The antimicrobial peptide gramicidin S permeabilizes phospholipid bilayer membranes without forming discrete ion channels.

We examined the permeabilization of lipid bilayers by the beta-sheet, cyclic antimicrobial decapeptide gramicidin S (GS) in phospholipid bilayers formed either by mixtures of zwitterionic diphytanoylphosphatidylcholine and anionic diphytanoylphosphatidylglycerol or by single zwitterionic unsaturated phosphatidylcholines having various hydrocarbon chain lengths, with and without cholesterol. In the zwitterionic bilayers formed by the phosphatidylcholines, without or with cholesterol, the peptide concentrations and membrane potentials required to initiate membrane permeabilization vary little as function of bilayer thickness and cholesterol content. In all the systems tested, the GS-induced transient ion conductance events exhibit a broad range of conductances, which are little affected by the bilayer composition or thickness. In the zwitterionic phosphatidylcholine bilayers, the effect of GS does not depend on the polarity of the transmembrane potential; however, in bilayers formed from mixtures of phosphatidylcholines and anionic phospholipids, the polarity of the transmembrane potential becomes important, with the GS-induced conductance events being much more frequent when the GS-containing solution is positive relative to the GS-free solution. Overall, these results suggest that GS does not form discrete, well-defined, channel-like structures in phospholipid bilayers, but rather induces a wide variety of transient, differently sized defects which serve to compromise the bilayer barrier properties for small electrolytes.

Kinetics of gramicidin channel formation in lipid bilayers: transmembrane monomer association.

Conducting gramicidin channels form predominantly by the transmembrane association of monomers, one from each side of a lipid bilayer. In single-channel experiments in planar bilayers the two gramicidin analogs, [Val1]gramicidin A (gA) and [4,4,4-F3-Val1]gramicidin A (F3gA), form dimeric channels that are structurally equivalent and have characteristically different conductances. When these gramicidins were added asymmetrically, one to each side of a preformed bilayer, the predominant channel type was the hybrid channel, formed between two chemically dissimilar monomers. These channels formed by the association of monomers residing in each half of the membrane. These results also indicate that the hydrophobic gramicidins are surprisingly membrane impermeant, a conclusion that was confirmed in experiments in which gA was added asymmetrically and symmetrically to preformed bilayers.

Platelet-activating factor is a general membrane perturbant.

Platelet-activating factor (PAF) is, at physiological (nanomolar) concentrations, a potent mediator of inflammation and coagulation. At pharmacological (micromolar) concentrations, PAF induces a variety of effects in diverse tissues. Here we show that PAF at micromolar concentrations is a membrane perturbant. Micromolar PAF alters the properties of channels formed by gramicidin A, and at concentrations greater than or equal to 4 microM disrupts the barrier properties of the host lipid bilayer. PAF thus can act as a detergent and non-specifically alter the behavior of membranes and membrane proteins. This may provide an explanation for some of the effects of PAF seen at high concentrations in vitro.

Proton block of rat brain sodium channels. Evidence for two proton binding sites and multiple occupancy.

The acid titration function of bilayer-incorporated batrachotoxin (BTX)-modified sodium channels was examined in experiments in which the pH was decreased symmetrically, on both sides of the membrane, or asymmetrically, on only one side. In an attempt to minimize interpretational ambiguities, the experiments were done in 1.0 M NaCl (buffered to the appropriate pH) with channels incorporated into net neutral bilayers. When the pH was decreased symmetrically (from 7.4 to 4.5), the small-signal conductance (g) decreased in accordance with the predictions of a simple (single-site) titration function with a pK of approximately 4.9. As the pH was decreased below 6.5, the single-channel current-voltage (i-V) relation became increasingly rectifying, with the inward current being decreased more than the outward current. When the pH was decreased asymmetrically (with the pH of the other solution being held constant at 7.4), the titration behavior was different for extra- and intracellular acidification. With extracellular acidification, the reduction in g could still be approximated by a simple titration function with a pK of approximately 4.6, and there was a pronounced rectification at pHs < or = 6 (cf. Woodhull, A. M. 1973. Journal of General Physiology. 61:687-708). The voltage dependence of the block could be described by assuming that protons enter the pore and bind to a site with a pK of approximately 4.6 at an apparent electrical distance of approximately 0.1 from the extracellular entrance. With intracellular acidification there was only a slight reduction in g, and the g-pH relation could not be approximated by a simple titration curve, suggesting that protons can bind to several sites. The i-V relations were still rectifying, and the voltage-dependent block could be approximated by assuming that protons enter the pore and bind to a site with a pK of approximately 4.1 at an apparent electrical distance of approximately 0.2 from the intracellular entrance. Based on the difference between the three g-pH relations, we conclude that there are at least two proton binding sites in the pore and that they can be occupied simultaneously.

Lysophospholipids modulate channel function by altering the mechanical properties of lipid bilayers.

Lipid metabolites, free fatty acids and lysophospholipids, modify the function of membrane proteins including ion channels. Such alterations can occur through signal transduction pathways, but may also result from "direct" effects of the metabolite on the protein. To investigate possible mechanisms for such direct effects, we examined the alterations of gramicidin channel function by lysophospholipids (LPLs): lysophosphatidylcholine (LPC), lysophosphatidylethanolamine (LPE), lysophosphatidylserine (LPS), and lysophosphatidylinositol (LPI). The experiments were done on planar bilayers formed by diphytanoylphosphatidylcholine in n-decane a system where receptor-mediated effects can be excluded. At aqueous concentrations below the critical micelle concentration (CMC), LPLs can increase the dimerization constant for membrane-bound gramicidin up to 500-fold (at 2 microM). The relative potency increases as a function of the size of the polar head group, but does not seem to vary as a function of head group charge. The increased dimerization constant results primarily from an increase in the rate constant for channel formation, which can increase more than 100-fold (in the presence of LPC and LPI), whereas the channel dissociation rate constant decreases only about fivefold. The LPL effect cannot be ascribed to an increased membrane fluidity, which would give rise to an increased channel dissociation rate constant. The ability of LPC to decrease the channel dissociation rate constant varies as a function of channel length (which is always less than the membrane's equilibrium thickness): as the channel length is decreased, the potency of LPC is increased. LPC has no effect on membrane thickness or the surface tension of monolayers at the air/electrolyte interface. The bilayer-forming glycerolmonooleate does not decrease the channel dissociation rate constant. These results show that LPLs alter gramicidin channel function by altering the membrane deformation energy, and that the changes in deformation energy can be related to the molecular "shape" of the membrane-modifying compounds. Similar alterations in the mechanical properties of biological membranes may form a general mechanism by which one can alter membrane protein function.

Monazomycin-induced single channels. I. Characterization of the elementary conductance events.

Monazomycin (a positively charged, polyene-like antibiotic) induces voltage-dependent conductance changes in lipid bilayer membranes when added to one of the bathing solutions. These conductance changes have generally been attributed to the existence of channels spanning the membrane. In this article we characterize the behavior of the individual conductance events observed when adding small amounts of monazomycin to one side of a lipid bilayer. We find that there are several apparent channel types with one or sometimes two amplitudes predominating. We find further that these fairly similar amplitudes represent two different states of the same fundamental channel entity, presumed to be the monazomycin channel. The current-voltage characteristics of these channels are weakly hyperbolic functions of applied potential. The average lifetimes are essentially voltage independent (between 50 and 400 mV). The average channel intervals, on the other hand, can be strongly voltage dependent, and we can show that the time-averaged conductance of a membrane is proportional to the average channel frequency.

Monazomycin-induced single channels. II. Origin of the voltage dependence of the macroscopic conductance.

The voltage dependence of the conductance induced induced in thin lipid membranes by monazomycin is shown here to be caused by voltage-dependent variations in the frequency of channel openings. We also experimentally demonstrate certain interesting properties of the channel activity that are predicted by a chemical kinetic model (Muller and Peskin, 1981), which successfully describes the macroscopic conductance. We conclude that two parallel mechanisms--one autocatalytic, the other simple mass action--exist that allow monazomycin to enter (or leave) the membrane so that the monazomycin molecules can be in a position to form channels.

On the importance of atomic fluctuations, protein flexibility, and solvent in ion permeation.

Proteins, including ion channels, often are described in terms of some average structure and pictured as rigid entities immersed in a featureless solvent continuum. This simplified view, which provides for a convenient representation of the protein's overall structure, incurs the risk of deemphasizing important features underlying protein function, such as thermal fluctuations in the atom positions and the discreteness of the solvent molecules. These factors become particularly important in the case of ion movement through narrow pores, where the magnitude of the thermal fluctuations may be comparable to the ion pore atom separations, such that the strength of the ion channel interactions may vary dramatically as a function of the instantaneous configuration of the ion and the surrounding protein and pore water. Descriptions of ion permeation through narrow pores, which employ static protein structures and a macroscopic continuum dielectric solvent, thus face fundamental difficulties. We illustrate this using simple model calculations based on the gramicidin A and KcsA potassium channels, which show that thermal atomic fluctuations lead to energy profiles that vary by tens of kcal/mol. Consequently, within the framework of a rigid pore model, ion-channel energetics is extremely sensitive to the choice of experimental structure and how the space-dependent dielectric constant is assigned. Given these observations, the significance of any description based on a rigid structure appears limited. Creating a conducting channel model from one single structure requires substantial and arbitrary engineering of the model parameters, making it difficult for such approaches to contribute to our understanding of ion permeation at a microscopic level.

Batrachotoxin-modified sodium channels in planar lipid bilayers. Ion permeation and block.

Batrachotoxin-modified, voltage-dependent sodium channels from canine forebrain were incorporated into planar lipid bilayers. Single-channel conductances were studied for [Na+] ranging between 0.02 and 3.5 M. Typically, the single-channel currents exhibited a simple two-state behavior, with transitions between closed and fully open states. Two other conductance states were observed: a subconductance state, usually seen at [NaCl] greater than or equal to 0.5 M, and a flickery state, usually seen at [NaCl] less than or equal to 0.5 M. The flickery state became more frequent as [NaCl] was decreased below 0.5 M. The K+/Na+ permeability ratio was approximately 0.16 in 0.5 and 2.5 M salt, independent of the Na+ mole fraction, which indicates that there are no interactions among permeant ions in the channels. Impermeant and permeant blocking ions (tetraethylammonium, Ca++, Zn++, and K+) have different effects when added to the extracellular and intracellular solutions, which indicates that the channel is asymmetrical and has at least two cation-binding sites. The conductance vs. [Na+] relation saturated at high concentrations, but could not be described by a Langmuir isotherm, as the conductance at low [NaCl] is higher than predicted from the data at [NaCl] greater than or equal to 1.0 M. At low [NaCl] (less than or equal to 0.1 M), increasing the ionic strength by additions of impermeant monovalent and divalent cations reduced the conductance, as if the magnitude of negative electrostatic potentials at the channel entrances were reduced. The conductances were comparable for channels in bilayers that carry a net negative charge and bilayers that carry no net charge. Together, these results lead to the conclusion that negative charges on the channel protein near the channel entrances increase the conductance, while lipid surface charges are less important.

Batrachotoxin-modified sodium channels in planar lipid bilayers. Characterization of saxitoxin- and tetrodotoxin-induced channel closures.

The guanidinium toxin-induced inhibition of the current through voltage-dependent sodium channels was examined for batrachotoxin-modified channels incorporated into planar lipid bilayers that carry no net charge. To ascertain whether a net negative charge exists in the vicinity of the toxin-binding site, we studied the channel closures induced by tetrodotoxin (TTX) and saxitoxin (STX) over a wide range of [Na+]. These toxins carry charges of +1 and +2, respectively. The frequency and duration of the toxin-induced closures are voltage dependent. The voltage dependence was similar for STX and TTX, independent of [Na+], which indicates that the binding site is located superficially at the extracellular surface of the sodium channel. The toxin dissociation constant, KD, and the rate constant for the toxin-induced closures, kc, varied as a function of [Na+]. The Na+ dependence was larger for STX than for TTX. Similarly, the addition of tetraethylammonium (TEA+) or Zn++ increased KD and decreased kc more for STX than for TTX. These differential effects are interpreted to arise from changes in the electrostatic potential near the toxin-binding site. The charges giving rise to this potential must reside on the channel since the bilayers had no net charge. The Na+ dependence of the ratios KDSTX/KDTTX and kcSTX/kcTTX was used to estimate an apparent charge density near the toxin-binding site of about -0.33 e X nm-2. Zn++ causes a voltage-dependent block of the single-channel current, as if Zn++ bound at a site within the permeation path, thereby blocking Na+ movement. There was no measurable interaction between Zn++ at its blocking site and STX or TTX at their binding site, which suggests that the toxin-binding site is separate from the channel entrance. The separation between the toxin-binding site and the Zn++ blocking site was estimated to be at least 1.5 nm. A model for toxin-induced channel closures is proposed, based on conformational changes in the channel subsequent to toxin binding.

Intrinsic characteristics of the proton pump in the luminal membrane of a tight urinary epithelium. The relation between transport rate and delta mu H.

A number of tight urinary epithelia, as exemplified by the turtle bladder, acidify the luminal solution by active transport of H+ across the luminal cell membrane. The rate of active H+ transport (JH) decreases as the electrochemical potential difference for H+ [delta mu H = mu H(lumen) - mu H(serosa)] across the epithelium is increased. The luminal cell membrane has a low permeability for H+ equivalents and a high electrical resistance compared with the basolateral cell membrane. Changes in JH thus reflect changes in active H+ transport across the luminal membrane. To examine the control of JH by delta mu H in the turtle bladder, transepithelial electrical potential differences (delta psi) were imposed at constant acid-base conditions or the luminal pH was varied at delta psi = 0 and constant serosal PCO2 and pH. When the luminal compartment was acidified from pH 7 to 4 or was made electrically positive, JH decreased as a linear function of delta mu H as previously described. When the luminal compartment was made alkaline from pH 7 to 9 or was made electrically negative, JH reached a maximal value, which was the same whether the delta mu H was imposed as a delta pH or a delta psi. The nonlinear JH vs. delta mu H relation does not result from changes in the number of pumps in the luminal membrane or from changes in the intracellular pH, but is a characteristic of the H+ pumps themselves. We propose a general scheme, which, because of its structural features, can account for the nonlinearity of the JH vs. delta mu H relations and, more specifically, for the kinetic equivalence of the effects of the chemical and electrical components of delta mu H. According to this model, the pump complex consists of two components: a catalytic unit at the cytoplasmic side of the luminal membrane, which mediates the ATP-driven H+ translocation, and a transmembrane channel, which mediates the transfer of H+ from the catalytic unit to the luminal solution. These two components may be linked through a buffer compartment for H+ (an antechamber).

Functional carboxyl groups in the red cell anion exchange protein. Modification with an impermeant carbodiimide.

Anion exchange in human red blood cell membranes was inactivated using the impermeant carbodiimide 1-ethyl-3-(4-azonia-4,4-dimethylpentyl)-carbodiimide (EAC). The inactivation time course was biphasic: at 30 mM EAC, approximately 50% of the exchange capacity was inactivated within approximately 15 min; this was followed by a phase in which irreversible exchange inactivation was approximately 100-fold slower. The rate and extent of inactivation was enhanced in the presence of the nucleophile tyrosine ethyl ester (TEE), suggesting that the inactivation is the result of carboxyl group modification. Inactivation (to a maximum of 10% residual exchange activity) was also enhanced by the reversible inhibitor of anion exchange 4,4'-dinitrostilbene-2,2'-disulfonate (DNDS) at concentrations that were 10(3)-10(4) times higher than those necessary for inhibition of anion exchange. The extracellular binding site for stilbenedisulfonates is essentially intact after carbodiimide modification: the irreversible inhibitor of anion exchange 4,4'-diisothiocyanostilbene-2,2'-disulfonate (DIDS) eliminated (most of) the residual exchange activity: DNDS inhibited the residual (DIDS-sensitive) Cl- at concentrations similar to those that inhibit Cl- exchange of unmodified membranes: and Cl- efflux is activated by extracellular Cl-, with half-maximal activation at approximately 3 mM Cl-, which is similar to the value for unmodified membranes. But the residual anion exchange function after maximum inactivation is insensitive to changes of extra- and intracellular pH between pH 5 and 7. The titratable group with a pKa of approximately 5.4, which must be deprotonated for normal function of the native anion exchanger, thus appears to be lost after EAC modification.

Effect of phloretin on the permeability of thin lipid membranes.

Phloretin dramatically increases cation conductances and decreases anion conductances of membranes treated with ion carriers (nonactin, valinomycin, carbonyl-cyanide-m-chlorophenylhydrazone [CCCP], and Hg(C6F5)2) or lipophilic ions (tetraphenylarsonium [tphAs+] and tetraphenylborate [TPhB-]). For example, on phosphatidylethanolamine membranes, 10(-4) M phloretin increases K+ -nonactin and TPhAs+ conductances and decreases CCCP- and TPhB- conductances 10(3)-fold; on lecithin: cholesterol membranes, it increases K+-nonactin conductance 10(5)-fold and decreases CCCP- conductance 10(3)-fold. Similar effects are obtained with p- and m-nitrophenol at 10(-2) M. These effects are produced by the un-ionized form of phloretin and the nitrophenols. We believe that phloretin, which possesses a large dipole moment, adsorbs and orients at the membrane surface to introduce a dipole potential of opposite polarity to the preexisting positive one, thus increasing the partition coefficient of cations into the membrane interior and decreasing the partition coefficient of anions. (Phloretin may also increase the fluidity of cholesterol-containing membranes; this is manifested by its two- to three-fold increase in nonelectrolyte permeability and its asymmetrical effect on cation and anion conductances in cholesterol-containing membranes.) It is possible that pholoretin's inhibition of chloride, urea, and glucose transport in biological membranes results from the effects of these intense intrafacial dipole fields on the translocator(s) of these molecules.

Steady-state gating of batrachotoxin-modified sodium channels. Variability and electrolyte-dependent modulation.

The steady-state gating of individual batrachotoxin-modified sodium channels in neutral phospholipid bilayers exhibits spontaneous, reversible changes in channel activation, such that the midpoint potential (Va) for the gating curves may change, by 30 mV or more, with or without a change in the apparent gating valence (za). Consequently, estimates for Va and, in particular, za from ensemble-averaged gating curves differ from the average values for Va and za from single-channel gating curves. In addition to these spontaneous variations, the average Va shifts systematically as a function of [NaCl] (being -109, -88, and -75 mV at 0.1, 0.5, and 1.0 M NaCl), with no systematic variation in the average za (approximately 3.7). The [NaCl]-dependent shifts in Va were interpreted in terms of screening of fixed charges near the channels' gating machinery. Estimates for the extracellular and intracellular apparent charge densities (sigma e = -0.7 and sigma i = -0.08 e/nm2) were obtained from experiments in symmetrical and asymmetrical NaCl solutions using the Gouy-Chapman theory. In 0.1 M NaCl the extracellular and intracellular surface potentials are estimated to be -94 and -17 mV, respectively. The intrinsic midpoint potential, corrected for the surface potentials, is thus about -30 mV, and the standard free energy of activation is approximately -12 kJ/mol. In symmetrical 0.1 M NaCl, addition of 0.005 M Ba2+ to the extracellular solution produced a 17-mV depolarizing shift in Va and a slight reduction in za. The shift is consistent with predictions using the Gouy-Chapman theory and the above estimate for sigma e. Subsequent addition of 0.005 M Ba2+ to the intracellular solution produced a approximately 5-mV hyperpolarizing shift in the ensemble-averaged gating curve and reduced za by approximately 1. This Ba(2+)-induced shift is threefold larger than predicted, which together with the reduction in za implies that Ba2+ may bind at the intracellular channel surface.

Energetics of gramicidin hybrid channel formation as a test for structural equivalence. Side-chain substitutions in the native sequence.

To determine whether amino acid side-chain substitutions in linear gramicidins after the structure of membrane-spanning channels formed by the modified peptides, we have developed a quantitative measure of structural equivalence of the peptide backbone among gramicidin channels based on functional (single-channel) measurements. The experiments exploit the fact that gramicidin channels are symmetrical dimers, and that channels formed by different gramicidin analogues can be distinguished on the basis of their single-channel current amplitudes or durations. It is thereby possible to determine whether hybrid channels can form between chemically dissimilar peptides, i.e. whether the peptides can adapt to each other. Further, since the relative rates of channel formation as well as the relative concentrations of pure and hybrid channel types can be measured in the same membrane, these experiments provide a quantitative measure of the energetic cost of hybrid channel formation relative to the formation of the pure channels. For a wide variety of different side-chains, we find that substitutions as extreme as glycine to phenylalanine at position 1, at the join between the two monomers in a membrane-spanning dimer, incur no energetic cost for channel formation, which implies that channels formed by each of the modified peptides are structurally equivalent. In addition, the average durations of the hybrid channels (except those having tyrosine or hexafluorovaline at position 1) are intermediate to the average durations of the respective pure channel types, thus providing further evidence for structural equivalence among channels formed by sequence-substituted gramicidins.

Energetics of heterodimer formation among gramicidin analogues with an NH2-terminal addition or deletion. Consequences of missing a residue at the join in the channel.

We examined the properties of membrane-spanning channels formed by gramicidin analogues that differ from [Val1]gramicidin A by having a single residue deletion or insertion at the formyl-NH terminus and of hybrid channels formed between such 14-, 15-, and 16-residue analogues. The channels' backbone structure, and helix sense, are not affected by the sequence modifications, because hybrid channels were observed for all combinations tested, and there was no excess energetic cost associated with hybrid channel formation. When hybrid channels form between analogues of different length the hybrid channel stability depends on the nature of the sequence dissimilarity. If two analogues differ by one residue (delta n = 1), the hybrid channels are destabilized by approximately 10 kJ/mol, because there is a defect (a "gap" in the peptide backbone) at the join between the two beta 6.3-helical monomers such that the dimer is stabilized by only five intermolecular C = O ... H-N hydrogen bonds rather than the usual six. This defect also alters the hybrid channels' permeability characteristics: the single-channel conductances are decreased, as if there were an additional barrier to ion movement through the channel. If the formyl-NH-terminal residue is Gly (and delta n = 1), the hybrid channels show multi-state behavior with voltage-dependent transitions between two conductance levels. If two analogues differ by two residues (delta n = 2), the hybrid channels are stabilized by 3 kJ/mol, indicating that structural continuity at the join between the monomers has been restored, as have the hybrid channels' permeability characteristics. The increased hybrid channel stability (when delta n = 2) may arise from altered membrane-channel interactions.

The conformational preference of gramicidin channels is a function of lipid bilayer thickness.

In order to understand how the material properties of lipid bilayers could affect integral membrane protein function, we examined the effect of a hydrophobic mismatch on the structure and function of membrane-spanning gramicidin channels. Changes in lipid bilayer thickness affect the conformational preference of membrane-spanning gramicidin A (gA) channels (single-stranded [SS] dimers <--> double-stranded [DS] dimers) and induces an additional conductance state in the standard (SS) beta6.3-helical channel. These results provide experimental evidence for the importance of energetic coupling between the bilayer and imbedded inclusions.

Design and characterization of gramicidin channels.

This article summarizes methods for the chemical synthesis and biophysical characterization of gramicidins with varying sequences and labels. The family of gramicidin channels has developed into a powerful model system for understanding fundamental properties, interactions, and dynamics of proteins and lipids generally, and ion channels specifically, in biological membranes.

Design and characterization of gramicidin channels with side chain or backbone mutations.

Mutations and chemical substitutions of amino acid side chains and backbone atoms have proved vital for understanding the folding, structure and function of gramicidin channels in phospholipid membranes. The channel's pore is lined by peptide backbone groups; their importance for channel structure and function is shown by a single amide-to-ester replacement within the backbone, which greatly reduces the resulting channel conductance and lifetime. The four tryptophans and the intervening leucines together govern the formation and dissociation of conducting channels from single-stranded subunits. Conducting double-stranded gramicidin conformations (channels) occur rarely in membranes--except when the sequence has been altered to permit special arrangements of tryptophans or (infrequently) in unusually thick membranes. The tryptophans anchor the single-stranded channels to the membrane/solution interface, and the indole dipoles promote cation transport through the channels. Removal of any indole dipole reduces ion conductance; whereas 5-fluorination of an indole, which increases its dipole moment, enhances ion conductance. Some sequence changes at the formyl-NH-terminus (in the membrane interior, away from the tryptophans), including fluorination of the formyl-NH-terminal valine, introduce voltage-dependent channel gating. Gramicidin channels are not just static conductors, but also dynamic entities whose structure and function can be manipulated by backbone and side chain modifications.

Nonoperative management of contained esophageal perforation.

Spontaneous perforation of the esophagus still carries a high rate of morbidity and mortality because of frequent delay in diagnosis, extensive mediastinal contamination, and inadequate surgical repair. We used a nonoperative approach in two patients in whom the perforation was well contained, with evidence of drainage back into the esophagus and few symptoms or signs of sepsis. Nonoperative management included administration of intravenous antibiotics and hyperalimentation. Both patients had a satisfactory outcome.