This first stage initiates with reversible attachment of bacteria to a favorable surface and is highly dependent on the physicochemical and electrostatic interactions between the bacterial envelope itself and the substrate. Attachment occurs seconds after the bacterial cells detect required environmental signals including changes in nutrients and nutrient concentrations (glucose, indole, polyamines), inorganic molecules (iron, phosphate), pH, antimicrobials, temperature, oxygen concentration, osmolarity, and host derived signals (bile acids, hydrogen peroxide) (O'Toole and Kolter, 1998; Aparna and Yadav, 2008; Karatan and Watnick, 2009). At this point bacterial cells usually exhibit a logarithmic growth rate.

Attachment is facilitated by different adhesive organelles, e.g., flagella and type IV pili play important roles in surface aggregation in Pseudomonas spp. and Vibrio cholerae, whereas fimbriae like type 1 pili, curli, and conjugative pili are important for biofilm formation in Escherichia coli (Thelin and Taylor, 1996; O'Toole and Kolter, 1998; Watnick and Kolter, 1999; Jackson et al., 2002). Curli fimbriae are also produced by other enteric bacteria such as Shigella, Salmonella, Citrobacter, and Enterobacter (Smyth et al., 1996). Following initial attachment, bacterial cells multiply to form aggregated microcolonies and can inter-communicate by producing quorum sensing molecules, which is one of the key events leading to biofilm development in some bacteria (Camilli and Bassler, 2006).

During the maturation stage cell aggregates begin to grow in layers in a three-dimensional matrix (Aparna and Yadav, 2008). The maturation stage still requires adhesive organelles, however, this stage is mostly characterized by cell-to-cell interactions and formation of important surface components that contribute to the structure of the biofilm (McLean et al., 1997; Davies et al., 1998; Holden et al., 1999; Pesci et al., 1999; Whiteley et al., 1999; De Kievit et al., 2001).

A general hallmark feature that determines the mature biofilm architecture is the presence of the extracellular matrix (EM) surrounding the resident biofilm bacteria. Besides its crucial role in maintaining biofilm structure, it enables bacteria to remain in close proximity to each other, protects embedded bacteria from desiccation, acts as a diffusion barrier, and allows bacteria to evade recognition by the host immune system. The biofilm matrix generally consists of up to 97% water, 2–5% microbial cells, 3–6% extra-polymeric substances (EPS) and ions (Sutherland, 2001). The EPS may account for 50–90% of the total organic carbon of biofilm and this is primarily composed of exopolysaccharides, but it also includes proteins (extracellular proteins and enzymes), DNA and RNA, which constitute less than 2% of the biofilm matrix (Flemming and Wingender, 2001; Sutherland, 2001; Donlan, 2002; Flemming et al., 2007). The polysaccharide composition along with other components such as proteins usually varies among different bacteria and even between strains of a single species, although there are some common polysaccharides produced by multiple species of bacteria. It has been proposed that after contact of bacteria with a surface, altered gene expression induces changes that initiate synthesis of extracellular polysaccharides since alginate, the EPS of P. aeruginosa biofilms, is up-regulated in recently attached cells in comparison with planktonic cells (Davies and Geesey, 1995). The systematic three dimensional development of mature V. cholerae biofilms following attachment, and specifically as this is related to synthesis of the EPS, Vibrio polysaccharide (VPS) and the three major EM proteins, RbmA, RbmC, and Bap1, has been captured in real time in elegant work done using advanced microscopy (Berk et al., 2012).

One of the most common and extensively studied matrix exopolysaccharides is the poly–N-acetylglucosamine (PGA or PNAG) that is utilized to construct the biofilm matrices (Wang et al., 2004; Izano et al., 2007, 2008; Parise et al., 2007) and is produced by diverse bacterial species, including E. coli, Staphylococcus epidermidis, Staphylococcus aureus, Yersinia pestis, Actinobacillus spp., Aggregatibacter actinomycetemcomitans, and Bordetella spp. Some bacteria are capable of producing multiple polysaccharides which usually confer different physiological properties to the biofilm matrix, e.g., P. aeruginosa makes biofilms constructed from 3 distinct exopolysaccharides (alginate, Pel, and Psl), Bacillus subtilis secretes 2 polymers (EPS and PGA) while E. coli principally synthesizes PGA, colanic acid, and cellulose. Cellulose is a commonly produced polymer among the enteric pathogens including Salmonella, Citrobacter, Enterobacter, and Shigella, where it has been strongly associated with the ability to form a rigid biofilm (Sutherland, 2001; Zogaj et al., 2001, 2003; Solano et al., 2002; Spiers et al., 2003; Da Re and Ghigo, 2006; Ude et al., 2006; Lopez et al., 2010).

The protein fraction of the EPS is generally quite large and an important component for biofilm formation that include structural proteins like adhesins, and other cell surface associated proteins like pili, flagella, curli and amyloid fibers (Prigent-Combaret et al., 2000; Toledo-Arana et al., 2001; Gohl et al., 2006; Flemming et al., 2007; Larsen et al., 2007), as well as a homologous group of large proteins, referred to as biofilm-associated proteins, found in, e.g., Staphylococcus, Enterococcus, Vibrio, and Salmonella spp. (Cucarella et al., 2001; Kristich et al., 2004; Latasa et al., 2006; Fong and Yildiz, 2007; Berk et al., 2012). The production of 3 biofilm-associated matrix proteins (RbmA, RbmC, and Bap1) and the Vibrio polysaccharide (VPS) are involved in V. cholerae biofilm formation, where the RbmA protein is specifically involved in cell-to-cell adhesion and Bap1 facilitates adherence of the biofilm to surfaces (Berk et al., 2012).

Extracellular DNA (eDNA) may act as a structural component of the biofilm matrix, where it can be found in varying quantities. eDNA is a major structural component in the biofilm matrix of S. aureus (Izano et al., 2008). Its role in biofilms was confirmed in P. aeruginosa, where DNAse was added to the culture medium and this resulted in dissolution of preformed biofilms (Whitchurch et al., 2002; Nemoto et al., 2003; Bockelmann et al., 2006, 2007). In fact, it has been reported that eDNA facilitates the self-organization of bacterial biofilm, as it coordinates the movement of cells and is required for their assemble into the intricate network of furrows that form the biofilm (Gloag et al., 2013).

In this stage these sessile communities (surface-attached) can give rise to planktonic (free-floating) bacteria that can rapidly multiply and disperse to colonize new surfaces. Some studies have shed light on the signals and signaling networks that lead to dispersal of biofilms. Most often the nutritional status of the environment dictates bacterial behavior, including the biofilm dispersal response that can result from both decreases and increases in environmental nutrients (Anderl et al., 2003; Walters et al., 2003; Sauer et al., 2004; Gjermansen et al., 2005; Thormann et al., 2005, 2006). Besides nutrient availability, there are other factors influencing the biofilm dispersion, such as the presence of oxygen, by-products of anaerobic metabolism, quorum sensing signaling, and levels of the chemical second messenger molecule cyclic-di-GMP (c-di-GMP). The exact mechanistic details of dispersal have not been elucidated. However, among the factors that influence biofilm detachment are synthesis of enzymes that degrade extracellular polymeric substances in the biofilm matrix, release of EPS and surface-binding proteins, induction of motility, surfactant production and cell lysis, hydraulic shear, sloughing, and erosion (Boyd and Chakrabarty, 1994; Lee et al., 1996; Allison et al., 1998; Jackson et al., 2002; Stoodley et al., 2002; Kaplan et al., 2003, 2004; Webb et al., 2003; Boles et al., 2005; Itoh et al., 2005; Purevdorj-Gage et al., 2005; Chambless and Stewart, 2007).

The processes described above for biofilm development are not necessarily synchronized throughout the whole biofilm, but are often localized so that at any time a small area on the surface may contain biofilm at each developmental stage. Even when cells in a biofilm are clonal populations, it has been shown that there are subpopulations of phenotypically dissimilar cell types, and that this cell heterogeneity is a consequence of alterations in gene expression in response to extracellular conditions encountered by the cells within a microenvironment (Lopez et al., 2010). Therefore, since the synthesis and assembly of biofilm matrix components is a costly and an exceptionally complex process, the transition from the planktonic state to the sessile state requires the coordinate expression and simultaneous regulation of many genes by complex genetic networks involving all levels of gene regulation.

Central to the mechanism of post-transcriptional regulation of biofilm formation is a tiny cyclic RNA chemical second messenger molecule, c-di-GMP. C-di-GMP is implicated in controlling various cellular functions including virulence, motility, and adhesion, although its principal role is controlling the switch from motile planktonic lifestyle to the sessile biofilm forming state. Elevated intracellular levels of c-di-GMP promote synthesis of exopolysaccharides and enhanced auto-aggregation and surface adhesion leading to biofilm formation (Simm et al., 2004). In contrast, reduced intracellular c-di-GMP concentrations are associated with decreased biofilm formation. The mechanistic details of how post-transcriptional gene regulation is mediated by the c-di-GMP molecule will be discussed below in section Riboswitches.

C-di-GMP is synthesized by GGDEF motif containing proteins that encode diguanylate cyclase (DGC) enzyme activity required to convert two molecules of GTP to c-di-GMP. Degradation of c-di-GMP is carried out by EAL or HD-GYP motif containing-proteins that encode phosphodiesterase activity leading to hydrolysis of c-di-GMP to pGpG (Ross et al., 1987; Tal et al., 1998; Romling et al., 2005). The GGDEF or EAL/HD-GYP protein encoding genes controlling c-di-GMP synthesis are present in multiple copies in bacterial genomes especially in those pathogens that infect multiple hosts, e.g., E. coli encodes 34, Salmonella enterica serovar Typhimurium encodes 27, V. cholerae encodes 63 (Galperin et al., 2001; Galperin, 2004) and P. aeruginosa encodes 39 such genes with GGDEF, EAL, or HD-GYP domains in their genomes (Kulasakara et al., 2006). Thus, it stands to reason that the c-di-GMP signal transduction is tightly synchronized and regulated to avoid interference between the functionally distinct c-di-GMP responsive systems.

Such distinct environmental niche-dependent regulation by c-di-GMP is exemplified in control of biofilm formation central to transmission of Y. pestis biofilm from the foregut of its flea vector. Y. pestis, the agent that causes bubonic plague, contains 10 genes encoding DGCs and PDEs in its genome (Sun et al., 2011). However, only two of the DGC enzymes, encoded by hmsT and y3730, have been shown to be functional in controlling synthesis of c-di-GMP to activate production of biofilm while only one phosphodiesterase (PDE), encoded by hmsP, is active and degrades c-di-GMP, reducing biofilm synthesis (Kirillina et al., 2004; Bobrov et al., 2005; Sun et al., 2011). Interestingly, despite their similar expression patterns during varied in vitro growth conditions and their equally robust expression in the flea gut, the production of biofilm in vitro appears to be mainly dependent on expression of hmsT while biofilm formation within the flea gut is mainly dependent on y3730 expression. The Y3730 DGC alone appears to promote autoaggregation indicating a potential role for this DGC in additional biofilm promoting functions (Sun et al., 2011). Additionally, Y3730 responds to a specific flea environmental cue detected by the gene product of the co-transcribed gene hmsC (Ren et al., 2013). The different predominant roles for each of these DGCs are driven by post-transcriptional control mechanisms still yet to be fully defined.

Another well-characterized example of how distinct phenotypes can be regulated by specific DGCs is found in P. aeruginosa. Mutational studies of the numerous DGCs encoded in the genomes of P. aeruginosa strains, PAO10 and PA14, indicate that while some DGCs impact biofilm formation, others either appear not to affect biofilm formation, or may require specific in vivo activating signals to initiate their regulation of biofilm formation; and yet other DGCs are essential for functions such as cytotoxicity and virulence (Kulasakara et al., 2006). Nevertheless, there is a correlation between increased amounts of c-di-GMP and biofilm formation in P. aeruginosa. Interestingly, unique localized pools of c-di-GMP may be found in the cell due to differential subcellular localization of DGCs and this may be intimately connected with specific signaling to cognate gene targets to produce biofilm (Merritt et al., 2010; Massie et al., 2012). This observation was made in P. aeruginosa when mutations in 2 DGCs, SadC and RoeA, showed insignificant changes in total intracellular c-di-GMP quantities despite their negative effects on biofilm formation. RoeA and SadC appeared to be differentially localized in the cell thus producing unique localized pools of c-di-GMP that likely specifically activate different gene targets for EPS production or increased swarming motility, to negatively impact biofilm formation (Merritt et al., 2010).

This group contains those proteins known to bind to mRNA, and by doing so regulate translation initiation, mRNA stability, and the half-life of the message. Below we will outline the functions of the most thoroughly studied RNA binding proteins CsrA and Hfq which are able to post-transcriptionally regulate biofilm formation in various bacterial pathogens.

The ncRNAs are a large group of small molecules of RNA (50–500 nts in length) that are not translated into proteins. They act as important trans-acting regulators of gene expression in bacteria, directly influencing protein synthesis at the post-transcriptional level (Gottesman and Storz, 2011) especially under stress conditions. These molecules carry out their regulatory function by base-pairing to a limited complementary sequence on the mRNAs of their cognate target gene, which leads to changes in mRNA translation or stability or both, thereby influencing the target gene expression. The ncRNAs can act by activating or repressing gene expression depending on the portion of the mRNA molecule they base-pair with (Waters and Storz, 2009). Employing ncRNAs to fine-tune the expression of genes, stabilize mRNA, and regulate biofilm formation is an energetically economical strategy (Waters and Storz, 2009). Synthesis of ncRNA is part of the on-going transcription process and likely does not require any additional enzymatic processing as far as is known, therefore defeating the requirement for additional proteins to be synthesized. This should be especially advantageous under limiting conditions of stress that normally trigger biofilm production.

CsrB and CsrC, are two well-studied examples of ncRNAs in E. coli that contain many CsrA-binding sites and function by antagonizing CsrA activity by binding and sequestering this protein, counteracting its translational repression activity (Liu et al., 1997; Romeo, 1998; Weilbacher et al., 2003; Babitzke and Romeo, 2007; Heroven et al., 2008; Lucchetti-Miganeh et al., 2008; Timmermans and Van Melderen, 2010). The study of CsrB and CsrC as regulatory factors for gene expression in bacteria has greatly increased in the recent years. Orthologs of CsrB/CsrC (and CsrA), have been identified in many other gammaproteobacteria (Lapouge et al., 2008), including Salmonella enterica (CsrB/CsrC, and CsrA) (Martinez et al., 2011), Pseudomonas species (RsmX/RsmY/RsmZ, and RsmA) (Kay et al., 2005), Vibrio species (CsrB/CsrC/CsrD, and CsrA) (Lenz et al., 2005), Erwinia carotovora (RsmB, and RsmA) (Cui et al., 2008), and Yersinia species (CsrB/CsrC and CsrA) (Heroven et al., 2012), where they also control the expression of many genes required for a wide variety of cellular functions, including biofilm (Lucchetti-Miganeh et al., 2008; Timmermans and Van Melderen, 2010). The role of these ncRNAs in biofilm production is at a post-transcriptional level by controlling the levels of free CsrA/RsmA protein resulting in activation of biofilm formation (Figure 1). It has been demonstrated that CsrB and CsrC have a redundant function, which means that CsrB levels exhibit a compensatory increase in response to CsrC disruption and vice versa (Weilbacher et al., 2003). In agreement with this, it has been shown in E. coli and Salmonella that only mutants lacking both csrB and csrC, but not the single mutants in csrB or csrC, show reduction in biofilm production (Wang et al., 2005; Teplitski et al., 2006). However, other unidentified ncRNAs may act similarly to CsrB/CsrC to counteract the effects of CsrA. This has been demonstrated by the ability of CsrA from C. jejuni to restore the affected CsrA phenotypes in E. coli, even though homologs to CsrB and CsrC or other proteins involved in the Csr pathway in E. coli have not been identified in C. jejuni (Parkhill et al., 2000; Hofreuter et al., 2006; Kulkarni et al., 2006; Fields and Thompson, 2012). It has been proposed that these two ncRNAs are in turn post-transcriptionally controlled by CsrD, a GGDEF/EAL protein that binds CsrB and CsrC leading to their decay by converting them to substrates for RNAse E degradation (Suzuki et al., 2006).

Although ncRNAs are widespread in nature and an increasing number of ncRNAs have been found to regulate critical pathways, the specific functions of many ncRNAs are still unknown. However, there is growing evidence that could support their role in biofilm production. These molecules often require the chaperone protein Hfq for their expression, stability, and/or function. The effect of Hfq on biofilm production varies among different bacteria and their conditions of growth, suggesting that Hfq-dependent ncRNAs specific to different environments could be the determining factors required to produce biofilm. Accordingly, studies have attempted to identify and characterize the arsenal of ncRNAs in different pathogens and under different conditions by using deep sequencing approaches and have found distinct predominating ncRNAs under different conditions. The ncRNAs that are produced in Y. pestis, for example, are distinct depending on growth environment alterations, e.g., temperature, medium and in vivo vs. in vitro (Koo et al., 2011; Beauregard et al., 2013; Yan et al., 2013).

It has been reported that iron concentrations can regulate biofilm formation via ncRNAs in the periodontal pathogen A. actinomycetemcomitans. Some ncRNAs whose expression is dependent on the ferric uptake regulator Fur, which functions to repress genes involved in iron uptake in iron-rich environments, were identified in this pathogen. It is proposed that these ncRNAs could be interacting with biofilm-associated genes, including the flp fimbrial operon and genes associated with EPS, according to in silico models (Amarasinghe et al., 2012). So far the ability of these iron-regulated ncRNAs to modulate biofilm formation has been demonstrated, but their functional target genes remain to be elucidated (Amarasinghe et al., 2012).

The ncRNAs are sometimes part of complex regulatory networks that finally induce biofilm formation by post-transcriptional control. In V. cholerae there are 4 redundant ncRNAs, called Qrr1-4 that are under control of the quorum sensing response (Lenz et al., 2004). They repress translation by binding and occluding the RBS of several mRNAs, including hapR, which encodes HapR, a transcription factor that controls the expression of genes involved in biofilm formation (Bardill et al., 2011). Each Qrr is predicted to base pair to the 5'-UTR of hapR aided by Hfq (Lenz et al., 2004). HapR indirectly represses the expression of the exopolysaccharide biosynthesis operon and alters the intracellular levels of c-di-GMP (Hammer and Bassler, 2003, 2009). Furthermore, Qrr ncRNAs may also facilitate degradation of hapR mRNA by cellular RNases as increased levels of the ncRNAs correlated with decreased levels of hapR transcript (Svenningsen et al., 2008). On the other hand, the ncRNAs Qrr1 and Qrr2 have been shown to base-pair to the 5'-UTR of the vca0939 gene to positively control its translation to a GGDEF motif containing protein and synthesis of c-di-GMP (Hammer and Bassler, 2007). It had been assumed that the four Qrr ncRNAs act in a similar way to induce the synthesis of Vca0939 as the Qrr/vca0939 pairing occurs through a region that is 100% conserved among them, which interestingly is the same region required to base-pair to the 5'-UTR of hapR (Zhao et al., 2013). All these studies demonstrate how biofilm formation is linked to the detection of cell density by multiple pathways, where at high density biofilm formation would be repressed because none of the Qrr ncRNAs would be present to repress HapR (Bardill et al., 2011).

We are beginning to understand that post-transcriptional control involving ncRNAs is highly complex and there is growing evidence for a role of these molecules as crucial regulators, alongside well-known global transcriptional regulators. Perhaps the current best characterized example of just how extremely complex and convoluted a post-transcriptional regulatory network can be is that of the regulatory network of CsgD, the global transcription factor that integrates signals to control biofilm formation in E. coli and Salmonella. CsgD induces the expression of the csgBAC operon, required for the production of curli fimbriae, and cellulose, two major adhesive factors required for biofilm, as well as the production of c-di-GMP. Furthermore, CsgD represses flagellar operons favoring the biofilm-formation phenotype (Ogasawara et al., 2011). CsgD can be post-transcriptionally repressed by 5 discrete Hfq–dependent small RNAs (OmrA, OmrB, GcvB, McaS, RprA) that respond to various stress conditions (Boehm and Vogel, 2012). These small RNAs directly target the 5'-UTR of the csgD mRNA hence blocking translational initiation. The ncRNAs OmrA and OmrB, are induced by the two component regulator, OmpR-EnvZ during high osmolarity conditions (Holmqvist et al., 2010). The multicellular adhesive ncRNA McaS, present only in E. coli, Enterobacter and closely related species, is an important small RNA element that disfavors biofilm by binding and exerting inverse control on csgD and flhCD mRNAs (Thomason et al., 2012). In the absence of McaS, csgD upregulation occurs; alternately, increased McaS expression activates the synthesis of flhDC, encoding the flagellar master regulator (Thomason et al., 2012; Jorgensen et al., 2013). In contrast to its occlusion of the SD region in the 5'-UTR of the csgD mRNA, McaS is able to unfold the stem-loop structure of the flhDC mRNA thus exposing the hidden SD sequence for translation initiation (Thomason et al., 2012). However, the McaS-dependent regulation of biofilm seems to be complex, as it also activates PGA, leading to the formation of biofilm that is independent of the CsgD-pathway. McaS-mediated activation of pga requires the presence of CsrA as McaS can be immunoprecipitated with CsrA in an E. coli lysate and the McaS transcript has several GGA motifs which are recognized targets for binding to CsrA (Holmqvist et al., 2010; Holmqvist and Vogel, 2013; Jorgensen et al., 2013). McaS is therefore the first example of a ncRNA that regulates gene expression through both CsrA and Hfq RNA binding proteins, acting by two different mechanisms: base-pairing and protein titration (Jorgensen et al., 2013).

RprA is the small RNA that responds to envelope stress mediated through the phosphorelay system Rcs proteins. RprA appears to fine tune expression of csgD by targeting several branches of the CsgD network in response to environmental cues (Mika et al., 2012). RpoS, the general stress response sigma factor is activated by RprA and in turn activates csgD expression and curli biosynthesis once cells transition into stationary phase, while RprA inhibits csgD expression through the DGC YdaM. RprA appears to modulate synthesis of colonic acid in the presence of curli and also cellulose biosynthesis, possibly to balance the expression of these EPS matrix components as necessitated by a specific environmental cue (Mika et al., 2012). RprA is another example of a ncRNA that plays a dual role to repress biofilm formation. The fifth known small RNA, GcvB is expressed in response to amino acid availability to repress csgD under these specific nutritional conditions (Boehm and Vogel, 2012; Jorgensen et al., 2012).

In Salmonella typhimurium, LuxS, the synthase enzyme required in quorum sensing response, is involved in biofilm production, as it has been shown that a luxS deletion mutant is impaired in biofilm production. This defect can be restored upon complementation only with luxS along with its native promoter region. This lead to the finding that adjacent to the luxS coding sequence is a ncRNA called MicA whose balanced concentration is essential for proper biofilm formation (Kint et al., 2010). The mechanism by which MicA exerts its function is still unknown (Kint et al., 2010).

Riboswitches are structured non-coding RNA domains that form part of the mRNA usually at the 5'-UTR where they can act in cis to control gene expression upon selectively binding ligands (Mandal and Breaker, 2004; Coppins et al., 2007; Roth and Breaker, 2009). Two main domains constitute the riboswitch: an aptamer molecule, which senses and binds a single ligand and an expression platform, usually located downstream of the aptamer that switches its secondary structure according to its ligand binding status to direct gene expression, by either transcriptional or translational mechanisms (Mandal and Breaker, 2004; Weinberg et al., 2007; Roth and Breaker, 2009). The mechanisms of riboswitches sensing and binding a metabolite are categorized into families and classes that are dependent on the required ligand, together with the secondary structure that is formed upon ligand binding.

The mechanism of post-transcriptional gene regulation by the c-di-GMP molecule is facilitated by recognition of a riboswitch. Here c-di-GMP acts as a ligand and mediates its regulation by binding to a riboswitch class in mRNA called GEMM (genes for the environment, for membranes, and for motility). This particular conserved RNA domain GEMM, is located upstream of genes encoding DGC and PDE proteins, as well as in other genes that are controlled by c-di-GMP (Sudarsan et al., 2008; Smith et al., 2010). The GEMM motif was first described in V. cholerae which carries two sequences of GEMM. One of the well-studied c-di-GMP GEMM riboswitches, Vc1, appears to regulate gbpA which encodes a chitin-binding protein required for adherence to chitin and epithelial cells and required for mammalian infection (Kirn et al., 2005; Sudarsan et al., 2008). The second Vc2 is located upstream of the tfoX-like gene vc1722 (Sudarsan et al., 2008), that has been shown to be up-regulated in V. cholerae mutants with the rugose phenotype, characterized in part by increased biofilm production (Lim et al., 2006; Beyhan et al., 2007).

Bacterial biofilms contain an increased prevalence of dormant cells known as persisters, which are characterized by up-regulation of genes known as toxin-antitoxin (TA) modules. These TA systems typically consist of two genes organized in an operon that encodes a stable toxin that disrupts an essential cellular process and a labile antitoxin (either RNA or protein) that prevents toxicity by binding to the toxin and forming a tight complex that neutralizes the toxin (Van Melderen and Saavedra De Bast, 2009). Depending on the nature of the interacting molecules, TA systems can be divided in 3 groups. In the Type I group, the antitoxin is a transcript that is antisense to the toxin mRNA and pairing of the two RNAs promotes mutual degradation. In the Type II group, both toxin and antitoxin are proteins and together they form a complex that masks the activity of the toxin. In the Type III group the toxin is a protein which is inhibited by an antitoxin RNA (Blower et al., 2011). Although the mechanism of toxicity at the molecular level is slightly different, the Type II toxins, MqsR, MazF, RelE, ChpB, YoeB, and YafQ prevent translation by cleaving RNAs (Zhang et al., 2003; Christensen et al., 2004; Gerdes et al., 2005; Brown et al., 2009; Prysak et al., 2009).

The role of TA systems are becoming less enigmatic and they have been implicated in several functions among which include, gene regulation, control of growth, persister cell formation, programmed cell arrest, programmed cell death, and anti-phage measures (Gerdes et al., 2005; Magnuson, 2007; Van Melderen and Saavedra De Bast, 2009). The role of TA systems in biofilm formation has been demonstrated as well as it is currently recognized that these systems can direct cells toward the formation of biofilm and persister cells (Kim and Wood, 2010). In mutant strains of E. coli lacking TA pairs MazF/MazE, RelE/RelB, YoeB/YefM, YafQ/DinJ and ChpB, biofilm formation has been demonstrated to be affected (Kim et al., 2009, 2010; Kolodkin-Gal et al., 2009). TA systems in cryptic prophages have been found to influence biofilm formation, e.g., deletion of toxin YpjF from the TA pair YpjF-YfjZ of the cryptic prophage CP4-57 of E. coli K-12 increased biofilm formation (Brown and Shaw, 2003; Wang et al., 2009, 2010).

The MqsR/MqsA pair of E. coli, was the first TA system found to be associated with biofilm formation when mqsR was identified among genes that were differentially regulated in biofilm cells (Ren et al., 2004). This system is conserved in 40 eubacteria and in many genera such as Y. pseudotuberculosis, Y. pestis, Bordetella bronchiseptica, and P. fluorescens (Kim et al., 2010). The toxin MqsR is an RNase belonging to the RelE family that cleaves mRNA at GCU and, to a lesser extent, GCA sequences. The antitoxin MqsA, which is located immediately downstream of mqsR in the same operon, is a DNA/binding protein that functions to neutralize MqsR toxicity (Brown et al., 2009). It has been shown that a deletion of mqsRA reduces biofilm formation in E. coli and this effect is due to MqsR favoring biofilm formation by affecting cellular motility through autoinducer-2 signaling (Gonzalez Barrios et al., 2006; Kasari et al., 2010). Furthermore, it also plays a role in E. coli biofilm and persistence by positively regulating the expression of toxin CspD and QseBC, the two component motility regulatory system (Gonzalez Barrios et al., 2006; Kim et al., 2010). On another level, MqsA is able to bind to its regulatory palindromic sequence found upstream of the CsgD promoter region, preventing CsgD transcription which in turn decreases curli biosynthesis and E. coli biofilm formation (Soo and Wood, 2013). Interestingly, this system is subject to an auto-post-transcriptional control, as free MqsR has the ability to degrade its own mRNA, and that of mqsA, thus alleviating the persister state and biofilm formation once the environmental stress is removed (Brown et al., 2013).

MqsRA can favor biofilm formation by indirectly controlling the level of c-di-GMP. This effect is the result of a mix of transcriptional and post-transcriptional mechanisms which start with the detection of stress conditions. In such conditions, the protease Lon degrades MqsA which acts as a negative regulator of rpoS by binding directly to its promoter (Wang et al., 2011). Once MqsA is degraded, the sigma factor encoded by rpoS, which controls up to 500 genes in E. coli (Hengge, 2008), is induced. This systematically leads to increases in expression of genes involved in c-di-GMP synthesis and repression of flhDC genes, leading to inhibition of motility, and induction of csgD, favoring both curli and cellulose production (Pesavento et al., 2008). Together all of these processes enhance the formation of biofilm (Wang et al., 2011).

Therefore, given that toxins are mRNA interferases that degrade mRNA with substrate specificity, they can be viewed as acting similarly to the global regulators like Hfq and CsrA or RNases (described below) that regulate gene expression at a post-transcriptional level by modulating mRNA decay.

RNases are enzymes that cleave RNAs, resulting in remarkably diverse biological consequences. Their elucidated role in biofilm production is mainly that of causing decay of ncRNAs that are involved in regulating this process. The primary endonuclease involved in mRNA decay is RNase E. The RNA-binding protein CsrA activity is indirectly regulated by RNase E. The expression of the two ncRNAs, CsrB and CsrC, that regulate CsrA activity in E. coli are regulated by the CsrD protein, which controls their degradation. This CsrD-mediated RNA decay requires RNase E, as demonstrated by the increased half-life of both CsrB and CsrC in rnaseE mutants. Despite CsrD containing a GGDEF/EAL motif, its function does not involve c-di-GMP synthesis or turnover (Suzuki et al., 2006). These results indicate that RNase E plays an important role in post-transcriptionally mediating biofilm formation, by degrading ncRNAs CsrB and CsrC (Suzuki et al., 2006).

Another RNase that acts similarly to degrade ncRNAs is RNase G which is implicated in mRNA maturation and turnover processing in E. coli. In P. aeruginosa, RNase G is involved in biofilm production by controlling the levels of the ncRNA RsmZ. In these bacteria, transition to later stages of biofilm formation is regulated by three two component regulatory systems, BfiSR, BfmSRR, and MifSR (Petrova and Sauer, 2009). BfiSR is required for transition to the irreversible attachment stage and it has been shown that it regulates biofilm development via the degradation of the ncRNA RsmZ through RNase G (CafA) (Petrova and Sauer, 2009, 2010). BfiR binds directly upstream cafA, encoding the RNase G, to activate its expression and once expressed, RNase G targets RsmZ under biofilm growth conditions. In P. aeruginosa, reduced RsmZ levels are essential to form biofilm. Thus, when cafA is inactivated, an increase in RsmZ levels occurs and a subsequent reduction in biofilm formation (Petrova and Sauer, 2010).

Rnase Y was found to regulate cleavage of the mRNAs of genes involved in biofilm formation in B. subtilis. Here, increasing quantities of the mRNA of the biofilm repressor gene sinR were observed with depletion of Rnase Y, resulting in less biofilm formation. Reciprocally, there appeared to be a correlation between increasing amounts of mRNAs of genes required for biofilm and concomitant increased biofilm production, with over-expression of the rny gene, encoding RnaseY (Lehnik-Habrink et al., 2011).

Although it is well known that 3'-UTRs of mRNA molecules in eukaryotes impact mRNA stability and translational efficiency, such experimental knowledge has been lacking for bacteria. However, Ruiz de los Mozos et al recently found that one third of the mapped mRNAs of S. aureus contains long 3'-UTRs (>100 nucleotides) (Ruiz de Los Mozos et al., 2013). Their specific investigation of the regulatory role of the long 3'-UTR of the icaR mRNA which codes for the repressor of the main exopolysaccharide compound of the S. aureus biofilm matrix, indicated that this 3'-UTR can base-pair with the SD sequence of the icaR mRNA and interfere with the translation initiation complex. This results in formation of a double stranded DNA substrate for RNaseIII and subsequent inhibition of biofilm development in S. aureus. This is a first report of such a mechanism involved in biofilm formation and further investigations into the presence of long 3'-UTRs in mRNAs of genes required for biofilm in other biofilm producing bacteria could reveal that this is a conserved mechanism of post-transcriptional regulation.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.