A 3D-Printed Customizable Platform for Multiplex Dynamic Biofilm Studies

Abstract

Biofilms are communities of microbes that colonize surfaces. While several biofilm experimental models exist, they often have limited replications of spatiotemporal dynamics surrounding biofilms. For a better understanding dynamic and complex biofilm development, this manuscript presents a customizable platform compatible with off-the-shelf well plates that can monitor microbial adhesion, growth, and associated parameters under various relevant scenarios by taking advantage of 3D printing. The system i) holds any substrate in a stable, vertical position, ii) subjects samples to flow at different angles, iii) switches between static and dynamic modes during an experiment, and iv) allows multiplexing and real-time monitoring of biofilm parameters. Simulated fluid dynamics is employed to estimate flow patterns around discs and shear stresses at disc surfaces. A 3D printed peristaltic pump and a customized pH measurement system for real-time tracking of spent biofilm culture media are equipped with a graphical user interface that grants control over all experimental parameters. The system is tested under static and dynamic conditions with Streptococcus mutans using different carbon sources. By monitoring the effluent pH and characterizing biochemical, microbiological, and morphological properties of cultured biofilms, distinct properties are demonstrated. This novel platform liberates designing experimental strategies for investigations of biofilms under various conditions.

1. Introduction

Conventional manufacturing utilizes formative and subtractive methodologies to realize a final product after several steps that often require costly infrastructure.^[1] Over the past few decades, 3D printing (3DP) has fundamentally revolutionized the materials, processes, and skills required to generate geometrically complex objects in a timely and cost-effective manner.^[2] For example, the advancement of 3DP techniques has enabled inexpensive and facile prototyping of microfluidic molds^[3] and microfluidic channels with adequate rigidity and reusability.^[4] With the development and widespread use of Computer Aided Design software to create 3D objects, the barrier to entry for new users to create and modify designs has been lowered. This lowered barrier to entry coupled with the development of a plethora of biocompatible materials^[5] has resulted in 3DP becoming particularly useful for the biomedical research community in areas such as implant design and tissue engineering.^[6,7] Consequently, 3DP is now regularly used to design custom experimental setups (bespoke devices, containment, and delivery systems^[8]) as well as experimental samples (composite biomaterials and bioinspired structures^[9]), making it an essential tool for a wide range of biomedical investigations.

In biomedical research, bacterial adhesion and subsequent biofilm formation processes have been studied in depth due to their significant implications on human health. Bacteria tend to colonize on a surface and form biofilms – organized aggregates or communities of microorganisms embedded in a self-produced matrix of extracellular polymeric substances (EPS).^[10,11] The intercellular interactions and properties of the matrix grant biofilms emergent properties, distinct from those of free-floating cells,^[12] which can lead to increased rates of gene exchange and antimicrobial resistance.^[13] These biofilms are often associated with persistent infections^[14],[15] and contamination of medical devices and implants.^[16] In particular, the development of biofilms is deeply associated with the pathogenesis of a variety of oral health conditions such as dental caries and periodontal diseases.^[17] Since biofilm formation involves complex physical, biological, and chemical processes, whose contributions are continuously altered throughout biofilm development and are also subject to prevailing environmental and hydrodynamic conditions,^[18] investigations into the formation of biofilms using various systems have been reported.

While several static models to study biofilms exist, they lack control over the spatiotemporal dynamics around the biofilms. Also, accumulated planktonic bacterial population and metabolic products over time in the system could affect biofilm development. Thus, a few semidynamic and dynamic models for biofilm studies have been developed.^[19–26] However, some of these models do not allow real-time monitoring of biofilm parameters from an individual sample or are not amenable with multiplexed, high-throughput experimental designs under various conditions at the same time. Also, they often lack freedom in testing biofilm formation under various fluid dynamic conditions or have difficulty differentiating active bacterial binding and sedimentation by gravity due to the horizontal orientation of biofilm specimens. For biofilm models to be relevant, they should incorporate experimental variables such as nutrient availability via flow, shear forces, and conditioning films on the substratum.^[27] Thus, it is crucial to develop a versatile biofilm experimental platform in accordance with various relevant scenarios to better understand biofilm formation and development processes.

Here, we present the study employing 3DP to develop a culture-mode switchable (static-to-dynamic or vice versa) in vitro biofilm model compatible with an off-the-shelf well plate setup. 3D printed parts were used to construct a new clamping system to hold any material in a stable, vertical position, allowing differentiation of active bacterial attachment versus settlement by gravity. These parts were integrated with tubing and a low-cost 3D printed peristaltic pump^[28] and were subjected to hydrodynamic simulations to allow nutrient media flow around the biofilm substrate surface at different angles. By combining programmable software and 3DP hardware, our system allows multiplexing and real-time monitoring of biofilm parameters. In this study, we configured our system for real-time monitoring of the effluent pH, providing insight into the acidogenic potential for biofilms over time. Additionally, we developed a simple graphical user interface (GUI) for communication with the pump and pH meter to allow users control over all experimental parameters during a biofilm investigation. As a proof-of-concept, we utilized our system for oral biofilm studies with the model cariogenic pathogen, Streptococcus mutans. By comparing the biofilm properties between different sugar sources under static and dynamic modes, we demonstrated the potential of our platform to serve as a realistic model for studying oral biofilms and the impact it can have in clinically relevant biofilm-related investigations of oral pathogens on teeth-like surfaces. Overall, we have developed a novel, customizable, and versatile platform for biofilm studies.

2. Results

2.1. Platform Design and Assembly

The platform developed in this study was designed to be compatible with instruments and apparatus generally used in laboratories regardless of manufacturer or model, specifically off-the-shelf 24-well plates and 12 mm diameter pH electrodes. The platform was an assembly of the following parts: 2-hole, 4-hole, or 6-hole trays to hold clamped discs (up to 12 samples) in 24-well plates, clamps to hold any biofilm substrate materials (such as hydroxyapatite (HA) discs), individual fluidic channel caps with an inlet and outlet for flow maintenance (for dynamic studies), covers without fluidic channels (for static studies), a downstream flow distributor, and a reservoir for effluent pH measurements. These parts and their assembly are represented in Figure 1. As depicted, multihole trays were designed to fit into 24-well plates to allow for parallel, multiplexed experiments (Figure 1A). Each tray had 8 symmetric engraved patterns for each hole to hold the disc clamp and fluidic channel cap (Figure 1B). The symmetrical structure of the multihole trays provided rotational freedom for attaching parts. Each fluidic cap could cover one clamped disc (Figure 1C) and fit into one hole of any multihole tray. The trays had banana plug-shaped holders for a snug fit into the interwell voids of a well plate (Figure 1D). Each disc clamp was intentionally designed to be smaller than the diameter of the HA disc (6 mm wide clamp for a 12 mm disc) to allow the elastic tension of the clamp arms to hold the discs stably in a vertical position (Figure 1E). Detailed schematics of the clamping system and 4-hole trays are depicted in Figures S1 and S2 (Supporting Information). To monitor pH changes in real-time, the effluent pH reservoir was designed to be compatible with a typical 12 mm diameter pH electrode (Figure 1F). A compressible O-ring polymer gasket (McMaster-Carr, Dash Number 206) between the reservoir and the pH electrode provided a secure, leak-proof seal, and a second O-ring in the screw cap supported the pH electrode in an upright position. The outlet for the reservoir was elevated with respect to the inlet to ensure that the bulb of the pH electrode was always completely submerged when flow through the reservoir was maintained. Actual printed parts placed on a 24-well plate are shown in Figure 1G. The 3D printed fluidic channel cap and disc clamp are shown in Figure 1H,I, respectively. Figure 1J,K shows the orientation of 12 and 7 mm discs held by the printed clamps. Parts for the pH electrode reservoir are shown in Figure 1L.

Figure 1.

Platform for performing dynamic biofilm investigations. Custom 3D printed parts as used in the experimental setup with a 24-well plate (clear), 4-hole trays (blue), 6-hole trays (purple), tubing (gray), fluidic channel caps (yellow), disc clamps (brown), and discs (white). A) Integrated assembly for the dynamic experiments. B) Assembly of clamping system with 4-hole tray and tubing. C) Assembled clamping system with inlet and outlet channels. D) 4-hole tray with banana plug-shaped holders for a sturdy fit with interwell voids in the plates. E) Compressing grip of the clamps to hold the discs vertically. F) Custom pH electrode reservoir with an elevated outlet for stable, continuous measurement of effluent pH. G) Assembled plate ready for experimental use. H) 3D printed fluidic channel cap. I) 3D printed clamp. J) 3D printed clamp with 12 mm disc. K) 3D printed clamp with 7 mm disc. L) 3D printed pH electrode reservoir parts.

Once the platform was designed and printed, it was ready to be integrated with tubing and a peristaltic pump for adaptive biofilm experiments. The general schematic for such biofilm experiments and actual experimental setups are shown in Figure 2. Briefly, nutrient media reservoirs with different carbon sources were placed upstream of a peristaltic pump system. Our integrated platform with 24-well plates (Figure S3, Supporting Information) was then connected downstream to the pump to allow media to flow through multiple wells in a high-throughput, multiplexed manner. To keep our system low-cost and easy to fabricate with flow rates amenable with millifluidic experiments, we printed a peristaltic pump specifically designed for such purposes.^[28] Following this, apparatus for real-time monitoring (in terms of effluent pH measurements as a proof-of-principle) was connected followed by a final collection site for analysis or waste collection (Figure 2A). Figure 2B shows an example of the setup for parallel experiments that can test 8 different conditions at a time (4 static and 4 dynamic modes). Our platform can also be used to test multiple replicates with different conditions at the same time by using a universal cap system that allows uniform flow in each channel (Figure 2C–E). For example, 4 replicates with 4 different conditions can be tested with one peristaltic pump using this universal cap system (Figure 2F). Finally, depending on the experimental need for sample collection and analysis, our system is compatible with a downstream flow distributor (Figure 2G) to segregate fluid for real-time monitoring and waste collection. Figure 2H shows a pH probe with tubing in the reservoir for downstream in-line measurements.

Figure 2.

General schematic of dynamic biofilm experiments and components for assembly. A) Schematic depicting dynamic experiments with multiple inlet sources pumping nutrient media into multiple wells that are then subjected to real-time effluent pH monitoring and collection for analysis or waste. B) Simultaneous runs in 8 different conditions (4 static and 4 dynamic modes). C) Schematic depicting the universal cap for testing multiple replicates under different conditions. D) Estimated uniform flow in each channel by computational fluid dynamics. E) 3D printed universal ca F) Simultaneous run with 4 replicates for 4 different conditions. G) 3D printed downstream flow distributor. H) pH probe with tubing in the customized reservoir for downstream in-line measurements.

2.2. Computational Fluid Dynamics around Discs in Biofilm Culturing Wells

An essential component of developing a dynamic biofilm model involves controlling the spatiotemporal dynamics around the surface on which the biofilm thrives. In our platform, we simulated the flow around discs by varying the input flow rate and angle between the disc and inlet/outlet. In particular, in this study, we intended to test our platform to mimic cariogenic biofilm formation in the human mouth. It is known that unstimulated salivary flow rates are in the range of 300–400 μL min⁻¹ and can fall at or under 100 μL min⁻¹ during sleep.^[29] Thus, we tested 3 input flow rates for nutrient media – 50, 100, and 300 μL min⁻¹. The angle between the inlet/outlet and the discs is another customizable parameter within our system. To encompass different flow patterns, we simulated input flow at 0°, 45°, and 90° to the disc surface. Streamlines around the disc and surface plots (along and across the disc) are shown in Figure 3A–C and these were used to identify an optimal angle between the discs and inlet/outlet that may maximize nutrient availability via adequate mixing without disrupting the growth of biofilms on the discs. As shown, the flow pattern created clearly contributed to the mixing of fluid around the discs (both along and across the disc surface). Additional flow rates and orientations are depicted in Figures S4–S11 (Supporting Information). Based on the simulations, the most optimal flow around the disc occurred when the flow was at 45° to the disc as opposed to 0° and 90°. Furthermore, we mapped shear stress at the disc surface at a flow rate of 100 μL min⁻¹ at 45° (Figure 3D) and the average shear stress on the disc was estimated to be 1.04 × 10⁻⁵ N m⁻². Shear stresses in other conditions ranged from 4.40 × 10⁻⁶ to 3.94 × 10⁻⁵ N m⁻². Mapped shear stresses and average values in each condition are depicted in Figure S12 (Supporting Information). In this study, we chose a flow rate of 100 μL min⁻¹ at 45° to the disc for all biofilm experiments.

Figure 3.

Simulations for flow around discs within the experimental setups. Computational Fluid Dynamics simulations for 100 μL min⁻¹ flow of media at 45° to disc. A) Streamline volume plots depicting fluid dynamics around the disc as XYZ, XY, YZ, and ZX views B) ZX surface velocity across the disc surface. C) YZ surface velocity plot around the disc. D) Shear stress plot across the disc surface.

2.3. Graphical User Interface

For ease of use, we developed an integrated GUI to control experimental parameters on our platform by communicating with the peristaltic pump and pH meter. The GUI was equipped with options to vary the measurement interval of the pH meter and the flow rate of input media by changing the speed of the peristaltic pump (Figure 4). We provided these options in the GUI to ensure researchers have greater freedom in experimental design for biofilm experiments – such as switching between static and dynamic modes, continuous versus pulsed flow of media, and real-time monitoring of environmental parameters at varying intervals throughout the experimental period as needed. Python files for these communication protocols are provided on GitHub (https://github.com/Ravi-ramjee/Well-Plate-Platform). It is important to acknowledge that different pH meters and peristaltic pumps may have different communication protocols with a computer. In this case, changes to our programs will be required to construct a similar interface with different components. For pH meters and stepper motors that use similar communication protocols, our platform should be usable as is.

Figure 4.

Schematic of GUI and communication with pump and pH meter.

2.4. Validation of Platform for Static and Dynamic Biofilm Experiments

Maintaining consistent vertical orientation of discs for biofilm experiments may reduce variation in data. Thus, prior to the main biofilm experiment, we compared our new clamping system to traditional wire-holder models^[30–32] to confirm the improved control over the stable placement of biofilm substrates. By performing static biofilm experiments, we observed that our new clamping system significantly reduced variation in all metrics used to harvest biofilms (end-point pH, dry weight, and CFU mL⁻¹; Figure S13, Supporting Information).

Then, to validate the versatility of our platform for biofilm experiments, we developed an experimental strategy of testing biofilm formation under different modes of operation (static or dynamic) and/or different types of nutrients in the context of dental caries. We utilized flow rates that are relevant to salivary flow rates, a substrate representing human enamel (i.e., vertically positioned HA disc), and dietary sugars (i.e., sucrose, glucose, and fructose) to mimic cariogenic biofilm formation in the human mouth. The rationale behind choosing these carbon sources for S. mutans was as follows: It is well known that S. mutans can produce copious amounts of insoluble EPS in the presence of sucrose.^[33] However, glucose and fructose metabolisms do not result in massive productions of insoluble EPS. Such distinct metabolism combined with biofilm culture environments could result in dramatically different biofilm properties. Thus, we performed biofilm experiments representing eight different conditions using our platform (i.e., 4 static and 4 dynamic conditions); 1) basic medium with no added sugar (No sugar), 2) basic medium + 0.5% glucose + 0.5% fructose (Glc+Frc), 3) basic medium + 1.0% sucrose (Suc), 4) basic medium + 1% sucrose + 0.5% glucose + 0.5% fructose (Suc+Glc+Frc)) and compared their properties.

Biofilm experiments with these carbon sources on a static model were performed using our platform with no flow and covers without fluidic channels. Biofilms cultured for 18 h were harvested to measure end-point pH of the well volume, dry-weight, and CFU of the biofilm and planktonic population (Figure 5A–D). S. mutans can metabolize a wide range of carbohydrates into organic acids.^[33] Thus, the end-point pH of the well volume was similar for sugar-supplemented groups (Glc+Frc, Suc, Suc+Glc+Frc; Figure 5A). In particular, pH values (4.2–4.4) for these groups were sufficiently acidic to produce an environment conducive to tooth demineralization (under pH 5.5).^[34] In contrast, the No sugar condition did not show a significant pH drop, resulting in neutral pH. The dry weight obtained was highest in Suc followed by Suc+Glc+Frc, while there were no significant differences between Glc+Frc and No sugar conditions (Figure 5B). Populations in biofilms also showed a similar trend to the one for dry weight; Suc and Suc+Glc+Frc showed ≈78-fold and ≈15-fold higher number of biofilm CFU compared to the No sugar condition (Figure 5C). However, this trend was reversed in the CFU counts for the planktonic phase; the No sugar condition was ≈8.5-fold and ≈3.5-fold higher than Suc and Suc+Glc+Frc conditions, respectively (Figure 5D). There were no significant differences between No sugar and Glc+Frc conditions both in biofilm and planktonic CFUs. To further understand the differences in biofilm properties between samples from different carbon sources, we investigated the tertiary structures of the biofilms using confocal microscopy (Figure 5E–H). As expected, the No sugar and Glc+Frc conditions had drastically smaller bacterial chains/clusters with negligible EPS (Figure 5E,F), while the Suc and Suc+Glc+Frc conditions had large bacterial clusters and plentiful EPS (Figure 5G,H).

Figure 5.

Validation of platform for static and dynamic biofilm experiments. Biofilm parameters after 18 h experiments with various sugar contents (basic medium with no added sugar (No sugar), basic medium + 0.5% glucose + 0.5% fructose (Glc+Frc), basic medium + 1.0% sucrose (Suc), and basic medium + 1% sucrose + 0.5% glucose + 0.5% fructose (Suc+Glc+Frc)) depicting A) End-point pH values B) Dry-weight C) Biofilm CFU, and D) Planktonic CFU for static experiments. E) Representative XY confocal images for static biofilms grown in E) No sugar, F) Glc+Frc, G) Suc, and H) Suc+Glc+Frc. Scale bars: 50 μm. Corresponding panels on the right side of the figure represent dynamic biofilm experiments I–P). Statistics: t-tests with * representing p < 0.05 for comparisons against no sugar control (n = 3).

For testing dynamic conditions, we conducted the same experiments by flowing 100 μL min⁻¹ of media into the wells for 18 h. All analyses for the biofilms were the same as in the static experiments apart from the addition of real-time monitoring of the effluent pH over 18 h. Since media with pH ≈7 was flowing into the wells throughout the experiment, it was expected that the end-point pH of the well volume would be affected. In the Suc condition, the bacteria were able to maintain a pH of 4.8 within the well volume despite fresh pH 7 media entering the well continuously for 18 h (Figure 5I). However, in all other conditions even including the Suc+Glc+Frc condition, pH values remained ≥ 6 (Figure 5I). Due to the continuous supply of nutrients, the dry weight of all biofilms in dynamic conditions was dramatically higher in comparison to the static model (Figure 5B,J); in particular, the dry weight of Suc was explosively increased (≈27-fold) compared with the No sugar condition, while Suc+Glc+Frc and Glc+Frc conditions were increased ≈8-fold and ≈6-fold, respectively. The broad trends for biofilm CFU (Figure 5K) were similar to those for the static model; biofilm CFU was ≈217-fold higher in the Suc condition and the one in the Suc+Glc+Frc condition was also ≈126-fold higher versus the No sugar condition, while biofilm CFU of Glc+Frc condition was only slightly higher (≈2-fold). In contrast, the planktonic CFU of Suc and Suc+Glc+Frc conditions were ≈11-fold and ≈8.5-fold lower than the No sugar condition, whereas the planktonic CFU for Glc+Frc condition was ≈20-fold higher (Figure 5L). These results were further corroborated by confocal imaging that depicted sparsely colonized EPS-free bacterial cells in the No sugar and Glc+Frc conditions (Figure 5M,N). In marked contrast, the Suc condition revealed large, thick, and EPS-rich biofilms across the surface (Figure 5O). The Suc+Glc+Frc condition also showed a large clump but the amount of EPS was limited and the vast majority of areas lacked biofilms (Figure 5P).

To accurately reflect the spatiotemporal dynamics around biofilms, it is imperative for a model to be amenable to real-time monitoring of biofilm parameters. As a proof-of-concept, we added a real-time pH tracking modality to our platform. For this, we 3D printed a custom effluent pH reservoir connected to a pH meter and collected pH data across the length of the experiment. Continuous pH profiles for all conditions are depicted in Figure 6. Interestingly, the pH profiles were dependent on the type of sugar rather than the number of bacteria in the system. Since the well volumes were continuously replaced with fresh media with pH 7 during the experiment, the pH of the well volumes could only fall if there was accumulation of produced acids. In our dynamic experiments, the entire well volume was replaced 36 times across the length of an 18 h experiment (at 100 μL min⁻¹). In the Suc condition, a drastic pH drop after ≈13 h of incubation was observed (red in Figure 6). Similarly, the Suc+Glc+Frc condition (green in Figure 6) also showed the dropping of pH over time but the magnitude was not as high as the Suc condition. The Glc+Frc condition showed a moderate reduction of pH (blue in Figure 6), while the No sugar condition was almost flat without significant reduction (black in Figure 6). Overall, we demonstrated the versatility of our platform in allowing several experimental modes and continuous monitoring of biofilm parameters.

Figure 6.

Continuous measurement of pH with dynamic biofilm experiments. Real-time pH of the effluent from dynamic biofilm experiments was continuously monitored for 18 h in 4 different conditions – no sugar (black), a combination of glucose + fructose (blue), sucrose (red), and a combination of sucrose + glucose + fructose (green) (n = 3).

3. Discussion

The advent of 3DP has fundamentally altered the tools available to biomedical researchers. In contrast to conventional manufacturing techniques that require substantial infrastructure, 3DP offers far simpler processes, far shorter timelines, and a much lower barrier to entry. Expectedly, this has led to extensive research in the development of printable materials that are biocompatible and hence functional in biological investigations.^[5]

Among diverse microbiological research fields, studying bacterial surface sensing and subsequent biofilm formation has been spotlighted because of its significant implication on industrial and medical settings.^[35] In a wide variety of environments, microorganisms tend to form complex biofilms adhered to abiotic and biotic surfaces. Such biofilms can sometimes be beneficial for the environment and some industries. In fact, biofilms have been utilized for microbial bioprocessing,^[36,37] microbial fuel cells,^[38,39] bioremediation,^[40,41] biosensor design,^[42,43] and biohydrogen production.^[44,45] However, a tendency of bacterial accumulation on surfaces is in charge of a diverse spectrum of local/systemic diseases and industrial bio-fouling.^[46] Since established mature biofilms are extremely difficult to eradicate,^[47,48] it is imperative to intervene in the biofilm formation process at an early stage. Despite the fact that biofilm properties can be dramatically altered depending on the culture conditions, most biofilm models (particularly medical biofilms) involve static studies of biofilms in vitro. Although these studies add value to the basic body of knowledge on microorganisms, they do not reflect the complex spatiotemporal dynamics around biofilms nor are they flexible to accommodate various experimental conditions. Thus, a new model is warranted for a better understanding of the dynamic factors affecting biofilm development which may lead to more targeted and effective biofilm disruption strategies.

Recently, there have been reports utilizing microfluidic approaches to develop high-throughput systems to study biofilms.^[49–51] While these allow fine control of fluid flow and are compatible with a plate reader or spectrophotometer, there is no freedom in selecting biofilm substrate. Typically, microfluidic approaches utilize coated polydimethylsiloxane, glass, or plastics as substrates for bacterial attachment. However, bacterial binding and biofilm properties could be significantly varied depending on the characteristics of a substrate surface.^[35] Furthermore, though the trend in microfluidics is toward more complex lab-on-a-chip tools,^[52] to date the approaches are still heavily rooted in microscope imaging. The sole use of imaging methods without traditional biochemical/microbiological assays (such as microbial counting and total biomass) may not provide precise data. Thus, enabling both the imaging of biofilm structure and analysis of biochemical/microbiological data as well as downstream monitoring of metabolic and environmental parameters on any biofilm substrates would be more useful to study biofilms. While this has been partially addressed previously in a microfluidic setup by determining released bacterial populations from biofilms,^[53] this was not completely applicable to holistic biofilm studies. Finally, microfluidic approaches are inherently prone to starvation without active flow due to rapid absorption or evaporation of the medium. Allowing switching between static and dynamic modes would provide complete freedom in designing experimental strategies across the static-dynamic spectrum.

Inspired by this need, we developed a platform to study the growth of biofilms in various scenarios. Here, we have utilized 3DP to tackle fundamental problems in biofilm research – the lack of key features such as multiplexing capability, switching between experimental modes, consideration of hydrodynamics, and real-time monitoring of biofilm parameters. Our system allows precise evaluation of governing parameters of biofilm formation by using a clamping system capable of holding any material in a stable, vertical position, subjecting these materials to nutrient flow at various angles and flow rates, and switching between static and dynamic modes mid-experiment. With the aid of a peristaltic pump with multiple channels and trays with multiple holes, we developed a high-throughput biofilm testing environment. While innovative, it is of the utmost importance to achieve data reproducibility. Traditionally, wire-holders have been extensively used to conduct biofilm experiments for several years,^[30–32] yet they often lead to an increased variation in biofilm data due to the variability in each hand-folded wire and the orientation of substrates. Our new clamping system completely removes this variation by eliminating the need for manual handling and adjustment of clamps and substrates while running biofilm experiments with multiple samples simultaneously. Indeed, we demonstrated high reproducibility of data generated by our platform, showing much smaller deviations in all measured biofilm properties (pH, CFU, and dry-weight) compared with biofilm properties cultured using a traditional wire-holder. These results were strong indications of our platform’s capability to generate reproducible, precise data in biofilm experiments with multiple samples.

Due to 3DP and our platform’s compatibility with basic laboratory supplies, any research group with access to a 3D printer can easily utilize and/or modify our platform to suit their experimental needs. For example, in this study, we designed and used clamps to fit 12 mm discs that are widely used for biofilm study. As shown in Figure 1K, however, it can be easily transformed into tighter clamps to grip 7 mm discs. In addition, the position and length of the inlet and outlet on the fluidic channel caps can be adjusted to modify the flow pattern around discs. Finally, 3DP parts employed in our platform are extremely cost-effective (<$0.04/clamp in material cost) and highly reusable (clamps lasted up to 3 autoclave cycles before becoming loose; all other parts were tested to last for more than 10 autoclave cycles). Together, our platform provides a cost-effective, realistic model of biofilm environments while offering researchers the ability to customize the design of the system to allow applicability over a broad range of potential studies.

With this platform, we observed interesting bacterial behavior by running 18 h static and dynamic biofilm experiments with S. mutans under various sugar nutrient conditions. Sucrose-rich condition (Suc) facilitated bacterial adhesion that preferentially remained attached to the disc surface rather than in the planktonic phase, which led to large bacterial clusters with abundant sticky EPS matrix (Figure 5). This can result in hypoxic regions within biofilms due to the limited diffusion of oxygen.^[54,55] S. mutans is known to be facultatively anaerobic and can switch to anaerobic fermentation in hypoxic conditions that can lead to increased acid production. Furthermore, produced acids could be retained within EPS-rich biofilm, causing a sustained acidic environment. This may explain how the pH profile for the Suc condition (red in Figure 6) dropped rapidly around the 13 h mark and maintained acidic pH until the end of the experiment. Unlike the Suc condition, the pH profile for other conditions failed to reduce pH below the critical pH (5.5) due to the lack of a thick EPS-matrix, resulting in relatively oxygen abundant conditions. In this case, S. mutans could remain in aerobic respiration and any pH drop would be continuously adjusted by the addition of fresh media of pH 7 Interestingly, however, the Suc+Glc+Frc condition was not able to form mature biofilm nor cause an acidic environment despite double the amount of total sugar content (vs Suc condition). This could be attributed to the preferential utilization of glucose over sucrose.^[56,57] Since there were plenty of preferable energy sources (i.e., glucose and fructose) in the Suc+Glc+Frc condition, those were preferably taken by S. mutans rather than sucrose, resulting in less production of EPS-matrix and mature biofilms compared with the Suc condition.

While we compared biofilm properties under various sugar source conditions as a proof-of-concept, there are several additional experimental designs that will work well with our system. For instance, our platform is designed to test different hydrodynamic conditions that are caused by the angle of biofilm substrate and/or flow rates. Given the tested flow rates and angles, we were able to see up to 9-fold differences of shear stress at the disc surface (Figure S12, Supporting Information). Since hydrodynamic conditions can affect biofilm architecture, composition, and mechanical strength by interfering or enhancing bacterial sensing on various surface properties,^[35] it would be interesting to investigate the influence of hydrodynamic conditions on bacterial adhesion and biofilm formation that has been often neglected. In addition, our platform is suitable to test the effect of complex nutrient changes (e.g., feast and famine conditions) or flow rate changes on biofilm formation and development. Our platform is also independent of the disc material. This makes it useful for rapid testing of biofilm properties on several different materials in the same experimental setting. Furthermore, our platform can be utilized for a more practical approach to testing antimicrobial efficacy – 1) treatment time can be adjusted easily and 2) multiple samples can be treated with the same or different antimicrobials at a time. It can be also used to test the efficacy of a drug delivery system with antimicrobials subject to continuous flow under dynamic conditions. Finally, our platform can mimic specific treatment conditions such as mouth rinsing motion by alternating between forward and backward flow. While currently designed for a 24-well plate setup, there are no physical restrictions to prevent scaling down to 48-well or 96-well plate compatible versions. This may further increase the throughput of the system and significantly reduce the use of a relatively large volume of biofilm culture medium compared with a microfluidic setting.

4. Conclusion

In summary, our platform represents the use of 3DP to enhance the usability of traditional in vitro setups for biofilm investigations. Our experimental platform represents a simple, yet powerful method to run biofilm experiments in dynamic mode with parallel testing of various parameters such as carbon sources, flow rates and patterns. Additionally, the introduction of a user-friendly GUI provides significant freedom to researchers to develop and test novel experimental strategies. With a proof-of-concept biofilm study using model oral biofilm forming organism S. mutans, we demonstrated the versatility of our platform that can capture various biofilm properties depending on varying experimental parameters (carbon source and flow rate). To demonstrate the amenability of our system with continuous monitoring, we tracked real-time effluent pH data from biofilms, revealing distinct pH profiles depending on sugar contents. In principle, in-line continuous monitoring can be expanded for other parameters such as oxygen sensing and secondary metabolite detection by simply adding those probes. We are currently designing experimental protocols to use our platform in these use-cases. Overall, we present a novel, customizable platform that will significantly impact biofilm investigations under various conditions.

5. Experimental Section

Platform Design:

The platform was designed using Onshape (Boston, MA) as an assembly of the following parts – 2-hole, 4-hole, or 6-hole trays to hold clamped discs in 24-well plates, clamps to hold biofilm substrate materials, individual fluidic channel caps with an inlet and outlet for flow maintenance (for dynamic studies), covers without fluidic channels (for static studies), a downstream flow distributor to collect spent media for analysis or as waste, and a reservoir for effluent pH measurements. The parts were designed to be compatible with the dimensions of a standard 24-well plate and a 12 mm diameter pH electrode.

Platform Construction:

All parts were 3D printed as STL files listed in GitHub (https://github.com/Ravi-ramjee/Well-Plate-Platform) with a Form 3 (Formlabs, Somerville, MA) low force stereolithographic printer at a layer thickness of 50 μm. For use in biological experiments, the Dental Surgical Guide Resin (Formlabs) was used, which is certified as both biocompatible and autoclavable.^[58] Printed parts were washed with 99% isopropyl alcohol (CleanPro IPA, Production Automation Corporation, MN) for 20 min followed by postcuring under UV (405 nm) in a Form Cure (Formlabs) for 60 min at 60 °C. After postcuring, the support structures were removed from the parts. Prior to experimental use, all parts were autoclaved at 121 °C and 15 psi.

Peristaltic Pump Construction:

The construction of the peristaltic pump was based on “The FAST Pump,” a low-cost peristaltic pump for research applications.^[28] The assembly procedure was identical except for the use of 4 × 16 mm hex stainless-steel screws on the pump body. The pump used a NEMA 17 USB stepper motor (Arcus Technology, Inc., Livermore, CA). A customizable drive program was provided for use with manufacturer software in GitHub (https://github.com/Ravi-ramjee/Well-Plate-Platform). Depending on the inner diameter of the tubing used, the flow rate range of the pump was 0.7–5750 μL min⁻¹.

Software Communication Protocol:

Python was used to construct a Windows GUI to adjust operating parameters on the pH meter and peristaltic pump. The GUI was equipped with options to vary the measurement interval of the pH meter and the flow rate of media by changing the speed of the peristaltic pump. Python files for the interface are provided in GitHub (https://github.com/Ravi-ramjee/Well-Plate-Platform). The pH meter used in the system (Mettler-Toledo SevenCompact Benchtop pH Meter; Mettler-Toledo, Columbus, OH) had an RS232 interface for printers by which a serial connection was established with a laptop. The Python library “pyserial” (https://github.com/pyserial/) was used to read data from the meter at an interval specified by the program. The storage destination from the meter was changed from local memory (default) to a printer. For collection and further analysis, data were written to a spreadsheet at the end of the experiment. The peristaltic pump motor was provided with manufacturer software to program basic motor movement but lacked the ability to change the speed and direction of the motor in an intuitive and quick manner. The Python library “pylablib” (https://github.com/AlexShkarin/pyLabLib/) was used to communicate over standard USB protocol with manufacturer provided dynamic link libraries (DLLs) to change the operating status of the pump motor using the GUI. The direction and speed of the motor were adjusted using text entries in the interface.

For combability with a 3DP flow distribution system, a proof-of-concept GUI was constructed by extending the traditional GUI functionality. Instances of a class that contained the attributes to maintain the status of distribution valves were created. In the GUI itself, each valve could be opened and closed by clicking on its respective buttons. This allows the user to segregate fluid from individual channels for real-time monitoring of a variety of parameters and waste collection. Based on the communication protocol of a flow distribution system, the code can be modified to program the opening and closing of each valve. The programs for this GUI are provided in GitHub (https://github.com/Ravi-ramjee/Well-Plate-Platform).

Fluid Flow Simulations:

The computational fluid dynamics (CFD) module of COMSOL Multiphysics 5.2 was used for stationary single-phase flow studies following the Navier-Stokes equation. Discs were subjected to 50, 100, and 300 μL min⁻¹ of flow with 0°, 45°, and 90° between the inlet and disc. Surface velocity plots were developed and visualized at varying planes and orientations using the following equations. Additionally, streamline plots were developed to visualize flow patterns around the discs at a given angle and flow rate

$$\rho (u\cdot \nabla )u=\nabla \cdot \left[-pI+\mu \left(\nabla u+{(\nabla u)}^T\right)\right]+F+\rho g$$ (1)

$$\rho \nabla \cdot (u)=0$$ (2)

where ρ is the density of fluid, u is the velocity vector, p is the pressure, I is the identity tensor, μ is the dynamic viscosity, the superscript T stands for the transpose, F is the volume force, and g is the gravity vector. The shear stress values on the disc surface for each condition were then estimated by multiplying the shear rate and the dynamic viscosity.

Strain, Culture Conditions, and Saliva Collection:

S. mutans UA159, a cariogenic dental pathogen and well-characterized EPS producer, was used for all static and dynamic biofilm experiments. Microbial stocks were stored at −80 °C in tryptic soy broth containing 50% glycerol. S. mutans was grown to mid-exponential phase (optical densities at 600 nm (OD₆₀₀) of 1.0 in ultrafiltered (10 kDa molecular mass cutoff; Millipore, Billerica, MA) yeast-tryptone extract broth containing 2.5% tryptone and 1.5% yeast extract (UFTYE; pH 7.0) with 1% (wt/vol) glucose at 37 °C and 5% CO₂, as described previously.^[31,47] Cells were harvested by centrifugation (6000 rcf, 10 min, 4 °C).

The protocol for saliva collection was reviewed and approved by the Institutional Review Board of the University of Pennsylvania (protocol number 818549) and written consent was obtained from the volunteers in this study. Saliva donation and storage were as described previously.^[31] Briefly, donors used unflavored paraffin wax to stimulate saliva production. Donated saliva was centrifuged (5500 rcf, 10 min, 4 °C) and then filter sterilized (0.22 μm; S2GPU01RE ultralow-binding protein filter; Millipore, Billerica, MA) before storage at 4 °C until use.

Static and Dynamic In Vitro Biofilm Model:

Biofilms were cultured using an established saliva-coated hydroxyapatite (sHA) model.^[32] Filtered saliva as described earlier was used to coat the HA discs (surface area of 2.7 ± 0.2 cm²; Clarkson Chromatography Products, Inc., South Williamsport, PA). First, to highlight the reduced variation in experimental results with the clamping system (designed and printed as previously described), 18 h static biofilm experiments were run with 2 × 10⁶ CFU of S. mutans mL⁻¹ in UFTYE media with 1% sucrose and compared results to a traditional wire-holder model. Biofilms were harvested to compare end-point pH, dry weight, and CFU mL⁻¹.

Next, static and dynamic experiments were conducted with the new clamping system. HA discs were suspended in a standard 24-well plate with the clamping system. Each well was inoculated with ≈2 × 10⁶ CFU of S. mutans mL⁻¹ in four types of media (UFTYE alone, UFTYE + 0.5% glucose + 0.5% fructose, UFTYE + 1.0% sucrose, UFTYE + 1% sucrose + 0.5% glucose + 0.5% fructose (wt vol⁻¹)). The biofilm was harvested at 18 h for analysis and imaging.

The initial setup for the dynamic model was identical to the static model. The HA discs were suspended vertically with the 3DP clamping system in a standard 24-well plate. As illustrated in Figure 2, silicone tubing (Masterflex L/S Precision Pump Tubing size 13, Cole-Parmer, Vernon Hills, IL) was attached to feed four individual media (UFTYE alone, UFTYE + 0.5% glucose + 0.5% fructose, UFTYE + 1.0% sucrose, UFTYE + 1% sucrose + 0.5% glucose + 0.5% fructose) into the wells via the peristaltic pump constructed as previously described. For continuous pH monitoring, the outlet from the well was connected to the effluent pH reservoir printed as previously described before collecting the media as waste. Each well was inoculated with 2 × 10⁶ CFU of S. mutans mL⁻¹ in media enriched with 1% (wt/vol) sucrose or 0.5% (wt/vol) each of glucose and fructose. After setting up the system, flow was turned on via the peristaltic pump’s drive program at a flow rate of 100 μL min⁻¹. Once the effluent pH reservoir was full of spent culture media, the pH meter started recording the pH at 90s intervals throughout the experiment. Flow was stopped at 18 h after inoculation and the biofilm was harvested for analysis and imaging. End-point pH values (0 and 18 h) were verified with another pH meter. The pH meter used for continuous measurement was always calibrated before and after each dynamic experiment. Variation in pH due to the temperature difference in the incubator (37 °C) and outside where the pH probe was placed (25 °C) was minimal, around the order of pH 0.1 in the case of pH standard buffers.^[59]

Microbiological and Confocal Microscopy Analysis of Biofilms:

Biofilms collected from static and dynamic experiments were subjected to standard microbiological analysis. At the end of the experimental period, biofilms were removed and homogenized via sonication, and the number of viable colony forming units (CFU mL⁻¹) was determined on blood agar plates (BD BBL Prepared Plate Media, Trypticase Soy Agar (TSA II) with Sheep Blood, Thermo Fisher Scientific, Waltham, MA).^[31,60] In parallel, an aliquot of the suspension was used to determine the dry weight of the harvested biofilm. After centrifugation (5500 rcf, 10 min, 4 °C), the pellet was washed twice with Milli-Q water and dried in an oven (105 °C, 2 h) and weighed. Three independent biofilm experiments were performed for each condition (static and dynamic) with four types of media.

In separate experiments, biofilms from each condition were visually examined using confocal laser scanning microscopy (CLSM) as described previously.^[31,47] Briefly, S. mutans were stained with 2.5 × 10⁻⁶ m SYTO 9 green-fluorescent nucleic acid stain (485/498 nm; Molecular Probes, Inc, Eugene, OR), while EPS were labeled with 1 × 10⁻⁶ m Alexa Fluor 647-dextran conjugate (647/668 nm; Molecular Probes, Inc, Eugene, OR). Using an upright single-photon confocal microscope (LSM800, Zeiss, Jena, Germany), the 18 h S. mutans biofilms formed on HA discs were imaged.

Statistical Analysis:

All experiments were conducted at least three times with all data represented as mean ± SD. GraphPad Prism was used for all statistical analyses. Significant differences in data were assessed using unpaired t-tests with a significance level set to p < 0.05.

Data Availability Statement

The data that support the findings of this study are available in the supplementary material of this article.