# Biochemical Production Systems using Photosynthetic Organismus

Photosynthetic microorganisms such as green microalgae are innovative cell factories for the sustainable production of chemicals from renewables. These microorganisms are capable of synthesizing a wide range of compounds from sunlight and carbon dioxide at higher volumetric and areal productivities compared to land-based crops. As an example for a photosynthetic systems we concentrate our work on β-carotene production from the microalga *D. salina*.

From the systems-theoretical point of view, modeling and control of micro-organic cultivations is a very challenging task: biological cell populations represent a complex dynamic system hierarchically organized on multiple temporal and spatial scales, of which we have only limited mechanistic knowledge (Figure 1). Additionally, biological systems exhibit strong variability in the phenotype, e.g. metabolite dynamics, which renders difficult experimental probing for model identification. Given the hierarchical multi-scale structure of a micro-organic cultivation, the **Elementary Process Functions** **Methodology**** **** (EPF) **is ideally suited for the optimal design of bioreactors. Therefore, we develop computational models and experimental methods to describe cellular metabolism, cell population and bioreactor dynamics, and finally, the plant level, including precultivation and downstream processing.

For an optimal bioprocess design, the predictive power of a dynamic-kinetic model, e.g. growth rate or yield, is an important feature and involves two aspects: a valid model structure and well identified model parameters. In order to derive predictive models, we use innovative experimental design methods to optimize our experimental data [1], which we analyze with several quantitative techniques including High Performance Liquid Chromatography (HPLC), Fourier Transformed Infrared Spectroscopy (FTIR) and Flow Cytometry (FC). With these tools we identified a comprehensive computational growth model of

*D. salina*covering nutrient uptake, light attenuation and pigmentation (Figure 2, [2]), which we use for robust design optimization and as a basis for a multi-scale growth model that also includes the metabolic level of

*D. salina*. The term robust refers to the consideration of natural biological variability and parameter uncertainties. Additionally, fluctuating operating conditions in combination with general process noise need to be considered to not render a purely deterministic process design suboptimal. Therefore we have developed a probabilistic process design approach based on a hybrid uncertainty propagation strategy (Figure 3) that accounts for uncertainties in the model parameters and variabilities in the process design [1, 3].

On the plant level, we developed an overall process model based on mass and energy balances to compare alternative optimal processing routes from the individual cell to the final product. By integrating real experimental results and economic aspects, the process model could reveal strengths and bottlenecks of the individual optimized processing steps. Primarily, we focused on the optimization of the biomass harvesting unit for potential reduction of operation cost, since on the one side this is a very challenging step given the low density and fragility of *D. salina*. On the other side, the optimal choice of a downstream route depends on the desired product [4]. For that reason we investigated chemical and physical flocculation as an innovative ß-carotene harvesting approach for microalgae (see Figure 4) in comparison to the conventional method of centrifugation. A reliable comparison of the different harvesting methods based on operation cost and energy demand of the overall process was achieved by integrating all experimental results into the process model [5]. This systems approach can be transferred to assess other process units, e.g. extraction of ß-carotene or post-processing by liquefaction of residual biomass. In consequence, the developed concept provides a general platform to evaluate microalgal process methods in a reliable manner on plant scale.

**Figure 1:** Multi-scale representation of an algal-based production system. The different length and time scales can be associated to the molecular, phase, process unit and plant level.

**Figure 2**: Growth model of *Dunaliella salina*.

(b) Identifiability analysis of 4 model parameters based on the profile likelihood (black solid lines c^{2}). Blue solid lines represent the profile likelihood level at fixed w* _{N,max}*. The red dotted line represents the critical c

^{2}value at significance level a = 0.05. The green asterisk indicates the best parameter estimate. Figure from [2].

**Figure 2**: Growth model of *Dunaliella salina*.

(a) Model simulation for high light (HL) and high light-nutrient depletion (HL-ND) scenarios compared to experimental data. Solid lines represent model simulations, points mark exp. data.

**Figure 3:** Scheme of hybrid uncertainty propagation strategy. The design *D* with its imperfect realizations (grey) and uncertain model parameters (dark gold) are mapped to the design objective using via sampling of sigma points and profile likelihood based confidence regions (CR). Figure from [3].

**Figure 4:** Simplified flow scheme of the overall process system for ß-carotene production using *D. salina* cultivation followed by multistage harvesting, extraction and optional liquefaction of the residual material. Figure from [5].

#### Recent Publications

[1] Flassig, R. J., Fachet, M., Rihko-Struckmann, L. and Sundmacher, K. (2015). Robust process design for the bioproduction of ß-carotene in green microalgae. In K.V. Gernaey, J.K. Huusom, R. Gani (Eds.), *Proceedings of 12**th **International Symposium on Process Systems Engineering and 25**th **European Symposium on Computer Aided Process Engineering*. 31 May – 4 June 2015, Copenhagen, Denmark, 2117-2122.

[2] Flassig, R. J., Migal, I., van der Zalm, E., Rihko-Struckmann, L. and Sundmacher, K. (2015). Rational selection of experimental readout and intervention sites for reducing uncertainties in computational model predictions. *BMC Bioinformatics*, **16**, 13.

[3] Pirwitz, K., Flassig, R. J., Rihko-Struckmann, L. K. and Sundmacher, K. (2015). Energy and operating cost assessment of competing harvesting methods for *D. salina* in a β-carotene production process. Algal Research, **12**, 161-169.

[4] Pirwitz, K., Rihko-Struckmann, L. and Sundmacher, K. (2015). Comparison of flocculation methods for harvesting Dunaliella. *Bioresource Technology*, **196**, 145-152**. **

[5] Fachet, M., Flassig, R. J., Rihko-Struckmann, L. and Sundmacher, K. (2014). A dynamic growth model of Dunaliella salina: Parameter identification and profile likelihood analysis. *Bioresource Technology*, **173**, 21-31.