Evolutionary analysis reveals the origin of sodium coupling in glutamate transporters

Reddy, Krishna D.; Rasool, Burha; Akher, Farideh Badichi; Kutlešić, Nemanja; Pant, Swati; Boudker, Olga

doi:10.1038/s41594-025-01652-z

Download PDF

Article
Open access
Published: 25 August 2025

Evolutionary analysis reveals the origin of sodium coupling in glutamate transporters

Nature Structural & Molecular Biology volume 32, pages 2564–2574 (2025)Cite this article

7988 Accesses
2 Citations
12 Altmetric
Metrics details

Subjects

Abstract

Secondary active membrane transporters harness the energy of ion gradients to concentrate their substrates. Homologous transporters evolved to couple transport to different ions in response to changing environments and needs. The bases of such diversification and, thus, principles of ion coupling are unexplored. Here, using phylogenetics and ancestral protein reconstruction, we investigated sodium-coupled transport in prokaryotic glutamate transporters, a mechanism ubiquitous across life domains and critical to neurotransmitter recycling in humans by excitatory amino acid transporters from the solute carrier 1 family. By inferring ancestral prokaryotic transporter sequences during a change in the ion-coupling mechanism, we found an evolutionary transition from sodium-dependent to independent substrate binding and transport. Structural and functional experiments on ancestral transporters suggest that the transition involved allosteric mutations, rendering sodium binding dispensable without affecting the ion-binding sites. Allosteric tuning of transporters’ energy landscapes might be a widespread route of their functional diversification.

Symport and antiport mechanisms of human glutamate transporters

Article Open access 04 May 2023

Glutamate transporters have a chloride channel with two hydrophobic gates

Article 17 February 2021

Structural basis of sodium ion-dependent carnitine transport by OCTN2

Article Open access 29 November 2025

Main

Transport of solutes across cell membranes against their concentration gradients is critical for cellular homeostasis. Secondary active membrane transporters catalyze this energetically unfavorable process using the energy of ions flowing down their electrochemical gradients. The ion-binding sites of many transporters are well characterized. However, the three-dimensional (3D) structural properties of transporters necessary to couple substrate and ion movements, and thereby how transporters harvest electrochemical energy, remain elusive.

In the dicarboxylate/amino acid:cation symporter family, the Na⁺-coupling mechanism persisted through millions of years, conserved between archaea and humans; mammalian excitatory amino acid transporters (EAATs; belonging to the solute carrier 1 (SLC1) family) use Na⁺ gradients to clear the neurotransmitter glutamate from the synaptic cleft. Many transporter families, including glutamate transporters, show functional diversification using Na⁺ or H⁺ gradients (Fig. 1a)^1,2. The cell’s environment and needs provide the selective pressure to use specific ions such that when cell membranes are H⁺-leaky or Na⁺ is abundant, Na⁺-coupled transport predominates. In contrast, in environments where Na⁺ is scarce, such as in fresh water, certain soils and fermented foods, Na⁺ coupling is no longer beneficial and is eliminated. We used phylogenetic analysis, ancestral protein reconstruction (APR), and structure-function characterization of inferred ancestral proteins to understand how prokaryotic glutamate transporters transitioned away from Na⁺ coupling. Contrary to our expectations, the transition did not involve promiscuous ion coupling^3,4 or a gradual loss of Na⁺-binding sites. Instead, allosteric changes first made Na⁺ binding dispensable to function in a distinct evolutionary step, and only then did the transporters lose their Na⁺-binding residues.

**Fig. 1: Mapping the evolutionary transition from sodium to proton coupling.**

Energetically linked substrate and Na⁺ binding underlies symport in Na⁺-coupled glutamate transporter homologs, including archaeal Glt_Ph and Glt_Tk. A higher extracellular Na⁺ concentration ensures a greater probability of substrate binding from the outside and release on the inside of the membrane, providing directionality to the transport cycle and leading to concentrative uptake^{5,6,7,8,9,10,11,12}. The binding sites for the substrate and three symported Na⁺ ions are located within a dynamic transport domain, which moves between extracellular and intracellular positions (outward-facing and inward-facing states, Extended Data Fig. 1) in an elevator-like motion relative to a stationary scaffold and trimerization domain^{13,14,15,16,17,18,19,20,21,22}. The transport mechanism and homotrimeric architecture are conserved across glutamate transporters^{23,24,25,26,27}. Na⁺ binding to two sites (Na1 and Na3) induces the formation of a high-affinity substrate-binding pocket, resulting in substrate and the third Na⁺ (Na2) binding and closure of the helical hairpin 2 (HP2) gate, which precedes translocation^{8,9,28,29,30,31,32} (Fig. 1b,c). Thus, ion-coupled binding and transport occur because the high-affinity substrate-binding state is a high-energy state only accessible as Na⁺ binds. Accordingly, mutations impacting all three Na⁺-binding sites diminish Na⁺-coupled substrate binding and transport^{3,10,33,34,35,36}.

Using APR, we reconstructed and characterized ancestral transporter sequences (Fig. 1d); we found that an intermediate ancestral transporter (Anc^Int) between Na⁺-coupled and H⁺-coupled ancestral transporters showed Na⁺-independent substrate binding and transport, despite preserved Na⁺-binding sites. Cryo-electron microscopy (cryo-EM) structures showed that Anc^Int accessed the high-affinity substrate-binding state spontaneously, suggesting that its energy is lower than in Na⁺-coupled transporters and explaining why Na⁺ binding was no longer necessary. Detailed evolutionary analysis revealed that the shifted energetic balance from the low-affinity to the high-affinity states was because of allosteric changes distant from the ion-binding and substrate-binding sites, and mutations impacting just two residues in Anc^Int restored Na⁺ coupling. We propose that allosterically changing the relative energy of functional states in transporters constitutes an evolutionary mechanism allowing the diversification of strictly specific ion coupling observed in present-day transporters. Our work reveals key principles underlying Na⁺ coupling, showcases the feasibility and power of structure–function APR in untangling the complex mechanisms of membrane transporters, and suggests a general evolutionary mechanism underlying adaptations to different ionic environments in secondary active transporters.

Results

An evolutionary intermediate between Na⁺-coupled and H⁺-coupled transporters

We used phylogenetic analysis and APR to understand how bacterial glutamate transporters might have changed their ion-coupling preference (Extended Data Fig. 2). The approach can reveal minimal subsets of stepwise, chronologically ordered changes required to alter functions, including epistatic changes, which are insufficient for the functional switch but are prerequisites for it³⁷. Subsequent inference of ancestral sequences along an evolutionary pathway provides a platform to test evolutionary and mechanistic hypotheses^{38,39,40,41,42,43,44}. We selected sequences in the Firmicutes/Bacillota phylum to recapitulate the transition from Na⁺ to H⁺ coupling, rooting the phylogenetic tree to archaeal Na⁺-coupled Glt_Ph (Methods and Extended Data Fig. 3a–d). We identified three major monophyletic clades in our tree (Fig. 1d), with two predictably displaying conserved signature sequences of Na⁺ and H⁺ coupling (Fig. 1e,f and Extended Data Fig. 3d). We refer to these present-day extant sequences as Ext^Na and Ext^H clades. Strikingly, we also observed an intermediate clade Ext^Int between Ext^Na and Ext^H clades. Only 16% of its sequences contain all Na⁺-coordinating residues, and none contain all H⁺-coupling signature residues (Fig. 1e,f and Extended Data Fig. 3d). This branch was robust to perturbation and reconstructed with high confidence on the basis of a UFBoot2 bootstrap value > 95% (Fig. 1d)⁴⁵. The extant sequences closest to Ext^Int with the full complement of Na⁺-coordinating residues are referred to as Ext*.

Next, we inferred ancestral sequences by determining posterior probabilities (PPs) of each amino acid occurring at each position (site) in the multiple-sequence alignment (Extended Data Fig. 4a–f). Each node’s final maximum likelihood (ML) amino acid sequence is constructed using the most probable amino acid at every site. We reconstructed sequences corresponding to the closest ancestor to the Na⁺-coupled root (Anc^Na) and the ancestor of the intermediate Ext^Int clade (Anc^Int); the average PPs per site were 0.87 and 0.85, respectively (Extended Data Fig. 4a,b), which is sufficiently high for APR studies^46,47. As expected, reconstructed Anc^Na sequences showed signature sequences of Na⁺-coupled transporters (Fig. 1c and Extended Data Figs. 4a and 5). Surprisingly, Anc^Int also featured the full complement of Na⁺-binding residues with high PPs, contrasting with their poor conservation among the descendent present-day transporters (Figs. 1c and 2a and Extended Data Figs. 4a and 5). These observations suggest a lack of selective pressure on Na⁺-binding residues in Anc^Int, raising the question of its functional properties.

**Fig. 2: Substrate transport by Anc^Int is independent of Na⁺ and H⁺.**

To design ancestral constructs for protein expression, we replaced poorly aligned highly variable loops and tails, excluded from the APR analysis, with those from an Ext* sequence with the highest identity to Anc^Int (Extended Data Fig. 5). To minimize the impact of these regions on our comparisons between ancestors, we used the same loops and tails for all ancestral constructs. Most of these were single-residue insertions, except for the C-terminus and the loop between transmembrane helix 3 (TM3) and TM4. Although these insertions might affect protein stability, flexibility and function, previous studies in Glt_Ph suggest that these regions are not directly involved in ion coupling. The C-terminus has been deleted in structural and functional studies¹³, and cleavage of the 3–4 loop affected transport kinetics but not ion coupling⁴⁸. We expressed ancestors in Escherichia coli with or without a C-terminal optimized GFP⁴⁹ and purified them (Fig. 2b and Supplementary Fig. 1a–c). While there are several bands above and below the primary band, they correlate with GFP fluorescence in gels and size-exclusion profiles. Therefore, these are likely misfolded or aggregated species. Despite extensive optimization, the ancestral proteins expressed ~10-fold lower than extant prokaryotic homologs, yielding ~20–100 µg of purified proteins per L of culture.

The ancestral intermediate transporter is uncoupled from Na⁺

When we purified and reconstituted Anc^Na into proteoliposomes, we observed a time-dependent accumulation of [³H]l-aspartate in vesicles in the presence of an inward Na⁺-gradient. In contrast, no [³H]l-aspartate was retained in the absence of Na⁺ (Fig. 2c). These results suggest that Anc^Na is adequately folded and recapitulates expected Na⁺ coupling. When we repeated the experiment with Anc^Int, we observed lower [³H]l-aspartate retention unaffected by the removal of Na⁺ but reduced in the presence of the competitive inhibitor dl-TBOA (Fig. 2c), suggesting that Anc^Int in proteoliposomes might be able to bind and transport l-aspartate independently of Na⁺. To demonstrate this, we measured [³H]l-aspartate accumulation in Anc^Int proteoliposomes under exchange conditions^1,50, with 500 µM unlabeled l-aspartate inside the vesicles and 0.5 µM [³H]l-aspartate outside. Anc^Int accumulated substrate independently of Na⁺ or pH, unlike the respective controls, Anc^Na and GltT_Bc (Fig. 2d–g). Together, this suggests that Anc^Int functions as a cation-independent exchanger.

We used microscale thermophoresis (MST) to measure substrate-binding affinity to Anc^Int. Anc^Int binds l-aspartate with similar micromolar affinities in the absence and presence of 100 mM Na⁺ (Fig. 3a). Notably, the MST signal-to-noise values upon l-aspartate binding ranged from 8.6 to 19.5, reflecting the high confidence of the binding isotherms (Supplementary Fig. 2a). We confirmed this finding using nano-differential scanning fluorimetry (nanoDSF) to measure ligand-dependent protein melting temperatures (T_m) as a qualitative proxy for ligand binding^51,52. Adding l-aspartate to Anc^Int in the absence of Na⁺ increased T_m by 7.1 °C, suggesting that, as expected, l-aspartate can bind without Na⁺ (Fig. 3b,c and Supplementary Fig. 2c). Notably, 10 mM Na⁺ also stabilized the protein by 5.3 °C without l-aspartate and by 13.5 °C in the presence of 1 mM l-aspartate, suggesting that the ions can still bind to the transporter regardless of substrate binding (Fig. 3c,d and Supplementary Fig. 2c). In Anc^Na, the MST signal-to-noise ratio was insufficient to accurately determine binding affinities, likely because of changes in measurable parameters, including charge, hydration shell or conformation⁵³. In nanoDSF l-aspartate titrations, much lower concentrations of l-aspartate were required to observe substrate binding in 10 mM Na⁺ as compared to 0 mM Na⁺ (Fig. 3e). We observed insignificant or small T_m increases when adding 1 mM l-aspartate or 10 mM Na⁺ separately compared to the ligand-free Anc^Na (Fig. 3f,h and Supplementary Fig. 2b). In contrast, we observed a dramatic 11.3 °C increase when adding 1 mM l-aspartate in the presence of 10 mM Na⁺(Fig. 3f,g). Thus, l-aspartate and Na⁺ bind to Anc^Na cooperatively, as reported for Na⁺-coupled transporters^{1,8,10,11,12,54,55}.

**Fig. 3: Anc^Int binds substrate independently of Na⁺.**

These results suggest that, unlike in Na⁺-coupled transporters, l-aspartate and Na⁺ can bind to Anc^Int independently; therefore, substrate and ion binding are ‘uncoupled’. Notably, while we did not observe any Na⁺ stimulation in either MST or exchange experiments, the ΔT_m of l-aspartate was larger in the presence of Na⁺ (15.3 °C versus 7.1 °C), suggesting residual cooperativity. However, thermal melts are not thermodynamic measurements and ΔT_m values cannot be directly related to ligand affinity. Overall, we conclude that the intermediate Ext^Int clade functionally diverged from the Na⁺-coupled Ext^Na transporters by uncoupling substrate from Na⁺ binding. We attempted to verify that the reconstructed Anc^Int was robust to phylogenetic uncertainty using ‘AltALL’ sequences where all ambiguous residues are varied⁴⁶ but these produced misfolded proteins. Instead, we validated the observed functional divergence by examining the necessary and sufficient sequence changes, as discussed below.

Anc^Int accesses high-affinity states without Na⁺

To visualize how Na⁺-independent aspartate binding occurs, we purified Anc^Int, reconstituted it into MSP1E3 lipid nanodiscs (Supplementary Fig. 3), exchanged it into a buffer containing no Na⁺ or any other alkali cations (20 mM HEPES pH 7.4 and 100 mM NMDG-Cl), and froze cryo-EM grids after and before adding 1 mM l-aspartate (that is, in bound and ligand-free conditions, respectively) (Supplementary Figs. 4 and 5). As a Na⁺-coupled control, we reconstituted Glt_Ph into MSP1E3 lipid nanodiscs and froze grids under ligand-free conditions (Supplementary Fig. 6).

First, we compared the structures of Glt_Ph and Anc^Int in ligand-free conditions. Cryo-EM imaging of Glt_Ph revealed the presence of two structural classes with final maps at 2.7-Å and 3.1-Å resolution (Extended Data Figs. 6 and 7, Supplementary Fig. 6 and Table 1). A highly conserved nonhelical loop N310-M311-D312 (NMD) motif, which breaks TM7 into TM7a and TM7b, is the nexus for solute binding and participates in coordinating Na1, Na2, Na3 and substrate through its side chain and main chain atoms. In both classes, the NMD motif is in a ‘flipped-out’ configuration, which distorts substrate-binding and Na⁺-binding sites (Fig. 4a,e). We call these conformations, characteristic of ligand-free Na⁺-coupled transporters, ‘low-affinity’ states^{9,15,16,30,56}. The two ligand-free Glt_Ph structural classes differ in the position of the transport domain. One corresponds to the outward-facing state (Fig. 4a,e) and the other corresponds to a previously observed intermediate^16,21,30,57 where the transport domain moves toward the inward-facing state by about one quarter of the way (Extended Data Figs. 7 and 8a). The HP2 gate is open in the outward-facing states, propped open by M311, and closed in in the intermediate, with M311 pointing out into the lipid from underneath (Fig. 4a and Extended Data Fig. 8a).

Table 1 Cryo-EM data collection, refinement and validation statistics

Full size table

**Fig. 4: Substrate-binding mechanism in Na⁺-coupled and uncoupled transporters.**

Previously determined Glt_Ph and Glt_Tk structures show that, upon Na⁺ binding to Na1 and Na3, Na⁺-coupled transporters restructure from the low-affinity ligand-free state to a ‘high-affinity’ state through allosterically coupled changes^8,9,28,10 (Fig. 4b,f). Their NMD motifs go from the flipped-out to the flipped-in configuration, repositioning M311 proximal to the substrate (Fig. 4b) and altering the tilt of TM7b³⁰. R397 moves from occupying the substrate-binding site to forming a cation–π interaction with Y317 in TM7b (Fig. 4b), with its guanidium group in place to salt-bridge with l-aspartate. Moreover, N401 repositions slightly to coordinate l-aspartate optimally³⁰. Together, these changes underlie high-affinity binding by optimally positioning residues coordinating the substrate and the additional Na2 (Fig. 4c,d,g). Following substrate binding, HP2 closes (Fig. 4g), facilitating translocation of the transport domain to the inward-facing state.

Cryo-EM imaging of ligand-free Anc^Int revealed three outward-facing structural classes, all refined to 3.0-Å resolution (Extended Data Fig. 7, Supplementary Fig. 5 and Table 1). One class resembles the low-affinity ligand-free Glt_Ph with a flipped-out NMD motif and an empty binding pocket; however, R397₃₆₉ is already in place to coordinate l-aspartate (hereafter, regular text and subscript correspond to Glt_Ph and Anc^Int numbering, respectively) (Fig. 4h). The Na⁺-binding sites are distorted similarly to ligand-free Na⁺-coupled transporters (Fig. 4l), except that T92₇₈ and [T/S]93₇₉ in the Na3 site are already correctly positioned.

However, unlike ligand-free Glt_Ph, we also observed two ligand-free Anc^Int classes in high-affinity states, with substrate-binding and Na⁺-binding site architectures like those of Glt_Ph after Na⁺ binding to Na1 and Na3 sites (Fig. 4i,m). The NMD motifs are flipped in, R397₃₆₉ and N401₃₇₃ are correctly oriented for l-aspartate coordination (Fig. 4i), and all residues in Na1 and Na3 sites are configured like in Na⁺-bound Glt_Ph (Fig. 4m). Water molecules likely occupy the sites in the absence of Na⁺ but the resolution is insufficient to model them, and computational studies will be needed to understand how they compensate for the missing ions. One of the classes shows an HP2 structurally identical to substrate-bound Anc^Int, except with an unresolved tip (the loop connecting the two helical arms of the hairpin) and an empty substrate-binding pocket (Extended Data Fig. 8d). In the other class, an ambiguous density occupies the pocket that is too large for l-aspartate and might correspond to a spuriously bound buffer component; this site might be generally attractive to anions²⁸ (Extended Data Fig. 8b). We excluded this structure and inward-facing classes from further consideration; the latter are poorly resolved and have notable regions of missing density in the transport domain (Extended Data Fig. 7). We expect similar structures of the substrate-binding and ion-binding sites in the inward-facing and outward-facing states on the basis of observations in other glutamate transporter family members^15,16,30.

Binding sites are unaltered in uncoupled transporters

Imaging of aspartate-bound Anc^Int in Na⁺-free conditions revealed a single outward-facing structural class, which we refined to 3.4-Å overall resolution (Extended Data Figs. 6 and 7, Supplementary Fig. 4 and Table 1). The only structural differences between aspartate-bound Anc^Int and high-affinity ligand-free Anc^Int are that its HP2 tip is well structured and closed, and the substrate-binding pocket is occupied by a strong nonprotein density, which we modeled as l-aspartate (Fig. 4j,k,n and Extended Data Fig. 8e). The substrate orientation and binding site architecture of Anc^Int in the absence of alkali cations were nearly identical to Na⁺/aspartate-bound Glt_Ph (Fig. 4d,k). The NMD motif is flipped in with the M311₂₈₉ side chain pointing into the substrate-binding pocket and forming van der Waals interactions with l-aspartate, which is coordinated by side chains of conserved T314₂₉₂, D394₃₆₆, R397₃₆₉ and N401₃₇₃ (Fig. 4k). R397₃₆₉ is the principal residue interacting with the side-chain carboxylate of l-aspartate and D394₃₆₆ coordinates the amino group; these residues determine the specificity of the transporter for the acidic over neutral amino acids and dicarboxylates, respectively^58,59,60. N401₃₇₃, one of the most conserved residues throughout the evolution of glutamate transporters, coordinates the main chain carboxylate of l-aspartate. At the current resolution, the architectures of all three Na⁺-binding sites in Anc^Int are identical to Na⁺/aspartate-bound Glt_Ph (Fig. 4n), although Na⁺ is absent.

Our results suggest that ligand-free Anc^Int spontaneously populates the high-affinity state with the correctly formed substrate-binding and ion-binding sites, facilitating substrate binding without Na⁺ binding. In contrast, this is a high-energy state in Glt_Ph and other Na⁺-coupled transporters, only accessible upon Na⁺ binding.

Allosteric mutations turn ion coupling back on

The amino acid sequence changes underlying the transition from Na⁺ coupling in Anc^Na to independence in Anc^Int are allosteric and lie outside the ion-binding sites. Anc^Na and Anc^Int are 78% identical over the transport domain sequence (Extended Data Fig. 4c), with a difference in 26 amino acids and PPs above 0.9 in at least one of the ML sequences (Extended Data Fig. 4d). The changes, however, are subtle, making it difficult to pinpoint those responsible for gaining independence from Na⁺ ions. Our phylogenetic tree shows that the evolution from Na⁺-coupled to uncoupled function occurred through several ancestral intermediates between Anc^Na and Anc^Int, which might have gradually acquired the mutations. To narrow down the search for changes necessary for the functional switch, we examined the ancestor immediately preceding Anc^Int, Anc*, and a present-day transporter most homologous to Anc*, Ext* (KJS87745.1) (Fig. 1d). Anc* and Ext* showed robust Na⁺-coupled binding and uptake, respectively, suggesting they are Na⁺ coupled (Fig. 5a and Extended Data Fig. 9).

**Fig. 5: TM7 is the allosteric determinant of Na⁺ coupling.**

There are differences in 18 amino acids between Anc* and Anc^Int (95.6% identical), and only two of these are located in the transport domain and have high PPs: S295₂₇₃P (0.969) and N327₃₀₅Q (0.996). These residues are positioned at the beginning and end of TM7, respectively (Fig. 5b). We hypothesized that the reverse substitutions in Anc^Int, P295₂₇₃S and Q327₃₀₅N, would be sufficient to switch ion coupling back on. Indeed, in proteoliposomes, the mutant exhibited Na⁺-dependent concentrative l-aspartate transport (Fig. 5c). In the thermal shift assays, l-aspartate stabilized the mutant when added with Na⁺ ions but not without them (Fig. 5d and Supplementary Fig. 7). Na⁺ alone induced a deeper and sharper first-derivative trough, consistent with ions binding and increasing the cooperativity of the unfolding process. Both substitutions appear to be necessary, as single mutants showed changes in the melting curves suggesting that the substrate binds without Na⁺ (Extended Data Fig. 9).

We also collected a cryo-EM dataset on Anc^Int P295₂₇₃S;Q327₃₀₅N in ligand-free conditions. At the achieved resolution of 5.0 Å, we could not model side chains and the NMD loop; nevertheless, the density of all observed structural classes aligned well with the TM3, TM7b, HP2 and TM8a regions of the low-affinity but not the high-affinity structures of Anc^Int (Supplementary Fig. 8). Together, these results suggest that substitutions can rescue Na⁺ coupling in Anc^Int by shifting the equilibrium of the ligand-free transporter back to the low-affinity state, thereby restoring coupled substrate and Na⁺ binding. Interestingly, introducing the forward substitutions (S295₂₇₃P and N327₃₀₅Q) into Anc^Na did not turn off Na⁺ coupling (Extended Data Fig. 9). Furthermore, introducing these substitutions in Anc* and Ext* resulted in a misfolded, unstable protein. Thus, TM7 mutations are necessary but insufficient to turn ion coupling off. Additional mutations might be required to compensate for negative fitness effects and shift energetics to complete uncoupling.

We compared the low-affinity and high-affinity states of Na⁺-coupled Glt_Ph (Fig. 5e–h) and uncoupled Anc^Int (Fig. 5i–l) to understand how allosteric mutations might affect their equilibrium. TM7 runs through the core of the transport domain with the NMD motif at its center (Fig. 5b,e,i). In Na⁺-coupled transporters Glt_Ph and human EAAT3, TM7b shows a displacement in the low-affinity ligand-free state compared to the high-affinity Na⁺-bound state (Fig. 5e,f and Supplementary Table 1), translating into repacking of the surrounding TM3, TM6, TM8 and HP2 (Fig. 5g,h). Packing differences in this region affect Glt_Ph affinity for l-aspartate and are, therefore, allosterically coupled to the changes at the binding site^3,61,62. In contrast, TM7b of Anc^Int shows minimal tilt changes and less helical repacking in the low-affinity state compared to the high-affinity state (Fig. 5i–k), suggesting that S295₂₇₃P and N327₃₀₅Q substitutions stabilize the high-affinity state. P295₂₇₃ is in the loop preceding TM7a and might rigidify the loop compared to serine (Fig. 5b,i). Q327₃₀₅ is the last residue of a highly conserved F-[I/V]-A-[N/Q] signature sequence motif in TM7b⁶³ (Fig. 5b,i). Q327₃₀₅ hydrogen-bonds to the top of TM3 (Fig. 5l), perhaps stabilizing TM7–TM3 interactions. By contrast, N327₃₀₅ in Glt_Ph is too short to reach TM3 (Fig. 5h). These analyses suggest that the structural flexibility of the central TM7 determines the equilibrium between the transporter states, and a more rigid TM7 in Anc^Int tightly packed to surrounding helices might eliminate the need for Na⁺ binding to achieve the high-affinity state.

Discussion

Na⁺-coupled glutamate transport is one of the most thoroughly characterized ion-coupled transport mechanisms, where Na⁺ binding is inextricably linked to substrate binding. Yet, how these binding sites are allosterically coupled, and thus the design principles of Na⁺-coupled transporters, were unknown. We gained unprecedented insights into this fundamental question by examining the evolution of prokaryotic glutamate transporters using phylogenetics and APR. Surprisingly, the transition from powering transport with Na⁺-to-H⁺ gradients occurred through an intermediate ancestral transporter Anc^Int, in which l-aspartate binding and transport were uncoupled from Na⁺ binding. Our reconstructed ancestors seem to show slow transport compared to extant transporters, although we did not systematically study this. Reconstructed ancestral proteins are generally more thermostable than extant proteins, which is an established artifact of APR^4,64. This thermostability might manifest as slow transitional kinetics and slow transport rates.

Unlike Na⁺-coupled homologs, Anc^Int does not require Na⁺ to form high-affinity substrate-binding and ion-binding sites. Thus, we propose that the relative free energies of the low-affinity and high-affinity states are the linchpin of ion coupling (Fig. 6a–c). In Na⁺-coupled transporters, the free energy of the low-affinity conformation is much lower than the high-affinity conformation (Fig. 6b). The substrate-binding energy alone is insufficient to overcome the penalty, and Na⁺ binding pays the energetic price, stabilizing the state. In uncoupled Anc^Int, the low-affinity state has a similar or higher energy, and the high-affinity state is populated without ions (Fig. 6c). Na⁺ ions can still bind, possibly with higher affinity because of preformed Na⁺ sites. However, they are unnecessary for substrate binding.

**Fig. 6: Evolution of the transporter energy landscape controls Na⁺ coupling.**

We found two high-probability allosteric mutations between uncoupled Anc^Int and the previous ancestor inferred to be Na⁺ coupled. In Anc^Int, these were sufficient to restore Na⁺-coupled binding and transport, likely by stabilizing the low-affinity state. However, it seems likely that before this switch, transporters accumulated other neutral epistatic mutations during the evolution from Na⁺-coupled Anc^Na to Anc^Int, which gradually increased the relative energy of the low-affinity state. Then, the final two changes were sufficient to make the high-affinity state accessible and function independent of Na⁺ (Fig. 6c). Such allosteric alterations of energy landscapes might underlie the evolutionary diversification of many proteins^65,66,67,68. This mechanism explains how a few mutations can turn off Na⁺ coupling without requiring several mutations to eliminate the three Na⁺-binding sites, which would diminish the transporter’s ability to bind substrate and overall cellular fitness.

Was the transition to uncoupled substrate binding a random evolutionary event or an adaptation required for organismal fitness? While functional studies of Anc^Int descendants will need to be carried out, most are unlikely to be ion coupled according to the lack of Na⁺-coupled or H⁺-coupled signature residues. Transporter copies or transporters from different families with similar functions could compensate for the loss of ion-coupled transport⁶⁹, which still raises the question of the physiological relevance of uncoupled transporters. Examining the gene neighbors of uncoupled transporters revealed that many are adjacent to enzymes involved in amino acid metabolism; we observed no similar preferences in the Na⁺-coupled or H⁺-coupled gene neighbors (Extended Data Figure 10). Thus, uncoupled transporters might mediate amino acid uptake paired with their enzymatic use, ultimately maintaining the inward concentration gradient required for continuous uncoupled transport. Such pairing would be reminiscent of glucose transporters, which pair uncoupled transport with subsequent glucose metabolism, a key interplay for insulin secretion⁷⁰.

In summary, we show that phylogenetic analysis and APR can uncover the evolutionary trajectory of functional diversification in transporters and pinpoint amino acid changes that alter their function. Comparative studies of ion-coupled and ion-independent proteins show that ion coupling can be allosterically regulated and turned on and off by mutations and perhaps environmental changes, which reshape the energy landscape of the transporter. We suggest that drugs could also allosterically modulate the extent of ion and substrate coupling in secondary active transporters, allowing a novel inhibitory modality in which transport is not abolished but its concentrative capacity is diminished. Such drugs could offer advantages in cases where complete inhibition might be harmful.

Methods

Sequence retrieval, curation and alignment

We designed a workflow to construct ML phylogenies capturing the Na⁺-coupled to H⁺-coupled (Ext^Na to Ext^H) evolutionary transition on the basis of previous APR studies⁴⁷ (Extended Data Fig. 2). Biochemically characterized Ext^H proteins feature signature sequence substitutions in the Na⁺-binding sites compared to Ext^Na transporters^2,25,50: [T/S]92[A/L/I/V/M/F], D405N and M311L (numbering is based on Ext^Na archaeal transporter Glt_Ph) (Fig. 1c). Putative Ext^H genes carrying these signature residues are widespread throughout bacterial phyla. We constructed a maximum-parsimony tree of all bacterial glutamate transporter homologs and found that putative Ext^H transporters cluster together regardless of phylum (Extended Data Fig. 3a). Although we cannot exclude the existence of H⁺-coupled transporters with completely different signature residues, we infer that the emergence of Ext^H was a single-gene duplication event, after which Ext^H genes spread through horizontal gene transfer.

Aiming to have a sequence set with a size amenable to APR, we retrieved 20,000 nonredundant Bacillota/Firmicutes sequences using PSI-BLAST (National Center for Biotechnology Information) because Ext^H sequences were predominantly found in this phylum. We used WP_011876993.1 as the original query, which we estimated would be close to both Na⁺-coupled and H⁺-coupled transporters^71,72. After clustering by 99% using CD-HIT⁷³, removal of sequences with ambiguous amino acid assignment and alignment by MUSCLE v5 (ref. ⁷⁴), a maximum-parsimony tree was generated by MP-boot⁷⁵ with the remaining 9,891 sequences. Subsets of sequences were selected on the basis of proximity to sequences with signatures of H⁺ coupling and nearby sequences up until those with signature Ext^Na residues (Extended Data Fig. 3b), resulting in 5,249 sequences. The remaining sequences were clustered to 90% using CD-HIT and the resulting 2,539 sequences were manually curated to exclude sequences with major gaps or insertions. Archaeal Glt_Ph (UniProt O59010) was included to be the tree root. The final sequence set (1,410 sequences) was aligned by MAFFT-DASH, which incorporates information from Protein Data Bank (PDB) structures into the alignment matrix⁷⁶. Finally, variable loops and tails were removed and the resulting alignment was manually adjusted to account for poorly aligned regions.

Phylogenetic analysis, ancestral protein reconstruction and construct design

ML trees and ancestral sequences were generated using IQ-TREE 2.1.4-beta (ref. ⁷⁷). Evaluation of all potential evolution models and rate heterogeneity parameters using ModelFinder revealed that LG + F + R was the best protein evolution model^78,79,80. FreeRate (F) heterogeneity considers site-specific differences in evolutionary rate, relaxing the constraint of equal site probabilities imposed by discrete gamma rate heterogeneity. The number of possible rate categories that could be assigned to a site (R) was empirically determined for each run using ModelFinder to query a range of 10–20.

As with other APR studies of microbial proteins, our tree topology is not reconciled to the microbial species tree, as this has poor confidence especially at lower taxonomic ranks⁸¹. We chose the outgroup-based approach of rooting the tree, using archaeal Glt_Ph as the outgroup because of the evolutionary distance between bacteria and archaea; Glt_Ph was also distant enough to the ingroups and did not introduce any long-branch attraction artifacts. Most archaeal transporters contain all Na⁺-coupling signature residues (Extended Data Fig. 3c), suggesting that the origin of glutamate transport was sodium coupled and the divergence into sodium-independent exchangers and subsequent proton-coupled transporters occurred after the divergence of bacteria and archaea. After performing over 60 replicate tree runs, we found that the best Firmicutes tree yielded robust UFBoot2 values, ranging between 91% and 100% for the Ext^H node (Fig. 1d). When we repeated the analysis for the Proteobacteria phylum, the best tree had 5–87% UFBoot2 values for the equivalent nodes, reflecting nonreproducible trees. Thus, we used the Firmicutes tree for all subsequent analyses.

The run yielding the best tree, as determined by log likelihood, was inspected for long branches, defined as >0.7 substitutions per site. These were removed and, in the second round of tree-searching with the final dataset of 1,153 sequences, the analysis was repeated with 254 independent tree search runs. The final tree had the best-fit model of LG + F + R14, chosen according to the Bayesian information criterion. Bootstrap values were determined by UFBoot2 using 1,000 replicates⁴⁵. Ancestral sequences were reconstructed from the best tree using the IQ-TREE implementation⁷⁷ based on Yang’s marginal reconstruction method^82,83,84. The ML sequences were generated using custom Python scripts, either written manually or generated using ChatGPT. The final ML amino acid sequence for each node was constructed using the amino acid with the highest PP at every site. All tree figures were generated using TreeViewer⁸⁵. Visual representations of sequence alignments and PPs were made using WebLogos⁸⁶.

Protein expression and purification

Because poorly aligned loops and tails are not reconstructed in our APR analysis, we had to graft them in for expression constructs. We took comparable loops and tails from an extant sequence close to Anc^Int (KJS87745.1); the 3L4 has two lysines that were substituted to histidines (K105H;K109H) to prevent potential proteolysis. All protein constructs were codon-optimized for E. coli, synthesized (GenScript) and cloned into a modified pBAD vector containing a C-terminal thrombin cleavage site followed by either an ALFA peptide tag⁸⁷ or a C-terminal thermostable GFP specialized for membrane proteins⁴⁹. An additional Twin-Strep tag and a 10×His tag were added to the C terminus of these constructs, all separated by glycine–serine (GSSS) linkers. The Glt_Ph construct used was the ‘CAT7’ construct, with seven histidines previously introduced for stability¹³. Vectors were transformed into E. coli BL21(DE3) or DH10B cells (New England Biolabs, C2527; Invitrogen, 12331013), cells were grown in 2×YT medium and expression was induced by adding 0.1% arabinose at 0.8–1.0 OD₆₀₀. Proteins were expressed for 2–3 h at 37 °C. Membranes were purified as previously described for Glt_Ph, with all steps being performed at 4 °C (ref. ³). Briefly, cells were broken using a high-pressure cell disruptor (Avestin) for three passes with cooled water continuously applied. The resulting lysate was spun at 17,400g for 15 min and the supernatant was spun at 186,000g for 1 h. The resulting membrane pellet was homogenized in 20 mM HEPES–NaOH pH 7.4, 10 mM EDTA and 10% sucrose. Finally, the slow and fast spins were repeated and the resulting pellet was frozen at −80 °C.

All protein purification steps were performed at 4–8 °C. Crude membranes were homogenized in 20 mM HEPES–NaOH pH 7.4, 200 mM NaCl and 1 mM monopotassium l-aspartate; solubilization was initiated by the addition of 40 mM n-dodecyl-β-d-maltopyranoside (DDM; Anatrace) and the solubilizing membranes were gently rocked for 2 h. Over the course of our experiments, we noticed that tags were being nonspecifically proteolyzed upon solubilization; to reduce this, we added 1 mM PMSF and 1 mM EDTA before adding DDM. After solubilization, samples were clarified by high-speed ultracentrifugation at 186,000g for 1 h.

For ancestral constructs, the supernatant was applied to a prepacked 1-ml Streptactin XT 4Flow high-capacity column (IBA Biosciences) at a constant flow rate of 0.5–1 ml min⁻¹. The resin was washed until the ultraviolet absorbance at 280 nm (UV₂₈₀) was flat, corresponding to 14 column volumes of 20 mM HEPES–NaOH pH 7.4, 200 mM NaCl, 1 mM l-aspartate and 1 mM DDM. Subsequently, the resin was eluted with the same buffer containing 50 mM biotin (IBA Biosciences) and proteins were concentrated.

For Glt_Ph, the supernatant was applied to a 1-ml HisTrap HP (Cytiva) at a constant flow rate of 1 ml min⁻¹. The resin was washed until the UV₂₈₀ was flat, corresponding to 30 column volumes of 20 mM HEPES–NaOH pH 7.4, 200 mM NaCl, 1 mM l-aspartate, 40 mM imidazole and 1 mM DDM. Subsequently, the resin was eluted with the same buffer containing increased imidazole (250 mM) and proteins were concentrated.

Reconstitution into proteoliposomes

Unilamellar liposomes were prepared essentially as previously described³. Briefly, a 3:1 (w/w) ratio of E. coli polar lipids and egg phosphatidylcholine (Avanti Polar Lipids) was dried, hydrated in 50 mM HEPES–Tris pH 7.4 and 200 mM KCl at a final concentration of 5 mg ml⁻¹, flash-frozen and stored at −80 °C. On the day of the reconstitution, liposomes were thawed, extruded through a 400-nm filter 13 times and destabilized with Triton X-100 using a 1:2 (w/w) ratio of Triton X-100 to lipid. For the counterflow and Na⁺-coupled uptake experiments, GFP-free proteins were purified by size-exclusion chromatography (SEC) using a Superdex 200 Increase 10/300 column (Cytiva) in 20 mM HEPES–NaOH pH 7.4, 200 mM NaCl, 1 mM l-aspartate and 1 mM DDM; proteins were subsequently concentrated using a 100-kDa-cutoff concentrator. Proteins in all comparative experiments (Fig. 2d–g) were purified and reconstituted alongside one another. To form proteoliposomes, the purified protein was mixed with destabilized liposomes (1:100, w/w) for 30 min; then, detergent was removed with BioBeads SM-2 (Bio-Rad) at 80 mg ml⁻¹, which were applied twice at 22 °C for 2 h, once overnight at 4 °C and once more at 22 °C for 2 h.

Proteoliposome uptake experiments

After detergent removal with BioBeads, proteoliposomes underwent the same general procedure for buffer exchange and concentration, regardless of experiment. Proteoliposomes were exchanged into the appropriate buffer using a PD-10 column (Cytiva), flash-frozen in liquid N₂, thawed, concentrated by ultracentrifugation (100,000g for 1 h), resuspended to a final concentration of 50 mg ml⁻¹, flash-frozen and stored at −80 °C. On the day of the experiment, the proteoliposomes were thawed and extruded through a 400-nm filter 23 times. To minimize sample dilution in the extruder, the proteoliposome volume was no less than 100 µl. All subsequent experiments were performed at 30 °C.

For uptake experiments, liposomes were prepared as described above to a final concentration of 50 mg ml⁻¹ in 20 mM HEPES–Tris pH 7.4, 200 mM KCl and 100 mM choline chloride. To initiate Na⁺-driven uptake in the presence of a negative membrane potential (−102 mV), liposomes were diluted 1:100 in 20 mM HEPES–Tris pH 7.4, 2 mM KCl, 198 mM choline chloride, 100 mM NaCl, 0.9 µM valinomycin and 1 µM l-aspartate. The l-aspartate was a mixture of cold monopotassium l-aspartate (Research Products International) and [³H]l-aspartate (PerkinElmer or American Radiolabeled Chemicals). For negative controls without the Na⁺ gradient, NaCl was replaced with choline chloride.

To stop the reactions, 200 µl of reaction mixture (100 µg of liposomes) was diluted in 2 ml of ice-cold quenching buffer (20 mM HEPES–Tris pH 7.4 and 200 mM lithium chloride). The quenched reaction was immediately filtered through a 0.22-µm filter membrane (Millipore Sigma) and washed three times with 2 ml of quenching buffer. Washed membranes were inserted into scintillation vials and the membranes were soaked overnight in 5 ml of Econo-Safe counting cocktail. Radioactivity in liposomes was measured using an LS6500 scintillation counter (Beckman Coulter). Technical replicates from an individual reconstitution were taken in duplicate and averaged and at least two separate reconstitutions were taken per point. Data were analyzed using GraphPad Prism 10.

Proteoliposome counterflow experiments (sodium-dependent)

For counterflow experiments, the proteoliposomes after BioBead detergent removal were similarly exchanged in 20 mM HEPES–Tris pH 7.4, 200 mM KCl and 100 mM choline chloride using PD-10 columns. The exchanged liposomes were then separated into tubes for sodium conditions and choline conditions. For sodium conditions, an equivalent buffer containing 100 mM sodium chloride instead of choline chloride was added after exchange so that the final sodium concentration was 10 mM and the final choline concentration was 90 mM. Choline conditions were treated similarly, with the original choline buffer instead of the sodium buffer. A final concentration of 0.5 mM monopotassium l-aspartate was added to both conditions. Finally, the liposomes were flash-frozen and thawed twice, concentrated by ultracentrifugation (100,000g for 1 h), resuspended in the same internal buffer at 50 mg ml⁻¹, flash-frozen and stored at −80 °C as described in the uptake experiments.

On the day of the experiment, proteoliposomes were thawed and freshly extruded, as described in the uptake experiments. Counterflow was initiated by diluting the liposomes 1:100 in an external buffer equivalent to the internal buffer, except that the cold 0.5 mM l-aspartate was replaced by 0.5 µM [³H]l-aspartate (American Radiolabeled Chemicals). Negative controls had an additional 0.5 mM cold l-aspartate in the external buffer. Reactions proceeded for the indicated time and were stopped and measured as described in the uptake experiments. Data were analyzed using GraphPad Prism 10.

Proteoliposome counterflow experiments (pH-dependent)

Proteoliposomes after BioBead detergent removal were exchanged in 50 mM potassium phosphate (KH₂PO₄–K₂HPO₄) buffer plus 0.5 mM l-aspartate (pH 6, 7 or 8) by three rounds of resuspension, two freeze–thaw cycles and ultracentrifugation (100,000g for 1 h). These were performed so that the total internal buffer exchange was greater than 99.5%. After the final round, proteoliposomes were resuspended to a final concentration of 25 mg ml⁻¹, flash-frozen and stored at −80 °C. On the day of the experiment, proteoliposomes were thawed and freshly extruded as described in the uptake experiments. Counterflow was performed as in the sodium-dependent counterflow experiments, except the dilution factor was 1:50 and the quenching buffer was 50 mM MES–KOH buffer pH 5.5. Data were analyzed using GraphPad Prism 10.

nanoDSF

All steps were performed on the same day of the protein solubilization and purification because we observed variance in results depending on the freshness of the protein. We opted to use GFP-tagged ancestral proteins because the T_m of the fused GFP is greater than 90 °C (ref. ⁴⁹), which does not interfere with the T_m values observed in our experiments. After affinity purification, the proteins with C-terminal GFP were concentrated using a 100-kDa-cutoff concentrator (MilliporeSigma). To remove ligands, the protein was dialyzed using a PD MiniTrap G-25 (Cytiva). For Anc^Na and Anc^Int P295₂₇₃S;Q327₃₀₅N, this buffer was 20 mM HEPES, 200 mM NMDG-Cl pH 7.4 and 1 mM DDM; for Anc^Int, this buffer was 50 mM KH₂PO₄–K₂HPO₄ pH 7 and 1 mM DDM. Different buffers were used for optimal first-derivative curves in ligand-free conditions and we observed identical T_m values in substrate-bound conditions regardless of dialysis buffer composition. The dialyzed, ligand-free protein was concentrated and diluted into the same respective buffers at a final concentration of ~1 mg ml⁻¹; for Anc^Int, the buffer contained 10 mM decyl-maltoside (DM) instead of 1 mM DDM. This was necessary to observe melting of Anc^Int, presumably by the shorter chain detergent destabilizing the overall protein. Thermal stability in the presence and absence of various potential ligands was measured using a Tycho NT.6 instrument (Nanotemper) and individual T_m values were calculated by determining the appropriate local minimum in the first derivative of the 350-nm:330-nm ratio. Data were analyzed using GraphPad Prism 10.

MST

After affinity purification, the proteins with C-terminal GFP were concentrated using a 100-kDa-cutoff concentrator and then SEC-purified on a Superose 6 Increase 10/300 column (Cytiva) in 50 mM KH₂PO₄–K₂HPO₄ buffer pH 7 and 1 mM DDM. For the MST experiment, monopotassium l-aspartate was serially diluted twofold in water for a total of 16 concentrations. Each aspartate dilution was mixed with in a 1:1 (v/v) ratio of protein samples so that the final protein concentration was 100–300 nM, the appropriate salt was 100 mM, the DDM was 0.4 mM and the KH₂PO₄–K₂HPO₄ was 10–20 mM. Measurements were taken using Monolith NT.115 Blue/Red MST (Nanotemper), high MST power, a blue detector (for GFP) and premium capillaries (Nanotemper).

For data analysis, obvious outliers caused by aggregates (as determined by instrument software, MO.Control 2) were discarded. The ideal response time for each individual experiment was determined by instrument software and this was always between 1.5–5 s. Individual binding titrations were normalized to fraction bound by K_D fits in MO.Affinity Analysis version 2.3. To obtain the final binding affinity value, the experiments were repeated three independent times, the data points were averaged and the K_D was calculated from the average values. Data were analyzed using GraphPad Prism 10.

Cryo-EM sample preparation

Membrane scaffold protein (MSP1E3) was expressed and purified as previously described⁸⁸ in the absence of Na⁺ or l-aspartate. A 3:1 (w/w) ratio of E. coli polar lipids and egg phosphatidylcholine (Avanti Polar Lipids) was dried, hydrated in 50 mM HEPES–Tris pH 7.4 and 200 mM KCl and solubilized in 72 mM DDM at final lipid concentration of 18 mM as previously described^15,88.

For Anc^Int, GFP-free transporter constructs were affinity-purified, concentrated and mixed with MSP1E3 and solubilized lipids (both containing no Na⁺ or l-aspartate) at a molar ratio of 1.5:1:50, aiming for a final concentration of lipid greater than 5 mM. This mixture was incubated at 22 °C for 30 min; then, 350 mg of BioBeads SM-2 (Bio-Rad) per ml of mixture (w/v) was added for 2 h at 22 °C with gentle shaking and another round of BioBeads was applied overnight at 4 °C with gentle shaking. The resulting MSP1E3 and protein preparations were cleared by ultracentrifugation at 100,000g for 30 min. The supernatant was applied to a Superose 6 Increase 10/300 column in 20 mM HEPES pH 7.4, 100 mM NMDG-Cl and the peak fractions were concentrated to 7–9 mg ml⁻¹, as measured by absorbance at 280 nm (A₂₈₀). For the Anc^Int substrate-bound dataset, 1 mM l-aspartate was added to the sample just before freezing. Reconstitution efficiency was verified using SDS–PAGE.

Glt_Ph was reconstituted into MSP1E3 as previously described¹⁵. Affinity-purified protein was further purified by SEC (Superdex 200 Increase, 10/300). Peak fractions were concentrated, mixed with MSP1E3 and solubilized lipids at a molar ratio of 0.75:1:50, aiming for a final concentration of lipid greater than 5 mM. This mixture was incubated and treated with BioBeads SM-2 in the exact same fashion as ancestral protein. The resulting MSP1E3 and protein preparations were exchanged in Na⁺-free buffer by three rounds of tenfold dilution and concentration and cleared by ultracentrifugation at 100,000g for 30 min. The supernatant was applied to a Superose 6 Increase 10/300 column in 20 mM HEPES–Tris pH 7.4 and 50 mM choline chloride, with the goal of mimicking previously published conditions. The peak fractions were concentrated to 8 mg ml⁻¹, as measured by A₂₈₀. Reconstitution efficiency was verified using SDS–PAGE.

To improve particle distribution, fluorinated Fos-Choline-8 (Anatrace) dissolved in water was added to the sample to a final concentration of 1.5 mM. UltrAuFoil R1.2/1.3 300-mesh gold grids (Quantifoil) were glow-discharged at 25 mA for 80 s. Then, 3 µl of protein was applied to each grid, incubated for 20 s under 100% humidity at 4 °C, blotted for 2.5 s under 0 blot force and plunge-frozen in liquid ethane using a Vitrobot Mark IV (Thermo Fisher Scientific).

Cryo-EM data acquisition and processing

Imaging was performed on a Titan Krios (FEI) equipped with a K3 camera (Gatan) using Leginon software. Specifics of each data collection and each individual map and model are provided in Table 1. Visual summaries of processing workflows and EM validation are provided in Extended Data Fig. 6 and Supplementary Figs. 4–6 and 8.

Initial processing pipeline for Anc^Int

For all datasets, raw videos were imported into cryoSPARC 4.0–4.4 and subjected to patch motion correction and patch contrast transfer function (CTF) estimation, where the videos were binned twice⁸⁹. Low-quality videos were removed, as determined by CTF fits, ice thickness, intensity, etc. Particles were manually picked to train Topaz⁹⁰, where, for each dataset, the optimal picking model and estimated number of particles per micrograph were empirically determined using the Topaz cross-validation tool in cryoSPARC. After picking and extraction with four binning rounds, an ab initio 3D model was constructed and high-quality particles were sorted using multiple heterogeneous refinement jobs, where ‘good’ particles sorted into the ab initio model class and ‘junk’ particles sorted into three pure noise ‘decoy’ volume classes (created by performing one iteration of ab initio). When greater than 95% of particles converged into the protein volume, the remaining particles were re-extracted to full box size and subjected to 3D classification without alignment and nonuniform (NU) refinement⁹¹. To maximize the number of particles obtained, the Topaz model was retrained on the clean particle stack after rounds of two-dimensional classification and heterogeneous refinement was repeated with the new particle stack and the high-resolution volume from NU refinement (replacing the ab initio volume). Remaining particles were re-extracted to full box size and subjected to NU refinement.

Anc^Int ligand-free dataset processing

Our ancestral protein datasets were highly heterogeneous and poorly symmetric; throughout the processing, we realized that they were reminiscent of other seemingly mobile membrane protein structures^92,93. To overcome this, we attempted to enrich particles with symmetric features so that we could classify individual protomers. In the ligand-free dataset, we converted particle coordinates to RELION format using PyEM⁹⁴. Particles were re-extracted in RELION using videos motion-corrected and binned twice using RELION’s implementation of MotionCor2 (ref. ⁹⁵). These particles were reimported into cryoSPARC and refined using NU refinement (C₁); the resulting particle images were converted to RELION format with PyEM. These particles were then subjected to two rounds of Bayesian polishing⁹⁶, classification with heterogeneous refinement and classification with NU refinement (C₁). To remove ‘disordered’ particles, we used 3D classification without alignment (C₁) in cryoSPARC. We further optimized symmetric particles by performing one round of ab initio with C₃ symmetry imposed. Using the best class, we achieved a resolution of 2.8 Å in C₃, which allowed us to resolve individual protomers using symmetry expansion and focused classification using a mask on the transport domain. Each class underwent local refinement masked on the entire protomer (including transport and scaffold domains) using pose and shift Gaussian priors during alignment (standard deviations of 3° and 2 Å).

Anc^Int substrate-bound processing

As with the ligand-free dataset, the substrate-bound dataset was similarly heterogeneous and asymmetric. Using the previous approach, we encountered trouble imposing C₃ symmetry, preventing focused classification of individual protomers. During our processing, we realized that we could more efficiently remove disordered particles using outputs from 3D variability (3DVA), where the first variability component reflects this ‘order–disorder’ transition between rigid, easily aligned particles and presumably mobile, difficult-to-align particles. We then took the first frame of the 3DVA video (disordered, bad volume), the last frame of the video (ordered, good volume) and two decoy noise volumes as inputs for heterogeneous refinement. Though a fundamentally similar approach to 3D classification used in the ligand-free dataset, the new approach allowed us to impose C₃ symmetry with good resolutions (3.4 Å). We then symmetry-expanded the particles, performed 3D classification without alignment on the masked transport domain and locally refined the protomers using pose and shift Gaussian priors during alignment (standard deviations of 3° and 2 Å), as with the ligand-free dataset.

Glt_Ph ligand-free processing

Raw videos were imported into cryoSPARC 4.4 and subjected to patch motion correction and patch CTF estimation, where the videos were binned twice⁸⁹. Low-quality videos were removed, as determined by CTF fits, ice thickness, intensity, etc. Particles were manually picked to train Topaz⁹⁰, where the optimal picking model and estimated number of particles per micrograph were empirically determined using the Topaz cross-validation tool in cryoSPARC. After picking and extraction, an ab initio 3D model was constructed. After one round of heterogeneous refinement with ab initio volumes, we performed NU refinement (C₃) on the remaining particles, symmetry-expanded the particles, performed 3D classification without alignment on the masked transport domain and locally refined the protomers using pose and shift Gaussian priors during alignment (standard deviations of 3° and 2 Å).

Anc^Int P295₂₇₃S;Q327₃₀₅N ligand-free dataset processing

Because of the substantially lower resolution of the dataset, previously described methods to improve the resolution after decoy classification (3DVA and 3D classification without alignment) did not work. Therefore, we immediately performed 3D classification (C₃) after decoy classification. We then performed symmetry expansion and 3D classification without alignment on the masked transport domain.

Model building and refinement

For model building, maps were sharpened using DeepEMhancer⁹⁷. We generated a homology model of Anc^Int in SWISS-MODEL, using an outward-facing state of Glt_Ph (PDB 2NWX) as a template^13,98. The template for Glt_Ph was PDB 6UWF (outward-facing) or PDB 6UWL (intermediate outward-facing). The model was adjusted into the density using ISOLDE⁹⁹ and the final model was iteratively real-space refined and validated in PHENIX using the unsharpened density and default parameters¹⁰⁰. All structural figures were made with ChimeraX¹⁰¹.

Gene neighborhood analysis

For each clade (Ext^Na, Ext^Int and Ext^H), a sequence similarity network and corresponding genome neighborhood was created using the Enzyme Function Initiative webserver¹⁰². Transporters without information regarding protein-coding gene neighbors were excluded. A transporter was considered paired to an enzyme if they were within three neighboring genes away from the transporter. Genes were considered to have amino-acid-modifying or dicarboxylate-modifying activity according to Pfam annotation.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

Atomic coordinates and EM densities were deposited to the PDB and EM Data Bank for the following structures: ligand-free Anc^Int low-affinity (PDB 9BGY, EMD-44526); ligand-free Anc^Int high-affinity (PDB 9BGZ, EMD-44527); l-aspartate-bound Anc^Int (PDB 9BH0, EMD-44528); ligand-free Glt_Ph intermediate outward-facing (PDB 9BH1, EMD-44529); ligand-free Glt_Ph outward-facing (PDB 9BH2, EMD-44530). Supplementary Data 2 includes all input and output data related to phylogenetic analyses, plasmid maps of all protein expression constructs used in this study and a ChimeraX session file containing comparison of Anc^Int models to Anc^Int P295₂₇₃S;Q327₃₀₅N maps (Extended Data Figure 9). Plasmid backbones designed in this study are available on Addgene (pBAD24-ALFA, 240248; pBAD24-TGP, 240249; pET15b-ALFA, 240252; pET15b-TGP, 240251). Previously published structural models are available from the PDB under accession codes 2NWX, 6UWF, 6X12, 6X15, 7AHK, 7RCP, 8CUI and 8CV2. All other plasmids and data are available upon request. Source data are provided with this paper.

References

Ryan, R. M., Compton, E. L. R. & Mindell, J. A. Functional characterization of a Na⁺-dependent aspartate transporter from Pyrococcus horikoshii. J. Biol. Chem. 284, 17540–17548 (2009).
Article CAS PubMed PubMed Central Google Scholar
Gaillard, I., Slotboom, D. J., Knol, J., Lolkema, J. S. & Konings, W. N. Purification and reconstitution of the glutamate carrier GltT of the thermophilic bacterium Bacillus stearothermophilus. Biochemistry 35, 6150–6156 (1996).
Article CAS PubMed Google Scholar
Reddy, K. D., Ciftci, D., Scopelliti, A. J. & Boudker, O. The archaeal glutamate transporter homologue GltPh shows heterogeneous substrate binding. J. Gen. Physiol. 154, e202213131 (2022).
Article CAS PubMed PubMed Central Google Scholar
Wheeler, L. C., Lim, S. A., Marqusee, S. & Harms, M. J. The thermostability and specificity of ancient proteins. Curr. Opin. Struct. Biol. 38, 37–43 (2016).
Article CAS PubMed PubMed Central Google Scholar
Trinco, G. et al. Kinetic mechanism of Na⁺-coupled aspartate transport catalyzed by GltTk. Commun. Biol. 4, 1–11 (2021).
Article Google Scholar
Groeneveld, M. & Slotboom, D.-J. Na⁺:aspartate coupling stoichiometry in the glutamate transporter homologue GltPh. Biochemistry 49, 3511–3513 (2010).
Article CAS PubMed Google Scholar
Zerangue, N. & Kavanaugh, M. P. Flux coupling in a neuronal glutamate transporter. Nature 383, 634–637 (1996).
Article CAS PubMed Google Scholar
Guskov, A., Jensen, S., Faustino, I., Marrink, S. J. & Slotboom, D. J. Coupled binding mechanism of three sodium ions and aspartate in the glutamate transporter homologue GltTk. Nat. Commun. 7, 13420 (2016).
Article CAS PubMed PubMed Central Google Scholar
Verdon, G., Oh, S., Serio, R. N. & Boudker, O. Coupled ion binding and structural transitions along the transport cycle of glutamate transporters. eLife 3, e02283 (2014).
Article PubMed PubMed Central Google Scholar
Boudker, O., Ryan, R. M., Yernool, D., Shimamoto, K. & Gouaux, E. Coupling substrate and ion binding to extracellular gate of a sodium-dependent aspartate transporter. Nature 445, 387–393 (2007).
Article CAS PubMed Google Scholar
Oh, S. & Boudker, O. Kinetic mechanism of coupled binding in sodium–aspartate symporter GltPh. eLife 7, e37291 (2018).
Article PubMed PubMed Central Google Scholar
Reyes, N., Oh, S. & Boudker, O. Binding thermodynamics of a glutamate transporter homolog. Nat. Struct. Mol. Biol. 20, 634–640 (2013).
Article CAS PubMed PubMed Central Google Scholar
Yernool, D., Boudker, O., Jin, Y. & Gouaux, E. Structure of a glutamate transporter homologue from Pyrococcus horikoshii. Nature 431, 811–818 (2004).
Article CAS PubMed Google Scholar
Reyes, N., Ginter, C. & Boudker, O. Transport mechanism of a bacterial homologue of glutamate transporters. Nature 462, 880–885 (2009).
Article CAS PubMed PubMed Central Google Scholar
Wang, X. & Boudker, O. Large domain movements through the lipid bilayer mediate substrate release and inhibition of glutamate transporters. eLife 9, e58417 (2020).
Article CAS PubMed PubMed Central Google Scholar
Arkhipova, V., Guskov, A. & Slotboom, D. J. Structural ensemble of a glutamate transporter homologue in lipid nanodisc environment. Nat. Commun. 11, 998 (2020).
Article CAS PubMed PubMed Central Google Scholar
Crisman, T. J., Qu, S., Kanner, B. I. & Forrest, L. R. Inward-facing conformation of glutamate transporters as revealed by their inverted-topology structural repeats. Proc. Natl Acad. Sci. USA 106, 20752–20757 (2009).
Article CAS PubMed PubMed Central Google Scholar
Ruan, Y. et al. Direct visualization of glutamate transporter elevator mechanism by high-speed AFM. Proc. Natl Acad. Sci. USA 114, 1584–1588 (2017).
Article CAS PubMed PubMed Central Google Scholar
Akyuz, N., Altman, R. B., Blanchard, S. C. & Boudker, O. Transport dynamics in a glutamate transporter homologue. Nature 502, 114–118 (2013).
Article CAS PubMed PubMed Central Google Scholar
Erkens, G. B., Hänelt, I., Goudsmits, J. M. H., Slotboom, D. J. & van Oijen, A. M. Unsynchronised subunit motion in single trimeric sodium-coupled aspartate transporters. Nature 502, 119–123 (2013).
Article CAS PubMed Google Scholar
Huang, Y. et al. Use of paramagnetic ¹⁹F NMR to monitor domain movement in a glutamate transporter homolog. Nat. Chem. Biol. 16, 1006–1012 (2020).
Article CAS PubMed PubMed Central Google Scholar
Grewer, C. et al. Individual subunits of the glutamate transporter EAAC1 homotrimer function independently of each other. Biochemistry 44, 11913–11923 (2005).
Article CAS PubMed Google Scholar
Groeneveld, M. & Slotboom, D.-J. Rigidity of the subunit interfaces of the trimeric glutamate transporter GltT during translocation. J. Mol. Biol. 372, 565–570 (2007).
Article CAS PubMed Google Scholar
Yernool, D., Boudker, O., Folta-Stogniew, E. & Gouaux, E. Trimeric subunit stoichiometry of the glutamate transporters from Bacillus caldotenax and Bacillus stearothermophilus. Biochemistry 42, 12981–12988 (2003).
Article CAS PubMed Google Scholar
Raunser, S. et al. Structure and function of prokaryotic glutamate transporters from Escherichia coli and Pyrococcus horikoshii. Biochemistry 45, 12796–12805 (2006).
Article CAS PubMed Google Scholar
Canul-Tec, J. C. et al. Structure and allosteric inhibition of excitatory amino acid transporter 1. Nature 544, 446–451 (2017).
Article CAS PubMed PubMed Central Google Scholar
Gendreau, S. et al. A trimeric quaternary structure is conserved in bacterial and human glutamate transporters. J. Biol. Chem. 279, 39505–39512 (2004).
Article CAS PubMed Google Scholar
Alleva, C. et al. Na⁺-dependent gate dynamics and electrostatic attraction ensure substrate coupling in glutamate transporters. Sci. Adv. 6, eaba9854 (2020).
Article CAS PubMed PubMed Central Google Scholar
Bastug, T. et al. Position of the third Na⁺ site in the aspartate transporter GltPh and the human glutamate transporter, EAAT1. PLoS ONE 7, e33058 (2012).
Article CAS PubMed PubMed Central Google Scholar
Qiu, B. & Boudker, O. Symport and antiport mechanisms of human glutamate transporters. Nat. Commun. 14, 2579 (2023).
Article CAS PubMed PubMed Central Google Scholar
Qiu, B., Matthies, D., Fortea, E., Yu, Z. & Boudker, O. Cryo-EM structures of excitatory amino acid transporter 3 visualize coupled substrate, sodium, and proton binding and transport. Sci. Adv. 7, eabf5814 (2021).
Article CAS PubMed PubMed Central Google Scholar
Garaeva, A. A., Guskov, A., Slotboom, D. J. & Paulino, C. A one-gate elevator mechanism for the human neutral amino acid transporter ASCT2. Nat. Commun. 10, 3427 (2019).
Article PubMed PubMed Central Google Scholar
Tao, Z., Zhang, Z. & Grewer, C. Neutralization of the aspartic acid residue Asp-367, but not Asp-454, inhibits binding of Na⁺ to the glutamate-free form and cycling of the glutamate transporter EAAC1. J. Biol. Chem. 281, 10263–10272 (2006).
Article CAS PubMed Google Scholar
Riederer, E. A., Moënne-Loccoz, P. & Valiyaveetil, F. I. Distinct roles of the Na⁺ binding sites in the allosteric coupling mechanism of the glutamate transporter homolog, GltPh. Proc. Natl Acad. Sci. USA 119, e2121653119 (2022).
Article CAS PubMed PubMed Central Google Scholar
Riederer, E. A. & Valiyaveetil, F. I. Investigation of the allosteric coupling mechanism in a glutamate transporter homolog via unnatural amino acid mutagenesis. Proc. Natl Acad. Sci. USA 116, 15939–15946 (2019).
Article CAS PubMed PubMed Central Google Scholar
Zhou, W., Trinco, G., Slotboom, D. J., Forrest, L. R. & Faraldo-Gómez, J. D. On the role of a conserved methionine in the Na⁺-coupling mechanism of a neurotransmitter transporter homolog. Neurochem Res. 47, 163–175 (2022).
Article CAS PubMed Google Scholar
Starr, T. N. & Thornton, J. W. Epistasis in protein evolution. Protein Sci. 25, 1204–1218 (2016).
Article CAS PubMed PubMed Central Google Scholar
Wilson, C. et al.Using ancient protein kinases to unravel a modern cancer drug’s mechanism. Science 347, 882–886 (2015).
Article CAS PubMed PubMed Central Google Scholar
Ugalde, J. A., Chang, B. S. W. & Matz, M. V. Evolution of coral pigments recreated. Science 305, 1433–1433 (2004).
Article CAS PubMed Google Scholar
Harms, M. J. & Thornton, J. W. Historical contingency and its biophysical basis in glucocorticoid receptor evolution. Nature 512, 203–207 (2014).
Article CAS PubMed PubMed Central Google Scholar
Gaucher, E. A., Govindarajan, S. & Ganesh, O. K. Palaeotemperature trend for Precambrian life inferred from resurrected proteins. Nature 451, 704–707 (2008).
Article CAS PubMed Google Scholar
Pillai, A. S. et al. Origin of complexity in haemoglobin evolution. Nature 581, 480–485 (2020).
Article CAS PubMed PubMed Central Google Scholar
Lim, S. A., Hart, K. M., Harms, M. J. & Marqusee, S. Evolutionary trend toward kinetic stability in the folding trajectory of RNases H. Proc. Natl Acad. Sci. 113, 13045–13050 (2016).
Article CAS PubMed PubMed Central Google Scholar
Hochberg, G. K. A. & Thornton, J. W. Reconstructing ancient proteins to understand the causes of structure and function. Annu. Rev. Biophys. 46, 247–269 (2017).
Article CAS PubMed PubMed Central Google Scholar
Hoang, D. T., Chernomor, O., von Haeseler, A., Minh, B. Q. & Vinh, L. S. UFBoot2: improving the ultrafast bootstrap approximation. Mol. Biol. Evol. 35, 518–522 (2018).
Article CAS PubMed Google Scholar
Eick, G. N., Bridgham, J. T., Anderson, D. P., Harms, M. J. & Thornton, J. W. Robustness of reconstructed ancestral protein functions to statistical uncertainty. Mol. Biol. Evol. 34, 247–261 (2017).
CAS PubMed Google Scholar
Orlandi, K. N., Phillips, S. R., Sailer, Z. R., Harman, J. L. & Harms, M. J. Topiary: pruning the manual labor from ancestral sequence reconstruction. Protein Sci. 32, e4551 (2023).
Article CAS PubMed PubMed Central Google Scholar
Mulligan, C. & Mindell, J. A. Mechanism of transport modulation by an extracellular loop in an archaeal excitatory amino acid transporter (EAAT) homolog. J. Biol. Chem. 288, 35266–35276 (2013).
Article CAS PubMed PubMed Central Google Scholar
Cai, H. et al. An improved fluorescent tag and its nanobodies for membrane protein expression, stability assay, and purification. Commun. Biol. 3, 753 (2020).
Article CAS PubMed PubMed Central Google Scholar
Groeneveld, M., Weme, R. G. J. D. O., Duurkens, R. H. & Slotboom, D. J. Biochemical characterization of the C₄-dicarboxylate transporter DctA from Bacillus subtilis. J. Bacteriol. 192, 2900–2907 (2010).
Article CAS PubMed PubMed Central Google Scholar
Majd, H. et al. Screening of candidate substrates and coupling ions of transporters by thermostability shift assays. eLife 7, e38821 (2018).
Article PubMed PubMed Central Google Scholar
Nji, E., Chatzikyriakidou, Y., Landreh, M. & Drew, D. An engineered thermal-shift screen reveals specific lipid preferences of eukaryotic and prokaryotic membrane proteins. Nat. Commun. 9, 4253 (2018).
Article PubMed PubMed Central Google Scholar
Seidel, S. A. I. et al. Microscale thermophoresis quantifies biomolecular interactions under previously challenging conditions. Methods 59, 301–315 (2013).
Article CAS PubMed Google Scholar
Ewers, D., Becher, T., Machtens, J.-P., Weyand, I. & Fahlke, C. Induced fit substrate binding to an archeal glutamate transporter homologue. Proc. Natl Acad. Sci. USA 110, 12486–12491 (2013).
Article CAS PubMed PubMed Central Google Scholar
Hänelt, I., Jensen, S., Wunnicke, D. & Slotboom, D. J. Low affinity and slow Na⁺ binding precedes high affinity aspartate binding in the secondary-active transporter GltPh. J. Biol. Chem. 290, 15962–15972 (2015).
Article PubMed PubMed Central Google Scholar
Jensen, S., Guskov, A., Rempel, S., Hänelt, I. & Slotboom, D. J. Crystal structure of a substrate-free aspartate transporter. Nat. Struct. Mol. Biol. 20, 1224–1226 (2013).
Article CAS PubMed Google Scholar
Verdon, G. & Boudker, O. Crystal structure of an asymmetric trimer of a bacterial glutamate transporter homolog. Nat. Struct. Mol. Biol. 19, 355–357 (2012).
Article CAS PubMed PubMed Central Google Scholar
Bendahan, A., Armon, A., Madani, N., Kavanaugh, M. P. & Kanner, B. I. Arginine 447 plays a pivotal role in substrate interactions in a neuronal glutamate transporter. J. Biol. Chem. 275, 37436–37442 (2000).
Article CAS PubMed Google Scholar
Scopelliti, A. J., Ryan, R. M. & Vandenberg, R. J. Molecular determinants for functional differences between alanine–serine–cysteine transporter 1 and other glutamate transporter family members. J. Biol. Chem. 288, 8250–8257 (2013).
Article CAS PubMed PubMed Central Google Scholar
Scopelliti, A. J., Font, J., Vandenberg, R. J., Boudker, O. & Ryan, R. M. Structural characterisation reveals insights into substrate recognition by the glutamine transporter ASCT2/SLC1A5. Nat. Commun. 9, 38 (2018).
Article PubMed PubMed Central Google Scholar
Ciftci, D. et al. Linking function to global and local dynamics in an elevator-type transporter. Proc. Natl Acad. Sci. USA 118, e2025520118 (2021).
Article CAS PubMed PubMed Central Google Scholar
Huysmans, G. H. M., Ciftci, D., Wang, X., Blanchard, S. C. & Boudker, O. The high-energy transition state of the glutamate transporter homologue GltPh. EMBO J. 40, e105415 (2021).
Article CAS PubMed Google Scholar
Seal, R. P. & Amara, S. G. A reentrant loop domain in the glutamate carrier EAAT1 participates in substrate binding and translocation. Neuron 21, 1487–1498 (1998).
Article CAS PubMed Google Scholar
Williams, P. D., Pollock, D. D., Blackburne, B. P. & Goldstein, R. A. Assessing the accuracy of ancestral protein reconstruction methods. PLOS Comput. Biol. 2, e69 (2006).
Article PubMed PubMed Central Google Scholar
Motlagh, H. N., Wrabl, J. O., Li, J. & Hilser, V. J. The ensemble nature of allostery. Nature 508, 331–339 (2014).
Article CAS PubMed PubMed Central Google Scholar
Saavedra, H. G., Wrabl, J. O., Anderson, J. A., Li, J. & Hilser, V. J. Dynamic allostery can drive cold adaptation in enzymes. Nature 558, 324–328 (2018).
Article CAS PubMed PubMed Central Google Scholar
Sailer, Z. R. & Harms, M. J. Molecular ensembles make evolution unpredictable. Proc. Natl Acad. Sci. 114, 11938–11943 (2017).
Article CAS PubMed PubMed Central Google Scholar
Chisholm, L. O., Orlandi, K. N., Phillips, S. R., Shavlik, M. J. & Harms, M. J.Ancestral reconstruction and the evolution of protein energy landscapes. Annu. Rev. Biophys. 53, 127–146 (2024).
Article CAS PubMed PubMed Central Google Scholar
Kandasamy, P., Gyimesi, G., Kanai, Y. & Hediger, M. A. Amino acid transporters revisited: new views in health and disease. Trends Biochem. Sci. 43, 752–789 (2018).
Article CAS PubMed Google Scholar
Thorens, B. GLUT2, glucose sensing and glucose homeostasis. Diabetologia 58, 221–232 (2015).
Article CAS PubMed Google Scholar
Altschul, S. F. et al. Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res. 25, 3389–3402 (1997).
Article CAS PubMed PubMed Central Google Scholar
Johnson, M. et al. NCBI BLAST: a better web interface. Nucleic Acids Res. 36, W5–W9 (2008).
Article CAS PubMed PubMed Central Google Scholar
Fu, L., Niu, B., Zhu, Z., Wu, S. & Li, W. CD-HIT: accelerated for clustering the next-generation sequencing data. Bioinformatics 28, 3150–3152 (2012).
Article CAS PubMed PubMed Central Google Scholar
Edgar, R. C. MUSCLE5: high-accuracy alignment ensembles enable unbiased assessments of sequence homology and phylogeny. Nat. Commun. 13, 6968 (2022).
Article CAS PubMed PubMed Central Google Scholar
Hoang, D. T. et al. MPBoot: fast phylogenetic maximum parsimony tree inference and bootstrap approximation. BMC Evolut. Biol. 18, 11 (2018).
Article Google Scholar
Rozewicki, J., Li, S., Amada, K. M., Standley, D. M. & Katoh, K. MAFFT-DASH: integrated protein sequence and structural alignment. Nucleic Acids Res 47, W5–W10 (2019).
CAS PubMed PubMed Central Google Scholar
Minh, B. Q. et al. IQ-TREE 2: new models and efficient methods for phylogenetic inference in the genomic era. Mol. Biol. Evol. 37, 1530–1534 (2020).
Article CAS PubMed PubMed Central Google Scholar
Soubrier, J. et al. The influence of rate heterogeneity among sites on the time dependence of molecular rates. Mol. Biol. Evol. 29, 3345–3358 (2012).
Article CAS PubMed Google Scholar
Le, S. Q. & Gascuel, O. An improved general amino acid replacement matrix. Mol. Biol. Evol. 25, 1307–1320 (2008).
Article CAS PubMed Google Scholar
Kalyaanamoorthy, S., Minh, B. Q., Wong, T. K. F., von Haeseler, A. & Jermiin, L. S. ModelFinder: fast model selection for accurate phylogenetic estimates. Nat. Methods 14, 587–589 (2017).
Article CAS PubMed PubMed Central Google Scholar
Parks, D. H. et al. A standardized bacterial taxonomy based on genome phylogeny substantially revises the tree of life. Nat. Biotechnol. 36, 996–1004 (2018).
Article CAS PubMed Google Scholar
Yang, Z., Kumar, S. & Nei, M. A new method of inference of ancestral nucleotide and amino acid sequences. Genetics 141, 1641–1650 (1995).
Article CAS PubMed PubMed Central Google Scholar
Hanson-Smith, V., Kolaczkowski, B. & Thornton, J. W. Robustness of ancestral sequence reconstruction to phylogenetic uncertainty. Mol. Biol. Evol. 27, 1988–1999 (2010).
Article CAS PubMed PubMed Central Google Scholar
Koshi, J. M. & Goldstein, R. A. Probabilistic reconstruction of ancestral protein sequences. J. Mol. Evol. 42, 313–320 (1996).
Article CAS PubMed Google Scholar
Bianchini, G. & Sánchez-Baracaldo, P. TreeViewer: flexible, modular software to visualise and manipulate phylogenetic trees. Ecol. Evol. 14, e10873 (2024).
Article PubMed PubMed Central Google Scholar
Crooks, G. E., Hon, G., Chandonia, J.-M. & Brenner, S. E. WebLogo: a sequence logo generator. Genome Res. 14, 1188–1190 (2004).
Article CAS PubMed PubMed Central Google Scholar
Götzke, H. et al. The ALFA-tag is a highly versatile tool for nanobody-based bioscience applications. Nat. Commun. 10, 4403 (2019).
Article PubMed PubMed Central Google Scholar
Ritchie, T. K. et al. Chapter 11—reconstitution of membrane proteins in phospholipid bilayer nanodiscs. Methods Enzymol. 464, 211–231 (2009).
Article CAS PubMed PubMed Central Google Scholar
Punjani, A., Rubinstein, J. L., Fleet, D. J. & Brubaker, M. A. cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination. Nat. Methods 14, 290–296 (2017).
Article CAS PubMed Google Scholar
Bepler, T. et al. Positive-unlabeled convolutional neural networks for particle picking in cryo-electron micrographs. Nat. Methods 16, 1153–1160 (2019).
Article CAS PubMed PubMed Central Google Scholar
Punjani, A., Zhang, H. & Fleet, D. J. Non-uniform refinement: adaptive regularization improves single-particle cryo-EM reconstruction. Nat. Methods 17, 1214–1221 (2020).
Article CAS PubMed Google Scholar
Vallese, F. et al. Architecture of the human erythrocyte ankyrin-1 complex. Nat. Struct. Mol. Biol. 29, 706–718 (2022).
Article CAS PubMed PubMed Central Google Scholar
Paknejad, N., Sapuru, V. & Hite, R. K. Structural titration reveals Ca²⁺-dependent conformational landscape of the IP3 receptor. Nat. Commun. 14, 6897 (2023).
Article CAS PubMed PubMed Central Google Scholar
Asarnow, D., Palovcak, E. & Cheng, Y. asarnow/pyem: UCSF pyem v0.5. Zenodo https://doi.org/10.5281/zenodo.3576630 (2019).
Zheng, S. Q. et al. MotionCor2: anisotropic correction of beam-induced motion for improved cryo-electron microscopy. Nat. Methods 14, 331–332 (2017).
Article CAS PubMed PubMed Central Google Scholar
Zivanov, J., Nakane, T. & Scheres, S. H. W. A Bayesian approach to beam-induced motion correction in cryo-EM single-particle analysis. IUCrJ 6, 5–17 (2019).
Article CAS PubMed PubMed Central Google Scholar
Sanchez-Garcia, R. et al. DeepEMhancer: a deep learning solution for cryo-EM volume post-processing. Commun. Biol. 4, 1–8 (2021).
Article Google Scholar
Waterhouse, A. et al. SWISS-MODEL: homology modelling of protein structures and complexes. Nucleic Acids Res. 46, W296–W303 (2018).
Article CAS PubMed PubMed Central Google Scholar
Croll, T. I. ISOLDE: a physically realistic environment for model building into low-resolution electron-density maps. Acta Crystallogr. D Struct. Biol. 74, 519–530 (2018).
Article CAS PubMed PubMed Central Google Scholar
Afonine, P. V. et al. Real-space refinement in PHENIX for cryo-EM and crystallography. Acta Crystallogr. D Struct. Biol. 74, 531–544 (2018).
Article CAS PubMed PubMed Central Google Scholar
Goddard, T. D. et al. UCSF ChimeraX: meeting modern challenges in visualization and analysis. Protein Sci. 27, 14–25 (2018).
Article CAS PubMed Google Scholar
Oberg, N., Zallot, R. & Gerlt, J. A. EFI-EST, EFI-GNT, and EFI-CGFP: Enzyme Function Initiative (EFI) web resource for genomic enzymology tools. J. Mol. Biol. 435, 168018 (2023).
Article CAS PubMed PubMed Central Google Scholar

Download references

Acknowledgements

We thank the following the people for their contributions: J. Thornton and members of the Thornton lab, M. Harms, L. Skrabanek and E. Fortea, for helpful discussions regarding bioinformatics and phylogenetics; members of the O.B. lab, especially A. Scopelliti, X. Wang, B. Qiu and Y. Lopez, for helpful discussions regarding biochemistry and cryo-EM; A. Accardi and C. Nimigean for critical reading of the manuscript; W. Eng for assistance with protein expression; GenScript for prompt and patient support with gene synthesis; V. Braband, B. Milorey and D. Leyva from NanoTemper for help with troubleshooting MST parameters; A. Frazier for input on the schematic figure. The work was supported by the National Institutes of Health (NIH) grants F32 NS102325 (to K.D.R.) and R37 NS134865 (to O.B.). O.B. is an investigator of the Howard Hughes Medical Institute. Phylogenetic analyses were performed on the Weill Cornell high-performance computing cluster maintained by the Scientific Computing Unit. ChatGPT was used to write some of the Python scripts involved in bioinformatic data analysis. All cryo-EM grids were initially screened at the Weill Cornell Cryo-EM Core Facility with help from C. Fluck. Ligand-free Anc^Int was collected at the New York University cryo-EM facility with help from B. Wang. Substrate-bound Anc^Int was collected at the Columbia University Medical Center Cryo-EM core with assistance from R. Grassucci. An earlier substrate-bound Anc^Int dataset (with assistance from K. Nehelu), the ligand-free Glt_Ph dataset (with assistance from A. Owji and A. Ishemgulova) and the ligand-free Anc^Int P295₂₇₃S;Q327₃₀₅N dataset (with assistance from G. Angulli and C. Hernandez) were collected at the Simons EM Center and National Resource for Automated Molecular Microscopy located at the New York Structural Biology Center, supported by grants from the Simons Foundation (349247), NYSTAR and the NIH National Institute of General Medical Sciences (GM103310), with additional support from Agouron Institute (F00316) and NIH S10 OD019994-01.

Author information

Burha Rasool
Present address: Department of Pharmacology, Yale University, New Haven, CT, USA
Nemanja Kutlešić
Present address: Cold Spring Harbor Laboratory Cancer Center, Cold Spring Harbor Laboratory, Cold Spring Harbor, NY, USA

Authors and Affiliations

Department of Physiology and Biophysics, Weill Cornell Medical College, New York, NY, USA
Krishna D. Reddy, Burha Rasool, Farideh Badichi Akher, Nemanja Kutlešić & Olga Boudker
Department of Molecular Medicine, University of South Florida, Tampa, FL, USA
Krishna D. Reddy
Department of Biochemistry, Weill Cornell Medical College, New York, NY, USA
Swati Pant
Howard Hughes Medical Institute, Weill Cornell Medical College, New York, NY, USA
Olga Boudker

Authors

Krishna D. Reddy
View author publications
Search author on:PubMed Google Scholar
Burha Rasool
View author publications
Search author on:PubMed Google Scholar
Farideh Badichi Akher
View author publications
Search author on:PubMed Google Scholar
Nemanja Kutlešić
View author publications
Search author on:PubMed Google Scholar
Swati Pant
View author publications
Search author on:PubMed Google Scholar
Olga Boudker
View author publications
Search author on:PubMed Google Scholar

Contributions

K.D.R. and O.B. conceptualized the project and wrote the manuscript. K.D.R. performed all evolutionary analyses. K.D.R., B.R. and F.B.A. performed the functional experiments, prepared the grids and processed the cryo-EM data. K.D.R. and B.R. prepared the figures. K.D.R., B.R., F.B.A. and O.B. built and refined the molecular models. N.K. and S.P., under the supervision of K.D.R., performed the foundational pilot experiments, including expression and purification optimization, and preliminary cryo-EM data collection and processing. All authors reviewed and edited the manuscript.

Corresponding authors

Correspondence to Krishna D. Reddy or Olga Boudker.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Structural & Molecular Biology thanks Rachelle Gaudet and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available. Primary Handling Editor: Katarzyna Ciazynska, in collaboration with the Nature Structural & Molecular Biology team.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended Data Fig. 1 Elevator mechanism of SLC1-type transporters.

Glt_Ph structures in the substrate-bound outward-facing state (OFS-Asp, PDB 6UWF) and substrate-free inward-facing open state (IFS ligand-free, 6X12). The rigid scaffold domain is in transparent wheat; the mobile transport domain is in blue; HP1 is in yellow; HP2 is in red; purple spheres are Na⁺; green molecule is L-aspartate. In a typical transport cycle, three Na⁺ ions and L-aspartate bind in the outward-facing state, resulting in the closure of the HP2 gate and translocation of the transport domain to the inward-facing state. The HP2 gate then opens, releasing solutes to the intracellular side.

Extended Data Fig. 2 APR workflow to capture Ext^Na to Ext^H evolutionary transition.

The diagram is based on Orlandi et al. (1) Based on previous biochemical and structural characterization of prokaryotic H⁺-coupled (orange; GltP_Ec, GltT_Bc, DctA_Bs) and Na⁺-coupled (purple; Glt_Ph, Glt_Tk) transporters, we identified signature residues that helped us sort putative sequences as either Ext^H or Ext^Na. (2) We retrieved homologs of transporters using PSI-BLAST. After clustering and quick sequence alignment, we removed sequences with large gaps or insertions. Finally, we used parsimony tree generation to keep sequences that encompass Ext^H, Ext^Na, and intermediate (green) transporters. (3) We carefully aligned the sequences using MAFFT-DASH, which incorporates structural information. Sequences with large gaps or insertions were removed. The remaining sequences were re-aligned, and poorly aligned loops and tails were stripped from the alignment. (4) The final sequence alignment was used as input for ML tree generation, using archaeal Glt_Ph as the root. The optimal model of protein evolution was selected using ModelFinder. Sequences that caused long branches were removed, and 254 parallel tree searches were performed to find the optimal branch lengths and topology, resulting in the ML tree with the highest possible likelihood. (5) To reconstruct ancestral sequences, every node was reconstructed using optimized ML parameters and protein evolution model; each site in each node is assigned probabilities for all 20 amino acids, and the highest-probability amino acid at a site is selected for the ML sequence. (6) To determine if a given node is creating a robustly reconstructed ancestor, bootstrap is performed (UFBoot2), which calculates branch supports by resampling the MSA 1000 times and determining the frequency of the branch re-occurring in the trees. Sequence sampling is critical in this step; when we retrieved sequences from Firmicutes/Bacillota, we obtained robust UFBoot2 values for Anc^Int and descendants, but not when retrieving sequences from Proteobacteria. (7) Ancestral sequences are engineered for expression and purification by inserting loops and tails from extant sequences (green), which were not considered during APR due to poor alignment. All relevant references here and elsewhere in the legends are in the Main Text or Methods. Created in BioRender. Boudker, O. (2025) https://BioRender.com/sl3la84.

Extended Data Fig. 3 Phylogeny, rooting, and conservation of prokaryotic glutamate transporter homologs.

(a) Maximum parsimony phylogeny of bacterial glutamate transporter homologs. The clade containing conserved signature residues in H⁺-coupled transporters is colored orange. (b) Selection of initial sequences for APR analysis from Firmicutes/Bacillota (99% clustered). Purple, green, and orange clades contain representative Ext^Na, Ext^Int, and Ext^H sequences, respectively. Gray sequences were unassigned in our analysis, and were eventually removed due to potential long-branch attraction artifacts. (c) Conservation of Na⁺- and H⁺-coupling signature residues in non-redundant archaeal glutamate transporters. (d) Conservation of the Na⁺-binding sites throughout the ML phylogeny. Top: sequences of Na⁺-coupled Glt_Ph and H⁺-coupled GltT_Bc encompassing Na⁺-binding sites of Ext^Na. The numbers above the sequences mark side chains in Glt_Ph that coordinate Na⁺ sites 1–3; A indicates side chains coordinating L-aspartate. Bottom: WebLogos of the relative amino acid frequencies in Ext^Na, Ext^Int, and Ext^H clades. The number of sequences per clade is 186, 161, and 786, respectively.

Extended Data Fig. 4 Robustness of the ancestral protein reconstruction.

(a) Reconstructed ancestral sequences in Na⁺-binding sites of Anc^Na, Anc^Int, and Anc^H. WebLogos visually represent the posterior probability per site in each ancestral sequence. The topmost residue at each site is the ML residue. Glt_Ph and GltT_Bc sequence alignments are shown above the WebLogos for reference, along with annotations for L-aspartate binding residues (A) or Na⁺-binding residues^1,2,3. (b) Average posterior probability (PP) per site, the natural logarithm of the total posterior probability of the entire ancestral sequence (Ln PP), and the fraction of the ambiguous sites (defined as two or more amino acids per site with a probability greater than 0.2) in each ancestor. Bold, larger values are for the entire sequence; values in parentheses are for the transport domain alone. (c) Pairwise identity (in %) matrix of ancestral and extant transporters. The alignment included only transport domains, defined as Glt_Ph residues 78–110, 228–416. (d) Visual representation of the changes between two ML ancestral sequences. In each bar, the first segment is the number of changes where the posterior probability is greater than 0.9 in only the first ancestor, that is, a robustly reconstructed site is ambiguously reconstructed in the second ancestor; the second segment is the number of changes where the posterior probability is greater than 0.9 in only the second ancestor, while the site is ambiguous in the first ancestor; the third segment is the number of changes where the posterior probability is greater than 0.9 in both ancestors with both sites confidently reconstructed. (e, f) Probable sequence changes between Anc^Na and Anc^Int (e) and Anc^Int and Anc^H (f), mapped onto the transport domain of Glt_Ph. Bottom: Labeling scheme for residues. Larger spheres represent sequence transitions; smaller spheres represent Na⁺ sites and structured waters observed in Glt_Ph (PDB 7RCP). ‘Unknown’ indicates when there is no clear physicochemical difference between the two residues. Ext^H signature residues are [T/S]92[A/L/I/V/M], M311L, and D405N; putative protonation site is S279E.

Extended Data Fig. 5 Sequence alignment of extant and ancestral glutamate transporter homologs.

(a) Alignment sequence numbering corresponds to Glt_Ph numbering. Extant sequences are Glt_Ph (UniProt O59010), GltT_Bc (P24944), and human EAAT3 (P43005). TMs, shown above the alignment, are colored wheat for the scaffold/trimerization domain, blue for the transport domain, yellow for re-entrant helical hairpin HP1, and red for re-entrant helical hairpin HP2. Dashed green lines represent residues grafted from an extant sequence into ancestral protein constructs in place of poorly aligned regions (see Methods). Filled purple circles mark residues in Glt_Ph that coordinate Na⁺ ions with their side chains, filled green squares mark residues in Glt_Ph that coordinate L-aspartate with their side chains, and empty circles represent allosteric mutations that can turn Na⁺-coupling on and off. Notable sequence changes in H⁺-coupled transporters compared to Na⁺-coupled transporters are shown in bold, and red residues are changes found at Na⁺-binding sites. (b) Structure of a Glt_Ph protomer in the outward-facing state (PDB 6UWF), colored as in (a). Purple spheres are Na⁺ ions (placed based on PDB 6×15), and the substrate is green.

Extended Data Fig. 6 Validation statistics for all structures.

Map validations of (a) ligand-free Anc^Int, low-affinity; (b) ligand-free Anc^Int, high-affinity; (c) L-aspartate-bound Anc^Int; (d) ligand-free Glt_Ph, low-affinity, iOFS; (e) ligand-free Glt_Ph, low-affinity, OFS. From left to right, map FSC from local refinement in cryoSPARC and local resolution estimation of the unsharpened map. All maps are contoured at 8 σ. The adjacent protomers (unused in refinement) are removed for clarity.

Extended Data Fig. 7 Global transport domain conformations of structures obtained in this study.

Anc^Int (green) or Glt_Ph (purple) structures were aligned on the scaffold domain (residues 1–75, 140–225) of the closest Glt_Ph structure in the PDB (translucent white). These Glt_Ph structures were either OFS (2NWX) or iOFS (6UWL). IFS states in Anc^Int were incompletely resolved, and models could not be built; therefore, the maps from 3D classification were contoured to a σ of 6 and aligned to the closest Glt_Ph structure, IFS-open (white, 6 × 12).

Extended Data Fig. 8 Alternate substrate binding configurations and AncInt HP2 densities.

(a) Substrate-binding sites of ligand-free iOFS Glt_Ph in the low-affinity configuration. (b) Substrate-binding site of ligand-free Anc^Int in the high-affinity configuration #2. The green density overlaps with the L-aspartate binding site but is too large and differently shaped to correspond to L-aspartate. (c–e) HP2 densities of ligand-free Anc^Int, low-affinity (c), ligand-free Anc^Int, high-affinity (d), and L-aspartate Anc^Int, high-affinity (e). Densities are unsharpened and contoured 10 σ. HP1 is in yellow, and HP2 is in red.

Extended Data Fig. 9 Functional characterization of mutant ancestral proteins.

(a) nanoDSF assays of Anc*, Anc^Na S295₂₇₃P/N327₃₀₅Q, and single Anc^Int mutations. Melting temperatures (T_m) in buffer alone (ligand-free), Na⁺, L-aspartate, and Na⁺/L-aspartate. Replicates are from at least two independent experiments, each with several technical replicates. All ligand concentrations are 10 mM Na⁺ and 1 mM L-aspartate, unless otherwise indicated. (b) Alignment of all 20 classes of Anc^Int P295₂₇₃S/Q327₃₀₅N protomers (obtained by 3D classification without alignment, additional detail in Methods) to the scaffold domains of either Anc^Int low-affinity or high-affinity states.

Source data

Extended Data Fig. 10 Genomic pairing of uncoupled transporters and enzymes.

Transporters were considered to have a paired enzyme if the enzyme was within 3 protein-coding genes away from the transporter and if the enzyme had putative free amino acid/dicarboxylate modifying function according to Pfam annotation. The fraction of sequences with genomic transporter/enzyme pairs is on the top of each bar. Only transporters with available genomic neighborhood information were considered; for the Ext^Na, Ext^Int, and Ext^H clades, this corresponds to 29/186, 38/161, and 170/786, respectively.

Supplementary information

Supplementary Information

Supplementary Figs. 1–8, Table 1 and Source data for Supplementary Figs. 1 and 3.

Reporting Summary

Peer Review File

Supplementary Data 1

Uncropped Coomassie gels and Supplementary Figs. 1 and 3.

Supplementary Data 2

All input and output data related to phylogenetic analyses, plasmid maps of all protein expression constructs used in this study, a ChimeraX session file containing comparison of Anc^Int models to Anc^Int P295₂₇₃S;Q327₃₀₅N maps (Extended Data Figure 9), a summary table of all structural maps and models and PDB validation reports.

Source data

Source Data Fig. 2

Statistical source data.

Source Data Fig. 2

Unprocessed gels.

Source Data Fig. 3

Statistical source data.

Source Data Fig. 5

Statistical source data.

Source Data Extended Data Fig. 9

Statistical source data.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Reddy, K.D., Rasool, B., Akher, F.B. et al. Evolutionary analysis reveals the origin of sodium coupling in glutamate transporters. Nat Struct Mol Biol 32, 2564–2574 (2025). https://doi.org/10.1038/s41594-025-01652-z

Download citation

Received: 28 June 2024
Accepted: 11 July 2025
Published: 25 August 2025
Version of record: 25 August 2025
Issue date: December 2025
DOI: https://doi.org/10.1038/s41594-025-01652-z

This article is cited by

A structural window into the evolution of secondary transport mechanisms
- Samuel P. Berry
- Rachelle Gaudet
Nature Structural & Molecular Biology (2025)

Subjects

Abstract

Similar content being viewed by others

Main

Results

An evolutionary intermediate between Na+-coupled and H+-coupled transporters

The ancestral intermediate transporter is uncoupled from Na+

AncInt accesses high-affinity states without Na+

Binding sites are unaltered in uncoupled transporters

Allosteric mutations turn ion coupling back on

Discussion

Methods

Sequence retrieval, curation and alignment

Phylogenetic analysis, ancestral protein reconstruction and construct design

Protein expression and purification

Reconstitution into proteoliposomes

Proteoliposome uptake experiments

Proteoliposome counterflow experiments (sodium-dependent)

Proteoliposome counterflow experiments (pH-dependent)

nanoDSF

MST

Cryo-EM sample preparation

Cryo-EM data acquisition and processing

Initial processing pipeline for AncInt

AncInt ligand-free dataset processing

AncInt substrate-bound processing

GltPh ligand-free processing

AncInt P295273S;Q327305N ligand-free dataset processing

Model building and refinement

Gene neighborhood analysis

Reporting summary

Data availability

References

Acknowledgements

Author information

Authors and Affiliations

Contributions

Corresponding authors

Ethics declarations

Competing interests

Peer review

Peer review information

Additional information

Extended data

Supplementary information

Source data

Rights and permissions

About this article

Cite this article

Share this article

This article is cited by

Search

Quick links

An evolutionary intermediate between Na⁺-coupled and H⁺-coupled transporters

The ancestral intermediate transporter is uncoupled from Na⁺

Anc^Int accesses high-affinity states without Na⁺

Initial processing pipeline for Anc^Int

Anc^Int ligand-free dataset processing

Anc^Int substrate-bound processing

Glt_Ph ligand-free processing

Anc^Int P295₂₇₃S;Q327₃₀₅N ligand-free dataset processing