In this study we focused on the production of 1,3-PDO from glycerol using two metabolic steps with 3-hydroxypropionic aldehyde (3-HPA) as an intermediate. The conversion of glycerol to 3-HPA can be achieved using two different types of dehydratases: (1) B12-dependent dehydratases and (2) B12-independent dehydratases (Knietsch et al., 2003; O'Brien et al., 2004; Rieckenberg et al., 2014; Liu et al., 2016). In this study to avoid the mechanism based inactivation of B12-dependent dehydratases that leads to vitamin B12 cleavage (Toraya et al., 1976; Yamanishi et al., 2012), we selected the B12-independent glycerol dehydratase from Clostridium butyricum (Raynaud et al., 2003). However, members of the B12-independent class of dehydratases are extremely oxygen sensitive and very difficult to handle in vitro (Raynaud et al., 2003). In our approach, the conversion of glycerol to 1,3-PDO requires: (1) the dehydration of glycerol to 3-HPA with the B12-independent glycerol dehydratase (dhaB1) and its activating protein (dhaB2) and (2) the hydrogenation of 3-HPA to 1,3 PDO with a 1,3-PDO dehydrogenase (dhaT) (Figure 1) (O'Brien et al., 2004; Lama et al., 2015). All enzymes considered in this study were selected from C. butyricum. These enzymes were expressed individually and as fusion proteins with a dockerin module from either Acetivibrio cellulolyticus, Bacteroides cellulosolvens, or Clostridium thermocellum (Figure 1). These fusion constructs are designed to bind to their cohesin binding partner on a synthetic protein scaffold through their strong natural and specific interactions. The scaffolds are composed of a CBM3a (family 3a carbohydrate binding module) able to specifically bind to cellulose for isolation and purification of the complex, and cohesin modules from different microorganisms that can specifically bind their corresponding dockerins found on the enzymes. All primers and constructs used in this study can be found in Supplementary Tables S1,S2.

Given that one of the key metabolic enzymes, the glycerol dehydratase activating enzyme (dhaB2), involved in the production of 1,3-PDO via 3-HPA is very difficult to express and purify (data not shown), we conducted most of the pathway and enzyme characterization in vivo before demonstrating the promise of our approach in vitro. We first confirmed the production of 1,3-PDO using the glycerol to 1,3-PDO pathway from C. butyricum with the glycerol dehydratase (dhaB1), dhaB1 activating protein (dhaB2), and the NADH-dependent 1,3-propanediol dehydrogenase (dhaT). The resulting E. coli strain encoding this new operon was able to achieve the production of more than 26 mM of 1,3-PDO in 72 h (Figure 2). Note that this strain was not optimized for 1,3-PDO production as this was not the goal of this study but was rather used as a benchmark for subsequent testing. The parent strain of E. coli used in this study was not able to produce 1,3-PDO from glycerol nor was a strain encoding for dhaB1 and dhaB2 even though E. coli possesses an alcohol dehydrogenase, YqhD, known to be promiscuous (Sulzenbacher et al., 2004; Jarboe, 2011) (with a function similar to that of dhaT) (Figure 2). The lack of production of 1,3-PDO in the latter strain indicates that the deletion of YqhD in the parent strains was not necessary to achieve a clean background for 1,3-PDO production.

We then systematically evaluated the impact of adding dockerin modules to these metabolic enzymes within the pathway. In all cases, the activity of the fusion enzymes that included the dockerin modules was drastically reduced with a corresponding production of 1,3-PDO barely reaching 3.0 mM after 72 h of incubation compared to 26 mM for the strain encoding for the native enzyme operon (Figure 3). There was almost no correlation between the origin of the dockerin module and the impact on activity as all enzymes were equally inhibited by dockerins (Figures 3A–C, and Supplementary Figures S2, S3). Specifically, Figure 3A shows that linking the AcDoc to dhaB1 results in a large activity loss with similar results when AcDoc is linked to dhaB2 (Figure 3B) and dhaT (Figure 3C).

Surprisingly, the activity of the enzyme and the pathway can be mostly restored when co-expressing a scaffoldin protein (Cohesin-CBM) (Figures 3A–C). Even in the most extreme case with each enzyme being expressed as fusion proteins with a dockerin module, the full length scaffoldin protein restores close to 80% of the 1,3-PDO production, producing 20.1 mM of 1,3-PDO compared to 26.5 mM for the native enzymes after 72 h (Figure 3D). It appears that the binding of the dockerin to its cohesin partner is responsible for this recovery. Another important aspect of the metabolic complex is the presence of a carbohydrate binding module that would allow one step purification of the complex on cellulose for cell-free applications.

Additionally, to evaluate the impact of the CBM module on the production of 1,3-PDO, we considered two different strains both encoding the three fusion enzymes and the protein scaffold but only one containing the CBM. We conducted longer fermentations with all three enzymes as dockerin fusions with the scaffoldin. As shown earlier, in the presence of scaffoldin, 1,3-PDO production was in the vicinity of 80% compared to the free enzyme reference after 72 h of incubation with glycerol but could actually achieve more than 90% recovery after 96 h (Figure 4A). Also, it appears that the impact of the CBM is negligible with achieving a titer of 24.1 mM of 1,3-PDO after 72 h compared to 22.7 mM in the case of the complex lacking the CBM and the same trend continues even after 140 h (Figure 4A).

To determine if other mechanisms were at play regarding this interesting activity recovery, we conducted other controls by simply co-expressing a carbohydrate binding module in the presence of a dhaB1-doc fusion enzyme or co-expressing the same dhaB1-doc fusion enzyme with an incompatible cohesin protein. In both cases the production of 1,3-PDO remained negligible with no significant changes after 72 h (Figures 4B,C). These results indicate that only specific binding of the enzyme-dockerin fusion to its cohesin partner allows for restored activity.

To assemble and isolate the metabolic complex, E. coli cells encoding for the co-expression of the three fusion enzymes and the full length scaffoldin proteins were grown anaerobically and the lysis, immobilization, and assays were also performed under anaerobic conditions (Figure 5). Plans for individual enzyme characterization were abandoned after it became clear that we could not successfully purify dhaB2 contrary to previous reports (O'Brien et al., 2004). Despite attempting anaerobic purification with and without iron-sulfur cluster reconstruction, aerobic purification with metals removed using EDTA followed by iron-sulfur cluster reconstruction, and various expression and purification conditions with or without affinity tags, we were not able to isolate an active dhaB2 (results not shown). Instead, the supernatant of the resulting cell lysate was incubated with cellulose to isolate the complex via interactions with the carbohydrate binding module. The cellulose was then washed extensively to remove all potential substrates, intermediates, and products produced during expression. The resulting immobilized metabolic pathway was incubated with glycerol for 48 h at room temperature (Figure 6). The same procedure was followed for the parent E. coli strain and the E. coli strain encoding for the free metabolic pathway. Three different glycerol concentrations (4/11/40 mM) were selected to determine how yield is affected with increasing glycerol concentration. The purified complex reached 99.5% conversion of glycerol to 1,3-PDO with an initial glycerol concentration of 4.2 mM indicating that the three dockerin-fused metabolic enzymes performed well when self-assembled in the cell on the scaffoldin. The two negative controls, namely the one with empty plasmid and the one with the three wild-type metabolic enzymes did not produce any or very negligible amounts of 1,3-PDO, respectively. The negligible production of 1,3-PDO from the strain encoding wild type enzymes is most likely due to nonspecific binding to the cellulose.

We also followed the kinetics of the conversion of glycerol to 1,3-PDO using the isolated enzyme complex (Supplementary Figure S1). Glycerol conversion reached 50% after 3.8 h and 70% after 8 h indicating the continuous production by the immobilized complex but also revealing that most of the conversion occurred during the first 8 h of incubation. The conversion ultimately reached 99.5% yield after 48 h (Figure 6A). When considering higher levels of glycerol and stoichiometric levels of the cofactor NADH, this pathway was significantly inhibited (Figure 6B). We hypothesized that providing NADH at stoichiometric levels could lead to inhibition of some of the enzymes in the complex. Therefore, we established a cofactor recycling strategy using a formate dehydrogenase and formate as a sacrificial substrate, to recycle NAD⁺ to NADH. Using this approach, the immobilized metabolic complex was clearly not limited by the initial amount of NADH (1 mM), owing to cofactor recycling using formate, and was able to produce 9.2 mM of 1,3-PDO with a yield of 80%. The immobilized complex was also able to reach 66% conversion of glycerol with an initial glycerol concentration of about 40 mM which further demonstrates the viability of our approach when using higher glycerol titers. However, follow up studies are needed to fully evaluate the impact of glycerol concentration and potential substrate inhibition on this system.

Genes used in this study were synthesized by Synbio Technologies (https://www.synbio-tech.com) or cloned from their genomic DNA. Gibson Assembly Cloning Kit (NEB, Ipswich, MA, United States) and other standard gene cloning protocols described previously were used for vector construction (Xu et al., 2017). Primers used in the gene cloning are listed in Supplementary Table S1. 5-alpha E. coli strains (NEB, Ipswich, MA, United States) were used for the gene cloning. pUC57-Kan vector was used for constructing new vectors with various operons related to the 1,3-PDO synthesis pathway. 1,3-PDO synthesis pathway genes were expressed in 5-alpha E. coli strains (NEB, Ipswich, MA, United States) with operons driven by a constitutive promoter from the C. acetobutylicum ATCC 824 thiolase gene. All operons used in this study are listed in Supplementary Table S2. pET vector was used to clone individual genes for their overexpression, and BL21 (DE3) was used for overexpression of individual genes. Cells were grown aerobically in LB supplemented with appropriate antibiotics at 37°C and 225 rpm overnight (seed culture). Twenty milliliters of the seed culture were inoculated into 1 L LB supplemented with appropriate antibiotics, and the cells were grown under the same growth conditions until optical cell density OD₆₀₀ reached 0.8. The culture was placed on ice for 3 h, then 0.3 mM IPTG was added, and finally grew overnight at 16°C and 225 rpm. The cells were harvested by centrifuge 10,000 xg for 10 min, and were further used for protein extraction and purification.

Frozen cells were lyzed by incubating for 30 min at room temperature followed by BeadBeater lysis with glass beads (Bio Spec Products Inc., Bartlesville, OK, United States; according to manufacturer instructions) in 50 mM Tris, pH 7.0, 100 mM NaCl, and 10 mM imidazole containing EDTA-free protease inhibitor (Thermo Scientific, Rockford, IL, United States; according to manufacturer instructions), 1 U/mL Pierce Universal Nuclease (Thermo Scientific, Rockford, IL, United States) and 5 mg/ml lysozyme (Sigma-Aldrich Corp. St. Louis, MO, United States). The lysate suspension was centrifuged at 15,000 g for 30 min and the resulting supernatant was purified using Ni-NTA chromatography (5 ml HisTrap FF crude column; GE Life Sciences, Piscataway, NJ, United States). Purification was completed with size exclusion chromatography using a HiLoad Superdex 75 (26/60) column (GE Life Sciences, Piscataway, NJ, United States) in 20 mM Tris pH 7.5 and 100 mM NaCl. Strep-tag purification was done according to manufacturer’s instructions for Strep-Tactin XT using a Twin-Strep-tag (IBA Lifesciences GmbH, Göttingen, Germany). When necessary (dhaB2) lysis and purification was done in an anaerobic chamber. Iron-sulfur cluster reconstruction was done several times for different dhaB2 constructs expressed, lyzed, and purified anaerobically or aerobically with and without EDTA removal of the iron-sulfur cluster using two separate protocols by O’Brien et al. (2004) (O'Brien et al., 2004) and Freibert et al. (2018) (Freibert et al., 2018).

Inclusion body preparation: Cells were lyzed as above in 50 mM Tris, pH 7.0, 100 mM NaCl, and 0.02% sodium azide, and the suspension was centrifuged at 15,000 xg for 30 min, its pellet was washed under the same buffer and centrifuge conditions for five times. The pellet suspension was used for SDS-PAGE analysis.

E. coli cells with operons were grown aerobically in LB medium in the presence of nitrilotriacetic acid (200 mg/L), FeSO₄.7H₂O (50 mg/L), K₂HPO₄ (0.5 g/L), Na₂SeO₄ (30 μg/L), supplemented with 40 g/L glycerol and appropriate antibiotic at 37°C overnight (seed culture), and then the seed culture was transferred to the same medium and continued to be grown anaerobically in an anaerobic chamber (COY, Grass Lake, MI, United States) filled with 5% hydrogen balanced by nitrogen at 37°C and 150 rpm. 2 ml aliquots were withdrawn at designated time points during the fermentation. Concentration of 1,3-PDO produced in the fermentation was measured by HPLC as described below, and the data was used to prepare curves of 1,3-PDO production vs. fermentation time. All fermentations were performed in duplicate, and the reported values are the average of duplicate assays.