Continuous phase separation of stable emulsions from biphasic whole‐cell biocatalysis by catastrophic phase inversion

The main bottleneck for the industrial implementation of highly promising multi‐phase whole‐cell biocatalytic processes is the formation of stable Pickering‐type emulsions, hindering efficient downstream processing. Especially for the crucial step of phase separation, state‐of‐the‐art processes require time‐consuming and costly process steps (excessive centrifugation/use of de‐emulsifiers). In contrast, using the phenomenon of catastrophic phase inversion (CPI), efficient phase separation can be achieved by addition of an excess dispersed phase within minutes. To show applicability of CPI as an innovative process step, a fully automated lab‐scale prototype was designed and constructed within this work. A simple mixer‐settler set‐up enabled a continuous phase separation using CPI termed applied catastrophic phase inversion (ACPI). Test runs were conducted using emulsions from biphasic whole‐cell biocatalysis (Escherichia coli JM101 and Pseudomonas putida KT2440 cells). Solvents used included n‐heptane, ethyl oleate or 1‐octanol as organic phase. These investigations revealed ideal process settings for a stable ACPI process (e.g., flow/stirring rates and volumetric phase ratios between organic and water phase). The knowledge of the CPI point is most crucial, as only the inverted state of emulsion is successfully destabilized.

F I G U R E 1 Scheme of CPI where an initially o/w-type emulsion stabilized by hydrophilic particles/cells (left) is inverted to w/o-type due to the addition of dispersed (organic) phase (middle), resulting in the loss of stabilizing effect in inverted emulsion as hydrophilic particles/cells remain in the now dispersed (water) phase. Without energy input, phases separate by gravity (right).
such emulsions have already been described in literature, including the biocatalytic epoxidation of styrene by Escherichia coli JM101 in a twoliquid (aqueous/organic) system with a volumetric phase ratio between aqueous and organic phase V o :V w = 1:1 by Kuhn et al. [3] Besides volumetric phase ratios of 1:1, also other phase ratios have been applied, for example, for the biocatalytic quinaldine hydroxylation by Pseudomonas putida in a two-liquid (aqueous/organic) systems with a phase ratio of V o :V w = 5:1. [4] Emulsion characteristics depend on the volumetric phase ratio between aqueous and organic phase (V o :V w ) and the wetting behavior of the cells to form o/w-(more hydrophilic cells) and w/o-type (more hydrophobic) emulsions. [5] In some cases also double or multiple emulsions can occur. [6,7] As much as emulsification is desired during the upstream processing (yielding in high surface areas and thus high mass transport rates), [1,3,8,9] downstream processing concepts often fail in processing the resulting (highly stable) Pickering-type emulsions. [10][11][12] Common concepts for the initial phase separation step of these emulsions, such as centrifugation, use of de-emulsifiers, filtration or membrane separations either fail, or require high effort in both costs and time. [13] In contrast to these common state-of-the-art phase separation concepts, phase separation can be achieved efficiently using the phenomenon of catastrophic phase inversion (CPI), first introduced 1988 by Salager. [14] Phase separation is herein achieved by continuous addition of an excess amount of dispersed phase until the emulsion inverts, thereby switching from oil-in-water (o/w) to water-in-oil (w/o) (or vice versa) ( Figure 1).

Emulsion phase separation using CPI
Depending on the wettability/wetting behavior of the particles/cells (expressed by the three-phase contact angle Θ ow ), the emulsion type that can be stabilized by these particles/cells is fixed. O/w-type emulsions are stabilized by hydrophilic particles/cells (Θ ow < 90 • ) and w/o-type (Θ ow > 90 • ) emulsions are stabilized by hydrophobic particles/cells. [6,7] The process of CPI is schematically shown in Figure 1 for an initially o/w-type emulsion stabilized by hydrophilic particles/cells. In this case, the addition of dispersed (that is organic) phase beyond a certain threshold results in a loss of stabilizing ability of the particles/cells, with the hydrophilic particles/cells residing in the now dispersed (water) phase. Without any further energy input, phase separation is easily achieved by, for example, gravimetric settling. [15] During the CPI process the apparent drop size within the emulsion is expected to change, as median drop diameters in o/w-type emulsions tend to be much higher than in w/o-type emulsions. [16] Moreover, multiple emulsions (e.g., o/w/o) are formed during the CPI process due to the addition of dispersed phase resulting in an increase of average dispersed phase drop size when approaching the CPI-point. [17] According to the work of Hohl et al., [18] a lower water content in w/o-Pickering-type emulsions did induce smaller drop sizes. Further effects on emulsion drop sizes are expected for the particle/cell concentration and hydrophobicity. Due to a higher surface coverage, higher particle concentrations result in smaller drops. Increasing particle/cell hydrophobicity led to increasing drop sizes. [18] Moreover, multiple emulsions (e.g., o/w/o) are formed during the CPI process due to the addition of dispersed phase resulting in an increase of average dispersed phase drop size when approaching the CPI-point. [17] Various batch experiments [5] already revealed the applicability of CPI as a tool for phase separation of various bioprocess-derived Pickering-type emulsions. Further, previous investigations within our group [5] also revealed, that the mechanisms of emulsions stabilization and destabilizing can be attributed/described by three key parameters, namely the particle size R, the three-phase contact angle Θ ow , and the interfacial tension γ ow . Furthermore, we developed a guideline to calculate/estimate the point at which CPI occurs (critical water volume fraction v w,crit. ) using these parameters. [5] This allows to define an appropriate solvent system and ACPI operating conditions for a given biphasic whole-cell biocatalytic process.

The concept of applied catastrophic phase inversion (ACPI)
Based on the batch experiments two drawbacks to the CPI concept ( Figure 1) remain: (1) The product is diluted, as it is preferentially located in the dispersed (mostly organic) phase (circumventing cell toxification). (2) Removing the excess organic phase reverses phase inversion and thus hinders gravimetric phase-separation. To circumvent these drawbacks, Glonke et al. [13] introduced the concept of ACPI, being the continuous sibling of the batch CPI concept using a continuous mixer-settler set up.
Within this work, we designed and constructed a fully automated lab-scale prototype for continuous phase separation adhering to the ACPI principle. We demonstrate the applicability of the concept for various long-term stable bioprocess-derived Pickering-type emulsions, investigating the influence of both, different organic solvents (n-heptane, ethyl oleate and 1-octanol) as well as biocatalysts (E. coli JM101 and P. putida KT2440). The critical volumetric phase ratio of organic to water phase (V o :V w ) which has to be applied to achieve phase inversion, was calculated based on the guideline developed in our previous work. [5] The process windows defined in this regard allowed for robust operating conditions for ACPI. To allow for a detailed investigation of the robustness of the process, we investigated the influence of process parameters (e.g., flow rates) on stability and success of the (continuous) phase separation. Furthermore, we investigated the robustness towards perturbations (e.g., fluctuation in water/organic phase ratio of the feed emulsion).

Strains, media, and process conditions
The Pickering-type emulsions investigated in this work were obtained by biphasic cultivation of E. coli JM101 as well as P. putida KT2440 in a KLF 2000 reactor (Bioengineering, Wald, Switzerland).
Key parameters for the biphasic cultivation are described in a previous work. [5] The biphasic system consisted of Riesenberg-medium [19] or Luria-Bertani-medium (LB) [20] as an aqueous (cultivation) phase and either n-heptane (Merck, Darmstadt, Germany 100%), ethyl oleate (Alfa Aesar, Kandel, Germany 70%), or 1-octanol (Alfa Aesar, 99%) as organic phase (V o :V w = 1:1). No biotransformation was performed within this work, thus after fed-batch cultivation, organic phase was added for about 1 hour to the aqueous cell broth. It was found in previous works (results not shown), [21] that the biotransformation itself only had a negligible influence on emulsion stability and highest values of overall emulsion stability are reached within the first hour of organic phase addition. [22]

Characterization of binary aqueous/organic system
For a first estimation of phase system properties (e.g., density and viscosity) of binary aqueous/organic systems, 0.05 M phosphate buffer (NaH 2 PO-K 2 HPO 4 ) at pH 7 was used as a sufficient simplification of an aqueous phased used during fermentation. 15 ml Falcon tubes were filled with aqueous (0.05 M phosphate buffer at pH 7 as model buffer) and organic phase (either n-heptane, ethyl oleate or 1-octanol) in equal parts. After mixing for 15 min, samples were equilibrated for 24 h at 25 • C. Thereby, two phases (aqueous-rich and organic-rich) settled within all samples (complete phase separation).
Density measurements of aqueous and organic phases were performed using a DMA 4100 M by Anton Paar GmbH (Graz, Austria).
Viscosity of both phases was measured using a temperature-controlled falling ball viscosimeter (LOVIS 2000 ME) by Anton Paar GmbH (Graz, Austria). All density and viscosity measurements were performed three times, ensuring reproducibility.

Monitoring of CPI via electrical conductivity
Within the ACPI prototype, phase behavior of the emulsion is monitored using conductivity sensors (CombiLyz AFI5, Baumer). As described in literature, [23,24]

Process concept of ACPI prototype
Based on a patent filled at TU Dortmund University (publication number EP 2 870 988 A1) and according to the work of Glonke et al. [13] a fully automated lab-scale prototype for continuous phase separation of long-term stable bioprocess-derived Pickering-type emulsions was constructed. Figure 2 shows the process flow diagram of the prototype including the measurement and control equipment.
The two tanks (tank 1 and tank 2 in Figure  Using a conductivity sensor (CombiLyz AFI5, Baumer) in tank 1 and the mixer, the apparent emulsion type can be determined through measuring conductivity, as only the o/w-emulsion type show significant values for electrical conductivity κ (see materials and methods section). [23,24] The feed emulsion is continuously pumped from tank  Figure 3 shows the fully constructed lab-scale prototype, constructed as described above.

Determination of emulsion stability and effectiveness of separation process
Three sampling points within the ACPI process are used to determine the emulsion stability and the effectiveness of phase separation. As shown in Figure 4, each sample is centrifuged (Eppendorf 5804R, rotor A-4-44, Hamburg, Germany) for 60 min at 4000 g and 25 • C.

Emulsion stability
The

Influence of emulsion characteristics on phase behavior and phase separation process
Various bioprocess-derived Pickering-type emulsions stabilized by E.
coli JM101 and P. putida KT2440 cells, containing either n-heptane, ethyl oleate or 1-octanol as organic phase were processed within the ACPI prototype.
To investigate the influence of emulsion characteristics on the separation behavior, the aqueous/organic base systems (no cells present) containing three different organic phases were characterized in terms of phase density ρ and viscosity η as well as interfacial tension γ ow . The latter namely between aqueous (0.05 M phosphate buffer) and organic (either n-heptane, ethyl oleate or 1-octanol) phase ( Figure 5). Further, the mutual solubility (here expressed as the water volume fraction v w , present in the water-rich and organic-rich phases) was determined F I G U R E 3 Picture of the lab-scale ACPI prototype as constructed described within this work.

F I G U R E 4
Scheme for collection and analysis of samples ensuring monitoring of emulsion stability and stream composition during the continuous ACPI process.
F I G U R E 5 (Top) Density ρ and viscosity η for the water-rich (blue squares) and the organic-rich (yellow circles) phases of the equilibrated binary 0.05 M phosphate buffer/organic systems at 25 • C and 1 bar. (Bottom, left) Mutual solubility of various organic compounds and water (left) at 25 • C and 1 bar. Data for water volume fraction v w in water-rich (blue squares) and organic-rich (yellow circles) phases were taken from literature: n-heptane, [25] ethyl oleate [26] , 1-octanol. [27] (Bottom, right) Interfacial tension γ ow between 0.05 M phosphate buffer and respective organic phase (grey circles) measured at 25 • C and 1 bar using pendant drop tensiometry. [5] ( Figure 5) for all binary base systems considered within this work. The impact of the low buffer concentration was found to be negligible and thus only pure water was considered for this data. and inverted emulsion in mixer). It is known from different studies, that the phase inversion behavior and success is highly affected by viscosity and interfacial tension. [24,28,29] Norato et al. [24] indicated that the ambivalence region, which separates the region of o/w-from the region of w/o-emulsion type, widens, if: (1) The viscosity of disperse increases and/or (2) the interfacial tension between the phases decreases. For the emulsions investigated with this work, widest ambivalence region and therefore the highest volumetric phase ratio needed for CPI is thus to be expected for emulsion containing 1-octanol as organic phase, whereas emulsions containing n-heptane are expected to result in the narrowest ambivalence region and therefore the lowest volumetric phase ratio required. It must be recognized that these estimations are based only on the properties of the aqueous/organic base system, with the cells having a decisive effect on the phase inversion process.

3.2
Process settings for ACPI prototype

Determining initial operating flow rates
Characterizing the phase inversion process in the presence of cells is crucial to define operating conditions and a process window for the ACPI prototype. The critical volumetric phase ratios V o :V w , required for phase inversion of the bioprocess derived Pickering-type emulsions processed within this work (either stabilized by E. coli JM101 or P. TA B L E 1 Overview of critical volumetric phase ratios V o :V w needed for phase inversion of emulsions originating from biphasic biocatalysis, which were determined experimentally in a previous work. [5] Biocatalyst putida KT2440 cells) to occur, were determined experimentally in a previous work [5] and are listed in Table 1.
Based on these values, volumetric phase ratios used within the ACPI process were chosen to be 10% to 20% higher than the critical volu-

Effect of flow rate on separation efficiency in the settler
The separation efficiency in the settler was investigated for flow rates between 60 and 120 ml min −1 . As shown in Figure 6  200 -300 μm by increasing the amount of organic phase. Median drop diameters in the inverted w/o-type emulsion are expected to be smaller. [16] Comparing drop settling velocities of water phase within the settler with the superficial velocities of organic phase (Table 2) Prior to steady state of continuous operation, a start-up routine (duration around 20 min), is performed, ensuring an inverted emulsion state in the mixer and phase separation in the settler. Therefore, depending on the volumetric phase ratio set for the ACPI process, the mixer is filled with stable Pickering-type emulsion and pure organic phase to a filling level of 500 ml. Afterwards, intensive mixing for about 2 min is applied to achieve the initial inverted state, which is monitored measuring electrical conductivity κ. When the κ value is near 0 mS cm −1 , inverted emulsion is pumped to the settler, where phase separation can be observed immediately. After complete separation, which is achieved within minutes, the process can be switched into its continuous (steady-state) mode.

Measurement and control concept
Monitoring and control of the fully automated prototype was ensured using LABVIEW. Each vessel within the process is therefore equipped with various sensors as described above. A possible process perturbation can result from fluctuating in the volumetric phase ratio of the emulsion feed (caused by inhomogeneous mixing or unequal amounts of phases in the upstream biocatalysis). To monitor the state of the feed emulsion, electrical conductivity κ is measured, and value is compared to the reference of κ at V o :V w = 1:1 (Table 3) Figure 8). If the increase in flow rate is too high, aqueous phase is rapidly pumped out of the settler, erroneously indicating the system to be in target state again. Therefore, the maximum increase is set to 50% (black circles in

Steady state processing
The measurement and control concept was applied to process var- inversion occurs [5] are included in Figure 9 as solid lines and serve as the cell wettability, measured in previous work. [5] Within the emulsion, typically two (stabilizing) cell fractions exist, exhibiting different wetting characteristics. [5,30] For the emulsion E. coli JM101/1octanol/Riesenberg the first cell fraction stabilize o/w-emulsion type but, as three-phase contact angle is lower 50.7 • no phase inversion to w/o-type is possible, whereas the second cell fraction of these emulsion preferentially stabilize w/o-type emulsions. Results of the conductivity measurements ( Table 3) show, that a significant values of κ (5.68 ± 0.17 mS cm −1 ) was detected, indicating o/w-emulsion type. Assuming the initial emulsion to be o/w-type, phase inversion by addition of organic phase result in w/o-type to be present the mixer, confirmed by significant decrease of κ (to about 2 mS cm −1 ) ( Figure 9).
As due to their wetting characteristics (Θ ow = 94 • [5] ) second fraction cells can also stabilize w/o-emulsion type, no phase separation is achieved. Same results were obtained using volumetric phase ratios

CONCLUSION
Based on the basic idea, described in the work of Glonke et al., [13] a fully automated prototype was designed and constructed, enabling efficient phase separation of long-term stable Pickering-type emulsions, based on the phenomenon of CPI in a mixer-settler set-up. ACPI ensures a wide applicability to different bioprocess-derived emulsions, requiring no other additives or co-solvents than originally used for biocatalysis. Therefore, ACPI is an innovative and universal tool overcoming the limitations of the drawbacks in classical downstream processing concepts used in state-of-the-art processing of bioprocess-derived Pickering-type emulsions.