Choosing a promoter to drive pforA expression
Thermoanaerobacterium saccharolyticum pforA and ferredoxin were integrated into the C. thermocellum genome in strain LL1319, at a location immediately downstream of Clo1313_2637, and before the putative promoter for the Clo1313_2635 gene (Additional file 1: Figure S1). This was done so as to locate the new operon close to the previously introduced T. saccharolyticum adhA-nfnAB-adhEG544D operon [4], without disrupting a putative peroxiredoxin two-gene cluster (Clo1313_2638-2637). Three promoters were tested to drive expression of the pforA–ferredoxin operon: the previously described C. thermocellum enolase promoter [31], the T. saccharolyticum pforA promoter, and the Athe_2105 promoter from Caldicellulosiruptor bescii [19] with a modified ribosome binding site [24]. Promoter sequences and predicted translation initiation efficiencies of pforA are reported in Additional file 1: Table S2.
T. saccharolyticum pforA and ferredoxin expression was observed in all three operon configurations (Fig. 2a), but not in wild-type C. thermocellum and the parent strain, LL1319, as expected. In general, we observed that ferredoxin expression was lower than that of the pforA gene. It was also observed that the C. thermocellum enolase promoter resulted in the highest level of gene expression, followed by the T. saccharolyticum pforA promoter, with the modified Athe_2105 promoter giving the lowest level of expression. However, we also observed that expression was more variable with the C. thermocellum enolase promoter than with the two heterologous promoters. The variation observed with the native enolase promoter is similar to what we have previously observed for lacZ expression [31].
Fermentation product profiles for the C. thermocellum strains using the native enolase and T. saccharolyticum pforA promoters were indistinguishable from the parent strain LL1319 on 20 g/L initial cellobiose. The strain that used the C. bescii Athe_2105 promoter, however, showed an unexpected decrease in ethanol production (Fig. 2b). On 52 g/L initial cellobiose, we again observed that the Athe_2105 promoter-containing strain exhibited reduced ethanol production compared to the parent strain LL1319 (Fig. 2c), unlike the other two strains that contained the T. saccharolyticum pforA and ferredoxin; this strain was thus excluded from further investigations. Between the two other strains, both showed comparable improvements to ethanol production over the parent strain LL1319, with the T. saccharolyticum pforA promoter-driven strain showing slightly higher ethanol titers. Given that the T. saccharolyticum pforA promoter resulted in the highest levels of ethanol production of the three promoters tested, and that its use avoids duplicating native DNA sequences (which can lead to unintended recombination events and complicate the analysis of resequencing data), we proceeded forward with this strain and designated it as strain LL1391 (Table 2). Subsequent fermentations confirmed that strain LL1391 produced more ethanol than strain LL1319 (Fig. 3b), which may be attributed to a significant increase in the BV:PFOR specific enzyme activity in strain LL1391 (unpaired two-tailed t test, p = 0.0045) (Fig. 3a).
The effects of T. saccharolyticum pforA and ferredoxin expression on ethanol production
Strain LL1319 contains four T. saccharolyticum genes from previous strain development [4]; we therefore investigated whether the improvements in ethanol production were dependent on the presence of the previously introduced T. saccharolyticum adhA, nfnAB, and adhEG544D genes. Strain LL1565 was created by integrating T. saccharolyticum pforA and ferredoxin, driven by the T. saccharolyticum pforA promoter (plasmid pSH106, Table 2) into strain AG929 (the parent strain of strain LL1319). A significant increase in BV:PFOR specific activity in strain LL1565 was observed, relative to strain AG929 (unpaired two-tailed t test, p = 0.0064); the measured specific activity of strain LL1565 was comparable to activity levels measured in strain LL1391 (Fig. 3a), indicating that the introduced T. saccharolyticum PforA protein was present and active in the strain. Ethanol production with strain LL1565 appeared to be slightly improved relative to parent strain AG929, although the metabolic yield and final titers were significantly lower than that of strain LL1319 and by extension that of LL1391 (Fig. 3b), suggesting that the improvement in ethanol production from introducing pforA is dependent on the presence of the other T. saccharolyticum ethanol production pathway genes.
To determine whether the T. saccharolyticum ferredoxin was necessary for the improvements in ethanol production, T. saccharolyticum pforA alone was integrated into strain LL1319 at the same locus as was done in strain LL1391 to create strain LL1566 (using plasmid pSH121, see Table 2). The BV:PFOR specific enzyme activity for strain LL1566 was no different from that observed in LL1391, as expected; fermentation products for the two strains were also similar, suggesting that the introduced T. saccharolyticum PforA protein was responsible for the improvements in ethanol production (Fig. 3). We attempted to introduce the T. saccharolyticum ferredoxin on its own into strain LL1319, but were not successful. Given that there appeared to be no detrimental effects in ethanol production due to the presence of T. saccharolyticum ferredoxin (Fig. 3), and that the ferredoxin is important in the production of ethanol as an electron carrier, we decided to retain it in subsequent strains (see Table 2 for strain lineage).
The effect of deleting native pfors on ethanol production
The introduction and expression of T. saccharolyticum pforA have thus far been associated with an increase in BV:PFOR specific activities, and an increase in ethanol titer and metabolic yield (Fig. 3). However, the strains evaluated so far still contain the five native Pfor-encoding genes and gene clusters (Table 1). To better determine whether the improvements in ethanol production were due to the introduced T. saccharolyticum PforA protein, we deleted all five C. thermocellum pfor gene clusters in an iterative manner.
Previous work suggested that pfor1 (Clo1313_0020-0023) and pfor4 (Clo1313_1353-1356) encoded for the main Pfor protein complexes in C. thermocellum [10]. Further support for pfor1 and pfor4 encoding for important Pfor complexes in C. thermocellum was found when it was observed that BV:PFOR specific activity decreased by ~ 80% relative to wild-type C. thermocellum (Additional file 1: Figure S2) when either pfor1 or pfor4 was deleted in C. thermocellum (strain LL1556 and LL1564). Strains containing a deletion of either pfor2 or pfor5 did not show any significant differences in BV:PFOR specific activity, suggesting that they were not important for PFOR activity in C. thermocellum, or that pyruvate was not the primary substrate for these enzymes. The deletion of pfor3, which bears the most similarity to the T. saccharolyticum pforA, also resulted in ~ 40% decrease in specific BV:PFOR activity. Given these observations, pfor1 and pfor4 were therefore the first targets for gene deletion in strain LL1391.
Starting with strain LL1391 (wt strain expressing T. saccharolyticum pathway, including pforA), deletion of pfor1 (strain LL1436) did not result in any significant effects on ethanol production or enzyme specific activity. Deletion of pfor4 (strain LL1437), however, showed a decrease in ethanol yield and a large decrease in titer (Fig. 3b), despite very little change in BV:PFOR activity (LL1391 vs. LL1437, unpaired two-tailed t test, p = 0.23) (Fig. 3a).
Resequencing analyses subsequently revealed that LL1437 contained a 1207G > T mutation in the coding sequence for the Clo1313_1483 gene that resulted in a G403* nonsense mutation in the amino acid sequence; excluding the targeted pfor deletions, there were no other differences between the genomes of the two strains. Clo1313_1483 is annotated as encoding a predicted pyrroloquinoline quinone-associated protein, and previous work suggests that it is expressed in C. thermocellum strain ATCC27405 [32, 33]; however, its function in C. thermocellum strain DSM1313 is unknown.
To compare the effect of the Clo1313_1483 mutation vs. pfor4 deletion, we constructed two pfor1/pfor4 double deletion strains: LL1438 (LL1436 with pfor4 deleted) and LL1567 (LL1437 with pfor1 deleted). Since neither of these strains showed any significant difference in ethanol production relative to their respective parent strains (LL1438 vs. LL1436 and LL1567 vs. LL1437), the difference in ethanol production between LL1436 and LL1437 is likely due to the Clo1313_1483 mutation, and not the effect of the pfor4 deletion.
The double pfor1/pfor4 deletion strain, LL1438, was still able to sustain the improved ethanol production observed in strain LL1391. To eliminate the possibility that this was due to the remaining three C. thermocellum Pfor enzymes compensating for the pfor1 and pfor4 deletions, the genes encoding for pfor3, pfor2, and pfor5 (see Table 1 for gene numbers) were iteratively deleted to create strain LL1570. Strain LL1570 was able to produce 424 ± 13 mM (~ 20 g/L) of ethanol from 50 g/L of cellobiose, with a metabolic ethanol yield of 80% of the theoretical maximum (Fig. 4b; also see Additional file 1: Table S3), an improvement over the reference strain LL1319, which was previously reported to have achieved a maximum ethanol yield of 74% of theoretical maximum on 20 g/L cellobiose, and a maximum ethanol titer of 326 mM (~ 15 g/L) on 60 g/L Avicel (120). The maximum specific growth rate of strain LL1570 on cellobiose was unaffected relative to the starting strain, LL1319 (Additional file 1: Table S3), suggesting that deleting the native pfor genes did not result in any growth defects, and that the introduced T. saccharolyticum pforA complemented the deletions of these five native C. thermocellum pfor genes.
In strain LL1570, we would have expected that deletion of pforA would eliminate acetate and ethanol production and divert flux to lactate production (similar to what was observed for pfor deletions in T. saccharolyticum [11]). Despite several attempts to delete pforA in this strain, we were not successful. Although not conclusive, this negative result suggests that PFOR activity is essential in this strain, and that pforA is the source of that activity.
Fermentation of high cellulose concentrations
To determine if replacing of the native pfors with the T. saccharolyticum pforA and ferredoxin had improved the maximum ethanol titer and maximum volumetric production rate of ethanol, and to evaluate the performance of the strain on a cellulosic substrate, strains LL1319 and LL1570 were grown on 100 g/L Avicel microcrystalline cellulose.
Strains LL1319 and LL1570 consumed about the same amount of Avicel (Figs. 3b and 4a, see also Additional file 1: Table S5), but produced different amounts of ethanol. Strain LL1319 produced ethanol to a titer of 486 ± 5 mM (~ 22 ± 0.2 g/L) (Fig. 4c), for a metabolic yield of 45% of the theoretical maximum; the ethanol titer observed here was higher than previously reported [4], although it should be noted that the media composition was different. In contrast, strain LL1570 produced 551 ± 32 mM (25 ± 1.5 g/L) of ethanol (Fig. 4d, Additional file 1: Figure S3), for a metabolic yield of 54% of theoretical maximum. The results provide further evidence that T. saccharolyticum pforA improved ethanol yield and titer (unpaired two-tailed t test; p = 0.003 for ethanol yield, p = 0.02 for ethanol titer). The differences in volumetric productivity of ethanol between strains LL1319 and LL1570 (0.66 ± 0.03 g L−1 h−1 and 0.70 ± 0.12 g L−1 h−1, respectively) were not statistically significant (unpaired two-tailed t test; p = 0.338) (Additional file 1: Table S5). The ethanol titer of 25 g/L for strain LL1570 was very similar to that produced by another engineered strain of C. thermocellum, LL1210 (Δhpt ΔhydG Δldh Δpfl Δpta-ack adhE(D494G)), which was generated by eliminating the native competing carbon and electron pathways, followed by strain adaptation over ~ 2500 generations to increase growth rate, and which was reported to produce 27 g/L of ethanol from 95 g/L of Avicel [34]. The byproduct concentrations (organic acids and total extracellular amino acids) in both strains LL1319 and LL1570 were similar, except for isobutanol production, which decreased to below our limit of quantification (0.1 mM) in strain LL1570 (Fig. 4d) (strain LL1319 produced a maximum isobutanol titer of ~ 14 mM; see Fig. 4c); this suggests that one of the deleted C. thermocellum pfors may be involved in the biosynthesis of isobutanol. Fermentation results from a set of C. thermocellum strains that contain a deletion of one of the five annotated pfors suggest that it is pfor4 that is associated with isobutanol production (Additional file 1: Figure S4).
Discussion and conclusion
In this work, we investigated the effects of T. saccharolyticum pforA and ferredoxin on ethanol production in C. thermocellum. There was no effect from expressing T. saccharolyticum ferredoxin in C. thermocellum. It is known that ferredoxins from one organism can often transfer electrons to proteins from another organism [35], so it would not be surprising if one of the native C. thermocellum ferredoxins was sufficient for electron transfer from T. saccharolyticum Pfor protein.
Introducing just the T. saccharolyticum pforA did not improve ethanol titer, but it slightly shifted the ethanol to acetate ratio in favor of ethanol production. When the T. saccharolyticum pforA was expressed alongside the previously introduced T. saccharolyticum adhA, nfnAB, and adhEG544D genes, ethanol production improved. These observations support the hypothesis that pforA is an important component of the T. saccharolyticum pyruvate-to-ethanol pathway. Furthermore, the pforA from T. saccharolyticum was able to functionally complement the deletion of the five annotated C. thermocellum pfor genes.
Isobutanol production in strain LL1570 was reduced below our limit of quantification (0.1 mM). Fermentation data from a set of C. thermocellum strains with single deletions of each of the five annotated pfors points to pfor4 being responsible for isobutanol production. Given the interest in producing isobutanol either for use as a biofuel or as a feedstock chemical [36]; the knowledge that pfor4 is necessary for isobutanol production could be beneficial to further improve its production from cellulosic substrates.
The ability to use a T. saccharolyticum promoter to drive the expression of the pforA-ferredoxin operon is relevant to future work involving gene expression in C. thermocellum. Whereas native C. thermocellum promoters have been characterized [31], with the enolase promoter being used successfully in this study to express T. saccharolyticum pforA, these promoters may still be subject to native transcriptional regulation, and therefore may not be suitable if constitutive gene expression is desired. It should be noted that other examples of heterologous promoters being used in C. thermocellum have been reported [19]. As heterologous gene expression becomes more prevalent in C. thermocellum [4, 22, 37], it will become increasingly necessary to develop libraries of non-native or synthetic genetic tools to avoid excessive duplicating of native DNA elements, which could contribute to genome instability.
Having now introduced six genes from the T. saccharolyticum pyruvate to ethanol pathway into C. thermocellum, we observe that there is still a ~ 40 g/L difference between the maximum ethanol titers achieved by engineered C. thermocellum (25–30 g/L) [34] and those achieved by engineered T. saccharolyticum (60–70 g/L) [5], suggesting that there remains more work to be done. With regard to the Pfor-catalyzed conversion of pyruvate to acetyl-CoA, one consideration is that the Pfor enzyme functions in tandem not only with ferredoxin, but also with ferredoxin:NAD(P)+ oxidoreductase (Fnor). It is possible that titer limitations in engineered C. thermocellum are due to an un-optimized Pfor–ferredoxin–Fnor module. Some promising directions for further research include characterizing the Pfor–ferredoxin–Fnor modules of C. thermocellum and T. saccharolyticum in more detail, or engineering the module with better enzymes or through protein engineering of the existing enzymes to overcome possible substrate or cofactor inhibitions [38, 39].