RESULTS AND DISCUSSION

3.1 Flow field prediction and CFD model validation

In order to test the reliability of CFD simulation and calculation algorithms, preliminary simulations were carried out, and the model was verified by comparing the KLa and \(P/V\) results of the simulated gas-liquid interface with the results calculated by empirical formulas in the literature. The grid cells of the microtiter plates were divided into approximately 550,000 cells, with verified independence(Fig.S1 ). The results of the simulation were shown inFigure 2 . It can be clearly seen in the Figure 2A and2B that the CFD simulation analysis results were very consistent with the results in the experimental literature. The predicted values differed from the results of the experiment, due to the slight differences in the fluid characteristics of the fermentation broth in experimental culture and the CFD. The liquid phase in the CFD simulation used the rheological properties of pure water, with a density (ρ) of 998.2 kg/m3, a viscosity (θ) of 0.01 m2/s, and a liquid surface tension (σ) of 0.0728 N/m. There was good consistency between the simulated value and the experimental value (p<0.05), indicating that the CFD model could realistically reflect the data in the experiment.
In addition, the flow field prediction diagram of microtiter plate sloshing was displayed by CFD to characterize the actual flow field at the gas-liquid interface. The specific movement of the liquid in the microplate was shown in Figure 2 , with blue parts representing the liquid phase and red parts representing the gas phase.Figure 2C displayed the contours plot on the cross-section, and Figure 2D showed a three-dimensional motion model of the microtiter plate. By tracking its movement during orbital vibration, both figures clearly illustrated the real situation of fluid movement in the experiment. During the rotational movement, due to the action of centrifugal force, the liquid moved towards the wall, and the gas-liquid interface area increased rapidly under strong vibration conditions. After the orbital cycle was complete, the liquid in the microplate circulated and reciprocated. Throughout the study, it was assumed that the gas-liquid interface was located on an iso-surface with a liquid phase volume fraction of 0.5. At the same time, the images of the movement track also verified the authenticity and reliability of the CFD model.

3.2 Performance evaluation of conventional microtiter plates

Most of the microtiter plates used in current studies have circular or square cross-sections. It has been proven that the oxygen transport efficiency of square cross-sectional microtiter plates is significantly higher than that of circular microtiter plates [40]. Therefore, only square cross-sectional microtiter plates were used in this section to explore and evaluate their performance during the culture and fermentation of strains. During experiments, the growth of microorganisms led to rheological changes in the fermentation broth. The changes in density, viscosity and surface tension must be considered during culture, in which the surface tension has a significant effect on the mass transfer efficiency between gas and liquid [41]. As shown in Table 1 , the surface tension of the fermentation broth increased from 0.076 N/m to 0.103 N/m during the entire fermentation process. This rise had a significant impact on mass transfer and mixing performance between gas and liquid. Therefore, the phenomenon must be discussed in depth to explore the applicability of microtiter plates.
In this study, the changes in rheological properties of the fermentation broth of strains were simulated and analyzed in ANSYS Fluent. The simulated environment was performed at 300 rpm and 15% liquid load, and the results were shown in Figure 3 . With the passage of culture time, the oxygen supply efficiency and mixing performance of the microtiter plate gradually decreased. The average turbulent dissipation rate (ε) also decreased from 0.45 m2/s3 at 0 h to 0.24 m2/s3 at 24 h, which greatly limited the growth of the microbes. The strain cultured in conventional microtiter plates was sampled after 24 hours, and it was observed microscopically that the biomass did not show a well-growing trend. The diameter of the cell was small, and the distribution was sparse. This growth condition resulted in low biomass and poor microbial activity, preventing the strains from meeting the needs of high-density fermentation, which would also have a negative impact on DHA production. Therefore, it is of great significance to optimize the existing microtiter plates to adapt to the culturing and screening of DHA producing strains.

3.3 New high-throughput bioreactor design

3.3.1 Reactor geometry design

Previous studies have demonstrated that the cross-sectional geometry of microtiter plates significantly affects their mass transfer and mixing properties [42]. In this section, the CFD technology was used to model microplates with different geometries and analyze them numerically, which explained the function and importance of geometry in microbial culture. Figure 4A presented a series of microtiter plate models with increasing edges, having a cross-sectional area limited to approximately 60 mm2 and a total volume of 1.2 mL per titer. A vibration frequency of 150 rpm and a liquid charge of 15% were used for the numerical solution to ensure the consistency of reaction parameters. The influence of different reactor geometries was analyzed. Real-time monitoring of the microtiter plate’s oxygen transport efficiency was performed using Equations 3 and 4 in Fluent, and its mixing performance was monitored in real-time using Equation 6. The results were output after 5 seconds of motion at a vibration radius of 3 mm, as shown in Figure 4 .
The results of the calculation for mass transfer related parameters KL, a, and KLa can be seen inFigure 4B . As the cross-sectional geometry of the microtiter plate changed from quadrilateral to hexagonal and gradually increased to circular, the data for oxygen transport efficiency fluctuated, with the relevant parameters reaching their highest values in the hexagonal microtiter plate. When the number of edges was increased from the conventional quadrangle to a regular hexagon, KLa rose significantly. Since KL and a in the miniature vibrating bioreactor contributed almost equally to KLa, the three parameters performed consistently. These parameters reflected the best performance in hexagonal microplates. In subsequent geometry modification, the mass transfer efficiency gradually decreased and finally reached a minimum oxygen transport efficiency in the round microtiter plate, which was also consistent with other studies. The oxygen transfer efficiency of the square cross-sectional microtiter plate was significantly higher than that of the round microtiter plate, and the mass transfer related parameters of the round microplate were far inferior to those of other geometric reactors. In addition, the turbulent flow kinetic energy (k), turbulent energy dissipation rate and volumetric power consumption of the mixing performance were also characterized in Figure 4C . Similar to the changes in mass transfer parameters, as the number of sides of the reactor gradually increased from quadrilateral to hexagonal, the mixing performance was gradually optimized. Subsequently, in the regular heptagonal shape, the mixing performance continued to improve slightly, because more edges and corners affected the intensity of the internal movement. However, this would lead to the splash of fermentation broth in the experiment, increasing the risk of contamination. In simulations that continued to increase the number of microplate edges, this short-lived improvement in mixing performance disappeared and was replaced by a similar trend to the mass transfer parameters, which reached a minimum in round microplates. In conclusion, hexagonal microtiter plates showed superior mass transfer and mixing performance in the simulation, which laid the groundwork for future development of screening and culturing of DHA producing strains.
Moreover, the contour diagram of the gas-liquid interface of microplates in motion was characterized in CFD-Post to demonstrate the superiority of hexagonal microtiter plates in mass transfer and mixing performance (Fig.S2 ). The figure showed the differences in the speed of liquid movement in different geometries of microplates. In hexagonal microplates, the liquid had the fastest rate of movement. Additionally, the liquid movement trend in several microplates showed consistency. The movement speed near the bottom of the microtiter was low, while it was relatively high at the upper end. The specific movement of the liquid was caused by the applied centrifugal force. The superiority of the new bioreactor was also verified in the velocity contours of the liquid, which were consistent with the previous simulations. However, modifying the geometric shape of the microtiter plate alone is far from sufficient to meet the conditions of culturing and screening the DHA producing strains, and further improvements will be necessary in subsequent work.

3.3.2 Structure and reaction parameter optimization

In order to obtain optimal growth conditions, multiple parameters need to be optimized for the growth of DHA producing strains. However, gradually optimizing reaction parameters in software or experiments is a complex and difficult task. The mathematical and statistical response surface methodology (RSM) technique is a widely used method that consists of a set of mathematical techniques to describe the relationship between input and output parameters for modeling and optimization purposes. This RSM model estimates the combination of the input parameters yielding an optimal response through fast-running approximation of the simulation process [43]. In this study, the rotational speed, liquid filling volume and the number of baffles were regarded as three independent variables A, B and C, KLa and P/V were used as the result variables, and the RSM model was established based on the numerical calculation results of CFD. A complete experimental design matrix was given, and CFD-based simulation numerical experiments were carried out according to the experimental plan. The final results showed the effectiveness of each parameter and the influence of its interaction on the mass transfer and mixing performance. In subsequent work, ANOVA was performed to assess the quality of the developed model, as shown in Table 2 . Based on a 95% confidence level, the results of the second-order quadratic models of KLa and P/V have significant F and P values (<0.05), which showed that the model constructed in this study was suitable for the range of variables studied. In the results of ANOVA in Table 2, the coefficients of determination R2 were 0.9803 and 0.9775, indicating that the model had a perfect correlation between the independent variables and the response. The corrected R2 were 0.9550 and 0.9486, indicating that only about 5% of the change could not be explained by the model.
The three-dimensional response surface was plotted based on the fitting second-order polynomial (Fig.S3 ). The influence of the baffle number, liquid charge and rotation speed on the changes of KLa and P/V was clear. Consequently, the interaction between the vibration velocity and the number of baffles significantly affects the mass transfer and mixing performance of the new bioreactor. The number of baffles and shaking frequency of the bioreactor were positively correlated with mass transfer and mixing. With the increase of the baffles and rotation speed, the performance of the bioreactor was gradually optimized. On the contrary, the liquid filling volume was negatively correlated with the mass transfer and mixing performance, and the performance of the bioreactor tended to decrease with the increase of liquid load. The optimal conditions recommended by the CFD-based RSM model included 6 baffles, 15% liquid load, 800 rpm, maximum KLa level of 0.61 s-1, and P/V of 2364 W/m3, which were sufficient to meet the growth conditions of DHA producing strains. The splashing phenomenon that might occur during the culture was also explained in subsequent model verification experiments. In summary, this study completed a method combined with the experimental design of the response surface on the framework of CFD technology to obtain the optimal culture conditions for DHA producing strains in a novel high-throughput bioreactor.
In this section, the superiority of the new high-throughput bioreactor for culturing and screening of DHA producing strains was established on the basis of CFD-RSM technology. Furthermore, simulation verification of CFD and experiments with solid microplates were conducted. The optimal reaction parameters given by the RSM model were numerically solved and compared for conventional quadrilateral microplates, hexagonal microplates, and the novel bioreactor using Fluent. The mass transfer and mixing levels of the new bioreactors were significantly higher than those of the other two microtiter plates. To further observe the superior performance of the new bioreactor, the liquid phase motion speed of the three microplates was derived in CFD-post. And planes of different heights were selected to reveal the distribution state of the liquid phase (Fig.5 ). Figure 5A showed that the motion state of liquid in the new bioreactor is more intense and uniform than that of the other two reactors, and the movement speed at parallel time is also significantly higher than that of the other two, which is conducive to the oxygen transport and nutrient absorption during the strain growth. In Figure 5B , several planes were selected inside the bioreactor. During the vibration movement, it was found that the liquid level height was higher in the new reactor. The closer it was to the ostiole, the better liquid phase distribution, compared with the other two reactors. The liquid phase distribution was more uniform in the new reactor rather than being limited to the wall surface, which attributed to the different geometric structure. All in all, the new bioreactor showed higher mixing and mass transfer levels than conventional reactors during simulations, and this improvement was sufficient to meet the culturing and screening needs of DHA producing strains.

3.4 Utilization of the novel microplate for screening DHA producing strains

Based on 3D modeling and printing technology, this study established and printed a new microtiter plate in equal scale. Three DHA producing strains stored in the laboratory were screened using the new microplate. The culture medium described above was used to cultivate the three strains in both conventional and novel microplates for 48 hours. The biomass of the fermentation broth was measured at 0, 12, 24, 36, and 48 hours. The results were shown in Figure 6A . The three species showed different growth states and trends over the culture time in the microplates. Under the same culture conditions, the three DHA-producing strains showed no significant difference in conventional microplates, and their biomass was lower than that of the novel microplates at the end of the culture. However, in the new microplates, the growth state and rate of Schizochytrium sp were significantly higher than those of the other two strains, and the biomass of Schizochytriumsp was 20% higher than that of Aurantiochytrium sp. andThraustochytrium sp. at the end of culture. This phenomenon indicated that the oxygen level and mixing capacity provided by the new microtiter plate were better than those provided by the conventional microtiter plate in the process of constant temperature oscillation culture. There was no liquid spatter phenomenon during the entire culture process, which facilitated the growth of microorganisms. In addition, the new bioreactor had a larger cell diameter and higher cell density, while the conventional reactor had a small cell morphology and sparse cell density, due to oxygen dissolved limitation and uneven mixing during the culture process. In conclusion, the performance of the new bioreactor is significantly superior to that of the conventional bioreactor in both simulation experiments and laboratory experiments. In the process of vibration culture, it provides sufficient oxygen level and high mixing efficiency for the growth of strains, which is conducive to the high-throughput culturing and screening of high-yield DHA strains.

3.5 Optimization of culture conditions by novel microplates

On the basis of screening out a high-quality strain, the new microplate was used to optimize the culture medium in order to increase the DHA yield of fermentation. Design-Expert software was used to optimize the content of monosodium glutamate (MSG), yeast extract (Yeast) and glucose (Glu). Nile red staining reagent was used to quantitively characterize the biomass, and the fluorescence intensity value was measured and used as the regulatory variable. The RSM model provided 17 kinds of experimental designs. Three groups of parallel experiments were conducted according to the recommended experimental design. After 48 h of culture, the samples of three groups were stained for measurement of fluorescence intensity (Table 3 ). Consequently, the optimal medium conditions included 20g/L MSG, 20g/L Yeast and 60g/L Glu. In order to verify the accuracy of the experimental design recommendation, at the end of the fermentation, the growth state of the strain was observed and recorded under a fluorescence microscope, as shown inFigure 6B . In the following work, Image J was used to process the pictures and calculate the average fluorescence density and the average diameter of the microorganisms under each different regulation medium state. The average fluorescence density represented the biomass of the strains, and the average diameter quantified the growth state of the strains. Considering the comprehensive comparison between the two parameters, the actual growth state of the strain under the regulation of the medium was observed, and the numerical results were shown inTable 3 . The fluorescence images of the strains were consistent with the predictions of the RSM model. Under the recommended experimental design, the density and diameter of the strains were generally larger than those of other experimental designs. This study efficiently optimized the growth parameters of DHA-producing strain culture by using a newly designed microplate combined with response surface design, which is of great significance for guiding industrial fermentation for DHA in the future.