Visualization of mRNA Expression in Pseudomonas aeruginosa Aggregates Reveals Spatial Patterns of Fermentative and Denitrifying Metabolism

ABSTRACT

Gaining insight into the behavior of bacteria at the single-cell level is important given that heterogeneous microenvironments strongly influence microbial physiology. The hybridization chain reaction (HCR) is a technique that provides in situ molecular signal amplification, enabling simultaneous mapping of multiple target RNAs at small spatial scales. To refine this method for biofilm applications, we designed and validated new probes to visualize the expression of key catabolic genes in Pseudomonas aeruginosa aggregates. In addition to using existing probes for the dissimilatory nitrate reductase (narG), we developed probes for a terminal oxidase (ccoN1), nitrite reductase (nirS), nitrous oxide reductase (nosZ), and acetate kinase (ackA). These probes can be used to determine gene expression levels across heterogeneous populations such as biofilms. Using these probes, we quantified gene expression across oxygen gradients in aggregate populations grown using the agar block biofilm assay (ABBA). We observed distinct patterns of catabolic gene expression, with upregulation occurring in particular ABBA regions both within individual aggregates and over the aggregate population. Aerobic respiration (ccoN1) showed peak expression under oxic conditions, whereas fermentation (ackA) showed peak expression in the anoxic cores of high metabolic activity aggregates near the air-agar interface. Denitrification genes narG, nirS, and nosZ showed peak expression in hypoxic and anoxic regions, although nirS expression remained at peak levels deeper into anoxic environments than other denitrification genes. These results reveal that the microenvironment correlates with catabolic gene expression in aggregates, and they demonstrate the utility of HCR in unveiling cellular activities at the microscale level in heterogeneous populations.

IMPORTANCE To understand bacteria in diverse contexts, we must understand the variations in behaviors and metabolisms they express spatiotemporally. Populations of bacteria are known to be heterogeneous, but the ways this variation manifests can be challenging to characterize due to technical limitations. By focusing on energy conservation, we demonstrate that HCR v3.0 can visualize nuances in gene expression, allowing us to understand how metabolism in Pseudomonas aeruginosa biofilms responds to microenvironmental variation at high spatial resolution. We validated probes for four catabolic genes, including a constitutively expressed oxidase, acetate kinase, nitrite reductase, and nitrous oxide reductase. We showed that the genes for different modes of metabolism are expressed in overlapping but distinct subpopulations according to oxygen concentrations in a predictable fashion. The spatial transcriptomic technique described here has the potential to be used to map microbial activities across diverse environments.

INTRODUCTION

Even in an isogenic population, bacteria can display a range of different phenotypes (1, 2). This phenotypic variation can form along microenvironmental gradients (3) but has also been observed in well-mixed cultures (4, 5). Heterogeneity in gene expression allows cells to fit different ecological niches and survive in ecosystems with fluctuating conditions, increasing the overall fitness of the population (4, 6, 7). Phenotypic heterogeneity is relevant in a variety of contexts, from rhizosphere colonization (8) to metabolic engineering for optimal biosynthesis of fuels and medicines (9). In chronic infections, phenotypic variation shields pathogens from antibiotics, allowing subpopulations with distinct metabolic activities to survive rounds of treatment (10, 11) and contribute to the development of antibiotic resistance (12).

While many technologies exist to study gene expression at the population (13) and single-cell (14) levels, few allow the mRNA output of individual bacteria to be observed while maintaining the microenvironmental context and without genetically modifying the cells (e.g., use of genetically encoded fluorescent proteins). Recently, a spatial transcriptomic method was introduced that enables visualization of hundreds of transcripts at a time in single cells (15), yet this method requires specialized instrumentation and analysis software, providing a barrier to widespread usage. In contrast, the hybridization chain reaction (HCR) is a more accessible technology that permits single-cell transcript visualization of a smaller number of genes using standard instrumentation (16, 17). HCR is a form of fluorescence in situ hybridization (FISH), where the binding of fluorescent probes to a target RNA of interest results in the binding of additional fluorescent probes, thus amplifying the signal (16). The technology first was developed in eukaryotes (18) but has been used in bacteria to image rRNA for the purposes of species identification and growth rate assessment in mixed communities (19, 20), as well as to visualize gene expression in heterogeneous populations (21).

Building on these studies, we reasoned that HCR was well suited to enabling a systematic exploration of the variation in catabolic activities within aggregate biofilms of the model biofilm-forming bacterium Pseudomonas aeruginosa. One environmental parameter that has profound effects on the metabolic state of bacteria is oxygen. For example, many infection sites have large regions of hypoxia, including those found in skin wounds, lungs, and the gut (22, 23). Additionally, opportunistic pathogens like P. aeruginosa can form aggregate biofilms during infection (20, 24, 25), and biofilms produce their own oxygen gradients, with biofilm-interior populations experiencing oxygen limitation (26). This has devastating consequences for human health because commonly used antibiotics such as tobramycin and ciprofloxacin are ineffective at killing oxygen-limited P. aeruginosa at the center of biofilms, while drugs that target anaerobic metabolism specifically kill oxygen-limited populations (27, 28). Accordingly, antibiotic treatments must take the microenvironment and the phenotypic heterogeneity experienced by opportunistic pathogens in the host into account to be successful (29, 30). Before we can rationally design better treatments, we must improve our understanding of the range of metabolic states that exist in biofilm populations spatiotemporally (31).

In the absence of oxygen, P. aeruginosa cells utilize different strategies to conserve energy, ranging from oxidative to substrate-level phosphorylation (32). Because chronic infections are often oxidant limited, how P. aeruginosa responds to electron acceptor availability is expected to lead to metabolic differentiation within these environments, which partly underpins its antibiotic tolerance and resistance (33). Under anoxic conditions, P. aeruginosa can use nitrate and other N-oxides as alternative terminal electron acceptors to oxygen by performing denitrification (34). The nitrate levels in chronic surgical wounds and cystic fibrosis lung infections are high enough to support denitrification in some regions, while the oxygen levels are low enough to induce denitrification (22, 35, 36). In the absence of nitrate, P. aeruginosa may survive under anoxic conditions for long periods by performing pyruvate fermentation, in which pyruvate is converted to lactate and acetate (37), or by fermenting arginine (38) or using phenazines to facilitate glucose fermentation (39). Aerobic, denitrifying, and fermentative metabolisms are all relevant to bacterial populations in chronic infection environments.

If we are to use HCR to map these metabolisms within biofilm aggregates, it is important to interpret any gene expression patterns we might see within an ecophysiological regulatory framework. In P. aeruginosa, the primary respiratory oxidase under highly aerobic conditions is cbb₃-1, encoded by cco1 (40). cco1 is constitutively expressed when oxygen is replete but slightly downregulated in the stationary phase (41, 42). Denitrification is catalyzed by a series of reductases, in particular, nar, nir, nor, and nos (34, 42). The main regulatory proteins that govern the expression of these genes are Anr, Dnr, NarXL, and NirQ (Fig. 1). Under anoxic conditions, Anr activates Dnr, a regulatory protein that is downregulated by phosphate availability, which activates NirQ. Where nitrate is available and oxygen is limited, Anr additionally activates NarXL, which, in turn, amplifies the activation of Dnr and NirQ. The nitrate reductase gene, nar, is upregulated by NarXL, and this upregulation is amplified by Dnr in the presence of nitric oxide (34). The nitrite reductase gene, nir, is highly upregulated under anoxic conditions and has higher differential expression from oxic to anoxic conditions than nar (22, 34, 41, 42). nir is slightly upregulated by Anr, causing a small amount of nitrite to be converted to nitric oxide, which triggers Dnr and increases the upregulation of nir in a feedback loop (43). NirQ and Dnr tightly coregulate nir with the nitric oxide reductase nor in response to nitric oxide availability, preventing nitric oxide from accumulating (34). The nitrous oxide reductase, nos, is upregulated by Dnr in response to nitric oxide availability (34). In contrast, the acetate kinase gene, ackA, responsible for ATP synthesis during pyruvate- and phenazine-mediated glucose fermentation, is induced by the regulatory proteins Anr and integration host factor (IHF) (37, 42).

FIG 1.

Regulatory diagram of target genes. This diagram was compiled from a literature search of transcriptomics reviews in P. aeruginosa (34, 37, 42).

We targeted the expression levels of cco1, narG, nirS, nosZ, and ackA mRNA molecules to investigate the spatiometabolic patterning of catabolic gene expression within and across P. aeruginosa aggregate biofilm populations. Our goal was to use a finite set of genes to identify the range of metabolic states adopted by multicellular aggregates under oxidant-limited conditions. Here, we show that HCR 3.0 not only enables spatiometabolic mapping in situ but that it allows us to infer the metabolic state and environmental conditions that single cells within these heterogeneous populations are experiencing.

RESULTS

HCR probes are specific for their mRNA targets.

To validate the probes designed for this study, we tested their specificity using wild-type (WT) and mutant strains in which the target gene had been cleanly deleted. To test narG and nirS probes, we grew WT, ΔnarGHJI, and ΔnirS strains aerobically in LB supplemented with 40 mM nitrate to late exponential phase to achieve oxygen limitation and activate Anr. To test ccoN1 probes, we grew WT and ΔccoN1 strains aerobically in the same medium to early exponential phase to achieve highly oxygenated conditions. To test nosZ probes, we grew WT and ΔnosZ strains anaerobically in the same medium to mid-exponential phase to achieve denitrifying metabolism. We were unable to observe ackA signal in any of the liquid culture conditions we tested, so, to test ackA probes, WT and ΔackA were grown in the agar biofilm block assay (ABBA) for 12 h in LB medium and 0.5% agar, supplemented with 40 mM nitrate. Briefly, in ABBA, P. aeruginosa cells are suspended in an agar medium that is allowed to solidify. Cells then grow as aggregate biofilms suspended in this agar matrix, and nutrient gradients are known to develop within individual aggregates as well as across the population of aggregates (see below).

The narG, nirS, nosZ, and ccoN1 probes showed an average mean intensity per cell that was an order of magnitude higher in the WT than the deletion strains, indicating that our probe sets are specific for each target mRNA (Fig. 2A and C). Additionally, the ackA probe showed significantly higher average mean intensity per ABBA aggregate than the deletion strain (Fig. 2B and D). In some of the deletion mutant controls, particularly for the nosZ and ccoN1 probe sets, we occasionally observed fluorescent puncta that did not overlap cells and appeared qualitatively different from the signal observed in the positive control. We therefore assume that these puncta were due to nonspecific binding to the slide and not biologically significant.

FIG 2.

mRNA probes are target specific. (A) Micrographs of single-cell deletion validation controls for mRNA probes. (B) Micrographs of aggregate deletion validation controls for mRNA probes. Images reflect the region 50 to 100 μm from the air-agar interface. (C) Quantification of mean mRNA probe intensities for single cells grown in liquid culture of wild-type and deletion mutants for target genes under upregulating growth conditions. Mean intensity was 10-fold higher in the wild type than in the deletion for all probe sets. Each boxplot summarizes approximately 10 images per replicate. Three replicates were performed per condition, and each colored box represents a different biological replicate. Whiskers represent 1.5 times the interquartile range, while diamonds on the boxplots represent outliers. (D) Quantification of mean mRNA probe intensities for aggregates grown in agar blocks wild-type and deletion mutants for ackA under upregulating growth conditions. Mean intensity was 2-fold higher in the wild type, and each boxplot summarizes 3 to 5 images per replicate. Three replicates were performed per condition.

Metabolite gradients are readily established in ABBA.

To explore the phenotypic heterogeneity of P. aeruginosa wild-type PA14, we used HCR to visualize the localization of narG, nirS, nosZ, ccoN1, and ackA expression across biofilm aggregates grown in the ABBA. We used a thresholded 16S rRNA intensity value to define aggregate boundaries and to control for baseline metabolic activity, and we quantified mRNA and rRNA signals within this mask. P. aeruginosa aggregates were grown in LB medium with 0.5% agar, supplemented with 40 mM nitrate. We chose these conditions to capture the effects of oxygen gradients on the expression of our genes of interest. Physiological levels of nitrate in infected cystic fibrosis lung samples and chronic wounds are ~400 μM (22, 44), but we chose a higher concentration of nitrate to ensure that it would still be available after 12 h of growth.

In an ABBA, an oxygen gradient develops, and the oxygen minimum migrates upwards toward the air-agar interface over the course of the incubation as cell density increases, and thus, oxygen availability decreases (Fig. 3A). The exact depths at which oxygen is minimal vary between agar blocks, but the pattern is conserved. A steep oxygen gradient occurs across the first ~100 μm directly below the surface, and regions deeper than ~250 μm are uniformly oxygen starved. Accordingly, deeper aggregates in the anoxic zone will have experienced anoxia for longer than the aggregates above them. As aggregates begin experiencing hypoxia, cells will upregulate denitrification genes and reduce nitrate such that a similar nitrate gradient is also expected to develop. The deepest aggregates are likely the first to begin consuming nitrate in their local environment, so deeper aggregates likely have less access to nitrate at this time point. As the denitrification pathway proceeds, nitric oxide is generated (45), so a greater amount of nitric oxide is expected at lower depths.

FIG 3.

Metabolic genes show distinct patterns across three-dimensional oxygen gradients. (A) Mean oxygen levels of ABBA samples grown with LB plus 40 mM nitrate for 12 h. The dark line is the mean, while the shading represents the standard deviation of 8 biological replicates grown under identical conditions. (B) Three-dimensional micrographs of probe signal in LB plus 40-mM nitrate ABBAs. Each image represents a 50-μm slice of agar, compiled from 8 individual z-slices with an interslice distance of 6.24 μm, viewed from the top of the block, with each sequential image from the top of the figure representing the section directly below the slice above it. The rRNA signal is colored cyan, while the mRNA signal is colored magenta. (C) Mean mRNA channel intensity per aggregate (x axis) plotted by depth. Each plot represents four images of one replicate each of an experimental and control condition, and each point represents one aggregate. Filled points represent the experimental condition, while the open circles represent the background autofluorescence intensity of aggregates imaged in the mRNA channel in a control condition where only rRNA probes were used.

Assuming these conditions and given the regulatory architecture controlling expression of these genes (Fig. 1), we predicted that narG would be expressed in regions with high nitrate and low oxygen concentrations, nirS would be expressed in regions with high nitrate and/or nitric oxide and low oxygen concentrations, and nosZ would be expressed in regions with nitric oxide present and low oxygen concentrations. We expected ccoN1 to be expressed most highly in regions with high oxygen and ackA to be expressed in regions that are limited for both oxygen and nitrate.

Metabolic genes show distinct expression patterns within ABBAs.

The oxidase ccoN1 was maximally expressed in the oxic region (0 to 50 μm from the air-agar interface at the top of the ABBA blocks) and steeply decreased in expression deeper into the block along the oxygen gradient, which is the same pattern shown by the 16S rRNA signal (Fig. 3B). In the larger aggregates closer to the surface, the signal for ccoN1 and 16S rRNA was higher along the aggregate exterior than in the centers, while the signal for the ackA gene was highly expressed in the interior (Fig. 3B and C; see Fig. S1 in the supplemental material). An elevated signal for the denitrification genes was not detected in this region.

In the region below the oxic portion of our experimental system (>50 μm from the air-agar interface at the top of the ABBA blocks), rRNA, ccoN1, and ackA expression was reduced, and expression of the denitrification genes was elevated, correlating with low oxygen concentrations that decreased with depth across this region (Fig. 3A to C) All three denitrification genes were expressed at their maximal levels in aggregates in this region. The nitrate reductase narG was upregulated under hypoxic conditions in the 40-mM nitrate samples, peaking in average mean intensity per aggregate at a depth of 100 to 150 μm from the air-agar interface (Fig. 3B and C). narG signal was restricted to the interior of aggregates, where oxygen availability is expected to be reduced compared to cells on the aggregate exterior (Fig. 3B; Fig. S1). The nitrite reductase, nirS, was upregulated under hypoxic and anoxic conditions, peaking between 100 and 200 μm from the surface. The nitrous oxide reductase nosZ was expressed in cells in the cores of some aggregates in the hypoxic and anoxic regions but was not uniformly expressed by aggregates at any depth (Fig. 3C). Oxygen levels reached their minimal levels between 150 and 250 μm from the air-agar interface (Fig. 3A).

Oxygen levels were uniformly low below 250 μm from the air-agar interface. All genes measured showed reduced expression with depth across this anoxic region, likely due to oxidant limitation, although nirS expression was elevated in the cores of aggregates at all depths. nirS signal was restricted to the interior of aggregates in the hypoxic zone but was expressed across entire aggregates further into the anoxic zone (Fig. 3B; Fig. S1). An appreciable narG signal was not detected in the deepest anoxic region (Fig. 3C). Data from two additional replicates of this experiment (mRNA and rRNA ABBA images with quantification) can be found in Fig. S2 to S7.

Differentially regulated metabolic genes show different expression patterns than rRNA.

By comparing the rRNA and mRNA signal per aggregate in each sample, we can detect where the mRNA signal diverges from the basal metabolic rate, reflected by the rRNA signal. The ccoN1 signal linearly and positively covaried with the rRNA signal across depths (Fig. 4), with a Pearson correlation coefficient (PCC) of 0.956. No other genes showed this pattern. All three denitrification genes showed maximal expression in regions where the rRNA signal was less than or equal to half the maximum, corresponding to the hypoxic transition region of the agar block. The PCC of narG signal with rRNA was 0.612, while for nosZ with rRNA, it was 0.678. The nirS signal showed the lowest covariance with the rRNA signal, with maximal expression occurring in aggregates with one-third the maximum rRNA signal and with a PCC of 0.080. The ackA signal showed positive, but nonlinear, covariance with the rRNA signal, showing the highest mRNA signal in the aggregates at the surface with the highest rRNA signal and linearly covarying throughout the oxic and hypoxic regions before a steep dropoff of mRNA signal in the anoxic region, with a PCC of 0.896. The different covariance patterns were most pronounced in the top 200 μm of the block (Fig. S1), as deeper regions had more uniformly low activity. The rRNA and ccoN1 signals were qualitatively highest at the aggregate edges, while the narG, nirS, and nosZ signals were highest in the aggregate cores (Fig. S1). This likely indicates a radial oxygen gradient from the edges to the cores of individual aggregates in addition to the vertical oxygen gradient measured across the agar block (Fig. 3A), as predicted based on models of oxygen diffusion in aggregate biofilms (46). The ackA signal was heterogeneously expressed in some cells throughout aggregates.

FIG 4.

Metabolic genes show distinct patterns compared to rRNA expression. Mean mRNA intensity per aggregate (x) plotted against mean rRNA intensity per aggregate (y) colored by depth. Each plot represents four z-stacks of one replicate, and each point represents one aggregate. Large points represent the experimental condition, and small points represent the control condition without mRNA probes. The Pearson correlation coefficient (PCC) of ccoN1 with rRNA was 0.956; for ackA with rRNA, it was 0.896; for narG with rRNA, it was 0.612; for nirS with rRNA, it was 0.080; and for nosZ with rRNA, it was 0.678.

DISCUSSION

In this work, we developed and validated a set of HCR probes targeting mRNA molecules for key denitrification, aerobic respiration, and fermentation genes in P. aeruginosa. This targeted probe set permits mapping of the metabolic states of cells in aggregate biofilms at high resolution across microenvironments characterized by steep oxygen gradients. Our results reveal distinct patterns of catabolic gene expression within these populations, underscoring that metabolic activity can be heterogeneous over small spatial scales. The demonstration that HCR can be applied to chart catabolism within a reductionist laboratory biofilm sets the stage for future studies of microbial responses to dynamic microenvironments in more complex contexts. Importantly, HCR permits the study of gene expression dynamics in complex WT populations without the need for generating reporter strains; though developed here for a model organism, this capability has obvious advantages for the study of strains that are not yet genetically tractable.

Our results show that the relationship between ccoN1 and rRNA signals for entire aggregates is linear and positive, indicating that changes in ccoN1 expression reflect changes in the metabolic rate at the population level (20). This is consistent with previous studies showing ccoN1 as constitutively expressed (47). Most genes in Gram-negative bacteria such as P. aeruginosa show basal levels of gene expression, even outside the upregulation condition (48), so the expression pattern shown by ccoN1 represents the null hypothesis of what expression should look like if the gene is not regulated by environmental conditions. If gene expression patterns diverge from the pattern shown by ccoN1 and rRNA, we can infer that they are upregulated and downregulated based on additional conditions other than the metabolic rate.

We observed that expression of the genes encoding the denitrification machinery in P. aeruginosa is spatially stratified across the ABBA system after 12 h. Previous studies have demonstrated that denitrification in liquid cultures occurs in successive steps over time (45). Recently, this sequential pattern was observed in structured communities of P. aeruginosa for narG and nirS using reporter strains (49), revealing that maximal expression of these genes moved outward from the center of particles suspended in bulk medium across time, suggesting aggregates progressively extinguished local oxygen, nitrate, and nitric oxide concentrations. Our findings confirm this spatial pattern in P. aeruginosa, showing that at a single time point, maximal narG expression occurs closer to the air-agar interface than maximal nirS and nosZ expression, as oxygen limitation and nitrate limitation increase with depth. Furthermore, we were able to visualize this spatial pattern in individual aggregates. In aggregates between 50 and 150 μm from the air-agar interface, narG expression occurs closer to the exterior of individual aggregates than nirS and nosZ expression, and nirS expression is present across entire aggregates at greater, mostly anoxic depths but limited to the cores of aggregates in hypoxic conditions.

Despite both being upregulated by Anr, fermentation and denitrification genes showed distinct expression patterns. The fermentation gene ackA was upregulated in the cores of the aggregates closest to the air-agar interface where oxygen concentrations are higher, while the denitrification genes were upregulated in the cores of the aggregates in the hypoxic regions. Both types of metabolism mark hypoxic/anoxic conditions, but they are regulated in response to distinct nitrate and nitric oxide conditions. Given that ackA is expressed in the absence of nitrate (37), the pattern of the denitrification and fermentation genes suggests that the cores of the aggregates at the top of the agar blocks may be conserving energy by fermentation due to a microenvironmental regulatory cue. Notably, the denitrification genes are directly upregulated by Dnr, which is activated by nitric oxide, while ackA is directly upregulated by Anr, which is deactivated by nitric oxide. We can infer from this regulatory relationship that nitric oxide signaling is occurring in the lower aggregates but not in the upper aggregates at this time point. Direct measurement of N-oxide concentrations in the ABBA system is important for understanding the relationship between microenvironmental conditions and metabolic state and will be the focus of future experiments.

Going forward, this set of probes may be expanded, and multiple high-signal probes may be used to determine correlation and anticorrelation more precisely. HCR may be used in two directions, read-out, in which specific regions of a sample are queried to learn which genes are coexpressed within them, and read-in, in which knowledge of genes whose coexpression is already understood is leveraged to infer something about the microenvironment of the region in which they are expressed (17). The former application allows us to choose a specific context in space and time and see which genes are being expressed without prior knowledge, whereas the latter approach allows us to apply our knowledge of gene regulation to infer something about the microenvironment, such as we did in this study to dissect the microenvironmental conditions of ABBA subpopulations. In the future, the probes developed in this study may be applied to observe metabolisms and infer potential microenvironmental conditions experienced by P. aeruginosa in clinical samples, opening the doors for greater insight into the metabolic processes sustaining cells in infections.

MATERIALS AND METHODS

Microbial strains and growth conditions.

P. aeruginosa strains UCBPP-PA14 (WT) and isogenic ΔnarGHJI (28), ΔccoN1 (47), ΔackA (39), ΔnirS, and ΔnosZ (50) were routinely grown in 5 mL lysogeny broth (LB) supplemented with 40 mM KNO₃ with shaking at 37°C. Aerobic cultures were incubated at 37°C with shaking at 250 rpm unless described otherwise. Anaerobic cultures were incubated in an anaerobic chamber at 37°C without shaking.

HCR v3.0 probe validation.

WT and deletion mutant cultures were grown overnight from freezer stocks in LB broth supplemented with 40 mM KNO₃. For the ackA mutant, cells were grown in the agarose biofilm block assay (ABBA) as described below. For other mutants, 20 μL overnight culture was inoculated into 5 mL of prewarmed and degassed media and incubated under conditions meant to induce upregulation of the target gene. For narG and nirS, cultures were grown aerobically as described above for 4 h to an approximate optical density at 500 nm (OD₅₀₀) of 1.1. For ccoN1, cultures were grown aerobically as described above for 2.5 h to an approximate optical density at 500 nm of 0.2. For nosZ, wild-type and deletion mutant cultures were grown anaerobically for 2.5 h to an approximate optical density at 500 nm of 0.2. We pelleted 1 mL of culture, and cells were fixed by resuspending in 1 mL of 4% paraformaldehyde in phosphate-buffered saline (PBS), pH 7.2, and incubating resuspensions at 25°C for 90 min. After fixing, cells were pelleted, washed with PBS, resuspended in 1 mL of a 1:1 ethanol/PBS solution, and stored at −20°C until HCR 3.0 experiments were performed.

Prior to performing HCR, cell suspensions were washed in PBS and treated with 1 mg/mL lysozyme (catalog number L6876; Sigma) in 10 mM Tris-HCl (pH 7.6) for 1 h at 37°C with shaking. After these treatments, cells were washed with PBS and resuspended in 1 mL of a 1:1 ethanol/PBS solution. Cell solutions were suspended in hybridization buffer for 1 h at 37°C. Probe solutions were prepared containing 5 nM appropriate HCR v3.0 initiator odd and even probe pairs in filter-sterilized hybridization buffer. We added 20 μL of probe solutions (Integrated DNA Technologies) to each cell solution. Pseudomonas aeruginosa rRNA and narG mRNA sequences were targeted using target sequences previously used in Jorth 2019. Probes specific to nirS, nosZ, ackA, and ccoN1 were designed in the study, and the target sequences are included in Table 1. Solutions were hybridized overnight at 37°C and then washed twice with 200 μL probe wash buffer at 37°C. Cells were then resuspended in 40 μL amplification buffer and preamplified for 30 min, while fluorescent hairpin probes (Molecular Technologies) were thawed, each denatured in separate tubes in a thermal cycler for 90 s at 95°C, and allowed to cool for at least 30 min at 25°C in the dark.

TABLE 1.

New HCR probe target sequences

Target	Sequence
nosZ 1	GAGCGACGACACGAAAAGCCCCCACGAAGAAACCCACGGCCTGAACCGCC
nosZ 2	GCGCAGCAAGGCCGAGGTCGCCCCCGGCGAACTGGATGAGTACTACGGGT
nosZ 3	GGAGCGGCGGACATTCCGGCGAAGTACGCGTGCTCGGCGTGCCGTCGATG
nosZ 4	GAGCTGATGCGCATACCGGTGTTCAACGTCGACTCGGCCACCGGCTGGGG
nosZ 5	GACCAACGAGAGCAAGCGGGTCCTCGGCGACAGCGCGCGCTTCCTCAACG
nosZ 6	ACTGCCACCATCCGCACATCTCGATGACCGACGGCAAGTACGACGGCAAG
nosZ 7	CATGAAATGCGACCGCATCGTCACCATTCCCAACGTCCAGGCGATCCACG
nosZ 8	TGCGCCTGCAAAAGGTGCCGCATACCCGCTACGTGTTCTGCAACGCCGAG
nosZ 9	ATCATCCCCCATCCCAACGACGGCTCGACCTTCGACCTGTCCGGCGACAA
nosZ 10	CTTCACCCTGTACAACGCCATCGACGCCGAGACCATGGAAGTGGCCTGGC
nosZ 11	TGATCGTCGACGGCAACCTCGACAACACCGACATGGACTACAGCGGCAGG
nosZ 12	GCCGCCTCCACCTGCTACAACTCGGAGAAGGCCGTCGACCTCGGCGGCAT
nosZ 13	AGATCAAGGCGAAGCGCTTCGTCACCCTCGGCGACTCGAAGGTGCCGGTG
nosZ 14	GACGGCCGGCGCAAGGACGGCAAGGACAGCCCGGTGACCCGCTACATCCC
nosZ 15	ACCGAAGAACCCCCACGGGCTGAACACCTCGCCGGACGGCAAGTACTTCA
nosZ 16	CCAACGGCAAGCTCTCGCCGACCTGCACCATGATCGCCATCGAGCGCCTC
nosZ 17	GACCTGTTCGCCGGCAAGCTGGCCGACCCGCGCGACGTGGTGGTGGGCGA
nosZ 18	GGAACTGGGCCTCGGCCCGCTGCACACCACTTTCGATGGCCGAGGCAACG
nosZ 19	ATACCACGCTGTTCATCGACAGCCAGTTGGTGAAGTGGAACCTGGCCGAC
nosZ 20	TGCCACGCCCTGCACATGGAAATGTGCGGGCGGATGCTGGTGGAAAAGGC
ccoN1 1	TCGCCGTTATGACGGTGGTCTGGGGGGTCATTGGAATGGGTCTCGGTGTC
ccoN1 2	TCGGTGGGTGCGCCCTCTTCGCCACCTCGTACTACGTGGTGCAACGTACC
ccoN1 3	CTGATTTCAGACACGCTGGCGGCCTTCACCTTCTGGGGTTGGCAGGCCGT
ccoN1 4	TCGTGGGCGCCGTGCTGACCCTGCCGCAGGGTTTCACCACCTCCAAGGAA
ccoN1 5	GCCGAACTGGAATGGCCGCTGGCCATCCTCCTGGCGATCGTCTGGATCAC
ccoN1 6	GCCTTCATCCTGGTGACGGCGATGCTGCACATCGTCAACCACATGTCGCT
ccoN1 7	ACTGGTCGCAAGCCACCCGGGCTTCATCGTGCGCATGATCGGCGGTGGTT
ackA 1	GCCCTCACGCAACATACTGGTGATCAACTGCGGCAGTTCGTCGATCAAGT
ackA 2	CCCTGGTCAACGAGGCCCACTCCCTGTTTCCCCTGCACGGCCTCGCCGAG
ackA 3	CTGGGCAGCCGCGATGCGGTGCTGCGCTGGAAGCGCGGCGGCGACAGCGA
ackA 4	CCTGATGATTCCCAACGCCGACCACCGCGCCGCCCTCGCCCAGTTGCTGC
ackA 5	TGGTGCAGAACGCCGCGGGCGGCAAGCTCCACGGCATCGGCCACCGGGTG
ackA 6	CATGGCGGCGAGCTGTTCACCCATGCCACGCGCATCGACGACCGGGTGGT
ackA 7	GGCGATCCGGGCCACCGCGCCGCTGGCGCCGCTGCACAACCCGGCCAACC
ackA 8	AAGGCATCGAGGCAGCGATGACGCTGTTTCCCAAGCTGCCCCACGTCGCC
ackA 9	TTCGACACCGCCTTCCACCAGAGCCTGCCGGAGCACGCCTACCGCTACGC
ackA 10	AGCATGGCGTGCGCCGCTACGGCTTCCACGGCACCAGCCACCGCTACGTC
ackA 11	GGCGACAGCAGTTGGCTCAGCGCCCACCTCGGCAACGGCAGCTCGACCTG
ackA 12	CATCGTCAACGGCCAGAGCCTCGACACCAGCATGGGCCTGACCCCGCTGG
ackA 13	GCCTGGTAATGGGCACCCGCAGCGGCGACGTCGACCCCAACCTGCACAGC
ackA 14	CTGGCGCGGACCCTGGGCTGGAGCCTGGAGCGCATCGACTCGATGCTGAA
ackA 15	CGAAAGCGGCCTGCTCGGCCTCTCCGACCTGTCCAACGACATGCGCACCC
ackA 16	AGCAGGAGCGCGAGCAGGGCCACCCCGGCGCGGCCCTGGCGATCGAGGTG
ackA 17	TGCTACCGCCTGGCCAAGTCCCTGGCGGCGATGAGCTGCGCCCTGCCGCA
ackA 18	GGACGGGGTGATCTTCACCGGTGGCATCGGCGAGAACTCGCCGCTGGTGC
ackA 19	CCAAGACCGCCGCCCACCTGCGGCTGTTCGACCTGCGCCTCGACCAGGAG
ackA 20	AACGCCCGCTGCGTGCGCGGCGTCGCCGGGCCGATCCAGGCCGCGGGACA
ackA 21	GCGGGTACTGGTGATCCCGACCAACGAAGAGCGGCAGATCGCCCTCGACA
nirS 1	GCCATTTGGCAAGCCACTGGTGGGCACCTTGCTCGCCTCGCTGACGCTGC
nirS 2	GCCTGGCCACCGCTCACGCCAAGGACGACATGAAAGCCGCCGAGCAATAC
nirS 3	GGTGCCGCTTCCGCCGTCGATCCCGCTCACGTGGTGCGCACCAACGGTGC
nirS 4	CGACATGAGTGAAAGCGAGTTCAACGAGGCCAAGCAGATCTACTTCCAAC
nirS 5	CCGGACATCACCCAGCAACGCGGCCAGCAATACCTGGAAGCGCTGATCAC
nirS 6	CGGCACCCCGCTGGGCATGCCGAACTGGGGCAGCTCCGGCGAGCTGAGCA
nirS 7	AACAGATCACCCTGATGGCCAAGTACATCCAGCACACCCCGCCGCAACCG
nirS 8	GAGTGGGGCATGCCGGAGATGCGCGAATCGTGGAAGGTGCTGGTGAAGCC
nirS 9	GGACCGGCCGAAGAAACAGCTCAACGACCTCGACCTGCCCAACCTGTTCT
nirS 10	TGACCCTGCGCGACGCCGGGCAGATCGCCCTGGTCGACGGCGACAGCAAG
nirS 11	ATCGTCAAGGTCATCGATACCGGCTATGCCGTGCATATCTCGCGGATGTC
nirS 12	TTCCGGCCGCTACCTGCTGGTGATCGGCCGCGACGCGCGGATCGACATGA
nirS 13	ACCTGTGGGCCAAGGAGCCGACCAAGGTCGCCGAGATCAAGATCGGCATC
nirS 14	GCGCGCTCGGTGGAAAGCTCCAAGTTCAAGGGCTACGAGGACCGCTACAC
nirS 15	CGCCGGCGCCTACTGGCCGCCGCAGTTCGCGATCATGGACGGCGAGACCC
nirS 16	AACCGAAGCAGATCGTCTCCACCCGCGGCATGACCGTAGACACCCAGACC
nirS 17	CACCCGGAACCGCGCGTGGCGGCGATCATCGCCTCCCACGAGCACCCCGA
nirS 18	CATCGTCAACGTGAAGGAGACCGGCAAGGTCCTGCTGGTCAACTACAAGG
nirS 19	TCGACAACCTCACCGTCACCAGCATCGGTGCGGCGCCGTTCCTCCACGAC
nirS 20	GGCTGGGACAGCAGCCACCGCTACTTCATGACCGCCGCCAACAACTCCAA
nirS 21	GGTTGCCGTGATCGACTCCAAGGACCGTCGCCTGTCGGCCCTGGTCGACG
nirS 22	GCAAGACCCCGCACCCGGGGCGTGGCGCCAACTTCGTGCATCCCAAGTAC

Hairpins with Alexa Fluor 488 were used for the rRNA probes, and hairpins with Alexa Fluor 647 were used for the mRNA probes. The B1 initiator/hairpin system from Molecular Technologies was used for rRNA, while B2 was used for nosZ and ccoN1 mRNA, B3 was used for narG mRNA, and B4 was used for nirS and ackA mRNA. Twenty-five microliters of each hairpin were suspended in 20 μL of amplification buffer and added to the cell solutions and incubated overnight in the dark at 25°C. The cells were then washed three times in 200 μL 5× saline sodium citrate buffer with 0.1% Tween 20 (5× SSCT) before being resuspended in 20 μL 5× SSCT and spotted onto a microscope slide (catalog number 3039-002; Thermo Fisher) and allowed to dry overnight in the dark at 4°C. Finally, slides were washed with ice-cold ~0°C ultrapurified water and then with −20°C ethanol, and slides were allowed to air dry at 25°C in the dark. Each dried cell spot was then covered with 2.5 mL ProLong Gold antifade mountant (catalog number P36930; Thermo Fisher), and slides were incubated for 48 h in the dark at 25°C before being stored and imaged using a Nikon Eclipse Ti2 fluorescence microscope at ×40 magnification with 50% laser power and 1 s exposure for the fluorescent channels. Image analysis was performed using Fiji (51). Phase images were thresholded to produce a mask and detect cells, the marked cells were thresholded by rRNA signal to eliminate nonbacterial debris from the data set, and then, mean intensity was determined for each labeled cell. Quantitative graphs were generated using Bokeh (52).

Hybridization, probe wash, and amplification buffers were acquired from Molecular Technologies (Caltech, Pasadena, CA).

ABBA preparation.

ABBA samples were prepared as described previously (21). Two sequential overnight cultures of wild-type P. aeruginosa PA14 were grown in the same liquid media as the final ABBAs, shaking aerobically at 37°C, and then diluted to an OD₅₀₀ of 0.001 in molten LB supplemented with nitrate, depending on experimental conditions, and 0.5% noble agar. The agar suspension was transferred to chambered coverglass slides and allowed to cool for 10 min before being transferred to a humidified chamber and incubated at 37°C for 12 h. After 12 h, blocks were carefully removed from their chambers using a sterilized metal spatula into prechilled 1 mL of 4% paraformaldehyde in PBS with ProtectRNA and incubated for 24 h at 4°C. Blocks were washed three times for 30 min each in 1 mL PBS, incubated for 1 h in 1 mg/mL lysozyme in 10 mM Tris-HCl at 37°C to lyse cell walls, and then washed again three times for 30 min each in 1 mL PBS and stored in 1 mL PBS. Blocks were checked on a microscope for proper growth patterning (large aggregates at the top and smaller ones in the bottom for WT in LB plus 40 mM nitrate) prior to continuing. For each condition and each gene target, three biological replicates were performed in addition to one blank (no cell inoculation) per condition.

ABBA oxygen profiling.

Oxygen profiles were determined from ABBA samples grown as described above. ABBAs were suspended in a sand bath at 37°C during oxygen measurement. The oxygen probe was a Clark-type amperometric electrode with a tip diameter of 25 μm, as described previously (28). The probe was connected to a high-sensitivity picoampere amplifier in a multimeter (Unisense, Denmark). The probe was calibrated using an anoxic solution of 0.1 M NaOH and 0.1 M sodium ascorbate at 37°C and an air-saturated solution of LB medium at 37°C. Depth was measured from the air-agar interface, defined as the last depth before the oxygen values diverged from atmospheric levels and labeled as 0 μm. The probe was positioned 75 μm or more away from the air-agar interface in a motorized micromanipulator. Measurements were taken using SensorTrace Pro 3.1.3 software at 25 μm steps for 1,000 μm, with 3 measurements at each step and 4-s increments between measurements. The oxygen profile data are averaged from 8 biological replicates and were reported for 500 μm of depth using seaborn (53).

ABBA HCR.

For each sample, 500 μL of hybridization buffer was mixed with 1 μL even and odd HCR initiator probes (2 μM stock) for 4 nM ultimate concentration. Each block was incubated in the buffer overnight at 37°C and then washed three times for 2 h in the 1 mL prewarmed wash buffer at 37°C. Each sample was then combined with 250 μL of preheated amplification buffer and 5 μL snap-cooled fluorescent hairpins (3 μM stock) for 60 nM final concentration. Samples were incubated overnight in the dark at room temperature, washed for 6 h in 1 mL 5× SSCT at room temperature in the dark, and then transferred to 500 μL PBS and stored at 4°C in the dark until imaging by using confocal fluorescence microscopy. Imaging was performed on a Zeiss LSM 800 in the Caltech Bioimaging Facility. Agar blocks were mounted on a glass-bottomed dish with the original air-agar interface oriented toward the glass bottom. Both the Alexa Fluor 488 rRNA and Alexa Fluor 647 mRNA channels were imaged simultaneously, with a laser power of 0.2% and a gain of 650 for the 488 channel and a laser power of 2% and a gain of 750 for the 647 channel. At least four image fields were imaged per sample using a 10× objective with 2× zoom for a final image field of 319.45 by 319.45 μm. For each image field, a 100-slice z-stack was imaged with an interslice distance of 6.24 μm. Images were framed such that the surface, which appears visible as a plane of rRNA signal due to a lawn of cells growing on the agar surface, was within 4 slices of the top of the image. Quantification of fluorescence signal per aggregate was performed using Imaris v9.5.1 (Imaris [RRID SCR_007370]). A batch pipeline was created to segment aggregates. The 488 signal intensity of each z-plane was normalized, and then, aggregates were segmented based on the thresholded 488 signal, with the threshold chosen manually per imaging date. The normalized signal was only used for the purpose of segmentation, and the mean signal intensity values reported here are based on the raw signal rather than the normalized signal. The mean 488 and 647 mRNA signal intensities, as well as aggregate size, were calculated and exported per aggregate. Depth was defined as the distance from the top of the image, and images were cropped for analysis such that the surface was at the top slice using native Imaris functions. Quantitative graphs were generated by using seaborn (53) and Bokeh (52). Raw images were displayed by using Fiji (51).