Structural characterization of the putative ABC–type 2 transporter from Thermotoga maritima MSB8

Abstract

This study describes the structure of the putative ABC-type 2 transporter TM0543 from Thermotoga maritima MSB8 determined at a resolution of 2.3 Å. In comparative sequence-clustering analysis, TM0543 displays similarity to NatAB-like proteins, which are components of the ABC-type Na⁺ efflux pump permease. However, the overall structure fold of the predicted nucleotide-binding domain reveals that it is different from any known structure of ABC-type efflux transporters solved to date. The structure of the putative TM0543 domain also exhibits different dimer architecture and topology of its presumed ATP binding pocket, which may indicate that it does not bind nucleotide at all. Structural analysis of calcium ion binding sites found at the interface between TM0543 dimer subunits suggests that protein may be involved in ion-transporting activity. A detailed analysis of the protein sequence and structure is presented and discussed.

Introduction

The superfamily of ATP-binding cassette (ABC) transporters comprises a large group of proteins playing different roles in a vast range of biological processes. For the majority of ABC systems, the main role of these proteins is catalysis and the transport of substrates across the membrane [18]. They are involved in the transport of nutrients, elimination of waste or side products from cells, energy generation, cellular signaling, and antibiotic and antifungal resistance. In humans, they are also associated with many genetic diseases caused by transport failure across the membrane, such as bleeding disorders and many liver and eye diseases [8, 15, 28, 40]. ABC transporters span a large group of paralogous protein families, and their representatives are found in all three domains of life: Bacteria, Archaea and Eukaryota. The typical transporter is a heterotetramer composed of two integral membrane proteins and two cytoplasmic ATPases, which drive the translocation of the substrate by the hydrolysis of ATP. Basic ABC type 2 efflux transporters are integral membrane proteins consisting of two domains: a transmembrane domain (TMD) that provides a translocation pathway, and a cytoplasmic, water-exposed nucleotide-binding domain (ATP-BD) that hydrolyzes ATP [43].

TMDs usually consist of six transmembrane (TM) α-helices. There is little sequence identity between the TMDs of different ABC transporters. ATP-BDs are far more conserved in comparison. They contain highly conserved nucleotide binding motifs, namely Walker A (GXXXXGS/T) and Walker B (hhhhD, where “h” is a hydrophobic residue) [54]. A number of isolated ATP-BDs were crystallized and their three-dimensional structures determined [4, 21, 31, 46, 59]. In addition, several complete structures of different ABC transporters have been reported [13, 19, 22, 23, 29, 39, 41]. Depending on the host organism, the ATP-BD and TMD may be present together in a single polypeptide chain (most prokaryotes), or two separate ones (most eukaryotes). Bacterial and archaeal ABC uptake systems contain an additional (fifth) domain, the “extracellular substrate-binding protein”, which binds the substrate and delivers it to the permease domain.

ABC transporters are classified in the TCDB (Transport Classification Database; http://www.tcdb.org/) under the entry number of 3.A.1: the ATP-binding Cassette (ABC) Superfamily. This superfamily consists of more than 200 families that are further divided into sub-groups. Sequence comparisons against this database showed that TM0543 from Thermotoga maritima MSB8 belongs to the subfamily of the Na⁺ efflux pump NatAB of ABC–type transporters (3.A.1.115.1), represented by a Bacillus subtilis (GI:1663528) transporter [5]. Efflux pumps from this subfamily are typically expressed by a pair of genes colocated in the same operon: one encoding ATP-binding proteins that are either components of an ABC-transport system or a traffic ATPase (natA), and one encoding permease substrate–binding proteins (natB). However, in the case of TM0543, both the NatA-like and NatB-like domains are encoded by a single gene.

Here, the structure of the predicted nucleotide-binding domain (ATP-BD) of the putative ABC type 2 transporter from Thermotoga maritima MSB8 (TM0543) is reported, which was determined by the single-wavelength anomalous dispersion (SAD) method and refined at a resolution of 2.3 Å. A model of the transmembrane domain (TMD) of TM0543 was also constructed using homology modeling. Structural and sequential comparisons of this protein to other known, well-described ABC transporters provide detailed information regarding conserved residues and other characteristic features of TM0543 from Thermotoga maritima MSB8.

Materials and methods

Cloning, expression and purification

The recombinant Thermotoga maritima MSB8 (TM0543) gene product, including a 6-residue polyhistidine (His6) tag at the N-terminus, was expressed in E. coli using the p15Tv lic vector. The cloning, expression and purification procedures were done as previously described [60]. Crystals of TM0543 were grown by the hanging-drop vapor-diffusion method at 295 K. For crystallization, a 3 μL drop consisting of 1.5 μL protein solution (10 mg/mL Se-Met-TM0543 in 10 mM HEPES buffer, pH 7.5 and 500 mM NaCl) and 1.5 μL well solution (28% v/v PEG400, 0.2 M CaCl₂ and 0.1 M HEPES buffer pH 7.5) was equilibrated against 100 μL of well solution. Crystals selected for data collection were transferred to paratone-N oil and flash-cooled in liquid nitrogen at 77 K.

Data collection, structure determination and refinement

Diffraction data were collected from a single crystal of TM0543 at the 19-BM beamline [42] of the Advanced Photon Source (APS) at Argonne National Laboratory. The diffraction data were processed and scaled with the HKL-2000 program suite [36]. X-ray data-collection statistics are summarized in Table 1. The TM0543 structure was solved by single-wavelength anomalous dispersion (SAD). Initial phases were calculated with 2.4 Å resolution data using the program HKL–3000 [33]. The HKL-3000 suite interacts with SHELXD, SHELXE [45], MLPHARE [35], DM [7], CCP4[57], ARP/wARP [38], and COOT [10]. Subsequent electron density map modification followed by initial model building was done using HKL-3000. The rest of the model was built manually with COOT. The experimental model, comprising residues 47–191 of TM0543, was then refined using REFMAC5 [34]. Solvent atoms were initially built with ARP/wARP, and later added or removed by manual inspection. The final R_work and R_free (calculated with a randomly selected 5% of reflections omitted from refinement) were 20.5% and 22.7%, respectively. The crystal structure contains one polypeptide chain per asymmetric unit. The refinement statistics and characteristics of the atomic models for TM0543 are given in Table 1. The model was validated with ADIT [58], MOLPROBITY and KING [30]. The atomic coordinates and structure factors for TM0543 have been deposited at the Protein Data Bank (PDB) with accession code 3CNI.

Table 1.

Crystallographic parameters, and data collection and refinement statistics.

	TM0543
Crystal Parameters
Space group	P 64 2 2
Cell dimensions:
a=b, c (Å)	65.6, 141.2
α, β, γ, (°)	90, 90, 120
Matthews coefficient (Å3/Da)	2.56
Solvent Content (%)	52
Data Collection
Wavelength (Å)	0.97934
Temperature, K	100
Resolution (Å)	56.8 - 2.3 (2.34 - 2.30)
Rsym (%)	5.0 (71.3)
No. of unique reflections	8587
No of unique reflections in Rfree set	591
Mean Redundancy	13.4 (10.6)
Overall completeness (%)	99.4 (99.2)
Mean I/σ I	65.7 (3.4)
Refinement Residuals (f ≥ 2σf):
Rfree (%)	22.3
Rwork (%)	20.5
Completeness (%)	99.6 (99.2)
Model Quality
RMSD bond lengths (Å)	0.022
RMSD bond angles (°)	1.8
MolProbity Ramachandran statistics†:
most favored (%)	98.6
allowed (%)	1.4
B-factor from Wilson Plot (Å2)	64.9
Mean overall B-factor (Å2)	51.5
Model Contents
Protomers in ASU	1
Protein residues	37-192
No. of protein atoms	1094
No. of calcium ions	2
No. of water molecules	31

Data for the highest resolution shell are given in parentheses. Ramachandran statistics were calculated with MOLPROBITY. R_merge was calculated for merged Friedel pairs.

^†Pro and Gly residues were excluded from calculation

Sequence analyses and clustering

PSI–BLAST [1] was used to search for similar sequences, and multiple sequence alignments were constructed using MUSCLE [9]. The alignments were used to generate a set of profile HMMs (Hidden Markov Models) using HHmake from the HHsearch package [47, 48] and these HHMs were used to search a database of profile-HHMs corresponding to alignments of protein families obtained from the COG, KOG [50], and PFAM [3] databases. This comparison of profile-HMMs (including protein sequences and consensus secondary structure) was done using HHsearch 1.5.0.1 with its default parameters (a maximum of 8 PSI-BLAST iterations, with activated modes of alignment and secondary structure scoring). Transmembrane regions were predicted using HMMTOP [51], TopPred [53], TMPred [17] and SOSUI [16]. Sequence motifs were determined both automatically with Motif Finder [2] and via manual analysis of a multiple sequence alignment.

To visualize pairwise similarities between and within protein families, CLANS (CLuster ANalysis of Sequences) [12], a Java utility that applies a version of the Fruchterman–Rheingold graph layout algorithm, was used. CLANS creates a three-dimensional graph where each vertex represents a sequence and each edge the pairwise similarity of two sequences. CLANS uses the p-values of high-scoring segment pairs (HSPs) obtained from an N × N BLAST search to compute attractive and repulsive forces between each sequence pair in a user-defined dataset. The attractive force between each pair of sequences is proportional to −log p for the pairwise interaction, so more similar pairs attract more strongly. To start, sequence vertices are randomly seeded in 3-D space, and iteratively moved within this environment by the algorithm according to the force vectors resulting from all pairwise interactions until the system converges.

Homology modeling

Both the full-length and individual domains of the target sequence were submitted to the GeneSilico MetaServer [25] to identify the TM0543 fold and to select the best template for homology modeling of the fragments of the protein missing from the crystal structure, namely the TMD. The localization of transmembrane regions (TM, residues 192–412) was predicted based on the amino acid sequence as described in the "sequence analyses and clustering" section. The alignments between the sequence of TM regions reported by the metaserver and the structures of selected templates, members of the aquaporin-like fold (SCOP f.19.1.1) as identified by Pcons, were used as a starting point for modeling the regions of TM0543 missing tertiary structure: the N-terminal fragment (residues 1–46) and the C–terminus. The two missing regions were modeled separately and the hybrid model was constructed using the "FRankenstein's Monster" method [24]. This was comprised of cycles of model building by MODELLER [11] and SwissModel, evaluation by MetaMQAPII [37], manual realignment of poorly scored regions and merging of the best scoring fragments (GDT_TS: 40.408, RMSD: 4.512). The quality of the final model (i.e. to which extent it approximates the real, unknown structure) was predicted using MetaMQAP and PROQ [55]. The TMD was docked to the ATP-BD using GRAMM [52].

Structure analysis

Structures were manipulated and modeled with SwissPDBViewer [14, 44]. Visualization and structure figures were generated with CCP4MG [32]. Structural similarity searches and superimpositions were done with DALI [20] and ProFunc [26].

Results and discussion

Multiple sequence analysis of TM0543

A multiple sequence alignment (MSA) was calculated for all sequences homologous to TM0543 collected by a database search. However, the MSA exhibited poor quality due to significant divergence between different paralogous families. In order to identify a group of more closely related (and thus more confidently alignable) members, all collected sequences of TM0543 homologs were clustered using CLANS.

The clustering analysis is presented in Figure 1. Most of the “typical” ABC transporters described in the literature by Reizer et al. [40] and Lee et al. [27] cluster together (green). There are a number of other clusters as well. COG1511 (yellow) contains predicted membrane proteins. COG0842 (brown) contains multidrug transport system permease proteins. PF01061 (purple) contains bacterial ABC type 2 membrane transporters that export drugs and carbohydrates. COG1668 (blue) contains NatAB Na⁺ efflux pumps, permease components and a putative ABC transporter domain from Porphyromonas gingivalis W83 (PDB structure 2P0S). The last three listed COGs also belong to the conserved domain superfamily cl00725 (ABC2 membrane superfamily in NCBI CDD database). Members of COG0842 form part of a defense mechanism that includes COG1668, PF01061, COG1277 (the permease component of the NosY – ABC– type transport system involved in multi–copper enzyme maturation) and COG1682 (TagG, ABC–type polysaccharide/polyol phosphate export systems—a permease component involved in carbohydrate transport and metabolism and cell envelope biogenesis in the outer membrane). The remaining cluster (teal) collects annotated NatAB-like sequences and the TM0543 sequence. This cluster includes other bacterial proteins, including the previously described Bacillus subtilis NatB [5]. The NatAB/TM0543 cluster is most closely connected to COG1511, PF01061 and COG1668.

Fig. 1. Clustering analysis of ABC transporters using CLANS.

COG0842 (brown) represents multidrug transport system permease components, COG1511 (yellow) predicted membrane proteins, PF01061 (red) ABC2 membrane transporters, and COG1668 (blue) NatB Na⁺ efflux pump permease components. Vertices shown in teal represent annotated NatB–like sequence and the sequence of 3CNI (the query sequence). “Typical” ABC transporters as described by Lee et al. [27] and Reizer et al [40] are shown in green.

The NatAB family includes all members of COG1668 and members of the family are thought to be the ABC-type Na⁺ efflux pump permease component implicated in energy production and conversion as well as the inorganic ion transport and metabolism. A multiple sequence alignment of TM0543 homologs from NatAB family members is presented in Supplementary Figure S1. Sequence analysis suggests that the most similar homolog of the TM0543 protein is the previously described NatAB system from Bacillus subtilis, and is considered to be a part of the ABC type 2 transport system (local sequence identity of 24% and similarity of 46%, as calculated by BLAST pairwise sequence alignment) [56]. Additionally, analysis of potential sequence motifs using the amino acid sequence only predicted the investigated Thermotoga maritima target to be a putative ABC type 2 transporter. Motif Finder predicted regions that correspond to 156–395 as an ABC type 2 transporter (PROSITE PROFILE method) domain. Residues 139–364 were assigned by PFAM to PF01061 with an expectation value (e-value) of 4.4 × 10⁻⁵ and the region comprising residues 173–325 were assigned to the same protein family (PF01061) by the PFAM_FS method with an e-value of 9.4 × 10⁻¹⁰.

According to previously published data [49], the ATP–binding domain of NatAB family members shows a greater degree of conservation than the membrane spanning domain.. Typically ATP-BD domains contain a classical Walker A motif (G-X-X-X-X-G-K-S/T) and Walker B motif (ending in h-h-h-h-D, where h represents any hydrophobic residue), which together comprise the presumed nucleotide binding fold [54]. However, the multiple sequence alignment of TM0543 with its homologs from NatAB family members (Supplementary Figure S1) suggests that the C-terminal TMD domain of the full-length TM0543 has a higher degree of conservation in than N-terminal ATP-BD domain. The region corresponding to the Walker A motif in the predicted ATP-BD domain of TM0543 is only weakly conserved. The Walker B motif is more strongly conserved (see Figure S1), which is defined only by a conserved aspartic acid (D55 in TM0543) and an V/L/I-X-V/L/I repeat.

Structure of the putative ABC transporter TM0543

The crystal structure of the putative ABC transporter TM0543 from Thermotoga maritima MSB8 was obtained by SAD at 2.3 Å resolution. The asymmetric unit of the crystal contains one molecule of the putative N-terminal domain of TM0543. The refined model of the putative domain consists of 145 residues (residues: 47–191) out of 156 presented in the purified protein construct (residues: 37-193). Thirty-three N– terminal residues and one C–terminal residue were omitted in the structure due to the lack of corresponding electron density. The putative TM0543 structure is composed of an 6-stranded antiparallel β-sheet comprising β1 (48–53), β2 (72–77), β3 (94–98), β4 (112–120) β5 (163–171), and β6 (174–176); surrounded by five α–helices and two short helical segments: α3 (58–70) α4 (80–91) α5 (102– 109), α6 (125–143) 3/10 α (146–150), 3/10 α (157–160) and α7 (181–188). Figure 2A shows the overall structure of the putative TM0543 domain.

Fig. 2. Structural representation of a putative TM0543 protein.

A. Ribbon diagram of the putative domain of ABC-type transporter TM0543 from Thermotoga maritima MSB8. Secondary structure elements are labeled. Residues involved in calcium ion (red sphere binding are shown in cylinder representation, colored by atom type (carbon in light green and oxygen in red); B. Ribbon diagram of TM0543 dimer; C. Electrostatic potential molecular surface representation of the TM0543 dimer. The surface was created by program CCP4MG and colored by surface potential charge scaled from negative in red (−0.5 V) to positive in blue (+0.5 V); D. Superposition of TM0543 protein structure (green) with structure of zinc transporter from Thermus thermophilus HB8 (gold); E. Superposition of TM0543 protein structure (green) with structure of a putative ABC transporter domain from Porphyromonas gingivalis W83 (light blue).

The crystal packing analysis suggests that putative domain of TM0543 forms a tightly associated dimer with symmetry equivalent molecule (Figure 2B). The total interface area of the surface between dimer subunits (1129 Ų) and predicted interaction energy (ΔG = −23.2 kcal/mol) calculated by PDBePISA server (www.ebi.ac.uk/pdbe/pisa) suggests that TM0543 dimer is most likely biological and important for the function of the protein. The interface area of the dimer is formed by helices α6, α8 and α9. There are in total 32 interface residues that form 5 hydrogen bonds.

The electrostatic potential molecular surface reveals that a cavity with volume of 504 ų is formed at the interface in the center of the TM0543 dimer (Figure 2C). There are two symmetry related lysine residues (K142) that create a small positively charged area inside the large negatively charged pocket. The inner surface of the cavity is lined by residues I116, W118, I134, S135, E139, K142, D161, Q167, S180, P181, and E182. The residues I116, I134, S180, and P181 are conserved among TM0543 homologs (Figure S1). The negative potential of the cavity suggests that the TM0543 protein may prefer to bind small positively charged ligands or ions. Two calcium ions related by 2-fold rotation axis were found inside the cavity of TM0543 dimer. Oxygen atoms of the side chains of S135, E182 and four water molecules coordinate calcium ion inside the cavity.

Additionally, we have found that a well-ordered calcium ion is also bound near the Walker B motif on the surface of the putative TM0543 protein. This calcium ion binds to the residue E61 and T56 from Walker B motif through water molecule. The Walker A and Walker B motifs correspond to ATP binding pocket in known structures of NatAB family member[54]. . Therefore, the presence of calcium ion in our structure could occupy the position of divalent metal required for ATP hydrolysis. Unfortunately, the region containing the Walker A motif is omitted in our structure. Therefore, it is difficult to predict the nucleotide-binding site and interactions between ATP and a predicted ATP-BD domain in the TM0543 putative structure. The overall fold of the predicted ATP-BD domain of TM0543 is different from that of any known protein from NatAB family. Due to the relative lack of sequence and structure similarity between ATP-BD domains of TM0543 homologs, it is not clear if TM0543 is capable of binding nucleotide. It is also possible that N-terminal domain of TM0543 protein serves a completely different function.

Structural comparison of TM0543 against the PDB database using the secondary structure matching program of the ProFunc server identified the structure of a zinc transporter from Thermus thermophilus HB8 [6] with PDBID 3BYP (a Z–score of 2.8 and RMSD 3.5 Å, as calculated by ProFunc) as having the most similar crystal structure (Figure 2D). Additionally, a search for structural homologs with the DALI server revealed another structure of a putative ABC transporter domain (Figure 2E) from Porphyromonas gingivalis W83 with PDBID 2P0S (a Z–score of 5.3 and RMSD 3.0 Å, as calculated by DALI). However, the target protein does not show significant sequence similarity to either of these transporters, and variations in the secondary structure elements at the N- and C-terminus between superimposed structures can be observed (Figure 2D, 2E). In the superimposed structures three β strands (β1-β3) and helix α6 display significant structural conservation. Additionally, the structure of the TM0543 dimer shows significant similarity in dimer architecture to an uncharacterized ABC transporter domain from Porphyromonas gingivalis W83, as compared to the dimer of the zinc transporter from Thermus thermophilus HB8. All proteins have different topology in each of their putative ligand binding cavities that are most likely related to their specific functions.

Model structure of the Transmembrane Binding Domain

To date, no (full length) crystal structure of a NatAB–like homolog highly similar to TM0543 has been deposited in the PDB. Since the TMD (residues 192–412) was not present in the crystallized protein, we tried to reconstruct a complete structure of TM0543, which might have provided guidelines for domain assembly of the TM0543 transporter. Sequence search analysis for the TMD domain of the TM0543 protein (residues: 192–412) predicted that it is composed almost entirely of transmembrane (TM) helices (see Materials and Methods). Predictions performed with the GeneSilico metaserver showed that the best modeling templates (according to the Pcons consensus method) are members of the aquaporin-like fold composed of a core of 8 helices (2 short helices surrounded by 6 long transmembrane helices; SCOP f.19.1.1). Therefore, the alignments between the sequence of TM regions of TM0543 and the structures of selected templates were used as a starting point for homology modeling of the ternary structure of the C- terminal region of TM0543 (as described in Materials and Methods). The resulting model of TMD was composed of six α-helices (TM1 (193–203), TM2 (225–250), TM3 (289–311), TM4 (320–339), TM5 (348–364), and TM6 (374–400) which together form an anti-parallel bundle (Figure S2). In the model the TM1 is the “linking” helix between the two domains (ATP-BD and TMD). Interactions between the two domains are mostly formed with helices α1 (residues 3–11) and α2 (17–41) of the predicted ATP-BD domain (Figure S2), which were not present in the crystallographic structure. These helices were also modeled.

Summary

In conclusion, through sequence-clustering analysis, we have determined that TM0543 is closely related to a family of ABC-type transporters of the sodium efflux pump permease component, in particular to members of the NatAB family. The multiple sequence alignment of TM0543 with its homologs shows that the predicted ATP-BD domain has lower sequence conservation then TMD domain. We have determined the crystal structure of the putative ATP-BD domain from TM0543 protein from Thermotoga maritima MSB8. The structure characterization shows that predicted ATP-BD domain of TM0543 has a different structural fold than any other NatAB family members of known structure. The structure of TM0543 ATP-BD also displays a unique dimer architecture among known ABC transporters and has a large negatively charged cavity formed at the interface between dimer subunits. The electrostatic surface analysis suggests that protein could be involved in transport of positively charged ions such as calcium ions, one of which was found inside the cavity.

List of abbreviations

TMD

transmembrane domain

ATP-BD

nucleotide–binding domain

TM

transmembrane α–helices

ABC Superfamily

ATP-binding Cassette Superfamily

TCDB

Transport Classification Database (http://www.tcdb.org/)

PDB

Protein Data Bank