A clinically-relevant polymorphism in the Na+/taurocholate cotransporting polypeptide (NTCP) occurs at a rheostat position

Conventionally, most amino acid substitutions at important protein positions are expected to abolish function. However, in several soluble-globular proteins, we identified a class of non-conserved positions for which various substitutions produced progressive functional changes; we consider these evolutionary “rheostats”. Here, we report a strong rheostat position in the integral membrane protein, Na+/taurocholate cotransporting polypeptide (NTCP), at the site of a pharmacologically-relevant polymorphism (S267F). Functional studies were performed for all 20 substitutions (“S267X”) with three substrates (taurocholate, estrone-3-sulfate and rosuvastatin). The S267X set showed strong rheostatic effects on overall transport, and individual substitutions showed varied effects on transport kinetics (Km and Vmax). However, the outcomes were substrate dependent, indicating altered specificity. To assess protein stability, we measured surface expression and used the Rosetta software suite to model structure and stability changes of S267X. Although buried near the substrate binding site, S267X substitutions were easily accommodated in the NTCP structure model. Across the modest range of changes, calculated stabilities correlated with surface-expression differences, but neither parameter correlated with altered transport. Thus, substitutions at rheostat position 267 had wide-ranging effects on the phenotype of this integral membrane protein. We further propose that polymorphic positions in other proteins might be locations of rheostat positions.


Introduction
Amino acid substitutions are commonly used to evaluate which amino acids in a protein contribute to function. Several decades of studies have led to conventional "rules" for mutational outcomes that are now included in many textbooks and are often implicitly or explicitly assumed in the design and interpretation of experimental studies. For instance, at "important" protein positions, only amino acids with biochemical properties similar to the wild-type are expected to allow function, whereas other amino acid substitutions are expected to abolish function or structure. However, the mutational studies that gave rise to these rules were primarily focused on evolutionarily conserved amino acid positions (Gray, Kukurba, & Kumar, 2012). When we performed substitution studies of less conserved positions, results were seldom consistent with expected outcomes. Instead of an "on/off" pattern, when non-conserved positions were substituted with a variety of amino acids, each substitution had a different outcome (e.g. Figure   2). The fact that one position could be substituted to access a continuum of functional outcomes is analogous to an electronic dimmer switch; therefore, these positions have been labeled as "rheostat" positions (Hodges, Fenton, Dougherty, Overholt, & Swint-Kruse, 2018;Meinhardt, Manley, Parente, & Swint-Kruse, 2013;Wu, Swint-Kruse, & Fenton, 2019).
At the onset of the current study, we had three areas of interest. First and foremost, we were curious whether rheostat positions were limited to the soluble-globular class of proteins in which they were discovered, or if they also exist in transmembrane proteins (Meinhardt et al., 2013). Second, if rheostat positions do exist in integral membrane proteins, we wondered whether the functional outcomes arising from various substitutions were dependent upon the substrate being transported. That is, we wished to explore the effects of substitutions at rheostat positions on substrate specificity. Third, we wanted to relate the continuum of functional outcomes to the complex conformational changes experienced by an integral membrane protein during transport.
To address these questions, we chose to use human Na + /taurocholate cotransporting polypeptide (NTCP) as a model membrane protein. NTCP is expressed at the basolateral membrane of human hepatocytes where it plays an important role in the enterohepatic circulation of bile acids (Hagenbuch & Dawson, 2004). In addition to conjugated bile acids such as taurocholate (TCA), NTCP mediates the uptake of other substrates into hepatocytes, including estrone-3-sulfate and several statins such as rosuvastatin (Claro da Silva, Polli, & Swaan, 2013).
Furthermore, several single nucleotide polymorphisms (SNPs) have been reported to alter the transport activity of NTCP (Ho, Leake, Roberts, Lee, & Kim, 2004). One of these SNPs (NTCP*2) leads to the missense amino acid substitution S267F and has an allele frequency of 7.5% in Chinese Americans. Previously published data for S267F indicated reduced transport of taurocholate, wild-type-like transport of estrone-3-sulfate, and increased transport of rosuvastatin in vitro (Ho et al., 2004;Ho et al., 2006;Pan et al., 2011). Clinically, this mutation resulted in severe hypercholanemia with total serum bile acid levels of about 15-to 70-fold above normal in homozygous pediatric patients (Deng et al., 2016;Dong et al., 2019); some of the patients also had elevated liver enzymes, jaundice and gallstones (Dong et al., 2019). In homozygous adult patients, NTCP*2 resulted in total serum bile acid levels 2-to 5-fold above normal (Liu et al., 2017).
We interpreted these previous observations as a potential indication that NTCP bile acid transport could be rheostatically modulated by substitutions at position 267. To test this, we assessed the function of wild-type NTCP and all 19 amino acid substitutions at position 267 with cellular uptake studies. We also determined whether substitution outcomes were substrate dependent, by measuring transport of taurocholate, estrone-3-sulfate, and rosuvastatin.
Additional experiments aimed to differentiate the effects of substitutions on protein surface expression and transport kinetics. Finally, we used homology modeling and energetic calculations to relate the rheostatic modulation of function to the complex conformational changes utilized by NTCP to transport bile acids and other substrates (Zhou et al., 2014).

Site-directed mutagenesis
Mutagenesis reactions were completed using the QuikChange Lightning Multi Site-Directed Mutagenesis Kit (Agilent, Santa Clara, CA). We used a previously-cloned His-tagged human NTCP in the pcDNA5/FRT vector for the mutagenesis template (Zhao et al., 2015).
Primers were designed to flank the S267 codon by approximately twenty nucleotides and were ordered from Invitrogen (Carlsbad, CA). Mutated cDNA was then transformed into One Shot TOP10 cells (Thermo Fisher Scientific, Waltham, MA) and bacterial colonies were randomly selected. The cDNA was isolated using QIAGEN (Hilden, Germany) mini prep kits and sequenced (GENEWIZ, South Plainfield, NJ). Constructs containing the appropriate amino acid replacements were transfected into HEK293 cells for functional or expression studies as described below.

Cell culture
HEK293T/17 cells obtained from (American Type Culture Collection, Manassas, VA) were plated on poly-D-lysine coated flat bottom 48-and 6-well TPP cell culture plates at densities of 100,000 and 800,000 cells per well respectively. Cells were cultured in Dulbecco's Modified Eagle's Medium (American Type Culture Collection) containing 100 U/ml penicillin, 100 g/ml streptomycin (Thermo Fisher Scientific) and 10% fetal bovine serum (Hyclone, Thermo Fisher Scientific) for 24 hours prior to transfection. Experiments were then completed 48 hours post transfection.

Transient transfection of HEK293 cells
Empty vector, wild-type NTCP, and variant NTCP plasmids were transfected into HEK293 cells 24 hours after plating. Transfections followed the FuGENE HD protocol obtained from Promega (Madison, WI). All plasmids were transfected in triplicates for functional studies on 48well plates. For surface biotinylation experiments completed on 6-well plates, single wells were transfected.

Initial uptake experiments
The uptake procedure has been optimized and is well established in our laboratory for functional studies (Zhao et al., 2015). Uptake buffer (142 mM NaCl, 5 mM KCl, 1 mM KH 2 PO 4 , 1.2 mM MgSO 4 , 1.5 mM CaCl 2 , 5 mM glucose, and 12.5 mM HEPES, pH 7.4) was used for all washes and to prepare uptake solutions. Cells were washed with warm uptake buffer and then incubated at 37°C with nanomolar concentrations of either [ 3 H]-taurocholate (30 nM), [ 3 H]estrone-3-sulfate (5.8 nM), or [ 3 H]-rosuvastatin (50 nM) for five minutes. Uptake was then terminated by washing with ice-cold uptake buffer and the cells were solubilized in a 1% TX-100 solution in phosphate buffered saline (PBS). Radioactivity was measured using a liquid scintillation counter and protein concentrations were determined using bicinchoninic acid assays (Thermo Fisher Scientific). Results were calculated by correcting for total protein, subtracting the uptake by cells expressing empty vector, setting wild-type NTCP as 100%, and then comparing all variants to wild-type as percent of control.

Surface biotinylation
HEK293 cells were plated on 6-well plates and transfected as described above. Forty-eight hours later, cells were incubated for 15 minutes on ice, and all solutions and buffers used were prechilled. Each well was washed with PBS and then incubated for one hour at 4°C while rocking with 1 mg/mL EZ-Link NHS-SS-Biotin (Thermo Fisher Scientific) in PBS. Cells were then washed with PBS and incubated for twenty minutes at 4°C with 100 mM glycine in PBS while rocking. After washing with PBS, cells were lysed using 10 mM Tris, 150 mM NaCl, 1 mM EDTA, 0.1% SDS, 1% Triton X-100, pH 7.5 (lysis buffer), containing protease inhibitors (cOmplete protease inhibitor cocktail, Sigma-Aldrich). The lysates were centrifuged at 10,000 x g for two minutes and the supernatant was incubated for one hour at room temperature using end-overend rotation with pre-washed NeutrAvidin Agarose Resin beads (Thermo Fisher Scientific). The beads were washed with lysis buffer, captured proteins were eluted with 1 X SDS sample buffer containing 5% β-mercaptoethanol and 1 X protease inhibitors at 70°C for 10 minutes and collected by centrifugation at 850 x g for five minutes.

Western blotting
Surface biotinylation samples were heated to 50°C for 10 minutes before separation using 4-20% polyacrylamide gradient gels (Bio-Rad, Hercules, CA). After separation, proteins were transferred to nitrocellulose membranes using Invitrogen's Power Blotter System. Blots were blocked with 5% milk in Tris-buffered saline containing 0.1% Tween 20 (TBS-T) for one hour at room temperature while rocking. Following blocking, blots were incubated overnight at 4°C with a mouse antibody against the  subunit of Na + /K + -ATPase (Abcam-ab7671, 1:2,000) and with a mouse antibody against the His-tag (Tetra·His Antibody, QIAGEN-Cat. No.34670, 1:2,000) in blocking solution on a rocker. The next day blots were washed with TBS-T and TBS before incubation with an HRP conjugated goat anti-mouse secondary antibody at 1:10,000 (Thermo Fisher Scientific-Cat. No. 31430) in 2.5% milk in TBS. After one hour, blots were washed with TBS and incubated with SuperSignal West Pico Chemiluminescent substrate (Thermo Fisher Scientific). Blots were visualized using a LI-COR Odyssey Fc (LI-COR, Lincoln, NE) and bands were quantified using their Image Studio Lite Quantification Software.

Time dependency and kinetics experiments
Initial linear rates were determined for wild-type, S267F, S267W, and S267N using multiple time points between 10 seconds and 10 minutes at low and high concentrations of each substrate: taurocholate at 0.1 μM and 100 μM; estrone-3-sulfate at 1 μM and 200 μM; and rosuvastatin at 5 μM and 500 μM (Table 1). Kinetics for wild-type, S267F, S267W, and S267N were determined using HEK293 cells plated on 48-well plates. Uptakes were performed using sodium containing and sodium-free uptake buffers (NaCl was replace by choline chloride) and after correction for total protein and surface expression, results were analyzed using GraphPad Prism 8 (Michaelis-Menten kinetics).  (Zhou et al., 2014). As with the templates upon which the models were based, the topology of each model comprises 9

Homology model for human NTCP
transmembrane (TM) helices linked by short loops into core (TM 3,4,5,8,9 and 10) and panel domains (TM 2,6 and 7), along with the substrate binding intracellular crevice ( Figure 1). The model of NTCP lacks the N-terminal sequence corresponding to TM1 of the ASBT crystal structures; to keep the helix numbering consistent with the known crystal structures, we have chosen to start the NTCP helix numbering with TM2.

Modeling S267 variants of human NTCP
To facilitate structural exploration in response to sequence variants, we began by using the "relax" protocol ( To build a structural model of a given NTCP sequence variant, we used the "ddG" protocol in Rosetta (Kellogg, Leaver-Fay, & Baker, 2011;Kuhlman et al., 2003). With respect to our study, this protocol was used to introduce the desired amino acid substitution at position 267 and then iterated between optimization of the nearby sidechains and optimization of the backbone. We applied this protocol 10 times to each of the 100 members of our (wild-type) structural ensemble to yield 1000 models of the desired sequence variant. To avoid potential sampling artifacts from drawing the conformation/energy from the single lowest-energy conformation sampled, we ranked all conformations for a given sequence variant on the basis for Rosetta energy and carried forward the 50 th -best conformation (i.e., 95 th percentile) as the representative.
The same process was repeated for each of the 20 amino acids, to generate all possible sequence variants at this position. While the starting models already included serine at position 267, the same protocol was nonetheless applied to introduce serine; this ensured that any structural/energetic changes were indeed due to sequence variations and not simply changes relative to the starting structure induced by the modeling protocol.
The same analysis was separately completed using the structural ensemble for the inward-open state and the structural ensemble for the outward-open state.
All Rosetta calculations were carried out using git revision 0e7ed9fd3cd610f2a7c9f3bdcaba64a9b11aab0d of the developer master source code.

Statistical analysis
Calculations were performed using GraphPad Prism 8 (GraphPad Software Inc., San Diego, CA). Correlation was evaluated using Pearson and Spearman correlation coefficients.
Significance was determined using One-way ANOVA followed by Dunnett's multiple comparisons.
Results were considered significantly different at p < 0.05.

Construction of an NTCP sequence alignment
A detailed description of the NTCP sequence alignment is given in the Supplemental Methods. The references associated with these methods are included in the Supplemental Material.

Cellular substrate transport of S267 variants
A property of the rheostat positions previously observed in soluble-globular proteins is that they were non-conserved throughout evolution (e.g. (Meinhardt et al., 2013)). NTCP position 267 was reported to be highly conserved among many different animals, including: primates, rodents, dogs, cats, horses, chicken, several fish, and marine chordate (Deng et al., 2016). However, when we expanded the sequence alignment to include 1,561 homologues of the Solute Carrier Family Notably, for all three substrates, some variants transported substrates better than wildtype whereas others had diminished transport. The range of observed changes spanned several orders of magnitude. Thus, position 267 exhibited definitive "rheostatic" substitution behavior.
Furthermore, the rank-order of the amino acid substitutions differed among the three substrates.
This is further discussed below, but here we particularly note that the NTCP*2 polymorphism, S267F, showed a significant decrease in taurocholate transport, a 2.0-fold increase for estrone-3-sulfate transport, and 2.5-fold increase for rosuvastatin transport. Previous reports showed decreased taurocholate transport, estrone-3-sulfate levels similar to wild-type, and increased rosuvastatin transport (Ho et al., 2004;Ho et al., 2006;Pan et al., 2011). The discrepancy for estrone-3-sulfate could arise from the slightly different uptake conditions including substrate concentrations, incubation time, and/or different cell lines used in the two studies.

Dissecting the composite cellular outcomes of S267 variants
Protein substitutions can alter substrate transport kinetics, substrate specificity, protein stability, and/or intra-cellular trafficking to the outer membrane. The cellular uptake assay is sensitive to changes in any of these parameters. Thus, we devised experiments to dissect the functional and structural contributions.
To assess the combined effects of trafficking and stability, we quantified differences in surface expression of NTCP variants using surface biotinylation experiments followed by western blotting ( Figure 3A). When normalized for the loading control, Na + /K + ATPase, the expression levels varied between 22.5% for S267Q and 153% for S267W. The expression of most of the other variants was similar to wildtype ( Figure 3B). Thus, NTCP appeared to accommodate different amino acid side chains at position 267 without significantly disrupting the overall structure or trafficking to the cell surface.
Next, initial uptake experiments from Figure 2 were corrected for the surface expression and results are shown in Figure 4. The overall rheostat-like behavior remained, showing that substitutions altered transport function. The rank orders of S267 substitutions were further analyzed using correlation plots to illustrate the effects on substrate specificity ( Figure 5, Supplemental Table 1). Strikingly, taurocholate uptake did not correlate with either estrone-3sufate or rosuvastatin uptake ( Figure 5A and B). Because the outcome of each substitution differed among the substrates, these data indicate that amino acid substitutions at rheostat position 267 altered substrate specificity (Tungtur, Schwingen, Riepe, Weeramange, & Swint-Kruse, 2019). In contrast, with the exception of S267C, the uptake of estrone-3-sulfate and rosuvastatin strongly correlated ( Figure 5C), suggesting a shared translocation pathway for these two substrates.
Based on the results presented in Figure 4, we performed a full kinetic analysis for select variants: wild-type, S267F (the NTCP*2 polymorphism), S267N, and S267W. Asparagine at position 267 was chosen because this substitution resulted in similar uptake as wild-type for estrone-3-sulfate and rosuvastatin but higher uptake for taurocholate. In contrast, the tryptophan substitution was chosen because it resulted in lower uptake for taurocholate but higher uptake for estrone-3-sulfate and rosuvastatin. For these four proteins, concentration dependent uptake was assessed under initial linear rate conditions using transiently transfected HEK293 cells. After normalizing for surface expression, we analyzed the results using the Michaelis-Menten equation and calculated K m and V max values ( Figure 6, Table 1). Both the K m and V max were frequently altered, indicating altered substrate affinity and transporter turnover.
To summarize these data, we calculated the capacity of each variant to transport the various substrates (V max /K m ) ( Table 1). For the most part, the capacity of the variants chosen agreed with the rank order shown in Figure 4. Of the four variants, wild-type NTCP had the highest capacity (134 ± 14 L/mg/min) for taurocholate ( Figure 6A; Table 1) but the lowest capacity for the other two substrates: 15 ± 1.3 L/mg/min for estrone-3-sulfate and 24 ± 5.1 L/mg/min for rosuvastatin. The lowest capacity for taurocholate (11 ± 2.1 L/mg/min) was determined to be for S267W ( Figure 6A; Table 1) which is in agreement with the single time point single concentration results ( Figure 4A). For estrone-3-sulfate and rosuvastatin ( Figure 6B and C, Table 1), S267F showed the highest capacity with 99 ± 11 L/mg/min and 112 ± 23 L/mg/min, respectively, which confirms the results presented in Figure 4B and C. In summary, amino acid substitutions at rheostat position 267 differentially altered both kinetic parameters for substrate transport.

Homology modeling of human NTCP structure
Both modeling and experiments suggest that human NTCP has 9 transmembrane (TM) helices (Doring, Lutteke, Geyer, & Petzinger, 2012;Hu et al., 2011) with an extracellular glycosylated amino terminus (Appelman, Chakraborty, Protzer, McKeating, & van de Graaf, 2017). These helices are arranged into "core" (TM 3,4,5,8,9 and 10) and "panel" domains (TM 2, 6 and 7) that flank the substrate-binding, intracellular crevice (e.g. Figure 1). Although no experimental structure is available for human NTCP, structural information could be derived from potential substitutions yield energies more favorable (more negative) than serine ( Figure 7A). This observation was striking because it is in stark contrast with typical results from such computation predictions, which have long found that the wild-type amino acid tends to score more favorably than any other substitution (Kuhlman & Baker, 2000).
To explore the structural basis for the observed energetics in this modeling experiment, we selected two substitutions that were stabilizing (S267I and S267W), one with little effect on stability (S267N), and one predicted to be slightly destabilizing (S267G). Inspection of the representative conformations for each of these variants revealed slight changes in the local environment; in particular, TM helices 2 and 7which face the sidechain presented at position 267respond in a slightly different manner to each variant ( Figure 7B).
In our model of the outward-open conformation, the wild-type S267 was more solvent exposed than in the inward-open conformation, with a solvent accessible surface area of 17.76 Å 2 as compared to 4.26 Å 2 . Remarkably though, the same unusual behavior emerged when probing stability differences in the outward-open conformation. Many substitutions were predicted to be more stable than the wild-type serine ( Figure 7C), and again the notable tolerance for alternate amino acids at position 267 could be rationalized by malleability of the local structure, this time the nearby TM helix 10 ( Figure 7D). In both cases, the small rearrangement of these helices understates the dramatic differences in sidechain conformations needed to accommodate these alternatively packed arrangements (Supplemental Figure 1).

Correlation of structure models and experimental data
We next compared the computational stabilities of each S267 variant to the experimental data. The most direct comparison should be to the NTCP surface expression, which would be decreased or increased by altered protein stability. To that end, we examined the effect between cellular expression levels and the calculated energies using ( Table 1).
Instead, we observed a correlation between surface expression and the difference in energy between the two conformations (Pearson and Spearman coefficients were -0.64 and -0.52, respectively; both correlation coefficients were non-zero with p < 0.05; Supplemental Table   1). There are several potential explanations for why sequence variants with more favorable conformation serves as a "trap" that disfavors membrane insertion and/or correct folding required for presentation on the cell surface. In such a model, the outward-open conformation is critical for transport but over-stabilization of this conformation would prove deleterious. As an alternate explanation, however, we note that the Rosetta energy function is parameterized primarily for study of soluble proteins. As a result, energy differences between sequence variants are computed relative to a reference state that may not be appropriate for membrane proteins (the

Discussion
Rheostat positions were first described in soluble-globular proteins (Hodges et al., 2018;Meinhardt et al., 2013) and have been predicted in a wide variety of proteins (Miller, Vitale, Kahn, Rost, & Bromberg, 2019). However, to our knowledge, no such positions have been biochemically/experimentally identified in transmembrane proteins. In the present study, we confirmed that rheostat positions can occur in transmembrane transport proteins such as NTCP.
The overall phenotype of each substitution at position 267 was primarily dominated by changes in altered transport kinetics. Full kinetic analyses with selected variants demonstrated that both K m and V max were differentially affected, along with substrate specificity. Nonetheless, both modeling results and experimental data indicated subtle changes in stability that were distinct from (not correlated with) changes in transport. Thus, changes at one rheostat position altered multiple functional and structural parameters, revealing a complex interplay that must be resolved to advance predictive pharmacogenomics. Since we also observed complex outcomesaffecting multiple functional parametersarising from substitutions at rheostat positions in human liver pyruvate kinase, this complexity is likely common in a variety of proteins (Wu et al., 2019).
Of the affected functional parameters in NTCP, we were particularly intrigued by the altered substrate specificity that was demonstrated by the changed rank-orders ( Figure 2, left side) and poor correlation ( Figure 5) of amino acid substitutions for the three substrates.
Historically, altered substrate specificity has been defined from the "perspective" of the proteins, as changes in either the rank-order of preferred substrate and/or changes in the fold-change of transport (Creighton, 1993;Tungtur et al., 2019). The results shown herefrom the perspective of the substrateprovide an orthogonal view of specificity.
The substrate-dependent effects became even more apparent when we compared the kinetics for selected variants ( Figure 6, Table 1). The substitution-dependent effects on substrate specificity could arise if the translocation pathway or binding pocket for taurocholate differed from that of the other two substrates. Distinct binding pockets within the translocation pathway were recently demonstrated for three substrates of the Organic Cation Transporter 1 (Boxberger, Hagenbuch, & Lampe, 2018). Different binding pockets or translocation pathways may be a common feature of multi-specific drug transporters. Alternatively, substrate-dependent substitution outcomes might arise if the different substrate/substitution combinations had different effects on the NTCP conformational and equilibrium dynamics, similar to the types of changes that arise from amino acid substitutions in beta-lactamase (Modi & Ozkan, 2018). When the rheostatic outcomes from the cellular uptake assays were further investigated, strong rheostatic effects were observed on transport, and the effects on stability of the inwardopen conformation (in which position 267 was mostly buried) were slightly rheostatic.
Furthermore, computational modeling and stability calculations were in agreement with the experimental measures of protein surface expression. This indicates that modeling will be a useful tool for identifying positions in functionally important regions that tolerate a wide range of substitutions, the hallmark characteristic of a rheostat position.
In conclusion, these combined results showed that NTCP position 267 is a rheostat position. Although individual substitutions had wide-ranging effects on the various aspects of function and stability that give rise to the overall phenotype of cellular transport, much of the change could be attributed to altered transport kinetics. Structurally, position 267 was strikingly tolerant to substitution, despite the fact that it was largely buried in the inward-open conformation.
Based on these results, we propose that other polymorphic positions in NTCP and in other proteins might also be locations of rheostat positions (Fenton, Page, Spellman-Kruse, Hagenbuch, & Swint-Kruse, 2020). Given the complex interplay between substitutions and substrate specificity observed at this NTCP rheostat position, it is imperative to expand recognition and understanding of rheostat positions in order to advance predictive pharmacogenomics.    the right-hand side shows the substitutions ordered from highest to lowest transport activity.
Results were calculated as percent of wild-type NTCP and are reported as the mean ± SEM from at least three independent experiments, each comprising three technical replicates.
Horizontal lines to aid visual inspection correspond to wild-type ("WT") values, which were set to 100%, and asterisks indicate a p < 0.05 level of significant difference from wild-type NTCP.   Figure 2B) and are presented with the substitutions rank-ordered from largest to smallest for each substrate. Horizontal lines indicate wild-type control, which was set to 100%. Error bars represents propagated SEM; asterisks indicate a p < 0.05 level of significant difference from wild-type NTCP.   Table 1). Individual points are labeled with letters to indicate their amino acid replacements; wild-type is indicated as "WT".  Structural details from the models underlying these energy differences. Four different sequence variants are compared with the wild-type S267; in each case, the conformation of TM helices 2 and 7 respond to changes in the amino acid at position 267, which is located on TM helix 9b.  Individual points are labeled with letters to indicate their amino acid replacements; wild-type is indicated as "WT". Proline is not shown because its predicted stability difference is off-scale.