Without 1α-hydroxylation, the gene expression profile of 25(OH)D3 treatment overlaps deeply with that of 1,25(OH)2D3 in prostate cancer cells

Recently, the antiproliferative action of 1,25(OH)2D3 (1,25D3), an active metabolite of vitamin D3, in the management of prostate cancer has been argued rigorously. In this study, we found that at a physiological concentration, 25(OH)D3 (25D3), the precursor of 1,25D3 and an inactive form of vitamin D because of its much weaker binding activity to the vitamin D receptor (VDR) compared with 1,25D3, had a gene expression profile similar to that of 1,25D3 in prostate cancer LNCaP cells. By immunocytochemistry, western blotting, and CYP27B1 and/or VDR knockdown by small interfering RNAs, we found that 10−7 M 25D3, which is within its uppermost physiological concentration in the bloodstream, induced VDR nuclear import and robustly activated its target genes in the virtual absence of CYP27B1 expression. Comprehensive microarray analyses verified 25D3 bioactivity, and we found that 25D3 target gene profiles largely matched those of 1,25D3, while the presence a small subset of 25D3- or 1,25D3-specific target genes was not excluded. These results indicated that 25D3 shares bioactivity with 1,25D3 without conversion to the latter. Metallothionein 2A was identified as a 1,25D3-specific repressive target gene, which might be a prerequisite for 1,25D3, but not 25D3, to exert its anti-proliferative action in LNCaP cells.


25D3 regulates CYP24A1 gene expression without CYP27B1 activity. Several researchers reported
that only a few prostate cells express abundant amounts of CYP27B1 protein, which can convert 25D3 to 1,25D3, while it was reported that LNCaP cells have no detectable CYP27B1 activity 24,31,34 . We reconfirmed the absence of CYP27B1 activity in LNCaP cells by using an ultrasensitive liquid chromatography-tandem mass spectrometry (LC-MS/MS) assay (see Supplemental Materials and Methods) (ASKA Pharmaceutical Co., Ltd., Tokyo, Japan). LNCaP cells were administered 10 −7 M of 25D3 for 24 h after incubating the cells with serum-free RPMI-1640 medium minus 25D3 for 24 h. Then, the cell pellets obtained from 5.0 × 10 6 cells and the culture medium were subjected to an LC-MS/MS assay for the quantification of vitamin D metabolites. The lower limit of detection of this assay was 1.2 × 10 −11 M. We found that there was essentially no 1,25D3 in either the cell lysate or medium (Supplemental Table S2), indicating that LNCaP cells do not have CYP27B1 activity.
To elucidate whether 25D3 exerted physiological activity similar to that of 1,25D3, we investigated the effects of 1,25D3 and 25D3 on CYP24A1 expression, which is a robust target gene of the 1,25D3-VDR axis, by qRT-PCR 7 . Endogenous CYP24A1 mRNA expression in LNCaP cells transfected with non-targeting siRNA as a negative control (siCT cells) was stimulated by 1,25D3 in a dose-dependent manner (11,377-fold for 10 −9 M, 88,243-fold for 10 −8 M, and 933,762-fold for 10 −7 M) ( Table 1). Of note, CYP24A1 gene expression was also up-regulated by 25D3 in a dose-dependent manner, although the dose of 25D3 required to obtain an effect comparable with that of 1,25D3 was 10-20 times higher, despite being used within its physiological range (see Discussion). Indeed, 10 −7 M of 25D3 exerted a stimulatory effect that was several hundred-fold stronger than at 10 −8 M and 70% weaker than at 10 −7 M of 1,25D3 (Table 1). In the siRNA experiment, we employed CYP27B1-specific siRNA, which further decreased the inherently negligible amount of CYP27B1 mRNA to 30% (Fig. 1A). In the cells transfected with siRNA specific for CYP27B1 (siCYP27B1 cells) to dampen 1,25D3 generation more thoroughly as shown above, a concentration of 25D3 as high as 10 −7 M again showed obvious stimulatory effects on CYP24A1 stimulation (Table 1). These results indicated that 25D3 exerts hormonal activity on CYP24A1 gene regulation even when there is no CYP27B1 activity in LNCaP cells.
25D3 induces nuclear localization of the VDR. To elucidate which nuclear receptor was involved in the activation of CYP24A1 by 25D3, we performed co-knockdown experiments using siRNAs specific to either the androgen receptor (AR) or VDR in addition to siRNA for CYP27B1. We confirmed the efficient knockdown of each nuclear receptor gene by qRT-PCR; AR was reduced to 25% and VDR was reduced to 26% (Fig. 1A). We administered 10 −7 M of 25D3 to each siRNA-transfected LNCaP cell line for 24 h, then examined CYP24A1 expression by qRT-PCR. We found that knockdown of the VDR alone in siCYP27B1 cells, but not the AR, contributed to the disappearance of the 25D3-dependent activation of CYP24A1 expression (Fig. 1B). These results validated our hypothesis that 25D3-dependent CYP24A1 expression is mediated by the VDR. Then, we analysed the effect of 25D3 on the intracellular localization of the VDR in siCYP27B1 cells by immunocytochemistry and western blotting experiments with a VDR-specific antibody (D2K6W). In vehicle-treated siCYP27B cells, intranuclear VDR staining was observed in a small number of cells (Fig. 1C). On the other hand, the VDR in cells administered 1,25D3 showed strong nuclear signals in almost all cells in a dose-dependent manner, which was reproducibly and consistently observed in 25D3-treated siCYP27B1-LNCaP cells (Fig. 1C). This nuclear accumulation of the VDR was seen when we used a dose as low as 10 −8 M of 25D3 and 1,25D3. Western blotting experiments from each hormone-treated LNCaP cell line also confirmed that 25D3 and 1,25D3 induced the nuclear accumulation of the VDR (Fig. 1D).
Exhaustive gene expression profiles reveal that the effects of 25D3 are almost exclusively mediated by the VDR. To reveal the gene regulatory profiles induced by 25D3, we performed a series of microarray analyses. Gene expression profiles were determined using total mRNA from either siCT cells or siVDR cells, together with or without the transient introduction of siCYP27B1 followed by treatment with 10 −7 M of 25D3 for 24 h. The Z-score and ratio of activity after 25D3 treatment were calculated. At first, we focused on the common genes induced by 25D3 in both siCT cells and siCYP27B1 cells. We isolated 25D3 target genes satisfying the Z-score criteria ≥2.0 or ≤−2.0 and a ratio ≥1.5 or ≤0.66 in both siCT and siCYP27B1 cells. As a result, 195 probes and 184 probes were isolated as 25D3 target genes in siCT cells and siCYP27B1 cells, respectively ( Fig. 2A: the gene list is presented in Supplemental Table S3A and B). A Venn diagram of the target genes indicated that 126 probes were shared between siCT cells and siCYP27B1 cells ( Fig. 2A: the gene list is presented in Supplemental Table S3C), while 69 targets were present specifically in siCT cells and 58 targets were present specifically in siCYP27B1 cells ( Fig. 2A: the gene list is presented in Supplemental Table S3D and E), although the implications of these results, such as the notion that a concentration of 10 −7 M of 25D3 was still too low in some cases, remain unresolved.
Then, to investigate how the VDR contributed to the gene regulation program of 25D3 in siCYP27B1 cells, we performed microarray analysis using mRNAs derived from LNCaP cells with doubly introduced siCYP27B1 and siVDR. From 184 probes of the genes regulated by 25D3 in siCYP27B1 cells ( Fig. 2A), we isolated genes satisfying our ratio criteria of ≤1.5 and ≥0.66 after treatment with 10 −7 M of 25D3 in doubly introduced siCYP27B1 + siVDR cells, and we assigned them as 25D3-VDR-mediated target genes. We found that 176 probes, 96% of the 184 species of 25D3 target genes in siCYP27B1 cells, were 25D3-VDR-mediated target genes ( Fig. 2B; the gene list is presented in Supplemental Table S4A). The 8 25D3 target genes that were seemingly not regulated by the VDR (Fig. 2B) are listed in Supplemental Table S4B. These results indicated that the VDR has a crucial role as a nuclear receptor in 25D3 signalling in LNCaP cells.
The 25D3 target gene profile is within the range of the 1,25D3 target gene profile. To verify the bioactivity of 25D3, comprehensive microarray analyses were performed using each mRNA sample from vehicle-vs 1,25D3-administered siCT cells and vehicle-vs 25D3-administered siCYP27B1 cells, respectively. We collected their target gene profiles satisfying a Z-score ≥ 2.0 or ≤−2.0 and a ratio ≥1.5 or ≤0.66. As shown in Supplemental Tables S3B and S5, 263 probes were identified as 1,25D3 target genes in siCT cells and 184 probes were identified as 25D3 target genes in siCYP27B1 cells. Gene ontology (GO) term analysis was performed using the term 'Biological Process' , revealing that 1,25D3 and 25D3 regulated diverse GO terms. For the 1,25D3 target genes, several metallothionein (MT) genes (MT1B, MT1E, MT1F, MT1G, MT1L, MT1X, and MT2A) formed the GO terms of 'Cellular response to zinc ion' , 'Negative regulation of growth' , and 'Cellular response to cadmium ion' (Table 2). On the other hand, for the 25D3 target genes, GO terms of 'Flavonoid biosynthetic process' and 'Flavonoid glucuronidation' were distinguished and consisted of the UGT2B gene family (UTG2B7, UGT2B10, UGT2B11, and UGT2B15), although UGT2B7 and UGT2B10 were also listed in 'lipid metabolic process' of the 1,25D3 GO term ( Table 2).   To reveal the comprehensive gene regulatory profiles, we applied more strict criteria (Z-score ≥ 2.0 or ≤−2.0 and ratio ≥2.0 or ≤0.5). As shown in Fig. 3A, 76 probes were identified as 1,25D3 target genes in siCT cells and 46 probes were identified as 25D3 target genes in siCYP27B1 cells (the gene list is provided in Supplemental Table S6A and B). These two groups shared 36 probes in common (the gene list is provided in Supplemental  Table S6C). Interestingly, the 25D3-specific population was small and consisted of 22% of the 25D3 target genes (  Table S6D). These results indicated that the gene profiles of 1,25D3 and 25D3 were largely overlapping, while 1,25D3 could independently regulate a certain set of genes which 25D3 could not regulate, even at the highest concentration used in the present study. We then reconfirmed the microarray results of some genes selected from each target group using qRT-PCR. For example, SLC39A10 and MT2A were screened as 1,25D3-specific targets, although even 10 −7 M of 25D3 had no discernible effect on their expression; PPFIBP2 and TSC22D3 were screened as common targets; and CHAC1 and G6PD were assessed as 25D3-specific targets (Fig. 3B). These results supported the hypothesis that 25D3-specific targets exist, but form a small population. Again, we raise the possibility that even 10 −7 M of 25D3 was still too low in some cases. The significance of the result that 10 genes, 2 of which (right panel of Fig. 3B) were analysed further by qRT-PCR, were 25D3-specific genes remains unsolved and is currently being investigated in our laboratory.    (Table 2). Especially, MT2A, which showed the highest expression among them in LNCaP cells (Supplemental Table S7), was repressed specifically by 1,25D3, but not 25D3 (Fig. 3B). We hypothesised that this repression of MT2A was involved in the 1,25D3-specific anti-proliferative function in LNCaP cells because recent reports suggested that the elevated expression of MTs was correlated with many types of cancer and their lethality 35,36 . Quantitative reduction of MT2A expression by introducing two different specific siRNAs for the MT2A gene was performed, resulting in its expression decreasing to 33% by siMT2A-1 and 32% by siMT2A-2 (Fig. 4A). Using the cultured cells after the introduction of each siRNA for 96 h, we investigated their proliferation. As a result, both of their relative growth rates were decreased significantly to 70% compared with the proliferation of siCT cells (Fig. 4B). Subsequently, we investigated the anti-proliferative action of 1,25D3 and 25D3 in these siMT2A cells. It has been reported that 1,25D3 shows an anti-proliferative effect on the growth of LNCaP cells 31 . We confirmed that 1,25D3 decreased cell proliferation in a dose-dependent manner, while 25D3 were selected from the 1,25D3-specific target genes in siCT cells. CHAC1 and G6PD were selected from the 25D3-specific target genes in siCYP27B1 cells. PPFIBP2 and TSC22D3 were selected as common target genes in both groups. Black and light grey bars indicate the results after vehicle (V) and 1,25D3 or 25D3 treatment, respectively. Each ratio calculated from the microarray data is presented below each graph. **P < 0.01 and *P < 0.05 compared with each vehicle. Data are shown relative to vehicle treatment and presented as the mean ± SD (n = 4).

Discussion
Prostate cancer is one of the most common malignancies in men. As the carcinoma advances, lethal metastasis often reduces quality of life. Aside from antiandrogens, one unanticipated, but promising, alternative is the use of vitamin D3, 1,25D3, or calcitriol, which exerts prodifferentiative, antiproliferative, and proapoptotic effects on prostate cancer cells both in vitro and in vivo 37 . To date, clinical trials of calcitriol, either alone or in combination with other chemotherapeutic agents have been conducted in patients with castrate-resistant prostate cancer [38][39][40][41][42][43][44] , resulting in limited antitumor benefits. More recently, the relationship between the VDR and prostate cancer has been investigated rigorously 45,46 . Conversely, the additive therapeutic role of vitamin D metabolites has been discussed widely in the treatment of oestrogen receptor-negative breast cancer in which the effects of anti-oestrogenic compounds seem to have some beneficial roles 47 . We hypothesised that 25D3 is an active vitamin D 3 ligand for the VDR in certain experimental conditions and cell lines. To strengthen our hypothesis, it is imperative to shut down the intracellular production of 1,25D3 by CYP27B1. It was reported that the mRNA expression and/or 1α-hydroxylase activity of CYP27B1 tend to decline during the progression of tumour development and the acquisition of higher malignant characteristics in prostate cancer 24 . Whitlatch and colleagues investigated CYP27B1 expression in several prostate tissues, including primary cultures of normal prostate, benign prostatic hyperplasia, and prostate cancer and its cell lines 34 . They found that normal prostate cells exhibited the highest expression of CYP27B1 among the tissues examined, while its expression was decreased in the following order: normal prostate, benign prostatic hyperplasia, and finally prostate cancer and its cell lines. These findings suggest that the malignant progression of prostate tissue certainly reduces CYP27B1 expression. Indeed, some reports have shown 24,48 , and we have also confirmed, that its activity was extremely low or non-detectable in LNCaP cells using an ultrasensitive LC-MS/MS assay (Supplemental Table S2), indicating that LNCaP cells lose CYP27B1 activity. We further treated LNCaP cells with siRNA for CYP27B1 to dampen further any residual CYP27B1 activity, if any. Thus, we concluded that the series of responses induced by 25D3 observed in this study were ascribed to bona fide 25D3 action.
We realised that there might be a cell type-specific outcome for the anti-proliferative effect of 25D3 on prostate cancer cells. We reconfirmed that 25D3, unlike 1,25D3, lacked the anti-proliferative function in LNCaP cells (Supplemental Fig. S1) reported by others 31 . On the other hand, PZ-HPV-7 cells, which were derived from normal prostate tissue, were demonstrated to be sensitive to the anti-proliferative action of 25D3; an analogue of 25D3, 25(OH)-19-nor-D3, which is rarely subjected to 1α-hydroxylation, and 25D3 itself showed anti-proliferative effects in this cell line 30,49 . Such discrepant sensitivity to 25D3 may be due to the fact that PZ-HPV-7 cells have acquired a distinct functional machinery along with specific proliferative apparatus in response to 25D3. Gene profiling of LNCaP cells revealed that the MT genes MT1B, MT1E, MT1F, MT1G, MT1L, MT1X, and MT2A were repressive target genes of 1,25D3 but not 25D3 (Table 2); especially, MT2A was found to be markedly repressed by 1,25D3 by using qRT-PCR (Fig. 3B). MTs, which form a gene family consisting of several MT1s, MT2A, MT3, and MT4, were identified originally as heavy metal ion binding proteins and responsible for the maintenance of cellular metal homeostasis 50,51 . Recently, the expression of MT genes was reported to be correlated with the carcinogenesis of several tumours and tumour progression, including prostate cancer 35,36 . The robust repression of MT2A by 1,25D3, but not by a supraphysiological concentration of 25D3, might contribute to its selective anti-proliferation activity. We confirmed that the quantitative reduction of MT2A expression by the introduction of its siRNA inhibited cell proliferation. In addition, neither 1,25D3 nor 25D3 exhibited further growth repressive effects in these cells (Fig. 4C). These results indicated that repression of MT2A might be a prerequisite for 1,25D3 to exert its anti-proliferative action. It has been reported that MT genes exert their proliferative function via multiple pathways in certain cancer cells. In human prostate cancer PC-3 cells, a lack of MT2A was shown to induce apoptosis with the down-regulation of c-MYC and BCL-2 using a sequence-specific ribozyme technique 52 . In other reports, MTs were demonstrated to interact directly with p53, which needs a zinc atom to perform its DNA-binding and transcriptional activity 53 . MTs function as metal chelators in this complex, resulting in the repression of p53 function 54 . From this perspective, the 1,25D3-dependent repression of MT2A expression might lead to a gain of p53 function, resulting in the induction of p53 signalling-dependent cell cycle arrest or apoptosis.
The cellular uptake of 25D3 is believed to be greater than that of 1,25D3 due to its higher hydrophobicity. Further, 25D3 is more stable than 1,25D3; the plasma half-life of 25D3 is approximately 2 weeks, while that of 1,25D3 is less than 4 h [55][56][57] . Given that the plasma concentration of 25D3 is 500-1000 times higher than that of 1,25D3, some authors have proposed that such a physiological concentration of 25D3 binds to and activates the VDR. However, the concept that a 1/500 ratio of total hormones can directly correspond to free active hormone is unlikely owing to the variable concentrations of vitamin D-binding protein (DBP). As it is known that DBP binds to 25D3 with a higher affinity than to 1,25D3, the ratio of free 1,25D3 to free 25D3 might be higher than the 1/500-1/1000 ratio of total 1,25D3 to 25D3. As such, an estimation based on both the concentration of DBP and the relative affinity of both metabolites to the VDR gives approximate concentrations of 10 and 1 pM for free 25D3 and 1,25D3, respectively 58 . Indeed, 10 −8 M of 1,25D3 had a stimulatory effect on CYP24A1 expression, which was only 3-4-fold lower than that after treatment with 10 −7 M of 25D3 (Table 1). Thus, from the physiological viewpoint, 25D3 can bind to and activate the VDR in vitro. In vivo, relevant evidence for a direct effect of 25D3 is provided by the ability of high-dose cholecalciferol to prevent osteomalacia and hypocalcaemia in 1-α-hydroxylase knockout mice 59,60 . In another example, it was shown that extracellular matrix mineralization was induced by 25D3, but not by 1,25D3, leading the authors to reach a conclusion that 25D3, despite its lower affinity to the VDR, can stabilise the VDR-ligand binding domain in its agonistic conformation in the same way as 1,25 D3 does 14 .
Hewison's group reported recently that high muscle strength was provided by 25D3, but not by 1,25D3 61 . In their report, there was no significant correlation between the serum levels of 25D3 and 1,25D3. In contrast, serum 25(OH)D 3 , 24,25(OH) 2 D 3 , and 3-epi-25(OH)D 3 levels were strongly correlated. They also reported that parallel analysis of CYP27B1 in muscle biopsies did not reveal the significant expression of this gene, suggesting that the localised conversion of 25(OH)D 3 to 1α,25(OH) 2 2D 3 is not a prominent feature of muscle, as was reported for the LNCaP cells shown here. Although they did not examine whether the 25D3-VDR axis functions in muscle, they found that out of the 92 genes examined, the expression of 24 skeletal muscle genes was strongly correlated with serum 25D3. These genes are involved in hormonal/intracellular signalling, cell stress, proteasomal activity, protein translation, amino acid metabolism, muscle contraction, myogenesis, and mitochondrial function. 1,25D3 was also associated with the expression of genes involved in skeletal muscle metabolism, protection against cell stress, and protein degradation. Thus, there might be another signal transduction system via which the effect of 25D3 is exerted by a non-VDR-mediated pathway in muscle.
Among the genes whose expression was affected by the single action of either 25D3 or 1,25D3 as well as their combined action in prostatic cancer LNCaP cells, we could not find any commonality, except that approximately 47% (Fig. 3A: 36/73) of the 1,25D3 target genes were also under the control of 25D3. We could not resolve the possibility that a higher dose of 25D3 than used in this study might complement the 1,25D3 target gene profile completely. CYP24A1 shown here is a typical example (Table 1). Altogether, in LNCaP cells, we compared the individual target gene profiles of 1,25D3 and 25D3 to clarify their interrelationship. We found that 25D3 largely regulated the expression of its target genes in common with 1,25D3 via the VDR when the VDR is expressed in prostate cancer cells, although the anti-proliferative activity of 25D3 was absent. We surmise that the ability to suppress the expression of MT genes may be one of the most important mechanisms for 1,25D3, but not 25D3, to exert its anti-proliferative action. Further study is warranted to examine the effectiveness of 25D3 in combating prostate cancer.

Methods
Cell cultures and hormones. Monolayer culture and maintenance of LNCaP cells were described previously 62 . For the proliferation assay, RPMI phenol red-free medium containing 10% charcoal-stripped foetal bovine serum with antibiotics (10% CS-FBS-medium) was used. 1α,25(OH) 2 D 3 (referred to as 1,25D3) and 25(OH)D 3 (referred to as 25D3) were supplied by DSM (Het Overloon, Netherlands). siRNA transfection and qRT-PCR. Non-targeting siRNA and a mixture of four different siRNAs specific for VDR were obtained from Dharmacon (Thermo Fisher Scientific, Waltham, MA). siRNA for CYP27B1 was obtained from Invitrogen (Thermo Fisher Scientific). siRNAs for MT2A (MT2A-1 and MT2A-2) were obtained from Sigma Aldrich (Merck KGaA, Darmstadt, Germany). Transfection of siRNA using Lipofectamine RNAiMAX (Thermo Fisher Scientific) was described previously 62 . Methods for the isolation of total RNA from cells, synthesis of cDNA, and qRT-PCR were also described previously 62,63 . The primer sets used in this study are shown in Supplemental Table S1. Data were normalised to the expression of TATA-binding protein (TBP) and presented relative to the result of vehicle ethanol treatment. Values are expressed as the mean ± standard deviation (SD) (n = 4).
Western blotting. The details of western blotting analysis were described previously 62 . For the detection of VDR-specific immunobands, a rabbit antibody against human VDR diluted at 1:1000 (D2K6W; Cell Signaling Technology, Danvers, MA) was used. The anti-VDR antibody used in this study recognised N-terminal amino acid residues, resulting in a reaction with both VDRB1 and VDRB2 isoforms.
Immunocytochemistry. The details of immunocytochemistry were described previously 62 . A rabbit antibody against human VDR diluted at 1:2000 (D2K6W; Cell Signaling Technology) was used.
Microarray analysis. Isolation of total RNA from siCT, siCYP27B1, or siVDR cells, which were treated previously with 10 −7 M of 25D 3 or 1,25D3 for another 24 h, was performed using an RNeasy Mini Kit (QIAGEN, Hilden, Germany). Using RNA samples, cDNA microarray analysis was performed by APRO Life Science Institute (Tokushima, Japan) and Cell Innovator (Fukuoka, Japan) as described previously 62 . The GEO accession number is GSE107438. We selected the probes that called the 'P' flag and the cut-off signal was smaller than 100. To identify regulated genes, we calculated the Z-score and the ratio (non-log scaled fold-change) for comparisons between each vehicle-and hormone-treatment sample.
Statistical analysis. Data are expressed as the mean ± SD of quadruplicate assay, unless otherwise stated.
Student's t-test (Fig. 1A) or one-factor analysis of variance and Tukey's test (Figs 1B, 3B and 4 and S1) were carried out.
Data availability. The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.