Caffeic acid, also known as 3,4-dihydroxy cinnamic acid, has attracted increasing attention due to its antioxidant , antivirus , anticancer [3, 4] and anti-inflammatory biological properties . Moreover, caffeic acid is an important precursor of plant-originated aromatic chemicals like rosmarinic acid, chlorogenic acid and caffeic acid phenethyl ester [6–8]. Therefore, it shows great potential in nutritional, pharmaceutical and cosmetics industries . Considering the environmental and economic benefits, biosynthesis of caffeic acid via engineering model microbes such as Escherichia coli and Saccharomyces cerevisiae provides a promising alternative to chemical synthesis or plants extraction .
The biosynthesis of caffeic acid starts from L-phenylalanine or L-tyrosine through the endogenous shikimate pathway . In plant, the deamination of L-phenylalanine is catalyzed by phenylalanine ammonia lyase (PAL) to produce cinnamic acid. The sequential two-step hydroxylation at the 4- and 3- positions of the benzyl ring of cinnamic acid is executed by two cytochrome P450 monooxygenases, cinnamate-4-hydroxylase (C4H) and p-coumarate 3-hydroxylase (Coum3H) , forming caffeic acid through p-coumaric acid. In recent years, reports have sporadically emerged regarding metabolic engineering for heterogeneous caffeic acid production in E. coli. However, the plant-originated P450 enzymes are difficult to express in microbial systems . Alternatively, tyrosine containing a 4-hydroxyl group could be directly converted to p-coumaric acid by microbial tyrosine ammonia lyase (TAL) . For further hydroxylation of p-coumaric acid, the sam5-encoded Coum3H from the actinomycete Saccharothrix espanaensis [15, 16] or the cytochrome P450 CYP199A2 from the bacteria Rhodopseudomonas palustris [17, 18] could be used, enabling caffeic acid formation in E. coli. By introducing RgTAL from Rhodotorula glutinis into E. coli together with expressing endogenous 4-hydroxyphenylacetate 3-hydroxylase (4HPA3H) and increasing the intracellular supply of tyrosine by overexpression of PEP synthase (encoded by ppsA) and transketolase (encoded by tktA) and feedback inhibition resistant 3-deoxy-D-arabino-heptulosonate-7-phosphate synthase (encoded by aroGfbr) and chorismate mutase-prephenate dehydrogenase (encoded by tyrAfbr), the highest caffeic acid production reached 766.7 mg/L from simple carbon sources in shake flasks . However, the cell growth and caffeic acid production still relied on phenylalanine supplement.
S. cerevisiae as GRAS (generally regarded as safe) organism with well-characterized genetic background, superior stress tolerance and excellent fermentation properties becomes an attractive microbial host for caffeic acid production. However, neither Coum3H nor CYP199A2 could enable caffeic acid biosynthesis in yeast . The bacterial 4-hydroxyphenylacetate 3-hydroxylases (4HPA3H) complex encoded by HpaB and HpaC was also found to effectively catalyze p-coumaric acid to caffeic acid (Fig. 1) . Expression of HpaB and HpaC from E. coli in S. cerevisiae led to caffeic acid production , and replacement of the E. coli enzymes with the combination of HpaB from Pseudomonas aeruginosa and HpaC from Salmonella enterica significantly improved the caffeic acid yield by 45.9-fold, leading to the highest production of caffeic acid (about 289.4 mg/L) in yeast . However, this process still relied on feeding of exogenous L-tyrosine as the precursor. Li Y et al reported that simultaneous expression of RcTAL from Rhodobacter capsulatus and the P450-dependent monooxygenase C3H together with its associated cytochrome P450 reductase CPR1 from Arabidopsis thaliana could enable de novo biosynthesis of caffeic acid from glucose in S. cerevisiae. However, low caffeic acid production (about 11.432 mg/L) was obtained, ascribed to the low activity of C3H . In both studies, episomal vectors were used for expression of caffeic acid pathway genes in S. cerevisiae. Considering that the yeast transformants harboring several plasmids are genetically unstable, integrating the caffeic acid pathway into the yeast genome may create a more stable cell factory.
As found with other tyrosine-derived products [21–23], the shortage of precursor supply may be another limiting factor of caffeic acid biosynthesis in S. cerevisiae. The critical step of shikimate pathway is the condensation of two starter units named phosphoenolpyruvate (PEP) and 4-erythritol phosphate (E4P) by isoenzymes Aro3 and Aro4 to produce 3-deoxy-D-arabino-heptulosonate-7-phosphate (DAHP) [23, 24]. In addition, chorismite, the last common intermediate for three aromatic amino acids, is transformed by the chorismate mutase Aro7 to generate prephenate, which is further divided into two branches, one towards L-phenylalanine and the other towards L-tyrosine (Fig. 1). In this pathway, the activity of Aro4 and Aro7 are feedback-inhibited by the end product tyrosine. Feedback-insensitive variant Aro4K229L and Aro7G141S have been created by rational design [25–27]. Either individual overexpression Aro4K229L or simultaneous expression of Aro4K229L and Aro7G141S could effectively increase intracellular shikimate, phenylalanine and tyrosine concentrations [24, 27]. On the other hand, the prephenate dehydrogenase (Tyr1) which catalyzes prephenate to α-keto acid 4-hydroxyphenylpyruvate (4HPP), the direct precursor of L-tyrosine, is transcriptionally inhibited by phenylalanine. Replacement of the native Tyr1 promoter with a constitutive one or expression of the feedback-insensitive cyclohexadienyl dehydrogenase TyrC from Zymomonas mobilis could both improve the production of tyrosine and its derivatives [28, 29]. Meanwhile, decarboxylation of 4HPP catalyzed by phenylpyruvate decarboxylases (encoded by Aro10, Pdc1, Pdc5, and Pdc6) would decrease the flux towards tyrosine, among which Pdc5 and Aro10 showed stronger decarboxylation activity than the others . Deletion of Aro10 and Pdc5 increased the intracellular tyrosine production by 5.7 folds  .
Taken together, alleviation of feedback inhibition and removal of competitive branches could improve the precursor supply and thus contribute to further enhancement of caffeic acid synthesis. In this study, we first constructed a controllable caffeic acid biosynthetic pathway in S. cerevisiae by employing a modified GAL system. The precursor supply was then strengthened by eliminating the feedback inhibition of aromatic amino acid and down-regulating the competitive pathways, and the rate-limiting enzyme TyrC was overexpressed to enhance the flux towards caffeic acid production.