Abstract
Aerolysin-like pore-forming proteins are an important family of proteins able to efficiently damage membranes of target cells by forming transmembrane pores. They are characterized by a unique domain organization and mechanism of action that involves extensive conformational rearrangements. Although structures of soluble forms of many different members of this family are well understood, the structures of pores and their mechanism of assembly have been described only recently. The pores are characterized by well-defined β-barrels, which are devoid of any vestibular regions commonly found in other protein pores. Many members of this family are bacterial toxins; therefore, structural details of their transmembrane pores, as well as the mechanism of pore formation, are an important base for future drug design. Stability of pores and other properties, such as specificity for some cell surface molecules, make this family of proteins a useful set of molecular tools for molecular recognition and sensing in cell biology.
This article is part of the themed issue ‘Membrane pores: from structure and assembly, to medicine and technology’.
1. Introduction
Cellular membranes are a vital part of all cells, and numerous mechanisms have evolved in order to destroy their integrity for purposes of defence, attack, hunting or digestion. One of the ancient mechanisms to affect membrane integrity is to punch holes by using pore-forming proteins (PFPs). Many of these proteins act as toxins (i.e. pore-forming toxins), which actually represent one of the largest known group of natural toxins to date [1,2]. The main principle of their action is that they are produced by their originating organisms as soluble units (i.e. monomers, dimers), which upon recognition of (specific) receptors bind to membrane surfaces, oligomerize on the surface and finally transfer part of their polypeptide chain across the lipid bilayer to form functional transmembrane pores. PFPs can be roughly classified into two major groups, α-PFPs or β-PFPs, which form pores by bundles of α-helices or by transmembrane β-barrels, respectively [1]. Although members of either group of PFPs share a common general mode of pore formation, several evolutionarily unrelated families can be distinguished according to the structures of their soluble monomers.
Three main families of β-PFPs are the α-hemolysin family found predominately in Staphylococcus aureus [3], the MACPF/CDC protein superfamily [4] and PFPs exhibiting similarity to aerolysin from the bacterium Aeromonas hydrophila [5]. In this review, we will focus on aerolysin-like β-PFPs (aβ-PFPs). The aβ-PFP family is named after the founding member, aerolysin, a well-studied toxin from the pathogenic bacterium A. hydrophila. Although the structure of soluble aerolysin has been known for more than 20 years [6], there have been few details available on the architecture of its transmembrane pore and its assembly, for this or any other member of this family until recently. New structures of pores of lysenin [7,8], an aβ-PFP from earthworm, and aerolysin [9] provide the first insights into this process and finally unravel molecular details of structural rearrangements connected with oligomerization and pore insertion.
2. Aerolysin-like β-pore-forming proteins
Aerolysin is the most studied member of this family and was also the first member to be associated with cytotoxicity. It was discovered in the Gram-negative bacterium A. hydrophila in 1975 [10]. Since then, many other members have been found in all kingdoms of life, from bacteria, archaea, fungi, animals and plants. Approximately 90% of the identified proteins were found in Proteobacteria, Firmicutes and Fungi [5]. Some well-known examples of bacterial aβ-PFPs are the septicum α-toxin from the highly virulent pathogen Clostridium septicum, ε-toxin and enterotoxin from Clostridium perfringens, insecticidal members produced by Bacillus sphaericus and Bacillus thuringiensis and others [11–14]. Bacterial aβ-PFPs are destined to kill cells of host organisms or have roles in interspecies relations [11,15]. Eukaryotic members of aβ-PFPs serve in defence against pathogens or parasites, such as enterolobin from plants [16] or lysenin from earthworm [17], or assist in prey digestion, such as hydralysins from hydra [18].
In addition to their important role in bacterial pathogenesis, aβ-PFPs have attracted a lot of attention recently in nanobiotechnological applications and as tools in cell biology. Pores formed by aβ-PFPs are extremely stable and have many useful features that may be exploited in sensing applications, in particular DNA sequencing [8,19]. Some aβ-PFPs are able to recognize specific molecules from the cellular surface. For example, lysenin can bind specifically to sphingomyelin [20,21], a predominant lipid from the outer leaflet of animal plasma membranes, and can be used as a specific molecular tool for sphingomyelin detection [17,22,23].
3. Structural features of aβ-pore-forming proteins
The crystal structure of proaerolysin solved in 1994 by Parker et al. [6] signalled the advent of structural information on aβ-PFPs. Since then, several structures of soluble monomeric aerolysin-type toxins have been determined. The size of proteins containing an aerolysin type of fold, rich in β-structure, is relatively small, between 30 and 60 kDa (figure 1). They fold into elongated structures, where one part of the molecule includes one or more receptor-binding domains (RBDs) and the other part builds the pore-forming module (PFM). In most of the known structures, RBDs are positioned at the N-terminus of the molecule and the PFM at the C-terminus (e.g. aerolysin, ε-toxin, haemolytic lectin LSL from Laetiporus sulphureus and parasporin-2 from B. thuringiensis), while the opposite architecture with RBDs at the C-terminus can be found in lysenin and C. perfringens enterotoxin. An exception is monalysin from Pseudomonas entomophila, whose single domain structure is basically globular and lacking the RBD [24]. Members of the aβ-PFPs are not limited just to a single PFM in their structure and can also contain various additional domains [5]. Many aβ-PFPs are synthesized in proforms and are proteolytically activated to form membrane pores. A propeptide can be attached either at the N-terminus (e.g. monalysin), or at the C-terminus (aerolysin), or at both termini (parasporin-2). However, the presence of the propeptide is not mandatory, like in Dln1 from zebrafish or lysenin. The aβ-PFP family is quite divergent in terms of amino acid sequence similarity, which is thus not the primary classification key for this family. Namely, the main common feature of aβ-PFPs is conserved structural elements of the PFM [5,25] (figure 1). On the other hand, RBDs which are important for receptor recognition and binding, as expected, exhibit limited structural similarity between different members of the family. A sequence variability and fusion events with other domains suggest that PFM, which is essential for pore formation, is widely used and may serve diverse functions in various biological contexts.
Structures of some of aβ-PFPs representatives. (a) Ribbon representation of some of the most studied representatives of aβ-PFPs, and their PDB-IDs are in the brackets below their names. RBDs are shown in grey and PFMs as: blue, conserved core; red, a variable loop; orange, a weakly conserved β-strand; cyan, an insertion loop. (b) Topological diagram of the PFM of three representatives of aβ-PFPs, colouring the same as in (a). β-Strands are numbered and coloured according to the consensus from Szczesny et al. [5].
Szczesny et al. [5] suggested a model that describes rather well the topology of the PFM domain: it consists of five β-strands, with an insertion loop between strands β2 and β3, and a variable loop between strands β4 and β5 which can range from only a few residues, as in LSL lectin or lysenin, to multiple secondary structure elements as found in aerolysin (figure 1b). The central insertion loop, also named a prestem loop or tongue, is amphipathic, and it has been long hypothesized that this loop spans the lipid bilayer and structurally reorganizes into the transmembrane β-barrel of the pore [13,26]. The structure of the lysenin pore revealed [8] that residues from strands β2 and β3 adjacent to the insertion loop can also be involved in the formation of the transmembrane β-hairpin. Namely, the membrane spanning β-hairpin contains altogether approximately 70 residues in all pore structures of aβ-PFPs; therefore, shorter insertion loops require more additional residues from β2 and β3 strands [7–9] (figure 2a,c,e,g). A special feature of all aβ-PFPs are alternating serine and threonine residues found in the insertion loop, as well as throughout the rest of the PFM. These polar residues are thought to participate in membrane binding [27] and oligomerization [28] and help the amphipathic loops in transmembrane pore formation [29].
Structural features of aβ-PFPs pores. (a–d) Lysenin pore (PDB-ID 5EC5). (a) Ribbon presentation of the lysenin pore, side view and (b) top view. Pore dimensions are shown, as well as membrane position (pink lines). (c) Superposition of the lysenin monomer (PDB-ID 3XZD) and a protomer from a pore, superposition on the C-terminal domain. (d) The electrostatic properties of the inner surface of the model of the wild-type lysenin pore [8]. The surface of the pore is coloured according to electrostatic potential. A cut-off of −5 kT e−1 was used for the negative potential (red) and of +5 kT e−1 for the positive potential (blue). (e–h) Aerolysin pore (PDB-ID 5JZT). (e) Ribbon presentation of the aerolysin pore, side view and (f) top view. Pore dimensions as well as membrane position (lines) are shown. (g) Superposition of the pro-aerolysin monomer (PDB-ID 1PRE) and a protomer from a pore, superposition on helices of domain D2. (h) The electrostatic properties of the inner surface of the model of the aerolysin pore [9]. The surface of the pore is coloured as in (d).
A number of regulatory mechanisms exist that control unnecessary pore formation by aβ-PFPs. Many aβ-PFPs are synthesized in inactive proforms which are later cleaved by proteases into active forms. Activity of the soluble form can be further restricted by the formation of a dimer, which dissociates upon receptor binding followed by the removal of the propeptide as in the case of aerolysin [29]. However, Dln1 also forms dimers, although in a different way than aerolysin, that can protect uncontrolled activity even in the absence of the propeptide [30]. Tightly bound Dln1 dimer, together with pH-dependent activation of this protein, prevents activation of the soluble form at the wrong time [30]. In contrast, lysenin includes neither the propeptide nor dimer to prevent its uncontrolled activation. Lysenin requires regions of high density of sphingomyelin for its tight association with the membrane as a first step in pore formation [21,23,31,32]. Yet another activation switch can be found in the case of monalysin that contains the propeptide at the N-terminus, but no RBD. Instead, this protein forms soluble doughnut-like 18-mers built of two flat nonamers, which probably helps concentrate the protein in the absence of the membrane receptor near the surface of the target membrane. As suggested by Leone et al. [24], upon finding the target membrane, and cleavage of the propeptide, this doughnut falls apart, resulting in the binding of the released nonameric discs to the membrane followed by the conformational changes leading to the formation of the transmembrane pore.
4. Structures of pores and mechanism of pore formation
aβ-PFPs form pores through stages, including membrane binding of soluble inactive forms, concentration at the membrane surface in a form of structured aggregates called prepores, followed by dramatic conformational changes that lead to the formation of the transducing channels built of β-barrels. aβ-PFP pores are relatively small, containing six to nine protomers, and have a diameter in the range 1–4 nm, depending on the protein involved [8,9,33]. While atomic structures of the monomeric aβ-PFPs have been emerging at a high rate (figure 1), the first atomic resolution structure of the pore of this family, i.e. the nonameric lysenin, was solved only recently [7,8]. A low-resolution cryo-EM structure has been reported also for Dln1 from zebrafish [30]. While low-resolution cryo-EM structures of aerolysin were reported in 1994 [6], the near atomic resolution structure of the prepore and pore were also determined only recently [9] (figure 2).
aβ-PFP pores exhibit a mushroom-like structure with a central stem built of a β-barrel from the top to the bottom of the pore (figure 2a,e). Such architecture is different from the rest of the PFPs: namely, it contains no vestibule at the top and no larger constriction but is rather evenly built from the top to the bottom like a tube.
The cap of the pore is built of a collar, which represents tilted β1, β4 and β5 strands of the conserved PFM (figures 1 and 2). β5 of one protomer is lined up with β1 of a neighbouring one. This enables formation of the collar, responsible for the extremely high stability of aβ-PFP pores, that serves as a nucleus for transmembrane β-barrel folding after the oligomer is formed [8]. β1 tends to bind other molecules also in the monomeric forms (sphingomyelin in the case of lysenin [20] and the C-terminal propeptide in the case of aerolysin [9]).
The typical alternating serine and threonine residues are distributed along β-hairpins of the entire barrel in both lysenin and aerolysin pores. Their side chains are mostly oriented into the lumen of the pore in the transmembrane region of the barrel; however, they can also be found in the extramembrane region, pointing into the space between the barrel and the cap. Their side chains probably contribute to the stability of the barrel, overall polarity and shape via their hydrogen bond network. The part of the β-barrel spanning the membrane is approximately 4 nm high and rich in hydrophobic residues, as expected for a membrane protein. The charge distribution in the interior of the barrel nicely explains cation selectivity of lysenin and weak anion selectivity of aerolysin (figure 2d,h) [7–9,34,35].
Formation of aβ-PFP pores is described in steps in figure 3. The prebinding and binding events differ slightly between the members of the family, depending on a membrane receptor requirement, cleavage of the propeptide and dissociation of soluble dimers before binding. However, as seen from the structures of the lysenin and aerolysin pore (figure 2), the RBD seems to have prestructured receptor-binding sites and thus does not significantly change the conformation upon membrane binding, as well as upon pore insertion. Following the binding of monomers to the membrane, membrane-attached oligomers are formed. In the case of lysenin, gradual formation of pores was observed via formation of arcs that grew with time into nonameric prepores [8,31]. However, such intermediates have not yet been reported for other members of the aβ-PFPs. The presence of conformational changes in individual protomers upon prepore formation seems to depend on the size of the oligomer. While conformational changes take place during prepore formation in the case of heptameric aerolysin, they seem to be absent in the case of nonameric prepores of lysenin and probably also in the case of nonameric monalysin and octameric Dln1. However, the insertion step leading to pore formation seems to be conserved between aβ-PFP members.
Mechanism of pore assembly by aβ-PFPs based on the structure of the lysenin monomer and its pore. Step 1: Soluble proteins (i.e. monomeric, dimeric) bind to their respective receptors (green circles/sticks for lysenin receptor sphingomyelin) on the membrane (pink) with their RBD domain (grey circle). Dimers of aerolysin or Dln1 dissociate upon binding. aβ-PFPs like monalysin do not require any specific receptor but instead form large soluble complexes in order to increase their concentration at the membrane. Step 2: Bound monomers start oligomerizing above the membrane to form ring-shaped prepores. Many aβ-PFPs contain propeptides, which need to be removed to initiate oligomerization. It has been shown for lysenin that the prepore grows via gradual elongation of membrane-bound arcs. The shape of the prepore depends on the subdomain structure of aβ-PFP, being cone-like for aerolysin and wreath-like for lysenin, resulting in heptameric and nonameric prepore, respectively. In the case of monalysin, the nonameric prepore seems to be already included in the doughnut-shaped soluble form that dissociates upon propeptide cleavage, resulting in nonameric disks landing on a membrane as prepores. Transition from the monomer structure to prepore can already require some conformational changes, as in the case of aerolysin, which seems not to be needed in lysenin. Step 3: Finally, large conformational changes in PFM domain enable insertion of approximately half of this domain into the membrane and formation of the pore with β-barrel channel. PFM is built of two parts (cyan for insertion loop and orange/blue for the twisted β-sheet that does not change its intrinsic secondary structure upon pore formation). Colours used are the same as in figure 1. In steps 2 and 3, only three protomers of the oligomer are shown for clarity. On average, the number of protomers in ring-like prepores and pores ranges between six and nine, depending on the protein involved.
Both lysenin and aerolysin have been shown to undergo large conformational changes in the transition between the soluble and transmembrane form, and this is most probably true for all other members of the aβ-PFPs. The structure of each aβ-PFP member can be divided into three parts. One part is the RBD, which is built of two domains at the N-terminus in aerolysin and one, the C-terminal domain, in lysenin. The PFM part can be divided into two parts, as shown in figures 2 and 3. While the RBDs do not change the internal structure of their fold, except for the change of the angle between the two domains in aerolysin, one half of the PFM unfolds during pore formation (figure 3, cyan). This includes disengagement and unfolding of the insertion loop that then refolds into a β-hairpin together with the preformed β-strands β2 and β3 already present in the soluble monomer. The other half of the PFM, a twisted β-sheet including strands β1, β4 and β5, remains structurally relatively intact in comparison to the monomeric protein. However, the angle between the twisted β-sheet and RBD strand changes dramatically upon pore formation, tilting β1, β4, β5 as a unit by 45° relative to the RBD. Consequently, these twisted β-sheets contributed by protomers now build a collar that connects the transmembrane part to the RBD domains in the cap of the pore. In the case of aerolysin, with a two subdomain RBD, this ends up in the formation of a heptamer and in the case of lysenin, the pore contains nine protomers. The presence of additional RBDs and thus stoichiometry may also affect the structure of the prepore, being cone shaped in the case of aerolysin and wreath shaped in the case of lysenin or monalysin. Additionally, this also affects the size of the collapse between the prepore and pore. In addition, as it has been clearly shown that (pre)pores formed by members of the CDC family grow via intermediate structures called arcs [2,36,37], it has been also shown for the aβ-PFP lysenin that construction of its prepore proceeds via arcs as intermediates, i.e. nonameric prepores grow in steps [8,31].
5. Concluding remarks
Evolutionary conservation of aβ-PFP PFMs throughout all kingdoms of life reflects the biological relevance of these proteins. aβ-PFP family members, which are mostly toxins, have important roles for the survival of organisms, being used in attack, defence, and also digestion via pore formation followed by cytolysis of targeted cells. The essential mechanism of pore formation seems to be remarkably conserved between the members of aβ-PFPs, despite the variability in RBD and type of activation. Knowing structures and functions of the aβ-PFPs is extremely important since many of the members of this family are of bacterial origin and in many cases major virulence factors. However, some of these proteins contain insecticidal properties and could be used as tools to prevent plant or animal infections. Further studies on monomeric and pore structures of different members of the aβ-PFPs will have important implications not only for the health of humans, animals and plants, but also for applications in nanobiotechnology and cell biology.
Authors' contributions
M.P., M.K. and G.A. drafted the manuscript, prepared the figures and gave final approval for publication.
Competing interests
We have no competing interests.
Funding
The authors thank the Slovenian Research Agency for financial support (programme grant no. P1-0391).
Acknowledgments
The authors thank the guest editors for including this review paper in the theme issue ‘Membrane pores: from structure and assembly, to medicine and technology'.
Footnotes
One contribution of 17 to a discussion meeting issue ‘Membrane pores: from structure and assembly, to medicine and technology’.
- Accepted October 14, 2016.
- © 2017 The Author(s)
Published by the Royal Society. All rights reserved.
