To design the microfluidic system, we used AutoCAD^® as the computer-aid tool. The system consists of 20 parallel channels with a constant width of 50 µm and length of 7 mm. There are square inlet and outlet reservoirs (4 × 4 mm) located at two ends of the channels. The design was sent for transparent mask fabrication (Fine Line Imaging, 50,000 dpi). By conducting standard photolithography technology (Bui et al., 2017; Bui et al., 2018; Hobbs et al., 2021), we generated a silicon mold (University Wafer, 3,774) with SU8-2025 negative photoresist features (MicroChem). The photoresist was spin-coated on a silicon wafer at 1,800 rpm for 30 s, then soft-baked at 65°C for 3 min and 95°C for 10 min. The photoresist in hard-contact with the mask was exposed for 10 s (KLOE UV-KUB 2, Power density 52.3 mW/cm²), post-baked at 65°C for 2 min, and 95°C for 10 min. The channel features appeared after being developed for 5 min. The desired thickness of the resulting photoresist was 50 μm, defining the height of the channels and the reservoirs. These features were replicated into PDMS (Dow Corning, SYLGARD™ 182 Silicone Elastomer Kit) via soft lithography technique (Figures 1A, B, D) (Bui et al., 2017; Bui et al., 2018; Hobbs et al., 2021).

To assemble a microfluidic system for experiments, the PDMS replicates, glass slides, tubings and connectors were decontaminated with ethanol 70% and washed three times with sterile deionized (DI) water inside a biosafety hood. Then the PDMS replicate was assembled to a glass slide to form closed channels. For achieving irreversible attachment and preventing fluid leakage, prior to assembly, both PDMS and the glass slides were ozone-treated for 5 min (Novascan, PSD Series System) (Bui et al., 2016; Bui et al., 2017). Tubings and connectors were also assembled to the inlet and outlet reservoirs. For flow study, the inlet was connected to a syringe pump (Fisherbrand™, 780100) to control the fluid movement through the channels.

Flow rate through the system inlet was set in the syringe pump (Fisherbrand™, 780100) at 40 μL/h. The flow was monitored by introducing 2 µm fluorescence polymer microspheres (R0200, Thermo Scientific) as tracer particles in low concentration (0.06% v/v), dispersed in DI water (Mina et al., 2016; Pollet et al., 2019). Videos of the tracers in flow at the channel’s entrance, middle and end were recorded in three channels (1, 10, 20). All videos were taken at ×20 objective for 10 s using a fluorescent microscope (Nikon Eclipse Ts2), and saved as TIFF files for analysis. The trajectories and velocities of the microspheres within the channels were analyzed using the Trackmate plugin (Mina et al., 2016; Pollet et al., 2019) in Fiji. Briefly the files were first converted into 8-bit and scale was set in micron. Using Trackmate, a thresholding detector was applied with the intensity adjusted from 30 to 40 to define individual microspheres. Once all microspheres in all frames were detectable, we then utilized the simple LAP tracker, the tool specific for non-branching tracks, to quantify the lateral (XY) trajectories and the mean velocity of the microspheres (Tsvirkun et al., 2017; Dubay et al., 2023). The plugin compiled all successive positions, constructed the track list, and exported data to an excel file for further analysis.

The flow profile within the microfluidic system was also computationally analyzed using COMSOL Multiphysics^® software. A single channel with a square cross-section of 50 μm × 50 μm and a length of 7 mm was generated to study variations in velocity, shear stress, and drag force. The simulation was run using the Laminar Flow module and Stationary study. Water was used as the fluid material; and a constant flow rate of 2 μL/h (i.e. 5.56 × 10⁻¹³ m³/s for water) was set at the inlet boundary condition (equivalent to the flow rate of 40 μL/h imposed by the syringe pump). The fluid dynamics inside the channel is governed by the following default Navier-Stokes equations in COMSOL for incompressible flow, where ρ is the density, u is the velocity vector, p is the pressure, μ is the dynamic viscosity, T is the temperature, and F is the external force applied on the fluid. Then, a fine meshing system for the channel was used to perform the simulation. Heatmap diagrams of the fluid velocity throughout the channel were generated by the software. ρ u ⋅ ∇ u = ∇ ⋅ − p I + μ ∇ u + ∇ u T + F ρ ∇ ⋅ u = 0

For a Newtonian fluid like water, shear stress (σ_s), which represents the tangential force per unit area exerted by the flowing fluid on a surface, is directly proportional to the shear rate (γ). This shear stress profile within the channel can be calculated and graphed in COMSOL via the following default equation. σ s = μ γ

The shear stress also attributes to the drag force applied on the bacteria that reside on the channel or move within the flow. Once the drag force overcomes the adhesive force, it removes adherent bacteria from the surface. To analyze the drag forces applied on a bacterium, we added a bacterial cell model with a diameter of 0.7 μm and an end-to-end length of 2.9 μm at the mid-wall of the channel (Figure 4C). The drag force (F_drag) was estimated using the equation below, where A indicated the area of the bacteria in contact with the fluid flowing through the channel; and σ_s denoted the shear stress acting on the bacterium (Persat et al., 2015; Yuan et al., 2023). F drag = A σ s

Synthesis of CMC-DOPA was performed via amidation reaction (Wang et al., 2021; Bian et al., 2023). Sodium carboxymethyl cellulose with 0.7 carboxymethyl groups per anhydroglucose unit (CMC, Sigma-Aldrich, 419273) was dissolved in 0.1 M MES buffer (pH 5.5) at concentration of 10 mg/mL. Pure ethanol (Sigma-Aldrich, 4450-500ML) was added at a ratio of ethanol:MES buffer 1:3 (v/v). Then, 1 M equivalent of 4-(4,6-Dimethoxy-1,3,5-triazin-2-yl)-4-methylmorpholinium Chloride (DMTMM, TCI America, TCD2919-5G) was introduced to activate the carboxyl group for 15 min, followed by introduction of 0.7 M equivalent of Dopamine hydrochloride (DOPA, Sigma-Aldrich H8502-25G). The mixture was protected from light and stirred under room temperature for 1 day, then purified by dialysis against 1% NaCl solution and DI water for 2 and 3 days, respectively. The CMC-DOPA product was lyophilized using a freeze dry system (Labconco FreeZone 6) and stored at −20°C until use.

Chemical structure of CMC-DOPA was evaluated by ¹H-NMR (Wang et al., 2021). Approximately 3 mg of CMC-DOPA was dissolved in Deuterium oxide (D₂O, Sigma-Aldrich, 1.13366.0025) for hydrogen-to-deuterium exchange, followed by freeze drying then dissolving it again in D₂O. ¹H-NMR was conducted with a Bruker Ascend™ 400 MHz NMR Spectrometer and data was analyzed by MNova software. The degree of substitution (DS) of DOPA, which represents the percentage of anhydroglucose in CMC containing DOPA, was quantified by UV-Vis spectrometry at 280 nm. A standard curve of different DOPA concentrations was first generated; and absorbance of the solutions was read using UV-Star^Ⓡ 96 well plates (Greiner Bio-One 655801) and a plate reader (BioTek, Synergy LX). Then CMC-DOPA was diluted in DI water, followed by measuring the absorbance. The DS value of DOPA in CMC-DOPA was calculated based on the absorbance measurement and standard curve.

The CMC amounts on the resulting CMC-DOPA were quantified via methyl blue assay (Yan and Chai, 2021). To create a standard curve, different solutions of CMC (0–100 mg/L) were prepared in DI water and mixed with 20 mg/L methyl blue solution at 1:1 v/v ratio. Methyl blue absorbance of all mixtures at 597 nm were measured on 96 well plates (100 µL per well); then the inverse of the absorbance values was used to generate the standard curve to show the linear effect of CMC on interfering the absorbance (data not shown). Absorbance spectra of methyl blue solution alone (MB) and when mixed with CMC (50 mg/L), CMC-DOPA (50 mg/L) and DOPA (1 mg/L, equivalent to the DS value) were measured.

The coating on the PDMS or glass surface was conducted under alkaline conditions. CMC-DOPA was dissolved in DI water to the desired concentrations (e.g., 5, 10, 20 mg/mL). The solution was alkalinized by adding 20 μL of 1 N NaOH per 1 mL of the CMC-DOPA solution. Coating solution was introduced onto plain PDMS and glass surfaces (static condition) or run through the channel using a syringe pump (Fisherbrand™, 780100) with low flow rate (10 μL/h) for 60 min with 5 min interval. The coated surfaces were incubated at 37°C overnight, followed by washing with DI water three times (static condition) or with PBS (flow condition) running through the channel for 2 h at 10 μL/h flow rate.

The coated surfaces were quantified via Calcofluor White stain (CFW) assay (Krause and Olsen, 2020). To create a standard curve, different CMC concentrations (0–200 mg/L) were prepared in DI water. CFW (Sigma-Aldrich, 18909) was diluted at 1:10 v/v ratio in DI water to make a staining solution. Then 100 µL CMC solution was mixed with 40 µL CFW staining solution on top of a well with either a glass or PDMS bottom. The mixture was incubated at room temperature for 15 min then imaged at ×10 objective using DAPI channel (Nikon Eclipse Ts2) with exposure time at 200 ms and 1X gain (n = 4 per concentration). All images were quantified for mean fluorescent intensity values using Fiji. The average of mean intensity values at each concentration was calculated and normalized to the values of 0 mg/L concentration in order to generate a standard curve (Supplementary Figure S1). The surfaces coated with CMC-DOPA (0, 5, 10, 20 mg/mL) were prepared and washed as above-mentioned. 100 μL DI water and 40 µL CFW staining solution were added onto the coated surfaces (n = 6–13 per concentration), followed by incubation and imaging. The resulting fluorescent intensity on each coated surface was quantified. Based on the standard curve, the amount of CMC on the coated surface was calculated.

The wettability of the coated surfaces was quantified via contact angle measurement using a Kruss DSA100 Drop Shape Analyzer. Glass and PDMS surfaces were coated with CMC-DOPA concentrations of 0, 5, 10, and 20 mg/mL, as described above. After rinsing with DI water, the surfaces were allowed to dry for 10 min inside a biosafety hood. A 20 µL volume of DI water was deposited onto each surface; the droplets were then imaged and analyzed using the DSA software. The sessile drop method was applied to measure left and right contact angles of the samples as well as the mean of the two angle values. Ten mean measurements for a single droplet were then recorded, and each coating concentration was assessed in four replicates. The results were plotted and statistical differences between the coated surfaces and the control (0 mg/mL) were determined using Excel and GraphPad.

Tomato cv. Bonny Best (wilt susceptible) and cv. Hawaii 7996 (wilt resistant) plants were grown at 28°C with a 12 h light/12 h dark photoperiod. Xylem sap was collected from five-week-old plants by decapitating the plants above the cotyledons. The cut stems were wiped with Kimwipes to remove phloem sap and cell debris. The xylem sap was collected for a period of 3 h, kept on ice, then filter-sterilized using 0.22-µm filters and frozen at −80°C until use (Khokhani et al., 2017).

The chemical structure of the synthesized CMC-DOPA was characterized by ¹H NMR (Figure 2A). The ¹H NMR spectrum revealed distinct peaks equivalent to the sugar ring of CMC (3.00–4.75 ppm), the CH₂ group (2.80 ppm) and the catechol group (dihydric phenol, 6.75 ppm) of DOPA (Wang et al., 2021). The presence of DOPA in the synthesized product was further demonstrated via the similar UV absorbance spectra and characteristic peaks of the catechol group at 280 nm observed in DOPA and CMC-DOPA (Figure 2C) (Chen et al., 2019). Also, using the absorbance data, we were able to generate a standard curve of DOPA concentrations (Figure 2B) and calculated the DS value of the synthesized CMC-DOPA, which was 2.53 (±0.16) % (n = 4 batches) and consistent batch to batch. Finally, as CMC showed its capability to linearly interfere with methyl blue (MB) absorbance via ionic association [Figure 2D, (Yan and Chai, 2021)], we then examined if CMC-DOPA retained CMC property by comparing the spectra of methyl blue solution alone and its mixture with either DOPA (1 mg/L), CMC (50 mg/L), or CMC-DOPA (50 mg/L). As a result, the DOPA did not cause notable changes in the MB absorbance spectrum while both CMC and CMC-DOPA reduced the absorbance significantly (Figure 2E). Note that the CMC-DOPA had less CMC compared to the CMC at the same concentration, which may affect the slight difference between the two spectra.

To assess the functionality of the CMC-DOPA coated surfaces, we first determined the coating efficiency on glass and PDMS substrates with different CMC-DOPA concentrations via Calcofluor White (CFW) staining and fluorescent image quantification. Note that the mean fluorescence intensity of CFW expressed a direct linear correlation with the increasing concentration of CMC (Supplementary Figure S1) due to the specific binding of CFW to cellulose (Wang et al., 2017; Deng et al., 2020; Krause and Olsen, 2020), which allowed us to directly deduce the CMC content on CMC-DOPA coated surfaces. As the result, we showed increases of detected CMC, 10.45 (±7.05), 22.06 (±12.57), 60.41 (±19.33) mg/L on glass and 18.95 (±8.91), 19.08 (±9.41), 100.13 (±30.25) mg/L on PDMS, with respect to increased CMC-DOPA coating concentrations of 5, 10, 20 mg/mL (Figures 3A, B). The CMC detected at 20 mg/mL was significantly higher than the other concentrations whereas between 5 and 10 mg/mL, there was no significant difference (Figure 3B). Finally, smaller standard deviations on glass surfaces might indicate a more uniform coating on the glass than on PDMS substrate.

The water contact angle test was utilized to assess the wettability of glass and PDMS substrates with and without CMC-DOPA coatings. As the CMC-DOPA concentration increased from 0 to 20 mg/mL, significant changes were observed in the average contact angles (Figures 3C, D). For the glass substrate, the water contact angle decreased from 46.38 (±5.18)° to 15.41 (±2.34)° as CMC-DOPA concentration increased from 0 to 20 mg/mL. Similarly, for the PDMS substrate, the water contact angle decreased from 92.53 (±3.85)° to 45.56 (±8.20)°. These results suggest that CMC-DOPA provided a robust adherent coating for both glass and PDMS substrates and the resulting coated surfaces became more hydrophilic. Note that the contact angle on the PDMS substrate was higher than that on glass due to the inherently higher hydrophobicity of PDMS as compared to glass (Yeo et al., 2006; Morent et al., 2007). In addition, the average of contact angle on PDMS surfaces coated with 5 mg/mL CMC-DOPA, 54.32 (±4.10)°, was slightly lower compared to those coated with 10 mg/mL, 57.58 (±9.31)°, a difference not observed on glass surfaces. There was also no significant difference between 5 and 10 mg/mL CMC-DOPA coatings on PDMS substrates (Figure 3D). Due to the hydrophobicity of PDMS, this phenomenon could be attributed to the coating baseline effect on PDMS at low CMC-DOPA concentrations (10 mg/mL or lower).

We observed that Rps formed biofilm poorly on glass and plastic surfaces such as PVC when it grew in xylem sap, even though this is the native condition Rps thrives in in planta (Supplementary Figure S2). To evaluate the effect of functionalized coatings on improving Rps biofilm formation, we coated the glass surface with 0, 5, 10, 20 mg/mL CMC-DOPA and imaged biofilm formation of Rps GMI1000 on these coated surfaces. We observed that while Rps formed thick biofilm that covered a large portion of the well surfaces in CPG, without CMC-DOPA coating, Rps only formed small microcolonies that covered approximately 10% of the surface in the presence of Bonny Best xylem sap (Figures 3E, F). Coating of glass surface with CMC-DOPA significantly enhanced Rps biofilm formation (Figure 3F), indicated by the increase in surface coverage of Rps biofilm stained with Crystal Violet (Figure 3E). At 20 mg/mL, CMC-DOPA coating increased the Rps biofilm surface coverage by approximately six times compared to control (no CMC-DOPA coating). Similar to coated-glass surfaces, we also found that on CMC-DOPA-coated PDMS, increasing the concentration of CMC-DOPA coating enhanced Rps biofilm surface coverage (Figures 3G, H). Interestingly, on the PDMS surface, the Crystal Violet-stained Rps biofilm appeared to be thicker, but more spatially constricted in the presence of xylem sap. In contrast, Rps formed a thin layer of biofilm covering the bottom of the PDMS wells when the bacterium was grown in CPG broth. These results showed that we have successfully developed a functionalized surface coating with CMC-DOPA, facilitating Rps attachment to surfaces and biofilm formation in the bacterium’s native condition: the tomato xylem sap environment.

The flow profile and its effect on a bacterium in a microchannel were examined both computationally and experimentally. The system flow rate was set as 40 μL/h based on the average xylem sap flow rate of 38.3 (±4.4) μL/hr measured directly from Bonny Best plants (Supplementary Figure S3). DI water was used as the fluid due to its similar viscosity to xylem saps and culture media (Supplementary Figure S4). The simulation confirmed uniform velocity profile throughout the length of the channel (Supplementary Figure S5A). The flow velocity followed the Hagen-Poiseuille parabolic profile with highest values in the center (4.86 μm/s) and smaller towards the channel wall. Moreover, 2 μm microspheres were introduced to the inlet of the xylem-mimicking channels to visualize the flow experimentally (Figure 4A). The microsphere movement through three channels (channel 1, 10, 20) at three different locations (entrance, middle and end) was monitored and quantified. The result showed that the mean velocities of microspheres were comparable among three channels: 2.00 (±0.87), 2.11 (±0.81), and 1.90 (±0.75) μm/s in channel 1, 10, 20, respectively. Also, the velocity ranges were uniform across three channels regardless of channel’s locations and were consistent with the simulation results (Figure 4B; Supplementary Figure S5A). For further understanding the effect of fluid flow on Rps, a bacterial cell model with diameter of 0.7 μm and an end-to-end length of 2.9 μm was introduced to the bottom wall of the channel (Figure 4C). We noticed that the varying shear stress values due to the fluid flow (Supplementary Figure S5B) affected the drag forces applied on the bacterium model. When the bacterium model was placed at the mid-wall (Figure 4C), where the simulated shear stress was found maximum, the drag force was observed to be highest (0.08 pN). As we gradually moved the bacterium model toward the corners of the channel, drag force also decreased (Figure 4D). In addition, when we increased the system flow rate from 0 to 100 μL/h, the drag force also increased linearly from 0 to 0.2 pN (Figure 4E) indicating a linear relationship between the system flow rate and the drag force.

To dissect the effect of seeding time and flow rates on bacterial attachment, we incubated CMC-DOPA coated channels with GFP-expressing Rps cells and monitored the attachment of bacterial cells in the microfluidic devices over a period of 8 h. We found that within the first 4 h, there was minimal attachment of Rps cells onto the surface of the channels (Figure 5A). At 6 h, it was evident that Rps cells attached significantly more along the wall of the channels. Bacterial attachment remained strong at 8 h compared to control, however, was not as robust as the 6 h time-point.

As we observed that the simulated bacterial cell experienced increased shear stress at higher flow rates (Figures 4C, D), we hypothesized that high flow rates not only reduced bacterial attachment to surfaces but also reduced bacterial biofilm formation in xylem vessels. To test this hypothesis, we seeded CMC-DOPA coated microfluidic devices with GFP-expressing Rps GMI1000 and observed the attachment of bacterial cells after 6 h in the devices at different flow rates. We observed that Rps cells attached more to the wall of the channels at lower flow rates (0, 10, and 20 μL/h) than higher flow rates (40 and 80 μL/h, Figure 5B). We found that Rps biofilm formation was higher at 10 μL/h than 0 μL/h, but was reduced significantly at higher flow rates (20, 40, and 80 μL/h) than at 10 μL/h (Figures 5C, D). Of note, at 80 μL/h flow rate, Rps biofilm failed to spread from the entrance of the channels to downstream positions. Surprisingly, Rps formed only microcolonies but not complex biofilm at 0 μL/h flow rate. This could be because of the limited amount of nutrient in tomato xylem sap available for Rps to grow and form biofilm in the small volume of the channels when there was no flow of fresh xylem sap into the channels over the course of the experiment.