All the strains used in this study are listed in Supplementary Table S1. E. coli JM109 (Sambrook, 1989) was used for plasmids construction and E. coli JM110 was used to eliminate the methylated plasmid in vivo (Palmer and Marinus, 1994). Luria Broth (LB) medium was used for E. coli culture at 37°C. When needed, 100 μg/mL of ampicillin was added to LB medium. H. mediterranei was cultivated at 37°C in nutrient-rich AS-168 medium (per liter, 5 g casamino acids, 5 g yeast extract, 1 g sodium glutamate, 3 g trisodium citrate, 2 g KCl, 20 g MgSO₄ ⋅ 7H₂O, 200 g NaCl, 5 mg FeSO₄ ⋅ 7H₂O, and 0.036 mg MnCl₂ ⋅ 4H₂O [pH 7.0]). AS-168SY medium was similar to AS-168 medium, except that yeast extract was excluded. Plasmids were transformed into H. mediterranei with the polyethylene glycol-mediated transformation method (Cline et al., 1989). When required, AS-168 medium was supplemented with 50 mg/mL uracil (Sangon, China) and 250 mg/mL 5-Fluoroorotic acid (Sangon, China) for counter-selection of the recombinants without pyrF marker. For lycopene production, a 1% (V/V) seed culture of H. mediterranei or its mutants was inoculated into a shake flask containing 50 mL of MG medium (per liter, 110 g NaCl, 20.51 g MgCl₂, 29.52 g MgSO₄, 5 g KCl, 1 g CaCl₂, 2 g NH₄Cl, 0.0375 g KH₂PO₄, 10 g glucose, 15 g PIPES, Fe(III) citrate, and 1 mL trace element solution SL-6 [pH 7.2]) (Antón et al., 1988) and cultured at 37°C and 200 rpm for 7 days.

The native candidate genes involved in lycopene synthesis were amplified with primers listed in Supplementary Table S2 from the H. mediterranei genomic DNA. The heterologous crt genes HAH_2563 (crtB_ha), HAH_1058 (crtI_ha) and OE_3093R (crtB_hs), OE_3381R (crtI_hs) were obtained via PCR from the genomic DNA of Haloarcula hispanica and Halobacterium salinarum, respectively. Amplified fragments were inserted into pWLR [derived from pWL502 by insertion of a strong promoter P_phaR (Cai et al., 2012)] digested with BamHI and XbaI, by using One Step Cloning Kit (Yeasen, Co., Ltd., China), to generate plasmids for gene overexpression under the control of promoter P_phaR (Supplementary Table S1).

All plasmids for gene knock-in or knock-out were constructed based on a suicide plasmid pHFX (Liu et al., 2011). A 583-bp DNA fragment located immediately upstream of crtB_hm was amplified with primer pair crtB-in-1/crtB-in-2 from H. mediterranei genomic DNA. Another 580-bp fragment containing promoter P_phaR and the partial 5′ region of crtB was amplified with primer pair crtB-in-3/crtB-in-4 from plasmid pW2547 used for crtB_hm overexpression (Supplementary Table S1). Then, the two PCR products were inserted into the plasmid pHFX to construct the integration plasmid of pHFXB, which was used to replace the native promoter of HFX_2547 (crtB_hm) in chromosome. For the heterologous crt gene integration, a 524-bp fragment up-stream of HFX_2549 and a 529-bp fragment down-stream of HFX_2549 were amplified by relevant primer pairs (Supplementary Table S2). Different crt gene fragments containing P_phaR were amplified by using crtB_ha, crtI_ha, crtB_hs, and crtI_hs overexpression plasmids pWHA2563, pWHA1058, pWOE3093, and pWOE3381 (Supplementary Table S1) as PCR template, respectively. The corresponding homologous arm fragments and different crt genes, containing promoter P_phaR, were assembled into pHFX to construct plasmids containing different crt genes used for their integration in chromosome (Supplementary Figure S1 and Supplementary Table S1). Similar to the plasmid construction method as described above, primer pairs HFX_2549-K1/HFX_2549-K2 and HFX_2549-K3/HFX_2549-K4 were used to construct the plasmid pHFX2549K for HFX_2549 knock-out. Primer pairs phaEC-K1/phaEC-K1 and phaEC-K3/phaEC-K4 were used for the construction of pHFXPK for phaEC knock-out (Supplementary Table S1).

Carotenoids in acetone extract obtained from different cultures were analyzed by thin layer chromatography (TLC), following the protocol described by Strand et al. (1997) with slight modifications. Briefly, after cultivation in MG medium for 7 days, the cells (1 mL) were harvested by centrifugation (12,000 × g for 5 min, at 4°C), and resuspended in acetone (1 mL) under a reduced light condition to prevent photo-bleaching and degradation (Alper and Stephanopoulos, 2008). The acetone supernatant containing carotenoids was collected and transferred to a new tube. This process was repeated until the pellets were totally white. Acetone extracted carotenoids (10 μL) were analyzed by TLC on a silica plate (GF254, Qingdao Haiyang Chemical, Co., Ltd., China) with acetone and n-heptane (1:1, v/v) as the development liquid in fume hood at room temperature. In addition, the visualized spot on the resulting TLC plate was scraped off, and extracted with 200 μL of acetone. The obtained supernatant was then scanned under 350–550 nm.

The lycopene content in the extract was determined by using a HPLC system (Agilent, 1260, United States) equipped with a ZORBAX Eclipse XDB-C18 column (4.6 mm × 150 mm, 5 μm) and a UV/VIS detector. The absorption was detected at 450 nm. The mobile phase consisted of methanol-isopropanol (65:35 V/V) with a flow rate of 1 mL/min at 30°C. Injection volume of sample was 20 μL. The lycopene concentration was calculated based on the calibration curve of lycopene (Macklin Biochemical, Co., Ltd., China).

The cells were collected by centrifugation at 10,000 × g, 4°C, 15 min and lyophilized. The lyophilized cells were treated with a mixture of chloroform and methanol containing 3% (v/v) sulfuric acid at 95°C for 4 h. The resulting hydroxyacyl methylesters were then analyzed by GC-6820 instrument (Agilent, United States) as described by Han et al. (2007).

The cells were cultured in AS-168 medium at 37°C for 12 h and subsequently harvested by centrifugation (12,000 × g, 4°C). The total RNA was extracted using TRIzol reagent (Invitrogen, United States) as previously described (Lu et al., 2008a). TURBO DNA-free^TM Kit (Thermo Fisher Scientific, United States) was used for removing DNA contamination. The cDNA was synthesized by reverse transcription with random hexamer primers from 1 μg of DNA-free total RNA using the Moloney Murine Leukemia Virus Reverse Transcriptase (M MLV-RT) (Promega, United States). The relative fold of gene expression was analyzed by ViiA^TM 7 Real-Time PCR System (Applied Biosystems, Inc., United States), using 7S RNA as an endogenous control to normalize the data of each sample. The primers used are listed in Supplementary Table S2.

The DNA sequences were obtained from National Center for Biotechnology Information (NCBI) Genome Database. The information about the most identified enzymes involved in MVA and lycopene synthesis pathway (supported by evidence at protein level), was accessed from UniProt Database¹. Sequence homology was assessed by BLASTN or BLASTP in NCBI (Altschul et al., 1990). Predictions of transmembrane helices in the proteins were performed by using the TMHMM Server v2.0² (Krogh et al., 2001).

The genome accession numbers deposited in GenBank are as following, H. mediterranei (CP001868.2), H. hispanica (NC_015948.1), and H. salinarum (AM774415.1).

Experiments were performed in triplicate and data was analyzed by the GraphPad Prism 7 software and represented as mean ± standard deviation. Statistical analysis was done using a two-tailed t-test. Statistical significance was defined as ^∗p < 0.05.

In silico metabolic pathway analysis reveals that H. mediterranei has a complete lycopene biosynthetic pathway, referring to the steps from isopentenyl-PP (IPP) and dimethylallyl-PP (DMAPP) to lycopene (Figure 2). Mevalonate (MVA) pathway provides the two important precursors, IPP and DMAPP, for lycopene synthesis. In MVA pathway, two molecules of acetyl-CoA are condensed to form acetoacetyl-CoA and a third acetyl-CoA molecule is then added to form 3-hydroxy-3-methylglutaryl-CoA (HMG-CoA) by Hydroxymethylglutaryl-CoA synthase (MvaB, HFX_2424). The next step involves the conversion of HMG-CoA to MVA by Hydroxymethylglutaryl-CoA reductase (HmgR, HFX_2609). MVA is then phosphorylated by Mevalonate kinase (Erg12, HFX_2773) to generate mevalonate phosphate (MVAP). Different from the classical MVA pathway, an alternative pathway is proposed for the IPP generation from MVAP, which is catalyzed by Diphosphomevalonate decarboxylase (DmD, HFX_1486) and Isopentenyl phosphate kinase (IpK, HFX_2774). IPP can be then isomerized to DMAPP by Isopentenyl-diphosphate delta-isomerase (IdI, HFX_2519). DMAPP and IPP are condensed to geranyl-PP (GPP) and then to farnesyl-PP (FPP) and finally to geranylgeranyl-PP (GGPP) by trifunctional prenyl diphosphate synthase (IdsA, HFX_2735). Two molecules of GGPPs are condensed by CrtB (HFX_2547) to form phytoene and then it undergoes four consecutive desaturation reactions catalyzed by CrtI (HFX_2550) to produce lycopene.

Although H. mediterranei possessed the complete lycopene synthetic pathway, we could not detect lycopene accumulation in this strain (Figure 3B). The extremely low lycopene production might be due to the rate-limiting steps in its lycopene biosynthesis. According to the previous studies about rate-limiting steps in MVA and lycopene synthesis pathway (Goldstein and Brown, 1990; Kang et al., 2005; Steussy et al., 2005; Anthony et al., 2009; Lombard and Moreira, 2011; Berthelot et al., 2012), we selected all the predicted genes as our candidates to be overexpressed in H. mediterranei (Figure 2 and Table 1). To rapidly identify the rate-limiting steps, we used a plasmid-based expression system for candidate genes overexpression under the control of a strong constitutive promoter P_phaR. The expression plasmid containing each gene was transformed into the strain DF50Δeps (Zhao et al., 2013) individually and correct transformants were confirmed by PCR and Sanger sequencing. Functional overexpression of the genes encoding rate-limiting enzymes made transformants orange or even red, so it was easy to identify the genes encoding rate-limiting enzymes. Obviously, only the overexpression of gene crtB_hm resulted in an orange colored phenotype (Figure 3A). Meanwhile the lycopene accumulation in this crtB_hm-overexpressed strain DF50-2547 was further confirmed by HPLC (Figure 3B), whereas no lycopene accumulation was detected in other gene overexpressed strains (Supplementary Figure S2). These results indicated that the step from GGPP to phytoene was the rate-limiting step in lycopene synthesis in H. mediterranei.

The plasmid-based overexpression of crtB_hm could reinforce the rate-limiting step and thus enhanced the production of lycopene in H. mediterranei. However, this plasmid-based system is not genetically stable and probably brings a metabolic burden. To address this issue, we therefore constructed the plasmid pHFXB (Supplementary Table S1) and used a two-step homologous recombination method to insert promoter P_phaR into the chromosome immediately up-stream of crtB_hm in DF50Δeps. The engineered strain, termed 50crtB, was easy to be distinguished visually, because of its orange color, which was different from the light pink color of its parental strain DF50Δeps (Figure 3C).

Next, we analyzed the transcription level of crtB_hm in strain 50crtB and DF50Δeps by qRT-PCR. The result showed that the insertion of promoter P_phaR dramatically increased the transcription level of crtB_hm by 245 times, when compared with the DF50Δeps strain (Figure 3C). As expected, the high CrtB_hm expression level significantly promoted the conversion from GGPP to phytoene and subsequently improved lycopene biosynthesis. After cultivation in shake flasks containing 50 mL MG medium for 7 days, a lycopene production of 6.05 ± 0.18 mg/g dry cell weight (DCW) (35.15 ± 0.43 mg/L) was achieved (Figure 4C).

Lycopene is the last shared intermediate in bacterioruberin and retinal biosynthesis in some haloarchaea (Peck et al., 2017). However, H. mediterranei lacks the genes, crtY (encoding lycopene cyclase) (Peck et al., 2002), brp and blh (encoding β-carotene dioxygenase) (Peck et al., 2001), involved in the retinal biosynthesis pathway and therefore, lycopene can only flux into bacterioruberin biosynthesis (Figure 1A). Thus, we next blocked bacterioruberin biosynthesis to enhance lycopene accumulation and its purity. Bioinformatic analysis revealed the presence of two genes (HFX_1501 and HFX_2549) potentially involved in bacterioruberin synthesis in H. mediterranei. Both of them are annotated as putative prenyltransferases, which can transfer 5-carbon prenyl groups to various substrates. HFX_1501 encodes a protein of 284 amino acid residues, which showed 30 and 28% identity to the LyeJ of H. salinarum and H. japonica, respectively (Supplementary Figure S3A). In the case of HFX_2549, it encodes a 292-amino acid protein exhibiting 64 and 61% identity to the LyeJ of H. salinarum and H. japonica, respectively (Supplementary Figure S3B). Moreover, it is located within the carotenoid biosynthetic gene cluster. Furthermore, a membrane topology analysis using TMHMM revealed that HFX_2549 encoded an integral membrane protein containing seven transmembrane domains (Supplementary Figure S3B), consistent with the LyeJ of H. salinarum or H. japonica. This suggested that HFX_2549 is likely to encode LyeJ in H. mediterranei.

We knocked out the gene, HFX_2549, and obtained the strain 50BΔ2549. The carotenoid components of the 50crtB and 50BΔ2549 strains were analyzed qualitatively by TLC. The sample from 50crtB displayed multiple spots on the silica plate (Figure 4A, Spots 1–4), which represented lycopene, bacterioruberin and its derivates. In contrast, the sample from 50BΔ2549 only contained a single spot 1 (Figure 4A, Spot 1). Subsequently, spot 1 on the silica plate from the 50BΔ2549 strain was recovered and extracted by using acetone. The UV-Vis absorbance spectrum of the extracted sample had a typical three-finger shape of lycopene at 442, 470 and 501 nm, similar to the absorption spectrum of lycopene standard (Figure 4B). These results indicated that 50BΔ2549 could not synthesize bacterioruberin and its derivates, and the reaction was terminated at the lycopene step. Consequently, HFX_2549 was the key enzyme for bacterioruberin synthesis in H. mediterranei. Lycopene production of 50BΔ2549 was further quantified by HPLC. The production significantly increased to 45.54 ± 1.23 mg/g DCW, which was about 6.5 times higher than that of 50crtB (Figure 4C).

To improve the yield of the target products, it is often necessary to enhance the availability of essential precursors (Xie et al., 2015a). In our study, overexpression of crtB_hm supplied more precursor, phytoene, for lycopene synthesis, and enhancement of the flux from phytoene to lycopene is another strategy for lycopene production improvement. Here, we strengthened the process by importing heterologous gene crtI_ha or crtI_hs from carotenogenic haloarchaea, H. hispanica or H. salinarum. First, we confirmed the function of crtI_ha and crtI_hs by transforming their expression plasmid pWHA1058 or pWOE3381 (Supplementary Table S1) into the 50BΔ2549 strain. The positive transformants with the functional expression of heterologous crtI showed enhanced color intensity (Figure 5A). Afterward, we constructed two plasmids, pHI6 and pHIH (Supplementary Table S1), for the integration of crtI_ha and crtI_hs into the chromosome of H. mediterranei (Supplementary Figure S1A). To avoid unpredictable effects brought by the insertion site, we used 50crtB as the host and integrated these two genes separately in the chromosome by replacing HFX_2549 (Figure 1B). The two resultant strains were named 50BI6 and 50BIH (Supplementary Table S1).

Lycopene production by the engineered strains were then determined by HPLC. Lycopene yields of 56.46 ± 0.74 mg/g DCW (292.21 ± 2.88 mg/L) and 49.37 ± 2.95 mg/g DCW (271.43 ± 6.07 mg/L) were obtained in the engineered strains, 50BI6 and 50BIH, respectively. This result showed that the heterologous expression of crtI_ha in H. mediterranei was more effective for enhancing lycopene synthesis compared to crtI_hs. Finally, 50BI6 got a 24.0% increase in lycopene production and 10.2% increase in lycopene titer compared to 50BΔ2549 (Figure 5C).

The high lycopene production strain 50BI6 contained two copies of crtI (crtI_hm and crtI_ha). To investigate whether CrtB was still the rate-limiting enzyme in 50BI6, another copy of crtB from H. hispanica or H. salinarum was introduced into 50BI6. We constructed two types of crtB-crtI expression cassettes, crtB_ha-crtI_ha and crtB_hs-crtI_ha (Figure 5B) and inserted these expression cassettes in the chromosome of 50crtB by replacing HFX_2549 to obtain new strains, 50B6I6 and 50BHI6 (Figure 1B and Supplementary Figure S1B). We analyzed the transcriptional status of crtB_ha, crtB_hs, and crtI_ha in these two strains and found that all the expected genes were successfully transcribed base on the RT-PCR analysis (Supplementary Figure S4). The lycopene yields of 50B6I6 and 50BHI6 were further increased to 68.95 ± 1.19 mg/g DCW (353.06 ± 2.39 mg/L) and 60.33 ± 1.56 mg/g DCW (297.05 ± 6.50 mg/L), respectively (Figure 5C). This suggested that the strains with co-introduced crtB_ha and crtI_ha could produce more lycopene. Lycopene titer displayed a 20.8% increase in 50B6I6 compared to 50BI6.

H. mediterranei can accumulate a large amount of PHBV when cultured in MG medium (Zhao et al., 2013) and acetyl-CoA is an important precursor for its biosynthesis (Don et al., 2006). Blocking acetyl-CoA flux to PHBV biosynthesis may be able to further maximize the lycopene production (Figure 1B). To prove this, we knocked out the PHBV synthase encoding genes, phaE and phaC (Lu et al., 2008b), in 50B6I6 and obtained a new strain, named 50B6I6ΔphaEC (Supplementary Table S1). It could not produce PHBV as determined by gas chromatography (Table 2) and this result suggested that the deletion of phaEC blocked PHBV synthesis. Moreover, HPLC analysis showed that 50B6I6ΔphaEC synthesized lycopene with a production level of 119.25 ± 0.55 mg/g DCW, which was 73.0% higher than that of 50B6I6. Meanwhile, lycopene titer of 50B6I6ΔphaEC increased to 429.41 ± 5.81 mg/L, which was 21% enhancement compared to 50B6I6. This result indicated that the disruption of PHBV biosynthesis could enhance the acetyl-CoA flux to lycopene biosynthesis.

In this work, all the genetic manipulation was based on the pyrF-deleted strain DF50Δeps (uracil auxotrophic mutant). Uracil addition was required to culture these engineered strains in MG medium. However, this approach was not suitable for high-cell density fermentation and increased the production cost. To address this issue, we restored the functional expression of pyrF in 50B6I6ΔphaEC. A linear DNA fragment containing pyrF and homologous arms, 500 bp in up-stream or down-stream of pyrF, was amplified by PCR using the genomic DNA of H. mediterranei as a template and transferred into the 50B6I6ΔphaEC strain. The screening process was carried out using AS-168SY medium, in which the negative colonies could not grow. We obtained the correct pyrF complementary strain 50FB6I6ΔphaEC and it gave a lycopene yield of 107.37 ± 2.37 mg/g DCW (396.70 ± 13.39 mg/L), while there was no difference in biomass between 50B6I6ΔphaEC and 50FB6I6ΔphaEC (Figure 6). Although the complementation of pyrF did not increase the biomass and led to a little decrease in lycopene content, 50FB6I6ΔphaEC could be cultured in MG medium without uracil, which was more feasible for industrial application.