Due to the increasing demand for molecules derived from natural resources, coupled with the heterogeneity of production crops, seasonality, and low yields leading to scarcity, the need for optimizing biotechnological processes for producing such molecules has grown. Key challenges in this area include the practical and economic feasibility of these production processes, often based on large-scale cultivation of high-yielding microorganisms (fermentation).
To capture the formed products during the production process, "In Situ Product Recovery" (ISPR) can be applied, for example, by adding an organic solvent during fermentation. ISPR also has the added benefit of preventing the accumulation of hydrophobic products in the culture medium to levels that inhibit the growth and production of the microorganisms. However, the use of organic solvents combined with surface-active compounds (SACs) poses challenges for product recovery after the stabilization of the culture medium.
The research described in this thesis was conducted within the Delft Integrated Recovery Column project (DIRC), which focuses on designing an integrated reactor for the production and recovery of hydrophobic compounds such as alkanes or isoprenoids.
This thesis explores the production of α-santalene and α-humulene, terpenes naturally found in the essential oils of Santalum album and Humulus lupulus, respectively. The production of these compounds in two different Escherichia coli BL21(DE3) strains was studied during fed-batch fermentations using glycerol as a growth-limiting substrate. Significant differences were observed between the two tested strains regarding cell viability and productivity.
Chapter 1 describes why sesquiterpenes (C15H24) are valuable non-polar molecules for various industrial sectors, based on their bioactivity or use as biofuels and fragrances. It also provides an overview of the biosynthetic pathways for terpene production, particularly the biosynthetic route for using glycerol as a carbon source. Finally, it reviews the literature on the advantages and limitations of Escherichia coli for the expression of heterologous products.
Chapter 2 presents the initial results of the biotechnological production of α-santalene, using ISPR (with the addition of dodecane) in 2 L bioreactors. Various process variations (such as different feed rates and induction strategies) were tested in fed-batch culture experiments with a genetically modified Escherichia coli BL21 (DE3) strain known for its high variability in sesquiterpene productivity. Moreover, the values of the black-box parameters (a, ms) from the Herbert-Pirt correlation for the specific substrate consumption rate (qs) were estimated based on the experimental data obtained.
Chapter 3 focuses on the same α-santalene-producing strain studied in Chapter 2, but instead investigates plasmid dynamics during an experiment without dodecane: while this organic solvent improves bioprocess performance, it interferes with qPCR assays.
The biotechnological production of α-humulene, also carried out in 2L bioreactors, is discussed in Chapter 4. This was done using a better-producing E. coli strain that our group gained access to in the final phase of this project. The productivity exceeded expectations compared to the findings described in the previous chapters. The influence of the substrate glycerol feed rate was examined, in combination with ISPR, as well as the impact of the absence of dodecane on specific productivity. Herbert-Pirt parameters were estimated for a fed-batch experiment conducted with a constant glycerol feed rate. Additionally, correlations between specific humulene productivity (qp) and specific growth rate (µ) were presented.
To base the modeling on realistic data, Chapter 5 determined and compared experimental oxygen transfer rates under actual extractive ("overlay") and non-extractive conditions for sesquiterpene fermentation with current insights into coalescing and non-coalescing systems. While the kLa values in extractive fermentation (ISPR) indicate non-coalescing behavior, the kLa of solvent-free fed-batch follows a general correlation for coalescing systems. Combined with the specific productivity results reported for the humulene-producing strain in Chapter 4, Chapter 5 outlines the design of scaled-up production processes, to be carried out in 1 m³ and 10 m³ bioreactors. Using a developed black-box model, the effects of two different glycerol feed strategies were compared: an exponentially increasing feed rate (Exponential Feed, EF) and a controlled feed rate aimed at a constant maximum oxygen consumption rate (Constant Oxygen Consumption Rate, COCR).
Chapter 6 summarizes the key findings and provides recommendations for further research and development.

The growing group of health-conscious consumers drives the development and sales in the market of low- or non-alcoholic beverages. In particular, sales in alcohol-free beer (AFB) has drastically gained in recent years, giving the incentive to further advance alcohol-free products. This thesis comprises the development of a new technology to produce an alcohol-free beer with an improved flavour, that is, an AFB with a significantly lower concentration in Strecker aldehydes. Despite the fact that Strecker aldehydes are present at trace concentrations of a few micrograms per litre, these compounds are majorly associated with the characteristic wort offflavour in alcohol-free beer. In the first stage, the origin of wort flavours is elucidated, and furthermore, the impact of the changed production process in comparison to regular beer and its resulting changes in the product is discussed. The most significant pathways of origin are the Maillard reactions and Strecker degradation, however, many other formation pathways have been suggested. While alcohol-free beers produced through regular fermentation and subsequent dealcoholisation are generally low in wort off-flavour, biologically produced AFBs exhibit a rather strong wortiness. This is because yeast’s activity during the fermentation is restricted to prevent alcohol formation, having the side effect that other compounds such as aldehydes are not converted to neutral or characteristic beer flavours. As it is impossible to prevent the formation of wort flavour, selective removal is required to produce an alcohol-free beer with improved flavour profile (balance). Subsequently, technologies applied to remove aldehydes from aqueous streams are investigated. Mechanisms that can be exploited for such removal are various – from separation based on size or volatility (distillation, pervaporation or membrane filtration) through sorption or conversion – past research has nearly studied all removal techniques available for aldehydes. However, most of the available techniques are not selective enough for a complex matrix such as beer, or not suitable for the application in the food industry. Nonetheless, adsorption is identified as the most promising technique to facilitate selective aldehyde removal. As a benchmark system, a dealcoholisation plant (spinning cone column) is studied with special focus on wort flavour removal. While volatile aldehydes such as 3-methylbutanal can relatively easily be reduced, methional, having a high boiling point, is not sufficiently separated from the end-product. Furthermore, due to the impact of heat, and ergo, the continuous formation and evaporation of these flavours, a plateau in their concentration is observed, where no further removal seems possible. This theory is proven by spiking a reactive sugar into the base and forming the mass balance over the system. Hence, in order to achieve a high reduction on high boiling wort flavours, another technology, such as adsorption, is required. To test, whether a suitable adsorbent can be identified, a wide screening experiment covering three adsorbent types (amine-functionalized polymers, hydrophobic resins and zeolites) and a total of 21 materials is performed. ZSM-5 type zeolites prove to have a superior selectivity as well as a relatively high capacity compared to other adsorbents. The choice of the right pore size and hydrophobicity is crucial for the separation success. Adsorption of the compounds of interest is non-competitive within the design space and robust against small changes in the beer matrix. The process is transferred to pilot scale, where it is shown that the zeolite effectively reduces wort off-flavour while maintaining the original character of the beverage. The reduced wort flavour concentration is also maintained over a period of 4 months ageing. Interestingly, the formation rates of aldehydes related to ageing are similar for the treated and reference product, with the exception of trans-2-nonenal, which is manifold lower in the treated AFB. A trained sensory panel confirms the difference in taste for both, the fresh and the aged product. In order to upscale this process to industrial scale, also the dynamics of the adsorption process on granular, binderless ZSM-5 zeolite is studied. The homogenous surface diffusion model is employed to regress the intraparticle diffusivity. The so obtained parameters indicate strong intraparticle mass transfer limitations as well as an inverse correlation of the effective diffusion coefficient to the molecules’ hydrophobicity. To avoid this process bottleneck, material improvements such as smaller crystal size and smaller granules are recommended. Finally, all obtained parameters are translated into a unit operation design. With a simple packed-bed operation of crushed particles, the production costs can be as much as 40 % lower than thermal dealcoholisation unit. Nonetheless, several assumptions still need to be verified, such as the material regenerability as well as other non- process-related factors such as material stability and food safety. Concluding, the newly developed technology for selective wort flavour removal in AFB in combination with restricted fermentation represents a great alternative to conventionally produced AFB.

Targeting to improve the utilization of lignocellulosic residues in the ethanol processing industry, this work aimed to test if the product inhibition of the enzymatic hydrolysis could be relieved by extractive reaction using aqueous two-phase systems (ATPS). The performance of enzymatic hydrolysis in ATPS is not well defined in literature. In this thesis, this extractive reaction was tested in terms of experimental conversion of sugarcane bagasse, simulations through conceptual process design and economic feasibility. A thermodynamic framework was developed in order to predict ATPS formation. The screening of ATPS and partition coefficient of the solutes were performed in a high throughput station. The ATPS were composed by polymer and salt. The enzymes were represented by the enzymatic cocktail Cellic CTec (Novozymes). The development of this platform consisted of two main parts: determination of phase diagrams (binodal curves and tie lines) and quantification of the solutes (sugar and proteins) in both top and bottom phases. The most promising ATPS were experimentally explored for enzymatic hydrolysis of sugarcane bagasse. Process design simulated two scenarios: hydrolysis occurring in the bottom phase and in the top phase. Topics such as the adsorption of phase forming components to the bagasse fibers and the influence of enzyme load on the hydrolysis were explored. The sugarcane bagasse hydrolysis in ATPS was conceptually assessed through the implementation of a model composed by two parts: hydrolysis and ATPS multi-batch separation. The designed case characterized by the ATPS hydrolysis was compared to the base case defined as conventional hydrolysis. Regarding the thermodynamic modelling of ATPS, the application of Flory-Huggins (FH) model to predict phase separation in polymer-salt systems was assessed. The implementation and analysis of FH theory involved the estimation of interchange energy (푤푖푗) and the calculation of phase diagrams. There were no statistical differences in determining the phase diagram in HTP platforms and bench-scale, verifying the reliability of methods and equipment suggested in this work. Moreover, tailored approaches to quantify the solutes were 8 presented, taking into account the limitations of techniques that can be applied with ATPS due to the interference of phase forming components with the analytics. This fast methodology proposes to screen up to six different polymer-salt systems in eight days and supplies the results to understand the influence of sugar and protein concentrations on their partition coefficients. Exploring experimentally the ATPS hydrolysis provided strategies on how to conduct extractive enzymatic hydrolysis in ATPS and how to explore the experimental results in order to design a feasible process. In the conceptual design of extractive enzymatic hydrolysis, one of the major bottlenecks identified was the partitioning of glucose to both phases. The resultant conceptual process design operates as a tool to evaluate ATPS hydrolysis and compare it to conventional one. On the other hand, the thermodynamic model could not quantitatively describe the data. This occurs mainly because of the strong influence of random experimental errors on the estimation of interchange energy, systematic errors when translating the observed data to calculated partition concentrations, and FH not being an exact description of phase separation in salt based ATPS. The high throughput screening methodology indicated ATPS able to partition sugar and enzymes. The selected ATPS presented no significant improvements to perform the enzymatic conversion of sugarcane bagasse compared to the conventional hydrolysis. The main reasons were the influence of phase forming components on the enzymatic activity and the low selectivity of sugars in the ATPS. To disclose the application of ATPS in the ethanol processing industry, the recovery and reuse of the phase forming components are imperative for economic feasibility. Moreover, the developed high throughput platform could be further employed to exhaustively screen systems to design effective ATPS for the partition of sugars and proteins in polymer-salt systems.

Trichoderma harzianum P49P11 was selected among several other microorganisms at the Biorenewables National Laboratory (LNBR, Brazil) and it was considered a promising strain to produce cellulase. Here in this project, a mathematical model and simulation platforms were developed as potential tools to be used for cellulase maximization using fed-batch mode (Chapter 2). Feeding strategies were simulated to maximize cellulase production, at first, only using cellulose as the substrate, and then using glycerol for cell growth and cellulose for cellulase production. Although the mathematical model and simulation platforms were built up for a wild type strain, these tools help to predict data and they can be adapted for optimized strains.
Chapter 3 evaluates the continuous production of enzymes using different carbon sources under carbon-limited conditions. It was found that glucose has a positive influence on the production of enzymes that can catalyse the hydrolysis of p-nitrophenyl-β-D-glucopyranoside (PNPGase). Sucrose and fructose seem to inhibit PNPGase synthesis; however, these substrates could also have a positive influence on the synthesis of other enzymes not evaluated in this project. Cells can uptake glucose without the need to synthesize extracellular enzymes like PNPGase. The increase in the production of PNPGase during the continuous culture using glucose as the carbon source indicates the presence of inducers. It was also discovered in this project that polysaccharides were present in the supernatant of all conditions using glucose, fructose/glucose and sucrose (Chapter 4 and Chapter 5). This suggests that the possible inducers could have come from fragments of the extracellular polysaccharides.
Sugar analysis showed the presence of sugar with the same retention time as gentiobiose in the supernatant of the conditions using glucose as the carbon source, which could be a fragment from polymers released from the cell wall. Gentiobiose could be acting as an inducer of enzymes. In addition, a mechanism was also proposed for continuous PNPGase production under glucose-limited conditions assuming that PNPGase includes beta-glucosidase (Chapter 4).
The carbon sources used under carbon-limited conditions influenced the PNPGase productivity and possibly the whole enzymatic cocktail secreted by the fungus. For this reason, shotgun proteomics and SDS-PAGE analysis were performed for the proteins present in the supernatant of the conditions using glucose, fructose/glucose and sucrose (Chapter 4). The shotgun proteomics analysis suggested that the different carbon sources used provided the production of different extracellular proteins including several uncharacterized proteins, which can also include different enzymes. This brings the possibility of creating a hypothesis that different carbon sources easily assimilated by the cells could lead to the synthesis of different inducers (fragments of extracellular polysaccharides), which could induce the synthesis of different enzymes under carbon-limited conditions.
Extracellular polysaccharides were the by-products discovered in this project during the production of enzymes under carbon-limited conditions. The behaviour of intracellular metabolites (glycolysis, citric acid cycle, pentose phosphate pathway and nucleotides) was evaluated under four different conditions in duplicate during the production of extracellular polysaccharides by Trichoderma harzianum under carbon-limited conditions (Chapter 5). This chapter has provided the first step for the optimization of the production of extracellular polysaccharides and the information about the behaviour of intracellular metabolites using this wild type strain is essential to the development of optimal strains.

Flavor-active components are key contributors to the profile of the final produced beer product. Their preservation and control during different stages of processing is crucial, since they might be lost during processing due to their volatile nature. In order to produce a final beer product with balanced flavor profile, which is acceptable by the consumer, the level of these components in the beer matrix should be adjusted and controlled. Various techniques can be applied for flavor control and recovery, such as distillation/stripping, pervaporation, supercritical extraction, and adsorption. Chapter two of this thesis discusses the recent advances in various techniques, which are applied for flavor recovery among which adsorption is a technique, which showed potential for selective removal and recovery of flavor/non-flavor-active components. This technique can be combined with heat processing, distillation/stripping, or can be used as a standalone technique. The focus of the work presented in chapter three of this thesis is on method development for selective removal and recovery of flavoractive volatile components mainly belonging to the group of esters, higher alcohols, and diketones, through adsorption technique. In order to investigate the single and competitive adsorption behavior of flavor-active components and their synergistic effects, high throughput experimentation technique is applied, improved for volatile components and isotherms are obtained using batch uptake experimentation. The competitive adsorption behavior of flavor-active components is investigated on various food-grade hydrophobic adsorbents, in order to study the influence of physical and chemical nature of the components and adsorbent properties on selectivity for each tested adsorbate over ethanol. Based on the results obtained from thermodynamic studies through various isotherm models, the appropriate adsorbent material is selected for further studies in the design stage. In the next step, deeper study is conducted on flavor-active esters, presented in chapters four, five, and six, which contribute to beer with a fruity taste and aroma. With adjusting their level in the final beer product and their fractionation, various products can be produced with fruity taste. Further investigation is performed on their competitive adsorption behavior both through batch uptake experimentation, and dynamic breakthrough analysis tests, discussed in chapters four and six respectively. Since ester components are present at low concentration level together with ethanol, which is present at higher concentration in comparison in various process streams, the influence of ethanol and temperature on their competitive adsorption is further investigated, discussed in chapter four. Physical properties such as isosteric heat, entropy, and Gibbs energy of adsorption, are calculated from performed thermodynamic studies, which contribute to our deeper understanding of the adsorption phenomena on the selected adsorbents. Considering the time-consuming steps, which are required to be followed for constructing the adsorption isotherms through batch uptake experimentation, the application of predictive models developed based on adsorbed solution theory, is evaluated in chapter five, for prediction of multicomponent adsorption isotherms for flavor-active esters from single-component adsorption isotherms, when experimental data for multicomponent behavior is not available. The predictive model developed based on IAST, was capable to predict the multicomponent adsorption behavior for the tested condition with accuracy and can be used as a tool for prediction of isotherms, when data on multicomponent adsorption is not available. Possibility for separation of flavor-active esters is further investigated in a fixed-bed column in lab-scale, discussed in chapter six, to study their breakthrough behavior and their separation under various process conditions (ethanol concentration and temperature). The results of the experimental tests obtained through breakthrough analysis and fractionation, are used for validation of simulations. Based on the results obtained in lab-scale, separation of flavor-active components is further investigated in a large-scale column, through simulation of various scenarios for separation. The performed experiments in the lab-scale and results of the simulations at large-scale give an insight on competitive adsorption behavior of these components, when present in a mixture. For a more detailed prediction of the adsorption behavior, future outlooks are discussed in chapter seven, to study the optimized condition considering the process conditions, and integration of the adsorption with other alternatives such as distillation/stripping.