Human RhCG Ammonia Conduction Mechanism

The structural differences between E. coli AmtB and the human RhCG as well as the Rh50 from Nitrosomonas europaea suggest different ammonia conduction mechanisms for the AmtB and the Rh proteins. This study investi-gates the mechanism differences by using molecular dynamics simulations on RhCG and Rh50NE structures. Unlike AmtB the Rh proteins lack the aromatic cage at the bottom of the periplasmic vestibule. This report establishes the periplasmic Glu residue as the 4 NH + binding site for Rh proteins, and the two Phe residues at the entrance of the pore as the NH 3 binding site. The 4 NH + molecule pushed by another ammonium releases one of its protons on its way to the phenyl gate. This study also discovers that, unlike AmtB, the Rh protein pores allow water molecules, which eventually facilitates the NH 3 conduction from periplasm to cytoplasm.

in the basolateral and apical membranes; RhBG has not yet been detected in those areas [7].
The previous computational studies for the AmtB mechanism [8] showed that the ammonium initially comes in contact with the charged residues at the periplasmic loops of Escherichia coli. Then, 4 NH + follows the negatively charged oxygen atoms of the side chain of the carboxyl groups and arrives at the aromatic matic cage, which is composecage at the bottom of the periplasmic vestibule. This arod of F103, F107 and W148, stabilizes the charged substrate at the entrance of the hydrophobic pore. AmtB from Escherichia coli has two tightly stacked phenylalanine rings at the periplasmic entrance of the pore. The synchronized opening and closing of these two phenylalanine rings allows charged ammonium to enter the pore and release its proton to the bulk of water, becoming a neutral NH 3 in the transfer [9].
Many studies have been done to decipher if ammonia is transported into the cell by RhCG as 4 NH + or as NH 3 . Measurement of apical plasma membrane 4 NH + and NH 3 permeability in the collecting duct of mice with global Rhcg deletion found decreased NH 3 permeability without any change in 4 NH + permeability [10]. Other studies reconstituted purified human RhCG into liposomes and demonstrated increased NH 3 permeability but no differences in 4 NH + permeability [5] [11]. Thus, the bulk of evidence suggests Rhcg/RhCG functions as a facilitated NH 3 transporter [12] [13].
The structural differences between AmtB and Rh proteins suggest a variation on this Am conduction mechanism [5]. First of all, the RhCG loops are bigger and contain more negatively charged residues, which could possibly yield better conduction rate. In addition, the W148 is replaced by L170, and F103 is completely missing as it does not have comparable residue in RhCG. The absence of an aromatic cage at the bottom of the vestibule raises the question of recruitment site. At the bottom of the periplasmic vestibule, A144 of AmtB is replaced by a conserved E166 in RhCG, which looks like a prime candidate for a charged substrate recruitment site. The crystal structures of Rh50 from Nitrosomonas europaea and human RhCG also show that the two PHE rings at the periplasmic vestibule exit are not stacked in Rh proteins (see Figure 1). Additionally, the cytoplasmic exit of the pore is not guarded by a PHE ring as in E. Coli AmtB; Human RhCG and Rh50NE have LEU and CYS residues instead. These structural differences at the cytoplasmic exits may result in differences in water's ability to access the central pores of AmtB and Rh proteins.
While this study focuses mainly on RhCG, we also included results from the simulations for Rh50NE to try and investigate general ammonia conduction mechanisms in Rh proteins.

No Synchronized Phenyl Motion for Rh Proteins
Throughout the known structure of the Amt/Mep/Rh protein family the periplasmic vestibule is separated from the pore with two phenylalanine residues. For the Amt protein, the phenylalanine residues are stacked parallel on top of each other and completely cover the channel. Rh proteins, on the other hand, have phenylalanines that create a right angle with each other with one in a vertical position while the other lies horizontal. In their experimental study, Javelle et.al found out that mutating both phenyls kills the function of AmtB [8]. On the other hand, for human RhCG, single mutation of the phenyl pair substantially stops the NH 3 conduction, but the double mutation keeps the channel almost as active as the WT [15]. The difference between these experimental studies suggests that the role of the phenyl gate is different between Rh proteins and AmtB.
The conserved twin phenyl residues that keep the gate of the hydrophobic pore closed are observed to flip open in a synchronized manner for E. Coli AmtB [9]. Additionally, the dihedral angle Ca-Cb-Cg-Cd1 distribution shows two energetically favorable configurations of AmtB phenyls. The π-π interaction between the phenyl side chains allows these two stable cases [9]. AmtB.
Our analysis on dihedral angles Ca-Cb-Cg-Cd1 shows that when there is no substrate in the vicinity of the phenyl gate, the F130 takes dihedral angles of 16 ± 23 or −170 ± 14 while F235 takes 87 ± 14 or −93 ± 12. These numbers represent the mean and the sigma of the Gaussian fit. Having exactly 180 degrees between these angles clearly shows that there is an only one energy minimum for phenyl pairs on human RhCG in the absence of substrate. However, the distribution of the same dihedral angles with the presence of

Water Leaks into the Central Pore of Rh Proteins
The crystal structures of Rh50NE [4] [14] (1.3 Å resolution) and human RhCG (2.1 Å resolution) have some distinct pore differences when compared to E. coli AmtB. The conserved phenyl rings at the pore entrance from the periplasmic vestibule are not stacked as tightly as in AmtB. The cytoplasmic exits of the pores on these structures do not have a phenyl gate either. This allows water molecules to have hydrogen bonds with central histidine residues, thus shortening the length of the hydrophobic pore to 11 Å. By contrast, the distance between the water molecules on both sides of the pore for AmtB is more than 20 Å. During long equilibrium MD simulations (equivalent of 540 ns simulations for a monomer) on a trimeric Rh50NE model, water molecules were observed to diffuse through pore in both directions. Although the F194 appears to keep the periplasmic entrance of Rh50NE pore closed, there is still enough space for water molecules to diffuse through. Throughout the simulations, a total of 16 water molecules leaked into the pore. Four of these water molecules diffused from the cytoplasmic vestibule to the periplasmic vestibule, and four others diffused in the opposite direction. The remaining 8 water molecules were observed to leak into pore from cytoplasm; after spending some time in pore they moved back to the cytoplasmic vestibule. The average time it took for a water molecule to diffuse from periplasm to cytoplasm was measured to be 2.4 ns. The cytoplasmic entrance of the pore is even less protected for Rh50NE proteins. The four water molecules conducted to the periplasmic vestibule of Rh50NE from cytoplasmic vestibule in average of 1.1 ns. However, the rest were seen going back to the cytoplasmic vestibule after spending an average of 1.5 ns within the pore. On two different occasions two water molecules entered together to Rh50NE pore.
The crystal structure of Rh50NE shows that the bottom of the periplasmic vestibule is divided into two parts by the I198 and F86. The narrow pathway created by A195, T196, A197, and I198 allows water molecules to create a water bridge between the pore entrance and the periplasmic vestibule. The water conduction from periplasm to pore is mediated through this narrow pore. Otherwise the water molecules located at the entrance of the pore have not been observed to be accessible from the F86 side of the periplasmic vestibule. Chérif-Zahar et al. [16] found that Rh50NE regulates NH 3 uptake, but that it appears to lack the primary proposed ammonium binding site found in AmtB and thus seems to function better in a higher pH environment. A computational study by Hub et al. [17] also reports that the Rh50NE should be partially active and suggests that its water conduction rate is 40 times slower than its conduction rate of NH 3  In regards to the human RhCG model, Baday et al. [18] believed the pore to be hydrophobic because of a conserved phenylalanine. Our simulations show that, although the water molecules can enter the channel, the human RhCG does not conduct water molecules. After approximately 450 ns of monomeric MD simulations performed with a trimeric human RhCG model, we observed 23 water molecules entering the pore from the cytoplasmic vestibule. All of these water molecules leaked into the pore from the cytoplasmic vestibule and returned back to cytoplasm. During the water leakages to the pore, the water molecule(s) tended to stay in the pore for 5 ns or less ( Figure 7). As is shown by the crystal structure, there is a water molecule at the top of the pore that has hydrogen bonds with H185, G179, G290 and N236 and does not leave. Unlike Rh50NE, this water molecule positioned at the pore and periplasm interface does not have any direct connection to the bulk of the water in the periplasmic vestibule. However, the cytoplasmic entrance of the human RhCG pore allows water molecules to leak through more freely. On many occasions during 450 ns monomeric MD simulations, two water molecules were observed to enter the pore together. In these cases, the water molecules were more stable inside the pore. Interestingly, in one occasion we observed a transiently stable chain of water molecules flooding the pore from cytoplasm all the way to F235 (Figure 8).

Conclusions and Discussion
In an attempt to better understand the mechanism of transport in RhCG transporters, we performed various simulations that affected different molecules in the system to better identify how they interact. Although the transport mechanism is still a mystery, our simulations suggest some viable theories as to what the process is. Because 4 NH + is attracted to E166 and NH 3 is attracted to the phenylalanines, it is logical to assume that there is a place of deprotonation somewhere between the two points. Another  . Water chain observed in the periplasmic pore of RHCG. The water molecules enter from cytoplasm but they cannot pass to periplasm. W490 was detected at this location on the x-ray structure of the protein [5].
molecule has been observed to come very close to the 4 NH + molecule already attracted to the E166. Since there are no proton acceptors around it, this suggests the deprotonation happens before the rings, possibly between E166 and the phenyl gate. The new molecule may serve as a catalyst to kick out the already attracted 4 NH + molecule, causing it to be forced over the energy barrier. Because of the large barrier of around 4 kcal/mol beyond E166, it is reasonable to believe that the 4 NH + molecule loses a proton before making the jump, considering it would be more energetically favorable to break through the barrier in an uncharged state. It would be very difficult for it to get over the barrier without doing so. After it loses the proton and becomes NH 3 , it is attracted to the phenylalanines at the top of the pore.
The phenylalanines are constantly moving and turning in the simulations. This might give the NH 3 molecule the opportunity to slip into the channel and navigate to the pocket with the structural water molecule. Once in the channel, we know that the NH 3 will act as the water does because it is a similar size, and will jump down past the histidines into the periplasm (Figure 9). During our simulations we substituted a NH 3 molecule entering the pore in place of the second water molecule, and observed that the NH 3 acted similarly to water and had jumped down the pore to the cytoplasmic side.
Our simulations have shown that the pore of RhCG is much less hydrophobic than that of AmtB. During our long equilibrium MD simulations (the equivalent of 740 ns for a single monomer) in an earlier study on trimeric AmtB, the author of this study reported only one water molecule entering the hydrophobic pore from cytoplasmic vestibule [9], unlike another study reporting hydrophilic AmtB pore [19]. In this study, we found that both the Rh50NE and human RhCG crystal structures had water molecules at the cytoplasmic exits of the pore. Furthermore, Zidi-Yahiaoui et al. report water conduction by human RhCG in HEK293E cells [14]. The reported conduction rate is ten times less than AQP1 and does not change with F130A/F235V double mutation. However, Gruswitz  These results are in agreement with Gruswitz et al. [5].
Further simulations and investigation will be necessary to illustrate the complete transport mechanism in RhCG. Nonetheless, our simulations suggest that 4 NH + is brought in and is converted into NH 3 , which then slips into the pore and diffuses in the same way the water molecules do. Figure 10 summarizes our proposed mechanism of human RhCG ammonia conduction.  NH + reaches the energy barrier, sheds a proton to become NH 3 and settles into the phenyl rings at its energy minimum; (d) As the rings fluctuate, the NH 3 works its way into the pore, and settles near the water shown in the crystal structure in the mildly unstable pocket; (e) NH 3 acts in a similar manner to water, and rushes out of the pore down the channel; (f) NH 3 is flushed out of the channel.
Hydrogen atoms were added to the structures with XPLOR-NIH [22].  [27] were used to keep the temperature and pressure constant at 310 K and 1 atm, respectively. The Van der Waals interaction cutoff was set at 12 Å in conjunction with switching at 8 Å. Full electrostatics was employed using the Particle Mesh Ewald (PME) method [28], with a grid spacing less than 1 Å. All simulations used an integration time step of 1.0 fs. The total charge of the system was neutralized by adding Cl − and Na + ions.
Equilibrium molecular dynamic simulations. To investigate the water diffusion into protein vestibules, separate initial models were built for each protein.
Before the long MD simulations, each system was first minimized for 2000 steps and then equilibrated for 2 ns with no constraints on the system. Total of 246 ns, 166 ns, and 164 ns of MD simulations were performed on AmtB, Rh50NE, and RhCG, respectively.
To determine how water diffuses into these proteins, the probability distribu-