O-Specific Antigen-Dependent Surface Hydrophobicity Mediates Aggregate Assembly Type in Pseudomonas aeruginosa

ABSTRACT Bacteria live in spatially organized aggregates during chronic infections, where they adapt to the host environment, evade immune responses, and resist therapeutic interventions. Although it is known that environmental factors such as polymers influence bacterial aggregation, it is not clear how bacterial adaptation during chronic infection impacts the formation and spatial organization of aggregates in the presence of polymers. Here, we show that in an in vitro model of cystic fibrosis (CF) containing the polymers extracellular DNA (eDNA) and mucin, O-specific antigen is a major factor determining the formation of two distinct aggregate assembly types of Pseudomonas aeruginosa due to alterations in cell surface hydrophobicity. Our findings suggest that during chronic infection, the interplay between cell surface properties and polymers in the environment may influence the formation and structure of bacterial aggregates, which would shed new light on the fitness costs and benefits of O-antigen production in environments such as CF lungs.

assessed. To resolve this, we evaluated aggregate formation in seven genetically diverse isolates sourced from heterogeneous populations of the P. aeruginosa strain PAO1 previously evolved in biofilms for 50 days (23). We assessed how each isolate formed aggregates in a polymer-enriched, spatially structured CF growth medium (SCFM2; containing mucin and extracellular DNA [eDNA] polymers), which has previously been used as an in vitro CF model to study the biogeography and physiology of P. aeruginosa (8,(24)(25)(26).
We found that the PAO1 ancestor and five isolates formed a stacked pattern (stacked aggregates), where cells closely packed lengthwise, similar to those identified in previous polymer-driven depletion-aggregation studies (27). Two isolates formed distinct disorganized aggregates of various sizes (clumped aggregates), similar to aggregates previously observed in CF sputum samples (10,28). Whole-genome sequencing showed that the two clumping isolates had alterations in the ssg gene (PA5001), which has previously been shown to be involved in lipopolysaccharide (LPS) core and O-antigen biosynthesis (29)(30)(31)(32). In P. aeruginosa, LPS contains three major components: the lipid A layer of the outer membrane, a core oligosaccharide, and O-antigen components. O-antigens are further subdivided into a D-rhamnose homopolymer found in most strains called the common polysaccharide antigen (CPA) (formerly A-band) and a variable heteropolymer of 3 to 5 sugars called the O-specific antigen (OSA) (formerly B-band) that confers serotype specificity (29,32).
We hypothesized that changes in O-antigen glycoforms capping LPS would lead to different aggregate assembly types by altering the physicochemical properties of the bacterial cell surface. This could result in new interactions (e.g., between the bacteria or the bacteria and polymers) that compete with the entropic force that otherwise leads to stacked aggregation in this polymer-rich environment. To elucidate the contribution of LPS-capping glycoforms to aggregate assembly, we assessed the aggregate formation of clean deletion mutants of OSA and CPA. We found that the loss of OSA and changes in the capped LPS core glycoforms led to increased hydrophobicity of the cell surface that overcame the entropic force imposed by host polymers, resulting in disorganized, irreversible clumping. Most importantly, we demonstrate that aggregate shape and structure are dependent on the interplay between the physical properties of the environment and the biological mediation of bacterial cell surface properties governed by the LPS core and OSA.

RESULTS
Distinct aggregate assembly types in genetically diverse P. aeruginosa isolates. Genetically and morphologically heterogeneous isolates of P. aeruginosa are commonly collected from expectorated CF sputum samples (20,21,33,34). Since it is known that several lineages of P. aeruginosa can stably coexist in CF airways, we tested whether population heterogeneity impacted aggregate formation. We chose seven distinct morphotypes isolated from a previous study where we evolved biofilms of PAO1 in synthetic polymer-free sputum medium (SCFM) for 50 days (23, 35, 36) (see Fig. S1A in the supplemental material). We assessed aggregate formation in a spatially structured iteration of SCFM termed SCFM2, which contains mucin and eDNA polymers (24). We identified two distinct types of aggregate assembly where PAO1 and five of the evolved isolates (A2, B8, B13, C25, and D4) were assembled into stacked aggregates, where cells were closely aligned side by side by entropic force, similar to previous reports (27) (Fig. 1A; Fig. S1B and C). In contrast, two of the evolved isolates (A9 and B9) formed clumped aggregates that appeared as disorganized small and large groups of cells, similar to bridging aggregation (15) (Fig. 1A). We also investigated the growth of A9 and B9 in SCFM (no addition of eDNA or mucin) and observed that both strains formed clumps even in a polymer-free environment, while the other isolates did not form any aggregates (Fig. S2A).
To identify the genetic determinants of clumping assembly in the evolved isolates, we performed whole-genome sequencing on each isolate using the Illumina MiSeq platform. We used breseq (0.34.0) for variant calling between the evolved isolates and the PAO1 ancestor (37). While we found differential levels of genetic variation in each isolate compared to PAO1, we observed that A9 and B9 each contained a 1-bp deletion in the ssg gene (PA5001) (Data Set S1). To confirm that mutation of ssg results in clumped aggregate assembly, we complemented A9 and B9 with an intact ssg gene in trans. We found that in both isolates, ssg complementation restored the stacked aggregate assembly seen in PAO1 ( Fig. 1B; Fig. S2B), suggesting that Ssg plays a role in the aggregate assembly type.
A distinct feature of the stacked versus clumped aggregates was the average volume. We found that stacked aggregate volumes were 2 to 4 times larger (median aggregate sizes of ;63 mm 3 and 89 mm 3 for PAO1 and A2, respectively, and ;200 mm 3 for ssg-complemented A9 and B9) than those of clumped aggregates (median aggregate sizes of ;23 and 30 mm 3 for A9 and B9, respectively) (Fig. 1B). To quantify these observed differences in the distribution of aggregate biomass in cells with stacked and clumped aggregate assembly types, we compared the distributions of biovolume (ratio of the aggregate volume to the surface area) for each type of aggregate assembly in SCFM2. We found that regardless of the size, the median biovolume in stacked aggregates was significantly larger than that in clumped aggregates (Table S1).
OSA and not other biofilm traits determines aggregate assembly type. The proposed function of Ssg is a glycosyltransferase, involved in LPS and exopolysaccharide (EPS) biosynthesis (30,38). P. aeruginosa strains with mutations in ssg have previously been shown to display decreased motility, enhanced phage resistance, and a lack of Oantigen (30,31,39). We next determined whether the different aggregate assemblies  (Fig. S3). We then determined the aggregate assembly type of the O-antigen mutants in SCFM2. We found that ssg, wbpL, and wbpM mutants with no OSA formed clumped aggregates, but the lack of CPA alone (PAO1 Drmd) did not change the aggregate assembly type ( Fig. 2A and B).
Biofilm formation by P. aeruginosa is regulated by several well-described mechanisms such as exopolysaccharide production, adhesins, and quorum sensing (QS) (6,9,42,43). To determine whether any of these factors interfered with stacked aggregation in SCFM2, we examined the aggregate assembly of defined mutants in exopolysaccharide production (PAO1 Dpel Dpsl) and lectins (PAO1 DlecA and PAO1 DlecB) and a mutant lacking a major QS regulator (PAO1 DlasR). Although the role of QS and biofilm formation remains controversial, we assessed this mutant because LasR regulates several pathways that could impact the aggregation type (7,44). We found that all these mutants displayed stacked aggregate assemblies like PAO1 (Fig. 3). This indicated that changes in the aggregate assembly type were not due to alterations in common phenotypes associated with biofilm formation; only the loss of OSA was important. Loss of OSA leads to clumped aggregates, independent of polymer and cell density. To determine differences in stacked and clumped aggregate assemblies in SCFM2, despite the presence of polymers, we monitored aggregate assemblies of PAO1 and PAO1 DwbpL over time. We found that an increase in the initial cell density resulted in the rapid formation of stacked aggregates, while there were no changes in clumped aggregate assembly (Movies S1 and S2).
When we monitored the aggregate formation of PAO1 and PAO1 DwbpL over 6 h, we found that there was a significant change in the PAO1 aggregate biovolume after 180 min of cell growth in SCFM2 and when the stacks were assembled (Fig. 4A), whereas regardless of cell density, the biovolume of PAO1 DwbpL aggregates remained constant over time (Fig. 4B). We also found that reducing the concentration of both polymers in SCFM2 led to the dissolution of stacked aggregates, as expected, while it did not affect the formation of PAO1 DwbpL clumped aggregates (Fig. S4). These findings suggest that the loss of OSA prevents entropically derived stacked aggregate assembly, and the associated mechanism is independent of the polymer concentration and/or cell density. This is a well-studied manifestation of the aggregation of hydrophobic particles in colloidal environments (45,46).
Aggregate assembly of P. aeruginosa is not serotype specific but dependent on cell surface relative hydrophobicity. There are 20 serotypes of P. aeruginosa based on the glycosyl groups of OSA (38). As our findings were limited to PAO1 (serotype O5), we examined the aggregate formation of PA14 (serotype O10), PAK (serotype O6), and STO1 (serotype O1), which all differ in the oligosaccharide units of OSA (38). In all serotypes, we observed a stacked assembly similar to that of PAO1, but in an STO1 strain lacking OSA (DwbpM), we identified small clumped aggregates, with the Regulation of Aggregate Assembly Type in P. aeruginosa ® restoration of stacks when wbpM was complemented in trans ( Fig. 5A and B). These data confirm that the aggregate assembly type is not serotype specific.
Previously, it has been shown that the lack of OSA increases the hydrophobicity of the P. aeruginosa cell surface (30). To determine whether hydrophobicity correlated with aggregate type, we assessed the relative surface hydrophobicity of OSA 1 and OSA 2 strains. We found a significant increase in surface hydrophobicity in OSA mutants (Fig. 6A), which corresponded with a clumping aggregate phenotype (Fig. 2). The loss of OSA is a common adaptive trait of P. aeruginosa in CF airways, possibly leading to an increase in cell surface hydrophobicity that could alter the spatial organization of P. aeruginosa cells in CF airways. We therefore evaluated the relative hydrophobicity of 11 P. aeruginosa isolates collected from the expectorated sputa of 2 individuals with CF. We observed heterogeneity in the cell surface relative hydrophobicity of the CF isolates across the two patients (Fig. 6B).

DISCUSSION
Despite a large body of work showing how P. aeruginosa adapts to the CF lung environment (19-21, 33, 47, 48) and population heterogeneity during chronic infection (20,21,23,49,50), little is known about the impact of this intraspecific heterogeneity on the formation of P. aeruginosa aggregates. To test whether genetic heterogeneity impacts aggregation, we investigated the aggregate formation of selected evolved isolates from a previous 50-day biofilm evolution experiment of PAO1 (23) in SCFM2 (24). We found that (i) there are two distinct types of aggregate assemblies formed by P. aeruginosa in SCFM2, (ii) the OSA impacts the aggregate assembly type, and (iii) the loss of OSA and LPS core 11 increases the hydrophobicity of the bacterial surface, which prevents depletion aggregation.
Previously, it was shown that the aggregation of P. aeruginosa cells in a polymerrich environment can be due to depletion forces, where the reduction of the free energy by increased entropy of the whole system leads to stacked aggregation of bacterial cells (27). The change of the polymer electrostatic properties in the same study altered the aggregate assembly to bridging assembly, suggesting that the aggregate assembly type is driven by physical properties of the environment and that the biological properties of the cells assume little or no role in the aggregate assembly type (27). However, we observed two distinct types of aggregate assemblies by genetically diverse PAO1 isolates in SCFM2. Stacked aggregation is dependent on the cell density and polymer concentration, and the increase in entropy is the driving force behind these assemblies. In contrast, clumping assembly is driven by changes in the surface properties of P. aeruginosa cells. This indicates that although aggregate formation is influenced by physical forces, OSA strongly impacts the spatial organization enforced by the physical properties of the environment.
Other studies have suggested that during CF infection, polymers like mucin can disperse cells in established biofilms (51), although our work suggests that polymers are more likely to influence the spatial arrangement of the cells. The clumping aggregate assembly was not influenced by factors that have previously been shown to be involved in biofilm and aggregate formation, including lectins (42,52), QS (6,9,53), and EPS (54-57). While we found that stacked aggregates had a larger biovolume than clumped aggregates and that this can be explained by depletion aggregation, other explanations also exist. First, stacked aggregates might be more likely to support growth within the aggregate, leading to larger aggregates. Second, as clumped aggregates become larger, they may be more prone to disaggregation, leading to them never achieving a larger size even if there was cellular division occurring in the aggregate.
It remains to be determined whether stacked aggregates are found during human CF infection, although a recent study demonstrated that in CF airways, eDNA surrounds aggregates, suggesting that in CF lungs, eDNA could increase entropic force and form aggregates via depletion aggregation (58). It is noteworthy that stacked Regulation of Aggregate Assembly Type in P. aeruginosa ® aggregates form in other species of bacteria such as Rhizobium leguminosarum, which was dependent on LPS core and O-antigen (59). Furthermore, honeycomb structures of P. aeruginosa (multilayers of stacks) were shown to form in the crops of Drosophila melanogaster, highlighting that stacked aggregates are found in some types of infection (60).
By assessing the relative hydrophobicity of cells forming different aggregate types, we found that P. aeruginosa cells that form clumped aggregates have surfaces with higher relative hydrophobicity. In agreement with previous studies (29,(61)(62)(63), we showed that this was due to a loss of OSA and the exposure of uncapped LPS core. Regulating the aggregate assembly type by directly altering OSA expression levels in environments containing differential levels of cells and polymers (such as CF sputum) may allow cells to better resist environmental stressors such as the host immune response, antibiotics, or phage. The loss of OSA has been reported in P. aeruginosa strains isolated from CF sputum, suggesting that finding small, clumped aggregates in CF sputum and CF airways (64) might be directly due to the loss of OSA. As our work was performed in vitro, further studies are required to determine how OSA and surface hydrophobicity impact aggregate formation in P. aeruginosa isolates growing in CF lungs.
Overall, our findings highlight that the surface properties of P. aeruginosa cells determine how they form aggregates in environments with different physicochemical properties, providing potential benefits from social interactions with highly related cells. Our findings also highlight that changes in the cell surface properties may influence how aggregates form in other species of bacteria and provide explanations as to how different P. aeruginosa strains and species can stably coexist in microbiomes.

MATERIALS AND METHODS
Bacterial strains and culture conditions. We selected 7 evolved isolates of PAO1 from 50-dayevolved populations in SCFM (23). We transformed all P. aeruginosa strains used in this study with the pME6032:gfp plasmid (65) using electroporation (66). Briefly, to prepare electrocompetent P. aeruginosa cells, we grew the bacterial cells in LB broth overnight, washed the cultures grown overnight with a 300 mM sucrose solution at room temperature, and then resuspended the bacterial pellets in 1 ml of 300 mM sucrose. We then electroporated 50 ml of electrocompetent cells with 2 ml of purified plasmid and recovered the cells by the addition of 950 ml of LB broth and incubation at 37°C at 200 rpm for 30 min. We selected the transformed cells by plating out the electroporated bacteria onto LB agar plates supplemented with 200 mg/ml of tetracycline. We obtained the clinical isolates from the Emory CF@LANTA Research Center. Patients in this study were aged between 21 and 29 years at the time of collection of the sputum samples. This study was approved by the Institutional Review Boards (IRBs) at the Georgia Institute of Technology and Emory University Hospital. A list of all bacterial strains used in this study is available in Table S2 in the supplemental material.
Determining diversity in colony morphologies. To determine diversity in colony morphology in evolved populations (23), we used a Congo red-based agar medium (1% agar, 1Â M63 salts [3 g monobasic KHPO 4 , 7 g K 2 PO 4 , and 2 g NH 4 Á2SO 4 , with the pH adjusted to 7.4], 2.5 mM magnesium chloride, 0.4 mM calcium chloride, 0.1% Casamino Acids, 0.1% yeast extract, 40 mg/liter Congo red solution, 100 mM ferrous ammonium sulfate, and 0.4% glycerol) (67). We inoculated each evolved isolate in LB broth, incubated the colony for 6 h at 37°C at 200 rpm, and then spotted 10 ml of the culture onto Congo red agar plates. We incubated the plates at 37°C for 24 h and a further 4 days at room temperature.
Genomic DNA extraction and whole-genome sequencing. We plated each of the selected evolved isolates on LB agar plates, picked single colonies of each isolate, inoculated the colonies in 5 ml of SCFM (35), and incubated the isolates overnight at 37°C at 200 rpm. We extracted the genomic DNA using the Qiagen DNeasy blood and tissue kit. We prepared sequencing libraries using the Nextera XT protocol (Illumina) and sequenced the libraries in a 24-plex format on the Illumina MiSeq platform to obtain an approximate calculated level of coverage of 50Â for each evolved isolate. For single nucleotide polymorphism (SNP) calling, we used breseq analysis (consensus mode) (37,68,69) and compared the genetic variation in each evolved isolate to the PAO1 ancestral strain.
Image acquisition and analysis. For imaging aggregates in SCFM2 (24), we inoculated each bacterial isolate into tryptic soy broth (TSB) supplemented with 200 mg/ml of tetracycline and incubated the mixture at 37°C at 200 rpm overnight. We inoculated 50 ml of the culture grown overnight into 5 ml of SCFM and incubated the culture at 37°C at 200 rpm for 5 to 6 h until the cultures reached mid-log phase (optical density at 600 nm [OD 600 ] = 0.5). We then adjusted the OD 600 to 0.05 in 400 ml of freshly made SCFM2 containing 0.6 mg/ml of DNA and 5 mg/ml of mucin (8,24). We incubated the cultures at 37°C for 16 h in chamber slides (Lab-Tek) before image acquisition. We used an LSM880 confocal microscope equipped with a 63Â oil immersion lens for image acquisition, scanned the aggregates using a diode laser at 488 nm, and collected the fluorescence emission between 480 and 530 nm for image acquisition. For imaging the cells grown in SCFM, we adjusted the OD 600 of cells from mid-log-phase growth to 0.05 in 400 ml of freshly made SCFM. We incubated the cultures at 37°C for 16 h in chamber slides before image acquisition. For image analysis, we used Imaris 9.0.1 image analysis software to analyze the morphological properties of the aggregates and measured the surface area and volume of each aggregate using a surface model algorithm. We used the same parameters for particle and voxel sizes. We measured the aggregate volume and surface area in 10 images acquired for each strain in three independent experiments (over 1,000 aggregates were measured in total under each condition). For time course experiments, we used the same image acquisition parameters, using the time series option and imaging as Z stacks every 20 min for up to 10 h. To assess the role of bacterial cell density in aggregation, we adjusted the OD 600 to 0.1 in 400 ml of SCFM2 and imaged the cells every 20 min for 6 h. We prepared time series videos using the three-dimensional (3D) plug-in in Fiji (70) and Adobe Lightroom.
Gene deletion and complementation. We used standard genetic techniques for the construction of P. aeruginosa mutants. To delete ssg, rmd, wbpL, wbpM, waaL, wzy, wzz1, and wzz2, we PCR amplified 600-bp DNA sequences flanking the open reading frame of each gene using Q5 DNA polymerase (New England BioLabs). We then cloned these sequences into EcoRI-XbaI-digested pEXG2 by Gibson assembly using NEBuilder HiFi assembly master mix (New England BioLabs) and transformed them into Escherichia coli S17 lpir. We verified cloned inserts by colony PCR and Sanger sequencing (Eurofins Genomics). We introduced the deletion constructs into PAO1 by electroporation and selected strains carrying singlecrossover insertions of the deletion constructs on LB agar plates supplemented with 100 mg/ml gentamicin. We cultured gentamicin-resistant colonies in LB without antibiotic and plated the colonies on LB agar plates with 0.25% NaCl and 5% sucrose. We then selected sucrose-resistant colonies, screened them for gentamicin sensitivity to ensure the loss of the pEXG2 construct, and assessed them for the desired gene deletion by colony PCR and Sanger sequencing of the PCR product. For ssg complementation, we PCR amplified the ssg coding sequence and 100 bp of upstream sequence (including the ssg native promoter) using Q5 DNA polymerase (New England BioLabs). We cloned this 1,057-bp product into KpnI-BamHI-digested pUC18T-miniTn7-Gent by Gibson assembly using NEBuilder HiFi assembly master mix (New England BioLabs) and transformed it into E. coli S17 lpir. We verified the cloned insert by colony PCR and Sanger sequencing (Eurofins Genomics). We cotransformed the complementation construct with the Tn7 helper plasmid pTNS3 into PAO1 Dssg, evolved isolates by electroporation, and selected isolates on LB agar plates supplemented with 100 mg/ml gentamicin. We verified the strains for ssg 1 complementation by colony PCR and for the loss of the pUC18-miniTn7-Gent vector and pTNS3 by screening for carbenicillin sensitivity.
LPS extraction. We isolated bacterial lipopolysaccharide by the hot phenol extraction method (71). Briefly, we pelleted 5 ml of cultures of PAO1 and PAO1-derived strains grown overnight in LB broth by centrifugation for 10 min at 4,200 Â g. We resuspended the pellets in 200 ml 1Â SDS buffer (2% b-mercaptoethanol [BME], 2% SDS, 10% glycerol, 50 mM Tris-HCl [pH 6.8]) and incubated them at 99°C for 15 min. Next, we added 5 ml of 20 mg/ml proteinase K (Sigma) to each tube and incubated the cell lysates at 59°C for 3 h. Next, we added 200 ml of ice-cold Tris-saturated phenol to each sample, vortexed the mixture for 10 min, added 1 ml diethyl-ether, and vortexed the mixture for a further 10 s. We centrifuged the samples for 10 min at 16,000 Â g and extracted the bottom layer. We performed a second Regulation of Aggregate Assembly Type in P. aeruginosa ® extraction with phenol and diethyl-ether as described above. We mixed an equal volume of the extracted LPS samples with an equal volume of 2Â SDS buffer and electrophoresed 10 ml of each sample on Novex 4 to 20% polyacrylamide gradient gels (Thermo Fisher) in Tris-glycine-SDS buffer. Following electrophoresis, we visualized LPS with a ProQ Emerald lipopolysaccharide staining kit (Thermo Fisher).
Assessing cell surface hydrophobicity. To assess the levels of cell surface hydrophobicity, we used hydrophobic interaction chromatography (29). Briefly, we grew bacterial cells for 6 to 8 h at 37°C at 200 rpm to reach mid-log phase. We harvested the cells, washed the cells three times with ice-cold 3 M NaCl (pH 7), and resuspended the cells in 3 M NaCl. We used octyl-Sepharose CL-4C beads (Sigma) to assess the interaction of hydrophobic cells with these beads compared to control Sepharose CL-4C beads (Sigma). We prepared bead columns by three washes of the beads with Milli-Q water and then three washes with 3 M NaCl (pH 7) (at 4°C). We then prepared 1-ml columns of both beads by using 3mm-diameter filter paper. We added 100 ml of the bacterial suspension and incubated the mixture at room temperature for 15 min. We measured the OD 450 of the flowthrough from each column. We calculated the relative hydrophobicity based on the ratio of the OD 450 octyl-Sepharose CL-C4 column flowthrough to that of the control column.
Statistical analysis. For statistical analysis of the aggregate volume and biovolume distribution, we used GraphPad Prism 8.0.
Data availability. The sequences in this study are available at the NCBI SRA database (accession number PRJNA702741).

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.