A Nonpump Function of Sodium Iodide Symporter in Thyroid Cancer via Cross-talk with PTEN Signaling

The sodium iodide symporter (NIS), encoded by the SLC5A5 gene (solute carrier family 5, member 5), is a plasma membrane protein expressed at high levels in the thyroid gland and the lactating breast (1–3). NIS mediates iodide uptake from the bloodstream into thyroid follicular cells for thyroid hormone biosynthesis, and iodide secretion into breastfeeding milk (4). NIS-mediated iodide uptake is the basis for diagnostic nuclear imaging and radioiodine therapy in thyroid-related diseases. In differentiated thyroid cancer (DTC), radioiodine-131 (I-131) is routinely utilized for remnant ablation and postsurgical adjuvant/targeted therapy (5). Therefore, although NIS is frequently studied in thyroid cancers, focus has been on its classical iodide-pump function.

Radioiodide uptake is generally reduced in thyroid cancer compared with normal thyroid tissue, and decreased NIS expression is widely believed to cause resistance (6). However, studies of NIS expression levels in DTC have yielded divergent data (2, 7–13). Studies reporting increased NIS levels show mostly intracellular localization, and thus associated with reduced radioiodide uptake in these cancers. Similarly, NIS has been reported to be overexpressed, but largely retained intracellularly in 70% to 80% of breast cancers (13, 14) and a number of other primary nonthyroidal cancers (15–17). We therefore hypothesized that in addition to the canonical iodide-pump function, NIS could have iodide pump–independent function when localized intracellularly in thyroid cancer cells. This hypothesis is important because the mainstay of treatment of advanced thyroid cancers remains radioiodine.

Interestingly, the two main cancers with reportedly elevated NIS, namely thyroid and breast cancers, are major phenotypic components of Cowden syndrome (CS). CS is an autosomal-dominant, difficult-to-recognize, and underdiagnosed disorder, characterized by high lifetime risks of thyroid, breast, and other cancers (18, 19). A subset of CS is caused by germline mutations in the tumor-suppressor gene PTEN (18, 20, 21). PTEN antagonizes the PI3K/AKT/mTOR signaling pathway and is involved in many different sporadic cancer-related signaling pathways (22). Though the direct association between PTEN alterations and NIS is unknown, PI3K signaling upregulation has been reported to be associated with reduced iodide uptake in thyroid cancer cells (23). We therefore hypothesized that PTEN alterations in thyroid cancer can affect NIS protein levels or subcellular localization, which can, in turn, promote tumorigenesis independent of its iodide-pump function. Hence, we investigated the nonpump function of NIS in human thyroid cancer, downstream cellular phenotypes, and how PTEN and downstream signaling regulate these functions.

We utilized BCPAP, 8505C, and FTC-133 thyroid cancer cell lines (Supplementary Table S1) stably expressing full-length human NIS (FL hNIS; ref. 24). BCPAP cells were grown in RPMI-1640 medium, and 8505C and FTC-133 cells were cultured in Modified MEM medium (Sigma M0325), supplemented with 10% FBS and 1% penicillin/streptomycin. All cell lines were maintained at 37°C and 5% CO₂ culture conditions and tested negative upon routine mycoplasma testing using the MycoAlert Mycoplasma Detection Kit (Lonza). All experiments were conducted with cells at passage numbers between 3 and 15. All cell lines were authenticated through the ATCC human cell authentication service (ATCC 135-XV) and were 100% matched to the reported shot tandem repeat profiles in the DSMZ database (test date April 19, 2018).

Tunicamycin and rapamycin were purchased from Sigma. LY294002 and MK-2206 were obtained from Selleckchem. Dimethyl sulfoxide (DMSO; Sigma) served as vehicle control for experiments involving deglycosylation drug or PI3K/AKT/mTOR inhibitor treatments. Rabbit anti-NIS (Pr 2890, Rb 4430) is an in-house generated and validated antibody (25).

RNA was extracted from the cell lines using the RNeasy Mini Kit (Qiagen), purified using Turbo DNase treatment (Life Technologies), and reverse transcribed using Superscript III reverse transcriptase (Life Technologies). Primers were designed for gene transcripts of interest and cDNA quantified using SYBR Green (Life Technologies). We utilized the Applied Biosystems 7500 Real-Time PCR System. Results were analyzed using the standard ΔΔC_t method.

Protein was extracted from whole-cell lysates using the Mammalian Protein Extraction Reagent M-PER (Thermo Scientific Pierce) supplemented with a cocktail of protease and phosphatase inhibitors (Sigma) and quantified through the BCA protein assay (Thermo Scientific Pierce). Lysates were separated by SDS-PAGE and transferred onto nitrocellulose membranes. We probed for rabbit anti-NIS at 1:4,000, anti-PTEN (6H2.1) mouse monoclonal (Cascade Bioscience) at 1:1,000, anti-LARG (leukemia-associated RhoA guanine exchange factor) mouse monoclonal (EMD Millipore) at 1:5,000, and anti-GAPDH rabbit monoclonal (Cell Signaling Technology #2118) at 1:20,000 dilution. Blots were scanned digitally using the GE Amersham Imager 600 (GE Healthcare Life Science). Densitometry was performed using ImageJ software. Uncropped versions of cropped immunoblots are included in Supplementary Fig. S1A–S1G.

Cells in 6-well plates were transfected with 0.2 μg GFP-tagged wild-type PTEN plasmid (pcDNA3.1 vector) or empty vector pcDNA3.1 with Lipofectamine 3000 reagent (Invitrogen) according to the manufacturer's instructions. Cells were grown for at least 48 hours before harvesting.

Rho activity was measured using the RhoA pull-down activation assay Kit (Cytoskeleton) as previously reported (18). Relative variations in normalized RhoA-GTP level is defined as the ratio of the level of activated GTP-bound forms of RhoA and the level of total RhoA in NIS stable cells, divided by the same ratio in control cells.

Whole-cell lysates were prepared by using M-PER as mentioned above and were precleared by incubation with Thermo Protein A/G Dynabeads (Thermo Scientific Pierce) for 3 hours at 4°C on a rotator. Precleared protein lysates were quantified with the BCA Protein Assay Kit, and 2 mg/mL lysates were prepared. We used anti-NIS, anti-LARG, and anti-RhoA antibodies for pull-down and immunoblotting at the recommended dilutions. Cell lysates were separated by SDS-PAGE and transferred onto nitrocellulose membranes. Blots were scanned digitally using the GE Amersham Imager 600 (GE Healthcare Life Science).

Cells in 6-well plates were transfected with NIS, PTEN, LARG, or DPAGT1 siRNA smartpool (Dharmacon) using Lipofectamine RNAiMAX (Thermo Fisher Scientific) according to the manufacturer's instructions. We used ON-TARGET nontargeting siRNA pool (Dharmacon) as a control (siNT). Cells were collected for downstream analysis 12 to 96 hours after transfection. We used Western blot and qRT-PCR analysis to confirm knockdown.

Cells grown on coverslips were fixed with 4% paraformaldehyde for 2 minutes and then incubated with 1% Triton X-100 for 2 minutes at room temperature (RT). Cells were blocked with 10% goat serum (Vector Laboratories Inc.) for 1 hour at RT, and incubated with rabbit anti-NIS and/or mouse anti-LARG (both at 1:500 dilution) overnight at 4°C. Secondary antibodies (Alexa 488/568) were incubated at 1:2,000 dilution for 1 hour in the dark at RT. Coverslips were mounted using ProLong Gold Antifade mountant with DAPI (Invitrogen). Slides were visualized and images obtained using the Leica TCS SP5 II confocal microscope (Leica Microsystems).

Cell surface protein was isolated using the Pierce Cell Surface Protein Isolation Kit (Thermo Fisher Scientific), as previously reported (26). The cytoplasmic fraction was isolated using the Mem-PER Plus Membrane Protein Extraction Kit (Thermo Fisher Scientific). Briefly, cell pellets were washed with wash buffer and incubated in 500 μL of permeabilization buffer for 10 minutes at 4°C with constant mixing. We centrifuged the permeabilized cells for 15 minutes at 16,000 × g. The resultant supernatant contains cytoplasmic proteins. NIS in cell surface and cytoplasmic fractions was analyzed by immunoblotting.

To test NIS protein stability (27), BCPAP-FL hNIS and 8505C-FL hNIS cell lines were seeded in 6-well plates and allowed to grow overnight. Cells were transfected with PTEN siRNA or nontargeting siRNA for 48 hours. At 48 hours, 100 μg/mL cycloheximide was added to cell media. Following 0, 1, 2, 4, 8, 12, 24, 48, and 72 hours, lysates were prepared with MPER and Western blot analysis performed with anti-NIS and anti-GAPDH antibodies.

For wound-healing assays, confluent-transfected cells' monolayers were scraped with a fine sterile pipette tip, placed in the cell incubator (5% CO₂, 37°C). Images were acquired at 0 and 20 hours after scraping using a Leica DMI3000B microscope (Leica Microsystems). For migration inhibitor tests, Rho-associated coiled-coil kinase (ROCK) inhibitor Y-27632 (Sigma) at 30 μmol/L was added at 1 hour before scraping. The wound margin area was determined by image processing using ImageJ software. The average wound-healing rates were calculated from three independent experiments.

Iodine uptake of BCPAP-FL hNIS and 8505C-FL hNIS cells transiently transfected with siPTEN or siNT was performed with a nonradioiodine uptake assay, as previously reported (28). The iodide uptake of experimental wells was determined after linear regression of the calibration curve and normalized to DNA amount determined in parallel wells. Non–NIS-mediated iodide uptake was examined by conducting parallel experiments in the presence of perchlorate (Sigma), a selective inhibitor for NIS-mediated iodide uptake.

Experimental data between control and experimental cells are given as mean ± SEM with n corresponding to the number of experiments performed. The Student t test was used for significance testing as indicated in figure legends. All statistical tests were two-sided, and P values < 0.05 deemed significant.

Because NIS is not expressed in vitro in most of the established human thyroid or thyroid cancer cell lines (Supplementary Fig. S2; ref. 29), we utilized a set of thyroid cancer cell lines stably transfected with FL hNIS for our experiments, including BCPAP and 8505C, which have wild-type PTEN, and FTC-133, which is PTEN null (Supplementary Table S1). NIS is a membrane glycoprotein whose function is regulated by posttranslational glycosylation (30). On immunoblot analysis, NIS displays a hyperglycosylated form of molecular weight of approximately 70 to 90 kDa and a hypoglycosylated form of 60 kDa (Supplementary Fig. S2). The hyperglycosylated NIS represents the mature and main form of iodide-pumping NIS on the plasma membrane (31).

To investigate whether NIS affects cell migration in a thyroid cancer context, we performed the wound-healing assay on BCPAP, 8505C, and FTC-133 cells. We found that overexpression of NIS increased cell migration rates, 1.3-fold in BCPAP cells, 1.8-fold in 8505C cells, and 2.3-fold in FTC-133 cells, compared with corresponding vector control cells (P = 0.006, 0.004, and 0.003, respectively; Fig. 1A). To further validate that this phenotype is NIS-dependent, we performed transient NIS knockdown through siRNA in 8505C-FL hNIS cells and compared the migration rate (change of wound-healing capacity). The migration rate was reduced following NIS knockdown at 48 hours and 72 hours (P = 0.017 and 0.0006, respectively; Fig. 1B). Taken together, these data provide evidence that NIS increases cell migration in thyroid cancer cells.

RhoA activation, which is controlled by Rho guanine-nucleotide-exchange factors (Rho GEF), is known to play a key role in cancer cell motility (32). One Rho GEF, LARG, has been reported to bind with NIS directly through a PDZ–PDZ domain interaction in breast and liver cancer cell lines (33). Here, we show that NIS–LARG interaction is also valid in thyroid cancer (Fig. 1C). By coimmunofluorescence analysis, we observed NIS expressed at the plasma membrane as well as the cytoplasm, whereas LARG predominantly localized to the cytoplasm (Supplementary Fig. S3A), with some at the leading (migrating) edges of the cells, suggesting that NIS and LARG localize in similar subcellular regions, thus facilitating their physical interaction.

Next, we investigated downstream activation of RhoA in thyroid cancer cells. We observed that the relative level of GTP-bound (activated) form of RhoA was significantly higher in BCPAP- and 8505C-FL hNIS cells than in control cells (1.95 and 2.32 fold, P = 0.018 and 0.001, respectively; Fig. 1D). Moreover, ROCK inhibitor Y-27632 suppressed the migration rate in cells with exogenous FL hNIS expression (in BCPAP-FL hNIS, fold change = 0.59, P = 0.031; Supplementary Fig. S3B), which further support that the RhoA-ROCK pathway is involved in the enhancement of cell migration by NIS.

To confirm that LARG is involved in the NIS-mediated RhoA activation and enhancement of cell migration, we measured relative RhoA-GTP levels in BCPAP-vector control and -FL hNIS cells after siLARG knockdown. In all cells, LARG knockdown decreased activated RhoA-GTP level, which was more pronounced in BCPAP-FL hNIS than in BCPAP-vector control cells (46% vs. 58%, P = 0.023; Supplementary Fig. S3C). Relatedly, the wound-healing rate also decreased significantly after LARG knockdown (Supplementary Fig. S3D). Taken together, these data show that NIS enhances RhoA activation and downstream cellular migration through interaction with LARG.

Following the observation that NIS increases cell migration in a thyroid cancer context, to test our hypothesis that PTEN might affect NIS levels or subcellular localization, we knocked down PTEN by siRNA in BCPAP- and 8505C-FL hNIS cells (both PTEN wild-type) over different time points. PTEN protein levels decreased starting at 12 hours after siPTEN transfection. We observed that NIS protein first decreased after PTEN knockdown at/before 24 hours compared with controls, then increased significantly after 24 hours (Fig. 2A; Supplementary Fig. S4A). Moreover, the increase of hypoglycosylated NIS was more prominent than that of hyperglycosylated NIS. For example, in BCPAP-FL hNIS cells, at the 72-hour time point, the fold increase of hypo- and hyperglycosylated NIS after PTEN knockdown was 2.17 versus 1.77 (P = 0.008; Supplementary Fig. S4B). Although the total NIS increased, the ratio of hyper- to hypoglycosylated NIS decreased compared with control (1.71 vs. 2.18, P < 0.001). Importantly, we did not observe differences in NIS mRNA levels following knockdown (Fig. 2B; Supplementary Fig. S4C). The latter is particularly pertinent because the thyroid cancer cell line models we utilized do not express endogenous NIS, and the exogenous FL hNIS is driven by a cytomegalovirus promoter. Hence, differences in NIS protein levels are independent of transcript expression and are almost certainly related to posttranscriptional phenomena.

Because canonical NIS iodide-pump function is dependent on plasma membrane localization, we then investigate cell surface and cytoplasmic NIS protein levels following PTEN knockdown. In the cytoplasmic fraction, hyper- and hypoglycosylated NIS levels were increased significantly after PTEN knockdown (1.84- and 2.22-fold, P = 0.006 and 0.003, respectively; Fig. 2C). In the cell surface fraction, we also observed that hyper- and hypoglycosylated NIS levels increased (1.33- and 1.18-fold, P = 0.029 and 0.032, respectively). Importantly, even though NIS protein levels increased after PTEN knockdown, we observed no differences in cellular iodide uptake (Fig. 2D; Supplementary Fig. S4D). Our data indicate that PTEN loss results in increased exogenous NIS protein levels after 24 hours independent of NIS transcription. Moreover, cytoplasmic NIS increases more significantly compared with plasma membrane NIS, indicating that PTEN signaling might also affect NIS subcellular localization and function.

Our data so far indicate that PTEN knockdown increases exogenous NIS protein levels independent of mRNA levels. Hence, we hypothesized that PTEN signaling might affect NIS protein stability. To test our hypothesis, we performed cycloheximide chase assay in BCPAP- and 8505C-FL hNIS cells transfected with siPTEN or siNT control. We observed that the hyperglycosylated NIS is more stable, but the hypoglycosylated NIS is comparatively unstable, with protein levels approaching zero at 4 hours after cycloheximide treatment (Fig. 3A). PTEN knockdown increased the half-life of both hyper- and hypoglycosylated NIS, from 60 to more than 72 hours and from 1.75 to 2.5 hours, respectively (Fig. 3A). Overall, these data indicate that NIS protein stability could potentially explain the increased NIS protein levels in thyroid cancer cells after PTEN knockdown.

To investigate whether LARG is involved in the increase of NIS protein levels resulting after PTEN knockdown, we studied NIS protein levels in BCPAP- and 8505C-FL hNIS cells transiently transfected with siNT, siPTEN, siLARG, and both siPTEN and siLARG for 72 hours. Similar to our previous observations, NIS protein level was increased significantly after PTEN knockdown (Fig. 3B). In contrast, NIS protein decreased compared with control cells following LARG knockdown, regardless of PTEN knockdown. The fold change of hyper- and hypoglycosylated NIS was 0.82 and 0.83, respectively, as compared with control (P = 0.013 and 0.007, respectively; Fig. 3B). These data show that the increase of NIS protein levels following PTEN knockdown is LARG-dependent.

Furthermore, PTEN knockdown also resulted in increased LARG protein levels, with time points of LARG increments consistent with NIS increments (both starting from 36 hours after PTEN siRNA knockdown; Fig. 3C; Supplementary Fig. S5). For example, LARG was increased about 1.29-fold at 72 hours compared with control (P = 0.012). Consistently, overexpression of PTEN in BCPAP-FL hNIS cells for 72 hours revealed 20% decrease in LARG protein levels compared with control (P = 0.018; Fig. 3D). Taken together, our data indicate that PTEN loss increases LARG protein levels, which in turn interacts with NIS to increase NIS protein stability and hence, NIS protein levels in thyroid cancer cells. Immunofluorescence analysis corroborates these observations: following PTEN knockdown, immunofluorescence intensity of NIS and LARG increased, as well as the overlap of staining representing colocalization of both proteins (Fig. 3E).

Next, to investigate the role of PTEN and NIS together in RhoA activation, we performed RhoA activity assay in BCPAP-vector control and -FL hNIS cells transiently transfected with siNT or siPTEN for 72 hours. First, following transfection with siNT, BCPAP-FL hNIS cells displayed a higher level of activated RhoA-GTP than vector control cells, validating our previous observations. In both of vector control and FL hNIS cells, PTEN knockdown induced a significant increase in activated RhoA-GTP (Fig. 3F). Expectedly, the increase was more pronounced in cells with FL hNIS than in vector control cells. These data show that NIS exaggerates the increase of activated RhoA-GTP levels induced by loss of function of PTEN in thyroid cancer cell lines.

Because we observed PTEN knockdown increases NIS protein levels, we then turned our attention to investigate if pharmacologic inhibition of signaling downstream of PTEN could also affect NIS protein levels or localization. To test our hypothesis, we used LY294002 (PI3Ki), MK-2206 (AKTi), and rapamycin (mTORi), and monitored NIS localization through immunofluorescence microscopy in BCPAP-FL hNIS cells. Treatment with LY294002, MK-2206, or rapamycin revealed increased fluorescence intensity of cell surface NIS compared with cells treated with DMSO (1.49-, 1.32-, and 1.67-fold, P = 0.029, 0.044, and 0.004, respectively; Fig. 4A). Furthermore, we observed decreased intracellular NIS after treatment with LY294002 and rapamycin, with a more significant decrease noted in cells treated with rapamycin (fold change = 0.38, P < 0.001; Fig. 4A). We observed the same results in 8505C-FL hNIS cells after treatment (Supplementary Fig. S6).

Because the glycosylation state of NIS has been reported to affect NIS trafficking and membrane localization in thyroid cells (30, 31, 34), we next investigated whether these inhibitors could also affect NIS glycosylation. We observed that PI3K/AKT/mTOR inhibition increases hyperglycosylated NIS levels, whereas that of hypoglycosylated NIS decreases. In cells cotreated with thapsigargin, which inhibits hypoglycosylated NIS modification into hyperglycosylated NIS by inducing ER stress, we found NIS protein levels did not change in cells treated with MK-2206 or rapamycin, but still decreased in cells treated with LY294002 (Fig. 4B). Expectedly, the mRNA level of NIS did not change after treatment with these drugs compared with controls (Fig. 4C). These data provide circumstantial evidence that PI3K/AKT/mTOR inhibitors enhance the process of hypoglycosylated NIS turning into the mature hyperglycosylated NIS. In addition, in cells treated with the LY294002, although NIS also decreased in cells cotreated with thapsigargin, the decrease was more pronounced in cells without thapsigargin (0.73-fold vs. 0.61-fold, P = 0.033; Fig. 4B). Moreover, we observed that LARG protein levels decreased after treatment with LY294002, but remained unchanged after MK-2206 or rapamycin treatments (Supplementary Fig. S7). These observations indicate that the PI3K inhibitor LY294002 might also decrease LARG protein levels to affect NIS protein levels.

We next investigated whether PTEN knockdown also affects NIS glycosylation. Indeed, the direct evidence of this association was observed in cycloheximide chase assay. After cycloheximide exposure arresting NIS translation, we observed decreasing hypoglycosylated NIS in cells with PTEN expression, yet a relative maintenance or increase in hyperglycosylated NIS (Fig. 5). By 4 hours, hypoglycosylated NIS was essentially null, whereas hyperglycosylated NIS started to decrease because of degradation. However, after PTEN knockdown, we observed hyperglycosylated NIS decreased immediately after treatment with cycloheximide. The normalized hyperglycosylated NIS at 4 hours in cells with siPTEN transfection was significantly decreased compared with siNT control (0.95-fold vs. 1.46-fold normalized to 0-time point, P = 0.036; Fig. 5). These data suggest that PTEN loss could be associated with inhibition of NIS glycosylation. We therefore postulate that PTEN knockdown may inhibit glycosylation, after which, cytoplasmic (non/hypoglycosylated and part of hyperglycosylated) NIS interacts with LARG with this NIS–LARG interaction stabilizing NIS.

We next investigated if the change of NIS glycosylation by altering PTEN signaling is through particular glycosylating enzymes. Because PI3K/AKT/mTOR inhibitors partly reversed the inhibition of tunicamycin (inhibitor of enzymes involved in the initial step of N-linked glycosylation) on NIS glycosylation (Supplementary Fig. S8), we hypothesized PTEN or downstream signaling could affect initial steps of NIS glycosylation such as through such glycosylating enzymes. After knockdown of UDP-N-acetylglucosamine-dolichyl-phosphate N-acetylglucosaminephosphotransferase (encoded by DPAGT1 gene), an enzyme involved in the first step of N-linked glycosylation (35), we observed a band shift of hyperglycosylated NIS (Fig. 6A), which is indicative of the deglycosylation of NIS. In cells cotreated with LY294002, MK-2206, or rapamycin, the shift distances were decreased compared with controls (0.61-, 0.80-, and 0.53-fold; P = 0.008, 0.037, and 0.007, respectively; (8505C-FL hNIS cells; Fig. 6A). These data indicate that PI3K/AKT/mTOR inhibitors reverse the deglycosylation of NIS caused by decreasing levels of DPAGT1. Interestingly, although DPAGT1 knockdown showed an effect on deglycosylation of NIS, it also increased NIS protein levels, which can be counteracted by PI3K, AKT, or mTOR inhibition (Fig. 6B). We also investigated NIS protein levels in BCPAP-FL hNIS cells transiently transfected with siNT, siPTEN, siDPAGT1, or both siPTEN and siDPAGT1 over different time points. Knockdown was confirmed by Western blot (Supplementary Fig. S9). At 24 hours after PTEN or DPAGT1 knockdown, we observed decreased NIS protein levels, with knockdown of both PTEN and DPAGT1 resulting in larger decrements (fold change of hypoglycosylated NIS: 0.67 vs. 0.63, P = 0.009; Fig. 6C). Taken together, these data provide further supportive evidence that PTEN affects NIS glycosylation, and this effect could potentially be mediated through the glycosylating enzyme DPAGT1.

NIS, a plasma membrane protein classically mediating iodide uptake in thyroid tissue, is generally decreased in thyroid cancer cells, resulting in reduced radioiodide uptake. Several studies reported decreased NIS mRNA levels in DTC compared with normal thyroid tissue (2, 7–10), with the former induced by PI3K or MAPK signaling activation, which plays critical roles in thyroid cancer initiation and progression (36). Interestingly, we found that PTEN loss, which activates PI3K and MAPK signaling, results in increased NIS protein levels in our thyroid cancer cell line models. Importantly, because exogenous FL hNIS expression is driven by a cytomegalovirus promoter, the effects on NIS protein levels are independent of transcript expression levels. Furthermore, our data suggest that increased NIS protein levels, preferentially in the cytoplasm, could have noncanonical pump-independent function associated with protumorigenic phenotypes.

Besides its canonical iodide uptake function, the nonpump function of NIS has remained elusive in thyroid cancer. Our observation that NIS significantly enhances cell migration provides direct evidence for the protumorigenic role of nonpumping NIS in thyroid cancer. Relatedly, it has been reported that the secondary structure of NIS contains a putative density-95/discs large/zona occludens-1 (PDZ) class 1 target motif (T/S-X-V/L) at the COOH-terminus, endowing it with the potential for binding proteins through PDZ protein–protein interactions (37). For example, NIS has been reported to interact with pituitary tumor transforming gene binding factor in the cytoplasm of COS-7 kidney and FRTL-5 thyroid cell lines (38). Importantly, a more recent study has shown that NIS interacts with LARG in breast and liver cancer cell lines (33). These studies further support that NIS has iodide-pump–independent function, including in organs other than the thyroid. As relevant to thyroid cancer, our data show that the NIS–LARG interaction is valid and promotes NIS cross-talk with PTEN signaling.

Germline PTEN mutations cause CS, characterized by a high risk of thyroid, breast, and other cancers (18, 20, 21). Somatic PTEN mutations occur in a wide variety of sporadic human cancers (22). It is also well-established that PTEN mechanistically functions as a PIP3 (phophatidylinositol-3,4,5-triphosphate) 3′-phosphatase to reduce the levels of intracellular PIP3 and antagonize PI3K-AKT signaling (22). The Rho GTPase family of proteins, which are controlled by Rho guanine-nucleotide-exchange factors (Rho GEF), plays a key role in cancer cell motility (39). Interestingly, PIP3 has been reported to recruit Rho GEFs to then activate Rho GTPase (32). Our data show that PTEN loss increases LARG (a Rho GEF), which can interact with NIS, coupled with an increase in downstream activated RhoA-GTP.

It remained intriguing to investigate how PTEN loss increases NIS protein levels. We found that PTEN knockdown increases NIS protein stability and protein levels in a LARG-dependent manner. Furthermore, PTEN knockdown increases LARG protein levels. Hence, we speculate that the NIS–LARG interaction can increase NIS protein stability. Indeed, LARG is a protein with a long half-life (60% protein retained 72 hours after knockdown), potentially explaining increased NIS stability as pertinent to the NIS–LARG interaction. Supportively, it has been reported that LARG binds to ATP-binding cassette transporter A1 (ABCA1) to then stabilize it (40). Moreover, we observed decreased NIS protein levels following treatment with the ROCK inhibitor Y-27632 (Supplementary Fig. S10), indicating that RhoA activation might also be involved in stabilizing NIS. In addition, our data indicate that PTEN loss or downstream PI3K/AKT/mTOR signaling activation inhibits NIS glycosylation and its classical membrane localization. Several studies have reported that the glycosylation state of NIS affects its trafficking and membrane localization in thyroid cells (30, 31, 34), further supporting our observations. Taken together, we postulate that PTEN loss can inhibit NIS glycosylation to then increase cytoplasmic NIS. Cytoplasmic NIS is then stabilized by LARG through NIS-LARG protein–protein interaction. In parallel, PTEN loss also increases LARG protein levels and RhoA activation, further contributing to NIS stabilization and downstream effects. Notably, we also observed that the cell surface NIS increases slightly after PTEN knockdown, compared with the more pronounced increase in cytoplasmic NIS. This observation could be due to the interaction with LARG, which localizes at the submembrane region, with the intracellular C-terminal region of transmembrane NIS to also stabilize it. Corroborating this observation, LARG has been reported to interact with a number of plasma membrane proteins such as the histamine-H1 receptor, the insulin-like growth factor-1 receptor, and the semaphorin 4D/plexin-B1 receptor (41–43).

We utilized knockdown of DPAGT1, an enzyme involved in the initial step of N-linked glycosylation, to further investigate a possible association between PTEN signaling and NIS glycosylation. We observed that PI3K/AKT/mTOR inhibitors can reverse the deglycosylation process caused by reduced DPAGT1. In contrast, PTEN knockdown results in a larger effect on deglycosylation through reduction of DPAGT1. However, whether PTEN is directly involved in regulating this and other glycosylating enzymes warrants further investigation. More interestingly, although NIS is deglycosylated by DPAGT1 knockdown, protein levels of NIS increase and can be counteracted by PI3K/AKT/mTOR inhibitors. Because we believe that deglycosylation should decrease NIS stability, we postulate that the subcellular localization could explain the NIS increase. However, this observation also warrants further exploration.

We utilized PI3K/AKT/mTOR inhibitors to investigate the effect of these pharmacologic drugs on NIS protein levels, glycosylation and membrane localization. Our data show that LY294002 (PI3Ki), MK-2206 (AKTi), and rapamycin (mTORi) increase the mature hyperglycosylated NIS but decrease the hypoglycosylated NIS, with consistent increase in cell surface NIS protein. Interestingly, LY294002 significantly induces NIS mRNA expression and iodide uptake in rat thyroid cells FRTL-5 and PCCL3, and LY294002 and Akti-1/2 also increase iodide uptake in exogenous NIS-expressing papillary thyroid cancer (PTC) cell lines (23). Relatedly, LY294002 and rapamycin increase NIS expression in PCCL3 cells; however, LY294002 and Akti-1/2 had little effect on cell surface or total NIS protein levels in HEK293 human kidney cells or COS7 monkey kidney cells expressing exogenous NIS (44). Thus, the modulation of NIS protein by PI3K/AKT/mTOR inhibitors appears to be cell-type and context-dependent.

Intriguingly, studies of NIS in breast cancer, another major CS component cancer, provide further supportive evidence that the activation of PTEN downstream signaling increases intracellular NIS. It has been reported that in the breast cancer MCF7 cell line, the activating mutation PI3K p110α^CAAX increases underglycosylated NIS, with predominant localization in the cytoplasm (45). Similarly, MEK activation has also been reported to be involved in maintaining NIS protein stability in human breast cancers (46). Importantly, a more recent study reported that NIS protein level in breast cancer is associated with poor prognosis and chemotherapy resistance (47). The association between NIS levels and prognosis in DTC remains inconclusive. Interestingly, we have previously reported that follicular thyroid cancer (FTC) is proportionately more prevalent in PTEN mutation–positive patients compared with the more common PTC in the general population (19). FTC is known to have poorer prognosis than PTC, with increased development of lung and bone metastases (48). Whether the PTEN-related NIS localization and associated noncanonical function is involved in these invasive phenotypes needs further study.

The ability to concentrate radioiodide (as related to the pump-function of NIS) is generally considered to indicate a more DTC phenotype (49). Redifferentiation studies in thyroid cancers generally focus on increasing iodide uptake by using some compounds such as retinoic acid, PPARγ agonists, and histone deacetylase inhibitors, which can increase endogenous NIS expression (49). However, recent clinical trials utilizing these compounds have shown disappointing results (49). As relevant to our observations, inhibition of mTOR signaling has been reported to induce NIS expression through upregulating the transcription factor TTF1 (50, 51). Our data suggest that PTEN and downstream signaling have an impact on exogenous NIS protein levels and localization, mimicking a change-of-function effect. We speculate that these changes functionally resemble a less differentiated phenotype, at least characterized by increased cellular migration. Certainly, investigating a comprehensive set of differentiation markers will be key to ascertain or refute this hypothesis. Hence, as a master regulator of PI3K/AKT/mTOR signaling, it is plausible that PTEN also regulates the differentiation of thyroid cancer upstream of mTOR signaling. Relatedly, research efforts have also been focused on inducing exogenous NIS expression for the theoretical possibility of radioiodide therapy in refractory thyroid cancers and some nonthyroidal cancers (6). However, although exogenous NIS could increase radioiodide uptake in some cancer cells, the therapeutic effect varies in different cancer cell types (6), presumably due to differential radiosensitivity to I-131. Our data provide evidence that exogenous NIS expression increases cell migration in thyroid cancer cells. It would hence be of clinical interest to investigate if the efficacy of NIS-transfer–facilitated I-131 therapy is associated with cytoplasmic NIS level and/or NIS-induced cell migration.

In summary, we found that in thyroid cancer cells, NIS interacts with LARG to subsequently crosstalk with signaling cascades downstream of PTEN to enhance cell migration. PTEN loss, a common somatic alteration and upregulation of PI3K/AKT/mTOR signaling pathway found in multiple cancer types, likely inhibits NIS glycosylation to impair NIS membrane localization, resulting in increased cytoplasmic NIS. Furthermore, PTEN loss increases LARG protein levels and RhoA activation, thus stabilizing NIS, especially the intracellular form, through NIS–LARG interaction (Fig. 7). Together, our data suggest that in thyroid cancers, NIS, particularly within the cytoplasm, has pump-independent protumorigenic roles and crosstalks with the PTEN-PI3K/AKT/mTOR pathway, a major signaling cascade in cancer. We suspect that this cross-talk may be generalizable in extra-thyroidal malignancies and suggest targeted therapies beyond attempting to increase NIS levels.

No potential conflicts of interest were disclosed.

Conception and design: F. Feng, C. Eng

Development of methodology: F. Feng, Y.S. Chang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F. Feng, C. Eng

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F. Feng, L. Yehia, C. Eng

Writing, review, and/or revision of the manuscript: F. Feng, L. Yehia, Y. Ni, S.M. Jhiang, C. Eng

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

Study supervision: C. Eng

This work was supported, in part, by the NCI (P01CA124570 and R01CA118989) and the American Cancer Society (RPG-02-151-01-CCE, Clinical Research Professorship; all to C. Eng). We thank Dr. Ritika Jaini for helpful discussions, Todd Romigh and Tammy Sadler for assistance in cell culture and Western blot analysis, and Shikha Gautam for technical advice related to FL hNIS stable cell lines and NIS antibody. L. Yehia is an Ambrose Monell Foundation Cancer Genomic Medicine Fellow at the Cleveland Clinic Genomic Medicine Institute. C. Eng is the Sondra J. and Stephen R. Hardis Chair of Cancer Genomic Medicine at the Cleveland Clinic and an American Cancer Society Clinical Research Professor.

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.