Microbial Interactions Shape Spatial Organisation and Transcriptional Responses in a Model Mixed-Species Biofilm

Abstract

Dynamic social interactions within bacterial biofilms drive distinct spatial organisation and transcriptional responses. Here, we combine fluorescence in situ hybridisation (FISH), confocal laser scanning microscopy (CLSM), and RNA sequencing (RNA-Seq) to investigate a model three-species biofilm community derived from a dairy pasteuriser, comprising Stenotrophomonas rhizophila, Microbacterium lacticum, and Bacillus licheniformis. CLSM revealed species-specific biovolume dynamics and stratified 3D structures over 24 h, with S. rhizophila as the dominant species and M. lacticum exhibiting the lowest abundance yet playing an essential role as the initial coloniser. Spatial patterns reflected known pairwise interactions – commensalism, exploitation, and neutral interaction. Transcriptomic profiling of S. rhizophila revealed extensive gene expression changes in dual-species biofilms with M. lacticum, including upregulation of genes related to flagellar motility, nutrient acquisition, energy metabolism, and TonB-dependent transport. In contrast, co-culture with B. licheniformis induced minimal transcriptional changes in S. rhizophila, consistent with a neutral interaction among the two. Our findings demonstrate how interspecies interactions govern both spatial topology and functional specialisation in mixed-species biofilms which is of relevance to microbial ecology, industrial biofilm control, and the targeting of keystone biofilm species.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00248-026-02701-w.

Introduction

A bacterial biofilm is a surface-associated or free-floating, aggregated multicellular community of bacteria embedded in a self-produced or shared matrix of extracellular polymeric substances (EPS), where bacterial growth and activity are impacted by chemical gradients and intercellular interactions, and augmented by environmental factors. Most biofilms in natural and industrial settings are composed of multiple bacterial species with varying metabolic capacities that interact with each other and the environment [1].

The confined arrangement of bacterial cells within biofilms facilitates metabolic exchanges [2], increases access to nutrients [3], and protects cells from antimicrobials [4–6], phage infections, and predation [7–10]. Biofilm structure and mechanical properties facilitate dynamic social interactions between different cells in biofilms, ranging from positive (e.g., cooperation [+/+] and commensalism [+/0]) to neutral and negative interactions, including competition [−/−], amensalism [0/−], and exploitation [+/−] [4, 11].

The biogeography, i.e. the spatial organisation of bacterial communities in mixed-species biofilms, is shaped by multiple factors, including cellular properties (e.g., size, adhesion), cell growth and motility, the structural characteristics of the extracellular polymeric substances (EPS), and the surrounding chemical and physical environment [3, 12].

In well-mixed systems, cellular interactions occur uniformly across the population. However, in spatially structured systems, the strength and impact of these interactions is influenced by the spatial proximity of cells and the diffusion range of exchanged chemicals [2, 3, 13]. On the other hand, bacterial interactions also govern spatial organisation in biofilms, where bacteria develop metabolic or physical dependencies, grow in aggregates, and create strong gradients of oxygen, pH, and metabolic substrates [14, 15].

Understanding the mechanisms of bacterial interspecies interactions and their spatial organisation in biofilms has significant implications for strategies to manipulate biofilm formation and behaviour. Advancements in imaging technologies have significantly improved our understanding of microbial interactions within multispecies biofilms. Microscale spatial organisation analysis by using fluorescence in situ hybridisation (FISH) offers an efficient approach to unravel biofilm dynamics, including spatial positioning and co-localisation patterns [16]. Meanwhile, RNA sequencing offers specific insights into functional dynamics, gene expression patterns, and metabolic interactions [17]. Recently, emerging spatial transcriptomic approaches such as PAR-seqFISH have demonstrated potential to bridge the gap between spatial imaging and transcriptional profiling, enabling simultaneous mapping of gene expression and microbial spatial arrangement at high resolution [18].

Model biofilms comprising a limited number of representative species provide experimentally tractable systems for studying spatial organisation in relation to interspecific interactions [19, 20]. Liu et al. [21] explored how bacterial interactions influence spatial organisation in multispecies biofilms. These findings indicate that increased biomass production in a four-species biofilm is an inherent community trait, arising from precise spatial optimisation facilitated by a balance of competitive and cooperative interactions.

Here, we link gene expression changes and spatial organisation patterns in a model three-member community to their previously characterised interspecies interactions, providing ecological and functional insights into these interactions. The community comprises strains of Microbacterium lacticum, Stenotrophomonas rhizophila, and Bacillus licheniformis, isolated from the surface of a dairy pasteuriser following cleaning and disinfection (C&D) [22, 23]. These three species were selected because they were among the isolates recovered from pasteuriser surfaces after C&D and, in our previous work, showed the greatest increases in biofilm biomass when tested within four-species combinations. Their consistent synergistic behaviour provides a robust and experimentally tractable model for dissecting interspecies interactions [22, 23]. A synergistic ~ 2.7-fold increase in biofilm biomass production was observed in the mixed-species biofilm community compared with the combined biofilm mass of individual species [11]. M. lacticum was identified as the potential keystone species, essential for the enhanced growth on stainless steel and plastic surfaces and biofilm formation of the other two species. Our previous studies showed that S. rhizophila benefitted in terms of enhanced growth, increased biofilm biomass, and protection from C&D in the presence of M. lacticum in a dual-species biofilm [4, 11]. B. licheniformis engaged in an exploitative interaction with M. lacticum. Here, we use the term exploitative interaction to describe a +/− relationship in which the growth of one organism increases in co-culture, while the partner shows reduced cell counts compared with their monocultures. In our previous work, B. licheniformis cell numbers increased, whereas M. lacticum cell numbers decreased in dual-species biofilms relative to monocultures [11]. Together, these interactions contributed to S. rhizophila becoming the dominant community member in the three-species biofilm with M. lacticum and B. licheniformis. Consistent with this pattern, S. rhizophila was also the most abundant species in a related four-species biofilm system that included M. lacticum, B. licheniformis, and either Kocuria salsicia or Calidifontibacter indicus [4, 11].

To investigate how interspecies interactions shape biofilm structure and transcriptional responses, we used RNA-Seq to profile the global gene expression of S. rhizophila in dual- and three-species biofilms relative to their monoculture biofilms. In parallel, we employed confocal imaging to resolve the spatial organisation of the three species in different combinations. These approaches were used to examine how community composition influences metabolic and structural adaptations within mixed-species biofilms.

Materials and Methods

Bacterial Strains and Growth Condition

Three bacterial species were used in this study: S. rhizophila (strain B68), B. licheniformis (strain B65), and M. lacticum (strain B30), which were previously isolated from the surface of a dairy pasteuriser following routine C&D [22]. For simplicity, throughout this study, the strains are referred to as S. rhizophila (SR), B. licheniformis (BL), and M. lacticum (ML). Hereafter, all species will be represented by the above-mentioned abbreviations when referring to combinations, and full names will be used when referring to individual species. The strains were cultured in Brain Heart Infusion (BHI) medium (BioRad, 3553664) at 30 °C and stored at − 70 °C as freezer stocks in 20% glycerol.

Biofilm Formation on Plastic Coupons

Bacterial cultures were incubated overnight (~ 16 h) under static conditions in 10 mL of BHI broth at 30 °C. The cultures were then appropriately diluted in fresh BHI to an OD595 value of 0.05 for all strains, as measured using a Multiskan™ FC Microplate Photometer (GENESYS 10 UV-Vis Spectrophotometer, Thermo Fisher Scientific). A total of 4 mL of the diluted and standardised overnight culture was added to each well of a six-well microtiter plate, each containing one sterile polycarbonate coupon with dimensions of 1 cm². For mixed-culture biofilms, equal volumes of the adjusted overnight culture from each of the three species were combined to reach a final volume of 4 mL in total and added to the wells. The microtiter plates were then incubated under static conditions at 30 °C for the required time periods (6 h, 12 h, 18 h, and 24 h). After the incubation period, the coupons were carefully removed with forceps and washed three times by submerging in sterile phosphate-buffered saline (PBS) to remove any loosely attached cells before fixing the samples for quantitative analysis. All biofilm experiments were performed using three independent biological replicates for each treatment and at each time point. For confocal microscopy, three distinct imaging sites were selected from each coupon per biological replicate. Quantitative measurements were calculated as mean ± standard deviation (SD).

Complementary Planktonic Co-Culture Assays

Complementary planktonic co-culture assays were performed for monocultures of S. rhizophila, M. lacticum and B. licheniformis, and for co-cultures SR + ML and BL + ML. We compared 24 h CFU counts between monoculture and mixed culture to assess whether the interaction effects were intrinsic to planktonic growth or arose from niche-specific mechanisms associated with biofilm organisation. Full experimental procedure is provided in Supplementary material S1.

16S rRNA Fluorescence in situ Hybridisation (FISH) and Confocal Laser-Scanning Microscopy (CLSM)

Bacterial biofilms were initially fixed with 4% (w/v) paraformaldehyde in PBS (pH 7.2) for 4 h at 4 °C. Following fixation, the coupons containing biofilms were resuspended in 200 µL of 1x PBS. To enhance the permeability of the bacterial cell wall, the coupons were immersed in a 1 mg/mL lysozyme solution (Cat. No. L6876, Sigma-Aldrich) and incubated for 10 min at room temperature. This was followed by two washes with PBS to remove any residual lysozyme. The samples were then dehydrated using a series of ethanol solutions (ethanolic series: 50%, 70%, 100%) for 3 min each at room temperature. Next, the samples were incubated with a hybridisation buffer containing 0.9 M NaCl, 20 mM Tris/HCl (pH 7.2), 30% (v/v) formamide, 0.01% (w/v) SDS, and fluorochrome-labelled oligonucleotide probes at a working concentration of 5 ng/µL at 46 °C for 3 h. The samples were then rinsed twice with 1 mL of pre-warmed (48 °C) washing buffer containing 20 mM Tris/HCl (pH 7.2), 5 mM Ethylenediaminetetraacetic acid (EDTA, pH 7.2), and 102 mM NaCl, followed by incubation in washing buffer for 15 min at 48 °C. The washing buffer was gently removed by pipetting, and the samples were rinsed with 1 mL of ice-cold dH₂O to eliminate residual salts derived from the washing buffer.

The three oligo probes were synthesised commercially by Eurofins Genomics (Eurofins Scientific, France) and 5’ labelled with three different fluorochromes: Cyanine 5 (Cy5) for S. rhizophila, Cyanine 3 (Cy3) for B. licheniformis, and 6-carboxyfluorescein (FAM) for M. lacticum. Details of the probes (including sequences) can be found in Supplementary Material S1. The composition of the hybridisation mixture was as follows: 18% (v/v) 5 M NaCl, 2% (v/v) 1 M Tris-HCl (pH 7.2), 30% (v/v) formamide, 0.2% (v/v) of a 2% (w/v) SDS solution, 1% (v/v) of the probes, with the remaining volume made up by water. The maximum excitation/emission wavelengths and actual emission wavelength ranges for Cy5, Cy3, and FAM were detected in four separate channels using a flexible detector (GaAsP-PMT) [4].

Confocal Image Processing and Quantification Analysis

All FISH images were captured using a CLSM (LSM 800, Zeiss) with a Plan-Apochromat 63x/1.4 oil-immersion objective. Each image contained 1024 × 1024 pixels, corresponding to physical dimensions of 101.4 × 101.4 μm. Z-stacks with 0.5 μm per layer were recorded using Axiocam 503 mono to obtain 3D images. Data were generated from three biological replicates at each time point, with each replicate comprising three images acquired from independent regions of the biofilm (n = 9). Acquired images in CZI format were initially processed using FIJI (ImageJ, version 2.1). The images first underwent noise reduction to improve signal clarity and accuracy during subsequent analysis. Following noise reduction, the images were converted to binary format and then separated into three distinct channels based on species-specific probes (red, yellow, and green) used. Thresholds for each channel were set using the “Otsu” algorithm [24]. To remove potential overlapping signals from the red fluorescent probes detected in the green channel, a ‘subtract’ algorithm in RCon3D was applied before quantification analysis. To quantify biomass volume, the number of voxels containing bacterial fluorescent signals was calculated across all channels. The biomass volume (in µm³) was determined by multiplying the voxel size (0.005 μm³ per voxel for FISH images) by the number of voxels containing bacterial signals across all 3D confocal images. Biomass volume quantification was performed using the R statistical programming language (version 4.3.0). The algorithms for absolute bio-volume quantification were implemented using the “quant” function in the RCon3D package (version 1.2.6), which is available on GitHub (www.github.com/Russel88/RCon3D version 1.2.6) and previously described [25]. To analyse the spatial location of each species, the biofilms’ three layers were identified using the ‘layer_split’ program: top layer (from the very top to the section comprising up to 45% of the stacks), middle layer (the middle 45% of the stacks), and bottom layer (the remaining 10% of the stacks). Pairwise abundance correlations between bacterial species were calculated using species-specific biovolume data extracted from individual biofilm layers at four distinct time points. These correlations were computed using the Pearson method implemented in the ggpubr package in R. Although the correlation analyses were based on data from separate layers, the abundance values were later pooled across all layers to derive composite species abundances per time point.

RNA-Seq Analysis of Biofilms

Mono- and mixed-species biofilms were grown in the same manner as on plastic coupons, on stainless steel (SS) coupons (AISI 304 grade) under static conditions for 24 h in BHI medium, as described previously in our work [11]. RNA-seq data were generated from three biological replicates collected on three independent occasions; for each biological replicate, biofilms from three wells (each containing an SS coupon) were pooled prior to RNA extraction. Prior to use, the SS coupons were degreased using a general laboratory detergent, thoroughly rinsed with distilled water, air-dried, and then sterilised by autoclaving at 121 °C for 15 min. After incubation, the coupons containing biofilms were gently washed with PBS to remove loosely attached cells. The coupons containing biofilms were then immediately immersed in 5 mL of RNAprotect Tissue Reagent (Qiagen, Germany) for 10 min to inactivate RNases and stabilise RNA. Total RNA was extracted from the biofilms suspended in PBS using the DNeasy PowerBiofilm Kit (Qiagen, Germany), following the manufacturer’s protocol. RNA degradation and contamination were assessed using 1% agarose gels, while RNA purity was evaluated with a NanoPhotometer^® spectrophotometer (IMPLEN, CA, USA). Sequencing libraries were prepared using the NEBNext^® Ultra™ RNA Library Prep Kit for Illumina^® (NEB, USA) according to the manufacturer’s instructions. After cluster generation, the library preparations were sequenced on an Illumina platform and paired-end reads were generated. Sequencing was performed as paired-end 150 bp runs (2 × 150 bp) on an Illumina NovaSeq 6000 platform. Raw FASTQ data were initially processed using fastp [26]. During this step, clean reads were obtained by trimming adapter sequences, removing poly-N sequences, and filtering out low-quality reads. Index codes were added to uniquely assign sequences to each sample. Both reference genome indexing and alignment of clean reads to the reference genome were performed using Bowtie2 [27]. A summary of total reads generated, mapped reads per genome, unique mapping percentages, and discarded reads for each sample is provided in Supplementary Table S1. FeatureCounts was used to quantify reads mapped to each gene [28]. Gene expression levels were then normalised for gene length and sequencing depth by calculating Fragments Per Kilobase of transcript per Million mapped reads (FPKM), providing a standardised measure of expression. Differential gene expression analysis was performed using DESeq2. For DESeq2, a negative binomial model was used, and adjusted p values (< 0.05, Benjamini-Hochberg) were applied to identify differentially expressed genes (DEGs). Genes with an adjusted p value < 0.05 and log₂ fold change > 0 were considered differentially expressed. Average gene expression values for pathways were calculated by taking the arithmetic mean of expression values for all genes assigned to a specific pathway. Only pathways with at least three genes were included in the analysis. Using the STRING online database (https://string-db.org), the protein-protein interaction (PPI) network was constructed with an interaction score > 0.4, based on annotations against the S. rhizophila database. The network was then visualised using Cytoscape (v3.8.2) [29].

Results

Specificity of the Primers

Species-specific FISH probes were designed through in silico analysis and validated experimentally. Probe performance was confirmed using mixed planktonic cultures and monoculture biofilms grown for 24 h. Each probe exclusively detected its target species without observable cross-hybridisation and produced minimal background fluorescence under optimised hybridisation conditions. These results demonstrate reliable target specificity and support the suitability of the probes for subsequent spatial analysis of multispecies biofilms (Supplementary Fig. S1).

Temporal Biovolume of the Three-Species Biofilm

The 3D visualisation of the three-species biofilm revealed distinct spatial organisation across time points (Fig. 1a). The biovolume of each species in a mixed biofilm was quantified at four time points: 6 h, 12 h, 18 h, and 24 h. At 6 h, S. rhizophila had the lowest biovolume (1.4 × 10³ µm³), while M. lacticum had the highest (1.1 × 10⁴ µm³) (p = 0.003). There was no significant (p = 0.98) difference between the biomass of M. lacticum and B. licheniformis at 6 h. After 24 h, a significant shift in relative biovolumes was observed, with S. rhizophila exhibiting the highest biovolume in the community (2.8 × 10⁴ µm³), exceeding that of M. lacticum (4.3 × 10³ µm³). The biovolume of S. rhizophila was significantly (p < 0.001) higher (~ 6.4-fold) than that of M. lacticum at 24 h (Fig. 1b).

Fig. 1.

(a) 3D visualisation showing the spatial localisation of species at different time points (6, 12, 18, and 24 h): Stenotrophomonas rhizophila (Cyanine 5 (Cy5), red), Bacillus licheniformis (Cyanine 3 (Cy3), yellow), and Microbacterium lacticum (6-carboxyfluorescein (FAM), green). (b) Bar charts depicting the biofilm biomass volume (µm³) shift of each species in the mixed-species community at different time points. SR, BL, and ML represent S. rhizophila, B. licheniformis, and M. lacticum, respectively. Error bars shown in (b) and (c) represent mean ± standard deviation (n = 9). Data based on three biological replicates at each time point, each replicate including three images acquired from three independent areas (n = 9)

Spatial Distribution of Species in Different Biofilm Layers

Biofilm sectioning has been successfully applied to analyse the spatial localisation patterns of species in our previous study [30]. To investigate the specific spatial distribution of the three species in this study, biofilm layers were defined as 10% bottom, 45% middle, and 45% top, and the relative abundance of each species in each layer was compared (Fig. 2a). The 10% layers chosen as the bottom mainly include cells that directly attach to the surface, most likely reflecting initial colonisers. The progressive increase in biofilm thickness over time reflected a temporal accumulation of biomass, with the most substantial biofilm mass (4.5 × 10⁴ µm³) and pronounced stratification observed at 24 h (Fig. 2b). All species were present in all layers of the biofilm at different time points; however, M. lacticum remained dominant in the bottom layer throughout the 24-h period. In contrast, S. rhizophila and B. licheniformis were more evenly distributed across the biofilm. M. lacticum initially formed the bottom layer upon which S. rhizophila grew, suppressing the growth of M. lacticum. Over time, relative S. rhizophila abundance increased in the top layers, indicating a competitive advantage ensuring optimal positioning regarding access to nutrients (medium) and electron acceptors (oxygen). Relative B. licheniformis abundance decreased in the top layers over time, whereas M. lacticum relative abundance in the top layers did not change during the 24-h growth period. In particular, slight variability in the relative abundance of each species was observed within the same layer at different time points. This may result from the use of independent samples for confocal microscopy. Alternatively, it is most likely that the spatial organisation of multispecies biofilms changes dynamically, leading to varying relative abundances of each species over time.

Fig. 2.

(a) Relative abundance of three bacterial species: Stenotrophomonas rhizophila (SR, red), Bacillus licheniformis (BL, yellow), and Microbacterium lacticum (ML, green) in different layers (Top, Middle, and Bottom) of a mixed-species biofilm grown on polycarbonate coupons over 24 h. The biofilm was divided into layers with the bottom layer comprising 10% of the biofilm thickness and the middle and top layers each accounting for 45%. (b) Relative abundance of each species across the biofilm, from the bottom to the top, with each layer representing 0.5 μm in thickness. Species distribution is shown at different time points (6 h, 12 h, 18 h, and 24 h), showing an increase in biofilm thickness over time, with the 24-h biofilm being the thickest. The species were identified using the following fluorescent probes: Cyanine 5 (Cy5) for S. rhizophila (red), Cyanine 3 (Cy3) for B. licheniformis (yellow), and 6-carboxyfluorescein (FAM) for M. lacticum (green). Data based on three biological replicates at each time point, each replicate including three images acquired from three independent areas (n = 9)

Pairwise Relationships Between the Relative Abundances of Three Bacterial Species

To examine the co-occurrence of different bacterial species within the biofilm structure, we performed pairwise correlation analyses based on species-specific biovolume data within each vertical stratum of the biofilm (Fig. 3). S. rhizophila and B. licheniformis exhibited a positive co-occurrence pattern across layers at 6 h, although the correlations at the other three time points did not reach statistical significance (Fig. 3a). The relative abundance of M. lacticum decreased as the abundance of S. rhizophila increased within each vertical stratum of the biofilm (Fig. 3b). This inverse relationship suggests vertical separation between the two species, with S. rhizophila primarily occupying the upper biofilm layers above M. lacticum. This spatial segregation is further supported by negative Pearson correlation coefficients observed at all time points. Meanwhile, B. licheniformis and M. lacticum showed a consistent negative correlation, indicating an inverse co-existence pattern within the biofilm structure (Fig. 3c).

Fig. 3.

Correlations between the abundance of Stenotrophomonas rhizophila (SR), Bacillus licheniformis (BL), and Microbacterium lacticum (ML) in a three-species biofilm grown on polycarbonate coupons over a 24-h period. Part (a) shows the correlation between SR and BL abundance, part (b) shows SR versus ML abundance, and part (c) shows BL versus ML abundance. Different colours represent time points (purple: 6 h, blue: 12 h, orange: 18 h, red: 24 h), reflecting species co-existence across different biofilm layers at these time points. Linear regression lines with confidence intervals illustrate species interactions, with Pearson correlation coefficients (R) and p values provided in each plot. Each dot represents a pairwise measurement of the abundance of two species at a specific time point (6 h, 12 h, 18 h, or 24 h) across different layers of the biofilm. Data based on three biological replicates at each time point, each replicate including three images acquired from three independent areas (n = 9)

Complementary Planktonic Co-Culture Assays

Planktonic co-culture assays showed that the interaction patterns observed in biofilms were not reproduced under well-mixed suspension conditions. When co-cultured with M. lacticum, S. rhizophila did not show any increase in CFU relative to monoculture, and M. lacticum cell numbers remained unaffected, indicating a neutral interaction in liquid culture. Likewise, in the M. lacticum and B. licheniformis co-culture, neither species showed a statistically significant advantage or reduction in CFU at 24 h compared with their respective monocultures. These findings contrast with the biofilm results, where S. rhizophila and B. licheniformis exhibited fitness benefits in the presence of M. lacticum. Detailed data are provided in Supplementary Figure S2 (Supplementary Material S1).

RNA-Seq Analysis of S. Rhizophila Gene Expression

The transcriptional response of S. rhizophila was monitored using RNA-Seq in different biofilm combinations (SR-BL, SR-ML, and SR-BL-ML) and compared to its monoculture biofilm. We focused exclusively on S. rhizophila for transcriptomic analysis because it emerged as the dominant species in the mixed-species biofilm and exhibited distinct spatial and growth advantages, particularly over M. lacticum, whose growth was suppressed in co-culture. Due to its overwhelming numerical dominance within the community (> 90% relative abundance), S. rhizophila accounted for the vast majority of RNA-seq reads across all mixed biofilm conditions. Consequently, transcriptional profiling was largely restricted to this species, and despite deep sequencing, our ability to resolve transcriptomic responses from the less abundant community members and to fully characterise interspecies interaction dynamics remained limited.

The results of the S. rhizophila gene expression analysis are shown in Fig. 4 (a-c). We observed pronounced transcriptional changes in S. rhizophila when co-cultured in biofilm with M. lacticum (SR-ML, Fig. 4a), with 475 genes upregulated and 205 genes downregulated compared to its monoculture biofilm. In contrast, biofilms formed with B. licheniformis (SR-BL, Fig. 4b) and the three-species combination (SR-BL-ML, Fig. 4c) showed limited S. rhizophila transcriptional responses, with 14 and 15 genes upregulated, and 67 and 105 genes downregulated, respectively. A correlation heatmap, global expression distributions and PCoA ordination (Supplementary Figure S3 a-c) collectively showed that S. rhizophila adopts distinct transcriptional programmes across SR-ML, SR-BL and SR-BL-ML biofilms. SR-ML displayed a markedly divergent expression state, with minimal similarity to SR-BL (r = 0.03) and weak similarity to SR-BL-ML (r = 0.35), consistent with its enhanced growth in the presence of M. lacticum. By contrast, SR-BL and SR-BL-ML were more closely related (r = 0.62), reflecting the broadly similar influence of B. licheniformis on S. rhizophila transcription. Global expression distributions further supported these distinctions (Supplementary Figure S3b). Violin plots revealed that SR-ML exhibited the most asymmetric distribution (skewness = 0.816; kurtosis = 3.84), consistent with a broader activation range, whereas SR-BL showed a more constrained profile (skewness = 0.470; kurtosis = 3.45), with SR-BL-ML intermediate (skewness = 0.755; kurtosis = 3.98). PCoA resolved three discrete clusters corresponding to each condition, confirming that partner species impose reproducible and condition-specific transcriptional states on S. rhizophila. The gene co-expression network of S. rhizophila in the SR-ML condition demonstrates a coordinated and functionally organised transcriptional response, with broad modulation of genes across discrete modules associated with motility, nutrient acquisition, core metabolism, and cell division (Fig. 5).

Fig. 4.

Differential gene expression of Stenotrophomonas rhizophila in three different biofilm combinations using RNA-Seq. Panel (a) shows gene expression changes in S. rhizophila when co-cultured with Microbacterium lacticum (SR–ML), (b) shows expression changes in co-culture with Bacillus licheniformis (SR–BL), and (c) shows expression changes in the three-species biofilm with both M. lacticum and B. licheniformis (SR–ML–BL). Coloured dots represent genes based on fold change and statistical significance: grey dots indicate non-significant genes for both log₂(fold change) and p value; red dots indicate statistical significance for p value (p < 0.05) but not log₂(fold change); green dots indicate statistical significance for both p-value and log₂(fold change), highlighting biologically relevant genes; and blue dots indicate genes that meet the log₂(fold change) threshold but do not reach statistical significance (p > 0.05)

Fig. 5.

Protein-protein interaction (PPI) network of Stenotrophomonas rhizophila genes that are differentially expressed in the SR-ML biofilm combination (ML = Microbacterium lacticum). Nodes represent genes with significant differential expression (adjusted p < 0.05), grouped by predicted biological function. Node outlines are coloured in black or red. Red outlines indicate downregulated genes, whereas black outlines indicate upregulated genes. Node fill colours denote functional clusters only (e.g. transport, stress response, chemotaxis, flagellar synthesis, cell wall and division, ribosomal proteins, fatty acid metabolism, electron transport, ATP production, amino acid and phosphate metabolism, nucleic acid metabolism, and metal/antibiotic resistance) and are used purely to aid visual grouping, not to represent expression changes

Functional Reprogramming of S. Rhizophila in Multispecies Biofilms

Heatmaps of S. rhizophila gene expression across three multispecies biofilm combinations (SR-ML, SR-BL, and SR-BL-ML), each computed relative to its monospecies biofilm, revealed coordinated transcriptional changes across multiple functional pathways, including (i) flagellar production and regulation (Fig. 6a), (ii) chemotaxis and cell division (Fig. 6b), (iii) stress and regulatory responses (Fig. 6c), (iv) energy metabolism and cytochrome activity (Fig. 6d), and (v) nutrient transport and metabolism (Fig. 6e). Overall, S. rhizophila exhibited the strongest transcriptional activation in the SR-ML combination, with consistent upregulation of motility-, metabolism-, stress-response-, and nutrient-uptake-related genes. In contrast, the presence of B. licheniformis (SR-BL and SR-BL-ML) led to marked suppression across these pathways.

Fig. 6.

Heatmaps showing log₂ fold-changes in gene expression for S. rhizophila grown in (i) dual-species biofilm with Microbacterium lacticum (SR–ML), (ii) dual-species biofilm with Bacillus licheniformis (SR-BL), and (iii) three-species biofilm with M. lacticum and B. licheniformis (SR-BL-ML), each relative to S. rhizophila monoculture biofilms. Genes are grouped into major functional categories: (a) flagella production and regulation; (b) chemotaxis and cell division; (c) stress response and regulatory pathways; (d) energy metabolism and cytochromes; and (e) nutrient transport and metabolic functions. Colour scale represents log₂ (fold-change) values, with red indicating upregulation and blue indicating downregulation. All genes shown were significantly differentially expressed (adjusted p < 0.05)

Flagellar biosynthesis-related transcripts were most abundant in SR-ML (average pathway expression value: 1.22), with intermediate levels in SR-BL-ML (0.87) and the lowest in SR-BL (0.28). Similarly, transcripts associated with chemotaxis were elevated in SR-ML (0.96), but reduced in SR-BL-ML (0.50) and SR-BL (−0.07), suggesting diminished chemotaxis-related signalling and environmental sensing by S. rhizophila in the presence of B. licheniformis. Genes involved in cell division followed the same trend, exhibiting the highest expression in SR-ML (1.19), moderate levels of upregulation in SR-BL-ML (0.40), and downregulation in SR-BL (−0.06).

The electron transport chain genes (nuoD, nuoE, nuoG, nuoH, nuoI, nuoL, nuoM, nuoN) were upregulated in SR-ML (average pathway expression value: 1.05), whereas their expression was lower in SR-BL-ML (0.12) and further downregulated in SR-BL (−0.26), suggesting a decline in respiratory activity. Similarly, TCA cycle genes (acnB, acnA, suhR, sucC, sucB, sucD) showed the highest expression in SR-ML (2.66), while being significantly reduced in SR-BL-ML (0.60) and SR-BL (−0.30), indicating a strong metabolic response of S. rhizophila in the SR-ML combination. Genes related to electron transport and cytochrome activity (cyoB, cyoC, cyoD, cydA, cydB, ccmE) were also highly upregulated in SR-ML (1.65), compared to SR-BL-ML (0.17) and SR-BL (−0.29). These findings suggest that S. rhizophila in the SR-ML biofilm maintains a metabolically active state, while, in SR-BL-ML and SR-BL, S. rhizophila exhibits reduced energy metabolism, likely reflecting distinct metabolic interactions and resource availability within the biofilm communities.

Stress response genes (HSP22.0, groL, htpG, dnaK, clpB, dnaJ, groS, cspA, groE) were upregulated in SR-ML (mean expression: 1.39) but showed reduced expression in SR-BL-ML (0.37) and SR-BL (0.05) compared to S. rhizophila monospecies biofilms. Similarly, nitrogen metabolism genes (glnA, gltB, gltD, gltA, hutU, hutH, hutI) had the highest expression in SR-ML (1.16), with lower levels in SR-BL-ML (0.17) and SR-BL (−0.08). Lipid metabolism genes (fadB, fadE, fadA) were also more expressed in SR-ML (1.79) compared to SR-BL-ML (0.00) and SR-BL (−0.38).

S. rhizophila transporters for vitamin B12 (btuB), iron-siderophore uptake (fpvA, cirA, fepA, fhuA), and the TonB-ExbB-ExbD system (tonB, exbB, exbD1), which provides energy for active nutrient import, exhibited the highest expression in SR-ML (mean expression: 1.55), compared to SR-BL-ML (0.19) and SR-BL (−0.20). These genes encode TonB-dependent transporters, which are outer membrane proteins that facilitate the uptake of essential nutrients such as iron and vitamin B12 across bacterial membranes. The increased expression of these genes in SR-ML suggests that M. lacticum likely enhances nutrient uptake in S. rhizophila through metabolic cross-feeding or increased bioavailability of micronutrients by improving transport across membranes, which likely induce energy metabolism, as reported above.

The top five downregulated genes in S. rhizophila in SR-ML compared to S. rhizophila monospecies biofilm included trpR (tryptophan biosynthesis repressor), pstS (phosphate uptake), phoB (phosphate starvation response), and hisD and hisG (histidine biosynthesis). These genes were more strongly downregulated in the SR-ML biofilm compared to SR-BL-ML and SR-BL. For instance, trpR was downregulated 2.57-fold in SR-ML relative to S. rhizophila monoculture, compared to 1.0-fold and 0.78-fold downregulation in SR-BL and SR-BL-ML, respectively. Similarly, hisG expression was reduced 2.06-fold in SR-ML versus S. rhizophila alone, while it was downregulated 1.0-fold in SR-BL and 0.43-fold in SR-BL-ML. The phosphate uptake gene pstS showed an 11.20-fold downregulation in SR-ML compared to S. rhizophila monoculture, whereas it was reduced by 1.0-fold and 2.77-fold in SR-BL and SR-BL-ML, respectively. Likewise, the phosphate starvation response regulator phoB was downregulated 11.98-fold in SR-ML relative to S. rhizophila monoculture, compared to 1.0-fold and 3.59-fold downregulation in SR-BL and SR-BL-ML, respectively. Finally, hisD was downregulated 2.10-fold more in SR-ML than in SR-BL and 1.69-fold more than in SR-BL-ML.

In summary, RNA-Seq analysis revealed that S. rhizophila exhibits the most pronounced transcriptional reprogramming when co-cultured with M. lacticum (SR-ML), with broad upregulation of genes involved in motility, metabolism, nutrient transport, and stress response, alongside strong downregulation of genes associated with phosphate and amino acid biosynthesis. In contrast, in biofilms with B. licheniformis (SR-BL) or in the three-species condition (SR-BL-ML), S. rhizophila showed relatively muted transcriptional responses. The distinctive S. rhizophila transcriptomic profile in SR-ML highlights the stimulatory role of M. lacticum, which appears to enhance nutrient availability and energy metabolism in S. rhizophila. In comparison, the transcriptional response of S. rhizophila in SR-BL was in line with their observed neutral interaction, whereas the presence of B. licheniformis in the three-species biofilm appears to suppress the effects seen in SR-ML. These findings confirm the previously observed exploitative interaction between B. licheniformis and M. lacticum, highlighting its role in reshaping the molecular and ecological dynamics of the biofilm.

Discussion

In mixed-species biofilms, bacterial spatial organisation is shaped by the complex interplay between microenvironmental dynamics and interspecies interactions [16]. Here, we explored how spatial organisation and transcriptional responses in the dominant community member, S. rhizophila, are influenced by previously characterised interspecies interactions within a dairy-origin model community. S. rhizophila exhibited a growth advantage in the presence of M. lacticum, while showing a neutral association with B. licheniformis [11]. These findings provide mechanistic insight into how microbial interactions can drive spatial and functional heterogeneity within biofilm communities.

In our previous work, M. lacticum was identified as a potential keystone species that enhanced the growth of all other members in a four-species biofilm, including S. rhizophila and B. licheniformis. Notably, B. licheniformis did not form a measurable biofilm in monoculture under the tested conditions, but exhibited increased biomass only in co-culture with M. lacticum, resulting in a synergistic increase in overall community biofilm formation [11]. Here, we observed that M. lacticum formed the initial biofilm layer, subsequently facilitating the abundant growth of S. rhizophila relative to its monoculture biofilm. This enhanced growth enabled S. rhizophila to dominate the community, ultimately reducing the growth of M. lacticum in the underlying biofilm layers. B. licheniformis and S. rhizophila were observed in close spatial proximity within the middle and upper biofilm layers from 12 to 24 h (Fig. 2), further reducing cell counts of M. lacticum, consistent with a previously described exploitative interaction between M. lacticum and B. licheniformis [11]. Complementary planktonic co-culture assays confirmed that the interaction patterns observed in biofilms were not due to intrinsic growth advantages in suspension. None of the species pairs reproduced the biofilm-associated fitness effects under well-mixed conditions, indicating that these interactions arise only when cells are spatially organised. This aligns with previous studies showing that microbial cooperation and competition often depend on biofilm-specific microenvironments, nutrient gradients, and physical structure rather than planktonic growth behaviour [10]. In contrast to our system, where interaction patterns were confined to biofilms, several studies have reported similar competitive outcomes in both planktonic and biofilm co-cultures – for example, Pseudomonas aeruginosa consistently outcompeting Staphylococcus aureus in vitro in both growth modes [31].

This corresponds well with previously reported interactions among the three species and verifies the potential of spatial structure analysis as a predictor of interspecies interactions. Yang and colleagues [25] investigated the dynamics of a four-species biofilm community on root surface over 15 days, demonstrating how Paenibacillus amylolyticus – initially a weak root coloniser with a relative abundance of 16.2% at day 5 in monoculture biofilms – became the second most relatively abundant species, increasing to 34.6% by day 15, through interactions with other community members when residing in the four-species community. Similarly, other studies have demonstrated the enhanced colonisation ability of bacterial strains in mixed-species biofilms. For instance, the colonisation ability of Escherichia coli O157 was shown to be significantly increased in dual-species biofilms with Acinetobacter calcoaceticus [32] and P. aeruginosa [33]. Similarly, another study showed enhanced surface colonisation (> 10-fold) and biofilm formation of Selenomonas sputigena in the presence of Streptococcus mutans [17].

The biovolume of M. lacticum in the three-species biofilm did not change significantly over time. This is consistent with our previous findings, which indicate that M. lacticum engages in a commensal relationship with S. rhizophila and is exploited by B. licheniformis in pairwise interactions [11]. Although the spatial organisation of S. rhizophila may enhance its growth and fitness by positioning on top of M. lacticum [19], it renders S. rhizophila more susceptible to antimicrobials [4]. The reduced growth of M. lacticum could be attributed to its spatial disadvantage and limited access to nutrients and oxygen, caused by the abundant growth of S. rhizophila on top of it. Additionally, the matrix produced by B. licheniformis and its exploitative relationship with M. lacticum could also be contributing factors [34]. The dominance of M. lacticum in the bottom layers could be due to its strong adherence capacity, robust biofilm-forming ability, and tolerance for lower oxygen levels or less strict nutrient requirements. Cellular arrangement within the biofilm influences resource accessibility and metabolic activity, contributing to antimicrobial drug resilience [35]. Since M. lacticum forms a layer at the bottom of the biofilm, with S. rhizophila growing on top, this arrangement likely creates nutrient-deficient conditions for M. lacticum, potentially inducing a low metabolic state that might have a role in its protection from antimicrobials. The spatial organisation may further enhance this protective effect [4].

A negative correlation between the biomass abundance of S. rhizophila and M. lacticum indicates that they occupy distinct spatial niches (Fig. 4). S. rhizophila preferentially colonises the biofilm surface, positioning itself above M. lacticum and driving their vertical separation within the biofilm. Interestingly, the positive yet statistically inconsistent correlation between S. rhizophila and B. licheniformis contrasts with recent in silico findings from a dual-species model biofilm of gut mucosal communities, where neutralism at the morphological level led to spatial segregation between bacterial species [28], although the study involved a different microbial system and experimental context. Furthermore, the negative correlation between B. licheniformis and M. lacticum (Fig. 4) supports the idea that exploitative interactions can lead to spatial partitioning, reinforcing the role of competitive dynamics in shaping biofilm architecture. It has been shown that a negative interaction between two bacterial species leads to spatial segregation within the biofilm [10, 36, 37].

Building on the distinct spatial organisation observed in the mixed-species biofilms – particularly the progressive dominance of S. rhizophila from 6 to 24 h on the pre-established layer of M. lacticum – we hypothesised that these structural dynamics reflect underlying physiological and transcriptional shifts in S. rhizophila in response to the presence of M. lacticum. To investigate this, we performed RNA-Seq to profile the global gene expression of S. rhizophila in dual-species (SR-ML, SR-BL) and three-species (SR-BL-ML) biofilms, each compared with its monospecies biofilm. Pairwise Pearson correlations further demonstrated that gene expression patterns of S. rhizophila diverged markedly when interacting with M. lacticum, whereas interactions involving B. licheniformis resulted in broadly conserved transcriptional programmes. Given the previously characterised antagonistic interaction in which B. licheniformis suppresses the growth of M. lacticum, the presence of B. licheniformis likely attenuates the interaction pressure exerted by M. lacticum on S. rhizophila, thereby reducing the extent of transcriptional reprogramming in S. rhizophila.

The transcriptional response of S. rhizophila in dual-species (SR-ML, SR-BL) and three-species (SR-BL-ML) biofilms revealed shifts in its physiology and metabolism, compared to its monospecies biofilm state, with the most pronounced gene expression changes occurring in the presence of M. lacticum, and the least changes observed in the presence of B. licheniformis, with which S. rhizophila appears to develop a neutral interaction (Figs. 5 and 6). Upregulation of flagella synthesis, chemotaxis, cell division, and metabolic pathways suggests that S. rhizophila adjusts its motility, positioning, and growth to colonise the surface of M. lacticum, as supported by its spatial organisation. Motility and chemotaxis play multiple roles in collective behaviours of bacteria including biofilm formation, swarming, and autoaggregation [38]. Contrary to reports suggesting that high levels of c-di-GMP suppress flagellar gene expression, and inhibit motility in biofilms, while promoting EPS biosynthesis [39], our findings reveal a role for flagellar machinery and chemotaxis in biofilm formation. Bacteria locate suitable sites for colonisation by integrating mechanical and chemical cues during surface sensing. Salemi et al. demonstrated how the flagellar accessory protein FssF (a homolog of the C-ring protein FliN) linked chemotaxis to surface sensing in Caulobacter crescentus [40]. Flagellar rotational switching, regulated by chemotaxis, has been shown to initiate biofilm formation in Helicobacter pylori as well [41]. Upregulation of flagella in Geobacter sulfurreducens was reported to serve as matrix scaffolds [42]. Similarly, flagella play an essential role in initiating biofilm formation in Vibrio cholerae by enabling cells to swim and detect surfaces [43]. The increased expression of genes associated with metabolic activity, including nitrogen and lipid metabolism in S. rhizophila within the SR-ML combination, as well as overall energy metabolism (e.g., the TCA cycle and electron transport chain) in S. rhizophila biofilms, likely reflects a response to nutrient limitation in deeper biofilm layers. This response was absent when S. rhizophila was grown with B. licheniformis or in the three-species combination. In general, cells within biofilms experience spatially distinct nutrient limitations and stress conditions, driving different metabolic programming. Nutrient-starved interior cells were reported to supply amino acids to peripheral cells, while peripheral cells exhibit reduced membrane potential and provide fatty acids to support cells in the interior. This suggests a division of metabolic labour across biofilm regions [2]. However, as the gene expression data represent population-averaged values rather than capturing layer-specific expression within the biofilm, this constitutes a limitation of the study. Although qPCR is sometimes used to validate individual transcriptional targets, our conclusions rely on broad pathway-level changes involving hundreds of genes rather than on any single low-expression gene. RNA-seq is considered sufficiently robust for such analyses, and independent qPCR validation is not routinely required when experiments follow state-of-the-art guidelines [44].

The upregulation of TonB-dependent receptors indicates potential metabolite exchange across membranes. These receptors are widespread outer membrane β-barrel proteins that enable the uptake of nutrients and bacteriocins. Their operation depends on the proton motive force, facilitated through an interaction with the ExbB/ExbD/TonB complex in the inner membrane [45]. Gram-negative bacteria acquire ferric siderophores and vitamin B12 through TonB-dependent outer membrane transporters [46]. The gene btuB, involved in vitamin B12 uptake, was the most upregulated gene in the genome of S. rhizophila and among all TonB-dependent receptors during biofilm formation with M. lacticum. The specific upregulation of TonB-dependent receptors and genes of the ExbB/ExbD/TonB complex in S. rhizophila in SR-ML combination suggests a possible role of S. rhizophila in exploiting its physical association with M. lacticum to acquire essential micronutrients, including iron and cobalamin, thereby enabling metabolic activation in the co-culture environment. Conversely, the downregulation of hisG, hisD, pstS, and phoB indicates a metabolic shift towards energy conservation and stress adaptation, with reduced investment in amino acid and phosphate biosynthesis pathways.

The alignment of the observed spatial organisation and previously identified interspecies interactions in this three-species biofilm community highlights the pivotal role of such interactions in shaping microbial community structure. Metabolic adaptation in S. rhizophila, as shown by changes in gene expression in different biofilm combinations confirms the bacterial ability to fine-tune metabolic processes in response to interspecies interactions for better resource utilisation, reflecting adaptive strategies that optimise survival and functional integration within the biofilm community. Whilst our RNA-seq analysis highlights pathways, including the upregulation of TonB-dependent receptors, that suggest metabolite-mediated interactions in these model biofilm organisms, the specific metabolites contributing to enhanced biomass remain unknown. Identifying these compounds will require targeted metabolomic or biochemical assays, and future work will experimentally validate these interactions to clarify the mechanistic basis of enhanced growth and biomass. Additionally, it is important to note that the transcriptomic profiles represent the biofilm state after 24 h, when community structure and interspecies interactions have stabilised. Earlier transcriptional events are likely to involve transient regulatory or stress responses and may differ from the mature biofilm state captured here. Investigating these temporal dynamics will require time-resolved transcriptomics to determine how interspecies interactions evolve during biofilm development.

Understanding these spatial relationships is essential for uncovering the mechanisms that drive biofilm formation and microbial community structure. Biofilms in healthcare, water systems, and industrial settings often exhibit similar spatial organisation and interspecies interactions to those observed in the dairy industry [47–50]. Insights from this study could inform targeted biofilm control strategies, such as focusing on keystone species, specific biofilm layers, or key metabolic pathways to control community dynamics in mixed-species biofilms. This is particularly relevant in the dairy industry, where early colonisers could be targeted to improve biofilm management and control.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contributions

FS and NY conducted the experiments and analysed the data related to CLSM. FS also carried out the RNA-Seq experiments and performed the corresponding data analysis. JG carried out the experiments related to bacterial growth in planktonic suspension. KDR, MB, and MH contributed by finalising the work plan, providing scientific guidance, and thoroughly reviewing the manuscript, shaping it into its current form.

Funding

Open access funding provided by Copenhagen University

Data Availability

Raw data for all CLSM images is available on Zenodo under the title ‘Metadata for Confocal Laser Scanning Microscopy Images of Monoculture and Mixed-Species Biofilms Formed by Bacterial Isolates of Dairy Origin’: https://zenodo.org/records/10659514. RNA-Seq data generated in this study have been deposited in the NCBI Sequence Read Archive (SRA) under BioProject accession number PRJNA1075650. The dataset includes transcriptomic profiles of *Stenotrophomonas rhizophila* under three co-culture conditions with three replicates for each: SRR27961408 (dual-species biofilm with *Microbacterium lacticum*), SRR27961407 (dual-species biofilm with *Bacillus licheniformis*), and SRR27961406 (three-species biofilm with *M. lacticum* and *B. licheniformis*).

Declarations

Competing interests

The authors declare no competing interests.

Disclaimer

The results and conclusions in this article reflect only the authors’ view. The funding Research Executive Agency (REA), delegated by the European Commission, is not responsible for any use that may be made of the information it contains.