Mechanism of anion selectivity and stoichiometry of the Naþ∕I− symporter (NIS)

Monika Paroder-Belenitskya, Matthew J. Maestasb, Orsolya Dohána,1, Juan Pablo Nicolaa,2, Andrea Reyna-Neyraa,2, Antonia Follenzic, Ekaterina Dadachovad, Sepehr Eskandarib, L. Mario Amzele, and Nancy Carrascoa,f,2,3

aDepartment of Molecular Pharmacology, fDepartment of , cDepartment of Pathology, dDepartment of Nuclear Medicine and Microbiology and Immunology, Albert Einstein College of Medicine, Bronx, NY 10461; bBiological Sciences Department, California State Polytechnic University, Pomona, CA 91768-4032; and eDepartment of and Biophysical Chemistry, Johns Hopkins University School of Medicine, Baltimore, MD 21205

Edited by Ramón Latorre, Centro de Neurociencias, Universidad de Valparaíso, Valparaíso, Chile, and approved September 9, 2011 (received for review May 24, 2011)

I− uptake in the , the first step in thyroid hormone bio- a lack of sequence homology (19). Consistent with our prediction, þ synthesis, is mediated by the Naþ∕I− symporter (NIS) with an elec- the Vibrio parahaemolyticus Na ∕galactose transporter (vSGLT), þ trogenic 2Naþ :1I− stoichiometry. We have obtained mechanistic which belongs to the same family as NIS (Na ∕solute cotranspor- information on NIS by characterizing the congenital I− transport ter family 5A), has the same fold as LeuT (20–22). defect-causing NIS mutant G93R. This mutant is targeted to the The NIS mutation G93R was identified in a patient who plasma membrane but is inactive. Substitutions at position 93 show developed goitrous hypothyroidism due to a compound hetero- that the longer the side chain of the neutral residue at this position, zygous G93R/T354P NIS mutation (23). Having shown that NIS the higher the K for the anion substrates. Unlike WT NIS, which translocates different substrates with different stoichiometries m þ mediates symport of Naþ and the environmental pollutant per- (3), here we now report startling changes in Na ∕anion stoichio- chlorate electroneutrally, G93T/N/Q/E/D NIS, strikingly, do it elec- metry and anion selectivity uncovered by the molecular charac- trogenically with a 2∶1 stoichiometry. Furthermore, G93E/Q NIS terization of various substitutions at position 93 in discriminate between anion substrates, a discovery with potential NIS. clinical relevance. A 3D homology model of NIS based on the structure of the bacterial Naþ∕galactose transporter identifies G93 Results as a critical player in the mechanism of the transporter: the changes G93R NIS Is Targeted to the Plasma Membrane but Is Inactive. from an outwardly to an inwardly open conformation during the Whereas COS-7 cells transiently transfected with WT NIS avidly − − transport cycle use G93 as a pivot. accumulated I upon incubation with a subsaturating [I ](20μM) K − μ ( mðI Þ of WT rNIS is approximately 30 M), and this activity − transport defect ∣ homology modeling ∣ radioiodide therapy ∣ was inhibited by ClO4 , G93R NIS was inactive (Fig. 1A,blue) − sodium solute cotransporter family even at near-saturating [I ](200μM) (Fig. 1A,red). Flow cytometry (FC) with an anti-rNIS-Ct Ab showed that G93R NIS was synthesized and expressed at levels similar to he iodine-containing T3 and T4 (triiodo- those of WT NIS (Fig. S2A). To determine the amount of NIS at Tthyronine and thyroxine) are essential for development and N maturation of the central nervous system, skeletal muscle, and the cell surface, we engineered an extracellular HA tag at the t lungs in the fetus and newborn, and they regulate intermediary of NIS. FC using an anti-HA Ab showed that G93R NIS was − targeted to the plasma membrane (Fig. 1B). This was further con- metabolism (1). I uptake into the thyroid, the first step in T3 þ − firmed by immunofluorescence (Fig. 1C). Therefore, the lack of and T4 biosynthesis is mediated by the Na ∕I symporter (NIS). By coupling the inward transport of Naþ down its electrochemical activity of G93R NIS was not due to impaired targeting. gradient to the translocation of I− against its electrochemical gradient, NIS avidly concentrates I− in the thyroid. NIS-mediated Lysine Is Tolerated at Position 93. Surprisingly, G93K NIS was as − þ active as WT NIS (Fig. 1A). Thus, the positive charge of Arg was I transport is electrogenic: Two Na are transported with − not the cause of G93R NIS lack of function, even though position each I (2). Remarkably, NIS translocates different substrates 93 was predicted to be located within TMS III. To examine the with different stoichiometries, as NIS-mediated transport of − − electrophysiological correlates of the uptake experiments de- perrhenate (ReO4 ) or (ClO4 ) is electroneutral þ − − scribed above, we expressed WTand G93 NIS mutants in Xenopus (1Na ∶1ReO ∕ClO ) (3). − 4 4 laevis oocytes. In control oocytes, I (≤5 mM) did not evoke an NIS activity has long played a role in the diagnosis and treat- electrogenic response, whereas in WT NIS-expressing oocytes it ment of thyroid disease, including the highly successful treatment evoked an inward current that represents NIS-mediated electro- of thyroid cancer with radioiodide after thyroidectomy (4). We þ − − genic Na ∕I symport (2). G93K NIS supported robust I - cloned NIS and have since extensively characterized it (2, 5–9). Human NIS (hNIS) and rat NIS (rNIS) share 84% amino acid identity and 93% similarity (10). The secondary structure model Author contributions: M.P.-B., M.J.M., O.D., J.P.N., A.F., S.E., L.M.A., and N.C. designed of NIS shows 13 transmembrane segments (TMS), an extracellu- research; M.P.-B., M.J.M., O.D., J.P.N., A.R.-N., A.F., S.E., L.M.A., and N.C. performed research; A.F. and E.D. contributed new reagents/analytic tools; M.P.-B., M.J.M., O.D., Nt lar amino terminus ( ), and an intracellular carboxy terminus J.P.N., A.R.-N., A.F., S.E., L.M.A., and N.C. analyzed data; and M.P.-B., M.J.M., J.P.N., S.E., (Ct) (7) (Fig. S1). L.M.A., and N.C. wrote the paper. − Congenital hypothyroidism due to an I transport defect (ITD) The authors declare no conflict of interest. is a rare autosomal recessive disorder (11, 12); 13 ITD-causing This article is a PNAS Direct Submission. NIS mutations have been reported to date (Fig. S1). The mutants 1Present address: National Institute of Oncology and Institute of Experimental Medicine, that have been studied in detail have provided key mechanistic Hungarian Academy of Sciences, Ráth György u. 7–9, 1122 Budapest, Hungary. – information on NIS (13 17). For example, substitutions at T354 2Present address: Department of Cellular and Molecular Physiology, School have revealed that this position requires an -OH group at the of Medicine, New Haven, CT 06510. þ β-carbon, which we showed is involved in Na binding/transloca- 3To whom correspondence should be addressed. E-mail: [email protected]. tion, and proposed a structural homology between the Aquifex This article contains supporting information online at www.pnas.org/lookup/suppl/ BIOCHEMISTRY þ aeolicus Na ∕leucine transporter (LeuT) (18) and NIS, despite doi:10.1073/pnas.1108278108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1108278108 PNAS ∣ November 1, 2011 ∣ vol. 108 ∣ no. 44 ∣ 17933–17938 Downloaded by guest on October 1, 2021 evoked currents, confirming that position 93 tolerates this resi- WT NIS-mediated Naþ∕I− symport is electrogenic with a due (Fig. 1D). 2Naþ :1I− stoichiometry (2, 3) (Fig. 2A). At 250 μMI−—close K − — K þ 104 16 G93A, N, S, and T were all active (Fig. 1 E and F). The mag- to its mðI Þ the G93T hNIS mðNa Þ was mM − μ K þ nitude of the currents reflects the level of expression of the (Fig. 2C). Raising the [I ] to 750 M, the G93T mðNa Þ de- relevant protein in a given oocyte and should not be taken to creased by approximately 20% (86 4 mM) (Fig. 2D). There- indicate the transport ability of a specific mutant as compared to fore, increasing the concentration of one substrate decreases K any other or to WT NIS. Such comparisons are effectively made the m of the cosubstrate in G93T NIS, as reported for WT NIS in the standardized kinetic analyses (Figs. 1G,2A,3D and E, (2) and other cotransporters (24, 25). and 4D). Notably, G93Q NIS showed currents only at very high [I−] (5 mM) (Fig. 1E). Consistent with this observation, there was Asn and Thr Substitutions at Position 93 Convert NIS-Mediated − − − no I accumulation in COS cells expressing G93Q NIS even at ReO4 (and ClO4 ) Transport from Electroneutral to Electrogenic. The − K − 61 7 4 6 μ 200 μMI (Fig. 1F), although the protein was targeted to the cell m of G93T NIS for ReO4 ( . . M) was similarly surface (Fig. S2B). Each of the other NIS mutants (A, S, T,and N) approximately 17-fold higher than that of WT hNIS 3 5 0 4 μ − displayed different levels of I− transport activity upon incubation ( . . M) (Fig. 3A and ref. 26); the levels of ReO4 trans- − port by G93T hNIS were considerably higher than those of WT with 20 μMI with a clear pattern: the longer the side chain, the þ − hNIS (Fig. 3 A and B). Strikingly, the Na ∕ReO4 symport lower the activity (Fig. 1F) and this was not due to different levels þ − mediated by G93T NIS showed sigmoidal Na dependence of expression (Fig. S2A). Interestingly, 200 μMI resulted in a (Hill coefficient ¼ 2) (Fig. 3B), suggesting that the substitution greater relative increase in transport by G93N and Tas compared þ − converted the 1Na ∶1ReO4 stoichiometry of WT NIS (3) to K − to WT NIS, suggesting a change in their mðI Þ (Fig. 1F). þ − electrogenic (2Na ∶1ReO4 ). Furthermore, in X. laevis oocytes expressing G93T NIS, ReO − elicited inward currents, whereas The K Values of G93N and T NIS for I− and Naþ Are Significantly 4 m there were no currents when WT NIS was expressed (Fig. 3C). K − Higher than Those of WT NIS. The mðI Þ values of G93A and K NIS − Moreover, the environmental pollutant ClO4 , which is structu- were 30 2 and 33 2 μM, similar to that for WT NIS [Fig. S2C − rally similar to ReO , also elicited currents in oocytes expressing and refs. 13–17, 19)]. In contrast, G93T and N NIS had signifi- 4 G93Tor N NIS (Fig. 3C), in contrast to WT NIS (2), implicating K − 282 44 358 69 μ cantly higher mðI Þ values ( and M, respec- position 93 as a key Naþ∕anion coupling link. Kinetic analysis tively) (Fig. S2D). Electrophysiologically, whereas G93K, A, and − − of G93T and N NIS-mediated ClO4 or ReO4 transport by K − S exhibited mðI Þ values comparable to those of WT NIS, G93N K − 17 2 μ electrophysiology revealed a mðClO Þ of M for G93N >18 K − 4 and T NIS exhibited -fold higher mðI Þ, and G93Q NIS had 18 4 μ K − 29 2 μ and M for G93T NIS and a mðReO4 Þ of M >200 K − an astonishingly -fold higher estimated mðI Þ (6.1 mM) for G93N and 24 3 μM for G93T NIS (Fig. S3A). K − − − (Fig. 1G). Similarly, the mutants with a high mðI Þ also had a þ No NIS-mediated ClO4 - or ReO4 -evoked currents were K þ higher mðNa Þ (Fig. 2A). The Hill coefficient for Na activation observed in G93R or K NIS (Fig. 3C). In contrast, both G93T þ − of NIS-mediated inward currents remained unchanged at 2 for and N NIS exhibited sigmoidal Na dependence with ClO4 − WT NIS and G93 mutants. (Fig. 3E) or ReO4 (Fig. S3B) (Hill coefficient ¼ 2), corroborat- To eliminate the experimental variation in transient transfec- ing the uptake data (Fig. 3B). Interestingly, although the Naþ∶ – − − tion, we transduced MDCK (Madin Darby canine kidney) cells ClO4 or ReO4 stoichiometry was different in G93T versus WT with a lentiviral vector to permanently express HA-tagged hNIS NIS, the relaxation kinetics of the presteady-state charge move- (WT or G93T). As G93N and T NIS behaved similarly (Fig. 1 F ments in G93N and G93T were similar to those of WT NIS, and G), further studies in MDCK cells focused on G93T hNIS. suggesting that the partial reactions corresponding to Naþ bind- I− transport kinetics in MDCK cells recapitulated those in COS-7 ing and binding-induced conformational changes were not K − 16 5 μ cells: WT hNIS mðI Þ was M (3, 17), whereas G93T significantly affected by the substitution (Fig. S4 and ref. 2). K − 269 32 μ hNIS mðI Þ was M (Fig. 2B). Furthermore, the transporter turnover rate, as determined by the

Fig. 1. Analysis of G93 NIS substitutions in COS-7 cells and X. laevis oocytes. (A) Steady-state I− transport in WT, G93R, or G93K NIS-transfected COS-7 cells. Cells − − − were incubated with 20 μMI in the absence (blue bars) or presence (light blue bars) of 80 μMClO4 , or with 200 μMI in the absence (red bars) or presence − − (pink bars) of 800 μMClO4 . Values shown (pmol I ∕μg DNA SD) are from one of at least five different experiments and corrected for transfection efficiency. (B) Flow cytometry under nonpermeabilized conditions with an Ab against the extracellular HA epitope at the NIS Nt , showing WT (19%) and G93R NIS (15%). (C) HA immunostaining of WT and G93R NIS under nonpermeabilized conditions. NIS is stained with Alexa 488 (green) and nuclei with DAPI (blue). Red scale bar ¼ 20 μm. (D and E) Current traces are shown in response to 5 mM I− in (D) control (water-injected) X. laevis oocytes or oocytes expressing WT, G93R, or G93K þ∕ − ¼ −50 − NIS, or (E) G93A, N, T, or Q NIS. The evoked inward currents represent NIS-mediated electrogenic Na I cotransport into the cell. V m mV. (F)I transport − ½ þ ¼ 100 as in A in COS-7 cells transfected with WT, G93A, N, Q, S, or T NIS cDNAs. (G) Kinetics of I transport by WT and G93K, A, S, T, N, and Q NIS. Na o mM and ¼ −50 − Vm mV. For each mutant, the current values were normalized to the maximum current obtained at saturating [I ]. The smooth lines are fits of the data to the Michaelis–Menten equation. Values represent the means SE from at least four oocytes.

17934 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1108278108 Paroder-Belenitsky et al. Downloaded by guest on October 1, 2021 Fig. 2. Kinetics of I− transport in X. laevis oocytes and MDCK cells. (A) Kinetics of Naþ dependence of I− transport by WT, G93T, and G93N NIS. ½I−¼5 mM and ¼ −50 þ V m mV. Current values were normalized to the maximum current obtained at saturating [Na ]. The smooth lines are fits of the data to the Hill equation. The Hill coefficient values were 1.9 0.2 for WT, 2.0 0.1 for G93T, and 1.9 0.1 for G93N NIS. Values represent the mean SE from at least four oocytes. (B) − − − ¼ 5 Initial rates (2-min time points) of I uptake were determined at the indicated [I ]s. KmðI Þs are indicated as an average SE (n experiments). The graph shown is a representative experiment. (C and D)Naþ dependence of I− uptake. Cells were incubated for 2 min with (C)250or(D) 750 μMI− and the indicated [Naþ]s. The graph is representative of >3 experiments. Isotonicity was kept constant with choline-Cl. In all flux experiments, the activity was standardized by expression levels at the cell surface, and background values obtained with nontransfected cells were subtracted (<5 in B and C; and 12 pmol∕μg DNA in D).

− I − ratio of the maximum I -evoked current ( max) to the maximum cating that G93K transports ReO4 electroneutrally, just like WT Q − presteady-state charge movements ( max), was approximately NIS. In contrast, G93R, which did not transport I in mammalian 35 s−1 in G93N and G93T mutants and was similar to that re- cells (Fig. 1A) or elicit currents in oocytes in the presence of − − − ported for WT NIS (2). I , ReO4 , or ClO4 (Figs. 1D and 3C), did not translocate − − ReO4 in MDCK cells either, even at high [ReO4 ]. Thus, it Glu93 and Gln93 Discriminate Between I− and ReO −∕ClO −. Even is not solely the presence of a positive charge at position 93 that − − 4 4 though ReO4 (or ClO4 ) did not elicit currents in G93K NIS- renders NIS nonfunctional, but more specifically the Arg side − μ K − chain. expressing oocytes (Fig. 3C)at3 M ReO4 (the mðReO4 Þ of WT − G93D NIS-expressing cells transported I−, although at sig- hNIS), G93K transported ReO4 in MDCK cells (Fig. 4A), indi- nificantly lower levels than WT NIS. On the other hand, G93E NIS-expressing cells showed no transport at 200 μMI− (Fig. 4B) or higher, though the protein was expressed at the plasma membrane (Fig. S5). I−-evoked currents mediated by G93E NIS were extremely small even at exceedingly high [I−] (5 mM, >160 K − times higher than the WT NIS mðI Þ) (Fig. 4C). Indeed, the K − >6 G93E NIS estimated mðI Þ was extraordinarily high ( mM), 220-fold higher than that of WT NIS (Fig. 4D), whereas G93D NIS-mediated robust I−-evoked currents, albeit with a highly K − μ increased mðI Þ of approximately 150 M (Fig. 4E). − G93D NIS-transfected cells transported ReO4 at 3 μM. In contrast, G93E NIS activity was barely detectable (Fig. 4A, light − green). On the other hand, at a 10-fold higher [ReO4 ], G93E − − NIS clearly exhibited ClO4 -sensitive ReO4 transport, although at lower levels than WT NIS (Fig. 4A, dark green). Electrophy- − − siologically, both G93D and E exhibited ClO4 - and ReO4 - evoked currents, suggesting a switch in transport stoichiometry þ − − leading to electrogenic Na ∶ClO4 ∕ReO4 symport (Fig. 4C). − − Although the apparent affinity of G93D for ClO4 and ReO4 K s ¼ 4 μ was high ( m and 8 M, respectively), that of G93E for − K 250 μ ClO4 was significantly reduced ( m approximately M) − − (Fig. 4 D and E). G93Q also displayed ClO4 - and ReO4 - elicited currents (Fig. 4C). The signal with G93E was too small for kinetic analysis; G93D (at 5 mM I−) yielded a sigmoidal re- K þ 23 3 ¼ 1 9 lationship with a mðNa Þ of mM (Hill coefficient . ) (Fig. S6). The striking observation that G93E and Q NIS trans- − − − port ReO4 and ClO4 even though I transport is severely

− impaired in these two mutants indicates that these two amino Fig. 3. ReO4 transport kinetics in MDCK cells and X. laevis oocytes. (A)In- − acids confer the ability to discriminate between substrates. itial rates (2 min) of ReO4 uptake in transduced MDCK cells were determined − at the indicated [ReO4 ]s. Background in nontransfected cells (<2 pmol∕μg þ − The NIS Homology Model Reveals Close Contact Between Gly93 and DNA) was subtracted. (B)Na dependence of ReO4 transport: Cells were − þ þ μ ð − Þ Trp255. The structures of four bacterial Na -driven transporters incubated for 2 min with 180 M ReO4 and the indicated [Na ]s. Km ReO4 s are given as an average SE of all experiments. The graph is a representative have been determined by X-ray crystallography: LeuT, a homo- experiment. Background obtained with NT cells (<2 pmol∕μg DNA) was sub- logue of eukaryotic neurotransmitter transporters such as the − − tracted. (C) Current traces in response to 1 mM ClO4 (Top Traces) or ReO4 norepinephrine, dopamine and serotonin, γ-aminobutyrate, and ¼ −50 (Bottom Traces) in oocytes expressing WT, G93R, K, N, or T NIS. Vm mV. − − glycine transporters (18, 27); vSGLT (20), a homologue of the For each mutant, ClO4 and ReO4 traces were obtained in the same oocyte. þ∕ − ½ þ ¼ eukaryotic Na glucose transporter (SGLT-1); the benzylhydan- (D) Kinetics of ClO4 transport by G93T and N NIS in oocytes. Na o 100 ¼ −50 þ toin transporter (Mhp1) from Moraxella liquefaciens (28), of the mM and V m mV. Data analyzed as in Fig. 1G.(E) Kinetics of Na dependence of anion transport by G93T and G93N NIS in oocytes with 1 mM nucleobase-cation-symport-1 family of transporters; and the þ − ¼ −50 ∕ BIOCHEMISTRY ClO4 Vm mV. Data were processed as in Fig. 2A. The Hill coefficient Na betaine symporter (BetP) from Corynebacterium glutami- values were 1.9 0.2 for G93N, and 2.2 0.2 for G93T NIS. cum, of the betaine/choline/carnitine transporter family (29).

Paroder-Belenitsky et al. PNAS ∣ November 1, 2011 ∣ vol. 108 ∣ no. 44 ∣ 17935 Downloaded by guest on October 1, 2021 − − − Fig. 4. G93E NIS does not transport I but transports ReO4 electrogenically. (A) Steady-state ReO4 transport assays in nontransfected (NT) or COS-7 cells − transfected with WT or G93K, D, or E NIS cDNAs. Cells were incubated for 1 h with 3 μM ReO4 in the absence (light green bars) or presence (yellow bars) of − − − − 120 μMClO4 , or with 30 μM ReO4 in the absence (dark green bars) or presence (olive bars) of 800 μMClO4 .(B) Steady-state I transport activity was assayed − − − in WT, G93D, or G93E NIS-expressing COS-7 cells as in Fig. 1A.(C) Current traces in response to 5 mM I (Left) or 1 mM ClO4 (Center)orReO4 (Right) in oocytes ¼ −50 expressing G93D (Top Traces), G93E (Middle Traces), or G93Q NIS (Bottom Traces). V m mV. For each mutant, all current traces were obtained in the same oocyte. (D) Kinetics of anion transport by G93D, E, and Q NIS, as in Fig. 1G.(E) Kinetic parameters.

The structure of LeuT was determined at the highest resolution (1.6 Å) (18). Surprisingly, although all four proteins share little sequence homology (<17%), all have the same fold—an inverted topology repeat and unwound helices in regions critical for substrate binding—and a similar way of coordinating Naþ. Using as a template the X-ray structure of vSGLT (20), we generated a 3D homology model of NIS (Fig. 5A). Alignment of the NIS and vSGLT sequences using BLAST shows significant identity (27% for NIS residues 50 to 476), a value comparable to that between vSGLT and its eukaryotic homologue SGLT1 (31%). Critically, the extent of the identity between NIS and vSGLT is similar to that between LeuT and mammalian neuro- transmitter transporters (20–25%) (18); even at this level of identity the LeuT structure has become a major model for eluci- dating mechanistic information on mammalian neurotransmitter transporters. Similarly, the NIS homology model provides vital information on the protein’s mechanism. In this model, TMS III2 and X9 (superscript Arabic numerals indicate the LeuT nomenclature) are at the edge of the molecule, defining the outer wall of a cavity that in vSGLT contains the galactose substrate (20). TMS VII6 crosses TMS III2 at an angle in such a way that G93 and W255 are in close contact, with the Cα of Gly abutting the indole ring of W255. They are at the opposite side of the cavity from the unwound portion of TMS VII6 (Fig. 5B), a common essential characteristic of Naþ-driven symporters. Residues in the next helical turns (N97 and Y259) also face each other, at the bottom end of the cavity. M90, from the previous turn of TMS III2, is inside the cavity and forms part of its “upper” roof. In vSGLT, galactose is bound in a cavity at the center of the core, shielded from the extracellular milieu by hydrophobic resi- dues (20). The equivalent cavity in the NIS model may contain Fig. 5. NIS model based on the vSGLT structure. The homology model, I− or I−∕Naþ in one of the conformations of NIS during the trans- including NIS residues 50 to 476, was generated as described in SI Text. – 2–5 – 6–10 port cycle. To determine whether the NIS residues (W255 and (A) Overall structure. TMS III VI are colored magenta and TMS VII XI Y259) play the key roles suggested by the model, we generated orange. G93 (red) abuts the indole ring of W255 (olive). The rest of the model is colored green. (B) Close-up of the NIS cavity depicting G93 and W255 various HA-tagged hNIS mutants. W255A was inactive even at 6 − − opposite the unwound portion of TMS VII . N97 and Y259 also face each high concentrations of I ; W255Yaccumulated I , albeit less than other at the bottom end of the cavity. M90 forms part of the upper roof of WT NIS. In contrast, Y259A and Y259F were both functional, the cavity. Transition between two conformations of NIS: (C) Superposition of although less than WT NIS (Fig. S7A). Cells expressing all these the inwardly open and outwardly open conformations showing the move- − mutants, except W255A, accumulated ReO4 (Fig. S7B). Differ- ment of the hairpins and the connecting residues. The inwardly open confor- ences in transport were not due to varying levels of expression or mation (vSGLT-like) is colored light green and the outwardly open plasma membrane targeting (Fig. S7 C and D). conformation (LeuT-like) red. The green arrow shows the position of TMS III2 in the inwardly open conformation and the red arrow the position of the outwardly open conformation. The black arrow indicates the point around Gly93 and Trp255 Form a Ball-and-Socket Joint. As substitutions which the connecting short helix pivots to allow the rotation of the two he- at and around position 93 have such dramatic effects on NIS lical hairpins. (D) Close-up of boxed region in C.(E) Schematic representation activity, we used the NIS homology model to shed light on the of D; TMS III and IV2–3 are depicted as cylinders, the short helix as a rectangle, participation of this position in the mechanism of NIS. Structural the unstructured connections between TMS III and IV2–3, and the rectangle as alignment of the NIS model with the four experimental struc- rods. Color scheme as in C and D.

17936 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1108278108 Paroder-Belenitsky et al. Downloaded by guest on October 1, 2021 tures, excluding vSGLT (the template for modeling NIS), showed exist as charged species, given the steep energy required for large deviations reflecting that the crystal structures, as described protonation at neutral pH (4.95 and 4.42 kcal∕mol, respectively). by the authors, captured different states in the transport cycle: If some of these residues are neutral in the environment of the whereas vSGLT, and by extension the NIS homology model, are protein, the energetic cost of protonation/deprotonation will be in the inwardly open conformation (20), the others are in various “paid” by a decrease in the overall stability of the mutant protein substates of the outwardly open conformation (18, 29). This ana- that may result in changes in the protein’s structure, ranging in lysis also shows that the coordination of the second Naþ (Na2) in severity from small local distortions to total misfolding. NIS is equivalent to that in LeuT (19), but the first Naþ (Na1) Further kinetic analyses in G93T hNIS-transduced MDCK K − must be coordinated differently in NIS. Among other changes, cells showed an approximately 18-fold higher m for both I 1 2 3 − the helical hairpins formed by TMS II and III , and TMS IV (Fig. 2B) and ReO4 (Fig. 3A), and an approximately 5-fold high- 4 K þ K þ and V , significantly change their relative orientations (Fig. S8 A er m for Na (Fig. 2C). This change in the m for Na is larger and B). In this rearrangement, the structure of the IV–V3–4 pair than those observed in substitutions of β-OH-containing residues remains highly conserved, whereas there are significant changes in TMS IX8 (19). Our 3D NIS model provides a rationale for in the other pair. Alignment of TMS IV3 and V4 of the NIS model the effects of the substitutions at position 93 on NIS activity. This with those of LeuT shows a low rmsd between the aligned Cα site appears to be a pivot around which occurs one of the major carbons (1.9 Å for 77 Cα). When this alignment—optimizing components of the conformational change between the inwardly the superposition of TMS IV3 and V4 between LeuT and NIS and the outwardly open conformations: the rotation of the helical —is applied to the portion of the NIS model that includes TMS hairpin formed by TMS III and IV2and3 (Fig. 5). Because it lacks a II–V1–4, the change in the relative orientation between the two side chain, Gly, the WTresidue, is ideally suited for this position. helical hairpins becomes evident (Fig. S8 A and B). The confor- Nevertheless, other substitutions at position 93 result in proteins mational change is best described as a 25–30° rotation of the TMS that not only are active but also in some cases produce additional II–III1–2 helical hairpin (Fig. 5C), which, strikingly, pivots around changes in key properties of the transporter. The conformational the contact between G93 and the indole ring of W255 (Fig. 5B). changes we propose are fully compatible with the model pro- Thus, the G93/W255 pair can be described as a ball-and-socket posed by Gouaux’s group (22, 30, 31) and by Forrest et al. (32), joint—the Cα H of Gly, the ball; the six-member ring of Trp, the but they highlight the importance of position 93. Faham et al. also socket. Residues 118–132, which connect to the helical hairpins, identified a Gly residue (G99 of vSGLT) as an important parti- provide a key structural element that permits the rotation of the cipant in the vSGLT transport cycle (20). two hairpins with small variations in their respective structures. In the NIS model, the Cβ of nonglycine residues points toward This region consists of a short helix (Fig. S8 A and B, in olive) the inside of the cavity (Fig. 5), which is occupied by galactose in flanked by two short loops. The change in distance between the vSGLT. By analogy, the substrates of NIS also likely occupy this ends of the two hairpins is easily accommodated by a rotation of cavity and may interact with the side chains of all substitutions at the short helix with small changes in the connecting pieces acting position 93. Arg is probably too large and rigid and may fill the as hinges [Fig. 5 C and D (close-up), and E (schematic represen- cavity and interfere with substrate binding. Lys, although longer 2 tation)]. The large displacement of the Ct of III “pushes,” via the than Arg, is more flexible and may extend upward toward the connecting rod, the short helix, resulting in a large displacement extracellular space. Thr, Asn, and Asp, being smaller, allow the of its Nt portion, whereas its Ct acts as a hinge for the rotation of substrates to enter the cavity and be transported, although with a 3 K – the helix and does not move significantly, allowing TMS IV and higher m (Figs. 1G and 2 B D). V4 to remain close to their original positions. TMS VII6 (in the The significance of G93 is underscored by the manner in − second half of NIS), which encompasses W255, the residue which WT NIS and the NIS mutants handle ClO4 , a transported against which G93 rests, runs at an angle with respect to TMS competitive inhibitor of NIS. The environmental and health 3 6 − IV . As the rotation takes place, TMS VII has to move away as impact of ClO4 has acquired a new sense of urgency, as the an- part of the overall conformational change. has been detected as a contaminant in public water supplies − Thus, the nature of the side chain at position 93 of NIS may (33). We have demonstrated that NIS actively transports ClO4 , þ − control the transition between the outwardly and the inwardly including translocating it into milk, and that the Na ∕ClO4 þ − open conformations and play a role in the kinetics as well as transport stoichiometry is electroneutral (1Na ∶1ClO4 ) (3). þ − þ the stoichiometry. This transition involves a change in not only Here we show that the stoichiometry of Na ∕ReO4 and Na ∕ − the relative sizes of the cavities open to either the cytosol or the ClO4 mediated by G93N/T/D/E/Q NIS is 2∶1 and thus electro- extracellular milieu but also the volume of the substrate-binding genic, in stark contrast to the electroneutral 1∶1 stoichiometry cavity at the center of the molecule (Fig. S8 C and D). of their transport by WT NIS. The reason for this change may − − þ be that in WT NIS, both ReO4 and ClO4 interfere with Na Discussion binding to the Na1 site. G93T/N/Q/D/E, rather than relieving Here we have characterized the NIS mutation G93R, which this interference (Fig. S9), probably provide an additional occurs in TMS III2 of the symporter. The lack of activity of this Naþ-coordinating group that shifts the position of Na1 away from − − mutant protein is not due to its having a positively charged resi- the bound ReO4 or ClO4 , but still participates in transport. due within the membrane, as G93K results in an active transpor- G93A, which cannot contribute to this site, elicited currents − − ter (Figs. 1 A, D, and G and 4A). In contrast to the many NIS barely detectable with ReO4 and ClO4 (Fig. S10). A possible mutants we have studied previously (13–17, 19), substitutions explanation is that this substitution allows a small leak current K − at position 93 show a significant change in the m for I . This that occurs only when transport takes place but is not thermo- is true for not only the neutral residues Thr, Asn, and Gln dynamically coupled to it. (Fig. 1G) but also Asp and Glu (Fig. 4D), indicating that NIS The lack of I− transport activity by G93R/Q/E NIS in mam- tolerates both, a strong basic (Lys) and a strong acidic residue malian cells probably results from the side chains being either (Asp) in the middle of TMS III2. Whether the side chains of these too large or incompatible with the chemical requirements. residues bear a charge at physiological pH depends on their G93E and Q exhibit the most surprising behavior: Although − respective pKas, a function of the electrostatic properties of G93D transports I , G93E and Q do so only at extremely high their microenvironment. Lys, and even more likely Arg, remains [I−] (Fig. 4 C–E), probably a consequence of the size difference positively charged, as the energetic cost of deprotonating these between the side chains. Strikingly, G93E and Q not only trans- − − þ two side chains at the experimental pH of 7.5 is 4.12 and port ReO4 and ClO4 but also do so with a 2∶1 Na ∶anion stoi- BIOCHEMISTRY 6.77 kcal∕mol, respectively. It is also probable that Asp and Glu chiometry. Although this observation is difficult to explain, it is

Paroder-Belenitsky et al. PNAS ∣ November 1, 2011 ∣ vol. 108 ∣ no. 44 ∣ 17937 Downloaded by guest on October 1, 2021 interesting that Asp, Glu, and Gln have the chemical character- positions equivalent to NIS G93 in other transporters may have istics necessary for coordinating Naþ, as we propose for the G93T a similar function. and N substitutions, whereas G93K displays electroneutral þ − − Na ∕ReO4 or ClO4 stoichiometry, just like WT NIS (Figs. 3C Materials and Methods and 4A). Vesicular stomatitis virus glycoprotein (VSV-G) pseudotyped, human immu- − nodeficiency virus-1-based, third-generation lentiviruses (35), one bearing The selective transport of ReO4 (Fig. 4) by G93E and Q NIS, which do not transport I−, may prove to be particularly useful WT hNIS and the other G93T hNIS, were generated using calcium phos- phate-mediated cotransfection of 293T cells with four plasmids: a CMV for NIS-based gene therapy. NIS has been used clinically for >60 promoter-driven packaging construct expressing the gag and pol genes, a y to successfully treat thyroid cancer with radioiodide, and Rous sarcoma virus promoter-driven construct expressing rev, a CMV promo- efforts to make non-NIS-expressing tumors susceptible to radio- ter-driven construct expressing the VSV-G envelope, and a self-inactivating therapy by introducing NIS are currently under way (34). Given transfer construct driven by the CMV promoter containing the human immu- that NIS is endogenously expressed in the thyroid, thyroid NIS nodeficiency virus-1 cis-acting sequences and an expression cassette for must be selectively down-regulated prior to radiotherapy by either WT or G93T hNIS. MDCK cells (105) were transduced by adding 500 μL administering either high concentrations of T4 or high doses of of viral supernatant per well in a six-well plate. Transduced cells were I−. However, I− cannot be used to down-regulate thyroid NIS if analyzed using flow cytometry. More details and associated references are the NIS molecule exogenously transfected into cancer cells also provided in SI Materials and Methods. transports I−. Thus, introducing G93E or Q NIS into tumors in- stead of WT NIS would allow the down-regulation of thyroid NIS ACKNOWLEDGMENTS. We thank the members of the Carrasco laboratory by I− without affecting NIS function in the tumor, as neither and Dr. Myles Akabas for critical reading of the manuscript and helpful − suggestions. M.P.B. was supported by Medical Scientist Training Program G93E nor Q NIS transports I . This may in turn permit successful Training Grant 5T32GM002788. M.J.M. was supported in part by the Howard 99 m − 188 − imaging with TcO4 and treatment with ReO4 . Hughes Medical Institute–Cal Poly Pomona Undergraduate Research Appren- The data here presented indicate that the side chain at NIS tice Program, and by a National Institutes of Health (NIH) training grant position 93 has a major effect on the size and chemical charac- (2R25GM061190). O.D. was supported in part by the American Thyroid Association. A.F. was supported in part by a grant from Ricerca Finalizzata teristics of the ion cavities as they undergo the transition from Sanitaria Regione Piemonte. This work was supported by NIH Grants the outwardly to the inwardly open conformation and also plays SC1GM086344 (to S.E.), NS-061827 (to L.M.A.), and DK-41544 and CA- a role in the kinetics and the stoichiometry of transport. The 098390 (to N.C.).

1. Dohan O, et al. (2003) The sodium/iodide symporter (NIS): Characterization, regula- 18. Yamashita A, Singh SK, Kawate T, Jin Y, Gouaux E (2005) Crystal structure of a tion, and medical significance. Endocr Rev 24:48–77. bacterial homologue of Naþ∕Cl–dependent neurotransmitter transporters. Nature 2. Eskandari S, et al. (1997) Thyroid Naþ∕I− symporter. Mechanism, stoichiometry, and 437:215–223. specificity. J Biol Chem 272:27230–27238. 19. De la Vieja A, Reed MD, Ginter CS, Carrasco N (2007) Amino acid residues in trans- þ − þ 3. Dohan O, et al. (2007) The Naþ∕I symporter (NIS) mediates electroneutral active membrane segment IX of the Na ∕I symporter play a role in its Na dependence transport of the environmental pollutant perchlorate. Proc Natl Acad Sci USA and are critical for transport activity. J Biol Chem 282:25290–25298. 104:20250–20255. 20. Faham S, et al. (2008) The crystal structure of a sodium galactose transporter reveals þ 4. Reiners C, Hanscheid H, Luster M, Lassmann M, Verburg FA (2011) Radioiodine for mechanistic insights into Na ∕sugarx symport. Science 321:810–814. ðþÞ remnant ablation and therapy of metastatic disease. Nat Rev Endocrinol 7:589–595. 21. Abramson J, Wright EM (2009) Structure and function of Na -symporters with – 5. Dai G, Levy O, Carrasco N (1996) Cloning and characterization of the thyroid iodide inverted repeats. Curr Opin Struct Biol 19:425 432. transporter. Nature 379:458–460. 22. Krishnamurthy H, Piscitelli CL, Gouaux E (2009) Unlocking the molecular secrets of – 6. Levy O, et al. (1997) Characterization of the thyroid Naþ∕I− symporter with an sodium-coupled transporters. Nature 459:347 355. anti-COOH terminus antibody. Proc Natl Acad Sci USA 94:5568–5573. 23. Kosugi S, Inoue S, Matsuda A, Jhiang SM (1998) Novel, missense and loss-of-function 7. Levy O, et al. (1998) N-linked glycosylation of the thyroid Naþ∕I− symporter (NIS). mutations in the sodium/iodide symporter gene causing iodide transport defect in – Implications for its secondary structure model. J Biol Chem 273:22657–22663. three Japanese patients. J Clin Endocrinol Metab 83:3373 3376. 24. Parent L, Supplisson S, Loo DD, Wright EM (1992) Electrogenic properties of the cloned 8. Tazebay UH, et al. (2000) The mammary gland iodide transporter is expressed during Naþ∕glucose cotransporter: I. Voltage-clamp studies. J Membr Biol 125:49–62. lactation and in breast cancer. Nat Med 6:871–878. 25. Loo DD, Eskandari S, Boorer KJ, Sarkar HK, Wright EM (2000) Role of Cl- in electrogenic 9. Riedel C, Levy O, Carrasco N (2001) Post-transcriptional regulation of the sodium/ Naþ-coupled cotransporters GAT1 and SGLT1. J Biol Chem 275:37414–37422. iodide symporter by thyrotropin. J Biol Chem 276:21458–21463. 26. Zuckier LS, et al. (2004) Kinetics of perrhenate uptake and comparative biodistribution 10. Smanik PA, et al. (1996) Cloning of the human sodium lodide symporter. Biochem of perrhenate, pertechnetate, and iodide by NaI symporter-expressing tissues in vivo. Biophys Res Commun 226:339–345. J Nucl Med 45:500–507. 11. Nicola JP, et al. (2011) Iodide transport defect: Functional characterization of a novel 27. Rudnick G (2007) What is an antidepressant binding site doing in a bacterial transpor- mutation in the Naþ∕I− symporter 5′-untranslated region in a patient with congenital ter? ACS Chem Biol 2:606–609. hypothyroidism. J Clin Endocrinol Metab 96:E1100–1107. 28. Weyand S, et al. (2008) Structure and molecular mechanism of a nucleobase-cation- 12. Spitzweg C, Morris JC (2010) Genetics and phenomics of hypothyroidism and goiter symport-1 family transporter. Science 322:709–713. due to NIS mutations. Mol Cell Endocrinol 322:56–63. 29. Ressl S, Terwisscha van Scheltinga AC, Vonrhein C, Ott V, Ziegler C (2009) Molecular 13. Reed-Tsur MD, De la Vieja A, Ginter CS, Carrasco N (2008) Molecular characterization basis of transport and regulation in the NaðþÞ∕betaine symporter BetP. Nature þ∕ − of V59E NIS, a Na I symporter mutant that causes congenital I- transport defect. 458:47–52. – 149:3077 3084. 30. Singh SK, Piscitelli CL, Yamashita A, Gouaux E (2008) A competitive inhibitor traps LeuT 14. De La Vieja A, Ginter CS, Carrasco N (2004) The Q267E mutation in the sodium/iodide in an open-to-out conformation. Science 322:1655–1661. symporter (NIS) causes congenital iodide transport defect (ITD) by decreasing the NIS 31. Singh SK, Yamashita A, Gouaux E (2007) Antidepressant binding site in a bacterial – turnover number. J Cell Sci 117:677 687. homologue of neurotransmitter transporters. Nature 448:952–956. 15. Levy O, Ginter CS, De la Vieja A, Levy D, Carrasco N (1998) Identification of a structural 32. Forrest LR, et al. (2008) Mechanism for alternating access in neurotransmitter trans- þ − requirement for thyroid Na ∕I symporter (NIS) function from analysis of a mutation porters. Proc Natl Acad Sci USA 105:10338–10343. that causes human congenital hypothyroidism. FEBS Lett 429:36–40. 33. Committee to Assess the Health Implications of Perchlorate Ingestion, National ðþÞ ð−Þ 16. Dohan O, Gavrielides MV, Ginter C, Amzel LM, Carrasco N (2002) Na ∕I symporter Research Council (2005) Health Implications of Perchlorate Ingestion (Natl Academic activity requires a small and uncharged amino acid residue at position 395. Mol Press, Washington, DC), pp 75–114. Endocrinol 16:1893–1902. 34. Hingorani M, et al. (2010) The biology of the sodium iodide symporter and its 17. De la Vieja A, Ginter CS, Carrasco N (2005) Molecular analysis of a congenital iodide potential for targeted gene delivery. Curr Cancer Drug Targets 10:242–267. transport defect: G543E impairs maturation and trafficking of the Naþ∕I− symporter. 35. Follenzi A, Naldini L (2002) Generation of HIV-1 derived lentiviral vectors. Methods Mol Endocrinol 19:2847–2858. Enzymol 346:454–465.

17938 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1108278108 Paroder-Belenitsky et al. Downloaded by guest on October 1, 2021