Adhesive interactions within microbial consortia can be differentiated at the single-cell level through expansion microscopy Authors: Pu-Ting Dong, Wenyuan Shi, Xuesong He, Gary G. Borisy Published: 2024 ABSTRACT SignificanceA single-cell understanding of microbe–microbe interactions is critical for unraveling the organization and dynamics of microbial communities. Through an unconventional application of expansion microscopy, we oppose the adhesive force holding microbes together by an expansion force pulling them apart, resulting in microbial separation dependent on the strength of microbial adhesion. Our approach establishes a proof of principle for differentiating adhesive interactions within microbial consortia at the single-cell level. Microbes in nature are frequently found as multispecies communities growing on surfaces as biofilms (1–4). The importance of spatial organization for understanding microbial ecosystems (5–8) has been highlighted in recent reviews on the gut microbiome (9, 10), oral microbiome (11–15), and polymicrobial infection (16, 17). Because many microbe–microbe interactions are short range, it is critical to understand spatial organization at the single-cell level. A key approach in meeting this objective has been simultaneous imaging of different bacterial taxa in a complex community through multiplexed fluorescence imaging (18–20). The information obtained from such studies establishes the spatial proximity relationship of individual microbes to other microbes or to host tissue. However, connecting this spatial information to function remains challenging. One aspect of function is the adhesion of microbes to each other or to host tissue. Adhesion is of fundamental importance in microbial biofilms subject to flow, such as the oral microbiome (21). Because of salivary flow, any oral microbe that does not adhere to host oral tissue or teeth either directly or indirectly via interacting with other adhered microbes will be flushed into the gastrointestinal tract. Adhesion is likely also instrumental to the building of oral biofilms (22) and the emergence of spatial organization as it dictates the stability of microbe–microbe associations. Coaggregation has been used with pure cultures in pairwise combinations in vitro to determine the relative tendency of bacterial species to adhere to each other (2, 23, 24). However, this tool while useful as an in vitro assay is not suited to evaluating adhesive interactions in a complex, multispecies community in situ. Alternative approaches are needed to fill this gap. Expansion microscopy is a novel approach introduced as a resolution-enhancing method (25). It is based on coupling target molecules or structures to a swellable, polyelectrolyte gel. Isotropic swelling of the gel causes the decrowding of target biomolecules and thus increased resolution with conventional microscopes. The approach has been used to achieve superresolution imaging of protein complexes (26), RNA complexes (27), and microtubule nanostructures (28). Of note, expansion microscopy also causes the samples to become nearly completely transparent with the refractive index of the expanded gel close to that of water, enabling deep tissue imaging (29). Expansion microscopy has also been applied to microbial communities after treatment with cell wall digesting enzymes (30). Although the purpose of that study was to mechanically resolve bacterial species through differences in their cell wall expandability, the results also showed that bacterial cells could be separated from each other independent of the extent to which their walls were expanded. In this paper, we describe an unconventional application of expansion microscopy to differentiate microbial adhesive interactions at the single-cell level. Whereas in conventional expansion microscopy, the objective is to render the gel-embedded tissue as uniform as possible through enzymatic digestion to achieve isotropic expansion, our goal is to maintain the integrity of the microbial cells while expanding the extracellular space between them. The key principle we introduce is heterotropic expansion—the application of expansion force to individual cells within microbial communities by embedding them into a swellable gel without swelling the cells themselves. The expansion process, driven by electrostatic repulsion of negatively charged groups on the gel polymer backbone, generates a pulling force on objects contained within the gel, termed as expansion force. Our rationale is that noninteracting or weakly interacting microbes will be pulled apart as manifested by an increase in their intermicrobial distance, whereas strongly adherent cells will resist the expansion force and remain close together (Fig. 1). Different from the standard expansion procedure, we tried to maintain the integrity of bacterial cell–cell adhesion by establishing minimally perturbing expansion conditions through entirely omitting protein digestion, fixation, and cross-linking. We first evaluated heterotropic expansion using mono- and dual-species in vitro biofilms. After establishing minimally perturbing expansion conditions, we applied our approach to investigate the physical interactions in a three-species consortium model consisting of Fusobacterium nucleatum and two Streptococcus species, Streptococcus mutans and Streptococcus sanguinis under conditions that facilitate interspecies coaggregation. We also investigated an epibiont association, a recently described episymbiosis between an ultrasmall bacterium Nanosynbacter lyticus strain TM7x, and its host bacterium, Schaalia odontolytica strain XH001 (31). In combination with spectral imaging, our approach establishes a proof of principle for differentiating adhesive interactions within microbial consortia at the single-cell level. The diagram illustrates application of expansion microscopy for the differentiation of microbe–microbe adhesive interactions within a microbial community. Strongly adhesive bacteria (green and yellow) are not pulled apart during expansion while weakly adhesive bacteria (blue vs green and yellow) are pulled apart. Results Evaluation of the Basic Methodology in a Monospecies Biofilm. Initial experiments were carried out to evaluate whether heterotropic expansion microscopy could be used to decrowd microbial biofilms. This would lay the groundwork for experiments to differentiate microbial adhesive strength within the microbial communities. As a test system, we used S. mutans which is known to self-aggregate (32), resulting in biofilms that are spatially heterogeneous. A key property that needed to be established was the behavior of noninteracting objects as a negative control for microbial interactions. For this purpose, fluorescent beads, 2 µm in diameter, were used as a proxy for noninteracting objects. Beads were mixed with S. mutans and became incorporated into the monospecies biofilm as the culture grew on saliva-coated glass substrate. The bead-containing biofilm was then embedded in a swellable gel and subjected to the expansion procedure as described in Material and Methods and reported by Chen et al. (25). In the absence of enzymes such as mutanolysin that specifically digest the bacterial cell wall (30), expansion microscopy can be carried out on microbial cells without compromising cell integrity. During expansion of the biofilm, the distance between beads increased and their concentration per unit area decreased (Fig. 2) as expected for a threefold expansion of the gel. This result is consistent with the lack of any significant adhesive interaction between the beads. Imaging of beads from the same field of view before and after expansion (SI Appendix, Fig. S1) confirmed that beads were retained within the gel after expansion and maintained their positions relative to each other. The expansion procedure separates bacteria away from each other in a monospecies biofilm. A representative image of monospecies biofilm before expansion (A) and after expansion (B). Monospecies S. mutans (SYTO nine-labeled; green) biofilms were cultured with fluorescent beads (magenta) serving as an internal standard for nonadhesive objects. Images obtained with 63× oil immersion objective (NA = 1.4). Before expansion, the close packing of S. mutans in the biofilm hindered visualization of individual cells (Fig. 2A). After expansion, S. mutans cells showed separation reflecting their initial heterogeneous spatial distribution (Fig. 2B). Aggregates of cells evident before expansion were still evident after expansion but were less densely packed as would be expected if the cells were weakly adherent. Importantly, cells were pulled away from each other sufficiently to allow for single-cell visualization. The integrity of the S. mutans cells was maintained in the expansion procedure as judged by retention of their size and shape. The separation of individual biofilm cells while maintaining microbial cell integrity establishes that expansion microscopy can be used to decrowd bacterial biofilms. Expansion Microscopy Analysis of a Dual-Species Biofilm. The next question we addressed was whether adhesive interactions between microbes in biofilms could be evaluated quantitatively. For this purpose, we evaluated the application of expansion microscopy to a dual-species biofilm system. When more than one species is present in a biofilm, reporter molecules are necessary to distinguish the different taxa comprising the biofilm. We used genetically encoded fluorescent proteins as reporter molecules for the dual-species biofilm experiments. Biofilms were grown from mCherry-expressing S. mutans and GFP-expressing S. sanguinis that were mixed 1:1 at an initial OD600 of 0.05 in Brain Heart Infusion (BHI) broth. Fluorescent blue beads were added to the mixture as an internal control for measuring the local expansion factor as described before. The fluorescence development of GFP and mCherry is known to require their correct protein folding which is dependent on oxygen (33). Therefore, dual-species biofilms were grown by incubating them aerobically at 37° C for 24 h. Imaging of the dual-species biofilms was carried out using spectral imaging followed by linear unmixing to separate the individual fluorescent signals (Materials and Methods). Although not strictly necessary for the dual-species biofilm, we employed the spectral approach in anticipation of subsequent experiments involving multispecies biofilms. Since the growth of S. mutans biofilms is known to be strongly dependent on sucrose concentration, we explored growth conditions to achieve a balanced biofilm of both S. mutans and S. sanguinis. At 1% (w/v) sucrose concentration, S. mutans overtook the coculture, while 0.1% (w/v) sucrose allowed the coexistence of both Streptococcus species. Thus, 0.1% (w/v) sucrose was used to grow the dual-species biofilm for expansion analysis. Using the same expansion procedure as used for the monospecies biofilm, dual-species biofilms were imaged before and after expansion. Before expansion, the biofilm of S. mutans and S. sanguinis contained closely packed aggregates. Although some individual cells could be discerned, many cells were too close to or overlapped neighbors to allow for unambiguous visualization (Fig. 3A). However, after expansion, both S. mutans and S. sanguinis could be visualized as individual cells in single confocal image planes (Fig. 3B), indicating that they were pulled away from each other in the X-Y plane. Confocal imaging at different axial planes showed that S. mutans was more abundant near the glass substrate while S. sanguinis was more abundant toward the outer periphery of the biofilm away from the glass bottom of the culture dish. Thus, cells were also pulled apart in the z direction, revealing different spatial distribution patterns for these two streptococci even within the same biofilm. The expansion procedure allows visualization of individual bacteria within a dual-species biofilm. A representative image along with a high-magnification image callout of dual-species biofilm before (A) and after expansion (B). Dual-member Streptococcus biofilms were cultured with fluorescent beads serving as an internal standard. Color annotation: fluorescent beads (magenta), mCherry-encoding S. mutans (yellow), and GFP-encoding S. sanguinis (cyan). Spatial proximity relationships of cells within the biofilm were quantified using a program developed for image analysis in microbial ecology, daime (34). Daime quantifies the spatial relationship of two populations by computing a pairwise correlation value as a function of the distance between individual objects. A correlation value larger than 1 means two populations tend to cluster together, less than 1 means they tend to repel each other, and equal to 1 means they are randomly distributed. Output of the analysis can include averaged pairwise correlation values with 95% CI (SI Appendix, Figs. S2 and S3). As shown in Fig. 4A, the spatial arrangement of fluorescent beads before and after expansion showed apparent repulsion at short distances and random distribution above a threshold distance. The apparent repulsion can be explained by the physical size of the beads which was 2 µm. Before expansion, the apparent repulsion below 2.5 µm is a consequence of two beads not being able to occupy the same space. After expansion, the threshold of 7.5 µm is a consequence of the linear expansion of approximately threefold as found before in the monospecies S. mutans biofilm experiments. The random distribution of the beads above these thresholds indicates that they had no intrinsic attraction to or repulsion from each other. Thus, the quantitative analysis of the bead distribution confirmed that the beads may considered as noninteracting, reference objects for interpreting the impact of expansion on the microbes of the biofilm. Linear dipole analysis quantifies spatial organization. Proximity analysis of fluorescent beads before and after expansion (A). Linear dipole analysis of images in Fig. 3 shows within-taxon autocorrelation (B and C); intertaxon correlation between S. mutans and S. sanguinis (D). Data: Mean (solid line) with 95% CI (shaded area). A pair correlation value of 1 is highlighted by orange dashed lines in the figure panels. Pairwise correlations were calculated using 500,000 random dipoles per field of view. Proximity analysis of microbes in the dual-species biofilm before expansion was ambiguous because many cells in the aggregates overlapped each other and could not be discerned as individual cells. However, after expansion, proximity analysis became unambiguous because individual cells could be clearly visualized. The pairwise correlation patterns of S. mutans with itself (Fig. 4B) and S. sanguinis with itself (Fig. 4C) both showed apparent repulsion below 1 µm and positive clustering at greater distances with a peak at about 2.5 µm. Apparent repulsion at short distance is a consequence of the cells, like the beads, not being able to occupy the same space. Positive association at greater distance reflects the existence of close aggregates before expansion which is consistent with the known properties of S. mutans and S. sanguinis to self-aggregate within biofilms (35). Since expansion was threefold and the diameters of S. mutans and S. sanguinis are about 0.8 to 0.9 µm, the expected peak in the pairwise correlation curves after expansion was approximately 2.4 to 2.7 µm, which agrees with the observed values of 2.5 to 3.0 µm. This result indicates that many S. mutans and S. sanguinis cells were in close association before expansion and were pulled apart by the expansion process. The pairwise correlation between S. mutans and S. sanguinis was less than 1 over most of the distance range, suggesting apparent repulsion (Fig. 4D). However, this negative association does not indicate an intrinsic repulsive interaction between these two Streptococcus species; rather, it is a mathematical consequence of the positive clustering behavior of each of the Streptococcus species taken individually. The dual-species biofilm confirmed important components of our expansion methodology. These included the compatibility with genetically encoded fluorescent proteins as reporters, the feasibility of imaging expanded gels by spectral imaging, and the quantification of spatial proximity relationships by pairwise correlation analysis. The results quantitatively demonstrate that microbial associations in a biofilm can be investigated at the single-cell level by expansion microscopy. Differentiation of Microbial Interactions in a Three-Species Consortium Model. The results with mono- and dual-species biofilms provided the groundwork for applying heterotropic expansion microscopy to analyze adhesive interactions within multispecies microbial communities. As a model system, we adopted an in vitro three-member (F. nucleatum–S. mutans–S. sanguinis) consortium under conditions that facilitate interspecies coaggregation. The binding of different bacteria to each other via specific adhesion molecules (36, 37) can be visually assayed by the resulting formation of flocculant precipitates (38). Coaggregation is thought to play an essential role in the development of multispecies communities, such as dental plaque biofilms (39, 40). However, coaggregation is a population-level phenotype which is not a direct measurement of the adhesive interactions between individual bacteria. Therefore, it would be significant to deconstruct the coaggregation phenotype at the single-cell level. To differentiate adhesive interactions within a multispecies bacterial system, we designed a minimally perturbing expansion protocol. All three bacterial species were labeled vitally. Wild-type F. nucleatum was labeled with a membrane-specific fluorophore, FM 1-43. The labeled F. nucleatum was mixed with mCherry-expressing S. mutans and GFP-expressing S. sanguinis at a ratio of 1:1:1 in coaggregation buffer (Materials and Methods). The mixture resulted in the formation of flocculant precipitate which settled at the bottom of the tube. Samples of the precipitated aggregates were transferred to a glass bottom dish, and gelation was conducted directly on the aggregates. This minimally perturbing protocol differed from conventional expansion protocols in three ways: No fixation was carried out; no Acryloyl-X molecular linker was employed; and no proteinase K digestion was carried out before expansion. Thus, in this protocol, the samples were essentially unfixed, unlinked, and undigested. Images before expansion were taken immediately after the gelation step. After gelation, expansion was driven by adding water to the aggregates-containing gel and changing the water three times over a period of 1 to 2 h. The size of the aggregates increased after expansion but were less densely packed as would be expected from the separation induced by the expansion process. Before expansion (Fig. 5A), even under a high-magnification 63× oil immersion objective (NA = 1.4) with a pixel size of 65 nm, we could not discern single bacteria inside the aggregates, let alone determine the three-dimensional architecture of these aggregates. However, after expansion, even with a low-magnification 20× objective (NA = 0.8), individual bacteria were clearly distinguishable in single-plane confocal images (Fig. 5B). We also examined the distribution of bacteria within the aggregates in the axial direction by acquiring images along the z-axis and representing the results through orthoslice views (Fig. 5C). The results show that all three bacterial species could be resolved individually in both lateral and axial directions. Expansion microscopy differentiates the interaction between F. nucleatum wild type with two Streptococcus species. (A). Spectral imaging (after linear unmixing) of F. nucleatum–S. mutans–S. sanguinis aggregates before expansion under a 63× oil immersion objective (NA = 1.4). (Scale bar, 10 µm.) (B). Spectral imaging (after linear unmixing) of F. nucleatum–S. mutans–S. sanguinis aggregates after expansion. 20× objective (NA = 0.8). (Scale bar, 50 µm.) (C). Orthoslice views of F. nucleatum–S. mutans–S. sanguinis aggregates in panel (B) in the axial and lateral directions. (D and E). High-magnification views (D) and single-cell image callouts (E) of merged view of F. nucleatum wild type with S. mutans and S. sanguinis. (F and G). Proximity analysis of image in (D) between F. nucleatum and different Streptococcus species aggregates after expansion. A pair correlation value of 1 is highlighted by orange dashed lines in the figure panels. Pairwise correlations were calculated using 500,000 random dipoles per field of view. Color annotation: F. nucleatum (white), S. mutans (red), and S. sanguinis (green). The capacity to resolve individual cells through expansion microscopy enabled us to perform quantitative proximity analysis at the single-cell level on the model consortium. Proximity analysis demonstrated differential interactions between F. nucleatum and S. sanguinis vs F. nucleatum and S. mutans. F. nucleatum showed a strong association with S. sanguinis at very short distances (<3 µm) but essentially a random association at greater distances (Fig. 5 D and F). In contrast, F. nucleatum showed a repulsion to S. mutans at very short distances, changing to a neutral interaction at greater distance (Fig. 5 D and G). The remarkable difference in behavior of F. nucleatum with respect to S. sanguinis and S. mutans was investigated further by acquiring high-magnification views of the aggregates after expansion. As shown in Fig. 5D, the close association of F. nucleatum and S. sanguinis was frequently manifested as overlapping cells. Overlapping S. sanguinis and F. nucleatum in the X-Y plane can be documented in two ways: First, from direct visualization of high magnification views, individual and merged panels show that F. nucleatum is juxtaposed with S. sanguinis (Fig. 5E); second, from the proximity analysis (Fig. 5 F and G), the high positive pairwise correlation value at short distances (<0.5 µm) indicates that the bacterial cells are closer together than a single cell diameter. In contrast, F. nucleatum and S. mutans were nearby or adjacent to each other but still clearly distinct and generally not overlapping. The exclusion at short distances accounts for the apparent repulsion of F. nucleatum and S. mutans shown by the proximity analysis. Whether differential adhesion between F. nucleatum and two Streptococcus species occurred axially as well as laterally was investigated more closely. To facilitate analysis of large fields of view in the expanded samples, we used a 20×/0.8NA objective. This objective was clearly adequate to determine whether cells were pulled apart or not in the X-Y plane. However, the theoretical ~3× lower optical resolution axially vs laterally (~1.2 µm vs 0.4 µm for the 20×/0.8NA objective) prompted experimental evaluation. Consequently, we carried out quantitative proximity analysis of orthogonal views. After threefold expansion, if pairs of cells were not tightly adherent, we would expect them to be separated by approximately 3 μm, whereas if they remained tightly bound to each other, their center-to-center distance would be approximately 1 μm. An axial resolution of 1.2 μm is sufficient to distinguish between these two alternatives. Similar to the results in the X-Y plane, the results of the orthogonal views (SI Appendix, Fig. S2) showed that F. nucleatum and S. sanguinis remained tightly associated axially after expansion, whereas F. nucleatum and S. mutans were pulled apart. The differential proximity results for two similarly sized Streptococcus species provide experimental validation that the 20×/0.8NA objective was sufficient to distinguish adhesive interactions between F. nucleatum and the two species of Streptococcus in three-dimensions. We interpret these images and the proximity analysis results to signify that F. nucleatum was in tight association with S. sanguinis but not with S. mutans. In short, our heterotropic expansion protocol indeed captured differences in the adhesion behavior of different bacterial species in our model system at the single-cell level. Evaluation of F. nucleatum RadD-dependent and -independent Adhesion to Different Streptococcus Species. The ability of F. nucleatum to adhere to other bacterial members of dental plaque has been attributed to a group of outer membrane adhesion proteins (41). Among these, RadD has been identified as the primary mediator of the interspecies adherence between F. nucleatum and various Streptococcus spp., including S. mutans and S. sanguinis (42, 43). Here, we evaluate the adhesion of wild-type F. nucleatum and F. nucleatum ΔradD which is depleted in RadD using heterotropic expansion microscopy. The three-species consortium containing F. nucleatum ΔradD, S. mutans, and S. sanguinis aggregated less in comparison to consortia with F. nucleatum wild type (Fig. 6A), confirming a role for RadD in coaggregation. Nonetheless, some aggregation occurred. The residual aggregation with the F. nucleatum ΔradD suggests a RadD-independent mode of coaggregation as well. This result is in accord with the literature that adhesin(s) in addition to RadD may also contribute to F. nucleatum–Streptococcus coaggregation (41). We then analyzed whether microbe–microbe interaction inside these non-RadD aggregates could be differentiated from those in the wild-type consortia. Individual cells were clearly distinguished after expansion inside these aggregates (Fig. 6B), which was further confirmed by images of orthogonal views. High magnification merged images of F. nucleatum ΔradD and the two Streptococcus species presented a similar association pattern as that of F. nucleatum wild type: A larger number of S. sanguinis cells overlapped spatially with F. nucleatum compared to S. mutans (Fig. 6 C and D). Proximity analysis (Fig. 6 E and F) showed that S. sanguinis demonstrated a close association with F. nucleatum ΔradD while the relationship between S. mutans and F. nucleatum ΔradD was repulsive at short distances. Expansion microscopy differentiates the interaction between F. nucleatum ΔradD with two Streptococcus species. (A). Coaggregation assay comparison between F. nucleatum wild type and F. nucleatum ΔradD with two different Streptococcus species. Images obtained by phase microscopy; insert shows the aggregation in test tube with heatmap to help visualize flocculant precipitate. (B). Spectral imaging (after linear unmixing) of F. nucleatum ΔradD–S. mutans–S. sanguinis aggregates after expansion under a 20× objective (NA = 0.8). (Scale bar, 50 µm.) (C and D). High-magnification views (C) and single-cell image callouts (D) of merged view of F. nucleatum ΔradD with S. mutans and S. sanguinis. (E and F). Proximity analysis of image in (C) between F. nucleatum ΔradD and different Streptococcus species aggregates after expansion. A pair correlation value of 1 is highlighted by orange dashed lines in the figure panels. Pairwise correlations were calculated using 500,000 random dipoles per field of view. Color annotation: F. nucleatum ΔradD (white), S. mutans (red), and S. sanguinis (green). To strengthen a quantitative comparison of the spatial arrangement within the aggregates from the three-species consortium containing F. nucleatum wild type or ΔradD in terms of the interaction with S. mutans and S. sanguinis, we randomly chose eight different fields of view to conduct the proximity analysis and then compared them statistically. As shown in SI Appendix, Fig. S3A, there was no apparent difference in the pairwise correlation values between F. nucleatum wild type/S. mutans and F. nucleatum ΔradD/S. mutans. Similarly, the pairwise correlation values between F. nucleatum ΔradD and S. sanguinis was almost the same as that of F. nucleatum wild type with S. sanguinis (SI Appendix, Fig. S3B) while the spatial relationship between S. mutans and S. sanguinis was essentially random (SI Appendix, Fig. S3C). Each of the three taxa, F. nucleatum, S. mutans, and S. sanguinis, showed self-correlation at short distances indicating a tendency to self-associate (SI Appendix, Fig. S3 D and F). Taken together, these data suggest that a non-RadD adhesin or adhesins contribute to the interaction between F. nucleatum and Streptococci in addition to the RadD adhesin. Furthermore, the non-RadD adhesin-mediated interaction between F. nucleatum and S. sanguinis was of sufficient strength that it was not disrupted by the expansion force. To understand whether the repulsive behavior between S. mutans and F. nucleatum (both wild type and radD mutant) at short distances in three-species consortium is due to the competitive association between S. sanguinis and F. nucleatum, we carried out coaggregation experiments and expansion analysis between F. nucleatum and individual Streptococcus species under the same experimental conditions. Proximity analysis (SI Appendix, Fig. S4) showed that both F. nucleatum wild type and F. nucleatum ΔradD displayed close association with S. sanguinis at short distances. In contrast, the association with S. mutans appeared to be repulsive at shorter distances and random at greater distances. Collectively, this evidence suggests that the close association between F. nucleatum and S. sanguinis is intrinsic, and the apparent repulsive behavior between F. nucleatum and S. mutans is not from competition with S. sanguinis. Expansion Microscopy Analysis of a Microbial Episymbiosis Model. Epibionts, by definition, are organisms that live on the surface of another organism and, therefore, have a close structural association. We reasoned that the association of a bacterial epibiont with its host cell might be particularly strong. As an additional illustration of the application of heterotropic expansion microscopy, we examined a two-member oral episymbiotic system in which the epibiont was ultrasmall Saccharibacteria N. lyticus strain TM7x and its host was a Schaalia odontolyticus strain XH001 (31, 44). Because TM7x is only 200 to 300 nm diameter, whereas its host is more than an order of magnitude larger, the two species can be readily distinguished morphologically (SI Appendix, Fig. S5). To query whether the expansion force could separate TM7x away from its host cell XH001, we applied the expansion procedure to cocultures vitally stained with SYTO-9 (Materials and Methods). Low magnification images of the coculture (Fig. 7A) show aggregates of cells in which individual XH001–TM7x complexes were hard to discern. After expansion, XH001–TM7x complexes were pulled away from each other but many TM7x cells remained closely attached to their host cells (Fig. 7B) suggesting that the interaction force between TM7x and its host cell is stronger than the expansion force. High-magnitude images confirmed that expansion did not detectably alter the physical association between TM7x and its host cell (Fig. 7B, Inserts). Expansion force does not pull apart the epibiont interaction pair: Saccharibacteria N. lyticus strain TM7x and S. odontolytica (XH001). Confocal fluorescence imaging of SYTO 9-labeled coculture XH001–TM7x complexes before (A) and after expansion (B). Two regions of interest (ROI), dashed white boxes, are shown at high magnification (Inserts). TM7x cells are highlighted by white arrows. Discussion Our study demonstrates the utility of an unconventional application of expansion microscopy to the analysis of microbial biofilms. We show how heterotropic expansion can “decrowd” dense biofilm communities, improve visualization of individual bacterial members, and enable differentiation of microbe–microbe adhesive interactions within microbial communities at the single-cell level. The application is unconventional in that its purpose is to exploit the force driving expansion to assess microbial adhesion rather than to achieve superresolution imaging. In conventional expansion microscopy, the biological matter is digested to enable uniform, isotropic expansion of the polymer gel. In contrast, in our application, the microbes remain intact while being entrapped in the polymer network. The space between the microbes is expanded while the space within the microbes is not, hence the term heterotropic expansion. Expansion of the swellable polymer gel generates a pulling force on the entrapped microbes. The result is a “tug-of-war” competition between the expansion force and microbial adhesive forces. Weakly adherent cells will be pulled apart while strongly adherent cells will remain together. Thus, heterotropic expansion enables the differentiation of microbe–microbe interaction within microbial communities at the single-cell level. Our application of expansion microscopy is complementary to the conventional coaggregation assay, which assesses the binding of pairs of bacteria to each other by whether they form visible flocculant precipitates. Heterotropic expansion microscopy deconstructs the coaggregation phenotype at the single-cell level. Specifically, through investigating the aggregates formed by the model community, F. nucleatum–S. sanguinis–S. mutans, we found that the spatial interaction between F. nucleatum and S. sanguinis was significantly different from that of F. nucleatum and S. mutans, a feature otherwise not evident in the coaggregation assays. The F. nucleatum–S. sanguinis adhesion was evidently stronger than the expansion force since cells remained in close proximity and many were juxtaposed upon each other even after expansion. In contrast, the aggregates formed between F. nucleatum and S. mutans were pulled apart by expansion, indicating that the F. nucleatum–S. mutans adhesion was weaker than the expansion force. The epibiont association between TM7 and S. odontolytica is another example of a tight association as the expansion process failed to pull apart the interacting partners. This result was not surprising since the association between an epibiont and its host was expected to be strong. Studies have shown that type IV pili trigger a close association between Saccharibacteria and its host bacteria (45). It has also been reported that pili help Neisseria meningitidis adhere to the endothelium and subsequently form intravascular aggregates (46). Another example is the interaction between the predatory bacteria Bdellovibrio and Staphylococcus aureus (47) which are known to form an epibiotic spatial association. Our approach might offer unique insights into the single-cell adhesion forces within these aggregates. RadD, a major outer membrane protein of F. nucleatum, has been reported to play a key role in the binding phenotype between F. nucleatum and Streptococci including S. mutans (48) and S. sanguinis (41) and, consequently, in the integration of F. nucleatum into the supragingival microbial community (42). We investigated whether RadD was essential for the strong adhesive interaction between F. nucleatum and S. sanguinis by testing a RadD deletion mutant of F. nucleatum. Although the three-species community, F. nucleatum ΔradD–S. sanguinis–S. mutans, showed a reduced level of coaggregation, expansion microscopy revealed that binding between F. nucleatum ΔradD and S. sanguinis remained tight. This result indicated that RadD was not essential for tight binding at the single-cell level and that a non-RadD adhesin or adhesins had sufficient binding force to underlie the tight binding observed. By either mechanism, the strong association between F. nucleatum and S. sanguinis supports the premise that this interspecies interaction serves as an important association in the organization of supra- and subgingival plaque (49, 50). Although RadD was not essential for the tight binding between F. nucleatum and S. sanguinis, deletion of RadD did lead to significant reduction in the overall amount of coaggregation. How can these apparently conflicting results be explained? One possible explanation arises from the likely multivalency of binding interactions. Each binding partner, F. nucleatum or S. sanguinis, bears multiple copies of adhesin ligands or receptors. The large aggregates and flocculant precipitate resulting from mixture of the two partners may be understood in a manner similar to an antibody–antigen precipitation reaction. Following the antibody–antigen paradigm, we varied the concentration ratio between F. nucleatum and S. sanguinis in the coaggregation experiment and found that both F. nucleatum ΔradD and F. nucleatum wild type had similar coaggregation profiles (SI Appendix, Fig. S6), peaking at a ratio (F. nucleatum/S. sanguinis) of one. However, the amount of coaggregation with F. nucleatum ΔradD was approximately half of that for F. nucleatum wild type. Coaggregation at a F. nucleatum ΔradD/S. sanguinis ratio of 1.0 was essentially the same as that of F. nucleatum wild type at a F. nucleatum/S. sanguinis ratio of 0.5. From this result, we infer that the reduced coaggregation observed after deletion of RadD may be a consequence of reduction in the number of adhesins per F. nucleatum and, therefore, its effective valency ratio to S. sanguinis. In sum, the reduced level of aggregates seen by coaggregation is fully consistent with the tight binding between F. nucleatum ΔradD and S. sanguinis as assayed at the single-cell level. Electrorepulsion among the negatively charged carboxylic groups in the swellable gel generates the expansion force according to Coulomb’s law. However, the net expansion force will become attenuated gradually as it is opposed by the elasticity of the polyelectrolyte backbone, resulting in a balance of force and a stable extent of expansion. It is noteworthy that the initial expansion force is adjustable depending on the concentration of monomer and cross-linker molecules during the gelation step (51). Therefore, we envision that gels with a series of expansion forces can be prepared by carefully choosing the gelation recipes. In this way, we can better differentiate microbial adhesive forces within microbial communities by varying the expansion force. Our approach for differentiating adhesive interactions within microbial consortia at the single-cell level has some limitations. First, microbial consortia often involve multiple bacterial species, where multiplexed fluorescence imaging is needed. Our study investigated adhesive interactions in a model microbial community consisting of three members. To investigate the feasibility of this single-cell functional platform within complex microbial communities, it will be necessary to evaluate the compatibility between heterotropic expansion microscopy and multiplexed fluorescence imaging. Second, our approach thus far achieved the differentiation of interspecies microbial adhesive strengths on a relative basis. However, the absolute binding strengths between microbes with their neighbors remain to be determined. For the quantification of microbial binding forces, a calibration method will be needed to quantify the gel expansion force. Finally, we need to take into consideration whether the expansion process affects the viability of microbes. Notwithstanding these limitations, our approach establishes a proof of principle for differentiating adhesive interactions within microbial consortia at the single-cell level. Materials and Methods Materials. Acryloyl-X SE (Invitrogen, A20770), Triton X-100 (Sigma-Aldrich, 9036-19-5), DMSO (Sigma-Aldrich, 67-68-5), EDTA, disodium (0.5 M, pH=8, Invitrogen, AM9260G), Tris HCl (1 M, pH=8, ThermoFisher Scientific, 15568025), NaCl (Fisher Scientific, AC447302500), proteinase K (1,500 U/ml, ThermoFisher Scientific, 25530049), UltraPure™ TEMED (ThermoFisher Scientific, 15524010), ammonium persulfate (APS, ThermoFisher Scientific, 17874), sodium acrylate (Sigma-Aldrich, 408220-25G), calcium chloride (Sigma-Aldrich, 10043-52-4), magnesium chloride (Sigma-Aldrich, 7786-30-3), sodium azide (Sigma-Aldrich, 26628-22-8), glycine (ThermoFisher Scientific, 36435.A1), sucrose (Fisher Scientific, AA3650830), acrylamide (40% (wt/vol), Fisher BioReagents, BP1402-1), N,N’-methylenebisacrylamide (Sigma-Aldrich, 110-26-9), Hi-Di™ formamide (Fisher Scientific, 44-407-53), phosphate-buffered saline, PBS (10× stock, Fisher BioReagents, BP3991), Poly-L-lysine solution (Sigma-Aldrich, 25988-63-0), 2-µm fluorescent beads (fluorescent blue, Sigma-Aldrich, L0280; fluorescent red, Sigma-Aldrich, L3030), Dextran, Alexa Fluor™ 647 (ThermoFisher Scientific, D22914), SYTO™ 9 (ThermoFisher Scientific, S34854), FM™ 1-43 (ThermoFisher Scientific, T3163), high-grid glass bottom dish (ibidi, 81148), and glass bottom dish (Fisher Scientific, NC0662883). Bacteria and Culture. Monospecies S. mutans biofilms. Planktonic S. mutans UA159 was cultured in Brain Heart Infusion (BHI) broth under aerobic (21% O2) conditions. To prepare S. mutans biofilms, S. mutans (OD600 = 0.05) was mixed with 20 µl of fluorescent blue beads into fresh BHI (supplemented with 0.1% sucrose) and added to saliva-precoated glass bottom dish. Robust biofilms were formed after overnight aerobic incubation. Dual-species Streptococcus spp. biofilms. GFP-expressing S. sanguinis and mCherry-expressing S. mutans were cultured in BHI broth under aerobic (21% O2) conditions to allow maturation of GFP and mCherry. Then, S. mutans (OD600 = 0.05) and S. sanguinis (OD600 = 0.05) were mixed with 20 µl of fluorescent blue beads into fresh BHI (supplemented with 0.1% and 1% sucrose, respectively) and added to saliva-precoated glass bottom dishes. Robust dual-species biofilms were formed after overnight aerobic cultivation. F. nucleatum–S. mutans–S. sanguinis coaggregation assay. F. nucleatum ssp. nucleatum strain ATCC 23726 was cultured in Columbia Broth (Fisher Scientific, DF0944-17-0) under anaerobic conditions (5% H2, 5% CO2, 90% N2) at 37° C. Then, F. nucleatum 23726 (F. nucleatum wild type) was stained with a membrane-labeling fluorophore, FM™ 1 to 43 for 30 min. Then, stained F. nucleatum were washed with 1 × PBS before combining with the two Streptococcus species (mCherry-expressing S. mutans and GFP-expressing S. sanguinis) in the coaggregation buffer (see recipe). Three bacterial species were then mixed together thoroughly at an OD600 of 2 for each bacterial species and allowed to stand in the dark. Flocculant precipitates appeared within 30 min after mixing. Precipitated aggregates were removed three hours after gently mixing and placed onto a glass bottom dish, semi-air-dried, followed by gelation, imaging before expansion, expansion by solvent exchange with water, and imaging after expansion. Solutions. Coaggregation buffer. Coaggregation buffer was prepared according to ref. 52. NaCl (150 mM), Tris⋅HCl (1 mM, pH=8.0), CaCl2 (0.1 mM), NaN3 [0.02% (w/v)], and MgCl2 (0.1 mM) were mixed in sterile water. Expansion-related solutions. Expansion-related solutions were prepared as reported (29). Acryloyl-X SE (AcX) stock solution. Five milligrams of acryloyl-X SE was dissolved in 500 µl of anhydrous DMSO. Aliquots were stored at −20° C. TEMED and APS stock solutions. Both chemicals were dissolved in water at a concentration of 10 g/100 ml and then divided into aliquots of 1 ml each and stored at −20° C. Monomer solution (Stock X). Sodium acrylate (9% w/v), acrylamide (2.7% v/v), N,N’-methylenebisacrylamide (0.16% w/v), and NaCl (12.4% w/v), in 1 × PBS. The Stock X solution was then divided into aliquots of 940 µl each and stored at −20° C. Gelation solution. Stock X solution plus TEMED (0.2% w/v) and APS (0.2% w/v). Digestion buffer. 0.5% w/v Triton X-100, 1 mM EDTA, 50 mM Tris HCl, NaCl (4.67% w/v), and proteinase K (15 U/ml). Digestion buffer was also divided into aliquots of 1 ml each and stored at −20° C; proteinase K was added to the remaining digestion buffer immediately before the digestion step. Gelation, Digestion, and Expansion on Biofilms. S. mutans UA159 biofilms and dual-species Streptococcus biofilms were fixed in 10% formalin (Sigma-Aldrich, HT501128-4L) for 1 h; then, glycine (100 mM) was added to quench the fixation for 5 min. After that, fixed biofilm samples were washed with 1 × PBS twice before the gelation step. S. mutans UA159 biofilms were then stained with SYTO9 for the visualization of S. mutans inside the biofilms. After incubation with AcX, biofilms were washed with PBS twice, each time for 15 min. Then, monomer solution (see recipe) was added to biofilms, and gelation was allowed to occur for 1 h at 37° C. Immediately afterward, digestion was carried out in the presence of proteinase K (10 U/ml, see recipe). After overnight digestion, expansion was induced by adding water to the digested samples, and exchange with water for three times, each time for 20 min. Part of the expanded samples were then carefully mounted to a poly-L-lysine-coated cover glass (VWR, 16004-348), and a few drops of water were added to the top of the gel before image acquisition to prevent sample dehydration. Gelation and Expansion of F. nucleatum Wild Type/F. nucleatum ΔradD–S. mutans–S. sanguinis Aggregates. Gelation in the presence of the monomer solution (Stock X) was immediately applied to the semi-dried three-member aggregates at 37° C for 1 h in a humidified chamber. Spectral imaging was then employed after gelation to acquire images before expansion. After imaging, water was directly added to the gels, and gel samples were exchanged with water three times, each time for 20 min. Part of expanded samples were then carefully mounted to a poly-L-lysine coated cover glass (VWR, 16004-348), and a few drops of water were added to the top of the gel before image acquisition to prevent sample dehydration over the process of image acquisition. Gelation and expansion of epibiont interaction pair Saccharibacteria N. lyticus strain TM7x/S. odontolytica (XH001). Preparation of epibiont pair XH001/TM7x was described in the previously published studies (31, 53). Coculture was then harvested by passaging for two times and washed with 1 × PBS. Cells were stained with SYTO 9 afterward and washed with 1 × PBS. SYTO 9-labeled XH001/TM7x were then dropped on to the glass bottom dish; gelation happened on top of the coculture aggregates. Imaging before expansion was acquired after gelation step, then water-induced expansion, and imaging after expansion. Spectral Imaging and Linear Unmixing. Spectral images were acquired using a Carl Zeiss LSM 780 confocal microscope with a Plan-Apochromat 20×/0.8 N.A. dry objective or 63×/1.4 N.A. oil immersion objective. Images were obtained using simultaneous excitation with 405, 488, and 561 nm laser lines for both dual-species Streptococcus biofilms and three-member F. nucleatum wild type/F. nucleatum ΔradD–S. mutans–S. sanguinis coaggregation assay. Spectral images were harvested at a bin size of 8.9 nm. Individual reference emission profiles were obtained under the same conditions. Raw spectral images were smoothed by a median filtering algorithm with a kernel pixel size of 3 through ZEN Black software (Carl Zeiss). Linear unmixing was then performed using ZEN Black software (Carl Zeiss) using reference spectra. Unmixed images were assembled and pseudocolored using Fiji software (54). Supplementary Material Appendix 01 (PDF)