Three strains of E. coli namely MG1655, Nissle 1917 and O157:H7 were used in this study. MG1655 was obtained from the Institute of Engineering and Biotechnology, O157:H7 (CICC 21530) was obtained from China Center of Industrial Culture Collection, Nissle 1917 (DM6601) was obtained from German Collection of Microorganisms and Cell Cultures.

Three bacterial strains were inoculated into Luria-Bertani (LB) broth medium and let to grow at 37°C, 220 rpm till logarithmic growth period, respectively. OD600 of the bacterial suspension was measured and adjusted to Abs = 1 (approximately 1 × 10⁹ CFU/mL) using fresh LB broth, and then it was further diluted 100 times to OD600 = 0.01. Then they were transferred into a 24-well cell culture plate (Corning, CLS3524) respectively. Then they were incubated at 25°C for 12, 24, 36, 48, or 60 h. After that, supernatant containing planktonic cells were removed and the wells were gently rinsed thrice with pure water.

Biofilms were stained with 0.1% (V/V) crystal violet for 20 min, rinsed three times with pure water, and the wells were dried completely. 95% (V/V) ethanol was added to each well and kept for 15 min to resuspend the stained biofilms completely. The resulting absorbance at 595 nm was measured using a spectrophotometer.

Bacterial cells with an initial concentration of OD600 = 0.01 in LB broth medium were transferred into glass-bottomed dishes (Cellvis, D29-14-1-N) and incubated at 25°C for 24 or 48 h to form biofilms, respectively. Subsequently, supernatant was removed and the biofilms attached on the dish were rinsed gently three times in PBS. The biofilms were then stained with SYTO 9 (Invitrogen, L13152) according to the manufacturer’s instructions, briefly, 3 μM SYTO 9 stain was added to the biofilms and incubate at room temperature in the dark for 15 min. Then it was examined using confocal laser scanning microscopy (ZEISS LSM 880) with an excitation of 488 nm. Three-dimensional images of the biofilms were generated through Z-stack imaging, utilized the Zen Black software (Zeiss, Oberkochen, Germany). Image J was subsequently used to calculate the integrated density.

Bacterial cells with an initial concentration of OD600 = 0.01 in LB broth medium were transferred into a 24-well cell culture plate respectively, with each well containing a piece of sterile cover slip. Then the plates were incubated in a 25°C incubator for 24 or 48 h. Biofilms would grow and attach on cover slips. Prior to observation, the biofilms were fixed using a 2.5% (V/V) glutaraldehyde solution for 4 h, followed by rinsing with PBS three times. Gradient dehydration was performed using ethanal concentrations (V/V) of 30, 50, 70, 80, 95, and 100% for ten minutes each. After CO₂ critical point drying and gold sputter coating, the samples were observed using a scanning electron microscope (Zeiss, Germany, EVO LS15).

Bacterial cells with an initial concentration of OD600 = 0.01 in LB broth medium were transferred into a 24-well cell culture plate respectively, with each well containing a piece of sterile CaF₂ window. Then the plates were incubated in a 25°C incubator for 48 h. Subsequently, the CaF₂ windows were rinsed three times with deionized water to eliminate any planktonic microorganisms. The windows were then immersed in a 4% paraformaldehyde solution for 30 min, followed by another wash with deionized water and left to dry at room temperature. Finally, FTIR microspectroscopy investigations were carried out at the beamline BL01B and BL06B of Shanghai Synchrotron Radiation Facility (SSRF). The aperture was set to 30 × 30 μm with a step size of 15 × 15 μm. The spectral range was 4,000–650 cm⁻¹ with a spectral resolution of 4 cm⁻¹, and each spectrum comprised 32 co-added scans.

Following IR mapping images, 9-point smoothing, baseline correction and normalization on CytoSpec were carried out. PCA (principal component analysis) and heterogeneity analysis were subsequently executed on Matlab R2021b, as detailed in our previous publications (Baker et al., 2014; Wang et al., 2023a,b).

Hypothesis testing was conducted to characterize the significance of differences between groups by GraphPad Prism7. Then, the IR peak positions of each biofilm were statistically analyzed and compared by two-sample t test.

The ability of the probiotic E. coli Nissle 1917 to form biofilms is compared with the opportunistic strain MG1655 and the pathogenic strain O157:H7. Crystal violet (CV) is a membrane-permeable basic trianiline dye that can be used to quantitatively assess biofilm biomass (Lianou and Koutsoumanis, 2012; Wilson et al., 2017). Three E. coli strains were inoculated into 24-well plates and incubated at 25°C each (Mathlouthi et al., 2018). After staining with CV, the formed biofilms were quantitatively analyzed by measuring the optical density at 595 nm. As shown in Figures 1A,B, the biofilm formation ability of MG1655 was the strongest, followed by Nissle1917, while O157:H7 was the weakest. The biomasses of biofilm formation of the strains were low at 12 h, then increased significantly from 24 h and increased along with time to reach the maximum value at 48 h, after which it started to decrease. This meant that the three E. coli biofilms started to enter the attachment stage at about 12 h, then enter the colonization and development stage from 24 h to reach the maturity stage at about 48 h, after which they started to disperse and colonize elsewhere, so the density of the organisms may start to decrease (Nag and Lahiri, 2021). Apart from this, when the biofilms are fully mature, the lack of sufficient space and nutrition will also lead to the death of microbial cells (Tahmourespour, 2012). The biofilm mass of the three strains started to differ significantly at 24 h, at which time the amount of biofilm formed in Nissle 1917 is 4.2 times less than MG1655, and 2.4 times more than O157:H7. At both 48 h and 60 h, the amount of biofilm formed by Nissle 1917 is approximately 5 times less than MG1655, and 2 times more than O157:H7. In conclusion, the biofilms of the three E. coli strains all enter the colonization and development stage from 24 h and enter the maturity stage at around 48 h, with MG1655 forming more the amount of biofilm mass, followed by Nissle 1917 and O157:H7. Note that, in a recent paper, Ramadevi et al. found that Nissle 1917 could form more the amount of biofilm compared to MG1655 at 37°C for 48 h, although two strains have a similar growth rate (Ramadevi et al., 2023). The difference may be due to the different temperatures used in our experiments. Gill et al. reported that Nissle 1917 forms biofilm in a temperature-dependent manner (Gill et al., 2019). Comprehensive analysis of the various physiological conditions, including temperature, nutrient, pH and osmotic stress, on the probiotic activity is necessary for its clinical application.

The morphologies of probiotic, opportunistic and pathogenic E. coli biofilms were observed using confocal laser scanning microscopy (CLSM), which has been used to map biofilms and quantify the distribution of extracellular proteins, polysaccharides, lipids, nucleic acids, and other biomolecules (Zhang et al., 2019). To visualize their growth characteristics and status, the biofilms (Nissle 1917, MG1655 and O157:H7) were allowed to grow for 24 h and 48 h, respectively, then the bacterial cells were stained with SYTO 9 and scanned under 488 nm excitation. The two-dimensional and layered three-dimensional scanning images of the three E. coli biofilms are shown in Figure 2. It can be seen that, at the 24 h, which corresponds to the colonization and development stage indicated in Figure 1, many bacterial cells adhere to the surface and form microcolonies in Nissle 1917 and MG1655, and the number of cells in MG1655 was higher than that in Nissle 1917, whereas only a few microcolonies adhere to the surface in O167:H7. After incubation for 48 h to the maturation stage of biofilms, Nissle 1917 colonies were connected over a large area forming thick communities of 3.53 μm; MG1655 has the thickest biofilm of 6.67 μm with many “island” shaped structures which indicating a more mature biofilm with complex three-dimensional communities; pathogenic O157:H7 has the thinnest biofilm of 1.96 μm, microcolonies were loosely dispersed. The integrated density was also calculated and shown in Figure 2 (right), it can be seen that at 24 h, the biomass of Nissle 1917 and MG1655 was greater than O157:H7; at 48 h, MG1655 had the most biofilm mass while O157:H7 had the least. These results are consistent with Figure 1, where biofilms formed by opportunistic pathogenic E. coli were the thickest, while pathogenic E. coli were the thinnest, and biofilms formed by probiotic E. coli were somewhere in between. It is well known that the attachment of the bacterial cell to a particular region is required for biofilm formation (Ramadevi et al., 2023). We can expect that the production of curli in probiotic E. coli will be modest.

SEM was also used to examine the differences in the spatial structure between the three E. coli biofilms (Figure 3). The results showed, all three bacteria adhered to the surface after 24 h. In addition, Nissle 1917 formed a single layer of microcolonies; MG1655 not only formed a few layers of microcolonies, but also started to produce extracellular polymeric substances (EPS); the number of attached microbial cells of O157:H7 is relatively less. After 48 h, the biofilms of all three bacteria were more mature: Nissle 1917 biofilms consisted of several layers of microbial cells and some EPS structures; MG1655 biofilms formed stable three-dimensional communities with channels inside; O157:H7 biofilms had fewer layers and less EPS compared to the other two. All the three E. coli strains formed complex bacterial communities in biofilm structures, through which they may improve their quorum sensing among microbes and increase their tolerance to survive in harsh environments (Annous et al., 2009).

The previous results indicated the differences in biomass and morphology between the three biofilms, to further explore the metabolic phenotypic differences, their chemical components were thoroughly analyzed. FTIR spectroscopy has been widely used to characterize the fatty acid, protein, polysaccharide and nucleic acid compositions in biomaterials without any complex sample preparation procedures (Cheeseman et al., 2021; Chirman and Pleshko, 2021; Sportelli et al., 2022; Wang et al., 2023b), so it was further applied to find out the biochemical molecular component differences between the three E. coli biofilms. MG 1655, Nissle 1917 and O157:H7 biofilms were imaged by FTIR microspectroscopy, and 140 individual IR absorption spectra of each biofilm were collected. The averaged spectra of each biofilm are shown in Figure 4A, with eight main absorption peak positions indicated on the graph. It can be seen that the absorption peaks of three biofilms were different, to find out whether the difference was statistically significant, all the 140 IR spectra of each biofilm were statistically analyzed, the results are shown in Figure 4B. Eight absorption bands were identified: 2,970–2,950 cm⁻¹ (C-H asymmetric stretching vibrations of methyl groups in the fatty acids), 2,945–2,920 cm⁻¹ (C-H asymmetric stretching vibrations of methylene groups in fatty acids), 1,657–1,650 cm⁻¹ (amide I of α-helical structures in proteins), 1,550–1,530 cm⁻¹ (amide II in proteins), 1,470–1,440 cm⁻¹ (C-H deformations of methylene groups), 1,410–1,385 cm⁻¹ (C=O symmetric stretching vibrations of COO⁻), 1,255–1,235 cm⁻¹ (P=O asymmetric stretching vibrations of >PO₂⁻), 1,105–1,075 cm⁻¹ (P=O symmetric stretching vibrations of >PO₂⁻) (Lasch and Naumann, 2015). The results showed that Nissle 1917 had significantly different absorption peaks with MG1655 and O157:H7 at 2934.1 cm⁻¹, indicating that the methylene groups in the fatty acids of the Nissle 1917 biofilm had different components from the other two biofilms. Similarly, O157:H7 had significantly different amide I components compared to Nissle 1917 and MG1655 biofilms. So far, we do not know the biological significance of the different chemical components between these strains. However, the infrared data provide useful clues to understanding the role of chemical molecules in the biofilm formation and probiotic properties. To the best of our knowledge, this is the first report on the comparison of microbial biofilms between different strains of the same species.

In order to visualize the differences in the metabolism of the three biofilms in a more intuitive way, a PCA based on whole molecular components (3,000–2,800 cm⁻¹, 1,800–900 cm⁻¹) was also performed, see Figure 4C. It can be seen that the three biofilms can be completely distinguished, indicating that the biochemical molecular components are significantly different. The loading plot of the PCA was shown in Figure 4D, in principle component 1 (PC1), the bands around 1,122 cm⁻¹ (corresponding to polysaccharides) contributed the most, while in PC2 the bands around 1,479 cm⁻¹ (corresponding to fatty acids), 1,085 cm⁻¹ (corresponding to nucleic acids) and 1,016 cm⁻¹ (corresponding to polysaccharides) contributed the most. To find out which biomolecules differed between the three biofilms, PCA was performed on three spectral regions: fatty acid (3,000–2,800 cm⁻¹), protein (1,800–1,500 cm⁻¹) and polysaccharide & nucleic acid (1,200–900 cm⁻¹), respectively (Figure 4E). It can be seen that, Nissle 1917 biofilm has significantly different fatty acid components, O157:H7 has significantly different nucleic acid and polysaccharide components, while their protein component shows partial similarities. Probiotic Nissle 1917 biofilms may secrete specific phospholipid components into the extracellular matrix, which may play an important role in establishing a healthy microenvironment in the human gut; whereas pathogenic O157:H7 biofilm may secrete specific exopolysaccharides and extracellular nucleic acids into the extracellular matrix to produce its toxicity and pathogenicity (Soliemani et al., 2022). Further analysis and identification of these intracellular or extracellular metabolites, for example by using mass spectrometry (MS), will help to elucidate the network of different metabolic pathways during biofilm formation.

The spatial distribution characteristics of the chemical composition of the three E. coli biofilms were further characterized by FTIR microspectroscopy; for each biofilm, the optical image and corresponding IR absorption images based on fatty acid, protein and nucleic acid are shown in Figure 5A. It can be seen that the spatial distribution of the biomolecules was inhomogeneous. The spatial heterogeneity of the biofilms was then evaluated quantitatively (Figure 5B). On the histogram, row 1 column 1, row 2 column 2 and row 3 column 3 indicated the within-group heterogeneity of Nissle 1917, MG1655 and O157:H7 biofilms, respectively; row 2 column 1 indicated inter-group heterogeneity of Nissle1917-MG1655, row 3 column 1 indicated Nissle1917-O157:H7, row 3 column 2 MG1655-O157:H7. The results showed that the inner-group heterogeneity of Nissle1917 and O157:H7 was relatively low, with a relatively sharp peak shape and peak positions less than 1. This proved that the spatial distribution of the biomolecules of the two biofilms was relatively uniform. The inter-group heterogeneity between Nissle1917-O157:H7 was greater than that between Nissle1917-MG1655 and MG1655-O157:H7, suggesting that the biochemical differences of biofilms between probiotic E. coli Nissle 1917 and pathogenic O157:H7 were much greater than those between Nissle 1917 and opportunistic MG1655. The clear biochemical heterogeneity between the three stains supports the idea that the spatial structures of the biofilms may be related to their biological activity. Further experiments are in progress.