Energy Conservation in Fermentations of Anaerobic Bacteria

Anaerobic bacteria ferment carbohydrates and amino acids to obtain energy for growth. Due to the absence of oxygen and other inorganic electron acceptors, the substrate of a fermentation has to serve as electron donor as well as acceptor, which results in low free energies as compared to that of aerobic oxidations. Until about 10 years ago, anaerobes were thought to exclusively use substrate level phosphorylation (SLP), by which only part of the available energy could be conserved. Therefore, anaerobes were regarded as unproductive and inefficient energy conservers. The discovery of electrochemical Na+ gradients generated by biotin-dependent decarboxylations or by reduction of NAD+ with ferredoxin changed this view. Reduced ferredoxin is provided by oxidative decarboxylation of 2-oxoacids and the recently discovered flavin based electron bifurcation (FBEB). In this review, the two different fermentation pathways of glutamate to ammonia, CO2, acetate, butyrate and H2 via 3-methylaspartate or via 2-hydroxyglutarate by members of the Firmicutes are discussed as prototypical examples in which all processes characteristic for fermentations occur. Though the fermentations proceed on two entirely different pathways, the maximum theoretical amount of ATP is conserved in each pathway. The occurrence of the 3-methylaspartate pathway in clostridia from soil and the 2-hydroxyglutarate pathway in the human microbiome of the large intestine is traced back to the oxygen-sensitivity of the radical enzymes. The coenzyme B12-dependent glutamate mutase in the 3-methylaspartate pathway tolerates oxygen, whereas 2-hydroxyglutaryl-CoA dehydratase is extremely oxygen-sensitive and can only survive in the gut, where the combustion of butyrate produced by the microbiome consumes the oxygen and provides a strict anaerobic environment. Examples of coenzyme B12-dependent eliminases are given, which in the gut are replaced by simpler extremely oxygen sensitive glycyl radical enzymes.


INTRODUCTION
Fermentation is a chemical reaction catalyzed by living cells with two different meanings. In biotechnology the bacterial production of a certain compound mainly from glucose under aerobic conditions is called fermentation, e.g., L-lysine fermentation. In this review, however, fermentation is defined as an anaerobic bacterial redox process of an organic substrate leading to different products, e.g., fermentation of glutamate to ammonia, CO 2 , acetate, butyrate and hydrogen. Thereby glutamate is oxidized to CO 2 and acetate; the resulting reducing equivalents are used to synthesize butyrate and H 2 to establish a balanced chemical equation. Inorganic electron acceptors such as nitrate, sulfate or Fe(III) are not involved; otherwise the process would be called respiration rather than fermentation. Thermodynamically, a fermentation is an exergonic process with a free energy of G • <−20 kJ/mol. G • is defined for 1 M substrates and products, 10 5 Pa for gasses and at a temperature of 298 K. Under biological conditions, in dilute aqueous solutions (55.5 M H 2 O) and at pH 7.0, G • is used. In this review, the G • values have been calculated from the data given by Thauer et al. (1977).
A well-established fermentation is the conversion of glucose by lactic acid bacteria to lactate at pH 7.0 with G • = −185 kJ/mol (eq. 1). In contrast, the aerobic oxidation of glucose in mammalian cells with O 2 to CO 2 and H 2 O yields G • = −2,872 kJ/mol (eq. 2).
C 6 H 12 O 6 → 2 CH 3 −CHOH−COO − + 2 H + ; Glucose oxidation via glycolysis, pyruvate dehydrogenase, Krebs' (citric acid) cycle and mitochondrial respiratory chain are very well analyzed processes leading to about 38 mol ATP/mol glucose; (−2,872 kJ/mol glucose): (38 mol ATP/mol glucose) = −76 kJ/mol ATP. Hence in a catabolic reaction with one or more irreversible steps, a free energy change of −76 kJ is required to generate 1 mol ATP, whereas under equilibrium conditions only about −50 kJ are necessary (Thauer et al., 1977;Buckel and Thauer, 2013). If this value is applied to the fermentation of glucose to 2 lactate, the ATP yield should be (−185 kJ/mol): (−76 kJ/mol ATP) = 2.4 ATP, which is 20% higher than 2.0 ATP via substrate level phosphorylation in glycolysis. This examples tells why energy conservation in fermentative anaerobic bacteria has been considered as inefficient. The discovery that anaerobic bacteria are able use in addition electrochemical ion gradients for ATP synthesis, catalyzed by ubiquitous H + or Na + dependent F 1 F 0 -ATPsynthases (von Ballmoos et al., 2009), changed this view. Work by Konings and colleagues showed that in the glucose fermenting Streptococcus cremoris, the export of a formed lactic acid molecule is accompanied by a proton, which establishes an electrochemical proton gradient (Otto et al., 1980). Since 2 lactic acids are produced from glucose, two additional protons are exported which can give rise to 0.5 ATP catalyzed by ATP synthase. Thus the involvement of electrochemical ion gradients allows bacteria to use energy increments that are smaller than that of 1 ATP, which makes anaerobes even more efficient than aerobes. Therefore, the amount of energy to synthesize 1 ATP in anaerobes is now considered as about −66 kJ rather than −76 kJ in aerobes; see section "Fermentation of Glutamate: 2-Hydroxyglutarate or 3-Methylaspartate Pathway?" In this review, I will show that also anaerobes have enzymes which are able to generate a Na + /H + motive force, the biotin-containing decarboxylases (see section "Biotin-Containing Decarboxylases as Sodium Ion Pumps") and the NAD:ferredoxin oxidoreductase (Rnf, see section "The Na+/H+ Pump Ferredoxin:NAD Oxidoreductase Also Called Rnf "). Sections "Flavin Based Electron Bifurcation" and "On the Mechanism of Electron Bifurcation Catalyzed by EtfAB-Bcd" will give an introduction to flavin based electron bifurcation, by which "energy rich" reduced ferredoxin and flavodoxin are obtained. In section "Fermentation of Glutamate: 2-Hydroxyglutarate or 3-Methylaspartate Pathway?" the 2-hydroxyglutarate and the 3-methylaspartate pathway of the fermentation of glutamate are introduced, which apply the biotin-containing decarboxylases and Rnf together with substrate level phosphorylation (SLP) to obtain the maximum theoretical yield of ATP. Section "Oxygen-Tolerant and -Intolerant Radical Enzymes" tackles the question why nature evolved two different pathways of glutamate fermentation. Finally, section "Butyrate Provides Anaerobiosis in the Gut" shows how butyrate converts the human gut to an extremely oxygen-free environment and an ideal environment for strict anaerobic bacteria.

BIOTIN-CONTAINING DECARBOXYLASES AS SODIUM ION PUMPS
Concomitant with the discovery of Konings, bacterial Na + pumps were detected and isolated, driven by decarboxylation (Buckel, 2001a). The first decarboxylase of this type was detected by Dimroth (1980) in Klebsiella pneumoniae and characterized in the following years as biotin containing integral membrane enzyme catalyzing the decarboxylation of oxaloacetate to pyruvate, G • ≈ −30 kJ/mol coupled to the translocation of 2 Na + . This breakthrough was followed by methylmalonyl-CoA decarboxylase from Veillonella aerogenes (Hilpert and Dimroth, 1982) and glutaconyl-CoA decarboxylases from Acidaminococcus fermentans, Peptostreptococcus asaccharolyticus (formerly called Peptococcus aerogenes; Ezaki et al., 1983) and Clostridium symbiosum Semmler, 1982, 1983). These enzymes share a common two step mechanism: the carboxyl group of the substrate is transferred to enzyme bound biotin. Addition of a proton to the carbonyl group of the formed N-carboxybiotin causes decarboxylation, which drives the translocation of Na + ions from the cytoplasm through the membrane to the outside of the bacterium. These biotin containing decarboxylases are composed of 3-5 subunits. The largest and hydrophilic αsubunit catalyzes the carboxy transfer from substrate to biotin and the extremely hydrophobic β-subunit is responsible for the decarboxylation of carboxybiotin and Na + transport. Biotin is covalently attached to a conserved lysine residue of the γ-subunit and the small δ-subunit connects the α-subunit in the cytoplasm with the β-subunit in the membrane (Vitt et al., 2020). Variations are observed with the γ-subunit. In oxaloacetate decarboxylase from K. aerogenes, a small domain of the α-subunit serves as γ-subunit (Schwarz et al., 1988;Xu et al., 2020), whereas glutaconyl-CoA decarboxylase from C. symbiosum contains two slightly different γ-subunits (Kress et al., 2009). The amount of Na + ions transported by these pumps could be determined for oxaloacetate decarboxylase from K. pneumoniae as up to 2 Na + per decarboxylation (Dimroth and Thomer, 1993). This is supported by the recent cryo-EM structure (4.5 Å resolution) of the trimeric βγsubunit complex of oxaloacetate decarboxylase from Salmonella typhimurium, where two Na + binding sites in one β-subunit were detected. Unfortunately, the coupling between decarboxylation of carboxybiotin and Na + -transport could not be elucidated (Xu et al., 2020).
Oxaloacetate decarboxylase from K. pneumoniae is involved in the fermentation of citrate to 2 acetates, 1.2 CO 2 and 0.5 formate. The pathway involves citrate cleavage to acetate and oxaloacetate, decarboxylation of the latter to CO 2 , pyruvate and µNa + , cleavage of pyruvate with CoA to acetyl-CoA and formate, substrate level phosphorylation (SLP) with acetyl-CoA to acetate, CoA and ATP. Generation of µNa + is used for citrate transport and for driving the endergonic reduction of NAD + for anabolic purposes by formate via ubiquinol (Pfenninger- Li and Dimroth, 1992). A balanced stoichiometry of citrate fermentation would give G • = −78.7 kJ/mol, which is the amount of energy required for the synthesis of about 1.2 ATP (eq. 3).
µNa + directly drives the Na + -dependent F 1 F o -ATP synthase (Laubinger and Dimroth, 1988).  (Schmehl et al., 1993). Similar genes were detected in the genome of C. tetani, a close relative of C. tetanomorphum, whose membranes catalyzed the oxidation of NADH with ferricyanide [hexacyanoferrate(III)] (Brüggemann et al., 2003;Imkamp et al., 2007). The enzyme complex in membranes from C. tetanomorphum, which contained the highest activity, could be solubilized with dodecylmaltoside and purified (Boiangiu et al., 2005;Jayamani, 2008). SDS-PAGE indicated that the complex was composed of six different subunits RnfCDGEAB (Figure 1), RnfCDGE are related to four subunits of the sodium pumping NADH:quinone oxidoreductase (Nqr) from Vibrio cholerae (Steuber et al., 2014a,b). The yellow-brownish native enzyme complex contained covalently bound flavin, non-covalently bound FMN, riboflavin and 23 ± 1 mol Fe per mol protein, most likely in six iron-sulfur clusters (Boiangiu et al., 2005). The oxidation of NADH with ferricyanide catalyzed by Rnf in membrane vesicles from C. tetanomorphum showed no Na +dependence; probably the electrons did not pass through all six subunits due to a short cut. Therefore, the reaction was performed in reverse. The reduction of NAD + was measured in an assay, in which 1 mM Ti(III)citrate was used to keep ferredoxin in the reduced state. Controls showed that Ti(III)citrate reduced NAD + without ferredoxin and membranes at pH ≥ 7.0. But at pH 6.8 again no Na + -dependence could be observed (Imkamp et al., 2007;Jayamani, 2008). The first Na + -transport with an Rnf was demonstrated with membrane vesicles from Acetobacterium woodii, NAD + , Ti(III)citrate and 22 Na + by Biegel and Müller (2010). The electrogenic transport was stimulated by valinomycin in the presence of K + and inhibited by 100 µM Na + -ionophore N,N,N' ,N'-tetracyclohexyl-1,2-phenylendioxydiacetamide (ETH 2120). Later Ti(III)citrate was replaced by purified carbon monoxide dehydrogenase/acetyl-CoA synthase and CO as reductant for ferredoxin, which resulted in much more reliable dependencies on Na + (Hess et al., 2013). In a similar manner reduced ferredoxin and flavodoxin quinol were generated with the electron bifurcating electron transfer flavoprotein (EtfAB) and butyryl-CoA dehydrogenase (Bcd) both from A. fermentans (see also next chapter, eq. 5). With membranes from A. fermentans it was shown that Rnf activity was Na + -dependent with apparent K m = 120 ± 20 µM Na + and K m = 280 ± 50 µM Li + at pH 6.8; in addition, K m = 2.0 ± 0.4 µM flavodoxin and 1.4 ± 0.1 µM ferredoxin were measured (Chowdhury et al., 2016). The values for Rnf from A. woodii were: K m = 201 ± 30 µM Na + at pH 6 and K m = 155 ± 39 µM Na + at pH 7.7 (Hess et al., 2013). Surprisingly, the K m values for Na + and Li + are much lower than those of the biotin-containing Na + pumps glutaconyl-CoA decarboxylase from A. fermentans (Buckel and Semmler, 1982) or oxaloacetate decarboxylase from K. pneumoniae (Dimroth and Thomer, 1986), which exhibit apparent K m values of 1.0-1.5 mM Na + and 25-100 mM Li + . Recently a survey showed that not all anaerobic bacteria contain Rnfs and not all Rnfs are Na +dependent. Thus the Rnf activity is absent in R. capsulatus and Escherichia coli, but present also in Bacteroides fragilis, Clostridium ljungdahlii, V. cholerae, and Clostridium kluyveri. Besides C. tetanomorphum, A. fermentans, and A. woodii, a Na + -dependence could be only detected in Rnf from B. fragilis (Hess et al., 2016).

FLAVIN BASED ELECTRON BIFURCATION
Ferredoxin-dependent electron bifurcation represents an important additional process to generate the energy rich molecule. Only a few enzymes, such as pyruvate ferredoxin: oxidoreductases (PFOR), reduce ferredoxin, the substrate of Rnf. A search for a reductant other than PFOR, took the butyryl-CoA dehydrogenases (Bcd) into consideration, because the difference between the reduction potentials of NAD/NADH, E 0 ' = −320 mV, and crotonyl-CoA/butyryl-CoA, E 0 ' = −10 mV, amounts to +310 mV, equal to G • = −n × F × E o ' = −2 × 96.5 × 0.310 Volt = −59.8 kJ mol −1 . Probably Bcd also reduces ferredoxin or pumps Na + driven by this large G • '. The hypothesis of a membrane bound butyryl-CoA dehydrogenase (Bcd) could be refuted by the already known isolation of a soluble green homotetrameric Bcd from A. fermentans which could be separated from the yellow heterodimeric electron transferring flavoprotein (EtfAB) (Buckel, 1990). A very similar green soluble Bcd has been known from Megasphaera elsdenii (formerly called Peptostreptococcus elsdenii) (Engel and Massey, 1971), which like A. fermentans belongs to the Negativicutes, a class with a Gram-negative cell wall of the, predominantly, Gram-positive phylum Firmicutes (Latin: cutis firma = strong outer skin) (Marchandin et al., 2010). In contrast, butyryl-CoA dehydrogenase from the Gram-positive C. tetanomorphum forms a tight complex with the heterodimeric EtfAB (EtfAB-Bcd). But immunogold labeling and electron microscopy clearly showed that this EtfAB-Bcd complex was also not associated with the membrane (Herrmann, 2008). Nevertheless, a hypothesis arose from these negative results that the reduction of crotonyl-CoA with NADH catalyzed by Bcd from C. tetanomorphum was somehow involved in the reduction of ferredoxin (E o ' = −420 mV). Furthermore, since 1969 it was known that cell extracts of Clostridium kluyveri catalyzed an acetyl-CoA dependent reduction of ferredoxin with NADH (Thauer et al., 1969). Since the cell extract contained all enzymes for the synthesis of butyrate ) as did C. tetanomorphum cells, it was ferredoxin that indeed should be reduced by NADH. Similar to complex III of the mitochondrial respiratory chain, it was postulated that NADH reduced EtfAB, which bifurcated one electron exergonically to δ-FAD of Bcd and the other endergonically to ferredoxin (eq. 5) . To reduce crotonyl-CoA with two electrons, repetition of the reaction is required: (5) This hypothesis was verified first with the purified Etf-Bcd complex from C. kluyveri (Li et al., 2008). The catalytic amounts of purified ferredoxin from Clostridium pasteurianum were regenerated with purified [FeFe]hydrogenase also from C. pasteurianum. If one of the components of this assay (NADH, crotonyl-CoA, Fd, hydrogenase or EtfAB-Bcd) was omitted, no reaction could be observed. Hence the electron bifurcation system is tightly coupled. The purified Etf-Bcd complex from C. tetanomorphum catalyzed exactly the same reaction as in eq. 5, as did the combination of Bcd and EtfAB from A. fermentans or M. elsdenii (Chowdhury et al., 2014(Chowdhury et al., , 2015.

ON THE MECHANISM OF ELECTRON BIFURCATION CATALYZED BY EtfAB-Bcd
To study the mechanism of electron bifurcation, EtfAB-Bcd was chosen, because it contains no iron-sulfur cluster and in the resting state, it is state stable under air; for other bifurcating systems see: Buckel and Thauer (2018a,b) and Kayastha et al. (2021). EtfAB-Bcd harbors three differently bound FAD molecules, αand β-FAD in EtfAB, and δ-FAD, one in each subunit of the tetrameric Bcd ( Figure 1). As source for EtfABs, A. fermentans (Chowdhury et al., 2014) and M. elsdenii (Chowdhury et al., 2015;Vigil et al., 2021) are the most suitable organisms, because their EtfABs form no tight complexes with their Bcds and can be studied separately. The crystal structure EtfAB from A. fermentans (Af -EtfAB) revealed its composition of 3 domains, similar to those of Etfs involved in fatty acid and anaerobic toluene oxidations (Figure 2, Bifurcation-like state) (Chowdhury et al., 2014;Vogt et al., 2019). Domains I and II form subunit A and domain III subunit B. α-FAD is located in domain II (subunit A) with its isoalloxazine ring at the interface to domain III (subunit B). β-FAD sits on domain III with its isoalloxazine ring at the interface between domains III and I. β-FAD is only present in bifurcating Etfs, whereas in non-bifurcating Etfs, e.g., those acting as electron acceptors of acyl-CoA dehydrogenases, AMP alone occupies the place of the AMP part of β-FAD; the place for the isoalloxazine part is empty (Roberts et al., 1996). In Af -EtfAB, the distance between αand β-FAD is 18 Å; the possible rotation of α-FAD on the flexible domain II toward β-FAD by about 10 • shortens the distance between the two FADs to 14 Å, the longest distance of fast electron tunneling (Moser et al., 1992). NADH is located close to β-FAD. Since the already known structure of Bcd from M. elsdenii did not tell anything about the interaction with Etf (Djordjevic et al., 1995), the structure of the tight EtfAB-Bcd complex from Clostridium difficile was solved (Figure 2, Dehydrogenase state) (Demmer et al., 2017). The structure revealed the homotetrameric Bcd with one EtfAB at each subunit, (EtfAB) 4 (Bcd) 4 . Furthermore, α-FAD on the flexible domain II has rotated by about 80 • in the opposite direction toward δ-FAD of Bcd, until 8 Å distance. Thus the two structures represent two states of the bifurcating system. The structure of EtfAB alone is close to the bifurcating state with a distance between α-FAD and β-FAD of 18 Å. In complex with Bcd, α-FAD swings into the dehydratase D-state, which brings it 34 Å apart from β-FAD (Figure 2). FIGURE 1 | An example of energy conservation in butyrate producing anaerobic bacteria via the generation of an ion motif force. The bifurcating EtfAB-Bcd complex reduces 2 Fd with NADH driven by the exergonic reduction of crotonyl-CoA to butyryl-CoA with a second NADH. The Rnf complex in the membrane produces 2 µNa + from the exergonic reduction of NAD + by 2 Fd − . The ATPase generates 1 ATP from 4 µNa + .
FIGURE 2 | Partial crystal structures of the recombinant Etf-Bcd complex from C. difficile produced in E. coli; EtfAB and 2 subunits of the tetrameric Bcd are displayed. The dehydratase state shows the structure as solved. In the bifurcation state domain II of Etf rotated CW by 80 • as found in Etf from A. fermentans (see red arrow); Etf, domains I + II (subunit A) in green, domain III (subunit B) in light brown, Bcd1 in yellow and Bcd2 in pinkish-brown. In the dehydratase state, α-FAD and δ-FAD are located close together (8 Å distance), ready for ET. In the bifurcation-like state, α-FAD and β-FAD are 18 Å apart; further rotation by 10 • CW would bring them closer together by 4 Å, ready for electron bifurcation (taken from Buckel and Thauer, 2018a).
The one-electron reduction potentials of β-FAD could not be measured by the applied method, because the half-life of the semiquinone is expected to be extremely short, T1 /2 ca. 10 ps (Lubner et al., 2017). In analogy to the long-known Q-cycle in the mitochondrial complex III, a reactive β-FAD •− with an extremely low stability constant (log K s <<0) is necessary for FIGURE 3 | Scheme of electron bifurcation in the EtfAB-Bcd complex. The different FADs are placed according to their approximate reduction potential. NADH reduces β-FAD to β-FADH − , which bifurcates. One electron goes to α-FAD − and the formed semiquinone, β-FAD •− , reduces ferredoxin (Fd) to Fd − . The formed hydroquinone (α-FADH − ) swings over to Bcd and transfers one electron to δ-FAD. Note the decrase of the reduction potential of α-FAD •− due to the change in location from Etf to Bcd. Repetition of this process yields δ-FADH − , which reduces crotonyl-CoA to butyryl-CoA. The FAD semiquinones are red, the quinones yellow and the hydroquinones colorless (Sucharitakul et al., 2021a,b). electron bifurcation (eq. 6) (Bergdoll et al., 2016). Therefore, the β-FAD/β-FAD •− reduction potential must be much more negative than that of β-FAD •− /β-FADH − , whereas in flavoproteins with stable semiquinones the opposite is the case (log K s > 0), e.g., flavodoxin from A. fermentans with reduction potentials for FAD/FAD − = −60 mV are much higher than that for FAD − /FADH − = −420 mV (Hans et al., 2002).
To convert K s to the one-electron reduction potentials of the bifurcating cofactor with a reduction potential of E 0 ' , the relationship of eq. 7 is used. A graphical plot of E' versus log K s gives two straight lines which cross at E' = E o ' and log K s = 0. At log K s > 0, the potentials are normal and at log K s < 0 the potentials are inverse or called "crossed" (Nitschke and Russell, 2012).
Probably the reader wonders why during electron bifurcation, α-FAD •− rather than α-FAD acts as high potential acceptor. There are three reasons, firstly the reduction potential of α-FAD with +134 mV is too high, whereas that of α-FAD •− is more in the right range of −36 mV. In electron bifurcation, the potential difference between β-FADH − and α-FAD •− with + 179−(−36) = +215 mV should be similar to the negative difference between the potentials of β-FADH − and Fd with − 719−(−390) = −329 mV, to ensure tight coupling. With α-FAD as acceptor, the difference of + 179−(+134) = +45 mV would be much too low. Secondly, the change of the one-electron reduction potential of α-FADH − to the two-electron reduction potential cannot occur at the state of α-FAD •− . Thirdly, as soon as EtfAB encounters NADH, either in vitro or in vivo, α-FAD is reduced to the semiquinone (Sucharitakul et al., 2021a). Finally, stopped flow measurements by Jeerus Sucharitakul directly showed that Etf with α-FADH − indeed transferred one electron to Bcd, whereas with α-FAD •− no electron transfer was observed (Sucharitakul et al., 2021b, unpublished).

FERMENTATION OF GLUTAMATE: 2-HYDROXYGLUTARATE OR 3-METHYLASPARTATE PATHWAY?
The previous chapters described enzymatic systems which are able to establish electrochemical Na + or H + gradients for energy conservation in anaerobic bacteria. In addition, electron bifurcation as a major source of reduced ferredoxin has been presented. This chapter shows how the two pathways of glutamate fermentation apply these systems to obtain the theoretical yield of ATP/glutamate (eq. 9). −314 kJ/5 glutamate: −66 kJ/ATP = 4.8 ATP/5 glutamate; 0.96 ATP/glutamate. (Buckel, 2001b).
The fermentations of glutamate via methylaspartate or 2-hydroxyglutarate lead to identical products and to identical amounts of conserved ATP. Therefore, the question arises why bacteria exclusively use one of these pathways. One would expect that bacteria prefer the 2-hydroxyglutarate pathway because the 3-methylaspartate pathway requires about 30 additional enzymes for the anaerobic biosynthesis of the coenzyme B 12 (Moore and Warren, 2012), whereas in the 2-hydroxyglutarate pathway only the ubiquitous cofactors NAD, FAD, CoA, pyridoxal-5phosphate and biotin are involved. These cofactors are not pathway specific, because they are also necessary for anabolism. In absence of vitamin B 12 , Fusobacterium varium uses the 2-hydroxyglutarate pathway, whereas after addition of 1 µM vitamin B 12 or just 1 µM CoCl 2 to the medium, the organism ferments glutamate via 3-methylaspartate (Ramezani et al., 2011). Hence, F. varium is able to synthesize coenzyme B 12 when Co 2+ is present and the argument that the expression of 30 additional genes for B 12 synthesis favors the 2-hydroxyglutarate pathway does not apply. Therefore, the reason for the existence of this pathway must be found somewhere else.
The organisms using the 2-hydroxyglutarate pathway, A. fermentans, P. asaccharolyticus, C. symbiosum, F. nucleatum, and F. varium were isolated from anaerobic niches of the human host, preferentially from the human large intestine. In contrast, the members of the 3-methylaspartate pathway C. tetanomorphum, C. tetani, C. cochlearium, C. lentoputrescens, C. limosum, C. malenominatum have been detected outside the gut on places like soil, where organic matter decomposes anaerobically (Buckel, 2001b). The employment of the 2-hydroxyglutarate pathway in the human gut could be due to the much lower oxygen concentration as compared to soil which encounters often exposures to air through plant roots, earthworms, moles, etc. Probably, the oxygen sensitivity of enzymes of an important pathway might be responsible for the ecology of the organism.
Both glutamate fermenting pathways contain oxygen sensitive radical enzymes. Whereas coenzyme B 12 -dependent glutamate mutase in the 3-methylaspartate pathway is only moderately oxygen sensitive, 2-hydroxyglutaryl-CoA dehydratase in the 2-hydroxyglutarate pathway immediately becomes inactive after exposure to air. Coenzyme B 12 -dependent mutases catalyze the reversible radical rearrangement of a methine radical to a methylene radical. The process is initiated by homolysis of the carbon-cobalt bond of coenzyme B 12 . The formed 5 -deoxyadenosyl radical abstracts the Si-hydrogen atom from the methylene group at C3 of (S)-glutamate and after rearrangement adds it back to the formed methylene radical yielding the methyl group of (2S,3S)-3-methylaspartate. FIGURE 6 | Butyrate nourishes the mucosa cells and makes the colon anaerobic. The major part of the food is digested and absorbed in the small intestine. The remaining fibers are hydrolyzed to monosaccharides and amino acids by the anaerobic bacteria and fermented the large intestine or colon to acetate, propionate and butyrate. The mucosa cells combust the produced butyrate with oxygen from blood and colon, whereby the colon becomes strictly anaerobic.
After each turnover, the radical disappears by reformation of the carbon-cobalt bond of the coenzyme, for a review see Buckel and Golding (1996). Thus, in the presence of air, the radicals are only transiently exposed to oxygen as the coenzyme B 12 -dependent methylmalonyl-CoA mutase in human mitochondria. In contrast, 2-hydroxyglutaryl-CoA dehydratase acts with a completely different radical mechanism (Buckel and Keese, 1995), for a review see Buckel (2019). The enzyme system is composed of two proteins. The homodimeric activator contains one ADP in each subunit and one [4Fe-4S] cluster coordinated by four cysteines, two from each subunit, similar to the iron protein (NifH) of nitrogenase (Locher et al., 2001). The heterodimeric dehydratase holds one [4Fe-4S] cluster in each subunit. Each cluster is coordinated by three cysteines; the fourth coordination is occupied in subunit A by a sulfur atom and in subunit B by water, which can be replaced by the thioester carbonyl of the substrate. The dehydration is initiated by ATP/ADP exchange of the activator and reduction of its [4Fe-4S] 2+ cluster by reduced ferredoxin (Fd − ). Driven by ATP hydrolysis, the extra electron is transferred to or "shot into" the [4Fe-4S] cluster of subunit A and further shifted to cluster B, where 2-hydroxyglutaryl-CoA is bound. The electron in cluster B reduces the thioester carbonyl to a ketyl radical, which due to its lower basicity is replaced from the iron by the hydroxyl group at C-2, a process called ligand swapping. Aided by the iron of cluster B, the nucleophilic ketyl eliminates the hydroxyl group to form an enoxy radical. The hydroxyl group at the iron of the cluster acts as base to remove the now acidic β-proton at C-3. The resulting allylic ketyl replaces the formed water by a second ligand swapping and returns the electron via cluster B to cluster A. The formed product glutaconyl-CoA is released and cluster B is able to accept the next substrate together with the electron form cluster A. The dehydratase is able to catalyze at least 1000 turnovers before another ATP molecule has to be hydrolyzed to continue catalysis . Proof of this mechanism was obtained by detection of the corresponding allylic ketyl radical of (R)-2-hydroxyisocaproyl-CoA dehydratase from C. difficile (Kim et al., 2008).
It has been shown experimentally that (R)-2-hydroxyacyl-CoA dehydratases are indeed much more oxygen sensitive than coenzyme B 12 -dependent carbon skeleton mutases. Whereas 2-hydroxyglutarate dehydratases had to be purified and assayed under strict exclusion of oxygen , glutamate mutase could be purified under air. During catalysis, however, after about 1 min a significant effect of air on the rate of 3-methylaspartate formation from glutamate was observed (Leutbecher et al., 1992). In the dehydratases, the electron never disappears during the catalytic cycle and the ironsulfur cluster of the activator is solvent accessible, which makes it extremely oxygen sensitive. The question arises why the gut, where (R)-2-hydroxyacyl-CoA dehydratases are active, contains such low oxygen concentrations despite its ample supply with oxygen-rich blood (see section "Butyrate Provides Anaerobiosis in the Gut").

OXYGEN-TOLERANT AND -INTOLERANT RADICAL ENZYMES
In the previous chapters it has been shown that the 3-methylaspartate pathway tolerates oxygen because after each turnover of glutamate mutase, the cobalt-carbon bond between the 5 -deoxyadenosyl radical and cob(II)alamin is reformed. Thus, only during catalysis the radical is exposed to oxygen. In contrast, in (R)-2-hydroxyglutaryl-CoA dehydratase of the alternative glutamate fermenting pathway, the radical is always present and exposed to the medium. In addition, the iron-sulfur clusters of the dehydratase and its activator are very oxygen-sensitive. Most likely, this is the reason, why organisms of the 2-hydroxyglutarate pathway are only found in the gut or in strictly anaerobic marine sediments. Furthermore, B 12 -dependent carbon skeleton mutases and eliminases as well as thiamin diphosphate dependent enzymes are substituted in the gut by glycyl radical enzymes ( Table 2). Similar to (R)-2-hydroxyacyl-CoA dehydratases, glycyl radical enzymes also require specific activating enzymes. These activases belong to the large family of radical SAM enzymes, which catalyze the one electron-reduction of S-adenosylmethionine (SAM) to methionine and the 5 -deoxyadenosyl radical. This radical irreversibly abstracts one hydrogen from a conserved glycine residue of the enzyme to give a stable protein-bound radical. Upon binding of substrate, the glycyl radical abstracts the sulfhydryl hydrogen form a nearby cysteine residue, which in turn removes the hydrogen atom from the substrate. After the rearrangement, the radical returns to the glycine residue and remains stable until the next turnover, unless it is attacked by oxygen (Knappe and Sawers, 1990;Shisler and Broderick, 2014). Many organisms including humans are able to degrade propionate via carboxylation of propionyl-CoA to (S)-methylmalonyl-CoA and racemization to (R)methylmalonyl-CoA, which is rearranged to succinyl-CoA mediated by coenzyme B 12 . The Krebs cycle converts succinyl-CoA to oxaloacetate, which enters gluconeogenesis or is degraded via pyruvate to acetyl-CoA. Propionibacteria use the reverse pathway to produce propionate from succinate. Veillonella alcalescens and P. modestum conserve energy with the biotin-dependent Na + -pump (S)-methylmalonyl-CoA decarboxylase (Hilpert and Dimroth, 1982;Schink, 1982). In contrast, Clostridium propionicum, isolated from marine mud (Cardon and Barker, 1946), converts alanine, cysteine and serine via pyruvate (Hofmeister et al., 1993) to (R)-lactate, which is dehydrated via (R)-lactyl-CoA to acrylyl-CoA (Schweiger and Buckel, 1984;Hofmeister and Buckel, 1992) and reduced to propionyl-CoA (Hetzel et al., 2003). Like 2-hydroxyglutaryl-CoA dehydratase, lactyl-CoA dehydratase is an extremely oxygen sensitive radical enzyme (Parthasarathy et al., 2010). The rumen micro-organism Megasphaera elsdenii reduces about half of the consumed lactate in the same way as C. propionicum; the other half is oxidized to acetyl-CoA, from which butyrate (C2 + C2), valerate (C2 + C3) and caproate (3 × C2) are formed (Lewis and Elsden, 1955).
Three eliminations of water from 1,2-diols are known, each of which is catalyzed by two different enzymes, either coenzyme B 12 -dependent or by the much more oxygen-sensitive glycyl radical enzymes ( Table 2). The coenzyme B 12 -dependent glycerol dehydratase catalyzes the removal of the central hydroxyl group of glycerol yielding 3-hydroxypropanal (Forage and Foster, 1982), which can be reduced to 1,3-propanediol, a building block for polyesters. The discovery of the glycyl radical dehydratase from Clostridium butyricum (O'Brien et al., 2004) though extremely oxygen-sensitive, paved the road to a more economic fermentation of 1,3-propanediol from glycerol, because the already established procedure with the coenzyme B 12 -dependent enzyme required feeding with the precious vitamin B 12 . The mechanism of coenzyme B 12 -dependent eliminations was first studied with propane-1,2-diol dehydratase from Aerobacter aerogenes by Janós Rétey and Dulio Arigoni in Switzerland (Rétey et al., 1966) as well as Robert H. Abeles (Zagalak et al., 1966) in the United States. In the human gut, degradation of L-fucose (6-deoxygalactose) and L-rhamnose (6-deoxymannose) affords propane-1,2-diol, which is dehydrated to propanal, catalyzed by the glycyl radical enzyme propane-1,2-diol dehydratase (Levin and Balskus, 2018). The third pair of coenzyme B 12 and glycyl radical enzymes catalyzing the same reaction are classes II and III of ribonucleotide reductase (Greene et al., 2020). The formation of the conserved thiyl radical in this enzyme is performed in class II by the 5 -deoxyadenosyl radical from coenzyme B 12 and in class III by the glycyl radical. Abstraction of the 3 -hydrogen of the ribonucleotide by the thiyl radical and deprotonation of the 3 -hydroxyl group affords a nucleophilic ketyl (radical anion), which eliminates the adjacent 2 -hydroxyl group (Buckel, 2019). The reduction of the formed enoxy radical differs between the classes. In class II the reductants are two cysteine residues which form a disulfide, whereas class III uses formate for this purpose. In the eliminations of ammonia from ethanolamine or trimethylamine from choline, carbon nitrogen bonds are broken. Interestingly, for the deamination of ethanolamine only a coenzyme B 12 -dependent enzyme is known (Shibata et al., 2010), whereas the elimination of trimethylamine from choline is exclusively catalyzed by a glycyl radical enzyme (Craciun et al., 2014). Perhaps the different elimination mechanisms, but not the oxygen sensitivity, are responsible for the kind of radical enzyme applied. It has been proposed theoretically (Buckel, 1996;Feliks and Ullmann, 2012;Kovacevic et al., 2018) and demonstrated experimentally (Levin and Balskus, 2018) that in the dehydration of propane-1,2-diol catalyzed by the glycyl radical enzyme, the hydroxyl group at C2 is eliminated. Coenzyme B 12 -dependent propane-1,2diol dehydratase and glycerol dehydratase, however, shift the hydroxyl group at C2 to C1, forming propane-1,1diol or 3-hydroxypropane-1,1-diol which dehydrate to the aldehydes. Because trimethylamine is a better leaving group than ammonia, chemistry might compel choline lyase to use the glycyl radical mechanism. Trimethylamine (TMA) is of medical importance, because in the liver it is oxidized to the N-oxide (TMAO), which may cause cancer and other diseases (Gatarek and Kaluzna-Czaplinska, 2021).
The oxidation of pyruvate to acetyl-CoA is a complex reaction for which nature has developed thiamin diphosphate, which causes Umpolung of the carbonyl group of the 2-oxo acid to enable decarboxylation. Most likely a concealed radical is involved in the reaction (Chen et al., 2018). The oxidant is either NAD + or ferredoxin. In aerobes with NAD + the reaction is irreversible, but reversible with ferredoxin in anaerobes (Eqs. 10, 11). Thus this reaction is an example of anaerobes being more efficient energy converters than aerobes.
Under strict anaerobic conditions, pyruvate dehydrogenase of E. coli (Eq. 10) is replaced by the glycyl radical enzyme, pyruvate formate lyase (PFL), which catalyzes the reversible cleavage of pyruvate to acetyl-CoA and formate (Eq. 12). The enzyme is also present in other enterobacteria and several clostridia.
The degradation of taurine (2-aminoethylsulfonate) in aerobes and facultative anaerobes affords sulfoacetaldehyde by amino transfer to pyruvate. The aldehyde is converted by the TDP-dependent sulfoacetaldehyde acetyltransferase Xsc to acetyl phosphate and sulfite (Ruff et al., 2003). Under the strict anaerobic conditions in the gut, the Bacillus Bilophila wadsworthia bypasses Xcs with an aldehyde reductase and the glycyl radical enzyme isethionate lyase. This enzyme catalyzes the radical diol dehydratase-like elimination of isethionate (2-sulfoethanol) yielding acetaldehyde and sulfite which is reduced to H 2 S (Peck et al., 2019). AdhE catalyzes the oxidation of acetaldehyde to acetyl-CoA, which as well as acetyl phosphate yields ATP by SLP. The genes iseG and iseH coding for isethionate lyase and its activating enzyme were also detected in several well-known sulfate reducing organisms (Dawson et al., 2021). The genes from Desulfovibrio vulgaris str. Hildenborough were expressed in E. coli and the produced proteins characterized as active isethionate lyase (Xing et al., 2019;Wei and Zhang, 2021).
In this review it has become apparent that radical reactions play an important part in the metabolism of anaerobes and most likely already in prebiotic chemistry (Dragicevic et al., 2015). Table 2 summarizes the pairs of oxygen-tolerant and intolerant enzymes which catalyze radical reactions leading to the same product or are key enzymes of alternative pathways. Whereas the oxygen-intolerant 2-hydroxyacyl-CoA dehydratases and glycyl radical enzymes have simple cofactors, the oxygen-tolerant coenzyme B 12 and TDPdependent enzymes use the most complex cofactors known in biochemistry (Jurgenson et al., 2009;Moore and Warren, 2012). Apparently, the oxygen-intolerant radical enzymes evolved first and prevailed under the primeval atmosphere until the oxygen concentration raised, which forced nature to develop the oxygen tolerant and much more complex cofactors coenzyme B 12 and TDP. The oxygen-intolerant radical enzymes survived only in strictly anaerobic places such as the human gut and marine sediments. Probably, these coenzyme B 12 and TDP dependent fermentations were not present at the origin of life, but emerged just before or during the Great Oxidation event at 2.1 billion years ago (Anbar et al., 2007).

BUTYRATE PROVIDES ANAEROBIOSIS IN THE GUT
Many obligate anaerobic bacteria in the human gut produce butyrate. The most common pathway is the condensation of two acetyl-CoA to acetoacetyl-CoA, followed by the first reduction to (S)-3-hydroxybutyryl-CoA, dehydration to crotonyl-CoA and the second reduction to butyryl-CoA as shown in the fermentation of glutamate via 3-methylaspartate. The first reductants are NADH or at low acetyl-CoA concentrations NADPH as found in Clostridium kluyveri. The stronger reductant NADPH helps to shift the unfavorable equilibrium of acetyl-CoA condensation to acetoacetyl-CoA toward (S)-3-hydroxybutyryl-CoA (Figure 5). The second reduction by NADH is coupled to the reduction of ferredoxin via electron bifurcation (Buckel and Thauer, 2013). Butyrate is released either by a CoA-transferase with acetate or by SLP via butyryl-phosphate (Louis and Flint, 2017). This pathway is used by most organisms fermenting carbohydrates, as are bacteria related to Roseburia intestinalis, Faecalibacterium prausnitzii, and Eubacterium hallii (all belong to Firmicutes). They are among the most abundant known butyrate-producing bacteria in human feces, comprising about 8% of the total flora (Hold et al., 2003). The majority of the bacteria in the human gut belong to Bacteroidetes Huttenhower and Gevers, 2012), which produce acetate and succinate rather than butyrate. As mentioned above A. fermentans, M. elsdenii, and C. difficile also use electron bifurcation for butyrate synthesis.
To the author's knowledge, electron bifurcation with crotonyl-CoA appears to be the only pathway to butyrate in anaerobes. Probably no anaerobe wants to abandon the large energy released during butyrate synthesis by taking another route. Some acetate assimilating aerobes, however, reduce crotonyl-CoA with NADH directly to butyryl-CoA without proceeding via Etf. Actually the enzyme acts as crotonyl-CoA carboxylase yielding ethylmalonyl-CoA. Only at low CO 2 concentrations the product is butyryl-CoA (Erb et al., 2007).
In the human gut, butyrate is the preferred nutrient of the mucosa cells, which form the inner wall of the large intestine, also called colonic epithelia cells. Butyrate activates the nuclear peroxisome proliferator-activated receptor-γ (PPAR-γ), a transcription factor, which in turn stimulates mitochondrial βoxidation. The very active β-oxidation of butyrate is fueled by oxygen from the blood and consumes any remaining oxygen inside of the intestine resulting in an oxygen-free medium, where anaerobes optimally thrive. PPAR-γ also inhibits expression of the NO2 gene, which codes for an enzyme that catalyzes the oxidation of arginine with oxygen to nitric oxide (NO). NO is further oxidized to nitrate, an electron acceptor for potentially pathogenic enterobacteria. For example, E. coli, the best known bacterium of the human gut, is well adapted to the strict anaerobic environment as demonstrated by its glycyl radical enzymes class III ribonucleotide reductase and pyruvate:formate lyase (PFL, eq. 11). However, in the presence of oxygen or nitrate the facultative anaerobe forms pyruvate dehydrogenase (eq. 10) and multiplies very fast, from about 1% of the anaerobic microbiome to a major inhabitant of the gut. In case, the quickly spreading E. coli strain is pathogenic, severe diseases can occur. To prevent such events, butyrate nourishes colonic epithelia cells, which provides an oxygen-free medium for the microbiome in the large intestine and protects the host from harmful bacteria (Cani, 2017). For a constant butyrate production, the microbiome must be well fed with carbohydrates and amino acids. However, glucose, lactose, starch, glycogen and most proteins are digested and resorbed already in the small intestine, and nothing is left for the microbiome in the gut. Therefore, the human diet must contain indigestible carbohydrate and protein fibers for the mircrobiome, which is able to hydrolyze almost every glycosidic and peptide bond. Hence, fibers for the human diet are necessary for a healthy gut. The microbiologists Erica and Justin Sonnenburg remind their children to eat enough fibers by saying "Feed your bugs!" (Sonnenburg and Sonnenburg, 2015).

AUTHOR CONTRIBUTIONS
The author confirms being the sole contributor of this work and has approved it for publication.