Multichannel Microfluidic Platform for Temporal-Spatial Investigation of Niche Roles of Pseudomonas aeruginosa and Escherichia coli within a Dual-Species Biofilm

ABSTRACT

In natural or man-made environments, microorganisms exist predominantly as biofilms forming surface-associated bacterial communities embedded in extracellular polymeric substances (EPSs). Often, biofilm reactors used for endpoint and disruptive analyses of biofilm are not suitable for periodic observation of biofilm formation and development. In this study, a microfluidic device designed with multiple channels and a gradient generator was used for high-throughput analysis and real-time monitoring of dual-species biofilm formation and development. We compared the structural parameters of monospecies and dual-species biofilms containing Pseudomonas aeruginosa (expressing mCherry) and Escherichia coli (expressing green fluorescent protein [GFP]) to understand the interactions in the biofilm. The rate of biovolume increase of each species in monospecies biofilm (2.7 × 10⁵ μm³) was higher than those in a dual-species biofilm (9.68 × 10⁴ μm³); however, synergism was still observed in the dual-species biofilm due to overall increases in biovolume for both species. Synergism was also observed in a dual-species biofilm, where P. aeruginosa forms a “blanket” over E. coli, providing a physical barrier against shear stress in the environment. The microfluidic chip was useful for monitoring the dual-species biofilm in the microenvironment, indicating that different species in a multispecies biofilm exhibit different niches for the survival of the biofilm community. Finally, we demonstrated that the nucleic acids can be extracted from the dual-species biofilm in situ after biofilm imaging analysis. In addition, gene expression supported that the activation and suppression of different quorum sensing genes resulted in the different phenotype seen in the biofilm. This study showed that the integration of microfluidic device with microscopy analysis and molecular techniques could be a promising tool for studying biofilm structure and gene quantification and expression simultaneously.

IMPORTANCE In natural or man-made environments, microorganisms exist predominantly as biofilms forming surface-associated bacterial communities embedded in extracellular polymeric substances (EPSs). Often, biofilm reactors used for endpoint and disruptive analyses of biofilm are not suitable for periodic observation of biofilm formation and development. Here, we demonstrate that a microfluidic device with multiple channels and a gradient generator can be useful for high-throughput analysis and real-time monitoring of dual-species biofilm formation and development. Our study revealed synergism in the dual-species biofilm, where P. aeruginosa forms a “blanket” over E. coli, providing a physical barrier against shear stress in the environment. Furthermore, different species in a multispecies biofilm exhibit different niches for the survival of the biofilm community. This study showed that the integration of microfluidic device with microscopy analysis and molecular techniques could be a promising tool for studying biofilm structure and gene quantification and expression simultaneously.

INTRODUCTION

In natural or man-made environments, microorganisms exist predominantly as biofilms forming surface-associated bacterial communities embedded in extracellular polymeric substances (EPSs). Biofilms are commonly found in various environments, and ~80% of the bacteria exist within a biofilm (1). Biofilms provide several benefits, such as mechanical resistance, protection from antibiotics and disinfectants, and adaptation to nutrient-limited environments (2). The development of a biofilm, including attachment of cells to a surface, multiplication, maturation, and production of a polymeric matrix, is an important survival strategy in natural and man-made niches (3).

Biofilms can attach themselves to virtually any surface, posing a serious problem in the health care, medical, and food industries (4–7). Within the biofilm, bacteria gain enhanced survivability toward harsh environments and the ability to exchange genetic information easily. Also, biofilms are believed to be a major contributor to the abundance of antibiotic resistance genes (ARGs) in the environment, leading to the rise in antibiotic resistance among the microbial communities in the environment (8). In the United States alone, antibiotic-resistant infections could cost up to $20 billion yearly, while biofilms on ships result in a 35 to 50% increase in fuel consumption due to drag (7, 9). However, investigation of biofilm development in vivo has been restricted due to the dynamic community within a complex structure, displaying spatial and functional heterogeneity (10). Therefore, in vitro biofilm study has been used to understand the formation and development of biofilm.

Biofilm studies have employed different reactors, such as microtiter plates, Robbins devices, the drip flow biofilm reactor, and rotary biofilm reactors to evaluate disinfectant strategies, rapid screening of antibiotics, effects of hydrodynamic conditions, gene expression, and enzymatic activities on biofilms (11). However, these biofilm reactors often resulted in endpoint analysis and are disruptive to the biofilm, and thus are not suitable for periodic measurements of biofilm formation and development. To study spatial and functional heterogeneity in a dynamic biofilm community, the fluorescence in situ hybridization (FISH) method has often been used (12, 13). FISH involves a sample processing step before imaging, increasing the probability of introducing artifacts into the results, thus, making it challenging to image a dynamic community. On the other hand, the microfluidics technique has been recognized as a promising method for studying microbial interactions within a biofilm. With the use of microfluidic devices, biofilm structures, such as water channels, tower structures, and densely packed cells, could be observed because of the precise control of the flow condition, accessibility to real-time observation, and high-throughput testing (14, 15). In addition, microfluidic devices have been applied with different configurations to understand biofilm formation and development in microenvironments. The feature-centered designs within microfluidic devices have been used with multiple channels for high-throughput experimentation, gradient mixing for dilution of reagents, real-time monitoring of biofilms to explore interactions, and gene quantification to explore the quantity of genetic materials (16–18). To our knowledge, no previous studies have reported on the development of a microfluidic device that integrates all of the necessary features for studying microbial interactions in dual-species biofilms. Furthermore, our device was designed with a bubble trap feature to minimize the occurrence of bubbles in the channels. This is a significant improvement over previous microfluidic designs, as bubbles can be a major issue in microchannels and interfere with image quality (19).

In this study, we present a microfluidic device with multiple channels for high-throughput analysis, a gradient generator, and real-time monitoring of biofilm formation and development (Fig. 1). Thus, the objectives of this study were (i) to validate our designed microfluidic device for biofilm formation and development, (ii) to demonstrate the development of monospecies and dual-species biofilms in real time for investigation of microbial interaction within the biofilm, and (iii) to evaluate the ARG abundance of the dual-species biofilm in the microfluidic device. Biofilm communities in the environment consist of many complex interactions resulting in antagonisms and synergisms, leading to difficulties in studying interactions in vivo (20). To overcome this limitation, we employed model microbial communities to investigate microbial interactions at the genetic and molecular levels (21, 22). Two bacteria were used in this biofilm study: Pseudomonas aeruginosa is widely used as a model organism to study biofilm formation (23), while Escherichia coli was less studied because most laboratory strains did not form a robust biofilm (24). This work demonstrates that the integration of microfluidics, microscopy, and molecular techniques is useful in studying the development and formation of dual-species biofilm.

FIG 1.

Schematic of the microfluidic design for biofilm study. (A) Microfluidic device design on CAD showing features such as the gradient mixers and the detection chamber; (B) growth of biofilm in the detection channel and fluorescence images obtained through confocal laser scanning microscopy (CLSM); (C) overall concept of the study to explore dual-species biofilm development through time by analyzing biovolume, coverage area, and roughness.

RESULTS

Validation and reliability of microfluidic device for a biofilm study.

To validate the microfluidic device for biofilm study, we examined the behaviors of fluids through microchannels and bacterial cell segmentation during image processing. Fluid parameters such as Reynolds number, fluid shear stress (FSS), fluid velocity, and gradient concentrations were diagnosed using COMSOL Multiphysics and ion chromatography (IC) analyses. Image segmentation for object recognition (i.e., microbial cells and cell aggregates) was conducted using FIJI.

The simulation analysis of the gradient concentrations demonstrated the proportional distributions of a reagent in each channel (Fig. 2). Increasing concentrations from 0 (blue) to 1 (brown) across channels C1 to C5 was determined by the simulation model (Fig. 2A). In addition, the ion chromatography analysis showed that the concentration of chloride at channel C1 was 0 mg/L, while 35.5 mg/L of chloride was measured at channel C5. Linear regression of the proportion of reagents obtained from the experimental and the simulation studies displayed a strong correlation between reagent concentrations and the channels from C1 to C5 (r² > 0.9). The percentage of deviation between the experimental and simulated results was relatively low (<10%), indicating the effectiveness of the gradient mixers in the microfluidic device. The average fluid velocity across the five channels from the simulation analysis was calculated. The velocities among all five channels were similar to u = 1.4 × 10⁻⁴ ms⁻¹ (Fig. 2D). Using u = 1.4 × 10⁻⁴ ms⁻¹, the Reynolds number (R_e) and the shear stress (τ) were calculated as R_e = 0.0248 and τ = 8.4 × 10⁻⁷ Pa, respectively, indicating the fluid in the microfluidic device followed the laminar flow (R_e < 2,000). Therefore, the fluid parameters illustrated that biofilm development in the microfluidic device could experience laminar flow and the same shear stress in all five channels. Furthermore, the gradient mixer in diluting the reagents also indicated that biofilms in an individual channel could be exposed to predetermined concentrations in a single experiment.

FIG 2.

Validation of the designed microfluidic chip through different fluid parameters. (A) Simulated concentration gradient across the 5 channels C1, C2, C3, C4, and C5; (B) simulated velocity across the 5 channels; (C) concentration gradient of chloride across the 5 channels obtained through IC (r² = 0.975) (n = 3); (D) simulated velocity across the 5 channels; (E) comparison between IC and COMSOL Multiphysics simulations (r² = 0.998) on the concentration gradient; (F) table showing the percentage of deviation between the IC results and the simulated results.

The initial image segmentation was conducted to analyze biofilm images taken from 0 to 96 h, as shown in Fig. 3. Low fluorescence and false signals interfered with the image analysis of biofilm in this study. For instance, high-quality biofilm images could not be obtained between 0 to 24 h due to the nondetection of the fluorescence from the biofilm. The biofilm images collected from 24 to 96 h were used for further analysis. Also, there were segmentation errors on images from 24 to 36 h, as shown by the red circles in Fig. 3. The segmentation was a critical step for identifying the regions of interest (ROI); the white regions corresponded to red and green for P. aeruginosa and E. coli, respectively. Since the bacteria in the biofilms were expected to increase in number as time progressed, the low fluorescence intensity between 0 and 48 h could be related to quorum sensing mechanisms, where the number of bacteria was too low to emit detectable fluorescence in this study. However, the biofilm images taken from 48 to 96 h enabled the subsequent analysis. The segmentation without errors suggested that the image analysis identified the ROI accurately. Nonetheless, the biofilm images taken from 48 to 96 h were used for downstream image processing.

FIG 3.

Validation of the segmentation step during image processing. The red circles show the area identified imprecisely by the segmentation step.

Structural parameters of monospecies biofilm.

Before exploring the interactions of dual-species biofilm, monospecies biofilm formation and development in the microfluidic device were characterized for in situ and real-time monitoring of biofilm structures in the microenvironment. Using structure parameters such as coverage area, biovolume, and surface roughness, we examined biofilm structures grown in microfluidic devices. Images from monospecies biofilm were analyzed by FIJI and BiofilmQ to calculate the parameters for illustrating biofilm structure. The change in biovolume demonstrated the dynamics of monospecies biofilms: there was a steeper change in E. coli biovolume between 52 h and 60 h, while there was a steeper change in P. aeruginosa biofilm between 80 h and 90 h (see Fig. S6B and S6C in the supplemental material). The different biovolume patterns suggest different growth patterns of each organism.

The biovolume of E. coli biofilm increased up to 29% from 48 to 67 h: values of 5.06 × 10⁶ μm³ and the peak biovolume of 1.82 × 10⁷ μm³ were observed at 48 and 67 h, respectively (Fig. 4A). The theoretical maximum biovolume in a channel from the biofilm image stack was calculated as 4.48 × 10⁷ μm³, based on the dimension (x = 800 μm, y = 800 μm, z = 70 μm) of the detection zone in the microfluidic chip. After the peak at 67 h, a 6% decrease in biovolume was observed at 96 h in E. coli biofilm. Additionally, there was also an increase in the coverage area from 1.45 × 10⁵ μm² at 48 h to the peak coverage area of 3.86 × 10⁵ μm² at 67 h (Fig. 4B). Also, the calculated maximum coverage area in each biofilm stack was 6.4 × 10⁵ μm², based on the dimension of the detection zone. We noticed a 38% increase in coverage area from 48 h to 67 h. After the peak, the colonized area of E. coli decreased up to 10% from 67 h to 96 h. It was also notable that E. coli biofilm colonized about 60% of the available space at 67 h. The surface roughness of E. coli biofilm increased from 0.10 at 48 h to the peak of 0.16 at 68 h (Fig. 4C). After the peak of surface roughness, it decreased to 0.14 at 96 h. Because the surface roughness was a dimensionless constant to determine the unevenness of the biofilm to represent the variation in thickness of the biofilm, the changes in surface roughness suggest that E. coli biofilm transited from a smoother biofilm to a rougher mushroom-like structure, before transiting back to a smoother biofilm.

FIG 4.

Structural parameters calculated for monospecies biofilms containing E. coli and P. aeruginosa. (A, B, and C) Biovolume, area coverage, and surface roughness, respectively, of E. coli biofilms; (D, E, and F) biovolume, area coverage, and surface roughness, respectively, of P. aeruginosa biofilms; (G) changes in proportion of bacteria as the biofilm develops. Red represents regions with a dense number of cells, while blue represents regions with a less dense number of cells. (H) Rate of change of biovolume and coverage area for each monospecies biofilm; (I) spatial distribution of each biofilm across time. The fluorescence is indicative of the density at the region, with a higher density of cells represented by yellow and a low density of cells represented by blue.

The structural parameters of biofilm were examined to investigate the biofilm formation and development of P. aeruginosa in a microfluidic device. The biovolume of P. aeruginosa biofilm increased from 1.11 × 10⁷ μm³ at 48 h to the peak biovolume of 2.54 × 10⁷ μm³ at 96 h (Fig. 4D). A decrease in biovolume was not observed between 48 h and 96 h. It was noted that 57% of the total volume within the detection region in the microfluidic chip was already occupied at 96 h. In addition, the coverage area increased from 2.91 × 10⁵ μm² at 48 h to the peak coverage area of 5.80 × 10⁵ μm² at 96 h (Fig. 4E). The surface roughness of P. aeruginosa biofilm increased from 0.21 at 48 h to the peak of 0.29 at 80 h and decreased to 0.06 at 96 h (Fig. 4F). The surface roughness parameter showed an overall decrease for both E. coli and P. aeruginosa, suggesting that high surface roughness corresponds to the period of low biovolume and coverage area. In contrast, low surface roughness corresponds to the period of high biovolume and coverage area. The surface roughness parameter might be the more important factor in early biofilm development as bacteria experienced opportunities of obtaining more resources effectively to increase the cell number and colonize more areas.

It was clear that the biovolume was closely related to the coverage areas from both P. aeruginosa and E. coli biofilms. There was also an increased rate of change in E. coli biofilm from 51 to 67 h, while the rate of P. aeruginosa biofilm increased between 77 h and 90 h. The biovolume and the coverage area plots suggested that monospecies biofilms followed a logistic population growth pattern, implying that a carrying capacity was present. For instance, E. coli biofilm reached carrying capacity at 67 h, while P. aeruginosa biofilm reached carrying capacity at 90 h. The images from Fig. 4G supported the observations of the logistic population growth pattern, where there were more red regions seen in E. coli biofilms than in P. aeruginosa biofilms at 72 h. This biofilm growth mode implied that the carrying capacity was reached earlier in E. coli biofilms than in P. aeruginosa biofilms.

An interesting observation was that P. aeruginosa tended to increase in the coverage area before increasing in thickness, while E. coli tends to increase in thickness before enlarging the coverage area (Fig. 4G). The fluctuation in biovolume and coverage area showed the dynamics of biofilm development (Fig. 4A, B, D, and E), suggesting that there were continuous changes to the colonized area and the cell number of the biofilm. Thus, the findings in monospecies biofilm suggested that the microfluidic device was capable of in situ real-time monitoring of biofilm and was suitable for interrogating biofilm formation development of both species in the microenvironment.

Biofilm structure and genetic quantification of dual-species biofilm.

To investigate the interaction of E. coli and P. aeruginosa biofilms, such as antagonism or synergism under a nutrient-rich environment, we examined the biofilm structure and development of dual-species biofilm in the microfluidic channels. In addition, structural and genetic quantifications of the biofilm were also interrogated to understand microbial interactions between E. coli and P. aeruginosa. The change in biovolume demonstrated the dynamics of dual-species biofilms (Fig. S6A), where there were differences in growth patterns compared to the monospecies biovolume. The different biovolume patterns suggest that interspecies interactions can alter the growth patterns of each organism.

The biovolume changes in P. aeruginosa biofilm fluctuated between 48 and 96 h, with two distinct peaks at 72 and at 88 h (1.58 × 10⁷ μm³ and 1.61 × 10⁷ μm³, respectively). On the other hand, there were three distinct low points at 52, 78, and 96 h. At the highest biovolume, P. aeruginosa occupied about 40% of the biofilm, while P. aeruginosa covered about 19% of the biofilm at the lowest biovolume. The biovolume of E. coli biofilm fluctuated from 48 h to 96 h; however, fewer distinct peaks were observed compared to P. aeruginosa. At the highest biovolume, E. coli occupies about 35% of the detection region, while E. coli presented about 12% of the detection region at the lowest biovolume.

Figure 5B shows that the coverage area of P. aeruginosa fluctuated, with two distinct peaks at 72 and 88 h, with coverage areas of 3.08 × 10⁵ μm² and 3.47 × 10⁵ μm², respectively. Similar to the biovolume, there were three distinct low points at 52, 78, and 96 h. In the highest coverage area, P. aeruginosa occupies about 54% and 30% of the detection region of the highest and lowest coverage areas, respectively. The trend of the coverage of E. coli also fluctuated; however, fewer distinct peaks were observed compared to P. aeruginosa. At the time points representing the highest coverage area, E. coli occupies about 52% of the detection region, while at time points representing the lowest coverage area, E. coli occupies about 27% of the detection region. Also, the coverage areas of P. aeruginosa and E. coli biofilms showed similar trends compared to the biovolume. P. aeruginosa occupied about 54% of the substratum in the highest coverage area, while at time points representing the lower coverage area, P. aeruginosa occupied about 30% of the substratum. On the other hand, E. coli resided in about 52% and 27% of the substratum of the highest and lowest coverage areas, respectively.

FIG 5.

Structural parameters calculated for dual-species biofilms containing E. coli and P. aeruginosa. (A and B) Biovolume and area coverage of each of the bacteria in a dual-species biofilm; (C and D) ratio of biovolume and area coverage of E. coli to P. aeruginosa, where the points above the red line are biased toward E. coli, while the points below are biased toward P. aeruginosa; (E) surface roughness of the dual-species biofilm; (F) rate of change of biovolume and coverage area for each of the bacteria in the dual-species biofilm; (G) distribution of bacteria and the proportion of bacteria between 48 h to 96 h; (H) spatial distribution of each biofilm across time. The fluorescence is indicative of the density at the region, with a higher density of cells represented by yellow and low density of cells represented by blue.

Furthermore, the ratios of the coverage areas and biovolumes in E. coli to P. aeruginosa were used to examine antagonism or synergism because a relatively large coverage area and biovolume could be advantageous for the bacterium. As shown in Fig. 5C and D, ratios above 1 might be considered advantageous to E. coli, while a ratio below 1 could be advantageous to P. aeruginosa. The cyclic pattern in Fig. 5C and D suggested that antagonism occurred due to the diffusion limitation of nutrients into the biofilm. However, there was an overall increase in both biovolume and coverage area for E. coli and P. aeruginosa between 48 h to 96 h. Thus, we propose that the cyclic pattern observed is a synergistic interaction to increase the survivability of both organisms.

Surface roughness indicated spatial competition in dual-species biofilm. The surface roughness decreased from 0.27 at 48 h to 0.19 at 76 h. An increase in surface roughness was observed, with a surface roughness of 0.32 until 96 h. In the dual-species biofilm, E. coli (green) was found nearer to the substratum, while P. aeruginosa (red) was found nearer to the surface of the biofilm (Fig. 5F), suggesting that P. aeruginosa experienced higher fluid shear stress (FSS) and higher concentration of nutrients than E. coli. Although having a higher nutrient concentration could be a factor desirable for both bacteria, the FSS might not be desirable to E. coli compared to P. aeruginosa, leading to P. aeruginosa protecting E. coli from it. Thus, this observation suggested that the more tolerant species could protect the less tolerant species in a dual-species biofilm. Also, we speculated that synergism was present between P. aeruginosa and E. coli in a dual-species biofilm, where P. aeruginosa provides protection to E. coli, while E. coli could provide structural support for P. aeruginosa. The Cube_volumeFraction function in BiofilmQ calculated the ratio of an occupied cube to the total cube volume (e.g., 1 was fully occupied, and 0 was unoccupied). The ratios of Paraview in the heat map showed the proportion of bacteria within the biofilm, with blue being less dense and red being denser. In the dual-species biofilm, the bacteria were less dense, as seen from the mostly blue regions, than monospecies biofilm (Fig. 5G), suggesting that the bacterial cells in dual-species biofilms developed more slowly than monospecies biofilms.

The gel electrophoresis results showed that five amplicons produced with the designed primers were confirmed, indicating that the microfluidic device could be used for gene analysis from the biofilm. The five primers were used to amplify gentamicin-resistant genes of P. aeruginosa, while the other primers were used to detect chloramphenicol-resistant genes in E. coli. Figure S7 shows that all designed primers successfully amplified the regions of antimicrobial-resistant genes as intended, with the same expected amplicon lengths as the ones in the supplemental material. Also, quantitative PCR (qPCR) results were tested using Wilcoxon signed-rank test at a 95% confidence interval, and resistant genes at both 48 and 96 h were statistically similar (P > 0.05). Similar to the biovolume, the gene abundance of the antibiotic resistance genes increased, Still, there was a statistically insignificant difference at 46 and 96 h, suggesting that the cells continued to grow within the dual-species biofilm.

Identification of genes of interest in dual-species biofilm using transcriptomics data.

To investigate the expression of various genes in a biofilm, the differentially expressed genes (DEGs) were mapped to the Kyoto Encyclopedia of Genes and Genomes (KEGG) database. The enrichment analysis will provide insights into the activation and suppression of genes in the development of dual-species biofilm.

There were 4,512 and 5,678 DEGs associated with E. coli and P. aeruginosa, respectively. The heat map with hierarchical clustering based on the DEGs demonstrated that genes such as grcA, csrC, and csrB showed lower expression levels in E. coli (Fig. S8A), while genes such as ssrA, oprF, and rnpB in P. aeruginosa (Fig. S8B) showed similar expression levels to the dual-species biofilm developed. The different patterns suggest that there were differences in gene expression due to the interspecies interactions between E. coli and P. aeruginosa. These DEGs revealed 19 KEGG pathways that were significantly enriched in E. coli (Fig. 6A), while 12 KEGG pathways were significantly enriched in P. aeruginosa (Fig. 6B). The biofilm-related pathway was observed to be enriched: for instance, “flagellar assembly” was activated and suppressed in E. coli and P. aeruginosa, respectively. The R package pathview was used to visualize the gene expression in the biofilm formation pathway of E. coli and P. aeruginosa (Fig. S4 and S5).

FIG 6.

(A) E. coli and (B) P. aeruginosa gene expression mapped to KEGG pathways. The plot shows significantly enriched pathways at a 95% level of significance between 48 h and 96 h. “GeneRatio” indicates the ratio of differentially expressed gene number to the total gene number in a certain pathway. The color and size of the dots represent the adjusted P value and the differentially expressed gene number, respectively.

The activation and suppression of different quorum sensing genes resulted in the different phenotype seen in the biofilm. Additionally, both E. coli and P. aeruginosa exhibited different quorum sensing systems. For instance, the suppression of the lsrR gene resulted in the activation of the wza gene and ag43 gene, encouraging colonic acid biosynthesis and adhesion of E. coli (Fig. 7A). P. aeruginosa on the other hand has a Pseudomonas quinolone signal (PQS) system, Rhl system, and Las system for quorum sensing. The Rhl system and Las system were suppressed, while the PQS system was activated; thus, there was less swarming motility and twitching motility seen in P. aeruginosa (Fig. 7B).

FIG 7.

Genes related to quorum sensing, where green represents suppressed genes, while red represents activated genes. (A) Quorum sensing genes in E. coli; (B) quorum sensing genes in P. aeruginosa.

DISCUSSION

In this study, we evaluated the microfluidic device as a tool for investigating dual-species biofilm development. The structural parameters of the monospecies and dual-species biofilms of P. aeruginosa and E. coli were calculated to understand interspecies interaction within biofilm in the microfluidic device. The quantity of antibiotic-resistant genes was determined and gene expression analysis was performed in the dual-species biofilm to evaluate the feasibility of detection and quantification methods in biofilms. Diffusion limitation becomes more significant in biofilms as the diffusion distance increases and fluid flow decreases (25). However, due to the laminar flow in the microfluidic device, diffusion is governed by the concentration gradient of the solute and the biofilm will be experiencing shear stress in the direction of the fluid. Photo-toxicity on the biofilm due to the use of a confocal microscope has been documented; however, the advantages of overcoming background fluorescence, reflections, and bacterial autofluorescence outweigh the disadvantage of the sensitivity of biofilm imaging (26). In addition, the imaging conditions were well within the limits of the previously reported study, indicating that the effects from photo-toxicity were insignificant (26). This study demonstrated that a microfluidic device integrated with confocal laser microscopy and molecular techniques could be suitable for studying dual-species biofilm development and interspecies interactions in microenvironments.

The structural parameters of biofilm such as coverage area, biovolume, and surface roughness described biofilm development in monospecies- and dual-species-specific biofilms. In this study, an inverse relationship between the biovolume and surface roughness was observed (Fig. 4D and F), suggesting that a rougher biofilm with mushroom-like structures is related to low cell number in the biofilm. The mushroom-like structures increased biofilm surface roughness and the contact area around the biofilm surface (27), leading to enhanced nutrient transfer. With the observed logistic growth pattern of the biovolume and coverage area of monospecies biofilms, the maximum was not reached due to the transfer limitation of resources to the biofilm. At early biofilm development, monospecies biofilm could enhance nutrient transfer by increasing surface roughness, thus, increasing bacterial numbers and colonized area. As the peak biovolume or coverage area was reached, a reduction in surface roughness indicated a decrease in nutrient transfers, leading to a decrease in bacterial numbers and the colonized area. Previous reports of mathematical modeling on biofilm supported our observations, where the decrease in nutrient transfer resulted in the development of pillar-like structures in the biofilm to increase roughness and increase nutrient transfer into the biofilm (28, 29). Thus, the surface roughness of the biofilm could be regulated by quorum sensing molecules as a response to cell density in the biofilm. Surface roughness directly influences nutrient transfer into the biofilm, causing an increase or decrease in bacterial growth and colonized area in the biofilm (27, 30).

The formation and development of dual-species biofilm demonstrated spatial competition and microbial interaction (Fig. 5C and D), where points above the red line show that E. coli outcompetes P. aeruginosa, while the opposite occurs with points below the red line. The formation of multispecies biofilm was reported in three main steps (30): (i) the attachment of the primary colonizer with EPS production on the substratum; (ii) the attachment of the secondary colonizers onto the microcolonies developed by the primary colonizer (30); and (iii) dispersion of the biofilm controlled by an environmental stimulus, such as changes in temperature and pH (31). EPS production from the primary colonizer was suggested as the most crucial step as it acted as an “intercellular cement” during multispecies biofilm development to encourage the attachment of the secondary colonizers (32). P. aeruginosa biofilm has been known as a strong biofilm former in previous studies (23, 33, 34). In this study, P. aeruginosa acted as a primary colonizer due to its ability to form robust biofilm by excreting EPSs containing three polysaccharides (e.g., alginate, Pel, and Psl), playing distinct roles in biofilm attachment (35). In contrast, E. coli served as the secondary colonizer to attach itself onto P. aeruginosa, as E. coli formed a less robust biofilm (36). In the process of the dual-species biofilm development from the substratum, P. aeruginosa eventually colonized the surface of the biofilm within the EPS network of P. aeruginosa biofilm, while E. coli colonized the substratum, with strong adhesion to the surface (Fig. 5G). It was previously reported that the dispersal events in P. aeruginosa biofilm actively evacuate the interior of the dual-species biofilm, revealing “hollow” spaces for E. coli to colonize (36), leading to the colonization of interior spaces by E. coli. Also, another study reported that the “blanket” effect by P. aeruginosa caused a decrease in Agrobacterium tumefaciens biomass, suggesting that this effect was linked to antagonism in a biofilm (37). These studies reported similar observations to our study, where P. aeruginosa formed a “blanket” and colonized the exterior of the dual-species biofilm. In addition, the structural parameters showed a decrease in biovolume and coverage areas of both species in the dual-species biofilm compared to the monospecies biofilm in this study. It was hypothesized that the greater the coverage area and biovolume, the more advantageous the biofilm is to the bacterium (38, 39). With the overall increases in biovolume and coverage area of both P. aeruginosa and E. coli in the dual-species biofilm, the “blanket” effect by P. aeruginosa suggested that the active evacuation of P. aeruginosa from the interior of the biofilm allowed E. coli to colonize the spaces, encouraging the overall growth of both species in a dual-species biofilm. Thus, antagonism between species in a dual-species biofilm was not observed in this study. When comparing the surface roughness of the dual-species biofilm to the monospecies biofilm, the surface roughness was higher in the dual-species biofilm (0.2 to 0.3). Thus, P. aeruginosa also played the role of being more efficient at obtaining nutrients, supporting the overall development of the dual-species biofilm.

Interspecies interactions in a multispecies biofilm induced various phenomena, such as improved attachment of biofilms (40), formation of denser biomass (41), encouraging the exchange of genetic materials (30), and influencing survivability of overall biofilm community (42). Interspecies interactions were characterized as antagonistic and synergistic, leading to various functionalities and organizations of multispecies biofilms (43). The lower biovolume of each species of bacterium in the dual-species biofilm than in the monospecies biofilm in our study contrasted with the previous study on P. aeruginosa and Candida albicans (44). This contrast was likely due to the different microfluidic environments, as it was suggested that higher retention of bacterial cells in the microfluidic chambers might be due to the decrease in active dispersal or disruption by fluid flow (44), thus, causing the discrepancy in biovolume of the biofilm between this study and the previous study. The previous work on the dual-species biofilm of Staphylococcus aureus and Candida albicans in a microfluidic device showed that a larger biofilm coverage area was observed than those of their respective monospecies biofilms (16). This study agreed with the findings on coverage area, where our dual-species biofilm showed a minimum coverage area of 57%, while the monospecies biofilm showed a minimum coverage area of about 31.3%. The comparison between the biovolume and coverage area for the individual organism in the monospecies biofilm and in the dual-species biofilm might be a better indication of interspecies interactions. For instance, each species showed higher biovolume and coverage area in a monospecies biofilm than in a dual-species biofilm, suggesting the synergism accommodated each other in the limited space of the biofilm. However, previous studies did not explore the spatial distribution of the microorganisms within a dual-species biofilm, which can provide information on the different roles that each organism plays in a dual-species biofilm (Fig. 5G). Spatial distribution data revealed distinct patterns between monospecies and dual-species biofilms, indicating that interspecies interaction plays an important role in multispecies biofilms in the environment (Fig. 4I and Fig. 5H). In monospecies biofilms, voids were attributed to EPSs due to the lack of fluorophore (Fig. 4I). Conversely, in a dual-species biofilm, the voids observed in P. aeruginosa biofilm were predominantly occupied by E. coli biofilm (Fig. 5H), suggesting niche roles of both organisms in a dual-species biofilm. In this study, P. aeruginosa colonized the surface of the biofilm as it was more efficient at obtaining resources, while E. coli colonized the surface of the substratum, suggesting that it played the role of creating structural support for the dual-species biofilm. Thus, our work revealed that each species in the biofilm played niche roles in biofilm development.

Oxygen availability influences the development of biofilms. It was reported that although an anaerobic condition leads to the inability of P. aeruginosa to develop a thick biofilm, it is essential for the overall development of the biofilm (45). The presence of oxygen increased the expression of type 1 pili in E. coli, leading to improved biofilm attachment onto surfaces (46), thus, suggesting that the oxygen gradient within monospecies biofilms plays an important role toward their development. However, studies on the influence of oxygen on dual-species biofilm are limited. While E. coli is a facultative anaerobe, P. aeruginosa is capable of arginine fermentation and pyruvate fermentation, which does not lead to growth (47). P. aeruginosa was observed to be distributed at the top of the biofilm, suggesting a greater availability of oxygen from the fluid flow. In the dual-species biofilm, the presence of interspecies interactions encourages the growth of bacteria in different locations in a biofilm, where E. coli can grow at the bottom of the biofilm, where there is less available oxygen, while P. aeruginosa can grow at the top of the biofilm, where there is more available oxygen.

Biofilm detection using gene quantification and gene expression was performed in the microfluidic platform to serve as a complementary tool for confocal microscope analysis (Fig. 5B, Fig. 6, and Fig. 7). The gentamicin-resistant genes in P. aeruginosa and chloramphenicol-resistant genes in E. coli were used for qPCR quantification on the biofilm. Although there were microfluidic devices reported for the use of qPCR, such as the microfluidic qPCR (MFQPCR) tool for the quantification of methanogens (18), and also high-throughput qPCR (HT-qPCR) for the quantification of food quality (48), there were no other studies to apply molecular techniques to biofilms grown in microfluidic devices. Dual-species biofilm was extracted from the microfluidic device, and subsequent DNA and RNA extraction was performed successfully. Antibiotic-resistant gene (ARG) expression was found to increase with the selection pressure caused by exposure to antibiotics (49). Since no antibiotics were added in this experiment, findings from ARG abundance show no significant differences, which was consistent with previous studies (50, 51). Gene expression analysis demonstrated that the suppression of the lsrR gene led to increased colanic acid biosynthesis in E. coli, which plays an important role in forming the complex three-dimensional (3D) structure of the biofilm (52). Our study supported the identification of three major quorum sensing systems in P. aeruginosa: specifically, the Las, Rhl and PQS systems. Of the three systems, only the PQS system was activated, suggesting that it may play an important role in regulating interactions between E. coli and P. aeruginosa and their assemblage in a dual-species biofilm (53).

In conclusion, our study demonstrated that the microfluidic device was capable of biofilm formation in parallel microchannels with a diffusive mixer for cells and media for dual-species biofilm monitoring. The presence of niche roles of each species in a dual-species biofilm was demonstrated through real-time monitoring of structural parameters. In addition, the microfluidic device integrated with molecular techniques can be a promising tool in studying gene quantification and expression during biofilm development. Slight alterations of the microfluidic device could be made to fit research needs: for instance, the effects of fluid flow rate could be explored by creating narrower channels in the detection channel (54). Future studies on the effects of antibiotics, temperature, disinfectant, fluid flow rate, inoculum concentration, and bacteriophages could be performed in a controlled environment provided by the microfluidic device to obtain insights into microbial interactions in the dual-species biofilm.

MATERIALS AND METHODS

Bacterial strains and growth media.

Pseudomonas aeruginosa PAO1 (41) and Escherichia coli SCC1 (55) were selected as bacterial models in this study. For the development of P. aeruginosa for this study, a pUC18-based vector plasmid was used for the construction of R6K replicon-based delivery plasmids. P. aeruginosa PAO1 was transformed through electroporation, and the fluorescence sequence was chromosomally inserted into the Tn7 region. P. aeruginosa maintained its mCherry fluorescence over 7 days in the absence of gentamicin selection.

E. coli SCC1 was developed using E. coli MG1655, with chromosomal insertion of P_A1/04/03 gfpmut3*. E. coli MG1655 was transformed through electroporation and chromosomally inserting pCCS167 plasmid between regions A and B. It was reported that the green fluorescent protein (GFP) fluorescence of E. coli SCC1 did not affect the coculture population relationships and was stable over exponential and stationary phases. Additionally, the pCCS167 plasmid contains a chloramphenicol resistance gene. The colonies of P. aeruginosa PAO1 and E. coli SCC1 were cultured in Luria-Bertani (LB) broth. All cultures were incubated at 30°C with shaking at 150 rpm. The culture was grown for 18 h before the experiment.

Microfluidic device design and fabrication.

The schematic diagram presented in Fig. 1 shows the design features of the microfluidic device used in this study. The microfluidic device was designed to dilute reagents and to perform high-throughput experimentation, including cell injection, a gradient generator, and detection channels. The microfluidic device adopted bubble trap features (19) and had a surface coating with polyvinyl alcohol (PVA) (56) to reduce the occurrence of bubbles because bubbles could affect the biofilm growth during experimentation within the microfluidic device (57). The microfluidic device was initially designed using CAD software (AutoCAD2019). The connecting channels and the gradient generator had a width and height of 100 μm, while the detection chambers were designed with a width of 800 μm and height of 100 μm. In addition, the bubble trap features had a hole of about 50 μm to prevent huge bubbles from entering the channels (19).

The AutoCAD design was printed onto a PET (polyethylene terephthalate) mask from Nanoimprint Tech(s) Pte., Ltd., and a photolithography process was then carried out by Mechanobiology Institute (MBI) to obtain the master mold of the microfluidic device. The microfluidic devices were then fabricated by pouring polydimethylsiloxane (PDMS) onto the mold. The PDMS was obtained from Dowsil Sylgard 184 and made with a 10:1 ratio of liquid elastomer to the curing agent provided. Briefly, the air bubbles were removed from PDMS by degassing for 1 h before curing the PDMS in the oven at 80°C for 1 h. The polymerized PDMS with the imprinted microfluidic features was then peeled off from the mold. A biopsy puncher was then used to create openings for the various inlets and outlets in the device. Following this, surface activation was carried out using a Harrick oxygen plasma cleaner for 1 min before pressing firmly on a glass coverslip. One percent PVA was then injected into the microfluidic device for 10 min before flushing the PVA using deionized (DI) water. The microfluidic device was then placed back into the oven at 80°C for about 18 h. The device was then cooled to room temperature before use.

Characterization and validation of the microfluidic device.

The Reynolds number was calculated to determine the fluid characteristic within the microfluidic device. In the case of a channel with a rectangular cross-section, the equivalent channel diameter is determined from equation 1 before substituting it into the Reynolds number equation in equation 2 (58):

$$D_z=\frac{4A}{O}$$ (1)

$$R_e=\frac{u{D}_z\rho }{\mu }$$ (2)

D_z is the equivalent channel diameter, A is the cross-sectional area of the channel, O is the wetted parameter of the channel, R_e is the Reynolds number, u is the liquid velocity, ρ is the liquid density, and μ is the liquid viscosity.

COMSOL Multiphysics modeling software v.5.5 was used to carry out the fluid dynamics simulation. A 2D microfluidic design was drawn under the geometry component of the software, and the fine mesh was constructed within the device. Simulations were performed using the laminar flow and the transport of diluted species module under a stationary study condition.

The Navier-Stokes equations governing the motion of fluid were used as the basis for the laminar flow module as described in equation 3, while equation 4 is the continuity equation. Equation 3 represents the conservation of momentum, while equation 4 represents the conservation of mass:

$$\rho \left(\frac{\partial u}{\partial t}\hspace{0.17em}+\hspace{0.17em}u\times \nabla u\right)=-\nabla p+\nabla \times \left(\mu \left(\nabla u\hspace{0.17em}+\hspace{0.17em}{\left(\nabla u\right)}^T\right)\hspace{0.17em}-\hspace{0.17em}\frac{2}{3}\mu \left(\nabla \times u\right)I\right)\hspace{0.17em}+\hspace{0.17em}F$$ (3)

$$\frac{\partial \rho }{\partial t}\hspace{0.17em}+\hspace{0.17em}\nabla \times \left(\rho u\right)=0$$ (4)

u is the fluid velocity, p is the fluid pressure, ρ is the fluid density, and μ is the fluid dynamic viscosity.

Fick’s law governs diffusion in the transport of diluted species modules. Fick’s first law of diffusion states that the molar flux due to diffusion is proportional to the concentration gradient, as represented in equation 5. By incorporating the conservation of mass, Fick’s second law is obtained, describing that the rate of change of concentration at a point in space is proportional to the second derivative of concentration with space, as represented in equation 6:

$$N_i=-D_i\nabla c_i$$ (5)

$$\frac{\partial c_i}{\partial t}=D_i{\nabla }^2{c}_i$$ (6)

For species i, N_i is the molar flux, D_i is the diffusion coefficient, and c_i is the concentration.

At the laminar flow module, no-slip boundary conditions were set on the walls of the channels, with a mean velocity of 1.67 × 10⁻² ms⁻¹ set at the inlets and zero pressure set at the outlet. At the transport of the diluted species module, 0 mol/m³ was set at one of the inlets and 1 mol/m³ was set at the other. Additionally, the fluid property was considered to be similar to water (μ = 1 mPa·s, ρ = 1,000 kg·m⁻³). A cut line was drawn at the center of the detection chambers, and the concentration of reagents and fluid velocity were calculated through the software.

To further validate the accuracy of the gradient generator, 1 mM potassium chloride (KCl) was prepared to be injected from one inlet, and DI water was injected from the other inlet. Both KCl and DI water were injected at 200 μL/min using a syringe pump (Fusion 101; Chemyx, Inc.). Ten milliliters of the resultant solution from each outlet was collected and prepared to test for the presence of chloride using ion chromatography (IC).

Since the expected fluid flow within the microfluidic device is laminar, the fluid shear stress (FSS) was calculated using equation 7 (59):

$$\tau =\frac{6Q\mu }{w{h}^2}$$ (7)

Q is the flow rate, μ is the dynamic viscosity, w is the width of the channel, and h is the height of the channel.

Experimental setup for biofilm development in microfluidic devices.

Before the start of the microfluidics experiment, food dye was added into 6 mL of 70% ethanol before injection into the device at a flow rate of 200 μL/min to check for defects and, at the same time, to sterilize the device. Next, 10 mL of DI water was injected into the device to remove the food dye and ethanol to prepare the device for the addition of bacterial cells. The overnight culture of the two bacteria, as described in the previous section, was washed and resuspended to an optical density at 600 nm (OD₆₀₀) of 1 in fresh LB broth.

To develop a monospecies biofilm, we injected a total of 6 mL of the diluted bacterial culture into the microfluidic device through the cell inlets on the device at a flow rate of 200 μL/min. The fluid flow was stopped for 2 h to allow the attachment of cells onto the substratum of the device. Subsequently, fresh LB broth was injected from the inlets of the device at a flow rate of 10 μL/min. LB broth flowed continuously for 48 h before subsequent hourly imaging from a confocal microscope (FluoView FV10i; Olympus) until the 96-h mark. The diluted bacterial culture was mixed in a fresh 50-mL tube with a 1:1 ratio of E. coli to P. aeruginosa to develop a dual-species biofilm. A total of 6 mL of the mixture of two bacteria was injected into the microfluidic device through the cell inlets on the device at a flow rate of 200 μL/min. The experimental preparation was the same as the monospecies biofilm setup.

Time-lapse confocal microscopy and image processing.

All time-lapse images for the spatial characteristics of the biofilms were taken using a confocal laser scanning microscope (FluoView FV10i; Olympus). At each time point, six images were taken as a z-stack, resulting in an ~70-μm-thick biofilm. The Olympus Fluoview version 4.1 software was used for the operation of the confocal microscope. mCherry (580-nm excitation, 610-nm emission) and Alexa Fluor 488 (499-nm excitation, 520-nm emission) were selected from the dye database in the software for the detection of the fluorescence protein. The .oif files were obtained at the end of each experiment and analyzed using image processing software. BiofilmQ was used mainly for the 3D visualization of the biofilm, while time-series quantification was achieved using FIJI (60) and Python3 (61). Ten independent flow cells from two separate experiments were evaluated for both monospecies and dual-species biofilms (×4 fields of view for each flow cell = 40 data points), respectively. Images were collected using a 20× lens objective.

Using BiofilmQ (62), images were aligned along the z axis and segmented using the Otsu threshold. Subsequently, local surface roughness data were obtained through parameter calculations. .vtk files were exported from BiofilmQ to be read in ParaView 5.10.1, where Cube_VolumeFraction was visualized as a heat map to understand the proportion of bacteria at a different layer of the biofilm. Furthermore, the distribution of E. coli and P. aeruginosa can also be differentiated and visualized in a dual-species biofilm. Biovolume and maximum coverage area were analyzed using FIJI, where segmentation was performed using the Otsu threshold, followed by the “Make Binary” function. The area of each image was then quantified and saved in a .csv file. Python was then used to calculate the biovolume using the sum of trapezoids (63) and to identify the largest coverage area among the biofilm layers. The ggplot2 package from R was used to plot different graphs in this article (64).

DNA extraction from microfluidic device and bacterial gene quantification.

Biofilm was extracted from the detection chamber by flushing at a flow rate of 500 μL/min into a 15-mL Falcon tube using DI water, collecting about 3 mL of the sample. DNA extraction was then performed on the sample using a blood and tissue kit to obtain the DNA required for PCR and qPCR. The bacterial genes of interest were the 16S rRNA gene and their respective antibiotic-resistant genes. E. coli contains a chloramphenicol-resistant gene, while P. aeruginosa contains a gentamicin-resistant gene, and their sequences were obtained from GenBank. Primers were designed from the sequence of the antibiotic-resistant genes using the IDT PrimerQuest Tool. The sequences of the primers targeting 16S rRNA are given in Table 1.

TABLE 1.

Forward and reverse primers used to target 16S rRNA in this study

Primer	Sequence	Amplicon length (bp)	Reference
BACT1369F	CGGTGAATACGTTCYCGG	123	63
PROK1492R	GGWTACCTTGTTACGACTT

The PCR for the 16S rRNA primers was run in the Veriti 96-well fast thermocycler with an initial denaturation at 95°C for 3 min, followed by 35 cycles of amplification (denaturation at 95°C for 15 s, annealing at 55°C for 15 s, extension at 72°C for 30 s) and a final extension at 72°C for 5 min. For the antibiotic-resistant primers, the PCR ran with an initial denaturation at 95°C for 3 min, followed by 30 cycles of amplification (denaturation at 95°C for 15 s, annealing at 60°C for 15 s, extension at 72°C for 30 s) and a final extension at 72°C for 5 min. All PCR analyses were performed in triplicates with a reaction volume of 25 μL. The reaction volume was made with 12.5 μL of GoTaq G2 green master mix (Promega) and 1 μL each of 5 μM forward and reverse primers, followed by 10.5 μL of nuclease-free water. Gel electrophoresis was made using TAE (Tris-acetate-EDTA) buffer with 1% agarose, followed by a voltage of 120 V for 30 min. A 100-bp DNA ladder was used as a reference to determine the length of the amplicon. Subsequently, the E-gel imager (Life Technologies) was used to observe the band images on the gel.

The qPCR for all of the primers was run in a Step One Plus real-time PCR system with an initial denaturation at 95°C for 10 min, followed by 40 cycles of amplification (denaturation at 95°C for 15 s, annealing at 57°C for 15 s, extension at 72°C for 30 s) and a final extension at 95°C for 15 s. All qPCR analyses were performed in triplicates with a reaction volume of 20 μL. The reaction volume was made with 10 μL of PowerUp SYBR green master mix (Applied Biosystems) and 1 μL each of 5 μM forward and reverse primers, followed by 8 μL of nuclease-free water. Standard curves for the assays were prepared using purified PCR amplicons amplified from E. coli (chloramphenicol resistance gene and 16S rRNA gene) and from P. aeruginosa (gentamicin resistance gene).

RNA extraction from microfluidic device for sequencing.

Biofilm was extracted from the detection chamber by flushing at a flow rate of 500 μL/min into a 15-mL falcon tube using DI water, collecting about 3 mL of the sample. Two independent samples were collected at 48 h, while five independent samples were collected at 96 h. RNA extraction was performed using PureLink RNA minikit (Invitrogen) following an initial guanidinium thiocyanate solution TRIzol step using TRI reagent (Sigma-Aldrich) to enhance the quality of RNA according to the manufacturer’s instructions.

Sequencing of RNA was performed by Singapore Centre for Environmental Life Science Engineering (SCELSE). The stranded total RNA prep with the Ribo Zero Plus library was prepared. Briefly describing the steps taken, rRNA was removed by using Ribo-Zero Plus depletion kit (Illumina). Barcodes were ligated to the RNA using TruSeq RNA CD indexes (dual barcoded). The library was sequenced on an Illumina HiSeqX (150-bp paired sequences).

Transcriptome analysis.

Data analysis was performed in the R statistical environment (v.4.2.2), and the data were prepared using Hisat2, followed by Stringtie (65). The FASTA file of the genome and GFF file of the genome annotation were obtained from NCBI for Pseudomonas aeruginosa PAO1 (NCBI Reference Sequence database accession no. NC_002516.2) and Escherichia coli MG1655 (NCBI Reference Sequence database accession no. NC_000913.3). The index file was built in Hisat2 using the Hierarchical Ferragina Manzini (HFM) index, where the RNA reads were aligned to the genome. Subsequently, commands in Stringtie were performed to assemble the transcripts.

The output count data set from Stringtie was prepared using DESeqDataSetFromMatrix() from the R package pasilla. Differential gene expression analysis (66) was performed with the R package DESeq2 using the command DESeq(). The differentially expressed genes (DEGs) were mapped to the Kyoto Encyclopaedia of Genes and Genomes (KEGG), pathway analysis was conducted, and the results were visualized using the clusterProfiler package in R (67). Briefly, the clusterProfiler package facilitates the mapping of the DEGs to the KEGG pathway using the command gseKEGG() at a P value cutoff of 0.05 to produce significant KEGG pathways. Visualization of the pathway was performed using the R package pathview (68).

Statistical analysis.

The wilcox.test() function in R (v.4.2.2) was applied to test for statistical differences between the compared groups. The Wilcoxon signed-rank test was performed with a 95% confidence interval, and statistical differences between the compared groups were determined with a P value of <0.05.

Data availability.

RNA sequencing data for monospecies and dual-species biofilms in this study have been deposited in the NCBI BioProject database under accession no. PRJNA980130.