Targeting Novel Sodium Iodide Symporter Interactors ADP-Ribosylation Factor 4 and Valosin-Containing Protein Enhances Radioiodine Uptake

Since the 1940s, radioiodine (RAI) treatment has been the central post-surgical therapy for patients with differentiated thyroid cancer (DTC) and ablative treatment with RAI is recommended in moderate and high-risk tumors (1). However, at least a quarter of patients with DTC do not uptake sufficient RAI for effective ablation (2, 3), which remains an urgent problem in metastatic disease. There are two cohorts of thyroid cancer patients: those who respond to RAI and have an excellent prognosis (radiosensitive tumors), and those who do not respond (radioresistant tumors) and whose outcome is dire (1). Despite efforts to improve outcomes, no substantial changes have been made to the way RAI is administered therapeutically. Troublingly, DTC is now the most rapidly increasing cancer in the UK and the USA, with 300,000 new cases reported worldwide per annum, and more than 40,000 deaths annually (4).

The sodium iodide symporter (NIS) is the sole human transporter responsible for iodide uptake (5), exploitation of which represents the first—and most specifically targeted—internal radiotherapy in existence. High-energy β-emitting ¹³¹I is utilized to destroy remaining thyroid cells post-surgery, and target metastases. More recently, the interest in NIS has been enhanced due to its use as a novel reporter gene in preclinical and translational imaging systems and in theranostic strategies in nonthyroidal tumors (6, 7). Breast tumors, for instance, can uptake RAI (8), with functional NIS expression confirmed in up to ∼80% of breast cancers (9). However, NIS is rarely localized to the plasma membrane (PM) in breast cancers, limiting its clinical utility (9, 10).

Decreased levels of NIS expression and/or diminished targeting of NIS to the PM represent the principal mechanisms behind radioiodine-refractory thyroid cancer (RAIR-TC; ref. 11). Numerous studies have addressed the common pathways of NIS regulation in vitro and in vivo (12–14), such as key transcriptional and epigenetic alterations, which silence thyroid-specific genes including NIS (or SLC5A5; refs. 15, 16). Clinical approaches to improve treatment of thyroid cancer have involved the use of retinoids (17), PPARγ agonists (18), MAPK pathway/BRAF inhibitors (19, 20), multitargeted kinase inhibitors (21), and histone deacetylase (HDAC) inhibitors (22). Multiple biologically targeted drugs have been evaluated in phase I, II, and III trials, with several agents including sorafenib (21), lenvatinib (23), and dabrafenib (24) showing promising responses and/or disease stabilization. However, issues of toxicity and drug resistance remain.

To actively transport iodide for thyroid hormone biosynthesis and radioisotopes, NIS must be present at the basolateral PM. However, little is known about the mechanisms that govern NIS trafficking. Thyroid-stimulating hormone (TSH) induces iodide uptake through upregulation of NIS expression and modulation of its subcellular localization (25). Yet, many thyroid cancers demonstrate reduced NIS activity through diminished PM retention (26, 27). BRAF-mutant tumors (60%–70% of thyroid cancers) are more likely to be resistant to RAI, partly due to decreased NIS expression (12), but also impaired PM targeting (13, 28), through mechanisms that remain ill-defined. Currently, PTTG1-binding factor (PBF) is the only protein shown to bind NIS and modulate its subcellular localization (29).

Here, we performed mass spectrometry (MS/MS) to identify previously undefined proteins that interact with NIS and regulate its trafficking or retention at the PM. We now report two proteins—ADP-ribosylation factor 4 (ARF4) and valosin-containing protein (VCP)—which specifically bind NIS and directly regulate its function in both breast and thyroid cancer models. Critically, the ability to inhibit VCP activity via FDA-approved drugs indicates a therapeutic possibility of systemically enhancing RAI uptake in patients. This provides new hope that NIS activity may be stimulated in RAIR-TC patients, but also supports the prospect of exploiting NIS trafficking and function to facilitate radioisotope uptake for in vivo imaging and therapy in nonthyroidal tumors.

Breast (MDA-MB-231) and thyroid (8505C, BCPAP, SW1736, TPC-1, and CAL62) cancer cell lines were maintained in RPMI-1640 (Life Technologies), whereas HeLa cervical cancer cells were maintained in DMEM (Sigma-Aldrich). Media were supplemented with 10% fetal bovine serum (FBS), penicillin (10⁵ U/L), and streptomycin (100 mg/L), and cell lines were maintained at 37°C and 5% CO₂ in a humidified environment. Cell lines were obtained from ECACC (HeLa and MDA-MB-231) and DSMZ (8505C and BCPAP), whereas SW1736 and TPC-1 cell lines were kindly provided by Dr. Rebecca Schweppe (University of Colorado, Denver, CO). Cells were cultured at low passage, authenticated by short tandem repeat analysis (NorthGene), and tested for Mycoplasma contamination (EZ-PCR kit; Geneflow).

Stable NIS-expressing MDA-MB-231 and TPC-1 cell lines were generated by lentiviral transduction, as per the manufacturer's instructions. In brief, ready-to-transduce lentiviral particles containing a precision lentiORF construct (pLOC) housing cDNA coding for red fluorescent protein (RFP; OHS5833) or the full-length human NIS cDNA without a stop codon within the open reading frame (ORF; OHS5900-224632369) were purchased from Dharmacon. For lentiviral transduction, the manufacturer's protocol was followed. In brief, 1 day after plating, MDA-MB-231 and TPC-1 cells were infected with the lentiviral vector containing NIS diluted in antibiotic-free and serum-free RPMI containing 8 μg/mL or 14 μg/mL polybrene, respectively. Cells transduced with the RFP-containing lentiviral vector served as the control. After 24 hours, medium was replaced with RPMI containing 10% FBS. After a further 48 hours, cells were maintained in RPMI containing 10% FBS, penicillin (10⁵ U/L), and streptomycin (100 mg/L) as well as 15 or 5 μg/mL blasticidin S for the MDA-MB-231 and TPC-1 cell lines, respectively. Twenty-four hours later, infection efficiency was assessed by fluorescence microscopy. Upon cell expansion and selection of single-cell colonies, NIS expression was assessed by quantitative PCR (qPCR) and Western blotting, with NIS function confirmed via the radioiodine uptake assay.

NIS interactors were isolated by coimmunoprecipitation (co-IP; anti-NIS antibody) and separated by SDS-PAGE. Proteins were reduced, alkylated, and trypsinized. Resulting peptides were separated on an acetonitrile gradient on the UltiMate 3000 HPLC (Thermo Fisher Scientific). Eluted peptides passed through the AmaZon ETD ion trap and tandem mass spectrometer (Bruker). Mass spectra were processed using the Bruker DataAnalysis software and analyzed using the Mascot search engine (Matrix Science). Data sets were filtered based on peptide number, DAVID functional classification (30, 31), and literature review (Supplementary Table S1; refs. 32–36).

Plasmids containing human NIS cDNA with an HA- or MYC-tag have been described (37) and GFP-tagged NIS was kindly provided by Dr. Takahiko Kogai (Dokkyo Medical University, Mibu, Japan). NIS mutants ₄₇₅ALAS₄₇₈ and ₅₇₄AAAK₅₇₇ were generated using the QuikChange Site-directed Mutagenesis Kit (Agilent Technologies). Untagged-ARF4 plasmid was purchased from Origene (#SC119092), whereas wild-type (WT) and mutant (QQ) VCP (rat) plasmid were kindly provided by Dr. Yihong Ye (National Institutes of Health, Bethesda, MD; refs. 11, 12). Human VCP cDNA was generated by site-directed mutagenesis. Further details on nucleic acids and siRNA are provided (Supplementary Tables S2 and S3). Plasmid DNA and siRNA transfections were performed with TransIT-LT1 (Mirus Bio) and Lipofectamine RNAiMAX (Thermo Fisher Scientific) following standard protocols in accordance with the manufacturer's guidelines.

All drugs were resuspended in dimethyl sulfoxide (DMSO), diluted in RPMI, and then added directly to cells at the appropriate final concentration. Cells were treated with VCP inhibitors: Eeyarestatin-1 (ES-1; Cayman Chemicals), NMS-873 (SelleckChem), astemizole (Sigma-Aldrich), clotrimazole (Sigma-Aldrich), and ebastine (Sigma-Aldrich) or the dynamin inhibitor dynasore (Sigma- Aldrich).

Western blotting, coimmunoprecipitation, cell-surface biotinylation assays and RAI (¹²⁵I) uptake assays were performed as described (29, 38) following dynasore treatment (2 hours), VCP inhibitor treatment (24 hours), DNA plasmid transfection (48 hours), or siRNA transfection (72 hours) as indicated.

Blots were probed with specific antibodies (Supplementary Table S4) and NIS expression quantified by densitometry in ImageJ relative to β-actin or Na⁺/K⁺ ATPase, as indicated.

Live-cell highly inclined and laminated optical sheet (HILO) microscopy was performed on an Olympus IX81 inverted fluorescence microscope. Images were acquired every 2 seconds for up to 5 minutes and then integrated at 5 frames/second using the Olympus xCellence build 3554 software (Supplementary Movies S1 to S3). Mean velocity and distance traveled of NIS-GFP was quantified using the TrackMate plugin (ImageJ). The Duolink in situ proximity ligation assay (PLA) was performed according to the manufacturer's instructions (Sigma-Aldrich).

Normalized gene expression data and clinical information (Supplementary Tables S5–S7) for papillary thyroid cancer (PTC) and breast cancer were downloaded from The Cancer Genome Atlas (TCGA) via cBioPortal (39, 40) and FireBrowse (41). In total, RNA-seq data for 501 papillary thyroid cancer (THCA) and 1,093 breast cancer (BRCA) TCGA samples were analyzed. Patient-derived poorly differentiated thyroid cancer (PDTC) and anaplastic thyroid cancer (ATC) samples were selected from the Memorial Sloan Kettering Cancer Center pathology department files from 1986 to 2015 according to the classification outlined in (42). In total, mRNA expression data from fresh-frozen tissue were obtained from 17 PDTC and 20 ATC samples via MSK-IMPACT targeted sequencing. All 37 tumor samples were also expression profiled via the Affymetrix U133 plus 2.0 array and Agilent SurePrint G3 CGH 1 × 1M arrayCGH platform to validate copy-number calls.

Data were analyzed using GraphPad Prism and Microsoft Excel (Supplementary Materials and Methods).

We performed MS/MS in MDA-MB-231 cells with lentivirally expressed NIS (NIS⁺) to identify interactors in whole-cell and PM extracts. NIS and its protein interactors were subjected to in-gel tryptic digest and resulting peptides were fingerprinted using the AmaZon ETD ion trap and tandem mass spectrometer prior to shortlisting (Fig. 1A and B; Supplementary Fig. S1A and S1B). Five proteins with established roles related to subcellular trafficking, endocytosis, PM targeting, and/or endosomal transport were selected (Supplementary Fig. S1A and S1B; Supplementary Table S1) and underwent an initial endoribonuclease-prepared siRNA (esiRNA) screen to investigate their impact upon NIS function (Fig. 1C and D). An important finding was that ARF4 depletion repressed iodine-125 (RAI; ¹²⁵I) uptake in MDA-MB-231 NIS⁺ cells, whereas VCP ablation resulted in a significant induction of RAI uptake (Fig. 1D). As a positive control, esiRNA knockdown of PBF, the only protein known to specifically modulate NIS subcellular localization and function (29), significantly increased ¹²⁵I uptake (Fig. 1C and D). In thyroidal TPC-1 NIS⁺ cells, treatment with ARF4, VCP, or PBF esiRNA also significantly altered RAI uptake (Supplementary Fig. S2A).

ARF4 and VCP were thus considered novel putative functional partners of NIS and further investigated. Subsequent ARF4 ablation in TPC-1 and MDA-MB-231 NIS⁺ cells, as well as in human primary thyroid cells, confirmed a ∼60% to 80% decrease in ¹²⁵I uptake (Fig. 1E) with different siRNA sequences (Supplementary Table S2). In contrast, transient ARF4 overexpression resulted in significantly increased ¹²⁵I uptake in all three cellular settings (Fig. 1F; Supplementary Fig. S2B), suggesting its impact on NIS function is bidirectional. A ∼50% to 80% increase in ¹²⁵I uptake was further validated in VCP-siRNA–depleted TPC-1 and MDA-MB-231 NIS⁺ cells, with human primary thyrocytes demonstrating a ∼35% increase (Fig. 1G; Supplementary Table S2). VCP induction by transient transfection reversed these effects, resulting in markedly repressed ¹²⁵I uptake in all three cell models (Fig. 1H; Supplementary Fig. S2C).

Thus, we identify two novel modulators of NIS function that alter RAI uptake both in thyroid and breast cells lentivirally transduced with NIS, and in human primary thyrocytes with endogenous NIS expression.

Having identified that manipulation of ARF4 and VCP expression altered RAI uptake, we sought to challenge our MS/MS data that indicated specific binding between each protein and NIS. Co-IP confirmed that NIS specifically interacts with ARF4 and VCP in both MDA-MB-231 and TPC-1 NIS⁺ cells (Fig. 2A and B). Control co-IPs further demonstrated specific interaction between NIS with ARF4 and VCP (Supplementary Fig. S3A–S3D). Additionally, PLA demonstrated specific binding between ARF4 and NIS in three different cell types, which appeared to occur generally throughout the cytoplasm with some binding within the endoplasmic reticulum (ER)/Golgi also suggested (Fig. 2C; Supplementary Fig. S3E and S3F). Similarly, PLA confirmed the VCP–NIS interaction, which occurred more generally within the cytoplasm of MDA-MB-231, TPC-1, and HeLa cells (Fig. 2C; Supplementary Fig. S3E and S3F).

We next performed cell-surface biotinylation assays to quantify whether ARF4 and VCP modulate the amount of NIS present at the PM. Exogenous expression of ARF4 in NIS⁺ cells demonstrated an approximate doubling of NIS protein in PM preparations compared with vector only (VO) controls (Fig. 2D and E). In contrast, VCP overexpression resulted in significantly diminished NIS protein PM abundance (Fig. 2F and G). Together, these data suggest that while ARF4 potentiates NIS presence at the PM, VCP inhibits its expression at its key site of symporter activity.

Given that ARF4 and VCP overexpression was associated with altered NIS PM localization, we next investigated their trafficking using HILO microscopy—a technique used to visualize and quantify molecular dynamics in cells. We identified ARF4-dsRED and NIS-GFP trafficking in coincident vesicles at the PM in HeLa cells (Fig. 3A and B; Supplementary Fig. S4A and S4B; Supplementary Movie S1). By contrast, no such cotrafficking was apparent for VCP-dsRED and NIS-GFP, suggesting the site of functional interaction between VCP and NIS was distant to the PM (Supplementary Fig. S4C; Supplementary Movie S2). Of particular significance, and in contrast to VCP, the presence of ARF4 led to an overall induction in both the mean velocity and distance traveled by NIS-GFP-positive vesicles (P < 0.001; Fig. 3C and D; Supplementary Fig. S5A–S5C).

In control experiments, we confirmed that NIS is endosomally trafficked in association with clathrin (Supplementary Fig. S5D; Supplementary Movie S3) and found a significant increase in the overall distance traveled by NIS-GFP in cells overexpressing clathrin (Supplementary Fig. S5A). We then appraised the impact of inhibiting clathrin-mediated endocytosis on NIS function using 100 μmol/L dynasore, a cell-permeable inhibitor of dynamin, which is required for the budding and formation of clathrin-coated vesicles. Treatment of NIS⁺ cells overexpressing ARF4 with dynasore resulted in significantly greater RAI uptake (Fig. 3E) without any effect on total NIS protein levels (Fig. 3F; Supplementary Fig. S5E), suggesting ARF4 is not primarily enhancing NIS recycling to the PM.

ARF4 binds to the terminal VXPX motif of Rhodopsin (43). We identified two putative VXPX motifs within the extracellular loop of NIS at positions 475–478 (sequence VLPS) and the C-terminus at positions 574–577 (sequence VAPK). Abrogation of both motifs (mutations ₄₇₅ALAS₄₇₈ and ₅₇₄AAAK₅₇₇) resulted in NIS proteins that retained endogenous functionality, but the ₅₇₄AAAK₅₇₇ NIS mutant could no longer be augmented in terms of RAI uptake by ARF4 overexpression in TPC-1 thyroid cells (Fig. 3G). Co-IP assays demonstrated that ₄₇₅ALAS₄₇₈-mutant NIS still avidly bound ARF4, whereas ₅₇₄AAAK₅₇₇-mutant NIS lost interaction (Fig. 3H). Hence, NIS has a C-terminal VXPX ARF4 recognition sequence, which is required for ARF4 potentiation of function.

Overall, our results indicate that ARF4 and VCP appear to have different modes of action; ARF4 is implicated in the trafficking of NIS to the PM, whereas VCP binds NIS and modulates NIS function elsewhere in the cell. In support of this, significantly altered total NIS protein levels were apparent in NIS⁺ cells after siRNA ablation or exogenous expression of VCP but not following modulation of ARF4 expression (Supplementary Fig. S6A and S6B).

VCP functions chiefly as a chaperone in disassembling protein complexes and facilitating the extraction and/or proteasomal degradation of proteins from the ER, but is also implicated in a wider range of other cellular actions (44). Several specific VCP inhibitors already exist that target different facets of VCP structure/activity (45). Critically, a significant induction of RAI uptake (Fig. 4A) was evident in NIS⁺ cells treated with the allosteric VCP inhibitor ES-1, which inhibits ER-cytosol dislocation and subsequent degradation of substrates (46). Similarly, the VCP inhibitor NMS-873, which is a potent and specific allosteric inhibitor of VCP (47), yielded a significant 3- to 4-fold increase in RAI uptake in NIS⁺ cells (Fig. 4B).

Importantly, there were no significant reductions in cell viability at the optimal doses of VCP inhibitors used (Supplementary Fig. S6C), nor any changes in VCP protein expression (Fig. 4C). PLA and co-IP assays also revealed that NIS and VCP retained the ability to bind in the presence of VCP inhibitors (Fig. 4D and E). Notably, we also observed a significant induction in RAI uptake after 24 hours of ES-1 and NMS-873 inhibitor treatment on native thyroidal cells (i.e., without lentiviral NIS) with different mutation profiles, including TPC-1 (Fig. 4F and G), CAL-62, 8505C, BCPAP, and SW1736 cells (Supplementary Fig. S6D–S6G). This important finding raises the possibility that VCP inhibitors can be used to induce RAI uptake in thyroid cells that have inherently repressed NIS function.

We next examined the dynamics of therapeutically targeting VCP and found that ES-1 was associated with increased detectable NIS protein after 12 hours, accompanied by a concomitant increase in RAI uptake by 24 hours (Fig. 4H). Similar data were apparent for NMS-873, although RAI uptake and NIS expression were detected at 6 hours posttreatment (Fig. 4I). To explore the dependency of ES-1 on the presence of VCP to modulate RAI uptake, we characterized VCP-ablated TPC-1 NIS⁺ cells, which demonstrated no induction of RAI uptake after ES-1 (Fig. 4J) or NMS-873 treatment (Supplementary Fig. S7A), thus confirming that inhibitors ES-1 and NMS-873 require VCP in order to exert their effects on NIS function.

VCP's canonical function lies in facilitating the extraction of misfolded proteins from the ER, a “segregase” process that generally requires ATPase activity. We next used an ATPase-deficient dominant-negative VCP mutant (QQ VCP) to investigate whether VCP required ATPase activity to modulate NIS function. The QQ VCP mutant behaved identically to WT VCP in repressing RAI uptake when overexpressed in multiple cell models (Fig. 4K) and retained the ability to decrease NIS localization at the PM in cell-surface biotinylation assays (Supplementary Fig. S7B). Additionally, ATP-competitive VCP inhibitors N2,N4-dibenzylquinazoline-2,4-diamine (DBeQ; ref. 48) and sorafenib, which inhibits VCP activity at the PM by blocking its phosphorylation, failed to increase RAI uptake in NIS⁺ cells (Supplementary Fig. S7C and S7D). Collectively, these results suggest that VCP functions in an ATPase-independent manner to affect NIS function, a facet that has been reported in VCP's ability to unfold proteins for subsequent proteasomal degradation (49).

Recently, three new drugs—astemizole, clotrimazole, and ebastine—have been identified as repurposed small molecules that specifically and allosterically inhibit VCP activity (50). Clotrimazole and ebastine are well tolerated in vivo and FDA approved (50). As ES-1 and NMS-873 may have limited clinical utility (51), we investigated whether astemizole, clotrimazole, and ebastine also enhance RAI uptake. All three drugs significantly induced RAI uptake (Fig. 5A) without affecting cell viability (Supplementary Fig. S6C). There was no induction of RAI uptake after treatment of VCP-ablated TPC-1 NIS⁺ cells, thus confirming these VCP inhibitors require VCP in order to exert their effect on NIS function (Fig. 5B and C). Additionally, VCP inhibitors increased NIS expression, particularly at the PM (Fig. 5D and E) but did not alter VCP protein levels (Fig. 5F).

We next progressed to our model of mouse thyroid function. Primary thyrocytes were isolated from C57BL/6 mice and treated with ebastine or clotrimazole; both drugs significantly enhanced RAI uptake and NIS protein expression (Fig. 5G to I). Finally, we tested our drugs in human primary thyroid cultures, which were confirmed TSH responsive. Here, both ebastine and clotrimazole significantly increased RAI uptake (Fig. 5J).

Having identified VCP and ARF4 as novel regulators of NIS function, we appraised their expression profiles and clinical relevance in thyroid cancer using TCGA papillary thyroid cancer (THCA) data set. Of significance, ARF4 was underexpressed in PTC compared with normal tissue (Fig. 6A), whereas VCP was significantly induced in PTC, PDTC, and ATC (Fig. 6A and B). Genetic drivers in PTC have distinct signaling consequences and have been categorized into BRAF-like and RAS-like PTCs according to distinct gene signatures (52). Interestingly, VCP mRNA expression did not differ between BRAF-like and RAS-like PTC, whereas ARF4 mRNA expression was lower in BRAF-like than RAS-like tumors across the THCA series (Fig. 6C).

In agreement with these findings, there was greater reduction of ARF4 expression in BRAF-mutant PTC versus the non–BRAF-mutant tumors (Fig. 6D). The frequency of BRAF alterations was also higher in PTC with low ARF4 expression compared with PTC with high ARF4 (Fig. 6E). By comparison, VCP expression was most significantly elevated in PTC with BRAF-mutant, RET fusion, and PAX8 fusion genes (Fig. 6F). In contrast to ARF4, there was a similar frequency of BRAF alterations in PTC with either high or low VCP expression (Fig. 6G).

We next evaluated whether VCP and ARF4 expression was associated with patient outcome, treatment (RAI vs. non-RAI) or disease classification. Overall, log-rank analysis using the entire cohort (n = 413) did not detect any difference in survival between PTC patients with high tumoral VCP compared with those with low VCP (P_L = NS), although a significant difference was evident at early (P_B = 0.016) and intermediate time points (P_T = 0.022; Fig. 6H). Similar to THCA, higher VCP expression was present in breast tumors versus normal tissue in the BRCA cohort (Supplementary Fig. S8A), which again failed to correlate with a significant reduction in survival (P_L = NS; Supplementary Fig. S8B).

A key finding, however, was that the subgroup of RAI-treated PTC patients with high tumoral VCP expression (n = 195) had significantly reduced disease-free survival (DFS) than those with low VCP (P_L = 0.016; Fig. 6I; Supplementary Fig. S8C). By comparison, there was no significant difference in DFS of patients who did not receive RAI treatment when stratified on median tumoral VCP (Fig. 6J). Cox regression analysis further highlighted that higher tumoral VCP in RAI-treated PTC patients was associated with an increased risk of recurrence (hazard ratio 3.57; Fig. 6K). Interestingly, the risk of recurrence was even greater for RAI-treated patient subgroups associated with BRAF-like (HR 4.63) or BRAF alterations (HR 4.97; Fig. 6K; Supplementary Fig. S8D–S8G). RAI-treated patients with lower tumoral ARF4 expression receiving a dose ≥100 mCi also had reduced survival (Fig. 6L and M; Supplementary Fig. S8H and S8I) and a higher risk of recurrence (HR 2.93; Fig. 6K). In contrast, there was no increase in the risk of recurrence of non–RAI-treated PTC patients with higher tumoral VCP or lower ARF4 expression (Fig. 6K). Overall, disease classification indicated that RAI-treated patients had significantly more advanced thyroid disease than non–RAI-treated patients (Supplementary Fig. S9A; Supplementary Table S5). Importantly, there were no significant differences in clinical staging attributes for RAI or non-RAI treatment groups stratified for VCP or ARF4 tumoral expression (Supplementary Fig. S9B and S9C; Supplementary Tables S6 and S7).

Collectively, we show that ARF4 and VCP are significantly dysregulated in PTC, with expression profiles that fit the repression of RAI uptake generally apparent in thyroid cancers. We propose that dysregulated ARF4 and VCP results in reduced trafficking of NIS to the PM via repressed ARF4 function and increased VCP activity in targeting NIS for degradation. We further identify that VCP and ARF4 are associated with poorer survival characteristics in RAI-treated patients and represent promising new drug targets in patients who are RAI refractory. A model of the proposed functional interaction of NIS with VCP and ARF4 is outlined in Fig. 7.

Extensive studies have sought to enhance NIS expression and function in patients with RAIR-TC and hence resensitize tumors to RAI therapy. This is essential because patients with RAIR-TC, particularly those with metastatic disease, have a life expectancy of 3 to 5 years and represent a group for whom there is a clear unmet medical need (2, 3). Most investigations so far have focused on “redifferentiation agents,” which stimulate the expression of thyroid-specific genes including NIS. Constitutive activation of the MAPK pathway in thyroid cancer results in dysregulated NIS expression and function, decreased RAI uptake, and a poor patient prognosis. A substantial number of studies have therefore focused on selective inhibitors of the MAPK pathway to restore RAI avidity (20, 53).

Importantly, new drug strategies, such as combining BRAF or MEK inhibitors with pan-PI3K inhibitors, are showing preclinical promise by rescuing NIS gene expression in thyroid cancer (19). However, we propose that augmenting NIS trafficking to the PM is fundamental to boosting the efficacy of radioisotope treatment, given that we are now close to having the necessary tools to restore NIS expression. With the advent of NIS-based in vivo imaging and therapeutic approaches in nonthyroidal malignancies and other disease modalities, the exploitation and enhancement of NIS function via FDA-approved VCP inhibitors have widespread clinical potential. For example, NIS expression can now be induced via administration of engineered viruses or mesenchymal stem cells in tumoral and nontumoral settings, including cardiac disease and tissue regeneration (6, 54, 55).

Here, we have identified five allosteric inhibitors of VCP that all enhanced RAI uptake in vitro by a minimum of 2-fold. Specific ARF4 agonists do not currently exist, and hence we pursued VCP inhibition as our central therapeutic strategy. To circumvent potential issues of toxicity, we evaluated alternative drugs to enhance RAI uptake, of which ebastine and clotrimazole are FDA approved and well tolerated in vivo (50). Given that RAI uptake was enhanced within 24 hours of ebastine or clotrimazole treatment, our findings necessitate the need for clinical trials to address whether patients receiving these drugs at the time of RAI therapy uptake more ¹³¹I. One important finding in support of this is that native, relatively dedifferentiated thyroid cells that have negligible endogenous RAI uptake in our hands showed measurable NIS function following VCP inhibition.

Previously PBF, which, unlike NIS, has a functional endocytosis motif (38), was the only protein shown to bind NIS and diminish its function due to altered subcellular localization via endocytosis (29). Now, our MS/MS, co-IP, and PLA assays have identified and validated ARF4 and VCP as novel NIS interactors. ARF4 binds NIS via a C-terminal ₅₇₄VXPX₅₇₇ motif to traffic NIS toward the PM and enhance NIS function. One surprising facet observed via new imaging technologies was the rapid movement of vesicles expressing ARF4 and NIS, which hint at a more dynamic process of NIS trafficking than previously suggested. ARF4 has a range of subcellular roles, from recruiting adaptor proteins for packaging proteins into vesicles destined for the PM, to recycling proteins from the PM through endosomes (56). Our data suggest ARF4 is involved in the vesicular transport of NIS to the PM, given that inhibition of endocytosis had no impact on the ARF4-mediated increase in RAI uptake.

An important clinical observation was that high tumoral VCP expression, which acts predominately in the ER, and hence early in the progression of NIS to the PM, was strongly correlated with DFS, both overall, but particularly in patients who received RAI treatment. Similarly, PTC patients with low tumoral ARF4 expression who received higher doses of RAI (≥100 mCi) also had a significantly worse DFS. In accordance with ATA guidelines (1), RAI-treated patients in TCGA had more advanced disease than non–RAI-treated patients. However, there was no significant difference in clinical staging attributes for RAI or non-RAI treatment groups stratified for VCP or ARF4 tumoral expression. Importantly, these findings suggest that altered expression of VCP or ARF4 do not contribute to the advancement of thyroid disease. Instead, we propose that high VCP or low ARF4 expression significantly influence outcomes following RAI treatment through modulating the function of NIS. Our hypothesis is that high tumoral VCP expression in patients with PTC results in increased NIS degradation, permitting less NIS to be trafficked to the PM by ARF4, which we show is repressed in PTC. Hence the function of NIS, which is already at very low expression levels or predominantly localized to intracellular compartments, in PTC, is further attenuated, resulting in worse DFS. In agreement with this, VCP or ARF4 expression did not significantly influence disease recurrence in patients who had not received RAI treatment.

Several gaps in our knowledge remain. As a protein expressed mainly in basolateral membranes, cell polarization is important to NIS function. No human polarized cell models are physiologically relevant to thyroid cell biology, however. Despite this, our studies in nonpolarized cell systems demonstrate a universal mechanism of NIS trafficking in transformed thyroid and breast cells, as well as primary human and mouse thyrocytes. Further, because our dedifferentiated TPC-1 thyroid cells, as well as our MDA-MB-231 breast cells, are unlikely to express physiologically relevant levels of the TSH receptor, we did not investigate the potential effects of the canonical regulator of NIS-TSH-on NIS trafficking (57). Our hypothesis is that TSH is more important to the expression of NIS than its subcellular trafficking. This reflects the surprising timescales of trafficking we observed in our HILO microscopy, whereby TSH is known to influence NIS function over the course of hours and days (57), in contrast to the vesicular movement of NIS close to the PM, which was unexpectedly dynamic. We thus propose that NIS function at the PM is considerably more rapid than currently envisaged.

Collectively, we now identify two novel NIS interactors, VCP and ARF4, with critical roles in modulating NIS function as well as correlating markedly with clinical outcome in PTC. Notably, VCP is specifically druggable with FDA-approved inhibitors, resulting in enhanced RAI uptake in breast and thyroid cancer models. Strategies that manipulate the function/expression of VCP or ARF4 will thus offer a promising new therapeutic strategy for RAIR-TC in addition to augmenting NIS function for novel in vivo imaging and therapeutic strategies across a broad disease spectrum.

No potential conflicts of interest were disclosed.

Conception and design: A. Fletcher, V.L. Poole, K. Boelaert, V.E. Smith, C.J. McCabe

Development of methodology: A. Fletcher, M.L. Read, V.L. Poole, A.S. Turnell, C.J. McCabe

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Fletcher, M.L. Read, C.E.M. Thornton, D.P. Larner, K. Brookes, H.R. Nieto, G.G. Lavery, J.A. Fagin, A.S. Turnell

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Fletcher, M.L. Read, H.R. Nieto, R.J. Thompson, I. Landa, J.A. Fagin, M.J. Campbell, V.E. Smith, C.J. McCabe

Writing, review, and/or revision of the manuscript: A. Fletcher, M.L. Read, G.G. Lavery, K. Boelaert, A.S. Turnell, V.E. Smith, C.J. McCabe

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Fletcher, M.L. Read, C.E.M. Thornton, V.L. Poole, M. Alshahrani

Study supervision: K. Boelaert, A.S. Turnell, V.E. Smith, C.J. McCabe

This work was supported by the Medical Research Council (MR/P000509/1 to C.J. McCabe, V.E. Smith, and K. Boelaert) and Wellcome Trust (RG_05-052 to A. Fletcher). C.J. McCabe also received funding from Cancer Research UK, Get A-Head Charitable Trust, and a PhD studentship from the Medical Research Council/University of Birmingham. We acknowledge the contribution to this study by the Human Biomaterials Resource Centre (University of Birmingham), A. Di Maio and D. Calebiro for expertise in HILO, and D. Nasteska and D. Hodson for C57BL/6 mice.

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.