Micro‐aerobic production of isobutanol with engineered Pseudomonas putida

Abstract Pseudomonas putida KT2440 is emerging as a promising microbial host for biotechnological industry due to its broad range of substrate affinity and resilience to physicochemical stresses. Its natural tolerance towards aromatics and solvents qualifies this versatile microbe as promising candidate to produce next generation biofuels such as isobutanol. In this study, we scaled‐up the production of isobutanol with P. putida from shake flask to fed‐batch cultivation in a 30 L bioreactor. The design of a two‐stage bioprocess with separated growth and production resulted in 3.35 gisobutanol L–1. Flux analysis revealed that the NADPH expensive formation of isobutanol exceeded the cellular catabolic supply of NADPH finally causing growth retardation. Concomitantly, the cell counteracted to the redox imbalance by increased formation of 2‐ketogluconic thereby providing electrons for the respiratory ATP generation. Thus, P. putida partially uncoupled ATP formation from the availability of NADH. The quantitative analysis of intracellular pyridine nucleotides NAD(P)+ and NAD(P)H revealed elevated catabolic and anabolic reducing power during aerobic production of isobutanol. Additionally, the installation of micro‐aerobic conditions during production doubled the integral glucose‐to‐isobutanol conversion yield to 60 mgisobutanol gglucose –1 while preventing undesired carbon loss as 2‐ketogluconic acid.

corrosive impact compared to ethanol [1,2]. However, regarding the need to reduce the human carbon footprint by establishing a circular economy, industry seeks for bio-based isobutanol production from renewable feedstocks. Accordingly, microbial isobutanol production from glucose has been established in several organisms [3,4] such as Escherichia coli [5,6], Bacillus subtilis [7], and Corynebacterium glutamicum [8,9] under aerobic or anaerobic conditions. Nevertheless, the toxicity of isobutanol to these organisms is a major drawback preventing their use for industrial scale production with even further improved key performance values [4,10].
Therefore, bacterial organisms such as Pseudomonas putida with a natural tolerance towards toxic compounds and alcohols like butanol and ethanol could be an alternative production host [11][12][13]. The Gram-negative soil bacterium P. putida KT2440, certified as HV1 [14] is genetically accessible [15] and endowed with a versatile metabolism that originates from its ability to colonize harsh environments [16,17]. Moreover, P. putida exhibits a broad range of substrate affinity which is essential for the consumption of complex harvest residuals such as lignocellulose and oils [18,19]. The cyclic nature of glucose oxidation via the combined EDEMP pathway [20,21] allows P. putida to adjust NADPH formation, a trait that is beneficial to counteract oxidative stress. As recently shown, P. putida is also well-endowed to withstand transient substrate starvation conditions that may occur in large-scale bioreactors thanks to the rapid access to intracellular 3-hydroxyalkanoate storage buffers [22].
These metabolic advantages render P. putida a suitable microbial platform for isobutanol production which was successfully shown in shake flask cultures [23]. The authors engineered the P. putida GN346 mutant [13], that is unable to degrade n-butanol, for the overproduction of isobutanol from glucose via the NADPH dependent Ehrlich pathway. Besides isobutanol, high accumulation of the by-product 2-ketogluconic acid (2-KG) (>0.4 g 2-KG g glc -1 ) was observed for P. putida Iso2 and engineered derivatives [23]. The problem of unwanted 2-KG formation also occurred with P. putida engineered for cis,cis-muconic acid production [24,25]. P. putida KT2440 is known to oxidize glucose to gluconate (glt) and to 2-KG in the periplasm via glucose dehydrogenase (Gcd) and gluconate dehydrogenase (Gad) [21,26]. However, the deletion of gcd in the isobutanol producer P. putida Iso2 not only stopped 2-KG production but isobutanol formation, too [23].

PRACTICAL APPLICATION
To meet the steadily growing demand for isobutanol as chemical precursor or fuel additive, biotechnological production approaches from renewable feedstocks are in the focus of research and industry. In contrast to bacterial strains that are sensitive to extracellular alcohol, Pseudomonas putida exhibits a natural tolerance towards solvents and other toxic chemicals. Therefore, we exploited this superiority with the previously engineered P. putida Iso2 strain to achieve high isobutanol titers with different fed-batch strategies. By installing a micro-aerobic production environment in the bioreactor, the balance between cellular NADPH formation and demand was improved resulting in enhanced isobutanol yields. Since carbon uptake and intracellular accumulation of succinic acid was observed during oxygen depletion, this study paves the way for innovative micro-aerobic applications using the strict aerobic microbe P. putida.
during micro-aerobic conditions in sealed shake flasks. The authors observed minimal metabolic activity even though P. putida KT2440 is not capable of anaerobic respiration to control its energy and redox balance in the absence of oxygen [15,27]. Different studies already aimed to enable P. putida for anoxic carbon conversion, e.g. by knocking-in a fermentative pathway [28] or by using a redox mediator in a bioelectrochemical system [29,30]. However, none of these attempts resulted in anaerobic growth which would require sophisticated strain engineering [31]. Further advances were achieved by creating an anaerotolerant P. putida strain that showed growth under micro-oxic conditions [32]. Our study focuses on the scaleup of isobutanol production with P. putida from shake flask to fed-batch cultivations in a 30 L stirred tank reactor. Aerobic and micro-aerobic conditions are installed to evaluate the supply of the critical electron donor NADPH as a key to improve substrate-to-isobutanol yields. Additionally, the cellular ATP management is quantified.

Strain and medium
The P. putida Iso2 strain that was engineered for isobutanol production from glucose [23] was applied in all experiments. The strain is based on the GN346 variant [13] of P. putida KT2440 with additional deletions of bkdAA and sthA. Furthermore, it harbours the arabinose inducible plasmid pIP02 (pNG413 araC P BAD kivD yqhD alsS ilvC ilvD) for overexpression of the Ehrlich pathway and the route from pyruvate to 2-ketoisovalerate (KIV) as part of the branched-chain amino acid pathway. The cultivation medium was adopted from Vallon et al. [

Cultivation
The seed train for each bioreactor fermentation was started with a preculture that was inoculated from the working cell bank (33% glycerol stock, stored at −70 • C). The first preculture was cultivated in a 500 mL baffled shake flask containing 50 mL of minimal medium (pH 7) with 6 g glucose, 10 g 3-morpholino-propanesulfonic acid (MOPS), and 0.5 g yeast extract (VWR International, Radnor, Pennsylvania) per liter. After approx. 10 h of incubation (130 rpm, 30 • C), cells from the first preculture were used to inoculate the second preculture in a 5 L shaking flask containing 500 mL minimal medium with 6 g glucose L −1 and 10 g MOPS L −1 . After incubation for 14 h, the total volume of the second preculture (about 0.8 g X L -1 ) was used to start batch cultivation in a 30 L stirred tank reactor (STR) (Bioengineering, Wald, Switzerland) containing 8 L minimal medium (inoculum included) with 15 g glucose L −1 , at 30 • C with total pressure of 1.5 bar. The initial aeration was started at 1 L min −1 and increased stepwise to 8 L min -1 during the batch phase. Also, the initial stirrer speed of 300 min -1 was adjusted to maintain the dissolved oxygen (DOT) above 15%. During fermentation, the pH was controlled at 6.9 using 25% NH 4 OH (Carl Roth, Karlsruhe, Germany). After initially supplied glucose was consumed an exponential feed consisting of 600 g glucose and 12.5 g MgSO 4 per liter controlled the biomass growth rate at a predefined value. Additionally, five-fold trace elements were supplemented during the fed-batch phase. Peristaltic pumps (120U, Watson-Marlow GmbH, Rommerskirchen, Germany) were used for liquid feeding. Furthermore, the plasmid was induced by adding 1.57 g arabinose L -1 to the medium. Three different process strategies were performed in this study comprising a reference process (Process R), a process with fast grow-ing cells (Process F) and a process with immediate microaerobic condition after plasmid induction (Process MA). The micro-aerobic environment (DOT < 1%) was installed by switching from submerse aeration to head-space aeration. Striped-out isobutanol was trapped from the exhaust gas in 10 L deionized water (adopted from Inokuma et al. [35]). This gas washing setup yielded an isobutanol recovery of 96.3% over 12 h in a reference experiment. The O 2 and CO 2 volume fractions in the off gas was determined with a combined O 2 /CO 2 gas sensor (BlueSens Gas Sensor GmbH, Herten, Germany).

Determination of biomass and extracellular metabolites
4 × 1 mL of biosuspension was centrifuged with 20,000 g for 5 min at 4 • C, washed twice with demineralized water, transferred into pre-weighed glass vials (1.5 mL, VWR International, Radnor, Pennsylvania) and eventually dried at 105 • C for 24 h. The weight of the remaining biomass was determined using a micro balance (XP26 Delta Range R , Mettler Toledo, Gießen, Germany). The organic acids, glucose, and gluconate in the supernatant were quantified using enzyme kits (r-biopharm AG, Darmstadt, Germany). 2-ketogluconic acid, 2-ketoisovalerate, and isobutanol were measured using an isocratic HPLC equipped with the RI detector (1200Series, Agilent, Santa Clara, CA, USA) and a Rezex ROA-Organic Acid H + (300 × 7.8 mm) column (Phenomenex, Aschaffenburg, Germany) at 50 • C. 0.4 mL min −1 of 5 mM H 2 SO 4 was used as mobile phase for the separation of the HPLC analytes. The extracellular concentrations of l-valine, l-isoleucine, and l-leucine were determined according to the amino acid detection protocol described by Buchholz et al. [36].

LC-MS based analysis of intracellular metabolites
Targeted metabolome analyses of intracellular P. putida extracts were based on previous HILIC-ESI-MS studies [37,38]. The inactivation of samples and preparation of cellular extracts was performed according following adapted procedure: 0.5 mL of biosuspension were directly mixed to 1.5 mL pre-cooled (-40 • C) quenching solution (60% methanol (v/v), 133 mM ammonium acetate/pH 9.2) and immediately centrifugated for 45 s at 20,000 g in a precooled rotor (-20 • C). Cell pellets were frozen in liquid nitrogen and stored at -70 • C. The frozen pellets were resuspended in pre-cooled (-20 • C) extraction buffer consisting of methanol (66% v/v), ammonium acetate (100 mM/ pH 9.2), l-norvaline (0.15 mM), and 3-mercaptopropionic acid (2.5 mM). Added volumes were adjusted to achieve constant biomass concentrations (c X = 10-15 g L -1 ). During resuspension, sample temperature was kept below -20 • C by rotational vortexing and chilling in a constantly cooled cryostat (-40 • C). Subsequently, the same volume of pre-cooled chloroform (-20 • C) was added and mixed properly. Next, the samples were incubated for 1 h at -20 • C and 1 h at room temperature in a rotary overhead-shaker and remaining cell debris were separated by centrifugation (10 min at 20,000 g and 4 • C). The upper H 2 O/methanol phase (polar metabolites) was stored at -70 • C until measurement.
Measurements were performed on an Agilent 6410B Triple-Quad LC-MS/MS system. Sample preparation and chromatographic separation by alkaline polymer-based zwitterionic hydrophilic interaction chromatography (ZIC-pHILIC) were performed as previously described [37]. Central carbon metabolites were analyzed with high selectivity in multiple reaction monitoring (MRM) mode with high selectivity and pre-optimized precursor-toproduct ion transitions with a mass resolution of 0.1 u and adapted MS/MS parameters [38]. Pyridine nucleotide factors were analyzed with higher sensitivity in selective ion monitoring (SIM) mode with a mass resolution of 0.3 u and analog MS parameters. System control, acquisition, and analysis of data were performed by usage of commercial MassHunter B.06.00 software.

Calculation of ATP formation and demand
The total ATP generation of the cell is based on the direct phosphorylation on substrate level q ATP,S (Equation 2) via carbon uptake q S,in (Equation 1) and on the oxidative phosphorylation via the respiratory chain q ATP,O (Equation 3). According to recent published carbon fluxes [39], a net formation of 2.67 mol ATP mol glucose -1 via direct phosphorylation is assumed from glucose based carbon uptake in P. putida. The P/O ratio (1.33 mol ATP mol oxygen -1 [40]) links the oxygen uptake to the ATP formation via oxidative phosphorylation through the electron transport chain and the proton gradient that fuels the ATP synthase. Electrons to reduce oxygen in P. putida cannot only be transferred by NADH or FADH 2 but also via pyrroloquinoline (PQQ) by oxidation of glucose to gluconate and 2-ketogluconic acid [41]. In accordance, the electron flux via glucose oxidation to 2-KG was shown in anaerobically cultivated P. putida F1 [29]. The authors illustrated that the incomplete oxidation of glucose to 2-KG is an alternative route for P. putida to generate ATP (q ATP,alt ). The enzyme complex in the electron transport chain creates a proton gradient through the electron flux which drives the ATP synthase. 2 eare released per molecule of glt and 4 eper molecule of 2-KG formed from glucose that can be used to finally reduce oxygen (Equation 4). In contrast to the ATP generation, the total growth related ATP demand can be determined by Equation 5 [40].

In silico calculation of NADPH demand
The total NADPH supply was estimated using the ratio of 1.7 mol NADPH mol glucose -1 , and a NADPH surplus of 14 to 21% was assumed for growth on glucose [21,39,42]. The additional NADPH demand in the isobutanol production pathway was calculated based on the extracellular uptake and formation rates of KIV, l-valine and isobutanol. If uptake of KIV was detected, it was assumed to serve as precursor for isobutanol and l-valine. Furthermore, the intracellular flux distribution from KIV to l-valine and to isobutanol was estimated to mirror the ratio of secreted isobutanol per l-valine during aerobic production. During micro-aerobic condition, no growth occurred, and stationary conditions were assumed. Then, NADPH demands for isobutanol production and 2-KG uptake via KguD should equal NADPH formation via the enzymes Zwf, Icd and MaeB. Recently published carbon fluxes [39] and the following stoichiometries were used for carbon and NADPH balancing:

Scale-up of microbial isobutanol production with P. putida Iso2
To increase biomass concentration and to monitor the growth and production performance of the metabolically engineered P. putida Iso2 strain [23] producing the nonnative metabolite isobutanol, the cultivation process was scaled-up from shake flask to a 30 L stirred tank bioreactor. Growth and production were separated in two stages to analyse the effect of plasmid encoded overexpression of the isobutanol pathway on the phenotype and metabolism. Cells were cultivated under controlled conditions in the bioreactor with 8 L of initial working volume. Three different process strategies were performed. As illustrated in Figure 1A, each process consists of a batch and fedbatch phase (growth phase) and a post-induction phase (production phase). Figure 1B shows the time course of different product concentrations and the isobutanol yield with respect to the time of plasmid induction (t = 0h). The reference process is entitled as "Process R" and was accomplished in the following manner: During the batch process (until -13 h), cells grew with a maximum specific growth rate of 0.29 ( ± 0.03) h -1 showing a biomass specific yield of 0.16 g X g glucose -1 (Table 1) and accumulated 1.7 g 2-KG L -1 . In the subsequent glucose limited fed-batch phase (from -13 to 0 h) cells grew with μ fb = 0.13 h -1 . Cell density reached 12 g X L -1 and 8.3 g L -1 of the precursor 2-ketoisovalerate (KIV) for the isobutanol pathway accumulated during the fed-batch phase. The feed profiles, glucose concentrations, and DOT levels are shown in the Figure S1. Next, the inducer arabinose (1.57 g L -1 ) was added to the broth to initiate the production of isobutanol (t = 0h). During production phase I, extracellular isobu-tanol accumulated to 1.7 g isobutanol L -1 , but the growth rate decreased to 0.02 h -1 within 8 h. Concomitantly, cells started to accumulate 22 g 2-KG L -1 , and 2.1 g L-valine L -1 . Extracellular KIV concentrations remained for the first 4 h of production but declined afterwards due to cellular consumption. In production phase II (from 8 to 12 h), the submerse aeration was switched to head-space aeration to install micro-aerobic conditions. The resulting oxygen deprivation led to only minimal secretion of 2-KG and l-valine while the isobutanol titer increased further from 1.7 to 2.4 g L -1 which was accompanied by consumption of KIV. In total, 943 g glucose were consumed to produce 23.1 g isobutanol within 38 h leading to a yield of 22 mg isobutanol g glucose -1 (refer to Table 1). Figure 2 illustrates the uptake and formation rates of the substrates and products in all three process strategies before (0 h) and after plasmid induction (2-12 h). In the reference Process R, the biomass specific glucose consumption rate rose from 1.95 to 2.95 mmol glucose g X -1 h -1 during production phase I even though the growth rate declined from 0.13 to 0.02 h -1 . Simultaneously, the 2-KG production rate increased significantly to 2.0 mmol 2-KG g X -1 h -1 which led to a persistent decline of carbon uptake (total of C-6 sugars). Nevertheless, the maximum isobutanol production rate of 0.39 mmol isobutanol g X -1 h -1 was observed within the first 2 h after induction, accompanied by a maximum l-valine excretion of 0.24 mmol valine g X -1 h -1 . The highest KIV consumption rate of 0.95 mmol KIV g X -1 h -1 was measured in the micro-aerobic production phase II in which isobutanol production rate (0.17 mmol isobutanol g X -1 h -1 ) was still half of the maximum rate.

High growth rates lead to increased product and by-product formation
With the aim to increase the biomass specific isobutanol production rate, fast cell growth of 0.22 h -1 that is close to μ max was installed during the fed-batch phase (from -10.9 to 0 h) in Process F by adjusting the glucose feed (see Figure 1). The plasmid encoded isobutanol production was started with 16.6 g X L -1 biomass concentration (t = 0 h). Within F I G U R E 1 (A) Schematic overview of the three different process strategies applied in this study. Process R serves as reference process. After the batch and fed-batch phase, plasmid was induced and isobutanol production was monitored. After 8 h, submerse aeration was switched to headspace aeration. In contrast, fast growing cells were induced in Process F and micro-aerobic condition was installed after 6 h. 4 h of production phase I, 3.0 g isobutanol L -1 accumulated. 2-KG formation increased by 3.3-fold, reaching a maximum medium concentration of 82 g 2-KG L -1 that accounts for 0.58 g 2-KG g glucose -1 . Consequently, the total glucose consumption increased by 41% still revealing an almost equal isobutanol yield of 25 mg isobutanol g glucose -1 compared to the reference Process R (Table 1). After reaching a peak at 18.3 g X L -1 biomass concentrations declined by more than 20% (from 2 to 6 h). Switching to micro-aerobic condition after 6 h only marginally affected isobutanol production. The maximum of 3.35 g isobutanol L -1 in production phase II coincided with the depletion of the extracellular precursor KIV.
As illustrated in Figure 2, the high growth rate during fed-batch in Process F resulted in a 1.7-fold enhanced maximum isobutanol production rate of 0.65 mmol g X -1 h -1 in relation to the reference Process R. Noteworthy, within production phase I in Process F, glucose consumption rate almost tripled from 2.7 to 7.6 mmol glucose g X -1 h -1 and 2-KG formation rate increased by 24-fold in contrast to the uninduced cells. Interestingly, despite an increased glucose supply and consumption in Process F, the combined carbon uptake of glucose, gluconate and 2-ketogluconate did not exceed 2.4 mmol g X -1 h -1 although a maximum carbon uptake of approx. 8 mmol g X -1 h -1 was observed during the batch phase. Even more, the growth rate declined drastically by nearly 70% within the first 2 h of production phase I followed by a phase of reduced cell density. During production of isobutanol, up to 0.28 mmol gluconate g X -1 h -1 was measured. Additionally, extracellular accumulated KIV was almost completely consumed during production phase I. To conclude, the set growth rate close to μ max led to increased production of isobutanol but also to exorbitantly high oxidation of glucose to 2-KG and a cessation of growth.

Metabolic constraints during aerobic overproduction of isobutanol
As illustrated in Figure 3A, approximately 30% of the total carbon uptake during the fed-batch phase in Process R and Process F accounts for the products KIV, isobutanol, lvaline, and CO 2 that are formed in the production pathway. Hereby, most of the carbon is excreted in form of KIV. During production phase I, the product per substrate carbon share in the isobutanol pathway is slightly decreased, but still above 20%. After 4 h of production phase I, a reduced isobutanol production rate resulted in less than 10% product formation per carbon uptake. Moreover, we calculated the additional NADPH that is consumed within the production pathway based on the product formation rates. The total NADPH supply was estimated from published NADPH production rates in P. putida KT2440 [21,39]. The authors found that glucose-grown P. putida consumes most of the created NADPH by anabolic reactions except of a surplus of 14 to 21%. As shown in Figure 3B, the NADPH surplus (17.4% as average of literature values) covers the additional NADPH demand from overexpressed enzymes during the fed-batch phase. However, during the first 4 h of production phase I the additional NADPH demand exceeds the theoretical NADPH surplus which coincides with a decline of the growth rate in both processes.
Furthermore, we estimated the ATP formation of the cells based on their carbon and oxygen uptake. Additionally, the oxidation of glucose to gluconate and 2-KG releases electrons that reduce the respiratory chain component PQQ [29,41]. In return, the proton gradient created by the electron transport chain drives the ATP synthase. Thus, the periplasmatic oxidation of glucose F I G U R E 4 Intracellular concentrations (in μmol g X -1 ) of the redox cofactors NAD + , NADP + , NADH and NADPH and the respective redox ratios during the production phase in the reference Process R (grey circle) and in Process F (black circle). Time point 0 h equals the reference condition before induction of the plasmid to 2-KG contributes to the overall ATP formation from oxidative phosphorylation. More precisely, Figures 3C,D illustrate the ATP formation from substrate level and oxidative phosphorylation before and after plasmid induction during aerobic production. The latter consists of electron transfers from NADH, FADH 2 , and periplasmatic glucose oxidation. The total ATP generation in Process R and Process F remains nearly constant during production phase I. Remarkably, the amount of ATP generated by electrons released from the 2-KG pathway increased from 0.6 to 5.7 mmol g X -1 h -1 in Process R and from 3.8 to 13.9 mmol g X -1 h -1 in Process F, resulting in an ATP coverage of 50% at the end of production phase I in Process F. Figure 4 shows absolute concentrations [μmol g X -1 ] of intracellular redox cofactors NAD + , NADH, NADP + and NADPH that were detected in the aerobic Process R and Process F. Reduced NAD(P)H analogues are often heavily degraded under standard cold-based quenching/extraction conditions [43]. However, alkaline conditions (pH 9.2) with varying buffer concentrations and the addition of the reducing agent mercaptoethanol stabilized reduced pyridine nucleotides in solution [44]. Thus, by testing different buffer and reducing agents, we identified the supplementation of ammonium acetate to methanol as optimal quenching solution and supplementation of ammonium acetate plus 3-mercaptopropioinic acid to methanol as ideal extraction solution in regard of reliable and stabilized NAD(P)H species. This method was suitable to detect oxidized as well as reduced nucleotides. Under aerobic isobutanol production in Process R and Process F, pools of intracellular oxidized cofactors NAD + and NADP + were found to be significantly decreased in production phase I and phase II (see Figure 4). Remarkably, the level of the reduced species NADH and NADPH remained nearly constant in both processes except for an NADH outlier after 6 h in Process R. Consequently, isobutanol producing P. putida show higher catabolic reduction charges (NADH/NAD + ) in both processes and slightly enhanced anabolic reduction charge in Process F. Further intracellular central carbon metabolites are shown in Figure S2.

Micro-aerobic conditions increased isobutanol yield and reduced by-product formation
As shown in the reference Process R, switching from aerobic to micro-aerobic condition significantly reduced the by-product formation with almost unchanged product formation at the same time. Therefore, Process MA was performed similarly to Process R, but micro-aerobic condition (DOT = 0%) was installed already 20 min after plasmid induction in production phase I. The results of this process were validated in a biological duplicate (refer to Table S1). The related process kinetics can be drawn from Figure 1. As expected, no extracellular accumulation of 2-KG was observed in the production phase I. However, the biomass concentration declined considerably from 12.4 to 10.2 g X L -1 during the micro-aerobic phase. Moreover, extracellular accumulated KIV was completely converted to isobutanol within 6 h after induction, resulting in a maximum titer of 2.67 g isobutanol L -1 . This process strategy yielded 60 mg isobutanol /g glucose within only 4 h, which is two times higher compared to Process F and Process R. Besides, total glucose consumed was 58% less than in the reference Process R. Furthermore, the extracellular TA B L E 2 Carbon balance derived from the three different process strategies Each value represents the carbon share (C-mol/C-mol) of the product based on the total glucose consumed at the end of the process.

A B
F I G U R E 5 Schematic overview of assumed metabolic fluxes that are related to the intracellular pools of ATP, NADH, and NADPH (A) before induction and (B) during isobutanol formation in P. putida Iso2. The thickness of the arrows indicates the intensity of the related flux by-product l-valine was approx. 50% lower than detected during both aerobic production strategies. As illustrated in Figure 2, only a minimal glucose consumption rate was observed during micro-aerobic condition in Process MA. Therefore, carbon uptake into the cells was only approx. 10% of aerobic carbon uptake. Concomitantly, the growth rate turned negative resulting in reduction of cell concentration. However, nearly equimolar conversion of extracellular KIV into isobutanol was observed within 4 h of oxygen limited production in Process MA, but isobutanol production ceased after complete consumption of KIV. The Figure S3 illustrates the intracellular concentrations of carbon metabolites and redox cofactor during micro-aerobic isobutanol production. Notably, succinic acid accumulated more than six-fold during oxygen depletion compared to aerobic growth. In addition, carbon balance revealed that 12.8% of glucose was fuelled into the l-valine and the Ehrlich pathway, of which 9.2% accounts for isobutanol, in case of the optimized Process MA (Table 2). In contrast, only 4% of glucose was channelled into isobutanol during aerobic production in Process R and Process F. Even more, aerobic, fast-growing P. putida cells (Process F) lost 58% carbon into 2-KG which does only occur to a minimal extent (0.6%) under micro-aerobic production.

NADPH limitation constrains the growth-coupled isobutanol production in P. putida
The overexpression of the recombinant enzymes IlvD and YqhD to produce isobutanol obviously leads to an excessive demand of the cofactor NADPH in P. putida Iso2. This resulted in metabolic constraints affecting the cellular pools of NADPH, NADH and ATP which we depicted in Figure 5. As shown by Kohlstedt and Wittmann [39], the metabolism of P. putida creates a surplus of NADPH in growing cells that is most likely converted into NADH by the transhydrogenase. Interestingly, isobutanol production in P. putida was only possible after knockout of the transhydrogenase stha [23]. This genetic manipulation enabled the ample supply of NADPH for the isobutanol production pathway. As shown in Figure 3, the cofactor surplus was sufficient to cover the overproduction of KIV and l-valine during the non-induced fed-batch phase. The expression of the relative enzymes before induction most likely mirrors leaky plasmid expression. However, the overexpression of Yqhd after induction created a further NADPH sink that exceeded the NADPH surplus. As a result, NADPH demand in the production pathway competed with anabolic processes which resulted in reduced growth in Process R. In fact, a 66% higher production rate in Process F coincided with cessation of growth. Furthermore, glucose uptake is accompanied by partial periplasmatic oxidation to gluconate and to 2-KG [21]. In contrast to recent approaches to prevent 2-KG secretion by deletion of the glucose dehydrogenase gcd [23][24][25], the glucose limited feeding strategy inhibited extracellular secretion of 2-KG before plasmid induction. However, after induction, the high NADPH demand in the isobutanol pathway competed with the NADPH dependent uptake of 2-KG and may partially explain the extracellular accumulation of this organic acid. In addition, to satisfy the extensive NADPH demand, an increased activity of the membrane bound transhydrogenase PntAB is likely to occur. This enzyme's favoured reaction is the reduction of NADP + at the expense of NADH via proton translocation [45,46] which leads to a reduced availability of NADH for respiratory ATP generation. To satisfy the cellular ATP demand, electrons for the respiratory chain are released by periplasmatic oxidation of glucose to 2-KG [29,41] which decouples ATP generation from the availability of NADH or FADH 2 as electron donor. This hypothesis is supported by the immense extracellular accumulation of 2-KG during aerobic isobutanol production that resulted in a carbon loss of up to 58%. Concomitantly, the electron supply from periplasmatic glucose oxidation contributed to total ATP generation with rising significance (Figure 3). The estimated total ATP generation in P. putida Iso2 is close to the theoretical demand of 15 mmol ATP g X -1 h -1 at μ = 0.13 h -1 and 24 mmol ATP g X -1 h -1 at μ = 0.22 h -1 for wildtype P. putida [40]. The crucial role of periplasmatic glucose oxidation during NADPH intensive product formation in P. putida is underlined by the lacking isobutanol production in the Δgcd strain [23].
These results show that cells are endowed to quickly adapt NADPH supply and energy management to current needs. Figure 4 outlines that physiological NADH and NADPH levels could be maintained for the sake of lower NAD + and NADP + levels which finally increased catabolic and anabolic reduction states. Still, the resulting redox ratios are in good agreement with published findings measured in P. putida KT2440 [47,48]. Since NADPH is also required to counteract oxidative [49] and chemical stress [50,51], the cellular need to maintain physiological levels or even increase the cofactor supply through rearranging metabolic fluxes [42] is important. Accordingly, the NADPH requirements for isobutanol compete with the 'life-style' of P. putida requiring for further systems metabolic engineering measures to succeed.
Even though the fed-batch cultivation significantly increased the isobutanol titer compared to shake flask (0.12 g L -1 ) [23] and exceeds the product formation of engineered Pseudomonas sp. strain [52], the production efficiency of P. putida Iso2 is still not competitive with high yield aerobic or anaerobic approaches using, e.g. E. coli or C. glutamicum [3].

Oxygen limited stationary conditions benefit NADPH dependent isobutanol formation
Processes R, F and MA vary by the installation of different post-induction aeration scenarios. In contrast to aerobic conditions that enabled growth-coupled isobutanol formation, micro-aerobic conditions limited cell growth leading to growth-decoupled isobutanol production. The dualphase cultivation strategy, comprising aerobic biomass formation followed by micro-aerobic isobutanol production in Process MA substantially improved the product yield and inhibited extracellular by-product formation. Moreover, the micro-aerobic process strategy optimized NADPH dependent product formation-within the cellular capacities as illustrated above-but hampered the growth of the strict aerobic microbe P. putida. The controlled limitation of oxygen supply forced the cells into a stationary state as their energy generation is mainly dependent on the electron transport chain with the final electron acceptor oxygen [39]. In this case, electrons cannot be transferred by the succinate dehydrogenase to reduce FAD + which explains the observed intracellular accumulation of succinic acid that is seven-fold higher in production phase I of Process MA compared to the pre-induction level. In accordance, accumulation of succinic acid was also observed for anaerobic production of isobutanol in C. glutamicum [9]. Noteworthy, P. putida Iso2 showed a substantially higher volumetric isobutanol productivity of 9.5 mmol L -1 h -1 during micro-aerobic conditions compared to 4.4 mmol L -1 h -1 observed for C. glutamicum [9]. Furthermore, the microaerobic production condition in Process MA resulted in direct conversion of KIV into isobutanol which doubled the formation rate in contrast to Process F and reduced the product related NADPH costs. We estimated a lower additional NADPH demand of 1.1 mol NADPH mol isobutanol -1 in Process MA compared to 1.5 mol NADPH mol isobutanol -1 and 1.9 mol NADPH mol isobutanol -1 in Process R and Process F, respectively. Nevertheless, metabolic activity of P. putida is a prerequisite for NADPH supply in a micro-aerobic environment and could potentially be monitored using exhaust gas analysis [53]. In agreement with Lai et al. [29] and Nitschel et al. [23], minimal carbon uptake to eventually generate NADPH through the central enzymes Zwf, Icd and MaeB [21,39] was detected during oxygen depletion. We estimated a total NADPH formation of 510 mmol NADPH from carbon uptake versus a total demand of only 300 mmol NADPH needed as cofactor for Yqhd to convert isobutyraldehyde into isobutanol during 4 h of micro-aerobic production in Process MA. Thus, a sufficient amount of cofactors was supplied for the conversion of KIV into isobutanol in stationary P. putida cells.
To conclude, stationary cells and sufficient precursor supply qualifies micro-aerobic isobutanol formation in P. putida Iso2 as more efficient in terms of energy and NADPH demand compared to aerobic production. Moreover, the formation of 2-ketogluconic acid appears to be a unique trait in P. putida to balance its energy supply as a countermeasure to cofactor limitation. The novel findings of oxygen limited succinate accumulation could pave the way for innovative biotechnological applications with P. putida. Since micro-aerobic isobutanol production only occurred until extracellular KIV was depleted, we assume a hampered carbon flux to the precursor KIV during oxygen depletion. This bottleneck and the role of the transhydrogenase PntAB should be addressed by subsequent metabolic engineering studies to facilitate NADPH demanding product formation with P. putida KT2440.