JOURNAL OF VIROLOGY, Nov. 1994, P. 7115-7123 0022-538X/94/$04.00+0 Copyright C 1994, American Society for Microbiology
Vol. 68, No. 11
An Amphipathic Peptide from the C-Terminal Region of the Human Immunodeficiency Virus Envelope Glycoprotein Causes Pore Formation in Membranes L. CHERNOMORDIK, A. N. CHANTURIYA, E. SUSS-TOBY, E. NORA,
AND
J. ZIMMERBERG*
Laboratory of Theoretical and Physical Biology, National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, Maryland 20892 Received 20 May 1994/Accepted 25 July 1994
The peptide fragment of the carboxy-terminal region of the human immunodeficiency virus (H1V) transmembrane protein (gp4l) has been implicated in T-cell death. This positively charged, amphipathic helix (amino acids 828 to 848) of the envelope protein is located within virions or cytoplasm. We studied the interaction of the isolated, synthetic amphipathic helix of gp4l with planar phospholipid bilayer membranes and with Sf9 cells using voltage clamp, potentiodynamic, and single-cell recording techniques. We found that the peptide binds strongly to planar membranes, especially to the negatively charged phosphatidylserine bilayer. In the presence of micromolar concentrations of peptide sufficient to make its surface densities comparable with those of envelope glycoprotein molecules in HIV virions, an increase in bilayer conductance and a decrease in bilayer stability were observed, showing pore formation in the planar lipid bilayers. These pores were permeable to both monovalent and divalent cations, as well as to chloride. The exposure of the inner leaflet of cell membranes to even 25 nM peptide increased membrane conductance. We suggest that the carboxy-terminal fragment of the HIV type 1 envelope protein may interact with the cell membrane of infected T cells to create lipidic pores which increase membrane permeability, leading to sodium and calcium flux into cells, osmotic swelling, and T-cell necrosis or apoptosis.
The onset of AIDS is correlated with a decrement of T-cell number and function. While T-cell depletion in vivo may result from a number of different mechanisms, T-cell death in vitro can be caused by acute human immunodeficiency virus (HIV) infection and is initially characterized by a ballooning of cells. There is good evidence that an increase in T-cell membrane permeability, particularly to calcium, is the cause of the osmotic swelling seen in the cytopathic effect (14). Increased intracellular calcium can lead to cell death directly, by necrosis (14), or indirectly through signalling apoptosis (27, 35). If the T-cell depletion seen in AIDS after HIV infection is initiated by a change in membrane permeability, it is important to determine the mechanism of this permeability change. HIV-induced cytopathology is hypothesized to involve the carboxy terminus of the viral transmembrane protein gp4l (residues 707 to 856) which is located exclusively in the intraviral and cytoplasmic compartment (17, 20, 23, 28, 30, 31). Some isolates of HIV type 2 (HIV-2) and simian immunodeficiency virus and mutants of HIV-1 truncated at the carboxyterminal region of the transmembrane protein are able to induce cell-cell fusion but lack significant cytotoxicity (see references 17, 26, and 28; but see also references 18, 40, and 47). A computer simulation of this viral protein fragment, in particular amino acids 828 to 848, reveals a high potential to form strongly amphipathic helices which present positively charged hydrophilic and hydrophobic residues on different sides of the helix (44). The structure of this gp4l fragment is similar to those of some cytolytic peptides whose interactions with membranes have been well studied (7, 13, 15, 16). The hypothesis that single-cell killing of T cells may be caused by the amphipathic helix of the gp4l carboxy terminus
directly interacting with the cell membrane was further substantiated by recent data on the membrane effects of the synthetic peptides of the amino acid sequence corresponding to that of the HIV envelope glycoprotein. A synthetic peptide whose sequence is that of amino acids 828 to 848 of gpl60 (P828) was found to bind strongly to lipid bilayers of negatively charged lipids (21). A peptide homolog of another span, from amino acids 828 to 855, alters the membrane permeability of lipid vesicles and both procaryotic and eukaryotic cells (30, 31, 42). However, it was unclear whether the concentrations of the peptide required to permeabilize membranes correspond to biologically relevant densities of gp4l molecules in the plasma membranes of host cells. In addition, the carboxy terminus of gp4l interacts with the cytoplasmic surfaces of host cells; therefore, it was important to test the membrane effects of the peptide when it was added into cells rather than into external medium (30, 31, 42). Finally, the structure of the permeable pathways (pores) remains unclear. P828 may act like a number of other amphiphilic peptides such as gramicidin, nystatin, and alamethicin which are known to directly form the ionic channels in membrane lipid bilayers (39). Ionic channels have a number of characteristic features related to the fact that the aqueous pathway inside them is surrounded by the peptide amino acids. In general, ionic channels are marked by welldefined selectivity, voltage dependence, and open channel conductance (39). Alternatively, the peptide can promote the local breakdown of the membrane bilayer structure, creating a pore with a surface formed completely or partially by polar lipid headgroups (5, 10, 15). The characteristics of these pores are
not as well defined as those of ionic channels and are more
dependent on membrane composition. Thus, both the character of the permeability changes and the degree of these changes for different cell lines may be quite different for channel-forming peptides and peptides that promote lipidic
* Corresponding author. Mailing address: Bldg. 10, Rm. 10D12, NICHD, Bethesda, MD 20892. Phone: (301) 496-6571. Fax: (301) 594-0813. Electronic mail address:
[email protected].
pore
7115
formation.
7116
J. VIROL.
CHERNOMORDIK ET AL.
To further investigate the role of the cytoplasmic tail of gp4l in HIV cytotoxicity, a number of biophysical and electrophysiological techniques were used. Addition of P828 to planar phospholipid bilayer membranes at surface peptide densities comparable to those of gp4l in viral particles caused an increase in conductance of lipid bilayers. By the patch clamp technique, qualitatively similar effects were found on cell membranes when peptide was perfused into cells to interact with the inner leaflet of the plasma membrane. Our data suggest that P828 promotes lipidic pore development rather than ionic channel formation. MATERIALS AND METHODS Phospholipid membrane formation, current measurements, and chemicals. Two methods of planar phospholipid membrane formation were used in the present study. To study membrane conductance, we used bilayers with relatively small areas formed by the Montal-Mueller technique (32). The side to which P828 was added is defined as the cis compartment. To save time and peptide, a dual bilayer chamber with one cis and two trans compartments was constructed. Membranes were formed across one 200- to 300-,m-diameter hole in each of two Teflon partitions in a Lucite chamber. Holes were pretreated with 10% hexadecane in pentane. A total of 3 to 6 R1 of a 1% (wt/vol) solution of bovine brain phosphatidylserine (PS), diphytanoyl phosphatidylcholine (DPC), or a PS plus DPC mixture in hexane was applied to the surface of the aqueous buffer in the cell, and the solution was allowed to dry for 10 min. Membrane thinning was monitored as capacitance. Ag/AgCl electrodes (In Vivo Metric, Healdsburg, Calif.), which were placed in saturated KCl, were connected to buffer through 1.5% agarose bridges. An electrode placed in the cis compartment was used for setting the potential on the membrane. Two identical electrodes placed in smaller compartments (trans) were connected to separate current/voltage converters. The solution in the cis compartment was stirred for 30 to 60 s after the addition of peptide. To change the ionic composition of the cis compartment or to remove unbound peptide, we perfused the cis compartment with 3 volumes of new solution. To measure membrane lifetimes under different voltages, a large number of membranes must be studied, since lifetime dispersions are great (2). For these measurements, MuellerRudin type planar phospholipid membranes (33) were used since they were easier to form. For these experiments and for measurements of the membrane surface potential, lipid bilayers were formed by applying a small amount of a lipid solution (10 mg/ml in n-decane) onto a 0.6-mm-diameter hole in a Teflon film separating two buffer-filled compartments. The hole in the chamber was pretreated twice with solution of the same lipid in heptane and allowed to dry. The areas of the bilayers were estimated using a binocular microscope to an accuracy of about 5%. Standard calomel electrodes with saturated junctions immersed directly into the solutions on either side of the membrane were used. All lipids were from Avanti Polar Lipids (Alabaster, Ala.). Dithiothreitol, heptane, decane, hexadecane, salts, and ethylene glycol-bis(P-aminoethyl ether)-N,N,N',N'-tetraacetic acid (EGTA) were purchased from Fluka (Fluka Chemika-BioChemika, Buchs, Switzerland); piperazine-N,N'-bis(2-ethanesulfonic acid) (PIPES) was purchased from Sigma Chemical Corp., St. Louis, Mo. Standard buffer (0.5 M KCl, 10 mM EGTA, 10 mM PIPES [pH 7.0]) and other solutions were prepared in
house-distilled, milli-Q-treated water. The P828 peptide (H3NArg-Val-Ile-Glu-Val-Val-Gln-Gly-Ala-Cys-Arg-Ala-Ile-Arg-HisIle-Pro-Arg-Arg-lle-Arg-COOH) was synthesized by Peptide Technologies, Inc. (Bethesda, Md.) and purified by high-performance liquid chromatography. Molecular weight was determined by time-of-flight mass spectroscopy. A 1.25 or 2.5 mM stock solution of the peptide in 0.1 M KCl-10% dithiothreitol was stored frozen in Eppendorf tubes at -70°C. After thawing, a portion of the peptide used for experiments was stored on ice for 1 to 2 days. All experiments were performed at room temperature, 20 to 220C. Measurements of surface tension and membrane stability. Membrane capacitance (c) was determined from the capacitive current in response to the application of a linear voltage sweep of 100 V/s with an amplitude of +20 mV. Surface tension (a-) was found by measuring the change in membrane capacitance under the given hydrostatic pressure gradient (43). Bilayer lifetimes (tj) were defined as the time from a step in voltage to the onset of irreversible membrane rupture of the membranes. Irreversible breakdown of bilayers in an electric field results from the development of lipidic pores with an overcritical radius, which tend to spontaneous expansion (2, 10). The experimental dependences of the mean t1 (averaged over no fewer than 10 measurements) on voltage applied (U) were fitted with the theoretical expression based on this model (2):
t, = A exp
IT-y2 [a( + C(E,/Em - 1)U2/2]kT
(1)
where A is a preexponential factor dependent on some additional model assumptions (2, 10), k is the Boltzmann constant, T is 300 K, y is the work of formation of the unit of the pore perimeter (linear tension of pore in the lipid bilayer), and &, = 80 and Em = 2, the dielectric permittivities of water and membrane, respectively. The least-squares method was applied to compute the values of y andA, providing the best agreement between calculated and experimental lifetimes. Measurement of the changes in membrane surface potential by a potentiodynamic technique. The capacitance of planar lipid bilayer depends on the transmembrane electric field (4). For asymmetric membranes (membranes after the addition of P828 to the cis chamber only), this field is the superposition of the difference of the boundary potentials A4b and of the external field U (1, 3, 11). We recorded and compared capacitative current curves with the application of triangular voltage pulses to the membrane (0, U1 > 0, U2 < 0, 0 and 0, -U1, -U2, 0). The sweep rate was 100 V/s, and IU11 and IU21 were in the range of 220 to 350 mV. The direction of the voltage sweep was changed by means of a switch which reversed the position of the electrodes in the circuit but not the position of the trace on the oscilloscope. In symmetrical systems, the capacitive current curves obtained from the forward and reverted sweeps coincide exactly. If the boundary potentials of the two sides of membrane were not equal, a constant bias potential (Uc) had to be applied to the membrane during the forward and backward voltage sweeps to match the two capacitive current curves. The value of Uc required for this is equal to A1b, with an accuracy of -2 mV to the measurement. To estimate the changes in surface charges of bilayers after the addition of P828, we used the Gouy-Chapman model (29) with a number of assumptions, including (i) that lipid and peptide charges are equally effective in changing the potential, (ii) that ions are considered as point charges, and (iii) that all charges are smeared uniformly over the membrane (29). The
7117
PORE FORMATION BY CYTOPLASMIC FRAGMENT OF HIV gp4l
VOL. 68, 1994
relation between electrostatic potential (4) at the surface of the membrane in 1:1 electrolyte of bulk concentration (Cb) and charge density (q) was as follows:
qI(8NeaIckT Cb) =
sinh(*/2kT)
(2)
where N is Avogadro's number, 8a is the dielectric constant, and eo is the permittivity of free space; this equation was applied to analyze our experimental data. We inserted into equation 2 the surface density q of 1 elementary charge per PS molecule with an area of 0.6 nm2 (0.267 C/M2) and an electrolyte concentration of 0.5 M to estimate the initial surface potential of the phosphatidylserine membrane to be -97 mV. Then, the same expression was applied to find the surface charge q' which would correspond to the value of the surface potential differing from the initial one by the A/4b measured at a given peptide concentration. The percentages of lipid molecules of PS and the DPC bilayer bound to P828 at different concentrations of the peptide were estimated assuming that each of the peptide molecules carried a net charge of +5 (21). Single-cell current recording. In order to compare results obtained with synthetic peptide to results from experiments planned with expression of env in a baculovirus vector system, we used Sf9 cells. The cells were grown using Grace's insect medium in an incubator at 27°C. Micropipets (Drummond; Thomas Scientific, Swedesboro, N.J.; resistance, 1.5 to 3.5 MfQ) were used for patch clamp recordings of single-cell currents in the whole-cell configuration (24). The pipet (internal) solution contained 170 mM K-glutamate, 6 mM KCl, 2 mM MgCl2, 5 mM EGTA, and 10 mM N-2-hydroxyethylpiperazine-N'-2ethanesulfonic acid (HEPES; pH 7.2). In the P828-involved experiments, the tip of the pipet was filled with peptide-free internal solution and then backfilled with internal solution containing or not containing P828 (25 to 100 nM). The external solution contained 8 mM glucose, 10 mM MgCl2, 20 mM MgSO4, 10 mM CaCl2, and 25 mM KCl. A List EPC-7 patch clamp amplifier was used for monitoring current changes (at a band width of 3 kHz) and for clamping at different voltages. The holding potential was varied in steps (in the range of -65 to +65 mV) to check the dependence of membrane conductance on voltage with and without P828. Membrane conductance of Sf9 cells was calculated from current measured during the first 1 to 2 min after onset of internal dialysis by obtaining whole-cell configuration at a holding potential of +65 mV with and without 25 to 100 nM P828 in the intracellular solution of patch pipet.
RESULTS
Binding of P828. An asymmetric addition of micromolar concentrations of P828 to the aqueous solution bathing a planar phospholipid bilayer membrane changed the boundary potential of the membrane monolayer exposed to peptide. The difference between this potential and the unchanged boundary potential of the other membrane monolayer, A*b, reached a new stationary level within 5 to 7 min. The magnitude of the potential asymmetry depended on the peptide concentration (Fig. 1), reflecting a change in membrane surface charge as a result of the binding of the strongly positive P828. Although P828 bound to the surface of neutral DPC bilayers (8.5 x 1011 molecules per cm2), there were =20 times more sites on negatively charged PS bilayers (2 x 1013 molecules per cm2). For PS bilayers in the presence of 2 ,uM of P828, we estimated
50r-
O PS 0
co
o o 00 0
40
c
0 00
E
o._.X c
0 0
0
30 Fo E: 00)
E 4c._ )
a) co -
20 [0
coC
-C.
)00nlo0
0 0. *I *
04
*_
.
: PC 0
_
12 4 8 [Peptide 828], FM
16
FIG. 1. Binding of the positively charged peptide to planar lipid bilayer characterized by the increase of the membrane boundary potentials. The increase of the boundary (surface plus dipole) potentials of planar lipid bilayers of PS (open circles) and DPC (filled circles) is plotted as a function of the concentration of the P828 added symmetrically.
that more than 60% of the membrane lipids were bound to
peptide.
Conductance of lipid membranes treated with P828. Addition of micromolar concentrations of P828 to the aqueous solution bathing a planar phosphatidylserine bilayer membrane led to an increase of membrane conductance in all experiments (n > 30). In approximately half of the experiments, we observed either short spikes of transmembrane current (Fig. 2A) or irregular increases of current up to 1 nA (Fig. 2B), with no steady-state current level established before membrane disruption (5 to 10 min after the addition of P828). In the other half of the experiments, we observed abrupt increases and decreases in current (Fig. 2C). Usually, it was possible to get some quasistable level of conductance after subsequent perfusion of the cell with peptide-free solution. The records were quite noisy and did not resemble recordings of well-defined ionic channels such as voltage-dependent anion channels of mitochondrial membranes shown for comparison in Fig. 2D. In contrast, the purely lipidic pores developed in the PS bilayer in the presence of lysophosphatidylcholine (LPC) (Fig. 2E) appear akin to those obtained in the presence of P828, albeit at lower amplitude. As expected from the large charge of P828, pore formation was dependent on applied potential. At concentrations of P828 of less than 10 ,uM, no changes of PS membrane conductance were observed at negative potentials. At potentials of +60 mV and higher, peptide-induced current fluctuations were observed (Fig. 3). Higher potentials were more favorable for the formation of long-living pores. Positive membrane potential was not an absolute requirement for P828-induced increases in membrane conductance. At higher peptide concentrations (two to four times), it was possible to register an increase in transmembrane current at low or negative potentials (Fig. 4). The amplitude histogram of long-lived pores, which are identified here as single jumps of membrane conductance with an amplitude of at least 0.5 nS
7118
J. VIROL.
CHERNOMORDIK ET AL.
A.
B.
OmV+4OmV
C.
1 OOpA 1 min
D. 5OpA 1 min.
OmV
E.
FIG. 2. Pores formed by P828 and other compounds. (A, B, and C) Current flowing through a PS planar lipid bilayer in the presence of the P828 (final concentrations, 3 jiM [record A] and 8 F.M [records B and C]) added to cis compartment at the times indicated by the arrows. Increased noise after peptide addition is due to stirring. The bathing solution contained 0.1 M KCI, 10 mM EGTA, and 10 mM PIPES (pH 7.03). The transmembrane potential was +40 mV. Descending and ascending arrows show moments when the potential was switched to 0. (D) An asolectin planar lipid bilayer in the presence of the voltage-dependent anion channel of rat liver outer mitochondrial membranes as an example of the well-defined ionic channel. The bathing solution contained 1 M KCl, 5 mM CaCl2, 1 mM PIPES, and 0.1 mM EDTA (pH 6.0). The transmembrane potential was +20 mV. (E) PS planar lipid bilayer in the presence of 20 ,uM palmitoyl LPC as an example of a lipidic pore. Increased noise starting just before the addition of LPC (shown by arrow) is due to stirring. The bathing solution consisted of 0.1 M KCl, 10 mM EGTA, and 10 mM PIPES (pH 7.03). The transmembrane potential was +25 mV.
and a lifetime of more than 2 s, shows a broad distribution with a mode (pore conductance which was observed the greatest number of times) at 4 nS (Fig. 5). For DPC membranes, the addition of P828 induced only short spikes of transmembrane current or, rarely, noisy current increases (data not shown). Conductance never reached steady-state levels because of membrane disruption soon after the addition of P828. Divalent and monovalent cation permeability. In symmetric monovalent cation solutions, the current/voltage characteristics of pores formed in the presence of P828 were linear while
open (Fig. 6). However, pores tended to close upon negative voltages. They showed cationic over anionic selectivity but no preference between potassium and sodium ions (Table 1). Membranes formed from the DPC-PS mixture in a ratio of 2:1 separating 2 from 10 mM MgCl2 were initially of very high resistance. After the addition of 10 to 15 ,uM P828, we reproducibly obtained increases in the transmembrane current. At an applied potential of + 100 mV, we observed discrete steps of conductance in the range of 50 to 300 pS. The reversal potential of -12 mV obtained with a MgCl2 gradient demonstrated that both magnesium and chloride ions permeated the
PORE FORMATION BY CYTOPLASMIC FRAGMENT OF HIV gp4l
VOL. 68, 1994
7119
25 r m Opening
CMClosing
20H LLI
500pA L 30 sec.
+ 120mV
15F
C0 0
10k 5
+4OmV + 2OmV
L-
z
OL
FIG. 3. The influence of membrane potential on pore formation by P828. P828 (final concentration, 4 FiM) was added to the cis compartment 10 min before the beginning of first record. The records were obtained for the same PS membrane in the standard buffer.
P828-induced pores. In addition, the superlinear relation between voltage and current indicates that at negative potential, magnesium current was actually greater than chloride current
1I
2 4 8 16 32 64 CONDUCTANCE, nS
FIG. 5. Amplitude histogram of long-living pores formed by 8 ,uM P828 in PS membranes. Pores were identified as a single jump of membrane conductance with an amplitude of at least 0.5 nS and a duration of more than 2 s. The total numbers of pores opening and closing that were recorded in different experiments were 43 and 23, respectively.
(Fig. 6B).
Lipid membrane stability and linear tension. A significant decrease in the lifetimes of PS and DPC membranes after the addition of peptide was noted during the conductance measurements described above. To quantify this effect, we measured the dependence of mean membrane lifetime on applied
P828 cis
tran
I+
2
+ 40mV 4
8
-I
irM
0. 5nA 2 min.
I
I
t
2 -
t
4
t 8
t 16
t
32
t
64
M
4OmV
FIG. 4. Current through the lipid bilayer versus time for various P828 concentrations. Upper record, -40 mV; lower record, +40 mV. In both experiments, P828 was added to the cis side of a PS membrane. Final concentrations are given in the figure. Standard buffer was used. An insert illustrates the configuration of the experimental setup, which is described in Materials and Methods.
voltage for planar PS bilayer membranes in the presence and absence of 2 ,uM P828 in both aqueous solutions bathing the membrane. In the range of voltages studied, peptide significantly decreased mean membrane lifetime (Fig. 7). We measured the specific capacitances and surface tensions of films to calculate the linear tensions of the pores (equation 1). The values of 4.23 ± 0.51 and 4.05 ± 0.55 mF/m2 for capacitance and 0.35 ± 0.1 and 1.32 ± 0.15 mN/m for surface tension were measured in the presence and absence of 2 ,uM peptide, respectively. Qualitatively, the decrease in surface tension of membranes in the presence of P828 is consistent with strong binding of P828 to membranes (see above). The least-squares fit to equation 1 describes well the voltage dependences of membrane lifetimes. From this fit, linear tensions were 6.6 x 10-12 and 9.8 x 10-12 N with and without peptide, respectively. P828 increases conductance of cell membranes. P828 was found to permeabilize membranes of Sf9 cells when it was added to the external medium. Treatment of cells with 100 ,uM P828 for 20 min at 21°C caused permeabilization of 55% ± 1.1% (n = 3) of cells for ethidium homodimer assayed by fluorescence microscopy (data not shown). To study the interaction of P828 with the cytoplasmic leaflet of biological membranes, whole-cell current was measured by the patch clamp technique (24). Internal dialysis of Sf9 cells with the peptide was found to cause an increased and unstable membrane conductance relative to Sf9 control cells (Fig. 8). The mean membrane conductance of Sf9 cells in the presence of P828 was 3,279 ± 507 pS (mean ± standard error [n = 27]) compared with a mean conductance of only 347 ± 84 pS (n = 26) which was found for control cells. As in planar lipid bilayers, cell membrane conductance was greater when the peptide-interacting side of the membrane was charged positively by application of positive voltage (Fig. 8). Interestingly, the P828 concentrations required to promote the increase in membrane conductance (25 to 100 nM) were at least 10 times less than those required to promote pore formation in planar lipid bilayers. Higher concentrations of
J. VIROL.
CHERNOMORDIK ET AL.
7120
TABLE 1. Ion selectivity of planar phospholipid bilayer membranes modified by P828 Bathing solutiona
Reverse potential' ± SD]) [mean (my;
Selectivit
KCI/KCl (0.5/0.1 M) KCIINaCl (0.5/0.5 M) MgCI2IMgCI2 (2/10 mM)
-21 ± 6 -1.5 ± 2 -12 ± 6
3.9 (PkIP.1) 1.1 (Pk/PNA) 0.5 (PMWIPC1)
2.0 l
A
§
0.5M KCI_ *O0.M KCI 3
1.5 _
a Compositions and concentrations of the electrolytes separated by the PS membrane are given as cis/trans. b A concentration or composition gradient of electrolytes was formed across a lipid bilayer in presence of 10 ,uM P828. Under these conditions, ions diffuse across the bilayer according to the ion selectivity of pores. Gradient of permanent ions will create potentials. These potentials were determined here by measuring the reverse potential, i.e., the external potential which had to be applied in order to null the current through the lipid bilayer. Reverse potentials were used to calculate the relative pore selectivities, defined as the ratios of the pore permeability for different ions (25).
1.0 _0 0 0.5 *
c
-
z
O i
n~~~~
LU
0
n
uC
-0.5
-
0
P828 in the pipet disrupted experiments too quickly to ascertain its effects.
-1.0 I
Il
I
I
-50 0 50 100 TRANSMEMBRANE VOLTAGE, mV
1
CI1
0.4 1 0.2
I
2mM cis 10OmM trans_
-
_
0.0r z
DISCUSSION Our data support the hypothesis that the carboxy terminus of the HIV envelope glycoprotein is involved in virus-induced cytopathology by triggering either necrosis (by disruption of osmotic regulation [14]), or apoptosis (via calcium entry [27, 35]). We found that the addition of P828 peptide synthesized from the sequence near the cytoplasmic end of gp4l causes considerable increase in the conductance of planar lipid bilayer membranes and cell membranes. The development of pores in lipid bilayers in the presence of P828 is consistent with the increase in liposome and cell membrane permeability upon addition of a peptide from an overlapping region of gp4l to the
LU
cc
100 r -082 4 *0
-
10k
-^P_^
L)
*
2pM of Peptide No Peptide
U1) CO)
.
-100 I
-50
0
50
1 100
.0
TRANSMEMBRANE VOLTAGE, mV FIG. 6. Current/voltage dependence of planar lipid bilayer modified by P828. (A) Line marked with open circles was obtained on planar lipid bilayer of PS after formation of 6 to 8 pores in a 0.1 M KCl-5 mM PIPES (pH 7.04) solution on both sides of the membrane. Line marked with filled circles was obtained in 0.5 M KCl-10 mM EGTA-10 mM PIPES (pH 7.03) solution on membrane with one pore active, which closed when the potential was lowered to +80 mV. Deviation from linearity at negative potentials is the result of irreversible closing of pores. Error bars represent fluctuation of transmembrane current around mean values at the given potentials. (B) Membrane was formed from the mixture of DPC and PS (2:1). cis and trans compartments contained 2 and 10 mM MgCl2, respectively. Both curves were obtained under the same experimental conditions in different experiments. In this particular experiment, we did not perfuse chamber with P828-free solution before recording the current voltage dependence but controlled possible formation of additional pores by returning the potential back to its initial value (+100 mV) after obtaining the curve. For the record presented in this figure, the conductance was unchanged.
0.1 LL
JL 0.01
0.001 100
200 300 400 500 TRANSMEMBRANE VOLTAGE, mV
600
FIG. 7. Decrease of the electromechanical stability of planar lipid bilayers in the presence of the P828. Voltage dependences of mean lifetimes of PS membranes in the presence and absence of 2 p.M P828 peptide are shown. All points are means ± standard errors (n = 10). Theoretical curves were calculated from equation 1 by using the following parameters for control membranes and for bilayers in the presence of P828, respectively. c, 4.05 and 4.23 mF/m2; a, 1.32 and 0.35 mN/m; A, 7.6 and -8.4; -y, 9.8 x 10-12 N and 6.6 x 10-12 N.
PORE FORMATION BY CYTOPLASMIC FRAGMENT OF HIV
VOL. 68, 1994
A. Control 20
5OpA
sec.
-65mV
rf--u + 65mV -65mV
B. 25nM
P828
20 sec.
OOpA
1
+ 65mV
FIG. 8. Increase of membrane conductance after internal dialysis of Sf9 cells with P828. Membrane currents of Sf9 cells were recorded in the absence (A) and in the presence of 25 nM P828 (B) at holding potentials of + 65 and 65 mV. The membrane conductance of control cells (no peptide added) was stable for at least 10 min. In the presence of P828, the conductance over time was unstable, and seals were usually lost during the first 10 min. -
extracellular medium (30, 31, 42). The main advantage of the electrophysiological assays applied here over permeability measurements is related to the remarkable sensitivity of conductance measurements that allows one to easily resolve a
single aqueous pore with a diameter of -0.4 nm (39). The effects of P828 on planar lipid bilayers depend on the composition of lipid bilayers and on the peptide concentration. Long-lived pores (lifetime of more than 2 s) that had developed in PS bilayer in the presence of 8 puM peptide had broadly distributed conductances with a mode at 4 nS. The estimated internal diameter of an aqueous pathway with a conductance of 4 nS conductance is about 2.7 nm (assuming a pathway length of 5 nm and solution conductivity of 20 fl x cm [25]). Such a large pore would be permeable to calcium and a variety of metabolites and small peptides. These results are consistent with the data reported earlier. The effects of the peptide env from amino acids 828 to 855 on an influx of sucrose were found to be significantly smaller than those on inulin influx (30). The diffusibility of a solute within a pore of comparable size is known to be significantly less than its value in bulk solution (6, 8). On the basis of molecular weights, pores formed by the env peptide from amino acids 828 to 855 (30) should have diameters of 1 nm for sucrose and 4 nm for inulin (8). This is consistent with an average pore diameter of 2.7 nm, estimated from the histogram of pore conductances (Fig. 5). The peptide fragment under consideration, without the transmembrane domain of the envelope glycoprotein molecule anchoring it to the membrane, is a water-soluble molecule. Thus, to compare the effective concentrations of the peptide with biologically relevant membrane densities of gp4l, we estimated how much peptide is actually bound to membranes at concentrations of P828 causing an increase in the conductance and a decrease in the stability of planar lipid bilayers. Our data show that P828 binds avidly to negatively charged planar phospholipid bilayer membranes, as reported for the binding of P828 to phospholipid vesicles using nuclear magnetic resonance (21). The potentiodynamic technique (1, 11) used here allowed us in addition to quantitate a 20-fold-lower
gp4l
7121
binding to neutral membranes. Despite the simplifications inherent in Gouy-Chapman theory (29), the two methods are in good agreement regarding the binding of P828 to PS. The maximum densities of bound P828 for DPC and PS bilayers were roughly estimated to be about 8.5 x 10" and 2 x 1013 molecules per cm2, respectively, which is not far from the density of envelope glycoprotein molecules in HIV virions, _10'2/cm2 (72 tetramers of gpl60 per HIV particle [22, 37]). Thus, the micromolar concentrations of P828 that we used in our experiments with planar lipid bilayers and the 0.2 to 3,uM concentration of the env peptide from amino acids 828 to 855 applied to liposomes (see reference 42) may correspond to biologically relevant densities of gp4l carboxy terminus in host cell membranes, especially during viral budding. Importantly, even lower concentration of P828 (25 to 100 nM) caused a significant increase in membrane conductance when the peptide was perfused into Sf9 cells to interact with the cytoplasmic surface of the membrane. In contrast, >100 ,uM concentrations of the gpl60 env peptide from amino acids 828 to 855 are required to permeabilize plasma membranes of Vero, HeLa, and Molt cells when it is added to the external medium (42), and 100,uM P828 had to be added to the external medium to significantly permeabilize Sf9 cells for the ethidium homodimer. Thus, there is a difference of 3 orders of magnitude between effective concentrations of the same peptide added to the opposite sides of a cell membrane. Even less P828 was required to permeabilize cell membranes when it was added to their inner leaflets (biologically relevant surface). Perhaps much of the strongly positive peptide added to the external medium binds to the negatively charged glycocalyx on the outer surface of the plasma membrane, dramatically decreasing the concentration of peptide available to interact with the membrane lipid bilayer. In addition, the inner leaflet of the plasma membrane is known to contain more negatively charged phospholipids than the outer leaflet (34), and, thus, it should be more sensitive to P828. What could be the structure of the pores that develop in the membrane lipid bilayers in the presence of the gp4l amphipathic peptide? In particular, are they ionic channels with walls formed directly by the peptide(s) or are they lipidic pores with pore walls formed, at least partially, by membrane lipids? The cationic selectivity of planar lipid bilayers treated by P828 in monovalent cation solutions (Table 1) does not support the idea that the wall of this pore (ionic channel) is formed by the peptide. Rather, one expects channels with anionic selectivity to be formed by positively charged peptides. However, cationic selectivity can easily be explained if the wall of the pore is formed, at least in part, by negatively charged PS. The formation of lipidic pores in membranes is thought to be the rate-limiting event in membrane rupture (2, 10); therefore, the theoretical expression for the average lifetime of a lipid bilayer includes a key parameter for lipidic pore development. This parameter is the linear tension of the pore, y. Since -y enters equation 1 exponentially, even a small decrement of linear tension should lead to the promotion of pores. The decrease of -y from 9.8 x 10-12 to 6.6 x 10-12 N after the addition of P828 can be compared with the decrease from 8.6 x 10-12 to 3.3 x 10-12 N (9) for phosphatidylcholine bilayers modified by LPC, the biological detergent known for its ability to promote lipidic pores formation (Fig. 2E) (9, 46). Thus, we propose that P828 promotes the formation of pores having a wall covered, at least partially, by lipid headgroups. P828 pore recordings with their variety of jump amplitudes, noisy baselines, and low ionic selectivities are reminiscent of recordings of membranes modified by some other peptides which form amphiphilic helices (13, 15, 16). The behavior of
7122
J. VIROL.
CHERNOMORDIK ET AL.
this group of membrane-active peptides differs significantly from that of ionic channel-forming peptides (alamethicin, nystatin, and gramicidin, etc.) that usually have high selectivity and a narrow range of channel amplitudes (39). The properties of pores formed by P828 in bilayers are also different from the highly specific natural channels of cell membranes in this regard (25). The rate of lipidic pore formation, as well as the sizes and lifetimes of pores, should depend more strongly on membrane lipid composition than the same properties of ionic channels. It follows that if HIV-induced cytopathology involves permeabilization of cell membrane by its interaction with the carboxy terminus of the envelope glycoprotein, the numbers and the sizes of lipidic pores formed in the plasma membranes of host cells should depend strongly on the membrane composition, even at the same density of gp4l. Thus, single-cell killing may be highly cell type dependent. The variability of the density of gp4l molecules and plasma membrane composition at different stages of infection could explain why the cytotoxicity of HIV bearing mutations in the C-terminal region of gp4l seems to depend on the cell type studied (18, 28). Of course, the fragment that we used in our study is only part of a much bigger molecule. Direct studies of the properties of the membranes of cells expressing gp4l have to be performed to verify the hypothesis under consideration and contrast it with an alternative hypothesis, that HIV-induced cytotoxicity involves the interaction of the C-terminal region of gp4l with calmodulin (41). We have noted above that P828 permeabilizes and destabilizes membrane lipid bilayers lowering lipidic pore energy in a manner similar to that of LPC, the lipid known to dramatically modulate the fusion reaction. Addition of the LPC between fusing membranes inhibits fusion (9, 12). Analogically, addition of the gp4l env peptide from amino acids 828 to 855 to the external medium was shown to inhibit gpl60-mediated cell-cell fusion (42). In contrast, addition of LPC to distal leaflets of membranes promotes fusion pore formation (9). This analogy between P828 and LPC effects allows us to hypothesize that the interaction of the carboxy-terminal portion of gp4l with inner leaflets of the viral or cell membrane may facilitate fusion pore development during virus entry and syncytium formation. This suggestion is consistent with a number of recent publications indicating a significant role of the transmembrane and cytoplasmic domains of the viral envelope glycoprotein in the membrane fusion step of virus entry and syncytium formation (19, 36, 38, 45). To conclude, a short, 20-amino-acid fragment of the cytoplasmic domain of the HIV envelope glycoprotein is shown here to interact strongly with phospholipid bilayer membranes and cell membranes, causing the formation of large pores at biologically relevant concentrations. One could hypothesize that similar increases in membrane permeability for Ca2" ions, a signal for apoptosis, may be caused by direct interaction of the gp4l carboxy terminus with the lipid bilayer of host cell membranes and may be a trigger for HIV-induced cytopathology.
ACKNOWLEDGMENTS We deeply appreciate the many contributions of Klaus Gawrisch to this study, including suggestion of the project, providing the initial synthetic P828, testing its molecular weight, and many useful discussions. We also thank Steven Gee and Evgenia Leikina for technical assistance. We appreciate the valuable help of Lynn Kelly in preparing
the manuscript. REFERENCES
1. Abidor, L. G., S. C. Aitian, L. V. Chernomordik, V. V. Cherny, and Y. A. Chizmadzhev. 1979. The determination of membrane bound-
2.
3. 4. 5.
6. 7.
ary potentials by potentiodynamic method. Dokl. AN SSSR 245: 977-981. Abidor, I. G., V. B. Arakelyan, L. V. Chernomordik, Y. A. Chizmadzhev, V. F. Pastushenko, and M. R. Tarasevich. 1979. Electrical breakdown of BLM: main experimental facts and their qualitative discussion. Bioelectrochem. Bioenerg. 6:37-52. Alvarez, O., and R. Latorre. 1978. Voltage-dependent capacitance in lipid bilayers made from monolayers. Biophys. J. 21:1-17. Babakov, A. V., L. N. Ermishkin, and E. A. Liberman. 1966. Influence of electric field on the capacity of phospholipid membranes. Nature (London) 210:953-955. Bashford, C. L., G. M. Alder, G. Menestrina, K. J. Micklem, J. J. Murphy, and C. A. Pasternak. 1986. Membrane damage by hemolytic viruses, toxins, complement, and other cytotoxic agents. A common mechanism blocked by divalent cations. J. Biol. Chem. 261:9300-9308. Bohrer, M. P., G. D. Patterson, and P. J. Carroll. 1984. Hindered diffusion of Dextran and Ficoll in microporous membranes. Macromolecules 17:1170-1173. Beschiaschvili, G., and J. Seelig. 1989. Melittin binding to mixed phosphatidylglycerol/phosphatidylcholine membranes. Biochemis-
try 29:52-58. 8. Bezrukov, S. M., I. Vodyanoy, and V.A. Parsegian. 1994. Counting polymers moving through a single ion channel. Nature (London) 370:279-281.
9. Chernomordik, L. V., G. B. Melikyan, and Y. A. Chizmadzhev.
1987. Biomembrane fusion: a new concept derived from model
studies using two interacting planar lipid bilayers. Biochim. Biophys. Acta 906:309-352. 10. Chernomordik, L. V., and Y. A. Chizmadzhev. 1989. Electroporation of bilayer lipid membranes: phenomenology and mechanism,
p. 181-192. In E. Neumann, A. Sowers, and C. Jordan (ed.), Electroporation and electrofusion in cell biology. Plenum Press, New York.
11. Chernomordik, L. V., G. B. Melikyan, N.
I.
Dubrovina, L. G.
Abidor, and Y. A. Chizmadzhev. 1984. Solvent free bilayers from squalene solutions of phospholipids. Bioelectrochem. Bioenerg.
12:155-166. 12. Chernomordik, L. V., S. S. Vogel, E. A. Leikina, A. Sokolof, H. 0. Onoran, and J. Zimmerberg. 1993. Lysolipids reversibly inhibit Ca2+, GTP, and pH dependent fusion. FEBS Lett. 318:71-76.
13. Christensen, B., J. Fink, R. B. Merrifield, and D. Mazerall. 1988. Channel forming properties of cecropins and related model compounds incorporated into planar lipid membranes. Proc. Natl. Acad. Sci. USA 85:5072-5076.
14. Cloyd, M. W., and W. S. Lynn. 1991. Perturbation of host-cell membrane is a primary mechanism of cytopathology. Virology
181:500-511. 15. Cruciani, R. A., J. L. Barker, S. R. Durell, G. Raghunathan, H. R. Guy, M. Zasloff, and E. F. Stanley. 1992. Magainin 2, a natural antibiotic from frog skin, forms ion channels in lipid bilayer membranes. Eur. J. Pharmacol. 226:287-296. 16. Diclohier, H., G. Molle, and G. Spach. 1989. Antimicrobal peptide magainin I from Xenopus skin forms anion permeable channels in planar lipid bilayers. Biophys. J. 56:1017-1021. 17. Fisher, A. G., L. Ratner, H. Mitsuya, L. M. Marselle, M. E. Harper, S. Broder, R. C. Gallo, and F. Wong-Staal. 1986. Infectious mutants of HTLV-III with changes in the 3' region and markedly reduced cytopathic effects. Science 233:655-659. 18. Gabuzda, D. H., A. Lever, E. Terwilliger, and J. Sodroski. 1992. Effects of deletions in the cytoplasmic domain on biological functions of human immunodeficiency virus type 1 envelope
glycoproteins. J. Virol. 6:3306-3315. 19. Gage, P. J., M. Levine, and J. C. Glorioso. 1993. Syncytiuminducing mutations localize to two discrete regions within the cytoplasmic domain of herpes simplex virus type 1 glycoprotein
B. J. Virol. 67:2191-2201. 20.
Garry, R. F. 1989. Potential mechanisms for the cytopathic prop-
erties of HIV. AIDS 3:683-694. 21. Gawrisch, K, K. H. Han, J. S. Yang, L. D. Bergelson, and J. A. Ferretti. 1993. Interaction of peptide fragment 828-848 of the envelope glycoprotein of human immunodeficiency virus type I with lipid bilayers. Biochemistry 32:3112-3118.
VOL. 68, 1994
PORE FORMATION BY CYTOPLASMIC FRAGMENT OF HIV gp4l
22. Gelderblom, H. R, M. Ozel, and G. Pauli. 1989. Morphogenesis and morphology of HIV. Structure-function relations. Arch. Virol. 106:1-13. 23. Haffar, 0. KI, D. J. Dowbenko, and P. W. Berman. 1991. The cytoplasmic tail of HIV-1 gpl60 contains regions that associate with cellular membranes. Virology 180:439-441. 24. Hamill, 0. P., A. Marty, E. Neher, B. Sakmann, and F. J. Sigworth. 1981. Improved patch-clamp techniques for high-resolution current recording from cells and cell-free membrane patches. Pflueg. Arch. 391:85-100. 25. Hille, B. 1992. Ionic channels of excitable membranes, p. 294-296. Sinauer Associates, Sunderland, Mass. 26. Hirsch, V., N. Reidel, and J. I. Mullins. 1987. The genome organization of STLV-3 is similar to that of the AIDS virus except for a truncated transmembrane protein. Cell 49:307-319. 27. Laurent-Crawford, A. G., B. Krust, Y. Riviere, C. Desgranges, S. Muller, M. P. Kieny, C. Dauguet, and A. G. Hovanessian. 1993. Membrane expression of HIV envelope glycoproteins triggers apoptosis in CD4 cells. AIDS Res. Hum. Retroviruses 9:761-773. 28. Lee, S.-J., W. Hu, A. G. Fisher, D. J. Looney, V. F. Kao, H. Mitsuya, L. Ratner, and F. Wong-Staal. 1989. Role of carboxyterminal portion of the HIV-1 transmembrane protein in viral transmission and cytopathogenicity. AIDS Res. Hum. Retroviruses 5:441-449. 29. McLaughlin, S. 1977. Electrostatic potentials at membrane-solution interfaces. Curr. Top. Membr. Transp. 9:71-134. 30. Miller, M. A., M. W. Cloyd, J. Liebmann, C. R. Rinaldo, Jr., K. R Islam, S. Z. S. Wang, T. A. Mietzner, and R. C. Montelaro. 1993. Alterations in cell membrane permeability by the lentivirus lytic peptide (LLP-1) of HIV-1 transmembrane protein. Virology 196: 89-100. 31. Miller, M. A., R F. Garry, J. M. Jaynes, and R. C. Montelaro. 1991. A structural correlation between lentivirus transmembrane proteins and natural cytolytic peptides. AIDS Res. Hum. Retroviruses 7:511-519. 32. Montall, M., and P. Mueller. 1972. Formation of biomolecular membranes from lipid monolayers and a study of their electrical properties. Proc. Natl. Acad. Sci. USA 69:3561-3566. 33. Mueller, P., D. 0. Rudin, T. Tien, and W. C. Westcott. 1962. Reconstitution of cell membrane in vitro and its transformation into an excitable system. Nature (London) 194:979-980. 34. Op den Kamp, J. A. 1979. Lipid asymmetry in membranes. Annu. Rev. Biochem. 48:47-71. 35. Orrenius, S., M. J. McCabe, Jr., and P. Nicotera. 1992. Ca2+dependent mechanisms of cytotoxicity and programmed cell death. Toxicol. Lett. 64/65:357-364.
7123
36. Owens, R. J., C. Burke, and J. K. Rose. 1993. Mutations in the membrane-spanning domain of the human immunodeficiency virus envelope glycoprotein that affect fusion activity. J. Virol. 68:570-574. 37. Pinter, A., W. J. Honnen, S. A. Tilley, C. Bona, H. Zaghouani, M. K. Gorny, and S. Zolla-Pazner. 1989. Oligomeric structure of gp4l, the transmembrane protein of human immunodeficiency virus type 1. J. Virol. 63:2674-2679. 38. Salzwedel, K., P. B. Johnston, S. J. Roberts, J. W. Dubay, and E. Hunter. 1993. Expression and characterization of glycophospholipid-anchored human immunodeficiency virus type 1 envelope glycoproteins. J. Virol. 67:5279-5288. 39. Sansom, M. S. P. 1991. The biophysics of peptide models of ion channels. Prog. Biophys. Mol. Biol. 55:139-235. 40. Sodroski, J., W. Haseltine, and M. Kowalski. 1991. Role of the human immunodeficiency virus type 1 envelope glycoprotein in cytopathic effect, p. 193-199. In N. Duzgunes (ed.), Mechanisms and specificity of HIV entry into host cells. Plenum Press, New York. 41. Srinivas, S. K., R. V. Srinivas, G. M. Anantharamaiah, R. W. Compans, and J. P. Segrest. 1993. Cytosolic domain of the human immunodeficiency virus envelope glycoproteins binds to calmodulin and inhibits calmodulin-regulated proteins. J. Biol. Chem. 268:22895-22899. 42. Srinivas, S. K., R V. Srinivas, G. M. Anantharamaiah, J. P. Segrest, and R. W. Compans. 1992. Membrane interactions of synthetic peptides corresponding to amphipathic helical segments of the human immunodeficiency virus type-1 envelope glycoprotein. J. Biol. Chem. 267:7121-7127. 43. Sukharev, S. I., L. V. Chernomordik, I. G. Abidor, and Y. A. Chizmadzhev. 1992. Effects of uranyl ions on the properties of bilayer lipid membranes. Bioelectrochem. Bioenerg. 9:133-140. 44. Venable, R M., R. W. Pastor, B. R. Brooks, and F. W. Carson. 1989. Theoretically determined three dimensional structures for amphipathic segments of the HIV-1 gp4l envelope protein. AIDS Res. Hum. Retroviruses 5:7-22. 45. Weiss, C. D., and J. M. White. 1993. Characterization of stable Chinese hamster ovary cells expressing wild-type, secreted, and glycosylphosphatidylinositol-anchored human immunodeficiency virus type 1. J. Virol. 67:7060-7066. 46. Weltzien, H. U. 1979. Cytolytic and membrane-perturbing properties of lysophosphatidylcholine. Biochim. Biophys. Acta 559:259287. 47. Wilk, T., T. Pfeiffer, and V. Bosch. 1992. Retained in vitro infectivity and cytopathogenicity of HIV-1 despite truncation of the C-terminal tail of the env gene product. Virology 189:167-177.
