Ten banana accessions were selected from the field germplasm plot at the National Agri-Food Biotechnology Institute (NABI), Mohali for the carotenoids estimation in ripe fruit-pulp (Table 1). The fruit sampling was performed during the summer season. Plant tissues such as pseudostem (outer, and inner layers), leaf (young, and mature stages), bract, fruit finger, fruit-peel and -pulp (unripe, and ripe stages) were collected from the contrasting cultivars, frozen in liquid nitrogen, and stored at −80°C until further use. For abiotic stress treatments, in-vitro grown Rasthali plants were exposed with ABA (100 μM), MeJA (250 μM) and SA (250 μM), as described previously (Sreedharan et al., 2012; Pandey et al., 2016). The control and treated plants were grown in plant growth chambers (Conviron, Canada) at 26°C under 16/8-h light (200 μmol photon m⁻² s⁻¹)/dark period with 60 ± 5% relative humidity (RH). To study the effect of dark condition, plants were kept under the dark for 72 h in a plant growth chamber (Conviron, Canada) at 26°C with 60 ± 5% RH. Leaves were collected at different time intervals, frozen immediately in the liquid nitrogen, and stored at −80°C till further use. All biochemical, standards, kits and reagents unless otherwise stated were molecular biology or cell culture grade from Sigma Chemical Company (St. Louis, MO) and Merck (India).

Carotenoid extraction was performed as described previously (Matsumoto et al., 2007; Flowerika et al., 2016) with some modifications. The lyophilized 1 g dry weight (DW) sample was homogenized in 50 ml solution containing 40% (v/v) methanol + 0.5 g magnesium carbonate and centrifuged. The insoluble fraction was extracted in diethyl ether and methanol (7:3, v/v) containing 0.1% (w/v) butylated hydroxy toluene (BHT) till the extract turned colorless. The extract was transferred in a separating funnel and 20 ml diethyl ether was added. The ether phase was washed three-times with 10 ml of saturated sodium chloride and 5 ml of 10% anhydrous sodium sulfate solution. The sample was evaporated using Rota vapor (Buchi Labortechnik AG) at 50°C at 100 rpm and 200 atmospheric pressures. The residue was dissolved in methyl tert-butyl ether (MTBE)-methanol (1:1, v/v) containing 0.1% (w/v) BHT, and centrifuged. The supernatant was filtered through 0.45-μm nylon membrane filter (Millipore, MA) and was used for High Performance Liquid Chromatography (HPLC). Analysis was performed in a liquid chromatograph binary gradient module pumps (Waters, Milford, MA, USA) equipped with photodiode array (Waters, 2998) and autosampler (Waters, 2767). Carotenoids were separated by using YMC C₃₀ column (YMC, Kyoto, Japan). The mobile phase was a gradient, prepared from 95% (v/v) methanol in HPLC-grade water (component A) and MTBE (component B). Carotenoid standards (β-carotene, lutein, zeaxanthin and lycopene) were used for calibration curve, comparison of retention time, and quantification of samples. Data was analyzed using MassLynx™ software and quantification was carried out at 450 nm.

To determine antioxidant activity, carotenoid extract was prepared from pulp tissue of Rasthali and Nendran as per procedure described above. Measurement of radical scavenging activity was carried out as per earlier reported 2,2-diphenyl-1-picrylhydrazyl (DPPH) method (Blois, 1958; Awasthi et al., 2016). Briefly, freshly prepared DPPH methanolic solution was added to 200 μl of methanol having different concentration of carotenoid extract. Reaction was incubated in dark for 30 min at room temperature. The absorbance was measured at 517 nm (UV-2700 UV-VIS Spectrophotometer, Shimadzu). Percent inhibition was calculated according to the following equation: % Inhibition=[(AbDPPH∗−AbS∗∗)/AbDPPH∗]×100,

^*Ab_DPPH: Absorbance of DPPH solution at 517 nm,

Experiment was performed in three replicates and ascorbic acid was used as positive control. IC₅₀ values (half maximal inhibitory concentration) were calculated as the concentration of extracts causing 50% inhibition of DPPH radical. Lower IC₅₀ value corresponds to a higher antioxidant activity of extract.

The whole genome sequence of banana (Musa acuminata) was downloaded from the Banana Genome Hub database (D'Hont et al., 2012; http://banana-genome.cirad.fr/) and used for the identification of PSY gene. BLAST search was performed using annotated PSY sequences of maize (Zea mays), rice (Oryza sativa) and arabidopsis (Arabidopsis thaliana) to explore similar sequences in the banana genome database. The reported PSY gene sequences of banana were also considered in BLAST analysis. The most similar sequences were retrieved from the banana genome database and designated as putative MaPSY genes. The percentage identity of the predicted protein sequences was noted. Selected coding (CDS) and genomic DNA sequences were aligned to identify the gene structures, and numbers of exons and introns in the sequences. Theoretical isoelectric point and molecular weight were predicted using the Compute pI/MW tool on the ExPASy server (Bjellqvist et al., 1994; http://web.expasy.org/compute_pi/). Transmembrane domains were analyzed on the TMHMM server v 2.0 (Krogh et al., 2001; http://www.cbs.dtu.dk/services/TMHMM-2.0/) and subcellular localization of deduced proteins predicted on the TargetP 1.1 server (Emanuelsson et al., 2007; http://www.cbs.dtu.dk/services/TargetP/). The presence of chloroplast transit peptides was predicted using the ChloroP 1.1 server (Emanuelsson et al., 1999). Mapping of the biosynthetic genes on banana chromosome was done by using the locus ID information available on Banana Genome Hub (http://banana-genome.cirad.fr/). The upstream regions (1.5 kbp) of all selected genes were extracted through a custom Perl script for the identification of cis-regulatory elements. This region was considered as the proximal promoter for the individual gene. The promoter sequences were placed in the Plant Care Database (Lescot et al., 2002) in which a brief description of all motifs were extracted and presented.

Total RNA was isolated from different tissues (pseudostem, leaf, bract, fruit finger, fruit-peel, and -pulp), and treated samples (ABA, MeJA, SA, and dark) according to previously described protocol (Asif et al., 2006). RNA sample was treated with DNase I Digest kit (Sigma-Aldrich, USA) to eliminate DNA contamination. The cDNA synthesis was performed using Revert Aid First Strand cDNA synthesis Kit (Fermentas, USA) from 3 μg of DNA-free total RNA following the manufacturer's instructions.

The quantitative real-time PCR expression was carried out by following SYBR Green chemistry at ABI PRISM 7700 Fast Real-Time PCR System (Applied Biosystems, USA). The invariant expression under the described experimental conditions was normalized by using two banana housekeeping genes, actin1 (GenBank Accession No. AF246288) and ubiquitin2 (GenBank Accession No. HQ853254). The relative fold expression was calculated by using 2^−ΔΔCT method (Schmittgen and Livak, 2008). The list of different primers used in the study is given in Supplementary Table S1.

Three MaPSY sequences retrieved from Musa acuminata reference genome were used for designing the gene specific primers (Supplementary Table S1). The full-length open reading frames (ORFs) of PSY were amplified from the ripe fruit-pulp of contrasting cultivars (Nendran and Rasthali). The genes isolated from Nendran and Rasthali are designated as NEN-PSY1, NEN-PSY2, NEN-PSY3, and RAS-PSY1, RAS-PSY2, RAS-PSY3 respectively. The amplicons containing BglII and NotI restriction sites were cloned in pBluescript SK+ vector (Stratagene, USA) and confirmed by sequencing. The conserved motifs and sites for PSY functional activity were determined by using NCBI CD-Search (Marchler-Bauer et al., 2011; http://www.ncbi.nlm.nih.gov/Structure/bwrpsb/bwrpsb.cgi) and ScanProsite (Hulo et al., 2008; http://prosite.expasy.org/scanprosite) tools. PSY protein sequence annotation analysis was accelerated using BioEdit sequence alignment editor (Hall, 1999; http://www.mbio.ncsu.edu/bioedit/bioedit). Phylogenetic analysis was conducted in MEGA6 (Tamura et al., 2013) for deduced protein sequences of arabidopsis, rice, maize and banana. The PSY sequences derived from Asupina (high provitamin A) and Cavendish (low provitamin A) banana cultivars (Mlalazi et al., 2012) were also considered in the study. Neighbor-joining method was used to infer homologs and evolution. Bootstrap value was set to 1000.

A heterologous bacterial complementation system was used to investigate the biological function of six MaPSY proteins derived from Nendran (NEN-PSY1, NEN-PSY2, NEN-PSY3), and Rasthali (RAS-PSY1, RAS-PSY2, and RAS-PSY3). The plasmids pAC-85b and pTrc were used to test the functional activity of PSY genes (Cunningham and Gantt, 2007; Fu et al., 2014). The E. coli cells transformed with these plasmids could not be able to synthesize carotenoid which resulted white colonies. The PCR products of six MaPSY genes comprising BglII and NotI restriction sites (Supplementary Table S1) were cloned into digested pTrc plasmid and named as pTrc-NEN-PSY1, pTrc-NEN-PSY2, pTrc-NEN-PSY3, pTrc-RAS-PSY1, pTrc-RAS-PSY2, and pTrc-RAS-PSY3. The competent E. coli cells were co-transformed with pAC-85b and pTrc plasmid having PSY genes. A single positive colony was inoculated in Luria Broth (LB) medium containing ampicillin (100 μg/ml) and chloramphenicol (50 μg/ml) antibiotics and grown at 37°C with 200 rpm. The relative activity of MaPSY proteins could be observed as variation in the color of cultures due to differential accumulation of carotenoid. The β-carotene was extracted from the bacterial cell pellets (0.6 OD) and analyzed by HPLC by the method described earlier (Matsumoto et al., 2007; Flowerika et al., 2016). The extractions were repeated at least three times from independent transformed bacterial cell cultures.

All experiments were repeated with three biological replicates and each experiment consisted of three technical replicates. The data are presented as mean ± SD and analyzed by Student's paired t-test. The mean values under each treatment were compared to determine significance (^*P ≤ 0.05; ^**P ≤ 0.001; ^***P ≤ 0.0001). The relationship between gene expression and individual carotenoid accumulation was calculated by Pearson correlation (r) analysis. R² values were considered significant (^*P ≤ 0.05; ^**P ≤ 0.005; ^***P ≤ 0.001) correlations. All statistical analyses were performed using Prism GraphPad software Ver. 5.01 (GraphPad Software Inc., San Diego, CA, USA).

Fruit-pulp collected at the ripening stage of 10 Indian banana cultivars were lyophilized and analyzed for carotenoids. The highest β-Carotene Equivalents (β-carotene + ½ α-carotene) was observed in Nendran (3700.57 μg/100 g DW) while the lowest was found in Rasthali (170.05 μg/100 g DW) (Table 1). The ripe pulp of Rasthali appeared creamy-white and Nendran was deep-yellow in color (Figure 1A). Nendran and Rasthali cultivars were studied for changes in carotenoids in fruit-peel and -pulp as collected at unripe and ripe stages (Figure 1B). Lutein, α-carotene and β-carotene were quantified, whereas lycopene and zeaxanthin were not present in detectable amounts in the tissues. In the pulp tissue, the higher content of all the carotenoids was noticed at the ripening stage. Nearly 28-fold higher β-carotene was present in Nendran (3011.94 ± 176.70 μg/100 g DW) as compared to Rasthali (105.35 ± 7.35 μg/100 g DW) (Figure 1B). Similarly, α-carotene was nearly 10-fold higher in Nendran as compared to Rasthali. Lutein was nearly 2-fold higher in Rasthali as compared to Nendran. Unripe pulp of both Nendran and Rasthali showed significantly low quantities of all carotenoids, compared to the ripe pulp (Figure 1B, Supplementary Figure S2). However, in the peel as compared to the pulp of both the cultivars, higher deposition of all carotenoids (except at the ripening stage of Nendran) was noticed (Figure 1B, Supplementary Figure S3). The greater deposition of α-carotene and β-carotene was found in ripe pulp of Nendran as compared to its ripe peel (Figure 1B).

In the present study, Nendran unripe/ripe pulp extracts showed the lower IC₅₀ value in comparison to Rasthali unripe/ripe extract (Table 4). Ripe pulp extract of both Nendran and Rasthali had lower IC₅₀ value as compared to the unripe pulp extract. The IC₅₀ values for the extracts of unripe and ripe pulp of Nendran and Rasthali are given in Table 4. The IC₅₀ value of all the extracts was lower than that of ascorbic acid (5.66 ± 0.86), which was used as positive control in the experiment.

Three homologs of PSY (MaPSY1-3) were identified in Banana Genome Hub database using known proteins from arabidopsis, maize and rice. Percentage identity of the predicted protein sequences of the banana PSY genes showed high homology with those in arabidopsis, maize and rice varies from 55 to 80% (Supplementary Table S2). Structural analysis of exons and introns of PSY from banana and other species is presented in Supplementary Table S3. The PSY homologs from banana, maize and rice had same number of exons (6) and introns (5) however, arabidopsis PSY was found to be without intron. The accession numbers of the genes, their positions on banana chromosome, predicted polypeptide length, isoelectric point, molecular mass, subcellular localization and transmembrane helices (TMH) are given in Supplementary Table S4. MaPSY1 and MaPSY2 were located on the chromosomes 6 and 9, respectively, while MaPSY3 was present on the un_random chromosome (uncharacterized scaffold). The MaPSY1 and MaPSY2 proteins had similar length of 397 amino acids whereas MaPSY3 was larger in length and contained 426 amino acids. The isoelectric point (9.38), and molecular weight (47.76 kDa) of MaPSY3 was higher than MaPSY1 (9.22, and 45.03 kDa), followed by MaPSY2 (9.12, and 44.35 kDa). All the MaPSY showed to be localized in chloroplast. MaPSY1 has shown the presence of transmembrane helix.

Expression analysis of three PSY homologs (MaPSY1, MaPSY2, and MaPSY3) in different tissues of contrasting cultivars (Rasthali and Nendran) was performed (Figures 2, 3). The normalization of transcripts was carried out against actin1 gene that showed most stable expression in various tissues. In general, the expression pattern of MaPSY homologs was similar in both cultivars, but they expressed at varying levels in different tissues. In the pseudostem of both cultivars, higher expression of MaPSY3 was detected in inner and outer layer (Figures 2A,B). The lower expression of MaPSY1 and MaPSY2 was observed in inner and outer layer of both cultivars, respectively. Two leaf developmental stages i.e., young and mature leaves were taken for the analysis (Figures 2A,B). The higher transcript level of MaPSY3 and MaPSY1 was noted in young leaf of Nendran and Rasthali, respectively. Mature leaf of both cultivars has shown higher expression of MaPSY1. The lower expression of MaPSY2 was observed in young and mature leaf of both cultivars. Bract and fruit fingers of both cultivars showed higher expression of MaPSY3. The lower expression of MaPSY2 and MaPSY1 was noted in bracts and fruit fingers of both cultivars, respectively (Figures 2A,B). These results suggest differential expression of MaPSY genes among different tissues as well as between the contrasting cultivars. The analysis indicated that all MaPSY gene homologs were highly expressed in leaf tissue of both cultivars (Figures 2A,B).

To understand the transcriptional regulation in the carotenogenesis in fruit tissues of two contrasting cultivars, expression of MaPSY homologs was studied in fruit-pulp (Figure 3A) and -peel (Figure 3B) at two developmental stages (unripe and ripe). The homologs were expressed in all fruit tissues of both cultivars, except MaPSY2. The transcript of MaPSY2 was not detected in unripe peel of Rasthali. In the unripe pulp, higher expression of MaPSY1 and MaPSY2 was noticed in Rasthali and Nendran, respectively. Whereas in ripe pulp, high transcript level of MaPSY2 and MaPSY1 was observed in Rasthali and Nendran, respectively (Figure 3A). In the peel at unripe stage, higher expression of MaPSY3 was recorded in Rasthali as well as in Nendran (Figure 3B). The higher expression of MaPSY1 was observed at the ripe stage of peel in both cultivars.

Pearson correlation analysis was carried out between relative fold expression of MaPSY transcripts and individual carotenoid content in fruit-pulp (Table 2) and -peel (Table 3) in the contrasting cultivars. In the Nendran pulp, significant positive correlation was observed for NEN-PSY1 (r = 0.9995, P < 0.05) and NEN-PSY2 (r = 0.9978, P < 0.05) with β-carotene and α-carotene, respectively. No significant correlation was noted for lutein in Nendran pulp. In Rasthali pulp, significant positive correlation was noted for RAS-PSY2 (r = 1.00, P < 0.005) and RAS-PSY1 (r = 0.9973, P < 0.05) with β-carotene. RAS-PSY2 (r = 0.9969, P < 0.05) showed significant positive correlation with α-carotene. Lutein accumulation in Rasthali showed significant positive correlation with both RAS-PSY1 (r = 0.9983, P < 0.05) and RAS-PSY2 (r = 0.9998, P < 0.05). In the Nendran peel, significant positive correlation was established for NEN-PSY1 (r = 1.00, P < 0.005) with β-carotene whereas, negative correlation was observed for NEN-PSY1 (r = −0.9997, P < 0.05) with lutein. No correlation was observed for α-carotene in Nendran. In the Rasthali peel, significant positive correlation was established for RAS-PSY1 with lutein (r = 0.9994, P < 0.05). However, with β-carotene significant negative correlation was noted for RAS-PSY2 (r = −0.999, P < 0.05). Besides, no significant correlation (r = 0.8262 to 0.9945, P ≥ 0.1031 to ≤ 0.381) was found between carotenoids and PSY3 homolog of both cultivars in pulp and peel (Table 2).

Various cis-regulatory elements were detected in the proximal promoter regions of three MaPSY genes (Supplementary Table S5). The cis-elements such as Skn-1 motif related to growth and development, Box4, and Sp1 motifs related to light response were present in all three MaPSY genes. The light responsive elements were mostly found in MaPSY1 and MaPSY3 while higher number stress responsive motifs were detected in MaPSY1 and MaPSY2, which suggested their possible role in these conditions.

To understand the effect of various environmental factors on transcriptional regulation of MaPSY genes, expression analysis was performed following different treatments in Rasthali plants. The differential expression of all three MaPSY genes was observed during the dark exposure to the plants. The analysis reveals that the dark exposure leads to significantly down-regulated expression of MaPSY1 and MaPSY3 genes (Figure 4A). Interestingly, nearly 26- and 7.5- fold up-regulated expression of MaPSY2 was noted after 12 and 48 h, respectively of incubation under the dark as compared to control (0 h) plants (Figure 4A). The expression of MaPSY2 was comparable to control (0 h) at 72 h. The expression patterns suggest a positive correlation with the presence of high number light-responsive cis-elements in MaPSY1 and MaPSY3 promoter regions.

The exogenous applications of ABA, SA and MeJA have modulated expression of MaPSY genes (Figures 4B–D). ABA exposure led to the significant enhancement in the expression of all three MaPSY genes at 4 h, but the differential modulation was noted for different genes at different time points (Figure 4B). MaPSY1 and MaPSY2 showed a significant up-regulation in the expression at 4 h which continued till 72 h. In case of MaPSY3, biphasic expression pattern was observed (Figure 4B). ABRE motif responsible for ABA response was solely present in the promoter region of MaPSY2 gene.

Exposure of SA led to the significant up-regulation in expression of MaPSY1 at 2 h and continued till 24 h. On the other hand, MaPSY2 transcript was found to be up-regulated at 2 h and then reduced to comparable levels of control (Figure 4C). MaPSY3 showed significant down-regulation in expression at different time points after treatment. Interestingly, TCA element that is responsible for SA response has found in three copies at the proximal promoter region of MaPSY1 whereas it was absent in MaPSY2 promoter. A single copy of TCA element was present in MaPSY3 promoter region.

The exogenous exposure of MeJA showed nearly 2-fold up-regulation in expression of MaPSY1 at 2 h which continued till 24 h (Figure 4D). MaPSY2 expression was up-regulated at 2 h and afterwards expression was comparable to the control (Figure 4D). The significant down-regulation in expression of MaPSY3 was found at different time points as compare to control. MeJA responsive elements (CGTCA and TGACG motifs) were present in the proximal promoter regions of MaPSY1 and MaPSY2, but absent in MaPSY3. These results suggest that MaPSY genes expression in response to different treatments largely correlates with the presence of cis-regulatory elements in their promoter regions.

Full-length gene sequences of Nendran (NEN-PSY1, NEN-PSY2, and NEN-PSY3), and Rasthali (RAS-PSY1, RAS-PSY2, and RAS-PSY3) were amplified for characterization. The sequencing analysis revealed that all the homologs of PSY in both the cultivars shares considerable (≥95%) homology at both nucleotide and protein levels. The phylogenetic relationship of MaPSY proteins with known banana and other plant species was constructed using neighbor joining method. Three MaPSY identified from each of Rasthali (RAS-PSY1, RAS-PSY2, and RAS-PSY3), and Nendran (NEN-PSY1, NEN-PSY2, and NEN-PSY3) showed general trend of clustering with other monocots (Figure 5). MaPSY homologs shared three distinct clades with other known banana PSY sequences. RAS-PSY1 and NEN-PSY1 were grouped into a clade with PSY2 proteins of Asupina (APSY2b) and Ladyfinger (LfPSY2). On the other hand, RAS-PSY2 and NEN-PSY2 grouped in a clade with PSY2a proteins of Asupina and Cavendish banana varieties. RAS-PSY3 and NEN-PSY3 were clustered in a different clade with PSY1 proteins of Asupina, Cavendish and Ladyfinger.

The deduced amino acid sequences of Nendran (NEN-PSY1-3) and Rasthali (RAS-PSY1-3) were compared to determine if differences in sequences could account for the differences in fruit carotenoids content between these two cultivars of banana (Figure 6). All PSY proteins contain a predicted chloroplast transient peptide at the N-terminal region. Conserved domain analysis revealed that all the six MaPSY proteins belong to class 1 superfamily of isoprenoid biosynthesis enzymes containing the conserved trans-isoprenyl diphosphate synthases, head-to-head (trans-IPPS_HH) domain. This domain has the binding sites for substrate and cofactor that catalyzes the production of phytoene by condensation of two molecules of GGPP (Geranyl geranyl pyrophosphate). ScanProsite analysis showed the presence of squalene/phytoene synthase signatures 1 and 2 (SQS-PSY 1 and 2) in all MaPSY proteins for PSY activity. The SQS-PSY1 motif (YCyyVAGTVGlmSvpV) and aspartate-rich motifs (ARM) (DELVD and DVGED) were found conserved in each PSY protein of Nendran and Rasthali (Figure 6). This observation was further confirmed by alignments with PSY sequences of several other plant species (Supplementary Figure S4). We noted that PSY1 of both the cultivars were 95% similar and differed by a total of 19 amino acid residues which included 4 in the transit peptide and 15 in the mature protein. Among these amino acid substitutions, four located in evolutionarily conserved domains of the mature protein. The substitutions from threonine to asparagine (N149T) and tyrosine to cysteine (C398Y) in the active site lid motifs (YAKTF and RAYV), valine to isoleucine (I292V) in the SQS-PSY2 (LGianQltNIlRDVgeDarrgRiYlP), and the tyrosine to aspartic acid (D169Y) in the GGPP substrate binding pocket were detected in the RAS-PSY1 (Figure 6). PSY2 and PSY3 of both the cultivars showed nearly 99% similarities and differed by 5 and 6 amino acids, respectively. The alanine (A307) and isoleucine (I312) residues were present in SQS-PSY2 motif of the PSY1 while these were changed by serine (S307A) and valine (V312I) in PSY2 of both the cultivars. The changes of lysine to glutamine (Q148K) in active site lid-motif and isoleucine to leucine (L292I) in SQS-PSY2 motif were found in the NEN-PSY3 protein (Figure 6).

To determine whether the six MaPSY (NEN-PSY1, NEN-PSY2, NEN-PSY3, RAS-PSY1, RAS-PSY2, and RAS-PSY3) proteins have functional activity, the ORF of each gene was cloned into the pTrc plasmid and co-transformed along with the pAC-85b into E. coli TOP10F. The visual observation suggested that co-transformed cells accumulated intense yellow color as compared to the control vector pAC-85b (no color) (Figure 7A), indicating that all six MaPSY accelerate the accumulation of β-carotene but with different levels of enzymatic activity. The pigments accumulated in the transformed E. coli cells were extracted and analyzed by HPLC. The expected product, β-carotene was confirmed by matching spectra and retention time. The highest β-carotene content was found in the bacteria transformed with pTrc-NEN-PSY1 (410 ± 15.45 μg/100 ml culture) while the lowest was recorded with pTrc-RAS-PSY3 (187 ± 9.59 μg/100 ml culture) (Figure 7B). The β-carotene was also not detected in individually transformed pTrc-NEN-PSY1–3 and pTrc-RAS-PSY1–3 plasmid controls. These results established that all six MaPSY proteins were enzymatically functional and could accelerate the biosynthesis of β-carotene at different levels.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.