For example, biofilms formed by Pseudomonas aeruginosa can be composed of alginate, Pel, or Psl polysaccharides (Branda et al., 2005; Ryder et al., 2007). Proteins or extracellular DNA

also appear to be important in stabilizing the matrix (Otto, 2008; La et al., 2010; Romero et al., 2010). Such variability can be due to the expression of select matrix genes under certain growth conditions, cell death, signaling pathway or simple fluctuations in the genetic background of strains being studied. The considerable diversity in biofilm EPS composition is one variable that has complicated the use of mathematical modeling to predict how biofilm structural changes arise as a consequence of physical parameters. (2) What is the contribution of phenotypic heterogeneity to biofilm formation? There are several different levels of genetic/phenotypic diversity within a biofilm, such as the variety of colonizing species, gene activation/repression, mutations, and more plastic phenotypic variations. Partially as a consequence of the level of details regarding the cell–cell signaling pathways (quorum sensing), GSK2118436 ic50 the discovery of second messengers such as bis-(3′-5′)-cyclic dimeric guanosine monophosphate, ‘social cheating’, as well as studies of the various mechanisms that protect the bacteria within the biofilm, phenotypic variation has moved to the

forefront of many studies (Parsek & Greenberg, 2005; Sandoz et al., 2007; Jonas et al., 2009; Hoiby et al., 2010). This introduces another difficulty in the theoretical studies. Although very detailed models can be created and analyzed for a single cell, coupling a realistic number of cells together through physical interactions while retaining the detailed microstructure and microenvironment leads to models that are computationally prohibitive (i.e. we do not currently have computational hardware and methods to attempt this). The different time scales for these events also compound the problem

(on the order of minutes for gene expression all the way Farnesyltransferase to the order of days for biofilm growth). This problem is similar to that in molecular biology where one is faced with the choice of molecular dynamics simulations, which are a faithful representation of almost all the forces/interactions, or a coarser model. The former simulations can be done for very short times (on the order of micro-milli seconds) while the latter can be done for much longer time periods (Balaban et al., 2004; Cogan, 2006). Several mathematical studies have focused on incorporating genetic expression via cell–cell communication or quorum sensing (Dockery & Keener, 2001; Anguige et al., 2004, 2005). From a mathematical standpoint, the minimal requirement for a diffusible signal to work is the existence of a mathematical ‘switch’ that turns specific gene pathways on or off. Reductionist models have been successful in predicting both the timing and physical consequences of the switching mechanisms.