Evidence that acetyl phosphate functions as a global signal during biofilm development



We used DNA macroarray analysis to identify genes that respond to the status of the intracellular acetyl phosphate (acP) pool. Genes whose expression correlated negatively with the ability to synthesize acP (i.e. negatively regulated genes) function primarily in flagella biosynthesis, a result consistent with observations that we published previously (Prüß and Wolfe, 1994, Mol Microbiol 12: 973–984). In contrast, genes whose expression correlated positively with the ability to synthesize acP (i.e. positively regulated genes) include those for type 1 pilus assembly, colanic acid (capsule) biosynthesis and certain stress effectors. To our knowledge, this constitutes the first report that these genes may respond to the status of the intracellular acP pool. Previously, other researchers have implicated flagella, type 1 pili, capsule and diverse stress effectors in the formation of biofilms. We therefore tested whether cells altered in their ability to metabolize acP could construct normal biofilms, and found that they could not. Cells defective for the production of acP and cells defective for the degradation of acP could both form biofilms, but these biofilms exhibited characteristics substantially different from each other and from biofilms formed by their wild-type parent. We confirmed the role of individual cell surface structures, the expression of which appears to correlate with acP levels, in fim or fli mutants that cannot assemble type 1 pili or flagella respectively. Thus, the information gained by expression profiling of cells with altered acP metabolism indicates that acP may help to co-ordinate the expression of surface structures and cellular processes involved in the initial stages of wild-type biofilm development.


Some bacteria form complex, sessile communities called biofilms that begin to develop when bacteria transit from a free-living, planktonic existence to a lifestyle in which they attach firmly to surfaces, interfaces and/or each other. This transition involves reversible adsorption of planktonic cells to a suitable surface often facilitated by flagella, followed by irreversible attachment and colonization, most often facilitated by pili and other adhesins. After the formation of an attached monolayer, three-dimensional structures develop. These structures comprise water-filled channels that separate mushroom-shaped microcolonies composed of cells and their products encased in a matrix of extracellular polysaccharides, also known as capsule (Stoodley et al., 2002).

Cells co-ordinate gene expression in response to environmental cues via a select subset of small molecules, including ppGpp, cAMP and leucine (Barker et al., 2001a,b; Bruckner and Titgemeyer, 2002; Chang et al., 2002; Hung et al., 2002). Several years ago, Wanner and others proposed that acetyl phosphate (acP), a high-energy, phosphorylated, activated acetate derivative, also plays a global regulatory role (Wanner and Wilmes-Riesenberg, 1992; McCleary et al., 1993). Yet, in contrast to most of its counterparts, the manner in which acP might function as a global signal remains an enigma.

Given its central location within carbon metabolism, acP is well suited to monitor nutritional status, a contention supported by the fact that acP levels respond to the quantity and quality of the nutritional environment (McCleary and Stock, 1994; Prüß and Wolfe, 1994). As the intermediate of the phosphotransacetylase–acetate kinase (Pta-AckA) pathway (Fig. 1) (Rose et al., 1954; Brown et al., 1977), acP plays a critical role during aerobic growth on excess carbon and during mixed acid fermentation. In the presence of oxygen, when the carbon flux into cells exceeds the amphibolic capacity of the tricarboxylic acid (TCA) cycle, cells recycle coenzyme A (CoASH) by directing excess acetyl-coenzyme A (acCoA) through the Pta-AckA pathway, excreting acetate and generating ATP (Brown et al., 1977; Chang et al., 1999). In the absence of oxygen, acCoA cannot enter the TCA cycle. To regenerate CoASH, the cells guide acCoA through the Pta-AckA pathway, again excreting acetate while generating ATP (el-Mansi and Holms, 1989). As a result, acP levels can reflect the nutritional status of the cell: a large acP concentration reflects either the presence of excess carbon and/or the lack of oxygen, whereas a small concentration reflects limited carbon in the presence of oxygen.

Figure 1.

Schematic of the phosphotransacetylase (Pta)–acetate kinase (AckA) pathway that converts acetyl-coenzyme A (acCoA) to acetate via an acetyl phosphate (acP) intermediate. Pi, inorganic phosphate; CoASH, coenzyme A; TCA, tricarboxylic acid cycle.

Here, we present evidence that the status of the intracellular acP pool affects the expression of flagella, type 1 pili, colanic acid and certain stress effector genes and influences the ability of Escherichia coli cells to form biofilms.


Identification of acP-responsive genes

To identify genes that respond to acP, we used DNA macroarrays to compare E. coli wild-type cells (strain AJW678) with cells lacking AckA (ackA mutant; strain AJW1939) or both Pta and AckA (pta ackA double mutant; strain AJW2013). Cells that lack AckA can convert acCoA to acP but cannot degrade acP to acetate efficiently (Fig. 1). Therefore, these cells accumulate acP to a higher level than do wild-type cells, especially when grown on a carbon source that feeds into the bottom half of the glycolytic pathway (e.g. serine present in tryptone). In contrast, cells that lack Pta and AckA do not synthesize acP from either acCoA or acetate (Fig. 1) (McCleary and Stock, 1994; Prüß and Wolfe, 1994). The extracellular acetate concentrations and intracellular acCoA and acP concentrations associated with the single pta mutant resemble those of the double pta ackA mutant (Prüß and Wolfe, 1994). Thus, we chose the double mutant over the single mutant to avoid AckA-dependent activation to acP of any trace quantities of acetate that may be generated by non-Pta-dependent mechanisms.

We aerated cells in tryptone broth (TB) at 37°C. Wild-type cells grew about 40% more rapidly (doubling time = 32 min) than either of the mutants. However, the mutants doubled at almost identical rates (44 and 46 min respectively). Wild-type cells evolved ammonia and consumed amino acids in the following preferential order: l-serine, l-aspartate, l-tryptophan, l-glutamate, l-threonine, l-alanine, as reported previously (Prüßet al., 1994). Both mutants evolved ammonia and consumed amino acids preferentially in a manner that resembled that of wild-type cells, but with some significant differences: most notably, slower rates of consumption for tryptophan, glutamate and threonine. However, both mutants consumed amino acids in the same order at comparable rates and evolved ammonia similarly (data not shown). These similar behaviours gave us confidence to compare the ackA and pta ackA mutants directly.

We harvested cells during early exponential growth (A590 = 0.1), when acP levels peak in wild-type cells (Prüß and Wolfe, 1994). We purified the RNA, generated cDNA, hybridized a paired set of macroarrays (Sigma-GenoSys) and, for each sample, analysed and processed the data as described previously (Conway et al., 2002). We normalized the data by expressing the intensity of each gene-specific spot as a percentage of the sum of spot intensities, calculated the ratio of gene expression for each mutant relative to the control expression level (wild- type) and used Student's t-test to determine the probability that the ratio is significant (P-value). Our statistical analysis established significantly regulated genes meeting two criteria: ratio values ≥ 2.5 standard deviations from the mean of the log ratios (99% confidence) with P-values ≤ 0.05 (95% probability). A total of 146 genes [≈ 3.5% of 4290 total open reading frames (ORFs) identified in E. coli] met these criteria. We performed subsequent analyses only on this subset of genes.

To distinguish genes that respond to the status of acP from those that respond to a slower rate of growth or to a general defect in acetate excretion, we plotted the log10 of the expression ratio pta ackA/ WT against the log10 of the expression ratio ackA/ WT (Fig. 2A). Using this strategy, genes that respond identically in both mutants when compared with their wild-type parent map to a line with a slope of 1 and a y-intercept of 0. Presumably, such genes respond either to a slower rate of growth or to a general defect in the Pta-AckA pathway, e.g. the accumulation of the pathway substrate (acCoA) or the absence of its product (acetate). In contrast, genes that respond negatively to acP levels map above and to the left of the line, whereas those that respond positively to acP map below and to the right of that line.

Figure 2.

Scatter plots displaying genes significantly overexpressed or underexpressed relative to wild-type cells (AJW678) by mutant cells defective for pta ackA (AJW2013) or ackA (AJW1939). Cells were aerated in TB at 37°C, harvested at A590 = 0.1 and subjected to DNA macroarray analysis. The dotted line (m = 1, b = 0) represents the predicted position of genes identically expressed in cells of both mutants.
A. All genes that are expressed differentially.
B. Of the genes shown in (A), only those negatively influenced by acP are shown. Numbers refer to those in Table 1.
C. Of the genes shown in (A), only those positively influenced by acP are shown. Open squares, genes involved in stress responses. Chequered squares, genes involved in type I pilus biosynthesis and assembly. Cross-hatched squares, genes involved in capsule biosynthesis. Grey squares, hypothetical genes and those of putative function. Numbers refer to those in Table 2.

Under the experimental conditions tested, 50 genes appeared to respond negatively to high acP levels; those that responded most significantly (≥5 SD) function exclusively in flagellar biosynthesis and assembly (Fig. 2B, Table 1). These include genes involved in assembly of the filament and hook (fliC, flgE, fliD, flgK, flgD, flgL), the basal body (flgC, flgB, flgG, flgA, flgF, fliL, flgJ), the switching mechanism (fliM), the chemotactic signal transduction pathway (cheY, tar) and transcription of those genes (fliA). We confirmed these results, using quantitative reverse transcription polymerase chain reaction (RT-PCR) to analyse fliA and fliM (Fig. 3). Both genes were expressed at highest levels in the pta ackA mutant, followed by the wild type and then the ackA mutant, consistent with the array results. Other flagellar genes responded significantly, but to a lesser degree (≥2.5 SD): they include those involved in hook length control (fliK), flagellar-specific export (flgN, flhB, fliI, fliJ), chemotactic signalling and energy transduction (cheB, cheZ, cheR, trg, motB), regulation (flhD, flhC) and unknown functions (fliY, fliZ). Non-flagellar genes that responded negatively to high acP levels encode outer membrane porins (ompF, ompC), other proteins associated with or predicted to associate with the envelope (rbsB, b1966, yqiH, glpD, rfbX, rbsD), chaperones (dnaK, mopB, mopA), the regulators of the maltose regulon (malT) and the carnitine operon (caiF) and an enzyme involved in carnitine metabolism (caiD) (data not shown, see http:www.ou.edumicroarray).

Table 1. . Genes significantly upregulated in a pta ackA mutant compared with an ackA mutant .ab
 Geneb no. pta ackA/ackA pta ackA/WT ackA/WT Gene product
  • a . Numbers refer to those in Fig. 2B.

  • b

    . Significance determined as having a ratio

  • 5 SD from the mean of the ratios and having a P-value

  • 0.05. Another 33 genes not listed exhibited a significant ratio between 2.5 and 5 SD from the mean of the ratios and had a P-value

  • ≤0.05.

 1 fliC 192330.91.511.81.1−2.6−0.4Flagellin, major filament component
 2 flgE 1076 8.91.0 4.00.6−2.3−0.4Hook
 3 fliD 1924 5.80.8 4.10.6−1.4−0.2Filament cap
 4 flgC 1074 5.50.7 3.10.5−1.8−0.3Basal body-rod
 5 flgB 1073 5.40.7 2.60.4−2.1−0.3Basal body-rod
 6 flgK 1082 4.10.6 3.20.5−1.3−0.1Hook–filament junction
 7 flgG 1078 4.10.6 2.50.4−1.7−0.2Basal body-rod
 8 flgD 1075 3.60.6 2.20.3−1.6−0.2Hook assembly initiation
 9 flgL 1083 3.40.5 2.40.4−1.4−0.2Hook–filament junction
10 fliA 1922 3.40.5 1.90.3−1.8−0.3Flagellar-specific sigma factor
11 fliL 1944 3.10.5 1.40.1−2.3−0.4Basal body-rod
12 flgA 1072 2.70.4 1.60.2−1.7−0.2Basal body assembly
13 flgF 1077 2.50.4 1.80.3−1.4−0.1Basal body-rod
14 fliM 1945 2.50.4 1.50.2−1.7−0.2Motor switch; energy transduction
15 cheY 1882 2.30.4 signalling
16 tar 1886 2.20.3 1.90.3−1.1−0.1Chemoreceptor
17 flgJ 1801 2.10.3 1.80.3−1.2−0.1Basal body-rod
Figure 3.

Mean log expression ratios ± SEM of a subset of genes. Cells were aerated in TB at 37°C, harvested at A590 = 0.1 and subjected to quantitative RT-PCR.

In contrast, 46 genes appeared to respond positively to high acP concentrations; the most significant (≥4.8 SD) function primarily in stress protection (osmB, ivy, osmY, osmC, dps, hslJ), colanic acid biosynthesis (rcsA, wcaE, gmd), type 1 pilus assembly (fimG, fimI, fimA, fimC, fimH, fimF) and the assembly of putative non-type 1 pili (b0943, sfmC) (Fig. 2C, Table 2). We confirmed these results, using quantitative RT-PCR to analyse osmY, osmB, osmC, fimA, fimG, rcsA, gmd and wcaE (Fig. 3). These genes were expressed at highest levels in the ackA mutant, followed by the pta ackA mutant and then by the wild-type parent, consistent with the array results. Apparently, these genes respond not only to acP but also, to a lesser degree, to slowed growth or a general pathway defect. Other genes responded significantly, but to a lesser degree (≥2.5 SD): among these are two other genes involved in colanic acid biosynthesis (wcaD, ugd) and one of unknown function located within the colanic acid biosynthetic locus (orf1.3); also included are genes that encode an outer membrane protein (ompX), a putative regulator of maltose metabolism (sfsA), a component of formate dehydrogenase-H (fdhF) and glutathione oxidoreductase (gor) (data not shown, see http:www.ou.edumicroarray).

Table 2. . Genes significantly upregulated in an ackA mutant compared with a pta ackA mutant. ab
 Geneb no. pta ackA/ackA pta ackA/WT ackA/WT Gene product
  • a . Numbers refer to those in Fig. 2C.

  • b

    . Significance determined as having a ratio

  • 4.8 SD from the mean of the ratios and having a P-value

  • 0.05. Another 21 genes not listed exhibited a significant ratio between 2.5 and 4.8 SD from the mean of the ratios and had a P-value

  • ≤0.05.

 1 osmB 1283−10.9−1.0−1.1<−0.0510.11.0Osmotically inducible lipoprotein
 2 ivy 0220−5.5−0.7  2.00.310.81.0Inhibitor of vertebrate lysozyme, previously yfkE
 3 osmY 4376−4.6−0.7  1.80.3 8.20.9Hyperosmotically inducible periplasmic protein
 4 yjbE 4026−3.6−0.6  1.0<0.05 3.60.6Hypothetical protein
 5 rcsA 1951−3.5−0.6  1.0<0.05 3.60.6Colanic acid biosynthesis (CAB), positive regulator
 6 wcaE 2055−3.5−0.5−1.1<−0.05 3.30.5CAB, putative glycosyl transferase
 7 yafP 0234−3.4−0.5  1.20.1 4.10.6Hypothetical protein
 8 gmd 2053−3.4−0.5−1.1−0.1 3.00.5CAB, GDP-d-mannose dehydratase
 9 b0943 0943−3.0−0.5−1.30.1 2.40.4Putative fimbrial protein
10 sfmC 0531−2.7−0.4  1.30.1 3.40.5Putative fimbrial chaperone
11 fimG 4319−2.7−0.4  1.80.3 4.90.7Type 1 pilus, morphology
12 b1172 1172−2.6−0.4  1.30.1 3.50.5Hypothetical protein
13 osmC 1482−2.6−0.4  1.40.1 3.70.6Osmotically inducible protein
14 ycfJ 1110−2.4−0.4  1.0<0.05 2.60.4Hypothetical protein
15 dps 0812−2.3−0.4  3.50.6 8.30.9Starvation, global regulator
16 yiaD 3552−2.3−0.4−1.3−0.1 1.90.3Putative outer membrane protein
17 hslJ 1379−2.3−0.4  1.40.1 3.10.5Heat shock protein
18 fimI 4315−2.3−0.4  1.60.2 3.70.6Type 1 pilus, minor protein
19 fimA 4314−2.2−0.3  1.70.2 3.70.6Type 1 pilus, major subunit, pilin
20 ybdG 0577−2.2−0.3−1.1<−0.05 1.90.3Putative transport protein
21 fimC 4316−2.1−0.3−1.1<−0.05 2.00.3Type 1 pilus, periplasmic chaperone
22 fimH 4320−2.1−0.3  1.50.2 3.20.5Type 1 pilus, d-mannose-specific adhesin
23 b2833 2833−2.1−0.3  1.1<0.05 2.20.3Hypothetical protein
24 fimF 4318−2.0−0.3  1.40.2 2.90.5Type 1 pilus, morphology
25 eco 2209−2.0−0.3−1.1<−0.05 1.90.3Ecotin, serine protease inhibitor

Twenty-seven genes appeared to respond positively to a slower growth rate or a general pathway defect; these genes function primarily in stress protection and the TCA cycle (Fig. 2A, see http:www.ou.edumicroarray). In contrast, 23 genes responded negatively; most remain hypothetical or putative, whereas others do not fall into clearly delineated functional categories (Fig. 2A, see http:www.ou.edumicroarray).

These results agree with our previous report that pta ackA mutants synthesize far more flagella than ackA mutants (Prüß and Wolfe, 1994). As predicted from our array data, ackA mutants produced mucoid colonies. In contrast, wild-type cells and pta ackA mutants produced non-mucoid colonies (Fig. 4A). Also as predicted, ackA mutants exhibited numerous pili, as do wild-type cells. In contrast, pta ackA mutants displayed few pili (Fig. 4B). At first glance, it might be puzzling that pta ackA mutants display substantially fewer pili than wild-type cells. Like ackA mutants, pta ackA mutants express all the fim structural genes (fimAIFGH) at higher levels than wild-type cells (Table 2, Fig. 2C). We suspect the difference lies with the behaviour of fimC, which encodes the type 1 pilus-specific chaperone that delivers pilus subunits to the pilus-specific usher, an assembly platform encoded by fimD (Sauer et al., 2000). Unlike ackA mutants, pta ackA cells apparently do not upregulate fimC (Table 2, Figs 2C and 5). This result leads us to suspect that an excess of pilus subunits might overload the FimC chaperone and that this overload results in reduced pilus assembly. If so, we do not suspect the usher, FimD, because both ackA and pta ackA mutants appear to express similar amounts of fimD (Fig. 5). Regardless of the underlying mechanism responsible for the pta ackA pilus assembly defect, taken together, our results support the argument that high acP levels enhance colanic acid production and type 1 pilus assembly, whereas low acP levels favour flagellar expression.

Figure 4.

A. Mucoidy phenotype of colonies formed by cells wild type (AJW678) or deficient for either ack (AJW1939) or pta and ackA (AJW2013). Cells were grown at 37°C on defined medium (M63) supplemented with 0.2% glucose.
B. Transmission electron micrographs of wild-type cells (AJW678), ackA mutant cells (AJW1939) or pta ackA mutant cells (AJW2013) grown at 37°C in TB and harvested at A590 = 0.4. Each bar represents 1 µm.

Figure 5.

Schematic showing expression profiles of the fim operon. P/A, log10(pta ackA/ackA); P/+, log10(pta ackA/WT); A/+, log10(ackA/WT). NS, not significantly different.

Pellicle and abiotic surface-adherent biofilm formation

Escherichia coli cells can simultaneously form two different structured communities during growth in static culture: a pellicle that forms on the surface of the culture medium and an abiotic surface-adherent biofilm that forms at the air–liquid interface. Formation of the pellicle depends, in part, upon the presence of intact type 1 pili that permit cell–cell interactions (Harris et al., 1990). As others have previously implicated type 1 pili and other adhesins, flagella, colanic acid and several stress responses in the formation of surface-adherent biofilms (reviewed by Stoodley et al., 2002), we compared the pellicle- and surface-adherent biofilm-forming ability of wild-type cells with that of cells that either accumulate acP (ackA) or do not synthesize it (pta ackA). For these experiments, we grew cells at 37°C in static TB cultures. At 2 h intervals, we monitored the A590 as a measure of biomass. Wild-type cells grew more rapidly (doubling time = 56 min) than either of the mutants. The mutants, however, doubled at identical rates (doubling time = 121 min). Wild-type cells consumed amino acids and evolved ammonia in a manner that resembles that observed under aerated conditions (data not shown). Although they did so more slowly than their wild-type parent, both mutants consumed amino acids and evolved ammonia similarly (data not shown). Biofilm development requires growth through exponential phase (Danese et al., 2001). Thus, the difference between wild-type and mutant growth rates may make direct comparison between wild-type and mutant biofilm development problematic. However, the virtually identical rates of growth and ammonia evolution and the similar profiles of amino acid consumption exhibited by the mutants permits their direct comparison.

To observe the ability of each strain to form a surface-adherent biofilm at the liquid–air interface, we stained 18 h cultures with crystal violet, and rinsed away non-adherent cells (Fig. 6A). The morphology of each biofilm appeared to be distinctly different. Both wild-type and pta ackA mutant cells produced substantial surface-adherent biofilms in contrast to ackA mutant cells, which constructed a sparse biofilm. To examine these biofilms more closely, we performed differential interference contrast (DIC) microscopy on biofilms grown on thin strips of microscope slides (Fig. 6B). After 6 h incubation, wild-type cells produced large microcolonies. In contrast, pta ackA mutants formed smaller microcolonies, whereas ackA mutants formed even smaller ones.

Figure 6.

A. Surface-adherent biofilms formed by cells wild-type (WT) or deficient for either ackA or pta ackA. Cells were grown without aeration at 37°C in TB. After 18 h incubation, the triplicate cultures were stained with crystal violet, the non-adherent cells were rinsed away, and the adherent cells that formed a biofilm at the air–liquid interface were visualized using an Alphaimager 2000 documentation and analysis system. Representative images are shown.
B. DIC micrographs of microcolonies formed by wild-type cells (AJW678), ackA mutant cells (AJW1939) or pta ackA mutant cells (AJW2013). Duplicate cultures were grown without aeration for 6 h at 37°C on thin strips of microscope slides partially immersed in TB. Representative images are shown.
C. Pellicles formed by cells wild type (AJW678) or defective for either ackA (AJW1939) or pta ackA (AJW2013). Cells were grown without aeration at 37°C in TB. After 18 h incubation, the triplicate cultures were photographed using a Wild dissecting microscope at a magnification of 25×. Representative images are shown. Note cracks at the bottom left of the wild-type pellicle and the upper right of the pta ackA pellicle.

The morphology of each pellicle also appeared to be distinctly different. Wild-type cells formed a film-like pellicle that covered the entire surface of the static culture (Fig. 6C). Upon this film, complex swirls of colony-like structures formed. When vortexed, the wild-type pellicle did not disperse easily; instead, it formed tight aggregates (data not shown). In contrast, ackA mutant cells formed isolated rafts of cells. When vortexed, these cells also formed aggregates (data not shown). Like wild-type cells, pta ackA mutant cells formed a film that covered the entire surface of the culture. In contrast to wild-type and ackA mutant pellicles, pta ackA mutant cells formed few colony-like structures and dispersed easily into a homogeneous suspension when vortexed (data not shown).

Finally, we began to explore the molecular basis that underlies the distinct characteristics displayed by biofilms and pellicles formed by ackA and pta ackA mutants. We introduced a mutant allele (fliA) that precludes the expression of flagella into cells either wild type or deficient for ackA alone or both pta and ackA. As reported previously (Pratt and Kolter, 1998), the lack of flagella eliminated the ability of otherwise wild-type cells (fliA) to form a biofilm (Table 3). Similarly, pta ackA mutants that cannot synthesize flagella (pta ackA fliA) formed, at best, very poor biofilms. In contrast, ackA fliA mutants formed biofilms indistinguishable from those produced by their ackA fliA+ parent. Thus, biofilm formation by the ackA mutant does not require flagella.

Table 3. . Biofilm phenotypes.
StrainRelevant genotypePhenotypea
  1. a . For images of phenotypes, see Fig. 6.

  2. 0, no biofilm.

AJW678Wild typeWild-type standard
AJW1939 ackA::Km ackA standard
AJW2067 ackA::TnphoA′-2 ackA standard
AJW2013Δ(ackA pta hisJ hisP dhu) pta ackA standard
AJW2145 fliA::Tn50
AJW2149 ackA::Km fliA::Tn5 ackA standard
AJW2153Δ(ackA pta hisJ hisP dhu) fliA::Tn5Trace
AJW2070 ackA::TnphoA′-2 ΔfimA::Km0
AJW2064Δ(ackA pta hisJ hisP dhu) ΔfimA::Km0
AJW2061 fimH::Km0
AJW2069 ackA::TnphoA′-2 fimH::Km0
AJW2062Δ(ackA pta hisJ hisP dhu) fimH::Km0

The tendency of wild-type and ackA mutant biofilms to resist dispersal by agitation correlates well with their expression of numerous pili, surface structures intimately involved in cell–cell contact. In contrast, the easy dispersal of pta ackA mutant biofilms reflects their relative lack of pili. Both mannose and its analogue methyl-α-d-mannopyranoside inhibit wild-type pellicle formation by binding a d-mannose-specific adhesin, encoded by fimH and used for cell–cell contact (Harris et al., 1990). Both inhibitors interfered with the formation of wild-type and ackA mutant pellicles and surface-adherent biofilms (Table 4), but not with those formed by pta ackA mutants. However, when we introduced mutant alleles that eliminate the expression of the entire type 1 pilus (Δfim) or simply the adhesin (fimH), cells formed neither pellicles nor surface-adherent biofilms, regardless of the status of ackA and/or pta (Table 3). Taken together, our observations suggest that type 1 pili must play some important role in biofilm development by pta ackA mutants, but that the small number they exhibit cannot resist dispersal.

Table 4. . Effect of mannose and a mannose analogue on pellicle formation. a
  1. a . Cells were grown in TB at 37 °C in the absence or presence of mannose or methyl- α- d-mannopyranoside (MADMP) across a range of concentrations (mM). After 18 h incubation, pellicle and surface-adherent (SA) biofilm formation were evaluated. The values (mode of at least three independent experiments each performed in triplicate) indicate the break-point concentration: that which resulted in the loss of biofilm formation (A) or most dramatic difference in morphology (B).

  2. ND, not done.

Wild type17.5 (A)17.5 (A)20 (B)50 (A)
ackA 17.5 (A)ND20 (A)30 (A)
pta ackA >100>100>100>100


Array analysis predicts that acP mediates biofilm initiation

DNA macroarray experiments designed specifically to identify acP target genes provide evidence that acP (i) positively regulates genes required for type 1 pilus assembly, colanic acid biosynthesis and protection against certain stresses and (ii) negatively regulates genes required for flagella assembly (see also Prüß and Wolfe, 1994). Each of these acP-responsive structures or processes contributes to pellicle or surface-adherent biofilm formation, developmental programmes that require these structures and processes to be available in the proper amount and in the appropriate order (Stoodley et al., 2002). We therefore tested whether mutants defective in their ability to metabolize acP could construct normal biofilms, and found that they could not.

acP metabolism mutants form aberrant biofilms

Both the ackA mutant that accumulates acP and the pta ackA mutant that cannot synthesize acP clearly form biofilms; however, these biofilms exhibit morphological and physical properties distinctly different from each other and their wild-type parent. Whereas the ackA mutant forms a poor surface-adherent biofilm composed of small microcolonies and a pellicle composed of isolated colony-like structures, the pta ackA mutant forms a more abundant surface-adherent biofilm and a rather featureless, film-like pellicle. Intriguingly, wild-type cells form pellicles that appear to incorporate both characteristics: ackA-like structures that sit upon a pta ackA-like film. Whereas the pta ackA mutant pellicle disperses easily upon agitation, the wild-type and ackA mutant biofilms resist dispersal. This makes sense, as type 1 pili contribute to cell–cell contact, pta ackA mutants appear to be defective in pilus assembly, and inhibition of the mannose-specific FimH adhesin exerts no great influence upon the formation of pta ackA mutant biofilms. However, pili must contribute to some degree, because their complete absence or simply that of FimH completely eliminates pta ackA mutant biofilm formation. In contrast, flagella clearly play a key role in the formation of biofilms by the hyperflagellated pta ackA cells and apparently an insignificant role in biofilm formation by the already poorly flagellated ackA cells. Taken together, these results suggest that pta ackA and ackA mutants form aberrant biofilms via somewhat different developmental pathways that rely to varying degrees upon structures (i.e. type 1 pili and flagella), the expression of which appears to be responsive to the status of the acP pool. In contrast, wild-type biofilm formation depends on all three structures (Stoodley et al., 2002).

acP as a signal of nutritional status

What causes the pta ackA and ackA mutants to synthesize flagella, type 1 pili and capsule differently enough to construct biofilms with distinctly different attributes? The characteristic phenotypes of both mutants did not vary under a number of laboratory conditions. For example, both mutants exhibit their characteristic phenotypes when grown in either TB or LB at either 37°C or 30°C (J. E. Seitz-Partridge and A. J. Wolfe, unpublished). When backcrossed to its wild-type parent, each mutant exhibits its characteristic behaviour (J. E. Seitz-Partridge and A. J. Wolfe, unpublished). These observations argue against a role for a temperature-sensitive structure or process or some unknown, uncharacterized mutation that might be influenced by media composition. Both mutants cannot recycle CoASH by the preferred method (i.e. the Pta-AckA pathway), both excrete little or no acetate, and both only modestly acidify their environment before evolving large amounts of ammonia (B. M. Prüß and A. J. Wolfe, unpublished). Thus, we must look elsewhere for the underlying cause of the mutants’ morphological and physical properties. The status of their acP pool stands out as the most obvious difference between the two mutants: whereas pta ackA mutants cannot synthesize acP, ackA mutants accumulate it in excess (McCleary and Stock, 1994; Prüß and Wolfe, 1994).

Its central role in intermediary carbon metabolism positions acP suitably to monitor nutritional status. The acP pool rises rapidly during exponential growth in an aerated culture replete with nutrients; the pool falls as the cells deplete those nutrients and begin their transition to stationary phase (Prüß and Wolfe, 1994). Direct measurement of acP in cells that have entered stationary phase provides evidence that the acP pool rises again after cells enter stationary phase (McCleary and Stock, 1994). This increase during stationary phase may occur in response to starvation and/or the lack of oxygen. The glucose starvation response, which includes increased levels of Pta and decreased levels of TCA cycle enzymes, depends upon Pta but not AckA, an observation that argues that acP helps to induce that response (Nystrom, 1994), presumably via ArcA, a two-component response regulator (Nystrom et al., 1996). A similar response occurs as cells transit from aerobic to anaerobic conditions that result in fermentation (Nystrom, 1994). Thus, it appears that acP concentrations rise in the absence of a functional TCA cycle, e.g. during starvation or fermentation, or whenever the accumulation of acCoA exceeds the TCA cycle's capacity, e.g. during rapid growth. In contrast, the acP pool diminishes whenever the acCoA pool does not exceed its capacity, e.g. during the transition from exponential growth to stationary phase.

acP as a modulator of biofilm development

These reported observations pertain to cells aerated in shake flask cultures. However, we have reason to believe that a similar sequence of events occurs during growth in static culture. First and foremost, in both aerated and static culture, pta ackA and ackA mutant cells exhibit strikingly similar growth, amino acid consumption and ammonia evolution profiles. Secondly, initiation of wild-type biofilm formation occurs before 6 h (Fig. 6B), an interval that coincides with the beginning of the transition phase. Assuming that wild-type cells grown in static culture co-ordinate metabolic and physiological events similarly to those grown in aerated culture, we expect that the acP pool decreases and flagella expression increases during the interval just before the transition phase. Later, as the local extracellular environment becomes increasingly anaerobic and depleted of nutrients, we predict that the acP pool increases with a concomitant increase in type 1 pilus and colanic acid biosynthesis. Thus, acP could serve as a signal to initiate pellicle and surface-adherent biofilm formation by modulating an orderly switch between the flagella-dependent, the pili-dependent and the capsule-dependent stages of biofilm formation, as has been established for wild-type biofilm development (Stoodley et al., 2002).

acP as a phosphodonor for response regulators

acP can act as a phosphodonor for some response regulators (McCleary et al., 1993; Wanner, 1995). Thus, we find it tempting to hypothesize that acP influences biofilm formation by acting as a phosphodonor for response regulators (FimZ, OmpR and RcsB) that control certain biofilm-associated genes. Phospho-FimZ and phospho-OmpR repress flagellar synthesis (Shin and Park, 1995; Clegg and Hughes, 2002), whereas phospho-FimZ activates type 1 pilus expression (Clegg and Hughes, 2002). Phospho-RcsB activates colanic acid biosynthesis (Stout, 1994) and at least one acP-responsive stress effector (osmC) (Davalos-Garcia et al., 2001).

Implications for biofilm development in the wild type

The results of DNA macroarray experiments designed to identify acP-responsive genes argue that the status of the acP pool modulates the expression of cellular components required for the normal development of biofilms. We envisage that acP modulates the initial steps of biofilm formation by linking the expression of flagella, pili and colanic acid to the nutritional status of the culture, presumably by means of several two-component response regulators. We expect that wild-type planktonic cells, experiencing the transition from exponential growth to stationary phase, possess a small acP pool, assemble several flagella and some pili and synthesize little colanic acid. Once these wild-type cells make contact with a suitable surface or with each other, however, we predict that the subsequent local depletion of nutrients and oxygen will cause intracellular acP levels to rise. In response, flagella synthesis will diminish, and first pili and then colanic acid biosynthesis will increase. Clearly, the dynamic relationships between nutritional environment, acP pool, acP-responsive signal transduction pathways and the targets of those pathways warrant further study.

Experimental procedures

Bacterial strains

All strains used in this study were derivatives of E. coli K-12 and are listed in Table 5.

Table 5. . Bacterial strains used in this study.
StrainRelevant genotypeSource or reference
AJW678 thi-1 thr-1(Am) leuB6 metF159(Am) rpsL136ΔlacX74 Kumari et al. (2000)
AJW1939AJW678 ackA::Km Kumari et al. (2000)
AJW2067AJW678 ackA::TnphoA′-2(P1)CP757→AJW678 (Tc/Ace–)
AJW2013AJW678 Δ(ackA pta hisJ hisP dhu) zej223-Tn10(P1)CP911→AJW678 (Tc/Ace–)
AJW2061AJW678 fimH::Km(P1)ORN133→AJW678 (Km)
AJW2069AJW678 ackA::TnphoA′-2 fimH::Km(P1)ORN133→AJW2067 (Km)
AJW2062AJW678 Δ(ackA pta hisJ hisP dhu) zej223-Tn10 fimH::Km(P1)ORN133→AJW2013 (Km)
AJW2063AJW678 ΔfimA::Km(P1)ORN172→AJW678 (Km)
AJW2070AJW678 ackA::TnphoA′-2 ΔfimA::Km(P1)ORN172→AJW2067 (Km)
AJW2064AJW678 Δ(ackA pta hisJ hisP dhu) zej223-Tn10ΔfimA::Km(P1)ORN172→AJW2013 (Km)
AJW2145AJW678 fliA::Tn5(P1)MG1665 fliA::Tn5→AJW678 (Km)
AJW2149AJW678 ackA::Km fliA::Tn5(P1)AJW2145→AJW1939 (Km)
AJW2153AJW678 Δ(ackA pta hisJ hisP dhu) zej223-Tn10 fliA::Tn5(P1)AJW2145→AJW2013 (Km)
CP750 thi-1 thr-1(Am) leuB6 metF159(Am) rpsL136 Shin and Park (1995)
CP757CP750 ackA::TnphoA′-2C. Park (KAIST, Korea)
CP911CP750 Δ(ackA pta hisJ hisP dhu) zej223-Tn10 Prüßet al. (1994)
MG1655 fliA::Tn5 fliA::Tn5F. Blattner (University of Wisconsin, Madison, WI, USA)
ORN133 fimH::KmP. Orndorff (North Carolina State University)
ORN172ΔfimA::KmP. Orndorff

Culture conditions

Cells were grown aerobically with 250 r.p.m. agitation at 37°C in tryptone broth [TB; 1% (w/v) tryptone and 0.5% (w/v) sodium chloride] in 250 ml fleakers (Corning). Cell growth was monitored spectrophotometrically (Beckman Instruments DU530) at 590 nm (A590) by removing 100 µl samples without interrupting culture agitation; no more than 5% of the culture was removed before harvesting cells for RNA isolation. Cultures were inoculated to an initial A590 of < 0.001 to ensure that the cells experienced at least 10 generations in the medium before reaching steady-state growth.

Amino acid analysis

Amino acid analysis of culture media was performed as described previously (Prüßet al., 1994).

RNA isolation

Cells for RNA isolation were harvested from the early exponential phase of growth; 10 ml of culture was pipetted into 10 ml of ice-cold RNA-Later (Ambion). Cells were removed from this mixture by centrifugation (8730 g for 20 min). RNA was isolated and purified using an RNeasy mini kit (Qiagen). Genomic DNA was removed after purification by DNase I (Ambion) treatment, with subsequent RNA repurification with an RNeasy column. RNA samples were quantified by 260 nm absorbance (Beckman DU530 spectrophotometer).

Labelled target synthesis

Labelled targets for hybridization were generated as described previously (Tao et al., 1999) with slight modifications. The initial preincubation (90°C for 2 min, then 42°C for 20 min in an MJ Research MiniCycler) was performed in 0.2 ml PCR tubes containing 1 µg of RNA template, dATP, dCTP and dGTP (0.33 mM each final concentration) and C-terminal primers (Sigma Genosys Biotechnologies) brought to a final volume of 23 µl with RNase-free dH2O. After preincubation, a labelling mix {20 µCi [α-33P]-dCTP (2000–3000 Ci mol−1, New England Nuclear), 200 units of Superscript II RNase H reverse transcriptase (BRL), 10 mM dithiothreitol (DTT; BRL) and 10 units of ribonuclease inhibitor (BRL)} was added, bringing the final volume to 30 µl. The cDNA reaction mixture was incubated for 2 h at 42°C. The labelled cDNA was denatured from the RNA template for 30 min at 65°C with 29 mM EDTA and 0.43 N NaOH. The denaturing reaction was neutralized with 160 mM Tris (pH 7.4) and 0.28 N HCl. The amount of radiolabel incorporation during labelling was determined by comparing scintillation counts of 1 µl of labelled cDNA (in 5.0 ml of scintillation fluid) before and after removal of unincorporated nucleotide by gel filtration through a G-50 Sephadex column (Sambrook et al., 1989). This procedure routinely supplied cDNA samples with 70–90% label incorporation.

Hybridization of arrays and image analysis

Panorama E. coli gene arrays (Sigma Genosys Biotechnologies), array prehybridization, hybridization and array washes were performed as described previously (Tao et al., 1999), with the following modifications. Before prehybridization, arrays were rinsed in 50 ml of 2× SSPE (1× SSPE is 0.18 M NaCl, 10 mM NaH2PO4 and 1 mM EDTA, pH 7.7), then prehybridized with 5 ml of hybridization solution (65°C for 1–4 h) and hybridized with 3.0 ml of hybridization solution supplemented with 400–600 µl of labelled cDNA probe (65°C for 16–18 h). The arrays were then wrapped in clear plastic food wrap (Saran Wrap) and exposed to a Kodak storage phosphor screen GP (Eastman Kodak) for 24 h. Longer exposure times were occasionally necessary to increase spot visibility and contrast. Exposed phosphor screens were visualized at a pixel density of 100 microns (10 000 dots cm−2) with a Storm 820 PhosphorImager (Molecular Dynamics). The array membranes were consecutively hybridized, stripped and rehybridized, a procedure that was repeated up to eight times. Arrays were stripped of hybridized probe by boiling the inverted array for 20 min in a microwave (100°C) in 200 ml of stripping solution (10 mM Tris, pH 7.5, 1 mM EDTA and 1% SDS). Phosphorimaging of a membrane array results in a TIFF image file that was processed for data analysis, as described previously (Conway et al., 2002). The raw spot intensities were represented in a row–column format and exported into Microsoft excel spreadsheets for further analysis.

Data analysis

The level of replication used in this study has been shown to provide robust statistical analysis and minimizes false positives (Conway et al., 2002). The raw data from each experimental replicate were exported to excel workbooks and processed as described previously (Conway et al., 2002). The duplicate spots on the arrays were treated as separate determinations. The raw data values were normalized by expressing each gene as a percentage of the sum of all the gene-specific spot intensities. The probability that the average of the experimental replicates is significantly different from the average of the control replicates was calculated by application of Student's t-test. The ratio of the experimental/control expression levels was calculated such that genes that are more highly expressed in the experimental condition are given a positive value and genes that are more highly expressed in the control condition are given a negative value. Statistical significance is established by consideration of two metrics, the P-value, which takes into consideration the uncertainty of the replicate measurements for both conditions, and the standard deviation of the mean of the log (base 10) ratios, which indicates the significance of the ratio. For consideration of genes affected by mutations resulting in altered intracellular concentrations of acP, we considered only those genes that were significantly up- or downregulated by> 2.5 standard deviations from the mean of the log ratios (99% confidence) with P-values ≤ 0.05 (95% probability).

Website and public access to analysis tools

A website has been constructed to permit users to download these analytical tools and protocols that describe their step-by-step implementation (http:www.ou.edumicroarray).

Quantitative RT-PCR

Real-time PCR was performed with RNA samples, prepared as described above. RNA was reverse transcribed using amino-allyl reverse transcription (http:www.cmgm.stanford.edupbrownprotocolsaadUTPCouplingProcedure.htm), and cDNA was purified. The PCR was performed with the SYBR green kit from PE Biosystems. The reaction mixture contained 100 ng of cDNA, 1× SYBR green buffer, 2.5 mM MgCl2, 0.25 mM each dNTP, 0.05 µM each primer, 0.01 U of AmpErase UNG and 1 U of Taq Gold polymerase. The reaction was performed with 50 cycles of 30 s at 94°C, 30 s at 55°C and 1 min at 72°C and monitored in an iCycler iQ real-time PCR detection system (Bio-Rad). A standard curve was derived from plasmid and used to convert threshold crossings to log copy numbers. Expression ratios were obtained by dividing copy numbers of wild-type cells by those of the mutants. All PCR fragments yielded a single band on agarose gel.

Biofilm and pellicle techniques

Cell growth was monitored spectrophotometrically (Beckman Instruments DU530) at A590. Cells were grown in microtitre wells without aeration at 37°C in TB. At 2 h intervals, the entire well was harvested. To ensure that all cells were harvested, the well was rinsed three times with fresh media. To ensure a homogeneous suspension, the cells were passed through a small bore tube connected at both ends to a syringe. Surface-adherent biofilms were stained with crystal violet as described previously (Pratt and Kolter, 1998) and visualized using an Alphaimager 2000 documentation and analysis system. Pellicles were photographed before harvesting, using a Wild (Heerbrugg) dissecting microscope.

Microscopic techniques

Light microscopy.

Glass strips cut from sterile microscope slides were suspended in static culture tubes containing 3 ml of TB so that the waterline bisected the strips. The tubes were inoculated and, at intervals during growth, strips were removed and examined along their length using a Nikon Microphot light microscope with differential interference contrast optics. Digital images were captured with a Spot RT digital camera.

Transmission electron microscopy.

Cells were grown in TB with aeration at 37°C to an A590 = 0.4. Whole cells placed on collodion-coated copper grids were negative stained by brief treatment with 2% uranyl acetate. Grids were observed and photographed on a Jeol 100CX transmission electron microscope.


We thank K. Visick, A. Driks, T. Gallagher and P. Orndorff for their advice and insight, and F. Blattner and P. Orndorff for strains. This work was supported by grants MCB-9630647 from the National Science Foundation and LU#11200 from the Loyola University Chicago Potts Foundation (A.J.W.) and grant RO1-AI-48945-01 from the NIH (T.C.).

Supplementary material

The following material is available from http:www.blackwellpublishing.comproductsjournalssuppmatmolemole3457mmi3457sm.htm

Table S1. Genes significantly upregulated in a pta ackA mutant compared to an ackA mutant.

Table S2. Genes significantly upregulated in an ackA mutant compared to a pta ackA mutant.

Table S3. Genes significantly upregulated in both mutants compared to wild type.

Table S4. Genes significantly downregulated in both mutants compared to wild type.

Fig. S1. A. A590 for wild-type cells (AJW678; □), or deficient for either ackA (AJW1939; ○) or pta ackA (AJW2013; ▵). Cells were grown without aeration at 37°C in TB.

B. Amino acid concentrations in the medium of cells deficient for ackA (AJW1939) or (C) for pta ackA (AJW2013). (▴) L-serine, (▪) L-aspartate, (▿) L-tryptophan, (□) L-glycine, (▾) L-glutamate, (•) L-threonine, (○) L-alanine. Amino acids that did not change significantly are not shown.

D. Ammonia concentrations in the medium of cells defective for ackA (AJW1939) (○) and pta ackA (AJW2013) (▵). Each point represents the mean of duplicate measurements from two independent experiments. The standard errors of the means were less than 20 µmoles l−1.