Depth distributions of signaling molecules in Pseudomonas aeruginosa biofilms mapped by confocal Raman microscopy

Abstract

Pseudomonas aeruginosa is an opportunistic human pathogen implicated in both acute and chronic diseases, which resists antibiotic treatment, in part by forming physical and chemical barriers such as biofilms. Here, we explore the use of confocal Raman imaging to characterize the three-dimensional (3D) spatial distribution of alkyl quinolones (AQs) in P. aeruginosa biofilms by reconstructing depth profiles from hyperspectral Raman data. AQs are important to quorum sensing (QS), virulence, and other actions of P. aeruginosa. Three-dimensional distributions of three different AQs (PQS, HQNO, and HHQ) were observed to have a significant depth, suggesting 3D anisotropic shapes—sheet-like rectangular solids for HQNO and extended cylinders for PQS. Similar to observations from 2D imaging studies, spectral features characteristic of AQs (HQNO or PQS) and the amide I vibration from peptide-containing species were found to correlate with the PQS cylinders typically located at the tips of the HQNO rectangular solids. In the QS-deficient mutant lasIrhlI, a small globular component was observed, whose highly localized nature and similarity in size to a P. aeruginosa cell suggest that the feature arises from HHQ localized in the vicinity of the cell from which it was secreted. The difference in the shapes and sizes of the aggregates of the three AQs in wild-type and mutant P. aeruginosa is likely related to the difference in the cellular response to growth conditions, environmental stress, metabolic levels, or other structural and biochemical variations inside biofilms. This study provides a new route to characterizing the 3D structure of biofilms and shows the potential of confocal Raman imaging to elucidate the nature of heterogeneous biofilms in all three spatial dimensions. These capabilities should be applicable as a tool in studies of infectious diseases.

INTRODUCTION

Bacterial populations often manifest as biofilms,^1,2 which are densely packed matrices that, depending on the bacterial species, stages of growth, and environmental conditions, can exhibit different morphological structures. Pseudomonas aeruginosa is an opportunistic bacterial pathogen for which biofilm occurrence has been shown to contribute to antibiotic resistance, thus exacerbating problems associated with related infections and diseases.^3–6 P. aeruginosa biofilms are typically heterogeneous in their arrangement of cells, metabolites, extracellular polymeric substances (EPS; polysaccharides, DNA, proteins, etc.), and other molecules.^1,2 Generally, P. aeruginosa biofilm formation begins with the attachment of cells, followed by irreversible adhesion of planktonic cells and secretion of EPS as the biofilm expands over the surface, developing a heterogeneous three-dimensional (3D) structure that supports the bacterial function. This heterogeneous three-dimensional structure composed of a complex, spatially and temporally organized collection of cells and secreted biomolecules, in which cells live in unique biological and chemical conditions, is a central distinguishing feature of biofilms.

Studies have shown that population density-dependent quorum sensing (QS) used by P. aeruginosa plays a role in coordinating actions with biofilms.^7–9 QS is a mechanism by which bacteria regulate their collective behavior by monitoring cell population density through the production of small signaling molecules, and it has been shown to play a role in various actions of different bacterial species relating to their growth, surface motility, virulence, antibiotic production, and biofilm development.^8,10–12 The four QS cascades used by P. aeruginosa, named Las, Rhl, PQS, and IQS, have interconnection and overlap.^13,14 The PQS system utilizes 2-heptyl-3,4-dihydroxyquinoline [Pseudomonas quinolone signal (PQS)] as its signal molecule. However, additional alkyl quinolones (AQs) are part of the PQS synthesis pathway, which notably include 2-alkyl-4(1H)-quinolones such as 4-hydroxy-2-heptylquinoline (HHQ; precursor to PQS) and 4-hydroxy-2-heptylquinoline N-oxide (HQNO), an AQ known for its antimicrobial activities.^15–17 Studies have shown that PQS and other AQs can aggregate on surface-growing P. aeruginosa and in stationary biofilms.^18–21 PQS production is spatially and environmentally dependent, a fact that can be used to explore the stress level experienced by bacteria.¹⁹ However, it is not clear why spatial distributions of AQs are so different from molecule-to-molecule and sample-to-sample. Furthermore, it is well known that biofilms frequently vary in thickness within the x–y plane, and since typical chemical imaging studies focus on in-plane distributions of the secreted factors, there is little information about how AQ production varies along the z-direction within the biofilm matrix.

A great deal has been learned about structural characteristics of bacterial biofilms by confocal scanning laser microscopy (CSLM),^22–24 a fluorescence microscopy method that can provide valuable information on labeled components, such as cells, structures, and tissues, but the need to provide a label limits the range of imaging experiments that can be performed. A number of different molecular spectroscopy and chemical imaging methods have been developed to augment CSLM for biofilm studies.²⁵ Among them, confocal Raman microscopy (CRM) has been used widely in different systems because it provides high spatial resolution and abundant molecular information, and it can be applied in situ, thus providing molecular distributions that can be used in tandem with phenotypic and genetic information to characterize the biofilm behavior.^18–20,26 Using CRM, spatial maps of important secreted molecules in microbial communities can be constructed both in the x–y plane and along the z-axis (depth), thus making it possible to construct 3D chemical distribution maps from the hyperspectral image data. 3D CRM imaging data can also be combined with chemometric techniques, such as principal component analysis (PCA), in order to extract even more chemical information.^19,27–29 CRM depth profiling has been used, for example, to detect structural variations at polymer–liquid interfaces for the estimation of axial spatial resolution and to detect the self-assembly of hybrid phospholipid bilayers.^30,31 This approach has been used in the spatiotemporal characterization of a 3D biofilm of E. coli, showing clear spatial and chemical variations among bacterial cells, proteins, and secreted polyhydroxybutyrate at different biofilm growth stages.³² 3D Raman imaging has been used to characterize differences in chemical compositions at different depths in bacteria-infected human skin tissues³³ and has also been used in genotoxicity and ecotoxicity testing by comparing the chemical and spatial distributions of biosynthetically modified polysaccharides with those produced by native microbes.³⁴

Here, we explore the use of 3D confocal Raman microscopy for the investigation on the spatial and chemical variations of secreted factors and signaling molecules in P. aeruginosa biofilms. We start by characterizing a fabricated multilayer-multicomponent polymer model system to mimic the stacked and heterogeneous structure in bacterial biofilms and thus define the 3D imaging characteristics of our laser scanning confocal Raman microscope. With these imaging characterization experiments in-hand, CRM is applied to the non-invasive, label-free acquisition of 3D spatial distributions of alkyl quinolones (PQS, HQNO, and HHQ) in P. aeruginosa biofilms. 3D chemical imaging reveals that AQ spatial profiles differ among different AQs and different locations (center vs edge) of the biofilm. Surprisingly, we find that the 3D PQS profiles display highly anisotropic vertically oriented sheets, while HQNO and HHQ typically exhibit rod-like structures. The difference in the 3D distributions of PQS as a function of x–y position in the biofilm may suggest that cells at the edge and the center experience different metabolic status and stress levels. These results represent the first 3D reconstructions of secreted factor distributions from P. aeruginosa and thus illuminate the relationship between the 3D biofilm structure and bacterial behavior.

RESULTS AND DISCUSSION

CRM 3D profiling performance

Raman Images, such as those shown in Figure 1, acquired in air from volumes internal to thick samples must be interpreted.^35,36 In order to address this point, the 3D imaging capabilities of the confocal Raman microscope were characterized using a custom model system prepared by transferring a monolayer of polystyrene (PS) beads (diameter = 3 µm) to a clean, hydrophilic silicon surface and then covering the PS beads with a thin (∼150 µm) polydimethylsiloxane (PDMS) film. This 3D multilayer system was designed to reproduce the optical and physical characteristics of a heterogeneous biofilm in all three spatial dimensions. The model system was constructed from silicon, PS, and PDMS, all of which have characteristic Raman spectra that can be easily distinguished (Fig. S1). Full Raman spectra were recorded at each image pixel at a given confocal position, and depth profiling was performed at 2–3 µm increments to acquire a total of seven to nine layers in the z-direction from bottom to top. Images acquired in studies of this model system, shown in Figs. S2 and S3, demonstrate high quality chemical imaging by CRM over depths up to 20 µm below the nominal surface and the ability to distinguish among chemically distinct components in all three spatial dimensions. Furthermore, fitting of the Raman images in the PS spectral window produces an estimate of the depth resolution, which is inferior to x–y resolution, as expected, due to a combination of the longitudinal point spread function and spherical aberration.

FIG. 1.

Raman images generated with different spectral windows at the edge of a 48 h P. aeruginosa biofilm. (a) Raman images obtained using spectral filters of 1330–1380 cm⁻¹ and 1640–1675 cm⁻¹ are displayed from 0 to 15 µm below the surface. Each image is 100 × 100 µm² and 80 × 80 pixel². (b) Composite Raman image obtained by overlaying AQ (1330–1380 cm⁻¹) and amide I (1640–1675 cm⁻¹) images at a depth of 0 µm. (c) Raman image of the AQ window (1330–1380 cm⁻¹) at z = 0 µm. The black line indicates the position of the cross-sectional cut, and the white dashed box indicates the region of interest in (d). (d) Raman image stack of six layers of AQ images, cross sectioned along the line indicated in (c). Scale bar = 20 µm in all panels.

3D spatial distributions of AQs in biofilms

Surface-attached biofilms of P. aeruginosa were investigated to gather information about the 3D spatial distribution of important alkyl quinolones. Biofilms of P. aeruginosa PAO1C were grown on a minimal medium containing 12 mM glucose solidified with 1.5% agar for 48 h,¹⁹ and CRM chemical imaging profiles were acquired at both the center and the edge of the biofilms to probe the effect of the biofilm boundary. Confocal depth profiling was conducted by stepping the z-axis confocal position in 3 µm steps through six layers, from 0 to 15 µm.

A spectral window of 1330–1380 cm⁻¹ was chosen for mapping because it encompasses the spectral region containing quinolone ring stretching vibrations from a variety of AQs, thereby identifying AQ production. Another 1640–1675 cm⁻¹ window was mapped to identify the spatial distribution of the amide I stretch associated with peptides or proteins. Raman images acquired over these windows acquired from the biofilm edge are compared in Fig. 1 for the AQ (left) and amide I (right) distributions at depths from 0 to 15 µm. Both AQ and peptide/protein features are evident in the Raman images at all depths with very similar x–y distributions within each window, indicating that within the longitudinal resolution of the confocal measurement, viz., Fig. S2, the features span a significant vertical depth. Moreover, the high intensity areas of the AQ and peptide/protein windows in the Raman images are not uniform but are locally concentrated. In addition, even though the local shapes of the features vary, being sharp and needle-like in the AQ window and more globular in the amide window, the AQ and amide features are clearly co-located throughout the entire 15 µm depth of the hyperspectral image stack, as has been observed previously in planar CRM images.¹⁸

Figure 1(b) depicts a composite image obtained by overlaying amide I (green channel) and AQ (red channel) Raman images at the top surface, i.e., depth = 0 µm. It is clear that the AQ and amide features are highly correlated in the x–y plane. Figures 1(c) and 1(d) show x–y (0 µm) and x–z images for the region of interest (ROI) indicated by the white dashed box in Fig. 1(c). The black line indicates the position of the cross-sectional cut shown in the stacked volume view AQ images in Fig. 1(d). The different AQ patterns in the cross-sectional images in Fig. 1(a) at different depths are consistent with the volume view in Fig. 1(d), indicating that AQ distributions vary as a function of z-depth and that, although AQs are aggregated at all depths, the aggregation pattern varies with depth.

The depth profiles obtained from the center of the biofilm at the same time point (48 h) (Fig. S4) show similarities and differences in the AQ and amide I profiles obtained at the biofilm edge. Changes in AQ distributions with depth are especially clear when comparing images at 0 and 15 µm. The anisotropic features in the upper right and lower left at 0 µm are absent at 15 µm, and there is a new feature in the lower right, the top of which begins at 9 µm. It is very clear that the AQ aggregates display a different shape (needle-shaped and long) than the more globular amide I aggregates. Figures S4(b)–S4(d) illustrate the correlation of AQ and amide I features and their variation with depth. In general, only a few locations showed a clear co-location, in contrast to the edge, where the co-location is more common. Figures S4(c) and S4(d) clearly confirm that the AQ aggregates vary with depth.

3D PCA analysis of AQ profiles

Raman images were analyzed by principal component analysis (PCA) to allow a more comprehensive observation of spatial and chemical information than available from Raman images alone. Smaller spatial subranges were analyzed to better depict the molecular spatial distributions in ROIs. Figure 1(c) indicates one such subrange ROI (white dashed box, 50 × 50 µm²) at the edge of a biofilm formed by wild-type P. aeruginosa, and the PCA results for this ROI are shown in Fig. 2. Z-score features corresponding to PQS vibrational bands (1159, 1245, 1372, 1466, and 1603 cm⁻¹) dominated principal component 1 (PC1) loading plots [Fig. 2 (left)] in addition to another feature at 1658 cm⁻¹, which corresponds to the amide I vibrational band. This observation confirms the co-localization of PQS with the peptide, as has been reported previously.¹⁸ The heat maps for PQS/peptide PC1 indicate the high abundance of PQS in aggregates that appear as ∼3 × 4 µm² globular shapes at several locations in the x–y plane. Interestingly, these aggregates, which appear globular in the x–y heat maps, extend over a depth of at least 15 µm, indicating that they are highly anisotropic (3 µm, 4 × ≥15 µm²) rod-shaped aggregates.

FIG. 2.

PCA analysis of Raman scans representing the edge of a P. aeruginosa biofilm. PC1 and PC2 loading plots and heat maps are shown at six z-depth positions. The scan direction goes from the top of the biofilm (z = 0 µm) to the bottom (z = 15 µm). The step size is 3 µm. Scale bar = 8 µm.

In contrast to PC1, PC2 loading plots exhibit major features (681, 717, 1359, 1435, 1451, and 1511 cm⁻¹) corresponding to HQNO, with some minor features similar to those in PC1, due to PQS. The PC2 heat maps are largely complementary to the PC1 heat maps at all depths. While the aggregates of PQS and HQNO are spatially correlated throughout the z-depth range, the high amplitude PC1 features tend to occur as globular features at the distal ends of long-needle-like PC2 features. Furthermore, the PC2 features decrease in intensity with depth. The HQNO-dominated PC2 aggregates are long and narrow in the x–y plots at all depths and are estimated to be 3 × 20 µm² (width × length), with a depth of ≥15 µm, indicating that HQNO is consistently observed at all depths.

Similarly, Fig. 3 and Fig. S4 show the PCA maps and Raman images, respectively, associated with the biofilm center. Like the edge profile, the center of the biofilm exhibits features characteristic of both PQS and HQNO at 48 h. Interestingly, in Fig. 3, PC1 at 0–9 µm shows features at the quinoline stretching frequency for both HQNO (1359 cm⁻¹) and PQS (1372 cm⁻¹) along with other PQS-related features (1466 and 1603 cm⁻¹) and a strong feature at 1658 cm⁻¹ characteristic of the amide I vibration of peptides. However, after 9 µm, only PQS-related features are observed. In the heat maps, PC1 is characterized by a globular aggregate (lower left) of ∼3 × 3 µm², with a depth of 15 µm or more, similar to the PQS features observed at the edge of the biofilm. The cylindrical rod associated with this feature is tilted relative to the z-axis, as can be seen from the apparent shift in position from 0 to 15 µm in the heat maps. Finally, the shift in presentation near 9 µm in depth is also evident in the 1330–1380 cm⁻¹ Raman images in Fig. S4. This region encompasses both HQNO and PQS features, but their x–y distributions shift dramatically from 6 to 12 µm.

FIG. 3.

PCA analysis of CRM microspectra acquired at the center of a P. aeruginosa biofilm. PC1 and PC2 loading plots and heat maps are shown at six depths from 0 (top) to 15 (bottom) μm. Heat maps plot the relative magnitudes of the PC as a function of spatial position by color according to the scales at the right of each map. Scale bar = 2 µm in all panels.

PC2 loading plots in Fig. 3 are clearly dominated by HQNO features (681, 717, 1359, 1435, 1451, and 1511 cm⁻¹). The PC2 spatial distribution heat maps indicate that the HQNO features are anisotropic (broadened versions of the needle-like structures seen for PC2 in Fig. 2) and decrease in deeper regions of the biofilm (z ≥ 12 µm). The HQNO-related features are ∼10 × 3 × 10 µm³ (length × width × depth). These HQNO features clearly differ from those at the edge, being both shallower and shorter. The differences in spatial profiles of features associated with HQNO and PQS shown in Fig. 3 suggest differences in the ways that the bacterial communities synthesize and secrete different AQs and that the x–y distributions are relatively insensitive to their location in depth. Those differences in shapes of the PC1 and PC2 features observed in Fig. 3 demonstrate the complexity of bacterial behaviors during biofilm construction, and these details of the 3D shape of different aggregates are revealed for the first time by confocal Raman imaging.

AQ spatial distributions for the lasIrhlI QS mutant

In order to broaden the understanding of the 3D secretion patterns in P. aeruginosa, changes in the spatial distributions of secreted factors were monitored in a QS mutant, lasIrhlI. The lasIrhlI mutant is deficient in production of both Las and Rhl cascade signals, and in this mutant, the PQS pathway halts at the precursor, HHQ, as confirmed by previous spectroscopic studies.³⁷ Although the three-dimensional distributions of HHQ in P. aeruginosa biofilms have not been previously studied, it is worthwhile to do so since it is a required precursor for PQS and thus plays an important role in quorum sensing.^16,38

The morphological aggregation patterns observed by bright field microscopy (Fig. S5) are very different at the center and edge for both the wild-type and mutant strains at 48 h post-incubation. Confocal Raman imaging results in Fig. S6 show that the AQs (1330–1380 cm⁻¹) are aggregated nonuniformly, showing dendritic features in the x–y plane starting 6 µm below the local surface. In contrast to biofilms formed by wild-type P. aeruginosa, the amide I spectral window exhibits a very similar spatial distribution, with both AQ and amide I windows showing increased intensities deeper (z ≥ 6 µm) in the biofilm.

PCA was then applied to elucidate the details of AQ secretion in the lasIrhlI mutant, as shown in Fig. 4. The ROI is 15 × 15 µm² and 30 × 30 pixel². In contrast to the wild-type data from both edge and center biofilm regions, the ROI in Fig. 4 exhibits a single dominant principal component, with the PC loading plots displaying distinct features at 709, 1176, 1354, 1503, 1554, and 1596 cm⁻¹, consistent with the vibrational spectrum of HHQ. An extended buried feature together with a small strong globular component is observed starting at 6 µm and extending to 15 µm in depth, illustrating the heterogeneity of HHQ production with depth in the biofilm. The size of the buried globular feature is 2–3 µm (length × width), beginning at a depth of 6 µm but maximizing intensity at 12 ± 3 µm. The relative sparseness of HHQ features in the 0–3 µm region is also reflected in the low signal-to-noise ratio of the z-score plots. The highly localized nature of this feature and its similarity to the size of a P. aeruginosa cell may suggest that the feature arises from HHQ localized in the vicinity of the cell from which it was secreted. This result is very different from the features observed in the PCA analysis of wild-type P. aeruginosa at both the edge (Fig. 2) and the center (Fig. 3) of the biofilm, where PQS-associated features (PC1) are smaller but deeper. These differences suggest that HHQ and PQS play different roles in quorum sensing, pointing out a potential method to relate the biofilm structure with AQ chemical profiles.

FIG. 4.

PC1 loading plots and heat maps of Raman scans at the edge of a biofilm formed by the lasIrhlI QS mutant of P. aeruginosa. Scale bar = 3 µm in all panels.

The spatial distribution of Raman intensities at the center of a lasIrhlI mutant biofilm is shown in Fig. S7 for both AQ and amide I regions. Similar to the edge region shown in Fig. S6, the images show clear co-localization of amide I and AQ features, as is evident from the yellow-coded areas in Fig. S7(b). However, the globular shapes of both AQ and amide I features in the center of the mutant biofilm are dramatically different than the dendritic distributions observed near the edge.

Figure S8 shows the results of PCA analysis of the center of a lasIrhlI mutant biofilm for an ROI defined by the white dashed box (15 × 15 µm² and 30 × 30 pixel²) in Fig. S7(c). Similar to the edge data shown in Fig. 4, the loading plots are dominated by features characteristic of HHQ at 709, 1354, 1554, and 1596 cm⁻¹ and the amide I feature at 1640 cm⁻¹. Similar to the edge images, the highest intensity area consists of two 2–3 µm globular features separated by ∼5 µm. The feature persists over a significant depth for the right-hand feature of the pair but is localized near z = 3 µm for the left-hand feature. This kind of pattern for HHQ is very similar to that observed for PQS from wild-type P. aeruginosa at both center and edge biofilm locations, consistent with the role of HHQ, as a precursor for PQS.

3D profiles of antibiotic-treated P. aeruginosa biofilms

Though it is known that P. aeruginosa reacts to many antibiotics by producing spatially varied AQ profiles in the x–y plane, there are no observations in the z-dimension. Therefore, wild-type P. aeruginosa biofilms were challenged with a proximal exposure to tobramycin, an antibiotic that is commonly used to treat P. aeruginosa infections,²⁰ as shown in Fig. S9, and 3D profiles of secreted components in the bacterial biofilms were acquired. Raman and PCA analysis focused on the differences between the edges of the P. aeruginosa colony proximal to, and distal from, the area of antibiotic placement.

Figure 5 shows both Raman images and PCA data from depths of 0–6 µm at the proximal edge of a tobramycin-treated biofilm, while Fig. S10 shows confocal imaging data from the distal edge. On the proximal edge, the signal intensities diminished dramatically at z ≥ 9 µm (data not shown). At depths 0–6 µm, the locations of amide I and AQ regions correlate, but their signal intensities vary. For instance, there is a very strong AQ feature at the right edge of the ROI, but it matches to a much weaker and more diffuse feature in the amide I heat map. On the distal side (Fig. S10), PQS and HQNO aggregates exhibit shapes similar to those observed in Fig. 1 on the edge of a native wild-type P. aeruginosa biofilm, and the features extend to a depth of ≥12 µm. Clearly, the distal-side features (6 × 4 × 15 µm³), similar to the PQS features in Fig. 1, are much thicker than those observed on the side proximal to tobramycin.

FIG. 5.

Raman images and PCA results from the proximal edge of a P. aeruginosa colony exposed to tobramycin at 48 h at different depths. (a) Raman images obtained using spectral windows at 1330–1380 cm⁻¹ and 1640–1675 cm⁻¹ are displayed from 0 to 6 µm below the surface. Each image is 50 × 50 µm² and 80 × 80 pixel², and the white dashed ROI used for PCA analysis is 18.75 × 18.75 µm² and 30 × 30 pixel². Scale bar = 10 µm. (b) PC1 and PC2 heat maps from the x–y ROI defined in (a). Scale bar = 4 µm. (c) Representative PC1 and PC2 loading plots.

Overall physical and chemical characteristics

In order to put the results obtained above into context, the overall features of the biofilms were characterized and are summarized here. Figure S11 shows the 3D image, obtained by Confocal Laser Scanning Microscopy, of a biofilm grown from a green fluorescent protein (GFP) tagged P. aeruginosa strain under the same growth condition as the non-tagged wild-type strain. The cross-sectional views show that the 48 h biofilm thickness at the edge is ∼15 µm, while the center, it is ∼35 µm.

Furthermore, the estimated spatial profiles of AQs obtained over the entire set of conditions studied in this work are summarized in Table S1. These measurements indicate that the wild-type P. aeruginosa populations secrete AQs in patterns that are sensitive to the position within the biofilm and the position relative to an antibiotic challenge. The lasIrhlI QS mutant secretes a different dominant AQ, HHQ, which is sensitive to the position as well.

Finally, as is evident from Figs. 5(a) and 5(b), information acquired from Raman images augments that acquired from heat maps of principal components. This is further illuminated in Fig. S12, which shows distinct individual spectra acquired from two different locations both of which contribute to high intensity pixels in the Raman image.

CONCLUSIONS

The activities of biofilms are critical for a variety of bacterial behaviors, but their 3D attributes are difficult to characterize chemically. Confocal Raman microscopy imaging, explored here, represents a nondestructive technique for characterization of the 3D spatially resolved composition of heterogeneous bacterial biofilms at the μm-scale. Data can be further elaborated with multivariate statistical tools, such as principal component analysis. In the context of CRM microspectra, loading plots can be compared to the vibrational spectra of standard compounds to identify principal components with one, or a few, dominant molecular species, and heat maps showing x–y spatial distributions of the components allow 3D reconstructions of secreted molecular components in biofilms. The results presented here show that CRM is capable of relaying information about the volume, shape, biochemical composition, and spatial distribution of aggregates of secreted compounds, such as the alkyl quinolones used extensively by P. aeruginosa.

The depth profiling and reconstructed 3D maps were used here to derive unique insights into the chemical architecture and composition of P. aeruginosa biofilms. Similar to observations from 2D imaging studies, spectral features characteristic of AQs (HQNO or PQS) and the amide I vibration from peptide-containing species were found to correlate across a broad range of depths. The anisotropic AQ and amide I features were observed to have a significant depth, suggesting 3D anisotropic shapes—sheet-like rectangular solids for HQNO and extended cylinders for PQS. Furthermore, the cylinder-shaped PQS features were typically located at the tips of the rectangular-shaped HQNO features. In the case of the QS-deficient mutant lasIrhlI, an extended feature was observed together with a small strong globular component. The highly localized nature of this feature and the similarity of its lateral extent to the size of a P. aeruginosa cell suggest that the feature arises from HHQ localized in the vicinity of the cell from which it was secreted. Furthermore, differences between HHQ and PQS further suggest that these two secreted factors play different roles in quorum sensing.

The difference in the shapes and sizes of the aggregates of the three alkyl quinolones in wild-type and mutant P. aeruginosa is likely related to the difference in the cellular response to growth conditions, environmental stress, metabolic levels, or other structural and biochemical variations inside biofilms. This may further suggest that the understanding of quorum sensing and the resulting regulation mechanisms involved in the development of P. aeruginosa biofilms may depend on the spatial distributions of signaling molecules and their overall number density.

SUPPLEMENTARY MATERIALS

See the supplementary material for additional experimental details and data analysis.

DATA AVAILABILITY

The data that support the findings of this study are available from the corresponding author upon reasonable request.