Cholesterol and 27-hydroxycholesterol promote thyroid carcinoma aggressiveness

Cholesterol mediates its proliferative and metastatic effects via the metabolite 27-hydroxycholesterol (27-HC), at least in breast and endometrial cancer. We determined the serum lipoprotein profile, intratumoral cholesterol and 27-HC levels in a cohort of patients with well-differentiated papillary thyroid carcinoma (PTC; low/intermediate and high risk), advanced thyroid cancers (poorly differentiated, PDTC and anaplastic thyroid carcinoma, ATC) and benign thyroid tumors, as well as the expression of genes involved in cholesterol metabolism. We investigated the gene expression profile, cellular proliferation, and migration in Nthy-ori 3.1 and CAL-62 cell lines loaded with human low-density lipoprotein (LDL). Patients with more aggressive tumors (high-risk PTC and PDTC/ATC) showed a decrease in blood LDL cholesterol and apolipoprotein B. These changes were associated with an increase in the expression of the thyroid’s LDL receptor, whereas 3-hydroxy-3-methylglutaryl-CoA reductase and 25-hydroxycholesterol 7-alpha-hydroxylase were downregulated, with an intratumoral increase of the 27-HC metabolite. Furthermore, LDL promoted proliferation in both the Nthy-ori 3.1 and CAL-62 thyroid cellular models, but only in ATC cells was its cellular migration increased significantly. We conclude that cholesterol and intratumoral accumulation of 27-HC promote the aggressive behavior process of PTC. Targeting cholesterol metabolism could be a new therapeutic strategy in thyroid tumors with poor prognosis.


Results
Low LDL-cholesterol levels in patients with high-risk PTC and dedifferentiated tumors. The serum lipid profiles of patients with different thyroid neoplasms are summarized in Table 1. Patients with highrisk PTC and PDTC/ATC tumors showed lower serum cholesterol levels (3.832 ± 0.198 mmo/L, p = 0.0041 and 3.793 ± 0.245 mmo/L, p = 0.0293, respectively) compared with those with BTT tumors (4.933 ± 0.207 mmo/L) and low/intermediate-risk PTC (4.581 ± 0.159 mmo/L). These changes were associated with a decrease in LDL cholesterol in high-risk PTC tumor patients (1.944 ± 0.222 mmo/L, p = 0.0515) and those with dedifferentiated PDTC/ATC tumors (1.775 ± 0.083 mmo/L, p = 0.0274) compared with those with BTT (2.551 ± 0.165 mmo/L) and low/intermediate-risk patients PTC (2.293 ± 0.129 mmo/L). The apoB levels were also decreased in patients with high-risk PTC tumors (0.748 ± 0.07 mmo/L, p = 0.0068) and PDTC/ATC (0.714 ± 0.064 mmo/L, p = 0.011). However, the HDL cholesterol levels did not differ among the groups analyzed. Moreover, the triglyceride levels were reduced in patients with all the types of malignant tumors, but significant changes were only found in the high-risk PTC group (0.824 ± 0.101 mmo/L, p = 0.0015).
The LDL receptor (LDLR) is upregulated in tumors with higher malignancy. The gene expression of the LDLR (specific receptor-mediated endocytosis of LDL) was upregulated in the PTC tumors with greater malignancy (high-risk PTC and PDTC/ATC) compared with the tumors with a minor degree of malignancy (low/intermediate-risk PTC and BTT; (Fig. 1). However, the tumorigenic-free cholesterol levels, esterified cholesterol, and triglycerides did not differ among all the groups analyzed ( Supplementary Fig. S1). Moreover, the LDL uptake analysis in primary cell cultures displayed that the LDL-derived cholesterol uptake in the PDTC tumors was higher (13 ± 1.4%) than that of high-risk PTC and BTT tumorigenic cells (8.9 ± 0.3% and 8.4 ± 0.1%, respectively, p ≤ 0.01 vs PDTC) ( Supplementary Fig. S2).
HMGCR and CYP7B1 are strongly downregulated in PDTC and ATC tumors compared with BTT. The Fig. 2A). However, no differences were found between PTC with different degrees of malignancy. A similar expression profile was observed in the CYP7B1 gene, which was also downregulated in all tumors, especially dedifferentiated PDTC/ATC (p < 0.001). The comparison between the malignant tumors exhibited a significant difference between the PTC (low/intermediate risk) and PDTC/ATC samples (p < 0.01; Fig. 2B). The gene expression analysis of the LXR alpha gene also displayed a similar gene expression pattern, although the downregulation detected was higher and significant in all groups analyzed (Fig. 2C). In contrast, the CYP27A1 gene expression did not differ significantly among the groups (Fig. 2D). The analysis of 24-Dehydrocholesterol reductase (24DHCR) and the scavenger receptor class B type 1 (SCARB1) genes showed that both genes tended to be downregulated when compared the most aggressive tumors with BTT group, although these changes were not significantly different (Supplementary Table 1).
To investigate whether LDL can produce similar effects in vitro to those found in vivo, two thyroid human cell lines with different genotypes (Nthy-ori 3.1 and CAL-62) were exposed to human LDL and the gene expression profile was analyzed. The comparison of both cell lines did not reveal significant changes in LDLR and CYP27A1 gene expression (Fig. 3A,C). However, a downregulation in HMGCR and CYP7B1 was detected in CAL-62 in comparison with Nthy-ori 3.1 (Fig. 3B,D). PDTC/ATC tumors show higher levels of 27-hydroxycholesterol. The percentage of tumors with detectable levels of 27-HC was higher in the malignant tumors analyzed, with 20% in the low/intermediate-risk PTC samples (6 positive of the 30 total samples), with a mean of 3.73 ng/mg protein; 44% in the high-risk PTC samples (8 positive of 18 total samples), with a mean of 10.07 ng/mg protein; and 89% in the PDTC/ATC (8 positive of 9 total samples), with a mean of 89.42 ng/mg protein. In contrast, the percentage of detection in the BTT samples was 5% (2 of 40 total samples), with a mean of 16.38 ng/mg protein (Fig. 4A). Moreover, we observed a negative correlation between the CYP7B1 gene expression and the 27-HC concentration, showing higher levels of this sterol in the most aggressive samples analyzed (Fig. 4B).
LDL cholesterol promotes proliferation and migration in anaplastic thyroid cell lines without changing in MAPK, PI3K and mTOR signaling pathways. The Nthy-ori 3.1 cell line, derived from normal human primary thyroid follicular epithelial cells, was immortalized with a plasmid containing an origin-defective SV40 genome (SV-ori). The CAL-62 cell line, derived from a human thyroid anaplastic carcinoma (ATC), is an epithelial-like cells stabilized in culture. These cells do not show thyroglobulin expression, as well as, do not present any BRAF mutation 29 . LDL promoted a dose-related response in cellular proliferation in both Nthy-ori 3.1 and CAL-62 (p < 0.01), without significance between them (Fig. 5A). Nevertheless, the evaluation of wound healing capacity showed that LDL cholesterol promoted cell mobility in the CAL-62 cell line in a dose-related response; in contrast, LDL cholesterol did not induce significant changes in the cell mobility in the Nthy-ori 3.1 cells (Fig. 5B). The concentration of 27-HC measured in the LDL treated cells and non treated cell was 3.73 ± 1.69 ng 27HC/mg prot and 0.37 ± 0.53 ng 27HC/mg prot, respectively (p = 0.01), in CAL-62 cell line. However, the concentration of 27-HC in the Nthy-ori 3.1 cells treated with LDL was lower in comparison with CAL-62 cell line (0.7 ± 0.25 ng 27-HC/mg prot) and it was undetectable in non treated cell. www.nature.com/scientificreports www.nature.com/scientificreports/ Finally, the effect of the LDL in the MAPK, PI3K and mTOR signaling was also investigated. A slight dose-dependent decrease of the P-Akt and ERK was found in both cell lines ( Supplementary Fig. S3). CYP7B1 overexpression and CYP27A1 silencing can modulate cellular proliferation rate. The overexpression of CYP7B1 in CAL-62 cell line treated with LDL arrested growth (p = 0.001) and decreased cellular migration with (p = 0.02) and without LDL treatment (p = 0.02) (Fig. 6A,B); nonetheless, Nthy-ori 3.1 cell line treated with 27-HC at 6 and 12 µM decreased the proliferation rate in both concentrations (p = 0.001), but promoted the migration rate (p = 0.001) without evidence of a dose-related response (Fig. 6C,D).
Furthermore, downregulation of CYP27A1 completely blocked LDL-mediated cell proliferation ( Supplementary Fig. S4). In contrast, DHEA did not promote proliferation in CAL-62 cell line, whereas a moderate arrest growth was detected in a doses dependent manner in Nthy-ori 3.1 cell line ( Supplementary Fig. S5).

Discussion
The role of cholesterol in cancer development is controversial; notwithstanding, several epidemiological studies have reported an association between serum cholesterol levels and a higher incidence of some types of solid tumors 18,26,30 . In this study, the patients bearing tumors with more aggressive behavior showed a significant decrease in serum LDL-cholesterol and apoB, without changing the HDL-cholesterol levels. These data are in consonance with different reports highlighting the importance of the uptake of systemic cholesterol as a mechanism modulating the tumoral process 31 . According to these data, we also detected overexpression of the LDLR in the subtypes of thyroid tumors with a more aggressive behavior. These data reinforce the hypothesis of these subtypes of tumors' higher systemic cholesterol uptake capacity in comparison with benign tumors. There is evidence in tumors that has demonstrated the role of LDL-cholesterol on the cell migration and metastasis 3,32 . The promotion of the migration process after incubation with LDL cholesterol in cells could support the idea of cholesterol's role in the tumor's metastatic promotion. On the other hand, different signaling pathways have been proposed as a mediator  www.nature.com/scientificreports www.nature.com/scientificreports/ in the control of development, progression, invasion and metastasis in thyroid cancer. Some of them, such as mitogen-activated protein kinase (MAPK), phosphoinositide 3-kinase (PI3K)/Akt, mTOR, has a strong relation with BRAF gene mutation, which are frequent in thyroid cancers 33 . The protein expression analyses did not show any change after LDL exposition in both CAL62 (wild-type BRAF and G12R KRAS mutation) and Nthy-ori 3.1 cell lines (no mutation has been reported). However, strong constitutive activation of these signaling transduction cascades under basal conditions were detected, indicating that the mechanisms by which LDL act are independent of these pathways. Additional cell lines harboring different oncogenic drivers should be investigated. In this context, SCARB1, which has been involved in tumor development 34,35 , tended to be downregulated when compared the most aggressive tumors with BTT group, although these changes were not significantly different. This result rather indicates that this receptor seems no to be associated with the tumor's aggressiveness.
The regulation of the cholesterol metabolism in tumoral cells is complex. It is known that cancer cells have the capacity to increase intracholesterol via a feedback regulation mechanism of the HMGCR enzyme, which controls de novo cholesterol synthesis 36,37 . Importantly, the expression of this gene was downregulated in association with the degree of malignancies of the tumors. These changes may be related with similar free cholesterol levels found in the more aggressive tumors compared with those of benign tumors. It is noteworthy that the analyses in normal and ATC cell lines revealed a gene expression pattern like that found in the thyroid tumors from the patients. In contrast, one of the metabolites of cholesterol with greater relevance in both inflammatory and tumor processes is 27-HC, the most abundant oxysterol in the systemic circulation. In 2013, Nelson et al. demonstrated that 27-HC acted as an estrogen receptor agonist in breast cancer, inducing tumor growth and metastasis. These effects were reproduced in mice by overexpressing the CYP27A1 gene, which regulates 27-HC synthesis 18 . In contrast, decreased expression of CYP7B1 triggers accumulation of 27-HC, and CYP7B1 was downregulated in breast cancer compared with normal breast tissue 19,20 . In line with these findings, CYP7B1 was strongly downregulated in the aggressive tumor tissues (PTC high risk and PDTC/ATC) in relation to the benign tumors, in close association with the higher concentration of 27-HC. These data indicate that the accumulation of 27-HC in the thyroid cells may be promoting the development and progression of TC. In fact, downregulation of CYP7B1 www.nature.com/scientificreports www.nature.com/scientificreports/ was also detected in the anaplastic cell line (CAL-62) and its overexpression reduced cell growth and migration. Furthermore, downregulation of CYP27A1 completely blocked LDL-mediated cell proliferation. However, 27-HC by itself only promoted cell migration in non tumoral cells. We also treated the cells with DHEA, which may be hydroxilated by CYP7B1 38 and it did not produce any effect in the cells with anaplastic phenotype. Nevertheless, in non-cancer cell (Nthy-ori 3.1 cells) we found a doses dependent decline in cell proliferation, as occurred with 27-HC. These results rather indicate a dual effect of 27-HC in cells with aggressive behaviors in contrast with immortalized cells with non tumoral phenotype. It should be noted that 27-HC increased ROS levels and reduced the antioxidant defense system levels in non pathological astrocytes, thereby affecting cell viability 39 . In addition, 27-HC may downregulate the expression of the nuclear factor E2-related factor 2 signaling pathways 40 . Moreover, in immortalized retinal pigment epithelial cell line (RPE cells), 27-HC also caused glutathione depletion, ROS generation, inflammation and apoptotic-mediated cell death 41 .
It should be noted that 27-HC is also involved in the regulation of the nuclear receptor LXR expression. This gene, which controls the cell cycle progression, is also implicated in the mechanism for controlling the metastasis of the tumors of epithelial lineage 42,43 . Moreover, the LXR gene controls the mechanism to promote liver LDLR degradation and cholesterol conversion to bile acids via CYP7A1 44,45 . The downregulation of LXR could also be directly affecting the LDLR levels in the aggressive tumors. The DHCR24, which catalyzes the conversion of desmosterol into cholesterol and is considered a tumoral biomarker involved in proliferation, adhesion, cell migration and apoptosis 46,47 , tended to be downregulated when compared the most aggressive tumors with BTT group, although these changes were not significantly different. Further investigation is needed to establish the role of desmosterol on thyroid cancer.
The expression of ATP Binding Cassette Subfamily A Member 1 (ABCA1), the main cholesterol efflux transporter from cells to HDL, was also determined. ABCA1 was upregulated in the most aggressive samples although this change was not significant in the PDTC group (Supplementary Table 1). We have recently reviewed the www.nature.com/scientificreports www.nature.com/scientificreports/ potential role of ABCA1 on tumor development. ABCA1 regulates plasma membrane cholesterol content and fluidity which are critical determinants of its metastatic capacity. However, ABCA1 also inhibits the development of tumors by inhibiting cellular proliferation thereby highlighting the dual role of ABCA1 in regulating proliferation and metastasis 48 . Further studies are needed to establish the role of ABCA1 on thyroid cancer.
Several studies have shown the relationship between the levels of 27-HC and the expression of estrogen receptors in the breast tumors, demonstrating that this oxysterol is a selective modulator of them, having a role as an agonist or antagonist in the function of the cellular type 18,19 . In relation to epithelial-type thyroid cancer, these have a higher incidence in females. This characteristic probably suggests its relationship with the expression of estrogen receptors in the follicular thyroid epithelium. There are not enough reported data identifying the expression of these sexual steroid receptors and their relationship with the neoplastic follicular cells 49 . We explored the expression of the estrogen receptors (alpha and beta) in a series of samples, and the results showed that both benign and malignant tumors express both types of receptors. However, their expression was observed in undifferentiated tumors (Supplementary Table S2). These data are consistent with those observed in breast cancer, where estrogen receptor (ER)-negative tumors tend to grow faster and have a worse prognosis, whereas a lack of ER isoforms β expression would lead to a decrease in apoptosis 50,51 . This result contrasted with the potential effect of 27-HC as an endogenous ER ligand that promotes ER-positive breast tumor growth 19 . In contrast, 27-HC treatment reduced cell growth in Nthy-ori3.1 cells (ERα/ß negative). This effect is in agreement with the reported in prostate cancer cells in which 27-HC reduced intracellular cholesterol accumulation independent of the androgen receptors status 52 . These results suggest that 27-HC could promote a negative feedback control of cholesterol biosynthesis in normal immortalized follicular cells. Studies of the potential effects of 27-HC on ER-independent tumorigenic processes are warranted.
Of note, some experimental studies have shown the antineoplastic efficacy of statins but statins were associated with thyroid carcinoma in some clinical studies 53 . Furthermore, hepatic PCSK9 expression is modified in several diseases, including in thyroid cancer 54 : however, no large genetic studies and controlled cholesterol-lowering clinical trials have been carried out yet. In summary, our findings showed that cholesterol could be a key molecule in the thyroid carcinogenesis process, especially in terms of 27-OH, which may influence the development and progression of TC (Fig. 7). Therefore, targeting the transport of cholesterol and synthesis of 27-HC may be a therapeutic strategy for controlling TC development in the future. The causal association of cholesterol/27-HC and TC should be further investigated.

Material and Methods
Data sharing. The data, analytical methods, and study materials will be available to other researchers for purposes of reproducing the results or replicating the procedure upon reasonably request.
Patients. This study included a cohort of patients (n = 89) with a diagnosis of TC who were recruited at Hospital de la Santa Creu i Sant Pau (HSCSP) in Barcelona, Spain. In these patients, the following clinical information was collected: sex, age, unhealthy habits, no dyslipidemic pathology or statin treatment, and baseline TSH values (0.5-5 mlU/L). The tumors were classified according to the histopathological diagnosis, as follows: benign thyroid tumor (BTT), well-differentiated malignancies (PTC), and advanced thyroid cancers which include poorly differentiated and anaplastic thyroid carcinoma (PDTC/ATC). The AJCC/UICC TNM staging system (TNM) was used to subdivide the patients into two groups (low/intermediate and high risk; NCCN Guidelines Version 1. 2018, Staging Thyroid Carcinoma; Table 2). This study was approved by the ethics committee at HSCSP, and all patients gave written informed consent to participate. All research was performed in accordance with relevant guidelines/regulations. Human tissue and serum samples. Surgical snap-frozen specimens and serum were obtained from patients recruited from January 2009 at HSCSP, as described above. These belonged to the sample collection registered at Instituto de Salut Carlos III (Spain; C.000281).
Lipid, lipoprotein, and apolipoprotein determinations. Serum lipid analyses were determined enzymatically using commercial kits adapted to a COBAS c501 autoanalyzer (Roche Diagnostics, Rotkreuz, Switzerland) 55 . High-density lipoprotein (HDL) cholesterol was measured in apoB-depleted serum obtained after precipitation with phosphotungstic acid and magnesium ions (Roche Diagnostics). The low-density lipoprotein (LDL) cholesterol was determined by the Friedewald formula 56 . The apoA-I and apoB concentrations were determined using nephelometric commercial kits (Roche Diagnostics) 55 .

Thyroid 27-hydroxycholesterol (27-HC) levels.
Frozen thyroid samples (n = 66) included in optimal cutting temperature (OCT) compound (Tissue-Tek, Sakura Europe, Alphen den Rijn, Netherlands) were sectioned and homogenized in 1 M NaOH. Protein concentration was measured using the BCA protein assay kit (Thermo Scientific S.L.U., Spain). Lipids were then extracted as described above, resuspended in 100 μl of methanol, and transferred to a vial with 6 μl of deuterium-labeled internal standard. The 27-HC levels were measured using high-performance liquid chromatography tandem mass spectrometry (HPLC-MS/MS) 57 . www.nature.com/scientificreports www.nature.com/scientificreports/ PBS Buffer, minced into small fragments and digested with collagenase type II (1 mg/mL, Gibco Invitrogen) in stir at 37 C for 1-3 h. The reaction was stopped by adding 10 mL of culture medium at room temperature and the digested tissues were centrifuged 1000 g for 10 min and washed two times in PBS Buffer salt solution. The cellular pellets were suspended with DMEM-F12 medium, transferred in cellular plates, and the medium was changed every two days. The cultures were maintained only six days. Finally, the cells were trypsinized, counted using an automated cell counter (TC10 Bio-Rad) and cell blocks with 10 5 -10 6 cells were prepared by plasma-thrombin method. In order to check cell thyroid purity, thyroglobulin immunostaining with sections from cell blocks was done by peroxidase-anti-peroxidase (PAP) technique.

Human lipoproteins isolation: Low-density lipoprotein (LDL) and high-density lipoprotein
LDL particles (1.019-1.063 kg/L) were isolated by sequential ultracentrifugation at 100,000 g for 24 h from normolipidemic human plasma obtained in EDTA-containing vacutainer tubes and radiolabeled with [1,2-3 H] cholesteryl oleate (Perkin-Elmer), as previously reported 58 . Subsequently, radiolabeled LDL (0.45 μCi [1,2-3 H] cholesterol oleate/50 µg LDL-apoB) was dialyzed against phosphate-buffered saline (PBS) by gel filtration chromatography and incubated with 10 5 cells in DMEN-F12 supplemented with 5% LPDS for 16 h. Then, the medium was collected and centrifuged at 10,000 g for 10 min, and the cells were lysed with 0.1 M NaOH. The radioactivity was determined in the medium and cell lysates. [1,[2][3] H] LDL cholesterol uptake was expressed as the percentage of the radioactivity collected in the cells relative to the sum of radioactivity in cells and medium per µg of protein per cell. To determine the cell viability rate, we used the thiazolyl blue tetrazolium bromide metabolic assay (MTT, Sigma, St. Louis, MO, USA). Absorbance was determined at 560 nm with a microplate reader (xMark, BIORAD). All conditions were performed in five replicates. Each experiment was repeated at least three times.
To determine cell migration, both cell lines (Nthy-ori 3.1 and CAL-62) were plated at high densities and grown to confluence at 90% O/N. Cells were scratched with a pipette tip (10 µl) and washed several times to remove the cellular debris.
The wounds were photographed in an inverted microscope at 0 h (t = 0), and this was done again in the same area after 16 h of cell incubation at 37 °C. The cultures were allowed to incubate in serum-free medium. Image J software was used to analyze the photographs. Here, the percentage of wound healing was determined based on three measurements of the wound area; each result was the mean of three independent experiments.  www.nature.com/scientificreports www.nature.com/scientificreports/ Transient transfection assay. Overexpression of CYP7B1 gene in CAL-62 cell line. Transfection in CAL-62 cell line was performed by JetPrime ™ (Polyplus transfection) according to the manufacturer's instructions.
Approximately 10,000 cells were plated in 96-well plates one day prior to transfection. The cells were transiently transfected with either 0.1 nM pre-designed CYP7B1 (NM_004820.3) DNA ORF clone or control (pcD-NA3.1 + /C-(K)DYK) purchased from GeneScript (OHu27221) and treated with LDL 200 µM or without LDL as a control. After incubation for 48 h, the transfected cells were tested for MTT assay. For wound-healing assays, 50 × 10 3 cells were plated in 4-well plates, transiently transfected with the same concentration described above and the wound-healing assays were performed as described below. Each assay was performed in triplicate in at least three independent experiments. Approximately 4,000 cells were plated in 96-well plates one day prior to transfection. The following day, the cells were transiently transfected with either 5 nM pre-designed short interfering RNA (siRNA CYP27A1) (ID106239) (Ambion, Life Technologies) or control siRNA purchased from Santa Cruz Biotechnology, Inc. (sc-37007). The transfected cells were treated with LDL 200 µM or without LDL as a control. After incubation for 48 h, the transfected cells were tested for MTT assay. Each assay was performed in triplicate in at least three independent experiments. Statistical analysis. Gene expression analysis was performed using the comparative CT (ddCt) method by DataAssist ™ Software, and the results were adjusted as the p-value using the Benjamini-Hochberg false discovery rate (FDR). Genes showing an FDR <0.15 and more than twofold difference in absolute values for both comparisons were considered differentially expressed. One-way analysis of variance (ANOVA) with a Tukey's multiple comparison post-test or Kruskal-Wallis multiple comparison with Dunn's post-test was used to compare differences among the groups. Correlations between variables were analyzed using Pearson's correlation analysis. GraphPad Prism software V5.0 was employed, and values of p < 0.05 were considered statistically significant.