The conserved lysine residue in transmembrane helix 5 is pivotal for the cytoplasmic gating of the L-amino acid transporters

Abstract

L-Amino acid transporters (LATs) play a key role in a wide range of physiological processes. Defects in LATs can lead to neurological disorders and aminoacidurias, while the overexpression of these transporters is related to cancer. BasC is a bacterial LAT transporter with an APC fold. In this study, to monitor the cytoplasmic motion of BasC, we developed a single-molecule Förster resonance energy transfer assay that can characterize the conformational states of the intracellular gate in solution at room temperature. Based on combined biochemical and biophysical data and molecular dynamics simulations, we propose a model in which the conserved lysine residue in TM5 supports TM1a to explore both open and closed states within the cytoplasmic gate under apo conditions. This equilibrium can be altered by substrates, mutation of conserved lysine 154 in TM5, or a transport-blocking nanobody interacting with TM1a. Overall, these findings provide insights into the transport mechanism of BasC and highlight the significance of the lysine residue in TM5 in the cytoplasmic gating of LATs.

Introduction

L-Amino acid transporters (LATs) (SLC7A5-11 and SLC7A13 in humans) play a key role in the cellular uptake and distribution of amino acids. LATs are thus relevant for a wide range of physiological processes, including protein synthesis, signal transduction, and the maintenance of cellular homeostasis. Defects in LATs lead to a range of disorders, including aminoacidurias and neurological disorders (1). In this regard, mutations in SLC7A7 and SLC7A9 cause the rare aminoacidurias lysinuric protein intolerance (LPI) (2) and cystinuria (3), respectively. Similarly, mutations in SLC7A8 are associated with age-related hearing loss (4) and cataracts (5), and the first mutation in SLC7A10 has recently been described to be associated with the neurological disorder hyperekplexia (6). Moreover, overexpression of SLC7A5, SLC7A8, and SLC7A11 has been observed in many types of cancer (7–9), where these transporters confer a metabolic advantage for tumor growth and progression. These examples underscore that a fundamental understanding of LAT transport mechanisms is highly relevant not only for understanding physiology but also for clinical applications.

LATs feature the so-called APC fold, which is characterized by a 5 + 5 inverted repeat of transmembrane (TM) helices (TM1–TM5 and TM6–TM10 in LATs; Fig. 1A). At the heart of the inverted repeat lie TM1 and TM6, each with an unwound segment, housing the main binding site of the substrates (10), dividing their TM helices into two parts (TM1a–TM1b and TM6a–TM6b from the N- to C-terminus). These half-TMs are part of cytoplasmic and extracellular gates, especially TM1a on the cytoplasmic side and TM6a on the extracellular side.

In recent years, structural studies have been complemented by biophysical techniques that can monitor changes in conformational dynamics even at room temperature in solution. A single-molecule Förster resonance energy transfer (smFRET) (11–15), for example, has been successfully used to study the dynamics of membrane transporters at the molecular level (16, 17–23). Relevant to this work, smFRET could help to show that the LeuT transporter, a bacterial homolog of the neurotransmitter sodium symporter (NSS) family with an APC fold (24), undergoes a rocking motion and switches from an outward-facing to an inward-facing conformations (25–30). The proposed rocking motion of the bundle domain (TM1, TM2, TM6, and TM7) versus the scaffold domain (TM3, TM4, TM8, and TM9) between outward-facing and inward-facing conformations is consistent with the general mechanism observed in other members of the APC transporter superfamily, including LAT structures (31–42).

Despite sharing similar fold and some mechanistic features, the NSS and LAT families of transporters show low sequence identity and distinct transport mechanisms. NSS transporters are sodium symporters (43), whereas LATs are amino acid exchangers (1) that transit between inward- and outward-facing conformations with substrate bound at the binding site. Notably, human LATs and the bacterial LAT BasC show lower K_m for substrates in the outward-facing conformation (μM range) than in the inward-facing conformation (mM range) (44, 45), allowing for the harmonization of substrate concentrations on both sides of the cell membrane (46, 47). In both BasC and human LATs, two conserved residues—a tyrosine in TM7 (Tyr236) and a lysine in TM5 (Lys154 in BasC)—occupy positions similar to those of the two sodium ions (Na1 and Na2) in NSS transporters, and both residues are crucial for the asymmetric substrate affinity (31). Na2 in LeuT is known to be involved in the opening and closing of the inner gate (48, 49). Interestingly, the conserved lysine residue in TM5 is the key for LAT transport activity. Moreover, the homologous lysine in SLC7A7 is mutated to glutamate (Lys194Glu) in a patient with LPI (50). Similarly, mutation of the homologous lysine to glutamate or alanine in BasC and human SLC7A10, or to glutamine in human SLC7A5 (Lys204Gln), results in a significant loss of function (31, 51). Whether this conserved lysine residue plays a role in the inner gating of LATs is unknown.

In the present work, to monitor cytoplasmic movement of BasC, we developed an smFRET assay that can characterize the conformational states of the ligand-free (apo) and ligand-bound (holo) forms. The assay uses fluorophore labeling of TM1a–TM12 and TM1a–TM4 (Fig. 1A) to track the motion of TM1a. Using specific nanobodies (Nbs) and point mutations of the conserved lysine in TM5 reveals a key role of this residue for gating at the cytoplasmic side of BasC. Based on the combined data and molecular dynamics (MD) simulations, we propose a model in which the conserved lysine residue in TM5 supports TM1a to explore both open and closed states within the cytoplasmic gate under apo conditions. This equilibrium can be altered by the substrate addition, mutation of the TM5 lysine, and blocked using a specific Nb that interacts with TM1a. Overall, these findings provide insights into the transport mechanism of BasC and highlight the significance of the lysine residue in TM5 in the cytoplasmic gating of LATs.

Results

BasC samples’ different conformations under apo conditions

To monitor the movement of the inner gate of BasC in the solution at room temperature, we developed an smFRET assay that tracks its conformational changes. To this end, two distinct double-cysteine variants of BasC were designed. Residue Ile7 in TM1a was present in both (Fig. 1A), because TM1a is predicted to move strongly during the closing of the inner gate in LATs. The equivalent position to Ile7 in BasC is Leu53 in human LAT SLC7A5, which differs by 11.5 Å between the inward-open (Protein Data Bank, PDB ID 6IRS) (32) and outward occluded structures (PDB ID 7DSQ) (41). Moreover, smFRET studies of LeuT seeking to track the inner gate movement used a similar position in TM1a (25, 27, 29, 30). In BasC, Ile7 in TM1a was combined with two other residues in the protein: Thr120 in TM4, forming the TM1a–TM4 pair, and Cys427, the only natural cysteine in BasC, in TM12, forming the TM1a–TM12 pair (Fig. 1A). Both BasC variants with cysteine residues in TM1a–TM4 (mutations Ile7Cys–Thr120Cys–Cys427Ala) and TM1a–TM12 (mutation Ile7Cys) retained ∼50% activity, in proteoliposomes (PLs), even when they were labeled with maleimide-conjugated fluorophores used later in FRET experiments (Fig. S1).

We used maleimide-conjugated FRET fluorophores to stochastically label the cysteine pairs in the two protein variants. Fluorophore-labeled proteins solubilized in detergent DDM were then studied in free diffusion using a confocal microscope built in-house (52). Figure 1B and C shows representative data of BasC for TM1a–TM12 (Fig. 1B, dye pair Alexa 546/Alexa 647) and TM1a–TM4 (Fig. 1C, dye pair SCy3/SCy5). We used alternating laser excitation, µsALEX (53, 54), to derive information on both the apparent FRET efficiency E* and brightness ratio S* (Fig. 1B and C). The 2D histogram allowed us to focus our analysis on donor–acceptor-labeled BasC by selecting an S* region between 0.3 and 0.8, indicated by a dashed rectangle.

In the TM1a–TM12 pair, we observed a bimodal distribution with two distinct E* populations (Fig. 1B). Fitting of the distribution with two Gaussian functions gave mean E* values of 0.52 (σ = 0.09) and 0.71 (σ = 0.06) for both states (Fig. 1B). The TM1a–TM4 pair showed only a single peak with a mean E* value of 0.56 (σ = 0.09; Fig. 1C). These results under ligand-free apo conditions suggest that the protein can adopt at least two distinct conformations (Fig. 1B). The observation of the single peak of the TM1a–TM4 pair (Fig. 1C) is attributed to a reduced sensitivity of the sCy3–sCy5 pair, since the connection axis of the residues in TM1a–TM4a does not lie directly within the direction of movement of TM1a. Additional evidence for this interpretation comes from the Cy3–sCy5 labeling of the TM1a–TM12 pair, where a broader distribution is also observed, yet not two discrete populations (Fig. S2A), showing the higher sensitivity of the Alexa dyes in the assay.

Substrate addition induces a closing movement of TM1a

In the next step, we used the smFRET assay to study the effect of the substrates L-alanine (L-Ala) and L-serine and the competitive inhibitor L-glutamine ((45) and Fig. S2C and D) on BasC. Upon the addition of L-Ala, we observed an increase in the high-FRET population and a slight shift in the low-FRET population for TM1a–TM12 labeling (Fig. 2A). Quantitatively, the substrate increased the high-FRET fraction, where TM1a and TM12 were closer, from ∼25 to 45% (Fig. 2B). Under similar conditions, the peak observed for TM1a–TM4 shifted toward higher E* values upon addition of both L-Ala and L-serine (Fig. 2C). In contrast, the addition of L-glutamine did not cause a shift (Fig. 2C). To obtain statistics from independent experiments, we normalized the ligand-induced E* shifts and calculated ΔE* as the difference between E* in the presence or absence of ligand (Fig. S3). Figure 2D shows the observed peak shifts (ΔE*) of TM1a–TM4, confirming a statistical increase in ΔE* > 0 for substrates but no change in ΔE* ≈ 0 for the inhibitor. The increase observed in E* indicates a decrease in the interprobe distances between the labeled positions in TM1a and TM4, consistent with a closing movement of the inner gate. Indeed, positions TM1a–TM12 labeled with SCy3 and SCy5 showed a similar tendency when substrate was added (Fig. S2A).

In addition, we measured the substrate-induced movement of the TM1a–TM4 pair under increasing concentrations of L-Ala. The response showed saturation >100 mM with an estimated EC50 around 40 mM (Fig. S2B). This value is very similar to the reported dissociation constant K_d of BasC (∼35 mM), determined by thermostability-based substrate-binding assays in detergent micelles (31). The smFRET results support the notion that substrate binding to BasC favors a closer conformation of TM1a.

Substrate-induced movement is prevented by specifically blocking TM1a

To confirm the involvement of TM1a in the substrate-induced movement detected by smFRET, we used three specific Nbs targeting the cytoplasmic side of BasC: Nb53, Nb71, and Nb74. These Nbs blocked the efflux of 10-µM L-Ala in BasC-reconstituted PLs (Fig. S4A and figure legend), and structural studies confirmed that they interact on the cytoplasmic side of the transporter in inward-facing conformation but at different regions (Fig. 3C and D and (31)).

The incubation of BasC with Nb71 resulted in the disappearance of the high-FRET population for the TM1a–TM12 pair (Fig. 3A) and fully abolished ΔE* effects of L-Ala in the TM1a–TM4 pair (Fig. 3B). In the low-resolution cryo-electron microscope (cryo-EM) volume of BasC complexed with Nb71, Nb71 established direct interactions with both TM1a and TM6b (Fig. 3C), explaining the complete reduction of movement of the inner gate. Interestingly, the incubation of BasC with Nb53 did not affect the ratio of the two FRET peaks in the BasC TM1a–TM12 pair (Fig. 3A) or alter the effects of L-Ala on the TM1a–TM4 pair (Fig. 3B). In cryo-EM 2.9 Å volume, Nb53 showed no interactions with TM1a or the rest of the bundle domain (Fig. 3D)—meanwhile, a substrate-compatible volume is present in the binding site (Fig. S6I and J), consistent with its lack of impact on the L-Ala-induced movement of TM1a. Additionally, the incubation of BasC with Nb74 only partially diminished the L-Ala-induced movement of TM1a within the intracellular gate (Fig. S4B). As previously reported, Nb74 did not interact with TM1a but did with TM6b in the bundle domain (PDB ID 6F2G (31)). Recent structures of the human LAT transporter Asc1 in the inward-facing conformation, with TM1a semi-occluded as the sole closed element of the inner gate (55, 56), suggest that the tilting of TM1a monitored here by smFRET analysis might represent the initial step in the gating of the LAT cytoplasmic gate.

The structural information is crucial for understanding the observed behavior of the three Nbs in smFRET experiments and confirms that the L-Ala-induced shift in smFRET experiments corresponds to the closing of TM1a, as part of the cytoplasmic gate. These results strongly support the notion that the substrate-induced changes in FRET efficiency E* and the distributions reflect the occlusion of TM1a when the substrate interacts with the binding site. Indeed, specific blocking of TM1a with Nb71 prevented this movement.

The fully conserved lysine residue in TM5 (Lys154) is needed for the closing of TM1a

The lysine residue in TM5, i.e. Lys154 in BasC, is a conserved residue among LATs that is described as critical in the transport mechanism (31). To elucidate the role of Lys154 in the substrate-induced movement of TM1a in BasC, we conducted additional experiments using a Lys154Ala variant (Fig. 4). smFRET experiments using the TM1a–TM12 and TM1a–TM4 pairs were performed within the Lys154Ala BasC variant. Interestingly, the two distinct populations observed in the TM1a–TM12 pair in wild-type (WT) BasC were not detected in the presence of Lys154Ala (K154A) under apo conditions or after substrate addition (Fig. 4A and B). Moreover, the addition of L-Ala failed to induce any shift in E* for the TM1a–TM4 pair when the Lys154Ala mutation was present (Fig. 4C and D). These results reveal that Lys154 plays a crucial role in TM1a: (i) allowing it to sample different conformational states of the inner gate even in the absence of substrate and (ii) mediating the closing of the inner gate in BasC upon substrate binding.

Lys154 stabilizes the unwound segment of TM1a

To allow further interpretation of our findings, we performed atomistic MD simulations (Fig. 5), which revealed how the Lys154Ala mutation destabilizes the substrate-binding site by increasing the distance between the unwound segments of TM1 and TM6 in apo conditions (Fig. 5A and B). This effect was attenuated in the holo structure (Fig. 5C), but remained insufficient to allow binding of the substrate in BasC. Indeed, we observed a reduction in the strength of hydrogen bonds between the substrate 2-amino isobutyrate (2AIB) and the residues belonging to the binding site (Fig. 5D) located on TM1 and TM6, as described in the solved inward-facing conformation bound to 2AIB (31). In particular, TM1 residues were more affected by this reduction, as reflected by the decrease in the percentage of hydrogen bond occupancies shown in Fig. 5D. As a result, the substrate was released from the binding site in two of the three replicas of the Lys154Ala mutant, whereas it remained within the binding site in the three trajectories for WT BasC (Fig. S7).

To dissect further the mechanism of action, we explored the capping effect of the positive charge of Lys154 on the C-terminal negative pole of TM1a helix. MD simulations performed with the nonprotonated uncharged ε-amine of Lys154 (Lys154(0)) showed the same behavior of the Lys154Ala mutation: destabilization of the unwound segment of TM1, substrate-binding weakness, and dissociation of the substrate (Figs. 5 and S7), in agreement of the MD simulations of mutant Lys204Gln in human LAT1 (57). Of note, the proximity of the εN atom of BasC Lys154 to the backbone of Ile18 in TM1 remained at H-bond distance to the O atom of the Ile18 backbone with protonated Lys154 (Lys154(+)), but the occupancy of this H-bond was low (∼20%) during MD trajectories (Fig. S8). In contrast, proximity of the side chain of Lys154 uncharged to Ile18 was lost in the simulations (Fig. S8). These findings support the notion that capping of TM1a by Lys154 is a major factor in stabilizing the unwound segment of TM1, allowing proper binding of the substrate and, probably, permitting the next step of the transport mechanism: the closing of the inner gate.

Discussion

Based on our combined dataset and MD simulations, we propose that the conserved lysine residue in TM5 helps TM1a to explore both open and closed states within the cytoplasmic gate under apo conditions, leading to an equilibrium with an open:closed ratio of ∼3:1 in detergent micelle, which can be altered by substrates, mutation of the conserved lysine in TM5, or blocked by Nb71 that binds TM1a. The increased population of the closed state upon substrate addition is consistent with smFRET studies of the gating of the immobilized symporter LeuT, where transitions between the two states occur dynamically (26). L-Ala substrate shifted the TM1a gating equilibrium toward the closed state in BasC (∼1:1 ratio of open to closed) in detergent micelle, whereas the LeuT substrate sodium pushes the equilibrium even more toward the closed state (ratio of 1:2 of open to closed) and dramatically reduces state transitions (26). The gating of TM1a reflects similarities and differences between LeuT and BasC. Both transporters share the APC protein fold (24, 31), but they use distinct transport mechanisms. LeuT, as a bacterial model of NSSs, is an obligatory amino acid:Na⁺ symporter, whereas BasC, as a bacterial model of LATs, is an obligatory amino acid exchanger (45).

In LeuT, the binding of sodium ions closes the cytoplasmic gate, stabilizes the outward-facing conformation (48) and facilitates the binding of the transported substrate (e.g. L-leucine) to trigger conformational transitions toward the occluded conformation. Then, partial unwinding of TM5 allows the dissociation of Na2, thereby promoting the opening of the cytoplasmic gate for the release of L-leucine into the cytoplasm (58).

In BasC, the lateral chain of Lys154, with ε-amino group occupying a similar position to Na2 in LeuT, is involved in establishing the dynamic equilibrium of open and closed states of TM1a. This equilibrium allows sampling of the closed state by the substrate (e.g. L-Ala), which is necessary to trigger amino acid efflux from the cell, and here, we establish that the highly conserved lysine residue in TM5 is central to this mechanism in LATs. Indeed, the Lys154Ala mutation in BasC significantly reduces V_max to <10% and increases the external K_m value for L-Ala 10-fold, without altering the cytoplasmic K_m value (31). These findings suggest that Lys154 plays a crucial role in a rate-limiting step in the BasC transport cycle, sustaining transport V_max. Interestingly, the homologous lysine residue in human LAT1 (Lys204) is key for the faster exchange (antiport) mode of transport but not for the slower facilitated diffusion (uniport) mode of transport. Mutation to glutamine (Lys204Gln) fully abolishes exchange transport activity but retains the facilitated diffusion mode of transport (57). Thus, in agreement with the role of Lys154 on the gating of TM1a in BasC, Lys204 is the key for the substrate-inducing conformational transitions in human LAT1. BasC, in contrast to human LAT1, does not present detectable facilitated diffusion (45) and mutation Lys154Ala drops amino acid exchange (antiport).

Consistent with this notion, Lys154Ala mutation freezes TM1a and the substrate is not able to stabilize the higher smFRET E* population, emphasizing the indispensability of Lys154 in mediating TM1a gating. The structural significance of Lys154 in BasC is mirrored in analogous roles within related transporters with an APC fold. The side chain of the LAT-conserved Lys residue in TM5 forms hydrogen bonds with TM1a residues, i.e. the carbonyl group of Gly15 and/or Ile18 in BasC, in all solved LAT structures (31–41). Moreover, in transporter structures with closed cytoplasmic gates, e.g. human LAT1 and the related bacterial cationic amino acid transporter GkApcT, the conserved TM5 Lys residue bridges TM1a and TM8 (41, 59), thereby pointing to a pivotal TM1a–TM5–TM8 interaction for cytoplasmic gate occlusion. Similarly, in NSS transporters, coordination bonds of Na2 link residues in TM1a, TM5, and TM8, stabilizing the outward-facing conformation (28, 49). This parallelism underscores the significance of Lys154 in orchestrating the inner gating of BasC. However, while this might explain the decreased L-Ala transport V_max, it is counterintuitive that defective closing of the inner gate would increase the extracellular K_m of the Lys154Ala mutant (31).

Interestingly, our MD simulations of the substrate-bound, inward-facing BasC conformation within a bacterial membrane model, although limited in capturing the tilting of TM1a (Fig. S8), provided valuable insights. During these simulations, removal of the hydrogen-bonding capacity of the Lys154 side chain (Lys154Ala mutant) or the protonation of Lys154 (Lys154(0)) led to changes in the conformation of the unwound segment of TM1, which is a key component of the substrate-binding site in BasC and mammalian LATs in inward- and outward-facing conformations (31, 32, 34, 36, 41, 42). Consequently, the Lys154Ala mutation and the uncharged Lys154 (Lys154(0)) weakens interaction with the substrate and leads to the release of the substrate to the cytoplasm. Further research will be required to ascertain whether the Lys154Ala mutation underlies the defective closing of the external gate, causing the observed increase in external K_m.

Collectively, our studies provide compelling evidence that the side chain of Lys154 is crucial for maintaining the stability of TM1, enabling its inherent dynamic gating between open and closed gate states, and facilitating that substrate-binding biases the conformational equilibrium toward the closed gate structure (Fig. 6). Akin to the role of Na2 in NSS transporters (28, 49), the closed conformation sampling is facilitated by the connection of the conserved Lys154 with TM1a, where capping of the negative pole of TM1a by Lys154 is emerging as the putative major interactor. When this interaction is disrupted, as seen in the Lys154Ala mutation, it impedes the natural tilting of TM1a and the closing induced by the substrate at the cytoplasmic gate. Whether Lys154 is also the key for the closing of the extracellular gating of BasC, which might be at the basis of the increased external K_m for L-Ala by mutation Lys154Ala remains to be studied.

Materials and methods

Mutagenesis

Point mutations in BasC were performed through QuickChange site-directed mutagenesis and verified by Sanger sequencing. TM1a–TM4 double-cysteine variants were generated on a Cys-less variant: Ile7Cys–Thr120Cys–Cys427Ala.

BasC expression and purification

A more detailed protocol of BasC purification was published in Ref. (45). Briefly, membranes of Escherichia coli–induced ON at 37 °C with BasC-3C-GFP-His plasmid were thawed and diluted with TB (20 mM of Tris, 150 mM of NaCl) until 3 mg/mL and incubated for 1 h at 4 °C with detergent (1% DDM for PL preparation and smFRET assays or 2% DM for structural analysis—X-ray/cryo-EM). Nonsolubilized membranes were discarded by ultracentrifugation (200,000 × g, 1 h, 4 °C). Solubilized BasC-GFP was bound to Ni-NTA Agarose (Qiagen) previously equilibrated with DM or DDM buffer (TB with 0.06% DDM or 0.17% DM, respectively). Resin was first washed 3 × 10 CV with DM or DDM buffer with increasing concentrations of imidazole (10, 20, and 30 mM, respectively). Next, protein was eluted with DM or DDM buffer with 300 mM of imidazole for PL preparation. On the other hand, 3C protease elution was performed on column for structural analysis and smFRET assays. Here, resin was equilibrated with 10 CV of 3C buffer (DDM buffer supplemented with 1 mM of DTT and 1 mM of EDTA). His-tagged HRV-3C protease was added to the resin and incubated for 3 h (for FRET experiments) or overnight (for structural studies) at 4 °C.

Nb generation, expression, and purification

Nbs were generated, as described previously (31, 60). One llama was immunized through 6 weekly injections of purified BasC reconstituted in E. coli polar lipid PLs. A panning experiment on BasC resulted in 29 specific Nb families. Nb53, Nb71, and Nb74 were selected based on their transport-blocking activity as explained before. All used Nbs were expressed and purified in pMESy4 as inducible periplasmic proteins in E. coli WK6 strain. They produced milligram amounts and were purified to >95% using immobilized Ni-NTA chromatography (Qiagen) from a periplasmic extract of a 1 L culture. Purified Nbs (5–10 mg/mL) in 20 mM of Tris–Base, 150 mM of NaCl, pH 7.4, were frozen in liquid nitrogen and stored at −80 °C before use. For structural purposes, we purified BasC complexed with Nbs as in Errasti-Murugarren et al. (31). Ni-NTA pure BasC and desired Nbs were incubated in a molar ratio of 1:1.4 (BasC:Nb) for 1 h on ice. The complex was separated from Nb excess by size exclusion chromatography (SEC) on a Superdex-200 10/300 column (GE Healthcare). When needed, we used protein concentrators with a cutoff of 100 kDa (Amicon) to avoid detergent concentration.

Labeling

On-column labeling started with the same purification protocol as described previously but with the addition of 10 mM of DTT from the lysis to labeling step. Before 3C elution, a 10 CV wash with DDM buffer (without DTT) was performed. Immediately after, maleimide fluorophores (Alexa Fluor 546 and Alexa Fluor 647 or the pair sulfo-Cy3 and sulfo-Cy5) were added in a BasC:dye molar ratio of 1:5. Incubation was carried out at 4 °C overnight in rotation and under light protection. To achieve balanced labeling, various donor:accepted ratios were tested. The resin was washed three times using 10 CV with DDM buffer to remove free dyes, and 3C protease elution was done as described before. Labeled protein was finally separated from remaining free dye by SEC in a Superdex 200 10/300 GL column using DDM buffer and measuring absorbance at 554, 647, 561, and 650 nm for sulfo-Cy3, sulfo-Cy5, Alexa 546, and Alexa 647, respectively.

ALEX-smFRET microscopy

A full description of the in-house confocal microscope and data analysis procedures for smFRET was provided previously and was adapted closely to Ref. (52). Briefly, all samples were examined by focusing the excitation/observation volume of a confocal microscope into an ∼100-μL sample drop containing ∼50 pM of BasC in DDM buffer on a glass coverslip. Before data recording, the coverslips were coated for >60 s with 1 mg/mL of bovine serum albumin to prevent fluorophore and/or protein adsorption. Each experimental condition (apo, ligand, etc.) was then studied for between 30 and 120 min, depending on the fraction of donor- and acceptor-containing BasC molecules in relation to donor- and acceptor-only ones. We also performed DNA controls with the same fluorophore pairs as for the proteins in DDM buffer to verify proper microscope alignment and exclude buffer interferences (Fig. S3).

For alternating laser excitation (53, 54), we used two laser sources for sample excitation with electronic modulation: a 532-nm diode laser at 60 μW (OBIS 532-100-LS, Coherent, USA) and a diode laser at 640 nm (OBIS 640-100-LX, Coherent, USA) at 25 or 30 μW for Alexa 546-Alexa 647 and sCy3–sCy5 dye-pairs, respectively. The laser light was guided into an epi-illuminated confocal microscope body (Olympus IX71, Hamburg, Germany) by a dual-edge beamsplitter (ZT532/640rpc, Chroma/AHF, Germany) and focused to a diffraction-limited excitation spot by a water immersion objective (UPlanSApo 60x/1.2w, Olympus, Hamburg, Germany). The emitted fluorescence was collected through the same objective, spatially filtered using a pinhole with a diameter of 50 μm, and spectrally split into donor and acceptor channels by a single-edge dichroic mirror H643 LPXR (AHF). Fluorescence emission was filtered (donor: BrightLine HC 582/75 [Semrock/AHF], acceptor: Longpass 647 LP Edge Basic [Semroch/AHF]) and focused onto avalanche photodiodes (SPCM-AQRH-64, Excelitas). The detector outputs were recorded by an NI-Card (PCI-6602, National Instruments, USA).

Data analysis and plotting were performed using a software package written in-house as described in Refs. (17, 21). Single-molecule events were identified using an all-photon-burst-search (61), with a threshold of 15, a time window of 500 μs, and a minimum total photon number of 50, with an additional per-bin with a minimum total photon number between 100 and 150 for inclusion of the burst in the histogram. Comparisons between absolute results and mean FRET efficiencies E*, which are setup- and alignment-dependent, were made only when data were collected on the same day.

Influx/efflux assays in BasC PLs

For transport measurements, PLs were prepared, as previously described (45). Imidazole-eluted BasC-GFP protein was reconstituted in E. coli polar lipid extract (Sigma-Aldrich). Lipids were dried under nitrogen flow and resuspended in Tris buffer by bath sonication. Purified BasC protein was added to reach the desired protein-to-lipid ratio of 1:100 (w:w). To destabilize the liposomes, 1.25% β-D-octylglucoside was added and the mixture was incubated on ice for 30 min. PLs were formed by dialysis for 24–48 h at 4 °C against 100 volumes of Tris buffer. PL suspensions were aliquoted, frozen in liquid nitrogen, and stored at −80 °C until use.

L-Alanine (A7627, Sigma-Aldrich), L-serine (S4500, Sigma-Aldrich), or L-glutamine (G8540, Sigma-Aldrich), along with 0.05 µCi/µL of L-[3H]-serine (Perkin Elmer) only for efflux experiments, was added to the PL suspension at the desired concentration. The mixture underwent six freeze/thaw cycles in liquid nitrogen to load the PLs. The excess liposomal solution was removed through ultracentrifugation (200,000 × g for 1 h at 4 °C), and the PLs were subsequently resuspended to one-third of the initial volume using Tris buffer. When amino acid concentrations exceeded 10 mM, either inside or outside the PLs, an equal concentration of D-mannitol (M4125, Sigma-Aldrich) was used on the opposite side of the PLs to maintain osmotic balance.

Influx and efflux assays were started after mixing 20 μL of PLs solution with 180 μL of influx transport buffer (10 μM of L-serine in Tris buffer with 0.5 µCi of L-[3H]-serine [Perkin Elmer] or 4 mM of cold amino acid) for efflux experiments. Transport was allowed to take place at room temperature and stopped in 5 s, adding 3 mL of ice-cold Tris buffer and by filtrating the PLs through 0.45 μm pore-size membrane filters (Sartorius Stedim Biotech). Filters were washed twice with 3 mL of Tris buffer and allowed to dry. Trapped radioactivity was then counted. Transport measured in PLs containing no amino acid was subtracted from each data point to calculate the net exchange. Transport measurements were normalized to the BasC protein concentration by sodium dodecyl sulfate–polyacrylamide gel electrophoresis.

Cryo-EM

The specimen was vitrified using 3 μL of freshly purified BasC-Nb71 or BasC-Nb53 complex in 0.17% DM applied to glow-discharged Quantifoil R 0.6/1 Cu 300 mesh grids (Electron Microscopy Sciences) or C-flat 1.2/1.3 Cu 300 mesh, at a concentration of 2.2 and 2.9 mg/mL, respectively. Grids were blotted for 2–3 s under 95% humidity and plunge frozen in liquid ethane using a Vitrobot Mark IV (Thermo Fisher Scientific). BasC-Nb53 complex was incubated with 100 mM of L-Ala 30 min before vitrification.

Cryo-EM datasets for BasC-Nb71 were collected on a Titan Krios electron microscope operating at 300 kV and equipped with Thermo Fisher Falcon 4 direct electron detector at the facilities of the University of Leeds. A total of 6,756 movies were collected in .eer format. The setup used a 130k magnification, which corresponded to 0.91 Å/pix. The defocus applied covered from −3.3 to −1.8 μm, and the total electron dose was 34.83 e−/Ų.

Cryo-EM datasets for BasC-Nb53 were collected firstly on a Glacios 200 kV transmission electron microscope with extreme field emission gun optics, equipped with Thermo Fisher Falcon 4 direct electron detector at the facilities of the ALBA synchrotron to select good grids. For higher resolution data, we collected 38,963 movies in tiff format in CM01 Titan Krios cryo-EM equipped with a K3 direct electron detector, a Quantum LS imaging filter, and a Volta-phase plate in ESRF (62). The setup used a 165k magnification, which corresponded to 0.51 Å/pix. The defocus applied covered from −1.0 to −2.4 μm, and the total electron dose per frame was 1.04 e⁻/Ų.

Cryo-EM image processing

For the BasC-Nb71 dataset (Fig. S5), movies were imported to RELION4 (63) and aligned using MotionCor2 (64) with 35 patches per image and 31 frames per group, applying dose weighting. Contrast transfer function (CTF) parameters were determined using the CTFFIND-4.1 approach (65). Particles were selected using Topaz, after training the neural network using an ∼1,000 manually selected particles in a subset of 100 randomly selected micrographs (66). Subsequent image processing was performed using RELION4, and cryoSPARC (67). Initially, 590,169 particles were extracted and subjected to several rounds of reference-free 2D classification, obtaining 409,984 particles. Selected particles were used to generate 3D reconstructions, leading to a consensus volume of >160,000 particles where all TMs were already visible. To recover good particles, iterative rounds of heterogeneous refinement were performed using CryoSPARC. Thus, an ab initio reconstruction was used to clean the particles in 3D, selecting a final subset of 169,672 particles. By nonuniform refinement (68), the density from the micelle was removed, and a map of ∼6.0 Å resolution was generated.

For the BasC-Nb53 dataset (69) (Fig. S6), CryoSPARC (68) was used to import movies and apply Patch Motion Correction and Patch CTF. A low-resolution model obtained previously from GLACIOS data from similar grids was used to use Template Picker and obtain 2,000,024 particles, which were extracted and subjected to several rounds of reference-free 2D classification, resulting in 1,343,403 particles. The selected particles were used to generate an ab initio model where all TM regions were already visible. Subsequent 3D classification yielded a volume class of good quality containing 381,139 particles, which were further refined with nonuniform refinement, local refinement, global CTF refinement, and reference-based motion correction, resulting in a map with an estimated global resolution of 2.9 Å.

Volumes were visualized in UCSF Chimera (70), which was also used to fit into the density the available BasC pdb (PDB ID 6F2G) and calculated models for Nb71 and Nb53 in Swiss-Model (71).

Model building for BasC-NB53 from these models was carried out in Coot (72), and PHENIX real space refinement (73) was used to model refinement and validation. See Table S1 for validation data.

Obtained volumes have been deposited in the Electron Microscopy Data Bank (EMDB) with EMDB accession codes EMD-50021 (74) and EMD-51912 (75) for BasC-Nb71 and BasC-Nb53, respectively. BasC complexed with NB53 and L-Ala model is also deposited in Protein Data Bank with PDB ID 9H76.

MD simulations and postprocessing analysis

All the simulated systems presented here were prepared starting from the BasC crystal structure (PDB ID: 6F2W) (31) in the open-inward conformation. By either mutating Lys154 or removing the substrate (2AIB), we obtained six distinct BasC systems: (i) and (ii) WT in apo and holo conditions; (iii) and (iv) Lys154Ala mutated in apo and holo conditions, and (v) and (vi) deprotonated neutral Lys154 mutated (Lys154(0)) in apo and holo conditions. All these systems were inserted in 150 × 150 Ų lipid membrane perpendicular to z-axis whose concentration mimics the E. coli polar lipid extract (Sigma-Aldrich) used in the PL transport assay (70% palmitoyloleoyl phosphatidylglycerol [POPG], 10% tetramyristoyl cardiolipin [TMCL], and 20% palmitoyloleoyl phosphatidylethanolamine [POPE]). Above and below the lipid bilayer, the simulation boxes were filled with TIP3P water molecules and a few Na⁺ and K^– ions to neutralize the systems and bring the total ion concentration to 150 mM. The resulting heights along the z-axis were 100 Å. We used the CHARMM-GUI webserver to setup the starting configurations (76). The simulations were performed using GROMACS 2020.2 code (77) and CHARMM 36 forcefield (78). LINCS (79) algorithm was used to constrain all the bonds involving hydrogen atoms, thereby allowing us to employ a 2-fs step to integrate the Newton equations. The Mesh Ewald method was used to account for long-range interactions with a real-space cutoff of 12 Å. The first minimization employing a steepest descendent algorithm (80) was followed by an equilibration step consisting of an NPT ensemble 100 ns run with a semi-isotropic Berendsen barostat (81) to maintain the pressure at 1 atm with a coupling constant of 0.6 ps. The temperature, set at 310 K, was controlled using the velocity rescaling algorithm with a 0.4-ps coupling constant (82). The production step was a 500-ns-long NPT ensemble simulation using a semi-isotropic Parrinello–Rahman barostat (83) with a 0.6-ps coupling constant and a Nose–Hoover thermostat (84) with a 0.4-ps coupling constant to fix pressure and temperature at 1 atm and 310 K, respectively. We performed three separate replicas for each system. All the analyses presented in this work were carried out using either VMD (85) software or in-house scripts.

Acknowledgments

The authors thank INSTRUCT-ERIC for providing Nbs (PID1176) and access to the 300 kV Titan Crios of Leeds, UK (PID17338). They thank Pablo Guerra from the IBMB-CSIC Cryo-EM Platform for assistance during the sample preparation and microscope data acquisition. The authors acknowledge funding from Project IU16-014045 (Cryo-TEM) from Generalitat de Catalunya and by “ERDF A way of making Europe,” by the European Union. They thank Nick Berrow (Protein Expression Core Facility; IRB Barcelona) for providing us with HRV-3C protease. They also thank Jasminka Boskovic and Johanne LeCoq from the Electron Microscopy Unit at CNIO for preparing and visualizing the grids and sending samples to Leeds. The authors acknowledge the European Synchrotron Radiation Facility for the provision of beam time on CM01 and thank Eaazhisai Kandiah for assistance. They gratefully acknowledge institutional funding from the Spanish State Research Agency of the Spanish Ministry of Science and Innovation—Programa Estatal de Fomento de la Investigación Científica y Técnica de Excelencia—Centres of Excellence “Severo Ochoa” CEX2019-000891-S and CEX2019-000913-S. O.L. laboratory also had the support from the National Institute of Health Carlos III to CNIO. IRB Barcelona is a member of the Centres de Recerca de Catalunya (CERCA) System of the Generalitat de Catalunya. J.F. is supported by a Centro de Investigación Biomédica en Red de Enfermedades Raras (CIBERER) contract, A.N.-A. was supported by FI-AGAUR Fellowship, and M.M.M. is supported by La Caixa (LCF/PR/HR20/52400017 to O.L. and M.P.). N.Z. acknowledges a postdoctoral fellowship from the Alexander von Humboldt Foundation. Research in the laboratory of T.C. was supported by the European Commission (ERC-STG 638536—SM-IMPORT), Deutsche Forschungsgemeinschaft (GRK2062, project C03 to T.C.; SFB863, project A13 to T.C. and the Center for Nanosicence). L.M. and M.O. acknowledge support from the following: Spanish Ministerio de Ciencia e Innovación (PID2020-116620GB-I00, RTI2018-096704-B-100, PID2021-122478NB-I00); Center of Excellence for HPC H2020 European Commission; BioExcel-3; Centre of Excellence for Computational Biomolecular Research (823830); Catalan SGR and the Instituto de Salud Carlos III Instituto Nacional de Bioinformatica (ISCIII PT 17/0009/0007 co-funded by the Fondo Europeo de Desarrollo Regional); European Regional Development Fund under the framework of the ERFD Operative Programme for Catalunya, the Generalitat de Catalunya AGAUR (SGR2017-134); European Union MDDB: Molecular Dynamics Data Bank; and the European Repository for Biosimulation Data (101094651). The project used computer resources from the Barcelona Supercomputing Centre. The authors thank Gabriel G. Moya-Munoz, Atieh Aminian Jazi, and Felix Pioch for their support for this study. Finally, they are grateful to Pol Torrents, Laia Bekius, and Javier Fernández for their help during their student internships. P.G-N. is a PhD student in the Biomedicine Program at the Universitat de Barcelona (UB).

Preprints

A previous version of this manuscript was posted on a preprint at: https://doi.org/10.1101/2024.03.26.586791

Supplementary Material

Supplementary material is available at PNAS Nexus online.

Funding

This work was funded by the Spanish Ministry of Science and Innovation (PID2021-122802OB-I00) and the Generalitat de Catalunya (grant 2021 SGR 01281).

Data Availability

All data are included in the manuscript and Supplementary material. The raw smFRET data supporting the findings of this study are available in Mendeley repositories with DOI: 10.17632/srb7pskmy3.1 (86). Scripts for generating MD histograms and Github: https://github.com/LucaMaggi19/Histogram-. Raw data for cryo-EM datasets can be found in ESRF repository (69), and cryo-EM volumes have been deposited in EMDB and PDB (74, 75) and are fully referenced in the manuscript.

References

Author notes