Cyp24a1 Attenuation Limits Progression of Braf^V600E-Induced Papillary Thyroid Cancer Cells and Sensitizes Them to BRAF^V600E Inhibitor PLX4720

Papillary thyroid cancer (PTC) is the most common type of thyroid cancer, accounting for more than 80% of thyroid cancer cases (1). The BRAF^V600E mutation is the most frequent genetic alteration in PTC, occurring in 28% to 83% of cases with an average rate of 44% (2–4). Constitutive activation of the RAS–RAF–MEK–ERK MAP kinase signaling pathway (MAPK) promotes the initiation and progression of PTC.

Vitamin D is mainly involved in bone and mineral metabolism. It has other important functions, such as the modulation of cell growth and immune function (5). Its antiproliferative effects have attracted great enthusiasm in recent years for its potential application as an anticancer agent. Significant antiproliferative effects have been observed in many human cancer cells, including thyroid, prostate, breast, colorectal, and lung cancers (6–9). Vitamin D receptor (VDR) knockout mice displayed a higher incidence of carcinogen-induced breast and skin tumors (10), and vitamin D deficiency promotes human breast cancer growth (11). Although clinical trials have shown the potential therapeutic effects of calcitriol in prostate cancer patients (12), the success has not been convincing regarding the clinical effects of vitamin D or its analogues in cancer treatment (13, 14). This may be due to the overexpression of CYP24A1 in many cancer patients.

Vitamin D 24-hydroxylase (CYP24A1) is the primary vitamin D-inactivating enzyme, which catabolizes 1α, 25(OH)₂D₃ (calcitriol) and, to a lesser extent, 25(OH)₂D₃ by 24 hydroxylation into inactive 1α, 24,25(OH)₃D₃ and 24,25(OH)₂D₃ (6). The calcitriol-mediated antiproliferative effects could be disrupted by CYP24A1 overexpression during tumor development (7). Indeed, CYP24A1 overexpression has been observed in many cancers, including thyroid (15, 16), lung (17), colon (18), esophageal (19), and breast (20), and has been linked to poor prognosis in patients with lung (21), esophageal (19), colon (22), and thyroid (16, 23) cancers. It has been proposed as a candidate oncogene due to its gene amplification in breast cancer (24). In patients with thyroid cancer, the serum calcitriol level was found to be significantly lower (25), although there was no significant difference in the serum 25(OH) D₃ level between thyroid nodule and thyroid cancer patients (25, 26), indicating that calcitriol might be converted to inactive 1α, 24,25(OH)₃D₃ by increased CYP24A1 expression. Although these data suggest that CYP24A1 overexpression could result in the abrogation of calcitriol-mediated growth arrest leading to tumor development and/or progression, there are no in vivo functional studies to support this hypothesis.

In our previous study, we demonstrated that CYP24A1 overexpression was associated with BRAF^V600E mutation and advanced stages of PTC (23). We also showed that BRAF^V600E induced CYP24A1 overexpression and the BRAF^V600E inhibitor PLX4720 significantly enhanced the antiproliferative effects of calcitriol in thyroid cancer cell lines (23). However, it is not clear to what extent CYP24A1 overexpression contributes to thyroid cancer development and progression in vivo, especially in the presence of BRAF^V600E. In the present study, we used a mouse model of Braf^V600E-induced PTC to investigate the role of Cyp24a1 in thyroid cancer progression. We observed that thyroid cancer growth was significantly reduced in the absence of Cyp24a1 expression.

The generation of TPO–Braf^V600E and Cyp24a1 knockout mice (Cyp24a1^null) have been described previously (27–29). TPO-Braf^V600E mice with wild-type Cyp24a1 (BVE^Cyp24a1-wt) developed PTC at approximately 5 weeks of age and were used as PTC tumor controls. TPO-Braf^WT mice with wild-type Cyp24a1 were used as normal controls. TPO-Braf^V600E mice with Cyp24a1 knockout (BVE^Cyp24a1-null) were obtained by several rounds of breeding among LSL-Braf^V600E(30) TPO-Cre (31), and Cyp24a1^+/− mice. Because 50% of the homozygous mutant Cyp24a1^null mice died before 3 weeks of age (29), the mice were kept in a heterozygous state (Cyp24a1^+/−). To knockout Cyp24a1 in TPO–Braf^V600E mice, Cyp24a1^+/− mice were first crossed with LSL–Braf^V600E or TPO–Cre mice to generate a Cyp24a1^+/−; Braf^V600E strain or TPO–Cre; Cyp24a1^+/− strain. Cyp24a1^+/−; Braf^V600E mice and TPO–Cre; Cyp24a1^+/− mice were then bred together to create TPO–Braf^V600E–Cyp24a1^{−/− or null} mice. Female athymic BALB/c-nu/nu mice (6–10 weeks of age) were acquired from The Jackson Laboratory. Mice were provided with autoclaved food and water ad libitum. The study was approved by the Animal Care and Use Committee of the institution and conducted in compliance with the Public Health Service Guidelines for the Care and Use of Animals in Research.

The genotyping of Cre-mediated recombination of the LSL–Braf^V600E targeted allele has been described previously (27). Briefly, the following primers were used to detect LSL–Braf^V600E recombination in the mouse tissue: primer A, 5′-AGTCAATCA TCCACAGAGACCT-3′; primer B, 5′-GCTTGGCTGGACGTAAACTC-3′; and primer C, 5′-GCCCAGGCTCTTTATGAGAA-3′. Primers A + C detected the wild-type allele (466 bp) and Cre-recombined Braf^V600E allele (518 bp). Primers B + C detected the LSL–Braf^V600E allele (140 bp). For genotyping the Cyp24a1-knockout mice, the following primers were used: primer 1, 5′-GCAGCATCTCCACAGGTTCACTGTC-3′; primer 2, 5′-AAGAT- CAACCCCTTCGCTCATCTCC-3′; and primer 3, 5′-CGCATCGCCTTCTATCGCCTTC-3′. Primers 1 + 2 detected the wild-type allele of 250 bp, and primers 1 + 3 detected the mutant allele of 600 bp. The PCR conditions were as follows: 94°C for 5 minutes followed by 35 cycles of amplification (94°C for 30 seconds, 58°C for 30 seconds, 72°C for 1 minutes) with a final extension at 72°C for 10 minutes.

Thyroid tumors were collected aseptically from donor mice (BVE^Cyp24a1-wt and BVE^Cyp24a1-null) using blunt dissection, then mechanically dissociated by mincing and passing through a 40-μm/mesh sterile screen, and suspended in DMEM/F12 growth medium (10% fetal bovine serum, 100 units/mL penicillin, 100 μg/mL streptomycin). The cells were further dissociated by incubation in growth medium containing 100 U/mL type I collagenase (Sigma-Aldrich) and 1.0 U/mL dispase I (Roche Diagnostics) at 37°C in a rocking water bath for 60 minutes. The cell suspension was washed twice with growth media and resuspended in a 10-mm culture dish with DMEM/F12 growth medium containing 4 mU/mL bovine TSH (Sigma-Aldrich) to establish BVE^Cyp24a1-wt and BVE^Cyp24a1-null cell lines. The established cell lines were propagated in DMEM/Ham's F12 growth medium. Two cell lines were established each from two separate BVE^Cyp24a1-wt or BVE^Cyp24a1-null primary thyroid tumors in 2015: BVE^Cyp24a1-wt-1, BVE^Cyp24a1-wt-2, BVE^Cyp24a1-null-1, and BVE^Cyp24a1-null-2. The thyroid origin of these cell lines was confirmed by genotyping in the lab as described above, which was performed 1 month before using the cell lines.

Total RNA was isolated from the thyroid tumor tissues of BVE^Cyp24a1-wt mice and BVE^Cyp24a1-null mice by the quanidinium thiocyanate–phenol–chloroform method (32). The integrity of the RNA was verified by denaturing gel electrophoresis. Two micrograms (μg) of each total RNA was reverse-transcribed to cDNA using the Promega RT system (Promega). The LightCycler DNA Master SYBR Green 1 Kit was used for quantitative real-time PCR analysis (33). The cDNA mix was diluted 10-fold, and 2 μL of the dilution was used for real-time PCR analysis. The PCR primers for the 126-bp Cyp24a1 cDNA fragment were: 5′-CATCGCAACGAAGCCTACGGG-3′ (sense, located in exon 2) and 5′-CTCATTGATTTTCTTGTCCAGC-3′ (antisense, located in exon 3). The sense primer spans over 756 bp intron 2 so that the contaminated genomic DNA would not be amplified. The Cyp24a1 cDNA fragment was verified by DNA sequencing. The mRNA level of the housekeeping gene Actb (β-actin) was used as an internal control, and a 180-bp PCR product was amplified using the following two primers: 5′-AAATCGTGCGTGACATCAAA-3′ (sense) and 5′-AAGGAAGGCTGGAAAA GAGC-3′ (antisense). The PCR conditions are 94°C for 30 seconds followed by 30 cycles of amplification (94°C for 10 seconds, 48°C for 5 seconds, and 72°C for 10 seconds). The resulting concentration of Cyp24a1 PCR products was normalized by comparison with β-actin and was used to determine the relative mRNA level of Cyp24a1 in the thyroid tumors (ddCt method; ref. 33).

Blood was collected by cardiac puncture. Serum thyroid stimulating hormone (TSH) was measured using the MILLIPLEX MAP Mouse Pituitary Magnetic Bead Panel following the manufacturer's instructions (EMD Millipore Corporation).

Histology and immunohistochemical staining were performed as described previously (34). Briefly, 4-μm-thick formalin-fixed paraffin-embedded tissue sections were prepared and stained with hematoxylin and eosin (H&E) or with a Ki67 antibody (1:100 dilution, ab16667, Abcam). A DAKO LSAB + kit using horseradish peroxidase (HRP) was used for immunostaining (DAKO). The sections were counterstained with Mayer's hematoxylin.

The Cyp24a1 cDNA was cloned into pcDNA3.1 as described previously (35). The expression construct was transfected into the BVE^Cyp24a1-null cell line using Lipofectamine (Invitrogen) and selected for 4 weeks with 400 μg/mL zeocin. Stable clones were pooled and used for subsequent experiments.

Cell lysates were obtained by extraction in RIPA buffer (20 mmol/L Tris-HCl, pH7.4, 150 mmol/L NaCl, 5 mmol/L EDTA, 1% NP-40) containing Pierce's Halt Protease Inhibitor Cocktail (Thermo Scientific). The protein concentration was determined by Bradford's assay using a Bio-Rad protein assay kit (Bio-Rad). The proteins (40 μg) were separated on a 12% SDS–polyacrylamide gel and then transferred to a PVDF membrane. Western blot analysis was performed using antibodies (1:1000 dilution, Cell Signaling Technology, Inc.) against phospho-Erk 1/2 (#4370), phosphor-Akt (#4060), p-Smad2 (#3101), E-cadherin (#3195), Snail (#3879), vimentin (#5741), and vitamin D receptor (#12550), or antibodies against Zeb1 (1:1000 dilution, sc-25388,Santa Cruz Biotechnology), or CYP24A1 (1:1000 dilution, ab 109632, Abcam).

Cells were seeded in 6-well plates (10⁵ cells/well), and a linear scratch was created with a sterile pipette tip when cells reached confluent monolayer. The cells were rinsed three times with medium to remove cellular debris. Cell migration or wound-healing was monitored by microscopy after 16-hour culture.

Cell proliferation was measured by a nonradioactive MTT assay kit according to the manufacturer's procedure (Promega Corp). Briefly, the cells were plated in triplicate into 96-well plates (10³ cells/well) in growth medium containing vehicle (0.5% DMSO), different concentrations of calcitriol (Sigma-Aldrich), PLX4720 (Selleck Chemicals), or both for up to 72 hours. For the final 4 hours of incubation, 20 μL of CellTiter 96 AQueous One Solution reagent was added into each well for the measurement of cell viability.

Cells were plated into 12-well plates (5 × 10² cells/well) and cultured for 14 days in the presence of different concentrations of calcitriol, PLX4720, or both. The cells were then fixed with methanol for 10 minutes and stained with 0.5% crystal violet dye (in methanol:deionized water, 1:5) for 10 minutes. After three washes with deionized water to remove excess crystal violet dye, the crystal violet dye was released from the cells by incubation with 1% SDS for 2 hours before optical density (OD)_{570 nm} measurement.

Total RNA from BVE^Cyp24a1-wt-1 and BVE^Cyp24a1-null-1 cell lines was isolated, and libraries were constructed using an Illumina TruSeq RNA Library Prep kit according to the manufacturer's procedure. All sequencing was performed on Illumina HiSeq 4000 with at least 20 million clean reads. The significant differentially expressed genes (DEG) were selected based on the following criteria: Log₂-fold change >2, false discovery rate (FDR) <0.001, and P value from difference test < 0.01. Selected DEGs were verified by qRT-PCR, and their primer sequences are provided in Supplementary Table S1.

Student t test (two-tailed) was used to compare two groups, and one-way ANOVA was used to compare multiple groups. A P value of 0.05 or less was considered statistically significant.

We examined Cyp24a1 expression by qRT-PCR in normal thyroid tissues from 8 TPO-Braf^WT mice and 2 thyroid tumors from different age groups of BVE^Cyp24a1-wt mice (from 1 to 6 months old). Consistent with our previous findings in human PTC, the Cyp24a1 expression was increased by more than 3 times in 1-month-old thyroid tumors from both male and female mice with Braf^V600E mutation. The Cyp24a1 expression was gradually increased as they aged up to 7 to 9 times in 6-month-old tumors as compared with normal thyroids (Fig. 1A, P < 0.001). To investigate the effect of Cyp24a1 overexpression on thyroid tumor growth in the presence of the Braf^V600E mutation, we knocked out the Cyp24a1 gene by cross-breeding BVE^Cyp24a1-wt mice with Cyp24a1^null mice. The Cyp24a1 knockout was confirmed by genotyping (Fig. 1B). As shown in Fig. 1C, thyroid tumor growth was significantly reduced in BVE^Cyp24a1-null compared with BVE^Cyp24a1-wt mice. Among 12 age- and sex-matched pairs (3 pairs in each group), the thyroid tumor load at different ages was reduced by an average of 4-fold in BVE^Cyp24a1-null mice (35.00 ± 11.73 mg vs. 127.5 ± 43.78 mg in BVE^Cyp24a1-wt mice, P < 0.05, Fig. 1C). The papillary architecture of thyroid tumors was lost and replaced by a more compact structure with reduced immunostaining of the cell proliferation marker Ki67 (Fig. 1D). These observations were not found in the heterozygous Cyp24a1-knockout mice carrying the Braf^V600E mutation (data not shown). The phenotype of BVE^Cyp24a1-null mice was similar to that of BVE^Cyp24a1-wt mice, and their body weight was approximately half of the normal mice. Both BVE^Cyp24a1-null and BVE^Cyp24a1-wt mice had severe hypothyroidism with elevated levels of serum TSH greater than 50,000 pg/mL (n = 5), which was beyond the detection limit of the assay and more than 100-fold higher than normal TPO-Braf^WT mice (n = 5, 394.3 ± 8.7 pg/mL). These data indicate that the loss of Cyp24a1 had no impact on hypothyroidism caused by the Braf^V600E mutation.

Due to impaired vitamin D catabolism, 50% of Cyp24a1^null mice died before 3 weeks of age (29, 36). A similar mortality rate was found in BVE^Cyp24a1-null mice, and the remaining mice could live up to 6 months. Because we could not maintain sufficient numbers of BVE^Cyp24a1-null mice to evaluate the long-term effect of Cyp24a1-knockout on tumorigenicity and survival, we instead established thyroid tumor cell lines from BVE^Cyp24a1-null and BVE^Cyp24a1-wt mice and injected them subcutaneously into nude mice (n = 5) to observe tumor growth. The BVE^Cyp24a1-null cells expressed neither functional Cyp24a1 nor responded to calcitriol stimulation to induce Cyp24a1 expression (Fig. 2A). The residual Cyp24a1 expression detected by qRT-PCR was likely truncated transcripts (without exon 9 and 10) undergoing nonsense-mediated decay. The CYP24A1 protein was not detected in the BVE^Cyp24a1-null cells and reappeared after transfection of Cyp24a1 cDNA (Fig. 2B). Cyp24a1 transcripts from BVE^Cyp24a1-wt cells were significantly increased following 16 hours of calcitriol stimulation (Fig. 2A). However, no change in the CYP24A1 protein level was observed following 100 nmol/L calcitriol stimulation for 48 hours. The elevated protein level was demonstrated after 72 and 96 hours of stimulation (Fig. 2B and C). The delay in protein synthesis after calcitriol stimulation may be due to the feedback inhibition by the already elevated CYP24A1 protein level and/or abrogated vitamin D signaling in the Braf^V600E-induced thyroid cancer cells. VDR expression was increased in all the cell lines after calcitriol stimulation. The basal level of VDR expression and its induction by vitamin D were reduced in BVE^Cyp24a1-null-1 cells (Fig. 2B). These data provided evidence that vitamin D signaling was intact and there was no amplification of vitamin D signaling as a result of Cyp24a1 deletion. As shown in Fig. 2D, the tumorigenicity of BVE^Cyp24a1-null cells from both BVE^Cyp24a1-null-1 and BVE^Cyp24a1-null-2 cell lines was significantly decreased. The tumor weight at 4 weeks after injection of 2 × 10⁶ cells was 0.10 ± 0.01 g from BVE^Cyp24a1-null-1 cells and 0.09 ± 0.02 g from BVE^Cyp24a1-null-2 cells vs. 1.90 ± 0.17 g from BVE^Cyp24a1-wt-1 cells and 1.97 ± 0.39 g from BVE^Cyp24a1-wt-2 cells (P < 0.001), a 19-fold reduction in the tumorigenic potential (Fig. 2D and E). The tumorigenic potential was partially recovered after reexpression of wild-type Cyp24a1 in the BVE^Cyp24a1-null-1 cells (0.1 ± 0.01 g in BVE^Cyp24a1-null-1 vs. 0.51 ± 0.09 g in BVE^Cyp24a1-null-1Cyp24a1, P < 0.001; Fig. 2D and E).

To investigate the mechanisms that resulted in the loss of tumorigenic potential, we studied the p-Erk, p-Akt, and p-Smad2 levels in the two BVE^Cyp24a1-null cell lines established from two separate tumors by Western blot analysis. Increased phosphorylation of these proteins was reported to be associated with tumor progression in thyroid cancer (37, 38). As shown in Fig. 3A, their phosphorylation levels were decreased in both BVE^Cyp24a1-null cell lines. We also found significant reduction of Snail and ZEB1 expression in the BVE^Cyp24a1-null cell lines (Fig. 3B). Both Snail and ZEB1 are zinc finger transcription factors that promote epithelial–mesenchymal transition (EMT) by downregulating the expression of the adhesion molecule E-cadherin (39, 40). As expected, E-cadherin expression was increased and the expression of the mesenchymal cell marker vimentin was not detected (Fig. 3B), indicating that EMT was absent in the BVE^Cyp24a1-null cells. Furthermore, cell migration was reduced in the BVE^Cyp24a1-null cells (Fig. 3C). To further confirm the reduced p-Erk, p-Akt, and p-Smad2 levels in the BVE^Cyp24a1-null cells was due to decreased Cyp24a1 expression, we transfected Cyp24a1 cDNA into the BVE^Cyp24a1-null-1 cell line (BVE^Cyp24a1-null-1Cyp24a1) to overexpress exogenous CYP24A1. As shown in Fig. 4D, the p-Erk, p-Akt, and p-Smad2 levels were increased, and E-cadherin expression was decreased in the BVE^Cyp24a1-null-1Cyp24a1 cells. The reexpression of Cyp24a1 did not rescue the expression of the EMT marker vimentin. It may take a longer time and/or need a higher level of CYP24A1 to induce its expression. These data demonstrated that CYP24A1 overexpression in Braf mutant cells could upregulate multiple signaling pathways to drive tumor progression.

Because CYP24A1 catabolizes calcitriol, we expected increased calcitriol-mediated growth arrest in BVE^Cyp24a1-null cells. To our surprise, we did not find any significant difference in cell proliferation between BVE^Cyp24a1-wt-1 and BVE^Cyp24a1-null-1 cells before or after calcitriol treatment (Fig. 4). This was confirmed by a nonradioactive MTS assay (200 nmol/L for up to 72 hours, data not shown). Next, we investigated whether the combination of calcitriol and PLX4720 could enhance the antiproliferative effect of PLX4720. Both BVE^Cyp24a1-wt-1 and BVE^Cyp24a1-null-1 cells were treated with calcitriol or PLX4720 alone or in combination for a short term (up to 72 hours) and long term (14 days). The short-term effects were determined by a nonradioactive MTS assay for cell proliferation, and the long-term effects were measured by a colony formation assay. The synergistic effects were not found during short-term culture (data not shown). A significant reduction in cell proliferation was, however, found during the long-term culture when BVE^Cyp24a1-null-1 cells were cultured in the presence of PLX4720 for 14 days (Fig. 4). BVE^Cyp24a1-null-1 cells were more sensitive to PLX4720 than BVE^Cyp24a1-wt-1 cells: 32% versus 75% viable cells after 8 μmol/L PLX4720 treatment (P < 0.0001). Although calcitriol alone had no significant effect, it synergized the antiproliferative effects of PLX4720, resulting in further reduction in cell viability to 9% in BVE^Cyp24a1-null-1 cells after combined treatment of 8 μmol/L PLX4720 and 200 nmol/L calcitriol versus 32% from single treatment of 8 μmol/L PLX4720 (P < 0.0001, Fig. 4). The synergistic effects could also be demonstrated in BVE^Cyp24a1-wt-1 cells: 18% viable cells after combined treatment of 8 μmol/L PLX4720 and 200 nmol/L calcitriol versus 75% viable cells after 8 μmol/L PLX4720 treatment alone (P < 0.0001, Fig. 4). Similar results were also observed in the BVE^Cyp24a1-wt-2 and BVE^Cyp24a1-null-2 cells (data not shown). To further confirm that CYP24A1 overexpression caused resistance to PLX4720, we tested BVE^Cyp24a1-null-1Cyp24a1 cell line, which overexpressed the CYP24A1 protein. As shown in Fig. 4, BVE^Cyp24a1-null-1Cyp24a1 cells became resistant to PLX4720 treatment: 50% viable cells versus 32% for BVE^Cyp24a1-null-1 cells after 8 μmol/L PLX4720 treatment (P < 0.0002). The calcitriol-mediated complementary effect was also reduced in BVE^Cyp24a1-null-1Cyp24a1 cells: 27% viable cells versus 9% after combined treatment with 8 μmol/L PLX4720 and 200 nmol/L calcitriol (P < 0.0002, Fig. 4). The synergistic effects of the combined treatment were still obvious when lower concentrations (2 μmol/L PLX4720 and 50 nmol/L calcitriol) were used in BVE^Cyp24a1-null-1 cells: 18% from a combined treatment versus 53% from a single PLX4720 treatment (P < 0.0001). In BVE^Cyp24a1-wt-1 cells, however, the synergistic effects of combined treatment were not significant at the concentration of 2 μmol/L PLX4720 and 50 nmol/L calcitriol: 62% from a combined treatment versus 72% from a single PLX4720 treatment (P = 0.0742, Fig. 4). The synergistic effects of combined treatment were demonstrated when higher concentrations (4 μmol/L PLX4720 and 100 nmol/L calcitriol) were used: 43% from a combined treatment versus 72% from a single PLX4720 treatment (P < 0.0013, Fig. 4). These data demonstrate the significant benefit of calcitriol in combination of PLX4720 for the treatment of Braf^V600E-positive PTC.

To investigate the impact of Cyp24a1 deletion on global gene expression of vitamin D and non–vitamin D-responsive genes, we performed RNA sequencing (RNA-Seq) analysis of BVE^Cyp24a1-wt-1 and BVE^Cyp24a1-null-1 cells before and after calcitriol stimulation (100 nmol/L for 16 hours). Differentially expressed genes (DEG) with log₂ ratio of BVE^Cyp24a1-wt-1 + D3/BVE^Cyp24a1-wt-1 > 2 (4-fold difference in gene expression after vitamin D stimulation) were selected as vitamin D–responsive genes. A total of 80 DEGs met the selection criteria as vitamin D–responsive genes and their transcript levels were compared among BVE^Cyp24a1-wt-1, BVE^Cyp24a1-wt-1+D3, BVE^Cyp24a1-null-1, and BVE^Cyp24a1-null-1+ D3. There were few significant changes in gene expression among vitamin D–responsive genes between BVE^Cyp24a1-wt-1 and BVE^Cyp24a1-null-1 cells: only 33 DEGs (17 down- and 16 upregulated) in BVE^Cyp24a1-null-1 cells (Fig. 5A; Supplementary Table S2). Several downregulated genes are known to be involved in tumor progression and metastasis: Aldh1l2, Efemp1, Gjb2, Krt17, Mmp13, Mmp17, and Notum (Supplementary Table S2). Among 16 upregulated genes, most of them had either unknown functions or their functions were paradoxically related to tumor progression (Mmp3, Kcnh1, and Krt16) except for Pdlim2 and Pigr, which might play a role in tumor growth inhibition (Supplementary Table S2). Pathway analysis showed potential interactions among Gjb2, Mmp13, Mmp17, Efemp1, and Pdlim2, affecting the extracellular matrix proteins, matrix metalloproteinase, and EMT (Fig. 6A). By contrast, 761 DEGs did not meet the selection criteria as vitamin D–responsive genes and were considered as non–vitamin D-responsive genes: 193 upregulated and 568 downregulated in BVE^Cyp24a1-null-1 versus BVE^Cyp24a1-wt-1 cells. We selected top 100 DEGs with log₂-fold change >6.57 (more than 95-fold difference in gene expression) for further analysis. A distinct pattern of gene expression was demonstrated between BVE^Cyp24a1-wt-1 and BVE^Cyp24a1-null-1 cells (Fig. 5B). Only 3 vitamin D–responsive genes were present: Efemp1, Hist2h2aa2, and Ppp1r2-ps3. The genes reported to be involved in tumor growth/progression or invasion/metastasis are listed in Table 1. Their relevant functions are provided in Supplementary Tables S2 and S3. We verified 13 DEGs from Table 1 by qRT-PCR and the results were consistent with the RNA-Seq data (Fig. 5C and D). Many of these genes are involved in the regulation of β-catenin/Wnt (Tff1↑, Olfm4↑, Fscn1↓, Cdh6↓, Dkkl1↓, and Notum↓, resulting in downregulation of β-catenin), TGF-beta (Nes↓, Dlx2↓), and Notch (Pdzrn4↓, Lnx1↓, and Msx1↑) signaling pathways, and may contribute to tumor regression. The significant downregulation of a group of EMT-promoting genes (Efemp1, Fscn1, Cdh6, Prrx1, Fut4, Hoxd9, Ctsz, Acp5, and Nes) may explain the loss of EMT in the BVE^Cyp24a1-null cells (Fig. 6). Interestingly, many of them are homeobox transcription factors: Msx1, Cdx2, Hoxd9, Prrx1, Hoxc5, En1, and Dlx2.

In the present study, we have demonstrated in vivo that CYP24A1 overexpression cooperates with oncogenic Braf^V600E to promote thyroid cancer progression. The oncogenic potential of Braf^V600E is significantly reduced after Cyp24a1 knockout. Furthermore, a synergy against thyroid cancer cells is observed between calcitriol and the BRAF^V600E inhibitor PLX4720, which may have immediate clinical translation for thyroid cancer patients, especially those who are resistant to targeted therapy by vemurafenib (PLX4032), which is the pharmacologic form of PLX4720 and has the same mode of action as PLX4720.

CYP24A1 overexpression has been reported in many cancers and can lead to the abrogation of growth control mediated by vitamin D. Multiple mechanisms are involved in its overexpression: gene amplification in breast and colon cancers (24, 41), miR-125b downregulation in breast cancer (42), and protein kinase CK2 activation in prostate cancer (43). Its overexpression is associated with poor prognosis in patients with lung (21), esophageal (19), colon (22), and thyroid (16, 23) cancers. These studies have clearly shown that CYP24A1 overexpression can be used as a biomarker for prognosis prediction, but its role in cancer development and progression has not been fully evaluated in vivo. The current study provides direct evidence that its overexpression promotes thyroid cancer growth and progression by upregulation of multiple signaling pathways, such as MAPK, PI3K/Akt, TGF-β, and EMT. Cyp24a1 encodes vitamin D 24-hydroxylase, which catabolizes calcitriol into inactive 1α, 24,25(OH)₃D₃. Calcitriol exerts its antiproliferative effects via cross-talk with other signaling pathways, including MAPK, PI3K/Akt, and TGFβ (7, 44). The upregulation of multiple signaling pathways may be due to increased degradation of calcitriol by Cyp24a1 overexpression, but a more plausible explanation would be that Cyp24a1 may have some intrinsic oncogenic functions apart from catabolizing calcitriol. This hypothesis is supported by the following observations: (i) Calcitriol treatment alone had no significant effect on cell proliferation in Cyp24a1-knockout cell line BVE^Cyp24a1-null; (ii) The tumor growth of BVE^Cyp24a1-null cells was dramatically reduced in nude mice without an increase in the serum calcitriol level; (iii) The tumor growth in nude mice was partially recovered by transfection of Cyp24a1 cDNA into BVE^Cyp24a1-null cells; (iv) No significant changes in the expression of vitamin D–responsive genes after Cyp24a1 knockout whereas many non–vitamin D-responsive genes were affected, including transcription factors, oncogenes, tumor suppressor genes, and genes involved in the immune surveillance and metabolism. These data indicate that calcitriol alone or elevated vitamin D signaling may not have significant impact on thyroid tumor growth of BVE^Cyp24a1-null cells. It is the Cyp24a1 deletion that likely contributes to the tumor regression. Although the hypothesis remains to be confirmed by further studies, the current study demonstrates that Cyp24a1 overexpression results in upregulation of multiple signaling pathways for thyroid cancer progression: Braf^V600E → Cyp24a1 overexpression → MAPK, PI3K/Akt, TGF-β, and EMT activation → thyroid cancer progression.

Thyroid tumor growth was reduced by 4-fold in BVE^Cyp24a1-null mice, whereas a 19-fold reduction was found in nude mice following transplantation of BVE^Cyp24a1-null cells. This was probably due to elevated serum TSH in the BVE^Cyp24a1-null mice, which could neutralize the beneficial effects of Cyp24a1 knockout (27). One should bear in mind that these tumors were subcutaneous and they were not in a similar microenvironment compared with the original tumors. Although tumor growth in nude mice could be affected by locally produced calcitriol, it is unlikely a major contributor for tumor regression given that we did not observe significant reduction of cell proliferation in Cyp24-null cells following calcitriol stimulation. The current study confirmed that Cyp24a1-knockout could not reverse the hypothyroidism induced by the Braf^V600E mutation. It is known that TSH stimulates thyroid cancer growth and progression, and higher serum TSH is associated with both thyroid cancer incidence and recurrence (45, 46). We have shown previously that chronic TSH stimulation leads to significant hyperplasia and goiter formation, and promotes Kras^G12D-mediated oncogenic transformation of thyroid follicular cells (34). Due to the high mortality caused by impaired vitamin D catabolism, we could not evaluate survival in the BVE^Cyp24a1-null mice. This needs to be performed in a mouse model with conditional Cyp24a1 knockout in the thyroid.

The antiproliferative effects of calcitriol were not demonstrated in the BVE^Cyp24a1-null cells. This is likely due to the presence of the Braf^V600E mutation, which antagonizes the antiproliferative effects of calcitriol. These effects were clearly shown when combined with the BRAF^V600E inhibitor PLX4720. The synergistic effects of calcitriol and PLX4720 have been reported in human thyroid cancer cell lines (23). Given that vemurafenib has recently been approved for the treatment of advanced PTC harboring BRAF^V600E mutation, combined treatment may offer better therapeutic outcomes for advanced PTC. A clinical trial may be warranted to test the efficacy of the combined therapy.

Finally, we have uncovered many DEGs that are associated with tumor growth or metastasis by RNA-Seq analysis, whose expression is significantly impacted by Cyp24a1 knockout. Most of them are non–vitamin D-responsive genes. Vitamin D–responsive genes or their signaling pathways are clearly involved in the antitumor synergy with PLX4720, even though they may not play a major role in the BVE^Cyp24a1-null tumor regression. Because the expression of many non–vitamin D-responsive genes regulating different cellular processes are impacted by Cyp24a1 knockout, the alteration of these gene is likely to cause downregulation of multiple signaling pathways and the loss of EMT, resulting in tumor regression. Such diverse impacts exerted by Cyp24a1-knockout indicate that Cyp24a1 is a good candidate for targeted cancer therapy.

No potential conflicts of interest were disclosed.

Conception and design: M. Zou, E.Y. Baitei, R.S. Parhar, Y. Shi

Development of methodology: M. Zou, E.Y. Baitei, R.S. Parhar, R. St-Arnaud, C. Pritchard, A.S. Alzahrani, Y. Shi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Zou, E.Y. Baitei, H.A. BinEssa, R.S. Parhar, A.M. Assiri, Y. Shi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Zou, E.Y. Baitei, H.A. BinEssa, F.A. Al-Mohanna, R.S. Parhar, B.F. Meyer, Y. Shi

Writing, review, and/or revision of the manuscript: M. Zou, F.A. Al-Mohanna, R.S. Parhar, S. Kimura, A.S. Alzahrani, A.M. Assiri, B.F. Meyer, Y. Shi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E.Y. Baitei, H.A. BinEssa, R.S. Parhar, S. Kimura, C. Pritchard, A.M. Assiri, Y. Shi

Study supervision: R.S. Parhar, Y. Shi

Other (overall discussion and ongoing supervision): R.S. Parhar, Y. Shi

We would like to thank Ms. Roua A Al-Rijjal and Mr. Wilfredo Antiquera for excellent technical support; Mr. Cong Li and Kai Huang from BGI for bioinformatics service.

This study is supported by KACST grant 13-MED1765-20.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.