A Chemical Biology Approach to Interrogate Quorum Sensing Regulated Behaviors at the Molecular and Cellular Level

SUMMARY

Small molecule probes have been employed extensively to explore biological systems and elucidate cellular signaling pathways. In this study, we utilize an inhibitor of bacterial communication to monitor changes in the proteome of Salmonella enterica serovar Typhimurium with the aim of discovering new processes regulated by AI-2-based quorum sensing (QS), a mechanism of bacterial intracellular communication that allows for the coordination of gene expression in a cell density-dependent manner. In S. typhimurium, this system regulates the uptake and catabolism of intracellular signals and has been implicated in pathogenesis, including the invasion of host epithelial cells. We demonstrate that our QS antagonist is capable of selectively inhibiting the expression of known QS-regulated proteins in S. typhimurium, thus attesting that QS inhibitors may be used to confirm proposed and elucidate previously unidentified QS pathways without relying on genetic manipulation.

INTRODUCTION

Quorum sensing (QS) is a process of microbial intracellular communication that relies on the exchange of small chemical signals called autoinducers. This allows bacterial populations to coordinate their gene expression, providing effective cooperation or competition with multicellular organisms. The QS systems of many bacterial species regulate behaviors that are detrimental to human health, exemplified by the formation of microbial biofilms and the expression of virulence factors.(Miller and Bassler, 2001; Parsek and Greenberg, 2005) As such, QS has emerged as an intriguing target for the development of new antimicrobial therapeutics, which in turn has stimulated the discovery of QS-regulated pathogenic phenotypes such as biofilm formation and virulence factor secretion.

One class of QS signals, termed autoinducer (AI)-2, has been suggested as an interspecies signal. Indeed, AI-2 systems regulate an array of microbial processes, including bioluminescence in Vibrio harveyi,(Bassler, et al., 1994) autoinducer uptake and catabolism in Salmonella enterica serovar Typhimurium (S. typhimurium) and Sinorhizobium meliloti,(Pereira, et al., 2008; Taga, et al., 2001) virulence factor production in Vibrio cholerae and Vibrio vulnificus,(Kim, et al., 2003; Miller, et al., 2002) and the formation of mixed biofilms by Streptococcus oralis and Actinomyces naeslundii.(Rickard, et al., 2006) The two thus far identified AI-2 signals, (Figure 1A), are derived from a common precursor, 4,5-dihydroxy-2,3- pentanedione (DPD), which is produced by the enzyme LuxS. Importantly, these AI-2 signals exist in a dynamic equilibrium, and as such the DPD-based autoinducer produced by one species may be recognized by a different species, thus allowing for interspecies communication via the production of a single chemical signaling species.(Globisch, et al., 2012; Meijler, et al., 2004; Miller, et al., 2004) The luxS gene has been identified in over 70 bacterial species, lending further credence to the notion of AI-2 as an interspecies signal. Additionally, recent studies have employed genomic analysis to identify AI-2 receptors in a variety of bacteria. For example, orthologs of the LuxP receptor, initially identified in V. harveyi, were found in various Vibrionaceae (Rezzonico and Duffy, 2008) while LsrB-type receptors have been identified in a variety of bacteria in the Enterobacteriaceae, Rhizobiaceae, and Bacillaceae families.(Pereira, et al., 2009; Rezzonico, et al., 2012) The ubiquity of the AI-2 production and detection systems, along with the interconversion of the AI-2 signals, presents an opportunity to manipulate a variety of bacterial behaviors with a compact toolbox of DPD analogs. In fact, a series of alkyl- DPD analogs has been developed that exhibit inhibitory effects in S. typhimurium and variable effects in V. harveyi based simply upon the addition of methylene units to the DPD core (Figure 1B).(Lowery, et al., 2009; Lowery, et al., 2008; Roy, et al., 2010; Tsuchikama, et al., 2012)

Figure 1.

Structures of (A) AI-2 signals and (B) propyl-DPD, a QS antagonist against S. typhimurium.

The investigation of AI-2 regulated phenotypes usually involves the generation of genetic ΔluxS mutants followed by observation of altered behaviors. In the case of unequivocally established AI-2 regulated processes, the addition of exogenous DPD serves to restore the original phenotype due to the AI-2 dependence. However, this is not always the case, as several reported AI-2 regulated phenotypes may be the result of metabolic defects stemming from the central role of LuxS in the removal of toxic intermediates produced in the bacterial activated methyl cycle (AMC).(Heurlier, et al., 2009; Holmes, et al., 2009; Wilson, et al., 2012) This pathway is responsible for methionine recycling and the generation of activated methyl groups for the methylation of proteins, nucleic acids, and metabolites.(Vendeville, et al., 2005) Thus, without proper validation, it remains unclear whether the observed phenotypic change arises from the loss of AI-2 mediated communication or merely the accumulation of toxic byproducts via compromised metabolism.

To discover new QS regulated behaviors while avoiding metabolic interference with the AMC, it is conceivable to employ the aforementioned DPD analogs for the observation of AI-2- dependent phenotypes in species with uncharacterized QS networks. In an analogous vein, the brominated furanone QS inhibitor, (5Z)-4-bromo-5-(bromomethylene)-3-butyl-2(5H)-furanone, was utilized to study cell density-dependent processes in Bacillus anthracis.(Jones, et al., 2005; Jones, et al., 2010) Nevertheless, this method may also give rise to “red herrings” in that certain QS inhibitors may also interfere with essential bacterial processes and thus affect cell viability.(Lowery, et al., 2009) For example, the brominated furanones inhibit the biofilm formation of S. typhimurium, a commonly QS-associated process, but do so independently of the known Salmonella QS system.(Janssens, et al., 2008) Furthermore, many of the QScontrolled phenotypes may not be as pronounced as bioluminescence or biofilm formation and therefore the plethora of biological assays to discover altered phenotypes has the potential to be labor intensive. Thus, the combination of QS inhibitor and broad view analysis of the resulting physiological effects, such as transcriptomics, or proteomics, might provide a more direct roadmap to uncover cellular communication pathways.

Mass spectrometry-based proteomics is rapidly developing as a powerful platform to unravel functional proteins responsible for executing complex biological processes. A major application of this technology is the comparison between two samples that differ in physiological states, such as healthy vs. diseased, or wild-type vs. genetic mutant. This comparative analysis has been particularly useful in the study of disease states, in which an infected or compromised cell is compared against a normal cell to monitor key differences in protein expression levels, allowing for the identification of new markers for disease.(Cravatt, et al., 2007) QS represents an appropriate biological process for the application of MS-based proteomic analysis, as different phenotypes may be induced by the addition or removal of QS signals. Proteomics has been used to study luxS-controlled protein expression in E. coli, V. vulnificus, and Neisseria meningitidis(Schauder, et al., 2005; Shin, et al., 2007; Soni, et al., 2007); however, these efforts are characterized by the use of luxS knockout strains and the aforementioned caveats associated with them. Herein, a previously reported and validated QS antagonist was employed in the proteomic analysis of the common food-borne pathogen S. typhimurium with the goal of both elucidating unrecognized and confirming suggested AI-2 regulated pathways. The obtained proteomics data were also correlated with mRNA levels and functional assays were performed to compare QS regulation at the molecular and cellular levels.

RESULTS

MudPIT Analysis of the Effects of Propyl-DPD treatment

Protein levels were compared between control cells and cells treated with the AI-2 inhibitor propyl-DPD using a Multidimensional Protein Identification Technology (MudPIT)-based approach. Protein samples were isolated, reduced to break disulfide bonds, and alkylated to prevent their re-formation. The alkylated protein samples were then digested into peptides with trypsin and analyzed using electrospray ionization (ESI) in combination with tandem mass spectrometry (MS/MS) and database searching to identify the peptide sequences.(Washburn, et al., 2001; Wolters, et al., 2001) To examine the proteome-wide effects of propyl-DPD, and the inhibition of AI-2-based QS, wild-type S. typhimurium cells were treated with 25 µM propyl-DPD or solvent control (0.0167 % DMSO) and grown to mid-exponential phase as at this time point the expression of the lsr operon has been initiated.(Taga, et al., 2003) This concentration was selected because in our previous report we observed nearly complete inhibition of QS in a reporter strain of S. typhimurium at 25 µM. In the control samples, 1,962 proteins were identified in the soluble fraction (i.e. proteins that were soluble in the cell lysis buffer) with a minimum of two peptides per protein and 2,074 proteins were identified in the insoluble fraction (i.e. proteins that were solubilized in 90% formic acid). In the propyl-DPD treated samples, 1,905 proteins were identified in the soluble fraction and 1,953 proteins in the insoluble fraction (Figure 2). For our comparative proteomic analysis, we focused only on proteins with a minimum of 10 spectral counts averaged over the three independent analyses; thus, in the soluble fraction 772 proteins were analyzed and in the insoluble fraction 807 proteins were analyzed (Tables S1 and S2). Using total summed protein spectral counts as a semi-quantitative measure of abundance we identified 15 proteins in the soluble fraction and 1 in the insoluble fraction that were downregulated by a factor of at least 2 with a p-value < 0.01. If the statistical cutoff is relaxed to p < 0.05, 80 proteins in the soluble fraction and 5 proteins in the insoluble fraction were downregulated by a factor of at least 2 (Shown in Tables S3 and S4). We were also interested in proteins up-regulated by treatment with propyl-DPD, i.e. QS-repressed proteins; however, we did not identify any proteins that were up-regulated at least 2 fold in a statistically significant manner.

Figure 2.

Comparison of proteins identified across three biological replicates.

Gratifyingly, of the down-regulated proteins that met our criteria, 4 of them were known AI-2 regulated proteins (LsrK, LsrB, LsrF, LsrA), supporting the quality of data yielded by our approach (Table 1). In fact, all of the known AI-2 regulated proteins with a minimum of 10 spectral counts were down-regulated by treatment with propyl-DPD. The Lsr family of proteins is responsible for the uptake and processing of AI-2 signals, presumably allowing S. typhimurium to interfere with the communication of other species. The observed down-regulation of the Lsr family of proteins is consistent with the effect of propyl-DPD in a reporter assay using a lacZ fusion to the lsr operon.(Lowery, et al., 2008) We also conclude that QS inhibition does not occur through down-regulation of AI-2 production as both LuxS, the AI-2 synthase, and Pfs, the enzyme responsible for producing the LuxS substrate S-ribosylhomocysteine from the toxic intermediate S-adenosylhomocysteine, were not affected. This finding is unsurprising, as transcription of luxS and pfs is not regulated by AI-2, but rather pfs is under the control of a regulatory network involved in methionine metabolism and luxS is constitutively expressed. Thus, AI-2 production is dependent on Pfs activity and consequently LuxS substrate availability; that is, increased expression of pfs results in greater levels of AI-2. As such, AI-2 production appears to depend on the metabolic state of the cell rather than bacterial population density.(Beeston and Surette, 2002)

Table 1.

Overview of QS related proteins affected by 25 µM propyl-DPD.

	Soluble Fraction		Insoluble Fraction
Protein Name	Expression Ratioa	T-test	Expression Ratio	T-test
LsrA	n.d.b	n.d.	− 5.55	0.01
LsrB	− 4.59	0.01	− 2.43	0.12
LsrF	− 1.86	0.17	− 3.17	0.03
LsrK	− 4.88	0.02	− 2.21	0.14

^aExpression ratio = (average spectral counts control) / (average spectral counts propyl). Ratios are represented as negative values as they represent down-regulation.

^bn.d. = not detected in the MudPIT experiments at a minimum value of 10 spectral counts

See Tables S1 and S2 for a complete list of proteins registering at least 10 spectral counts.

In addition to AI-2 biosynthesis, LuxS and Pfs are involved in the AMC of many bacteria. Although LuxS and Pfs were not directly affected, other proteins in the AMC were down-regulated: for example, MetH, the enzyme responsible for the conversion of homocysteine to methionine, was 4 fold down-regulated. Additionally, three S-adenosylmethionine(SAM)-dependent transferases, Mod (4 fold), RsmH (3 fold), and UbiE (2 fold) were also down-regulated. These enzymes are involved in a variety of metabolic pathways, and transfer a methyl group from SAM to produce S-adenosylhomocysteine, which in turn is the substrate for Pfs. As such, it is evident that treatment with propyl-DPD interferes with the AMC, the process required for AI-2 production. Although the AMC is often critical for proper bacterial growth, treatment with propyl-DPD does not affect the growth of wild-type S. typhimurium (Figure S1).(Lowery, et al., 2008; Roy, et al., 2010) Furthermore, in the case of S. typhimurium, the deletion of luxS does not result in any growth deficiencies compared to the wild-type strain.(Beeston and Surette, 2002)

Besides the known AI-2 regulated proteins and enzymes involved in the AMC, several other proteins were down-regulated in response to propyl-DPD, most of which are involved in metabolic processes (Figure 3).(Gillespie, et al., 2011) Several proteins involved in nucleotide metabolism were affected: PurU, PurB, PurR, DnaN, Mfd, and YjeQ are involved in purine metabolism and DnaN, Cmk, DeoA, and Dut are involved in pyrimidine metabolism. Proteins involved in translation were also affected by propyl-DPD. For example, two components of the 30S ribosomal subunit were down-regulated, RpsQ and RpsG. Two other proteins, PrfA and PrfB, which are involved in the termination of translation, were also down-regulated upon treatment with propyl-DPD.

Figure 3.

Breakdown of proteins affected by a QS antagonist, propyl-DPD. Protein classifications were determined using the PATRIC database. (Gillespie, et al., 2011) See also Tables S3 and S4 for a complete list of proteins affected by a factor of 2 with a p-value ≤ 0.05.

We were also interested in the effects of propyl-DPD on the proteins involved in the pathogenesis of S. typhimurium, as previous reports have suggested a link between AI-2-based QS and invasion.(Choi, et al., 2012; Choi, et al., 2007) An examination of proteins in the Salmonella pathogenicity island 1 (SPI-1) revealed that, in general, proteins involved in bacterial invasion and protein secretion are unaffected. However, the invasion proteins InvA, InvC, and InvJ and the transcriptional regulator HilD are all down-regulated at least 2 fold, albeit with p-values between 0.05 and 0.10.

RT-PCR analysis of selected targets

Because the data obtained from the MudPIT experiments pertain to changes at the protein level, we also sought to characterize the effects of propyl-DPD on gene transcription. Towards this end, we treated wild-type S. typhimurium with 25 µM propyl-DPD as for the protein analysis. In these assays bacteria were also grown to mid-exponential phase for comparison to the proteomics data as well as the fact that the expression of invF, a gene encoding one of the primary regulators of SPI-1 genes, significantly increases during exponential growth.(Choi, et al., 2007) Total RNA was isolated, and real-time PCR was performed to measure the transcription levels of several genes corresponding to proteins identified above, implicated in quorum sensing, or associated with S. typhimurium pathogenesis (Table 2). First, lsrA was chosen to confirm the effects of propyl-DPD on the established AI-2-regulated lsr operon. As anticipated, lsrA was down-regulated 3 fold in the presence of propyl-DPD, a result that was expected based on the proteomics data as well as the inhibition of gene transcription in the lsr operon by other known AI-2 inhibitors.(Garner, et al., 2011) The flgN, purB, and yeiG genes were selected in an attempt to correlate down- and up-regulated proteins with the corresponding genes. The absence of activity against purB and flgN expression was unanticipated based on the proteomics results, and may speak to a post-transcriptional level of regulation of the corresponding proteins (vide infra). Additionally, the yeiG gene, selected because its protein product was only one of a few up-regulated proteins affected in a manner approaching statistical significance; however, it was also unaffected. In this case, the lack of activity of propyl-DPD likely speaks to the fact that the corresponding protein levels were not affected in a statistically significant manner.

Table 2.

RT-PCR data of selected targets.

Gene	mRNA expression levela	Standard Deviation	RT-PCR p-value	Protein expression level	MudPit p-value
lsrA	−3.33	0.57	0.003	−5.55c	0.01c
flgN	+1.14	0.14	0.04	−3.13b −1.29c	0.02b 0.32c
flhD	+1.38	0.22	0.02	n.d.d	n.d.
fliC	+1.52	0.65	0.09	+1.02b −1.41c	0.94b 0.54c
invA	−1.49	0.16	0.0003	−4.00b +1.05c	0.07b 0.84c
invF	+1.16	0.22	0.27	n.d.	n.d.
purB	−1.35	0.22	0.008	−4.00b	0.01b
sicA	+1.15	0.15	0.63	−1.48b −1.11c	0.32b 0.78c
yeiG	−1.13	0.11	0.27	+3.08c	0.07c

^aNegative Expression level = control sample / propyl-DPD sample. Positive Expression level = propyl-DPD sample / control sample. All expression levels were normalized to rrsH levels and represent the average of at least 3 replicates.

^bProteomic expression level and p-value from the soluble fraction.

^cProteomic expression level and p-value from the insoluble fraction.

^dn.d. = not detected in the MudPIT experiments at a minimum value of 10 spectral counts

We also sought to explore the effects of propyl-DPD on genes involved in the pathogenesis of S. typhimurium. A series of recent publications has linked the regulation of flagella as well as SPI-1-localized genes to luxS-dependent QS.(Choi, et al., 2012; Choi, et al., 2007) In these studies, fliC, sicA, and invF were down-regulated in a ΔluxS mutant, while fliC, flhD, sicA, and invF were also demonstrated to be under the control of LsrR, a transcription regulator that represses the lsr operon at low cell density. In our analysis, we also included invA as another representative of SPI-1 genes; although the corresponding protein levels were not affected in a statistically significant manner (−4 fold, p = 0.074), this was deemed a suitable target as the InvF protein was not detected in our proteomic analysis. As such, we reasoned that because propyl-DPD serves to inhibit the lsr operon, we would observe a down-regulation of these genes involved in virulence and motility. However, we did not observe a significant effect on flhD, fliC, invA, invF, or sicA. This lack of propyl-DPD activity posits different mechanisms of gene regulation in the lsrR and luxS mutants compared to the effects exerted by propyl-DPD.

Propyl-DPD affects invasion but not type III secretion in vitro

From our proteomic and genomic data, it would appear that inhibition of the AI-2 QS system does not impact the pathogenesis of S. typhimurium. Nevertheless, we sought to investigate the effect of Lsr inhibition on several pathogenic phenotypes of S. typhimurium to assemble a complete blueprint of QS inhibition at the transcriptional, translational, and phenotypic levels. Towards this end, we evaluated the effects of propyl-DPD on S. typhimurium adhesion and invasion of mammalian cells, as well as the secretion of virulence factors. To test the effect of QS inhibition on these behaviors, S. typhimurium (14028s) cells were treated with 25 µM propyl-DPD and adhesion and invasion of HeLa cells was monitored. It should be noted that 14028s strain was used in these assays whereas LT2 was used in the proteomics and RT-PCR experiments; however, we have not observed a difference in virulence between these two strains (S.I.M., personal communication). In these assays, propyl-DPD had no effect on adhesion, but a 35–50% reduction in invasion was observed (Figure 4A). As such, this finding not only links, albeit tenuously, AI-2-based QS and the invasion of epithelial cells, but also implicates propyl-DPD as a potential lead for the development of novel invasion inhibitors.

Figure 4.

Inhibition of Salmonella invasion and secretion by propyl-DPD. A) Inhibition of S. typhimurium invasion of epithelial cells upon treatment with 25 µM propyl-DPD. Data represents 5 independent experiments, each performed in triplicate. B) Secretion profile of S. typhimurium upon treatment with 25 µM propyl-DPD. C) Western blot analysis to monitor the effect of 25 µM propyl-DPD on SipC secretion.

Another prominent role of SPI-1 is type III secretion (T3S), which is responsible for the delivery of virulence factors into eukaryotic cells. AI-2-based QS has also been implicated in this process (Choi, et al., 2007) and again we sought to compare our proteomic findings with phenotypic observations. From our proteomics results, the proteins involved in secretion were not affected by treatment with 25 µM propyl-DPD. In accord with these results, propyl-DPD, even at concentrations up to 250 µM, had no effect on the Salmonella secretion profile (Figure 4B). To get a more focused vantage on the T3S, we examined the effects of propyl-DPD on the secretion of the effector protein SipC, which is required for the translocation of other T3S proteins into epithelial cells. Again, propyl-DPD did not affect the secretion of this specific protein (Figure 4C).

DISCUSSION

The AI-2 based QS system of S. typhimurium has been the subject of thorough investigation and a clear picture of its major genetic component, the lsr operon, has been established through a variety of genetic and biochemical studies. The lsr operon encodes proteins responsible for the uptake and catabolism of the AI-2 signal, which may serve to interfere with the communication of competing bacteria.(Xavier and Bassler, 2005) In addition to this well-defined connection between AI-2 QS and regulation of the lsr operon, several other processes have been linked to AI-2-based QS in S. typhimurium including virulence and biofilm formation.(Choi, et al., 2007; De Keersmaecker, et al., 2005) The potential for a broad regulatory role of QS in S. typhimurium, especially as it relates to pathogenesis, has spurred several efforts to map cellular pathways influenced by AI-2. One effective approach has been the use of microarray technology, (DeLisa, et al., 2001; Wilson, et al., 2012) but this only provides information at the genomic level rather than the functional protein level. MS-based proteomic analysis is a potent complement to genomic studies and has been effectively employed in the study of QS systems to decipher the molecular mechanisms of bacterial communication in several human pathogens including S. typhimurium.(Di Cagno, et al., 2011) However, these studies have relied on genetic luxS mutants to monitor proteins controlled by AI-2-based QS and proteins affected may be a result of a metabolic defect rather than the loss of intercellular communication. As such, we envisioned that by deploying the small molecule QS inhibitor propyl-DPD in the analysis of the Salmonella proteome we could monitor the affected proteins and establish a connection between these proteins and AI-2 regulation.

Because the activity of propyl-DPD was initially discovered and developed in reporter assays, we first sought to characterize its action on wild-type Salmonella to ensure that its activity stems from inhibition of the Lsr system rather than off-target effects. Indeed, the largest family of proteins that were down-regulated in this analysis was the Lsr family of proteins. At the time point used in our experiments, it is likely that the repressor LsrR is still active in the repression of the lsr operon.(Thijs, et al., 2010) While this may result in a lower total level of Lsr proteins observed in the treated and control samples, we expected to observe diminished Lsr protein levels in the propyl-DPD sample based on the proposed mechanism of the alkyl-DPD inhibitors, in which the analog binds to and stabilizes the LsrR repressor, resulting in lower QS activity.(Roy, et al., 2010)

Outside of the known QS regulated proteins, a variety of additional proteins were down-regulated upon treatment with propyl-DPD, the majority of which are involved in metabolic processes. This is in contrast with a study from Soni et al, who observed an up-regulation of proteins involved in carbohydrate transport and metabolism in a ΔluxS mutant.(Soni, et al., 2008) In a different report, only 4 proteins were differentially expressed, 1 of which was LuxS itself, and 2 others were unidentified proteins.(Kint, et al., 2009) The differences between the current study and the two previous proteomics-based studies of S. typhimurium are readily interpreted. For one, employing ΔluxS mutants in the previous studies compared to the use of a QS inhibitor in our study is a distinct difference. As discussed previously, it is possible that the deletion of luxS significantly alters metabolic fitness, resulting in differential protein expression.(Vendeville, et al., 2005) However, from our analysis, it is also apparent that propyl- DPD affects cellular metabolism, but neither deletion of luxS nor treatment with propyl-DPD has an obvious effect on cell viability (Figure S1).(Taga, et al., 2003) As such, we cannot explicitly conclude that the differentially expressed proteins are a direct result of QS inhibition rather than off-target effects, a result of metabolic deficiencies in the cell, or a compensation mechanism for compromised communication. We also applied a less stringent statistical analysis to our data to gain a broader insight into potential AI-2 regulated pathways that would subsequently be confirmed using both RT-PCR and functional assays. Using this enabled logic we cite a study by Kint et al, in which a p-value < 0.01 and fold increase/decrease of 1.5 was applied to minimize false-positives, whereas we applied a p-value < 0.05 and fold increase/decrease of 2.0 in attempt to gain a contrasting overview of affected proteins. Finally, it should be noted that these previous two studies utilized 2D gel electrophoresis, which does not allow for the identification of as great a number of proteins as MudPIT, to determine relative protein levels.(Washburn, et al., 2001)

For a more complete depiction of the AI-2 regulatory network, we sought to correlate and validate the proteins that were differentially expressed with the corresponding mRNA levels. However, when we examined the mRNA levels that correspond to several of the differentially expressed proteins, only lsrA was significantly affected. The fact that the other mRNAs were unaffected is somewhat surprising, as diminished protein levels would seemingly correlate with diminished mRNA levels. As such, it may be that propyl-DPD acts at a post-transcriptional level. One potential avenue for this level of control is via small RNA (sRNA) regulators. It is well established that sRNAs are intimately involved in the QS networks of V. harveyi and V. cholerae, and regulatory RNAs are known to control biofilm formation in Pseudomonas aeruginosa and thus play a critical role in the QS of several species.(Lenz, et al., 2004; Ventre, et al., 2006) Furthermore, in E. coli, an organism with an lsr system similar to S. typhimurium, the deletion of the QS regulators LsrR and LsrK affects the levels of several sRNAs.(Li, et al., 2007) In our studies, propyl-DPD reduces Salmonella LsrK levels, which may also affect the action of sRNA species. Interestingly, three down-regulated proteins (ArgT, STM4351, DppA) are under the control of the sRNA termed GcvB.(Hebrard, et al., 2012) Although a direct link has not been validated between the lsr operon and GcvB, it is intriguing to hypothesize such a connection based on our data.

Another plausible explanation for the lack of correlation between mRNA and protein levels lies in the down-regulation of proteins involved in translation. In this scenario, either the inhibition of the lsr operon or the action of propyl-DPD itself results in the down-regulation of these proteins, which in turn modulate the levels of several of the proteins observed in our study. Regardless of the propyl-DPD mechanism of action, these observations point to the need to examine QS networks at the genomic and proteomic levels to more fully understand intricate bacterial communication networks.

The examination of QS at both protein and mRNA levels may still prove insufficient for a complete picture of AI-2 regulatory pathways, as treatment with propyl-DPD did not have a clear effect on invasion proteins or mRNA levels but did result in diminished invasion of epithelial cells by S. typhimurium. As such, propyl-DPD is the first example of an antagonist of AI-2-based QS that affects S. typhimurium invasion. However, it is also evident that QS inhibition does not have an obvious effect on the other pathogenic processes of S. typhimurium, at least not under the conditions examined. In this light, it may be that propyl-DPD affects invasion at a level other than AI-2-based regulation.

These findings also expose a series of conflicting reports on the role of AI-2 in S. typhimurium virulence. On one hand, it has been demonstrated that both LuxS and LsrR are required for invasiveness and expression of SPI-1,(Choi, et al., 2012; Choi, et al., 2007) while another study concluded that LuxS-dependent signaling does not have any effect on invasion, nor does it play a role in type III secretion.(Perrett, et al., 2009) Furthermore, a recent report demonstrated that the AI-2-dependent transcriptional regulator LsrR directly acts only on the lsr operon and not any gene targets associated with Salmonella virulence.(Thijs, et al., 2010) As we did not observe a clear connection between QS inhibition and pathogenesis, our data support the notion that the only unequivocally defined AI-2 regulated behavior in S. typhimurium remains AI-2 uptake and catabolism, at least in the LT2 strain under the conditions examined herein. At this point it must be acknowledged that LT2 is less virulent than the wild type strains SL1344 and 14028, so the possibility remains that the observed variation between proteomic/genomic vs. phenotypic data may be due to strain differences.(Garcia-Quintanilla and Casadesus, 2011)

This dynamic continues to beg the question as to why an organism would release a chemical signal only to destroy it later? One current hypothesis is that the removal of AI-2 from the environment by S. typhimurium and other enteric bacteria serves to interfere with the communication of neighboring bacteria, thus providing a competitive advantage to the consumers.(Xavier and Bassler, 2005) Another possibility is that the detection of AI-2 allows the bacterial population to monitor extracellular conditions such as diffusion rates and spatial cellular distribution.(Hense, et al., 2007; Platt and Fuqua, 2010; Stacy, et al., 2012) In this scenario, S. typhimurium may alter cellular processes in response to AI-2, but in an indirect manner rather than regulation through the AI-2-dependent lsr operon. Finally, it cannot be ruled out that the lsr system of S. typhimurium is simply acting to recycle a metabolite that has been excluded from the cell to alleviate high intracellular concentrations. (Williams, et al., 2007; Winzer, et al., 2002; Winzer, et al., 2003) For example, the Lsr system in S. typhimurium produces 2-phosphoglycolic acid, which is a useful intermediate in general metabolism as demonstrated in E. coli.(Teresa Pellicer, et al., 2003; Xavier, et al., 2007) Furthermore, in S. typhimurium production of AI-2 does not correlate with cell density but rather with cellular metabolism (vide supra), lending further credence to a metabolic role of AI-2 in this organism.(Beeston and Surette, 2002)

SIGNIFICANCE

We have verified that one of our small molecule AI-2 inhibitors selectively down-regulates proteins involved in the well-characterized lsr-regulated QS system of S. typhimurium, serving as proof-of-principle that QS inhibitors developed in a reporter strain are also effective in wild-type strains. Furthermore, the selective targeting of the Lsr family of proteins points to the approach of using AI-2 antagonists in proteomic analyses to effectively map out AI-2 regulated processes in species with uncharacterized systems. Notably, this does not require genetic manipulation, which does irrevocably affect the overall physiological state of the bacteria, namely the introduction of a selection marker that might exert an additional metabolic burden onto the cell, but rather just a compact toolbox of chemical implements. While these synthetic molecule probes may still effect overall bacterial physiology, they represent useful tools as they allow for temporal control of QS modulation and the ability to examine differential levels of modulation based on inhibitor concentration rather than simply an on/off status achieved with genetic manipulation. Nevertheless, due to the intertwined nature of AI-2-based QS and metabolism, it is evident that there is not a single approach that can be employed that is free from auxiliary metabolic side effects. Furthermore, at least in S. typhimurium, it is apparent that there is not necessarily an obvious connection between inhibition of transcription, protein translation, and phenotypic observations. As such, this obscurity speaks to the complexity of the AI-2 regulatory system and the necessity to take a holistic approach to the study and discovery of previously uncharacterized AI-2-controlled pathways.

EXPERIMENTAL PROCEDURES

Multidimensional Protein Identification Technology (MudPIT) of Bacterial Proteomes

An overnight culture of S. typhimurium LT2 (ATCC 700720) was diluted 1:100 into 20 mL of LB medium in 50 mL Falcon tubes. To the medium was added 100 µL propyl-DPD (synthesized as previously reported(Lowery, et al., 2008)) from a 5 mM stock in 0.5 M phosphate buffer containing 3.33 % DMSO (Final concentration: 25 µM propyl-DPD, 0.0167 % DMSO, 2.5 mM phosphate buffer). A solvent control sample containing 0.0167 % DMSO and 2.5 mM phosphate buffer was also prepared. The samples were incubated at 37°C with shaking and grown to an OD₆₀₀ of 1.3–1.5 (about 4 h). At this point, the tubes were centrifuged at 2400 × g at 4°C, the supernatant was discarded, and the cell pellet was resuspended in ice-cold PBS buffer (pH 7.4). The centrifugation and washing process was repeated two more times, and the cell pellet was isolated and stored at -80°C until MudPIT analysis.

Proteomic samples were generated by taking bacterial samples and first treating them to three rounds of freezing-thawing in the presence of protease inhibitors. Samples were then resuspended in 50 mM Tris (pH 8.0) with protease inhibitors, sonicated on ice, and centrifuged at 20,000 × g for 30 minutes at 4°C to generate the soluble and insoluble proteome samples. 25 µg of soluble proteome sample in a final solution of 8M urea/50 mM Tris (pH 8.0) were used for MudPIT analysis. First, soluble proteome samples were reduced with 10 mM tris (2-carboxylethyl) phosphine (TCEP) and alkylated with fresh 12 mM iodoacetamide, in the absence of light, for 30 minutes each. The concentration of urea was reduced to 2 M by dilution with 50 mM Tris (pH 8.0) for digestion. Digestions were performed for 12 hours by incubation with trypsin (5 µL of a 0.5 µg/µL solution) in the presence of 2 mM CaCl₂ at 37°C. Tryptic peptide samples were then acidified to a final concentration of 5 % formic acid and loaded onto a biphasic (strong cation exchange/reverse phase) capillary column for MudPIT analysis. Peptides were separated and analyzed by 2D LC in combination with MS/MS as previously described using an 11-step gradient,(Washburn, et al., 2001) and MS/MS analysis was performed using an LTQ-mass spectrometer (Thermo Scientific). Peptide identifications were yielded from the MS/MS spectra by searching each against a protein database using the search algorithm SEQUEST. Data were then compiled and organized using DTASelect software.

RT-PCR Analysis

S. typhimurium LT2 samples were treated with 25 µM propyl-DPD or solvent control as described above in the proteomics experiments, incubated at 37°C with shaking, and grown to an OD₆₀₀ of 1.3–1.5 (about 4 h). Cells were then harvested by centrifugation and total RNA was isolated using the RNAeasy Mini Kit (Qiagen). Reverse transcription was performed using the QuantiTect Reverse Transcription Kit (Qiagen). Relative quantification analyses were performed using the QuantiTect SYBR Green PCR Kit (Qiagen) with an Applied Biosystems 7900HT Fast Real-Time PCR System and the housekeeping gene rrsH as a reference.(Choi, et al., 2012) The primers used are listed in Table 3.

Table 3.

Primers used in this study.

Primer	Oligonucleotide sequence (5’ → 3’)
lsrA forward	GGCAATGGTGCGGGTAAATCAACA
lsrA reverse	GAGGCGTTAAATGACTGCAACGCA
flgN forward	AGCGCAAACGATGACATTGCAGAG
flgN reverse	GCGCCTGTTGATTACGCTCGATTT
flhD forward	CCTCGGTATCAACGAAGAGATG
flhD reverse	GATGATCGTCAAACCGGAAATG
fliC forward	TGAACGAAATCGACCGTGTA
fliC reverse	CGATAGTTTCACCGTCGTT
invA forward	ATTATCGCCACGTTCGGGCAATTC
invA reverse	ACGATAAACTGGACCACGGTGACA
invF forward	GCAGGATTAGTGGACACGAC
invF reverse	TTTACGATCTTGCCAAATAGCG
purB forward	ACTGGCATCAGTTCAGCGAAGAGT
purB reverse	AAAGCGCGCGATACAGTCAAACAG
sicA forward	ATTTGGGATGCCGTTAGTGAAG
sicA reverse	TAAACCGTCCATCATATCTTGAGG
yeiG forward	ACGCTAAACTGCGCCATGACATT
yeiG reverse	AATACAATGCCCAGTTCTGCTGCG
rrsH forward (16s rRNA, housekeeping)	GAGCAAGCGGACCTCATAAA
rrsH reverse (16s rRNA, housekeeping)	GGCATTCTGATCCACGATTACTA

Invasion assays

The presence (25 µM) or absence of propyl-DPD was kept constant throughout the assay. Cultures of S. typhimurium 14028s grown to exponential phase were used to infect HeLa cells at a multiplicity of infection of 10:1. Invasion was allowed to proceed for 15 min at 37°C with 5 % CO₂. HeLa cells were washed twice with Dulbecco’s modified Eagle medium containing 10 % heat-inactivated fetal bovine serum and gentamycin (15 mg/ml) and further incubated for 1 h in the same medium. After washing twice in 1× PBS, cells were lysed in 1× PBS containing 0.1 % Triton X-100. Bacterial cells were enumerated by plating dilutions on LB agar plates.

Analysis of secreted proteins

Secreted proteins were purified from S. typhimurium 14028s overnight cultures grown in LB in the presence of 0, 25 or 250 µM propyl-DPD. After centrifuging the cultures at 15,000 × g for 5 min, the supernatants were filtered through a 0.2 µM syringe filter. Pre-chilled TCA was added to a final concentration of 10%, samples were placed on ice for 30 min and then centrifuged at 15.000 × g for 30 min. After drying, the pellets were dissolved in SDS sample buffer, boiled for 10 min and analyzed by SDS-PAGE or Western blot using anti- SipC antibodies (ABIN335178).

HIGHLIGHTS.

• Known quorum sensing proteins were down-regulated specifically in response to a QS antagonist

• Developed an approach to map QS networks without reliance on genetic manipulation

• No observation of clear link between QS signaling and virulence in S. typhimurium

• Demonstrates the necessity of examining QS networks at genomic, proteomic, and phenotypic levels