Comparative RNA-Seq reveals conserved and species-specific biofilm-associated pathways across clinically relevant airway bacteria

Biofilms are structured microbial communities associated with chronic airway disease and increased tolerance to antimicrobial treatments. While biofilm-associated physiology has been studied in individual species, the extent to which transcriptional responses are conserved across diverse airway pathogens remains poorly defined. We performed comparative RNA sequencing across eight clinically relevant bacterial species under three conditions: planktonic growth, biofilm formation, and perturbation of established biofilms using doxycycline or a biocidal agent (Zoono GermFree24). Ortholog clustering and fold-change–based prioritisation were integrated to identify conserved transcriptional responses across species. Directional conservation was quantified to assess the extent of shared regulation within orthologous gene clusters. Biofilm formation induced widespread but modest transcriptional changes across species, with partial conservation observed in functional categories related to metabolism, transport and stress response. However, gene-level conservation was limited, with only a subset of ortholog clusters showing consistent directional change. In contrast, transcriptional responses to perturbation were highly heterogeneous, with substantial variation in both magnitude and direction across species. While doxycycline preferentially affected information processing pathways and Zoono induced broader functional shifts, both perturbations exhibited limited cross-species conservation. Quantitative analysis confirmed that most ortholog clusters displayed intermediate or inconsistent directional agreement, with strongly conserved responses representing a minority. These findings demonstrate that biofilm formation is associated with partially conserved functional adaptation across airway pathogens, whereas responses to antimicrobial perturbation are largely species-specific. This distinction highlights the importance of functional constraints during biofilm establishment and regulatory flexibility under stress, with implications for the development of broadly effective strategies to target biofilm-associated infections.

Bacterial biofilms are microbial communities that are embedded within a self-produced extracellular matrix that confers resistance to environmental, immune and antimicrobial stress. As a result, biofilms are implicated in persistent infections across clinical settings, including chronic airway diseases such as chronic rhinosinusitis, cystic fibrosis and non-cystic fibrosis bronchiectasis (Peters et al., 2012; Lopes et al., 2014; Menberu et al., 2021; Herzberg and van Hasselt, 2025). These airway diseases are also characterized by complex microbial communities rather than single-pathogen infections, making it important to understand which biofilm-associated responses are conserved across taxa and which are likely to vary between community members (Broderick et al., 2025).

Despite increasing recognition of their role, studying airway biofilms remains challenging due to limited accessibility of the respiratory tract and the complexity of polymicrobial communities. Several airway taxa including Staphylococcus aureus, Pseudomonas aeruginosa, Streptococcus pneumoniae, Moraxella catarrhalis and Staphylococcus epidermidis have been noted to form biofilms in a range of conditions including chronic rhinosinusitis (Koefoed et al., 2023; Liu et al., 2024). While Corynebacterium accolens and Dolosigranulum pigrum are commensal bacteria often reported to modulate or reduce pathogenicity of other bacteria, they have rarely been reported to either produce biofilms themselves or interact with other those of other species (Lopes et al., 2014; Menberu et al., 2021).

In vivo, airway biofilms are unlikely to exist as isolated single-species structures. Rather, they are commonly embedded within polymicrobial communities in which neighbouring taxa may influence biofilm architecture, metabolic activity, stress tolerance, antimicrobial susceptibility, and persistence (Peters et al., 2012). These interactions can be cooperative or competitive, involving mechanisms such as metabolic cross-feeding, competition for resources, quorum sensing, extracellular matrix production and antimicrobial interference (Herzberg and van Hasselt, 2025). As a result, the behaviour of a species within a polymicrobial biofilm may not be predictable from its behaviour in monoculture alone.

Moreover, while many airway taxa can form biofilms in vitro, confirming biofilm formation in situ remains difficult, as detection typically requires preservation of spatial structure and direct imaging approaches such as confocal microscopy, scanning electron microscopy, or fluorescence in situ hybridisation (Barbosa et al., 2023). Direct imaging of biofilms in tissue or mucus samples can overcome some of these challenges, but imaging techniques are time consuming, expensive, potentially invasive and still requires molecular targets to distinguish between simple clusters of bacteria and the more ordered and structured biofilm growth phase (UniProt, 2025). In addition, conserved molecular markers for biofilm detection that are applicable across diverse bacterial species remain limited, reflecting that while biofilm formation may require several similar phenotypic and regulatory changes these are likely specific for each bacterial species rather than a single universal gene or regulator (Koefoed et al., 2023; Liu et al., 2024).

Although biofilm resistance to antibiotics is well established, less is known about whether transcriptional responses to antimicrobial perturbation are conserved across species. A key challenge for interpreting polymicrobial biofilms is distinguishing transcriptional responses that reflect broadly conserved biofilm-associated physiology from those that are species-specific. This distinction is important because mixed-species biofilm behaviour emerges from the combined responses of individual community members, but the extent to which these responses are shared across taxa remains unclear. Comparative analysis of monoculture biofilms therefore provides an important baseline for identifying conserved functional constraints that may also be relevant within polymicrobial communities, while recognising that interspecies interactions may substantially modify these responses.

Here, we performed a cross-species transcriptomic analysis of clinically relevant airway bacteria to (i) identify conserved functional signatures of biofilm formation, (ii) compare responses to antibiotic and biocidal perturbation, and (iii) establish a comparative baseline for interpreting shared versus species-specific transcriptional behaviours in polymicrobial airway biofilms.

Bacterial strains and biofilm establishment

Biofilms were established from eight ATCC reference strains: methicillin-susceptible Staphylococcus aureus (ATCC 6538), methicillin-resistant S. aureus (ATCC 33593), P. aeruginosa (ATCC 27853), S. epidermidis (ATCC 12228), M. catarrhalis (ATCC 25238), C. accolens (ATCC 49726), D. pigrum (ATCC 51524), and S. pneumoniae (ATCC 49619). Species specific culture conditions are provided in Supplementary Table 1. Growth conditions were individually optimised for each species to maximise recovery of reproducible surface-associated biomass on the Calgary Biofilm Device. The objective was not to generate identical biofilm architectures across taxa, but to establish biologically relevant surface-associated growth states suitable for comparative transcriptomic analysis. Following biofilm establishment, biofilms were exposed to one of the following conditions: untreated biofilm, consisting of a further 24-hour incubation with fresh growth medium; doxycycline-treated biofilm, consisting of growth media supplemented with 0.27 µg/mL doxycycline, selected to reflect reported sinonasal mucus concentrations after oral administration; or biocide-treated biofilm, where biofilms were treated with Zoono for species-specific durations determined through preliminary optimisation experiments designed to produce measurable but non-complete reductions in biofilm-associated viability. Treatment durations ranged from 20 seconds to 5 minutes depending on species (Supplementary Table 2).

Following removal of planktonic bacteria, biofilms were recovered through use of a cell scraper and RNA was extracted using the AllPrep DNA/RNA kit (Qiagen) according to the manufacturer’s protocol with subsequent DNAse treatment with PCR and Qubit RNA High Sensitivity assay used to assess yield and quality. Approximately 200 ng RNA samples (n = 32) were processed at Auckland Genomics (University of Auckland) using the Zymo-Seq RiboFree Total RNA Library Kit and sequenced on an Illumina NovaSeq X Plus platform. A total of 32 RNA samples were sequenced, corresponding to eight bacterial species under four experimental conditions (planktonic, untreated biofilm, doxycycline-treated biofilm, and Zoono-treated biofilm). As the primary objective of this study was to compare transcriptional responses across diverse species, a single RNA sample was sequenced for each species-condition combination.

Bioinformatic processing

Raw sequencing reads were quality assessed, filtered and trimmed using FastQC and Cutadapt. HISAT2 was used to align trimmed reads to species-specific reference genomes or closely related reference assemblies when exact genomes were not available. Alignments were processed using SAMtools, and gene-level counts were generated using featureCounts. Because biological replication was not available for individual species-condition combinations, formal differential expression testing and false-discovery-rate-based inference were not performed. Instead, genes were ranked within each species according to the magnitude of transcriptional change and the top 10% retained as fold-change-prioritised candidates. This approach was intended to identify the strongest transcriptional responses within each organism for downstream cross-species comparison and should not be interpreted as a measure of statistical significance.

To enable cross-species comparison, protein sequences were predicted using Prodigal and clustered using MMseqs2. A minimum sequence identity threshold of 30% and coverage threshold of 50% were selected to identify putative orthologous protein families across phylogenetically diverse airway-associated bacteria. As the aim was to capture functional conservation rather than strict one-to-one orthology, these thresholds were chosen as a compromise between sensitivity and specificity, allowing detection of related protein families despite substantial sequence divergence between taxa. To assess the robustness of downstream conclusions, sensitivity analyses were also performed using increasingly stringent clustering thresholds of 40% identity/60% coverage and 50% identity/70% coverage (Supplementary Tables 4A, B). The final high-confidence candidate set was then reassigned under each clustering scheme to determine whether cross-species candidate structure was retained at stricter thresholds. Clusters were further restricted to those present in ≥5 species and supported by large fold-change responses. Functional annotation was performed using Swiss-Prot (UniProt, 2025) and Pfam (Mistry et al., 2021), and grouped into broad biological processes using a keyword-based framework (Supplementary Table 3). Categories were defined to capture core cellular functions and included metabolism/energy, transport, information processing, regulation/stress, cell envelope/membrane, and unknown/other annotated functions. This approach enables consistent cross-species comparison while minimising bias arising from variable annotation depth and coverage across genomes.

To further characterise cross-species transcriptional responses, directional agreement was assessed at the level of ortholog clusters. For each cluster and comparison, the number of species in which the cluster was detected and the direction of transcriptional change (up- or down-regulation) were determined. Directional agreement was defined as the proportion of species exhibiting the majority direction of change within a cluster and was used as a descriptive measure of concordance rather than a formal test of conservation. Clusters were grouped according to the proportion of species sharing the majority response, and directional agreement analyses were performed for all ortholog clusters, with summary visualisation restricted to clusters present in at least five species to focus on broadly shared responses. To assess whether observed directional agreement exceeded that expected by chance, agreement values were compared with a binomial null expectation based on random assignment of transcriptional direction.

Sequencing quality and overview

Across samples, a median of 85.1 million fragments (range 6.0–168.4) were processed, with an average alignment rate of 73.1% (range 8.8–87.1%). These metrics indicate robust transcriptome coverage across all samples and species, supporting downstream comparative analysis.

Ortholog clustering and candidate definition

Ortholog clustering identified 1,503 clusters shared across multiple species, while a fold-change–based approach defined 6,991 candidate genes within species. Of these, 875 genes were supported by both approaches, representing a high-confidence subset with evidence of both strong expression change and cross-species conservation (Figure 1A).

Most clusters were present in a limited number of species, with only 29 detected across all taxa, indicating that gene-level conservation was restricted (Figure 1B). However, clusters shared across subsets of species were common, particularly among closely related organisms. Together, these results establish a spectrum of transcriptional responses ranging from species-specific to conserved across taxa. Sensitivity analyses demonstrated that increasing MMseqs2 clustering stringency reduced the number of broadly shared ortholog clusters, as expected, but did not eliminate cross-species structure. Under the primary clustering threshold (30% identity/50% coverage), 823 raw clusters were present in at least five species and 248 were shared across all eight species. Increasing stringency to 50% identity/70% coverage reduced these values to 277 and 48 clusters, respectively. Reassignment of the final high-confidence candidate gene set under each clustering scheme yielded similar conclusions, with 138 candidate-associated clusters remaining linked to raw clusters present in at least five species at the strictest threshold. These findings indicate that the conserved transcriptional patterns identified in this study are not solely dependent on a single clustering threshold and remain detectable across a range of orthology definitions (Supplementary Tables 4A, B).

Transcriptional responses to biofilm formation

To address whether conserved transcriptomic signatures of biofilm formation can be identified across upper airway pathogens, we first examined cross-species transcriptional responses during the transition from planktonic to biofilm growth.

Biofilm formation induced widespread but generally modest transcriptional changes across all species, spanning metabolism/energy, transport, regulation/stress, and information processing (Figure 2A). However, the relative contribution of these functional categories varied substantially between species, indicating that while biofilm formation is a shared phenotypic state, the underlying transcriptional responses are highly heterogeneous. The proportion of shared ortholog candidates retained within the top 10% fold-change set varied by species and functional category, further supporting partial rather than uniform conservation of the biofilm response (Figure 2B).

Despite variability, subsets of coordinated functional shifts were observed. Several species exhibited enrichment in specific pathways, including increases in transport and cell envelope-associated functions in C. accolens, and upregulation of regulation/stress-associated pathways in M. catarrhalis and S. epidermidis. In contrast, other species, including P. aeruginosa, D. pigrum, and S. pneumoniae, displayed more distributed transcriptional responses without a single dominant functional category. Responses in S. aureus strains, were similarly variable, with contributions from poorly characterised (“other annotated”) functions and reductions in transport-related pathways in some cases.

Consistent with ortholog clustering analysis, gene-level conservation across species was limited. Only a small subset of clusters exhibited shared transcriptional responses, and these were primarily associated with transport and metabolic functions (Figure 2C). Notably, even among these shared clusters, effect sizes were generally modest and directional agreement was incomplete across species.

The limited gene-level conservation suggests that conserved biofilm signatures are more likely to be detected at the level of broad functional processes rather than individual genes.

Transcriptional responses of established biofilms to perturbation

To assess whether antimicrobial responses are conserved across airway pathogens, we examined transcriptional changes in established biofilms following doxycycline and a biocide Zoono GermFree24 (Zoono Ltd., Auckland, New Zealand), an aqueous non-alcoholic handsanitiser in common use exposure.

Perturbation induced more heterogeneous transcriptional responses than those observed during biofilm formation, with substantial variation in both magnitude and direction across species (Figure 3). Although changes spanned core functional categories including metabolism/energy, transport, regulation/stress, and information processing, no consistent cross-species pattern was observed and agreement within functional groups was limited.

Species-specific responses dominated under both doxycycline exposure and Zoono perturbation. For example, C. accolens showed decreases in transport and membrane-associated functions, while S. pneumoniae exhibited increases in transport alongside reductions in regulation/stress pathways. P. aeruginosa showed reduced information processing activity, whereas S. epidermidis demonstrated increased cell envelope/membrane-associated responses. In several species, including D. pigrum and Staphylococcus aureus strain 6538, responses were more diffuse, with modest changes distributed across multiple functional categories. This variability indicates that, in contrast to biofilm formation, transcriptional adaptation to external stress is not governed by a common functional cluster.

Consistent with ortholog clustering results, conserved transcriptional responses to perturbation were limited. Only a small number of shared clusters exhibited consistent directional changes across species, primarily involving functions related to information processing and metabolism (Figure 3). However, most responses showed incomplete agreement in both magnitude and direction, indicating that adaptation to external stress is largely species-specific.

While both doxycycline and Zoono induced structured transcriptional responses, their effects differed in functional emphasis. Doxycycline exposure was more frequently associated with changes in information processing functions, consistent with disruption of translational machinery. In contrast, Zoono exposure induced broader but less consistently directed shifts across functional categories. Despite these differences, both perturbations were characterised by substantial heterogeneity and limited conservation across species.

Directional conservation of transcriptional responses across functional categories

To quantify the extent of cross-species conservation, we assessed directional consistency of transcriptional change across high-confidence ortholog clusters stratified by functional category and comparison (Figure 4A), and across the broader ortholog cluster landscape for clusters present in ≥5 species (Figure 4B).

During biofilm formation, a subset of functional clusters exhibited partial directional agreement across species, particularly within information processing and transport functions (Figure 4A). However, agreement was generally incomplete and most clusters fell within intermediate agreement ranges. Comparison with a binomial null expectation demonstrated that mean directional agreement differed only slightly from values expected under random direction assignment (biofilm 70.3% observed versus 67.2% expected; doxycycline 69.0% versus 67.2%; Zoono 66.0% versus 67.2%). These findings indicate that strongly conserved transcriptional responses were uncommon despite the presence of shared functional themes.

In contrast, perturbation responses exhibited limited directional agreement across species. While both doxycycline and Zoono treatments produced occasional clusters with high agreement, these were uncommon and unevenly distributed across functional categories. Mean agreement values were close to those expected under a random-direction null model, particularly for Zoono-associated responses, indicating that conserved transcriptional responses to perturbation were limited.

Analysis of the full ortholog cluster landscape further demonstrated this limited conservation. Most clusters were present in only a subset of species and exhibited intermediate levels of directional agreement, with few showing both high species representation and strong directional consistency.

Together, these findings indicate that while conserved transcriptional responses can be identified, they represent a minority of the overall response. Instead, transcriptional adaptation across both biofilm formation and perturbation is dominated by partial conservation and species-specific variation, even within shared functional categories.

Together, these findings support a model in which biofilm formation is governed by shared functional constraints, whereas responses to external stress are shaped by species-specific regulatory pathways, limiting the predictability of antimicrobial responses across taxa. These findings provide a comparative baseline for interpreting polymicrobial biofilms by identifying which transcriptional responses are conserved across species in monoculture and which are inherently species-specific. While interspecies interactions are expected to modify these responses in mixed communities, the present analysis defines the underlying species-level transcriptional landscape from which such interactions emerge. In polymicrobial biofilms, interspecies interactions can alter biofilm architecture, metabolic activity, spatial organisation, and antimicrobial susceptibility. Therefore, limited conservation of perturbation responses across species suggests that mixed-species communities may contain multiple, partially independent stress-response programs, complicating attempts to predict or disrupt community-level behaviour using a single conserved molecular target.

Conserved functional adaptation during biofilm formation

The transition to biofilm growth was associated with coordinated transcriptional changes across multiple species. While broad functional categories such as metabolism/energy production, transport systems, and stress-associated functions were repeatedly implicated, conservation at the level of individual ortholog clusters was limited. Directional agreement analyses demonstrated that most shared clusters exhibited only partial concordance across species and generally differed little from random expectation. These findings suggest that biofilm formation is characterised by convergence at the level of broad functional adaptation rather than strict conservation of individual transcriptional responses (Ramond et al., 2025).

Consistent with previous studies, biofilm-associated growth involved metabolic reprogramming, altered membrane composition, and activation of stress-response pathways (Chukwudi, 2016). The partial conservation observed here extends these findings by demonstrating that such adaptations can emerge across taxonomically distinct pathogens when examined at the level of shared functional pathways.

However, this conservation was not universal. Even within high-confidence shared clusters, variability in the magnitude and occasionally direction, of change was observed. This suggests that while biofilm formation imposes common functional constraints, the transcriptional routes used to achieve this state remain flexible. This balance between constraint and flexibility aligns with broader ecological models of microbial functional redundancy, in which similar community- or phenotype-level functions may be maintained despite variation in the taxa, genes, or regulatory pathways responsible (Ramond et al., 2025).

Limited conservation and increased heterogeneity under perturbation

In contrast to biofilm formation, transcriptional responses to perturbation were dominated by species-specific effects. Both doxycycline and Zoon o exposure induced measurable shifts across multiple functional categories, but these responses showed limited agreement in direction or magnitude across species. Quantitative analysis of directional conservation (Figure 4) confirmed that only a small proportion of ortholog clusters exhibited consistent responses, even among those present in multiple species.

This limited conservation likely reflects the diversity of regulatory networks governing stress responses in different organisms. While biofilm formation represents a relatively constrained adaptive state, perturbation introduces selective pressures that can be resolved through multiple, organism-specific strategies. These may include modulation of efflux systems, altered membrane composition, activation of stress-response regulators, changes in metabolic state, and other mechanisms associated with biofilm tolerance and antimicrobial resistance (Liu et al., 2024).

Importantly, the cluster-level conservation landscape (Figure 4B) demonstrates that even broadly shared orthologs are rarely associated with uniform transcriptional responses. Instead, most clusters exhibit intermediate levels of directional agreement, indicating partial or inconsistent responses across species. This reinforces the conclusion that transcriptional adaptation to external stress is inherently variable and not easily captured by a single conserved response pathway.

Distinct but non-conserved responses to doxycycline and biocidal exposure

Although both perturbations resulted in heterogeneous responses, clear differences in the structure of these responses were observed. Doxycycline exposure was more frequently associated with changes in information processing functions, consistent with its mechanism of action as a tetracycline-class antibiotic that binds the bacterial 30S ribosomal subunit and inhibits protein synthesis (Chukwudi, 2016). However, these effects were not consistently conserved across species, indicating that downstream transcriptional consequences of translational inhibition vary substantially between organisms.

In contrast, Zoono exposure induced broader transcriptional shifts spanning multiple functional categories, including metabolism, transport, and stress response pathways. Despite this breadth, directional agreement across species remained limited, suggesting that biocidal stress elicits diffuse and relatively non-specific responses rather than a coordinated transcriptional pathway.

Taken together, these findings suggest that antibiotic and biocidal perturbations differ in functional emphasis but converge in their lack of cross-species conservation. Doxycycline induces more targeted functional disruption, whereas Zoono was associated with broader but less coordinated transcriptional changes. In both cases, adaptive responses are largely species-specific.

Functional constraints versus regulatory flexibility

The contrasting patterns observed between biofilm formation and perturbation responses support a conceptual framework in which microbial transcriptional adaptation is governed by two interacting forces: functional constraint and regulatory flexibility. During biofilm formation, shared environmental pressures constrain the range of viable adaptive states, leading to partial convergence in functional responses across species. Under perturbation, these constraints are relaxed, allowing species-specific regulatory networks to drive divergent transcriptional outcomes. This interpretation is consistent with microbial ecology frameworks describing functional redundancy, where similar ecological or phenotypic functions can be maintained despite variation in the taxa, genes, or regulatory pathways responsible (Louca et al., 2018; Ramond et al., 2025).

This framework aligns with emerging views of microbial community behaviour, in which conserved ecological functions can be maintained despite substantial variability at the gene or regulatory level. It also highlights the importance of analysing transcriptional responses at the level of functional categories and orthologous groups, rather than relying solely on gene-level comparisons (Charlebois et al., 2016; Louca et al., 2018).

Implications for treatment of biofilm-associated infections

The limited conservation observed under perturbation has important implications for the treatment of chronic biofilm-associated infections, such as those seen in chronic rhinosinusitis and other airway diseases. While biofilm-associated processes may represent broadly targetable pathways across pathogens, the species-specific nature of stress responses suggests that broadly effective interventions may be more difficult to identify than targets associated with biofilm establishment.

The absence of a conserved transcriptional response to perturbation also raises the possibility that sublethal or transient exposures, which are common in clinical settings, may promote heterogeneous adaptive states within microbial communities. These may include modulation of efflux systems, altered membrane composition, activation of stress-response regulators, changes in metabolic state, and other mechanisms associated with biofilm tolerance and antimicrobial resistance (Uruén et al., 2020; Liu et al., 2024).

Representative candidate clusters and limitations of gene-level conservation

Analysis of representative candidate clusters illustrates both the strengths and limitations of cross-species comparisons. While certain clusters, such as those associated with ATP-dependent helicases and membrane transport systems, were identified across multiple comparisons, their transcriptional responses were not uniformly conserved. This demonstrates that even widely distributed orthologs do not necessarily exhibit consistent regulation across species.

Accordingly, individual candidate genes should be interpreted as indicators of broader functional trends rather than definitive markers of conserved responses. This reinforces the value of integrating gene-level analysis with higher-level functional and conservation metrics when interpreting cross-species transcriptomic data.

Limitations and future directions

Prior work has characterised biofilm transcriptomes in individual airway pathogens and, in some cases, compared dual-species or community biofilms; however, to our knowledge, studies explicitly quantifying ortholog-cluster-level directional concordance of biofilm-associated transcriptional responses across a broader panel of airway-associated bacteria have not been reported. Nevertheless, several limitations should be considered when interpreting these findings. First, the absence of biological replication within species-condition combinations prevented estimation of within-condition variance and precluded formal differential expression testing. Consequently, the transcriptomic responses described here should be interpreted as exploratory and hypothesis-generating, with candidate genes and functional patterns requiring validation in future replicate-supported studies; likewise, directional conservation analyses were intended to provide descriptive measures of concordance across species rather than formal evidence of conserved regulatory mechanisms. Second, although all species were cultured under conditions selected to promote surface-associated growth and biofilm formation, detailed phenotypic validation of biofilm architecture was not performed as part of the present transcriptomic study. Future work incorporating microscopy, crystal violet quantification, or related biofilm characterisation methods would strengthen interpretation of the observed transcriptional responses. Third, the study compared transcriptional responses across species grown under individually optimised biofilm-forming conditions. Differences between species may therefore reflect both underlying biological variation and differences in the structure or maturity of the resulting surface-associated communities. Similarly, perturbation conditions were selected to generate biologically relevant, non-lethal stress rather than equivalent pharmacodynamic exposures across species. Doxycycline was applied at a common concentration across all organisms rather than using species-specific MIC-normalised exposures, while Zoono exposure durations were optimised individually for each species. Consequently, observed differences between species may reflect variation in effective treatment intensity as well as biological differences and should therefore be interpreted as comparative transcriptional responses under species-optimised conditions rather than direct measures of treatment equivalence. Fourth, the analysis relies on ortholog clustering to define shared functional units, which may not fully capture differences in gene regulation or functional redundancy across species. Although sensitivity analyses demonstrated broadly similar ortholog structures across a range of clustering thresholds, alternative clustering strategies may yield different candidate genes and cluster memberships. Finally, the use of bulk RNA sequencing limits the ability to resolve subpopulation-level heterogeneity, which may be particularly relevant under perturbation conditions.

Future work integrating transcriptomic data with proteomic, metabolomic, and phenotypic measurements will be important to further define the functional consequences of these transcriptional patterns. As all analyses were performed using monoculture biofilms, the findings should be interpreted as species-intrinsic transcriptional responses and not as direct representations of behaviour within polymicrobial communities, where interspecies interactions may substantially alter transcriptional activity. Future work should extend this framework to defined mixed-species airway biofilms and ex vivo or in vivo systems to determine the extent to which monoculture transcriptional responses are retained, modified, or replaced by interspecies interactions. In particular, pairing comparative transcriptomics with spatial imaging, targeted perturbation, and species-resolved abundance measurements would help determine whether conserved functional responses identified here are maintained, amplified, or suppressed within polymicrobial communities. This will be important for identifying anti-biofilm strategies that target shared functional constraints while accounting for species-specific stress responses.

This study demonstrates that biofilm formation across respiratory pathogens is associated with partially conserved functional adaptation, whereas responses to perturbation are highly heterogeneous and species-specific. These findings highlight the importance of distinguishing between conserved functional constraints and flexible regulatory responses when interpreting microbial transcriptomic data. Understanding this balance provides a framework for interpreting transcriptional behaviour in more complex polymicrobial biofilm systems and may assist future efforts to identify broadly applicable anti-biofilm strategies.