Correspondence: Carlos Cervantes, Instituto de Investigaciones Químico-Biológicas, Universidad Michoacana, Edificio B-3, Ciudad Universitaria, 58030 Morelia, Michoacán, México. Tel./fax: 52 + (443) 326 5788; e-mail: firstname.lastname@example.org
Short-chain monodomain family comprises pairs of membrane proteins of about 200 amino acid residues each that belong to the chromate ion transporter (CHR) superfamily. The short-chain CHR homologous pair Chr3N/Chr3C from Bacillus subtilis strain 168 confers chromate resistance only when both proteins are expressed. Membrane topology of the Chr3N and Chr3C proteins was determined in Escherichia coli by the analysis of translational fusions with reporter enzymes alkaline phosphatase and β-galactosidase. Each short-chain CHR protein was found to consist of five transmembrane segments with antiparallel orientation between them. The C terminus of Chr3N is located in the cytoplasm, whereas the C terminus of Chr3C is located in the periplasm. In silico analyses suggest that this antiparallel arrangement is shared by all protein members of the short-chain CHR3 subfamily and that the two Chr3N/Chr3C proteins might carry out distinct functions for the transport of chromate.
The best-studied bacterial chromate resistance system is that of the Pseudomonas aeruginosa ChrA protein, which functions as a chemiosmotic pump that extrudes chromate ions from the cytoplasm using the proton motive force (Alvarez et al., 1999). ChrA belongs to the chromate ion transporter (CHR) superfamily (Nies et al., 1998; Nies, 2003), which includes hundreds of homologues from all three life domains (Díaz-Pérez et al., 2007; Henne et al., 2009). The CHR superfamily is composed of two families of sequences: (1) short-chain monodomain family made up of proteins of about 200 amino acid (aa) residues and (2) long-chain bidomain family of about 400 aa (Díaz-Pérez et al., 2007). Genes encoding short-chain CHR proteins are organized mainly as homologous tandem pairs (Díaz-Pérez et al., 2007).
Several proteins of the long-chain CHR family have been demonstrated to function as membrane transporters able to extrude chromate ions from the cytoplasm (reviewed in Ramírez-Díaz et al., 2008), and paired genes encoding short-chain CHR proteins from Bacillus subtilis strain 168 were also shown to confer resistance to chromate by chromate efflux when expressed in Escherichia coli (Díaz-Magaña et al., 2009). With respect to membrane topology, the long-chain ChrA protein from Cupriavidus metallidurans has been reported to have 10 transmembrane segments (TMSs), in an unusual 4 + 6 arrangement (Nies et al., 1998). Another long-chain CHR member, the ChrA protein from P. aeruginosa, possesses 13 TMSs in an unusual 6 + 1 + 6 arrangement, with one extra TMS inserted in the middle of the two homologous domains (Jiménez-Mejía et al., 2006). This last arrangement in P. aeruginosa ChrA protein determines that amino and carboxyl terminal domains are oppositely oriented with respect to the membrane. Interestingly, the 4 + 6 arrangement of the C. metallidurans protein suggests that amino and carboxyl terminal domains possess the same orientation. Based on the distribution of positively charged lysine and arginine residues among predicted hydrophilic loops of short-chain CHR proteins, Díaz-Pérez et al. (2007) proposed that short-chain CHR protein pairs possess opposite membrane orientation. However, the number of TMSs in SCHR proteins is uncertain. In an attempt to understanding the functioning of this protein family, PhoA and LacZ translational fusions of paired B. subtilis Chr3N/Chr3C proteins were constructed and used to obtain insights on short-chain CHR membrane topology. Our results showed that the structure of short-chain CHR protein pairs consists of five TMSs each, with antiparallel orientation in the membrane.
Materials and methods
Bacterial strains and growth conditions
Escherichia coli strains XLBlue (recA1 endA1 gyrA96 thi-1 hsdR17 supE44 relA1 lac [F′ proA+B+lacIqZ∆M15 Tn10 (TetR)] (Stratagene, La Jolla, CA), and CC118 [araD139 Δ(ara leu) 7697 ΔlacX74 phoAΔ20 galE galK thi rpsE rpoB argE(Am) recA1] (Manoil & Beckwith, 1986) were utilized as hosts for plasmids. Cells were routinely grown at 37 °C with shaking in LB broth (Sambrook et al., 1989). For the preparation of solid medium, 1.5% agar was added to LB broth.
Plasmid pUCywrB_A (Díaz-Magaña et al., 2009) was utilized as a source of B. subtilis chr3N and chr3C genes. The pJET1.2 blunt vector (Fermentas, Glen Burnie, MD) was used to clone PCR-amplified DNA fragments. PhoA and LacZ expression vectors pUCPphoA and pUCPlacZ (Jiménez-Mejía et al., 2006), respectively, were employed for construction of translational fusions. These vectors have a lac promoter upstream the phoA or lacZ genes, respectively, as well as an intervening kanamycin-resistance gene between KpnI and XbaI endonuclease restriction sites (Nies et al., 1998; Jiménez-Mejía et al., 2006).
Plasmid DNA was isolated from cultures grown in LB broth by an alkaline lysis method and visualized following electrophoresis in 1% agarose gels in TAE buffer (Sambrook et al., 1989). Plasmids were purified with Wizard plus SV miniprep DNA purification system (Promega, Madison, WI) according to the supplier's instructions.
DNA manipulations and sequencing
Endonuclease restriction enzymes were purchased from Promega, and DNA was digested following standard procedures (Sambrook et al., 1989). Polymerase chain reaction (PCR), DNA ligations, and electrotransformation of E. coli strains were conducted as described (Sambrook et al., 1989). DNA sequencing was carried out at the Department of Genetics, CINVESTAV, Irapuato, México.
Construction of Chr-PhoA and Chr-LacZ fusions
Amplification of DNA fragments of the chr3N and chr3C genes from pUCywrB_A plasmid was carried out by PCR with oligonucleotides designed to yield translational PhoA/LacZ fusions within hydrophilic loops of the Chr3N and Chr3C proteins, according to a topological model based on hydropathic profiles and secondary-structure prediction programs. Oligonucleotides, obtained from Invitrogen (Carlsbad, CA), were designed with either KpnI [forward primers, ODSN1(+) for Chr3N, and ODSC1(+) for Chr3C, common to all fragments] or XbaI (reverse primers) restriction endonuclease sites (Supporting Information, Table S1). PCR fragments were purified, cloned into the pJET1.2 blunt vector, and transferred to the E. coli XLBlue strain. Isolated recombinant plasmids were then digested with KpnI plus XbaI, and the resulting fragments were subcloned into the corresponding sites of either pUCPphoA or pUCPlacZ vectors. This cloning strategy created in-frame fusions of the different chr3N/chr3C regions with the corresponding reporter gene. Correct reading frames at fusion sites were confirmed by DNA sequencing, utilizing the direct primer.
To measure the expression of reporter genes in the fusions, recombinant plasmids were transferred by electroporation into E. coli CC118 strain (lacking phoA and lacZ genes). PhoA and LacZ enzyme activities were determined utilizing chromogenic substrates in permeabilized cells, as previously described (Jiménez-Mejía et al., 2006). Enzyme activities of control cells containing only the vectors (< 5% of highest values) were subtracted from values determined in the fusions. Activities were normalized by adjusting the highest value measured to 100%; only values higher than 15% were considered as significant. Measurements were repeated at least three times by duplicate assays.
Protein sequence analysis
Orthologous Chr3N/Chr3C amino acid sequences were retrieved by Blastp searches at the UniProt site (Jain et al., 2009) (http://www.uniprot.org/blast). Phylogenetic analyses performed using the mega5 software (Tamura et al., 2011) (http://www.megasoftware.net/) were used to identify protein sequences as members of the short-chain CHR3 subfamily. Progressive multiple protein sequence alignments were calculated using clustal x ver. 2 (Larkin et al., 2007) (http://www.clustal.org/). DS Gene v1.5 software suite (Accelrys Inc., San Diego, CA) was used to generate hydropathic profiles [calculated according to Kyte & Doolittle (1982), with a window of 21 amino acid residues], and von Heijne transmembrane plots (von Heijne, 1992). Free energy for membrane insertion of potential transmembrane helices was calculated using the ΔG prediction server v1.0 (Hessa et al., 2007) (http://dgpred.cbr.su.se) and Membrane Protein Explorer (Snider et al., 2009) (http://blanco.biomol.uci.edu/mpex/). Topology models were generated using the consensus web server topcons (Bernsel et al., 2009) (http://topcons.cbr.su.se/), which uses a number of prediction programs (octopus, pro-tmhmm, prodiv-tmhmm, and scampi-single and scampi-multi), to produce a consensus result, thus improving the reliability of predictions.
To determine more precisely the membrane topology structure of proteins from the short-chain CHR family, the B. subtilis Chr3N/Chr3C protein pair was employed. Translational PhoA/LacZ fusions were constructed in predicted hydrophilic regions (Fig. S1a), as described under 'Materials and methods'.
Membrane topology of Chr3N
Topology models predicted that the N-terminal end of B. subtilis Chr3N was located in the periplasm, just about 12 residues distal of TMS1 (Fig. S1b). Fusions were not constructed in this short hydrophilic region because Chr3N-PhoA recombinant proteins would remain in the cytoplasm by lacking a TMS that might translocate PhoA to the periplasm. The shortest Chr3N fusion, made in residue Gly24 (predicted to reside within TMS1, close to the cytoplasm), yielded high LacZ activity and no significant PhoA activity (Fig. 1a). Thus, the presence of TMS1 could not be clearly demonstrated, and we rely on the prediction of the topology models to suggest that the N-terminal end of Chr3N is located in the periplasmic space (Fig. S1b). Fusions located in amino acids Asn37, Ile50, and Lys74 showed LacZ activity and null PhoA activity (Fig. 1a), indicating that this region is situated in the cytoplasm; this location is in agreement with prediction models (Fig. S1b), which showed large hydrophilic (cytoplasmic) regions between residues 50 and 90. Fusions at residues His106, Leu137, Ile161, and Ser189 yielded alternating high and low PhoA activities (Fig. 1a), indicating that these regions have corresponding alternate periplasmic and cytoplasmic locations; this location was confirmed by the fact that these four fusions also yielded alternating low and high LacZ activities (Fig. 1a). The topology at this region, which spans the last four TMSs of Chr3N, is in complete agreement with prediction models (Fig. S1b). Together, these results suggested a topology of five TMSs for Chr3N, with the N-terminal end in the periplasm and the C-terminal end in the cytoplasm (Fig. 1b).
Membrane topology of Chr3C
Topology models predicted that the N-terminal end of B. subtilis Chr3C was located in the cytoplasm (Fig. S1b). Accordingly, fusions located in amino acids Tyr36 and Met47 showed both high PhoA activity and low LacZ activity (Fig. 1c), indicating that this region was situated in the periplasm; a TMS should be present distal of Tyr36 to allow for this region to be translocated to the periplasm and to yield PhoA enzyme activity. These data confirmed that the N-terminal of Chr3C is located in the cytoplasm. Topology models predicted a large hydrophilic (periplasmic) Chr3C region spanning residues 50 through 90 (Fig. S1b). However, fusions at Val66 and Ala70 displayed unexpectedly low and null PhoA activity, respectively (Fig. 1c); the Ala70 fusion showed low LacZ activity, indicating that it was not at the cytoplasm. As fusion at Gly109 showed significant LacZ activity, a TMS must be present between residues 70 and 109, as predicted (Fig. S1b); this means that the 66–70 upstream region must be located in the periplasm. The anomalous behavior of fusions Val66 and Ala70 suggests that the 50–70 aa region may not be free in the periplasmic space, but probably in close association with the membrane, or might be subject to a reorientation process, but in a location that inhibits reporter enzyme activity. Fusions at residues Gly109, Gly133, Lys157, and Tyr177 yielded alternating low and high PhoA activities (Fig. 1c), indicating that these regions have corresponding alternate cytoplasmic and periplasmic locations; this location was confirmed by fusions Gly109, Gly133, and Lys157 also yielding alternate high and low LacZ activities (Fig. 1c). The topology of this region, which spans the last four TMSs of Chr3C, was in complete agreement with prediction models (Fig. S1b). Together, these results suggested a topology of five TMSs for Chr3C, with the N-terminal end in the cytoplasm and the C-terminal end in the periplasm (Fig. 1d).
In conclusion, membrane topology of the B. subtilis Chr3N/Chr3C homologous pair, as determined by translational fusions, consists of five TMSs in antiparallel orientation, with the N-terminal end of Chr3N located in the periplasm and the N terminus of Chr3C located in the cytoplasm (Fig. 1b and d).
Sequence analysis of short-chain CHR proteins
Eighty-two amino acid sequences, retrieved during Blastp searches at the UniProt site, were identified as members of the short-chain CHR3 subfamily (orthologous Chr3N/Chr3C) by phylogenetic analyses with the mega5 software. All chr3N/chr3C genes found are organized as tandem pairs and belong mainly to bacteria from the phylum Firmicutes (Bacillales; 76 protein sequences) and the γ-proteobacteria (Oceanospirillales; six protein sequences) group. Table S2 shows all Chr3N/Chr3C amino acid sequences studied in this work.
Membrane topology prediction of short-chain CHR proteins
A multiple protein sequence alignment was constructed with the 82 orthologous Chr3N/Chr3C sequences. Kyte-Doolittle hydropathic profiles, von Heijne transmembrane profiles, and free energy (ΔGapp) for membrane insertion of potential transmembrane helices were calculated for each sequence and are shown in Fig. S1a. Profiles for Chr3N and Chr3C are very similar, suggesting that both types of proteins possess the same number of TMSs. Figure S1a shows five evident local minima of calculated ΔGapp values that represent candidate TMSs (shaded areas). Additional local minima weakly supported are indicated by empty areas. As expected, these local minima corresponded with local maxima of hydrophobicity, supporting the existence of the abovementioned putative TMSs.
ΔG prediction server v1.0 (Hessa et al., 2007) recognized a range from three to six TMSs for each identified Chr3N/Chr3C protein sequences. Thus, TMS3 and TMS4 were recognized, with no exceptions, in all short-chain CHR3 subfamily members; TMS5 and TMS6 were predicted in the majority of analyzed Chr3N/Chr3C sequences, and TMS1 was recognized in all of Chr3C sequences and in the majority of Chr3N sequences (Table 1). In contrast, TMS2 (indicated by empty areas in Fig. S1a) was recognized only in one Chr3N and in none Chr3C sequences (Table 1). These data agree with calculated values of average ΔGapp for membrane insertion of each of the six potential TM helices for Chr3N and Chr3C proteins (Table 1). Barely recognized TMS2 possesses the most positive ΔGapp value for membrane insertion, questioning their ability to insert into the membrane. Nevertheless, Table 1 shows additional positive ΔGapp values for other TMSs. It has been reported that a relatively large fraction of the TM helices in multi-spanning membrane proteins have ΔGapp values > 0 kcal mol−1 and are thus not expected to insert efficiently by themselves (Hessa et al., 2007); this suggests that those TM helices may depend on interactions with neighboring TM domains for proper partitioning into the membrane (White & von Heijne, 2008). Indeed, examples where membrane protein folding takes place even after translation in the ribosome–translocon complex have been described. Aquaporin 1, initially synthesized in the membrane with four TMSs, undergoes an internal reorientation to acquire its mature ‘six-spanning’ structure (Buck et al., 2007); and in cystic fibrosis transmembrane conductance regulator, TMS2 initiates translocation after TMS1 emerges from the ribosome and subsequently directs TMS1 translocation to span the membrane in a post-translational event (Sadlish & Skach, 2004).
Table 1. Number of recognized TMSs and average ΔGapp (kcal/mol) for potential membrane insertions predicted for Chr3N and Chr3C proteins
Protein sequences (41 Chr3N and 41 Chr3C) were analyzed using the ΔG prediction server v1.0 (Hessa et al., 2007) for each of the six potential TMSs (local minima of calculated ΔGapp values).
Average ΔGapp (± standard error mean) for membrane insertion was calculated using the ΔG prediction server v1.0 and/or Membrane Protein Explorer (Snider et al., 2009).
topcons algorithm initially predicted a topology model with six TMSs for each identified Chr3C sequence. In contrast, topcons predictions for Chr3N sequences yielded topologies with five (39%) or six TMSs (61%). However, considering the dubious existence of TMS2, it is clear that prediction algorithms need additional experimental data to resolve between five- and six-TMS models. Constraining topcons predictions with the C-terminal locations of Chr3N/Chr3C yielded 41 topological models (one per each sequence), from which 38 were structures with five TMSs, and 35 of them corresponded to the model illustrated in Fig. S1b. Predictions for Chr3N proteins yielded, with no exception, a topology structure with five TMSs as that shown in Fig. S1b. Experimental C-terminal location was used for constraint predictions because positive ΔGapp values for membrane insertion of some TMSs (see Table 1) suggested that proper partitioning of Chr3N/Chr3C into the membrane may depend on interactions with neighboring TM helices and therefore may undergo internal reorientation(s) to acquire its mature structure. Thus, overall constrained-topcons results support the existence of a topological structure with five TMSs, and the absence of TMS2, in both Chr3N and Chr3C proteins.
Because constrained-topcons results also suggested that the Chr3N/Chr3C protein pair possesses antiparallel orientation in the membrane (Fig. S1b), distribution of positively charged amino acid residues in hydrophilic regions was analyzed as this parameter has been established as the most important determinant of TMS orientation (von Heijne, 1992). Distribution patterns of arginine/lysine residues in hydrophilic loops of selected Chr3N and Chr3C proteins revealed that predicted inside loops possess a higher (K + R) content than do periplasmic loops (Fig. 2; see also Fig. S2 for a complete analysis). This opposite distribution is compatible with the antiparallel arrangement of Chr3N/Chr3C shown in Fig. 1 for the B. subtilis protein pair and in Fig. S1b for the short-chain CHR protein family. The loops in Fig. S1b also show the average of positively charged residues (K + R)/loop per sequence, calculated from the complete alignment with 82 Chr3N/Chr3C sequences (Fig. S2). Thus, for both Chr3N and Chr3C, all abovementioned data point out to an antiparallel topology structure with five TMSs.
The monodomain short-chain CHR family belongs to the CHR superfamily of transporters (Díaz-Pérez et al., 2007) and is constituted by polypeptide pairs of about 200 aa each. The only short-chain CHR protein member whose function has been experimentally established is the B. subtilis Chr3N/Chr3C transporter pair, which confers resistance to chromate by the active efflux of chromate ions from the cell cytoplasm (Díaz-Magaña et al., 2009). Expression of both Chr3N and Chr3C proteins was found to be necessary for chromate resistance (Díaz-Magaña et al., 2009).
Díaz-Pérez et al. (2007) proposed that short-chain CHR protein pairs possess opposite membrane orientation. However, the number of TMSs in short-chain CHR proteins remained uncertain. It is interesting to observe that membrane topology prediction with the topcons algorithm initially yielded topology models with six TMSs for Chr3C, and with five or six TMSs for Chr3N proteins. However, constraining topcons prediction with the experimentally determined location of C-terminal yielded five-TMS topology models with opposite orientations for Chr3N and Chr3C proteins. This clearly shows that predicted models can be improved by providing just a little additional experimental data.
Results obtained with translational fusions indicated a membrane topology of five TMSs for both Chr3N and Chr3C (Fig. 1b and d). A previous topology model suggested weak hydrophobic regions for predicted TMS2, involving residues 50–70 in both Chr3N and Chr3C, giving rise to a six-TMS topology. A vestige of this region is probably still present in Chr3C and generates an α helix that is probably unable to span the lipid bilayer and may be instead located in the periphery of the periplasmic side of the membrane (Fig. 1d). Amino acid sequences in the large loops between TMS1 and TMS2 in both Chr3N and Chr3C show high identity and similarity (53% and 89%, respectively, in a 45-residue span), but a clear difference in positively charged residues content (six in Chr3N vs. two in Chr3C). These results support a distinct location of these hydrophilic regions. Data from PhoA/LacZ fusions in the Corynebacterium glutamicum LysE transporter led to removal of predicted TMS2, shifting from a six-TMS predicted model to a five-TMS predicted structure (Vrljic et al., 1999).
Membrane topology of Chr3N and Chr3C is antiparallel. The C-terminal end of Chr3N is located in the cytoplasm, whereas the C terminus of Chr3C lies in the periplasm (Fig. 1b and d). Jiménez-Mejía et al. (2006) reported a 13-TMS topology for P. aeruginosa ChrA protein, a member of the long-chain CHR family of the CHR superfamily. The two homologous halves of ChrA, formed by six TMSs each, displayed antiparallel membrane topology between them. It was proposed that this structure arose from the duplication of an equally oriented six-TMS ancestral protein domain followed by insertion of a central TMS (TMS7); this insertion might have caused the repeated domains to adopt the opposite orientation in a native parallel structure (Jiménez-Mejía et al., 2006).
Topologic inversion of halves of membrane proteins has been widely reported and is considered a common evolutionary process for these polypeptides (Ichihara et al., 2004; Rapp et al., 2006). It was proposed that membrane proteins with two antiparallel domains arose from ancestral monodomain proteins with dual topology (Rapp et al., 2006), that is, proteins that may insert into the membrane in either orientation (a ‘flip-flopping’ protein; Bowie, 2006). This dual topology ancestor may form homodimers displaying opposite orientation in the membrane. Gene duplication followed by sequence divergence would result in heterodimeric proteins with subunits of fixed but opposite orientation. Experimental evidence supporting this evolutionary pathway has been obtained from the analysis of proteins of the small multidrug resistance (SMR) family (reviewed in Bay et al., 2008). Antiparallel arrangement of E. coli homodimeric EmrE transporter has been widely reported (see Chen et al., 2007), although a parallel structure has also been claimed (Steiner-Mordoch et al., 2008). Another SMR family member, the EbrAB protein pair, has also been assigned antiparallel membrane topology (Kikukawa et al., 2007). Closely homologous proteins RnfA and RnfE from E. coli (Saaf et al., 1999) and NqrD and NqrE from Vibrio cholerae (Duffy & Barquera, 2006), both pairs being NADH-oxidoreductases constituted by six-TMS monomers, showed a completely opposite membrane topology. Members of several 10-TMS transporter families are also constituted by 2 five-TMS repeat units arranged in opposite membrane orientations (Saier, 2003; Lolkema et al., 2005). Aquaporins (Murata et al., 2000), ClC chloride channels (Dutzler et al., 2002), AmtB ammonia transporters (Khademi et al., 2004), and members of the DUF606 family of bacterial transporters (Lolkema et al., 2008) are all additional examples of proteins composed of two repeated halves with opposite membrane orientations. Indeed, the antiparallel domain organization is observed more frequently in the 3D structures of membrane proteins than the parallel domain organization (Lolkema et al., 2008).
Inverted membrane topology might suggest that the N- and C-terminal halves of a transporter carry out different functions. In agreement with this hypothesis, members of the major facilitator superfamily show higher sequence similarity among their N-terminal halves than at their C-terminal moieties; it was proposed that the N-terminal half of these carriers is essential for energization of transport, whereas the C-terminal half is involved in substrate specificity (Paulsen et al., 1996). Also, the finding of successive genes encoding N- and C-terminal domains of a full-length CHR protein suggests a distinct function for each protein half (Nies et al., 1998). Random mutagenesis of the P. aeruginosa chrA gene, selecting for mutants that lost chromate resistance, revealed that most essential residues are located at the amino half of the protein (Aguilera et al., 2004). Moreover, phylogenetic analysis showed that sequences of N-terminal halves in 77 putative ChrA homologues are significantly more conserved than those from C-terminal domains (Díaz-Pérez et al., 2007). These data further suggest that the two halves of Chr3N/C proteins have different roles in their function as chromate transporters.
It has been suggested that inverted topology in membrane transporters may be important for their function because it allows the arrangement of two conformational states (inward and outward) in a symmetric form with respect to both sides of the membrane, because of the structural symmetry of each inverted repeat domain (Forrest & Rudnick, 2009; Radestock & Forrest, 2011). Moreover, it was proposed that inverted topology in small heterodimeric transporters, with fixed but opposite membrane topology, may increase the stability of each monomer in the membrane by allowing formation of stable and functional heterodimers (Kolbusz et al., 2010).
This work was supported by grants from Coordinación de Investigación Científica (UMSNH; 2.6), CONACYT (México; 79190), and Dirección General de Asuntos del Personal Académico (UNAM; IN208510). R.M.-V. and G.R.-C. were supported by graduate and undergraduate student fellowships, respectively, from CONACYT.