Indole as an intercellular signal in microbial communities

Abstract

Bacteria can utilize signal molecules to coordinate their behavior to survive in dynamic multispecies communities. Indole is widespread in the natural environment, as a variety of both Gram-positive and Gram-negative bacteria (to date, 85 species) produce large quantities of indole. Although it has been known for over 100 years that many bacteria produce indole, the real biological roles of this molecule are only now beginning to be unveiled. As an intercellular signal molecule, indole controls diverse aspects of bacterial physiology, such as spore formation, plasmid stability, drug resistance, biofilm formation, and virulence in indole-producing bacteria. In contrast, many non-indole-producing bacteria, plants and animals produce diverse oxygenases which may interfere with indole signaling. It appears indole plays an important role in bacterial physiology, ecological balance, and possibly human health. Here we discuss our current knowledge and perspectives on indole signaling.

Introduction

In most environmental niches, multiple bacterial species coexist as dynamic communities. Bacteria have developed intercellular signaling to adapt and survive in natural communities because of nutritional limitation, competition with other bacteria, and the host defense system. Many bacteria secrete small diffusible signal molecules to sense the local environmental conditions, including their own population, and to synchronize multicellular behaviors (Fuqua et al., 1994; Waters & Bassler, 2005; Keller & Surette, 2006; Hughes & Sperandio, 2008). A variety of intercellular signal molecules, such as the most studied N-acyl-homoserine lactones (AHLs) in Gram-negative bacteria, autoinducer 2 (AI-2) in both Gram-negative and Gram-positive bacteria, and signal peptides in Gram-positive bacteria, among others, have been discovered over the last 20 years (Waters & Bassler, 2005; Keller & Surette, 2006; Dong et al., 2007). The intercellular signal molecules coordinate the gene expression for bioluminescence, sporulation, plasmid conjugal transfer, competence, virulence factor production, antibiotic production, and biofilm formation (Fuqua et al., 1994; Comella & Grossman, 2005; Waters & Bassler, 2005; Nadell et al., 2008). There has been widespread research on new signal molecules and their signaling mechanisms in this area, and canonical signaling systems of AHL quorum sensing have been extensively reviewed elsewhere (Waters & Bassler, 2005; Keller & Surette, 2006; Dong et al., 2007; Williams et al., 2007). Recently, the concept of intercellular signaling has been broadened, and is no longer restricted to cell–cell communication such as quorum sensing (Monds & O'Toole, 2008; Ryan & Dow, 2008) or limited to intraspecies signaling, but now includes interspecies and interkingdom signaling (Bassler, 1999; Shiner et al., 2005; Diggle et al., 2007a; Hughes & Sperandio, 2008; Ryan & Dow, 2008). Among the bacterial signal molecules, indole has recently received much attention due to its diverse biological roles in several bacterial strains. This review focused primarily on indole signaling.

In 1897, reports showed that Bacillus coli (Escherichia coli) and Asiatic cholera (Vibrio cholerae) produced indole during a stationary cell growth phase (Smith, 1897), and the indole test has been regularly used as a diagnostic marker for the identification of E. coli (Smith, 1897; Wang et al., 2001). Table 1 shows a compilation from the literature of indole-producing organisms and organisms having homologues of the TnaA tryptophanase (Deeley & Yanofsky, 1981) from E. coli responsible for the biosynthesis of indole. The table demonstrates that a variety of both Gram-positive and Gram-negative bacteria (more than 85 species), including many pathogenic bacteria such as Bacillus alvei, pathogenic E. coli, several Shigella strains, Enterococcus faecalis, and V. cholerae, can produce indole.

Although the mechanism of indole biosynthesis in E. coli has been investigated over the past several decades (Newton & Snell, 1965; Botsford & DeMoss, 1971; Yanofsky et al., 1991), the real biological functions of indole have only recently started to be revealed. After Gerth et al. (1993) first suggested that indole was an autoinducer in Stigmatella aurantiaca (Gerth et al., 1993), many groups have reported diverse functions of indole. These include an extracellular signal in E. coli (Wang et al., 2001), drug resistance in E. coli (Hirakawa et al., 2005), plasmid stability in E. coli (Chant & Summers, 2007), virulence control in pathogenic E. coli (Anyanful et al., 2005; Hirakawa et al., 2009), and biofilm formation in E. coli (Di Martino et al., 2003; Lee et al., 2007b) and V. cholerae (Mueller et al., 2009). Owing to the limited information on indole-related phenotypes, indole was initially not considered to be a cell-to-cell signal molecule (Winzer et al., 2002). However, several lines of new evidence suggest that indole acts as an intercellular signal molecule (Chant & Summers, 2007; Lee et al., 2007a; Jayaraman & Wood, 2008; Monds & O'Toole, 2008; Ryan & Dow, 2008), which is discussed here in detail. Additionally, we address the fact that indole and AI-2 (one of the most studied signal molecules) have many similar as well as divergent characteristics in biosynthesis and function.

Interestingly, indole signaling is dynamic in multispecies communities. For example, indole decreases the cell growth of a fungus, Aspergillus niger (Kamath & Vaidyanathan, 1990), attenuates the virulence in Pseudomonas aeruginosa (Lee et al., 2009a), and increases drug resistance in Salmonella enterica (Nikaido et al., 2008), which cannot produce indole. Additionally, many non-indole-producing bacteria and eukaryotes encode various oxygenases that can modify/degrade indole and produce indole derivatives (Ensley et al., 1983; Rui et al., 2005; Wikoff et al., 2009). Indole derivatives are widely distributed in the human body, but their functions have not yet been revealed (Gillam et al., 2000; Crumeyrolle-Arias et al., 2008, 2009; Wikoff et al., 2009). Therefore, indole is more important than originally thought, and many non-indole-producing species have developed specific mechanisms to metabolize indole and interfere with indole signaling. This review also discusses the perspectives of current indole research to study its bacterial physiology, pathogenesis, and ecology.

TnaA and indole biosynthesis

In E. coli, indole is produced by tryptophanase (TnaA; EC 4.1.99.1), which can reversibly convert tryptophan into indole, pyruvate, and ammonia (Newton & Snell, 1965) (Fig. 1) in the tryptophan pathway in E. coli (Pittard, 1996; Lee et al., 2007b). However, no pathway for indole degradation is known for this bacterium (Chant & Summers, 2007). A comprehensive review of TnaA regarding its enzyme activity, structure, and mechanisms of action, can be found elsewhere (Snell, 1975). It has been long thought that bacteria utilize TnaA to synthesize tryptophan from indole as a carbon source (Gong & Yanofsky, 2002). However, the equilibrium of the reaction favors the production of indole from tryptophan (Tewari & Goldberg, 1994; Monds & O'Toole, 2008). In fact, the exogenous addition of indole (1–6 mM) does not increase the cell density because of its toxicity in E. coli (Chant & Summers, 2007). Escherichia coli, the most extensively studied organism for indole biosynthesis, uses several mechanisms (repression, transcriptional attenuation, and feedback inhibition) to regulate the expression of tryptophan operon (trpABCDE) and tna operon (tnaCAB) in the tryptophan metabolism (Fig. 1) (Yanofsky et al., 1991; Gong & Yanofsky, 2002; Lee et al., 2007b). In tryptophan-deficient conditions, the expression of the trp operon is elevated, whereas the expression of the tna operon consisting of TnaC (24 amino acid leader peptide, previously called TnaL), TnaA (tryptophanase), and TnaB (permease) is repressed, as transcription-terminating factor (Rho)-dependent termination occurs in the tna operon. As a result, the expression of TnaA and TnaB and indole production are repressed when the level of tryptophan is low (Yanofsky et al., 1991; Gong & Yanofsky, 2002). In tryptophan-rich conditions, Rho-dependent transcriptional termination is eliminated so that indole production is elevated. Hence, extracellular tryptophan and other amino acids directly influence indole production (Gong & Yanofsky, 2002). Additionally, three permeases (Mtr, TnaB, and AroP) play different roles in tryptophan transport in different environment compositions and thus can directly influence the levels of indole (Fig. 1) (Yanofsky et al., 1991; Gong & Yanofsky, 2002). The Mtr permease is principally responsible for transporting indole, and the TnaB permease is critical for tryptophan uptake (Yanofsky et al., 1991). Although the production of tryptophan is costly (Yanofsky et al., 1991), cells still utilize the tryptophan pathway to produce and secrete indole in large quantities.

Indole generated from tryptophan can be transported through cell membrane proteins (Fig. 1). For example, E. coli and V. cholerae can excrete indole up to 0.6 mM in a rich medium (Kobayashi et al., 2006; Lee et al., 2007b; Mueller et al., 2009). In E. coli, the efflux proteins AcrEF are partially responsible for exporting indole, as the indole excretion of the acrEF mutant was lower than that of its wild-type strain (Kawamura-Sato et al., 1999). The Mtr permease is primarily responsible for importing indole, as indole is not taken up by the mtr mutant (Yanofsky et al., 1991). However, it was recently suggested that indole could be directly diffused through the cell membrane due to its hydrophobic nature (Gaede et al., 2005). Hence, it is imperative to gain a clear understanding of how indole is imported and exported.

Indole-producing bacteria

Many Gram-positive and Gram-negative bacteria encode a single copy of the tnaA gene in their chromosome and produce indole (Table 1). Although most organisms contain the tryptophan biosynthesis pathway, to date, only bacteria encoding tnaA can synthesize indole, and so far, no eukaryotic cells have been shown to produce indole. A ncbi-blastp search was performed using the E. coli TnaA protein. blastp 2.2.20+ was used with a hit list size of 2000 and all of the other parameters were the default parameters in the program. Protein sequences showing <30% identity and the putative protein sequences of unknown function were discarded. As a result, more than 67 species contain TnaA homologues in a wide range of sequence identities (40–99% sequence identity with the E. coli TnaA in Table 1). Data available in the literature show that 54 bacterial species produce indole from these 67 species (Table 1). Further analysis of the literature shows that 31 more species produce indole, although the genomic sequence of the respective strains has not been reported (indicated as ‘NA’ in Table 1). In total, we found data supporting that, to date, 85 bacterial species have been shown to produce indole. Importantly, these include many pathogenic bacteria, for example Vibrio vulnificus (Dalsgaard et al., 1999), Haemophilus influenzae (Stull et al., 1995), Pasteurella multocida (Clemons & Gadberry, 1982), four Shigella strains (Rezwan et al., 2004), Klebsiella planticola (Liu et al., 1997), and Proteus vulgaris (DeMoss & Moser, 1969) (Table 1). Notably, several Gram-positive strains including B. alvei (Hoch & Demoss, 1965) and E. faecalis (Schleifer et al., 1984) also produce indole.

Intriguingly, several bacteria, such as Aeromonas salmonicida, Pseudovibrio sp., Shigella sonnei, Vibrio vulnificus and Yersinia kristensenii have lost the ability to synthesize indole, although these strains have a tnaA gene homologue in their chromosomes (Table 1). For example, unlike normal Shigella strains, some Shigella with point mutations, insertion, and/or deletion in the tna operon do not produce indole, possibly due to some adaptive advantage (Rezwan et al., 2004). Although speculative, some individuals may avoid the cost of producing indole by exploiting the signal from the local bacterial consortia in line with the kin selection model (Diggle et al., 2007b). Nevertheless, the main question remains as to why many bacteria produce indole in large quantities.

Environmental factors controlling indole biosynthesis

The accumulation of extracellular indole could be critically affected by environmental factors, such as the cell population, carbon sources, temperature, and pH. In fact, several lines of evidence support this possibility. First, the extracellular indole concentration is cell population density dependent where E. coli and V. cholerae start to produce indole in the early exponential phase. The concentration reaches a maximum level (up to 0.6 mM indole in a rich medium) in the stationary phase, and is stably maintained during the stationary phase (Kobayashi et al., 2006; Mueller et al., 2009). Secondly, in 1919, it was reported that glucose repressed indole biosynthesis (John & Wyeth, 1919). The catabolic repression of TnaA was confirmed because the transcription of tnaA gene was repressed during carbon limitation (Botsford & DeMoss, 1971). Additionally, the tnaAB operon was activated by catabolite regulation protein cyclic AMP complex in E. coli (Deeley & Yanofsky, 1982). Hence, E. coli produces a relatively large quantity of indole when its population is high and carbon source has dwindled (Fig. 1).

Temperature and pH are also important environmental factors that affect indole biosynthesis in E. coli (Fig. 1). The gene expression of tnaAB was induced in E. coli by temperature shifting from 30 to 43 °C (Li et al., 2003), but E. coli lost the ability of indole biosynthesis at 44.5 °C (Bueschkens & Stiles, 1984). Also, the effect of indole signaling was more significant at a lower temperature (30 °C) compared with 37 °C in the control of gene expression, biofilm formation, and antibiotic resistance in E. coli (Lee et al., 2008). Additionally, a low pH inhibits indole production in E. coli (John & Wyeth, 1919), and TnaA was one of the most induced proteins at pH 9.0 (Blankenhorn et al., 1999). Therefore, the environmental conditions such as cell population density, carbon source, temperature, and pH directly control the concentration of extracellular indole. It would be interesting to investigate whether the mechanism of indole biosynthesis in other indole-producing bacteria is similar to that in E. coli.

Functions of indole signal

Indole plays diverse biological roles in several bacteria including spore formation, drug resistance, virulence, plasmid stability, and biofilm formation (Fig. 1, Table 2). Historically, it was first speculated that indole was a possible autoinducer, as indole induced a spore formation of S. aurantiaca (myxobacteria, Table 1) in fruiting bodies (Gerth et al., 1993). The identity of indole as a possible extracellular signaling molecule was investigated in E. coli with an elegant genetic screen (Baca-DeLancey et al., 1999). As a result, four genes (astD, cysK, gabT, and tnaB) were shown to be activated by the accumulation of self-produced extracellular signals during the stationary phase (Baca-DeLancey et al., 1999). In a follow-up study, indole was confirmed as the extracellular signaling molecule required for the activation of astD, gabT, and tnaB (Wang et al., 2001).

More recently, it has been shown that indole increases drug resistance by inducing intrinsic xenobiotic exporter genes (mdtEF and acrD) in E. coli, where indole acts via two-component signal transduction systems (BaeSR and CpxAR) (Hirakawa et al., 2005) (Fig. 1). Hence, it is possible that these two-component signal systems can be used as indole sensors. The result corroborates another study in which indole induces the expression of spy (spheroplast protein Y) gene via BaeSR and CpxAR (Raffa & Raivio, 2002; Nishino et al., 2005). Furthermore, it has been suggested that GadX (AraC-type transcription factor), Hfq (global regulator of sRNA function), and RpoS (stress and stationary phase sigma S) are essential for indole-induced mdtEF expression (Kobayashi et al., 2006). Hence, indole may interact with a variety of global regulators.

Indole and the tnaA gene also affect the virulence of pathogenic bacteria. Tryptophanase activity is linked to the killing of nematodes by enteropathogenic E. coli, as tryptophanase activity is required for the full activation of the LEE1 promoter (Anyanful et al., 2005). Moreover, indole increases secretion of virulence-related EspA and EspB proteins (LEE4 gene products) and formation of attaching and effacing lesions in enterohemorrhagic E. coli (Hirakawa et al., 2009). In V. cholerae, indole and the tnaA gene increases grazing resistance to phagocytic eukaryote Dictyostelium discoideum, probably by inducing the virulence-associated secretion proteins (Mueller et al., 2009). Additionally, among isolates of H. influenzae, most serotypes (94–100%) are indole-positive, compared with only 70–75% of harmless isolates. The result indicates that indole production is necessary but not sufficient for virulence to this strain (Martin et al., 1998).

Indole enhances plasmid stability in E. coli (Chant & Summers, 2007). The study demonstrates that small noncoding RNAs from the E. coli plasmid ColE1 bind to TnaA and help preventing plasmid loss, and indole delays cell division (Chant & Summers, 2007). Also, indole in E. coli decreases acid resistance by repressing the acid-resistance genes such as gadABCEX, hdeABD, and ymgB (Lee et al., 2007b, c). Escherichia coli may turn off the acid resistance genes in the presence of indole in the weak basic gut flora, as acid resistance proteins are no longer needed after survival through the acidic stomach (Lee et al., 2007b). Additionally, indole is a chemo-repellent and decreases motility, possibly due to cell division interference, whereas the eukaryotic hormones epinephrine and norepinephrine are a chemo-attractant and increase motility in E. coli O157:H7 (Bansal et al., 2007). Furthermore, indole decreases cell adherence to epithelial cells, whereas epinephrine and norepinephrine increase cell adherence (Bansal et al., 2007). As bacterial adherence and colonization to epithelial cells are important for infection, it was hypothesized that these molecules would also differentially impact the bacterial virulence (Bansal et al., 2007). Additionally, the TnaA-dependent production of 3-nitrosoindole compounds takes place in E. coli under anaerobic growth, although the function of 3-nitrosoindole compounds in signaling has not been established (Kwon & Weiss, 2009).

Moreover, indole increases drug resistance in S. enterica (non-indole-producing bacteria) by inducing the efflux pump system (AcrAB) via the transcriptional regulator RamA that belongs to the AraC transcriptional activator family (Nikaido et al., 2008). The RamA-binding sites are located in the upstream regions of acrAB and tolC (Nikaido et al., 2008). Also, indole decreases the production of virulence factors in P. aeruginosa (non-indole-producing bacteria) by altering gene expression, in contrast to AHLs; for example, indole represses genes encoding the mexGHI-opmD multidrug efflux pump, phz operon, pqs operon, pch operon, and pvd operon, whereas AHLs induce these genes (Lee et al., 2009a). Therefore, indole influences several phenotypes of non-indole-producing bacteria as well as indole-producing bacteria.

Indole and biofilm formation

Indole controlling the biofilm formation is a key example of group behavior in E. coli (Di Martino et al., 2003; Domka et al., 2006; Lee et al., 2007b), V. cholerae (Mueller et al., 2007, 2009), and P. aeruginosa (Lee et al., 2009a). Initially, it was reported that E. coli S17-1 tnaA mutant reduced biofilm formation and indole restored the tnaA biofilm (Di Martino et al., 2003) in Luria–Bertani (LB) medium at 37 °C (Di Martino et al., 2002, 2003). In contrast, indole decreased the biofilm formation of nine nonpathogenic E. coli (BW25113, BW25113 bhsA, BW25113 bssR, BW25113 bssS, BW25113 tnaA, ATCC25404, JM109, TG1, and XL1-Blue) as well as pathogenic E. coli O157:H7 (Domka et al., 2006; Bansal et al., 2007; Lee et al., 2007b; Zhang et al., 2007) in LB glucose medium at 37 °C or in LB at 30 °C. The latter studies demonstrated that indole decreases E. coli biofilms by reducing motility (Domka et al., 2006; Bansal et al., 2007; Lee et al., 2007b), repressing acid resistance genes (Lee et al., 2007b), reducing chemotaxis (Bansal et al., 2007), and reducing attachment to epithelial cells (Bansal et al., 2007). This discrepancy of indole effects on biofilm formation between the studies of Di Martino et al. (2002, 2003) and others (Domka et al., 2006; Bansal et al., 2007; Lee et al., 2007b; Zhang et al., 2007) could be caused by the different experimental conditions and different E. coli strains used (Table 2). For example, the presence of glucose turns off endogenous indole production (Botsford & DeMoss, 1971) so that the effect of exogenous indole may be more significant in the presence of glucose because it reduces background indole (Lee et al., 2007b). Also, the effects of indole on biofilm formation, antibiotic resistance, cell division, and the ability to control gene expression are more significant at temperatures below 37 °C (Lee et al., 2008). Therefore, there is the question of why indole decreases its own biofilm formation in E. coli. A possible explanation can be found in the known biology of V. cholera, in which quorum-sensing molecules negatively regulate biofilm formation to redirect V. cholera growth to a less cell-dense environment (Nadell et al., 2008).

In contrast to E. coli, V. cholerae tnaA mutant resulted in decreased biofilm formation in LB medium at 37 °C, and the addition of indole complemented the tnaA biofilm (Mueller et al., 2007). Indole activates the genes involved in Vibrio polysaccharide (VPS) production, which is essential for V. cholerae biofilm formation (Mueller et al., 2009). Hence, it appears indole positively influences the biofilm formation of V. cholerae (Mueller et al., 2007), whereas other quorum-sensing molecules negatively regulate its biofilm (Hammer & Bassler, 2003; Nadell et al., 2008). It was also reported that indole increased the biofilm formation of the non-indole-producing P. aeruginosa (Lee et al., 2009a). Although many bacteria produce indole (Table 1), indole signaling has only been investigated in a few bacteria, mainly E. coli, S. aurantiaca, and V. cholerae. Hence, it would be interesting to investigate indole signaling in other indole-producing bacteria (Table 1).

Several transcriptomics studies have demonstrated that indole regulates the expression of many genes in E. coli (Nishino et al., 2005; Bansal et al., 2007; Lee et al., 2007b; Lee et al., 2008), P. aeruginosa (Lee et al., 2009a), and V. cholerae (Mueller et al., 2009). Although whole-transcriptome profiling may not reveal some aspects of heterogeneous cells such as biofilm cells (An & Parsek, 2007), DNA microarrays can be used as a starting point toward a better understanding of indole signaling.

Indole as an intercellular signal molecule

It has been controversial whether indole is an intercellular signal molecule or not. A number of criteria for the requirement of a quorum-sensing signal molecule have been suggested. Winzer et al. (2002) suggested four criteria, which were satisfied for indole as followed:

1. The putative signal must be produced during a specific stage. Indole is produced primarily in the stationary phase (Wang et al., 2001; Mueller et al., 2009).

2. The putative signal must accumulate extracellularly and be recognized by a specific receptor. The chemical nature of indole is well known and in most cases of indole regulation, chemical complementation was demonstrated, where indole accumulates during the stationary phase and is a known extracellular signal (Wang et al., 2001; Mueller et al., 2009). This receptor is exported by AcrEF (Kawamura-Sato et al., 1999) and imported by Mtr (Yanofsky et al., 1991). Moreover, several possible indole receptors are currently being investigated (such as SdiA and two-component systems); this matter will be discussed more in detail in the next section.

3. The putative signal must accumulate and generate a concerted response. Indole has been shown to control spore formation (Gerth et al., 1993) and biofilm (Lee et al., 2007b; Mueller et al., 2009).

4. Importantly, the putative signal must elicit a response that extends beyond the physiological changes required to metabolize or detoxify the signal. Indole has been shown to control virulence (Hirakawa et al., 2009), biofilms (Lee et al., 2007b; Mueller et al., 2009), and plasmid stability (Chant & Summers, 2007), which are not related to indole metabolism.

Therefore, according to these criteria, indole has the potential to be a quorum-sensing molecule (Chant & Summers, 2007; Lee et al., 2007a).

As the above four criteria only fit the canonical quorum-sensing signals such as AHLs, Monds & O'Toole (2008) added two more criteria that are more generally applicable and of practical value:

1. The physiologically relevant concentration of the signal required for the phenotypic changes is not toxic to the cell. Escherichia coli and V. cholerae produce up to 0.6 mM indole that is not toxic to the cell (Chant & Summers, 2007; Mueller et al., 2009) and control many phenotypes at that physiological concentration as outlined in the previous section.

2. The signal network is adaptive at the level of the community. Although hard to prove due to the difficulty of evolutionary experiments, we hypothesized that the intercellular signal indole may be beneficial to the microbial community even though the production of indole is costly to the individual.

For example, indole increased plasmid stability (Chant & Summers, 2007), drug resistance (Hirakawa et al., 2005), and grazing resistance to phagocytic eukaryote D. discoideum (Mueller et al., 2009) in indole-producing bacteria E. coli and V. cholerae. In contrast, in non-indole-producing bacteria, indole decreased cell growth of a fungus (Kamath & Vaidyanathan, 1990) and virulence of P. aeruginosa by interfering with the quorum-sensing system (Lee et al., 2009a). Hence, we speculate that indole-producing bacteria may use indole to survive against other bacteria and eukaryotes.

Indole-binding regulatory proteins

Canonical cell-to-cell signal systems, such as AHLs, include a signal synthase and a cognate transcriptional regulator such as LuxR-type proteins that bind to the accumulated signal molecules (Fuqua et al., 1994; Patankar & Gonzalez, 2009). Unlike the AHL systems, there is no LuxR-type protein that directly binds to indole or AI-2. Intriguingly, the E. coli LuxR-homologue SdiA is involved in indole signaling (Lee et al., 2007b, 2008, 2009b), although E. coli does not naturally produce AHLs (Ahmer, 2004). In these studies, SdiA was necessary for indole signaling in E. coli, as in the sdiA mutant, the effect of indole on biofilm formation was lost, and no significant changes of gene expression with a response to indole were shown (Lee et al., 2008). Additionally, the mutation of SdiA by the directed evolution influenced the indole production and biofilm formation in E. coli (Lee et al., 2009b). However, so far, there is no proof of the direct binding of indole and SdiA, and how indole and SdiA interact and work is still unclear and remains to be investigated further.

Indole also acts on the sensor kinases, BaeS and CpxA, and interacts with GadX (a transcriptional activator for acid resistance) to control drug resistance in E. coli (Hirakawa et al., 2005). In V. cholerae, indole can directly interact with the RNA polymerase regulator protein DksA, the dnaK suppressor protein and indole activates the genes involved in VPS production through the DksA and the VPS regulator, VpsR, a distant homologue of SdiA (Mueller et al., 2009). In S. aurantiaca, indole binds to pyruvate kinase (PykA) to induce spore formation in the fruiting bodies (Stamm et al., 2005). Hence, it is interesting to find indole-interacting proteins in other bacterial species using the indole affinity matrix to identify indole binding PykA of S. aurantiaca (Stamm et al., 2005).

Comparison of AI-2 signaling and indole signaling

Bacteria often produce multiple signal molecules and differentially respond to each signal molecule (Waters & Bassler, 2006; Williams & Cámara, 2009). To date, the only signal molecule shared by both Gram-positive and Gram-negative bacteria is AI-2, which is synthesized by the enzyme LuxS (Xavier & Bassler, 2003; Hardie & Heurlier, 2008). The AI-2 system is found in over 55 species and E. coli senses AI-2 that is produced by Vibrio harveyi to assess any changes in its cell population in batch culture on rich laboratory medium (Xavier & Bassler, 2005b), which suggests that AI-2 is an interspecies signal molecule (Xavier & Bassler, 2003; Vendeville et al., 2005). AI-2 is one of the most studied quorum-sensing signals because it is involved in the regulation of bioluminescence and virulence-associated traits in Vibrio (Xavier & Bassler, 2003; Xavier & Bassler, 2005b). AI-2 (Fig. 2) binds to its periplasmic receptor LuxP and interacts with a histidine kinase LuxQ in V. harveyi (Neiditch et al., 2005), and AI-2 in E. coli is internalized by the Lsr ABC transporter (Xavier & Bassler, 2005a). Unlike AHL systems, there is no report of a LuxR-type regulator that binds directly to either AI-2 or indole. Hence, AI-2 and indole signaling may not require a direct binding of LuxR-type protein.

As shown in Table 1, 85 species of both Gram-positive and Gram-negative bacteria produce indole, which supports a possible role of indole as an interspecies signal. Thus, in addition to AI-2, indole is another signal molecule shared by both Gram-positive and Gram-negative bacteria. Notably, E. coli, Photorhabdus luminescens, Porphyromonas gingivalis, Rhizobium leguminosarum bv., Shigella flexneri, V. cholerae, and V. vulnificus, produce both AI-2 and indole (Table 1). Additionally, V. harveyi and Vibrio parahaemolyticus produce multiple signal molecules, such as HAI-1 (an AHL), CAI-1, AI-2 (Waters & Bassler, 2006), and even indole (Table 1). Hence, bacteria may use various signaling mechanisms to gather, process and transduce diverse environmental information, such as pH, temperature, nutrient availability, osmolarity, and cell density to survive in dynamic microbial communities (Xavier & Bassler, 2003; Hardie & Heurlier, 2008; Williams & Cámara, 2009).

Interestingly, indole and AI-2 have divergent characteristics for the biosynthesis of molecules in various environmental conditions (Table 3). AI-2 accumulation is induced in the presence of glucose (Surette & Bassler, 1998), whereas indole synthesis is inhibited in the presence of glucose (John & Wyeth, 1919; Botsford & DeMoss, 1971). AI-2 accumulation was maximal at a low pH (pH 5.0) (Surette & Bassler, 1999), whereas indole synthesis was inhibited at pH levels below 4.3 (John & Wyeth, 1919) and TnaA was one of the most induced protein at pH 9.0 in E. coli (Blankenhorn et al., 1999). AI-2 exists transiently during the exponential phase and is a heat-labile compound (Surette & Bassler, 1998), while indole is stably maintained during the stationary phase and is heat stable in E. coli (Kobayashi et al., 2006) and V. cholerae (Mueller et al., 2009). Furthermore, the signal modifications are different in that AI-2 can be phosphorylated by LsrK kinase and further degraded by LsrF and LsrG (Xavier & Bassler, 2005a), whereas indole can be oxidized by oxygenases from other species (Gillam et al., 2000; Lee et al., 2007a), but is not degraded by its own species (Chant & Summers, 2007).

In E. coli (Table 3), exogenously supplied AI-2 increases biofilm formation at 37 °C (González Barrios et al., 2006), whereas indole decreases biofilm formation more significantly at 30 °C than at 37 °C (Lee et al., 2007b, 2008). AI-2 is chemo-attractant and increases motility and cell adherence to epithelial cells (Bansal et al., 2008), whereas indole is chemo-repellent and decreases swimming motility and cell adherence to epithelial cells (Bansal et al., 2007). The addition of (S)-4,5-dihydroxy-2,3-pentanedione (DPD, AI-2 precursor) to the AI-2-deficient (luxS) mutant leads to more extensive differential gene expression at 37 °C than at 30 °C, whereas the addition of indole to the indole-deficient (tnaA) mutant leads to more extensive differential gene expression at 30 °C than at 37 °C (Lee et al., 2008). As a result, DPD induces the expression of uracil-related genes (carAB, pyrLBI, pyrC, pyrD, pyrF, upp, and uraA) at 37 °C, but no induction at 30 °C, whereas indole represses the expression of the same uracil-related genes at 30 °C, but not at 37 °C (Lee et al., 2008). Also, the indole-derivative isatin (Fig. 2) seems to mimic AI-2 in stimulating biofilm formation (Lee et al., 2007a) because both AI-2 and isatin induce the same flagella genes and increase motility, and isatin decreases the indole production (Ren et al., 2004; Lee et al., 2007a; Bansal et al., 2008). Therefore, bacterial species may utilize different signaling systems (or redundant signal molecules) to sense diverse environmental conditions and to adapt in a new environment. Indole and AI-2 signalings seem to be intertwined.

Understanding how AI-2 and indole signaling are connected to each other is challenging but stimulating. In other strains, V. harveyi uses shared regulatory components to discriminate between multiple autoinducers, and LuxR controls the quorum-sensing-regulated genes (Waters & Bassler, 2006). Additionally, multiple quorum-sensing systems in P. aeruginosa are integrated and subject to the prevailing environmental conditions (Williams & Cámara, 2009). In E. coli, SdiA (a LuxR homologue) recognizes AHLs, although E. coli cannot produce AHLs (Yao et al., 2006). The indole signaling mechanism requires SdiA (Lee et al., 2007b; Lee et al., 2008) and the directed-evolution of the SdiA protein influences the production of indole (Lee et al., 2009b). It was also suggested that AI-2 might work together with SdiA (DeLisa et al., 2001). Therefore, it is possible that SdiA may be associated with multiple signals, such as AHLs, AI-2, and indole.

Interference of indole signaling

In environmental niches, bacteria coexist in multispecies communities with other bacteria and their hosts, while competing for resources and spaces. Signal molecules can be interfered intrinsically and extrinsically by signal degradation, inhibition of signal molecule biosynthesis, the reduction of receptor proteins, and structural modification (Zhang & Dong, 2004). Although indole is stably present during the stationary phase in indole-producing E. coli and V. cholerae, many non-indole-producing bacteria can metabolize indole through some dioxygenases and monooxygenases found in microbial communities. For example, Pseudomonas putida PpG7 (Ensley et al., 1983), Ralstonia pickettii PKO1 (Fishman et al., 2005), Pseudomonas mendocina KR1 (Tao et al., 2004), and Burkholderia cepacia G4 (Rui et al., 2005) readily convert indole to oxidized indole compounds such as 2-hydroxyindole, 3-hydroxyindole, 4-hydroxyindole, isatin, indigo, isoindigo, and indirubin (Rui et al., 2005). In some cases, non-indole-producing bacteria can utilize indole as a carbon source. For example, P. aeruginosa Gs isolated from mangrove sediment (Yin et al., 2005) and Pseudomonas sp. ST-200 from soil (Doukyu & Aono, 1997) grow on indole and thus can remove indole from the environment.

A previous report showed that indole inhibited cell growth of a fungus, A. niger, that could degrade indole (Kamath & Vaidyanathan, 1990). Indole enhances the virulence of enterohemorrhagic E. coli (Anyanful et al., 2005; Hirakawa et al., 2009). Moreover, indole diminishes the virulence of P. aeruginosa that may threaten E. coli and its host (Lee et al., 2009a), whereas P. aeruginosa can easily degrade indole (Lee et al., 2009a). In a dual species culture of E. coli and Pseudomonas fluorescens expressing toluene-o-monooxygenase that can oxidize indole, the presence of toluene-o-monooxygenase decreases the level of indole and controls biofilm formation of E. coli (Lee et al., 2007b). Hence, non-indole-producing bacteria have developed a defense system to cope with indole; therefore, competition and interference for indole signal appear to be intense in multispecies bacterial consortia.

A proposed model of interference for indole signaling in an animal intestine is shown in Fig. 3. A variety of animal species contain indole-positive bacteria in their intestinal tract (DeMoss & Moser, 1969). The concentration of fecal indole in the average Asian woman is between 10 and 30 μg g⁻¹ dried stool depending on her diet (Fujisawa et al., 2006; Ishikado et al., 2007), which indicates that gut bacteria produce copious amounts of indole in humans. Non-indole-producing bacteria with oxygenases can oxidize indole into diverse oxidized indoles, which can be further dimerized into insoluble indigoid compounds (Rui et al., 2004) (Fig. 3). Hence, some bacterial oxygenases have been proposed to have evolved to regulate the concentration of indole signal by removing it via precipitation (Lee et al., 2007b). Indole generated from gut bacteria can be absorbed into the human body in substantial amounts and cytochrome P450 enzymes can oxidize indole into a variety of oxidized indoles (Gillam et al., 2000). Moreover, metabolomic analysis reveals that the production of several indole metabolites in mice blood, such as indolyl sulfate and indole-3-propionic acid, is completely dependent on gut microbial communities (Wikoff et al., 2009).

Roles of indole derivatives in multispecies community

Bacteria-originated indole derivatives may play biological roles in microbial consortia as well as eukaryotic cells. Oxidized indole derivatives showed distinctive effects on biofilm formation, cell motility, and gene expression in the pathogenic E. coli O157:H7 (Lee et al., 2007a). For example, isatin increases biofilm formation by repressing indole production and increasing motility, whereas 5-hydroxyindole and 7-hydroxyindole (Fig. 2) are more potent in decreasing biofilm formation of E. coli O157:H7 compared with indole (Lee et al., 2007a). Also, 7-hydroxyindole diminishes the virulence of P. aeruginosa by repressing quorum-sensing-related genes in a manner opposite that of AHLs and decreases swarming motility and colonization of P. aeruginosa in guinea pigs (Lee et al., 2009a). Hence, 7-hydroxyindole (Lee et al., 2009a) could potentially be used for antivirulence therapies (Lesic et al., 2007; Cegelski et al., 2008), which are also known as antipathogenic drugs (Rasmussen & Givskov, 2006). Antivirulence compounds are an important tool for fighting infectious diseases because, unlike antimicrobials, antivirulence compounds such as 7-hydroxyindole do not affect cell growth (Lee et al., 2009a), so there is less chance of developing resistance (Hentzer et al., 2002; Lesic et al., 2007).

Many plants have developed bacterial defense systems that interfere with bacterial cell signaling using signaling inhibitors, such as halogenated furanones from Delisea pulchra (Hentzer et al., 2003) and other quorum-sensing inhibitors from carrots, garlic, chili, and water lily (Rasmussen & Givskov, 2006). Natural indole-like compounds also influence bacterial physiology. For example, indole-3-carbinol and 3,3′-diindolylmethane (Fig. 2) from cruciferous vegetables show antimicrobial, antiviral, and anticancer activity (Higdon et al., 2007; Fan et al., 2009). As a preliminary, indolylacetonitrile (Fig. 2) inhibited biofilm formation of E. coli O157:H7 up to 10-fold in a dose–response manner (0, 10, 30, 75, and 100 μg mL⁻¹) without affecting the cell growth (J.-H. Lee & J. Lee, unpublished data). Additionally, a synthetic indole derivative (CBR-4830) has been shown to inhibit P. aeruginosa growth through a multidrug efflux pump, mexAB-oprM (Robertson et al., 2007). These results suggest that indole derivatives (natural and synthetic) are potential natural biofilm inhibitors as well as antimicrobial agents to control pathogenic bacteria.

Moreover, indole and indole derivatives may have some biological functions in the human body. Humans and intestinal bacteria have developed a commensal relationship over a long period, and some bacteria are crucial for nutrient assimilation and are beneficial to the human immune system (Hooper & Gordon, 2001; Wikoff et al., 2009). Recently, an MS-based metabolomics study demonstrated that the production of indoxyl sulfate and indole-3-propionic acid in animal blood completely depends on enteric bacteria (Fig. 2) (Wikoff et al., 2009). Indole-3-propionic acid is a powerful antioxidant (more potent than melatonin) and a possible treatment for Alzheimer's disease (Chyan et al., 1999). Furthermore, gut bacteria-derived isatin (Fig. 2) was detected in blood, peripheral tissues, urine, and even brain up to a relative high level (70 μM), although its biological targets remain poorly characterized (Crumeyrolle-Arias et al., 2009). Additionally, 5-hydroxyindole (Fig. 2) produced from gut bacteria was detected in blood, plasma, and brain, where the concentration of 5-hydroxyindole fluctuated photoperiodically, although its role is unknown (Crumeyrolle-Arias et al., 2008).

The study of how gut epithelial cells react to indole and indole derivatives is of interest. Recently, it was shown that indole significantly affects the gene expression of epithelial cell-junction and cytokines (Bansal et al., 2009). Interestingly, bacterial signal indole and eukaryotic hormones, epinephrine and norepinephrine, exert divergent effects on bacterial physiology of enterohemorrhagic E. coli (Bansal et al., 2007). Therefore, indole and indole derivatives (Fig. 2) may influence the balance of microbial flora in the human body.

Concluding remarks

We demonstrate here that both Gram-positive and Gram-negative bacteria including many pathogens produce a relatively large quantity of indole (Table 1) in spite of the high metabolic cost, and that indole represents a new class of intercellular signal molecules having diverse biological functions in ecological niches. To date, indole signaling has been studied using only few strains, and it offers an exciting challenge to investigate the roles of indole in various species including human cells. More importantly, further studies are essential to determine the genetic regulatory mechanism of indole signaling, which is still elusive.

As indole biosynthesis can be influenced by several environmental factors (Fig. 1, Table 3), the experimental conditions are critical in the study of indole signaling (Table 2). For example, if one wants to mimic the high indole environment of the large intestine, a glucose-free medium should be chosen. Additionally, a physiologically relevant concentration of indole is required to study indole signaling (note that E. coli and V. cholerae produce extracellular indole up to 0.6 mM in a rich medium). A high concentration of indole (above 2 mM) apparently decreases cell growth in E. coli (Chant & Summers, 2007; Lee et al., 2009b), probably due to the blocking of cell division (Chant & Summers, 2007; Lee et al., 2009b), the disruption of the bacterial envelope (Raffa & Raivio, 2002), and/or oxidant toxicity (Garbe et al., 2000). Consequently, a high dose of indole may affect the overall cellular metabolism and may lead to a pleiotropic effect. Additionally, it is difficult to measure the environmental concentration of indole in dynamic microbial niches, especially inside a biofilm and in a multispecies bacterial consortium. Therefore, a wide range of indole concentrations should be carefully investigated. For example, AI-2 concentration critically influences (increases or decreases) biofilm formation of dual-species of Actinomyces naeslundii T14V and Streptococcus oralis 34 (Rickard et al., 2006).

As microbial communities dynamically react with multiple signal molecules such as AHLs, AI-2, and indole, temporal study of signal molecules is necessary instead of a single time point study. It was hypothesized that multiple AHL signals in P. aeruginosa were believed to regulate the gene expression in a specific temporal order (Schuster et al., 2003), and a temporal transcriptomic study of AI-2 signal showed the temporal expression of virulence genes (Bansal et al., 2008). Table 3 indicates a possible interconnection between AI-2 and indole, which suggests that bacteria may utilize multiple signals to respond dynamically to diverse environmental stimuli.

Many bacteria may use the indole signal to thrive over other bacteria in multispecies communities, whereas other bacteria have acquired diverse defense systems, such as monooxygenases, dioxygenases, and P450 family, to metabolize indole. Furthermore, animals possibly utilize indole derivatives originated from gut microbial communities for their immune systems (Wikoff et al., 2009). Currently, there is much interest in ‘postbiotics’ as well as ‘probiotics’ as new therapeutic strategies (Neish, 2009). Postbiotics are defined as the supplementation of the human gut with bacteria products such as butyrate and other short-chain fatty acids (Neish, 2009). Indole derivatives have a potential as postbiotics, as indole-3-propionic acid is a powerful antioxidant (Wikoff et al., 2009), and 7-hydroxyindole diminishes virulence and colonization of pathogenic P. aeruginosa (Lee et al., 2009a). To develop a new therapeutic agent, it will be necessary to screen more natural and synthetic indole derivatives that are nontoxic and cannot be easily metabolized by pathogens.

Indole signaling appears to be important in microbial consortia and may influence the digestive and immune system in humans. Research into indole signaling has recently been begun and much remains to be investigated. Understanding indole signaling will help develop effective antimicrobial or antivirulence strategies and biotechnology applications.

Acknowledgements

This research was supported by the Yeungnam University research grant (to J.L.). J.-H.L. was supported by the Brain Korea 21 Project from the Ministry of Education and Human Resources, Korea. We would like to thank Thomas K. Wood (Texas A & M University) for his help in the study of indole signaling.

References