Bioreactor Chambers Manufacturing and Processing
AIMS was composed of three main components: (1) interchangeable bioreactor chambers, (2) a control unit, and (3) a custom-built positive displacement pump (Fig. 1, S. Figure 1)12,15.
In total, four types of interchangeable bioreactor chambers (chambers) were designed and manufactured in this study. All chambers were modeled in Autodesk Inventor 2020 Pro (Autodesk. Inc, USA) based on prior work and with an objective to incorporate different generalized fluid flow directionality depending on catheter location within the ventricles. Chamber materials were investigated in a pilot study that led to the development of two types made of resin (single-inlet and a “dome” with multi-directional inlets), one of polydimethyl siloxane (PDMS, silicone), and one out of polyethylene terephthalate glycol (PETG). These chambers were designed to incorporate protein, single cells, cell monolayers, multilayers, or co-cultures, and viable tissue structures. (Fig. 2).
Resin chambers were manufactured using an Anycubic Photon 4K MSLA printer (Anycubic Technology CO., Limited, Hong Kong). The resin chambers (n = 45) were divided into two broad subcategories single inlet chambers for unidirectional flow (Resin) and multidirectional inlet chambers (Dome). Four iterations of the single inlet resin chambers with 0°, 45°,90°, and 135° inlet angles relative to the ventricular catheter were developed and printed to simulate CSF flow mixing specific to catheter positioning relevant to CSF bulk flow or pulsation production (N = 10/ per inlet angle) (S. Figure 2).
The high-throughput silicone chambers were modeled based on our prior work where the incorporation of proteins and single cells in suspension or attached to the catheter surface was imperative16. The silicone chambers were assembled using luer-type barb connectors on the inlet and outlet ports, silicone tubing, and a glass outer shell. Forty chambers were manufactured in batches of 10. Each batch was contained in a custom holding tray (Fig. 2).
High-throughput polyethylene terephthalate glycol (PETG) chambers were specifically designed for the incorporation of cell layers, tissue, or tissue-like structures for investigation of cell attachment and migration easily identifiable with real-time confocal and brightfield microscopy techniques without deconstructing the setup. These chambers were fabricated utilizing a vacuum forming machine (A3, Vacucu 3D, China) where a 3D printed resin mold was employed to transfer the chamber's geometry onto transparent PETG sheets. The geometry of the resin mold was iteratively modified based on computational Fluid Dynamics (CFD) analysis to mimic CSF flow patterns through the catheter holes in enlarged, fixed ventricles. Overall, 50 PETG chambers were manufactured for this study. All the manufactured chambers were visually inspected post-processing for potential malformations and geometrical defects using a Trinocular stereo zoom microscope (SM-4TZ-144A, AmScope, USA) with a 0.5X Barlow lens.
Flexible MRI-based models of patients’ ventricular systems were produced in three stages: filling, rotational molding, and demolding. Anycubic Kobra Max Fused Deposition Modeling printer (Anycubic Technology CO., Limited, Hong Kong) was utilized to produce a three-dimensional mold of the patient ventricular systems. The molds were designed such that the internal volume of the mold matched the volume of the patient ventricular system. In total 50 mL of uncured silicone was injected into the hollow molds through a 12-gauge stainless steel needle. The ventricles were exposed to biaxial rotation for three hours followed by an additional eighteen hours of stationary rest to complete the curing process. After the completion of the curing process, the molds were gently removed to expose the silicone ventricle model.
Computational Fluid Dynamics
The composite geometry in the Standard for the Exchange of Product Data (STEP) format was imported into ANSYS Fluent computational fluid dynamic software and was discretized into a nonuniform unstructured rigid grid. Finite element grid refinement was performed at the catheter drainage holes, lumen, and the inlet and outlet boundary face for each chamber variation. Adaptive uniform boundary layers were applied on all rigid faces of the computational model. Polyhedral mesh elements were used to capture curvature and proximity with the flow field regiment at the entrance of each drainage hole. Final mesh skewness ranged from 0.35–0.40 based on the bioreactor model simulated. In Fluent a laminar flow model was selected, and the kinematic viscosity of water was set equal to \(0.75\times {10}^{-6} \frac{{m}^{2}}{s}\) 17. A non-slip flow condition was applied at the walls of the computational domain. The inlet boundary conditions were set to match both the clinical measurement of pulsatile CSF outflow through EVD in the pre-and post-operation state. A constant zero pressure was specific at the outlet. A transient solver using the Navier-Stokes equations was selected to simulate a full cardiac cycle of 1 second. Vector velocity profiles were captured in the chamber and catheter domains using Fluent Post Processing to observe different velocity flow fields with respect to the reactor model (Fig. 3, S. Figure 3).
The Control Unit, Reciprocating Positive Displacement Pumps, and the Fluidic Components
The Control Unit consisted of an Arduino-based controller board and a custom-built, proprietary Python program with a custom user interface. The Control Unit interfaced with up to ten Reciprocating Positive Displacement Pumps (pumps) simultaneously, assembled according to our previously published study12,15. Briefly, five 3-ml syringes in each pump were actuated forwards and backward by a drive unit to induce flow. The pumps acted as intermediaries between the fluidic components and the computer software. Thus, the pumps interfaced with a stopcock or a series of check valves (Duckbill Check Valve, Qosina, USA) through silicone tubing. The pumps, check valves, bioreactor chambers, and flow sensors (SLF3s-1300F Series, Sensirion, Switzerland) were all placed in series in a circuit; flow instigated by the pumps traveled through the connected components and eventually drained into a separate reservoir (Fig. 1). AIMS was designed such that each one of the 50 pump channels was independent of the other bioreactor chambers. Specifically, there was no exchange of solution or flow interference between adjacent channels. The AIMS setup’s automated priming feature facilitated air bubble removal and constant fluid contact throughout the experiments. During the priming process, saline flowed through the chambers continuously for twenty minutes. The displaced air in the fluidic components was released in the reservoir and was forced out of the circuit through a filter with 0.22 \(\mu m\) pore size.
AIMS was developed to recapitulate a wide range of physiologic and pathophysiologic CSF flow patterns with varying pulse amplitude (0.05-55 \(\frac{mL}{min}\)), pulsation rate (1-400 \(\frac{Beats}{min}\)), and bulk flow rate (0.01-5 \(\frac{mL}{min}\))12. These variables were specified in the user interface to match clinical CSF flow measurements. The Python program transformed the user specifications into real-time commands that manipulated the pump’s output. Flow data were collected using the Sensirion USB Sensor Viewer program on a Hewlett-Packard desktop computer running Windows 10Pro operating system.
Capacity and Consistency Analysis
In addition to flow simulation capabilities, AIMS was also utilized as a modular platform for chamber testing and quality control (S. Figure 1A-C). A pressure manometer (Extech HD700 Differential Pressure Manometer − 2PSI, Teledyne FLIR LLC, USA) was utilized to quantitatively analyze the capacity of the manufactured chambers under differential pressure. Each chamber underwent ten pressurization cycles, with the maximum gauge pressure reaching 140.7 \(cm{H}_{2}O\). (13.8 kPa) (S. Figure 1A). The pressure was induced by the slow forward motion of the reciprocating pumps. The pressure curve was visualized and recorded via HD700 Data Acquisition software (Teledyne FLIR LLC, USA) during and after pressurization. The pressure curves were closely monitored in real-time and post hoc for signs of sudden depressurization or pressure loss due to a leak or chamber rupture. Overall, 1350 maximum pressure cycles were recorded (N = 10, 135 individual chambers).
In an alternative configuration of the setup, the absolute maximum amplitude induced by the AIMS at a 0.3 ml/min output volume rate was investigated (S. Figure 1B). A flow sensor was placed immediately after the stopcock. This experiment was also repeated with check valves to investigate the dampening effects of the valves. A similar setup was utilized to induce high amplitude (60 BPM, 0.34 ml/min bulk flow rate) pulsatile flow and to investigate the fluidic consistency of the pumps and the chambers. The pulsation rate (60 \(\frac{Beats}{min}\)) and bulk flow rate (\(0.34 \frac{mL}{min}\)) were selected for their physiologic relevance to CSF production, while the peak amplitudes (50 \(\frac{mL}{min}\)) were chosen to maximize system volatility and to “stress test” the setup and the chambers. A baseline of the pulsatile output was obtained by recording 10 separate 1-minute measurements of the pump output. The peak amplitude of ten consecutive peaks was analyzed per recording (N = 10, 10 measurements).
All high-throughput chambers, Resin chambers (N = 40), silicone chambers (N = 40), and PETG chambers (N = 50) were manufactured in batches of 10 chambers and were analyzed with their respective batches as independent product lots. The peak amplitudes of batches were statistically compared across chamber types. The fluidic consistency of five individual Dome chambers was also investigated (N = 5). Overall, 1,450 peak amplitude data points were recorded with the flow sensors (S. Figure 1C). As previously reported, check-valves were essential to the automatic refilling functionality of AIMS. Therefore, check valves replaced the stopcocks for the remainder of the experiments.
Long-Term Fluidic Performance
The consistency of the AIMS setup’s volumetric output was evaluated from 0 to 30 days. The volumetric analysis of bulk output was performed by weighing the output at 0, 15, and 30-day intervals using analytic balance Mettler Toledo AT261 DeltaRange Analytical Balance (Mettler-Toledo, LLC, USA). AIMS was programmed to run a pulsatile 70 BPM, 0.3 ml/min profile uninterruptedly for the 30-day duration. A total of eight ten-minute weight measurements were collected across 15 pump channels per volume profile (n = 15) at each time point yielding 360 ten-minute measurements. The output of each channel was collected and weighed in individual beakers (S. Figure 1D). The pump output flow pattern was also recorded at each time point.
Flow Simulation and Compliance
Compliance (C) was defined as the change in volume (V) relative to the change in pressure (P). As previously reported, the compliance of inline fluidic components such as the chambers (\({C}_{Bioreactor Chambers})\) had an impact on the amplitude and the waveform of the individual pulsations. A known volume of air was added to the pumps to augment the compliance (\({C}_{Augmented})\) in the circuits to better recapitulate the flow waveform of clinical measurements. This air volume was entrapped within the reciprocating portion of the pumps that were specifically accounted for in the program, thereby influencing the compliance of the system without interacting with the chambers throughout all experiments. The combined impact of augmented compliance and the inherent compliance of inline fluidic components (\({C}_{Total})\) on peak amplitude and pulsatile flow profile was investigated; to this end, 10 consecutive pulsations were recorded at varying amounts of augmented compliance (56.25 µL, 112.5 µL, and 225 µL air volume). Additionally, the flow profile of the setups with augmented compliance was compared to circuits with no augmented compliance (0 µL).
$${C}_{Total}\cong {C}_{Bioreactor Chambers}+{C}_{Augmented}$$
In a subset of samples, the CSF outflow from a pediatric patient’s external ventricular drain (EVD) was referenced and then simulated via AIMS with all four chamber types independently18. An overlay of these data was used to compare the pulsations with total compliance adjusted. Pre-operation (pathophysiologic) and post-operation (physiologic) flow patterns were simulated with augmented compliance and without augmented compliance in the Fig. 1A setup. The volume of air was iteratively adjusted for each chamber type to replicate the clinical flow pattern measurements as closely as possible.
In a subset of samples, an adjustable valve was added to the setup to simulate the measurements of a noninvasive wireless monitoring device that measured CSF flow through shunt systems in patients (S. Figure 1E)19. Additionally, a custom flow profile was produced to replicate the in-vivo measurements with the pumps directly. In this setup, the check valves were replaced with a stopcock, and the shunt valve was removed. The in-vivo measurements were replicated with the pumps directly using a custom profile; effectively replicating the combined impact of physiologic input and the mechanical action of shunt valves. This unique capability leveraged the setup’s programmability to produce complex, non-heterogenous flow patterns. This feature was also utilized to simulate clinical measurement of aqueduct flow patterns in a healthy volunteer and a hydrocephalic patient20.
Finally, a Watson Marlow 323 DU peristaltic pump (Wilmington, USA) equipped with a 314MC attachment and 0.38mm bore/1.02mm bore peristaltic tubing was utilized to investigate the compatibility of the bioreactor chambers with alternative sources of flow induction such as those from peristaltic pumps (S. Figure 1F). This setup was used to simulate the EVD pre-recovery and post-recovery flow patterns simulations with all types of bioreactor chambers and was compared to the same output from AIMS.
Biocompatibility
The biocompatibility of chamber materials was considered by quantifying the growth of human astrocytes on each material type over five days. To assess the biocompatibility of resin, silicone, and PETG, a puck with a surface area equivalent to 50% of a 24-well plate bottom was secured to the bottom of the respective wells using a silicone-based adhesive. In total, eight wells within a 96-well microplate were prepared for each experimental condition: 1) negative control (formaldehyde), 2) neutral (uncoated well-plate), 3) Resin, 4) silicone, 5) PETG, 6) Silicone-based glue.
Initial cell count was made using a hemocytometer and a trypan blue exclusion assay. A media stock solution was prepared by combining astrocyte media (ScienCell, USA), supplement kit, 10 mL fetal bovine serum (FBS), 5 mL penicillin/streptomycin, and 5 mL astrocyte growth serum, (ScienCell, USA), and an additional 10 mL FBS (ScienCell, USA). After the 5 days, a 2.5mg/mL aliquot of DiD was diluted to a 20 µg/mL solution using astrocyte media, and a 5mM stock solution of pluronic F127 was diluted to a 100 µM solution using stock media. After removing existing media from wells, the wells were rinsed with 50uL of Hank’s Balanced Salt Solution (HBSS). Then, a 20uL of DiD and 20uL pluronic F127 solution were added to each well. The samples were incubated for 2 hours. This process was repeated three times. The wellplates were imaged on a Leica TCS SP8 confocal laser scan microscope using a Leica 10× magnification dry lens.
Imaging and Optical Output
Fluorescent particles, specifically Green Fluorescent particles (G0500, Thermo Fisher Scientific, USA), characterized by a diameter of 5 µm and a density of 1.06 \(\frac{g}{mL}\), were introduced into a 3 mL saline solution. The flow rate was set to post-operation clinical measurement of CSF flow in a pediatric patient 18. These particles were subsequently directed into a chamber for observation using Resonance scanning confocal microscopy (RS-G4), installed on an upright microscope platform (Caliber ID, Andover, MA, USA). The imaging protocol utilized a wavelength of 488 nm to ensure stable signal acquisition from the fluorescent microspheres as they traversed the flow field proximal to the drainage holes. Single z-stack frames from a 5-second recording of microsphere movement near each lateral hole were exported. The pictures were sequenced and overlapped to produce long exposure views of microsphere motion near the lateral holes of unused and patient-explanted catheters from our biorepository.
Six digital microscopes (Plugable Technologies, USA) were utilized to construct an imaging setup to produce brightfield microscopic and macroscopic time-lapses of the PETG chambers in a high throughput manner. The setup also featured a custom-built chassis that held up to 24 chambers in total, and a linear actuation mechanism that interfaced with the chassis and propelled the chambers forward such that the next row of catheter holes could be imaged. This process continued until all 24 chambers were imaged. An ultraviolet (UV) light source and fluorescent dye were utilized to record a macroscopic view of the fluorescent dye entering through the inlet of the PETG chamber and leaving through the ventricular catheter.
Statistical Analysis
Statistical analyses were performed using SPSS Statistics Suite (IBM Corporation, USA). The Levene Test of Homogeneity of Variances was performed to evaluate the homoscedasticity of the datasets. Parametric data were analyzed using One-way analysis of variance (ANOVA) or repeated measures ANOVA tests. A linear regression analysis was performed as a simple predictive analysis to investigate the correlation between variables and the coefficient of determination (\({R}^{2}\)) was calculated. For all tests, a confidence interval was set at 0.95 (α = 0.05) A post hoc Tukey test was performed when the null hypothesis (no difference in the group means) was rejected.