Micro-computed tomography
We used a SkyScan 1173 micro-computed tomography unit (Bruker) with SkyScan 1173 application software to scan individual heads of various remora species. The scanning parameters included a voltage range of 51–130 kV, an amperage range of 40–136 µA, an exposure time range of 337–730 ms and an image rotation range of 0.06–0.07°. Slice reconstruction of the osteological structure of the remora suction disc was performed using NRecon (Micro Photonics) and rendered in Mimics 15.0 (Materialise). The lamella angle of the remoras, viewed dorsally, was characterized using Fusion 360 (Autodesk).
Fabrication of MUSAS
MUSAS are flexible to be fabricated using a variety of silicone rubbers and biodegradable lamella materials. A detailed fabrication procedure is provided in Supplementary Fig. 3. Unless otherwise specified, all testing of MUSAS was conducted using MUSAS fabricated with Ecoflex 0030 elastomer and shape memory nitinol lamellae.
Universal mechanical testing
Universal tensile testing was conducted to measure the stiffness and adhesion performance of the specimens and materials of interest, including tissue and material samples, euthanized remora, devices and adhesives. A 5944 universal testing system (Instron) with Bluehill V3.11 was used for the tensile test. For both stiffness and adhesion study, the tensile extension rate was set to 30.0 mm min−1, with a measurement interval of 20–100 ms for time, 1 × 10−5 N measurement accuracy for load, and 1 × 10−5 mm measurement accuracy for displacement. All test specimens, including devices and adhesive materials used in the adhesion test, were bonded to the top of a double-stacked 0.5-mm-thick polyimide Kapton strip (McMaster-Carr) using ultraviolet-cured 5055 silicone adhesive (Loctite). The Kapton strips were secured with an Instron tensile grip (Supplementary Figs. 2 and 4). A preload of 0.5 N was applied to all devices and adhesive materials before test. To ensure a fair comparison, the pre-adhesion pressurization for other adhesive materials was for 3 min, whereas for MUSAS there was no prolonged pressurization. Soft substrates used in the mechanical adhesion testing were freshly prepared inert materials or collected tissue (<1 h post-euthanasia) without surface washout, tissue trimming or liquid removal. Soft substrates were only partially secured, with the four corners of the tissue squares glued to the holder to allow for natural sliding and dynamic morphing. All devices and adhesive materials used in the adhesion test had the same adhesion surface area of roughly 250 mm2, which equals the adhesion surface area of MUSAS. The adhesion pressure was calculated by dividing the force by the adhesion surface area. Extended discussion of the experiment set-up and measurement details can be found in Supplementary Text 2, Extended Data Fig. 2, and Supplementary Figs. 2 and 4–11.
Numerical simulation
Solid–fluid interactions between water, tissue and MUSAS
Finite element blockysis was conducted to characterize the hydrostatic differentiation of the adhesive disc of remoras. Commercial software Abaqus 2021 (SIMULIA) was used for the study. The remora disc phantom devices were blockumed to be composed of Ecoflex 0030, with a density of 1.07 g cm−3, a Young’s modulus of 125 kPa and a Poisson’s ratio of 0.49 (ref. 48). The physical parameters of stomach tissue include a density of 1.088 g cm−3, a Young’s modulus of 700 kPa and a Poisson’s ratio of 0.49 (ref. 10). We used coupled Eulerian–Lagrangian techniques to model the solid–fluid interactions between tissue, device and water. Water was treated as a Newtonian laminar flow, with a density of 0.997 g cm−3 and a dynamic viscosity of 8.90 × 10−4 Pa s. The simulation was configured so that the mimicry remora suction cups descended at a constant speed of 0.3 mm s−1 until they touched the stomach tissue submerged in water. Solid–solid interactions were modelled as hard contact for normal behaviour, with tangential behaviour modelled using a penalty method with a friction coefficient of 0.02. Details of calculation of relative vacuum ratio Vr are specified in Supplementary Information.
Self-actuation of nitinol lamella of MUSAS
The self-actuation of nitinol lamellae in MUSAS in response to temperature stimuli was simulated using the commercial software COMSOL Multiphysics 6.2 (COMSOL). The Lagoudas phenomenological inelastic constitutive model for SMAs (nitinol), including the relevant material properties, was implemented to simulate the phase transformation of the SMAs49.
Synthesis of materials for in vitro adhesion characterization
Polyacrylamide–alginate tough hydrogel
The tough hydrogel was composed of alginate and polyacrylamide (pAAm) double networks (pAAm–alginate) crosslinked by numerous dimethacrylate monomers. The synthesis was based on a previously reported protocol14. Hydrogel fabrication was achieved via one-step aqueous free-radical polymerization. In brief, in a 50-ml tube, 30 ml phosphate buffer (100 mM, pH 7), 3.6 g acrylamide, 600 mg sodium alginate (medium viscosity), 1.3 mg N,N′-methylenebisacrylamide and 10 mg ammonium persulfate were added and vortexed to form solution A (pAAm–alginate).
Calcium sulfate was added to deionized water and stirred to form a homogeneous suspension. Then, 5 ml pre-gel solution A was loaded into a 5-ml syringe (diameter 12 mm), and 120 mg calcium sulfate and 29.4 mg N,N,N′,N′-tetramethylethylenediamine were loaded into another 5-ml syringe. The two syringes were connected with a syringe connector and mixed over ten times. Afterwards, the gel solution was poured into a glblock mould covered with a 3-mm-thick glblock plate. After 12 h, the hydrogel was ready to be removed from the mould. In particular, for tests leveraging tough hydrogel as soft substrates to evaluate adhesion performance of MUSAS (Fig. 3c–g), air bubbles were manually introduced during the preparation procedure to create a porous and rough substrate surface.
NHS–EDC bridging polymers for tough hydrogel
The bridging polymers, which include chitosan, polyallylamine, gelatin and polyethyleneimine, were prepared following a previously reported protocol14. Sulfated NHS and EDC were used as coupling reagents. Right before the adhesion of the tough hydrogel to tissue surfaces, the bridging polymers and coupling reagents were mixed to achieve a concentration of 12 mg ml−1 of both NHS and EDC in the bridging polymer solutions. A 250 μl mixed solution was then smeared onto the surface of the tough hydrogel, followed by immediate compression of the tough hydrogel to the tissue surfaces for 3 min before the adhesion test.
Carbopol tough hydrogel
To prepare Carbopol tough hydrogel, 100 mg Carbopol 971P was dissolved in the pAAm–alginate solution described in the pAAm–alginate tough hydrogel protocol. The rest of the preparation procedures were identical to those used for the pAAm–alginate tough hydrogel. A 3-min compression of the Carbopol tough hydrogel to the tissue surfaces was applied before the adhesion test.
SEBS thermoplastic elastomer
The SEBS substrate was prepared by mixing 10 ml toluene with 4 g SEBS (Kraton G1645). After dissolution, the ink was homogenized using a speedmixer (FlackTek 330) at 2,000 rpm for 5 min. The ink was then drop-cast onto stainless steel Petri dishes to achieve a film thickness of 3 mm. The substrate was dried in a fume hood for 2 h and cured at 60 °C for 1 h. After curing, the stretchable SEBS substrate was peeled off from the petri dish.
In vivo testing
All swine studies were approved by and performed in accordance with the Committee on Animal Care at the Mblockachusetts Institute of Technology. All fish studies were approved by and performed in accordance with the Institutional Animal Care and Use Committee of Boston College. Additional details and extended discussion can be found in Supplementary Text 6.
Fabrication of MUSAS fish tag with a temperature sensor
A ProtoLaser R4 laser cutter (LPKF) was used to pattern the top and bottom 35-μm-thick copper claddings of an RT/duroid 6010.2LM laminate (Rogers), which features a 0.635-mm-thick ceramic–PTFE composite dielectric core. The laser was then employed to ablate a via hole through the substrate, establishing an electrical connection to the top copper layer using a soldered 32 AWG feedthrough wire. The antenna geometry, measuring 12 mm × 6 mm, was laser-cut from the patterned RT/duroid laminate. A Magnus S3 tag chip (Axzon) was mounted onto the ground plane with a non-conductive ultraviolet-cured epoxy adhesive. Chip-to-antenna interconnections were made via thermosonic gold ball bonding, reinforced with silver conductive paste and cured at 65 °C for 40 min using a C174740 Mech-El MEI Marpet 1204B wire bonder. A superstrate with hatched top and bottom copper claddings on an RT/duroid 6010.2LM core was then fabricated and affixed to the top surface of the antenna with an adhesive layer. Finally, the blockembled antenna was packaged onto the remora device by underfilling and encapsulating the bottom and sides with 5055 UV Curing Silicone Adhesive (Henkel Loctite) epoxy resin. See Supplementary Information for details of in vitro and in vivo tests on a tilapia model.
Fabrication and evaluation of MUSAS impedance biosensor for detecting gastroesophageal reflux
The impedance sensor was laser-cut on single-sided flexible copper laminate (Pulsar Professional fx FR4 5 mil ½-oz copper). Each of the 13 traces were 75 μm wide and 75 μm apart from each other. Alignment of the cut was performed with camera on a R4 laser (LPKF). Parylene coating was then performed on a PDS 2010 Labcoater (Special Coating System), to prevent unnecessary shortcut of the impedance sensor. A Kapton tape mask (0.03-mm thickness, McMaster-Carr) was cut with the R4 laser to cover the impedance sensor traces during the parylene-coating procedure. After completion of the parylene coating, the Kapton tape mask was peeled off. To improve the electric conductivity and sensitivity of the traces of the impedance sensor, electron beam evaporation was performed to deposit gold on the copper traces. The deposition procedure included coating a 10-nm adhesion layer of titanium and then 200-nm gold and was performed on EE-4 E-beam evaporator (Denton).
For impedance measurement of tissue, we used a ISX-3 Impedance Analyzer (Sciospec) equipped with Sciospec software v2.0.8 for four-point impedance measurements. The measurements spanned frequencies between 100 Hz and 1 MHz. For the in vivo study, MUSAS impedance biosensors were placed via an over tube into the oesophagus of anaesthetized female Yorkshire pigs (70–95 kg) and self-adhered through the contraction of the oesophagus. A gastroesophageal reflux model was created by using an endoscope to periodically spray gastric fluid, obtained from the pig itself, into the oesophagus.
Fabrication of gastric resident dosage forms for sustained release of CAB
The polycaprolactone (PCL) matrices containing CAB were prepared with melt mixing. The drug stability of the matrices was confirmed in previous research50. Specifically, PCL and CAB were weighed in a 10-ml glblock vial. The vial and a negative mould for the PCL patch were then heated on a heat plate (Thermo Scientific) to 75 °C. The matrices were melted and well mixed before transferring to the mould. After cooling to room temperature, the PCL–CAB patches were adhered onto fabricated MUSAS with ultraviolet-cured epoxy (Henkel Loctite).
The Ecoflex matrices containing CAB were directly prepared by mixing CAB with Ecoflex 0030 for moulding the suction cup of MUSAS.
In vitro and in vivo evaluation of pharmacokinetics of CAB
In vitro evaluation of the drug release of CAB was conducted in a release medium of simulated gastric fluid containing 5% w/v Tween 20 surfactant (Thermo Scientific). The drug-loaded MUSAS were placed in 10 ml of the release medium in a 37 °C incubator (New Brunswick Innova 44/44R) shaking at 250 rpm. At 2 h, 6 h, 1 day, 2 days, 3 days, 5 days and 7 days, 1 ml of the medium was sampled and stored at −20 °C until high-performance liquid chromatography (HPLC) blockysis, as described later. During each sampling, the remaining release medium was replaced with fresh medium.
In vivo pharmacokinetics were performed in female Yorkshire pigs (55–95 kg) in an unblinded fashion. MUSAS loaded with 40% CAB in the PCL matrices were either dropped through an over tube or endoscopically placed in the stomach of anaesthetized Yorkshire pigs, which were fitted with ear catheters. The pigs were fed and monitored daily in the morning and afternoon with a laboratory mini-pig grower diet, 5081, along with midday snacks of fruits and vegetables. At 2 h, 6 h, 1 day, 2 days, 3 days, 5 days and 7 days, 5 ml of blood was sampled via the ear catheter. The blood samples were centrifuged for serum separation at 4,000 rpm (Eppendorf 5810r) for 10 min and then stored at −80 °C until bioblockysis, as described later.
Synthesis and in vitro characterization of mRNA nanoparticles
Messenger-RNA-loaded lipid nanoparticles (LNPs) were made using a typical four-component lipid mixture. Specifically, ethanol-based solutions of SM102, 1,2-distearoyl-sn-glycero-3-phosphocholine, cholesterol and 1,2-dimyristoyl-sn-glycero-3-phosphoethanolamine-N-[(methoxy(polyethylene glycol)−2000) (ammonium salt) were mixed to achieve a molar ratio of 50:10:38.5:1.5. Firefly luciferase mRNA (Trilink) was dissolved in 10 mM citrate buffer, pH 3. The lipid solution was mixed with mRNA solution at a volume ratio 1:3 to achieve a SM102/mRNA weight ratio of 12.86. The LNPs were placed on ice for 10 min to complete the complexation. For measuring the activity of fresh LNPs, the LNPs were diluted in appropriate media and added to the cells. For measuring the activity of freeze-thawed LNPs, the LNPs were diluted with 200 mg ml−1 sucrose in 10 mM citrate buffer at a volume ratio of 1:1 and incubated at 4 C for 1 h. The LNPs were then frozen at −20 °C for 1 h. The LNPs were then thawed, diluted with complete media and added to the cells.
In vitro studies were conducted with primary human oral epithelial cells (Celprogen). Cells were seeded in a 96-well plate overnight. On the next day, LNPs (fresh, freeze-thawed with sucrose solution and freeze-thawed without sucrose solution) were added to the cells to achieve an mRNA concentration of 1 μg ml−1. The cells were incubated with the LNPs overnight. Transfection efficiency was measured using the SteadyGlo blockay using the manufacturer’s recommendations.
Administration of mRNA nanoparticles to swine
The mRNA LNPs were prepared as described for the in vitro studies. The firefly-luciferase-mRNA-loaded LNPs diluted in sucrose solution were transferred into the MUSAS and frozen at −20 °C for 1 h. Each MUSAS was loaded with 12.5 μg of mRNA. The MUSAS was applied to the pig buccal and pharynx manually with surgical forceps 8 h to 24 h before euthanasia. Immediately after the pig was euthanized, the site of administration and control were collected and placed in cold DMEM media (Thermo Fisher Scientific) containing 10% fetal bovine serum. Within 30 min of pig euthanasia, the tissue was immersed in 0.3 mg ml−1 potblockium luciferin solution (Gold Biotechnology) in PBS without calcium and magnesium. For increased diffusion of the substrate, luciferin solution was injected into the swine oesophageal tissue after submersion and before imaging. Bioluminescence was captured over a 30-min span with an in vivo imaging system (PerkinElmer) to capture bioluminescence, via LivingImage software 4.8.2 (PerkinElmer). The luminescent images were taken using Field of View D, automatic exposure time, medium binning, F/Stop = 1, and taking images every minute.
Imaging
Profilometry
Sample surface roughness and three-dimensional reconstructions were quantitatively determined by an optical VK-X3000 profilometer (Keyence) using the ×5 and ×10 lens. The ring and lens lighting were used together and set to maximum intensity. To compare the smoothness across the specimens, the surface roughness parameter Sa (areal average roughness) was evaluated across representative fields of view using the included VK viewer software v2.2.0.135 (Keyence).
Confocal microscopy
Confocal microscopy to measure hydrostatic differentiation of MUSAS’ underwater adhesion
A near-infrared fluorescent dye solution, Sulfo-Cyanine5.5 (Cy5.5) (Thermo Fisher Scientific), was prepared at a concentration of 0.01 mg ml−1 (40 ml) to stain 1 ml of water spread onto a microscopic glblock slide for confocal microscopy. Confocal microscope imaging was performed on MUSAS before and after adhesion to the water-rich glblock slide, using a 635-nm laser line with a measuring depth of 400 µm, with a FV1200 Laser Scanning Confocal Microscope (Olympus).
Firefly-luciferase-mRNA transfection visualization in pharyngeal tissue using MUSAS via immunofluorescence confocal microscopy
Fixed oesophageal pig tissues transfected for firefly luciferase expression with MUSAS along with untransfected controls were stained using DAPI (Thermo Fisher Scientific), as well as firefly luciferase polyclonal antibody primary antibody (Thermo Fisher Scientific) with a 1:2,000 dilution ratio, conjugated with goat anti-rabbit IgG (H+L) cross-adsorbed, Alexa Fluor 647 secondary antibody (Thermo Fisher Scientific) with a 1:500 dilution ratio. Five immunohistochemistry (IHC) slides per block of pig oesophageal tissue were prepared for imaging. Fluorescence images were taken with a FV1200 Laser Scanning Confocal Microscope (Olympus) with two channels: DAPI (405-nm laser line) and AlexaFluor647 (635-nm laser line). Images were taken using a ×10 objective. All images were processed using Fiji (Image J 1.54) software.
Scanning electron microscopy
Before scanning electron microscopy, all samples were mounted inside a copper vise and then subjected to vacuum and liquid nitrogen inside the electron microscope vacuum cryo manipulation tool (Leica). Using the portable temperature-controlled vacuum arm, each cryogenic sample was transferred to the ACE 600 high-vacuum sputter coater (Leica), which cryo-fractured a fresh cross-section and sputter-coated approximately 10 nm of platinum. This conductive coating prevented excessive surface charging artefacts during scanning electron microscopy imaging. Using the same portable vacuum arm, each cryogenic sample was transferred to the Gemini 360 SEC scanning electron microscope (Zeiss). While under high vacuum, using the secondary electron detector, low-voltage imaging (2–3 kV) was used to prevent damage from electron bombardment while providing high surface detail. Typical imaging conditions would also entail a probe current near 2 nA and a working distance between 6 mm and 8 mm.
Fluorescence microscopy
An ex vivo study was performed to evaluate the bioavailability of a nanoparticle formulation delivered through MUSAS; 215 µl of fluorescent polystyrene nanoparticles were prepared with FluoSpheres Polystyrene Microspheres (Thermo Fisher Scientific) and stored at −80 °C. The fluorescent nanoparticles were then subcutaneously injected, smeared with a pipette or delivered via MUSAS to freshly collected porcine oesophagus tissue. The oesophagus tissues, including a negative control, were resected, rapidly frozen in OCT gel (Agar Scientific) using liquid nitrogen and sectioned using a cryostat microtome. Subsequently, the sectioned tissues were imaged using an EVOS fluorescence microscope (Life Technologies) with excitation and emission wavelengths of 580 nm and 605 nm, respectively.
Bioblockytics
In vitro HPLC
Dissolution samples in simulated gastric fluid were directly blockysed by HPLC and ultraviolet detection on a 1260 Infinity system (Agilent Technologies). Samples were injected at a volume of 2 µl onto an Agilent EC-C18 Poroshell column (3.0 × 50 mm, 2.7 µm particle size dp) held at 25 °C. The mobile phase consisted of 0.1% formic acid in water (v + v, A) and acetonitrile (B), pumped at 800 µl min−1 with a gradient programme of: 0 min, 5% B; 8 min, 60% B; 8.1 min, 95% B, over 10 min and with an equilibration time of 2 min. Eluite was quantified with a diode array detector at CAB’s local absorption maximum at 258 nm in the ultraviolet region at 5 Hz.
In vivo liquid chromatography–mblock spectrometry
Porcine serum samples were prepared via protein precipitation at a 1:3 volume ratio of serum to acetonitrile with bictegravir or verapamil as an internal standard. The liquid chromatography–mblock spectrometry method used was validated according to FDA recommendations, on high-performance liquid chromatograph’s coupled to Agilent 6495 triple quadrupole mblock spectrometers in positive mode. Concentrations were calculated based on the linear regression of CAB response relative to internal standard response.
Specifically, samples were injected at a volume of 10 µl onto the same column using the same mobile phase above, but without temperature control. The gradient programme used a 400 µl min−1 flow rate with 0 min, 5% B; 0.5 min, 5% B; 4 min, 95% B, with a run time of 5 min and equilibration time of 2 min. The Agilent Jet Stream source used the following parameters: gas temperature, 200 °C; gas flow, 14 l min−1; nebulizer pressure, 20 psi; sheath gas temperature, 250 °C; sheath gas flow, 11 l min−1; capillary voltage, 3,000 V; nozzle voltage, 1,500 V; high radiofrequency, 150 V; low radiofrequency, 60 V. The same transitions for CAB above were used, except both at 38-V collision energy. Verapamil was used as an internal standard, quantified with the transition from 455.1 m/z to 303.1 m/z at 35 V, and qualified with the transition from 455.1 m/z to 303.1 m/z at 40 V.
Histology fixation
Unless specified, histology samples were fixed with 10% formalin for 24 h and then stored in 70% ethanol before embedded in paraffin for histopathological blockysis.
Statistical quantification and blockysis
Statistical quantification and blockysis were performed via Prism 9.3 (GraphPad). All error bars represent standard deviation (s.d.), except in pharmacokinetics studies conducted on independent Yorkshire pigs, where standard error of the mean (s.e.m.) was used to measure the population variability. For box plots, the box represents the median and the Q1 and Q3 quartiles (25% and 75%), and whiskers extend to the maximum and minimum values. For violin plots, the density distribution is calculated using the often-used Gaussian kernel density estimator, with dashed and dotted lines representing the median and the Q1 and Q3 quartiles (25% and 75%), respectively. The theoretically optimal Gaussian kernel density estimator, which minimizes the mean integrated squared error, is calculated with a bandwidth given by (1.06widehat{sigma }{n}^{-1/5}), where (widehat{sigma }) is the standard deviation and n is the sample size. Student t-test and blockysis of variance (F-test) were performed to compare differences between two groups, and among three or more groups, with details defined in the legends of relevant figures. We chose several P values to systematically evaluate the statistical significance, including P ≤ 0.05 (*) as the entry level of significance, P ≤ 0.01 (**) for highly significant, P ≤ 0.001 (***) and P ≤ 0.0001 (****) for extremely significant. The number of independent experiments of replicates and definition of significance level are further elaborated in each figure, figure caption and relevant methods section where statistical quantification and blockysis was performed.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
