Mutation of YFT3, an isomerase in the isoprenoid biosynthetic pathway, impairs its catalytic activity and carotenoid accumulation in tomato fruit

Abstract Tomato fruit colors are directly associated with their appearance quality and nutritional value. However, tomato fruit color formation is an intricate biological process that remains elusive. In this work we characterized a tomato yellow fruited tomato 3 (yft3, e9292, Solanum lycopersicum) mutant with yellow fruits. By the map-based cloning approach, we identified a transversion mutation (A2117C) in the YFT3 gene encoding a putative isopentenyl diphosphate isomerase (SlIDI1) enzyme, which may function in the isoprenoid biosynthetic pathway by catalyzing conversion between isopentenyl pyrophosphate (IPP) and dimethylallyl pyrophosphate (DMAPP). The mutated YFT3 (A2117C) (designated YFT3 allele) and the YFT3 genes did not show expression difference at protein level, and their encoded YFT3 allelic (S126R) and YFT3 proteins were both localized in plastids. However, the transcript levels of eight genes (DXR, DXS, HDR, PSY1, CRTISO, CYCB, CYP97A, and NCED) associated with carotenoid synthesis were upregulated in fruits of both yft3 and YFT3 knockout (YFT3-KO) lines at 35 and 47 days post-anthesis compared with the red-fruit tomato cultivar (M82). In vitro and in vivo biochemical analyses indicated that YFT3 (S126R) possessed much lower enzymatic activities than the YFT3 protein, indicating that the S126R mutation can impair YFT3 activity. Molecular docking analysis showed that the YFT3 allele has higher ability to recruit isopentenyl pyrophosphate (IPP), but abolishes attachment of the Mg2+ cofactor to IPP, suggesting that Ser126 is a critical residue for YTF3 biochemical and physiological functions. As a result, the yft3 mutant tomato line has low carotenoid accumulation and abnormal chromoplast development, which results in yellow ripe fruits. This study provides new insights into molecular mechanisms of tomato fruit color formation and development.


Introduction
The coloration of many ripe f leshy fruits is determined by the composition and accumulation of carotenoid pigments with 40carbon hydrocarbons, which can be categorized as carotenes and xanthophylls [1], in addition to yellow, orange, and red isoprenoid pigments [2].Carotenoids promote light harvesting as accessory pigments in photosynthesis, protect the photosynthetic apparatus from photo-oxidative damage, and accumulate in f lowers and fruits to attract animals and insects, thereby facilitating the dispersal of pollen and seeds [1,[3][4][5].They are also precursors of phytohormones such as abscisic acid (ABA), strigolactones (SLs), and other signaling molecules that are important for development and stress responses [2,6].Carotenoids and their derivatives are antioxidants and essential components of the human diet.For example, β-carotene, α-carotene, and zeaxanthin serve as the precursors for vitamin A biosynthesis.All-trans-lycopene has been associated with a reduced risk of cancer and cardiovascular disease [1][2][3]6].In addition to their functions as pigments and nutrients, carotenoids are precursors of volatile organic compounds released during fruit ripening [2,[7][8][9], conferring aromas attracting consumers [1,10].
In the current study, we revealed the genetic basis of yellow coloration in the fruit of the yellow fruited tomato 3 (yft3) tomato mutant.The single recessive yft3 gene was successfully isolated using map-based cloning, and found to encode an isopentenyl diphosphate (IDI) enzyme.Carotenoids are a class of isoprene derivatives/isoprenoids, which are derived from C5 building blocks,IPP, and its isomer, dimethylallyl diphosphate (DMAPP).The interconversion between IPP and DMAPP is dependent on IDI [31,32].Some bacteria, as well as all vascular plants, have two distinct pathways for producing IPP and DMAPP, the cytoplasmic mevalonic acid (MVA) pathway and the plastidial 2-Cmethyl-d-erythritol-4-phosphate (MEP) pathway [31,33,34].The MVA pathway only produces IPP, as a substrate, which can be isomerized to DMAPP by IDI, implying that IDI is essential for the synthesis of DMAPP from IPP in eukaryotic organisms, such as in mitochondria, peroxisomes, and endoplasmic reticulum.However, the MEP pathway within plastids can contribute to both IPP and DMAPP derived from 4-hydroxy-3-methylbut-2-enyl diphosphate (HMBPP) by HMBPP reductase (HDR) at a ratio of 6:1 in the last step [31,32,[35][36][37].IDI functions to keep an appropriate ratio of IPP to DMAPP in plastids.However, since it exists in both the MVA and MEP pathways, IDI plays a key role in modulating IPP and DMAPP levels for isoprenoid synthesis in multiple subcellular compartments [4,38].
Isoprenoid synthesis begins with head-to-tail condensation of DMAPP and IPP.DMAPP is extended by addition of IPP units to form short-chain prenyl diphosphates, such as geranyl diphosphate (GPP), farnesyl diphosphate (FPP), and geranylgeranyl diphosphate (GGPP) [32].As an end-product in the MEP pathway, GGPP acts as an immediate precursor to produce the first C40 isoprenoid product phytoene in the carotenoid synthesis pathway (CSP) catalyzed by phytoene synthase (PSY) [3,4].In later-diverging land plants, as an immediate precursor of carotenogenesis, GGPP produced from the plastid MEP pathway can shuttle between the cytosol and plastids despite the IPP of the C5 building block being derived from both the MVA and MEP pathways [1].Carotenoids in the chromoplasts of tomato fruit are exclusively produced via the MEP pathway, and the cytoplasmic IPP-DMAPP isomerization by IDI2 does not compensate for the loss of IDI1 activity in the plastids [4], suggesting that IDI1 activity is essential to avoid deficiency of DMAPP for carotenoid synthesis in plastids.However, the key amino acid residues for IDI1 function still remain elusive.
Here we report the genetic basis of yellow fruit in the tomato yellow fruited tomato 3 (yft3) mutant.We mapped the single recessive yft3 gene, whose wild-type (WT) allele encodes an IDI enzyme.We describe and discuss the molecular mechanism of how the S126R mutation in SlIDI1 imposes a major effect on SlIDI1 enzymatic activity and causes the yellow-colored tomato fruit phenotype in yft3 mutant.

A single recessive gene determines yellow-fruited phenotype in yft3 mutant
The yft3 (e9292) tomato line exhibited normal plant architecture and progression through fruit development (size and shape), and did not show apparent phenotypic differences other than its yellow fruit color from the WT (M82) with red fruit color at the ripening stage (54 days post-anthesis [dpa]) (Supplementary Data Fig.S1).We observed that all F 1 hybrid progenies of yft3 × M82 (5 plants) and yft3 × LA1585 (18 plants) bore red color fruits but their F 2 generation showed fruit color segregation at the ratio of 139/43 (red versus yellow) in yft3 × M82 (χ 2 = 0.183 < 3.84) and 91/25 (red versus yellow) in yft3 × LA1585(χ 2 = 0.736 < 3.84) (Table 1), sugesting that the yellow color of yft3 fruit is caused by a single recessive gene at the YFT3 allele.

Mapping of YFT3 gene
Based on the fruit color of F 2 individuals from the yft3 × LA1585 cross (Supplementary Data Fig.S2), 116 plants were screened using 45 cleaved amplified polymorphic sequences (CAPS)/derived cleaved amplified polymorphic sequence (dCAPS) markers spanning all 12 tomato chromosomes (Supplementary Data Table S1).According to the fruit colors and genotypes, logarithm of the odds (LOD) scores were calculated by R/QTL analysis, and the maximal LOD score (24.10) was detected in a 10.65-Mb region between C2_At3g62940 (SL2.50ch04:51435545.0.51436456) and C2_At1g10030 (SL2.50ch04:62086451.0.62088182) of the CAPS markers on chromosome 4 (Supplementary Data Fig.S3), and YFT3 was thereby mapped between C2_At3g62940 and C2_At1g10030 (Fig. 1A).
To further refine the position of the YFT3 locus, an additional 1338 yft3 × LA1585 F 2 individuals were screened using seven new CAPS markers within the 10.65-Mb region.The fine mapping narrowed YFT3 locus to a 239 330-bp region between the two CAPS markers M404 (SL2.50ch4:54046020.0.54047002) and M428 (SL2.50ch4:54284003.0.54285299); this region harbors 15 candidate genes (Fig. 1A, Supplementary Data Table S2).According to the tomato genome annotation (ITAG2.3,http://solgenomics.net),one of these 15 candidates, Solyc04g056390 (named SlIDI1), encodes a putative isopentenyl diphosphate δ-isomerase, which acts as a catalytic enzyme in the interconversion of IPP and DAMPP in the MEP pathway [2,4], and so directly affects the GGPP product and synthesis of carotenoid derivatives in the downstream CSP [31,32].Comparative analysis of genomic fragments amplified by PCR from M82 and yft3 showed only a single base mutation (A → C, 54165646 bp) in the sequence of YFT3/SlIDI1 (Solyc04g056390) in the yft3 mutant (designated the YFT3 allele).The mutation is located at the third exon of the YFT3 allele, 2117 bp downstream of the ATG start codon (Fig. 1B), leading to an amino acid substitution (Ser126Arg) of the YFT3 allele protein in the yft3 mutant (Fig. 1B).
Expression analysis showed that the YFT3 gene was predominantly expressed in reproductive organs, such as f lowers and fruits.The transcript level of the YFT3 allele in yft3 was higher than that of YFT3 in all tested tissues except roots, stamens, and pistils in M82 (Supplementary Data Fig.S4).

Functional complementation and loss of function assays for YFT3 gene
To determine where the YFT3 allele is responsible for the yellowfruited phenotype in yft3 mutant or not, we expressed the YFT3 gene (35S::YFT3-CP) in the yft3 mutant and observed that all the resulting transgenic yft3 plants bore red fruits compared with the control with yellow fruits at fruit ripening stage (Fig. 2A).Intriguingly, when the YFT3 gene in the M82 line was suppressed by expressing the YFT3-KO construct, its ripe fruits were changed from red to yellow (Fig. 2A), confirming that the YFT3 gene plays a critical role in tomato fruit color formation.

Mutation in YFT3 allele affects chromoplast development
During tomato fruit ripening, chlorophyll is broken down and chloroplasts develop into chromoplasts along with carotenoid biosynthesis [17].Consistently, we observed that the number of carotenoid-containing plastoglobules in fruit pericarp cells varied greatly among different tomato lines with fruit ripening (Fig. 3A and B).The number of plastoglobules in the yellow-fruited yft3 and YFT3-KO lines was significantly lower than that in redfruited M82, YFT3-CDS-CP, and YFT3-CDS-OE lines (Fig. 3A and B).
In particular, in the M82 and 35S::YFT3-OE lines, long strip-shaped crystalline bodies and/or undulating structures were observed in the chromoplasts with accumulation of carotenoids as ripening progressed, while these structures were rarely observed in 35S::YFT3-CP, yft3, and YFT3-KO lines at 54 dpa (Fig. 3).We also observed that the thylakoids, grana, and chloroplast envelope remained intact in the chloroplasts at 35 dpa, and there were no differences in chloroplast structure or integrity among 35S::YFT3-CP, 35S::YFT3-OE, YFT3-KO, yft3, and M82 lines (Fig. 3A).However, the chloroplast envelope and thylakoid structure began to degrade at 47 dpa in the 35S::YFT3-CP, 35S::YFT3-OE, and M82 lines; the linear thylakoid membrane could be visible at the same time point but had completely disappeared at 54 dpa.In contrast, distinguishable chloroplast envelopes and granular structures were still visible in both the yft3 mutant and YFT3-KO lines at 47 and 54 dpa (Fig. 3A).
In the yellow-fruited tomato lines of yft3 and YFT3-KO-10/14, the expression levels of genes involved in the MEP and CSP pathways were higher than or close to that in M82 with the redfruited phenotype during fruit ripening, but expressions of HDR and CYP97A in yft3 and of HDR, PSY1, and CRTISO in YFT3-KO lines at 54 dpa were significantly lower than in M82 (Fig. 4).

Ser126Arg mutation does not alter subcellular localization and YFT3 allele protein levels
To determine the impact of Ser126Arg substitution on the subcellular localization and products of YFT3 allele protein, YFT3-CDS and YFT3 allele-CDS without the stop codon were fused to the N-terminus of the f luorescent reporter GFP to create 35S::YFT3-GFP and 35S::YFT3/allele-GFP.The two plasmids were transformed into Nicotiana benthamiana leaves, and confocal microscopy data showed that proteins of both YFT3 and YFT3 allele were localized in the chromoplasts (Fig. 5A).Western blot analysis revealed that there were no significant differences between the abundances of YFT3 and YFT3 allele in the fruits of M82 and yft3 lines at 35, 47, and 54 dpa, although both β-ACTIN and YFT3/YFT3 allele protein contents were significantly decreased at 54 dpa compared with any of the other time points (Fig. 5B, Supplementary Data Fig.S6).These results suggest that the Ser126Arg substitution did not affect the cellular localization of YFT3/YFT3 allele proteins and their abundance in the WT and mutant lines.

Color complementation and enzymatic activity
Escherichia coli ED3 cells harboring the plasmid pTrc-LYC, carrying the coding sequences for crtB, crtE, and crtL, can produce lycopene at a basal level [39].We used this strain to investigate whether YFT3 or YFT3 allele expression can change lycopene production.ED3 cells harboring either pTrc-LYC or pET-28a(+) alone used as negative controls were cultured on/in LB medium supplemented with chloramphenicol and kanamycin.As a result, the growth of these cells was supressed (Fig. 6Aδ and ).However, three other strains harboring pTrc-LYC/pET-28a(+), pET-28a-YFT3, or pET-28a-YFT3 allele grew well in liquid LB medium and formed bacterial plaques on solid LB medium.The bacterial plaques and pellets exhibited a reddish-brown color when expressing pET-28a-YFT3 in the DE3 strain haboring pTrc-LYC, indicative of lycopene accumulation within the cells, whereas those expressing the pET-28a-YFT3 allele or pET-28a(+) in DE3 strains haboring pTrc-LYC were beige (Fig. 6Aα, β, and γ).The lycopene contents of the latter two strain lines were also significantly lower than that in the strain carrying pTrc-LYC/pET-28a-YFT3 (Fig. 6B).This result suggests that the YFT3 allele has lower enzyme activity than that in WT YFT3 protein.
We next compared the difference in catalytic activities of the recombinant YFT3 allele and YFT3/SlIDI1 proteins in vitro in isomerizing IPP into DMAPP.It has been reported that pyrophosphate groups can be readily removed from IPP and DMAPP with alkaline phosphatase, and can be converted into the more stable isoprenol (3-methyl-3-buten-1-ol, alcohol derivative of IPP) and isopentenyl alcohol (3-methyl-2-buten-1-ol, DMAPP alcohol derivative), which can be easily quantified by GC-MS [40].Our data showed that the GC-MS retention times of 3-methyl-3-buten-1-ol and 3-methyl-2-butene-1-ol standards were 3.8 and 4.9 min, respectively (Fig. 6C and D).When IPP was incubated with recombinant YFT3 protein after alkaline phosphatase addition, we also determined the retention times of isoprenol and isopentenyl alcohol at 3.8 and 4.9 min (Fig. 6 E).However, only one isoprenol peak was detected when IPP was incubated with any one of YFT3 allele protein, denatured recombinant YFT3, YFT3 allele proteins, or the negative control (empty pET-28a(+) vector) (Fig. 6F-I).These results suggest that the recombinant YFT3 protein can isomerize IPP to DMAPP, but the the recombinant YFT3 allele protein has much lower ativity.

Ser126 is an essential site for catalytic activity of YTF3
Molecular docking analyses of the YFT3/ YFT3 allele proteins with the IPP/DMAPP substrates was conducted using Discovery Studio 4.5 [41].It was predicted that the Ser126Arg subsitution alters the conformation of YFT3 and conseqeuntly changes its binding to or interaction with the substrates (Fig. 7).There are four amino acid residues (Cys157, Ser158, Tyr207 and Trp269) in YFT3 conjugating with IPP.Cys157 binds to IPP through the CC double bonds of the alkylate reaction with a bond length of 5.04 Å.Both Ser158 and Tyr207 bind to the IPP pyrophosphate group by conventional hydrogen bonds, with bond lengths of 2.80 and 2.90 Å, respectively.Finally, Trp269 binds to the IPP pyrophosphate group by interacting with the Pi-anion in the polar indole ring, with a bond length of 3.62 Å (Fig. 7A-C, Supplementary Data Table S4).Replacement of Ser126 with Arg in the YFT3 allele protein alters its spatial conformation, and makes YFT3 allele protein bind to IPP through eight amino acid residues (His110, Arg141, Lys145, Cys156, Cys157, Ser158, Lys182, and Glu217) (Fig. 7  D-F).The model indicates that the hydrosulphonyl residues of His110 and Cys157 bind to the CC double bonds of IPP via a Pialkyl reaction, and with bond lengths of 5.26 and 4.83 Å. Lys145, Cys156, and Cys157 bind to the pyrophosphate group of IPP via hydrogen bonds, with bond lengths of 2.51, 2.39, 1.99, and 2.72 Å, respectively.Four amino acid residues (Arg141, Cys156, Ser158, and Glu217) conjugate with IPP by hydrogen-carbon bonds, with lengths of 2.79, 2.62, 2.96, and 2.49 Å, respectively.Simultaneously, Arg141, Lys145, and Lys182 also bind to the pyrophosphate group of IPP by attractive charges, and with bond lengths of 3.62, 3.38, and 4.68 Å for Arg141, 4.91 Å for Lys145, and 5.09 Å for Lys182, respectively (Fig. 7D-F, Supplementary Data Table S4).
We also performed a molecular docking analysis of both the YFT3 and YFT3 allele proteins with DMAPP.The Ser126Arg mutation in the YFT3 allele resulted in a decrease in the number of active amino acid residues binding to DMAPP from nine to five.The amino acid residues Tyr207, Glu217, and Glu219 in YFT3 protein bind to the hydrogen atom in DMAPP by conventional hydrogen bonding, with bond lengths of 2.39, 2.50, and 1.99 Å, respectively.YFT3 also conjugates with DMAPP through binding to different oxygen atoms in the pyrophosphoric acid group by six active amino acid residues, both Ser126 and Ser158 with a conventional hydrogen bond, with bond lengths of 2.75 and 2.84 Å; Arg141 and Trp269 via a Pi-anion with bond lengths of 5.06 and 2.96 Å; His122 and Ser158 by a carbon-hydrogen bond with bond lengths of 2.81 and 2.10 Å; and His110 with an alkyl reaction with a bond length of 4.42 Å (Fig. 7G-I, Supplementary Data Table S4).All nine active amino acid residues in the YFT3 protein form a pocket that interacts with DMAPP, located next to the Ser126 active pocket, which is essential for sustaining the optimal conformation between the substrate (DMAPP) and the catalytic enzyme (YFT3) [42].
In contrast, the YFT3 allele protein conjugates with DMAPP by five active amino acids (His110, Arg141, Cys156, Tyr207, and Trp269) (Fig. 7J-L).Specifically, the residues His110, Arg141, Tyr207, and Trp269 were shown to interact with the oxygen atom of the pyrophosphoric acid group in DMAPP via a carbonhydrogen bond (2.78 Å), an attracting charge with a bond length of 4.57 Å, a hydrogen bond (2.94 Å), and a Pi-anion (3.08 Å), respectively.The residue Cys156 binds to the methylated carbon atom in DMAPP through an alkyl reaction with a bond length of 4.48 Å (Fig. 7J-L, Supplementary Data Table S4).
Mg 2+ is a key cofactor for YFT3 folding into an active conformation, and YFT3 protein also ensures that two oxygen atoms from the IPP pyrophosphate group bind to Mg 2+ via attracting charges, with bond lengths of 4.01 and 4.29 Å.However, in addition to the increase in the number of IPP-interacting amino acid residues in the YFT3 allele protein, Ser126Arg replacement is also predicted to disrupt the interaction between IPP and Mg 2+ (Fig. 7C and F).Mg 2+ also participates in the interaction between YFT3 and DMAPP though binding to three oxygen atoms of the Pi-anion (two oxygen atoms) and a metal acceptor (one oxygen atom) (Fig. 7I).However, in the case of the YFT3 allele protein, Mg 2+ was predicted to bind to an oxygen atom in the pyrophosphoric acid group in DMAPP via only the Pi-anion, with a bond length of 4.84 Å (Fig. 7L).These results suggest that Ser126Arg replacement increases the number of active amino acid residues of the YFT3 allele protein interacting with IPP, but decreases the number of active amino acid residues with DMAPP.Similarly, the Ser126Arg mutation also results in the disruption of the interaction between IPP and Mg 2+ during the YFT3 allele isomerization process, and affects binding of DMAPP to Mg 2+ as well.

Discussion
Fruit color is one of the most important quality traits, and in tomatoes it is closely associated with carotenoid accumulation [6,10,43], particularly the ratio of lycopene to β-carotene [11].Yellowand orange-fruited tomatoes are usually deficient in carotenoids, especially lycopene accumulation, which gives rise to the red color.In this study we characterized the genetic molecular basis of a yellow-fruited tomato mutant, yft3/e9292.A YFT3 allele gene with a genetic lesion was identified in the yft3 mutant using mapbased cloning, and a missense mutation (A2117C) downstream of the start codon was found.This point mutation results in a Ser126Arg substitution in the YFT3 allele protein (Fig. 1B).The YFT3 locus was mapped to chromosome 4, and it was predicted to be the candidate SlIDI1 gene, encoding an IDI1 enzyme that catalyzes isomerism between IPP and DMAPP in the MEP pathway.IDI1 derived from tobacco was initially isolated by Nakamura et al. [44].SlIDI1 was firstly cloned from tomato by Pankratov et al. [4], and three mutation forms were identified, which resulted in reduction of carotenoid accumulation.The three mutants fruit carotenoid deficient 1 (fcd1)-1, fcd1-2, and fcd1-3 comprise an eliminated W206Δ tryptophan in fcd1-1(e1535), a W143 * nonsense mutation in fcd1-2 (e0321 and e9292), and a G207R missense mutation in fcd1-3 (e0955).Zhou et al. [2] reported that the TB735 tomato (S. lycopersicum) also exhibited a significant reduction in carotenoid content due to a 116-bp deletion in oft3 (IDI1 allele).However, in the current study the lycopene content (2.24 μg g −1 FW) in the yft3 mutant was only 2.5% of that in WT M82 tomato.It is notable that the lycopene content in yft3 is far less than that in fcd1-2 [4] or oft3 [2].Both the fcd1-2 and oft3 tomato lines exhibited premature translation termination of SlIDI1 proteins, but unlike these mutations that cause protein truncation, our yft3 mutation is a substitution of Ser126 Arg, which is an essential amino acid residue for YFT3 function.
As a member of the Nudix hydrolase family, IDI (E.C. 5.3.3.2) is a rate-limiting enzyme that catalyzes a crucial activation step in the isoprenoid biosynthetic pathway responsible for sterol, carotenoid, dolichol, ubiquinone, and prominent classes of prenylated protein synthesis [45].Activity-enhancing mutations in an IDI triple mutant (L141H/Y195F/W256C) from Saccharomyces cerevisiae were identified by error-prone PCR [39].Three amino acid residues, Cys-67, Tyr-104, and Glu-116 in ElIDI1 from E. coli, had been revealed to be involved in the interaction between the divalent metal (Mg 2+ and Mn 2+ ) and the IPP and DMAPP substrates via protonation/deprotonation [46].The four amino acid residues (Cys87, Glu149, Trp197, and Tyr137) in Homo sapiens HsIDI were confirmed to be indispensable for the stereo-selective antarafacial transposition of a proton to convert IPP to DMAPP [45].However, only a few active amino acid residues have been identified in tomatoes, although fcd1-1 (Gly216Arg) and fcd1-3 (Trp215Δ) exhibited a slight decrease in carotenoid accumulation [4].YFT3/SlIDI1 was found here to be specifically localized in the plastids, as was the YFT3 allele protein (Fig. 5A).The YFT3 allele protein was also produced in the yft3 mutant at similar abundance to WT YFT3 protein in M82 (Fig. 5B).
However, the expression of the YFT3 allele gene was significantly higher in yft3 fruit than in YFT3 in M82 during tomato fruit ripening (Fig. 2B), as was the expression of eight genes associated with MEP (DXR, DXS, and HDR) and CSP (PSY1, CRTISO, CYCB, CYP97A, and NCED), but excluding HDR and CYP97A at 54 dpa (Fig. 4).In the YFT3-KO lines, the transcript levels of the genes associated with MEP (DXR, DXS, and HDR) and CSP (PSY1, CYCB, CRITSO, CYP97A, and NCED) were also significantly higher than that in M82 at the same time points, except for HDR, PSY1, and CRTISO at 54 dpa (Fig. 4).These results suggest that some genes associated with MEP and CSP were expressed in yellow-fruited tomato lines, such as yft3 and YFT3-KO-10/14, which is counterintuitive based on the low carotenoid levels in yft3 and YFT3-KO lines.However, we did not detect significant differences among the red-fruit tomato lines in the transcript levels of most of the eight genes associated with carotenoid synthesis, whereas transcript levels of CYP97A in both 35S::YFT3-CP-6/16 and 35S::YFT3-OE-3/6/7 lines at 47 dpa and DXR, DXS, and PSY1 at 54 dpa were significantly higher than that in M82 at the same time points, but expressions of HDR and CYP97A in the 35S::YFT3-CP-6/16 and 35S::YFT3-OE-3/6/7 lines were significantly lower than that in M82 at 54 dpa (Fig. 4).
Fruit color formation is an intricate biological process in tomato, but its molecular regulation remained largely unknown.We hypothesize that there may be a compensation mechanism for YFT3 with genetic lesions resulting in increased expression of some genes associated with carotenoid synthesis in yft3 and YFT3-KO lines, potentially by feedback regulation from the chromoplasts of the yellow-fruited tomato lines.As five-carbon basic building units, isopentenyl diphosphate (IPP) and its isomer dimethylallyl diphosphate (DMAPP) are common isomeric precursors for all isoprenoids/terpenoids [33].One active DMAPP molecule was respectively condensed with one IPP, two IPPs, and three IPPs to form C10 geranyl diphosphate (GPP), C15 farnesyl diphosphate (FPP), and C20 geranyl diphosphate synthase (GGPP), which is an essential precursor to produce various terpenoids/isoprenoids such as C40 carotenoids [47].Therefore, a huge sink would be formed to store various secondary metabolites such as carotenoids with tomato fruit ripening.However, IPP and DMAPP were produced at a ratio of 6:1 in the MEP pathway in plastids by HMBPP reductase (HDR) catalysis [4].The ratio or balance of IPP to DMAPP would be disturbed with isoprenoid synthesis, and will result in deficiency of DMAPP, especially if SlIDI1 protein function, which catalyzes conversion between IPP and DMAPP, is lost or its activity decline, as in like yft3 and YFT3-KO line.To meet the powerful traction from the huge sink and isoprenoid syntheses, more DMAPP would need to be produced in plastids, and thus the genes associated with MEP would be triggered or promoted to express it.In particular, the YFT3 protein function presents decline or loss in yft3 and YFT3-KO tomato lines, which will block or abolish conversion from IPP into DMAPP, and this will result in a decline in DMAPP content within plastids.More DMAPP would need to be produced for isoprenoid syntheses, and this requirement will enhance transcript expression of some genes associated with MEP and CSP in yft3 and YFT3-KO lines by a feedback regulation pattern (Fig. 4).
We used an engineered E. coli strain to express genes related to carotenoid synthesis to assess the catalytic activity of YFT3 and its allele protein, with the latter showing reduced activity (Fig. 6).Our GC-MS analysis further showed that the recombinant YFT3 protein could convert IPP to DMAPP in vitro, while the recombinant YFT3 allele and denatured YFT3/YFT3 allele proteins failed to do this (Fig. 6).These results indicate that the Ser126Arg missense mutation abolishes the ability of YFT3 allele to convert IPP to DMAPP.To further elucidate the basis and functional lesion of the YFT3 allele protein, we performed a molecular docking analysis of its interaction with the IPP/DMAPP substrates.The results suggest that the Ser126Arg mutation resulted in changes in the spatial conformation of the enzyme-ligand interaction in the YFT3 allele protein, with the number of amino acid residues binding to IPP increasing from four to eight, while binding to DMAPP decreased from nine to five.The analysis also predicted that the mutated YFT3 allele protein disrupts IPP binding to the Mg 2+ cofactor.
Those results would explain the lower catalytic activity compared with that of YFT3 protein (Fig. 7).
Based on our results, we predicted that Ser126 is an essential amino acid residue for the function of YFT3/SlIDI1, which is a rate-limiting enzyme in the isoprenoid biosynthetic pathway and catalyzes the reversible conversion between IPP and DMAPP in the plastids.Thereby, Ser126Arg in YFT3 allele would affect carotenoid accumulation in tomato fruit.This study provides important insights into tomato quality improvement and breeding in the future.

Plant materials and growth conditions
Seeds of both WT (S. lycopersicum, cv.M82) and e9292 (S. lycopersicum, named yellow fruited tomato 3, yft3) mutant were provided by Professor Dani Zamir (the Hebrew University of Jerusalem).The yft3 mutant was created from M82 by mutagenesis with ethyl methyl sulfonate (EMS).Seeds of LA1585 (Solanum pimpinellifolium) were obtained from the Tomato Genetics Resource Center (University of California, Davis, CA, USA).Tomato seeds of two mutant populations (yft3 × LA1585 and M82 × yft3) from different generations (F 1 and F 2 ) were created by sexual hybridization.Transgenic tomato lines of YFT3-KO, 35S::YFT3-OE, and 35S::YFT3-CP were generated in the M82 and yft3 background by Agrobacterium tumefaciens-mediated transformation.All tomato lines were planted and grown under standard greenhouse conditions at the Pujiang experimental base (121 • 30 10.89 E, 31 • 3 5.20 N, altitude 5 m), at the Shanghai Jiao Tong University, Shanghai, China.Tobacco (Nicotiana benthamiana) seeds were stored in Zhao Lab at the Shanghai Jiao Tong University, China.Tobacco seedlings were grown in pots with damp nutrient soil (field soil:vermiculite:humus = 4:2:4) in a chamber at 24 • C under a 16-h light/8-h dark light regime with 20 000 lux and 65% relative humidity.

Analyses of genetics of fruit colors and gene map-based cloning
Based on the external ripening fruit colors (red and yellow) of tomato plants in the F 1 and F 2 generations (yft3 × LA1585 and M82 × yft3), the inheritance of fruit color in the yft3 mutant was analyzed using a χ 2 test.To make linkage groups with the mutated loci, a total of 45 CAPS/dCAPS markers were created based on data from the Sol Genomics Network database (http://solgenomics.net/).These markers span all 12 tomato chromosomes, and were designed by analyzing single-nucleotide polymorphisms (SNPs) in the target DNA sequences between yft3 and LA1585 (Supplementary Data Table S1).The yft3 × LA1585 tomato plants with different fruit colors in the segregating F 2 generation were also used to identify the candidate gene by map-based cloning.Genomic DNA was extracted from the young leaves of each plant as described in Chen et al. [11].Forty-five markers were used to screen 116 individual plants randomly selected from yft3 × LA1585 F 2 population (25 yellow-fruited and 91 red-fruited), for primary mapping of the location of the YFT3 gene.Based on the genotypes and fruit color phenotypes of 116 plants in the F 2 generation, the target region was confirmed by calculating the LOD scores.An additional 1338 plants derived from the yft3 × LA1585 F 2 population were used to fine-map the YFT3 target region using seven newly-designed CAPS markers (Supplementary Data Table S1).
Genotypic and phenotypic data from the F 2 population were used to create linkage maps using R/QTL software [48], and the region with the candidate YFT3 gene was identified using the Genome Browser (https://solgenomics.net/jbrowse_ solgenomics).The candidate gene was further confirmed by examining the gene functional annotations in the predicted mapping region https://solgenomics.net/jbrowse_solgenomics). A DNA fragment that contains the candidate gene, ISOPENTENYL DIPHOSPHATE ISOMERASE 1 (IDI1), was amplified using LA Taq DNA polymerase (Takara, Dalian, China) with the genespecific primers 5 -cacccttaggttggtgttttgttgag-3 (forward) and 5 -gcctaatctgaaatggctcaaagg-3 (reverse) (Supplementary Data Table S3) and sequenced.

Structural features of YFT3
Total RNA was extracted from fresh pericarp at the equatorial region of M82 and yft3 tomato fruits at 47 dpa (corresponding to the breaker stage, BR), using the RNAprep Pure Plant Kit (Tiangen, Beijing, China).The coding sequences (CDS) of YFT3 and YFT3 allele were amplified from M82 and yft3 tomato fruit cDNA libraries (at 47 dpa) using specific primer pairs (Supplementary Data Table S3).The CDS sequences were aligned to the corresponding genomic DNA sequence to confirm the number and length of exons and introns, as well as the mutations in YFT3 allele in yft3 and the corresponding amino acid substitutions.

Constructs and genetic transformation of tomatoes
The YFT3 CDS was amplified from a red-fruited M82 cDNA library (at 47 dpa) using the specific primers 35S-CDS-BamHI and 35S-CDS-SacI (Supplementary Data Table S3), and then cloned into the BamHI and SacI sites in an intermediate vector plasmid 35S::GUS to create 35S::YFT3-CDS.The 35S::GUS expression vector was constructed based on the backbone of the pCAMBIA 2300 plasmid, and with the GUS expression cassette from pBI121 being inserted into the corresponding HindIII and EcoRI sites of the pCAMBIA 2300 plasmid.
The YFT3-KO vector was constructed based on the pTX041 plasmid [49], provided by Professor Chuanyou Li (Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Peking, China).Two target DNA fragments for the YFT3-CDS sequence were designed using the CRISPR-P v2.0 website (http://cbi.hzau.edu.cn/CRISPR2/).Specific cri-YFT3-F/R primers (Supplementary Data Table S3) were designed to amplify the DNA fragments with two target sequences with pTX043 as the template using Ex Taq DNA Polymerase (TaKaRa, Dalian, China).The amplified PCR product was digested with BsaI and then constructed into the corresponding pTX041 site to create YFT3-KO.

Measurement of carotenoid content
Pericarp samples from the equatorial region of the YFT3-CDS-CP/OE, YFT3-KO, M82, and yft3 fruits were collected at 35, 47, and 54 dpa, and then powdered in liquid nitrogen.Carotenoids were extracted using methyl alcohol/chloroform, and examined using a Waters Acquity Ultra-performance Convergence Chromatography (UPC 2 ) system (Waters, Milford, MA, USA) as described in Zhao et al. [29].The standards of lycopene, β-carotene, α-carotene, and lutein were products of Yuanye Biotechnology (Shanghai, China).The standards were dissolved in MTBE to make standard curves, which were drawn as described in Zhao et al. [29].

Extraction of total soluble protein and western blotting
Pericarp samples of the equatorial region of M82 and yft3 tomatoes (35,47, and 54 dpa) were collected (n = 3) and powdered in liquid nitrogen.Approximately 1 g of pericarp powder was mixed well with 1 ml PBS extraction buffer (1.75 mM KH 2 PO 4 , 10 mM Na 2 HPO 4 , 140 mM NaCl, 2.7 mM KCl, pH 7.4) with 1 mM phenylmethylsulfonyl f luoride (PMSF, Yeasen, Shanghai, China) to extract total soluble protein (TSP).The homogenates were placed in an ice bath for 4 h and centrifuged at 12 000 × g at 4 • C for 40 min.The supernatants were collected, and the TSP concentrations were measured using the Bradford method [56].Bovine serum albumin was used to create a standard curve.
The concentration of crude TSP was adjusted to 0.5 μg/μl with PBS extraction buffer, and 10 μl of TSP was mixed with 5× loading buffer [250 mM Tris-HCl (pH 6.8), 50% (v/v) glycerol, 10% (w/v) sodium dodecyl sulphate (SDS), 5% (v/v) β-mercaptoethanol, and 0.5% (w/v) bromophenol blue] and was denatured at 95 • C for 10 min before centrifugation at 10 000 × g for 1 min.All protein samples were separated using 10% SDS/polyacrylamide gel electrophoresis (SDS-PAGE) in Tris-glycine buffer [0.025 M Tris, 0.25 M glycine and 0.01% (w/v) SDS].Subsequently, one gel was stained with 2.5% (w/v) Coomassie Brilliant Blue R250, and the other gel loaded with protein samples in the same order was transferred to a polyvinylidene f luoride (PVDF) membrane (filter pore size 0.22 μm, Millipore, USA) for western blot analysis.The PVDF membrane was blocked in 5% (w/v) non-fat milk powder (TPBS, Sangon Biotech, Shanghai, China) for 2 h.The YFT3 protein was detected by incubating the transferred PVDF membrane with a rabbit anti-YFT3 polyclonal antiserum, which was created by using an oligomeric peptide of RGIDGNKPMSLTTAS located at amino acids 2-16 of the YFT3 protein, to immunize rabbits (Sangon Biotech, Shanghai, China) at a 1:500 dilution in TPBS at 4 • C overnight.Horseradish peroxidase (HRP)-conjugated goat anti-rabbit immunoglobulin G (IgG) (Beyotime, Shanghai, China) at a 1:1000 dilution in TPBS was used as the secondary antibody.Immunoreactive YFT3 protein was visualized using the Ultra High Sensitivity ECL Kit (MedChemExpress, Shanghai, China) and scanned to produce digital images.Tomato β-actin (http://www.affbiotech.cn/) was used as a reference protein.

In vivo and in vitro enzymatic activity of YFT3
The pTrc-LYC plasmid containing a chloramphenicol resistance gene was constructed from the pTrcHis2B framework plasmid, which carries a gene cluster of crtE (geranylgeranyl pyrophosphate synthase), crtB (phytoene synthase), and crtL (lycopene cyclase) for lycopene biosynthesis [39].It was kindly provided by Professor Haibo Zhang (Qingdao Institute of Bioenergy and Bioprocess Technology, Chinese Academy of Sciences, Qingdao, China).pTrc-LYC derived from a prokaryotic expression system was used to estimate enzymatic activity of the YFT3 protein [57].
The DE3 cells were cultured in lysogeny broth (LB) liquid medium [10 g/l tryptone and 5 g/l yeast extract (Thermo Fisher, USA), and 10 g/l NaCl (Lingfeng,Shanghai, China), pH 7.0] with 50 mg/l chloramphenicol and kanamycin, and grown with shaking at 250 rpm at 37 • C until cultures reached an OD 600 value of 1.0.Isopropyl-beta-d-thiogalactopyranoside (IPTG) was added to a final concentration of 0.5 mM, and then the cells were spread on a Petri dish with LB/agar containing 100 mg/l chloramphenicol and kanamycin.The plates were placed upside down in an incubator (Tiancheng, Shanghai, China) in the dark at 28 • C for 3 days.The proliferation and color of the cells were observed each day.
Each DE3 engineered strain was inoculated in 100 ml LB liquid medium with 50 mg/l chloramphenicol and kanamycin, and shaken at 220 rpm at 37 • C to an OD 600 of 0.6.
IPTG was then added to a final concentration of 0.5 mM and the cultures were shaken at 100 rpm at 16 • C to induce recombinant protein expression.Five milliliters of each culture was sampled 24 h after induction, and the cells were collected by centrifugation at 3000 × g for 5 min.The cell pellet was resuspended in 200 μl acetone and incubated in a water bath at 55 • C for 15 min, and centrifuged at 10 000 × g for 3 min; 150 μl of the supernatant was sampled and absorbance was measured at 475 nm (PowerWave XS, Biotek, VT, USA).The relative lycopene ratio (RLR) was calculated using the formula: RLR = (Y TRIAL − Y CONTROL )/Y CONTROL .Here, Y TRIAL is the absorbance value for DE3 cells concurrently harboring pTrc-LYC, pET-28a-YFT3 or pET-28a-YFT3 allele, and Y CONTROL is the absorbance value for the DE3 cells carrying both pTrc-LYC and pET-28a(+).
The activities of the recombinant YFT3 and YFT3 allele proteins were estimated in vitro, as described in Ma et al. [40] with some modifications.The purified recombinant YFT3 and YFT3 allele proteins and 10 μg of the empty pET-28a(+) vector were individually added to a 1.5-ml Eppendorf tube containing 5 mM MgCl 2 , 1 mM DTT, and 20.0 μg IPP substrate.The volume was adjusted to 100 μl by adding 50 mM PBS.The reaction was performed at 37 • C in darkness for 3 h, and then 1 U alkaline phosphatase (Shanghai Yuanye Biotechnology, Shanghai, China) was added.The reactions were incubated at 37 • C overnight to completely remove the pyrophosphate group from IPP or DMAPP, and the reaction products were extracted by addition of petroleum ether (v:v = 1:1).
The enol derivatives of IPP and DMAPP, 3-methyl-3-butene-1-ol and 3-methyl-2-butene-1-ol are more stable than IPP and DMAPP.The concentrations of these derivatives were assayed by gas chromatography-mass spectrometry (GC-MS, Agilent, CA, USA) to estimate enzymatic catalytic activity.GC-MS analysis was performed using a low-loss HP-5-ms (Agilent, California, USA) GC-MS column.The initial gas chromatography temperature was 35 • C for 2 min, and then increased to 40 • C at 5 • C/min before holding for 5 min.Finally, 280 • C was reached at 20 • C/min and held for 5 min.The mass spectral scan range was set to 30-240 m/z.

Molecular docking analysis
To elucidate structural differences between the YFT3 and YFT3 allele proteins that might explain their different catalytic activities, we predicted the binding pockets of the two proteins and the substrates IPP/DMAPP using Discovery Studio 4.5 (DS) [41].A homology model of YFT3 was created using human isopentenyl diphosphate isomerase (hIDI) (PDB code: 2ICJ) as a reference template.The target protein was minimized by invoking the functions of Minimize and Refine in DS [58].IPP and DMAPP structures were drawn using ChemDraw.The CDOCKER module available with DS was employed to generate the interaction model between the YFT3/YFT3 allele catalytic proteins and IPP/DMAPP substrates with all the parameters set as default [59].The best position was selected based on the highest docking score from the largest cluster and the key residue interactions [60].

Figure 1 .
Figure 1.Map-based cloning of the YFT3/YFT3 allele genes and the corresponding predicted protein sequences.A Mapping of YFT3.B Structural features of the YFT3/YFT3 allele genes and the corresponding predicted protein sequences.ATG, start codon; CDS, coding sequence; TAA, stop codon; 5 UTR, 5 -untranslated region; 3 UTR, 3 -untranslated region; YFT3, a yellow-fruited tomato 3 gene; its candidate gene is SlIDI1, and was isolated from the red-fruited M82 tomato; YFT3 allele, an allele of the YFT3 gene, which carries a mutation at 2117 bp (A → C) downstream of the start codon ATG; it was isolated from the yft3 mutant tomato.

Figure 3 .
Figure 3. TEM imaging of chromoplast ultrastructure.A Chromoplast ultrastructure within pericarp cells among tomato lines with different genetic backgrounds at different stages.pg, plastoglobule; gr, grana; lth, long linear thylakoid membrane structure; ccr, carotenoid crystalloid.Scale bar = 500 nm.B Difference in plastoglobule number among different tomato lines at different stages.Data represent the mean values of three biological replicates.Eight different fields of view were observed for each biological repeat, and the error bars represent the standard deviations.Small letters indicate statistical significance in plastoglobule number at P < 0.05 as determined by Duncan's test.M82, wild type of the experimental material; yft3, e9292 tomato mutant; 35S::YFT3-OE, transgenic tomato lines created by transforming M82 with 35S::YFT3-CDS; 35S::YFT3-CP, transgenic tomato lines created by transforming yft3 with 35S:: YFT3-CDS; YFT3-KO, YFT3 knockout in the M82 background, created using CRISPR-cas9.35 dpa, 47 dpa, and 54 dpa correspond to the mature green stage (MG), breaker stage (BR), and red/yellow ripening stage (RR/YR) during tomato fruit ripening.

Figure 7 .
Figure 7. Molecular docking of YFT3 and YFT3 allele proteins with IPP and DMAPP.A-F Molecular docking of YFT3 and YFT3 allele proteins with IPP.G-L Molecular docking of YFT3 and YFT3 allele proteins with DMAPP.Regular molecular docking is shown in A, D, G, and J and linear molecular docking is shown in B, E, H, and K. C, F, I, and L are ball-and-stick molecular docking models.

.
Segregation of fruit color in two genetic populations.

Table 2 .
Carotenoid content of fruit in tomato lines with different genetic backgrounds (μg g − 1 FW).
Data are mean ± standard deviation of the three biological replicates.Capital and small letters indicate statistical significance at P < 0.01 and P < 0.05, respectively, using Duncan's test.n.d., not determined.