Computational fluid dynamic analysis reveals the underlying physical forces playing a role in 3D multiplex brain organoid cultures

Organoid cultivation in suspension culture requires agitation at low shear stress to allow for nutrient diffusion, which preserves tissue structure. Multiplex systems for organoid cultivation have been proposed, but whether they meet similar shear stress parameters as the regularly used spinner flask and its correlation with the successful generation of brain organoids, has not been determined. Herein, we used computational fluid dynamics (CFD) analysis to compare two multiplex culture conditions: steering plates on an orbital shaker and the use of a previously described bioreactor. The bioreactor had low speed and high shear stress regions that may affect cell aggregate growth, depending on volume, whereas the CFD parameters of the steering plates were closest to the parameters of the spinning flask. Our protocol improves the initial steps of the standard brain organoid formation, and organoids produced therefrom displayed regionalized brain structures, including retinal pigmented cells. Overall, we conclude that suspension culture on orbital steering plates is a cost-effective practical alternative to previously described platforms for the cultivation of brain organoids for research and multiplex testing. Highlights Improvements to organoid preparation protocol Multiplex suspension culture protocol successfully generate brain organoids Computational fluid dynamics (CFD) reveals emerging properties of suspension cultures CFD of steering plates is equivalent to that of spinner flask cultures


Abstract 25
Organoid cultivation in suspension culture requires agitation at low shear stress to 26 allow for nutrient diffusion, which preserves tissue structure. Multiplex systems for 27 organoid cultivation have been proposed, but whether they meet similar shear stress 28 parameters as the regularly used spinner flask and its correlation with the successful 29 generation of brain organoids, has not been determined. Herein, we used 30 computational fluid dynamics (CFD) analysis to compare two multiplex culture 31 conditions: steering plates on an orbital shaker and the use of a previously described 32 bioreactor. The bioreactor had low speed and high shear stress regions that may affect 33 cell aggregate growth, depending on volume, whereas the CFD parameters of the 34 steering plates were closest to the parameters of the spinning flask. Our protocol 35 improves the initial steps of the standard brain organoid formation, and organoids 36 produced therefrom displayed regionalized brain structures, including retinal pigmented 37 cells. Overall, we conclude that suspension culture on orbital steering plates is a cost-38 effective practical alternative to previously described platforms for the cultivation of 39 brain organoids for research and multiplex testing. 40

Introduction 51
Three-dimensional (3D) cerebral organoids generated from human pluripotent stem 52 cells (hPSCs) are complex structures that partly reproduce fetal brain development in 53 vitro, making them powerful tools for the study of human development and disease 54 (Lancaster and Knoblich, 2014a). The self-organization that occurs during hPSC 55 differentiation in cerebral organoids allows for the appearance of complex structures, 56 including those recapitulating regions of the cerebral cortex, ventral forebrain, midbrain, 57 hindbrain, hippocampus, and retina (Eiraku et al., 2011;Monzel et al., 2017;Quadrato 58 et al., 2017). Several research groups have used this model to study the development 59 of diseases such as microcephaly, lissencephaly (Bershteyn et al., 2017;Lancaster et 60 al., 2013), and Zika infection (Garcez et al., 2016), as well as for drug testing (Dakic et 61 al., 2017). 62 Organoids are grown in 3D suspension culture, which enables efficient nutrient delivery 63 to 3D organized tissue. Historically, cerebral organoids have been cultured in spinner 64 flasks (Lancaster and Knoblich, 2014b). These flasks have the advantage of providing 65 a low-shear environment (Wang et al., 2013), which is important because hPSCs have 66 been shown to be sensitive to shear stress (Nampe et al., 2017;Wang et al., 2013). 67 However, spinner flasks have the disadvantage of requiring a high volume of cell 68 culture media for cultivation, increasing the costs of experiments thus being limiting to 69 drug testing and other multiplex experiments including comparison of multiple patients 70 and controls. Recently, Qian et al. (2016) (Qian et al., 2016) proposed the use of a 3D-71 printed scalable mini-bioreactor, the SpinΩ, which would be cost effective and provide 72 a feasible, reproducible platform for chemical compound testing. However, cultivation in 73 the SpinΩ still requires an initial investment and the availability of 3D-printing 74 equipment and other materials, which might make it infeasible for most laboratories. 75 The use of orbital shaker plates described originally (Lancaster and Knoblich, 2014b) is 76 a multiplex alternative to the often cost-prohibitive use of spinner flasks. However, 77 The medium was changed weekly until day 60 of culture. They were imaged with an 134 EVOS cell imaging system (Thermo Fisher Scientific) in brightfield. The area, diameter, 135 and circularity of individual cerebral organoids were quantified using a custom macro in 136 ImageJ. 137

2.3.
Computational fluid dynamics simulation 138 CFD simulations were performed for the flows imposed by the SpinΩ impeller and the 139 orbital shaker using the finite element commercial code COMSOL Multiphysics ® . The 140 rotational speed used for the SpinΩ was 60 rpm and that used for the orbital shaker 141 was 90 rpm, with consideration of a 9.5-mm radius. The geometry and finite element 142 meshes used in the simulations involved 3 ml suspension culture media for both mixing 143 techniques (Supp. Fig. 2). The finite element meshes used for the bioreactor and 6 well 144 plates simulations contained 390,000 and 125,000 elements, respectively. The 145 numbers of finite elements and time steps used in the simulations were selected after 146 grid refinement analyses; the differences between the velocity components computed 147 with the meshes used in this work and with less refined meshes were less than 3% at 148 Newtonian behavior. The flows were considered to be isothermal, after experimental 159 evidence demonstrated that the work imposed by the impeller in the bioreactor was 160 negligible (temperature variations over 24 h were less than 0.2°C; data not shown). 161 Analysis of the orbital shaker required that transient states were simulated until a 162 quasi-steady-state regime was reached, when the flow became periodic. Liquid flow 163 was analyzed 0.5, 14, and 15 s after the start of movement 164

Proteomic analysis 180
Two independent pools of four organoids each were used in the experiments. Protein 181 digestion, peptide fractionation, mass spectrometric analysis, and raw data processing 182

Statistical analysis 186
Statistical testing was performed using two-tailed t-test with GraphPad Prism 6 187 software. Statistical significance was defined as p<0.05 unless otherwise stated in 188 figure legends. Correlation analysis was done comparing the R square of a non-linear 189 fit (Exponential fit in Fig. 1c and a Gaussian fit in Fig. 1e) for the two conditions, SpinΩ 190 and orbital shaker. (data not shown). Therefore, the cells were passaged manually before the EB 202 formation step. Immediately after treatment for cell dissociation and before 203 centrifugation, 10 µM ROCKi was added to the trituration solution. This step improved 204 cell morphology after dissociation (Supp. Fig. 1a and b) (Horiguchi et al., 2014). The 205 centrifugation step significantly improved the circularity of organoids on day 1 of growth 206 (Supp. Fig. 1e), which was correlated with a significant increase in the observed areas 207 of organoids in the two conditions (Supp. Fig. 1f). However, after 10 days, no 208 significant difference was seen between specimens treated with and without 209 centrifugation (Supp. Fig. 1f), suggesting that this potentially negative effect was 210 temporary. During the EB stage, no significant growth was observed (Fig.1a, b) and the 211 morphology of aggregates did not change (Fig. 1b). Growth during the neuroinduction 212 stage also was not significant (Fig. 1a, b). In the neuroinduction stage, protrusions of 213 developing organoids started to expand; these continued to grow over time (Fig. 1b)  214 and formed neuroepithelium-like tissue (see also Fig. 3b). The pattern of organoid 215 growth resembled an exponential curve (Fig. 1c), with R square values of 0.9811 for 216 the SpinΩ and 0.8587 for the orbital shaker. Growth in the orbital shaker was initially 217 more rapid than that in the SpinΩ (Fig. 1d), but no significant difference was observed 218 at the 30-day timepoint. A histographic analysis of organoids grown in the SpinΩ and 219 the orbital shaker at 30 days showed that the size distribution of organoids grown on 220 the shaker more closely resembled a Gaussian fit (Fig. 1e), suggesting more 221 homogeneity in the shaker.  The SpinΩ analysis is presented for different plane cuts. The velocity and shear stress 240 fields are presented in Figure 2b. The highest velocities occurred at the edge of the 241 impeller, with values around 0.05 m/s (Fig. 2a). The velocities decreased with distance 242 from the impeller and rotating shaft, being null at the well walls due to the non-slip 243 conditions. In particular, lower velocities at the bottom of the well did not favor the 244 mixture required for the enhanced growth of organoids. Our attempts to form 245 aggregates from single cells in the SpinΩ created bodies with disparate sizes. Single 246 cells accumulated in the low-speed area of the bioreactor and formed a large 247 aggregate, while adjacent cells formed smaller bodies (Supp. Fig. 4). Large differences 248 in the velocity fields implied differences in nutrient mixing, which could in part explain 249 the delayed growth in the SpinΩ on days 17-19 (Fig. 1d) and the wide area distribution 250 shown on the histogram at day 30 for SpinΩ (Fig. 1e). 251 The maximum shear stress was 0.56 Pa at the edge of the impeller due to the large 252 difference between velocity gradients in this region (Fig. 2b). Shear stress in the bulk 253 fluid was lesser, with magnitudes on the order of 10 -3 to 10 -2 Pa. 254 Velocity and shear stress are correlated parameters. In this study, however, the 255 absolute velocity magnitudes were greater for the orbital shaker than for the SpinΩ, The gold standard for organoid protocols, a spinning bioreactor has been reported to 260 sustain organoid growth in culture for more than 8 months (Lancaster and Knoblich, 261 2014b;Quadrato et al., 2017). Comparison with previous literature on the CFD of a 262 spinning bioreactor (Wang et al., 2013) showed that shear stresses of the steering 263  Table 2). The maximum shear stress of the SpinΩ (0.56 Pa) was one order of 268 magnitude greater. 269 The velocity fields of the steering plates (maximum, 0.12 m/s) were of the same order 270 of magnitude as those of the spinner bioreactor (maximum, 0.277 m/s), whereas those 271 of the SpinΩ were lower (0.05 m/s) (Supp. Table 2). 272 Computational simulations indicated that the use of the 6 well plates on the orbital 273 shaker was more suitable for the growth of organoids than was the use of the 274 bioreactor. First, the SpinΩ has regions of low velocity at the bottom of the well, where 275 fluid mixing is poor and particles deposition is likely to happen. Second, shear stress is 276 lesser in the orbital shaker, which could be better for the preservation of organoid 277 structures in long-term culture. 278 279

The SpinΩ reactor and orbital shaker derived structured organoids 280
We examined the maturation of organoids with a focus on the transition from 281 predominantly neuroprogenitor stem cells to the development of neuroepithelial 282 regions. Nestin staining of neural stem cells, performed at 10, 14, and 30 days of 283 culture, showed similar decreases over time for the orbital shaker and SpinΩ 284 treatments (Fig. 3a). These results are consistent with the start of differentiation of 285 progenitor cells into neurons. At 30 days, the organoids had developed ventricular-like 286 regions and neuroepithelium-like structures that were positive for MAP2 and TBR2 287 (Fig. 3b). MAP2 staining levels were similar in organoids cultivated in the orbital shaker 288 and SpinΩ, but TBR2 staining levels were significantly stronger for those cultivated in 289 the SpinΩ. As TBR2 labeled neuron progenitors in sub-ventricular zones, we examined 290 whether cell proliferation was increased under our culture conditions through phospho-291 histone-3 staining on day 30. However, no difference in the number of proliferating cells 292 was detected between the two conditions (Fig. 3c). 293

3.4.
Organoids generated in suspension cultures presented markers for 294 distinct brain regions 295 296 Organoids grown in the SpinΩ and orbital shaker displayed very similar morphology 297 and developmental profile. We decided to focus on organoids grown on the orbital 298 plates to provide further characterization of the organoid generation pipeline, because 299 the organoids grown in SpinΩ have been described elsewhere (Qian et al., 2016). We 300 observed that 30-day organoids from orbital shaker cultures were positive for FOXG-1 301 (forebrain), PAX-6 (dorsal telencephalon), OTX-2 (retinal cells and midbrain), and islet-302 1 (hindbrain; Fig. 4a) showing diversification and development consistent with previous 303 reports (Quadrato et al., 2017). We observed that, at 45 days, the organoids had 304 pigmented regions (Fig. 4b, c), which were previously described to reproduce the 305 formation of retinal pigmented epithelium (Quadrato et al., 2017). The pigmented 306 regions were positive for the retinal cell marker glycogen synthetase (GS) (Fig. 4c). 307 Electrophysiological recordings of 45-day organoids detected neural oscillations below 308 10 Hz in a pigmented organoid, but no firing was detected at the same developmental 309 age in non-pigmented organoids (Supp. Fig. 5). formation of the pigmented regions which were observed later, at 45 days (Fig. 4c). 320 321

Discussion 322
Brain organoids present cytoarchitecture that recapitulates brain tissue organization, 323 offering a complex in vitro model for the study of brain normal and pathological 324 development (Lancaster et al., 2013;Mariani et al., 2015;Paşca et al., 2015). Although 325 brain organoid cultivation presents challenges related to the lack of reproducibility and 326 scalability, we achieved high reproducibility of early-stage organoid size and growth by 327 adding steps to a standard protocol (Lancaster and Knoblich, 2014b). This included the 328 use of higher ROCKi concentrations and plate centrifugation in the EB formation step, 329 which have been previously demonstrated to improve EB formation, but have never 330 been applied to organoid formation protocol (Hookway et al., 2016;Horiguchi et al., 331 2014). 332 Scalability was achieved by applying two multiplex platforms: steering plates on an 333 orbital shaker and the SpinΩ bioreactor (Qian et al., 2016). The SpinΩ was 3D printed 334 according to the blueprints provided by Qian et al. (Qian et al., 2016). We encountered 335 the following issues with SpinΩ use: 1) manual handling, as medium changes involved 336 disassembly of a combination of pieces; 2) the need for sterilization for consecutive 337 use; and 3) the maintenance of sterile conditions, as the equipment has 12 gears that 338 could not be cleaned properly during the course of the experiment. These issues make 339 the SpinΩ dependent on user skills, rendering it more prone to error and susceptible to 340 contamination over the long timeframe of brain organoid cultivation (up to 8 months) 341 (Lancaster and Knoblich, 2014b;Quadrato et al., 2017), when compared with the use 342 of steering plates on an orbital shaker. 343 We suggest that the lower velocities of the SpinΩ, which affect nutrient mixing, may 344 explain the decreased organoid growth seen on days 17 and 19, and the wide size 345 distribution of organoids observed on day 30. Overall, our CFD analysis indicated that 346 the fluid dynamic variables examined in this study are closer to those of the spinner 347 bioreactor for the steering plate on the orbital shaker. Therefore, this method should be 348 preferentially selected as a multiplex alternative to the use of a spinner bioreactor. 349 The appearance of diverse brain regions and pigmented regions labelled with the 350 retinal epithelium marker GS has been previously described (Quadrato et al., 2017) 351 and related to a regional differentiation in organoids. Organoids produced using our 352 protocol presented pigmented regions positive for GS suggesting that the technique 353 described here may be appropriate for studies involving the complexity of early brain 354

development. 355
In addition, proteomic analysis confirmed the organoids (produced in this study) show a 356 protein profile that is compatible with several differentiated brain regions. Altogether, Organoid morphology on selected days of the 30-day culture period, with arrows 456 indicating the length of exposure to each medium condition. On day 14, organoids were 457 divided into growth in the orbital shaker (red) and in the SpinΩ (blue). Scale bar = 458 1,000 μm c. Area growth curves (μm 2 ) from day 1 to day 30, from n = 2 independent 459 tests. Each test replicate contained at least 12 individual brain organoids. Lines