Digitalis-like Compounds Facilitate Non-Medullary Thyroid Cancer Redifferentiation through Intracellular Ca²⁺, FOS, and Autophagy-Dependent Pathways Free

Patients diagnosed with non-medullary thyroid cancer generally have a good prognosis as a result of implementing treatment regimens consisting of thyroidectomy and ¹³¹I radioactive iodide (RAI) ablation. In contrast, treatment success and overall survival of patients with RAI-refractory thyroid cancer is poor as no other efficient treatments are available, leading to recurrences, persistent disease, and thyroid cancer–related mortality. Mechanistically, it is well established that RAI resistance is accompanied by severe tumor dedifferentiation (1, 2). Strategies for restoring thyroid-specific gene expression, designated as redifferentiation, are therefore considered as a promising treatment modality to improve clinical responses to RAI treatment.

Autophagy is a ubiquitously expressed degradation machinery for recycling of intracellular content such as cytoplasmic organelles, proteins, and macromolecules (3). In addition, autophagy plays important roles in cancer initiation and progression by influencing proliferation, differentiation, and anticancer therapy resistance (4, 5). Interestingly, loss of autophagy activity was shown to be associated with thyroid cancer dedifferentiation and reduced clinical response to RAI therapy (6). On the basis of these observations, we hypothesized that pharmacologic activation of autophagy could facilitate restoration of hNIS expression and, hence, could establish reactivation of iodide uptake in RAI-refractory thyroid cancer. To explore this hypothesis, we initiated a screening of pharmacologic compounds for their capacity to induce autophagy and redifferentiation. Selection of autophagy-activating compounds was based on systematic high-throughput screenings that have identified FDA-approved autophagy-activating small molecules, which was demonstrated by a number of standardized readout assays (7, 8). For the current study, we selected 15 small molecules for their effect on thyroid cancer redifferentiation. Furthermore, we have investigated the underlying mechanism of thyroid cancer redifferentiation induced by autophagy activators through complementary approaches at both the cellular and biochemical level.

Thyroid cancer cell lines BC-PAP (papillary, BRAFV600E mutation), FTC133 (follicular, PTEN deficient), and TPC-1 (papillary, RET/PTC rearrangement) were obtained from the sources previously described and were authenticated by short tandem repeat profiling (9). Cell lines FTC133 and TPC-1 were cultured in DMEM (Invitrogen) supplemented with gentamicin (10 μg/mL), pyruvate (10 mmol/L), and 10% FCS (Invitrogen). Cell line BC-PAP was cultured in RPMI medium (Invitrogen) supplemented with gentamicin (10 μg/mL), pyruvate (10 mmol/L), and 10% FCS (Invitrogen). Cells were incubated with DMSO (vehicle control), the mTOR inhibitor rapamycin (Biovision), or with the selected autophagy-activating compounds (Supplementary Table S1) for the indicated time points and concentrations. For all compounds, stock concentrations ranged from 50 to 100 mmol/L. Selection of the compounds and applied concentrations are derived from earlier studies (7, 8). Distribution coefficients (cLogD) at pH = 7.0 for estimating lipophilicity and polarity of DLCs was determined in silico by ChemBioOffice software package. For mRNA expression knockdown of ATF3 and FOS, SMARTpool siRNA oligonucleotides targeting scramble, ATF3, or FOS (Thermo Fisher Scientific) were transfected with FuGene HD (Promega) according to the manufacturer's instructions.

Thyroid cancer cell lines were treated with TRIzol reagent (Invitrogen), and total RNA purification was performed according to the manufacturer's instructions. Isolated RNA was subsequently transcribed into cDNA using iScript cDNA synthesis kit (Bio- Rad Laboratories) followed by quantitative PCR using the SYBR Green method (Life Technologies). The following primers were used: hNIS (SLC5A5) forward 5′-TCC-TGT-CCA-CCG-GAA-TTA-TCT-3′ and reverse 5′-ACG-ACC-TGG-AAC-ACA-TCA-GTC-3′; ATF3 forward 5′-CCT-CTG-CGC-TGG-AAT-CAG-TC-3′ and reverse 5′-TTC-TTT-CTC-GTC-GCC-TCT-TTT-T-3′; FOS forward 5′-GGG-GCA-AGG-TGG-AAC-AGT-TAT-3′ and reverse 5′-CCG-CTT-GGA-GTG-TAT-CAG-TCA-3′; JUN forward 5′-TCC-AAG-TGC-CGA-AAA-AGG-AAG-3′ and reverse 5′-CGA-GTT-CTG-AGC-TTT-CAA-GGT-3′; TTF1 forward 5′-AGC-ACA-CGA-CTC-CGT-TCT-C-3′ and reverse 5′-GCC-CAC-TTT-CTT-GTA-GCT-TTC-C-3′, TTF2 forward 5′-CAC-GGT-GGA-CTT-CTA-CGG-G-3′ and reverse 5′-GGA-CAC-GAA-CCG-ATC-TAT-CCC-3′; and PAX8 forward 5′-AGT-CAC-CCC-AGT-CGG-ATT-C-3′ and reverse 5′-CTG-CTC-TGT-GAG-TCA-ATG-CTT-A-3′. Data were corrected for expression of the housekeeping gene β2 microglobulin, for which the primers forward, 5′-ATG-AGT-ATG-CCT-GCC-GTG-TG-3′, and reverse, 5′-CCA-AAT-GCG-GCA-TCT-TCA-AAC-3′, were used.

For Western blotting of hNIS protein, 5 × 10⁶ cells were lysed in 40 μL of lysis buffer [50 mmol/L Tris (pH 7.4), 150 mmol/L NaCl, 2 mmol/L EDTA, 2 mmol/L EGTA, 10% glycerol, 1% Triton X-100, 40 mmol/L β-glycerophosphate, 50 mmol/L sodium fluoride, 200 mmol/L sodium vanadate, 10 mg/mL leupeptin, 10 mg/mL aprotinin, 1 mmol/L pepstatin A, and 1 mmol/L phenylmethylsulfonyl fluoride]. The homogenate was frozen and then thawed and centrifuged at 4°C for 3 minutes at 14,000 × g, and the supernatant was mixed with a loading buffer containing dithiothreitol, incubated at 37°C for 30 minutes (no boiling), and taken for Western blot analysis. Equal amounts of protein were subjected to SDS-PAGE using 10% polyacrylamide gels. After SDS-PAGE, proteins were transferred to nitrocellulose membrane (0.2 mm). The membrane was blocked with 5% (w/v) milk powder in TBS/Tween 20 for 1 hour at room temperature, followed by incubation overnight at 4°C with an hNIS antibody (1:500, 250552; Abbiotec) in 5% milk powder in TBS/Tween 20 or with an actin antibody (loading control, 1:1,000, A2066; Sigma) in 5% milk powder in TBS/Tween 20. After overnight incubation, the blots were washed three times with TBS/Tween 20 and then incubated with horseradish peroxidase–conjugated swine anti-rabbit antibody at a dilution of 1:5,000 in 5% (w/v) milk powder in TBS/Tween 20 for 1 hour at room temperature. After being washed three times with TBS/Tween 20, the blots were developed with enhanced chemiluminescence (GE Healthcare) according to the manufacturer's instructions.

For measurement of proliferation, MTT assays have been performed according to the manufacturer's instructions (Sigma-Aldrich). In brief, cells were plated in 96-well plates with a cell density of 1,000 cells per well. After the indicated time points and treatments, MTT substrate was added and the amount of converted substrate was measured by a plate reader at 570 nm. In addition, at the same time points, activity of lactate dehydrogenase (LDH) has been measured in cell culture supernatants for assessment of cell death in accordance with the manufacturer's protocol (CytoTox 96 kit, Promega).

In vitro [¹²⁵I]-iodide uptake was determined as described previously (10). During 72 hours before the assessment of iodide uptake, cells were cultured with rapamycin (50 μg/mL), digoxin, strophanthin K, lanatoside C, digoxigenin, or proscillaridin A, all in 50 μmol/L concentration. Cells were incubated for 30 minutes with 1 kBq Na¹²⁵I (Perkin-Elmer) and 20 μmol/L nonradioactive NaI as a carrier, with or without 80 μmol/L of sodium perchlorate to control for hNIS-specific uptake. The radioactive medium was aspirated and the cells were washed with ice-cold PBS and lysed in 0.1 mol/L NaOH buffer at room temperature. Radioactivity was measured in the cell lysates in an automatic gamma counter (Wizard², Perkin-Elmer). In parallel experiments, DNA was isolated from the cells by standard procedures (Puregene kit, Gentra Systems) and quantified by Nanodrop measurements (Thermo Fisher Scientific). Accumulated radioactivity was expressed as cpm per nanogram of DNA.

For RNA sequencing analysis, 5 × 10⁶ cells were cultured for either 24, 48, or 72 hours with 50 μmol/L hNIS-inducing DLCs. Thyroid cancer cell lines were treated with TRIzol reagent (Invitrogen) and total RNA purification was performed according to the manufacturer's instructions. Successively, total RNA was subjected to (i) ribosomal RNA depletion by using riboZero (Illumina), (ii) DNAse treatment, (iii) RNA fragmentation, (iv) first- and second-strand cDNA synthesis with random hexamers, (v) library preparation by applying Kapa Hyper Prep (Kapa Biosystems), and (vi) RNA sequencing by Illumina HiSeq 2000. Raw unmapped RNA sequencing data were mapped on the UCSC hg19 human genome assembly and were analyzed by Tuxedo tools TopHat and CuffDiff in the RNA sequencing module on the GenePattern server of Broad Institute (www.genepattern.broadinstitute.org) to analyze differential expression. Processed data were visualized by the Rstudio CummeRbund tool (11). Venn diagrams were constructed by using Venny v2.1 (http://bioinfogp.cnb.csic.es/tools/venny/). Protein networks were built by applying STRING v10 (http://string-db.org/; ref. 12). Raw RNA sequencing data are deposited in the GEO database under accession number GSE79400.

For assessment of intracellular Ca²⁺ accumulation evoked by DLCs, cells were loaded with Fura-2 (Thermo Fisher Scientific, dilution 1:500 in culture medium without phenol red) for 30 minutes, washed, and administered to 50 μmol/L or 5 μmol/L concentrations of DLCs. Immediately after addition of DLCs, cells were consecutively imaged at 340 and 380 nm by a BD Pathway 855 Robotic high content microscope every 30 minutes for a total duration of 7.5 hours. An automatic threshold algorithm was applied, background subtractions were performed, and ratios of 340 nm:380 nm were calculated by FIJI software.

Affinity of thyroid cancer cell lines for Na⁺/K⁺ ATPase–dependent DLC binding was measured by competition assays with [³H]-ouabain (Perkin-Elmer). Cells (5 × 10⁵/well) were seeded in 24-well plates and were simultaneously incubated for 90 minutes at 37°C with 10 nmol/L [³H]-ouabain and different concentrations of DLCs in RM-medium (20 mmol/L Hepes-NaOH, 130 mmol/L NaCl, 0.5 mmol/L MgCl₂, 0.2 mmol/L Na₂HPO₄, 0.4 mmol/L NaH₂PO₄, pH 7.4) supplemented with 1.5 mmol/L sodium vanadate. To control for specific binding to Na⁺/K⁺ ATPases, 5 mmol/L KCl was added. Afterwards, cells were washed with cold RM-medium and lysed in 50% methanol. The amount of [³H]-ouabain radioactivity was measured in cell lysates by a liquid scintillation counter after addition of 4 mL OptiFluor (Canberra Packard). IC₅₀ was determined by calculating the DLC concentration corresponding to 50% of maximal ³H-ouabain–binding capacity that was measured in the absence of unlabeled DLCs.

Statistical significance was tested with Student t test, Mann–Whitney U test, Spearman rank correlation coefficient or by Wilcoxon matched-pairs signed rank test, when appropriate. P values below 0.05 were considered statistically significant. For RNA sequencing data, a false discovery rate of 0.05 was incorporated.

To assess whether autophagy-activating compounds are capable of reactivating hNIS expression in dedifferentiated thyroid cancer cell lines BC-PAP, FTC133, and TPC-1, cells were incubated for 72 hours with DMSO vehicle, 15 different autophagy-activating compounds that were previously identified as such (Supplementary Table S1; ref. 7), or with the mTOR inhibitor rapamycin as positive control for activating hNIS expression (Fig. 1A). Five of 15 compounds were shown to upregulate hNIS in at least one of the cell lines, some only at 50 μmol/L concentrations and some also at 5 μmol/L, all of which are digitalis-like compounds (DLC); four compounds (digoxin, digoxigenin, proscillaridin A, strophantin K) in FTC133, lanatoside C in TPC-1, and proscillaridin A in BC-PAP. Gene expression data were confirmed by Western blotting for detection of hNIS protein, demonstrating similar or higher expression of hNIS after 72 hours, both the inactive 56 kDa as the glycosylated 87 kDa active form, as compared with the positive control rapamycin (Fig. 1B). Functional hNIS expression was assessed by measurement of iodide uptake capacity. After 72 hours of culturing in the presence of DLCs, [¹²⁵I]-iodide uptake was increased comparable with rapamycin and was efficiently inhibited by the hNIS high-affinity substrate sodium perchlorate (Fig. 1C).

Additional sets of experiments were conducted to investigate the effects of DLCs on cell survival, metabolic activity, and proliferation. First, MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) assays were performed to measure metabolic activity of BC-PAP, FTC133, and TPC-1 after administration of DLCs. The results indicate that metabolic activity is completely inhibited upon DLC treatment after 48–72 hours of culture, for most DLCs at and above concentrations of 1 μmol/L (Fig. 2A). Second, to determine the contribution of either cell death or cell-cycle arrest pathways in blocking metabolic activity, lactate dehydrogenase (LDH) activity was measured in the respective cell culture supernatants. Although LDH release was induced to some extent by DLCs in all cell lines, 30%–90% of cells survived after 72 hours of treatment (Fig. 2B). As these surviving cells exhibit inactive metabolism, cell-cycle arrest has likely occurred.

To provide mechanistic insights into the molecular pathways driving hNIS reactivation by DLCs, RNA sequencing of thyroid cancer cell lines was performed after 24, 48, and 72 hours of DLC administration (proscillaridin A for BC-PAP; digoxin, digoxigenin, proscillaridin A, and strophantin K for FTC133; lanatoside C for TPC-1) and was compared with vehicle-treated cells. High quality RNA sequencing data, on average 3 × 10⁷ reads per sample, was obtained as determined by the construction of dispersion, density, and volcano plots. Furthermore, greatly distinct gene expression patterns were observed between vehicle- and DLC-treated cells by performing principal component analysis (Supplementary Figs. 1–6). Total number of identified transcripts and total number of differentially expressed genes reaching statistical significance [based on uncorrected P values ≤0.05 and with expression difference of at least 2 fold (2log ≥ 1 or 2log ≤ −1)] was on average 20,406 and 5,168 for BC-PAP, 19,777 and 5,946 for FTC133, 19,880 and 7,609 for TPC-1, respectively. For subsequent analyses, the mentioned gene sets only containing differentially expressed genes reaching statistical significance were selected. Significantly, differently expressed genes were further analyzed in Venn diagrams to depict overlapping genes between 24-, 48-, and 72-hour time points for over- and underexpressed genes separately and per single cell line–DLC combination (Supplementary Figs. 1–6). Furthermore, heatmaps were constructed to depict the 25 most pronounced up- and downregulated genes per cell line–DLC combination (Fig. 3A–C). To identify common players involved in hNIS reactivation in thyroid cancer cell lines by DLC treatment, genes that were identified as consistently up- or downregulated at all three time points, represented as overlapping centres of Venn diagrams in Supplementary Figures 1–6 comprising on average 508 over- and 900 underexpressed genes per thyroid cancer cell line–DLC combination, were selected for comparison between thyroid cancer cell line–DLC combinations. Before composing overall Venn diagrams comprising all three cell lines, other Venn diagrams were drawn for comparison of FTC133 datasets separately based on treatment with different DLCs (digoxin, digoxigenin, proscillaridin A, and strophantin K). Between these FTC133 datasets, 391 and 496 genes were commonly up- or downregulated, respectively (Fig. 3D). Ultimately, based on total 27 independent transcriptomics datasets, Venn diagrams were built to identify overlapping genes that are differentially expressed in all thyroid cancer cell line–DLC combinations at all 24, 48, and 72-hour time points (Fig. 3D) and resulted in concise gene lists of 128 over- and 125 underexpressed genes that are listed in Supplementary Table S2 and Supplementary Table S3, respectively. To visualize gene networks and to perform Gene Ontology (GO) pathway analysis based on the lists of commonly regulated genes (128 up and 125 down), functional protein–protein interaction networks were generated by applying STRING v10 (http://string-db.org/; ref. 12). These analyses revealed broad protein interaction networks of both up- and downregulated genes (Supplementary Fig. S7) and identified upregulated FOS, ATF3, EGR1, CDKN1A, and autophagy genes and downregulated AKT1 as central molecular players within these networks. Furthermore, GO Biological Processes pathway analysis demonstrated statistically significant enrichment scores (FDR = 0.05) for numerous upregulated pathways, especially those involving autophagy (Supplementary Fig. S8). In contrast, no significantly downregulated GO pathways were identified (data not shown).

For functional validation of RNA sequencing data, additional experiments were performed into the role of the early-response genes FOS and ATF3, as both are consistently upregulated by DLCs in all cell lines and both have previously been recognized as potential drivers of hNIS expression through hNIS upstream enhancer (hNUE) binding in cooperation with the transcription factor PAX8 (13, 14). First, modulation of ATF3 and FOS gene expression has been determined in BC-PAP, FTC133, and TPC-1 by treatment with DLCs for 48 and 72 hours (Fig. 4). As anticipated, in all cell lines, both ATF3 and FOS expression was increased by a number of DLCs. Interestingly, except for the BC-PAP cell line, DLCs capable of upregulating hNIS also most potently induced ATF3 and FOS expression, that is, digoxin, digoxigenin, strophantin K, and proscillaridin A for FTC133 and lanatoside C for TPC-1, also reflected by Spearman rho tests and corresponding P values. Follow-up experiments of ATF3 and FOS gene silencing by siRNA revealed that FOS, but not ATF3, represents the transcription factor driving hNIS expression in all cell lines tested (Fig. 5). On the basis of RNA sequencing data, another hNIS transcription factor, JUN, was the highest upregulated gene in hNIS-positive BC-PAP cells and could therefore be of specific importance as cofactor of FOS for proscillaridin A–induced hNIS expression in BC-PAP. Indeed, JUN expression was increased most potently in BC-PAP treated with proscillaridin A as compared with the other treatment conditions, again corroborated by Spearman rho statistics. In contrast, for FTC133 and TPC-1, no significant correlations were observed between expression of JUN and hNIS (Supplementary Fig. S9). Importantly, expression of thyroid transcription factors TTF1, TTF2, and PAX8 was not modulated by DLC treatment in any of the cell lines (Supplementary Fig. S10).

Pharmacologically, DLCs selectively bind to membranous Na⁺/K⁺ ATPases, thereby inhibiting Na⁺ and K⁺ fluxes across the plasma membrane resulting in intracellular Ca²⁺ accumulation facilitated by the Na⁺/Ca²⁺ exchanger (NCX). To investigate the mechanism underlying differential responses of thyroid cancer cell lines BC-PAP, FTC133, and TPC-1 to the different DLCs tested, intracellular Ca²⁺ concentrations were measured by labeling of cells with the ratiometric fluorescent dye Fura-2 which binds to free intracellular Ca²⁺. After loading cells with Fura-2, cells were exposed to DLCs in 50 μmol/L or 5 μmol/L concentrations for up to 7.5 hours and Fura-2 ratios were determined by measuring fluorescence at 340 and 380 nm every 30 minutes. Significant changes in intracellular Ca²⁺ concentrations were observed by comparing the effects of different DLCs. In FTC133 and TPC-1, DLCs that are capable of upregulating FOS and hNIS induced the highest elevation of intracellular Ca²⁺, whereas this was less increased by the hNIS-negative compounds in the respective cell lines. In BC-PAP, however, intracellular Ca²⁺ was increased at early time points by proscillaridin A and remained stable afterwards, whereas hNIS-negative DLCs evoked increased intracellular Ca²⁺ only at later time points ultimately resulting in higher concentrations of intracellular Ca²⁺ (Fig. 6). These data suggest that in a cell type–specific manner the magnitude of intracellular Ca²⁺ accumulation induced by DLCs determines whether reactivation of hNIS expression occurs.

To further explore the pharmacologic processes responsible for DLC-specific responses in thyroid cancer cell lines, cellular binding affinity of DLCs was assessed by [³H]-ouabain competition assays and was analyzed in relation to Na⁺/K⁺ ATPase expression. Proscillaridin A, digoxigenin, and strophantin K were shown to bind with highest affinity to all thyroid cancer cell lines, whereas for digoxin and lanatoside C, binding affinity was about 10-fold lower (Fig. 7A). Expression of Na⁺/K⁺ ATPase isoforms and subunits, of which only ATP1A1, ATP1B1, ATP1B3, and FXYD5 could be detected, revealed marginal differences in both naïve and DLC-treated cells, although DLC treatment mostly upregulated Na⁺/K⁺ ATPase expression (Fig. 7B). Other players influencing DLC pharmacokinetics are membrane transporters facilitating influx, that is, organic anion transporters such as SLCO4C1, and efflux by ABC transporters including multidrug resistance gene (MDR1/ABCB1) of most, but not all, DLCs. Heatmaps based on RNA sequencing data were constructed to compare expression of organic anion transporters as potential active influx mechanism and ABC transporters representing putative efflux channels. Whereas in FTC133 several ABC transporters are highly expressed and SLC transporters relatively low, in BC-PAP and TPC-1, high expression of SLCO4A1 is observed in vehicle-treated cells (Supplementary Fig. S11). In addition to cellular features that determine DLC efficacy, also physicochemical properties of DLCs could modulate its bioavailability for Na⁺/K⁺ ATPase inhibition. Importantly, between tested DLCs, large differences in lipophilicity and polarity were predicted, with proscillaridin A as most lipophilic and lanatoside C as least lipophilic (Supplementary Table S4).

Mechanisms responsible for dedifferentiation and concomitant development of RAI refractoriness in thyroid cancer patients are incompletely understood, although recent findings indicate that interactions between oncogenic and autophagy pathways are highly relevant (1, 15, 16). Furthermore, molecular targeting of MAPK and mTOR kinases, with autophagy as potential effect modifier (6), have emerged as promising treatment modalities leading to, although partially, restored sensitivity to RAI therapy by inducing redifferentiation (17–19). We therefore hypothesized that autophagy could represent a target for adjunctive therapy to restore RAI-sensitivity in dedifferentiated thyroid cancer. Accordingly, the current study demonstrates that DLCs, a specific class of autophagy-activating agents, induce robust redifferentiation of thyroid cancer cell lines irrespective of the tumor genetic background being either mutated BRAF, PTEN loss, or RET/PTC rearrangement. In all tested thyroid cancer cell lines, at least one of five DLCs restored functional hNIS expression and increased capacity for iodide uptake up to 400-fold. Genome wide transcriptomic profiling by RNA sequencing and functional assays revealed intracellular Ca²⁺-dependent expression of the transcription factor FOS, with in specific conditions JUN as potential cofactor, as prerequisite for functional hNIS expression. In addition, DLCs were confirmed to activate autophagy, were shown to evoke CDKN1A/p21 expression, and to repress AKT1, culminating in cell-cycle arrest and, to some extent, cell death.

DLCs, also termed cardiac glycosides, are well characterized Na⁺/K⁺ ATPase inhibitors that block Na⁺ efflux and, consequently, accelerate passive transport through the Na⁺/Ca²⁺ exchanger (NCX) and promote accumulation of cytoplasmic Ca²⁺ (20). Because of this, DLCs are effective drugs for treatment of cardiac failure or arrhythmia by increasing cardiomyocyte contractility, for which digoxin is still in clinical use (21, 22). Furthermore, regulation of cytoplasmic Ca²⁺, at the level of the plasma membrane as well as intracellular storage compartments, has been previously demonstrated to influence a broad spectrum of cellular processes implicated in both physiologic and pathologic conditions. Pathways of differentiation are among the most prominently influenced mechanisms in this respect, which has been shown for an extensive collection of cell types (23–27). Changes in intracellular Ca²⁺ have also been linked to carcinogenesis, proliferation, and dedifferentiation grade of malignant cells, with autophagy as important pathway involved in processes of UPR and ER stress and in regulation of Ca²⁺ transport between different intracellular Ca²⁺ storage compartments (28–32). DLCs have been extensively investigated as repositioned anticancer treatment based on their antiproliferative and apoptotic effects, selectively induced in malignant cells only, and was revealed to result from perturbations of multiple pathways (reviewed in ref. 33). In addition, systematic high-throughput screening studies have demonstrated DLCs as potentially effective treatment modalities for myelodysplastic syndrome (34) and prostate cancer (35).

In the current study, not all tested DLCs reactivated hNIS expression in all thyroid cancer cell lines, which was demonstrated to be associated with their capacity to sufficiently elevate intracellular Ca²⁺ concentrations and, for BC-PAP specifically, the capacity to enable long-term Ca²⁺ accumulation within a certain range. Pharmacologically, this could potentially be due to differential binding affinity of DLCs to BC-PAP, FTC133, and TPC-1 based on expression profiles of Na⁺/K⁺ ATPase genes or to differences in DLC pharmacokinetics. Although minor differences in ATPase gene expression were present between cell lines, binding affinities for DLCs were highly similar. Pharmacokinetics of DLCs is complex and at least involves a combination of active and passive transport across the cell membrane, with Na⁺/K⁺ ATPase internalization (36), SLCO4C1-mediated influx (37), and ABCB1/MDR1-dependent efflux (38, 39) as active transport mechanisms. Whereas ABCB1/MDR1 and SLCO4C1 are minimally expressed by thyroid cancer cell lines, other ABC transporter homologs (ABCB2/TAP1, ABCB3/TAP2, ABCB7, ABCB8) are especially expressed by FTC133 and the organic anion transporter SLCO4A1 by specifically BC-PAP and TPC-1. These differences could potentially confer differential responses of thyroid cancer cell lines to DLCs by modulating their extracellular concentration in time, although it remains to be proven whether these transporters are involved in DLC transport. In addition, molecular lipophilicity and polarity properties of DLCs are important factors proportionally related to both passive and active transport across the cell membrane, thereby influencing its bioavailability for Na⁺/K⁺ ATPase inhibition (40). By in silico molecular modeling, proscillaridin A was predicted as most lipophilic and lanatoside C as least lipophilic of all tested DLCs. On the basis of these observations, one could envision that variable configurations of DLC lipophilicity combined with relative contributions of active and passive DLC transport, overall determining in vitro DLC bioavailability for Na⁺/K⁺ ATPase inhibition, are responsible for the differential responses of thyroid cancer cell lines to DLC treatment. Differences in in vitro DLC bioavailability also has important implications for the interpretation of DLC dosages applied here, as the biologically available dose for Na⁺/K⁺ ATPase inhibition is probably variable and could differ per DLC and per thyroid cancer cell line.

The proto-oncogene FOS is an important player in the physiologic response of thyroid follicular cells to thyroid-stimulating hormone (TSH), mediating signals of proliferation and thyroid-specific gene expression downstream of the second messenger cyclic adenosine monophosphate (cAMP). For hNIS expression specifically, FOS has been identified as transcription factor binding to the hNIS upstream enhancer (hNUE) in conjunction with PAX8 (14). Accordingly, FOS expression has been associated with benign thyroid neoplasms and differentiated thyroid cancer, suggesting FOS as a marker of the thyroid differentiation state (41, 42). In both physiologic and malignant cell types, it is well established that FOS expression is regulated by its transcription factor cAMP response element-binding protein (CREB) that is phosphorylated under the influence of intracellular Ca²⁺ with calmodulin, cAMP, RAS, and ERK as intermediate factors (illustrated in Fig. 8) (43–45). As opposed to FOS-induced hNIS expression, proliferative effects of FOS are strongly inhibited by DLC treatment. One major explanation for the observed inhibition of proliferation is concomitant activation of autophagy, that is likely to be elicited by DLCs through both overlapping and distinct mechanisms regulated by intracellular Ca²⁺ involving ERK signaling, mTOR inhibition by AMPK and TFEB signaling (illustrated in Fig. 8) enabled by activation of calcineurin, that is also known to inhibit BRAF (46–49). Furthermore, DLCs induce CDKN1A/p21-dependent cell-cycle arrest, confirming previous findings (50–54), and to some extent cell death occurs that could be attributed to either autophagic cell death or apoptosis (55, 56).

Of note, as opposed to the mechanism of hNIS induction by mTOR inhibition (19), TTF1 mRNA expression remained unaffected after DLC treatment, confirming a differential underlying mechanism for restoration of hNIS expression by DLCs through elevation of intracellular Ca²⁺ and FOS expression. This is reinforced by the observation that lanatoside C restores functional hNIS expression in TPC-1 despite its TTF1 deficiency, whereas this cell line remains negative for TTF1-dependent hNIS expression upon mTOR inhibition (19). As opposed to DLCs, the other selected autophagy-activating compounds listed in Supplementary Table S1 were not able to reactivate hNIS expression, indicating that autophagy activation could be necessary but is not sufficient for thyroid cancer redifferentiation. This notion is reinforced by the crucial role of a Ca²⁺-driven second pathway culminating in FOS-dependent thyroid cancer redifferentiation as we have demonstrated for DLCs (Fig. 8), which are the only compounds in Supplementary Table S1 known to directly influence intracellular Ca²⁺ concentrations.

This is the first report to reveal the ability of DLCs to facilitate tumor redifferentiation and, hence, uncovers additional therapeutic benefits of DLCs as anticancer treatment besides their well-established effects on proliferation and cell survival. These findings need confirmation in subsequent studies involving mouse models and human cohorts. Importantly, because of the narrow therapeutic index of DLCs, the dosage and chemical structure of DLCs required for thyroid cancer redifferentiation in vivo in relation to its toxicity remains an important aspect to be addressed. In conclusion, DLCs exhibit multiple beneficial effects as potential treatment of RAI-refractory thyroid cancer by restoring hNIS expression and iodide uptake capacity and by inducing autophagy and cell-cycle arrest, all elicited at least in part by intracellular Ca²⁺ accumulation. Therefore, DLC treatment emerges as a promising adjunctive therapy for patients with RAI-refractory thyroid cancer.

No potential conflicts of interest were disclosed.

Conception and design: M.G. Netea, J.W.A. Smit, R.T. Netea-Maier, T.S. Plantinga

Development of methodology: M.H. Tesselaar, H.G. Swarts, O.C. Boerman, J.W.A. Smit, R.T. Netea-Maier, T.S. Plantinga

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.H. Tesselaar, T. Crezee, D. Gerrits, O.C. Boerman, H.G. Stunnenberg, R.T. Netea-Maier, T.S. Plantinga

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.H. Tesselaar, T. Crezee, H.G. Swarts, D. Gerrits, J.W.A. Smit, R.T. Netea-Maier, T.S. Plantinga

Writing, review, and/or revision of the manuscript: M.H. Tesselaar, O.C. Boerman, M.G. Netea, J.W.A. Smit, R.T. Netea-Maier, T.S. Plantinga

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T.S. Plantinga

Study supervision: J.W.A. Smit, R.T. Netea-Maier, T.S. Plantinga

Other (development and interpretation of Na,K-ATPase/digitalis-like compound interactions assay): J.B. Koenderink

T.S. Plantinga was supported by a Veni grant of the Netherlands Organization for Scientific Research (NWO) and by the Alpe d'HuZes fund of the Dutch Cancer Society (KUN2014-6728).

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.