The Cyclic AMP Receptor Protein Regulates Quorum Sensing and Global Gene Expression in Yersinia pestis during Planktonic Growth and Growth in Biofilms

Bacterial pathogens have evolved extensive signaling pathways to translate environmental signals into changes in gene expression. While Crp has long been appreciated for its role in regulating metabolism of carbon sources in many bacterial species, transcriptional profiling has revealed that this protein regulates many other aspects of bacterial physiology. The plague pathogen Y. pestis requires this global regulator to survive in blood, skin, and lungs. During disease progression, this organism adapts to changes within these niches. In addition to regulating genes for metabolism of nonglucose sugars, we found that Crp regulates genes for virulence, metal acquisition, and quorum sensing by direct or indirect mechanisms. Thus, this single transcriptional regulator, which responds to changes in available carbon sources, can regulate multiple critical behaviors for causing disease.

Crp-regulated genes dependent upon growth in biofilms and under glucose-limiting conditions.
To recapitulate human infection conditions, Y. pestis was grown at 37°C with shaking in defined liquid culture media with glucose to mimic early pneumonia or with glycerol to mimic later infection after glucose is consumed. As the cultures were aerated by shaking, Y. pestis formed a biofilm at the air-liquid interface, thereby facilitating comparisons of levels of gene expression of the planktonic state representing early infection and the biofilm state of later infection (Fig. 1B). In addition, a Δcrp mutant was included in parallel to facilitate identification specifically of Crp-regulated genes under these conditions. The combined set of six experimental conditions allowed identification of glucose-, Crp-, and biofilm-dependent genes that may play a role during pneumonic plague infection.
Carbon source and growth state do not affect cAMP requirement for Crp binding to DNA. A potential concern with this experimental setup is that differences in expression of Crp-regulated genes could be affected by changes in the activity of CyaA and the phosphotransferase system (28), leading to differences in Crp activity rather than in expression of crp. To control for this possibility, we measured cAMP concentrations in Y. pestis under all experimental conditions. In contrast to the expected increase, we observed a decrease in cAMP concentrations in Y. pestis grown in glycerol (Fig. 1C). This was also reported in a study of Vibrio fischeri, in which intracellular cAMP levels were reduced but total concentrations of cAMP (including extracellular cAMP) were higher (29). Deletion of the cyaA adenylate cyclase gene reduced cAMP levels, while deletion of crp had no effect. cAMP concentrations were much higher in cells grown at 37°C versus 26°C, suggesting the importance of Crp regulation during mammalian infection. cAMP concentrations were not significantly affected in comparisons between plankton-and biofilm-grown cells (Fig. 1D). These data indicate that any changes in expression of Crp-regulated genes observed by RNA sequencing (RNA-seq) would not be due to differences in the cAMP levels.
Another potential caveat concerning this experimental setup is the possibility that Crp could bind and regulate genes in Y. pestis entirely independently of cAMP. Earlier mutational studies on Escherichia coli showed that the specific double mutant variant T128L/S129I had extremely high cAMP-independent DNA binding affinity, comparable with the activity of cAMP-bound wild-type Crp (30). Y. pestis Crp shares 99% sequence homology with Crp from E. coli, with differences located on the C helix at positions 119, 123, and 127. We considered that the sequence differences might affect cAMP binding or dimerization and thus cause changes in DNA binding. To investigate this, we carried out structural studies using protein crystallography and solved the crystal structure of Y. pestis Crp in complex with cAMP. Crp was crystallized as a dimer, and each monomer had cAMP bound to the binding site ( Fig. 2A). We compared the resulting structure to the structure of E. coli Crp bound to cAMP (PDB identifier [ID] 4R8H) (30). The structures aligned closely, with root mean square deviation (RMSD) values of 0.8 Å for 185 pruned C␣ atom pairs and 1.1 Å for all 200 pairs (Fig. 2B). Notably, in the Y. pestis structure, C helix residues S119, N123, and I127 (unlike S129 in the constitutive binding mutant of E. coli Crp) are directed away from cAMP, suggesting that these altered amino acids should not impact cAMP binding (Fig. 2C). The side chain of residue N123 is involved in dimer formation and directly interacts with the side chain of E78 from another chain. The N123 exchange for arginine, which occurs in Y. pestis, disrupts this interaction, but the S119 replacement for alanine restores the hydrogen bond to E78; thus, these two mutations do not alter the stability of the dimer. V127 in Crp from E. coli is pointed away from dimerization interface, and it is not involved in the cAMP binding, so isoleucine at this position in Y. pestis does not have any impact that contributes to binding of cAMP or to dimer formation. As confirmation of our findings, an electrophoretic mobility shift assay (EMSA) performed in the presence of cAMP (see Fig. 5A) indicated that Crp requires cAMP to bind a DNA fragment that corresponds to the pla promoter. This cAMP dependence is also consistent with studies of Crp from other Y. pestis isolates and from Yersinia pseudotuberculosis (23,31,32).
Thus, cAMP concentration data, structural comparisons, and EMSA data suggest that Crp function is controlled by cAMP in Y. pestis but that the cAMP concentrations did not  vary dramatically under the conditions surveyed in this study. Thus, we expect that variations in gene expression noted in this study would be linked specifically to Crp. This expectation is consistent with our prior finding that regulation of pla (between planktonic and biofilm-grown cells) depends on differential expression of Crp as opposed to changes in cAMP (26).
Global analysis of gene expression. We subsequently performed RNA-seq on total RNA isolated from Y. pestis pCD1 and the Δcrp mutant under all experimental conditions. RNA from Y. pestis Δcrp growing planktonically and in biofilms was collected to identify Crp-regulated genes. Deep sequencing reads were mapped to the Y. pestis chromosome and plasmids pPCP1 and pMT1. Reads were also mapped to the location of annotated small RNAs (sRNAs) in Y. pestis (33). The raw RNA-seq data sets were deposited into GenBank (accession no. GSE135228). The cutoff value for significantly differentially expressed genes was set at log2 fold change greater than 1 or less than Ϫ1, with a false-discovery-rate (FDR) P value of Ͻ0.05. Complete lists of differentially expressed genes can be found in Data Sets S1 and S2 in the supplemental material. Principal-component analysis revealed close associations of most of the replicates across all conditions (see Fig. S1 in the supplemental material). Crp and carbon source altered expression of thousands of Y. pestis genes in planktonic and biofilm growth states ( Fig. 3A to D). Indeed, 1,200 unique protein-coding genes (713 Crp-activated and 487 Crp-repressed genes) were impacted by the presence of crp whereas the presence of glucose altered expression of 1,872 unique genes between the planktonic and biofilm growth states.
Only 47 genes were differentially expressed strictly between plankton-and biofilmgrown cells (Fig. 3F). The low number of differentially expressed genes might be due to overlap of planktonic cells with biofilm cells at the air-liquid interface or to heterogeneous expression of genes throughout the biofilm. Genes required for biofilm formation such as those encoding diguanylate cyclases, e.g., hmsT and Y. pestis 0449 (ypo0449), were not differentially expressed between planktonic and biofilm cells (12,34,35). A similar observation in a study reported previously by Vadyvaloo et al. revealed that these genes were also not significantly different between planktonic cells and biofilms in flow cells (36). Genes previously known to influence biofilm formation, namely, crp (12), csrA (13), rcsAB (37), rovM and rovA (38), phoPQ (39), fur (40), and hfq (35), were also not differentially expressed. These data more likely suggest that genes for regulating biofilm formation are controlled by environmental factors, such as temperature and growth in the flea (41,42). Carbon source is also known to play a role (13), and while we observed no difference in the levels of expression of the diguanylate cyclases in biofilms, they were significantly upregulated in glycerol compared to glucose (Data Set S1). An additional possibility is that the biofilms observed in vitro are the result of adhesion and autoaggregation of Y. pestis cells to each other and to the flasks due to the presence of the Ail outer adhesin. Expression of the ail gene was downregulated in Y. pestis biofilms in glucose and glycerol as shown by RNA-seq, but we cannot discount the possibility that sufficient Ail was still present to account for the biofilm or aggregation.
Significant genes were categorized and enriched by biological process with Gene Ontology (GO) at GeneOntology.org by the use of Y. pestis as a reference list (Data Set S3) (43,44). Categorizing Crp-activated genes in planktonic or biofilm cells returned results that demonstrated enrichment of DNA-templated transcriptional regulators involved in a wide range of biological processes, including catabolism of carbohydrates (malT and araC), amino acid metabolism (leuO), and quorum sensing (ypeR and yspR). This category was also enriched in glucose-repressed (i.e., glycerol-activated) genes with overlap of multiple transcriptional regulators, including malT and ypeR. We anticipated overlapping of Crp-activated and glucose-repressed genes, as well as overlapping of Crp-repressed and glucose-activated genes, on the basis of the function of cAMP-Crp. By filtering these data sets for genes increased in glycerol compared to glucose and increased in wild-type Y. pestis compared to Y. pestis Δcrp, we were able to better identify cAMP-Crp-activated (or repressed) genes (see Table S2 in the supplemental material). Importantly, the pla gene, known to be Crp-activated and glucoserepressed and active during the end stages of pneumonic plague (6,23), filtered into the Crp-activated category. Several Crp-regulated genes from a previous microarray of Y. pestis (pim, pst, ptsG, araF, rpoH, yfiA, and ompC) (32) were also identified. While GO analysis identified quorum sensing transcriptional regulators yspR and ypeR as Crp activated, the corresponding acyl-homoserine lactone synthetase genes (yspI and ypeI) also were identified as CRP activated (Table S2).
Furthermore, while we observed few differentially expressed genes in comparisons of Y. pestis in biofilms to Y. pestis in planktonic cells, genes and enriched pathways classified as activated or repressed by Crp or glucose differed between planktonic and biofilm cells. In other words, the determination of which genes or pathways are turned on by Crp or respond to changes in carbon source depends on whether Y. pestis is growing in a planktonic or biofilm state. We subsequently focused on identifying Crp-regulated genes that may play a role during the progression of pneumonic plague as Y. pestis forms biofilms and the environment switches from Crp repressive to Crp active.
We also observed overlapping of Crp-regulated genes identified here for Y. pestis and previously published expression profiling for Y. pseudotuberculosis compared to its isogenic Δcrp mutant (45). In total, 160 genes were shared across these data sets as differentially expressed in a manner dependent upon Crp, despite differing culture conditions and cutoffs for significance (Fig. 3E). An additional five genes (rseC, rpsL, rpsG, rpmA, and ybiT) were Crp activated in one species but Crp repressed in the other. Differences between the two data sets may also have resulted from the differing manners in which the crp gene is regulated among the two species. The small RNA chaperone Hfq is required for full production of Crp in Y. pestis but not in Y. pseudotuberculosis (6,31). The PhoP response regulator is an activator of crp expression in one strain of Y. pestis (21), but variation in the DNA-binding domain of PhoP between Yersinia strains alters transcription of its target genes (46). In addition, these differences in regulation of the crp gene and in the extent to which genes are regulated by Crp likely result from changes in the promoter region of genes and reflect adaptations to the different environments that Y. pestis and Y. pseudotuberculosis inhabit.
Crp indirectly represses expression of genes for yersiniabactin biosynthesis and uptake. Among all the genes identified as controlled by Crp in the RNA-seq data set, it was noted that Crp represses expression of genes for metal acquisition (Data Set S1). These included genes for biosynthesis and uptake of the iron siderophore yersiniabactin (Ybt), such as fyuA, the gene for the Ybt receptor, and ybtA, the gene for the transcriptional regulator that controls expression of the Ybt locus (Table 1) (47). Consequently, the levels of expression of the irp genes, required for Ybt production, and genes for heme transport were also decreased, but the genes for the Yfe and Feo transport systems were not significantly differentially expressed (Table 1). Crp also represses expression of ybtX (irp8 in strain CO92), a known virulence factor and trigger of inflammation during pneumonic plague (48,49).
Reverse transcription-quantitative PCR (qRT-PCR) results supported the RNA-seq results, as transcript levels of ybtA and fyuA were reduced in glycerol-grown cultures and were also increased in the Δcrp mutant, albeit not to a statistically significant degree ( Fig. 4C and D). Crp repression was observed in both the biofilm and planktonic states. EMSAs were also performed to determine whether Crp directly binds to the promoters of these genes. In contrast to Crp binding to the DNA sequence corresponding to the pla promoter in a cAMP-dependent manner (Fig. 5A), Crp did not bind to sequences corresponding to the ybtA or fyuA promoters ( Fig. 5B and C). Thus, the control of Ybt by Crp is likely indirect. Crp did not alter expression of fur or of the RyhB sRNAs, known regulators of iron acquisition (Data Sets S1 and S2).
It is surprising to find that ybt genes are Crp repressed whereas other virulence factors, such as pla and psa, are Crp activated. Iron acquisition is critical for Y. pestis pathogenesis in multiple infectious routes (50) and would have been predicted to be required in the potential iron-limiting environment of the lung. It is possible that Ybt is necessary early in pneumonia and expressed when glucose is plentiful and crp expression is low. As the pneumonia progresses to the proinflammatory phase and host cells die of pyroptosis (51), iron is acquired by other means or is more available and Crp turns off expression of the Ybt genes indirectly.
Crp is required for growth on nonglucose sugars and directly binds to the malT promoter. Another potential important set of genes controlled by Crp during infection consists of those essential for the use of alternative carbon sources as glucose is depleted in the lung. Many of the Crp-activated genes identified by RNA-seq and enriched pathways identified by GO analysis are involved in metabolism or transport (Data Sets S1 and S3). This is not surprising given the historical role of Crp in regulating the lac operon and genes for alternative sugar metabolism mechanisms (52). In our RNA-seq data, we found that Crp activated expression of ptsG, which is a gene that is required for acquiring glucose during pneumonia and that is also important during bubonic plague (26). Crp was also found to increase expression of malT, the transcriptional activator of maltose metabolism (53). Genes ptsG and malT were confirmed to be controlled by Crp, as transcript levels were lower in the Δcrp mutant than in wild-type Y. pestis as shown by qRT-PCR ( Fig. 4E and F), but transcript levels were not increased when grown in glycerol. This regulation was direct, as EMSAs demonstrated that Crp directly bound to the promoter sequences for ptsG and malT ( Fig. 5D and E). In addition, the binding of Crp to these promoters required cAMP.
The direct regulation of the malT promoter by Crp suggested that maltose could be an important alternative to glucose to support growth of Y. pestis. Indeed, while Y. pestis grew well in thoroughly modified Higuchi (TMH) medium with maltose (Fig. 6A), the Δcrp mutant did not grow with maltose as the sole carbon source. Deletion of malT also reduced growth, although this mutant grew better than Y. pestis Δcrp, suggesting that Crp activates expression of additional genes involved in maltose metabolism. Complementation of crp and malT restored growth of the Δcrp and ΔmalT mutants, respectively. In addition to the results seen with maltose, Y. pestis grew better in TMH medium supplemented with galactose than in TMH medium alone (Fig. 6B). Genes for galactose catabolism were not significantly different between wild-type Y. pestis and Y. pestis Δcrp, but genes in the araF operon for arabinose catabolism were found to be dependent on crp for expression (Data Set S1). In TMH medium supplemented with glucose, Y. pestis formed a biofilm at the air-liquid interface that resulted in lowered observed growth in this assay (Fig. S2A). Such robust biofilm formation was not observed in the presence of the other carbon sources. In contrast, Y. pestis Δcrp grew only in TMH medium with glucose (Fig. 6C), suggesting that Crp-activated genes are necessary for metabolism of nonglucose carbon sources. However, which carbon sources Y. pestis uses in the lungs after glucose is consumed is unknown. Crp directly and indirectly regulates expression of quorum sensing genes. A final class of potentially important Crp-regulated genes identified in this study consists of those associated with the LuxIR-based acyl-homoserine lactone (AHL) quorum sensing systems. The gene pairs in the Y. pestis genome are 100% identical to the AHL synthetase genes (yspI and ypeI) convergently transcribed with the receptors (yspR and ypeR) (Fig. 7A). Crp positively controlled expression of both the AHL synthetase genes and the AHL receptors (Table 2). qRT-PCR demonstrated that expression of yspI and ypeR in glycerol was increased in wild-type Y. pestis compared to Y. pestis Δcrp, similarly to the positive-control pla (Fig. 4A and G to H; not statistically significant). Moreover, expression of the AHL synthetase yspI gene was further increased during biofilm growth. A cAMP-Crp dependent shift of the DNA sequence corresponding to the ypeR promoter was observed by EMSA (Fig. 5E). However, Crp did not bind to yspI promoter sequence (Fig. 5F), suggesting that regulation of this gene is indirect.
In Y. pseudotuberculosis, expression of the same quorum sensing genes is autoregulated and regulated by each other and and Crp activated (31,54). Both sets of synthetases and receptors are expressed from separate promoters antiparallel to each other (Fig. 7A). To determine if a similar pattern occurs in Y. pestis, green fluorescent protein (GFP) reporters containing the promoter regions of yspR, yspI, ypeI, and ypeR were integrated into wild-type Y. pestis, Y. pestis Δcrp, Y. pestis with deletion of ypeI plus ypeR (ΔypeIR), Y. pestis with deletion of yspR plus yspI (ΔyspRI), or Y. pestis with deletion of all four AHL genes (ΔyspRI ΔypeIR). Expression of all four reporters was reduced approximately 50% in Y. pestis Δcrp, confirming RNA-seq and qRT-PCR data showing that Crp stimulates expression of the ype and ysp AHL genes (Fig. 7B to E). Expression of the P ypeI -GFP reporter showed no dependence upon ysp or ype genes. The P ypeR -GFP reporter expression was reduced in the ΔypeIR and ΔyspRI ΔypeIR mutants, suggesting that ypeR expression may be positively autoregulated, but the differences were not statistically significant. Exactly the opposite occurs in Y. pseudotuberculosis, wherein the ypeIR (ypsIR) genes repress their own expression at 37°C (54). Expression of the P yspI -GFP reporter was reduced 2-fold in Y. pestis ΔypeIR, suggesting that the reduction in yspI expression in Y. pestis Δcrp could have been due in part to reduced expression of ypeR (Fig. 4G). Similarly, the yspI homolog in Y. pseudotuberculosis is dependent on ypeR (ypsR in Y. pseudotuberculosis) (54) and expression of ypeIR and yspIR homologs in Y. pseudotuberculosis YPIII may also be Crp regulated (31).
Taken together, these data suggest direct regulation by Crp at the ypeR promoter and indirect regulation of yspI through ypeR resulting in stimulation of genes in glucose-limited media in biofilm. Despite the coding sequences of these genes being Crp regulates production of AHLs in vitro. Given the decrease in expression of yspI and ypeI in Y. pestis Δcrp, production of AHLs should also have been reduced. Cell-free supernatants from Y. pestis strains were collected during planktonic growth in vitro and Results of GFP reporter assays performed using the (B) P ypeI -GFP, (C) P ypeR -GFP, (D) P yspI -GFP, and (E) P yspR -GFP reporters integrated into wild-type Y. pestis and the Δcrp, ΔypeIR, ΔyspIR, and ΔypeIR ΔyspIR mutant strains. Background fluorescence was subtracted, and the remaining signal was normalized to wild-type Y. pestis, the value for which was set to 1. Data represent the means and SEM of results from three independent experiments. *, P Ͻ 0.05; **, P Ͻ 0.01 (from one-way ANOVA with Bonferroni's multiple-comparison test).

Crp-Dependent Global Gene Expression in Yersina pestis
® were used to stimulate the Rhizobium radiobacter AHL bioreporter. Levels of AHLs produced by Y. pestis increased with cell density during exponential-phase growth as expected for quorum sensing (Fig. S3A). In contrast, AHL production in Y. pestis Δcrp increased only slightly despite a 6-fold increase in bacterial density. Because the growth rate of Y. pestis Δcrp was found to be reduced, AHL concentrations were also compared between strains at similar cell densities. Even accounting for cell density, AHL production was reduced in the Δcrp strain (Fig. 8A). These data suggest that the reduction in AHL production was due to reduced ypeI and yspI expression and not to differences in cell density or growth of Y. pestis Δcrp. To determine if Crp alters production of specific AHLs, extracts were developed by thin-layer chromatography (TLC). Three spots were observed in all three strains that corresponded to 3-oxo-C6-homoserine lactone (3-oxo-C6-HSL), C6-HSL, and C8-HSL (Fig. 8C, top to bottom) as reported previously for Y. pestis (55). However, the areas of each of the spots from Y. pestis Δcrp were smaller. No spots were detectable in extracts collected from the ΔypeIR ΔyspRI mutant strain (ΔAHL, Fig. 8D).
To determine whether AHL molecules are produced during lung infection, bronchoalveolar lavage fluid (BALF) was collected from mice infected with fully virulent Y. pestis. AHLs were undetectable in BALF from mock-infected mice and mice at 24 h postinfection (hpi) (Fig. 8B). Only low levels of AHLs were detectable in BALF at 48 hpi, while BALF contained a 10-fold increase in AHLs at 72 hpi. Similar trends were present for AHLs from BALF developed by TLC plates (Fig. 8D). The significant increase in AHLs at 72 hpi correlates with the time frame during pneumonia when Y. pestis depletes glucose from the lungs, forms large biofilms, and, as a result, expresses crp at a high level (Fig. 1A) (26). The formation of biofilms in the lungs may provide a favorable environment for amplification of quorum sensing-based gene regulation. Indeed, yspI was more highly expressed in vitro in glycerol-grown biofilms than in glucose-grown biofilms (Fig. 4G). Taken together, these data provide strong evidence for Crpdependent production of AHLs and induction of quorum sensing during lung infection. Transcriptome studies have linked Y. pestis quorum sensing mutants to defects in  (75). Deletion of both synthetase-receptor pairs impairs the ability of Y. pestis to make biofilms and metabolize maltose (55,56). Both of these phenotypes are also Crp dependent (Fig. 6) (12) and suggest that connecting the Crp regulon to the quorum sensing regulon may allow Y. pestis to cooperatively regulate gene expression under conditions in which glucose is depleted and bacterial density is high.
While multiple measurements indicated that Crp regulates AHL-based quorum sensing, autoinducer-2 (AI-2)-based or LuxS-based quorum sensing is not Crp regulated in the planktonic state (Table 2; see also Data Set S1). The levels of AI-2 secretion were not significantly different between wild-type Y. pestis and Y. pestis Δcrp at equivalent cell densities in planktonic culture ( Fig. S3B and C). In contrast to AHLs, the concentration of AI-2 in BALF samples increased in a stepwise manner from 24 to 48 to 72 hpi as Y. pestis density grew within the lungs even prior to the period of increased expression of crp (Fig. S3D) (26). These data suggest that AI-2 production is not dependent on Crp but rather that bacterial cell density increases between 24 and 72 hpi.
Conclusions. In this study, we solved the crystal structure of Y. pestis Crp while determining its regulon under planktonic and biofilm conditions. While crp is structurally similar to the E. coli Crp gene, the modes of regulation of the crp gene differ among pathogenic Yersinia spp., which colonize distinct environmental niches. Overall, transcription profiling conducted in this study revealed comprehensive insight into how Y. pestis adapts to growth in the biofilm state and the role of Crp and carbon sources during this growth. This is specifically relevant in the lungs, where Crp, through sensing of depletion of glucose, may serve as a switch turning on or off multiple behaviors of Y. pestis as disease progresses (26) (Fig. 9).
A unique result revealed by our transcriptional profiling was the observed indirect repression of Ybt genes by Crp (Table 1; see also Table S2). The Ybt and irp genes are critical for iron acquisition, virulence, and inflammation during pneumonic plague (48)(49)(50). Expression of these genes may be higher during the first 48 hpi, when expression and activity of Crp are low (Fig. 1A). After 48 hpi, Crp activates a second set of virulence factors such as pla and psa as disease progresses, correlating with the biphasic nature of pneumonic plague (23,27,32). The regulation of iron acquisition genes is unique to Y. pestis compared to Y. pseudotuberculosis. Even though these closely related species share 100% identical Crp proteins, the expression of the crp gene and the environments inhabited by these species may afford different sets of regulated genes.
We have expanded the pool of directly Crp-activated genes in Y. pestis to include genes for quorum sensing and alternative metabolisms. As glucose becomes a limiting resource, genes for carbohydrate catabolism and uptake are activated to support growth of Y. pestis. Within biofilms in the lungs, AHL-dependent quorum sensing is likely amplified due to the close proximity of cells. Here, we report production of AHLs in the lungs of plague-infected mice correlating with the time during which expression of crp is highest and Y. pestis is growing within biofilms of the lungs. Thus, Crp is directly regulating not only its own genes but also other transcriptional regulators and regulons (Fig. 9). These additional pathways may direct additional Yersinia behaviors and may serve as a link to the examples of indirect regulation observed in our data set. Future experimentation performed on the differentially expressed mRNAs and sRNAs identified in this study will provide additional information on biofilm-regulated and Crpregulated genes and behaviors in Y. pestis and on their importance during pneumonia.

MATERIALS AND METHODS
Bacterial strains, media, and growth conditions. The bacterial strains, plasmids, and primers used in this study are listed in Tables S3 and S4 in the supplemental material. Y. pestis strains (6, 74) were passaged on Difco brain heart infusion (BHI) agar (Becton, Dickinson) and into 2 ml of BHI broth or of the defined media, TMH (57), supplemented with 0.2% glucose for overnight growth at 26°C. Unless otherwise stated, overnight cultures in TMH medium were subcultured at an optical density at 620 nm (OD 620 ) of 0.1 into 10 ml of TMH medium with 0.2% glucose in 125-ml Erlenmeyer flasks incubated at 37°C with shaking at 250 rpm for measuring planktonic and biofilm cell differences. Planktonic cells were removed from the flask with a serological pipette. Biofilms were scraped and resuspended in phosphatebuffered saline (PBS) warmed to 37°C. For growth assays, overnight cultures were subcultured at an OD 620 of 0.1 into 200 l of fresh TMH medium with 0.2% of indicated carbon sources at 37°C in a 96-well plate. Absorbance was measured with a Molecular Devices Spectramax M5 microplate reader. E. coli strains were passaged on Luria-Bertani (LB) agar or in LB broth at 37°C. Vibrio harveyi MM32 (75) was cultured in ATCC 2034 medium or LB with kanamycin. R. radiobacter was passaged on BHI medium with gentamicin. Ampicillin (100 g/ml), kanamycin (50 g/ml), and gentamicin (15 g/ml) were added as necessary.
Construction of new strains and plasmids. Y. pestis mutants containing deletions in crp, ypeIR, yspRI, and luxS were generated by lambda red recombination (6). The ΔypeIR ΔyspRI double mutant was generated by deletion of the ypeIR locus from the ΔyspRI mutant. Y. pestis Δcrp and ΔmalT were complemented in trans by cloning the crp or malT codons plus 500 bp flanking the genes into pUC18-R6k-miniTn7t by Gibson assembly. The resulting plasmid was integrated into the Tn7 site (58). GFP reporters were also integrated into the Tn7 site. Amplification of 500 bp upstream of the promoters for yspI, yspR, ypeI, and ypeR was performed by PCR, and the amplification products were adjoined to the codons of gfp by overlap extension PCR as described previously (35).
RNA extraction, sequencing, and qRT-PCR. Total RNA was isolated from Y. pestis cultures as described previously (59) except that RNA protect reagent (Qiagen) was added to planktonic or biofilm cells in PBS. RNA was treated with 1 l of Turbo DNase (Invitrogen Life Technologies) for 30 min at 37°C followed by addition of another 1 l for 30 min following the manufacturer's instructions.
Next-generation RNA sequencing was carried out at the Northwestern University NUSeq Core Facility. Total RNA was checked for quality and quantity on an Agilent Bioanalyzer 2100 and a Qubit fluorometer. An Illumina TruSeq stranded total RNA library preparation kit was used to prepare sequencing libraries from 500 ng of total RNA samples according to the manufacturer's instructions without modifications. This procedure includes depletion of rRNA by the use of a Ribo-Zero rRNA removal kit (bacteria), purification and fragmentation of the remaining RNA, cDNA synthesis, 3=-end adenylation, Illumina adapter ligation, and library PCR amplification and validation. An Illumina NextSeq 500 sequencer was used to sequence the libraries with the production of single-ended, 75-bp reads.
The quality of DNA reads, in fastq format, was evaluated using FastQC. Adapters were trimmed, and reads that were of poor quality or that aligned to rRNA sequences were filtered out. The cleaned reads were aligned to the Y. pestis genome (NC_003143.1) and plasmids pPCP1 (NC_003132.1) and pMT1 (NC_003134.1) using STAR (60). Read counts for each gene were calculated using htseq-count (61) in conjunction with a gene annotation file for Y. pestis. Read counts for sRNA were obtained using bedtools and the annotated locations reported previously by Schiano et al. (33). Normalization and differential expression were determined using DESeq2 (62). The cutoff for determining significantly differentially expressed genes was a false-discovery-rate (FDR)-adjusted P value of less than 0.05. Raw and processed data were deposited into the GEO database at the NCBI (see below).
For validation of RNA sequencing results, qRT-PCR was performed using a SuperScript III Platinum SYBR green one-step qRT-PCR kit (Thermo Fisher Scientific) with 25 ng of DNased RNA on a Bio-Rad iQ5 cycler with melt curve analysis following the manufacturer's instructions.
Purification of Y. pestis Crp. The amplified crp gene corresponding to residues 1 to 210 from Y. pestis CO92 (www.csgid.org; accession no. IDP97063) was cloned by Gibson assembly into the SspI site of expression vector pMCSG7 (63,76), which contains an N-terminal polyhistidine tag followed by the tobacco etch virus (TEV) protease cleavage site and the start codon of the crp gene. The resulting plasmid was transformed into E. coli BL21(DE3) (Magic). The bacteria were grown at 37°C and 200 rpm in 3 liters of Terrific broth until the OD 600 reached 1.6. Protein expression was induced with 0.6 mM IPTG (isopropyl-␤-D-thiogalactopyranoside), and cells were grown overnight with the shaking rate reduced to 180 rpm and the temperature to 22°C. Cells were harvested by centrifugation. The resulting cell pellet was resuspended in 120 ml of lysis buffer [10 mM Tris-HCl (pH 8.3), 500 mM NaCl, 1 mM Tris(2carboxyethyl)phosphine hydrochloride (TCEP), 10% (vol/vol) glycerol, 0.01% (vol/vol) IGEPAL CA630, Roche EDTA-free protease inhibitors (1 tablet/100 ml buffer)], and the suspension was frozen at -20°C until purification. The frozen suspension was thawed under the cold running water, sonicated, and centrifuged. The protein was purified in two steps using nickel (II) affinity chromatography followed by size exclusion chromatography as described previously (65). The polyhistidine tag was removed by incubation of the tagged protein with the recombinant TEV protease for overnight at 20°C. The resulting 58 mg of pure protein was at a final concentration of 13.7 mg/ml. Structure determination of Y. pestis Crp. For crystallization screening, we used a protein solution with a concentration of 6.6 mg/ml in a mixture containing 10 mM Tris-HCl (pH 8.3), 500 mM NaCl, 5 mM TCEP, and 1 mM cAMP. Crystallization drops were equilibrated at a 1:1 protein/reservoir solution ratio against 96 conditions/screen using commercially available PACT, PEG and PEG II suites (Qiagen). Diffraction-quality crystals of Crp were grown under PACT suite condition H12.
Prior to flash-cooling in liquid nitrogen, crystals of Crp were transferred into a 5-l drop of the reservoir solution that they had grown from. Data were collected on the LS-CAT 21-ID-F beamline at the Advanced Photon Source (APS) at Argonne National Laboratory. A total of 350 images were indexed, scaled, and integrated using HKL-3000 (66). Data collection and data processing statistics are listed in Table S1. The structure of Y. pestis Crp in complex with cAMP was solved by molecular replacement using Phaser (67) from the CCP4 suite (68). The structure E. coli Crp (PDB entry 3RYP) was used as a search model. The initial solution was processed by several rounds of refinement in REFMAC v.5.5, residues were mutated, and cAMP was added in Coot (69). Water molecules were generated using ARP/wARP (70), and the model refinement was continued in REFMAC. Translation-libration-screw (TLS) groups were created by the use of the TLSMD server (71) (http://skuld.bmsc.washington.edu/~tlsmd/), and TLS corrections were applied during the final stages of refinement. MolProbity (64; http://molprobity.biochem.duke.edu/) was used for monitoring the quality of the model during refinement and for the final validation of the structure. The final model and diffraction data were deposited in the Protein Data Bank (https://www .rcsb.org/; see below). The final model consists of two polypeptide chains which form a dimer. Chain A contains Crp residues 6 to 208, and chain B contains residues 9 to 208. The crystal structure includes one cAMP molecule bound to each monomer, 3 chloride ions, and 292 water molecules. Refinement statistics and the quality of the final model are summarized in Table S1. Molecular graphics procedures and alignment with E. coli Crp bound to cAMP (PDB ID 4R8H) (30) were performed with the UCSF Chimera package.
Gel shift assays. EMSAs were carried out using a LightShift chemiluminescent EMSA kit (Thermo Fisher Scientific). Biotinylated primers were used for PCR amplification of gene-specific promoters (Table S4). PCR products were concentrated and gel purified using Wizard SV Gel and PCR cleanup protocol (Promega). Binding reactions were performed with the manufacturer's binding buffer, 100 ng poly(dI-dC), and 20 fmol of biotinylated promoter fragment in a 20-l reaction mixture. Reaction mixtures contained increasing amounts (6.25 to 100 ng) of purified Crp protein and cAMP (Sigma) (2 mmol). Binding reactions were carried out following the manufacturer's instructions. Nylon membranes (Hybond-Nϩ; 30 cm) were developed using X-ray film.
GFP reporter assays. For GFP reporter assays, 2-ml volumes of overnight Y. pestis cultures in TMH medium with 0.2% glucose were subcultured at an OD 620 of 0.1 into 2 ml of TMH medium incubated at 37°C in a rotary drum for 6 h. Cultures were diluted to an OD 620 of 0.25 to 0.40, and 200-l aliquots were added in duplicate to a 96-well plate. Fluorescence was read on a Tecan Safire II plate reader, and results are reported as previously described (6).
AHL and AI-2 bioreporter assays. R. radiobacter carrying pZLR4 was grown overnight at 30°C. Cultures were diluted to an OD 620 of 0.1 in 1 ml of BHI medium. A 1-l volume of filter-sterilized supernatants of Y. pestis or 10 l of filter-sterilized BALF, previously collected from mice intranasally infected with wild-type Y. pestis (26), was added. Tubes were incubated at 30°C with shaking at 250 rpm for 5 h. Cultures were pelleted and resuspended in 500 l of Z-buffer (60 mM Na 2 HPO 4 , 40 mM NaH 2 PO 4 , 10 mM KCl, 1 mM MgSO 4 , pH 7.0). A 25-l volume of chloroform and a 12.5-l volume of 0.1% SDS were used to lyse the bacteria for 5 min. A 100-l volume of 2-nitrophenyl-␤-D-galactopyranoside (ONPG; 4 mg/ml) was added to initiate the galactosidase reaction, which was stopped by the addition of 250 l of 1 M Na 2 CO 3 . The reaction mixtures were centrifuged for 5 min at 13,000 rpm, and 500 l was added to a cuvette to record the OD 420 . Relative beta-galactosidase activity levels were expressed as ratios of OD 420 /OD 620 .
For TLC, 100-l volumes of culture supernatants or BALF were extracted twice with 100 l of ethyl acetate and concentrated to 20 l in a 60-Hz Savant SpeedVac DNA 100 concentrator (Thermo Fisher Scientific). Volumes of 1 l of culture supernatants or 7.5 l of BALF samples were spotted onto aluminum-backed C18-W silica plates (Sorbent Technologies) and developed in 60% methanol-40% water as described previously (55).
To measure AI-2 concentrations, overnight cultures of MM32 were diluted 1:500 and grown at 30°C for 1 h. A 180-l volume was added to a 96-well plate with 20 l of either culture supernatant or BALF. Plates were incubated at 30°C with shaking for 5 h. Optical density and luminescence were recorded on a Molecular Devices Spectramax M5 microplate reader. Reported data are from 3 h of incubation (culture supernatants) and 4 h of incubation (BALF).
Measurement of intracellular cAMP concentration. Overnight cultures of Y. pestis were diluted to an OD 620 of 0.1 in TMH medium with 0.2% glucose or 0.2% glycerol and grown for 6 h at 26°C or 37°C. Alternatively Y. pestis was grown in TMH medium overnight in 125-ml Erlenmeyer flasks to form biofilms. Volumes consisting of 0.2 OD 620 equivalents were centrifuged and lysed in 0.1 M HCl-0.5% Triton X-100 for 10 min. Cellular debris was pelleted by centrifugation, and the supernatant was stored at -20°C. A Direct cAMP ELISA kit (72) was used following the acetylation protocol. cAMP concentrations were normalized to protein content measured using a Pierce bicinchoninic acid (BCA) protein assay kit (Thermo Fisher Scientific).
Statistical analysis and graphing. Statistical analyses, including Student's t tests and one-way analysis of variance (ANOVA) with Bonferroni multiple-comparison tests, were performed using Graph-Pad Prism, version 5.0, with a P value of Ͻ0.05 as a cutoff for significance. Venn diagrams and volcano plots were generated in R (Version 9.0) using the VennDiagram (Version 1.6.20) and tidyverse (Version 1.1.1) packages.
Gene ontology analysis. RNA-seq data were processed at GeneOntology.org (43,44) using GO enrichment analysis (73). The PANTHER overrepresentation test (released 11 July 2019) was conducted with the GO Ontology database (released 3 July 2019). The reference list used was that of Yersinia pestis.
Data availability. Raw and processed data were deposited in the GEO database at the NCBI (GEO accession number GSE135228). The final model and diffraction data for Crp were deposited in the Protein Data Bank (https://www.rcsb.org/) with the assigned PDB entry 6DT4.