Redox Balance in Lactobacillus reuteri DSM20016: Roles of Iron-Dependent Alcohol Dehydrogenases in Glucose/ Glycerol Metabolism

Lactobacillus reuteri, a heterofermentative bacterium, metabolizes glycerol via a Pdu (propanediol-utilization) pathway involving dehydration to 3-hydroxypropionaldehyde (3-HPA) followed by reduction to 1,3-propandiol (1,3-PDO) with concomitant generation of an oxidized cofactor, NAD+ that is utilized to maintain cofactor balance required for glucose metabolism and even for oxidation of 3-HPA by a Pdu oxidative branch to 3-hydroxypropionic acid (3-HP). The Pdu pathway is operative inside Pdu microcompartment that encapsulates different enzymes and cofactors involved in metabolizing glycerol or 1,2-propanediol, and protects the cells from the toxic effect of the aldehyde intermediate. Since L. reuteri excretes high amounts of 3-HPA outside the microcompartment, the organism is likely to have alternative alcohol dehydrogenase(s) in the cytoplasm for transformation of the aldehyde. In this study, diversity of alcohol dehydrogenases in Lactobacillus species was investigated with a focus on L. reuteri. Nine ADH enzymes were found in L. reuteri DSM20016, out of which 3 (PduQ, ADH6 and ADH7) belong to the group of iron-dependent enzymes that are known to transform aldehydes/ketones to alcohols. L. reuteri mutants were generated in which the three ADHs were deleted individually. The lagging growth phenotype of these deletion mutants revealed that limited NAD+/NADH recycling could be restricting their growth in the absence of ADHs. Notably, it was demonstrated that PduQ is more active in generating NAD+ during glycerol metabolism within the microcompartment by resting cells, while ADH7 functions to balance NAD+/NADH by converting 3-HPA to 1,3-PDO outside the microcompartment in the growing cells. Moreover, evaluation of ADH6 deletion mutant showed strong decrease in ethanol level, supporting the role of this bifuctional alcohol/aldehyde dehydrogenase in ethanol production. To the best of our knowledge, this is the first report revealing both internal and external recycling for cofactor homeostasis during 3-HPA conversion in L. reuteri.


Introduction
The current study was initiated to explore the diversity of ADH enzymes in Lactobacillus species and L. reuteri DSM20016 in particular, with a subsequent focus on the iron-dependent group III ADHs and their role in the redox balance in glycerol/glucose metabolism.

Bacterial strains, plasmids and primers
The bacterial strains and plasmids used and the primers constructed in this study are listed in S1 Table. Escherichia coli strains DH5α was used as an intermediate cloning host and was grown at 37˚C in Luria Bertani (LB) broth. For preparation of inoculum of L. reuteri, wild type as well as mutants, 20 mL of 55 g/L MRS broth (Difco) supplemented with 20 mM glycerol was added to the 30 mL serum bottles, boiled, and bubbled with nitrogen gas, after which the bottles were closed with rubber stoppers, and autoclaved at 121˚C for 10 min. After inoculation of the medium, the cells were grown anaerobically without shaking for 8 h at 37˚C. The stock culture was stored at -80˚C in MRS broth containing 20% v/v glycerol. For selective cultivation of the mutants, appropriate antibiotics were supplemented to the media. For recombinant E. coli, 10 μg/mL chloramphenicol and 250 μg/mL erythromycin were used, while for recombinant L. reuteri 10 μg/mL chloramphenicol and 30 μg/mL erythromycin were used.

Chemicals and reagents
All restriction enzymes, X-Gal and isopropyl-β-D-1-thiogalactopyranoside (IPTG) were purchased from Fermentas. T4 DNA ligase and Taq DNA polymerase were from New England Biolabs, while BugBuster protein extraction reagent was from Novagen. β-Nicotinamide adenine dinucleotide, reduced disodium salt hydrate (NADH), β-Nicotinamide adenine dinucleotide (NAD + ), and all antibiotics were purchased from Sigma-Aldrich. Primers and other kits for DNA extraction and purification were from Thermo Fisher Scientific.

Bioinformatics analysis
Geneious 9.1.2 was utilized for detection of enzymes in the Protein DataBank having sequence resemblance to ADHs. Multiple sequence alignments for identification of conserved amino acids were analyzed by ClustalW2 tools (EMBL-EBI) and visualized in Geneious. The derived Lactobacillus ADH sequences for phylogenetic analysis were extracted from protein database UniProt. Homology modeling of the L. reuteri ADHs was performed with the aid of Program DeepView (Swiss PDB-Viewer) [21]. Homology model quality assessment was performed by utilizing the SWISS-MODEL workspace. Chimera developed by the UCSF Resource (http:// www.cgl.ucsf.edu/chimera) was used as a platform for the study of dockings with NADH using the tool AutoDock Vina [22,23]. Figures were generated using either Chimera or PyMol (http://www.pymol.org) and raytraced images are produced with POV-Ray. Global alignment (Cost Matrix: Blosum 62, Gap open penalty:12, Gap extension penalty:3) was adapted for building distance matrix by using Neighbor-Joining (NJ) as the tree bulid method and Jukes-Cantor as genetic distance model (bootstrap number: 100). No outgroup was chosen for rooting the phylogenetic tree.

Construction of plasmids and mutant strains
The ADH genes were amplified from the genomic DNA of L. reuteri DSM20016. The approach to obtain gene deletion variants was based on double-crossover integration method reported earlier by Jolamda et al. [24]. Generally, about 1-kb fragment of the sequences upstream and downstream of the target locus were amplified and cloned into the SwaI and SrfI blunt-end restriction sites of pNZ5319, respectively. Colony PCR was performed to identify the colonies harboring the anticipated insert in the desired orientation.
Using this approach, the PduQ mutagenesis vector pLCH007 was constructed by successive cloning of both 5'-and 3'-flanking regions of PduQ gene (lreu_1734). Clones that harbored the desired inserts were identified by PCR using primer sets 13-16 (S1 Table). Likewise, vectors pLCH008 and pLCH009 were successfully constructed to harbor the 5'-and 3'-flanking regions of ADH6 and ADH7 gene inserts, respectively. Approprate primer sets 17-24 were designed to check the correct insertion.
The gene-specific mutagenesis vectors (1-4 μL) were transformed into L. reuteri DSM20016 in order to engineer lox66-P 32-cat-lox71 gene replacement by electroporation as previously described [25], with slight modification. An overnight culture of the transformed L. reuteri strains was used to inoculate 15 mL MRS broth containing 2% glycine and 0.5 M sucrose, and incubated at 37˚C. The absorbance of the culture at 600 nm reached 0.2 after 3 h, after which the cells were separated by centrifugation and then washed twice with 5 mL cold distilled water followed by a 10 min incubation in 50 mM EDTA solution (pH 8.0). After washing with cold distilled water and 0.3 M sucrose, the cells were re-suspended in 100 μL of 0.3 M sucrose solution as electroporation buffer. Then the suspension was transferred into a disposable cuvette (Bio-Rad Laboratories) and subjected to an electric pulse using a MicroPulser (Bio-Rad) under the following conditions: Capacitance = 25 μF; Resistance = 400 ohm; Voltage = 2.0 kV.
In order to cut off the P 32 -cat cassette from the chromosome, 1-4 μL of cre expression vector pNZ5348 was transformed into the mutants. Em r (erythromycin resistance) colonies were tested by PCR for desired Cre-mediated recombination, using primer pairs of 25:26, 27:28 and 29:30 for PduQ, ADH6 and ADH7, respectively. The pNZ5348 vector was cured from relevant colonies of L. reuteri mutants by growth without erythromycin selection pressure for about 10 generations.
All the above inserts were verified by DNA sequencing. Other molecular cloning and DNA manipulation procedures, not mentioned here, were essentially done as described by Sambrook et al. [26].

Recombinant expression and production of L. reuteri PduQ and ADH7
The gene sequences coding for PduQ and ADH7 genes were amplified by using appropriate primer pairs (S1 Table). Initial denaturing at 95˚C for 3 min was followed by 30 cycles of denaturing at 95˚C for 30 s, annealing at 54˚C for 30 s and elongation at 72˚C for 1 min. A final elongation was performed at 72˚C for 15 min. The PCR product was separated and purified by agarose gel electrophoresis and inserted directly into the expression vector pET21a after digesting both vector and PCR product with the restriction enzymes BamHI and XhoI. The resulting plasmids pET21a-pduQ-His 6 / pET21a-ADH7-His 6 were introduced in E. coli BL21 (DE3) and the recombinant strains were designated as E. coli PduQ-WT and ADH7-WT, respectively.
The recombinant E. coli strains were then grown in 100ml LB medium of 500 mL flask containing 40 μg/mL Ampicillin under aerobic conditions at 37˚C with shaking at 220 rpm. The gene expression was induced by addition of 0.1mM IPTG when the bacterial growth had reached mid-exponential phase (OD 620 around 0.5) with subsequent incubation for 24 h at 15˚C and 160 rpm. The cells were then harvested and washed with 20 mM sodium phosphate buffer pH 7.4 containing 0.5 M NaCl.
Lysis of the cells was performed by re-suspending and incubating the cells in 5 ml Bugbuster protein extraction reagent and 10 μL lysonase bioprocessing reagent (Novagen) according to manufacturer´s instructions. The soluble protein fraction was separated from cell debris by centrifugation for 40 min at 16 000 x g, 4˚C, and subjected to immobilized metal ion affinity chromatography (IMAC) for purification of the recombinant enzymes. The enzyme solution was loaded on 5 ml Ni-NTA HisTrap™ FF crude column equilibrated with binding buffer at a flow rate of 5 mL/min. After washing the column with the equilibration buffer, the bound protein was eluted using elution buffer containing 500 mM imidazole solution.
The eluted enzyme was concentrated 10 fold at 4˚C in Vivaspin 15R (MWCO 10kDa) centrifugal concentrator tubes using a Sigma 3-16PK centrifuge at 3000 g for about 60 minutes. The concentrated enzyme solution (volume%1mL) was mixed with 900 μL of pure glycerol and 20 μL of antioxidant stock solution (1M ascorbic acid, 50 mM copper sulphate), and stored in an airtight anaerobic glass vial at -20˚C. Purity of the recombinant PduQ and ADH7 was analysed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) on a gel containing 12% acrylamide. The protein bands were stained with Coomassie Brilliant Blue R-250.

Analyses
Cell growth was monitored at 600 nm using a Ultrospec 1000 spectrophotometer (Pharmacia Biotech) and correlated with cell dry weight (CDW) as described previously [12]. For CDW determination, the culture broths (5 mL) were harvested and centrifuged (4000 x g) in a preweighed tube. After removing the supernatant, cell pellets were dried overnight in an oven at 110˚C and then weighed again. The difference in weight is then refered to the cell dry weight in 5 mL cell culture.
Glucose, glycerol, lactate, ethanol, 1,3-PDO and 3HP concentrations were determined by HPLC (JASCO, Tokyo, Japan) equipped with refractive index detector. Separation of the compounds was performed on Aminex HPX-87H chromatographic column using sulfuric acid (50 mM) as mobile phase. The modified colorimetric method of Circle et al. [27] was used for measuring 3HPA concentration using acrolein as standard [28].
The activity of PduQ and ADH7 was determined as described by Reid et al. [20] with some modification. Aldehyde reduction was measured by following the decrease in absorbance of NADH at a wavelength of 340 nm (Δε 340 NADH = 6.22mM -1 cm -1 ) in a reaction mixture containing 0.2 mM NADH, 100 mM 3-HPA and appropriate amount of enzyme in 50 mM potassium phosphate buffer pH 7.0, if not mentioned otherwise. For determining alcohol oxidation, the reaction mixture contained 5 mM NAD + , 500 mM 1,3-PDO and the enzyme in 50 mM Tris-HCl pH 9.0. All spectrophotometric measurements were made using UV-1650 PC Spectrophotometer (Shimadzu, Kyoto, Japan) at 30˚C. 3-HPA was produced by biotransformation of glycerol in aqueous solution using resting cells of L. reuteri as described by Sardari et al. [29]. One Unit (U) of the enzyme activity is defined as the amount of enzyme catalyzing the formation or consumption of 1 μmol NADH per minute under the standard assay conditions. Kinetic parameters were obtained by non-linear curve fitting to the Michaelis-Menten equation v = V max [S]/(K m +[S]) using GraphPad Prism6 software (http://www.graphpad.com).
carbohydrates into the typical three-carbon intermediates [30]. In general, homofermentative LAB convert carbohydrates into lactate using the Embden-Meyerhof pathway (EMP), whereas heterofermentative LAB produce lactate, acetate and ethanol using the phosphoketolase pathway (PKP) by the enzymes listed in Table 1.

Alcohol dehydrogenases (ADHs) from lactobacilli
To date, Lactobacillus is the largest genus of the LAB group, with over 50 species in total [32].
In order to analyse diversity amongst Lactobacillus ADHs and related proteins, all their amino acid sequences (2533 in number) available from public databases (UniProt, NCBI and PDB) were compiled and aligned using Geneious19.1.2 for the purpose of constructing a phylogenetic tree using the average distance method and bootstrap analysis (Fig 3). The resulting phenogram revealed that within the NAD(P)-dependent group, the Lactobacillus ADHs are clustered in order according to enzyme type, i.e. zinc-dependent ADHs, short-chain ADHs, and iron-containning ADHs, irrespective of the species of Lactobacillus. It is worth mentioning at this point that ADHs from closely related species are often clustered together. Using this method, unknown Lactobacillus ADHs can be classified via phylogenetic relatedness into their  respective ADH types, and as observed in the phylogenetic tree more than 70% of known Lactobacillus ADHs are grouped into Zn-dependent ADH type. Normally, zinc-containing long/ medium-chain enzymes (group I ADHs) and metal-free short-chain enzymes (group II ADHs) are approximately 350 or 250 amino acids, respectively, in size, while the iron-containing ADHs (group III ADHs) exhibit a size of around 385 amino acid residues per subunit ( Fig 3B). Typically, screening of all ADHs in Lactobacillus shows a group of ADHs in the range of 850 aa to 900 aa, referred as the bifunctional alcohol/aldehyde dehydrogenases. All ADHs found in L. reuteri are listed in Table 1. A genome-wide transcription analysis of L. reuteri cells grown on glucose with differentially expressed genes in response to glycerol revealed 2.7 and 1.9 times higher expression of ADH7 and PduQ genes in exponential phase, and during stationary phase the expression levels were 2.9 and 1.5 fold higher, respectively [31]. In contrast, ADH5 and ADH6 genes were down-regulated in the presence of glycerol. Particularly, ADH6 gene expression dropped to 25% and 1% of the original in exponential and stationary phase, respectively, during glycerol feeding.
All the L. reuteri ADHs are metal-ion dependent enzymes, where the metal-ion can have a catalytic and/ or a structure stabilizing effect on the enzyme. Addtionally, homology modeling for the prediction of the tertiary structure of these ADHs were obtained through the online sever Swiss-model and structural analysis were performed using Pymol (Fig 4). The modeled monomer of all ADHs contains about 320-400 amino acid residues except for the ADH6 with 878 aa residues. In general, the monomeric protein of ADH6 (alcohol dehydrogenase part), ADH7 and PduQ fold into two domains that are separated by a deep cleft. The N-terminal domain has a so-called Rossmann-fold architecture, which is a typical protein structure motif that binds nucleotide cofactor such as FAD + , NAD + or NADP + . The C-terminal domain consists of α-helical structures, with an up-and-down bundle architecture well-known as a dehydroquinate synthase-like α-domain. Particularly, the binding site for cofactor NAD(P) + resides in the deep hydrophilic pocket between the two domains. The parameters used for homology modeling and the outcome are listed in S2 Table. Our laboratory studies have shown these ADH enzymes to have low affinity and catalytic efficiency towards ethanol as substrate (Unpublished data). It has been speculated that most group III ADHs are mainly involved in aldehyde reduction in bacterial species rather than alcohol turnover. Nevertheless, the other alcohol dehydrogenases (ADH1-ADH5, and ADH8) in L. reuteri are more likely to play a role in ethanol transformation as reported in other microorganisms [33].
To identify the genes involved in 3-HPA to 1,3-PDO conversion, a BLASTP analysis using PduQ as query was conducted among putative alcohol dehydrogenases in the genome of L. reuteri (NC_009513). The results indicated the presence of two other putative ADHs, i.e. ADH6 (lreu_0321, E-value: 2e -53 ) and ADH7 (lreu_0030, E-value: 5e -46 ). To identify features unique to these ADHs, protein sequence alignment of PduQ, ADH6 (truncated) and ADH7 was carried out after being aligned and adjusted by gap insertion. Similarities (Blosum62 with threshold 1) ranging from 42.9% to 48.4% were observed (Fig 5). The conserved amino acid sites marked by boxes revealed the common features for substrate/NADH/iron binding as in the iron-dependent ADHs. The specificity of all the enzymes towards NAD(P) + is determined by motif G-G-G-S-X-X-D-X-X-K. In addition, five conserved metal-ion coordination residues, Asp207, His211, His276, His280 and His290 were found to be involved in the Fe 2+ -binding.

Characteristics of the ADH mutant L. reuteri cells
To elucidate the iron-dependent alcohol dehydrogenase activity of ADH6, ADH7 and PduQ, three deletion strains were constucted using Cre-lox-based system resulting in strains with curing LCH010 (Δlreu_1734), LCH011 (Δlreu_0321) and LCH012 (Δlreu_0030), respectively (S1 Table). The strategy for construction of mutant for removal of lreu_1734 gene, lreu_0321 gene and lreu_0030 gene is illustrated in Fig 6. Addition of glycerol to mutant cells is expected to achieve regeneration of NAD(P) + to various levels that in turn will have an impact on other coupled physiological reactions in the central metabolism. Therefore, tests on growth characterisitics of mutants were carried out by growing both wild type cells and mutants in MRS medium with or without glycerol supplementation. The growth rate without glycerol was observed to be similar for wild type (0.71 h -1 ) and LCH010 (0.70 h -1 ), but slightly decreased for LCH012 (0.63 h -1 ) and LCH011 (0.53 h -1 ) (Fig 7A). With the supplemented glycerol, the maximum growth rate of all cells increased by    lreu_0321 gene (B) and lreu_0030 gene (C). Firstly, the gene-specific mutagenesis vector was transformed into cells, after which the target gene was replaced by a lox66-P32-cat-lox71 cassette according to homologous recombination. Then, the double-crossover mutant was selected by proper antibiotics, after which the lox66-P32-cat-lox71 cassette would be resolved to a single double-mutant lox72 site by transient Cre expression from the curable plasmid. Primers used were indicated as black arrowheads (see S1 Table). about 28%, 16%, 66% and 16% for wild type cells, LCH010, LCH012 and LCH011, respectively. Particularly for LCH012, the higher growth rate (0.88 h -1 ) suggests higher ATP production rate, and as a consequence end-product yield in this mutant can be partly replenished by the use of glycerol as the electron acceptor. Therefore, quantification of the metabolic end-products at an early stationary phase would provide more insight for the mutant behavior on NADH/NAD + homeostasis.
As shown in Fig 7B, when grown in the presence of glycerol (20 mM), there was no notable difference in the glucose and glycerol comsumption by the wild type and all mutant strains. However, the clear differences were observed in end-product concentration, i.e. lactate, ethanol and 1,3-PDO. The wild type, LCH010 and LCH012 produced around 20 mM lactate, whereas LCH011 produced 26.4 mM lactate, which is 32% higher compared to the others. This could suggest that LCH011 had to resort to the use of LDH activity for regeneration of NAD + . More interestingly, nearly similar production of ethanol by wild type cells (16.9 mM) and LCH010 (18.5mM) was observed, while LCH012 produced relatively high amount of ethanol (28.2 mM) and ADH6 knockout strain LCH011 showed distinctly lower ethanol production (8.4 mM). This observation accords with the idea that ADH6 enzyme is a bifunctional alcohol/ aldehyde dehydrogenase; similar enzymes are found in many other fermentative cells, and Roles of Iron-Dependent Alcohol Dehydrogenases in Lactobacillus reuteri participate in ethanol generation by catalysing the conversion of an acyl-coenzyme A to an alcohol via aldehyde as an intermediate. Higher level of ethanol and lower 1,3-PDO production obtained in case of LCH012, obviously suggests the preference of the bifunctional ADH6 to regenerate NAD + and also lower activity of PduQ for converting 3-HPA to 1,3-PDO. Indeed, up to 13 mM 1,3-PDO is produced by the wild type strain, LCH010 and LCH011, whereas production of 1,3-PDO in LCH012 was much lower (5.3 mM), which strengthened the assumption that ADH7 was most likely involved in the conversion of 3-HPA to 1,3PDO outside of Pdu microcompartment.
3-HPA, 3HP and 1,3-PDO production was then studied in a two-step process, in which L. reuteri cells were first cultivated anaerobically in MRS medium supplemented with 20 mM glycerol, followed by using the cells for transformation of 300 mM glycerol. The wild type and LCH011 strain produced nearly similar amounts of 3-HPA, i.e. 138.6 mM and 137.3 mM, respectively, while strain LCH010 produced 152.5 mM and LCH012 produced only 115.3 mM of 3-HPA (Fig 7C). The amounts of by-products 1,3-PDO and 3-HP formed by the wild type L. reuteri were 18.9 mM and 23.7 mM, respectively. The strain LCH010 produced 7.7 mM 1,3-PDO and 16.2 mM 3-HP, which results in a significantly higher ratio of 3-HPA/3-HP and 3-HPA/1,3-PDO as described in Fig 7D, showing that the ADH encoded by PduQ gene is mainly active in non-growing cells.

Kinetic analysis of recombinant PduQ and ADH7 activities
Since PduQ and ADH7 seemed to catalyse 3-HPA reduction, both the enzymes were produced recombinantly and purified for further characterization. Studies with PduQ-His 6 showed that the iron content of the enzyme (measured by ICP-MS) as well as the activity was lower in the enzyme exposed to air as compared to that under anaerobic conditions, confirming the iron dependence of the enzyne (unpublished data). Kinetic characterization of PduQ-His 6 and ADH7-His 6 with varied concentrations of aldehyde and alcohol substrates, respectively, obtained from GraphPad Prism6 showed the enzymes to exhibit significantly higher activity (K cat /K m ) for aldehyde reduction than the oxidation of the corresponding alcohol, attributed to lower K m as well as higher V max with the aldehyde and NADH (Table 2). Propionaldehyde was a better substrate than 3HPA (not shown). Measurements at varying concentrations of cofactor with 3HPA or 1,3-propanediol as substrate revealed that the enzyme activity increased with increase in NADH or NAD + concentration up to a certain extent. Maximum specific activity of PduQ-His 6 for aldehyde reduction with NADH as cofactor (12.6± 0.7 U/mg) was about three-fold higher than that obtained for alcohol oxidation with NAD + (3.9± 0.2 U/mg). On the other hand, the difference between the V max-NADH and V max-NAD+ with the purified ADH7-His 6 was eight-fold (33±2.1 vs 4.1±0.2 U/mg).

Discussion
Several microorganisms possess the ability to metabolize glycerol both reductively and oxidatively, e.g. Klebsiella, Clostridium or Enterobacter [34]. These species can utilize the NAD +dependent enzyme glycerol dehydrogenase for the production of dihydroxyacetone (DHA), followed by subsequent phosphorylation to DHAP that gets funneled into glycolysis. However, L. reuteri does not grow on glycerol as the sole carbon and energy source, apparently due to an inactive glycerol dehydrogenase (lreu_1840, Fig 1). This assumption could also be confirmed by functional genomics analysis carried out by Santos et al. [31]. In their work, glycerol dehydrogenase gene was observed to be down-regulated 3.3-fold in the cells grown in the presence of glycerol compared to that without glycerol, and the extent of down-regulation was almost 10.8-fold once the cells entered early stationary (non-growth) phase. L. reuteri possesses theoretically an alternative glycerol-utilization pathway involving initial reactions catalyzed by glycerol kinase (lreu_1065) and glycerol-3-phosphate dehydrogenase (lreu_0371), and the obtained dihydroxyacetone phosphate (DHAP) is converted to intermediate glyceraldehyde 3-phosphate by triose-phosphate isomerase (lreu_0171, lreu_0237) before entering the glycolytic pathway (Fig 1). However, no significant evidence has been found so far to prove the feasibility of this pathway in L. reuteri; a possible reason could be the reversible redox conversion catalyzed by the key enzyme glycerol-3-phosphate dehyderogenase (GPDH), which plays a vital role in lipid biosynthesis and has been shown to allow the prompt dephosphorylation of glycerol 3-phosphate into glycerol [35].
Accorrding to Talarico et al. [5], addition of glycerol (20 mM) reduced the generation time of L. reuteri from 66 min to 42 min, and as a consequence increased the final cell mass yield about 3-fold. Although further increase in glycerol concentration to 40 mM did not influence the generation time, the cell mass yield was consistently increased. Indeed, supplementation of the growth medium with glycerol leads to an obvious increase in acetate and a significant decrease in ethanol production, as also observed in our study.
These observations suggest that the Pdu pathway is the important route for glycerol metabolism in L. reuteri, and that glycerol, serving as an electron acceptor, could compete with acetyl phosphate in glucose fermentation, resulting in a shift in metabolite production from ethanol to acetate and the accumulation of 1,3-PDO (Fig 1).
Earlier studies have indicated that an alcohol dehydrogenase encoded by the Pdu operon is involved in glycerol/1,2-PD metabolism in the microcompartment in L. reuteri as well as in Salmonella enterica, and the role of the enzyme is mainly to regenerate NAD(P) + from NAD (P)H [13,36]. It has also been shown that L. reuteri is able to excrete large amounts of 3-HPA, an antimicrobial compound, to the outside of microcompartment and into the extracellular medium while still maintaining viability at concentrations that are inhibitory to other microorganisms [37]. Hence, the self-defence mechanisms by 3-HPA detoxification in L. reuteri, as the producer strain, was assumed to exist by habouring other conversion routes of 3-HPA.
Some microorganisms have mutiple ADHs, sometimes of different types for different functions. For example, there is one well-characterised and 12 putative ADHs of one type found in Sulfolobus solfataricus [38], while Pyrococcus furiosus has two different types of ADHs [39]. The presence of mutiple ADHs within one organism conceivably reflects the environment to which the organism has been exposed and adapted. Therefore, each ADH may have a different role and specificity for survival.
In this study, we have identified nine different genes encoding alcohol dehydrogenases in L. reuteri DSM20016, three of which were hypothesized as genes encoding iron-dependent ADHs including PduQ present in the Pdu microcompartment. Deletion mutants of L. reuteri incorporating gene deletions of the iron-dependent ADHs ORFs (PduQ, ADH6, ADH7), respectively, displayed different behaviours in the production of metabolites. Supplementation with glycerol results in a higher growth of both L. reuteri wild type strain and mutants to different extents, but with little effect on glucose consumption. The mutants LCH010 (Δlreu_1734, PduQ deletion mutant) and LCH012 (Δlreu_0030, ADH7 deletion mutant) displayed 15.7% and 15.8% increased growth rate, respectively, in the presence of glycerol, compared to the wild type with 28% higher growth, implying an affected NAD + regeneration for 3-HPA reduction (Fig 7). Most strikingly, LCH011 (Δlreu_0321, ADH6 mutant) displayed 66% increased growth rate with glycerol. Due to the lowest growth rate without glycerol (0.53 h -1 ) and almost similar growth as the wild type L. reuteri in the presence of glycerol (0.88 h -1 ), ADH6 is not likely to be involved in 3-HPA reduction but is responsible for another reaction in glucose fermentation.
Indeed, shift in metabolites production was further detected in the wild type L. reuteri and mutants. Although all the strains showed similar comsumption behaviour for glucose and glycerol, LCH012 produced 7.5 mM less 1,3-PDO but 11.3 mM higher ethanol concentration compared to the wild type (Fig 7). What was most suprising was that LCH010, the PduQ deletion mutant, displayed similar production behaviour as the wild type. In contrast, the role of ADH6, shown by LCH011 mutant, seems to be different. With 25.1% increased production of lactate and 50.5% reduced ethanol production, ADH6 is assumed to play a role in cofactor regeneration in the last step of glucose fermentation by converting acetyl-CoA to ethanol instead of 3-HPA conversion.
Previous genetic tests illustrated that a PduQ deletion mutant grew slower than wild-type Salmonella on 1,2-PDO and could not be complemented by co-expressing a non-microcompartment associated Adh2 from Zymomonas sp. that catalyzes the same reaction [36]. This phenotype proposes that PduQ keeps a microcompartment-specific function. The overarching function of Pdu microcompartment is to optimize metabolic pathways by restricting the efflux of toxic or volatile intermediates while allowing enzyme substrates, products and cofactors to pass [40]. In another report, Cheng et al. reported that it could be possible to restore the growth of a PduQ deletion mutant to wild-type levels by genetically disrupting the MCP shell [36], which was ascribed to facilitating the free diffusion of the metabolites and NAD + /NADH recycling.
The bifunctional alcohol/aldehyde dehydrogenase (ADHE) enzymes, which catalyse the conversion of an acyl-coenzyme A to an alcohol via an aldehyde intermediate, by coupling with oxidation of two NADH molecules to maintain the NAD + pool during fermentative metabolism, are found in many microorganisms [41]. In L. reuteri, ADH6 is the bifuctional dehydrogenase, which seems to catalyse oxidation of acetyl-CoA to ethanol, and has little activity with 3-HPA substrate. The ADH-domain structure of ADH6 enzyme has been understood to consist of an Rossmann-fold architecture and an α-helical domain containing residues coordinating a metal ion, which is structurally similar to ADH7 and PduQ as shown by the homology-model. In accordance with the data for the E. coli ADHE [42], the activity of the L. reuteri ADH6 is only stimulated by the presence of Fe 2+ and not by other metal ions (unpublished data), which is also the case for enzymes reported for Streptococcus bovis and Entamoeba histolytica [43,44].
The exact mechanism of iron-dependent alcohol dehydrogenases catalysed reduction of 3-HPA to 1,3-PDO with concomitant NAD + generation remains to be elucidated. Nevertheless, the divalent metal ion in active site of the metal-dependent alcohol dehydrogenase (ADH) is believed to aid polarization of the aldehyde carbonyl O atom, facilitating reduction by NADH to proceed [37]. Furthermore, due to differences in the strength of the carbonyl group polarization, the charge density of the metal ion can have an effect on the catalysis rate [42].
Cumulatively, our studies have confirmed that the PduQ enzyme can convert NADH to NAD + to supply the cofactor needed for oxidation of 3-HPA (in a reaction catalysed by PduP) internally within the MCP, while ADH7 is mainly responsible for providing NAD + via 3-HPA to 1,3-PDO outside of the MCP. In fact, the latter enzyme exhibited faster reaction kinetics compared to PduQ (Table 2).

Conclusions
The study reveals the diversity of alcohol dehydrogenases in Lactobacillus species, and their significance in energy and redox balance is exemplified in case of L. reuteri that uses glycerol as an electron acceptor and transforms it to 1,3-PDO via 3-HPA. Although glycerol is not used as a carbon source by L. reuteri for cell growth, it influences metabolism and improves growth through cofactor regulation; as the cofactor requirement is different for growing and resting cells, glycerol plays different roles. It is yet to be determined to what extent the distribution of energy carriers, such as NAD + /NADH through the microcompartment shells contribute to the regulation.
The findings presented here are the first to show the 3-HPA reduction to 1,3-PDO is not limited just to the activity of PduQ located in the bacterial microcompartment but also to ADH (ADH7) outside of the microcompartment, however with distinct roles in cofactor regeneration. Moreover, ADH6, the bifunctional alcohol/aldehyde dehydrogenase, does not play a role in 3-HPA reduction but is responsible for cofactor regeneration in the last step of glucose fermentation by converting acetyl-CoA to ethanol. Further studies on the individual enzymes are ongoing for gaining understanding of structure-function relationship as well as potential in biocatalysis.