Gramicidin,
VDAC, porin and perforin
channels
D. Levitt Department
of Physiology, Current
Opinion
University in Cell
This review covers the miscellaneous channels that do not fall w=ithin the domain of the other reviews in this section. Three different channel systems will be reviewed: the gramicidin A channel; the large weakly selective channels that are present in the outer membrane of mitochondria (VDACs) and bacteria (porin); and the large channels whose function is to kill cells by colloid osmotic hemolysis (perform).
2:689-694
High resolution X-ray crystallography has shown that when gramicidin is crystallized from bulk solutions, it forms a different ‘channel-like’ structure referred to as the gramicidin ‘pore’, in which the two peptide chains are in an antiparatlel double-stranded helix (Langs, Science 1988, 241:188-191; Wallace and Ravikumar, Science 1988, 241:182-187). Unfortunately, as these ‘pore’ sttuctures cannot be related to a physical channel conductance, their usefulness is limited. It is possible that they represent the low-conductance ‘mini-channels’ that are seen at frequencies varying from 5 to 50% of that of the large channel [ 11. Despite considerable effort, little is known about these mini-channels. It seems clear that they are a different conformation of the gramicidin channel and are not the result of impurities in the preparation. Indirect evidence suggests that these mini-channels re-
The gramicidin A channel is the simplest, best characterized ion channel (for reviews, see articles in Truqbort Gurikrs,
1990,
Channels and Pumps,
edited by Pulknan et al. Kluwer Acudemic Publtiers, 1988). It is formed from two identical 15-amino-acid peptides (Fig. 1). Although there is still no direct (i.e. crystallographic) proof of channel structure, a large num-
I
Cl’*ytotoxic
T lymphocytes;
@
Abbreviations Nh4R-nuclear magnetic resonance; VDAC-voltage-dependent anion channel.
Current
Biology
USA
Nati Acud Sci USA 1971, 68:672476).
A
through Membranes:
Biology
Minneapolis,
ber of different approaches have provided strong evidence that the conducting channel in lipid membranes is a head-to-head left-handed, single-stranded P-helix with 6.3 residues per turn as originally proposed by Urry (Proc
Introduction
Gramicidin
of Minnesota,
Ltd
ISSN
0955+674
Fig. 1. Model of the gramicidin channel. It IS formed from an NHz-terminal to NH*-terminal dimer of 8-helices The rihbon indicates the peptide backbone that lines the channel, while the solid black region corresponds to the hydrophobic side chains. Published by permission (Andersen et al., In Transport through Membranes: Carriers, Channels and Pumps; edited by Pullman et al. 1988, p. 118).
W-pore-forming
protein;
690
Membrane
peryeability
suit from very small changes in the dominant conducting channel, and thus are probably not the double-stranded ‘pore’ structure described earlier (Busath and Szabo, Bio plys J 1988, 53689-695). The fmding of Sawyer et al [l] that low levels of detergent increase mini-channel frequency is the lirst success in being able partially to control their number. The gramicidin channel may not be a good model for biological ion channels for several reasons. It is a uniform cylinder, about 0.42 nm in diameter and 2.5 nm long, that is just large enough to hold a row of about nine water molecules. The channel is so narrow that the ion and water cannot get past each other and so the movement of an ion through the channel must displace all the water ions. This is one example of the ‘single-file’ effect which dominates the transport kinetics in this channel (Levitt, Cuw Tqo Memb Tran 1984,21:181-197). In contrast, many biological channels have relatively wide mouths, tapering to short narrow, selectivity regions (Dani, Curr Opin Cell Biol 1989, 1:753-764). These structural differences can produce qualitative differences in the transport kinetics. For example, the movement of an ion within the gramicidin channel may be limited by the water-wall interaction of the nine water molecules that it must displace (Dani and Levitt, Bic@ys J 1981, 35:501-508) while ion kinetics in biological channels may be limited by ion-wall interactions, especially in the narrow selective region. In addition, the gramicidin channel is uncharged and its cation selectivity presumably arises from interactions with the carbonyl oxygens that form the pore wall. This interaction may be relatively unimportant in many biological channels where ion selectivity is probably dominated by fixed charges. Whatever the biological relevance of the channel, the combination of its extreme simplicity and the knowledge of its (probable) structure means that, for now and for the foreseeable future, it is the only channel in which detailed theoretical models can be tested. Gramicidin can be regarded as the proving ground for biophysical theories of channel conductance. One factor which must be included in any physical model of the channel is the electrostatic potential that arises from ion-ion interaction and from the low dielectric of the channel wall and lipid. Simple continuum theories provide estimates of these potential energy barriers that are in reasonable agreement with the experimental data (Jordan, J Pbys C5em 1987, 91:65826591). In the past, these continuum calculations have either neglected the presence of ions other than the transport ion or have included them in a very approximate manner (Levitt, Bio p&s J 1985, 48:19-31). Recently, Jordan et al [2] have performed a much more accurate and complete treatment of the influence of the other ions. As one might expect, these ions tend to screen each other and lower the potential. Because the narrow gramicidin channel only holds one or two ions, these effects are small within the channel but can become significant outside the channel. The major limitation of these calculations is that the system must be close to equilibrium. Thus, they are not applicable to the important case of large net fluxes.
The biophysics of gramicidin is simple enough that, in theory, it can be solved from first principals, i.e. the inter- and intramolecular forces of the channel, membrane lipid, water and ion are simulated and the classic movement of the ion in these forces is solved exactly. This is the molecular dynamic approach that has been applied to the gramicidin channel by several groups (Polymeropoulos and Brickmann, Ann Reu Biopbys Cbm 1985, 14:315-330). The use of molecular dynamics has now become a standard tool for studying protein movements and structures and commercial software packages are now available. Chiu et al [3] have applied the GROMOS package (Hermans et al, Biopo&mers 1984, 23:23-84) to the gramicidin channel. They investigated the ion-free channel and concentrated on the movements of the carbonyl oxygens and the interaction and structure of the channel water molecules. The carbonyl oxygens flex and rotate slightly out into the channel where they form hydrogen bonds with the water. The most dramatic result is that all the channel water molecules seem to align themselves with each other, with all their dipoles pointing in the same direction. The reorientation of the dipoles in the opposite direction is a very rare event occurring only once in the 70~s simulation. This is surprising as one would expect one-dimensional water-water interactions to be relatively short range (Landau and hfshitz, Statkticul Pbyszb, Pergamon Press, 1958, p. 482) and water-wall interactions to dominate. Mackey et al. (Bioply J 1984,46:229-248) saw essentially the same effect as Chiu et aL. [3], while Lee and Jordan (Biopbys J 19,46:80%319), using a very different model of the water and channel did not observe this correlation. These results illustrate the major limitation of the molecular dynamic approach: most of the theoretical predictions cannot be experimentally tested. Calculating the channel conductance is a particularly difficult problem. Accurate estimates of ion conductance require that the ion channel potential profile be known to about 1 kT while the molecular dynamic calculations require knowledge of intermolecular potentials that are several orders of magnitude larger. Potentials that have been found to be adequate for calculations of protein structure may lead to large errors in the estimates of ion channel conductance. One approach which could offer some direct feedback on the molecular dynamic results is to combine them with nuclear magnetic resonance (NMR) measurements (Cl-h et al, Biopky J 1990, 57:lOla). The limitations of the molecular dyamic approach are illustrated in the evaluation of equilibrium cation potentials in the channel. Because cations have very long-range forces, they pose special problems. Previous estimates of the equilibrium potential have differed markedly from experimental estimates. &vist and Warshel [4] have developed a new approach to this problem that yields the first reasonable estimate of these potentials. They use a highly simplified model of the dielectric properties of the water and lipid that allows them to take account of the longrange forces.
Cramicidin,
Another factor limiting the usefulness of molecular dynamic calculations is the considerable uncertainty about the relevant experimental quantities needed to evaluate them. For example, from molecular dynamics one can calculate the equilibrium potential profile in the channel. Estimating this equilibrium potential from the experimental channel conductance requires an assumption of a detailed kinetic model, and different models can lead to markedly different estimates of the potential profile (using the same conductance data). Recently, direct measurements of equilibrium binding using NMR techniques have been made [ 51 and the future extension of this approach should provide better experimental data.
VDAC
and porin
These two classes of channels, the voltage-dependent anion channel (VDAC) and porin are discussed together because they have similar general function and structure. Both are large-conductance large-diameter ion channels with thin walls formed by a P-sheet structure and located in the outer membrane of either mitochondria (VDAC) or bacteria (porin). Because of these similarities, VDAC has been referred to as ‘mitochondrial porin’. However, this nomenclature is confusing because it obscures the fact that these two classes have no ammo acid homology and that they are clearly different proteins with major differences in structure and function. VDAC is probably related to some mitochondrial inner membrane transport proteins (Mannella and Auger, BzbpbysJ 1986, 49:272a). VDAC forms a large (about 3nm in diameter) slightly anion-selective channel with complex voltage dependence (reviewed by Colombini [6], see also the mini-review series in J Bioenerg Biomemb 1989, 21:417-507). Compared with other channel proteins, it is relatively
VDAC, porin
and perforin
channels
Levitt
small: the yeast protein consists of 283 ammo acid residues (molecular weight 29883). A low-resolution structure of the channel has been determined by electron microscopic imaging on two-dimensional crystals [7]. The channels have an outer diameter of approximately 5 nm and an inner diameter of between 2.5 and 3 nm. What differentiates these channels from the structure of most other ion channels is that they have very thin walls, about the thickness of one ammo acid chain. The amino acid sequence has a pattern of alternating polar and non-polar side chains that suggests that it forms a P-sheet structure in which the polar residues point into the channel and the non-polar residues contact the lipid. This suggestion has been directly confirmed by sitedirected mutagenesis on the yeast protein (81. Colombini and his colleagues [8] found that changing the charge at some locations produced little or no change in conductance, but marked changes in ion selectivity. Assuming that a residue in which a charge change alters selectivity must line the channel wall, the authors screened the entire sequence to find the segments that contributed to the wall structure. (This structural assignment may be in slight error because charges at the pore mouth should also influence ion selectivity [lo] .> The results suggest that each VDAC molecule consists of 12 transmembrane ~-strands and a transmembrane u-helix. A model that summarizes the structural and functional properties of the channel is shown in Fig. 2. The channel is shown as a dimer of VDAC, although there is no direct evidence for this. The voltage-dependent closure results from the transfer of a portion of the protein from the channel wall to the membrane surface, decreasing the membrane diameter and changing its charge and ion selectivity. The structure of bacterial porin differs significantly from that of VDAC. The low-resolution (about 23A) structure of one type of pot-in (OmpF) has been determined by
p sheet
I
u helix
open
closed
Fig. 2. Proposed model of the open and closed form of the voltagedependent anion channel NDAC) channel. The channel is a homodimer with monomers oriented in opposite directions, each monomer consisting of a psheet (thin-wall region) and one a-helix. When the channel closes, an a-helix and part of the p-sheet region move from the channel wall to the membrane surface. (Kindly provided by M. Colombini.)
691
692
Membrane
permeability
electron microscopic imaging techniques on negatively stained two-dimensional crystals (Engel @ al, Nature 1985, 643-645). It has a remarkable structure, consisting of a trimer in which three channels at the outer surface converge and merge into a single opening at the inner surface. This unusual structure leads to correlated single channel openings and closings. Using similar technology to Engel et al, (1985), Jap [9] has determined the structure of another porin (PhoE) to a resolution of about 18A A model that summarizes the data is shown in Fig. 3. It is also a trimer with three openings on the outer surface that converge but, unlike the OmpF protein, do not merge at the inner surface. Although the resolution of the images is not high enough to determine the detailed structure of the channel walls, they are shown as P-sheet structures in Fig. 3, as suggested by a large number of other physical measurements.
ment proteins (C5, C6, C7, C8 and C9) aggregate to form the pore, referred to as the ‘membrane attack complex’ (Mtiller-Eberhard, Ann Ret) Immunoll986, 4:503-528). Because of the complexity of this process, little is known about the structure of this channel (Stanley, Curr Top Micrab Immunoll988,140:49-65). Not even the stoichiometry of the complex is agreed upon. This is one reason for the excitement greeting the discovery that cytotoxic T lymphocytes (CTL) use a protein that is homologous to complement C59 to kill cells (reviewed in [ 161. The purified protein, referred to as perforin, cytolysin or lym phocyte PFP, can produce channels on cells, vesicles and black lipid membranes, indicating that the channel is a homopolymer and should be more amenable to structural studies than the membrane attack complex.
The porin PhoE is anion-selective, while the porins OmpF and OmpC are cation-selective. By investigating fusion products of the proteins (OmpC-PhoE and OmpF-PhoE), Benz et al. [lo] were able to determine the amino acid residues responsible for the selectivity Most of the selectivity was accounted for by a short 74amino-acid sequence at the N-terminal end. The anionselective PhoE was converted into the cation-selective OmpC by substituting two negative aspartic acids for a positive lysine and neutral serine. It is thought, based on evidence from other studies, that these amino acids are positioned at the pore mouth. One can expect that the continued application of molecular biological techniques to this system should soon provide a detailed map of the amino acids that influence the channel selectivity.
Extracellular
Periplasmic
Complement
and perforin
There are a large number of channels whose function is to produce cell membrane holes that lead to cell death by colloid osmotic hemolysis (reviewed by Bhakdi and Tranum-Jensen, Ret1 Pbysiol Biocbem Pharmucol 1987, 107:147-233). The definitive test for this mechanism of cell death is that it can be prevented by adding a large impermeant molecule to the bathing solution that balances the osmotic pressure of those cellular components too large to pass through the hole. In some cases this test fails because the holes are so large that even highmolecular-weight dextrans can pass. This channel class, referred to collectively as pore-forming proteins (PFPs), includes such disparate compounds as staphylococcal atoxin [ 111, streptolysin (which is also related to the toxin from the mushroom amanita), toxins from Gram-negative bacteria [12] and bee venom (melittin) [ 131 and toxins made by amoeba [ 14,151 and possibly responsible for amoebic dysentery. Because of their clinical importance there is most interest in the PFPs of the immune system. It has been known for many years that PFPs are the primary mechanism of complement-mediated cell death. Five of the comple-
VW
Fig. 3. A schematic model of the PhoE channels. The channel walls are shown as P-sheet structures. On the extracellular side, there is large vestibule, about 3.5 nm long that has an elliptical opening with major and minor axis of 1.8 and l.Onm. At the periplasmic side, the channel narrows to an inner diameter of 1 nm with a length of 1 nm. The outer leaflet of the membrane is lipopolysaccharides and the inner leaflet is phospholipids. Published by permission 191.
The pores formed by perforin are very large, with a diameter of 915 M-I. At high concentrations, ring-like ‘tubular’ membrane lesions with diameters of about I6 nm can be observed under the electron microscope. These are not observed at lower concentrations when smaller, ‘nontubular’ channels are probably formed [ 161. This is supported by the fact that, in bilayer studies, a heterogeneous population of single-channel conductances is observed (Young et al, J Exp Med 1986, 164:144-155). The structural characterization of the channel in terms of the amino acid sequence is only at a very prelim inary stage. The perforin sequence is predominantly hydrophilic with a short stretch of about 60 hydrophobic amino acids that could form two membrane-spanning a-helices [ 171. There are two other amphiphilic regions that might form an a-helix or P-sheet structure. The per-
Gramicidin,
VDAC, porin
and perforin
channels
Levitt
C8a
Fig. 4.
A simplified diagram of the amino acid sequences of the protein complement components that form the chan-
Perforin
N
Cal+
piij
nel complex K6,C7,C8,C9) and perforin. The heavy solid line indicates the region with high homology. ‘***’ indicates the
C
region bound
535
of C9 which is labeled by lipidprobes. Published by permission
[18l.
forin channel may not follow the usual rules in which the channel wall is formed primarily from hydrophobic amino acids. The wall might consist of a single thickness of either a P-sheet (e.g. porin, VDAC) or a-helix, in which case one would expect to find amphiphilic amino acids, with the hydrophilic groups in contact with the channel water and the hydrophobic groups in contact with the lipid. One approach to interpreting the amino acid sequence is to look for homologies with the complement proteins. With the sequencing of C6, all the components of the ‘membrane attack complex’ have now been sequenced [18,19]. Fig. 4 shows a simplified diagram of the homology between G-9 and perforin. They all share a contiguous region of about 370 amino acids. If one assumes that perforin and complement have similar membranespanning structures while differing in the other domains, then this similarity identifies the transmembrane regions of the sequence. The complement protein C9 has another, more surprising, homology with melittin, the 26. amino-acid protein that is the main component of honey bee venom (there are two sequences of four and six residues that are homologous). Furthermore, antibodies against these sequences crossreact between melittin and C9 [20]. Because of the small size of melittin, one might expect that it would be easier, at least compared with complement or perforin, to determine the structure of this channel. However, at present, little is known about melittin and it is not even certain that the melittin-induced cell lysis is produced by f&d channels [ 131. The discov-
ery of the sequence similarity should stimulate renewed interest in this question.
Conclusions The study of these ‘other’ channels is a useful reminder of the diversity of channel structure. The structures of gramicidin (P-helix), and porin and VDAC (P-sheet) are well established and differ from the hydrophobic a-helical structure that is the conventional channel building block. In addition, the ion selectivity of gramicidin arises from neutral amino acids while that of porin and VDAC is produced by charged amino acids. One can also expect that other aspects of channel function, such as the importance of diffusion limitation, ion-ion interaction and ion-water interaction will vary from channel to channel.
Annotated
references
and recommended
reading 0 *a
Of interest Of ountanding
1.
SWAYER DB.
0
tance
interest KOEPPE RE. ANMRSEN
heterogeneity
28:65X 4,583. The number of low conductance the addition of low concentrations
in gramicidin
OS: Induction of conducchannels. BibcAm 1989,
‘mini-channels’ of detergent.
can be increased b> It is suggested that this
693
694
Membrane
permeability
is the mectkism among results
responsible from d&rent
for the variation laboratories.
in ‘mini’ frequency
found
2.
JORDAN PC, BACQUET RJ. MCC~ION JA, T~LW P: How electrolyte shielding influences the electrical potential in transmembrane ion channels. Bio&a/ 1989, 55:1041-1052. The first rigorous calculation of the importance of electrolyte shielding on channel conductance. The conclusions are limited to the near-equilibrium regime. l e
3.
CHtll
l
and polypeptide
S. JAKOB~~ON E. MCCAM~~ON conformations in the gramicidin
S, SlIBRAMANVUI
JA: Water
channel.
Bisbhys J 1989, 56:253261. . Application of the GROMOS molecular dynamic package to gramicidin. It is limited to the ion-free channel and focuses on the structure of water in the channel and the channel-water interaction. I(QVIST J, WARSHEL AZ Energetics of ion permeation through membrane channels. Bi@vs J 1989, 56:171-182. The first direct calculation of the eqiilibrium potential energy of an ion in the gramicidin channel that yields a result consistent with the experimental data. Long-range electrostatic forces are incorporated using a very simple approximation for the aster and lipid. 4.
l e
5.
HI~V~ON JF, JERNANDU.
l
namic parameters for the binding of the divalent cations to gramicidin A incorporated into a lipid environment by Tl205 nuclear magnetic resonance. Bi~s.11989. 55:27-330.
JQ,
SHIJNG~
C. MILLET
Measurement of the entropy and enthalpy lent and divalent ions to gramicidin channels vesicles is summarized. 6.
. Review 7. l e
FS: Thermody-
of the binding of monova in Iysophosphatidyicholine
This is the highest resolution obtained for any ion channel. nels have thin walls and internal diameters of about 3 nm. BLACHl.Y-DI’SOii
changes structural
memof 2D
The
than-
M. FORTE F: Selectivity ion channel: Science 1990, 247:123.+1236. Each of the positive bsine residues was .separJteh converted to a nega tive residue and the influence on the channel ion xkctiviv wz meaured. A model of the channel was formed based on the assumption that those sites that produced a change m ion selecti\iv must line the channel wdll. 9.
E. PENI) S. cOmhlHINI
in site-directed implications.
JAP BK: Molecular
mutants
design
of the VDAC
of PhoE
pot-in
l
BENZ R. %CtihlII) A. VAN DER LIZI, P. TOhlhbGSEI‘: J: Molecular basis of porin selectivity: membrane experiments with
OmpC-PhoE
and OmpF-PhoE
hybrid
proteins of Escherichia 1989, 981:%1-1.
coli K-12. Bixhim Ric&s Am Most of the ion selectivi~ was as.sociated with a short sequence at the N-terminal end. The results suggest that the fixed charges that produce the ion selectivity are krJkd at the channel mouth. 11.
FORTI 5. IM~~~~~~~
A: Staphylococcal a-toxin of liquid vesicles by cholesterol
l
permeability
increases and pH
ROPELE M. MENEST~UNA G: Electrical properties and molecular architecture of the channel formed by Escherichia coli hemolysin in planar Lipid membranes. BiodMm BiOpqs AC&~
1989. 985+18. The channel has a conductance of about 300 pS in 0.1 M KCI. The cation selectivity is reduced by acid pH, suggesting that tixed negative charges are invohred. KATSLI T. KUROKO M, MORIKAWA T. SANCHIKA K, FLIJ~TA Y, YMLIKA H, UDA M: Mechanism of membrane damage induced by the amphipathic peptides gramicidins and melittin. Biocbim Bicpkys Aclu 1989. 983:135141. Two different mechanisms have been proposed for the melittin-induced permeability: discrete channels and disordering of the lipid. This paper proposes a third mechanism invohing membrane fragmentation. 13. l
14. l
YOUNG JD, ILXVREY DM: Biochemical terization of a membrane-associated from the pathogenic ameboflagellate
and functional characpore-forming protein Nuegletia fowled J
Biol Uxnr 1989, 264:1077-1083. The pore has a single channel conductance varying from 150400 pS in 0.1 M NaCl and a channel diameter of about 5 nm. It is immunologically distinct from perforin and the PFP from Enf~~rnoeba bisto!)dca [ 151, 15.
K!ZUER F, HANKE W, TR~SSL D, BAKKER-GRLINWAID T: Pore-forming protein from Entamoeba histolytica forms voltage and pH controlled multi-state channels with properties similar to those of the barrel-state aggregates. Biccbem BiqLys
Acfu 1989, At low voltages ( < about 100% pS is ductance channels multimer.
982:89-93. 50 mV), a dominant single channel conductance of seen. At higher voltages a multitude of lower conis seen, suggesting a break up of the one dominant
JD: Killing of target cells by lymphocytes: A mechview. PLyiol Ret1 1989, 691250-314. An excellent review of an area in which there has been an exponential growth in the number of publications in recent years. 16.
YOUNG
l
artistic
1’.
Kwo~: BS, WAK~ICHIR M, Ltu C. PERSECHINI PM, TRAPANI JA, &Q AK, KIM Y. YOUNG JD: The structure of the mouse lymphocyte pore-forming protein perforin. Biocbenz BiqLys Res Comm 1989s 58:1-10. The authors present the sequence of mouse perforin and discuss the possible implications of the sequence for the channel structure. l
18.
and its functional
l e consequences. .I ,llol Rio/ 1989. 205:-107+19. Three-dimensional structure of the channel at a rewlution of about IS& determined hy negative stain imaging of two-dimensional crystals. This low-resolution structure provides the tirst direct images of this channel type. The channel is a trimer in which the three channels converge but do not merge in going from the outer to inner surface of the membrane.
10.
l
channel,
htANNEUA CA: Structure of the mitochondrial outer brane channel derived from electron microscopy crystals. J Biomq Biomrmb 1989. 213.127437.
l e
12.
l
COLOMBIM M: Voltage gating in the mitochondrial VDAC J ,+k& Rio/ 1989. 111:10~111. of the VDAC channel.
8.
pendent assembly of oligomeric channels. E~tr J Biocbm 1989, 181~767-773. The formation of the channel in vesicles is a relatively slow process involving assembly of G-10 monomers. Assembly requires conformational changes which are influenced hy the pH.
the de-
DISCIPIO RG. Hucu TE: The molecular architecture of human complement component C6. J Biol Gem 1989, 264:16197-16206. The sequence of complement C6 is presented. This complete the sequencing of all the factors that form the channel. The relation of this factor to the other complement factors and perforin is discussed. l e
19. l e
0iAtiKAvAR-n HJ: Structural
other
channel
DN. Cn%aAvAan homology of
forming
B, PARRA CA, MLIUR-EBERHARII complement protein C6 with
protein
of complement.
Acud Sci USA 1989, 86:279$L2803. A paper published simultaneousiy with [ 18) with discussion. 70. l
sirmlar
Proc
Null
results
and
LUNE RO. ESS~R AF: Identitication of the discontinuous epitape in human complement protein C9 recognized by antimelittin antibodies. J fnmunol 1989, 143:553557.
A demonstration that antibodies against homologous in C9 and m&tin cro.ssreact, have similar channel StPJCNreS.
the peptide regions suggesting that both
that are proteins
VDAC, porin and perforin
channels
D. Levitt Department
of Physiology, Current
Opinion
University in Cell
This review covers the miscellaneous channels that do not fall w=ithin the domain of the other reviews in this section. Three different channel systems will be reviewed: the gramicidin A channel; the large weakly selective channels that are present in the outer membrane of mitochondria (VDACs) and bacteria (porin); and the large channels whose function is to kill cells by colloid osmotic hemolysis (perform).
2:689-694
High resolution X-ray crystallography has shown that when gramicidin is crystallized from bulk solutions, it forms a different ‘channel-like’ structure referred to as the gramicidin ‘pore’, in which the two peptide chains are in an antiparatlel double-stranded helix (Langs, Science 1988, 241:188-191; Wallace and Ravikumar, Science 1988, 241:182-187). Unfortunately, as these ‘pore’ sttuctures cannot be related to a physical channel conductance, their usefulness is limited. It is possible that they represent the low-conductance ‘mini-channels’ that are seen at frequencies varying from 5 to 50% of that of the large channel [ 11. Despite considerable effort, little is known about these mini-channels. It seems clear that they are a different conformation of the gramicidin channel and are not the result of impurities in the preparation. Indirect evidence suggests that these mini-channels re-
The gramicidin A channel is the simplest, best characterized ion channel (for reviews, see articles in Truqbort Gurikrs,
1990,
Channels and Pumps,
edited by Pulknan et al. Kluwer Acudemic Publtiers, 1988). It is formed from two identical 15-amino-acid peptides (Fig. 1). Although there is still no direct (i.e. crystallographic) proof of channel structure, a large num-
I
Cl’*ytotoxic
T lymphocytes;
@
Abbreviations Nh4R-nuclear magnetic resonance; VDAC-voltage-dependent anion channel.
Current
Biology
USA
Nati Acud Sci USA 1971, 68:672476).
A
through Membranes:
Biology
Minneapolis,
ber of different approaches have provided strong evidence that the conducting channel in lipid membranes is a head-to-head left-handed, single-stranded P-helix with 6.3 residues per turn as originally proposed by Urry (Proc
Introduction
Gramicidin
of Minnesota,
Ltd
ISSN
0955+674
Fig. 1. Model of the gramicidin channel. It IS formed from an NHz-terminal to NH*-terminal dimer of 8-helices The rihbon indicates the peptide backbone that lines the channel, while the solid black region corresponds to the hydrophobic side chains. Published by permission (Andersen et al., In Transport through Membranes: Carriers, Channels and Pumps; edited by Pullman et al. 1988, p. 118).
W-pore-forming
protein;
690
Membrane
peryeability
suit from very small changes in the dominant conducting channel, and thus are probably not the double-stranded ‘pore’ structure described earlier (Busath and Szabo, Bio plys J 1988, 53689-695). The fmding of Sawyer et al [l] that low levels of detergent increase mini-channel frequency is the lirst success in being able partially to control their number. The gramicidin channel may not be a good model for biological ion channels for several reasons. It is a uniform cylinder, about 0.42 nm in diameter and 2.5 nm long, that is just large enough to hold a row of about nine water molecules. The channel is so narrow that the ion and water cannot get past each other and so the movement of an ion through the channel must displace all the water ions. This is one example of the ‘single-file’ effect which dominates the transport kinetics in this channel (Levitt, Cuw Tqo Memb Tran 1984,21:181-197). In contrast, many biological channels have relatively wide mouths, tapering to short narrow, selectivity regions (Dani, Curr Opin Cell Biol 1989, 1:753-764). These structural differences can produce qualitative differences in the transport kinetics. For example, the movement of an ion within the gramicidin channel may be limited by the water-wall interaction of the nine water molecules that it must displace (Dani and Levitt, Bic@ys J 1981, 35:501-508) while ion kinetics in biological channels may be limited by ion-wall interactions, especially in the narrow selective region. In addition, the gramicidin channel is uncharged and its cation selectivity presumably arises from interactions with the carbonyl oxygens that form the pore wall. This interaction may be relatively unimportant in many biological channels where ion selectivity is probably dominated by fixed charges. Whatever the biological relevance of the channel, the combination of its extreme simplicity and the knowledge of its (probable) structure means that, for now and for the foreseeable future, it is the only channel in which detailed theoretical models can be tested. Gramicidin can be regarded as the proving ground for biophysical theories of channel conductance. One factor which must be included in any physical model of the channel is the electrostatic potential that arises from ion-ion interaction and from the low dielectric of the channel wall and lipid. Simple continuum theories provide estimates of these potential energy barriers that are in reasonable agreement with the experimental data (Jordan, J Pbys C5em 1987, 91:65826591). In the past, these continuum calculations have either neglected the presence of ions other than the transport ion or have included them in a very approximate manner (Levitt, Bio p&s J 1985, 48:19-31). Recently, Jordan et al [2] have performed a much more accurate and complete treatment of the influence of the other ions. As one might expect, these ions tend to screen each other and lower the potential. Because the narrow gramicidin channel only holds one or two ions, these effects are small within the channel but can become significant outside the channel. The major limitation of these calculations is that the system must be close to equilibrium. Thus, they are not applicable to the important case of large net fluxes.
The biophysics of gramicidin is simple enough that, in theory, it can be solved from first principals, i.e. the inter- and intramolecular forces of the channel, membrane lipid, water and ion are simulated and the classic movement of the ion in these forces is solved exactly. This is the molecular dynamic approach that has been applied to the gramicidin channel by several groups (Polymeropoulos and Brickmann, Ann Reu Biopbys Cbm 1985, 14:315-330). The use of molecular dynamics has now become a standard tool for studying protein movements and structures and commercial software packages are now available. Chiu et al [3] have applied the GROMOS package (Hermans et al, Biopo&mers 1984, 23:23-84) to the gramicidin channel. They investigated the ion-free channel and concentrated on the movements of the carbonyl oxygens and the interaction and structure of the channel water molecules. The carbonyl oxygens flex and rotate slightly out into the channel where they form hydrogen bonds with the water. The most dramatic result is that all the channel water molecules seem to align themselves with each other, with all their dipoles pointing in the same direction. The reorientation of the dipoles in the opposite direction is a very rare event occurring only once in the 70~s simulation. This is surprising as one would expect one-dimensional water-water interactions to be relatively short range (Landau and hfshitz, Statkticul Pbyszb, Pergamon Press, 1958, p. 482) and water-wall interactions to dominate. Mackey et al. (Bioply J 1984,46:229-248) saw essentially the same effect as Chiu et aL. [3], while Lee and Jordan (Biopbys J 19,46:80%319), using a very different model of the water and channel did not observe this correlation. These results illustrate the major limitation of the molecular dynamic approach: most of the theoretical predictions cannot be experimentally tested. Calculating the channel conductance is a particularly difficult problem. Accurate estimates of ion conductance require that the ion channel potential profile be known to about 1 kT while the molecular dynamic calculations require knowledge of intermolecular potentials that are several orders of magnitude larger. Potentials that have been found to be adequate for calculations of protein structure may lead to large errors in the estimates of ion channel conductance. One approach which could offer some direct feedback on the molecular dynamic results is to combine them with nuclear magnetic resonance (NMR) measurements (Cl-h et al, Biopky J 1990, 57:lOla). The limitations of the molecular dyamic approach are illustrated in the evaluation of equilibrium cation potentials in the channel. Because cations have very long-range forces, they pose special problems. Previous estimates of the equilibrium potential have differed markedly from experimental estimates. &vist and Warshel [4] have developed a new approach to this problem that yields the first reasonable estimate of these potentials. They use a highly simplified model of the dielectric properties of the water and lipid that allows them to take account of the longrange forces.
Cramicidin,
Another factor limiting the usefulness of molecular dynamic calculations is the considerable uncertainty about the relevant experimental quantities needed to evaluate them. For example, from molecular dynamics one can calculate the equilibrium potential profile in the channel. Estimating this equilibrium potential from the experimental channel conductance requires an assumption of a detailed kinetic model, and different models can lead to markedly different estimates of the potential profile (using the same conductance data). Recently, direct measurements of equilibrium binding using NMR techniques have been made [ 51 and the future extension of this approach should provide better experimental data.
VDAC
and porin
These two classes of channels, the voltage-dependent anion channel (VDAC) and porin are discussed together because they have similar general function and structure. Both are large-conductance large-diameter ion channels with thin walls formed by a P-sheet structure and located in the outer membrane of either mitochondria (VDAC) or bacteria (porin). Because of these similarities, VDAC has been referred to as ‘mitochondrial porin’. However, this nomenclature is confusing because it obscures the fact that these two classes have no ammo acid homology and that they are clearly different proteins with major differences in structure and function. VDAC is probably related to some mitochondrial inner membrane transport proteins (Mannella and Auger, BzbpbysJ 1986, 49:272a). VDAC forms a large (about 3nm in diameter) slightly anion-selective channel with complex voltage dependence (reviewed by Colombini [6], see also the mini-review series in J Bioenerg Biomemb 1989, 21:417-507). Compared with other channel proteins, it is relatively
VDAC, porin
and perforin
channels
Levitt
small: the yeast protein consists of 283 ammo acid residues (molecular weight 29883). A low-resolution structure of the channel has been determined by electron microscopic imaging on two-dimensional crystals [7]. The channels have an outer diameter of approximately 5 nm and an inner diameter of between 2.5 and 3 nm. What differentiates these channels from the structure of most other ion channels is that they have very thin walls, about the thickness of one ammo acid chain. The amino acid sequence has a pattern of alternating polar and non-polar side chains that suggests that it forms a P-sheet structure in which the polar residues point into the channel and the non-polar residues contact the lipid. This suggestion has been directly confirmed by sitedirected mutagenesis on the yeast protein (81. Colombini and his colleagues [8] found that changing the charge at some locations produced little or no change in conductance, but marked changes in ion selectivity. Assuming that a residue in which a charge change alters selectivity must line the channel wall, the authors screened the entire sequence to find the segments that contributed to the wall structure. (This structural assignment may be in slight error because charges at the pore mouth should also influence ion selectivity [lo] .> The results suggest that each VDAC molecule consists of 12 transmembrane ~-strands and a transmembrane u-helix. A model that summarizes the structural and functional properties of the channel is shown in Fig. 2. The channel is shown as a dimer of VDAC, although there is no direct evidence for this. The voltage-dependent closure results from the transfer of a portion of the protein from the channel wall to the membrane surface, decreasing the membrane diameter and changing its charge and ion selectivity. The structure of bacterial porin differs significantly from that of VDAC. The low-resolution (about 23A) structure of one type of pot-in (OmpF) has been determined by
p sheet
I
u helix
open
closed
Fig. 2. Proposed model of the open and closed form of the voltagedependent anion channel NDAC) channel. The channel is a homodimer with monomers oriented in opposite directions, each monomer consisting of a psheet (thin-wall region) and one a-helix. When the channel closes, an a-helix and part of the p-sheet region move from the channel wall to the membrane surface. (Kindly provided by M. Colombini.)
691
692
Membrane
permeability
electron microscopic imaging techniques on negatively stained two-dimensional crystals (Engel @ al, Nature 1985, 643-645). It has a remarkable structure, consisting of a trimer in which three channels at the outer surface converge and merge into a single opening at the inner surface. This unusual structure leads to correlated single channel openings and closings. Using similar technology to Engel et al, (1985), Jap [9] has determined the structure of another porin (PhoE) to a resolution of about 18A A model that summarizes the data is shown in Fig. 3. It is also a trimer with three openings on the outer surface that converge but, unlike the OmpF protein, do not merge at the inner surface. Although the resolution of the images is not high enough to determine the detailed structure of the channel walls, they are shown as P-sheet structures in Fig. 3, as suggested by a large number of other physical measurements.
ment proteins (C5, C6, C7, C8 and C9) aggregate to form the pore, referred to as the ‘membrane attack complex’ (Mtiller-Eberhard, Ann Ret) Immunoll986, 4:503-528). Because of the complexity of this process, little is known about the structure of this channel (Stanley, Curr Top Micrab Immunoll988,140:49-65). Not even the stoichiometry of the complex is agreed upon. This is one reason for the excitement greeting the discovery that cytotoxic T lymphocytes (CTL) use a protein that is homologous to complement C59 to kill cells (reviewed in [ 161. The purified protein, referred to as perforin, cytolysin or lym phocyte PFP, can produce channels on cells, vesicles and black lipid membranes, indicating that the channel is a homopolymer and should be more amenable to structural studies than the membrane attack complex.
The porin PhoE is anion-selective, while the porins OmpF and OmpC are cation-selective. By investigating fusion products of the proteins (OmpC-PhoE and OmpF-PhoE), Benz et al. [lo] were able to determine the amino acid residues responsible for the selectivity Most of the selectivity was accounted for by a short 74amino-acid sequence at the N-terminal end. The anionselective PhoE was converted into the cation-selective OmpC by substituting two negative aspartic acids for a positive lysine and neutral serine. It is thought, based on evidence from other studies, that these amino acids are positioned at the pore mouth. One can expect that the continued application of molecular biological techniques to this system should soon provide a detailed map of the amino acids that influence the channel selectivity.
Extracellular
Periplasmic
Complement
and perforin
There are a large number of channels whose function is to produce cell membrane holes that lead to cell death by colloid osmotic hemolysis (reviewed by Bhakdi and Tranum-Jensen, Ret1 Pbysiol Biocbem Pharmucol 1987, 107:147-233). The definitive test for this mechanism of cell death is that it can be prevented by adding a large impermeant molecule to the bathing solution that balances the osmotic pressure of those cellular components too large to pass through the hole. In some cases this test fails because the holes are so large that even highmolecular-weight dextrans can pass. This channel class, referred to collectively as pore-forming proteins (PFPs), includes such disparate compounds as staphylococcal atoxin [ 111, streptolysin (which is also related to the toxin from the mushroom amanita), toxins from Gram-negative bacteria [12] and bee venom (melittin) [ 131 and toxins made by amoeba [ 14,151 and possibly responsible for amoebic dysentery. Because of their clinical importance there is most interest in the PFPs of the immune system. It has been known for many years that PFPs are the primary mechanism of complement-mediated cell death. Five of the comple-
VW
Fig. 3. A schematic model of the PhoE channels. The channel walls are shown as P-sheet structures. On the extracellular side, there is large vestibule, about 3.5 nm long that has an elliptical opening with major and minor axis of 1.8 and l.Onm. At the periplasmic side, the channel narrows to an inner diameter of 1 nm with a length of 1 nm. The outer leaflet of the membrane is lipopolysaccharides and the inner leaflet is phospholipids. Published by permission 191.
The pores formed by perforin are very large, with a diameter of 915 M-I. At high concentrations, ring-like ‘tubular’ membrane lesions with diameters of about I6 nm can be observed under the electron microscope. These are not observed at lower concentrations when smaller, ‘nontubular’ channels are probably formed [ 161. This is supported by the fact that, in bilayer studies, a heterogeneous population of single-channel conductances is observed (Young et al, J Exp Med 1986, 164:144-155). The structural characterization of the channel in terms of the amino acid sequence is only at a very prelim inary stage. The perforin sequence is predominantly hydrophilic with a short stretch of about 60 hydrophobic amino acids that could form two membrane-spanning a-helices [ 171. There are two other amphiphilic regions that might form an a-helix or P-sheet structure. The per-
Gramicidin,
VDAC, porin
and perforin
channels
Levitt
C8a
Fig. 4.
A simplified diagram of the amino acid sequences of the protein complement components that form the chan-
Perforin
N
Cal+
piij
nel complex K6,C7,C8,C9) and perforin. The heavy solid line indicates the region with high homology. ‘***’ indicates the
C
region bound
535
of C9 which is labeled by lipidprobes. Published by permission
[18l.
forin channel may not follow the usual rules in which the channel wall is formed primarily from hydrophobic amino acids. The wall might consist of a single thickness of either a P-sheet (e.g. porin, VDAC) or a-helix, in which case one would expect to find amphiphilic amino acids, with the hydrophilic groups in contact with the channel water and the hydrophobic groups in contact with the lipid. One approach to interpreting the amino acid sequence is to look for homologies with the complement proteins. With the sequencing of C6, all the components of the ‘membrane attack complex’ have now been sequenced [18,19]. Fig. 4 shows a simplified diagram of the homology between G-9 and perforin. They all share a contiguous region of about 370 amino acids. If one assumes that perforin and complement have similar membranespanning structures while differing in the other domains, then this similarity identifies the transmembrane regions of the sequence. The complement protein C9 has another, more surprising, homology with melittin, the 26. amino-acid protein that is the main component of honey bee venom (there are two sequences of four and six residues that are homologous). Furthermore, antibodies against these sequences crossreact between melittin and C9 [20]. Because of the small size of melittin, one might expect that it would be easier, at least compared with complement or perforin, to determine the structure of this channel. However, at present, little is known about melittin and it is not even certain that the melittin-induced cell lysis is produced by f&d channels [ 131. The discov-
ery of the sequence similarity should stimulate renewed interest in this question.
Conclusions The study of these ‘other’ channels is a useful reminder of the diversity of channel structure. The structures of gramicidin (P-helix), and porin and VDAC (P-sheet) are well established and differ from the hydrophobic a-helical structure that is the conventional channel building block. In addition, the ion selectivity of gramicidin arises from neutral amino acids while that of porin and VDAC is produced by charged amino acids. One can also expect that other aspects of channel function, such as the importance of diffusion limitation, ion-ion interaction and ion-water interaction will vary from channel to channel.
Annotated
references
and recommended
reading 0 *a
Of interest Of ountanding
1.
SWAYER DB.
0
tance
interest KOEPPE RE. ANMRSEN
heterogeneity
28:65X 4,583. The number of low conductance the addition of low concentrations
in gramicidin
OS: Induction of conducchannels. BibcAm 1989,
‘mini-channels’ of detergent.
can be increased b> It is suggested that this
693
694
Membrane
permeability
is the mectkism among results
responsible from d&rent
for the variation laboratories.
in ‘mini’ frequency
found
2.
JORDAN PC, BACQUET RJ. MCC~ION JA, T~LW P: How electrolyte shielding influences the electrical potential in transmembrane ion channels. Bio&a/ 1989, 55:1041-1052. The first rigorous calculation of the importance of electrolyte shielding on channel conductance. The conclusions are limited to the near-equilibrium regime. l e
3.
CHtll
l
and polypeptide
S. JAKOB~~ON E. MCCAM~~ON conformations in the gramicidin
S, SlIBRAMANVUI
JA: Water
channel.
Bisbhys J 1989, 56:253261. . Application of the GROMOS molecular dynamic package to gramicidin. It is limited to the ion-free channel and focuses on the structure of water in the channel and the channel-water interaction. I(QVIST J, WARSHEL AZ Energetics of ion permeation through membrane channels. Bi@vs J 1989, 56:171-182. The first direct calculation of the eqiilibrium potential energy of an ion in the gramicidin channel that yields a result consistent with the experimental data. Long-range electrostatic forces are incorporated using a very simple approximation for the aster and lipid. 4.
l e
5.
HI~V~ON JF, JERNANDU.
l
namic parameters for the binding of the divalent cations to gramicidin A incorporated into a lipid environment by Tl205 nuclear magnetic resonance. Bi~s.11989. 55:27-330.
JQ,
SHIJNG~
C. MILLET
Measurement of the entropy and enthalpy lent and divalent ions to gramicidin channels vesicles is summarized. 6.
. Review 7. l e
FS: Thermody-
of the binding of monova in Iysophosphatidyicholine
This is the highest resolution obtained for any ion channel. nels have thin walls and internal diameters of about 3 nm. BLACHl.Y-DI’SOii
changes structural
memof 2D
The
than-
M. FORTE F: Selectivity ion channel: Science 1990, 247:123.+1236. Each of the positive bsine residues was .separJteh converted to a nega tive residue and the influence on the channel ion xkctiviv wz meaured. A model of the channel was formed based on the assumption that those sites that produced a change m ion selecti\iv must line the channel wdll. 9.
E. PENI) S. cOmhlHINI
in site-directed implications.
JAP BK: Molecular
mutants
design
of the VDAC
of PhoE
pot-in
l
BENZ R. %CtihlII) A. VAN DER LIZI, P. TOhlhbGSEI‘: J: Molecular basis of porin selectivity: membrane experiments with
OmpC-PhoE
and OmpF-PhoE
hybrid
proteins of Escherichia 1989, 981:%1-1.
coli K-12. Bixhim Ric&s Am Most of the ion selectivi~ was as.sociated with a short sequence at the N-terminal end. The results suggest that the fixed charges that produce the ion selectivity are krJkd at the channel mouth. 11.
FORTI 5. IM~~~~~~~
A: Staphylococcal a-toxin of liquid vesicles by cholesterol
l
permeability
increases and pH
ROPELE M. MENEST~UNA G: Electrical properties and molecular architecture of the channel formed by Escherichia coli hemolysin in planar Lipid membranes. BiodMm BiOpqs AC&~
1989. 985+18. The channel has a conductance of about 300 pS in 0.1 M KCI. The cation selectivity is reduced by acid pH, suggesting that tixed negative charges are invohred. KATSLI T. KUROKO M, MORIKAWA T. SANCHIKA K, FLIJ~TA Y, YMLIKA H, UDA M: Mechanism of membrane damage induced by the amphipathic peptides gramicidins and melittin. Biocbim Bicpkys Aclu 1989. 983:135141. Two different mechanisms have been proposed for the melittin-induced permeability: discrete channels and disordering of the lipid. This paper proposes a third mechanism invohing membrane fragmentation. 13. l
14. l
YOUNG JD, ILXVREY DM: Biochemical terization of a membrane-associated from the pathogenic ameboflagellate
and functional characpore-forming protein Nuegletia fowled J
Biol Uxnr 1989, 264:1077-1083. The pore has a single channel conductance varying from 150400 pS in 0.1 M NaCl and a channel diameter of about 5 nm. It is immunologically distinct from perforin and the PFP from Enf~~rnoeba bisto!)dca [ 151, 15.
K!ZUER F, HANKE W, TR~SSL D, BAKKER-GRLINWAID T: Pore-forming protein from Entamoeba histolytica forms voltage and pH controlled multi-state channels with properties similar to those of the barrel-state aggregates. Biccbem BiqLys
Acfu 1989, At low voltages ( < about 100% pS is ductance channels multimer.
982:89-93. 50 mV), a dominant single channel conductance of seen. At higher voltages a multitude of lower conis seen, suggesting a break up of the one dominant
JD: Killing of target cells by lymphocytes: A mechview. PLyiol Ret1 1989, 691250-314. An excellent review of an area in which there has been an exponential growth in the number of publications in recent years. 16.
YOUNG
l
artistic
1’.
Kwo~: BS, WAK~ICHIR M, Ltu C. PERSECHINI PM, TRAPANI JA, &Q AK, KIM Y. YOUNG JD: The structure of the mouse lymphocyte pore-forming protein perforin. Biocbenz BiqLys Res Comm 1989s 58:1-10. The authors present the sequence of mouse perforin and discuss the possible implications of the sequence for the channel structure. l
18.
and its functional
l e consequences. .I ,llol Rio/ 1989. 205:-107+19. Three-dimensional structure of the channel at a rewlution of about IS& determined hy negative stain imaging of two-dimensional crystals. This low-resolution structure provides the tirst direct images of this channel type. The channel is a trimer in which the three channels converge but do not merge in going from the outer to inner surface of the membrane.
10.
l
channel,
htANNEUA CA: Structure of the mitochondrial outer brane channel derived from electron microscopy crystals. J Biomq Biomrmb 1989. 213.127437.
l e
12.
l
COLOMBIM M: Voltage gating in the mitochondrial VDAC J ,+k& Rio/ 1989. 111:10~111. of the VDAC channel.
8.
pendent assembly of oligomeric channels. E~tr J Biocbm 1989, 181~767-773. The formation of the channel in vesicles is a relatively slow process involving assembly of G-10 monomers. Assembly requires conformational changes which are influenced hy the pH.
the de-
DISCIPIO RG. Hucu TE: The molecular architecture of human complement component C6. J Biol Gem 1989, 264:16197-16206. The sequence of complement C6 is presented. This complete the sequencing of all the factors that form the channel. The relation of this factor to the other complement factors and perforin is discussed. l e
19. l e
0iAtiKAvAR-n HJ: Structural
other
channel
DN. Cn%aAvAan homology of
forming
B, PARRA CA, MLIUR-EBERHARII complement protein C6 with
protein
of complement.
Acud Sci USA 1989, 86:279$L2803. A paper published simultaneousiy with [ 18) with discussion. 70. l
sirmlar
Proc
Null
results
and
LUNE RO. ESS~R AF: Identitication of the discontinuous epitape in human complement protein C9 recognized by antimelittin antibodies. J fnmunol 1989, 143:553557.
A demonstration that antibodies against homologous in C9 and m&tin cro.ssreact, have similar channel StPJCNreS.
the peptide regions suggesting that both
that are proteins
