Mathematical modeling of cell seeding in xenograft blocks
A clinically established and FDA approved bone substitute material was used as a model for all the simulations and experiments (Bio-Oss® Block, Geistlich Pharma AG, Switzerland). According to the manufacturer, it is composed of deproteinized and sterilized cancellous bovine bone with a cuboid dimension of 10 mm × 10 mm × 20 mm, a bimodal pore structure with mean pore diameters of d1 = 17 nm and d2 = 1.07 × 105 nm and a porosity of 69%20. Xenograft blocks are physically and chemically comparable to the mineralized matrix of human bone rendering by osteoconductive properties and suitable for bone regeneration.
The 3D scaffold geometry of Bio-Oss blocks was obtained by means of micro-computed tomography (Skyscan 1172, Bruker microCT, Belgium) at a resolution of 9.5 μm. Computational fluid dynamics (CFD) analysis (ANSYS Fluent, ANSYS Inc., Canonsburg, PA, USA) of fluid flow in xenograft blocks were conducted as already described20. Briefly, for the numerical simulations steady-state flow conditions, an inlet flow rate of 1 ml/min, an outlet pressure of 0 Pa and no-slip-conditions on the walls were applied. Based on experimental investigations at 37 °C using a Haake RheoStress 1 rotational rheometer equipped with a double cone setup (Thermo Fisher Scientific Inc., Newington, NH, USA), dynamic viscosity and density of the cell suspension (ρ = 1 × 106 cells/ml) were set to 1.176 mPas, and 1004 kg/m3 respectively. In the numerical simulations, the necessity of a chamber with a form-locking fit with respect to the scaffold block was postulated. Connector geometries for the inflow and outflow of the perfusate were simulated and compared with regard to the homogeneity of flow within the scaffold. Firstly, the influence of different diameters (D1 = 1 mm, D2 = 10 mm) of the cylindrical connector geometry was investigated (Fig. 1A). Secondly, cylindrical and conical connector geometries were simulated with in- and outflow lengths ranging from L = 1 mm to L = 10 mm, respectively (Fig. 1A, B). To quantitatively compare the homogeneity of the flow through the different bioreactor designs, the Hoover coefficient was calculated at 24 cross-sectional planes with a constant spacing of 1 mm (Fig. 1A, B).
Schematic representation of numerically investigated cylindrical (A) and conical (B) connector geometries. For the cylindrical geometry, the influence of different connector diameters (D1 = 1 mm, D2 = 10 mm) as well as lengths (L1 = 1 mm, L2 = 2 mm, L3 = 3 mm, L4 = 4 mm, L5 = 5 mm, L6 = 10 mm) was investigated. For the conical connector geometry, the influence of different connector lengths (L1, L2, L3, L4, L5, L6) with a constant connector diameter (D2) was investigated. The Hoover coefficient was calculated at 24 cross-sectional planes perpendicularly aligned to the longitudinal axis of the scaffold.
Design and manufacture of bioreactors
The design of the perfusion chambers was developed using Autocad Fusion 360 (Autodesk, Inc., CA, USA). Perfusion chambers were printed with a Formlabs Form2 3D printer (Formlabs Inc., MA, USA) using the Dental SG 1 L resin (class 1 biocompatible synthetic resin (EN-ISO 10,993–1:2009/AC:2010, USP class 4) by stereolithography. A layer thickness of 25 µm was used. The dimensions of the perfusion chambers are 1 cm in width, 2 cm in height and 1 cm in depth, which guaranteed that the Bio-Oss blocks perfectly matched to the chambers. Type 3 of the chambers has the same dimensions, was only integrated horizontally into the perfusion module. The chamber closing mechanism was ensured by a conventional rubber seal on the module cover, which was firmly closed with the perfusion module using 4 screws per perfusion chamber (Fig. 2C). The biocompatible silicone tubing material for flow applications was purchased by ibidi GmbH, Gräfelfing, Germany (#10841). All perfusion related materials were preheated at 37 °C in 5% CO2 and 95% air-humidified incubator at least for 24 h.
Bioreactor prototypes. (A) Schematic presentation of the different bioreactor types (longitudinal section). Black: shape of the perfusion chamber and the inflow/outflow nozzles. Blue: scaffold block. Red: silicone encasement around the scaffold. The perfusion chambers of type 1 and type 2 are 1 cm in width and 2 cm in height, type 3 is 2 cm in width and 1 cm in height. The presence or absence of the encasement is indicated by enc+ or enc−, respectively. (B) Design of the four bioreactors integrated in a single 3D-printed block drawn by Autocad Fusion 360. The different inlet and outlet geometries are visible above and below the scaffold chambers. Arrows indicate separate inflow and outflow lines of the chambers. Central connectors are needed for the water circulation system to regulate the temperature. Bioreactor types 2 and 3 with encasement harbor a silicone sealing (red) for the scaffolds (blue). (C) 3D view of the bioreactor without the silicon encasement.
Based on the results of the numerical simulations, different connector geometries were applied using different types of lateral encasement and orientation of the scaffold (Fig. 2A). The following connector dimensions for inflow and outflow were used: (i) 1 mm cylindrical shaped connector geometry (type 1) as well as (ii) conical shaped connector geometry with 15 mm in height x 15 mm in width (type 2) and 15 mm in height × 25 mm in width (type 3). The conically shaped connectors (type 2, 3) are firmly attached to the scaffold housing, so that the cell inflow is made possible over the entire upper scaffold surface. On the other hand, the cylindrical connector only allows punctual inflow through the 1 mm wide opening.
In the experimental investigations of type 1 and 2 connector geometries, the scaffold was perfused lengthwise (along the longitudinal axis). In contrast, the scaffold of type 3 connector geometry was transversally perfused to its longitudinal axis. Whereas the perfusion chamber itself was initially designed loose-fit to the scaffold block (type1/2_enc−), it was equipped with a press-fit encasement to the scaffold block (type1/2/3_enc+) in later experiments manufactured from dental grade silicone (Wagnersil 9N, Wagner Dental, Hückelhoven, Germany) with a Shore Hardness of 9 to 35 Shore. It was individually prepared to fit inside the perfusion chambers and tightly enclose the scaffold block, allowing the direct perfusion through the scaffold and preventing a flow over the lateral surfaces. For this purpose, a template of the scaffold was created by 3D printing to cast the silicone encasement around the scaffold. After curing and sterilization, the scaffold was placed directly into the silicone encasement.
For the micro PET/CT experiments, four perfusion chambers were integrated in a single 3D-printed bioreactor block in order to perform multiple experiments and measurements simultaneously in a single PET/CT session (Fig. 2B). Perfusion was realized by negative pressure from the outflow side by NE-4000 double syringe pump (modified for up to four syringes; New Era Pump Systems Inc., East Farmingdale, NY, USA) or Ibidi Pump System (ibidi GmbH, Gräfelfing, Germany). The cell suspension was gently stirred in a 250 ml flask and was routed via individual feed lines to the inflow side of each perfusion chamber. Each perfusion chamber in the bioreactor block was separately controlled by one pump/syringe. To maintain constant temperatures in the micro PET/CT experiments, a temperature-control system based on a water circulation system operated by Medres heated pump and equipped with a Carel Ir33 temperature controller (Carel Industries S.p.A., Brugine, Padova, Italy) was implemented. As oxygen sensor, the PreSens O2 Flow-Through Cell FTC-PSt3 in conjunction with the PreSens Oxy 10 (PreSens Precision Sensing GmbH, Regensburg, Germany) was used to monitor oxygen content within the bioreactor. One sensor each was placed at the inlet and outlet of the connectors. A representation of the bioreactor setup in combination with the perfusion pump is shown in Supplementary Fig. 1.
Isolation, propagation and characterization of adipose-derived stem cells (ASC)
Lipoaspirate samples were collected from patients undergoing liposuction or lipofilling procedures at the Rostock University Hospital or Plastic Surgery Clinic in Rostock, Germany, with approval from the Ethics Committee at the University Medical Center Rostock No. A 2014–0092. All patients provided informed consent. The procedure for isolating adipose-derived stem cells (ASC) has been described elsewhere21. Briefly, the samples were washed with PBS, and transferred in a 50 ml Falcon Tube containing 10 ml PBS with 6 mg/ml collagenase NB4 (SERVA Electrophoresis GmbH, Heidelberg, Germany). After 30 min tissue digestion using a Lab Rotator at 37 °C, the digestate was filtered through a 100 μm cell strainer (Becton Dickinson, Franklin Lakes, NJ, USA) by adding 10 ml PBS including 10% Hyclone Newborn Calf Serum (NCS; Sigma Aldrich Chemie GmbH, Munich, Germany) and was finally centrifuged at 1000 rpm for 10 min. The pellet was washed in 10 ml PBS/10% NCS, centrifuged again and resuspended in 10 ml PBS/10% NCS. Adipose-derived stromal cells were kept at 37 °C in 5% CO2 and 95% air-humidified incubator. Cell culture medium constituted of 45% Iscove’s Modified Dulbecco’s Medium, 45% Gibco® F-12 Nutrient Mixture, 10% NCS and supplemented with 0.1 ml Gibco™ Penicillin–Streptomycin (all from Thermo Fisher Scientific GmbH, Regensburg, Germany) and 10 µg basic fibroblast growth factor (recombinant human bFGF; Millipore Merck KGaA, Darmstadt, Germany) per 1 l of cell culture media. Once 80% confluency was reached, flasks were passaged in a 1:3 ratio. Adherent ASCs were assessed for mycoplasma contamination using a PCR-based assay and DAPI staining (Life Technologies GmbH, Darmstadt, Germany). The ASCs were used up to the 3rd passage. The ability for adipogenic, osteogenic, and chondrogenic differentiation of these ASCs as evidence of mesenchymal stem cell character has been described elsewhere22,23. Moreover, expression of CD73, CD90, and CD105 was proven by flow cytometry using BD FACSAria™ IIIu (BD Biosciences, San Jose, CA, USA). Therefore, all primary antibodies (HLA-DR, CD105, CD90, CD45, CD73, CD34, CD19, CD14) were purchased from BD Biosciences and results were calculated using FlowJo (https://www.flowjo.com).
Comparison of cell concentrations and perfusion rates
Prior to micro-PET/CT experiments, a first series of experiments with cell imaging by confocal laser-scanning microscopy (LSM) was performed in order to test optimal seeding parameters like cell concentration and perfusion rate using the bioreactor type2_enc+ . Prior to perfusion, stem cells were labeled with the PKH26 Red Fluorescent Cell Linker Kit (Sigma-Aldrich Chemie GmbH) following manufacturer’s instructions. Either the unidirectional or the oscillating perfusion mode was selected. Unidirectional means that the cells were only perfused through the scaffold from top to bottom. In the oscillating mode, the orientation was switched every 5th minute. The perfusion chambers including scaffold blocks were perfused with warmed cell suspension (37 °C) with a priming volume of 10 ml over 60 min and washed with warmed PBS (w/o Ca2+ and Mg2+; Biochrom GmbH, Berlin, Germany) thereafter. First, perfusion was performed with cell concentrations between 0.06 and 1.22 × 106 cells/ml at a constant perfusion rate of 0.5 ml/min in order to check the influence of the cell concentration on the seeding efficiency and homogeneity (10 steps). In a second series, the influence of the perfusion rate was tested by using 0.25, 0.5, 1.0, and 2.0 ml/min at a constant cell concentration of 0.4 × 106 cells/ml (n = 3). After retrieving the scaffolds, bioreactors and tubes were washed with 100% alcohol (cleaning solution). Number of residual cells was measured in the outflow, the wash out, and the cleaning solution. Results were related to the inflow concentration to calculate the seeding efficiency.
LSM imaging and 3D image reconstruction
Confocal laser-scanning microscopy (LSM) was used to analyze the three-dimensional (3D) spatial distribution of cells within the scaffold blocks to check calculated seeding efficiency. After cell perfusion, scaffolds were fixed in 4% paraformaldehyde (Santa Cruz, Dallas, USA), washed with PBS, and cut in into 1.5 mm slices using an Exakt Diamant Bandsäge 300 CL saw (EXAKT Advanced Technologies GmbH, Norderstedt, Germany). PKH26-labeled cells were investigated with an inverted confocal laser-scanning microscope (LSM780, Carl Zeiss Microscopy GmbH, Jena, Germany). The following fixed settings were used for scanning: 40 µm, 3.2 AU, Alexa Fluor555, resolution 1024 × 1024 Pixel per tile, 10 × magnification Tiles (20 × 20). Overlay images of z-scans were assembled by scanning the cells on the surface at 0.8 µm intervals (software ZEN black, Carl Zeiss Microscopy GmbH). 3D reconstruction of the scaffold blocks was generated with the software Fiji ImageJ (https://imagej.net/Fiji/), a self-written program and visualized with VisIt (https://visit.llnl.gov/) (Supplemental Fig. 2). The distribution of the fluorescence within the 3D reconstructed blocks served as a qualitative measure for the homogeneity of the seeding.
[18F]FDG micro-PET/CT imaging
[18F]FDG-labeling and micro-PET/CT examination was used in the main series of experiments for the real-time imaging of cell distribution in the scaffold blocks. Prior to this, [18F]FDG uptake of ASC was tested up to 180 min after administration of [18F]FDG. 2 × 105 ASCs of the 3rd passage were trypsinized, centrifuged, and resuspended in 1 ml cell culture media. [18F]FDG has been synthesized fully automated at the GMP laboratory of the Department of Nuclear Medicine using a Fastlab module in combination with corresponding FDG-citrate cassettes (GE Healthcare). The required [18F]fluorine has been obtained from the in-house cyclotron MiniTrace PT 700 (GE Healthcare). Cells were incubated with 1 MBq [18F]FDG at 37 ℃ under continuous stirring. After 10, 30, 60, 90, 120 and 180 min, cells were washed and trypsinized followed by repeated washing steps with PBS. Residual activity of the cells was measured using a gamma counter (WIZARD2 10 detector Gamma Counter; Perkin Elmer, Waltham, MA, USA). The measurements were repeatedly performed (n = 6). Based on these results (Supplementary Fig. 3), an incubation time of 60 min was considered to be adequate for [18F]FDG-labeling, because no significant differences in [18F]FDG-uptake were calculated after 60 min, and used in the following perfusion experiments. [18F]FDG-labeled cells were washed twice with PBS and diluted to a concentration of 0.4 × 106 cells/ml. Perfusion of scaffold blocks in the bioreactors was conducted as described above and imaging studies were performed using a preclinical micro-PET/CT system (Inveon®, Siemens Healthcare Erlangen, Germany). A different number of experiments were performed depending on the bioreactor type: loose-fit perfusion chambers without silicone encasement (bioreactor type 1 (n = 1) and bioreactor type 2 (n = 2)), bioreactors with silicon encasement (bioreactor type 1 to 3 (n = 3 each) and bioreactor type 3 with oscillating perfusion direction (n = 3). The total PET scan time was approx. 100 min including unidirectional perfusion with [18F]FDG-labeled stem cells in the first 60 min. For further 40 min, seeded scaffolds were observed during a wash-out-phase with PBS to clean the bioreactor of remaining free FDG and non-adherent cells. In case of bioreactor type 2 with encasement, an additional oscillating mode was tested with changing the perfusion direction once after 30 min (type2_enc+ _osc).
Each PET data set was corrected for random coincidences, dead time and attenuation. PET images with dynamic framing every 5 min were reconstructed using a 3D Ordered Subset Expectation Maximization with Maximum A Priori Shifted Poisson algorithm (3DOSEM / SP MAP). For quantitative results, regions of interest (ROI) were placed around the whole scaffold block, in the upper, middle and lower third as well as in the outflow tract to keep track of cell sedimentation outside the scaffold. Data were averaged and cell distribution was compared regarding bioreactor types and perfusion algorithms. The Hoover coefficients (0 representing total equal distribution and 1 representing maximal inequality) were calculated for every bioreactor type as a measure of the inhomogeneity of cell distribution within the scaffold block. In order to validate calculated Hoover indices using [18F]FDG-labeled cells, results were compared to 3D reconstructions from LSM evaluation using PKH26-labeled cells. Both experiments were performed with a bioreactor type2_enc+ . In addition, cell distributions within the scaffold blocks were compared to the in silico calculations generated from numerical simulations.
Cell adherence in scanning electron microscopy (SEM)
To investigate cell adherence after perfusion, scaffold blocks were incubated in culture medium for 3, 6 or 24 h after the one-hour perfusion (t0) with 0.4 × 106 cells/ml in the bioreactor type 2 with encasement. Scaffold blocks were washed with PBS, fixed with 2.5% glutaraldehyde in 0.05 M HEPES buffer, dehydrated in an ascending ethanol series and critical point dried using CO2 as an intermedium with the EMITECH 850 critical point dryer (Emitech Ltd., Ashford, UK). Carbon coating was done with the carbon coater SCD500 (Leica Microsystems GmbH, Wetzlar, Germany). Images were taken with a QUANTA FEG 250 field emission scanning electron microscope (FEI Company, Hillsboro, OR, USA) using an acceleration voltage of 5–10 kV and a sample chamber pressure of p < 1.5 × 10–2 Pa (‘high vacuum’ mode) for imaging.
Long-term perfusion and initial osteogenic differentiation
For the long-term experiments, Bio-Oss blocks were perfused in bioreactor type 2_enc+ with 0.4 × 106 cells/ml for 60 min using the Ibidi pump system (ibidi GmbH). An oscillating perfusion rate of 0.5 ml/min was chosen. After perfusion, scaffolds were incubated in a roller culture flask for 14 days and the cell culture medium was changed every three days. All experiments were performed in triplicates. The initial osteogenic differentiation was analyzed by Alizarin red staining according to manufacturer’s protocol (Osteogenesis assay kit, #ECM815, Millipore, Burlington, MA, USA). Briefly, cells were washed with PBS, fixed with 4% paraformaldehyde for 10 min, washed again, and incubated with 0.1% Alizarin red for 30 min. Finally, cells were washed until red stained structures were clearly visible. Visualization and imaging were carried out with the Axioscope A1 microscope using AxioVision Imaging Software 4.8.2.0 (both from Carl Zeiss Microscopy GmbH).
Immunohistochemistry (IHC)
After cell perfusion, scaffold blocks were fixed in 4% buffered formalin (Formafix®, Global Technologies Ltd., Düsseldorf, Germany) further dehydrated and infiltrated with embedding resin (Technovit 9100, Kulzer, Wehrheim, Germany), cut to 200 µm slices, grind to 35–50 μm with sandpaper discs (1200 grit, EXAKT micro-grinding system), and final-polished with 4000 grit polish paper (EXAKT Advanced Technologies GmbH, Hamburg, Germany). These slices were stained with primary antibodies against collagen I (#ab34710), osteopontin (#ab8448), and bone sialoprotein (#ab52128; all from Abcam, Cambridge, UK). Secondary labeling was performed using an Alexa Fluor 488 goat anti-rabbit IgG (Thermo Fisher Scientific) combined with Hoechst counterstaining (PanReacAppliChem, Darmstadt, Germany). Stainings were investigated using the LSM780. Notably, images were taken at identical device settings to guarantee comparable results. The image processing was carried out using ZEN 2011 and 3D picture reconstruction was realized by taken z-scans with increments of 0.5 µm using ZEN black (both from Carl Zeiss Microscopy GmbH). A part of the sections of each block were treated with 10% hydrogen peroxide for 10 min after embedding in Technovit 7200 instead of 9100. Sections were stained with Giemsa’s azur eosin methylene blue solution (#109204; Merck KGaA, Darmstadt, Germany) for 30 min and toluidine blue-methylene blue-pyronin solution (#12796; Morphisto, Frankfurt (Main), Germany) for 10 min. Slides were scanned with an Axio Imager.M2 microscope equipped with an AxioCam MRc5 camera (both from Carl Zeiss Microscopy GmbH) that was connected to an automatic scanning table (M-686K011, Wienecke & Sinske GmbH, Gleichen, Germany).
Statistical analysis
Statistical analyses were performed with GraphPad Prism 5 software (GraphPad Software, San Diego, CA, USA). Regarding cell enrichment and cell distribution, differences in [18F]FDG activity and Hoover coefficient were assessed using student’s t-test, respectively, and considered statistically significant at P-values of *P < 0.05, **P < 0.01, ***P < 0.001. In case of [18F]FDG uptake experiments, normal data distribution was tested by Kolmogorov–Smirnov test followed by Kruskal–Wallis one-way analysis of variance and Mann–Whitney U-test as post-hoc test. According to not normally distributed data, graph displays box-and-whisker diagram. Boxes include 25th and 95th percentiles as well as the median. Whiskers represent 10th and 90th percentiles. Pearson correlation coefficient (R2) was calculated to test linear correlation between seeding efficiency and cell concentration/perfusion rate. A correlation coefficient of -1 demonstrates a total negative linear correlation, whereby 0 represents no linear correlation.
