Growth media (Lysogeny broth, LB), polyethylene glycol MW 2,000 and 35,000 Da, and antibiotics were purchased from Sigma. Strains and their sources are listed in Table 1 .

For PEG-induced depletion aggregation, bacteria were added at the indicated densities to either LB diluted 4:6 with distilled water or LB diluted with 50% PEG 35 kDa (w/vol) prepared in distilled water to ensure that nutrient concentrations were the same in dispersed and aggregated conditions. LB was diluted with water or 50% w/vol PEG 35 kDa for all experiments described unless noted otherwise. Cultures were then incubated on a roller (60 rpm) at 37°C unless indicated otherwise.

The indicated bacterial strains in Figure 2 were grown overnight in LB at 37°C with shaking. One hundred µl of overnight cultures were used to inoculate 3 ml of LB+PEG 35 kDa. After 18-h of growth, 100 µl of the indicated cultures were removed to a 1.5 ml tube containing 900 µl of either 1x PBS or PBS supplemented with 30% w/vol PEG 35 kDa and vortexed. PBS was used to facilitate imaging. Imaging was performed on 50 µl culture aliquots pre- and post-dilution using a Leica DM1000 LED microscope by spotting onto a glass slide. Aggregate dispersal was either scored i) by eye as either aggregated or dispersed by comparing to undiluted control cultures. Aggregate reversal assays shown in Figure 3 were performed as follows: P. aeruginosa PAO1 or ΔpelA/pslBCD/algD constitutively expressing YFP were grown overnight in LB at 37°C, spun down, and washed and resuspended in PBS at 1x10⁹ CFU/ml. Bacteria were added at a 1:1 ratio to either an 8% (w/vol) solution of mucin (porcine gastric mucin) and 4 mg/ml DNA (HMW, salmon sperm DNA) in PBS or a 30% (w/vol) solution of PEG 35k in PBS. Cultures were incubated on a roller (60 rpm, 37°C) for 15 or 120 minutes. Bacteria (50 ul) were diluted with 200 ul PBS and mixed by inverting the tube. 20ul of diluted (or undiluted) cultures were placed onto a slide and bacteria were imaged (YFP) such that aggregates were bright and had a distinct defining border from any background. Aggregate area was calculated using Velocity’s (Improvision) find object tool using intensity and a minimum aggregate size of 16.64 µm² (40 pixels). The mean are ± SD of at least 100 aggregates per replicate was then calculated.

S. aureus SH1000 (Horsburgh et al., 2002) and P. aeruginosa PAO1 (Holloway et al., 1979) were grown overnight at 37°C with shaking in LB broth. S. aureus and P. aeruginosa were pelleted and resuspended at 10⁸ CFU/ml in fresh LB broth. One hundred µl of each culture was added to 2 ml LB supplemented with either 30% w/vol PEG (35 kDa or 2 kDa) where indicated. Bacteria were grown in co-culture for 18 h and viable bacteria were enumerated by serial dilution and plating on LB plates. For experiments investigating the effects of quorum-regulated antimicrobials on S. aureus killing, P. aeruginosa PAO1 or ΔlasR/rhlR (Siehnel et al., 2010) were grown overnight at 37°C with shaking in 50 ml LB broth in a 250 ml flask. Bacteria were removed by centrifugation (10 minutes, 9,000 x g) and supernatants were filter sterilized using bottle top vacuum filters with 0.2 µm pore size (Millipore). PEG 2 kDa or 35 kDa was added to these supernatants to a final concentration of 30% w/vol where indicated. S. aureus was inoculated into P. aeruginosa supernatants at 10⁸ CFU/ml and cultured for 6 h at 37°C on a roller at 60 rpm. Viable S. aureus were enumerated by serial dilution and plating onto LB agar plates. To investigate TSS mediated killing, P. aeruginosa PAO1, ΔclpV1 (Mougous et al., 2006), and B. thailandensis E264 (Yu et al., 2006) were grown overnight at 37°C with shaking in LB broth. Bacteria were resuspended in fresh LB at 10⁹ CFU/ml. One hundred µl containing 1x10⁸ CFU P. aeruginosa PAO1 or ΔclpV1 and 100 µl containing 2.0x10⁷ CFU B. thailandensis were added to 800 µl LB or the indicated polymer solutions and incubated in co-culture for 24 h at 37°C on a roller at 60 rpm. Viable bacteria were enumerated by serial dilution and plating on LB plates. For fluorescent imaging of aggregates, strains PAO1 or ΔclpV1 constitutively expressing GFP (PAO1 attTn7::GFP (Choi and Schweizer, 2006),) were co-cultured with B. thailandensis E264 attTn7::mCherry for 24 hours (LeRoux et al., 2012).

S. aureus SH1000 carrying the fluorescent reporter pCE-SarA-mCherry (Malone et al., 2009), P. aeruginosa PAO1 attTn7::GFP, PAO1 attTn7::TFP (Zhao et al., 2013), PAO1 attTn7::YFP (Zhao et al., 2013), E. coli carrying pUCP18-mCherry (Irie et al., 2012), B. cenocepacia K56-2 attTn7::GFP (Varga et al., 2013) and B. thailandensis E264 attTn7::mCherry were co-cultured as indicated. Depletion aggregates assembled from dead bacteria were prepared by washing and resuspending overnight cultures of PAO1 YFP or PAO1 TFP in PBS at a concentration of 10⁹ CFU/ml. Formaldehyde (16%, Thermo) was added slowly to bacteria while vortexing to a final concentration of 4% vol/vol. Bacteria were allowed to fix for 30 minutes with constant mixing to prevent bacteria from clumping. Cells were then centrifuged for 10 minutes at 9,000 x g, washed twice with PBS, and resuspended in 1 ml PBS. Complete bacterial killing was confirmed by plating fixed bacteria on LB agar. One hundred µl of the indicated fixed strains were added to 2 ml PBS or PBS+30% PEG 35 kDa. Bacteria were incubated in a 37°C in a roller at 60 rpm. Samples were removed and visualized on a glass slide at the indicated times using a Zeiss LSM 510 confocal laser-scanning microscope. Image series were processed using Volocity (Improvision).

P. aeruginosa encodes three exopolysaccharides: Pel is a cationic polymer composed of partially acetylated N-acetylgalactosamine and N-acetylglucosamine (Jennings et al., 2015), Psl is a neutral polymer containing glucose, mannose, and rhamnose (Byrd et al., 2009), and alginate is a negatively-charged polymer composed of mannuronic and guluronic acid (Pedersen et al., 1992; Gibson et al., 2003).

We first tested wild-type P. aeruginosa PAO1 that encodes all three exopolysaccharides (Colvin et al., 2012; Wiens et al., 2014). As seen previously, wild-type P. aeruginosa exposed to the model polymer PEG (35 kDa) rapidly aggregated via the depletion mechanism, but disaggregated when polymers were diluted by adding PBS ( Figure 2A ). Adding PEG did not disperse aggregates, implicating polymer dilution rather than physical disruption in disaggregation ( Figure S1 ). Notably, wild-type P. aeruginosa aggregates held together by PEG exposure for as long as 18 hours disaggregated upon polymer dilution with PBS ( Figure 2A and Figure S2 ). Reversibility with dilution is a hallmark of depletion aggregation, as it is driven by a reduction in crowding effects of environmental polymers. Thus, in the conditions tested, wild-type P. aeruginosa did not activate bacterially-driven adhesive mechanisms to maintain aggregation.

Expression of exopolysaccharides is a key step in surface adherence and aggregation in surface-associated biofilms (Mann and Wozniak, 2012), so we reasoned that strains overproducing exopolysaccharides might remain aggregated after polymer dilution. To test this, we aggregated P. aeruginosa PAO1 overproducing alginate, Pel, or Psl, and investigated whether PBS dilution caused dispersal. Alginate overproduction was achieved via a mutation in an anti-sigma factor gene regulating alginate (PAO1 mucA22), and Pel or Psl overproduction was achieved using Pel or Psl genes on an inducible promoter (PAO1 P_BAD-Psl and PAO1 P_BAD-Pel).

After 18 hours of depletion aggregation, wild-type PAO1 and PAO1 overproducing alginate (PAO1 mucA22) readily dispersed after polymer dilution (i.e. PBS addition) ( Figures 2A, B ) whereas the strains over-expressing Pel and Psl did not ( Figures 2C, D and Figure S2 ). These findings indicate that cells aggregated by the depletion mechanism that have Pel and Psl expression induced can remain aggregated after depletion promoting-conditions are reversed.

To determine if the differential effects of alginate verses Pel and Psl on aggregate stability were generalizable to strains other than PAO1, we studied P. aeruginosa clinical isolates taken from people with CF. CF strains can evolve exopolysaccharide over-expression phenotypes (Smith et al., 2006). Pel or Psl overexpression is known to produce a rugose small-colony morphology (Starkey et al., 2009) whereas strains that over-produce alginate are mucoid (Pedersen et al., 1992).

All 10 P. aeruginosa CF clinical isolates tested that had a rugose colony morphology formed dilution-resistant aggregates in PEG ( Figure 2E and Table 2 ), whereas all (9/9) alginate-overproducing clinical isolates (i.e. mucoid strains) had a reversible aggregation phenotype ( Figure 2F and Table 2 ). These results with exopolysaccharide-overproducing clinical isolates are consistent with findings using engineered PAO1 strains (see above) and suggest that induced expression of Pel and Psl, but not alginate, enable aggregates formed by the depletion mechanism to remain intact after depletion-promoting conditions are reversed. Different chemical compositions or physical properties of bacterial exopolysaccharides such as charge may explain these differences.

Previous work shows that non-adsorbing biological polymers can produce depletion aggregation of bacteria like PEG does (Dorken et al., 2012; Secor et al., 2018). However, polymers found at infection sites can also induce biological responses in bacteria that have consequences for aggregation whereas PEG is considered to be relatively inert (Banerjee et al., 2012). For example, exposure to mucin can induce the expression of P. aeruginosa genes important in infection pathogenesis (Lory et al., 1996; Wang et al., 1996). Our previous work indicates that depletion aggregates formed in vitro by exposing P. aeruginosa to mixtures of mucin and DNA are comparable in size to aggregates formed by PEG (Secor et al., 2018). These observations led us to investigate whether depletion aggregates induced by biological polymers exhibit dispersal after polymer dilution, like aggregates induced by PEG.

To test this, we induced depletion aggregation using a mixture of mucin and DNA, which are major polymers in lung secretions (i.e. sputum) from people with CF. In these experiments we used concentrations found similar to those in vivo (mucin at 4% w/vol and DNA at 2 mg/ml) (Secor et al., 2018), and fluorescently-tagged bacterial strains because mucin/DNA mixtures are opaque (PEG is transparent), and assayed several hundred aggregates per condition. Similar to PEG, mucin and DNA mixtures aggregated wild-type PAO1 and PBS addition 15 minutes later caused aggregate dispersal ( Figure 3A ; Figure S3 ). However, when we extended the period of polymer aggregation to 120 minutes, wild-type PAO1 that had been aggregated in mucin/DNA mixtures remained intact after PBS dilution and mixing by vortexing ( Figure 3B ; Figure S3 ), whereas those that had been aggregated in PEG dispersed ( Figure 3B ). These results suggest that the aggregates that survive dilution by PBS and mixing are stable and in a steady state.

Our finding that induced expression of Pel and Psl makes depletion aggregates dilution-resistant led us to investigate whether self-produced exopolysaccharides mediated the dilution-resistant phenotype of aggregates induced by mucin and DNA. We tested this using PAO1 in which biosynthetic genes of all three exopolysaccharides had been inactivated (PAO1ΔpelA/pslBCD/algD) and found that the mutant lacking exopolysaccharides genes dispersed upon dilution with PBS regardless of whether aggregation was induced by PEG or the mucin/DNA mixture ( Figures 3C, D ). Collectively, these results suggest that depletion aggregation can actuate cell-cell adhesion by some P. aeruginosa exopolysaccharides, and that depletion aggregation by polymers present at infection sites can initiate the formation of aggregates that remain intact after depletion-mediating conditions are reversed provided exopolysaccharide genes are intact.

Theory predicts that bacteria aggregated by the depletion mechanism will be arranged to minimize the amount of volume occupied (Poon, 2002), as efficient packing will increase the space available for polymers and the concomitant entropy gains. This effect should cause bacteria with similar shapes to be arranged together, and bacteria with different shapes to separate, unless other external forces or bacterial activity intervene. To test this hypothesis, we mixed P. aeruginosa, Burkholderia cenocepacia (rod), Escherichia coli (rod), and Staphylococcus aureus (a coccus) bearing different florescent labels in various combinations in PEG and examined species distribution by microscopy.

Polymer-mediated depletion aggregation caused cocci-shaped species (S. aureus) to segregate from rods (P. aeruginosa and B. cenocepacia). Two patterns of segregation were observed. In some cases, entire aggregates appeared to be composed of a single species (i.e either rods or cocci) without an appreciable presence of the differently shaped species ( Figure 4A ). In other cases, sections of mixed-species aggregates were composed primarily of either the rod or cocci-shaped species, as shown with S. aureus and B. cenocepacia ( Figure 4B ). Similar results were seen using mixtures of formalin-killed P. aeruginosa and S. aureus, and with P. aeruginosa mixed with 2 µm diameter spherical beads similarly sized as S. aureus ( Figures S4A, B ). Thus, bacterial activity is not required for species segregation under the conditions tested.

In contrast, depletion aggregation caused bacteria with similar cell shapes (i.e. differentially labeled P. aeruginosa with P. aeruginosa, or P. aeruginosa with E. coli) to intermix ( Figures 4C, D ). These experiments, along with previous work using inert particles (Adams et al., 1998), show that physical forces mediating depletion aggregation cause like-shaped bacteria to intermix, and differently shaped bacteria to separate. The physical arrangement of bacterial species in aggregates can affect competitive and cooperative interactions (see below).

Our finding that depletion aggregation can determine the physical arrangement of species within aggregates led us to investigate its effects on interspecies interactions. P. aeruginosa and S. aureus are often co-isolated from CF airways (Harrison, 2007; Hauser et al., 2011) and wounds (Kirketerp-Moller et al., 2008; DeLeon et al., 2014) for long durations. However, in laboratory co-cultures, P. aeruginosa rapidly inhibits S. aureus by quorum-regulated antimicrobials such as rhamnolipids, hydrogen cyanide, phenazines, quinolones, and others (Mavrodi et al., 2001; Deziel et al., 2004; Mashburn et al., 2005; Palmer et al., 2005; Schuster and Greenberg, 2006). Because aggregation can increase antimicrobial tolerance (Haaber et al., 2012; Staudinger et al., 2014), we hypothesized that depletion aggregation could enhance the ability of S. aureus to co-exist with P. aeruginosa.

Similar to previous studies (Mavrodi et al., 2001; Deziel et al., 2004; Mashburn et al., 2005; Palmer et al., 2005; Schuster and Greenberg, 2006), we found that wild-type P. aeruginosa severely inhibited S. aureus in non-aggregated broth co-cultures ( Figure 5A ), and inhibition was diminished if quorum sensing was genetically inactivated (i.e. using ΔlasI/rhlI PAO1) ( Figure 5A , compare white bars). However, in co-cultures where PEG or mucin/DNA was used to induce depletion aggregation, the competitive index of wild-type P. aeruginosa over S. aureus was reduced by greater than 10-fold ( Figure 5A , gray and black bars).

Previous work indicating that depletion aggregation caused marked tolerance of P. aeruginosa to pharmaceutical antibiotics (Secor et al., 2018) led us to investigate whether depletion aggregation could cause S. aureus to become insensitive to antimicrobials produced by P. aeruginosa. We tested this by exposing dispersed and depletion-aggregated S. aureus to filter-sterilized P. aeruginosa planktonic culture supernatants. Supernatants from wild-type P. aeruginosa killed ~10-fold more dispersed S. aureus than aggregated S. aureus ( Figure 5B ), whereas supernatants from P. aeruginosa ΔlasI/rhlI did not kill dispersed or aggregated S. aureus ( Figure 5C ). Control experiments indicate that PEG did not diminish the antimicrobial activity of wild-type P. aeruginosa supernatants ( Figure S5 ). These results suggest that depletion aggregation may promote co-existence of P. aeruginosa and S. aureus by enhancing S. aureus tolerance to quorum-regulated antimicrobials secreted by P. aeruginosa. It is also possible that decreased production of antimicrobial factors by aggregated P. aeruginosa contributes to species co-existence in aggregates.

In addition to secreted factors, P. aeruginosa and other bacteria also possess competitive mechanisms that depend upon direct cell-to-cell contact. One mechanism is type VI secretion (TSS) in which a needle-like apparatus delivers toxins and effectors into neighboring cells (Mougous et al., 2006). Our finding that depletion aggregation causes like-shaped bacterial cells to intermix in aggregates led us to hypothesize that depletion aggregation could promote TSS-mediated bacterial antagonism.

To test this, we mixed P. aeruginosa (which is capable of TSS antagonism) with Burkholderia thailandensis, a TSS-susceptible rod-shaped Gram-negative bacterium (LeRoux et al., 2012) at a 5:1 ratio following previously established protocols (LeRoux et al., 2015). In dispersed conditions, no P. aeruginosa-B. thailandensis antagonism was apparent over 24 hours, as the ratio of P. aeruginosa to B. thailandensis remained unchanged at 5:1 ( Figure 6A ). In contrast, P. aeruginosa outcompeted B. thailandensis in depletion aggregates as measured by viable counts ( Figure 6A ) and visually assessing differentially-labeled species ( Figure 6B ). Notably the aggregation-induced competitive advantage of P. aeruginosa was eliminated by genetically inactivating TSS [i.e. PAO1 ΔclpV1 ( Figure 6C )]. The reduced fluorescent signal of B. thailandensis could be due to cell death or inhibition. Taken together, these results demonstrate that depletion aggregation can facilitate contact-dependent mechanisms of bacterial antagonism.