Abstract
Nutrient-transporting channels are found throughout mature Escherichia coli biofilms, however the influence of environmental conditions on intra-colony channel formation is poorly understood. We report the effect of different substrate nutrient concentrations and agar stiffness on the structure and distribution of intra-colony channels in mature E. coli colony biofilms using fluorescence mesoscopy and quantitative image analysis. Intra-colony channel width was observed to increase non-linearly with radial distance from the centre of the biofilm and channels were, on average, 50% wider at the centre of carbon-limited biofilms compared to nitrogen-limited biofilms. Channel density also differed in colonies grown on rich and minimal medium substrates, with the former creating a network of tightly packed channels and the latter leading to well-separated, wider channels with easily identifiable edges. We conclude that intra-colony channel morphology in E. coli biofilms is influenced by both substrate composition and nutrient availability.
Introduction
Biofilms are biological structures formed by microorganisms in a variety of environments, and they consist of microbial cells embedded in a self-secreted extracellular matrix [1]. Biofilm-forming organisms have evolved numerous mechanisms to protect their constituent cells from external biological and physio-chemical stresses, resulting in populations of cells that exhibit increased resistance to a wide range of deleterious agents when compared to planktonic cells [2].
Interactions between bacterial aggregates and external substrates are crucial to biofilm formation and development, with surface properties affecting cell size [3], colony shape [4], growth rate [5, 6], expansion rate [7], motility [8] and wetting capability [9]. Biofilm-surface interactions can trigger the exchange of biomolecules between different species [10], and have been shown to influence pathogenicity through mechanical deformation of soft substrates and epithelial cell monolayers [11]. As a result, biofilms are a major burden to public health and industrial sectors [12], though the mechanisms of biofilm formation and metabolic behaviours have only recently been explored across spatial scales relevant to these settings [13]. It is therefore essential to gain a deeper understanding of the mechanisms that govern how biofilms form and grow, and how the growth environment influences their structure.
Bacterial growth rate and biomass formation depend on the type [14] and concentration [5] of nutrients available [15]. Altering the nutrient substrate upon which a biofilm grows can illicit profound changes in biofilm morphology, growth dynamics, and mechanical properties [16, 17, 18, 19]. Changing the bulk carbon or nitrogen concentration of growth media can alter pattern formation within biofilms [4, 20, 21], and recent studies on single-species biofilms have identified cross-feeding mechanisms for acetate [22] and alanine [23]. Additionally, varying substrate stiffness has been linked to altered attachment [24, 25], motility [8], growth dynamics [26], and expansion [27] of biofilm constituent cells. Water and nutrient transport mechanisms have been found inside mature biofilms [28, 29, 30], yet how these may be altered by changing environmental conditions is currently unknown.
The complex channel networks recently reported by Rooney et al. [31] were found to facilitate nutrient transport throughout mature E. coli biofilms, however it is uncertain if the architecture of those intra-colony channels is influenced by the substrate composition. It was hypothesised that intra-colony channel morphology changes according to nutrient availability and stiffness of the growth substrate. This hypothesis was tested using the Mesolens [32], an optical mesoscope which enables sub-micron resolution across multi-millimeter-size live biofilms. A custom image analysis workflow was also designed to measure intra-colony channel width under different environmental conditions to establish the role of the growth substrate on E. coli intra-channel morphology.
Results
Nutrient availability affects biofilm morphology at both the microscale and the mesoscale
To test how biofilm morphology is affected by nutrient availability, specific carbon and nitrogen limiting conditions were determined for E. coli JM105-miniTn7-gfp (Supplementary Figure 1). Carbon and nitrogen were limited by controlling the nutrient concentrations of the growth media (glucose and ammonium chloride, respectively). The limiting carbon concentration was 15 ± 5 mM (C:N ratio of 1:1.25) and the limiting nitrogen concentration was 2.5 ± 1.5 mM (C:N ratio 27:1). These are broadly compatible with previous E. coli growth experiments in minimal medium [33].
We investigated the impact of nutrient concentration on biofilm morphology by measuring the global biofilm property of base area (Figure 1). Biofilms grown on carbon-limited substrates and those grown on nitrogen-limited substrates were similar in size, with average base areas measuring 1.374 ± 0.118 mm2 (n = 16 colonies) and 1.668 ± 0.166 mm2 (n = 18 colonies) for carbon- and nitrogen-limited colonies respectively. As expected, biofilms grown on carbon-rich substrates were larger than those grown under carbon limitation (p = 0.00088, t-value = 2.20, 11 degrees of freedom). Carbon excess also led to larger biofilms than nitrogen excess (p = 0.00090, t-value = 2.20, 11 degrees of freedom): biofilms grown on 200 mM carbon substrates had an average base area of 3.738 ± 0.510 mm2 (n = 11 colonies), whereas those grown on 10 mM nitrogen had an average base area of 1.370 ± 0.129 mm2 (n = 11 colonies). Errors given correspond to the standard error on the mean across each dataset. The variation in base area measurements is a direct consequence of colony size variation across the substrate, which is affected by the relative proximity of adjacent competing colonies on the surface as well as by differences in metabolic activity between individual biofilms [34]. Our results show that carbon is the limiting factor governing biofilm base area during growth on minimal medium substrates.
The maximum intensity projections (Figure 2) which are colour-coded by depth also indicate that not all biofilms have the uniform dome-shaped structure typical of E. coli biofilms [7, 35]. Instead, in Figure 2c we observe that the thickest region of the biofilm is located at an intermediate radius, between the centre and the edge of the colony. A phenomenon similar to colony sectoring is observed on carbon-limited substrates (Figure 2a), where radial sections of the colony have a considerably lower fluorescence signal than the rest of the colony. The pattern of intra-colony channels is also distributed heterogeneously inside the biofilm depending on the nutrient availability. Channels grown on nutrient-limited substrates appear to expand radially outwards from the centre in approximately straight lines, whereas on nutrient-rich substrates the channels often change direction sharply and do not follow a straight line. These observations confirm that the morphology of E. coli biofilms is strongly determined by nutrient availability.
Intra-colony channel width increases non-linearly with radial distance from the centre of the colony
Channel width was measured along whole circumferences in E. coli JM105-miniTn7-gfp biofilms grown on M9 minimal medium substrates with limitation and excess of both carbon and nitrogen. The average channel width was plotted against radial distance from the centre of the colony, revealing a non-linear increase in width with radius. Representative plots of n = 3 repeats for each nutrient condition are shown in Figure 3. The minimum observed channel width was between 10 µm and 15 µm, and corresponded to the innermost region of the biofilm (at 200 µm radius). This value was not limited by the resolution of the image datasets, but rather by the smallest radial distance at which channel width could be accurately measured.
The modality of increase in intra-colony channel width was explored by applying exponential fits to each dataset. R-squared values of exponential fits were obtained across 3 replicates, and their averages were 0.778 for 15 mM carbon, 0.664 for 200 mM carbon, 0.805 for 2.5 mM nitrogen and 0.870 for 10 mM nitrogen. By verifying that the increase in channel width along the colony radius is non-linear, we ensured that the change was not just a result of polar geometry.
Carbon-limited colonies possess wider channels than nitrogen-limited colonies
To investigate the effect of nutrient availability on intra-colony channel width, colonial biofilms were grown on substrates with a range of nutrient concentrations: 15 mM carbon (n = 8 colonies), 200 mM carbon (n = 7 colonies), 2.5 mM nitrogen (n = 9 colonies) and 10 mM nitrogen (n = 4 colonies). Intra-colony channel width was compared across nutrient conditions at three normalised radial positions, revealing that channels were approximately 50% wider on carbon-limited substrates than on nitrogen-limited substrates inside the biofilm. This increase was most significant at the mid-radius region of each colony, where channel widths measured on average 21.25 ± 1.26 µm under carbon limitation and 14.38 ± 0.71 µm under nitrogen limitation (p = 2.49×10-6, t-value = 1.96, 776 degrees of freedom).
These data suggest that carbon- and nitrogen-based nutrients mechanistically affect channel morphology in different manners. The number of channels detected inside the biofilm also varied differently depending on the nutrient source, and was largest at mid-radius.
Substrate stiffness affects colony size and channel density
The effect of substrate stiffness (agar concentration) on internal colony morphology was investigated by imaging biofilms grown on soft and hard agar substrates, in both rich and minimal medium (Figure 4). We observed an increase in out-of-focus fluorescence for biofilms grown on rich medium substrates compared to biofilms grown on minimal medium substrates. On minimal medium, channel borders were better resolved due to the higher contrast relative to the rest of the biofilm. On rich medium, the density of intra-colony channels increased with decreasing substrate stiffness. On 0.5% agar LB substrates (Figure 4a) channels were densely packed in the whole biofilm, while some widely-separated channels progressively appeared on portions of biofilms grown on stiffer substrates (1% and 2% agar concentration, Figures 4b and 4c respectively).
Varying agar concentration was also associated with a change in biofilm base area for both rich and minimal medium substrates (Figure 5). On rich LB medium substrates, decreasing agar concentration resulted in an increase in biofilm base area, which measured on average 1.733 ± 0.245 mm2 on 2% agar (n = 12 colonies), 3.080 ± 0.477 mm2 on 1% agar (n = 15 colonies) and 3.854 ± 0.772 mm2 on 0.5% agar (n = 13 colonies). Interestingly, the effect was opposite on M9 minimal medium substrates, where the average base area was 2.253 ± 0.224 on 2% agar (n = 18 colonies), 2.434 ± 0.184 mm2 on 1% agar (n = 12 colonies) and 1.387 ± 0.142 mm2 on 0.5% agar (n = 9 colonies). Two-tailed t-tests were performed between the 0.5% and the 2% agar concentration data for both rich and minimal medium, and revealed that agar stiffness had a greater effect on biofilm base area on minimal medium (p = 0.0032, t-value = 2.06, 25 degrees of freedom) than on rich medium (p = 0.0202, t-value = 2.14, 14 degrees of freedom). The difference in base area between biofilms grown on rich and minimal medium substrates was most significant at 0.5% agar concentration (p = 0.0078, t-value = 2.16, 13 degrees of freedom).
Discussion
This study constitutes the first quantitative analysis of environmental modulation of intra-colony channels in mature E. coli biofilms. Using a simple custom image analysis pipeline based on the open-source software FIJI and the Python programming language, it was possible to quantify biofilm properties at both the mesoscopic scale (by measuring biofilm base area) and at the microscale (by measuring intracolony channel width). Both quantities were used to investigate the effect of environmental conditions on biofilm morphology, which was found to be affected by both the type of substrate (rich or minimal medium agar) and by substrate properties (stiffness and nutrient concentration). Thanks to its simplicity, our pipeline can easily be adapted to the study of internal patterns of other biofilm-forming species. Our results show that intra-colony channel architecture is heavily affected by substrate composition.
The mesoscopic effects of nutrient limitation were studied by calculating biofilm substrate area. Nutrient abundance in the substrate, achieved by increasing the amounts of either glucose or ammonia, led to larger biofilms than nutrient limitation. Biofilm base area was most significantly affected by the carbon concentration in the growth substrate and was 2.7 times larger for biofilms grown on carbon excess than on carbon limitation. On the other hand, increasing the nitrogen concentration on a substrate with sufficient carbon availability did not significantly affect biofilm base area. These findings agree with the reported increase of biofilm formation with glucose concentration in Staphylococcus species [36] and the increase in colony size with glucose concentration for E. coli [37] and yeast [38].
Biofilm base area was also affected by substrate stiffness, but in opposite manners depending on the medium. An increase in agar stiffness in rich medium was associated with a reduction in colony size, due to slower radial expansion of biofilms on the surface brought by stronger frictional forces [7]. The trend of increased base area with decreased agar stiffness has previously been observed in Vibrio cholerae biofilms grown on LB medium with different agar concentrations [4]. The opposite trend was observed on minimal medium: biofilms grown on stiff substrates had a base area on average 1.6 times larger than those grown on semi-solid substrates. This could be due to faster assimilation of nutrients occurring on semi-solid minimal media with respect to stiff minimal media, causing a halt to radial expansion. Finally, colony size was on average almost 2.5 times larger on semi-solid rich medium substrates than on semi-solid minimal medium substrates. The slower growth of E. coli in minimal medium with respect to rich medium (with average doubling times of 18 ± 0.5 minutes in LB and 38 ± 1.1 minutes in M9-glucose [39]) can explain this difference in biofilm base area only in part, since we observed no significant difference in base area between rich and minimal medium substrates at higher agar concentrations of 1% and 2%.
Environmental conditions also altered the distribution of intra-colony channels inside biofilms. Channels formed on substrates with limiting carbon or nitrogen concentrations expanded radially outwards, in approximately straight lines. This could be because bacteria growing on nutrient-depleted substrates adhere less strongly on the surface of attachment, and expand more rapidly [40], or may be indicative of nutrient foraging behaviour. On the other hand, channels formed on nutrient-rich substrates formed complex fractal patterns. It is hypothesised that this type of structure emerges due to rapid cell growth and larger cell dimensions, thereby resulting in denser colonies creating a tighter network of channels. This in turn enables a greater proportion of cells to access the nutrients that are transported within the biofilm. A recent study by Fei et al. [4] highlighted two distinct types of patterns forming on the surface of V. cholerae biofilms, namely radial stripes and zigzag herringbone patterns. Conversely, the biofilms in our study only involved the same type of fractal patterns: intra-colony channel structures had homogeneous morphology, with only width and spacing varying depending on environmental conditions. Nonetheless, it is important to remember that the wrinkle patterns of V. cholerae and B. subtilis are substantially different from intra-colony channels in E. coli, since the former are surface features arising during biofilm expansion because of frictional forces, whereas the latter permeate the whole biofilm volume and have a complex 3D architecture and nutrient uptake function.
Furthermore, we found that substrate stiffness affected intra-colony channel density on rich medium substrates: channels were tightly packed and had a uniform, narrow width on semi-solid substrates, but as the agar concentration increased, wider channels appeared in portions of the biofilm. This observation appears to be in contrast with previous work on V. cholerae biofilms, where homogeneity in radial feature distribution and spacing at the edge of mature biofilms increased with agar stiffness [41]. An increase in out-of-focus fluorescence was also observed in widefield epi-fluorescence images of biofilms grown on rich medium compared to those grown on minimal medium. The fluorescence signal inside channels was likely due to bacterial cells being present inside them, as a result of movement inside the biofilm, collective reorientation [42] or shedding of cells from the inner walls of the channels. Previous studies on V. cholerae [43] confirmed that expansion in the depth direction is dependent on friction between the substrate and the expanding cells, with softer substrates leading to flatter biofilms than stiff substrates. This could explain the reduction in contrast between channel border and constituent cells observed on stiff rich medium substrates. A variation in internal biofilm architecture along the depth direction has also previously been reported for E. coli growing on agar substrates [44] and was attributed to differences in extracellular matrix components assembly and organisation inside the biofilm volume.
Finally, the variation in channel width inside mature biofilms was investigated: channel width increased non-linearly along the radial direction from the centre of the biofilm under all nutrient conditions. We hypothesise that channel width is larger at the edge of the colony due to rapid cell growth in nutrient sufficient conditions, resulting in non-uniform radial expansion, as previously observed for B. subtilis [40] and V. cholerae [4]. Remarkably, channels were on average 50% wider at the centre of carbon-limited media than inside nitrogen-limited media. Bacteria regulate their cell size and growth rate in response to nutrient availability [45], suggesting that the overall growth rate and cell dimensions affect the manner in which channels form across the colony, resulting in variations in channel dimensions as an emergent property of biofilm expansion.
Studying biofilm internal patterns at the cellular level is a task mostly relegated to theoretical modelling and computer simulations [7, 46, 47, 48, 49]. Experimental approaches in imaging biofilms at single-cell resolution [50, 51, 52] have been limited in the total biofilm size that can be imaged at once (under 100 µm2). This limitation has recently been overcome thanks to the Mesolens, which has enabled the discovery and characterisation of intra-colony channels inside E. coli biofilms and insight into how the environment shapes their formation.
Methods
Materials and strains
The non-pathogenic E. coli strain JM105 (endA1, recA1, gyrA96, thi, hsdR17 (rk–, mk+), relA1, supE44, ∆(lac-proAB), [F’ s traD36, proAB, laqIqZ∆M15]) containing mini-T7 gfp [53], enabling GFP fluores-cence, was used throughout the study. Liquid cultures were grown in Lysogeny Broth (LB) [54]. Minimal medium growth was in M9 medium [54], prepared in a 5x concentrated solution (Na2HPO4 30 g/L, KH2PO4 15 g/L, NH4Cl 5 g/L, NaCl 2.5 g/L) then diluted to 1x with distilled deionised water and supplemented with 1 mM MgSO4· 7H2O, 0.2% (w/v) glucose and 0.00005% (w/v) thiamine. Solid substrates were made by adding agar prior to autoclaving, in concentrations of 20 g/L (LB agar) and 15 g/L (M9 agar) for nutrient concentration experiments. For agar stiffness experiments, LB and M9 media were prepared as described above, but the agar concentration was progressively reduced for soft substrates (giving 5 g/L, 10 g/L and 20 g/L for both LB and M9 media).
Imaging chamber design and 3D-printing
The imaging chamber to be used with the Mesolens was designed on AutoCAD (Autodesk, USA) by modifying the original design of Rooney et al. [31]. The chamber was designed to mimic a Petri dish, and consisted of a 120 mm x 100 mm x 12 mm plate with a 60 mm diameter, 10 mm deep well at its centre to hold the agar substrate (Supplementary Figure 2). After 3D printing the chamber in black ABS plastic (FlashForge, Hong Kong) using a FlashForge Dreamer 3D printer (FlashForge, Hong Kong), the corners were smoothed out using a scalpel. This reduced the movement of the imaging chamber inside the square plate. The chamber was sterilised with a 70% (v/v) ethanol solution, and then under UV light for 20 minutes immediately before the addition of sterile solid growth medium.
E. coli biofilm specimen preparation
Liquid cultures of E. coli JM105-mini-Tn7-gfp were prepared in LB medium supplemented with 25 µg/mL gentamicin and incubated overnight at 37 °C. Overnight cultures were diluted (1:100) into fresh LB medium and incubated at 37 °C until they reached an optical density of 0.5 (mid-exponential growth phase).
To grow biofilms on rich medium, a liquid culture with density 1×104 colony-forming units (CFUs) per mL was inoculated on the LB agar substrate inside the imaging chamber. The specimen was incubated at 37 °C for 24 hours in darkened condition prior to imaging. For minimal medium with varying nutrient concentration, mid-exponential phase liquid cultures were washed three times with 1x M9 salts. They were then resuspended in M9 medium with appropriate amounts of glucose and ammonium chloride (for carbon and nitrogen variation, respectively). M9 agar substrates cast into the 3D-printed imaging chambers were inoculated at a concentration of 1×104/mL, which ensured colonies were sufficiently spaced out and did not have to compete for nutrients with others in their proximity. It was also ensured that only one colony was visible in the field of view of the Mesolens at once, which prevented background signal from nearby colonies from reaching the detectors.
E. coli growth characterisation under different nutrient conditions
Mid-exponential growth phase liquid cultures of E. coli JM105-mini-Tn7-gfp were prepared as described above, then washed and resuspended in 1x M9 salts. The cultures were diluted to an OD600 of 0.04, split in individual tubes and supplemented with appropriate amounts of the nutrient of interest. The nutrient concentrations were chosen to be between 0 mM and 80 mM - for comparison, the carbon and nitrogen concentrations in nominal M9 salts are 66.6 mM and 18.7 mM, respectively. Aliquots (200 µL) of liquid culture of each investigated concentration was added in triplicate to a black Nunc MicroWell 96-well optical-bottom plate with polymer base (ThermoFisher Scientific, US). Absorbance (OD600) measurements were performed every 15 minutes for 24 hours, with the plate being held at 37°C and shaken continuously.
Growth curves were produced for each nutrient concentration thanks to the Gen5 microplate software (BioTek, USA), and exported to MATLAB (MathWorks, USA) for analysis. The average of the first absorbance value for each concentration was used as a baseline and subtracted from each growth curve. The y axis was displayed in logarithmic scale, in order to identify the exponential growth phase (the linear portion of the plot). A linear fit to this region was applied using MATLAB’s Curve Fitting Toolbox by selecting an exponential function of the form y = a exp(bx), with the specific growth rate corresponding to the coefficient b. While error analysis was produced internally in MATLAB as R-squared, SSE and RMSE values, the error bars on the specific growth rate plots were calculated as the standard deviations across biological repeats for the same nutrient concentration. Note that the duration of the exponential growth condition varied depending on the nutrient concentration, and lasted between 6 and 18 hours, hence the data points corresponding to exponential growth were selected manually. Growth curves and specific growth rates were plotted in Python using matplotlib.
Widefield epi-fluorescence and confocal laser scanning Mesolens imaging
All the data in this work was acquired using the Mesolens, a custom-made optical microscope combining high numerical aperture (0.47) with low magnification (4x). These characteristics allow simultaneous imaging over a field of view of 6 mm x 6 mm with sub-cellular lateral resolution throughout (700 nm). Images can be acquired over a total thickness of 3 mm with an axial resolution of 7 µm. For widefield epi-fluorescence Mesolens imaging, GFP fluorescence was excited by a 490 nm pE-4000 LED (CoolLED, UK), and emitted fluorescent light was made to pass through a 540 10 nm bandpass filter before being detected by a VNP-29MC sensor-shifting CCD camera (Vieworks, South Korea). The camera uses sensorshifting to acquire 9 images for each pixel (each shifted by 1/3 of a pixel width from the other, forming a square grid), effectively increasing the resolution to 259.5 MP [55]. In confocal laser scanning mode, GFP fluorescence was excited using a 488 nm laser (Multiline Laserbank, Cairn) at 5mW power. Fluorescence emission was filtered through a 525/39 nm bandpass filter (MF525-39, Thorlabs, USA) before being detected using a PMT (PMM02, Thorlabs, USA). The emission path also included a 505 nm longpass dichroic mirror (DMLP505R, Thorlabs, USA). Both widefield and confocal laser scanning microscopy were performed with the lens in water immersion, to match the refractive index of the mounting media (LB and M9), except the 0.5% M9 agar datasets which were acquired with air immersion. This was needed to preserve biofilm structure, as colonies grown 0.5% M9 agar were less adherent to the substrate and detached from it after addition of liquid mounting medium. The intra-colony channels formed by these biofilms remain evident even though the lower refractive index resulted in poorer spatial resolution (Figure 4d). The Mesolens collars were adjusted to minimise spherical aberrations.
Image analysis
Confocal z-stacks were displayed as a maximum intensity projection, colour coded by depth across the z-direction using the “Fire” lookup table from FIJI. Where necessary, median-filtered widefield epifluorescence Mesolens images (filter radius: 2 pixels) were deconvoluted using the proprietary Huygens Professional version 19.04 software (Scientific Volume Imaging, Netherlands). Deconvolution was performed after an in-built manual background subtraction and a theoretical point spread function estimation on Huygens, using the Classic Maximum Likelihood Estimation method with 50 iterations, a signal-to-noise ratio of 40 and a quality threshold of 0.01.
To calculate colony base area, widefield Mesolens images were opened in FIJI, and thresholded using the “adjust threshold” tool, using the mean of gray levels as the threshold value. The “Wand (tracing) tool” was then used to select the biofilm mask, and the area was calculated using the “Measure” function. The base area measurements were systematically underestimated due to the thresholding method used by FIJI to create the binary mask. Nonetheless, by using the same thresholding method on all images we ensured that this limitation had comparable effects on all datasets. Biofilm thickness was calculated by plotting z-axis profiles of confocal z-stacks in FIJI and calculating the distance between the two minima on each profile.
Mesolens images were brightness- and contrast-adjusted using FIJI for presentation purposes [56].
Intra-colony channel width calculation
Intra-colony channel width was calculated using the image analysis workflow outlined in Figure 6. The pipeline made use of the FIJI plugin “Polar transformer” [57], which performed an image transformation from polar to cartesian coordinates, as well as of two custom Python scripts. Initially, the coordinates of the centre of the colony were calculated using the “Measure” function on FIJI applied on an oval selection of the whole colony. The Polar Transformer plugin was then launched, and the origin of transformation was manually entered as the coordinates of the colony centre. The number of pixels for each degree in angular coordinates was set to 7200 (corresponding to 20 pixels for each degree). The polar-transformed image was contrast-adjusted using the “CLAHE” (Contrast Limited Adaptive Histogram Equalization) feature on FIJI with block size 60, maximum slope 3 and 256 histogram bins to facilitate channel identification. The image was then despeckled using the “Despeckle” function on FIJI in order to remove noise. The image was also inverted, making intra-colony channels appear light and cells dark, which facilitated the rest of the analysis. Vertical line selections were taken at different x positions on the polar-transformed image, corresponding to circumferences around the colony taken at different radial distances from its centre. Signal profiles were obtained for each line selection using the “Plot profile” feature on FIJI, and were exported to Python, where signal peaks were located thanks to a custom script using the “find peaks()” function. For each dataset, peak thresholding was performed in order to exclude noise: peak prominence was chosen as 20% of the difference between maximum and minimum signal, and a minimum distance between adjacent peaks was selected as 9 pixels, which corresponds to the average length of an E. coli cell (2 µm) on a Mesolens widefield epifluorescence image. Intra-colony channel widths were then calculated as the full-width at half-maximum of each peak. The actual width of the channels in µm was converted according to the radial distance from the centre, using polar coordinate geometry.
Variation of intra-colony channel width along the radial dimension
Intra-colony channel width was measured along whole circumferences at radius intervals of 50 µm, starting from an initial radius of 200 µm. This initial radius value was chosen because at lower values, polar transformed images were distorted and signal analysis was not reliable. Because of the ellipticity of the colonies, polar-transformed images appeared as rectangles with a non-straight right side (Figure 6), corresponding to the colony perimeter. Regions beyond which a full circumference could not be obtained for the colony were excluded from the analysis.
Channel widths were measured for colonies grown on the four nutrient conditions described in the Results section. Three biological replicates were used for each condition, giving a total sample size of 12 colonies. An exponential fit was applied to each plot of channel width against radius, and R-squared values for the fit were computed for each dataset. The number of channels identified by the script varied between radial positions, and was highest for the intermediate radii. This is likely due to the fractal nature of intra-colony channels, which have sharp changes in direction and fold on themselves inside the colony.
To compare channel widths across different nutrient concentrations of the substrate, intra-colony channel width was calculated at three normalised radial positions (maximum radius, mid-radius and 20% radius) to investigate the effect of nutrient concentration. Datasets were acquired for each of the following nutrient concentrations: 15 mM carbon (8 colonies, giving 1062 channels), 200 mM carbon (7 colonies, giving 1424 channels), 2.5 mM nitrogen (9 colonies, giving 1524 channels) and 10 mM nitrogen (4 colonies, giving 507 channels).
Statistical analysis
The comparison of colony base areas and channel widths under different nutrient availabilities and agar stiffness values was performed by means of a two-tailed t-test, assuming unequal variances between each pair of datasets. The test was performed using the Data Analysis package in Microsoft Excel 2016. P-values smaller than 0.05 were considered statistically significant. T-values and degrees of freedom were quoted for each t-test.
Data and code availability
The Python script used to calculate intra-colony channel width was written on Python version 3.7, and is deposited on GitHub (https://github.com/beatricebottura/biofilm_channel_width, DOI: 10.5281/zenodo.5786305). An example dataset to be analysed with the script is also available in the same repository. The remaining data used to generate channel width results is available upon reasonable request.
Author Contributions
BB conducted all experiments and analysed all data. All authors were responsible for the experimental design. All authors prepared the manuscript.
Competing interests
The authors declare no competing interests.
Acknowledgements
We thank Dr Ainsley Beaton (John Innes Centre, formerly University of Strathclyde) for the kind gift of the JM105-miniTn7-gfp strain. We also acknowledge Dr John Munnoch (University of Strathclyde), Dr Rebecca McHugh (University of Glasgow) and Dr Morgan Feeney (University of Strathclyde) for helpful discussions. Beatrice Bottura is supported by a University of Strathclyde Student Excellence Award. Gail McConnell is supported by the Medical Research Council [MR/K015583/1] and Biotechnology & Biological Sciences Research Council [BB/P02565X/1, BBT011602]. Paul A. Hoskisson is supported by funding from BBSRC [BB/T001038/1 and BB/T004126/1] and the Royal Academy of Engineering Research Chair Scheme for long term personal research support [RCSRF2021\11\15].
References
- [1].↵
- [2].↵
- [3].↵
- [4].↵
- [5].↵
- [6].↵
- [7].↵
- [8].↵
- [9].↵
- [10].↵
- [11].↵
- [12].↵
- [13].↵
- [14].↵
- [15].↵
- [16].↵
- [17].↵
- [18].↵
- [19].↵
- [20].↵
- [21].↵
- [22].↵
- [23].↵
- [24].↵
- [25].↵
- [26].↵
- [27].↵
- [28].↵
- [29].↵
- [30].↵
- [31].↵
- [32].↵
- [33].↵
- [34].↵
- [35].↵
- [36].↵
- [37].↵
- [38].↵
- [39].↵
- [40].↵
- [41].↵
- [42].↵
- [43].↵
- [44].↵
- [45].↵
- [46].↵
- [47].↵
- [48].↵
- [49].↵
- [50].↵
- [51].↵
- [52].↵
- [53].↵
- [54].↵
- [55].↵
- [56].↵
- [57].↵