Spatiotemporal Dynamics of Molecular Messaging in Bacterial Co-Cultures Studied by Multimodal Chemical Imaging

Abstract

Microbial community behavior is coupled to a set of genetically-regulated chemical signals that correlate with cell density – the quorum sensing (QS) system – and there is growing appreciation that the QS-regulated behavior of bacteria is chemically, spatially, and temporally complex. In addition, while it has been known for some time that different species use different QS networks, we are beginning to appreciate that different strains of the same bacterial species also differ in their QS networks. Here we combine mass spectrometric imaging (MSI) and confocal Raman microscopy (CRM) approaches to investigate co-cultures involving different strains (FRD1 and PAO1C) of the same species (Pseudomonas aeruginosa) as well as those involving different species (P. aeruginosa and E. coli). Combining MSI and CRM makes it possible to supersede the limits imposed by individual imaging approaches and enables the spatial mapping of individual bacterial species and their microbial products within a mixed bacterial community growing in situ on surfaces. MSI is used to delineate the secretion of a specific rhamnolipid surfactant as well as alkyl quinolone (AQ) messengers between FRD1 and PAO1C strains of P. aeruginosa, showing that the spatial distribution and production rate of AQ messengers in PAO1C far outstrips that of FRD1. In the case of multiple species, CRM is used to show that the prolific secretion of AQs by the PAO1C strain of P. aeruginosa is used to mediate its interaction with co-cultured E. coli.

I. Introduction

Microbial Biofilms.

Most bacteria in natural and clinical settings grow as surface-attached biofilms — highly structured communities encased by a matrix composed of extracellular polymeric substances (EPS). Eradication of these biofilms is extremely difficult, as biofilm bacteria can be up to 1000 times more resistant to antibiotics than isolated bacteria.¹ In addition to being more resistant to treatment, biofilm bacteria can cause harm in numerous other ways. For example, the bacterium Pseudomonas aeruginosa is an opportunistic human pathogen that causes skin, eye, lung, and gastrointestinal infections in susceptible individuals. In many clinical situations, including the lungs of cystic fibrosis patients, P. aeruginosa is the dominant organism, existing predominantly as attached-growth biofilms, in which its pathogenicity and resistance to antibiotics are significantly enhanced.^2–4 While much has been learned about select factors that regulate biofilm formation in vitro and in animal models, the specific mechanisms by which multispecies groups of cells form biofilms are not yet understood. Further complicating the picture, much of what we know regarding the chemical profile of bacterial communities arises from homogenized samples and extrapolation of nucleic acid profile data - an approach which obliterates spatial relationships - yet we know that most organisms occupy 2D and 3D space heterogeneously.^{5, 6}

It is widely appreciated that microbial community behavior is strongly coupled to a set of genetically-regulated chemical signals – a system known collectively as quorum sensing (QS) – that correlate with cell density. There is also growing appreciation that the QS-regulated behavior of bacteria is complex – chemically, spatially, and temporally. For example, even at the level of different strains within the same species, P. aeruginosa exhibits highly variable QS networks.⁷ For example, QS activation of virulence factors was recently shown to act not just upon localized communities after reaching a QS density threshold, but QS signals were additionally shown to influence the size of the community as well as the phenotype of other regional bacteria.⁸ Further highlighting the inter-connected web of chemical relationships that determine bacterial behavior, Shrout and coworkers showed that nutritional and surface conditions can strongly influence QS-mediated activity, by regulating factors that guide biofilm development.^9–11 While it is widely appreciated that QS is important to multi-species interactions,¹² it is not currently possible to predict how specific bacterial strains use QS to guide their development into niches, for example those that cause infection.⁷ This realization disputes the notion that pathogenic differences among strains are easily explained by the presence or absence of one or a few virulence proteins, or that the response to exogenous agents, like antibiotics, is linear.

Interactions among Species.

Biofilms associated with infections, such as those encountered in CF and otitis media, are typically polymicrobial in nature (bacterial, fungal, and viral).^{13, 14} The microbes inhabiting mixed-species biofilms exhibit complex interactions with each other, including cell-cell signaling, metabolic cooperation, and competition among the members for available nutrients.¹⁵ For example, P. aeruginosa, Burkholdaria cepacia and additional species can form multi-species biofilms in the CF lung and communicate with each other via acyl homoserine lactone cell-cell signaling, which is thought to synergistically enhance their collective virulence.¹⁶

Hogan and Kolter showed that P. aeruginosa secretes virulence factors and toxins to kill filaments of Candida albicans.¹⁷ This work illustrates the competitive nature of interactions among microbes in multi-species biofilms, and indicates that virulence factors associated with infections in humans are also implicated in regulating bacteria-fungal interactions. Furthermore, competition for available resources also occurs between sibling colonies of the same strain of bacteria, indicating that under conditions of limited nutrients, bacteria can resort to cannibalism to survive.¹⁸ Considering the role of multi-species biofilms in disease and the inherent differences in virulence among strains of the same bacteria, it is pivotal to understand the complex inter-species and inter-strain interactions between microbial biofilm community members to elucidate the underlying mechanisms implicated in virulence, pathogenesis, and disease progression.

In an early demonstration of the potential utility of chemical imaging to understand inter-species interaction (vide infra), Moree et al. analyzed colonies of P. aeruginosa and the pathogenic fungus Aspergillus fumigatus cultured side-by-side on the same agar plate using matrix-assisted laser desorption/ionization Fourier-transform ion cyclotron resonance mass spectrometry imaging (MALDI-FTICR MSI).¹³ The resultant maps of the secretome showed that phenazine metabolites produced by P. aeruginosa are transformed by A. fumigatus into other compounds possessing different chemical properties. These increased toxicity and stimulated the production of fungal siderophores.

Multimodal Chemical Imaging.

Metagenomic studies of multispecies communities are limited in their ability to describe the activity and function of different strains and/or species. As a result, species-species biomolecular interactions in 2D/3D space are largely undefined. To address this knowledge gap, we are exploring how the behavior and competition of pathogens changes with spatial arrangement, thus permitting us to move beyond cataloguing the secretome to defining the spatial and temporal distributions of chemical messengers that mediate interspecies communication. For example, by examining the spectra of multi-species biofilms to distinguish among bacterial constituents, it is possible to illuminate how differences in genome composition, pathogenesis, and QS networking evolve in co-culture. Given that pathogens like P. aeruginosa, Staphylococcus aureus, and others can infect multiple host cell types, it is critically important that we better understand how the biosecretome of these pathogenic bacteria evolves as they exert dominance.

Multimodal chemical imaging – an innovative set of experiments that target the spatial and temporal organization of chemical signaling in microbial communities – constitutes a radical new approach to understanding microbial behavior that addresses the fundamental dichotomy between the spatially heterogeneous way that microbial cells exist and function in nature, and the principal laboratory tools that are used to study them - tools that largely rely on homogenizing bacterial cultures.^19–30 Initial efforts to realize the potential of this approach in our laboratories have addressed three primary areas. First, we developed mass spectrometric imaging/confocal Raman microscopy, MSI/CRM, multimodal chemical imaging, in which secondary ion mass spectrometry (SIMS) imaging and confocal Raman microscopy (CRM) are combined to analyze the chemical composition of cultured Pseudomonas biofilms.²⁰ Precise spatial correlation between SIMS and CRM images – carried out on different instruments hundreds of miles apart – was achieved using a chemical microdroplet array, allowing us to relocate regions of interest, and align image data with μm-scale precision. We also developed novel approaches to reconcile the differences in sensitivity, dynamic range, and noise susceptibility between the MSI and CRM instruments. Second, we applied these tools to the spatiotemporal mapping of chemical messengers and correlated the spatial and temporal distributions of these secreted components to the development of phenotype in communities of P. aeruginosa.²¹ Finally, we have also applied these approaches to the spatially- and temporally-organized responses of P. aeruginosa to antibiotic exposure.²⁸

Although P. aeruginosa is well-known to secrete substantial amounts of three classes of extracellular compounds: rhamnolipids (RHL), acyl homoserine lactones (AHL), and the nitrogen containing heterocyclic aromatic 2-alkyl-4-quinolones (AQs), Scheme 1, the AQs have occupied a substantial fraction of our efforts.³¹ Pseudomonas quinolone signal (PQS), 2-heptyl-3-hydroxy-4-quinolone, is the most-studied example of this quinolone/quinoline molecule class in the literature.^31–35 By examining the distinct chemical image information from MSI and CRM, we have been able to show that P. aeruginosa biofilms respond to an antibiotic challenge in a way that is: (a) chemically specific (to the antibiotic), (b) spatially-heterogeneous, and (c) which exhibits orders of magnitude dynamic range in chemical response across a community.²⁸ Clearly, these capabilities that enable the identification of individual bacterial strains and their biochemical products with sub-μm resolution are extraordinarily useful. Not only do they enhance the understanding of host-associated microbial community development, but they also inform the long-term goal of designing diagnostics for bacterial growth, cell-cell signaling, and host colonization. These can, in turn, be translated to improved therapeutics in support of personalized medicine strategies.

Scheme 1.

Alkyl quinolones (AQs) secreted by P. aeruginosa. *PQS abbreviation often includes both C7-PQS and C9-PQS.

II. Results and Discussion

Strain-Strain Interactions.

Understanding intra-species behavior, especially interactions among different strains of the same microbe, is critically important. For example, specific classes of mutants are frequently identified in abundance in chronic lung and wound infections. The selective pressure by which these mutant strains become dominant is poorly understood and is likely due to both nutrient availability and species-species interaction.³⁶ We have applied both CRM and SIMS imaging to characterize the spatial patterns of biomolecules secreted by FRD1 and PAO1C - two representative strains of P. aeruginosa which have drastically different growth traits and are expected to compete in infection environments. The FRD1 strain was originally isolated from cystic fibrosis sputum, and is characterized by its mucoid phenotype, in which the polysaccharide alginate is overproduced as a dominant component of the EPS. The PAO1C strain was originally isolated from a burn wound and is more mobile than FRD1. Each strain of P. aeruginosa produces varying amounts of AQ and rhamnolipids during the biofilm growth period on agar, and as such exhibit characteristic spatiochemical distributions and interactions.

Here, CRM and MSI were employed to study the interactions between P. aeruginosa strains PAO1C and FRD1 in co-culture biofilm communities. Colony biofilms were grown on the surface of FAB media solidified with 1% noble agar with glucose as the carbon source. For the co-culture biofilm assay the FAB-agar plates were simultaneously inoculated with overnight broth cultures of FRD1 and PAO1C (OD = 1 at 600 nm) at approximately the same distance from the plate center¹⁸ and incubated at 37 °C until the desired time of growth had elapsed. For MSI analysis the agar containing the bacterial growth was carefully excised and transferred to an MSI plate covered with copper tape, dried under a stream of nitrogen for ~3 h and stored under vacuum until analysis.²⁷

Consistent with the inherent differences in motility between FRD1 and PAO1C, the area occupied by the FRD1 colony biofilm is much smaller than that occupied by PAO1C. Optical images of 3- and 5-day co-cultures of FRD1 and PAO1C, Figure 1, show that PAO1C grows around and eventually envelops the FRD1. Furthermore, the FRD1 colony exhibits a mucoid phenotype which is not observed in the PAO1C colony. C₆₀-SIMS MSI, Figure 2, reveals that PQS/HQNO are secreted preferentially in the core and at the outer edges of the PAO1C colony, while dense, highly concentrated AQs are present nearly uniformly in the FRD1 biofilm. SIMS imaging and SIMS product ion imaging reveal that FRD1 produces alkyl-quinolones from all three major classes at much higher abundance than PAO1C, but these quinolones are confined to a small thick biofilm region. In contrast, PAO1C expresses quinolones at both the center and the periphery of the biofilm and produces more rhamnolipids, as evidenced by ion images of an unidentified component (possibly a rhamnolipid), which are spatially ubiquitous around the PAO1C region but excluded from the core of the FRD1 region.

Figure 1.

Optical images of FRD1 and PAO1C strains in co-culture colony biofilms cultured for (a) 3 and (b) 5 days on the surface of agar.

Figure 2.

SIMS ion images showing distribution of the main quinolone classes for (a) 3- and (b) 5-day FRD1-PAO1C co-cultures. A putative RHL (m/z 525.26) and an unidentified ion (m/z 296.35), produced by FRD1 in response to the presence of PAO1C, are also shown. False-color ion intensities are arbitrarily scaled from 0 (black) to 1 (red) for each image.

Additionally, during the interaction of these two strains, an unidentified upregulated component at m/z 296.35 is observed, as shown in Figure 2. This component is produced exclusively by FRD1. Interestingly, several additional unidentified components (m/z 294.35, m/z 322.38, m/z 324.38) are also observed in the C₆₀-SIMS ion images generated from the FRD1 and PAO1C 3 and 5-day co-culture colony biofilms. Just as for the m/z 296.35 component, these unidentified ions are present almost entirely in and around the FRD1 colony biofilm. In fact, the unknown with m/z 296.35 is produced specifically by FRD1, since it is present in FRD1 pure culture colony biofilms, even in the absence of PAO1C (data not shown). The spatial segregation of the unknown compounds around the FRD1 and away from the PAO1C colony strongly suggests that they play a role in modulating biofilm formation in FRD1 and, consequently, in developing the response of FRD1 to the presence of nearby PAO1C.

Inter-species Interactions.

The inter-strain behaviors noted above during the interaction of FRD1 and PAO1C colonies of P. aeruginosa suggest that more interesting information can be obtained by examining interactions between colonies of different species. For example, a similar strategy can be employed to study interactions between S. aureus and P. aeruginosa. These studies would be very informative from a clinical perspective, since co-culture communities of P. aeruginosa and S. aureus are relevant to infections in the CF lung.^{37, 38} Because of the limitations of metagenomic approaches to study multispecies communities and provide information on biomolecular interactions in 2D/3D space, we are exploring how the behavior and interaction of pathogens changes with spatial arrangement. We are particularly interested in the spatial range of inter-species communication and how surface chemistry directs and organizes the response of one species to a non-cognate, co-inhabitant species. This work moves beyond cataloguing the secretome and will permit us to define the spatial and temporal distributions and contributions of chemical messengers that mediate inter-species communication.

P. aeruginosa and S. aureus are pathogens that can infect multiple human organs, including the lung, skin, and surgical sites and implants. Both are specifically affiliated with hospital-associated infections,³⁹ and while each may become the dominant species in an infection, we do not currently understand why one bacterial pathogen dominates the other over time.

Starting at the level of community morphology, P. aeruginosa-S. aureus interspecies interactions have been explored using dual-culture assays on soft agar. Surprisingly, the non-motile P. aeruginosa PAK strain exhibits swarming motility when co-cultured with S. aureus USA300, Figure 3. As unexpected as this result is, it is even more curious that the P. aeruginosa PA14 strain exhibits no change in swarming when co-inoculated with S. aureus, cf. top and bottom rows in Figure 3. At present, this result remains unexplained, however PAK remains motile even when co-cultured in the presence of an S. aureus agr-QS mutant, which lacks the gene to produce the only known S. aureus surfactant. These results clearly indicate the upregulation of a motility-inducing factor in the presence of S. aureus, thus illustrating the complexity of interspecies interactions and highlighting the need for a spatial mapping of secreted biomolecular signals.

Figure 3.

Surface spreading phenotypes of P. aeruginosa strains co-cultured with S. aureus. The position of the S. aureus inoculation is indicated by SA.

As a model system to explore the application of CRM to detailed studies of bacterial co-cultures, we started with the interactions of P. aeruginosa PAO1C co-cultured with E.coli DH5α. The PAO1C strain of P. aeruginosa is competent to swarm over the surface, as illustrated by time-dependent photographic and CRM microspectral images at 24, 48, and 96 h, Figure 4. A typical set of data, acquired at the longest incubation time, 96 h after co-inoculation, is shown in Figure 5. These 96 h images clearly show that the edges of the PAO1C and DH5α colonies merge, offering the opportunity to characterize the spatial distribution of signaling molecules.

Figure 4.

Time series of images showing the progress of PAO1C P. aeruginosa (PA, inoculated on the left) and DH5α E.coli (EC, inoculated on the right) co-inoculated at t = 0

Figure 5.

Interactions of P. aeruginosa PAO1C co-cultured with E.coli DH5α. (a) Bright-field (10x objective) image covering the entire sample. Attention is focused on the growing edge of P. aeruginosa colony towards E.coli, area encompassed by the red circle. (b) Loading plot constructed from the CRM image, showing the highest contribution to variability from principal component analysis (PCA), PC1, in the region of the P. aeruginosa-E. coli interface. Loading plots under these conditions display substantial similarities to the underlying Raman spectra. (c) Raman image (integrated over 1330-1380 cm-1) of the region near the P. aeruginosa-E. coli interface. Scale bar, 10 μm. (d) Heat map of PC1 over the same region.

Under the microscope, bacterial cells can be observed to secrete clusters/aggregates of signaling molecules, which vary over time and space. After sufficient time these secreted molecules can exceed the local solubility limit and exhibit ordered, crystalline structures (not shown). At the earliest time point observed, 8 h after inoculation, although the bacteria began to spread across the surface, no clusters/aggregates were detected. However, at the 24 h and 48 h post-inoculation time points, distinct clusters/aggregates were observed, although the total concentrations were sufficiently small that the spatially averaged Raman microspectra were weak.

However, at 96 h the colonies are sufficiently evolved, Figure 5, displaying a number of interesting and informative characteristics. The CRM microspectra are sufficiently strong to render strong, assignable features in the loading plot, Figure 5(b). Features at 718 cm⁻¹, 1210 cm⁻¹, 1438 cm⁻¹, and 1513 cm⁻¹ align with Raman bands typically observed in samples exhibiting strong AQ secretion. In particular, the strong feature at 1359 cm⁻¹ is the characteristic quinolone ring vibration of a specific AQ signaling molecule, AQNO. The C₇- and C₉-congeners (HQNO and NQNO, Scheme 1) cannot typically be distinguished by inelastic light scattering, except in the most favorable cases, and it is likely that both are secreted at the bacterial interface in the PAO1C-DH5α sample. The dominance of the AQNO signaling feature can further be observed by comparing the Raman image integrated over 1330-1380 cm⁻¹, Figure 5(c), to a heat map of the strongest principal component (PC), PC1, from a principal component analysis of the CRM image in the bacterial interface region. The spatial distributions clearly match each other to a high degree indicating that: (a) the AQNO signal is the dominant molecular component measured by inelastic light scattering in this region, and (b) DH5α E. coli is the principal cellular component, although the cellular Raman signals were obscured by the strong AQNO signals observed . Table 1 summarizes the results of a global PC analysis including both PC1 and PC2 from 8 to 96 h post-inoculation, clearly indicating that signaling molecules reach the periphery of the E. coli DH5α colony by 24 h, well before the colonies are observed to merge visually. One cannot discount the presence of E. coli DH5α signaling molecules counter-propagating to reach the PAO1C P. aeruginosa colony, however these have not been observed in the present study.

Table 1.

Principal Components Identified at Various Time Points

	8 h	24 h	48 h	96 h
PC1	cell components	cell components	cell components	AQNO + cell components
PC2	not detectable	AQNO	AQNO	AQNO + unidentified components

III. Conclusions

Most bacteria in natural and clinical settings grow as surface-attached biofilms, which are microbial communities that have self-assembled into an encased and protective matrix of extracellular polymeric substances (EPS). The inter-cellular signaling process known as quorum sensing (QS) plays an important role in biofilm formation and development, both critical to the establishment of an infection. For the opportunistic pathogen, P. aeruginosa, QS regulates the expression of many genes important to biofilm initiation, EPS production, and virulence. However, while a number of the key actors have been chemically identified, their overall distributions and patterns of movement have not. To address this problem, the multimodal mass spectrometric and Raman chemical imaging approaches described here allow us to circumvent the limits imposed by previous technology that allows examination of only a small number of cells or entire cell populations that have been removed from the conditions of interest. These methods allow the determination and spatial mapping of individual bacterial species and their microbial products within a mixed bacterial community growing in situ on surfaces.

With this work, we have demonstrated the application of these tools to two types of mixed bacterial communities, those comprised of two different strains of the same species (inter-strain), and those in which the bacteria come from different species (inter-species). The examples presented here utilize SIMS MSI and CRM in order to assess the extent and time course for the secretion of chemical messengers. These two tools work well together, in as much as they produce complementary information that can be combined ex post facto to yield a more complete picture of the behavior of microbial communities than either one would provide alone. As applied in the current scenarios these have permitted the unique behaviors of specific strains and species to be identified by the differences in their secretion of chemical messengers. This dichotomy is evidenced specifically by the differences in the production of rhamnolipids and PQS/HQNO between the PAO1C and FRD1 strains of P. aeruginosa picked up by SIMS MSI. The use of MSI here is particularly important, as it is able to detect numerous individual signaling components and to differentiate among the C₇- and C₉-variants of the AQ signaling molecules shown in Scheme 1.

In contrast, CRM is able to distinguish easily between isomers, such as PQS/HQNO, both at m/z 260.17, and the corresponding pair of C₉ AQs, C₉-PQS/NQNO at m/z 288.20. As it pertains to the interaction between P. aeruginosa PAO1C and E. coli DH5α, CRM was able to clearly establish that chemical messaging occurs over lengths scales that can be much larger than the visible size of the colony, the chemical identity of the major messenger, AQNO, and the time course of its presence distal from the original P. aeruginosa inoculation. We know that the chemical products produced by bacterial pathogens are critical to host colonization, infection, and virulence not just by virtue of their identity, but also by the nature of their distribution in space and time. These latter characteristics are now beginning to be discerned in the context of multiple, co-cultured strains and species.