Ingestible active capsule for gastrointestinal microbiome sampling
Mohammed Hadi Shahadha, Andreas Voigt, Denise Gruner, Uwe Marschner, Maxime Le Floch, Jochen Hampe, Frank Brauer, Sebastian Schostek, Marco Luniak, Karlheinz Bock, Andreas Richter
TL;DR
A new ingestible capsule can collect gut microbiome samples from specific parts of the gastrointestinal tract, improving the accuracy of microbiome studies.
Contribution
The development of a modular, wireless sampling capsule that enables segment-specific microbiome collection in variable gastrointestinal conditions.
Findings
The capsule's actuator module successfully collected and preserved liquid samples in vitro using a stool simulant and porcine intestine.
The opening valve activated within 3.6 ± 0.5 seconds, and the closing valve sealed the inlet completely within 10 minutes.
The design functions effectively in liquids with varying pH, chemical composition, viscosity, and particle size.
Abstract
Various gastrointestinal disorders have been linked to gut microbiome dysbiosis, as it plays a critical role in immune regulation, metabolism, nutrient digestion, and pathogen suppression. However, the microbiome’s spatial variability across gastrointestinal segments and its intra- and interindividual differences complicate its study and clinical interpretation. While fecal DNA analysis is commonly used, stool samples only capture an accumulated signal and miss the spatial dynamics of microbial populations. To address this, we propose a modular sampling capsule capable of wirelessly collecting liquid. The capsule consists of two main modules: (i) an actuator module integrating a polymer-based microfluidic system with meltable wax-based opening valve, screen-printed microheater, cellulose membrane-based closing valve, evacuated sampling chamber with dried sample preservative material,…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 10
Figure 11
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Figure 9
Figure 12- —Technische Universität Dresden (1019)
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsColorectal Cancer Screening and Detection · Bacterial Identification and Susceptibility Testing · Gut microbiota and health
Introduction
The gut microbiota is distributed throughout the entire 7–9 m long gastrointestinal (GI) tract, colonizing both the digesta and the mucus layer (Donaldson et al. 2015; Ursell et al. 2012). Its composition varies between the gastrointestinal segments in terms of microbial abundance, species diversity, and functional capabilities, playing a critical role in maintaining gastrointestinal and metabolic health (Arumugam et al. 2011; Del-Rio-Ruiz et al. 2024; Folz et al. 2023; Leite et al. 2020; Procházková et al. 2024; Ursell et al. 2012). This composition is highly dynamic and exhibits significant inter- and intraindividual variability, which makes it challenging to study and interpret clinically (Arumugam et al. 2011; Del-Rio-Ruiz et al. 2024). Disruptions to the normal microbial balance, known as dysbiosis, have been linked to the pathogenesis of various disorders, including inflammatory bowel disease (IBD), irritable bowel syndrome (IBS), small intestinal bacterial overgrowth (SIBO), type 2 diabetes mellitus, colorectal cancer, metabolic syndrome, obesity, immune dysregulation, and gastrointestinal infections (Arumugam et al. 2011; Hua 2020; Ursell et al. 2012; Waimin et al. 2020). The gut microbiome is not static; it can be modulated by external factors such as antibiotic use and dietary composition (Jang et al. 2020; Ursell et al. 2012). Notably, microbes ingested through food can contribute to the diversification of the gut microbiome, potentially enabling novel metabolic functions, including the digestion of previously indigestible foods (Ursell et al. 2012). In addition to its roles in digestion, immune regulation, and gastrointestinal diseases, the gut microbiome has recently been studied for its associations with psychoneurological disorders, such as autism, stress, anxiety, depression, and mood instability, as well as its relationship with dietary patterns and nutritional health (Foster and McVey Neufeld 2013; Jayachandran et al. 2020; Mayer et al. 2014; Naseribafrouei et al. 2014; Rehan et al. 2022). A study has also shown that the composition of the gut microbiota and microbiota-dependent metabolism, such as enzyme activity, can significantly be affected by gut physiology and environmental conditions, such as segmental transit time and gut pH (Procházková et al. 2024). The small intestine harbours key microbial species that are essential for nutrient digestion and absorption as well as immune function, which underscores the importance of characterizing microbial composition and function in this segment of the GI tract (Arumugam et al. 2011; Leite et al. 2020; Ying et al. 2023). Overall, a comprehensive analysis of gut microbiota, including its composition, spatial distribution, and functional attributes, is essential for advancing diagnostics in gastrointestinal health. Moreover, the gut microbiome holds promise as a biomarker for disease detection and monitoring, and as a potential target for therapeutic interventions (Sarnaik et al. 2025).
Gut microbiota profiling is commonly performed using molecular diagnostic assays targeting fecal DNA or via fecal metabolite analysis (Del-Rio-Ruiz et al. 2024; Kim et al. 2020; Nechvatal et al. 2008). Stool sampling remains the standard method due to its non-invasive nature, ease of collection, and low cost. However, this method provides a cumulative output of the entire gut, which precludes site-specific sampling of microbiota along different GI segments (duodenum, jejunum, ileum, colon) and limits the detection of microbes from the upper gastrointestinal tract (Cummins 2021; Eckburg et al. 2005; Hillman et al. 2017). This reduces spatial resolution and the clinical utility of microbiome data (Wu et al. 2019). Additionally, previous studies have shown that the microbiome profile in the stool samples significantly differs from that in the intestinal liquid samples (Wang et al. 2024). Therefore, to improve diagnostic accuracy, localized sampling from specific gut regions is essential, while avoiding downstream contamination. Although procedures such as colonoscopy and gastroscopy enable targeted sampling, they are invasive and limited in reach due to the gut’s length (7–9 m) and variable diameter in the gut. They also require clinical infrastructure (Chen et al. 2020; Rehan et al. 2022; Valdastri et al. 2012; Waimin et al. 2020). Recently, ingestible smart capsules have emerged as a minimally invasive alternative that allows for site-specific microbiota collection, reducing patient discomfort and eliminating the need for specialized clinical settings.
A capsule designed for site-specific gut microbiota sampling must accomplish three key steps: (1) aperture (inlet) opening, (2) gut liquid ingress, and (3) inlet closure. Recently, numerous autonomous, battery-free sampling capsules have been developed to improve the collection of microbiota and biomarkers for gastrointestinal diagnostics. As an opening mechanism, many of these capsules utilize degradable, pH-sensitive enteric coatings to seal the capsule aperture, which dissolve at specific regions of the GI tract characterized by particular pH ranges (5.4–7.4 in the small or large intestine) (Nejati et al. 2022; Rezaei Nejad et al. 2019; Salem et al. 2022; Sarnaik et al. 2025; Waimin et al. 2020). However, variability in gut pH between individuals may alter the dissolution time of the enteric coating, which could compromise precise, site-specific opening (Sarnaik et al. 2025). Furthermore, the pH-dependent dissolution mechanism requires continuous exposure to gut liquids, which may be hindered by air bubbles or undigested food residues covering the aperture, thereby delaying or preventing the opening process altogether.
For optimal sampling, efficient and immediate liquid ingress following aperture opening is critical. Various mechanisms such as capillary action, osmosis, and imbibition via superabsorbent materials have been employed to drive liquid into the capsule (Nejati et al. 2022; Rezaei Nejad et al. 2019; Salem et al. 2022; Waimin et al. 2020). All the passive capsules referenced above use these slow, gut-environment-sensitive methods of driving liquid. In contrast, vacuum-driven liquid aspiration offers faster and more reliable sampling than other methods where liquid transport may be impeded or delayed significantly due to the viscous and heterogeneous nature of gut chyme. Recent articles have employed vacuum either by pressing a soft sampling chamber or integrating a pressure controller in the capsule (Ding et al. 2021; Du et al. 2018). However, a pressed soft chamber might not provide sufficient suction pressure to aspirate highly viscous intestinal liquids, and integrating a pressure controller increases the cost and energy demand of the capsule.
Immediate closure of the inlet following liquid sampling is also an essential step to ensure that the collected sample originates exclusively from the targeted region of the GI tract and is not contaminated by contents from distal sections. Previous studies have explored passive closure mechanisms employing superabsorbent, swellable, gut environment-sensitive materials such as hydrogels (Nejati et al. 2022; Salem et al. 2022; Waimin et al. 2020). While these materials offer simplicity and low-cost fabrication, they are subject to limitations, including variability in swelling behavior and closure time. These inconsistencies primarily arise due to the highly variable physicochemical properties of gut chyme such as pH, viscosity, and particulate composition, which differ significantly between individuals. Furthermore, the large apertures (diameter 2–5 mm) (Nejati et al. 2022; Salem et al. 2022; Waimin et al. 2020) used in these autonomous capsules allow large food particles to enter the sampling reservoir, obstructing inlet closure through the swellable material. This could lead to contamination from downstream sections, which would compromise the spatial precision of microbiota sampling. In conclusion, while passively actuated capsules are easy to fabricate and cost-effective, they may perform unreliably under the dynamic and heterogeneous conditions of the gastrointestinal environment.
To achieve optimal, cost-effective, and precise site-specific sampling of the intestinal gut microbiota, we addressed, in this work, the aforementioned technical challenges by envisioning a modular capsule system based on a fluidic microsystem (Fig. 1). The device comprises two main modules: (i) an actuator module incorporating the sampling reservoir and miniaturized controllable active opening and passive closing mechanisms, and (ii) an electronic control module providing power, gastrointestinal segment localization signal, and wireless communication with an external receiver. The main goal of this work is to consider design, actuation, and material aspects to design, fabricate, and validate the actuator module that can efficiently perform opening and liquid aspiration processes as well as complete immediate reliable closure, under the varying physicochemical conditions of the intestinal liquids (pH, viscosity, chemical composition, and particle size). This will eventually ensure a precise-site-specific sampling of intestinal liquids. We mean with precise-site specific that the sampling process time including capsule opening, liquid aspiration, and capsule closing should be instant and reproducible under various physicochemical properties upon activation to ensure liquid sampling at the activation site before the capsule is transited into a distal region in the small intestine. Also, for this purpose, fast localization and communication modules for sampling activation should be incorporated in the controlling electronic module. The capsule can serve as a research tool for collecting high-quality intestinal samples and investigating various intestinal biomarkers for gastrointestinal diagnostics, such as metabolites, epithelial cells (cancer biomarkers), pus, and red blood cells, in addition to microbiota.
As illustrated in Fig. 1, the opening mechanism uses a thermally meltable wax valve (depicted in green) surrounded by a screen-printed microheater (highlighted in magenta). Once the targeted intestinal segment is reached, the microheater is activated wirelessly by applying current via the control module, generating localized heat that melts the wax valve. Due to the pre-applied vacuum in the sampling reservoir, the intestinal liquid is aspirated immediately upon valve opening. We completely evacuate the rigid sampling chamber to a pressure of 100 mbar, enabling an inexpensive and efficient aspiration method. This ensures liquid aspiration at a flow pressure of 900 mbar even with viscous intestinal liquid samples. Using a controllable wax-based opening mechanism that operates independently of the chyme’s physicochemical nature, and can be activated within seconds, enhances the precision and repeatability of site-specific sampling. Capsules with active controllable opening mechanisms based on heaters and meltable materials such as wax or low-melting-point metals have already been developed (Ding et al. 2021; Du et al. 2018). However, using expensive metal-based heaters such as nichrome wire, which are integrated through pick-and-place assembly, makes the fabrication costly and increase integration time. Here, we use screen-printed heaters (printed silver particle paste printed on a substrate) because they enable precise, parallel, high-throughput production, and provide a fast, inexpensive method of in-capsule integration.
The closing mechanism is based on a miniaturized, swellable cellulose membrane (shown in orange), that is integrated into a 3D fluidic channel. Upon contact with the intestinal liquid, the membrane undergoes vertical expansion, effectively sealing the inlet to the sampling reservoir. Importantly, the cellulose membrane is chemically stable, which minimizes the influence of chyme chemical variability on the swelling time of the membrane (Vázquez-Rivas et al. 2025). The material and design parameters of the closing and opening mechanisms ensure consistent and reliable sealing and opening processes across the diverse physiochemical conditions of the intestinal chyme.
To prevent large food particles from obstructing the fluidic system or interfering with the actuation mechanisms, we integrated a filter membrane (indicated in grey) into the capsule’s aperture. To our knowledge, this is the first time this has been done in an ingestible capsule. For sample stabilization, particularly to preserve microbial DNA and RNA, we also integrated dried preservative material (shown in dark blue) into the sampling reservoir. This material dissolves upon contact with the collected intestinal liquid, ensuring the sample is fixed immediately. Once the capsule has been excreted and retrieved, the sample is extracted via a syringe through a designated extraction channel for microbiome diagnostics and profiling. Table 1 summarizes the advantages of the envisioned capsule relative to previously reported capsule technologies.
Table 1. Comparison of reported passive and active intestinal liquids sampling capsules mentioned above with the proposed capsule, outlining key design, actuation, fabrication, and material considerations affecting reliable and reproducible in vivo, site-specific intestinal liquid samplingCategoryPassive capsulesActive capsulesOur capsuleKey advantage of our capsuleOpening valveMaterialpH- and physicochemical variations-sensitive dissolvable materialsMeltable material (wax)Meltable material (wax)Wax is insensitive to pH and intestinal liquid physicochemical variability, enabling reproducible opening timesDesign & IntegrationpH-sensitive dissolvable material integrated directly in lateral or central capsule´s aperturepH-sensitive dissolvable material integrated directly in lateral aperture or separatedWax separated from central capsule apertureReduced interaction of the material with intestinal liquids improves actuation reliability and response timeActivationPassive dissolutionNichrome wire or metal-based macro-heaterMiniaturized screen-printed microheaterLower power consumption and precise activation.ManufacturabilityLimited design flexibility and stabilityBulky heater limits integrationOn-foil screen-printed heaterScalable, low-cost, and robust integrationClosing ValveMaterialPassive superabsorbent hydrogelsPassive superabsorbent resin ballsPassive swellable pressed cellulose membraneHigher mechanical and chemical stability; less sensitivity to physicochemical variationsDesign & IntegrationLarge superabsorbent material in sampling chamberLarge superabsorbent material in fluidic channelMiniaturized cellulose membrane in fluidic channelFast swelling and reliable sealing even if very small liquid volumes with air bubbles are aspirated into the capsule; secured contact with liquidActivationBigger liquid volume requiredLess liquid volume requiredLeast liquid volume requiredManufacturabilityRequires complex fabrication process of superabsorbent materialsRequires complex fabrication process of superabsorbent materialsCommercially availableCheap filter membrane that can be easily structured through punching or laser machiningLiquid AspirationCapillary, osmosis, and hydrogel imbibition forcesVacuum via pressed elastic chamber or electronic controllerPre-evacuated sealed chamber (~ 100 mbar)Strong, reproducible aspiration across variable liquid viscositiesParticle FiltrationNot availableNot available150 μm filter membrane at aperturePrevents food particles interference with the opening and closing mechanismsSample PreservationNot availableNot availableIntegrated dried preservativeMaintains sample DNA/RNA integrity
Fig. 1. Modular capsule system for site-specific sampling of microbiota-containing intestinal liquid. The capsule is composed of two functional modules: (i) an actuator module, which integrates a wax-based, controllable opening mechanism, a passive closing mechanism based on a swellable cellulose membrane, a preloaded sampling reservoir containing dried preservative, and a filter membrane designed to prevent large food particles from entering the intestinal chyme; and (ii) an electronic control module, responsible for power supply and wireless communication with an external receiver. Upon arrival in the small intestine, the microheater is wirelessly activated via the control module. Current is supplied to the microheater, generating localized heat that melts the wax and opens the valve. Due to the negative pressure in the evacuated sampling reservoir, the intestinal liquid is immediately aspirated, dissolving the integrated preservative to stabilize the collected sample. Subsequently, the cellulose membrane absorbs the intestinal liquid and swells, passively seal the reservoir and prevent further intake or contamination. After gastrointestinal transit, the capsule is excreted and retrieved. The liquid sample is then extracted from the capsule via a dedicated extraction channel using a syringe. Samples are then processed for microbiome profiling and diagnostics using molecular techniques such as polymerase chain reaction (PCR) and next-generation sequencing (NGS). Image of GI tract taken from “Complete digestive apparatus” by Servier Medical Art https://smart.servier.com/smart_image/complete-digestive-apparatus/
License: CCBY4.0 https://creativecommons.org/licenses/by/4.0/.
Materials and methods
Opening mechanism design and fabrication
A normally closed, actively actuated, meltable opening valve was designed and manufactured to enable the wireless activation and sampling of intestinal liquid within the small intestine (Fig. 2a-c). It is based on solid wax integrated in a hole (H 325 μm x Ø 600 μm) surrounded by a screen-printed microheater with an electrical resistance. This microheater separates two adjacent microfluidic channels (W 800 μm, H 425 μm). In its default state, the valve remains sealed due to solid wax obstructing the flow of liquid and gas from the valve inlet to outlet. Upon activation, an electrical current is applied to the embedded microheater generating heat and inducing wax melting. This phase change opens the flow path, allowing controlled passage of liquids or gases. Complete melting of the wax at the edges of the hole is necessary to optimally open the valve and collect the required amount of intestinal liquid. If the microheater does not generate enough heat, however, only partial melting of the wax may occur, resulting in small openings in the wax hole. This increases the hydrodynamic resistance when aspirating the intestinal liquid, and the small particles in the intestinal liquid may also block these openings completely. The valve outlet of the whole capsule is connected to a sampling reservoir that is evacuated to a pressure of 100 mbar. Consequently, the wax is subjected to a pressure difference of 900 mbar between valve inlet and outlet. Therefore, the solid wax must be stable under this high-pressure difference, and it must be thoroughly sealed in order to maintain the vacuum in the sampling chamber.
The valve is constructed from multiple layers that are stacked and structured using a compact laser micromachining system RDX500 (Pulsar Photonics GmbH, Herzogenrath, Germany). The layers are then assembled and bonded together using a 25 μm-thick acrylic adhesive film (Adhesive 9969, 3 M, Minnesota, United States) (Fig. 2a). The bottom layer is made of 175 μm thick poly (methyl methacrylate) (PMMA, Plexiglas 478987, König Folienzentrum GmbH, Wuppertal, Germany) and serves as a cover of the channels in the second layer. The second layer is a 375 μm-thick PMMA layer (Plexiglas 478989, König Folienzentrum GmbH) laminated with adhesive film on both sides (total layer thickness of 425 μm), containing the 800 μm-wide valve inlet channel. The third layer is made up of 300 μm-thick PMMA (Hesaglas, Topacryl AG, Schönenwerd, Switzerland) that has been laminated with an adhesive film on its underside only, giving a total thickness of 325 μm. This layer contains a microstructured hole (Ø 600 μm) filled with wax (Tudamelt 60/62 321 DAB 61 °C) of melting temperature of 61 °C (indicated in green, see Fig. 2a-c). The wax was first melted up with hot air at 80 °C on a spatula, then manually integrated into the wax hole while the PMMA layer was heated to 65 °C. Any remaining wax on the layer was cut off using a sharp scalpel. To ensure an optimal sealing of the wax valve, the wax was melted at 60 °C for 1–4 min (depending on chip thickness) in a reflow process, which sealed any potential openings between the wax and the edge of the hole. This layer also incorporates an integrated star-shaped microheater (Ø 2 mm, theoretical line width in the mask of 150 μm), which was patterned using screen printing technology with silver paste (LTR3501, Heraeus Electronics, Hanau, Germany) and hardened at 80 °C for 60 min (indicated in magenta, Fig. 2a). To avoid generating high temperatures in this region, the resistance line between the conducting pads and the star-shaped heater was dimensioned to be 250 μm wide in the mask. The microheater was printed using a SPM model screen printer (MPM, Massachusetts, United States) with the following parameters: a squeegee pressure of 60 N, squeegee speed of 15 mm/s, and a snap-off distance of 700 μm. The screen mask was made of SD 75/36 stainless steel mesh (mesh size 75 μm, wire diameter 36 μm, Spörl KG, Sigmaringendorf, Germany). The pattern of the microheater was created on the mesh using photolithography and the MS 40 capillary film resist (thickness 40 μm, Murakami, California, United States). The fourth layer is made using the same stack of layers as the first layer but is structured to form a valve outlet channel (W 800 μm, H 425 μm) and a chamber (Ø 2.3 mm) that covers the microheater and the conducting pads. The final top layer is made of 175 μm-thick PMMA, that is laminated with adhesive acrylic film on both sides and contains structured valve inlet and outlet holes (Ø 800 μm), as well as small chambers (W 1.36 mm x L 1.95 mm) for housing the microheater conducting pads (total layer thickness 225 μm). The opening valves were characterized by using a microfluidic measurement station as presented in Kutscher et al. (2023)(Kutscher et al. 2023). The 3D morphology, the height, and the width of the star-shaped microheater were characterized using a confocal microscope (µSurf, NanoFocus AG, Oberhausen, Germany). Electrical and functional characterization of the microheater was conducted using a power supply, and the generated heat was measured using a thermal camera (VarioCAM HR, InfraTec GmbH, Dresden, Germany) and IRBIS 3 Plus software.
The opening time of the valve can be controlled by various geometric and operational parameters. For example, the distance between the wax and the microheater affects the efficiency of heat transfer, thereby influencing the melting rate upon activation. Additionally, the volume of the integrated wax significantly impacts the valve response time, as larger volumes require more time to melt completely. The melting point of the wax is another critical factor: waxes with lower melting temperatures facilitate faster phase transitions, resulting in quicker valve actuation. Among the operational parameters, the amount of heat generated by the microheater significantly influences the wax melting rate. Higher electrical currents result in increased heat generation, thereby accelerating the melting process and reducing the valve’s response time.
Fig. 2. Opening mechanism. (a) CAD representation illustrating the design and assembled layers and elements of the opening valve. Grey – structured PMMA layers, Magenta – screen-printed microheater, Green – solid wax integrated in a hole. (b) Fabricated opening valve showing all structured layers aligned and bonded together. The circular structures in the microheater conducting pads represent holes serving for electrical connection of the microheater with the power supply module integrated in the capsule. (c) Wax integrated in a hole surrounded by screen-printed microheater
Closing mechanism design and fabrication
A cylindrical shaped and passively actuated closing valve (Fig. 3a-d) was engineered and built using a multilayer stacking and bonding method with the four layers described in Sect. 2.1. The valve comprises three fundamental structural elements: a 1.88 mm diameter valve seat, patterned in the 425 μm-thick second layer; a cellulose-based swellable membrane (Ø 1.8 mm–1.7 mm), manually mounted in its dried state in the valve seat, and a valve inlet hole (Ø 800 μm) patterned in the 325 μm-thick third layer, which covers the valve seat. First, a cellulose membrane (390 μm thick, pore size 6 μm, Whatman 3, Cytiva, Marlborough, United States) with dimensions 25 mm x 25 mm was pressed with an embossing machine at 32 kN down to a thickness of 200 μm (± 10 μm). Then, the adhesive film (Adhesive 9969, 3 M, Minnesota, United States) was laminated onto it and pressed at 4 kN. This composite layer was then cut and structured using a laser micromachining system. The valve inlet hole, which has a diameter 800 μm, is smaller than the valve seat´s diameter and is aligned over the cellulose membrane, leaving a distance to the membrane surface (step size, st). In this state, the valve is at its inherently open state. When a liquid is applied, the membrane absorbs the liquid at the membrane-liquid interface (liquid contact area, ca.), which initiates vertical volumetric expansion via osmotic swelling. This expansion continues until the membrane completely occludes the valve inlet hole, thereby obstructing liquid transit.
In order to ensure sustained valve closure, the swollen membrane must exert a counteracting hydrostatic pressure that is equivalent to the applied flow pressure. The actuation dynamics of the valve, specifically its time to closure, are controlled by multiple interdependent parameters. These include the geometric dimensions of the system, particularly the membrane height (Hm) and the valve seat height (Hvs), as well as the step size (st) and the intrinsic swelling kinetics of the cellulose membrane. The step size (Fig. 3d) is defined as the vertical distance between the valve inlet hole and the membrane, and is determined by the relative heights of the membrane and the valve seat. A larger step size requires a greater vertical volumetric expansion of the membrane in order to achieve complete closure of the valve inlet. The intrinsic swelling kinetics of the membrane are characterized by parameters such as the rate of volumetric expansion, which is influenced by the cooperative diffusion coefficient (D_coop_) of the cellulose polymer matrix, resulting from the membrane´s porosity and degree of hydrophilicity. Another parameter is the contact area (ca.) between the cellulose membrane and the liquid. For example, for a membrane with a given volume, the greater the contact area, the higher the absorption rate is and the faster the valve will close. Furthermore, the physicochemical properties of the liquid process medium, such as its dynamic viscosity and the size of additional particles in the liquid, significantly influence the swelling behavior of the membrane by modulating the kinetics and extent of actuation. For the valve to function optimally, the contact area must be large enough to allow the membrane to absorb the required volume of intestinal liquid within a reasonable time, enabling complete vertical volumetric expansion and, eventually, complete closure of the valve.
To improve the valve’s resistance to high flow pressures, an acrylic adhesive layer (Adhesive 9969, 3 M, Minnesota, United States) with a thickness of 25 μm can be laminated onto the surface of the cellulose membrane (highlighted by the blue line, Fig. 3d). During vertical volumetric expansion of the membrane, the adhesive layer is compressed against the valve inlet hole, resulting in a secure stable seal. As illustrated in Fig. 3d and e, the adhesive layer limits contact with the process medium (highlighted by the green line) primarily to the periphery of the membrane. To increase the contact area between the membrane and the applied liquid, the membrane-adhesive composite can be inverted so that the side not covered with adhesive faces the valve inlet hole as shown in Fig. 3e.
Fig. 3. Closing mechanism. (a) CAD representations illustrating the assembled key components of the closing valve, including the valve seat and inlet and outlet channels (grey), cellulose membrane (orange) with a laminated acrylic adhesive layer (blue), and the valve inlet hole (grey). (b) Fabricated closing valve demonstrating the integrated cellulose membrane with laminated adhesive layer (white). (c) A closing valve filled with an ink solution showing how the membrane absorbs the ink solution and vertically expands to close the valve inlet hole. (d and e) CAD Cross-sectional views of the valve highlighting the structural components, geometric design parameters and the contact area with the liquid (indicated by the green line). Design parameters include the height of the valve seat (Hvs), the height of the membrane (Hm), and the step size (st). A blue line indicates the adhesive layer on the cellulose membrane
Capsule design, assembly and fabrication
We envisioned a modular capsule concept (Fig. 4, L 30 mm x Ø 11.4 mm) comprising two modules: (1) an actuator module integrating mainly opening and closing mechanisms and the sampling reservoir, and (2) a controlling electronic module housing the control and communication systems. The actuator module was designed using computer-aided design (CAD) software (Inventor Professional 2025) and is composed of four distinct parts. The largest component (part 1, L 5.3 mm x Ø 11.4 mm) was fabricated via 3D printing (Phrozen Sonic Mini 8 K Resin 3D Printer, Hsinchu City, Taiwan) using colorless, water washable resin (Anycubic, Shenzhen, China). This part includes the sample reservoir (H 6.5 mm x Ø 4 mm, theoretical volume 133 µL), the microfluidic structures for closing and opening valves, and a port for subsequent intestinal liquid extraction (Ø 1.3 mm). In addition, it incorporates holes (Ø 1.2 mm) for alignment with other capsule parts as well as holes (Ø 1.2 mm with constriction of a diameter of 0.55 mm) for integrating contact springs (indicated in red, model UEBK-13296, Uwe electronics GmbH, Unterhaching, Germany). A vacuum pipe (highlighted in dark blue) connects to the sample reservoir through an aperture (Ø 0.8 mm) to evacuate the reservoir. First, a volume of 80 µL of stool sample preservative (OMR-205, DNAgenotek Inc, Ottawa, Canada), which preserves microbial DNA and RNA, was loaded into the reservoir chamber and then dried at 40 °C for 24 h, or at 60 °C for six hours. These dried chemical preservatives will be later dissolved in the intestinal liquid sample after its collection. Finally, the cellulose membrane (H 0.2 mm x Ø 1.7 mm, indicated in orange) with laminated acrylic adhesive film was manually integrated and bonded to the bottom of the valve seat (H 0.43 mm x Ø 1.8 mm).
As described in Sect. 2.1, part 2 consists of multilayered substrates that are structured and bonded together. It incorporates mainly the wax-filled hole (H 0.325 mm x Ø 0.6 mm, indicated in green), the star-shaped, screen-printed microheater (Ø 2 mm, shown in magenta), a chamber to cover the microheater (H 0.425 mm x Ø 2.3 mm), the conducting pads (W 1.36 mm x L 1.95 mm x H 0.425 mm), the microfluidic channels (W 0.8 mm, H 0.425 mm), a chamber (H 0.755 mm x Ø 2 mm) to house the vacuum pipe, and structures for the closing and opening valves. First, part 2 was aligned with part 1 using the three notches (radius 0.6 mm) at the edge of part 2, the alignment holes in part 1 and the alignment pins in part 3. Then, part 2 was manually pressed into place. To optimize the sealing of the wax valve, the two assembled parts were heated at 60 °C for four minutes on a hot plate to allow wax to reflow. Through-holes (Ø 0.8 mm) in the center of the microheater’s conducting pads established an electrical connection between the microheater and contact springs via a conductive adhesive (indicated in olive green color, Elecolit 336, Panacol, Steinbach, Germany). First, the small chambers housing the contacting pads were filled with a conductive adhesive, after which the contact springs were inserted through the holes in part 1. The conductive adhesive was then fully cured at 40 °C for three hours. The vacuum pipe (indicated in dark blue, TYGON-ND 100 − 80, Saint-Gobain, Courbevoie, France) was then inserted into the designated hole in part 1, and the interface between the hole´s wall and the pipe was sealed with an epoxy resin (indicated in golden yellow, 5-minute epoxy, R&G Faserverbundwerkstoffe GmbH, Waldenbuch, Germany). The vacuum pipe was then used to evacuate the sampling reservoir for three minutes at a pressure of 100 mbar, after which it was thermally sealed using a hot tweezer (heated with hot air at 200 °C for 25 s). Finally, to ensure the pipe was securely sealed, it was completely covered with the 5-minute epoxy.
Part 3 is manufactured using 3D printing with TR-300 Ultra-High Temp resin (Phrozen, Hsinchu City, Taiwan). It contains a central inlet channel (H 4.1 mm x Ø 1 mm) covered by a 3D-printed filter membrane, a lateral extraction channel (H 3.7 mm x Ø 1.2 mm) for collecting liquid samples, three alignment pins (H 3 mm x Ø 1 mm) for aligning with part 1 and part 2, and cavities (H 1.3 mm x Ø 1.15 mm) for housing the contact springs ends. It also contains a cavity (H 2.5 mm x Ø 2 mm) for housing the vacuum pipe covered with the 5-minute epoxy. The filter membrane has a diameter of 2.6 mm and a thickness of 0.275 mm and features slit-shaped apertures with a width of 150 μm. It is fabricated via 3D printing together with a ring-shaped spacer of 0.825 mm height and a ring thickness of 0.33 mm. As shown in Fig. 4, the membrane is mounted in a dedicated chamber (H 0.835 mm x Ø 3.5 mm) using a press-fit fastening method. This leaves a clearance of 0.550 mm from the capsule’s inlet channel. The membrane prevents large food particles from entering the system, which could potentially obstruct the fluidic channels or interfere with the opening and closing mechanisms. To prevent the extraction channel from being filled with gastric or colonic contents, it is sealed with an adhesive film (a 175 μm-thick PMMA with laminated acrylic adhesive film, diameter 1.6 mm). This ensures that the collected sample originates specifically from the small intestine.
Part 4 functions as a DC-DC converter PCB, which is placed into a 3D-printed adapter (H 2.31 mm x Ø 11.4 mm, colorless water washable resin, anycubic, Shenzhen, China) and fixed using the 5-minute epoxy resin. The PCB converter is aligned in the adapter via two alignment pins (H 4 mm x Ø 1.05 mm) incorporated in the adapter structure, and two notches (radius 0.6 mm) on the edge of the PCB converter. The adapter with the mounted PCB converter was then aligned and fixed with 5-min epoxy with part 1 allowing alignment and physical connection of the converter’s conducting pads with the contact springs. The DC-DC converter converts the 4.65 V from the button cells in the electronic control module to 0.8 V and is based on the Texas Instruments TPSM82821SILR integrated circuit. Conversion efficiencies of about 75%−80% can be achieved. This enables currents of up to approximately 80 mA to be used for the heater, while connecting the button cells directly to the heater might result in too high current being drawn. The PCB contains a BM23FR0.6-16DS-0.35 V socket that connects to the electronic module that contains the main electronics via power lines and an activation signal line. A separate PCB with a BM23FR0.6-16DS-0.35 V interface was manufactured as a test board for the actuator module. This board can be operated by button cells or a power supply unit and enables push-button activation of the opening valve.
The target design parameters for the electronic module include a battery lifetime exceeding 6 h, which is sufficient to allow transit through the small intestine, and periodic communication with an on-body external receiver, for example at one-minute intervals. The current dimensional target for the integrated capsule is a length of 30 mm and a diameter of 11.4 mm, which is considered to represent the upper limit for safe swallowability. The capsule mass is expected to be below 8 g; however, given the intended functionality, the exact mass is not a critical parameter.
Fig. 4. Design and assembly process of the actuator module. (a) Computer-aided design (CAD) representation of the capsule (Ø 11.4 mm, L 30 mm) and its four constituent components, along with the assembly sequence of the actuator module. Part 1 comprises the sampling reservoir (theoretical volume 133 µL), microfluidic channels and apertures for the opening and closing valves, three alignment holes, a liquid sample extraction port, a vacuum pipe inlet, a swellable cellulose membrane that functions as a closing valve (depicted in orange), and spring contacts (shown in red) that provide electrical connection of the DC-DC converter (part 4) with the microheater embedded in part 2. Part 2 is fabricated from structured multilayer components bonded with an acrylic adhesive. It includes the opening valves consisting of a screen-printed microheater (highlighted in magenta) and solid wax (highlighted in green), as well as microfluidic channels and apertures, sample extraction port, alignment notches, and a chamber for housing fast-curing epoxy resin (5-minute epoxy, indicated in gold-yellow). Part 2 is bonded and pressed onto part 1, and the vacuum pipe is inserted through the designated hole in part 1. After reservoir evacuation, the pipe is thermally sealed and covered with 5-minute epoxy. A conductive adhesive (highlighted in olive green) is used to establish an electrical connection between the spring contacts and the microheater’s conductive pads. Part 3 is the capsule’s hemispherical shell. It features three alignment pins, a main central liquid inlet channel, a lateral sample extraction channel, a filter membrane, a chamber for housing the vacuum pipe, which is sealed with epoxy, a lid for closing the extraction channel, and two holes for housing the spring contact terminals. Part 4 contains the DC-DC converter and an adapter for its alignment and integration. (b) Photos of the fabricated parts and the fully assembled actuator module
Actuator module validation for intestinal microbiome collection
DNA extraction and purification were carried out on intestinal liquid sample (150 µL) collected from an uncleaned porcine intestine using a commercial microbiome DNA extraction kit (ZymoBIOMICS DNA Miniprep Kit, Zymo Research Europe GmbH, Freiburg, Germany). The intestinal liquid sample was first retrieved from the actuator module, and then total microbial DNA was extracted, purified, and eluted in 80 µL of nuclease-free water. The resulting DNA served as a template for polymerase chain reaction (PCR) amplification of a 262-base pair (bp) fragment specific to the Escherichia coli genome.
Each PCR reaction had a final volume of 25 µL, containing 1× PCR buffer, 0.2 mM dNTPs (deoxynucleotide triphosphates), 0.4 µM primer mix, 2 mM MgCl₂, 0.06 U/mL HotStart Taq Plus DNA polymerase (QIAGEN, Hilden, Germany), and 2 µL of the extracted DNA. Amplified PCR products were analyzed by gel electrophoresis on a 2% agarose gel (Agarose, low EEO molecular biology grade, type D1, Apollo Scientific, Manchester, England). A 100 bp DNA ladder (DirectLoad PCR 100 bp Low Ladder; Merck Group, Darmstadt, Germany) was used as a molecular size marker. For nucleic acid staining, PeqGreen DNA/RNA dye (6× concentration) was incorporated into the agarose gel. Electrophoresis was performed in 0.5× TBE buffer (Tris–Borate–EDTA) at 180 V, 100 mA, and 17 W for 25 min. Visualization of DNA fragments was conducted under ultraviolet illumination using a UV transilluminator (Blue/White Light Table, Serva, Heidelberg, Germany).
Contamination evaluation of the collected liquid sample
Potential contamination of the collected samples within the actuator module was evaluated. Three actuator modules (P1, P2, and P3) were activated in a diluted stool sample (330 mg/mL), and samples were collected in the sampling chamber. The modules were subsequently immersed for 2 days at room temperature in an E. coli broth culture carrying a plasmid containing a cytomegalovirus (CMV) promoter (pCMV) sequence. After incubation, the module´s surface as well as the sample extraction channels were washed with distilled water followed by 70% ethanol to remove bacteria adhering to the module surface and in the sample extraction channel. The collected stool samples were then processed for plasmid DNA extraction using the NucleoSpin DNA Plasmid Miniprep kit (Macherey-Nagel, Düren, Germany). PCR amplification and agarose gel electrophoresis were subsequently performed as stated in Sect. 2.4. PCR primers specific to pCMV sequence were used for DNA segment amplification. Contamination of the collected samples with E. coli containing the CMV promoter was indicated by the presence of an amplified DNA fragment of 290 bp on the gel. Four controls were included: positive control 1 (PC1; 150 µL diluted stool sample mixed with 350 µL E. coli broth culture(pCVM)), positive control 2 (PC2; 150 µL distilled water mixed with 350 µL E. coli broth culture (pCVM)), a negative control (NC; 150 µL diluted stool sample mixed with 350 µL distilled water), and a PCR non-template control (NTC).
Assessment of the integrated DNA/RNA preservative
The efficacy of a dried preservative integrated into the sampling chamber of the actuator module was evaluated. Briefly, 500 µL of a stool DNA/RNA preservative solution was dried in a 1.5 mL microcentrifuge tube at 40 °C for 3 days. The dried preservative was subsequently redissolved in a mixture consisting of 500 µL of E. coli lysate and 500 µL of diluted stool sample (330 mg/mL). Samples without preservative were prepared by combining identical volumes of E. coli lysate and diluted stool sample. The E. coli lysate was generated by heat lysis at 95 °C for 8 min, followed by centrifugation to remove unlysed cells and cellular debris. The resulting supernatant, containing E. coli gDNA and RNA, was used for all subsequent analyses. A control sample consisted solely of 1 mL of E. coli lysate serving as a reference of genomic DNA (gDNA) and RNA segment size as well as of E. coli lysate quality was also prepared. A second control containing only preservative redissolved in 1 mL water was also used. Samples with and without preservative were prepared in triplicate and incubated at room temperature for 0, 5, and 10 days to assess long-term preservation performance and gDNA and RNA integrity. For gDNA and RNA segment integrity analysis, all samples were first centrifuged at 4000 rpm for 4 min and then a 20 µL of the supernatants were loaded in a 1% agarose gel and left for migration during gel electrophoresis at 120 V, 40 mA, and 5 W for 35 min.
Results
Opening mechanism
Initial characterization focused on evaluating the integration, sealing performance, stability, and functionality of a meltable wax-based opening valve under high pressure in the absence of an integrated microheater. For system closure to be ensured prior to valve activation and vacuum integrity to be maintained within the liquid sampling reservoir, effective adhesion of the wax to the inner wall of the hole is essential.
To evaluate the sealing capability of the manually integrated wax, the inlet channel was pressurized to 900 mbar, while the outlet channel was filled with air. Under these conditions, continuous formation of water droplet at the wax–hole interface was observed in the outlet channel (Fig. 5a), indicating incomplete sealing. To improve seal performance, a reflow step was introduced, during which the valve was heated to 60 °C for one minute to remelt and redistribute the wax. Post-reflow inspection revealed no visible droplet formation, confirming an effective seal (Fig. 5b).
The valve’s opening behavior was then evaluated by heating the valve chip on a hot plate at 70 °C. As shown in Fig. 5c and in the video (online resource 1), wax melting initiated at the wax–hole interface and proceeded until the valve fully opened, allowing liquid passage (Fig. 5d). The opening time was found to depend on whether both the inlet and outlet channels are filled with liquid or only one of them. When both the inlet and outlet channels were filled with air, the valve opened in 13.4 ± 0.3 s at an applied pressure of 500 mbar. In contrast, when both channels were filled with water, the opening time increased to 28.0 ± 1.2 s at the same pressure. This delay is likely due to the thermal absorption of water, which slows heat transfer and thereby delays wax melting. In the final capsule design, air will be trapped in the inlet channel, enabling a faster response time of the valve.
Fig. 5. Integration and performance characterization of the wax-based opening valve without an integrated microheater. *(*a) Initial integration of wax without reflow, showing incomplete sealing at the wax–hole interface as evidenced by continuous water droplet formation. (b) Improved sealing after a reflow step at 60 °C for one minute, with no droplet formation observed, indicating successful sealing of the interface. (c) Onset of wax melting at the wax–hole interface during heating at 70 °C. (d) Complete opening of the valve following wax melting, allowing water to flow from the inlet channel into the outlet channel
The screen-printed microheater, which is based on electrical resistance, was subjected to optical and functional characterization (Fig. 6a-c). Key parameters, including fabrication parameters, electrical resistance, and power requirements for effective heat generation were systematically measured and optimized. The primary functional requirement of the microheater was to reliably exceed the melting temperature of wax (61 °C) while minimizing power consumption. First, an optical characterization was conducted using confocal microscopy. As shown in Fig. 6a and b, the morphology of the printed heater was examined, and the height and width of the resistive line were determined to be 24.8 ± 4.5 μm and 82.4 ± 3.7 μm, respectively.
Electrical characterization was performed at a dedicated measurement station using a power supply and contact needles. The resistance of the heater was found to be 6.9 ± 0.5 Ω, with a contribution of 1.2 Ω from the measurement setup. With an applied voltage of 0.6 V and a current of 90 ± 10 mA, the heater consumed 52 ± 3 mW of power and reached a steady-state temperature of 62.3 ± 2.1 °C within 20 s. To assess performance at higher power levels, the heater was also tested at 0.8 V with a current of 120 ± 10 mA, corresponding to a power input of 96 ± 8 mW. Under these conditions, the heater achieved a temperature of 101.0 ± 3.7 °C within 20 s (Fig. 6c). Operating at 120 mA successfully generated heat that exceeded the wax melting point without surpassing the glass transition temperature of the heater carrier (PMMA, 120 °C), thereby ensuring the wax-based opening mechanism was activated effectively without compromising the structural integrity of the device.
Following the successful individual characterization of the wax component and the microheater, the complete valve was fabricated and evaluated under a flow pressure of 900 mbar. At the time of valve activation, both the inlet and outlet channels were empty. Actuating the normally closed valve (Fig. 6d with a current of 90 ± 10 mA), the valve opened after 9.5 ± 1.0 s (Fig. 6h). In contrast, at a higher activation current of 120 ± 10 mA, the valve opened significantly faster, within 3.6 ± 0.5 s (Fig. 6h). These rapid response times demonstrate the valve’s suitability for immediate and localized liquid sampling from the small intestine. Figure 6d-g as well as the video (online resource 2) show the entire opening mechanism of the valves.
It is crucial that the wax at the hole´s edge melts rapidly and completely prior to the arrival of intestinal liquid at the valve opening site, as contact between the intestinal liquid and melted wax can lead to localized cooling and subsequent solidification of the wax. Insufficient melting may prevent full valve actuation, which could result in significant delays in the sampling process. Furthermore, as the valve is designed to operate when both the inlet and outlet channels are initially empty, it is expected that its behavior remains consistent with these results when subsequently operated inside the capsule.
Fig. 6. Characterization of the microheater and the opening mechanism. (a) Surface morphology profile obtained by confocal microscope measurement of the printed star-shape microheater showing its height and width. (b) A 3D surface profile of the microheater. (c) Thermal image of the microheater activated at 90 mA. The heater generated a heat of 72.5 ± 1.3 °C at the core of heater after 20 s of activation. (d) Whole normally closed opening valve with integrated wax and microheater. (e) Activated microheater at 120 mA showing the melting process of wax at 0 mbar (pressure difference between valve inlet and outlet). Wax at the hole´s edge was completely melted within 2–4 s. (f) Opening valve showing the opening process activated at 120 mA and at flow pressure of 900 mbar. After 2–4 s of microheater activation, the ink solution flowed through the wax hole into the outlet channel. (g) The valve with opened and cleared hole after microheater activation. The resolidified wax was pushed away from the valve hole due to the applied flow pressure. (h) Valve opening time at two different current values (three measurements each) demonstrating the dependency of the opening time on the applied current. A higher current generates higher heating and an earlier opening of the valve. Q – flow rate, t - time
Closing mechanism
The closing valve utilizing a swellable cellulose membrane was fabricated as described in Sect. 2.2. Its closing time and the maximum flow pressure were evaluated experimentally. To verify the functionality of the valve and investigate its performance, the geometrical parameter known as ‘step size’, which is defined as the vertical distance between the membrane surface and the valve inlet hole, was systematically varied by adjusting both the valve seat height and the membrane thickness. In the initial experiment, the valve seat height was fixed at 300 μm with a diameter of 1.8 mm. The surface of the compressed cellulose membrane, with a thickness of 200 ± 10 μm, was covered with either one or three layers of acrylic adhesive film (each 25 μm thick), yielding step sizes of 75 μm and 25 μm, respectively. In both valves, the adhesive film faced the valve inlet hole, as shown in Fig. 4d. Since both valve seat and membrane diameters were the same, liquid-membrane contact area (0.06 mm^2^) was restricted only at the connection point between outlet channel and valve seat as shown in the video (online resource 3). Valve performance was characterized by using water under a constant flow pressure of 900 mbar. As shown in Fig. 7a, the valve with the larger step size (75 μm) exhibited a delayed closure, achieving a flow rate of 6.6 ± 2.8 µL/min after six seconds, and complete closure (0 µL/min) after 12.1 ± 3.5 s, with a total dosed volume of 56.0 ± 6.6 µL. By contrast, the valve with the smaller step size (25 μm) closed more quickly, reaching a flow rate of 4.0 ± 2.8 µL/min within three seconds and complete closure after 5.8 ± 0.4 s, with a total dosed volume of 5.7 ± 0.6 µL. The results indicated that valves with both step sizes achieved near-complete closure within four seconds. However, due to the continuous application of a flow pressure of 900 mbar during the closing phase, a minimal amount of liquid continued to pass through the valve. Complete closure, characterized by a flow rate of 0 µL/min, required additional time to allow full membrane swelling. This enabled the membrane to generate sufficient counterpressure to overcome the applied flow pressure.
The volume dosed in the initial experiment was insufficient, falling significantly below the intended target volume of approximately 150 µL for collection within the capsule. To address this issue, the step size was increased by employing a higher valve seat (425 μm), and the diameter of the cellulose membrane was reduced to 1.7 mm to ensure sufficient contact area between the membrane edge and the liquid as shown in the video (online resource 4). The cellulose membrane with two layers of laminated adhesive, with a total thickness of 250 ± 10 μm, was prepared. Here, the adhesive film also faced the valve inlet hole. These modifications resulted in a step size of 175 ± 10 μm and a contact area of 1.068 mm^2^. Complete valve closure within 10–20 min was considered acceptable for the intended application, since no more liquids will be aspirated into the reservoir once it is fully filled within two seconds, even if the valve is not yet fully closed. It is important to note that intraluminal pressure within the small intestine varies by region but generally does not exceed 250 mbar, as reported by Farmer et al. (2013) (Farmer et al. 2013). Based on this, the valve’s pressure resistance and closing time were assessed at a flow pressure of 150 mbar at time intervals of 10, 20, and 30 min. To accurately replicate the fluid dynamics involved in in-capsule intestinal liquid sampling, water was introduced into the valve at a pressure of 900 mbar for two seconds. This condition simulated the pressure differential between the evacuated sampling reservoir (100 mbar) and the ambient pressure (1000 mbar) surrounding the capsule, as well as the time required to fill the sampling reservoir with a driving pressure of 900 mbar. At each time interval, this pressure was maintained for 3–5 s while the resulting flow rate was recorded. As shown in Fig. 7b, the flow rate measured after 20 min was 1.75 ± 1.75 µL/min, indicating a valve closure efficiency of approximately 99%. The total dosed volume obtained after two seconds at 900 mbar was also measured to be 202.5 ± 19.5 µL.
The closing valves were evaluated using the same geometrical design and experimental conditions with a diluted stool sample (330 mg/mL) to more closely simulate the physicochemical properties of intestinal liquid, which is similar in nature to stool. Two distinct configurations of a cellulose membrane with two laminated adhesive films were integrated into the valve: (i) with the adhesive film facing the valve inlet, thereby restricting contact area (1.068 mm^2^) with the stool sample to the membrane’s periphery (Fig. 3d), and (ii) with no adhesive film on the membrane surface to allow an increased contact area (3.338 mm^2^) ((Fig. 3e). To minimize interference with valve function from large stool particulates, and to simulate the function of the filter membrane at the capsule´s aperture, the stool sample was first centrifuged at 1.2 relative centrifugal force (RCF) for one minute. Subsequently, 150 µL of the supernatant was manually injected into the valve using a syringe. Flow rates were then measured at four time points: 5, 10, 20, and 30 min.
As shown in Fig. 7c, even after 30 min, the valve with the limited contact area did not fully close and exhibited significant variability. This incomplete closure was probably caused by insufficient liquid absorption by the cellulose membrane, which prevented the vertical expansion required to fully seal the valve. The reduced absorption can be attributed to both the high viscosity of the stool sample and the limited membrane–sample contact area. In highly viscous liquids, it is more difficult for the liquid to penetrate the narrow space between the membrane’s edge and the valve wall, as well as the small pores of the cellulose membrane, thus significantly impairing liquid uptake. In contrast, the configuration with an increased contact area achieved complete valve closure within 10 min and demonstrated improved reproducibility. These findings emphasize the importance of maximizing the contact area between the membrane and the sample in ensuring proper valve function, particularly when dealing with viscous biological liquids such as intestinal contents.
Fig. 7. Characterization of Valve Closure Dynamics. (a) Closing profiles and times of two valves with different actuation step sizes (75 μm and 25 μm) and were evaluated using water under a constant flow pressure of 900 mbar. Both adhesive covered-membrane and valve seat have the same diameter (1.8 mm) restricting the contact area (0.06 mm^2^) at the connection point between valve outlet channel and valve seat. (b) The effect of a larger actuation step size (175 μm) on valve closure time was investigated. Water was introduced at 900 mbar for 2 s, followed by closure time assessment at three-time intervals using a reduced pressure of 150 mbar for 3–5 s. The increased step size resulted in a significantly slower closure. In this configuration, the adhesive-covered membrane diameter was reduced to 1.7 mm resulting in an increased contact area (1.068 mm^2^) at the membrane periphery. (c) Valve closing dynamics were further assessed using a diluted stool sample (330 mg/mL) to simulate the rheological properties of stool-like intestinal liquid. Two valve configurations were tested: one with a smaller membrane–liquid contact area (membrane surface laminated with adhesive film, indicated in blue line, limiting the contact to the membrane periphery, 1.068 mm^2^), and one with a larger contact area (membrane without adhesive film, indicated in orange line, 3.338 mm^2^). A volume of 150 µL of the stool sample was manually introduced into each valve using a syringe. Valve closure was then investigated at four-time intervals under 150 mbar for 3–5 s. The configuration with the larger contact area achieved complete closure within 10 min and demonstrated consistent performance, while the valve with the smaller contact area failed to close even after 30 min. Ca – contact area, Q – flow rate, t - time
In vitro validation of the whole actuator module
The actuator module was fabricated as outlined in Sect. 2.3 and was then evaluated in vitro using different liquid samples. Initial testing involved connecting the microheater to a power supply (0.8 V, 120 ± 10 mA) via either a cable or a contact needle (Fig. 8). For testing with a diluted stool sample (330 mg/mL) or ink-colored water, a liquid sample container was mounted on the capsule hemisphere to ensure direct contact between the capsule’s aperture and the liquid. As shown in Fig. 8a–c, three actuator modules were tested using the diluted stool sample. All modules operated as intended: the opening valve actuated within two seconds, and the sampling reservoir was filled within one to two seconds as shown in the video (online resource 5). Following sample collection, the contents were extracted through the designated extraction channel using a syringe (Fig. 8d). The collected sample volume was measured as 144.1 ± 5.1 µL.
The actuator module can also be used for either in vivo or in vitro validation in the intestine under operational physiological conditions. In such cases, the microheater must be connected to an external power supply via a cable (Fig. 8e). For more controlled and simplified in vitro assessments of the actuator module with liquid samples, the actuator module can be evaluated alternatively using a test board equipped with a BM23FR0.6-16DS-0.35 V connector as shown in Fig. 8f and in the video (online resource 6). In this configuration, the DC-DC converter and the contact spring are integrated to enable electrical connection to the test board. As shown in Fig. 8f, the actuator module functioned as intended and the liquid was collected. A small bubble was formed in the sampling reservoir due to the empty central inlet channel in the hemisphere (part 3) of the actuator module.
To further evaluate the actuator module with a liquid sample exhibiting different physicochemical characteristics, the module was tested for liquid collection in a freshly excised, uncleaned porcine intestine (Fig. 8g). The intestine was intentionally left uncleaned to preserve the natural properties of the intestinal liquids. Considerable variability in liquid viscosity was observed across the intestinal tract; therefore, the module was examined in two different segments characterized by high and low viscosities. In both cases, the actuator module reproducibly functioned as intended with the diluted stool sample and ink-colored water, successfully collecting porcine intestinal liquid samples (Fig. 8g). Microbial DNA was subsequently extracted from the collected samples and subjected to PCR amplification targeting a specific genomic region of E. coli. Gel electrophoresis revealed the expected 262 bp amplicon (PCR product), thereby confirming successful intestinal microbiome sampling. In brief, the system has efficiently performed for a sample collection with liquids of variable physicochemical characteristics. The pH values of all liquid samples tested using the actuator module were measured to be 7.5 for ink-colored water, 7 for the porcine intestinal liquid, and 6.2 for the diluted stool sample. To further confirm complete closure of the actuator module’s closing valve following 10 min of sampling, as previously conducted in Fig. 7, the capsule inlet channel was connected to a fluidic pressure source (150 mbar) while the sample extraction channel was opened. The closing valve also exhibited full valve closure with all tested samples, with a measured flow rate of 0 µL/min. In addition, potential evaporation of the collected sample at elevated temperatures was assessed by incubating the actuator module at physiological temperature (37 °C) for three days. No significant sample evaporation was observed within the sampling chamber, confirming the effective sealing performance of the closing valve. Throughout the validation and verification processes, the actuator module demonstrated reliable performance across repeated tests, including experiments conducted with ink-colored water (n = 8), diluted stool samples at a concentration of 330 mg/mL (n = 16), and porcine intestinal liquid (n = 3).
Fig. 8In vitro validation of the actuator module. (a) Triplicate actuator modules tested with a diluted stool sample (330 mg/mL). The microheater was operated using contact needles at 0.8 V and 120 ± 10 mA. (b) Initiation of the sampling process, demonstrating aspiration of the stool sample into the sampling reservoir. (c) The actuator modules triplicate before (empty) and after activation of the sampling process (filled). (d) Retrieval of the collected stool sample via the extraction channel using a syringe. (e) Electrical connection of the actuator module’s microheater to a power supply via a cable, enabling potential in vivo validation. (f) Actuator module under test using a dedicated test board. (g) Validation of the actuator module conducted in porcine intestines at two distinct segments with different liquid viscosities. DNA was extracted from the collected intestinal liquid samples, followed by PCR amplification targeting a specific DNA segment in intestinal E. coli genome. A PCR product of 262 bp, corresponding to the expected E. coli DNA fragment, was detected, thereby confirming successful sampling of intestinal bacteria. Triplicate actuator modules have been validated (piece 1 – piece 3). P - Piece, NTC – none-template control, marker – DNA fragment size reference
Potential contamination of the collected liquid in the actuator module was also evaluated. Here, three pieces of the actuator module were activated in a diluted stool sample (330 mg/mL), and the samples were collected in the sampling chamber. To investigate if the module is completely closed, the modules were immersed for 2 days at room temperature in an E. coli broth culture carrying a sequence of cytomegalovirus promotor (pCMV). Possible contamination of the collected stool sample with E. coli carrying CMV promotor can be detected if a PCR product of a DNA segment size of 290 bp is detected after performing DNA extraction and PCR and gel electrophoresis of the collected sample. As seen in Fig. 9, highly faded DNA segment (290 bp) was detected in one of the actuator modules (P3) indicating very minimal contamination of the collected stool sample.
Fig. 9. Evaluation of potential contamination of samples collected in the actuator modules. Three actuator modules (P1, P2, and P3) containing collected diluted stool samples (330 mg/mL) were immersed for 2 days at room temperature in an E. coli broth culture carrying a plasmid with a cytomegalovirus (CMV) promoter sequence (pCMV). The collected samples were subsequently processed for DNA extraction, followed by PCR amplification and agarose gel electrophoresis to assess contamination. Faint DNA band at 290 bp was observed with only module 3 (P3), indicating only minimal potential contamination of the collected sample with E. coli harboring the CMV promoter sequence. Four controls were used: positive control 1 (PC1; diluted stool sample mixed with E. coli broth culture (pCMV)), positive control 2 (PC2; distilled water mixed with E. coli broth culture (pCMV)), a negative control (NC; diluted stool sample mixed with distilled water), and a PCR non-template control (NTC)
To improve the sensitivity and precision of the downstream diagnostic assays for gut microbiome profiling, it is crucial to preserve the intestinal liquid samples in the sampling reservoir for at least five days at room temperature. To achieve this, a commercial stool sample preservative solution (OMR-205, DNAgenotek Inc) was integrated into the sampling reservoir by drying the solution directly within it. This preservative solution can maintain the integrity of microbial DNA and RNA for up to10 days. Initial evaluation of the integration and subsequent resuspension of the preservative in a diluted stool sample (330 mg/mL) was conducted in both an open chamber and the sampling chamber of the actuator module, as illustrated in Fig. 10. First, a volume of 80 µL of the preservative solution was introduced into the chambers and dried at 60 °C for six hours. Then, 150 µL of the diluted stool sample was added to assess the resuspension behavior. Successful reconstitution of the dried preservative and its homogenization within the stool solution are critical for maintaining sample integrity. Although the dried preservative dissolved completely within 40 min, uniform distribution and diffusion were hindered by the high viscosity of the stool sample. Consequently, mechanical agitation (e.g. stirring) was necessary to ensure homogeneous mixing. As demonstrated in Fig. 10, the integration and resuspension of the preservative material in the sampling chamber of the actuator module was successfully achieved. Here, the actuator module was gently rotated and tilted several times after the stool sample was collected, to simulate capsule movement in the small intestine. As can be seen, an acceptable amount of dried preservative was dissolved in the diluted sample within 40 min.
Fig. 10. Integration of the DNA/RNA preservative material within the sampling reservoir. A volume of 80 µL of preservative solution was deposited and dried onto the inner wall of the sampling chamber of the actuator module or an open chamber. To evaluate the dissolution behavior, 150 µL of a diluted stool sample (330 mg/mL) was added. The dried preservative started dissolving upon contact with the stool sample; however, it did not distribute uniformly within the stool sample. Subsequent mixing steps were therefore required to achieve homogenous reconstitution of the preservative within the sample
The efficacy of the preservative following dissolution in a diluted stool sample (330 mg/mL) was evaluated. An E. coli lysate containing genomic DNA (gDNA) and RNA was mixed with the diluted stool sample, and the integrity of gDNA and RNA was assessed in the presence and absence of the preservative after incubation periods of 0, 5, and 10 days at room temperature. Long-term nucleic acid integrity was analyzed by gel electrophoresis. As a control, E. coli lysate alone was first loaded onto the gel to verify lysate quality and to serve as a reference for intact gDNA and RNA. Also, preservative alone has been loaded onto the gel to visualize potential background signals. As shown in Fig. 11, distinct gDNA and RNA bands were observed in E. coli lysate, confirming successful bacterial lysis. Subsequently, test samples were analyzed to evaluate preservative performance and long-term nucleic acid integrity. Samples lacking the preservative exhibited clear E. coli RNA degradation after 5 days and significant E. coli gDNA degradation after 10 days. In contrast, samples containing the preservative maintained good integrity of E. coli gDNA throughout the 10-day incubation period, while E. coli RNA was not able to be clearly visualized due to the strong bright background signal of the preservative material. The smeared regions observed between E. coli gDNA and RNA in the absence of preservative may indicate the presence of degraded nucleic acids derived from the stool samples. These smeared regions were not able to be visualized in the presence of preservative due to the strong dark background signal in these regions. These results demonstrate that the preservative functions properly after drying and subsequent re-dissolution, enabling effective preservation of nucleic acids in collected intestinal liquids when the capsule will be implemented in vivo.
Fig. 11. Assessment of the efficacy of the dried DNA/RNA preservative following re-dissolution for long-term nucleic acid preservation. E. coli lysate was incubated with diluted stool samples (330 mg/mL) in the presence or absence of the re-dissolved preservative for 0, 5, and 10 days at room temperature. Long-term DNA and RNA integrity was evaluated by gel electrophoresis. In samples lacking preservative, degradation of E. coli RNA was observed after 5 days, while degradation of E. coli gDNA was significant after 10 days. In contrast, E. coli gDNA maintained good integrity after 10 days of incubation in the presence of the preservative. However, E. coli RNA and potentially degraded stool-derived nucleic acids could not be visualized due to strong background signals originating from the preservative
Discussion
Achieving precise-site specific sampling requires fast, reliable, and reproducible performance of all actuation mechanisms (opening, liquid aspiration, and closing) at the activation site under various physicochemical properties (pH, viscosity, chemical composition, and particle size) of the intestinal liquids. This ensures that the liquid sample was collected only at the activation site enhancing the spatial resolution of intestinal microbiome analysis. An actuator module based on a microfluidic microsystem incorporating opening, closing and liquid aspiration has been fabricated and validated for this purpose.
In this work, we present a controlled meltable opening valve that demonstrates a rapid response time. Activation of the microheater initiates valve opening with minimal power consumption (96 ± 8 mW; 120 ± 10 mA). As illustrated in Fig. 6e, complete melting of the wax at the hole´s edge occurs within 2–4 s of heater activation. The temperature reached 101 ± 4 °C within 20 s of microheater activation, but the valve opened in just 3.6 ± 0.5 s. This suggests that a temperature above the wax melting point (61 °C) was reached within less than 5 s. Therefore, the in-capsule wireless microheater will only be activated for 5 s later. During subsequent in vivo capsule validation, the opening valve is expected to actuate slightly earlier when operated at physiological temperature (37 °C), as the target actuation temperature (> 61 °C) is reached more rapidly when heating is initiated from 37 °C rather than from room temperature. Nevertheless, valve opening remains effectively instantaneous and it will not have an impact on the performance of the application. Subsequently, the negative pressure in the sampling reservoir facilitates the movement of both melted and partially melted wax into the outlet channel, thereby clearing the hole and enabling intestinal liquid to enter the reservoir. Upon contact with the incoming liquid, the molten wax rapidly solidifies, and the wax mass is completely displaced from the valve hole (Fig. 6g). This leaves the wax hole completely open, ensuring a low hydrodynamic resistance and a fast flow rate during intestinal liquid aspiration. This process ensures rapid and effective valve operation, supporting eventually the acquisition of enough intestinal liquid in the targeted region. The high reproducibility and rapid opening of the valve were also attributed to the integration of wax beyond the filter membrane rather than at the capsule aperture, in contrast to opening valves based on pH-sensitive dissolvable materials. Unlike dissolvable systems, wax activation does not require direct contact with liquid. Consequently, the influence of intestinal liquids on the valve opening time is minimized, as liquid contact during activation was previously shown to delay opening due to heat dissipation into the surrounding medium.
To activate the sampling process at a specific target segment, the control electronic module should incorporate localization capabilities. These may include pH, temperature, gas, and pressure sensors, which collectively characterize gastrointestinal segments and enable determination of the capsule’s position within the gastrointestinal tract. In addition, a miniaturized camera may be incorporated to visualize the gastrointestinal wall and facilitate targeted sampling at specific locations. Such capabilities would be particularly advantageous for collecting biomarkers and bacteria from pathological sites, such as tumors or regions of inflammation, for diagnostic applications.
As demonstrated in Fig. 7c, the miniaturized swellable cellulose membrane-based closing valve operated as intended when tested with a diluted stool sample (330 mg/mL). The valve achieved complete closure within 10 ± 0 min under an applied pressure of 150 mbar. Based on these results, we anticipate that the valve will perform similarly in the small intestine when exposed to intestinal liquid, for two primary reasons. First, cellulose membranes exhibit great chemical stability minimizing sensitivity to variations in gut pH and chyme chemical composition, resulting in more reproducible swelling behavior and, consequently, consistent valve closing times. This characteristic is critical for achieving reliable and reproducible site-specific liquid sampling and valve closure, especially across different patients with variable intestinal conditions. Second, the design of the valve and the way the miniaturized swellable cellulose membrane is integrated into the small fluidic channel ensure reliable and instant contact between the membrane and the intestinal liquid. The valve is highly miniaturized, requiring only a small volume of liquid to initiate swelling. Specifically, the membrane has a volume of 0.45 mm³, with a diameter of 1.7 mm and a height of 200 μm. This compact configuration enables sufficient liquid uptake, even when small air bubbles are aspirated and flow through the fluidic channel during sampling. This ensures that enough liquid still reaches the membrane, initiating the swelling process. Figure 7c demonstrates the importance of sufficient membrane contact with an adequate amount of intestinal liquid. As the results showed, covering the cellulose membrane with laminated adhesive film limited the contact area to the periphery, which affected both the swelling time and valve reproducibility. It appeared that a very small peripheral contact area did not always provide sufficient contact with the viscous stool sample. Therefore, it was essential to remove the adhesive film from the membrane surface to ensure a reliable and reproducible closure of the closing valve.
In the previously developed passive capsules, environment-sensitive hydrogels were primarily integrated into the base of a large sampling reservoir, far from the capsule’s inlet aperture. This configuration may restrict the hydrogel’s exposure to liquid, particularly in systems that depend on capillary action or imbibition rather than vacuum-driven liquid ingress. Furthermore, the high viscosity of intestinal liquids can impede sufficient liquid absorption or transport into the sampling reservoir. This can lead to slow or incomplete hydrogel swelling, resulting in improper closure of the sampling inlet. The miniaturized swellable valve based on a cellulose membrane presented here overcomes these limitations. Our design ensures that the cellulose membrane consistently absorbs enough liquid to rapidly initiate and complete swelling. This guarantees precise, site-specific sampling of intestinal liquids and prevents contamination from downstream regions of the intestine.
To optimize the performance of the capsule’s actuation mechanisms and to prevent occlusion of the fluidic channels by intestinal food particles, a filter membrane (150 μm size exclusion, Ø 3.3 mm) was integrated into the design of the ingestible capsule. With a diameter of 3.3 mm, the filter membrane covers the capsule’s inlet channel, which measures 1 mm in diameter. Both the membrane and the inlet channel are centrally positioned at the apex of the capsule’s hemisphere. This central placement minimizes the likelihood of direct contact with the intestinal wall, thereby reducing the risk of membrane obstruction and ensuring unobstructed liquid aspiration. The large membrane surface area increases the likelihood of maintaining open flow paths for liquid entry, even when substantial food particle accumulation occurs. Food particles larger than 150 μm are unable to enter the system, thereby avoiding their possible accumulation at the opened wax hole (valve hole diameter of 600 μm) or on the cellulose membrane (valve step size of 175 μm). Additionally, membranes with slit-shaped openings demonstrated superior performance compared to those with circular pores when tested with a diluted stool sample of 330 mg/mL. Specifically, membranes with circular pores (150 μm pore size, Ø 3.3 mm) became completely occluded during testing with the same concentration of the diluted stool, resulting in an inability to collect the required sample volume. In contrast, actuator modules equipped with slit-shaped membrane openings consistently maintained functionality under identical test conditions, without experiencing complete blockage.
The diluted stool sample at a concentration of 330 mg/mL consisted of untreated human stool and exhibited high viscosity, along with a substantial presence of large, undigested food particles of variable sizes, including fibrous material. Frequent blockage of a 1000 µL pipette tip (orifice diameter: 1 mm) during handling of this sample further confirmed the presence of large particulate matter and served as a practical reference for particle size and density. The filter membrane demonstrated robust and consistent performance during in vitro testing with the diluted stool sample as well as with porcine intestinal liquids with varying viscosities and particle size. During actuator module testing, a consistent sampling delay of 1–2 s was observed across all tested samples for the collection of a defined volume. This consistent delay indicates that the slit-shaped filter membrane operated efficiently and reproducibly without complete occlusion by undigested food particles. The reproducibility of the short sampling delay time is also due to the high negative pressure (100 mbar) that was maintained in the sampling reservoir. These results indicate that the membrane can function effectively in liquids with diverse physicochemical properties and particle size distributions.
The selected concentration of diluted stool (330 mg/mL) is considered a proper simulant for chyme concentration in the small intestine, the intended site of sample collection. Rough calculations were performed to estimate the chyme concentration in the small intestine at the time of capsule activation, based on values reported by Sensoy et al. (Sensoy 2021). The cumulative daily volume of ingested water and gastrointestinal secretions entering the small intestine is approximately 8,500 mL, while the total mass of food is about 800 g. Assuming that 80% of the liquid phase is absorbed in the small intestine, the remaining fluid volume is estimated to be 1,700 mL. Under these conditions, 800 g of food distributed in 1,700 mL corresponds to a chyme concentration of approximately 470 mg/mL. As it is planned for capsule implementation that patients drink a large volume of water (e.g., 1–1.5 L) prior to capsule activation, the chyme concentration at the time of activation is expected to decrease to approximately 250–300 mg/mL. However, accurately estimating and reproducing the physiological properties of chyme remains challenging, as they depend on the quantity and composition of ingested food and liquids.
Evaluation of potential contamination in the actuator module after stool sample collection yielded promising results. Among the three actuator modules tested, only one module (P3) exhibited minimal contamination, as indicated by the presence of a faint PCR amplicon corresponding to the CMV promoter sequence (Fig. 9). These findings demonstrate that the system is effectively sealed and that the collected sample is unlikely to be contaminated by downstream intestinal contents following activation of the actuator module at the targeted segment when it is implemented in vivo. The minimal contamination observed in module 3 (P3) can be considered negligible, as it is unlikely to significantly alter the overall microbiome profile of the collected sample. The observed contamination may have originated from the capsule inlet channel, which was potentially exposed to E. coli (pCMV) broth culture. During sample retrieval from the sampling chamber using a syringe, the application of high aspiration pressure on the closing valve may have facilitated the unintended passage of E. coli broth from the inlet channel into the sampling chamber, resulting in minor sample contamination.
The integration of preservative material into the reservoir chamber, and then dissolving it in the diluted stool sample, has shown promising results. As can be seen in Fig. 10, a mixing step was necessary to homogenize the redissolved preservative in the stool sample. It is hypothesized that similar agitation and mixing will naturally occur within the gastrointestinal environment due to peristalsis and the anatomical complexity of the small intestine (i.e. its twisting nature). Additionally, following liquid sampling (see Fig. 8f), a gas bubble was observed to form within the reservoir, resulting from the empty inlet channel in part 3. Both the intestinal motility and the presence of this bubble are expected to facilitate the mixing and homogenization of the dissolved preservative within the collected sample. It is suggested that the commercial preservative will preserve intestinal liquid samples as well as the stool samples.
Safety considerations for in vivo implementation of the sampling capsule were carefully addressed. The capsule form, with a standardized capsule size (30 mm x 11.4 mm) was designed without sharp edges on its external surface to minimize the risk of tissue injury or intestinal wall perforation. To ensure biocompatibility, the capsule surface will be coated with parylene C, a well-established biocompatible polymer. Also, this coating provides surface smoothing, which reduces friction against the intestinal wall and facilitates capsule transit driven by intestinal peristalsis. To prevent occlusion of the filter membrane during the polymer deposition process, the membrane will be protected by a temporary blocking layer. The filter membrane should then be printed using a biocompatible resist. Potential risks associated with the heating process generated by the miniaturized microheater (Ø 2 mm) were also considered. Owing to the short duration of localized heating (< 5 s, maximum energy of 480 mJ) and the effective thermal insulation between the heater and the intestinal wall, no adverse thermal effects on intestinal tissue are expected during operation. To ensure safe swallowability and unobstructed transit through the small intestine, the capsule dimensions were specified as 30 mm in length and 11.4 mm in diameter, which can be considered as the upper dimensional limit of ingestible capsules.
The proposed system may encounter several challenges and limitations during in vivo intestinal liquid sampling. One potential issue is the accidental presence of a gas bubble at the capsule aperture, which could result in gas aspiration and subsequent failure to collect intestinal liquid. To mitigate this risk, the volume of the sampling chamber could be increased to allow liquid aspiration following initial gas intake. Alternatively, redundancy may be introduced by integrating two actuator modules within the capsule, positioned on opposite hemispheres, to compensate for potential sampling failure of a single module. Another potential limitation is the presence of large food particles or sheet-like aggregates that could completely obstruct the filter membrane (Ø 3.3 mm). To reduce the risk of full membrane blockage by large particles or agglomerates in the small intestine, future design change may incorporate an enlarged, curved filter membrane covering up to half of the capsule hemisphere. Such a configuration would increase the probability of maintaining an unobstructed liquid pathway for effective aspiration. Additionally, to further minimize the risk of system obstruction and ensure reliable actuation, minimal gastrointestinal tract preparation prior to sampling may be implemented. This could include the consumption of easily digestible foods and adequate liquid intake to reduce the presence of concentrated food residues at the capsule aperture.
Conclusion
The idea was to create a modular capsule for site-specific sampling of intestinal liquid for gut microbiome analysis. The capsule comprises two fundamental functional units: (i) an actuator module with opening and closing mechanisms, an evacuated sampling chamber preloaded with dried preservative, and a filter membrane to exclude large food particles, and (ii) an electronic control module with communication, segment localization, and power supply components. In this study, the actuator module was designed, fabricated, and validated in vitro using a diluted stool sample (330 mg/mL) and intestinal liquids with two different viscosities in uncleaned porcine intestine. Design-, actuation-, and material-related challenges were addressed to ensure robust and reliable sampling under variable physicochemical properties (pH, chemical composition, viscosity, and particle size) of the intestinal liquids.
The opening mechanism consists of a wax-based valve (melting temperature: 61 °C) actuated by a screen-printed microheater (resistance: 6.9 ± 0.5 Ω). The heat generated after 20 s of activation was measured to be 101.0 ± 3.7 °C. Complete valve opening was achieved within 3.6 ± 0.5 s at 120 ± 10 mA of microheater activation, meaning that the temperature reached a value above wax melting temperature in less than 5 s. The valve maintained integrity against a 900 mbar pressure (differential pressure between the sampling chamber and intestinal lumen). Closure of the sampling chamber was achieved via a swellable cellulose membrane integrated into a 3D microfluidic channel, sealing the inlet within 10 min of wetting and thereby preventing downstream contamination of the collected sample.
The developed actuator module demonstrated reliable and reproducible performance during *in vitro *validation across all tested liquids, including a diluted stool sample (330 mg/mL) mimicking the physicochemical properties of human intestinal chyme, two porcine intestinal liquids with different viscosities, and ink-colored water. Upon activation, samples were filtered through the integrated filter membrane (150 μm size exclusion) and aspirated into the sampling chamber within a delay of 1–2 s. The collected sample volume was 144 ± 5 µL, and no clogging of the filter membrane was observed with either stool or porcine intestinal liquids. The integrated filter membrane ensured reliable operation of the opening and closing mechanisms by preventing large particles from interfering with valve function. Furthermore, the high negative pressure generated within the evacuated chamber (100 mbar) enabled efficient aspiration of highly viscous liquids. The dried preservative fully dissolved within 40 min under gentle rotation and tilting, simulating capsule movement during intestinal transit, demonstrated a good preservation of nucleic acid after 10 days of incubation with diluted stool samples and E. coli lysate. In vitro validation of the actuator module in an uncleaned porcine intestine, followed by PCR and fragment analysis, confirmed reliable collection of intestinal E. coli. The assessment of potential contamination of the liquid sample collected within the actuator module yielded favorable results, confirming effective sealing of the sampling chamber by the closing valve.
The tested liquid samples exhibited a wide range of physicochemical properties, including pH, viscosity, chemical composition, and particle size. This variability did not affect the performance or reproducibility of the actuator module, demonstrating that all actuation mechanisms: opening, closing, and liquid aspiration, as well as the integrated filter membrane, operated reliably and consistently. Across all conditions, the system showed rapid and well-defined activation kinetics, with capsule opening occurring in less than within 5 s, liquid aspiration within 1–2 s, and capsule closing completed within 10 min. These activation times ensure precise and reproducible site-specific sampling upon microheater activation, without allowing sufficient time for capsule transit into distal intestinal segments when the capsule will be implemented in vivo. Collectively, these results indicate that the proposed actuator module can enable robust and precise intestinal liquid sampling for microbiome diagnostics across individuals with diverse intestinal liquid conditions, which may vary due to factors such as diet, metabolic state, and intestinal health.
Outlook
Several design refinements can be made to finalize the actuator module and complete the capsule system. For example, increasing the diameter of the filter membrane can reduce the risk of the capsule aperture becoming completely blocked and improve performance in highly viscous intestinal liquids containing large, concentrated food particles. Although we believe that the concentration of the diluted stool sample used in this work (330 mg/mL) can be a proper simulant for the physiological chyme concentration at the capsule activation moment, the efficacy of the filter membrane can be further evaluated using diluted stool samples with concentrations exceeding 330 mg/mL.
The electronic components of the control module, including the communication and power supply units, were tested and verified using a dedicated test board. A localization unit will also be defined, tested, and integrated in the whole controlling module. These components will subsequently be integrated into a compact, foldable printed circuit board (PCB), which is designed to fit within one half of the capsule shell (L 15 mm x Ø 11.4 mm). The electronic module will then be electrically and mechanically connected to the actuator module to enable complete in vivo capsule validation. A mechanical joining method will be developed, either temporary or permanent; a snap-fit mechanism is being considered as a potential solution for coupling the electronic and actuator modules. Furthermore, the DC–DC converter circuitry (part 4 of the actuator module) is to be integrated into the control module PCB, thereby reducing the total number of components within the actuator module.
Initial in vivo validation of the complete capsule will be carried out in an animal model to validate the performance and safety. After the capsule is excreted, intestinal liquid samples will be collected and subjected to microbiome analysis. The target bacterial species that colonize the small intestine will be defined as biomarkers and indicators for the sampling of intestinal liquid at specific sites.
In order to reduce costs and enable frequent microbiome analysis and gut health monitoring, particularly in animal models, the electronic control module, which represents the most expensive component of the system, can be reused multiple times instead of being designated for single use. Following capsule excretion, the control module can be decoupled and sterilized before being coupled with a new actuator module for the next analysis cycle. Alternatively, to eliminate the need for sterilization, the control module or the entire capsule can be encased in a thin, removable, biocompatible protective layer prior to use. After excretion, this layer can be manually replaced with a new one before redeployment. Furthermore, the control module can be adapted to integrate with other functional modules for diagnostic applications beyond intestinal liquid sampling, such as tissue biopsy, and for the use of pressure and temperature sensors, gas sensors, and other biosensing modalities. This modularity enables comprehensive physiological assessment and enables large-scale data collection, thereby supporting more precise and effective gastrointestinal diagnostics.
Depending on the sensitivity and type of assay employed for intestinal microbiome profiling, nucleic acids alone or in combination with intact microbial cells may be required as biomarkers for microbiome analysis. When intact intestinal microbiota are included in the analysis, preservation of the microbiome and stabilization of bacterial abundance in the collected samples are essential to accurately maintain the native intestinal microbiome structure.
Supplementary Information
Below is the link to the electronic supplementary material.
Supplementary Material 1 (MP4 1.30 MB)
Supplementary Material 2 (MP4 298 KB)
Supplementary Material 3 (MP4 1.08 MB)
Supplementary Material 4 (MP4 458 KB)
Supplementary Material 5 (MP4 593 KB)
Supplementary Material 6 (MP4 378 KB)
Supplementary Material 7 (DOCX 14.1 KB)
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1P.B. Eckburg, E.M. Bik, C.N. Bernstein, E. Purdom, L. Dethlefsen, M. Sargent, S.R. Gill, K.E. Nelson, D.A. Relman, Diversity of the human intestinal microbial flora. Science. 308(5728), 1635–8 (2005). 10.1126/science.1110591
- 2L.K. Ursell, J.L. Metcalf, L.W. Parfrey, R. Knight, (2012). Defining the human microbiome. Nutrition Reviews, 70(SUPPL. 1). 10.1111/j.1753-4887.2012.00493.x
- 3G. Wang, S. Menon, L. Wilsack, R. Rehak, L. Lou, C. Turbide, J. Auger, A. Tremblay, O. Mathieu, S. Binda, T.A. Tompkins, S. Bruehlmann, C.N. Andrews, (2024). Spatially and temporally precise microbiome profiling in the small intestine using the SIMBA capsule with X-ray tracking. Frontiers in Microbiomes, 3. 10.3389/frmbi.2024.1321624