Bacterial biofilms are aggregates of bacteria enmeshed in a self-produced extracellular matrix (ECM) and represent the primary mode of microbial life. 1,2 This community lifestyle protects bacteria against environmental stresses such as desiccation, UV radiation, disinfectants, mechanical stress, starvation, protozoan grazing, salinity, and metal exposure. 3 Biofilms are ubiquitous in nature and play integral roles in the biogeochemistry of ecosystems through carbon and nitrogen fixation, for example. Humans utilize biofilms to produce foods (kimchi and kombucha), to generate energy (microbial fuel cells), 4 and to clean wastewater (bioremediation). Unfortunately, pathogenic bacteria also form biofilms, and these are often associated with serious and chronic infections. Currently available antibiotics often fail to eradicate biofilms, necessitating multiple antibiotic treatment regimens that drive the evolution of resistant pathogens and exhaustion of last-resort antibiotics. 5,6 Understanding biofilm structure, assembly, function, and pathogenicity will help to best utilize beneficial biofilms and to combat biofilmassociated infectious diseases.
The biofilm ECM usually consists of a complex mixture of biomolecules that can include proteins, polysaccharides, lipids and/or extracellular DNA. E. coli and Salmonella species form remarkable biofilm assemblies that recruit two fascinating types of biopolymers: curli, which are functional amyloid fibers, and phosphoethanolamine (pEtN) cellulose, a recently discovered chemically modified form of cellulose. [7][8][9][10][11][12][13] Both species produce striking wrinkled macrocolony biofilm architectures when grown on agar and form pellicle biofilms at the air-liquid interface. Biofilm morphologies of multiple human and environmental isolates of E. coli and Salmonella have been extensively documented since the identification of curli as adhesive fibers in E. coli in 1989. 7 Macrocolonies from growth on Congo red (CR)-supplemented agar have been scored according to distinctions in their coloring and textured appearance, e.g. the red dry and rough (rdar) and brown dry and rough (bdar) phenotypes. [13][14][15][16][17][18] Dyes, such as Calcofluor, are also widely used to qualitatively determine whether cellulosic material is produced by strains, and a new methodology using CR fluorescence distinguishes between modified and unmodified cellulose. 19 The investigation of more quantitative parameters of ECM composition and architecture, on the other hand, typically poses a larger challenge to biochemical analysis due to the insolubility and complexity of ECM assemblies. The combination of molecules present, such as insoluble polysaccharides and proteins, often requires harsh methods for only partial liberation of components and can alter structures, preventing a quantitative determination of chemical composition. We previously addressed this challenge and developed an integrated approach to defining ECM composition that combined new ECM isolation protocols with protein analysis, electron microscopy, and solid-state NMR spectroscopy. Solid-state NMR is nonperturbative and uniquely suited to the study of insoluble materials, cellular assemblies, and even intact tissues or organisms. [20][21][22] For 13 C NMR analysis, samples can be prepared without the use of isotopic labeling, simply observing 13 C spins at natural abundance levels, which permits analyses of samples grown from any nutrient medium. In work with curli-associated biofilms formed by a uropathogenic E. coli strain, UTI89, inspection of the extracted ECM by electron microscopy revealed remarkably woven basket-like structures that surround bacteria. 23 We developed a bottom-up sum-of-the-parts cross-polarization magic-angle spinning (CPMAS) solid-state NMR approach to determine the composition of the ECM. Our analysis of the insoluble ECM assemblies extracted from UTI89, isolated curli fibers, and the purified cellulosic component revealed that the insoluble ECM consisted of two major insoluble components in a defined ratio: 85% curli by mass with the other 15% attributed to the cellulosic component. 23 The 13 C CPMAS NMR spectrum of the purified cellulosic component (from cells that do not produce curli) had additional unanticipated contributions that distinguished it from standard cellulose. Indeed, E. coli and other organisms, including Salmonella enterica, produce pEtN cellulose (Figures 1A and 1B), in which approximately half of the glucose units are modified by pEtN. 9 Cellulose is the most abundant biopolymer on earth, and this presented the first definitive experimental discovery of a chemically modified cellulose produced in nature. We also determined the genetic and molecular basis for installation of the pEtN modification.
The identification of pEtN cellulose follows from decades of research on the genetic, molecular, and structural basis for bacterial cellulose assembly, regulation, and function in biofilm formation. 11,24 Two divergent operons in E. coli encode the cellulose synthesis machinery, yhjR-bcsQABZC and bcsEFG (Figure 1C). The proteins BcsA and BcsB form a transmembrane complex that polymerizes glucose and transports the growing cellulose strand into the periplasm. [25][26][27] BcsC forms a pore in the outer membrane to export cellulose outside the cell. 25 BcsQ is required for cellulose synthesis; 28 and BcsZ has endoglucanase activity and is required for maximal cellulose synthesis. 29,30 The bcsEFG operon is found primarily in gamma and beta proteobacteria, 25 and these genes are considered essential for cellulose synthesis in Salmonella. 31 We uncovered the function of BcsG as a transmembrane protein that functions as a pEtN transferase. 9 A panel of bacterial two-hybrid assays identified protein-protein interactions between BcsE and BcsF, BcsF and BcsG, and BcsG and BcsA, suggesting the presence of a novel regulatory pathway for pEtN cellulose synthesis. 9 BcsE is known to interact with c-di-GMP, 32 which also positively regulates cellulose synthesis by the BcsA-BcsB machinery. Recently, two groups reported crystal structures of the periplasmic C-terminal domain of BcsG and confirmed its catalytic ability to remove pEtN from phospholipid head groups 18 and to add pEtN to a carbohydrate substrate. 33 Curli biosynthesis involves another set of divergently transcribed operons encoding the curli synthesis genes (csg) -the csgBAC and csgDEFG operons [34][35][36] (Figure 1D). CsgA and CsgB are secreted outside the cell. CsgB serves as a nucleator protein and promotes the assembly of CsgA into amyloid fibers that are often closely cell-associated, although not covalently attached to cell surface structures. CsgG is an outer membrane lipoprotein that oligomerizes to form a secretion pore for the export of CsgA, CsgB, and CsgF. [37][38][39][40] CsgE and CsgF facilitate the transport and secretion of CsgB and CgsA and the proper assembly of amyloid fibers at the cell surface. 8,[39][40][41][42][43] CsgD is a master regulator of both curli and cellulose biosynthesis. 14,15 Curli promote adhesion to mammalian [44][45][46] and plant cells, 17,47,48 promote biofilm formation, 7,17 and are implicated in pathogenicity. 17,49 Curli are also antigenic and elicit an immune response. The cellulosic polysaccharide has been demonstrated to reduce curli-associated antigenicity, perhaps by masking curli recognition by the immune system. 50,51 We recently revealed an additional molecular role for pEtN cellulose in strengthening the association of curli at the bacterial cell surface and enhancing curli-mediated adhesion of E. coli to bladder epithelial cells. 44 Questions abound regarding E. coli's biofilm assembly principles. How do E. coli utilize curli and pEtN cellulose to weave the intricate basket-like structures that encase individual cells, 23 and how do they further enmesh the community? How do these polymeric assemblies influence functions such as adhesion to host cells and abiotic surfaces? Indeed, the two ECM components create a mechanically robust ECM scaffold that enables remarkable, cohesive macrocolony biofilm architectures. 9,23 The pEtN modification on cellulose is essential for the hallmark wrinkled macrocolonies of E. coli and Salmonella grown on agar 9,18 and for the elaboration of long continuous cellulosic filaments in E. coli macrocolonies. 9 Does the pEtN modification on cellulose influence biofilm formation at the liquid-air interface (i.e. pellicle formation) or plastic-associated biofilm formation? Might the ratio of curli and pEtN cellulose produced by a bacterium impact biofilm phenotypes and properties?
Here, we report results from the sum-of-the-parts 13 C CPMAS NMR analysis to define the curli-to-pEtN cellulose ratio in the ECM of the E. coli laboratory K12 strain, AR3110. We compare and contrast this compositional analysis and associated biofilm phenotypes for AR3110 and the clinical isolate, UTI89. We examine these two strains and a panel of mutants in three biofilm models: macrocolonies on agar, pellicles at the liquid-air interface, and biomass adherence on plastic. We propose structure-function relationships with connections between the molecular components and physico-chemical implications for adhesion, cohesion, and biofilm phenotypes.
Strains used in this study include uropathogenic strain UTI89 (O18:K1:H7) 52 and mutants UTI89∆csgA, 53 UTI89∆bcsA, 54 UTI89∆bcsG (this study), UTI89∆bcsG∆csgA (this study), UTI89∆bcsA∆csgA (this study). Nonpathogenic, K-12 derived lab strains W3110, 55 MC4100 56,57 and AR3110 11 and its mutants, AR3110∆csgBA, 11 AR3110∆bcsA∆csgB, 11 AR3110∆bcsG∆csBA, 9 AR3110∆bcsG, 9 and AR3110∆bcsA 11 are also used. Bacterial stock cultures were prepared on LB agar and single colonies were inoculated into either 4 mL of YESCA broth (yeast extract 1 g/L, casamino acids 10 g/L) or LB broth (yeast extract 5 g/L, tryptone 10 g/L, NaCl 10 g/L) and grown at 37°C. Deletion mutants of UTI89∆bcsG, UTI89∆bcsG∆csgA, and UTI89∆bcsA∆csgA were generated using the techniques described by Datsenko and Wanner 58 and Murphy and Capellone. 59 Nonpolar deletion of the gene bcsG in the strain UTI89 was performed using λ Red recombinase-mediated recombination as described previously 59 using the primers bcsG_3KO (TTACTGCGGGTAAGGCACCCAGTCGCCGCCGTTCAGGCGAACGTACGGTTG-TGTAGGCTGGAGCTGCTTC) and bcsG_5KO (ATGACTCAATTTACGCAAAATACCGCCATGCCTTCTTC-CCTCTGGCAATAATGGGAATTAGCCATGGTCC). The mutant was verified using the primers that flank the bcsG sequence bcsG_upstream (AAAGCCAGGGCAACCAAAAA) and bcsG_downstream (GAACGAAAAAGGCCGCAGAG).
Nonpolar deletion of the gene csgA in the strains UTI89∆bcsG and UTI89∆bcsA using λ Red recombinase-mediated recombination. Primers used were csgA_3KO (TTAGTACTGATGAGC-GGTCGCGTTGTTACCAAAGCCAACCTGAGTGACGTGTGTAGGCTGGAGCTGCTTC) and csgA_5KO (ATGAAACTTTTAAAAGTAGCAGCAATTGCAGCAATCGTATTCTCTGGTAGCATATG AATATCCTCCTTAG). The mutant was verified using the primers flanking csgA: csgA_20_upstream (CAATCCGATGGGGGTTTTAC) and csgA_30_downstream (GCGCCCTGTTTCTGTAATAC).
ECM isolation was performed as in McCrate et al. 23 Non-isotopically labeled NMR samples of ECM and cellulosic materials were prepared from AR3110, UTI89, and UTI89ΔcsgA grown on YESCA agar (1 g/L yeast extract, 10 g/L casamino acids, 20 g/L agar) supplemented with 25 µg/mL Congo red. Curli were isolated from MC4100 grown on YESCA agar without Congo red. Bacterial cells were grown at 26°C for 60 hours, scraped into a 10 mM Tris, pH 7.4, and sheared using an OmniMixer homogenizer for five cycles of one-minute shear and two-minute rest. Cells were pelleted by centrifugation at 5,000 g at 4°C for 10 minutes and washed and pelleted two additional times in the Tris buffer. 5M NaCl was added to the resulting supernatants to achieve a final concentration of 170 mM NaCl. The ECM or cellulosic material was then pelleted by centrifugation at 13,000 g for one hour. Pellets were subjected to 4% SDS treatment overnight and subsequently washed to remove all SDS. The samples were lyophilized.
The 13 C CPMAS experiments 60 were performed using an 89-mm wide-bore Varian magnet at 11.7T (500.92 MHz for 1 H, 125.52 MHz for 13 C), a Varian console with VNMRJ software and a home-built four-frequency transmission-line probe with a 13.66-mm-long, 6-mm inner diameter sample coil, and a Revolution NMR MAS Vespel stator. Samples were spun in thin-wall 5mm outer diameter zirconia rotors (Revolution NMR, LLC) at 7143±2 Hz using a Varian MAS control unit. 13 C CPMAS spectra were obtained with a contact time of 1.5 ms, with spin-echo detection with a π -pulses of 10 µs, and a recycle delay of 2.0 s. Field strength for 13 C crosspolarization was 50 kHz. A 10% linear ramp centered at 57 kHz was employed for 1 H. 1 H TPPM decoupling was performed at 72 kHz during acquisition. The 13 C spectra were referenced to TMS as 0.0 ppm, which was determined relative to an adamantine standard at 38.5 ppm. The free induction decay was multiplied by a decaying exponential function with line broadening of 80 Hz.
Bacteria were grown on YESCA agar. Samples were prepared by centrifuging the appropriate volume of the bacterial culture (to give OD600 1.0 when resuspended to 1 mL) at 10,000g for 10 minutes to pellet the cells, followed by aspiration of the supernatant. Samples were treated with 100 µL >98% formic acid to dissociate CsgA. Formic acid was removed with vacuum centrifugation, and samples were resuspended in 200 µL of sodium dodecyl sulfatepolyacrylamide gel electrophoresis (SDS-PAGE) loading buffer. Samples were heated at 95˚C for 10 minutes, allowed to cool, and protein gel electrophoresis was carried out using 12% SDS-PAGE gels (NuPAGE). Proteins were transferred onto 0.2 µm nitrocellulose transfer membranes (GE Healthcare) at 47V for 3 hours at 4°C, and then blocked overnight at 4°C. The polyclonal rabbit antiserum to a peptide sequence from CsgA (CGNGADVGQGSDDSS) was used as the primary antibody, and an Alexa Fluor 488-conjugated goat anti-rabbit antibody (Molecular Probes) was used as the secondary antibody. Both primary and secondary antibodies were diluted 1:10,000.
Overnight cultures were made in LB broth. Macrocolony biofilm formation was initiated by spotting10 𝜇l of overnight bacterial culture onto YESCA agar plates. Plates were incubated at 26 °C. Colony morphology for each strain was observed and imaged after 48 h and 72 h of growth. Images were taken using the Leica S6D Stereomicroscope and Leica MC Camera. Images were placed in PowerPoint and adjusted for brightness and contrast.
Overnight cultures were made in LB broth. Pellicle formation was initiated by inoculating 2 𝜇L of an overnight culture into 2 mL YESCA broth in 24-well-plate wells and incubated at 26 °C for 48 or 72 hours. Pellicle formation was inspected visually and assessed by perturbation with a pipet tip. Images were taken using the Leica S6D Stereomicroscope and Leica MC Camera. Images were placed in PowerPoint and adjusted for brightness and contrast.
We performed a modified version of the CV assay 61 as in Boruki et al. 62 Overnight cultures were prepared in 4 mL of LB broth shaking at 200 rpm at 37°C for 14 hours . The 96 well PVC plates were irradiated under UV light in a biosafety cabinet for 30 minutes to sterilize. The overnight cultures were diluted 1:1000 in YESCA broth, and 150 𝜇L of culture was added to each well. The plates were incubated at 26°C for 48 hours. Then, the media and unattached cells were removed. The wells of the PVC plate were washed three times with 170 𝜇L water. The PVC plate dried in the inverted position for 30 minutes at room temperature. Then, 170 𝜇L of 0.1% crystal violet was added to each well and incubated for 45 minutes at room temperature. The crystal violet was removed, and each well is washed three times with 170 𝜇L of water. The CV was then extracted with 170 𝜇L of 95% ethanol for 30 minutes at 4°C. 100𝜇L of each well was transferred to a sterile polystyrene microtiter plate, and CV absorbance was measured at 595 nm in a microplate reader (SpectraMax M5, Molecular Devices).
Diverse E. coli strains producing curli and pEtN cellulose exhibit similarities in biofilm phenotypes, including the production of highly wrinkled macrocolony morphologies. Yet, strains display reproducible differences in their macrocolony architectures and in other biofilm phenotypes, including pellicle formation and plastic-associated biofilm formation. We hypothesized that strains could differ in the amounts of curli and pEtN cellulose they produce and in the ratio of these matrix polymers in the ECM. We sought to identify molecular correlates for differences in biofilm phenotypes by investigating UTI89, AR3110, and panels of relevant matrix mutants in each strain.
We previously introduced CPMAS solid-state NMR approaches to enable quantification of the composition of the insoluble curli-pEtN cellulose-containing ECM materials and discovered that the UTI89 ECM contains 85% curli and 15% pEtN cellulose by mass (or a 6:1 mass ratio of curli and pEtN cellulose). 9,23,63 AR3110 is another well-studied E. coli strain generated by Hengge and coworkers. 11 It is derived from the laboratory K-12 strain, W3110, through the correction of a domesticating SNP in bcsQ to restore expression of the bcsQ gene and thus enable pEtN cellulose synthesis. 11,28 AR3110 and UTI89 each exhibit wrinkling phenotypes with visually striking buckling up of biofilms, 11,53 yet display reproducibly different macrocolony morphologies as shown here for growth on YESCA (Yeast extract/CASamino acids) nutrient agar (Figure 2A). While UTI89 displays a highly entangled network of wrinkles spanning its entire surface, AR3110 elaborates dense wrinkles in the center of the macrocolony with separated and defined ridges extending radially from the center.
To compare curli production by UTI89 and AR3110, we employed Western blot analysis to detect cell-associated CsgA, the major subunit of curli fibers, for each strain when grown on YESCA agar and normalized by cell density (Figure 2B). CsgA Western blot results from previous studies of clinical isolates have revealed that some E. coli isolates can exhibit variation in the extent of curli production, while also displaying variation in macrocolony wrinkling. 64 In this case, CsgA levels for UTI89 were detectably slightly higher than AR3110 when harvested from YESCA agar after 48 hours of growth. Curli production in these strains is also comparable to that in MC4100 and W3110, widely studied curli-producing/cellulose-deficient strains (Figure 2B).
We next sought to measure the relative amounts of curli and pEtN cellulose in the extractable ECM of AR3110 and to compare this with UTI89 using the solid-state CPMAS NMR approach. The ECM from AR3110 was isolated using the previously reported ECM extraction protocol for cells grown on YESCA agar supplemented with Congo red (CR), a commonly used indicator dye that binds to both curli and cellulosic polymers 63,[65][66][67][68][69] and aids in precipitation of ECM material. 23 We obtained the AR3110 ECM spectrum and compared it to the UTI89 ECM, normalized to the prominent peak near 175 ppm that corresponds to the carbonyl carbons exclusively in curli. The AR3110 ECM clearly differed from UTI89 with added intensity in the regions associated with pEtN cellulose, the anomeric carbon centered at 103 ppm and the other main sugar carbons between 60-90 ppm (Figure 2C). CR which is not removed by SDS in the ECM isolation specifically remains ECM-associated due to interactions with pEtN cellulose, as previously determined. 63 The significant increase in CR carbon intensity parallels the increase in pEtN cellulose, further supporting the difference in AR3110 ECM composition. We next performed a sum-of-the-parts analysis to quantify the compositional differences observed in the AR3110 ECM. As for UTI89, the spectral sum of purified curli and pEtN cellulose could recapitulate the 13 C CPMAS NMR spectrum of the isolated ECM from AR3110 (Figure 2D). However, the results of the spectral sum analysis for AR3110 revealed a major difference in the ratio of curli and pEtN cellulose that contribute to the ECM than for UTI89. The ECM spectral sum is obtained by normalizing each full individual spectrum according to its mass and then determining the scaling factor for each spectrum needed to recapitulate the intact ECM as described. 23 The analysis was previously validated with an additional sample containing a physical mixture of the two polymers in a defined ratio plus remaining CR in the sample. 23 Taking into account the CR contribution to the pEtN cellulose sample (~ half the mass), 23 the curli:cellulose:CR contributions in the AR3110 ECM are 60:20:20 Therefore, in terms of curli and pEtN cellulose, the AR3110 ECM contains approximately 75% curli and 25% pEtN cellulose, or a 3:1 ratio of curli to pEtN cellulose.
Thus, we discovered that the AR3110 ECM has higher pEtN cellulose content with respect to curli than the UTI89 ECM, which contains 15% pEtN cellulose by mass, i.e. a 6:1 ratio of curli to pEtN cellulose. The higher ratio of pEtN cellulose to curli content could plausibly impact function. For example, curli fibers are known to be antigenic, 49,51,70 and cellulose helps to mask curli's antigenicity. 50,71 A higher proportion of pEtN cellulose in the ECM could help dampen curli's antigenicity to better evade the immune system. Curli also mediate bacterial adhesion to mammalian and plant cells and abiotic surfaces. [44][45][46] Curlimediated adhesion of UTI89 to bladder epithelial cells is also stronger when curli are coproduced with pEtN cellulose, wherein pEtN cellulose appears to have a Velcro-like role in keeping curli cell-associated. 44 Different strains of E. coli may have optimized the curli-pEtN cellulose ratio for their specific environmental niche; or perhaps, a regulatory mechanism could exist to modulate the proportion of proteinaceous and polysaccharide ECM components according to changes in environmental signals. We performed multi-pronged biofilm analyses to begin to define and to directly compare biofilm behaviors in these two strains and relevant mutants.
The observed differences in macrocolony wrinkling patterns between the WT E. coli strains (Figure 2A) led us to examine the morphologies of a panel of mutants in AR3110 and UTI89 that are deficient in one or more of the insoluble ECM components. We previously reported that the pEtN modification of cellulose was required for typical macrocolony wrinkling in AR3110. 9 Here, we more closely examined macrocolony architecture in the bcsG mutant, producing unmodified cellulose, and sought to determine if the pEtN modification of cellulose was also required for UTI89 macrocolony wrinkling.
AR3110∆bcsG and UTI89∆bcsG, which produce unmodified cellulose, do not develop a full rugose morphology akin to the WT strains, but instead produce ridges in a concentric ring pattern at the outer edge of the colony (Figures 3, S1 and S2). Thus, the pEtN modification is essential for full macrocolony wrinkling in both AR3110 and UTI89. This is compatible with a model in which pEtN cellulose plays an important structural role in E. coli biofilms, likely through intermolecular interactions with curli and acting as a type of glue to keep curli cellassociated. 44 Macrocolonies of the curli deficient strains, AR3110∆csgBA and UTI89∆csgA, display a distinct and shallow puckering morphology, in agreement with previous reports. 9,11,54,72 We further observed that the "buckling up" architecture of curli deficient strains specifically requires the pEtN modification. AR3110∆csgBA∆bcsG 9 and UTI89∆csgA∆bcsG (this study) produce only unmodified cellulose and lose the puckered or textured appearance. Similarly and as anticipated, bacteria producing only curli in the absence of cellulose do not generate macrocolony wrinkling by 72 hours; 23,54,72,73 AR3110∆bcsA and UTI89∆bcsA are smoothsurfaced colonies that resemble AR3110∆csgBA∆bcsA and UTI89∆csgA∆bcsA (Figure 3).
Similar connections between ECM composition and agar colony morphology are even observed in Gram-positive microorganisms that elaborate an ECM from polysaccharide and amyloid or amyloid-like proteinaceous fibers. For example, Bacillus subtilis produces an exopolysaccharide (eps) and amyloid-like fibers composed of TasA protein subunits. The tasA mutant displays a puckered macrocolony texture with less robust wrinkling than the wild type, 74 similar to curli-deficient E. coli macrocolonies. Likewise, all wrinkling is abolished in the B. subtilis eps mutant and the smooth colony resembles the double mutant, lacking TasA and polysaccharides. 74 Thus, our results on E. coli, in conjunction with the Salmonella and B. subtilis comparisons, suggest the presence of specific structural roles attributed to amyloid or amyloidlike fibers and polysaccharides across bacterial species.
Unlike the solid-air interface, biofilm formation at the liquid-air interface, i.e. pellicle formation, is less commonly studied despite its prevalence in both natural and industrial settings. 75,76 Among different environmental parameters, high oxygen levels in air and nutrients in the liquid phase create a favorable environment for growth, 76,77 and this lifestyle may be advantageous for survival outside of the host. 78 For example, Salmonella pellicles better withstand exposure to a common disinfectant, sodium hypochlorite, than planktonic cells. 78 E. coli pellicle formation begins with flagellar-mediated motility to facilitate aggregation and surface association. Flagella are primarily detected in planktonic populations of cells beneath pellicles and not considered to be major structural components of resulting pellicle biofilms in E. coli. 12 This is consistent with distinct regulation of flagella and curli production, wherein cyclicdi-GMP activates the master biofilm regulator, CsgD, and upregulates curli production and suppresses flagella synthesis. 79,80 Flagellar mutants form macroscopic cell clumps 81 (described as rosette biofilms in E. coli by Hung et al. 12 ), but do not generate a typical mat pellicle structure, 82 consistent with a lack of motility and insufficient number of bacteria accessing the liquid-air interface.
Curli is required for pellicle formation in E. coli. 12,46,53,73,82 Small-molecule inhibitors, termed curlicides, have been introduced and are effective in inhibiting pellicle formation by UTI89. 53,83,84 In addition, increased curli content in UTI89 pellicles yields a mechanically more robust film with increased elasticity. 73,85 The cellulosic component contributes to cohesive integrity of UTI89 pellicles, 12 and one recent study of BcsG indicated that pEtN modification does contribute to pellicle integrity in AR3110, although in different growth conditions at 30°C. 33 Thus, we sought to address the question of how pellicle formation in UTI89 and AR3110 might differ, knowing that the composition of the isolated ECM above revealed variation in the ratio of pEtN cellulose and curli. Then, we explored the influence of the pEtN modification in AR3110 and UTI89 pellicle formation and evaluated bcsA mutants.
We first compared UTI89 and AR3110 pellicles formed through static growth in YESCA broth for 72 hours at 26°C. The AR3110 pellicle has a remarkable morphology with intricate spiral patterning (Figure 4). The UTI89 pellicle has a comparably plain and relatively unpatterned surface as observed in previous work. 12,53,73 Poking the AR3110 pellicle with a pipet tip cleanly removed it from the side of the well (Figure 4). Interestingly, the detached biomass sank into the broth and remained an intact sheet, demonstrating the cohesive nature of the pellicle (Figure 4). In contrast, UTI89 pellicles are more brittle than AR3110; perturbation with a pipet tip punctured the UTI89 pellicle and formed a hole, while the majority of the biomass remained attached to the plastic well (Figure 4). These physical observations of AR3110 and UTI89 pellicles correlate well with our quantitative molecular determination of curli and pEtN cellulose in the ECM. The AR3110 strain is associated with a higher pEtN cellulose-to-curli ratio than UTI89 (for ECM isolated from agar-grown cells), which is consistent with a more cohesive and less adhesive pellicle than UTI89, if ECM ratios are comparable in pellicles. Likewise, the greater proportion of curli in the UTI89 ECM corresponds to a more brittle pellicle with enhanced adhesion to plastic. As anticipated, UTI89∆bcsA∆csgA and AR3110∆bcsA∆csgBA do not form pellicles and simply grow planktonically in the broth, increasing turbidity (Figures 4 and S3). Collectively, we observed major differences in pellicle phenotypes for two strains which both produce curli and pEtN cellulose. These results emphasize that the simple presence or absence of the matrix polymers does not solely dictate pellicle phenotypes; the relative proportions of matrix components should also be considered in understanding function.
We next sought to determine whether the pEtN modification is specifically required for pellicle formation, with power in comparison of the two strains, AR3110 and UTI89. The entirely cellulose-deficient mutant, UTI89∆bcsA, can form a weak pellicle. 12 We observed that biomass during attempted UTI89∆bcsA pellicle formation remains attached to the sides of the well, consistent with curli-mediated adhesion to plastic, while the center of the pellicle breaks away after it begins to form and falls to the bottom (Figures 5 and S4). AR3110∆bcsA also results in a collapsed pellicle that sinks to the bottom of the well (Figures 5 and S3). However, its biomass cleanly breaks from the sides instead of failing at its center, suggesting that enhanced curli production in UTI89 may be associated with increased adhesiveness, even in the cellulosedeficient mutant (Figure 5). Intriguingly, the pEtN modification mutants, UTI89∆bcsG and AR3110∆bcsG, are both nearly identical to UTI89∆bcsA and AR3110∆bcsA that produce curli but completely lack cellulose. Thus, robust curli-associated pellicle generation requires the pEtN modification in AR3110 and UTI89.
Results of one final set of mutants yielded surprising results. AR3110∆csgBA∆bcsG displayed a unique floating pellicle that did not require attachment to the sides of the well (Figure 5 and S3). This indicates that unmodified cellulose, in the absence of curli, promotes formation of a buoyant biofilm. The phenotype was most pronounced in AR3110. The observed pellicles for UTI89∆csgA∆bcsG exhibited variation in phenotypes ranging from no surfaceassociated bacteria (Figure 5) to a weak surface-associated pellicle (Figure S4). Komagataeibacter xylinus is a bacterium widely studied and utilized for its production of highly crystalline cellulose. The tightly packed crystalline cellulose of K. xylinus is thought to contribute to the buoyancy of its pellicle. 86,87 K. xylinus mutants that produce less dense networks of cellulose do not stay afloat. This behavior is attributed to added space in the loose fibril network, enabling liquid to penetrate and cause pellicle sinking. 87 E. coli pEtN cellulose is noncrystalline, 9,28 and the pEtN modification adds mass, increases the width/size of the polymer, and could influence multiple intermolecular interaction opportunities either with other pEtN cellulose chains or other cell-surface biomolecules. It is plausible that the unmodified cellulose secreted in the absence of curli by AR3110∆bcsG∆csgBA may be able to form a denser network with tight hydrogen bonding common to cellulosic materials and allow the microbial mat to float, as observed in K. xylinus.
In summary, remarkable structural variations are exhibited by AR3110 and UTI89 pellicles. The combination of pEtN cellulose and curli is essential for robust pellicular development in both AR3110 and UTI89, and we observe that pEtN cellulose in pellicles seems to play a cohesive role similar to biofilms at the solid-air interface.
Finally, we examined biofilm formation at the solid-liquid interface. Plastic-associated biofilms are especially relevant when considering pathogenic biofilms such as those on catheters and implanted medical devices. Catheter associated urinary tract infections, in particular, account for 75% of all UTIs acquired in the hospital. 88 Thus, we aimed to understand how the individual and combined contributions of curli, pEtN cellulose, and/or unmodified cellulose affect plastic associated E. coli biofilm formation. AR3110, UTI89, and the five matrix mutants in each strain were evaluated for biomass accumulation on poly vinyl chloride (PVC) 96 well plates in YESCA broth for 48 hours at 26°C using the crystal violet assay to quantify biomass accumulation. 61 Cellulose production (both pEtN cellulose and unmodified cellulose) appeared dispensable for biomass accumulation in both strains. The cellulose-deficient mutants, AR3110∆bcsA and UTI89∆bcsA, and the pEtNdeficient mutants, AR3110∆bcsG and UTI89∆bcsG, have near wild type levels of measurable biomass attached to wells (Figure 6). This indicates that, unlike the solid-air and liquid-air biofilm models, neither the pEtN modification nor the cellulose macromolecule are required for plastic-associated biofilms. Instead, curli is the crucial ECM component for attachment to PVC in this static broth-based assay. Curli deficient E. coli mutants (UTI89∆csgA, UTI89∆bcsG∆csgA, UTI89∆bcsA∆csgA, AR3110∆csgBA, AR3110∆bcsG∆csgBA and AR3110∆bcsA∆csgBA) exhibit severely reduced crystal violet retention and do not adhere well to plastic (Figure 6), in agreement with previous work in some strains, including UTI89. 53,64,82,89- 91 Thus, to the extent to which plastic associated biofilm is evaluated in a standard biomass accumulation assay for cells grown in YESCA broth, biofilm formation is dominated by curli production in UTI89 and AR3110.
It is notable that AR3110 displays almost twice as much biomass accumulation as UTI89 and, importantly, the cellulose mutants in each strain also exhibit this strain-dependent difference in curli-dependent adhesion. We previously reported differences in biomass accumulation of curli-dependent adherence across a large panel of clinical isolates, with one isolate even exhibiting three times the CV retention as UTI89, 64 but this did not involve mutants of the isolates. Our detailed comparisons of AR3110 and UTI89 suggest that other cell-surface components, besides curli, contribute to the extent of plastic adherence in E. coli. One hypothesis is that lipopolysaccharide (LPS) differences could alter curli cell-association and hence biomass accumulation. Alternatively, strains may exhibit altered curli production when attached to plastic. Elucidating the factors that contribute to these differences among strains in curlidependent plastic association remain a point for future work.
A more complete understanding of how bacteria function uniquely in biofilm communities requires improved descriptions of matrix composition, cell-surface structures, genetic networks, and regulation mechanisms. Direct comparisons across strains and biofilm models will further our understanding of the functions of individual and combined ECM components. We investigated the ECM composition of AR3110 and determined the ratio of curli and pEtN cellulose of the isolated matrix. The extracted AR3110 ECM has a 3:1 mass ratio of curli-to-pEtN cellulose. This is twice as much cellulose relative to curli compared to the 6:1 curli-to-pEtN cellulose ratio of the UTI89 ECM. We present the first quantitative comparison of this type and reveal that the relative amount of curli and pEtN cellulose assembled in the ECM can quantitatively differ for biofilms of the same species. These discoveries provide fundamental parameters for future studies to determine whether individual E. coli strains regulate the composition of the ECM in response to environmental cues.
We fully explored biofilm phenotypes in these two strains and investigated potential correlations between composition and function. Indeed, we discovered variations, especially in macrocolony morphology and macroscopic pellicle patterning between AR3110 and UTI89. These differences are correlated with their altered prevalence of curli and pEtN cellulose. Future work will determine whether the increased pEtN cellulose content in the isolated ECM of AR3110 versus UTI89 is uniquely correlated with these striking phenotypic differences by examining additional strains or manipulating this ratio in other ways. We also expanded our understanding of plastic-associated biofilm formation in E. coli. Plastic-associated biofilm formation appeared to be independent of cellulose production, whether it is modified or unmodified. However, significant biomass accumulation on PVC required curli, which is in agreement with prior studies.
Collectively, our results further differentiate the roles of curli and pEtN cellulose. Curli appears to be more important for adhesive properties, enabling attachment of bacteria to plastic surfaces; while pEtN cellulose contributes more to cohesion, particularly as observed in pellicle phenotypes. These results provide an exciting foundation for new avenues of study and for identifying other roles of these matrix components, especially in protecting resident bacteria from external stresses and for colonization and persistence in the host and other environments. Figure 1. (A) Chemical structure of cellulose. (B) Representation of chemical structure of pEtN cellulose, here shown with one modified glucose unit adjacent to an unmodified glucose. (C) E. coli bacterial cellulose synthesis (bcs) gene operon schematic, where bcsG (blue) encodes the pEtN transferase. (D) E. coli curli synthesis genes (csg) operon schematic. 13 C CPMAS spectral comparison of isolated ECM from AR3110 and UTI89 reveals significantly enhanced pEtN cellulose content relative to curli in the AR3110 ECM. (D) A spectral sum analysis of pEtN cellulose and purified curli reveals that the pEtN cellulose content in AR3110 is approximately twice that in UTI89, consistent with the approximate doubling of cellulosic peaks apparent in the AR3110 to UTI89 overlay in panel C. Asterisks indicate carbonyl peak spinning sidebands resulting from magic-angle spinning. The sample sizes were 26 mg for AR3110 ECM, 134 mg for UTI89 ECM, 102 mg for curli and 77 mg for pEtN cellulose. AR3110 spectrum was acquired with 44,000 scans, and all others were obtained with 32,768 scans.