The production of plastics from petroleum-based raw materials extensively contributes to global pollution and CO₂ emissions. While bio-based polymers from crops produce significantly lower carbon emissions and have lower energy production requirements (Meereboer et al., 2020), their scalability is limited by competition with food industries, acreage of land, and consumption of water and nutrients (Chia et al., 2020). Additionally, climate change severely affects the agricultural industry by extreme weather anomalies (Beniston et al., 2007); in particular, the cultivation of crops such as grains and maize, as well as vegetable and fruit production was affected. Consequently, it is essential for a biotechnological industry to use agricultural feedstocks more efficiently. One main strategy is the shift toward feedstocks that are not competitive with human and animal nutrition. This includes the use of industrial sidestreams as well as lignocellulose-derived sugars like xylose and arabinose (Wendisch et al., 2016). Most industrial platform strains are not able to naturally utilize these nonfood feedstocks or produce value-added compounds in high titers in their wild-type form (Wendisch et al., 2016). Metabolic engineering and systems biology were indispensable to establish microbial cell factories independent from glucose (Dai and Nielsen, 2015; Erb et al., 2017). In recent years, the model organism C. glutamicum was extensively engineered to broaden its substrate spectrum according to the flexible feedstock concept (Wendisch et al., 2016). Thereby, access to the lignocellulosic pentoses arabinose (Kawaguchi et al., 2008; Schneider et al., 2011; Meiswinkel et al., 2013) and xylose (Kawaguchi et al., 2006; Jin et al., 2020) for growth and production of amino acids was enabled. Recently, it could also be demonstrated that production of sarcosine from the pentose xylose was more efficient than that from glucose (Mindt et al., 2019b). Additionally, less processed substrates such as lignocellulosic residuals, rich straw, wheat bran and plant mass hydrolysates, chitin, the biorefinery sidestream pyrolysis water, and wheat sidestream concentrate (WSC) from the starch and paper industries could be harnessed by C. glutamicum as carbon sources (Gopinath et al., 2011; Lange et al., 2017; Mindt et al., 2019a; Sasaki et al., 2019; Prell et al., 2021a; Vortmann et al., 2021).

A special focus lies on the fermentative production of amino acids by C. glutamicum, which has been well established over the past decades (Wendisch, 2020). In general, amino acids have a wide spectrum of applications in the food, feed, and pharmaceutical industries. The annual production of l-glutamate and l-lysine reached 3,210,000 and 2,600,000 metric tons, respectively (Wendisch, 2020). The production of l-lysine from alternative carbon sources has been successfully established from whey (Barrett et al., 2004) and starch (Seibold et al., 2006). The high industrial impact of the amino acid production by C. glutamicum formed the basis for the synthesis of amino acid–based derivatives (Pérez-García et al., 2017, 2018; Mindt et al., 2018; Veldmann et al., 2019). Some prominent derivatives of the amino acid l-lysine are the bifunctional monomers cadaverine, 5-aminovalerate (5AVA), and glutarate, which are monomeric precursors for the production of bio-polyamides. As plastics are primarily synthesized from natural gas and petroleum, the worldwide interest in bio-based production of the functional monomers is steadily increasing. Polyamides can be either obtained by anionic ring-opening polymerization of lactams, the cyclization products of ω-amino acids, or by condensation of dicarboxylic acids with diamines (Radzik et al., 2020). The production of these monomeric building blocks for polyamides has been established in metabolically engineered C. glutamicum and E. coli (Chae et al., 2020; Wendisch, 2020). Diamines like putrescine (Qian et al., 2009; Schneider et al., 2012) and cadaverine (Na et al., 2013; Kind et al., 2014) as well as the ω-amino acids γ-4-aminobutyrate (GABA) (Choi et al., 2015; Jorge et al., 2017a) and 5AVA (Jorge et al., 2017b; Haupka et al., 2020), and the dicarboxylic acids succinate and glutarate (Okino et al., 2008; Pérez-García et al., 2018; Han et al., 2020) were successfully produced in high titers.

The C5-ω-amino acid 5AVA, the monomeric precursor of the polyamide 5 (PA 5) (Chae et al., 2017; Zhang et al., 2017), can be derived from l-lysine by four different (synthetic) pathways. In the first pathway, l-lysine is converted to 5AVA by l-lysine-α-oxidase (RaiP) from Scomber japonicas by oxidative deamination and a spontaneous decarboxylation step (Cheng et al., 2020). Recently, an alternative synthetic route was established starting with RaiP, but the intermediate 2-keto-6-aminocaproate is converted by α-ketoacid decarboxylase (KivD) from Lactococcus lactis and aldehyde dehydrogenase (PadA) from Escherichia coli to 5AVA (Cheng et al., 2021). The second pathway to 5AVA combines oxidative decarboxylation by l-lysine monooxygenase (DavA) using molecular oxygen followed by desamidation by γ-aminovaleramidase (DavB) from P. putida (Adkins et al., 2013). The third and the fourth pathways are both initiated by decarboxylation of l-lysine catalyzed by l-lysine decarboxylase (LdcC) from E. coli (Jorge et al., 2017b; Haupka et al., 2020). The third pathway employs putrescine oxidase Puo from Rhodococcus qingshengii, which requires molecular oxygen and NAD-dependent γ-aminobutyraldehyde dehydrogenase PatD from E. coli for conversion of cadaverine to 5AVA (Haupka et al., 2020). The fourth pathway does not require molecular oxygen and converts cadaverine to 5AVA through 2-oxoglutarate-dependent putrescine/cadaverine transaminases PatA and PatD from E. coli (Jorge et al., 2017b). Later, this pathway was also used with alternative enzymes, namely l-lysine decarboxylase CadA from E. coli, 2-oxoglurate-dependent putrescine transaminase KpcPA and NAD-dependent γ-aminobutyraldehyde dehydrogenase KpcPD from Klebsiella pneumoniae (Wang et al., 2021). The transaminase reaction, e.g., in the LdcC-PatA-PatD cascade, yields l-glutamate from 2-oxoglutarate. Flux enforcement in a metabolic setup in which growth requires production of 5AVA can be achieved by deletion of gdh, the gene coding for the major ammonium assimilating enzyme l-glutamate dehydrogenase as shown for glutarate production (Pérez-García et al., 2018; Haupka et al., 2020). In a comparison of two cascades with either one (LdcC-PatA-PatD-GabT-GabD) or two transaminase reactions (LdcC-Puo-PatD-GabT-GabD), a 1:1 stoichiometry proved to be superior to coupling with two transaminases providing l-glutamate (Pérez-García et al., 2018; Haupka et al., 2020).

To produce bio-based polyamides from sustainable resources, C. glutamicum was engineered to produce cadaverine from starch (Tateno et al., 2007, 2009) and from xylose (Buschke et al., 2013). Also, other model organisms like E. coli have been successfully engineered, and cadaverine could be produced from soybean hydrolysates (Guo et al., 2021). Aside from cadaverine, ω-amino acids GABA and 5AVA were produced in C. glutamicum from empty fruit bunch sugars (Baritugo et al., 2018) and from a Miscanthus and a rice straw hydrolysate (Joo et al., 2017; Sasikumar et al., 2021), respectively. As hydrolysates are mainly obtained directly from grains, a new perspective would be the processing of sidestreams from industrial processes. It was shown before that WSC from the paper industry can be applied as a medium for the production of the dicarboxylic acid glutarate and the trifunctional molecule l-2-hydroxyglutarate (Prell et al., 2021a). In this study, the WSC and different hydrolysates derived from it were investigated for production of l-lysine and its derivatives cadaverine and 5AVA (Figure 1).

In this study, WSC from the paper industry was shown to be a sustainable feedstock for the fermentative production of the amino acid l-lysine, the diamine cadaverine, and the ω-amino acid 5AVA. Hydrolysis of the sidestream and enabling access to pentose utilization improved production. Deletion of gdh reduced formation of the by-product glutamate and coupled growth to production by flux enforcement. Strain AVA1-G produced 5.9 ± 0.2 g L⁻¹ 5AVA from WSCH.

Utilization of hydrolysates from alternative feedstocks has been demonstrated for C. glutamicum before for products such as l-glutamate and l-lysine (Gopinath et al., 2011), N-ethylglycine (Mindt et al., 2019a), and succinate (Mao et al., 2018). Similar to those studies, among the tested hydrolysates of WSC with acid, base, and water, it was found to be most efficient with 4% (w/v) H₂SO₄ as judged from the yield of glucose, xylose, and arabinose. During acidic hydrolysis, starch and maltose are degraded to glucose (BeMiller and Mann, 1966; Abbadi et al., 1998), which explains the high glucose concentration in WSCH_H2SO4. Hydrolysis of cellulose and hemicellulose from lignocellulose could have led to the release of glucose, xylose, and arabinose (Debiagi et al., 2020). Typically, lignocellulosic hydrolysates contain growth inhibitors like furfural, hydroxymethylfurfural, formic acid, and phenolic compounds (Palmqvist and Hahn-Hägerdal, 2000), to which C. glutamicum possesses moderate tolerance, and in growth-arrested C. glutamicum R cells, productivity has been shown to be little or not affected in the presence of various inhibitors (Sakai et al., 2007). In this study, a growth-inhibiting effect of high WSCH concentrations was observed (data not shown), likely due to sugar-derived inhibitors resulting from hydrolysis of lignocellulose. In order to increase WSCH concentrations, reduce lag phases, and reach higher growth rates, hydrolysates would need to be detoxified. Many approaches for detoxification of hydrolysates have been developed. In C. glutamicum S9114, several oxidoreductases were identified which helped to increase growth and inhibitor conversion. Overexpression of CGS9114_RS01115 led to increased conversion of five inhibitory aldehydes including furfural, hydroxymethylfurfural, 4-hydroxybenzaldehyde, vanillin, and syringaldehyde (Zhou et al., 2019). Koopman et al. identified the furfural and hydroxymethylfurfural degradation pathways of Cupriavidus basilensis HMF14 and enabled Pseudomonas putida to detoxify the aldehydes in situ and utilize them as a carbon source by degrading them to 2-oxoglutarate (Koopman et al., 2010). In contrast to targeted gene expression for in situ detoxification, several studies reported increased tolerance to inhibitors in lignocellulosic hydrolysates by UV mutagenesis of Scheffersomyces shehatae (Senatham et al., 2016) or adaptive laboratory evolution as shown for C. glutamicum S9114 in corn stover hydrolysates (Wang et al., 2018). However, these approaches should be accompanied by whole genome re-sequencing to exclude mutational changes at the expense of production. As an alternative to increase tolerance against inhibitors in hydrolysates, hydrolysates can be prepared or treated enzymatically to reduce the accumulation of inhibitors. Treatment with laccase and lignin peroxidase removed phenolic compounds almost completely (Chen et al., 2019). Preparation of hydrolysates with the cellulolytic enzyme reagent Cellic^® Ctec2 instead of acidic hydrolysis prevented the accumulation of inhibitors (Mao et al., 2018). Detoxified WSCH could improve the growth of production strains and ultimately, productivity, without the need for heterologous expression of additional genes or adaptive laboratory evolution.

As the chemical composition in wheat and wheat hydrolysates varies in different years and batches (Rosicka-Kaczmarek et al., 2013), hydrolysates from two batches were compared in this study (Figure 3). WSCH Lot2 contained more xylose and arabinose than Lot1, which could be utilized for enhanced l-lysine and cadaverine production by the expression of the genes xylA _Xc B _Cg and araBAD _Ec . Apart from sugars, hydrolysates may contain aromatic compounds, which can be consumed by C. glutamicum, e.g., benzoate, phenol, 3-hydroxybenzoate, protocatechuate, vanillate, benzyl alcohol, and ferulate (Shen et al., 2012). Strikingly, the strain LYS not only produced 55 ± 4 mM l-lysine in WSCH Lot2, but also additional 23 ± 1 mM l-glutamate, while producing almost no l-glutamate in Lot1. This cannot be attributed to a difference in the initial l-glutamate concentrations in Lot1 (5 mM) and Lot2 (6 mM) and, therefore, must be due to regulatory effects. In C. glutamicum, l-glutamate production is triggered, i.a., by biotin limitation (Eggeling and Sahm, 1999). It has been reported that biotin availability in different batches of wheat varies and that the content of free biotin and its bioavailability are much lower in wheat compared to that in corn, barley, and oats (Frigg, 1976; Anderson et al., 1978; Bryden et al., 1991). Thus, biotin limitation might play a role here, other than in the biotin-rich corn stover hydrolysate, for which C. glutamicum S9114 had to be metabolically engineered in order to overcome inhibition by biotin and to produce l-glutamate (Wen and Bao, 2019). The key enzyme to regulate l-glutamate overproduction, the 2-oxoglutarate dehydrogenase complex (ODHC), is the target of phosphorylation by the inhibitor OdhI, and OdhI is phosphorylated by PknG. On the one hand, WSCH Lot2 might contain inhibitors of ODHC like cis-aconitate, oxalacetate, and pyruvate (Shiio and Ujigawa-Takeda, 1980), and potentially, OdhI and PknG might be influenced by effectors in WSCH as well with the effect of ODHC repression, leading to increased l-glutamate production. On the other hand, under l-glutamate producing conditions, the global posttranslational protein acetylation in C. glutamicum decreases and succinylation increases, affecting proteins in carbohydrate metabolism, translation, and amino acid metabolism (Mizuno et al., 2016). If perhaps increased succinylation status is triggered in WSCH Lot2 compared to that in Lot1, this might lead to increased l-glutamate production.

When the l-lysine pathway was extended to cadaverine and 5AVA, incomplete conversion of l-lysine was observed in WSCH (Figures 5, 6, 8), making the decarboxylation of l-lysine via LdcC the limiting step in the pathway. Since this was not observed in minimal medium, either LdcC is inhibited by an unknown inhibitor in WSCH or its cofactor pyridoxal 5′-phosphate (PLP) is limiting. PLP and l-glutamate are synthesized from d-ribose 5-phosphate, d-glyceraldehyde 3-phosphate, and l-glutamine via the PLP synthase complex PdxST (Jochmann et al., 2011). Since l-glutamate accumulates in WSCH, l-glutamine synthesis via glutamine synthetase GlnA might be hampered, resulting in low l-glutamine content and, consequently, low PLP synthesis. Whether PLP limitation, in fact, is the cause for incomplete l-lysine decarboxylation could be investigated by PLP addition to the WSCH medium. In E. coli whole-cell biocatalysis, supplementation of PLP or heterologous expression of pdxS and pdxT from Bacillus subtilis increased cadaverine productivity by 2.9 folds (Ma et al., 2015).

Flux enforcement can be used as a means to couple production to another condition to increase the flux through the targeted pathway. By deletion of the l-glutamate dehydrogenase–encoding gene gdh, synthesis of l-glutamate is impaired, providing the opportunity to couple growth to production of 5AVA via 2-oxoglutarate-dependent putrescine/cadaverine transaminase PatA. Haupka et al. have shown that flux enforcement is more effective with one coupling site than with two (Haupka et al., 2020). Deletion of gdh in the strain LYS-G resulted in 42% decreased l-lysine production (28 ± 2 mM) in minimal medium (Figure 7) due to a missing coupling site to synthesize l-glutamate. Prell et al. showed that nitrogen starvation was partially triggered by gdh deletion, leaving the GS/GOGAT system as a l-glutamate replenisher with an increased energy demand compared to Gdh, which resulted in decreased growth rates in glutarate-producing strains (Prell et al., 2021b). In AVA1-G, however, which possesses a coupling site for l-glutamate synthesis via PatA, 40 ± 2 mM 5AVA was produced, thereby exceeding the l-lysine production in LYS-G, which hints at the effectiveness of flux enforcement. This is in line with the lower 5AVA production in AVA2-G, which has no coupling site due to use of the putrescine oxidase gene puo. In WSCH, the production of l-lysine and 5AVA by the strains LYS-G, AVA1-G, and AVA2-G exceeded production in minimal medium (Figures 7, 8) with AVA1-G as the highest 5AVA producer in WSCH (50 ± 2 mM). Even without the flux enforcement, the deletion of gdh improved production in WSCH as 2-oxoglutarate is not converted to l-glutamate, but further driven through the TCA cycle toward oxaloacetate, the precursor for l-lysine biosynthesis. Since WSCH is rich in amino acids including l-glutamate, the nitrogen starvation response, regulated by AmtR (Schulz et al., 2001), might have been canceled despite of gdh deletion in these strains, resulting in an escape of the metabolic deficit that is accompanied by deletion of gdh and the nitrogen starvation response. In a glutarate-producing C. glutamicum strain carrying a gdh deletion, which was evolved by adaptive laboratory evolution, RNAseq analysis revealed that many genes of the AmtR regulon were downregulated in comparison with the unevolved strain due to the amino acid exchange E686Q in the large subunit GltB of GOGAT (Prell et al., 2021b). The mechanism for nitrogen sensing in C. glutamicum is not known yet, but it is conceivable that the nitrogen starvation response in AVA1-G is downregulated in WSCH compared to that in minimal medium as it is for the evolved glutarate producer.

The fundamental establishment of WSCH as a medium for production of 5AVA in C. glutamicum was shown here; however, a scale-up approach is necessary in order to reach industrial relevance. C. glutamicum has been employed for bioreactor fermentation with several hydrolysates. Rice straw hydrolysate was used as a carbon source in CGXII medium for the production of 1.6 g L⁻¹ N-ethylglycine in a fed-batch cultivation with a 2 L starting volume (Mindt et al., 2019a). Miscanthus hydrolysate was utilized as a carbon source in CG100 medium for the production of 12.5 g L⁻¹ 5AVA in a 500 mL batch cultivation (Joo et al., 2017). For the production of succinate in 3.5 L batch fermentation, C. glutamicum cells were immobilized in the bioreactor using a porous polyurethane filler, and cassava bagasse hydrolysate was used yielding 22.5 g L⁻¹ succinate (Shi et al., 2014). WSCH can substitute most of the components of CGXII minimal medium; therefore it is likely that for upscaling of 5AVA production, a similar mixture of WSCH and ammonium sulfate can be utilized as it has been used here, with the exception of MOPS buffer, since pH is regulated automatically during bioreactor fermentation.

Future applications of WSCH may target glutarate (Pérez-García et al., 2018) and l-2-hydroxyglutarate (Prell et al., 2021a) since the pathway only needs to be extended by additional two and three reactions, respectively. Both compounds were produced from unhydrolyzed WSC (Prell et al., 2021a). Furthermore, reactive extraction has been proven to be an effective method for downstream processing of glutarate from CGXII fermentation broth (Prell et al., 2021b). In future, it has to be studied if the developed reactive extraction/re-extraction methods can be applied to WSCH-based production processes.