Membrane binding of zebrafish actinoporin-like protein: AF domains, a novel superfamily of cell membrane binding domains

Abstract

Actinoporins are potent eukaryotic pore-forming toxins specific for sphingomyelin-containing membranes. They are structurally similar to members of the fungal fruit-body lectin family that bind cell-surface exposed Thomsen–Friedenreich antigen. In the present study we found a number of sequences in public databases with similarity to actinoporins. They originate from three animal and two plant phyla and can be classified in three families according to phylogenetic analysis. The sequence similarity is confined to a region from the C-terminal half of the actinoporin molecule and comprises the membrane binding site with a highly conserved P-[WYF]-D pattern. A member of this novel actinoporin-like protein family from zebrafish was cloned and expressed in Escherichia coli. It displays membrane-binding behaviour but does not have permeabilizing activity or sphingomyelin specificity, two properties typical of actinoporins. We propose that the three families of actinoporin-like proteins and the fungal fruit-body lectin family comprise a novel superfamily of membrane binding proteins, tentatively called AF domains (abbreviated from actinoporin-like proteins and fungal fruit-body lectins).

INTRODUCTION

The protein world is classically divided into soluble and membrane proteins, the latter comprising approx. 30% of all proteins. There is, however, another large group that spans both definitions. Peripheral or amphitrophic proteins are soluble proteins that transiently interact with the lipid membrane [1]. Such interactions are fundamental for many biological processes, as temporary and permanent binding to cellular membranes elicits different biological functions. Many examples of protein-membrane interactions have been described at the molecular level. Proteins that are involved in cell signalling events are by far the largest and best-known group of peripheral proteins [2]. Examples include PKC (protein kinase C) conserved domains 1 and 2, PH (pleckstrin homology) domains, PX (Phox) domains, Fab1, YOTB, Vac1 and EEA1 domains (VYFE) [2] etc. Other examples of amphitrophic proteins are enzymes that act on or at the membrane [3], intracellular proteins involved in the transport of lipids between cellular compartments [4], annexins [5] and proteins that participate in antimicrobial defence [6].

Furthermore a large number of protein toxins attach to membrane surfaces during the first step of transmembrane pore formation [7]. Creation of transmembrane pores is a very efficient way of killing cells, as the lipid membrane is the first barrier that needs to be overcome by invading molecules. So-called PFTs (pore-forming toxins) are thus one of the biggest and most important groups of natural toxins [8]. PFTs are extremely heterogeneous in terms of their lipid or protein receptor specificity. Some of them require a particular lipid, for example, actinoporins from sea anemones [9] require SM (sphingomyelin) for optimal activity. Actinoporins are a distinct eukaryotic PFT family of basic proteins with molecular masses of approx. 20 kDa, which fold into a compact β-sandwich flanked on each side by α-helices [10–12]. The most studied representative is EqtII (equinatoxin II) from the sea anemone Actinia equina. The unique molecular mechanism of actinoporin action has been unravelled. In the first step, they attach to SM-containing membranes [12–15], then they transfer an amphipathic N-terminal helix to the lipid/water interface [16,17] and, finally, monomers aggregate on the membrane surface to form a (probably tetrameric) final pore [16,18,19]. The suggested pore is composed of four α-helices, one from each monomer, and is very likely to also include membrane lipids [20,21]. Actinoporins thus belong to the α-helical PFT group [22].

For a long time it was believed that actinoporins represented an isolated protein family exclusive to sea anemones that was dissimilar to any other protein family. This assumption was first challenged by the discovery of a haemolytic toxin from the salivary gland of a marine gastropod with high sequence similarity to actinoporins [23]. The recently reported structures of the fungal lectins XCL (Xerocomus chrysenteron lectin) [24] and ABL (Agaricus bisporus lectin) [25] have revealed that they have remarkable structural similarity to actinoporins despite having less than 15% sequence identity. They both belong to a fungal fruit-body lectin protein family (Pfam code PF07367). Members of this family have insecticidal [26] and antiproliferative properties [27]. Both XCL and ABL selectively bind T-antigen (Thomsen–Friedenreich antigen), a Galβ1-3GalNAc disaccharide present on glycoproteins of some malignant cells [25,28].

In the present study, we have performed extensive searches of public sequence databases and found a large number of actinoporin-like sequences, which arise from three metazoan and two plant phyla. We found that the conserved domain corresponds to the C-terminal part of EqtII including the putative SM-binding site. Actinoporin-like sequences were classified in three families according to phylogenetic analysis. We hypothesize that these sequences are not toxins but intracellular proteins that are able to attach to lipid membranes. To test this hypothesis, we have expressed a representative from zebrafish and found it to bind to lipid membranes in vitro. We propose that the three families of actinoporin-like proteins, together with the fungal fruit-body lectin family, form a new superfamily of cell membrane binding domains, tentatively called AF domains (actinoporin-like proteins/fruit-body fungal lectin domains).

EXPERIMENTAL

Materials

DOPC (1,2-dioleoyl-sn-3-glycerophosphocholine), DOPG [1,2-dioleoyl-sn-3-glycerophospho-rac-(1-glycerol)] and bovine brain SM were from Avanti Polar Lipids (Alabaster, AL, U.S.A.). ANS (8-anilinonaphtalene-1-sulfonic acid) was from Molecular Probes (Eugene, OR, U.S.A.). GdnHCl (guanidine hydrochloride) was from Merck (Darmstadt, Germany). All other chemicals were from Sigma unless stated otherwise.

Bioinformatics

Public databases were searched using BLASTP and TBLASTN algorithms [29]. EqtII sequence was used initially as a probe. When similar sequences were retrieved, databases were also searched with species-specific ALP (actinoporin-like protein) sequences. All available protein and nucleotide databases (as of 5 January 2006) at NCBI were searched, as well as species-specific genomic databases on the Ensembl Genome Browser (http://www.ensembl.org) for Fugu, zebrafish and other vertebrates (human, mouse, Xenopus) and Genoscope (http://www.genoscope.cns.fr) for Tetraodon nigroviridis. ESTs (expressed sequence tags) were assembled by the contig assembly program EST assembler (http://bio.ifom-firc.it/ASSEMBLY/assemble.html) [30]. Overlap length was taken as 30 and overlap percentage identity was 90. Singletons or contigs were translated in all six frames to retrieve the ALP sequence. All sequences obtained in this way have a low E value, typically approx. 10⁻⁸. Signal peptides were predicted by SignalP [31]. Sequences were also analysed for the presence of other protein domains using tools from the InterPro database. Secondary structure content was estimated from the amino acid sequence using PSIPRED [32]. Estimation of secondary structure content from far-UV CD spectra was done on DICHROWEB [33] using CDSSTR [34].

Phylogenetic analysis

The final alignment of 98 protein sequences used for the phylogenetic analysis was obtained in three steps. First, an alignment created by MUSCLE [35] was used to obtain a preliminary neighbour-joining tree. Secondly, a subset of sequences representing all main branches were aligned taking into account known DSSP-defined secondary structures with the aid of PRALINE [36]. Small corrections were made by hand to incorporate the known secondary structure information of the actinoporins [10–12]. Thirdly, the ‘profile’ option of MUSCLE was used to align the rest of the sequences to the secondary structure based alignment. Alternatively, all sequences were aligned using PRALINE, taking into account pre-defined secondary structures without manual interference. This approach introduced slightly less gaps (alignment lengths were 326 and 315 respectively). Both alignments yielded trees with largely identical topologies indicating that the alignment-sensitive positions did not bear a substantial phylogenetic signal. An optimal model of amino acid evolution was selected using Modelgenerator (http://bioinf.nuim.ie/software/modelgenerator). The model preferred by Akaike Information Criterion two and Bayesian Information Criterion was a Whelan–Goldman [37] model with a gamma-distributed rate heterogeneity among sites. The search for optimal trees was conducted under the maximum-likelihood criterion using the hill-climbing algorithm implemented in PHYML [38]. The shape parameter of a discrete (four rates) approximation of the gamma distribution was optimized via maximum likelihood. In the absence of suitable outgroups, midpoint rooting was used to tentatively root the tree.

Cloning and expression

The coding region of the predicted Dr1 (Danio rerio ALP) from zebrafish (Danio rerio) was amplified by PCR from the IMAGE EST clones 7267252 (GenBank CN510193) and 7038734 (GenBank CF998026) using oligonucleotides Dr1+ (5′-CGCGGATCCACTGAGTCTGCCGAGG-3′) and Dr1− (5′-GGAATTCACGCGTTAGCAATCCATCTGAGCC-3′). The PCR product was cleaved with BamHI and MluI, and inserted into a precleaved T7-promoter-based expression vector. A modified version of pET8c vector was used that leaves a His₆-tag with a thrombin recognition sequence at the N-terminal to allow easier purification. The correct construction of the plasmid was verified by nucleotide sequencing. Dr1 was expressed in an Escherichia coli Origami (DE3) pLysS strain (Novagen). Overnight culture (20 ml) was used to inoculate 0.5 l of M9:Luria–Bertani medium with 100 μg/ml ampicillin and 25 μg/ml chloramphenicol. Expression of Dr1 was induced at a D₆₀₀ of approx. 0.8 by the addition of isopropyl β-D-thiogalactoside at a final concentration of 1 mM. Cells were grown for an additional 3 h, centrifuged at 2800 g for 10 min at 4 °C and frozen at −20 °C. From the frozen cells, 7 g were thawed into 17.5 ml of 50 mM sodium phosphate buffer (pH 8.0), 300 mM NaCl, 20 mM 2-mercaptoethanol, 0.5 mg/ml lysozyme, 10 μg/ml DNase, 20 μg/ml RNase, 0.5 mM PMSF and 1 mM benzamidine. They were incubated on ice for 45 min with occasional vigorous shaking. Bacteria were sonicated twice at full power during the incubation for 2 min by using a microtip (MSE 150W ultrasonic disintegrator). Broken cells were centrifuged for 30 min at 4 °C at 11000 g (Sigma). The supernatant was stored at 4 °C and the pellet was washed with 7.5 ml of the above buffer. Following incubation for 20 min on ice, the pellet was subjected to a further 2 min sonication and centrifugation using the same conditions as before. The supernatants were pooled and applied to a 1 ml Ni-NTA (nitrilotriacetate) column (Qiagen, Crawley, U.K.), which had been equilibrated with 50 mM sodium phosphate buffer (pH 8.0), 10 mM imidazole and 20 mM 2-mercaptoethanol. Unbound proteins were eluted with the above buffer and the bound Dr1 was eluted by 50 mM NaH₂PO₄, pH 8.0, 300 mM imidazole and 20 mM 2-mercaptoethanol. The protein was dialysed three times against 2 l of 5 mM Tris/HCl, pH 8.0 at 4 °C. The His₆-tag was cleaved from Dr1 with thrombin according to the supplier's specifications (Novagen). Ion-exchange chromatography on an Åkta™ FPLC™ system (Amersham Biosciences) was used as a final purification step. The sample was applied to a MonoQ column equilibrated with 20 mM Tris/HCl, pH 8.0 and bound Dr1 was eluted from the column by a gradient of the same buffer with 2 M NaCl. Fractions shown to contain Dr1 by SDS/PAGE (12% gels), were merged and stored at −20 °C. Molar absorption coefficients, ϵ^0.1% (for 1 g/l of protein solution), calculated from the sequence at ExPASy Proteomics tools Internet site (http://us.expasy.org/tools/), for Dr1 and His-tagged Dr1 used were 1.442 and 1.347 respectively.

Fluorescence measurements

All fluorescence measurements were performed on a Jasco FP-750 spectrofluorimeter (Jasco Corporation, Japan). The sample compartment was equipped with a Peltier thermostatted single-cell holder. All experiments were performed at 25 °C with constant stirring. Excitation wavelength was fixed at 295 nm to eliminate the contribution of the tyrosines, and the emission spectra were recorded between 310 and 400 nm. Excitation and emission slits were set at 5 nm. Proteins were preincubated in 20 mM DTT (dithiothreitol) before spectra acquisition. Protein concentration in the cuvette was 1 μM in a final volume of 1400 μl. Buffer was 10 mM Hepes, 200 mM NaCl, 20 mM DTT, pH 7.5, and, where indicated, 6 M GdnHCl was included in the buffer. For the iodide quenching experiments the concentration of Dr1 was 1 μM in the above buffer. Spectra were recorded without and in the presence of increasing iodide concentrations. The stock of iodide was composed of 2.5 M KI and 1 mM Na₂S₂O₃. All spectra were corrected with the corresponding spectra of the buffer alone and for the dilution caused by the addition of iodide. No further correction for wavelength-dependent sensitivity was done. The values of the collisional quenching constant were obtained from the Stern–Volmer equation [39]:

where F₀ is the fluorescence intensity in the absence of iodide, F is the fluorescence intensity in the presence of iodide, K_SV is the collisional Stern–Volmer constant and Q is iodide concentration. The fraction of fluorophores accessible to the soluble quencher (f_a) was calculated by using the Stern–Volmer equation modified for multiple emission centres [39]:

where f_a is the fraction of accessible tryptophans and K_eff is the effective quenching constant.

For the ANS binding assay, concentration of Dr1 and ANS in 10 mM Hepes, 200 mM NaCl, 10 mM DTT, pH 7.5 was 15 μM. Measurements were taken between 12 °C and 90 °C at a heating rate of 1 °C/min. Excitation wavelength was 370 nm and emission was measured at 488 nm. Excitation and emission slits were set at 5 nm. The T_m was calculated as the mean value between the temperature at which fluorescence starts to increase and the temperature at which maximum fluorescence signal is reached.

CD spectroscopy

CD spectra were measured by Aviv CD spectrometer (Lakewood, N.J., U.S.A.). Bandwidth was set to 1 nm and 0.5 nm for far-UV and near-UV CD spectra respectively. The signal was averaged for 5 s at each wavelength at a temperature of 20 °C. The protein concentration in 20 mM sodium phosphate buffer (pH 8.0), was 8 μM in a 1 mm pathlength cuvette and 30 μM in a 1 cm pathlength cuvette for far-UV and near-UV CD spectra respectively. Spectra were scanned from 250–190 nm in the far-UV CD region and from 320–240 in near-UV CD region. Dr1 concentration was 6 μM for unfolding experiments in a 1 mm pathlength cuvette. The CD signal at 208 nm was followed at 0.5 °C increments. For each data point the signal was averaged over 5 s. Temperature was increased at a rate of 1 °C/min.

Surface pressure measurements

Surface pressure measurements were carried out with a MicroTrough-S system from Kibron (Helsinki, Finland) at room temperature (22 °C) under constant stirring. The aqueous subphase consisted of 500 μl of 10 mM Hepes, 200 mM NaCl, pH 7.5. For monolayer insertion experiments the appropriate lipid mixture, dissolved in chloroform/methanol (2:1, v:v), was gently spread over the subphase to create a monolayer. The desired initial surface pressure was attained by changing the amount of lipid applied to the air/water interface. After 10 min, for solvent evaporation, the protein that had been preincubated in the presence of 10 mM DTT was injected through a hole connected to the subphase. The final concentration of Dr1in the Langmuir trough was 2 μM. The increment in surface pressure against time was recorded until a stable signal was obtained.

Surface plasmon resonance

Surface plasmon resonance measurements were performed on a Biacore X (Biacore, Sweden) apparatus at 25 °C. An L1 chip was equilibrated in 20 mM Tris/HCl, 140 mM NaCl, 1 mM EDTA, pH 8.5. Large unilamellar vesicles were prepared by extrusion as described previously [13]. The liposome-coated chip surface was prepared as described previously [40]. Briefly, liposomes were passed at 0.5 mM lipid concentration across the chip for 10 min at 1 μl/min and were washed with two injections of 100 mM NaOH for 1 min at 30 μl/min. Non-specific binding sites were blocked by one injection of 0.1 mg/ml BSA for 1 min at 30 μl/min. For the binding experiment Dr1 was injected at the desired concentration for 90 s at 30 μl/min. Four different concentrations of Dr1 were used in each experiment. After the sensorgrams were corrected for the contribution of blank injections of buffer, association (k_a) and dissociation (k_d) rate constants were globally fitted to a 1:1 Langmuir model using BIAevaluation 3.2 software (Biacore AB). The first 10 s of the injection and of the dissociation were not included in the fit due to refractive index effects. The dissociation constant (K_D) was calculated as follows:

RESULTS

Identification of ALP families

We were able to retrieve 72 sequences by using EqtII sequence as a probe in exhaustive searches of public databases (Figure 1, Table 1). All retrieved sequences had a significant E value that was typically around 10⁻⁸–10⁻⁹. Sequences were from three animal (chordates, cnidarians and molluscs) and two plant (mosses and ferns) phyla. The majority of the sequences were EST clones from teleost fishes (36 sequences) (Table 1). Twelve sequences were actinoporins, twelve further non-actinoporin sequences were found in Cnidaria and the rest were from mosses (3), ferns (1), Xenopus (4), chicken (2), skate (1) and the marine gastropod Monoplex echo (1). No hits with significant E value were found in other organisms, such as bacteria, Caenorhabditis elegans, Drosophila, rat, mouse, human or other plants. The ALP sequences are not expressed in large copy number, as the number of EST found for all organisms was very low (Table 1), i.e. typically from 0.002% (15 EST ALP clones from a total of 689581 in zebrafish; dbEST release 111105, 11 November 2005) to 0.045% (80 EST ALP clones from a total of 179615 in Hydra).

Figure 1. Alignment of actinoporin-like proteins.

Selected examples from three families of ALP are presented. An empty line separates each family from top to bottom: hydrozoan ALP, actinoporins, and ALP found predominately in vertebrates. Selected examples of fungal fruit body lectin family are shown for comparison below the line of asterisks, which highlights highly conserved residues in ALPs. Alignment is shaded according to conservation. Black, amino acids conserved in more than 90% of sequences; dark grey with light letters, conserved in more than 60% of sequences; light grey with black letters, conserved in more than 40% of sequences. Alignment was prepared and coloured using GeneDoc (http://www.psc.edu/biomed/genedoc/). The numbers in parentheses on the left-hand side denote the number of residues before the conserved domain. The size of insertion between helix and the fifth β-strand is shown in parentheses within the alignment. The number of residues after the domain is shown in parentheses on the right-hand side of the alignment. The total number of amino acids for each protein is written on the far right-hand side. The numbering above the sequence is according to the sequence of mature EqtII. AGAbis, Agaricus bisporus; Avt1, actinoporin from Actineria villosa; Dr1, Danio rerio ALP; Eqt2, equinatoxin II from Actinia equina; Fr, Fugu rubripes; Ga, Gasterosteus aculeatus; Gg, Gallus gallus; Hc, Haplochromis chilotes; He, Hydractinia echinata; HlyME; haemolysin from Monoplex echo; Hm, Hydra magnipapillata; Hv, Hydra vulgaris; Le, Leucoraja erinacea; Ma, Misgurnus anguillicaudatus; MARpol, Marchantia polymorpha; NEUcra, Neurospora crassa; Ol, Oryzias latipes; Om, Oncorhynchus mykiss; OrA, actinoporin from Oulactis orientalis; PHAchr, Phanerochaete chrysosporium; PLEcor, Pleurotus cornucopiae; Pp, Physcomitrella patens; Ss, Salmo salar; StII, sticholysin II from Stichodactyla helianthus; TORrur, Tortula ruralis; Tn, Tetraodon nigroviridis; Tr, Tortula ruralis; XERchr, Xerocomus chrysenteron; Xl, Xenopus laevis; Xt, Xenopus tropicalis.

Table 1. The distribution of AF superfamily members in public databases.

Number of different sequences in a single organism and percentage of EST clones, when available, is reported in parentheses.

Family/organism
1. Actinoporin-like proteins found predominately in vertebrates
Cnidaria
Hydractinia echinata (1; 0.01%)
Mollusca
Monoplex echo* (1)
Vertebrata
Chondrichthyes
Leucoraja erinacea (1; 0.009%)
Teleostei
Cyprinus carpio (1; 0.009%), Danio rerio (2; 0.002%), Fugu rubripes† (5),
Gasterosteus aculeatus (2; 0.002%), Haplochromis chilotes (5; 0.06%),
Misgurnus anguillicaudatus (1; 0.009%), Oncorhyncus mykiss (5; 0.02%),
Oryzias latipes (7; 0.03%), Paralichthys olivaceus (1; 0.026%), Salmo salar
(2; 0.008), Tetraodon nigroviridis† (5)
Amphibia
Xenopus tropicalis and X. laevis (4; 0.0003%)
Aves
Gallus gallus (2; 0.01%)
2. Hydrozoan actinoporin-like proteins
Hydra magnipapillata (10; 0.025%), Hydra vulgaris (1; 0.66%)
3. Actinoporins
Cnidaria
Actineria villosa* (1), Actinia equina* (3), Heteractis crispa* (1), Heteractis
magnifica* (1), Phyllodiscus semoni* (1), Oulactis orientalis* (2), Sagartia
rosea* (1), Stichodactyla helianthus* (2)
Plants
Bryophyta
Physcomitrella patens (1; 0.014%), Tortula ruralis (2; 0.62%)
Pteridophyta
Selaginella lepidophyla (1; 0.191%)
4. Fruit-body lectins
Fungi
Agaricus bisporus* (1), Arthrobotrys oligospora* (1), Aspergillus fumigatus†
(1), Aspergillus terreus† (2), Coccidioides immitis† (1), Gibberella
moniliformis† (1), Gibberella zeae (1; 0.012%), Neosartoria fischeri† (1),
Neurospora crassa (1; 0.004%), Paxillus involutus (3; 0.1%), Phanerochaete
chrysosporium† (2), Pleurotus cornucopiae* (2), Podospora anserina† (1),
Xerocomus chrysenteron* (1)
Bryophyta
Marchantia polymorpha (3; 0.85), Tortula ruralis (1; 0.02)

*Protein sequence was obtained by protein sequencing or was deduced from mRNA.

†Sequences were retrieved from the genome data.

The similarity with EqtII is confined to the C-terminal EqtII region 83–179 (Figures 1 and 2A), which can be described as a fingerprint of three separate regions with high local similarity. All fish sequences have an insertion of five amino acids between the first two parts (Figure 1). This insertion corresponds to the loop between strands β6 and β7 of the EqtII structure. The second part is the most conserved of all, possessing an invariant hydrophobic residue (Val¹⁰² in EqtII) and a pattern P-[WYF]-D (Figure 1) located on the broad loop between strands β7 and β8 (Figure 2A). The most variable part of the domain by length is the insertion between the second and third parts. It is 13–53 residues long and in EqtII comprises the loop between helix B and strand β10 (including short strand β9). The last motif of the domain encompasses EqtII strands β11 and β12. It is the least conserved of all three and possesses only two residues that are conserved in more than 90% of all ALP sequences. The conserved region is thus structurally well defined, encompassing almost half of the EqtII molecule (Figure 2A) and is arranged in two layers which flank helix B. Some of the most conserved residues of the domain help maintain its three-dimensional structure. Thus, the space between the two layers is filled with side chains of the hydrophobic amino acids, that is Phe¹⁰⁴, Val¹⁰², Val¹¹⁹, Leu¹⁰⁰ and Ile¹²¹ from the second part are in close contact with Ile¹⁷⁴ and Met¹⁶⁴ from the third part of the domain (Figure 2A). Similarly, Leu¹⁰³ from β7 and Leu¹³⁶ from helix B are in close contact and position helix B on one side of the layer (Figure 2A). Other conserved residues together with P-[WYF]-D pattern overlap well with the actinoporin POC (phosphocholine) binding site (Figures 1 and 2A) [12].

Figure 2. Structural properties of actinoporin-like proteins.

(A) Structural features of the conserved domain using the known structure of EqtII. The structure of the whole molecule of EqtII (PDB code 1IAZ) is presented on the lefthand side. The conserved region is coloured blue. The side chains of residues that participate in POC binding, which are highlighted in the alignment in (C), are shown as red sticks. The tyrosine and aspartic acid residues from the conserved motif P-[WYF]-D are shown in green as sticks. POC is coloured orange. The structural elements conserved in ALP are also labelled as strands (β6–β8 and β10–β12) and an α-helix B. The next image to the right shows T-antigen binding site in ABL (PDB 1Y2X and 1Y2W). Amino acids that participate in T-antigen binding, which are highlighted in the alignment in (C), are presented as sticks. T-antigen is coloured green. Two far righthand images show the CPK presentation of side chains of some of the residues that are more than 90% conserved in all ALPs. (B) Structural organization of ALP. Red, predicted signal peptide; green, cnidocyst sorting signal [49]; yellow, amphipatic helix needed for permeabilizing activity [14,16,17]; blue, the conserved ALP region. (C) Structural alignment of actinoporins and fungal fruit body lectins. Elements of secondary structure as defined by DSSP are presented above the sequences. Residues that participate in POC binding (POC) in StII and residues that participate in T-antigen binding (Tant) in ABL are shown below the sequences. PDB codes are written beside the names of proteins. The numbering is according to mature proteins. Colour scheme; green with white characters: hydrophobic amino acids (Val, Leu and Ile); green with black characters: aromatic amino acids (Trp, Tyr and Phe); green with red characters: small hydrophobic amino acids (Ala and Gly); yellow: polar amino acids (Asn, Gln, Met, Thr, Ser and Cys); blue: positively charged amino acids (Lys, Arg and His); red: negatively charged amino acids (Asp and Glu); white: proline.

No other known protein domains were found in ALP sequences according to searches using the InterPro database (Figure 2B). Furthermore, only a fraction of sequences possess a predicted signal peptide. Those that do include all the actinoporins, seven sequences from Cnidaria, the haemolysin from Monoplex echo and only one fish sequence, Hc1.

The similarity with fungal lectins was first revealed by comparison of their three-dimensional structures [24]. We did not retrieve sequences of the fungal fruit-body lectin family from databases by using EqtII or other ALP sequences as a probe, since sequence identity with actinoporins is only 11–14%. Nevertheless, there is a high degree of structural conservation with a root mean square deviation value of 1.38 Å over 68 Cα atoms of equivalent residues [24] (Figure 2A). Crucially, a binding site for cell-surface T-antigen in ABL is located on the same side of the β-sandwich as the POC site of actinoporins. Structurally equivalent amino acids form the binding site in both cases (Figure 2C) [24,25]. The fungal fruit-body lectin family is highly conserved. XCL and other family members have 19–65% identical residues. Apart from sequences presented in the alignment of Figure 2 in Birck et al. [24], we were able to find members of this family in six other fungal species and also plants Marchantia polymorpha and Tortula ruralis, by using XCL as a probe (Table 1). Interestingly, Tortula ruralis is at the moment the only organism to possess both ALP and fungal fruit-body lectin family sequences.

The retrieved actinoporins and actinoporin-like proteins, along with fungal fruit-body lectins, constitute a highly divergent group of hypothesized homologues. Particularly the lectins, which have only 6–17% of amino acids in identical positions compared with other sequences, making a purely sequence-based alignment practically impossible. However, the highly conserved, nearly identical secondary structure (Figure 2C) [24] enabled a clear homology assessment between different regions of the proteins.

Phylogenetic analysis revealed four distinct groups of proteins (Figure 3). Between-group sequence similarity (12–17%) was substantially lower than similarity within groups (45–69%). On the basis of these results, we tentatively assign these groups to four protein families: (1) ALPs found predominantly in vertebrates; (2) hydrozoan ALPs; (3) actinoporins; and (4) fungal fruit-body lectins. The first three groups are referred to as ALP families. The gap in average sequence similarity between the three ALP groups (27%) and lectins (14%) favour a separate phylogenetic position for the latter. The relationships between the ALP families are less obvious, although an outermost position of hydrozoan ALPs seems to be indicated. Strikingly, a small group of proteins from mosses and ferns are phylogenetically so closely associated with actinoporins that they can be considered as members of the same family.

Figure 3. Phylogenetic relationships of actinoporins, ALPs and fungal fruit-body lectins inferred by maximum likelihood using a Whelan–Goldman gamma model of amino acid evolution.

Numbers at nodes are bootstrap values in percentages (200 replicates); n.s. indicates non-supported nodes with a bootstrap value below 50%. Abbreviations of the protein and species names are as follows: ARToli, Arthrobotrys oligospora; ASPfum, Aspergillus fumigatus; ASPter, Aspergillus terreus; Cc, Cyprinus carpio; COCimm, Coccidioides immitis; GIBmon, Gibberella moniliformis; GIBzea, Gibberella zeae; HMgIII, an actinoporin from Heteractis magnifica; NEOfis, Neosartoria fischeri; PAXinv, Paxillus involutus; Po, Paralichthys olivaceus; PODans, Podospora anserine; PsTx20A, an actinoporin from Phyllodiscus semoni; RTXA, an actinoporin from Heteractis crispa; Sagros, an actinoporin from Sagartia rosea; Sl, Selaginella lepidophyla; SORmac, Sordaria macrospora. Other names are abbreviated as shown in Figure 1.

Membrane-binding behaviour of ALPs

On the basis of the sequence similarity and conservation of residues that participate in POC binding, we hypothesized that ALPs would all have membrane-binding behaviour. To check for this, we chose one of the ALP from fishes, zebrafish Dr1, and checked its membrane-binding ability with various in vitro systems.

Preparation and characterization of Dr1

Dr1 was expressed in E. coli as a His₆-tagged protein, purified from the bacterial cytoplasm and cleaved with thrombin to remove the His-tag. This procedure yielded pure protein of the expected size 21 kDa (Figure 4). First we needed to confirm that Dr1 is folded. As the ALP domain represents more than half of this protein we assumed that Dr1 structural characteristics are similar to EqtII. PSIPRED estimation of secondary structure content from sequence is H11/S39/T+R50 (percentage of helix/strand/turn + random) (Figure 4). The far-UV CD spectrum of Dr1 confirmed its folded nature and was strikingly similar to EqtII [17]. It has a negative maximum at 218 nm, indicating a high content of β-structure (Figure 4). Indeed, estimation of the secondary structure content by CDSSTR gave H7/S37/T24/R32 in agreement with the estimation from the sequence. Thus, Dr1 is a mainly β-sheet protein, very similar to EqtII in secondary structure content. The spectral features of Dr1 were lost when the protein was denatured in 6 M GdnHCl (Figure 4). The spectrum in the presence of GdnHCl is typical of denatured proteins, indicating that Dr1 in normal buffer is folded. Furthermore, Dr1 also has a well defined, and complex, near UV-CD spectrum, indicating a fixed chiral environment of aromatic residues (Figure 4).

Figure 4. Isolation and secondary structure of Dr1.

Top panel, 1 μg of protein was resolved by SDS/PAGE (12% gel) and stained with Coomassie Blue. M, molecular mass marker; lane 1, EqtII; lane 2, His₆-tagged Dr1; lane 3, thrombin-cleaved Dr1. Middle panel, prediction of secondary structure elements of Dr1 by PSIPRED [32]. Bold italics, helix; bold, strand; normal typeface, other (turns, irregular). Lower case letters, confidence of prediction 0–4 as reported by the PSIPRED program; upper case letters, confidence 5–9. ALP domain is shaded. Elements of the EqtII secondary structure determined from the three-dimensional structure are denoted by underlining with a full line for β-strands and dashed line for the α-helix. Bottom panel, Far-UV CD spectra of purified Dr1. The concentration of protein in 20 mM sodium phosphate buffer (pH 8.0) was 8 μM (solid line). Protein was denatured by incubation in 6 M GdnHCl for 90 min (dashed line). The spectra of EqtII from [17] is shown for comparison (broken line). The inset shows near-UV CD spectra of 30 μM Dr1 in 20 mM sodium phosphate buffer (pH 8.0). All spectra were recorded at 20 °C.

The fluorescence spectra of Dr1 in buffer and in the presence of 6 M GdnHCl are shown in Figure 5. The emission maximum of Dr1 in solution is 342 nm, typical of a protein, which excludes some of its fluorophores from the solvent. The emission maximum of the spectra in the presence of GdnHCl shows a red-shift to 354 nm, indicative of unfolding and exposure of its three tryptophans to the solvent. We further quenched tryptophan fluorescence of Dr1 by iodide. Stern–Volmer plots for the quenching in the absence and presence of 6 M GdnHCl are shown in Figure 5. Quexnching constant, K_sv, obtained from the slope, is 3.71 M⁻¹ for the unfolded sample. The data for Dr1 in buffer deviated from linearity and gave K_sv of 1.72 M⁻¹. This indicates that in folded Dr1 a fraction of the tryptophans are hidden within the protein fold, being inaccessible to the iodide. When using a modified Stern–Volmer plot for multiple emitting centres (Figure 5, inset) we obtained the fraction of tryptophans accessible to the iodide (f_a) of approx. 43%, which means that at least one of the tryptophans is fully exposed to the solvent while the other two are buried or partially buried in the hydrophobic core of the protein.

Figure 5. Tryptophan fluorescence and unfolding of Dr1.

Top panel, tryptophan fluorescence of Dr1 in 10 mM Hepes, 200 mM NaCl, 20 mM DTT, pH 7.5, without (solid line) or in the presence of 6 M GdnHCl (dashed line). Protein concentration was 1 μM. Excitation wavelength was 295 nm. Spectra were measured at 25 °C and with constant stirring. Middle panel, Stern–Volmer plots for the quenching with iodide in 10 mM Hepes, 200 mM NaCl, 20 mM DTT, pH 7.5, without (●) or in the presence of 6M GdnHCl (○). The modified Stern–Volmer plot for multiple emitting centres is shown in the inset for the protein in the buffer without denaturant. Other conditions are the same as the top panel. Bottom panel, denaturation of Dr1 followed by ANS fluorescence (dashed line) or CD spectroscopy (dots and solid line). Protein and ANS concentrations were 15 μM. Buffer was 10 mM Hepes, 200 mM NaCl, 20 mM DTT, pH 7.5. Excitation and emission wavelengths were fixed at 370 and 488 nm respectively, and slits for both were 5 nm. Experiment was performed in a temperature-controlled holder with constant stirring. For unfolding followed by CD spectroscopy Dr1 concentration was 6 μM in 20 mM sodium phosphate buffer (pH 8.0). The CD signal at 208 nm was measured at 0.5 °C increments. For each data point the signal was averaged over 5 s. The bandwidth was 1 nm. Temperature was increased in both cases at a rate of 1 °C/min. The data from CD spectroscopy was fitted to a sigmoidal curve using Origin software, yielding a midpoint of transition at 47.6±0.6 °C.

Finally we characterized the thermostability of Dr1 by ANS binding assay and CD spectroscopy. ANS increases quantum yield when it binds to newly exposed hydrophobic patches of temperature-denatured proteins (Figure 5). The denaturation temperature was calculated as the mean value between the temperature at which fluorescence starts to increase and the temperature at which maximum fluorescence signal was reached, which in the case of Dr1 was 46.3 °C. A similar transition was observed with CD when the signal was followed at 208 nm. In this case the T_m obtained was 47.6 °C, in close agreement with the ANS data.

On the basis of the above results we concluded that Dr1 is folded with at least some of the tryptophans buried in the interior of the protein. Although the melting temperature of Dr1 is lower than EqtII [17,41], the other spectral characteristics are similar. Therefore, we continued to characterize the properties of Dr1 and initially checked whether it possessed permeabilizing activity similar to EqtII. As predicted Dr1, which lacks a permeabilizing amphipatic helix, could not release the fluorescent probe calcein from DOPC/SM liposomes, typically used to monitor Eqt activity (results not shown). We then checked whether it could insert into lipid monolayers or bind to immobilized liposomes.

Surface-pressure measurements

First, we measured the ability of Dr1 to partition to the air/water interface. Dr1 readily interacts with this lipid-free surface in a concentration dependant manner (Figure 6). The maximum increment in the surface pressure was approx. 6 mN/m. Saturation was obtained at 2 μM protein, so this was the concentration we used in the assays in the presence of lipids. Dr1 is also able to insert into lipid monolayers of different compositions. Figure 6 shows the kinetics of insertion of Dr1 into lipid monolayers composed of DOPC, DOPC/SM (1:1, mol/mol), and DOPC/DOPG (1:1, mol/mol), at an initial pressure of 15 mN/m. Dr1 inserts readily in all three monolayers reaching a final value of approx. 25 mN/m. We then performed similar kinetics at different initial pressures and by plotting the increment in the surface pressure against the initial pressure we calculated the critical pressure, that is, the initial pressure at which no protein can insert in the monolayer (Figure 6). The values of the critical pressures were similar for all three compositions: 27.9 mN/m, 27.2 mN/m and 27.0 mN/m for DOPC, DOPC/SM and DOPC/DOPG monolayers respectively.

Figure 6. Interaction of Dr1 with lipid monolayers.

Top panel, accumulation at the air/water interface. Different concentrations of Dr1 were injected into a subphase composed of 10 mM Hepes and 200 mM NaCl, pH 7.5, and changes in the surface pressure were registered with a surface-pressure microbalance. Dr1 was always preincubated with 10 mM DTT. Experiments were carried out under constant stirring at room temperature. Middle panel: insertions into lipid monolayers at an initial pressure of approx. 15 mN/m. Concentration of Dr1 in the subphase was 2 μM. Other experimental conditions are the same as the top panel. Bottom: Critical pressure plots for different lipid monolayers. DOPC monolayer (■), DOPC/SM monolayer (●) and DOPC/DOPG monolayer (▲).

Surface plasmon resonance measurements

The binding of Dr1 was tested on supported liposomes by using an L1 chip on a Biacore X. Dr1 readily binds to liposomes composed of either DOPC, DOPC/SM (1:1, mol/mol), and DOPC/DOPG (1:1, mol/mol) (Figure 7). Blanks of the buffer without protein were subtracted from the raw sensorgrams and curves were then fitted with 1:1 Langmuir model. Reasonable fits with low χ² values typically around 1 were obtained (Figure 7). Again, in these surface plasmon resonance experiments Dr1 showed a similar level of binding for all lipid compositions tested, in agreement with monolayer insertion data. Kinetic parameters were quite similar for DOPC and DOPC/SM membranes (Table 2). For DOPC/DOPG the association rate was slightly lower and the dissociation rate was higher resulting in a K_D twice that of DOPC liposomes.

Figure 7. Interaction of Dr1 with supported liposomes.

Liposomes were immobilized on the surface of an L1 chip as described previously [40]. The flow-rate was 30 μl/min. The running buffer was 140 mM NaCl, 20 mM Tris/HCl and 1 mM EDTA, pH 8.5. The injection lasted 90 s and the dissociation was followed for 5 min. Each panel shows sensorgrams for 2.5, 5, 10 and 20 μM Dr1 (from bottom to top). Sensorgrams were fitted to a 1:1 Langmuir binding model, fits are shown as thick lines superimposed on sensorgrams.

Table 2. Kinetic parameters for the binding of Dr1 to supported liposomes as determined by SPR.

The buffer was 140 mM NaCl, 20 mM Tris/HCl and 1 mM EDTA, pH 8.5. n=2 or 3; means±S.D.

Liposomes	ka (M−1s−1)×103	kd (s−1)×10−3	KD (M)×10−6
DOPC	1.7±0.2	3.2±1.5	1.9±1.0
DOPC/SM (1:1, mol/mol)	1.3±0.2	2.7±0.3	2.1±0.6
DOPC/DOPG (1:1, mol/mol)	1.2±0.3	3.8±0.5	3.7±1.2

DISCUSSION

ALP families

A large number of sequences with a considerable degree of similarity to EqtII have been retrieved from databases. Phylogenetically, they can be attributed to three distinct families. ALP is obviously an ancient domain, as it spans the metazoan kingdom from diploblasts to vertebrates and is present in the genomes of some representatives of mosses and ferns. To further discuss phylogenetic and evolutionary scenarios of ALP sequences we need to keep in mind the taxonomic bias of the databases from which most sequences were retrieved. This bias notwithstanding, it is remarkable that despite the wide range of taxa sharing ALP sequences, many large groups with extensive available genomic information lack these domains altogether. Particularly noteworthy is their absence from protozoans, arthropods, nematodes, mammals and flowering plants. The phylogenetic pattern emerging on the maximum likelihood tree (Figure 3) seems to be consistent with a pre-metazoan origin followed by substantial subsequent differentiation in cnidarians, giving rise to the actinoporins and hydrozoan ALPs. The third monophyletic group and ALP family includes largely triploblastic animals, with invertebrates heavily under represented. This under representation appears to be real and not just a consequence of taxonomic bias.

ALP sequences are most abundantly expressed in teleost fishes (Table 1), where up to seven isoforms were observed in the single species, Medaka (Oryzias latipes), while some of them, common carp, zebrafish or salmon, have only one or two copies. The medaka fish paralogues are scattered among all five distinct teleost clades indicating that a high genomic diversity might be ancestral to this group. This finding is in line with recently discovered large-scale genome-level duplications within the stem lineage of the Actinopterygii (ray-finned fishes) [42]. There are two possible explanations for the lack of diversity in some teleost species and the generally low number of EST clones: first, secondary loss following reduced selection pressure and, secondly, under representation due to methodological bias. The fact that some of them are expressed in specific developmental stages (see below) supports the latter explanation.

As a general observation it can be concluded that few groups of metazoans display a diverse assemblage of ALPs, particularly anthozoans, hydrozoans, teleosts, and possibly frogs and birds. In some other groups ALPs seem to present as genomic relics (e.g. in snails and cartilaginous fishes), whereas in the majority of well-studied metazoan taxa these genes are lost. Such an unbalanced taxonomic distribution suggests that the biological functions conducted by ALPs are both diverse and specialized. This is underscored by actinoporins found in mosses and ferns, where their function might be fundamentally different from that in sea anemones. Furthermore, in the moss Tortula ruralis ALP domains are quite abundantly expressed (they are present in up to 0.6% of total available EST clones), indicating an important functional role. At the present stage of taxonomic sampling it is difficult to offer a plausible explanation for the isolated occurrence of plant ALPs amidst the large metazoan assemblage.

All ALP sequences, apart from actinoporins and the previously characterized echotoxin from the marine snail Monoplex echo [23], represent as yet undefined, novel proteins. They are mostly intracellular proteins, since only a few possess a predicted signal peptide. Of those, actinoporins and echotoxin are secreted toxins that also possess an additional amphipatic α-helix at their N-terminal, which is used for efficient pore-formation (Figure 1) [14,43]. In addition, a large fraction of the sequences from Hydra also have a predicted signal peptide, and it was proposed that they act as allomones for preying or defence using a mechanism similar to actinoporins [44].

The fact that there are so many different fish ALP makes it likely that at least some of them are functional copies. In fact, the expression of two of them was confirmed in medaka fish by a large in situ hybridization screen of stages 18, 24 and 32 [45]. Ol3 (MEPD clone 631-136-06-O) was abundantly and homogeneously expressed in all stages, while Ol4 (MEPD 631-135-09-M) was differentially expressed in the enveloping layer in stages 18 and 24. The exact biological roles of the majority of ALPs still needs to be determined. It has not escaped our notice that ALP sequences are presented mostly in aquatic organisms. This observation along with their abundance and preservation through evolution in widely separated taxonomic groups suggests functional importance.

Properties of Dr1

On the basis of EqtII similarity we wanted to check whether these proteins use the conserved ALP domain for membrane attachment. To test this, we cloned Dr1 and checked its membrane binding ability by biophysical methods. According to the structural characterization, recombinant Dr1 is a folded protein with a measured secondary structure content close to the predicted structure and is very similar to EqtII. It, however, is not haemolytic and does not permeabilize liposomes. Also, it does not show the SM specificity typical of actinoporins, as the kinetic parameters for the binding to immobilized liposomes or insertion into monolayers composed of DOPC or DOPC/SM membranes were very similar (Table 2). Electrostatic interactions do not contribute significantly to the binding, as the presence of excess negative charge in DOPC/DOPG membranes does not improve the binding. Estimated equilibrium dissociation constants are two orders of magnitude larger than for EqtII binding to the tethered lipid membrane composed of DOPC/SM (1:1, mol/mol) [14]. This can be explained in part by the contribution of the N-terminal amphipathic helix that is not present in Dr1 and, furthermore, Dr1 also lacks the equivalent of EqtII Trp¹¹², which is replaced by a leucine residue. This exposed tryptophan in EqtII contributes significantly to the binding [14] and was the only one that showed changes in ¹⁹F-NMR chemical shift upon addition of SM to dodecylphosphocholine micelles in a recent study, where EqtII tryptophans were labelled with ¹⁹F [15]. In the lipid monolayer experiments, Dr1 was close to, but below the threshold for deeper insertion into natural membranes, taking into account that the lateral pressure of the lipids in a membrane bilayer is approx. 30 mN/m [46]. In a DOPC monolayer, EqtII shows critical pressure values around 25 mN/m [47], similar to those observed with Dr1, and reversible association with DOPC membranes [47], that is, by using a sub-optimal lipid composition EqtII behaves in a similar way to Dr1.

To summarize, Dr1 can insert into lipid monolayers at low initial surface pressures and can reversibly associate with immobilized liposomes, which suggest that Dr1 can associate weakly with lipid membranes and therefore can be assigned as a peripheral membrane protein. Such behaviour is not uncommon in peripheral proteins, a large fraction of yeast PX and PH domains exhibit low affinity and were proposed to require additional domains within the same protein or other protein partners for efficient membrane binding [48].

AF domains, a novel superfamily of membrane binding domains

On the basis of sequence, structural and functional similarities described above we propose that three ALP families and the fruit-body fungal lectins are the tentative basis for a novel superfamily, the AF membrane binding domains. So far, members of this family were shown to bind various molecules of cell membranes, specifically the lipids SM [18,19] and GM₁ ganglioside [43] or carbohydrates [25,28]. The common link between ALP and fungal fruit-body lectins is a highly conserved structure at almost the full-length extent with weak sequence similarity, suggesting a homeomorphic protein superfamily. On the other hand the absence of structural information for hydrozoan ALP and ALP found predominately in vertebrates precludes any firm conclusions. If highly similar local regions in ALP are considered as separate domains, we can describe the AF domain as a homology-domain superfamily. The observed widespread distribution may be explained by this alternative. However, the number and order of highly similar local regions within sequences of ALP is strictly conserved, thus favouring the first possibility. It is quite possible that sequences presented here are only a subset of the bigger family and that many members exist that do not have sufficient sequence similarity to be recognized. Since the AF domain superfamily is ancient and highly divergent, it will be difficult to recognize novel members and trace the evolutionary history of ALP families unless more structural information is available. Currently, the AF superfamily is characterized by a structure rich in β-sheets, which are organized as a stable β-sandwich core flanked by an α-helix and binding determinants for different ligands located on one of the broad loops of the sandwich and α-helix (Figure 2).