Topologically correct synthetic reconstruction of pathogen social behavior found in deep tissue sites

Within deep tissue sites, extracellular bacterial pathogens often replicate in clusters that are surrounded by immune cells. Disease is modulated by interbacterial interactions as well as bacterial-host cell interactions resulting in microbial growth, phagocytic attack and secretion of host antimicrobial factors. To overcome the limited ability to manipulate these infection sites, we established a system for Yersinia pseudotuberculosis (Yptb) growth in microfluidics-driven microdroplets that regenerates microbial social behavior in tissues. Chemical generation of nitric oxide (NO) in the absence of immune cells was sufficient to reconstruct microbial social behavior, as witnessed by expression of the NO-inactivating protein Hmp on the extreme periphery of microcolonies, mimicking spatial regulation in tissues. Similarly, activated macrophages that expressed inducible NO synthase (iNOS) drove peripheral expression of Hmp, allowing regeneration of social behavior observed in tissues. These results argue that topologically correct microbial tissue growth and associated social behavior can be reconstructed in culture.


45
A variety of bacterial pathogens colonize and replicate within tissues despite the presence 46 of the host immune system (Carter & Collins, 1974;Cheng et al., 2009;Simonet et al., 1990). 47 Growth in tissue sites involves the formation of distinct foci of replication, which can develop 48 into either abscesses, granulomas, or poorly defined clusters of bacteria (Cheng et al., 2011;49 Pagan & Ramakrishnan, 2018). Extracellular bacterial pathogens, in particular, can establish a 50 tissue niche and replicate to high numbers. These clusters of bacteria in tissues are often clonal, 51 result in distinct microcolonies, and are surrounded by host innate immune cells. 52 Yersinia pseudotuberculosis (Yptb) is an enteric pathogen that replicates in the intestinal 53 lumen and regional lymph nodes, with the potential for disseminating via a poorly characterized 54 route into deep tissue sites such as the liver or spleen (Barnes et al., 2006). Once colonized, Yptb 55 establishes extracellular foci of replication, resulting in the formation of microcolonies that 56 develop into lesions that are densely populated by immune cells (Simonet et al., 1990). Within 57 the murine spleen, the bacterium sets up a beachhead in which distinct microcolonies are derived 58 from a single seeding bacterium (Davis et al., 2015). Surrounding the bacterial microcolony, 59 which contains between 50-5000 bacteria, are strata of immune cells. In direct contact with the 60 population center are neutrophils, which have cytoskeletal elements that are paralyzed by the 61 abutting bacteria that translocate Type III Secretion System (TTSS) effectors. As a consequence, 62 phagocytosis of the pathogen by surrounding neutrophils is severely disrupted and reactive 63 oxygen species (ROS) production is greatly reduced (Songsungthong et al., 2010). The resulting 64 neutrophil-bacterium interface appears to result in a stable relationship, in which frustrated tissue sites. There has been limited analysis of the dynamics of bacterial growth in tissue, 90 investigation of the spatial relationship between immune cells and the bacteria, or identification 91 of inter-bacterial communication within the growing microbial populations in tissues. To bridge 92 the gap between animal infection models and tissue culture models, we developed a chemical 93 strategy to incorporate the architecture of inflammatory sites in a tissue culture system. 94 In this report, we develop an in vitro system that uses droplet-based microfluidics to

106
Droplet gels support clonal growth of Yersinia pseudotuberculosis microcolonies. 107 Yptb microcolonies in the murine spleen are derived from single isolated bacteria that 108 replicate as extracellular clusters and become surrounded by innate immune cells (Davis et al.,109 2015; Simonet et al., 1990). In order to accurately model the growth of Yptb in deep tissue sites, 110 a single bacterium needs to be isolated and grown into a microcolony in a matrix that supports 111 this 3-dimensional (3D) topology. To encase single bacterial cells in matrix, we utilized droplet- Droplets containing encapsulated Y. pseudotuberculosis were cultured at 26 o C, and microcolonies were visualized at indicated timepoints by phase contrast and fluorescence microscopy. Microcolony areas were determined by image analysis (Materials and Methods). Each timepoint is median +/-95% confidence interval (CI) of 3 biological replicates of 50 microcolonies. G. Representative images of microcolonies from (F) at the noted times visualized by phase contrast and fluorescence microscopy. Scale bar: 50μm. 5 based microfluidics, which is commonly used for isolating single mammalian cells (Koster et al.,113 2008; Macosko et al., 2015;Mazutis et al., 2013). 114 Yptb was added to a matrix consisting of molten 1% ultra-low melt agarose containing 115 25% HyStem ® -C Hydrogel. HyStem ® -C Hydrogel has thiol-modified hyaluronan and gelatin, 116 allowing on-demand polymerization of these two components controlled by the addition of a 117 thiol-reactive crosslinker (Fig. 1A). The biomatrix was included in the droplet mixture to add were about 8 μm in diameter ( Fig. 1C and 1D). After droplet generation, HyStem ® -C Hydrogel 128 was crosslinked within the agarose using Extralink ® , a thiol-reactive crosslinker, at room 129 temperature and oil was removed from the droplets (Materials and Methods). During oil 130 removal, the population of small droplets was lost, and only the larger droplets remained.

131
In tissue, Yptb microcolonies form clusters (Simonet et al., 1990). We wanted to confirm 132 that Yptb grows as a cluster inside the agarose/ HyStem ® -C Hydrogel droplets. To this end Yptb 133 expressing GFP (Ptet::gfp, expressed from pACYC184) was encapsulated in droplets, oil was 134 removed, and the droplets were incubated at 26 o C in 2xYT broth with rotation. Microcolony 135 6 growth was determined by identifying a threshold that defines edges of microcolonies in the GFP 136 channel and determining the number of pixels in the region of interest (ROI). The data were then 137 converted to metric scale using a stage micrometer and displayed as μm 2 . The area of the 138 microcolony increased over time with roughly logarithmic kinetics over a 10-hour period, 139 indicating that the agarose/ HyStem ® -C Hydrogel droplets support efficient Yptb microcolony 140 formation (Fig. 1F). The replicating bacteria also grew in clusters, with similar appearance to Y. 141 enterocolitica grown in collagen gels (Fig. 1G) (Freund et al., 2008). Altogether, these results

142
show that agarose/ HyStem ® -C Hydrogel droplets support microcolony formation of Yptb, 143 mimicking growth and topological constraints observed in tissue.  DETA-NONOate. Microcolony area was measured over time through a single plane. In parallel, 159 the two strains were grown in broth culture. The growth of the WT and ∆hmp microcolonies was 160 similar in the droplets in the absence of DETA-NONOate ( Fig. 2A and 2D). The microcolonies 161 derived from WT were relatively resilient to NO insult, with complete blockage of growth only 162 observed at extreme concentrations (40mM DETA-NONOate) ( Fig. 2A). In contrast, the ∆hmp 163 strain was blocked from replicating in droplets at all concentrations of DETA-NONOate tested, 164 indicating that Hmp is essential for growth under these conditions (Fig. 2D). Although 8 hours of 165 20mM DETA-NONOate exposure showed some interference with replication in microdroplets, 166 the WT microcolonies were clearly larger than Δhmp mutant microcolonies ( Fig. 2B and 2E).    The WT and ∆hmp strains harboring gfp + Phmp::mcherry were cultured in either droplet 190 microcolonies or broth and exposed to 0,1, 2.5, 5, and 10 mM of DETA-NONOate for 0, 4 and 8  The heterogeneous mcherry reporter fluorescence in WT was dependent on the hmp 205 promoter, as GFP expression driven by the unrepressed Tet promoter was uniform in bacteria 206 that originate from broth culture and droplet microcolonies ( Fig. 3C and S1B). Similar to the 207 observations with mcherry there was a small population representing approximately 5% of the 208 bacteria that failed to fire gfp perhaps due to the dispersal and fixation procedure. These results 209 argue that microcolony topology is a critical factor that drives heterogeneity observed in Spatial regulation of hmp expression in tissues can be reproduced in droplet culture. 221 We hypothesized that the heterogenous distribution of Phmp::mcherry fluorescence of 222 WT Yptb as seen by flow cytometry analysis is indicative of spatial regulation of the promoter.

223
To determine if the spatial regulation of Phmp observed in tissue can be reproduced in droplet 224 culture, WT and Δhmp strains carrying the gfp + and Phmp::mcherry reporters were encapsulated 225 in droplets, grown into microcolonies, and exposed to DETA-NONOate for 4 hours (Fig. 4A).  The fluorescence for the defined PLI is also calculated in the same fashion, and the periphery verses PLI ratio is determined (Supplemental Data). B. Example of identification of periphery, mask generation and definition of peripheral ROI. The GFP channel defines the periphery, allowing mask to be set. Within the mask, the periphery is defined as extending 12 pixels into the microcolony. The PLI within the mask is defined as the area of 5-pixel radius with the lowest intensity in the mCherry channel.

252
Flow cytometry analysis of WT Yptb originating from droplet microcolonies exposed to 253 these low DETA-NONOate levels revealed heterogenous expression when compared to bacteria 254 that were not exposed to DETA-NONOate ( Fig. 6C: left). This argues that the broad distribution 255 of mcherry fluorescence intensity is a consequence of spatially regulated Phmp firing (Fig. 6A).  Model for droplet microcolonies. Bacteria are grown in situ within the droplets prior to challenge with iNOS + cells to mimic Y. pseudotuberculosis microcolony structure in tissue. Black arrows note there is space between iNOS+ cells and cluster of bacteria. mimicking of action-at-a-distance by iNOS + cells, as the microcolonies in the engineered system 274 are situated at similar distances relative to the droplet surface (Fig. 7A, iNOS + Cell Recruitment).

275
To recapitulate tissue dynamics, therefore, BMDMs were activated to produce iNOS + and added 276 to droplets containing pregrown microcolonies (Fig. 7B). The agarose/ HyStem ® -C Hydrogel 277 droplet served as the buffer zone between the bacterial microcolony and the iNOS + cell layer 278 (Fig. 7B). Pregrowth of the microcolony prior to exposure to BMDMs is envisioned to 279 recapitulate the kinetics observed from histology of spleen tissues (Fig. 7A).  (Fig. 8F). This is consistent with activated macrophages providing 294 soluble growth-promoting substances at a distance, including either NO3or oxidized cofactors 295 as a consequence of Hmp activity. To show that peripheral expression of hmp was dependent upon RNI, the NOS inhibitor, 297 L-NMMA, was introduced into this system (Fig. 8F, G). BMDMs were incubated in the presence 298 or absence of 2mM L-NMMA prior to challenging droplet microcolonies, and expression of the from BMDMs in the absence of activation (Fig. 8G). As expected, there was no spatial 305 regulation of Phmp::mcherry in the strain lacking hmp activity, and the presence of NMMA had 306 no effect (Fig. 8G). These results indicate that peripheral expression of Phmp::mcherry is 307 dependent on the presence of intact Hmp protein and in response to RNI generated by iNOS 308 activity.

309
To demonstrate that there was heterogeneous hmp expression in WT microcolonies on a 310 population level, WT and ∆hmp microcolonies challenged with unprimed, LPS/IFNγ-primed, or 311 L-NMMA treated-LPS/IFNγ-primed BMDMs were dispersed and analyzed by FACS (Fig. 8H). showing that NO secreted by macrophages can penetrate the entire microcolony ( Fig. 8H: right).   We have addressed this gap in the field by reconstructing Yptb inflammatory sites using 341 microfluidic technology to generate a 3D model that allows interbacterial interactions to be 342 studied as well as action-at-a-distance by immune cells, with topology mimicking a tissue 343 infection. Yptb is an intestinal pathogen that can establish extracellular foci in deep tissue sites 344 after translocation across the intestine into either the bloodstream or regional lymph nodes 345 (Barnes et al., 2006). In the course of establishing an infectious niche within the murine spleen, 346 Yptb forms microcolonies tightly associated with recruitment of neutrophils that directly contact 347 the cluster of bacteria, which are, in turn, encased by layers of more neutrophils followed by 348 macrophages and monocytes that largely do not contact bacteria directly (Davis et al., 2015). The differs greatly from the droplet analysis. It does argue that the amount of RNI generated by 1 390 mM DETA-NONOate is more than sufficient to explain the images generated from the mouse 391 spleen.

392
One of the strengths of the synthetic droplet system is that we were able to disrupt the 393 droplets to free the bacteria and analyze bacterial populations in bulk, a task that is often difficult   to each syringe and then inserted into the appropriate ports on the microfluidics device (Fig. 1).

19
The pump holding the oil-phase was set to 700 µl/hour while the other pump was set to 400 435 µl/hour and droplets were collected in a 1.5 ml Eppendorf tube through tubing that had been 436 inserted into the droplet collection port.  Nitric oxide experiments. To analyze the response of colonies to exogenous NO, droplets 452 containing Y. pseudotuberculosis were generated, oil was removed, and the droplets were rotated 453 for 7 hours at 26 o C in 2xYT broth to allow colony formation. After the 7-hour growth period, 454 droplets were subjected to centrifugation for 30 seconds at 250 RCF and washed twice in PBS.

455
50 µl of droplets were transferred to a 1.5 ml Eppendorf tube and resuspended in 1 ml 2xYT 456 broth and exposed to DETA-NONOate (Sigma #AC32865) during growth at 26 o C with aeration 457 for the indicated timepoints. For experiments performed at 37 °C with 5% CO2 (Fig. 5), 50 µl of 458 droplets were resuspended in 1 ml RPMI 1640 (Gibco, USA) supplemented with 10% FBS and 2 459 mM glutamine media in 12-well non-tissue culture treated plates for the indicated times.

512
of droplet size, images were captured by phase contrast microscopy using a 20x lens, and images 513 were analyzed to identify 300 droplets, with analysis in Volocity TM . Droplet size was determined 514 by identifying a threshold that defines edges of droplets and determining the diameter in pixels.

515
The data were then converted to metric scale using a stage micrometer and displayed as a and zero at all other points. This matrix was morphologically dilated with a structuring element 547 of radius 5 to sample a region of lowest intensity rather than a single pixel. The number of pixels 548 in this resulting mask are then recorded and the total intensity of the green channel was recorded.

549
The mask was applied to the raw image data in the red channel, and the total intensity recorded.

550
Total intensity was normalized to number of pixels to achieve an average intensity for the entire 551 PLI region in the red and this value was normalized to the average intensity of the green channel.