plasma is very close to that found in seawater (270), and it is believed that the evolution of the calcium homeostatic system took place as animals emerged from the sea into fresh water and further onto land. Very likely the dependence of a number of life's essential functions on calcium occurred because of the constancy and abundance of calcium in seawater. Among them are the neural transmission, muscle contraction (and relaxation), exocrine secretion, blood clotting, and the adhesion of cells to each other. The presence of calcium in abundance in seawater also made understandable the use of calcium in the construction of structural elements such as the skeleton. As animals emerged into fresh water, resulting in a drop in ambient calcium concentration, it immediately brought into need the ability to mobilize calcium to meet the needs of the very critical functions such as neural transmission and muscle contraction. This only intensified as animals emerged from fresh water onto land where calcium availability was even more limited than in fresh water. Furthermore, the gravitational forces applied to the terrestrial animals must have increased the need for a structurally sound skeleton. Thus the evolved mammal had many problems to solve before it could live as we now know life. It must be able to aggressively acquire environmental calcium when required (81 80 130 270 283). It had to have a constant source of calcium available to plasma to support nerve and muscle functions. It also had to be able to construct a skeleton of considerable strength to protect the organism and to provide for motility. Finally, the reproductive needs of the animals had to be satisfied, including provision of calcium during embryonic and postembryonic development including construction of an entirely new skeleton of the unborn animals. It is, therefore, clear that the calcium homeostatic system is a very complex one (see Fig. 5) that involves many hormones, most of which are known but some that are not. From this discussion, it will become apparent that the vitamin D endocrine system is the basic one in managing calcium of plasma, with equally important roles for the parathyroid hormone and calcitonin.

B. Role of Parathyroid Gland and Its Hormone

For many years it has been clearly recognized that the parathyroid gland is the calcium-sensing organ in the body (266 270 283). Thus, in response to even slight hypocalcemia, the parathyroid glands react within seconds to secrete the 84-amino acid peptide hormone PTH (302). This hormone then initiates the sequence of events that results in the mobilization of calcium to replace that which has been taken from plasma.

There has been a great deal of recent effort expended in the direction of the calcium receptor for the parathyroid glands. The calcium receptor is well known (39 225) and appears to act in a cAMP-dependent mechanism to facilitate the secretion of PTH. This calcium receptor may play an important role in other tissues as well. Interested readers are directed elsewhere (301). The PTH has a lifetime in plasma that can be measured in minutes if not seconds (238). The receptor for the PTH is known and has been cloned (289). This receptor is found throughout the length of the nephron of kidney, is not found in the intestine, and is found in the osteoblasts but not osteoclasts of the skeleton (2). In the kidney, the PTH plays an important role in many functions (101). Well known is that it blocks reabsorption of phosphate causing a phosphate diuresis (39). In the proximal convoluted tubule cells, it activates the 25-hydroxyvitamin D-1α-hydroxylase (25-OH-D₃-1α-OHase) that converts 25-hydroxyvitamin D₃ to the active hormone, 1,25-(OH)₂D₃ (109 298 340). As already discussed in section i, the 25-OH-D-1α-OHase has now been cloned by three research groups (298 316 332) that will undoubtedly result in our understanding of the molecular mechanism whereby PTH activates the 1α-OHase. At the present time, it is known that PTH through cAMP (141) activates the 1α-OHase by increasing the mRNA encoding for this important enzyme (298). Whether it acts at the transcriptional level or elsewhere remains to be determined. At the same time the PTH through cAMP activates the 1α-OHase, it markedly suppresses the 25-OH-D-24-OHase, the major enzyme involved in destruction of the vitamin D hormone as described in section i (297). Again, the mechanism of suppression of 24-hydroxylase (24-OHase) by PTH remains unknown, although there is clearly a decrease in the mRNA encoding for the 24-OHase. These two actions result in a marked elevation of plasma levels of 1,25-(OH)₂D₃ (341).

C. Physiological Actions of 1α,25-Dihydroxyvitamin D₃

The consequences of an elevation of this major calcium mobilizing hormone are as follows: 1,25-(OH)₂D₃ acts by itself to initiate active intestinal calcium transport in the small intestine (35). This system has a relatively long lifetime, being measured in days (123), whereas the other actions of 1,25-(OH)₂D₃ are much shorter. 1,25-(OH)₂D₃ also activates osteoblasts. The result of this activation is to either stimulate the osteoclast to resorb bone and/or to activate the reverse transport of calcium from the bone fluid compartment to the plasma compartment (137 328 329 339). The final result is that calcium is mobilized by the skeleton into the plasma compartment by the action of the vitamin D hormone and the PTH. Of considerable importance is that vitamin D-deficient animals having abundant calcium in their bones will not mobilize calcium from the skeleton in response to PTH unless vitamin D is provided (110 271). Similarly, parathyroidectomized animals cannot mobilize calcium in response to 1,25-(OH)₂D₃ unless PTH is provided (110). Therefore, the presence of both the PTH and the vitamin D hormone are required for this system to operate in vivo. Whether this mechanism is through osteoclastic-mediated bone resorption or a membrane transport phenomenon remains to be determined. In the distal renal tubule, another mechanism proceeds that is dependent on both the PTH and the vitamin D hormone (186 362). Again, these two hormones acting in concert cause the reabsorption of the last 1% of the filtered load of calcium into the plasma compartment. These sources of calcium then cause a rise in serum calcium that then clears the sensing point of the calcium receptor. This then shuts down the secretion of the PTH. It does not appear that the PTH-related protein (PTHrP) functions in this system but may play a role in abnormal calcium mobilization as, for example, in malignancy (212).

D. Role of Calcitonin

The danger to hypercalcemia is calcification of soft tissues especially kidney, heart, aorta, and intestine, causing organ failure and death. To guard against hypercalcemia, not only is the shut-off of the parathyroid gland important but also the turning on of the C cells or the parafollicular cells of the thyroid to secrete the hormone calcitonin. This is a 34-amino acid peptide hormone that is responsible for lowering serum calcium by its action on the skeleton (57). It directly acts on osteoclasts and osteocytes reducing the calcium mobilizing activity and shutting down calcium coming from the skeleton (57). Although other actions of calcitonin have been described in kidney and intestine, by far the most important in the regulation of serum calcium is that which occurs at the skeleton. There have been reports of calcitonin regulating vitamin D metabolism (16 105); however, these largely are secondary to changes in parathyroid secretion, and there is no convincing evidence that calcitonin plays any direct role on regulation of the vitamin D hormonal levels.

E. Vitamin D Metabolites and Other Hormones

There has been considerable interest in other metabolites of vitamin D playing an important role in the regulation of calcium mobilization in suppression of hypercalcemia, or in bone growth. Of particular importance are the many studies carried out on 24,25-(OH)₂D₃ . This compound has been alleged to be important in the regulation of calcium homeostasis (245) or in the formation of the skeleton (33 230 255) or in counteracting the hypercalcemia activity of 1,25-(OH)₂D₃ (201). By now, extensive studies have been carried out with fluoro analogs, namely, 24,24-F₂-25-OH-D₃ to show that hydroxylation on the 24-position has no functional significance (38 149). Thus animals grown for two generations with 24,24-F₂-25-OH-D₃ as their sole source of vitamin D illustrate that 24-hydroxylation is not required either for skeletal formation, maintenance of calcemia, or growth of bone. More recently, a knockout of the 24-hydroxylase has been reported, and the results are not at all clear as to whether 24-hydroxylation plays a role except in the destruction of the potent hormone 1,25-(OH)₂D₃ and its precursor, 25-OH-D₃ (313). Further work using these models is required before any conclusions can be reached.

Other hormones such as estrogen and glucocorticoids have significant effects on bone and calcium metabolism, but they do not appear to be directly involved in regulating serum calcium concentration. This seems largely to be the role of the vitamin D hormone, the PTH, and calcitonin.

F. Intestinal Calcium and Phosphate Absorption

From a historical point of view, the role of 1,25-(OH)₂D₃ in intestinal absorption of calcium is perhaps best known. Orr et al. (256) discovered that vitamin D is required for intestinal calcium absorption many decades ago. This was reaffirmed by the work of Nicolaysen and Eeg-Larsen (239), who further demonstrated that the need for calcium increased the ability of the animal to absorb calcium. He postulated the existence of an endogenous factor that would inform the intestine of the skeletal needs for calcium. This basic observation was then shown to be primarily the vitamin D endocrine system, and 1,25-(OH)₂D₃ has been thought to be the agent that stimulates intestinal calcium absorption to meet the needs of the skeleton (34). It is, therefore, abundantly clear that intestinal calcium absorption remains as one of the basic functions of 1,25-(OH)₂D₃ . Quite independently of calcium is the role of 1,25-(OH)₂D₃ in stimulating intestinal absorption of phosphate (126 180). Both are active calcium transport mechanisms, but they appear to be independent of each other (63 126 180). The molecular mechanism of action of 1,25-(OH)₂D₃ in stimulating intestinal calcium absorption and intestinal phosphate absorption remains unknown, despite many efforts by many investigators. 1α,25-Dihydroxyvitamin D₃ stimulates the production of calbindin D9k in mammals (342) and calbindin D28k in birds (68) to appear in the intestine. In the case of the 28k protein in birds, it is absent in deficiency and present in large amounts after stimulation by vitamin D (357). A vitamin D-responsive element has been demonstrated to be present in the calbindin D9k promoter in mammals (77) and the 28k mammalian gene (113) but that has not been shown for the calbindin D28k protein of birds. The exact molecular mechanism for initiating production of the calbindin D28k remains to be determined. Furthermore, if one studies the time course of appearance of the 28k in birds as related to calcium absorption, there is no clear-cut correlation (125 312). The appearance of this protein and calcium transport coincide as a function of time in response to vitamin D or 1,25-(OH)₂D₃ . However, calcium absorption diminishes while calbindin D28k remains high in the gut. Therefore, there appears to be at least some discrepancy between calbindin D28k and calcium transport. It has suggested that some other protein or proteins are involved. This led to an analysis of a calcium pump in the basolateral membrane that is induced by 1,25-(OH)₂D₃ (357). However, the degree of induction in the time course of its appearance is not certain to account for the role of vitamin D in calcium transport. In short, the molecular mechanism of action of 1,25-(OH)₂D₃ in inducing intestinal calcium and phosphate transport is largely unknown. Wasserman and Feher (357) believe there are multiple sites of action of 1,25-(OH)₂D₃ in intestinal calcium absorption. Considerable work will be required before one can demonstrate the exact role of the calbindin proteins and the calcium pump in the vitamin D-induced calcium transport.

G. Vitamin D and Bone Calcium Mobilization

Even more poorly understood is the role of vitamin D in bone resorption or bone mobilization. The idea that vitamin D could result in the mobilization of calcium from bone was derived from the early work of Bauer et al. (14). A vitamin D-deficient animal on a zero-calcium diet will provide an increase in serum calcium at the expense of skeleton when given vitamin D. This mechanism requires the presence of PTH (110). Furthermore, 1,25-(OH)₂D₃ is clearly a stimulator of osteoclastic bone resorption in culture (269 319). However, the osteoclast has neither a receptor to the PTH nor a receptor to the vitamin D hormone (223). Instead, a signal appears to arise from interaction of these two hormones with the osteoblast. This signal causes the osteoclast to resorb bone (221 222 328). There is also the possibility that upon 1,25-(OH)₂D₃ and/or PTH signaling, the osteoblast may cause the transport of calcium from the bone fluid compartment to the plasma compartment (333). The nature of the signal arising from stimulation by the PTH and by the vitamin D hormone on the osteoblast has not been determined.

On the other hand, a great deal of progress has been made in understanding the role of vitamin D and osteoclastic formation and differentiation (328). Through the work of Abe et al. (1) and Tanaka et al. (334) has come the discovery that 1,25-(OH)₂D₃ plays an important role in causing the differentiation of promyelocyte to the monocyte. The monocyte is considered the precursor of the giant osteoclast. Further differentiation of the osteoclast precursor is catalyzed by the osteoclast differentiation factor that is produced by the osteoblast in response to 1,25-(OH)₂D₃ (328).

The action of vitamin D in stimulating osteoclastic bone resorption may be to provide bone calcium for the plasma but more likely to cause bone resorption in preparation for coupled formation to complete the bone-remodeling process. This would argue that the vitamin D hormone is involved in this important function that strengthens bone and repairs microfractures that may have occurred during bone usage. At this stage, therefore, 1,25-(OH)₂D₃ may be viewed as an important hormone that not only plays a role in bone calcium mobilization when required but also plays an important role in initiating bone remodeling and modeling systems required for shaping bone and required for repairing bone (107 319).

H. 1α,25-Dihydroxyvitamin D₃ Regulates the Parathyroid Gland

A new addition to the calcium homeostatic system was the discovery that the parathyroid gland is a target of vitamin D action. The initial work carried out by Stumpf et al. (324) showed that the parathyroid gland is a site of localization of highly labeled 1,25-(OH)₂D₃ . This was followed by two groups who demonstrated the presence of VDR in that tissue (133 145). Finally, a 1,25-(OH)₂D₃ suppression of parathyroid gland proliferation and of parathyroid production became known. A suppression of parathyroid secretion could be demonstrated in dialysis patients treated with intravenous 1,25-(OH)₂D₃ . The parathyroid gene was cloned (303), and the VDRE was demonstrated in the promoter region of the gene (85 184). This VDRE acts clearly in a mechanism to suppress transcription of the parathyroid gene. Thus a new loop in the calcium homeostatic system was discovered. Figure 5 shows all of these mechanisms that work together in the regulation of serum calcium concentration.

III. MOLECULAR MECHANISM OF ACTION AT TARGET CELLS

A. Overall Mechanism of Transcriptional Regulation by Vitamin D

The mechanism by which 1,25-(OH)₂D₃ exerts its effects on transcription is rapidly being uncovered. It is becoming clear that the vitamin D system shares many features with other ligand-activated nuclear receptors such as the retinoic acid receptor (RAR) and thyroid hormone receptor (TR) (95), including many of the same transcriptional cofactors. A summary of what is currently known about the activation process is shown in Figure 6.

In the first place, activation of vitamin D target genes has been shown to require a specific nuclear receptor protein, the VDR. Target genes are upregulated through binding of the VDR protein to specific DNA sequences termed response elements in the promoter regions of these genes. As shown at the top of Figure 6, binding of 1,25-(OH)₂D₃ to the receptor increases heterodimerization of VDR with a cofactor, the retinoid X receptor (RXR), on a response element (169). Binding of the heterodimer to the response element induces a bend in the DNA of the promoter (151). Binding of 1,25-(OH)₂D₃ to the receptor also appears to change the conformation at the COOH terminus of the VDR, permitting a region termed the AF-2 domain to interact with other transcription factors, including coactivator proteins such as SRC-1 (216 253 279 356). Recent exciting work indicates that coactivator proteins possess intrinsic histone acetylase activity (59 345). These coactivators bind to transcriptional “integrators” such as calcium-binding protein (CBP) and p300 (59 152 165), which, in addition to other functions, have also been shown to possess histone acetylase activity (179). Thus recruitment of coactivators to a promoter by a liganded receptor appears to result in remodeling of DNA structure through acetylation of histones and their subsequent release from DNA. This in turn leads to opening of the promoter to the transcriptional machinery. The net result of binding of liganded receptor to an upregulated target promoter is therefore to increase the rate of transcription of the gene, leading to increased production of the corresponding protein. It must be stressed at this point that a number of the details of this proposed mechanism are unclear at this point, including the exact sequence of events after receptor binding to the promoter, so the order of events presented in Figure 6 is somewhat speculative.

In contrast to genes that are upregulated by vitamin D, several genes, including PTH (85 251) and interleukin (IL)-2 (6), have been shown to be downregulated by 1,25-(OH)₂D₃ . The means by which this downregulation is carried out is not clear in all cases, and at least two possibilities exist. The first is that, as has been proposed for the IL-2 promoter, VDR may bind to a downregulatory response element and disrupt the binding of upregulatory transcription factors, leading to a decrease in transcription (6). For other downregulated genes, the situation may be quite different, and binding of VDR to an inhibitory response element may lead to interactions with repressor proteins that decrease transcription of the gene. Interestingly, corepressor proteins such as nuclear receptor corepressor (NCoR) (142) and silencing mediator for retinoic acid receptor and thyroid hormone receptor (SMRT) (60) have recently been found to bind, through intermediary proteins, to histone deacetylase enzymes (131 233). Presumably in this situation the deacetylated histones then bind to the promoter of the downregulated gene and shut off transcription. The VDR has not yet been reported to bind to corepressor proteins, but given the rate at which such proteins are currently being discovered, it is possible that a corepressor that binds to VDR will be uncovered.

Phosphorylation of the receptor may also play a role in the induction of transcription by the VDR (73), perhaps through modulation of the affinity of VDR for the various cofactor proteins involved in transcription. Rapid phosphorylation of the VDR has been shown to occur in organ culture systems upon addition of ligand (41). The phosphorylated residues have been localized to the ligand-binding domain of the protein. The exact functional consequences of this phosphorylation have been difficult to determine. Estrogen receptor phosphorylation by mitogen-activated protein kinase has been shown to have a direct and measurable effect on transcription (167), but this phosphorylation was mapped to the NH₂-terminal AF-1 region of the estrogen receptor, which is lacking in VDR. Effects of phosphorylation on activation of the AF-2 domain, at the COOH terminus of the protein, have not yet been clearly shown for the VDR. Indeed, even the kinase (or kinases) responsible for the phosphorylation in vivo has yet to be determined. Some studies have suggested that serine-208 of the VDR can be phosphorylated by casein kinase II and that phosphorylation of this residue may cause increased transcriptional activity (162). Other work has revealed that mutation of this serine failed to affect transcription, although alternate phosphorylation on adjacent serines was noted (135). The exact effects of phosphorylation must await determination of the precise amino acids phosphorylated in vivo and the functional consequences that follow from this.

B. Vitamin D Receptors

The VDR is a member of a superfamily of nuclear receptors (95). Within this family, the VDR has the highest similarity to the subfamily that includes retinoic acid, thyroid hormone, and peroxisome proliferator activator receptor (PPAR) receptors, to which it has sequence and structural resemblance. The VDR has been cloned from several species (12 42 92 164 219) and shows considerable similarity between species in size and sequence. In the rat, for example, the VDR protein consists of 423 amino acids, with a molecular mass of ∼50 kDa, whereas in the human the protein has an additional 4 amino acids at the NH₂ terminus, for a total of 427.

Like the other nuclear receptors, the VDR can be divided by function into several domains. An illustration of the different domains of the VDR is shown in Figure 7. At the NH₂ terminus is a truncated A/B domain of ∼20 amino acids, to which little function has yet been ascribed for the VDR. After this, the DNA-binding domain, termed the C domain, is located between amino acids 20 and 90. A D or hinge domain is located approximately between amino acids 90 and 130, followed by the COOH-terminal E or ligand-binding domain between amino acids 130 and 423. The ligand-binding domain of the protein is a complex region of the protein, responsible for high-affinity binding of ligand, for dimerization with RXR, and for binding to transcription factors (95). It should be noted that exact delineation of the division between the hinge region and the ligand-binding domain is somewhat uncertain and is based on deletion analysis.

Knowledge of the structural and functional properties of the receptor proteins has increased dramatically in recent years, after the advent of systems which allow expression and purification of large quantities of receptor protein from bacteria. This has allowed a piecemeal approach to determination of receptor domain structures; the DNA-binding and ligand-binding domains of several receptors have been expressed separately and purified, and structures have been determined by NMR or X-ray crystallography (32 272). In contrast, to this point, information concerning the VDR has come primarily from site-directed mutagenesis, in which the receptor cDNA is mutated and the mutant protein studied by transfection into mammalian cells (236). No structural data are yet available concerning the VDR, but the other receptor domains may serve as models on which to base hypotheses about the VDR and its properties.

Structures for the DNA-binding domains of other receptors, including the RXR, have been obtained using both NMR and X-ray crystallography. The RXR structure showed that the DNA-binding domain, comprised of two zinc finger motifs, consists of two α-helixes oriented at approximately right angles to one another (194). One helix, termed the orientation helix, thought to be critical for recognition of the receptor response element was proposed to fit into the major groove of the DNA and bind to a specific DNA sequence in the response element. These domains contain two zinc atoms tetrahedrally coordinated to eight conserved cysteine residues. Both the zinc atoms and the cysteine residues are necessary to maintain the three-dimensional structure required for response element recognition and DNA binding. The amino acid sequence of the DNA-binding domain is similar between members of the receptor superfamily, suggesting that these structures can be used as a model for that of the VDR.

There is as yet no direct evidence for the structure of the VDR-RXR complex bound to DNA. However, the crystal structure of the TR and RXR DNA binding domains complexed to a DR4 response element has been determined (272). With the use of this structure as a model, the VDR DNA-binding domain was hypothesized to have specific amino acid contacts with RXR: between the asparagine residue 14 of VDR and RXR residues glutamine-49 and arginine-52. In addition, lysine-68 and glutamate-69 of VDR were modeled to form salt bridges with RXR residues aspartate-39 and arginine-38, respectively. These interactions were thought to account for the optimal spacing of three residues between the direct repeats of the vitamin D response element (DR3).

The crystal structures of the ligand-binding domains of the RXR-α (without ligand) (32), the RAR-γ (with ligand) (279), and the TR-α₁ (with ligand) (356) have been determined. The ligand-binding domains of the RXR and RAR had been previously predicted to have a high content of α-helix (64 204), and this was confirmed by structural analysis. All three ligand-binding domains were found to share a common secondary structure of 12 α-helixes, with a small content of β-sheet. A comparison of the structures of the RAR and TR with that of the VDR is shown in Figure 8. The COOH-terminal portion of the proteins, termed the AF-2 domain, has been determined to be critical for transcription. Removal of this portion of the protein results in decreased ligand-binding affinity and loss of transcriptional activation. The structural data indicate that the helixes 11 and 12 of the RAR and TR may undergo a large conformational change in response to ligand binding, folding up around the ligand in what the authors of the RAR paper term a mouse trap action to form a hydrophobic pocket (279). This brings the amino acid residues on helix 12 which make up the AF-2 domain into position to interact with other transcription factors. The VDR shows sequence similarity to the RAR and TR in this region and may be expected to possess a similar three-dimensional structure and to undergo a similar conformational change upon ligand binding.

The wild-type VDR binds its ligand, 1,25-(OH)₂D₃ , with extremely high affinity, in the range of 10⁻¹⁰ M (208 282). Both the 1α- and 25-hydroxyl groups are critical for high-affinity binding, and the absence of either results in approximately a 500-fold decrease in affinity of ligand for the receptor (82). The exact amino acid residues in the receptor that contact the ligand remain unknown, although some work has been done recently on affinity labeling of the binding site (273). Previous work with transfected cells has shown that deletion of the NH₂-terminal 116 amino acids of the VDR left a protein with measurable ligand binding activity, whereas deletion of the NH₂-terminal 160 amino acids did not (220). More recent work with fusion proteins has shown that deletion of the NH₂-terminal 124 amino acids gave rise to a functional ligand-binding protein, whereas deletion to amino acid 172 did not (242). From the other end, removal of the COOH-terminal 20 amino acids of the VDR resulted in a 10-fold loss of affinity for ligand, whereas removal of more than this number resulted in complete loss of ligand binding (236). Thus a core domain of ∼300 amino acids was shown to be required for the protein to bind ligand with wild-type affinity. This is in accordance with the ligand-binding domains of the related receptors that have been crystallized, which were all expressed in bacteria as proteins of between 250 and 300 amino acids.

An alignment of the VDR ligand-binding domain sequence with that of the RAR and TR supports the possibility that amino acids in the VDR directly in contact with ligand begin at approximately amino acid 220. In Figure 8, amino acids in contact with ligand in the RAR and TR crystal structures are shaded. Inspection of this figure shows that the contact amino acids begin at helix 3 in both the RAR and TR. Based on sequence alignment, the same region of the VDR begins at approximately amino acid 220, with a leucine residue that is conserved in all three proteins. Interestingly, work in which 10 amino acid segments of the VDR were sequentially deleted found that impairment of ligand binding did not occur until approximately amino acid 230, which is in accordance with this possibility (154).

Site-directed mutagenesis of the VDR ligand-binding domain has been performed recently on several of the cysteine residues in the human protein (235). Alteration of cysteine-288 to glycine resulted in severe attenuation of ligand binding at room temperature, whereas the same mutation at cysteine-337 resulted in a smaller decrease in affinity. The exact significance of this result is uncertain, although a contact between cysteine-237 and carbon-13 of retinoic acid was noted in the binding of retinoic acid to the RAR-γ (279). Interestingly, in Figure 8, the corresponding cysteine residue (cysteine-284) of the rat receptor aligns exactly with ligand-contacting amino acids in the RAR and TR, suggesting that this cysteine may in fact directly contact the ligand.

The VDR is generally expressed at relatively low levels in vivo. Target tissues, such as bone, kidney, and especially intestine, may have relatively high levels of receptor (3,000–6,000 fmol/mg protein), but in other tissues, the levels are generally much lower (72). The receptor has been shown to be present in most tissues that have been examined, including activated immune cells such as T cells, where it may play a role in modulating the levels of cytokines such as IL-2 (6). In contrast to other receptors, such as the glucocorticoid receptor, which is associated with a number of proteins, including heat shock proteins, in the cytoplasm before hormone binding (146), studies with radiolabeled 1,25-(OH)₂D₃ have shown that the VDR is predominantly nuclear (325). Little information is available concerning whether any proteins associated with the VDR before DNA binding. There are some data suggesting that many hormone-binding receptors, including VDR, contain a binding site for calreticulin in the DNA-binding domain (43). Some reports indicate that calreticulin may bind to VDR and inhibit activation of vitamin D target genes (358), but the physiological significance of this is as yet unclear.

There have been numerous reports in recent years of rapid nongenomic effects of vitamin D on calcium transport, termed transcaltachia (241). The physiological importance of these data is unclear. Whether these effects are mediated by the VDR or a different protein is also unclear. The development recently of a mouse model in which the VDR has been ablated (365) may provide some insight into the relevance of these reports, and into whether the VDR plays any role in them.

Recently, the promoters of the mouse (150) and human (226) VDR have been isolated. Some studies have suggested that the VDR is upregulated by treatment with agents such as forskolin, which induce protein kinase A (183). Further study of these promoters will shed more light on the means by which VDR protein levels are regulated inside target cells.

C. Retinoid X Receptors and Other Coactivators

In the absence of cofactor proteins, the VDR is unable, at physiological concentrations of protein, to bind to most response elements that have been described. This has become clear from both in vitro gel retardation assay experiments and from experiments involving receptor expression in yeast (155 231 310). Work with the VDR and with related receptors indicated that the RXR is the required cofactor.

The RXR was isolated several years ago, and initially its ligand and functional importance were unknown (211). Subsequent work identified 9-cis-retinoic acid as a ligand for RXR (134). It also became clear that RXR plays a critical role in binding to DNA of several different receptors, including the VDR, the TR, the RAR, and the PPAR, among others (98 173). In the absence of RXR, it appears that none of these receptors binds efficiently to their response elements. Like other receptors, the RXR can be divided into functional domains, including an NH₂-terminal A/B domain, a C domain which binds to DNA, and a COOH-terminal DE domain that binds ligand and activates transcription (58 210). Heterodimerization with other receptors appears to be mediated through two interfaces: one in the DNA-binding C domain (366) and another in the ligand-binding E domain (359). Interestingly, in structural studies, the RXR ligand-binding E domain was found to crystallize as a dimer, and the dimer interface occurred along helices 9 and 10, exactly where mutagenesis studies suggest RXR heterodimerizes with other receptors (32 261). This region of RXR from amino acids 389 to 429, termed the I box, is sufficient for strong interaction between RXR and TR, but the same sequence provides only weak interaction with VDR. Additional NH₂-terminal RXR sequence is required for strong interactions with VDR (261), indicating a potential distinction between VDR and related receptors.

Most vitamin D response elements consist of two half-sites of the sequence AGGTCA separated by three bases (see sect. iii D). Because RXR is required for binding to the response element, the question arose as to whether one half-site was preferred by RXR, and if so, whether this preference was for the upstream or downstream site. For VDR, as for the other receptors that heterodimerize with RXR, it appears that RXR binds to the 5′-half-site and VDR the 3′-half-site (189 366), although there may be some exceptions to this rule. The polarity of this binding has been shown to be important for maximal gene activation, with significantly lowered transcription rates occurring if the response element orientation is reversed relative to the promoter start site. Binding of the VDR-RXR heterodimer to a response element is greatly increased by 1,25-(OH)₂D₃ when salt concentrations are in the physiological range, i.e., 100–150 mM (169). Interestingly, for many of the receptors that heterodimerize with RXR, the natural ligand of RXR, 9-cis-retinoic acid, may not enhance DNA binding. Some findings indicate that RXR may be unable to bind ligand while in a complex with another receptor (98). In addition, some in vitro reports indicate that the VDR-RXR complex is disrupted by addition of 9-cis-retinoic acid (205). This contrasts with work in transfected cells indicating that both 1α,25-(OH)₂D₃ and 9-cis-retinoic acid enhanced reporter gene expression from a vitamin D-responsive promoter (291). The reason for these conflicting results is unclear; it may be an experimental artifact, or it may be that the additional proteins that interact with the VDR-RXR heterocomplex in vivo, such as the coactivator/corepressor proteins, affect the conformation of the RXR and permit ligand to bind.

The VDR has been shown to interact directly with a growing number of other transcription factors, including transcription factor IIB (TFIIB). The interaction of VDR with TFIIB has been characterized in several reports (27 206). TFIIB is a 30-kDa protein that was originally isolated as a cofactor associated with TATA-binding protein (120). A region of VDR of ∼70 amino acids in the D domain has been reported to bind to a 43-residue NH₂-terminal region of TFIIB (27 203). The sequence of TFIIB in this region is similar to that of a zinc finger, suggesting that the zinc finger motif may function in protein-protein interactions as well as DNA binding. Interestingly, one report suggests that binding of ligand to VDR causes dissociation of TFIIB from the receptor, which may indicate a complex interaction between proteins before the initiation of transcription (217).

Interactions between the VDR and other factors are less well characterized, although this work is progressing, particularly with regard to the coactivator proteins. Recent work indicates that VDR binds to SUG1/TRIP1, a nuclear protein that also binds to other receptors (355). It should be noted that the function of SUG1 in VDR-mediated transcription is unclear. SUG1 binds to the AF-2 domain of the VDR (216 355), but because SUG1 has been found to be a component of the proteasome complex in the nucleus, it may function primarily in receptor degradation.

The VDR has also been shown to bind to a growing number of coactivator proteins such as SRC-1 and TIF-1 (216 355). Other coactivator proteins that have been discovered recently, including ACTR (59) and p/CIP (345), join the rapidly enlarging group of such proteins. The coactivators seem to fall into a family with many features in common, including size (∼1,400 amino acids), structure (NH₂-terminal Per-Arnt-Sim/basic helix-loop-helix domains, central receptor interaction domains) and activity (COOH-terminal histone acetylase domains). It seems likely that most, if not all, of the coactivator proteins will be found to interact with VDR and activate transcription. For example, both SRC-1 and TIF-1 have been shown to bind to the VDR (216 355), whereas nuclear receptor coactivator protein (ACTR) has been shown to enhance transcription from a vitamin D target gene in transfected cells (59). However, it is not yet clear whether all coactivators interact with VDR equally well; some may function more effectively than others in the vitamin D system. An additional possibility is that tissue-specific expression may be a significant factor in determining which coactivator works with the VDR. For example, the coactivator ACTR was shown to be expressed at relatively low levels in kidney, an important vitamin D target tissue (59). This may indicate that other coactivators play a more important role in transcriptional activation of vitamin D-controlled genes, at least in this tissue.

Whether VDR interacts with corepressor proteins to bring about downregulation of target genes is currently unclear. Like the coactivator proteins, the corepressor proteins appear to have many features in common. The corepressors SMRT and NCoR (142), for example, have considerable sequence similarity, although NCoR possesses a large NH₂-terminal region lacking in SMRT. In addition, both proteins appear to repress transcription through interactions with the Sin3 proteins, which bind to histone deacetylase enzymes (131 233). However, unlike related receptors such as retinoic acid and TR, it is unclear whether the VDR is bound by corepressor proteins. At least one corepressor protein, NCoR, has been reported not to bind to VDR at all. NCoR binds to a conserved region of the TR and RAR, termed the CoR box, to exert its effects; binding of ligand to the TR or RAR causes dissociation of NCoR from receptor. NCoR does not bind to the CoR box region of the VDR, despite the similarity of the VDR sequence in this region to that of TR and RAR (142). This may explain why the VDR does not seem to exhibit dominant negative silencing seen with both the TR and RAR (262 286). The possibility remains that other corepressor proteins will be found to bind to the VDR and that this interaction is responsible for silencing at least some of the genes that are downregulated by vitamin D.

D. Vitamin D Responsive Elements

As mentioned in section iii A, responsive elements are the sequences of DNA, isolated from the promoters of vitamin D-responsive genes, that are bound by the VDR. One of the first such response elements isolated was that of the rat osteocalcin gene (84). Work with this and other vitamin D response elements, and with the response elements of other receptors, led to the demonstration that many response elements consisted of two repeats of the half-site sequence AGGTCA separated by several nonspecified bases (75 173 349). The variable that determines receptor specificity seems to be the number of nucleotides separating the half-sites. The RXR has been shown to have a preference for two half-sites separated by a single nucleotide, termed a DR-1 response element, whereas the VDR binds to a DR-3 element, the TR to a DR-4 element, and the RAR to a DR-5 element (349). Interestingly, almost all of the naturally occurring VDRE isolated from genes upregulated by the VDR have fallen into the DR-3 category. The negatively regulated genes, such as those for PTH and IL-2, may not follow this rule. Initial work with the PTH gene indicated that the response element in this case consisted of only a single half-site. However, more recent work indicates that the PTH VDRE may in fact consist of two half-sites (76), which may leave only a few exceptions, such as the IL-2 gene, to this rule (6).

One interesting aspect of the interaction between receptor and response element is the bend in the response element DNA that is induced by binding of the VDR-RXR heterodimer (151). This has been reported for binding of related receptors, such as the RAR, to their response elements (71). The degree of bending induced by the protein complex is reportedly on the order of 30°. Although this bend is much lower in magnitude than that reported for binding of the TATA-binding protein to DNA (240), it is still appreciable. It is interesting to speculate on the possible role of this induced bending, particularly in light of the finding that coactivator proteins and their cofactors acetylate histones. It may be that the combination of DNA bending and histone dissociation acts to make the promoter of the target gene more accessible to transcription factors before transcriptional activation.

Some recent reports have suggested that a greater degree of flexibility exists for binding of VDR to response elements. Work in vitro has suggested that VDR can bind to DR-3 response elements, such as the osteopontin VDRE, as homodimers (291). Other work suggests that TR and RAR can act as heterodimeric partners with VDR on a response element, instead of RXR (292). In addition, response elements that do not fall into the DR-3 category have been suggested to bind VDR, including DR-6 elements (264) and one element composed of an inverted repeat of two half-site elements separated by nine bases (292). At present, the significance of this work is unclear. For instance, it has been reported that although VDR can bind to the osteopontin response element as a homodimer, the homodimer is transcriptionally inactive and the VDR-RXR heterodimer is required for transcriptional activity (197). Thus the results of such in vitro experiments must be interpreted with caution and, if possible, in vivo. For example, the report of VDR heterodimerization with TR and RAR may potentially be evaluated through the use transgenic animals in which RXR has been ablated, to determine whether VDR-responsive genes can still be expressed. Unfortunately, ablation of RXR-α, the preferred partner for VDR, is a lethal mutation (90), but it is possible that inducible knockouts of the RXR-α will provide insight into whether these other receptors can heterodimerize with VDR and rescue the expression of VDR target genes.

E. Vitamin D-Dependent Genes, Their Roles, and Gene Complexity

Many different genes have been shown to be responsive to vitamin D. Most of these genes play a direct role in calcium endocrinology or bone formation and were the earliest examples of vitamin D response elements isolated. Examples of these include osteocalcin (84), osteopontin (243), PTH (251), the hydroxylases CYP24 (368) and CYP1α (298 316 332), and the calbindin genes (69). Many of these have been reviewed recently (75). The roles of many of these genes in calcium endocrinology have become more evident in recent years. For example, the bone protein osteocalcin is secreted by osteoblasts and represents the single greatest protein constituent of bone by mass. Expression of this protein is upregulated by vitamin D, and it may play a significant role in maintaining bone integrity (73). On the other hand, several genes have been shown to be downregulated by vitamin D, including the gene for PTH, which plays a critical role in controlling the levels of 1,25-(OH)₂D₃ in serum.

Several of the hydroxylase enzymes that act directly on 1,25-(OH)₂D₃ formation and degradation have been shown to be controlled at the transcriptional level by 1,25-(OH)₂D₃ . In particular, the hydroxylase enzyme CYP24 is one of the most highly regulated genes that responds to vitamin D. This enzyme represents the major means by which 1,25-(OH)₂D₃ is catabolized in vivo, and its importance is emphasized by recent gene knockout experiments in which loss of the enzyme was found to be lethal to at least 50% of the recipient mice (313). The CYP24 gene is highly upregulated by 1,25-(OH)₂D₃ and contains two sets of VDRE in its promoter (61 168 368). Both sets of elements are required for the gene to display maximum inducibility by vitamin D (168 368). It therefore appears that the presence of multiple copies of response elements are one way in which vitamin D target genes can be efficiently upregulated.

In contrast to the preceding example, the hydroxylase enzyme CYP1α, which is responsible for catalyzing the formation of 1,25-(OH)₂D₃ , is highly downregulated by its product. This enzyme, expressed in the proximal convoluted tubule of the kidney, is regulated with extreme stringency, although the exact details of the way in which this regulation is exerted are unclear. The recent cloning and expression of this enzyme open the door to a more complete understanding of its regulation, and details of the way in which the gene and protein expression are regulated are sure to be forthcoming (298 316 332).

Another set of genes that have been shown to be regulated by vitamin D are the calbindins. In mammals, the calbindin D₉K is expressed primarily in intestine, whereas the calbindin D₂₈K is expressed in kidney and other tissues. The calbindin D₉K may play a role in absorption of calcium from the gut, whereas the calbindin D₂₈K may be involved in reabsorption of calcium from glomerular filtrate, although this has not been definitively shown. The promoters of both of these genes have been cloned (74 153 331), and both have been reported to contain VDRE. Recent work on targeted gene disruption of the D₂₈K has recently been reported in the mouse (311). This is not a lethal mutation, and the effects of this knockout on regulation of calcium levels will be of great interest in deciphering the role of this protein in calcium transport.

More recent and somewhat more surprising discoveries of genes that are regulated by vitamin D include the discovery of VDRE in the promoters for a number of genes not traditionally considered vitamin D targets. These include potential response elements in the promoters of several genes, including integrin β₃ (52), fibronectin (264), atrial natriuretic factor (163 198), c-fos (48 290), PTHrP (182), and p21 (199). The significance of the vitamin D regulation of many of these genes is unclear and may prove an interesting area of study. In this regard, the recently developed mouse strain in which the receptor has been ablated may provide an interesting vehicle for assessing the contribution of the VDR to regulation of these genes (365).

F. Target Cell Metabolic Enzymes

The metabolism of vitamin D by target cells has been shown to occur in a number of tissues, including kidney, intestine, and bone. In kidney and intestine in particular, upregulation of the 24-hydroxylase enzyme in response to 1,25-(OH)₂D₃ treatment is rapid, occurring within 4 h, as shown by both Northern and enzymatic analyses (297). Cultured bone cells have also been shown to possess highly inducible 24-hydroxylase activity (202). These cells have been used to isolate catabolic products of 1,25-(OH)₂D₃ that can be placed in a logical order on a degradative pathway leading from the active compound to inactive excretion products (82). Thus it appears that the target cells for vitamin D contain the means to regulate its activity at the cellular level. The additional level of control that this target cell metabolism provides over local concentrations of 1,25-(OH)₂D₃ may be an important means by which tissues regulate the responsiveness of genes to vitamin D.

IV. RECENTLY DISCOVERED FUNCTIONS OF 1α,25-DIHYDROXYVITAMIN D₃

A. Discovery of New Target Organs for 1α,25-Dihydroxyvitamin D₃

The first insight into possible new functions of 1,25-(OH)₂D₃ beyond its actions in the regulation of calcium and phosphorus as described in section ii came as a result of studies involving the cellular and subcellular localization of radiolabeled 1,25-(OH)₂D₃ at high specific activity. This allowed the administration of physiological doses of 1,25-(OH)₂D₃ and its detection by frozen section autoradiography in target tissues (324 325). The result of this study in the late 1970s illustrated that not only was radiolabeled 1,25-(OH)₂D₃ localized in the nuclei of the enterocyte of the small intestine, the distal renal tubule of the kidney, and the osteoblasts of bone, but were found in tissues not previously considered targets of vitamin D action. These localizations were found specific inasmuch as the localizations could be prevented by the administration of unlabeled 1,25-(OH)₂D₃ but not 25-OH-D₃ . Of great interest was the finding of ³H-labeled 1,25-(OH)₂D₃ in the nuclei of the islet cells of the pancreas, keratinocytes of skin, ovarian tissue, mammary epithelium, epithelial cells of the epididymis, certain neuronal tissue, promyelocytes, macrophages, and T lymphocytes. This unexpected development focused on the possibility that 1,25-(OH)₂D₃ may carry out functions not previously appreciated and in tissues not previously considered targets of vitamin D action. This result was followed by the demonstration of the VDR in the parathyroid gland (133 145), in skin (305), in the thymus (267), and in ovarian cells (88). Following this lead came the demonstration that the vitamin D hormone carries out functions in many of these tissues, and still other functions remain to be discovered.

B. Role of Vitamin D Hormone in the Parathyroid Gland

Perhaps the most well-established new function of 1,25-(OH)₂D₃ is in the parathyroid gland. Specific localization of 1,25-(OH)₂D₃ in the parathyroid gland (325) and the presence of VDR was provided by two independent laboratories (133 145). However, there had been considerable controversy as to whether the 1,25-(OH)₂D₃ could suppress secretion or production of the PTH (114 115 302). Early work failed to provide support for the idea that 1,25-(OH)₂D₃ could, in fact, suppress parathyroid secretion in isolated parathyroid glands. Furthermore, 25-OH-D₃ was continuing to be used as a suppressant of secondary hyperparathyroidism in renal failure patients on dialysis (274). Notably, however, large amounts of this compound were needed to provide any degree of suppression, usually of the order of 20 μg/day. Suppression of parathyroid levels in such patients with orally administered 1,25-(OH)₂D₃ was also not dramatic. However, when an intravenous route was used, excellent treatment of secondary hyperparathyroidism of renal failure resulted (308). These results strongly suggested that 1,25-(OH)₂D₃ may have a direct action through its receptor in the parathyroid glands. Thus PTH secretion by isolated parathyroid glands or cells could be suppressed by the direct administration of 1,25-(OH)₂D₃ (82–84). This suppression required some time to achieve. With the cloning of the preproparathyroid gene came a search for the possibility that this gene might be under control of 1,25-(OH)₂D₃ through its receptor. Analysis of mRNA encoding the PTH by Northern blot suggested a transcriptional suppression by 1,25-(OH)₂D₃ (284 296 304). Subsequently, work by Demay et al. (85) and Silver et al. (303) have shown unequivocally the presence of a VDRE in the promoter region of the parathyroid gene. Although this was first thought to be a single 6-base response element, subsequent work has revealed that in actual fact it represents a DR-3 in which two repeat sequences are indeed found at 123–108 bases 5′ from the transcriptional start site (76). This responsive element system causes a suppression of expression of the preproparathyroid gene promoter. Details of the mechanism are discussed in section iii.

In addition to the role of the vitamin D hormone in suppressing the parathyroid gene is the idea that the vitamin D hormone also suppresses proliferation of the parathyroid gland cells. The discovery by Suda and co-workers (1 334) that 1,25-(OH)₂D₃ can suppress the growth and stimulate cellular differentiation of the promyelocytes brought forth the idea that 1,25-(OH)₂D₃ might also suppress proliferation of the parathyroid gland cells. This rationale provides further impetus to the idea that renal osteodystrophy patients should be treated with 1,25-(OH)₂D₃ to prevent the development of hypertrophied parathyroid glands as well as to suppress secretion of the PTH.

C. Role of Vitamin D Hormone in Skin

After the discovery by Suda and co-workers (1 334) of the differentiative action of 1,25-(OH)₂D₃ on the promyelocytes came the application of this finding to the keratinocytes. Hosomi et al. (144) probably provided the first report that 1,25-(OH)₂D₃ induces keratinocyte differentiation. These results were found by the in vitro addition of 1,25-(OH)₂D₃ . Similar results were reported by Holick and co-workers (309), and it is now common knowledge that the keratinocytes are induced to differentiate by the in vitro addition of 1,25-(OH)₂D₃ . Exactly how important this differentiation effect of 1,25-(OH)₂D₃ is in vivo is difficult to assess. Certainly, vitamin D-deficient animals do not have a problem with keratinocyte differentiation. Thus hyperproliferation of the keratinocyte and failure to differentiate is not found in vitamin D-deficient animals (79; DeLuca, personal observations). It is possible that the differentiation may be triggered by more than one system providing a redundancy and hence the failure to observe a lesion in skin as a result of vitamin D deficiency. Together with the differentiation of the keratinocyte comes an inhibition of proliferation. This inhibition of proliferation has been utilized in the treatment of hyperproliferative diseases of skin as, for example, psoriasis (144 309). Both 1,25-(OH)₂D₃ and analogs especially MC903 of Leo Pharmaceuticals (also termed Dovenex) can be used as a significant therapy against psoriasis with as many as 70% patients responding to this treatment (44).

A number of studies have been carried out on the mechanism whereby 1,25-(OH)₂D₃ brings about the differentiation of the keratinocyte. During the differentiation process, there is an increase in the protein involukrin (326). There is an increase in transglutamanase activity (326) and cornified envelope formation (263). However, exactly how 1,25-(OH)₂D₃ induces differentiation of the keratinocyte and inhibits proliferation is unknown and remains to be investigated.

Finally, it has been proposed that the keratinocyte functions in a paracrine fashion in which 1,25-(OH)₂D₃ is produced by the keratinocyte itself, and this compound serves as a paracrine to stimulate differentiation of the keratinocyte (21). If this is true, it remains to be explained how vitamin D-deficient animals are able to obtain mature and functional keratinocytes. There have been very dramatic statements made that the keratinocyte is fully able to metabolize 25-OH-D₃ to 1,25-(OH)₂D₃ but that it lacks the ability to 25-hydroxylate 1α-OH-D₃ (21). In work carried out in the rat by two different laboratories (277 299), there was a failure to detect any ³H-labeled 1,25-(OH)₂D₃ in skin or other tissues in nephrectomized rats given the highest specific activity (160 Ci/mmol) of ³H-labeled 25-OH-D₃ available. Thus, in normal rats, the keratinocyte was unable to produce significant amounts of 1,25-(OH)₂D₃ in vivo. On the other hand, there are a number of scientists who believe that this does, in fact, occur in vivo (21). Thus the role of 1,25-(OH)₂D₃ in the keratinocyte under normal physiological circumstances is unknown, and certainly the idea that the keratinocyte in vivo is able to produce 1,25-(OH)₂D₃ is not universally accepted. Recently, three laboratory groups have succeeded in cloning the 25-OH-D-1α-OHase (102 103 298 316 332), and one report used PCR techniques to clone it from human keratinocytes (102). This could support the presence of this enzyme in keratinocytes; however, Northern analysis of mouse skin (330) did not confirm 1α-OHase mRNA in this tissue. Additional work will be required before this area is clarified. However, certainly one has to consider the keratinocyte as a potential target of 1,25-(OH)₂D₃ to the extent that analogs are therapeutically effective against the hyperproliferative keratinocyte disease psoriasis.

D. Role of 1α,25-Dihydroxyvitamin D₃ in the Immune System

The presence of the VDR in activated T lymphocytes was reported by Provvedini et al. (267), who also showed that resting lymphocytes do not express the VDR. Certainly the thymus, which is a repository of immature lymphocytes, is a source of VDR, since it was used as a source of receptor for measurements of 1,25-(OH)₂D₃ (138). Among the antigen-presenting cells, the macrophage that is derived from the monocyte has receptors for 1,25-(OH)₂D₃ (20 264). There is no doubt, therefore, that the immune system does indeed possess receptors to 1,25-(OH)₂D₃ . These results suggest a role for 1,25-(OH)₂D₃ in the immune system. However, the role is just now beginning to be defined. There have been many reports using peripheral blood lymphocytes that suggest that production of IL-2, tumor necrosis factor-α (TNF-α), and interferon-γ is suppressed by 1,25-(OH)₂D₃ (193). Furthermore, there are reports of increased synthesis of transforming growth factor-β in a number of immortalized cell lines in response to 1,25-(OH)₂D₃ in vitro (177). In fact, there are many conflicting reports regarding the in vitro response of the lymphocytes to 1,25-(OH)₂D₃ . This review focuses on what is known concerning the in vivo responses of the immune system in the absence and presence of 1,25-(OH)₂D₃ and its analogs.

In a study of delayed hypersensitivity, which is a T-helper cell lymphocyte-mediated response, vitamin D deficiency markedly reduces the ability of the mouse to respond (364). However, large amounts of 1,25-(OH)₂D₃ and its analogs will also suppress the delayed hypersensitivity response (363). These results illustrate that the T-helper cell lymphocyte system is vitamin D responsive but that both immunostimulation and immunosuppression can be found in the in vivo situation. Currently, there is little or no evidence to support the idea that B cell-mediated immunity is affected by 1,25-(OH)₂D₃ .

The most dramatic results obtained to date in the immune system are those found in several autoimmune diseases. Most notably is the autoimmune disease known as experimental autoimmune encephalomyelitis (EAE) that can be induced in B10.PL mice or SJL mice by the injection of myelin basic protein together with small amounts of pertussis toxin (49). This treatment induces a multiple sclerosis-like disease (EAE). This disease is aggravated by agents that stimulate T-helper (Th)-1 cells and that increase interferon-γ and TNF-α secretion (280). Injections once every other day of 1,25-(OH)₂D₃ in some cases produced a reduction in the severity and in other cases produced very slight responses (36 196). However, since 1,25-(OH)₂D₃ has a lifetime measured certainly in hours, the administration of this compound three times a week or every other day would not be expected to be sufficient (116). More recently, EAE in B10.PL mice could be completely prevented by the addition of 1,25-(OH)₂D₃ to the diet of such mice. Furthermore, if the hormonal substance was begun after the EAE had begun, the development of symptoms could be prevented (Fig. 9) (49). Current results strongly suggest that the vitamin D hormone and its analogs are functioning by stimulating Th-2 T-helper cells to produce transforming growth factor-β1 and IL-4 that might serve to suppress the TNF-α and interferon-γ production by Th-1 cells (51). Similar results were obtained with another autoimmune disease, rheumatoid arthritis induced by Borrelia burgdorferi or Lyme's disease (50) or arthritis produced by the injection of collagen into susceptible mice (50). Some improvement has also been reported for the autoimmune induction of diabetes in the nonobese diabetic mouse (218). The idea, therefore, that the vitamin D compounds may play a role in modulating the immune system is strongly supported. Furthermore, there is the idea that 1,25-(OH)₂D₃ and its analogs might become useful in the prevention or treatment of autoimmune diseases.

Of some interest is the idea that the immunomodulatory action of vitamin D might be useful in the management of transplant rejection. Initial work indicates that the vitamin D hormone and its analogs would only be marginally useful in prevention of transplant rejection and might only be considered as an adjunct to the well-established cyclosporin method of preventing transplant rejection (195). However, more recently, the administration of 1,25-(OH)₂D₃ and its analogs in the diet and thus their provision in a continuous manner has shown that it can be much more effective than cyclosporin itself in prevention of transplant rejection (D. A. Hullett, M. T. Cantorna, C. Redaelli, J. Humpal-Winter, H. W. Sollinger, and H. F. DeLuca, unpublished data). It has been used effectively to prolong transplant survival beyond that achieved by cyclosporin in mouse embryonic cardiac transplants and in a vascularized cardiac transplant model in the rat. Of great interest is that the prevention of transplant rejection by the vitamin D compounds was not accompanied by a bone loss phenomenon as is the case with cyclosporin A and, furthermore, was not accompanied by the danger of opportunist infection (M. T. Cantorna, D. A. Hullett, C. Redaelli, C. R. Brandt, J. Humpal-Winter, H. W. Sollinger, and H. F. DeLuca, unpublished data). However, the exact role of how the vitamin D hormone affects T-lymphocyte cell populations, cytokine secretion, and production is not entirely known. There has been a suppressive element reported for the IL-2 gene for the VDR. Whether this is a significant mechanism or not remains to be determined.

E. Islet Cells of the Pancreas

The intriguing finding of a nuclear localization of 1,25-(OH)₂D₃ in the islet cells of the pancreas has led to a speculation that vitamin D might be involved in glucose regulation of insulin secretion (324 325). The presence of a VDR in these cells by now is well accepted, but it is unclear as to what, if any, role vitamin D plays in the functioning of the islet cells. Initial results reveal that vitamin D-deficient rats were unable to respond to a glucose challenge by secreting appropriate amounts of insulin (65). Although this could be corrected by the administration of 1,25-(OH)₂D₃ , other studies suggested that this was mediated by the action of vitamin D in raising plasma calcium concentration (66). The problem, therefore, is that it is unclear whether the regulation of insulin secretion may have a vitamin D component because of the direct actions of calcium concentration on the functioning of the islet cells. Nevertheless, the idea that inappropriate responses of the islet cells to a glucose challenge suggests that a role of the vitamin D hormone is involved. These findings must also be coupled to the experiments carried out in the nonobese diabetic mouse in which the onset of diabetes is delayed or retarded by the administration of 1,25-(OH)₂D₃ (218). The relationship between vitamin D and diabetes is certainly worthy of additional investigation.

F. Role of Vitamin D and Reproduction

Because of the focus of scientists on the role of vitamin D in calcium homeostasis and regulation of phosphate metabolism, the possibility that vitamin D might be involved in reproduction has been largely ignored in the past. However, during the course of producing vitamin D-deficient embryos came the observation that female reproduction is markedly and significantly diminished in vitamin D deficiency (122). Thus a reduction in fertility of 80% was found and could not be corrected by correcting the hypocalcemia (191). This defect, therefore, is quite clearly one related to an absence of the vitamin D molecule. Similar female reproductive failure was noted in the transgenic receptor knockout mice described by Yoshizawa et al. (365). Why there is a failure in the female reproductive system is not known except that it is clear that ovarian cells contain VDR (88). Furthermore, ovarian cells in vivo also accumulate 1,25-(OH)₂D₃ (324 325). Other organs involved in female reproduction, i.e., the hypothalamus, also contain VDR. It has been suggested that the ovary therefore is a target of vitamin D action. In experiments carried out in the authors' laboratory, the VDR levels of the ovary are increased during estrus and diminished during proestrus (348). These results suggest that female rats may not ovulate properly. The infertility brought about by vitamin D deficiency in the female rat can be easily corrected by the administration of 1,25-(OH)₂D₃ (191). These results suggest a role for the vitamin D hormone and its receptor in ovarian function; however, the mechanism remains to be found.

In the case of male reproduction, vitamin D deficiency also reduces the effectiveness of the male (190). However, this diminished male fertility can be corrected by merely providing additional calcium, raising plasma calcium concentration which in turn restores fertility (190). Thus there does not appear to be a role for the vitamin D hormone in spermatogenesis and male reproduction.

G. Does Vitamin D Play an Essential Role During Embryonic Development?

This question intrigued investigators in this field a couple of decades ago. The interest in this possibility was intensified when Suda and colleagues (1 334) demonstrated for the first time the role of 1,25-(OH)₂D₃ in promyelocyte differentiation. This was followed by work which illustrated that cancer cell lines also responded by differentiating in response to 1,25-(OH)₂D₃ (91). This then provided the basis for believing that the vitamin D hormone is also a developmental hormone. However, vitamin D-deficient rats are able to reproduce albeit at only 20% of normal, but the embryos develop approximately normally to parturition (121). Furthermore, the provision of sufficient nutrients to the pups allows them to develop approximately normally until weaning (37). It is at this time that the dependency on vitamin D becomes very obvious. These results strongly suggest that vitamin D does not play an essential role in differentiation and development during embryogenesis. It does not exclude, however, that the vitamin D hormone plays an important role in terminal differentiation of specific cell types, as for example the differentiation and formation of osteoclasts. It does argue, however, that 1,25-(OH)₂D₃ is not involved in the development of major organ systems. These results are supported very strongly by the receptor knockout mice and by the vitamin D-dependency rickets type II cases which are the human example of receptor knockouts. In these cases, the fetuses develop normally and, in fact, appropriately fed, the animals will survive to day 55 (365). If in the absence of receptor the major organ systems develop normally, it provides little support for the idea that vitamin D is an important developmental vitamin. One might argue that there are nongenomic actions of 1,25-(OH)₂D₃; however, there is insufficient evidence for this concept at this time as well (195). Therefore, it is the view of these authors that 1,25-(OH)₂D₃ is not a major developmental hormone at least involved in the development of major organ systems during embryogenesis.

H. Summary

Scientists are just beginning to investigate the new and previously unappreciated functions of the vitamin D hormone. They include functions in the parathyroid glands, the keratinocytes of skin, the T cells and macrophages of the immune system, the islet cells of the pancreas, and ovarian cells of the female. There is the possibility that additional functions will be found in such systems as the islet cells of the pancreas, mammary cell epithelium, and cells in the hair follicle system.

V. VITAMIN D ANALOGS

A. Development of New Analogs of 1α,25-Dihydroxyvitamin D₃

The discovery of 1,25-(OH)₂D₃ and the other metabolites of vitamin D in the early 1970s immediately led to their chemical synthesis and to the first generation of compounds suitable for hormone replacement therapy (13 294). In addition to 1,25-(OH)₂D₃ , most notable among these were the prodrugs 1α-OH-D₃ and 1α-OH-D₂ , which were designed to overcome the need for a functional kidney containing the 1α-hydroxylase enzyme, instead, depending only on the need for 25-hydroxylation (193 258). Synthetic 1,25-(OH)₂D₃ and 1α-OH-D₃ proved to be valuable calcemic agents capable of the correction of hypocalcemia and bone abnormalities (rickets and osteomalacia) resulting from a variety of causes from simple vitamin D deficiency to chronic renal failure. There were early indications that patients with osteoporosis might benefit from the use of an active vitamin D preparation. As a result, it is not surprising that clinical trials of 1α-OH-D₃ (129 254), 1α-OH-D₂ (106), and 1,25-(OH)₂D₃ (7 9 108 257 343 351) have been undertaken. Modest gains in bone mineral density and reductions in fracture rates were reported in many of these studies, and this subject has been reviewed recently by Cali and Russell (44) and Seeman et al. (293). Consequently, 1,25-(OH)₂D₃ and 1α-OH-D₃ are currently used to treat osteoporosis, particularly in Japan and Europe, where dietary calcium intakes are often lower and hypercalcemic side effects not so prevalent.

Although vitamin D analogs continued to be synthesized in the late 1970s, the emergence of the newer nonclassical functions of 1,25-(OH)₂D₃ , particularly the discovery of its role in controlling cell proliferation and differentiation (228 252) (see sect. iv), led to an acceleration in the pace of the search for new analogs. The chemical synthesis of new analogs of 1,25-(OH)₂D₃ became a major priority with the hope that the calcemic properties of 1,25-(OH)₂D₃ might be separated from the antiproliferative cell-differentiating properties (119 157). Hundreds of molecules have now been synthesized in pursuit of this goal, and a couple of recent reviews have attempted to classify or list all published vitamin D analogs in a detailed fashion (20 31 47 104). For the sake of brevity, we have selected some of the more promising of these that are under development by the pharmaceutical industry in Table 1.

Those selected are not the only analogs being developed, but all are at some advanced stage of animal or clinical testing. From perusal of the compounds in Table 1, the reader will immediately notice that the side chain appears to be a popular target for modification, and there is a rational basis for this, namely, that the VDR is relatively tolerant of changes in this part of the molecule. However, since 1980, the synthetic chemist has experimented with modification of every part of the vitamin D molecule in an attempt to accentuate either the antiproliferative or calcemic properties. The full extent of these synthetic modifications is best displayed as the annotated structure illustrated in Figure 10. Probably the most successful 1,25-(OH)₂D₃ analog developed thus far is the Leo antipsoriatic drug calcipotriol, also known as MC903, which was the first vitamin D analog to reach the market for any condition other than hypocalcemia or bone disease (45). When given orally, calcipotriol is ineffective due to the fact that it is rapidly broken down (24 354). When given topically as an ointment, calcipotriol survives long enough to cause improvement in 70% of psoriasis patients (18 181). Both 1,25-(OH)₂D₃ and calcipotriol are believed to be effective in psoriasis because they block hyperproliferation of keratinocytes, increase differentiation of keratinocytes, and help suppress local inflammatory factors through their immunomodulatory properties. Like several other analogs [1,24S-(OH)₂D₂ and 1,24-(OH)₂D₃] currently being pursued, calcipotriol has no 25-hydroxyl group but instead possesses a 24-hydroxyl group which appears to act as a surrogate 25-hydroxyl function. Other analogs being tested, such as 22-oxacalcitriol (OCT) (232), 1,25-(OH)₂-16-ene-23-yne-D₃ (11), and EB1089 (23 161), all contain modifications at C-22 and C-23 which affect the metabolic stability of the molecules without adversely altering the VDR binding. 22-Oxacalcitriol is in development for use in secondary hyperparathyroidism, while EB1089 is in clinical trial for treatment of breast cancer. Another successful modification has been the addition of carbons to the backbone or terminus of the side chain (termed homologation), examples being EB1089 and KH1060 (124). The latter analog epitomizes the latest generation of vitamin D analogs, in that it contains a combination of modifications; in the case of KH1060, four modifications: an extra C at C-24a; additional terminal carbons at C-26a and C27a; a 22-oxa group; and epimerization at the C-17 to C-20 bonds giving a side chain with the opposite orientation. The synthesis of these 20-epi-analogs (46 320), as they are termed, represents a successful experiment in vitamin D design because it led to a family of compounds, many with improved VDR binding compared with the natural side chain. Very recently, several groups (188 350) have used this experience to go one step further and synthesize so-called “butterfly” or “Gemini” analogs with both 20-epi (20S) and normal (20R) side chains. Although little is known at this time about the biological success of this experiment, it is certain to lead to other more exotic double-side chain molecules with symmetrical and asymmetrical structures around C-20 (Fig. 11). In the asymmetrical double-side chain analog (Fig. 11) featuring one 23-oxa side chain and one normal side chain, X-ray structural analysis (188) has revealed the orientation of the two side chains and a measure of the range of conformational structures that can be accommodated by the ligand binding pocket of the VDR, since it is known to bind both 20S- and 20R-analogs.

Equally interesting and perhaps more practical from a biological activity perspective are analogs with modifications of the cis-triene, either changing the conformation of the double bonds to create molecules with an altered cis-triene (250) or analogs that lack the C-10 to C-19 methylene group altogether such as 19-nor-1,25-(OH)₂D₃ or 19-nor-1,25-(OH)₂D₂ (260). In a sense, these latter compounds represent an extension of the experiment begun over 60 years ago (97) when German chemists successfully stabilized the labile cis-triene of one of the photoirradiation products of vitamin D and created the world's first vitamin D analog, dihydrotachysterol (DHT). This analog DHT has an A-ring rotated 180° and has lost the double bond between C-10 and C-19 but retains the C-19 as a methyl group not a methylene, as is the case in almost all other vitamin D compounds. Dihydrotachysterol is metabolized to 25-hydroxylated and 1,25-dihydroxylated forms, both of which bind to VDR and are biologically active, explaining why DHT was once the analog of choice for hypocalcemia, particularly that associated with chronic renal failure (158 268). The experiments with 19-nor-1,25-(OH)₂D₂ and DHT show that the 10–19 double bond and the C-19 carbon are probably superfluous for vitamin D-dependent gene expression. As with several other analogs, 19-nor-1,25-(OH)₂D₂ appears effective in the treatment of hyperparathyroidism associated with chronic renal failure (307).

Further excursions into A-ring or C- and D-ring modifications are currently underway. One molecule into clinical trial for the treatment of osteoporosis is the Chugai analog ED-71, which is modified at C-2 of the A ring by the addition of a hydroxypropoxy group (241), a feature which appears to extend its lifetime in vivo. An extensive collaboration between the De Clerq and Bouillon laboratories has generated many new analogs with modification in the C and D rings, although at present many of these are only partially tested biologically (29 78). One interesting spin-off from the latter research is the synthesis of novel E-ring vitamin D analogs in which the C-8–14–13–17 backbone is retained with the loss of the C and D rings of the vitamin D nucleus but involving the creation of a new stabilizing E ring utilizing C-13, C-12, C-21, C-20, and C-17. The resulting 1,25-dihydroxylated E-ring analog KS176 has 10-fold lower VDR binding and 5-fold lower DBP binding than 1,25-(OH)₂D₃ but is virtually inactive in calcemic assays in vivo (29). It will be interesting to see why E-ring analogs show good in vitro activity but appear to be devoid of in vivo activity.

B. Factors That Alter the Action of Vitamin D Analogs

From the synthesis of literally hundreds of analogs has emerged a clearer understanding of the factors that are important in changing the action of the specific analog relative to 1,25-(OH)₂D₃ . These are listed in the following sections.

1. Activating enzymes

The natural hormone of vitamin D is activated by sequential steps in the liver and kidney as described in section i. The distribution and pharmacokinetics of 1,25-(OH)₂D₃ depend on these steps occurring efficiently at substrate concentrations that circulate bound to DBP and/or are available in the respective tissues. The strategy with most synthetic analogs is to synthesize a molecule with both 25- and 1α-hydroxyl groups and miss out the need for these activation steps. The disadvantage of this approach is that the body loses its ability to regulate the amount of the active principle in the blood and target cell, with the success of the therapy being heavily dependent on the pharmacokinetics of the specific analog.

On the other hand, the use of prodrugs such as 1α-OH-D₃ circumvents some of these problems by introducing a step of activation back into the process, albeit by a different regulatory enzyme. This activating enzyme is in essence a “slow-release” mechanism for delivering the same “active form,” 1,25-(OH)₂D₃ , with altered pharmacokinetics. This is because the activating enzyme has been changed from the tightly regulated renal 1α-hydroxylase to the loosely regulated liver 25-hydroxylase. Although this was believed to be a simple strategy to get around dependence on the easily damaged renal enzyme when it was conceived in the 1970s, the emergence of CYP27 as one of the enzymes responsible for the 25-hydroxylation of vitamin D₃ and 1α-OH-D₃ has complicated this story. CYP27 assumes an even greater importance for the activation of 1α-OH-D₃ because several studies have shown that this cytochrome P-450 prefers 1α-OH-D₃ as a substrate over vitamin D₃ itself (8 119). CYP27 is found in extrahepatic sites including kidney, bone, and endothelial lining cells. In theory, this makes the activation of 1α-OH-D₃ possible in many sites of the body in addition to the liver and opens up the possibility that 1,25-(OH)₂D₃ may be formed from 1α-OH-D₃ in target cells like the osteoblast (147) to act locally on vitamin D-dependent processes. This theory remains to be proven in vivo.

Another aspect of the activation strategy that was overlooked at first glance was the potential of producing multiple active forms from the same administered prodrug. The analog 1α-OH-D₂ has consistently proven to be less toxic than its D₃ counterpart in animal studies and clinical trials (93 306). The explanation for these findings remains obscure but may be due to the fact that 1α-OH-D₂ gives rise to two biologically active products, 1,25-(OH)₂D₂ and 1,24-(OH)₂D₂ (320), each with its own set of slightly different pharmacokinetic parameters compared with 1,25-(OH)₂D₃ (175). CYP27 is capable of the efficient synthesis of 1,24-(OH)₂D₂ (119). The enzyme responsible for the synthesis of 1,25-(OH)₂D₂ remains unknown. Whether this strategy of delivering a prodrug which because of its structure is activated by different enzymes with altered distribution or gives rise to more than one active metabolite offers therapeutic advantages remains moot at this time.

2. Binding to DBP

It has been recognized since the discovery of the various vitamin D metabolites that the affinity of an analog for DBP is inversely correlated with its rate of clearance from the bloodstream. Those metabolites with the strongest affinity for DBP [e.g., 25-OH-D₃ , 24,25-(OH)₂D₃ , and 25-OH-D₃-26,23-lactone] possess the longest half-life values in the blood on the order of days (17 26). Those metabolites such as 1,25-(OH)₂D₃ have lower affinity and half-life values on the order of hours. These basic concepts were confirmed by our experience with analogs of 1,25-(OH)₂D₃ .

Clearly, some modifications are detrimental to DBP binding such as the addition of a 1α-hydroxyl group. Other modifications such as the addition of the double bond between C-22 and C-23 or the addition of a C-24 methyl group, as in the metabolites of vitamin D₂ , do little to affect binding to mammalian DBP but dramatically lower the binding of these compounds to avian DBP (17). However, the findings with D₂ compounds with mammalian DBP go against the general rule, and most 1α-hydroxylated, side chain-modified analogs have inferior DBP binding compared with 1,25-(OH)₂D₃ . Calcipotriol, OCT, EB1089, KH1060, and 20-epi-1,25-(OH)₂D₃ all bind DBP with an affinity at least one order of magnitude less than 1,25-(OH)₂D₃ (89 171). One analog, the A ring-substituted, 1α-hydroxylated compound ED-71, binds DBP more strongly than the natural hormone, a fact which dramatically increases its survival time in the blood and hence its usefulness (241).

The consequences of altered DBP binding are altered tissue distribution of analogs and changes in the rates of their metabolic clearance and cell uptake. Reduction in the affinity of DBP binding can lead to transport on alternate carriers, either albumin or lipoproteins. It has been claimed that OCT derives part of its usefulness for treatment of hyperparathyroidism from its accumulation in parathyroid tissue in vivo, possibly caused by an altered plasma transport on lipoprotein (176).

Second, the rate of metabolic clearance is affected by altering DBP affinity. Consistent with the findings for vitamin D metabolites, a reduction in DBP binding for a given vitamin D analog leads to even more rapid excretion than that observed for 1,25-(OH)₂D₃ . Calcipotriol binds DBP with an affinity two to three orders of magnitude lower than 1,25-(OH)₂D₃ (171). When given systemically, calcipotriol is metabolized and excreted so rapidly that it has no biological effects even at 50 times the dose used for the natural hormone. As described above, it is only effective because it is topically administered in an ointment and is rapidly cleared (and metabolized) before it can reach tissues involved in calcium homeostasis. Thus DBP binding is an important parameter in dictating in vivo biological activity.

Third, the affinity of the analog for DBP influences the rate of cell uptake. Given that it is the “free” fraction of the plasma-borne ligand which is believed to gain access to the target cell and function by binding to the receptor (VDR) in the classical steroid hormone model (see sect. iii), it is therefore no surprise that this might also be the case for vitamin D analogs. Many analogs have the same affinity for VDR as 1,25-(OH)₂D₃ , which when combined with a lower affinity for DBP than 1,25-(OH)₂D₃ creates an even greater downhill gradient into target cells. This has been clearly demonstrated in cultured cells in vitro where the composition of the extracellular fluid can be easily modified. Whereas DBP in fetal calf serum slightly retards entry of 1,25-(OH)₂D₃ , it has no discernible effect on the entry of 20-epi-1,25-(OH)₂D₃ , a notoriously poor binder to DBP (86). There is no doubt that poor DBP binding contributes positively to the promising profile of many vitamin D analogs in biological assays in vitro.

On the other hand, when this effect of DBP, greater access to target cells, is considered along side its other effect, more rapid clearance at the liver, in the animal in vivo, the net effect is usually not so advantageous. This is presumably because these two effects are counteractive and the vitamin D analog cannot be active unless it reaches the target cell in sufficient concentrations to gain greater access and produce its target cell biological effects. In the case of analogs such as ED-71, chemists have created a longer lasting analog that may give reduced biological activity but for a longer period of time. In essence, they have altered the pharmacokinetics again.

3. Binding to VDR-RXR-VDRE complexes

Section iii of this review has established that 1,25-(OH)₂D₃ is able to work through a VDR-mediated genomic mechanism to stimulate transcriptional activity at vitamin D-dependent genes. Whether 1,25-(OH)₂D₃ works through other nongenomic mechanisms to produce physiologically relevant effects is a theory that, in our opinion, remains unproven (244).

Much evidence exists to support the viewpoint that vitamin D analogs operate by mimicking 1,25-(OH)₂D₃ and using a genomic mechanism. The first clue that vitamin D analogs can work through a VDR-mediated transcriptional mechanism came 15 years ago from the bone resorption studies reported by Stern (318). Stern (318) showed that there exists a strong correlation between chick intestinal VDR binding of an analog and its potency in a ⁴⁵Ca rat bone resorption assay. This suggests that a vitamin D analog is only as good as its affinity for the VDR. Certainly, there are some apparent exceptions to this general rule, but these are not discrepancies that cannot be explained by considerations such as DBP binding or pharmacokinetics. Preliminary results with the analogs KH1060, EB1089, and 20-epi-1,25-(OH)₂D₃ (25) suggested that they might be active in immunoregulatory roles at concentrations orders of magnitude below their affinities for the VDR (e.g., at as low as 10⁻¹⁵ M for KH1060 whereas it binds VDR at 10⁻¹¹ M). More recent results suggest that KH1060 (55) is consistently only one or two orders of magnitude more potent than 1,25-(OH)₂D₃ in gene transactivation assays, a difference that could be explained by using the transcriptional model of analog action in conjunction with pharmacokinetic considerations rather than discarding the genomic hypothesis all together.

With the finding that the liganded VDR probably functions transcriptionally as a vitamin D-VDR-RXR heterodimer (128 322) but can form VDR-VDR homodimers (53 54) has come the question as to whether vitamin D analogs might differ from 1,25-(OH)₂D₃ and favor association with VDR-VDR homodimers. Although it remains possible that vitamin D analogs bound to VDR-VDR homodimers might preferentially act at certain VDRE in selected vitamin D-dependent genes (e.g., those associated with immunoregulatory or cell regulatory responses), currently the data do not support this (128 322). It should be noted that in a recent review (53) it was theorized that 1,25-(OH)₂D₃ (and thus vitamin D analogs) might be able to act through as many as 14 different heterodimeric response element combinations to modulate gene expression, but the physiological relevance of the majority of these has yet to be established.

To add to the complexity of the target cell action of 1,25-(OH)₂D₃ , and thus that of vitamin D analogs, is the type and context of the VDRE involved (40). The work of Williams and co-workers (40 166 360) shows that compared with 1,25-(OH)₂D₃ , KH1060 and EB1089 show different patterns of gene activation in bone marrow, osteoblastic cells (ROS17/2, ROS25/1, and UMR106), and intestine (HT-29 and Caco-2) which appear to be gene and cell specific. These data are probably explained by the changes in available transcription factors acting at VDRE in different tissues and at different stages of cell differentiation (317). Furthermore, there have been reports of a nonclassical VDRE in the promoter of the c-fos gene (48) that could theoretically respond differently to “noncalcemic” and “calcemic” analogs, but this has not been demonstrated experimentally. Furthermore, Morrison and Eisman (229) showed that a noncalcemic analog such as calcipotriol is capable of transactivating a calcemic VDRE such as the human osteocalcin promoter VDRE placed in front of the chloramphenicol acetyltransferase reporter gene and stably transfected into ROS17/2 cells. One interpretation of this experiment is that a noncalcemic analog with good VDR affinity is just as calcemic as 1,25-(OH)₂D₃ if it can be delivered to the target cell.

A more recent hypothesis put forward to explain the unique properties of certain vitamin D analogs is that they generate different conformations of the ligand-VDR-RXR heterodimer which result in altered binding at the VDRE. Such subtle changes in protein conformation may still be detected by indirect biochemical means such as altered sensitivity to protease digestion. Recent studies have revealed differences between 1,25-(OH)₂D₃-VDR-RXR and 20-epi-1,25-(OH)₂D₃-VDR-RXR or KH1060-VDR-RXR complexes to trypsin digestion, suggesting subtle changes in the conformation of the ligand binding domain after ligand binding (259 352). The increased resistance to trypsin digestion provided to VDR by KH1060 binding is mimicked by the most active metabolites of KH1060, suggesting that certain receptor conformations might correlate with increased biological activity (353). The use of truncated VDR mutants has allowed the work on KH1060 to be extended and to potentially reveal the amino acid residues involved in these conformational differences and in ligand binding (200). In the future, it is to be expected that the generation of large quantities of the ligand-binding domain of the VDR using bacterial or baculovirus-generated material (323) will permit the determination of the three-dimensional structure with and without ligand [1,25-(OH)₂D₃ or vitamin D analog] by NMR or X-ray crystallographic analysis. For now, we can draw on the experience gained from related fields. If the findings from X-ray crystallographic studies of various retinoid receptors (apo-RXR and holo-RAR ) are relevant to VDR, then the orientation of the COOH terminus containing the AF-2 domain will turn out to be dramatically altered upon vitamin D analog binding and to be very important to the interaction with the other proteins of the transcription initiation complex (279). Moras and co-workers (361) have already used computer modeling based on his RAR model to predict the orientation of ligands including vitamin D analogs in the binding pocket of the VDR. He predicts that 1,25-(OH)₂D₃ and various analogs bind with side chain deep into the ligand-binding pocket and A ring close to the putative “trap-door” (AF-2 domain). Therefore, the observed differences in protease resistance induced by potent 20-epi compounds such as KH1060 (259 352) could be important clues that the AF-2 domain of the heterodimer-analog complex is in a slightly different conformation to that of the heterodimer-1,25-(OH)₂D₃ complex. As a result, these complexes may recruit additional transcription factors (see sect. iii) that may result in qualitative or quantitative differences in gene expression.

There is some evidence from surface plasmon resonance studies that a series of Roche analogs causes differences in the association and dissociation rates of VDR-RXR heterodimers to artificial VDRE-containing oligonucleotide anchored to gold-coated chips (67). Whether this translates into increased stability of specific complexes on the vitamin D-dependent gene promoter in vivo remains to be proven. In summary, researchers continue to probe the many steps of the gene transcription activation process in the hope of finding a fundamental difference between 1,25-(OH)₂D₃ and its analogs that may explain their improved potency or their action at only selected genes. Thus far, they have more questions than answers.

4. Target cell and liver catabolic enzymes

As stated in section i, much evidence has accumulated to support the hypothesis that 1,25-(OH)₂D₃ is subject to target-cell catabolism and side chain cleavage via a 24-oxidation pathway to calcitroic acid. CYP24 carries out multiple steps in the C-24 oxidation process, is vitamin D inducible, and is present in many (if not all) vitamin D target cells. We have postulated that the purpose of this catabolic pathway is to desensitize the target cell to continuing hormonal stimulation by 1,25-(OH)₂D₃ (203). One can also ask the question: Are vitamin D analogs subject to the same catabolic metabolism? And, if not, are other drug-catabolizing systems present within vitamin D target cells to inactivate the vitamin D analog?

We now have a good idea as to the answer to these questions, but the answers are analog specific. Certainly, there are metabolism-sensitive analogs such as calcipotriol and OCT that are metabolized very rapidly by target cell enzymes to clearly defined and unique metabolites (213 215) that resemble products of the 24-oxidation pathway for 1,25-(OH)₂D₃ . The provision of such features as a 22-oxa group seems to accelerate the rate of metabolism of OCT, and the 24-hydroxyl group of calcipotriol is a site for deactivation. The rapid metabolism of calcipotriol by cultured keratinocytes in vitro (215) correlates well with the in vivo findings that it is rapidly broken down when administered topically and fails to reach tissues involved in calcium homeostasis (24). One does not need to invoke hepatic metabolism as the principal site of inactivation, even though this tissue can also break down calcipotriol (24 215).

Analogs that appear to be slowly metabolized by target cells include EB1089 and 1,24S-(OH)₂D₂ (156 170 295). Both are considered more calcemic than either calcipotriol or OCT but not excessively so. These data are consistent with metabolism-resistant analogs having a longer half-life inside the target cell as compared with 1,25-(OH)₂D₃ and a longer half-life in pharmacokinetic studies (30 175), and possibly a higher biological activity in vivo, although this is tempered by inferior DBP binding and more rapid clearance from the bloodstream. Interestingly, both EB1089 and 1,24S-(OH)₂D₂ are blocked at the C-24 position and cannot enter the C-24 oxidation pathway, leaving C-26 as the new site of hydroxylation, albeit at a slower rate (156 170 295). Another potent analog blocked in the C-24 position is 1,25-(OH)₂-16-ene-23-yne-D₃ , which is also subject to 26-hydroxylation (288). Studies with EB1089 using transfected human CYP27 suggest that this cytochrome is not responsible for the 26-hydroxylation observed (295). However, Suda and co-workers (227) have concluded that the other candidate cytochrome P-450, CYP24 may be responsible for 26-hydroxylation of the 24-blocked analog 24,24-F₂-1,25-(OH)₂D₃ inside kidney cells.

Another metabolic complication brought on by the use of vitamin D analogs is the potential for creating “active” rather than “inactive” products as a result of target cell enzymes. Three examples of this have now been documented, and this may be more common than we think. The potent calcemic analog 26,27-F₆-1,25-(OH)₂D₃ is converted both in vivo and in vitro, presumably by the cytochrome CYP24 into 26,27-F₆-1,23,25-(OH)₃D₃ . Sasaki et al. (287) have shown that this metabolite which accumulates in bone in vivo is fivefold more active than 1,25-(OH)₂D₃ in a reporter gene expression system. Thus a metabolite may explain the long-lasting effects of the parent analog. In a second example, Dilworth et al. (87) recently showed that KH1060 is metabolized very rapidly in vitro and in vivo into as many as 22 metabolites but that some of these metabolites retain significant biological activity in similar reporter gene assay systems. The most abundant of these metabolites include 26-hydroxy and 24α-hydroxy KH1060, which can be detected in the blood in vivo, are stable in vitro, and as stated earlier can form trypsin-resistant complexes with VDR (87 353). At the very least, these active metabolites of KH1060 may explain its high potency when compared with 1,25-(OH)₂D₃ using in vitro assays. In addition, the findings may also explain the disappointing performance of KH1060 in vivo, since the metabolites are not allowed to accumulate and are rapidly cleared from the bloodstream. The third example of possible activation of an analog is 3-epimerization (275). Using in vitro cultured cell and perfused kidney systems where artificial media and high analog concentrations are employed, several researchers have observed the 3-epimerization of vitamin D analogs [e.g., 1,25-(OH)₂D₃ itself] (214 224 275). As with the example of KH1060, the claim is not that the metabolic product 3-epi-1,25-(OH)₂D₃ is more active than the parent compound, just that it retains significant biological activity and is more stable, not being subject to the C-24 oxidation pathway which is the fate that befalls the hormone. The nature of the enzyme involved is currently undefined but is reminiscent of the behavior of steroidal 3α- and 3β-oxidoreductases. Moreover, the formation of 1α- and 1β-hydroxylated DHT in rats and humans in vivo, referred to in section v A, resembles the 3-epimerization of vitamin D analogs and suggests that the process may be more than just an in vitro artifact. Nevertheless, the main hurdle for proponents of the 3-epimerization hypothesis is that they must show the metabolic step can occur in vivo.

The final metabolic consideration is hepatic modification. The liver is believed to be devoid of VDR and the inducible cytochrome P-450; CYP24 and cultured liver cells (Hep G2 and Hep 3B) do not appear to degrade the hormone 1,25-(OH)₂D₃ very rapidly (213 215). However, the story with vitamin D analogs appears to be quite different. Analogs such as calcipotriol and OCT are broken down by liver enzymes, these presumably contributing to the rapid metabolic clearance observed in vivo (213 215). As the analog structure is modified away from the vitamin D prototype, it becomes vulnerable to other more general catabolic enzymes found in the liver. One interesting example is the enzyme that saturates the double bond between C-22 and C-23 in the side chain of calcipotriol but does nothing to the same bond in the side chain of vitamin D₂ (215). The simple replacement of the 24-methyl of the D₂ side chain with a 24-hydroxyl group in calcipotriol is enough to allow the saturase enzyme to function. As vitamin D analogs continue to deviate more from the classical structure, they become more and more susceptible to general purpose or unrelated enzyme systems.

It should be noted with regard to molecular mechanisms of action at the target cell level that metabolism is often disregarded or given too little emphasis. Metabolic assumptions are made when testing biological activity in vitro that are not always valid. These include 1) the analog is biologically active as administered and 2) the analog has the same stability as 1,25-(OH)₂D₃ in the in vitro target cell model used, whether in in vitro organ culture, cultured target cell, or host cell/reporter gene construct. The validity of this approach is made even more tenuous when data acquired with different in vitro models, where metabolic considerations may or may not apply, are compared with data acquired in vivo, where metabolic considerations do apply. The reader is cautioned that invalid comparisons of in vivo and in vitro data abound in this field. The outcome is that a number of promising in vitro-tested vitamin D analogs have been found to be inactive when tested in vivo.

5. Relative importance of key factors on vitamin D analog action

It is evident that vitamin D analogs differ from 1,25-(OH)₂D₃ in a variety of ways: activating enzymes, DBP affinity, VDR-RXR-VDRE-gene association, and catabolic enzyme susceptibility. Any one or a combination of these differences could be important in making a specific analog appear calcemic or noncalcemic or work in vitro but not in vivo, or vice versa. Although it is attractive at this time to consider as most important the subtle VDR conformational differences induced by analogs as compared with 1,25-(OH)₂D₃ , there is a wealth of information that argues that metabolism, transport, and pharmacokinetics play a major role in vivo.

With our existing in vitro cell systems, it is difficult to study any one of these factors to the total exclusion of all others. Thus it seems unlikely that we shall be able to determine the relative importance of all of the factors listed until we have a fully functional cell-free in vitro model for transcription so that analogs can be tested without the complication of metabolic enzymes.

C. Future Directions in Vitamin D Drug Design

The growing availability of recombinant cytochromes P-450 will allow for a search for potential inhibitors. Such specific inhibitors of CYP24 may be of value in blocking 1,25-(OH)₂D₃ catabolism. Modeling of the vitamin D binding pocket of VDR, DBP, and the three vitamin D-related cytochrome P-450 will become a major goal now that all these specific proteins have been cloned and overexpressed. Although the full-length proteins are slightly beyond the current limits of NMR or X-ray crystallography, the ligand-binding domains or pockets are not. It is likely that technical problems with these procedures will be overcome shortly and the full-length proteins can be tackled. The membrane-associated region of cytochromes P-450 poses problems, but enormous strides have been made based on models built with soluble prokaryotic isoforms. Although the first priority is to study 1,25-(OH)₂D₃ binding to the VDR, the binding of a variety of analogs to all these proteins is likely to follow. In the process, we will be able to explore the permitted boundaries for optimal protein binding. More rational vitamin D analog design will follow to take advantage of structural idiosyncrasies of all of these key proteins. Meanwhile, the not-so-rational synthesis of new analogs is likely to continue.

FOOTNOTES

• We were supported by National Institutes of Health Program Project Grant DK-14881 (to H. F. DeLuca), by a fund from the Wisconsin Alumni Research Foundation (to H. F. DeLuca), and by a grant from the Medical Research Council of Canada (to G. Jones).

• ¹The term 27-hydroxylation has been suggested by a consortium of cytochrome P-450 specialists to describe terminal hydroxylation of steroids given that methyl groups at C-26 and C-27 are indistinguishable. The old nomenclature for this was 26-hydroxylation, and the literature contains numerous references to 26-hydroxylation of vitamin D compounds giving rise to 26-hydroxylated vitamin D metabolites, e.g., 25,26-(OH)₂D₃ .