Microaerobic growth‐decoupled production of α‐ketoglutarate and succinate from xylose in a one‐pot process using Corynebacterium glutamicum

Lignocellulosic biomass is the most abundant raw material on earth. Its efficient use for novel bio‐based materials is essential for an emerging bioeconomy. Possible building blocks for such materials are the key TCA‐cycle intermediates α‐ketoglutarate and succinate. These organic acids have a wide range of potential applications, particularly in use as monomers for established or novel biopolymers. Recently, Corynebacterium glutamicum was successfully engineered and evolved towards an improved utilization of d‐xylose via the Weimberg pathway, yielding the strain WMB2evo. The Weimberg pathway enables a carbon‐efficient C5‐to‐C5 conversion of d‐xylose to α‐ketoglutarate and a shortcut route to succinate as co‐product in a one‐pot process.

can be envisioned; thus, extending the network of value chains starting from cheap and renewable D-xylose.

K E Y W O R D S
α-ketoglutarate, Corynebacterium glutamicum, succinate, Weimberg pathway, xylose INTRODUCTION Lignocellulosic biomass is the most abundant raw material and potential waste stream on earth. [1,2] To be integrated into existing value chains, it needs to be converted into chemical building blocks, versatile molecules that can be combined into more complex products. Sustainable production of such chemical building blocks increases the sustainability of the entire production chain. Organic acids are promising candidates for this concept, as they can be produced by microbial processes and have versatile functional groups. [3] The oxocarboxylic acid α-ketoglutarate can be used as a dietary supplement, [4] in the synthesis of active pharmaceutical ingredients, [5,6] and to form biodegradable elastomers. [7] Currently, it is mainly produced by chemical synthesis from succinic acid and oxalic acid diethyl esters with a yield of 75%. [6] These processes use hazardous chemicals such as cyanohydrines or acyl cyanides and produce toxic waste. Bioprocesses using bacterial and yeast strains have been developed to provide more sustainable alternatives. [5] In terms of titer and yield, processes using the yeast Yarrowia lipolytica as production host and rapeseed oil as substrate show the best performance. Here, titers of up to 134 g L −1 and yields of up to 1.3 g g −1 were achieved. [5,8] Using Corynebacterium glutamicum as production host, titers of up to 5 g L −1 were achieved in a patent by inactivating the enzyme glutamate dehydrogenase and using a production medium with excess ammonium. [9] The dicarboxylic acid succinate is currently used as chelator, surfactant, additive in the agricultural and food industries, in the pharmaceutical industry, and as a precursor molecule for numerous applications in the chemical industry. [10,11] Prominent examples of the latter application are the synthesis of tetrahydrofuran [12] and the esterification with 1,4-butanediol to provide the thermoplastic polymer polybutylene succinate. [13] A variety of bioprocesses for the production of succinate have been described. [14][15][16][17][18][19] Most of them use refined Dglucose or glycerol as feedstocks. Titers as high as 134 g L −1 have been achieved using C. glutamicum as the production host. [14] Recent studies also focused on the use of lignocellulosic biomass. [20] Here, specialized production hosts such as Actinobacillus succinogenes were used to utilize corn fiber hydrolysate and titers of 35.4 g L −1 were reached. [21] Most processes for organic acid production rely on D-glucose or glycerol as substrate. The substrate is utilized via glycolysis and passed through the tricarboxylic acid (TCA) cycle, for example, to form αketoglutarate. This mode inevitably results in carbon losses during the conversion of D-isocitrate to α-ketoglutarate. The Weimberg pathway represents a promising, carbon-efficient alternative: First discovered in the bacterium Pseudomonas fragi, [22] this pathway theoretically allows C5-to-C5 conversion of D-xylose to α-ketoglutarate with-out carbon loss. D-xylose is an abundant carbohydrate in lignocellulosic biomass and one of the most abundant raw materials on earth. [23] Utilizing this carbohydrate instead of D-glucose, issues such as competition over arable land can be avoided. [24,25] The gram-positive bacterium C. glutamicum is an established host for the production of organic acids. High titers have been achieved, for example, for succinate (1.1 mol L −1 , [14] ) and lactate (1.3 mol L −1 , [26] ).
These processes are usually performed under microaerobic or anaerobic conditions to foster carbon flux into such products, those formation is coupled to the recycling of reduced cofactors (e.g., NADH and MQH 2 in C. glutamicum) when oxidative phosphorylation becomes limiting.
Recently, the Weimberg pathway was introduced into C. glutamicum. A codon-optimized version of the xylXABCD genes encoding the relevant enzymes of the Weimberg pathway from Caulobacter crescentus were cloned into an appropriate vector system and transformed into C. glutamicum to allow growth on D-xylose as sole carbon and energy source. [27] Adaptive Laboratory Evolution (ALE) was used to increase substrate uptake and growth rates, yielding strain WMB2 evo . [28] Using the findings from the ALE study, several rational engineered strains were constructed. The strain C. glutamicum P O6 iolT1 ΔodhA pEKEx3-xylXD Cc -opt was able to accumulate 7.92 ± 0.13 g L −1 α-ketoglutarate. [29] In this study, the ability of strain WMB2 evo for combined production of α-ketoglutarate and succinate from D-xylose under microaerobic conditions is demonstrated. To verify the direct conversion of D-xylose to α-ketoglutarate and succinate via the Weimberg pathway in this strain, a comparative 13 C-labeling experiment was employed. Specifically, 1-13 C-D-xylose was applied as isotopic tracer and the enrichment patterns in metabolites of central carbon metabolism of strain WMB2 evo were compared to a C. glutamicum strain harboring the wellknown isomerase pathway for D-xylose utilization. The effect of oxygen limitation on co-production was investigated in fed-batch bioreactor cultures. Finally, by choosing a robust microaerobic condition, the established one-pot process was further intensified by applying higher substrate concentrations in an optimal feeding scheme.
If not stated otherwise, all chemicals were of analytical grade and purchased from Sigma (Steinheim, Germany), Merck (Darmstadt, Germany) or Roth (Karlsruhe, Germany).
Complex BHI medium was prepared by dissolving 37 g L −1 brainheart infusion (BHI) in deionized water, followed by autoclaving at 121 • C for 20 min.

Biomass quantification methods
Depending on the experiment, one or more of the following biomass quantification methods were used: Optical density at a wavelength of 600 nm was measured using a UV-spectrophotometer (UV-1800, Shimadzu) and 1 mL cuvettes. 0.9% (w v −1 ) NaCl solution was used as a reference. Absorption of samples was ensured to be between 0.05 and 0.3 a. u. by dilution. If not stated otherwise, measurements were done in analytical triplicates.
Cell dry weight was determined by transfer of a 2 mL cultivation sample in a pre-dried and pre-weight tube. The sample was centrifuged for 10 min at 13,000 × g in a tabletop centrifuge. Subsequently, the supernatant was removed, and the pellet was washed using 1 mL 0.9% (w v -1 ) NaCl solution. The sample was centrifuged again for 10 min at 13,000 × g in a tabletop centrifuge and the supernatant discarded. The sample was dried in a dry oven at 80 • C for 12 to 24 h and subsequently weight. . External standards were used for linear regression within the linear dynamic range. Estimation of measurement errors was done by parametric bootstrapping as previously described. [32]

Quantification of organic acids in culture supernatant
Samples were prepared by passing the supernatants through a cellu- For external calibration, standards of organic acids or carbohydrates (supplied by Sigma-Aldrich, Steinheim, Germany) were applied in the linear dynamic range. Estimation of measurement errors was done by parametric bootstrapping as previously described. [32] 2. 6 Analysis of 13

C-labeled intermediates
Each sampled biomass pellet was processed and extracted following the procedure described previously. [ The targeted fragment ions for each metabolite and their structural identity was taken from. [34]

RESULTS AND DISCUSSION
3.1 C. glutamicum WMB2 evo accumulates α-ketoglutarate and succinate during dynamic microaerobic cultivation on D-xylose C. glutamicum is a known producer of organic acids under microaerobic and anaerobic conditions, where it switches from aerobic respiration to fermentative metabolism. While the predominant product under these conditions is L-lactate, extracellular accumulation of acetate and succinate also occurs, but to a lesser extent. This property has been successfully exploited in the past to establish production processes for Llactate [26] and succinate. [14] Most studies about fermentative processes in C. glutamicum deal with the wild type (ATCC 13032) and D-glucose as carbon and energy source, which enters the catabolism via glucose-6-phosphate as first intermediate of the Embden-Meyerhof-Parnas pathway (EMP). Strain WMB2 evo on the other hand carries a functional Weimberg pathway for utilizing D-xylose as sole carbon and energy source for aerobic growth ( Figure 1A). Here carbon enters the catabolism at the αketoglutarate node and is further metabolized via the TCA cycle, leading to a potential excess of NADH. Therefore, we questioned whether this strain can utilize D-xylose for growth under oxygen limiting conditions, and how the resulting phenotype differs from that of the C. glutamicum wild type on D-glucose.
To answer this question, a microaerobic batch cultivation with WMB2 evo using defined CGXII medium with D-xylose as sole carbon and energy source was performed ( Figure 1B). The cultivation conditions were deliberately chosen so that the culture was no longer sufficiently supplied with oxygen after reaching a certain biomass concentration. During the first 14 h, sufficient oxygen was present for unlimited aerobic growth (phase I). This was followed by a microaerobic growth phase (phase II). Dissolved oxygen concentration was 0%, indicating oxygen limitation, but cells continued to grow. During this phase, α-ketoglutarate, succinate, and acetate were formed. In phase III, cell growth stopped, even though substrate was still present, indicating a severe oxygen limitation. Despite this, succinate production even accelerated, and at the end of the cultivation, no more acetate could be found in the supernatant. Most likely it was taken up again, indicated by the dip in the dissolved oxygen signal at t = 61 h.
This first experiment revealed that D-xylose is utilized by strain WMB2 evo under microaerobic conditions. The metabolic footprint differs from growth on D-glucose: α-ketoglutarate and succinate, both intermediates of the TCA cycle ( Figure 1B), were accumulated at the end of the cultivation with titers of 0.83 and 28.64 mmol, respectively. The rate of their formation depended on the availability of oxygen. Acetate was formed as a by-product but was also taken up again as soon as oxygen limitation was removed. Noteworthy, no L-lactate could be found in the supernatant samples, which is a significant difference from the wild type strain under such conditions. Furthermore, no D-xylonate was found as a by-product of D-xylose utilization via the Weimberg pathway. [28]

Verification of D-xylose utilization mode by 13 C-isotope labeling
The biosynthesis of α-ketoglutarate and succinate by oxidation of Dxylose via the Weimberg pathway is associated with the generation of excess energy equivalents, for example, 3 mol NADH and 1 mol ATP per mol of D-xylose in the case of succinate (cf. Figure 1A). Especially under the microaerobic conditions applied, it is questionable whether this excess is fully utilized by the cells for growth and maintenance. As an alternative explanation, it is conceivable that another by-product is formed that allows cofactor recycling and that was not detectable by the applied analytics. As a second hypothesis, one could speculate on a different, previously unknown mode of D-xylose assimilation in the evolved WMB2 evo strain, linking the formation of succinate to a reductive operation of the TCA cycle, as is the case in the wild type under microaerobic conditions.
To verify the exclusive operation of the Weimberg pathway for Dxylose assimilation in WMB2 evo , an isotope labeling experiment was performed. 1-13 C-D-xylose was used as the isotopic tracer and the resulting α-ketoglutarate pool should consist mainly of the isotopomer with a labeled carbon atom at the first position. In addition, succinate should be found mainly unlabeled, because the labeled carbon atom is released as CO 2 upon oxidation of α-ketoglutarate to succinyl-CoA.
For a direct comparison of the resulting labeling patterns, a C. glutamicum strain that utilizes D-xylose via the isomerase pathway (ISO) [30] was analyzed in parallel. A much broader distribution of labeled carbon was expected for this strain, as D-xylose is introduced into the pen-tose phosphate pathway (PPP) at the xylulose-5-phosphate level and further metabolized via the oxidative PPP and the lower EMP (cf. Figure 2).
The experiment was performed with shake flask cultures sampled in the mid-exponential phase. Cold methanol quenching in combination with tandem mass spectrometry was performed to analyze the labeling patterns of key intracellular metabolites ( Figure 2). As expected, the pools of D-xylonate, α-ketoglutarate, and L-glutamate in strain WMB2 evo consisted mainly of the single labeled isotopomer (>89%) introduced by the substrate 1-13 C-D-xylose. Furthermore, succinate was predominantly unlabeled. However, a small portion of this pool (12%) also contained a single labeling that was higher than the natural isotopic abundance (≈4.5%). A likely explanation is that anaplerotic reactions present in C. glutamicum can recapture labeled carbon that is released as CO 2 in the reaction leading to succinyl-CoA. This could also explain the other smaller fractions of labeled 3-phosphoglycerate and glyceraldehyde-3-phosphate (cf. Figure 2). By contrast, in the ISO strain only D-xylonate shows a narrow labeling pattern, proving that D-xylonate is also formed from D-xylose in this strain. This is in good agreement with the previous finding that the endogenous enzyme IolG F I G U R E 3 Impact of oxygen availability on α-ketoglutarate and succinate productivity. (A) Exemplary bioreactor screening process using C. glutamicum WMB2 evo and defined CGXII medium. In phase I, biomass was generated using 40 g of D-glucose as sole carbon and energy source. In phase II, microaerobic conditions were induced by fixing the stirrer rate to 900 rpm, the aeration rate to 0.2 vvm, and varying the oxygen content in the inlet gas. As substrate for the co-production of both organic acids, D-xylose was fed into the bioreactor via a peristaltic pump at a rate of 48 mmol h −1 for 6 h. In phase III, the feed was stopped and monitoring continued. (B) Productivity and titers of all eight bioreactor screening processes. The oxygen concentration in the inlet gas was varied between 8 and 21 vol.%, resulting in an oxygen transfer rate plotted on the X-axis. Productivity was calculated by linear regression of the data points within each production phase. No productivity could be calculated for α-ketoglutarate for the 14 and 21 vol.% conditions because no production phase could be defined. also drives D-xylose oxidation in C. glutamicum. [38] In summary, the labeling pattern found for C. glutamicum WMB2 evo verifies the exclusive utilization of D-xylose via the Weimberg pathway in this strain.

Impact of oxygen availability on organic acid production performance
To further investigate, the influence of oxygen availability on the coproduction of α-ketoglutarate and succinate, a dedicated microaerobic fed-batch screening process was set up ( Figure 3A). In phase I, biomass was produced using a defined amount of D-glucose and fully aerobic conditions ensured by DO-controlled stirring and gassing with pure air.
From phase II onwards, microaerobic conditions were achieved by lowering the stirring speed as well as the volumetric gas flow rate and mixing nitrogen into the inlet gas. Depending on the proportion of nitrogen added, the oxygen supply could be varied in a controlled manner.
In phase II, D-xylose was also added as a substrate for organic acid production. In this procedure, only α-ketoglutarate and succinate were produced in quantifiable amounts; acetate production was absent. This is likely due to the rapid change between fully aerobic and microaerobic conditions in this setup, which also leads to growth decoupling of the production process. In phase III, the supply of D-xylose was stopped, but since not all D-xylose was taken up by the cells at this time, organic acid production continued. Interestingly, the accumulation of α-ketoglutarate stopped rapidly after the termination of feeding, whereas the production of succinate continued until all the Dxylose was used up.
In total, eight different levels of oxygen percentage in the inlet gas flow were tested ( Figure 3B, Figures S1-S3). The resulting OTRs ranged from 22 to 45 mmol L −1 h −1 . For α-ketoglutarate, the highest amount, 74.0 mmol, was achieved by the processes with the lowest and second lowest oxygen levels. The same processes also reached the highest productivities of 11.85 and 11.84 mmol h −1 , respectively. For succinate both the highest amount, 226.8 mmol, and the highest productivity, 16.7 mmol h −1 , were achieved by the processes with 12 vol.%.
A general trend seems to be that a higher degree of oxygen limitation leads to higher α-ketoglutarate titers and productivities. This is in good agreement with the first experiment (cf. Figure 1A). Little to no α-ketoglutarate was found under the higher oxygen conditions. It is noteworthy that varying the oxygen supply does not cause a shift in the carbon flux between the two organic acids, but rather a lower oxygen supply results in a higher amount of both products. Carbon balances revealed that a substantial amount of carbon is still converted to CO 2 , hinting to the need for further strain optimization. Note, the validity of carbon balances for these processes could be limited due to issues regarding CO 2 measurement (Figures S2 and S3).
In conclusion, the screening experiments showed that αketoglutarate and succinate can be produced in a growth-decoupled manner. Lower oxygen availability increased the accumulation of α-ketoglutarate and succinate, but below an OTR of 30 mmol L −1 h −1 productivity for succinate began to decline.

Process intensification by evaluation of different feeding strategies
Based on previous findings, two strategies for enhanced co-production of α-ketoglutarate and succinate were evaluated: A fed-batch process with an extended feed phase and a pulsed fed-batch in which D-xylose is rapidly added at discrete time points. Because of the growth-decoupled production mode, potential substrate inhibition was assumed to be less critical, so high D-xylose concentrations were used.
A total of 1440 mmol (reactors 1, 2, and 3) and 720 mmol (reactor 4) were added. For all reactors, biomass was generated as described before.
Continuous feeding resulted in a linear increase of α-ketoglutarate in the supernatant, shortly after feeding was started ( Figure 4A Table 1). Production occurred throughout the feeding phase and beyond, resulting in an accumulation of 140 mmol (R1) and 146 mmol (R2) α-ketoglutarate, respectively. The rate of product formation remained constant throughout the production phase, indicating that no inhibition occurred with this feeding regimen. Succinate also accumulated rapidly but reached a maximum at t = 68 h with amounts of 190 mmol (R1) and 160 mmol (R2), respectively.
Pulsed feeding also resulted in accumulation of α-ketoglutarate and succinate ( Figures 4B and S4,  Space time yield was calculated from start of feeding until the end of the cultivation.
feeding, the accumulation of α-ketoglutarate was faster at the beginning, indicating a positive effect of the high substrate concentrations.
In reactor 3 fed with 1440 mmol D-xylose, a slightly lower amount of 128 mmol α-ketoglutarate was produced, compared to continuous feeding. However, due to the lower amount of base added to this reactor, the titer is higher. Reactor 4 received only a 720 mmol substrate pulse. Nevertheless, more than half the amount of α-ketoglutarate (75 mmol) was produced, indicating possible product inhibition. Succinate accumulated at similar rates compared with continuous feeding.
Interestingly, accumulation also stopped at similar amounts, and there was little difference between reactor 3 (two pulses) and reactor 4 (one pulse). Probably, the cells were not able to export more succinate under these process conditions. With this feeding mode, biomass decreased significantly during the production phase, an effect that was not as pronounced with continuous feeding and could explain the lower amount of base added.
For all processes, yields for α-ketoglutarate and succinate were in the range of 0.05 -0.14 mol mol −1 , meaning that only 15 -24% of the substrate was converted to the target products. Additionally, carbon balances showed a substantial gap between carbon input and carbon quantified ( Figure S4). It is therefore likely, that at least one other by-product was formed. For example, the amino acid Lglutamate, which is directly derived from α-ketoglutarate, could accumulate intracellularly. It is known that C. glutamicum is able to store large amounts of this amino acid intracellularly, [39,40] and the required reduction of α-ketoglutarate might also be driven by an excess of NADH.
F I G U R E 4 Evaluation of different feeding strategies for enhanced co-production of α-ketoglutarate and succinate. (A) Fed-batch bioreactor process using C. glutamicum WMB2 evo and defined CGXII medium. In phase I, biomass was produced using 40 g of D-glucose as sole carbon and energy source. In phase II, microaerobic conditions were induced by fixing the stirrer rate to 900 rpm, the aeration rate to 0.2 vvm, and the oxygen content in the inlet gas to 12 vol.%. As substrate for α-ketoglutarate production, D-xylose was fed into the bioreactor via a peristaltic pump with at a rate of 48 mmol h -1 for 30 h. (B) Pulsed fed-batch bioreactor process using C. glutamicum WMB2 evo and defined CGXII medium. Biomass production and microaerobic conditions were identical to A. Substrate addition was in pulses, with reactor 3 receiving two pulses of 720 mmol D-xylose and reactor 4 receiving one pulse of 720 mmol D-xylose.
Testing of different feeding strategies revealed that high substrate concentrations lead to higher product formation rates, but possibly also to higher byproduct accumulation. In addition, despite growthdecoupled production, cells were more affected by pulsed feeding than by continuous feeding. Pulsed feeding resulted in slightly higher titers for both α-ketoglutarate and succinate, which may be beneficial for downstream applications. Moreover, pulsed feeding could be performed using the substrate as solid, further reducing dilution and potentially increasing product titers.

CONCLUSIONS
Despite not being optimized for that purpose, C. glutamicum WMB2 evo is able to co-produce substantial amounts of α-ketoglutarate and suc-cinate from the highly abundant lignocellulosic sugar D-xylose. By screening different microaerobic conditions and feeding strategies, titers of up to 78.4 mmol L −1 (11.45 g L −1 ) for α-ketoglutarate and 96.2 mmol L -1 (11.36 g L −1 ) for succinate were reached in a growthdecoupled process.
It is most likely that these performance indicators can be further improved by metabolic engineering, for example, by improving D-xylose uptake [41,42] or reducing by-product formation. Future work will focus on tailor-made down-stream processing of both organic acids from the fermentation broth to enable their application as building blocks in chemical syntheses. Alternatively, direct conversion of one or both acids via whole-cell or cell-free enzymatic approaches can be envisioned; thus, extending the network of value chains starting from cheap and renewable lignocellulosic feedstocks.