Production of the Marine Carotenoid Astaxanthin by Metabolically Engineered Corynebacterium glutamicum

Astaxanthin, a red C40 carotenoid, is one of the most abundant marine carotenoids. It is currently used as a food and feed additive in a hundred-ton scale and is furthermore an attractive component for pharmaceutical and cosmetic applications with antioxidant activities. Corynebacterium glutamicum, which naturally synthesizes the yellow C50 carotenoid decaprenoxanthin, is an industrially relevant microorganism used in the million-ton amino acid production. In this work, engineering of a genome-reduced C. glutamicum with optimized precursor supply for astaxanthin production is described. This involved expression of heterologous genes encoding for lycopene cyclase CrtY, β-carotene ketolase CrtW, and hydroxylase CrtZ. For balanced expression of crtW and crtZ their translation initiation rates were varied in a systematic approach using different ribosome binding sites, spacing, and translational start codons. Furthermore, β-carotene ketolases and hydroxylases from different marine bacteria were tested with regard to efficient astaxanthin production in C. glutamicum. In shaking flasks, the C. glutamicum strains developed here overproduced astaxanthin with volumetric productivities up to 0.4 mg·L−1·h−1 which are competitive with current algae-based production. Since C. glutamicum can grow to high cell densities of up to 100 g cell dry weight (CDW)·L−1, the recombinant strains developed here are a starting point for astaxanthin production by C. glutamicum.


Introduction
Carotenoids are natural pigments with yellow-to-red coloring properties, found ubiquitously in plants, algae, fungi, and bacteria. These pigments form a subfamily of the large and diverse group of terpenoids with more than 55,000 different structures. Terpenoids are natural secondary metabolites composed of isoprene units, which typically exhibit flavoring, fragrance and coloring properties. Carotenoids and their derivatives have become more and more important for the health care industry due to their beneficial effects on human and animal health and their possible pharmaceutical, medical, and nutraceutical applications. For example, carotenoids are suggested to have beneficial effects on the human immune system and to protect against degenerative diseases and cancer [1][2][3]. Astaxanthin is a marine, red, cyclic C40 carotenoid and the third most important carotenoid on the global market after β-carotene and lutein, with a predicted sales volume of 670 metric tons valued at 1.1 billion US$ in 2020 [4]. Currently, astaxanthin is primarily used as a food and beverage colorant, in animal feed Lycopene β-cyclase, crtZ: β-Carotene hydroxylase (3,3ʹ-beta-ionone ring hydroxylase), crtW: β-Carotene ketolase (4,4ʹ-beta-ionone ring ketolase). Endogenous genes are shown in grey boxes and their overexpression indicated by green arrows. Heterologous genes are highlighted in colored boxes.

Construction of a β-Carotene Producing C. glutamicum Base Strain
C. glutamicum was metabolically engineered for plasmid-independent lycopene overproduction ( Table 1). Chromosomal integration of the synthetic operon crtEBI under the control of the endogenous promoter of the gene coding for the translational elongation factor (Ptuf) in the crtYeYfEb deletion mutant of C. glutamicum MB001 (LYC3) [37] was performed in order to improve the expression of prenyltransferase CrtE, phytoene synthase CrtB and phytoene desaturase CrtI encoding genes. Thereby, the flux from the precursor molecules IPP and DMAPP to lycopene was enhanced and an 8-fold higher lycopene titer resulted for strain LYC4. When dxs, encoding the first enzyme of the MEP-pathway, was additionally overexpressed by chromosomal exchange of its natural promoter by the strong promoter Ptuf, the lycopene titer was further improved by 34% and the respective strain LYC5 produced 0.43 ± 0.02 mg•(g•CDW) −1 ( Table 1).

Construction of a β-Carotene Producing C. glutamicum Base Strain
C. glutamicum was metabolically engineered for plasmid-independent lycopene overproduction ( Table 1). Chromosomal integration of the synthetic operon crtEBI under the control of the endogenous promoter of the gene coding for the translational elongation factor (P tuf ) in the crtY e Y f Eb deletion mutant of C. glutamicum MB001 (LYC3) [37] was performed in order to improve the expression of prenyltransferase CrtE, phytoene synthase CrtB and phytoene desaturase CrtI encoding genes. Thereby, the flux from the precursor molecules IPP and DMAPP to lycopene was enhanced and an 8-fold higher lycopene titer resulted for strain LYC4. When dxs, encoding the first enzyme of the MEP-pathway, was additionally overexpressed by chromosomal exchange of its natural promoter by the strong promoter P tuf , the lycopene titer was further improved by 34% and the respective strain LYC5 produced 0.43˘0.02 mg¨(g¨CDW)´1 (Table 1). Strain LYC5 was converted to a β-carotene producing strain (Table 2) by heterologous expression of the lycopene β-cyclase gene crtY from P. ananatis. Plasmid-borne expression of crtY under the control of the isopropyl β-D-1-thiogalactopyranoside (IPTG) inducible tac promoter (pEKEx3_crtY Pa ) allowed for β-carotene production. Constitutive expression of crtY under control of the P tuf promoter from the newly constructed expression and shuttle vector pSH1 resulted in a comparable production titer. Similarly, a β-carotene titer of 6.5 mg¨g´1 was achieved by BETA3, a strain having crtY Pa under the control of P tuf integrated into the genome of C. glutamicum strain LYC5 (Table 2). Table 2. β-Carotene production in recombinant C. glutamicum strains. Cells were grown in glucose CGXII minimal medium for 24 h induced by 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG). Means and standard deviations of three replicates are given.

Design of the Combinatorial Gene Assembly and Library Construction for Engineering Astaxanthin Production in C. glutamicum
Metabolic flux in a synthetic pathway may require well-adjusted activities of the enzymes involved. Prediction of the flux from gene expression is rather difficult, hence, a combinatorial gene assembly was used to screen for balanced expression of the β-carotene ketolase and β-carotene hydroxylase encoding genes with respect to astaxanthin production. Since crtY from P. ananatis has previously been expressed successfully in C. glutamicum for production of β-carotene, the β-carotene hydroxylase crtZ gene from this organism was chosen. However, P. ananatis lacks β-carotene ketolase, and therefore the β-carotene ketolase gene crtW from Brevundimonas aurantiaca was used, which on the contrary lacks a crtZ gene. CrtW from B. aurantiaca and crtZ from P. ananatis were combined in an artificial operon under the control of the constitutive P tuf promoter in the vector pSH1. Gene expression was varied by combining different ribosome-binding sites (RBS) and start codons separated by spacers of different lengths ( Figure 2). The theoretical translation initiation rates were calculated using the RBS calculator [43] and ranged from 14 to 33,626 for crtW and from 40 to 30,731 for crtZ. A library of combinatorially assembled crtW and crtY genes was generated and the constructed library of pSH1_crtW Ba _crtZ Pa plasmids was used to transform the β-carotene accumulating strain C. glutamicum BETA1. Strain LYC5 was converted to a β-carotene producing strain (Table 2) by heterologous expression of the lycopene β-cyclase gene crtY from P. ananatis. Plasmid-borne expression of crtY under the control of the isopropyl β-D-1-thiogalactopyranoside (IPTG) inducible tac promoter (pEKEx3_crtYPa) allowed for β-carotene production. Constitutive expression of crtY under control of the Ptuf promoter from the newly constructed expression and shuttle vector pSH1 resulted in a comparable production titer. Similarly, a β-carotene titer of 6.5 mg•g −1 was achieved by BETA3, a strain having crtYPa under the control of Ptuf integrated into the genome of C. glutamicum strain LYC5 (Table 2). Table 2. β-Carotene production in recombinant C. glutamicum strains. Cells were grown in glucose CGXII minimal medium for 24 h induced by 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG). Means and standard deviations of three replicates are given.

Design of the Combinatorial Gene Assembly and Library Construction for Engineering Astaxanthin Production in C. glutamicum
Metabolic flux in a synthetic pathway may require well-adjusted activities of the enzymes involved. Prediction of the flux from gene expression is rather difficult, hence, a combinatorial gene assembly was used to screen for balanced expression of the β-carotene ketolase and β-carotene hydroxylase encoding genes with respect to astaxanthin production. Since crtY from P. ananatis has previously been expressed successfully in C. glutamicum for production of β-carotene, the β-carotene hydroxylase crtZ gene from this organism was chosen. However, P. ananatis lacks β-carotene ketolase, and therefore the β-carotene ketolase gene crtW from Brevundimonas aurantiaca was used, which on the contrary lacks a crtZ gene. CrtW from B. aurantiaca and crtZ from P. ananatis were combined in an artificial operon under the control of the constitutive Ptuf promoter in the vector pSH1. Gene expression was varied by combining different ribosome-binding sites (RBS) and start codons separated by spacers of different lengths ( Figure 2). The theoretical translation initiation rates were calculated using the RBS calculator [43] and ranged from 14 to 33,626 for crtW and from 40 to 30,731 for crtZ. A library of combinatorially assembled crtW and crtY genes was generated and the constructed library of pSH1_crtWBa_crtZPa plasmids was used to transform the β-carotene accumulating strain C. glutamicum BETA1.  For each gene four different RBS, three different spacer lengths, and two different translational start codons were chosen. These were introduced by the forward primers and equimolar mixture of these primers and one reverse primer by PCR. The resulting DNA products were gel-extracted and combined by cloning via Gibson Assembly [44] in pSH1. Thus, theoretically 24 different constructs per gene resulted ( Figure 2). With this approach 576 different combinations of crtW and crtZ genes are theoretically possible and the event of creating a specific combination of the two genes follows the Poisson distribution [45] with a probability of 1/576 (Equation (1)). To cover with 99% probability that a single specific combination is present at least once in the library, approximately 2650 clones are required (Equation (1)). The necessary number of transformants for creating a library with each of the 576 combinations can be calculated employing the path rules [45]. For creating a library that includes each of the 576 specific combinations at least once with a 99% probability, approximately 6315 transformants are required (Equation (2)). Preliminary experiments showed that correct assembling of an insert with the restricted vector via Gibson assembly occurs in about 90% of the events. Consequently, the number of transformants had to be corrected by multiplication by 1.11, thus, a minimum of 7000 transformants had to be screened.
Equation (1): Poisson distribution. λ = n * p; n: library size; p: probability of one specific gene assembly of crtW and crtZ, k: number of one specific gene assembly in library with size n. p all pk ě 1q "´1´e´λ¯N (2) Equation (2): Path rules. λ = n * p; n: library size; p: probability of one specific gene assembly of crtW and crtZ, k: number of one specific gene assembly in library with size n; N: number of possible gene assemblies.
Around 8000 transformants were visually color-screened on plates and 46 colonies with different colors ranging from yellow to red were selected for further analysis. The plasmid DNA was isolated and sequenced to identify the sequences (RBS, spacer, translational start codon) of crtW and crtZ. The set of 46 transformants represented 20 of the 24 possible variants for crtW and 19 of 24 variants of crtZ. Furthermore, three plasmids harbored only the crtW gene and two plasmids harbored only the crtZ gene.

Combinatorial Engineering Covered Vastly Different Astaxanthin, β-Carotene, Zeaxanthin and Canthaxanthin Titers
To evaluate which of the gene combinations was best in terms of astaxanthin production, the 46 selected transformants referred to as COMB strains, were characterized with respect to carotenoid production. After growth in CGXII minimal medium with 100 mM glucose, appropriate antibiotics and 1 mM IPTG in a Biolector micro fermentation system (Figure 3), carotenoids were quantified by HPLC using standards for β-carotene, canthaxanthin, zeaxanthin, and astaxanthin.
As expected, the parental strain BETA1 ( Figure 3) produced β-carotene (6.7 mg¨(g¨CDW)´1), but no further carotenoids. The 46 COMB strains could be categorized in six groups according to their carotenoid production profiles (group I: only lycopene, group II: only β-carotene, group III: β-carotene and zeaxanthin, group IV: β-carotene, zeaxanthin and astaxanthin, groupV: β-carotene and canthaxanthin, group VI: β-carotene, canthaxanthin and astaxanthin; Figure 4). For all COMB strains, the TIRs were calculated with the RBS calculator tool [41], which takes (amongst others) the free binding energy of the RBS and the 16S rRNA into consideration as well as the free energy of secondary structures of the mRNA itself.

Combinatorial Engineering Covered Vastly Different Astaxanthin, β-Carotene, Zeaxanthin and Canthaxanthin Titers
To evaluate which of the gene combinations was best in terms of astaxanthin production, the 46 selected transformants referred to as COMB strains, were characterized with respect to carotenoid production. After growth in CGXII minimal medium with 100 mM glucose, appropriate antibiotics and 1 mM IPTG in a Biolector micro fermentation system (Figure 3), carotenoids were quantified by HPLC using standards for β-carotene, canthaxanthin, zeaxanthin, and astaxanthin. As expected, the parental strain BETA1 ( Figure 3) produced β-carotene (6.7 mg•(g•CDW) −1 ), but no further carotenoids. The 46 COMB strains could be categorized in six groups according to their carotenoid production profiles (group I: only lycopene, group II: only β-carotene, group III: βcarotene and zeaxanthin, group IV: β-carotene, zeaxanthin and astaxanthin, groupV: β-carotene and canthaxanthin, group VI: β-carotene, canthaxanthin and astaxanthin; Figure 4). For all COMB strains, the TIRs were calculated with the RBS calculator tool [41], which takes (amongst others) the free binding energy of the RBS and the 16S rRNA into consideration as well as the free energy of secondary structures of the mRNA itself. . Carotenoid profiles and calculated translational initiation rates (TIRs) for C. glutamicum strains expressing crtW from B. aurantiaca and crtZ from P. ananatis with varied translation initiation signal. TIRs were calculated by applying the RBS calculator tool [46] on the mRNA sequence. TIRs were classified as follows: TIRs <200: low; 200 < TIRs < 2000: medium; TIRs >2000: high. Production of β-carotene, zeaxanthin, canthaxanthin and astaxanthin was determined after 24 h of cultivation in CGXII + 100 mM glucose in Biolector micro fermenter. COMB 40 ( Figure 4) produced none of the cyclic carotenoids, but about as much lycopene (0.39 mg•(g•CDW) −1 ) as LYC5 (0.5 ± 0.1 mg•(g•CDW) −1 ), the parental strain of BETA1. Sequencing of pEKEx3_crtYPa isolated from COMB 40 revealed a deletion of 11 base pairs in the coding region of crtY, hence, β-carotene production was not possible in this strain. By contrast, the other 45 strains produced β-carotene with a titer of at least 1 mg•(g•CDW) −1 (Figure 4). For about 24% of the strains, β-carotene was the only cyclic carotenoid being produced. In these cases, the calculated TIRs of crtW and/or crtZ were rather low (less than 200 for at least one gene; Figure 4). Zeaxanthin, one of the intermediates in the pathway towards astaxanthin, was detected in only four strains (COMB 14, Figure 4. Carotenoid profiles and calculated translational initiation rates (TIRs) for C. glutamicum strains expressing crtW from B. aurantiaca and crtZ from P. ananatis with varied translation initiation signal. TIRs were calculated by applying the RBS calculator tool [46] on the mRNA sequence. TIRs were classified as follows: TIRs <200: low; 200 < TIRs < 2000: medium; TIRs >2000: high. Production of β-carotene, zeaxanthin, canthaxanthin and astaxanthin was determined after 24 h of cultivation in CGXII + 100 mM glucose in Biolector micro fermenter. COMB 40 ( Figure 4) produced none of the cyclic carotenoids, but about as much lycopene (0.39 mg¨(g¨CDW)´1) as LYC5 (0.5˘0.1 mg¨(g¨CDW)´1), the parental strain of BETA1. Sequencing of pEKEx3_crtY Pa isolated from COMB 40 revealed a deletion of 11 base pairs in the coding region of crtY, hence, β-carotene production was not possible in this strain. By contrast, the other 45 strains produced β-carotene with a titer of at least 1 mg¨(g¨CDW)´1 ( Figure 4). For about 24% of the strains, β-carotene was the only cyclic carotenoid being produced. In these cases, the calculated TIRs of crtW and/or crtZ were rather low (less than 200 for at least one gene; Figure 4). Zeaxanthin, one of the intermediates in the pathway towards astaxanthin, was detected in only four strains (COMB 14, COMB 26, COMB 30, and COMB 35) and these strains exhibited very diverse TIRs for crtZ (from 81 to 5887) ( Figure 4). The highest production of zeaxanthin was detected in group IV for strain COMB 30 with 0.3 mg¨(g¨CDW)´1, although this strain possessed a low TIR for crtZ. The highest titers of canthaxanthin and astaxanthin were observed among the strains of the large group VI (39%) and these strains co-produced β-carotene along with canthaxanthin and astaxanthin ( Figure 4). The intermediate canthaxanthin was detected in 30 strains with strain COMB 42 showing the highest titer for canthaxanthin (0.6 mg¨(g¨CDW)´1; Figure 4). In average, the TIR for crtW of these strains was high (10,299). Astaxanthin was identified in 20 of the 46 strains, but only two strains, COMB 44 and COMB 48, exhibited reasonably high astaxanthin yields (approximately 0.3 mg¨(g¨CDW)´1). These two strains exhibited high crtW TIRs (33,626) and medium to high crtZ TIRs (5813 and 1377) ( Figure 4).
In general, it was found that the higher the TIR of crtW the higher was the astaxanthin production, with three exceptions, COMB 11, COMB 12, and COMB 45 ( Figure 4). In the latter three strains, however, the TIRs for crtZ were 3 to 145-fold lower than in the best astaxanthin producing strains COMB 44 and COMB 48. For strains COMB 44 and COMB 48, a spacing length of six base pairs, the RBS sequence GAAAGGAGG, and the translation start codon ATG was found for crtW. The crtZ gene in COMB 44 showed the consensus RBS sequence, a spacer length of eight base pairs and ATG as translational start codon. The crtZ gene variant of COMB 48 had a slightly lower TIR and possessed the RBS sequence GAAAGAAGG, six base pairs of spacing and ATG as start codon.
Three strains (COMB 37, COMB 3 and COMB 19) did not express crtZ due to an incorrect gene assembly. Strains COMB 3 and COMB 19 accumulated canthaxanthin besides β-carotene, while COMB 37, which also showed a low TIR for crtW, only accumulated β-carotene ( Figure 4). Strains COMB 25 and COMB 14 did not express crtW due to an incorrect gene assembly. While COMB 25 only produced β-carotene probably because of a very low TIR for crtZ (Figure 4), strain COMB 14 produced zeaxanthin besides β-carotene.
Taken together, widely varied carotenoid production was represented by the library, but none of the combinations tested yielded high astaxanthin product levels.

In Silico Analysis of the Carotenogenic Genes crtZ and crtW from Marine and Non-Marine Bacteria for
Heterologous Expression in C. glutamicum In the above described experiments the bacteria B. aurantiaca and P. ananatis were chosen as sources for crtW and crtZ, respectively, although not producing astaxanthin themselves. B. aurantiaca lacks crtZ, but possesses the crtG gene coding for a 2,2'-beta-ionone ring hydroxylase and produces canthaxanthin and 2-hydroxycanthaxanthin as main carotenoids. P. ananatis lacks crtW and produces glycosylated zeaxanthin involving CrtZ. Thus, on the basis of available genome sequences, reported carotenoid production and biological diversity, four alternative prokaryotic natural carotenoid producers were selected as donors for crtW and crtZ. Since crt genes of Brevundimonas species were reported to show a high potential for heterologous carotenoid productions [47] two different Brevundimonas strains were selected: Brevundimonas vesicularis, a non-marine bacterium suggested to be a suitable gene donor for astaxanthin production [47,48], and as alternative Brevundimonas bacteroides [49]. The marine bacterium Fulvimarina pelagi was chosen due to its promising brownish-yellow color as a consequence of carotenoid accumulation [50] and the evolutionary distance to Brevundimonas. In addition, the red-pigmented marine bacterium Sphingomonas astaxanthinifaciens was selected since experimental evidence that astaxanthin is the major carotenoid produced by this bacterium has been reported [51,52].
The organization of carotenogenic gene clusters of the considered donors B. aurantiaca, B. bacteroides, B. vesicularis, F. pelagi, P. ananatis, and S. astaxanthinifaciens was analyzed on the basis of the partly available genome sequences/contigs or scaffolds in GenBank: the carotenoid gene cluster of B. bacteroides, an orange-red pigmented bacterium, comprises crtW and crtZ as well as the gene idi encoding the IPP isomerase of the MEP-pathway and several other genes encoding for enzymes of the astaxanthin biosynthesis pathway, however, a crtE gene is not present in its genome. The genome of B. vesicularis DC263, a red-pigmented soil bacterium, possesses a large carotenoid gene cluster with 10 coding sequences, eight of which encode enzymes for the biosynthesis pathway of astaxanthin or the terpenoid precursors IPP and DMAPP. In addition, a second hydroxylase CrtG is encoded, which is responsible for further hydroxylation of astaxanthin leading to 2-hydroxyastaxanthin. Carotenogenic genes of F. pelagi, a Mn(II)-oxidizing bacterium [53], are found in at least four different loci of the genome. Genes encoding for enzymes of the astaxanthin biosynthesis and glycosylation as well as enzymes for the spirilloxanthin biosynthesis (CrtC, CrtD, CrtF) are present. Furthermore, two genes coding for an ABC-transporter and a MFS-transporter are located next to the carotenogenic genes crtZ and crtY. S. astaxanthinifaciens, producing astaxanthin and its glycosides, has at least two carotenoid gene clusters in its genome also including farnesyl pyrophosphate synthase. Moreover, a gene encoding a putative carotenoid transporter is located in this cluster.

High
Astaxanthin Production by C. glutamicum Strains Expressing crtW and crtZ from F. pelagi β-Carotene ketolase and hydroxylase genes (crtW and crtZ, respectively) from B. aurantiaca, B. bacteroides, B. vesicularis, F. pelagi, P. ananatis, and S. astaxanthinifaciens were expressed in the plasmid-free β-carotene overproducing C. glutamicum strain BETA4. The affinities of the β-carotene ketolases and hydroxylases for the various substrates and intermediates of the branched astaxanthin biosynthesis pathway may vary and it is conceivable that astaxanthin production proceeds e.g., only via canthaxanthin or only via zeaxanthin. However, also various other routes via hydroxyechinenone are possible (Figure 1). Thus, in a first step only either crtW or crtZ was expressed in the parental strain BETA4 that produced~12 mg¨(g¨CDW)´1 β-carotene with a productivity of~3.4 mg¨L´1¨h´1 in 24 h of cultivation and a growth rate of 0.32˘0.01 h´1. Zeaxanthin accumulated (0.52 and 1.1 mg¨(g¨CDW)´1, respectively) when crtZ from P. ananatis or F. pelagi were expressed (data not shown). Canthaxanthin accumulated (0.34 to 1.0 mg¨(g¨CDW)´1) when crtW from S. astaxanthinifaciens, F. pelagi or B. aurantiaca were expressed, while only traces were observed as consequence of expression of crtW from B. bacteroides or B. vesicularis (data not shown). Thus, crtW from S. astaxanthinifaciens, F. pelagi or B. aurantiaca and crtZ from P. ananatis or F. pelagi appeared suitable for further analysis.
Subsequently, combinations of crtW from S. astaxanthinifaciens and B. aurantiaca with crtZ from F. pelagi were co-expressed in strain BETA4 using the two expression vectors pSH1 and pEC-XT99A. In addition, the crtW and crtZ genes from species known to synthesize astaxanthin (B. bacteroides, B. vesicularis, F. pelagi, and S. astaxanthinifaciens) were co-expressed in BETA4. Carotenoids of these strains were extracted and analyzed in the stationary growth phase 24 h after inoculation. Transformants harboring the genes from B. vesicularis grew poorly and were not analyzed further.

Discussion
In this study, Corynebacterium glutamicum was engineered for the production of the marine carotenoid astaxanthin. C. glutamicum grows fast to high cell densities [54] and, thus, is suitable for production of carotenoids and other compounds that are stored within the cell. Here, C. glutamicum was shown to produce β-carotene to about 12 mg¨(g¨CDW)´1 within 24 h at a volumetric productivity of about 3.4 mg¨L´1¨h´1. Growth and production of carotenoids by C. glutamicum is monophasic and strains BETA4 and ASTA1 showed growth rates of 0.32˘0.01 h´1 and 0.29˘0.05 h´1, respectively. This is in contrast to biphasic growth/production of carotenoids e.g., by the alga Haematococcus pluvialis [55]. As a consequence, the volumetric productivity for β-carotene exceeds that reported for the industrially used microalga Dunaliella bardawil [56] or the yeast Saccharomyces cerevisiae [57] by about a factor of three.
Combined expression of the genes coding for β-carotene ketolase and hydroxylase from microorganisms that do not synthesize astaxanthin (B. aurantiaca and P. ananatis) in a β-carotene producing C. glutamicum led to astaxanthin production. However, astaxanthin was not the main carotenoid being produced. Since a balanced expression of the β-carotene ketolase and hydroxylase genes are essential for an efficient astaxanthin production [48,58] we assumed that the activities of the respective enzymes in the tested recombinants were not matched. Therefore, translation initiation rates of the respective genes, crtW and crtZ, were varied in a combinatorial approach. However, a strict correlation between TIR and production titers was not observed. As tendencies, the lower the TIRs of both crtW and crtZ the lower were the canthaxanthin and astaxanthin titers, and the higher the TIR of crtW the higher were astaxanthin titers (Figure 4).
In E. coli astaxanthin biosynthesis from β-carotene was reported to proceed more efficiently via zeaxanthin rather than canthaxanthin since ketolated intermediates did not accumulate [48,58]. Both ketolase and hydroxylase compete for their substrates and accept β-carotene as well as canthaxanthin and zeaxanthin, respectively, as substrates [59,60]. Independently induced expression of crtZ from P. ananatis and crtW148 of Nostoc puntiforme PC73102 revealed that hydroxylation occurred fast with β-carotene, echinenone, adonirubin, and canthaxanthin [58]. In their system, CrtW148 was identified as the limiting step in conversion of zeaxanthin to astaxanthin [58]. Expression of crtZ from P. ananatis in β-carotene producing C. glutamicum also yielded zeaxanthin [38] as did expression of crtZ from F. pelagi in this study (data not shown). Varying expression levels of crtW Ba and crtZ Pa led to accumulation of zeaxanthin only if TIR for crtW Ba was low (Figure 4). On the other hand, canthaxanthin accumulated as intermediate typically if TIR of crtW Ba was medium to high ( Figure 4). Canthaxanthin accumulation may be explained best by the assumption that β-carotene ketolase CrtW from B. aurantiaca did not accept the non-natural substrate zeaxanthin well. It is likely that astaxanthin production by this approach was not only limited by an imperfect match between expression levels of the β-carotene ketolase and hydroxylase genes, but rather by imperfect compatibility of the substrate spectra of the chosen β-carotene ketolase and hydroxylase enzymes.
Consequently, crtW and crtZ genes from marine and non-marine bacteria known to synthesize astaxanthin were examined in the second approach. Astaxanthin was produced in combinations of CrtZ from the marine bacterium F. pelagi and CrtW from either F. pelagi, S. astaxanthinifaciens or B. aurantiaca. F. pelagi was isolated from ocean surface water, an aerated environment at least transiently exposed to high solar radiation [45]. It is hypothesized that carotenoids play an important role as antioxidants for survival of F. pelagi under these conditions [50]. Analysis of the codon usage of crtW and crtZ from F. pelagi revealed a good fit to the codon usage of C. glutamicum, which is in compliance with the achieved astaxanthin titers of the recombinants. Co-expression of crtW from B. aurantiaca and crtZ from F. pelagi led to comparable astaxanthin titers, but considerable β-carotene amounts accumulated as side-product (Table 3), co-expression of crtW and crtZ from F. pelagi, instead, yielded astaxanthin as major carotenoid (80%; Table 3).
As compared to β-carotene production of about 12 mg¨(g¨CDW)´1 by the parent strain BETA4, the astaxanthin titers were at least seven fold lower (Table 3). Thus, conversion of β-carotene to astaxanthin is incomplete; however, other carotenoids besides canthaxanthin and residual β-carotene did not accumulate to significant titers (data not shown and Table 3). The partial conversion of β-carotene to astaxanthin may, thus, indicate that astaxanthin and/or intermediate(s) of its biosynthesis are inhibitory. This is in line with our finding that overexpression of only crtW from F. pelagi resulted in 0.5 mg¨(g¨CDW)´1 canthaxanthin and 1.7 mg¨(g¨CDW)´1 remaining β-carotene. Similarly, overexpression of only crtZ yielded 1.1 mg¨(g¨CDW)´1 zeaxanthin and 5.6 mg¨(g¨CDW)´1 β-carotene remained. Similarly, heterologous expression of crtW148 and crtZ in the β-carotene-producing E. coli strain reduced the overall formation of carotenoids, indicating that the formation of the carotenoid precursors were affected [58].
High product purities and titers are beneficial for downstream processing. The astaxanthin producing C. glutamicum strain overexpressing crtW and crtZ from F. pelagi accumulated astaxanthin (about 1.6 mg¨(g¨CDW)´1) as major (about 80%) carotenoid. The fact that little β-carotene and canthaxanthin accumulated (about 0.3 and 0.1 mg¨(g¨CDW)´1, respectively) may be an important advantage for downstream processing. Nevertheless, higher product purities can be obtained by algae with 95% of total carotenoids being astaxanthin [58]. Purification of astaxanthin from the cell walls of algae and red yeasts is challenging since algae like H. pluvialis accumulate astaxanthin in response to stress and heavily walled cysts are formed in the red stage [55]. Extraction of carotenoids from microalgae does not only require the removal of chlorophyll [61], but also efficient cell breakage technology [55]. Ethoxyquin or other antioxidants are added to the cells in order to minimize oxidation of the carotenoids during drying and cracking [58]. Because of laborious and time-consuming extraction processes of astaxanthin from algal systems, its production by a prokaryotic host, Escherichia coli, has emerged for substitution [62]. It has to be noted that H. pluvialis produces esterified astaxanthin, which is more stable than the free form astaxanthin as it does not cross react with proteins and e.g., lipoproteins [8], and which is incorporated easier by marine animals [63]. But hydrolysis of the ester narrows the bioavailability of astaxanthin e.g., to salmon [64]. The rigid cell walls of the red yeast X. dendrorhous also requires cell breakage prior to astaxanthin extraction [65,66]. In contrast to that, a simple methanol-acetone extraction was sufficient to recover astaxanthin from C. glutamicum cells at lab scale.
The volumetric productivities of up to about 0.4 mg¨L´1¨h´1 obtained in simple shaking flask cultures by the recombinant C. glutamicum strains compare favorably with those reported for the commercially used production hosts such as the green microalgae H. pluvialis [55,67] and the red yeast Xanthophyllomyces dendrorhous (formerly Pfaffia rhodozyma) [6,68] under similar conditions as well as recombinant E. coli [58]. Under optimal conditions, astaxanthin titers obtained e.g., with H. pluvialis are very high (up to about 40 mg¨(g¨CDW)´1), but slow growth, biphasic growth (green stage) and production (red stage) properties and the low final biomass concentrations reduce the maximal volumetric productivity [55]. After the non-productive green phase (about 4 days), the volumetric productivity for astaxanthin in the red stage is about 1 mg¨L´1¨h´1 and can be maintained for extended periods [55]. Although astaxanthin product titers from red yeasts such as X. dendrorhous are generally lower than from algae [69], higher growth rates and easier cultivation conditions argue in favor of these yeasts [70]. After optimization of a glucose-based fed-batch process a volumetric productivity of about 5 mg¨L´1¨h´1 was achieved [65,71]. Can it be envisioned that comparably high volumetric productivities can be obtained using the recombinant C. glutamicum strains described here? In pressurized high-cell-density fed-batch cultivations C. glutamicum grows to biomass concentrations of about 220 g¨CDW¨L´1 within 24 h [54]. If this growth could be achieved with the C. glutamicum strains accumulating astaxanthin to titers of about 1.6 mg¨(g¨CDW)´1, theoretically volumetric productivities of about 14 mg¨L´1¨h´1 may be achieved. Future work focused on process intensification, however, needs to be performed in order to evaluate if scale-up to such high astaxanthin volumetric productivities can be realized with C. glutamicum.

Bacterial Strains, Media and Growth Conditions
The strains and plasmids used in this work are listed in Table 4. C. glutamicum ATCC 13032 was used as wild type (WT), for metabolic engineering the prophage-cured C. glutamicum MB001 [72] was used as platform strain. Pre-cultivation of C. glutamicum strains was performed in LB medium or LB with 50 mM glucose. For cultivation in CGXII medium [73], pre-cultivated cells were washed once with CGXII medium without carbon source and inoculated to an initial OD 600 of 1. Glucose was added as carbon and energy source to a concentration of 100 mM. Standard cultivations of C. glutamicum were performed at 30˝C in a volume of 50 mL in 500 mL flasks with two baffles shaking at 120 rpm. The OD 600 was measured in dilutions using a Shimadzu UV-1202 spectrophotometer (Duisburg, Germany). Alternatively, cultivations were performed in 1 mL volume in micro-titerplates at 1100 rpm at 30˝C using Biolector ® micro fermentation system (m2p-labs GmbH, Baesweiler, Germany). For cloning, E. coli DH5α was used as host and cultivated in LB medium at 37˝C. When appropriate, kanamycin, tetracycline or spectinomycin was added to concentrations of 25, 5, and 100 µg¨mL´1, respectively. Gene expression was induced by addition of 1 mM IPTG, at inoculation of the main culture.  Km R ; E. coli/C. glutamicum shuttle vector for construction of insertion and deletion mutants in C. glutamicum (pK18 oriV Ec sacB lacZα) [82] pK19mobsacB-cg0725 pK19mobsacB with a cg0725 deletion construct -pK19mobsacB-P tuf -dxs pK19mobsacB derivative with a tuf promoter region (200 bp upstream of the coding sequence of the tuf gene(cg0587) construct for the promoter exchange of dxs [42] pK19mobsacB-IntcrtEBI pK19mobsacB derivative containing the artificial operon crtE_crtBI under the control of the P tuf promoter with an addition ribosome binding site in front of crtB for integration in the cgp2 cured region of C. glutamicum MB001 this work pVWEx1-crtEBI pVWEx1 derivative for IPTG-inducible expression of crtE, crtB and crtI from C. glutamicum containing artificial ribosome binding sites in front of crtE and crtBI [38] pK19mobsacB-IntcrtY pK19mobsacB derivative containing crtY of Pantoea ananatis under the control of the P tuf promoter for integration in the cgp1 cured region of C. glutamicum MB001 this work

Recombinant DNA Work
Plasmids were constructed in E. coli DH5α from PCR-generated fragments (All-in HiFi, highQu, Kraichtal, Germany) and isolated with the Plasmid GeneJET Miniprep kit (Thermo Fisher Scientific, Schwerte, Germany). Oligonucleotides used in this study were obtained from Metabion (Planegg/Steinkirchen, Germany) and are listed in Table 5. Standard reactions like restriction, and PCR were performed as described previously [83]. Besides cloning by restriction, Gibson assembly was applied for the construction of plasmids [44]. If applicable, PCR products were purified using the PCR clean-up and gel extraction kit (Macherey-Nagel, Düren, Germany). For transformation of E. coli DH5α, the RbCl method was used and C. glutamicum was transformed via electroporation [84] at 2.5 kV, 200 Ω, and 25 µF. All cloned DNA fragments were confirmed by sequencing.

Construction of Expression Vector pSH1
The plasmid pSH1 was constructed based on the expression vector pVWEx1 [81]. The backbone of pSH1 was amplified from pVWEx1 omitting the lacIq and P tac region using the oligonucleotides pV1-fw and pV6962-rv (Table 5) with All-in HiFi polymerase (highQu, Kraichtal, Germany). The promoter of the C. glutamicum tuf gene (cg0587) was amplified using the primers pV_P tuf -fw and pV_P tuf -rv (Table 5). Both fragments were assembled using the Gibson method [44]. Vector sequence was confirmed via sequencing to exclude mutations.

Deletion and Exchenge Mutagenesis in the Genome of C. glutamicum
For targeted deletion of cg0725, which encodes a transcriptional regulator and is part of the carotenogenesis gene cluster of C. glutamicum, the suicide vector pK19mobsacB was used [82]. Genomic regions flanking cg0725 were amplified from genomic DNA of C. glutamicum WT using primer pairs cg0725-A/B and cg0725-C/D (Table 5), respectively. Subsequently the purified PCR products were linked by crossover PCR using the primer pair cg0725-A/D ( Table 5). The resulting amplificate was cloned into pK19mobsacB resulting in the construction of deletion vector pK19mobsacB-cg0725 ( Table 4). Deletion of cg0725 via two-step homologous recombination as well as the selection for the first and second recombination events were carried out as described previously [85]. Successful removal of cg0725 was verified by PCR analysis of the constructed mutant using primer pair cg0725-E/F ( Table 5).
The integration of the synthetic operon crtEBI and the lycopene cyclase gene of Pantoea ananatis crtY, respectively, was conducted by using the suicide vector pK19mobsacB [82]. Operon crtEBI consists of the carotenogenic genes crtE (cg0723), crtB (cg0721) and crtI (cg0720) and was amplified from the expression vector pVWEx1-crtEBI [38] using the oligonucleotides crtEBI-Int5 and crtEBI-Int6. The P tuf promoter region was amplified using the oligonucleotides crtEBI-Int3/4 or crtY-Int3/4, respectively. Genomic regions flanking the selected insertion region were amplified from genomic DNA of C. glutamicum MB001 using primer pairs crtEBI-Int1/2 and crtEBI-Int7/8 for integration in the cgp2 cured region in the case of the crtEBI operon, or crtY-Int1/2 and crtY-Int7/8 for integration of crtY in the CGP1 cured region (Table 5), respectively. CrtY was amplified from genomic DNA of P. ananatis using the primer pair crtY-Int5/6. The purified PCR products were directly combined together with the plasmid by Gibson assembly [44]. The final assembly of the insert with linearized pK19mobsacB led to the construction of the respective integration vectors pK19mobsacB-IntcrtEBI and pK19mobsacB-IntcrtY (Table 4). The following integration of the operon by two-step homologous recombination was performed according to the deletion of genes. The integration in the cgp1 or cgp2 region was verified by PCR using the primers cgp1-E/F and cgp2-E/F, respectively.
The plasmid pK19mobsacB-P tuf -dxs was constructed to replace the native dxs promoter with the tuf promoter region from C. glutamicum WT as described earlier [42]. The promoter exchange was verified by PCR using the primers dxs E and 33, and sequencing of the PCR product.

Combinatorial Gene Assembly, Library Construction and Overexpression of Carotenogenic Genes
The combinatorial assembly of genes crtW Ba and crtZ Pa was performed with Gibson Assembly [44]. The crtW gene was amplified from the genomic DNA of Brevundimonas aurantiaca in a one-pot-PCR containing an equimolar mixture of forward primers (N1-N24) and a reverse primer (N49) ( Table 5). The crtZ gene was amplified from the genomic DNA of Pantoea ananatis in a one-pot-PCR containing an equimolar mixture of forward primers (N25-N48) and a reverse primer (N50) ( Table 5). PCR products of both genes were gel-extracted (Macherey-Nagel) and cloned in BamHI-restricted pSH1 applying Gibson Assembly. The transformation of E. coli DH5α was done as described above. 1/10 of the transformed cells were plated on selective agar plates for colony number calculation while 9/10 of the transformants were grown in selective liquid medium for plasmid isolation. Isolated plasmids were used for transformation of C. glutamicum strain BETA1.

Extraction and Quantification of Carotenoids
For extracting carotenoids from C. glutamicum, 1 mL of the culture was harvested by centrifugation for 7 min at 14,000 rpm. Carotenoid pigments were extracted with 800 µL methanol:acetone (7:3) containing 0.05% BHT at 60˝C for 15 min with careful vortexing every 5 min. Cell debris was spun down for 7 min at 14,000 rpm and the supernatant was used for high performance liquid chromatography (HPLC) analysis. For HPLC the Agilent 1200 series system (Agilent Technologies Sales & Services GmbH & Co. KG, Waldbronn, Germany) was used. The UV/visible (Vis) spectrum was recorded with a diode array detector (DAD). The quantification of carotenoids was performed by the integration of the extracted wavelength chromatogram at λ max 470 nm for every maximum and by the analysis of the appropriate UV/Vis profiles. Standard calibration curves were generated with lycopene (Sigma-Aldrich), β-carotene (Sigma-Aldrich), canthaxanthin (Sigma-Aldrich), zeaxanthin (Sigma-Aldrich) and astaxanthin (Sigma-Aldrich) to quantify carotenoid titers. All standards were dissolved in chloroform according to their solubility and diluted in methanol:acetone (7:3) containing 0.05% BHT.
As column system, a precolumn (10ˆ4 mm MultoHigh 100 RP18-5, CS Chromatographie Service GmbH, Langerwehe, Germany) and a main column (ProntoSIL 200-5 C30, 250ˆ4 mm, CS Chromatographie Service GmbH) were used. The HPLC protocol ensured a gradient elution for 10 min and a mobile phase composition of (A) methanol and (B) methanol/methyl tert-butyl ether/ethyl acetate (5:4:1) starting from 10% to 100% of eluent B, followed by 20 min of isocratic elution with 100% B. After that, the eluent composition was set back to 10% B for 3 min. The injection volume was 100 µL and the flow rate was kept constant at 1.4 mL/min.