Molecular basis for the polyamine-OmpF porin interactions: inhibitor and mutant studies

doi:10.1006/jmbi.2000.3599 available online at http://www.idealibrary.com on J. Mol. Biol. (2000) 297, 933±945

Molecular Basis for the Polyamine-OmpF PorinInteractions: Inhibitor and Mutant Studies

Ramkumar Iyer1, Zhiqian Wu2, Patrick M. Woster2 and Anne H. Delcour1*

1Department of Biology and ofBiochemistry, University ofHouston, HoustonTX 77204-5513, USA2Department of PharmaceuticalSciences, Wayne StateUniversity, DetroitMI 48202, USA

E-mail address of the [email protected]

Abbreviations used: 1,12-D3D, 1,azadodecane; 1,12-D4D, 1,12-diamin1, 12-D5D, 1,12-diamino-5-azadodecIPTG, isopropyl b-D-thiogalactopyraspermine; SPD, spermidine; WT, w

0022-2836/00/040933±13 $35.00/0

By testing the sensitivity of Escherichia coli OmpF porin to various naturaland synthetic polyamines of different lengths, charge and other molecularcharacteristics, we were able to identify the molecular properties requiredfor compounds to act as inhibitors of OmpF in the nanomolar range.Inhibitors require at least two amine groups to be effective. For diamines,the optimum length of the hydrocarbon spacer was found to be of eightto ten methylene groups. Triamine molecules based on a 12-carbon motifwere found to be more effective that spermidine, an eight-carbon triva-lent derivative. But differences in inhibition ef®ciencies were also foundfor trivalent compounds depending on the relative position of theinternal secondary amine group with respect to the terminal groups.Finally, quaternary ammonium derivatives had no effect, suggesting thatthe nature of the terminal amine is important for the interaction. Fromthese observations, we deduce that inhibition ef®ciency in the nanomolarrange requires a 12-carbon chain triamine with terminal primary aminegroups and replacement of the eighth methylene by a secondary amine.The need for this type of molecular architecture suggests that inhibitionis governed by interactions between speci®c amine groups and proteinresidues, and that this is not simply due to the accumulation of chargesinto the pore. Together with previous observations from site-directedmutagenesis studies and inspection of the crystal structure of OmpF,these results allowed us to propose three residues (D113, D121 and Y294)as putative sites of interaction between the channel and spermine.Alanine substitution at each of these three residues resulted in a loss ofinhibition by spermine, while mutations of only D113 and D121 affectedinhibition by spermidine. Based on these observations, we suggest amodel for the molecular determinants involved in the porin-polyamineinteraction.

# 2000 Academic Press

Keywords: E. coli; patch-clamp; modulation; channel; mutants
*Corresponding author
Introduction

Polyamines form a subset of biologically activemolecules ubiquitous to both prokaryotic andeukaryotic cells (Tabor & Tabor, 1984). These com-pounds are positively charged at physiological pH,which enables associations with various cellular

ing author:

12-diamino-3-o-4-azadodecane;ane; CON, control;noside; SPD,

ild-type.

macromolecules. The naturally occurring polya-mines include putrescine, cadaverine, spermidineand spermine, of which spermine is exclusive toeukaryotic cells.

Polyamines have been shown to play a modula-tory role on a variety of eukaryotic channels, inparticular those of neural cells (Scott et al., 1993;Johnson, 1996). Spermine, for example, interactswith the NMDA receptor, in events such as earlyneural synaptogenesis and excitotoxic shock fol-lowing neural injury (Scott et al., 1993). It has beenshown that speci®c glutamate and aspartate resi-dues of the NMDA receptor may be important inthe modulation of the NMDA receptor by sper-mine, and may form part of a spermine binding

# 2000 Academic Press

mailto:[email protected]

http://www.idealibrary.com

934 Polyamine-OmpF Porin Interactions

site (Williams et al., 1995; Kashiwagi et al., 1996a).The inward recti®er K� channel, which is involvedin heart excitability, is blocked in a voltage-depen-dent manner by polyamine molecules (Lopatinet al., 1994, 1995; Ficker et al., 1994; Johnson, 1996).Spermine, spermidine and putrescine also blockthe ryanodine receptors of the cardiac sarcoplasmicreticulum (Uehara et al., 1996). Recently, Weigeret al. (1998) characterized the action of some di-and polyamines on maxi Ca2�-activated K� chan-nels. Therefore, it appears that polyamines areeffective modulators of different channel types ineukaryotes. In addition, polyamine toxins havecomplex effects on a variety of ion channels ofeukaryotic cells (Scott et al., 1993; Lee et al., 1999;Jayaraman et al., 1999). Regulation also encom-passes a variety of enzymes including the proteinkinase CK2 (Leroy et al., 1997a).

Molecular models of the interactions betweenpolyamines and proteins have been provided forthe periplasmic spermidine-binding protein PotDof Escherichia coli and for the protein kinase CK2.The binding of spermine to glutamic acid residuesin the acidic stretch of the b subunit of CK2 kinaseis shown to produce a sharp conformationalchange of the protein, leading to kinase stimulation(Leroy et al., 1997b). The crystal structure of a sper-midine-PotD complex was elucidated and revealedthe architecture of the binding site (Sugiyama et al.,1996b). The importance of charged, polar and aro-matic residues in the interaction with spermidinewas further con®rmed by site-directed mutagenesis(Kashiwagi et al., 1996b). Binding of the aminoends of spermidine to PotD is achieved by hydro-gen bonding, salt-bridge formation and electro-static interaction with discrete glutamine, aspartateand glutamate residues. In addition, ionic inter-actions are also formed between the internal aminogroup of spermidine and an aspartate residue inPotD, while the intervening methylene groupsmake van der Waals contacts with the aromaticside-chains of tyrosine and tryptophan residues(Sugiyama et al., 1996b).

The roles of polyamines within the E. coli cell arenot well characterized. They have been implicatedin the control of DNA replication and of transcrip-tion and translation of certain genes (Tabor &Tabor, 1984; Canellakis et al., 1990). They havebeen shown to associate with the outer membrane(Koski & Vaara, 1991), and their concentrations inthe periplasm or in the external milieu may varyaccording to environmental conditions (Buch &Boyle, 1985; Olson, 1993). Recently, it has beenshown that excretion of cadaverine induced atacidic pH leads to a decrease in porin-mediatedouter membrane permeability (Samartzidou &Delcour, 1999a).

Porins are trimeric transmembrane channels thatregulate the ¯ux of solutes and b-lactam antibioticsacross the outer membrane of E. coli (Nikaido,1996). Recent electrophysiological experimentsdemonstrated that all four naturally occurringpolyamines promote channel closures of the non-

speci®c E. coli porins, OmpC and OmpF, in aconcentration and voltage-dependent manner(delaVega & Delcour, 1995; Iyer & Delcour, 1997).This inhibition involves an interaction of polya-mine molecules with the inside of the pore, assuggested by the voltage-dependence of inhibition(delaVega & Delcour, 1995; Iyer & Delcour, 1997),the apparent correlation between ion selectivityand sensitivity to modulators (Samartzidou &Delcour, 1999b), and the polyamine-insensitivebehavior of an OmpC porin where the pore-exposed D105 residue of the L3 loop has beenreplaced by a glutamine residue (Liu et al., 1997).

Here, we establish correlations between the ef®-ciency and the molecular structure of variousinhibitors to gain further insight into the molecularnature of the polyamine-porin interaction. We useour results, the published structure of polyamine-protein complexes (Sugiyama et al., 1996b; Leroyet al., 1997b), the molecular modeling of polya-mines (Weiger et al., 1998) and the knowledge ofthe atomic structure of OmpF (Cowan et al., 1992)to design site-directed mutants in OmpF. Theresults from the mutagenesis study lead to sugges-tions for the architecture of a polyamine bindingsite in the pore.

Results

Length effects

Our previous characterization of the concen-tration and voltage-dependent inhibition of OmpFporin by the natural polyamines revealed thefollowing relative strengths of the four polyamines:spermine > spermidine > cadaverine > putrescine(delaVega & Delcour, 1995; Iyer & Delcour, 1997).To assess whether length and/or charge of themolecule were crucial in determining modulatoref®ciency, it was essential to determine the contri-bution of each parameter in isolation.

We ®rst isolated the effects solely due to lengthby using the diamines shown in Table 1. These dia-mines have a varying hydrocarbon spacer lengthbetween the terminal amino groups, but a constantcharge of �2 at physiological pH. For all the exper-iments presented here, the chemicals were appliedto the bath-side of the excised patches. Becauseporins display an identical asymmetric voltagedependence in patches made from live cells andfrom reconstituted liposomes, we have determinedthat the side of the patch in contact with the bathcorresponds to the periplasmic side (Buechner et al.,1990; Samartzidou & Delcour, 1998).

The gating activity of porins is assessed in con-trol conditions and after polyamines have beenapplied to the same patch. Figure 1(b) shows atypical current recording of OmpF activity in theabsence of modulator. The trace is clearly differentfrom that obtained with porin-free liposomes madeonly of phospholipids (Figure 1(a)). The baseline(BL) represents the current passing through all pre-dominantly open porins (�20-80 usually) in the

Table 1. List of compounds used in this study

Chemical formulae Name Charge

NH2-CH2-CH2-CH2-CH2-NH2 1,4-Diaminobutane(putrescine)

�2

NH2-CH2-CH2-CH2-CH2-CH2-NH2 1,5-Diaminopentane(cadaverine)

�2

NH2-CH2-CH2-CH2-CH2-CH2-CH2-NH2 1,6-Diaminohexane �2NH2-CH2-CH2-CH2-CH2-CH2-CH2-CH2-CH2-NH2 1,8-Diaminooctane �2NH2-CH2-CH2-CH2-CH2-CH2-CH2-CH2-CH2-CH2-CH2-NH2 1,10-Diaminodecane �2NH2-CH2-CH2-CH2-CH2-CH2-CH2-CH2-CH2-CH2-CH2-CH2-CH2-NH2 1,12-

Diaminododecane�2

NH2-CH2-CH2-CH2-NH-CH2-CH2-CH2-CH2-NH2 Spermidine �3NH2-CH2-CH2-CH2-CH2-CH2-CH2-CH2-NH-CH2-CH2-CH2-CH2-NH2 1,12-Diamino-5-

azadodecane (1,12-D5D)

�3

NH2-CH2-CH2-CH2-CH2-CH2-CH2-CH2-CH2-NH-CH2-CH2-CH2-NH2 1,12-Diamino-4-azadodecane (1,12-D4D)

�3

NH2-CH2-CH2-CH2-CH2-CH2-CH2-CH2-CH2- CH2-NH-CH2-CH2-NH2 1,12-Diamino-3-azadodecane (1,12-D3D)

�3

NH2-CH2-CH2-CH2-NH-CH2-CH2-CH2-CH2-NH-CH2-CH2-CH2-NH2 Spermine �4NH2-C-NH-CH2-CH2-CH2-CH2-NH2 Agmatine �2

kNH

(CH3)3 N�-CH2-CH2-CH2-CH2-CH2-CH2-N�(CH3)3 Hexamethonium �2

(CH3)3 N�-CH2-CH2-CH2-CH2-CH2-CH2-CH2-CH2-CH2-CH2-N�(CH3)3 Decamethonium �2

Polyamine-OmpF Porin Interactions 935

patch. The upward de¯ections are transient closingevents. An increase in the frequency of closuresis clearly observed when 1,8-diaminooctane(Figure 1(e)) or 1,10-diaminodecane (Figure 1(f)) isapplied to the periplasmic side of the excisedpatch. To quantify the extent of modulation, wehave measured the fold increase in the total num-ber of closures during a 20 second recording whenthe concentration of polyamine in the bath solutionis switched from 0 (control) to 1 mM. The numberof porins in the patch is variable and directly in¯u-ences the observed closing frequency in any givencondition. For this reason, it is not meaningful toreport the absolute number of closures countedfrom control and modulated traces averaged overseveral experiments. However, relative changesfrom control to modulated conditions are consist-ent between patches, and therefore we focus onreporting the fold-increase in the total number ofclosures induced by the modulator, rather than theabsolute value. The average fold increase in thenumber of closures obtained from three patcheswere 2.81(�0.08) and 3.12(�1.05), for 1,8-diamino-octane and 1,10-diaminodecane, respectively. Amilder effect is obtained in the presence of cada-verine (Figure 1(c)) or 1,6-diaminohexane(Figure 1(d)), with fold-increases in the number ofclosures of 1.99(�0.35) for cadaverine (three exper-iments) and 59(�0.18) for 1,6-diaminohexane (fourexperiments). Strikingly, the periplasmic appli-cation of 1,12-diaminododecane (Figure 1(g)) didnot result in any modulation (fold-increase of1.00(�0.20), from three experiments). Thus, we®nd that the extent of the modulation by the dia-mines is dependent on hydrocarbon chain length

with ef®ciency in the order C-10 � C-8 > C-6 �C-5 > C-12. Putrescine (C-4) was previously shownto be ineffective (Iyer & Delcour, 1997). As withthe natural polyamines, the modulation displayeda clear asymmetric voltage dependence, with inhi-bition at negative pipette voltages but little or noeffects at positive pipette potentials (data notshown). As documented by Iyer & Delcour (1997),this is not due to the orientation of porin moleculesin the membrane, but to the electrophoretic attrac-tion of bath-applied polyamines to the channel atnegative pipette potentials.

We had only limited success in testing the effectof higher concentrations of 1,8-diaminooctane or1,10-diaminodecane because they caused somemembrane instability. Lower concentrations pro-duced milder effects, and the onset of modulationis seen �0.1-0.3 mM for both compounds; this isslightly lower than that for cadaverine and 1,6-dia-minohexane, with threshold concentrations of 1-3 mM. Finally, we performed control experimentswith these polyamines and those described belowthat show that none of the compounds affect therecordings made on porin-free patches at the con-centrations lower than 1 mM.

Charge effects

To assess the impact of charges in modulatoref®ciency, we compared the effects of spermineand spermidine with their respective derivativesthat differ from each other solely by the magnitudeof charge.

Spermidine, and its lesser-charged counterpart1,8-diaminooctane, are equally effective in their

Figure 1. Current traces from a porin-free lipid patch(a) or OmpF-containing patches in the absence (b) or thepresence of 1 mM bath-applied diamines ((c)-(g)). Thepipette voltage was ÿ60 mV. The number of methylenegroups between terminal amine groups is representedas C-#. The preferred current level is marked as BL(baseline). Upward de¯ections from the baseline corre-spond to transient closures of one to several monomers(single monomer current of �2.0 pA). For example, in(b), the two left-most upward de¯ections represent clo-sures of single monomers, while the largest upwardtransition (�1.2 second after the start of the trace) corre-sponds to the closure of six monomers. Other tran-sitions, such as those at 580 ms and 1.64 second,correspond to the closures of three and four monomers,respectively. For (b)-(g), the patches contain � 50-60channels. Note the strong modulatory effects (enhancedclosing frequency) of 1,8-diaminooctane and 1,10-diami-nodecane.


ability to increase OmpF closures. The averagefold-increase in total number of closures obtainedfrom three separate experiments for each com-pound was 2.47(�0.50) in the presence of 1 mM

spermidine, and 2.81(�0.09) in the presence of1 mM 1,8-diaminooctane. Submillimolar concen-trations of spermidine or 1,8-diaminooctane pro-duced an identical mild increase in closingfrequency (data not shown).

We also compared the effects of the divalent 1,8-diaminooctane with its corresponding monovalentderivative (n-octylamine) which has only one term-inal amine. This was important in order to test iftwo terminal amino groups were essential in deter-mining the strength of the modulator. We foundthat n-octylamine was ineffective at the concen-trations tested (0.1 to 3 mM), even at a higher pip-ette voltage (ÿ80 mV) than used with thepolyamines (data not shown). We did not test con-centrations higher than 3 mM due to problemswith membrane stability. We also found that 1-aminoethanol (charge, �1) was ineffective, even inthe tens of millimolar range (data not shown). Wedid not test monoamines with longer hydrocarbonchains to avoid detergent-like effects from theseamphipathic molecules.

Synthetic derivatives with a charge of �3, butcomparable carbon chain length to 1,12-diamino-dodecane (�2) and spermine (�4), were also usedto study charge effects. The synthetic compounds(1,12-D3D; 1,12-D4D; 1,12-D5D) exert a strongmodulation of OmpF. In order to quantify the rela-tive differences between compounds and makecomparisons with our previous results with naturalpolyamines (Iyer & Delcour, 1997), we use anoperational parameter, htBLi, which represents theaverage time for which the current trace dwells atthe baseline level, in-between closing events. Thisparameter decreases as the frequency of closuresincreases, and hence is an indication of themodulator strength. To alleviate the problem of thedependency of htBLi on the number of channels inthe patch and the variability of channel numberfrom patch to patch, we calculate the ratio of htBLiobtained from traces in the presence of polyaminesto that calculated from the control trace (no polya-mine) in the same patch. Thus, we are able to poolthe data from many patches, because only relativecomparisons were made. As documented by Iyer& Delcour (1997), polyamines also induce adecrease in macroscopic current due to the closingof many channels. This inhibition appears distinctfrom the kinetic modulation because it typicallyoccurs in a higher concentration range than thekinetic effects and involves a different set ofmolecular determinants (Liu, 1999). Thus, for thecalculation of htBLi, we have restricted ouranalysis to patches and polyamine concentrationswhere macroscopic current inhibition was less than15 %.

In Figure 2, we plotted the concentration depen-dence of the ratio of htBLi in the presence to that inthe absence of polyamines. We did not intend toobtain quantitative information on the af®nity ofthe polyamines from such plots, since the datawere not obtained from patches containing a singlechannel. The graph simply illustrates the phen-

Figure 2. Concentration dependence of the OmpFclosing frequency depicted as the ratio of ht BLi in thepresence of a polyamine to that in its absence (control).The broken and continuous lines correspond to sper-mine and spermidine, respectively and are taken fromIyer & Delcour (1997). Symbols represent the averagesfrom three experiments obtained with 1,12-D4D (circles)or 1,12-D3D (triangles), or 1,12-D5D (squares). The errorbars are SD. The pipette voltages were ÿ60 mVthroughout.

Figure 3. Consecutive current traces obtained in thepresence of 0.1 mM spermine (a) or 1,12-D4D (b) high-lighting the different kinetic signatures promoted bythese two polyamines. Upward de¯ections from thebaseline (BL) represent closures. Arrows in (a) markprolonged closures involving many pores, that are typi-cal of modulation by spermine (or spermidine, or cada-verine), but are lacking in the ¯icker-inducing 1,12-D4D.The pipette voltage was ÿ60 mV. The third trace in (a)is the segment encompassing the ®rst two long closuresof the second trace, shown on an expanded time andcurrent scale. Horizontal lines were drawn throughwell-de®ned current levels by visual inspection. Aster-isks mark two well-resolved current levels separated by2.2 pA. The scale bar in (b) applies to all traces: the yaxis is 20 pA for the main traces and 9.2 pA for theexpanded trace; the x axis is 50 ms for the main tracesand 14 ms for the expanded trace.


omenological differences in ef®ciency of thecompounds. The plot for 1,12-D4D (circles) issuperimposable to that of spermine (broken line),as determined by Iyer & Delcour (1997). The othertwo synthetic derivatives, 1,12-D3D and 1,12-D5Dappear equally potent, in a concentration rangethat is intermediate between that of spermidineand the spermine/1,12-D4D pair.

The plots in Figure 2 indicate that differences inpotencies exist among the trivalent compounds(spermidine, 1,12-D4D, 1,12-D3D and 1,12-D5D).This observation precludes the possibility that themodulation of porin kinetics arises from a mereaccumulation of charges at the mouth of the pore,and hints that speci®c interactions between theprotein and the modulator underlie the inhibition.The results suggest a direct but non-linear depen-dence of inhibition ef®ciency on charge. While aminimum charge of �2 sets the basis for inhibition,other parameters, such as position of the internalamine and overall molecule length, might haveoverriding effects. Thus a comparison among theC12 series (1,12-diaminododecane, 1,12-D3D, 1,12-D4D, 1,12-D5D and spermine) indicates that thedrastic increase in modulation that accompaniesthe transition from �2 to �3 appears to arise fromthe single internal amino group which ``rescues''the inert phenotype of 1,12-diaminododecane.However, introducing one more charge in 1,12-D4D, to convert it into spermine, does not create asigni®cant difference in the ef®ciency. A modelreconciling these observations will be presented inthe Discussion.

Heterogeneity of kinetic signatures

A hallmark of the effect induced by cadaverine,spermidine and spermine is the appearance of pro-longed closures of many pores in a cooperativemanner (delaVega & Delcour, 1995; Iyer &Delcour, 1997). Examples of such behavior are thetransitions marked by arrows in the consecutivetraces of Figure 3(a). It is noteworthy that theseprolonged events typically involve the simul-taneous closing followed by re-opening of many

Figure 4. Comparison of effects by 30 mM 1,6-diami-nohexane (b) and 1 mM 1,10-diaminodecane (d) andidentical concentrations of their respective quaternaryammonium derivatives hexamethonium (c) and deca-methonium (e). A control trace is shown for comparison(a). The baseline current level is marked as BL and clo-sures are upward transitions. Pipette voltages are ÿ60mV throughout.


channels. The third trace of Figure 3(a) displays anexpanded segment of the second trace (®rst twolong closures). The horizontal lines have beendrawn through well-de®ned current levels. Tenintervals are shown on this trace, and the averageinterval current value is 2.2 pA. In particular, it isnoteworthy that the difference in current levels oflarge transitions (such as those marked by theasterisks) is also �2.2 pA. These differences cannotbe ascribed to electronic noise because they arealways greater than the rms noise obtained fromporin-free patches (�0.5-0.7 pA at ÿ 60 mV). Wepostulate that 2.2 pA represents the monomericcurrent (and not a substate), because there is nofavored large event that might represent the fullsingle-channel conductance (see Materials andMethods). The trace clearly shows a large array ofconductance levels that were attained by the simul-taneous closure of multiple unit conductances. Evi-dence for cooperative behavior of porins in E. coliand other organisms has also been given in otherpublications (Delcour et al., 1989a,b; Berrier et al.,1992; delaVega & Delcour, 1995; Iyer & Delcour,1997; Besnard et al., 1997).

This consistent high level of cooperativity is dif-®cult to reconcile with an inhibitory mechanismsolely based on physical blocking of the open pore,since it is unlikely that many channels would fre-quently be blocked and unblocked at exactly thesame time. Thus, we have proposed that polya-mines regulate the kinetics of gating between openand closed states (delaVega & Delcour, 1995; Iyer& Delcour, 1997). Surprisingly, 1,12-D4D neverpromotes such long-lived closed states, but onlyinduces a ¯ickering activity (Figure 3(b)). The samegating pattern was also observed with 1,12-D3Dand 1,12-D5D, but at slightly higher concentrations(data not shown).

Specificity of the interaction

Polyamines are smaller than the 600 Da cut-offfor OmpF porin, and are likely to permeate thepore. Is the observed inhibition of porin by polya-mines simply due to the fact that they penetratethe pore and block it? This is unlikely because per-meating solutes such as glucose, lysine and aspar-tate did not produce any increased frequency ofclosures (data not shown). In light of the resultspresented above, one might argue that these com-pounds are not large enough or charged enough toinhibit porins. To address this issue, we havetested two quaternary ammonium salts, hexa-methonium (hexane-1,6-bis[trimethylammonium])and decamethonium (decane-1,10-bis[trimethylam-monium]), and compared them with their aliphaticcounterparts, 1,6-diaminohexane and 1,10-diamino-decane, respectively. While 1,10-diaminodecaneand 1,6-diaminohexane were effective at sub-milli-molar and tens of millimolar concentrations,respectively, decamethonium and hexamethoniumhad no effect at comparable concentration rangesfor the same applied pipette voltage of ÿ60 mV

(Figure 4). The loss in effect is more pronounced inthe comparison of 1,10-diaminodecane with deca-methonium, as 1,6-diaminohexane is by itself notas effective in inducing closures in OmpF(Figure 4).

We have also tested a different type of polya-mine, agmatine, which is a decarboxylation pro-duct of arginine. It is a ®ve-carbon chain, muchlike cadaverine, but with one of the terminal pri-mary amino groups replaced by the guanidinomoiety (see Table 1). Thus, agmatine has a lengthand a divalent charge comparable to that of cada-verine (as the guanidino group behaves as a singletitratable charge), but an enhanced ability for mul-tiple hydrogen bonding due to the guanidinogroup. Agmatine can produce inhibition of porinsbut is less effective than cadaverine. Concen-trations of at least 60 mM are required to see a 1.6-fold increase in total number of closures, althoughlower concentrations can induce an appreciabledecrease in the total number of open pores. At1 mM, agmatine is not an inhibitor with a1.19(�0.17)-fold increase in total number of

Figure 5. (a) Cartoon representation similar to that byLeroy et al. (1997a) of some of the polyamines used inthis study. The molecules have been aligned at one ofthe terminal amine groups to highlight the molecularlength and the relative position of the amine groups,shown as large gray circles (methylene groups areshown as small black circles). The 9 AÊ and 4 AÊ lengthsare those of 1,8-diaminooctane and 1,3-diaminopropanefrom Weiger et al. (1998). The abbreviations are: SPD,spermidine; D4D, 1,12-D4D; and SPM, spermine. (b)Cross-section of the constriction zone of OmpF viewedfrom the periplasmic side (from Cowan et al., 1992),showing the location of the mutated residues. For sakeof clarity some extracellular loops have been clipped.Y294 is not protruding in between the barrel and theloop, but is exposed to the lumen in a slightly moreextracellular plane than D121. (c) Side-view of an OmpFmonomer showing the location of the mutated residues(D113 and D121 are on the intraluminal L3 loop). Thebarrel wall in the foreground has been eliminated fromthe picture to obtain a clear view of the inside of thepore. The three residues are directly accessible to thepore lumen.


closures, compared to a �twofold increase in thecase of cadaverine.

Mutant studies

The results presented above suggest that speci®cinteractions exist between the porin protein andthe inhibitory polyamine molecule. Based on thedifferent inhibition ef®ciencies observed for thepolyamines reported here, and the published con-®gurations of polyamine binding sites on the pro-tein kinase CK2 (Leroy et al., 1997a,b) and theE. coli spermidine-binding protein PotD (Sugiyamaet al., 1996a,b), we propose that ionic or H-bondinteractions occur between the amine groups of thepolyamine molecule and polar or negativelycharged residues of porins. Since 1,12-D4D andspermine appear equally effective, we propose thatminimally the binding site should comprise resi-dues that can interact with amine groups of 1,12-D4D. This would place three main sites of inter-action, separated by distances of 9 AÊ (the lengthbetween the terminal amines of diaminooctane)and 4 AÊ (the length between the terminal aminesof diaminopropane) (Weiger et al., 1998)(Figure 5(a)).

To test this hypothesis, we have targeted resi-dues in OmpF for site-directed mutagenesisaccording to the following rationale. We have evi-dence that polyamines interact with the proteininside the pore (Iyer & Delcour, 1997; Liu et al.,1997; Samartzidou & Delcour, 1999b): (1) polya-mines are effective from either side of the mem-brane; (2) inhibition is voltage-dependent; (3) thereis a correlation between ion selectivity and sensi-tivity to modulators; and (4) modulation is abol-ished in OmpC when a pore-exposed residue onthe L3 loop (D105) is mutated to a glutamine resi-due. Therefore, we postulate that one site of inter-action is likely to be D113 on the L3 loop (which ishomologous to D105 in OmpC), and we engin-eered a D113A mutant in OmpF. Other residues tobe mutated were chosen from inspection of theknown crystal structure (Cowan et al., 1992) andthe following criteria: (1) the residues are directlyexposed to the pore; (2) the residues are 9 AÊ or 4 AÊ

or 13 AÊ away from D113; (3) the candidate resi-dues are either negatively charged or capable ofhydrogen-bonding. We have found only two otherresidues that ful®lled these criteria: D121 is located9 AÊ away from D113, and Y294 is located 4 AÊ

away from D121. The positions of all three candi-date residues are shown in Figure 5.

We converted each of these amino acids to ala-nine by site-directed mutagenesis. The traces ofFigure 6 show that wild-type channels are stronglyinhibited by 100 mM spermine (average fold-increase in number of closures: 13.8(�4.9), in fourexperiments), but porins from the D113A andD121A mutants are insensitive to 1 mM spermine,an even greater concentration. A mild spermineeffect is observed on the Y294A channels (averagefold-increase in number of closures: 2.3(�1.6), in

Figure 6. Representative current traces obtained from wild-type (WT) and mutant channels, at the indicated sper-mine (SPM) and spermidine (SPD) concentrations. Upward de¯ections are closures from the baseline level highlightedby a right tick mark. The pipette voltage is ÿ60 mV.


three experiments). Spermidine (3 mM) produces astrong effect on wild-type channels (average fold-increase in number of closures: 4.9(�2.0), in threeexperiments), but is ineffective in the D113A andD121A mutants The average fold-increase innumber of closures from three separate exper-iments was 1.3(�0.3) and 1.2(�0.3), for D121A andD113A, respectively. Modulation of Y294Achannels by 3 mM spermidine remains clearlyobserved, as shown in the traces of Figure 6 (aver-age fold-increase in number of closures: 2.5(�1.0),in four experiments).

Discussion

We previously found that natural polyaminesinhibit porin function by stabilizing closed, non-ionconducting states. The voltage-dependence of inhi-bition (delaVega & Delcour, 1995; Iyer & Delcour,

1997), the apparent correlation between ion selec-tivity and sensitivity to modulators (Samartzidou& Delcour, 1999b), and the study of porin mutantsthat are insensitive to polyamines (Liu et al., 1997)lead us to propose that the site(s) of interactionbetween polyamines and porins reside(s) withinthe transmembrane pore. In order to understandbetter the nature of the binding site and the mol-ecular mechanism of action, we studied the effectof a variety of amine bearing compounds on thegating kinetics of OmpF.

We found the following set of requirements formolecules to behave as inhibitors of OmpF porinin the submillimolar range: (1) at least a chargeof � 2; (2) the need for primary or secondaryamines, but not quaternary amines; (3) an opti-mum molecular length between two chargedamines of at least the equivalent of a C-8 to C-10hydrocarbon chain. These properties suggest that


the modulator might bridge two charged or polarloci within the pore separated by a distance cor-responding to the length of the C-8 and C-10diamines (9-11 AÊ , see Weiger et al., 1998). Anincreased ef®ciency is seen with spermine and thesynthetic polyamines of the C-12 series, possiblybecause these compounds have an extra hydro-carbon chain with terminal amine tagged to thebasic 1,8-diaminooctane motif. This additionalmethylene chain may interact with yet anothercharged locus in the same binding pocket. Themolecular modeling of Weiger et al. (1998) revealsthat the spermidine and the C-2 to C-10 diaminesare relatively rigid, while spermine and 1,12-diami-nododecane are more ¯exible, and can assumebend-over con®gurations. However, by themselves,¯exibility and terminal amine substitution on the12th carbon atom are not suf®cient to make apotent compound, because 1,12-diaminododecaneis completely inert. In their fully extended confor-mation, spermine and 1,12-diaminododecane havesimilar length and ¯exibility (Weiger et al., 1998).But the long hydrocarbon intersegment of 1,12-dia-minododecane confers a much higher hydrophobi-city than in spermine (and even in 1,10-diaminodecane) (Weiger et al., 1998). Although thehydrophobic character of the molecule appears tobe required for modulation of some types of chan-nels (Weiger et al., 1998), it is likely that here itplays a major role in preventing 1,12-diaminodode-cane from reaching the polyamine binding sitewithin the pore, since porins are water-®lled chan-nels that typically exclude hydrophobic solutes(Nikaido, 1996). Any decrease in the extent ofhydrophobicity, such as a smaller hydrocarbonspacer (as in the case of 1,10-diaminodecane) or thepresence of internal amines (as in the case of 1,12-D4D or spermine), would allow a better access ofthe polyamine to the channel interior, and hencewould confer ef®ciency.

Charge or length itself is not unique in determin-ing the ef®ciency of the modulator, but severalother features may be needed for interaction. Theinert phenotype of the quaternary compoundscould potentially arise from: (i) a loss of hydrogen-bonding ability at the terminal nitrogens (due toreplacement of the hydrogens); (ii) lesser ¯exibilityof the molecule (Weiger et al., 1998); or (iii) stericeffects. Similarly, the bulkiness contributed by theguanidino group of agmatine might also accountfor the deviation of this compound from a cadaver-ine-like ef®ciency, despite the potential additionalability to hydrogen bond with candidate residueswithin the pore. Importantly, the lack of effect bysome charged and/or bulky compounds stronglysuggests that that the basis for modulation is notsimply the physical occlusion of the pore or theaccumulation of charges into the pore, as shownfor polyanions acting on VDAC (Mangan &Colombini, 1987), but that speci®c interactionsbetween polyamine molecules and proteinic resi-dues are likely to occur.

Bridging between negatively charged acidicamino acids and a polyamine molecule has beenpresented in a model of the spermine binding siteof the protein kinase CK2 (Leroy et al., 1997a) anddemonstrated in crystals of the E. coli protein PotDbound to spermidine (Sugiyama et al., 1996a). Byanalogy with these two systems and on the basisof different inhibition ef®ciencies observed for thepolyamines reported here, we minimally proposethree sites on OmpF separated by distances of 9 AÊ

and 4 AÊ , and which would interact with aminegroups of spermine. Inspection of the OmpF struc-ture and previous studies (Liu et al., 1997) revealedthree candidate residues. D113 and D121 arelocated on the periplasmic and extracellular sidesof the L3 loop, respectively. Y294 belongs to thebarrel wall adjacent to the L3 loop, and is slightlymore extracellular than D121. All residues areexposed to the pore lumen, as shown in the trans-verse section of Figure 5(b).

Alanine substitution at each residue has greatlyimpaired inhibition by spermine. These resultssuggest that through its interaction with these resi-dues, spermine may act as a linker between thetwo branches of the L3 loop and the barrel wall.The spermine molecule would enter the narrowestpart of the pore in an end-on conformation, whichhas been suggested for other channels (Lopatinet al., 1995), and, being perpendicular to the planeof the membrane, would make contact with thepore-exposed residues. The ¯exibility of spermine(Weiger et al., 1998) could allow it to adopt a some-what bend-over conformation necessary to bridgethe three residues that are not perfectly aligned.The L3 loop residues D113 and D121 also play animportant role in the inhibition by spermidine,since their conversion to alanine has suppressedinhibition by this compound. The 9 AÊ distancebetween D113 and D121 ®ts with the distancebetween the terminal amines of spermidine, andthe results suggest that these two residues providethe main points for anchoring a spermidine mol-ecule. Consequently, it is likely that Y294 plays alesser role in spermidine binding, and this is corro-borated by a level of inhibition of the Y294Amutant that is very close to the wild-type level.

The fact that D113 and D121 are main players inspermine/spermidine inhibition is substantiated byour previous ®ndings that mutations of other nega-tively charged residues of the L3 loop altered onlyslightly or not at all the sensitivity of the channelto polyamines (Liu et al., 1997; Liu, 1999). The abil-ity of a polyamine molecule to bridge optimallybetween these residues might provide an expla-nation for the observed differences in polyamineef®ciencies. The short cadaverine is not longenough to effectively bridge the sites separated by9 AÊ , and this may explain the lower ef®ciency ofcadaverine compared to diaminooctane. Putrescineis even shorter (Weiger et al., 1998), and unable tomodulate closing kinetics (Iyer & Delcour, 1997).Diaminooctane and spermidine have similar lengthand ef®ciencies. Polyamines of the C-12 series


would have a higher ef®ciency because the 4 AÊ

spacer provides an additional point of attachmentat the terminal amine. The 1,12-D3D/1,12-D5Dpair is slightly less effective than the 1,12-D4D/spermine pair, possibly because 1,12-D3D and1,12-D5D have internal amines that do not alignwith those of spermine or 1,12-D4D. However,these two compounds are still much better porinmodulators than spermidine, possibly because thepresence of the 4 AÊ -tag compensates for the slightmismatch at the internal site.

It is noteworthy that spermidine and spermineare distinct from 1,8-diaminooctane and the syn-thetic polyamines of the C-12 series because theyboth have an interruption of the 9 AÊ -spacer by anamine. This interruption of the octamethylenespacer may favor the appearance of prolongedclosed states of many cooperative channels, since itexists in two compounds that promote such longclosures (spermidine and spermine), but is missingin compounds that do not. It is conceivable thatthe presence of this internal amine group leads to adifferent type of interaction with the protein,resulting in different sets and/or lifetimes of chan-nel conformations. But the molecular basis forthese interactions is unclear at this point. It appearsthat 1,8-diaminooctane and the synthetic polya-mines of the C-12 series behave as simple open-channel blockers, since a ¯ickering channel activityis often the indication of oscillations between openand blocked states (Hille, 1994). By promoting longclosures as well as ¯ickering activity, spermine andspermidine appear to alter the kinetic rate con-stants that govern intrinsic spontaneous oscil-lations between closed and open states, and thusact as modulators of the gating kinetics and aschannel blockers.

If the binding site is indeed within the pore, howcan polyamines promote cooperative closures?Even in the absence of modulators, porins displaya highly cooperative behavior (Schindler &Rosenbusch, 1981; Delcour, 1997; Delcour et al.,1989b; Berrier et al., 1992; Samartzidou & Delcour,1998). The molecular basis for this intriguingphenomenon is unknown. Surprisingly, there isevidence that pore mutations can affect cooperativ-ity (Delcour et al., 1991; Liu & Delcour, 1998; Liu,1999). The L3 loop in particular has been postu-lated to play a role in maintaining the structuralintegrity of the barrel (Fourel et al., 1994; Schmidet al., 1998). Although there is evidence that move-ment of the L3 loop is not involved in the voltage-driven inactivation process (Phale et al., 1997;Eppens et al., 1997; Bainbridge et al., 1998), somemotion of the loop during spontaneous gating hasbeen proposed on the basis of mutant studies (Liu& Delcour, 1998). Thus, any modi®cation of theloop's position or ¯exibility during gating, orthrough mutations or binding of modulators,might affect interactions between barrels, and thuscooperative behavior. Our hypothesis is that thenatural polyamines stabilize closed states by tem-pering with the gating mechanism, thus promoting

closures that are themselves intrinsically coopera-tive in nature, and possibly enhancing cooperativ-ity as well through long-range interactions withthe barrel.

At this point, we cannot dismiss the possibilitythat multiple polyamine molecules interact withthe channel at the same time. This possibility issupported by our ®nding of mutations that elimin-ate the polyamine-dependent increase in closingactivity, but do not affect suppression of openingsor macroscopic current reduction (Liu, 1999; Liuet al., 1997). Answers to this question and thoserelating to the nature of the binding site will beobtained either from crystallography of porin-poly-amine complexes or from a combination ofapproaches, such as binding studies, site-directedmutagenesis, and molecular modeling studies. Theknowledge of the architecture of the polyaminebinding site might ultimately have implications inthe design of an optimal drug against porins andrelated proteins in other pathogenic Gram-negativebacteria.

Materials and Methods

Chemicals

Putrescine, cadaverine, spermidine, spermine, agma-tine and 1,6-diaminohexane were purchased from Sigmaor Aldrich Chemical Co. as the hydrochloride forms. 1,8-Diaminooctane, 1,10-diaminodecane and 1,12-diamino-dodecane were purchased from Sigma as the amineforms. Concentrations higher than 1 mM of 1,10-diami-nodecane and 1,12-diaminododecane were dissolved byaddition of 300 to 400 ml of 1 M HCl. The quaternaryammonium compounds were obtained either as chlorideor bromide salts from Research Biochemicals Inter-national. In all cases, the pH of the ®nal solution wasadjusted to 7.2. All reagents used for the synthesis of1,12-D3D, 1,12-D4D, and 1,12-D5D were purchased fromAldrich Chemical Co. and were used without furtherpuri®cation. Reaction solvents were purchased in anhy-drous form, or dried using established procedures forsolvent puri®cation.

Synthesis of 1,12-diamino-3-azadodecane(1,12-D3D), 1,12-diamino-4-azadodecane (1,12-D4D)and 1,12-diamino-5-azadodecane (1,12-D5D)

The synthesis of 1,12-diamino-4-azadodecane (Table 1)was accomplished using previously described syntheticprocedures (Saab et al., 1993; Bellevue et al., 1996). A por-tion of 1,8-diaminooctane was treated with 2.2 equiva-lents of mesitylenesulfonyl chloride partitioned betweendichloromethane and 10 % (v/v) aqueous NaOH (Yajimaet al., 1978) to afford the corresponding bis-mesitylatedamine. Treatment of this protected diamine with 1.1equivalents of 3-bromopropylphthalimide in the pre-sence of sodium hydride then afforded the correspond-ing protected triamine. Sequential removal of thephthalimide and mesitylene protecting groups usingmethanolic hydrazine and then 30 % HBr in acetic acid(Yajima et al., 1978), respectively, afforded the requisiteproduct, 1,12-diamino-4-azadodecane, as the trihydro-bromide salt. The analog was found to be >99 % pure, asestimated by proton NMR spectroscopy: 1H NMR (2H2)


ppm d 3.00 (m, 8H, H1, H3, H5, H12), 2.00 (m, 2H, H2),1.59 (m, 4H, H6 and H11), 1.29 (m, 8H, H7, H8, H9, H10).

The synthesis of 1,12-diamino-5-azadodecane (1,12-D5D) was accomplished using an analogous pathway.bis-Mesitylation of diaminoheptane afforded the corre-sponding bis-mesitylated amine, which was alkylated asdescribed with bromobutylphthalimide to produce theprotected intermediate triamine. Deprotection thenresulted in the formation of the desired product as thetrihydrobromide salt. The analog was found to be >99 %pure, as estimated by proton NMR spectroscopy: 1HNMR (2H2O)ppm d 3.0 (m, 8H, H1, H4, H6,H12), 1.83 (m,4H, H2 and H3), 1.74 (m, 4H, H7 and H11), 1.29 (m, 6H,H8, H9, H10).

The synthesis of 1,12-diamino-3-azadodecane (1,12-D3D) was accomplished according to a similar scheme.bis-Mesitylation of diaminononane afforded the corre-sponding bis-mesitylated amine, which was alkylated asdescribed with bromoethylphthalimide to produce theprotected intermediate triamine. Deprotection thenresulted in the formation of the desired product as thetrihydrobromide salt. The analog was found to be >99 %pure, as estimated by proton NMR spectroscopy: 1HNMR (2H2O)ppm d 3.5 (m, 4H, H1 andH2), 3.2 (t, 2H,H4), 3.06 (t, 2H, H12), 1.75 (m, 4H, H5 and H11), 1.41 (m,10H, H6, H7, H8, H9, H10).

Bacterial strains and site-directed mutagenesis

For experiments on wild-type cells, E.coli K12 strainAW738 expressing OmpF only (Ingham et al., 1990) wasused. A kanamycin-resistant plasmid containing a clonedOmpF gene (pNLF10) was used as the template to intro-duce site-directed mutations into OmpF with the uniquesite elimination method (U.S.E., Pharmacia) as pre-viously described (Liu et al., 1997). All mutations(D113A, D121A, Y294A) were con®rmed by DNAsequencing of the whole gene. The mutated plasmid wasintroduced into HS111, an E. coli K12 strain deleted forOmpC and OmpF from the chromosome, as previouslydescribed by Samartzidou & Delcour (1999a). Enzymesused in molecular biology protocols were purchasedfrom either Gibco or Promega. The DNA sequencing kitwas either from United States Biotechnology or PerkinElmer, and the DNA puri®cation kit was from Promega.

Membrane preparation and electrophysiology

Cells expressing wild-type or mutated OmpF weregrown in T broth (1 % (w/v) Tryptone (Difco Labora-tories) and 0.5 % (w/v) NaCl) to mid-log phase, with0.7 mM IPTG if required. After harvesting the cells,outer membrane fractions were puri®ed by sucrose gra-dient, as described by Delcour et al. (1989a). Patch clampexperiments were performed on blisters induced fromgiant liposomes containing the reconstituted outer mem-brane fractions (Delcour et al., 1989b). Protein:lipid ratiosof 1600-1750 typically yielded seals of �0.5-1.0 G, dueto the presence of multiple open porin channels in thepatch. Patches were excised by air exposure, resulting inan inside-out orientation with the extracellular side ofthe channel exposed to the pipette. Control experimentswere done in solutions of 150 mM KCl, 5 mM Hepes,0.1 mM K-EDTA, 0.01 mM CaCl2 (pH 7.2), in both thepipette and the patch clamp chamber (bath). Solutions of10-15 ml polyamine in the same buffer were thenapplied to the periplasmic side of the patch with bathperfusion. All solutions were ®ltered through a 0.2 mm

®lter. Currents were ®ltered at 2 kHz (FrequencyDevices) and recorded with an Axopatch-1D ampli®er(Axon Instruments). Continuous recordings were madeon VCR tapes.

Identification of OmpF channels and data analysis

Although we conduct our experiments on outer mem-brane fractions and not on puri®ed proteins, we haveascertained that the recording activities originate fromOmpF because (1) we are using a strain that expressesOmpF only as the major porin; (2) the activity is absentin a mutant lacking OmpF and OmpC and no otherchannel activities are observed (Delcour et al., 1992). Wepurposefully did not purify OmpF because puri®cationand reconstitution procedures can affect porin behavior(Saxena et al., 1989; Lakey & Pattus, 1989; Buehler et al.,1991).

Closing transitions of OmpF channels are too infre-quent and transient to yield sizeable peaks in amplitudehistograms. Therefore, the conductance of a single mono-mer is deduced from the amplitude of individual events.The amplitudes of the closing transitions typically clusteraround values that are integer multiples of the smallestobserved value. Because there is no favored conductancelevel, we have made the working hypothesis that thetransitions of the smallest amplitude represent a singlemonomer (rather than a sub-conductance state), and thelargest transitions are those of monomers gating coop-eratively. Although the single-channel conductance wecalculated is smaller than that reported in planar lipidbilayer studies, an accurate comparison with otherarticles is dif®cult because of the great difference in timeresolution between patch-clamp and planar lipid bilayertechniques.

For analysis, the data were re-®ltered at 1 kHz anddigitized at 100 ms sampling intervals (Instrutech). Dataacquisition and analysis were done with Axobasic pro-grams developed in the laboratory. For kinetic analysisof the closures, the algorithm uses the half-amplitudecriterion (Liu et al., 1997) to classify events lasting formore than 300 ms as closures of 1, 2, . . . , N channels. Thenumbers and average durations of such events are com-puted, as well as the average time spent at the baselinelevel, htBLi.

Acknowledgments

We thank Nazhen Liu for cloning the wild-type OmpFgene into the pNLF10 plasmid, and Linda Guynn fortechnical assistance. This work was supported by NIHgrants AI34905 (A.H.D.) and CA63552 (P.M.W.), and agrant-in-aid of research from the Sigma Xi Foundation(to R.I.).

References

Bainbridge, G., Mobasheri, H., Armtsrong, G. A., Lea,E. J. A. & Lakey, J. H. (1998). Voltage-gating ofEscherichia coli porin: a cystine-scanning mutagen-esis study of loop 3. J. Mol. Biol. 275, 171-176.

Bellevue, F. H., Boahbedason, M. L., Wu, R. H., Casero,R. A., Jr, Rattendi, D., Lane, S., Bacchi, C. J. &Woster, P. M. (1996). Structural comparison ofalkylpolyamine analogues with potent in vitro anti-


tumor or antiparasitic activity. Bioorg. Med. Chem.Letters, 6, 2765-2770.

Benz, R. (1988). Structure and function of porins fromGram-negative bacteria. Annu. Rev. Microbiol. 42,359-393.

Berrier, C., Coulombe, A., Houssin, C. & Ghazi, A.(1992). Fast and slow kinetics of porin channelsfrom Escherichia coli reconstituted into giant lipo-somes and studied by patch-clamp. FEBS Letters,306, 251-256.

Besnard, M., Martinac, B. & Ghazi, A. (1997). Voltage-dependent porin-like ion channels in the ArcheonHaloferax volcanii. J. Biol. Chem. 272, 992-995.

Buch, J. K. & Boyle, S. M. (1985). Biosynthetic argininedecarboxylase in Escherichia coli is synthesized as aprecursor and located in the cell envelope.J. Bacteriol. 163, 522-527.

Buechner, M., Delcour, A. H., Martinac, B., Adler, J. &Kung, C. (1990). Ion channel activities in the Escheri-chia coli outer membrane. Biochim. Bioph. Acta, 1024,111-121.

Buehler, L. K., Kusumoto, S., Zhang, H. & Rosenbusch,J. P. (1991). Plasticity of Escherichia coli porin chan-nels. J. Biol. Chem. 266, 24446-24450.

Canellakis, E. S., Kyriakidis, D. A., Huang, S. C. &Panagiotidis, C. A. (1990). In The Biology and Chem-istry of Polyamines (Goldemberg, S. H. & Algranati,I. D., eds), pp. 105-114, Oxford University Press,USA, NY.

Cowan, S. W., Schirmer, T., Rummel, G., Steiert, M.,Ghosh, R., Pauptit, R. A., Jansonius, J. N. &Rosenbusch, J. P. (1992). Crystal structures explainfunctional properties of two E. coli porins. Nature,358, 727-733.

delaVega, A. L. & Delcour, A. H. (1995). Cadaverineinduces closing of E. coli porins. EMBO J. 14, 6058-6065.

Delcour, A. H. (1997). Function and modulation of bac-terial porins: insights from electrophysiology. FEMSMicrobiol. Letters, 151, 115-123.

Delcour, A. H., Martinac, B., Kung, C. & Adler, J.(1989a). A modi®ed reconstitution method used inpatch-clamp studies of Escherichia coli ion channels.Biophys. J. 56, 631-636.

Delcour, A. H., Martinac, B., Kung, C. & Adler, J.(1989b). Voltage-sensitive ion channel of Escherichiacoli. J. Memb. Biol. 112, 267-275.

Delcour, A. H., Adler, J. & Kung, C. (1991). A singleamino acid substitution alters conductance and gat-ing of OmpC porin of Escherichia coli. J. Memb. Biol.119, 267-275.

Delcour, A. H., Kung, C., Adler, J. & Martinac, B. (1992).Membrane-derived oligosaccharides (MDO's) pro-mote closing of an E. coli porin channel. FEBS Letters,304, 216-220.

Eppens, E. F., Saint, N., Van Gelder, P., van Boxtel, R. &Tommassen, J. (1997). Role of the constriction loopin the gating of outer membrane porin PhoE ofEscherichia coli. FEBS Letters, 415, 317-320.

Ficker, E., Taglialatela, M., Wible, B. A., Henley, C. M.& Brown, A. M. (1994). Spermine and spermidineas gating molecules for inward recti®er K� chan-nels. Science, 266, 1068-1072.

Fourel, D., Bernadac, A. & PageÁs, J.-M. (1994). Involve-ment of exposed polypeptide loops in trimeric stab-ility and membrane insertion of Escherichia coliOmpF porin. Eur. J. Biochem. 222, 625-630.

Hille, B. (1994). Ionic Channels of Excitable Membranes,2nd edit., Sinauer, Sunderland, USA, MA.

Ingham, C., Buechner, M. & Adler, J. (1990). Effect ofouter membrane permeability on chemotaxis inEscherichia coli. J. Bacteriol. 172, 3577-3783.

Iyer, R. & Delcour, A. H. (1997). Complex inhibition ofOmpF and OmpC bacterial porins by polyamines.J. Biol. Chem. 272, 18595-18601.

Jayaraman, V., Usherwood, P. N. & Hess, G. P. (1999).Inhibition of nicotinic acetylcholine receptor by phi-lanthotoxin-343: kinetic investigations in the micro-second time region using a laser-pulse photolysistechnique. Biochemistry, 38, 11406-11414.

Johnson, T. D. (1996). Modulation of channel functionby polyamines. Trends Pharmacol. Sci. 17, 22-27.

Kashiwagi, K., Fukuchi, J., Chao, J., Igarashi, K. &Williams, K. (1996a). The aspartate residue in theextracellular loop of the N-methyl-D-aspartate recep-tor controls sensitivity to spermine and protons.Mol. Pharmacol. 49, 1131-1141.

Kashiwagi, K., Pistocchi, R., Shibuya, S., Sugiyama, S.,Morikawa, K. & Igarashi, K. (1996b). Spermidine-preferential uptake system in Escherichia coli. Identi-®cation of amino acids involved in polyamine bind-ing in PotD protein. J. Biol. Chem. 271, 12205-12208.

Koski, P. & Vaara, M. (1991). Polyamines as constituentsof the outer membranes of Escherichia coli andSalmonella typhimurium. J. Bacteriol. 173, 3695-3699.

Lakey, J. H. & Pattus, F. (1989). The voltage-dependentactivity of Escherichia coli porins in different planarbilayer reconstitutions. Eur. J. Biochem. 186, 303-308.

Lee, J.-K., John, S. A. & Weiss, J. N. (1999). Novel gatingmechanism of polyamine block in the stronginward recti®er K channel Kir2.1. J. Gen. Physiol.113, 555-563.

Leroy, D., Filhol, O., Delcros, J. G., Pares, S., Chambaz,E. M. & Cochet, C. (1997a). Chemical features of theprotein kinase CK2 polyamine binding site. Bio-chemistry, 36, 1242-1250.

Leroy, D., Heriche, J., Filhol, O., Chambaz, E. M. &Cochet, C. (1997b). Binding of polyamines to anautonomous domain of the regulatory subunit ofprotein kinase CK2 induces a conformationalchange in the holoenzyme. A proposed role for thekinase stimulation. J. Biol. Chem. 272, 20820-20827.

Liu, N. (1999). Structure-function relationships of E. ColiOmpC Porin. The effects of site directed mutationson porin channel function. PhD. thesis, Universityof Houston.

Liu, N., Benedik, M. J. & Delcour, A. H. (1997). Disrup-tion of polyamine modulation by a single aminoacid substitution on the L3 loop of the OmpC porinchannel. Biochim. Biophys. Acta, 1326, 201-212.

Liu, N. & Delcour, A. H. (1998). The spontaneous gatingactivity of OmpC porin is affected by mutations ofa putative hydrogen bond network or of a salt-bridge between the L3 Loop and the barrel. ProteinEng. 11, 797-802.

Lopatin, A., Makhina, E. N. & Nichols, C. G. (1994). Pot-assium channel blocking by cytoplasmic polyaminesas the mechanism of intrinsic recti®cation. Nature,372, 366-369.

Lopatin, A., Makhina, E. N. & Nichols, C. G. (1995). Themechanism of inward recti®cation of potassiumchannels: ``long-pore plugging'' by cytoplasmicpolyamines. J. Gen. Physiol. 106, 923-955.

Mangan, P. S. & Colombini, M. (1987). Ultrasteep vol-tage dependence in a membrane channel. Proc. NatlAcad. Sci. USA, 84, 4896-4900.

Nikaido, H. (1996). Outer membrane. In Escherichia coliand Salmonella. Cellular and Molecular Biology


(Neidhardt, F. C., ed.), pp. 29-47, ASM Press,Washington, DC.

Olson, E. (1993). In¯uence of pH on bacterial geneexpression. Mol. Microbiol. 8, 5-14.

Phale, P. S., Schirmer, T., Prilipov, A., Lou, K.-L.,Hardmeyer, A. & Rosenbusch, J. P. (1997). Voltagegating of Escherichia coli porin channels: role of theconstriction loop. Proc. Natl Acad. Sci. USA, 94,6741-6745.

Saab, N. H., West, E. E., Bieszk, N. C., Preuss, C. V.,Mank, A. R., Casero, R. A. & Woster, P. M. (1993).Synthesis and evaluation of unsymmetrically susbti-tuted polyamine analogues as inhibitors of spermi-dine/spermine-N1-acetyltransferase (SSAT) and aspotential antitumor agents. J. Med. Chem. 36, 2998-3004.

Samartzidou, H. & Delcour, A. H. (1998). E. coli PhoEporin has an opposite voltage-dependence from thehomologous OmpF. EMBO J. 17, 93-100.

Samartzidou, H. & Delcour, A. H. (1999a). Excretion ofendogenous cadaverine leads to a decrease inporin-mediated outer membrane permeability.J. Bacteriol. 181, 791-798.

Samartzidou, H. & Delcour, A. H. (1999b). Distinct sen-sitivities of OmpF and PhoE to charged modulators.FEBS Letters, 444, 65-70.

Saxena, R. K., Ishii, J. & Nakae, T. (1989). Modi®cationof the porin function by the membrane componentsused for the reconstitution of model membranes.Curr. Microbiol. 19, 189-191.

Schindler, H. & Rosenbusch, J. P. (1981). Matrix proteinin planar membranes: Clusters of channels in nativeenvironment and their functional reassembly. Proc.Natl Acad. Sci. USA, 78, 2302-2306.

Schmid, B., Maveyraud, L., KroÈmer, M. & Schulz, G. M.(1998). Porin mutants with new channel properties.Protein Sci. 7, 1603-1611.

Scott, R. H., Sutton, K. G. & Dolphin, A. C. (1993). Inter-actions of polyamines with neuronal ion channels.Trends Neurosci. 16, 153-160.

Sugiyama, S., Vassylyev, D. G., Matsushima, M.,Kashiwagi, K., Igarashi, K. & Marikawa, K. (1996a).Crystal Structure of PotD, the primary receptor ofthe polyamine transport system in Escherichia coli.J. Biol. Chem. 271, 9519-9625.

Sugiyama, S., Matsuo, Y., Maenaka, K., Vassylyev, D. G.,Matshushima, M., Kashiwagi, K., Igarashi, K. &Morikawa, K. (1996b). The 1.8-AÊ X-ray structure ofthe Escherichia coli PotD protein complexed withspermidine and the mechanism of polyamine bind-ing. Protein Sci. 5, 1984-1990.

Tabor, C. W. & Tabor, H. (1984). Polyamines in micro-organisms. Annu. Rev. Biochem. 53, 749-790.

Uehara, A., Fill, M., VeÂlez, P., Yasukochi, M. &Imanaga, I. (1996). Recti®cation of rabbit cardiacryanodine receptor current by endogenous poly-amines. Biophys. J. 71, 769-777.

Weiger, T. M., Langer, T. & Hermann, A. (1998). Exter-nal action of di- and polyamines on maxi calcium-activated potassium channels: an electrophysiologi-cal and molecular modeling study. Biophys. J. 74,722-730.

Williams, K., Kashiwagi, K., Fukuchi, J. & Igarashi, K.(1995). An acidic amio acid in the N-methyl-D-aspartate receptor that is important for sperminestimulation. Mol. Pharmacol. 48, 1087-1098.

Yajima, H., Takeyama, M., Kanaki, J., Nishimura, O. &Fujino, M. (1978). Studies on peptides. LXXX. NG-mesitylene-2-sulfonylarginine. Chem. Pharm. Bull. 26,3752-3757.

Edited by G. von Heijne

(Received 1 December 1999; received in revised form 8 February 2000; accepted 9 February 2000)

Molecular basis for the polyamine-OmpF porin interactions: inhibitor and mutant studies

Documents

Transcript of Molecular basis for the polyamine-OmpF porin interactions: inhibitor and mutant studies