Community behavior and spatial regulation within a bacterial microcolony in deep tissue sites serves to protect against host attack

Summary

Bacterial pathogens express virulence-specific transcriptional programs that allow tissue colonization. Although phenotypic variation has been noted in the context of antibiotic exposure, no direct evidence exists for heterogeneity in virulence-specific transcriptional programs within tissues. In a mouse model of Yersinia pseudotuberculosis infection, we show that at least three subpopulations of bacteria develop within a single tissue site in response to distinct host signals. Bacteria growing on the exterior of spleen microcolonies responded to soluble signals and induced the nitric oxide (NO)-detoxifying gene, hmp. Hmp effectively eliminated NO diffusion and protected the interior bacterial population from exposure to NO-derived inducing signals. A third subpopulation, constituting the most peripherally-localized bacteria, directly contacted neutrophils and transcriptionally upregulated a virulence factor. These studies demonstrate that growth within tissues results in transcriptional specialization within a single focus of microbial replication, facilitating directed pathogen counterattack against the host response.

Introduction

Phenotypic variation within microbial populations has been described for decades in the context of antibiotic exposure (Bigger, 1944; McCune Jr and Tompsett, 1956). Studies of microbial growth in tissue sites, however, have largely relied on averaging microbial behavior over an entire population. Recent advances in imaging, cell sorting, and sequencing technologies have enabled the identification of microbial subpopulations during disease (Ackermann et al., 2008; Burton et al., 2014; Dandekar et al., 2012; Diard et al., 2013; Helaine et al., 2014; Uliczka et al., 2011). The importance of analyzing subdivided populations during growth in a single tissue site is emphasized by the behavior of the Salmonella enterica serovar Typhimurium type-III secretion system-1 (SPI-1) after oral inoculation. The secretion system is expressed in a subset of bacterial cells, even though it is a central virulence factor (Diard et al., 2013; Sturm et al., 2011; Thiennimitr et al., 2011). It is not clear if the presence of distinct subpopulations is a consequence of growth in spatially disparate niches, or if host signals drive subdivision of bacteria. Individual bacterial subpopulations have an impact on virulence within invertebrates (Raymond et al., 2012), but whether such subpopulations are required to perform specialized functions driving diseases in mammalian hosts is unknown (Burton et al., 2014).

During systemic disease by invasive enteric pathogens, microbes set up residence in organs such as the liver and spleen. This can be modeled in mice using Yersinia pseudotuberculosis, which spreads systemically from intestinal tissues, and replicates extracellularly in the spleen and liver (Simonet et al., 1990). In these tissues, Y. pseudotuberculosis forms clusters of bacterial microcolonies surrounded by host phagocytes. The type III secretion system (T3SS) and associated Yop effector proteins promote extracellular growth by intoxicating host cells, preventing phagocytosis and interfering with reactive oxygen species (ROS) production (Songsungthong et al., 2010). Expression of the T3SS, and many additional bacterial factors, are required for virulence, although it remains unclear what additional pathways allow the bacterium to battle the host immune response.

Toxic reactive nitrogen species (RNS) and reactive oxygen species (ROS) play major roles in protecting the host from bacterial infections, by directly killing bacteria (Shiloh et al., 1999; Shiloh and Nathan, 2000). ROS are primarily generated by host phagocytes to eliminate pathogens during phagocytosis, and can also be released from host cells. In contrast, a larger number of host cells are capable of generating RNS in the form of nitric oxide (NO), which is a freely diffusible gas (Bogdan et al., 2000). Pathogens have multiple pathways to resist ROS and RNS, either by directly interfering with their production, by expressing detoxifying genes, or expressing repair enzymes (Poole, 2005; Shiloh and Nathan, 2000; Wagner et al., 2001). The bacterial responses to RNS and ROS contribute to virulence during infection with Yersinia pestis and Yersinia enterocolitica, respectively, indicating Yersinia pseudotuberculosis may also sense and respond to these stresses (Roggenkamp et al., 1997; Sebbane et al., 2006).

In this report we identify phenotypically distinct subpopulations of bacteria within single sites of replication, and demonstrate that a specialized subpopulation supports growth of individual microcolonies. Therefore, the spatial location of an individual bacterial cell drives gene expression and subpopulations of bacteria collaborate to allow proliferation in tissues.

Results

Y. pseudotuberculosis phenotypically specializes within microcolonies

Y. pseudotuberculosis replicates within tissues to form microcolony structures (Crimmins et al., 2012). It remains unclear if microcolonies are clusters of clonal bacteria, similar to Y. enterocolitica (Oellerich et al., 2007), or if bacteria seed independently and then proceed to cluster together. To analyze clonal behavior, we injected mice intravenously with equal mixtures of Y. pseudotuberculosis expressing mCherry (yopE::mCherry; chromosomal insertion of mCherry downstream of the yopE gene) or GFP (P_tet::gfp; expressed from the low copy plasmid, pACYC184, downstream of P_tet) and visualized the composition of microcolonies in the spleen by fluorescence microscopy. Clonal microcolonies should be composed of either red or green fluorescent bacteria, while a mixture of red and green fluorescent bacteria would indicate that multiple bacteria established the site. Microcolonies exclusively contained either red or green fluorescent bacteria (Figure 1A, 1B), indicating that spleen microcolonies are likely clonal.

Figure 1. Spatial control of Y. pseudotuberculosis gene expression during growth in tissues.

C57BL/6 mice were inoculated intravenously with 10³ WT Y. pseudotuberculosis, and spleens were harvested at day 3 post-inoculation (PI). A) Inoculated with an equal mixture of Yptb GFP⁺ (P_tet::gfp expressed from pACYC184, lacking tetO) and Yptb mCherry⁺ (yopE::mCherry, integrated downstream of yopE⁺). Sections stained with Hoechst and visualized by fluorescence microscopy. 7 mice were analyzed, with 3–11 centers/mouse. B) Three distinct microcolonies (arrows), low power magnification. C) Low and high power magnification of H&E stained sections from a mouse infected with WT, centered on area of inflammation (dashed box, I). R: red pulp region, I: area of inflammation. D) Bacterial transcripts quantified by qRT-PCR. Values expressed relative to 16S. Each bar represents an individual mouse (numbered, black bars) or the inoculum (white bars). E) Nos2 transcripts quantitified by qRT-PCR, relative to Gapdh, and relative to uninfected controls. Each bar represents an individual mouse (numbered), and corresponds to the mice in panel D. F) Mice inoculated with the WT gfp⁺ P_hmp::mCherry strain. Frozen splenic sections stained with Hoechst to detect host nucleic acids. Ratiometric image of the mCherry (P_hmp) signal/GFP signal (range: .2–2) displayed using rainbow scale (Experimental Procedures). See also Figure S1.

Bacteria around the periphery of microcolonies appeared to directly contact host cells, based on Hoechst staining (Figure 1A, 1B) and the presence of neutrophils within areas of inflammation, where microcolonies likely reside (Figure 1C). We hypothesized that bacteria may sense different microenvironments within a single center of replication because interior bacteria lack host cell contact. Host cells release reactive oxygen species (ROS) and reactive nitrogen species (RNS) into the extracellular environment, which may be sensed preferentially by peripheral bacteria. To determine if RNS and ROS were sensed by Y. pseudotuberculosis during systemic infection, we measured global transcription levels of genes for the bacterial detoxifying proteins nitric oxide dioxygenase (hmp) and superoxide dismutase (sodA), which are upregulated in response to nitric oxide (NO) and superoxide, respectively. Mice were infected intravenously with Y. pseudotuberculosis, bacterial RNA was isolated from spleens, and bacterial transcripts were detected by qRT-PCR. Transcription of hmp was detected in each spleen, indicating bacteria were exposed to NO (Figure 1D). sodA transcript levels in tissue were similar to levels observed with broth-grown bacteria, consistent with analysis of Y. pestis during bubo (lymph node) colonization which demonstrated no induction of the superoxide response (Sebbane et al., 2006). Additional endogenous controls (rpoC and gyrA) supported the results obtained using 16S rRNA values for normalizing transcription (Figure S1A, S1B, S1C) (Takle et al., 2007). There was considerable mouse-to-mouse variation in hmp transcript levels, so we then determined if this variation correlated with NO levels in individual mice. Transcription of Nos2, the murine gene for inducible nitric oxide synthase (iNOS), relative to Gapdh was used to approximate NO levels within splenic tissue. Although host Nos2 transcript levels varied between mice, increased Nos2 levels did not correlate with heightened bacterial hmp transcription (Figure 1E). Additional endogenous normalization controls (ß-actin and Rps9) supported this result (Figure S1D).

The non-uniformity of hmp transcription levels between individual mice suggested that there could be heterogeneous expression of this gene. To visualize hmp expression at the single bacterium level, we generated a P_hmp::mCherry transcriptional reporter plasmid, and transformed this into gfp⁺ (P_tet::gfp) bacteria, to compare hmp expression relative to constitutive gfp (under control of tetP lacking tetO). Mice were intravenously inoculated with the WT gfp⁺ P_hmp::mCherry strain, and microcolonies were visualized by fluorescence microscopy within the spleen. The hmp gene was strongly expressed by bacteria around the periphery of microcolonies, while hmp expression was approximately 10 fold lower within interior bacteria (Figure 1F). The fluorescent proteins used in these experiments were stable, so the total fluorescence observed represents an accumulation of signal rather than a dynamic, recent response. Nevertheless, these results clearly indicate that peripheral and interior bacteria sense and respond to distinct microenvironments within a microcolony.

Hmp activity is required for spatial regulation of P_hmp

The bacterial Hmp protein detoxifies NO by converting it to NO₃ (Poole et al., 1996; Robinson and Brynildsen, 2013), which may be important for Yersinia growth in tissues, based on the attenuation of a Y. pestis Δhmp strain in a bubonic plague model of infection (Sebbane et al., 2006). This raised the possibility that the expression of hmp on the periphery of microcolonies reduces the concentration of tissue-derived NO, protecting interior bacteria. To determine if Hmp prevents NO diffusion into the microcolony center, we constructed a Y. pseudotuberculosis Δhmp strain harboring the P_hmp::mCherry reporter to detect the presence of NO. Mice were inoculated intravenously with WT gfp⁺ P_hmp::mCherry or Δhmp gfp⁺ P_hmp::mCherry strains, and microcolonies were visualized within the spleen. P_hmp reporter expression was high in bacteria located on the periphery of WT microcolonies, while Δhmp microcolonies showed no locale-specific expression of the P_hmp reporter, which was detected throughout the Δhmp microcolonies (Figure 2A). The Δhmp microcolonies often appeared altered relative to WT, so comparisons were made between WT and Δhmp microcolonies with similar sizes and morphologies. These results argue that the subset of peripheral Hmp-expressing bacteria effectively eliminated NO diffusion across WT microcolonies.

Figure 2. Hmp activity is required for spatial regulation of P_hmp.

C57BL/6 mice intravenously inoculated with WT gfp⁺ P_hmp::mCherry or Δhmp gfp⁺ P_hmp::mCherry strains, and spleens harvested at day 3 PI. Frozen sections prepared, and bacterial reporter expression visualized by fluorescence microscopy. A) Size-matched representative images of four WT and four Δhmp microcolonies. B) Colorimetric depiction of signal intensity (scale bars) in each channel (top: mCherry, bottom: GFP) from the bottom images in (A). C) hmp(mCherry)/GFP ratios generated by dividing signal intensity of each channel at the centroid and periphery of each replication center. Each dot represents an individual microcolony, 2–12 centers analyzed/mouse, 5 mice. The dotted line: signal intensity equivalent for each channel. D) Δhmp rescued by integrating hmp::mCherry construct into chromosome. Representative microcolony shown alongside hmp(mCherry)/GFP ratios, generated as described in C. 1–10 centers analyzed/mouse, 5 mice. Statistics: Mann-Whitney (comparison between centroids), Wilcoxon matched pairs (comparison between centroid and periphery), ***p<0.001, n.s.: not significant. Scale bars: 20μm. See also Figure S2.

To visualize the expression gradient of the P_hmp reporter in microcolonies, 3D histograms were generated based on the signal intensity from each fluorescent channel in the bottom representative images of Figure 2A. At the extreme edge of the microcolony, there was upregulation of the P_hmp reporter in the WT strain that was not apparent in either the Δhmp strain or for constitutive GFP expression (Figure 2B). The signal intensity ratios were quantified and compared at both the centroid and the periphery of individual microcolonies, and P_hmp expression was significantly higher at the periphery of WT colonies (Figure 2C). In contrast, P_hmp expression was similar at the centroid and periphery of Δhmp microcolonies, with the centroids of Δhmp microcolonies having significantly higher P_hmp reporter expression than WT. The Δhmp strain was then rescued by the integration of an hmp⁺::mCherry construct into the bacterial chromosome, in which intact hmp⁺ is located upstream of mCherry. The hmp::mCherry reporter was expressed specifically in peripheral bacteria, showing the importance of the Hmp protein for this phenotype (Figure 2D).

The WT YPIII strain used in these experiments lacks a functional PhoP protein, which is important for replication within macrophages (Grabenstein et al., 2004). Although Y. pseudotuberculosis replicates extracellularly in the spleen, we wanted to ensure that the absence of PhoP or some other feature specific to YPIII was not the cause of peripheral P_hmp expression. Therefore, we repeated P_hmp reporter experiments using another clinical isolate, IP2666 (Balada-Llasat and Mecsas, 2006). WT IP2666 also had high expression of the P_hmp reporter specifically in peripheral bacteria, while the P_hmp reporter was expressed diffusely in size-matched IP2666 Δhmp microcolonies (Figure S2). Therefore, the small peripheral subpopulation of hmp-expressing bacteria effectively detoxifies nitric oxide during infection, and eliminates NO diffusion across microcolonies, based on results from multiple strain backgrounds.

yopE and hmp expression define distinct subpopulations of peripheral bacteria within microcolonies

One of the major virulence factors found in all Yersinia species is the type III secretion system (T3SS), and its associated translocated proteins (Yops) (Biboud and Bliska, 2005). In bacteriological culture, this system is most highly expressed at 37° C. In cell culture, it is upregulated in response to cell contact, which can be mimicked in Ca²⁺-depleted bacteriological medium (Gemski et al., 1980; Pettersson et al., 1996; Portnoy et al., 1981). Microcolonies are surrounded by inflammatory cells (Figure 3A), which could directly contact peripheral bacteria and lead to heightened T3SS expression. To determine if there is spatial regulation of T3SS expression within a microcolony, mice were intravenously inoculated with a Y. pseudotuberculosis gfp⁺ yopE::mCherry reporter strain (Crimmins et al., 2012), and yopE reporter expression was visualized. Bacteria around the periphery of microcolonies expressed higher levels of the yopE reporter than interior bacteria at the centroid, consistent with upregulation due to host cell contact (Figure 3B, 3C).

Figure 3. Distinct subpopulations of peripheral bacteria within microcolonies.

A) Increasing magnifications of area of inflammation (I, dashed box) from splenic section, stained with H&E. W: white pulp region, R: red pulp region, I: area of inflammation. B) C57BL/6 mice intravenously inoculated with the WT gfp⁺ yopE::mCherry or Δhmp gfp⁺ yopE::mCherry strains, and spleens harvested at day 3 PI. Frozen sections prepared and visualized by fluorescence microscopy. C) yopE(mCherry)/GFP ratios generated as described in Figure 2C. 1–11 centers analyzed/mouse, 5 mice/infection. D) C57BL/6 mice intravenously inoculated with WT P_hmp::gfp yopE::mCherry or Δhmp P_hmp::gfp yopE::mCherry strains, and spleens harvested at day 3 PI. Frozen sections prepared and visualized by fluorescence microscopy. Representative images shown with colorimetric depiction of signal intensity for each channel. E) hmp(GFP)/yopE(mCherry) ratios generated as described in Figure 2C. 2–11 centers analyzed/mouse, 6 mice/infection. F) Nitrogen stress genes induced with NO₂ (NO₂, +), and/or type-III secretion system genes induced by low Ca²⁺ conditions (T3SS, +), in cultures of the WT strain. Transcription levels expressed relative to 16S. Each bar represents four replicates, error bars depict the median and range. Statistics: Mann-Whitney (in vitro transcription), Wilcoxon matched pairs (for comparison between centroid and periphery), ***p<0.001, ** p<0.01, *p<0.05, n.s.: not significant.

Peripheral bacteria displayed heightened expression of both yopE and hmp, consistent with the bacteria responding to both signals. To determine if NO diffusion was affecting T3SS expression, yopE reporter expression was visualized in a Δhmp strain. Δhmp microcolonies still had increased yopE::mCherry reporter expression around the periphery when compared to the centroid of microcolonies, indicating that NO has little effect on tissue regulation of T3SS (Figure 3B, 3C).

To compare the fluorescent reporters directly, WT and Δhmp strains were generated containing both yopE::mCherry (integrated) and P_hmp::gfp plasmid constructs. Visualization of the WT strain indicated that P_hmp expression was specific to peripheral bacteria, while the yopE reporter was expressed throughout microcolonies, but induced to higher levels in peripheral bacteria contacting host cells, indicating that a subset of peripheral bacteria responded to both signals (Figure 3D). In the Δhmp background, P_hmp reporter expression was detected throughout the microcolonies, while yopE expression remained peripheral. Ratiometric comparisons of P_hmp and yopE reporter expression indicated that the centroids of WT microcolonies have higher yopE expression than hmp, while both yopE and hmp are expressed at the periphery (Figure 3E). In contrast, the hmp reporter is expressed at the centroid and throughout microcolonies in the Δhmp background, with yopE expression enriched in the periphery (Figure 3E).

To provide further evidence that T3SS and hmp expression are controlled by distinct environmental signals, bacteria were grown in broth culture with the addition of inducing signals, and transcription was measured by qRT-PCR. Growth of bacteria at 37°C and depleted Ca²⁺, which induces T3SS transcription (Figure 3F, T3SS, +), resulted in increased yopE transcription, which did not significantly change with nitrogen stress (Figure 3F; NO₂, +). In contrast, hmp transcription increased with nitrogen stress (Fig. 3F; NO₂, +), but was not affected by T3SS induction (Figure 3F, T3SS, +). Together, these data support the conclusion that the nitrogen stress response and the T3SS are induced by distinct signals.

Bacteria do not directly contact iNOS producing host cells

Inducible nitric oxide synthase (iNOS, the product of Nos2) generates nitric oxide (NO), and can be expressed by a wide variety of cell types (Pautz et al., 2010). To determine if iNOS was driving hmp reporter expression, Nos2^−/− mice were challenged with the WT gfp⁺ P_hmp::mCherry, and reporter expression was visualized in the spleen. All microcolonies in Nos2^−/− mice lacked hmp reporter expression, despite recruitment of neutrophils (Ly6G⁺) and macrophages (CD68⁺), indicating iNOS drives hmp reporter expression (Figure 4A). In Nos2^+/+ mice, the iNOS⁺ population was primarily composed of T cells (CD4⁺ and CD8⁺) (Pautz et al., 2010; Yang et al., 2013), with smaller populations of iNOS⁺ B cells (B220⁺), dendritic cells (CD11c⁺ CD11b⁺), neutrophils (Ly6G⁺ CD11b⁺), and macrophages (F4/80⁺ CD68⁺) (Figure 4B; Figure S3A). Infection increased the fraction of iNOS⁺ cells within dendritic cell (CD11c⁺ CD11b⁺), neutrophil (Ly6G⁺ CD11b⁺), and macrophage (F4/80⁺ CD68⁺) populations, indicating that these phagocyte populations were sensing the presence of bacteria (Figure 4C) (Durand et al., 2010; Marketon et al., 2005).

Figure 4. Bacteria do not directly contact iNOS producing host cells.

A) Nos2^−/− mice infected with WT gfp⁺ P_hmp::mCherry strain at day 3 PI. Spleen frozen sections prepared, stained with antibodies for indicated host markers and visualized by fluorescence microscopy. 23 centers were analyzed across 2 mice. B) C57BL/6 mice intravenously inoculated with WT (black bars, 5 mice), or PBS (white bars, 3 mice). Splenocytes were detected by flow cytometry. Values represent % of each cell type within total population of iNOS⁺ cells. C) Splenocytes stained and detected by flow cytometry. Values represent % of iNOS⁺ cells within each cell type. D) Low and high power magnification of an area of inflammation (dashed box, I), from splenic section of a mouse infected with WT. W: white pulp region, R: red pulp region, I: area of inflammation. E) & F) C57BL/6 mice intravenously inoculated with GFP⁺ WT strain, splenic frozen sections prepared at day 3 PI, and stained with antibodies. E) Low power magnification. F) High power magnification. Inset: zoom of colony interface. G) C57BL/6 mice intravenously inoculated with GFP⁺ WT, splenic sections prepared at day 3 PI, and antibody stained. Nine representative images shown. Panels A, E, F, & G all depict distinct microcolonies, 115 microcolonies were analyzed from 14 mice. See also Figure S3.

Areas of acute inflammation, which contained microcolonies, were present in red pulp regions (Figure 4D). To determine if iNOS⁺ phagocytes were localized near or contacting bacteria, mice were inoculated with GFP⁺ bacteria and splenic tissue sections were analyzed. Neutrophils (Ly6G⁺) and macrophages (CD68⁺) were located proximal to microcolonies (Figure 4E, 4F), whereas dendritic cells were not detected near microcolonies. Neutrophils appeared to interact with bacteria directly (Fig. 4E, 4F, top) while macrophages were located outside neutrophil populations (Fig. 4E, 4F, bottom). iNOS staining localized with macrophages and some neutrophils, which was consistent with the flow cytometry data that indicated ~50% of macrophages and ~25% of neutrophils express iNOS during infection (Figure 4C). Interestingly, there was scant evidence for iNOS staining in the neutrophils directly contacting bacteria (Figure 4E, 4F inset, 4G), indicating that NO may originate in host cells not directly in contact with microcolonies. Y. pseudotuberculosis T3SS effector proteins can inhibit host signaling pathways (Schesser et al., 1998), which could eliminate NO production in neutrophils directly contacting bacteria. Activated neutrophils did not increase Nos2 expression in response to Y. pseudotuberculosis lacking type-III secretion (P⁻) (Figure S3B, S3C), indicating the lack of iNOS induction may be due to intrinsically low responses by neutrophils.

The absence of Hmp causes reduced fitness

Expression of inducible NO synthase (iNOS) by host cells follows detection of bacterial products, consistent with iNOS controlling bacterial growth after bacteria have already established a foothold in tissues (Nathan and Cunningham-Bussel, 2013; Shiloh et al., 1999; Vasquez-Torres et al., 2000). As hmp is only expressed in the presence of NO (Bang et al., 2006; Poole et al., 1996; Tucker et al., 2008), then it is likely that Hmp acts to promote bacterial survival after a colony has been established. To determine the effects of RNS penetration into microcolonies on bacterial survival, mice were intravenously inoculated with the WT or Δhmp strain, and colony-forming units (CFUs) were quantified at day 3 and day 5 post-inoculation (PI). Balb/c mice were used because at late timepoints (day 5), C57BL/6 mice would have succumbed. At day 3 PI, the WT and Δhmp strains had similar CFUs and had similar microcolony areas (Figure 5A). At day 5 PI, there was a decrease in Δhmp CFUs and microcolony areas compared to WT, with visible disintegration of Δhmp microcolonies (Figure 5B). CFUs of the WT strain showed a small decrease between day 3 and day 5, without a change in microcolony area, indicating there may be some bacterial death in the WT infection as well. Based on in vitro experiments with an NO-donor molecule (DETA/NO), degradation of Δhmp microcolonies is likely due to increased sensitivity of the Δhmp strain to diffusing RNS (Figure S4A).

Figure 5. The absence of hmp causes reduced fitness.

A) Balb/c mice intravenously inoculated with the GFP⁺ WT or GFP⁺ Δhmp strain, and spleens harvested at days 3 and 5 PI. Microcolony areas (μm²) were quantified (Experimental Procedures) in 5 mice, with 6–10 total centers/mouse. B) WT and Δhmp microcolonies at day 5 PI. C) C57BL/6 mice intravenously inoculated with an equal mixture of mCherry⁺ (yopE::mCherry) WT and GFP⁺ Δhmp strains, and spleens harvested at day 3 PI. Competitive index: ratio of Δhmp/WT spleen CFUs at day 3, divided by the ratio of Δhmp/WT CFUs in inoculum. Each dot: individual mouse. Dotted line: each strain has equal fitness. WT and mutant microcolony areas (μm²), quantified (Experimental Procedures) in 7 mice, with 6–10 total centers/mouse. D) Representative WT and Δhmp microcolonies from same organ shown at day 3 PI. Statistics: Mann-Whitney, ***p<0.001, **p<0.01. Scale bars: 50μm. See also Figure S4.

To compare differences in the relative fitness of each strain, C57BL/6 mice were intravenously inoculated with equal amounts of WT mCherry⁺ and Δhmp GFP⁺ strains, and spleens were harvested at day 3 PI to quantify CFUs and microcolony areas during growth within the same organs. The WT strain had increased fitness relative to the Δhmp strain, and the median area of WT microcolonies was significantly larger than Δhmp microcolonies (Figure 5C). Additionally, Δhmp microcolonies showed some evidence of breakdown, even at this earlier timepoint (day 3) in C57BL/6 mice (Figure 5D). These results indicate that diffusion of RNS into microcolonies is antimicrobial in the spleen, and supports the model that hmp expression by peripherally-localized bacteria protects the microcolony.

Bacteria within microcolonies remain in exponential phase

Peripheral bacteria simultaneously sense cell contact and nitrogen stress, which could collaborate to arrest bacterial growth. As a measure of growth arrest, we analyzed markers of entry and establishment of stationary phase. Gram-negative bacteria transcribe the dps gene as they transition out of exponential phase as well as during stationary phase (Almiron et al., 1992; Nair and Finkel, 2004). The KatG catalase/peroxidase is similarly known to be upregulated during the transition into stationary phase (Uhlich et al., 2012; Wakamoto et al., 2013). Low amounts of dps and katG transcripts were detected by qRT-PCR in spleens, at levels similar to exponential phase cultures (Figure 6A), indicating that only a small population of bacteria behaves similarly to post-exponential cells. Additional endogenous controls (rpoC and gyrA) support the results obtained using 16S normalization values (Figure S5A, S5B).

Figure 6. Bacteria within microcolonies fail to activate post-exponential genes.

A) C57BL/6 mice intravenously inoculated with WT strain and spleens harvested at day 3 PI. Bacterial RNA isolated, reverse transcribed, and transcripts detected by qRT-PCR. Transcription levels expressed relative to 16S. Bars: individual mice, or mean of culture condition (6 replicates/culture condition). B) C57BL/6 mice intravenously inoculated with WT P_katG::gfp yopE::mCherry strain, and spleens harvested at day 3 PI. Frozen sections prepared and visualized by fluorescence microscopy. 2–8 centers analyzed/mouse, 5 mice. Three representative images are shown. Scale bars: 20μm. See also Figure S5.

To determine if peripheral bacteria upregulate dps or katG expression, P_dps::gfp and P_katG::gfp transcriptional reporter plasmids were generated and microcolonies were visualized. P_dps reporter expression was not detected within microcolonies, however the P_katG::gfp fusion could be detected in a few individual bacteria in each microcolony (Figure 6B). katG reporter expression within microcolonies appeared to be random, and lacked spatial control, a property reminiscent of Mycobacterium smegmatis colonies formed in culture (Wakamoto et al., 2013). These results indicate that the stress encountered by peripheral bacteria is not sufficient for a transition into post-exponential phase.

Discussion

Multiple types of eukaryotic cells commonly work together to form tissues within higher organisms. Bacterial populations also cooperate to form three-dimensional structures during environmental growth on surfaces (Berk et al., 2012; Serra et al., 2013; Vlamakis et al., 2008), but very little is known regarding microbial cooperative behavior during disease. Several recent publications have described the existence of two populations of bacteria during gastrointestinal infection (Ackermann et al., 2008; Diard et al., 2013), based on expression of a virulence-associated regulon. It is unclear if these populations cooperate to ensure virulence, if the bacterial populations can be spatially distinguished, or if the presence of multiple regulatory populations occurs in other tissue sites. The studies presented here argue that the presence of bacterial populations with distinct regulatory programs is the result of each population residing in clearly distinguishable locales. Furthermore, the topological relationship of these individual populations could facilitate cooperative behavior, allowing the activities of peripheral bacteria to promote survival of interior bacteria.

Our results indicate that in host tissues, bacterial gene expression programs are driven by both contact -independent and –dependent signals. Bacteria localized around the periphery of microcolony structures come into direct contact with host cells, and upregulate virulence genes (T3SS) to prevent phagocytosis, while NO diffusion promotes expression of nitrogen stress genes (hmp) in peripheral bacteria, a subset of which are in host cell contact. These signals define three distinct populations within a single site of replication: interior bacteria, peripheral bacteria responding to NO, and peripheral bacteria responding to both NO and host cell contact. Peripheral bacteria responding to both signals likely grow more slowly than interior bacteria and perform essential detoxifying cell functions that are critical to the survival of interior bacteria.

Bacteria located on the periphery of microcolonies respond to host cell contact by upregulating expression of the T3SS (Pettersson et al., 1996), which inhibits phagocytosis by neutrophils as a consequence of multiple translocated Yops (Spinner et al., 2010). Recruited neutrophils appear to express low levels of low molecular weight antimicrobials based on low iNOS expression by these cells, the observation that Yersinia species block NADPH oxidase activity (Songsungthong et al., 2010; Spinner et al., 2010), and the marked absence of a bacterial response to host-derived reactive oxygen intermediates. Therefore, neutrophils could be providing a defensive barrier for microcolonies, partially protecting peripheral bacteria from direct contact with cell types expressing high levels of NO. It remains unclear if iNOS induction within proximal neutrophils is intrinsically low, or if it is directly inhibited by the T3SS of Y. pseudotuberculosis. Our attempts to demonstrate bacterial inhibition of iNOS induction using activated ex vivo neutrophils have consistently failed, so it appears more likely that recruited neutrophils simply express low levels of iNOS, in keeping with a number of observations in the literature (Pautz et al., 2010; Tsuda et al., 2004). In addition, high levels of NO may directly inhibit neutrophil signaling cascades, thus preventing cells from responding to the presence of bacteria (Bogdan et al., 2000).

Based on the localization of inducible NO synthase, multiple host cell types simultaneously attack a single microbe, with neutrophils directly engaging the bacterium, and a second set of cells located at a distance delivering diffusible antimicrobial molecules. The distance at which host cells are capable of attacking bacteria via RNS diffusion is not clear, but based on the hypothesis that most of these products are generated by iNOS-expressing cells (Pautz et al., 2010), RNS must diffuse across the width of a neutrophil before encountering the bacteria. Detailed analysis of the sources of NO would be greatly aided by the development of strategies to directly measure NO concentrations in tissues.

Based on our results, we hypothesize that location relative to host cells may play a major role in driving bacterial gene expression of other pathogens as well. During growth in broth culture of Salmonella enterica serovar Typhimurium, a subset of bacteria expresses the T3SS (SPI-1) in a stochastic manner, and this behavior is also observed during intestinal infections (Ackermann et al., 2008; Diard et al., 2013). One hypothesis for this phenomenon is that Salmonella interacting directly with the intestinal epithelium could express higher levels of SPI-1 compared to bacteria in the intestinal lumen, paralleling our results with yopE expression. T3SS expression is also controlled by oxygen tension changes in Shigella flexneri and Pseudomonas aeruginosa, and likely other bacterial pathogens, which could lead to differential expression depending on the location of a bacterium within the intestine (Marteyn et al., 2010; O’Callaghan et al., 2011). Variable expression of stress response pathways has also been studied in a S. typhimurium model of systemic infection (Burton et al., 2014). Intracellular S. typhimurium was associated with a variety of host cell types within the spleen, which may explain its non-uniform stress responses.

The presence of functionally specialized subpopulations of bacteria is not limited to host cues during infection. In fact, bacterial-driven quorum sensing is well characterized in pathogens, leading to specialized subpopulations of Pseudomonas aeruginosa during in vitro growth (Dandekar et al., 2012). It is not known if spatial differences in gene expression within P. aeruginosa biofilms mimic disease processes, or whether different transcriptional subpopulations reside within distinct anatomical locations of the lung.

Peripheral bacteria appear to be actively replicating, although the growth of this population is likely to be slower than those in the interior of the microcolony. Our basis for this argument is that we have evidence that the nitrogen stress response slows the growth of Y. pseudotuberculosis. A ΔnsrR strain, which constitutively expresses hmp and other nitrogen stress-related genes (Filenko et al., 2007; Partridge et al., 2009) (Supplemental Figure 4B), competes poorly with a WT strain during single strain infections and co-infections (Supplemental Figure 4C–D). Therefore, individual peripheral bacteria appear to respond to NO at a fitness cost to themselves, mirroring data on the NsrR regulon in S. typhimurium (Gilberthorpe et al., 2007). This result is consistent with the model that virulence gene expression causes a fitness cost to the individual, for the benefit of the total population (Jansen and van Baalen, 2006; Diard et al., 2013; Sturm et al., 2011). This point was strikingly demonstrated with the Salmonella T3SS expression. The slow growth of bacteria expressing the secretion system can be compensated by a subset of faster growing bacteria with low TTSS expression (Ackermann et al., 2008; Diard et al., 2013; Sturm et al., 2011). We argue that the interior bacteria in a Yersinia microcolony similarly reap the benefits of peripheral bacteria, while contributing faster microbial proliferation in tissues. Not clear is whether the interior bacteria also impact the fitness of the peripheral population. Taken together, we believe that although cooperation may be costly for some individual bacteria, it is required for continued maintenance of an established bacterial community within a host organism, laying the basis for phenotypic specialization within a microcolony.

Experimental Procedures

Bacterial strains & growth conditions

The WT Y. pseudotuberculosis strain, YPIII, was used throughout. For mouse infections, bacteria were grown overnight into stationary phase at 26° C. Exponential phase cultures were sub-cultured and grown an additional 2hrs. For yopE induction experiments, cultures were grown at 37° C in media containing sodium oxalate and MgCl₂ to deplete Ca²⁺, and sodium nitrite was added to induce nitrogen stress. See also Supplemental Experimental Procedures.

Generation of reporter constructs

The WT Y. pseudotuberculosis gfp⁺ strain and WT gfp⁺ yopE::mCherry strains have been previously described (Crimmins et al., 2012). All gfp⁺ strains had P_tet::gfp inserted in the plasmid pACYC184, which lacks tetO, leading to constitutive expression. Transcriptional fusions were constructed by fusing promoters to a fluorescent protein gene (gfp or mCherry) by PCR, and were expressed from plasmids. See also Supplemental Experimental Procedures.

Generation of mutant strains

Deletion constructs were amplified with flanking sequence on each side, cloned into the suicide vector, pSR47S, and transformed into Y. pseudotuberculosis. Sucrose selection was used to select for bacteria that had incorporated the mutant construct by double recombination (Crimmins et al., 2012). See also Supplemental Experimental Procedures.

Integrated hmp reporter construction (Δhmp rescued strain)

The hmp::mCherry construct was generated by inserting mCherry immediately downstream of hmp. This fragment was cloned into the suicide vector, pSR47S, and transformed into the Δhmp strain. Sucrose selection was used to isolate strains that had recombined the hmp construct. See also Supplemental Experimental Procedures.

Murine model of systemic infection

Six to 8-week old female C57BL/6 and Balb/c mice were obtained from Jackson Laboratories (Bar Harbor, ME). All animal studies were approved by the Institutional Animal Care and Use Committee of Tufts University. Mice were injected intravenously with 10³ bacteria for all experiments. For co-infection experiments, mice were inoculated with 5 × 10² CFU of each strain, for a total of 10³ CFUs. At the indicated timepoints post-inoculation (PI) (3 days or 5 days), spleens were removed and processed (Crimmins et al., 2012).

qRT-PCR to detect transcription in broth-grown cultures

Bacterial RNA was isolated using the RNeasy kit (QIAGEN), reverse transcribed using M-MLV reverse transcriptase (Invitrogen), and cDNA was used as a template in reactions with SYBR Green (Applied Biosystems). Reactions were carried out using the StepOnePlus Real-Time PCR system, and relative comparisons were obtained using the ΔΔC_T or 2^−ΔCt method (Applied Biosystems). See also Supplemental Experimental Procedures.

qRT-PCR to detect transcription from mouse tissues

Spleens were harvested at day 3 PI, homogenized, and RNA was isolated using the RNeasy kit (QIAGEN). Bacterial RNA was enriched using the MICROBEnrich kit (Ambion), and mouse tissue transcripts were detected from total RNA samples, prior to enrichment. qRT-PCR was performed as described above. See also Supplemental Experimental Procedures.

Flow cytometry

Spleens were harvested at day 3 PI, and single cell suspensions were prepared and fixed in 4% PFA. Non-specific binding was blocked using a rat anti-mouse antibody against the FcγIII/II receptor (CD16/CD32) (BD), and the following rat anti-mouse cell surface antibodies were used: B220-PE-Cy7 (BD), CD4-FITC (eBioscience), CD8a-PE-Cy5 (BD), CD11c-APC (eBioscience), CD11b-PE-Cy7 (eBioscience), Ly6G-PE (BD), and F4/80-PE-Cy5 (eBioscience). For intracellular staining, the following antibodies were applied: rat anti-mouse CD68-Alexa700 (BioRad), rabbit anti-mouse iNOS (AbCam) with goat anti-rabbit-Alexa488 (Invitrogen). Fluorescent signals were detected using a LSRii flow cytometer (BD). See also Supplemental Experimental Procedures.

Histology/Immunofluorescence

Spleens were harvested at day 3 PI and fixed in 4% PFA. Tissue was frozen-embedded in Sub Xero freezing media (Mercedes Medical) and cut by cryostat microtome. The following antibodies were applied to sections: rat anti-mouse Ly6G-PE (BD), rabbit anti-mouse iNOS (AbCam) with either goat anti-rabbit Cascade blue (Molecular Probes) or donkey anti-rabbit AMCA (Jackson Immunoresearch), rat anti-mouse CD68 (AbCam) with goat anti-rat-Texas Red (Invitrogen). Coverslips were mounted using ProLong Gold (Life technologies). Tissue was imaged with either 20x or 63x objectives, using a Zeiss Axio Observer.Z1 (Zeiss) fluorescent microscope with Colibri.2 LED light source, an Apotome.2 (Zeiss) for optical sectioning, and an ORCA-R² digital CCD camera (Hamamatsu). See also Supplemental Experimental Procedures.

Hemotoxylin & Eosin (H&E) staining

Frozen splenic sections were prepared as described above. Sections were stained with hematoxylin, washed, and incubated in 95% ethanol. Sections were then stained with eosin, dehydrated in ethanol, and cleared in xylene prior to mounting coverslips. Tissue was imaged with 4x, 10x, 40x, and 60x objectives, using a Nikon Eclipse TE2000-U inverted light microscope with a Nikon color CCD camera (Nikon).

Image analysis

Openlab software was used to generate the ratiometric image shown in Figure 1F. Image J software was used to generate 3D histograms, which are a colorimetric depiction of the signal intensity of each fluorescent channel across a 2D image. Image J was used to quantify the signal intensity of each channel at the centroid and periphery of microcolonies. Thresholding defined the area of each microcolony, the centroid was calculated, and 0.01 pixel² squares were defined to calculate values at the centroid. Peripheral measurements depict bacteria in contact with host cells. Image J or Volocity software were used to quantify microcolony areas. See also Supplemental Experimental Procedures.

Statistics

Statistical comparisons were computed using the Mann-Whitney U test (non parametric, 2-tailed test) or the Wilcoxon matched-pairs test (non-parametric, paired, 2-tailed test) as denoted in figure legends (Prism 4, GraphPad Software). A P value of less than 0.05 was considered significant.