Investigations of Dimethylglycine, Glycine Betaine, and Ectoine Uptake by a Betaine-Carnitine-Choline Transporter Family Transporter with Diverse Substrate Specificity in Vibrio Species

Vibrio parahaemolyticus possesses at least six osmolyte transporters, which allow the bacterium to adapt to high-salinity conditions. In this study, we identified several additional osmolytes that were utilized by V. parahaemolyticus. We demonstrated that the compound DMG, which is present in the marine environment, was a highly effective osmolyte for Vibrio species. We determined that DMG is transported via BCCT family carriers, which have not been shown previously to take up this compound. BccT1 was a carrier for GB, DMG, and ectoine, and we identified the amino acid residues essential for the coordination of these compounds. The data suggest that for BccT1, GB is more easily accommodated than ectoine in the transporter binding pocket.

ABSTRACT

Fluctuations in osmolarity are one of the most prevalent stresses to which bacteria must adapt, both hypo- and hyperosmotic conditions. Most bacteria cope with high osmolarity by accumulating compatible solutes (osmolytes) in the cytoplasm to maintain the turgor pressure of the cell. Vibrio parahaemolyticus, a halophile, utilizes at least six compatible solute transporters for the uptake of osmolytes: two ABC family ProU transporters and four betaine-carnitine-choline transporter (BCCT) family transporters. The full range of compatible solutes transported by this species has yet to be determined. Using an osmolyte phenotypic microarray plate for growth analyses, we expanded the known osmolytes used by V. parahaemolyticus to include N,N-dimethylglycine (DMG), among others. Growth pattern analysis of four triple-bccT mutants, possessing only one functional BCCT, indicated that BccT1 (VP1456), BccT2 (VP1723), and BccT3 (VP1905) transported DMG. BccT1 was unusual in that it could take up both compounds with methylated head groups (glycine betaine [GB], choline, and DMG) and cyclic compounds (ectoine and proline). Bioinformatics analysis identified the four coordinating amino acid residues for GB in the BccT1 protein. In silico modeling analysis demonstrated that GB, DMG, and ectoine docked in the same binding pocket in BccT1. Using site-directed mutagenesis, we showed that a strain with all four residues mutated resulted in the loss of uptake of GB, DMG, and ectoine. We showed that three of the four residues were essential for ectoine uptake, whereas only one of the residues was important for GB uptake. Overall, we have demonstrated that DMG is a highly effective compatible solute for Vibrio species and have elucidated the amino acid residues in BccT1 that are important for the coordination of GB, DMG, and ectoine transport.

IMPORTANCE Vibrio parahaemolyticus possesses at least six osmolyte transporters, which allow the bacterium to adapt to high-salinity conditions. In this study, we identified several additional osmolytes that were utilized by V. parahaemolyticus. We demonstrated that the compound DMG, which is present in the marine environment, was a highly effective osmolyte for Vibrio species. We determined that DMG is transported via BCCT family carriers, which have not been shown previously to take up this compound. BccT1 was a carrier for GB, DMG, and ectoine, and we identified the amino acid residues essential for the coordination of these compounds. The data suggest that for BccT1, GB is more easily accommodated than ectoine in the transporter binding pocket.

INTRODUCTION

In order to grow in high-osmolarity environments, bacteria accumulate compounds called compatible solutes (osmolytes) within the cytoplasm of the cell, via either uptake or biosynthesis (1–4). These compounds balance the internal osmolarity with that of the environment and maintain the turgor pressure of the cell (3, 5, 6). Osmolytes also protect proteins, nucleic acids, and other vital molecular machinery by increasing the hydration shell around these molecules (7). Osmolytes fall into several classes of compounds, including sugars (trehalose), polyols (glycerol and mannitol), free amino acids (proline and glutamine), amino acid derivatives (ectoine [ect]), and quaternary amines (glycine betaine [GB], choline, and l-carnitine) (4, 5, 8–12).

Biosynthesis of compatible solutes is energetically costly, and therefore, bacteria encode compatible solute transporters to scavenge available osmolytes from the environment (12–14). Compatible solute transporters include members of the multicomponent ATP binding cassette (ABC) family, such as ProU (proVWX) in Escherichia coli, OpuC in Bacillus subtilis, and OpuC (proVWX) in Pseudomonas syringae (15–19), and the single-component betaine-carnitine-choline transporter (BCCT) family, which are Na⁺ or H⁺ dependent. Members of the BCCT family include BetT in E. coli, which transports choline with high affinity, and GB transporters in B. subtilis (OpuD) and Corynebacterium glutamicum (BetP), among many others (20–23).

BCCTs are energized by sodium- or proton-motive force symport and are organized into 12 transmembrane (TM) segments (20, 22, 24, 25). Aromatic residues found in TM4 and TM8 make up the GB binding pocket in BCCT family transporters examined to date. These residues are highly conserved in BCCTs that transport trimethylammonium compounds such as GB, l-carnitine, and γ-butyrobetaine (21, 24, 26, 27). An additional tryptophan residue is present in TM8 just outside the binding pocket and is thought to participate in the coordination of substrates during conformational changes that occur during transport (26).

Vibrio parahaemolyticus is a halophilic bacterium present in marine and estuarine environments in association with plankton, fish, and shellfish (28–33). Vibrio parahaemolyticus is the leading bacterial cause of seafood-related gastroenteritis worldwide, frequently associated with consuming raw or undercooked seafood (30, 34). Vibrio parahaemolyticus is a halophile that grows optimally in 3% NaCl (∼500 mM), which is approximately the salinity of seawater, but can grow at salinities of up to 9% NaCl (∼1.5 M) (35). Vibrio parahaemolyticus encounters a range of salinities in tidal habitats and possesses four BCCTs, encoded by bccT1 (VP1456), bccT2 (VP1723), bccT3 (VP1905), and bccT4 (VPA0356), and two ProUs, encoded by proVWX (ProU1) (VP1726 to VP1728) and proXWV (ProU2) (VPA1109 to VPA1111), for the uptake of compatible solutes (36). In addition to compatible solute transporters, V. parahaemolyticus also possesses biosynthesis systems for the compatible solutes ectoine, ectABC-asp_ect (VP1719 to VP1722), and GB, betIBA-proXWV (VPA1112 to VPA1114), whose expression is induced at high salinity (36). It was demonstrated that the expression of these biosynthesis systems in Vibrio species was under tight regulation, controlled by the quorum sensing regulators OpaR and AphA as well as CosR, a global regulator of the osmotic stress response (37–40). Previously, we reported that in V. parahaemolyticus, bccT1, bccT3, bccT4, and both proU operons are responsive to salinity (40, 41). BccT1 had the broadest substrate specificity, transporting GB, choline, proline, and ectoine, while BccT2 and BccT3 transport GB, choline, and proline, and BccT4 transports only choline and proline (Fig. 1) (41). Interestingly, the number of osmolyte transporters present among Vibrio species varies, which suggested differences in the osmolytes utilized and osmotolerance (36). It was shown that Vibrio alginolyticus contained four BCCTs and two ProUs, Vibrio harveyi and Vibrio splendidus possessed six BCCTs and two ProUs, and Vibrio vulnificus possessed a single ProU and a BccT3 homolog, whereas Vibrio cholerae possessed only a BccT3 homolog (36).

FIG 1.

(A) BCCT transporters present in V. parahaemolyticus RIMD2210633 and their known substrates. bccT1, bccT3, and bccT4 are induced by high salinity. (B and C) Structures of BCCT substrates with methylated headgroups highlighted in blue boxes (B) or cyclic compounds (C). AA, amino acids.

In this study, we examined a range of osmolytes utilized by V. parahaemolyticus using an osmolyte phenotypic microarray plate, which identified 14 potential osmolytes. We determined the ability of several Vibrio species to utilize dimethylglycine (DMG), one of the most effective osmolytes that we identified and a compound not previously shown to be an osmolyte in Vibrio. We examined V. parahaemolyticus transport of DMG using several osmolyte transporter mutants and showed that BCCT carriers were required, representing a new transporter family for the uptake of DMG. In V. parahaemolyticus, BccT1 has the broadest substrate uptake ability in terms of the number and diversity of compounds (methylated head groups and cyclic compounds). Our in silico modeling analysis demonstrated that GB, DMG, and ectoine docked in the same binding pocket in BccT1. We investigated via mutagenesis and functional complementation the amino acid residues that are required for the coordination of GB, DMG, and ectoine. This analysis describes for the first time the residues that coordinate DMG and ectoine in a BCCT family transporter.

RESULTS

V. parahaemolyticus can utilize a wide range of compatible solutes.

To determine the range of compatible solutes that can be utilized by V. parahaemolyticus, a Biolog 96-well PM9 osmolyte phenotypic microarray plate was used. Growth analyses were performed using a V. parahaemolyticus ΔectB deletion mutant, which is unable to synthesize ectoine de novo and therefore has a growth defect at high salinity in the absence of exogenous osmolytes (35, 36). The ΔectB mutant was grown in 23 unique osmolytes, 14 of which rescued the growth of the ΔectB mutant (see Fig. S1 in the supplemental material). Previously unrecognized osmolytes for this species included N,N-dimethylglycine (DMG), γ-amino-N-butyric acid (GABA), trimethylamine-N-oxide (TMAO), glutathione, dimethylsulfoniopropionate, morpholinepropanesulfonic acid (MOPS), creatine, N-acetyl l-glutamine, and octopine, in addition to those already known to provide osmoprotection, such as trehalose, β-glutamic acid, GB, ectoine, and l-proline (Fig. S1).

To confirm the data from the phenotypic microarray analysis, we performed growth analyses with wild-type (WT) V. parahaemolyticus RIMD2210633 and the ΔectB mutant in the presence of the compatible solute DMG, GABA, or TMAO in M9 minimal medium supplemented with glucose and 6% NaCl (M9G 6% NaCl) (Fig. 2). In the wild-type strain, in the absence of DMG, there was an ∼4-h lag phase with a growth rate of 0.04 h⁻¹, whereas in the presence of DMG, the lag phase was <1 h, and cells had a higher growth rate (0.06 h⁻¹ in the presence of DMG), indicating that DMG is a very effective compatible solute for V. parahaemolyticus (Fig. 2A). The growth of the ΔectB mutant was rescued by the addition of exogenous DMG, with a lag phase of less than 1 h (Fig. 2A). The growth of the wild-type strain in the presence of GABA had a slightly increased growth rate of 0.045 h⁻¹ (Fig. 2B). The growth of the wild-type strain in the presence of TMAO was not changed, with a growth rate of 0.04 h⁻¹ (Fig. 2C). However, the presence of exogenous GABA or TMAO rescued the growth of the ΔectB mutant, which confirmed that V. parahaemolyticus can utilize these as compatible solutes (Fig. 2). The ΔectB mutant was rescued to a greater extent with exogenous DMG, with a growth rate of 0.07 h⁻¹, than with GABA (0.02 h⁻¹) or TMAO (0.01 h⁻¹) (Fig. 2). In addition, growth analyses of the wild type in M9G 6% NaCl supplemented with either GB or ectoine confirmed that both are effective compatible solutes (Fig. S2) (35).

FIG 2.

Growth analyses of the V. parahaemolyticus RIMD2210633 wild type (WT) and a ΔectB mutant in M9G 6% NaCl. Growth medium was supplemented with DMG (A), GABA (B), or TMAO (C). The optical density (OD₅₉₅) was measured every hour for 24 h; the means and standard errors from at least two biological replicates are shown.

DMG is an effective compatible solute for Vibrio species.

To investigate whether DMG can also act as an osmolyte in other Vibrio species, we grew four species representing divergent clades of Vibrio in M9G plus 4% NaCl or 5% NaCl with and without DMG at 37°C for 24 h. Growths of V. harveyi 393, V. cholerae N16961, V. vulnificus YJ016, and V. fluvialis ATCC 33809 were all rescued by exogenous DMG (Fig. S3A to D). In the absence of DMG, V. harveyi grew with a 9-h lag phase, while this strain grew with a 3-h lag phase and an increased growth rate, from 0.04 h⁻¹ to 0.06 h⁻¹, in the presence of DMG (Fig. S3A). In the presence of DMG, the growth rate of V. cholerae was increased from 0.02 h⁻¹ to 0.04 h⁻¹ (Fig. S3B). V. vulnificus is unable to grow in M9G 4% NaCl in the absence of DMG, but growth was rescued in the presence of DMG with a 10-h lag phase and a growth rate of 0.04 h⁻¹ (Fig. S3C). Vibrio fluvialis grown in the presence of DMG also had a reduced lag phase (Fig. S3D). To demonstrate that the changes in the Vibrio species growth patterns were not due to the catabolism of DMG, we grew all strains in M9 medium supplemented with DMG as the sole carbon source. None of the species strains tested grew in the presence of DMG (Fig. S4). Thus, DMG is a bona fide compatible solute for these Vibrio species strains examined.

BCCTs are responsible for the transport of DMG.

We examined whether the BCCTs in V. parahaemolyticus were required for the transport of DMG, and to accomplish this, we used a bccT-null mutant (quadruple ΔbccT1-ΔbccT3-ΔbccT4-ΔbccT2 mutant). We grew the wild type and the bccT-null mutant in M9G 6% NaCl with and without DMG at 37°C for 24 h (Fig. 3). The bccT-null mutant did not exhibit a reduced lag phase or an increased growth rate in the presence of DMG, while the wild-type strain grew with a reduced lag phase of <1 h (Fig. 3), which indicated that a BCCT carrier is required for the transport of DMG. Although V. parahaemolyticus encodes two ABC family compatible solute transporters, ProU1 and ProU2, which are present in the bccT-null mutant, these do not appear to be involved in DMG uptake.

FIG 3.

Growth analysis of V. parahaemolyticus RIMD2210633 wild-type and bccT-null (ΔbccT1-ΔbccT3-ΔbccT4-ΔbccT2) mutant strains in M9G 6% NaCl with and without DMG. The optical density (OD₅₉₅) was measured every hour for 24 h; means and standard errors from at least two biological replicates are displayed.

To determine which of the four BCCT family carriers present in V. parahaemolyticus transports DMG, a set of four triple-bccT mutants, each possessing a single functional bccT gene, was utilized in growth assays. The triple ΔbccT2-ΔbccT3-ΔbccT4 mutant, which contains only bccT1, had a slightly reduced lag phase and a slightly higher growth rate through exponential phase in M6G 6% NaCl supplemented with DMG (Fig. 4A). This suggested that BccT1 transported DMG with low efficiency. In contrast, the triple ΔbccT1-ΔbccT3-ΔbccT4 mutant, which contains only bccT2, had a reduction in lag phase similar to that of the wild type (Fig. 4B). This suggested that BccT2 transported DMG as efficiently as the wild-type strain. The triple ΔbccT1-ΔbccT2-ΔbccT4 mutant, which contains only bccT3, grew similarly to the bccT1-only strain, with a marginally reduced lag phase, and grew marginally better through exponential phase in the presence of DMG (Fig. 4C). This suggested that DMG is transported by BccT3 with low efficiency. The triple ΔbccT1-ΔbccT2-ΔbccT3 mutant showed no difference in growth in the absence or presence of DMG, which indicated that BccT4 does not transport DMG into the cell (Fig. 4D).

FIG 4.

Growth analysis of V. parahaemolyticus RIMD2210633 wild-type and triple mutant ΔbccT2-ΔbccT3-ΔbccT4 (A), ΔbccT1-ΔbccT3-ΔbccT4 (B), ΔbccT1-ΔbccT2-ΔbccT4 (C), or ΔbccT1-ΔbccT2-ΔbccT3 (D) strains in M9G 6% NaCl with and without DMG. The optical density (OD₅₉₅) was measured every hour for 24 h; means and standard errors from at least two biological replicates are displayed.

In order to examine DMG uptake by the BCCT family transporters further, we used E. coli MKH13, a mutant strain that has deletions in betIBA-betT, proU, proP, and putP and cannot grow in M9G 4% NaCl (42). This strain has been utilized successfully in previous studies to assess transporter function in several species (26, 41, 43). We cloned each of the bccT genes into an expression plasmid, used each construct to complement E. coli MKH13, and examined growth in M9G 4% NaCl supplemented with DMG (Fig. 5). Escherichia coli MKH13 strains complemented with bccT1, bccT2, and bccT3 grew in the presence of DMG, confirming that these three are carriers for DMG (Fig. 5). The bccT4 complemented strain and the empty expression plasmid did not grow, which indicated that BccT4 cannot take up DMG (Fig. 5).

FIG 5.

E. coli strain MKH13 complemented with the empty vector, bccT1 (pBAVP1456), bccT2 (pBAVP1723), bccT3 (pBAVP1905), or bccT4 (pBAVPA0356) and grown for 24 h with DMG. The final optical density (OD₅₉₅) was compared to that of a strain harboring pBAD33. Means and standard errors from at least two biological replicates are shown. Statistics were calculated using Student’s t test (*, P < 0.05; ***, P < 0.001).

BccT1 sequence homology to structurally characterized BCCTs.

From a structure/function standpoint, BccT1 is of particular interest as it can transport GB, DMG, and ectoine, among other substrates (Fig. S5). While GB and DMG have similar structures with methylated head groups, ectoine is a cyclic compound and may require coordination by different amino acid residues in the BCCT transporter (24). Furthermore, BCCT family transporters of ectoine typically do not possess the conserved aromatic residues located in TM4 and TM8 that coordinate GB binding (24). The hydropathy profile of BccT1 was aligned with that of the BCCT transporter BetP from C. glutamicum (CgBetP), whose structure has been studied extensively. We found that BccT1 possessed 12 TM segments (TM1 to TM12) along with N- and C-terminal tail extensions. BccT1 shared matched positions with 89% of residues in CgBetP, which indicated that the structures were highly conserved (Fig. S6). In CgBetP, the residues that form the GB binding pocket are Trp189, Trp194, and Tyr197 in TM4 and one residue located in TM8, Trp374. An additional residue located in TM8 below the binding pocket (Trp377) is thought to be involved in substrate coordination during conformational changes (26). In addition, we aligned the C. glutamicum BCCT family transporters LcoP (CgLcoP) and EctP (CgEctP), which were reported to take up GB and ectoine (44, 45). CgBetP, CgLcoP, and BccT1 possessed amino acids identical to the residues corresponding to TM4 and TM8 described above (Fig. S7).

Structural modeling of BccT1.

The structural models of BccT1 consist of 11 transmembrane helices (TM2 to TM12) and a periplasmic helix (H7) (Fig. 6A). The first 108 amino acids of BccT1, including TM1, could not be modeled because of their poor sequence similarity with CgBetP. Nonetheless, the BccT1 models are very similar to the CgBetP structure, with root mean square deviations (RMSDs) of 0.26 Å (399 C_α atoms) and 0.22 Å (391 C_α atoms) for closed-state (C_cS) and open substrate-bound-state (C_iS) models, respectively. The residues involved in the central binding pocket, cytoplasmic and periplasmic gates, as well as the rest of the substrate pathway are mostly conserved in BccT1. More specifically, a comparison of BccT1 models with the GB-bound structure of CgBetP showed that all the substrate binding residues are conserved and located in TM4 and TM8 (Fig. S7).

FIG 6.

Predicted overall structure of BccT1 and its active-site interactions with different substrates. (A) Side view of the overall structure of BccT1. α-Helices are depicted by cylinders, and the directionality of the polypeptide chain is shown by a color gradient from blue (N terminus) to red (C terminus). Eleven transmembrane helices (TM2 to TM12) and a periplasmic helix (H7) make the core of the structure. Two small helices, IH1 and EH2, connect TM4-TM5 and TM9-TM10, respectively. (B to D) Interactions of the active-site residues of BccT1 with the substrates. Interacting residues of the BccT1 C_cS state with docked glycine betaine (B) and DMG (C) as well as those in the CiS state with docked ectoine (D) are represented as sticks. Substrates and BccT1 residues are shown in orange and gray, respectively. Dashed lines show hydrogen-bonding distances between BccT1 residues and various atoms in the substrates. Illustrations were prepared using PyMOL (PyMOL Molecular Graphics System, version 2.0; Schrödinger, LLC).

The closed state (C_cS) represents a transition state between outward- and inward-facing open states where the substrate is bound in a central cavity with both substrate entry and exit ports occluded. Unlike open states, the substrate makes optimal interactions with the active-site residues in the C_cS state (46); therefore, the C_cS model of BccT1 was used for our substrate docking studies. The GB-docked BccT1 model showed a binding free energy change (ΔG) of −4.8 kcal/mol (Table 1). Briefly, the aromatic rings of Trp380, Trp381, and Trp384 form a small hydrophobic pocket that binds GB in this BccT1 model (Table 1 and Fig. 6B). The quaternary amine group of GB binds to this pocket by cation-π and van der Waals interactions. Other aromatic residues, Trp203, Trp208, and Tyr211, of BccT1 also contribute to the hydrophobic interactions with GB. The GB carboxylic group interacts with the peptide backbones of Ile160 and Gly161 as well as the side chain of Ser266 via hydrogen bonding (Fig. 6B).

TABLE 1.

Results of the docking study showing the free energy change upon substrate binding and BccT1 residues mediating interactions with each substrate

Substrate	Binding free energy (kcal/mol)	Residues involved in hydrogen-bonding interactions	Residues involved in van der Waals interactions
Glycine betaine	−4.8	Ile160 Gly161, Ser266	Trp203, Trp208, Tyr211, Trp380, Trp381, Trp384
DMG	−4.4	Ile160 Gly161, Ser266	Trp203, Trp208, Tyr211, Trp380, Trp381, Trp384
Ectoine	−6.1	Trp203, Tyr211	Trp208, Trp380, Trp381, Trp384

Docking analysis of DMG and ectoine binding residues in BccT1.

Next, we sought to determine the DMG and ectoine binding residues in BccT1 by docking. Docking of DMG to the BccT1 C_cS model (Fig. 6C) predicted a mode of interaction identical to that of GB with a favorable ΔG of −4.4 kcal/mol (Table 1). This was expected since DMG and GB have very similar chemical structures.

To identify the mode of ectoine binding to BccT1, ectoine was first docked to the active site of the C_cS model of BccT1. However, the ΔG of ectoine binding to the C_cS model of BccT1 was too low (−2.8 kcal/mol), likely because of the spatially constrained fitting of the large ectoine ring into the smaller hydrophobic pocket comprised of the aromatic rings of Trp380, Trp381, and Trp384. Therefore, we turned our attention to the C_iS model of BccT1 ectoine docking, keeping all other docking parameters constant. The substrate binding site in the C_iS state differs from that in the C_cS state in the relative orientations of the Trp380 and Trp381 side chains remodeling the hydrophobic pocket. More specifically, during the C_cS-to-C_iS transition, aromatic rings of Trp380 and Trp381 are known to flip 90°, opening up the hydrophobic pocket (46). Interestingly, ectoine docking to the C_iS state of BccT1 was accompanied by a significant ΔG of −6.1 kcal/mol (Table 1). The ectoine binding in this BccT1 state is stabilized by hydrogen-bonding interactions between the ectoine carboxyl group and the peptide backbone of Ala157 as well as the side chain of Tyr211 (Fig. 6D). Furthermore, the ectoine ring nitrogens are involved in hydrogen bonding with Tyr211 and the peptide backbone of Trp203. The bound ectoine also shows van der Waals interactions with the aromatic residues Trp203, Trp208, Trp381, and Trp384 (Fig. 6D).

Site-directed mutagenesis of BccT1 uncovers the ectoine binding pocket.

Our structural modeling and docking experiments predicted BccT1 residues that are likely involved in the binding and transport of GB, DMG, and ectoine. Therefore, we selected these residues in BccT1, namely, Trp203, Trp208, Tyr211, and Trp384, for a mutagenesis study. We utilized functional complementation of E. coli MKH13 as a readout for the uptake of GB, DMG, and ectoine by BccT1 with the following single-amino-acid substitutions: Trp203Cys, Trp208Leu, Tyr211Leu, and Trp384Leu. We also tested a BccT1 mutant with all four of these residues replaced by alanine, which is chemically inert and possesses a nonbulky methyl functional group. The replacement of Trp203 resulted in reduced growth of the E. coli MKH13 strain in the presence of GB, while strains harboring the other three single-replacement mutants Trp208Leu, Tyr211Leu, and Trp384Leu grew similarly to a strain expressing WT BccT1 (Fig. 7A). This indicates that Trp203 is important, but not essential, for the uptake of GB by BccT1. In the absence of each of the other three residues 208, 211, and 384, GB may be accommodated by an alternate residue. The replacement of all four residues with alanine resulted in no growth of the E. coli MKH13 strain (Fig. 7A). This indicated that these four residues are involved in the coordination of GB in BccT1.

FIG 7.

Functional complementation of E. coli MKH13 with BccT1 (WT) and five BccT1 mutants grown in GB (A), DMG (B), or ectoine (C). Strains were grown for 24 h, and the final optical density (OD₅₉₅) was compared to that of the strain expressing wild-type BccT1 (pBAVP1456). Means and standard errors from at least two biological replicates are shown. Statistics were calculated using Student’s t test (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001).

The replacement of Trp203 or Trp384 resulted in no uptake of DMG, as strains expressing these mutants did not grow (Fig. 7B). The replacement of Tyr211 resulted in a reduced ability to take up DMG, as evidenced by the growth of this strain to a lower final optical density (OD) (Fig. 7B). The replacement of all four residues resulted in no growth of E. coli MKH13, which indicated that these residues make up the binding pocket for DMG (Fig. 7B).

The replacement of residue Trp203, Tyr211, or Trp384 individually was sufficient to completely abolish the uptake of ectoine by E. coli MKH13, as the strains expressing these mutants were unable to grow (Fig. 7C). However, the replacement of Trp208 did not result in a statistically significant difference in growth from that of a WT BccT1-expressing strain, indicating that Trp208 is not required for the uptake of ectoine (Fig. 7C). The replacement of all four residues with alanine also resulted in the abrogation of ectoine transport (Fig. 7C). Together, these results indicated that DMG and ectoine share a binding pocket with GB. The coordination of DMG requires Trp203 and Trp384, while the coordination of ectoine requires Trp203, Tyr211, and Trp384. These results suggest that the coordination of DMG and ectoine cannot be as easily accommodated by alternate residues as can GB.

Distribution of BccT1.

The BccT1 protein is 562 amino acids long and is present in all sequenced V. parahaemolyticus genomes (>800 genomes). A highly homologous protein is also present in all V. alginolyticus, V. antiquarius, and V. diabolicus strains (95% sequence identity); V. natriegens and V. nereis (90% sequence identity); as well as V. harveyi and V. campbellii sequenced strains (86% sequence identity) (Fig. S8). In general, phylogenetic analysis indicated that BccT1 is conserved within the Harveyi clade, and when present in other clades, it is present in all strains of each species. Overall, within the family Vibrionaceae, BccT1 is present in 35 Vibrio species and in 11 Photobacterium species (Fig. S8). It is of interest to note that in all species, the residues corresponding to Trp203, Trp208, Tyr211, and Trp384, which coordinate substrates, are conserved (data not shown), suggesting an ability to take up a range of osmolytes in these species.

DISCUSSION

Here, we demonstrated that the halophile V. parahaemolyticus can utilize at least 14 compatible solutes, which included DMG, GABA, and TMAO. The transporters BccT1, BccT2, and BccT3 were carriers for DMG, a highly effective osmoprotectant for V. parahaemolyticus, V. vulnificus, V. harveyi, V. cholerae, and V. fluvialis. DMG is the N-dimethyl derivative of glycine, while GB is the N-trimethyl derivative. Some halophilic bacteria were shown to accumulate DMG, an intermediate compound produced during de novo biosynthesis (47–55). DMG was found to be suitable for osmotic adaptation on its own, without modifications by bacteria (47). Additionally, DMG is an intermediate of aerobic GB catabolism and is available in the environment (56–60). None of the Vibrio species strains examined in this study could grow on DMG as a sole carbon and energy source. Several species have been shown to possess genes for the conversion of GB to DMG and the conversion of DMG to sarcosine, which can then be catabolized to glycine (58, 61–66). These genes have not been described within the family Vibrionaceae, and none of the species strains under study contained homologs of these genes.

DMG is known to be transported by major facilitator superfamily (MFS) transporters and ATP binding cassette (ABC) family transporters (56, 67). Here, we demonstrated that the BCCT family of transporters can also take up DMG. BccT1 and BccT2 transport DMG when expressed in a heterologous E. coli background (Fig. 5). However, in the native background, a V. parahaemolyticus strain expressing only BccT2 is fully rescued to wild-type levels by the presence of DMG, while a strain expressing BccT1 is only partially rescued (Fig. 4A and B). This difference may be explained by the fact that bccT2 is not induced by high salinity and has a basal level of transcription in the cell, whereas bccT1 expression is repressed under low-salinity conditions (40, 41). Therefore, in a strain containing only bccT2, expression is constitutively active at low salinity, allowing for more rapid adaptation to high salinity via the uptake of DMG, resulting in a reduced lag phase. In C. glutamicum, the transporter EctP is not induced by external osmolarity, and the EctP transport capacity is maximal at low osmolarity and may act as a rescue system that is available at low osmolarity, ensuring that cells can respond to osmotic stress (45). The same could be true of BccT2 in V. parahaemolyticus, acting as an immediate rescue system to scavenge available osmolytes in the event of osmotic stress.

DMG was not shown previously to be transported by the BCCT family of carriers, and the amino acid residues important for the uptake of DMG by a BCCT are unknown. Our in silico docking study showed that GB was coordinated by residues in the binding pocket of BccT1 (Fig. 6B) identical to those previously reported for CgBetP (26, 27). The coordination of DMG by BccT1 was identical to that of GB (Fig. 6C). Likewise, an ABC transporter, OpuAC, from B. subtilis has been reported to interact very similarly with GB (Protein Data Bank [PDB] ID 2B4L) and a sulfonium analog of DMG, dimethylsulfonioacetate (DMSA) (PDB ID 3CHG) (18, 19). Our mutagenesis studies demonstrated that GB, DMG, and ectoine are coordinated in the same binding pocket of BccT1, but the residues required for coordination are strict for DMG and ectoine, while GB may be accommodated by alternate residues in single-amino-acid mutants (Fig. 7). It was demonstrated previously in an ABC-type transporter that while the replacement of a single aromatic residue in the binding pocket results in decreased affinity of the protein for GB, the substrate is still coordinated with reasonable affinity. Only when any combination of two residues was mutated was transport completely abolished (68). We did not see a reduction in the ability of E. coli MKH13 to grow when only one residue is mutated, and the uptake of GB was completely abolished only when all four residues were mutated. This is likely due to the coordination of GB at alternate positions within the binding pocket that do not affect the overall growth ability of E. coli MKH13 but may affect the affinity of BccT1 for GB. Comparative protein analyses also demonstrated that in other Vibrio species with BccT1 homologs, the residues that coordinate GB, DMG, and ectoine are conserved. There is a very high percent identity shared with BccT1 among BCCT family transporters in these species, and therefore, the ability of these transporters to take up a broad range of substrates is most likely conserved.

All 14 substrates that V. parahaemolyticus was demonstrated to take up are available in the marine environment, including GB, choline, DMG, and ectoine (4, 57, 58, 69). Five of the six compatible solute transporters encoded by V. parahaemolyticus are induced by high salinity and collectively take up a broad range of compatible solutes (40, 41). Although there is redundancy in the compounds taken up by each BCCT, the ability to take up many different compatible solutes likely provides V. parahaemolyticus and other Vibrio species with a fitness advantage in the marine environment that allows them to thrive and grow optimally under high-salt conditions.

Members of the BCCT family are widespread among bacteria, present in both Gram-positive and Gram-negative bacteria as well as Archaea. For example, using the InterPro database (http://www.ebi.ac.uk/interpro/) accession no. IPR000060 for BCCT as a search, 23,000 BCCT proteins fall within the domain Bacteria, 604 BCCT proteins fall within the Archaea, and 78 BCCTs fall within the Eukaryota. Of the 604 Archaea representatives, 593 were from the Stenosarchaea group, of which 549 were within the Halobacteria, suggesting an important function in osmotolerance. Surprisingly, there have been no studies on BCCT function from representatives of Archaea or Eukaryota to date, with the exception of a recent study on the distribution of BCCTs in corals (70).

MATERIALS AND METHODS

Bacterial strains, media, and culture conditions.

All strains and plasmids used in this study are listed in Table 2. Vibrio parahaemolyticus strains were grown in either lysogeny broth (LB) (Fisher Scientific, Fair Lawn, NJ) with 3% (wt/vol) NaCl (LB3%) or M9 minimal medium (47.8 mM Na₂HPO₄, 22 mM KH₂PO₄, 18.7 mM NH₄Cl, 8.6 mM NaCl; Sigma-Aldrich) supplemented with 2 mM MgSO₄, 0.1 mM CaCl₂, and 20 mM glucose as the sole carbon source (M9G) and NaCl as indicated. Dimethylglycine (DMG) was used at a final concentration of 20 mM when supplied as a carbon source. E. coli strains were grown in either LB supplemented with 1% NaCl (LB1%) or M9G supplemented with 1% (wt/vol) NaCl (M9G 1% NaCl). All strains were grown at 37°C with aeration. Chloramphenicol (Cm) was added to the media at 25 μg/ml when necessary.

TABLE 2.

Strains and plasmids used in this study

Strain or plasmid	Genotype or description	Reference(s) or source
Strains
V. parahaemolyticus
RIMD2210633	O3:K6 clinical isolate; Strr	87, 88
RIMD2210633 ΔectB	RIMD2210633 ΔectB (VP1721) Strr	36
SOYBCCT124	RIMD2210633 ΔVP1456 ΔVP1723 ΔVPA0356 Strr	41
SOYBCCT123	RIMD2210633 ΔVP1456 ΔVP1723 ΔVP1905 Strr	41
SOYBCCT134	RIMD2210633 ΔVP1456 ΔVP1905 ΔVPA0356 Strr	41
SOYBCCT234	RIMD2210633 ΔVP1723 ΔVP1905 ΔVPA0356 Strr	41
SOYBCCT1342	RIMD2210633 ΔVP1456 ΔVP1723 ΔVP1905 ΔVPA0356 Strr	41
V. cholerae N16961	O1, El Tor strain, Bangladesh, clinical, 1975	89
V. vulnificus YJ016	Clinical isolate	90
V. fluvialis ATCC 33809	Synonym, NCTC 11327 or 606; clinical isolate	91
V. harveyi 393	Isolated from barramundi in Australia	92
Escherichia coli
DH5α λpir	Δlac pir	Thermo Fisher Scientific
β2155 λpir	ΔdapA::erm pir for bacterial conjugation	93
MKH13	MC4100 (ΔbetTIBA) Δ(putPA)101 Δ(proP)2 Δ(proU) Spr	42
Plasmids
pBAD33	Expression vector; araBAD promoter; Cmr; p15a origin	71
pBAVP1456	pBAD33 harboring full-length VP1456	This study
pBAVP1723	pBAD33 harboring full-length VP1723	41
pBAVP1905	pBAD33 harboring full-length VP1905	41
pBAVPA0356	pBAD33 harboring full-length VPA0356	This study
pBAVP1456W203C	pBAD33 harboring full-length VP1456 with Trp203Cys mutation	This study
pBAVP1456W208L	pBAD33 harboring full-length VP1456 with Trp208Lys mutation	This study
pBAVP1456Y211L	pBAD33 harboring full-length VP1456 with Tyr211Lys mutation	This study
pBAVP1456W377L	pBAD33 harboring full-length VP1456 with Trp377Lys mutation	This study
pBAVP1456-3	pBAD33 harboring full-length VP1456 with Trp203Ala, Trp208Ala, and Tyr211Ala mutations	This study
pBAVP1456-4	pBAVP1456-3 harboring full-length VP1456 with Trp377Ala mutation	This study

Growth analysis.

Vibrio parahaemolyticus and an in-frame deletion mutant of ectB were grown overnight in M9 minimal medium supplemented with 1% NaCl and 20 mM glucose as the sole carbon source. Cultures were subsequently diluted 1:50 into fresh medium and grown for 5 h. Cultures were pelleted, washed two times with 1× phosphate-buffered saline (PBS) to remove excess salt, and then diluted 1:50 into M9G, and 100 μl was then added to each well of a 96-well Biolog PM9 plate containing different osmolytes and/or salt concentrations (Biolog, Inc., Hayward, CA). The plates were incubated at 37°C with intermittent shaking in a Tecan Sunrise microplate reader, and the optical density at 595 nm (OD₅₉₅) was measured every hour for 24 h. The area under the curve (AUC) was calculated using Origin 2018 for the wild type (WT) and the ΔectB mutant. Statistics were calculated using Student’s t test; growth in 6% NaCl in the presence of a compatible solute was compared to growth in 6% NaCl with no exogenous compatible solutes.

For growth analysis with individual compatible solutes, wild-type (WT), ΔectB mutant, and quadruple ΔbccT1-ΔbccT3-ΔbccT4-ΔbccT2 mutant (bccT-null) strains were grown overnight in M9G 1% NaCl. Cultures were subsequently diluted 1:50 into fresh medium and grown for 5 h to late exponential phase. Exponential-phase cultures were then diluted 1:40 into 200 μl of M9G 6% NaCl with and without exogenous compatible solutes in a 96-well microplate and grown at 37°C with intermittent shaking for 24 h. The compatible solute N,N-dimethylglycine (DMG), trimethylamine-N-oxide (TMAO), γ-amino-N-butyric acid (GABA), glycine betaine (GB), or ectoine (ect) was added to a final concentration of 500 μM. Growth analysis with each compatible solute was repeated according to the above-described procedure with each of four triple-bccT mutants, ΔbccT2-ΔbccT3-ΔbccT4, ΔbccT1-ΔbccT3-ΔbccT4, ΔbccT1-ΔbccT2-ΔbccT4, and ΔbccT1-ΔbccT2-ΔbccT3.

Growth analyses of V. cholerae N16961, V. harveyi 393, V. vulnificus YJ016, and V. fluvialis ATCC 33809 (synonym, NCTC 11327 or 606) were conducted by growing strains overnight in M9G supplemented with 2% NaCl (M9G 2% NaCl). Strains were diluted 1:40 into 200 μl of M9G 4% NaCl, or M9G 5% NaCl for V. fluvialis, with and without exogenous DMG, in a 96-well microplate and grown at 37°C with intermittent shaking for 24 h. To test DMG as a carbon source, V. cholerae, V. vulnificus, V. fluvialis, and V. parahaemolyticus were grown overnight in LB1%, and V. harveyi was grown in LB2%. Cells were pelleted, washed two times with 1× PBS, and diluted 1:40 into 200 μl of M9 medium with 20 mM DMG as the sole carbon source and 1% NaCl (2% NaCl for V. harveyi). Strains were grown in M9G 1% NaCl (2% NaCl for V. harveyi) as a control. Strains were grown in a 96-well microplate as described above.

Functional complementation of E. coli strain MKH13 with VP1456, VP1723, VP1905, and VPA0356.

Full-length VP1456 (bccT1) or VPA0356 (bccT4) was amplified from the V. parahaemolyticus RIMD2210633 genome using primers listed in Table 3. All primers were purchased from Integrated DNA Technologies (Coralville, IA). The Gibson assembly protocol using NEBuilder HiFi DNA assembly master mix (New England BioLabs, Ipswich, MA) was followed to ligate the VP1456 or VPA0356 fragment with the expression vector pBAD33 (71), which had been linearized with SacI. Regions of complementarity for Gibson assembly are indicated by lowercase type in the primer sequences in Table 3. The resulting expression plasmid, pBAVP1456 or pBAVPA0356, was transformed into E. coli DH5α for propagation. Plasmids were then purified, sequenced, and subsequently transformed into E. coli strain MKH13, which has large deletions that include all compatible solute transporters (putP, proP, and proU) and the choline uptake and GB biosynthesis loci (betT-betIBA) (42). E. coli MKH13 strains containing pBAVP1456, pBAVP1723, pBAVP1905, or pBAVPA0356 were grown overnight in minimal medium supplemented with 1% NaCl and 20 mM glucose (M9G1%) and subsequently diluted 1:40 into M9G supplemented with 4% NaCl (M9G 4% NaCl). E. coli MKH13 strains containing pBAVP1456, pBAVP1723, pBAVP1905, or pBAVPA0356 were grown overnight in M9G 1% NaCl with chloramphenicol and subsequently diluted 1:100 into M9G 4% NaCl with 500 μM the indicated compatible solute and chloramphenicol for plasmid maintenance. The expression of each BccT protein was induced with 0.01% arabinose, and functional complementation was determined by measuring the OD₅₉₅ after 24 h of growth at 37°C with aeration. Growth was compared to that of an MKH13 strain harboring empty plasmid pBAD33, which cannot grow in M9G 4% NaCl without exogenous compatible solutes. Statistics were calculated using Student’s t test.

TABLE 3.

Primers used in this study

Use and primer	Sequence (5′–3′)a	Length (bp)
Expression
VP1456 Fwd	tgggctagcgaattcgagctTTTGCTGTAAATATGCAATAAAAGTG	1,740
VP1456 Rev	ggatccccgggtaccgagctTTAGCGGTAAGCGGAAAG
VPA0356 Fwd	tgggctagcgaattcgagctAGCGGCTTTTTGAACATC	1,671
VPA0356 Rev	ggatccccgggtaccgagctTTAAACCAAACCTTGATCC
Mutagenesis
BCCT1 TRP203CYS F	CATGTTTCACtgcGGTGTTCACGGTTGGAG	7,096
BCCT1 TRP203CYS R	GTCGCGCCCATCGCGAGT
BCCT1 TRP208LEU F	TGTTCACGGTctgAGTATTTACGC	7,096
BCCT1 TRP208LEU R	CCCCAGTGAAACATGGTC
BCCT1 TYR211LEU F	TTGGAGTATTctgGCCCTTGTTGCGTTG	7,096
BCCT1 TYR211LEU R	CCGTGAACACCCCAGTGA
BCCT1 TRP384LEU F	GTGGGTATCTctgTCTCCGTTTG	7,096
BCCT1 TRP384LEU R	CAAGCCCAGTAGAACACTG
BCCT1 203208211ALA F	tgcgagtattgcgGCCCTTGTTGCGTTGGCG	7,096
BCCT1 203208211ALA R	ccgtgaacacccgcGTGAAACATGGTCGCGCC
BCCT1 TRP384ALA F	GTGGGTATCTgcgTCTCCGTTTG	7,096
BCCT1 TRP384ALA R	CAAGCCCAGTAGAACACTG

^aRegions of complementarity for Gibson assembly are indicated by lowercase type.

Site-directed mutagenesis was performed on pBAVP1456 with a Q5 site-directed mutagenesis kit (New England BioLabs, Ipswich, MA) and primers listed in Table 3. Primers were designed to create nucleotide substitutions resulting in the following amino acid changes: Trp203Cys, Trp203Ala, Trp208Leu, Trp208Ala, Tyr211Leu, Tyr211Ala, Trp384Leu, and Trp384Ala. Site-directed mutagenesis was performed according to the manufacturer’s protocol to create single-amino-acid substitutions Trp203Cys, Trp208Leu, Tyr211Leu, and Trp384Leu. Residues 203, 208, and 211 were mutagenized to encode alanines, and residue 384 was subsequently mutagenized in this plasmid backbone to encode an alanine.

Homology modeling of BccT1 and docking of GB, DMG, and ectoine.

A BLAST search of the BccT1 sequence against the Protein Data Bank (PDB) as the search database showed the highest sequence identity with the 3.2-Å X-ray crystal structure of a GB transporter from C. glutamicum (CgBetP) (PDB ID 4AIN). Homology modeling of BccT1 was done with the SWISS-MODEL server using the CgBetP structure (PDB ID 4AIN) as a template (72). CgBetP is an active trimer correctly represented in its three-dimensional (3D) structure with the three protomers (chains) showing different stages of the substrate transport cycle with an alternating-access mechanism (46). More specifically, chain A of the CgBetP structure represents the closed apo state (C_c), whereas chains B and C represent a closed substrate-bound (C_cS) and an open substrate-bound (C_iS) state, respectively. Chain B and chain C of CgBetP were used as the templates to generate C_cS- and C_iS-like homology models of BccT1, respectively. The energy-minimized models of BccT1 were verified using a 3D profile method of Verify3D that evaluates the correlation between the amino acid sequence (one dimensional [1D]) and the model (3D) by comparing it to the other known structures in the database. Residues showing poor stereochemical properties or close contacts in each monomeric model were fixed manually in COOT (73). Models were then subjected to energy minimization using the 3Drefine server to reduce any other structural restraints (74). The quality of the resulting models was finally verified with Verify3D and PROCHECK (75, 76). The C_cS and C_iS models of BccT1 showed Verify3D scores of 88.6% and 84.0%, respectively, which indicated good quality of the models (see Fig. S9 in the supplemental material). The stereochemical properties of the models were examined by a Ramachandran plot using PROCHECK. Both the C_cS and C_iS BccT1 models contain 99.2% of all the residues in the allowed region and none in the disallowed region (Fig. S10).

For docking studies, ligand models for GB, DMG, and ectoine were obtained from Chemical Entities of Biological Interest (ChEBI) EMBL (77). Polar hydrogens were added to the ligand structures in PRODRG (78). AutoDock tools v1.5.6 was used to assign the rotatable bonds in the ligands and to add all polar hydrogens in the BccT1 models to prepare them for the docking. The ligand-centered maps for BccT1 models were assigned a grid size of 20 by 20 by 20 ų. The docking experiments of GB, DMG, and ectoine ligands with the BccT1 receptor model were performed using AutoDock Vina (79). The binding free energies and BccT1 residues making interactions with the ligands are listed in Table 1. The structural illustrations were prepared using PyMOL (PyMOL Molecular Graphics System, version 2.0; Schrödinger, LLC).

Bioinformatics and phylogenetic analyses.

Transmembrane helix probabilities of C. glutamicum BetP (GenBank accession number CAA63771.1) and BccT1 (GenBank accession number Q87PP5.1) were generated using OCTOPUS and aligned via the AlignMe program (http://www.bioinfo.mpg.de/AlignMe) (80–82). The sequences of the V. parahaemolyticus protein BccT1 (GenBank accession number Q87PP5.1), CgBetP (GenBank accession number CAA63771.1), CgEctP (GenBank accession number CAA04760.1), and CgLcoP (GenBank accession number ASW14702.1) were downloaded from the NCBI database and aligned using the ClustalW algorithm (83). Aligned sequences were displayed and annotated using ESPript (http://espript.ibcp.fr/ESPript/cgi-bin/ESPript.cgi) (84).

Phylogenetic analysis was conducted using the BccT1 (VP1456) protein as a seed to identify all homologs within the family Vibrionaceae with completed genome sequences available. Unique protein sequences that had >95% sequence coverage and >70% amino acid identity with BccT1 were obtained from the NCBI database and aligned using the ClustalW algorithm (83). The evolutionary history of BccT1 was inferred by using the maximum likelihood method and the Le_Gascuel_2008 model as determined by best-fit model selection in MEGAX (85, 86). The tree with the highest log likelihood (−10,453.23) is shown. The percentages of trees in which the associated taxa clustered together are shown next to the branches. The initial tree(s) for the heuristic search was obtained automatically by applying neighbor-joining and BioNJ algorithms to a matrix of pairwise distances estimated using a Jones-Taylor-Thornton (JTT) model and then selecting the topology with the superior log-likelihood value. A discrete gamma distribution was used to model evolutionary rate differences among sites (5 categories [+G, parameter = 0.4143]). The tree is drawn to scale, with branch lengths measured in the number of substitutions per site. This analysis involved 49 amino acid sequences and a total of 525 positions in the final data set.