Several pathogens associated with chronic infections, including Pseudomonas aeruginosa in cystic fibrosis pneumonia, Haemophilus influenzae and Streptococcus pneumoniae in chronic otitis media, Staphylococcus aureus in chronic rhinosinusitis and enteropathogenic Escherichia coli in recurrent urinary tract infections, are linked to biofilm formation. Biofilms are usually defined as surface-associated microbial communities, surrounded by an extracellular polymeric substance (EPS) matrix. Biofilm formation has been demonstrated for numerous pathogens and is clearly an important microbial survival strategy. However, outside of dental plaques, fewer reports have investigated biofilm development in clinical samples. Typically biofilms are found in chronic diseases that resist host immune responses and antibiotic treatment and these characteristics are often cited for the ability of bacteria to persist in vivo. This review examines some recent attempts to examine the biofilm phenotype in vivo and discusses the challenges and implications for defining a biofilm phenotype.
Bacteria in most environments form organized communities of aggregated cells embedded in a hydrated matrix of extracellular polymeric substances (EPS) called biofilms (Costerton et al., 1999; Donlan and Costerton, 2002). Biofilm development is an ancient prokaryotic adaptation (Hall-Stoodley et al., 2004) and represents a mode of growth that allows bacteria to survive in hostile environments and to colonize new niches by various dispersal mechanisms (Hall-Stoodley and Stoodley, 2005; Purevdorj-Gage et al., 2005; Mai-Prochnow et al., 2008). Biofilm bacteria demonstrate coordinated behaviour with the formation of complex three-dimensional structures (Fig. 1) and functionally heterogeneous bacterial communities (Stoodley et al., 2002; Hall-Stoodley et al., 2004). This phenotypic heterogeneity, or localized specialization in biofilms is striking. For example, populations of bacteria within biofilms exhibit differences in the expression of surface molecules, antibiotic resistance, nutrient utilization and virulence factors (Bagge et al., 2004; Vuong et al., 2004; Pearson et al., 2006; Jurcisek and Bakaletz, 2007; Lenz et al., 2008; Zhang and Mah, 2008). Bacteria in biofilms also coordinate behaviour by cell–cell communication using secreted chemical signals. Cell signalling allows bacteria to sense and phenotypically respond to their environment, for example, by assessing cell density (also called quorum sensing or QS) or environmental cues. In Pseudomonas aeruginosa QS has been linked with biofilm formation and the production of virulence factors (Jensen et al., 2007). This extraordinary ability to adapt to and modify different microniches allows bacteria in biofilms to weather diverse and stressful conditions facilitating survival at a population level.
Biofilm infections are recalcitrant to antibiotic treatment and host defences
Although the matrix may not inhibit the penetration of antibiotics into the biofilm altogether, it may retard the rate of penetration enough to induce the expression of genes within the biofilm that mediate resistance (Jefferson et al., 2005). Also, the charge of polymers (Walters et al., 2003) and antibiotic-degrading enzymes (Bagge et al., 2004) in the matrix may lead to binding and/or deactivation. Naturally dead cells within the biofilm may even ‘dilute’ antibiotics on a per cell basis (Mai-Prochnow et al., 2008). Gillis et al. (2005) showed that unlike planktonic cells, P. aeruginosa in biofilms were resistant to azithromycin (AZM) due to an efflux pump, which was only expressed in the presence of the antibiotic. Different genes in the efflux pump operon were expressed in different parts of the biofilm and disruption of two gene clusters was required to induce AZM tolerance indicating that P. aeruginosa has redundant mechanisms for AZM resistance. Furthermore the efflux pump did not affect the ability of P. aeruginosa to resist other antibiotics such as tobramycin. However, tolerance to tobramycin, gentamicin and ciprofloxacin was recently linked to another efflux pump expressed to a greater extent in biofilms (Zhang and Mah, 2008). Thus, there appear to be multiple mechanisms for antibiotic recalcitrance in biofilm cells, depending on the type of bacterium and the type of antibiotic.
In addition to reduced antibiotic susceptibility, biofilms exhibit resistance to host phagocytic defences. Early evidence of resistance came from transmission electron micrographs of clinical specimens, which showed biofilms surrounded, but not penetrated, by antibodies and inflammatory cells (Lam et al., 1980). Several recent studies support the resistance-to-phagocytes hypothesis. Specifically, leukocytes penetrated channels, but not bacterial cells within the EPS matrix in Staphylococcus aureus biofilms and exhibited an unstimulated morphology with no evidence of motility or phagocytosis (Leid et al., 2002). In Staphylococcus epidermidis biofilms, protection from phagocytosis by polymorphic neutrophils (PMNs) and innate immune mechanisms was mediated by the intracellular adhesin, PIA, controlled by the ica locus (Vuong et al., 2004). PMNs were also immobilized in the EPS matrix in P. aeruginosa biofilms, and although phagocytosis occurred, PMNs exhibited diminished oxidative potential measured by the generation of H2O2 (Jesaitis et al., 2003). This effect may be QS-dependent, since PMNs penetrated a QS-deficient P. aeruginosa biofilm but did not penetrate a wild-type biofilm (Bjarnsholt et al., 2005). Furthermore, alginate was shown to protect P. aeruginosa in biofilms from phagocytosis by IFN-γ-activated macrophages (Leid et al., 2005). Recently QS-induced P. aeruginosa production of a PMN toxin, rhamnolipid B, was shown to kill neutrophils (Jensen et al., 2007). Moreover, when biofilms of P. aeruginosa and S. aureus were observed in chronic wounds, host phagocytic cells did not penetrate bacterial aggregates, with P. aeruginosa distinctly sequestered from host cells (Kirketerp-Møller et al., 2008). Interestingly, the presence of PMNs augmented P. aeruginosa biofilm development and both DNA and actin polymers facilitated binding and matrix cohesion likely mediated by type IV pili binding to DNA (Walker et al., 2005; Allesen-Holm et al., 2006). These findings suggest that P. aeruginosa may opportunistically take advantage of ‘neutrophil extracellular traps’ (NETs), which are thought to trap bacterial cells (Brinkmann et al., 2004). PMNs also enhanced the ability of Candida albicans to form biofilms, while planktonic cells were vulnerable to this host response (Chandra et al., 2007). In Escherichia coli, Antigen 43, a surface protein that mediates aggregation, regulated both uptake by PMNs and its intracellular survival (Fexby et al., 2007).
Moreover the contribution of antibody in the protection against biofilm aggregates may be inadequate. Diffusion studies showed that IgG remained bound to the periphery of biofilm clusters, and failed to penetrate the EPS matrix (de Beer et al., 1997) or failed to mediate opsonic killing of biofilm cells (Cerca et al., 2006). However, antibody to an outer membrane protein in P. aeruginosa was shown to inhibit biofilm formation by interfering with adhesion to the surface (Tashiro et al., 2008), suggesting that antibody may play a role in preventing biofilm development. Antibodies also inhibited biofilm formation by the yeast Cryptococcus neoformans (Martinez and Casadevall, 2005).
Finally, in a report that studied polymicrobial interactions and innate immunity, exposure to H2O2 produced by Streptococcus gordonii induced an outer membrane protein on another oral pathogen, protecting it from killing by human serum during co-culture (Ramsey and Whiteley, 2009). Co-culture also protected Streptococcus pneumoniae against β-lactam antibiotics in a polymicrobial biofilm containing a β-lactamase-producing strain of Moraxella catarrhalis (Budhani and Struthers, 1998). The polymicrobial nature of oral biofilms associated with dental plaque and periodontitis has made oral biofilms a pioneering model of interspecies interactions and highlights yet another level of complexity in biofilm research (Kuramitsu et al., 2007).
The in vitro study of biofilms is important because it allows the investigation of complex bacterial phenotypes that are not readily interpreted in an in vivo system, including how specific molecular mechanisms affect biofilm development and how biofilms resist killing. Although clinical specimens rarely demonstrate a specific mechanism, they remain important in the examination of biofilm infections in the context of human disease. For example, fluorescent in situ hybridization (FISH) was used to determine the spatial distribution of P. aeruginosa in the sputum of cystic fibrosis (CF) patients (Yang et al., 2008). Patients colonized with mucoid variants demonstrated actively growing cell clusters encased in an extracellular matrix in the airway mucus while sputum from patients colonized with non-mucoid strains indicated that P. aeruginosa was distributed as dispersed cells. This study indicates that biofilms may not always be attached to a surface within the host, but may be associated with a surrounding mucus layer or fluid. For example, in a case history reporting evidence of biofilms in a failed elbow arthroplasty, the most extensive S. aureus cell clusters were found in aggregates in the joint fluid (Stoodley et al., 2008b). It was not possible to tell whether the biofilm aggregates had grown in the fluid or had detached during surgery from a surface-associated biofilm. Similar detached biofilm aggregates were observed from biofilms of a methicillin-sensitive strain of S. aureus (MSSA) in an in vitro catheter model (Fux et al., 2004). The decreased susceptibility to oxacillin of these aggregates suggested that bacteria exhibited a biofilm phenotype, and in terms of the development of gradients of consumed nutrients and waste products creating microniche environments, as well as gradients of secreted diffusible signal molecules, bacterial aggregates will almost certainly behave functionally as attached biofilms. Thus aggregates of bacteria seen in clinical specimens may have different origins: growth of aggregates in a fluid, or as naturally detached emboli from biofilms. In support of this, P. aeruginosa formed biofilm macrocolonies more readily in dehydrated mucus and mucin stimulated aggregation and biofilm formation (Landry et al., 2006; Matsui et al., 2006). It is possible that aggregation, uncoupled from attachment per se, may influence host–pathogen interactions. For example C. neoformans exited macrophages as biofilm-like aggregates, which may protect against dissemination, but increase the potential for localized inflammatory damage (Alvarez et al., 2008).
Biofilms in human disease and the criteria for biofilm infections
The complexity of biofilm research resulting from the diversity of microorganisms in different adaptive environments and the intricacies of biofilms at the host–pathogen interface defies a simple restrictive definition. Nevertheless, biofilm formation is now recognized as causing or exacerbating numerous chronic infections. These include periodontitis, device-related infections, CF pneumonia, chronic urinary tract infections (UTI), recurrent tonsillitis, chronic rhinosinusitis, chronic otitis media (OM) and chronic wound infections. Thus, biofilm infections add a further level of complexity to the existing multifactorial process of biofilm development and the disease process itself. As with any infectious disease, biofilm infections reflect the interplay between pathogen and host genetic programmes. Many biofilm infections occur at epithelial sites within the host and it is useful to first draw a distinction between biofilm infections and colonization. Biofilm infections are associated with pathogenic or opportunistic bacteria coupled with a chronic condition (i.e. the infection is recurrent or long-lasting in spite of host immune responses and antibiotic therapy). Organisms such as P. aeruginosa, Haemophilus influenzae, S. pneumoniae and S. aureus may be found in asymptomatic hosts and finding adherent organisms in vivo is commonly referred to as a colonization, rather than an infection. Typically colonization (microbial presence) is differentiated from an infection by evidence of a pathological process in the host such as an overt inflammatory response in the case of infection and its absence in colonization. This issue is particularly challenging, in view of some studies that suggest that biofilms are also present in control specimens (e.g. chronic rhinosinusitis). One problem with clinical specimens is defining an appropriate control group, since it is rarely possible to obtain these from persons without disease. Another problem is the paucity of animal colonization models. Therefore, determining corollaries for the presence of opportunists in normal carriage versus a disease process remains difficult. However, an interesting possibility is that biofilm formation may result from the carriage state, making biofilm development a ‘persistence factor’ as opposed to a virulence factor per se.
Biofilms are also troublesome to diagnose because culture, although generally sufficient in acute disease, is not an accurate predictor of chronic biofilm infection. While diagnostic parsimony may indicate no infectious aetiology with culture-negative results, recurrent signs and symptoms of infection argue otherwise. For example in chronic OM with effusion, detection of bacteria by culture is poorly correlated with infection compared with acute OM (Post et al., 1995; Rayner et al., 1998). Detection of bacteria by culture methods was positive for OM pathogens in children undergoing tympanostomy tube (TT) placement less than 30% of the time, compared with polymerase chain reaction (PCR) and FISH which were positive ∼80–90% of the time (Hall-Stoodley et al., 2006). Similarly there was no correlation between culture methods and the detection of P. aeruginosa by FISH in chronic wounds, in spite of demonstrated biofilm aggregates in the wound (Kirketerp-Møller et al., 2008). Exactly why culture is less reliable in biofilm disease may be due to inadequate sampling, inadequate incubation times or inadequate nutrient composition in the media.
The difficulty of identifying biofilms infections in vivo has led to the outline of specific criteria for diagnosing biofilm infections from clinical specimens by Parsek and Singh and others (Fux et al., 2003; Parsek and Singh, 2003) (see Box 1). Briefly, the ‘Parsek–Singh criteria’ include (1) surface-associated, (2) microbial aggregates, (3) localized infections that are (4) intractable to antibiotic treatment despite the demonstrated susceptibility of planktonic bacteria. Fux et al. further suggested that diagnosis based on culture results alone was insufficient without giving some thought to sampling techniques since localized bacteria in a biofilm may be missed in a conventional blood sample or aspirate. Consistent with this last point, multiple consecutive lavage samples demonstrated that the bacterial burden of biofilm associated non-typeable H. influenzae (NTHi) was much greater than that shown by single middle-ear effusion samplings in the chinchilla model of OM (Leroy et al., 2007). Moreover, Tunney et al. associated biofilms with orthopaedic implant infections by observing bacterial aggregates by immunostaining and 16S reverse transcription polymerase chain reaction (RT-PCR) (Tunney et al., 1999) and diagnosis of orthopaedic infection was markedly improved by direct sampling of the prosthesis surface and sonication (Trampuz et al., 2007).
Box 1. Diagnostic criteria for biofilm infections.
1Pathogenic bacteria are associated* with a surface.
2Direct examination of infected tissue demonstrates aggregated cells in cell clusters encased in a matrix, which may be of bacterial and host origin.
3Infection is confined to a particular site in the host.
4Recalcitrance to antibiotic treatment despite demonstrated susceptibility of planktonic bacteria**.
5Culture-negative result in spite of clinically documented high suspicion of infection (since localized bacteria in a biofilm may be missed in a conventional blood sample or aspirate).
6Ineffective host clearance evidenced by the location of bacterial cell clusters (macrocolonies) in discrete areas in the host tissue associated with host inflammatory cells.
*Note the use of the term ‘associated’ allows the inclusion of aggregates not necessarily firmly attached.
**We propose the following modification of criteria ‘4’, since antibiotic susceptibility of planktonic cultures can only be tested upon positive culture from a clinical specimen. We suggest that in the absence of culture, recalcitrance to antibiotic therapy may be inferred from the presence of live bacterial cells in the biofilm from in situ viability staining or reverse transcription polymerase chain reaction (RT-PCR).
Since antibiotic susceptibility of planktonic cultures can only be tested when clinical specimens result in a positive culture, we suggest a modification of criteria ‘4’. In the absence of culture, resistance to antibiotic therapy might be inferred by demonstrating the presence of live aggregated bacteria using methods such as direct viability staining coupled with molecular methods such as RT-PCR or FISH (Stoodley et al., 2008b). We also propose a sixth criterion for biofilm-associated infections: ineffective host clearance indicated by the location of bacterial cell clusters in discrete areas in the host tissue associated with host inflammatory cells. Evidence of PMNs and macrophages surrounding aggregated bacteria in situ considerably increases the suspicion of infection. However, in the absence of biofilm-specific markers and appropriate animal models for many biofilm infections, fulfilling ‘Koch’s postulates' for biofilm infections remains a challenge.
Biofilms are demonstrated locally associated with tissues such as oral, respiratory or urogenital epithelium
Biofilms have now been associated with several chronic infections in the respiratory tract by examining clinical samples or biopsies. Sputum and bronchiolar lavage samples and lung biopsies (BAL), often in conjunction with in vitro experiments, have been useful in examining the distribution of bacterial aggregates in CF by P. aeruginosa and NTHi in chronic bronchitis in order to examine the distribution of bacterial in situ in the lung during chronic infections (Lam et al., 1980; Singh et al., 2000; Worlitzsch et al., 2002; Starner et al., 2006; Martínez-Solano et al., 2008; Yang et al., 2008). These studies have observed aggregated bacteria, suggesting a biofilm infection. P. aeruginosa was found in bacterial aggregates enclosed in a matrix within O2-depleted mucus from CF patients, although not necessarily adherent to the airway epithelium (Worlitzsch et al., 2002). The role that insufficiency of mucociliary clearance plays as a contributing factor in biofilm infections needs to be further explored since a stagnant mucin-rich effusion that promotes bacterial proliferation has been demonstrated in the chinchilla model of OM (Miyamoto and Bakaletz, 1997) and mucin (MUC) gene polymorphism in patients with chronic OM may contribute to the inadequate host clearance of H. influenzae, S. pneumoniae and M. catarrhalis (Kerschner, 2007; Ubell et al., 2008).
Chronic OM has been identified as a biofilm infection from studies using both animal models and clinical specimens. NTHi formed mucosal biofilms demonstrated by electron and confocal microscopy in the Chinchilla laniger model of OM, whereas no biofilm was found in sham-infected controls (Ehrlich et al., 2002). Moreover, biofilm aggregates of S. pneumoniae, H. influenzae and M. catarrhalis were present on the epithelium of the middle-ear mucosal biopsies in children undergoing TT placement for the treatment of chronic or recurrent OM, but not on middle-ear mucosal biopsies obtained from patients undergoing surgery for cochlear implantation (Hall-Stoodley et al., 2006). Antibiotic resistance was inferred since children had been treated for multiple OM infections and pharmacokinetic penetration studies of antibiotics commonly used in OM indicate that bactericidal concentrations are readily achievable in the middle-ear space and sufficient to kill planktonic pathogenic bacteria (Rosenfeld and Post, 1992).
These studies also suggest a primary role for the epithelium in the defence against biofilm infections. While the ability of pathogens to exploit opportunities that arise in deficient host defences is likely to be multi-factorial, the association of pathogenic bacteria with epithelial tissues raises an interesting question about a fundamental criterion of biofilm and whether aggregated bacteria must be attached to a surface. The evidence with P. aeruginosa argues for aggregates of bacteria within the mucus layer overlaying the epithelium being an important determinant for a biofilm infection particularly in concert with impairment of multiple components of innate immunity (Matsui et al., 2006; Döring and Gulbins, 2009). Since non-attached aggregates of bacteria retain the antibiotic recalcitrance seen in biofilms, cell aggregation may be a key indicator of a biofilm infection (Fux et al., 2004). Nevertheless, the examination of several chronic infections in the upper respiratory tract such as recurrent tonsillitis and chronic rhinosinusitis in human clinical specimens suggests that both attachment and aggregated bacteria are present (Chole and Faddis, 2003; Psaltis et al., 2007; Prince et al., 2008). Chole and Faddis used electron microscopy and culture to show that biofilms were associated with the mucosal epithelium of tonsils in 73% of tonsils removed for tonsillitis and 75% of those tonsils removed due to hypertrophic tonsils alone. These results suggest that biofilms are present associated with mucosal epithelium in disease and in hypertrophic lymphoid tissues.
More recently molecular in situ techniques have demonstrated that biofilms, formed by pathogens (sometimes in polymicrobial communities), also occur on the mucosal surface of human adenoids, associated with epithelial cells (Kania et al., 2008; Nistico et al., 2009). Sections from the same adenoid biopsies that demonstrated bacterial pathogenic biofilms associated with the epithelium by FISH showed increased resistance to azithromycin and the presence of a carbohydrate matrix in situ (L. Nistico, P. Stoodley, and L. Hall-Stoodley, unpubl. obs.). Therefore, using an algorithm of tests in the case of clinical samples which includes specific probes to determine the distribution of pathogenic bacteria and whether they are aggregated will be more effective in diagnosis of a biofilm infection. FISH in conjunction with pathogen-specific 16S probes represents a particularly good method for detecting specific pathogens in situ compared with electron and light microscopy. However pathogen-specific antibody staining is also compatible with transmission electron microscopy or immunofluorescence on clinical samples. The ability of FISH to be used in conjunction with antibodies to specific human cell markers can also give specific information about the location of the biofilm in tissue (Nistico et al., 2009). The real value of these studies is that it is possible to see the microbial host interfacein situ, which may be an important guide for future exploration of specific mechanisms used by pathogens in biofilm infections.
Adherence of pathogens to host cells and intracellular bacteria in biofilm infections
Recently one of the best examples of the evolution of the concept of a biofilm infection has come from the investigation of recurrent UTI. Work by Hultgren and others has demonstrated that uropathogenic E. coli (UPEC) adhere to and invade the bladder epithelium during the early stages of infection in a mouse model of UTI (Anderson et al., 2003; Justice et al., 2006). Prior to this work, UPEC was thought to reside extracellularly on the epithelial surface of the urinary tract and, clinically, the recurrence of UTI was thought to be due to re-infection from the gastrointestinal tract. However the observation that UPEC existed in large aggregates of bacteria in mouse bladder epithelial cells suggested another mechanism for chronic recurrent infection of the bladder epithelium (Anderson et al., 2003). Remarkably, intracellular UPEC undergo a complex developmental process into biofilm aggregates within the epithelial cell cytoplasm (Justice et al., 2006). These aggregates exhibit biofilm ultrastructure (large cell clusters of E. coli surrounded by a matrix) and resistance to antibiotic and host killing. Recently another UTI pathogen, Klebsiella pneumoniae, was shown to develop intracellular biofilms linked to the expression of type 1 pili (Rosen et al., 2008) and intracellular UPEC exhibiting the same phenotype in epithelial cells present in human UTI infections has been described (Rosen et al., 2007).
Interestingly, intracellular S. pneumoniae, H. influenzae, M. catarrhalis and S. aureus in situ is observed in adenoids from children undergoing adenodectomy for the treatment of chronic OM or hypertrophic adenoids using FISH (Forsgren et al., 1994; Nistico et al., 2009). H. influenzae and intracellular S. pneumoniae have also been found in the bronchial biopsies in patients with chronic bronchitis and in middle-ear mucosal biopsies in children with chronic OM (Bandi et al., 2001; Coates et al., 2008). H. influenzae and S. pneumoniae have been shown to invade epithelial cells by numerous molecular mechanisms (Swords et al., 2000; St Geme, 2002; Hammerschmidt et al., 2005; Pracht et al., 2005). S. aureus is found intracellularly in the nasal mucosal of patients with recurrent sinusitis (Clement et al., 2005). Thus biofilm aggregates attached to the epithelial surface, residing in the mucus layer or sequestered in intracellular aggregates may all be mechanisms for persistence in mucosal infections. It is interesting to speculate that pathogens such as S. aureus, S. pneumoniae, H. influenzae reside extracellularly in an aggregated biofilm phenotype that protects bacteria from the host response and antibiotic therapy, but also as intracellular sequestered aggregates in respiratory epithelial cells. Both extracellular and intracellular aggregation would facilitate the persistence of pathogens at epithelial interfaces. The phenotypic heterogeneity exhibited by bacteria in biofilms may result in an invasive phenotype under certain conditions in the host (Garcia-Medina et al., 2005; Li et al., 2005) and could influence the induction of adaptive host immune mechanisms. However, the role of epithelial cells in the protection against infection is only beginning to be investigated in the human host.
Biofilm development is an important component of bacterial survival. However understanding how it contributes to infectious disease in the human host is not easily investigated by a reductionistic approach. This is due to the multifaceted nature of biofilm research which examines biofilm development as a process of intrinsic genomic programmes and adaptive plasticity exhibited by each organism in response to different host environments. At the host level, we are only beginning to understand how biofilm infections caused by well-known pathogens affect the host response. Moreover, we have barely considered the further complexity that polymicrobial infections may present in terms of both the responses to other organisms and in the overall host response. By investigating the distribution of bacterial pathogens as they are found in various chronic infections we can better understand the relevant microenvironments and the limitations of the host responses to various pathogens. We think that assessing putative biofilm infections using clinical specimens and biopsies ex situ will be useful towards understanding the relationship and interplay between biofilm bacteria and the inflammatory and innate immune response relevant to a specific tissue. These studies will help build a more comprehensive and coherent model of biofilm-related pathogenesis and provide data for the refinement of existing treatment strategies for biofilm infections as well as novel treatment strategies.
We would like to thank the reviewers for their comments and thoughtful suggestions. L.H.-S. and P.S. are supported by Allegheny-Singer Research Institute and P.S. is supported by Phillips Oral Healthcare, Snoqualmie, WA. We thank Dr Joseph Kerschner for the clinical specimen of the adenoid used in Fig. 1.