|Home | About | Journals | Submit | Contact Us | Français|
Previous work has shown that Burkholderia cenocepacia produces the diffusible signal factor (DSF) family signal cis-2-dodecenoic acid (C12:Δ2, also known as BDSF), which is involved in the regulation of virulence. In this study, we determined whether C12:Δ2 production is conserved in other members of the Burkholderia cepacia complex (Bcc) by using a combination of high-performance liquid chromatography, mass spectrometry, and bioassays. Our results show that five Bcc species are capable of producing C12:Δ2 as a sole DSF family signal, while four species produce not only C12:Δ2 but also a new DSF family signal, which was identified as cis,cis-11-methyldodeca-2,5-dienoic acid (11-Me-C12:Δ2,5). In addition, we demonstrate that the quorum-sensing signal cis-11-methyl-2-dodecenoic acid (11-Me-C12:Δ2), which was originally identified in Xanthomonas campestris supernatants, is produced by Burkholderia multivorans. It is shown that, similar to 11-Me-C12:Δ2 and C12:Δ2, the newly identified molecule 11-Me-C12:Δ2,5 is a potent signal in the regulation of biofilm formation, the production of virulence factors, and the morphological transition of Candida albicans. These data provide evidence that DSF family molecules are highly conserved bacterial cell-cell communication signals that play key roles in the ecology of the organisms that produce them.
The Burkholderia cepacia complex (Bcc) comprises a group of currently 17 formally named bacterial species that, although closely related, are phenotypically diverse (17, 22, 23). Strains of the Bcc are ubiquitously distributed in nature and have been isolated from soil, water, the rhizosphere of plants, industrial settings, hospital environments, and infected humans. Some Bcc strains have emerged as problematic opportunistic pathogens in patients with cystic fibrosis or chronic granulomatous disease, as well as in immunocompromised individuals (17). The clinical outcome of Bcc infections ranges from asymptomatic carriage to a fulminant and fatal pneumonia, the so-called “cepacia syndrome” (12, 17). Although all Bcc species have been isolated from both environmental and clinical sources, B. cenocepacia and B. multivorans are most commonly found in clinical samples (16).
Many bacterial pathogens have evolved a cell-cell communication mechanism known as quorum sensing (QS) to coordinate the expression of virulence genes. In spite of their genetic differences, most Bcc species produce N-acylhomoserine lactone (AHL) QS signals (25). More recently, another QS signal molecule, cis-2-dodecenoic acid (BDSF), has been identified in B. cenocepacia (3). Subsequent studies showed that BDSF plays a role in the regulation of bacterial virulence (6, 20). Interestingly, the two QS systems appear to act in conjunction in the regulation of B. cenocepacia virulence, as a set of the AHL-controlled virulence genes are also positively regulated by BDSF (6). Furthermore, mutation of Bcam0581, which is required for BDSF biosynthesis, results in substantially retarded energy production and impaired growth in minimal medium (6), highlighting the dual roles of the QS system in the physiology of and infection by B. cenocepacia.
BDSF is a structural analogue of cis-11-methyl-2-dodecenoic acid, which is a QS signal known as diffusible signal factor (DSF) originally identified from the plant bacterial pathogen Xanthomonas campestris pv. campestris (2, 24). Evidence is accumulating that DSF-type fatty acid signals represent a new family of QS signals, which are widespread among Gram-negative bacteria (10, 24). For example, DSF and seven structural derivatives were identified in supernatants of Stenotrophomonas maltophilia (8, 11), 12-methyl-tetradecanoic acid was shown to be produced by Xylella fastidiosa (18), and cis-2-decenocic acid was found to be synthesized by Pseudomonas aeruginosa (5). In addition, DSF-like activity has also been reported in a range of Xanthomonas species, including X. oryzae pv. oryzae and X. axonopodis pv. citri (1, 2, 4, 24), but the chemical structures of these DSF analogues remain to be determined. Unlike other known QS signals, such as AHL and AI-2 family signals, DSF and its analogues, including BDSF, are fatty acids and these fatty acid signals were collectively designated DSF family signals for the convenience of discussion (10). Considering the fact that the list of DSF family signal is expanding, we propose to designate cis-11-methyl-2-dodecenoic acid (DSF) 11-Me-C12:Δ2 and cis-2-dodecenoic acid (BDSF) C12:Δ2. This nomenclature is based on one of the fatty acid nomenclatures (13, 19) where the methyl (Me) substitution and its position are indicated first (for example, 11-Me indicates a methyl group on C-11 of the fatty acid carbon chain), followed by the length of the fatty acid carbon chain (C12 represents a 12-carbon fatty acid chain), and then the position of the double bond in the fatty acid chain (Δ2 indicates a double bond in the cis configuration at site 2, i.e., between C-2 and C-3 of the fatty acid carbon chain). In this way, it is convenient to say that 11-Me-C12:Δ2 and C12:Δ2 have identical 12-carbon fatty acid chains with a cis bond at the same site but differ in a methyl substitution on C-11. Following this nomenclature system, 12-methyl-tetradecanoic acid and cis-2-decenocic acid can be referred to as 12-Me-C14 and C10:Δ2, respectively.
DSF family signals have emerged as important factors in the regulation of virulence and biofilm formation in a wide range of bacterial pathogens (10). In this study, we have investigated the production of the DSF family signals in nine Bcc species. It is demonstrated that C12:Δ2 is conserved in members of the Bcc and that 11-Me-C12:Δ2 and a novel DSF family signal were also produced by some, but not all, of the Bcc strains investigated. This new signal was identified as cis,cis-11-methyldodeca-2,5-dienoic acid (11-Me-C12:Δ2,5) by nuclear magnetic resonance (NMR) analysis and mass spectrometry. We have also investigated the biological significance of 11-Me-C12:Δ2,5 in intraspecies and interspecies communication.
The Bcc strains used in this work are listed in Table Table1.1. These strains were grown at 28°C or 37°C, as indicated, with shaking at 250 rpm in Luria-Bertani (LB) broth. X. campestris pv. campestris strain 8004 and its rpfF deletion mutant 8004dF were described previously (9, 24). X. campestris pv. campestris strains were maintained at 30°C in YEB medium (26). B. cenocepacia J2315 and its Bcam0581 deletion mutant d0581 were described previously (3). For the cultivation of static biofilms of B. cenocepacia, bacteria were grown at 30°C in basal salt medium (pH 7.2) containing 20 mM citrate and 0.5% Casamino Acids (3). The following antibiotics were added as supplements when necessary: rifampin, 50 μg ml−1; gentamicin, 100 μg ml−1; tetracycline, 10 μg ml−1; trimethoprim, 400 μg ml−1 (B. cenocepacia) or 1.5 mg ml−1 (Escherichia coli). Candida albicans SC5314 was grown in GMM medium consisting of 6.7 g of Bacto yeast nitrogen base (Difco, Sparks, MD) and 0.2% glucose (pH 7.2) (3). The DSF family signal molecules were added to the medium at a final concentration of 5 μM unless indicated otherwise.
Overnight bacterial culture supernatants (250 ml) were extracted with acidified ethyl acetate at a 1:1 ratio. The organic phase was dried using a rotary evaporator, and the residues were dissolved with 200 μl of methanol. An aliquot of 5 μl extracts was spotted onto a silica gel TLC plate (20 by 20 cm; Merck) and separated with ethyl acetate-hexane (20:80, vol/vol) as running solvents. Subsequently, the plates were dried under airflow and overlaid with 100 ml of NYG medium (20 g glycerol, 5 g peptone, and 3 g yeast extract per liter), which was supplemented with 0.8 g agarose, 250 μg of 5-bromo-4-chloro-3-indolyl-β-d-glucoside, and 4 ml of the biosensor strain FE58 at an optical density at 600 nm (OD600) of 1.8 (24). The TLC plate was incubated overnight at 28°C, and DSF activity was indicated by the presence of a blue spot.
To isolate and identify C12:Δ2 and its analogues from supernatants, 1-liter cultures of Bcc strains were grown to an OD600 of about 3.0 and centrifuged. The supernatants were acidified to a pH of 4.0 with diluted HCl and extracted with ethyl acetate (1.0, vol/vol) twice. Following evaporation of the ethyl acetate, the residue was dissolved in methanol, subjected to flash chromatography on normal-phase silica gel, and eluted consecutively with 2 bed volumes of hexane, 2 bed volumes of 10% ethyl acetate in hexane, and 4 bed volumes of 25% ethyl acetate in hexane. The active fractions, which were detected using the DSF sensor FE58 (24), were combined for high-performance liquid chromatography (HPLC) profiling analysis on a reverse-phase column (Phenomenex Luna, 5 μm C18, 250 by 4.60 mm) and eluted with 80% methanol in H2O at a flow rate of 1 ml min−1. Peaks were monitored with a UV detector (λ = 210 and 254 nm). Fractions were collected at 1-min interval and assayed using the DSF biosensor FE58.
The 1H, 13C, and heteronuclear multiple quantum coherence (HMQC) NMR spectra in CDCl3 solution were obtained using a Bruker DRX500 spectrometer operating at 500 MHz for 1H or 125 MHz for 13C. High-resolution electrospray ionization mass spectrometry (ESI-MS) was performed on a Finnigan/MAT MAT 95XL-T mass spectrometer under the conditions described previously (24).
The coding region of Bmul5121 was amplified from B. multivorans via PCR using primers BMUL5121-F (5′-TGCTCTAGAGCAATGCAGCTCCAATCACATCCC) and BMUL5121-R (5′-CCCAAGCTTGGGTCACACCGTGCGCAACTTC). The product was digested with XbaI and HindIII and ligated at the same enzyme sites under the control of the S7 ribosomal protein promoter in plasmid vector pMSL7 (15). After sequence verification, the resulting construct, pMSL7-Bmul5121, was introduced into the DSF-negative mutant d0581 by triparental mating. Transconjugants were selected on LB agar plates supplemented with gentamicin and trimethoprim. Likewise, plasmid pMSL7-Bcam0581 was used to complement mutant d0581. Plasmids pMSL7-Bcam0581 and pMSL7-Bmul5121 were also used to heterologously express Bcam0581 and Bmul5121 in E. coli strain DH5α.
For quantification of extracellular polysaccharide (EPS) production, 10-ml volumes of overnight YEB cultures at an OD600 of 3.0 were centrifuged at 12,000 rpm for 20 min. The supernatants were mixed with 2.5 volumes of absolute ethanol, and the mixture was incubated at 4°C for 30 min. The precipitated EPS was isolated by centrifugation and dried overnight at 55°C before the determination of dry weights.
The formation of biofilms was investigated as follows. Cultures of the X. campestris pv. campestris wild-type strain and the DSF-negative mutant 8004dF were grown overnight in 5 ml of YEB medium with or without a signal molecule. Methanol was used as a solvent control. After overnight incubation, bacterial cells were visualized by phase-contrast microscopy (Olympus BX50) and images were taken with an Olympus DP70 digital camera.
Plasmid pMLS7-egfp (15) was used to tag B. cenocepacia J2315 and its DSF-negative derivative d0581 with green fluorescent protein. Five-microliter samples of bacterial cultures grown overnight to an OD600 of about 3.0 were inoculated in duplicate into sterile six-well tissue culture plates containing 3 ml of basal salt medium. The plates were incubated without agitation at 30°C for 3 days. The biofilms that formed at the air-liquid interface were analyzed by confocal scanning laser microscopy using a Carl Zeiss LSM510-Axiovert 100M confocal microscope.
To test the effects of signal molecules on the expression of the zmpA gene, which encodes a zinc metalloprotease important for pathogenicity, a PzmpA-lacZ gene fusion was employed as described previously (6) and the promoterless fusion construct pMLS7-lacZ was used as the negative control (6). For measurement of β-galactosidase activity, bacterial cells were grown in LB medium at 37°C with shaking at 250 rpm. When required, signal molecules were added to a final concentration of 5 μM as indicated. Bacterial cells were harvested, and the β-galactosidase activities were assayed as described previously (14).
To test the effect on C. albicans germ tube formation, an overnight culture of C. albicans strain SC5314 grown in GMM medium was diluted 20-fold in fresh GMM medium (3). Signal molecules were then added separately as indicated, and the cells were induced for 3 h at 37°C. Visualization and quantification of germ tube formation were performed using a phase-contrast microscope (Olympus BX50) by counting about 400 fungal cells per sample. Microphotography was done with an Olympus DP70 digital camera.
Our previous study has shown that BCAM0581 of B. cenocepacia J2315 is the enzyme responsible for the synthesis of C12:Δ2, which shares about 37% identical amino acids with the DSF synthase RpfF of X. campestris pv. campestris (3). A BLAST search revealed that BCAM0581 is conserved in B. lata, B. multivorans, B. cenocepacia, B. vietnamiensis, B. dolosa, and B. ambifaria. Sequence alignment of the BCAM0581 homologues showed that they are highly conserved, with more than 94% amino acid identity (Fig. (Fig.1A).1A). A synteny analysis showed that the downstream region is highly conserved in all six strains while the upstream region is variable, with only one conserved gene, Bcam0582 (Fig. (Fig.1B).1B). Sequence analyses of the conserved neighboring genes suggested that Bcam0580 encodes a PAS-GGDEF-EAL multidomain fusion protein, Bcam0578 encodes a putative 5-oxoprolinase, and Bcam0582 encodes a transglutaminase. These orthologous genes show about 80 to 91% amino acid identity.
To test whether the strains are capable of producing C12:Δ2, we analyzed the crude solvent extracts of nine Bcc strains by TLC. After separation, the DSF-like signals were visualized by overlaying the TLC plate with the DSF sensor strain FE58, which contains a GUS gene under the control of a DSF-inducible promoter (24). A blue spot with an Rf value similar to that of C12:Δ2 was detected in all nine strains, suggesting that they produce a DSF-like signal(s) (Fig. (Fig.1C1C).
For purification, the solvent extracts of the culture supernatants were first subjected to flash chromatography. The active fractions, as identified by aid of the DSF sensor strain FE58, were then combined for reverse-phase HPLC analysis. Bioactive peaks with a retention time indicative of C12:Δ2 (3) were observed with all of the strains investigated (Fig. (Fig.2A).2A). Intriguingly, additional peaks with DSF-like activity were observed with B. multivorans, B. stabilis, B. anthina, and B. pyrrocinia (Fig. (Fig.2A).2A). The structures of these DSF-like molecules were identified as exemplified for B. multivorans, which produces three peaks with DSF-like activity (Fig. (Fig.2B).2B). Active fractions were collected and analyzed by ESI-MS. Fractions a, b, and c showed peaks at m/z values of 211.27, 197.27, and 209.27, respectively (Fig. 3A to C). These m/z values are in agreement with the molecular formulas C13H23O2, C12H21O2, and C13H21O2, respectively. The compounds in fractions a and b were unambiguously characterized as cis-11-methyl-2-decenoic acid (11-Me-C12:Δ2) and cis-2-dodecenoic acid (C12:Δ2) by NMR analysis (Fig. (Fig.3D;3D; Table Table2).2). These molecules were originally identified in culture supernatants of X. campestris and B. cenocepacia (3, 24).
In the case of fraction c, the 1H spectrum indicated that there are two pairs of ethylenic protons (Table (Table2).2). The coupling constants between the protons in each pair are lower than 11 Hz. This indicates that the two double bonds are both in the cis configuration. The two methylene protons at δH 3.45 suggest that this methylene carbon is connected with the two double bonds. The overlapping signals of two doublet methyl groups at δH 0.87 indicate a branched structure similar to that of 11-Me-C12:Δ2 (24). 13C spectra revealed that one of the double bonds is conjugated with a carbolic acid (Table (Table2).2). Therefore, the second double bond in the molecule is at C-5 (Table (Table2).2). Collectively, the 1H, 13C, and HMQC data established the structure of this novel DSF family member as cis,cis-11-methyldodeca-2,5-dienoic acid, which is structurally identical to 11-Me-C12:Δ2, except for an extra double bond between C-5 and C-6 (Table (Table2).2). For consistency and convenience, this newly identified molecule was designated 11-Me-C12:Δ2,5. Using a combination of HPLC analysis and bioassays, we showed that 11-Me-C12:Δ2 was only produced by B. multivorans, whereas 11-Me-C12:Δ2,5 was detectably produced by the three Bcc species B. stabilis, B. anthina, and B. pyrrocinia (Fig. (Fig.2A2A).
Quantitative analysis showed that the Bcc species investigated here differed not only in the quantity of but also in the variety of DSF-like molecule species they produced (Fig. (Fig.2A).2A). Given that the Bcc species we investigated have similar but not identical BDSF synthase genes (Fig. (Fig.1A),1A), it was of importance to test whether the ability to produce different DSF-like signals is related to variations in the BDSF synthase genes. To this end, the BDSF synthase gene Bcam0581 of B. cenocepacia and its homologue Bmul5121 of B. multivorans were cloned and expressed in the Bcam0581 deletion mutant d0581. As expected, overexpression of Bcam0581 in d0581 rescued C12:Δ2 biosynthesis (Fig. (Fig.2C).2C). Interestingly, in trans expression of Bmul5121 in d0581 also led to the production of C12:Δ2 but not of 11-Me-C12:Δ2 or 11-Me-C12:Δ2,5 (Fig. (Fig.2C).2C). However, heterologous expression of both Bcam0581 and Bmul5121 in E. coli DH5α resulted the production of 11-Me-C12:Δ2 and C12:Δ2 (Fig. (Fig.2D).2D). These data suggest that it is the genetic background of the Burkholderia strains used rather than variation in the BDSF synthase genes that affects the spectrum of DSF signal molecules.
To evaluate the biological relevance of C12:Δ2 and 11-Me-C12:Δ2,5, we tested whether they can serve as signal molecules in interspecies communication. In the plant pathogen X. campestris, 11-Me-C12:Δ2 is required as an antiaggregation factor (7). While the DSF-producing wild-type strain 8004 grows planktonically, with cells being well dispersed (Fig. (Fig.4A),4A), the DSF-negative mutant 8004dF forms large aggregates (9) (Fig. (Fig.4B).4B). Similar to the effect of 11-Me-C12:Δ2 (Fig. (Fig.4C),4C), addition of 5 μM C12:Δ2 and 11-Me-C12:Δ2,5 completely dispersed these cell aggregates (Fig. 4D and E).
We then quantitatively compared the biological activity of C12:Δ2 and 11-Me-C12:Δ2,5 with that of their analogue on the production of EPS in the DSF-negative X. campestris rpfF mutant 8004dF. These experiments showed that addition of 5 μM 11-Me-C12:Δ2, C12:Δ2, and 11-Me-C12:Δ2,5 to cultures of 8004dF increased their EPS production to 78%, 69%, and 78.9% of the wild-type level, respectively (Fig. (Fig.4F).4F). These results establish C12:Δ2 and 11-Me-C12:Δ2,5 as an effective signal in bacterial interspecies communication and suggest that the three signal molecules are functionally interchangeable.
Previous studies in our laboratory have shown that C12:Δ2 and 11-Me-C12:Δ2 are able to modulate the morphological transition of C. albicans (3, 24). To test whether the extra double bond in 11-Me-C12:Δ2,5 influences its potency in bacterium-fungus communication, this newly identified signal molecule was added to fresh fungal yeast cells. We used 11-Me-C12:Δ2 and C12:Δ2 as positive controls along with methanol as a solvent control. After incubation at 37°C for 3 h, the majority of the C. albicans cells in the solvent control formed germ tubes (Fig. (Fig.5A),5A), whereas the fungus grew mainly in the form of yeast cells when the culture medium was amended with 5 μM 11-Me-C12:Δ2,5, 11-Me-C12:Δ2, or C12:Δ2 (Fig. 5B to D). A quantitative analysis using serial dilutions of the signal molecules showed that C12:Δ2 was the most potent inhibitor of C. albicans germ tube formation, followed by 11-Me-C12:Δ2,5 and 11-Me-C12:Δ2 (Fig. (Fig.5E5E).
The fact that 11-Me-C12:Δ2 plays a key role in the negative regulation of cell aggregate formation by X. campestris (7, 9) encouraged us to examine the role of the DSF family signals in biofilm formation by B. cenocepacia. When grown statically, B. cenocepacia formed a thin layer of pellicle-like biofilm at the liquid-air interface. Microscopic examination of the biofilms formed by the wild-type strain B. cenocepacia J2315 revealed a smooth surface with only a few small cell aggregates (Fig. (Fig.6A).6A). However, the surface of the biofilms formed by the DSF-negative mutant was uneven, with several large protrusions, as indicated by the lumpy x- and y-axis cross sections (Fig. (Fig.6B).6B). The biofilm structure was restored to that of the wild-type strain when the mutant was grown in the presence of 11-Me-C12:Δ2, C12:Δ2, or 11-Me-C12:Δ2,5 (Fig. 6C to E, respectively).
Our previous study has shown that inactivation of the DSF synthase gene Bcam0581 in B. cenocepacia J2315 resulted in decreased expression of virulence genes and that this reduction could be rescued by the addition of C12:Δ2 (6). To test whether 11-Me-C12:Δ2,5 and 11-Me-C12:Δ2 are functional analogues of C12:Δ2, cultures of the DSF-negative B. cenocepacia mutant d0581 harboring a fusion of the zmpA metalloprotease promoter with the lacZ reporter gene on a plasmid were supplemented with DSF family signal molecules. Compared with that in the wild type, the zmpA promoter activity was decreased by 53% in mutant d0581 (Fig. (Fig.6F).6F). This reduction was rescued not only by the addition of C12:Δ2 but also by supplementation with 11-Me-C12:Δ2,5 and 11-Me-C12:Δ2 at the same concentration (Fig. (Fig.6F6F).
The results of this study showed that C12:Δ2 is a conserved signal molecule in the Bcc, with at least nine species producing C12:Δ2 as the major DSF family signal molecule (Fig. (Fig.2A).2A). Whether the eight more recently described Bcc species (22, 23) also produce DSF-like molecules remains to be determined. In addition to its role in interspecies signal communication (3), C12:Δ2 has recently been shown to play a role in the regulation of B. cenocepacia virulence gene expression (6, 20). In addition, evidence is emerging that B. cenocepacia appears to utilize both AHL- and DSF-dependent QS systems to coordinate the expression of virulence factors (6). Interestingly, similar to C12:Δ2, the AHL-type QS signal N-octanoyl-l-homoserine is also produced by many members of the Bcc (21, 25). These findings, together with the results of this study, suggest that the two QS systems may have coevolved in the Bcc to concertedly modulate bacterial physiology and virulence.
Surprisingly, while five of the Bcc species investigated, namely, B. lata, B. cenocepacia, B. vietnamiensis, B. dolosa, and B. ambifaria, only produced C12:Δ2, B. multivorans, B. stabilis, B. anthina, and B. pyrrocinia were shown to also synthesize 11-Me-C12:Δ2,5 (Fig. (Fig.2A),2A), which is a new member of the family of DSF-like signal molecules. Moreover, B. multivorans was found to also produce 11-Me-C12:Δ2 (Fig. 2A and B), which was originally identified in supernatants of the plant pathogen X. campestris pv. campestris (10, 24). What may account for the structural diversity of DSF family molecules in these bacterial species? The findings of this study are consistent with the notion that the variation in the DSF synthases is not responsible for the differences in the product spectra. First, only the C12:Δ2 molecule was detected when the DSF synthase genes from B. cenocepacia or B. multivorans were overexpressed in a DSF-negative mutant of B. cenocepacia (Fig. (Fig.2C).2C). Second, heterologous expression of Bcam0581, as well as Bmul5121, in E. coli DH5α resulted in the production of 11-Me-C12:Δ2 and C12:Δ2 (Fig. (Fig.2D).2D). Importantly, a bioinformatic analysis of the B. multivorans genome sequence with the protein sequence of the DSF synthase Bmul5121 did not reveal the presence of paralogues. Taken together, these data suggest that the observed variations in the production of DSF molecules is likely due to the availability of different precursors in the different Bcc species.
Structural analysis characterized 11-Me-C12:Δ2,5 as cis,cis-11-methyldodeca-2,5-dienoic acid (Fig. (Fig.3;3; Table Table2),2), which differs from 11-Me-C12:Δ2 by an extra double bond in the cis configuration at the C-5-C-6 position (Fig. (Fig.3D).3D). Functional characterization of this newly identified signal molecule showed that 11-Me-C12:Δ2,5 is not only active in interspecies communication (Fig. (Fig.44 and and5)5) but also potent in the regulation of virulence gene expression and biofilm development in B. cenocepacia (Fig. (Fig.6).6). In agreement with the previous finding that the methyl group substitution at C-11 contributes to the biological activity in the regulation of virulence gene expression (24), we found that 11-Me-C12:Δ2,5 was superior to C12:Δ2 in the induction of EPS production in X. campestris pv. campestris (Fig. (Fig.4F).4F). The conserved distribution of DSF family signals in the Bcc provides further evidence that this group of related molecules represents a novel class of widely conserved bacterial cell-cell communication signals (10).
The funding for this work was provided by the Biomedical Research Council, Agency of Science, Technology, and Research (A*Star), Singapore.
Published ahead of print on 28 May 2010.