Overview of the cryo-EM structures.

Density maps and corresponding atomic models are shown for each mutant. Fab molecules are shown in orange (light chain) and red (heavy chain) with YiiP colored in cyan, blue, purple and green, depending on the mutant. The homo-dimers adopt C2 symmetry for WT (A), D51A (B) and D287A (D) mutants, but a bend between TMD and CTD break this symmetry for D70A (C). The D287A/H263A mutant (E) forms a dimer of dimers in which the Fab molecules are rather disordered. Two conformations were observed for D70A, both of which are shown: D70A_sym on the left and D70A_asym on the right. Location Zn2+ binding sites, membrane boundaries as well as the topology of the YiiP protomer (inset) are shown in panel A; rectangles and arrows represent α-helices and β-sheets, respectively. Insets in panels B-D show the coordination geometry at the individual sites as seen in the WT structure.

Structure determination of YiiP mutants

Comparisons of atomic structures.

(A) Overlay of the TMD for WT (cyan) and D51A (blue) structures with a view prioritizing protomer A. Core helices (M2, M3, M6) align well but that there are substantial changes to the peripheral helices (M1, M4, M5), presumably due to lack of Zn2+ binding at the site bridging M2 and M5. (B) Overlay of the TMD for WT (cyan) and D70A_sym (purple) with a view along the two-fold axis shows a very similar configuration of helices and an intact dimeric interface. (C) Structure of D70A_asym viewed along the membrane plane showing the asymmetry between the two TMD’s. The protomer on the left (chain B) adopts an IF conformation, whereas the protomer on the right (chain A) adopts a novel occluded conformation that includes a reconfigured TM2/TM3 loop (blue) making a novel interaction with the CTD. (D) Overlay of the TMD for WT (cyan) and D70A_asym (plum) showing the occluded conformation adopted by protomer A (top) and the IF conformation adopted by protomer B (bottom). L199 and L154 make van der Waals interactions that appear to stabilize the occluded conformation. (E-F) Overlays of D70A_asym (light purple) and D70A_sym (dark purple) structures after alignment of core helices TM3 and TM6. Structures are viewed parallel to the membrane plane in E and along the two-fold axis in F. A significant shift in the position of the CTD is apparent in E and bottom panel in F. Despite this shift and substantial conformational changes in the occluded protomer A, the dimer interface in the TMD (TM3 and TM6) is well aligned (F, top panel). (G) Domain swapped dimer of dimers adopted by the D287A/H263A construct. Dimerization of TMD’s involves interaction between one dark-green and one lightgreen molecule (e.g., TMDa and TMDb, where “a” and “b” refer to chain ID’s), whereas dimerization of CTD’s involves interaction between either two dark-green or two light-green molecules (e.g., CTDa and CTDd). The linker between M6 and the CTD adopts a long straight helix in chains b and d, but remains an unstructured loop in chains a and c. (H) Overlay of the TMD for WT (cyan) and D287A/H263A structures viewed along the dimer axis shows a good match, indicating that disruption of site C affects mainly the configuration of the CTD.

Zn2+ removal from site B in the TM2/TM3 loop.

(A) Per-residue RMSF of the WT, holo structure with Zn2+ present (cyan) and absent (purple) at site B demonstrates a notable increase in fluctuations in the TM2/TM3 loop. The dashed line demarcates the boundary between TMD and CTD. (B) Angle between the TMD and CTD in three individual simulations in the absence (purple) and presence (cyan) of Zn2+ at site B. The distribution of angles, on the right, highlight greater mobility when site B is empty. (C) Per-residue RMSF for simulations with empty site B using three different alignment schemes: the entire molecule (“All), transmembrane domain (“TMD”), or C-terminal domain (“CTD”). Analogous data for the WT, holo structure have previously been published (Fig. 3i in Lopez-Redondo et al., 2021). (D) Distributions of distance between Asp72 in one chain and Arg210 in the opposite chain during simulations of the WT, holo structure, the apo structure with Zn2+ absent from all sites, and the structure with site B empty. The sharp peak at ~2 Å from the holo structure suggests a salt bridge that is disrupted in other structures. (E,F) Structure of the WT, holo YiiP dimer showing global C2 symmetry about a vertical axis and juxtaposition of the TM2/3 loop (blue) with the TM6/CTD linker from the opposing protomer (orange-to-red). (G,H) D70A_sym structure shows a kink between TMD and CTD and disordering of the TM2/3 loop. Both protomers are in the IF state. (I,J) D70A-asym structure showing further twisting of the CTD and asymmetry of the TMD’s. The TM2/3 loop is disordered in the protomer on the left (chain B), but adopts a novel interaction with the CTD in the protomer on the right (chain A). (K,L). Electrostatic surface of D70A_asym showing a negatively charged cavity leading to site A on the left, but an occluded cavity with positive charge on the right. Rainbow colors progress from blue to red moving from the N-terminus of one protomer to the C-terminus of the other protomer. Note that L is at an oblique angle looking down on the M2-M3 loop.

Binding affinity for individual Zn2+ sites measured by MST* or deduced by the MST inference algorithm§

pH dependence of Zn2+ binding affinity.

MST was used to experimentally determine Zn2+ binding affinity at each site (panels A, B and C). These data were combined with a thermodynamic model, represented in panel H, with parameters derived either from CpHMD (panels D and E) or MST inference (panels F and G). (A) For MST studies of site A, the D70A/D287A/H263A construct was used with Zn2+ buffered either with NTA (pH 7.0 and 7.4) or with citrate (pH 5.6, 6.0 and 6.5). Curves represent the law of mass action with Kd values listed in Table 3. The relatively poor fit at pH 7.4 may reflect the fact that the Kd (1 nM) is lower than the minimum protein concentration (8 nM) supported by the assay. (B). For MST studies of site B, the D51A/D287A/H263A construct was used and Zn2+ was buffered with citrate. (C) For MST studies of site C, the D51A/D70A construct was used and Zn2+ was buffered with NTA at pH 7 or with citrate at the other pH’s. (D,E) Curves represent predictions of a thermodynamic model with pKa values for either site A residues (Asp47, Asp51, His153, Asp157) or site B residues (D70, H73, H77) taken from CpHMD simulations. These pKa values are listed in Table 4 and the distributions of protonation states are shown in Suppl. Figs. 10 and 11. Symbols represent experimental MST data as in A and illustrate a poor fit using these parameters. Thermodynamic modeling was not possible for site C because of instability of CpHMD simulations in the absence of Zn2+. (F,G). Curves represent predictions of the MST inference algorithm with corresponding pKa values listed in Table 4. Again, symbols represent experimental MST data and show the excellent fit using these refined parameters. (H) Schematic representation of the microscopic thermodynamic model for site A. Titratable residues (D47, D51, D159, H155) are represented either as black squares for the protonated state, or white squares for the deprotonated state. Binding of one Zn2+ ion to the site is indicated by a filled magenta circle. Transitions are only possible between states connected by lines (protonation/deprotonation) or corresponding states connected by double-sided arrows (Zn2+ binding/release). An analogous model for site B comprised D70, H73 and H77.

Binding affinity and Hill coefficients derived from MST data*

pKa values of residues determined by CpHMD simulations (Hill-Langmuir) and MST inference

Transport cycle for YiiP.

According to the alternating access paradigm, YiiP toggles between the OF and IF states via an intermediate occluded state. We assume that these changes occur independently in each protomer based on asymmetry seen in structures of soYiiP and Znt8 (Xue et al., 2020). Thus, our model depicts changes in the right-hand protomer while the left-hand protomer remains in a resting IF state. Although these intermediate states are informed by current and past structural work, they do not precisely conform to solved structures but represent hypothetical states that we believe to exist under physiological conditions. (A) Zn2+ is released to the periplasm. The TM2/TM3 loop is depicted interacting with the CTD in a Zn2+-free, extended conformation, as seen in our D70A_asym structure. (B) The release of Zn2+ is facilitated by the low pH of the periplasm and results in protonation of two residues in site A, or potentially three residues at lower pH. (C-D) The protonated form transitions to the IF state via an occluded state. In the IF state, the Zn2+-free TM2/TM3 loop is disordered. (E) Zn2+ is recruited to site B, inducing an ordered conformation of the TM2/TM3 loop that folds onto the CTD. (F) Zn2+ is transferred from a relatively low affinity site B to the much higher affinity site A via a negatively charged access channel, thus displacing two protons. (G) This transfer induces a Zn2+-bound, occluded conformation in which the CTD tilts toward the occluded protomer and interacts with the TM2/TM3 loop in its Zn2+-free, extended conformation. Features of the model are illustrated in the middle, boxed panel with desaturated colors. Zn2+ ions are depicted as magenta spheres and protons with a “+”. The CTD is pink with two Zn2+ ions constitutively bound at site C. The scaffolding membrane helices (TM3 and TM6) are blue and the transport domain (TM1,2,4,5) is yellow. The TM2/TM3 loop is blue and depicted with dashed lines in the disordered state. Created with biorender.com.

© 2024, BioRender Inc. Any parts of this image created with BioRender are not made available under the same license as the Reviewed Preprint, and are © 2024, BioRender Inc.

MD simulations

MST convergence criteria

Determination of WT and D51A structures by cryo-EM.

(A,B) SDS-PAGE, left, and elution profile from SEC purification. Molecular weight standards are shown on the right at 116, 66, 45, 35, 25, 18.4, 14.4 kDa, with the YiiP monomer running as expected at 32.5 kDa. (C,D) The cryo-EM workflow starts with correction of beam-induced motion and contrast transfer function (CTF) of movies, followed by template-based particle picking. The particle set was subjected to two rounds of 2D classification and ab initio structure determinations to remove false positives. (E,F) Hetero-refinement jobs were used to look for multiple conformations and to maximize the homogeneity of particles for a final non-uniform refinement job, in this case with C2 symmetry. (G,H) The final structure was characterized by local resolution, represented by the colored surface shown from two orthogonal directions, and the Fourier Shell Correlation.

Density at Zn2+ sites in the cryo-EM maps.

An overview of each cryo-EM structure is shown on the left, followed by densities at sites A, B and C, respectively. Contour levels used to display the various maps were as follows. WT: 7 σ, D51A: 7 σ, D287A, 6.5 σ for sites A and C, 5.5 σ for site B, D70A_asym: 9 σ, D70A_sym: 9 σ, D287A/H263A: 10 σ. Side-chains from selected residues only are shown, for clarity. For D70A_sym, site B is disordered in both protomers. In D70A_asym, the TM2/TM3 loop forms a Zn2+-free association with the CTD (H1 and C-terminus of M6) in one protomer and is disordered in the other protomer. In D287A/H263A, one of the TM2/TM3 loops associates with CTD’s from opposing protomers (CTDb, CTDc and TMDd refer to chains B, C and D, respectively). The other TM2/TM3 loop associates with the TM6/CTD linker as shown in Suppl. Fig. 5.

Determination of D70A structures by cryo-EM

(A) SDS-PAGE and elution profile for SEC purification; molecular weight markers are at 116, 66, 45, 35, 25, 18.4, 14.4 kDa. The workflow for image processing was complex, involving multiple attempts to segregate distinct classes via ab initio and hetero-refinement jobs. This chart is a summary of the final, productive steps leading to structures presented. (B) Initial steps included correction of beam-induced motion and CTF followed by template-based particle picking, rounds of 2D classification and ab initio structure determination. (C) Hetero-refinement was used to segregate two major classes: class 1 comprising the expected dimer and class 2 comprising a dimer of dimers. The dimer of dimers involved interactions between Fab domains and thus did not appear to affect the conformations adopted by YiiP. (D) Particles from class 1 were subjected to further hetero-refinement to segregate into an additional two classes: one with symmetric TMD’s (purple) and the other with asymmetric TMD’s (blue) as described in the text. (E) Particles from class 2 (dimer of dimers) were exported to RELION where symmetry expansion and signal subtraction was used to generate a larger set comprising isolated dimers. (F) These symmetry-expanded particles were grouped together with class 1 particles for hetero-refinement using reference volumes with symmetric and asymmetric TMD’s as well as a junk collector (not shown). (G, H) The resulting final classes were then used for non-uniform refinement to produce D70A_sym and D70A_asym structures, each colored according to local resolution and shown from orthogonal angles.

Determination of D287A and D287A/H263A structures by cryo-EM

(A,B) SDS-PAGE and elution profiles from SEC purifications. Images come from a single gel with the molecular weight markers (116, 66, 45, 35, 25, 18.4, 14.4 kDa). Position and presence of the double peak is consistent with the higher order oligomerization seen during image analysis. (C,D) Corrections for motion and CTF were followed by two rounds of 2D classification and ab initio structure determination. (E,F) Hetero-refinement was used to select a homogeneous set of particles for non-uniform refinement. (G,H) Final structures are characterized by FSC and by local resolution, illustrated by the plots and by surface coloring, respectively.

Interactions between the TM2/TM3 loop and the CTD.

(A) In the WT structure, the TM2/TM3 loop (blue) contacts the TM6/CTD linker from the opposing protomer with proximity of D72 and R210. (B) In the D70A_asym structure, the Zn2+-free TM2/TM3 loop extends to interact with the H1 helix in the CTD of the opposing protomer. (C) In the D287A/H263A structure, the TM2/TM3 loop extends towards CTD’s of two different protomers (e.g., TM2/TM3 loop from chain B inserts between CTD’s of chains A and D). (D) In the other monomer of D287A/H263A, the TM2/TM3 loop interacts with the TM6/CTD linker of an opposing chain (e.g., the loop from chain A interacts with the linker from chain B). In these chains, the TM6/CTD linker has refolded into one, long continuous helix.

Measuring Zn2+ affinity by MST

(A-D) SEC elution profiles for YiiP constructs. WT protein (peak at 12.4 ml) is included for comparison but was not analyzed by MST. The site A mutant, D70A/H263A/D287A, site B mutant, D51A/H263A/D287A, and site C mutant, D51A/D70A, produced peaks at at 12.6 ml, 12.3 ml and 12.6 ml respectively. (E-G) Raw data from MST for the titrations at pH 7 plotted in Fig. 4. The blue rectangles correspond to the time window used for initial fluorescence levels and the red rectangles correspond to the time window used for the thermophoresis rates. Each plot includes all Zn2+ concentrations, which were performed in triplicate.

Walk of replica simulations in CpHMD pH-replica exchange (REX) through pH space.

Each panel shows how one replica simulation (Rep:0 to Rep:29) changes over the course of the simulation (Frame number) its current pH state, as indicated by the “pH replica” on the ordinate. The pH replicas range from pH 1.5 (pH replica 0) to 11.5 (pH replica 29). Multiple replicas sample most of the pH range and many move across the especially relevant range between pH replica 9 (pH 4) and pH replica 22 (pH 9), indicating that the REX procedure samples pH space well.

Convergence of the deprotonated fraction for titratable residues in CpHMD simulations.

Residues in site A (D47, D51, H155, D159) and site B (D70, H73, H77) from both protomers (A and B) are shown. The instantaneous deprotonated fraction S is plotted as a function of simulation time and pH value, sampled across all replicas in the replica exchange simulation. The deprotonated fractions generally converge to a stable value after about 8 ns with similar behavior in both protomers A and B, thus indicating sufficient sampling during the CpHMD simulations.

Titration curves for titratable residues in CpHMD simulations.

Residues in site A (D47, D51, H155, D159) and site B (D70, H73, H77) are shown from both protomers (A and B). The unprotonated fraction S from the end of the CpHMD simulations is shown as a function of pH (black points). The Hill equation (generalized Henderson-Hasselbalch equation) is fitted to the data (black line). The estimated pKa is indicated as a red line at S=0.5. As described in the text, H73/H77 were considered to be coupled and were analyzed in aggregate; the two pKas for the combined system (H73AH77A and H73BH77B) are represented by the pair of red lines in the last two panels.

Site A Protonation and Zn2+-binding states by CpHMD and MST inference.

(A). Populations of protonated states obtained directly from CpHMD simulations for site A, in the absence of Zn2+. Data for protomer A and protomer B were combined. Color coded definitions of individual states are shown below with a “1” indicating the protonated state and absence of a number the unprotonated state of each corresponding residue. (B) Populations of protonated states from an inverse Multibind model based on inferring microscopic pKa values directly from the CpHMD populations in (A). (C) Population of protonated states resulting from MST inference, which relies on the MC method to refine microscopic model parameters based on experimental MST data. Populations represent the Zn2+-free states although the model incorporates the whole range of Zn2+ bound states [see (F)]. Only states S0, S3, S7, S11 have appreciable (>0.0001 maximum probability) although all states contribute to aggregated state probabilities in D and E. (D) MST inference results represented as populations of aggregated states defined by the total number of protons bound in the absence of Zn2+. (E) Deprotonated fraction of each site A residue as a function of pH, averaged over all possible states of the MST inference model in the absence of Zn2+. Data points (symbols) were generated from the model; then the Hill-Langmuir equation was fit to the generated data in order to obtain an effective per-residue pKa (listed in Table 4). (F) Population of states derived from the MST inference model as a function of Zn2+ concentration at various pH values. S0Zn (dotted line) denotes the fully deprotonated, Zn2+-bound state, which is the only populated Zn2+-bound state predicted by the MST inference.

Site B Protonation and Zn2+-binding states by CpHMD and MST inference.

(A). Populations of protonated states obtained directly from CpHMD simulations for site B, in the absence of Zn2+. Data for protomer A and protomer B were combined. Color coded definitions of individual states are shown below with a “1” indicating the protonated state and absence of a number the unprotonated state of each corresponding residue. (B) Populations of protonated states from an inverse Multibind model based on inferring microscopic pKa values directly from the CpHMD populations in (A). (C) Population of protonated states resulting from MST inference. For this analysis, a symmetry restraint was imposed on H73 and H77 as described in the text. Populations are shown in the absence of Zn2+ although the model incorporates the whole range of Zn2+ concentrations [see (G)]. (D) MST inference results represented as populations of aggregated states defined by the total number of protons bound in the absence of Zn2+. (E) Deprotonated fraction of each site B residue as a function of pH, averaged over all possible states of the MST inference model in the absence of Zn2+. Data points were generated from the model and the Hill-Langmuir equation was fit to D70 data (solid black line) to obtain an effective per-residue pKa; data for H73 and H77 are not correctly modelled by a Hill fit (not plotted). (F) H73 and H77 were treated as a coupled system and the total deprotonated fraction as a function of pH was fit with the “coupled titration model” (solid black line), resulting in two effective pKa values. (G) Population of states from the MST inference mode as a function of Zn2+ concentration at various pH values. Unlike site A, multiple Zn2+-bound states (S0Zn, S1Zn, S2Zn, S3Zn) are populated at lower pH values.

Summary of structures from CDF transporters.

Cartoon representations of most structures so far determined. The structures have been grouped according to conformational state: symmetric IF, symmetric OF and asymmetric states. PDB accession codes are indicated for previously published structures together with concentrations of Zn2+ present in the solution and the number of ions observed in the structures. Accession codes for current work are listed in Table 1. Although the sites been named differently for Znt7 and Znt8, the A,B,C nomenclature has been used to simplify the comparison. The helices in the TMD are shown in rainbow colors starting with blue for TM1 and red for TM6. CTD’s are represented as a triangle. Zn2+ ions as magenta spheres. Hydrophobic residues forming the so-called hydrophobic gate (Leu154 and Leu199 in SoYiiP) are shown as sticks. Disordered in TM2/TM3 loops is indicated by a blurred line.