Improved Production and In Situ Recovery of Sesquiterpene (+)-Zizaene from Metabolically-Engineered E. coli

The sesquiterpene (+)-zizaene is the direct precursor of khusimol, the main fragrant compound of the vetiver essential oil from Chrysopogon zizanioides and used in nearly 20% of men’s fine perfumery. The biotechnological production of such fragrant sesquiterpenes is a promising alternative towards sustainability; nevertheless, product recovery from fermentation is one of the main constraints. In an effort to improve the (+)-zizaene recovery from a metabolically-engineered Escherichia coli, we developed an integrated bioprocess by coupling fermentation and (+)-zizaene recovery using adsorber extractants. Initially, (+)-zizaene volatilization was confirmed from cultivations with no extractants but application of liquid–liquid phase partitioning cultivation (LLPPC) improved (+)-zizaene recovery nearly 4-fold. Furthermore, solid–liquid phase partitioning cultivation (SLPPC) was evaluated by screening polymeric adsorbers, where Diaion HP20 reached the highest recovery. Bioprocess was scaled up to 2 L bioreactors and in situ recovery configurations integrated to fermentation were evaluated. External recovery configuration was performed with an expanded bed adsorption column and improved (+)-zizaene titers 2.5-fold higher than LLPPC. Moreover, internal recovery configuration (IRC) further enhanced the (+)-zizaene titers 2.2-fold, whereas adsorption velocity was determined as critical parameter for recovery efficiency. Consequently, IRC improved the (+)-zizaene titer 8.4-fold and productivity 3-fold from our last report, achieving a (+)-zizaene titer of 211.13 mg L−1 and productivity of 3.2 mg L−1 h−1. This study provides further knowledge for integration of terpene bioprocesses by in situ product recovery, which could be applied for many terpene studies towards the industrialization of fragrant molecules.


Introduction
The biotechnological production of chemicals by engineered microorganisms is a potential alternative for the production of terpenes from renewable resources [1]. Recent advances in metabolic engineering have made possible the production of terpenes by microbial platforms at economically-feasible titers (over grams of terpene per liter of broth), reaching the industrial scale, such as artemisinin, β-farnesene, and squalane [2][3][4]. Fragrant sesquiterpenes used in the cosmetic industry are potential candidates to be produced by biotechnological systems, such as the sesquiterpenes contained in the vetiver essential oil (VEO) from the grass Ch. zizanioides. VEO is an important component for the formulation of cosmetics and it has been used in nearly 36% of Western perfumes and 20% of men's fragrances [5] with a total world production of 300−350 tons per year [6]. VEO is composed of a mixture of sesquiterpenes and their hydroxylated derivatives, with a characteristic Figure 1. The mechanism for the in situ recovery of terpenes produced from metabolically engineered E. coli. (A) Liquid-liquid phase partitioning cultivation is carried out by liquid extractants (organic solvents), which extract the terpenes by hydrophobic interactions. (B) Solid-liquid phase partitioning cultivation utilizes solid extractants (adsorbers) and recovers the terpenes by adsorption. The tripartite efflux pump AcrAB-TolC is used as an example of a secretion system for hydrophobic molecules and its components are properly described.
In our previous report, we demonstrated the production of (+)-zizaene by a metabolically-engineered E. coli and optimized the fermentation conditions at shake flask scale, comprising the induction, media, pH, and growth temperature, reaching a (+)-zizaene titer of 25.09 mg L −1 and a productivity of 1.05 mg L −1 [10]. However, the scale-up for the bioprocess of (+)-zizaene The mechanism for the in situ recovery of terpenes produced from metabolically engineered E. coli. (A) Liquid-liquid phase partitioning cultivation is carried out by liquid extractants (organic solvents), which extract the terpenes by hydrophobic interactions. (B) Solid-liquid phase partitioning cultivation utilizes solid extractants (adsorbers) and recovers the terpenes by adsorption. The tripartite efflux pump AcrAB-TolC is used as an example of a secretion system for hydrophobic molecules and its components are properly described.
In our previous report, we demonstrated the production of (+)-zizaene by a metabolicallyengineered E. coli and optimized the fermentation conditions at shake flask scale, comprising the induction, media, pH, and growth temperature, reaching a (+)-zizaene titer of 25.09 mg L −1 and a productivity of 1.05 mg L −1 [10]. However, the scale-up for the bioprocess of (+)-zizaene requires a suitable ISPR configuration to circumvent product loss and further improve the recovery of (+)-zizaene.
In this study, we developed an integrated ISPR configuration for the microbial production of (+)-zizaene at a 2 L bioreactor scale with adsorber extractants in order to improve the (+)-zizaene recovery. For that, the product loss by volatilization was analyzed and the (+)-zizaene recovery was improved by using LLPPC at shake flask scale. Furthermore, distinct polymeric adsorbers were evaluated, in terms of selectivity and recovery ratio, as potential extractants to be applied at bioreactor scale. The desorption process was studied by a comparative assessment of organic solvents. The bioprocess was scaled up to 2 L bioreactors using the fed-batch culture technique and three ISPR configurations by direct contact mode were tested: ERC, IRC, and IRC with gas stripping. As a result, the (+)-zizaene titers and productivity were improved significantly.

Product Volatilization Measurements and (+)-Zizaene Recovery by LLPPC
In our previous report, we demonstrated the microbial production of (+)-zizaene by engineering the MEV pathway and the ZS synthase in a multi-plasmid E. coli strain [9]. To further improve the production of (+)-zizaene, the bioprocess development requires the scale-up to stirred-tank bioreactors to reach HCDC, and an efficient downstream procedure. However, the recovery of terpenes involves special considerations due to their physicochemical properties. For the production of (+)-zizaene, loss of product could be expected during cultivation due to volatilization or microbial degradation. Moreover, toxicity to the E. coli cells could occur because of (+)-zizaene accumulation in the fermentation.
Initially, we analyzed the loss of (+)-zizaene at shake flask scale without extractants by cultivating an induced E. coli TZS+MevZS strain, grown for 24 h. After removing the cells from the culture broth, (+)-zizaene measurements were taken time-wise from cell-free media. As a result, half of the (+)-zizaene amount was volatilized after 1 h and only traces were detected after 4 h ( Figure 2). This demonstrates that, as for most terpenes, volatilization is a major constraint for the (+)-zizaene production. Eventually, the application of ISPR could provide solutions for these shortcomings, taking advantage of the hydrophobicity of (+)-zizaene (log P = 5.10), which could be extracted simultaneously during cultivation whether by liquid-or solid-phase recovery. requires a suitable ISPR configuration to circumvent product loss and further improve the recovery of (+)-zizaene.
In this study, we developed an integrated ISPR configuration for the microbial production of (+)-zizaene at a 2 L bioreactor scale with adsorber extractants in order to improve the (+)-zizaene recovery. For that, the product loss by volatilization was analyzed and the (+)-zizaene recovery was improved by using LLPPC at shake flask scale. Furthermore, distinct polymeric adsorbers were evaluated, in terms of selectivity and recovery ratio, as potential extractants to be applied at bioreactor scale. The desorption process was studied by a comparative assessment of organic solvents. The bioprocess was scaled up to 2 L bioreactors using the fed-batch culture technique and three ISPR configurations by direct contact mode were tested: ERC, IRC, and IRC with gas stripping. As a result, the (+)-zizaene titers and productivity were improved significantly.

Product Volatilization Measurements and (+)-Zizaene Recovery by LLPPC
In our previous report, we demonstrated the microbial production of (+)-zizaene by engineering the MEV pathway and the ZS synthase in a multi-plasmid E. coli strain [9]. To further improve the production of (+)-zizaene, the bioprocess development requires the scale-up to stirred-tank bioreactors to reach HCDC, and an efficient downstream procedure. However, the recovery of terpenes involves special considerations due to their physicochemical properties. For the production of (+)-zizaene, loss of product could be expected during cultivation due to volatilization or microbial degradation. Moreover, toxicity to the E. coli cells could occur because of (+)-zizaene accumulation in the fermentation.
Initially, we analyzed the loss of (+)-zizaene at shake flask scale without extractants by cultivating an induced E. coli TZS+MevZS strain, grown for 24 h. After removing the cells from the culture broth, (+)-zizaene measurements were taken time-wise from cell-free media. As a result, half of the (+)-zizaene amount was volatilized after 1 h and only traces were detected after 4 h ( Figure 2). This demonstrates that, as for most terpenes, volatilization is a major constraint for the (+)-zizaene production. Eventually, the application of ISPR could provide solutions for these shortcomings, taking advantage of the hydrophobicity of (+)-zizaene (log P = 5.10), which could be extracted simultaneously during cultivation whether by liquid-or solid-phase recovery.  To prove these hypotheses, a comparative test comprising a LLPPC with a solvent overlay and a negative control (without extractant) was performed with the E. coli TZS+MevZS strain at shake flask scale. As shown in Figure 3A, the (+)-zizaene production on the LLPPC and negative control tests followed similar kinetics, reaching the maximum peak at 48 h and dropping afterward. The LLPPC was nearly 4-fold higher when compared to that of the negative control at the highest production peak. Therefore, the loss of (+)-zizaene was estimated at nearly 27 mg L −1 due to volatilization at 48 h. Similarly, cell growth was higher on the LLPPC (OD 600 5.7 and biomass 2.1 g DCW L −1 ) than the negative control (OD 600 5.1 and biomass 1.7 g DCW L −1 ) at 48 h, suggesting a toxic effect due to (+)-zizaene accumulation in the control tests ( Figure 3B). To prove these hypotheses, a comparative test comprising a LLPPC with a solvent overlay and a negative control (without extractant) was performed with the E. coli TZS+MevZS strain at shake flask scale. As shown in Figure 3a, the (+)-zizaene production on the LLPPC and negative control tests followed similar kinetics, reaching the maximum peak at 48 h and dropping afterward. The LLPPC was nearly 4-fold higher when compared to that of the negative control at the highest production peak. Therefore, the loss of (+)-zizaene was estimated at nearly 27 mg L -1 due to volatilization at 48 h. Similarly, cell growth was higher on the LLPPC (OD600 5.7 and biomass 2.1 gDCW L -1 ) than the negative control (OD600 5.1 and biomass 1.7 gDCW L -1 ) at 48 h, suggesting a toxic effect due to (+)-zizaene accumulation in the control tests (Figure 3b

Screening of Polymeric Adsorbers for SLPPC
Although LLPPC has been used widely for isoprenoid recovery, SLPPC with solid extractants can be advantageous towards the scale-up of bioprocesses due to the following features: reusability, bio-compatibility, cost reduction of organic solvents, non-emulsion formation, and simple separation from the aqueous phase [15,37,38]. Thus, an in vivo adsorber screening analysis was carried out based on Halka [11] at shake flask scale. The tested polymeric adsorbers were chosen according to their affinity to adsorb hydrophobic molecules, as demonstrated in previous terpene recovery studies [25,39]. A negative control without extractants and a LLPPC control were also included.
As a result, the adsorbers showed similar product selectivity, where the terpene profile for all the tested adsorbers by GC-MS showed approximately a product ratio of 90% of (+)-zizaene, 9.5% of β-acoradiene, and traces of hydrocarbons ( Figure S1, S2).
Besides the Amberlite IRA400 Cl test, the rest of the adsorber tests recovered most of the (+)-zizaene amounts from the adsorbers (75.2%-92.5%), followed by the cell-free media (4.5%-18.7%) and in lower amounts from the cells (3.1%-6.1%; Table S2). As expected, only low (+)-zizaene amounts were measured from cell-free media by the negative control, confirming a loss of (+)-zizaene by volatilization. Besides, the highest amount of insoluble (+)-zizaene protein (inclusion bodies) was observed on the negative control, whereas most of the tested adsorbers and LLPPC showed low amounts of inclusion bodies ( Figure S3).

Screening of Polymeric Adsorbers for SLPPC
Although LLPPC has been used widely for isoprenoid recovery, SLPPC with solid extractants can be advantageous towards the scale-up of bioprocesses due to the following features: reusability, bio-compatibility, cost reduction of organic solvents, non-emulsion formation, and simple separation from the aqueous phase [15,37,38]. Thus, an in vivo adsorber screening analysis was carried out based on Halka [11] at shake flask scale. The tested polymeric adsorbers were chosen according to their affinity to adsorb hydrophobic molecules, as demonstrated in previous terpene recovery studies [25,39]. A negative control without extractants and a LLPPC control were also included.
As a result, the adsorbers showed similar product selectivity, where the terpene profile for all the tested adsorbers by GC-MS showed approximately a product ratio of 90% of (+)-zizaene, 9.5% of β-acoradiene, and traces of hydrocarbons ( Figures S1 and S2).
All of the tested adsorbers obtained similar cell growth between OD600 3.4-3.9 and biomass 0.96-1.17 gDCW L -1 from the E. coli TZS+MevZS strain (Figure 4b). On the other hand, the negative control (without extractant) had a lower cell growth when compared to all the other tests, with an OD600 of 2.7 and biomass of 0.7 gDCW L -1 , suggesting a toxic effect from the (+)-zizaene amounts in the culture broth. Because of the high recovery performance from the Diaion HP20 between adsorbers and comparable results with the LLPPC control, the Diaion HP20 was selected for further cultivations.

Assessment of Organic Solvents for the Desorption of (+)-Zizaene
After the screening of adsorbers for the SLPPC, an evaluation was done to analyze the elution performance of different organic solvents, which ideally should have a high partition coefficient and high selectivity towards (+)-zizaene [19]. Consequently, distinct solvents with high log P values were tested comprising decane, dodecane, pentane, ethyl acetate, isopropanol, isooctane, and acetonitrile (Table S1).
As evidenced in Figure 5, the organic solvent tests displayed significant desorption differences, with (+)-zizaene titers between 15-25 mg L −1 approximately. The eluents isooctane, decane, and ethyl acetate recovered the highest (+)-zizaene titers (25.7, 24.7, and 24.0 mg L -1 , respectively), followed by dodecane, pentane, acetonitrile, and isopropanol. As a consequence, isooctane was selected as the elution solvent and further used for scale-up tests. Besides the Amberlite IRA400 Cl test, the rest of the adsorber tests recovered most of the (+)-zizaene amounts from the adsorbers (75.2%-92.5%), followed by the cell-free media (4.5%-18.7%) and in lower amounts from the cells (3.1%-6.1%; Table S2). As expected, only low (+)-zizaene amounts were measured from cell-free media by the negative control, confirming a loss of (+)-zizaene by volatilization. Besides, the highest amount of insoluble (+)-zizaene protein (inclusion bodies) was observed on the negative control, whereas most of the tested adsorbers and LLPPC showed low amounts of inclusion bodies ( Figure S3).
All of the tested adsorbers obtained similar cell growth between OD 600 3.4-3.9 and biomass 0.96-1.17 g DCW L −1 from the E. coli TZS+MevZS strain ( Figure 4B). On the other hand, the negative control (without extractant) had a lower cell growth when compared to all the other tests, with an OD 600 of 2.7 and biomass of 0.7 g DCW L −1 , suggesting a toxic effect from the (+)-zizaene amounts in the culture broth. Because of the high recovery performance from the Diaion HP20 between adsorbers and comparable results with the LLPPC control, the Diaion HP20 was selected for further cultivations.

Assessment of Organic Solvents for the Desorption of (+)-Zizaene
After the screening of adsorbers for the SLPPC, an evaluation was done to analyze the elution performance of different organic solvents, which ideally should have a high partition coefficient and high selectivity towards (+)-zizaene [19]. Consequently, distinct solvents with high log P values were tested comprising decane, dodecane, pentane, ethyl acetate, isopropanol, isooctane, and acetonitrile (Table S1).

Integration of In Situ Recovery of (+)-Zizaene to Fermentation at Bioreactor Scale
The fermentation was scaled up to 2 L stirred-tank bioreactors using the fed-batch cultivation method by feeding glucose continuously to reach HCDC. Because the use of liquid extractants is challenging for industrial-scale bioreactors [15,33], we tested the (+)-zizaene recovery with solid extractants. Thus, the external, internal, and internal with gas stripping in situ recovery configurations ( Figure 6) were integrated to the fermentation with the best performing adsorber (Diaion HP20) and eluent (isooctane), in order to maximize product recovery and avoid cell toxicity, which could be more problematic at bioreactor scale due to the higher production of (+)-zizaene. The Diaion HP20 adsorber was used at 50 g L -1 , whereas higher amounts resulted in mixing problems and reading disturbances from the probes of the bioreactors. The optimal fermentation conditions for (+)-zizaene production at shake flask scale were previously determined [9] and used as the initial point for the bioreactor cultivations (ADM, pH 7.0), including induction at 20 °C to avoid formation of inclusion bodies from the ZS protein and to reduce volatilization of (+)-zizaene. Early induction was applied in all cultivations (OD600 5-7) because late induction (OD600 > 15) was

Integration of In Situ Recovery of (+)-Zizaene to Fermentation at Bioreactor Scale
The fermentation was scaled up to 2 L stirred-tank bioreactors using the fed-batch cultivation method by feeding glucose continuously to reach HCDC. Because the use of liquid extractants is challenging for industrial-scale bioreactors [15,33], we tested the (+)-zizaene recovery with solid extractants. Thus, the external, internal, and internal with gas stripping in situ recovery configurations ( Figure 6) were integrated to the fermentation with the best performing adsorber (Diaion HP20) and eluent (isooctane), in order to maximize product recovery and avoid cell toxicity, which could be more problematic at bioreactor scale due to the higher production of (+)-zizaene.

Integration of In Situ Recovery of (+)-Zizaene to Fermentation at Bioreactor Scale
The fermentation was scaled up to 2 L stirred-tank bioreactors using the fed-batch cultivation method by feeding glucose continuously to reach HCDC. Because the use of liquid extractants is challenging for industrial-scale bioreactors [15,33], we tested the (+)-zizaene recovery with solid extractants. Thus, the external, internal, and internal with gas stripping in situ recovery configurations ( Figure 6) were integrated to the fermentation with the best performing adsorber (Diaion HP20) and eluent (isooctane), in order to maximize product recovery and avoid cell toxicity, which could be more problematic at bioreactor scale due to the higher production of (+)-zizaene. The Diaion HP20 adsorber was used at 50 g L -1 , whereas higher amounts resulted in mixing problems and reading disturbances from the probes of the bioreactors. The optimal fermentation conditions for (+)-zizaene production at shake flask scale were previously determined [9] and used as the initial point for the bioreactor cultivations (ADM, pH 7.0), including induction at 20 °C to avoid formation of inclusion bodies from the ZS protein and to reduce volatilization of (+)-zizaene. Early induction was applied in all cultivations (OD600 5-7) because late induction (OD600 > 15) was The Diaion HP20 adsorber was used at 50 g L −1 , whereas higher amounts resulted in mixing problems and reading disturbances from the probes of the bioreactors. The optimal fermentation conditions for (+)-zizaene production at shake flask scale were previously determined [9] and used as the initial point for the bioreactor cultivations (ADM, pH 7.0), including induction at 20 • C to avoid formation of inclusion bodies from the ZS protein and to reduce volatilization of (+)-zizaene. Early induction was applied in all cultivations (OD 600 5-7) because late induction (OD 600 > 15) was proven deleterious for the (+)-zizaene production, where most of the ZS protein was overexpressed as inclusion bodies (data not shown).

Bioreactor Cultivation with an Integrated ERC
The ERC was analyzed because external loops can facilitate product removal from large bioreactors and enable semi-continuous product recovery [15,38]. Moreover, the EBA chromatography was preferred as the extractant unit over the conventional packed-bed adsorption chromatography, because it allows operations at higher flow rates [40]. For this purpose, a stainless steel EBA column was built with the feature of a sampling port to ease the process monitoring and adsorber renovation ( Figure 6A). The EBA was loaded with 75 g of Diaion HP20 and exchanged with 75 g of fresh resin every 24 h of operation.
The fed-batch stage started at 7 h of growth by lowering the temperature to 20 • C, and after 1 h the culture was induced with 0.5 mM ITPG ( Figure 7B). Afterwards, the recirculation through the external loop (EBA) was initiated at a flow rate of 150 ml min −1 . Thereafter, the production of (+)-zizaene and cell growth increased proportionally from 12 to 72 h of culture ( Figure 7A), showing a coupling between both variables with a Pearson coefficient (PI) of 0.99. After 72 h of growth, the maximum production peak was observed, reaching a (+)-zizaene titer of 93.4 mg L −1 and an OD 600 of 43.0. Therefore, the bioprocess improved the (+)-zizaene titer 2.5-fold and the cell growth 7.5-fold when compared to that of the LLPPC at shake flask scale. At the end of the cultivation (72 h), the recovery ratio (94.6%) from the adsorbers was similar to that obtained from the LLPPC test from Section 2.2 (94.4%), followed from the cell-free media (3.0%) and from the cells (2.4%). proven deleterious for the (+)-zizaene production, where most of the ZS protein was overexpressed as inclusion bodies (data not shown).

Bioreactor Cultivation with an Integrated ERC
The ERC was analyzed because external loops can facilitate product removal from large bioreactors and enable semi-continuous product recovery [15,38]. Moreover, the EBA chromatography was preferred as the extractant unit over the conventional packed-bed adsorption chromatography, because it allows operations at higher flow rates [40]. For this purpose, a stainless steel EBA column was built with the feature of a sampling port to ease the process monitoring and adsorber renovation ( Figure 6A). The EBA was loaded with 75 g of Diaion HP20 and exchanged with 75 g of fresh resin every 24 h of operation.
The fed-batch stage started at 7 h of growth by lowering the temperature to 20 °C, and after 1 h the culture was induced with 0.5 mM ITPG (Figure 7b). Afterwards, the recirculation through the external loop (EBA) was initiated at a flow rate of 150 ml min -1 . Thereafter, the production of (+)-zizaene and cell growth increased proportionally from 12 to 72 h of culture (Figure 7a), showing a coupling between both variables with a Pearson coefficient (PI) of 0.99. After 72 h of growth, the maximum production peak was observed, reaching a (+)-zizaene titer of 93.4 mg L -1 and an OD600 of 43.0. Therefore, the bioprocess improved the (+)-zizaene titer 2.5-fold and the cell growth 7.5-fold when compared to that of the LLPPC at shake flask scale. At the end of the cultivation (72 h), the recovery ratio (94.6%) from the adsorbers was similar to that obtained from the LLPPC test from Section 2.2. (94.4%), followed from the cell-free media (3.0%) and from the cells (2.4%).

Bioreactor Cultivation with an Integrated IRC
The IRC was evaluated as an alternative to improve the (+)-zizaene recovery further, whereas the configuration was designed to recover (+)-zizaene directly from the culture broth and improve the adsorption velocity ( Figure 6B). To maintain the same amount of resin as the ERC, the bioreactor was loaded with 75 g of adsorbers.
Similar to the ERC, a positive correlation between (+)-zizaene levels and cell growth was observed from 12 to 72 h (PI = 0.98) (Figure 8b). Moreover, the production of (+)-zizaene increased after 24 h, achieving the highest (+)-zizaene titer of 207.8 mg L -1 and cell growth (OD600 48.9 and biomass 10.3 gDCW L -1 ) at the end of the fermentation (Figure 8a). Thus, the IRC improved the (+)-zizaene titers 2.2-fold when compared to that of the ERC. In addition, the (+)-zizaene recovery ratio from adsorbers was improved to 98.4% when compared to the ERC and reduced the amounts of (+)-zizaene from cell-free media (0.9%) and cells (0.7%).

Bioreactor Cultivation with an Integrated IRC
The IRC was evaluated as an alternative to improve the (+)-zizaene recovery further, whereas the configuration was designed to recover (+)-zizaene directly from the culture broth and improve the adsorption velocity ( Figure 6B). To maintain the same amount of resin as the ERC, the bioreactor was loaded with 75 g of adsorbers.
Similar to the ERC, a positive correlation between (+)-zizaene levels and cell growth was observed from 12 to 72 h (PI = 0.98) ( Figure 8B). Moreover, the production of (+)-zizaene increased after 24 h, achieving the highest (+)-zizaene titer of 207.8 mg L −1 and cell growth (OD 600 48.9 and biomass 10.3 g DCW L −1 ) at the end of the fermentation ( Figure 8A). Thus, the IRC improved the (+)-zizaene titers 2.2-fold when compared to that of the ERC. In addition, the (+)-zizaene recovery ratio from adsorbers was improved to 98.4% when compared to the ERC and reduced the amounts of (+)-zizaene from cell-free media (0.9%) and cells (0.7%).

Bioreactor Cultivation with an Integrated IRC+GS
As discussed in Section 2.1, there was a loss of (+)-zizaene by volatilization at shake flask scale, which could be even higher at bioreactor scale due to the gassing of air through the bioreactor vessel. Therefore, the (+)-zizaene recovery from the off-gas of the bioreactor could be considered as an additional recovery source to enhance the accumulative (+)-zizaene recovery from the fermentation. For that, a variant from the IRC was performed with the addition of a column loaded with 75 g L −1 of adsorbers and installed in the off-gas of the bioreactor ( Figure 6C).
Similar to the other bioreactor configurations, the correlation between cell growth and (+)-zizaene production on the IRC+GS was positive with PI = 95%, indicating a coupling between both variables. The results were similar to the IRC test, where after 72 h of culture, the maximum cell growth (OD600 45.3 and biomass 9.7 gDCW L -1 ) and (+)-zizaene production (203.4 mg L −1 ) were reached (Figure 9a,b). The (+)-zizaene amounts were not detected on the ethanol trap, demonstrating efficient adsorption of (+)-zizaene from the resins in the off-gas column with a (+)-zizaene recovery ratio of 0.6%. These low amounts of (+)-zizaene detected on the off-gas suggest an efficient (+)-zizaene trapping from the adsorbers in the culture broth, showing a recovery ratio from adsorbers of 97.6% and a low recovery ratio of 0.9% from both cell-free media and cells, similar to the IRC.

Accumulative (+)-Zizaene Production from the Bioreactor Configurations
The contribution to the accumulative (+)-zizaene amount from the IRC+GS was low (less than 1%). In consequence, the IRC and the IRC+GS showed similar accumulative (+)-zizaene levels during the course of the fermentation, whereas the difference between the ERC and both IRCs increased after 12 h of cultivation ( Figure 10).

Bioreactor Cultivation with an Integrated IRC+GS
As discussed in Section 2.1, there was a loss of (+)-zizaene by volatilization at shake flask scale, which could be even higher at bioreactor scale due to the gassing of air through the bioreactor vessel. Therefore, the (+)-zizaene recovery from the off-gas of the bioreactor could be considered as an additional recovery source to enhance the accumulative (+)-zizaene recovery from the fermentation. For that, a variant from the IRC was performed with the addition of a column loaded with 75 g L −1 of adsorbers and installed in the off-gas of the bioreactor ( Figure 6C).
Similar to the other bioreactor configurations, the correlation between cell growth and (+)-zizaene production on the IRC+GS was positive with PI = 95%, indicating a coupling between both variables. The results were similar to the IRC test, where after 72 h of culture, the maximum cell growth (OD 600 45.3 and biomass 9.7 g DCW L −1 ) and (+)-zizaene production (203.4 mg L −1 ) were reached ( Figure 9A,B). The (+)-zizaene amounts were not detected on the ethanol trap, demonstrating efficient adsorption of (+)-zizaene from the resins in the off-gas column with a (+)-zizaene recovery ratio of 0.6%. These low amounts of (+)-zizaene detected on the off-gas suggest an efficient (+)-zizaene trapping from the adsorbers in the culture broth, showing a recovery ratio from adsorbers of 97.6% and a low recovery ratio of 0.9% from both cell-free media and cells, similar to the IRC.

Bioreactor Cultivation with an Integrated IRC+GS
As discussed in Section 2.1, there was a loss of (+)-zizaene by volatilization at shake flask scale, which could be even higher at bioreactor scale due to the gassing of air through the bioreactor vessel. Therefore, the (+)-zizaene recovery from the off-gas of the bioreactor could be considered as an additional recovery source to enhance the accumulative (+)-zizaene recovery from the fermentation. For that, a variant from the IRC was performed with the addition of a column loaded with 75 g L −1 of adsorbers and installed in the off-gas of the bioreactor ( Figure 6C).
Similar to the other bioreactor configurations, the correlation between cell growth and (+)-zizaene production on the IRC+GS was positive with PI = 95%, indicating a coupling between both variables. The results were similar to the IRC test, where after 72 h of culture, the maximum cell growth (OD600 45.3 and biomass 9.7 gDCW L -1 ) and (+)-zizaene production (203.4 mg L −1 ) were reached (Figure 9a,b). The (+)-zizaene amounts were not detected on the ethanol trap, demonstrating efficient adsorption of (+)-zizaene from the resins in the off-gas column with a (+)-zizaene recovery ratio of 0.6%. These low amounts of (+)-zizaene detected on the off-gas suggest an efficient (+)-zizaene trapping from the adsorbers in the culture broth, showing a recovery ratio from adsorbers of 97.6% and a low recovery ratio of 0.9% from both cell-free media and cells, similar to the IRC.

Accumulative (+)-Zizaene Production from the Bioreactor Configurations
The contribution to the accumulative (+)-zizaene amount from the IRC+GS was low (less than 1%). In consequence, the IRC and the IRC+GS showed similar accumulative (+)-zizaene levels during the course of the fermentation, whereas the difference between the ERC and both IRCs increased after 12 h of cultivation ( Figure 10).

Accumulative (+)-Zizaene Production from the Bioreactor Configurations
The contribution to the accumulative (+)-zizaene amount from the IRC+GS was low (less than 1%). In consequence, the IRC and the IRC+GS showed similar accumulative (+)-zizaene levels during the course of the fermentation, whereas the difference between the ERC and both IRCs increased after 12 h of cultivation ( Figure 10). Table 1. The accumulative (+)-zizaene titer and productivity between the IRC and IRC+GS were more than 2-fold higher than the ERC at 72 h of cultivation. Although differences between cell growth and product yield (Y X/S ) were not so drastic between IRCs and ERC, the production of soluble ZS protein from the IRCs was roughly 4-fold higher when compared to the ERC, suggesting a relation between the soluble ZS protein and the (+)-zizaene levels. In consequence, the best configuration for the in situ recovery of (+)-zizaene at bioreactor scale was the IRC; achieving an accumulative (+)-zizaene titer of 211.13 mg L −1 and productivity of 3.2 mg L −1 h −1 . Moreover, these results improved the (+)-zizaene titers 8.4-fold and productivity 3-fold when compared to those from our last report [10]. The comparison of performance between bioreactor configurations is summarized in Table 1. The accumulative (+)-zizaene titer and productivity between the IRC and IRC+GS were more than 2-fold higher than the ERC at 72 h of cultivation. Although differences between cell growth and product yield (YX/S) were not so drastic between IRCs and ERC, the production of soluble ZS protein from the IRCs was roughly 4-fold higher when compared to the ERC, suggesting a relation between the soluble ZS protein and the (+)-zizaene levels. In consequence, the best configuration for the in situ recovery of (+)-zizaene at bioreactor scale was the IRC; achieving an accumulative (+)-zizaene titer of 211.13 mg L −1 and productivity of 3.2 mg L −1 h −1 . Moreover, these results improved the (+)-zizaene titers 8.4-fold and productivity 3-fold when compared to those from our last report [10].

Discussion
For the microbial production of (+)-zizaene, significant product volatilization was observed when the recombinant E. coli strain was cultured without extractants, similar to that observed with the sesquiterpenes α-humulene [41]. Regardless of the low vapor pressure (5.1 kPa at 20 °C) and boiling point (288 °C) of (+)-zizaene, the volatilization of (+)-zizaene could be explained due to its low aqueous solubility (0.1289 mg L -1 [42]), which practically remains immiscible in the aqueous broth and tends to volatilize. As a solution, we tested the LLPPC at shake flask scale and improved the (+)-zizaene recovery and cell growth, reducing then the cell toxicity. Similar results were shown for the recovery of amorpha-4,11-diene, where the titers from the LLPPC tests were 8.5-fold higher than the controls without extractants [43]. Such improvements can be explained because the organic solvents extract the isoprenoids during culture due to hydrophobic interactions, and partition from the aqueous phase (culture broth) as demonstrated in many terpene studies [1,44,45].  94.60 ± 0.14 98.40 ± 0.21 97.60 ± 0.12 1 Y P/X : product/biomass yield. Y X/S : 2 Biomass/substrate yield. 3 Y P/S : product/substrate yield. Data correspond to the mean of three sample replicates with ± SD.

Discussion
For the microbial production of (+)-zizaene, significant product volatilization was observed when the recombinant E. coli strain was cultured without extractants, similar to that observed with the sesquiterpenes α-humulene [41]. Regardless of the low vapor pressure (5.1 kPa at 20 • C) and boiling point (288 • C) of (+)-zizaene, the volatilization of (+)-zizaene could be explained due to its low aqueous solubility (0.1289 mg L −1 [42]), which practically remains immiscible in the aqueous broth and tends to volatilize. As a solution, we tested the LLPPC at shake flask scale and improved the (+)-zizaene recovery and cell growth, reducing then the cell toxicity. Similar results were shown for the recovery of amorpha-4,11-diene, where the titers from the LLPPC tests were 8.5-fold higher than the controls without extractants [43]. Such improvements can be explained because the organic solvents extract the isoprenoids during culture due to hydrophobic interactions, and partition from the aqueous phase (culture broth) as demonstrated in many terpene studies [1,44,45].
Accordingly, the Diaion HP20 obtained the highest (+)-zizaene recovery ratio from all tested adsorbers. In comparison with the other adsorbers, Diaion HP20 has the largest pore size (290 Å) and pore volume (1.3 mL g −1 ) and it is used commonly for the adsorption of relatively large molecules such as small proteins. Although the sesquiterpene (+)-zizaene is not considered a large molecule (204.35 g mol −1 ), the large pores from the Diaion HP20 could possibly favor the (+)-zizaene adsorption. This idea could be supported due to the low recovery from the Lewatit 1064 MD, which has the lowest pore size of all the tested adsorbers (50 Å). Similar to our findings, the Diaion HP20 resin obtained the highest recovery ratio for the fragrant benzaldehyde (106.12 g mol −1 ) and L-phenylalanine (165.19 g mol −1 ) in bioreactor cultivations of Pycnoporus cinnabarinus [46], as well as for the prodigiosin-like red pigment (323 Da) from Serratia sp. KH-95 [47], whereas the Amberlite XAD16 adsorber (pore size 200 Å) showed the lowest recovery for both cases.
Additionally, all of the tested adsorbers demonstrated higher cell growth when compared to the negative control without extractants. Such results were expected since it is known that most of the adsorbers do not affect cell growth due to their synthetic polymeric composition [37,48]. In addition, the low cell growth from the negative control suggests a toxic effect due to (+)-zizaene accumulation, similar to the toxic effect of linalool and linalool oxides accumulation on the fermentation of Corynespora cassiicola [27]. Eventually, further tests are required to measure the toxicity threshold of (+)-zizaene in fermentations.
Concerning the screening of solvents for the desorption of (+)-zizaene, the results were as expected, where solvents with log P values higher than 0.73 eluted the highest sesquiterpene amounts due to their capability to trap hydrophobic compounds [15,19]. Although many studies used ethyl acetate [11,25] for terpene desorption, in our case we chose isooctane (log P = 3.80) because of the similar product recovery when compared to ethyl acetate but at a lower cost. In addition, isooctane has been proven suitable for the recovery of other terpenes such as limonene-1,2-diol produced by Rhodococcus erythropolis DCL14 [49]. Moreover, isooctane was used successfully for the extraction of (+)-zizaene on in vitro biotransformation reactions, as shown in our previous report [10].
To recover efficiently the (+)-zizaene at bioreactor scale, distinct ISPR configurations were integrated into the fermentation, using Diaion HP20 as adsorber and isooctane as elution solvent. Although these were chosen due to their recovery performance, other factors such as their low cost and ease for implementation were considered for the selection criteria.
Initially, the ERC improved the (+)-zizaene production 2.5-fold when compared to that of the LLPPC at shake flask tests. Eventually, the minimal growth by feeding glucose maintained the acetate levels at a minimum, allowing the fermentation to reach a higher cell growth (OD 600 of 43.0), and avoiding an overflow metabolism. Similar cell growth was also achieved by fed-batch fermentation of a metabolically-engineered E. coli strain for the production of (-)-α-bisabolol, fed with glycerol after 72 h of culture [14]. Besides, the external loop of the ERC allowed the semi-continuous product recovery, reaching (+)-zizaene titers of 93.4 mg L −1 . Such improvements have also been obtained by the use of ERC in other terpenes studies, such as monoterpene carvone (225 mg L −1 ) [37] and diterpene cembratriene-ol (78.9 mg L −1 ) [50].
The (+)-zizaene production was enhanced 2.2-fold further by the use of the IRC when compared to the ERC. Besides, the (+)-zizaene amounts from cell-free media from the IRC were lower than the ERC. This demonstrates a higher (+)-zizaene recovery ratio from the IRC (98.4%) than the ERC (94.6%), which resulted in higher cell growth and soluble ZS protein synthesis, suggesting cell toxicity from (+)-zizaene accumulation in the culture broth of the ERC. This could be explained due to the residence time of the adsorbers on the culture broth, whereas in the ERC, the adsorbers have less time in contact with the culture broth (recirculation rate 6 h −1 ) than in the IRC, in which it resides constantly. Similar results were obtained for the production of prodigiosin-like red pigment from Serratia sp. KH-95, where the ERC obtained lower amounts than the IRC due to the lower contact of the adsorbers in the culture broth [51]. Consequently, terpenes need to be recovered rapidly from the culture broth before becoming volatilized. This was confirmed by the IRC+GS, in which less than 1% of the (+)-zizaene amount was volatilized due to the efficient (+)-zizaene recovery from the adsorbers inside the vessel (97.6%). Hence, the adsorption velocity plays an important role in the ISPR of terpenes.
The IRC showed lower (+)-zizaene amounts from the cells than the ERC; demonstrating no correlation between cell growth and (+)-zizaene amount from cells. Possibly, the constant contact of the E. coli cells with the Diaion HP20 adsorbers from the IRC could improve the secretion and trapping of (+)-zizaene by an adsorption mechanism (chemisorption) as described in Figure 1. Similar behavior was observed in the production of a prodigiosin-like red pigment, where the compound was bound to the cell wall from Serratia sp. KH-95 and it was adsorbed towards the Diaion HP20 adsorbers, dispersed in the culture broth [51].

Strain and Pre-Cultures
The metabolically-engineered strain used in all the experiments of this research was the multi-plasmid E. coli Tuner TZS+MevZS strain, as described in our previous report [9].
All pre-cultures were prepared in 5 mL LB broth with 30 mg L −1 kanamycin and 34 mg L −1 chloramphenicol from glycerol stocks, and cultivated at 37 • C in a rotatory incubator at 150 rpm. For shake flask experiments, pre-cultures were grown overnight and inoculated to main cultures consisting of 35 mL of a modified Aparicio defined medium (ADM, [9]) with 5 g L −1 glucose in sealed glass-baffled shake flasks, to initiate at an OD 600 of 0.1, and grown with the same conditions as mentioned before. Induction was performed when cultures reached OD 600 0.6-0.8 by lowering the temperature to 20 • C and adding isopropyl-β-D-thiogalactoside (IPTG) to a final concentration of 0.5 mM.

Product Volatilization and LLPPC Experiments
Shake flask cultures without extractants and induced for 24 h were centrifuged for 20 min (10,000 × g at 4 • C) and the supernatant was filtered through a 0.2 µm filter. Cell-free broth was transferred to sterile shake flasks and incubated at 20 • C and 150 rpm. Samples were taken from the cell-free broth, further extracted and terpene products were measured via gas chromatography coupled with a flame ionization detector (GC-FID).
For the LLPPC evaluation, cultures were prepared as in Section 4.2 and 10% isooctane (v/v) was added promptly after the addition of IPTG. No extractant was added to the negative control. Cultures were grown for 72 h and every 12 h samples were taken for growth kinetics and terpene analytics.

Screening of Polymeric Adsorbers
The testing of adsorbers was performed using an in vivo method as described by Halka [11]. For that, distinct polymeric adsorbers ( Table 2) were conditioned by washing them with water, isopropanol, isooctane, and finally water. After autoclaving shake flasks with 50 g L −1 of the respective adsorbers, 35 mL of ADM was added and inoculated with pre-culture broth to an initial OD 600 of 0.1. Growth conditions and induction procedures were according to Section 4.2 and samples were analyzed after 24 h of induction. Terpene products were extracted from adsorbers, cell-free broth, and cells according to Section 4.7.3. A negative control without extractants was included, in which only cell-free broth and cells were analyzed. Identification of terpene products was carried out via gas chromatography with mass spectrometry (GC-MS) and quantification of (+)-zizaene was carried out via GC-FID. To measure the effect of the tested adsorbers on cell growth, the OD 600 and the dry cell weight (DCW) biomass were analyzed from the distinct resins. The (+)-zizaene recovery ratio from adsorbers was calculated as the (+)-zizaene titer from adsorbers between the accumulative (+)-zizaene titer (recovered from adsorbers, cells, and cell-free media). Data sets were analyzed by ANOVA according to Section 4.8. Table 2. Main properties of the polymeric adsorbers used for the in situ recovery of (+)-zizaene 1 .

Evaluation of Organic Solvents for the (+)-Zizaene Desorption
Organic solvents with a high log P value were evaluated for the desorption of (+)-zizaene comprising decane, dodecane, isooctane, ethyl acetate, isopropanol, acetonitrile, and pentane (Table  S1). The Diaion HP20 adsorbers were prepared as described in Section 4.4 and the tested organic solvents were used respectively for the conditioning of adsorbers.
Microbial cultures were prepared as described in Section 4.2 without extractants and after 48 h of cultivation, 2 mL of culture broth was transferred to sterile vials with 0.5 g of Diaion HP20 adsorbers. Afterwards, vials were incubated for 4 h at 20 • C and 150 rpm. Furthermore, the culture broth was discarded and adsorbers were extracted with the tested solvents as described in Section 4.7.3 (+)-Zizaene concentrations were measured via GC-FID and statistical analyses were carried out according to Section 4.8.

Bioreactor Cultivations with In Situ Recovery of (+)-Zizaene
Pre-cultures were prepared as described in Section 4.2, followed by a third pre-culture that consisted of a 100 mL ADM shake flask culture. After 12 h of cultivation, pre-cultures were inoculated to bioreactors to an OD 600 of 0.3-0.4.
2 L stirred-tank bioreactors (Biostat A+, Sartorius, Göttingen, Germany) were used for the cultivations, with 1.5 L working volume consisting of an ADM with 5 g L −1 glucose and respective antibiotics. The pH was controlled automatically with 1 M HCl and 25% NH 4 OH (v/v) solutions. The Diaion HP20 adsorbers were conditioned according to Section 4.4 with isooctane as the solvent.

IRC Bioreactor
The IRC was prepared with 50 g L −1 of Diaion HP20 dispersed in the culture broth, inside the bioreactor vessel. The IRC with gas stripping (IRC+GS) was carried out similarly to the IRC and a 200 mL column was added to the off-gas line loaded with 75 g of Diaion HP20, followed by an ethanol trap cooled with dry ice and two 0.2 µm sterile filters.

ERC Bioreactor
The ERC was configured without extractants inside the bioreactor vessel, and the culture broth was recirculated to an external recovery loop through a stainless steel EBA column of 200 mL inner volume with 75 g of Diaion HP20 as the fluidized bed. For that, a silicone tubing of 4 mm inner diameter (i.d.) was connected to the EBA column (4.6 cm i.d., 12 cm length for inner chamber) with outlets of 4 mm i.d. and a sampling port on the lateral side. The EBA column had two stainless steel meshes (6 cm diameter of 500 µm) on both terminal sides, fixed with gaskets to recirculate cells and media through the resins while retaining the adsorbers inside the column. The external loop operated continuously at a flow rate of 150 mL min −1 (recirculation rate = 6 h −1 ) by a SciLog Expert peristaltic pump (Wisconsin, USA).
For all bioreactor configurations, the batch cultivation settings were dissolved oxygen >30%, agitation 400-700 rpm, temperature 37 • C, pH 7.0, and gas flow rate 1.0 vvm. When the glucose was exhausted, the fed-batch stage was initiated by dropping the temperature to 20 • C and initiating the feeding. The fed-batch medium was composed of glucose 100 g L −1 , NaCl 1.2 g L −1 , CaCl 2 0.14 g L −1 , MgSO 4 ·7H 2 O 0.6 g L −1 , FeSO 4 ·7H 2 O 0.001 g L −1 , and CuSO 4 ·5H 2 O 0.001 g L −1 with respective antibiotics. The induction was performed after 1 h of feeding by adding 0.5 mM IPTG and in the case of the ERC, the recirculation towards the external loop was activated. Samples for growth kinetics, soluble ZS protein, and terpene analysis were properly taken, measured according to Section 4.7, and plotted with Origin 9.5.5. (Northampton, MA, USA).

Growth Kinetics Analysis
To assess the growth kinetics from cultivations, the cell growth was analyzed by measuring the optical density from fermentation samples at 600 nm using a Biochrom Libra S50 UV-Vis spectrophotometer (Cambridge, UK) and biomass by the dry cell weight method. Glucose consumption was measured from cell-free broth samples with the Biochemistry Analyzer YSI 2900 (Yellow Springs, Greene County, OH, USA). Acetate was measured via high-pressure liquid chromatography as described elsewhere [55].

Soluble ZS Protein Fraction Analysis
Broth samples were normalized to an OD 600 of 2.0 and extracted as described in our previous report [9]. Soluble ZS protein fractions were analyzed on a 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis with a calibration curve of bovine serum albumin. Quantification of soluble ZS protein was performed by a densitometric method [56,57], measuring the intensity of the ZS protein bands at 78 kDa from stained gel images by the GelAnalyzer 2010 (developed by Istvan Lazar).

Sample extraction
Samples for terpene analyses were extracted from distinct sources during cultivations. Samples from cell-free media, known also as supernatant, were prepared by transferring 2 mL of culture broth to 10 mL glass vials. After centrifugation for 5 min at 10,000× g, the supernatant was transferred to other glass vials and extracted vigorously thrice with 0.5 mL of isooctane. Organic phases were obtained by centrifugation and transferred to GC vials for (+)-zizaene measurements via GC-FID. Samples from cells were prepared similar to Section 4.7.2. After extracts were ultrasonicated, 300 µL of isooctane was added to 300 µL of cell extract and extracted vigorously. Organic phases were separated by centrifugation and transferred to GC vials for (+)-zizaene measurements via GC-FID.
Samples from adsorbers were extracted vigorously three times by transferring 300 mg of resins to 10 mL glass vials and adding 1 volume of isooctane (or tested solvent for Section 4.5). Organic phases were transferred to GC vials for further product identification via GC-MS and (+)-zizaene quantification via GC-FID. The quantification of (+)-zizaene was calculated as α-cedrene equivalents by a calibration curve of α-cedrene (standard grade) due to the lack of a commercial (+)-zizaene standard, as demonstrated in a previous report [10].

Statistical Analysis
Data from Sections 4.4 and 4.5 were analyzed by analysis of variance (ANOVA) and mean comparison tests to assess statistical differences. Adsorbers and organic solvents were used as factors respectively and (+)-zizaene titer was used as the response variable. Data sets were analyzed by Minitab 16 (Pennsylvania, USA) with the ANOVA module and the Bonferroni test was applied for mean comparison test with a 95% confidence level.

Conclusions
The results achieved in this study demonstrated the improvement of the microbial production of (+)-zizaene compared to previous studies by enhancing the recovery of (+)-zizaene. Initially, the loss of (+)-zizaene by volatilization was measured and further reduced by LLPPC at shake flask scale. Furthermore, the Diaion HP20 resin obtained the highest (+)-zizaene recovery after screening distinct adsorbers by SLPPC. After evaluating distinct solvents for the desorption process, the isooctane was selected as a suitable eluent and the SLPPC reached (+)-zizaene titers comparable to those obtained by the LLPPC. The scale-up to bioreactors by integrated product recovery configurations improved dramatically the (+)-zizaene production, whereas the IRC demonstrated higher (+)-zizaene titers (211.13 mg L −1 ) and productivities (3.2 mg L −1 h −1 ) than the ERC. Consequently, the successful application of ISPR proved a greener extraction method, which reutilizes the extractant material (polymeric adsorbers), reduces the number of extraction reagents (only one solvent is required), reduces the energy input and quantity of chemical wastes, and improves the recovery ratio of (+)-zizaene over 98%.

Supplementary Materials:
The following are available online: Figure S1: Terpene profile between distinct polymeric adsorbers; Figure S2: Mass spectra of terpene products; Figure S3: Comparison of soluble and insoluble ZS protein amounts between distinct polymeric adsorbers; Figure S4: Off-line analytics from bioreactor configurations; Table S1: Physicochemical properties of the organic solvents used for the desorption of (+)-zizaene; Table S2: Recovery ratio of (+)-zizaene from distinct polymeric adsorbers.
Author Contributions: Funding acquisition and conceptualization, T.S. and S.B.; methodology, S.B. and F.A.; experimentation, data analysis, and manuscript draft preparation, F.A.; manuscript revision, T.S. and S.B. All authors agreed to the final version.