Gram-negative bacteria and mitochondria are both covered by two distinct biological membranes. These membrane systems have been maintained during the course of evolution from an early evolutionary precursor. Both outer membranes accommodate channels of the porin family, which are designed for the uptake and exchange of metabolites, including ions and small molecules, such as nucleosides or sugars. In bacteria, the structure of the outer membrane porin protein family of β-barrels is generally characterized by an even number of β-strands; usually 14, 16 or 18 strands are observed forming the bacterial porin barrel wall. In contrast, the recent structures of the mitochondrial porin, also known as VDAC (voltage-dependent anion channel), show an uneven number of 19 β-strands, but a similar molecular architecture. Despite the lack of a clear evolutionary link between these protein families, their common principles and differences in assembly, architecture and function are summarized in the present review.
- β-barrel protein
- outer membrane protein
- voltage-dependent anion channel (VDAC)
Membranes are the natural barriers that enable compartmentalization and development of an individual lifestyle in bacteria and the distribution of functions to organelles in eukaryotic cells. Exchange of nutrients over the outer membranes of Gram-negative bacteria and mitochondria is orchestrated by narrow and water-filled protein pores of high abundance, which give these membranes a sieve-like appearance [1,2]. These channels, initially described for the outer membrane of Gram-negative bacteria, were termed porins, given their pore shape and function, and were first characterized by electron microscopy and conductance measurements in planar lipid membranes [3,4]. Subsequently, a channel protein discovered in MOMs (mitochondrial outer membranes), similar in architecture and function, was entitled the mitochondrial porin or VDAC (voltage-dependent anion channel) . VDAC isoforms are a unique group of porin-like molecules in MOMs and typically show a tissue-specific expression in higher eukaryotes . By contrast, in bacteria the ensemble of porins is variable and diverse due to the individual requirements of the bacterium for nutrient uptake [7,8]. The expression profiles of porins are often regulated by environmental conditions, in particular pH changes, starvation or adaption to changing nutrient conditions .
The endosymbiotic theory forms the basis of our current understanding for the process of how mitochondria and Gram-negative bacteria are related by evolution. According to this theory, the mitochondrial organelle originated from a prokaryotic organism and independently developed over a time range of approx. 1.5 billion of years. The overall structure has not changed dramatically and both species have a similar size range of 1–10 μm with some rare exceptions. The protein class of porins forms the major protein constituents of bacterial and MOM. Although several bacterial genes encoding porin molecules are observed in the genome (e.g. for Escherichia coli OmpG, OmpF, PhoE, OmpC and maltoporin, where Omp is outer membrane protein), often one porin variant (e.g. OmpF or maltoporin in E. coli) is strongly abundant. In bacteria the number of porin copies was determined to be up to 100000 per cell wall , whereas in mitochondria the number of VDAC channels was estimated to be in the range 1000–10000, depending on the organelle's origin. For both bacteria and mitochondria, the topography of isolated native membranes has recently been studied by atomic force microscopy, and these images acquired at molecular resolution returned a fascinating picture of the protein in native membranes. The method, applied to selected bacterial and mitochondrial membranes (e.g. Roseobacter denitrificans and tomato), provided a picture of these membranes demonstrating an unexpected density of porin ensembles [11,12]. Whereas mitochondrial membranes are covered up to 30% by the mitochondrial porin, bacterial membranes contain up to 70% or more (e.g. almost 100% of the P100 protein in Thermus thermophilus) of a largely abundant porin representative [11,13–15].
Bacterial porins have been studied in great detail for many years and a first porin reconstruction, achieved using electron microscopy methods, appeared in 1988 and clearly showed the principles of porin architecture . The first high-resolution X-ray structure was published in 1990/1991 from the major outer membrane protein of R. capsulatus. The structure of this trimeric porin showed the archetypical fold of 16 tilted β-strands, all of which are connected by extraplasmic loops and periplasmic turns with the particularly long loop L3 folded inside the barrel (Figures 1A and 1C) [16,17]. Many additional porin structures have been determined since then and, for a long time, there emerged a picture of a high similarity in architecture with smaller variations, mainly in loop topology and surface charges (Figure 1B) [18–21]. From these structures and supplemental functional analysis, two main groups of porins were classified as 16-stranded unspecific and 18-stranded specific subgroups, both of which are trimeric (Figure 1A) [22,23]. This concept has been proved useful for many years and provided a clear structure-based classification into e.g. sugar-specific channels, such as ScrY or LamB of the 18-stranded class, or channels of the 16-stranded class designated for the uptake of small and mostly inorganic molecules (e.g. Omp32, OprP and PhoE) [24–26]. However, during recent years and owing to the discovery of new bacterial cell envelopes and outer membrane protein structures, the group of porins became more diverse and a simple classification today seems more difficult. Some of the ‘new’ porin members express a differing fold and quaternary structure, as shown for OmpG or CymA (e.g. 14-stranded monomeric proteins) [27,28]. In addition, a number of porins previously classified as unspecific were later discovered to bind small, typically negatively charged (organic acids or phosphates) ligands, thereby providing a structural and functional rationale for the facilitated small metabolite uptake in bacteria and their specific lifestyle conditions [24,25]. Among the large group of porin channels even more ‘exotic’ molecules such as the P100 protein from T. thermophilus are observed. Initially this protein was reported to form the S-layer (surface protection layer) of this bacterium . In later studies the P100 protein turned out to form a densely packed structure of porin molecules in these native membranes and consisted of a multi-domain structure built by a N-terminal peptidoglycanbinding domain, a long coiled-coil domain and an unusually large porin domain [13,30].
The mitochondrial porin VDAC has been studied structurally since the early 1970s by two methods: electron microscopy and, more recently, by atomic force microscopy. These techniques applied to protein arrays, mostly isolated from natural membranes, led to the estimation of its pore dimensions as approx. 3 nm in diameter and 4.5 nm in height with a quaternary oligomerization state of six molecules in natural membranes. However, this approach could only be used to approximate the number of β-strands forming the underlying architecture [14,31–34]. On the basis of bioinformatic and antibody epitope studies of VDAC proteins several two-dimensional models have been generated and 12–19 β-strands were predicted with model-dependent localization of the N-terminal helix either as part of or outside the barrel [35–39]. More recently the precise architecture of the first MOM protein, VDAC1 from mammals, was revealed at atomic resolution. Three structures independently conducted by X-ray crystallography, NMR and a combination of both methods, are all identical with respect to their topology and show a monomeric 19-stranded β-barrel with the N-terminal helix enclosed in the barrel pore [40–42].
Functionally, mitochondrial and bacterial porins are considered to operate in a similar way due to their similar pore dimensions, architecture and passive diffusion properties; most of their specific parameters have been determined by identical techniques. For both groups, experiments have been conducted to give a biophysical characterization, and to determine the molecular sieve properties and structural biology [5,43–46]. Sugar-specific porin structures, together with indirect experimentation, led to the estimation of molecular diffusion limits as between 600 and 800 Da (with some exceptions in the bacterial kingdom of up to 6000 Da), depending on the channel under investigation, the method or the substrate applied [22,23,47–49]. The newly discovered VDAC structures now also allow, for the first time, a direct structural comparison between both protein groups.
It has been the topic of speculation for a long time that bacterial porins may have formed the evolutionary precursors for mitochondrial porins [50,51]. However, this analogy has not been shown on either the gene or protein sequence level, although a large number of sequences from both bacteria and mitochondria are now available. On the basis of this finding the present review aims to summarize the current knowledge of porins from both kingdoms of life in order to compare their structural and functional features. General structural principles needed to develop the stable pore architecture along with the natural constraints given by membranes are dissected. Finally, individual adaptive elements, which allow for their specific functional divergence are dissected.
THE β-HAIRPIN MOTIF AS A BUILDING BLOCK AND PRINCIPLES OF BACTERIAL PORIN ARCHITECTURE
The evolution of bacterial OMPs is likely to have occurred by the multimerization of a small gene fragment with a length of a β-hairpin motif as the common repetitive unit [52–54]. This hypothesis has long been raised given the high regularity of the OMP protein structures determined to date; these are in the range of a four- (an autotransporter, which trimerizes to form a 12-stranded β-barrel), eight- and up to 24-stranded β-barrel (all form singular β-barrels) [55–57]. Recently, a β-hairpin signature motif has been identified as the principal evolutionary building block of most OMPs by bioinformatic techniques, and confirms previous speculations on how OMP proteins, including the large porin family evolved .
Porin structures typically contain 14, 16 or 18 β-sheets depending on the definition of the protein family . The structure of the smallest representative, the 14-stranded porin OmpG has been determined by X-ray crystallography [28,59]. Owing to a missing loop L3, present in the archetypical porins, the pore opening, at approx. 1.3 nm, is considerably large. Additional 14-stranded barrel proteins have been investigated, most of which have a specific function not related to the diffusion of hydrophilic molecules. These proteins are instead designed for the exchange of hydrophobic molecules, such as lipids or hydrophobic substances (e.g. toluene) and are typically closed by loop or cork-domain-like structures [60,61].
The majority of porins studied so far belong to the 16- or 18-stranded bacterial porins, are typically of oval shape and have an overall dimension of laterally ~3–3.5 nm and ~ 5 nm in height for the monomer (Figure 1A). All these proteins share similar properties, such as their high abundance in native membranes, an increased thermal stability and the regular content of β-sheet structures (in the range of 60%), as well as their specific conductance profiles in planar membranes . The general motif of structural architecture is the closure of the barrel by pairing of the first and last β-strand in an anti-parallel way (Figure 1E). Furthermore, all strands are connected by eight or nine long loops, facing the extracellular side, with seven or eight small turn-like structures in the periplasmic space, as shown for the trimeric Omp32 protein from Comamonas acidovorans (Figure 1A). The third loop L3 of both specific and unspecific bacterial porins is folded inside the oval barrel, thereby restricting the size to as less than ~ 1×1 nm, depending on the particular porin (Figure 1D).
Another architectural principle, independent of the OMP topology, is the formation of aromatic girdles with tyrosine and phenylalanine residues that are predominantly located at the outer and inner membrane boundaries. Tyrosine residues are more frequent at the extraplasmic sites, whereas phenylalanine residues are located at the periplasmic side of the highly asymmetric membrane (Figure 1D) . These aromatic girdles are also present in α-helical membrane proteins and presumably adjust secondary structure elements at the borders of natural membranes . Residues located between these girdles and facing the hydrophobic lipid environment mainly have high Kyte–Doolittle values (e.g. leucine, valine and isoleucine). Additional girdles, often charged, such as observed in Omp32, exist in bacterial porins and may have some implications in a tighter interaction to lipid head groups or LPS (lipopolysaccharide) molecules (Figure 1E) . At the very C-terminus almost all porins have a phenylalanine residue, which is an important prerequisite for proper import and folding in the outer membrane [63–65].
VDAC: STRUCTURAL ADAPTION TO THE MOM FORCES A DIFFERENT ARCHITECTURE
In mitochondria the composition of two lipid bilayers (forming two of the four mitochondrial compartments) has been preserved since the uptake of a α-proteobacterium into an early eukaryotic precursor cell [66,67]. However, the protein distribution and structure of both membranes varies significantly as mitochondria have adapted to a stable cellular environment, whereas bacteria face a variable and often inhospitable surrounding. Therefore, the mitochondrial outer membrane does not contain a shielding surface (comprising a LPS layer) and instead forms the platform for an exchange of metabolites prepared in the mitochondrial matrix space and transported by passive diffusion. Structurally, the mitochondrial porins are not divergent and all are 19-stranded β-barrels, as illustrated by more recent three-dimensional structures, secondary structure prediction methods and the high sequence similarity between each other (Figure 2A) . Functional divergence has only been created by gene duplication and variation to give closely related VDAC isoforms with divergent functions [41,68]. The number of VDAC isoforms varies considerably between independent species. In fungi, such as Neurospora crassa or Saccharomyces cerevisiae, one or two isoforms are observed respectively. In mammals there are typically three different isoforms, whereas in plants up to five isoforms have been identified [51,69].
The β-barrel pore of mammalian VDAC has an almost ideal roundish shape with all the β-strands following a similar angle of inclination . The first and last strand pairs in a parallel way and, in contrast with bacterial porins, all strands are connected by short loop-like structures of two to seven residues, which are slightly elongated towards the cytosolic space (Figure 2A). Variations with mutations/deletions in the individual loop lengths between VDAC sequences are rare, but short extensions at the very N-terminus between isoforms are known [41,70,71]. One major structural determinant protruding the pore is the N-terminal α-helix, which causes a significantly decreased pore diameter. The pore interior carries an overall positive net charge, which is obviously correlated with the major substrates ADP/ATP being translocated by passive diffusion [72,73]. Very surprisingly, a conserved glutamate residue (Glu73) in the VDAC barrel located in strand β4 of the structure is pointing into the membrane interior (Figure 2A). This hydrophilic ‘finger’, located midway of the two aromatic girdles (Figures 2A and 2B), was reported to influence binding of hexokinase in a negative way when mutated, which implies that hexokinase may interact with VDAC via the outer surface .
An evolutionary link between pro- and eukaryotic outer membrane proteins has been predicted, but was so far has only been confirmed for the integral protein Sam50 (Sam is sorting and assembly machinery), which is homologous with BamA (where Bam is β-barrel assembly machinery, previously known as Omp85 or YaeT; identity/similarity 15%/30% for E. coli compared with S. cerevisiae). Paralogues of VDAC have appeared several times during evolution, which implies the absence of an ancient precursor protein, but does not necessarily imply the absence of a bacterial precursor . The gene structure of the individual mammalian VDAC1 isoforms is rather consistent, but there are differences between the different isoforms [e.g. hVDAC1 (human VDAC1) and hVDAC2]. Most vertebrate VDAC open reading frames contain eight exons and seven introns.
In contrast, bioinformatic algorithms have surprisingly shown that on the protein sequence level a repetitive element can be demonstrated. According to such analyses, the hVDAC1 barrel may have evolved by the pentamerization of fragments of approx. ~ 50 residues in length (Figures 2C–2E). These repeats of ~50 residues have been detected for a number of MOM proteins, including Tom40 (translocase of outer mitochondrial membrane 40) sequences (K. Zeth, unpublished work). These segments when mapped on to a three-dimensional structure are highlighted in Figure 2(E) and comprise an arrangement of 3/4/4/4/4 β-strands with the N-terminal β1-strand slightly extended. Although, the exon structure matches the protein repeats at some level in size and location of elements, the overall consistency (protein compared with gene sequence) is relatively weak (not shown).
PORIN MATURATION: COMMON THEMES AND DIFFERENCES
Mitochondrial proteins originated from the engulfment of a α-proteobacterium into another precursor cell and resulted, aside from a significant overlapping ensemble of proteins, in the presence of only a small remaining mitochondrial genome. Hence, more than 99% of the protein ensemble has to be imported into mitochondria, among which are many ‘old gene products’ with significant relationship to the bacterial precursor, but also new gene products. This feature is particularly visible for proteins with functions that have not existed in the bacterial precursor cell e.g. proteins of the TOM or TIM23 (translocase of inner mitochondrial membrane) complex, which together form the protein import machinery for matrix proteins .
VDAC proteins are imported via the TOM proteins and inserted into the outer membrane by the SAM complex [76,77]. Interestingly, a recent paper describes the import of bacterial OMPs into mitochondria by the same TOM- and SAM-dependent mechanism pointing to a potentially conserved insertion mechanism between the two systems . This discovery corroborates previous observations that the outer membrane protein PorB from Neisseria gonorrhoeae was observed in MOMs after cell infection [78–80]. However, in higher eukaryotic cell lines, it appears that the same PorB protein is predominantly observed in the inner membrane, pointing to the existence of a potentially differing targeting mechanisms .
Bacterial porins are all inserted via BAM, a complex comprising five proteins, four of which belong to the large lipoprotein family in bacteria with BamA (also termed YaeT or Omp85; the Sam50 homologue from bacteria) as the membrane anchor . These findings indicate the conservation of the same assembly mechanisms from bacteria to mitochondria . Notably, whereas the bacterial import machinery relies on the presence of lipoproteins, in particular the BamD (YfiO) protein, mitochondrial sorting and assembly is dependent on the two accessory proteins of the SAM complex; the Sam35 and Sam37 protein .
IS Tom40 AN EVOLVED MITOCHONDRIAL PORIN?
VDAC and Tom40 share a significant sequence similarity, of ~30%, with each other . In agreement with the high conservation between VDAC species, mutations and deletions occur very rarely in loop-like regions. In contrast, the N- and C-terminal portions of the proteins show significant extensions. However, sequence comparisons between pro- and eukaryotic porins or Tom40 do not reveal an extended conserved pattern, other than the analogy detected through the amphipathic pattern distribution . Hence, it is tempting to speculate that both proteins, VDAC and Tom40, originate from a newly evolved precursor protein . This finding was not unexpected given that the TOM complex has no functional homologues in bacteria; protein synthesis and secretion is organized in a ‘vectorially differing’ manner given that protein synthesis takes place in the cytoplasm over the MOM, followed by insertion through the SAM machinery, whereas in bacteria it is carried out within the inner membrane border. Moreover, Tom40 and the TOM complex proteins fulfil functions that did not exist in bacteria and therefore it can be speculated that under a similar functional pressure a VDAC channel co-evolved to allow for substrate transport over the MOM. This would be consistent with the repetitive gene structure of VDAC, which is present, although less pronounced, in Tom40 (Figures 2C and 2D). In bacteria, there is evidence for an ancient β-hairpin peptide, e.g. the dimerization and trimerization of an eight-stranded β-barrel yielded the expected 16- and 24-stranded folded pore . However, this oligomerization event has not been proven experimentally for the smaller peptide repeats derived from mitochondrial proteins of β-barrel topology.
OUTER MEMBRANE PORINS: TWO ARCHITECTURES, ONE MECHANISM?
How similar are the architectures of bacterial and mitochondrial porins? Both classes follow building principles that allow an unusually stable arrangement within the membrane, with an even increased stability for bacterial porins, which is likely to be achieved through oligomerization (Table 1) . This stability for both protein types is largely derived from a similar number of inter-strand hydrogen bonds yielding the common overall β-sheet content of ~60% . The charge distribution in bacterial porins varies more significantly according to the substrates to be translocated by passive diffusion . In mitochondria, this charge distribution is likely to be adapted primarily for the binding of the negatively charged ADP/ATP molecules. Moreover, mitochondrial porins also have developed the potential for interactions to different classes of protein molecules, such as ANT (adenine nucleotide translocator), hexokinase and tBid (truncated BH3-interacting death domain) [83–85]. Translocation of molecules related to energy is one of the major functions, but regulation of apoptosis and more cellular functions are of similar importance. Therefore regulation of the protein, not only by expression, but also through post-translational events (e.g. phosphorylation) can occur, which is absent from bacterial porins [86,87].
The structure superposition of porin members from both classes shows the more obvious varying structural features (Figure 3A). Whereas most bacterial porins have a significantly oval shape, the mitochondrial porin is almost ideally circular, but both have very similar dimensions. The structurally well-conserved part in the superposition is the C-terminal portion, including approx. 12 β-strands, with both terminal β-strands overlapping, e.g. β16 from Omp32 and β19 of mVDAC1 (mouse VDAC1) (Figure 3A). These C-terminal strands encounter the same environment, i.e. they face the hydrophobic membrane environment. Oligomeric bacterial porins, e.g. Omp32, by contrast build trimeric architectures in particular via their N-terminal strands. These strands often adopt a different tilt angle relative to the normal axis of the membrane. It is a striking and obvious feature that the common architecture of porins of both classes overlaps in a part which is not involved in oligomerization (strands ~8–19 in VDAC; Figure 3A). It seems likely that owing to the structure constraints of barrel formation within a native membrane VDAC has adopted a similar folding pattern as for the bacterial porin family with a similar tilting angle of ~40–45 °. These constraints are stronger for barrel folds with a similar number of β-strands (14–18), but deviate for smaller (8–10 stranded) and larger (22-stranded) OMP proteins, which encapsulate smaller domains and are typically used for specific substrate uptake .
Another obvious structure similarity formed by different portions is extended into the central structure elements narrowing the pore (Figure 3B). Both the N-terminal helix of VDAC and the L3 loop of bacterial porins nicely align and may have similar implications for conductance . The structural motif of the aromatic girdle is present in VDACs and in bacterial porins, although it is not as strongly pronounced as in some of the bacterial porins, such as OmpF (Figure 2B) .
ELECTROPHYSIOLOGICAL MECHANISMS: ADAPTION TO FUNCTIONALITY AND ENVIRONMENT
Porins can be subdivided in two different classes according to functionality: (i) general diffusion pores and (ii) substrate-specific porins. The first group, 16-stranded porin molecules, the general diffusion pores, such as OmpF of E. coli , mainly sort according to the molecular mass of the solutes and show a linear relationship between translocation rate and solute concentration gradient. Porins of this class often contain binding sites for charged molecules, including small organic acids or phosphate molecules [90,91]. Some of these channels contain ladder-like structures of charged residues (e.g. arginine and lysine residues in OprP), which span the entire pore to provide the ideal structural basis for ion translocation [89,92]. The second group are the 18-stranded porins, originally described as substrate-specific porins, such as LamB/maltoporin of E. coli, which contains a substrate-binding site inside the channel and exhibits Michaelis–Menten kinetics for the transport of linear sugar molecules via the alignment of aromatic residues located in the pore interior (the ‘greasy slide’) [93,94].
The biophysical characterization of porins has been performed using planar lipid techniques . Three characteristics are typically used to describe the current through pores: (i) the conductance through a single pore at a certain salt concentration; (ii) the ion selectivity towards anions or cations; and (iii) the voltage-dependent behaviour of pore constriction.
Prokaryotes display a huge range of single channel conductance values for a variety of porins [89,92]. As an example, for the nucleoside-specific channel Tsx from E. coli a small single channel conductance of 10 pS in 1 M KCl has been determined . On the other hand, channel conductances of up to 3.2 nS in 1 M KCl for the fully open trimer of the general diffusion pore OmpF have been determined . Outstanding huge single channel conductances are described for the P66 porin of Borrelia (11 nS in 1 M KCl) and P100 from T. thermophilus (20 nS in 1 M KCl) [30,98]. Many of the bacterial porins with a conductance in the low picosiemens range contain specific binding sites for substrates to allow the rapid uptake of certain classes of molecule, such as carbohydrates [99,100], nucleosides  or phosphate . Most of substrate-specific porins belong to the group of 18-stranded specific β-barrels. However, there are some rare exceptions, which demonstrate the heterogeneity of these classes, e.g. Omp32 forms pores with a single channel conductance of 510 pS 1 M KCl, but is a specific 16-stranded β-barrel . Mitochondrial porins mainly form pores with a large open-state single channel conductance of 5 nS (hVDAC1) or 3.5 nS (VDAC2 from bovine spermatozoa) in 1 M KCl. Most of single channel conductance values that have been determined are for the VDAC1 isoform, and although they have been determined for various species they show only small deviations in single channel conductance.
Notably, the different conductance values of all porins are usually consistent with an increased pore diameter of the barrel pore, e.g. is ~1.0 nm for Omp32 , ~1.5 nm for OmpF  and ~ 1.5 nm for VDAC (in the full open state) . The influence of the open state diameter of VDAC porins has also been described in theoretical works by Engelhardt et al.  and Menzel et al. . However, a precise relationship of channel conductance to pore size is not necessarily true, as there are several effects that may influence ion conductance through a channel, such as structural subunits extending into the channel lumen (loop L3 of bacterial porins and the N-terminal helix of VDAC), the action of image force and a more hydrophobic interior or channel friction, which may hinder ion movement [106,107]. The recordings shown in Figure 4 show the typical stepwise increase in conductance after addition of OmpF (Figure 4A), Omp32 (Figure 4B) and hVDAC1 (Figure 4C) to the artificial lipid bilayer membrane.
Selectivity towards ions can be quantified by determination of the permeability ratio between cations over anions (Pc/Pa). Dependent on the asymmetric distribution of charges in the channel and the resulting net charge, the selectivity towards cations or anions can be a distinct characteristic for prokaryotic porins, e.g. the Omp32 porin, which has a large surplus of positively charged residues in the pore, is strongly anion-selective with a Pc/Pa of 0.06 in KCl [103,108]. Other porins show distinct cation-selectivity, such as OmpU from Vibrio cholerae with Pc/Pa of ~14 . In comparison, VDAC is only slightly selective for anions in the high conductance or open state  and shows a small cation-selective behaviour at voltages above 30 mV. VDAC from Paramecium mitochondria was shown to be approx. 7-fold more permeable to chloride than to potassium ions  and VDAC2 has a Pc/Pa of 0.8 in KCl .
Many prokaryotic porins show voltage-dependent closure at higher voltages (approx. 100–200 mV), despite the fact that no voltage-dependent closure has been observed so far in in vivo experiments [111,112]. For OmpF, the voltage-dependent behaviour has been investigated in detail. After reconstitution of one OmpF trimer into a planar lipid bilayer, subconductive closing events of single monomers can be observed at voltages above 100 mV (Figure 4A). This behaviour was similarly observed at positive and negative membrane potentials [97,113]. In comparison, as shown in multichannel experiments, Omp32 channel conductance decreases asymmetrically at voltages higher than +40 mV and lower than −70 mV, with positive voltages triggering a more drastic conductance decrease than negative ones (Figure 4B) . So far, no equivalent states have been determined for the eukaryotic porins. In contrast, the conductance of VDAC exhibits reversible transitions to partially closed states at small transmembrane voltages with a conductance difference of approx. 2.0 nS between open and closed states [36,105]. Characteristically, all eukaryotic VDAC porins are voltage-gated . At membrane potentials of 10 mV closing events for reconstituted hVDAC1 channels are rare, whereas at voltages larger than 20 mV the number of closed channels increases and the membrane current decayed to a smaller value. At voltages of approx. 60 mV approximately 50% of hVDAC1 switched to a closed substrate . These VDAC subconducance states were shown to have a reduced permeability and a reversed selectivity .
In addition to the voltage-dependent channel conductance behaviour, VDAC can easily be gated (reversibly closed) by e.g. decreasing the pH below 5. CD studies indicated that VDAC can undergo major conformational changes involving decreased β-sheet and increased α-helical content. It is probable that the N-terminus forms a mobile α-helix that normally resides in a groove in the lumen wall and gating stimuli favour its displacement, destabilizing the β-barrel [33,114]. Considering the observed gating mechanism, it has been proposed that the cell controls VDAC permeability via membrane potential, which could be derived from a Donnan potential raised by impermeant proteins of opposite charge present on the two sides of the outer membrane . These specific gating properties are characteristic for VDAC and have not been reported to occur in bacterial porins.
SIMILAR STRUCTURE, SIMILAR FUNCTION?
Clear differences are visible in some of the individual characteristics of porins in planar lipid membranes, which are mostly independent of pore size, e.g. the voltage-dependent decrease of VDAC conductance . Moreover, the selectivity towards ions changes during the voltage-induced transformation and led to the definition of the mitochondrial porin as the ‘voltage-dependent’ anion channel. These features are unique to most VDAC proteins and have not been reported for any of the bacterial porins [8,116]; in bacteria many of the voltage-dependent channels show complete closure behaviour at higher voltages of around 100 mV, which have been attributed to the movement of certain loop structures. For VDAC conductance, changes may originate from as yet undetermined movement of secondary elements or side chains. Owing to the smaller pore diameter of bacterial channels, the selectivity expressed can be significantly strong towards an ion group [108,109].
The work of our laboratory is supported by the Deutsche Forschungsgemeinschaft (DFG) [grant number Ze522/2–2]; and by the Max-Planck Society.
Abbreviations: ANT, adenine nucleotide translocator; BAM, β-barrel assembly machinery; LPS, lipopolysaccharide; MOM, mitochondrial outer membrane; OMP, outer membrane protein; SAM, sorting and assembly machinery; tBid, truncated BH3-interacting death domain; TOM, translocase of outer mitochondrial membrane; VDAC, voltage-dependent anion channel; hVDAC1, human VDAC1; mVDAC1, mouse VDAC1
- © The Authors Journal compilation © 2010 Biochemical Society