*3.1. Easy Colorimetric Screening of Production of Carotenoids*

Carotenoids contain chromophores absorbing visible light and appear as being yellow (e.g., Ά-carotene) to red (e.g., lycopene), which benefits carotenogenic gene mining and engineering upon carotenoid synthesis pathway. To date, many carotenoid biosynthetic genes have been cloned from plants, bacteria, and fungi based on their abilities to render different colors to the host [68–70]. This merit has been vigorously implemented for random mutagenesis, directed evolution, and proofof-principle experiments in synthetic biology. Moreover, cellular carotenoids can be easily extracted into an organic solvent and differentiated in a sensitivity of submilligrams per liter with a linear correlation between carotenoid contents and color intensity [71,72]. This provides an easy and high-throughput way to evaluate the performance of newly built synthetic circuits or methodologies for improved biosynthesis of carotenoid (Figure 5A). 

**Figure 5.** Technologies and efforts to improve carotenoid production. ( **A**) Colorimetric screening of desired producer; ( **B**) Pathway engineering approaches for production improvement; ( **C**) Genetic modifications for host strain development; ( **D**) Protein engineering for enzyme and pathway improvement. 

*Mar. Drugs* **2014**, *12*, 4810–4832

*3.2. Pathway Engineering for Production of Carotenoids* 

Carotenoid biosynthesis emerges from the central isoprenoid pathway, either the MEP pathway or the MVA pathway, existing in all organisms [19,22]. The expression of carotenogenic genes can yield carotenoids of interest in a heterologous organism [16,73–75]. The early attempts led to the production of lycopene, Ά-carotene, and astaxanthin in *Saccharomyces cerevisiae* and *Candida utilis* by the expression of carotenogenic enzymes from *Pantoea ananatis* [74,76]. *Corynebacterium glutamicum* is a native producer of decaprenoxanthin and its glucosides, and it has been engineered to synthesize C50 carotenoids C.P.450 and sarcinaxanthin [77]. To date, there have been many exemplary illuminations to achieve high carotenoid titers from non-native producers. Carotenogenic enzymes from different sources exhibit different capacities in carotenoid biosynthesis. A two-fold higher lycopene production is obtained in *E*. *coli* by the expression of carotenogenic enzymes from *<sup>P</sup>*. *agglomerans* (27 mg/L) than from *<sup>P</sup>*. *ananatis* (12 mg/L) [78]. Metabolic engineering approaches allow the assembly of genes from different organisms for production purposes or for building new carotenoids [32,79,80]. Ά-Carotene production has been improved by hybrid expression of carotenogenic genes from *<sup>P</sup>*. *agglomeras* and *P*. *ananatis* in *E*. *coli* [81]. In another example, expression of Ά-end ketolase from *Agrobacterium aurantiacum* extends the zeaxanthin Ά-D-diglucoside pathway from *<sup>P</sup>*. *ananatis*, and synthesizes novel astaxanthin Ά-D-diglucoside and adonixanthin Ά-

D-diglucoside [29]. Generally, a sufficient precursor supply is a prerequisite for highyield production of carotenoids. Overexpression of the rate-limiting enzymes 1- deoxy-D-xylulose-5-phosphate synthase and reductoisomerase led to a 3.6-fold increase in lycopene production in *E*. *coli* when compared with the native MEP pathway for IPP and DMAPP supply [71]. Overexpression of the rate-limiting enzyme 3-hdroxy-3-methyl-glutaryl-coenzyme A (HNG-CoA) reductase of the MVA pathway from *Xanthophyllomyces dendrorhous* significantly increased Ά-carotene production in *S*. *cerevisiae* [82]. A great effort in metabolic engineering of the central carotenoid building block pathway is the introduction of a hybrid MVA pathway of *Streptococcus pneumonia* and *Enterococcus faecalis* into *E*. *coli*, which enables the recombinant host to produce 465 mg/L of Ά-carotene [83]. With more available genetic tools, microbial organisms such as *Pseudomonas putida* and *Bacillus subtilis* have also been developed as platform hosts for carotenoid production [84,85]. 

Carotenoids synthesis involves multiple enzymes [2]. The expression level of all the components of a multigene circuit should be orchestrated to optimize metabolic flux and to gain a high yield (Figure 5B) [86]. A random approach is screening of the best orchestra from numerous combinatorial assemblies of required genes and control elements. BioBrick™ paradigm is capable of rapidly assembling a biosynthetic pathway in a variety of gene orders from different promoters in plasmids with different copy numbers [87]. It is possible to build a hybrid carotenoid pathway wherein each enzyme possesses a right turnover number, however, BioBrick™ assembly is still not in a high throughput to create vast combinatorial expression constructs for the best combination of carotenogenic genes. Recently, several advanced assembly methods using homologous recombination, such as sequence and ligation-independent cloning (SLIC), Gibson DNA assembly and reiterative recombination, have been applied to construct multigene circuits [88–90]. These advances promise to randomize all genetic components, including genes, promoters, ribosome binding sites, and other control modules to build a large number of individual genetic circuits for screening purposes. A so-called "randomized BioBrick assembly" approach has been applied to the optimization of the lycopene synthesis pathway wherein the expression construct was designed to independently express each enzyme from its own promoter, which resulted in an increase by 30% in lycopene production [91]. A longer and more complicated pathway can be modularized into subsets, which contain pathway enzymes with similar turnover numbers. Modulating these subsets would be more convenient and efficient than regulating all components of the entire pathway for improved production [92]. By using this multivariate modular metabolic engineering (MMME) approach, recent work achieved a 15,000-fold increase in production of taxadiene, a precursor of the anti-cancer drug taxol [93]. There are also a variety of promising approaches, such as tunable promoters, tunable intergenic regions, and ribosome binding site design, which can be applied to fine tuning the expression of modules [94–96]. In the other approaches, a multi-genic operon is transcribed into a single  polycistronic mRNA, and then the large transcript can be spliced to small monocistronic transcripts through post-transcriptional RNA processing such as ribozyme cleavage and *clustered regularly interspaced short palindromic repeats* (CRISPR) editing. Thus, the stability of the monocistronic transcripts can be independently modulated to differentiate the expression level of each enzyme even in a multi-gene operon. These RNA processing tools have been developed as insulating elements between operonic genes to reduce the context dependence of the genes in a polycistronic transcription unit [97]. The diffusion of pathway intermediates can decrease the effective concentrations of intermediates for following enzyme reactions and some intermediates may serve for competing pathways. By learning from Mother Nature, synthetic biologists spatially organize enzymes of the MVA pathway by protein scaffolds in *E*. *coli* to minimize diffusion limitation and achieve a 77-fold increase in mevalonic acid production [98]. The propanediol utilization machinery of *Citrobacter freundii* has been heterologously recasted in *E*. *coli* [99]. Some intermediates of carotenoid synthesis such as isoprenyl diphosphates are toxic when they accumulate over the concentration threshold [100]. To avoid the accumulation of toxic intermediates, genetic sensors can potentially be coupled with gene expression cassettes to regulate the intermediate flux in a dynamic manner. The native *E*. *coli* promoters that respond to the toxic FPP have been successfully used to dynamically regulate the amorphadiene synthesis pathway and improve the production by two-fold over common inducible promoters and constitutive promoters [101]. The Ntr regulon has been engineered to control lycopene synthesis in response to glycolytic flux dynamics, resulting in an 18-fold increase in lycopene production [102]. 

## *3.3. Genome Engineering for Strain Development*

For the most efficient carotenoid production, the biological system of the host organism also needs to be optimized, by, for example, redirecting cellular carbon flux to the carotenoid synthesis pathway. The *de novo* synthesis of carotenoids is initiated from acetyl-coA by the MVA pathway or glycolytic metabolites pyruvate and glycraldehyde-3-phosphate (G3P) by the MEP pathway. The direct efforts are focused on the modification of associated genes to these pathways. Deletion of pyruvate kinases PykFA can balance the availability of pyruvate and G3P for the MEP pathway, and increase lycopene production by 2.8-fold in *E*. *coli* [103]. The deletion of glucose-6-phospahte (G6P) dehydrogenase Zwf, which branches G6P to pentose phosphate pathway results in an increase by 30% in lycopene production [104]. Deletion of carbohydrate phosphotransferase system yields a seven-fold increase in lycopene production in another study [105]. Replacement of native promoters of the rate-limiting genes of the MEP pathway with the T5 promoter has been carried out for enhancement of the targeted pathway flux, which results in a 4.5-fold increase in Ά-carotene production [106]. 

A heterologous pathway is not just an independent entity. It communicates with the native cellular metabolism and is therefore governed by the global regulation of the host organisms. Adaptive laboratory evolution is a traditional route for strain engineering to achieve desirable industrially relevant phenotypes. Owing to the antioxidant properties of carotenoids, adaptive evolution has been successfully applied to an engineered *S*. *cerevisiae* with periodic hydrogen peroxide shocking, resulting in a three-fold increasee of Ά-carotene production. Subsequent transcriptome analysis indicates that some genes related with lipid biosynthesis and MVA pathways are up-regulated in the adopted strains [107]. It also suggests that carotenoid production can be improved by modifications (knock-out or overexpression) of distant genes, which are responsible for the overall regulation of the metabolic network or the physiological fitness of the host (Figure 5C). In a genome-wide screening of yeast deletion collection, 24 deletions exhibit significant higher carotenoid levels than the wild type. The triple deletion of *ROX1*, *YJL064W*, and *YJL062W* shows an almost four-fold increase in total carotenoid production [108]. Gene deletions of *hnr*, *yjfP*, and *yjiD* related to the improvement of lycopene production have been identified from a global transposon *E*. *coli* mutant library [109]. Other gene deletions such as *gdhA*, *cyoA*, *ppc*, *gpmA*, *gpmB*, *eno*, *glyA*, aceE, *talB*, and *fdhF* have been *in silico* identified using a stoichiometric model [110]. The triple mutation of *gdhA*, *aceE* and *fdhF* was validated to increase lycopene production by nearly 40% in *E*. *coli* over the engineered parental strain. A similar set of gene deletions *dhA*, *cyoA*, *gpmA*, *gpmB*, *icdA*, and *eno* have been also *in silico* identified using different metabolic network models [111]. Overexpression of some genes encoding global regulatory proteins AppY, Crl, RpoS, and ElbAB, oxidoreductases TorC, YdgK, and YeiA, and hypothetical proteins YedR and YhbL, result in a significant increase in lycopene production in *E*. *coli* [112]. With a profound understanding of the landscape of genome manipulation, all these knocked-out and overexpressed alleles have been combined and optimized to generate high-fitness host strains for lycopene production [113,114]. ATP and NADPH are also important cofactors for the production of carotenoids. Using engineering ATP synthesis, pentose phosphate and TCA modules, recent work has shown the highest Ά-carotene production of 2.1 g/L by a fed-batch fermentation process in *E*. *coli* [115]. The advances in synthetic biology greatly boost genome manipulation on a large scale. Multiplex automated genome engineering (MAGE) simultaneously targets many locations on the chromosome for modification in a single cell or across a population of cells by directing ssDNA to the lagging strand of the replication fork during DNA replication [116]. The modifications can cover gene inactivation, expression regulations, and so on. Aforementioned twenty genes related to lycopene production have been targeted to tune their expression using a complex pool of synthetic DNAs, and lycopene production is increased more than five-fold. A complementary method called trackable multiplex recombineering (TRMR) has been developed to simultaneously map genome modifications that affect a trait of  interest, which combines parallel DNA synthesis, recombineering and molecular barcode technology to enable rapid modification of all *E*. *coli* genes in an *a priori* knowledge-independent way [117]. 

Metabolic engineering for the production of valuable compounds often heavily relies on plasmid-based expression of the synthesis pathway in a heterologous host. Although plasmids are easily manipulated and allow strong expression of targeted enzymes, the plasmid-based systems suffer from genetic instability such as plasmid loss, an additional antibiotic cost, and a potential risk of antibiotic marker spreading to other organisms [118]. Accordingly, chromosomal integration of the production pathway promises the host to achieve stable overproduction of the desirable chemicals including carotenoids. By Ώ-Red homologous recombination, plasmidfree engineered *E*. *coli* strain has been developed to produce lycopene and astaxanthin [119]. The expression cassettes can be integrated into different loci to increase the number of gene copies. P1 transduction usually plays a role in transfering the different alleles between host strains. Recently, an intelligent strategy called chemically inducible chromosomal evolution (ClChE) has been developed to reduce the daunting repeated one-at-a-time tasks in the chromosomal integration of target genes [88]. ClChE allows the host to acquire a high gene copy (up to 40 copies) expression of integrated pathways with increasing concentration of selective chemicals, and the increased copy number is stabilized by the removal of the *recA* gene. With this approach, lycopene production has been increased by 60% from single copy integrated strain. The ClChE strategy has been further modified to eliminate antibiotic marker for environmental safety and health issue after the evolution of the recombinant host strain [120]. 

## *3.4. Protein Engineering for Improvement of Carotenoid Production Enzymes*

Pathway engineering for efficient production of desired chemicals is often challenged by limitations associated with the pathway enzymes themselves, such as low turnover numbers and promiscuities generating unwanted by-products [121]. Protein engineering provides a powerful solution to improve specific activity and substrate specificity of enzymes, and even to create new activity. Methods of protein engineering include directed evolution and computer-assisted rational design (Figure 5D) [122,123]. Directed evolution is an iterative process that imitates Darwinian evolution in the laboratory to select or screen a desired phenotype from mutagenesis. Typically, error-prone polymerase chain reaction (PCR) is used to generate mutant libraries, and DNA shuffling is carried out to recombine existing mutations. It can be performed in a blind manner with limited information on target enzymes, such as structures and reaction mechanisms, but it relies on an effective screening strategy. It is practical for the evolution of carotenogenic enzymes due to the innate traits of carotenoid pigments. Six mutants ((H96L, R203W, A205V, A208V, F213L and A215T) have been isolated to improve the catalytic activity of Ά-carotene  ketolase from *Sphingomonas sp*. [124]. Three mutations (L175M, M99V, and M99I) of ketolase from *Paracoccus sp*. result in the improvement if its specificity of to synthesize astaxanthin [125]. *Staphylococcus aureus* dehydrosqualene (C30) synthase has evolved to synthesize lycopene by mutation F26L or F26S [126]. DNA shuffling of phytoene desaturases from *<sup>P</sup>*. *agglomerans* and *P*. *ananatis* results in the isolation of a variant favoring the production of fully conjugated tetradehydrolycopene [127]. Rational design of proteins is based on the *in silico* simulation and the prediction using *a priori* enzyme information, which greatly liberates biologists from onerous screening task. This strategy requires adequate information to predict specific targeted amino acid mutations, which can confer desired enzyme traits [128]. Unfortunately, the limited information on carotenogenic enzymes leads to few achievements using such a method. 

As aforementioned, carotenoids are derived from the central isoprenoid pathway, which is also employed to synthesize several essential and secondary metabolites in nature. The carotenoid-based colorimetric screening has been developed for evolution of other isoprenoid pathway enzymes. Mutations of GGPP synthase are hypothesized to affect the binding efficiency of the magnesium ions needed for substrate anchoring and improve its catalysis. An error-prone PCR library of *Tsuga canadensis* GPPS has been screened using the lycopene synthesis pathway as a colorimetric reporter. The GPPS variant (S239C and G295D) is created to increase levopimaradiene production with a 1.7-fold increase over the wild type in *E*. *coli* [129]. Augmentation of one pathway can tamper with other pathways, which utilize the same substrate in one organism. Based on this fact, mutagenesis libraries of terpene synthases have been screened by depigmentation of colonies due to the competition between terpene synthases and carotenoid synthases for isoprenyl diphosphates, since the weakened carotenoid color intensity indicates an improvement of terpene synthase activity [130]. 
