Supplementary Figure S1. Urea-mediated buffering mechanism of H. pylori. Gastric urea is funneled to a cytoplasmic urease that is presumably attached

Transcription

1 Supplementary Figure S1. Urea-mediated buffering mechanism of H. pylori. Gastric urea is funneled to a cytoplasmic urease that is presumably attached to HpUreI. Urea hydrolysis products 2NH 3 and 1CO 2 are exported to buffer the periplasm to ph ~6.1.

2 Supplementary Figure S2. A: Folding simulation of the modeled periplasmic loop 1. The charged histidine residues are highlighted in green. They transmembrane section of the protein was harmonically restrained to the crystal structure coordinates. Bilayer and water molecules have been removed for clarity. B: Pore collapse of the monomeric HpUreI protomer when unrestrained. The figure shows the structural overlap of the initial (0 ns) and final configuration (200 ns). The bilayer has been removed for clarity.

3 Supplementary Figure S3. A: Pore occupancy of urea vs. simulation time in the HpUreI hexamer. On average the hexamer is occupied 58% of the time with an average of 1.3 ± 0.6 urea molecules in a pore when occupied. Each protomer is occupied for only ~10% of the time with an average of ~1 urea in the pore when occupied. B: Trans-bilayer density distribution of urea, averaged over the 1 μs hexamer simulation.

4 Supplementary Figure S4. A: Simulation system of the HpUreI monomer (148 POPC lipids, ~53,000 atoms, CHARMM force field). Phosphate groups are orange, chloride ions red, and sodium ions blue. Urea is shown in space-filling representation. B: The 7 μs achieved with this system capture 10 spontaneous transport events summarized in Table S4.

5 Supplementary Figure S5. Inter-loop contacts of the hexamer are restricted to periplasmic loop 1 (PL1, residues 55-75). For each protomer, residues interact with residues in the neighboring loop. This arrangement might facilitate cooperative opening/closing of the channel during ph gating.

6 Supplementary Figure S6. Urea conduction through the HpUreI protomer cavity. A: Urea molecules along various stages of the conduction pathway. Helix 1 has been removed for clarity. The channel surface is color-coded: blue = positively charged, red = negatively charged, green = polar, and grey = hydrophobic. B: Urea molecules have very few specific interactions with channel lining residues, which occur mainly at the constriction.

7 Supplementary Figure S7. Free energy profile of urea transport through the channel pore for the monomer and protomer F of the hexamer simulation. The profiles were derived from the occupancy of urea in the channel. The monomer is based on 10 conduction events occurring over 7 μs, while the hexamer result is based on 2 conduction events occurring over 1 μs, and therefore not fully converged.

8 Supplementary Figure S8. Structural overlay of the crystal structure conformation (red) and the equilibrated conducting conformation (white, after 480 ns) of chain F. The view from the periplasmic side (A) shows that the loops are folded away, clearing the entrance to the pore, while the helices are spaced a little further apart. At the site of the periplasmic constriction C P helices A, C, and F tilt outwards by 1, 2, and 2.5 Å, respectively, while helix D moves inwards by ~1.5 Å with respect to crystal structure, resulting in a widening of C P by ~2 Å, which is sufficient to permanently open this constriction to urea flux. The side view (B) shows that the overall structural changes are minimal. In the crystal structure the periplasmic constriction (C P ), formed by the residues Leu6 (grey), Leu9 (iceblue), Phe84 (yellow), Trp146 (pink), and Trp149 (white), was found to be too narrow for conduction of urea or water (C, top panel). The widening of the periplasmic vestibule in combination with a rotation of Phe84 (yellow) results in an opening of C P (C, lower panel). In contrast, the cytoplasmic constriction (C C ) shows no significant rearrangement (D).

9 Supplementary Figure S9. Structural relaxation of the protomer (chain F) during the one microsecond simulation. After a rapid initial rise the RMSD converges at ~2.8 Å after ~600 ns.

10 Supplementary Figure S10. Simulations of the monomeric HpUreI protomer before the opening of the periplasmic constriction. A: In order to prevent pore collapse in the absence of the other protomers (Figure S2), the C α -carbons were restrained to the crystal coordinates using 1 kcal/mol/å 2. B: Restraining the channel prevents the opening of the Cp constriction, with no urea transfer observed in 1 µs. The colored lines are the z-positions of individual urea molecules. C: The channel remains closed even if the restraints are reduced to 0.1 kcal/mol/å 2. A further softening of the restraining potential resulted in partial collapse of the channel, similar to the one observed for the unrestrained monomer simulation (Figure S2).

11 Supplementary Figure S11. Change in channel hydration for mutants. Six mutants were chosen and modeled as one hexamer for 1 microsecond in the membrane. Water super-position graphs show the average channel hydration over a typical 50 ns period. Each mutated residue is shown in space-filling representation.

12 Supplementary Figure S12. A: Sampling of sodium (blue) and chloride (red) ions, showing that the ions generally do not enter the membrane or HpUreI channels during the simulations. The only exception is a chloride ion that enters the cytoplasmic vestibule of the channel but does not cross the cytoplasmic constriction C C. B: Time evolution of the ions, showing that Na + enters the bilayer more deeply, due to hydrogen bonding interactions with the phosphate head groups. At ~5 μs a single chloride ion enters the cytoplasmic vestibule for a brief period of 2 ns.

13 Supplementary Figure S13. Pore surface analysis by polarity. Positively charged residues are blue, negatively charged residues are red, polar residues are yellow, and hydrophobic residues are white. The cytoplasmic vestibule surface is less hydrophobic than the periplasmic vestibule surface.

14 Supplementary Figure S14. Time evolution of ammonium ions (NH 4 + ), showing that the ions do not enter the membrane or HpUreI channels during the simulation.

15 Supplementary Figure S15. A: Snapshot from a simulation containing urea, NH 3, CO 2, and NH + 4. The CO 2 and NH 3 molecules can be seen traversing the lipid bilayer freely, while urea and NH + 4 do not enter the membrane. B: Trans-bilayer flux of CO 2. C: Trans-bilayer flux of NH 3. Both the NH 3 and CO 2 fluxes are many orders of magnitude higher than urea transport at similar concentrations. Urea conducts only through HpUreI channels, while NH + 4 was not found to conduct either through the membrane or through the HpUreI channels. Unidirectional conduction rates, obtained by fitting to straight lines, are ~9 x 10 8 s -1 for 100 mm CO 2 and ~3 x 10 7 s -1 for 150 mm NH 3. Adjusted for 5 mm concentrations both solutes conduct at rates >10 6 s -1, much higher than the rate of urea transport.

16 A B Supplementary Figure S16. A: Docking model of membrane-embedded HpUreI (green) to the cytoplasmic urease dodecamer (grey), viewed perpendicular to membrane (pink). Loop residues regulating urea access to active site are related by 3-fold symmetry (blue). Docking results were obtained by using the urease monomer (PDB code 1e9y) as the receptor and HpUreI hexamer as the ligand using the ClusPro 2.0 online server 52. The docked structure was aligned to the urease dodecamer and found to lie almost exactly on the urease 3-fold axis. The figures were constructed using epmv plugin in Cinema4D 53. B: View through HpUreI from the periplasm. The 3-fold symmetry axis is indicated by an orange triangle.

17 Supplementary Figure S17. During transport through the cytoplasmic constriction, urea is pressed against a hydrophobic wall of highly conserved residues (Leu6, Val9, and Leu13) by Trp153. Protonated Glu177 and Tyr88 offer orthogonal hydrogen bonding partners.

18 Supplementary Figure S18. Folding of the two periplasmic loops PL1 and PL2 during the 400 ns loop equilibration simulation. The charged histidine residues are highlighted in green and are required for ph gating of the channels.

19 Supplementary Figure S19. Free energy profile of urea transport across POPC lipid bilayers. The profile was determined by the WHAM advanced sampling technique 4. The barrier of urea transport is ~10 kcal/mol (~6 kt). Urea shows a slight attraction towards the lipid head groups (near ±18 Å along the membrane normal) compared to bulk water. Desolvation of urea in the hydrophobic membrane core (shaded in grey) is highly unfavorable.

20 Supplementary Figure S20. Orientation vector definition for the aromatic plane normal. Theta (θ) is measured between the aromatic ring normal and the membrane normal, while phi (φ) is measured between the projection of aromatic ring normal vector onto the x,y plane and the x-axis. This distinguishes between orientations that open the pore (θ 90 and φ > 180 ) and orientations that close the pore.

21 System Length [ns] [C] urea [mm] [C] NaCl [mm] Comment Mono loop model Mono unrestrained, pore collapses Mono unrestrained, pore collapses Mono unrestrained, pore collapses Mono5 1, restrained: 1 kcal/mol*a 2, loop closes channel Mono6 6, restrained: 1 kcal/mol*a 2, loop restrained Mono restrained: 1 kcal/mol*a 2 (Cp closed) Mono restrained: 0.1 kcal/mol*a 2 (Cp closed) Hexa1 1, unrestrained Hexa2 1, mutants a Hexa mutants and solutes b Hexa mutants and solutes c Hexa mutants and solutes d Supplementary Table S1. Summary of equilibrium simulations. NPT ensemble (T = 37 C at 1 bar pressure) for all systems. The OPLS all-atom protein and united-atom lipid force fields were used for Mono1-3 and Hexa1-5; Mono4-8 used the all-atom CHARMM27 protein and all-atom CHARMM 36 lipid force field. System sizes are: Mono1-4 = ~21,000 atoms (protein = 195 residues, 92 POPC, 4324 H 2 O, 18 NaCl, 13 urea), Mono5-8 = ~53,000 atoms (protein = 195 residues, 148 POPC, 9747 H 2 O, 38 NaCl, 71 urea), Hexa1-5 = ~110,000 atoms (protein = 6x195 residues, 370 POPC, H 2 O, 303 urea, 172 NaCl). Position restraints on the transmembrane helix backbone atoms only (C α, C, O, & N); the loops are flexible. Position restraints on the transmembrane helix backbone atoms and the backbone loop atoms (C α, C, O, & N); a Mutants: A: W149Y, B: W149F, C: W153F, D: W153A, E: Y88F, F: Y84L. b 250 mm CO 2 and 500 mm NH 3. c 150 mm urea, 100 mm CO 2 and 150 mm NH 3. d 150 mm urea, 100 mm CO 2 and 150 mm NH 3, and 75 mm NH 4 +.

22 Time t dwell Direction [ns] [ns] [p c] Supplementary Table S2. Summary of the spontaneous urea conduction events of the 7 μs HpUreI monomer simulation. The average time a conducting urea molecule spends inside the channel (t dwell ) is 7±5 ns. Direction of urea molecule: 1 = from the periplasm to the cytoplasm, 0 = from the cytoplasm to the periplasm.

23 Residue Position pka State ASP 25 cytoplasmic loop 4.02 charged ASP 100 cytoplasmic loop 1.47 charged ASP 126 periplasmic loop 3.30 charged ASP 129 periplasmic loop 3.80 charged ASP 130 periplasmic loop 4.19 charged ASP 140 periplasmic loop 4.99 charged GLU 60 periplasmic loop - charged GLU 63 periplasmic loop - charged GLU 138 periplasmic loop 4.25 charged GLU 159 cytoplasmic 2.74 charged GLU 177 channel/buried 6.47 neutral HIS 54 periplasmic loop 6.43 charged HIS 70 periplasmic loop - charged HIS 71 periplasmic loop - charged HIS 95 cytoplasmic 3.02 neutral HIS 123 periplasmic 6.75 charged HIS 131 periplasmic loop 6.52 charged HIS 193 periplasmic 6.74 charged Termini cytoplasmic/buried charged Supplementary Table S3. Protonation state of residues in the molecular dynamics simulations. The ph was chosen to be identical to the crystallization conditions (ph = 5.1), where the channel is open. Solution side chain pka values are pka Asp = 3.9, pka Glu = 4.3, pka Arg = 12.0, pka Lys = 10.5, pka His = 6.1, pka Cys = 8.3, pka Tyr = 10.1, and the carboxyl ( COO ) and amino ( NH 3 + ) termini are charged for 2.2 < ph < 9.4. Periplasmic loop residues (PL1: residues 55-74, PL2: residues ) not resolved in the crystal structure were chosen to have solution pka values.

24 System Solute Length [C] NaCl ΔG barrier Comment [ns] [mm] [kcal/mol] Mono2 urea 55 x not converged Mono2 thiourea 55 x not converged Mono2 H 3 O + 55 x not converged POPC urea 55 x ± 0.2 Supp. Fig. S19 Supplementary Table S4. List of potential of mean force (PMF) simulations. soft position restraints (20 kj/mol/nm 2 ) on transmembrane helix backbone atoms only (C α, C, O, & N), loops are flexible. T = 37 C. Forcefield = OPLS-AA.

25 Supplementary Discussion Monomer simulations with closed periplasmic constriction Simulations of single protomers were initially carried out by restraining to the starting configuration of protomer F after opening of the periplasmic constriction C P. This configuration, which was taken from the hexamer simulation (t = 480 ns), proved to conduct urea at high efficiency (Supplementary Fig. S4). Control simulations were carried out with protomer structures restrained to the initial crystal structure coordinates, where C P was closed. For the first simulation restraints of 1 kcal/mol/å 2 were applied. This prevents the channel from collapsing (Supplementary Fig. S2), but also prevented the opening of the periplasmic constriction C P, with no urea transfer observed over the 1 µs simulation time (Supplementary Fig. S10). Comparison with the urea motion in the channel with open C P (Supplementary Fig. S4) reveals that urea does not enter the periplasmic vestibule of the channel. C P remains closed even if the restraints are reduced to 0.1 kcal/mol/å 2 (Supplementary Fig. S10). Further softening of the restraining potential results in partial collapse of the channel, as observed for the unrestrained monomer (Supplementary Fig. S2). These simulations suggests that the opening of the C P is essential for allowing urea to enter the periplasmic vestibule and conduct through the channel pore. Comparison of monomer and hexamer conduction Comparison of urea conduction through the monomeric and hexameric protomers showed similar behavior for urea molecules during transit as well as for the key channel lining residues Phe84, Tyr88, Trp149, and Trp153. Phe84, Tyr88, and Trp153 are generally restricted to two orientations, while Trp149 alternates between four configurations (Figure 4A), all of which are observed in both the monomeric and hexameric protomer simulations. Motion of Trp153, which is essential for urea conduction, closely follows the motion observed in the hexamer (see movies). Comparison of the motion of urea between both simulations is difficult since urea is contained in a hydration shell for most of the time and highly mobile inside the wide section of

26 the pore cavity (Supplementary Fig. S6). Comparison of the free energy of urea transport derived from the urea pore occupancy of protomer F of the hexamer (two conduction events) with the single protomer in the monomer simulation (ten conduction events) is shown in Supplementary Fig. S7. While the figure shows some differences, particularly in the barrier height of the periplasmic constriction, both systems capture the double barrier profile of the channel. The overall agreement of the profiles is reasonable, given the seven times shorter simulation timescale and lower number of conduction events of the hexamer system, suggesting that the simulations sample the same principal urea translocation pathway through the pore. The key difference between the simulations is an increased rate of water transport observed for the monomeric system, which is due to differences in the periplasmic loop structures and might also be influenced by the force field employed (OPLS-AA for the hexamer and CHARMM27 for the monomer). Mutant simulations Urea flux through HpUreI is too low for reliable determination of selectivity in wild-type and mutant channels from microsecond equilibrium simulations, which are the current de facto simulation limit for complex membrane protein systems of this size. Nevertheless, the fact that water flux is much higher than urea flux, and urea passes through the channel in the presence of water suggests that urea conductivity is likely to be strongly correlated with water flux. This means that a change in water conductivity for a given mutation will likely reflect a corresponding change in urea flux. Since water flux through HpUreI can be accurately determined via long-timescale equilibrium simulations, we performed a microsecond simulation of a HpUreI hexamer with each protomer carrying a different single mutation for which the conduction rate and urea/thiourea selectivity ratios are known experimentally (Supplementary Fig. S11) 5. Measurement of the water flux and structural analysis of the channel structure shows that F84L created a hydrophobic plug that disrupts the water channel at C P, explaining why the mutant does not conduct. W149Y was found to orient itself horizontally across the pore, blocking the pathway in agreement with experimental flux measurements. Why the W149F

27 mutation was not functional is not clear, this system will need longer simulations. Y88F was found to inhibit conduction by dehydrating C P, again in agreement with experimental data. The W153 mutants are the most interesting as they are functional but modulate selectivity: In the simulations W153A completely opens both constrictions resulting in a continuous water filled channel where urea can enter deeply (Supplementary Fig. S11). This suggests that the experimentally observed loss of selectivity is due to widening of water channel and the fact that Ala is too small to pin urea against the hydrophobic wall of Leu6, Leu9, and Leu13. Experimentally, W153F also showed decreased selectivity, which can be attributed to the widening of C C observed in the simulations. However, C P is still partially constricted for this mutant, potentially explaining the lower flux and higher selectivity compared to W153A.

28 Supplementary References 52 Kozakov, D. et al. Achieving reliability and high accuracy in automated protein docking: ClusPro, PIPER, SDU, and stability analysis in CAPRI rounds Proteins 78, , doi: /prot (2010). 53 Johnson, G. T., Autin, L., Goodsell, D. S., Sanner, M. F. & Olson, A. J. epmv embeds molecular modeling into professional animation software environments. Structure 19, , doi: /j.str (2011).