Conformational changes in Kir2.1 channels during NH4+-induced inactivation.

We have shown previously that NH(4)(+) binding to the external pore of a Kir2.1 channel induces channel inactivation possibly through conformational changes. In this study, we performed further biophysical analyses of the NH(4)(+)-induced inactivation modeled by a refined kinetic scheme. Also, we investigated the conformational change hypothesis by examining whether the chemical modification of single-cysteine substitution of amino acids located at the internal pore alters the kinetics of the NH(4)(+)-induced inactivation. In addition, we examined whether the mutation of amino acids located at various parts of a Kir2.1 channel influences the NH(4)(+)-induced inactivation. Kir2.1 channels were expressed in Xenopus oocytes and studied using patch-clamp techniques. The gating of the NH(4)(+)-induced inactivation was affected by mutation of several amino acids located at various regions of the Kir2.1 channel. These results suggest that amino acids from different parts of a Kir2.1 channel are involved in the channel closure. Furthermore, internal chemical modification of several cysteine mutants resulted in the block of inward currents and changes in the on and off rate for the NH(4)(+)-induced inactivation, suggesting that the internal pore mouth is involved in the closure of a Kir2.1 channel. Taken together these results provide new evidence for conformational changes affecting the NH(4)(+)-induced inactivation in the Kir2.1 channel.

Ion channels are membrane proteins that interact closely with permeant ions. Therefore, it is conceivable that the structures of these proteins may assume different conformations during different functional states, such as in the opening and closing of ion channels. In inward rectifier K ϩ channels, the most studied gating mechanism is the membrane voltage (V m ) 1dependent channel block by internal Mg 2ϩ and polyamines (1)(2)(3)(4). This V m -dependent block results in inward rectification, which contributes to the physiological functions of these channels. However, little is known whether structural changes are involved in the functions of inward rectifier K ϩ channels. It has been shown that the gating of these channels depends on per-meant ions (5,6), and most recent evidence supports that gating may be attributed to conformational changes resulting from the interaction between permeant ions and the backbone carbonyls in the selectivity filter of the cloned Kir2.1 channels (7). Our previous study demonstrated that external NH 4 ϩ induces the Kir2.1 channel into fast inactivation during hyperpolarization (8). We showed that the NH 4 ϩ -induced inactivation is not because of the NH 4 ϩ block of Kir2.1 channels. Furthermore, studies in the R148Y mutant suggest that one or both of the two binding sites located at the external pore mouth are involved in the NH 4 ϩ -induced inactivation. Because the binding site is located outside the electrical field (9), and yet the inactivation is V m -dependent, we propose that a V m -dependent process occurs within the pore to effect channel closure.
Even though evidence supports the involvement of conformational changes in the gating of the Kir2.1 channel, there remain doubts challenging the hypothesis because of the lack of an intrinsic V m sensor in these channels. In this study we further analyzed the biophysical properties of the NH 4 ϩ -induced inactivation. We found that its gating shares several similarities with that of a cloned Cl Ϫ channel, ClC-0. The gating in both channel types depends on V m , concentrations of permeant ions, and is described by a Boltzmann distribution with a non-zero offset (10 -12). It has also been demonstrated that an intrinsic V m sensor is not required in the gating of the ClC-0 channel. The V m dependence of gating arises from an intrinsically V m -dependent conformational change induced by the V m -independent binding of Cl Ϫ to the channel (12). Based on the model for the gating in the ClC-0 channel, we propose a non-equilibrium kinetic scheme to account for the NH 4 ϩ -induced inactivation in the Kir2.1 channel.
Also, we further suggest that if global conformational changes are indeed involved in the NH 4 ϩ -induced inactivation, several amino acids lining the pore of a Kir2.1 channel should participate in the V m -dependent process preceding to channel closure. We tested this hypothesis using two approaches. First, we examined whether the mutation of amino acids located at different parts of the Kir2.1 channel influences the gating of the NH 4 ϩ -induced inactivation. Second, we investigated whether the kinetics of the NH 4 ϩ -induced inactivation is altered by MTSET modification of cysteine mutants whose mutation is located at the internal pore. Our results show that the mutation of several amino acids located at different part of a Kir2.1 channel indeed changed the gating of the NH 4 ϩ -induced inactivation. In addition, chemical modification at the internal pore mouth reduced the kinetics of the NH 4 ϩ -induced inactivation. Kinetic studies have provided us with bountiful information on how the inward rectifier K ϩ channels operate to serve their functions. However, we have very little information on their underlying structures mainly because of technical limitations. The NH 4 ϩ -induced inactivation provides us with a model to study the structural-functional relationship within the Kir2.1 channels.

Molecular Biology and Preparation of Xenopus
Oocytes-Site-directed mutations were generated in the wild-type channel (IRK1 clone) using the Altered Sites II in vitro mutagenesis systems (Promega, Madison, WI). The cysteine mutants were constructed in the IRK1J clone. Purification of cDNA, and in vitro T7 or SP6 transcription reactions (mMessage mMachine; Ambion, Dallas, TX) were performed as described previously (13). Xenopus oocytes were isolated by partial ovariectomy from frogs anesthetized with 0.1% tricaine (3-aminobenzoic acid ethyl ester). The incision was sutured, and the animal was monitored during the recovery period before it was returned to its tank. Following the last oocyte collection, frogs were anesthetized with 0.1% of tricaine and sacrificed by decapitation. All surgical and anesthetic procedures conformed to national ethics committee guidelines.
The command V m and data acquisition functions were processed using a Pentium computer, a DigiData board, and pClamp6 software (Axon Instruments, Foster City, CA). Data sampling rates were 2.5-5 kHz, and the data were filtered at 0.5-1 kHz with an 8-pole low pass filter (Frequency Devices, Rochester, NY). In the experiments of V m -dependent inactivation, the holding potential was 0 mV, prepulses ranged from Ϫ200 to ϩ100 mV, and the test V m was Ϫ120 mV. Single-channel currents were recorded at Ϫ140 mV from a holding potential of 0 mV. The open and closed events obtained from voltage steps have exactly the same distributions as those obtained from the steady state (8,17). Capacitive currents were corrected using the built-in capacitance neutralization in the Axopatch 200A amplifier.
Data Analysis-Instantaneous currents were determined by fitting monoexponential functions to the currents at the test pulses and by extrapolating them to their beginnings. Histograms of the duration of time that channels remained open and closed were constructed with square root-log ordinates (13). The histograms were fit to monoexponential functions with the maximum log likelihood method, and, in general, biexponential functions did not provide significantly better fits (p Ͼ 0.05) than monoexponential functions, as judged by the maximal likelihood ratio test (4). Substates were observed, but they did not occur frequently. Transitions between the open state and substate, as well as between the closed state and substate, were not included in data analysis.
The time course of current inhibition of mutants by MTSET followed single exponential decay. Time constants for MTSET modification were obtained by fitting the time courses of current inhibition. The apparent second-order rate constants for MTSET modification were then calculated as the reciprocal of the respective time constants divided by the concentration of MTSET. Results are presented as mean Ϯ S.E.

Biophysical Properties of the NH 4
ϩ -induced Inactivation-Previously, we have examined NH 4 ϩ -induced inactivation with the voltage protocol described in Fig. 1A. In this study we analyzed further the gating properties of the NH 4 ϩ -induced inactivation. Fig. 1B illustrates the representative currents recorded in 10 and 100 mM symmetrical [NH 4 ϩ ], respectively. Inward currents inactivated during strong hyperpolarization and the rate of the inactivation was higher in 100 mM [NH 4 ϩ ] than that in 10 mM [NH 4 ϩ ]. Note that some residual capacitive currents were not corrected in the currents recorded in 10 mM [NH 4 ϩ ]. The degrees of inactivation, however, were actually the same for both 10 and 100 mM [NH 4 ϩ ] (see Fig. 2A). Instantaneous tail currents were recorded after prepulses to various test voltages. The steady-state open probability was quantified by normalizing these tail currents to the maximal one obtained following the most depolarizing prepulse potential (normalized I). Fig. 2A shows that normalized I was smaller at more negative V m . Changes in symmetrical [NH 4 ϩ ] did not affect the normalized I-V m relationship. The normalized I-V m relationship was fitted with a Boltzmann distribution containing a non-zero offset. The effective gating charge is around 0.7, and the non-zero offset is about 0.2 for all [NH 4 ϩ ] tested. Because the time course of inactivation could be fitted to a monoexponential function, the rate of inactivation was calculated as the reciprocal of the time constant (). The rate of inactivation depended on V m and [NH 4 ϩ ] (Fig. 2B). The single Kir2.1 channel demonstrates one open and one closed state in our previous study where we proposed that the ؉ through Kir2.1 channels. A, voltage protocol used to record the steady-state inactivation of currents through Kir2.1 channels. The holding potential was 0 mV, prepulses ranged from Ϫ200 to ϩ100 mV, and the test V m was Ϫ120 mV. B, current traces obtained from two different inside-out patches exposed to 10 and 100 mM symmetrical [NH 4 ϩ ] as indicated. The horizontal lines indicate zero current levels throughout this study. ϩ -induced inactivation could be described by Scheme 1 (8). We then calculated k on and k off from macroscopic currents recorded in various symmetrical [NH 4 ϩ ] using Equations 1 and 2, shown below, where f is the fraction of channels remaining open at steady state.
Fig. 2C demonstrates that k on increased exponentially with hyperpolarization with an effective gating charge Ϸ 0.4. The rate of inactivation increased with elevated [NH 4 ϩ ] and saturated in ϳ100 mM [NH 4 ϩ ], indicating that NH 4 ϩ binding is involved in the inactivation. Fig. 2D shows that the dose dependence of k on was similar at different V m . Because NH 4 ϩ is itself a permanent ion through the Kir2.1 channel, the entering and leaving of the ion at both sides of the membrane complicate the analysis of k on . However, the V m dependence of the k on is the same at both low and saturating [NH 4 ϩ ] levels, suggesting that the effective gating charge (0.4) is intrinsic to the on rate of the NH 4 ϩ -induced inactivation. On the other hand, k off did not show V m dependence (Fig. 2E). Increasing symmetrical [NH 4 ϩ ] also accelerated k off to the same degree as it did to k on (Fig. 2F). Because steady-state open probability is equal to k off /(k on ϩ k off ), the same [NH 4 ϩ ] dependence of k on and k off accounts for the same normalized I-V m relationships in various symmetrical [NH 4 ϩ ] shown in Fig. 2A. Fig. 2, D and F shows that the dependence of k on and k off on [NH 4 ϩ ] are about the same, and both are not V m -dependent, suggesting again that the NH 4 ϩ binding site affecting inactivation is located outside the electrical field (possibly at the external pore mouth according to our previous study (8)).
In summary, our results show that the NH 4 ϩ binding site affecting inactivation is located outside of the electrical field yet the inactivating process is V m -dependent. Also, k on and k off both depend on [NH 4 ϩ ]. Previously, we demonstrated that blocking rate would be too slow to account for NH 4 ϩ acting as a permeant blocker, and the inactivation is dependent on external rather than on internal NH 4 ϩ (8). Taken together our data suggest that the external NH 4 ϩ -induced inactivation is because of the conformational changes of the Kir2.1 channels. NH 4 ϩ -induced Inactivation Is Affected by Mutation of Amino Acids Ranging from the External to Internal Pore Mouth-To test whether conformational changes are involved in the NH 4 ϩinduced inactivation, we first examined whether the NH 4 ϩinduced inactivation is affected by mutation of amino acids located at different parts of a Kir2.1 channel. We first examined whether the amino acids (Glu-125, Ile-137, and Thr-141) involved in Ba 2ϩ binding within the Kir2.1 channel (18,19) may function in the NH 4 ϩ -induced inactivation. Fig. 3 shows that the NH 4 ϩ -induced inactivation in the E125N and I137L mutants is similar to the wild-type channels whereas it was greatly reduced in the T141V mutant. We also included our previous recording in the R148Y mutant (8), which shows little NH 4 ϩ -induced inactivation. To test whether the NH 4 ϩ -induced inactivation involves amino acids located at different parts of a Kir2.1 channel, we next recorded currents through mutants whose mutation was located at the internal pore mouth. Both Asp-172 and Glu-224 have been shown to be accessible to internal polyamines and Mg 2ϩ and thus are reckoned to be located at the internal pore mouth (3,4,20). Fig. 3 shows that the rate of the NH 4 ϩ -induced inactivation in the D172N mutant was increased (see also Fig.  4C). Both the rate and degree of the NH 4 ϩ -induced inactivation were reduced in the E224G mutant. Fig. 4 summarizes the normalized I-V m relationship (Fig. 4A) and the V m dependence of the kinetic parameters (Fig. 4, B-D) obtained in the wild type and mutants. Except for the T141V mutant whose kinetics of macroscopic currents was not analyzed, k on of all mutants showed similar V m dependence as the wild-type channels. Fitting the k on -V m relationship with a Boltzmann equation, we obtained k on and the effective gating charge (z on ), which are listed in Table I. The k on value at 0 mV, k on (0), was decreased about 2-fold in the E125N and E224G mutants. Values of z on ranged from 0.32 to 0.45 and do not seem to change dramatically although the change in the E125N mutant is statistically significant compared with the wild-type channels.
Table I also lists the parameters obtained from fitting the normalized I-V m relationships to Boltzmann distributions. The major findings shown in Fig. 4 and Table I are as follows. First, the degree of the NH 4 ϩ -induced inactivation was dramatically reduced in the T141V mutant. Second, the rate of NH 4 ϩ -induced inactivation was accelerated in the mutant D172N but was decreased in the E224G mutant. Third, the gating charge was significantly decreased in the E224G mutant. Fourth, V 0.5 was shifted toward negative V m in the E125N and E224G mutants. Table I shows that the gating charge in the E224G mutant was reduced significantly. Recently, E224 has been shown to screen surface charge, thereby affecting ion conduction (21). We next examined whether surface-charge screening affects the gating properties of the NH 4 ϩ -induced inactivation by comparing the inactivation in different ionic strengths. Fig. 5A shows the currents recorded from the wild types exposed to symmetrical 15 mM [NH 4 ϩ ] (the minimal [NH 4 OH] added to keep the pH at 7.4 in the presence of 5 mM EDTA) and 200 mM sucrose. The normalized I-V m (Fig. 5B) relationship was identical to those exposed to solutions with larger ionic strength ( Fig. 2A). These results suggest that the decrease in gating charge in the E224G mutant is not because of the reduced screening of surface charge. However, the rate of inactivation (Fig. 5C), k on (Fig. 5D), and k off (Fig. 5E) were all larger than those obtained in the wild types exposed to 10 -50 mM [NH 4 ϩ ] plus 100 mM N-methyl-D-glucamine (Fig. 2), suggesting that reducing ionic strength (less surface-charge screening) increases the kinetics of the NH 4 ϩ -induced inactivation. In sum-   Table I. in the T141V mutant. These results suggest that several amino acids ranging from the external to the internal pore mouth of the Kir2.1 channel are involved in the NH 4 ϩ -induced inactivation, which is thus likely because of global conformational changes of channel structure.
Chemical Modification of Substituted Cysteines in the Inner Pore Reduces Current and Alters the Kinetics of NH 4 ϩ -induced Inactivation-To further confirm that the internal pore is involved in the NH 4 ϩ -induced inactivation initiated by external NH 4 ϩ binding, we next examined whether chemical blocking in the internal pore changes the kinetics of the NH 4 ϩ -induced inactivation. Lu et al. (22) have shown that the wild-type Kir2.1 channel is sensitive to modifications by MTS reagents. Thus, an MTS-insensitive channel, IRK1J (C54V, C76V, C89I, C101L, C149F, and C169V), was constructed (22). Several single-cysteine substitutions (constructed in IRK1J) in the M2 domain of a Kir2.1 channel are sensitive to internal MTSET modification in 140 mM [K ϩ ], indicating that the substituted residues are located at the internal pore mouth (22). Therefore, we carried out experiments in these M2 cysteine mutants, as well as in an E224C mutant. Fig. 7A shows that, in 100 mM symmetrical [NH 4 ϩ ] MTSET did not reduce the current through IRK1J nor did it affect the inactivation process. Except for the I171C mutant, the MTSET modification increased the steadystate open probability during the test pulse (Ϫ120 mV) in all the other cysteine mutants. Also, MTSET modification signifi-  cantly decreased the rate of NH 4 ϩ -induced inactivation in the Q164C, G168C, V169C, D172C, and I176C mutants. Fig. 7, B and C summarize the effects of MTSET modification on the k on and k off (calculated using Equations 1 and 2) of the NH 4 ϩ -induced inactivation in the cysteine mutants. MTSET modification reduced k on in all the mutants whose NH 4 ϩ -induced inactivation is affected, i.e. in all but the I171C mutant. Also, MESET modification reduced k off of the NH 4 ϩ -induced inactivation in the Q164C, V169C, and D172C mutants but enhanced k off in the I176C and E224C mutants. These results suggest that MTSET modification at the internal pore mouth may alter the flexibility of the Kir2.1 channel at the internal pore mouth thereby changing the kinetics of the NH 4 ϩ -induced inactivation. Also, our previous study provides evidence for the involvement of the external pore mouth in the NH 4 ϩ -induced inactivation. Together these results suggest that the global changes of channel structure may be involved in the NH 4 ϩinduced inactivation.
State Dependence of MTSET Modification of Substituted Cysteines in the Inner Pore- Fig. 7, B and C shows that the MTSET bound in the inner pore at 0 mV (at which there is no NH 4 ϩ -induced inactivation) affects the kinetics of the NH 4 ϩinduced inactivation in some cysteine mutants. However, it remains to be determined whether the structure of the inner pore is changed during the inactivation such that the accessibility of the MTSET to the substituted cysteines is state-de-pendent. To further probe the structural changes of the inner pore during the NH 4 ϩ inactivation, we next measured the statedependent rates of MTSET modification in different conformational states of the Kir2.1 channel. Data shown in Fig. 7 were obtained by measuring MTSET modification mainly during an open state (at least 95% of time in the open state with a pulse frequency of 0.5 Hz and duration of 100 ms). Next the rates of MTSET modification during the inactivated state were estimated by holding the patches at 0 mV and stepping to Ϫ120 mV (100 ms) at 5 Hz. Using this protocol the channels were thus held at Ϫ120 mV for 50% of the total recording time. Shown in Fig. 8 are these experiments, which were carried out in the cysteine mutants whose kinetics of the NH 4 ϩ -induced inactivation are greatly affected by MTSET modification (Fig.  7, B and C). Fig. 8A shows the time courses of MTSET modification obtained with a pulse frequency of 0.5 and 5 Hz, respectively, in the Q164C mutant exposed to 100 mM symmetrical [NH 4 ϩ ]. The rate of MTSET modification was slightly higher with a pulse frequency of 5 Hz. Because MTSET is positively charged it is conceivable that its accessibility to the substituted cysteine located in the pore is affected by V m . As a control, we also measured the MTSET modification in the cysteine mutants exposed to 100 mM symmetrical [K ϩ ], in which all the cysteine mutants do not show inactivation during hyperpolarization (data not shown) (23). Fig. 8B illustrates the time courses of MTSET modification obtained with a pulse frequency of 0.5 and 5 Hz, respectively, in the Q164C mutant exposed to 100 mM symmetrical [K ϩ ]. The rate of MTSET modification was slower with a pulse frequency of 5 Hz. Fig. 8C shows the averaged rates of MTSET modification obtained in the Q164C, D172C, and I176C mutants in 100 mM [NH 4 ϩ ]. An increase of pulse frequency from 0.5 to 5 Hz did not significantly accelerate the rates of MTSET modification in the Q164C mutant (p Ͼ 0.11) although the rate seemed to be slightly higher at 5 Hz. The rates of MTSET modification in the D172C and I176C mutants  were not significantly changed (p Ͼ 0.45) by an increase of pulse frequency. The rate of MTSET modification was greatly accelerated in the D172C mutant exposed to 100 mM [NH 4 ϩ ] compared with 100 mM [K ϩ ]. The effect is not related to the NH 4 ϩ -induced inactivation and is currently under investigation in our laboratory. Fig. 8D shows the averaged rates of MTSET modification in 100 mM symmetrical [K ϩ ]. An increase of pulse frequency significantly decreased the rates of MTSET modification in the Q164C, D172C, and I176C mutants. MTSET modification in all the cysteine mutants used in this study could not be reversed by washout in the control solutions (100 mM [NH 4 ϩ ] and 100 mM [K ϩ ]). Considering the following factors, the rates of MTSET modification during the NH 4 ϩ -induced inactivation may be increased in the Q164C, D172C, and I176C mutants. First, hyperpolarization significantly decreases the accessibility of MTSET to the pore of the Kir2.1 channel (Fig.  8D). Second, the channels spend only 50% of the entire recording time at Ϫ120 mV where the channels inactivate. Third, not all of the channels are in the inactivated state at Ϫ120 mV (40% for Q164C, 60% for D172C, 70% for I176C). Together these results suggest that MTSET modification may be state-dependent to a certain degree in the Q164C, D172C, and I176C mutants.

Refinement of the Kinetic Scheme for NH 4
ϩ -induced Inactivation-Scheme 1 is simplified from Scheme 2, where K O is the dissociation constant for NH 4 ϩ binding in the open state, ␣ is the on rate, and ␤ is the off rate between the open and inactivated transition. The simplification was made by assuming that the O and O⅐NH 4 ϩ states have the same conductance and that the binding step is very rapid compared with any subsequent conformational change. k off is also dependent on [NH 4 ϩ ], suggesting that an empty inactivated state may exist. Scheme 2 is therefore modified as shown in Scheme 3, where K I is the dissociation constant for NH 4 ϩ binding in the inactivated state. The fitting of the normalized I-V m relationships to Boltzmann distributions with non-zero offsets suggests that the system is not an equilibrium one. Previous studies have demonstrated theoretically (24) and experimentally (10) that a non-equilibrium distribution of conformational states is created, if there exists a coupling of ion translocation and confor- FIG. 7. MTSET modification of IRK1J and the indicated cysteine mutants. A, each panel shows two current traces recorded from a voltage step from 0 to Ϫ120 mV under control (solid lines) and complete modification (dotted lines in IRK1J, Q168C, and V169C mutants) or 63% of peak current block (dotted lines in the Q164C, I171C, D172C, I176C, and E224C mutants). MTSET was 0.2 mM for the E224C mutant and 2 mM for all the other cysteine mutants. B and C, bar graphs of the calculated k on and k off for the NH 4 ϩ -induced inactivation under control and MTSET modification (n ϭ 4 -6). In the cysteine mutants, in which chemical modification produced complete current block (Q164C, I171C, D172C, I176C, and E224C), k on and k off were calculated from traces obtained at 63% of current block. Asterisks indicate that the MTSET groups were significantly different from the control groups. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. mational transitions. An essential requirement for the coupling of ion translation and conformational transition is that transitions between the two states can take place both in the empty and occupied state of the binding site (24). In other words, it is essential to introduce a transition between the O and I state. Therefore, Scheme 3 is further modified as shown in Scheme 4. Scheme 4 predicts that the observed k on and k off have the following relationships with ␣ 1 and ␣ 2 and with ␤ 1 and ␤ 2 , respectively, as illustrated by Equations 3 and 4.
Fitting data shown in Fig. 2, D and F to Equations 3 and 4, we obtained one ␣ and one ␤ (data not shown). Because the gating is induced by NH 4 ϩ binding ␣ 2 should be much less than ␣ 1 . Thus, ␣ 2 is equal to zero. On other hand, we propose that ␤ 1 ϭ ␤ 2 to incorporate the coupling of ion translocation and state transitions in Scheme 4. Fig. 2A shows that the gating charge is around 0.7, and the non-zero offset is about 0.2. Both values are similar to that of ClC0 channels (12). The effective gating charge is severalfold smaller than that for voltage-gated cation channels (25). Similar to previous discussion (12), we also suspect that the source of the gating charge is the NH 4 ϩ ion itself, moving inwards during the (O⅐NH 4 ϩ 3 I⅐NH 4 ϩ ) transition rather than the movement of charge intrinsic to the protein. The reason for this hypothesis is 2-fold. First, there are few charged amino acids lining the pore of a Kir2.1 channel. So far, D172 and E224 are the only two charged residues that are known to be located in the electrical field of the Kir2.1 channel. Neutralization of Asp-172 to a non-polar residue did not affect the gating charge (Table I). The gating charge in the E224G mutant is decreased but not completely eliminated. Second, the gating depends on NH 4 ϩ ions, which are permeant ions and thus efficient gatingcharge carriers. Thus, we propose that the conducting ions move through the Kir2.1 channel, entering and leaving on both sides, and keeping the cycle under consideration above out of equilibrium. Unlike the ClC0 channel, which shows a clear time-asymmetric single-channel record arising from the double-barreled nature of the channel (10), we could not actually observe a time-asymmetric single-channel record in the Kir2.1, which is a single-pore channel. However, we can reason that a channel may assume two states. One state has the binding site accessible only from the external side, and the other state has the binding site accessible only from the internal side. In this case, in the presence of a large electrochemical gradient, the channel exhibits a carrier-like behavior. An NH 4 ϩ moves from high to low electrochemical potential with each turn of the conformational cycle. Scheme 5 illustrates such an imaginary process. NH 4 ϩ (o) and NH 4 ϩ (i) denote the NH 4 ϩ at the external and  internal space, respectively. The energy profile above the I 1 ⅐NH 4 ϩ state indicates that the barrier to internal space is very high, and thus the binding site is mainly accessible from the external side. The energy profile above the I 2 ⅐NH 4 ϩ shows the opposite energy-barrier profile for the NH 4 ϩ binding site. Scheme 5 is evolved from the accumulated biophysical data of the NH 4 ϩ -induced inactivation. Further experiments need to be designed to test its validity. Also, we should emphasize that the NH 4 ϩ -induced inactivation is a non-equilibrium system. The parameters obtained by fitting data to Boltzmann equations thus do not have physical meanings. However, these parameters are useful as a first approximation for understanding the mechanism underlying the gating in the Kir2.1 channel.
Global Conformational Changes Are Involved in NH 4 ϩ -induced Inactivation-In this study, we provide two new pieces of evidence for conformational changes involved in the NH 4 ϩ -induced inactivation in Kir2.1 channels. The first line of evidence is that the NH 4 ϩ -induced inactivation during hyperpolarization is affected by mutation of amino acids located at various parts of the channel protein. These results suggest that global conformational changes are involved in the V m -dependent closure of Kir2.1 channels. In the following, we discuss the mutants whose gating for the NH 4 ϩ -induced inactivation differs dramatically from the wild types.
Glu-125 has been shown as a Ba 2ϩ binding site that is presumed to be located at the external pore mouth (19). Neutralization of this site indeed decreases both k on and k off to the same degree (ϳ30%) at all V m . The dissociation constants for k on and k off are about the same in the wild-type channels (Fig.  2, D and F). According to Scheme 5, k off depends on the ion translocation of NH 4 ϩ , which in turn depends on NH 4 ϩ binding affinity (K O for k on ). It is therefore likely that NH 4 ϩ dependence of k off is affiliated with the NH 4 ϩ dependence of k on . Thus, the effects of E125N on the kinetics of the NH 4 ϩ -induced inactivation may result from its effect on the NH 4 ϩ binding to the channel.
In the T141V mutant, which is located within the pore of the Kir2.1 channel (18), k on is increased slightly, but k off is enhanced by 10-fold. Also, the degree of the NH 4 ϩ -induced inactivation (see Fig. 3 and Fig. 4A) is greatly reduced. T141 is located close to the K ϩ selectivity filter GYG, which, according to the structure of KcsA channel, constitutes to the narrowest part to the channel (26). Therefore, it is possible that Thr-141 may also be part of the narrow pore filter and thus stabilizes the inactivated state in the Kir2.1 channel. Furthermore, the single-channel current of the T141V is larger than that of the wild type (Fig. 6). As shown in Scheme 5, the NH 4 ϩ -induced inactivation may be gated by NH 4 ϩ itself. The conducting ions move through the Kir2.1 channel, entering and leaving on both sides, and keeping the cycle under consideration out of equilibrium. An increase in single-channel conductance may further drive the cycle out of equilibrium in the direction of prompting the exit of the T141V mutants from the inactivated state. In other words, k off depends on the ion translocation of NH 4 ϩ . Therefore, the increase in k off may result from the increase of single-channel conductance in the T141V mutant.
Of all the mutants tested, E224G is the only one that shows a significant decrease in the gating charge. Recently,  has been shown to screen surface charge, thereby affecting ion conduction (21). However, Fig. 5 shows that the gating properties are the same for the wild-type channels exposed to 15 mM [NH 4 ϩ ] and 200 mM sucrose, as well as for those exposed to 10 -50 mM [NH 4 ϩ ] plus 100 mM N-methyl-D-glucamine. Therefore, the effect of E224G mutant on gating charge cannot be attributed to a decrease in surface charge screening. It is possible that Glu-224 contributes directly to the gating charge in the NH 4 ϩ -induced inactivation. On the other hand, the permeability for K ϩ in the E224G mutant has been shown to decrease (20). Therefore, it is also possible that the effect of Glu-224 on the gating charge is because of the change in conductance for NH 4 ϩ , which may be the gating charge itself. The mutation at Glu-224 (E224G) decreases both the degree and k on but does not affect k off of the NH 4 ϩ -induced inactivation. Fig. 5, D and E shows that a decrease in ionic strength increases both k on and k off . Therefore, the decrease of k on cannot be attributed to a reduction in the screening of surface charge in the E224G mutant. Because NH 4 ϩ binding is located at the external pore, the effects of mutation at position 224 are likely to be because of a decrease in the transition rate from the open to inactivated state instead of a decrease in NH 4 ϩ binding. The second line of evidence for the conformational changes hypothesis is that the MTSET modification decreases k on and k off in several cysteine mutants whose mutation is constructed at the internal pore. Our results are consistent with the hypothesis stating that the NH 4 ϩ -induced inactivation is because of conformational changes of Kir2.1 channels. MTSET modification changes the flexibility of the Kir2.1 channels, which then close and reopen in a rate that is different from the unmodified channels. Is it possible that the changes of k on and k off are because of the interaction of MTSET and NH 4 ϩ in the pore? For example, an effect of k on can be because of competition of the MTSET with the NH 4 ϩ bound at the external site. However, we consider this an unseemly possibility for the following reasons. First, we have shown previously that the NH 4 ϩinduced inactivation is inconsistent with the permeant ion block mechanism. Second, the NH 4 ϩ binding site is located at the external pore mouth, yet the cysteine mutation is positioned within the internal pore. Thus, it is unlikely that MT-SET would compete with NH 4 ϩ within the pore to decrease k on . Third, in all the cysteine mutants where k off are decreased, k on values are also decreased. Thus, our results are inconsistent with a direct competition (k off should not be affected) or knockoff (in that case, k off should increased by MTSET modification).
According to Scheme 5, k on can be because of variations in NH 4 ϩ binding affinity or the on-rate for inactivation or both. However, the cysteine-replacement is at the internal pore mouth, and the NH 4 ϩ binding site is at the external pore mouth, so the effect of k on observed in the cysteine mutants seems to indicate changes of the on-rate for the NH 4 ϩ -induced channel closure.
State-dependent modification of ion channels by MTS reagents has been used previously to probe the conformational changes of proteins in different states (27)(28)(29). We showed here that the MTSET modification may also been state-dependent to a certain degree in the cysteine mutants located in the inner pore of the Kir2.1 channel. However, the rates do not seem to be affected dramatically during the NH 4 ϩ -induced inactivation, indicating that the major structural changes of Kir2.1 channels during the NH 4 ϩ -induced inactivation may be located further SCHEME 5. Illustration of how an NH 4 ؉ moves from high to low electrochemical potential with each turn of the conformational cycle.
externally to site 164, e.g. close to the selectivity filter (7) and site 141. The effects then of MTSET modification during the open state (Fig. 7) on the NH 4 ϩ -induced inactivation may be propagated to the narrow pore whose closure is restricted, because the wider inner vestibule is held in a fixed place. CONCLUSIONS In this study, we performed further biophysical analyses of NH 4 ϩ -induced inactivation. We find that the NH 4 ϩ -induced inactivation is a non-equilibrium system. The gating properties are similar to those of the Cl Ϫ -dependent activation for the ClC0 channel. Also, we provide further evidence that conformational changes are probably proceeding to the closure of the Kir2.1 channels during the NH 4 ϩ -induced inactivation based on the following results. First, the mutation of several amino acids located at different parts of a Kir2.1 channel changes the gating of the NH 4 ϩ -induced inactivation. Second, chemical modification at the internal pore mouth reduces the rate of the NH 4 ϩinduced inactivation, which is initiated by the binding of NH 4 ϩ at the external pore mouth.
Although we have provided additional evidence supporting the relationship between structural changes and gating mechanism, more conclusive evidence still awaits a direct probe of the conformational changes in the Kir2.1 channel during inactivation. Furthermore, it remains to be defined how the amino acids, which are involved in the NH 4 ϩ -induced inactivation, move to effect changes of the kinetics of the inactivation.