Gating of the Kir2.1 channel at the bundle crossing region by intracellular spermine and other...

Gating of the Kir2.1 Channel atthe Bundle Crossing Region byIntracellular Spermine andOther CationsCHIUNG-WEI HUANG1

AND CHUNG-CHIN KUO1,2*1Department of Physiology, National Taiwan University College of Medicine, Taipei, Taiwan2Department of Neurology, National Taiwan University Hospital, Taipei, Taiwan

In the Kir2.1 channel, the flow-dependent blocking effect of intracellular spermine (SPM) strongly indicates coupled movement of ions in asegment of the pore. We have shown that the bundle crossing region of M2 constitutes this critical segment of the pore. Moreover, thissegmentmay undergo opening/closing conformational changes mimicking channel gating. In this study, we further investigate these “gating”conformational changes and relevant controlling mechanisms at this critical segment. We demonstrate that A184R mutation in the innerend of the bundle crossing region not only abolishes the inward rectifying features of SPM block but also tends to close the channel pore,which can then only be opened by intracellular (e.g., Naþ, or equally effectively, Kþ) but not extracellular cations. We also found that theexit (back to the intracellular milieu) of the blocking in the deep site is facilitated rather than deterred by the presence of the other SPM inthe superficial site.We conclude that intracellular SPMmay bind to a deep site in the pore and serve as a flow-dependent blocker. The SPMin the superficial site, on the other hand, serves both as a docking form ready for permeation to the deep site, and as a gating particle capableof opening the bundle crossing region. This inner end of the bundle crossing region of the Kir2.1 channel pore thus constitutes a pivotalsegment, which, in collaboration with intracellular SPM and Kþ ions, closely couple channel gating to (inward rectifying) ion permeation.J. Cell. Physiol. 229: 1703–1721, 2014. © 2014 Wiley Periodicals, Inc.

Inward-rectification Kþ currents have been expressed andrecorded in either Xenopus oocytes or mammalian cells(Hagiwara et al., 1976; Matsuda, 1988; Chang et al., 2005). TheKir2.1 channel is widely expressed in the cardiovascular andnervous systems (Karschin et al., 1996; Miyashita andKubo, 1997; Pruss et al., 2005; Howe et al., 2008), and is one ofthe strongest inward rectifying Kþ channel, giving rise toapparently much larger inward than outward currents inphysiology conditions (Hagiwara and Takahashi, 1974;Hille, 2001; Lu, 2004). The mechanism underlying the intriguinginward rectification phenomenon in the Kir channel has beenascribed mostly to spermine (SPM) or Mg2þ block of the pore(Hagiwara and Takahashi, 1974; Lu and MacKinnon, 1994;Matsuda et al., 2003; Huang and Kuo, 2014). The I–V curve ofthe Kir2.1 channel, obtained in the absence of internal blockers,shows a linear relationship (Lopatin et al., 1995; Xie et al., 2003;Huang and Kuo, 2014). Intracellular SPM preferentially inhibitsthe outward currents in a concentration-dependent manner(Lopatin et al., 1995; Matsuda et al., 2003, 2010; Huang andKuo, 2014). The major blocking effect of intracellular SPM isclosely correlated with current flow, a phenomenon stronglyindicative of the existence of coupled ionic movement in a flux-coupling pore region (Lopatin and Nichols, 1996; Guo andLu, 2003; Shin and Lu, 2005; Kurata et al., 2007; Huang and Kuo,2014). In this regard, D172 in the central cavity of the pore hasbeen identified as the major ligand marking the blocking site ofSPM (Kubo and Murata, 2001; Shin and Lu, 2005; Kurataet al., 2010, 2013; Matsuda et al., 2010; Huang and Kuo, 2014).Actually D172 is probably located close to the external end ofthis flux-coupling segment, because there is an adjacentexternal barrier which deters the passage of SPM but not Kþ atS165 (Kubo and Murata, 2001; Fujiwara and Kubo, 2002;Kurata et al., 2010, 2013). On the other hand, themore internalpart of this flux-coupling segment is less well characterized.Among the residues internal to D172, residues E224 and E299in the cytoplasmic domain are the most frequently reportedtargets of mutagenesis which disrupts rectification of ionic flow

and/or narrows the pore (Kubo and Murata, 2001; Changet al., 2003, 2005; Clarke et al., 2010; Huang and Kuo, 2014).The cytoplasmic domain, however, is a much widened part ofthe pore (Tao et al., 2009; Clarke et al., 2010; Hansenet al., 2011). The diagonal distances between any two of thefour E224 or the four E299 were estimated to be 15–17Å(Pegan et al., 2005), and the distances between D172 and E224or E299 were estimated to be 36–38Å (according to molecularhomology modeling based on the crystallographic data on theKir2.2 channel [Pegan et al., 2005; Tao et al., 2009; Hansenet al., 2011]). It is thus hard to envisage E224 and E299 as part ofthe critical flux-coupling segment, which in theory shouldcontain single-file ionic sites separated by insignificant energybarriers and physically contiguous or connected to D172. E224and E299 mutations have accordingly been implicated to affectrectification via allosteric mechanisms (Xie et al., 2003; Peganet al., 2005; Tao et al., 2009; Huang and Kuo, 2014).

Contract grant sponsor: National Health Research Institutes,Taiwan;Contract grant numbers: NHRI-EX101-10006NI, NHRI-EX102-10006NI.Contract grant sponsor: National Science Council, Taiwan;Contract grant number: NSC95-2320-B-002-064-MY3.

*Correspondence to: Chung-Chin Kuo, Department of Physiology,National Taiwan University College of Medicine, No. 1, Jen-AiRoad, 1st Section, Taipei 100, Taiwan.E-mail: [email protected]

Manuscript Received: 10 December 2013Manuscript Accepted: 12 March 2014

Accepted manuscript online in Wiley Online Library(wileyonlinelibrary.com): 14 March 2014.DOI: 10.1002/jcp.24616

ORIGINAL RESEARCH ARTICLE 1703J o u r n a l o fJ o u r n a l o f

CellularPhysiologyCellularPhysiology

© 2 0 1 4 W I L E Y P E R I O D I C A L S , I N C .

Based on the amino acid sequence of the Kir2.1 channel andthe crystallographic data on the Kir2.2 channel, residues I176to A184 may be responsible for the pore segment immediateinternal to D172 (Tao et al., 2009; Clarke et al., 2010; Hansenet al., 2011). The possible role of this part of the pore in inwardrectification, however, has not yet been examined in detail. Inthis regard, one may note that I176 to A184 also constitute the“bundle crossing” region, a region responsible for theactivation/deactivation gate and capable of open/closingconformational changes in many other ion channels (Kitaguchiet al., 2004; Kurata et al., 2010; Khurana et al., 2011). If theKir2.1 channel indeed chooses this relatively mobile part of thepore to make inward rectification, which relies on flux couplingand thus presumably a delicate arrangement of ligands on thewall of the pore, it would be desirable to explore the molecularinteractions between SPM and this bundle crossing region inmore detail. In this study, we show that A184 in the bundlecrossing region is essential for the appropriate conformation ofthis critical pore segment. Introduction of an arginine residueto this position would markedly increase the peak of anasymmetrical permeation barrier for SPM, andmake the bundlecrossing region “closed” if there are not enough cations(e.g., Kþ, or equally effectively, Naþ) present intracellularly.Accordingly, we also demonstrate that internal SPM is not onlya rectifying blocker of the Kir2.1 channel at the bundle crossingregion, but also a “gating particle” capable of opening thisregion of the pore.

Materials and MethodsMolecular biology and expression of Kir2.1 channel

This part of the methods is basically the same as that described inthe previous study (Huang and Kuo, 2014). Briefly, mutant channelcDNAs (mouse macrophage Kir2.1 DNA subcloned into Blue-script II SK1) were made using the QuickChange site-directedmutagenesis kit (Stratagene, La Jolla, CA). Each mutant DNA ofKir2.1 channel was conformed by automatic DNA sequencer(Applied Biosystems, Foster City, CA, 3730x1 DNA Analyzer).cRNA was made from purified linearized cDNA with in vitro T7transcription kits (mMESSAGE mMACHINE, Ambio, Austin, TX).Xenopus oocytes were defolliculated by wash for�40min with 1%collagenase (Sigma Chemical Co., St. Louis, MO) in ND 96 solutionwhich contains (in mM): 96 NaCl, 2 KCl, 1.8 MgCl2, 1.0 CaCl2, and5 HEPES (pH 7.4). After defolliculation, Xenopus oocytes weremicroinjected �55 nl Kir 2.1 cRNA (Nanojector II, DrummondScientific, Inc., Sarasota, FL) and weremaintained at 18°C inND 96solution within 1–5 days. Prior to inside-out giant patch recording,theXenopus oocyte wasmoved to a hypertonic solution containing(mM): 500 Sucrose, 72 KCl, 8 K2HPO4, 2 KH2PO4, and 5 K2EDTA(pH 7.4) for �3min to separate the membrane from the vitellinemembrane, which was then removed manually. The Xenopusoocyte was used for excised inside-out giant patch recording rightafter wash in the bath solution containing (mM): 72 KCl, 8K2HPO4, 2 KH2PO4, and 5 K2EDTA (pH 7.4). All animalexperiments were performed in accordance with animal welfareguidelines and were approved by the Institutional Animal Care andUse Committee (IACUC) of National Taiwan University Collegeof Medicine.

Electrophysiological recordings

Macroscopic currents were recorded from excised inside-outgiant patches from microinjected Xenopus oocytes with anAxopatch 200A amplifier and pClamp 6.0 software (AxonInstruments, Inc., Sunnyvale, CA). The electronic pipettes werepulled from borosilicate micropipettes (a tip internal diameter of10–60mM) and fire polished (Narishige Scientific Instruments, Inc.,Tokyo, Japan). The pipette resistance was 100–600 kV when filledwith extracellular solutions. The voltage error due to series

resistance was estimated to be no larger than 2–5mV in mostcases, and thus no corrections were made in all of the records.Unless otherwise specified, the extracellular (recording) andintracellular (bath) solutions both contained (in mM): 72 KCl,8 K2HPO4, 2 KH2PO4, and 5 K2EDTA (the “100mMKþ solution”),titrated to pH 7.4 with KOH. The intracellular solutions alsocontained 25mM L-a-phosphatidylinositol-4,5-bisphosphate(PIP2), 0.5–3.0mM SPM (both from Sigma Chemical Co.). For theexperiments in symmetrical 10mM Kþ, both intracellular andextracellular solutions contained (in mM): 4.6 KCl, 0.9 KOH, 2K2HPO4, 0.5 KH2PO4, and 1 EDTA (the “10mM Kþ solution”),pH 7.4. For the experiments in symmetrical 20mM Kþ, bothintracellular and extracellular solutions contained (in mM):8.6 KCl, 2.4 KOH, 4 K2HPO4, 1 KH2PO4, and 3 EDTA(the “20mM Kþ solution”), pH 7.4. For the experiments insymmetrical 30mM Kþ, both intracellular and extracellularsolutions contained (in mM): 15 KCl, 2 KOH, 4 K2HPO4, 1KH2PO4, and 2 K2EDTA (the “30mM Kþ solution”), pH 7.4.Data were filtered at 1–2 kHz and digitized at 1–10 kHz throughDigidata 1200 interface (Axon Instrument, Inc.).

After a seal was formed, the patch was moved in front of thethree-barrel square glass tube (internal diameter�0.6mm)emittingdifferent intracellular recording solutions. The glass tube holderwasconnected to a stepper control (SF-77B perfusion system, WarnerInstruments, Inc., Hamden, CT) which enabled rapid switchbetween different glass tubes. The speed of solution changewith thesystemwas carefully examined in the presence (Fig. 1A) and absence(Fig. 1B) of a membrane patch at the pipette tip. The very similarkinetics in Figure 1A,B demonstrate that the rate-limiting factors ofsolution change in this case aremost likely intrinsic properties of thesystem, and thus the existenceof a giantmembrane patch containingKir 2.1 channels does not cause additional delay or slowing insolution change. Because the mean latent period before the currentchange to 5% solution exchange from the electronic command was�75msec, and to 95% solution exchange was �180msec (Fig. 1),we always fitted the decay/recovery phases of SPM block/unblock200msec after the electronic command of solution change wasgiven. This 200msec delay was deliberately chosen to assure that>95% solution change was achieved (Fig. 1). In the meanwhile, theSPM concentration was also deliberately chosen to assure <20%current change after 200msec based on an estimate with amonoexponential process and the time constant from the fit (i.e.,>80% of the total current change for the fitting process). All dataare expressed as mean� SEM. For comparisons between groups,the Student’s t-test was used, and P< 0.05 was considered to bestatistically significant.

Homology modeling of the Kir2.1 channel

This part of the methods is basically the same as that described inthe previous study (Huang and Kuo, 2014). Briefly, homologymodels of the Kir2.1 channel were built from the X-ray crystalstructure of the chicken Kir2.2 and mouse Kir2.1 channel (Peganet al., 2005; Tao et al., 2009; Hansen et al., 2011). Specially, thestructure of cytoplasmic domain and transmembrane domainwerecreated from the structure template of chicken Kir2.2 (PDB ID:3SPI) and mouse Kir2.1 channel (PDB ID: 1U4F). The alignedsequences were presented to the Discovery Studio v2.5 program(DS v2.5; Accelrys, Inc., San Diego, CA) to generate relativepositions of the chosen residues and the secondary structure ofthe Kir2.1 model.

Molecular dynamics simulation

Molecular dynamics simulation for modeling of theWT and A184Rmutant channels were also performed using Discovery Studio v2.5programs (Accelrys, Inc.) and the Chemistry at HARvardMolecular Mechanics (CHARMm) force field (Shiau et al., 2002;Lou et al., 2003). This simulation system provides an appropriate

JOURNAL OF CELLULAR PHYSIOLOGY

1704 H U A N G A N D K U O

environment for the interaction-site water model. Supplementarycounter-ions were added to ensure that the overall net charge onthe system was zero. Subsequent to a “steepest descent” energyminimization with positional restraints, Kþ (or Naþ) ions wereintroduced by the replacement of water molecules at the highestelectrostatic potential in the bundle crossing region, the centralcavity, and the cytoplasmic domain of theWT, E224Q, and A184Rmutant channels. A second conjugated gradient energy minimi-

zation was then performed until no significant energy change couldbe detected. During this equilibration process, the watermolecules and Kþ (or Naþ) were free to move. Berendsencoupling was applied for the simulation parameters to maintain aconstant temperature of 300°K and a constant pressure of 1 bar(Berendsen, 1984). The Van der Waal’s force was modeled with acutoff value of 10Å. The Leapfrog Verlet procedure (advanceddynamic integrator) was used to integrate the equation of motion.

Fig. 1. Examination of the speed of solution change. A: An excised inside-out giant patch expressing Kir 2.1 channel was held at þ10mV andwas moved in front of the three-barrel square glass tube (internal diameter �0.6mm) emitting different intracellular recording solutions. Thesolution on the intracellular side was changed from 100mM Kþ solution to a solution containing 100mM N-methyl-D-glutamine (NMDG) butno Kþ. The time course of change in the current amplitude is taken as an indicator of the speed of internal solution change with the perfusionsystem. The blue dotted lines are the regression lines of the initial �30msec and last �30msec of the current and define the range of thecurrent amplitude change. Arrow a indicates the time point at which the electronic command of solution change is given. Arrow b and cindicate the time points at which the current reaches 5%, and 95% of its total change, respectively. There is at first a latent period (�75msec)before the point at which the current starts to change (from the time point when the computer command is given to 5% change of thecurrents). It then takes �100msec from 5% to 95% of total current change, respectively. B: An open-tip glass pipette containing no oocytemembrane patch was moved in front of the three-barrel square glass tube emitting different intracellular recording solutions. The protocoland arrow has the same meaning as in part A. In this case, the junctional current changes are directly driven by the junctional potentialchanges, illustrating the speed of solution change in the absence of a membrane patch containing Kir 2.1 channels. C: Cumulative results areobtained from five patches with the same protocol described in part A and from five open-tip glass pipettes in part B. The times (from point ain part A and B) required for the current decay to reach 1%, 5%, 50%, 90%, and 95% of its total currents are rather similar in both cases. It isevident that >95% solution change would be completed in most instances with an elapsed time of 200msec, from the electronic command.


S P M B L O C K O F K i r 2 . 1 C H A N N E L I N T H E B U N D L E C R O S S I N G R E G I O N 1705

In the presence of ~40watermolecules, 1–7 Kþ (orNaþ) ions and/or 2 SPM molecules were applied to the cavity region of WT,E224Q, and A184R mutant channels to select the one with thelowest potential energy for equilibration from the 10 candidatemodels.

ResultsThere is a large discrepancy between the kinetics ofinternal SPM block measured with or without pre-existing SPM

It has been reported that the bundle crossing region seems tobe capable of gating-like conformational changes in the otherKþ channels (Kitaguchi et al., 2004; Khurana et al., 2011), wetherefore explored whether such gating conformationalchanges are also present in the WT Kir2.1 channel. Figure 2Ashows that 0.3–1.0mM SPM markedly inhibits the outward butnot the inward Kir 2.1 currents, a rectifying blocking effectcharacterizing the primary action of SPM on the channel. Thedecay time constants in the presence SPM are fitted with amonoexponential function. The reciprocals of the timeconstants at þ10mV are plotted against SPM concentration inFigure 2B, where a linear fit gives a binding (association) rateconstant of SPM (kon) of 2.3� 10þ9M�1 sec�1 (the slope of thelinear fit) to the first blocked state, and an unbinding(dissociation) rate of SPM of 20 sec�1 (the Y-intercept of thelinear fit).

The association (kon) and dissociation rate (koff) constants ofSPM interacting with the binding site are also simplisticallyderived based on the following functions by Holmgren(Holmgren et al., 1997):

kon ¼ 1� ft � ½SPM� ð1Þ

koff ¼ ft

ð2Þ

where f is the relative current given by the ratio between thesteady-state current in SPM and that in control, and t is themacroscopic decay time constant of the current being blockedby internal SPM. The derived SPM binding and unbinding ratesare �2� 10þ9M�1 sec�1 and �15 sec�1 at þ10mV (Fig. 2C),reasonably close the results in Figure 2B. Similar values of konand koff could also be derived using Equations (1) and (2) andexperiments repeated in different concentrations (0.3–30mM)of SPM (data not shown). One may also note that the kineticdata remain rather constant when there is preponderantoutward currents (þ20 to þ100mV), but start to showsignificant changes (decrease in kon and increase in koff) whenthe preponderant direction of current flow turns inward (þ10to �20mV, Fig. 2C). On the other hand, the kinetics of SPMaction are very different if measured with a rapid exchange ofthe internal solution (Fig. 2D). When the patch containingKir2.1 was held at þ10mV and moved from the controlsolution into a solution containing 0.001 and 1.0mM SPM, thedecay time constant were �350 and �150msec, respectively,much slower than that in Figure 2A. The cumulative results inFigure 2E indicate a binding rate constant nearly 1,000-foldslower than that in Figure 2B,C. In contrast, the unbinding ratesin Figure 2D,E are just�10 times slower than that in Figure 2B,C. These very large discrepancies between the binding kineticsmeasured with and without pre-existing SPM (Fig. 2F) and thesmaller discrepancies between the unbinding kinetics (in Fig. 2Fthe ambient SPM was existent for a longer time when the offrates thanwhen the on rates weremeasured) imply a significantinteraction between SPM and the Kir2.1 channel pore, or morespecifically, a possible pore-opening effect of SPM (see below).

The discrepancy between the two differentmeasurements gets smaller with fewer charges andsmaller size of the blocker

We also explore whether other polyamines (e.g., spermidine[SPD] and putresine [PUT]) and Mg2þ ions have similardiscrepancies to SPM. Figure 3 demonstrates that the bindingrates of these blockers still show marked discrepancies whenmeasured with the two foregoing different methods. However,it seems that the discrepancies are the largest for SPM, and thendecrease with fewer positive charges on the blocker (þ3 forSPD, þ2 for PUT and Mg2þ). Besides, Mg2þ shows an evensmaller discrepancy than PUT, suggesting less blocker-channelinteraction or pore-opening effects of Mg2þ (which has arelatively smaller size and simpler conformation) than PUT.

The discrepancy in SPM blocking kinetics is markedlydecreased or even is negligible in specific mutant Kir2.1channels

If the foregoing discrepancy is ascribable to the pore openingeffect of (pre-existing) SPM, then it may become less evident inspecificmutant channels bearing a constricted pore. E224G andE299S mutations are known to allosterically constrict theKir2.1 channel pore (Chang et al., 2003, 2005; Huang andKuo, 2014). Also, the bundle crossing region of the pore (e.g.,M183 and A184) has been shown to undergo gating-likeconformational changes (Clarke et al., 2010; Huang andKuo, 2014). When the patch containing the E224G mutantchannel was stepped from�100 toþ20mV (Fig. 4A–C) or wasmoved in front of different tubes emitting the control solutionand the solution containing SPM (Fig. 4D,E), the binding ratesbecome essentially the same with the two different experi-mental approaches. Similarly, the discrepancy is markedlydecreased or even negligible in the E299S, M183N, and A184Rmutant channels (Fig. 4F). These results suggest that E224 andE299 may allosterically keep the bundle crossing region (e.g.,M183 and A184) open or prone to open, making a stage for theeffect of SPM on ion permeation and gating-like conformationchanges.

The blocking SPM in the deep site unbinds faster withhigher intracellular SPM

When the WT Kir2.1 channel is tested with a two-pulseprotocol (Fig. 5), a “slow tail” emerges in the intervening�60mV pulse in the presence but not in the absence of internalSPM. The slow tail thus is highly likely ascribable to the slowunbinding of SPM from a (deep) binding site. It is interesting thatboth the decay and the development of the slow tail,presumably signaling the unbinding and binding rates of SPM,respectively, are internal SPM concentration-dependent. Forexample, the decay time constants are �0.8 and �0.3msec, ininternal 0.01 and 100mMSPM at�60mV, respectively (Fig. 5A,B). Although the difference between 0.8 and 0.3msec is notlarge, a dose–response relationship between the SPMconcentration and SPM unbinding rates in the WT channelcould be clearly demonstrated and described by a one-to-onebinding process with a Kd of �3.5mM (Fig. 5C). A similar butless pronounced phenomenon could also be seen in the E224Gmutant channel, although with a much larger Kd of �30mM(Supporting Information Fig. S1). It is intriguing that increasedambient concentrations of SPM could accelerate the unbindingof the bound SPM. This finding strongly implicates that thechannel must be able to accommodate at least two SPMmolecules simultaneously (Lopatin et al., 1995). Moreover,SPM binding to the superficial site very likely would “open” thebundle crossing region to facilitate the access to or exit fromthe deep site in the pore. It is also interesting to note that these



Fig. 2. Markedly different kinetics of internal SPM block of the WT channel when measured with different methods. A: The voltage acrossthe patch membrane was first held at 0mV and was hyperpolarized to �100mV for 15msec to remove any residual block, and then stepped toa test potential at þ10mV for 25msec in the absence (control) or presence of SPM. The decay phases of the outward current at the þ10mVtest potential in 0.3 and 1.0mM SPM are well-described monoexponential functions with t¼ 2.3msec and 0.55msec, respectively. The greendotted line indicates the zero current level. B: The reciprocal of the time constant (mean�SEM, n¼ 3–6) from part A (1/tau), is plottedagainst [SPM] (spermine concentration) (0.01, 0.03, 0.1, 0.3, 1.0, and 3.0mM) for þ10mV. The dashed line is a linear regression fit to the datawith a slope of 2.3� 10þ9M�1 sec�1� [SPM], and a Y-intercept of 20 sec�1. C: The cumulative results of kon and koff from calculations based onexperiments in part A and Equations (1) as well as (2) are plotted against the test voltage (n¼ 5). D: SPM block and unblock with rapid solutionchange in theWT channel. The patch was held at þ10mV, and quickly moved into or out of the solution containing 1.0mM internal SPM. Boththe onset and recovery phases (200msec after the electronic command was given, see Materials and Methods Section) of SPM block/unblockare well fitted with monoexponential functions, and the time constants are �150msec and �7 sec, respectively (red lines). E: The reverses ofthe blocking and unblocking time constants from part D are plotted against [SPM] (spermine concentration). The dashed line is a linearregression fit to the data with a slope of 4.2� 10þ6M�1 sec�1� [SPM], where [SPM] denotes SPM concentration, and the dotted line is ahorizontal line denoting 1/tauoff¼ 0.3 sec�1 (n¼ 5–8). F: A comparison between the binding and unbinding of kinetics measured with differentmethods (see parts A and D) at þ10mV. There is a large discrepancy between the binding rate constant (kon) measured with and without pre-existing SPM. The discrepancy between the unbinding rates (koff), on the other hand, is much smaller.



measured Kd values of SPM (in terms of pore opening) is verymuch close to the Kd2 (Supporting Information Fig. S2), whichsignals SPM binding to a shallow binding site in the WT andE224G mutant channels. The result of molecular dynamicssimulation is also consistent with the view that the Kir2.1channel could accommodate at least two SPM molecules inthe pore simultaneously (Fig. 6B,C). Furthermore, we foundthat the unbinding of the blocking SPM in the deep site isfacilitated rather than deterred by the presence of the otherSPM in the superficial site. These findings indicate that SPMmayact not only as a pore blocker when binding to the deep site,but also as a gating particle capable of opening the pore atthe bundle crossing region (e.g., A184, Fig. 6D andsee below) when binding to the superficial site in the WTKir2.1 channel.

Residue A184 plays a critical role in both inwardrectification and channel gating

In the previous figures we have noted that SPM may act as agating particle “opening” the Kir2.1 channel pore at A184 at theintracellular end of the bundle crossing region. We thereforeexamined whether A184 also plays a direct role in the inwardrectification associated with internal SPM block. We made afew substitutions (C, W, G, Q, E, and R) for this residue, andgot the A184R mutant channels showing enough currents foranalysis and most different characters of internal SPM block.From �100 to þ100mV, the Kd values in the A184R mutantchannel are always higher (the affinity is lower) than that in theWT channel (Figs. 7 and S2). Most interestingly, the Kd–Vm plotno longer shows a dramatic change around the reversalpotential. The IR index of the mutant A184R channel (�2) thusis much smaller than that of the WT channel (Fig. 7C),substantiating that A184 in the bundle crossing region isresponsible for a pore conformation critical for the flow-dependent block by internal SPM, the essential feature of theKir2.1 channel.

A184R, E224G, and E224Qmutations increase the heightof the same permeation barrier in the pore

We further investigate the molecular effect of A184R, E224G,and E224Q mutations. Both inward and outward currentsthrough the WT channel are discernible and have a roughlylinear I–V relationship in the absence of SPM (Fig. 8A,B). TheI–V plots of the E224G and E224Q mutant channels, however,appear less linear. A close examination of the I–V plots ofE224G and E224Q reveals that, unlike the case of SPM block(of the WT channel), the conductance does not show a sharpchange around the reversal potential. This is because theinward currents in the E224G and E224Qmutant channels havea larger increase with a fixed amount of hyperpolarization (i.e.,have a larger conductance) toward more negative potentials.This property is even more manifest in the A184R mutantchannel, but remains roughly the same in A184R/E224G doublemutations, implying that E224G and A184R mutations act onthe same sites of the pore to make this intriguing property.

Figure 8C shows that the conductance ratio is fairly constantin each mutant channel between�100 and�150mV. The ratiois �1 (and thus a roughly linear I–V relationship) in the WTchannel, but is reduced to �0.8 for the A184R and E224G/A184R double mutant channels (the currents of E224Q/A184Rmutant were so small that the I–V plot can not be reliablydefined). According to the model of Mott and Gurney (Mott,1940), the rate for an ion to go through the channel pore isinversely correlated with the height of the major (or rate-limiting) energy barrier it must cross and is proportional toexp(�U/RT), where U represents the height of the energybarrier. The rate constant for the Kþ ion to cross the energybarrier can then be expressed as A� exp[�(U�ZdFV)/RT],where A is the pre-exponential factor, Z (for Kþ ions) is 1, d isthe electrical distance between the rate-limiting energy barrierpeak and the adjacent upstream energy well, V is themembranepotential, and R, T, and F have their usual thermodynamicmeanings. The conductance ratio is then given by:

DI1ðV�V�10ÞDI2ðV�10�V�20Þ

¼exp �U

RT þ ZdFðV�10ÞRT

h i� exp �U

RT � ZdFVRT

� �

exp �URT � ZdFðV�20Þ

RT

h i� exp �U

RT � ZdFðV�10ÞRT

h i ð3Þ

where DI1 and DI2 denote the current amplitude change frommembrane potentials (inmV) V to V� 10, and from V� 10 toV� 20, respectively. With the conductance ratio of �0.8(A184R single mutant and E224G/A184R double mutantchannel, Fig. 8C) and Equation (3), one can readily have a newmajor (rate-limiting) permeating barrier with an external slopeof 0.5 (the electrical distance, d, from the peak to theimmediately externally locatedwell in the pore) in the presenceof A184R mutation at �100 to �150mV. In comparison, thisbarrier is unlikely the rate-limiting one for Kþ permeation inthe WT channel, where a conductance ratio of �1 (Fig. 8C)would signal a relatively “sharp” rate-limiting energy barrierwith an external slope d of �0 (see Discussion for moredetails). In this regard, it is interesting that E224Q mutationcreates a new permeation barrier with an external slope dof 0.4, a value very close to the d in the A184R mutantchannel. Moreover, double mutations E224G/A184R do notfurther change the d value. These findings would stronglyindicate that A184R and E224G mutations increase the heightof the same permeation barrier in the Kir2.1 channel pore. Inother words, these findings are consistent with an idea thatthere is a permeation barrier at the internal end of thebundle crossing segment (�A184). The height of the barrier islow and thus has no significant effect on Kþ permeation inthe WT channel, but could be markedly increasedallosterically with E224 mutations or directly with A184

Fig. 3. The binding kinetics of other internal polyamines and Mg2þ

onto the WT Kir2.1 channel. Summary of the kon values for theblock of WT channel by internal SPM, SPD, PUT, and Mg2þ ions.The kon values (mean�SEM; n¼ 4–5) were obtained with the sameexperimental protocols in Figure 2A,D. The circles connected by adotted line denote the ratios between the kon (at þ20mV) valuesobtained from the two methods in Figure 2, and are �1,000, �500,�300, and �190 for SPM, SPD, PUT, and Mg2þ, respectively.



Fig. 4. The discrepancy in SPM blocking kinetics is markedly decreased or even negligible in specific mutant channels. A: The voltage across apatch membrane containing the E224G mutant channel was first held at 0mV and then hyperpolarized to �100mV for 10msec, followed by atest pulse of þ20mV for 20msec in the absence (control) or presence of SPM. The E224G mutant channel apparently needs higherconcentrations of SPM to inhibit the outward currents. The decay phase of the outward current upon stepping the membrane voltage from�100 to þ20mV in 300mM SPM is well fitted by a monoexponential function with a time constant of �25msec. The green dotted line indicatesthe zero current level. B: The reciprocal of the time constant (mean�SEM, n¼ 3–5) from part A (1/tau), is plotted against [SPM] (spermineconcentration) (10, 30, 100, 300, and 1,000mM) for þ20mV. The dashed line is a linear regression fit to the data with a slope of1.5� 10þ5M�1 sec�1� [SPM], and Y-intercept of 0.01 sec�1. C: The cumulative results of kon from calculations based on experiments in part Aand Equation (1) is plotted against the test voltage (n¼ 5). D: SPM block and unblock with rapid solution change in the E224Gmutant channel.The patch was held at þ20mV, and quickly moved into or out of the solution containing 1.0mM internal SPM. Both the onset and the recoveryphases of SPM block/unblock are fitted with monoexponential functions, and the time constants are �1.5 and �120 sec, respectively. E: Thereverses of the time constants of the onset and recovery phase from part D are plotted against [SPM] (spermine concentration). The dashedline is a linear regression fit to the data with a slope of 1.3� 10þ5M�1 sec�1� [SPM], where [SPM] denotes SPM concentration (n¼ 5–8). F:Summary of the kon values of SPM (at þ20mV) obtained with the two different methods for the WT, E224G, E299S, M183N, and A184Rmutant channels. The kon values are �1,000-, �1.1-, �1.0-, �13-, and �40-fold different between the two methods for the WT, E224G, E299S,M183N, and A184R mutant channels, respectively.



Fig. 5. Faster SPM unblock in the presence of higher concentrations of internal SPM. A: A two-pulse protocol was used to measure thebinding and unbinding kinetics of internal SPM. The patch was held at 0mV and stepped twice to �100 and �60mV for 5msec each, withgradual lengthening of the intervening þ100mV period in control and in the presence of 0.01 or 100mM internal SPM in the WT Kir2.1channels. The development of the slow tail in the second �60mV pulse is used as a measure of SPM binding to the slow unbinding siteatþ100mV. Consistently, there is no discernible slow tail at �60mV in control (in the absence of SPM). Part of the current sweeps in thesecond �60mV pulse is enlarged to demonstrate the different decay kinetics of the slow tails in internal SPM. The decay phase of the slow tailcurrents is fitted with monoexponential functions, and the time constants are �0.8 and �0.3msec for 0.01 and 100mM SPM, respectively. Thegreen dotted line indicates the zero current level. B: Cumulative results were obtained for the experiments described in part A. The timeconstants of decay of the slow tail currents at �60mV are 0.9� 0.17msec (n¼ 4) and 0.3� 0.06msec (n¼ 8) in the 0.01 and 100mM SPM,respectively. *P< 0.001 when compared with data from 0.01mM SPM. C: The inverse of the decay time constant of the slow tail isplotted against SPM concentration for the WT channel (n¼ 4–8). The dotted line is a fit to the data points of the form: 1/taudecay(msec�1)¼ {A1/(1þ [SPM]/Kd)}þ {A2� ([SPM]/Kd)/(1þ [SPM]/Kd)}, with A1¼ 1.14, A2¼ 3.74, and Kd¼ 3.5mM.



Fig. 6. Molecular dynamics simulation determines SPM is not only a blocker of the Kir2.1 channel but also a gating modifier capable ofopening the bundle crossing region. A: The homology model is based on sequence alignment and crystal structures of the Kir2.1 (cytoplasmicdomain only) and Kir2.2 channel using Discovery studio v2.5 software (see Materials and Methods Section for more details). Side view of theKir2.1 channel showing the transmembrane domain of the two subunits. The transmembrane domain is colored green and yellow. The sidechains of A184 are indicated in CPK model (carbon, gray; nitrogen, blue; and oxygen, red). In the narrowed pore portion, the predicted inter-residue diagonal distance (from side chain tip to tip, the black line) between residues of A184 is �11Å. B: In the presence of 40 watermolecules and 7Kþ ions, two SPM molecules are applied to the central cavity of the channel. Two subunits of the channel are shown in solid-ribbon model. The two SPM molecules in the deep and superficial binding sites are presented in CPK model. C: Representative structure ofthe transmembrane and cytoplasmic domains of one subunit of the WT Kir2.1 channel is shown, with a regional view of the residues D172,M183, A184, E224, and E299 in the boxed part (CPK model: carbon, gray; nitrogen, blue; sulfur, orange; and oxygen, red). The two SPMmolecules in the deep and superficial binding sites, and the side chains of the foregoing residues are presented in CPK model. D: In thepresence of 40 water molecules and 7Kþ ions, two SPM molecules are applied to the central cavity of the channel. The Kir2.1 tetramer isviewed from the intracellular side. The four different subunits are colored green, blue, yellow, and pink. The side chains of A184 residue arepresented in CPK model (scale bar¼ 10Å). The white-dotted square (from Ca to Ca of residues A184) indicates the evident pore-openingeffect of SPM at the level of A184.



mutations to create a new rate-limiting step of Kþ permeation.These aforementioned mutations seem to result in majorconformational changes at the bundle crossing region, as theynot only significantly increase the barrier height but alsodecrease the Kþ sites (or decrease the coupling betweenKþ sites) to eliminate the inward rectification produced by SPMin the pore.

The inward currents in the A184R mutant channeldevelop slowly in low internal Kþ

In view of the potentially changeable conformation of thebundle crossing region in the Kir2.1 channel, we explored thepossible controllers of this imperative region of the pore. Wefound an intriguing development phase in the inward currents

Fig. 7. Inhibition of A184R mutant currents by different concentrations of internal SPM in symmetrical 100mM Kþ. A: Sample sweepsdemonstrating inhibition of A184R mutant currents from the same patch by internal SPM in symmetrical 100mM Kþ. The patch containingA184R mutant channel was first hyperpolarized from 0mV holding potential to �100mV for 50msec, and then stepped to different testvoltage between �200 and þ100mV for 450msec in 10mV increments. The patch was then step to �100mV again for 50msec beforereturning to the holding potential of 0mV. The green dotted line indicates the zero current level. Inset: The current traces in A184R mutantchannel elicited by a test voltage of þ100mV with or without 300mM internal SPM (the red and black traces, respectively) are superimposedto facilitate a direct comparison. The green dotted line indicates the zero current level. B: The relative current is defined as the ratio betweenthe steady-state current amplitude in internal SPM and in control, and is plotted against SPM concentration for the A184R mutant channel atdifferent voltages (n¼ 5). The lines are fits with the following equation to obtain Kd at different voltages: f¼ {1/[1þ ([SPM]/Kd)]}, where f isthe relative current, [SPM] is the concentration of SPM, and Kd is the apparent dissociated constant of SPM. The Kd values are �3,300,�1,600, �1,130, �1,030, �940, and �870mM at �100, �60, �20, þ20, þ60, and þ100mV, respectively. C: The Kd from part B is plotted againstvoltage. The Kd data of the WT channel (red line) are taken from Figure S2D and shown here for comparison. The IR index (defined by theratios between the Kd values at �30 and þ30mV) is �2 for the A184R mutant channel, but is �140,000 for the WT channel.



through the A184R mutant channel (Fig. 9A). The kinetics ofthis phase mimicking channel activation apparently carry littlevoltage-independence (see also Fig. 10A,B), and can beobserved only in low but not in high internal Kþ. On the otherhand, the external Kþ does not seem to have a similar effect

(Fig. 9B). Figure 9C shows that the percentage of the slowcomponent can be described by a one-to-one binding curve.This implies that a Kþ binding site with a Kd of �5–8mM isresponsible for the opening of the A184R mutant channel at�160 to�200mV.While internal Naþmay have a comparable

Fig. 8. Intrinsic inward rectification of WT, A184R, E224G/A184R, E224G, and E224Q mutant channels. A: The sample sweeps wererecorded in symmetrical 100mM Kþ solutions containing no SPM for the WT, A184R, E224G/A184R, E224G, and E224Q mutant channels.The voltage across the patch membrane was first held at 0mV and was hyperpolarized to �100mV for 10msec, and then stepped to differenttest voltages between �150 and þ100mV for �100msec in 10mV increment. The green dotted line indicates the zero current level. B:Normalized IV curves for the WT, E224Q, E224G, E224G/A184R, and A184R mutant channels (The current amplitude at each voltage wasnormalized to that at �150mV) (n¼ 5–8 for each different channels). C: The increase of current for each 10mV of increasedhyperpolarization is documented for �100 to �150mV. Each current increase is divided by the increase of next 10mV to give the conductanceratio, which is plotted against the voltage (see text for details). The lines are linear fits with Y-intercept of 0.97� 0.02 (WT); 0.93� 0.03(E224G); 0.84� 0.04 (E224Q); 0.79� 0.02 (E224G/A184R); and 0.78� 0.03 (A184R). On the other hand, the slope of these linear fits are all veryclose to 0 (�0.0012 and 0.0004), indicating the essentially voltage-independent nature of the conductance ratio for each channels (n¼ 5–8).



Fig. 9. The critical role of internal but not external Kþ and Naþ in “opening” of the A184R mutant channel. A: The patch which expressedA184R mutant channels was first hyperpolarized from the 0mV holding potential to �100mV for 10msec, and then stepped to different testvoltages between �200 and 100mV for �150msec in 10mV increment in symmetrical 10–100mM Kþ solution. Note the slow developmentphase of the inward currents at negative potentials in low but not in high Kþ solution. The green dotted line indicates the zero current level. B:The sample sweeps demonstrate that the slow development of inward currents is dependent on the internal rather than external Kþ in A184Rmutant channel. The pulse protocol is the same as that in part A. The green dotted line indicates the zero current level. C: The fraction of theslow component of the inward currents is defined by 1� [(initial current amplitude)/(steady-state current amplitude)] at �160, �180, and�200mV, and is smaller in higher concentration of Kþ (n¼ 6, 5, 3, 4, and 6 for 100, 50, 30, 20, and 10mM symmetrical Kþ

o/Kþi, respectively).

The dashed lines are the best fit of the form with a function: percentage of slow-developing current¼ 1/[1þ ([Kþi]/Kd)], where the [Kþ

i] is the(internal) Kþ concentration (in mM, the transverse axis), and Kd is 4.6, 7.0, and 8.2mM at �160, �180, and �200mV, respectively. D: Thesame pulse protocol as that in part A was applied to elicit A184R mutant channel currents in solutions containing 100mM external Kþ and100mM internal Naþ (and 0Kþ) or 50mM internal Naþþ 50mM internal Kþ. Naþ seems to have a similar effect on the slow-developmentinward currents to Kþ. The green dotted line indicates the zero current level. E: The sample sweeps were record in solutions containing100mM external Kþ and 100mM internal Kþ or 200mM internal sucrose for the WT and A184R mutant channels, and the pulse protocol isthe same as that in part A. TheWT data are from the same patch. The IV curves for theWT and A184R mutant channels are presented in therightmost part. Note the absence of either inward or outward currents if all internal cations are replaced by sucrose in A184R mutant channelbut essentially the same macroscopic conductance at negative potentials (e.g., �150 to �200mV, solid lines) in the WT channel whether theinternal cations are present or not (�24.1 and �24.3 pS in symmetrical 100mM Kþ and 100mM external Kþþ 200mM internal sucrose,respectively). The green dotted line indicates the zero current level.



Fig. 10. The voltage dependence of the “opening” and “closing” kinetics of the A184R mutant channel in symmetrical 10–20mM Kþ. A: Insymmetrical 10mM Kþ, the patch containing A184R mutant channels was first hyperpolarized from the 0mV holding potential to �100mVfor 10msec, followed by a step depolarization to þ100mV for 60msec, and then stepped to different test voltages between �200 and�100mV for �100msec in 10mV increment. A short (10msec) depolarizing pulse to þ100mV was subsequently given for 1msec, after whicha second test pulse to different test voltages between �200 and �100mV for �100msec in 10mV increment was given again. Note the slightlydifferent initial current amplitude but very similar slow-developing kinetics of the inward currents on the two sides of the short (1msec)depolarizing pulse. The green dotted line indicates the zero current level. B: The reciprocals of the time constants obtained from thedevelopment phase of inward currents in the first test pulse of part A are plotted against membrane voltage (between �200 and �150mV) ina semilogarithmic scale (n¼ 4). The dashed line is a linear regression fit of the form: 1/tauopen¼ 0.108� exp(0.16V/25)msec�1, where V is themembrane voltage in mV, demonstrating little voltage-dependence of the kinetics of development of the inward currents. C: The patchcontaining A184R mutant channel was first hyperpolarized from the 0mV holding potential to �200mV for 100msec, followed by a graduallylengthened gap depolarization to þ100mV, and then stepped back �200mV for �100msec. The sweeps are arranged so that the currents inthe second pulse are gradually shifted rightwards as the gap depolarization is gradually lengthened (by 1msec between each sweep). Thegradually growing inward current at the second �200mV pulse is used as a measure of the closing kinetics of A184R mutant channel at þ100to �60mV, and is described by a monoexponential function. The dotted line in the enlarged part of the figure is the monoexponential fit to thedata with a time constant of �1.8msec (at þ20mV in symmetrical 20mM Kþ solution). The green dotted line indicates the zero current level.D: The reciprocals of the time constants obtained from part C are plotted against membrane voltage in a semilogarithmic scale (�60 toþ100mV) in symmetrical 10 and 20mM Kþ solutions (n¼ 4–7). The dashed lines are fits of the form: 1/tauclose¼ 0.25� exp(0.49V/25)msec�1,and 0.46� exp(0.41V/25)msec�1 for 10 and 20mM Kþ, respectively, where V is the membrane voltage in mV.



effect to internal Kþ (Fig. 9D), the A184R mutant channel cannot even “open” to conduct any discernible (inward) Kþ

currents down to �200mV in the absence of internal cationssuch as Kþ or Naþ. In contrast, the ionic conductances are notsignificantly altered in the absence of intracellular cations in theWT channel (Fig. 9E). This finding strongly supports that thedevelopment of inward currents and disappearance of outwardcurrents in the A184R mutant channel is a gating phenomenonrather than a direct unblock/block of the channel pore byextrinsic particles. In this regard, the ionic site responsible forthe opening of the A184Rmutant channel is not so selective forKþ in view of the similar effect of internal Naþ. We alsodemonstrate that the A184R mutant channel does not “openimmediately” upon membrane potential change to get inwardcurrents in symmetrical 10mM Kþ solution (Fig. 10A). Thedevelopment phase of the inward currents could be describedwith an apparently voltage-independent time constant whichremains almost the same between �150 and �200mV(Fig. 10B). On the other hand, there is a modest but definitevoltage dependence that the channel closes e-fold faster per50–60mV depolarization (Fig. 10C,D). Moreover, the closingkinetics are in general�1.8-fold faster in 20mM than in 10mMsymmetrical Kþ. These latter findings once more support theview that the development of inward currents anddisappearance of outward currents are gating phenomenarather direct unblock/block of the channel pore by extrinsicparticles, otherwise the “blocking” kinetics should be slower inhigher than in lower ambient Kþ in view of possiblecompetition between the permeating and blocking ions.

The slow component of inward currents in the A184Hmutant channel is much more evident in pHi 6.4 than 8.4

To pursue the origin of the voltage dependence of A184Rmutant channel closure (e-fold faster per 50–60mVdepolarization), we examined the gating kinetics of A184Hmutant channel in pHi 6.4 and 8.4. The A184H mutant channelopening also carries no significant voltage dependence in bothpHi (Fig. 11A). On the other hand, there is a mild but definitevoltage dependence of A184H channel closure in pHi 6.4 (e-fold faster per 125mV depolarization) (Fig. 11B,C). However,the voltage dependence is greatly decreased in pHi 8.4. Thesefindings demonstrate that the voltage-dependent gating inA184R mutant channel is mostly ascribable to the charge onresidue 184. It is plausible that the charged side chain of residue184 takes a position at d up to 0.5 (from inside) in the closedpore, but may be introduced into a position with a d� 0 in theopen channels. The closed to open channel transition thus mayinvolve an inward displacement (toward the cytoplasmicdomain of the pore) of the side chain of residue 184 (or anequivalent change of the electrical field across the pore) withthe highest energy transitional state happening also at a d closeto 0.5 (and hence the voltage dependent closing kinetics butvoltage-independent opening kinetics). The large changes of dexperienced by the protonated histidine in position 184 is alsoconsistent with the idea that the position is located very closeto or right at a pore region capable of significant caliberchanges.

DiscussionColocalization of intrinsic as well as extrinsic rectificationand gating functions of the Kir2.1 channel pore at �A184

The apparent inward rectification in the Kir2.1 channel can beeither extrinsic or intrinsic in origin. The extrinsic rectificationis associated with a flow-dependent block (e.g., the “long-poreeffect” of the permeating ion on the blocking ion SPM) andwould have a dramatic change at the reversal potential wherethe preponderant ionic flow changes its direction. The

extrinsic rectification thus is best characterized by the IR index(a comparison of the blocking effects right on the two sides ofthe reversal potential, e.g., Fig. 7C). On the other hand, theintrinsic rectification is most likely ascribable to anasymmetrical barrier for the permeate ions to pass. Theintrinsic rectification thus would become more manifest if thepeak energy of the asymmetrical barrier becomes higher and/or the two slopes flanking the peak are more asymmetrical (interms of electrical distance). In this regard, it is most interestingthat single point mutation A184R essentially abolishes the flow-dependent (flux-coupling) extrinsic rectification but markedlyintensifies the voltage-dependent intrinsic rectification.Because A184 is located at the inner end of the bundle crossingregion (Fig. 6A), a pore segment markedly narrower than andthus more suitable for the generation of flux-couplingphenomenon than the cytoplasmic domain (where E224 islocated), the same permeation barrier that A184 and E224mutation act on to abolish extrinsic rectification and toincrease intrinsic rectification is most likely at A184. In otherwords, it is plausible that there are crucial interactionsoccurring in the cytoplasmic domain to allosterically keep theconformation of the pore at residue A184 in the bundlecrossing region (Fig. 6B–D). Residue A184 thus may be viewedas the other critical structure beside the selectivity filter in theKir2.1 channel. In this regard, it is interesting to note that thebundle crossing region has been reported to function as theactivation/deactivation pore gate in many channels (Kurataet al., 2010; Khurana et al., 2011), and that A184R pointmutation tends to close the pore, which can be opened byintracellular but not extracellular cations. Unlike the selectivityfilter, which is located at the external pore mouth and isresponsible for the Kþ ion selectivity of the pore, the ionic siteinvolving A184 is considerably less selective for Kþ. This isbecause SPM can also permeate through this site, and becauseNaþ displays a fairly similar gating effect to Kþ when binding tothe sites (Fig. 9, also see below). The intriguing colocalization ofthe mechanisms underlying intrinsic and extrinsic inwardrectification, and even channel gating, in the same pore segmentmay be related to the evolution of molecular design of ionchannels (see below).

Contribution of the cations internal to the bundlecrossing region to the opening of Kir2.1 channel pore

We demonstrate the potential gating movement of the bundlecrossing region in the A184R mutant channel by altering theinternal Kþ concentration (Figs. 9 and 12). Our findings indicatethat Kþ binding to a site internal to the bundle crossing regionwith an apparent dissociation constant of 5–8mM will help tostabilize the A184R mutant channel pore in an “open”conformation (Fig. 9), although no similar effect could bedemonstrated experimentally in the WT channel. This site ofpotential gating control is essentially nonselective for differentcations such as Kþ and Naþ and may be partly reminiscent ofthe opening of the selectivity filter region of Kþ channels byexternal Kþ (Doyle et al., 1998; Jiang et al., 2002), although inthe latter case the opening effect is highly selective for Kþ.Consistent with these findings, molecular dynamics simulationshow that E224Q and especially A184R mutations maydecrease the pore caliber in the bundle crossing region(Fig. 12A). The inter-residue distance of A184 in the absence ofintracellular Kþ ions is �13 and �8Å for the WT and A184Rmutant channels, respectively (Fig. 12B–D). This inter-residuedistance increases greatly in the presence of Kþ or Naþ ions inthe A184R mutant channel pore (Fig. 12C,D). In other words,although the WT Kir2.1 channel does not seem to needintracellular cations to remain open from the experimentaldata (Fig. 9E), cations in the cytoplasmic domain may still have asmall but definite effect on the pore caliber at the bundle



crossing region in the WT channel based on moleculardynamics simulation. Accordingly, we demonstrate thatinternal SPMmay acts as a gate opener possibly by binding to itssuperficial site (Fig. 5), although inward rectifying block of Kþ

flow undoubtedly comprises the most prominent effect of SPM

on the Kir 2.1 channel currents. In this regard, the SPM at thesuperficial site could also serve as a docking form ready for thepermeation to the deep site, where SPM block is very muchflowdependent and thus greatly contributes to the inwardrectification feature of the channel. These actions of the SPM in

Fig. 11. The different opening kinetics in A184H mutant channel in different internal pH. A: Similar experiments to that in Figure 10A wererepeated in the A184H mutant channel in symmetrical 10mM Kþ at pHi of 6.4 and 8.4 (n¼ 5–8). The dashed lines are linear regression fits ofthe form: 1/tauopen¼ 0.09� exp(�0.22V/25)msec�1 (10mM Kþ, pHi 8.4), or 0.29� exp(�0.12V/25)msec�1 (10mM Kþ, pHi 6.4), where V is themembrane voltage in mV. B: Similar two-pulse protocol to that in Figure 10C were repeated in the A184H mutant channel in symmetrical10mM Kþ (n¼ 5–8), except that the gap depolarization was fixed atþ100mV. Note that the slow tail of SPM unblock is present when pHi iseither 6.4 or 8.4 in the A184Hmutant channel. The green dotted line indicates the zero current level. C: The reciprocals of the time constantsobtained from part B are plotted against membrane voltage in semilogarithmic scale (�60 to þ80mV) in symmetrical 10mM Kþ. The dashedlines are linear regression fits of the form: 1/tauclose¼ 0.25� exp(0.1V/25)msec�1 (pHi 8.4), or 1.18� exp(0.2V/25)msec�1 (pHi 6.4), where Vis the membrane voltage in mV, demonstrating more manifest voltage-dependence at pHi 6.4 than pHi 8.4 in the A184H mutant channel.



Fig. 12. Homology modeling of the Kir2.1 and Kir2.2 channels. A: The homology model protocol is the same as that in Figure 6. Two subunitsof the channel are shown in solid-ribbon model (left upper part). A regional view of residues A184 and E224 are shown in the boxed parts.Note the decrease in pore caliber (diagonal distance from Ca to Ca) at positions 184 and 224 caused by E224Q or A184R single mutation. B:In the presence of 40 water molecules, 0–7 Kþ ions were applied to the central cavity of the channel. The Kir2.1 tetramer is viewed from theintracellular side. The bundle crossing region at residue A184 is shown in CPK model. The four different subunits are colored green, blue,yellow, and pink. The side chains of A184 residues are presented in CPK model. C: The homology model of the A184R mutant channel. Notethe very prominent pore-opening effect of Kþ ions. D: In the WT and A184R mutant channel, the diagonal distance between residues of 184(from Ca to Ca or from tip to tip) is plotted against the number of intracellular Kþ ions. Note the distance is gradually increased with moreKþ ions up to �4Kþ, and then reaches an apparent plateau value in the A184R mutant channel. E: In the presence of 40 water molecules, 0 or7Kþ ions were applied to the central cavity of the A184R mutant channel. There is a prominent repulsion effect “abducting” the side chains ofarginines to an externally deflected position and thus increasing the Ca–Ca diagonal distance. The side chain of the A184 residue is shown inCPK model, and the hydrogen atoms are colored white, nitrogen, blue; oxygen, red; and carbon, gray.



Fig. 13. A schematic model illustrating the pivotal role of residue A184 in both inward rectifying and gating features in the bundle crossingregion of the Kir2.1 channel pore. A: The WT Kir2.1 channel consists of transmembrane and cytoplasmic domains. The ion conductionpathway is marked by the arrow. The E224Q mutation would allosterically make the bundle crossing region of the pore close to A184 nolonger stay in a stably “open” conformation, and thus decrease the rate of ion passage across this point and the inward rectification featuresof, for example, SPM block of the channel. B: A184 in the bundle crossing region is potentially the most critical point because A184R mutantchannel can completely stop the current flow through the pore by “closing” this region. Most interestingly, internal Kþ, or equally effectivelyNaþ can open this narrowest region of the A184R mutant channel pore. The lack of additive effect of concomitant E224 mutation in thisregard further indicates the same site of actions of these mutations. C: SPM may enter the superficial site close to or involving E224 (andE299) in the WT Kir2.1 channel pore in a linear form, and then binds to the deep site involving D172 with a more or less curly form. The SPMat the superficial site may not only be a docking form for the subsequent binding to the deep site, but also be a gating particle capable of“opening” the bundle crossing region of the pore in the WT Kir2.1 channel. E224 mutations therefore decrease the differences in the SPMbinding rates (kon, Fig. 4F) or the unbinding rates (1/taudecay, Figs. 5C and S1C) between the conditions with and without pre-existing or thepresence of ambient SPM. In this regard, the Kir2.1 channel pore must be able to accommodate at least two SPM molecules simultaneously(Lopatin et al., 1995).



the superficial site may altogether explain the discrepancy inthe blocking kinetics of SPM measured with or without pre-existing SPM in the system (Figs. 2–4).

The significance of an alanine residue at the criticalposition 184

A small-size residue, such as alanine, at position 184 seems to beimportant to secure an open conformation of the pore at thebundle crossing region (Fig. 6A,B). Mutations of alanine intomedium-size residue such as glutamate or glutamine produce nosignificant gating changes, whereas mutations into residue withlarge side chains such as leucine or tryptophan would result inchannels expressing little or no currents (data not shown). In thisregard, it is interesting that mutation of alanine into residue withlarge but positively charge side chains still results in a conductingpore, if only there are enough cations in the intracellular milieu(cytoplasmic domain). It seems that the potential electrostaticrepulsions would “abduct” the side chains to a more externallydeflected position to increase the Ca–Ca distance andtherefore the pore caliber (Fig. 12E). Because the bundlecrossing region or position A184 is also involved in the flux-coupling segment responsible for the flow-dependent block ofSPM in the Kir2.1 channel, the inward rectifying feature ofinternal SPM block is completely disrupted with A184 mutation.Moreover, the positively charged side chain could be situatedvery close to the flux-coupling segment. The preponderantlyinward flux could then assist in the side chain abduction, whichmay not yet reach a saturation level in low intracellular cations.The assisted movement of the side chain and consequentwidening of the pore seem to proceed with a time constant of�50msec (Figs. 9 and 10). The functional localization of theactivation–deactivation gate at the bundle crossing regionprobably is a common design in the two-transmembrane-segment ion channel superfamily, including voltage-dependentNaþ, Kþ, and Ca2þ channels and the ionotropic glutamatereceptors (Doyle et al., 1998; Kitaguchi et al., 2004; Sunamiet al., 2004). In this regard, our data also indicate that the E224Gmutation allosterically makes the bundle crossing region of thepore close to A184 no longer stay in a stably “open”conformation, and thus decreases the rate of ion passage acrossthis point as well as weakens the inward rectification of theKir2.1 channel (Figs. 4, 8, and S2A, see also Fig. 13A,B). Thepore-opening effect of the SPM in the superficial site in the WTchannel (with an alanine at position 184) may also be related tothe allosteric (conformational) effect of the cytoplasmic domainon the bundle crossing region of the pore (Figs. 6B–D and 13B).The Kir2.1 pore is thus well adapted to the intracellular milieuand even sophisticatedly uses intracellular molecules toaccomplish not only direction-specific permeation block(and consequently inward rectification), but also a gated accessto the ion conduction pathway (Fig. 13C).

Implication on the evolution of functional design of Kirchannels

The Kir channels are well known and functionally characteristicfor their inward rectification features of ionic transport, whichwould effectively keep the cell membrane potential close to thereversal potential of Kþ at rest but avoid losing too muchintracellular Kþ during prolonged depolarization (Hagiwaraet al., 1976; Ishikawa et al., 1993; Nichols and Lopatin, 1997;Oliver et al., 2000; Lu, 2004). In this regard, it is interesting whythe channel would have the core structure responsible for thiskey inward rectification reside at the bundle crossing region,which may also be the gate of the channel (Clarke et al., 2010;Khurana et al., 2011; Huang and Kuo, 2014). A channel gatetends to have variable conformations, which would berelatively incompatible with the delicate functional coupling

between the ionic sites responsible for inward rectification. Kirchannels are evolutionarily old proteins and thus are found inmany phylogenetically different organisms (Kuffler andNicholls, 1966; McKinney and Gallin, 1988; Greger, 1990). SPMand Mg2þ block of the pore probably are also old phenomena,evolving at least before the make of the ionotropic glutamatechannel, considering that the ionotropic glutamate channelshave their pore structure (M1-pore-M3) directly taken fromthe two-transmembrane-segment Kþ channel (M1-pore-M2),and are also characterized by the same SPM and Mg2þ blockwith prominent flux-dependent features (Yang et al., 2010).We would propose that SPM block of Kir channels might beviewed as a very early, if not the earliest, endeavor to put asubstantial gating control in the most ancient or prototypicalion channels, the two-transmembrane-segment Kþ channel(Kurata et al., 2007, 2010; Clarke et al., 2010; Khuranaet al., 2011). The cytoplasmic domain, which contains E224/E299 and R228/R260, probably was originally added to theM1-pore-M2 motif for the control of the conformation of thebundle crossing region, that is, for gating (or activation–deactivation) control rather than for making the sequentialsites for inward rectification. When SPM was picked from theintracellular milieu as a key gating controller, its binding sites inthis gating region were gradually modified through evolutionalselection. Finally a stretch of contiguous ionic sites wereformed in the pore to fit the regularly spaced charges in SPM.This stretch in the pore, because it was made for an intimateinteraction with SPM, does have a narrow caliber and thusmake a flux-coupling segment. Some of these prototypicalpores bearing functional coupling between these single-fileionic sites and thus flux-coupling features were again selectedfor energy conservation associated with inward rectification(less intracellular Kþ loss during cellular activities). As theevolutionary advantage of inward rectification (energy con-servation) became more evident, some of these Kir channelslargely “gave up” their original attempt of gating control at thisbundle crossing region, and more or less “fixed” theconformation of this region to assure the inward rectificationfeature of ion permeation. This is perhaps why there is adefinite, albeit modest, gating control by internal SPM in thebundle crossing region of the WT Kir2.1 channel. Thecoexistence of the key structures for the inward rectificationand channel gating in the same pore region is therefore verylikely a molecular evolutionary aftermath sculpted by theinteractions between the Kir channel protein and itsenvironment.

Conclusion

Point mutation A184R not only renders SPM block no longerflow-dependent or inward-rectifying, but also tends to closethe channel pore. The closed pore of A184R mutant channelcan be opened by intracellular but not extracellular cationssuch as Kþ and Naþ. Moreover, an asymmetrical barrierpotentially contributory to the intrinsic inward rectification ofthe Kir2.1 channel pore is mademanifest with A184Rmutation.In the meanwhile, the blocking SPM in the deep site unbindsfaster with higher intracellular SPM. SPM thus is not only a poreblocker but also a gating particle capable of opening the pore atthe bundle crossing region.We conclude that A184 at the innerend of the bundle crossing region marks a pivotal point of theKir2.1 channel, collaborating channel gating and ionpermeation, and making both extrinsic and intrinsic inwardrectification features of the pore.

Acknowledgments

This work was supported by National Health ResearchInstitutes, Taiwan, grant NHRI-EX101-10006NI and NHRI-



EX102-10006NI (to C.C.K.), and also by National ScienceCouncil, Taiwan, grant NSC95-2320-B-002-064-MY3 (to C.C.K.). We are grateful to Dr. Ru-Chi Shieh (Academic Sinica,Taiwan) for the gifts of the Mouse macrophage Kir2.1DNA clone.

Literature Cited

Berendsen HJC, Postma JPM, Vangunsteren WF, Dinola A, Haak JR. 1984. Molecular-dynamics with coupling to an external bath. J Chem Phys 81:3684–3690.

ChangHK, Yeh SH, Shieh RC. 2003. The effects of spermine on the accessibility of residues inthe M2 segment of Kir2. 1 channels expressed in Xenopus oocytes. J Physiol 553:101–112.

ChangHK, Yeh SH, Shieh RC. 2005. A ring of negative charges in the intracellular vestibule ofKir2. 1 channel modulates Kþ permeation. Biophys J 88:243–254.

Clarke OB, Caputo AT, Hill AP, Vandenberg JI, Smith BJ, Gulbis JM. 2010. Domainreorientation and rotation of an intracellular assembly regulate conduction in Kirpotassium channels. Cell 141:1018–1029.

Doyle DA,Morais Cabral J, Pfuetzner RA, KuoA, Gulbis JM, Cohen SL, Chait BT, MacKinnonR. 1998. The structure of the potassium channel: Molecular basis of Kþ conduction andselectivity. Science 280:69–77.

Fujiwara Y, Kubo Y. 2002. Ser165 in the second transmembrane region of the Kir2.1 channel determines its susceptibility to blockade by intracellular Mg2þ. J Gen Physiol120:677–693.

Greger R. 1990. An electrophysiological approach to the study of isolated perfused tubules.Methods Enzymol 191:289–302.

Guo D, Lu Z. 2003. Interaction mechanisms between polyamines and IRK1 inward rectifierKþ channels. J Gen Physiol 122:485–500.

Hagiwara S, Takahashi K. 1974. The anomalous rectification and cation selectivity of themembrane of a starfish egg cell. J Membr Biol 18:61–80.

Hagiwara S, Miyazaki S, Rosenthal NP. 1976. Potassium current and the effect of cesium onthis current during anomalous rectification of the egg cell membrane of a starfish. J GenPhysiol 67:621–638.

Hansen SB, Tao X, MacKinnon R. 2011. Structural basis of PIP2 activation of the classicalinward rectifier Kþ channel Kir2. 2. Nature 477:495–498.

Hille B. 2001. Ionic channels of excitable membrane. Sunerland, MA: Sinauer assoc.Holmgren M, Smith PL, Yellen G. 1997. Trapping of organic blockers by closing of voltage-dependent Kþ channels: Evidence for a trap door mechanism of activation gating. J GenPhysiol 109:527–535.

Howe MW, Feig SL, Osting SM, Haberly LB. 2008. Cellular and subcellular localization ofKir2. 1 subunits in neurons and glia in piriform cortex with implications for Kþ spatialbuffering. J Comp Neurol 506:877–893.

Huang CW, Kuo CC. 2014. The bundle crossing region is responsible for the inwardlyrectifying internal spermine block of the Kir2. 1 channel. Pflugers Arch 466:275–293.

Ishikawa T, Wegman EA, Cook DI. 1993. An inwardly rectifying potassium channelin the basolateral membrane of sheep parotid secretory cells. J Membr Biol 131:193–202.

Jiang Y, Lee A, Chen J, Cadene M, Chait BT, MacKinnon, R. 2002. The open poreconformation of potassium channels. Nature 417:523–526.

Karschin C, Dissmann E, Stuhmer W, Karschin A. 1996. IRK(1-3) and GIRK(1-4) inwardlyrectifying Kþ channel mRNAs are differentially expressed in the adult rat brain. J Neurosci16:3559–3570.

Khurana A, Shao ES, Kim RY, Vilin YY, Huang X, Yang R, Kurata HT. 2011. Forced gatingmotions by a substituted titratable side chain at the bundle crossing of a potassiumchannel. J Biol Chem 286:36686–36693.

Kitaguchi T, Sukhareva M, Swartz KJ. 2004. Stabilizing the closed S6 gate in the Shaker Kvchannel through modification of a hydrophobic seal. J Gen Physiol 124:319–332.

Kubo Y, Murata Y. 2001. Control of rectification and permeation by two distinct sites afterthe second transmembrane region in Kir2. 1 Kþ channel. J Physiol 531:645–660.

Kuffler SW, Nicholls JG. 1966. The physiology of neuroglial cells. Ergeb Physiol 57:1–90.

Kurata HT, Akrouh A, Li JB, Marton LJ, Nichols CG. 2013. Scanning the topography ofpolyamine blocker binding in an inwardly-rectifying potassium channel. J Biol Chem288:6591–6601.

Kurata HT, Cheng WW, Arrabit C, Slesinger PA, Nichols CG. 2007. The role of thecytoplasmic pore in inward rectification of Kir2. 1 channels. J Gen Physiol 130:145–155.

Kurata HT, Zhu EA, Nichols CG. 2010. Locale and chemistry of spermine binding in thearchetypal inward rectifier Kir2. 1. J Gen Physiol 135:495–508.

Lopatin AN, Makhina EN, Nichols CG. 1995. The mechanism of inward rectification ofpotassium channels: “long-pore plugging’‘ by cytoplasmic polyamines. J Gen Physiol106:923–955.

Lopatin AN, Nichols CG. 1996. [Kþ] dependence of polyamine-induced rectification ininward rectifier potassium channels (IRK1, Kir2. 1). J Gen Physiol 108:105–113.

Lou KL, Huang PT, Shiau YS, Liaw YC, Shiau YY, Liou HH. 2003. A possible molecularmechanism of hanatoxin binding-modified gating in voltage-gated Kþ-channels. J MolRecognit 16:392–395.

Lu Z. 2004. Mechanism of rectification in inward-rectifier Kþ channels. Annu Rev Physiol66:103–129.

Lu Z, MacKinnon R. 1994. Electrostatic tuning of Mg2þ affinity in an inward-rectifierKþ channel. Nature 371:243–246.

Matsuda H. 1988. Open-state substructure of inwardly rectifying potassium channelsrevealed by magnesium block in guinea-pig heart cells. J Physiol 397:237–258.

Matsuda H, Hayashi M, OkadaM. 2010. Voltage-dependent block by internal spermine of themurine inwardly rectifying Kþ channel, Kir2. 1, with asymmetrical Kþ concentrations.J Physiol 588:4673–4681.

Matsuda H, Oishi K, Omori K. 2003. Voltage-dependent gating and block by internalspermine of the murine inwardly rectifying Kþ channel, Kir2. 1. J Physiol 548:361–371.

McKinney LC, Gallin EK. 1988. Inwardly rectifying whole-cell and single-channel K currentsin the murine macrophage cell line J774. 1. J Membr Biol 103:41–53.

Miyashita T, Kubo Y. 1997. Localization and developmental changes of the expression of twoinward rectifying K(þ)-channel proteins in the rat brain. Brain Res 750:251–263.

Mott NF, Gurney RW. 1940. Electronic process in ionic crystals. Oxford, UK: OxfordUniversity.

Nichols CG, Lopatin AN. 1997. Inward rectifier potassium channels. Annu Rev Physiol59:171–191.

Oliver D, Baukrowitz T, Fakler B. 2000. Polyamines as gating molecules of inward-rectifierKþ channels. Eur J Biochem 267:5824–5829.

Pegan S, Arrabit C, Zhou W, Kwiatkowski W, Collins A, Slesinger PA, Choe S. 2005.Cytoplasmic domain structures of Kir2. 1 and Kir3. 1 show sites for modulating gating andrectification. Nat Neurosci 8:279–287.

Pruss H,Derst C, Lommel R, Veh RW. 2005. Differential distribution of individual subunits ofstrongly inwardly rectifying potassium channels (Kir2 family) in rat brain. Brain Res MolBrain Res 139:63–79.

Shiau YS, Lin TB, Liou HH, Huang PT, Lou KL, Shiau YY. 2002. Molecular simulationreveals structural determinants of the hanatoxin binding in Kv2. 1 channels. J Mol Model8:253–257.

Shin HG, Lu Z. 2005. Mechanism of the voltage sensitivity of IRK1 inward-rectifierKþ channel block by the polyamine spermine. J Gen Physiol 125:413–426.

Sunami A, Tracey A, Glaaser IW, Lipkind GM, Hanck DA, Fozzard HA. 2004. Accessibility ofmid-segment domain IV S6 residues of the voltage-gated Naþ channel tomethanethiosulfonate reagents. J Physiol 561:403–413.

Tao X, Avalos JL, Chen J, MacKinnon R. 2009. Crystal structure of the eukaryotic stronginward-rectifier Kþ channel Kir2. 2 at 3. 1 A resolution. Science 326:1668–1674.

Xie LH, John SA, Weiss JN. 2003. Inward rectification by polyamines in mouse Kir2. 1channels: Synergy between blocking components. J Physiol 550:67–82.

Yang YC, Lee CH, Kuo CC. 2010. Ionic flow enhances low-affinity binding: A revisedmechanistic view into Mg2þ block of NMDA receptors. J Physiol 588:633–650.

Supporting Information

Additional supporting information may be found in the onlineversion of this article at the publisher’s web-site.



Gating of the Kir2.1 channel at the bundle crossing region by intracellular spermine and other...

Documents

Transcript of Gating of the Kir2.1 channel at the bundle crossing region by intracellular spermine and other...