Combined pH ratiometry and fluorescence lectin-binding analysis (pH-FLBA) for microscopy-based analyses of biofilm pH and matrix carbohydrates

Stimulated saliva was collected and pooled from 38 healthy volunteers after written informed consent was obtained. The collection of pooled saliva was considered exempt from ethical review by the Ethical Committee of Region Midtjylland (1-10-72-109-23). All donors refrained from eating or drinking for at least 30 min before saliva collection. The volunteers were asked to chew on a paraffin tablet (Sigma-Aldrich, Brøndby, Denmark) to stimulate saliva flow, and the produced saliva was collected in centrifuge tubes (Corning Life Sciences, Corning, NY, USA). Half of the collected saliva was centrifuged for 5 min at room temperature (RT) at 1,150 × g and pooled and stored at −80°C in a 1:1:1 mix of saliva:glycerol:phosphate-buffered saline (PBS). The remaining saliva was centrifuged for 1 min (RT) at 1,150 × g and used for the preparation of sterile saliva, as detailed elsewhere (21). Briefly, the saliva samples were pooled, diluted 1:1 with demineralized sterile water, and mixed with dithiothreitol (BDH Biochemicals, Poole, UK) to a final concentration of 2.5 mmol. Subsequently, the samples were centrifuged at 48,000 × g for 10 min (4°C), sterilized by ultrafiltration using 0.45 µm pre-filters (Jet Biofil, Guangzhou, China) and 0.22 µm filters (Jet Biofil), and stored at −20°C.

For in vitro biofilm growth, the channels of flow cells with a non-fluorescent polymer coverslip bottom (uncoated μ-slide VI, Ibidi, Munich, Germany) were inoculated with 10 µL of whole saliva in 90 µL of medium that contained 80% sterile saliva, 20% fetal bovine serum, and either 1% sucrose (SUC) or no sucrose (NoSUC). Each channel had the following dimensions: 0.4 mm (height), 17 mm (length), and 3.8 mm (width). Biofilms were grown anaerobically under static conditions at 37°C. After 24 h of growth, SUC biofilms were too thick for confocal microscopy examination and FLBA. Therefore, SUC biofilms were grown for 16 h and NoSUC biofilms for 24 h, which resulted in similar biofilm thicknesses for both growth conditions. After biofilm growth, the flow cells were immediately subjected to confocal microscopy analysis. All experiments were performed in biological triplicates. The experimental procedures are illustrated in Fig. S1.

The extracellular pH in the biofilms was monitored using the ratiometric pH-sensitive dye C-SNARF-4 (Thermo Fisher Scientific, Roskilde, Denmark) (2). The calibration of the dye was performed using MES buffer solutions (50 mM; Sigma-Aldrich) titrated to pH 4.0–8.0 in steps of 0.2 pH units. Buffer solutions were mixed with C-SNARF-4 in the wells of a 96-well plate to a final dye concentration of 30 µM and allowed to equilibrate for 5 min at 35°C. An inverted confocal microscope (Zeiss LSM 700 Axio Observer, Jena, Germany) coupled with a 63×/1.40 oil-immersion objective (alpha Plan-Apochromat, Zeiss) was used for image acquisition. The microscope was set to a pinhole size of 1.76 AU (1.3 µm optical slice thickness), an image size of 512 × 512 pixels (101.61 × 101.61 µm), a pixel dwell time of 12.61 µs, zoom of 1×, and an 8-bit intensity resolution. Images of the buffer solutions were acquired at a 5 µm distance from the bottom of the wells in three randomly chosen fields of view (FOVs) using an excitation wavelength of 555 nm and simultaneous detection from 300 to 618 nm (green channel) and 618 to 800 nm (red channel). The detector gain for both channels was kept constant throughout the experiments. Additionally, images with the laser turned off were acquired for each pH value to correct for detector offset (background). All red and green channel images were exported to ImageJ (22) as TIF files, the background was subtracted, and the fluorescence intensity ratios between the two channels (green/red) were calculated. Finally, the intensity ratios (r) were plotted against the respective pH values and fitted to a symmetrical sigmoidal curve (R² = 0.9998) using online software (www.mycurvefit.com) (Fig. S2). The resulting equation (1) allowed the conversion of fluorescence emission ratios to pH values:

The flow cell was mounted onto the microscope sample holder (universal mounting frame K, Zeiss), and the medium was carefully removed by inserting filter paper strips (8 × 1.2 mm; Melitta, Minden, Germany) in the outlet of each channel. Thereafter, the biofilms were washed with 100 µL of sterile saline (pH 7.0). The biofilm pH response to sucrose was monitored in 100 µL of saline (pH 7.0) containing 0.4% (wt/vol) sucrose and C-SNARF-4 (30 µM). For each biofilm, five FOVs, at least 1,000 µm apart from each other, were selected at random and marked in the microscope software (Zen 2012 SP5, Zeiss). Images were acquired at the bottom of the biofilms, 5 min (T1) and 15 min (T2) after the addition of sucrose, with the same microscope settings as during pH calibration.

Glycoconjugates and polysaccharides in the biofilms were visualized by fluorescence lectin binding analysis (FLBA using Aleuria aurantia lectin (AAL; SMS Gruppen, Rungsted Kyst, Denmark) and Morus nigra agglutinin G (Morniga-G/MNA-G; SMS Gruppen), both labeled with fluorescein isothiocyanate (FITC). AAL specifically binds fucose-containing carbohydrate components, whereas MNA-G targets galactose in the biofilm matrix (Table S1). After pH measurements, the ratiometric dye C-SNARF-4 was gently removed from the flow cell channels using absorbing filter paper strips. The biofilms were then washed two times with 100 µL of PBS (pH 7.4) to remove the remaining dye and minimize background fluorescence signals. Thereafter, the biofilms were incubated with 100 µL of the respective FITC-labeled lectin (100 µM) for 30 min (RT), washed once with PBS, and counterstained for 10 min with SYTO 60 (10 µM; Molecular Probes, Thermo Fisher, Carlsbad, CA, USA) to visualize microbial cells. To investigate the relationship between the biofilm pH profiles and the abundance and spatial distribution of matrix carbohydrate components at the microscale, FLBA images were acquired in the same FOVs previously imaged during pH ratiometry. In each FOV, six-sliced z-stacks spanning the height of the biofilms were acquired, both with zoom 1× (512 × 512 pixels; 101.61 × 101.61 µm) and zoom 0.5× (512 × 512 pixels; 203.23 × 203.23 µm). Therefore, carbohydrate components could be quantified within the biofilm areas investigated with pH ratiometry and also in the immediate surroundings. Images were acquired with a pixel dwell time of 2.5 µs, and the pinhole size was set to 1.6 AU (1.3 µm optical slice). FITC-labeled lectins were excited at 488 nm and detected from 300 to 628 nm. SYTO 60 was excited at 639 nm and detected from 644 to 800 nm. Laser power (488 nm) and detector gain were kept constant for each fluorescently labeled lectin and across all biofilm samples. Experiments were carried out in biological triplicates.

Extracellular pH in the biofilms was determined as described previously (23). Briefly, the C-SNARF-4 images were imported into the software daime (digital image analysis in microbial ecology, v. 2.2) (24) and segmented to remove microbial cells. In ImageJ (22), the green and red channel images were divided by each other to obtain the fluorescence intensity ratios for all extracellular areas. The average ratios (±SD) were calculated for each FOV and converted to pH values according to equation (1).

The biovolumes of microbial cells and lectin-targeted carbohydrates in the biofilms were estimated in each six-sliced z-stack using the software daime (19). First, all green- and red-channel images were imported into daime as TIF files, and the areas covered by microorganisms (red) and carbohydrates (green) were identified using intensity threshold-based segmentation. The microbial and carbohydrate biovolumes in the bottom layer (slices one and two), in the middle layer (slices three and four), and at the top of the biofilms (slices five and six), as well as the total biovolumes (all six slices), were determined for each FOV using the Cavalieri principle (25), by multiplying the interslice distance with the respective areas covered by microbial cells or lectin-targeted carbohydrates. The carbohydrate biovolumes in each z-stack were quantified as a percentage of the total number of voxels and as normalized biovolumes (% of the microbial biovolumes). The digital image analysis procedures are illustrated in Fig. S3 and S4.

For bacterial 16S rRNA sequencing, biofilms were grown in flow cells as previously described and then removed by sonication for 10 min at 37°C in an ultrasonic cleaner (Branson 5510, Branson Ultrasonics Co., Danbury, CT, USA). The collected material from technical duplicates was pooled, and DNA extraction was performed using the DNeasy PowerLyzer PowerSoil DNA Isolation Kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. The V3–V4 region of bacterial 16S ribosomal RNA genes was amplified using primers Bac 314F and Bac 805R (26), as described previously (20). Paired-end sequencing (2 × 300 bp) was done on an Illumina MiSeq sequencer using the V3 sequencing kit (Illumina, Inc., San Diego, CA, USA). Experiments were performed in biological triplicates.

All sequence analyses were performed in R v. 4.3.0 (27). Sequences were trimmed to remove barcodes and primers using cutadapt v. 0.2.0 (28). Error correction, amplicon sequence variant (ASV) calling, chimera removal, and taxonomic classification were performed with the R package “DADA2” v. 1.27.1 (29). The classification was done according to the Silva SSU reference database nr. 138 (30). ASVs were filtered to only include those classified as bacteria. Nucleic acid extraction blanks and PCR negatives were used for decontaminating the data using the R package Decontam v. 1.19.0 (31). Contaminants were found using the prevalence method with a threshold of 0.1 and subsequently removed from the data.

Differentially abundant genera between the SUC and NoSUC treatments were identified using the R package ANCOM-BC2 v2.1.4 (32). The R packages Phyloseq v. 1.43.0 (33), Microbiome v1.21.0 (34), vegan v2.6.4 (35), ggplot2 v3.4.2 (36), and custom R scripts were used for all further analyses. For calculating alpha diversity (Shannon index), the number of reads was subsampled to a common read depth of 33,555 reads. All other analyses were performed without subsampling. Relevant ASVs were blasted against the extended Human Oral Microbiome Database (http://www.homd.org) to obtain a species-level classification, and only sequences with 100% identity were retained (Table S2). ASV tables with read counts, relative abundances, and relevant metadata can be found in Table S3. Raw sequences were submitted to NCBI under the BioProject PRJNA1004439.

Differences in average biofilm pH after 5 and 15 min of sucrose challenge were compared using paired t-tests, while differences in pH, cell density, and biofilm thickness between SUC and NoSUC biofilms were assessed using unpaired t-tests. pH variances at the FOV level were compared between groups using the F-test. The lectin-stained biovolumes in SUC and NoSUC biofilms were compared using paired t-tests. Correlations between biofilm pH and biofilm thickness were assessed using Pearson or Spearman correlation coefficients. Data were checked for normal distribution and homogeneity of variance using the Shapiro–Wilk and Levene tests, respectively. Statistical analyses were performed with the software R v. 4.3.0 (27) and GraphPad Prism v. 10 (GraphPad Software Inc., San Diego, CA, USA) with the level of significance set at α = 0.05.

Biofilm formation in both the presence and absence of sucrose was robust. The biofilms exhibited areas of high and low cell density, and thick protruding cell clusters were often present. Sixteen-hour SUC and twenty-four-hour NoSUC biofilms were similar in thickness (P = 0.11) and cell density (P = 0.49), with average thicknesses (±SD) of 31.76 (±3.66) µm and 26.46 (±6.22) µm and average cell densities (% pixels) of 15.13 (±9.2) and 13.61 (±7.8), respectively. If grown for 24 h, SUC biofilms increased in thickness to 157.37 (±53.22) µm, which prevented FLBA imaging of the biofilms due to the limited lectin penetration (data not shown).

Biofilm composition for both treatment groups was highly reproducible across biological replicates (Fig. 1). SUC biofilms were dominated by Streptococcus spp. (mean relative abundance of 96%); all other bacterial genera had mean relative abundances below 0.6%. NoSUC biofilms were more diverse than SUC biofilms (Shannon indices of 3.3 ±0.09 SD and 1.6 ±0.05 SD, respectively). The most abundant genera in NoSUC biofilms were Streptococcus spp. (mean relative abundance of 32%) and Porphyromonas spp. (mean relative abundance of 28%), with lower contributions from Fusobacterium spp., Haemophilus spp., and Gemella spp. (mean relative abundances of 5%–10%). The microbial composition differed markedly between groups, with the genera Rothia and Streptococcus being significantly more abundant in SUC biofilms (P < 0.05) (Fig. 2). Interestingly, the presence of sucrose during growth also shifted the dominant streptococcal ASV in the biofilms from the Streptococcus mitis/oralis group to the Streptococcus salivarius group (Fig. 1; Table S2).

The washing of the flow cells with PBS after pH recordings successfully removed C-SNARF-4 from the microbial cells and the extracellular matrix, and no remaining background fluorescence signals interfered with the subsequent FLBA (Fig. S5). The carefully executed washing procedures and the protective geometry of the flow cell preserved the biofilm structure through the sequential staining and washing steps, such that identical FOVs could be re-identified and re-imaged (Fig. 3).

The extracellular pH of both SUC and NoSUC biofilms dropped significantly after 5 and 15 min of sucrose challenge (P < 0.001) (Fig. 4A). The average pH (±SD) for NoSUC biofilms was 6.56 (±0.15) after 5 min and 6.38 (±0.15) after 15 min. Biofilm pH drops were more pronounced for SUC compared to NoSUC biofilms at both time points, with pH values dropping to 6.28 (±0.14) (P = 0.01) and 6.15 (±0.13) (P = 0.02) after 5 and 15 min, respectively. pH varied moderately between different FOVs inside the biofilms, with differences ranging from 0.05 to 0.54 pH units between the least and most acidogenic FOVs. In NoSUC biofilms, pH variance was significantly more pronounced than in SUC biofilms (0.017 ±0.011 SD vs 0.007 ±0.002 SD; P = 0.004). In both groups, biofilm pH did not correlate with biofilm thickness (NoSUC: r = −0.16, P = 0.41; SUC: r = −0.9; P = 0.65). Figure 4B shows the pH developments over time in two FOVs representative of SUC and NoSUC biofilms.

The biovolumes visualized by fluorescently labeled lectins were quantified and normalized against microbial biovolumes in the biofilms (Fig. 5A). Overall, both lectins stained high biovolumes, particularly in sucrose-grown biofilms, where they clearly exceeded the microbial biovolume. Both AAL and MNA-G stained significantly larger biovolumes (±SD) in SUC biofilms compared with NoSUC biofilms (206.1% ±82.0 vs 70.0% ±44.8 and 453.2% ±56.0 vs 153.4% ±133.8, respectively; P < 0.05). The biovolumes stained by MNA-G were significantly larger than those visualized by AAL in SUC (P = 0.02), but not in NoSUC biofilms (P = 0.40).

In addition, we evaluated the distribution of biovolumes (microbial cells and matrix carbohydrates) across biofilm layers (Fig. 5B). The proportion of lectin-stained biovolumes compared to microbial biovolumes was increased in SUC biofilms compared to NoSUC biofilms for both MNA-G and AAL. In SUC biofilms, MNA-G was distributed evenly across the layers of the biofilms, whereas AAL was more abundant in the middle and top layers. In NoSUC biofilms, both AAL and MNA-G were progressively more abundant towards the bottom of the biofilms.

The relationship between biofilm pH and the amount of lectin-stained carbohydrates in SUC and NoSUC biofilms at the FOV level for each biological replicate is shown in Fig. S6.