Alone at last! – Heterologous expression of a single gene is sufficient for establishing the five-step Weimberg pathway in Corynebacterium glutamicum

Corynebacterium glutamicum can grow on d-xylose as sole carbon and energy source via the five-step Weimberg pathway when the pentacistronic xylXABCD operon from Caulobacter crescentus is heterologously expressed. More recently, it could be demonstrated that the C. glutamicum wild type accumulates the Weimberg pathway intermediate d-xylonate when cultivated in the presence of d-xylose. Reason for this is the activity of the endogenous dehydrogenase IolG, which can also oxidize d-xylose. This raised the question whether additional endogenous enzymes in C. glutamicum contribute to the catabolization of d-xylose via the Weimberg pathway. In this study, analysis of the C. glutamicum genome in combination with systematic reduction of the heterologous xylXABCD operon revealed that the hitherto unknown and endogenous dehydrogenase KsaD (Cg0535) can also oxidize α-ketoglutarate semialdehyde to the tricarboxylic acid cycle intermediate α-ketoglutarate, the final enzymatic step of the Weimberg pathway. Furthermore, heterologous expression of either xylX or xylD, encoding for the two dehydratases of the Weimberg pathway in C. crescentus, is sufficient for enabling C. glutamicum to grow on d-xylose as sole carbon and energy source. Finally, several variants for the carbon-efficient microbial production of α-ketoglutarate from d-xylose were constructed. In comparison to cultivation solely on d-glucose, the best strain accumulated up to 1.5-fold more α-ketoglutarate in d-xylose/d-glucose mixtures.


Introduction
The Gram-positive bacterium Corynebacterium glutamicum has a long history in the industrial production of proteinogenic amino acids. In particular L-glutamate and L-lysine are produced at million ton-scale with this microorganism (Eggeling and Bott, 2015;Lee and Wendisch, 2017). Furthermore, C. glutamicum strains for more than 70 biotechnologically interesting compounds such as alcohols, organic acids or polyphenols have been engineered over the last years (Becker et al., 2018;Kallscheuer et al., 2016Kallscheuer et al., , 2017Vogt et al., 2016;Wieschalka et al., 2013). However, all large-scale applications for amino acid production with C. glutamicum use D-glucose from starch hydrolysates or D-fructose (and sucrose) from molasses and the substrate spectrum of C. glutamicum variants engineered for other small molecules is also for the most part limited to these hexoses (Blombach and Seibold, 2010).
More recent studies focus on engineering C. glutamicum for the utilization of lignocellulose-derived pentoses D-xylose and L-arabinose as C. glutamicum cannot naturally catabolize these sugars (Kawaguchi et al., 2006(Kawaguchi et al., , 2008. In case of D-xylose, two different metabolic routes have been individually added to the catabolic repertoire of C. glutamicum. In the Isomerase pathway, D-xylose is first converted to D-xylulose by a heterologous D-xylose isomerase (encoded by xylA from either Escherichia coli or Xanthomonas campestris) and subsequently phosphorylated by an endogenous D-xylulokinase (encoded by xylB) yielding D-xylulose-5-phosphate, which can be rapidly metabolized (Kawaguchi et al., 2006;Meiswinkel et al., 2013). Several C. glutamicum strains, capable of utilizing D-xylose via the Isomerase pathway have been engineered for the production of succinate, ethanol, lysine, glutamate, ornithine, putrescine and 1,5-diaminopentane (Buschke et al., 2011;Jo et al., 2017;Meiswinkel et al., 2013). In contrast, functional introduction of the xylXABCD operon from Caulobacter crescentus enabled C. glutamicum to grow on D-xylose as sole carbon and energy source via the five-step Weimberg pathway (Radek et al., 2014). In this pathway, D-xylose is initially oxidized to 1,4-D-xylonolactone via a xylose dehydrogenase (XylB) and subsequently hydrolyzed by a D-xylonolactonase (XylC) yielding D-xylonate (Fig. 1). Two subsequent dehydration reactions, catalyzed by a D-xylonate dehydratase (XylD) and a 2-keto-3-deoxyxylonate dehydratase (XylX), lead to α-ketoglutarate semialdehyde, which is finally oxidized by an α-ketoglutarate semialdehyde dehydrogenase (XylA) to the tricarboxylic acid (TCA)-cycle intermediate α-ketoglutarate. However, C. glutamicum WMB1 as the first engineered strain having the Weimberg pathway allowed only for a growth rate of μ ¼ 0.07 h À1 on D-xylose containing defined medium. Adaptive laboratory evolution improved D-xylose utilization by 260 % yielding the strain C. glutamicum WMB2 evo (μ max ¼ 0.26 h À1 ) (Radek et al., 2017). Genome sequencing of this strain revealed a functional loss of the transcriptional regulator IolR, which controls the expression of 22 genes for the most part believed to be involved in myo-inositol metabolization (Klaffl et al., 2013). Among these genes is iolT1 encoding for the myo-inositol/proton symporter IolT1, which turned out to also contribute to D-xylose uptake in C. glutamicum . By rationally introducing two point mutations into the IolR-binding site of the iolT1-promoter yielding C. glutamicum P O6 iolT1, this effect could be successfully mimicked. Furthermore, an endogenously encoded D-xylose dehydrogenase (IolG) contributing to the oxidation of D-xylose in C. glutamicum could be identified, which was subsequently employed for the carbon efficient production of D-xylonate with C. glutamicum .
These studies show that the C. glutamicum wild type, although not capable of D-xylose utilization via the Weimberg pathway or any other catabolic strategy by nature, does already possess individual Weimberg pathway components enabling D-xylose transport and initial D-xylose oxidation. This causes one to wonder whether there are additional endogenous enzymatic activities contributing to D-xylose utilization, which would help to reduce the number of heterologous genes required for establishing the Weimberg pathway in this bacterium.
In this study, we performed an analysis of the C. glutamicum genome in combination with systematic reduction of the xylXABCD operon to identify such enzymes. Furthermore, we exploited the Weimberg pathway for the direct conversion of D-xylose to α-ketoglutarate and could show that this represents a promising strategy for the microbial production of α-ketoglutarate with C. glutamicum.

Bacterial strains, plasmids, media and growth conditions
All used bacterial strains and plasmids including their characteristics and sources are listed in Table 1. Escherichia coli DH5α, used for cloning purposes only, was routinely cultivated on a rotary shaker (170 rpm, 37 C) in reaction tubes with 5 mL Lysogeny Broth (LB) medium (Bertani, 1951) or on LB agar plates (LB medium with 1.8 % [wt/vol] agar). All C. glutamicum strains are derived from C. glutamicum ATCC 13032 (Abe et al., 1967) and were aerobically cultivated on a rotary shaker either in reaction tubes (170 rpm, 30 C) or in baffled shake flasks (130 rpm, 30 C). As cultivation medium, brain heart infusion (BHI) medium (Difco Laboratories, Detroit, USA) or defined CGXII medium (Keilhauer et al., 1993) supplemented with different D-glucose/D-xylose mixtures were used. For plasmid propagation, kanamycin was added to final concentrations of 25 μg mL À1 (C. glutamicum) or 50 μg mL À1 (E. coli). Where appropriate, the antibiotic spectinomycin was added to a final concentration of 100 μg mL À1 . Induction of gene expression was achieved by isopropyl β-D-thiogalactoside (IPTG) supplementation to a final concentration of 1 mM. In general, growth of bacterial strains, cultivated in baffled shake flasks, was followed over time by measuring the optical density at 600 nm (OD 600 ). Cultivations in the microtiter plate format were performed in Flower Plates with optodes using the microbioreactor BioLector (m2plabs, Baesweiler, Germany), enabling online determination of backscatter, pH and dissolved oxygen. BioLector cultivations were routinely inoculated to an OD 600 of 1 and incubated at 30 C, 1300 rpm and 80 % humidity. The total culture volume was always 1 mL and the backscatter gain was set to 15.

Plasmid and strain construction
All enzymes were purchased from Thermo Scientific (Schwerte, Germany) whereas codon-optimized synthetic genes for expression in (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) C. glutamicum were obtained from Life Technologies (Darmstadt, Germany). Oligonucleotides were synthesized by Eurofins genomics (Ebersfeld, Germany) and are listed in Table 2. For molecular cloning work, standard protocols, e.g. PCR and Gibson were used (Gibson et al., 2009;Sambrook and Russel, 2001). Verification of the constructed plasmids was performed either by restriction analysis or colony PCR. DNA sequencing was conducted at Eurofins Genomics (Ebersberg, Germany). E. coli DH5α was routinely transformed using the RbCl-method, whereas C. glutamicum was always transformed by electroporation followed by an additional heat shock at 46 C for 6 min (Eggeling and Bott, 2005;Hanahan, 1983). In-frame deletion of odhA and cg0535 (kasD) was performed by two-step homologues recombination using the plasmids pk19mobsacB-ΔodhA and pk19mobsacB-Δcg0535 as previously described (Sch€ afer et al., 1994).

Microbial production of α-ketoglutarate
For initial biomass formation, all constructed C. glutamicum strains were cultivated in 50 mL BHI medium with 10 g/l D-glucose in 500 mL baffled shake flasks at 130 rpm and 30 C on a rotary shaker. Cells were harvested by centrifugation at 4000 rpm for 10 min, resuspended in defined CGXII medium with either 4 % D-glucose or a 1 % D-glucose/3 % D-xylose mixture and then further cultivated for 40 h at 130 rpm and 30 C on a rotary shaker. For α-ketoglutarate production, defined CGXII medium with either 4 % D-glucose or a 1 % D-glucose/3 % D-xylose mixture was inoculated to an OD 600 of 4. If appropriate, gene expression was induced by adding IPTG to a final concentration of 1 mM.

Heterologous expression of Cg0535 in E. coli and protein purification
The plasmid pET-28b(þ)-cg0535 was transformed into E. coli BL21 for heterologous gene expression of cg0535. Cultivations for this purpose were performed in 10 mL 2xYT medium in baffled shake flasks for 15 h at 37 C and 130 rpm on a rotary shaker. 1 mL of this culture was used to inoculate an expression culture in 100 mL 2xYT medium with 50 mg L À1 kanamycin and cultivated at 37 C and 130 rpm. At an optical density of OD 600 ¼ 1.5, gene expression was induced by the addition of 0.5 mM IPTG and then further incubated at 18 C and 130 rpm for 18 h. Cells were harvested by centrifugation for 30 min at 6000 rpm and the cell-free supernatant was discarded. Cell pellets were routinely stored at À80 C if not further processed the same day. In order to avoid protein degradation, all subsequent steps for protein isolation were performed at 4 C. Frozen cell pellets were first thawed on an ice-water mixture and resuspended in 15 mL lysis buffer (50 mM Tris-HCl pH 7.6, 100 mM NaCl, 10 mM Imidazole, 5 % Glycerin and 1 mM DTT). Crude cell extracts were obtained by using a Branson Sonifier 250 (intensity, 7; duty cycle, 40 %, 6 min; Branson Ultrasonics, Danbury, USA). After removal of the cellular debris by two centrifugation steps (30 min at 6000 rpm and 45 min at 50,000 rpm) Cg0535 was purified from the protein fraction by affinity chromatography using a GE € Akta pure chromatography system (GE Healthcare Life Sciences, Chicago, USA).

Quantification of D-xylose
For quantification of D-xylose, a commercial enzyme assay kit was used according to the manufacturer's instructions (Xylose Assay Kit, Megazymes, Wickow, Ireland). A set of different D-xylose concentrations served as external standards.

HPLC analysis
Identification and quantification of metabolites was performed using a High Performance Liquid Chromatography (HPLC) 1260 Infinity system (Agilent, Waldbronn, Germany). Separation was achieved by using an Organic acid H þ column (8 %, 300 mm by 7.80 mm; Phenomenex, Torrance, CA, USA) at 80 C with an isocratic elution program using 5 mM sulfuric acid. For detection of organic acids and D-glucose, a diode array detector (DAD) at 210 nm or a refraction index (RI) detector was used, respectively. Data acquisition and analysis was performed using the Agilent OpenLAB Data Analysis -Build 2.200.0.528 software (Agilent, Derivative of C. glutamicum ATCC 13032 with two point mutations in the promotor of iolT1, relative to the start codon at position À113 (A→G) and À112 (C→G) respectively ) This study This study Spec r ; C. glutamicum/E. coli shuttle vector for regulated gene expression; (P tac , lacI Q , pBL1 oriVCg, pUC18 oriVEc) (Gande et al., 2007) pEKEx3-xylXABCD Ccopt Spec r ; pEKEx3 derivative for the regulated expression of xylXABCD Cc of C. crescentus Radek et al. (2017) pEKEx3-xylXAD Cc -opt Waldbronn, Germany).

Identification of an endogenous α-ketoglutarate semialdehyde dehydrogenase activity
Presence of the endogenous dehydrogenase IolG oxidizing D-xylose to 1,4-D-xylonolactone and the observation that hydrolyzation of this lactone can occur spontaneously, indicates that heterologous expression of the xylose dehydrogenase (encoded by xylB) and the D-xylonolactonase (encoded by xylC) from C. crescentus might not be required for establishing the Weimberg pathway in C. glutamicum. Since reduction of the Weimberg pathway encoding operon has not been tried yet, a synthetic operon comprised of codon-optimized genes for 2-keto-3-desoxyxylonate dehydratase (xylX), xylonate dehydratase (xylD) and the α-ketoglutarate semialdehyde dehydrogenase (xylA), all originating from C. crescentus, was constructed. The resulting pEKEx3-xylXAD Cc -opt plasmid was then transferred to C. glutamicum P O6 iolT1, which is characterized by deregulation of the myo-inositol/proton symporter gene iolT1. Growth of the resulting strain C. glutamicum P O6 iolT1 pEKEx3-xylXAD Cc -opt was compared to that of C. glutamicum P O6 iolT1 pEKEx3-xylXABCD Cc -opt bearing the complete xylXABCD operon from C. crescentus ( Fig. 2A). Surprisingly, growth of both strains was indistinguishable (μ max ¼ 0.26 AE 0.006 h À1 , μ max ¼ 0.26 AE 0.004 h À1 , respectively), indicating that heterologous expression of the xylose dehydrogenase (encoded by xylB) and the xylonolactonase (encoded by xylC) is neither necessary nor beneficial for growth of C. glutamicum.
Motivated by these results, we performed a genome-wide search based on sequence similarity to identify genes potentially encoding for enzymes with XylX-, XylD-or XylA-activity in C. glutamicum ATCC 13032. These analyses suggested that the gene cg0535 could encode for an enzyme having a α-ketoglutarate semialdehyde dehydrogenase activity. However, the predicted protein Cg0535 shares only 25 % sequence identity with XylA of C. crescentus. For a better assessment, secondary structures of Cg0535 and XylA were calculated and aligned using PROMALS3D (PROfile Multiple Alignment with predicted Local Structures and threedimensional constraints) ( Supplementary Fig. S1) (Pei et al., 2008). This in silico analysis revealed a striking resemblance between both proteins with regard to their secondary structure triggering further investigations.
To the best of our knowledge, nothing about regulation and expression of cg0535 in C. glutamicum is known. Nonetheless, C. glutamicum P O6 iolT1 pEKEx3-xylXD Cc -opt with a further reduced operon was constructed to find out whether heterologous expression of xylA from C. crescentus is required for establishing the Weimberg pathway in C. glutamicum.
Comparative cultivation of C. glutamicum P O6 iolT1 pEKEx3-xylXD Cc -opt and C. glutamicum P O6 iolT1 pEKEx3-xylXAD Cc -opt revealed that C. glutamicum indeed does have an endogenous α-ketoglutarate semialdehyde dehydrogenase as both strains exhibited the same growth rate (μ max ¼ 0.26 AE 0.008 h À1 , μ max ¼ 0.26 AE 0.006 h À1 , respectively) ( Fig. 2A). Subsequently, cg0535 was deleted in the genome of C. glutamicum P O6 iolT1, yielding C. glutamicum P O6 iolT1 Δcg0535. After transformation of this strain with pEKEx3-xylXD Cc -opt, the resulting strain C. glutamicum P O6 iolT1 Δcg0535 pEKEx3-xylXD Cc -opt was compared to its parent strain C. glutamicum P O6 iolT1 pEKEx3-xylXD Cc -opt. These experiments showed that deletion of cg0535 completely abolished growth of C. glutamicum P O6 iolT1 Δcg0535 pEKEx3xylXD Cc -opt confirming that Cg0535 indeed has α-ketoglutarate semialdehyde dehydrogenase activity (data not shown). The inability of the cg0535 deletion mutant to grow on D-xylose also indicates that Cg0535 appears to be the only endogenous enzyme of C. glutamicum significantly contributing to α-ketoglutarate semialdehyde oxidation, at last under the cultivation conditions tested.
With the aim to characterize Cg0535 in more detail, the cg0535 gene was isolated from the genome of C. glutamicum ATCC 13032 by PCR and cloned into the pET-28b(þ)-vector for heterologous expression in E. coli BL21 (DE3). Gene expression in E. coli at 100 mL-scale and subsequent protein purification by affinity chromatography yielded 0.6 mg Cg0535 protein. Subsequently, we performed in vitro dehydrogenase assays using α-ketoglutarate semialdehyde as substrate and NAD or NADP as cofactors to determine selected kinetic parameters of Cg0535. These in vitro experiments confirmed the assumed α-ketoglutarate semialdehyde dehydrogenase activity of this enzyme and furthermore revealed a preference for the cofactor NAD as the calculated specific activity (U mg À1 ) with NAD was three times higher compared to the activity with NADP (51.8 μmol min À1 mg À1 and 15.8 μmol min À1 mg À1 , respectively) (Supplementary Fig. S2). Therefore, depending on the cofactor used, different Michaelis constants (K m ) for α-ketoglutarate semialdehyde could be calculated (NAD, 0.87 mM; NADP, 0.21 mM). Considering these findings, we would like to introduce the designation ksaD (α-ketoglutarate semialdehyde dehydrogenase) for cg0535 of C. glutamicum.

Expression of xylD or xylX enables growth on D-xylose
Analysis of the C. glutamicum genome did not identify any hitherto unknown dehydratases potentially catalyzing the two subsequent dehydration reactions of the Weimberg pathway. Noteworthy, enzyme assays conducted with the dehydratases XylX and XylD of C. cresentus showed that both dehydratases accept D-xylonate as substrate (Dahms and Donald, 1982). Since both dehydratase substrates D-xylonate and 2-keto-3-deoxy-xylonate of the Weimberg pathway, are chemically quite similar, it makes one wonder why two separate enzymes appear to be necessary (Fig. 1). Unfortunately, no experimental data shedding more light on this interesting aspect are available for the enzymes of C. crescentus. However, a conducted comparison of both enzymes as part of this study revealed only a low sequence identity (18 %) and an analysis using PROMALS3D suggested two very different secondary structures (data not shown). Nevertheless, driven by curiosity, the plasmids pEKEx3-xylX Cc -opt and pEKEx3-xylD Cc -opt were constructed and individually introduced into C. glutamicum P O6 iolT1. The resulting strains C. glutamicum P O6 iolT1 pEKEx3-xylX Cc -opt and C. glutamicum P O6 iolT1 pEKEx3-xylD Cc -opt were compared with regard to growth to C. glutamicum P O6 iolT1 pEKEx3 (Fig. 2B). As a result, both strains expressing either xylX or xylD could grow on this defined medium with D-xylose as sole carbon and energy source, whereas C. glutamicum P O6 iolT1 could not. The growth rates of C. glutamicum P O6 iolT1 pEKEx3-xylX Cc -opt and C. glutamicum

respectively). Apparently, both dehydratases
can complement for each other in C. glutamicum and heterologous expression of either xylX or xylD from C. crescentus is sufficient for enabling D-xylose utilization via the Weimberg pathway in C. glutamicum P O6 iolT1.

α-ketoglutarate synthesis via the Weimberg pathway
The Weimberg pathway represents a shortcut to the biotechnologically interesting TCA-cycle intermediate α-ketoglutarate without loss of carbon as compared to α-ketoglutarate synthesis starting from D-glucose (Fig. 1) (Jo et al., 2012). Nevertheless, microbial production of this dicarboxylic acid from D-xylose via the Weimberg pathway with C. glutamicum has not been investigated, yet. Within the TCA-cycle of C. glutamicum, the large multienzyme α-ketoglutarate dehydrogenase complex (ODHC) is responsible for the oxidative decarboxylation of α-ketoglutarate (Usuda et al., 1996;Bott, 2007). ODHC is comprised of three subunits: E1o (α-ketoglutarate decarboxylase, OdhA), E2 (dihydrolipoamide acetyl/succinyl-transferase, AceF) and E3 (dihydrolipoamide dehydrogenase, Lpd). It could be shown previously, that deletion of odhA results in the accumulation of α-ketoglutarate (Asakura et al., 2007). With the aim of establishing microbial α-ketoglutarate production from D-xylose via the Weimberg pathway in C. glutamicum, odhA was also deleted in C. glutamicum P O6 iolT1. Initially, the resulting strain C. glutamicum P O6 iolT1 ΔodhA was cultivated in defined CGXII medium supplemented with 40 g L À1 D-glucose as the sole carbon and energy source to find out if this is able to overproduce α-ketoglutarate from this hexose. Within 120 h, this strain accumulated 5.76 AE 0.06 g L À1 (39.43 AE 0.4 mM) α-ketoglutarate in the supernatant (Fig. 3). In contrast, the parent strain C. glutamicum P O6 iolT1 without deletion of odhA accumulated only 0.05 AE 0.00 g L À1 (0.37 AE 0.03 mM) α-ketoglutarate.
Subsequently, C. glutamicum P O6 iolT1 ΔodhA was transformed with pEKEx3-xylXABCD Cc -opt to find out whether the resulting strain accumulates more α-ketoglutarate in D-glucose/D-xylose mixtures. Noteworthy, deletion of odhA interrupting the TCA-cycle renders cultivation on D-xylose as sole carbon and energy source impossible. Here, C. glutamicum P O6 iolT1 ΔodhA pEKEx3-xylXABCD Cc -opt accumulated 7.92 AE 0.13 g L À1 (54.21 AE 0.86 mM) α-ketoglutarate in the supernatant when cultivated in defined CGXII medium with 10 g L À1 D-glucose and 30 g L À1 D-xylose (Fig. 3). In comparison to cultivation of C. glutamicum P O6 iolT1 ΔodhA in defined medium containing only D-glucose, the product titer could be increased 1.5-fold. Motivated by these findings, C. glutamicum P O6 iolT1 ΔodhA pEKEx3-xylX Cc -opt and C. glutamicum P O6 iolT1 ΔodhA pEKEx3-xylD Cc -opt were also constructed and characterized with regard to their α-ketoglutarate production capabilities on D-glucose/D-xylose mixtures. Interestingly, both strains accumulated pEKEx3-xylXABCD Cc -opt (black), C. glutamicum P O6 iolT1 pEKEx3-xylXAD Cc -opt (brown) and C. glutamicum P O6 iolT1 pEKEx3-xylXD Cc -opt (red); (B) C. glutamicum P O6 iolT1 pEKEx3-xylX Cc -opt (cyan), C. glutamicum P O6 iolT1 pEKEx3-xylD Cc -opt (green) and C. glutamicum P O6 iolT1 pEKEX3 (orange). All strains were cultivated in a BioLector microbioreactor system using defined CGXII medium with 40 g L À1 D-xylose as sole carbon and energy source. All data represent mean values from three biological replicates. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) much less α-ketoglutarate in the supernatant compared to the strain with the full xylXABCD-operon (1.27 AE 0.1 g L À1 (8.71 AE 0.7 mM) and 1.26 AE 0.0 g L À1 (8.62 AE 0.0 mM, respectively). This was somewhat surprising, as these results hint towards a limitation of the flux through the Weimberg pathway during product formation, which was not observable during growth experiments with C. glutamicum strains without odhA-deletion. Subsequent construction and characterization of C. glutamicum P O6 iolT1 ΔodhA pEKEx3-xylXD Cc -opt, bearing the plasmid for expression of both dehydratase genes from C. crescentus, supports the hypothesis of a restricted flux through the Weimberg pathway in this strain background because an increased α-ketoglutarate concentration of 3.30 AE 0.09 g L À1 (22.61 AE 0.65 mM) could be determined in the supernatant.

Discussion
Functional introduction of a pathway from another organism or implementation of a novel synthetic pathway usually means addition of new enzymatic activities to the catalytic repertoire of the respective host organism. However, sometimes the "new" enzymes have overlapping substrate specificities with endogenous enzymes rendering their introduction unnecessary. This could be already shown for C. glutamicum R and C. glutamicum ATCC 13032 when establishing the two-step Isomerase pathway for D-xylose utilization as both strains already have a xylulokinase (XylB) and thus only require a heterologous gene encoding for a xylose isomerase (Kawaguchi et al., 2006). In case of Pseudomonas sp., it could be demonstrated that the periplasmic glucose dehydrogenase (Gcd) also contributes to D-xylose utilization via the Weimberg pathway (K€ ohler et al., 2015;Meijnen et al., 2009). Similarly, an enzymatic study not directly connected to growth on D-xylose revealed that the two endogenous myo-inositol dehydrogenases IolG1 and IolG2 of Lactobacillus casei BL23 can convert D-xylose to D-xylonate similar to IolG of C. glutamicum (Aamudalapalli et al., 2018).
The endogenous α-ketoglutarate semialdehyde dehydrogenase (KsaD) of C. glutamicum discovered in the context of this study is characterized by a high specific activity (51.8 μmol min À1 mg À1 ), which is comparable to that of the enzyme with the same activity in P. putida (53 μmol min À1 mg À1 ) (Adams and Rosso, 1967). In the genome of C. glutamicum, the open reading frame of ksaD overlaps with that of cg0536 encoding for a putative 5-dehydro-4-deoxyglucarate dehydratase. This finding hints towards a potential role of KsaD in a putative oxidative pathway for the utilization of sugar acids such as D-galacturonic acid or D-glucuronic acid as theses pathways require α-ketoglutarate semialdehyde dehydrogenase-and 5-dehydro-4-deoxyglucarate dehydratase activities (Richard and Hilditch, 2009;Pick et al., 2016). It sounds reasonable that C. glutamicum has such a catabolic pathway as these sugar derivatives typically to be found in pectin-rich fruits and vegetables such as grapes, apples, bean sprouts should be readily available in the natural habitat of this soil bacterium (Li et al., 2016).
Heterologous expression of either xylX or xylD in C. glutamicum P O6 iolT1 enables growth in D-xylose containing media, indicating that both dehydratases from C. crescentus catalyze both dehydration reactions of the Weimberg pathway in C. glutamicum. In contrast, a P. putida S12 strain equipped with the Weimberg pathway from C. crescentus inevitably requires the expression of xylD whereas heterologous expression of xylX alone is not sufficient for enabling growth on D-xylose (Meijnen et al., 2009). In this case, it was assumed that the endogenous dehydratase PP2836 of P. putida S12 exhibiting 57 % sequence identity to XylX from C. crescentus renders heterologous xylX expression unnecessary. However, this is somewhat puzzling as it would mean that the two dehydratases from C. crescentus cannot complement for each other. Unfortunately, the importance of having both dehydratase has not been studied in C. crescentus as the natural source of both enzymes yet. A detailed kinetic characterization of both dehydratases could shed more light on this important aspect. Noteworthy in this context, the archaeon Haloferax volcanii, naturally having the Weimberg pathway, requires the activity of both dehydratases (HVO_B0038A and HVO_B0027) for growth on D-xylose containing media (Johnsen et al., 2009).
In our experiments, microbial synthesis of α-ketoglutarate from a Dglucose/D-xylose mixture with engineered C. glutamicum strains having the Weimberg pathway turned out to be more beneficial for product formation compared to cultivations using D-glucose as only substrate. This could be a direct consequence of the carbon efficiency of the Weimberg pathway offering a theoretical product yield of 100 %. In contrast, α-ketoglutarate synthesis from D-glucose is always accompanied by loss of carbon as CO 2 during isocitrate oxidation in the TCA-cycle, which eventually only allows for a maximum theoretical yield of 83 %. However, we could observe that reduction of the xylXABCD-operon also reduced final product concentrations in the constructed odhA-deletion strains. At this stage, we can only speculate that deletion of odhA, necessary for the accumulation of significant amounts of α-ketoglutarate, causes this effect as this is the only genetic difference to the other Dxylose consuming C. glutamicum strains evaluated in the context of xyl-XABCD-operon reduction. However, this indicates that heterologous expression of the whole pentacistronic xylXABCD-operon might not be necessary for growth of C. glutamicum in D-xylose containing defined medium, but is beneficial for product formation via the Weimberg pathway, especially in more engineered strains.

Conclusions
Reduction of the Weimberg pathway encoding operon from C. crescentus revealed that sole expression of xylX (2-keto-3-deoxy-xylonate-dehydratase) or xylD (xylonate dehydratase) is sufficient for establishing this five-step pathway in C. glutamicum. Reason for this is that C. glutamicum is already equipped with two dehydrogenases conferring the capacity to oxidize D-xylose and α-ketoglutarate semialdehyde. A lactonase converting 1,4-D-xylonolactone to D-xylonate is not required as hydrolyzation of this lactone can occur spontaneously. Conducted experiments employing the carbon efficient Weimberg pathway for the microbial synthesis of α-ketoglutarate indicate that D-xylose might represent a more suitable substrate for the production of this organic acid compared to D-glucose. Fig. 3. Accumulation of α-ketoglutarate during shake flask cultivations of different C. glutamicum strains in defined CGXII medium supplemented with either 40 g L À1 D-glucose or a mixture of 10 g L À1 D-glucose and 30 g L À1 Dxylose. C. glutamicum P O6 iolT1 (orange, 40 g L À1 D-glucose), C. glutamicum P O6 iolT1 ΔodhA (purple, 40 g L À1 D-glucose), C. glutamicum P O6 iolT1 ΔodhA pEKEx3-xylXABCD Cc -opt (black, 10 g L À1 D-glucose and 30 g L À1 D-xylose), C. glutamicum P O6 iolT1 ΔodhA pEKEx3-xylXD Cc -opt (red, 10 g L À1 D-glucose and 30 g L À1 Dxylose), C. glutamicum P O6 iolT1 ΔodhA pEKEx3-xylX Cc -opt (cyan, 10 g L À1 Dglucose and 30 g L À1 D-xylose), C. glutamicum P O6 iolT1 ΔodhA pEKEx3-xylD Cc -opt (green, 10 g L À1 D-glucose and 30 g L À1 D-xylose). The data represent mean values and standard deviations obtained from three independent cultures. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)