Aerotaxis (oxygen-seeking) behaviour in Escherichia coli is a response to changes in the electron transport system and not oxygen per se. Because changes in proton motive force (PMF) are coupled to respiratory electron transport, it is difficult to differentiate between PMF, electron transport or redox, all primary candidates for the signal sensed by the aerotaxis receptors, Aer and Tsr. We constructed electron transport mutants that produced different respiratory H+/e– stoichiometries. These strains expressed binary combinations of one NADH dehydrogenase and one quinol oxidase. We then introduced either an aer or tsr mutation into each mutant to create two sets of electron transport mutants. In vivo H+/e– ratios for strains grown in glycerol medium ranged from 1.46 ± 0.18–3.04 ± 0.47, but rates of respiration and growth were similar. The PMF jump in response to oxygen was proportional to the H+/e– ratio in each set of mutants (r2 = 0.986–0.996). The length of Tsr-mediated aerotaxis responses increased with the PMF jump (r2 = 0.988), but Aer-mediated responses did not correlate with either PMF changes (r2 = 0.297) or the rate of electron transport (r2 = 0.066). Aer-mediated responses were linked to NADH dehydrogenase I, although there was no absolute requirement. The data indicate that Tsr responds to changes in PMF, but strong Aer responses to oxygen are associated with redox changes in NADH dehydrogenase I.
Motile bacteria swim towards an oxygen concentration that is optimal for growth. This behavioural response, named aerotaxis, is a key determinant in the ecology of many bacteria. Sampling of the water column in a marine or estuarine environment reveals horizontal veils of bacteria, each species aggregated at a preferred oxygen concentration (Jorgensen, 1982; Canfield and Des Marais, 1991; Donaghay et al., 1992).
The Aer protein includes a PAS domain with FAD as a cofactor (Bibikov et al., 1997; 2000; Repik et al., 2000). PAS domains constitute a superfamily of sensory transduction modules that, in bacteria, sense oxygen, reduction potential and light (Taylor and Zhulin, 1999; Zhulin et al., 1997a). The PAS domain is believed to be the sensory input module of Aer, because FAD binding is essential for aerotaxis. An attractive model for Aer sensing proposes that a change in the redox state of FAD-PAS initiates the aerotaxis signal. The redox change could be produced by cytosolic electron donors, or by direct interaction of the PAS domain with the electron transport system. The latter scenario places the PAS domain in close proximity to the membrane, which may be the case, given that the PAS domain is tethered to the inner surface of the cytoplasmic membrane (Amin et al., 2006). In such a system, the redox state of Aer would be coupled to a component of the electron transport system which changes redox state upon changes in respiration (Taylor et al., 2001; 2003).
To distinguish between electron transport sensing and PMF sensing in E. coli, we compared aerotaxis in a series of electron transport mutants. The E. coli electron transport system is composed of three main components (Gennis and Stewart, 1996): (i) substrate-specific dehydrogenases that are able to accept electrons from a wide variety of organic compounds and pass them on to quinones, (ii) lipid-soluble quinones (ubiquinone, menaquinone and demethyl-menaquinone), and (iii) terminal reductases that transfer the electrons from a quinone to the terminal electron acceptor. By varying which dehydrogenase, quinone and reductase is expressed, bacteria are able to tailor their respiratory components to utilize various electron donors and acceptors present in the environment.
The aerobic electron transport system that oxidizes NADH in E. coli has two NADH dehydrogenases and two quinol oxidases (Fig. 1). NADH dehydrogenase I, product of the nuo operon, is similar to complex I in the mitochondria of eukaryotes and translocates two protons per electron (2 H+/e–) (Wikstrom, 1984; Bogachev et al., 1996; Gennis and Stewart, 1996; Galkin et al., 1999; Friedrich and Scheide, 2000). NADH dehydrogenase II, a simpler, single subunit protein expressed from the ndh gene, does not translocate H+ (Matsushita et al., 1987; Hayashi et al., 1989). The quinol oxidases include cytochrome bo oxidase, which is expressed from the cyo operon and translocates approximately 2 H+/e–, and cytochrome bd oxidase, which is expressed from the cyd operon and translocates approximately 1 H+/e– (Ingledew and Poole, 1984; Calhoun and Gennis, 1993; Trumpower and Gennis, 1994; Gennis, 1998). The protons translocated across the cytoplasmic membrane form the PMF. We constructed a series of electron transport mutants that expressed binary combinations of one NADH dehydrogenase and one cytochrome oxidase. An aer or tsr mutation was introduced into each mutant to create two sets of electron transport mutants. We were then able to correlate aerotaxis responses with both PMF and electron transport under conditions where the H+/e– ratio was varied. Tsr-mediated aerotaxis responses in E. coli correlated with changes in PMF. Aer-mediated aerotaxis did not correlate with PMF or respiration but was strongly influenced by the redox activity of NADH dehydrogenase I.
Strategy for differentiating between electron transport and PMF
In E. coli, the H+/e– ratio is the number of protons transported from the cytosol into the periplasm for each electron passed through the electron transport system. This ratio reflects the efficiency of the electron transport system in converting redox potential into a proton gradient. The ratio and efficiency varies depending on which electron transport components are expressed. In the aerobic respiratory pathway from NADH to oxygen, the components include two NADH dehydrogenases (coded by the nuo and ndh genes) which oxidize NADH, and two quinol oxidases (cyo and cyd) which reduce oxygen (Fig. 1). We constructed E. coli strains expressing one respiratory NADH dehydrogenase and one quinol oxidase by inactivating nuo or ndh and cyo or cyd in each strain. This yielded four strains expressing different binary combinations of these respiratory components. We expected that these strains would translocate different numbers of protons in response to an oxygen pulse, because binary combinations of these respiratory components have different predicted H+/e– ratios (Gennis and Stewart, 1996). If so, each strain could exhibit a unique jump in PMF in response to oxygen, and might be used to differentiate PMF sensing from electron transport sensing if the rate of respiration remained relatively constant.
To test this concept, we chose two aerotaxis receptors (Aer and Tsr) that sense PMF- and/or electron transport-mediated changes in response to oxygen. We and others have proposed that Aer may sense oxygen-mediated redox changes in the electron transport system (Bibikov et al., 1997; 2000; Rebbapragada et al., 1997; Repik et al., 2000; Taylor et al., 2001), while Tsr may sense oxygen-mediated changes in PMF (Rebbapragada et al., 1997; Repik et al., 2000; Taylor et al., 2001). As these receptors have overlapping phenotypes, we deleted the aer gene in one set of strains and deleted tsr in the other (Table 1). This meant that in addition to the two mutations in the respiratory components, each strain had a third mutation in either the aer or tsr gene, so that each strain expressed only one receptor for aerotaxis. For example, E. coli BT3401 (aer nuo cyo) expressed Tsr, NADH dehydrogenase II and cytochrome bd oxidase. In this article, we specify the relevant proteins expressed in a strain instead of the mutations present.
Table 1. Summary of relevant properties of E. coli electron transport mutants.a
The H+/e– ratio was measured as described in the text. Values shown are the mean of six or more determinations.
The rate of oxygen consumption was measured as described in the text (n = 6).
The membrane potential (Δψ) was measured at pH 7.6 as described in the text. The values shown are the increase in Δψ when anaerobic bacteria were exposed to air (n = 9).
The results are the mean of two independent experiments with three replicates in each.
Time (s) required for 50% of the bacteria to return to prestimulus behaviour after perfusion with N2 gas.
Time (s) required for 50% of the bacteria to return to prestimulus behaviour after exposure of an anaerobic culture to air (21% oxygen).
Abbreviations: Cyt bd, cytochrome bd oxidase; Cyt bo, cytochrome bo oxidase; NDH-1, NADH dehydrogenase I; NDH-2, NADH dehydrogenase II.
2.06 ± 0.20
5.90 ± 0.67
−40.8 ± 9.0
31.2 ± 7.2
59.2 ± 12.4
Tsr NDH-2 Cyt bd
1.46 ± 0.18
7.04 ± 0.82
−23.2 ± 8.2
27.7 ± 10
16.3 ± 2.4
Tsr NDH-2 Cyt bo
2.33 ± 0.50
6.70 ± 0.39
−42.9 ± 9.3
27.7 ± 6.1
52.7 ± 5.4
Tsr NDH-1 Cyt bd
1.77 ± 0.15
7.03 ± 1.93
−30.6 ± 11.8
22.7 ± 5.2
32.8 ± 4.4
Tsr NDH-1 Cyt bo
2.64 ± 0.20
6.76 ± 0.92
−48.0 ± 14.9
27.2 ± 14.2
58.3 ± 10.9
Aer NDH-2 Cyt bd
1.51 ± 0.25
7.42 ± 0.32
−27.8 ± 10.5
35.0 ± 4.3
17.0 ± 4.1
Aer NDH-2 Cyt bo
2.88 ± 0.41
6.30 ± 0.48
−55.3 ± 14.4
19.8 ± 9.2
29.8 ± 7.5
Aer NDH-1 Cyt bd
1.97 ± 0.44
6.97 ± 1.12
−39.9 ± 6.7
20.8 ± 2.6
56.2 ± 10.9
Aer NDH-1 Cyt bo
3.04 ± 0.47
6.75 ± 0.91
−62.3 ± 13.6
16.3 ± 3.8
57.2 ± 8.5
Characterization of strains
The phenotypes of all mutant strains were verified by measuring enzyme activity, difference spectra and chemotaxis. NADH dehydrogenase I can be distinguished from NADH dehydrogenase II by comparing the oxidation of NADH and deamino-NADH. NADH dehydrogenase I oxidizes deamino-NADH and NADH at similar rates whereas NADH dehydrogenase II has little deamino-NADH oxidase activity. As predicted, in cell extracts, the ratio of NADH oxidation compared with deamino NADH oxidation was approximately 1.0 for ndh strains that expressed only NADH dehydrogenase I, and greater than 4.0 for nuo strains that expressed NADH dehydrogenase II (results not shown).
We measured the expression of cytochrome bd by difference spectra. The reduced-minus-oxidized spectrum of haem d has an absorbance peak at 630 nm, which shifts 7 nm towards the red (637 nm) in the reduced(CO)-minus-reduced spectrum (Castor and Chance, 1959; Miller and Gennis, 1983; Bogachev et al., 1993). This spectrum was absent in all cyd mutants but present in all strains expressing cytochrome bd oxidase (Edwards, 2005). In contrast, the difference spectra for cytochrome bo (compare Ingledew and Poole, 1984; Gennis and Stewart, 1996) were not adequately resolved in cyd mutants. In these strains, we inferred that cytochrome bo was expressed at normal levels by confirming that respiration was normal.
The presence or absence of the tsr and aer genes was confirmed by the polymerase chain reaction (PCR) and by spatial gradient assays on tryptone (tsr) or succinate (aer) semisolid agar. Western blots were also used to confirm the absence of Aer protein expression in aer strains (Repik et al., 2000).
Respiration and growth of mutants is relatively unaffected by different H+/e− ratios
For growth at 30°C in H1 minimal salts medium with glycerol as a carbon source, the doubling time for wild type (RP437, 66 min) and Tsr-expressing (67–75 min) strains were similar. Growth rates for electron transport mutants were not significantly altered by H+/e– ratios, indicating the mutants are able to compensate for different H+/e– ratios, as reported previously (see Discussion in Unden and Bongaerts, 1997; Unden and Schirawski, 1997; Tran and Unden, 1998). Of note, doubling times were slightly longer for tsr strains expressing Aer (78–91 min), but the tsr strains entered stationary phase at an optical density at 600 nm of approximately 1.4 compared with 0.9 for wild-type and Tsr-expressing strains. We investigated whether these differences were specific to the tsr deletion or another factor.
The tsr deletion (Δtsr-7021; Callahan et al., 1987) includes 5 kb of flanking DNA sequences in addition to the tsr gene. To determine whether the change in growth was due to deletion of sequences other than tsr, BT3409 (Δtsr-5550:erm) was constructed with an in-frame deletion between the fourth and last codons of tsr and an erythromycin-resistance cassette inserted in its place. The growth of BT3409 cells was indistinguishable from the growth of wild-type cells. Therefore, the slower growth of RP5882 may be due to the extended deletion in Δtsr-7021, which includes the cstA and yjiA genes, both of which are involved in carbon starvation responses (Kasahara et al., 1991; Matin, 1991; Schultz and Matin, 1991; Marschall et al., 1998). However, the growth effects of the Δtsr-7021 deletion did not appear to impair Aer-mediated aerotaxis in this study or in our previous studies (Rebbapragada et al., 1997; Repik et al., 2000).
To estimate electron transport in wild-type and electron transport mutants, we measured the rate of respiration in intact cells. The respiration rate varied minimally between strains (Table 1). The only statistically significant difference in respiration was between BT3402 cells and BT3404 cells (compare Table 1). However, RP437, which is wild type for chemotaxis, had a significantly slower respiration rate than the strains used in this study.
All strains were grown aerobically to minimize differences in expression of electron transport components
Both the ndh and cyo genes are repressed under anaerobic conditions (Green and Guest, 1994) and expressed maximally under highly aerobic conditions. In contrast, cyd expression levels are repressed under highly aerobic conditions and are highest when oxygen is approximately 7% of air saturation (Tseng et al., 1996; Salmon et al., 2003). Under anaerobic conditions, cyd expression is 40% of maximum levels. To minimize transcriptional changes in these components, all strains were grown aerobically to OD600 = 0.35–0.5 and harvested. The effect of anoxia and re-aeration on these cells was studied by rapidly changing the ambient oxygen concentration of the harvested cells, unless specified otherwise. No significant changes in the transcription of the investigated proteins were expected within this time frame.
The in vivo H+/e− ratios were altered in electron transport mutants
To determine the H+/e– ratio, we quantified the number of protons translocated across the inner membrane during transport of a known number of electrons through the respiratory chain (Fig. 2). Briefly, cells in a temperature-controlled chamber equipped with a pH-sensitive electrode, were made anaerobic by flushing with argon gas. Air-saturated water (100 μl) was injected into the chamber, causing a transient burst of respiration that released H+ into the medium and lowered the extracellular pH. Thiocyanate in the medium collapsed membrane potential (Δψ) (Reenstra et al., 1980; Bogachev et al., 1996) and prevented the rapid re-uptake of the released H+. The recording was calibrated by injecting 10 μl aliquots of anoxic 5 mM HCl. The ratio of H+ translocated per O2 molecule injected was calculated and converted to an H+/e– ratio by dividing by 4, the number of electrons required to reduce one O2 molecule.
The method was first validated by measuring the H+/e– ratio for E. coli RP437 wild-type cells grown anaerobically with glycerol as the energy source and dimethylsulphoxide (DMSO) as the electron acceptor (Bogachev et al., 1996) (data not shown). After extrapolation to correct for re-uptake of H+, the calculated H+/e– ratio was 2.3, which is similar to the published value of 2.5 (Bogachev et al., 1996).
The H+/e– ratios for the series of Aer- and Tsr-expressing constructs in glycerol medium were similar to published in vivo values (Jones et al., 1975; Poole and Haddock, 1975; Bogachev et al., 1996) (Table 1). A typical recording from an H+/e– assay of aerobically grown RP437 is shown in Fig. 2. For mutant strains, ratios between 1.51 and 3.04 (Aer-expressing strains), and 1.46 and 2.64 (Tsr-expressing strains) were observed (Table 1). The most important determinant for higher H+/e– ratios was cytochrome bo oxidase. In all isogenic strains, the mutant expressing cytochrome bo oxidase had a significantly higher H+/e– ratio than the mutant expressing cytochrome bd oxidase (P ≤ 0.002) The data are consistent with the reported translocation of 2 H+/e– by cytochrome bo oxidase and 1 H+/e– by cytochrome bd oxidase (Wikstrom, 1989; Calhoun et al., 1993; Gennis, 1998; Rich et al., 1998; Umemura et al., 2002).
Proton motive force analysis
Proton motive force consists of a chemical potential (ΔpH) and an electrical potential (Δψ) (Mitchell, 1961; Gennis and Stewart, 1996). To estimate PMF, we determined the membrane potential at a pH identical to that of the cytosol (pH 7.6). At this pH, the ΔpH component of the PMF is zero, so the entire PMF is in the form of membrane potential (Δψ) (PMF = Δψ − 59ΔpH). The Δψ was estimated by measuring the cellular uptake of the lipophilic cation, tetraphenyl phosphonium (TPP+), which concentrates in the cell due to the negatively charged interior of the membrane. The drop in extracellular TPP+ after it was taken up by the cells was measured with a TPP+-sensitive electrode as described previously (Bespalov et al., 1996; Zhulin et al., 1996). The steady-state PMF in aerobic cells was similar in all strains. For example, the average PMF in cells expressing Tsr was −130 mV (BT3401), −146 mV (BT3403), −141 mV (BT3405) and −152 mV (BT3407).
To determine PMF during anoxia, cells grown aerobically (as described above) were added to a temperature-controlled chamber and flushed with argon until they became anaerobic. The PMF (Δψ) was measured when the cells were anaerobic, and after re-equilibration of the medium with air (Table 1). The increase in Δψ upon aeration (ΔψΟ2 − ΔψΝ2) was an average of −22 mV higher in strains that expressed cytochrome bo oxidase, compared with strains that expressed cytochrome bd oxidase (P < 0.01). NADH dehydrogenase I (2 H+/e–) contributed little towards PMF, unlike the terminal oxidase (Table 1). This is consistent with reports showing that NADH dehydrogenase I in wild-type cells is a minor contributor to PMF during aerobic respiration, ostensibly due to the much higher NADH dehydrogenase II activity (see Discussion in Unden and Bongaerts, 1997; Unden and Schirawski, 1997). In the present study, the contribution of NADH dehydrogenase I towards PMF was low, even in the absence of NADH dehydrogenase II, perhaps because alternative dehydrogenases are more active in introducing electrons into the electron transport system. Twelve additional substrate-specific dehydrogenases are known to donate electrons to the quinone pool of the E. coli electron transport system (Gennis and Stewart, 1996). Overall, the increase in Δψ on aeration closely correlated with the measured H+/e– ratios (Table 1).
The minor variations we observed between the aerobic PMF of mutant and wild-type strains were similar to observations previously reported by others. There was no correlation between the aerotaxis response and the aerobic steady-state PMF. This finding was anticipated. The earliest studies of mammalian sensory systems determined that the sensory response is proportional to the relative change in stimulus, rather than the magnitude of the stimulus per se. This sensitivity is likely related to the adaptive mechanisms of sensory receptors, which reset their baseline response to zero by reversible covalent modification (see Kehry et al., 1983; Terwilliger et al., 1986; Nowlin et al., 1987). Our working hypothesis was that an aerotaxis receptor would detect the change in PMF or electron transport when the ambient oxygen concentration changed.
Quantitative behavioural analysis in temporal assays
Temporal aerotaxis assays are both quantitative and independent of metabolism (Laszlo and Taylor, 1981). This is in contrast to spatial aerotaxis assays in semisolid agar or capillaries, which depend on consumption of nutrients/oxygen to form a gradient of chemoeffectors (Laszlo and Taylor, 1981). Temporal responses to the change in oxygen concentration were measured in each strain by placing a drop of culture in a microchamber. The chamber was perfused with N2 gas until the bacteria were anaerobic and then re-exposed to air. The behaviour of the bacteria was observed in a dark-field microscope and videotaped for analysis. When aerobic bacteria were exposed to N2 gas they tumbled constantly until they adapted and returned to random motility. When re-exposed to oxygen, the bacteria suppressed tumbling (directional changes) and swam smoothly until they adapted to the oxygen and returned to random motility (Laszlo and Taylor, 1981; Rebbapragada et al., 1997). We quantified the time interval for half of the bacteria to return to the prestimulus tumbling frequency.
The responses of RP437 (wild type) to an oxygen increase (59 s) or decrease (31 s) were similar to those published previously (Rebbapragada et al., 1997). Responses of the other strains used in the study are shown in Table 1. As evident in this Table, the most relevant protein for maximal smooth-swimming responses in Aer-expressing strains was NADH dehydrogenase I. For Tsr-expressing strains the most relevant protein was cytochrome bo oxidase.
To determine quantitative relationships between the parameters listed in Table 1[i.e. H+/e–, respiration, PMF jump (ΔψO2 − ΔψN2) and aerotaxis responses], we analysed two-dimensional plots of all possible combinations of these parameters. Of note, H+/e– ratios were proportional to the rise in PMF (ΔψO2 − ΔψN2) (Fig. 3, upper plots), showing highly significant correlations (r2 = 0.986–0.996) in both the Aer-expressing and Tsr-expressing strains. However, the smooth-swimming (temporal) aerotaxis responses in these strains did not have the same correlation patterns. Tsr-mediated temporal responses showed a highly significant correlation (r2 = 0.989) with the rise in PMF in response to air (Fig. 3, lower right), but Aer-mediated responses did not correlate with the rise in PMF (r2 = 0.297) (Fig. 3, lower left), nor did they correlate with the observed respiration rate (r2 = 0.066, data not shown). The simplest interpretation of these results is that Tsr, but not Aer, mediates aerotaxis by monitoring changes in PMF when the oxygen concentration increases.
Behavioural responses in spatial gradients
In addition to the quantitative studies described above, we also made qualitative observations of colony morphologies formed by mutant and wild-type strains. In semisolid agar, motile bacteria consume nutrients at the inoculation site and swim outwards in response to the gradients of consumed metabolites (Adler, 1966; Wolfe and Berg, 1989). Visible rings of bacteria form, and continue to expand over time. Each ring results from chemotaxis to a specific molecule. For wild-type E. coli and aer strains in tryptone semisolid agar, the outer ring is formed by bacteria sensing serine via the chemoreceptor Tsr, and a second inner ring by bacteria sensing aspartate via the chemoreceptor Tar (Adler, 1966; Hedblom and Adler, 1980; 1983) (Fig. 4A and B). The contribution of Tsr was evident in colonies formed by a tsr mutant: the serine ring was missing (Fig. 4C). This ring was also absent in an aer tsr strain (BT3312) (Fig. 4D), which had a smaller colony size, despite the presence of an aspartate ring. For unknown reasons, the aspartate ring was faint in strains lacking NADH dehydrogenase I (Fig. 4E), as reported previously for such mutants (Falk-Krzesinski and Wolfe, 1998).
The three-dimensional shape of the colonies on tryptone agar varied significantly in electron-transport mutants. Wild-type cells (RP437) formed wide, dome-shaped colonies with a central ring of dense growth at the surface and walls that extended to the bottom of the semisolid agar. Strains without cytochrome bd oxidase (e.g. BT3407) formed colonies with walls that did not reach the bottom of the agar where oxygen levels would be lowest. The requirement of cytochrome bd for taxis to the bottom of the agar was similar for Aer-expressing and Tsr-expressing strains. Two factors may play a role in this phenotype: (i) cytochrome bd oxidase has a higher affinity for oxygen than cytochrome bo oxidase (the Kd for oxygen is 0.2 μM for cytochrome bd oxidase and about 2 μM for cytochrome bo oxidase) (Rice and Hempfling, 1978; Anraku and Gennis, 1987; Gennis and Stewart, 1996), and (ii) cytochrome bo oxidase expression is suppressed in an anaerobic environment while cytochrome bd oxidase is expressed at a wider range of oxygen concentrations, reaching maximum expression at microaerobic oxygen levels (Salmon et al., 2003).
On succinate semisolid agar, wild-type (RP437) and RP5882 (tsr) strains formed dome-shaped colonies. The bacteria at the outer edge formed a distinct ring that was visible from above (Fig. 4F and G). Aer-mediated succinate taxis was dependent on NADH dehydrogenase I but not on cytochrome bo or cytochrome bd (Fig. 4H–J). An aer mutation reduced swarming on succinate plates, but not motility, consistent with previous studies showing that an aer mutant does not mediate taxis to succinate (Bibikov et al., 1997). The basis for this defect lies not with the Tsr receptor, which can in fact mediate succinate taxis. Rather, it is caused by the presence of the Tar receptor, which inhibits colony expansion as aspartate is excreted from the cells under these conditions (Bibikov et al., 2000).
Glycerol taxis in wild-type E. coli (RP437) yielded dome-shaped colonies with an outer ring and shape similar to colonies on succinate semisolid agar (Fig. 4L; Greer-Phillips et al., 2003; Zhulin et al., 1997b). Unlike succinate taxis, glycerol taxis on semisolid agar was mediated by Aer or Tsr (Figs 4M and N). Tsr-mediated glycerol taxis required cytochrome bo oxidase and was enhanced by NADH dehydrogenase I (Fig. 4O–R). These effects were not related to motility, as cyo strains that were deficient in glycerol taxis (BT3401 and BT3405) had normal motility when viewed under the microscope. Unlike Tsr-mediated glycerol taxis, the major determinant of colony size with Aer-mediated glycerol taxis was NADH dehydrogenase I (Fig. 4T–V). Interestingly, BT3312 (aer tsr) showed some chemotaxis on glycerol semisolid agar (Fig. 4S). Whether this was mediated by Tar or the other MCPs is not known, although glycerol taxis is eliminated in strains lacking all chemoreceptors (Greer-Phillips et al., 2003).
Behavioural responses in capillaries
In culture-filled glass capillaries, bacterial respiration creates an oxygen gradient with the highest oxygen concentration at the air–liquid interface. Bacteria sense this gradient with aerotaxis receptors and accumulate at their preferred oxygen concentration. Wild-type E. coli accumulate in a visible band at 5% oxygen (Fig. 5I; M. Johnson, unpubl. obs.). In Luria–Bertani (LB) medium, the Aer-expressing strains (Fig. 5A–D) formed bands closer to the air interface than wild-type cells (Fig. 5I), whereas Tsr-expressing strains formed bands further from the interface, as reported previously (Rebbapragada et al., 1997). Cytochromes bo and bd had additional effects on the placement of Aer strains and density of the band. The capillaries in Fig. 5 are arranged with the mutant strains that produced the largest rise in Δψ at the top (Fig. 5A and E) and those with the smallest rise in Δψ at the bottom (Fig. 5D and H). Aer strains expressing cytochrome bo (Fig. 5A and B) formed denser bands, and banded closer to the meniscus than isogenic strains expressing cytochrome bd (Fig. 5C and D). The band placement is likely related to each cytochrome's Kd,O2 which is higher for the ‘meniscus-seeking’ cytochrome bo than for the ‘interior-seeking’ cytochrome bd. The finding that Aer-mediated bands also formed in strains without NADH dehydrogenase I (BT3404 and BT3402; Fig. 5B and D) was unexpected because these strains were not aerotactic in semisolid agar (Fig. 4), and had shorter responses in temporal assays (Table 1, Fig. 3). This indicated that NADH dehydrogenase I was not absolutely necessary for Aer to orchestrate aerotaxis in a spatial assay, and also highlights subtle differences that can be observed with these aerotaxis assays.
Unlike Aer-expressing strains, the formation of a band by Tsr-expressing strains required the presence of cytochrome bo as the terminal oxidase (Fig. 5E and F), correlating with a larger rise in Δψ (Table 1). The band was diffuse on the distal surface. Tsr-expressing strains containing cytochrome bd oxidase were negative for aerotaxis, and the cells were distributed uniformly throughout the capillary (Fig. 5G and H). These strains were also negative for glycerol taxis on semisolid agar.
Aerotactic responses in E. coli do not require oxygen per se; rather, they require a step change in electron flow through the electron transport system. (Taylor et al., 1979; Taylor, 1983). This requirement can be demonstrated with the alternative electron acceptors, nitrate and fumarate, which generate substantial PMF from anaerobic respiration (Tran and Unden, 1998) and cause an aerotaxis-type response when introduced under anaerobic conditions (Taylor et al., 1979). These alternative electron acceptors can also competitively inhibit the responses of anoxic bacteria to oxygen (Laszlo and Taylor, 1981). Presumably, this inhibition results from a smaller step increase in respiration when oxygen is introduced (see Tran and Unden, 1998), as inhibition requires the terminal reductase (nitrate reductase or fumarate reductase) of the anaerobic electron transport system (Laszlo and Taylor, 1981). The present study was undertaken to determine how the Aer and Tsr receptors in E. coli sense changes in the electron transport system.
Separating PMF from electron transport
The discovery that Aer binds an FAD cofactor (Bibikov et al., 1997) led to a hypothesis that Aer senses redox changes in the electron transport system (Bibikov et al., 1997; 2000; Rebbapragada et al., 1997; Repik et al., 2000; Taylor et al., 2001). In contrast, the Tsr receptor does not have a similar redox centre, but did require a functional electron transport system, so we proposed that Tsr sensed a component (ΔpH or Δψ) of the PMF (Rebbapragada et al., 1997). In this investigation, we explored the role of the electron transport system in signalling using electron transport mutants with different respiratory H+/e– ratios, in an effort to independently vary PMF and electron transport.
We created four electron transport mutants that expressed different binary combinations of one NADH dehydrogenase and one cytochrome oxidase. These mutants exhibited different step increases in PMF in response to air, even though respiration rates were similar (Table 1). The rise in membrane potential in response to air closely mirrored the values determined for the H+/e– ratios in the same strains (Fig. 3). This reflects the close coupling between these two parameters. If the respiration rates had risen markedly in strains with lower H+/e– ratios, increased respiration could theoretically have produced a larger increase in PMF in response to air. But, this was not observed in the present studies.
The Tsr-mediated smooth-swimming response correlated directly with the rise in membrane potential that occurred when air was introduced (Fig. 3). That is, a larger PMF jump elicited a longer smooth-swimming response. This response did not correlate with the membrane potential under steady-state anaerobic or aerobic conditions (data not shown). In this respect, aerotaxis is comparable to chemotaxis in which Tsr responds to the relative change in serine concentration (Springer et al., 1979).The Tsr receptor is precisely tuned to steady-state conditions by a methylation system, which methylates up to five glutamyl residues in the signalling domain of the receptor (Kehry et al., 1983; Bibikov et al., 2004; for a review see Bourret and Stock, 2002). This allows the receptor to respond to subtle changes in stimuli, and contributes to the ‘perfect adaptation’ exhibited by bacterial chemotaxis sytems (Yi et al., 2000).
Although the rise in membrane potential that accompanied the transition between anaerobic and aerobic conditions was relatively small compared with the steady-state potentials of these strains under aerobic conditions (∼140–160 mV), the increases were apparently above the threshold for PMF sensing by Tsr. The length of the smooth-swimming response to air increased from 16 s for a 25 mV jump in PMF, to 58 s for a 48 mV jump (Fig. 3; Table 1). The sensing elements for the Tsr-mediated response are not known, but likely candidates include charged residues on either side of the membrane-spanning segments of the protein.
The Aer-mediated aerotactic responses did not correlate with the change in PMF (Fig. 3) or respiration (r = 0.066, data not shown), leaving redox changes as the most likely signal for aerotaxis. There are two possible pathways for signalling to the Aer protein: Aer could be reduced directly by a specific dehydrogenase or molecule (such as ubiquinone) in the respiratory complex, or it could be reduced by a cytosolic electron donor (e.g. NADH) or diffusible redox component which itself is reduced by the electron transport system (e.g. FAD). The current study cannot exclude either scenario, although Aer-mediated sensing was strongly influenced by NADH dehydrogenase I (Table 2). Interestingly, it is NADH dehydrogenase II activity that is predominant in aerobic E. coli (Unden and Bongaerts, 1997; Unden and Schirawski, 1997), so it is noteworthy that aerotaxis is linked to NADH dehydrogenase I in aerobic cells. Although the activity of NADH dehydrogenase I was lower than NADH dehydrogenase II (data not shown), it did not limit respiration in aerobic cells when NADH dehydrogenase II was absent.
Table 2. Summary of requirements for an NADH dehydrogenase and quinol oxidase in aerotactic responses transduced by the Aer and Tsr receptors.
Cyt, cytochrome; NDH, NADH dehydrogenase; NR, no response.
NDH-1 > NDH-2
Cyt bo > Cyt bd
NDH-1 > NDH-2
NADH dehydrogenase I could reduce Aer directly by interacting with a partial PAS-domain sequence that is part of the NuoE subunit of NADH dehydrogenase I (I. Zhulin, pers. comm.). Other PAS domains are known sites for dimerization (Taylor and Zhulin, 1999). However, if NADH dehydrogenase I can reduce Aer directly, it is not the sole electron donor, as there was not an absolute requirement for this dehydrogenase, except on plate assays. In temporal assays, smooth-swimming responses were longer when NADH dehydrogenase I was expressed, but in capillary assays, NADH dehydrogenase I was not absolutely required (Fig. 5). Taken together, the results indicate that NADH dehydrogenase I is an important segment of electron transport for Aer-mediated responses, and that strong redox signals originate in this segment.
Role of the quinol oxidases
Cytochrome bo oxidase had a prominent role in PMF (Tsr)-mediated responses (Tables 1 and 2). Presumably, this was due to the fact that cytochrome bo oxidase translocated 2 H+/e– compared with 1 H+/e– for cytochrome bd oxidase. Strains lacking cytochrome bo were defective in Tsr-mediated responses on plate assays (Fig. 4) and capillary assays (Fig. 5), and exhibited shorter responses in temporal assays (Fig. 3; Table 1). This was not true for Tsr strains that expressed cytochrome bo oxidase, regardless of which NADH dehydrogenase was expressed. These strains formed colonies of similar sizes on glycerol motility plates (Fig. 4O and Q) and formed bands in capillary assays (Fig. 5E and F). From these data, we infer that the quinol oxidase segment of electron transport is the more important segment for PMF (Tsr)-mediated responses (Table 2), probably because NADH dehydrogenases did not contribute substantially to the PMF under aerobic conditions.
The type of terminal oxidase also altered the preferred oxygen concentration to which E. coli migrated (Fig. 5). The bacteria were attracted to a lower concentration of oxygen when cytochrome bd was the functional oxidase, apparently reflecting the higher affinity of cytochrome bd for oxygen.
Significance of the findings
The possibility of constructing E. coli mutants with a range of H+/e– ratios (Fig. 1) was discussed in principle (Calhoun et al., 1993) but, to our knowledge, this is the first investigation in which H+/e– ratios were measured in vivo for constructed strains with all possible combinations of NADH dehydrogenases and quinol oxidases. These strains are a resource for future studies of bioenergetics in E. coli. The observed H+/e– ratios in the mutants constructed for this study ranged from 1.5 to 3.0 (Table 1). The constructed strains with the lowest H+/e– ratios grew at a similar rate to strains with the highest H+/e– ratios, despite the difference in efficiency of oxidative phosphorylation. The range of respiration in the different strains growing in glycerol medium was only 5% (Tsr) and 17% (Aer), respectively, indicating that homeostasis can compensate for a 50% decrease in the H+/e– ratio.
The collective findings of this study indicate that Tsr senses changes in PMF. Elucidating this step completes the sensory transduction pathway for Tsr-mediated aerotaxis, a pathway that converts a change in oxygen concentration to a change in reversal frequency of the flagellar motors (Fig. 6). For Aer-mediated responses, the nature of the link between the electron transport system and Aer has not been established, although the current study offers more clues. Data indicating (i) that changes in PMF and respiration were not sensed by Aer, and (ii) that NADH dehydrogenase I strongly influenced Aer-mediated responses, support the hypothesis that Aer is a redox sensor.
Unless noted otherwise, E. coli was grown at 30°C with vigorous aeration in either LB broth (Davis et al., 1980) containing thiamine (1 μM) or H1 minimal salts medium supplemented with auxotrophic requirements and a specified carbon source (Adler, 1973).
Escherichia coli UU1117 (aer) and RP5882 (tsr), isogenic strains of RP437, were used as parental strains to create mutants with binary combinations of the electron transport system components (see Table 3). P1 transductions were performed using the method of Miller (1992). For in-frame ndh gene replacements, the temperature sensitive plasmid, pKJW1 (K. Watts), was derived from pKO3 (Link et al., 1997). An 800 bp region upstream (including ndh codons 1–4) and a 760 bp region downstream of the ndh gene (including the three nucleotides following the stop codon) were amplified by PCR and cloned into the SmaI site of pKO3. The 982 bp erythromycin cassette was then amplified by PCR and cloned into an engineered BglII restriction site between the upstream and downstream components. Allelic exchange with this new vector, was performed as described previously (Link et al., 1997; Yu et al., 2002). For in-frame tsr gene replacements, another temperature sensitive vector, pRK1 (R. Korsen), was derived from pKO3. A 735 bp segment upstream (including tsr codons 1–4) and a 1016 bp segment downstream of the tsr gene (including the last tsr codon), were amplified by PCR and cloned into the XmaI site of pKO3. The 982 bp erythromycin cassette was then amplified by PCR and cloned into an engineered PstI restriction site between the upstream and downstream components. The new vector, pRK1, was used for tsr gene replacements.
Table 3. Bacterial strains used in this study.
Reference or construction (parents; relevant selection)
The tsr gene was replaced by allelic exchange with the temperature sensitive plasmid pRK1, constructed by R. Korson.
The ndh gene was replaced by allelic exchange with the temperature sensitive plasmid pKJW1, constructed by K. Watts.
All mutations were verified by the size of the product produced by PCR (all strains) and the phenotype was confirmed. The tsr strain was also tested for serine taxis on tryptone swarm plates, whereas cyo and cyd mutations were confirmed by difference spectra of the constructed strains (Ingledew and Poole, 1984).
Bacteria were grown overnight in 2 l of LB medium, and resuspended in 2 ml of 50 mM K+PO4 buffer, pH 7.6, containing 0.3% lysozyme, 1 μg ml−1 DNase I (Worthington Biochemical Corporation, Lakewood, New Jersey) and one tablet Complete Mini protease inhibitor cocktail (Roche Diagnostics, Mannheim, Germany). Cells were freeze/thawed five times and then sonicated (Branson Sonic Power Company, Danbury, CT) three times at 50% power, 40% duty cycle for 20 pulses. The lysate was centrifuged at 12 000 g and the supernatant was recentrifuged at 480 000 g at 3°C. The pellet, containing the membrane fraction, was resuspended in 1.5 ml of a 50 mM K+PO4, pH 7.6 buffer and stored on ice.
A cuvette containing one sample was reduced with a small amount of Na+-dithionite while a cuvette with an identical sample was shaken to aerate the sample. For the CO-reduced samples, the cuvette was flushed with CO after reduction with Na+-dithionite. Reduced minus oxidized and reduced[CO] minus reduced spectra were recorded from 400 nm to 700 nm with a dual-beam spectrophotometer Cary 300 Bio UV-Vis Spectrophotometer (Varian, Walnut Creek, CA).
Determination of H+/e− ratios
Bacteria were grown in H1 minimal salts medium with 0.5% glycerol to an OD600 of 0.35–0.5, harvested and washed twice in 150 mM KCl, pH 6.5. Cells were resuspended to an OD600 of 20–25 in reaction buffer containing 100 mM KCl, 50 mM KSCN, 0.5 mM MES, 10 mM glycerol at pH 6.5. Glycerol was omitted or replaced with 10 mM malate or 10 mM succinate in the reaction buffer to measure H+/e– ratios generated by endogenous or alternate energy sources.
Proton extrusion by suspended bacteria was measured in a closed 3 ml chamber containing four ports [World Precision Instruments (WPI), Sarasota, FL; Cat WPI# NOCHM-4]. An H+-sensitive ion-selective electrode (WPI; KWIKH-2) and a compatible Dri-Ref reference electrode (WPI; DRIREF-2) were inserted into two of the side ports. An argon gas line was inserted through the top port and a Hamilton syringe was inserted into an injection port to deliver anaerobic 20 mM KOH, anaerobic 5 mM HCl or air-saturated H2O. The argon line was split into three gas lines: one to the chamber, and two lines to perfuse a 20 mM KOH solution and a 5 mM HCl solution. A Faraday cage of copper mesh was constructed around the apparatus and the cage was electrically grounded. Mixing was by a magnetic stirrer bar in the chamber. The change in extracellular pH, measured in millivolts, was recorded with a MacLab MKIII data acquisition system (Analog Digital Instruments, Milford, MA).
The chamber with 1.5 ml cell suspension was flushed with argon for 10 min to ensure anaerobiosis (Fig. 2A). Although the pH of the buffer was 6.5, the cells lowered the pH considerably, and the pH was re-adjusted to pH 6.5 by injecting argon-flushed 20 mM KOH (Fig. 2B). After the pH stabilized, 100 μl (25 nmol O2; Robinson and Cooper, 1970) of air-saturated water was injected into the chamber (Fig. 2C) and the change in millivolts was recorded. The pH was readjusted to ≥ 6.5 (Fig. 2D), and a standard curve of mV versus pH was created by injecting 10 μl (50 nmol) aliquots of argon-flushed 5 mM HCl (Fig. 2E). The H+/e– ratio was derived by dividing the H+/O2 ratio by 4, which is the number of electrons accepted by one oxygen molecule. H+/e– measurements were made in duplicate on a minimum of three experiments (n = 6).
Proton motive force measurements
Proton motive force was determined in bacteria at pH 7.6, which is also the intracellular pH of E. coli (Slonczewski et al., 1981). The ΔpH is then zero and the PMF is composed solely of the membrane potential. Membrane potential was quantified by measuring the partition of the permeating lipophilic cation, TPP+ in response to the charge difference across the inner membrane, as described previously (Bespalov et al., 1996). The change in external TPP+ concentration was measured by a TPP+-sensitive electrode. PMF measurements were made in triplicate on a minimum of three experiments (n = 9).
Bacteria were grown at 30°C in H1 minimal salts medium with 0.5% glycerol as energy source. Cells were harvested at an OD600 of 0.35–0.5, washed twice and resuspended to an OD600 of 0.4 in a buffer containing 10 mM K+PO4, pH 7.6, supplemented with 20 mM glycerol. The cells were kept on ice until respiration was determined in a chamber at 30°C using a Clarke-type oxygen electrode, as described previously (Bespalov et al., 1996). Respiration measurements were made in duplicate on a minimum of three experiments (n = 6).
Swarm plate assays
Assays were performed on tryptone semisolid agar (0.28%) (Armstrong and Adler, 1969; Wolfe and Berg, 1989), or on semisolid agar with H1 minimal media containing 15 mM carbon source (compare Bibikov et al., 1997; Greer-Phillips et al., 2003). The composition of succinate semisolid agar used in this study differed from that used by Bibikov and Parkinson (Bibikov et al., 1997). Consequently, there may be differences in the swarm morphology. For each carbon source, strains were twice subcultured in H1 minimal medium supplemented with the carbon/energy source and the second culture used to inoculate (3 μl) fresh swarm plates which were grown at 30°C in a humid environment. Colonies were illuminated with indirect light and rings were viewed from the top. Images were captured with an Alpha Innotech imaging system (Alpha Innotech Imaging, CA).
Aerotaxis capillary assays
Bacterial cultures were grown to mid-log phase at 30°C in LB medium. An optically flat, open-ended, 0.1 mm glass capillary tube (VitroCom, Mt. Lakes, NJ) was inserted into the culture for several minutes to allow the bacterial suspension to rise into the capillary tube, and the capillary tube was placed on a microscope slide. An oxygen gradient was allowed to form near the air–liquid interface in the capillary (15–30 min) before viewing the capillaries with a dark-field microscope (Leitz Dialux, Wetzler, Germany) at a 62.5× magnification. The microscope was fitted with a video camera (Cohu, San Diego, CA) attached to a Power Macintosh 8500/150 computer equipped with Apple Video Player Version 1.7.1 image acquisition software (Apple Computers, Cupertino, CA).
Temporal aerotaxis assays
The response of bacteria to an increase or decrease in oxygen concentration was measured in a gas flow chamber on a microscope stage, as described previously (Laszlo and Taylor, 1981). For each sample, a drop of bacteria was made anaerobic by perfusion with N2 gas, then the perfusate was switched to air and the response of the bacteria was video recorded. This was repeated twice and each strain was studied in triplicate on three separate days (n = 9). Video records were reviewed by eye and, for each change in oxygen concentration, the time for 50% of the bacteria to return to prestimulus tumbling frequency was recorded. Response times were independently confirmed by two people.
Statistical differences were determined using a two tailed, homoscedastic t-test and P-values < 0.05 were considered significant. Correlation coefficients (r) and coefficients of determination (r2) were calculated by a linear correlation test.
We thank I. Zhulin for suggesting the use of electron transport mutants and I. Zhulin, K. Watts, R. Gennis, S. Greer-Phillip, J.S. Parkinson and A. Wolfe for helpful discussions. We thank K. Watts and R. Korson for plasmid construction and A. Wolfe, R. Gennis, K. Watts and J.S. Parkinson for donation of strains. S. Fry and N. Abraham provided technical assistance. This investigation was supported by a grant from the National Institute of General Medical Sciences (GM29481) to B.L. Taylor and a National Medical Test Bed award to M.S. Johnson.