To assess whether pantoate and its derivatives are likely substrates for ASBT_NM, we first used a dye-based stability assay in which stability is used as a surrogate for binding (Alexandrov et al., 2008). Whereas pantoate stabilised the protein to a similar amount to TCH, neither ketopantoate nor pantothenate had any effect under the conditions tested (Figure 2a). To verify binding and obtain a more reliable estimate of the affinity of pantoate for ASBT_NM isothermal calorimetry (ITC) was then used, giving a measured K_D of 127 μM (Figure 2b). Although we were unable to obtain a reliable estimate of the K_D of TCH using ITC due to its detergent-like properties, the K_D of pantoate is similar to the K_M reported for TCH (50 μM) (Hu et al., 2011).

To understand how pantoate binds to ASBT_NM we solved the structure of the protein in the presence of pantoate using X-ray crystallography (ASBT_NM(Pan)). Crystals were grown using the in meso method of crystallisation and the structure was solved by molecular replacement and refined at a resolution of 2.3 Å (Table 1). Density consistent with pantoate, and sodium ions in both ion binding sites is evident in the resulting maps (Figure 3B, Figure 3—figure supplement 1). There is also evidence of a lipid-like molecule within the binding site. The transporter adopts an inward-facing conformation as seen for the structure of ASBT_NM with TCH present (ASBT_NM(TCH); 3ZUX) (Hu et al., 2011) and the two structures can be superposed with a root mean square deviation (RMSD) of 0.6 Å for 263 out of 308 C_α atoms within 2 Å after superposition (see Materials and methods). The most substantial difference in the conformation of the two structures is seen for TM1. In the ASBT_NM(TCH) structure, TM1 bounds one side of the crevice between the panel and core domains (Figure 1A). In this structure the helix is kinked at residue Thr 14 so that it splays out and enlarges the cavity on the inward-facing side of the protein (Figure 1A). In the ASBT_NM(Pan) structure TM1 is still kinked although the whole helix has moved as an approximate rigid body by ~75° pivoting around Ile 11 such that residues 1–10 move over the cytoplasmic entrance to the cavity to partially occlude it from the inward-facing side, and residues 12–28 move away from TM10. This creates an opening into the crevice from the membrane between the panel and core domains (Figure 1B and C, Figure 1—video 1).

The pantoate binds between the two cross-over helices TM4b and TM9b (Figure 3). The carboxylic acid of the pantoate interacts with the main chain nitrogen atoms of Thr 112 and Ala 113 of TM4b and the 2-hydroxyl oxygen is within hydrogen bonding distance of the main chain nitrogen of Gly 267 of TM9b and the amino oxygen of Asn 265. The hydroxyl oxygen of the methyl-propanol moiety also interacts with His 294 and Asn 295, which reside on TM10. These residues are all within the core domain. The closest residues to the pantoate on the panel domain are Ile 203 and Ile 47 that interact with the methyl-propanol moiety. The sodium ion binding sites in ASBT_NM are also located at the cross-over region of the two helices, behind the pantoate when viewed from the crevice between the core and panel domains (Figure 3C). The ions are clearly defined in the electron density (Figure 3—figure supplement 1A) and there is very little change in their coordination in the ASBT_NM(Pan) structure with respect to that of ASBT_NM(TCH) (Figure 3—figure supplement 1B).

Within the region of the sodium and pantoate binding sites the most obvious change in the pantoate-bound structure relative to ASBT _NM(TCH) is that the main chain nitrogen of Thr 112 is displaced by ~1 Å and the C_γ by 2.4 Å (Figure 3D). In fact, there is a slight movement of the whole of TM4b, which includes the sodium ion ligands, Ser 114 and Asn 115, towards the pantoate (Figure 3D, Figure 1—video 1). On the panel domain Ile 203, located at the centre of TM7, is also displaced slightly (~1.2 Å) (Figure 3D and E) enabling the pantoate to be accommodated easily. The conformational change of TM1 may be triggered by this displacement given that Ile 203 would clash with Phe 15 if TM1 had adopted the same conformation as in ASBT_NM(TCH) (Figure 3E).

Given that the subtle conformational changes between the pantoate and TCH-bound structures would be consistent with mechanistic changes upon substrate binding, with ASBT_NM(TCH) representing a non-substrate bound structure, we also solved the structure without TCH or pantoate present (ASBT_NM(ns)) at 2.1 Å using the in meso method of crystallisation (Table 1). Overall, ASBT_NM(ns) is very similar to ASBT_NM(TCH) with an RMSD of 0.5 Å for 293 out of 308 C_α atoms (see Materials and methods) and an almost identical coordination of the sodium ions (Figure 4, Figure 4—figure supplement 1). There are only two regions where there are slightly larger changes. The first, again, centres on TM1. However, the change is much more subtle than to the pantoate-bound structure, pivoting ~15° at Thr 14 (Figure 4A and B). The second is in the loop between TM5 and TM6, which links the core to the panel domain where the loop takes a conformation more similar to that seen in the ASBT_NM(Pan) structure. In flexing between the inward- and outward-facing structures, as reported for ASBT_YF (Zhou et al., 2014), this loop changes conformation as it allows the panel to move with respect to the core. Overall, therefore, it appears that the binding of pantoate, rather than either the absence of TCH or the difference in crystallisation method, causes the change in position of TM4b.

To probe the specificity of binding we used two approaches. Firstly, we mutated the two residues for which the side chains are within hydrogen bonding distance of the pantoate in the ASBT_NM(Pan) structure and tested the affinity of pantoate for the resultant proteins by ITC. Whereas the mutation of Asn 265 to alanine caused the binding of pantoate to be abolished (Figure 5A), mutation of Thr 112 to either valine or alanine surprisingly resulted in an increase in affinity to 87 μM (Figure 5B) and 11 μM (Figure 5C) respectively, though noticeably the latter was entropy driven. Secondly, because much of the molecular recognition involves the main chain atoms, we used a structure-activity relationship approach, testing whether a panel of similar compounds would stabilise the protein (Figure 5—figure supplement 1). None of the compounds tested stabilised the protein as much as pantoate, showing the importance of the hydroxy-acetate group, which interacts with the main chain atoms. Just replacing the hydroxyl oxygen with a ketone as in ketopantoate appears to disrupt binding, likely due to an unfavourable interaction with Asn 265. The only other residues that possess the hydroxy-acetate moiety are isocitrate and D-malate. Both compounds have additional charged groups that may make them less favourable for binding.

To gain insight into the effect of sodium ions on the binding of pantoate, MD simulations were carried out. Over 0.5 µs of MD simulations, pantoate remains in the crystallographic binding position (Figure 6). With sodium ions present in both the Na1 and Na2 sites, the hydrogen bonds between pantoate and the main chain nitrogen atoms of T112 and A113 at the N-terminus of TM4b and G267 at the N-terminus of TM9b remain intact (Figure 6A) with more fluctuating interactions with the side chain atoms. On the other hand, in the absence of sodium ions the pantoate is more mobile (Figure 6B, Figure 6—figure supplement 1A) with the interactions with the main chain atoms more intermittent (Figure 6A). Over the course of the simulations with the sodium ions, the ions remain stably bound in or close to the Na1 and Na2 sites although the simulations indicate that there is an alternative sodium ion binding position close to the crystallographic Na1 site (labelled Na1* in Figure 6C). In the simulations without bound sodium ions, ions enter the inward-facing funnel and approach the Na1* binding site, but do not settle into the same binding mode seen in the bound simulations (Figure 6—figure supplement 2), although complete binding events may occur on longer time scales.

ASBT_NM (Hu et al., 2011) was subcloned into a modified version of the expression vector, pWaldo GFPd (Drew et al., 2006) in which the TEV protease site had been altered to a 3C protease recognition site (Hatton et al., 2022). Site-directed mutations were introduced by PCR (Quikchange II, Agilent Technologies). Cultures were grown in Lemo21 (DE3) cells in PASM-5052 media following the MemStar protocol (Lee et al., 2014). Briefly, the cells were grown at 37°C with shaking at 200 rpm. At an OD₆₀₀ of 0.5, 0.4 mM IPTG and 0.25 mM L-rhamnose were added and the temperature was decreased to 25°C for overnight induction. Cell pellets were harvested by centrifugation at 5000 × g for 15 min at 4°C and resuspended in PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.8 mM KH₂PO₄) with 1 mM MgCl₂, DNaseI, and 0.5 mM Pefabloc (Roche). Cells were lysed by passing them twice through a cell disruptor at a pressure of 25 kpsi. Unbroken cells and cell debris were pelleted by centrifugation at 15,000 × g for 13 min and the supernatant was subjected to ultracentrifugation at 200,000 × g at 4°C for 1 hr to pellet the membranes. Membrane pellets were resuspended in PBS, 15 ml per 1 l of culture, snap frozen in liquid nitrogen, and then stored at –80°C.

Membranes were solublised in 1× PBS, 150 mM NaCl, 10 mM imidazole, and 1% (wt/vol) DDM supplemented with 0.5 mM Pefabloc (Roche) for 2 hr at 4°C. Insolubilised material was removed by centrifugation at 200,000 × g for 45 min and the supernatant was added to HisPur Ni-NTA superflow agarose, (Thermo Fisher) (1 ml per 1 mg GFP-tagged protein). The slurry was gently stirred for 3 hr to allow binding and then loaded into a glass Econo-Column (Bio-Rad). The column was washed with 10 column volumes (CV) of wash buffer (1× PBS, 150 mM NaCl, 0.1% DDM) containing 20 mM imidazole, followed by 10 CV with the imidazole augmented to 30 mM. 3C protease (1:1 stoichiometry with ASBT_NM-GFP) was added to the resin and cleavage was performed overnight at 4°C. The protein was eluted with 20 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.03% DDM, and passed over a 5 ml HisTrap HP column (GE Healthcare) equilibrated with the same buffer. The flow-through was collected and concentrated to 6–10 mg/ml using a 100 kDa molecular weight cutoff centrifugal concentrator (Sartorius) and loaded onto a Superdex 200 Increase 10/300 GL column equilibrated with 20 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.03% DDM. Fractions containing protein were pooled together and concentrated to ~25 mg/ml as above.

Screening of compounds for binding was carried out using a stability assay based on binding of 7-diethylamino-3-(4’-maleimidylphenyl)-4-methylcoumarin (CPM) to the protein (Alexandrov et al., 2008; Sonoda et al., 2011). CPM (Thermo Fisher) was dissolved in DMSO to a final concentration of 4 mg/ml. The assay was performed in 0.2 ml non-skirted low profile 96-well PCR plates (Thermo Fisher). 50 μl of protein (2.5 μg in 20 mM Tris-HCl [pH 7.5], 150 mM NaCl, 0.03% DDM) was added to each well supplemented with 1 μl (final concentration 1 mM) of each of the compounds of interest and the plate incubated for 30 min at room temperature. The CPM dye was diluted 1:100 in 20 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.03% DDM, and 2.5 μl of the diluted dye was added to each well. The assay was performed using a Stratagene Mx3005P Real-Time PCR machine (Strategene) and samples were heated from 25°C to 95°C in 1°C/min steps. Data were analysed using GraphPad Prism.

The protein sample was dialysed overnight at 4°C against 20 mM Tris-HCl (pH 7.5), 150 mM NaCl, and 0.03% DDM and centrifuged at 16,000 × g at 4°C for 30 min. ITC experiments were performed on a MicroCal PEAQ-ITC (Malvern Panalytical, UK). The protein solution (220 μM) was filled into the sample cell and the pantoate solution (5 mM in the dialysis buffer) into the syringe. The cell temperature was set to 10°C with a stirring speed of 750 rpm and a reference power of 10 μcal/s. 20 injections were performed with an initial delay of 250 s. The initial injection was performed for 0.8 s with an injection volume of 0.4 μl. The later injections were performed for 4 s with an injection volume of 2 μl. 180 s spacing was left between each injection. The data were analysed using the ‘one set of sites’ model within the MicroCal PEAQ-ITC software (Malvern) iterated using the Lavenberg-Marquardt algorithm after subtraction of the control experiment (pantoate titrated into buffer). The thermodynamic and binding parameters were derived from the nonlinear least squares fit to the binding isotherm.

The pantoate-bound ASBT structure was embedded in in a 80:20 POPE:POPG bilayer and solvated with neutralising ions (0.15 M NaCl) to a final box size of 9.1 × 9.1 × 9.6 nm³ using CHARMM-GUI (Jo et al., 2008; Lee et al., 2016; Wu et al., 2014). An initial structure was generated without bound sodium. Two sodium ions were moved back to the binding sites manually to generate a sodium-bound initial structure.

From each starting structure, simulations were performed using Gromacs 2018.6 (Abraham et al., 2015) with the CHARMM-36 forcefield (Best et al., 2012) and TIP3 water. Parameters for pantoate were generated using CGenFF (Vanommeslaeghe et al., 2010) and converted to a Gromacs format using the cgenff_charmm2gmx.py script. Energy minimisation and 5 ns multi-step equilibration were performed following the CHARMM-GUI protocol, followed by three 500 ns production runs using different initial velocities. The simulation timestep was set to 2 fs; temperature and pressure were maintained using the stochastic velocity rescaling thermostat (Bussi et al., 2007) (at 303.15 K) and the Parrinello-Rahman semi-isotropic barostat (Parrinello and Rahman, 1981) (at 1 atm), respectively. The particle-mesh Ewald method (Darden et al., 1993) was used for long-range electrostatic interactions, and non-bonded interactions were reduced from 1 nm to a 1.2 nm cutoff using potential shift. Bonds involving hydrogens in ASBT, lipids and pantoate were constrained using the LINCS algorithm (Hess et al., 1997); all bonds in the rigid TIP3 water molecules were constrained with SETTLE (Miyamoto and Kollman, 1992). Three repeats of 500 ns were carried out for both the structures with no sodium bound (-Na⁺) and sodium ions bound in both the Na1 and Na2 sites (+Na⁺). All simulation analysis was performed using MDAnalysis (Gowers et al., 2016; Michaud-Agrawal et al., 2011). For hydrogen bond analysis, a 3.5 Å distance and 145^o angle cutoff were used. Visualisations of structures were made using VMD (Humphrey et al., 1996).