The Hydrophilic N-terminal Domain Complements the Membrane-anchored C-terminal Domain of the Sensor Kinase KdpD of Escherichia coli *

The putative turgor sensor KdpD is characterized by a large, N-terminal domain of about 400 amino acids, which is not found in any other known sensor kinase. Comparison of 12 KdpD sequences from various micro-organisms reveals that this part of the kinase is highly conserved and includes two motifs (Walker A and Walker B) that are very similar to the classical ATP-binding sites of ATP-requiring enzymes. By means of photoaffinity labeling with 8-azido-[ a - 32 P]ATP, direct evidence was obtained for the existence of an ATP-bind-ing site located in the N-terminal domain of KdpD. The N-terminal domain, KdpD/1–395, was overproduced and purified. Although predicted to be hydrophilic, it was found to be membrane-associated and could be solubilized either by treatment with buffer of low ionic strength or detergent. The membrane-associated form, but not the solubilized one, retained the ability to bind 8-azido-[ a - 32 P]ATP. Previously, it was shown that the phosphatase activity of a truncated KdpD, KdpD/ D 12– 395, is deregulated in vitro (Jung, K., and Altendorf, K. (1998) J. Biol. Chem. 273, 17406–17410). Here, we demonstrated that this effect was reversed in vesicles containing both the truncated KdpD and the N-terminal domain. Furthermore, coexpression

The membrane-bound sensor kinase KdpD and the soluble response regulator KdpE comprise a sensor kinase/response regulator system of Escherichia coli, which regulates expression of the kdpFABC operon (1) encoding the high affinity K ϩ -translocating complex KdpFABC (see Ref. 2 for review). KdpD consists of a long hydrophilic N-terminal domain, four putative transmembrane domains (TM1-TM4), 1 and an extended C-terminal transmitter domain (see Fig. 1

) (3).
KdpD undergoes autophosphorylation at a histidine residue in the presence of ATP, and subsequently, the phosphoryl group is transferred to the response regulator KdpE (4,5). Phosphorylated KdpE exhibits an increased affinity to bind upstream of the kdpFABC promoter region and consequently triggers transcription of this operon (6). The enzymatic activities of KdpD have been shown with purified protein reconstituted into proteoliposomes and purified KdpE (7). Moreover, KdpD is found to be the only protein that mediates the dephosphorylation of purified KdpEϳP (7). Recently, we have shown that KdpD is a homodimer and that there is no change of the oligomeric state upon phosphorylation (8). Based on studies with KdpD proteins, in which different Arg residues were individually replaced with Gln, an electrostatic switch mechanism within the protein is proposed, which regulates the ratio of kinase to phosphatase activity (9). The concept of this is supported by the fact that KdpD kinase activity is dependent on negatively charged phospholipids (10).
The stimulus, which KdpD senses, is believed to be a decrease in turgor or an effect thereof. Expression of kdpFABC is induced under K ϩ -limiting growth conditions (less than 2 mM), where the constitutive K ϩ -translocating systems TrkG, TrkH, and Kup of E. coli are unable to maintain the required intracellular pool of K ϩ . In mutants lacking these K ϩ transport systems, kdpFABC is already expressed in media containing less than 50 mM K ϩ . Therefore, KdpD seems to sense the cell's "need" for K ϩ to maintain turgor (11)(12)(13). Control by turgor is supported by the finding that a sudden increase in medium osmolarity, which reduces turgor, transiently turns on kdp-FABC expression. This model has been challenged by the finding that under some conditions expression of kdpFABC is only significantly induced when the osmolarity of the medium is increased by salt and not in case of sugar (9,14,15). Based on kdpD mutants, which give rise to K ϩ -independent kdpFABC expression but regain the ability to respond to changes in medium osmolarity independently of the solute, it is suggested that KdpD senses two stimuli: increase in osmolarity and K ϩ limitation (16). This presumption is supported by the differentiated response of KdpD proteins with individual substitutions of Arg residues toward K ϩ limitation and osmotic upshock (9).
Compared with other known sensor kinases, KdpD is the only protein that has a large, hydrophilic N-terminal domain of about 400 amino acids in length (17,18) (Fig. 1). Sequence comparison of 12 KdpD sequences of various bacteria reveals * This work was supported by the Deutsche Forschungsgemeinschaft (SFB 431) and the Fonds der Chemischen Industrie. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  that this domain is more conserved then other regions of KdpD. This N-terminal domain includes two sequence motifs, which are very similar to those of a typical ATP-binding site (Walker A and B motifs) (18) (Fig. 2). Sequences of short N-terminal versions of KdpD have been found in Synechocystis sp. (19) and in Anabaena sp. L-31 (GenBank™ accession number AF213466). It is suggested that such a fragment may form a functional KdpD complex with a C-terminal fragment that includes the membrane-spanning domains.
It has been shown recently that the phosphatase activity of KdpD is significantly increased in the presence of ATP or nonhydrolyzable ATP analogues, whereas other nucleotides have no effect. Truncated KdpD proteins lacking different parts of the N-terminal domain, including amino acids 12-128 are characterized by phosphatase activities independent of the presence of ATP. This finding suggests that an ATP-binding site has a regulatory role within this domain (18).
Here, we show that the N-terminal domain of E. coli KdpD can be produced separately allowing both the functional and biochemical characterization of this domain.

EXPERIMENTAL PROCEDURES
Materials-[␥-32 P]ATP, NAP-5 gel filtration columns and thrombin were purchased from Amersham Pharmacia Biotech. 8-Azido-[␣-32 P]ATP was from ICN Biomedicals. Goat anti-(rabbit IgG)-alkaline phosphatase and anti-(mouse IgG)-alkaline phosphatase conjugates were purchased from Biomol. Ni-NTA agarose and penta-His antibody were from Qiagen. All other reagents were reagent grade and obtained from commercial sources.
Preparation of Inverted Membrane Vesicles-E. coli strain TKR2000 transformed with plasmids pPV5-3, pBD, or pBD3 carrying different kdpD mutations was grown aerobically at 37°C in KML complex medium (1% tryptone, 0.5% yeast extract, and 1% KCl) supplemented with ampicillin (100 g/ml) or with chloramphenicol (34 g/ml) and ampicillin (100 g/ml). Overexpression of genes under control of the arabinose promoter was achieved by addition of 0.2% arabinose to the medium. Cells were harvested at an absorbance at 600 nm of ϳ1.0. Inverted membrane vesicles were prepared as described previously (7) with the exception that vesicles containing the N-terminal domain were not washed with buffer of low ionic strength. Vesicles were resuspended in 50 mM Tris-HCl, pH 7.5, containing 10% (v/v) glycerol, frozen in liquid nitrogen, and stored until use at Ϫ80°C.
All samples were immediately subjected to SDS-polyacrylamide gel electrophoresis (PAGE) (26). Gels were dried, ATP binding was detected by exposure of the gels to a phosphor screen, and images were analyzed using the PhosphorImager system (Molecular Dynamics). To confirm that equal amounts of protein were used, gels were stained with Coomassie Blue.
Dephosphorylation of KdpE-KdpEϳ 32 P was obtained as described (9). Dephosphorylation was initiated by addition of inverted membrane vesicles (1 mg/ml) containing KdpD derivatives and, when indicated, 20 mM MgCl 2 and 20 M ATP␥S. At different times, aliquots were removed, and the reaction was stopped by addition of SDS sample buffer (26). All samples were immediately subjected to SDS-PAGE. Shortly before stopping SDS-PAGE, an [␥-32 P]ATP standard was loaded on the gels. Gels were dried, and phosphorylation of KdpE was detected and quantified as described above.
Analytical Procedures-Protein was assayed by a modified Lowry method (27), using bovine serum albumin as a standard. Immunodetection of KdpD, KdpFABC, and His-tagged proteins was performed with polyclonal antibodies against KdpD and KdpFABC, respectively, and monoclonal penta-His antibodies according to a previous study (4).

RESULTS
Overproduction, Localization, and Purification of KdpD/1-395(6His)-A short version of KdpD comprising the N-terminal domain (amino acids 1-395) followed by six His residues (KdpD/1-395(6His)) was generated by insertion of six codons for His and three stop codons at the appropriate site in kdpD. This gene was overexpressed in E. coli strain TKR2000, which was transformed with plasmid pPV5-3/1-395(6His) (data not shown). Localization studies indicated that KdpD/1-395(6His) was membrane-associated (Fig. 3, lane 2); none of the protein was found in the cytoplasm (Fig. 3, lane 3). KdpD/1-395(6His) could either be completely solubilized by detergent (data not shown) or partially detached from the membrane under low ionic conditions (Fig. 3, lanes 4 and 5) (see also "Experimental Procedures"). Protein obtained by both methods was used for purification. As illustrated in Fig. 4, KdpD/1-395(6His) binds either in the presence or absence of detergent specifically to Ni-NTA agarose and can be eluted from the resin by increasing the imidazole concentration to 100 mM. Because under low ionic buffer conditions only membrane-associated proteins are released, purification of KdpD/1-395(6His) reaches about 95% homogeneity, whereas in the presence of detergent two major contaminating proteins are detectable (Fig. 4).
Detection of ATP Binding to the N-terminal Domain by Photoaffinity Labeling-The N-terminal domain is highly conserved among all thus far known KdpD sequences of various microorganisms. Moreover, it contains a highly conserved region, which is very similar to a classical ATP-binding site (Walker A and Walker B motifs) (Fig. 2). Direct biochemical evidence for binding of ATP to this site was obtained by photoaffinity labeling with 8-azido-[␣-32 P]ATP. The purified and reconstituted full length sensor kinase, carrying a thrombin cleavage site at amino acid position 395 (KdpD/Th(6His)) to separate the ATP-binding sites of the N-terminal and C-terminal domain, was first labeled with 8-azido-[␣-32 P]ATP and then treated with thrombin. The protein fragments were loaded on an SDS gel, and the corresponding autoradiography was performed (see Fig. 5). Besides the labeled and uncleaved KdpD, two labeled fragments were detectable; however, their size did not correspond to the expected fragment size after thrombin cleavage. Determination of the N-terminal sequences of these fragments confirmed that the large (ϳ29 kDa) fragment repre- sents amino acids 1-240 and thus compromises the ATP-binding site within the N-terminal domain of KdpD. The amount of protein of the large (ϳ35-kDa) fragment was too low to determine the N-terminal sequence, thus the corresponding amino acid composition of this fragment is unknown. The same experiment was performed after preincubation of the protein with an excess of nonradioactive ATP. As shown in Fig. 5 (lane 3), labeled full length KdpD but no labeled fragments were detectable indicating the specificity for binding of ATP. Labeling of KdpD, in which the ATP-binding site was inactivated (KdpD/ G37A, K38A, T39C) was detectable, although the value reached only 10% compared with that of wild-type KdpD (data not shown). As shown in Fig. 5 (lane 4), the isolated N-terminal domain retains the ability to bind 8-azido-ATP when it is membrane-associated. In contrast, no labeling of the solubilized N-terminal domain was observed, neither after solubilization by detergent nor after treatment with low-ionic buffer (lane 5).
Complementation of kdpD/⌬12-395 with kdpD/1-395 in trans Configuration-The importance of the N-terminal do-main for the function of the sensor kinase was shown by complementation experiments as described under "Experimental Procedures." For this purpose, an Escherichia coli strain TKW22692 (17) carrying an in-frame deletion within kdpD encoding KdpD/⌬12-395 was used. kdpFABC expression was monitored by Western blotting using antibodies directed against the transport complex, KdpFABC. Using a polyclonal antiserum against KdpFABC, only subunit KdpB is easily detectable (Fig. 6). In confirmation of earlier data, which were based on a transcriptional fusion of the kdpFABC promotor and lacZ (17), it is found that KdpFABC production is drastically reduced in this strain in comparison to a control strain (Fig. 6,  lanes 1 and 2). However, transformation of E. coli TKW22692 with plasmid pBD/1-395, encoding the N-terminal domain of KdpD, KdpD/1-395, gave rise to a significant increase in the amount of produced KdpFABC complex (Fig. 6, lane 3). The same effect was detectable for the His-tagged N-terminal domain, KdpD/1-395(6His), indicating that the His-tag does not disturb complementation (data not shown). Remarkably, no increase in KdpFABC production was observed when the ATPbinding site in the N-terminal domain was inactivated by amino acid replacements (Fig. 6, lane 4). These results clearly indicate the importance of the ATP-binding site in the hydrophilic N-terminal domain of KdpD and provide evidence for the first time of an interaction of this domain with the rest of the protein.
Influence of KdpD/1-395 on the Phosphatase Activity of KdpD/⌬12-395-The sensor kinase KdpD is characterized by two enzymatic activities: kinase and phosphatase activity. In vitro, phosphatase activity is high in the presence of ATP or nonhydrolyzable ATP analogues, whereas in the absence of ATP dephosphorylation of KdpEϳP is very slow. However, KdpD constructs lacking the ATP-binding site of the N-terminal domain (KdpD/⌬12-228 or KdpD/⌬12-395) by truncation are characterized by a phosphatase activity that is independent of the presence of ATP (18), (Fig. 7A). To test whether the N-terminal domain can complement the truncated KdpD in vitro, plasmids pBD6-92, encoding KdpD/⌬12-395, and pBD3/ 1-395(6His), encoding the N-terminal domain of KdpD, were cotransformed in E. coli strain TKR2000. Cells were cultivated, the two kdpD derivatives were overexpressed, and KdpD phosphatase activity was determined after preparation of inverted membrane vesicles. As shown in Fig. 7B, inverted membrane vesicles containing both the truncated KdpD as well as the N-terminal domain are characterized by a KdpEϳP phosphatase activity, which shows a restored dependence on ATP. These data further support the assumption of a direct interaction of the N-terminal domain (KdpD/1-395) with the rest of the protein and stress the importance of the ATP-binding site located in this region. DISCUSSION The stimulus that KdpD senses to activate signal transduction, which results in induction of kdpFABC expression, is believed to be a decrease in turgor or some effect thereof, reflecting the role of K ϩ as an important cytoplasmic solute (12). However, there is an ongoing debate as to whether KdpD is a turgor sensor or not (13)(14)(15)(16). Despite this, the nature of the primary stimulus for KdpD is still unknown. Although KdpD shares high similarity to transmitter domains of other histidine kinases, it is unusual in regard of the large input domain. This input (sensor) domain comprises an extended N-terminal cytoplasmic region, four transmembrane domains, and part of the C-terminal domain (Fig. 1). Three regions within the input domain have been identified to be important for the activity of KdpD: (i) A cluster of positively charged Arg residues close to transmembrane domain four (TM4) is involved in the shift between kinase and phosphatase activities of KdpD (9). (ii) The four transmembrane domains seem to be important for the correct positioning of N-and C-terminal domains to each other as well as for perception and/or signal transduction (7,17,22). (iii) Amino acids 12-128, including a slightly modified Walker A site, seem to be important for the regulation of the phosphatase activity of KdpD (18). Here, we describe the overexpression and purification of the N-terminal domain of KdpD as well as its functional importance.
Although the Walker A motif found in the N-terminal domain of KdpD is slightly different from a classical motif (an additional amino acid is found between the first and second Gly), photoaffinity labeling by 8-azido-[␣-32 P]ATP shown here provides direct evidence that the N-terminal domain binds ATP. This has been shown for the full length sensor kinase and for the separately produced domain. The other known example for the same modification of the Walker A motif is found in the dethiobiotin synthetase (28), for which ATP binding was shown. ATP binding was detected only for the membrane- associated N-terminal domain, but not for the solubilized peptide. Although the N-terminal domain is mainly composed of hydrophilic amino acid residues, there is a stretch of 17 hydrophobic amino acid residues around the Walker A site (3). From the differences found in the ability to bind azido-ATP, it is concluded that only the membrane-associated domain retains the proper folding to accommodate the nucleotide. It has been shown for the ATP-sensitive channel (K ATP channel), which is blocked by intracellular ATP, that binding affinity changes significantly depending on the membrane phospholipid phosphatidylinositol-4,5-bis-phosphate (29). Because negatively charged phospholipids influence the kinase activity of KdpD (10), it is conceivable that alterations in the attachment of this domain to the lipid bilayer influences the binding to ATP.
The truncated KdpD proteins KdpD/⌬12-395 and KdpD/ ⌬12-228 but not KdpD/⌬128 -391 diminish kdpFABC expression significantly under K ϩ -limiting growth conditions (18). Interestingly, these are the only known examples of KdpD derivatives, which influence the response to K ϩ limitation. Although other amino acid replacements within KdpD alter the response to osmolarity and/or to an increase in the external K ϩ concentration, none of them influence the response to K ϩ limitation (9). In the present study, we have shown that separate expression of the N-terminal domain significantly increases kdpFABC expression under K ϩ -limiting growth conditions. This result indicates the importance of this domain for signal transduction and provides evidence for an interaction between the N-terminal domain and the C-terminal transmitter domain of KdpD.
It has been described before that truncation of amino acids 12-128 causes deregulation of the phosphatase activity in vitro. In the presence of ATP or nonhydrolyzable ATP analogues, dephosphorylation of KdpEϳP is fast, whereas in the absence of ATP only a very slow rate is observed (18). Truncated KdpD derivatives, in which amino acids 12-128 are missing, are characterized by an ATP-independent phosphatase activity (18). Here, we demonstrated that an ATP dependence of the dephosphorylation of KdpEϳP by KdpD/⌬12-395 is restored in the presence of the separately produced N-terminal domain.
The physiological role of the ATP-binding site within the N-terminal domain is still puzzling. Interestingly, under conditions of an osmotic upshock (30) as well as under K ϩ -limiting growth conditions 3 a sudden increase of the intracellular ATP concentration is observed. Because ATP binding affects the phosphatase activity of KdpD, a role of the intracellular ATP concentration as a primary stimulus for KdpD to regulate the ratio between kinase and phosphatase activities is conceivable. However, the direct link between the increased intracellular ATP concentration and binding of ATP to the N-terminal domain is unproved.
In summary, the results presented here indicate that the N-terminal domain is essential for modulating the phosphatase activity of KdpD and thus proper signal transduction. The data provide direct evidence for the existence of an ATP-binding site within this domain. The role of this binding site remains unclear. KdpD activity can be restored in vivo as well as in vitro from two separately produced domains providing indirect evidence for the interaction between the N-terminal domain and the rest of the protein. Finally, the purification protocol for the N-terminal domain presented here provides the prerequisite to characterize this interaction in more detail.