We have used Escherichia coli as a model system to investigate the initiation of biofilm formation. Here, we demonstrate that E. coli forms biofilms on multiple abiotic surfaces in a nutrient-dependent fashion. In addition, we have isolated insertion mutations that render this organism defective in biofilm formation. One-half of these mutations was found to perturb normal flagellar function. Using defined fli, flh, mot and che alleles, we show that motility, but not chemotaxis, is critical for normal biofilm formation. Microscopic analyses of these mutants suggest that motility is important for both initial interaction with the surface and for movement along the surface. In addition, we present evidence that type I pili (harbouring the mannose-specific adhesin, FimH) are required for initial surface attachment and that mannose inhibits normal attachment. In light of the observations presented here, a working model is discussed that describes the roles of both motility and type I pili in biofilm development.
Genetic and molecular approaches to understanding the growth, metabolism, adaptability and physiology of bacteria have predominantly focused on studies of planktonic cells living in batch culture. It is important to note that, although these studies have provided extensive information describing the basic molecular mechanisms controlling bacterial growth, many bacteria live primarily in sessile communities, referred to as biofilms. The term biofilm is used to describe matrix-enclosed bacterial populations adherent to each other and/or to surfaces (Costerton et al., 1995). As many of the world's bacteria live within biofilms, it is critical to our understanding of bacterial life to understand the development of these communities. Biofilms can become hundreds of microns in depth, are difficult to treat with antibiotics and can clog industrial pipes (Hoyle and Costerton, 1991; Costerton et al., 1995; Finlay and Falkow, 1997). Thus, gaining a molecular understanding of the regulation and formation of biofilms has industrial, medical as well as ecological relevance.
Over the past two decades, much has been learned about the structures of diverse biofilms. Numerous bacterial species have been reported to form biofilms, and it seems clear that bacterial biofilm formation is favoured under most nutrient-sufficient environments. In addition, biofilm structure has been shown to be extremely complex, consisting of more than bacteria simply adhered to a surface (Costerton et al., 1995). Biofilms can be formed by single or multiple species (Costerton et al., 1995), and Pseudomonas aeruginosa biofilm bacteria form an extensive exopolysaccharide matrix (Davies et al., 1993; Hoyle et al., 1993; Davies and Geesey, 1995). It has been proposed that biofilms reflect complex communities of microcolonies separated by water-filled channels. These channels have been compared to circulatory systems, leading to the emerging view that biofilms can be considered primitive multicellular organisms (Costerton et al., 1995).
In contrast to the extensive characterization of biofilm structure described above, little is understood about the molecular details of biofilm formation. To gain a molecular understanding of the events leading to the initiation of biofilm formation we asked the following questions. What factors enable a bacterium to first reach a surface? Once reached, what factors enable a bacterium to adhere to the surface? After initial bacterium-surface contact is established, what gene products are required for the development of a mature biofilm? To begin to answer these questions, we have isolated and characterized mutants of Escherichia coli that are defective in biofilm formation on an abiotic surface. Here, we describe the characterization of a subset of the mutants isolated and present a model describing the role of motility and of type I pili in the first stages of biofilm development.
Escherichia coli forms biofilms in a nutrient-dependent fashion
We tested the ability of the well-characterized, Gram-negative bacterium Escherichia coli to initiate biofilm formation on abiotic surfaces. To assay for such attachment, we used a modified version of a previously described protocol (Fletcher, 1977). Cells were first grown for either 24 or 48 h at room temperature without shaking in microtitre dishes or glass tubes. To remove any unattached cells, the microtitre dishes (or glass tubes) were rinsed thoroughly with water and subsequently stained with 1.0% crystal violet (CV) for ≈20 min. This staining procedure allowed us to visualize cells that had attached to the surface as such cells are stained purple with CV, whereas abiotic surfaces are not stained with CV. We found that a number of motile laboratory strains of E. coli were able to attach to multiple abiotic surfaces when grown in Luria Bertani broth (LB). Specifically, E. coli 2K1056 formed biofilms on all surfaces tested, including polyvinyl chloride (PVC), polypropylene, polycarbonate, polystyrene and borosilicate glass.
Importantly, the ability to form such biofilms was strongly influenced by the nutritional environment. Biofilm formation could be visualized with CV after as little as 2 h of growth in LB (data not shown). Similarly, biofilm formation was supported by various minimal media containing casamino acids (CAA) (Fig. 1). In contrast, minimal media without CAA (with either glucose or glycerol as a carbon and energy source) did not support biofilm formation that was visible after staining with CV (Fig. 1).
Screen for Escherichia coli mutants defective in biofilm formation
To identify genes required for biofilm formation, we screened for mutants defective in forming biofilms in LB on PVC plastic. Strain 2K1056 was subjected to insertion mutagenesis (Kleckner et al., 1991) with a mini Tn10cam, and insertion mutants were selected on LB agar containing 30 μg ml−1 chloramphenicol. Chloramphenicol-resistant colonies were picked and grown at room temperature in 96-well PVC microtitre dishes containing glucose minimal medium with 30 μg ml−1 chloramphenicol. After 48 h, the cells were subcultured into corresponding wells in a 96-well PVC microtitre dish containing LB with 30 μg ml−1 chloramphenicol. The cultures were grown at room temperature for another 48 h and then rinsed thoroughly with water to remove any planktonic cells. The wells were stained with CV as described in Experimental procedures, rinsed and potential biofilm-defective mutants were chosen based on decreased staining compared with a wild-type control. Any potential biofilm-defective mutants were isolated from the original, corresponding well in the microtitre dish containing glucose minimal medium with chloramphenicol. These candidate mutant strains were streaked for single colonies on LB agar and retested for their ability to form biofilms. Each of the insertion mutations that appeared to confer a defect in biofilm formation was transferred into a fresh 2K1056 background via P1vir transduction and retested. Of 10 000 such insertion mutations analysed, 177 were found to confer a decrease in biofilm formation.
Initial classification and mutant identification
It is possible that a mutant strain isolated in the above screen could exhibit decreased biofilm formation because it harbours a mutation that either (i) confers a non-specific growth defect that indirectly affects biofilm development or (ii) interferes in the formation of biofilms without interfering with the growth rate. To distinguish between these possibilities, mutant strains were grown in LB and their growth rates were compared with the wild type. Only strains exhibiting growth rates indistinguishable from the wild type are discussed in the following sections.
The mutant strains displayed a wide array of phenotypes with respect to the severity in their decreased ability to form biofilms. The macroscopic phenotypes ranged from wells that displayed subtle decreases in CV staining to wells that appeared completely clear after CV treatment. Previous studies with biofilm-forming bacteria have implicated flagella and/or motility and/or chemotaxis as important for biofilm formation (Lawrence et al., 1987; Korber et al., 1989; 1994; DeFlaun et al., 1994; Graf et al., 1994; Mills and Powelson, 1996; Vidal et al., 1998). Thus, as an early step in characterization of the mutants, each was analysed for its ability to swarm on LB motility agar (0.3% agar). Approximately one-half of the mutants (87 out of 177) displayed a decreased ability to swarm, whereas the remaining mutants formed swarms that were indistinguishable from the wild type. The majority of the Swarm− mutants were severely defective in their ability to form biofilms (i.e. clear wells after staining with CV). Such swarm assays do not always allow one to distinguish between defects in flagellar biosynthesis, motility and/or chemotaxis. Thus, the following central question arose. Which of these three aspects of bacterial flagella/movement is critical to biofilm formation?
Among the remaining Swarm+ mutants, 23 displayed macroscopic phenotypes similar to those observed with Swarm− mutants (i.e. clear wells after staining with CV; see examples of mutants that display the clear well phenotype in Fig. 2), whereas the others displayed less severe phenotypes. For the purposes of this initial study, we focused on the 23 Swarm+ mutants with the strongest phenotypes. Genetic linkage analysis of these 23 mutants was performed by using a P1vir lysate that had been grown on a pool of cells containing transposons randomly inserted throughout the chromosome (Kleckner et al., 1991). Through this approach, a Tn10 was isolated that was tightly linked to all 23 Swarm+ mutants. The precise locations of eight of the 23 insertion mutations within this linkage group were identified using arbitrarily primed PCR (Caetano-Annoles, 1993) followed by DNA sequence analysis (see Experimental procedures). All eight insertions were located in genes encoding for the regulation or synthesis of type I pili. Specifically, independently isolated insertions were found in fimB (two alleles), fimA, fimC, fimD (three alleles) and fimH (Fig. 2). Thus, a second question arose. What is the role of type I pili in Escherichia coli biofilm formation?
Motility, not chemotaxis, is critical for biofilm formation
Although flagella, motility and/or chemotaxis have been implicated as important for biofilm formation in other organisms (Lawrence et al., 1987; Korber et al., 1989; 1994; DeFlaun et al., 1994; Graf et al., 1994; Mills and Powelson, 1996; Vidal et al., 1998), their precise roles in this process have not yet been defined. We reasoned that there are three mechanisms through which flagella might be required. First, it is possible that flagella could be directly required for attachment to abiotic surfaces, thus facilitating the initiation of biofilm formation (e.g. as with tethered cells). Alternatively, motility could be necessary to enable a bacterium to reach the surface (e.g. to move through surface repulsion present at the medium–surface interface). Also, motility might be required for the bacteria within a developing biofilm to move along the surface, thereby facilitating growth and spread of the biofilm. Finally, it is possible that chemotaxis is required for the bacteria to swim towards nutrients associated with a surface.
As flagellar synthesis, motility and chemotaxis have been extensively studied in E. coli (Macnab, 1996; Stock and Surette, 1996), well-defined mutations that inhibit each of these three aspects of flagellar function are available. Accordingly, we obtained the following mutations: (i) fliC ::kan (strains harbouring this allele are unable to synthesize flagellin) and flhD ::kan (a master regulator of flagellar synthesis whose absence confers an inability to synthesize flagella); (ii) ΔmotA, ΔmotB and ΔmotAB (lesions that do not inhibit flagellar biosynthesis but render cells non-motile or paralysed); and (iii) ΔcheA-Z ::kan (strains harbouring this lesion are motile but non-chemotactic).
Each of these alleles was moved into 2K1056 via P1vir transduction, and the resulting strains were analysed for their ability to form biofilms. Construction of these strains provided us with the tools required to distinguish between the possible roles of flagella/motility/chemotaxis that were detailed above. First, the simple microtitre dish assay used in our screen revealed that motile cells that are non-chemotactic (ΔcheA-Z ::kan) appear to form biofilms indistinguishable from their wild-type counterpart. In contrast, cells either lacking flagella (fliC ::kan, flhD ::kan) or possessing paralysed flagella (ΔmotA, ΔmotB, or ΔmotAB ) were severely defective in biofilm formation (Fig. 2). When biofilm formation was quantified over time (Experimental procedures), it became very clear that, under these conditions, chemotaxis is completely dispensable for normal biofilm formation (Fig. 3). In contrast, cells either lacking complete flagella (fliC ::kan) or possessing paralysed flagella (ΔmotA, ΔmotB, or ΔmotAB ) are severely hindered in the initial stages of biofilm formation (Fig. 3).
More detailed analysis of the defects conferred by these alleles was obtained through microscopic analysis of cells attached (or the absence of such attached cells) to PVC after growth in LB. As illustrated in 4Fig. 4A and B, motile cells that are non-chemotactic are able to form biofilms that are indistinguishable at the cellular level from the biofilms formed by wild-type cells. In contrast, non-flagellated or paralysed cells attach poorly to PVC. Moreover, the few cells that do attach are often located in small, dense clusters of cells (Fig. 4D).
Type I pili are critical for initial attachment to abiotic surfaces
As mentioned above, the macroscopic analysis of biofilm formation of fim mutants was analogous to that observed with the motility-defective mutants (i.e. clear wells after staining with CV) (Fig. 2). However, microscopic analysis of these mutants revealed distinct phenotypes. Specifically, fim mutants are even more dramatically defective in initial attachment than are the paralysed and non-flagellated cells. As illustrated in 4Fig. 4C, most microscopic fields had no cells attached at all, and only infrequently were a few attached cells observed. This observation indicated that type I pili are critical for initial interaction with abiotic surfaces such as PVC.
α-Methyl-D-mannoside inhibits attachment to abiotic surfaces
One of the insertions in the fim gene cluster is located in the final gene of the operon, fimH. Lesions in fimH have been reported to affect the length of the tip (fibrilla) of type I pili (Ottemann and Miller, 1997a). In addition, FimH functions as a mannose-specific adhesin, allowing E. coli to interact specifically with mannose residues on eukaryotic cells facilitating infections such as cystitis (Old, 1972; Maurer and Orndorff, 1987; Hanson and Brinton, 1988; Low et al., 1996). Consequently, it is possible that the altered structure of the fibrilla of type I pili in fimH mutants could interfere with normal attachment to abiotic surfaces. Alternatively, the mannose-specific adhesin may play a more direct role in attachment. To address further the role of FimH in biofilm formation, we tested whether the presence of a non-metabolizable mannose analogue, α-methyl-D-mannoside, affected the ability of the wild-type strain, 2K1056, to form biofilms on PVC. As illustrated in Fig. 5, α-methyl-D-mannoside inhibits biofilm formation in a concentration-dependent fashion. Importantly, α-methyl-D-mannoside does not inhibit growth rates (data not shown). As a specificity control, we have shown that although mannose also has a similar effect as α-methyl-D-mannoside, glucose does not inhibit biofilm formation, and neither mannose nor glucose inhibits growth (data not shown). It is also important to note that α-methyl-D-mannoside inhibits biofilm development on all other abiotic surfaces tested, including polycarbonate, polystyrene and borosilicate glass (data not shown). It is reasonable to assume that these various surfaces do not resemble mannose.
We have used the well-studied and genetically tractable organism, E. coli, to rapidly identify genes required for the initial stages of biofilm formation. The mutations isolated affect factors required for flagellar biogenesis, motility and the regulation and biogenesis of type I pili. It is well established that flagellar-mediated motility and the ability to produce a number of pili contribute to the virulence of pathogenic bacteria (Finlay and Falkow, 1997; Ottemann and Miller, 1997b). This leaves us with the suggestive overlap of functions essential for both biofilm formation and functions needed for pathogenicity. In this regard, screens such as the one described here may prove useful in the identification of gene products important for the pathogenicity of a variety of bacteria. In addition, the work with E. coli may serve as a paradigm for the study of bacteria less amenable to genetic and molecular approaches. Although we predict similarities in the molecular mechanisms used by other biofilm-forming bacteria, distinguishing details have begun to arise (O'Toole and Kolter, 1998a,b). Such distinctions should be especially informative about the particular mechanisms used by bacteria that occupy diverse environmental niches.
Despite prior implications in biofilm function, the particular aspect(s) of flagellar biogenesis and function that is needed for biofilm formation in any species of bacteria had not yet been clearly defined (Lawrence et al., 1987; Korber et al., 1989; 1994; DeFlaun et al., 1994; Graf et al., 1994; Mills and Powelson, 1996; Vidal et al., 1998). Flagella could potentially perform four, non-mutually exclusive roles: (i) flagellar-mediated chemotaxis could function to enable planktonic cells to swim towards nutrients associated with a surface or towards signals generated by cells attached to an abiotic surface; (ii) flagellar-mediated motility could enable bacteria to initially reach a surface, perhaps by overcoming repulsive forces at a surface; (iii) flagellar-mediated motility could enable attached, dividing bacteria to spread along a surface; and (iv) flagella could function in a direct fashion by physically adhering to an abiotic surface.
We have made the surprising discovery that, under the conditions used in this study, chemotaxis is dispensable for the initiation of E. coli biofilm formation. In contrast, motility is critical for normal biofilm formation. This result is illustrated in Figs 2–4, as cells defective in motility attach poorly to PVC, and the few cells that do attach are often located in small, dense clusters (Fig. 4D).
As chemotaxis is dispensable for the initial stages of biofilm development, it may be that motility is required to overcome repulsive forces at the surface–medium interface. In addition, the observation of small clusters of cells from the paralysed (Fig. 4D) or non-flagellated (data not shown) strains suggests that motility may also play an important role in the initial spread of a biofilm by facilitating movement of growing cells along an abiotic surface.
Although it is possible that flagella also play a direct role in adhering to abiotic surfaces, several observations indicate that these structures are not the primary factors involved in this process. First, there is no phenotypic difference observed in the attachment between paralysed cells and non-flagellated cells. Moreover, the absence of fully functional type I pili results in virtually no cells attaching to the surface; this microscopic phenotype is markedly more dramatic than that displayed by cells defective in flagellar biosynthesis or function (Fig. 4C and D). Therefore, if flagella do function directly in adhering to abiotic surfaces, this interaction depends largely on the integrity of type I pili.
In sum, our data suggests a dual role for flagellar-mediated motility in E. coli biofilm development. We propose that, under the conditions used in these studies, motility promotes initial cell-to-surface contact and may also contribute to the spread of a growing biofilm along an abiotic surface.
As previously noted, flagella, motility and/or chemotaxis have been implicated as important for biofilm formation in other organisms (Lawrence et al., 1987; Korber et al., 1989; 1994; DeFlaun et al., 1994; Graf et al., 1994; Mills and Powelson, 1996). However, the lack of molecular characterization of these strains left the possibility that these strains contained pleiotropic defects. Also, without a molecular description of the lesions conferring biofilm defects, it is difficult to clearly define the roles (adherence, motility and/or chemotaxis) that flagella play in biofilm development. O'Toole and Kolter (1998a,b) have isolated a non-motile strain of P. aeruginosa that contains an insertion mutation in a flgK homologue and non-motile strains of Pseudomonas fluorescens (fliP and flaE ); these strains are unable to form biofilms. Although it is not yet possible to define which aspect(s) of flagellar structure/function are important in biofilm development for these organisms, the isolation and characterization of these alleles confirm earlier hypotheses that flagella are indeed required for P. aeruginosa and P. fluorescens biofilm development.
Importantly, the phenotypes conferred by the flagellar mutations in P. aeruginosa and P. fluorescens are distinct from the phenotypes conferred by flagellar mutations in E. coli. Under certain conditions, some aspect of flagella, motility and/or chemotaxis is essential for both P. aeruginosa and P. fluorescens to reach and/or interact with an abiotic surface. For example, a P. aeruginosa flgK mutant does not attach to PVC and no microcolonies are formed (O'Toole and Kolter, 1998a). Recall that cultures of E. coli flagellar mutants do attach to PVC, albeit poorly, and some microcolonies are formed. Thus, although flagella are important for biofilm formation in E. coli, P. aeruginosa, and P. fluorescens, the role flagella play are clearly different. It is noteworthy that P. aeruginosa possess a single, polar flagellum and swim in a straight/stop/turn mode (Takahito and IIno, 1980), whereas E. coli possesses peritrichous flagella (about 5–10 flagella per cell originating at random points on the cell surface) and moves via a random walk involving alterations between smooth swimming and tumbling (Stock and Surette, 1996). Given the very different nature of the flagella present on E. coli versus P. aeruginosa, it is not surprising that the roles these structures play in biofilm development are quite distinct.
Our findings reveal that the presence of type I pili is essential for the initial attachment of E. coli. Cells harbouring lesions in genes encoding for the regulation or biogenesis of type I pili simply do not attach to PVC. In addition, we discovered that attachment is inhibited by the presence of mannose or α-methyl-D-mannoside. Type I pili contain the mannose-specific adhesin, FimH, which plays a role in facilitating pathogenesis through specific interactions between FimH and mannose oligosaccharides present on eukaryotic cell surfaces (Old, 1972; Hanson and Brinton, 1988; Low et al., 1996). The observation that FimH is also critical for attachment to abiotic surfaces was surprising and leads us to assign a novel role to type I pili. We propose that the FimH–surface interaction is direct and involves a region of FimH involved in ‘non-specific’ binding to abiotic surfaces. If this is the case, then the binding of mannose to FimH may somehow alter its conformation, masking the FimH region that interacts with abiotic surfaces.
The mannose inhibition of E. coli biofilm formation on abiotic surfaces may have general applications to other biofilm-forming bacteria. Bacteria that form biofilms on surfaces in medically and/or industrially relevant environments may also require the integrity of adhesins analogous to E. coli 's requirement for FimH. Thus, it is possible that the formation of problematic biofilms could be blocked through treatment with innocuous materials such as mannose.
The observations described here lead to the following model to describe the initiation of E. coli biofilm formation in rich media. Motility, but not chemotaxis, is important for cells to first come in contact with an abiotic surface. This requirement may reflect a necessity to overcome repulsive forces present at an abiotic surface to be colonized. Once a surface is reached, type I pili are required to achieve stable cell-to-surface attachment. The presence of the FimH adhesin, when it is not bound to mannose, promotes such stable adherence to abiotic surfaces. Finally, we hypothesize that motility facilitates the development of a mature biofilm by allowing movement along a surface, thereby promoting spread of the biofilm (Fig. 6).
O'Toole and Kolter (1998a) have found insertion mutations in genes required for functional type IV pili that interfere with normal P. aeruginosa biofilm formation. Importantly, the phenotype associated with P. aeruginosa strains lacking type IV pili differs dramatically from E. coli mutants lacking type I pili. Specifically, P. aeruginosa strains lacking type IV pili are able to form a monolayer of cells attached to PVC, but do not proceed past this stage. Unlike their wild-type counterparts, the cells do not develop into a multilayered biofilm and no microcolonies are formed. These observations have led O'Toole and Kolter to propose the following role to type IV pili in P. aeruginosa biofilm development. Unlike type I pili in E. coli, P. aeruginosa type VI pili are not essential for initial attachment to PVC. Instead, the twitching motility associated with the presence of type IV pili is proposed to facilitate movement along an abiotic surface. This twitching-mediated motility contributes to the formation of microcolonies within a developing biofilm.
Vent exo− DNA polymerase, β-agarase and associated buffers were obtained from New England Biolabs. NuSieve GTG low-melting-temperature agarose was purchased from FMC Bioproducts. All primers were purchased from Genosys. Crystal violet and α-methyl-D-mannoside were obtained from Sigma. Ninety-six-well polycarbonate dishes were purchased from USA Scientific; 96-well Pro-bind Assay Plates, non-tissue culture-treated polystyrene dishes were purchased from Becton Dickenson; 96-well polyvinyl chloride dishes (Falcon 3911 microtest III flexible assay plates) were also obtained from Becton Dickenson; polypropylene 1.5 ml Eppendorf tubes were provided by Marsh Biomedical Products; and borosilicate glass culture tubes were purchased from VWR.
Bacterial strains, bacteriophage, media and genetic techniques
2K1056 (E. coli K-12 F− λ−IN(rrnD–rrnE)1 rph-1Δ(argF-lac) U169) was used as the parental strain; all strains described are either 2K1056 or derivatives of this strain. The media used have been previously described (Pardee et al., 1959; Silhavy et al., 1984), and casamino acids were used at a concentration of 0.5%. Generalized transduction using P1vir was performed as previously described (Silhavy et al., 1984).
After strain construction involving alleles that affect flagella, motility and/or chemotaxis, the presence (or absence) of flagella was confirmed using a simple staining procedure that has been described previously (Heimbrook et al., 1989). Motility and chemotaxis were analysed using both swarm assays (Adler, 1966; Wolfe and Berg, 1989) and phase-contrast microscopy of living cells. λNK1324 was used for insertion mutagenesis of 2K1056 as previously described (Kleckner et al., 1991).
Quantification of CV-stained attached cells and growth curves
Attached cells were quantified as described previously with a few modifications (Genevaux et al., 1996; O'Toole and Kolter, 1998b). After wells had been stained with 125 μl of 1.0% CV, rinsed and thoroughly dried, the CV was solubilized by the addition of 200 μl of ethanol–acetone (80:20). An 80 μl sample of the solubilized CV was removed and added to a fresh polystyrene 96-well dish, and OD600 or OD570 was determined using either a Series 700, Microplate Reader from Cambridge Technology or an MR 700 Microplate Reader from Dynathech Laboratories.
Growth curves were determined by subculturing (1:100) the relevant strain into the appropriate medium and growing the culture at room temperature without shaking. OD600 readings were taken over time using a spectronic 20D+ from Spectronic Instruments.
PCR amplification and DNA sequence analysis
The locations of transposon insertion mutations were identified using a previously described arbitrarily primed PCR method (Caetano-Annoles, 1993), with a few minor modifications (O'Toole and Kolter, 1998b). This method involves two rounds of PCR amplification using arbitrary primers to prime from the chromosome and primers specific to the minitransposon (Caetano-Annoles, 1993; O'Toole and Kolter, 1998b). The first round of PCR reactions used the following primers: ARB1 (GGCCACGCGTCGACTAGTACNNNNNNNNNNGATAT) or ARB6 (GGCCACGCGTCGACTAGTACNNNNNNNNNNACGCC) and OUT1-L (CAGGCTCTCCCGTGGAG). The second round of PCR reactions used the following primers: ARB2 (GGCCACGCGTCGACTAGTAC) and PRIMER1L (CTGCCTCCCAGAGCCTG).
After the second round of PCR amplification, PCR products were separated a 1.0% low-melt agarose gels, and bands were excised from the gel. The agarose was digested with β-agarase, and the DNA was subjected to DNA sequence analysis uisng PRIMER1L. Sequence analysis was carried out at the Biopolymers Laboratory of the Department of Biological Chemistry and Molecular Pharmacology of Harvard Medical School.
We thank Paul Danese, Howard Berg, Mike Manson and Mike Surette and for strains. We are grateful to members of the Kolter laboratory for helpful discussions and to Paul Danese and George O'Toole for critical reading of the manuscript. This work was supported by NIH GM58213 to R. Kolter and a Fellowship from The Jane Coffin Childs Memorial Fund for Medical Research to L. Pratt.