Pseudomonas aeruginosa forms diverse matrix-enclosed surface-associated multicellular assemblages (biofilms) that aid in its survival in a variety of environments. One such biofilm is the pellicle that forms at the air–liquid interface in standing cultures. We screened for transposon insertion mutants of P. aeruginosa PA14 that were unable to form pellicles. Analysis of these mutants led to the identification of seven adjacent genes, named pel genes, the products of which appear to be involved in the formation of the pellicle's extracellular matrix. In addition to being required for pellicle formation, the pel genes are also required for the formation of solid surface-associated biofilms. Sequence analyses predicted that three pel genes encode transmembrane proteins and that five pel genes have functional homologues involved in carbohydrate processing. Microscopic and macroscopic observations revealed that wild-type P. aeruginosa PA14 produces a cellulase-sensitive extracellular matrix able to bind Congo red; no extracellular matrix was produced by the pel mutants. A comparison of the carbohydrates produced by the wild-type strain and pel mutants suggested that glucose was a principal component of the matrix material. Together, these results suggest that the pel genes are responsible for the production of a glucose-rich matrix material required for the formation of biofilms by P. aeruginosa PA14.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Wherever multicellular assemblages form, be it in eukaryotes or prokaryotes, extracellular matrix production appears to be a common strategy to provide structural support and facilitate surface attachment (Mackie, 1985; Nieduszynski, 1985; Sutherland, 2001). In animals, connective tissue is composed largely of an extracellular matrix in which a network of collagen fibres is dispersed within a mixture of connective tissue polysaccharides (Vogel, 1994). In plant tissues, structure is conferred by a complex extracellular matrix composed of components such as cellulose fibres, pectin and glycoproteins (Mackie, 1985). In microbial biofilms, the constituent cells are embedded in matrices composed of diverse extracellular polymeric substances (Costerton et al., 1995; Sutherland, 2001).
Biofilm formation appears to be a widespread attribute of bacteria and may allow increased survival ability under stressful conditions such as low nutrients or antimicrobial treatments (Costerton et al., 1995; Mah and O’Toole, 2001). Pseudomonas aeruginosa, like most bacteria, forms different biofilms depending on the environment. Among these biofilms are the pellicles that form at the air–liquid interface of a standing culture, the solid surface-associated (SSA) submerged biofilms that form in flow cells and the colonies that form on agar plates. P. aeruginosa is one of the most studied microbes in the context of biofilms (Costerton et al., 1999; Gottenbos et al., 1999; Costerton, 2001) , yet the genetic basis for the formation of extracellular matrices in different P. aeruginosa biofilms remains poorly defined. There are strong indications that this organism forms multicellular aggregates within sites of infection, e.g. in the lungs of cystic fibrosis patients or on the surfaces of contaminated catheters (Deretic, 2000; Singh et al., 2000). Thus, identification of constituents of the extracellular matrix might enable the development of new therapeutic strategies. Chemical analyses of P. aeruginosa SSA biofilms formed in flow cells suggest that the matrix that encases the cells in such biofilms is composed of water, exopolysaccharide, DNA, RNA, proteins and ions (Sutherland, 2001; Whitchurch et al., 2002). However, the genes involved in the production of matrix material have not yet been described for this organism. Alginate is generally referred to as an important component of the extracellular matrix of SSA biofilms. These assertions are most probably based on the fact that P. aeruginosa strains isolated from the lungs of cystic fibrosis patients are mucoid and overproduce alginate (Dogget, 1969; Baltimore, 1993; Govan and Deretic, 1996; Poschet et al., 2001). However, most strains of P. aeruginosa are non-mucoid and have not been shown to produce significant amounts of alginate. Most importantly, Hentzer et al. (2001) and Wozniak et al. (2003) recently demonstrated that alginate is not a significant component of the extracellular polysaccharide present in the matrix of biofilms formed by P. aeruginosa strains PA01 and PA14 under commonly used laboratory growth conditions.
In this paper, we present the results of a genetic screen that identified pellicle-defective mutants of P. aeruginosa PA14. The mutants were also defective in the formation of SSA biofilms and displayed altered colony morphologies, indicating that the mutated genes are important in diverse settings of biofilm formation. Sequence analyses suggest that these genes are involved in the production of an extracellular carbohydrate-rich substance. Scanning electron microscopy (SEM) of wild-type and mutant samples implicates these genes in the production of the biofilm matrix. Enzymatic treatment combined with carbohydrate analyses of the wild-type matrix and comparison of the carbohydrates present in wild-type and mutant samples suggest that the P. aeruginosa PA14 pellicle matrix is rich in glucose. We propose that the newly identified genes participate in the production of the biofilm matrix and that the biofilm matrix in P. aeruginosa PA14 contains a glucose-rich component.
Isolation of mutants defective in pellicle formation
A pellicle is a biofilm that assembles at the air–liquid interface of a standing liquid culture. P. aeruginosa PA14 can form robust pellicles when grown in standing culture (Fig. 1A). After inoculation of standing T broth at low cell densities (OD600 = 0.0025), the bacteria grew exponentially in the planktonic phase of the culture. After 1 day of growth, a thin pellicle began to form at the air–liquid interface. By 48 h, the pellicle developed a distinctive architecture with cells embedded in an opaque matrix visible with the naked eye. After 5 days of incubation, the pellicle had acquired extremely rigid properties and could not be dispersed even by extensive vortexing and boiling. Pellicle formation depended on the nutrient composition of the growth medium; robust pellicles were observed in both rich and minimal media (see Experimental procedures).
In order to study the process of pellicle formation in P. aeruginosa PA14, we screened a newly made library of ≈ 10 000 P. aeruginosa PA14 Tn5-B21 mutants for defects in pellicle formation. Individual mutant colonies were inoculated into standing liquid cultures and allowed to grow for 72 h at room temperature. Mutants that grew as turbid cultures that never gave rise to pellicles were isolated and characterized further (for phenotype, see Fig. 1A). Arbitrary polymerase chain reaction (PCR) was used to determine the precise location of the transposon insertion (O’Toole et al., 1999). The sequence from the arbitrary PCR-derived fragments was compared with the P. aeruginosa PAO1 genomic sequence. Eleven mutants were found to contain insertions in four genes of a seven-gene cluster (PA3058 to PA3064) (Fig. 1B). Subsequently, we generated two independent in frame deletions, one in PA3064 and one in PA3060 (blue boxes in Fig. 1B). Both deletions resulted in mutants with phenotypes indistinguishable from the transposon mutants. The fact that multiple insertions in the locus and in vitro-constructed in frame deletions in two genes all gave rise to indistinguishable phenotypes strongly indicates that the phenotypes observed are the result of the mutations at hand. Thus, complementation analyses were not carried out. We designate these genes (PA3064–PA3058) as pelABCDEFG for the fact that they are involved in pellicle formation.
Properties of the pellicle-defective mutants
We isolated the pel mutants as a result of their defects in pellicle formation under a particular set of environmental conditions. To test whether P. aeruginosa PA14 required the pel genes for biofilm formation under other conditions, we analysed the mutants’ ability to form both pellicles and SSA (solid surface-associated) biofilms under a variety of environmental conditions. P. aeruginosa PA14 forms robust pellicles when grown in glycerol–glutamate minimal medium (MSgg) or rich media such as T broth or LB over a wide range of temperatures (15–37°C), but the thickness and architecture of the pellicle vary depending on growth medium and temperature. Importantly, the pel mutants are defective in pellicle formation in all growth media and at all temperatures that we have tested to date.
Our laboratory carried out a previous genetic analysis of SSA biofilm formation by P. aeruginosa PA14 (O’Toole and Kolter, 1998a). Applying the widely used crystal violet (CV) staining method to detect biofilm formation, our laboratory reported the isolation of strains containing mutations in genes required for the initiation of SSA biofilm formation in P. aeruginosa PA14. The conditions used previously were specifically designed to analyse the initiation of biofilm formation because washes were performed gently and after only 8 h of incubation (O’Toole and Kolter, 1998a). Three classes of mutants defective in the initiation of SSA biofilms were isolated in the previous study: mutants defective in growth, mutants defective in flagellar motility and mutants defective in type IV pili. Mutations in the pel genes were not obtained in the previous screen. We thus wanted to investigate whether the pel mutants had a phenotype when assayed using CV staining to measure the initiation of SSA biofilm formation.
To determine whether the pel genes are required for the initiation of biofilm formation, we tested the pel mutants under conditions identical to those described by O’Toole and Kolter (1998a). The pel mutants formed CV-staining rings in 96-well polyvinyl chloride (PVC) microtitre dishes just like the wild-type strain (Fig. 2A). As expected, flagellar motility mutants (e.g. flgI) show significantly less CV staining (Fig. 2A). These experiments were performed using two different media, T broth and M63 with glucose plus casamino acids (CAA), and at a range of temperatures (room temperature, 30°C and 37°C). These environmental changes did not alter the trend in the results. Therefore, we conclude that the pel mutants are not defective in the initiation of SSA biofilms, consistent with the fact that the pel genes were not identified in the previous screen. The fact that the pel mutants are not defective in the initiation of biofilm formation is consistent with the observations that they are indistinguishable from the parent strain with respect to growth rate and flagellar and twitching motilities.
A simple modification of the wash conditions used before CV staining allowed us to distinguish between the initiation of biofilm formation and the formation of a mature, matrix-encased biofilm by P. aeruginosa PA14. Mild washing by simply submerging the microtitre dish in water before and after CV staining allows the detection of cells that are only loosely associated with the surface, such as those present during the initiation phases of biofilm formation. Washing the microtitre wells with running water before and after CV staining led to the removal of loosely associated cells and resulted in virtually no staining of the wells during early incubation periods even when using the wild-type strain (Fig. 2B). After maturation (about 24 h under the incubation conditions used in Fig. 2B), the SSA biofilm formed by the wild-type strain became resistant to these harsher washing conditions. Figure 2B demonstrates that the pel mutants were defective in the formation of mature SSA biofilms. Wild-type P. aeruginosa PA14 formed robust, vigorous, washing-resistant SSA biofilms on three different surfaces: polystyrene, PVC and borosilicate glass (Fig. 2C). In every case, the pel mutants did not form mature SSA biofilms on any of these surfaces; staining was reduced by at least 75% in every case (Fig. 2C). Therefore, the pel genes are required for both SSA biofilm formation and pellicle formation by P. aeruginosa PA14.
Analysis of the PA3058–PA3064 sequence
We used sequence analysis of the pel gene cluster in an attempt to predict a putative function for these genes in mature biofilm formation. The pel gene cluster contains seven open reading frames (ORFs), PA3058–PA3064, spanning a 12.2 kb region of the P. aeruginosa PAO1 genome. The ongoing sequencing of the P. aeruginosa PA14 genome (Montgomery et al., 2002; accession numbers AABQ06000000–AABQ06000008) allowed us to compare this region of P. aeruginosa PAO1 with the corresponding region of P. aeruginosa PA14. The 12.2 kb region in PAO1 shares 98% nucleotide sequence identity with the corresponding PA14 sequence. Most importantly, the predicted protein products from the seven ORFs are identical in P. aeruginosa PA14 and P. aeruginosa PAO1. We used PCR analyses to detect the presence of these genes in seven clinical and five environmental P. aeruginosa strains from our strain collection. All the strains tested contained the seven genes (PA3058–PA3064) as judged by PCR analyses. These results suggest that this gene cluster is highly and widely conserved in P. aeruginosa.
All seven predicted proteins from this gene cluster are currently annotated as proteins of unknown function (Stover et al., 2000). However, homologues appear to be widely conserved among diverse microbes (for protein accession numbers, see Experimental procedures). The entire gene cluster is conserved in Ralstonia solanacearum, Ralstonia metallidurans, Geobacter metallireducens and Burkholderia fungorum. However, in none of these cases has a function been assigned to any of the predicted proteins. Homologues of three of the seven genes are found together in Clostridium acetobutylicum ATCC824 and Magnetococcus sp. MC-1. Homologues to individual genes within this gene cluster are found in a wide range of microbes, including Deinococcus radiodurans R1 (PA3064), Sulfolobus tokodaii strain 7 (PA3064), Xanthomonas campestris pv. campestris ATCC3391 (PA3063), Methanosarcina acetivorans C2A (PA3063), Xanthomonas axonopodis pv. citri 306 (PA3063), Fusobacterium nucleatum ATCC 25586 (PA3058) and Thermotoga maritima MSB8 (PA3058).
To date, sequence comparisons of the predicted Pel protein products using the psi-blast program have not revealed any protein homologues of known function. However, in-depth sequence analyses revealed regions of limited but intriguing sequence similarity to proteins involved in polysaccharide processing. Figure 3 indicates the regions of amino acid sequence similarity within each protein that correspond to sequences of proteins with known functions in polysaccharide processing. The C-terminal region of PA3059 shows sequence similarity to a family of glycosyltransferases that contains the COG0438, RfaG, glycosyltransferase motif and the pfam00534, glycosyltransferase group 1 motif (Mulder et al., 2003).
The pel mutants are defective in the production of an extracellular matrix
The sequence of the pel genes led us to hypothesize that they are involved in the production of a putative carbohydrate-rich product. Carbohydrate-rich substances such as colanic acid (Escherichia coli), cellulose or cellulose-like materials (Salmonella enteritidis, Salmonella typhimurium, Pseudomonas fluorescens and E. coli), Vps (Vibrio cholerae) and the intercellular polysaccharide adhesin-PIA (Staphylococcus epidermidis and S. aureus) have been implicated in biofilm formation (Mack et al., 1996; Cramton et al., 1999; Watnick and Kolter, 1999; Yildiz and Schoolnik, 1999; Danese et al., 2000; Zogaj et al., 2001; Solano et al., 2002; Spiers et al., 2002; Dobinsky et al., 2002). These substances are thought to be important components of the biofilm matrix. Chemical analyses of P. aeruginosa SSA biofilms formed in flow cells suggest that the matrix that encases the cells in such biofilms is composed of water, exopolysaccharide, DNA, RNA, proteins and ions (Sutherland, 2001; Whitchurch et al., 2002). As the carbohydrate component of the P. aeruginosa PA14 biofilm matrix is currently undefined and the pel mutant phenotype is defective in biofilm formation, we hypothesized that the pel gene cluster might be required for the production of a carbohydrate-containing component of the biofilm matrix.
Simple macroscopic and microscopic observations support the notion that there is a defect in biofilm matrix production in the pel mutants. For one, the pel mutants remained as turbid cultures composed of independent planktonic cells that were not held together by a matrix (Fig. 1A). In addition, wild-type and mutant colonies had dramatically different properties. The cells from wild-type colonies appeared to be encased in matrix material because the colonies could be removed as a single piece from the agar and the colony did not dissolve when submerged in water. In contrast, pel mutant colonies could not be removed from the agar surface in a single piece and the constituent cells dispersed easily in water. We used scanning electron microscopy (SEM) to visualize the colony at the cellular level. Figure 4 shows SEM images of a wild-type colony and a pel mutant colony. Although both colonies contain rod-shaped cells, it is apparent that the wild-type cells are embedded in an extracellular matrix whereas the pel mutant cells are not. These results support the hypothesis that the pel gene cluster is involved in the production of the biofilm's matrix.
Further analyses of wild-type and pel mutant colonies provided additional support for the hypothesis that the pel genes are involved in the production of an extracellular matrix. Congo red has been shown to bind extracellular matrix components in numerous other bacteria, among them Salmonella enterica serovar Typhimurium, S. enteritidis, E. coli and Yersinia pestis (Perry et al., 1990; Hinnebusch et al., 1996; Zogaj et al., 2001; Darby et al., 2002; Solano et al., 2002). When grown on agar plates containing Congo red, P. aeruginosa PA14 colonies were dark red whereas the pel mutants were pale pink (Fig. 5). Solano et al. (2002) showed that colonies that stain red under these conditions do so because of Congo red adsorption. The wild-type colonies also had a wrinkled or ‘rugose’ morphology, whereas the pel mutant colonies were smooth (Fig. 5). Rugose colonies in V. cholerae result from the overproduction of an exopolysaccharide (Yildiz and Schoolnik, 1999; Yildiz et al., 2001). Similar changes in colony morphology in Salmonella have also been correlated with the presence or absence of an extracellular cellulose-like material (Zogaj et al., 2001; Solano et al., 2002), and a cellulose-like material is required for the ‘wrinkly spreader’ colony morphology in P. fluorescens (Spiers et al., 2002). Thus, the smooth phenotype of the pel mutant colonies is consistent with the loss of an extracellular matrix component.
Alginate is not required for biofilm formation under the conditions tested
As mentioned above, exopolysaccharides are often shown to be required in biofilm matrix formation. One possible explanation for the phenotype of the pel mutants could have been that they were involved in the same genetic pathway of a previously characterized exopolysaccharide. As alginate is a well-characterized exopolysaccharide produced by mucoid variants of P. aeruginosa, we wanted to determine whether alginate-defective mutants had similar phenotypes to the pel mutants. We found that, in contrast to the defects in biofilm formation displayed by pel mutants, a mutant completely defective in alginate production (ΔalgD) formed pellicles (not shown) and SSA biofilms indistinguishable from those formed by the parent wild-type strain (Fig. 6). Our results are consistent with those of others, who have reported that P. aeruginosa PAO1 and PA14 defective in alginate production produce SSA biofilms just like the wild-type strains (Hentzer et al., 2001; Wozniak et al., 2003). Taken together, all these results show that alginate is not required for biofilm formation under the diverse conditions tested in different laboratories. P. aeruginosa PA14 ΔalgD also produced red and wrinkled colonies on agar plates containing Congo red, just like the wild-type strain (data not shown). All these results strongly suggest that, under the conditions studied thus far, alginate is not required for the production of a matrix that enables pellicle formation and SSA biofilm formation.
The cupA fimbriae participate in pellicle formation
Aside from exopolysaccharides, other surface structures such as fimbriae have been implicated in the formation of SSA biofilms and wrinkled colonies, as well as in Congo red binding (Prigent-Combaret et al., 2000; Romling et al., 2000). Two fimbrial structures, type IV pili and cupA fimbriae, have been shown to be involved in the initiation of SSA biofilm formation in P. aeruginosa (type IV pili in strains PAO1 and PA14 and cupA in strains PAK and PAO1) (O’Toole and Kolter, 1998a; Vallet et al., 2001; D’Argenio et al., 2002). Twitching motility and phage susceptibility experiments indicate that the pel mutants produce functional type IV pili. Currently, there are no simple assays to determine whether the cupA fimbriae are present and functional in P. aeruginosa PA14, so we used a cupA2 transposon insertion mutant (that we isolated independently) in P. aeruginosa PA14 to test the role of the cupA fimbriae in pellicle formation and Congo red binding. Figure 7 shows Congo red-stained pellicles formed by the wild-type and a cupA2 mutant. The cupA2 gene was not required for Congo red staining or pellicle formation. However, the structure of the cupA2 mutant pellicle differs greatly from that of the wild-type pellicle (Fig. 7). The wild-type pellicle had a consistency similar to wet paper; the material could be moved by tweezers from one container to another, it wrinkles and bends and could only be torn apart by significant force (Fig. 7). The cupA2 mutant pellicles had an elastic consistency; they could be moved but became very thin in some regions and broke easily (Fig. 7). In addition, the cupA2::Tn5 ΔpelA double mutant had a pel mutant phenotype, suggesting that the mutation in cupA2 is not leading to the activation of a novel pellicle formation pathway (data not shown). Taken together, these results indicate that, although the cupA fimbriae participate in the establishment of wild-type pellicle, a mutation in the cupA2 gene that appeared to alter the physical structure of the pellicle did not lead to a complete elimination of the biofilm matrix component that binds to Congo red. Our results, coupled with the aforementioned facts, suggest that the pel genes are involved in the production of a carbohydrate-containing component matrix in P. aeruginosa PA14 biofilms.
Enzymatic and biochemical treatment of the biofilm matrix suggests the presence of a glucose-rich matrix component
Wild-type P. aeruginosa PA14 pellicles had a paper-like consistency that was not easily disrupted. Simple procedures such as boiling and vortexing were not sufficient to release the cells from the biofilm matrix. Enzymatic treatments with DNase, RNase and Proteinase K alone or in combination were not sufficient to release the cells from the biofilm matrix or impact the structure of the pellicle. We found that three other treatments broke down the paper-like structure of the wild-type pellicle. Sodium periodate and sodium hydroxide broke the pellicle down into single cells and small cell clusters (see Experimental procedures). Sodium periodate oxidizes carbohydrate rings, suggesting that the biofilm matrix requires some intact carbohydrate-containing component for its integrity (Fatiadi, 1974; Sussich and Cesàro, 2000). Cellulase treatment of the wild-type pellicle resulted in a breakdown of the paper-like structure of the pellicle into small fragments (Fig. 8). As we showed that the cupA fimbriae appear to be involved in the structural integrity of the pellicle (see Fig. 7), we decided to test the more fragile cupA2 mutant pellicle for cellulase sensitivity. Cellulase treatment of the cupA2::Tn5 mutant pellicle results in dissolution to single cells and cell clusters (Fig. 8). These results suggest that the P. aeruginosa PA14 biofilm matrix contains a carbohydrate-rich component, possibly similar in structure to cellulose.
Although the pellicle matrix was susceptible to digestion with cellulase, three lines of evidence suggest that the cellulase-susceptible matrix material is not cellulose per se. (i) Calcofluor binding of bacterial colonies is often associated with the production of a cellulose-like material (Zogaj et al., 2001), yet the pellicle and wild-type colonies of P. aeruginosa PA14 did not fluoresce when stained with Calcofluor. (ii) The P. aeruginosa PA14 genome sequence does not contain a cluster of genes that is orthologous to other cellulose biosynthesis gene clusters. (iii) We analysed the supernatant from wild-type and cupA2::Tn5 pellicles after cellulase digestion and found no significant increase in free glucose. Together, these lines of evidence suggest that the P. aeruginosa PA14 matrix contains a carbohydrate-rich component that, although cellulase sensitive, is not cellulose. One explanation for the ability of cellulase to break down non-cellulose material is that glucosidic linkages in such material may be required to maintain structural integrity. If cellulase can hydrolyse key structural bonds, the material may become water soluble without the release of detectable amounts of free glucose. The commercially available cellulase enzyme preparation that we used contains a mixture of enzymatic activities including glucosidic activities against both 1-4- and 1-6-linked glucose polymers. Therefore, the carbohydrate-containing matrix component need not necessarily be a 1-4-linked cellulose-like polymer; it may contain 1-6 linkages. The above results led us to hypothesize that the P. aeruginosa PA14 matrix contains a carbohydrate-rich material with glucosidic linkages necessary for structural integrity.
Carbohydrate analyses reveal glucose as a major component of the biofilm matrix
In order to determine the carbohydrate content of the pellicle's matrix material, it was necessary to separate the cells from the biofilm matrix. We developed a crude purification procedure for the matrix material based on the fact that it could be solubilized with 1 M sodium hydroxide (see Experimental procedures). After dissolving the matrix, we removed the cells by ultracentrifugation followed by filtration of the supernatant through a 0.2 µm pore size filter. We then performed carbohydrate analysis using gas chromatography/mass spectrometry (GC/MS) on the ethanol-precipitable components of the supernatant. The results that we obtained from total carbohydrate analyses were consistent with the presence of carbohydrate-rich material. One milligram of purified matrix material contained, on average: 62.4 µg of rhamnose, 37.6 µg of glucose, 28.7 µg of unknown amino sugars, 19.8 µg of N-acetyl quinovosamine (QuiNAc), 6.5 µg of N-acetyl glucosamine (GlcNAc), 6.4 µg of kdo and 3.6 µg of ribose. Most of these sugars are components of the P. aeruginosa PA14 lipopolysaccharide (LPS) (Rocchetta et al., 1999). However, carbohydrate linkage analysis revealed glucosidic linkages not expected from LPS alone, suggesting other carbohydrate components in addition to LPS. To carry out the linkage analysis, the matrix material was subjected to methylation by preparing partially methylated alditol acetate derivatives. The glycosyl linkage analysis of the purified matrix material revealed 18% 3-linked rhamnopyranosyl residues and 2% 3,4-linked rhamnopyranosyl residues. Of the glucopyranosyl residues, 15% were terminal linkages, 13% were 4-linked, 10% were 3-linked, 10% were 3,4-linked, 9% were 3,6-linked, 9% were 4,6-linked and 3% were 6-linked. Not only do these data support the presence of additional carbohydrate-rich material, they also help to explain the observation that the matrix material is cellulase sensitive. The diversity of glucosidic linkages suggests that the material is not likely to be cellulose. In summary, carbohydrate analyses of crude matrix material are consistent with the presence of a matrix material that contains a carbohydrate-rich component different from cellulose and LPS, and support the hypothesis that the matrix contains a glucose-rich substance.
As the pel mutants do not produce significant amounts of biofilm matrix, we compared carbohydrate content from wild-type pellicles with pel mutant cultures to obtain an indication of the carbohydrate composition of the matrix. To perform these analyses, wild-type pellicles of P. aeruginosa PA14 were grown, washed extensively with water and lyophilized. These samples were compared with samples from cultures of the ΔpelA mutant grown under the same conditions. We used wild-type pellicles compared with pel mutant cultures, rather than colonies, because the pellicle is composed of a larger proportion of matrix material compared with colonies. Analyses of the carbohydrate content of each sample were performed as described in the Experimental procedures. The analyses performed could detect rhamnose, ribose, glucose, kdo, QuiNAc and GlcNAc (Fig. 9). Of these carbohydrates, we detected a 3.74-fold increase in a single sugar, glucose, in the wild-type sample compared with the mutant sample. These results support our hypothesis that the pel genes are required for the production of a glucose-rich matrix component of the pellicle.
Genes required for the production of the P. aeruginosa PA14 biofilm matrix
A biofilm's matrix encases the constituent cells, providing the scaffold for the biofilm's architecture. The matrix also gives the cells the opportunity of a sedentary lifestyle, protecting the cells from currents or shear forces ubiquitous in fluid environments; the mere presence of a matrix helps to keep the cells in place. By virtue of the larger size of the resulting matrix-enclosed multicellular aggregate, cells can escape predation in the environment or engulfment by phagocytic cells within a mammalian host. The matrix can also serve a protective function that alleviates nutritional scarcity and environmental assaults. By sequestering nutrients, ions and water, the matrix can provide a safe haven in times of environmental stress. Given the importance of biofilm matrices in microbial survival, we hope to understand the molecular details of matrix production.
Most multicellular aggregates, including mammalian tissues and bacterial biofilms, are encased in self-synthesized matrices (Mackie, 1985; Nieduszynski, 1985; Sutherland, 2001). The components of these extracellular matrices are diverse but usually include polymeric substances such as polysaccharides, nucleic acids and proteins. In this paper, we present the identification and characterization of the pel genes, required for biofilm matrix formation in P. aeruginosa PA14 under diverse environmental conditions. The pel genes are required for the production of pellicles and the SSA biofilms that form on plastic or glass. The pel mutants are able to attach to solid surfaces, but shear forces produced by washing remove the cells from the surface compared with the wild-type sample which, after the initiation phase of biofilm formation on the surface, is able to resist such shear forces. The pel mutants grow, twitch and swim like wild-type P. aeruginosa PA14, indicating that previously identified contributors to biofilm formation are not defective in these mutants (O’Toole and Kolter, 1998b). SEM comparisons of wild-type and pel mutant colonies suggest that the pel genes are required for the formation of the matrix material that encases the cells.
What do the pel gene products synthesize?
Five of the seven predicted protein products of the pel genes have amino acid sequence similarities with proteins of known function in carbohydrate processing (Fig. 3). PelA has N-terminal similarity to endo-α-1,4-polygalactosaminidase and C-terminal similarity to an oligogalacturonide lyase (Tamura, 1993; Thomson et al., 1999). PelC has a region of similarity to a galactoside-2-α-l-fucosyltransferase (Piau et al., 1994). PelD has similarity to an l-lactate permease (Dong et al., 1993). PelE has a region of similarity to a sucrose synthase (Komatsu et al., 2002), and PelF has a recognizable glycosyltransferase motif (Belanger et al., 1999; Bateman et al., 2000). These regions of sequence similarity suggest that the pel genes might be involved in the production of a carbohydrate-containing substance. PelD, PelE and PelG are predicted to contain transmembrane helices and could therefore be involved in export of the carbohydrate-containing substance.
The simplest hypothesis for the role of the pel gene cluster in biofilm matrix production is that the pel genes may produce a carbohydrate-rich component required for the structural integrity of the biofilm matrix. However, the pel genes could produce a precursor to the biofilm matrix or a regulatory molecule required for the initiation of production of the biofilm matrix by a series of redundant genes.
What is the main structural component of the P. aeruginosa PA14 biofilm matrix?
Searches for altered colony morphology mutants in strains that display elaborate colonial phenotypes have identified several of the components required for the formation of such complex structures. In this way, thin aggregative fimbriae in S. enterica serovar Typhimurium, curli fimbriae in E. coli and CupA fimbriae in P. aeruginosa PAO1 were identified as extracellular structures involved in colony morphology and/or biofilm formation (Hammar et al., 1995; Romling et al., 1998; 2000; D’Argenio et al., 2002). Disruption of fimbrial structures results in a partial disruption of the multicellular structures in these organisms. For example, wrinkled red (Congo red staining) colonies of S. enterica serovar Typhimurium or E. coli become smooth and pink as a result of the loss of fimbriae (Hammar et al., 1995; Romling et al., 2000). Yet, in those cases, mutants lacking fimbriae retain some structure because those colonies are not easily disrupted. This holds true in P. aeruginosa PA14 as well. Although type IV pili and CupA fimbriae have been implicated in the initiation of SSA biofilm formation in P. aeruginosa, we describe here how these fimbriae are not absolutely required for pellicle formation or matrix production in P. aeruginosa PA14.
For several bacterial species, smooth white (no Congo red binding) colonies that can be easily washed away result when the matrix is severely compromised (Romling et al., 2000; Zogaj et al., 2001). This is correlated with the disruption of putative exopolysaccharide biosynthetic genes (Romling et al., 2000; Zogaj et al., 2001; Darby et al., 2002). Analyses using colony morphology mutants have led to the discovery of genes presumed to be involved in exopolysaccharide synthesis in S. enterica serovar Typhimurium, E. coli, P. fluorescens and V. cholerae (Wai et al., 1998; Yildiz and Schoolnik, 1999; Zogaj et al., 2001; Spier et al., 2002). On Congo red-containing medium, the pel mutants of P. aeruginosa PA14 grow as smooth pale pink colonies that can be easily washed away similar to the phenotype described above. As their sequence suggests that they are involved in the production of a carbohydrate-containing product, the pel genes could be involved in the production of a putative exopolysaccharide.
What is the carbohydrate component of the P. aeruginosa PA14 biofilm matrix?
Glucose-rich polymers such as cellulose, starch and β-1-3-glucans play a structural role in holding cells together in many organisms including plants, algae and bacteria (Mackie, 1985). Morphology studies, involving both colonies and pellicles, of P. fluorescens, E. coli, S. enterica serovar Typhimurium and Y. pestis multicellular aggregates have revealed conserved exopolysaccharide synthesis genes involved in Congo red binding, Calcofluor binding and the production of cellulase-sensitive matrix material (Zogaj et al., 2001; Darby et al., 2002; Spiers et al., 2002). The genes required for the production of the matrix material are, in every case, similar to the genes in the cellulose operons present in Acetobacter xylinum (Wong et al., 1990; Saxena et al., 1994) and Agrobacterium tumefaciens (Matthysse et al., 1981). Such genes are not present in P. aeruginosa PA14 or PA01. This and the lack of Calcofluor binding argue against cellulose production by P. aeruginosa. Nonetheless, the only significant difference in carbohydrate content detected between wild-type and pel mutant samples was a nearly fourfold increase in the amount of glucose in the wild-type sample. Our results thus suggest the presence of a novel (not cellulose) glucose-rich structural component in the P. aeruginosa PA14 matrix.
The identification of the pel genes and the discovery of a glucose-rich matrix material in P. aeruginosa PA14 should prove helpful in future studies aimed at determining the role of a biofilm's matrix. The pel genes are also potential targets for the control and possible eradication of biofilms formed by P. aeruginosa and related organisms. By understanding how the cells regulate the expression of the pel genes, we may determine the environmental cues that are sensed in diverse environments and trigger some pathways of biofilm formation.
Strain and growth conditions
Pseudomonas aeruginosa PA14 was used for all biofilm studies. All biofilms were grown at room temperature (21–27°C), 30°C or 37°C as specified per experiment. P. aeruginosa PA14 was grown in T-broth medium (10 g l−1 bacto tryptone, 5 g l−1 NaCl), LB, tryptone broth (10 g l−1 bacto tryptone), M63 supplemented with 0.2% glucose and 0.5% casamino acids, 1 mM MgSO4, MSgg minimal medium (2.5 mM potassium phosphate buffer, pH 7.0, 100 mM MOPS, pH 7.0, 50 µM FeCl3, 2 mM MgCl2, 50 µM MnCl2, 1 µM ZnCl2, 2 µM thiamine, 50 mg of tryptophan, 50 mg off phenylalanine, 0.5% glycerol, 0.5% glutamate and 700 µM CaCl2) unless otherwise stated. Antibiotics were added at the following concentrations for E. coli: tetracycline (Tc), 15 µg ml−1; nalidixic acid (Nal), 20 µg ml−1; gentamicin (Gm), 10 µg ml−1; for P. aeruginosa: Tc, 150 µg ml−1; Gm, 60 µg ml−1.
The following strains were used in this paper. P. aeruginosa PA14: ΔpelA, ΔpelE pelA::Tn5, pelB::Tn5, pelD::Tn5, pelG::Tn5, ΔalgDΔpelA, flgI::Tn5, cupA2::Tn5 (isolated and described within this paper); ΔalgD (Yorgey et al., 2001), pilA::Tn5, pilB::Tn5, pilC::Tn5 (O’Toole and Kolter, 1998a).
Approximately 10 000 transposon mutants were generated with Tn5-B21 (Tcr) using a modification of published protocols (Simon et al., 1989) as described by O’Toole and Kolter (1998b). The resulting transposon mutants were screened as described below.
Pellicle formation assay
Standing cultures containing 6 ml of T broth were grown at room temperature (20–27°C) in an 18 × 150 mm DurexTM borosilicate glass tube (VWR). Pellicles were assayed by visual inspection of the air–liquid interface of the standing culture. Complete coverage of the surface of the culture by an opaque layer of cells and matrix material is considered to be pellicle formation.
Screen for mutants defective in pellicle formation
Screening of the tetracycline-resistant mutants was carried out by inoculating 2 ml of T broth plus Tc in a 13 × 100 mm DurexTM borosilicate glass tube (VWR) with individual mutants and letting the cultures grow at room temperature. The standing cultures were then assessed at 72 h, and reassessed at 86 h, for the presence or absence of a pellicle at the air–liquid interface. Candidates defective in pellicle formation were retested using the pellicle formation assay. From ≈ 10 000 individual mutants screened, we obtained 13 pellicle-defective mutants. Eleven contained mutations in the PA3058–PA3064 gene cluster. The other two independent insertions were found outside the gene cluster, one in PA3180 and one in PA4907, and they were not characterized any further as they were single insertions.
Abiotic SSA biofilm formation assay
Assays were started by two methods: (i) overnight cultures rolling at 37°C were diluted 1:100 in matching growth medium; (ii) bacteria grown overnight on agar plates were resuspended in matching liquid medium and diluted to a final OD600 of 0.0025. Cultures were transferred to a standing culture vessel: 96-well microtitre plates: PVC = 100 µl per well (Falcon); polystyrene = 150 µl per well; or 13 × 100 mm DurexTM borosilicate glass tubes = 2 ml per tube (VWR) and allowed to stand at room temperature, 30°C and 37°C for the specified amount of time.
Staining of SSA biofilms
For the initiation of biofilm formation, assay culture vessels were washed by dunking the vessel into a container filled with standing water and gently tapping the wash into a waste container. For the mature biofilm assay, culture vessels were washed rigorously under warm to hot tap water before staining. Samples were stained by the addition of 1% crystal violet solution into each well above the initial inoculation level and allowed to sit for 20 min before washing. After staining, vessels were washed with the respective wash condition. Crystal violet stain was measured after the addition of dimethyl sulphoxide (DMSO) to each dry well. The samples then sat for 20 min and were measured at OD590 on a plate reader.
All samples were tested in at least eight independent wells. In order to determine consistency between 96-well plates, eight wells of wild-type PA14 and the ΔpelA mutant were included on all plates. No significant differences were detected between wild-type PA14 and ΔpelA mutants compared with similar samples over all the plates tested.
DNA sequences flanking the transposon mutants were determined using arbitrary PCR, as described by O’Toole et al. (1999), with the following modifications. Arbitrary PCR was performed on purified DNA using the GC-rich PCR system (Roche Diagnostics), with the addition of 5 µl of 25 mM MgCl2 and 5 µl of GC-rich solution per 50 µl reaction.
Construction of P. aeruginosa pelA and pelE deletion mutants
Flanking sequences of the pelA gene were amplified by PCR from PA14 chromosomal DNA using primers corresponding to the PA3064 predicted ORF sequence. The deletion construct was created by sewing the 5′ and 3′ regions of PA3064 together in two rounds of PCR. The first round of PCR created two products using the following primers. For ΔpelA: 5′-CCGGGCCTCAAGCTGTTTTTCAAT-3′ (LF114), 5′-AGAAG T G GTAGTACA GGTGCAGGCCGGGACTCGGCA GCGGC A CAAGACGC-3′ (LF115); 5′-GTCTTGTGCCGCTGCCGAG TCCCGGCCTGCACCTGTACTACCACTTCT-3′ (LF116) 5′-CAATCGGCGCGTCTGCTCCTCA-3′ (LF117). For ΔpelE: 5′-TTCACCCGCGCCCATCACATCCTG-3′ (LF142), 5′-CAGTAT CTCGCCAGGGCCGCGAAGACGCTCGCCCAACTGCTCA-3′ (LF141); 5′-TGAGCAGTTGGGCGAGCGTCTTCGCGGC CCTGGCGAGATACTG-3′ (LF144) 5′-GCCGACCAGCCCGA CCAC-3′ (LF145). The second round of PCR used the following primers to generate a fusion between the 5′ and 3′ flanking regions: ΔpelA: 5′-CTAGCACTAGTAGCCCCGGCCAC GTCTACCTC-3′ (LF118) 5′-CTAGGAGCTCACGCGCCTGC TCGAAGTCACC-3′ (LF119). ΔpelE: 5′-CTAGCACTAGTGCG AGCTGCCGGGCGATGAGG-3′ (LF143) 5′-CTAGGAGCT CGGATTGGCGGCGCGATAGGTG-3′ (LF146). The second-round primers were constructed with SacI and SpeI sites in the respective primers. The fusion PCR products were digested with SacI and SpeI (New England Biolabs) and cloned into the pEX18Gm plasmid, which was digested with SacI and XbaI (Hoang et al., 1998). For ΔpelA, pEX18Gm plasmid contains 579 bp from the 5′ region of the gene and 829 bp from the 3′ region of the gene. For ΔpelE, pEX18Gm plasmid contains 894 bp from the 5′ region of the gene and 838 bp from the 3′ region of the gene. These constructs were mated into P. aeruginosa PA14 and used to replace the wild-type copy of pelA or pelE, respectively, in PA14 and PA14DalgD, as described previously (Donnenberg and Kaper, 1991; Hoang et al., 1998). The resulting pelA deletion contains the first 512 amino acids in frame with the last 194 amino acids of PA3064. Overall, the PelA protein is missing 243 amino acids in the centre of PA3064. The resulting pelE deletion contains the first 23 amino acids in frame with the last seven amino acids of PA3060. Overall, the PelE protein is missing 299 amino acids in the centre of PA3060. The constructs were confirmed by PCR amplification of chromosomal DNA.
Genomic sequence analysis
The predicted protein sequences PA3058–PA3064 were analysed using the psi-blast, blocks and pfam sequence comparison programs (Altschul et al., 1997). Regions of sequence identity with proteins of known function were identified based on the psi-blast program, and regions with similarity to polysaccharide biosynthesis pathways are shown in Fig. 3.
Pairwise alignments were performed on PA3064–PA3058 against all the genomes in the CMR (via TIGER) and GenBank databases (using psi-blast). Identity between sequences was calculated as the percentage of identical residues in an alignment. psort was used tentatively to predict the cellular location and transmembrane domains of the predicted proteins (Nakai and Horton, 1999).
The homologous proteins follow. Ralstonia solanacearum: RSc2276, RSc2275, RSc2274, RSc2273, RSc2272, RSc2271 and RASc2270; Ralstonia metallidurans Reut_12: Reut0105,Reut6029, Reut0588, Reut0589,Reut0591, Reut0592 and Reut0593; Geobacter metallireducens: Gmet 1874, Gmet 1872, Gmet 1870, Gmet 1869, Gmet 1868, Gmet 1867 and Gmet 1866; Burkholderia fungorum: Bcep5552, Bcep5553, Bcep5554, Bcep5555, Bcep5556, Bcep5557 and Bcep5558; Clostridium acetobutylicum ATCC824: CAC0736, CAC0735 and CAC0734; and Magnetococcus sp. MC-1: Mmc13277,Mmc13274 and Mmc13270.
Time course assays and biofilm formation on glass and plastic
All biofilm experiments were started by scraping bacteria from a fresh overnight plate grown at room temperature into 500 µl of T broth. This culture was then diluted to a final OD600 = 0.0025 and distributed into the appropriate container at time zero.
Biofilm formation on 96-well plates was assayed on two types of plastics. Non-tissue culture-treated MicroTestTM flat-bottomed polystyrene plates and Falcon 3912 MicroTest IIITM flexible assay PVC plates (Falcon). For growth curves, sterile, covered 96-well non-tissue culture-treated MicroTestTM flat-bottomed polystyrene plates were used. Eight individual samples of 100 µl each were assessed for each mutant at each time point, and the experiment was repeated twice. The OD was read on a plate reader at 600 nm for the growth curve. The samples were emptied from the wells, and the wells were washed with running water and stained by the addition of 200 µl of 1% crystal violet solution for 20 min. The wells were then washed thoroughly with running water and allowed to dry. The dye was quantified after the addition of 200 µl of DMSO to each well and read at an OD of 580 nm.
Biofilm formation on glass tubes was assayed by inoculating each tube with 3 ml of T broth with bacterial sample at an OD600 of 0.0025. Each tube was allowed to sit for 24, 48 and 72 h, then the tubes were washed with water and stained with crystal violet as described above.
Flagellar swimming assays were performed by inoculating T-broth plates made with 0.3% Difco agar as described by O’Toole et al. (1999). Swimming was assessed after 16 h of growth at room temperature. Swimming was also tested at 37°C. A Tn5 transposon insertion in the P. aeruginosa PA14 flgI gene was isolated in the pellicle screen and used as a motility-defective negative control.
To assess twitching motility of P. aeruginosa PA14 wild-type and pellicle-defective mutants pelA, pelD, pelG and pelH, zones of twitching were measured and compared. The P. aeruginosa pilB and pilC mutants were used as controls (O’Toole and Kolter, 1998a). A sample of 5 ml of T broth containing 1.5% agar was poured into 15 mm plates. The plates were inoculated by stabbing through the agar with a pipette tip. After 48 h of growth at room temperature, the agar was removed, and the plate was stained with 1% crystal violet for 5 min. The plates were washed with water and allowed to air dry. Purple-stained zones represent the zones of twitching.
Scanning electron microscopy
Bacterial colonies were gently scooped from a plate and fixed in a solution of 2.5% glutaraldehyde, 2.5% paraformaldehyde in 1 mM CaCl2, 0.1 M cacodylate, pH 7.4, for 2 h at room temperature. The samples were rinsed once in the same buffer and post-fixed in 1% osmium tetroxide in 0.1 M cacodylate, pH 7.4, for 1 h at room temperature. Samples were dehydrated by increasing concentrations of ethanol (30%, 50%, 70%, 90%, 100%). The samples were dried at a critical point in a CO2 atmosphere. The dried samples were fixed to stubs with conductive self-sticking adhesive tabs, coated with gold film sputtering and used for analysis by SEM.
The wild-type samples stayed in a single piece; however, the mutant sample was spun at 3000 r.p.m. for 2 min as the cells instantly suspend in liquid.
Congo red assay
Tryptone broth (10 g l−1) without salt with the addition of Congo red (40 µg ml−1) and Coomassie brilliant blue (20 µg ml−1) was used to judge pellicle morphology and colour (Romling et al., 1998). The addition of 1.0% agar was used to create Congo red plates. Samples of 1, 5 and 10 µl of OD600 = 0.025 in T broth no salt were plated and grown at room temperature to assess colony morphology. Congo red-stained pellicles were grown in T-broth medium containing 40 µg ml−1 Congo red.
Cellulase and sodium periodate treatment
Sections of pellicles from a 100 ml standing culture of the respective wild-type or cupA2::Tn5 culture were rinsed in 0.05 M sodium acetate, pH 5.0, 0.1% sodium azide and 60 µg ml−1 gentamicin. The liquid was removed and replaced with the same medium with or without the addition of 5 mg ml−1 or 100 µg ml−1 cellulase (Sigma; EC 22.214.171.124 from Trichoderma viride). The samples were incubated at 37°C. Control samples containing Kimwipes (Kimberly Clark) were completely dissolved by 24 h. Samples were followed for 7 days checking every 12 h and repeated more than six times.
Sections of pellicles were placed in a solution of 10 mg ml−1 sodium m-periodate (Sigma S-1878) dissolved in dH2O. The wild-type P. aeruginosa PA14 pellicle dissolves after 5 min at room temperature.
Crude pellicle matrix isolation
Standing cultures with 1 l of T broth in a 2 l flask were inoculated with plate-grown bacteria to an OD600 = 0.0025. The cultures were left undisturbed at room temperature for 7 days. For the wild-type P. aeruginosa PA14 pellicle sample, a 50 ml pipette was used to gather the pellicle from the top of the culture removing as little medium as possible (<5 ml). The pellicle was washed once in sterile dH2O. An aliquot of 1 M NaOH was added, and the sample was vortexed every 2 min for 15 min. The sample was spun at 39 000 r.p.m. in an ultracentrifuge for 1 h at 4°C. The supernatant was removed and filtered through a 0.2 µm filter. The filtrate was neutralized with concentrated HCl, precipitated by the addition of ethanol to 70% and placed at −20°C overnight. The precipitate was collected by centrifugation at 9000 r.p.m. for 30 min at 4°C. The pellet was washed with 70% ethanol, allowed to dry for 45 min and resuspended in water, then lyophilized. The lyophilized material was resuspended in water, dialysed against water and lyophilized before carbohydrate analysis.
Culture comparison isolation procedure
Standing cultures with 1 l of T broth in a 2 l flask were inoculated with plate-grown bacteria to an OD600 = 0.0025. The cultures were left undisturbed at room temperature for 7 days. For the wild-type P. aeruginosa PA14 pellicle sample, a 50 ml pipette was used to gather the pellicle from the top of the culture removing as little medium as possible (<5 ml). The ΔpelA mutant sample was allowed to grow at the same time under identical conditions. As no pellicle forms in the ΔpelA mutant sample, the entire sample was centrifuged for 20 min at 7000 r.p.m. in a Sorval RC 5C plus the cells were transferred to a 50 ml Falcon tube. Both samples were washed five times with water by resuspending the pellet in water and then centrifuging for 7 min at 7000 r.p.m. The bacterial pellets were lyophilized overnight.
Carbohydrate composition and linkage analysis
Glycosyl composition analysis was performed by combined gas chromatography/mass spectrometry (GC/MS) of the per-O-trimethylsilyl (TMS) derivatives of the monosaccharide methyl glycosides produced from the sample by acidic methanolysis.
Methyl glycosides were first prepared from 0.5 mg of dry sample by methanolysis in 1 M HCl in methanol at 80°C (18–22 h), followed by re-N-acetylation with pyridine and acetic anhydride in methanol (for the detection of amino sugars). The samples were then per-O-trimethylsilylated by treatment with Tri-Sil (Pierce) at 80°C (0.5 h). These procedures were carried out as described previously (Merkle and Poppe, 1994; Wozniak et al., 2003). GC/MS analysis of the TMS methyl glycosides was performed on an HP 5890 GC interfaced to a 5970 MSD, using a Supelco EB1 fused silica capillary column. Sample analysis was performed by the Complex Carbohydrate Research Center at the University of Georgia, Athens, GA, USA. Sample analysis was performed in triplicate for both wild-type pellicle and the ΔpelA mutant. Sample analysis was performed in duplicate for the crude matrix preparation.
Glycosyl linkage analysis was performed as described BY Wozniak et al. (2003) AT the Complex Carbohydrate Research Center at the University of Georgia, Athens, GA, USA.
We thank R. Schalek for performing the scanning electron microscopy, P. Yorgey and H. Schweizer for strains, D. Fraenkel for helpful discussions, and Kolter laboratory members for helpful discussions and comments on the manuscript. Special thanks to J., E. and H. Friedman for editorial and loving support. This work was supported by grants from the NIH (GM58213), CFF (LORY00V0), DOE (DE-FG02-02ER63445) and the Ellison Medical Foundation (ID-SS-0248-02). L.F. was the recipient of a Canadian Institutes of Health Research (CIHR) Postdoctoral Fellowship. The SEM work was performed at the Center for Imaging and Mesoscale Structures, Harvard University, from instrumentation funded by NSF grant number 0099916, Division of Biological Infrastructure. The carbohydrate analysis was done by the Complex Carbohydrate Research Center, University of Georgia by P. Azadi and coworkers who are supported in part by the Department of Energy-funded (DE-FG02-93ER-20097) Center for Plant and Microbial Complex Carbohydrates.