Determinants of chemotactic signal amplification in Escherichia coli

doi:10.1006/jmbi.2000.4389 available online at http://www.idealibrary.com on J. Mol. Biol. (2001) 307, 119±135

Determinants of Chemotactic Signal Amplification inEscherichia coli

Catherine Kim, Marilyn Jackson, Renate Lux and Shahid Khan*

Laboratory of CellularBioenergetics, Department ofPhysiology and BiophysicsAlbert Einstein College ofMedicine, BronxNY 10461, USA

E-mail address of the [email protected]

Abbreviations used: CW, clockwicounterclockwise; MCP, methyl-accprotein.

0022-2836/01/010119±17 $35.00/0

A well-characterized protein phosphorelay mediates Escherichia coli che-motaxis towards the amino acid attractant aspartate. The protein CheYshuttles between ¯agellar motors and methyl-accepting chemoreceptor(MCP) complexes containing the linker CheW and the kinase CheA.CheA-CheY phosphotransfer generates phospho-CheY, CheY-P. Aspartatetriggers smooth swim responses by inactivation of the CheA bound tothe target MCP, Tar; but this mechanism alone cannot explain theobserved response sensitivity. Here, we used behavioral analysis ofmutants deleted for CheZ, a catalyst of CheY-P dephosphorylation, orthe methyltransferase CheR and/or the methylesterase CheB to examinethe roles of accelerated CheY-P dephosphorylation and MCP methylationin enhancement of the chemotactic response. The extreme motile bias ofthe mutants was adjusted towards wild-type values, while preservingmuch of the aspartate response sensitivity by expressing fragments of theMCP, Tsr, that either activate or inhibit CheA. We then measuredresponses to small jumps of aspartate, generated by ¯ash photolysis ofphoto-labile precursors. The stimulus-response relation for �cheZmutants overlapped that for the host strains. �cheZ excitation responsetimes increased with stimulus size consistent with formation of anoccluded CheA state. Thus, neither CheZ-dependent or independentincreases in CheY-P dephosphorylation contribute to the excitationresponse. In �cheB�cheR or �cheR mutants, the dose for a half-maximalresponse, [Asp]50, was ca 10 mM; but was elevated to 100 mM in �cheBmutants. In addition, the stimulus-response relation for these mutantswas linear, consistent with stoichiometric inactivation, in contrast to thenon-linear relation for wild-type E. coli. These data suggest that responsesensitivity is controlled by differential binding of CheR and/or CheB todistinct MCP signaling conformations.

# 2001 Academic Press

Keywords: sensory transduction; histidyl-aspartyl phosphorelay; videomotion analysis; domain liberation; caged compounds
*Corresponding author
Introduction

Escherichia coli chemotaxis is likely to be the ®rstsensory response to be completely understood atthe molecular level. It is mediated by a simple,intracellular, biochemical network. The motileresponse can be readily quanti®ed and atomicstructures for most of the protein components areavailable.1 The E. coli chemotactic phospho-relayconsists of a small, cytoplasmic protein, CheY,

ing author:

se; CCW,epting chemotaxis

which shuttles between methyl-accepting chemo-taxis protein (MCP) receptor signal complexes and¯agellar motors. CheY is phosphorylated by thehistidine kinase CheA that forms part of the signalcomplexes. Dephosphorylation of phosphorylatedCheY, CheY-P, is promoted by CheZ. The swim-ming motility of E. coli alternates between runsand tumbles. Runs occur when a bundle of coun-terclockwise (CCW) rotating ¯agella pushes thebacteria forward. Tumbles result when a suf®cientnumber of ¯agella within a bundle reverse rotationsense, causing its breakup.2 Addition of chemo-attractants increases CCW rotation intervals; theirwithdrawal increases clockwise (CW) intervals.3

Migration up spatial gradients of attractant occurs

# 2001 Academic Press

http://www.idealibrary.com

mailto:[email protected]

120 Chemotactic Signaling in E. coli.

primarily by increase in the length of swimmingruns.4

Explanation of the high sensitivity of the chemo-tactic response in terms of the signal pathway bio-chemistry has emerged as the outstanding problemof bacterial chemotaxis. CheA inactivation is theonly mechanism for signal generation by attractantligands revealed thus far by biochemical studies.However, stoichiometric inactivation of the CheAbound to the attractant-MCP complex predicts alower response sensitivity than observed.5,6

Here, signal processing in mutants deleted forchemotaxis proteins that lie off the signal pathwayhas been analyzed to identify circuitry responsiblefor the unexplained sensitivity. The proteins areCheZ, the MCP methylesterase, CheB and theMCP methyltransferase, CheR (Figure 1). It is poss-ible that CheZ is associated via the short form ofCheA, CheAs, with MCP complexes.7,8 The smoothswim chemotactic signal (i.e. CheY-P decrease)might then be increased by aspartate-dependentactivation of CheZ that is part of such extendedreceptor signal complexes. It has also been shownthat the CheA activity of ternary Tar signal com-plexes depends upon MCP methylation level. Highmethylation levels increased CheA activity butreduced aspartate-Tar af®nity.9 ± 11

We have used an assay based on computer-assisted motion analysis of sub-saturation popu-lation responses of swimming E. coli to smalljumps of photo-released aspartate.12 The aberrantmotile bias of these strains needed to be adjustedtowards wild-type values for this purpose. Weused inducible, plasmid-encoded expression of theTsr290-470 signal domain to adjust the motor

Figure 1. Signal propagation from an E. coli Tar complexthe linker CheW (W), forms stable complexes with the MCPaspartate occupancy inactivates CheA, blocking CheA toinduces a CCW motor response. Three chemotaxis proteins (lesterase, CheB (B); the methyltransferase, CheR (R); and Cmutants lacking one or more of these proteins are the subjreduced by increased aspartate occupancy (ÿ) (not depicted)to a slower, adaptive increase in Tar methylation.

rotation bias, mb, of these mutants.13,14 The nativeTsr290-470 fragment, henceforth termed TsrSCW,stimulates CheA. A single point mutation (413AV)converts it to a CheA inactivating form, henceforthtermed TsrSCCW. TsrSCW and TsrSCCW change mbby increasing or decreasing CheY-P levels as sche-matized (Figure 2). Our results provide novelinsights into the timing and ampli®cation of thechemotactic signal.

Results

The stimulus-response relation in host��tsr strains

We ®rst measured the stimulus-response relationof the host �tsr strain RP5700.15 We wished todetermine whether the absence of Tsr, whichmakes up almost half of the total MCPs in wild-type E. coli, affected Tar-mediated aspartateresponses. Responses of swimming populationswere measured as the change in population rcd;where the rcd for a single cell path is the absoluteangular rate of change of direction of the cell cen-troid. �rcd obtained from sub-saturation popu-lation responses (Figure 3) was converted to �mb(mb � fCCW/( fCW � fCCW) where fCCW and fCW arethe counterclockwise and clockwise rotating teth-ered cell fraction, respectively, over the measuredtime interval). Stimulus strength was measured asthe apparent change in aspartate occupancy, �Rocc,estimated from adaptation times obtained in rapid-mixing experiments.16 These and other symbolsfrequently used here are listed in Table 1.

to a ¯agellar motor. The CheA kinase (A), together withs (e.g. Tar). The chemotactic signal induced by increasedCheY phosphotransfer (ÿ). The drop in CheY-P levelsblack circles) lie off the signal pathway: the MCP methy-heZ (Z), a catalyst of CheY dephosphorylation. Deletionect of this study. CheA to CheB phosphotransfer is also. The drop in CheB-P levels, the active CheB form, leads

Figure 2. Adjustment of motile bias by TsrS expression. (a) Summary of previous and present work. IPTG induc-tion of plasmid-encoded TsrSCCW was used to decrease CheY-P levels ([Y-P]), hence increase mb, in wild-type and�tsr strains. TsrSCW was used to increase CheY-P levels in wild-type, �tsr and �cheR strains, hence decrease mb, allas denoted by black arrows. In the present study, the mb of �cheZ and �cheB strains was decreased, in addition, byTsrSCCW (grey arrow). These bias-adjustments allowed decreases in mutant CheY-P levels induced by small aspartatestimuli to be evident as changes in mb. (b) Schematic explanation of how TsrS fragments change [CheY-P]. Lettersdenote proteins as in Figure 1. TsrSCCW (SCCW) binds CheA and CheW to decrease the number of MCP signalcomplexes. TsrSCW (SCW) in addition stimulates CheA. Since plasmid-encoded TsrSCW may be made at higher levelsrelative to chromosomally encoded MCPs, CheY-P production, hence CheY-P increase.

Chemotactic Signaling in E. coli. 121

Following previous work on wild-type E. coli,6

semi-log plots were used to assess the relationbetween �mb and �Rocc. Values for the �tsr strain

Table 1. Nomenclature used in this work

Symbol Definit

KD Tar-aspartate dissociation constantH Apparent Hill co-efficient for CheYkp

a CheY phosphorylation ratekÿp

a CheY-P dephosphorylation ratercd (deg. sÿ1) (Absolute frame-to-frame rate of chmb Tethered cell motor rotation bias�(�mb mb change estimated from the meast1/2 Time to reach half-maximal �rcdtr Adaptation recovery times to rapidtrmax Adaptation recovery time to a max�Rocc Apparent change in Tar occupancykex Excitation response rate�(ln 2)/t1/2

�mbexptda Predicted �mb

A Signal amplification during process�[Asp](mM) Step [Asp] change from negligible p[Asp]50(mM) �[Asp] required for a half-saturatio

Symbols repeatedly used in the text, their de®nitions and referenare de®ned where they ®rst occur.

a Formulae for these symbols are given in the Appendix.

were scattered around the best ®t for the wild-typestimulus-response relation (Figure 3, inset). Thus,the absence of Tsr does not noticeably affect aspar-

ion Reference

-motor binding

ange of direction)�(frames/s) 49fCCW/( fCW�fCCW)) 12ured �rcd 12

50-mix/photorelease stimuli 50imal stimulus 16�tr/trmax 16

16This work

ing��mb/�mbexptd This workre-stimulus value This workn response This work

ces where they were ®rst introduced. These and other symbols

Figure 3. �tsr stimulus-response relation. Example of a sub-saturation response, measured as the population rcdchange (�rcd) in response to a step stimulus of photo-released aspartate (�Rocc � 0.09, kex � 6.5(�2.4) sÿ1). The arrowdenotes the time at which aspartate was photo-released from NPEC-caged aspartate. Reference lines indicate the pre-stimulus population rcd (broken line) � frame-to-frame standard deviation (dotted line). The continuous line indicatesthe rcd value for complete smooth swimming. Inset: The �tsr data is scattered around the best ®t (broken line) to thewild-type stimulus-response data (�mb � 0.45 � 0.0665(ln(�Rocc)).

6 Continuous and dotted lines denote relations pre-dicted for stoichiometric inactivation with H � 3.5 and 10.5, respectively. The predicted �mb values decrease forassumed mb values > 0.6, as the difference between the corresponding pre-stimulus CheY-P level and the CheY-P/motor KD increases.


tate responses. Perhaps, this absence is compen-sated for by increase in other MCP levels.

Two additional parameters of the stimulus-response relation were de®ned, with reference toFigure 3. (1) Dose for half-maximal response,[Asp]50 � aspartate concentration jump from zeropre-stimulus aspartate, �[Asp], required to elicit ahalf saturation response. For wild-type E. coli,mb � 0.6 (see16). Therefore a maximal smooth swimresponse will have �mb of 1-0.6 � 0.4, while a half-maximal response will have �mb of 0.2.

(2) Chemotactic signal ampli®cation, A � ratio ofthe measured �mb over that predicted by stoichio-metric inactivation, �mbexptd. The latter was com-puted given knowledge of the stimulus strength,�Rocc and the apparent Hill coef®cient, H, forCheY-P motor interaction (see the Appendix).�mbexptd is a function, most directly, of �[Asp].Conversely, �[CheY-P] was estimated from �mbwith CheY-P � 12.5 mM for mb � 0.5.17,18 Recentwork indicates that H is 10.5 rather than 3.5, thesmaller values reported previously being supposedto be an artifact due to the population-average

determination of CheY level.19 However, even withthe higher H value, there is a substantial discre-pancy between observation and expectation(Figure 3). A assesses ampli®cation during signalprocessing, ampli®cation due to signal reception(i.e. change in motor CheY-P occupancy) beingcontained in �mbexptd values.

Perturbation of chemotactic signaling byTsrS expression

Next, we evaluated whether expression of TsrSfragments in the �tsr strain affected its chemotacticresponse sensitivity. TsrS expression levelsincreased linearly with inducer (IPTG) concen-tration from 0 to 0.5 mM, as determined in immu-noblots using Tsr antibody. The IPTGconcentration dependence of TsrSCW and TsrSCCW

expression levels was the same. Culture-to-culturevariation of rcd values with TsrS induction waswithin 25 %, provided that cultures were inocu-lated and harvested as described (Materials andMethods).


The swimming motility of the �tsr strain(rcd � 650 � sÿ1) was tuned by TsrSCW expressionto values where rcd remained sensitive to mb (upto 950 � sÿ1) (see Figure 2 of Khan et al.12 for therelationship between rcd and mb). Equivalent indu-cer levels affected a similar increase in RP8611, astrain deleted for most MCPs (Table 2); indicatingthat the rcd increase in �tsr/TsrSCW was due tostimulation of soluble CheA. At low levels(<0.2 mM IPTG), signaling was unimpaired(Figure 4(a),i). At high levels (>0.4 mM IPTG),there was little response regardless of signalstrength (data not shown). This result is probablydue to two factors. First, the decrease in [CheY-P]due to ligand-induced inactivation of MCP-boundCheA should be proportional to the activity ratioof MCP-bound CheA/total CheA. This ratioshould decrease upon TsrSCW stimulation of thesoluble CheA. Second, increased CheA activity willresult in elevated CheB-P, as well as CheY-P levels.The increased CheB activity would, in turn, lowerMCP methylation and, hence, activity of the MCP-bound CheA.

The �tsr strain became smooth-swimming uponTsrSCCW expression. In order to measure chemotac-tic responses, its bias was compensated backtowards wild-type values by acetate. Guttedstrains become overwhelmingly CW upon additionof 25 mM acetate,20 due to both phosphorylationand acetylation of CheY.21 The acetate effect wasmore modest in the �tsr strains; the rcd increasedlinearly over the 0 to 75 mM acetate concentrationrange from a basal value of ca. 650 to 950 � sÿ1.Nevertheless, acetate effectively compensatedagainst TsrSCCW expression. Presumably the rate ofacetate-based CheY modi®cation was signi®cantrelative to the rate of CheY-P generation in �tsr/TsrSCCW strains, expected to be lower than in wild-type E. coli due to the reduced number of ternaryMCP complexes (see Figure 2). High TsrSCCW indu-cer levels (0.2 to 0.5 mM IPTG) or acetate did notaffect response sensitivity (Figure 4(a),ii), in con-trast to TsrSCW expression. This difference could be

Table 2. Strains and plasmids used in this work

Strain Relevant genotype

RP437 wild-typeRP8611 �(tsr)7028�(tar-tap)5201�(trg)100zbd::Tn5RP5700 �(tsr)7028RP5099 zea::Tn10-3RP1254 �(cheR)58-13RP4971 �(cheB)63-216LLR413 �(cheB)63-216 zea::Tn10-3RP2859 �cheR-cheB �2241RP1616 �(cheZ)67-25KLR205 �(cheZ)67-25 zea::Tn10-3KLR300 �(cheZ)67-25�(tsr)7028 zea::Tn10-3KLR301 �(cheR)58-13�(tsr)7028 zec::Tn10-2KLR302 �(cheB)63-216�(tsr)7028 zea::Tn10-3PlasmidspPA56 lacO�.tsr290-470

�

pPA58 lacO�.tsr290-470413AV�

due to lower CheB activity in �tsr/TsrSCCW com-pared to �tsr/TsrSCW E. coli. The difference isexpected, since acetate-based CheY phosphoryl-ation in �tsr/TsrSCCW E. coli is due to conversionof acetate to acetyl phosphate, a phosphate donorthat does not phosphorylate CheB.20. There isno evidence, thus far, for acetate-based CheBacetylation.

Responses of �tsr/TsrSCW bacteria, at lowTsrSCW expression, and �tsr/TsrSCCW bacteria,compensated with acetate, were comparable.Response sensitivity remained substantially greaterthan predicted from stoichiometric inactivationover a substantial �Rocc range. The stimulus-response relation for �tsr/TsrS strains deviatedfrom that for wild-type and �tsr E. coli for�Rocc < 0.1 (Figure 4(b)). At �Rocc � 0.1, A is ca 10for both wild-type and �tsr/TsrS strains; but it isca 40 and 24, respectively, at �Rocc � 0.03(H � 3.5).

In addition, we examined whether signal-proces-sing times were affected. Excitation response rates,kex values, were determined from single exponen-tial ®ts. kex � (ln 2)/t1/2, where t1/2 is the half-response time. kex values and their dependence onstimulus strength were similar for wild-type andthese mutant strains. Thus, neither lack of Tsr norexpression of TsrS affected processing of signalsgenerated by aspartate stimuli. We proceeded,therefore, to construct bias-adjusted che mutants(Materials and Methods) and to evaluate theirsignaling properties.

The chemotactic signal is amplified in��cheZ mutants.

Saturation responses of �cheZ mutants to ionto-phoretically released aspartate,23 photo-releasedserine12 or glucose, acting as phosphoenol pyru-vate transport system substrate,24 have beenmeasured. The saturation kex was 50-fold lowerrelative to values obtained for wild-type E. coli,due to the absence of CheZ-catalyzed CheY-Pdephosphorylation. In this study, the responses of

Reference

511313J. S. Parkinson, pers. comm.J. S. Parkinson, pers. comm.J. S. Parkinson, pers. comm.24J. S. Parkinson, pers. comm.J. S. Parkinson. pers. comm.P1.RP5099�RP1616P1.KLR205�RP5700P1.RP1254�RP5700P1.LLR413�RP5700

1313

Figure 4. Excitation signaling in TsrS bias-adjusted hosts. (a) Excitation responses of:i, &tsr/TsrSCW (0.1 mM IPTG.�Rocc � 0.2, kex � 8.6(�1.5) sÿ1); ii, �tsr/TsrSCCW � acetate (0.2 mM IPTG, 50 mM acetate, �Rocc � 0.2 kex

� 17.9(�4.4) sÿ1) to photo-released aspartate. Arrows and reference lines are as in Figure 3. (b) Stimulus-responserelation. �tsr/TsrSCCW � acetate (triangles), �tsr/TsrSCW (inverted triangles). �Rocc values were computed as forFigure 3. The broken line is the ®t to the wild-type E. coli data. The continuous line is the best ®t to the combined�tsr/TsrS data (�mb � 0.537 � 0.118(ln(�Rocc))).


both bias-adjusted and non bias-adjusted mutantsto large step stimuli of photo-released aspartatewere measured (Figure 5(a),i) and found to besimilar to each other and to the responses reportedpreviously.

CheZ mutants were reported to be ``dif®cult toexcite''.25 We measured sub-saturation responses ofbias-adjusted �cheZ mutants (�cheZ�tsr/TsrSCCW)to small step stimuli (Figure 5(a), ii and iii), inorder to evaluate whether the measured Adepended on CheZ. Since �cheZ mutant signalprocessing times are of the order of seconds, a¯ash-lamp was used to photorelease aspartate over

the entire sample and, thus, produce concentrationjumps that were stable for many tens of seconds.16

In all cases, excitation responses initiated withoutmeasurable latency, showing that in �cheZ, as inother mutants, signal generation remained rapidrelative to downstream signal processing. The[Asp]50 for both the bias-adjusted mutant and thehost strains was 0.15 mM. Furthermore, the stimu-lus-response relation of the bias-adjusted �cheZmutants was indistinguishable from that of thebias-adjusted �tsr/TsrS hosts (Figure 5(b)). There-fore, CheZ does not affect A.

Figure 5. Excitation signaling in bias-adjusted �cheZ mutants. (a) Excitation responses to photo-released aspartate.Arrows and reference lines are as in Figure 3. i, �Rocc � 0.32, kex � 0.27(�0.09) sÿ1; ii, �Rocc � 0.21, kex � 0.57(�0.11) sÿ1; iii, �Rocc � 0.05, kex � 3.2(�1.2) sÿ1. (b) Stimulus-response relation. �Rocc values were computed as forFigure 3. Responses of bias-adjusted �cheZ mutants (®lled circles) compared against host �tsr/TsrSCW and �tsr/TsrSCCW � acetate strains (open circles). The continuous line is the best ®t to the host data. Dotted lines indicate 95 %con®dence limits. (c) Relationship between excitation response rates (kex) and amplitudes (�mb). Bars denote standarderrors. The line indicates the ®t (correlation co-ef®cient, r2 � 0.72) obtained for [CheY]T � 30 mM, H � 3.5,kÿ0 � 0.3 sÿ1, k1 � 1.7 sÿ1 and K � 0.4 (see the text).


The stimulus strength-dependence of ��cheZexcitation response times is consistent with aligand-induced occluded CheA conformation

In addition, we measured the dependence of kex

upon stimulus strength. As detailed below, quanti-®cation of this relation can discriminate amongstpossible signal generation mechanisms. In �cheZmutants, kex values should be cleanly limited byCheY-P turnover, given by the sum of the sum ofCheY phosphorylation (kp) and dephosphorylation(kÿp) rates. If CheA inactivation is the sole mode ofsmooth-swim signal generation, kex will decreasewith increased stimulus (�Rocc) or response (�mb)size, since kp decreases and kÿp is unchanged. Thekex values, were determined as before from single

exponential ®ts. While double exponential func-tions improved ®ts somewhat, presumably bytaking account of the onset of adaptive recovery,this improvement was not marked enough to merittheir use. As anticipated, kex increased withdecreasing �mb (Figure 5(c)), in marked contrast towild-type E. coli.16

However, the extent of the increase could not beexplained as due only to the decrease in kp. To rea-lize this point, note that kex � kÿp at saturationresponse, given negligible MCP-independent CheYphosphorylation. kÿp, as determined from the sat-uration kex was 0.3 sÿ1, comparable with the CheY-P auto-phosphatase rate, kÿ0.26 Bias-adjusted �cheZmutants had rcd values between 750 (mb � 0.7) to950 � (mb � 0.5) sÿ1, so their [CheY-P]pre levels lie


between 9 and 17 mM.17 ± 19 In the pre-stimulusstate, kp/kÿp � [CheY-P]pre/[CheY]pre � 13/17 for30 mM total CheY, CheYT. Therefore, kex for a closeto threshold response, kex

�mb � 0 � (kp � kÿp) � (13/17)kÿp � kÿp � (30/17) � 0.3 � 0.53 sÿ1. This value,clearly not compatible with our data, would beeven lower for signi®cant MCP-independent CheYphosphorylation.

One resolution for this anomaly would be ifreversed phosphotransfer from CheY-P to CheAcontributed signi®cantly to the pre-stimulus kÿp,and attractant ligands placed MCP-bound CheA inan occluded state, as proposed on the basis ofphosphate-exchange data.27 In this case, the post-stimulus mb, mbpost (�pre-stimulus mb � �mb )was used to compute [CheY-P]post (see theAppendix). The post-stimulus CheY phosphory-lation rate, kp

0 � kÿ0 ([CheY-P]post/(CheYT

ÿ (1 � K)[CheY-P]post), where K � k1/kÿ1. kex

(�kp0(1 � K) � kÿ0), thus obtained as a function of

�mb for different K values, provided a reasonable®t to the data (Figure 5(c)).

In wild-type E. coli, there is a slight but signi®-cant increase in kex with stimulus strength.16 Onepossibility to explain this increase was that kÿp,hence kp, increases with stimulus strength due toaccelerated CheY-P dephosphorylation. This possi-bility, may now be ruled out. The alternativepossibility, that CheY-motor dissociation partlylimits excitation signal processing in wild-typebacteria, merits examination.

The chemotactic signal is not amplified in��cheR��cheB, bias-adjusted ��cheR or��cheB mutants

�cheR�cheB mutants were known to have closeto wild-type motile bias, but a response sensitivitythat is greatly impaired relative to wild-typebacteria.12,23 �cheB mutants were dif®cult to stimu-late iontophoretically25 and required abnormallylarge photo-release of serine or aspartate for mea-surable responses.24 Responses of �cheR strains toattractant stimuli had not been measured pre-viously. In the present study, we quanti®ed thestimulus - response relation of �cheR�cheB andbias-adjusted �cheB (�cheB�tsr/TsrSCCW) and�cheR (�cheR�tsr/TsrSCW). mutants. We wishedto know whether sensitivity was impaired in allthree mutants and whether this was a result of ahigher threshold and/or a lower slope for thestimulus response relation. The mutants are alldefective in adaptation; the impaired adaptationrecovery aided determination of excitationresponse amplitudes.

Saturation smooth-swim responses wereobtained upon aspartate photo-release in thesemutants, but at higher aspartate concentrationscompared to wild-type E. coli, consistent with pre-vious studies (Figure 6(a)). The [Asp]50 was 10 mMfor the �cheR�cheB strain. It was elevated to100 mM for the bias-adjusted �cheB strain. Thisshift in the [Asp]50 values correlated with the

reduction in Tar-aspartate af®nity with increasedmethylation level, as estimated from the CheAactivity in in vitro assays.10,11

Close to saturation smooth-swimming responseswere obtained in the bias-adjusted �cheR strain.The stimulus response relation for this strain([Asp]50 � 8 mM) was close to that for �cheR�cheBmutants. The wild-type mb of �cheR�cheB strainsis due to the two MCP glutamine residues. Theseresidues are not modi®ed to glutamate in thismutant, due to the absence of CheB, which func-tions as a deamidase as well as methylesterase.Amide groups increase the activity of the MCP-bound CheA comparable to glutamate methylesters.1 In comparison, the activity of the MCP-bound CheA should be much lower in the �cheRstrain, an expectation consistent with the free-swimming phenotype of this mutant (Materialsand Methods). However, the �cheR mutant is notcompletely smooth swimming. This indicates thatthe MCPs are not completely deamidated in thisstrain, the residual presumably re¯ecting a balancebetween MCP synthesis and CheB deamidaseactivity. The estimated Tar-aspartate KD valuesfor 1/4 and 1/2 amidated Tar ternary complexeswere 9(�1) and 23(�4) mM, respectively;11 i.edifferent by only a factor of 2. Thus, the similarstimulus-response curves of the �cheR�cheB andbias-adjusted �cheR strain may be explained if thefractional MCP amidation in the latter is ca 1/4.

In addition, the stimulus-response relation forthese mutants was not logarithmic, in contrastto wild-type E. coli. The mutant data could not be®t by any equation of the form�mb � b((log(�[Asp]) ÿ log (�[Asp] � KD)). Theaspartate jump (�[Asp])-dependence of responseamplitudes (�mb's) could be ®t by a single bindingisotherm when the chemotactic signal wasassumed to be generated solely by stoichiometricinactivation (Figure 6(b)). The best-®t KD (13.5 mM)for the ®t to the combined �cheR�cheB and bias-adjusted �cheR data was close to the best-®t singlesite Tar-aspartate KD (10.5 mM), determined fromthe concentration-dependence of adaptation recov-ery in wild-type E. coli,16 as well as the KD valuesfor the 1/4-1/2 amidated Tar complexes.11 Therelations all seem somewhat more co-operativethan predicted from stoichiometric inactivation,consistent with the modest positive co-operativityfor aspartate-induced CheA inactivation reportedby Bornhorst & Falke.11 Thus, the chemotacticresponse sensitivity of these mutants is in goodagreement with current biochemical knowledge(i.e. A � 1).

The mutants have widely differing MCP amida-tion/methylation levels. Therefore, this parameterper se cannot be responsible for the observed lossof signal ampli®cation in these strains. Excitationresponse rates (kex) may be ruled out as a par-ameter affecting signal ampli®cation. kex values for�cheR�cheB, bias-adjusted �cheR or �cheBmutants varied between 17 and 3 sÿ1, increasingslightly with �Rocc as observed for wild-type

Figure 6. Excitation signaling inMCP methylation/demethylationenzyme mutants. (a). Excitationresponses. Arrows and referencelines are as in Figure 3. i,�cheR�cheB (�[Asp] � 32 mM,kex � 16.9(�3.0) sÿ1); ii, �cheR�cheB(�[Asp] � 8 mM, kex � 7.4(�1.8)sÿ1); iii, bias-adjusted �cheR(0.1 mM IPTG, �[Asp] � 32 mM,kex � 9.3(�1.0) sÿ1); iv, bias-adjusted�cheB (0.2 mM IPTG,�[Asp] � 230 mM, kex � 4.6(�-0.5) sÿ1). (b) �[Asp]-dependence ofexcitation response amplitudes(�mb). �cheR�cheB (open circles),bias-adjusted �cheR (gray circles),bias-adjusted �cheB (black circles).The dotted line denotes the best ®t(r2 � 0.7, H � 3.2, KD � 13.5 mM) tothe combined �cheR�cheB, bias-adjusted �cheR data for the stoichio-metric inactivation model, H and KD

being allowed to vary. The data setswere combined, since the differencebetween them cannot be meaning-fully distinguished given culture-to-culture variations. H will be under-estimated in the ®t, however, sincethe combined data are spread over awider �[Asp] range. The dashedline denotes the corresponding best®t (r2 � 0.86, KD � 177 mM) to thebias-adjusted �cheB data with Htaken as 3.5.


E. coli (Figure 6(a)) in spite of the vastly differentresponse sensitivities.

Bias adjustment does not restore chemotaxis

Finally, we evaluated whether bias-adjustmentrestored chemotaxis as assessed by swarming onsemisolid agar. �cheR�cheB and bias-adjusted�cheR are known to swarm better than �cheR or�cheB mutants alone. This improvement resultsfrom ``pseudo-taxis'', namely improved ability ofthe bacteria to migrate in semi-solid medium due

to restoration of wild-type swim-tumble bias.14 Inaddition we measured swarm rates of bias-adjusted �cheB and �cheZ mutants, as well as ofthe �tsr/TsrSCW control.

For all mutants, swarming was maximal whenmb was altered to match wild-type values (0.1 to0.2 mM IPTG induction). However, maximalmutant swarm rates remained substantially belowvalues obtained for the �tsr/TsrS hosts (Figure 7).This result was expected for �cheR�cheB and bias-adjusted �cheB, �cheR mutants consistent withtheir greatly reduced response sensitivity. Interest-

Figure 7. Migration of bias-adjusted strains on semisolid agar. Swarm rates of the bias-adjusted strains were nor-malized by division with rates obtained when TsrS locked in the opposite mode was expressed in the strains as con-trol. Swarm rates were constant for the control strains over the 0 to 0.4 mM IPTG range, as reported,14 indicating thatTsrS expression at these inducer levels was not deleterious. Induction by 1 mM IPTG reduced swarm size. At 0 mMIPTG, swarm rates were 0.56(�0.03) mm/hour, 0.64(�0.03) mm/hour and 0.66(�0.03) mm/hour for bias-adjusted�cheR, �cheB and �cheZ mutants, respectively. Swarm rates were 2.6(�0.1) mm/hour (0 mM IPTG) and3.1(�0.2) mm/hour (0.1 mM IPTG) for �tsr/TsrSCW and 2.35(�0.2) mm/hour (0 mM IPTG) for �tsr/TsrSCCW hoststrains.


ingly, however, the improvement in swarm abilityof the bias-adjusted �cheZ mutants was also slight,though the response sensitivity of these mutantswas comparable to that of the �tsr/TsrS hosts.Conversely, the �tsr/TsrSCW strain when inducedto have mb comparable to the tumbly �cheZmutant out-performed the bias-adjusted version inswarm assays. Thus, the chemotactic ability of�cheZ mutants in swarm assays is limited by fac-tors other than steady-state mb or response sensi-tivity. �cheZ mutants should have lower CheY-Pturnover as opposed to the host strain, due to theabsence of CheZ-catalyzed CheY-P dephosphoryla-tion, even for close-to-threshold responses. Inaddition, they may be impaired in adaptation.28

One or both of these factors may affect migrationin semi-solid agar.

Discussion

The present study is the ®rst where liberation oflocked receptor domains has been used to separateeffects of mutations on chemotactic signaling fromassociated changes in steady-state motility. Assess-ment of the strategy as a general tool for analysisof chemotaxis mutants and the implications of our®ndings for smooth-swim signal generation, timingand ampli®cation are detailed below.

Strategies for bias adjustment.

The large genetic mutant database has proveddecisive in making possible current understandingof bacterial chemotaxis. It is likely to play a similar

pivotal role as focus shifts from atomic leveldescription of the individual components1 to dis-section of their dynamic interplay that results inthe integrated behavioral response. Chemotaxismutants typically have an extreme motile bias thatprevents or hinders evaluation of their signalingproperties. It is thus important for such dissectionto evolve strategies for separation of the effectson steady-state motility from those on chemotacticsignaling.

A number of strategies have been used to restoremotile bias. One approach is to combine mutationsthat generate opposing motile phenotypes.29 Singlemutant properties cannot be analyzed using thisapproach. Another approach has been the use ofCW or CCW biased motor mutants.26 Thesemutants typically have impaired responsesensitivities,12 presumably due to impairment ofthe motor switch. In addition, mb has beenincreased by CheY overproduction.30,31 CheY over-production cannot be used to decrease mb and, inaddition, will have multiple effects due to alteredCheY interactions with CheA, CheZ and themotor. Finally, swim-tumble bias may be con-trolled by expression of receptors with altered sig-naling activity. Thus, chromosomal expression ofwild-type and smooth-swim mutant Salmonellatyphimurium Tar was used to demonstrate that che-motaxis proceeded ef®ciently from different initialtumbling frequencies.32

Here, liberation of a receptor cytoplasmic signaldomain (TsrS) rather than intact receptors wasused to modulate bias. TsrS lacks periplasmic andtransmembrane domains, one methylation segment


and the CheR tether peptide. Its biochemical inter-actions are thus more restricted and have beencharacterized.13,14 We have found that controlledTsrS expression modulates swim-tumble bias with-out greatly perturbing signal processing or ampli®-cation. kex values varied with amplitude in the�tsr/TsrS hosts in a manner similar to that ofwild-type E. coli; A was comparable above�Rocc > 0.1 (Figure 3). Thus, sensitivity was littleaffected by bias adjustments, mediated by TsrScw

or TsrSccw/acetate, that caused modest (two- tofourfold) decreases in the fractional contribution ofMCP-bound CheA to maintenance of steady-stateCheY-P. The parameters that control sensitivitymust differ in this regard from those that set stea-dy-state motile bias, since the latter was appreci-ably altered by the same TsrS expression levels.

Figure 8. Mechanisms for signal ampli®cation. (a) Modeattractant ligand is greater than computed from ligand-recepII, ligand binding places the receptor-coupled CheA in a conIII, the inactive CheA conformation is coupled to an activ(b) Fits (dotted lines) to the stimulus-response relation (brokedispersed MCP signal complexes � 1, N � 100, ni � 15. ii, Mnotation and formulae.

Mechanistic implications forsignal amplification

Current ideas regarding chemotactic signalampli®cation may be broadly divided into twogroups; those exhibiting enhanced CheA inacti-vation or CheY-P destruction (Figure 8).

Within the ®rst group, inactivation of CheA isthe sole smooth swim signal, but ligand bindingeffects a greater change in CheY-P levels than pre-dicted from stoichiometric inactivation. Confor-mational spread, in which a single ligand-bindingevent inactivates many clustered MCP signal com-plexes is a model of this type.33 The occurrence ofMCP clusters,34 in trans inactivation of CW-lockedTar receptor complexes in vesicle preparations10

and signaling by hetero-dimeric receptors with a

l I, the fraction of CheA (A) inactivated by binding oftor binding af®nity due to co-operative sensing.33 Modelformation where it acts as a CheY-P phosphatase. Modelated CheZ (Z) conformation, possibly via CheAs (As).n line) of wild-type E. coli. i, Model I, ratio of clustered/

odel III, kz0 � 20.kz, a � 0.1, b � 6. See the AppendixX for


single cytoplasmic domain, most likely also intrans35 are observations consistent with receptorcross-talk (model I).

Conceptually, the most straightforward ampli®-cation mechanism based on CheY-P destruction isutilization of attractant ligand-binding energy forreversal of CheA phospho-transfer (model II).Phospho-histidine and, more so, phospho-aspartateresidues have high transfer free energies relative toATP and must be stabilized by the proteinenvironment.36 Therefore, phospho-transfer reac-tions may, in principle, be substantially altered byligand-induced protein conformational transitions.Reverse CheA-CheY phosphotransfer has beendemonstrated. Forward E. coli CheA-CheY phos-pho-transfer is favored in vitro;37 but E. coli CheAP1 fragments accept phosphate from CheY presentin molar excess,38 while a Rhizobium meliloti CheYhomolog is dephosphorylated by CheA.39 CheY-Pdestruction models involving accelerated CheZ-catalyzed CheY dephosphorylation provide a moresophisticated alternative. Compared to reversedCheA-CheY phospho-transfer schemes, these havethe advantage that CheY-P destruction is separablefrom CheA inactivation. In vitro, complexes ofCheZ with CheAs have enhanced CheY-P depho-sphorylation activity.8 In addition, CheAs/CheA/CheW complexes have been isolated.7 Regulationof signal ampli®cation may thus be achieved byenvironmental modulation of the association ofCheZ via CheAs with MCP signal complexes; adesirable strategy for an organism that encountersdiverse feast or famine regimes (model III).

We ®rst considered whether the non-linear exci-tation stimulus-response relation determined forwild-type E. coli favored a particular model. Forthe ®ts, the kinetic formalism presented earlier,6

was extended by incorporation of the equationdescribing conformational spread33 or addition of aCheZ activation term (see the Appendix). Wefound that for conformational spread, a 1:1 ratio ofdispersed to clustered MCPs was required to ®tthe data over the entire range, consistent with theconclusion that MCP dispersal is advantageous forextending response range.33 The simulations indi-cated, however, that activation of CheZ-catalyzedCheY-P dephosphorylation, comparable to thatobserved experimentally,8 also ®ts the relation.Alternatively, ligand-induced reversed phospho-transfer ®ts this relation, even though biochemicaldata to constrain parameter values are not avail-able for this case. In conclusion, therefore, the non-linear stimulus-response relation determined forwild-type E. coli cannot discriminate among signalampli®cation mechanisms.

Signal amplification does not involveaccelerated CheY-P destruction

Our data show that a large fraction of the ampli-®cation measured for wild-type E. coli is present in�cheZ mutants. In addition, they provide evidencethat chemotactic signaling in these mutants

involves an occluded CheA state incapable ofphosphate transfer, as ®rst deduced by Borkovich& Simon.26 Accordingly, models II and III, basedon CheA-mediated CheY-P destruction, may beruled out for responses at the low to negligiblebackground levels used in our photo-releaseassays.

Amplified CheA inactivation due to CheRand/or CheB

The chemotactic signal ampli®cation measuredin the photo-release assays must, by elimination,be due to enhancement of ligand-induced CheAinactivation. This ampli®cation is abolished by lossof either CheR or CheB, as established by examin-ation of �cheR�cheB, bias-adjusted �cheR and�cheR mutants. This result has the followingimportant implications. First, it establishes that thebulk of the signal ampli®cation measured in ourassays cannot be due to CheY-motor binding co-operativity, as argued earlier on theoreticalgrounds (Figure 3, inset), since there is no evidencethat either CheR or CheB participate in or in¯uencethis interaction. Second, similar reasoning furtherrules out a role for CheY-CheZ interactions.Finally, as noted earlier, differences in MCP amida-tion level per se are ruled out as a possible sourceof signal ampli®cation.

Alternative explanations must be sought based,most directly, on interactions of CheR and CheBwith MCP ternary complexes. One possibility isfacilitation of conformational spread.40 If so, thiscannot be simply due to increase in number or sizeof the MCP clusters seen by immuno-electronmicroscopy. MCP clusters in �cheR�cheB mutantsare indistinguishable from those in wild-typeE. coli.41 In addition, CheR, a small (33 kDa), com-pact molecule42 relative to the MCP complexes(280 kDa), is present in too few copies per cell (ca850 compared to 1500 for Tar alone, and ca 4000for CheA and CheW each (Table 1 of1) to form acontinuous meshwork required for such a structur-al role. CheR and CheB must facilitate confor-mational spread in more subtle ways.

Biochemical evidence and simulations suggestthat CheB and CheR bind preferentially to activeand inactive MCP signal complexes respectively,enabling precise adaptation.43 If so, either absenceof CheR or disabled CheB-MCP interactions couldbe responsible for the loss of ampli®cation in�cheR/TsrSCW strains. Even though CheR and/orCheB - MCP interactions have not been consideredin conformational spread models, recent analysesthat model the modulation of receptor confor-mational lifetimes by ligand or downstream circuitcomponents40,44 provide clues to their possible role.Speci®cally, CheB and CheR molecules tethered toMCPs within a cluster may serve as antennae byprolonging lifetimes of the MCP conformation towhich they bind in trans. Differential bindingwould allow the proteins to play a complementaryrole to both extend response extent and enhance


response sensitivity. Thus, stabilization by CheBcould increase the MCP fraction in the active sig-naling conformation, thereby increasing the extentto which CheA may be inactivated by attractantligands. CheR binding to the inactive MCP confor-mation induced by attractant ligand could enhanceits spread within a cluster, thereby effecting ampli-®cation. The proteins would reverse roles forresponses to repellent stimuli. Ampli®cation wouldbe a property of the spatio-temporal conformation-al ¯uctuations, rather than static organization, ofthe MCP receptor clusters.

Materials and Methods

Strains and plasmids

Table 1 lists the strains used. Double mutants wereconstructed by moving che genes into the �tsr strainRP5700 using P1 transduction as previously described.24

Transductants were selected on the basis of tetracyclineresistance carried on linked Tn10 markers. The linkagefrequency between the Tn10 marker and the meche oper-on was 20-30 %. Single colonies were isolated and theirmotility in semi-solid swarm agar (0.35 % agar) andliquid media characterized (see below). Colonies thattested positive were saved for further study as glycerolstocks at ÿ20 �C.

Plasmids pA56 and pA58 are derived from thepBR322-based expression vector pTM30.45 They code forTsrS290-470 (TsrSCCW) and TsrS290-470413AV (TsrSCW)respectively, under lac control. Ampicillin-resistant trans-formants were screened following Ames et al.14 In brief,a pool of about 100 colonies was streaked on swarmagar. Motile ¯ares emerging from the streaks wereplated on hard agar to yield single colonies. Swarmingand swimming behavior of three to ®ve such colonieswas examined in detail. Colonies exhibiting inducer-dependent adjustment of swim-tumble bias were savedfor further analysis.

Chemicals and media

Antibiotics and IPTG were from Sigma Chemicals (StLouis, MO). Tetracycline and ampicillin were used at®nal concentrations of 25 mg/ml. Motility buffer was20 mM Na2HPO4/KH2PO4 (pH 7.0 � 0.2), 10 mM pot-assium chloride, 0.1 mM EDTA, 5 mM lithium lactate,125 mM methionine, 5 mM dithiothreitol}16 plus 0.1 mMIPTG. Caged HPTS, the 1-(2-nitrophenyl)ethyl ether ofHPTS (8-hydroxypyrene-1,3,6-trissulphonic acid), andNB-caged aspartate, the b-2,6-dinitrobenzyl ester ofL-aspartic acid have been described.16 Caged HPTS has aproduct quantum yield, Qp, of 0.13 and a photolysis rateof 550 sÿ1 at pH 7.0, 0.1 M ionic strength and 22 �C. NB-caged aspartate has Qp of 0.21 and photolysis rate of630 sÿ1 under the same conditions. NB-caged aspartatewas used in a number of experiments on �tsr and �tsr/TsrS hosts in order to relate responses of these strains tothose documented for wild-type E. coli.6 The cagedHPTS was used as a caged ¯uorophore to de®ne theamplitude and temporal stability of the aspartate jumpsadministered by photolysis of the caged aspartate.All caged compounds were handled as describedpreviously.16

N-1-(2-Nitrophenyl)ethoxycarbonyl-L-aspartate (NPEC-caged aspartate) was synthesized essentially as described

for N-1-(2-nitrophenyl)ethoxycarbonyl-L-serine.12 Then325 mg (2.47 mmol) of L-aspartic acid was added withstirring to 19.8 ml of ice-cold, aqueous 0.25 M NaOH.Separate solutions of 4.49 mmol 1-(2-nitrophenyl)ethylchloroformate in 22 ml of tetrahydrofuran and of 12.8 mlof 0.5 M aqueous KOH were added simultaneously, tomaintain the pH between 8 and 9. The solution was stir-red overnight at 22 �C at which time the pH was 5.95. Thesolution was near-saturated with NaCl and extractedwith diethyl ether. Back-washing of the ether phase usinga saturated NaCl solution and distilled water was carriedout in case some caged aspartate had been extracted intoether. Thin-layer chromatographic (TLC) analysis of thecombined aqueous phases using K6F silica plates (What-man) and n-butanol/glacial acetic acid/distilled water(10:2:3 by vol.) showed a single UV-absorbing spot, andno free aspartate (ninhydrin assay).

The combined aqueous phases were adjusted to pH 2.5with aqueous HCl and extracted with ethyl acetate. Thecombined ethyl acetate phases were washed with satu-rated NaCl solution and then dried with magnesium sul-fate, ®ltered and evaporated to dryness. Water (50 ml)was added and the solution was neutralized withNaOH. An aliquot containing 250 mmol of the cagedaspartate was applied at 5 �C to a DEAE cellulose col-umn (2.2 cm � 30 cm) equilibrated with 10 mM triethy-lammonium bicarbonate (TEAB), pH 7.4, and elutedwith a linear 10 mM-300 mM TEAB gradient (totalvolume 2000 ml). The product eluted at ca 180 mMTEAB. HPLC analysis (Whatman Partisphere SAX col-umn, mobile phase 0.1 M (NH4)2HPO4 adjusted to pH 6.0with HCl) showed a single peak absorbing at 260 nm,with a retention time of four minutes. The pooled frac-tions were rotary-evaporated under vacuum and re-evaporated (three times) from methanol to removeexcess TEAB, then dissolved in a small amount of waterand stored at ÿ20 �C. The yield, determined by UVabsorbance spectroscopy at 263 nm (e 4700 Mÿ1 cmÿ1)was 49 % (calculated on the basis of the whole reactionmixture being processed). A portion was converted tothe potassium salt with K �-Dowex 50. The productquantum yield for photolysis of NPEC-caged aspartatewas taken to be 0.65 (�0.05) by analogy with otherNPEC-caged amino acids.12,46 NPEC-caged aspartatephotolyzed at 17 sÿ1 in 20 mM KH2PO4/K2HPO4

(pH 7.0), 15 mM KCl, 1 mM dithiothreitol buffer at22 �C. NPEC-caged aspartate was characterized by the1H-NMR spectrum of its K � salt, and was a clean,hydrolytically stable, homogeneous product. It contained0.10 % L-aspartic acid contamination, as determined byamino acid analysis, in contrast to 1.5 % for NB-cagedaspartate.16 Importantly, this allowed generation of largeaspartate jumps from low or negligible pre-stimulusvalues. The concentration jump, �[Asp], could then beused as a direct measure of stimulus strength. NPEC-caged aspartate was used for most experimentsdescribed in this study. Responses for a given �Rocc

obtained with either NB or NPEC-caged aspartate weresuperimposable.

Motility assays

Swarming, free-swimming and tethered cell motilityof the bacteria was characterized. Swarming on semiso-lid agar at 30 �C was determined. Samples (1 ml) fromovernight cultures were spotted onto the plates: 12 hourslater, measurements of swarm diameter out from theinoculated spots were started and continued every twohours for the next eight to ten hours.

Figure 9. Adjustment of motile bias upon TsrS expression. i, �tsr�cheZ mutants. TsrSCCW expression (0.2 mMIPTG). ii, �tsr�cheR mutants. TsrSCW expression (0.1 mM IPTG). Arrows denote absorbance at the time of IPTGaddition. rcd values were recorded at 30-45 minute intervals after addition (®lled symbols, �IPTG; open symbols,ÿIPTG). Bars denote sample measurement error. The rcd values were mapped onto (CCW/(CW � CCW)) rotationbias, mb, using an empirically determined relation, linear over the 0.35 to 1.00 mb range.12


For swimming and tethered cell assays, innocula fromovernight cultures of the bacteria, at 1/100 dilution,were grown in a 10 ml volume of tryptone broth at 35 �Cin a side-arm ¯ask to an optical density (A600) of 0.1 to0.2. At this stage, cultures were divided into two ali-quots, IPTG added to one aliquot and the motility inboth aliquots followed with time. Sample measurementerror was small compared to sample-to-sample or cul-ture-to-culture variations. The rcd increased towards lateexponential phase, consistent with the reported increasein tumbling frequency.47 Modest TsrSccw induction(0.2 mM IPTG) blocked the growth phase dependent rcdincrease in �cheZ and �cheB mutants. Higher TsrSccw

levels obtained upon induction with 0.3 to 0.5 mM IPTGdid not notably affect the rcd further. The population rcdof �cheR mutants increased upon 0.1 mM IPTG induc-tion of the TsrSCW domain (Figure 9).

Immunoblot analysis

TsrS expression levels were quanti®ed by enhancedchemi-luminescence (ECL) utilizing published proto-cols.48 Antibody against TsrS was a gift from Dr J.S.Parkinson.13

Photo-release assays

The assays were conducted as described previously16

with the following modi®cations. (1) For compensationof the �tsr/TsrSccw host strain by acetate, the bacteriawere washed and re-suspended in motility buffer con-taining acetate. (2) An elliptical ¯ash-lamp (model JML-E. Gert Rapp Opto-Electronics, Hamburg, Germany) wasused. This circumvented use of the liquid light guideemployed previously,6 allowing about ®ve times moreper ¯ash to be photo-released globally in the sample. (3)For tumbly mutant responses, a low-power 10� objec-tive (Nikon CF-¯uor) was used to maintain the bacteriain the depth of ®eld during tumbling events.

Data analysis

The output from the CCD video-camera (MAC-HSC180) was digitized directly using a VP320 processor(Motion Analysis Inc.) or recorded onto videotape for

later analysis off-line. Centroid and path ®les were com-puted from the digitized video data. Rcd, spd and avelmotion analysis operators (ExpertVision software,Motion Analysis Inc., Santa Rosa, CA) were used tomeasure population swimming and tethered cellresponses using algorithms that have been described.12,16

Excitation response times and amplitudes were deter-mined from the plots of population rcd with time as pre-viously detailed.6,16 Model simulations were performedutilizing programs written in Microsoft BASIC (version3.0). Non-linear least-squares ®ts (Sigmaplot 4.0, JandelScienti®c. San Rafael, CA) were made to the data.

Acknowledgements

We thank Dr David R. Trentham and Gordon P. Reidfor synthesis of the NPEC-caged aspartate; Dr John S.Parkinson for strains, plasmids and advice on their use;and Dr J. Corrie for comments on the manuscript. Thiswork was supported by National Institutes of Healthgrant GM RO1-49319 (to S.K.).

References

1. Falke, J. J., Bass, R. B., Butler, S. L., Chervitz, S. A.& Danielson, M. A. (1997). The two-component sig-naling pathway of bacterial chemotaxis: a molecularview of signal transduction by receptors, kinases,and adaptation enzymes. Annu. Rev. Cell. Devel. Biol.13, 457-512.

2. Macnab, R. M. & Ornston, M. K. (1977). Normal-to-curly ¯agellar transitions and their role in bacterialtumbling. Stabilization of an alternative quaternarystructure by mechanical force. J. Mol. Biol. 112, 1-30.

3. Larsen, S. H., Reader, R. W., Kort, E. N., Tso, W. W.& Adler, J. (1974). Change in direction of ¯agellarrotation is the basis of the chemotactic response inEscherichia coli. Nature, 249, 74-77.

4. Berg, H. C. & Brown, D. A. (1972). Chemotaxis inEscherichia coli analyzed by three-dimensional track-ing. Nature, 239, 500-504.


5. Spiro, P. A., Parkinson, J. S. & Othmer, H. (1997).A model of excitation and adaptation in bacterialchemotaxis. Proc. Natl Acad. Sci. USA, 94, 7263-7268.

6. Jasuja, R., Lin, Y., Trentham, D. R. & Khan, S.(1999b). Response tuning in bacterial chemotaxis.Proc. Natl Acad. Sci. USA, 96, 11346-11351.

7. McNally, D. F. & Matsumura, P. (1991). Bacterialchemotaxis signaling complexes: formation of aCheA/CheW complex enhances autophosphoryla-tion and af®nity for CheY. Proc. Natl Acad. Sci. USA,88, 6269-6273.

8. Wang, H. & Matsumura, P. (1996). Characterizationof the CheAs/CheZ complex: a speci®c interactionresulting in enhanced dephosphorylating activity onCheY-phosphate. Mol. Microbiol. 19, 695-703.

9. Dunten, P. & Koshland, D. E., Jr (1991). Tuning theresponsiveness of a sensory receptor via covalentmodi®cation. J. Biol. Chem. 266, 1491-1496.

10. Borkovich, K. A., Alex, L. A. & Simon, M. I. (1992).Attenuation of sensory receptor signaling bycovalent modi®cation. Proc. Natl Acad. Sci. USA, 89,6756-6760.

11. Bornhorst, J. A. & Falke, J. J. (2000). Attractantregulation of the aspartate receptor-kinase complex:limited cooperative interactions between receptorsand effects of the receptor modi®cation state.Biochemistry, 39, 9486-9493.

12. Khan, S., Castellano, F., Spudich, J. L., McCray, J. A.,Goody, R. S., Reid, G. P. & Trentham, D. R. (1993).Excitatory signaling in bacteria probed by cagedchemoeffectors. Biophys. J. 65, 2368-2382.

13. Ames, P. & Parkinson, J. S. (1994). Constitutivelysignaling fragments of Tsr, the Escherichia coli serinechemoreceptor. J. Bacteriol. 176, 6340-6348.

14. Ames, P., Young, A. & Parkinson, J. S. (1996). Meth-ylation segments are not required for chemotacticsignaling by cytoplasmic fragments of Tsr, themethyl-accepting serine chemoreceptor of Escherichiacoli. Mol. Microbiol. 19, 737-746.

15. Ames, P. & Parkinson, J. S. (1988). Transmembranesignaling by bacterial chemoreceptors: E. coli trans-ducers with locked signal output. Cell, 55, 817-26.

16. Jasuja, R., Keyoung, J., Reid, G. P., Trentham, D. R.& Khan, S. (1999a). Chemotactic responses of Escher-ichia coli to small jumps of photoreleased L-aspartate.Biophys. J. 76, 1706-1719.

17. Alon, U., Camarena, L., Surette, M. G., Aguera, Y.,Arcs, B., Liu, Y., Leibler, S. & Stock, J. B. (1998).Response regulator output in bacterial chemotaxis.EMBO. J. 17, 4238-4248.

18. Scharf, B. E., Fahrner, K. A., Turner, L. & Berg, H. C.(1998). Control of direction of ¯agellar rotation inbacterial chemotaxis. Proc Natl Acad. Sci. USA, 95,201-206.

19. Cluzel, P., Surette, M. & Leibler, S. (2000). An ultra-sensitive bacterial motor revealed by monitoringsignaling proteins in single cells. Science, 287,1652-1655.

20. Dailey, F. E. & Berg, H. C. (1993). Change in direc-tion of ¯agellar rotation in Escherichia coli mediatedby acetate kinase. J. Bacteriol. 175, 3236-3239.

21. Ramakrishnan, R., Schuster, M. & Bourret, R. B.(1998). Acetylation at Lys-92 enhances signaling bythe chemotaxis response regulator protein CheY.Proc. Natl Acad. Sci. USA, 95, 4918-4923.

22. Lukat, G. S., McCleary, W. R., Stock, A. M. & Stock,J. B. (1992). Phosphorylation of bacterial responseregulator proteins by low molecular weight phos-pho-donors. Proc. Natl Acad. Sci. USA, 89, 718-22.

23. Segall, J. E., Block, S. M. & Berg, H. C. (1986). Tem-poral comparisons in bacterial chemotaxis. Proc. NatlAcad. Sci. USA, 83, 8987-8991.

24. Lux, R., Munasinghe, V. R. N., Castellano, F.,Lengeler, J., Corrie, J. E. T. & Khan, S. (1999). Eluci-dation of a PTS-carbohydrate chemotactic signalpathway using a time-resolved behavioral assay.Mol. Biol. Cell, 10, 1133-1146.

25. Block, S. M., Segall, J. E. & Berg, H. C. (1982).Impulse responses in bacterial chemotaxis. Cell, 31,215-226.

26. Hess, J. F., Oosawa, K., Kaplan, N. & Simon, M. I.(1988). Phosphorylation of three proteins in thesignaling pathway of bacterial chemotaxis. Cell, 53,79-87.

27. Borkovich, K. A. & Simon, M. I. (1990). Thedynamics of protein phosphorylation in bacterialchemotaxis. Cell, 63, 1339-1348.

28. Blat, Y., Gillespie, B., Bren, A., Dahlquist, F. W. &Eisenbach, M. (1998). Regulation of phosphataseactivity in bacterial chemotaxis. J. Mol. Biol. 284,1191-1199.

29. Liu, J. D. & Parkinson, J. S. (1989). Role of CheWprotein in coupling membrane receptors to the intra-cellular signaling system of bacterial chemotaxis.Proc. Natl Acad. Sci. USA, 86, 8703-8707.

30. Clegg, D. O. & Koshland, D. E., Jr (1984). The roleof a signaling protein in bacterial sensing: behavioraleffects of increased gene expression. Proc. Natl Acad.Sci. USA, 81, 5056-5060.

31. Wolfe, A. J., Conley, M. P., Kramer, T. J. & Berg,H. C. (1987). Reconstitution of signaling in bacterialchemotaxis. J. Bacteriol. 169, 1878-1885.

32. Weis, R. M. & Koshland, D. E., Jr (1990). Chemotaxisin Escherichia coli proceeds ef®ciently from differentinitial tumble frequencies. J. Bacteriol. 172, 1099-1105.

33. Bray, D., Levin, M. D. & Morton-Firth, C. J. (1998).Receptor clustering as a mechanism to control sensi-tivity. Nature, 393, 85-88.

34. Maddock, J. R. & Shapiro, L. (1993). Polar locationof the chemoreceptor complex in the Escherichia colicell. Science, 259, 1717-1723.

35. Gardina, P. J. & Manson, M. D. (1996). Attractantsignaling by an aspartate chemoreceptor dimer witha single cytoplasmic domain. Science, 274, 425-426.

36. Stock, J. B., Surette, M. G., Levitt, M. & Park, P.(1995). Two-component signal transduction systems:structure-function relationships and mechanisms ofcatalysis. In Two-component Signal Transduction(Hoch, J. A. & Silhavy, T. J., eds), pp. 25-51,American Society for Microbiology Press,Washington, DC.

37. Li, J., Swanson, R. V., Simon, M. I. & Weis, R. M.(1995). The response regulators CheB and CheYexhibit competitive binding to the kinase CheA.Biochemistry, 34, 14626-14636.

38. Garzon, A. & Parkinson, J. S. (1996). Chemotacticsignaling by the P1 phosphorylation domain liber-ated from the CheA histidine kinase of Escherichiacoli. J. Bacteriol. 178, 6752-6758.

39. Sourjik, V. & Schmitt, R. (1998). Phosphotransferbetween CheA, CheY1 and CheY2 in the chemotaxissignal transduction chain of Rhizobium meliloti. Bio-chemistry, 37, 2327-2335.

40. Duke, T. A. J. & Bray, D. (1999). Heightened sensi-tivity of a lattice of membrane receptors. Proc. NatlAcad. Sci. USA, 96, 10104-10108.

41. Lybarger, S. R. & Maddock, J. R. (1999). Clusteringof the chemoreceptor complex in Escherichia coli is


independent of the methyltransferase CheR and themethylesterase CheB. J. Bacteriol. 181, 5527-5529.

42. Djordojevic, S. & Stock, A. M. (1997). Crystal struc-ture of the chemotaxis methyltransferase CheRsuggests a conserved structural motif for bindingS-adenosylmethionine. Structure, 5, 545-558.

43. Morton-Firth, C. J., Shimizu, T. S. & Bray, D. (1999).The Tar receptor signaling complex: a two-statemodel of a processing unit. J. Mol. Biol. 286, 1059-1074.

44. Shea, L. & Linderman, J. (1997). Mechanistic modelof G-protein signal transduction: determinants ofef®cacy and effect of coupled receptors. Biochem.Pharmacol. 53, 519-530.

45. Morrison, T. B. & Parkinson, J. S. (1994). Liberationof an interaction domain from the phosphotransferregion of CheA, a signaling kinase of Escherichia coli.Proc. Natl Acad. Sci. USA, 91, 5485-5489.

46. Corrie, J. E. T., DeSantis, A., Katayama, Y.,Khodakhah, K., Messenger, J. B., Ogden, D. C. &Trentham, D. R. (1993). Postsynaptic activation atthe squid giant synapse by photolytic release of L-glutamate from a ``caged'' L-glutamate. J. Physiol.465, 1-8.

47. Amsler, C. D., Cho, M. & Matsumura, P. (1993).Multiple factors underlying the maximum motilityof Escherichia coli as cultures enter post-exponentialgrowth. J. Bacteriol. 175, 6238-6244.

48. Zhao, R., Amsler, C. D., Matsumura, P. & Khan, S.(1996). FliG and FliM distribution in the Salmonellatyphimurium cell and ¯agellar basal bodies.J. Bacteriol. 178, 258-265.

49. Sundberg, S. A., Alam, M. A. & Spudich, J. L.(1986). Excitation signal processing times in Halobac-terium halobium phototaxis. Biophys. J. 50, 895-900.

50. Khan, S., Spudich, J. L., McCray, J. A. & Trentham,D. R. (1995). Chemotactic signal integration in bac-teria. Proc. Natl Acad. Sci. USA, 92, 9757-9761.

51. Parkinson, J. S. (1978). Complementation analysisand deletion mapping of Escherichia coli mutantsdefective in chemotaxis. J. Bacteriol. 135, 45-53.

Appendix

Simulations of Signal Ampli®cation Models

CheY !kp

kÿp

CheY-P

Pre-stimulus rate of CheY phosphorylation,kp � {k0 � csa [SA-P]}[CheY]

Pre-stimulus rate of CheY dephosphorylation,kÿp � {kÿ0 � cÿsa[SA] � cz[Z]}[CheY-P] where [SA-P] and [SA] are molar activities of the phosphory-lated and unphosphorylated forms of MCP-CheA-CheW signal complexes ([SA]T � [SA-P] � [SA]);[CheY]T � 30 mM;1,2 [CheY-P] is molar activity ofphospho-CheY � [CheY]T - [CheY]; [Z] �molaractivity of CheZ � 10-20 mM (see Jasuja et al.3).

k0 is the rate for CheY phosphorylation catalyzedby other sources (e.g. small-molecule phosphatedonors, soluble CheA). kÿ0 is the rate for CheY-Pautophosphatase activity k1 � csa[SA-P] � rate forSA-catalyzed CheY phosphorylation kÿ1 � cÿsa

[SA-P] � rate for SA-catalyzed CheY dephosphory-lation. [SA-P]/[SA] will be decreased, in addition,by CheA phosphotransfer to CheB. cz is the rateconstant for CheZ-catalyzed CheY-P dephosphory-lation.

The expected change in motor rotation bias,�mbexptd, may be computed from the predictedstimulus-induced change in CheY-P level, usingthe general form of the equation employed by Bray& Bourret:4

�mbexptd � mbpost ÿmb

� fa��Chey-P��H=f�a��CheY-P��H�

� ��CheY-P�post�Hgg ÿ �a=�a� 1��

Conversely, for observed response �mb, [CheY-P]post , may be calculated from the equation:

�CheY-P�post��a��CheY-P��H��1ÿmbpost��=�mbpost�1=H

We assume that CheY-P is the sole signal whoselevels are modulated by attractant stimuli.�[CheY-P], hence �mbexptd may be computed as afunction of stimulus strength (�Rocc) as follows:

��CheY-P� � �CheY-P� ÿ �CheY-P�post

� f�kp=�kp � kÿp�� ÿ �k0p=�k0p � k0ÿp��g�CheY�Twhere kp

0 and kÿp0, the post-stimulus CheY phos-

phorylation and dephosphorylation rates respect-ively, are derived below as a function of �Rocc fordifferent signal ampli®cation models.

Stoichiometric inactivation.5

k0p � k0 � k1�1ÿ�Rocc�

� k0 � �KD=��Asp� � KD��k0ÿp � kÿp

Reverse phosphotransfer from CheY-P to CheA(kÿ1) was ignored, since this makes a small contri-bution to kÿp when CheZ is present.

Conformational spread within a MCP cluster.6

k0p � k0 � k1�1ÿ�Rinf�k0ÿp � kÿp

Here, the ratio of (ligand occupied/total) SAcomplexes, Rocc � na/N; while that of (infected/total) SA complexes, Rinf � 1 ÿ N�k � 1 ÿ nA{1 ÿ (ni/(N ÿ k � 1))}. N, nA and ni are independent vari-ables where N � SA complexes per cluster;


nA � SA complexes occupied by ligand per cluster;ni � SA complexes infected per occupied complex.

Ligand induced CheY-CheA transfer

Energy derived from ligand binding(�G � ÿ RTln(KD); 10 kcal/mol for KD � 1 mM)decreases k1 and increases kÿ1 respectively by [exp-(�G/2RT)] (Figure 7, model II). Therefore:

k0p � k0 � k1�1ÿ �1� exp -��G=2RT��Rocc:k0ÿp

� kÿ0 � kÿ1�1ÿ �1- exp��G=2RT��Rocc:

Accelerated CheZ catalyzed dephosphorylation

CheZ activity of SA-associated CheZ is transi-ently increased (cz

0 > cz) during the excitationphase (Figure 7, model III). As before,kp0 � k0 � k1(1 ÿ �Rocc). In addition:

k0ÿp�kÿ0�kÿ1�SA��f�cz��CheZ�ÿ�ab�SA�T��Rocc��

� c0zab�SA�T��Rocc�gcz0, a and b are independent variables; where

a � [SA]T/[CheZ]. b � CheZ copies bound perMCP-CheA-CheW complex.

References

1. Alon, U., Camarena, L., Surette, M. G.,Aguera y Arcs, B., Liu, Y., Leibler, S. & Stock, J. B.(1998). Response regulator output in bacterial che-motaxis. EMBO. J. 17, 4238-4248.

2. Scharf, B. E., Fahrner, K. A., Turner, L. &Berg, H.C. (1998). Control of direction of ¯agellarrotation in bacterial chemotaxis. Proc Natl. Acad.Sci.USA, 95, 201-206.

3. Jasuja, R., Keyoung, J., Reid, G. P., Trentham,D. R. & Khan, S. (1999a). Chemotactic responses ofEscherichia coli to small jumps of photoreleasedL-aspartate. Biophys. J. 76, 1706-1719.

4. Bray, D., & Bourret. R. B. (1995). Computeranalysis of the binding reactions leading to a trans-membrane receptor-linked multi-protein complexinvolved in bacterial chemotaxis. Mol. Biol. Cell. 6,1367-1380.

5. Jasuja, R., Lin, Y., Trentham, D. R. & Khan,S. (1999b). Response tuning in bacterial chemo-taxis. Proc. Natl. Acad. Sci. USA. 96, 11346-11351.

6. Bray, D., Levin, M. D. & Morton-Firth, C. J.(1998) Receptor clustering as a mechanism to con-trol sensitivity. Nature, 393, 85-88.

Edited by I. B. Holland

(Received 2 October 2000; received in revised form 11 December 2000; accepted 14 December 2000)

Determinants of chemotactic signal amplification in Escherichia coli

Documents

Transcript of Determinants of chemotactic signal amplification in Escherichia coli