Functional interactions between the subunits of the lactose transporter from Streptococcus thermophilus

Although the quaternary state has been assessed in detail for only a few members of the major facilitator superfamily (MFS), it is clear that multiple oligomeric states are represented within the MFS. One of its members, the lactose transporter LacS from Streptococcus thermophilus assumes a dimeric structure in the membrane and in vitro analysis showed functional interactions between both subunits when proton motive force ( D p)-driven transport was assayed. To study the interactions in further detail, a covalent dimer was constructed consisting of in tandem fused LacS subunits. These covalent dimers, composed of active and completely inactive subunits, were expressed in Escherichia coli , and initial rates of D p-driven lactose uptake and lactose counterﬂow were determined. We now show that also in vivo , both subunits interact functionally; that is, partial complementation of the inactive subunit was observed for both transport modes. Thus, both subunits interact functionally in D p-driven uptake and in counterﬂow transport. In addition, analysis of in tandem fused LacS subunits containing one regulatory LacS-IIA domain showed that regulation is primarily an intramolecular event.


Introduction
The major facilitator superfamily (MFS) forms the largest family of secondary transporters. Its members, found in all living organisms, catalyze the translocation of a variety of substrates among which are sugars, peptides, anions, cations and hydrophobic compounds. Most MFS proteins consist of 12 a-helical transmembrane-spanning segments (TMS) that are connected by loop regions ranging from a few amino acid residues to several dozens of residues. 1 Although data on functional and structural aspects of these proteins are produced at a great pace, information on the quaternary state(s) is scarce. For both the lactose transporter LacY 2 and the sugar-phosphate antiporter UhpT 3 from Escherichia coli, the monomeric unit is sufficient for transport, and earlier claims for LacY dimers are thought to be based on improper experimentation and inaccurate data interpretation. For the anion-exchanger AE1 4,5 and glucose transporter GLUT1 6 from human, and the tetracycline transporter TetA 7,8 from E. coli, on the other hand, there is compelling evidence that these transporters form higher oligomeric structures, and functional interactions between the subunits of the oligomers have been shown.
For secondary transporters outside the MFS, oligomeric states higher than the monomer have been proposed (e.g. the Na C /H C antiporter NhaA 9 and the multidrug exporter AcrB 10 from E. coli, the Na C /glycine betaine symporter BetP 11 from Corynebacterium glutamicum, and the human glial glutamate transporter hEAAT2 12,13 ). For most of these transporters, it remains to be determined whether functional interactions between subunits take place. Furthermore, in those cases where cooperativity between subunits has been shown, the role of the oligomeric structure in the mechanism of transport is still far from clear.
Members of the galactoside-pentoside-hexuronide transporter family (GPH family) of the MFS 14 are of pro-or eukaryotic origin and catalyze the transport of sugars and sugar 0022-2836/$ -see front matter q 2005 Elsevier Ltd. All rights reserved.
E-mail address of the corresponding author: b.poolman@rug.nl derivatives in symport with cations. 15,16 These transporters consist of 12 a-helical transmembrane segments that span the membrane in a zig-zag fashion. Detailed analysis of the oligomeric state has been conducted only for the lactose transporter LacS from Streptococcus thermophilus, and this system forms a structural and functional dimer. For the xyloside transporter XylP from Lactobacillus pentosus, a dimeric structure has been reported. On the other hand, the projection structure of another GPH family member, the melibiose permease MelB from E. coli, does not comply with a dimeric state. 17 Both XylP and LacS have been analyzed in the detergent-solubilized state using blue native gel electrophoresis and analytical ultracentrifugation, and shown to be in a dynamic monomer to dimer equilibrium. 13,18 The extent of dimerization could be manipulated by varying the detergent type or concentration. The membrane-embedded oligomeric state of XylP and LacS has been determined using freeze-fracture electron microscopy, which suggested that both proteins are present in the bilayer as dimers only. 13,18 Additional evidence for this dimeric state of LacS in the membrane came from saturation-transfer electron spin resonance studies. 19 Chemical cross-linking of cysteine substitution mutants of LacS in situ suggested that on the extracellular side of LacS, TMS V and VIII, and on the cytoplasmic side TMS VI and VII, are located near the centre of the LacS dimer. 20 By studying LacS heterodimers of active (LacS(C320A)) and (conditionally) inactive (LacS(E67C/C320A)) species in proteoliposomes, it has been shown that the subunits within the dimer cooperate. 21 Cooperativity was observed during lactose accumulation driven by the proton motive force (Dp), but not for the lactose exchange mode of transport.
Within a LacS subunit, two domains can be discerned. The N-terminal membrane-embedded carrier domain catalyzes the translocation event. The C-terminal hydrophilic LacS-IIA domain, unique for the LacS subfamily within the GPH family, 15 is homologous to IIA Glc domains of the PEP phosphotransferase system and resides at the cytoplasmic face of the membrane. The LacS-IIA domain is not essential for transport, but serves a regulatory role. Phosphorylation of LacS-IIA by HPr(HiswP) enables the domain to interact with the carrier domain and modulate the transport activity (our unpublished results). 22,23 In order to increase our insights into the functional role(s) of the subunit interactions, heterodimers were formed by fusing two different LacS carrier domains in tandem. In whole E. coli cells, functional interactions between a LacS(D71C/ C320A) subunit, which is inactive in all modes of transport but still adopts a conformation capable of substrate-binding, 24 and the active LacS(C320A) subunit were analyzed. In addition, the regu-lation of the LacS carrier domain by LacS-IIA was studied in an asymmetric covalent dimer, in which only one subunit was equipped with a LacS-IIA domain.

Results
A covalent dimer to study subunit interactions Functional interactions between subunits in Figure 1. Comparison between the calculated amount of heterodimers and total activity for free and forced association of two species. (a) The black line depicts the percentage of heterodimers formed as a function of different ratios of active over inactive species, assuming random association. The circles connected by the gray line show the percentage of heterodimers if the subunits are covalently linked. (b) The summed activity of all species as a function of different ratios of active over inactive species. The dotted line represents the activity observed if the subunits function independently. Functional dominance of one type of subunit will lead to the continuous curves a and b, representing negative and positive dominance, respectively. Upon fusion of both subunits, these scenarios are represented by the circles connected by the broken lines. Note that for the forced association of the subunits, only three and no intermediate situations exist. dimeric proteins have been studied by mixing active and inactive species in different ratios and characterizing the resulting heterodimers, 21,[25][26][27] which constitute at most 50% of all dimers; that is, at an equal ratio of both species, as shown in Figure 1(a) (continuous line). If the association is random, each of the homodimers makes up 25% of all dimers at an equal ratio of both species. If two species do not interact or function independently, the total activity is determined only by the percentage of active species and will decrease linearly ( Figure 1(b), dotted line). However, if both species do interact functionally and the phenotype of one of the species dominates the activity of the heterodimer, the activity will follow a quadratic relationship, as shown in Figure 1(b) (continuous lines); lines a and b are obtained when a subunit has a negative-or positive-dominant effect on the opposing subunit, respectively.
To determine if the subunits function in a cooperative manner, the specific activity of the heterodimer needs to be resolved. When the exact ratio of the two species is known (e.g. in a proteoliposomal system) this can be determined from the summed activity of all species. However, in whole cells, the ratio of two separately expressed subunits is difficult to control, because of variations in the expression and inaccuracies in the determination of protein levels from, for instance, immunoblots. By covalent coupling of subunits, the ratio is known beforehand, which has the additional advantage that the maximum percentage of heterodimers can be elevated from 50% to 100% (see Figure 1(a)). This increases the signal that discriminates between independent functioning and negative or positive-dominant effects of subunits, as shown in Figure 1(b).
In order to control the total amount of protein in a reproducible way, expression of the covalent dimer in E. coli MC1061 was governed by the arabinoseinducible P BAD promoter, which proved to be a more convenient and reliable expression system than the vector wherein LacS expression was controlled by its endogenous promoter (our unpublished results).

Construction and functional expression of a covalent LacSDIIA 2 dimer
The minimal unit of LacS able to catalyze substrate translocation is the membrane-embedded carrier domain (our unpublished results). 22 To study the functional interactions of the carrier domains in a defined manner, the covalent LacS dimer comprised two joined LacSDIIA subunits rather than two LacS subunits. LacSDIIA lacks the C-terminal LacS-IIA domain that can be phosphorylated by HPr(HiswP) (our unpublished results), 22,23 but shows equal rates of Dp-driven lactose transport and lactose counterflow as unphosphorylated full-length LacS (our unpublished results).
Twenty-eight amino acid residues follow the predicted end of TMS 12 of the first LacSDIIA(C320A) subunit. An artificial sequence of 15 amino acid residues was used to link the two subunits. At the DNA level, the linker region contains four endonuclease restriction sites and a sequence coding for a TEV protease recognition site, yielding the protein sequence GSGDQENLYFQGTSA. Together with the 18 amino acid residues preceding the predicted start of the first TMS of the second LacS-DIIA(C320A) subunit, the total linker region connecting both subunits comprises 61 amino acid residues ( Figure 2). This covalent dimer, in which both subunits contain the C320A mutation, was designated LacSDIIA 2 (CC).
As shown in Figure 3, LacSDIIA 2 (CC) catalyzed Dp-driven lactose uptake in whole E. coli MC1061 cells, demonstrating its functional expression and membrane insertion. Maximal transport activity of LacSDIIA 2 was observed when cells were induced with 2!10 K3% (w/v) L-arabinose, whereas induction with 1!10 K3% (w/v) L-arabinose yielded maximal transport activity of LacSDIIA (results not shown). At these optimal concentrations of inducer, the initial rates of Dp-driven lactose uptake for LacSDIIA 2 (CC) and LacSDIIA(C320A) were 2.6 nmol of lactose/mg of protein per minute and 11 nmol of lactose/mg of protein per minute, respectively. Analysis of the levels of LacSDIIA 2 Figure 2. Topology model of the LacSDIIA 2 dimer. Membrane topology of the LacSDIIA subunits is based on the model of MelB. 33 The gray horizontal lines indicate the membrane interfaces. Asp71 in the second transmembrane segment (TMS) of each subunit is depicted in black. The LacSDIIA subunits are coupled via a linker of 61 amino acid residues (gray circles). The sequence of the artificially introduced stretch of 15 residues in the linker is shown. In all subunits, the endogenous Cys320 in TMS IX (also shown in black) was replaced by an alanine residue.
(CC) and LacSDIIA(C320A) showed that the former was indeed expressed several-fold lower (data not shown).
Inactive D71C/C320A subunits are complemented by active C320A subunits within LacSDIIA 2 during Dp-driven and counterflow transport Heterodimeric LacSDIIA 2 derivatives were constructed, comprising the D71C mutation in either the first or the second subunit and using the C320A background. The D71C mutation renders LacS inactive in all modes of transport, but does not affect the overall structure of the transporter, as the capacity of LacS(D71C/C320A) to bind substrate was retained. 24 The LacSDIIA 2 derivatives containing the D71C/C320A mutation in the first, last, or both subunits were designated LacSDIIA 2 (DC), LacSDIIA 2 (CD), and LacSDIIA 2 (DD), respectively. The mutant variants of LacSDIIA 2 were expressed to comparable levels ( Figure 4(a)), enabling a direct comparison of the transport activities.
Like the strain containing a control plasmid, cells expressing LacSDIIA 2 (DD) showed no significant uptake of lactose in whole E. coli MC1061 cells ( Figure 3). The transport rates of LacSDIIA 2 (CD) and LacSDIIA 2 (DC) were equal, indicating that both halves of the forced dimer are correctly inserted in the membrane. The observed rates of Dp-driven lactose uptake of LacSDIIA 2 (CD) and LacSDIIA 2 (DC) were approximately 80% of the initial rate of uptake of LacSDIIA 2 (CC) (Figures 3 and 5). This result suggests that the D71C/C320A subunit is partially complemented by the active C320A subunit.
Lactose counterflow transport by the LacSDIIA 2 derivatives showed a profile similar to that observed for Dp-driven lactose uptake ( Figure 5). LacSDIIA 2 (DD) was completely defective in transport, and both   . Initial rates of Dp-driven lactose uptake and lactose counterflow by LacSDIIA 2 derivatives in E. coli MC1061. Counterflow and Dp-driven lactose transport rates were measured on cell suspensions derived from the same culture. To measure lactose counterflow transport, de-energized cell suspensions were preloaded with 10 mM lactose and diluted into buffer containing 100 mM [ 14 C]lactose. Cells used to measure lactose transport driven by the Dp were pre-energized for two minutes by incubation with 10 mM D-Li-lactate and aeration. Subsequently, lactose accumulation was started by the addition of 50 mM [ 14 C]lactose. Each rate reflects the average from two independent experiments. Open and hatched bars reflect initial lactose uptake rates for Dpdriven and lactose counterflow transport, respectively. LacSDIIA 2 (CD) and LacSDIIA 2 (DC) showed approximately 80% of the initial rate of lactose counterflow by LacSDIIA 2 (CC), indicating that for counterflow transport also the D71C/C320A subunit is partially complemented by the active C320A subunit.
Covalent fusion of LacSDIIA subunits differently affects counterflow and Dp-driven lactose transport Under identical experimental conditions, noncovalently linked LacS(C320A) showed approximately half the rate of Dp-driven transport compared to lactose counterflow (9.0 nmol of lactose/mg of protein per minute and 20 nmol of lactose/mg of protein per minute, respectively), while LacSDIIA(C320A) showed comparable rates for both modes of transport (11 nmol of lactose/mg of protein per minute and 8.6 nmol of lactose/mg of protein per minute). This is most probably caused by a stimulation of the counterflow reaction of LacS(C320A) by HPr(HiswP) mediated phosphorylation of LacS-IIA (our unpublished results).
Whereas the ratio of the initial rates of Dpdriven transport over counterflow was 0.5 and 1.3 for LacS and LacSDIIA, respectively, this ratio was approximately 3 for the CC, CD and DC derivatives of LacSDIIA 2 (summarized in Table 1). Since the initial rates of both Dp-driven and counterflow lactose transport were determined on cells derived from the same culture, the discrepancy in relative activities cannot be caused by variation in expression levels. Effects of variations in internal pH between both transport modes on LacSDIIA 2 derivatives could be excluded, because the pH of the buffer during lactose counterflow transport was adjusted to 7.7, which was equal to the internal pH of cells during Dp-driven transport. 28 Furthermore, by charging the cells with [ 14 C]lactose, it was shown that the MC1061 cells containing LacSDIIA 2 were equilibrated with lactose to levels similar to that of cells containing non-covalently linked LacSDIIA(C320A), ruling out limiting intracellular lactose concentrations as a cause for the decreased counterflow transport rates. We, therefore, conclude that the threefold difference in initial rates of Dp-driven lactose transport and lactose counterflow transport by the CC, CD, and DC derivatives of LacSDIIA 2 is a genuine property of these dimers.
A more detailed kinetic analysis of lactose exchange transport, which comprises the same kinetics steps as counterflow transport and differs from counterflow only in substrate concentrations and the location of the 14 C isotope of lactose (inside for exchange, outside for counterflow), was conducted ( Figure 6). All LacSDIIA 2 derivatives showed similar lactose exchange kinetics, again reflecting that all subunits (irrespective of the order of active C320A and inactive D71C/C320A subunits) were affected equally by the covalent linkage. Moreover, the apparent affinity constants of the LacSDIIA 2 derivatives for lactose were maximally increased only twofold compared to the K m app of LacSDIIA, suggesting that the overall structure of both subunits in the LacSDIIA 2 derivatives is conserved. Summarizing, it seems that both subunits within the LacSDIIA 2 derivatives adopt a correct conformation. The LacSDIIA C320A and D71C/C320A subunit are presented as a rectangle, and a rectangle filled with a cross, respectively. The LacS-IIA domain is depicted as a flattened sphere. The initial transport rates were determined as described in the legend to Figure 5. a Values within parentheses are percentages.
The LacS-IIA domain interacts primarily with the subunit of LacSDIIA-LacS to which it is attached Upon HPr(HiswP)-mediated phosphorylation of His552 in the LacS-IIA domain, the LacS-IIA domain interacts with the carrier and thereby stimulates lactose counterflow (our unpublished results). 23 To determine whether the regulation of the LacS-IIA domain occurs inter-or intramolecularly, a tandem fusion was constructed in which the last subunit has a C-terminal LacS-IIA domain attached to the carrier domain, creating a topology equal to the full-length LacS protein. This construct is designated LacSDIIA-LacS. The mutant variants derived from this construct, harboring the D71C mutation, were not expressed to an equal level, as shown in Figure 4(b). Constructs containing the D71C substitution in the second subunit were expressed to a higher level than subunits without a D71C mutation. This variation in expression levels allowed only a qualitative analysis, since exact quantification of the amount of functional transporters was not possible at this stage, due to the lack of a suitable ligand-binding assay.
The rate of Dp-driven lactose uptake of LacSDIIA-LacS(CC) was in the same range as the rate of LacSDIIA 2 (CC) (2.8 nmol of lactose/mg protein per minute and 2.6 nmol of lactose/mg protein per minute, respectively) ( Figure 7). Furthermore, the kinetics of lactose exchange by LacSDIIA-LacS(CC) yielded an apparent affinity constant for lactose similar to that of the other LacS derivatives ( Figure 6). In line with the observations on the LacSDIIA 2 derivatives, both LacSDIIA-LacS(CD) and LacSDIIA-LacS(DC) were active in Dp-driven transport, whereas LacSDIIA-LacS(DD) was inactive ( Figure 7). As described above, the ratio of the initial rates of Dp-driven transport over counterflow was 0.5 and 1.3 for LacS and LacSDIIA, respectively, while this ratio was close to 3 for the CC, CD and DC derivatives of LacSDIIA 2 (summarized in Table 1). Both LacSDIIA-LacS derivatives that had a LacS-IIA domain attached to the active C320A subunit (LacSDIIA-LacS(CC) and LacSDIIA-LacS(DC)) showed a ratio of Dp-driven transport over counterflow of approximately 0.6. In contrast, this ratio was elevated to 1.7 for LacSDIIA-LacS(CD). Since the ratio differs for the LacSDIIA-LacS(CD) and LacS-DIIA-LacS(DC) constructs, this suggests strongly that the IIA domain is restricted in its opportunity to interact with the carrier domains within the LacS dimer. Most likely, it cannot interact with any nearby subunit but only with the subunit to which it is attached.

Discussion
Several independent studies led to the conclusion that the detergent-solubilized LacS protein is in monomer to dimer equilibrium, 18,20,29 whereas  . Initial rates of Dp-driven lactose uptake and lactose counterflow by LacSDIIA-LacS derivatives in E. coli MC1061. Counterflow and Dp-driven lactose transport rates were measured on concentrated cell suspensions derived from the same culture as described in the legend to Figure 5. Each rate reflects the average from two independent experiments. Open and hatched bars reflect initial lactose uptake rates for Dp-driven and lactose counterflow transport, respectively. membrane-embedded LacS is dimeric. [18][19][20] Veenhoff et al. showed by in vitro analysis of membranereconstituted (conditionally) inactive LacS(E67C/ C320A) and active LacS(C320A) species that, within this structural dimer, functional interactions between the subunits take place in Dp-driven uptake but not in counterflow transport. 21 Here, we report the first in vivo demonstration of functional interactions between the subunits within the LacS dimer and show that both subunits functionally interact in Dp-driven transport and in counterflow. We show that: (i) the activity of the D71C mutant is (partially) restored by an active subunit; (ii) covalent linkage of two LacS subunits increases the rate of Dp-driven uptake relative to counterflow transport; and (iii) phosphorylation of the LacS-IIA domain stimulates the cis rather than the trans subunit of dimeric LacS.
Due to its uncoupled phenotype in whole cells, resulting in rapid efflux of lactose down the concentration gradient, LacS(E67C/C320A), previously used to demonstrate functional interactions within the LacS dimer in vitro, 21 could not be used to study subunit interactions in whole cells. Instead, the translocation-defective but lactose bindingcompetent LacS(D71C/C320A) was employed. Both Glu67 and Asp71 are located in TMS II, and acidic residues at these positions are highly conserved throughout the GPH family. Glu67 has been proposed to be involved in coupling proton and galactoside transport, whereas Asp71 is thought to contribute to the proton binding site. 15 In addition, both residues are in a region that is conformationally active upon substrate-binding. 24 Membranereconstituted LacS(E67C/C320A) catalyzes lactose counterflow transport at w80% of the rate of LacS(C320A) and after modification of Cys67 with NEM w30% activity remained. LacS(E67C/C320A) is completely defective in Dp-driven lactose transport. 21 In contrast, LacS(D71C/C320A) is inactive in both Dp-driven lactose transport and lactose counterflow, but the overall structure of the protein is retained as substrate-binding still occurs. Several second-site suppressor mutations of Cys71 have been isolated, among which R230C, which could restore lactose counterflow transport but not Dpdriven lactose transport. 24 By employing LacS(D71C/C320A) in tandem constructs with LacS(C320A), the effect of a subunit defective in all transport modes could be determined.
The LacSDIIA 2 derivatives were expressed functionally, although the levels were lower than those of the free subunits, as observed before for other in tandem fusion proteins. 2,26 The expression levels of the different LacSDIIA 2 derivatives were equal, as judged by immunoblotting (Figure 4(a)), but this technique lacks the sensitivity to demonstrate small variations (!20%) in expression levels. The LacSDIIA 2 (CC) was functional in both Dp-driven and lactose counterflow transport, and both subunits are membrane-inserted, as LacSDIIA 2 (CD) and LacSDIIA 2 (DC) showed similar transport rates. LacSDIIA 2 (DD), which resembles the LacS(D71C/C320A) homodimer, is inactive, indicating that the mere coupling of the subunits is not sufficient for the restoration of the transport capacity. The difference in the ratio between the initial rates of Dp-driven lactose transport and lactose counterflow for the LacSDIIA 2 derivatives when compared with LacSDIIA and LacS, suggest that either Dp-driven lactose transport benefits or lactose counterflow activity deficits from the fusion of the subunits (Table 1). Kinetic analysis of lactose exchange transport showed that substrate-binding of the LacSDIIA 2 derivatives was affected only slightly, because the K m app was at most twofold higher than the K m app of the LacSDIIA dimer. Taken together, the covalent coupling of subunits introduces only some moderate changes within the LacS dimer and seems a valid system to study subunit interactions.
For independent operating subunits within the LacS dimer, the heterodimers LacSDIIA 2 (CD) and LacSDIIA 2 (DC) would be expected to show half the initial transport rates of LacSDIIA 2 (CC). A negative dominant phenotype of the D71C subunit would yield inactive LacSDIIA 2 (CD) and LacSDIIA 2 (DC) heterodimers. Clearly, both scenarios do not apply, since the initial rates of LacSDIIA 2 (CD) and LacSDIIA 2 (DC) for both Dp-driven lactose transport and lactose counterflow were approximately 80% of the rates of LacSDIIA 2 (CC) ( Table 1). Rather, these transport activities support the contention that the C320A subunit has a positive-dominant effect, since it is able to partly restore the activity of the LacSDIIA(D71C/C320A) subunit.
Our new findings differ in some aspects from previous observations, using a proteoliposomal system, which demonstrated functional interactions between the LacS subunits for Dp-driven lactose transport only. 21 In proteoliposomes, increasing the percentage of impaired LacS(E67C/C320A) over LacS(C320A) resulted in a linear decrease in lactose counterflow transport, suggesting that the subunits function independently during this mode of transport. Furthermore, heterodimers composed of inactive LacS(E67C/C320A) and active LacS(C320A) were shown to be completely inactive in Dp-driven lactose transport, indicating negative dominance of the LacS(E67C/C320A) subunit. In the present study, negative dominance of the LacSDIIA(D71C/ C320A) subunit in LacSDIIA 2 (CD) or (DC), is clearly not observed. Rather than interpreting these discrepancies in terms of differences in the experimental context (proteoliposomes versus intact cells), we feel that they can be explained by the different phenotypes of the LacS(E67C/C320A) and LacS (D71C/C320A) mutants. Most likely, LacS(D71C/ C320A) is locked in one conformation, since it has lost its ability to reorient its substrate-binding sites as required for transport while retaining the ability to bind substrate. 24 An opposing functional subunit could enable the LacSDIIA(D71C/C320A) to overcome this locked conformation, leading to (partial) complementation. LacS(E67C/C320A), on the other hand, kept the ability to reorient its binding sites and thereby catalyze lactose counterflow transport. Only its ability to catalyze Dp-driven lactose transport was impaired. As the extent of the impairment is different for both mutants, this may explain the lack of complementation of LacS(E67C/ C320A) by LacS(C320A).
Apart from the relative activities of the different LacSDIIA 2 derivatives for either Dp-driven lactose uptake or lactose counterflow, the ratio of the initial rates of Dp-driven lactose transport over counterflow reveals interesting features. This ratio is approximately 3 for the CC, CD and DC derivatives of LacSDIIA 2 , while it is close to 1 for LacSDIIA. It is likely that the difference in the ratio of the initial rates for LacSDIIA 2 compared to LacSDIIA is a specific property of the fusion protein, caused by the presence of the linker connecting both subunits. A possible consequence of the covalent coupling could be the decreased ability of the subunits to separate transiently.
Additionally, the mere presence of the LacS-IIA domain to the second subunit decreased the ratio of Dp-driven transport over counterflow from w3 to 0.6-1.7, indicating that phosphorylation of this domain affects the subunit interactions. Since the absolute rates of Dp-driven lactose transport of LacSDIIA 2 (CC) and LacSDIIA-LacS(CC) are similar, it seems that the rate of lactose counterflow transport is increased upon addition of the LacS-IIA domain. Therefore, it is tempting to speculate that the kinetic step that is impaired by the coupling of the subunits is the same as the one that is stimulated by the phosphorylation of the LacS-IIA domain. Thus, the phosphorylated LacS-IIA domain could induce a transient conformation of the carrier domain that is less strongly interacting with the opposing subunit.
Under the conditions used, the IIA-domain of LacS is phosphorylated in E. coli MC1061, resulting in an increase in the rate of counterflow transport and a decrease in the ratio of Dp-driven transport over counterflow (our unpublished results). Phosphorylation of the LacS-IIA domain in the LacSDIIA-LacS derivatives resulted in a ratio of Dp-driven transport over counterflow near 0.6 if the LacS-IIA domain was associated with the LacSDIIA(C320A) subunit (Figure 7), and near 1.7 if it was associated with the inactive D71C subunit. The difference in the ratio of both transport modes observed for LacSDIIA-LacS(CD) and LacSDIIA-LacS(DC) suggests that the LacS-IIA domain prefers to functionally interact with the subunit to which it is attached. Although the rate of lactose counterflow is increased relatively for LacSDIIA-LacS(CD), the increase is smaller than that observed for the CC and DC derivatives of LacSDIIA-LacS, indicating that the D71C/C320A subunit cannot be stimulated by phosphorylated LacS-IIA to the same extent as the C320A subunit.
Taken together, the data presented here strengthen the conclusion that the two subunits of the LacS dimer interact functionally, as separate proteins in vitro and in tandem fusions in vivo. The functional interactions within the LacS dimer take place during both Dp-driven lactose transport and lactose counterflow, and can be both positive and negative of nature. Furthermore, the LacS-IIA domain primarily stimulates transport through intramolecular interactions with the carrier domain.

Materials and Methods
Bacterial strain E. coli JM110 30,31 was used for intermediate cloning steps. The final constructs were expressed in E. coli MC1061 (relevant genotype: DlacZY, araBADC K ). 32 Both strains were cultivated at 37 8C on Luria broth under vigorous aeration. When appropriate, the medium was supplemented with 50 mg/ml of ampicillin.
In order to construct the plasmid containing the second subunit, pNlacSC320Ahis was digested with BclI-NcoI and the 389 bp fragment was replaced by a linker of two annealed oligonucleotides (linkerBN and linkerNB; see Table 2) with extensions resembling a BclI or an NcoI overhang, yielding pBADsub2C320ACIIA. The sequence coding for the LacS-IIA-domain was removed by exchange of the 2229 bp AatII-XbaI fragment for the 1723 bp AatII-XbaI fragment from pSKlacSC320A (/D71C)DIIA, producing pBADsub2C320A(/D71C)DIIA.
To link both subunits, the 34 bp BclI-XbaI fragment from pBADsub1C320A(/D71C) was exchanged for the 2333 bp BclI-XbaI fragment from pBADsub2C320A (/D71C)DIIA containing the second subunit, yielding pBADlacSDIIAlacSDIIA. Four derivatives of pBADlacS-DIIAlacSDIIA were generated, containing the D71C mutation in the first, the last, or both subunits.

pBADlacSDIIAlacSCIIA
The construction of the vector harboring a gene coding for a fusion between a LacS carrier domain and fulllength LacS was similar to the construction of the vector for two fused LacS carrier domains. Instead of pBAD-sub2C320A(/D71C)DIIA, plasmid pBADsub2C320A (/D71C)CIIA was used. The D71C mutation was added by exchanging the 2229 bp AatII-XbaI fragment of pBADsub2C320ACIIA for the 2229 bp AatII-XbaI fragment of pSKE8EhisC320A/D71C.

Whole cell transport assays
Cultivation E. coli MC1061 cells were cultivated, washed and concentrated as will be described elsewhere. Cultures were induced with 1!10 K3 % and 2!10 K3 % (w/v) L-arabinose when LacS derivatives or LacS-LacS fusion proteins were expressed, respectively. Cells used for lactose exchange transport were induced with 1.5!10 K4 % (w/v) L-arabinose when LacS or LacSDIIA were expressed. Concentrated cell preparations were kept on ice until lactose transport was assayed.

Transport assays
General handlings involved in lactose transport in E. coli MC1061 cells and the preparation of cells to be used for Dp-driven lactose uptake and lactose counterflow transport will be described elsewhere.

Lactose exchange transport
Cells were concentrated to 45 mg protein/ml and incubated overnight in KPM (50 mM KPi (pH 7.7), 2 mM MgSO 4 ) plus 5 mM [ 14 C]lactose. The next day, cells were de-energized by incubation with the protonophore SF6847 (50 mM) plus 30 mM NaN 3 for two hours. Lactose exchange was assayed at 20 8C by 100-fold dilution of the cells into KPM, supplemented with 50 mM SF6847. The external lactose concentration varied from 50 mM to 20 mM.

Membrane vesicle isolation
Inside-out membrane vesicles from E. coli MC1061 cells were prepared as described. 20 Membrane vesicles were resuspended in 50 mM KPi (pH 7), plus 3 mM DTT, frozen in liquid nitrogen and stored at K80 8C. The protein concentration was determined using the DC protein assay (Bio-Rad).

Purification and immunodetection of LacS-LacS fusion proteins
All steps during the purification were performed at 4 8C. E. coli MC1061 membrane vesicles (approximately 6 mg of total protein) containing LacS fusion proteins were deprived of DTT by washing and subsequently solubilized as described. 20 Next, the insoluble fraction was removed by centrifugation at 267,000 g for 15 minutes) and the supernatant was mixed with 0.25 ml of Ni-NTA resin that was washed with ten volumes of MilliQ water and two volumes of elution buffer (200 mM imidazole (pH 7.0), 10% (v/v) glycerol) and pre-equilibrated with four volumes of solubilization buffer (15 mM imidazole (pH 8.0), 100 mM NaCl, 10% (v/v) glycerol) supplemented with 0.05% (w/v) DDM. The mixture was incubated for one hour with continuous mixing. After that, the column was drained, washed with 30 volumes of solubilization buffer plus 0.05% DDM and eluted with elution buffer plus 0.05% DDM.
Samples were analysed by SDS-PAGE, semi-dry electroblotting and subsequent immunodetection with a primary antibody directed against a hexa-His-tag (Amersham Pharmacia Biotech) as described. 20 LinkerBS and LinkerSB constitute a double-stranded linker with artificial BamHI and SpeI overhangs. LinkerBN and LinkerNB constitute a double-stranded linker with artificial BclI and NcoI overhangs. The sequence of both double-stranded linkers partially overlaps; double-stranded linkers were created by mixing both oligonucleotides in equal ratios, boiling for five minutes and annealing by cooling slowly to room temperature.