|Home | About | Journals | Submit | Contact Us | Français|
Protein dynamics is essential for virtually all protein functions, certainly for gating mechanisms of ion channels and regulation of enzyme catalysis. Ion channels usually feature a gate in the channel pore that prevents ion permeation in the closed state. Some bifunctional enzymes with two distant active sites use a tunnel to transport intermediate products; a gate can help prevent premature leakage. Enzymes with a buried active site also require a tunnel for substrate entrance; a gate along the tunnel can contribute to selectivity. The gates in these different contexts show distinct characteristics in sequence, structure, and dynamics, but they also have common features. In particular, aromatic residues often appear to serve as gates, probably due to their ability, through side chain rotation, to effect large changes in cross section.
Ion channels can be viewed as the gateways into cells. A typical ion channel shuttles between an open state and a closed state; the conformational transition between open and closed states is referred to as gating. Channel opening is triggered by a specific stimulus such as the voltage or pH difference across the cell membrane or the binding of a ligand [1–5]; upon opening the channel is selective in the type(s) of ions that can pass through its pore. In the closed state the channel must be tightly shut; ion permeation is prevented by energy barriers along the pore which arise from physical occlusion or the hydrophobic nature of residues lining the pore. The barrier region is appropriately referred to as the gate. Notably, amino acid substitutions in this region can lead to abnormal channel kinetics .
A different kind of channel exists in bifunctional enzymes in which the product in one active site is the substrate of a distant, second active site. Although the intermediate product can, in principle, diffuse through the bulk solution to the final destination, a tunnel through the interior provides a direct passage in a large number of enzymes [7, 8]. Here a gate might serve to prevent the passage of unintended substrates. Upon forming the intermediate product, the gate might open to allow its passage.
In some monofunctional enzymes the active sites are buried [9–12]. A tunnel then provides an entrance for the substrate and an exit for the product. A gate along the tunnel could contribute to substrate selectivity .
The gates in these different contexts form the focus of the present review. A main aim is to expose ion channel scientists to the knowledge of gates in enzymes, and vice versa. Although gates and gating mechanisms of different ion channels show distinct characteristics, they share certain common features. Some of these features also appear to be shared by enzymes which contain intramolecular tunnels.
Potassium (K+) channels are among the best studied ion channels. Structure determination and functional studies of mutant proteins have contributed significantly to the identification of channel gates and the motions leading to channel opening. The pore of a K+ channel, formed at the center of four identical subunits, is lined by the selectivity filter on the extracellular half and the inner helix on the intracellular half (Figure 1). It is generally accepted that the selectivity filter moves little during channel gating [14, 15]. The gate therefore resides in the intracellular half of the inner helix.
At least three modes of gating motions have been proposed. The first consist of rigid-body motions of the pore-lining helices. For example, it was initially proposed that  cyclic nucleotide-gated channel 1 (CNG1) opens as the intracellular end of each inner helix rotates around the pore axis while the extracellular end remains fixed (Figure 2a). The rotation opens the channel by widening the pore at the crossing of the inner helices, which occurs within the intracellular half of the inner helix. The inner-helix crossing thus forms the gate here. Similarly, the gating motions of the mechanosensitive channel of large conductance (MscL) are now thought to involve a rigid-body tilt of the pore-lining M1 helices away from the pore axis [17–19].
In the second mode (Figure 2b), the inner helices are, upon activation, assumed to kink at a hinge around the middle; again the inner-helix crossing in the closed state forms the gate. By comparing the structures of the closed KcsA K+ channel and the open MthK K+ channel (Figure 1a, b), Jiang et al.  identified a kink around a Gly residue in the open state. This Gly is highly conserved (Figure 1d), suggesting that helix kinking is a common gating motion among many K+ and related channels. In particular, substitution of the corresponding Gly (Gly466) into Ala renders the Shaker K+ channel non-functional, whereas an additional substitution, of the next residue, Val467, into Gly, restores channel function . (Another hinge point, within the Pro-Val-Pro sequence further toward the intracellular side, has also been proposed .) The helix-kinking mode was also prescribed to G protein-sensitive inwardly rectifying potassium (GIRK; also known as Kir3) channels [22, 23] and later to CNG1 . The helix-kinking motion has been captured by recent X-ray structures of the open and closed forms of the nonselective cation channel NaK [15, 25].
The third mode of gating motions adds a new element to the second mode: an aromatic side chain within the intracellular half of the inner helix projects into the channel pore to seal it off in the closed state. In the open state, the inner helix kinks and the aromatic side chain simultaneously retracts to widen the pore. The role of such an activation gate was assigned to Phe146 in KirBac1.1  and the corresponding Tyr132 in KirBac3.1 . The corresponding residue, Phe181, in a variant of mouse Kir3.1 occludes the channel pore in the closed structure (in this variant, the sequence of positions 83–177 was replaced by that of KirBac1.3) . Two aromatic side chains, Phe203 and Tyr215, seem to play a similar role in the bacterial MlotiK1 channel; Ala substitutions of these residues led to significant increases in the rate of ion uptake . Interestingly, whereas the residues around the inner-helix crossing, notably Gln103, form the gate in the closed state of NaK , the side chain of Phe92 projects into the pore to form a new constriction in the open state; Ala substitution of this residue also caused an increase in ion-conductance rate .
It is worth noting that the third mode of gating motions is very similar to one recently proposed for the M2 proton channel of influenza A virus , which is also homo-tetrameric and is activated by low pH on the extracellular side. Structure determination in the past decade for this protein [2, 30, 31] has not provided a mechanism of channel activation. Molecular dynamics simulations, however, led to the proposal that Gly34 serves as a hinge for helix kinking, and the side chain of Trp41 provides an aromatic gate [29, see also 32]. Uptake of a proton from the extracellular side by His37 increases the propensity of helix kinking, which in turn leads to a reorientation of the Trp41 indole from being perpendicular to the pore axis to being parallel. These conformational changes poise the protonated His37 to release its proton to the intracellular side.
Recent structure determination of a bacterial pentameric ligand-gated ion channel in the open state [4, 5] along with an earlier structure of a related channel in the closed state  has provided molecular details of the gating motions of this family of channels which includes acetylcholine receptors. In these pentametic channels, each subunit has four transmembrane helices, and the channel pore is lined by the second (referred to as M2) of these helices. The membrane-spanning residues of the M2 helix are assigned positions −2′ to 20′, going from the intracellular side to the extracellular side. Relative to the closed state, the M2 helix in the open state is tilted around an axis parallel to the membrane plane. This tilt widens the extracellular end of the pore, which is constricted by three rings of hydrophobic residues (at positions 9′, 13′, and 16′) in the closed state, and slightly narrows the intracellular end of the pore. The gating motions shown by comparing the structures of the open and closed states were, in part, suggested by earlier computational studies. In particular, Yi et al.  proposed the tilt of the M2 helix as the gating motion of the α7 acetylcholine receptor. This motion was proposed to be produced by a pull of the M1 helix by the β10 strand of the ligand-binding domain, whose propensity for an upward motion was suggested to be increased by agonist binding. However, this putative role of the M1 helix does not seem to be supported by the structural data, which suggests that the M2 helix is fixed during gating. At the quaternary structure level, the five subunits of the transmembrane domain undergo a clockwise twist in the open state relative to the closed state . Such a quaternary twist was predicted by normal mode analysis [35, 36].
The three rings of hydrophobic residues comprise the gate , in the sense that they constitute the narrowest part of the closed channel pore. A number of amino acid substitutions which lead to disorders in neuromuscular transmission have been found within or bordering the hydrophobic rings .
Some enzymes catalyze two ordered reactions at two separate active sites, with a product from the first reaction becoming a substrate in the second reaction. For such bifunctional enzymes, efficient transfer of the intermediate product from one active site to another is critical. In some bifunctional enzymes, such transfer occurs via a tunnel which connects the two active sites [7, 8, 37, 38]. The situation then is similar to that of an ion channel: here the two active sites are similar to the two sides of the cell membrane; the intermediate product is like the permeating ion; and the tunnel is the channel pore. The tunnel might initially be closed to prevent the passage of unintended substrates. Upon completion of the first reaction, the tunnel is opened and the intermediate product is then transferred to the second active site, where the second reaction can finally proceed. In addition to its efficient transfer, sequestration within a tunnel provides a means of preventing a potentially toxic intermediate product from being released to the solvent .
The role of the tunnel is well illustrated by glucosamine-6-phosphate synthase, which catalyzes the conversion of D-fructose-6-phosphate (Fru6P) into D-glucosamine-6-phosphate using L-glutamine as a nitrogen source (Figure 3a) . The functional unit of this enzyme is a dimer; each subunit consists of two domains (consisting of residues 1–239 and residues 249–608, respectively) (Figure 3b). The two active sites of this bifunctional enzyme are located on the N-terminal and C-terminal domains, respectively, and connected by a 18-Å tunnel (Figure 3c). At the N active site, the side chain of L-glutamine is hydrolyzed to produce L-glutamate and ammonia. The ammonia produced is then transferred through the tunnel to the C active site, where it is incorporated into Fru6P to produce the final product, D-glucosamine-6-phosphate.
The first step of the two coupled enzymatic reactions is the binding of Fru6P to the C active site. The structure of the enzyme bound with Fru6P was determined by Mouilleron et al. . They also determined a second structure, in which glucose-6-phosphate (a Fru6P analog) is bound at the C active site and 6-diazo-5-oxo-L-norleucine is bound at the N active site. The bound 6-diazo-5-oxo-L-norleucine mimics the intermediate produced at the N active site immediately after the production of ammonia. A comparison of these two structures shows how the tunnel is reorganized and opened by glutamine binding and the subsequent reaction at the N active site. The walls of the tunnel are mainly formed by Trp74 and residues 601–607 of the C-terminal tail. Upon binding 6-diazo-5-oxo-L-norleucine, the Q-loop (consisting of residues 73–81) closes the N active site. Importantly, in the Fru6P-bound structure, the Trp74 side chain seals off the tunnel, possibly explaining why this enzyme cannot use exogenous ammonia as a nitrogen source . Along with the closure of the Q-loop, the χ1 torsion angle of Trp74 changes by 75°, leading to the opening of the tunnel.
Another bifunctional enzyme, tryptophan synthase, has an α2β2 subunit composition . The α subunit catalyzes the cleavage of indole 3-glycerol phosphate into indole and D-glyceraldehyde-3-phosphate, and the β subunit catalyzes the condensation of indole and L-serine into L-tryptophan (with a water molecule as a byproduct). The quaternary arrangement of tryptophan synthase shares some similarity with that of glucosamine-6-phosphate synthase (Figure 3b). Whereas the two C-terminal domains of glucosamine-6-phosphate synthase are stacked side-by-side, the two β subunits of tryptophan synthase are stacked back-to-back. The two α subunits, like the two N-terminal domains of glucosamine-6-phosphate synthase, are located at the opposite ends. The two active sites are connected by a 25-Å tunnel, which provides a passage way for indole, the intermediate product. Midway along the tunnel, the walls are lined in part by Tyr279 and Phe280 of the β subunit, which adopt alternative conformations, resulting in an equilibrium between open and closed forms of the tunnel . There is evidence that substrate binding to the α subunit leads to an increase in the closed form of the tunnel by altering the conformational preferences of these residues. Upon cleavage of the substrate, the equilibrium between open and closed forms is presumably restored, so that the transfer of indole to the β subunit can proceed.
The active sites in some monofunctional enzymes are buried; a tunnel fluctuating between alternative conformations can provide transient substrate access and product release. The burial of active sites might prevent the catalyzed reactions, such as hydrolysis, of inappropriate substrates [13, 42]. In addition, transfer of a substrate, typically a small molecule such as oxygen, through a tunnel could afford temporal and regiocontrol of its reaction with a larger substrate. The monooxygenase ActVA-Orf6, which catalyzes the oxidation of a phenolic compound, 6-deoxydihydrokalafungin, at the C-6 position into the corresponding quinone, dihydrokalafungin, provides an example of regiocontrol . The enzyme-bound 6-deoxydihydrokalafungin is well shielded from the solvent except for the reactive C-6 position. Oxygen can access this carbon through a narrow tunnel gated by a mobile loop consisting of residues 34–48.
Cholesterol oxidases contain tunnels, which allow for both oxygen access and hydrogen peroxide release. These enzymes are classified according to whether the cofactor flavin adenine dinucleotide (FAD) is covalently linked to the enzyme. In the type-I cholesterol oxidase of Streptomyces sp. SA-COO, the side chains of a large number of residues adopt two sets of alternative conformations, one of which yields an open tunnel connecting the active site to the exterior . Among the residues lining the tunnel entrance is Phe359; in the F359W mutant, the larger aromatic side chain is observed only in the closed conformation (whereas other tunnel-lining residues are found in two alternative conformations) . This observation shows that the equilibrium between open and closed forms of the tunnel can be easily perturbed by the size of key residues. In the type-II cholesterol oxidase of Brevibacterium sterolicum, only three tunnel-lining residues (Ile423, Glu475, and Arg477) display alternative conformations . The open conformation of Arg477 is stabilized by a salt bridge with Glu311; a Glu311 substitution seems to disrupt the open tunnel, altering oxygen binding kinetics .
In both types of cholesterol oxidases, cholesterol binding induces tunnel closure [11, 12]. The coupled oxidation of cholesterol and reduction of FAD (into FADH−) presumably triggers the re-opening of the tunnel, allowing the access of oxygen to oxidize FADH− back to FAD (the byproduct hydrogen peroxide is likely released through the same tunnel). Therefore the substrate, oxygen in the present case, reaches the active site through the tunnel at the desired time, illustrating the principle of temporal control.
Yet another potential advantage of gated access to a buried active site, which we first proposed , is substrate selectivity. In acetylcholinesterase (AChE), a 20-Å tunnel leading to the active site is gated by a ring of aromatic side chains [9, 13, 46]. The fluctuations of these side chains are such that the tunnel remains blocked for the substrate, acetylcholine (ACh), most of the time, but occasionally opens wide enough for its passage (Figure 4). Remarkably, however, the catalytic efficiency of this enzyme for ACh (kcat/KM ~ 109 M−1s−1) is among the highest. This apparent conflict was resolved within a theory for the substrate binding kinetics that accounts for both the gating dynamics of the tunnel and the diffusion of the substrate toward the active site . This theory predicts that when the switch between the closed and open forms of the tunnel, e.g., rotations of the aromatic side chains in the case of AChE, is sufficiently fast (relative to the diffusion of the substrate), substrate binding through the gated tunnel is effectively unhindered even when the tunnel remains closed most of the time. Such a scenario is exactly what was suggested by molecular dynamics simulations of AChE for the passage of ACh [13, 46].
By contrast, for slightly larger substrates such as butyrylcholine (BCh; which harbors two additional methylene groups), the open fraction of the tunnel is substantially reduced; the effectively unhindered scenario would only occur in times scales of the gating dynamics that are too short to be physically possible. As a result, the binding rate of BCh to the AChE active site is predicted to be orders-of-magnitude lower than that of ACh, in agreement with the observed disparity in catalytic efficiency between these two substrates [48, 49]. The dynamics of the gating aromatic side chains thus affords AChE substrate selectivity (ACh over BCh) without sacrificing the catalytic efficiency of the intended substrate, ACh. The role of the aromatic side chains in guiding substrate selectivity is further supported by observed increase in catalytic efficiency for BCh when some were substituted with smaller side chains [48–50]. A gated tunnel to the active site of catalase, which converts two hydrogen peroxide molecules into one oxygen and two water molecules, has also been suggested to afford similar substrate selectivity .
Ho recently presented an interesting discussion of the diversity of roles of intramolecular transfer of substrates and products . This discussion focuses on photosystem II, the large, transmembrane multisubunit enzyme that converts water to oxygen molecules, thus providing protons for ATP production and electrons for photosynthesis. This process requires the delivery of water molecules to the oxygen-generating center, located at the heart of the enzyme, and the directional delivery of protons from this center to the appropriate side of the thylakoid membrane to maintain the proton gradient needed for ATP synthesis. The resulting oxygen molecules must also be released from the center in a controlled fashion to avoid harmful side-reactions. Each of these transfer processes is thought to involve a specific intramolecular tunnel. Particularly interesting in the present context is the suggested tunnel for water molecules, which appears to be gated in part by aromatic residues, Tyr161, Phe186, and His190. A possible function of the gate is to moderate the flow of water molecules to the oxygen-generating center. Experimental evidence suggests that excessively rapid delivery of water molecules to this center can ‘short-circuit’ the catalytic process, leading to the generation of harmful species such as hydrogen peroxide.
ATP-dependent proteases are responsible for degrading the majority of unwanted proteins in cells. The proteolytic active sites of HslVU, a prokaryotic homolog of the proteasome, are sequestered in the central chamber of HslV, which is a dodecamer of two back-to-back stacked hexameric rings . Protein substrates are unfolded and translocated to the proteolytic chamber through HslU, which forms a hexameric ring and binds to one or both ends of HslV. HslU is activated by ATP hydrolysis. Its central pore forms the translocation tunnel, which is gated by a highly conserved sequence, GVYG. Based on structures of HslU bound with different nucleotides, it was proposed that conformational changes induced by ATP hydrolysis are propagated to the gating sequence. In particular, the Tyr side chain undergoes a 180° change in direction, which is coupled to substrate unfolding and translocation. Substitutions of this residue, except with Phe, significantly impair the ability of HslVU to degrade proteins .
In the previous examples of transmembrane ion channels, intramolecular tunnels in bifunctional enzymes, and tunnels connecting active sites to the exterior in monofunctional enzymes, aromatic residues (phenylalanine, tyrosine, and tryptophan) often appear as gates. As we noted previously , “The aromatic rings are distinguished by their oblateness, so a small rotation leads to a large change in the width of the gate.” Such rotations have been demonstrated by alternative side-chain conformations in X-ray structures and observed in numerous molecular dynamics simulations. In particular, simple libration of the χ2 torsion angle yields a substantial change in the cross section for an aromatic side chain in a channel or tunnel. Libration of the χ1 torsion angle, or larger-scale collective motions, can swing the entire ring aside. All of these types of displacements are seen in the gating dynamics of acetylcholinesterase, for example, not only in the primary tunnel to the active site, but also on secondary pathways such as the ‘back door’ [46, 54, 55]. The ability of aromatic side chains to form cation–π and π– π interactions with ionic and aromatic groups of substrates has also been noted .
These types of motions can be driven by dynamic processes on different temporal/spatial scales, possibly with different functional consequences. Local dynamic processes include simple, transient packing defects that can drive rapid fluctuations of the aromatic ring orientations; these could be sufficient for opening and closing gates for the displacement of small molecules such as oxygen [13, 56, 57]. Coupling of the side chain motions to collective dynamics of the protein is likely to be involved in the gated transport of larger ligands [e.g., 52]. Such coupled local and collective motions might also provide a mechanism by which gated transport and ligand selectivity are mediated by the binding of effectors elsewhere in the protein , similar to what is seen in ligand-gated ion channels .
The gates of ion channels and of tunnels in enzymes are central to activation and ligand selectivity. Their composition ranges from a small number of side chains to segments of secondary structures. An important lesson, drawn from the study of a monofunctional enzyme with a tunnel leading to a buried active site, but applicable to gated tunnels and channels in general, is that both the magnitudes and the time scales of the motions of gates are essential for determining the efficiency of transport. Attention to such dynamic aspects will be essential for gaining molecular insight into gating mechanisms of ion channels and regulation of enzyme catalysis.
HXZ’s research was supported by NIH and JAM’s research was supported by NSF, NIH, HHMI, CTBP, and NBCR. We thank Sanbo Qin for preparing some of the figures.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Huan-Xiang Zhou, Department of Physics and Institute of Molecular Biophysics, Florida State University, Tallahassee, FL 32306, USA.
J. Andrew McCammon, Departments of Chemistry and Biochemistry and of Pharmacology, Center for Theoretical Biological Physics, and Howard Hughes Medical Institute, University of California at San Diego, La Jolla, CA 92093-0365, USA.