Exposing the Three-Dimensional Biogeography and Metabolic States of Pathogens in Cystic Fibrosis Sputum via Hydrogel Embedding, Clearing, and rRNA Labeling

ABSTRACT Physiological resistance to antibiotics confounds the treatment of many chronic bacterial infections, motivating researchers to identify novel therapeutic approaches. To do this effectively, an understanding of how microbes survive in vivo is needed. Though much can be inferred from bulk approaches to characterizing complex environments, essential information can be lost if spatial organization is not preserved. Here, we introduce a tissue-clearing technique, termed MiPACT, designed to retain and visualize bacteria with associated proteins and nucleic acids in situ on various spatial scales. By coupling MiPACT with hybridization chain reaction (HCR) to detect rRNA in sputum samples from cystic fibrosis (CF) patients, we demonstrate its ability to survey thousands of bacteria (or bacterial aggregates) over millimeter scales and quantify aggregation of individual species in polymicrobial communities. By analyzing aggregation patterns of four prominent CF pathogens, Staphylococcus aureus, Pseudomonas aeruginosa, Streptococcus sp., and Achromobacter xylosoxidans, we demonstrate a spectrum of aggregation states: from mostly single cells (A. xylosoxidans), to medium-sized clusters (S. aureus), to a mixture of single cells and large aggregates (P. aeruginosa and Streptococcus sp.). Furthermore, MiPACT-HCR revealed an intimate interaction between Streptococcus sp. and specific host cells. Lastly, by comparing standard rRNA fluorescence in situ hybridization signals to those from HCR, we found that different populations of S. aureus and A. xylosoxidans grow slowly overall yet exhibit growth rate heterogeneity over hundreds of microns. These results demonstrate the utility of MiPACT-HCR to directly capture the spatial organization and metabolic activity of bacteria in complex systems, such as human sputum.

identification after PACT) to permit the study of diverse bacterial pathogens residing in cystic fibrosis patient sputum. While PACT preserves spatial and molecular information and allows for efficient clearing as well as protein and transcript labeling via use of fluorescent probes, we incorporated key modifications to ensure (i) stabilization of amorphous sputum samples, (ii) high retention of bacteria, and (iii) efficient labeling of bacterial rRNA via hybridization chain reaction (HCR) (14,15) and fluorescence in situ hybridization (FISH). Though developed in the context of CF, MiPACT-HCR can be readily applied to diverse host-microbe systems.
Patients with CF accumulate obstructive sputum plugs in their lung airways that can harbor an array of opportunistic pathogens (16). Sputum buildup and the resultant chronic infections lead to severe lung damage and eventual respiratory failure (17). CF patients routinely expectorate infected sputum, which provides tractable samples for in situ analysis of pathogens (6,18,19). Until recently, Pseudomonas aeruginosa was the most prevalent pathogen isolated from CF patients, and P. aeruginosa colonization is well known to correlate with disease progression in CF (16,20,21). Therefore, the majority of studies addressing the biogeography of CF have focused on P. aeruginosa. FISH analysis of thin sections of CF lung or smears of CF sputum have revealed that P. aeruginosa can exist both as single cells and in large clusters and that P. aeruginosa grows more slowly in situ than in typical laboratory cultures (6,7).
While P. aeruginosa plays an important role in CF pathogenicity in many patients, other microbes also colonize the CF lung and contribute to exacerbations, or increase disease severity (16). Indeed, culture-independent studies have revealed that individuals harbor a distinct microbial ecosystem whose species composition can vary over time and treatment regimens (10,22). Though recent studies have attempted to gain a perspective on the distribution of particular clone types as a function of lung geography, these studies have been herculean, requiring microdissection, cultivation, and sequencing of thousands of regional isolates (5,19,23). Recognizing the need to study CF pathogens in situ to gain information relevant to the design of accurate in vitro models, we sought a method that would permit rapid scanning of large spatial areas at various magnifications, as well as one that would permit microbial identification and study at the single-cell level. Here, we describe our usage of MiPACT-HCR to study three important attributes of diverse pathogens in CF sputum: aggregation patterns, bacterium-host interactions, and growth rates.

RESULTS AND DISCUSSION
We obtained seven sputum samples (numbered 1 to 4 and 5.1, 5.2, and 5.3) with consent from five patients at the Children's Hospital of Los Angeles (CHLA). Samples 1, 4, 5.2, and 5.3 were collected during an exacerbation, while samples 2, 3, and 5.1 were collected during outpatient well visits. Disease states varied between patients, with patients 1 and 3 having FEV1% (percent forced expiratory volume in 1 s, a measure of lung function) values of 48 and 44 (moderate obstruction), respectively, while the remaining patients had FEV1% values greater than 70 (mild to normal).
When fixed in paraformaldehyde (PFA; 4%) and embedded in A 4 P 0 (4% acrylamide, 0% PFA), sputum completely dissolved during clearing. To provide more structural stability, we replaced acrylamide with 4% 29:1 acrylamide:bis-acrylamide (29A:1B) 4 P 0 , providing additional cross-linking (Fig. 1a). Use of (29A:1B) 4 P 0 preserved sputum integrity and allowed for clearance in SDS (Fig. 1b). Samples took 3 to 14 days to fully clear (Fig. 1b). Because sputum is composed largely of host-derived DNA and mucins (24), we labeled DNA with 4=,6-diamidino-2-phenylindole (DAPI) and mucins with rhodamine-conjugated lectin (wheat germ agglutinin [WGA]) after clearing to obtain a structural context. Imaging revealed a high degree of compositional variation between samples (Fig. 1c). For example, sputum samples from patients 1 and 2 were composed largely of lectin-stained mucin, with interspersed DAPI-bright host cells. Sputum 5 was composed almost entirely of polymorphonuclear neutrophils (PMNs), consistent with findings that PMNs are a major component of CF patient sputum (25). PMN cell boundaries were outlined by a network of extracellular DNA (Fig. 1d), potentially a result of neutrophil extracellular traps (NETs) (26). While intersample heterogeneity was evident, sampling different regions of a single sputum sample revealed that intrasample composition was relatively homogenous (see Fig. S1 in the supplemental material).
We next verified that the common CF pathogens P. aeruginosa and Staphylococcus aureus could be retained and visualized under the same embedding and clearing conditions required for retaining sputum integrity. There was no significant loss of DAPIstained logarithmic-or stationary-phase bacteria after clearing of pure cultures embedded in (29A:1B) 4 P 0 hydrogel blocks (see Fig. S2a in the supplemental material). FISH staining with saturating probe concentrations of the universal bacterial probe EUB338 after MiPACT (see Fig. S3a in the supplemental material) revealed that the Gram-positive microbe S. aureus required treatment with lysostaphin after clearing via SDS, while the Gramnegative P. aeruginosa did not require lysozyme treatment (see Fig. S3b). In sputum, autofluorescence makes bacteria, particularly slowly growing cells, difficult to demarcate by FISH (see Fig. S4 in the supplemental material). Therefore, we employed HCR, a FISH amplifying technique which has previously been used to fluorescently label RNA in zebrafish embryos, brain tissue, and environmental microbes (15,27,56). HCR entails hybridizing target RNA with a DNA probe that triggers amplification of fluorescently labeled DNA hairpins into polymer chains via a specific initiator region (14,15). To directly compare FISH and HCR, FISH with a dilabeled AlexaFluor 594 EUB338 probe and HCR with an initiator EUB338 probe and AlexaFluor 594 hairpins were performed separately on stationary-phase cultured cells embedded in (29A:1B) 4 P 0 and cleared for 5 days. HCR increased the average fluorescence intensity of P. aeruginosa cells by~68-fold and S. aureus cells by~42-fold above levels obtained with FISH.
HCR hybridizations in sputum were optimized such that (i) the EUB338 probe bound and nucleated hairpin polymerization, (ii) samples did not fluoresce when incubated with both NON338, the reverse complement of EUB338, and fluorescent hairpins, and (iii) class/genus-specific probes did not cross-react with other relevant bacteria (see Fig. S5 and S6 in the supplemental material). The Betaproteobacteria probe BET42a, used for Achromobacter xylosoxidans, had weak cross-reactivity with P. aeruginosa and was therefore not used in P. aeruginosa culture-positive samples. Some species-specific probes tested, including those specific for A. xylosoxidans, were excluded due to their inability to withstand the stringent hybridization and wash conditions necessary for HCR specificity (see Materials and Methods). Object-based colocalization analysis after HCR multiplexing was performed to further validate HCR specificity. Greater than 90% of objects (dis-DePas et al. crete HCR-identified cells or aggregates with a size of Ͼ4 voxels) in sputum were concurrently identified by using two separate universal probes (see Fig. S6). Moreover, Ͼ90% of objects identified by the class/genus-specific probes used in sputum colocalized with EUB338 but not with other class/genus-specific probes (see Fig. S6). HCR allowed multiscale visualization of bacteria; low magnification (e.g., ϫ10) enabled broad surveying of the sample (Fig. 1d), and increased magnification (e.g., ϫ25) enabled singlecell resolution and revealed the spatial organization of bacteria and host cells (Fig. 1d).
Once optimized for retention and identification of bacteria in sputum, we utilized MiPACT-HCR to measure bacterial aggregation in situ. Bacterial aggregates are thought to contribute to the persistence of pathogen populations in chronic infections, including those in CF patients (1,(28)(29)(30), yet direct evidence for this is sparse (6,19,31). We examined distribution patterns of Staphylococcus sp. in sputum sample 5.1 (culture positive for S. aureus and A. xylosoxidans) by using a Staphylococcus-specific probe. Cultured bacteria were analyzed in parallel with magnification ϫ25 sputum surveys to calibrate our expectations for the signal size of single bacterial cells (see Fig. S7 in the supplemental material). The mean fluorescence volume of objects in stationary-phase cultures of S. aureus was 12.1 m 3 . In sputum, Staphylococcus cells existed in a range of intermediate aggregates, with only 6% of objects being greater than 1,000 m 3 (see Fig. S7a). The Staphylococcus size distribution in sputum cleared for 5 or 14 days was similar, signifying that clearing preserves a range of bacterial aggregate sizes (see Fig. S2b in the supplemental material). Taking advantage of the large-scale surveying enabled by MiPACT, we next acquired ϫ10 magnification Z-stacks of sputum sample 5.1, analyzing thousands of objects in sputum volumes of~0.1 to 0.3 mm 3 . Like the ϫ25 magnification surveys, Z-stacks at ϫ10 magnification revealed that Staphylococcus was chiefly visible as small to medium aggregates (85% of objects ranged in size from 50  intensity projection of a Z-stack after HCR and staining with DAPI (blue) and WGA (orange), acquired with a 10ϫ objective. The second panel is a maximum intensity projection of a separate Z-stack acquired with a 10ϫ objective while only collecting HCR signal (Staphylococcus-specific probe mix with B4 amplifier and AlexaFluor 488-conjugated B4 hairpins) (7,910 objects analyzed). Each object identified in the second panel's Z-stack was binned according to proportional object volume (each object's fluorescent volume relative to the total fluorescent volume for that Z-stack; shown in the graph on the right). The top right panel is a maximum intensity projection of a Z-stack acquired with a 25ϫ objective, highlighting a representative region from the same sputum sample. (b to d) The same analysis was applied to sputum 5.1 using a Betaproteobacteria-specific probe with B4 amplifier and AlexaFluor 488-conjugated B4 hairpin (21,255 objects analyzed) (b), to sputum 1 with a Pseudomonas-specific probe mixture with B4 amplifier and AlexaFluor 647-conjugated B4 hairpins (9,520 objects analyzed) (c), or a Streptococcus-specific probe mixture with AlexaFluor 488-conjugated B4 hairpins (4,603 objects analyzed) (d).
to 1,000 m 3 ) (see Fig. 2a and S7a). In contrast, Betaproteobacteria showed very little aggregation; 71% of objects fell in the smallest bin (Ͻ50 m 3 ) ( Fig. 2b; see also Fig. S7b in the supplemental material). Because S. aureus is the most common pathogen cultured from CF patients (21), we monitored aggregation in samples from three distinct areas of sputum 5 (5.1A, 5.1B, and 5.1C) (see Fig. S8a in the supplemental material). Also, three temporal samples from patient 5 (5.1, 5.2 [103 days after 5.1], and 5.3 [1 day after 5.2]) and a sample from patient 4, also culture positive for S. aureus, were analyzed. All samples demonstrated a similar pattern of small-to medium-sized aggregates (see Fig. S8b). Next, we took advantage of the straightforward multiplexing enabled by HCR (15) to concurrently probe Betaproteobacteria and Staphylococcus sp. in different regions of sputum 5.1 (see Fig. S9 in the supplemental material). Both were present in all areas of sputum 5.1 tested, but their relative abundance differed between regions (see Fig. S9).
The only organism for which sputum 1 was culture positive was P. aeruginosa, and surveying at ϫ10 and ϫ25 magnifications with a Pseudomonas-specific probe mixture revealed small objects  (6,19). While surveying sputum sample 1 with the EUB338 probe, we unexpectedly found bacteria with a distinctive filamentous morphology. Patient 1 had previously produced a sputum sample that was culture positive for Streptococcus anginosus, and probing sputum 1 with a Streptococcus-specific probe mixture revealed a dense bacterial population (Fig. 2d). The largest proportion of Streptococcus signal volume (which ranged from 10 to 300,000 m 3 at ϫ10 magnification) came from large (Ͼ1,000 m 3 ) aggregates (Fig. 2d). Streptococcus is often missed in routine clinical culturing, highlighting the gap that is often observed between culture-dependent and culture-independent techniques (32,33).
An important advantage of surveying large volumes at low magnification is the ability to quickly identify key areas that can benefit from higher magnification. After performing ϫ10 surveys in sputum, we focused on large bacterial aggregates with a 25ϫ objective (Fig. 3). Multiplexing of sputum 1 for both Streptococcus and Pseudomonas revealed that aggregates were mostly monospecies, with little visible interaction (Fig. 3a). Pseudomonas aggregates existed in a range of sizes, with large biofilms having diameters up to~50 m (Fig. 3b). PMNs could be seen surrounding, and in some cases within, the biofilm structure (Fig. 3b). With finer resolution, it became apparent that the large Streptococcus aggregates visible at ϫ10 had morphologies indicative of association with an interior substrate ( Fig. 3c and d). To determine the substrate, we stained samples of sputum 1 with DAPI and WGA after HCR with Streptococcus-specific probes. Staining revealed that the areas inside Streptococcus aggregates were in fact host cells with single-lobed nuclei (Fig. 3d). Each host cell boundary stained brightly with WGA, potentially indicative of polysaccharide moieties on the host cell surface (Fig. 3d and e). These results exemplify the ability of MiPACT-HCR to identify novel bacterium-host interactions.
While the importance of aggregative or biofilm modes of growth in chronic infection is well appreciated (1,3,4,(28)(29)(30)34), the role of growth rate is less so. Recent studies demonstrated slow in situ S. aureus growth rates in CF sputum (35) and slow-growth-specific regulation networks in P. aeruginosa (36), underscoring the importance of careful growth measurements in situ for designing in vitro models that faithfully recapitulate in vivo physiology. Many species show a linear relationship between growth rate and rRNA abundance (37), but a number of challenges impede the calculation of precise growth rates in sputum from FISH data alone: rRNA abundance can be completely decoupled from growth rate in some species (37), at low growth rates rRNA abundance ceases to linearly correlate with growth rate (7), and sputum autofluorescence can overwhelm signals from slowly growing cells (see Fig. S4 in the supplemental material). To address these problems, we refrained from estimating specific growth rates of individual cells, instead opting to describe the growth rates of bacterial populations with respect to logarithmic-and stationary-phase standards, analyzed in parallel. We first verified that the FISH signal of both S. aureus and A. xylosoxidans decreased in stationary phase (see Fig. S10a in the supplemental material). We then determined that FISH signal from logarithmic cells did not substantially decay even after 14 days of clearing (see Fig. S10b). Lastly, we used HCR to distinguish bacterial signals from background autofluorescence and to select for the desired genus in a mixed population. For analysis, HCR-identified objects were outlined and EUB338 FISH fluorescence (the proxy for growth rate) within the outlines was quantified (Fig. 4a). EUB338 was chosen as the FISH probe due to its robustness and hybridization to a separate rRNA locus, preventing probe competition.
The growth rate measurements described above were performed on objects identified with a Staphylococcus-specific HCR probe from portions of sputum 5.1 taken from distinct areas of the sample (5.1A, -B, and -C, with subsamples 5.1A1, 5.1A2, etc.) ( Fig. 4a and b). We then calculated the percentage of objects that crossed a threshold above which 90% of logarithmic-phase cultured cells fell (corresponding to a doubling time of Ϸ1 h). Regions 5.1A1, 5.1B1, and 5.1C1 demonstrated mostly low growth rates, with 0%, 4.3%, and 0.2% of objects reaching signal threshold (Fig. 4c). Interestingly, subsample 5.1C2 demonstrated an increase in growth rate compared to its neighbors, with 20.8%, of objects reaching threshold (no objects in 5.1A2 and 7.0% of objects in 5.1B2 reached threshold) (Fig. 4d). Temporal samples from patient 5 (sputum samples 5.2 and 5.3) did not reach threshold (Fig. 4e). Samples 4A and -B and -C, from a different S. aureus culture-positive patient, contained slowly growing bacteria as well, with only 2.7%, 4.1%, and 2.0% of objects, respectively, above threshold (Fig. 4f). In order to determine if aggregate size correlated with our proxy for growth rate, we separated S. aureus objects (from Fig. 4c to f) into quartiles with respect to object size and plotted against mean fluorescence intensity from FISH. Fluorescence intensity increased significantly with increasing object size, signifying that larger aggregates may have experienced higher growth rates (see Fig. S10c and d in the supplemental material). This was possibly due to the greater susceptibility of single cells to antibiotics (34). Interestingly, cultured, planktonic S. aureus cells also showed a positive correlation between object size and fluorescence intensity (see Fig. S10e). This is consistent with previous studies showing that RNA abundance and cell size increase with higher growth rates (38). Further study would be needed to determine what, if any, correlation exists between aggregate size and cell size in situ.
Lastly, as A. xylosoxidans is subject to the same in vivo conditions as S. aureus in these samples, we assayed growth rates of  Betaproteobacteria in sputum sample 5.1. All populations were growing slowly, with 0.1% of objects from 5.1A2, 0.5% from 5.1B2, and 0.8% from 5.1C2 reaching the threshold set by logarithmic-phase A. xylosoxidans standards (doubling time of Ϸ0.6 h) (Fig. 4g).
We have shown that MiPACT-HCR is effective at retaining and visualizing bacteria in complex samples after optical clearing, and enables the rapid survey of large volumes of these samples. In our CF patient sputum samples, P. aeruginosa, Streptococcus sp., A. xylosoxidans, and S. aureus aggregation patterns varied, suggesting that broader, species-specific cellular interaction trends occur in vivo. Our results also reinforced that in vivo CF pathogen aggregates, particularly in regards to S. aureus and A. xylosoxidans, are considerably smaller than typical laboratory biofilms (34), an important observation when attempting to model biofilms for this context in the laboratory. We have also demonstrated that MiPACT-HCR, in combination with FISH, provides an accessible method for assessing growth rates in situ. Using this strategy, we found low growth rates for S. aureus and A. xylosoxidans, consistent with the few in situ measurements existing for these CF pathogens (6,7,35). While most of the populations surveyed were slow growers, there were pockets of relatively fast growth, illustrating the heterogeneity in in vivo growth rates that is beginning to be described in the literature (35). Indeed, due to concomitant expectoration of sputum plugs from different airways, and from the gradients of nutrients and oxygen existing with various sputum plug geometries (18), heterogeneity within the same expectorated sample is expected. Furthermore, particular sputum environments may favor one bacterial species or another, leading to the compositional heterogeneity that we observed in the same sputum sample.
MiPACT provides a widely accessible technique to characterize biogeography in situ, in three dimensions (3D) and at imaging depths not previously practical, and has the potential to reveal the range of microbes growing slowly in a wide variety of contexts. Through systematic application of MiPACT and fluorescent hybridization techniques such as HCR, patterns may emerge that will lead to new insights into heterogeneous polymicrobial communities, conditions conducive to promoting health (the human microbiome), treating disease (bacterial infections), or understanding important ecological interactions.

MATERIALS AND METHODS
Strains and growth conditions. The following strains were used in this study: Pseudomonas aeruginosa PA14, Staphylococcus aureus MN8, and a clinical isolate of Achromobacter xylosoxidans generously donated by CHLA. Strains were grown aerobically in lysogeny broth (LB) at 37°C with shaking at 250 rpm.
Growth curves studies and embedding bacteria in acrylamide-based hydrogel blocks. Five-milliliter cultures of S. aureus, P. aeruginosa, and A. xylosoxidans were inoculated with single colonies from LB agar plates. Once cultures reached late exponential phase, the cultures were diluted 1:1,000 into glass culture tubes with 10 ml of medium. The optical density at 500 nm (OD 500 ) or the OD 600 of these cultures was tracked using a Thermo Spectronic 20Dϩ system during aerobic growth, with shaking, at 37°C. At intervals denoted in Fig. S10 in the supplemental material, cells were removed from the culture, normalized to an OD 500 or OD 600 of 1 in LB, and then PFA (to 2% [vol/vol]; EMS 15713) was added. Samples were slowly rotated at 4°C overnight. The next day, after washing with phosphate-buffered saline (PBS), fixed cells were diluted 1:10 into 4% (vol/vol) 29:1 acrylamide:bis-acrylamide (catalog number 161-0146; Bio-Rad) and 0.25% (wt/vol) VA-044 hardener (catalog number 27776-21-2l; Wako) in 1ϫ PBS for polymerization. After leaving samples open, but covered, in an anaerobic hood for 5 min to decrease oxygen in the headspace, blocks were polymerized in a 37°C water bath for 3 h, without shaking, and then cut to~1 mm 3 . Unless otherwise noted, blocks were cleared for 5 days in 8% SDS and then processed for FISH as described below. All solutions were sterilized with a 0.2-m filter.
Sputum sample collection. Sputum samples were collected at CHLA in accordance with study CCI-13000211, which was approved by the CHLA IRB. Immediately upon expectoration, sputum samples were placed into 50-ml conical tubes with 25 ml of 4% paraformaldehyde solution in 1ϫ PBS (pH 7.2). Samples were incubated for 24 h at 4°C, gently washed 3 times in 50 ml 1ϫ PBS (pH 7.2), and then stored in 1ϫ PBS with 0.01% (wt/vol) sodium azide at 4°C.
MiPACT processing of sputum. For sputum samples, small sections roughly 5 mm in diameter were removed under sterile conditions with a scalpel and placed in a 1.5-ml culture tube. Samples were incubated overnight in 4% (vol/vol) 29:1 acrylamide:bis-acrylamide and 0.25% (wt/vol) VA-044 hardener in 1ϫ PBS, made fresh and filter sterilized. After overnight incubation, samples were moved into an anaerobic hood and left open, but covered, for 5 min to remove headspace oxygen. Samples were polymerized in fresh solution for 3 h at 37°C in a water bath, without shaking. Under sterile conditions, samples were routed to a solution of 8% SDS, pH 8.0, at 37°C, with shaking until cleared. Generally, samples took from 3 to 14 days to fully clear (average, 5 days). After clearing, samples were washed 3 times in 50-ml conical tubes in 1ϫ PBS (ml volumes) to remove SDS. Once cleared, samples were stored in 1ϫ PBS with 0.01% (wt/vol) sodium azide and 1ϫ ProtectRNA RNase inhibitor (catalog number R7397; Sigma) at 4°C. PACT has a flexible formulation with application-specific recommendations regarding inclusion or exclusion of PFA and bis-acrylamide (see the troubleshooting instructions of Treweek et al. in reference 12).
Lysozyme and lysostaphin digestion. Before lysozyme/lysostaphin treatment, samples were trimmed and sectioned with ethanol-sterilized razor blades to~1-mm 3 blocks. All sputum and cultured cells in acrylamide-based hydrogel blocks were incubated in 1.5-ml microcentrifuge tubes in 500 l of a sterile solution of lysozyme (1 mh/ml; catalog number L6876; Sigma) and lysostaphin (0.05 mg/ml; catalog number L7386; Sigma) in 10 mM Tris-HCl (pH 7.6) for 3 h at 37°C with shaking. Samples were then washed 2 times for 30 min each in 50 ml of 1ϫ PBS.
FISH. To ensure adequate binding site saturation of our target rRNAs, we evaluated using FISH over a range of probe concentrations, with loss of binding site saturation occurring at around 0.008 ng/l (1.2 nM) (see Fig. S4 in the supplemental material). To accommodate potential sample variability, we chose a higher concentration, 1 g/ml (150.7 nM), for subsequent FISH experiments. Therefore, unless otherwise noted, samples were hybridized with 1 g/ml (150.7 nM) probe at 46°C, with shaking, overnight in 15% formamide for cultured cells in acrylamide-based hydrogel blocks or 25% formamide for sputum samples (or growth rate standards). Each~1-mm 3  and 0.01% sodium azide) with 10 g/ml DAPI at room temperature (RT) with gentle shaking, protected from light, for at least 24 h before imaging.
HCR. HCR involves a hybridization step with an unlabeled DNA probe. This probe contains a specific sequence tag (the specific sequences used here were termed B1, B3, and B4) that triggers the oligomerization of pairs of fluorescently labeled DNA hairpins (the amplification step; for more details, see Choi et al. [15]). (ii) Amplification. Before amplification, hairpin pairs were heated to 95°C for 1.5 min in a thermocycler in separate PCR tubes. Hairpins were then cooled at room temperature (RT) for at least 30 min while protected from light. Each hairpin in a pair was added at 1:25 from a 3 M stock to a final concentration of 120 nM in HCR amplification buffer (100 l 20ϫ SSC, 100 mg dextran sulfate, and Milli-Q H 2 O to 1 ml) for acrylamidebased hydrogel blocks and sputum 1, or at 1:12.5 to a final concentration of 240 nM for sputum from patients 5 and 4. A 120-l volume of amplification buffer with the appropriate hairpin mixture was then added to each sample in a 1.5-ml centrifuge tube. Samples were incubated at RT with gentle shaking for 48 h. For sputum samples from patients 4 and 5, samples were amplified for 48 h. For sputum samples from patient 1, samples were amplified for 24 h. After amplification, samples were washed in 50 ml of 337.5 mM FISH wash buffer at 48°C for 3 h in a water bath, without shaking. Samples were then incubated in 250 l RIMS with 10 g/ml DAPI (1:1,000 from 10-mg/ml stock solutions in dimethyl sulfoxide) at RT with gentle shaking for at least 24 h before imaging.
FISH probes were dilabeled with the indicated fluorophores, with one fluorophore at the 5= end and one at the 3= end. Three HCR initiator/hairpin systems were used in this study: B1, B3, and B4. For B1 initiator probes, the sequence 5=-TATAGCATTCTTTCTTGAGGAGGGCAGCAAACGGGAAG AG-3= was added to the 3= end of the indicated DNA probe. For B3 initiator probes, 5=-TAAAAAAGTCTAATCCGTCCCTGCCTCTATATCTCCACT C-3= was added to the 3= end of the indicated DNA probe.
For B4 initiator probes, 5=-ATTTCACATTTACAGACCTCAACCTA CCTCCAACTCTCAC-3= was added to the 3= end of the indicated DNA probes. DNA hairpins conjugated to either AlexaFluor 488, AlexaFluor 594, or AlexaFluor 647, as indicated, were used with the appropriate initiator probe sets. Hairpins conjugated to fluorophores were purchased from Molecular Instruments.
Lectin staining. When indicated, lectin staining was performed immediately before incubation in RIMS/DAPI. WGA conjugated to rhodamine (vector RL-1022) was used for lectin staining. Samples were incubated in 1 ml of 50 g/ml WGA in 1ϫ PBS at RT, with shaking, for 24 h. They were then washed for another 24 h at RT, with shaking, in 1 ml of 1ϫ PBS before incubation in RIMS/DAPI.
Imaging. Prior to imaging, samples were incubated at RT overnight, with shaking, in RIMS with 1 g/ml DAPI. Samples were then mounted on slides in 0.9 mm or 1.7 mm CoverWell perfusion chambers (Electron Microscopy Services) with a coverslip on the top. Imaging was performed using a Zeiss LSM 780 confocal microscope or a Zeiss LSM 880 confocal microscope with either a Plan-Apochromat 10ϫ/0.45-numerical aperture M27 objective (working distance [wd], 2.0 mm) or an LD LCI Plan-Apochromat 25ϫ/0.8-numerical aperture Imm Corr DIC M27 multiimmersion objective (wd, 0.57 mm), using glycerol as the immersion fluid. All images and Z-stacks were collected in 12-bit mode, with at least a 1,024-by-1,024 scan format and a line averaging of 2. Image reconstructions were made with Imaris imaging software (Bitplane) or the FIJI distribution of ImageJ (39,40). Image analysis was chiefly performed using the 3D object counter plug-in (41) in FIJI to identify and quantify fluorescently labeled objects. The R package was used to make box plots and for t test analysis (42).
Aggregation analysis. Aggregation measurements were performed from HCR-stained samples using probes listed in Table 1 (probes that did not withstand the HCR wash step are also shown). Laser power and gain settings were adjusted for each sample so that the brightest objects were just below saturation. For image analysis, the 3D object counter from ImageJ was utilized to record the fluorescence volume of each object. Objects were then binned according to volume. The fluorescence volume of each object in a given bin was summed, and each bin sum was divided by the total fluorescent volume of the entire Z-stack to obtain a proportional volume value for each bin.
Growth rate analysis. For growth rate measurements, HCR was performed on sputum samples and, in parallel, on logarithmic or stationaryphase bacteria (of the appropriate species) embedded in acrylamide-based hydrogel blocks that had been cleared for 5 days (unless otherwise noted). All HCR for growth rate measurements was performed with hairpins conjugated to Alexafluor488. After the typical HCR hairpin wash, all samples were hybridized with EUB338 di-labeled with CY5 in 25% formamide for 24 h. Samples were then washed in 84 mM FISH wash buffer for 3 h at 48°C in a water bath with no shaking. Samples were incubated with RIMS/ DAPI for at least 24 h at RT with gentle shaking before imaging. During image acquisition, laser power and gain settings were adjusted for each sample in the HCR channel (Alexafluor488) so that the brightest objects were just below saturation. Laser power and gain settings for the FISH channel (CY5) were adjusted for so that the brightest objects in logarithmic phase culture standards were just below saturation. Once adjusted for logarithmic standards, the FISH settings were kept constant for the stationary-phase standard and for all sputum samples. For image analysis, the 3D object counter from ImageJ was used to perform segmentation from the HCR channel. The redirect option was used to measure fluorescence of each HCR-identified object in the FISH channel. The threshold and minimum size settings in 3D object counter were kept constant for all samples in a set. Objects were binned by average fluorescence intensity, and then the relative frequency of each bin was calculated by dividing the number of objects in each bin by the total number of objects for each sample. Histograms were then created from the relative frequency of each bin. Colocalization analysis. For colocalization analysis, HCR was performed on sputum samples with two separate probes and two corresponding hairpin sets, one conjugated to AlexaFluor488 and one conjugated to Alex-afluor647. Z-stacks were obtained with a 25ϫ objective. The 3D object counter plug-in was used for each channel to obtain threshold images and identify objects with a minimum size of 5 voxels. The two binary Z-stacks (one from each channel) were multiplied together, and objects in the product Z-stack (with a size of at least 3 voxels) were counted with a 3D object counter. The number of objects in the product Z-stack was divided by the number of objects in each original Z-stack to yield the percent colocalization.

ACKNOWLEDGMENTS
A portion of the imaging was performed in the Caltech Biological Imaging Facility, with the support of the Caltech Beckman Institute and the Arnold and Mabel Beckman Foundation. We thank Elise Cowley, Ajay Kasi, and the pulmonary clinic team and patients from Children's Hospital of Los Angeles (CHLA) for their assistance with sample collection and support of this study; Harry Choi (and the entire Beckman Institute Molecular Instruments team), Brittany Belin, Antti Lignell, Bin Yang, Alex Persat, and Octavio Mondragón-Palomino for technical advice, and members of the Newman and Gradinaru groups for helpful feedback and discussions. This study was funded by grants from the NIH (grant no. 5R01HL117328-04) to D.K.N., from HHMI to D.K.N., and the Center for Environmental Microbial Interactions (CEMI) at Caltech to D.K.N. and V.G. V.G. is a Heritage Principal Investigator supported by the Heritage Medical Research Institute. This work was also funded by the NIH Director's New Innovator program (IDP20D017782-01) and PECASE (V.G.) and the Beckman Institute for the Resource Center on CLARITY, Optogenetics and Vector Engineering for technology development and broad dissemination.

FUNDING INFORMATION
This work, including the efforts of Viviana Gradinaru and Dianne K. Newman, was funded by Center for Environmental Microbial Interactions. This work, including the efforts of Dianne K. Newman, was funded by HHS | National Institutes of Health (NIH) (5R01HL117328-04). This work, including the efforts of Viviana Gradinaru, was funded by HHS | National Institutes of Health (NIH) (IDP20D017782-01 and PECASE). This work, including the efforts of Dianne K. Newman, was funded by Howard Hughes Medical Institute (HHMI).