The porous structure induced heterogeneous and localized failure of the biofilm in microfluidic channels

In nature, biofilm is a community of bacteria that can be either superficially attached or free-floating. The bacteria are held together by extracellular polymeric substances secreted by the bacteria (Nadell et al. 2008; Flemming et al. 2016). Bacteria can survive almost anywhere (Conrad & Poling-Skutvik 2018), and soil is the most complex component of the earth's ecosystem and is suitable for microbial growth. Therefore, it is necessary to understand the formation process and characteristics of biofilm communities in soil.

The porous structure of soil affects the spatial distribution and colony behavior of bacteria (Raynaud & Nunan 2014; Coyte et al. 2016). The formation of biofilms in porous media has been studied by many researchers. Scheidweiler et al. (2019) found that biofilms differentiate into an annular base biofilm coating the microcolumns and into streamers protruding from the microcolumns into the pore space. The shape and distribution of streamers are influenced by multiple factors, such as flow velocity (Valiei et al. 2012), channel size, and flow direction (Marty et al. 2012). In addition, Hassanpourfard et al. (2015) show that bacterial flocs (i.e. bacterial aggregates encapsulated in extracellular polymeric substances) can lead to the formation of streamers through large deformation processes. The secondary flow formed by tortuous channels promotes the development of streamers (Marty et al. 2014). Under certain flow conditions, the biofilm attached to the porous medium forms a network of streamers that capture the separated cells, resulting in uneven distribution and even clogging of the biofilm (Drescher et al. 2013). This heterogeneity causes biofilm separation by increasing local fluid shear forces (Karimifard et al. 2021). Hassanpourfard et al. (2016) observed that the clogged bacterial biomass exhibited an instability phenomenon marked by localized streamer breakage and failure leading to the formation of extended water channels, but the origin of this mode of failure is not yet fully understood.

The finite element method is widely used to provide better understanding and explanation of biofilm growth in fluid environments. Traditionally, biofilms have been studied as impermeable domains. In such models, no water can enter biofilms, and contaminants can only enter via molecular diffusion (Bottero et al. 2013; Peszynska et al. 2016). Benioug et al. (2017) studied the biofilm flow in porous media using the Lattice–Boltzmann method and simulated biofilm growth. However, other researchers believe that this assumption is incorrect, and experimental work has shown that biofilms have uneven morphology, including voids and channels, and there are both flowing and stagnant water in biofilms (Lewandowski 2000), so biofilms are permeable. Some researchers have considered the effect of biofilm permeability on the flow of porous media. Deng et al. (2013) studied the effect of permeable biofilm on the flow of porous media and developed a model to predict the overall permeability based on the biofilm permeability and biofilm volume ratio (i.e. the ratio of biofilm volume to preinoculation pore-space volume). The model proves that biofilm permeability affects the shear stress distribution. Even though much work has been done on biofilms in porous media, most of them only focus on the structure of biofilms. Microfluidic platforms have not been used to study how different pore sizes affect the spatial distribution and detachment of biofilms.

In this paper, we used microfluidic platforms for research. Microfluidic platforms enabled us to have better control over the flow of the bacterial solution. We used the microscope to observe the solution for a long time. These devices were disposable and prevented cross-contamination during the experiment. Most importantly, researchers can easily change the section shape and various sizes of the channel to meet the experimental requirements (Rusconi et al. 2011). More and more researchers used this method to explore the dynamic changes of biofilm. In this work, we designed and manufactured microfluidic chip embedded with a micropost array, and the pores between the microposts simulate the interval among soil particles. Our devices made us observe the movement of bacteria in a porous structure that mimics the soil topography. Bacillus subtilis was cultured in these microfluidic channels at different volume flow rates. We described the evolution of the initial spatial distribution of biofilms in porous media and related it to hydrodynamic factors. We found the bacterial transport velocity gradients across pore space have influence on the initial biofilm formation. However, as the biofilm grew, the channels were gradually blocked, and the biofilm was locally unstable and detached. Considering permeability and porosity, we conducted computational fluid dynamics simulations by coupling the Navier–Stokes equations with Brinkman equation and proposed a mode of biofilm detachment.

Bacillus subtilis can be isolated from various environments (soil, marine habitat, etc.) (Yan et al. 2016; Yahya et al. 2021) and has become a model bacterium for studying biofilm formation (Lemon et al. 2008). So, we used Bacillus subtilis 3610 in our experiments, as it is one of the most studied (Gram-positive bacteria) and most widely used to study biofilm growth (Dervaux et al. 2014; Douarche et al. 2015; Gingichashvili et al. 2022). We cultured biofilms by feeding the minimal-salts-glutamate-glycerol (MSgg) solution at OD₆₀₀ = 0.4 with a flow rate of Q = 10 and 25 μL h⁻¹ in microfluidic channels. We cultured at a flow rate of Q = 10 μL h⁻¹ for 25 h; on this basis, culture was carried out at a flow rate of Q = 25 μL h⁻¹. MSgg is composed of 5 mM potassium phosphate (pH 7), 100 mM 4-morpholinepropanesulfonic acid (MOPS) (pH 7), 2 mM MgCl₂, 700 μm CaCl₂, 50 μm MnCl₂, 50 μm FeCl₂, 1 μm ZnCl₂, 2 μm thiamine, 0.5% glycerol, 0.5% glutamate, 50 μg/ml tryptophan, 50 μg/ml phenylalanine, and 50 μg/ml threonine. The MSgg used in the experiment had been sterilized. All experiments were repeated five times.

We used the COMSOL software to carry out related simulation in fluid flow and used CAD software to draw the two-dimensional shape of porous media. Fluid space was set to have the material properties of water, with the inlet having a velocity boundary condition of fully developed flow and the outlet having a pressure boundary condition of zero Pa. The mesh used was a free triangular mesh.

F_D represents the fluid drag force on the particle, C_D represents the drag coefficient of the particle, and d_p represents the diameter of the particle.

The particles enter at the same time as the water flows, and the particles were released at the inlet boundary. The release time was set at 0.05 s, and 200 particles were released each time. In the experiment, the number of bacteria is not easy to be quantified. The number of simulated particles does not correspond to the number of bacteria but only qualitatively studies the movement law of bacteria, and we simulated a different number of particles and found that their motion laws were consistent.

As the experiment progressed, the biofilm continued to grow and formed large clusters with compact and dense structures that blocked the pores in the inlet (Figure 2(b) and 2(d)) (Xiao et al. 2020; Ke et al. 2021). Although the biofilm adhesion in both devices occurred for 5 min, the distribution was different. The initial formation of biofilm in the non-uniformly distributed micropost was mainly located in the region with small microposts (Supplementary Movie S1). On the contrary, only a few biofilms formed in the region with bigger microposts, showing an asymmetric distribution. However, the initial formation of biofilm in uniformly distributed microposts was more uniform in space (Supplementary Movie S2).

The front micropost column of the channel was not completely covered by biofilm (Figure 6), and the selected section including the biofilm not completely covered micropost column was named as morphology 2 and showed the morphology 2 evolution containing the biofilm detachment (Figure 6(a)–6(d)).

To further analyze the mechanism of biofilm detachment, we used finite element simulation to obtain the flow velocity field and shear stress distribution in morphology 1 and morphology 2, which had the same size of 380 μm × 600 μm (Figure 7). The channel width was 600 μm, and the segment along the channel length was 380 μm, as the formation of biofilm was a dynamic process and the growth of biofilm followed the same evolutionary trend. The selected section with a length of 380 μm represents the biofilm dynamics along the whole channel.

As for biofilms in morphology 1, at lower biofilm permeability, K_b = 1 × 10⁻¹⁵ m², the fluid flows through the biofilm very slowly, while the flow velocity was large in the area without the biofilm, that is because the lower biofilm permeability induces high pressure differences in the interface between the biofilm and the fluid (Figure 8(d)).

When K_b = 1 × 10⁻¹² m², the flow becomes easy to permeate through the biofilm; however, the flow velocity of the fluid in the rear was smaller and the flow path did not change significantly, compared to the situation with K_b = 1 × 10⁻¹⁵ m² (Figure 8(c)).

When K_b = 1 × 10⁻⁹ m², the flow becomes much easier permeating the biofilm, and the flow path was significantly altered and the fluid velocity is largely reduced, resulting in almost no velocity gradient (Figure 8(b)).

As for biofilms in morphology 2, at lower biofilm permeability, K_b = 1 × 10⁻¹⁵ m², a clear water fluid flows through the preferential path emerged, there was an obvious velocity gradient, and the water fluid flow barely conducted in the biofilm because the liquid could hardly permeate through the thick biofilm (Figure 8(h)).

When K_b = 1 × 10⁻¹² m², the flow path was not significantly changed, but the velocity was relatively reduced (Figure 8(g)).

When K_b = 1 × 10⁻⁹ m², the flow path was relatively unchanged, the flow velocity decreased further, and the flow also proceeded in the biofilm region (Figure 8(f)).

As for biofilms in morphology 1, at lower biofilm permeability, K_b = 1 × 10⁻¹⁵ m², biofilms were virtually impermeable but cause higher shear stress (maximum = 5.88 Pa) inside biofilms relative to that at the liquid biofilm interface, detaching was more likely to occur as an internal detach than as an external erosion (Figure 9(d)).

When K_b = 1 × 10⁻¹² m², fluid flow was still not easily permeable to biofilms, at which point the shear stress inside the biofilm was not obviously different from that at the liquid biofilm interface, which can proceed either as an internal detach or as an external erosion (Figure 9(c)).

When K_b = 1 × 10⁻⁹ m², the fluid flow was easily permeable to the biofilm, and the shear stress inside the biofilm was similar to that at the liquid biofilm interface, resulting in both the internal detach and interfacial erosion (Figure 9(b)).

As for biofilms in morphology 2, at lower biofilm permeability, K_b = 1 × 10⁻¹⁵ m², a clear shear stress gradient was found, with a large shear stress at the liquid biofilm interface (maximum = 4.14 Pa) than that at the interior of the biofilm, at which point the biofilm was more likely to detach in the form of an external erosion (Figure 9(b)).

When the permeability increased to 1 × 10⁻¹² and 1 × 10⁻⁹ m², a shear stress gradient was also found, shear stress at the liquid biofilm interface was still greater than that at the interior of the biofilm (Figure 9(f)–9(g)), so it was more likely that biofilms undergo detaching in the form of external erosion.

In this work, we studied the biofilm formation and detachment dynamics in porous structed microfluidic channels. Combined with experimental observation and mathematical modeling, we obtained the following conclusions:

• It showed that the bacterial transport velocity gradients across pore space influenced the initial biofilm growth, and the biofilm clogging preferentially occurred in the area with small microposts. Previous studies have also shown that rapid bacterial aggregation occurs at the pore throat (i.e. the region of channel contraction and expansion), partly due to the large velocity gradient (Lee et al. 2023).

• The study showed that the detaching of biofilms was related to its morphology and permeability (Karimifard et al. 2021; Wang et al. 2022), which cause biofilm detaching through shear stress with a value of about 5 Pa. In our previous study, about the biofilm growth in the microfluidic channel with a single micropost inside (Liu et al. 2022), we found that the shear stress threshold that is suitable for the biofilm adhesion was 0.3 Pa and proposed that there are two detaching modes of biofilm: internal detachment and external erosion.

The authors would like to thank Professor David A. Weitz and Professor Shmuel Rubinstein from Harvard University for their experimental support. The authors would like to thank the the National Natural Science Foundation of China for funding support (12372321, 11972074, 11772047 and 11620101001).

Y.T., C.T., Z.Z., S.L., F.D., D.Z., J.Z. and X.W. contributed to the study conception and design.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.