Role of Multicellular Aggregates in Biofilm Formation

ABSTRACT In traditional models of in vitro biofilm development, individual bacterial cells seed a surface, multiply, and mature into multicellular, three-dimensional structures. Much research has been devoted to elucidating the mechanisms governing the initial attachment of single cells to surfaces. However, in natural environments and during infection, bacterial cells tend to clump as multicellular aggregates, and biofilms can also slough off aggregates as a part of the dispersal process. This makes it likely that biofilms are often seeded by aggregates and single cells, yet how these aggregates impact biofilm initiation and development is not known. Here we use a combination of experimental and computational approaches to determine the relative fitness of single cells and preformed aggregates during early development of Pseudomonas aeruginosa biofilms. We find that the relative fitness of aggregates depends markedly on the density of surrounding single cells, i.e., the level of competition for growth resources. When competition between aggregates and single cells is low, an aggregate has a growth disadvantage because the aggregate interior has poor access to growth resources. However, if competition is high, aggregates exhibit higher fitness, because extending vertically above the surface gives cells at the top of aggregates better access to growth resources. Other advantages of seeding by aggregates, such as earlier switching to a biofilm-like phenotype and enhanced resilience toward antibiotics and immune response, may add to this ecological benefit. Our findings suggest that current models of biofilm formation should be reconsidered to incorporate the role of aggregates in biofilm initiation.

changing glucose concentration by a factor of 10,000 should change the relative fitness of aggregates with 11 respect to single cells. Instead, we found that there was no difference in growth rates between aggregates 12 and single cells regardless of the glucose concentration (P = 0.4525; Figure S7). These results indicate that 13 glucose is not a limiting resource in our experiments. 14

Generation of Circular Aggregates 15
To generate bacterial aggregates to seed the surface in the simulations, circular areas containing 16 approximately 100 cells were cut out from previously grown (simulated) biofilms (see Figure S1B). The 17 centre of the circular area was chosen in the middle of the biofilm to ensure no overlaps between 18 neighbouring bacteria, and to ensure that there was a uniform number density of bacteria in the aggregate. 19 The aggregate was generated by computing all bacteria that lay within a radius R = 20 μm of the centre O 20 ( Figure S1B). This procedure gave rise to approximately 100 cells in the aggregate. The resulting aggregate 21 was then placed on the surface as shown in Figure (S1A). Surrounding cells were then inserted at random 22 positions excluding the region occupied by the aggregate. The initial number of these competitor cells 23 either side of the aggregate were determined by the specified surface density (cell μm -1 ).

Supplementary Material on Simulation
Algorithm and Rate Equations

Simulation implementation
To model biofilm growth, starting from configurations such as those shown in Figure S1, we use the agent-based microbial simulation package iDynoMiCs [1]. In iDynoMiCs, individual bacteria, represented as spheres (or discs in 2D), grow at a rate that is dependent on the local nutrient concentration, which is in turn altered by the resulting growth dynamics of the bacteria. Consumption of the nutrients by bacteria, leads to growth and proliferation of the microbial community, giving rise to local stresses that are subsequently alleviated via inter-cellular "shoving". During each global time-step of the simulation, the order in which cells are selected to grow and divide is stochastic. The location of each daughter cell within the area surrounding the mother cell upon division is also computed stochastically.
In the simulation, nutrient is assumed to diffuse towards the biofilm from a bulk region far from the top of the growing biofilm, with the concentration of nutrient in this bulk region set to a constant value. In the biofilm region, the diffusion of nutrients is hindered relative to regions outside the biofilm. Periodic boundary conditions are imposed on both the nutrient concentration field and the particle coordinates in the horizontal direction, while the condition of zero-flux is imposed at the surface.
Mathematically, the dynamics of the nutrient, which is represented as a concentration field, are governed by the reaction-diffusion equation where O(x) is the space (x)-dependent oxygen concentration, D O (x) is the diffusion coefficient of oxygen, and r O (x) is the consumption rate of the oxygen by the bacteria. The rate of oxygen consumption, r S (x) is related to the growth rate of the bacteria, dX/dt, via X(x) is the local biomass density, and Y X/O is the yield coefficient that describes the amount of substrate required to produce one unit of biomass, X.
The growth rate of each bacterium is governed by the equation This is the well-known Monod function for bacterial growth. Here, µ max is the maximum specific growth rate of the bacteria, and K O is the concentration of oxygen, O, at which the growth rate is half maximal. In our simulations, kinetic growth constants from empirical and simulation studies on Pseudomonas aeruginosa were used as input growth parameters. Note that the growth rate parameters Y X/O , µ max , and K O are the same for both the aggregate cells and the surrounding competitor cells on the surface (see Table 1 of main manuscript). It is practical, due to the different timescales for oxygen diffusion and cell growth, to assume a pseudo-steady state oxygen concentration with respect to biomass growth. Therefore one can remove the time dependency in Equation S1 to give where s G and s O are the glucose and oxygen concentrations respectively, and k G and k O are constants. The constant k O in Equation S5 is not the same as the K O (the concentration of oxygen, O, at which the growth rate is half maximal) in Equation S3. In the main manuscript we set K O = k O ln 2 = 8.12 × 10 −4 g L −1 . To derive the expression K O = k O ln 2, we start by assuming that s G >> k G in the Tessier equation above, i.e., glucose is in abundance. Therefore the factor (1 − e −s G /k G ) → 1, giving The growth rate is half maximal at an oxygen concentration of K O