Products of Vitamin D3 or 7-Dehydrocholesterol Metabolism by Cytochrome P450scc Show Anti-Leukemia Effects, Having Low or Absent Calcemic Activity

Background Cytochrome P450scc metabolizes vitamin D3 to 20-hydroxyvitamin D3 (20(OH)D3) and 20,23(OH)2D3, as well as 1-hydroxyvitamin D3 to 1α,20-dihydroxyvitamin D3 (1,20(OH)2D3). It also cleaves the side chain of 7-dehydrocholesterol producing 7-dehydropregnenolone (7DHP), which can be transformed to 20(OH)7DHP. UVB induces transformation of the steroidal 5,7-dienes to pregnacalciferol (pD) and a lumisterol-like compounds (pL). Methods and Findings To define the biological significance of these P450scc-initiated pathways, we tested the effects of their 5,7-diene precursors and secosteroidal products on leukemia cell differentiation and proliferation in comparison to 1α,25-dihydroxyvitamin D3 (1,25(OH)2D3). These secosteroids inhibited proliferation and induced erythroid differentiation of K562 human chronic myeloid and MEL mouse leukemia cells with 20(OH)D3 and 20,23(OH)2D3 being either equipotent or slightly less potent than 1,25(OH)2D3, while 1,20(OH)2D3, pD and pL compounds were slightly or moderately less potent. The compounds also inhibited proliferation and induced monocytic differentiation of HL-60 promyelocytic and U937 promonocytic human leukemia cells. Among them 1,25(OH)2D3 was the most potent, 20(OH)D3, 20,23(OH)2D3 and 1,20(OH)2D3 were less active, and pD and pL compounds were the least potent. Since it had been previously proven that secosteroids without the side chain (pD) have no effect on systemic calcium levels we performed additional testing in rats and found that 20(OH)D3 had no calcemic activity at concentration as high as 1 µg/kg, whereas, 1,20(OH)2D3 was slightly to moderately calcemic and 1,25(OH)2D3 had strong calcemic activity. Conclusions We identified novel secosteroids that are excellent candidates for anti-leukemia therapy with 20(OH)D3 deserving special attention because of its relatively high potency and lack of calcemic activity.

Unfortunately, the toxicity (hypercalcemia) of high levels of vitamin D largely prevents the use of pharmacological doses of 1,25(OH) 2 D3 for either the prevention or treatment of cancer. Therefore, there is a continuing search to find vitamin D analogues (for example, Gemini analogs) that retain antiproliferative activity but that are non-calcemic, acting as partial receptor agonists for the VDR [18][19][20]. In addition, more than 30 years ago Holick et al showed that shortening the side chain of vitamin D or eliminating it (producing 20-hydroxypregnacalciferol (20(OH)pD), attenuates or eliminates the calcemic effect [21].

Novel secosteroids have low calcemic effect
One of us (MFH) has documented that elimination of all but two carbons of the side chain in vitamin D3, leading to production of 20(OH)pD and pD, eliminates the calcemic effect of the secosteroid [21]. Since P450scc hydroxylates the side chain of D3 and 1(OH)D3 at C20, we tested the calcemic effects of the resulting products in rats at a doses 0.1 mg/kg-3 mg/kg of body weight (Fig. 1). 20(OH)D3 at dose as high as 3.0 mg/kg had no calcemic activity (calcium = 10.461.5 vs 9.361.3 mg/dL for control), whereas 1,25(OH) 2 D3 at the same dose had the expected strong calcemic effect raising calcium to 16.061.2 mg/dL. Interestingly, addition of a 1a-hydroxyl group to 20(OH)D3 promoted calcemic activity raising calcium to 13.960.8 mg/dL. Thus for 20(OH)D3, the lack of 1a-hydroxylation underlies its inability to influence serum calcium. 1,20(OH) 2 D3 displays a calcemic effect, however less than that of 1,25(OH) 2 D3. This is consistent with a fundamental role of the hydroxyl group at the 1 alpha-position of vitamin D for the regulation of serum calcium levels [8,21].

Novel secosteroidal products show anti-leukemic activity
Anti-leukemic effects of 20(OH)D3, 20,23(OH) 2 D3, 1,20(OH) 2 D3, pD, 20OHpD and pL in comparison to 1,25(OH) 2 D3 and to precursor steroidal 5,7-dienes (for structures see Figure S1), were tested in K562 human chronic myeloid and MEL mouse erytholeukemia and human HL-60 promyelocytic and U937 promonocytic leukemia cells. The choice of 10 27 M as the concentration of drugs to use in the majority of experiments was based on initial experiments (not shown) as well as on already published data [26,27,31,34], demonstrating that this concentration is optimal for comparative studies.
In human HL-60 promyelocytic and U937 myeloid leukemia all of the secosteroids and steroidal 5.7-dienes showed statistically significant inhibitory effects, however, their relative potency were different for both cell lines (
When the level of differentiation was determined from the relative concentration of hemoglobin as hemin measured spectrophotometrically in an equal number of cells, all compounds tested induced significant cell differentiation. 1,25(OH) 2 D3 was more potent than other compounds with the sequence of potency being: 1,25(OH) 2 Table S2B, C).

Molecular modeling studies
To understand the potential binding mode of 20S(OH)D3 in VDR, we selected the crystal structure of the VDR complex with calcitriol (PDB code: 1DB1) [35]and performed molecular modeling studies using Schrodinger Suite 2009 (Schrodinger Inc., New York, NY). The docking program successfully reproduced the original binding pose of the native ligand with an excellent docking score (213.5, Supplemental Fig. S4). All of the six hydrogen bonding interactions between the ligand and the residues in VDR are reproduced in this validation: 3-OH to Ser274/Tyr143; 1-OH to Ser233/Arg270; and 25-OH to His393/HIP301.
Next we examined the potential binding of 20S(OH)D3 to the VDR. The 20S(OH) isomer (the metabolite generated by P450scc enzymes) overlaps with the native ligand very well with a docking score 211.4. We also performed docking calculations for a variety of D3 metabolites shown in Figure S1. The results are listed in Table S3. Overall, all the metabolites have good docking score, suggesting they may bind well to the VDR. In general, metabolites having 1 alfa-OH have better docking scores than those that do not have. Shorter chain analogs (pD3 analogs) generally have worse docking scores, suggesting that the removal of the side chain may be detrimental to the VDR binding. While we are fully aware of the limitations with molecular modeling, these docking studies suggest that there are high binding affinities of these compounds for the VDR, and the biological activity of these metabolites may still be generated via their interactions with the VDR.
In vivo data shown in Fig. 1 indicate that the calcemic effect of vitamin D3 derivatives can largely be separated from their other properties (inhibition of proliferation and stimulation of differentiation) as clearly demonstrated for 20(OH)D3. Interestingly, removal of the key hydroxyl group at C-1 whose presence is known to play a role in calcium regulation [5,6,20], has only a slight or no influence on the induction of differentiation of monocytes and leukemia cells (20,23 [26,34]. Importantly, gene silencing technology used on human keratinocytes has clearly demonstrated that the phenotypic action of 20(OH)D3 and 20,23(OH) 2 D3 are mediated through activation of VDR [26,27,34]. Therefore, because of this intriguing dissociation between phenotypic effects seen in leukemia cells (present studies) and keratinocytes (parallel studies) we have examined the potential binding of 20(OH)D3 to the VDR ( figure  S4). We have found that the 20S(OH) isomer (the metabolite generated by P450scc [22,23]) overlaps with the native ligand (calcitriol) very well having comparable docking score ( figure S4). This opens a new and exciting area to study on the correlation between structure and activity of P450scc and CYP27A1 and CYP27B1 generated active vitamin D3 hydroxyderivatives Since these compounds also differ in the presence or absence of a hydroxyl group at C-25, we suggest that a C-25 hydroxyl group, in addition to C-1 (see above), is necessary for strong induction of CYP24 gene expression. Future studies on this subject are mandatory, since CYP24 is the major enzyme causing inactivation of vitamin D, and respectively slower rates of inactivation could be predicted for the P450scc-derived secosteroids, which should be of advantage for therapeutic use of these compounds.
The observed activities of the various secosteroids and steroidal 5-7dienes we tested were influenced by the genetic background of the cell lines used, suggesting the existence of cell-type specific factors that can modulate activities. Such factors might include the cocktail of transcription factors present, and the concentration of vitamin D metabolizing enzymes in the cells such as CYP24, CYP27A1 and CYP27B1. Despite this variability between cell types, the non-calcemic 20(OH)D3 and its hydroxyl-derivatives generally exerted effects on cell proliferation and differentiation comparable to those of 1,25(OH) 2 D3. Furthermore we observed functional similarity with the effects induced by 20(OH)D3 and 20,23(OH) 2 D3 in epidermal keratinocytes, e.g., G1/G0 arrest connected with induction of their differentiation program [26,34]. The shortening of the side chain resulted in decreased activity, which is in agreement with lower docking score in comparison to 1,25(OH) 2 D3 (Table S3). Nevertheless, these compounds, pD3 and 20-OH pD3, do induce leukemia cell differentiation, similarly as in melanoma cells [29,31]. This suggests that the mechanism of action of the tested compounds is conserved among different cell types including leukemias (cell of hematopoietic origin), normal skin keratinocytes [26,34] and malignant melanoma cells [29,31]. Acute myeloid leukemias (AML) are high grade malignancies originating from blood cell precursors in bone marrow. Depending on the lineage of leukemic cells they can be divided into myeloid, monocytic, erythroid and megakaryocytic. AML is characterized by accumulation of immature forms (blasts) in bone marrow and subsequently in peripheral blood and are rapidly and universally fatal if not treated. Developed over the last three decades, chemotherapeutic regimens to treat AML markedly improved the rate of remission and survival in AML patients. However, many leukemias show primary resistance to current therapy and many patients suffers from AML relapse following initial remission. Relapsed disease often becomes resistant to currently available  chemotherapeutic drugs and short of bone marrow transplant such relapsed patients become incurable and die. These findings necessitate a continuing search for new drugs or adjuvant therapeutic approaches to treat AML.It is well established that hematopoietic cells and their malignant counterparts (leukemic blasts) express VDR. Moreover, in vitro studies of myeloid leukemia cells show that there is strong response to the active form of vitamin D ranging from induction of differentiation to induction of apoptosis or autophagy, with resulting attenuation of proliferation [10,[12][13][14]36]. Unfortunately in vivo use of these promising drugs is limited by their toxic (calcemic) activity. In this study we report that newly identified and characterized derivatives of vitamin D that have low or no calcemic activity still retain their potent anti-leukemic effects. Furthermore, the above compounds have the potential to be normally generated in skin (from their 5,7dienes precursors) exposed to solar radiation [29][30][31]37]. Poten- tially these derivatives can be synthesized in other tissues in the body expressing P450scc, for example the adrenal gland [22,23,25,30,38]. Interestingly the number of tissues in which expression of P450scc has been identified is growing and includes bone, thymus, skin and brain as examples [30,[39][40][41][42][43][44], raising the possibility of local production of the ligands identified in this study in the bone marrow environment. Therefore, we believe that we have identified a family of non-toxic and physiological compounds that can be used in therapy of acute leukemias [22,23,[28][29][30][31].
In summary, novel secosteroidal products of P450scc metabolism of vitamin D3 and pro-vitamin D3, show excellent antileukemia activity in vitro with 20(OH)D3 deserving special attention because of its relatively high potency and lack of calcemic activity. Moreover, since these compounds potentially can be synthesized in vivo in adrenal gland or generated by skin and theoretically may be produced by many other tissues including bone, further studies are required to define their potential physiological and pathophysiological role in bone marrow homeostasis and leukemiagenesis.

Steroids and secosteroids
Chemical structures of the compounds tested are shown in Figure S1. 1,25(OH) 2 D3 was purchased from (Fluka). 20(OH)D3 and 20,23(OH) 2 D3 were biochemically synthesized as described  previously using an in vitro reconstituted P450scc system with vitamin D3 (Sigma) as substrate [23,30]. 1,20(OH) 2 D3 was biochemically synthesized from 1a-hydroxyvitamin D3 (Sigma) as described previously [28].The products were purified by TLC followed by RP-HPLC and identities were confirmed based on mass and UV spectra, as well as on retention times in comparison to standards previously characterized by NMR [23,28,30]. The compounds were aliquoted, dried and stored at 280uC until use.
7DHP and 20(OH)7DHP were synthesized and purified as described previously [31,32]. After NMR confirmation of their structures, the steroids were purified by HPLC, aliquoted and stored for further use or subjected to photochemical transformation using a Biorad UV Transilluminator 2000 (Biorad, Hercules, CA). Spectral characteristics of the UVB (280-320 nm) source were published previously [45,46] and its strength (4.860.2 mW cm 22 ) was routinely measured with a digital UVB Meter Model 6.0 (Solartech Inc., Harrison Twp, MI). Irradiation was followed by incubation for 14 h at room temperature then selected products were purified by RP-HPLC chromatography as described [31,32]. The major products were identified on the basis of their retention time and characteristic UV absorption. Initial identification was confirmed by means of MS and NMR measurements [31,32]. The resulting secosteroids including pregnacalciferol (pD), pregnalumisterol (pL) and 20-hydroxy-pregnacalciferol (20(OH)pD) were RP-HPLC purified, aliquoted and stored at 280uC [31,32].

Testing on leukemia cell lines
MEL mouse erythroleukemia [47], human K562 chronic myeloid leukemia (purchased from American Tissue Culture Collection, USA), and human HL-60 human promyelocytic [48] and U937 promonocytic leukemia [49] cells were cultured in RPMI 1640 containing 10% fetal bovine serum (FBS) (Atlanta Biologics)) and 1% penicillin/streptomycin/amphotericin antibiotic solution (Sigma) at 37uC in 5% CO 2 . Test compounds were dissolved in ethanol and added to the cultures to reach final concentrations as listed. Ethanol at a concentration of 0.1% was used as a vehicle (negative) control, while 2% DMSO or 32 nM TPA served as positive controls for studies on cell differentiation. U937 and MEL cell lines were generous gift from Dr S. Hanissian (University of Tennessee HSC) and HL-60 from Dr M. Radic (University of Tennessee HSC).
Cell proliferation. Cells at concentrations of 2610 6 cells/ml (K56 and MEL lines) and 10 7 cells/ml (HL-60 and U937 lines) were inoculated into T25 flasks (TPP, Midwest Scientific) containing 10 ml of RPMI supplemented with charcoal treated 10% FBS. The test compounds were added at concentrations of 10 27 M every day with media being changed every 72 hours. After 7 days days the cells were stained in 0.4% trypan blue (Sigma) and the viable cells counted with a hemacytometer.
Cell cycle analysis. Cell cycle analysis was performed by flow cytometry following standard protocols used in our laboratories [26,50]. Briefly, HL-60 cells were cultured in RPMI 1640 plus charcoal treated 10% FBS (HyClone). Test compounds were added to a final concentration 10 26 M in 0.1% ethanol every day with media being changed every third day. On day five the cells were washed in PBS, fixed in 70% cold ethanol and stained with propidium iodine (Sigma). DNA content analysis was performed with a FACS Calibur flow cytometer.
Cell differentiation. To estimate erythroid differentiation (production of hemoglobin), first we evaluated the number of benzidine positive cells after 5 and 7 days in culture. Cells were centrifuged and washed four times with PBS and resuspended in 1 ml of fresh PBS. For hemoglobin determination, a benzidine staining solution was freshly prepared by mixing one part of 30% hydrogen peroxide, one part of base stock solution of 3% benzidine in 90% acetic acid, and 5 parts of water. The solution was diluted 1:10 with the cell suspension and 250 ml aliquots added to 4 wells of a 24-well plate. After 10 min of incubation at room temperature, benzidine-positive cells were counted under the microscope with a minimum of 200 cells scored. Second, to define the relative content of hemoglobin spectrophotometrically [51], equal number of cells (7610 6 ) were washed with cold PBS and lysed for 20 min in 100 ml of lysis buffer (0.2% Triton X-100 in 100 mM potassium phosphate buffer, pH 7.8). The lysates were centrifuged for 15 min at 1500 r.p.m. and 100 ml of the supernatant was incubated with 2 ml of benzidine solution (5 mg/ml in glacial acetic acid) and 2 ml 30% H 2 O 2 for 10 min at room temperature in the dark. Absorption was measured at 600 nm. Data are shown as fold increases in comparison to the level of hemoglobin in vehicle-treated cells.
Differentiation of HL-60 or U937 cells toward monocytes-like morphology and NBT-reduction was assessed after 5 and 7 days. Cells (2610 6 ) were washed with PBS four times and resuspended in 200 ml of NBT solution (4 mg/ml). After the addition of 200 ml of TPA solution (2 mg/ml in PBS) cells were incubated at 37uC for 60 min in 24-well plates. Cell differentiation was assessed by intracellular blue formazan deposits. The NBT positive and negative cells were scored under light microscopy examination (206) with a minimum of 200 cells scored. [52]. For spectrophotometric analysis the cells were washed twice with buffer containing cold bovine serum albumin solution (17 mg/ml BSA, 137 mM NaCl, 5 mM KCl, 0.8 mM MgSO 4 , 10 mM, HEPES, pH 7.4) to remove unreacted NBT, and the insoluble formazan deposits in the resulting pellets were solubilized in 1 ml of 90% DMSO, 0.1% SDS and 0.01 mM NaOH by vigorous vortexing. The samples were centrifuged 5 min at 1,500 x g to remove the cellular debris, and then the absorbance of supernatants measured at 715 nm. Data are expressed as change in A 715 /10 6 cells [53].
CD11b expression. Staining of the CD11b cell differentiation marker was carried out according to the manufacturer's protocol as follows. Cultured cells were resuspended in cold wash buffer (PBS supplemented with 1% FBS and 1% human serum to block non-specific F c -mediated binding) at a concentration of 2610 6 /ml. After centrifugation, the cell pellet was resuspended in 100 ml wash buffer. Aliquots (50 ml) of this were incubated with 50 ml of diluted antibody, either phycoerythrin-conjugated mouse anti-human CD11b antibody or phycoerythrin-conjugated mouse IgG isotype as control (both from BD Pharmingen, San Diego, California). After 30 min the cells were centrifuged at at 400 x g and resuspended in 1 ml of wash buffer. This step was repeated with cells finally being resuspended in 200 ml of wash buffer for flow cytometry [54].
Real-time RT PCR. The RNA from HL-60 cells treated as described above was isolated using the Absolutely RNA Miniprep Kit (Stratagene, La Jolla, CA). Reverse transcription (1 mg RNA/ reaction) was performed using the Transcriptor First Strand cDNA Synthesis Kit (Roche, Mannheim, Germany). Real-time PCR was performed using undiluted cDNA and a TaqMan PCR Master Mix. Reactions (in triplicate) were performed at 50uC for 2 min, 95uC for 10 min and then 50 cycles of 95uC for 15 sec, 60uC for 30 sec and 72uC for 30 sec. The primers and probes were designed with the universal probe library (Roche). Data were collected on a Roche Light Cycler 480. The amount of amplified product for each gene was compared to that for b actin using a comparative C T method. A list of the primers used for RT-PCR DNA amplification is shown in Table S4.

Testing of calcemic effect
The calcemic effects of 20(OH)D3 and 1,20(OH) 2 D3 were compared with 25(OH)D3 and 1,25(OH) 2 D3. Briefly, weanling rats were obtained from Holtzman and fed a vitamin D deficient diet for 3 months before they were divided into 13 groups (6 animals per group) and injected with either vehicle (propylene glycol) or 3 concentrations (0.3 mg/kg, 1.0 mg/kg, 3.0 mg/kg) of 25(OH)D3, 1,25(OH) 2 D3, 20(OH)D3 or 1,20(OH) 2 D3 dissolved in propylene glycol for 7 consecutive days. A day after the final dosing, blood was obtained by heart puncture and serum was prepared. Serum calcium was determined by using a kit (DICA-500 QuantiChrom TM ) from BioAssay Systems (Hayward, CA). Data were analyzed using a student's 1-tailed independent t-test with a p value indicating significance at p,0.05.

Statistical analysis
Data were analyzed with GraphPad Prizm Version 4.0 (GraphPad Software Inc., San Diego, CA, USA) using the t-test or one way ANOVA with appropriate post-hoc tests. Differences were considered significant when p,0.05. The data are presented as means 2/+ SE.  Author Contributions