Construction of a heterologous FA biosynthesis pathway
The exogenous genes tal (Genbank No. CP033447.1), sam5 (Genbank No. HE804045.1) and comt (Genbank No. EF413031.1) comprising FA biosynthesis pathway were selected based on previous studies (Berner et al. 2006; Ma and Xu 2008; Watts et al. 2004). Jendresen et al. reported that highly efficient tyrosine ammonia-lyases (TALs) from diverse origins enable enhanced production of aromatic compounds in bacteria and Saccharomyces cerevisiae (Jendresen et al. 2015). The tal gene (Genbank No. CP033447.1) performed well in our previous study on the de novo biosynthesis of resveratrol in E. coli (Wang et al. 2014), and the expression of the FA biosynthesis pathway using BL21(DE3) as the host cell resulted in p-coumaric acid accumulation (Fig. 2A). Rodrigues et al. compared p-coumarate 3-hydroxylase from Saccharothrix espanaensis (sam5) and Rhodopseudomonas palustris (CYP199A2) for the conversion of p-coumaric acid produced from tyrosine into caffeic acid in E. coli. The CYP199A2 enzyme was more active, but it requires two redox partners, which may complicate its heterologous expression (Haslinger and Prather 2020; Rodrigues et al. 2015a). COMT from A. thaliana (GenBank No. AY062837) was previously used to catalyze caffeic acid to FA in several reports (Choi et al. 2011; Heo et al. 2017). Here we tested COMT from wheat (Triticum aestivum L. cv. H4564, GenBank No. EF413031.1); this enzyme was named TaCM in original literature, and was proposed to methylate phenol substrates containing aldehyde, flavonoid and CoA moieties. Thus it may have a broad substrate preferences, and can not only involves in converting caffeic acid to ferulic acid and 5-hydroxyferulic acid to sinapic acid, but also in the conversion of caffeoyl-CoA to feruloyl-CoA, and 5-hydroxyferuloyl-CoA to sinapoyl-CoA (Ma and Xu 2008). Notably the protein sequence of TaCM has 100% identity with TaOMT2 (Genbank No. DQ223971), which showed the highest activity specifically to tricetin in substrate preference test against a number of phenolic compounds, and was proposed to accept caffeic acid and 5-hydoxy-ferulic acid as substrates (Zhou et al. 2006, 2009). Here, we used TaCM in the FA biosynthesis pathway and confirmed its ability to convert caffeic acid into FA. To our best knowledge, this is the first report of the caffeic acid O-methyltransferase activity of TaCM/TaOMT2.
E. coli BL21(DE3) is generally a more effective expression host than JM109(DE3), and the latter is a high acetate producer (Shiloach et al. 1996), while acetate accumulation can reduce the growth rate and recombinant protein synthesis (Noronha et al. 2000). In a study by Kang et al. E. coli C41(DE3) [mutant derivative of BL21(DE3)] was engineered as a tyrosine over-producing chassis for the biosynthesis of phenylpropanoic acid (Kang et al. 2012). Similarly, Huang et al. used E. coli BW25113 (mutant derivative of E. coli K-12) for the biosynthesis of caffeic acid (Huang et al. 2013; Lin and Yan 2012). In this study, the heterologous biosynthetic pathway under the control of the strong T7 promoter was tested in E. coli JM109(DE3) and BL21(DE3), and the results showed that it could convert L-tyrosine into FA, with a small amount of residual caffeic or p-coumaric acid (Fig. 2a). Interestingly, the FA yields were higher when using JM109(DE3) as the host strain than with BL21(DE3), so JM109(DE3) was selected as heterologous expression host cell for further experiments.
To increase the FA titer, we firstly attempted to improve soluble protein expression using the T7 promoter by decreasing the temperature form 37–22 °C. SDS-PAGE analysis showed that TAL and SAM5 were mostly expressed in the form of inclusion bodies, and their solubility was slightly improved at the lower temperature (Additional file 1: Figure S1A–C). We next replaced the strong T7 promoter with the weaker T5 promoter, and protein expression was changed greatly. Although the total protein expression level decreased, the solubility was improved, especially for TAL and SAM5. Based on this, we replaced the T7 promoter in the FA expression cassette with the T5 promoter, and the FA titer in M9Y medium increased to 130 mg/L (Fig. 2C). This result suggests that the FA yield could be optimized by tuning the gene expression to an appropriate level.
Pathway optimization by copy number and promoter strength tuning
Using different combinations of plasmid replicons and promoter strengths is convenient to tune the expression levels of pathway genes. Modulating diverse expression level of up- and downstream pathways of taxadiene synthesis resulted in a remarkable 15,000-fold increase of taxadiene titer in an engineered E. coli strain. Here our results of different plasmid replicon (p15a, about 10 copies; pBR322, about 20 copies) and promoter strength (T7, the relative strength is 4.97; T5, the relative strength is 2) combinations showed strain p15a-T5 yield the highest FA titer, 180.5 ± 10 mg/L FA, 5 ± 0.1 mg/L caffeic acid mg/L. The following two were strain pBR322-T5 and pBR322-T7, and the production of strain p15a-T7 was the lowest. The order of the corresponding replicon copies and promoter strength combinations is 10–2 > 20–2 > 20–4.97 > 10–4.97, FA titer decline in turn. The numbers refers to replicon copies and promoter strength in Ajikumar’s work (Ajikumar et al. 2010). More work is needed for even refined modulation, like screening for combinations of promoters and terminators, modular controlling via different promoters, using RBS libraries that covered a broad range of translational initiation rates, and so on (Zhang and Hong 2020), and it is becoming a universal approach to achieve high yield for heterologous biosynthesis.
It is worth noting that, production of strain pBR322-T7 (118.8 ± 11 mg/L FA, 14.4 ± 1.0 mg/L caffeic acid) is higher than that previous fermentation carried in TB, indicating culture medium composition may affect the yield.
Improving FA production by NADPH regeneration enzyme
In the FA biosynthetic pathway, p-coumarate 3-hydroxylase (SAM5) consumes NADPH. Thus, the overexpression of heterologous pathway can greatly affect the intracellular redox homeostasis. The overexpression or introduction of a heterologous cofactor regeneration system is an important strategy for engineering redox homeostasis, and this approach has been successfully used for the biosynthesis of many NAD(P)-dependent products (Liu et al. 2018; Zhao et al. 2017). Here, we tested five NADPH regenerating enzymes that were proved to increase the NADPH/NADP+ ratio in E. coli (Huang et al. 2017), covering almost all NADPH regeneration systems for E. coli. The glucose-6-phosphate dehydrogenase ZWF and 6-phosphogluconate dehydrogenase GND are involved in the PP Pathway. Similarly, the glyceraldehyde-3-phosphate dehydrogenase GapN from Clostridium acetobutylicum ATCC 824 is involved in the glycolysis pathway, and the isocitrate dehydrogenase ICD participates in the TCA cycle. By contrast, the transhydrogenase PntAB catalyzes the transfer of reducing power from NADH to NADP+ (Fig. 1C). The strain co-expressing the pntAB genes with the FA biosynthesis pathway (T5FA + pntAB) exhibited a significant increase of FA production (191.9 ± 9.6 mg/L; Fig. 4). However, the other NADPH regenerating enzymes showed no effect on the FA titer. Similarly, pntAB overexpression to mitigate the redox imbalance significantly increased the product titer and yield in the biosynthesis of shikimic and glycolic acid in E. coli (Cabulong et al. 2019; Cui et al. 2014), as well as L-lysine, acetic and succinic acid in Corynebacterium glutamicum (Kabus et al. 2007; Yamauchi et al. 2014). Sauer et al. found that PntAB produced 35–45% of the NADPH consumed during the exponential batch growth phase on glucose, while the pentose phosphate pathway and isocitrate dehydrogenase contributed 35–45% and 20–25%, respectively (Sauer et al. 2004). Based on this, PntAB contributed the most among the five tested enzymes, which may explain why the strain co-expressing the pntAB genes with the FA biosynthesis pathway showed a significant increase of the product titer, which was not observed with the other tested genes. However, glycerol was used as the carbon source in this work, and the effect may be different if glucose is used in the case of the other NADPH regenerating enzymes, especially zwf and gnd, which are involved in the pentose phosphate pathway.
We noted that the FA production of the control strain E. coli T5FA + pCL1920-T7 (140 mg/L FA and 30 mg/L caffeic acid, Fig. 4) was slightly higher than that of E. coli T5FA, which carries only FA biosynthesis pathway plasmid(130 mg/L FA and 33 mg/L caffeic acid, Fig. 2C). The product titers of the strains expressing the other candidate genes (icd, zwf, gnd, gapN) were also lower than that of the control strain. As an amino glycoside antibiotic, spectinomycin binds to the bacterial 30S ribosomal subunit and blocks protein synthesis. In the strain carrying the pCL1920 plasmid, the protein synthesis maybe affected due to the addition of spectinomycin to the culture medium, which may reduce the burden of protein synthesis, and improve the yield of FA. Our results also showed that the strain co-transfected with a plasmid carrying a spectinomycin resistance gene exhibited better growth and FA production than the strains without this plasmid.
Improving FA production by non-native SAM synthetase
The transformation of caffeic acid into FA consumes S-adenosylmethionine (SAM) at the same time (Fig. 1). Similar as redox cofactor NADPH, SAM is a ubiquitous intracellular methyl or methylate donor, previous work reported improving heterologous polyketide production in E. coli by overexpression of an non-native S-adenosylmethionine synthetase (SsmetK) gene (Wang et al. 2007). Increased intracellular SAM availability is beneficial to increase FA production in theory. Endogenous S-adenosylmethionine synthetase gene metK (SAMs, catalyzing the reaction of ATP and L-Methionine to form SAM, Fig. 1C) in E. coli is inhibited by L-methionine (Zocchi et al. 2003), for this reason, we chose the non-native metK from Streptomyces spectabilis and tested in our heterologous biosynthesis pathway.
The FA production of strain p15aT5 + metK, strain p15aT5 + gntAB + metK, and strain p15aT5 + gntAB is significantly higher than the control strain p15aT5 + pCL1920-T7, the highest one is p15aT5 + gntAB + metK, producing 212.5 ± 11.4 mg/L FA with 11.9 ± 0.5 mg/L caffeic acid residue. Caffeic acid accumulation was lowest in strain p15aT5 + metK among these strains, which may attribute to supplement of SAM. Compared with strain T5FA + pntAB (191.9 ± 9.6 mg/L FA and 30 ± 10 mg/L caffeic acid), strain p15aT5 + gntAB (207.4 ± 2.9 mg/L FA and 13 ± 0.7 mg/L caffeic acid) produced more FA and left less caffeic acid residual, which is consistent with the expression level optimization objective. While there is no significant difference of FA production between these three strains (Fig. 5), which means that the effect of gene metK on FA production is almost equal with gene pntAB. The strain p15aT5 + pntAB + metK showed no increase of the FA production further, not as expected a possible superimposed promotion effect. We also tried adding L-methionine in culture medium directly, but resulted in decreased FA production (Additional file 1: Table S4). Kunjapur et al. reported deletion of metJ coupled with expression of feedback-desensitized variants of metA* and cysE*, genes involved in methionine biosynthesis, improved de novo vanillate titers by 33% in an engineered E. coli K-12 MG1655 strain RARE that serves as a platform for aromatic aldehyde biosynthesis. While overexpression of mtn and luxS, genes involved in S-adenosylhomocysteine (SAH) recycling improved de novo vanillate titers by 25%, it is supposed to work by the mechanism of increasing SAM availability, since SAH is a potent inhibiter of SAM-dependent methyltransferases. And vanillate production improved further upon supplementation with methionine (Kunjapur et al. 2016). So, it seems that methylation improvement in biosynthetic pathway in E. coli could be achieved by methods of improving SAM availability like introducing heterologous methionine adenosyltransferase (SsmetK), engineering methionine biosynthesis related genes (metJ, metA* and cysE*) and SAH recycling related genes (mtn and luxS). As pointed out in Kunjapur et al.’s work, adding methionine to E. coli decreases flux in the Met and SAM biosynthetic pathway due to feedback inhibition of MetA. The effect of supplying methionine exogenously may vary from case to case. SAM availability in FA biosynthesis may improve and gain increased FA titer by combining all these strategies in future work. So, the highest FA titer in our work is 212 mg/L, although the FA titer of Kang et al 2012 is 196 mg/L in 36 h, and our data are 212 mg/L in 5 days, it seems that the productivity of the former is better, but the carbon supplement of the former is much higher and the host cell is BL21(DE3) (Additional file 1: Table S1). In parallel with our work, Rodrigues et al. reported even higher FA production (257 mg/L) in their work of curcuminoids biosynthesis in E. coli BL21(DE3) recently. They constructed TAL from Rhodotorula glutinis and C3H/SAM5 from S. espanaensis in pCDF-duet backbone (CloDF13 ori, PT7lac), and COMT from A. thaliana in pRSF-duet backbone (RSF1030 ori, PT7lac) simultaneously. Except for different source of TAL and COMT with our work, BL21(DE3) was chosen as host cell and two plasmids were used to construct FA biosynthesis pathway in their work (Rodrigues et al. 2020). The genes they chose were reported as most the efficient ones, while gene tal and gene comt were not the same version as those used in our work, and FA biosynthesis pathway was constructed in a combination of two plasmids, pCDF-duet_TAL (20–40 copies, PT7lac) and pRSF-duet_C3H_COMT (10 copies, PT7lac). Both the plasmid copy number and promoter strength are higher than the ones used in our work, but reached a balance in their work, although introduction of multiple plasmids is inconvenient for follow-up operations, indicating FA production could be further improved by combined application of two strategies.
In this study, we constructed a heterologous FA biosynthesis pathway by introducing genes from three different species, and the caffeic acid O-methyltransferase activity of TaCM was shown for the first time. FA production was tested in two common E. coli strains, and using JM109(DE3) as the host strain resulted in a higher FA yield than using BL21(DE3). Tuning down the promoter strength of the expression cassette significantly increased the final FA titer, so we continued to optimize the pathway expressing level using combinations of different replicons and promoter strengths, which further increased FA production. To further increase the production, the endogenous NADPH regeneration genes pntAB and the heterologous SAM formation enzyme SsmetK were used to improve the supply of the key cofactors NADPH and SAM, which greatly increased the FA titer from 130 to 212 mg/L. Further studies are needed to test for synergistic effects of simultaneous NADPH and SAM supplementation, and relevant methods such as increasing methionine biosynthesis and SAH recycling to improve SAM availability can be adopted. Strategies that can more finely modulate the flux of pathway metabolites, such as splitting the FA biosynthesis pathway into two modules and regulating each module separately, may unlock greater productivity of the host cells. Based on all these efforts, the widely used heterologous expression host BL21(DE3) with glucose as carbon source may still show good productivity in the heterologous biosynthesis of FA. A tyrosine over-producing strain or chromosomal integration of the biosynthesis pathway can also tested to reduce the fermentation cost. This study lacks fed-batch fermentation data, but the shake-flask results were comparable with the literature, indicating good application prospects. Our work offers a basis for further studies on the heterologous biosynthesis of phenolic natural products.