Biofilms form naturally under flowing conditions that require active surface attachment of microbial cells, in riverbeds and industrial water systems, in biological tissues and on medical devices. Fluid flow plays a crucial role in biofilm development by influencing mass transfer, nutrient diffusion, and local shear gradients within the biofilm matrix. Microstructures that develop under flow conditions, such as wrinkles and channels, aid the mass transfer of solutes and nutrients from the bulk fluid into deeper biofilm layers (Chen et al., 2024). Shear forces generated by fluid flow have been identified as one of the key factors driving structural adaptations in biofilms that cannot be replicated in static conditions (Liu and Tay, 2002). Enhanced shear stress leads to progressively thinner and denser biofilms characterized by a more compact and stable structure (Chang et al., 2020; Kwok et al., 1998; Liu and Tay, 2002; Pechaud et al., 2022; Zhang et al., 2022). Directly correlated is also the stimulated secretion of extracellular polymeric substances (EPS) under shear, benefiting the adhesion and structural integrity of biofilms (Hou et al., 2018; Liu and Tay, 2002). Incorporating flow into in vitro biofilm models revealed that biofilm phenotypes and gene expression patterns (particularly those related to virulence or adhesion) are fundamentally altered compared to their statically grown counterparts (Nailis et al., 2010). Given these profound flow-dependent effects on biofilm properties, accurate in vitro modeling necessitates the integration of representative hydrodynamic conditions.

Over the past 2 decades, increasingly sophisticated in vitro models have been developed that more closely mimic clinically relevant in vivo conditions. These provided various systems with distinct advantages and limitations. Static models cultivate bacteria directly in multi-well plates, employing different approaches: Biofilms can be formed on the well surface itself (Sharma et al., 2005), on substrate-discs (e.g., hydroxyapatite) placed on the bottom of the wells (Guggenheim et al., 2001), or on pegs inserted into the wells from above by employing a Calgary Biofilm Device (Ceri et al., 1999). While offering simplicity and high-throughput potential, these static models suffer from nutrient limitation and accumulation of metabolic byproducts (Coenye and Nelis, 2010), cannot incorporate any flow dynamics and often raise the question of whether the cells actively attach to presented surfaces or simply sediment down (Ramachandra et al., 2023).

Semi-dynamic models are primarily drip-fed systems, where growth medium is supplied by dripping it onto a substrate to cultivate biofilms under a thin fluid interface (Abdullah and Jakubovics, 2022). Using a drip-flow biofilm reactor, fluid flow is achieved by simply tilting the system to drain the liquid over the substrate surface (Goeres et al., 2009). In the case of the “constant depth film fermenter” (CDFF), excess biomass is removed by rotating scraper blades that sustain drip-fed biofilms at a uniform thickness (McBain, 2009; Pratten, 2007). While drip-fed systems achieve more consistent biofilm growth by constantly replenishing nutrients, their incorporation of flow, particularly in a controlled manner, remains limited. In contrast, advanced flow-through systems such as flow cells (Palmer and Caldwell, 1995), microfluidic chips (Tang et al., 2022) or microfluidic devices (Benoit et al., 2010) facilitate the simulation of fluid dynamics by precisely controlling flow conditions when passing bacterial culture through small channels. These systems excel at investigating biofilm formation and behavior on a microscopic level, with the benefit of real-time observation (Azeredo et al., 2017). However, their restricted working volumes and channel dimensions render them inadequate for creating entire biofilms on a macroscopic scale and across diverse substrates (Yawata et al., 2016), as often required for a comprehensive assessment of medical applications.

Lastly, the CDC (Centers for Disease Control and Prevention) biofilm reactor and the Modified Robbins device (MRD) represent fully dynamic biofilm systems on a larger scale, both harboring removable coupons or plugs, exposing a surface of interest to continuous culture flow (Abdullah and Jakubovics, 2022). The CDC reactor operates as a stirred-vessel system, where biofilm surfaces are located directly within the cultivation vessel and shear is generated through the rotational stirring motion (An et al., 2022; Rudney et al., 2012). While this design favors the optimization of cultivation conditions to achieve a steady-state culture, the exact flow exposure is difficult to determine, and the non-uniform mixing pattern created by the stirring motion leads to a certain heterogeneity between the different biofilm sites. The MRD, also called flow-over reactor, features removable plugs along a flow channel, presenting the biofilm surfaces into the flow under defined conditions (Azeredo et al., 2017; Luo et al., 2022). While traditional models connected the MRD in-line with bacterial culture or fresh medium (Kharazmi et al., 1999), recent models connected it to the efflux of a continuous culture to provide optimal biofilm growth conditions (Blanc et al., 2014). This configuration comes with a critical limitation: The flow rate within the MRD is directly coupled to the drainage outflow of the bioreactor. This coupling restricts experimental flexibility, as overly high flow rates in the MRD risk a washout of strains from the steady-state culture. At the same time, low flow rates may lead to nutrient starvation within the culture and/or biofilms.

To address these practical limitations in current methodologies, we developed an in vitro biofilm model that decouples flow dynamics from cultivation conditions. Our system improves previously reported configurations and enables the independent adaptation of both parameters, leading to a high reproducibility of the resulting biofilms. Employing a dual-cycle approach allows for maintaining precise control of flow velocity over target surfaces within the MRD during biofilm formation. At the same time, in- and outflow of the bioreactor can be individually regulated to achieve desired culture conditions, such as retaining steady-state conditions for optimal growth or inducing nutrient limitation. To demonstrate the capabilities of our decoupled flow-cultivation system, we selected oral biofilms on dental implants as a representative model system that combines the technical challenges of anaerobic multi-species cultivation with clinically relevant flow dynamics. Such relevant dynamic in vitro models are crucial to facilitate the development and evaluation of critically needed novel treatments for dental implant infections.

Dental implants have become the standard for tooth replacement in modern dentistry (Buser et al., 2012), improving not only oral health but also patients’ overall quality of life (Johannsen et al., 2012). Concurrently, the increasing prevalence of implant placement has led to a corresponding rise in peri-implantitis (Atieh et al., 2013; Warreth et al., 2015), infections characterized by inflammation and bone loss around osseointegrated implants. The main drivers for these infections are bacterial biofilms, which form on over 40% of dental implants (Algraffee et al., 2012).

These oral biofilms consist of complex, highly structured sessile microbial communities, self-embedded in extracellular polymeric matrix (Marsh, 2005), whose maturation is sequentially organized: Initial adhesion is facilitated by the presence of a pellicle (Hojo et al., 2009). This pellicle is formed by salivary-derived proteins (including mucins), which promote surface attachment of early colonizers such as Streptococci and Actinomyces (Chahal et al., 2022; Kolenbrander et al., 2002). Serving as “anchors” for subsequent colonization, the latter provide attachment sites for secondary/intermediate colonizers, including Veillonellae and Fusobacteria. These bridging species support the colonization of further species and form co-aggregates and mutualistic interspecies relationships, creating a metabolic network within the biofilm (Kolenbrander et al., 2010; Zhou et al., 2021). Eventually, late colonizers and potential pathogens, such as Porphyromonas gingivalis and Aggregatibacter actinomycetemcomitans, can integrate into the mature biofilm (Socransky et al., 1998). While all previous species are common within a healthy periodontal microbiome, the increased incorporation of these periodontal pathogens (dysbiosis) is strongly associated with periodontal disease, as they are a key driver of inflammatory processes (Hajishengallis et al., 2012; Socransky et al., 2002).

Besides their complex multi-species composition, another crucial factor to consider when studying these oral biofilms is their exposure to the highly dynamic environmental conditions within the oral cavity. Oral biofilms are typically challenged by constantly changing hydrodynamic and environmental conditions due to salivary fluid flow, which also transports planktonic bacteria between attachment sites (Hall et al., 2017). Beyond these general factors, the peri-implant environment presents even more complex flow dynamics: Connective tissue fibers orient only parallel to the implant, rather than inserting perpendicularly as they do with natural teeth. The absence of a natural periodontal ligament results in an enlarged sulcular space around the implant (Berglundh and Lindhe, 1996; Emecen-Huja et al., 2013). This sulcular space is flushed with an inflammatory secretion known as peri-implant sulcular fluid (PISF), analogous to gingival crevicular fluid (GCF) around natural teeth (Javed et al., 2011). Both PISF and GCF demonstrate significantly increased volumetric flow rates during inflammatory conditions (Bevilacqua et al., 2016): The flow of crevicular fluid is closely correlated with the severity of periodontal inflammation, serving as a direct indicator of tissue response; it increases markedly due to enhanced vascular permeability and epithelial ulceration at inflamed sites. While healthy gingiva produces minimal GCF, inflammation leads to a substantial rise in both volume and compositional complexity, transforming it into an exudate characteristic of inflammatory processes. In addition to the effects of increased fluid flow on biofilm formation, the associated enhancement in nutrient delivery promotes accelerated plaque accumulation and maturation at inflamed gingival margins (Quirynen et al., 1991). Yamamoto et al. (2021) reported a ∼88% increase in blood flow in the peri-implantitis group compared with natural teeth. Similar increases have been reported in other animal models and other groups (Katz et al., 2025; Mayr et al., 2024; Rodríguez-Martínez et al., 2006; Sugiyama et al., 2012), highlighting the importance of a model system that permits flow rate regulation.

Recognizing and understanding the complex interplay between flow conditions and biofilm growth is crucial for developing critically needed novel therapeutic approaches against peri-implantitis. These are traditionally evaluated in animal studies, whose increasing numbers raise significant ethical concerns, alongside practical limitations regarding cost and scalability (Blanc-Sylvestre et al., 2021; Faggion, 2015; Faggion jr et al., 2016; Rovida, 2009). An alternative is to implement in vitro biofilm models that replicate key aspects of the oral environment, simulating bacterial colonization of implant surfaces under controlled laboratory conditions.

When examining biomedical treatments such as debridement applications, mechanically robust biofilms are essential for accurate testing. Weakly attached biofilms would be non-specifically removed during routine washing steps of controls, masking genuine treatment effects. Higher rigidity in biofilms, making them suitable for rigorous experimental testing, can be achieved by applying shear stress during cultivation. For instance, Stoodley et al. (2002) demonstrated that subgingival biofilms formed under high-shear conditions exhibited stronger surface attachment and greater cohesive strength than their low-shear counterparts. This principle guided our approach to using our dynamic in vitro model to investigate how varied flow velocities affect the mechanical properties of biofilms on implant surfaces, aiming to identify optimal conditions for debridement testing. The inclusion of six common and representative species from subgingival plaque (Socransky et al., 1998) aimed to reveal the effects of flow on biomass, structural integrity, and species composition in our mixed microbial community.

The central aim of this study was to develop and characterize a dynamic in vitro biofilm model that decouples biofilm flow exposure from bioreactor cultivation conditions. The versatility of this system extends beyond the oral biofilm demonstration, which was presented to illustrate the model’s capabilities. An independent control of hydrodynamic and cultivation parameters enables systematic investigation of flow effects across any biofilm-forming community, from single-species industrial biofilms to polymicrobial infections. By replicating physiological flow conditions while optimizing growth conditions to specific microbial consortia, the model is particularly suited to simulate clinical scenarios on various biomaterials: From low-shear orthopedic implants to high-shear urinary catheters, vascular grafts or prosthetic heart valves. Cultivation conditions can also be adjusted to investigate how nutrient availability or environmental factors modulate biofilm properties under flow. This approach shifts biofilm research from the constraints of coupled or in-line systems towards flexible, physiologically relevant in vitro models.

A six-species bacterial consortium was selected for the in vitro model, representing each stage of the typical colonization sequence in natural oral biofilms and implant-related infections: The early colonizers Streptococcus oralis and Actinomyces naeslundii were incorporated to facilitate initial adhesion to the implant-like titanium surface. As representative middle colonizers, Veillonella parvula and Fusobacterium nucleatum were chosen to establish a bridging network that supports further bacterial recruitment and inter-species metabolic relationships within the developing biofilm. The consortium was completed with the late colonizers and periodontal pathogens Porphyromonas gingivalis and Aggregatibacter actinomycetemcomitans, which integrate into mature biofilm structures and drive inflammatory responses during in vivo peri-implant infections. This consortium has been described previously by Sánchez et al. (2011) and Blanc et al. (2014) as an appropriate species selection to produce representative biofilms under both, static and dynamic conditions.

For the reactor-inoculum, species ratios were adapted from Blanc et al. (2014). However, species concentrations were proportionally increased to achieve an initial OD₆₀₀ of 0.1 in the reactor culture. This elevated inoculum density was implemented to prevent species washout during culture establishment, accounting for the short reactor residence time required to maintain the mixed culture under optimal continuous growth conditions. All selected strains demonstrated successful co-cultivation in the reactor culture and formed stable mixed-species biofilms on titanium surfaces, with species diversity maintained throughout cultivation as confirmed by qPCR detection of all consortium members (see section 3.3).

In the present study, we describe the successful development and validation of a dual-cycle biofilm cultivation system to independently control flow dynamics and cultivation parameters for in vitro biofilms. This approach addresses a fundamental challenge in biofilm research: Due to the recognized importance of fluid flow for the representative imitation of natural biofilm behavior in in vitro models, the state-of-the-art has moved away from purely static models to dynamic systems. Yet existing models compromise between either precise flow exposure of biofilms or optimal regulation of culture conditions. The here presented model overcomes these constraints by decoupling the biofilm formation under defined flow parameters in a MRD from the cultivation of a microbial consortium in a small-scale bioreactor under steady-state conditions. Reducing the reactor volume to a minimum while simultaneously recirculating the MRD flow through back into the system brings the additional advantage of minimizing the required medium throughput for long-term experiments. The power of this in vitro model was demonstrated by cultivating six-species oral biofilms, which emulated peri-implant infections, under low-, moderate- and high-shear conditions. The calculated Reynolds numbers (Re ≈ 20–100) fall within the laminar regime, consistent with physiological flow conditions in the gingival crevice and peri-implant sulcus, where viscous forces dominate fluid movement, reflecting the hydrodynamic conditions relevant for oral implant surfaces (Krastev and Filipov, 2020 for detailed flow rate estimation see Supplementary Material 1). By examining produced biofilms regarding their biomass, nucleic acid content and species composition by qPCR, we could demonstrate that flow velocity distinctly shapes biofilm architecture, robustness and microbial community patterns.

Variation in flow velocities exerted a profound influence across multiple dimensions of the biofilms created with our model. Total biomass exhibited a maximum at low to moderate flow velocities (8–16 cm ⋅ min^–1) and then decreased significantly with increasing flow. Comparable results were demonstrated by Nairn et al. (2024) in a multi-well plate model, where applied shear stress promoted adhesion and biofilm accumulation of single-species S. gordonii and ex vivo plaque biofilms until an optimal threshold, above which further shear stress reduced total biomass.

The observed discrepancy between total biomass (measured with a crystal-violet-based assay) and cellular biomass (determined by DNA content quantification) at moderate flow velocities (16 cm ⋅ min^–1) indicates a differential flow response in EPS production. The high biomass but comparatively low DNA content achieved under moderate flow indicates an enhanced proportion of extracellular components in these biofilms. In accordance with this observation, previous studies have demonstrated how shear stress fosters adhesion and structural integrity of biofilms by stimulating EPS secretion (Hou et al., 2018; Liu and Tay, 2002). A molecular basis for these observations is given by Rodesney et al. (2017), who identified mechanosensory elements in P. aeruginosa that respond to applied shear forces by activating cyclic-di-GMP signaling pathways known to stimulate biofilm formation and EPS production (Liu et al., 2022; Simm et al., 2004; Steiner et al., 2012; Valentini and Filloux, 2016). Moreover, previous studies concluded that an impaired nutrient availability under low shear stress constrains EPS production, resulting in fragile biofilms prone to detachment (Nilsson et al., 2025; Wang et al., 2020). This supports our observations of greater biofilm rigidity under moderate compared to low flow conditions. It has to be acknowledged that direct visualization of biofilm architecture, thickness and EPS distribution by Confocal Laser Scanning Microscopy (CLSM) would complement these functional measurements. However, CLSM analysis of biofilms on rough titanium surfaces presents known technical challenges: The optical properties of the (structured) metallic substrate, particularly opacity and reflection/scattering, limit axial resolution and complicate the localization of the biofilm-substrate interface (Hirai et al., 2025; Kleine et al., 2019). These constraints were confirmed in our preliminary experiments and necessitated the quantitative approach employed here. In future studies we plan to overcome these limitations by complementary techniques such as scanning ion conductance microscopy (SICM) for topographical analysis on opaque substrates, as suggested by Hirai et al. (2025).

The use of qPCR, paired with the de novo development of primers to target the single-copy rpoB gene, allowed us to examine the precise species distribution at varying hydrodynamic conditions. This approach was chosen since the single-copy target allows for a direct translation of quantified DNA to cell abundance and thus maintains comparability between the different species. In contrast to the remaining strains, A. naeslundii, V. parvula and A. actinomycetemcomitans form highly persistent multicellular aggregates, even in planktonic cultures, making a comparison based on CFU not feasible or at least heavily biased, as described previously (Harmsen et al., 2000; Wagner et al., 1993).

Resulting species abundances in our reference biofilms grown under static conditions showed a generally comparable community structure to the static model described by Sánchez et al. (2011), using identical species but on hydroxyapatite (HA) surfaces. The main difference constituted the higher relative abundance of the early colonizers S. oralis and A. naeslundii in our model (comparing equivalent cultivation periods). This could be attributed to the generally higher CFU concentrations in our inoculum, or to surface-dependent colonization dynamics, whereby early colonizers may colonize the rough titanium surfaces more successfully than the smoother HA. Different DNA-based species quantification techniques used in both studies (qPCR vs. T-RFLP) might also introduce a methodological bias. In particular, A. naeslundii is known to resist standard DNA extraction protocols (Barsotti et al., 1987; Li et al., 2020), an issue we addressed in our extraction approach, potentially accounting for its higher abundance in our model. This species-specific extraction bias represents an important methodological consideration for any multi-species biofilm study, when assessing species abundances. Several studies have shown that extraction efficiencies across methods can vary substantially between species, particularly with extraction methods that rely on commercial kits, resulting in altered community compositions (Abusleme et al., 2014; Corcoll et al., 2017; Li et al., 2020). Our own experiments revealed that standard DNA extraction kits produced highly variable recovery across our six species (data not shown). To minimize this bias, we employed a phenol-chloroform based protocol, known to significantly increase DNA recovery (Rosenbaum et al., 2019), combined this with an extensive multi-step lysis to equalize efficiency across species as far as possible. While residual extraction bias cannot be entirely excluded, our data presents a strong correlation of biomass measurements and DNA quantification, as well as consistency with published community compositions. Researchers adapting our system for quantitative multi-species studies should therefore carefully select and optimize the DNA extraction method for their target species.

The introduction of flow resulted in a distinct shift in the biofilm community composition within our model. Lower flow conditions (8 cm ⋅ min^–1) promoted middle-colonizer dominance, characterized by a substantial over-representation of F. nucleatum (> 60%). This predominance aligns with previous observations of in vitro and in vivo studies: Guggenheim et al. (2001) reported F. nucleatum accounting for more than 50% of the total cell load in their static model, while Moore and Moore (1994) identified it as one of the most prevalent isolates from gingival-crevice plaque. More recently, F. nucleatum was identified as the most abundant species which, alongside Veillonella spp., forms part of the shared core microbiome of healthy and diseased peri-implant sites (Belibasakis and Manoil, 2021; Chen et al., 2022).

The elevated abundance of late colonizers observed under lower flow conditions in our model (P. gingivalis: ∼5%; A. actinomycetemcomitans: ∼8%) is potentially facilitated by the established function of F. nucleatum as a bridging organism. Adhering to both early and late colonizers, F. nucleatum enables coaggregation and mutualistic relationships that promote pathogen integration into developing biofilms (Chen et al., 2022). The substantial F. nucleatum presence under low shear exposure likely provided attachment and integration for late colonizer establishment in our model.

The progressive reduction of late colonizers with increasing shear observed in our study can be attributed to multiple factors. A structural analysis of a static 10-species subgingival Zurich biofilm model (Ammann et al., 2012) by fluorescence in situ hybridization (FISH) localized the pathogenic late colonizers, including P. gingivalis, loosely attached in the top layer of biofilms. This spatial distribution renders them particularly vulnerable to detachment under elevated hydrodynamic forces. Additionally, late colonizer establishment requires potentially extended maturation periods: Sánchez et al. (2011) observed low relative abundances for P. gingivalis during initial biofilm development in their static model, reaching peak abundance only after 144 h of cultivation. A cultivation period of 84 h, as in our flow-exposed biofilm model, may have been insufficient for late colonizers to achieve substantial representation under the additional selective pressure of elevated shear stress. In general, a low pathogen abundance is entirely common for in vivo biofilms on dental implants: A systematic analysis of implant infections by Jia et al. (2024) reported relative abundances of Porphyromonas spp. between 6.1% under peri-implantitis and 2.6% under peri-implant mucositis conditions. These proportions align with the keystone pathogen hypothesis, stating that low-abundance pathogens can orchestrate inflammatory diseases by transforming commensal microbial communities into dysbiosis (Hajishengallis et al., 2012). The flow of gingival crevicular fluid (GCF) is closely correlated with the severity of periodontal inflammation, serving as a direct indicator of tissue response; it increases markedly due to enhanced vascular permeability and epithelial ulceration at inflamed sites. While healthy gingiva produces minimal GCF, inflammation leads to a substantial rise in both volume and compositional complexity, transforming it into an exudate characteristic of inflammatory processes (Barros et al., 2016).

The balanced microbial community we observed under moderate flow conditions (16 cm ⋅ min^–1), with early and middle colonizers exhibiting comparable abundances, likely reflects an optimal balance between hydrodynamic forces and bacterial colonization dynamics. As described by Beyenal and Lewandowski (2002), enhanced flow velocities increase the mass transfer, allowing nutrient transport also into deeper biofilm layers. This more uniform nutrient distribution could support the maintenance of early colonizer populations alongside middle colonizers. They further describe how flow velocities beyond a certain threshold reinforce denser biofilm structures to resist shear stress at the expense of internal nutrient transfer, potentially explaining the recurrent suppression of early colonizers under higher flow conditions (32 cm ⋅ min^–1). The reduced diversity under high flow aligns with previous findings that high shear stress reduces biofilm diversity and slows maturation (Rochex et al., 2008). Across all flow rates, the balanced community observed under moderate flow corresponds the closest to in vivo peri-implant microbial profiles: In clinical biofilm compositions reported by Jia et al. (2024), the genera Streptococcus and Fusobacterium exhibited the highest abundances of comparable magnitude, while also Actinomyces and Veillonella presented slightly lower yet relatively similar ratios. However, a direct comparison with in vivo situations remains challenging, due to substantially higher microbial complexity and diversity, infection stage-dependent variations and highly patient-specific compositions (Dieckow et al., 2024; Jia et al., 2024).

We assessed the structural integrity of flow-exposed biofilms by applying mechanical stress in the form of standardized washing steps. This revealed the presence of a persistent basal layer within all biofilms, presenting similar DNA content across all flow rates but significantly enhanced total biomass at moderate flow conditions (16 cm ⋅ min^–1). These results suggest that shear exposure up to a certain threshold not only promotes EPS production but also strengthens the cohesive properties of the extracellular matrix. Consistent with these findings, Gloag et al. (2020) described how shear stress enhances both rigidity and also elasticity of biofilms, proposing an improved interaction between EPS compounds as a shear response. Our observation of a resilient basal biofilm layer has likewise been documented by Simões et al. (2022) for biofilms of P. fluorescens formed under various shear stress conditions. They demonstrated that even after combined mechanical and chemical treatment, a firmly adherent basal layer of predominantly viable cells was retained on the surface, whereby biofilm resilience increased proportionally with shear exposure during formation.

While the mechanical resistance of basal biofilm layers has been well described for environmental or industrial settings (Lemos et al., 2015; Möhle et al., 2007; Paul et al., 2012; Simões et al., 2022), its compositional stability in oral biofilms has received limited attention. The relative species distribution in the basal layer of our model biofilms remained largely uniform across all flow rates. Characterized by a clear middle colonizer dominance, the basal layer showed flow-dependent changes mainly in the progressive reduction of late colonizers with increasing flow, as also observed for untreated biofilms. This suggests that the species composition of this foundational community is less influenced by shear-induced selection, but rather by initial surface colonization dynamics. The presence of middle colonizers such as F. nucleatum directly at the substrate surface has been previously observed by Ammann et al. (2012), who described a stratified structure within their static 10-species subgingival biofilm model. In connection with peri-implant treatment strategies, observing such a resilient basal layer is highly relevant, as it could serve as a persistent reservoir for biofilm reformation after debridement.

Our system successfully demonstrates the production of representative oral biofilms, robust enough to endure potential treatment evaluations. The creation of more resilient in vitro biofilms through adapted flow exposure also facilitates a more realistic assessment of treatment efficacy: Longer maturation periods and higher diversity with numerous interspecies interactions grant in vivo biofilms cross-protections against antimicrobials such as hydrogen peroxide (Dieckow et al., 2024), which can hardly be recreated by static models. Notably, our system achieved stable multi-species biofilms without requiring pooled human saliva, which is used in most oral biofilm models (Darrene and Cecile, 2016), despite its inherent donor-dependent inter-individual variability (Quintana et al., 2009).

While our model enables valuable insights into flow-dependent variations of biofilm development, mechanical properties and microbial communities, the standard qPCR analysis has potential inherent limitations. Quantifying the total DNA content, including non-viable cells, does not allow for a conclusion about the proportion of metabolically active cells. We explored several viability assessment approaches and identified technical challenges, arising from the anaerobic nature of our samples, that should be considered by researchers planning to adapt this system: Without a suitable anaerobic workstation, the exposure of biofilms to oxygen during sampling and processing compromises standard viability measurements. Obligate anaerobic species potentially perish during handling of metabolic activity assays (e.g., tetrazolium reduction) or CFU enumeration. Conventional Live/Dead fluorescent staining faces the additional issue that the sudden availability of oxygen as electron acceptor can elevate the membrane potential of viable facultative anaerobes enough to take up propidium ions (“dead stain”) through intact membranes (Kirchhoff and Cypionka, 2017). Future studies will address this by utilizing an anaerobic workstation to perform PMA-qPCR (“viability” qPCR), as well as species localization by FISH and CLSM. The decoupled flow control of our system will also permit to challenge biofilms with controlled shear pulses or flow ramps, without impairing culture stability. Other potential applications include sequential sampling of biofilms at multiple timepoints to investigate flow-dependant colonization kinetics and species succession, as well as the customization of flow conditions to resemble the dynamic flow changes in the oral cavity throughout the day. This could even improve biofilm formation, as Tsagkari et al. (2022) demonstrated that non-uniform flow patterns can enhance biomass accumulation.

Overall, our dual-cycle approach provides a versatile platform, with the ability to independently control flow exposure and cultivation conditions. This approach addresses specific limitations experienced with existing dynamic biofilm systems while expanding experimental flexibility: The CDC biofilm reactor employs rotational stirring to create a flow that is characterized by unsteady turbulent structures and produces a non-uniform shear exposure, whereby exact flow characterization requires computational fluid dynamics modeling (Johnson et al., 2021). In comparison to this, our MRD configuration provides uniform, quantifiable flow across all sample surfaces, with a gentle tilting motion to prevent gas and debris accumulation while preserving laminar flow characteristics. Traditional in-line MRD-chemostat configurations couple the flow rate through the biofilm device directly to the culture efflux (Blanc et al., 2014; Jass et al., 1995), creating a fixed relationship between biofilm shear exposure and culture turnover rate. This coupling restricts experimental flexibility, since the flow through cannot be fully regulated without altering steady-state culture conditions. Maintaining steady-state culture conditions, however, is critical for physiological relevance, as these conditions are required for bacteria to express cell wall proteins and respond to environmental stimuli similarly to in vivo cells, as opposed to batch-grown cultures (Drake and Brogden, 2014). The parallel recirculating design presented here eliminates this trade-off, enabling independent modulation of both parameters across wider experimental ranges. This parametric independence also makes high-shear studies feasible: An in-line system, connected to either a chemostat or a buffered batch culture, which operates at the flow velocities examined here (8–32 cm⋅min^–1) would consume 50–80 L of medium/culture over 84 h, compared to ∼4 L with our recirculating configuration where medium consumption is independent of the MRD flow velocity. The low media consumption of our system is also supported by the small-scale reactor volume (50 mL culture volume compared, e.g., to 350 mL minimum working volume for CDC reactors) that further reduces the required medium feed to maintain steady-state conditions for fastidious anaerobes. This efficiency is particularly important for studies that require costly culture media supplements. Beyond this resource efficiency, the dual-loop design provides new experimental degrees of freedom that were impractical with previous systems. In future studies, multiple MRDs can be connected in parallel to a single chemostat (using separate pumps), allowing the simultaneous examination of different flow conditions while excluding variability from different culture conditions that might occur across independent experimental runs. These capabilities make this configuration readily adaptable to various scenarios. Culture conditions can be altered by nutrient addition or limitation to mimic specific stages of infections or inflammatory responses. The system provides a generalizable tool to simulate diverse biomaterial-associated infections, including urinary catheters, orthopedic implants, vascular grafts or prosthetic heart valves.