Cell, Vol. 63, 1339-1348,

December

21, 1990, Copyright

0 1990 by Cell Press

The Dynamics of Protein Phosphorylation in Bacterial Chemotaxis Katherine A. Borkovich and Melvin Division of Biology California institute of Technology Pasadena, California 91125

I. Simon

The chemotaxis signal transduction pathway allows bacteria to respond to changes in concentration of specific chemicals (ligands) by modulating their swimming behavior. The pathway includes ligand binding receptors, and the CheA, ChM, Chew, and CheZ proteins. We showed previously that phosphorylation of CheY is activated In reactions containing receptor, Chew, CheA, and CheY. Here we demonstrate that this activation signal results from accelerated autophosphorylation of the CheA kinase. Evidence for a second signal transmitted by a ligand-bound receptor, which corresponds to inhibition of CheA autophosphorylation, is also presented. We postulate that CheA can exist in three forms: a “closed” form in the absence of receptor and CheW; an ‘bpen” form that results from activation of CheA by receptor and CheW, and a %equestered” form in reactions containing ligand-bound receptor and Chew. The system’s dynamics depends on the relative distribution of CheA among these three forms at any time. Introduction Bacterial chemotaxis is an excellent model system for the study of the mechanisms involved in signal transduction (for recent reviews see Stewart and Dahiquist, 1987; Macnab, 1987; Koshland, 1988; Taylor and Lengeler, 1990). Bacteria such as Escherichia coli and Salmonella typhimurium respond to changes in the concentration of chemicals in their environment by altering their swimming behavior. The bacteria accumulate in regions with higher concentrations of compounds called attractants and lower concentrations of compounds termed repellents. Swimming behavior is modified in response to changing chemical concentration by changes in the direction of rotation of helical flagella. When the flagella rotate counterclockwise, the bacterium exhibits smooth swimming behavior and makes net progress in one direction. The switch to clockwise flagellar rotation causes the cell to tumble, allowing a reorientation of the direction of swimming. Changes in swimming behavior are initiated by the binding of specific subclasses of attractants and repellents to four transmembrane receptor proteins: Tar, Tsr, Trg, and Tap. Binding of attractants favors smooth swimming, while binding of repellents causes tumbly behavior. A schematic representation of our current view of the excitation portion of the signal transduction pathway in bacterial chemotaxis is given in Figure 1. The products of four chemotaxis genes, cheA, cheY chew, and cheZ, are re-

quired to transduce binding events at the receptor into an intracellular signal that changes the direction of flageliar rotation. All of these proteins have been purified and their phosphorylation reactions characterized in vitro. CheA is an autophosphoryiating protein kinase that transfers phosphate to CheY (Hess et al., 1987; Wylie et al., 1988; Hess et al. 1988c). Mutations in the cheA or cheYgene that result in defects in this enzymatic step also give rise to cells that have a smooth swimming phenotype, suggesting that CheY phosphoryiation is required to generate the tumble signal (Oosawa et al., 1988; Bourret et al., 1990). CheA is phosphorylated on histidine residue 48 (Hess et al., 1988b), whereas CheY is phosphorylated on aspartate residue 57 (Sanders et al., 1989). CheZ accelerates the dephosphorylation of CheY-phosphate (Hess et al., 1988a, 1988c), an activity that correlates with its role as an antagonist of CheY function in vivo (Parkinson et al., 1983; Kuo and Koshland, 1987; Wolfe et al., 1987). We recently demonstrated coupling of the transmembrane receptor Tar and the soluble Chew protein to the phosphate transfer reactions (Borkovich et al., 1989). In the presence of Tar receptor-containing membranes and the CheA and Chew proteins, the level of CheY-phosphate was increased approximately 300-fold. This stimulation of CheY-phosphate formation was dependent on the nature of the added receptor; a wild-type or tumble mutant receptor, which causes the cells to tumble continuously and is insensitive to ligand, caused activation, i.e., acceleration of CheY-phosphate formation. On the other hand, a smooth mutant receptor, which causes the cells to swim continuouslywithout tumbling, did not activate. In addition, aspartate, the attractant that binds to Tar, inhibited activation of CheY-phosphate synthesis in reactions containing wildtype but not tumble mutant receptor. These results suggested that sensory transduction involves the modulation of the rate of CheY-phosphate formation. The mechanism of receptor-mediated phosphorylation of CheY-like regulatory proteins is the basis not only of bacterial chemotaxis but also of signal transduction in a large number of bacterial systems that result in responses to environmental changes (for recent reviews, see Bourret et al., 1989; Stock et al., 1989). These two-component systems range from the regulation of interactions between plant and mammalian pathogens and their hosts to the regulation of gene expression in critical metabolic pathways. Protein phosphorylation in the chemotaxis system is characterized by the rapidity of its response to environmental changes in chemical concentration. The time required for generation of a response after application of attractant or repellent to a cell is approximately 200 ms (Segall et al., 1982). Furthermore, the cells are able to respond in a coordinated fashion to complex gradients of a number of different attractants and repellents, suggesting that the signal transduction system is capable of integrating multiple inputs (Tsang et al., 1973; Adler and Tso, 1974; Berg and Tedesco, 1975; Segall et al., 1988). Part of the sophistication of this system is due to a sec-

CEll 1340

Aspartate. Maltose -

Results

Dipeptides

Figure 1. A Model Based on Previous Results from a Coupled for the Phosphorylation Reactions of Bacterial Chemotaxis

Assay

Model based on data in Sorkovich et al. (1989). The exact nature of the interactions between CheA. receptors, and Chew in activating CheYphosphate formation is not known. Addition of attractant ligand to receptor inhibits activation of CheY phosphorylation. See text for details.

ond process in signal transduction, i.e., adaptation (Goy et al., 1977). In addition to the excitation response (described above), which leads to switches in flagellar rotation, there is a system that reversibly modifies the receptor by carboxymethylation, extending the cellular response to complex chemical gradients. Adaptation is mediated by the products of the cheR and che6 genes (Springer and Koshland, 1977; Stock and Koshland, 1978). CheR is a methyltransferase that esterifies specific glutamate residues on the receptor (Springer and Koshland, 1977). CheB is a carboxymethylesterase that demethylates the receptor8 and can be activated by phosphorylation by CheA (Stock and Koshland, 1978; Stock et al., 1938; Lupas and Stock, 1989; Stewart et al., 1990); this provides direct feedback to the receptor through CheA. In the absence of these gene products, the cell display8 rudimentary chemotaxis-like behavior that is driven by excitation, i.e., in the presence of high concentrations of attractant it swim8 smoothly and when attractant is removed it tumbles (Weis and Koshland, 1988; Wolfe and Berg, 1989). In this study we focus our attention on the excitation pathway. In this paper, we address several questions regarding the mechanism of activation of CheY-phosphate formation, and the way in which signals from several receptors in various signaling mode8 are integrated to give a concerted response to stimuli. Which step(s) of the reaction is activated by the addition of receptor and Chew? Which step is inhibited by attractant ligand? What is the nature of the smooth signal, i.e., can a receptor bound to a ligand inhibit activation of CheY-phosphate formation by a stimulatory receptor? Finally, we develop evidence for a general model to explain the excitation process that regulate8 the direction of flagellar rotation.

The Tar Chemoreceptor and the Chew Protein Activate Autophosphorylation of the CheA Kinase We have shown that in the presence of activated receptor (receptor in the absence of ligand), and the Chew and CheA proteins, there is an extremely rapid increase in the level of CheY-phosphate (Borkovich et al., 1989). To determine whether this was the result of the activation of the kinase by the receptor, we further fractionated the system and looked for direct activation of CheA autophosphorylation by the receptor. Addition of wild-type receptor and the Chew protein to a reaction containing CheA resulted in an approximately lo-fold stimulation in the level of CheAphosphate formed within 5 8 (Figure 2A). This stimulation absolutely required the presence of both Tar and Chew. The level of phosphorylated CheA produced in reactions lacking either Tar or Chew showed a slow but steady increase over 15 s, reaching a maximum after approximately 10 min (data not shown); this is characteristic of CheA autophosphorylation (Hess et al., 1988c). In contrast, in reactions containing both Tar and Chew, the level of CheA-phosphate reached a maximum in less than 5 s and remained constant after this time (Figure 2A). Reactions containing tumble mutant receptor but not the smooth mutant receptor also showed stimulation of CheAphosphate accumulation (Figure 28) indicating that stimulation of CheA-phosphate formation requires the presence of a receptor capable of generating tumbles in vivo. We attempted to measure the rate of transfer of phosphate from CheA to CheY in both the coupled and uncoupled systems. However, in both cases, the rate was too rapid to follow using 5 s time points (Hess et al., 1988c; data not shown). Rapid-quench kinetic experiments are currently being designed to measure these rates. CheA-Phosphate Formed in the Presence of Receptor and Chew Is Dynamic The steady-state level of CheA-phosphate attained in the presence of Tar and Chew can be explained in a number of ways. It could result from the rapid formation of a stable phosphoprotein that reaches a saturated level by 5-10 s, from constant synthesis and turnover of a relatively labile phosphoprotein, or by continual ATP-ADP exchange catalyzed by CheA. To investigate the stability of CheAphosphate formed in reactions containing the activators, a series of pulse-chase analyses was performed. CheAphosphate was allowed to form for 10 s, at which time excess ATP, ADP EMA, or both EDNA and ATP were added, and the effect of the chase on the level of CheA-phosphate was analyzed (Figure 3A). Addition of excess ATP or ADP resulted in almost complete loss of label from CheAphosphate by the first time point of the chase. Chasing with EDTA alone or in concert with ATP resulted in maintenance of prechase levels of CheA-phosphate, indicating that a divalent cation, presumably Mg2+, is required for the lability of activated CheA-phosphate.

Phosphorylation 1341

in Bacterial

Chemotaxis

A + +

0

Chase

CheA CheA+CheW CheA+Tar CheAKheW+Tar

+

ImMADl’

+

5mMADP

+

5mMAll’

+

15mM

+

15mMEDTA

EDTA,

5mM

All

10

Time kec) 20

Time

30

(set)

-

Water

-

1mMATP

+

1mMADP

0 Wild

Type

Tumble

Smooth -w

Source of Tar Figure 2. Activation of CheA-Phosphate ence of Both a Stimulatory Receptor

Formation and Chew

Requires

the Pres-

(A) Dependence on individual components. Reactions contained 10 pmol Of CheA. and 0.1 mM ATP (22,800 cpmlpmol), with or without 80 pmol of Chew. Either wild-type Tar-containing membranes (24 ng of membrane protein) or negative control membranes (32 ug of membrane protein) were added, as indicated. Membranes were derived from strain HCB437 (B) Wild-type and tumble receptor-containing membranes stimulate phosphorylation of CheA. All reactions contained 10 pmol of CheA, 80 pmol of Chew, and 0.1 mM ATP (3380 cpmlpmol), and receptorcontaining membranes of the indicated type from strain HCB437 (22 ug of membrane protein). Reaction time was 5 s.

The behavior of activated CheA-phosphate formed in the presence of receptor and Chew (Figure 3A) contrasts with that of nonactivated CheA-phosphate (Figure 38). CheA-phosphate formed in the presence of Chew but in the absence of receptor is relatively stable during a chase with excess unlabeled ATP (Figure 38). However, activated and nonactivated CheA-phosphate behave similarly during a chase using unlabeled ADP; the rate of disappearance of both phosphoproteins is rapid (Figures 3A and 38). Preformed purified CheA-phosphate undergoes virtually no hydrolysis when exposed to Chew and receptor-containing membranes (data not shown); thus, the membrane preparations do not contain appreciable CheAphosphate phosphatase activity, and the lability observed for activated CheA-phosphate during a chase with ATP is dependent on its formation in the presence of the receptor and Chew. It is apparent that CheA-phosphate formed in the pres-

0

10

20

Time Figure 3. Comparison CheA-Phosphate

of the Properties

30

4”

50

Csec) of Activated

vs. Nonactivated

(A) Activated CheA-phosphate is labile. Reactions contained 5 pmol of CheA, 40 pmol of Chew, wild-type Tar-containing membranes from strain HCB721 (34 pg of membrane protein), and 0.1 mM ATP for labeling (12,400cpmlpmol). At lOs, chase mixtureswereadded as indicated. The results of several other experiments indicate that 1 mM ATP is as effective as 5 mM ATP during a chase in reactions containing CheA in the presence of receptor and Chew (data not shown; Figure 48). (B) CheA-phosphate made in the absence of receptor is stable during a chase with ATP. Reactions contained 10 pmol of CheA, 80 pmol of Chew, and membranes lacking receptor from strain HCB721 (35 ug of membrane protein). Labeling was for 10 s using 0.1 mM ATP (23,600 cpmlpmol); after this time chase mixtures containing 1 mM ATP 1 mM ADP or water were added, as indicated.

ence of the receptor and Chew is a dynamic species, clearly distinguishable in its properties from nonactivated CheA-phosphate. The phosphate on the activated or “open” form of CheA is rapidly chased by both ATP and ADP i.e., a steady state is reached in less than 5 s. The data are consistent with a rapid exchange mechanism whereby activated CheA equilibrates a pool of ATP and ADP via formation of a phosphorylated intermediate in an Mg2+-dependent manner. Alternatively, the results could be explained if open CheA-phosphate is unstable and hydrolyzes to produce CheA and orthophosphate, and if

Cell 1342

1

AL

I Wild

L

Type

1 Tumble

Source

of Tar

10

Time

20

kec)

C

ChaSe .x!oo-

9

--

ADP

--e

ADP+aspartate

10

m

Time (set) Figure 4. Aspartate Inhibits Activation of CheA-Phosphate and Stabilizes It during a Chase Using ATP or ADP

Formation

(A) Aspartate prevents activation of CheA-phosphate formation by wildtype but not tumble mutant receptor. Reactions contained 10 pmol of CheA, 80 pmol of Chew, and wild-type (13 ug of membrane protein) or tumble mutant (11 ftg of membrane protein) receptor-containing membranes from strain HCB437, 0.1 mM ATP (~23,000 cpmlpmol), and 1 mM aspartate or water (control). Reaction time was 5 s. (B) Aspartate stabilizes activated CheA-phosphate in the presence of ATP Reactions contained 10 pmol of CheA, 90 pmol of Chew, and either wild-type (12 ug of membrane protein) or tumble mutant (13 ttg of membrane protein) receptor-containing membranes from strain HCB437 ATP at 0.1 mM was used for labeling (9940 cpm/pmol). At 5 s, the chase mixtures indicated were added. The residual level of CheAphosphate remaining after a chase in the absence of aspartate in these reactions is due to a subsaturating level of receptor: the “sheIF can be lowered by addition of more receptor-containing membranes, and is presumably due to the uncoupled reaction (see also Figures 3A and 4C). (C) Activated CheA-phosphate cannot be chased using ADP in the presence of aspartate. Reactions contained 5 pmol of CheA, 40 pmol

CheA can then rapidly rephosphorylate using ATP in the presence of Chew. The nonactivated or “closed” form of CheA slowly catalyzes autophosphorylation using ATP and its phosphate is rapidly chased using ADP but not ATP Addition of the Ligand Aspartate Inhibits Further Formation of Activated CheA-Phosphate but Stabilizes Existing Activated CheA-Phosphate We have previously shown that inclusion of aspartate in coupled assays containing Tar, Chew, CheA, and CheY results in loss of amplification of CheY-phosphate formation (Borkovich et al., 1989). Sensitivity to aspartate was exhibited in reactions containing the wild-type but not the tumble mutant receptor, and showed the same concentration dependence as that for modulation of swimming behavior in vivo. To determine if aspartate acts to inhibit formation of CheY-phosphate at the level of autophosphorylation of the CheA kinase, the effect of aspartate on CheA-phosphate formation was measured. When added to reactions at time zero, aspartate inhibited activation of CheA-phosphate (Figure 4A). When the tumble mutant receptor, containing adominant mutation that fixes it in the tumble-generating mode, was used as the source of receptor, phosphorylation was insensitive to aspartate, while reactions with the wild-type receptor showed inhibition. Since activated, or open, CheA can rapidly exchange phosphate, we could test whether aspartate addition affected this property. To this end, a pulse-chase experiment was performed. CheA-phosphate made in reactions containing CheA, Chew, and either wild-type or tumble mutant Tar was chased using a lo-fold excess of unlabeled ATP in the presence of either aspartate or a control compound, succinate (a poor ligand for this receptor). In reactions containing the tumble mutant receptor, CheAphosphate exhibited the same relative lability in the presence of both succinate and aspartate (Figure 48). CheAphosphate was also labile during a chase with succinate in reactions containing the wild-type receptor. However, when reactions with the wild-type Tar were chased with aspartate, CheA-phosphate was stable, remaining at a constant level similar to that seen using chases containing EDTA as described above (Figure 3A). When a similar chase experiment was performed in reactions with the wild-type receptor using ADP instead of ATP, the same result was obtained (Figure 4C). Thus, the addition of ligand to active (wild-type) receptor immediately converts CheA from the open state to a sequestered state. Sequestered CheA-phosphate is not chased by either ADP or ATP a trait that distinguishes it from both the open and closed forms of CheA. Since the behavior of CheA depends on ligand binding to the receptor, the CheA protein must either be in direct physical contact with the receptor or be connected by a series of mobile equilibria to the state of the receptor. of Chew, and wild-type Tar-containing membranes from strain HCB721 (35 ug of membrane protein). Labeling was for 5 s using 0.1 mM ATP (17,200 cpmlpmol), after which chase mixtures containing either 1 mM ADP in water or 1 mM ADP in 1 mM aspartate were added.

Phosphorylation 1343

in Bacterial

Chemotaxis

-

x .z ;: 5 > E 2

CPM CheAPi (before additIonI ChcY CheY+aspartate

4 +

Tumble Tar / WT Tar+Aspartate

+

tl” -

WT Tsr , WT Tar+Aspartate WT Tar / Smooth Tar

60 -

40-

‘ii $

20-

-I

a I

0 n

10

Time

20

0

1

~1 Inhibitory

2

3

Receptor-Containing

4

5

Membranes

(set)

Figure 5. CheA-Phosphate Formed in the Presence of Aspartate Can Transfer Phosphate to CheY, but Is Incapable of Further Activated Phosphorylation Reactions contained 5 pmol of CheA, 40 pmol of Chew, and 35 pg of membrane protein from strain HCB721 containing the wild-type Tar receptor. ATP at a concentration of 0.1 mM was used for labeling (*17,200 cpmlpmol). At 5 s, 100 pmol of CheY in the presence or absence of 1 mM aspartate was added to the reactions as indicated. Since no unlabeled ATP was included in the addition, production of phosphorylated protein before and after the addition was at the same specific radioactivity, i.e., there should have been no effect on phosphorylation rates by dilution. The level of CheA-phosphate present before the additions is shown by the closed triangle. CheAPi = CheAphosphate.

To explore further the effects of adding ligand to receptor in the coupled reaction, we added excess CheY in the presence or absence of aspartate to autophosphorylation reactions already in progress, and measured phosphate transfer from CheA to CheY. After the addition of CheY, phosphate is rapidly transferred from preformed CheAphosphate to CheY in the presence or absence of aspartate, but a low level of CheA-phosphate is maintained in reactions lacking aspartate (data not shown). There is no new synthesis of CheA-phosphate in reactions containing aspartate (data not shown). Accordingly, the level of CheYphosphate rises rapidly and increases with time in reactions lacking aspartate, while reactions containing aspartate show a much lower level of CheY-phosphate that is maximal at the first time after the addition (Figure 5). The level of CheA-phosphate (in cpm) at 5 s is approximately equal to the maximum level of CheY-phosphate (in cpm) observed in reactions to which aspartate has been added (Figure 5). In contrast, the maximum level of CheY-phosphate in the absence of aspartate is lo-fold higher than that of CheA-phosphate before the addition, indicating multiple phosphorylation and transfer events occurred after addition of CheY (Figure 5). The lack of new rounds of synthesis of CheY-phosphate ObSWWd in reactions containing aspartate cannot be explained by aspartate causing CheY-phosphate to turn over more rapidly, because the tI12 calculated for CheY-phosphate under these conditions (20-30 s; data not shown) is actually greater than the value (6 s) determined for CheY-phosphate in the absence of receptor and Chew (Hess et al., 1988c). These data are consistent with a model in which CheA-phosphate formed in the presence of aspartate can transfer its phosphate to

Figure 6. Attractant-Bound Receptor hibits Activation of CheY-Phosphate Receptor

or a Smooth Formation

Mutant Receptor Inby a Stimulatory

All reactions contained 40 pmol ofCheW, 300 pmol of CheY, 0.4 mM ATR and other components as indicated below. Membranes devoid of receptor were added to appropriate reactions to give the same final membrane protein concentration, and all membranes were isolated from strain HCB721. On the figure, the stimulatory receptor is indicated to the left of the slashes, while the inhibitory receptor is shown on the right. On they axis, % maximum velocity is defined as the % maximum cpm CheY-phosphate formed per s, with the maximum rate being that found in the absence of added inhibitory receptor (at 0 11). WT = wild type. Circles: reactions contained 5 pmol of CheA, 1 mM aspartate, 9 pg of membrane protein from tumble mutant Tar-containing membranes (13.5 lg of membrane protein equivalent to wild type, based on receptor activity), and the indicated volume of wild-type Tar-containing membranes (4 wg/pl membrane protein). Squares: reactions contained 5 pmol of CheA, 1 mM aspartate, 0.4 vg of Tsr-containing membrane protein (1.2 pg of protein equivalent to wild type; see Results), and the indicated volume of wild-type Tar-containing membranes (4 Kg/PI membrane protein). Triangles: reactions contained 2 pmol of CheA, 35 wg of wild-type Tar-containing membrane protein, and the indicated volume of smooth mutant Tar-containing membranes (6 pg/wl membrane protein). Reaction time was 5 s.

CheY, but the dephosphorylated CheA is incapable rounds of receptor-activated phosphorylation.

of new

Integration of Signals from Multiple Receptors in Different Signaling States Occurs at the Level of Activation of CheA-Phosphate Formation Receptors can exist in at least two states, liganded and unliganded. In the absence of ligand, the receptor can activate CheA phosphorylation and subsequent CheYphosphate formation. However, the question remains whether the receptor with ligand bound can inhibit tumble formation, i.e., can it influence the production of phosphorylated CheA and CheY by a tumble-generating recep tor? A related question is how the signals from several receptors, in various signaling states, are integrated by the cell to give an appropriate response to the environment. To address these questions, a series of in vitro competition assays was performed, testing the effect of adding receptor bound to attractant in increasing amounts to reactions containing wild-type or mutant tumble receptors. CheY-phosphate formation was followed using a relatively low level of CheA, near-saturating levels of Chew, and saturating CheY, in addition to the receptor-containing membranes in the assays. We used both the wild-type and

Cell 1344

-f-

2col-

w-r Tar +Tsr+serine WTTar+Tsr

0 0

1

2

3

~1 Tsr-Containing

4

5

Membranes

+ +

WI’ Tar +Trr+serine WTTar +Tsr

l

II 0

1

2

~1 Tsr-Containing

3

4

5

Membranes

Figure 7. Receptor Bound to Attractant Inhibits Activation of CheYPhosphate Formation by a Stimulatory Receptor at the Level of CheA Autophosphorylation (A) Wild-type Tsr bound to serine inhibits activation of CheY-phosphate formation by the wild-type Tar receptor. Reactions contained 2 pmol of CheA, 40 pmol of Chew, 300 pmol of CheY, 0.4 mM ATP (4530 cpmlpmol), 35 pg of wild-type Tar-containing membrane protein, and the indicated volume of wild-type Tsr-containing membranes (4 pglpl membrane protein, which is equivalent to 12 pg/ul wild-type Tar-containing membrane protein in terms of receptor activity; see Results). In addition, the indicated reactions contained 10 mM serine. All reactions were brought to the same final membrane protein concentration using membranes devoid of receptor. Strain HCS721 was the source of all membranes. WT = wild type. (E) Wild-type Tsr bound to serine inhibits activation of CheA-phosphate formation by the wild-type Tar receptor. Conditions were the same as in (A), except that the specific activity of the ATP was 3150 cpmlpmol and CheY was omitted from reactions. A Phosphorimager was used for initial quantitation of CheA-phosphate produced, after which the Phosphorimager units were converted to cpm using known radioactive standards (2.205 x low4 cpm per Phosphorimager unit). Reaction time was 5 s.

tumble mutant of Tar, in the absence and presence of its inhibitory ligand aspartate. For some experiments the wild-type Tsr receptor, which binds the attractant ligand serine, was used. Reactions containing the Tsr receptor showed activation of CheA and CheY-phosphate formation in the absence, but not in the presence, of serine (data not shown). In the first series of experiments, the wild-type Tar receptor in the presence of aspartate or the smooth mutant Tar receptor was used as the inhibitory receptor in reactions containing either the tumble mutant Tar, the wild-type Tsr

receptor, or wild-type Tar. The tumble mutant Tar receptor is itself insensitive to aspartate (Borkovich et al., 1989) however, addition of wild-type Tar receptor in the presence of aspartate inhibited by 80% the initial rate of formation of CheY-phosphate in reactions containing this receptor by 80% (Figure 8). In an analogous fashion, addition of increasing amounts of wild-type Tar plus aspartate to reactions containing wild-type Tsr caused a 83% reduction in the initial rate of CheY-phosphate formation activated by the Tsr receptor (Figure 8). The smooth mutant Tar proved to be the most potent inhibitory receptor under our experimental conditions; addition of this receptor to reactions containing wild-type Tar inhibited CheY-phosphate formation by 88% (Figure 8). Inhibition of wild-type Tar activated phosphoprotein production by the wild-type Tsr receptor bound to attractant was chosen for more detailed study. As in the cases described above, increasing amounts of Tsr plus serine inhibited activation of CheY-phosphate formation catalyzed by the Tar receptor (Figure 7A). When serine was omitted, the initial rate of CheY-phosphate production was increased in reactions with a higher level of total receptor present (Figure 7A). When the effect on CheA-phosphate accumulation at 5 s of reaction was analyzed in analogous reactions lacking CheY, the same relative patterns of inhibition or amplification were seen depending on the experimental conditions (Figure 78). These data indicate that a”smooth” receptor can reduce the activation of phosphoprotein production mediated by a”tumble” receptor, by acting at the level of CheA autophosphorylation. The amount of receptor protein necessary for inhibition varied with the receptor type. Wild-type Tar was able to inhibit tumble mutant Tar at approximately stoichiometric levels. Likewise, both Tsr and the smooth mutant Tar inhibited wild-type Tar at a 1:l ratio of receptor protein. However, a 5 to Ffold molar excess of wild-type Tar protein was necessary to inhibit Tsr to the same degree. Since Tsr is approximately 3 times more active (on a per protein basis) in stimulating the rate of CheY-phosphate formation than either the tumble or wild-type Tar (data not shown), this may explain the need for an excess of Tar to inhibit Tsr. These data are consistent with a mechanism of inhibition that involves sequestration of CheA and/or Chew by the liganded receptors. In addition, the observation that liganded Tsr and Tar are equally effective inhibitors, yet Tsr is a better activator of phosphorylation, suggests that the mechanisms involved in the inhibition and activation processes may be different. There does not appear to be a strict correlation between the ability of a receptor to activate CheA autophosphorylation and the amount of it required for inhibition when bound to attractant ligand. Discussion We have previously shown that in an in vitro reconstituted system, the rate of CheY phosphorylation is coupled to the state of the chemotaxis receptor. In the current work we further explored the mechanism of this activation. We can explain our data by proposing that the CheA kinase behaves as if it exists in three different forms (see Figure 8A).

Phosphorylation

in Bacterial

Chemotaxis

1345

Figure

8. The Three

Postulated

Forms

of CheA

The properties of closed (formed in absence of receptor and CheW), open (activated in presence of receptor and CheW), and sequestered (associated with liganded receptor and Chew) CheA molecules are shown. See text for detailed explanation.

In the presence of an activated receptor (unliganded Tar receptor or tumble mutant receptor), Chew, and ATP the kinase is primarily in the open form. In the absence of any of these components, CheA is found primarily in the closed form and can be readily converted to the open form. In the presence of the smooth mutant receptor or receptor with ligand and Chew, CheA is in the sequestered form. The open form of CheA is distinguished from the closed and sequestered forms by the following operational criteria. First, in the presence of ATP, the open form is rapidly autophosphorylated (many times faster than the closed form of CheA). Second, the autophosphorylated open form exchanges covalently bound phosphate extremely rapidly with a pool of free ATP or AD!? Third, the open form is tightly coupled to the state of the receptor in the presence of (stoichiometric amounts of) Chew; thus, the addition of an appropriate ligand to the receptor results in the immediate conversion of the open form of CheA to the sequestered form. Fourth, in the open form, CheA is able to very rapidly phosphorylate CheY. This appears to be the result of an acceleration in the rate at which CheA, once it has transferred phosphate, is recycled to an active autophosphorylated form (see Figure 5). The sequestered form of CheA-phosphate is differentiated from both the open and closed forms in that it requires the presence of liganded receptor and does not show rapid exchange of its phosphoryl group upon the addition of ATP or ADP (see Figures 46 and 4C). We suggest that the unliganded receptor exists in a conformation that can interact with CheA and Chew. We do not know if a relatively stable receptor-CheA-Chew complex exists or if there is a series of rapid equilibrium binding events that connects the activated receptor to the CheA and Chew proteins. However, Liu and Parkinson (1989) have obtained genetic evidence consistent with the existence of receptor-CheA-Chew complexes in vivo, and CheA-Chew complexes can be isolated from crude cell extracts in vitro (D. McNally and I? Matsumura, personal communcation). Therefore, the function of the un-

liganded receptor may be to stabilize the interaction between CheA and Chew, increasing the lifetime of a complex containing CheA in the open form. In any event, the evidence clearly indicates that the open form of the CheA kinase is tightly coupled to the state of the receptors, since the addition of an appropriate ligand such as aspartate immediately changes the behavior of CheA from that of the open form to that of the sequestered form. We do not know the nature of the physical correlates of the open and sequestered forms of CheA. These may correspond to changes in multimerization or to complexes formed between CheA, receptor, and Chew. Evidence suggests that two kinds of excitation signals exist to activate chemotaxis: a tumble signal generated by the receptor in the absence of ligand, and a smooth signal generated by the receptor in the presence of ligand. Parkinson and coworkers, for example, have isolated Tsr dominant signaling mutants that are locked in the smooth chemotaxis form (Ames and Parkinson, 1988). Corresponding Tar mutations that confer a smooth dominant chemotaxis phenotype have also been isolated (Mutoh et al., 1988). Our data suggest that a “smooth” signal is initiated when a receptor in the presence of attractant ligand converts CheA to the sequestered form, effectively depleting the pool of kinase available for activation. We have observed this inhibitory activity, or smooth signal, using the smooth mutant Tar receptor, the Tar receptor bound to aspartate, and the Tsr receptor associated with serine. Thus, in order for a receptor to generate either a tumble or a smooth signal, it must interact with CheA. Our in vitro results using the unmodified receptors in the coupled assay are in good qualitative agreement with in vivo results reported by two different laboratories (Weis and Koshland, 1988; Wolfe and Berg, 1989). Bacterial cells devoid of CheB and CheR tumble in the absence of added ligand (Weis and Koshland, 1988; Wolfe and Berg, 1989). At a concentration of >25 t.tM aspartate, 100% of the population is smooth swimming and only returns to the tumbly mode after removal of the aspartate (Weis and Koshland, 1988). In our coupled assay, the unliganded receptor stimulates an increase in the level of CheA autophosphorylation; apparently the basal state of the unmodified receptor is the form that can produce open CheA-phosphate and tumbles. Addition of aspartate inhibits the activation of autophosphorylation and presumably converts CheA to the sequestered form, leading to smooth swimming behavior. Amplification in chemotaxis could result from the ability of a single receptor to stabilize CheA in the open or catalytic form, which can then rapidly phosphorylate many CheY molecules. The role of the other proteins in the chemotaxis system also becomes clearer. CheZ plays a critical role in regulating CheY-phosphate levels. The CheZ protein returns the system to its basal level by facilitating the rapid dephosphorylation of CheY-phosphate. This allows any subsequent activation of CheA, and hence, increased level of CheY-phosphate, to be interpreted as a new signal. Amplification of the attractant signal could result from the ability of the receptor-ligand complex to convert multiple molecules of CheA to the sequestered form.

Cell 1346

It is further possible that CheZ function is regulated. For example, if CheZ were activated by sequestered CheA or by the components of the motor, it could use this feedback to limit tumbles more effectively. A variety of other regulatory interactions and feedback loops almost certainly occur in this system. By modifying the receptor, and therefore changing its signaling bias, CheR and CheB can have a dramatic effect on the kinetics and characteristics of the signaling pathway. Nonetheless, the primary pathway in the excitation response involves the complex interactions of CheA, Chew, and the various forms of the receptor. This model provides an explanation of how integrated signaling processes can occur in bacteria in only 200 ms. The rapid signaling that occurs during chemotaxis is in coriirast to the slower transcriptional activation responses mediated by other two-component regulatory systems of bacteria. This difference in speed may result from slower autophosphorylation of the kinase, slower transfer of phosphate to the regulator protein, production of a more stable phosphorylated regulator, or incorporation of novel phosphate-transfer activities into the pathway. In the case of the nitrogen assimilation regulatory system, one difference is an enhanced stability of the phosphorylated response regulator; NtrC-phosphate has a half-life of approximately 210-300 s (Keener and Kustu, 1988; Weiss and Magasanik, 1988), as compared with 6 s for CheYphosphate (Hess et al., 1988a). Analogies between bacterial chemotaxis and the signal transduction systems found in eukaryotic organisms are striking. Many hormone receptors function by coupling the binding of ligand to a receptor with the activation of exchange of GDP noncovalently bound to a G protein for GTP (for reviews see Gilman, 1987; Casey and Gilman, 1988; Lochrie and Simon, 1988; Ross, 1989). This nucleotide exchange activates the G protein and allows it in turn to interact with second messenger-generating systems. The general principle is that ligand stabilizes a specific state of the receptor, which activates nucleotide exchange on the heterotrimeric G protein, “opening” the a subunit of the G protein and stabilizing it in an activated form. In chemotaxis, the unliganded form of the receptor activates a complex of CheA and Chew by allowing rapid exchange of a covalently bound phosphoryl group with ATP and ADP In both systems, rapid feedback occurs by covalent modification of the receptor (Springer and Koshland, 1977; Stock and Koshland, 1978; reviewed in Sibley et al., 1987). Comparison of these systems suggests that further study of signal transduction in bacterial chemotaxis will uncover the strategies used by both prokaryotic and eukaryotic organisms in the design of information-processing circuits. Experlmental

Procedures

Purlflcatlon of Chemotaxis Proteins and Preparation of Membranes The CheA, CheY, and Chew proteins were purified as previously described (Matsumura et al., 1984; Stock et al., 1987; Hess et al., 1988~). Plasmid pJC3, which overexpresses the wild-type Tsr chemoreceptor under control of a tat promoter, was obtained from J. Chen and J. S. Parkinson, University of Utah, Salt Lake City, Utah. Plasmids used for overexpression of the wild-type Tar chemoreceptor (pNT201), tumble

mutant Tar (pNTZOl-NlOl). or smooth mutant Tar (pNT201-N15) have been previously described (Borkovich et al., 1989). These receptorencoding plasmids were maintained in the Che- deletion strains HCB437 (A(W) 7021 A(Wg) 100 A(ChaA-CheZJ 2209) (Wolfe et al., 1987) or HCB721 (A(w) 7021 trg::TnlO A(chwl-cheY)::Xho (Tn5)) (Wolfe et al., 1988), as indicated. Membranes containing the various chemotaxis receptors were isolated as previously described (Borkovich et al., 1989), except that 2 M NaCl was substituted for 2 M KCI for the high-salt wash step in some cases. This modification of the washing procedure caused no change in the stimulatory properties of the receptor(s) (data not shown). Protein concentrations were determined using the Bio-Rad protein reagent concentrate with bovine serum albumin as a standard. The relative levels of receptors in various membrane preparations were estimated by both visual inspection and densitometry of the Coomassie-stained band corresponding to the receptor after sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) analysis (Laemmli, 1970). The experiments described were performed in a system that has been simplified to exclude the effects of receptor adaptation. Thus, the receptor used in all of the studies presented in this paper was prepared from a strain deleted for the che6 and cheR genes. Control experiments with receptor from strains that were ccypetent for modification (containing the CheB and CheR gene products) showed some quantitative differences; however, there were no major qualitative differences in the results obtained with modified vs. unmodified receptor (data not shown). Phosphorylation Assays Phosphorylation reactions were performed at room temperature as previously described (Borkovich et al., 1989) in a20 ~1 solution containing 50 mM Tris-HCI (pH 7.5), 50 or 100 mM KCI, 5 mM MgClz, various concentrations of [P~~P]ATP, purified proteins, and washed membranes as indicated in the figure legends. Depending on the experiment, the concentration of ATP u .>d for labeling ranged from 0.1-0.4 mM. The reaction samples were subjected to SDS-PAGE as described (Borkovich et al., 1989). Incorporation of Ij2Plphosphate into CheA or CheY was determined by excising the radioactive band out of the dried gel and quantitating in scintillation fluid or by analysis of the intact gel using a Phosphorimager (Molecular Dynamics, Sunnyvale, CA) and comparing with known radioactive standards. Because the autophosphorylation reaction reached a steady state so quickly, it was not possible to determine initial rates of CheA-phosphate formation in the presence of receptor and Chew (see Results). Therefore, the amount of CheA-phosphate formed is always reported as the amount present after 5-10 s of reaction. On the other hand, at lower concentrations of ATP (