The only structurally characterized DASS transporter, VcINDY, has exclusively been observed in a C_i-Na⁺-S state (Figure 1—figure supplement 1e; Mancusso et al., 2012; Nie et al., 2017). To obtain the structure of a DASS protein in a substrate-free, Na⁺-bound state, we purified VcINDY in 100 mM Na⁺, but in the absence of a substrate, for single-particle cryo-electron microscopy (cryo-EM). The total mass of a VcINDY dimer is only 96 kDa, and almost that entire mass is embedded in the membrane. This posed a challenge for single-particle analysis, and therefore we used two separate strategies for cryo-EM sample preparation. The first was to increase the effective particle mass using a synthetic antibody fragment (Fab). Using VcINDY reconstituted in lipid nanodiscs, we screened for Fabs using phage display technology (Figure 1—figure supplement 2a; Fellouse et al., 2007; Miller et al., 2012). In this way we identified Fab84, which was subsequently overexpressed, purified, and mixed with VcINDY in nanodiscs (Figure 1—figure supplement 2b and c). The VcINDY-Fab84 structure was determined to 3.15 Å resolution (Figure 1b, Figure 1—figure supplement 2d,e and h, and Table 1). In a parallel approach, we used amphipol polymer (Huynh et al., 2018) to preserve the transporter protein (Figure 1c, Figure 1—figure supplement 2f,g & k, and Table 1), which allowed data collection on particularly thin ice to maximize signal-to-noise. This VcINDY structure in amphipol, without Fab, was determined to 3.16 Å resolution. Both maps are at a sufficiently high resolution and quality to allow direct model building, and the structures are nearly identical (r.m.s.d. 0.750 Å). Finally, to obtain clear density of substrate in the binding site of VcINDY, we crystallized the protein in complex with sodium and terephthalate, and solved the X-ray structure to 3.92 Å resolution by molecular replacement (Figure 1d and Table 2). With two carboxylate moieties at approximately the same distance as those of native substrates, we hypothesized the terephthalate would bind within the VcINDY binding site, while its large benzene ring would provide stronger electron density than succinate, fumarate, or malate.

To obtain the structure of a DASS protein in its C_o conformation, we screened various VcINDY homologs and chose a DASS-E protein from Lactobacillus acidophilus (LaINDY) (Figure 1—figure supplement 1c). To characterize the function of LaINDY we performed in vivo complementation and transport assays. LaINDY was able to support aerobic growth on αKG as the sole carbon source for an E. coli strain in which the only aerobically expressed αKG transporter was knocked out (Baba et al., 2006; Figure 2—figure supplement 1a). E. coli transformed with LaINDY accumulated radioactive succinate from an external solution, likely in exchange for endogenous internal dicarboxylates (Figure 2—figure supplement 1b). Subsequent addition of excess external succinate or αKG reduced the amount of internalized radioactive succinate due to LaINDY-facilitated dicarboxylate exchange (Figure 2a).

Like other DASS proteins, purified LaINDY formed a dimer in detergent solution (Figure 2—figure supplement 1d & e). Observing that the VcINDY structures were nearly identical in amphipol and nanodiscs, we surmised that amphipol was suitable for structure determination of another DASS protein, LaINDY. The LaINDY map obtained in this way was at 3.09 Å resolution, allowing for direct model building (Figure 2b, Figure 2—figure supplement 1g–i, and Table 1). Notably, the LaINDY structure was in an outward-facing C_o state.

Next, we attempted to determine the structure of LaINDY in its substrate-bound state. To identify substrates suitable for structure determination of such a complex, we examined the thermostability and monodispersity of LaINDY at elevated temperatures in the presence of various potential substrates and substrate analogs (Figure 2—figure supplement 1f). Following identification of malate and αKG as stabilizers, we determined cryo-EM maps of LaINDY in complex with each to 3.36 Å and 3.71 Å, respectively (Figure 2c & d, Figure 2—figure supplement 1j & k, and Table 1). We also solved a 2.85 Å X-ray structure of LaINDY in the presence of both malate and αKG (Figure 2e and Table 2), using the LaINDY-αKG cryo-EM structure as the search model for molecular replacement. As with the apo LaINDY structure, all three LaINDY complex structures were in an outward-facing C_o-S state.

The LaINDY and VcINDY maps were all of sufficient quality to build side chains (Figure 2—figure supplement 2a–g), though side-chain rotamers were not always clear at these resolutions. However, we recognized that the significance of comparing structures would be dependent upon the models’ accuracy. This accuracy is a concern particularly as two structural methods and multiple instruments were used. We therefore used the models’ C_α to C bond lengths as an internal benchmark, as we expect this bond to be insensitive to variations in amino acid sequence, secondary structure, or local environment. The C_α – C bond lengths in our cryo-EM structures, as well as those in the PDB, were typically smaller than the ideal and those determined by X-ray crystallography (Figure 2—figure supplement 2h–j). This discrepancy was systematic and independent of map resolution, suggesting it is not the result of microscope calibration errors. Still, the cryo-EM models’ accuracy are sufficient to allow us to interpret the observed side chain movements of 1–2 Å.

Finally, to characterize the C_o-S to C_i-S transition of a DASS protein, we used a custom biased MD simulation protocol (Figure 6—figure supplement 1a and Supplementary Note) to drive LaINDY to a C_i-S target model suggested by LaINDY’s internal inverted repeat topological symmetry (Video 1 and Supplementary Note). We bookended this induced transition with unbiased MD simulations of LaINDY’s C_o-S and C_i-S states (Video 2), which enabled us to characterize LaINDY’s equilibrium structural dynamics.

Along with the previous VcINDY structures (Mancusso et al., 2012; Nie et al., 2017), the newly determined structures of the C_o and C_i conformations in apo and substrate bound states, along with the MD simulations of the C_o to C_i transition, allow us to examine the reaction cycle of DASS transporters. We will begin by describing the outward-facing apo state of LaINDY, and subsequently characterize the structural changes associated with substrate binding to the C_o state, the C_o to C_i transition that carries substrate across the membrane and, finally, substrate release into the cytosol (Figure 1a, and Figure 1—figure supplement 1a and b).

The 3.09 Å cryo-EM map of LaINDY determined in the absence of substrate shows the transporter in its C_o apo state (Figure 3a and b). In agreement with LaINDY’s apparent mass in detergent solution (Figure 2—figure supplement 1e), the map shows a transporter dimer. Each protomer consists of a scaffold domain and a transport domain. The transmembrane topology and domain organization of LaINDY resemble that of VcINDY (Mancusso et al., 2012), with the scaffold domain being formed by transmembrane α-helices TMs 1–4 and 7–9, while the transport domain consists of TMs 5, 6, 10 and 11, as well as the helix hairpins HP_in and HP_out (Figure 3a-c). However, for both hairpins, a bend is found in the second helix, at Val163 in HP_inb and Ala404 in HP_outb. Another new structural feature of LaINDY is an extrusion near the dimer interface into the periplasmic space formed by a sequence insertion between TM3 and TM4.

The scaffold and transport domains are linked via two horizontal helices and two loops near the membrane surface, H4c and L4-HP_in on the cytosolic side and H9c and L9-HP_out on the extracellular side (Figures 3c and 4a). A third linker is formed by a cytoplasmic helix between TM6 and TM7 (H6b), where the equivalent region in VcINDY exists as a long loop (Figure 4b). The domain interface is largely formed by branched or short hydrophobic residues, with only two hydrogen bonds. This results in a smooth domain interface, similar to that observed in the elevator transporter Glt_Ph (Reyes et al., 2009).

While LaINDY’s topology is similar to that of VcINDY, the relative domain positions are different. The transport domain is oriented toward the extracellular side with its substrate binding site facing the periplasm, yielding a C_o conformation (Figure 3a). The individual LaINDY and VcINDY domains exhibit strong structural homology, with backbone r.m.s.d.s of 2.897 Å and 2.044 Å for the scaffold and transport domains, respectively (Figure 3d and e). Compared with VcINDY, the transport domain in LaINDY is repositioned by 13.0 Å towards the periplasm with a 37.4° rotation (Figure 4). The domains’ structural conservation and relative positions agree with the notion that DASS proteins operate via a rigid-body, elevator-type movement of the transport domain (Figure 1a).

The structures of LaINDY determined in malate, αKG and in the malate/αKG mixture are in an outward-facing, substrate-bound (C_o-S) state (Figure 2e and Figure 2—figure supplement 1j & k). The architecture of the LaINDY binding site is similar to that of VcINDY. Notably, within the binding site, density corresponding to substrate is seen in the X-ray omit map of LaINDY-malate-αKG, and in the LaINDY-αKG cryo-EM difference map, in a similar mode to substrate binding in VcINDY (Figure 5—figure supplement 1e & f). All three C_o-S structures are not only similar to each other but, importantly, also to the LaINDY C_o apo structure, with pairwise r.m.s.d.s of 0.41–0.52 Å (Figure 3f), indicating that substrate binding introduces little conformational change.

In contrast to the similarity at the substrate binding site, major differences between LaINDY and VcINDY are seen at the cation binding sites. At the Na1 site of VcINDY (Figure 5—figure supplement 1f), an arginine (Arg159) is found in LaINDY, stabilized by a salt bridge with Glu146 (Figure 5a and Figure 5—figure supplement 1a & c). Similarly, a histidine (His392) in LaINDY is located at the equivalent of the VcINDY Na2 site, with another histidine (His401) located 4 Å on its extracellular side. Indeed, these residues are conserved in DASS exchangers but absent in cotransporters (Figure 1—figure supplement 1c). Furthermore, the cation binding sites are more completely enclosed by the surrounding loops in LaINDY compared to VcINDY, facilitated by DASS exchanger specific insertions in the HP_in and HP_out and L10ab regions (Figure 1—figure supplement 1d). Based on these well enclosed and sterically occupied cation binding sites, it is unlikely that sodium ions bind to LaINDY as Na1 and Na2 do in VcINDY. Rather, we hypothesize these two positively-charged residues in exchangers act as permanent surrogates of the Na⁺ ions in cotransporters. Such Na⁺ ion substitutions have previously been observed in other transporters, which makes their substrate transport independent of sodium (Shaffer et al., 2009; Kalayil et al., 2013).

The structure determination of LaINDY and VcINDY in multiple states provides an opportunity to characterize conformational changes of a DASS protein during the C_o-S to C_i-S transition. MD simulations of succinate-bound LaINDY revealed how the transition between its C_o-S and C_i-S states is realized. The transition consists of a 39° rigid-body rotation and an 8.3 Å translation of the transport domain relative to the scaffold domain (Figure 6—figure supplement 1b–d and Video 2). The two ‘arm’ helices, H4c and H9c, are fixed in space with respect to the scaffold domain and cradle the transport domain during the C_o-S to C_i-S transition (Figure 6c–e and Video 3). Such arm rigidity agrees with the conserved salt bridge, between Arg122 and Glu283, and bulky-residue interactions at the elbows connecting H4c and H9c to the scaffold domain (Figure 6a & b). The importance of the Arg122 to Glu283 salt bridge is also consistent with recent observations in NaCT, where transport activity was abolished by mutations of the equivalent arginine (Khamaysi et al., 2020), probably by disrupting the conserved salt bridge between H4c and TM7. In contrast to the rigidity at the elbows, flexibility of the hinge loops L4-HP_in and L9-HP_out at the other end of the arm helices allows transport domain movement (Figure 4). During the C_o-S to C_i-S transition the angle between the arm helices and the hairpin helices changes by approximately 30° at both hinges (Figure 6f), allowing the overall rotation and translation of the transport domain.

Despite the large domain movements of LaINDY during MD simulations of the C_o-S to C_i-S transition, succinate stably binds without significant changes within the binding site (Figure 5b), though the depth of substrate binding correlates with the orientation of Asn156 (Figure 6—figure supplement 1g,h & l). Furthermore, the conformations and interactions of the Na⁺ surrogate side chains, Arg159 and His392, are generally stable through the conformational transition (Figure 5—figure supplement 1b & d).

The two cryo-EM structures of VcINDY are in a substrate-free, inward-facing (C_i-Na⁺) state (Figure 1b & c) and provide an opportunity to describe the conformational changes that occur upon substrate release. The C_i-Na⁺ VcINDY structures are very similar to each other, with an r.m.s.d of 0.75 Å (Figure 7—figure supplement 1b). However, compared with the C_i-Na⁺-S structures, the substrate-free structures display local changes throughout the protein.

At the substrate binding site of both cryo-EM VcINDY structures, Pro422 at the N-terminus of TM10b in the substrate free structures moved by up to 1.5 Å (Figure 7a). Concurrently, on the cytoplasmic surface, His432 at the C-terminus of the same helix is rotated by 73° (Figure 7—figure supplement 1a). This movement of His432 causes a steric clash with its neighbor, Tyr178 of TM5a, inducing a rotation in that side chain by 41°. One helix turn away, the side chain of Arg175 moves closer to Glu437 of TM11, forming a salt bridge. These rearrangements lead to movements of HP_ina, HP_inb and TM5a by 1.4 Å toward the scaffold domain. Supporting the importance of these interactions, human NaCT’s transport activity is abolished when the equivalent of Glu437 is mutated to histidine, resulting in SLC13A5 Deficiency (Hardies et al., 2015).

Comparison of the previous C_i-Na⁺-S (Mancusso et al., 2012; Nie et al., 2017) and amphipol preserved C_i-Na⁺ VcINDY structures also revealed prominent changes on the periplasmic surface (Figure 7b and Video 4). The loop connecting HP_outb and TM10a, from Ala395 to Pro400, has moved on the periplasmic surface. The C-terminus of HP_outb also unwinds by one turn (Val392 – Glu394) in the amphipol structure. As a result, the conserved salt bridge between Glu394 with Lys337 of H9b from the scaffold domain breaks, and Phe396 moves away from its contact with H9c. Such structural changes agree with observations of human NaCT mutations, at positions equivalent to Pro400 and Val401 in VcINDY, which abolish transport and cause SLC13A5 Deficiency (Thevenon et al., 2014; Klotz et al., 2016). Finally, the side chain of Trp461 is inserted between TM6, TM10a and TM11, packing against another conserved aromatic residue, Phe220. Notably, when bound to Fab84, the loop connecting HP_outb and TM10a in VcINDY is midway between the C_i-Na⁺-S and C_i-Na⁺ structures. This agrees with Fab84’s epitope covering the periplasmic surface of VcINDY and binding both apo and substrate-bound states (Figure 7—figure supplement 1c & d). Also, a complete lipid molecule from the periplasmic leaflet was found at the interface between the transport and scaffold domains, interacting with TM1, TM2 and HP_outa, and therefore may be involved in regulating conformational changes (Figure 7—figure supplement 1e).

As a target seed VcINDY was nominated to the cloning core of the New York Consortium of Membrane Protein Structure for the cloning of its homologs (Love et al., 2010). The homologous protein from Lactobacillus acidophilus (Uniprot: Q5FKK5_LACAC, LaINDY) was found to give the highest expression levels in E. coli BL21 DE3 cells when transformed with pET-LaINDY. The transport activity of LaINDY was characterized in E. coli whole cells following a published protocol with minor modifications (Pos et al., 1998; Kim and Unden, 2007). E. coli BL21 DE3 cells were transformed with a modified pET vector (Love et al., 2010) encoding N-terminal 10x His-tagged LaINDY (pET-LaINDY) and grown at 37°C until cells reached OD₅₉₅ of 0.7. Cells were induced with IPTG and growth was continued for 3 hr. Cells were harvested by centrifugation and resuspended at OD₅₉₅ = 10 in 10 mM NaCl, 100 mM choline chloride, 50 mM Tris pH 7.5. Cells were kept on ice until needed, and warmed to 30°C five mins prior to transport assay. The uptake reaction was initiated by addition of reaction buffer 10 mM NaCl, 100 mM choline chloride, 50 mM Tris pH 7.5, 1 μM ³H-succinate to the cell suspension at 1:10 (reaction buffer:cells) volumetric ratio. For the exchange reaction, 10 mM succinate or 10 mM α-ketoglutarate (αKG) was added 90 s after initiation of the uptake reaction. Aliquots were collected at fixed time points, with the reaction terminated by collecting the cells on pre-wetted 0.45 mm nitrocellulose filters mounted on a Hoeffer vacuum manifold and immediately washing with 4 mL ice cold 100 mM potassium phosphate buffer pH 7.5. The filters were incubated for 10 mins in scintillation fluid before measuring radioactivity using a Wallac 1450 Microbeta Plus liquid scintillation counter (Shelton, CT) (Law et al., 2007; Law et al., 2008; Law et al., 2009).

The E. coli strain JW2571 (Baba et al., 2006), in which the only endogenous αKG transporter under aerobic conditions KgtP (Seol and Shatkin, 1991) was knocked out, was transformed with a pGEM-5Zf(+) vector encoding LaINDY. Transformed cells were grown in LB broth with ampicillin to an OD₅₉₅ ~0.8, then diluted to an OD₅₉₅ of 0.1. Dilute cells were inoculated into eM9 media supplemented with 50 mM αKG (Rhie et al., 2014), and cell growth monitored using a Tecan SPECTRAFluor Plus microplate reader (Männedorf, Switzerland) incubated at 37°C.

LaINDY was expressed by autoinduction (Studier and Moffatt, 1986) at 25°C overnight in the E. coli strain BL21 DE3 transformed with the pET-LaINDY plasmid. Cells were harvested and lysed in a buffer of 50 mM Tris pH 8.0, 400 mM NaCl, 10 mM Imidazole. Ligand, either 10 mM Na⁺ αKG or 10 mM Na⁺ DL-malate, was added to the lysis buffer and all subsequent purification steps. Membranes were resuspended in a buffer of 50 mM Tris pH 8.0, 200 mM NaCl, 10 mM Imidazole, and solubilized in 1.2% dodecyl-maltoside (DDM), and protein was purified on a Ni²⁺-NTA affinity column. The decahistidine tag was removed by overnight digestion at 25°C with TEV protease, followed by preparative size exclusion chromatography (SEC) in a buffer containing 25 mM Tris pH 8.0, 150 mM NaCl, 20% glycerol, and 0.075% DDM.

A theremostability assay was used to search for compounds that stabilized LaINDY (Auer et al., 2001; Mancusso et al., 2011). Size exclusion chromatography purified LaINDY was dialyzed into a buffer of 20 mM Tris pH 8.0, 150 mM NaCl, 10% glycerol, 0.15% DM. Aliquots of 100 mg LaINDY were incubated with 100 mM test compounds at 42°C for 2 hr, and subsequently injected onto a Shodex KW804 analytical SEC column (Thomson, Clear Brook, VA) on HPLC (Shimadzu, Columbia, MA) in a buffer containing 200 mM Na₂SO₄, 50 mM Tris 7.5, 3 mM NaN₃, and 0.05% DDM. The height of the SEC peak for the 4°C control was used as a reference for normalization.

Purified LaINDY sample (50 μL) was injected onto a Shodex KW803 analytical SEC column on a Waters HPLC (Milford, MA) and eluted with the buffer containing 0.05% DDM at a rate of 0.5 mL/min. The mass of the LaINDY protein was determined using a Wyatt miniDAWN TREOS three angle-static light scattering detector (Santa Barbara, CA), a Wyatt Optilab rEX refractive index detector and a Waters 2489 UV absorbance detector (Slotboom et al., 2008; Waight et al., 2010). The differential refractive index (dn/dc) for DDM, 0.128 mL/g, was calculated using the refractive index detector. The size of the protein–detergent conjugate was deconvoluted following the published method (Kendrick et al., 2001), in which contributions from co-purifying lipids were not distinguished from those of the detergent.

Expression and purification of VcINDY was carried out according to our previous protocol (Mancusso et al., 2012). Briefly, E. coli BL21-AI cells (Invitrogen) were transformed with a modified pET vector (Love et al., 2010) encoding N-terminal 10x His tagged VcINDY. Cells were grown at 30°C until OD₅₉₅ reached 0.8, protein expression occurred at 19°C following IPTG induction, and cells were harvested 16 hr post-induction. Cell membranes were solubilized in 1.2% DDM and the protein was purified on a Ni²⁺-NTA column. VcINDY was further purified by SEC in buffer containing 25 mM Tris pH 8, 100 mM NaCl, 10 mM Na⁺-succinate, 5% glycerol and 0.075% DDM unless otherwise indicated.

SEC purified His-tagged VcINDY protein was reconstituted into MSP2N2 nanodiscs (Grinkova et al., 2010) at a molar ratio of VcINDY: MSP: lipids of 1 (9 mM): 8 (72 mM): 277 (2.5 mM) (Bayburt and Sligar, 2010; Schuler et al., 2013). E coli polar lipids in chloroform (Avanti) were vacuum-dried and rehydrated in nanodisc buffer containing 20 mM Tris pH 7.4, 100 mM NaCl, 0.5 mM EDTA and 0.1 mM TCEP to a concentration of 20 mg/mL. VcINDY, lipids and MSP2N2 protein were mixed in the nanodisc buffer to a final volume of 500 mL. Subsequently, 600 mg of Bio-Beads were added and incubated at 4° C overnight. The Bio-Beads were then removed and the solution was filtered through a 0.2 mm centrifugal filter. The nanodiscs containing VcINDY were purified by Ni²⁺-NTA chromatography. The sample was incubated with a 1:4 molar ratio of Fab84 for 1 hr at 4°C before SEC.

The plasmid pFab101, a modified pFab007 (Miller et al., 2012) which includes a N31T mutation in the VH region, was used as the template to construct a Fab phage-display library termed NYC1 containing 1.1 × 10¹¹ sequences, following a previously described design (Miller et al., 2012). Fab library sorting was performed as previously described (Fellouse et al., 2007; Dominik and Kossiakoff, 2015) with minor modifications. In each round, phage solution was prepared in 50 mM Tris HCl buffer pH 7.5 containing 100 mM NaCl, 1% bovine serum albumin and 0.1 mM TCEP with or without 2 mM succinate, and was first incubated with streptavidin-coated magnetic beads harboring biotinylated nanodisc that did not contain an embedded protein (‘empty’ nanodisc). The supernatant of this reaction was incubated with 100 nM VcINDY embedded in biotinylated nanodisc, and phages bound to the VcINDY-nanodisc complex were captured using streptavidin-coated magnetic beads. A total of four rounds of library sorting were performed. Enriched clones were individually tested using phage ELISA (Fellouse et al., 2007; Sidhu et al., 2000).

Target Fab84 was subcloned into the Fab expression vector Ptac_Fab_accept_tagless (Burioni et al., 1998). After rigidification (Bailey et al., 2018), the subsequent Fab-containing plasmid was transformed into E. coli 55244 cells. Cells were grown in TGB media at 30°C for 22 hr and harvested. Fab84 protein was purified by injection onto a Protein G column (GE healthcare) and eluted with 100 mM glycine-pH 2.7. Protein was collected in 1 mL fractions containing 2 M Tris pH 8. Fractions containing protein were dialyzed against 50 mM sodium acetate pH 5.3 before purification using a Resource S column (GE healthcare).

After nickel column purification PMAL-C8 (Anatrace, Maumee, OH) was added to the detergent-purified transporter protein at a 1:5 protein:amphipol weight ratio (Huynh et al., 2018). The mixture was incubated at 4° C overnight with nutating. To remove detergent, Bio-Beads were incubated with sample at a 1:100 detergent:Bio-Beads weight ratio for 2 hr at 4°C with gentle agitation. The Bio-Beads were then removed by centrifugation at 4,500 rpm and the sample was further purified by SEC in buffer containing 25 mM Tris pH 7.5, 100 mM NaCl and 0.1 mM TCEP.

All cryo-EM grids were prepared by applying 3 mL of protein at ~3 mg/mL to a glow-discharged QuantiAuFoil R1.2/1.3 300-mesh grid (Quantifoil) and blotted for 2.5 to 4 s under 100% humidity at 4°C before plunging into liquid ethane using a Mark IV Vitrobot (FEI).

Cryo-EM data of apo LaINDY were acquired on a Titan Krios microscope with a K2 direct electron detector (Gatan), using a GIF-Quantum energy filter with a 15 eV slit width. Leginon (Suloway et al., 2005) was used for ice thickness targeting and automated data collection. Each micrograph was dose-fractioned over 50 frames, with an accumulated dose of 75 e^-/Ų.

Cryo-EM data of LaINDY-malate were acquired on a Titan Krios microscope with a K2 direct electron detector, using a GIF-Quantum energy filter with a 20 eV slit width. SerialEM was used for automated data collection (Schorb et al., 2019). Each micrograph was dose-fractioned over 60 frames, with an accumulated dose of 50 e^-/Ų.

Cryo-EM data of LaINDY-αKG were acquired on a Talos Arctica microscope with a K3 direct electron detector. Leginon was used for ice thickness targeting and automated data collection (Rice et al., 2018). Each micrograph was dose-fractioned over 56 frames, with an accumulated dose of 46 e^-/Ų.

Cryo-EM data of VcINDY-Na⁺-Fab84 in nanodiscs were acquired on a Talos Arctica microscope with a K2 direct electron detector. SerialEM was used for automated data collection. Each micrograph was dose-fractioned over 50 frames, with an accumulated dose of 44 e^-/Ų.

Cryo-EM data of VcINDY-Na⁺ in amphipol were acquired on a Talos Arctica microscope with a K2 direct electron detector. SerialEM was used for automated data collection. Each micrograph was dose-fractioned over 48 frames, with an accumulated dose of 40 e^-/Ų.

Motion correction, CTF estimation, particle picking, 2D classification, ab initio model generation, heterogenous and non-uniform refinement, and per particle CTF refinement were all performed with cryoSPARC (Punjani et al., 2017). Each dataset was processed using the same protocol, except as noted. All maps were sharpened using Auto-sharpen Map in Phenix (Adams et al., 2010).

Micrographs underwent motion correction and the CTF estimated, and those with an overall resolution worse than 8 Å excluded from subsequent steps. An ellipse-based particle picker identified particles used to generate initial 2D classes. These classes were used for template-based particle picking. Template identified particles were extracted and subjected to 2D classification. A subset of well resolved 2D classes were used for the initial ab initio model building, while all picked particles were subsequent used for heterogeneous 3D refinement. After multiple rounds of 3D classification (ab initio model generation and heterogeneous 3D refinement with two or more classes), a single class was selected for non-uniform 3D refinement, resulting in the final map.

For the VcINDY-Na⁺-Fab84 sample in nanodiscs, refinement of an initial 3.36 Å map indicated oblique views were rare in the particle set. Projections of this map were therefore used as templates for an additional round of particle picking, with subsequent 3D classification, per particle CTF refinement and non-uniform 3D refinement.

LaINDY protein purified in NG and αKG was concentrated to 2 mg/mL using a centrifugal filtration device and crystallized at 18°C using the hanging drop vapor diffusion method by mixing equal volumes of concentrated protein and well solution of 30% Jeffamine ED-2001, 100 mM HEPES pH 6.8. Crystal quality was improved by the addition of 10 mM malic acid prior to size exclusion chromatography. Crystals were flash frozen in liquid nitrogen with crystallization solution serving as cryoprotectant. Crystals were of the P2₁ space group with unit cell dimensions of a = 91.3 Å, b = 76.6 Å, c = 96.9 Å, β = 90.5° and contained two molecules in the asymmetric unit, which form a single transporter dimer. Data were collected at the Advanced Light Source beamline 5.0.2. Diffraction data were processed and scaled in HKL2000 (Otwinowski and Minor, 1997).

VcINDY purified in DM was incubated with 10 mM sodium terephthalate (TTP) on ice for 30 min prior to crystallization with a protein concentration of 4 mg/mL at 4°C by hanging-drop vapor diffusion. Crystals were grown in 30% PEG 550, 50 mM NaCl and 100 mM Tris pH 8.0, and were subsequently frozen in liquid nitrogen with the crystallization solution serving as the cryoprotectant. X-ray data were collected at the Advanced Light Source Beamline 5.0.2. Data processing and scaling were performed using HKL2000. Crystals were of space group P2₁ with unit cell dimensions around a = 108 Å, b = 103 Å, c = 174 Å, β = 96°, and contained four molecules per asymmetric unit.

Cryo-EM models were built in Coot (Emsley and Cowtan, 2004) and refined in Phenix real space refine (Adams et al., 2010). The LaINDY-apo model was manually built, and subsequently used as an initial model for the LaINDY-malate and LaINDY-αKG structures. VcINDY models were built using the VcINDY X-ray structure (PDB: 5UL9), with sodium and citrate removed, as an initial model. The Fab was rebuilt from an initial model of a homologous X-ray structure (PDB: 5EII) with the variable loops removed. In both VcINDY and LaINDY maps some discontinuous densities could be attributed to the aliphatic chains of lipids, detergents, or amphipols and were modeled as alkanes for simplicity.

The LaINDY X-ray structure was determined by molecular replacement using the LaINDY-apo cryo-EM model as a starting model, followed by model building in Coot and refinement in Phenix.

The VcINDY-TTP X-ray structure was determined by molecular replacement using the structure of VcINDY (PDB ID: 5UL9), with sodium and citrate removed, as the initial search model. The model was refined using group B-factors, and NCS, secondary structure and reference model restraints.

DASS family orthologs were aligned in Promals3D (Pei et al., 2008), including VcINDY (PDB ID: 5UL9) and CitS (PDB ID: 5A1S) as structural models. As a non-member of the DASS family, CitS was used as the outgroup when calculating distances using FastTree (Price et al., 2009).

The distance between C_α and C carbons was calculated using a custom Biopython script (Cock et al., 2009). Median C_α to C distances were calculated for 2045 single particle cryo-EM and 64,676 X-ray structures deposited in the Protein Data Bank between 2013 and 2020, containing only L-type polypeptides, and resolved to at least 4 Å.

Molecular dynamics (MD) simulation systems of the C_o-S and C_o states of LaINDY were constructed from its X-ray crystal structure. Succinate was added to the X-ray crystal structure of LaINDY by aligning the cryo-EM structure of LaINDY with α-ketoglutarate bound and mutating the α-ketoglutarate into succinate. Force field parameters for succinate were determined using the online CGenFF interface (Vanommeslaeghe and MacKerell, 2012; Vanommeslaeghe et al., 2012). The protonation states of the titratable residues of LaINDY were determined using PROPKA (Søndergaard et al., 2011; Olsson et al., 2011). Two independent POPE (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine) membranes were prepared using the Membrane Builder (Wu et al., 2014; Jo et al., 2009; Jo et al., 2007; Lee et al., 2019) tool of CHARMM-GUI (Jo et al., 2008). The orientation of LaINDY in the membrane was determined by aligning the protein’s symmetry axis with the membrane-normal axis, and the insertion depth in the membrane was determined by applying the PPM (Positioning of Proteins in Membrane) Server (Lomize et al., 2012) to a model (Mulligan et al., 2016) of the C_o-Na⁺-S state of VcINDY. Once the membrane and protein were combined, all lipids within 1 Å of the protein were removed from the systems. Water was added to the systems using the VMD (Humphrey et al., 1996) Solvate plugin. The systems were neutralized with Cl⁻, and 0.15 M NaCl was added to the systems using the VMD Autoionize plugin.

All simulations were performed with NAMD (Phillips et al., 2005) using the CHARMM36m (Huang et al., 2017; Klauda et al., 2010; Vanommeslaeghe et al., 2010; Yu et al., 2012; Jorgensen et al., 1983; Venable et al., 2013) force field at a constant temperature of 310 K and a constant pressure of 1 atm. Constant temperature was maintained using Langevin dynamics for all non-hydrogen atoms with a damping coefficient of 1 ps⁻¹. Constant pressure was maintained using a Nosé-Hoover Langevin piston (Martyna et al., 1994; Feller et al., 1995) with a period of 100 fs and a damping time scale of 50 fs. The cut-off distance for both electrostatic and van der Waals interactions was set to 12 Å, and a switching function was applied at 10 Å. Periodic boundary conditions were applied in all simulations, and long-range electrostatic interactions were calculated using the particle mesh Ewald method (Darden et al., 1993; Essmann et al., 1995) with a grid point density of 1 Å⁻¹.

After building the systems, harmonic restraints were applied to the constituent atoms of the protein, membrane, and bound succinate (i.e., only the bulk water and ions were left unrestrained). A force constant of 1.0 kcal/mol/Ų was used, and an integration time step of 1 fs was used. Three thousand steps of conjugate gradient energy minimization were performed, after which the systems were simulated with MD for 1 ns. The restraints on the membrane were then released, and the systems were minimized for three thousand steps and then simulated for 10 ns. Next, the restraints on the protein and bound succinate were released, and the systems were minimized for three thousand steps and then simulated for 10 ns. With initial equilibration complete, the integration time step was changed to 2 fs, and an additional 100 ns of unbiased equilibrium simulation was performed for each system to sample the dynamics of the C_o-S and C_o states of LaINDY.

Biases were then applied with the goal of inducing a transition from the C_o-S state to an approximate target model of the C_i-S state. These biases and the approximate C_i-S target are described in the ‘Biasing Protocol’ section below. To simplify comparison between the transitions sampled by the two unique protomers of LaINDY and between independent trials of the simulation, a 10 ns transition was first induced to a convenient reference C_o-S state: the X-ray crystal structure. The system was then restrained at this reference C_o-S state for 10 ns to allow the membrane and aqueous environment to equilibrate around this new protein state. Next, a 100 ns transition to the approximate C_i-S target was induced, and the system was subsequently restrained at the approximate C_i-S target for 50 ns again to allow the environment to equilibrate around the new protein state. Finally, all restraints were released, and a 100 ns unbiased equilibrium simulation was performed to examine the stability of and to sample the dynamics of the newly generated model of the C_i-S state of LaINDY. See Figure 6—figure supplement 1a for a graphical overview of the simulations performed.

The transition to the C_i-S state was first induced without any biases applied to the bound succinate. A successful transition was sampled by protomer B, but succinate spontaneously unbound during the early stages of the transition sampled by protomer A. To increase sampling, an additional trial of this transition was performed with biases applied to the bound succinate (see ‘Biasing Protocol’). Throughout this article, data from the simulation without biases applied to succinate are labeled ‘protomer B₁.’ Data from the simulation with biases applied to succinate are labeled ‘protomer A₂’ and ‘protomer B₂.’.

In transporters with an inverted repeat topology, it has been shown that the C_o-S and C_i-S states can be generated from each other by swapping the conformations of the transporter’s repeats (Forrest et al., 2011). The internal symmetry in these transporters means that the fundamental global structural differences between their C_o-S and C_i-S states can be observed simply by rotating one state 180° about any axis in the plane of the membrane. When this approach is applied to LaINDY, it is immediately apparent that the fundamental global differences between the C_o-S X-ray crystal structure and the newly generated approximate C_i-S target model are: (1) the orientation of the transport domains relative to the scaffold domain and (2) the position of the transport domains relative to the scaffold domain along the membrane-normal axis. To describe these differences quantitatively and to induce a transition to the approximate C_i-S target, the Collective Variables (Colvars) module (Fiorin et al., 2013) in NAMD was used.

In the Colvars module, collective variables are mathematical functions used to describe collective properties of interest of selections of atoms within a system, and transitions are induced by applying moving harmonic potentials to the collective variables. In this case, one collective variable was defined for each of the fundamental global differences observed between the C_o-S state and approximate C_i-S target. Specifically, for the orientations of the transport domains, spinAngle collective variables were used, and distanceZ collective variables were used for the positions of the transport domains relative to the scaffold domain along the membrane-normal axis. Helix C_α atoms of the transport domains and transmembrane helix C_α atoms of the scaffold domain were used in the definitions of these collective variables. Schematic depictions of these collective variables are shown in Figure 6—figure supplement 1b.

To determine the rotation axis used in the definition of the spinAngle collective variables, the rotation matrix needed to structurally align the transport domains of the C_o-S state with the transport domains of the approximate C_i-S target was calculated using the measure fit command of VMD. Importantly, the order option was used to allow the first repeat of the C_o-S state to be compared to the second repeat of the approximate C_i-S target at the same time that the second repeat of the C_o-S state was compared to the first repeat of the approximate C_i-S target. This calculation also determined the change in angle (i.e., amount of spin about the spin axis) that was needed to induce the transition to the approximate C_i-S target. To induce the transition to the approximate C_i-S target using biased MD simulations, moving harmonic restraints with a force constant of 8 kcal/mol/(°)² were applied to the spinAngle collective variables.

Calculating the change in the distanceZ collective variables required to induce the transition to the approximate C_i-S target was straightforward, as the approach used to generate the approximate C_i-S target simply changes the direction of the displacement along the membrane-normal axis of the transport domains relative to the scaffold domain without affecting its magnitude. Moving harmonic restraints with a force constant of 80 kcal/mol/Ų were applied to the distanceZ collective variables. Video 1 demonstrates how the approximate C_i-S target was generated, and it shows the VMD-based quantification (described above) of the global structural differences between the C_o-S state and the approximate C_i-S target used to define the spinAngle and distanceZ collective variables.

In addition to the spinAngle and distanceZ collective variables, harmonic restraints were applied to up to two additional collective variables during the induced transition. First, the orientation of the scaffold domain was restrained to prevent the overall orientation of LaINDY from changing. Specifically, orientation collective variables were used on the scaffold domains (defined as above), and harmonic restraints with a force constant of 100,000 kcal/mol were applied to them. Second, in one trial of the transition (see ‘Simulations Performed’), succinate was restrained to the binding sites of the transport domains using distance collective variables. One distance collective variable per succinate was defined using the non-hydrogen atoms of the succinate and the C_α atoms of the residues that were consistently closest to the bound succinate during the unbiased simulations of the C_o-S state of LaINDY (i.e., Pro154-Arg159, Thr208-Pro212, Ala396-Thr399, and Asn442-Pro444). A half-harmonic restraint centered at 2.0 Å and with a force constant of 1.0 kcal/mol/Ų was applied to these collective variables.

Cryo-EM maps and models have been deposited in the Protein Data Bank and EMDB database, respectively, for VcINDY-Na⁺ in amphipol (6WU3, EMD-21904), VcINDY-Na⁺-Fab84 in nanodisc (6WW5, EMD-21928), LaINDY-apo (6WU1, EMD-21902), LaINDY-αKG (6WU4, EMD-21905), and LaINDY-malate (6WU2, EMD-21903). X-ray derived models and diffraction data have been deposited in the Protein Data Bank for LaINDY-malate-αKG (6WTW) and VcINDY-TTP (6WTX). Coordinates of representative LaINDY structures from MD simulations of the C_o-S and C_i-S states have been made publicly available on Zenodo (DOI 10.5281/zenodo.3965996). Bond length analysis code is available at Sauer, 2020; https://github.com/DavidBSauer/bond_length_analysis (copy archived at https://github.com/elifesciences-publications/bond_length_analysis).