Microbial cell factories for the sustainable manufacturing of B vitamins

Vitamins are essential compounds in human and animal diets. Their demand is increasing globally in food, feed, cosmetics, chemical and pharmaceutical industries. Most current production methods are unsustainable because they use nonrenewable sources and often generate hazardous waste. Many microorganisms produce vitamins naturally, but their corresponding metabolic pathways are tightly regulated since vitamins are needed only in catalytic amounts. Metabolic engineering is accelerating the development of microbial cell factories for vitamins that could compete with chemical methods that have been optimized over decades, but scientific hurdles remain. Additional technological and regulatory issues need to be overcome for innovative bioprocesses to reach the market. Here, we review the current state of development and challenges for fermentative processes for the B vitamin group.


Introduction
Vitamins are essential organic compounds functioning mainly as coenzymes of metabolic reactions in all organisms. The 13 vitamins are classified as fat-soluble (A, D, E, K) or water-soluble (C, B group). Plants and microorganisms naturally produce vitamins, while humans and animals need to consume them [1 ]. Vitamin demand is increased by pathological conditions, malnutrition, poor diet, high exercise, pregnancy, stress and drug abuse [1 ].
Moreover, food processing and preservation methods can decrease vitamin content, as vitamins are sensitive to light, heat, oxygen and pH [1 ]. Consequently, vitamins are produced for the feed, food, cosmetics, chemical and pharmaceutical industry.
Vitamins are currently produced by chemical synthesis and/or biotechnology [1 ]. Organic synthesis often employs non-renewable chemicals and produces hazardous waste. Besides concerns of sustainability, economics has been the major driving factor for the development of bioprocesses during the last decades. Vitamin B 2 is a remarkable example. Its organic synthesis employs toxic compounds [2 ], yet iterative rounds of mutagenesis and screening of Ashbya gossypii and Bacillus subtilis strains resulted in significant vitamin production. The global market volume of fermented vitamin increased from 5 to 75% from 1999 to 2012 [3 ]. Overall, the bioprocess reduced the environmental footprint and production costs by 43% [4].
Bioprocesses can have a positive impact on both the economics and the environment, but the question is whether other vitamins will follow. Thanks to an improved understanding of the vitamin biosynthetic steps and regulation (mainly in microbes), and novel metabolic engineering tools [5 ], it should be possible to engineer high-yielding vitamin-producing strains. Additionally, high yield and purity vitamin purification from the fermentation broth must be achieved at an industrially acceptable cost. Here, we review how metabolic engineering challenges could be addressed for all B vitamins, and briefly discuss the importance of technical, environmental and regulatory issues. Only peer-reviewed work is considered. We do not cover B vitamin history, discovery, sources, detailed chemical synthesis, physiological role or deficiency [1 ,2 ].

Current biotechnological status on B vitamins
The global B vitamin market is large and growing by >4% annually. 5 Market size and cost depends on the vitamin and its application. By volume, the largest consumption is in feed, followed by the food, pharmaceutical and chemical industries, with most vitamins currently being manufactured by chemical synthesis. B vitamin fermentations   (Table 1).
Microbial fermentations have met limited success [15]. Before the pathway was fully elucidated, B. subtilis production strains with mutations in three genes were reported [16 ] (Table 2). Knocking out thiN increases thiamine production over TMP/TPP. Recently described TMP phosphatases [17] could further increase the dephosphorylated product. Aspergillus oryzae was engineered to produce 4-fold intracellular TPP by overexpressing three genes [18] (Table 2).
A riboswitch-based biosensor enabled the discovery [19 ] and engineering [20 ] of thiamine transporters, and an increased thiamine production in E. coli overexpressing thiFSGHCE and thiD or thiM combined with transposon mutagenesis [21 ]. This identified genes involved in iron/sulfur metabolism, consistent with the fact that ThiC/ThiH are Fe-S cluster-dependent enzymes catalyzing complex rearrangements. These enzymes are inhibited by S-adenosyl-methionine (SAM) metabolites and represent potential bottlenecks, with ThiC having very low activity (k cat = 0.14 min À1 ) [22 ,23 ,24]. Another hurdle is the branched nature of the pathway that requires balanced expression of many genes. Since thiamine fermentation will require high yields due to its low cost, major metabolic engineering breakthroughs will be needed.

Vitamin B 2
By 2018, riboflavin is exclusively produced by fermentation using A. gossypii and B. subtilis (Table 2). A comprehensive compilation of production strains is available [3 ] as well as systems biology data for B. subtilis [25 ].
The bacterial, archaeal and fungal biosynthesis pathway starts with one molecule of guanosine triphosphate (GTP) and two of ribulose 5-phosphate (R5P) (Figure 1b). GTP and phosphorylated pentoses are abundant in cells, and the biosynthetic enzymes for these riboflavin precursors are efficient. Conversely, the riboflavin metabolic enzymes are very slow [26 ], and it is thus necessary to strongly and stably overexpress the corresponding genes. Mutations in the flavokinase/ flavin adenine dinucleotide (FAD) synthetase gene and mRNA stabilization by modified flavin mononucleotide (FMN) riboswitches are important features of modern production strains.
However, there are still unresolved issues: (1) A superordinate regulator protein (RibR) was recently discovered in B. subtilis [27 ]. RibR affects the activity of FMN riboswitches and could limit production. (2) The phosphatase(s) of the riboflavin pathway has not been identified yet (early intermediates are phosphorylated) [28]. (3) Flavins are highly reactive molecules [29 ]. Intracellular flavins (and corresponding flavoproteins) can be a major source for reactive oxygen species (ROS), thus lowering yields by creating oxidative damage to cells. A way to reduce ROS-related cellular stress could be to introduce a riboflavin binding dodecin [30 ]. (4) The hydrophobic dimethylbenzene portion of the isoalloxazine ring system of riboflavin probably supports diffusion of riboflavin over the cytoplasmic membrane in the absence of a dedicated transport system. The B. subtilis cell appears not to be able to actively export flavins (in contrast to A. gossypii) [31] and introduction of the riboflavin transporter gene ribM from Streptomyces davaonensis enhanced riboflavin production [32]. Roseoflavin from S. davaonensis negatively affects FMN riboswitches and flavoenzymes [33], and it is used to select riboflavin overproducing bacteria [34]. This antimetabolite could be used to further improve vitamin production.   There are three forms of vitamin B 3 , all of which are converted to the essential cofactors enzymatically [35]. Although the commercial forms are niacin (NA) and the corresponding niacinamide (NAM), nicotinamide riboside (NR) has gained attention recently in treatment of age-related diseases [35,36]. Figure 1c illustrates the biosynthetic pathway to NAD + in E. coli and Saccharomyces cerevisiae. NA and NAM are breakdown products or intermediates in salvage pathways. NR can also be synthesized from NAD + using salvage enzymes [37,38].
There is no commercial fermentation process for NA or NAM [6] but biocatalytic routes exist that use3-cyanopyridine as a starting material that is hydrolysed to NA by a nitrilase or hydrated to NAM by a nitrile hydratase [7]. The enzymatic hydrolysis is catalyzed by whole-cell biocatalysts of engineered Rhodococcus rhodochrous [39]. Recent reports describe the use of recombinant E. coli expressing R. rhodochrous nitrile hydratase for NAM production [40]. A yeast strain producing extracellular NR by the native NAD + breakdown process (>8 mg/L NR with media supplementation of NA) was constructed by knocking out the NR importer Nrt1 [41 ] ( Table 2). With the high-yielding biocatalytic process in place there is little motivation for a fermentative process to NA or NAM, while fermentation of NR could be appealing if a market demand emerges.

Vitamin B 5
Pantothenic acid (pantothenate) is a precursor of coenzyme A (CoA) and the acyl carrier protein (ACP) that play essential roles in fatty acid metabolism [42,43]. The active form is pantothenate (panthenol and pantetheine are two more vitamers), and it is stable as a calcium salt, which is the commercial form [43].
De novo synthesis of vitamin B 5 occurs in microorganisms, plants and fungi. In B. subtilis and E. coli, it involves seven enzymatic reactions starting from pyruvate and aspartate [42]. Pantoate is generated from two molecules of pyruvate by the acetolactate synthase complex and 4 enzymes (Figure 1d). Pantoate and b-alanine (generated via decarboxylation of aspartate by PanD) are condensed to pantothenate by PanC.
Most microbial engineering efforts target genes involved in the metabolic pathway and cofactor generation, for example, inactivating threonine dehydratase (ilvA) and overexpressing ilvBNCD/panBC enhance biosynthesis of ketoisovalerate/pantothenate, resulting in 18 mg/L/h vitamin B 5 using C. glutamicum [44] (Table 2). Glycine accumulation was suggested to affect regeneration of 5,10-methylenetetrahydrofolate (Met-THF) [45], a cofactor of PanB. Hence, Met-THF supply seems to be an important bottleneck. Production could be further enhanced to <35 mg/L/h by maintaining the ketoisovalerate-consuming aminotransferase IlvE at low levels with b-alanine [44] ( Table 2). The highest titre was reported in B. subtilis (Table 2), and it could be improved by manipulating pantothenate kinase (coaA) to reduce pantothenate consumption [47 ].

Vitamin B 6
Pyridoxine (PN), pyridoxal (PL), pyridoxamine (PM) and their respective 5 0 -phosphate derivatives PNP, PLP, PMP compose this vitamin. Pyridoxine hydrochloride, the commercial form, is synthesized from PN. In vivo, PLP is the most important vitamer, serving as a cofactor for a plethora of enzymes mostly involved in amino acid metabolism [49].
Since the PdxST complex is catalytically inefficient, some organisms have been genetically engineered for producing PN/PL via the DXP-dependent pathway ( Table 2). A B. subtilis strain carrying the respective enzymes PdxA and PdxJ from E. coli and Sinorhizobium meliloti produced 900 mg/L/h PN from 4-hydroxy-L-threonine (4HT) ( Table 2) [51]. 4HT can be converted to 4HTP by homoserine kinase (ThrB) [51]. It seems that PdxJ is a bottleneck in the DXP-dependent pathway because it is an inefficient enzyme [52] that more efficient orthologs or protein engineering might improve. Little is known about PNP and PLP phosphatases, which may facilitate vitamers export and prevent accumulation of toxic intermediates like 4HTP [55]. Unfortunately, production levels should exceed 10 g/L in 48 h for being commercial relevant [55], so major metabolic engineering efforts are needed.

Vitamin B 7
D-(+)-Biotin (or biotin) is an essential cofactor for carboxylation reactions. Biotin intermediate pimelic acid is synthesized by two alternative routes [56]. In E. coli, fatty acid biosynthesis is hijacked by O-methylation of the malonyl-ACP by BioC, elongation by the fatty acid cycle and hydrolysis of the ester by BioH to form pimeloyl-ACP [57 ]. In B. subtilis a cytochrome P450-dependent enzyme (BioI) oxidatively cleaves a long-chain acyl-ACP to pimeloyl-ACP [58]. Pimeloyl-ACP feeds into the last four-enzyme pathway part (Figure 1f). Variations in the methylase/esterase (BioCH) pair and downstream (BioFADB) genes have been identified [59]. Biotin production is mainly regulated by a transcription factor that can function as a biotin ligase [60 ].
In the past, efforts for engineering biotin production strains using random mutagenesis and antimetabolites met limited success [61 ]. The highest biotin titer reported is 500 mg/L with Serratia marcescens in a $10-day fermentation ( Table 2). One challenge seems to be a high demand of the co-factor SAM by BioC, BioB and BioA, although these enzymes are strongly inhibited by SAM byproducts [62]. SAM levels can be increased via classical [63 ] or emergent [64 ] strain engineering methods, while non-feedback inhibited homologs could be found in metagenomic screens [48 ]. In B. subtilis, BioA is replaced by BioK, which uses lysine as amino donor instead of SAM. Lysine feeding to an engineered B. subtilis strain produced >600 mg/L desthiobiotin (direct biotin precursor) but only up to 21 mg/L biotin, indicating that the last step catalyzed by BioB is a bottleneck [65 ] ( Table 2).
BioB is a very slow enzyme (k cat = 0.12 min À1 ) and one of its Fe-S cluster needs to be regenerated and the other one repaired for each catalytic turnover [66 ]. Overexpressing BioB can result in growth inhibition [67], so a better understanding in the regeneration and repair of Fe-S clusters might be worthwhile [68 ]. To develop a biotin bioprocess, major metabolic (SAM supply) and protein (BioB) engineering efforts are needed.

Vitamin B 9
Folates play important roles in the metabolism of amino acids and nucleotides. Folic acid, folinic acid and 5-methyltetrahydrofolate (THF) are the only vitamers. There are no objections in engineering a folic acid overproducer because the biological precursors GTP, phosphoenolpyruvate, D-erythrose-4-phosphate and glutamate are readily available in many microorganisms. One of the initial biosynthetic steps is catalysed by GTP cyclohydrolase I, which competes with GTP cyclohydrolase II (initial step in riboflavin biosynthesis) for GTP. In B. subtilis, the folate-related transcription units carry 4-6 biosynthetic genes (Figure 1g), whereas in E. coli seven fol genes are spread across the chromosome [69]. Microbial production of folates has been reported in the fortification of dairy products using engineered lactic acid bacteria or in yeasts strains, but A. gossypii appears to be the most promising host [70 ] (Table 2). However, fermentation of folic acid appears to not be economic because its chemical synthesis is very cheap and its demand low, so it will be challenging to establish a bioprocess unless environmental regulations force .

Vitamin B 12
Cobalamins are composed of adenosyl-(AdoB 12 ), methyl-(CH 3 B 12 ), hydroxy-(OHB 12 ) and cyanocobalamin (CNB 12 ). AdoB 12 /CNB 12 are the active forms, while OHB 12 can be an intermediate (or co-factor of dehalogenases) and CNB 12 is the most stable and commercial form due to its longer shelf life. Cobalamin chemical synthesis involves $70 reactions (Table 1), whereas microbes can produce it using fewer steps by a de novo or salvage pathway. The former is composed of $30 genes starting from glutamate and can be aerobic or anaerobic (Figure 1h), whereas the latter pathway has $12 genes (e.g., E. coli) but it needs dedicated ATP-driven ABC transporters for importing cobinamide, a cobalamin analog [71 ].
High titers of $200 mg/L were reported for Propionibacterium shermanii, but Pseudomonas denitrificans, which produces similar titers is the current industrial host (Table 2) [72 ]. The engineering of the de novo Bacillus megaterium pathway [73] or the transfer of 22 upstream genes from P. dentrificans into E. coli enabled de novo B 12 production [74], while a riboswitch-based screening system was developed to screen for high producers in S. meliloti [75 ] but the yields are very low (Table 2).
Since both pathways are highly regulated at different levels [71 ,72 ,73,76 ], the main challenges are to remove the inhibition of mRNAs of cysG and the cbi operon by the cobalamin riboswitch as well as enzymes including HemA by protoheme, HemE by coproporphyrinogen III, CysG by cobalamin and CobA by its own substrate [71 ,77]. Other issues are the need to supply sufficient cobalt to the broth and generate enough SAM co-factor for methylation reactions (e.g., CysG).

Technological, environmental and regulatory issues
To develop efficient bioprocesses, genetic tools must be available [5 ] and the host must be suitable for fermentation [81 ]. The current B 12 bioprocess is suboptimal: it involves long fermentation cycles, expensive media and lack of genetic tools in commercial hosts [71 ]. Additionally, down-stream processing (DSP) is an important step. B vitamins have different water-solubility (Table 1). Riboflavin's low solubility enabled an efficient DSP: a large fraction of the product crystallizes during fermentation; the crystals are washed and separated [3 ]. Similar DSPs could be developed for vitamins B 7 and B 9 . Cobalamin fermentation produces a mixture of AdoB 12 , MeB 12 and OHB 12 . During DSP, the vitamers are converted to CNB 12 with potassium cyanide at high pressure/temperature [72 ]. The biocatalytic conversion of cyanopyridine to NAM is practically quantitative, thus simplifying its purification [7]. DSPs for vitamins B 1 , B 3 and B 6 are unknown. The DSP of vitamin B 5 is well-established [82]. Finally, there are also environmental and regulatory issues (Box 1).

Conclusion
Relatively small cellular requirements for vitamins mean that their biosynthesis has not evolved to be high yielding in nature. Yet, industrial bioprocesses for vitamin B 2 and B 12 have successfully been established through rational and classical metabolic engineering, and the question is whether the six other B vitamins will follow. While low current manufacturing cost combined with highly complex biosynthetic pathways and sloppy enzymes makes industrially relevant cell factories extremely challenging, several factors points towards a future transition towards bio-based processes: (i) our fundamental understanding of vitamin biosynthesis increases; (ii) as the synthetic biology toolbox expands, our ability to engineer microbes is rapidly advancing; (iii) environmental restrictions are already today increasing cost of polluting chemical processes; and (iv) in certain segments, for example, food supplements and cosmetics, we observe a consumer demand for naturally produced biochemicals.

Conflict of interest statement
LSG, MM and HJG have filed patent applications that are related to the fermentation of some B vitamins.

Box 1. Sustainability and regulatory issues
Aside from cost, sustainability is a motivating factor for switching to bioprocesses. However, predicting environmental impact is challenging. In the case of riboflavin, a life-cycle assessment (LCA) determined that the fermented process scored better than the chemical one in 4 out of 5 indicators (as measured by equivalents of) [47 ]: Reduction of potential of global warming by 30% (CO 2 ), acidification by 50% (SO 4 ), ozone creation by 60% (NO 3 ) and cumulated energy consumption by 40%; the eutrophication potential, however, increased by 40% (PO 4 ). The LCA of vitamin B 2 can be a guide, but the exact specification of the chemical and fermentative processes are important to calculate their actual impact. Similar to vitamin B 2 , the conventional chemical routes to B 1 , B 5 , B 6 and B 7 rely on non-renewable sources, produce toxic waste and have high-energy demands. The synthesis of vitamin B 3 has switched to a more sustainable process. While vitamin B 12 is currently produced by fermentation, the conversion of the product mixture to CNB 12 using potassium cyanide is not sustainable. Performing this step biologically could result in a greener bioprocess.
Additives of food and feed produced with natural hosts or genetically modified microorganisms (GMM) in closed systems are not evaluated under the same regulations as genetically modified food and feed in the European Union (EU) [83]. In most cases, the biologically produced vitamin can be identical to the chemically produced vitamin. However, the production method will affect the presence of impurities. For example, aniline is a non-desired impurity in the chemically produced riboflavin that is absent in the fermented version [15]. Conversely, it is unlikely to find DNA in chemically produced additives, which could be an issue in fermented products. The European Food Safety Authority (EFSA) states that it is mandatory that neither GMM nor its recombinant DNA can be present in the final product that is placed on the EU market as non-GMM food or feed additive [83,84]. Further, several regulatory steps need to be taken to replace a chemical production process of a vitamin with a biological process, and these steps differ depending on the vitamin application, the host organism, the purity level obtained, the stability of the product, and the specific impurities present in the product. One step involves showing that the new product is safe -generally done by toxicity studies and/or by proving that it is near-identical to the previously approved product, for example, by demonstrating that it conforms to the specifications of the relevant pharmacopoeia monographs (e.g. European pharmacopoeia in the EU). In the US, it is standard practice to notify the FDA that an ingredient is Generally Recognized As Safe (GRAS), for example, riboflavin biosynthesized by Eremothecium ashbyii has been designated as GRAS. 41