Muons penetrate much further than X-rays, they do essentially zero damage, and they are provided for free by the cosmos.
Image credit: Science Source/Digital Globe.
Edited by Natasha V. Raikhel, Center for Plant Cell Biology, Riverside, CA, and approved March 2, 2017 (received for review December 8, 2016)
Significance
Microbial interactions in the rhizosphere, the microenvironment surrounding plant roots, are a major research area for both plant biologists and microbial ecologists. A significant obstacle in this field is the difficulty in studying root–bacteria interactions in real time. Here, we developed a microfluidics-based device that allows dynamic imaging of root–bacterial interactions at previously unattainable spatiotemporal resolutions. The intimate interaction of Bacillus subtilis with plant roots is accurately traced in high resolution to the root elongation zone. We further reveal how bacterial behavior leads to rapid root colonization and exclusion of competing bacteria. Hence, this approach has the potential to transform our understanding of root–bacteria interactions and the modulation of bacterial communities colonizing the root surface.
Abstract
Plant roots play a dominant role in shaping the rhizosphere, the environment in which interaction with diverse microorganisms occurs. Tracking the dynamics of root–microbe interactions at high spatial resolution is currently limited because of methodological intricacy. Here, we describe a microfluidics-based approach enabling direct imaging of root–bacteria interactions in real time. The microfluidic device, which we termed tracking root interactions system (TRIS), consists of nine independent chambers that can be monitored in parallel. The principal assay reported here monitors behavior of fluorescently labeled Bacillus subtilis as it colonizes the root of Arabidopsis thaliana within the TRIS device. Our results show a distinct chemotactic behavior of B. subtilis toward a particular root segment, which we identify as the root elongation zone, followed by rapid colonization of that same segment over the first 6 h of root–bacteria interaction. Using dual inoculation experiments, we further show active exclusion of Escherichia coli cells from the root surface after B. subtilis colonization, suggesting a possible protection mechanism against root pathogens. Furthermore, we assembled a double-channel TRIS device that allows simultaneous tracking of two root systems in one chamber and performed real-time monitoring of bacterial preference between WT and mutant root genotypes. Thus, the TRIS microfluidics device provides unique insights into the microscale microbial ecology of the complex root microenvironment and is, therefore, likely to enhance the current rate of discoveries in this momentous field of research.
Plant roots are among the most productive ecosystems in the topsoil layer. The unique conditions formed within the rhizosphere (i.e., the few millimeters of soil extending from the root surface) foster numerous interkingdom interactions shaped through long evolutionary history (1⇓–3). Plant roots receive a multitude of signals and depending on their physiological state, respond by secretion of various exudates through their outer layer (4, 5). These exudates lead to chemical and physical modifications of the rhizosphere microenvironment. The unique conditions found within the rhizosphere exert a strong selection pressure on the diverse microbial communities associated with the surrounding soils, resulting in the enrichment of specific microbial populations that have evolved to colonize the root surface (2, 6, 7). Although many past studies described pathogenic interactions (8), recent work revealed multiple beneficial effects of root-associated bacteria. In agricultural crops, such beneficial bacteria have the potential to positively impact crop yields by improving plant fitness or through the elimination of root pathogens (9⇓⇓–12). Thus, improved understanding of root–bacteria interactions is of major interest to both basic and applied plant research.
Conditions within the rhizosphere are highly heterogeneous in both space and time. Metabolic profiling of Arabidopsis thaliana root exudates revealed a diverse chemical repertoire, including organic acids, amino acids, dipeptides, and secondary metabolites (13). Single-cell type analysis of Arabidopsis root metabolites showed that different tissues exhibit unique metabolic profiles (14). Thus, root exudate composition likely varies along the length of the root and over the course of root development (15⇓–17). Correspondingly, studies of root bacterial colonization detected specific bacterial species associated with different parts of the embryonic root tissue (15). Nevertheless, current approaches for studying root–bacteria interactions are limited in their capacity to track changes in bacterial colonization of the root surface over space and time and even more to examine bacterial behavior underlying these colonization patterns.
Microfluidic approaches in combination with advanced microscopy provide a powerful set of tools for studying biological systems at the microscale (18). System miniaturization enables precise control over environmental parameters within the system, while allowing direct observation of dynamic biological processes at high spatial–temporal resolutions. Several microfluidic approaches have recently been adapted for use in the plant sciences (reviewed in ref. 19), particularly for the study of root development and physiology (20⇓–22), offering a promising alternative to conventional methods used in the field. Only a single study investigated plant root–microorganism interaction with microfluidics, observing nematode root penetration and hyphae attraction to roots (23). However, root–bacteria interaction dynamics was not observed in such a setup to date. Application of microfluidics-based technologies for investigating bacterial behavior and colonization dynamics at the root interface will provide an indispensable tool in the rapidly evolving field of rhizosphere microbiology (24, 25).
Here, we developed a relatively simple and efficient microfluidic-based system for tracking root–bacteria interaction in real time. The microfluidic device, which we termed tracking root interactions system (TRIS), allows up to nine independent assays of Arabidopsis seedlings. Roots of individual seedlings grow into individual microfluidic chambers, in which they are inoculated with selected bacterial strains and imaged using wide-field or confocal microscopy. This platform allows direct observation and study of bacterial behavior with a living plant root under specific physical and chemical environmental conditions. It, therefore, provides unique insights into the microscale microbial ecology of this complex environment.
The applicability of our method was shown by real-time imaging of the interaction between the soil-borne bacterium Bacillus subtilis and actively growing Arabidopsis roots. B. subtilis is an established model for root-associated bacteria, well-known for its ability to colonize plant roots and induce beneficial effects on plant growth (26⇓–28). Using the TRIS microfluidic device, we observed a unique pattern of B. subtilis attraction directed toward a specific region of the growing root. We further discovered that B. subtilis actively excludes Escherichia coli from the root surface, indicating a possible protection mechanism against root pathogens. Finally, we assembled and applied a modified version of the TRIS device, allowing the simultaneous tracking of two root systems in a single chamber built to study bacterial preference between WT and mutant root genotypes. These experiments showed that the combination of microfluidics and live-imaging microscopy is a powerful method for studying root–bacteria interactions.
Results
Microfluidic Device Assembly and Experimental Setup.
To visualize root–bacteria interaction in real time, we explored the possibility of developing a microfluidic-based system that allows tracking bacteria in high resolution at the root interface. We fabricated TRIS, a microfluidic device that, after it is coupled to the appropriate microscopy and image analysis techniques, provides high spatial and temporal resolution information vis-à-vis bacterial behavior in a root environment. The TRIS device allows monitoring of individual Arabidopsis roots in nine separate channels (Fig. 1). To prevent cross-contamination between channels throughout bacterial inoculation, each channel contains independent inlet and outlet ports. A third port, located at one end of the channel, is used for introduction of the germinated Arabidopsis root. Channels are etched by soft lithography into a single slab of polydimethylsiloxane (PDMS) bonded to a glass microscope slide (29, 30). The design consists of an elongated hexagon with a short extension at one end, with outer dimensions of 13 × 4 mm (Fig. S1). The total internal volume of the channels is ∼6.4 µL, and their height is set to 160 µm, providing ample space for growth of roots having a typical diameter of 100 µm. Roots of up to 8 mm in length could be grown in each channel (i.e., up to 10 d from germination in the case of A. thaliana). Arabidopsis seeds are first germinated in agar-filled pipette tips, which are subsequently placed at the dedicated port of each channel, and the device is incubated for up to 5 d under appropriate conditions (Materials and Methods). After root elongation, the device is placed on a microscope stage, with polyethylene tubes connected to the inlet and outlet ports for liquid handling and inoculation with bacteria (Fig. 1 and Fig. S2). A custom humidity chamber is used to minimize liquid evaporation during long-term imaging. This setup allows for capturing the dynamic behavior of bacteria at and around the root surface under controlled conditions with high spatial and temporal resolutions.
Fig. 1.
TRIS: a microfluidic device for tracking root–bacteria interactions. (A) Illustration of the TRIS device mounted on the microscope stage (dark rim). (Inset) Schematic longitudinal section of a microfluidic channel containing root and bacterial cells (red; not drawn to scale). (B) Top view of Arabidopsis seedlings growing in plastic pipette tips attached to the TRIS device. Roots are visible as thin white lines extending from tip ends. (Scale bar: 1 cm.) (C) Microscopic view of nine Arabidopsis roots growing inside the TRIS microfluidic device captured using bright-field illumination at 10× magnification. Arrows: a, inlet port.; b, tip in a root-dedicated port with c, root extending toward the outlet port; d, outlet port. (Scale bar: 5 mm.)
Fig. S1.
Mask design for generating the TRIS microfluidic device. Microfluidic device mask design used for the photolithography process. (A) Array of nine single microfluidic channels. *Magnification of one single-channel diminutions. (B) Array of six double microfluidic channels. **Magnification of one double-channel diminutions. Permeable divider indicated by the black arrow.
Fig. S2.
The TRIS microfluidic device placed on a microscope stage. View of the TRIS device mounted onto a microscope stage. The device contains microfluidics channels with Arabidopsis seedlings, and the entire setup includes a tubing system, 1-mL syringes, a medium reservoir, and a humidity chamber made of acrylic plastic. (Scale bar: 5 cm.)
TRIS Enables Real-Time Imaging of Bacterial Attraction and Root Surface Colonization.
B. subtilis is known to effectively colonize the surface of growing roots, including that of Arabidopsis (31). To study the initiation of this process, we injected 106 cells per 1 mL fluorescently labeled B. subtilis into the TRIS microfluidic system using preloaded syringes. Bacterial behavior was studied over the next 30 min using video microscopy with dark-field illumination. We consistently observed rapid bacterial accumulation at an area just above the root tip, likely driven by chemotaxis toward high local concentrations of root exudates (Fig. 2A). This behavior is evident already 20 min postinoculation, a period considerably shorter than the bacterial doubling time of ∼24 min in a rich media (Fig. S3), indicating that bacterial accumulation at this stage is not driven by bacterial growth. This bacterial “hotspot” remained noticeable over at least 30 min, with bacterial cells continuing to accumulate throughout the imaging period.
Fig. 2.
Live imaging of B. subtilis cells interaction with Arabidopsis root. (A) Accumulation of B. subtilis cells near the root elongation zone of Arabidopsis seedlings (white arrows) during the first 30 min of coincubation. Each image is an MIP from a short video (11 s; ∼100 frames) obtained using dark-field microscopy. The rapid accumulation suggests a chemotactic response of bacterial cells toward high local concentrations of root exudates. (B) Selected images from time-lapse confocal microscopy of a WT Arabidopsis root (bright field) incubated with red fluorescent-labeled B. subtilis (mKate) for 12 h. Images are representative of nine independent inoculation experiments. (C) A B. subtilis biofilm defective mutant (eps) shows similar attraction as the WT strain but with little or no subsequent biofilm formation. Images are representative of four independent inoculation experiments. The white area plots (on the right side of the mKate fluorescent images) show normalized bacterial quantification. These area plots could be compared between the experiments in B and C; x axes in all area plots have the same length and range between zero and one normalized intensity. (Insets in B and C) Higher magnification view of the area marked by a dashed box in the mKate view. The same area is marked by a dashed box in the bright-field view. Magnification in B shows biofilm formation at a mature part of the root compare to C. White arrows in B and C indicate lower bacterial accumulation, likely because of a lower amount of motile cells at 12 h (compared with the same area at 6 h). (Scale bars: 200 µm.)
Fig. S3.
B. subtilis cells cannot grow in plant media. Growth curve of B. subtilis cells growing in a 96-well plate in Lysogeny broth bacterial media (LB; red line) and a plant media: Murashige and Skoog (MS; blue line). Values are normalized means of bacterial density measured at OD600 (n = 5; independent inoculations for each treatment). Bars indicate ±SD of five independent replicates.
To extend the observation of root colonization over periods of several hours, we combined the TRIS microfluidic setup with laser-scanning confocal microscopy. Using time-lapse microscopy, we followed bacteria–root interactions of nine independent inoculations at 30-min intervals over a period of 12 h. Bacterial accumulation immediately above the root tip was observed already 30 min from inoculation exactly. Bacterial density at the same position increased significantly by fivefold during the first 6 h postinoculation (Fig. 2B and Movie S1). Interestingly, after 6 h, we observed a gradual decrease of bacterial density above the root tip and concomitant bacterial accumulation appearing as large bacterial plugs (likely showing biofilm formation) on the root surface farther from the growing tip (Fig. 2B). Interestingly, the time window of the observed reduction is in line with a recent study, suggesting that the primary adhesion of bacterial cell to root surface would likely be reversible before cellular specification in production of ECM (32).
On surface interaction, bacteria are known to secrete ECM largely composed of exopolysaccharides (EPSs), proteins, and nucleic acids that lead to formation of a biofilm layer (33). To examine more precisely the time and location of biofilm formation on the root surface in the course of root–bacteria interactions, we used an EPS bacterial mutant (eps) that has a defect in its ability to form biofilm. Cells of the eps mutant showed strong and specific attraction above the growing tip, similar to that displayed by the WT strain. However, bacterial colonization over the entire root length, observed during the later stages of the original experiment (from 3 to 4 h and on), was significantly reduced (Fig. 2C and Movie S2).
Another aspect of using the TRIS-based method for root interaction studies is the possibility to simultaneously visualize different fluorescent reporters derived from roots as well as the interacting organism: here, bacterial cells. We used a set of GFP-labeled Arabidopsis lines, each marking a different root cell type, to determine precisely the root zone targeted by B. subtilis attraction. Using the same setup portrayed above, we tracked bacterial attraction to the roots of six Arabidopsis cell type marker lines expressing a GFP reporter, including COBRA-LIKE9 (COBL9; delineating the root hairs), CORTEX (cortex), WER (nonhair epidermis), SCARECROW (SCR; endodermis-quiescent center), WOODEN LEG (WOL; vasculature), and PET111 (columella) (Fig. 3). Results from these experiments specified that B. subtilis cells are primarily attracted toward the border region between the root elongation and early maturation zones. The upper limit of this zone was delineated by the GFP fluorescence of the COBL9 line as well as the higher GFP intensity of the CORTEX marker. This observation was further supported by tracking the upper part of the WER signal, displaying preferential expression pattern in the lateral root cap cells adjacent to the elongation zone (34) (Movies S3–S5). The lower border of the B. subtilis cells attraction zone was defined by the root meristematic cell niche marker SCR as well as the columella marker (PET11), both located below the root elongation zone. Interestingly, this section of the root marked by the two reporters, SCR and PET111, is initially not colonized by B. subtilis cells over the first 2 h of the experiment. The latter observation was supported by the WOL and WER (its lower part) markers. SCR cells are located in the midplane of the root, one cell layer below the WOL cells, and medial to the cells marked by WER, marking the meristematic zone in the Arabidopsis root. By integrating the information from these cell type markers, we could confidently define the hotspot for B. subtilis attraction as the root elongation zone. Moreover, the maximum ratio of bacterial density of the root “young part” vs. the “mature part” (approximately twofold) was observed at 6 h postinoculation (Fig. S4) as detected in the independent experiment described above (Fig. 2). In addition to the significant accumulation of bacteria at the elongation zone, we also detected low but significant accumulation at lower root regions during later stages of these experiments (after 4–5 h postinoculation) (Fig. 3 and Movies S6–S8), indicating an additional root–bacteria interaction process.
Fig. 3.
B. subtilis is attracted to the root elongation zone. Single images extracted from a time-lapse image series of six Arabidopsis strains with a constitutive GFP reporter (green) marking different root cell types. Each strain was incubated with fluorescently labeled (mKate) B. subtilis cells (red). GFP labeling is in the following root cell types: root hairs (COBL9), cortex (CORTEX), nonhair epidermis (WER), endodermis-quiescent center (SCR), vasculature (WOL), and columella (PET111). The presented images were obtained 2 h after introducing labeled B. subtilis into the TRIS device. The two horizontal dashed lines in the merge row denote the region of primary bacterial colonization. The white arrow in the PET111 GFP image points to the green fluorescent signal coming from the columella cell type. One set of images was presented of three independent experiments for each reporter line. (Scale bars: 200 µm.)
Fig. S4.
Bacterial distribution between different parts of Arabidopsis root. Quantification of bacteria at different parts of the Arabidopsis root. (A) Normalized B. subtilis intensity (red lines) and cobl9 cell layer marker normalized intensity (green lines) as a function of root length over 12 h after bacterial inoculation. Small blue lines mark the borders between the elongation zone and the mature zone at different time points. (B) Normalized bacterial density (bacterial intensity per root length) ratio between the young part and the mature part of the root, indicating specific bacterial colonization to the root tip zone up to 5–6 h postinoculation. The cyan solid line is a fitted line showing ratio change pattern. The presented image is a cobl9 root cell layer captured at 2 h after bacterial inoculation. (Scale bar: 200 µm.)
Bacterial Competition for Domination of the Root Surface.
The nutrient-rich microenvironment created by plant roots drives a fierce competition between cohabiting soil microorganisms (35⇓–37). This type of interaction is of particular importance when considering the infiltration of the rhizosphere microbiome by root pathogens. We used the TRIS system to examine in real-time bacteria–bacteria interaction in the presence of roots. Thus, Arabidopsis roots were inoculated with fluorescently tagged cells of E. coli with or without the addition B. subtilis. E. coli is a Gram-negative bacterium and can be found in temperate soil or as a result of manure fertilization (38, 39). Moreover, E. coli strains isolated from plant roots are found to harbor traits indicative for root association, such as biofilm formation and utilization of aromatic compounds likely found in plant tissues (40). When inoculated alone, E. coli seemed to have a markedly different root colonization pattern compared with the one of B. subtilis. Cells of E. coli are typically dispersed homogeneously along the entire root length, with minor accumulation at the root cap starting at 3 h postinoculation (Fig. 4A and Movie S9). However, this pattern changed dramatically after E. coli and B. subtilis were coinoculated. Although B. subtilis followed a similar colonization pattern to the one described above, cells of E. coli were effectively excluded from the root surface within 3 h of inoculation. The bulge-shaped E. coli “exclusion” area started at the root elongation zone, where B. subtilis predominantly colonizes to the highest densities. By the end of the experiment (12 h postinoculation), the exclusion zone was further extended to over 100 μm from the root surface (Fig. 4B and Movie S10). Relative quantification of bacterial fluorescence intensity showed a strong negative correlation between B. subtilis and E. coli concentrations (Fig. 4C, SI Materials and Methods, and Fig. S5). Importantly, cell densities of B. subtilis within the exclusion zone were similar to those measured beyond it, suggesting that the exclusion of E. coli is likely mediated by diffusible agents exuded by either colonized root or B. subtilis in a biofilm structure.
Fig. 4.
Bacterial competition for root surface colonization. Real-time imaging of two bacterial strains at the root zone. (A) WT Arabidopsis root (bright field) inoculated with E. coli cells (GFP) and imaged every 30 min for 12 h. Images are representative of five independent inoculation experiments; Movie S9 shows the complete experiment. (B) WT Arabidopsis root (bright field) coincubated with a mixture of GFP-labeled E. coli (green) and mKate-labeled B. subtilis (red). White dashed lines in the GFP images delineate the exclusion zone of the E. coli from the root. Images are representative of four independent coinoculation experiments. (C) Quantification of normalized bacterial fluorescence intensity in B at each time point of B. subtilis (red–yellow surface) and E. coli (green–blue surface) as a function of the distance from the root surface over 12 h. (Scale bars: 200 µm.)
Fig. S5.
Quantification of coinoculation of bacterial fluorescence intensity. Bacterial density as a function of time and distance from root surface: B. subtilis (red–yellow surface) and Escherichia coli (green–blue surface). 2D graphs show bacterial density as a function of distance from root surface at 0, 6, and 12 h postinoculation.
Bacterial Choice Assays in a Double-Channel TRIS Device.
A modified double-channel version of our original TRIS microfluidic device was manufactured, consisting of a wider chamber partitioned by a semipermeable divider. This microfluidic device enables the simultaneous observation of two Arabidopsis roots growing on each side of the divider in a single background. The divider (exhibiting a pore size of 50 µm) prevents entanglement of the two roots to avoid physical communication between the two plants, while allowing the free flow of bacterial cells and diffusible signaling molecules between the two root systems (Fig. S1B).
To show the effectiveness of the double-channel microfluidic device, we tested the preference of B. subtilis to either WT Arabidopsis or two known root architecture mutants: the hairless caprice/triptychon (cpc/try) and the excessive root hair werewolf/myb23 (wer/myb23) genotypes (41, 42). When comparing two WT plants, B. subtilis accumulation around the two root systems was similar in time, with only a minor difference in intensity (Fig. S6 A and D and Movie S11). In contrast, maximum bacterial accumulation around the two root architecture mutants was reproducibly detected approximately 1 h earlier than for the WT genotype (Fig. S6 B, C, E, and F and Movies S12 and S13). Interestingly, significant bacterial accumulation on the hairless mutant root (cpc/try) was 30% higher than on the WT root when grown side by side in the double-channel device. These results showed the utility of the double-channel device in performing real-time bacterial choice assays and tracking bacterial behavior when exposed to adjacent root systems growing at identical surroundings.
Fig. S6.
Bacterial choice assays in the root environment using the double-channel TRIS device. Time-lapse confocal imaging of two Arabidopsis roots growing in parallel in a double TRIS microfluidic channel separated by a semipermeable divider (dashed bars in bright-field images) (Fig. S1) inculcated with fluorescent B. subtilis cells (mKate). (A) Two WT roots (wt.:wt.) incubated with B. subtilis, (B) WT root (wt.; left) and root of a hairless mutant (cpc try; right), and (C) WT root (wt.; left) and root of a hairy mutant (wer myb23; right). The white arrow in wt.:cpc try in B points to a significant increase in the bacterial “cloud” size at the hairless mutant root. Images are representative of three independent inoculation experiments for each group. (Scale bars: 200 µm.) (D–F) Normalized bacterial intensity at the root elongation zone of each root in three groups: (D) wt.:wt., (E) wt.:cpc try, and (F) wt.:wer myb23. Left and right roots’ intensities are indicated as blue and red lines, respectively. Circles are real data points; solid lines are fitted, and dotted lines are boundaries at 95% confidence intervals.
SI Materials and Methods
TRIS Experiment and Quality Control.
Arabidopsis root diameter of 100 µm is already above the thickness limit of confocal microscope (55). To overcome this limitation and minimize the time required for capturing large images (i.e., imaging area of 6 × 4 mm) and the requirement to track global bacterial behavior around the root, images were acquired at three root longitudinal sections. The root midplane z coordinate (the longitudinal section that intersects with the center of the root) was manually adjusted to be the medial height (z = 0) for each root, whereas the remaining two planes were adjusted to be (+)40 and (−)40 µm relative to the midplane. Conceptually, captured images from one-half of the root are sufficient to describe the bacterial behavioral around the root because of the symmetry of the concentric rings organization of the different root cell layers. As the root continues growing in the course of the experiment, the focused points of the acquired planes with the roots are shifted. Longitudinal sections from both sides of the root midline together with the midline section can provide a set of a representative focus spots for most of the root regions. Focused elements of the growing root in each imaging channel were captured by at least one of the z stacks and used to construct a single focused image. Generally, each experiment with confocal microscopy was composed of a calibration phase, a run phase, and quality control. In the calibration phase, the TRIS device was mounted onto a microscope stage, connected with the input and output tubing, and filled carefully with plant media to prevent air bubbles trapping in the system. The midplane z coordinate was adjusted manually for each root. Single-time point imaging was performed before introducing the bacteria to the TRIS device to ensure the scanned area boundaries and the brightness of the output images. Images passing through the midplane of each root were tested and corrected individually. Before bacterial inoculation, culture OD and swimming ability (tested with dark-field microscopy) were estimated. Moreover, cultures having a low ratio of motile to static cells were excluded as well as samples containing bacterial clumps. The run phase included inoculation of bacteria (106 cells per 1 mL) from preloaded syringes to the inlet of the TRIS device and initiation of imaging. The system was handled without liquid flow during image acquisition. Root–bacterial interaction was imaged every 30 min at three z heights for 12 h. Each image contained a bright-field layer and a red fluorescence layer for the labeled Bacillus subtilis. A green fluorescence layer was added in case of imaging Arabidopsis root cell markers or additional green-labeled bacteria. A strict quality control process was performed manually for every imaged root. In cases of unhealthy and dead Arabidopsis seedlings as well as impaired image sequence, root channels were excluded from additional analysis.
TRIS Design and Assembly.
TRIS microchannels were fabricated in PDMS (Dow Croning) using lithography and replica molding. By photolithography, we patterned the microchannels on a layer of the negative tone resist SU-8 2100 (Microchem) on a silicon wafer, which is referred to as the master mold. By using a mask aligner equipped with a UV source (365 nm), we exposed the layer of SU-8 2100 (160-µm thick) through a transparent mask designed by AutoCAD (Autodesk), printed at high resolution of 50,000 dots per inch (FineLine), and therefore, selectively polymerized the resist (the mask scheme is in Fig. S1) (CAD file can be provided on request). During developing, the unexposed resist was washed away, and the exposed, cross-linked resist remained on the wafer to form the microchannels in the master. The PDMS and the curing agent were mixed at a ratio of 10:1 (10 g PDMS and 1 g curing agent) according to the manufacturer’s instructions. Next, we poured the silicon polymer PDMS on top of the master after degassing, which was subsequently cured by heating to 60 °C overnight. The PDMS formed a negative mold of the device and was then carefully cut and lifted away from the master. Solidified microfluidic device was processed by a puncher (1 mm in diameter) to make the inlet, outlet, and root holes and then mounted to a glass microscope slide.
Metabolomics Analysis of the TRIS Device Content.
The liquid content of a TRIS channel containing WT Arabidopsis root growing for 10 d was collected into a tube by introducing 200 µL miliQ water through the inlet hole facilitated by syringe pump (New Era Pump Systems, Inc.). Collected samples were lyophilized up to full dryness and subjected to the derivatization procedure described previously by Malitsky et al. (56), with small modification. Briefly, to the dried samples, 20 µL 20 mg/mL methoxylamine hydrochloride in pyridine was added, and samples were shaken at 37 °C for 90 min. Subsequently, 50 µL MSTFA (Sigma) was added, and samples were agitated at 37 °C for 30 min. An Agilent 7090A gas chromatograph combined with a time of flight (TOF) Pegasus IV mass spectrometer (Leco) was used for analysis; 1 µL was injected, and carrier gas was set as constant flow of 1.2 mL/min. Chromatography was performed on an Rtx-5Sil MS column (30 m × 0.25 mm i.d.; 0.25 µm; Restek). The GC oven temperature program was 80 °C initial temperature with 2-min hold time and ramping at 15 °C/min to a final temperature of 330 °C with 6-min hold time. The temperature of transfer line and source temperature were 280 °C and 260 °C, respectively. After a solvent delay of 200 s, mass spectra were acquired at 20 scans per second with a mass range from 45 to 800 m/z. Peak detection and mass spectrum deconvolution were performed with ChromaTOF software (Leco). Compound identification was performed by comparison of collected mass spectra and retention index with standard reference mass spectra and retention index available in the National Institute of Standards and Technology library.
Calculating Bacterial Intensity at Different Root Segments.
Bacterial fluorescence signal was calculated based on a longitudinal physical root map generated from each image based on root length as determined from corresponding bright-field images. Displayed are normalized values of summed bacterial intensities to the accumulated bacterial fluorescence during the experiment time (area plots are in Fig. 2 and Fig. S4A). The end of root elongation zone in Fig. S4 was determined based on the minimum GFP fluorescence at the region of the meristematic zone of the root. Bacterial density in Fig. S4B was calculated as normalized bacterial intensity at a specific root segment divided by the root segment length.
Extended Depth of Focus.
FStack function from the File Exchange repository (MathWorks) was used to generate focused images from the captured z stacks at each time point. FStack was modified to process root images as an input considering file size, resolution, and object features for gray images. Modification also includes setting up additional filters and starting parameters. Fig. S8 shows an example of output images of extended depth of focus (EDF) compared with MIP. Selection of focused image information from each of the acquired stacks using the EDF algorithm prevented data multiplication and noise accumulation.
Measuring Bacterial Signal Intensity from the Root Surface.
Root objects were separated from the background of each EDF image by applying edge detection after smoothing followed by contrast enhancement. The same settings were applied to the entire sequence for each experiment. Root edges were emphasized and tracked manually over all of the images in the sequence to ensure perfect root segmentations (Fig. S9). Layers of distance map were produced using contour function to generate a logical matrix each 10 µm from the root surface. To calculate bacterial signal intensity at a certain distance
Discussion
The TRIS microfluidic device reported here offers a robust platform for microscopy-based studies of the interactions between roots and associated microorganisms. Real-time microscopic study of roots is typically limited by the opaque soil surrounding the root, and hence, microfluidics offers an attractive solution to such limitation. Although several platforms for live imaging of plant roots have been reported, they were restricted to observations of root systems (43), but none of them showed their immense value in tracking the dynamics of root interactions with other organisms, particularly bacteria. The TRIS device exhibits two key variations compared with the previously reported RootChip platform (22). The RootChip design uses a two-layer design with a built-in valve system to divide a single fluid input into the different channels. Although this design has several advantages in terms of liquid handling, it does not allow liquid separation between the different channels. We thus opted for a different design, with each channel receiving an independent input, enabling the inoculation of roots with different microbial communities with no cross-contamination. This design also allowed us to implement the device in a single PDMS layer, which in turn, permits easier fabrication, experiment assembly, and operation. Thus, TRIS offers a robust microfluidic platform for studying root–bacteria interactions that can be implemented in a typical experimental biology laboratory.
Experiments involving different organisms (i.e., a plant root and bacteria) spanning three orders of magnitude in scale makes image acquisition a challenging task. Furthermore, bacterial cells are motile, able to move on surfaces because of swarming, and act as a community. Therefore, an optimized scanning procedure is required to follow maximal bacterial behavioral dynamics on the one hand, while scanning multiple roots in parallel (up to nine in our typical setup) on the other hand. Whereas shorter time intervals between frames will permit the analysis of fewer roots per experiment, longer time intervals will result in less accurate measurement of bacterial behavior surrounding the root. Nevertheless, speeding up image acquisition can be likely achieved by the use of light-sheet microscopy through a lower sample laser exposure rate (44).
The application of the TRIS platform was shown for studying three fundamental questions in root–bacteria interactions. Using our principal microfluidic design, we investigated the initiation of bacterial colonization of the root surface as well as competition between different bacterial species. We used a second fluidics design, incorporating two root systems in a single chamber, to examine the ability of bacteria to distinguish between roots of different plant genotypes. These preliminary experiments provided unique insights into the microscale features of the root interaction domain. Perhaps the most robust of these observations was the initial aggregation of B. subtilis at the elongation zone of Arabidopsis roots. This accumulation is strikingly observed already 20 min postbacterial inoculation, much faster than the expected generation time of B. subtilis under the incubation conditions used in these experiments. We thus put forward the hypothesis that this process is primarily driven by bacterial chemotaxis toward exudates secreted from the root surface. The fact that aggregation is only seen at the root elongation zone suggests increased secretion of specific infochemicals from this part of the root, which mediate this attraction (45). Preliminary metabolomics analysis of root exudates performed in the course of this study by analyzing the content of the TRIS device identified several amino acids known to serve as chemoattractants for B. subtilis cells (Fig. S7) (46, 47). This result is consistent with a significant reduction in attraction to Arabidopsis root shown by B. subtilis mutants lacking the chemoreceptors McpB and McpC, known to bind primary metabolites, including amino acids, sugars, and sugar alcohols (32). A thought-provoking link exists between our observations here and a recent report showing Ca2+ flux at the root elongation zone on exposure to bacterial proteins (48), signifying that bacteria–root interactions at this localization may involve a two-way communication.
Fig. S7.
TRIS device enables single-root exudate metabolomics. Selected ion chromatogram of a representative metabolite profile of single-root exudates collected from a 10 d WT Arabidopsis root growing in the TRIS device and analyzed by GC coupled to TOF MS. Numbers indicate selected metabolites identified based on authentic standards: 1, phosphoric acid; 2, isoleucine; 3, glycine; 4, succinic acid; 5, pyroglutamic acid; and 6, myoinositol.
Coinoculation experiments showed active exclusion of E. coli from the root surface after its colonization by B. subtilis. Interestingly, although B. subtilis cells were most concentrated on the root itself, E. coli cells were excluded to a distance exceeding 100 µm from the root surface. This pattern indicates that a diffusible agent released from the colonized root or the colonizing bacteria likely mediates the antagonistic interaction. An induction of either production or release of this agent on root colonization may account for the observed expansion of the exclusion zone over the first 4 h of the experiment. One possible mechanism of E. coli exclusion could be through diffused molecules released from colonizing B. subtilis cells or alternatively, ones secreted from the root on B. subtilis colonization. B. subtilis biofilms are known to secrete the lipopeptide surfactin, which has antimicrobial effect that might generate the observed E. coli exclusion zone (49⇓–51). A recent study showed that an extracellular death factor, a hexapeptide (RGQQNE), secreted by B. subtilis cells during chemotaxis was sufficient to activate the toxin–antitoxin system and induce programed cell death in E. coli cells (52). Additional investigation using the TRIS platform, including chemical and proteomic analyses of root exudates as well as assays with surfactin defective mutants, will likely clarify the nature of the observed bacterial species interaction in the presence of roots.
Using the double-channel design, we could measure bacterial preference for specific root genotypes. Bacterial preference can be regarded as a measurement of the ability of a plant to recruit microbes from the environment. Here, we show the effect of the two main regulatory transcription factors, wer myb23 and cpc try, that control root hair formation on bacterial accumulation. B. subtilis showed an increased affinity toward the architectural mutant roots represented in the early colonization and higher bacterial intensity developed by time, in particular, cpc try, compared with the WT. The elongation zone is well-described as the location at which cells will either develop as a root hairs or remain as epidermis cell (53, 54). Thus, changes in root hair differentiation might be responsible for the altered bacterial attraction that we observed in the double TRIS assay. More profound research is required to identify the mechanisms underlying the observed differences between bacterial attraction and accumulation on surfaces of mutant plants altered in root architecture.
The use of the TRIS device here represents only a small portion of the expected applications of this method in studying root interactions. This system may be extended for studying, for example, plant–plant and not merely plant–bacteria interactions. Endless combinations of bacterial strains and plant genotypes could be examined with multiple fluorescent reporters. Altogether, we anticipate that the reported method will open the way for extensive high-resolution spatial–temporal studies of bacterial dynamics in the root environment.
Materials and Methods
Plant Material and Growth.
WT A. thaliana ecotype Columbia (Col-0) seedlings were grown as described by Grossmann et al. (22) with minor modifications. In short, seeds were surface sterilized in chlorine gas for 2 h in a closed desiccator and transferred aseptically to a plastic pipette tip (one seed per tip) containing one half-strength Murashige and Skoog basal salt mixture (Duchefa Biochemie) supplemented with 0.8% (wt/vol) plant agar (Duchefa Biochemie) and without sugar supplement. The seed-containing tips were preincubated at 4 °C for 48 h in the dark and transferred to a 16-h light/8-h dark period at 23 °C for 3–4 d. Plastic tips with growing seedlings were placed in the TRIS device and incubated vertically for an additional 4–5 d before imaging. All mutant lines and cell layer markers were in the Col-0 background.
Microbial Strains and Culture Conditions.
B. subtilis strain NCIB 3610 and the EPS mutant (eps; strain 3610 background) were mKate labeled, and E. coli OKN3 GFP-labeled was inoculated in 1% (wt/vol) tryptone and 0.5% (wt/vol) NaCl, pH 7 (Sigma; TB buffer) overnight at 37 °C, diluted 1:100 into fresh TB buffer the next day, and incubated again in the same conditions to reach OD600 of 0.6. The cells were washed twice gently with plant media (Murashige and Skoog), suspended in Murashige and Skoog, and kept 30 min at room temperature before applying to Arabidopsis roots in the TRIS device. B. subtilis growth under relevant conditions was assessed by incubation in either Lysogeny broth or Murashige and Skoog. Bacterial density was measured at OD600 every 15 min for 13 h in a plate reader (BioTek; n = 5 for each individual point).
TRIS Design and Assembly.
TRIS microchannels were fabricated in PDMS (Dow Croning) using lithography and replica molding. SI Materials and Methods has technical details.
Setup of the TRIS Assay.
Autoclaved mounted microfluidic devices were filled with liquid plant media (one half-strength basal salt Murashige and Skoog without agar and sugar supplements) carefully to eliminate air bubbles trapped in the system in sterile conditions. Plastic tips containing growing seedlings were placed in the root hole of each channel of the microfluidic device and incubated for 4–5 d in the growth room. On the day of imaging, each microchannel was connected with two polypropylene tubes (1.2 mm o.d.) to the inlet and outlet holes. The tubes were filled with Murashige and Skoog media before connecting them to the TRIS device using 18-gauge needles. Plastic syringes (1-mm volume) filled with washed bacterial cell suspension were connected to the inlet tube to inoculate the chambers. Reservoir tubes (Eppendorf) filled with Murashige and Skoog were connected to the other end of the second tube. Finally, the fully connected TRIS device was mounted onto a microscope stage and covered with a plastic cover containing a transparent window to maintain a humid environment for the Arabidopsis seedlings. The room was temperature controlled and set to 23 °C (Fig. 1 and Fig. S2). After setting the channel position coordinates in the microscope software, the z axis was adjusted to the midplane of each root. Mounting the microfluidic device to a standard microscope slide makes it suitable for different imaging platforms. Images were taken with either dark-field or confocal microscopy. After the microfluidic device was fixed to the microscope stage and imaging parameters were adjusted, bacterial cells were introduced to the TRIS device. Quality control and other experimental tips are in SI Materials and Methods.
Image Acquisition and Dark-Field Microscopy Analysis.
An inverted optical microscope (Olympus) equipped with a 10× dark-field objective was used to record bacterial dynamics. After the TRIS device was properly mounted as mentioned above, root midplane was selected manually for image acquisition. B. subtilis cells were injected into the TRIS device, denoted as t = 0. Short videos (∼100 frames) were captured at a frame rate of approximately nine frames per second, each 5 min, up to 30 min postinoculation. Image analysis was done in two steps using built-in functions in the cellSens software (Olympus): (i) background subtraction from all acquired frames and (ii) generation of a maximum intensity projection (MIP) image of the captured videos for each time point.
Confocal Image Acquisition and Analysis.
An inverted laser-scanning confocal microscope (Nikon Corporation) Ti-eclipse body equipped with a 10× objective lens (CFI Plan Fluor 10×; N.A. 0.3) and an automated stage was used to track long-term fluorescent bacterial behavior. The microscope was programed to scan multiple roots at three z slices (40-µm distance from one to the other) of large images (6.5 × 7.5 mm) each 30 min postbacterial inoculation (t = 0) for 12 h in dark light conditions. Laser illumination emission at 488 nm coupled with a 525/50-nm excitation filter was used for GFP fluorescence, and laser illumination of 561 nm coupled with a 595/50-nm excitation filter was used for mKate fluorescence. Images were exported as tiff files and processed in Matlab (Mathworks). Image analysis is in SI Materials and Methods and Figs. S8 and S9.
Fig. S8.
Generating focused images using the EDF procedure. (A) Raw images of WT Arabidopsis roots (bright field) incubated with red fluorescent-labeled B. subtilis cells (mKate) at three z heights (−40, 0, and +40 µm). (B) MIP and (C) EDF outputs of the raw images shown in A. (D) Binary map used to generate the EDF images in C colored according to three z heights. Shown images are 3 h postinoculation. (Scale bars: 200 µm.)
Fig. S9.
Quantifying bacterial intensity on the root surface. An example of bacterial fluorescence quantification as a function of distance from the root surface. (A) Bright-field image of WT Arabidopsis root. (B) Automated root edges selection (blue line) highlighting the root surface boundaries. (C) Binary mask of the root object. (D) Computer-generated distance map mask layers from the root surface, with a final thickness of 10 µm per layer. Shown roots are 10 d old. (Scale bars: 200 µm.)
Metabolomics Analysis of the TRIS Device Content.
Information is in SI Materials and Methods.
Acknowledgments
We thank I. Kolodkin-gal for fruitful discussions and providing the B. subtilis lines; O. Bitton, L. Tunik, and A. Yoffe for assistance in the nanofabrication facility; and P. Benfey for cell layer reporters and the root mutants. We also thank the Adelis Foundation, the Leona M. and Harry B. Helmsley Charitable Trust, the Jeanne and Joseph Nissim Foundation for Life Sciences, the Tom and Sondra Rykoff Family Foundation Research, and the Raymond Burton Plant Genome Research Fund for supporting the activity of the laboratory of A.A. This research was supported as part of a PhD funded by a Planning & Budgeting Committee of the Council of Higher Education of Israel personal grant (to H.M.). A.A. is the incumbent of the Peter J. Cohn Professorial Chair.
Footnotes
- ↵1To whom correspondence may be addressed. Email: orr{at}agri.gov.il or asaph.aharoni{at}weizmann.ac.il.
Author contributions: H.M., O.H.S., and A.A. designed research; H.M., E.K., and S.M. performed research; H.M., O.H.S., and A.A. analyzed data; and H.M., O.H.S., and A.A. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
See Commentary on page 4281.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618584114/-/DCSupplemental.
References
Live imaging of root–bacteria interactions
Article Classifications
- Biological Sciences
- Plant Biology
Sign up for Article Alerts
Jump to section
You May Also be Interested in
Researchers are exploring whether these ubiquitous fluorinated molecules might worsen infections or hamper vaccine effectiveness.
Image credit: Shutterstock/Dmitry Naumov.
One protein seems to play a key role in touch sensitivity for flytraps and other meat-eating plants.
Image credit: Shutterstock/Kuttelvaserova Stuchelova.
Adam Mastroianni and Daniel Gilbert explore why conversations almost never end when people want them to.
A study finds that giant pandas roll in horse manure to increase their cold tolerance.
Image credit: Fuwen Wei.
See related content: