- Email: [email protected]
Genetics of chemotactic behavior in bacteria
Cell, Vol . 4, 183-188, March 1975, Copyright©1975 by MIT
Genetics of Chemotactic Behavior in Bacteria
John S . Parkinson Department of Biology University of Utah Salt Lake City, Utah 84112
Many kinds of bacteria exhibit a simple behavior called chemotaxis in which the organism moves toward or away from various chemicals . As a general rule, chemotactic bacteria are attracted to nutrient compounds and repelled by potentially harmful ones. Recently, much attention has centered on the chemotactic systems of Escherichia coli and its close relative Salmonella typhimurium because of their suitability for genetic and biochemical studies . Genetic analysis, in particular, is proving to be a valuable tool in understanding the molecular details of chemoreception and sensory transduction in these bacteria . This review of chemotaxis will concentrate on genetic studies of the E . coli-S . typhimurium system . More extensive reviews by Adler (1975), Berg (1975), and Koshland (1975) should be consulted for information about other approaches to the problem . A complete list of references can be found in those reviews ; only the ones most pertinent to genetic studies will be cited here . Swimming Behavior E . coli and Salmonella swim by means of 5-8 flagella distributed over the cell surface which form a bundle at the rear of the cell and act in concert to propel the organism through its liquid environment . Instabilities of the flagellar bundle cause the organism to wobble or tumble momentarily and bring about changes in swimming direction . In the absence of chemical stimuli, tumbling takes place at random intervals and results in nearly random reorientations of swimming direction . The normal swimming pattern thus approximates a random walk in three dimensions . Chemotaxis is mediated through biases imposed on the random walk behavior to achieve net migrations in chemical gradients . Using an automated tracking microscope to follow the movements of individual bacteria in attractant gradients, Berg and Brown (1972) found that the mean run length of bacteria moving up-gradient was greater than of bacteria swimming down-gradient . The difference in run lengths accounted for net movement toward the attractant . Since the mean run length downgradient was identical to that in the absence of a gradient, run lengths are evidently biased by suppressing tumbles . Studies of a different sort with repellents also suggest that net migration depends on tumble suppression that occurs whenever the organism moves away from a repellent (Tsang,
Review
Macnab, and Koshland, 1973) . Movement toward attractants or away from repellents therefore is enhanced relative to other swimming directions by decreasing tumble probability whenever the organism swims in the preferred direction . Chemotaxis thus involves modulation (suppression) of tumble frequency as a function of swimming direction in a gradient . Gradient Detection E . coli and Salmonella are attracted to a variety of sugars and amino acids and are repelled by noxious substances like indole or ethanol . Adler and his colleagues (Mesibov and Adler, 1972 ; Adler, Hazelbauer, and Dahl, 1973 ; Tso and Adler, 1974) have shown that gradients of such chemicals are detected by specific chemoreceptors, each of which can sense a group of closely similar compounds . The specificity of each receptor has been determined by competition experiments . For example, if two compounds are detected by the same receptor, then a large uniform concentration of either one should inhibit chemotaxis in a gradient of the other owing to saturation of that receptor . There will be no competition if the compounds are detected by different receptors . About twenty different chemoreceptors have been defined this way and some of them are listed in Table 1 . Adler (1969) also showed that the chemical itself is detected, not the beneficial or harmful effects it may have on the organism, since uptake and metabolism are neither necessary nor sufficient for chemotaxis . This suggests that the receptors are located within or near the cytoplasmic membrane, the main permeability barrier of the cell . Chemoreceptors detect chemical gradients by measuring concentration changes over time . This was first demonstrated by Macnab and Koshland (1972), who subjected bacteria to sudden changes in attractant concentration and noted the effect on swimming behavior . Increases in attractant concentration suppressed tumbling, just as movement toward attractants suppresses tumbling in spatial gradients . Tumbling is also suppressed by a decrease in repellent concentration (Tsang et al ., 1973) . These effects can persist for several minutes, depending on the size of the temporal stimulus . In spatial gradients, movement away from attractants or toward repellents does not alter tumbling behavior . However, with temporal stimulation, both attractant decreases and repellent increases can cause a brief increase in tumbling rate . Since these responses are very brief and require stimuli too large to be encountered in spatial gradients, they do not contribute to migration in spatial gradients .
Cell 184
Table 1 . Chemoreceptors in Escherichia coli Attractants
Repellents
Chemoreceptor Name
Major Compounds Detectedb
Chemoreceptor Name
Major Compounds Detectedb
Aspartate
L-Aspartate, L-Glutamate, L-Methionine
Alcohol
Iso-Propanol, Ethanol
Fructose
D-Fructose
Aromatic
Benzoate, Methylbenzoate
Galactose
D-Galactose, D-Glucose, D-Fucose
Fatty acid
Propionate, Acetate
Glucose
D-Glucose, Methyl-a-D-glucoside
Hydrogen ion
Maltose
Maltose
Hydrophobic amino acid
H+ L-Leucine, L-Isoleucine, L-Valine, L-Tryptophan, L-Phenylalanine, L-Histidine
Mannitol
D-Mannitol
Hydroxyl ion
Mannose
D-Mannose, D-Glucose
Indole
Indole, Skatole
Ribose
D-Ribose
Metal cation
Ni++, Co++
Serine
L-Serine, L-Glycine, L-Alanine
Salicylate
Salicylate
OH -
oChemoreceptors are usually named after the compound they detect most efficiently . This is not a complete list of all the known chemoreceptors. bCompounds are listed in order of decreasing efficiency (that is, increased thresholds) in eliciting a chemotactic response .
Chemoreceptors thus can modulate tumble frequency in both directions, however, the machinery for doing so operates in a highly asymmetric manner which eventually must be explained . Flagellar Rotation The flagella of E . coli seem to be rigid, helical filaments that propel bacteria by rotating like the screws of a ship . Silverman and Simon (1974) were able to demonstrate flagellar rotation by observing latex beads attached to the flagella with antibodies . In a slightly different method, monoflagellated cells (produced by various growth regimes) can be tethered to the surface of a microscope slide with antibodies directed against the flagellar filament . The antibody-coated filament sticks to the glass slide, causing the cell body to rotate around the base of the anchored flagellum . Temporal stimulation of tethered wild type cells and experiments with tumbling and nontumbling mutants (see below) indicate that swimming corresponds to counterclockwise rotation as viewed from above the tethered cell and tumbling to clockwise rotation (Larsen et al ., 1974) . Since tethered wild type bacteria exhibit frequent reversals in the direction of rotation, tumbling is probably produced by a "tumble generator" that is able to initiate reversals of rotation independent of actual swimming movements . The Stimulus-Response Sequence Our current view of chemotaxis can be summarized as follows. Changes in chemical concentration (stimuli) are detected by chemoreceptors, which then modify tumbling rate, presumably by controlling the tumble generator, to elicit the appropriate swimming pattern (response) . Experiments using temporal stimulation show that the tumble genera-
for can be suppressed or excited by appropriate stimuli . When conflicting stimuli are presented, the response depends on the relative strengths of the two stimuli (Tsang et al ., 1973 ; Adler and Tso, 1974) . This means that signals from different receptors can be integrated either by the tumble generator or by transmission components linking the receptors to the tumble generator. Moreover, receptor signals are transient, since bacteria recover from temporal stimulation in a short time . How do chemoreceptors control the tumble generator? What is the nature of chemoreceptor signals ; how many kinds are there, and how are they transmitted? How are conflicting signals integrated? How does the tumble generator control the direction of flagellar rotation? These and many other questions have not yet been answered, although the genetic studies discussed below have begun to supply tantalizing clues about the workings of the chemotaxis machinery . Measuring Chemotaxis and Isolating Mutants The chemotactic potential of a strain can be evaluated by temporal stimulation of tethered or freely swimming cells ; however, mutant selections have depended on methods for actually measuring chemotactic migration . One method, based on the pioneering work of Pfeffer in the late 1800s, has been developed into a quantitative chemotaxis assay (Adler, 1973) . In the Pfeffer assay, an attractant-filled capillary tube is inserted into a suspension of motile bacteria . As attractant diffuses from the capillary, chemotactic bacteria are attracted to the mouth of the capillary tube and begin to accumulate inside . After a fixed time, bacteria inside the capillary are counted, usually by colony assay, to determine the chemotactic response . Repellent
Genetics of Bacterial Chemotaxis 185
taxis can be measured by mixing repellent with the bacteria and determining the number of bacteria that swim into a buffer-filled capillary for refuge . Since only a small proportion of the chemotactic bacteria in a Pfeffer assay ever enter the capillary, this method has not been used to select chemotaxis-deficient mutants . It can be used, however, to select revertants and mutants that are chemotactic under conditions where the parent is not, simply by selecting cells that end up in the capillary tube . Another method for estimating chemotactic ability is easier but less quantitative than the Pfeffer assay, and depends on the ability of bacteria to swim and to establish attractant gradients in semisolid agar "swarm" plates (Adler, 1966) which contain one or more attractant compounds as nutrient sources . Bacteria establish steep attractant gradients as they grow and consume these nutrients . Chemotactic bacteria migrate outward from the site of inoculation (the origin) because they follow the gradients that they have made . Nonchemotactic and nonmotile bacteria remain at the origin and therefore can be distinguished from wild type by their colony morphology . Several techniques for obtaining mutants are based on swarm plate methods . Nonchemotactic bacteria have been obtained by repeatedly cycling cells that remain at the origin (Armstrong, Adler, and Dahl, 1967) . Another method involves screening for mutant colonies by their morphology in swarm agar (Ordal and Adler, 1974a ; J . S . Parkinson, submitted for publication) . Since each colony can be initiated from a single bacterium, mutant bacteria give rise to mutant colonies that can be detected by their failure to produce a swarm . Since in these methods bacteria must establish gradients for themselves, nonswarming mutants can include ones with metabolic defects as well as those with chemotaxis defects . Swarm plate methods cannot be used to obtain repellent mutants or mutants with defects in essential functions . A third class of techniques for measuring chemotaxis employs preformed gradients . Tso and Adler (1974) have described several methods for measuring repellent taxis using diffusion gradients in agar or liquid . Several of these methods were used to select mutants defective in repellent taxis . Dahiquist, Lovely, and Koshland (1972) developed a novel method for measuring chemotaxis in mechanically preformed gradients in which migration of the bacterial population could be followed by laser scattering . Recently, Aswad and Koshland (submitted for publication) and Miller and Dahlquist (personal communication) have employed similarly formed gradients to isolate generally nonchemotactic mutants . These methods eliminate several drawbacks in swarmplate selections and should prove
useful in isolating conditional chemotaxis mutants that may involve essential functions . Chemoreceptor Mutants Mutants with a chemoreceptor defect are unable to respond to compounds detected by a single receptor species, but responses mediated by other receptors are unimpaired . Among the attractants, mutants of the galactose (Hazelbauer, Mesibov, and Adler, 1969), maltose (Hazelbauer, 1975), ribose (Aksamit and Koshland, 1974), and mannose and glucose (Adler and Epstein, 1974) receptors have been studied . Among repellents, apparent mutants of the leucine, indole and salicylate receptors, are known but have not been studied in detail (Tso and Adler, 1974) . The response patterns of these mutants are consistent with single receptor defects and have served to confirm receptor classifications based on competition studies . Attempts to isolate aspartate or serine receptor mutants have been unsuccessful, perhaps because there may be several distinct receptors contributing to these responses . Most of the sugar receptor mutants are defective in transport of their respective substrates . In fact, many of these mutants were first isolated as uptake mutants and shown subsequently to have chemotaxis defects . An uptake defect per se is not responsible, however, for loss of chemotaxis, since other types of transport mutants are still chemotactic (Adler, 1969) . The receptor mutants appear to lack a protein that functions as the primary recognition component for both transport and chemotaxis . Each of these proteins can bind to its respective attractant molecules with affinities that parallel the effectiveness of those compounds in eliciting a chemotactic response . For example, the galactose binding protein has greater affinity for galactose than fucose, and the response threshold for chemotaxis occurs at lower concentrations of galactose than fucose (Hazelbauer and Adler, 1971) . In fact, chemotaxis to galactose (and presumably other compounds as well) only occurs over a limited concentration range centered about the Km for binding of galactose by the galactose binding protein (Mesibov, Ordal, and Adler, 1973) . Using an enzyme system to generate a continuous temporal attractant gradient, Brown and Berg (1974) showed that the chemotactic response probably depends on the rate of change in the proportion of bound receptors . The binding protein components of these sugar receptors appear to be associated with the cytoplasmic membrane . The glucose and mannose binding proteins are part of the phosphotransferase sugar transport system which is known to reside in the cytoplasmic membrane (Adler and Epstein, 1974) . The galactose, maltose, and ribose binding
Cell 1 86
proteins can be released from the cell by osmotic shock, and cells so treated are no longer attracted to these compounds (Hazelbauer and Adler, 1971) . These shock-releasable proteins are thought to reside on the surface of the cytoplasmic membrane or perhaps in the "periplasmic space" between the inner and outer membranes . Chemoreceptors contain other components in addition to a binding protein . In an extensive study of galactose taxis mutants, Ordal and Adler (1974a,b) discovered two or possibly three genes among 45 independent isolates . Their study suggested that additional transport components were involved in chemotaxis either directly or through interaction with the binding protein to control its conformation for chemoreception . A similar situation is known for the glucose and mannose receptors ; other components of the phosphotransferase system which interact with the binding proteins are needed for chemotaxis (Adler and Epstein, 1974) . Whether any of these components participate directly in converting concentration data into signals that modulate the tumble generator is not known . Flagellar Mutants The bacterial flagellum consists of a long, helical filament that is connected to a basal body by a short hook (Figure 1) . The basal body contains four rings or discs attached to a central rod . The two outer rings are associated with the outer membrane and peptidoglycan layers of the cell wall, the innermost ring is associated with the cytoplasmic membrane
Figure 1 . A Model of the Basal Portion of the E . coli Flagellum Based on Electron Microscopic Studies (DePamphilis and Adler, 1971 a, b) The dimensions are given in nanometers . (Figure courtesy of J . Adler .)
(DePamphilis and Adler, 1971 a,b) . The machinery for generating rotation and reversals is probably located in the basal body . In E . coli, at least 18 genes are required for assembly and function of the flagellum (Hilmen, Silverman, and Simon, 1974) . One gene (hag) is the structural gene for the filament protein, flagellin ; another (flaE) controls the length of the hook ; and a third (flal) somehow regulates synthesis of the entire flagellum . Of the remainder, 14 genes have a nonflagellated mutant phenotype, while one (mot) produces paralyzed flagella that appear morphologically normal . At least five basal body components have been detected by SDS-acrylamide gel electrophoresis ; however, none of these has yet been correlated with a specific gene product . Communication Mutants The chemoreceptors and flagella are linked by a communication network . For convenience, any mutant that is motile but defective in more than one response will be called a communication mutant . Such mutants define components of the chemotaxis machinery that are needed by more than one receptor . Two types of communication defects have been identified . Several chemotaxis mutants are known that lack some but not all responses . These mutants have provided information about branch points in the communication pathway . One class of mutants has essentially normal taxis to all compounds except ribose and galactose (trg mutants) . The trg mutations do not map near any known chemotaxis gene and therefore define a new chemotaxis locus whose product is either a component of both the galactose and ribose receptors or a common component of the communication system that transmits ribose and galactose signals to the tumble generator (Ordal and Adler, 1974b) . Another class of mutants, called tsr, have lost serine and several repellent responses while retaining all others (Tso and Adler, 1974) . Based on additional evidence discussed below, the tsr product appears to modulate the tumble generator in response to signals from the serine receptor and some repellent receptors . The second type of communication mutant is generally nonchemotactic, and although motile, is unable to respond to any stimuli (Armstrong et al ., 1967) . These mutants have been studied extensively in both E . coli and S . typhimurium, and thus far every mutant of this type has had an aberrant tumbling pattern which suggests that many of the common components of the communication system are involved in generating or controlling tumbles . In E . coli, more than 200 independent mutants have been characterized (Armstrong and Adler, 1969 ; Parkinson, submitted for publication) ; in S . typhimurium,
Genetics of Bacterial Ghemotaxis 187
57 have been examined (Aswad and Koshland, submitted for publication) . The E . coli mutants have been analyzed by complementation tests using F' elements to construct stable partial diploids, and four genes have been identified . The properties of these four genes, cheA, cheB, cheC, cheD, are summarized below (Parkinson, 1974) . -All cheA mutants fail to tumble while swimming and rotate only counter-clockwise when tethered . Since they cannot be induced to tumble by temporal stimulation, cheA mutants probably lack an essential component of the tumble generator . -Mutants of the cheB gene have two distinct phenotypes . Some cheB mutants cannot tumble, whereas others tumble incessantly . The nontumbling mutants represent the greater loss of function, since they are more frequent than and recessive to the incessantly tumbling isolates . Since nontumbling cheB mutants do not respond to temporal stimulation, the cheB product must also be an essential part of the tumble generator . However, the tumbling cheB mutants suggest that the cheB product also plays a role in regulating tumble frequency . Tumbling cheB mutants respond to temporal stimulation, but their response thresholds are considerably greater than wild type and usually vary for different stimuli . This suggests that chemoreceptor signals are unable to control properly the tumble generator of these mutants . Since the mutants tumble uncontrollably, receptor signals are believed to be inhibitory rather than excitatory, and it may be that some basal level of inhibition in unstimulated cells accounts for the characteristic tumble frequency of wild type bacteria . In tumbling cheB mutants, the tumble generator is only poorly coupled to its controlling signals and therefore "runs wild" like an engine with a defective governor . -The cheC product is probably part of the flagellum (Silverman and Simon, 1973) . Most cheC mutants, including nonsense mutations and deletions, are nonflagellated . However, a few cheC mutants are motile but unable to tumble, and therefore generally nonchemotactic . Since these mutants are rare and also partially dominant, they evidently retain considerable function that permits assembly of a flagellum but prevents clockwise rotation . -cheD mutants cannot tumble and are fully dominant to wild type . This implies that a suitably altered cheD product can inhibit the tumbling machinery and thereby lead to a generally nonchemotactic phenotype . The available genetic evidence indicates that cheD mutants actually represent a special type of tsr defect . For example, nonsense mutations in the cheD gene abolish dominance of cheD mutants by preventing formation of the tumbleinhibiting product . These mutations map very close to the original cheD mutation and resemble tsr mu-
tants in all respects (Parkinson, submitted for publication) . This suggests that the tsr product normally interacts with the tumble generator, functioning as a switch to control tumbling in response to serine signals and several repellent signals . In support of this explanation is evidence for interaction of the tsr product with the cheB product based on studies of tsr cheB double mutants (Parkinson, in preparation) . Information Processing during Chemotaxis A model of the chemotactic process, based on mutant studies, is presented in Figure 2 . The tumble generator is controlled by inhibitory signals whose level is regulated by the chemoreceptors . Upon detection of an increase in attractant concentration or a decrease in repellent concentration, chemoreceptors raise the level of inhibitory signals to suppress tumbling . Increased tumbling rates that can be caused by certain temporal stimulations are probably effected by a decrease in the inhibitor level . The various receptors control at least two types of signal, one that is modulated by the serine receptor and several repellent receptors and transmitted by the tsr product, and one or more additional signals from the other receptors that are not transmitted via tsr product . All signals ultimately reach the tumble generator and suppress tumbling by interacting with the cheB component . The point at which conflicting stimuli are integrated must depend on the particular stimuli . For example, galactose and ribose signals may be summed by the trg component . Other combinations of stimuli could be integrated by the tsr or cheB components or any other components that constitute the first common part of two receptor pathways . Conclusions The chemotaxis machinery of bacteria is designed to carry out processes like stimulus detection, transduction, and integration that resemble those in more complex behavioral systems. Genetic methods are useful in revealing the details of these events . Although our analysis has only begun, three main points can be made . Based on the variety of mutants now available, the system promises to be complex . More than 30 gene products are known to be required for chemotaxis. Although many of these are also needed for motility, more chemotaxis-specific functions are likely to emerge as better mutant isolation procedures are developed . The information pathways of chemotaxis can be dissected genetically in much the same way that metabolic pathways are studied with nutritional mutants . It is clear that information from all the
Cell 1 88
RECEPTORS
FLAGELLA
TUMBLE GENERATOR
aspartate cheA product
I salicylate maltose
cheC product
ribose galactose
I.
nproduct
serine indole
I
hydrogen ion
Figure 2 . A Model of Information Flow during Chemotaxis The overall pattern of data transmission from receptors to flagella converges at the tumble generator . Information from the various receptors (only a few are shown) is passed through different components . The tsr product handles signals from the serine, indole, hydrogen ion, aromatic, fatty acid, and hydrophobic amino acid receptors . Other signals may interact directly with the cheB product . Tumbling impulses produced by the cheA and cheB components are transmitted to the flagella and the cheC product.
chemoreceptors is channeled through a convergent communication network to the tumble generator and flagella . However, we do not yet know the nature of these signals, how they are modulated and transmitted, and how many kinds there might be . There is very little biochemical information about chemotaxis . Investigations of the sugar receptors show that binding proteins are involved in gradient detection . The remainder of the chemotactic process, between binding proteins and the eventual response, is poorly understood . It is hoped that genetic analysis will pave the way for future biochemical studies . References Adler, J . (1966) . Science 153, 708 . Adler, J . (1969) . Science 166, 1588 . Adler, J . (1973) . J . Gen . Microbiol . 74, 77 . Adler, J . (1975) . Ann . Rev . Biochem ., in press . Adler, J ., and Epstein, W . (1974) . Proc . Nat . Acad . Sci . USA 71, 2895 . Adler, J ., and Tso, W.-W. (1974) . Science 184, 1292 . Adler, J ., Hazelbauer, G . L ., and Dahl, M . M . (1973) . J . Bacteriol . 115, 824 . Akasmit, R ., and Koshland, D . E ., Jr . (1974) . Biochem . 13, 4473 . Armstrong, J . B ., and Adler, J . (1969) . Genetics 61, 61 . Armstrong, J . B ., Adler, J ., and Dahl, M . M . (1967) . J . Bacteriol . 93, 390 . Berg, H . C . (1975) . Ann . Rev . Biophys . Bioeng ., in press . Berg, H . C ., and Brown, D . A. (1972) . Nature 239, 500 .
Brown, D . A ., and Berg, H . C . (1974) . Proc . Nat Acad . Sci. USA 71, 1388 . Dahlquist, F . W ., Lovely, P ., and Koshland, D . E ., Jr . (1972) . Nature New Biol . 236, 120 . DePamphilis, M . L ., and Adler, J . (1971a) . J . Bacteriol . 105, 384 . DePamphilis, M . L., and Adler, J . (1971b) . J . Bacteriol . 105, 396 . Hazelbauer, G . L. (1975) . J . Bacteriol ., in press . Hazelbauer, G . L., and Adler, J . (1971) . Nature New Biol . 230, 101 . Hazelbauer, G . L ., Mesibov, R . E ., and Adler, J . (1969). Proc . Nat . Acad. Sci. USA 64, 1300 . Hilmen, M ., Silverman, M ., and Simon, M . (1974) . J . Supramolecular Structure 2, in press . Koshland, D . E ., Jr. (1975). Mosbach Symposium, in press. Larsen, S . H ., Reader, R . W ., Kort, E . N ., Tso, W .-W ., and Adler, J . (1974) . Nature 249, 74 . Macnab, R ., and Koshland, D . E ., Jr. (1972) . Proc . Nat . Acad . Sci . USA 69, 2509 . Mesibov, R ., and Adler, J . (1972) . J . Bacteriol . 112, 315 . Mesibov, R ., Ordal, G . W., and Adler, J . (1973). J . Gen . Physiol . 62, 203 . Ordal, G . W ., and Adler, J . (1974a) . J . Bacteriol . 117, 509 . Ordal, G . W., and Adler, J. (1974b) . J . Bacteriol . 117, 517 . Parkinson, J . S . (1974) . Nature 252, 317 . Silverman, M ., and Simon, M . (1973) . J . Bacteriol. 116, 114 . Silverman, M ., and Simon, M . (1974) . Nature 249, 73 . Tsang, N ., Macnab, R ., and Koshland, D . E ., Jr . (1973) . Science 181, 60 . Tso, W .-W ., and Adler, J . (1974) . J . Bacteriol . 118, 560 .
Genetics of Chemotactic Behavior in Bacteria
John S . Parkinson Department of Biology University of Utah Salt Lake City, Utah 84112
Many kinds of bacteria exhibit a simple behavior called chemotaxis in which the organism moves toward or away from various chemicals . As a general rule, chemotactic bacteria are attracted to nutrient compounds and repelled by potentially harmful ones. Recently, much attention has centered on the chemotactic systems of Escherichia coli and its close relative Salmonella typhimurium because of their suitability for genetic and biochemical studies . Genetic analysis, in particular, is proving to be a valuable tool in understanding the molecular details of chemoreception and sensory transduction in these bacteria . This review of chemotaxis will concentrate on genetic studies of the E . coli-S . typhimurium system . More extensive reviews by Adler (1975), Berg (1975), and Koshland (1975) should be consulted for information about other approaches to the problem . A complete list of references can be found in those reviews ; only the ones most pertinent to genetic studies will be cited here . Swimming Behavior E . coli and Salmonella swim by means of 5-8 flagella distributed over the cell surface which form a bundle at the rear of the cell and act in concert to propel the organism through its liquid environment . Instabilities of the flagellar bundle cause the organism to wobble or tumble momentarily and bring about changes in swimming direction . In the absence of chemical stimuli, tumbling takes place at random intervals and results in nearly random reorientations of swimming direction . The normal swimming pattern thus approximates a random walk in three dimensions . Chemotaxis is mediated through biases imposed on the random walk behavior to achieve net migrations in chemical gradients . Using an automated tracking microscope to follow the movements of individual bacteria in attractant gradients, Berg and Brown (1972) found that the mean run length of bacteria moving up-gradient was greater than of bacteria swimming down-gradient . The difference in run lengths accounted for net movement toward the attractant . Since the mean run length downgradient was identical to that in the absence of a gradient, run lengths are evidently biased by suppressing tumbles . Studies of a different sort with repellents also suggest that net migration depends on tumble suppression that occurs whenever the organism moves away from a repellent (Tsang,
Review
Macnab, and Koshland, 1973) . Movement toward attractants or away from repellents therefore is enhanced relative to other swimming directions by decreasing tumble probability whenever the organism swims in the preferred direction . Chemotaxis thus involves modulation (suppression) of tumble frequency as a function of swimming direction in a gradient . Gradient Detection E . coli and Salmonella are attracted to a variety of sugars and amino acids and are repelled by noxious substances like indole or ethanol . Adler and his colleagues (Mesibov and Adler, 1972 ; Adler, Hazelbauer, and Dahl, 1973 ; Tso and Adler, 1974) have shown that gradients of such chemicals are detected by specific chemoreceptors, each of which can sense a group of closely similar compounds . The specificity of each receptor has been determined by competition experiments . For example, if two compounds are detected by the same receptor, then a large uniform concentration of either one should inhibit chemotaxis in a gradient of the other owing to saturation of that receptor . There will be no competition if the compounds are detected by different receptors . About twenty different chemoreceptors have been defined this way and some of them are listed in Table 1 . Adler (1969) also showed that the chemical itself is detected, not the beneficial or harmful effects it may have on the organism, since uptake and metabolism are neither necessary nor sufficient for chemotaxis . This suggests that the receptors are located within or near the cytoplasmic membrane, the main permeability barrier of the cell . Chemoreceptors detect chemical gradients by measuring concentration changes over time . This was first demonstrated by Macnab and Koshland (1972), who subjected bacteria to sudden changes in attractant concentration and noted the effect on swimming behavior . Increases in attractant concentration suppressed tumbling, just as movement toward attractants suppresses tumbling in spatial gradients . Tumbling is also suppressed by a decrease in repellent concentration (Tsang et al ., 1973) . These effects can persist for several minutes, depending on the size of the temporal stimulus . In spatial gradients, movement away from attractants or toward repellents does not alter tumbling behavior . However, with temporal stimulation, both attractant decreases and repellent increases can cause a brief increase in tumbling rate . Since these responses are very brief and require stimuli too large to be encountered in spatial gradients, they do not contribute to migration in spatial gradients .
Cell 184
Table 1 . Chemoreceptors in Escherichia coli Attractants
Repellents
Chemoreceptor Name
Major Compounds Detectedb
Chemoreceptor Name
Major Compounds Detectedb
Aspartate
L-Aspartate, L-Glutamate, L-Methionine
Alcohol
Iso-Propanol, Ethanol
Fructose
D-Fructose
Aromatic
Benzoate, Methylbenzoate
Galactose
D-Galactose, D-Glucose, D-Fucose
Fatty acid
Propionate, Acetate
Glucose
D-Glucose, Methyl-a-D-glucoside
Hydrogen ion
Maltose
Maltose
Hydrophobic amino acid
H+ L-Leucine, L-Isoleucine, L-Valine, L-Tryptophan, L-Phenylalanine, L-Histidine
Mannitol
D-Mannitol
Hydroxyl ion
Mannose
D-Mannose, D-Glucose
Indole
Indole, Skatole
Ribose
D-Ribose
Metal cation
Ni++, Co++
Serine
L-Serine, L-Glycine, L-Alanine
Salicylate
Salicylate
OH -
oChemoreceptors are usually named after the compound they detect most efficiently . This is not a complete list of all the known chemoreceptors. bCompounds are listed in order of decreasing efficiency (that is, increased thresholds) in eliciting a chemotactic response .
Chemoreceptors thus can modulate tumble frequency in both directions, however, the machinery for doing so operates in a highly asymmetric manner which eventually must be explained . Flagellar Rotation The flagella of E . coli seem to be rigid, helical filaments that propel bacteria by rotating like the screws of a ship . Silverman and Simon (1974) were able to demonstrate flagellar rotation by observing latex beads attached to the flagella with antibodies . In a slightly different method, monoflagellated cells (produced by various growth regimes) can be tethered to the surface of a microscope slide with antibodies directed against the flagellar filament . The antibody-coated filament sticks to the glass slide, causing the cell body to rotate around the base of the anchored flagellum . Temporal stimulation of tethered wild type cells and experiments with tumbling and nontumbling mutants (see below) indicate that swimming corresponds to counterclockwise rotation as viewed from above the tethered cell and tumbling to clockwise rotation (Larsen et al ., 1974) . Since tethered wild type bacteria exhibit frequent reversals in the direction of rotation, tumbling is probably produced by a "tumble generator" that is able to initiate reversals of rotation independent of actual swimming movements . The Stimulus-Response Sequence Our current view of chemotaxis can be summarized as follows. Changes in chemical concentration (stimuli) are detected by chemoreceptors, which then modify tumbling rate, presumably by controlling the tumble generator, to elicit the appropriate swimming pattern (response) . Experiments using temporal stimulation show that the tumble genera-
for can be suppressed or excited by appropriate stimuli . When conflicting stimuli are presented, the response depends on the relative strengths of the two stimuli (Tsang et al ., 1973 ; Adler and Tso, 1974) . This means that signals from different receptors can be integrated either by the tumble generator or by transmission components linking the receptors to the tumble generator. Moreover, receptor signals are transient, since bacteria recover from temporal stimulation in a short time . How do chemoreceptors control the tumble generator? What is the nature of chemoreceptor signals ; how many kinds are there, and how are they transmitted? How are conflicting signals integrated? How does the tumble generator control the direction of flagellar rotation? These and many other questions have not yet been answered, although the genetic studies discussed below have begun to supply tantalizing clues about the workings of the chemotaxis machinery . Measuring Chemotaxis and Isolating Mutants The chemotactic potential of a strain can be evaluated by temporal stimulation of tethered or freely swimming cells ; however, mutant selections have depended on methods for actually measuring chemotactic migration . One method, based on the pioneering work of Pfeffer in the late 1800s, has been developed into a quantitative chemotaxis assay (Adler, 1973) . In the Pfeffer assay, an attractant-filled capillary tube is inserted into a suspension of motile bacteria . As attractant diffuses from the capillary, chemotactic bacteria are attracted to the mouth of the capillary tube and begin to accumulate inside . After a fixed time, bacteria inside the capillary are counted, usually by colony assay, to determine the chemotactic response . Repellent
Genetics of Bacterial Chemotaxis 185
taxis can be measured by mixing repellent with the bacteria and determining the number of bacteria that swim into a buffer-filled capillary for refuge . Since only a small proportion of the chemotactic bacteria in a Pfeffer assay ever enter the capillary, this method has not been used to select chemotaxis-deficient mutants . It can be used, however, to select revertants and mutants that are chemotactic under conditions where the parent is not, simply by selecting cells that end up in the capillary tube . Another method for estimating chemotactic ability is easier but less quantitative than the Pfeffer assay, and depends on the ability of bacteria to swim and to establish attractant gradients in semisolid agar "swarm" plates (Adler, 1966) which contain one or more attractant compounds as nutrient sources . Bacteria establish steep attractant gradients as they grow and consume these nutrients . Chemotactic bacteria migrate outward from the site of inoculation (the origin) because they follow the gradients that they have made . Nonchemotactic and nonmotile bacteria remain at the origin and therefore can be distinguished from wild type by their colony morphology . Several techniques for obtaining mutants are based on swarm plate methods . Nonchemotactic bacteria have been obtained by repeatedly cycling cells that remain at the origin (Armstrong, Adler, and Dahl, 1967) . Another method involves screening for mutant colonies by their morphology in swarm agar (Ordal and Adler, 1974a ; J . S . Parkinson, submitted for publication) . Since each colony can be initiated from a single bacterium, mutant bacteria give rise to mutant colonies that can be detected by their failure to produce a swarm . Since in these methods bacteria must establish gradients for themselves, nonswarming mutants can include ones with metabolic defects as well as those with chemotaxis defects . Swarm plate methods cannot be used to obtain repellent mutants or mutants with defects in essential functions . A third class of techniques for measuring chemotaxis employs preformed gradients . Tso and Adler (1974) have described several methods for measuring repellent taxis using diffusion gradients in agar or liquid . Several of these methods were used to select mutants defective in repellent taxis . Dahiquist, Lovely, and Koshland (1972) developed a novel method for measuring chemotaxis in mechanically preformed gradients in which migration of the bacterial population could be followed by laser scattering . Recently, Aswad and Koshland (submitted for publication) and Miller and Dahlquist (personal communication) have employed similarly formed gradients to isolate generally nonchemotactic mutants . These methods eliminate several drawbacks in swarmplate selections and should prove
useful in isolating conditional chemotaxis mutants that may involve essential functions . Chemoreceptor Mutants Mutants with a chemoreceptor defect are unable to respond to compounds detected by a single receptor species, but responses mediated by other receptors are unimpaired . Among the attractants, mutants of the galactose (Hazelbauer, Mesibov, and Adler, 1969), maltose (Hazelbauer, 1975), ribose (Aksamit and Koshland, 1974), and mannose and glucose (Adler and Epstein, 1974) receptors have been studied . Among repellents, apparent mutants of the leucine, indole and salicylate receptors, are known but have not been studied in detail (Tso and Adler, 1974) . The response patterns of these mutants are consistent with single receptor defects and have served to confirm receptor classifications based on competition studies . Attempts to isolate aspartate or serine receptor mutants have been unsuccessful, perhaps because there may be several distinct receptors contributing to these responses . Most of the sugar receptor mutants are defective in transport of their respective substrates . In fact, many of these mutants were first isolated as uptake mutants and shown subsequently to have chemotaxis defects . An uptake defect per se is not responsible, however, for loss of chemotaxis, since other types of transport mutants are still chemotactic (Adler, 1969) . The receptor mutants appear to lack a protein that functions as the primary recognition component for both transport and chemotaxis . Each of these proteins can bind to its respective attractant molecules with affinities that parallel the effectiveness of those compounds in eliciting a chemotactic response . For example, the galactose binding protein has greater affinity for galactose than fucose, and the response threshold for chemotaxis occurs at lower concentrations of galactose than fucose (Hazelbauer and Adler, 1971) . In fact, chemotaxis to galactose (and presumably other compounds as well) only occurs over a limited concentration range centered about the Km for binding of galactose by the galactose binding protein (Mesibov, Ordal, and Adler, 1973) . Using an enzyme system to generate a continuous temporal attractant gradient, Brown and Berg (1974) showed that the chemotactic response probably depends on the rate of change in the proportion of bound receptors . The binding protein components of these sugar receptors appear to be associated with the cytoplasmic membrane . The glucose and mannose binding proteins are part of the phosphotransferase sugar transport system which is known to reside in the cytoplasmic membrane (Adler and Epstein, 1974) . The galactose, maltose, and ribose binding
Cell 1 86
proteins can be released from the cell by osmotic shock, and cells so treated are no longer attracted to these compounds (Hazelbauer and Adler, 1971) . These shock-releasable proteins are thought to reside on the surface of the cytoplasmic membrane or perhaps in the "periplasmic space" between the inner and outer membranes . Chemoreceptors contain other components in addition to a binding protein . In an extensive study of galactose taxis mutants, Ordal and Adler (1974a,b) discovered two or possibly three genes among 45 independent isolates . Their study suggested that additional transport components were involved in chemotaxis either directly or through interaction with the binding protein to control its conformation for chemoreception . A similar situation is known for the glucose and mannose receptors ; other components of the phosphotransferase system which interact with the binding proteins are needed for chemotaxis (Adler and Epstein, 1974) . Whether any of these components participate directly in converting concentration data into signals that modulate the tumble generator is not known . Flagellar Mutants The bacterial flagellum consists of a long, helical filament that is connected to a basal body by a short hook (Figure 1) . The basal body contains four rings or discs attached to a central rod . The two outer rings are associated with the outer membrane and peptidoglycan layers of the cell wall, the innermost ring is associated with the cytoplasmic membrane
Figure 1 . A Model of the Basal Portion of the E . coli Flagellum Based on Electron Microscopic Studies (DePamphilis and Adler, 1971 a, b) The dimensions are given in nanometers . (Figure courtesy of J . Adler .)
(DePamphilis and Adler, 1971 a,b) . The machinery for generating rotation and reversals is probably located in the basal body . In E . coli, at least 18 genes are required for assembly and function of the flagellum (Hilmen, Silverman, and Simon, 1974) . One gene (hag) is the structural gene for the filament protein, flagellin ; another (flaE) controls the length of the hook ; and a third (flal) somehow regulates synthesis of the entire flagellum . Of the remainder, 14 genes have a nonflagellated mutant phenotype, while one (mot) produces paralyzed flagella that appear morphologically normal . At least five basal body components have been detected by SDS-acrylamide gel electrophoresis ; however, none of these has yet been correlated with a specific gene product . Communication Mutants The chemoreceptors and flagella are linked by a communication network . For convenience, any mutant that is motile but defective in more than one response will be called a communication mutant . Such mutants define components of the chemotaxis machinery that are needed by more than one receptor . Two types of communication defects have been identified . Several chemotaxis mutants are known that lack some but not all responses . These mutants have provided information about branch points in the communication pathway . One class of mutants has essentially normal taxis to all compounds except ribose and galactose (trg mutants) . The trg mutations do not map near any known chemotaxis gene and therefore define a new chemotaxis locus whose product is either a component of both the galactose and ribose receptors or a common component of the communication system that transmits ribose and galactose signals to the tumble generator (Ordal and Adler, 1974b) . Another class of mutants, called tsr, have lost serine and several repellent responses while retaining all others (Tso and Adler, 1974) . Based on additional evidence discussed below, the tsr product appears to modulate the tumble generator in response to signals from the serine receptor and some repellent receptors . The second type of communication mutant is generally nonchemotactic, and although motile, is unable to respond to any stimuli (Armstrong et al ., 1967) . These mutants have been studied extensively in both E . coli and S . typhimurium, and thus far every mutant of this type has had an aberrant tumbling pattern which suggests that many of the common components of the communication system are involved in generating or controlling tumbles . In E . coli, more than 200 independent mutants have been characterized (Armstrong and Adler, 1969 ; Parkinson, submitted for publication) ; in S . typhimurium,
Genetics of Bacterial Ghemotaxis 187
57 have been examined (Aswad and Koshland, submitted for publication) . The E . coli mutants have been analyzed by complementation tests using F' elements to construct stable partial diploids, and four genes have been identified . The properties of these four genes, cheA, cheB, cheC, cheD, are summarized below (Parkinson, 1974) . -All cheA mutants fail to tumble while swimming and rotate only counter-clockwise when tethered . Since they cannot be induced to tumble by temporal stimulation, cheA mutants probably lack an essential component of the tumble generator . -Mutants of the cheB gene have two distinct phenotypes . Some cheB mutants cannot tumble, whereas others tumble incessantly . The nontumbling mutants represent the greater loss of function, since they are more frequent than and recessive to the incessantly tumbling isolates . Since nontumbling cheB mutants do not respond to temporal stimulation, the cheB product must also be an essential part of the tumble generator . However, the tumbling cheB mutants suggest that the cheB product also plays a role in regulating tumble frequency . Tumbling cheB mutants respond to temporal stimulation, but their response thresholds are considerably greater than wild type and usually vary for different stimuli . This suggests that chemoreceptor signals are unable to control properly the tumble generator of these mutants . Since the mutants tumble uncontrollably, receptor signals are believed to be inhibitory rather than excitatory, and it may be that some basal level of inhibition in unstimulated cells accounts for the characteristic tumble frequency of wild type bacteria . In tumbling cheB mutants, the tumble generator is only poorly coupled to its controlling signals and therefore "runs wild" like an engine with a defective governor . -The cheC product is probably part of the flagellum (Silverman and Simon, 1973) . Most cheC mutants, including nonsense mutations and deletions, are nonflagellated . However, a few cheC mutants are motile but unable to tumble, and therefore generally nonchemotactic . Since these mutants are rare and also partially dominant, they evidently retain considerable function that permits assembly of a flagellum but prevents clockwise rotation . -cheD mutants cannot tumble and are fully dominant to wild type . This implies that a suitably altered cheD product can inhibit the tumbling machinery and thereby lead to a generally nonchemotactic phenotype . The available genetic evidence indicates that cheD mutants actually represent a special type of tsr defect . For example, nonsense mutations in the cheD gene abolish dominance of cheD mutants by preventing formation of the tumbleinhibiting product . These mutations map very close to the original cheD mutation and resemble tsr mu-
tants in all respects (Parkinson, submitted for publication) . This suggests that the tsr product normally interacts with the tumble generator, functioning as a switch to control tumbling in response to serine signals and several repellent signals . In support of this explanation is evidence for interaction of the tsr product with the cheB product based on studies of tsr cheB double mutants (Parkinson, in preparation) . Information Processing during Chemotaxis A model of the chemotactic process, based on mutant studies, is presented in Figure 2 . The tumble generator is controlled by inhibitory signals whose level is regulated by the chemoreceptors . Upon detection of an increase in attractant concentration or a decrease in repellent concentration, chemoreceptors raise the level of inhibitory signals to suppress tumbling . Increased tumbling rates that can be caused by certain temporal stimulations are probably effected by a decrease in the inhibitor level . The various receptors control at least two types of signal, one that is modulated by the serine receptor and several repellent receptors and transmitted by the tsr product, and one or more additional signals from the other receptors that are not transmitted via tsr product . All signals ultimately reach the tumble generator and suppress tumbling by interacting with the cheB component . The point at which conflicting stimuli are integrated must depend on the particular stimuli . For example, galactose and ribose signals may be summed by the trg component . Other combinations of stimuli could be integrated by the tsr or cheB components or any other components that constitute the first common part of two receptor pathways . Conclusions The chemotaxis machinery of bacteria is designed to carry out processes like stimulus detection, transduction, and integration that resemble those in more complex behavioral systems. Genetic methods are useful in revealing the details of these events . Although our analysis has only begun, three main points can be made . Based on the variety of mutants now available, the system promises to be complex . More than 30 gene products are known to be required for chemotaxis. Although many of these are also needed for motility, more chemotaxis-specific functions are likely to emerge as better mutant isolation procedures are developed . The information pathways of chemotaxis can be dissected genetically in much the same way that metabolic pathways are studied with nutritional mutants . It is clear that information from all the
Cell 1 88
RECEPTORS
FLAGELLA
TUMBLE GENERATOR
aspartate cheA product
I salicylate maltose
cheC product
ribose galactose
I.
nproduct
serine indole
I
hydrogen ion
Figure 2 . A Model of Information Flow during Chemotaxis The overall pattern of data transmission from receptors to flagella converges at the tumble generator . Information from the various receptors (only a few are shown) is passed through different components . The tsr product handles signals from the serine, indole, hydrogen ion, aromatic, fatty acid, and hydrophobic amino acid receptors . Other signals may interact directly with the cheB product . Tumbling impulses produced by the cheA and cheB components are transmitted to the flagella and the cheC product.
chemoreceptors is channeled through a convergent communication network to the tumble generator and flagella . However, we do not yet know the nature of these signals, how they are modulated and transmitted, and how many kinds there might be . There is very little biochemical information about chemotaxis . Investigations of the sugar receptors show that binding proteins are involved in gradient detection . The remainder of the chemotactic process, between binding proteins and the eventual response, is poorly understood . It is hoped that genetic analysis will pave the way for future biochemical studies . References Adler, J . (1966) . Science 153, 708 . Adler, J . (1969) . Science 166, 1588 . Adler, J . (1973) . J . Gen . Microbiol . 74, 77 . Adler, J . (1975) . Ann . Rev . Biochem ., in press . Adler, J ., and Epstein, W . (1974) . Proc . Nat . Acad . Sci . USA 71, 2895 . Adler, J ., and Tso, W.-W. (1974) . Science 184, 1292 . Adler, J ., Hazelbauer, G . L ., and Dahl, M . M . (1973) . J . Bacteriol . 115, 824 . Akasmit, R ., and Koshland, D . E ., Jr . (1974) . Biochem . 13, 4473 . Armstrong, J . B ., and Adler, J . (1969) . Genetics 61, 61 . Armstrong, J . B ., Adler, J ., and Dahl, M . M . (1967) . J . Bacteriol . 93, 390 . Berg, H . C . (1975) . Ann . Rev . Biophys . Bioeng ., in press . Berg, H . C ., and Brown, D . A. (1972) . Nature 239, 500 .
Brown, D . A ., and Berg, H . C . (1974) . Proc . Nat Acad . Sci. USA 71, 1388 . Dahlquist, F . W ., Lovely, P ., and Koshland, D . E ., Jr . (1972) . Nature New Biol . 236, 120 . DePamphilis, M . L ., and Adler, J . (1971a) . J . Bacteriol . 105, 384 . DePamphilis, M . L., and Adler, J . (1971b) . J . Bacteriol . 105, 396 . Hazelbauer, G . L. (1975) . J . Bacteriol ., in press . Hazelbauer, G . L., and Adler, J . (1971) . Nature New Biol . 230, 101 . Hazelbauer, G . L ., Mesibov, R . E ., and Adler, J . (1969). Proc . Nat . Acad. Sci. USA 64, 1300 . Hilmen, M ., Silverman, M ., and Simon, M . (1974) . J . Supramolecular Structure 2, in press . Koshland, D . E ., Jr. (1975). Mosbach Symposium, in press. Larsen, S . H ., Reader, R . W ., Kort, E . N ., Tso, W .-W ., and Adler, J . (1974) . Nature 249, 74 . Macnab, R ., and Koshland, D . E ., Jr. (1972) . Proc . Nat . Acad . Sci . USA 69, 2509 . Mesibov, R ., and Adler, J . (1972) . J . Bacteriol . 112, 315 . Mesibov, R ., Ordal, G . W., and Adler, J . (1973). J . Gen . Physiol . 62, 203 . Ordal, G . W ., and Adler, J . (1974a) . J . Bacteriol . 117, 509 . Ordal, G . W., and Adler, J. (1974b) . J . Bacteriol . 117, 517 . Parkinson, J . S . (1974) . Nature 252, 317 . Silverman, M ., and Simon, M . (1973) . J . Bacteriol. 116, 114 . Silverman, M ., and Simon, M . (1974) . Nature 249, 73 . Tsang, N ., Macnab, R ., and Koshland, D . E ., Jr . (1973) . Science 181, 60 . Tso, W .-W ., and Adler, J . (1974) . J . Bacteriol . 118, 560 .