Carbohydrate Kinase (RhaK)-Dependent ABC Transport of Rhamnose in Rhizobium leguminosarum Demonstrates Genetic Separation of Kinase and Transport Activities

Abstract

In Rhizobium leguminosarum the ABC transporter responsible for rhamnose transport is dependent on RhaK, a sugar kinase that is necessary for the catabolism of rhamnose. This has led to a working hypothesis that RhaK has two biochemical functions: phosphorylation of its substrate and affecting the activity of the rhamnose ABC transporter. To address this hypothesis, a linker-scanning random mutagenesis of rhaK was carried out. Thirty-nine linker-scanning mutations were generated and mapped. Alleles were then systematically tested for their ability to physiologically complement kinase and transport activity in a strain carrying an rhaK mutation. The rhaK alleles generated could be divided into three classes: mutations that did not affect either kinase or transport activity, mutations that eliminated both transport and kinase activity, and mutations that affected transport activity but not kinase activity. Two genes of the last class (rhaK72 and rhaK73) were found to have similar biochemical phenotypes but manifested different physiological phenotypes. Whereas rhaK72 conferred a slow-growth phenotype when used to complement rhaK mutants, the rhaK73 allele did not complement the inability to use rhamnose as a sole carbon source. To provide insight to how these insertional variants might be affecting rhamnose transport and catabolism, structural models of RhaK were generated based on the crystal structure of related sugar kinases. Structural modeling suggests that both rhaK72 and rhaK73 affect surface-exposed residues in two distinct regions that are found on one face of the protein, suggesting that this protein's face may play a role in protein-protein interaction that affects rhamnose transport.

INTRODUCTION

Cells require a specific and regulated way to transport substrates across membranes. One of the largest families of transporters is the ATP binding cassette (ABC) transporters. ABC transporters utilize free energy from ATP hydrolysis to transport substrates across a membrane. They are widely distributed in all domains of life and are involved in transport that affects diverse biological functions (1–3).

Members of the ABC superfamily are defined by the ATPase protein that contains the Walker A and Walker B motifs, along with an LSGGQ conserved consensus sequence (4, 5). Functional ABC importers generally consist of two proteins that have transmembrane domains consisting of six membrane-spanning regions (permeases), two ABC proteins that are cytoplasmically localized and contain the ATP binding domains, and a periplasmically localized substrate binding protein for the function of the transport system (6). Relatively few sequence similarities are found between binding proteins for different substrates as the periplasmic substrate binding protein plays a central role in substrate specificity (7).

Gram-negative bacterial ABC transport systems responsible for the import of carbohydrates can be broken into two main classes: carbohydrate uptake transporters 1 and 2 (CUT1 and -2, respectively) (8). Most work on bacterial carbohydrate ABC importer systems has been done using the CUT1 Escherichia coli maltose transport system as a model, and many aspects of its function are understood at the atomic level, giving rise to a clear understanding of the mechanism and kinetics of sugar uptake systems (9–13). In contrast, less work has been carried out using CUT2 transporters as models.

CUT2 transporters are classified primarily on the basis of the ABC protein (8). The ABC protein tends to be about 500 amino acids, which contrasts with about 300 amino acids generally found in the CUT1 family (8). This protein appears to have arisen by a fusion of two ABC domains (14). Although the ABC protein contains two ATP binding sites, it is not clear that both nucleotide binding motifs are functional in ATP hydrolysis since a key lysine residue in the C-terminal Walker A motif is replaced with an arginine (6, 8). This type of mutation has been shown to adversely affect activity (15). Due to its size, it is not clear whether CUT2 transporters function with one or two copies of this larger ABC protein (8, 14).

The genomes of Rhizobium species tend to have high numbers of genes encoding ABC transport systems. Sinorhizobium meliloti, Rhizobium leguminosarum, and Mesorhizobium loti contain 200, 269, and 216 annotated ABC genes, respectively (16–19). In contrast, Pseudomonas aeruginosa, which has a comparable genome size and can also be found as a soil organism, contains about 124 (20). Since rhizobia do have a high number of ABC transport systems and since a number of these strains are being used as model systems, studying them in these organisms may lead to findings that give insight into nuances associated with ABC transport.

In R. leguminosarum strain Rlt100, the inability to transport or catabolize rhamnose leads to a decreased ability to compete for nodule occupancy relative to the wild-type strain (21). Transport of rhamnose was shown to be encoded by a CUT2-type ABC transporter encoded by rhaSTPQ (22). Strains carrying a mutation in the sugar kinase (rhaK) that is necessary for the catabolism of rhamnose were unable to transport labeled rhamnose into the cell even though components of the transporter being transcribed were translated and localized to the membrane (23), suggesting that RhaK affects ABC transport in addition to having kinase activity. In this paper we address the hypothesis that the kinase activity of RhaK is distinct from its ability to affect the transport of rhamnose.

MATERIALS AND METHODS

Bacterial strains and culture conditions.

E. coli strains used in this work were grown on LB as a complex medium at 37°C (24). R. leguminosarum strains were grown at 30°C on TY (tryptone-yeast extract) as a complex medium (25) or on Vincent minimal medium (VMM) as a defined medium (26). Carbon sources were filter sterilized and added to a final concentration of 15 mM. Antibiotics were added at the following concentrations for solid media: tetracycline (Tc), 5 μg/ml; streptomycin (Sm), 200 μg/ml; neomycin (Nm), 200 μg/ml; kanamycin (Kan), 50 μg/ml; and ampicillin (Amp), 100 μg/ml. For growth in liquid medium, antibiotic concentrations were halved (Table 1).

Table 1.

Strains and plasmids

Strain or plasmid	Relevant genotype	Reference or source
R. leguminosarum strains
Rlt100	W14-2 Smr, wild type	47
Rlt144	Rlt100 rhaK50::Tn5-B20	22
Rlt106	Rlt100 rhaT2::Tn5-B20	21
E. coli strains
DH5α	λ− ϕ80dlacZΔM15 Δ(lacZYA-argF)U169 recA1 endA1 hsdR17(rK− mK−) supE44 thi-1 gyrA relA1	48
MT616	MT607(pRK600)	49
Plasmids
pRK7813	Broad-host-range vector, Tcr	50
pRK600	pRK2013 npt::Tn9, Cmr	49
pMR110	pRK7813 with rhaK+ expressed from Plac promoter	23
pMR178	rhaK with N-terminal His6 tag in pRK7813	This work
pDR170	pRK7813 with rhaK72 with an N-terminal His tag expressed from Plac promoter	This work
pLit28i	High-copy-number colEI cloning vector, Apr	New England Biolabs
pDR1	rhaK in pLit28i	This work
pDR22	pRK7813 with rhaK66 expressed from the Plac promoter	This work
pDR70	pRK7813 with rhaK65 expressed from the Plac promoter	This work
pDR72	pRK7813 with rhaK70 expressed from the Plac promoter	This work
pDR73	pRK7813 with rhaK72 expressed from the Plac promoter	This work
pDR74	pRK7813 with rhaK73 expressed from the Plac promoter	This work
pDR77	pRK7813 with rhaK88 expressed from the Plac promoter	This work
pDR78	pRK7813 with rhaK74 expressed from the Plac promoter	This work
pDR79	pRK7813 with rhaK91 expressed from the Plac promoter	This work
pDR80	pRK7813 with rhaK84 expressed from the Plac promoter	This work
pDR81	pRK7813 with rhaK76 expressed from the Plac promoter	This work
pDR82	pRK7813 with rhaK78 expressed from the Plac promoter	This work
pDR144	pRK7813 with rhaK69 expressed from the Plac	This work
pDR146	pRK7813 with rhaK79 expressed from the Plac promoter	This work
pDR147	pRK7813 with rhaK85 expressed from the Plac promoter	This work
pDR155	pRK7813 with rhaK92 expressed from the Plac promoter	This work
pDR157	pRK7813 with rhaK95 expressed from the Plac promoter	This work
pDR158	pRK7813 with rhaK97 expressed from the Plac promoter	This work
pDR159	pRK7813 with rhaK115 expressed from the Plac promoter	This work
pDR160	pRK7813 with rhaK114 expressed from the Plac promoter	This work
pDR161	pRK7813 with rhaK96 expressed from the Plac promoter	This work
pDR162	pRK7813 with rhaK104 expressed from the Plac promoter	This work
pDR163	pRK7813 with rhaK103 expressed from the Plac promoter	This work
pDR164	pRK7813 with rhaK109 expressed from the Plac promoter	This work
pDR165	pRK7813 with rhaK116 expressed from the Plac promoter	This work
pDR166	pRK7813 with rhaK111 expressed from the Plac promoter	This work
pDR167	pRK7813 with rhaK117 expressed from the Plac promoter	This work
pDR168	pRK7813 with rhaK119 expressed from the Plac promoter	This work
pDR169	pRK7813 with rhaK121 expressed from the Plac promoter	This work
pDR181	pRK7813 with rhaK122 expressed from the Plac promoter	This work
pDR182	pRK7813 with rhaK64 expressed from the Plac promoter	This work
pDR183	pRK7813 with rhaK90 expressed from the Plac promoter	This work
pDR184	pRK7813 with rhaK94 expressed from the Plac promoter	This work
pDR185	pRK7813 with rhaK98 expressed from the Plac promoter	This work
pDR186	pRK7813 with rhaK99 expressed from the Plac promoter	This work
pDR187	pRK7813 with rhaK100 expressed from the Plac promoter	This work
pDR188	pRK7813 with rhaK112 expressed from the Plac promoter	This work
pDR189	pRK7813 with rhaK118 expressed from the Plac promoter	This work
pDR191	pRK7813 with rhaK105 expressed from the Plac promoter	This work
pDR192	pRK7813 with rhaK106 expressed from the Plac promoter	This work
pDR193	pRK7813 with rhaK80 expressed from the Plac promoter	This work

Genetic manipulations.

Conjugations between E. coli strains and R. leguminosarum were performed as described previously using the mobilizing strain MT616 (27).

DNA manipulations and constructions.

Standard techniques were used for gel electrophoresis, restriction enzyme digests, and isolation of plasmid DNA (28). Nucleotide sequencing was carried out by cycle sequencing using a Big-Dye, version 3.1, kit. Sequencing reactions were carried out as recommended by the manufacturer and resolved using an ABI 3130 sequencer.

Construction of pDR1.

The rhaK coding region was removed from pMR110 using BamHI and HindIII and ligated into pLit28i. The resultant construct was confirmed by sequencing and named pDR1 (Table 1). To overexpress an RGS (arginine-glycine-serine)-His₆-tagged variant of RhaK72 in R. leguminosarum, a C-terminally tagged version was constructed. pDR73 was used as a template. The primers 5 HindRBS (5′ATATAAGCTTGGAGAACTGCAGATGACCGCCAGTTCCTATC-3′; containing an HindIII restriction site, a ribosome-binding site [RBS], and the 5′ annealing portion of rhaK) and 3His 6B (5′-ATGGATCCCTAATGATGATGATGATGATGCGATCCTCTTGCCATCGCCGCGTACC-3′; containing a BamHI restriction site, a C-terminal RGS-His₆ tag, and the 3′ annealing portion of rhaK) and rhaK72 as a template were used to create a C-terminally tagged RhaK72. The PCR product was subsequently cut with BamHI and HindIII and cloned into pRK7813 such that it was expressed from the P_lac promoter. The construct was sequenced and named pDR170.

Generation of Tn7 in-frame insertional rhaK alleles.

In-frame insertion mutations were generated using a New England BioLabs GPS-LS Mutagenesis Kit. Briefly, in vitro mutagenesis was carried out using pDR1, pGPS5 (transposon donor), and TnsABC (transposase). The reaction product was transformed into library-grade competent cells. Transformants carrying insertions within the coding sequence of rhaK were identified by restriction digestion using BamHI and HindIII. The point of insertion of each allele was determined by sequencing using the suggested primers that were designed to the ends of the Tn7 transposon. Inserts that were within the coding sequence were subsequently recloned into pRK7813. The Tn7 construct had PmeI sites placed such that the Tn7 DNA could be deleted by cutting with PmeI and religating. This would leave behind a 15-bp insert (10 bp from the transposon [TGTTTAAACA] and a 5-bp duplication at the site of insertion) that could generate either a 5-amino-acid insert or a stop codon in the final protein. Each allele was confirmed by sequencing and conceptually translated to determine the nature of the insertion.

Rhamnose transport assay.

Transport assays were carried out as previously described with slight modifications (22, 23, 29). Radioactive [³H]rhamnose (5 Ci/mmol) was purchased from American Radiolabeled Chemicals, Ltd. (St. Louis, MO). Strains were grown overnight in 5-ml TY cultures and then subcultured into 5 ml of VMM-glycerol-rhamnose. Cells were harvested by centrifugation (5,000 × g for 10 min) at mid-log phase (optical density at 600 nm [OD₆₀₀] of 0.5 to 0.7), washed twice in VMM-glycerol, and then resuspended in defined salts medium to a final OD₆₀₀ of 0.3 in a total volume of 2 ml. Transport assays were initiated by the addition of [³H]rhamnose to a final concentration of 2 μM, and aliquots of 0.5 ml were withdrawn at appropriate time points and rapidly filtered through a Millipore 0.45-μm-pore-size HV filter on a Millipore sampling manifold. Filtered cells were immediately washed with 5 ml of defined salts medium to wash away unincorporated label. Label incorporation was quantified using a liquid scintillation spectrophotometer (Beckman LS6500). Samples were taken every 20 s, and sampling continued for up to 2 min. Transport rates were generally linear over the first minute of the assay. Longer-term accumulation assays were carried out in a similar fashion except that incubation times with label were extended.

Biochemical assays.

An assay for rhamnose kinase activity was developed by adapting a fructose kinase assay that has been previously described (30, 31). Briefly R. leguminosarum cells were grown, and cell-free lysates were prepared as previously described (32–34). The assay was initiated with the addition of rhamnose. Rates were determined from linear portions of NADH oxidation data that were corrected for background oxidase activity and were proportional to the amount of extract assayed. Malate dehydrogenase assays were carried out as previously described (35).

Isolation of cell lysate membrane fraction.

Strains were grown on defined medium to mid-log phase and harvested, and cell extracts were prepared. Membranes were isolated by subsequent ultracentrifugation (Beckman Coulter MaxE) of the cell extracts using a TLA 100.3 rotor (45,000 × g for 30 min at 4°C). Membrane pellets were subsequently resuspended in extraction buffer. Isolated membrane fractions were routinely monitored for cytoplasmic cross-contamination by assaying for malate dehydrogenase.

Western blotting.

Cell fractions were analyzed by SDS-PAGE (36) and transferred to a nitrocellulose membrane for Western blot analysis using a His₆ monoclonal antibody as a primary antibody and a goat anti-mouse secondary antibody linked to horseradish peroxidase (HRP). HRP was detected colorimetrically with an opti-4CN substrate detection kit (Bio-Rad Laboratories) as previously described (29, 36).

Bioinformatic analysis and molecular modeling of RhaK.

Basic protein alignments were carried out using Clustal-X (37) and Praline (38). Structural models of RhaK were constructed by using Phyre² (39). The final scaffold that was used to model RhaK was gluconate kinase from Lactobacillus acidophilus (Protein Data Bank [PDB] file 3GBT). Final figures were generated in PyMol (The PyMOL Molecular Graphics System, version 1.5.0.4, Schrödinger LLC).

RESULTS

RhaK can associate with the cytoplasmic membrane.

Strains lacking the kinase RhaK were shown to be unable to transport rhamnose (22, 23). The inability to transport rhamnose was not due to the lack of expression or translation of the ABC transporter RhaSTPQ (23). Conversely, insertional mutants lacking the ABC transporter components RhaTPQ, the mutarotase RhaU, and RhaK did not affect rhamnose kinase activity when rhaK was expressed from a plasmid (data not shown). It was hypothesized that if RhaK affected rhamnose transport, it might interact with either the cytoplasmic membrane or the components of the ABC transporter.

To address this hypothesis, two RhaK variants with an RGS-His₆ tag on either the C or N terminus were generated such that the localization of RhaK could be followed by Western blotting. Each variant was capable of complementing rhaK mutants for both growth using rhamnose as a sole carbon source and transport (data not shown). Western blot analysis using a monoclonal antibody to the RGS-His₆ epitope resulted in a single band of approximately 55 kDa (Fig. 1). Cell lysates were further separated by ultracentrifugation into cytoplasmic and membrane fractions. Assaying these fractions for malate dehydrogenase activity showed that the membrane fraction did not contain detectable activity, suggesting that the membrane fractions were devoid of cytoplasmic contamination (data not shown).

Fig 1.

Western blot analysis of RhaK in cellular fractions. (A) His-tagged wild-type RhaK expressed in wild-type (Rlt100) and rhaK50 (Rlt144) strains. Cleared cell lysates (L) were fractionated into membrane (M) and cytoplasmic (C) fractions. Equal volumes of each fraction were separated by SDS-PAGE, blotted, and detected. The numbers below each lane indicate the proportions of total rhamnose kinase activity from the lysate found in each fraction. (B) Salt washes of membrane-bound RhaK. Membrane samples were run on SDS-PAGE gels, blotted, and detected. NaCl wash concentrations are indicated above the panel. (C) Western blot analysis of His-tagged RhaK expressed in Rlt106 (rhaT2). Designations are as described for panel A.

The data showed that RhaK could be found in the membrane fraction (Fig. 1). Since rhamnose kinase activity is directly related to the amount of RhaK detected by Western blot analysis, RhaK activity was used to determine the proportion of RhaK that was associated with each of the fractions (Fig. 1A). When the proportion of the total RhaK activity associated with each fraction was calculated, it was found that about 7% of the total activity was associated with the membrane fraction. Similar results were obtained in both a wild-type and an rhaK strain background (Fig. 1A).

To determine how tightly RhaK was associated with the membrane, isolated membrane fractions were subjected to salt washes. Whereas a 50 mM NaCl wash did not appear to give results that were qualitatively different from an unwashed membrane, the majority of the RhaK signal was lost with a 150 mM wash. This suggests that a proportion of RhaK can interact with the cytoplasmic membrane in a manner that is consistent with it being a peripheral membrane protein (Fig. 1B).

The ABC rhamnose transporter in R. leguminosarum consists of the periplasmically localized solute binding protein RhaS, the integral membrane proteins RhaP and RhaQ that are the permease components, and the cytoplasmic protein RhaT, which is the ABC component. Since an absence of RhaK results in a lack of transport activity, we hypothesized that RhaK may be directly interacting with components of the ABC transporter.

To address this hypothesis, the plasmid pMR178, which contains the tagged version of RhaK, was mated into Rlt106. Rlt106 contains a Tn5 insertion within rhaT, thus eliminating the rhaTPQ as well as the mutarotase rhaU (29) and the chromosomal copy of rhaK. Western blot analysis of the membrane and cytoplasmic fractions shows that the distribution of His-tagged RhaK in this background is comparable to that seen in both the wild-type background and the rhaK strain background. This suggests that the RhaK membrane interaction is not dependent on the presence of the rhamnose ABC transporter components since RhaK localizes to the membrane fraction even in the absence of the transporter components (Fig. 1C).

Generation of rhaK mutant alleles.

The characterization of RhaU demonstrated that the structure of this protein was consistent with its function as a cytoplasmic rhamnose mutarotase (23). These data strongly suggest that the ability of RhaK to affect transport and its ability to phosphorylate its substrate can be uncoupled physiologically.

To address this hypothesis, in-frame insertional mutagenesis of RhaK was carried out. The rhaK open reading frame was cloned into pLit28i as a BamHI/KpnI fragment, yielding pDR1, which was subsequently mutagenized. Putative transposon insertional mutants were screened by restriction to identify those that generated an insert band that was larger than rhaK, indicating the presence of the transposon. Thirty-nine unique insertional mutants within rhaK were generated in this manner. The exact location of each insertion was determined by sequencing. Since there are two PmeI sites located in the Tn7, the internal portion of the transposon was removed by cutting with PmeI and religating. The resultant construct had a 15-bp insertion within rhaK. These insertional mutants were subsequently cut with BamHI and HindIII and then recloned into pRK7813 (Table 1).

Not including inserts that were found in identical positions, the average spacing between inserts was 12 amino acids (high of 50 amino acids; low of 1). Ten of the insertions generated stop codons, leaving 29 inserts that generated 5-amino-acid inserts. In addition to the 39 inserts presented, 15 insertions that were in identical positions (generating identical insertions) were also isolated (data not shown). We note that the largest gap in our coverage appears to occur between alleles rhaK72 and rhaK109 near the C terminus of the protein (Fig. 2).

Fig 2.

Linker-scanning mutagenesis. Sites and sequences of the mutants generated are shown. Insertion sites are denoted by triangles. Filled triangles denote inserts that generate truncations. The conceptual translation generated by each insert is shown on the top of the triangle. Highlighted amino acids denote residues deducted to play a role in either ATP binding (red), Mg²⁺ binding (yellow), or substrate binding (green) or are predicted to be a dimerization domain (blue) based on other related carbohydrate kinases.

Screening of linker-scanning rhaK alleles.

All the alleles that were generated were screened for their ability to complement the rhaK mutant strain Rlt144 for growth on rhamnose. In addition, since the buildup of a phosphorylated intermediate has also been correlated with the inability to complement (40), the alleles were also screened on medium containing rhamnose and glycerol (Table 2). The results show that 25 of the insertions were not affected in their ability to complement Rlt144 (Table 2), and thus these were not further examined. Thirteen of the remaining insertions were unable to complement Rlt144 for growth using rhamnose as a sole carbon source. Eleven of these insertions were predicted to generate stop codons that would yield a truncated protein. The two remaining alleles, rhaK73 and rhaK111, were of particular interest for further analysis.

Table 2.

Complementation analysis of rhaK50 using rhaK linker-scanning alleles

Strain	Relevant genotypeb	Growth on the indicated mediuma
		Rhamnose	Rhamnose-glycerol	Glycerol
Rlt100	Wild type	+	+	+
Rlt144	rhaK50	−	+	+
Rlt144(pDR22)	rhaK66	−	+	+
Rlt144(pDR70)	rhaK65	+	+	+
Rlt144(pDR72)	rhaK70	+	+	+
Rlt144(pDR73)	rhaK72	+/−	+	+
Rlt144(pDR74)	rhaK73	−	+	+
Rlt144(pDR77)	rhaK88	−	+	+
Rlt144(pDR78)	rhaK74	+	+	+
Rlt144(pDR79)	rhaK91	−	+	+
Rlt144(pDR80)	rhaK84	−	+	+
Rlt144(pDR81)	rhaK76	+	+	+
Rlt144(pDR82)	rhaK78	−	+	+
Rlt144(pDR144)	rhaK69	−	+	+
Rlt144(pDR146)	rhaK79	−	+	+
Rlt144(pDR147)	rhaK85	−	+	+
Rlt144(pDR155)	rhaK92	+	+	+
Rlt144(pDR157)	rhaK95	+	+	+
Rlt144(pDR158)	rhaK97	+	+	+
Rlt144(pDR159)	rhaK114	+	+	+
Rlt144(pDR160)	rhaK115	−	+	+
Rlt144(pDR161)	rhaK96	+	+	+
Rlt144(pDR162)	rhaK104	+	+	+
Rlt144(pDR163)	rhaK103	+	+	+
Rlt144(pDR164)	rhaK109	+	+	+
Rlt144(pDR165)	rhaK116	+	+	+
Rlt144(pDR166)	rhaK111	−	+	+
Rlt144(pDR167)	rhaK117	+	+	+
Rlt144(pDR168)	rhaK119	+	+	+
Rlt144(pDR169)	rhaK121	+	+	+
Rlt144(pDR182)	rhaK64	−	+	+
Rlt144(pDR183)	rhaK90	+	+	+
Rlt144(pDR184)	rhaK94	+	+	+
Rlt144(pDR185)	rhaK98	−	+	+
Rlt144(pDR186)	rhaK99	+	+	+
Rlt144(pDR187)	rhaK100	+	+	+
Rlt144(pDR188)	rhaK112	+	+	+
Rlt144(pDR189)	rhaK118	+	+	+
Rlt144(pDR191)	rhaK105	+	+	+
Rlt144(pDR192)	rhaK106	+	+	+
Rlt144(pDR193)	rhaK80	+	+	+

^aComplementation was scored by judging growth, on defined medium, of Rlt144 carrying each of the alleles. Growth is scored as follows: +, wild-type level; +/−, slow growth; −, no growth.

^bRlt144 carries rhaK50::Tn5-B20. For clarity, only the allele being tested is listed.

A single insert, rhaK72, was initially scored as unable to complement Rlt144 following 3 days of incubation. It was found that colonies would appear after extended periods of growth (6 days). To determine whether this growth was due to second site mutations arising in the culture, single colonies were isolated, purified, and tested for regrowth. It was found that, upon restreaking, these colonies exhibited the same slow-growth phenotype, suggesting that the growth that was exhibited was due to the rhaK72 allele and not a second site mutation.

Biochemical properties of linker-scanning alleles.

We hypothesized that with respect to affecting rhamnose transport and kinase activity, four types of mutant alleles should be possible, with the following results: presence of both activities, absence of both activities, presence of transport and the lack of kinase activity, and, finally, presence of kinase and lack of transport activity. Each of the remaining alleles was introduced into Rlt144 and screened for the ability to confer transport and kinase activity. The majority of these alleles did not show any appreciable differences from Rlt144 or were obviously nonfunctional because the insertion generated a stop codon (Fig. 2). Thus, representative alleles from these subsets were chosen for further analysis.

The alleles chosen were as follows: two insertions (rhaK64 and rhaK85) that represent the longest truncated versions of RhaK and three insertions (rhaK70, rhaK92, and rhaK97) from the N-terminal, middle, and C-terminal regions of RhaK, respectively, that were indistinguishable from the wild type based on complementation analysis. The remaining alleles (rhaK72, rhaK73, and rhaK111) had complementation phenotypes that were consistent with affecting kinase and/or transport activity.

The results showed that alleles that encode premature stop codons did not confer either kinase or transport activity (Fig. 3). We note that although rhaK64 does not confer either transport or kinase activity, insertions that do not generate stop codons in this area (rhaK109 and rhaK97) do not impair RhaK function (Table 2 and Fig. 3; also data not shown). Not surprisingly, alleles that retained the ability to complement an rhaK mutation retained kinase activity and conferred the ability to transport rhamnose at nearly wild-type levels (Fig. 3).

Fig 3.

Transport and kinase rates of representative rhaK alleles. RhaK variants expressed from plasmids in Rlt144 were assayed for transport and kinase activities. Whole-cell assays were used to measure uptake of radiolabeled rhamnose, whereas cleared lysates were used to measure kinase activity. Transport and kinase rates are presented as percentages of the wild-type levels. Error bars represent standard deviations of the means of at least three biological replicates. Typical wild-type kinase rates were approximately 600 μmol/min/mg of protein. Typical wild-type transport rates were about 10 nmol/min/mg of protein.

Whereas the rhaK111 allele was unable to confer either rhamnose transport ability or kinase activity, rhaK73 had appreciable rhamnose kinase activity (Fig. 3). Assays of Rlt144 carrying the rhaK72 allele showed that this allele encoded a variant that had kinase activity that was not significantly different from the wild type, but it could not confer the ability to transport rhamnose over background levels (Fig. 3). Western blot analysis using His-tagged versions of rhaK72 showed that it localized identically to the wild-type gene (data not shown).

Alleles that uncouple transport and kinase activity have different complementation phenotypes.

rhaK72 and rhaK73 both uncouple the ability to confer transport and kinase activity (Fig. 3). In an attempt to gain new insight into how these alleles might be affecting the ability of the cell to utilize rhamnose, a growth experiment was carried out to quantitate growth. Rlt100 growing on defined medium under our growth conditions typically has a doubling time of between 8 and 10 h (Table 3) (29). If Rlt144 carrying the rhaK50 allele is inoculated into defined medium with rhamnose as a sole carbon source, no measurable doubling occurs. Rlt144 complemented with the wild-type rhaK on a plasmid clearly grows; however, it is not unusual to note a slightly longer doubling time (Table 3). Introduction of either rhaK72 or rhaK73 on a multicopy plasmid into Rlt144 shows significantly different growth rates compared with the strain carrying the wild-type rhaK allele or with each other. Whereas the doubling time of Rlt144 carrying the rhaK72 allele was calculated to be about 21 h, the doubling time could not be calculated from Rlt144 carrying the rhaK73 allele because there was very little growth (Table 3).

Table 3.

rhaK72 and rhaK73 alleles confer different growth rates when they are expressed in Rlt144 (rhaK50::Tn5)

Strain	Relevant genotype		Doubling time (h)a
	Chromosomal	Plasmid
Rlt100	Wild type		9.5 ± 0.9
Rlt144	rhaK50		NG
Rlt144(pMR110)	rhaK50	rhaK+	12.7 ± 0.1
Rlt144(pDR73)	rhaK50	rhaK72	21.1 ± 1.1
Rlt144(pDR74)	rhaK50	rhaK73	NG

^aDoubling times were calculated from cultures growing in defined medium over an 16-h time interval as follows: doubling time = ln(2)/[ln(N₂/N₁)]/T, where N₁ is the initial OD₆₀₀, N₂ is the final OD₆₀₀, and T is time (h). All cultures were in mid-log phase over the entire time interval. Values represent the means ± standard deviations of three independent cultures. NG, no detectable growth.

Initial rhamnose transport assays that were carried out with rhaK72 or rhaK73 in Rlt144 did not show transport activity that was over the level of background in a 2-min assay (Fig. 3). To determine if residual rhamnose uptake could be detected, extended incubations of Rlt144 carrying rhaK72 and rhaK73 were carried out. Consistent with the growth data, Rlt144 carrying rhaK72 was able to accumulate about 59 nmol/mg of protein, whereas Rlt144 carrying rhaK73 accumulated approximately 6 nmol/mg of protein (Fig. 4). Both of these values are significantly greater than the value that was determined for Rlt144 (Fig. 4).

Fig 4.

Accumulation of rhamnose in Rlt144 carrying either rhaK72 or rhaK73. Rlt144 or Rlt144 carrying either rhaK72 or rhaK73 was incubated with labeled rhamnose for 60 min. Whole-cell assays were used to measure uptake of radiolabeled rhamnose. Data are presented as nmol of rhamnose accumulated/mg of protein. The corresponding wild-type transport rate for this experiment was 10.2 ± 0.8 nmol/mg/min. Error bars represent standard deviations of the means of three biological replicates.

Molecular architecture of RhaK.

Bioinformatic, genetic, and biochemical data support that RhaK is a sugar kinase. Our initial analysis was capable of identifying a limited number of domains and motifs associated with RhaK (Fig. 2). Briefly, it was shown that RhaK belonged to a family of FGGY-type kinases, that a conserved P-loop motif used for the binding of ATP was found close to the N terminus, and that based on Clustal and Praline alignments to the E. coli GlpK (41, 42), the active-site residue necessary for the catalysis was D248 (23).

The generation of linker-scanning insertions that uncoupled the protein kinase and transport activities necessitated a more in-depth analysis of the protein to develop hypotheses regarding how RhaK might have such disparate activities. Initial experiments focused on overexpressing His-tagged variants in E. coli with the ultimate goal of generating an experimentally determined structural model. These experiments were complicated by the number of rare arginine codons that are found in RhaK. Expressing rhaK in either R. leguminosarum or S. meliloti circumvented some of the low-yield issues that were encountered; however, the isolated protein did not show in vitro activity although the constructs were capable of complementing strains carrying an rhaK mutation. To gain insight into how the insertion alleles might be changing RhaK, we instead opted to generate a model of RhaK using the Phyre² server.

The Phyre² server predicts the secondary structure of the query protein and matches this pattern to that of proteins that have been crystallized (39). The output data modeled RhaK to 20 different crystallized proteins. The top 19 ranked hits were all carbohydrate kinases. The percent sequence identity varied from 14 to 39%. The typical alignment coverage was 97%, and the confidence level for the predicted model was 100% in each case. The top hit was a gluconate kinase from Lactobacillus acidophilus (PDB 3GBT), with 98% alignment coverage, 15% amino acid identity, and 100% confidence level in its prediction. Visual analysis of the 19 models that were generated to the carbohydrate kinases showed a high degree of similarity in the overall structure. A common anomaly that we note is a “floating helix” at the C terminus of the protein. This type of anomaly appears to be a result of the difference in lengths between the protein being modeled and the protein to which it is being scaffolded. Since there was general agreement with all the models, we chose to use the model that was generated by scaffolding RhaK onto the top hit.

The predicted model shows a two-lobed protein (Fig. 5A). The active-site residue is found at the base of a long cleft; the ATP binding site and substrate binding sites are also found along the N-terminal lobe that makes up the active-site cleft. Other notable features include a pair of beta sheets that act as a likely dimerization domain (43). Moreover, the entire shape and structure of the model predict a carbohydrate kinase that has a great number of conserved structural features. Most notably there is a high degree of agreement on which amino acids appear to be surface exposed (Fig. 5A).

Fig 5.

Predicted structure of wild-type RhaK and variants. Structural models were generated using Phyre². The RhaK amino acid sequence was scaffolded to gluconate kinase from Lactobacillus acidophilus (PDB 3GBT). (A) Overall predicted structure of RhaK. (B) RhaK72 variant model superimposed onto RhaK. The region highlighted in panel A corresponds to the region depicted in panel B. The orange ribbon indicates the amino acid insertion, whereas the pink ribbon depicts regions where the structures differ. (C) RhaK73 variant model superimposed onto RhaK (top view relative to panel A). The red ribbon depicts the amino acid insertion. (D) Composite model (top view relative to panel A) depicting surface residues that are predicted to be changed in RhaK73 (red) and RhaK72 (pink).

The rhaK72 insertion occurs immediately following A363 in RhaK and generates a 5-amino-acid insertion (Fig. 2, VFKHH). The region from R357 to I380 in RhaK is predicted with a high degree of confidence (9 on scale of 9) to form an alpha-helical region. The inclusion of the 5 amino acids in the rhaK72 variant within this helix is not predicted to disrupt the helix since the secondary structure prediction across this area still maintains a high confidence in the probability of an alpha helix forming. The net result is a helix of 23 amino acids in the rhaK72 variant, whereas the wild-type contains a 24-amino-acid alpha helix. Modeling this into a three-dimensional space predicts that this is a helix that leads to a surface-exposed loop. Overlaying the predicted rhaK72 structure with the wild-type structure suggests that this insertion leads to a change in the surface loop region of the variant (Fig. 5B).

The rhaK73 insertion occurs immediately following W56 in RhaK and generates the 5-amino-acid insertion CLNTF (Fig. 2). The region of I52 to A69 of RhaK is predicted with high confidence levels to form a 17-amino-acid alpha helix. With the inclusion of the 5 amino acids generated from the insertion, the secondary structure in this area is still predicted with high confidence levels to be an alpha helix. Placement of this helix in the RhaK model predicts this to be a surface-exposed helix that is found in the N-terminal lobe of RhaK (Fig. 5C).

Taking these observations together, both inserts that have uncoupled the kinase activities from the transport of rhamnose into the cell map to two distinct regions of the protein. It is noteworthy that the predicted changes for both of the insertions result in changes on the same protein face (Fig. 5D), suggesting that perhaps this protein face plays a role in the protein-protein interaction of RhaK with other protein(s).

Based on the generated model and the positioning of the rhaK72 allele, it was noted that closely related RhaK orthologs contained a conserved amino acid sequence (R358, E359, E360, and R361). To test if this region is directly involved in a protein-protein interaction, amino acids 358 to 361 were replaced with alanines, yielding allele rhaK122. Testing of this construct showed that the growth phenotype of Rlt144 carrying rhaK122 was indistinguishable from that of the wild type. In addition, kinase and transport rates were also comparable to those of the wild type, indicating that no significant effect was generated by this substitution (data not shown).

DISCUSSION

In this paper we have initiated a characterization of how RhaK affects transport of rhamnose into the cell. Two hypotheses regarding how RhaK may affect transport were previously put forward (23): the first is that rhamnose itself may be a negative regulator that affects its own ABC transporter, and the second is that RhaK, in addition to the known catabolic activity, also affected the activity of the ABC transporter. The isolation of two alleles of rhaK that retain kinase activity and have lost the ability to confer in vivo transport demonstrates that these activities can be uncoupled (Fig. 3). Moreover, these data provide direct evidence that the mechanism by which RhaK affects rhamnose transport in R. leguminosarum is not due to a negative regulatory effect of the rhamnose on its own ABC transporter.

In an attempt to demonstrate direct interaction of RhaK with its transporter, we were able to demonstrate that a small proportion of RhaK that had rhamnose kinase activity could be found loosely associated with the membrane fraction (Fig. 1). We note that this association was not dependent on the presence of the ABC transporter components and that this was only tested with rhaK expressed from a low-copy-number plasmid. The significance of this result is not clear at present and may be resolved with further study.

Prior to this work, the only RhaK variant that was constructed was in the P-loop region which is used for the binding of ATP (23). These variants eliminated both transport and kinase activity (23). Assuming that the ability to affect kinase and transport were independent, we had anticipated isolating four mutant classes from our mutagenesis of rhaK; these were variants that had neither activity affected, variants that had both activities affected, and variants that had one or the other affected. The results of our analysis show that we did not isolate alleles that encoded variants that retained the ability to transport but lost kinase activity. Although we achieved our anticipated coverage of the open reading frame following our mutagenesis, we note that a number of our inserts generated stop codons, leaving the possibility that this class of variant may still exist. Alternatively, it could be possible that the ability to affect transport may be linked to the ability of RhaK to bind ATP or carry out kinase activity. It is not unprecedented for protein-protein interactions to be dependent on phosphorylation state (41, 42).

The alleles rhaK72 and rhaK73 confer distinct phenotypes when they are used to complement an rhaK mutant (Tables 2 and 3 and Fig. 3 and 4). It could be that these two mutations affect two separate interactions or that both alleles affect a single protein-protein interaction. Although the growth differences are clear (Table 3), it could be that the rhaK72 allele allows some residual transport activity that is not above the level of detection in our assays. We do not have data that would allow us to favor either of these scenarios at this time.

Two alleles, rhaK72 and rhaK73, encoded variants that retained kinase activity but were unable to complement transport activity in vivo (Fig. 3). Mapping of these inserts showed that the 5-amino-acid insertion (CLNTF) that defined rhaK73 followed W56, whereas the rhaK72 insertion (VFKHH) was closer to the C-terminal quarter of the protein after H362. Aligning of RhaK to closely related kinases using Clustal-X (37) and Praline (38) alignments was useful in identifying residues that were involved in substrate binding and catalysis compared to the well-studied glycerol kinase (GlpK) (44). Both rhaK72 and rhaK73 insert 7 and 8 amino acid residues prior to conserved leucine residues. Structural modeling of RhaK using the Phyre² server produced a model that had 97% coverage and 100% confidence levels compared to GlpK. Mapping of these two insertional variants suggests that both appear to be in regions that are strongly predicted to be helices and that the resulting changes appear as changes on the same surface of the protein. Further construction and testing of the REER motif that is conserved in closely related RhaK orthologs, however, did not change RhaK's ability to affect rhamnose transport. Whereas the insertional variant rhaK72 moved residues farther apart, the rhaK122 allele targeted a group of charged amino acids and changed these to uncharged neutral amino acids. It could be that amino acid spacing rather than charge is more important for the function of this area. Alternatively, this particular region is not important to transport activity. These hypotheses are currently being addressed.

This work was carried out to test the hypothesis that kinase and transport activities are affected by RhaK. Isolation of the rhaK72 and rhaK73 alleles of R. leguminosarum rhaK clearly shows that these activities can be separated. Our working hypothesis is that RhaK interacts via protein-protein interaction to affect the rhamnose ABC transporter. It is not clear whether this will occur as a direct or indirect interaction. Precedent for interacting proteins affecting transport exists. In particular, MalK has been shown to interact with EIIA^Glc (45) and the maltose regulator MalT (13, 46).

Our work is now focusing on identifying proteins that may interact with RhaK to affect rhamnose transport. In addition, we have embarked on generating a crystal structure for RhaK to be used for more detailed structure-function studies to define the motif(s) on RhaK that is used to affect rhamnose transport by the CUT2 ABC transporter RhaSTPQ.