Rapid Prototypable Biomimetic Peristalsis Bioreactor Capable of Concurrent Shear and Multi-Axial Strain

The involuntary contraction and relaxation of smooth muscle layers causes peristalsis, or wave-like movements through various organs, including the gastrointestinal tract, uterus, and ureters [Huizinga, 1999; Kunz and Leyendecker, 2002; Hosseini et al., 2018]. Forces associated with peristalsis regulate diverse biological functions including nutrient absorption, transport during reproductive processes, and urine dynamics. Peristalsis plays a central role in physiology and consists of two main mechanical components: multi-axial strain, and fluid shear (online suppl. Fig. 1; for all online suppl. material, see www.karger.com/doi/10.1159/000521752) [Yokoyama and Ozaki, 1990; Gayer and Basson, 2009]. Multi-axial strain is composed of radial and axial strain; in the gastrointestinal system and ureters, strain is induced by the contractions of the circular and longitudinal muscle layers [Mittal, 2011; Vahidi and Fatouraee, 2012]. In the uterus, only the circular muscles are responsible for uterine contractions [Kunz and Leyendecker, 2002]. In most organ systems, peristalsis is meant to aid in the transport (propulsion) of material through the organ. Flow of fluid or luminal content that accompanies this propulsion creates the induced shear [Fung and Yih, 1968; Aranda et al., 2015; Waldrop and Miller, 2016]. Within the organs that experience peristalsis, the cells are therefore continuously exposed to both multi-axial strain and shear stress. These mechanical signals are integrated by the cells into biochemical responses, through mechano-transduction [Cei et al., 2016]. Aberrant peristalsis, or dysperistalsis, is a patho-physiological condition associated with functional gastrointestinal disorders and endometriosis [Leyendecker et al., 1996; Iwakiri et al., 2017; Patel and Thavamani, 2021]. Given the central role of peristalsis in homeostasis, and the role aberrant peristalsis plays in the development and sustenance of patho-physiology, it is advantageous to include peristalsis and its associated kinematics in in vitro studies of disease and development.

The kinematics of peristalsis is complex, and hence challenging to replicate holistically in vitro without the use of bioreactors or microfluidic approaches. Currently, there are several examples of peristalsis bioreactors associated with the gastrointestinal tract at the microfluidic and the macro bioreactor levels [Kim et al., 2012; Cei et al., 2016; Costello et al., 2017; Wang et al., 2018; Zhou et al., 2018]. However, several of these approaches to study peristalsis, along with ureter and uterine models, often simplify the kinematics of peristalsis into perfusion-based fluid shear alone, or cyclic strain or cyclic uniaxial strain alone [Basu et al., 2012; da Rocha and Smith, 2012; Wang et al., 2014; Vardar et al., 2015; Hellstrom et al., 2017; Elad et al., 2020a; Elad et al., 2020b; Kim et al., 2020]. Neither of these two simplified stimuli adequately capture the concurrent multi-axial strain and shear of peristalsis. Appropriately mimicking both forces concurrently is important, because several cell types respond differently to shear stress and strain [Jufri et al., 2015; Ostrowski et al., 2016; Meza et al., 2019].

The ability to apply concurrent multi-axial strain and shear stress was therefore a central motivator to our new peristalsis bioreactor design. Our approach circumvents the limitations of current systems that either rely heavily on perfusion or uniaxial tension. Here we present a novel bioreactor design where a biocompatible membrane experiences cyclic multi-axial strain produced by a rotating screw drive, while simultaneously experiencing pulsatile fluid flow-induced shear stress. We integrated predictive finite element modeling and rapid prototyping to generate reproducible peristalsis bioreactors. We validated the cellular effect from the mechanical forces of our system through the use of human mesenchymal stem cells (hMSCs) due to their inherently mechanosensitive nature [Hoey et al., 2012; Kim et al., 2015; Kuo et al., 2015; Liu et al., 2015]. Overall, the objective of this work was to develop a bioreactor capable of mimicking the kinematics of peristalsis and elucidate changes in the cellular response to this unique mechanical event (Fig. 1).

Cell culture reagents utilized in this study including validated bone marrow-derived hMSCs and their nutrient medium was purchased from Lonza (Walkersville, MD, USA). According to the manufacturer’s certification of analysis, hMSCs were positive for CD73, CD90, and CD105, tested via flow cytometry analysis. Polydimethylsiloxane (PDMS) was purchased from DOW Chemical (Midland, MI, USA). All other chemical reagents were purchased from Sigma Aldrich (St. Louis, MO, USA), unless otherwise indicated. Antibodies used for cellular staining were purchased from Santa Cruz Biotechnology (Dallas, TX, USA), unless otherwise indicated. Custom-made oligos were purchased from Integrated DNA Technologies (Coralville, IA, USA). All other molecular biology-grade reagents were purchased from Thermo Fisher Scientific (Waltham, MA, USA).

Computer aided design (CAD) models of the bioreactor prototype were designed and edited in SolidWorks (Dassault Systèmes, Waltham, MA, USA). These models were used for conceptual design and computational fluid dynamics (CFD) analysis. The bioreactor parts include: (1) bioreactor top; (2) bioreactor bottom; (3) rotating screw drive; and (4) motor support. These parts are annotated in Figure 2a.

The bioreactor was designed in such a way that a screw drive was housed in the bioreactor bottom, connected to an actuating DC motor that caused rotation of the screw drive. A PDMS membrane was used to create the “wall” (component 5 of the bioreactor, Fig. 2a). The PDMS membrane was placed abutting the screw drive (component 3), and clamped down by the bioreactor top (component 1). The biocompatible membrane was held in place by the pressure exerted from clamping down the bioreactor top to the bioreactor bottom. When clamped, the screw drive caused a 1.6-mm indentation in the 3-mm thick PDMS. The bioreactor top had a raised cell chamber (component 7), and an inlet (component 8) and outlet (component 9) for nutrient medium flow, enabled by a peristaltic pump (illustrated in Fig. 1). Cells were seeded on a constant area in the PDMS “wall.” The rotation of the screw drive and simultaneous peristaltic flow of nutrient medium was used to mimic the kinematics of peristalsis on the cell seeded monolayer, examined via computational modeling in the following section.

Rapid prototyping and manufacturing were executed using a Prusa iMK3 fused deposition modeling 3D printer. By utilizing CAD modeling in conjunction with 3D printing, modifications to the bioreactor design were efficiently manufactured and implemented into testing. For the final prototype, polylactic acid filament was chosen as the main substrate material as it was less prone to warping during the printing process [Pei et al., 2015]. CAD models were sliced using the PrusaSlicer software using the “quality” print condition of 0.1 mm. The models were printed at 100% infill to remove interstitial space and eliminate leakage. CAD renderings created on SolidWorks are demonstrated in Figure 2b, which resulted in a final print product, photographed in Figure 2c.

The “wall” (component 5) of the bioreactors was fabricated using PDMS membranes, manufactured (10:1 pre-polymer base to crosslinker) by pouring into custom membrane molds and degassed for 1 h. Custom membrane molds were designed to produce PDMS membranes of uniform dimensions (7 × 4 × 0.3 cm). Membranes were cured at 60°C for 4 h and allowed to cool to room temperature overnight. Once PDMS membranes were cured, they were prepared for cell seeding using protocols outlined in the following sections.

The cell chamber (component 7, a part of component 1 in Fig. 2a) was designed as a CAD model on SolidWorks, as outlined in the section above. The model was then uploaded into ANSYS (Canonsburg, PA, USA) Fluent CFD software, where the CAD model was meshed and boundaries were defined as follows: (i) inlet, (ii) outlet, (iii) wall, and (iv) cell seeding area on the wall. The inlet (component 8) and outlet (component 9) represented the vertical channels from which media was pumped into and out of the cell chamber (component 7) on a constant mass-flow basis from the peristaltic pump. The “wall” was defined as all surfaces in contact with the fluid flow, excluding the wall membrane (shown in green, online suppl. Fig. 2a). The “cell seeding area” was defined as the luminal surface of the PDMS membrane “wall” that interfaced with the fluid flow, as well as the area where cells would be seeded during in vitro cell studies (shown in red, online suppl. Fig. 2a). This area remained constant across all experiments to ensure uniformity (2.79 cm²).

In order to determine the mass-flow rate in the CFD analysis, a theoretical flow rate was calculated for a fluid shear of 0.4 Pa through a rectangular cross-section. The shear of 0.4 Pa was chosen based on shear stresses experienced by various peristaltic organs, as reported in the literature [Kublickiene et al., 2000; Avvisato et al., 2007; Kimura et al., 2018]. The calculated optimized mass-flow rate to produce a minimum shear of 0.4 Pa was 28 mL/min.

The flow rate was integrated into the ANSYS software to model the flow across the “wall” itself and with the cell characteristics exhibited in the bioreactor. The cell was modeled as a spherical object with 30% embedded into the membrane wall while maintaining a perfect spherical conformation. The 30% embedding was intended to represent a more accurate cell anatomy experiencing the flow while approximating the cell-membrane interaction as rigid bonding. Using this method, the average shear over the artificial surface was assessed and quantified.

A finite element method was used to estimate the strain distribution induced in the membrane as the results of the screw indenting the membrane followed by its rotation. The simulations were performed using Abaqus/Standard Software (Abaqus, Johnston, RI, USA) implementing the implicit solution method. Deformations were considered quasi-static and inertia forces were assumed to be negligible. The screw was modeled as a rigid body. The membrane was modeled as a third-order Ogden hyperplastic material with material properties listed in online supplementary Table 1, derived from tensile testing of 10:1 PDMS membranes (online Suppl. Fig. 3).

The boundary condition was considered as follows: (i) the surface-to-surface contact interactions were applied via penalty friction formulation; (ii) the membrane geometry included a “wall” (shown in green) and the cell seeding area (shown in red; online suppl. Fig. 2a). The water seal ridge (yellow) and the outermost edge of the bioreactor were replaced by a clamped boundary condition in the simulation. The margin around the “wall” (green) was constrained by fixing the z-direction (preventing out-of-plane displacements), but remained free to have in-plane deformations (online suppl. Fig. 2b). The cell seeding area (red) was allowed to freely deform. The loading consisted of two steps: (i) the screw indenting the membrane in the z-axis for 1.6 mm (denoted by t = 0 s), and (ii) the screw rotating at 12 RPM around the x-axis while being held at the indented depth. One complete cycle, with the physical time duration of 5 s, was simulated.

Strain was experimentally measured using a custom polyvinyl alcohol-polydopamine (PVA-PDA) hydrogel strain sensor following protocols established for strain measurement [Liu et al., 2018; Wang et al., 2019]. Briefly, 10% (w/v) PVA solution, and a 160-mg/mL PDA solution were mixed and sonicated to produce a homogenous mixture. Then, 0.04 g/mL of sodium tetraborate solution was added to induce gelation of the PVA-PDA mixture, to result in a PVA-PDA hydrogel. The PVA-PDA hydrogel was then desiccated for 2 h and put through 2 cycles of a 12-h interval freeze-thaw process. This hydrogel was then inlayed into a strain sensor mold made from 3-mm thick VHB tape (3M, Maplewood, MN; USA) and copper foil placed along the X-direction. The piezoelectric, PVA-PDA strain sensor was then inserted into a modified bioreactor that would securely fit the strain sensor in the area representative of the cell seeding area on the PDMS membrane. It was also secured using cable ties to better simulate the in vitro bioreactor setup conditions. In the strain calibration set up, the screw drive was replaced with fixed, calibrated strain inducers to induce 0, 10, 26, and 41.7% strain maximally in the radial direction. Changes in electrical resistivity were measured using a digital multimeter, and a calibration curve was generated (online suppl. Fig. 4). To calculate the actual strain experienced by the screw drive, the hydrogel strain sensor was strained using the screw drive in the bioreactor and the resistivity was measured. The resulting strain was extrapolated from the calibration curve according to the measured resistivity (star, online suppl. Fig. 4).

hMSCs (Lonza Bioscience) were cultured in mesenchymal stem cell growth medium supplemented with MSCGM^TM SingleQuots^TM Supplement Kit (Lonza Bioscience). hMSCs were evaluated for trilineage differentiation capacity according to the manufacturer (online suppl. Fig. 5). Cells were cultured on tissue culture dishes and used up to Passage 5 as recommended by the manufacturer to avoid spontaneous differentiation. In order to maximize seeding of hMSCs on PDMS, Collagen I was used to coat the cell seeding area of the PDMS (1.8 cm²) at 200 μg/mL. Collagen-coated PDMS was stored at 4° until use. hMSCs were seeded on the PDMS at 200,000 cells/mL and incubated overnight to allow for attachment to the PDMS.

All items used in bioreactor set-up were sterilized using ethanol and exposure to ultraviolet light for at least 10 min. Prior to full assembly, the remaining media from hMSC seeding on each PDMS membrane was removed and 25–30 mL of media was added to each bioreactor’s media reservoir. The bioreactor bottom was assembled by placing the screw drive, motor, and motor support in their respective locations (see Fig. 2a). Silicone grease was added to the top of the screw drive to minimize friction between the PDMS membrane and screw drive. The PDMS membrane was then transferred atop the screw drive and placed cell side up in the bioreactor bottom. The bioreactor top was then clamped down on to the membrane, and sealed using commercially available zip ties. The peristaltic pump tubing was connected to the bioreactor and media bottle to create a closed loop between the media, pump, and bioreactor. The assembled bioreactor, nutrient medium reservoir, and pump were then placed into the incubator. The motor and pump were connected to the Arduino that ran the pre-programmed code. A schematic of the full assembly is provided in Figure 1, step 4. The bioreactor ran for 24 h and cells were then collected for various downstream analyses. Three conditions were programmed with the Arduino: strain, shear, and peristalsis. The corresponding motor and pump speeds can be seen in Table 1. The differences seen between the motor speeds and flow rates across conditions were not significant. For the control condition, hMSC-seeded PDMS membranes were incubated statically for 24 h.

hMSC-seeded PDMS membranes from static and mechanically stimulated conditions were cut along the cell seeding area for ease of staining. Membranes were rinsed with PBS and fixed in 4% formalin for 45 min. Samples were blocked using 0.15% Triton X and 5% fetal bovine serum at room temperature for 1 h. Cells were incubated with fluorescently tagged primary antibodies for 1 h at room temperature. The primary antibodies used included Ki67 (Alexa Fluor 647 conjugated), phalloidin (Alexa Fluor 488 conjugated), and smooth muscle actin (FITC conjugated). A nuclear counterstain (DAPI) was also included in the primary antibody incubation. Unbound antibodies were rinsed using PBS, and cell-seeded PDMS membranes were mounted using an antifade mounting reagent. Fluorescence was observed using a Leica SP8 Confocal Microscope, with 5 independent, non-overlapping regions for analysis. NIH ImageJ was utilized to perform fluorometry.

Ki67⁺ cells were identified by the presence of red fluorescence within DAPI counterstained nuclei, and were quantified manually. Analysis was performed to determine the number of actively proliferating cells (Ki67 antigen in the nuclei) versus the total cell number (number of nuclei) to determine the percentage of proliferating cells.

hMSCs were stained with fluorescently conjugated phalloidin to visualize cytoskeletal networks and actin alignment at low and high magnifications. To analyze actin filament angle orientation, green channels from the higher magnification images were isolated in ImageJ. A single-pixel-line filter was applied at sequential rotational intervals (0°, 1°, 2°, etc.) and the brightness of the remaining pixels were summed [Püspöki et al., 2016]. Then, the summed brightness at each angle was computed into a fraction. The totality of the 180 fractions equaled 1. Angles were then grouped into increments of 30° by adding the fractions from 1° to 30°, 30° to 60°, etc. to assess the angle distribution trends.

All high-magnification actin filament images were also assessed for nuclei circularity using the DAPI channel. The blue channel was isolated in ImageJ to remove any Phalloidin stains from interfering with the circularity measurement. A filter was applied to the selected DAPI image to mask out everything except the cell nuclei. The circularity of each nucleus is measured by Eq. 1.

Circularity = 4 * π * (area/perimeter²)

A perfectly circular nuclei would result in a value of 1, and an elongated nuclei would return a value approaching zero.

Using fluorescently conjugated smooth muscle actin, the alpha isotype of smooth muscle was visualized in hMSCs exposed to the bioreactor. To analyze the presence of smooth muscle actin, the amount of green fluorescence was compared to the total blue fluorescence, normalized to a constant imaging area. This resulted in a normalized value for the relative expression of smooth muscle actin for each condition.

Alizarin Red S, Alcian Blue, and Oil Red O stains were used to visualize Osteocyte, Chondrocyte, and Adipocyte differentiation, respectively, following well-established protocols [Eggerschwiler et al., 2019]. Membranes were rinsed with PBS and fixed in 4% formalin for 15 min. Alizarin Red S and Alcian Blue were rinsed with DI water and incubated at 37°C for 45 min and at room temperature for 15 min, respectively. Oil Red O was rinsed with PBS and incubated at room temperature for 20 min. Unbound stain was rinsed using DI water or PBS, and cell-seeded PDMS membranes were mounted using an antifade mounting reagent. All stains were observed using a Zeiss Axio Observer D1 (Carl Zeiss, Jena, Germany) inverted microscope with Achroplan 10×/0.32 objective. Images were captured using a Tucsen IS500 camera.

RNA extraction was performed on harvested hMSCs using the RNeasy Mini Kit (Qiagen). RNA concentration and purity were evaluated using a NanoDrop OneC (ThermoFisher Scientific), and stored at −80°C until ready to use. Reverse transcription was performed following manufacturer’s protocols using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). qPCR was performed with a CFX96 Real-Time System (Bio-Rad) using the Applied Biosystems PowerSYBR Green PCR Mastermix (Thermofisher Scientific) for detection. Genes that were investigated include CNN1 (calponin), CALD1 (caldesmon), ACTA2 (smooth muscle actin), SM22α (transgelin), BGLAP (osteocalcin), ALPL (alkaline phosphatase), SPP1 (osteopontin), ACAN (aggrecan I), COL2A1 (collagen II), SOX9 (SRY-box transcription factor 9), COMP (cartilage oligomeric matrix protein), ADIPOQ (adiponectin), and LEP (leptin). The primer sequences used for each gene are shown in Table 2. Genes are categorized as follows: myocyte: CNN1, CALD1, ACTA2, and SM22α; osteocyte: BGLAP, ALPL, and SPP1; chondrocyte: ACAN, COL2A1, SOX9, and COMP; and adipocyte: ADIPOQ and LEP. Changes in gene expression were calculated using the 2ΔΔC_t method, with GAPDH as the housekeeping control [Livak and Schmittgen, 2001]. qPCR experiments were run in triplicates, with 3 independent biological replicates. Comparisons were drawn between control hMSCS that were not exposed to mechanical stimulus and each of the three applied stimuli.

Statistical analysis was performed on GraphPad Prism 9. All reported values are means ± SEM, resulting from 3 to 5 biological replicates. All qPCR data were normalized to control conditions within each experimental set and performed in triplicates over at least 3 biological replicates. Image analysis and morphometry included 5 non-overlapping fields of view from 3 to 5 biological replicates. ANOVA-based hypothesis testing was performed where appropriate, and statistical significance is indicated within each experimental data set, with associated p values.

The design of the peristalsis bioreactor as envisioned, along with its geometry and dimensions, resulted in shear stresses that fell within the range of reported values for peristalsis. The mass flow rate of 28 mL/min set on the pump produced a homogeneous velocity profile with a laminar flow over the PDMS membrane surface (Fig. 3a). When modeled solely with the characteristics of the PDMS membrane, i.e., the absence of cells, 28 mL/min produced an average shear of 0.3627 Pa on the surface of the membrane (Fig. 3b). When the cell was modeled individually as a spherical object embedded within the PDMS membrane, the computational model predicted an average surface shear of 0.4387 Pa (Fig. 3c). Based on the shear maps produced within the ANSYS Fluent software, both the membrane “wall” and cell model experienced a similar shear as indicated by the color mapping. The cumulative data from the shear simulations provided sufficient evidence of the bioreactor’s capability of producing at least the minimum target shear value of 0.4 Pa with the given design and flow rate of 28 mL/min.

The strain induced within the bioreactor was validated through both experimental and computational methods. The PVA-PDA piezoelectric hydrogel measured resistivity changes when indented with calibration strain inducers, in the range of 0–41.7%. The change in resistivity was fit to a linear regression line with R² = 0.93 (online suppl. Fig. 4). Using this calibration curve, the hydrogel strain sensor was used to measure experimental strain arising from the screw drive. Changes in resistivity values when indented within the bioreactor with the screw drive correspond to 15.9%, in the direction of measurement X, or E₁₁.

A whole range of multi-axial strains were computationally predicted using finite element simulations. In the computational determination of strain, proper peristaltic patterning was observed. Contours for the normal logarithmic strains E₁₁, E₂₂, and E₃₃, at the maximum indentation point, right before any rotation (i.e., t = 0 s), were estimated at the cell seeding side of the membrane (Fig. 3d). The strain fields indicated a periodic pattern resembling peristalsis waves. In particular, E₁₁ was greater at two ends of the screw (along the x-axis) and less pronounced in the middle, while E₂₂ showed a wave-like distribution along the perpendicular (y) direction, qualitatively capturing a peristaltic motion along x.

The computational determination of the strain also showed that maximum strain magnitudes consistently occurred at the half-rotation cycle (t = 2.5 s; Fig. 3e). The strains along the length of cell area (x-axis) were always maximum at the contact points of two crests with the membrane. The contact points monotonically moved away from each other and generated deeper indentations to t = 2.5 s (Fig. 3e) followed by an opposite pattern in the second half-cycle. The axial strain (E₁₁) measured by the hydrogel piezoelectric strain sensor at the mid-plane was in excellent agreement with the finite element estimation as indicated by the black marker in the E₁₁ graph (Fig. 3e).

In order to understand the effect of the kinematics produced in the peristalsis bioreactor on cells, we utilized hMSCs. hMSCs were seeded on to a defined area within the bioreactor (with maximal contact with the screw drive, and hence its associated strain). hMSCs subjected to static control, perfusion, strain, or peristalsis were processed for immunofluorescence following 24 h of stimulation. Sections of PDMS were stained with a proliferation marker (Ki67, red fluorescence; Fig. 4a). Ki67 antigen expression within the DAPI counterstained nucleus was considered a positive marker for proliferation [Ohta and Ichimura, 2000]. After 24 h of exposure to the respective conditions, strain (33.21 ± 4.89%), perfusion (25.23 ± 4.99%), and peristalsis (41.98 ± 4.80%) had significantly lower proliferation compared to the control (****p < 0.0001, one-way ANOVA, Fig. 4b). Peristalsis also had a 16.74% decrease in proliferation compared to perfusion (**p < 0.01, one-way ANOVA, Fig. 4b), indicating that the combined application of shear stress and strain in peristalsis influenced hMSC proliferation uniquely.

hMSCs were maintained as static controls or subject to strain, perfusion, or peristalsis in the bioreactor (Table 1). After 24 h, the PDMS was collected from bioreactors and stained with an actin filament marker to assess the actin alignment within the hMSCs (phalloidin, green fluorescence, Fig. 5a). The nuclei of the stained phalloidin images were assessed for circularity. All three mechanical conditions significantly (*p < 0.05 and ***p < 0.001, one-way ANOVA) differed from the control with increased circularities, implying that mechanical stimuli increase nuclear circularity. Strain and perfusion increased from the control by 1.38- to 1.6-fold, respectively (Fig. 5b). hMSCs subject to peristalsis were the most circular (compare 0.71 ± 0.017 in peristalsis to 0.4 ± 0.062 in static controls; ****p < 0.0001, one-way ANOVA, Fig. 5b). Overall, the cellular nuclei were more circular in conditions with increased mechanical stimuli, specifically peristalsis.

Phalloidin images were also evaluated based on actin filament angle orientation to quantify cellular alignment changes in response to mechanical stimulation compared to static controls. hMSCs in static controls were randomly oriented, with no apparent frequency distribution in any of the 30° increment bins (Fig. 5c). In fact, random distributions of orientations ranging from 9.7 to 19.7% was observed in all the bins in static controls. In contrast, the application of perfusion resulted in greater alignment along the 60° and 120° bins (holding between 17.7 and 19.5% of cells). The application of strain promoted approximately 18.7% fiber alignment along the 30° axis, in line with documented evidence that cells were likely to align perpendicularly in response to strain [Ghazanfari et al., 2009]. Interestingly, the application of peristalsis resulted in an even distribution of fiber alignment uniformly across 0–180°, with the range of cells falling into each bin between 13.2 and 16.5%.

In order to further investigate the changes observed through image analysis, we used qPCR to identify underlying changes in gene expression. hMSCs exposed to mechanical stimuli were compared and normalized to static controls maintained for the same duration (24 h), indicated by the dotted line at 1 (Fig. 6a). Notably, peristalsis had increased expression compared to strain in CNN1, CALD1, ACTA2, SM22α, and ALPL (****p < 0.0001), and BGLAP and SPP1 (*p < 0.05; two-way ANOVA, Fig. 6a). Peristalsis also had increased expression compared to perfusion in CNN1, CALD1, ACTA2, SM22α, and ALPL (****p < 0.0001) and BGLAP (***p < 0.001; two-way ANOVA, Fig. 6a). Along with those genes, hMSCs stimulated with strain showed increased expression relative to peristalsis in ACAN, COL2A1, SOX9, ADIPOQ, and LEP (****p < 0.0001) and COMP (***p < 0.001; two-way ANOVA, Fig. 6a). Perfusion demonstrated an increase in expression in COMP (****p < 0.0001) compared to both peristalsis and strain (two-way ANOVA, Fig. 6a).

Notably, four myogenic markers (CNN1, CALD1, ACTA2, and SM22α) were significantly increased in hMSCs stimulated with peristalsis, providing strong evidence that hMSCs were differentiating toward the myogenic lineage. In order to validate qPCR findings pointing towards myogenic differentiation, hMSCs exposed to strain, perfusion, or peristalsis were stained with α-smooth muscle actin. These results indicated an increased expression of α-smooth muscle actin in hMSCs stimulated with peristalsis (6.409 AFU) compared to perfusion and control (3.841 AFU and 3.917 AFU, ****p < 0.0001) and strain (4.820 AFU, ***p < 0.001; one-way ANOVA, Fig. 6c).

In addition to α-smooth muscle actin staining, we also confirmed the absence of staining for trilineage osteogenic, adipogenic, and chondrogenic differentiation (Fig. 6d). Static controls served as negative controls, while online supplementary Figure 5b demonstrates positive controls for trilineage differentiation of hMSCs using widely validated protocols [Eggerschwiler et al., 2019]. hMSCs subject to mechanical stimulus in the bioreactor did not demonstrate positivity for osteogenic (Alizarin Red S), chondrogenic (Alcian Blue), or adipogenic (Oil Red O) staining (Fig. 6d).

Taken together, our data suggest that an hMSC differentiation event was occurring. Depending on the kind of mechanical stimulation the cells were exposed to within the bioreactor, the orientation of differentiation lineage varied, with peristalsis promoting myogenic differentiation in hMSCs.

Peristalsis is a nuanced mechanical stimulus, occurring across several smooth muscle organ systems including the intestines, uterus, airways, and ureters [Huizinga, 1999; Kunz and Leyendecker, 2002; Hosseini et al., 2018]. Peristalsis, and thereby dysperistalsis, is central to physiology and patho-physiology [Leyendecker et al., 1996; Iwakiri et al., 2017; Patel andThavamani, 2021]. With a growing body of evidence suggesting that mechanotransduction plays a vital role in development and disease, it is advantageous to include peristalsis in the study of cells, tissues, and organs [Jaalouk and Lammerding, 2009; Maurer and Lammerding, 2019]. Biomanufacturing of cells for cellular therapy or regenerative engineering of tissues and organs will all benefit from having received mechanical cues during in vitro bioprocessing. In several current in vitro studies, the kinematics of peristalsis is often simplified to fluid shear, or cyclic uniaxial strain. However, peristalsis is a summation of multi-axial strain that propels luminal content, which them simultaneously results in shear stress. The mechanics associated with peristalsis therefore involve concurrent multi-axial strain and shear stress.

Several current designs mimic peristalsis favorably, especially those associated with the gastrointestinal tract at the microfluidic and the macro bioreactor levels [Kim et al., 2012; Cei et al., 2016; Costello et al., 2017; Wang et al., 2018; Zhou et al., 2018]. However, a majority of the studies intended to study peristalsis-associated mechanotransduction simplify the kinematics of peristalsis to shear stress or cyclic strain alone. In one example, both longitudinal and circular contractions were incorporated to closely simulate a physiological intestine [Cei et al., 2016]. However, there was no additional shear stress added to the device, besides the fluid stress created by the contractions. Instances of peristalsis bioreactors within the reproductive tract are much fewer and far between [da Rocha and Smith, 2012; Hellstrom et al., 2017; Elad et al., 2020a], with models relying on fluid shear stress from perfusion [Elad et al., 2020b], or uniaxial cyclic tensile stretch [Kim et al., 2020]. In urothelial tissue engineering, bioreactors are used to precondition ureteral grafts prior to implantation, which results in urothelial cellular organization [Cattan et al., 2011; Janke et al., 2019]. The mechanical stimulus however is either fluid flow or cyclic stretching, neither of which are biomimetic of ureteric peristalsis [Basu et al., 2012; Wang et al., 2014; Vardar et al., 2015]. All of these devices incorporate very important aspects of shear or strain but still fall short of a complete peristaltic model. In comparing our device to those currently available, the full effect of peristalsis can be explored with the inclusion of the simultaneous administration of shear and strain in our novel model.

In order to assess the peristaltic capabilities of our device, the fluid shear was simulated via CFD software and the strain was computationally and experimentally determined. The shear simulations produced an average shear of 0.3627 Pa (membrane) and 0.4387 Pa (cell surface on the membrane). This falls well within the desired range of reported values for kinematics of peristalsis. Mainly, during peristalsis, intestinal shear ranges from 0.2 to 3.2 Pa, ureter shear ranges from 0.002 to 0.006 Pa, and uterine shear ranges from 0.7 to 2.9 Pa [Kublickiene et al., 2000; Avvisato et al., 2007; Kimura et al., 2018]. Currently, a flow rate of 28 mL/min produces a shear of 0.4 Pa. This flow rate is tunable, and can therefore produce a range of shear stresses that makes this peristalsis bioreactor.

The average of the minimum values from each of the three ranges previously reported is 0.3 Pa. The value of 0.4 Pa falls near the average of the shear experienced during peristalsis in the body, making it the target point for our device for initial proof of concept studies. The resulting computational analysis indicated that our peristalsis bioreactor was capable of mimicking shear stresses within the range of those reported in the literature. By modeling the target shear stress, our device can therefore be applied to a broad range of organ systems.

In addition to the shear analysis, the strain was computationally and experimentally analyzed. There is a large range of peristalsis-induced strain forces found in the body; strain can be anywhere from 1% to upwards of 20% depending on the organ system [Forouhar et al., 2006; Fujita et al., 2010; Kim et al., 2012, 2020]. The experimental strain value of 15.9% was within these reported ranges. The results of computational modeling also indicated that the screw drive acted as an effective method for controlling the magnitude and frequency of oscillating strain to generate the desired peristalsis waves.

With the evaluation of the shear and strain meeting physiological values, hMSCs were applied in the use of the bioreactor to determine the effect of various mechanical stimuli (perfusion, strain, and peristalsis) compared to static controls. hMSCs are known to respond to mechanical stimulation, lending them as good models to study changes in cellular response to mechanical stimulus. Direct mechanical stimulation, in the form of oscillatory fluid flow, of hMSCs in vitro has been documented to have a strong role in influencing stem cell behavior and differentiation [Hoey et al., 2012]. Specifically, hMSCs are well documented for how they respond to shear and strain forces [Simmons et al., 2003; Friedl et al., 2007; Zhang et al., 2012; Yuan et al., 2013]. When strain forces are applied to hMSCS, they have a tendency to differentiate toward an osteogenic, tenogenic, or myogenic phenotype and align perpendicularly to the applied strain [McClarren and Olabisi, 2018; Nam et al., 2019]. In terms of hMSC response to perfusion, hMSCs have an increase in gene expression related to differentiation and, at high rates of perfusion, an overall decrease in proliferation [Zhao et al., 2007; Becquart et al., 2016].

Our data indicate a shift in hMSCs toward differentiation when exposed to mechanical stimuli in both the immunofluorescence (Fig. 4, 5) and gene expression analyses (Fig. 6a). There is also a significantly different response to peristalsis compared to strain, perfusion, and static controls. Specifically, in the proliferation study, the different effects of each force, perfusion, strain, and the combined peristalsis, indicate a connection between mechanical stimuli and the proliferative capacity of hMSCs. In hMSC studies with strain or perfusion, extended time points, like 24 h, and increased shear rates have decreased overall proliferation, lending them toward differentiation [Song et al., 2007; Zhao et al., 2007; Yourek et al., 2010]. In our data, with the combination of both the perfusion and strain stimuli in peristalsis, there was a large decrease in proliferation, consistent with the findings in literature for strain and shear individually (Fig. 4b) [Songet al., 2007; Zhao et al., 2007]. With such a decrease in proliferation, we hypothesized that differentiation was likely happening and not proliferation. This idea was largely supported by the gene expression findings. With the increased expression of genes associated strongly with smooth muscle cells, like CNN1, CALD1, ACTA2, and SM22α [Naritaet al., 2008], in peristalsis, a clear differentiation toward myogenic lineage was occurring (Fig. 6a) [Lin and Lilly, 2014; Brunet al., 2015]. CNN1 and ACTA2 were also upregulated in perfusion, along with the osteocyte marker, SPP1, the chondrocyte marker, COL2A1, and the adipocyte marker, ADIPOQ. While there was no direct lineage differentiation in the perfusion condition, it is evident the hMSCs were expressing differentiation markers with a decreased proliferative capacity, consistent with the literature for a decrease in proliferation and maintenance of lineage potential [Franket al., 2016]. hMSCs exposed to strain also demonstrated an increase in gene expression in COL2A1, SOX9, COMP, ADIPOQ, and LEP. With the upregulation of two genes for both chondrocyte and adipocyte, there again was not a clear differentiation lineage. However, like with perfusion, these results corroborate a decrease in proliferation still allowing for lineage differentiation to occur. With each of the genes evaluated, the effect of, and subsequent response to, peristalsis, strain, and perfusion can be used to support the idea of increased differentiation and decreased proliferation.

Furthermore, the differentiation of hMSCs exposed to each condition was evaluated with a smooth muscle actin fluorescence stain and trilineage differentiation stains (Fig. 6b, d). Consistent with the qPCR data (Fig. 6c), the smooth muscle actin fluorescent images indicated an increase in actin expression compared to control, strain, and perfusion which is also consistent with myogenic differentiation reported in hMSCs [Liu et al., 2013].

Along with differentiation characteristics, cell shape was evaluated. The alteration in nuclear morphology can be seen in the actin fiber alignment results (Fig. 5c). The application of peristalsis resulted in an even distribution of fiber alignment uniformly across 0–180°, while strain and perfusion aligned across one or two axes. Specifically, hMSCs have been shown to align perpendicularly to strain which is evident in our results with the majority of strain cells aligning along the 30° axis [Nam et al., 2019]. In addition to actin alignment in response to mechanical stimulation, our results also indicate an increase in nuclear circularity, particularly in peristalsis (Fig. 5b). hMSCs in their stem-like state have an inherently elongated nuclei which would output a value less than 1 [Ankam et al., 2018], while differentiating hMSCs approach 1. Ultimately, in comparing our findings with the hMSCs to well-established data, we have shown that the mechanics within the peristalsis bioreactor impact cells differently, dependent on the mechanical stimuli (perfusion, strain, or peristalsis), in terms of gene expression, α-smooth muscle actin expression, alignment and nuclear morphology [Ito et al., 2007; Song et al., 2007; Zhao et al., 2007; Mishra et al., 2008; Robin et al., 2013; Elsafadi et al., 2016; Nam et al., 2019].

Peristalsis is central to many smooth muscle organs, including the gastrointestinal tract, uterus, and ureters, and is comprised of multi-axial strain (radial and axial strain) and shear stress. Our new peristalsis bioreactor design combining finite element modeling with rapid prototyping delivers multi-axial strain and concurrent shear stress to cells seeded on a biocompatible membrane. hMSCs, widely used in tissue engineering and regenerative medicine, sensed and responded to peristalsis within the bioreactor uniquely. A bioreactor such as ours will allow the kinematics of peristalsis to be replicated in several in vitro processes, including regenerative engineering and biomanufacturing. Furthermore, the bioreactor is tunable and amenable to the study of mechano-transduction-related processes involved in development and disease across several organ systems.

The results reported in this manuscript did not involve human or animal subjects.

The authors have no conflicts of interest to disclose.

This work was supported by the Texas A&M University Department of Biomedical Engineering and the Texas A&M Engineering Experiment Station (S.R.), and NIH R00HL138288 (R.A.).

S.R. and L.Z.C. designed the bioreactor, L.Z.C. and D.N. prototyped the bioreactor, performed computational models of shear stress and experimentally validated strain. A.J.C., L.Z.C., and D.N. contributed to design refinements, prototyping process parameter refinements supervised by S.B.M. D.N. performed all operational programming of microprocessors and bioreactor control, and additionally performed all automated cellular morphometry and analysis thereof. A.J.C. performed all biological experiments and downstream analysis of cells exposed to bioreactors. H.B. performed all the finite element modeling, supervised by R.A. S.R. conceived the project, designed experiments, and oversaw operational analysis, data acquisition and interpretation. A.J.C., L.Z.C., D.N., and S.R. wrote initial drafts of the manuscript, which were critically reviewed and revised by R.A., H.B., and S.B.M. All authors agreed on the final submission of this manuscript.

Data supporting the findings of this study are available from the corresponding author upon reasonable request.