A Model-Structure of a Periplasm-facing State of the NhaA Antiporter Suggests the Molecular Underpinnings of pH-induced Conformational Changes^*

Structural Modeling

We obtained a sequence alignment between TMs III-V and TMs X-XII from a structural alignment via SKA (34), and manually aligned the sequences of TMs I-II and TMs VIII-IX. The conformation of TM VI-VII in the model was retained from the crystal structure. We adjusted the sequence alignment to obtain a model that is consistent with two generic traits of membrane proteins (35): the hydrophobicity pattern (36, 37) and evolutionary conservation profile (38, 39), that is, hydrophobic and variable positions face the lipids, while conserved and/or strongly hydrophilic residues are embedded in the interior, mediating helix-helix and substrate interactions (supplemental Figs. S1 and S9). Constraints were applied to maintain the helicity of the helical segments of the crystal structure. We selected the lowest energy model from 2,000 candidate conformations, generated using Modeler, version 9v8, which uses the CHARMM22 force field (40).

Normal Mode Analysis using Elastic Network Models

In GNM and ANM (27, 29, 31, 41), the collective modes of motion are determined by the particular structural topology, represented by a contact matrix. The fluctuations of each residue and their inter-correlations in the slowest modes were obtained from GNM calculations, performed with an in-house program with the commonly used cutoff of 10 Å to indicate contact between Cα atoms. To predict the direction of the motion, we carried out ANM computations via the HingeProt webserver using the default cutoff of 15Å (42). We matched the corresponding GNM and ANM slow modes based on their mean-square fluctuations, to associate the size and direction of detected motions (supplemental Fig. S6). More details are available as supplemental data.

Intra-molecular Cross-linking at the Cytoplasmic End of NhaA in Situ

Site-directed intramolecular cross-linking was conducted in situ on membrane vesicles isolated from TA16 (43) cells expressing the various NhaA double-Cys replacements, as previously described (44). Membranes (2 mg of protein) were resuspended in a buffer (3 ml) containing one of the following homo-bifunctional cross-linkers, freshly prepared: 2 mm BMH (1,6-bis-maleimidohexane) or 1 mm pPDM (N,N′-p-phenylenedimaleimide) and processed at pH 6.5. Mutants expressing the double Cys replacements L138C/H256C and A142C/H256C were processed at pH 8.5 as well, whereas mutants expressing S146C/E252C were processed also at pH 9. The affinity-purified protein was treated with trypsin at pH 9 and resolved on SDS-PAGE to identify the proteolytic products according to (45). Because the only trypsin-cleavable site (Lys-249) is located between the two Cys replacements, trypsinolysis of untreated samples results in two tryptic peptides with faster mobility than that of the intact protein (17.24 and 32.5 kDa, designated high frequency and low frequency, respectively). On the other hand, intramolecular cross-linking results in one fragment whose mobility is equal to that of the intact protein (46). Each experiment was repeated at least twice with practically identical results. Notably, there was no intermolecular cross-linking that could have been recognized by a band of a mobility of the NhaA dimer.

Modeling the Periplasm-facing State of NhaA

In accordance with prior studies (8, 9), we postulated that NhaA consists of two pairs of pseudo-symmetric segments (Fig. 1A). This type of structural relationship can be used for modeling an alternate state if there is a distinct difference between the internal conformations of the pseudo-symmetric segments in the known state (22, 24, 25). While TMs III-V and TMs X-XII superimposed very well (Figs. 1A and supplemental Fig. S1A, referred to as segments 2 and 4), superimposing TMs I-II and TMs VIII-IX (Fig. 1A, segments 1 and 3, respectively) with the SKA program did not reveal a structural relationship suitable for modeling an alternate state (supplemental Fig. S1B). However, the comparison did reveal an intriguing difference; specifically, although TM II and TM IX are approximately of the same length, there is a difference in their conformations: TM II is only slightly kinked, while TM IX is more banana-shaped (Figs. 1A and supplemental Fig. S1D). Since in the crystal structure TMs II and IX line the cytoplasmic funnel, and TM II also contributes to the periplasmic funnel (8, 19), we reasoned that this conformational difference might contribute to the fact that in the crystal structure the funnel in the cytoplasmic side is larger than that in the periplasmic side. We also considered TMs I-II and III-V as one non-contiguous structural repeat, and TMs VIII-IX and X-XII as a second structural repeat, with each repeat containing five TM helices. It can be seen that these two repeats differ in the relative translation of the first two helices of each repeat with respect to the last three helices of the repeat (supplemental Fig. S1C).

In line with the aforementioned hypothesis, we were able to obtain a model-structure of NhaA in a distinct conformation by swapping the conformations of the structural repeats (Figs. 1B and supplemental Fig. S2). The two helical domains, the first consisting of TMs I, II, VIII, and IX and the second bundle including TMs III, IV, V, X, XI, and XII, are rotated relative to one another in comparison to their positioning in the crystal structure. This reflects the different relative positions of segments 1 or 3 with respect to segments 2 or 4, respectively (Figs. 1B and 2). This relative rotational change can be most clearly seen when superposing the two structures using only segments 2 and 4 (supplemental Fig. S3C) or only segments 1 and 3 (supplemental Fig. S3B). The remaining peripheral TMs VI and VII retained the same conformation in the model as in the crystal structure, and their exact positioning in the model is of lower confidence (supplemental text).

The primary consequence of the rotation of the two domains, and of the more local changes in the conformation of several helices, is that, in comparison to the crystal structure, the model is more open to the periplasm, whereas its cytoplasmic funnel is smaller in volume (Figs. 1, C and D and 2). In both conformations, the cytoplasmic funnel is lined by TMs II, IVc, V, and IX (Fig. 2A). In the model-structure, the domain rotation changed the positions of these TM segments relative to each other; this in turn results in closure of the cytoplasmic funnel. In particular, the cytoplasmic ends of TM II and TM IX pack up against IVc, displacing TM I, which now also packs against TM XII (Fig. 2A). Bending of TM II and straightening of TM IX may also contribute to the closure (Fig. 3). Interestingly, we also observe a decrease in the tilt of TMs IVc and V relative to the membrane normal, concurrent with an increased tilt of TM XII. As a result, a separate pathway from the cytoplasmic funnel on the cytoplasmic side, leading to the unwound region of the TM IV-XI assembly, is enlarged because of an increase in the distance between the cytoplasmic end of TM XII and those of TMs IVc and V (Figs. 2A and 3A). Notably, as TMs IV and V line both the cytoplasmic funnel and the pathway to the TM IV-XI assembly, the change in their relative orientations contributes to the closure of the funnel and also to the enlargement of the pathway at the model's cytoplasmic end.

On the periplasmic side, the funnel (Fig. 2B) expands due to the rotation of the domains and the altered conformations of TM II and TM IX in the model (supplemental Fig. S1D). The periplasmic funnel, lined mainly by TMs II and XIp in the crystal structure, is formed between TMs II, XIp, IX and XII in the model (Figs. 2B and 3B). In addition, a separate pathway to the periplasm, located within the TM VI-XI assembly, is enlarged somewhat by separation of TMs IVp and XIp (Figs. 2B and 3B). Since it lines both the periplasmic funnel and the putative pathway to the TM IV-XI assembly, TM XIp mediates both the opening at the periplasmic funnel and the expansion of the pathway toward the assembly.

Normal Mode Analysis of the Crystal Structure

We investigated the global, cooperative (correlated) dynamics of the crystal structure using normal mode analysis. We focused on the two slowest GNM modes, i.e. those with the most significant contribution to the overall motion (supplemental Fig. S4), and obtained the mean-square fluctuations of the residues and inter-residue cooperative dynamics. The directions of these motions were inferred from the corresponding ANM modes (supplemental Fig. S5).

The Slowest Mode of Motion

The slowest GNM mode (GNM1) revealed hinges that reside at, or in the vicinity of, evolutionarily conserved and functional residues, (supplemental Fig. S6A and Table S1), as is typical of hinges (47). Almost all the detected hinges were located along the putative axis that separates NhaA into its two structural domains (supplemental Fig. S6B). The GNM cooperative dynamics were concentrated around hinges situated at “inter-domain” connection points. The proposed motions within each helical bundle were positively correlated, while being anti-correlated to the proposed motions of the other bundle (Fig. 4C). The corresponding ANM mode (ANM1, supplemental Fig. S5A) manifested a rotational or twisting motion of the domains relative to one another, around the central axis (Fig. 4, A and B and supplemental Movie S1). These deformations altered the general shape of the two funnels, while preserving their volume. Accordingly, residues that originally lined the pore still retained their roles, albeit changed their exact positioning around the funnels.

The Second Slowest Mode

The hinge regions in the second slowest GNM mode (GNM2, supplemental Fig. S6C) were also enriched with highly conserved residues that are important for function (supplemental Table S1). The hinges in the TM IV-XI assembly and flanking helices, along with TM IX, roughly in the membrane midplane, separate these TM segments into cytoplasmic and periplasmic rigid parts (Fig. 5A). As the modes are orthogonal, independent of one another (27, 29, 31, 41), the hinge plane in GNM2 (Fig. 5A) is distinct from that of GNM1 (supplemental Fig. S6A). The dynamics of the cytoplasmic rigid parts were positively correlated, while anti-correlated with the dynamics of the periplasmic ends (Fig. 5B). Three positively correlated, highly fluctuating, rigid segments lined the cytoplasmic funnel; these are the cytoplasmic ends of TMs II-III, TMs IVc-V, and TMs VIII-IX (Fig. 5, A and B). In the corresponding ANM2 mode (supplemental Fig. S5B) the TMs II-III segment moved away from the center of the funnel, and the cytoplasmic ends of TMs IVc-V and TMs VIII-IX approached one another at the center of the funnel (Fig. 5, C and D, supplemental Movies S2 and S3).

Agreement between the Two Computational Approaches

By superimposing the crystal structure and the periplasm-facing model, we were able to derive a rough estimate of the conformational changes leading to the transition between them (Fig. 2). We compared these trends of motion to the two slow modes of motion, keeping in mind that normal mode analysis is capable of revealing the relative magnitude of the fluctuations and their directions but not their precise amplitudes (31). Remarkably, we detected an overall qualitative agreement between our two independent computational results (Figs. 6, supplemental Fig. S7 and Movies S4–S6). According to both methods, the cytoplasmic ends of TMs IV-V and TMs VIII-IX approach each other near the center of the cytoplasmic funnel, effectively closing it off (Figs. 2A, 5C, and 6 and supplemental Movie S4). Notably, straightening of TM IX was predicted by both the second mode of motion and the periplasm-facing model (supplemental Fig. S7). Likewise, a rotational motion suggested by ANM1, which could be described as twisting around an axis running through the two funnels and separating the two domains (Fig. 4, A and B and supplemental Movie S1), can also be detected when examining the transition between the crystal structure and model (Fig. 2 and supplemental Movies S5 and S6). Still, two features of the periplasmic-facing model were not suggested by the normal mode analysis, namely, opening of the periplasmic funnel, and moderate opening at the cytoplasmic and periplasmic entrances of TMs IV-XI (Figs. 2 and 3). These differences may be attributed to the limitations of normal mode analysis, discussed below.

Biochemical Study of Conformational Changes in Situ

Tests of accessibility to various SH-reagents of single Cys replacements and cross-linking between double Cys replacements, as a function of changes of various conditions, are widely-used to identify conformational changes of membrane proteins in situ (48, 49). For NhaA, such tests have been conducted in the physiological pH range (pH 6.5–8.5) when NhaA is activated, measuring response to pH level and/or different ligand concentrations, and the results have been compared with the pH 4 crystal structure.

Accessibility Measurements as a Function of pH

As reactivity of the membrane-permeant SH reagent N-ethyl maleimide (NEM) requires water, accessibility of Cys replacements to NEM within a protein implies the presence of a water-filled cavity connected directly or not with the periplasm or cytoplasm. The complementary reagents 2-sulfonatoethyl methanethiosulfonate (MTSES) and 2-(trimethyl ammonium)-ethylmethanethiosulfonate (MTSET) are charged, membrane-impermeable, and comparable in size to hydrated Na⁺ (19). Therefore, modification by these reagents indicates the presence of solvent-accessible funnels within the protein through which the reagents can reach the Cys replacement. Accessibility data for these reagents as a function of pH are available for Cys replacements on TMs II, IV, V, VIII, IX, X, and XI. The model agrees with most of the experimental data, as follows: (i) residues on TM II line the cation pathway (19); several positions accessible to the periplasmic side at physiological pH are not exposed in the crystal structure at pH 4, but are exposed in the periplasm-facing model (residues Asn-64, Asp-65, Met-68, Phe-71, and Phe-72) (Fig. 7). (ii) Except for the N-terminal residues of TM VIII (G223C, V224C, and H225C), no TM VIII Cys mutants were accessible to MTSES in intact cells, implying that this helix lines only the entrance to the periplasmic funnel (50). Furthermore, the accessibility to MTSES was not affected by pH. We observed that, in agreement with these data, no TM VIII positions, other than G233, V224, and H225, were significantly accessible from the periplasm, either in the model or in the crystal structure (Fig. 7). (iii) In TM IX, pH-dependent conformational changes were identified at the cytoplasmic side at Lys-249 and Glu-252, and the cytoplasmic funnel deepens at physiological pH (16), which should correspond to a cytoplasmic-facing, rather than a periplasm-facing, conformation, as depicted by the model. From the periplasmic end of TM IX, no MTSET-accessible residues were detected up to and including Phe-267 (16). Of these residues, only Phe-267 lines the periplasmic funnel in the model-structure (Fig. 7B). However, it is located at the bottom of the funnel, and in a narrow region, which would be inaccessible to the large reagent. (iv) Several studies examined the accessibility of residues in the TM IV-XI assembly and surrounding helices. Accessibility assessments of TM IV residues (51), as well as scans of TMs IV and V in HPNhaA, a close homologue from Helicobacter pylori (52, 53), showed that only the equivalents of Ile-128 and Asp-164 are accessible to the periplasmic side (52, 53). A few positions in the center of TM X were accessible to reagents from the cytoplasmic end at pH 7.5 (44). Last, cross-linking was obtained between several residue pairs in the proximal TM helices IV and XI (55) indicating that positions in TMs IV and XI within the membrane core are exposed as well. While none of these positions is directly accessible in the crystal or in the model, the cytoplasmic and periplasmic pathways to the assembly region are enlarged in the model compared with the crystal structure (Fig. 3A), potentially exhibiting a trend of motion toward the formation of accessible pathways.

Cross-linking

To examine the predicted conformational changes of TMs IVc-V and TM IX, we conducted cross-linking experiments at two pH conditions between three Cys pairs on TM IVc or the TMs IVc-V loop, and TM IX (L138C-H256C, A142C-H256C, and S146C-E252C). For the former two pairs, we conducted the experiments in pH 6.5 and 8.5, whereas for the latter pair the alkaline range was set to 9, as the E252C causes an alkaline shift in the pH activity range of NhaA (56). Two bifunctional cross-linking reagents were used: PDM and BMH, with spacing arms of 9.2–12.3 Å and 3.5–15.6 Å, respectively. According to the Cα-Cα distances of these residue pairs in the crystal structure, in the periplasm-facing model, and in the crystal structure after the most extreme deformation along the ANM2 motion, only one pair, namely S146C-E252C, is predicted to change sufficiently that the difference would be detected by cross-linking (supplemental Table S2 and Fig. 8, C–E). Indeed, S146C and E252C cross-linked at pH 9 but not 6.5, while the other pairs cross-linked at both pH levels (Fig. 8, A and B). These results suggest that upon pH change from 6.5 to alkaline range, TMs IVc and IXc approach each other, in agreement with both the suggested periplasm-facing model and ANM motion (Fig. 8, C--E), indicating that the conformational changes predicted herein are pH-induced.

Limitations of the Computational Methodologies

The major structural alteration predicted herein could probably not be accessed using molecular dynamics simulations, nor using other commonly used approaches that are based on all-atom description, because of the large size of the transporter, the restricted simulation time currently available, and the complexity of the conformational change (10, 57). Thus, we were compelled to use coarse-grained approximations that have their own limitations. The comparative modeling procedure used to construct the periplasm-facing conformation predicts the overall architecture of this state, at an approximate accuracy resolution of ∼3 Å (58). It also cannot take into account any additional structural rearrangements beyond those contained within the pseudo-symmetry. The GNM/ANM analyses predict the general trends of motion, while overlooking the details (28, 30, 31, 47). Moreover, the ANM allows examination of the individual modes in three-dimensional, but the integrated motion, consisting of a superimposition of all modes of motion, cannot be predicted, as this requires correctly combining each mode's phase and amplitude. Additionally, the motion revealed by normal mode analysis covers a limited scope of the conformational landscape; taking the NhaA structure as the energetic minimum, we survey its surrounding equilibrium dynamics, but cannot cross high energetic barriers (59). Lastly, although the two slowest modes appear to describe the transition well, some contribution of other modes may be required to accomplish the whole transition. Nevertheless, as several lines of experimental evidence support the validity of both analyses, we maintain that the computations are useful for deriving hypotheses to be tested in experiments.

Motion, Transition, and Alternating Access

Transporters alternate between inward- and outward-facing conformations, often illustrated as rocking of two independent domains (32, 33). In this study, we detected two types of motions. Directly consistent with this alternating-access concept, in the periplasm-facing model and in the ANM2 motion, segments lining the cytoplasmic funnel approach its center in a coordinated manner (Figs. 2A and 5C and supplemental Movies S2-S4), while the model-structure also shows simultaneous opening of the periplasmic funnel (Figs. 2B and 7). In GNM1, the two domains exhibit opposite dynamical correlations, each rotating around a putative symmetry axis in the corresponding ANM1 motion (Fig. 4, A and B and supplemental Movie S1), similar to the relative rotation of the two domains observed when comparing the periplasm-facing model with the crystal structure (supplemental Fig. S3, B and C). Interestingly, measuring the positioning of the two domains revealed a rotation of 22° in the model relative to the crystal structure. This is comparable to the helical rotations observed when comparing different structural conformations of LeuT (ranging between 9° and 45°, with most of the helices rotating ∼20°) (60), and those measured for the inter-domain rotations between the crystal structures and models of GltPh (37°, (24)) and LacY (51°, (25)). The division into two domains is reminiscent of the independent domains illustrated in classical alternating-access, and we thus suggest that the rotational motion could be related to switching between states as well. Although this might not be the intuitive motion one would imagine in order to switch between conformations, this could be an interesting alternative to a simple rocking (Fig. 4, C and D).

Structural Implications of the Periplasm-facing Scaffold

In the periplasm-facing model, the cytoplasmic funnel adopts a closed conformation (Figs. 1D and 2A), in agreement with cross-linking data (Fig. 8). While the periplasmic funnel in the model is larger than that in the crystal, it does not widen as much as the cytoplasmic funnel in the crystal structure (Figs. 2 and 7). Nevertheless, the model is in excellent agreement with TM II accessibility data, which show that Phe-71 and Phe-72 are accessible only to NEM (19), indicating that they reside in a narrower cavity than preceding residues, and that the periplasmic funnel ends at Phe-72 (19).

Interestingly, the structural model also predicts some opening of putative pathways on the cytoplasmic and periplasmic sides that lead into the TM IV-XI assembly (Figs. 2 and 3). The location of these pathways is consistent with observations from MD simulations, which suggest separate cation and proton pathways through the funnels and via the TM IV-XI assembly region, respectively (10). It is also consistent with the observation that several residues in the centers of TMs IV, X, and XI were exposed under physiological conditions (44, 51). Nevertheless, the opening of the periplasmic pathway in the model (Fig. 3B) is insufficient to expose the functionally important Asp-164 (8, 10–12) to the periplasm, although relatively small additional conformational changes, enlarging the pathway, would increase the periplasmic accessibility of this residue. Notably, the HpNhaA equivalent of Asp-164 was accessible to the periplasm upon Cys mutation (53). As no other TM V positions were accessible to SH-reagents from the periplasmic side upon Cys mutation, it is apparent that an entire face of TM V does not line the periplasmic funnel, an observation consistent with our periplasm-facing model (Figs. 1, C and D and 2). We thus propose that Asp164 could be exposed to the periplasmic side through pathway within the TM IV-XI assembly, suggesting how the residue may participate in cation binding.

pH-induced Conformational Changes?

Studies involving diverse experimental approaches have demonstrated that NhaA undergoes conformational changes in response to pH alteration (13–21). Two general views of this pH-dependent activation have been postulated, although they are not mutually exclusive. According to the first view, the apparent pH-induced activation results from direct competition between H⁺ and Na⁺ (or Li⁺) (61). According to the second view, the protons are not only ligands, but are also involved in pH regulation of NhaA's activity (13–21).

Consistent with the latter view, recent studies have shown two distinct conformational changes: one in response to pH, and the other, at alkaline pH, in response to ligand addition (17, 21). CryoEM projection maps of two-dimensional crystals of NhaA indicated conformation changes induced by pH change (4–8) at the funnel region, and other ligand-induced changes at alkaline pH in the TM IV-XI assembly region, ascribed to TM IVp (17). Correspondingly, a site-directed fluorescence study showed that TM IVc is affected by the pH-induced conformational change, and TM XI (at the active site) by ligand-induced changes (21). We suggest that the periplasm-facing model and the deformations depicted by the slowest modes of motion correspond to the pH-induced conformational change observed in these prior studies. This suggestion is supported by several lines of evidence. First, the extent of structural changes at the cytoplasmic funnel of the model-structure is larger than in the periplasmic funnel (Figs. 2 and 7), which can account for the increase in density detected by the cryoEM maps upon increasing pH (see Fig. 7 in Ref. 17). Second, site-directed Trp fluorescence revealed a pH-induced conformational change at TM IVc (21), in agreement with the model-structure (Figs. 1, C and D and 2A) and second mode of motion (Fig. 5C). Third, our cross-linking data were pH-dependent (Fig. 8A), indicating closure at the cytoplasmic funnel at pH 9. Further supporting evidence includes the accessibility of TM II residues at alkaline pH (19), which is consistent with the periplasm-facing model (Fig. 7). Moreover, mutations in various residues in TMs II and IX shifted the pH dependence profile (16, 19), as did intermolecular cross-linking of V254C of TM IX (62). The fact that TMs II and IX undergo substantial structural changes in our model further implies that the modeled conformation corresponds to pH-induced alterations. Finally, Lys-249 is the only cleavable site of all putative sites in NhaA, and cleavage was shown to be pH-sensitive, occurring in alkaline pH and not in acidic pH (13, 46). The extensive rearrangements predicted at the cytoplasmic ends of the helices by our analysis might also explain the pH-dependent sensitivity of Lys-249 of TM IX to trypsin cleavage (13, 46): although the accessibility of the residue itself remains similar in both the crystal structure and model, the extensive rearrangements at the cytoplasmic end may potentially affect the ability of trypsin to bind to the cleavage site.

Previous experimental data indicate that solvent-accessible pathways are formed around the TM IV-XI assembly at physiological pH (16, 51–53, 55), possibly creating ion-accessible pathways (these pathways have also been observed in MD simulations (10)). The experimental data can be grouped into three types: (i) accessibility to SH-reagents of some residues in TMs IV (51, 52) and V (53) in NhaA and in its homologue from Helicobacter pylori; most of these residues are situated at the periplasmic ends and accessible from the periplasmic side; (ii) accessibility to SH-reagent of a few residues in TM X in NhaA (16); (iii) cross-linking between residues from TMs IV and XI, indicating also that these residues are exposed (55). The ligand-induced conformational changes detected in the TM IV-XI assembly region by cryoEM (17) and by site-directed tryptophan fluorescence (21) can potentially account for these accessibility measurements. While the conformational changes in the periplasm-facing model were larger in the funnels than in the TM IV-XI assembly and flanking helices, we did observe structural rearrangements in the latter region as well (Figs. 2 and 3). Although the current conformation cannot account for the above mentioned accessibility of residues in TMs IV, V, X, and XI (16, 51–53, 55), it does suggest a clear trend of opening toward the assembly region from both sides of the membrane (Fig. 3). Assuming that ligand-induced changes indeed induce solvent-accessible pathways within the assembly region, we conclude that the model is essentially compatible with these data.

Previous studies have suggested that a “pH sensor” perceives the pH signal (a change in intra-cellular pH) that is then transduced to activate NhaA (64). Based on the crystal structure (8), computation (65), and experimental data (reviewed in Ref. 12), the pH sensor was identified at the cytoplasmic half of TM IX, loop VIII-IX, and the cytoplasmic end of TM II. Consistent with this suggestion, the cytoplasmic ends of TMs II and IX indeed undergo pronounced changes in their internal structural conformation in the model compared with the crystal structure. Our cross-linking data indicate that the modeled structural changes are indeed pH-induced.

On the Transport Mechanism

Computational studies have previously suggested that the evolutionarily conserved residues Glu-78 and Glu-82 of TM II are involved in pH sensing (57, 65). Furthermore, the mutation to Cys of each of these acidic residues, as well as of the conserved Glu-252 of TM IX, increased the apparent K_m for Na⁺ and caused an alkaline shift in the activity profile (19, 56). A previous computational study calculated the pK_a of charged residues in the crystal structure, predicting values of 5.4, 10.4, >15, >15 and 2.7 for the acidic Glu-78, Glu-82, Asp-163, Asp-164, and Glu-252, respectively, along with values of 15.4 and 9.3 for the basic Arg-81 and Lys-153, respectively (65). Interestingly, in contrast to the crystal structure, the modeling procedure placed the funnel-lining residues Glu-78, Glu-82, and Glu-252 in close proximity to the highly conserved Lys-153 of TM V, which faces into the funnel from the other side (supplemental Fig. S8). Electrostatic interactions among these residues could stabilize the conformation depicted by the model-structure. In light of these structural features, in conjunction with prior computational and experimental data, we propose a working hypothesis for the pH-induced conformational changes in NhaA within the context of an alternating-access mechanism (Fig. 9). At acidic pH, one or more of the conserved Glu-78, Glu-82, and Glu-252 residues are presumably protonated, potentially forming hydrogen bonds with the highly conserved Arg-81 of TM II (supplemental Figs. S8A and 9A). The model predicts that upon pH elevation, TMs II and IX change their conformations; concurrently, the glutamates are expected to become charged, attracting in turn Lys-153 (supplemental Figs. S8B and 9B). In this conformation, which would correspond to the periplasm-facing model, Asp-163 and Asp-164 are buried in the protein core, presumably remaining protonated. A second, ligand-induced change could expose one or both aspartates to the cytoplasm via motion of TM IVp, inducing rapid exchange of protons for Na⁺ (Fig. 9C), followed by swapping into an Na⁺-bound, periplasm-facing conformation (Fig. 9D). Upon the exchange of Na⁺ for protons, the antiporter could return to the partly-occluded periplasm-facing state (Fig. 9B). Overall, this mechanism connects the large conformational changes of TMs II and IX, predicted by our calculations, to the evidence indicating a response of specific residues in these TM helices to pH changes (19, 56, 57, 65), and it also connects these changes to pH-induced conformational changes in TM IV, and to the ligand-induced response detected at the TM IV-XI assembly (17, 21). The actual mechanism may be more complex and include more states.