Metabolite profiling and its genetic basis in tomato

By combined metabolomic approach (see Method), we detected and quantified 92 metabolic traits in red fruit from the association panel (302 inbred lines) harvested in two environments open field (E1) and greenhouse (E2) (S1–S3 Tables). Most metabolites had coefficients of variation (CV) was greater than 40% (S1A Fig). Among the 49 repeatedly detected traits (48 primary metabolites including ascorbate, SSC), 46.9% (23 out of 49) displayed broad-sense heritability (H²) greater than 0.3, and 22.4% had heritability values >0.5 (S1B Fig). A wide range of variation was observed for some metabolites in each species and subgroup over the years (S4 Table). Across the groups, several sugars (sucrose, fructose, and myo-inositol) and organic acids (citric acid, malic acid, ascorbic acid and pentanoic acid) were present at the highest levels in PIM, as were two amino acids (alanine and L-glutamic acid) and three flavor-related components (SSC, total sugar and total acid) (S2 Fig), suggesting selection during domestication for these metabolites.

We performed GWAS using a compressed mixed linear model (CMLM) to reveal the metabolic regulatory mechanisms in tomato fruit under different environments (E1 and E2). Using a Bonferroni correction based on the effective numbers of independent markers [29], the P-value thresholds were set at 2.4 × 10⁻⁷ (suggestive) and 1.2 × 10⁻⁸ (significant). The detected 388 lead SNPs (including 262 suggestive SNPs and 126 significant SNPs) across the two environments (S3 Fig and S5 Table), including repeated detection of 103 SNPs (Fig 1). We identified six potential mGWAS hotspots (density>0.015), on chromosomes 1, 3, 5, 9, and 11. These loci were often frequently by metabolites that are biochemically related; for instance, a hot spot on chromosome 1 was identified for approximately half (46.7% or 7/15) of the sugars detected in this study (S5 Table). Candidate genes underlying these loci might encode central regulators of these pathways and/or influence rate-limiting reactions. The percentage of phenotypic variation explained by each locus ranged from 2.4 to 36.1% and from 2.1 to 30.3% in the two replicates, with mean values of 10.2 and 8.6%, respectively (Table 1). To test for possible interactions between high significance QTL loci (P≤1.2 × 10⁻⁸), we investigated the pairwise epistatic interactions between the QTLs of each metabolic trait in each environment. We detected 27 significant interactions (P<0.05) for 4 metabolites (4-hydroxyproline, quininic acid, lactic acid and galactose oxime) whose levels were controlled by more than one QTL in both environments (S6 Table). The epistatic effect (i.e., sum of two-locus interaction effects) on metabolic variation ranged from 2.2 to 51.2%, with an average of 8.03%, suggesting that epistasis plays an important role for the four metabolites. For lactic acid, the epistatic effect was weaker compared with the main effects of the loci (i.e., sum of single-locus effects), while for the other three metabolites, the effect was comparable to or greater than the predominant effect (S4 Fig) suggesting diverse pathways or interconnected mechanism.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Overview of mGWAS results in this study.

(A) The strength of the association of metabolites is indicated as the negative logarithm of the P value for the CMLM model. All metabolite–SNP associations with P values < 2.4 × 10⁻⁷ are plotted against genome location at 1-Mb intervals. Red and blue horizontal dashed lines in the manhattan plots indicate the genome-wide suggestive (P values < 2.4 × 10⁻⁷) and significant (P values < 1.2 × 10⁻⁸) threshold, respectively. Amino acids are indicated by olive green circles. Organic acids are indicated by light green circles. Sugars are indicated by light blue circles. Other compounds are indicated by pink circles. (B) Distribution density of mGWAS signals across the tomato genome (SL2.50). mGWAS, metabolomic genome wide association analyses.

https://doi.org/10.1371/journal.pgen.1008149.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Summary of significant loci–trait associations identified in GWAS population.

https://doi.org/10.1371/journal.pgen.1008149.t001

Key genes involved in primary metabolism in ripe tomato fruit

The high-density linkage disequilibrium map generated in this study helped us narrow down association signals to regions close to or directly on genes that have been reported previously [3, 27, 30–35]. For each significant loci identified in this study, candidate genes within 200kb (<average LD of tomato) of the lead SNP are listed in S5 Table, providing a database for investigation of specific metabolites of interest.

Taking advantage of the mGWAS results, we searched for candidate genes based on (i) gene annotation, (ii) prior knowledge, (iii) gene expression, and (iv) structural variation to identify the most likely causal genes of the identified loci for metabolites measured here. The 37 candidate genes listed in S7 Table, which are located within 32 significant loci for 27 metabolites, are potentially causative for the identified association signals (S5 Fig). Among these 37 candidate genes, four significantly associated loci (ch02_44249907, ch05_1882379, ch06_36888571, and ch09_3306649) on chromosomes 2, 5, 6, and 9, with P values of 1.4 × 10⁻⁷, 2.8× 10⁻⁸, 3.4 × 10⁻⁹ and 3.9 × 10⁻¹¹ were identified by GWAS for SSC. Three of these four loci correspond to HT1 (Hexose transporter 1), SUT2 (Sucrose transporter LeSUT2), and Lin5 have been reported previously [3, 31, 32]. Ch06_36888571 is a novel locus within LD of Solyc06g054270, which encodes the sugar transporter STP11. We further validated the association between its allelic variations and SSC via mQTL mapping (S6 Fig). We generated an experimental F₂ population by crossing a tomato accession with low SSC, HG22 (an elite inbred line in China, relative SSC of 4.8%) with TS-21, which confers high SSC (a Solanum pimpinellifolium accession from Peru, relative SSC of 9%). By the bulk segregant analysis (BSA), the causal locus of fruit SSC accumulation was mapped to two intervals, 2.07 Mb on Chr 2 and 7.84 Mb on Chr6 (from ~38.8 to 40.8 Mb on Chr2 and 36.1 to 43.9 Mb on Chr6, respectively), with the peak centered on the mapping interval identified in our GWAS analysis (S6A and 6H Fig). The results of linkage mapping in the F₂ population from TS-21×HG22 were consistent with the results of GWAS analysis, further supporting the notion that STP11 on chromosome 6 is a candidate gene driving the observed natural variation in fruit SSC (S8 Table). This demonstrates that our metabolite profiling and GWAS analysis could provide accurate genetic architecture of tomato primary metabolites. We further investigate the putative loci associated with ascorbate concentration in tomato fruit. Three loci (SL2.50ch07_1311726 with P value = 8.0 × 10⁻⁸, SL2.50ch07_60983724 with P value = 4.6 × 10⁻¹⁰ and SL2.50ch07_65942036 with P value = 4.1 × 10⁻¹¹) on chromosome 7 and one loci (ch09_64101874 with P value = 3.1 × 10⁻¹¹) on chromosome 9 were identified by the GWAS of AsA.

Identification of TFA9 by GWAS

Among those loci associated with ascorbate concentration in tomato fruit, one loci was co-localized with a previously reported AsA large-effect QTL on chromosome 9 [36], we designated this locus as TOMATO FRUIT ASCORBATE ON CHROMOSOME 9 (TFA9) (Fig 2). The SNP with the highest association to fruit AsA content explained 15.9% of the total variance. Two major haplotypes based on the lead SNP (ch09_64101874) of the association signal—High-AsA haplotype (HAH) and Low-AsA haplotype (LAH)—were associated with high-AsA and low-AsA phenotypes in tomatoes, respectively (Fig 2C).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Genome-wide association results for ascorbate content in tomato fruits.

(A) Manhattan plot displaying the GWAS results for ascorbate content in fruits (CMLM, N = 302). Negative log₁₀-transformed P values from the compressed mixed linear model are plotted on the y-axis. Horizontal dashed line indicates a genome-wide suggestive threshold of 2.4×10⁻⁷. (B) Quantile-quantile plot for ascorbate content in the GWAS population. (C) Box plot for ascorbate content, plotted as a function of genotypes at SNP ch09_64101874. The metabolic data for ascorbate were log₂ transformed. (D) Detailed plot from 63.1–65.1 Mb on chromosome 9 (x-axis). Lead SNP is indicated in purple. A representation of pairwise r² values (a measure of LD) among all SNPs in region 63.1–65.1 Mb, where the colour of each box corresponds to the r² value according to the legend. (E) A representation of the pairwise r² values (a measure of LD) among all polymorphic sites in the 830 kb genomic region corresponding to (D), where the darkness of the color of each box corresponds to the r² value according to the legend. The 145 haploblocks are represented by inverted triangle. Haploblock 20 (marked by red star) contains Lead SNP associated with fruit AsA content. (F) Gene structure of two genes in the haploblock 20. Filled grey, blue and blacklines represent coding sequence, promoter & 3’UTR and introns respectively. (G, H) The ascorbate content (G) and relative expression of candidate gene (SlbHLH59) (H) in fruits from different selected accessions. Data represent means ±s.d. (n = 3). The dashed horizontal lines represent the average ascorbate content and expression levels of SlbHLH59 in fifteen low-AsA accessions and fifteen high-AsA accessions.

https://doi.org/10.1371/journal.pgen.1008149.g002

There was a total of 19 genes within 100-kb on either side of ch09_64101874 (S9 Table). But given the estimated LD decay rate of more than 800 kb in BIG group tomato [26], we carefully analyzed the pairwise LD distance within the 2-Mb interval centered on the lead SNP (ch09_64101874) from the GWAS (Fig 2D). All significant SNPs (P value ≤ 1.2 × 10⁻⁸) fall into an 830 Kb region for 63.95 Mb to 64.77. A haplotype analysis of the region spanning all the significant SNPs on chromosome 9 (830 Kb) identified 145 haploblocks, and many significant SNPs including lead SNP (ch09_64101874) trace back to haploblock 20 (SL2.50ch09_64,095,308-SL2.50ch09_64,109,883) (Fig 2E). The Haploblock 20 (14.575 kb; Fig 2F) spans two genes, a bHLH transcription factor (Solyc09g065820, 64,095,481–64,100,029bp) and NADH ubiquinone oxidoreductase (Solyc09g065830, 64,101,025–64,109,232bp) and contains 47 SNPs, 19 of which show P values were less than 12.2× 10⁻⁷ (S10 Table), suggesting their potential role in AsA accumulation. To identify the casual gene for AsA content in tomato fruit, we randomly selected 15 PIM accessions of high-AsA and 15 BIG accessions of low-AsA and measured expression of both genes in fruit by quantitative RT-PCR. The expression of Solyc09g065820 showed significantly higher expression in fruit with high-AsA as compared to low-AsA accessions (Fig 2G and 2H). No significant difference was observed in the expression of NADH ubiquinone oxidoreductase (S7 Fig). Basic helix-loop-helix (bHLH) proteins are a large superfamily of transcription factors functioning in a wide range of metabolic, physiological, and developmental processes in plants [37, 38]. Based on these results, Solyc09g065820 gene (referred to hereafter as SlbHLH59) is the likely candidate underlying TFA9.

An 8-bp InDel in the promoter of SlbHLH59 influences its expression

To investigate functional allelic variation at the SlbHLH59 locus, we analyzed the nucleotide sequence of SlbHLH59 in 369 tomato accessions with diverse AsA content. Sequence analysis suggested that the SlbHLH59 genotype can be classified into four different haplotypes (Hap 1, Hap 2, Hap 3 and Hap 4) by a total of 11 polymorphisms (Fig 3A and S11 Table), including one InDel (InDel_8, ->TCTCTTTC variant at position -1324) and three SNPs (SNP1, T>C variant at position -983; SNP2, A>G variant at position -402; SNP3, C>T variant at position -399) in the promoter region, six intron SNPs (SNP4, T>C variant at position 958; SNP5, A>G variant at position 1419; SNP6, C>T variant at position 1922; SNP7, T>A variant at position 2030; SNP8, T>C variant at position 2718; SNP10, T>A variant at position 3438), and one nonsynonymous polymorphism in exon 4 (SNP9, A>G variant at position 2903, with amino acid change from I to V). Interestingly, Hap 1 (mainly consists of BIG accessions), Hap 2 (consisting of BIG and CER accessions) and Hap 3 (consisting of BIG and CER accessions) showed lower AsA content than Hap 4 (mostly PIM accessions) (Fig 3B), and line with the result mentioned above where PIM tomato accessions showed higher AsA content than CER and BIG tomato accessions (S2B Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Haplotype and functional analysis of SlbHLH59 in tomato.

(A) Structural variations of the four SlbHLH59 haplotypes. Blue represent the promoter and 3’ UTRs, grey boxes represent coding sequences and the line between the grey boxes represents the intron. The 11 nucleotide polymorphisms are indicated at their corresponding positions. Four haplotypes of SlbHLH59 are examined in 367 tomato accessions. (B, C) The relative ascorbate content (B) and expression of SlbHLH59 (C) in fruits of four SlbHLH59 haplotypes. (D) Transient expression assay of promoter activity in Arabidopsis protoplasts. The SlbHLH59 promoter fragments were inserted into the reporter vector (pCAMBIA1304-GUS). Both vectors of reporters and internal control (pGreen II 0800, 35S+LUC) were transformed into Agrobacterium cells and used for Arabidopsis protoplast transformation. Left, constructs with site-directed mutations at the four polymorphisms in the promoter region. Right, relative GUS/LUC values. Values are represented as means ± SD (n = 3). (E, G, I) The relative expression of SlbHLH59 in SlbHLH59^SNP9A overexpression (E) SlbHLH59^SNP9G overexpression (G) and SlbHLH59 RNAi (I) transgenic tomato lines. (F, H, J) The ascorbate content in SlbHLH59^SNP9A (F) SlbHLH59^SNP9G (H) and SlbHLH59 RNAi (J) transgenic tomato lines. The background of SlbHLH59 overexpression and RNAi transformation is TS-186 and TS-265 respectively. Asterisks indicate significant differences by t test: **P< 0.01.

https://doi.org/10.1371/journal.pgen.1008149.g003

All accessions in Hap 4 with nine consensus polymorphisms (InDel_8, SNP1, SNP2 and SNP3 in promoter; SNP4, SNP6, SNP7, SNP8 and SNP10 in intron) exhibited higher AsA content than other haplotypes suggesting variation in AsA content among the SlbHLH59 haplotypes was attributed to polymorphisms in the promoter (Fig 3A). The two SNPs (SNP1 and SNP2) do not change the known cis elements in the promoter of SlbHLH59 but SNP3 resides at a box-1 cis element (light responsive element) according to PLACE (Plant cis-acting regulatory DNA elements) analysis. Notably, the InDel_8 in the promoter of SlbHLH59 led to the formation of a 5’UTR Py-rich stretch motif (TTTCTCTCTTTCTC) associated with elevated expression of downstream genes [39, 40], and consistent with increased SlbHLH59 expression in the high-AsA accessions that contain this motif versus the low-AsA accessions without this motif (Fig 2H). Using published data [41], we observed that SlbHLH59 showed significantly higher expression in Hap 4 accessions than in other haplotypes (Fig 3C). Moreover, we conducted transient assays using site-mutated promoter fragments of SlbHLH59 in Nicotiana benthamiana to test the effects of the four polymorphisms (three SNPs and InDel _8) under the promoter region on SlbHLH59 expression (Fig 3D). Expression was significantly higher in the promoter fragments with the 8bp insertion than not, supporting its role in the expression of SlbHLH59.

In total of 159 tomato bHLH proteins, 68 SlbHLHs (43%) showed I and 60 SlbHLHs (38%) showed V at the SNP9 which is located on the second helix of the conserved bHLH domain, suggesting the conservation of this amino acid residues in tomato (S8A Fig) [42]. The non-consensus SNP9 with “A” allele in Hap 1 and “G” allele in Hap 2, 3 and 4 showed irrelevance with AsA content in tomato (Fig 3A). To further functionally characterize the role of SNP9on SlbHLH59 expression and AsA content, two overexpression (OE) constructs containing allele SlbHLH59^SNP9A and SlbHLH59^SNP9G were separately introduced into TS186 (a low AsA accession). The two SlbHLH59^SNP9A transgenic plants, A-OE5 and A-OE9, with higher SlbHLH59 expression, showed significantly enhanced total and reduced AsA content compared to wild type (Fig 3E and 3F). Similarly, fruits from G-OE4 and G-OE6, two SlbHLH59^SNP9G overexpression transgenic lines exhibited enhanced AsA content than the control, with the comparable AsA levels as A-OE5 and A-OE9 fruits (Fig 3G and 3H). These results indicate that the SNP9A allele and SNP9G allele of SlbHLH59 are both functional, consistent with the results of haplotype analysis indicating SNP9 was non-causal for AsA content (Fig 3A). The differential AsA accumulation within tomato is therefore more likely attributed to nucleotide differences in the promoter region. To further test the function of SlbHLH59 in AsA biosynthesis, an RNA interference (RNAi) vector was constructed and transformed into TS-265 (a high AsA accession). Down regulation of SlbHLH59 in TS-265 resulted in significant reduction AsA content (Fig 3I and 3J). All of these results suggested that the InDel_8 in the promoter of SlbHLH59 and the resulting absent/present 5’UTR Py-rich stretch motif, is the major cause underlying the QTL TFA9 on variation in AsA levels and attributable to altered gene expression.

Expression pattern of SlbHLH59

Phylogenetic analysis showed that SlbHLH59 belongs to the basic helix-loop-helix (bHLH) family transcription factors and showed highest amino acid similarity with UNE12 (UNFERTILIZED EMBRYO SAC 12) which is responsible for the regulation of fertilizationin processes in Arabidopsis [43] (S8B Fig and S1 File). We investigated the spatial and temporal expression patterns of SlbHLH59 in high and low AsA content accessions (high-AsA accession TS-265 and low-AsA accession TS-186). SlbHLH59 showed high expression in leaves but low in different fruit developmental stages, with the transcript level of SlbHLH59 higher in most tissues of TS-265 versus TS-186 (Fig 4A), supporting a role of SlbHLH59 in positively regulating AsA content in tomato. A previous study revealed that light plays a critical role in regulating AsA metabolism and accumulation [44]. To assess whether SlbHLH59 is involved in light-dependent ascorbate metabolism, we analyzed the expression of SlbHLH59 under successive illumination circulation. Interestingly, we observed light-suppressed SlbHLH59 expression, i.e. it rapidly decreased in the light and increased under dark (S9 Fig). These results explained in part the molecular mechanism of light-dependent accumulation of AsA in tomato.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. The characterization of SlbHLH59.

(A) The transcript levels of SlbHLH59 indifferent tomato organs: R, root; S, stem; L, leaf; FL, flower; IMG, immature fruit; MG, mature green fruit; BR, breaker stage fruit; YR, yellow stage fruit; RR, red ripe stage fruit. TS-186 is a low-AsA accession; and TS-265 is a high-AsA accession. (B, C) Subcellular co-localization of transiently expressed SlbHLH59^SNP9A-YFP (B) and SlbHLH59^SNP9G-YFP (C) fusion protein with a nuclear marker (ERF) in N. benthamiana leaves. Bars = 50 μm. (D-F) Phenotype comparisons of SlbHLH59-OE and wild-type plants after methyl viologen (MV) treatment. The SlbHLH59-OE transgenic plants had less yellow lesion (D, E) and H₂O₂ accumulation (F) than WT after MV treatment. Bar = 10 cm (D), 5cm (E) and 2cm (F).

https://doi.org/10.1371/journal.pgen.1008149.g004

In order to investigate SlbHLH59 cellular localization, we created SlbHLH59^SNP9A-YFP and SlbHLH59^SNP9G-YFP fusion proteins, which were transiently expressed in Nicotiana benthamiana. Fluorescent signals of YFP overlapped with that of ERF-RFP, a marker for the nucleus, suggesting that both SlbHLH59^SNP9A and SlbHLH59^SNP9G were located in the cell nucleus (Fig 4B and 4C).

When plants were exposed to oxidative stress, AsA acts as an antioxidant protecting cells from oxidative damage by scavenging excessive reactive oxygen species [10]. To evaluate whether overexpressed SlbHLH59 in tomato can increase tolerance to oxidative stress, 1-month-old A-OE lines (A-OE5 and A-OE9) and wild type (TS-186) were subjected to oxidative stress by treatment with 75 μM methyl viologen (MV) for 2 days. DAB staining showed that there was no significant difference between TS-186 and A-OE lines under normal conditions (treatment with ddH₂O), but more brown spots in the leaves of TS-186 were observed than in the leaves of A-OE after MV treatment (Fig 4D–4F). Additionally, the content of chlorophyll and MDA were not significantly altered in the A-OE lines, but significantly decreased and increased in the wild-type after MV treatment, respectively (S10 Fig). These results demonstrate that SlbHLH59 induces AsA accumulation facilitating increased oxidative stress tolerance.

SlbHLH59 regulates AsA biosynthesis by directly modulating the expression of SlPMM, SlGMP2 and SlGMP3

bHLH transcription factors have previously been reported to recognize and bind to the E-box cis-element (CANNTG), thereby affecting the expression of downstream genes [45]. To test whether the expression of structural genes in the AsA biosynthesis pathway were modulated in the SlbHLH59 transgenic plants, we performed qPCR analysis. The expression of SlPMI, SlPMM, SlGMP1, SlGMP2, SlGMP3, SlGMP4 and SlGME1 were higher and lower in the fruits of the SlbHLH59-OE and SlbHLH59-RI lines, respectively, when compared with wild-type fruits, suggesting transcriptional regulation mediated by SlbHLH59 on the AsA biosynthetic genes (Fig 5A and 5B). Using PLACE (http://www.dna.affrc.go.jp/PLACE/signalscan.html) and PlantCARE (http://bioinformatics.psb.ugent.be/webtools/plantcare/html/) software, we analyzed the cis-elements in the promoters of SlPMI, SlPMM, SlGMP1, SlGMP2, SlGMP3, SlGMP4 and SlGME1, and determined that all of the genes were predicted to harbor the E-box cis-elements (S12 Table). We speculated that SlbHLH59 might bind and regulate these genes of the D-Man/L-Gal pathway. To test this hypothesis, we conducted yeast one-hybrid (Y1H) analysis to test the binding activity of SlbHLH59 protein to the promoters of SlPMI, SlPMM, SlGMP1, SlGMP2, SlGMP3, SlGMP4 and SlGME1. Cis-elements from the SlPMM, SlGMP2 and SlGMP3 promoters were bound by SlbHLH59 (Fig 5C and 5D). Also, the physical interactions between SlbHLH59 and the promoter fragments derived from SlPMM, SlGMP2 and SlGMP3 were detected by using the dual luciferase system (Fig 5E). These results indicate that SlbHLH59 can directly bind to the SlPMM, SlGMP2 and SlGMP3 promoter to modify their expression, and thereby positively regulate tomato fruit AsA content.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Identification of SlPMM, SlGMP2 and SlGMP3 as direct targets of SlbHLH59.

(A, B) Relative expression of ascorbate (AsA) biosynthesis, recycling and oxidation-related genes in fruits of SlbHLH59 overexpression (A) and RNAi (B) lines. (C) The promoter cis-elements of SlPMM, SlGMP2 and SlGMP3. The sequences in black boxes represent E-box cis-elements. (D) Yeast-one hybrid (Y1H) assay of SlbHLH59 binding to promoter fragments of SlPMM, SlGMP2 and SlGMP3. The bait vectors SlPMM_pro, SlGMP2_pro and SlGMP3_pro (the fragments containing the region 3000 bpupstream from initiation codon) and the prey vector containing SlbHLH59 were introduced into yeast strain Y187, and interaction between bait and prey enhanced ABA resistance. Yeast cells spread on SD-Leu media with various concentrations of ABA (0, 10, 15, and 20 mM). The bait vector (SlPMM_pro, SlGMP2_pro and SlGMP3_pro) + pGADT7 were also transformed into Y187 as a negative control. (E) Binding of SlbHLH59 to promoters of SlPMM, SlGMP2 and SlGMP3 assayed by dual luciferase system. The SlbHLH59 ORF was cloned into the effector vector (pGreen II62-SK) and promoter fragments of SlPMM, SlGMP2 and SlGMP3 were inserted into the reporter vector (pGreen II 0800 LUC). Both vectors of effectors and reporters were transformed into Agrobacterium cells and used to infiltrate tobacco leaves. LUC, firefly luciferase activity; RLU, Renilla luciferase activity; PG, the empty vector of pGreenII 62-SK. The promoters of SlPMM, SlGMP2 and SlGMP3 plus PG were used as control. (F-J) Dynamics of ascorbate content and SlbHLH59 expression during tomato fruit development. Ascorbate (AsA) concentration (F) and expression of SlbHLH59 (G), SlPMM (H), SlGMP2 (I), SlGMP3 (J) at different fruit developmental stages of wild-type (TS-186) and SlbHLH59 transgenic lines. Experiments were performed in immature (IMG), mature green (MG), breaker (BR), yellow ripe (YR) and red ripe (RR) fruit, respectively, with three replicates. All data in the graphs are presented as means ± SE. Asterisks indicate significant differences by t test: *P < 0.05; ** P< 0.01.

https://doi.org/10.1371/journal.pgen.1008149.g005

To test how SlbHLH59 affects SlPMM, SlGMP2 and SlGMP3 during the fruit development in tomato, we determined total AsA and the expression of SlPMM, SlGMP2 and SlGMP3 in immature green stage (IMG), green mature stage (MG), breaker stage (BR), yellow ripe stage (YR) and red ripe stage (RR) fruits. The G-OE4 showed higher total AsA content than TS-186 throughout the whole fruit development stages and the greatest difference was observed at MG stage, consistent with the dynamic change of SlbHLH59 transcript level (Fig 5F and 5G). The expression of SlPMM, SlGMP2 and SlGMP3 were all significantly higher in G-OE4 than in TS-186 except SlPMM and SlGMP2 in IMG fruits, and the greatest difference was observed at MG stage, consistent with the dynamic change of AsA content in TS-186 and SlbHLH59 overexpression line (Fig 5H–5J). This result supports the notion that SlbHLH59 contributes to AsA biosynthesis by directly regulating SlPMM, SlGMP2 and SlGMP3 during fruit development with the highest effect at the MG stage just prior to the onset of ripening.

Human selection of SlbHLH59^InDel_8 in tomato

Since InDel_8 represents a functional polymorphism of the TFA9 locus for AsA biosynthesis via SlbHLH59 expression in tomato fruit, we investigated InDel_8 variants in 540 accessions, including 333 BIG, 141 CER, 54 PIM and 12 accessions of wild tomato species (Fig 6A and S13 Table). Only 6 accessions in the BIG group carried insertion_8, as TS-265, and all other accessions carried deletion_8, as TS-186. Significant differences in both total and reduced AsA were detected between the eight insertion_8 BIG accessions (BIG^insertion_8) and TS-186 (BIG^deletion_8) (Fig 6B). All 12 accessions of wild species carried insertion_8 but the deletion_8 was detected in 9 of 54 tested PIM accessions, suggesting the deletion of InDel_8 occurred in early domestication during the time the PIM group differentiated from wild species. Thirteen of the 23 CER^insertion_8 carrying accessions originated from South America, including those from Ecuador and Peru where wild tomato relatives originated. Thus, it is likely that the deletion_8 initially occurred in South America and subsequently spread into other neighbouring countries.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Nucleotide diversity analysis of TFA9 loci during tomato domestication and improvement.

(A) Frequency of derived and ancestral InDel_8 allele in tomato subpopulations. n = number of accessions. (B) Comparison of ascorbate content of Hap 1 accession TS-186 and six Hap 4 accessions in BIG subgroup. The data are presented as means ± SE (n = 3). The asterisks represented significant differences from TS-186 (Hap 1), as indicated by the t-test (** P< 0.01). (C) Distribution of nucleotide diversity (π) of the PIM (red line), CER (purple line), and BIG (blue line) within the 10-kb region of TFA9. Grey lines indicate the locations of indel_8 (TCTCTTTC/-). The X-axis denotes the position of SlbHLH59 and the Y-axis indicates average p values. (D) The ratio of nucleotide diversity (π) is calculated from the SlbHLH59 sequence of PIM with CER (red line) and CER with BIG (blue line). In total, 331 tomato accessions were used for analysis, including 53 PIM accessions, 112 CER accessions, and 166 BIG accessions.

https://doi.org/10.1371/journal.pgen.1008149.g006

The high fruit AsA-associated insertion_8 was present at a highest frequency of 83.3% (45/54) in wild progenitor variety PIM, but the frequency sharply decreased in CER accessions (16.3%, 23/141) that were domesticated from the wild progenitor variety PIM and BIG landraces (1.8%, 6/333) (Fig 6A). The result revealed that the less fruit AsA content was selected during domestication. To examine the evolutionary history of the TFA9 locus, DNA sequence variation in the genomic region spanning 10 kb upstream fragment and coding region of the SlbHLH59 was investigated. On average, the nucleotide diversity (π) in the coding and promoter region of SlbHLH59 was much lower in BIG (π = 0.11×10⁻³) than in CER (π = 0.65×10⁻³) and PIM (π = 1.32×10⁻³) (Fig 6C). The highest ratios of nucleotide diversity in PIM to CER (π_PIM/π_CER, 4.03) and CER to BIG (π_CER/π_BIG, 13.74) occurred at -2 kb to -1 kb region of SlbHLH59, where the indel_8 was located on (Fig 6D). This result was consistent with previous report that SlbHLH59 was pointed out as a domestication and improvement sweep at the whole genome level [26]. Moreover, Tajima’s D of the region from -2 kb to -1 kb was negative in BIG (-1.53) subgroup but positive in the CER (0.46) and PIM (2.33) subgroup, respectively (S14 Table). Taken together, these results suggested that TFM6 region has been subjected to human selection during the domestication of tomato.

Plant materials and growth conditions

A total of 302 tomato accessions, including 171 Solanum lycopersicum (BIG), 104 S. lycopersicumvar.cerasiforme (CER) and 27 S. pimpinellifolium (PIM) accessions that were selected from the previously described 360 accessionsand used for GWAS in this study. The GWAS was conducted at two sites: E1 (Spring 2013, open field at Huazhong Agricultural University, Wuhan, China) and E2 (Spring 2016, greenhouse at the Agri-Academy of Sciences of Wuhan, China). A F₂ population of 1,587 individuals was derived from a cross between TS-21 (high SSC) and HG22 (low SSC) conducted in an open field at Huazhong Agricultural University in spring 2014. For GWAS and BSA, at least three fruits from at least three plants per line were harvested at the ripe stage. For fruit development analysis, flowers were tagged at the full-bloom stage to synchronize developmental stages. The fruits were harvested at immature green (IMG, 21 DAF), mature green (MG, 37 DAF), breaker (BR, 40 DAF), yellow ripe (YR, 42 DAF), and red ripe (RR, 49 DAF) developmental stages. Three biological replicates of each developmental stage were analyzed. After tissue selection, the outer pericarp was bulked (five fruits) and stored at –80°C for metabolic and transcript profiling. The remaining fresh red fruits were used directly to measure SSC, total sugars, total acids, and sugar/acid ratio.

Sample preparation and metabolite profiling

Red fruit metabolite profiling of the 302 tomato accessions was performed by GC-MS using the method as described previously [27]. Tomato fruit juice samples were used to measure the SSC, total sugars, total acids, and sugar/acid ratio. SSC was determined within the GWAS population and F₂ segregation population using a hand saccharimeter (B429335, ATAGO). The total sugar and total acid contents were measured using a Brix-Acidity Meter (PAL-BX|ACID3, ATAGO), followed by calculation of sugar/acid ratios. Total sugars and SSC in the fruit juice were measured by directly dropping the sample onto the meter and recording the value. The fruit juice was then diluted 50-fold with ddH₂O water and used to determine total acids. The AsA levels were measured as previously described [44].

Data processing and statistical analyses

The relative content of each metabolite was obtained by comparing the peak area of each metabolite with the peak area of internal standard (ribitol). For each phenotype, normal distribution of the data was tested using a Shapiro-Wilk test. The normality test revealed that 34 of the 92 phenotypes (37%) were not normally distributed and were Box-Cox transformed. The coefficient of variation values was calculated independently for each metabolite (using the mean of the biological replicates of the untransformed m-trait data) as follows: σ/μ, where σ and μ are the s.d. and mean of each metabolite in the population, respectively. Broad-sense heritability (H²) was calculated using the following equation by treating accessions as a random effect and the biological replication as a replication effect using one-way ANOVA: H² = Var_(G) /(Var_(G) + Var_(E)), where Var_(G) and Var_(E) are the variance derived from genetic and environmental effects, respectively. Finally, differences in the metabolic traits among the six subgroups (PIM, SA_CER, NSA_CER, UO_CER, NP_BIG, and P_BIG) were analysed by ANOVA tested. Significance was declared at P <0.05. The LD heatmap surrounding candidate gene in the GWAS was constructed using Haploview 4.2 with default parameters [54], indicating r² values between pairs of SNPs multiplied by 100.

Association mapping

To facilitate SNP identification and genotype imputation, two sequencing data sets were used in this study. The first was downloaded from a diverse global collection of 360 tomato accessions (NCBI Sequence Read Archive [SRA] under accession number SRP045767) [26]. Additional sequence data from 398 varieties were generated by Tieman et al. [20] was downloaded from the National Center for Biotechnology Information BioProject site under accession number PRJNA353161. SOAP2 was used to map all sequencing reads from each accession to the tomato reference genome (Version SL2.50) with the following parameters: -m 100, -x 888, -s 35, -l 32, -v 3. Mapped reads were filtered to remove PCR duplicates. Both paired-end and single-end mapped reads were used for SNP calling throughout the entire collection of tomato accessions using SOAPsnp with the following parameters: -L 100 -u -F 1(23). After imputation, SNPs with missing rates of less than 20% were selected, resulting in a total of 4,180,023 SNPs (MAF > 0.05, the number of varieties with the minor allele ≥ 6) for further analysis. Detailed information on called SNPs can be viewed and downloaded from Sol Genomic Network (https://solgenomics.net/).

Association analyses were performed using the compressed MLM [55] with TASSEL 4.0 [56]. Suggestive (1/n, ≤2.4×10⁻⁷) and significant (0.05/n, ≤1.2×10⁻⁸) P-value thresholds were defined to control the genome-wide type 1 error rate (n = total number of markers used) [51, 57]. We used Haploview software to perform local LD analysis [54] and calculate LD accordingly with modification [20]. Briefly, the average linkage decay for each 0.5 Mb region of the whole genome was evaluated with the following parameters: -maxdistance 2000 -minMAF 0.05 -hwcutoff 0. Pairwise LD between the suggestive/significant SNPs for each metabolic trait was calculated. The physical locations of the SNPs were identified based on tomato genomic sequence version SL2.50 (http://solgenomics.net/).

Linkage mapping

An F₂ population of 1,587 individuals derived from a cross between TS-21 (a high-SSC accession) and HG22 (a low-SSC accession) was planted in the Spring of 2014 in an open field at Huazhong Agricultural University, China. For each individual, the average SSC of three representative fruits was recorded, and genomic DNA was isolated from fresh leaves using the CTAB method. For bulked segregant analysis, bulk DNA samples for high- and low-SSC accessions were constructed by mixing equal amounts of DNA from 50 individuals showing extremely high and low SSC, respectively. Subsequently, 34.97× genome sequences for TS-21, 35.56× genome sequences for HG22, and roughly 60× genome sequences for each bulk sample (high-SSC fruit and low-SSC fruit) were generated by BIOMARKER Company (Beijing, China). The SLAF (Specific-Locus Amplified Fragment) label was located on the reference genome using SOAP, and labels that were sequenced <5× in the parent were filtered out. Short reads were aligned against the reference genome (release SL2.50) using the Burrows-Wheeler Aligner (BWA). The ΔSNP index was obtained by subtracting the SNP index of the low-SSC bulk sample from that of the high-SSC bulk sample. The average SNP index for the high-SSC and low-SSC bulk samples was calculated using a 1,000-kb sliding window with a step size of 10 kb. The statistical confidence intervals of the ΔSNP index were calculated under the null hypothesis of no QTLs, and 0.32 was then set as the threshold.

Epistasis analysis

For each metabolic trait in each environment, the pairwise additive-by-additive epistatic interactions were investigated for all identified loci. Epistatic interactions were determined by two-way analysis of variance (ANOVA) (using P< 0.05 as a significance threshold) using all significant loci in pairwise combinations. The proportion of variance explained by epistasis was tested by comparing the residual of the full model containing all single-locus effects and two-locus interaction effects with that of the reduced model containing all single-locus effects but excluding two-locus interaction effects [58].

DNA sequencing

To detect the variation in SlbHLH59 gene region (chromosome 9: 64,094,000–64,101,000, release SL2.50), DNA sequences of SlbHLH59 in 30 tomato accessions (15 high-AsA and 15 low-AsA accessions) were amplified by PCR using primers listed in (S15 Table). The PCR products were sequenced and compared against the reference genome for polymorphism analysis. In addition, DNA sequences of SlbHLH59 in 367 tomato accessions (250 BIG accessions, 94 CER accessions and 23 PIM accessions) and the data for genotype analysis of InDel_8 in 540 tomato accessions were downloaded from the public database (National Center for Biotechnology Information BioProject site under the accession PRJNA353161).

Gene cloning, vector construction, and transformation

For overexpression construct, completed SlbHLH59 open reading frame (ORF) were amplified from the cDNA of tomato (TS-186 for SlbHLH59^SNP9A and TS-265 for SlbHLH59^SNP9G), and then incorporated into the pHELLSGATE8 vector using homologous recombination (ClonExpress II One Step Cloning Kit, Vazyme). For RNAi construct, a 200-bp fragment of SlbHLH59 was amplified by SlbHLH59-RI primers (S15 Table), and then cloned into the pHGRV vector using BP Clonase according to the manufacturer’s instructions (Invitrogen, USA). All the recombinant constructs were transformed into Agrobacterium strain C58 by electroporation and subsequently transformed into the tomato genome (TS-186 for overexpression and Ts-265 for RNAi) using cotyledon explants as described previously [59]. Transgenic plants were confirmed by PCR using CaMV35S promoter forward primer and SlbHLH59 specific reverse primer (S15 Table).

RNA isolation and gene expression analysis

Total RNA was extracted from different accessions and transgenic lines using TRIZOL reagent (Invitrogen, USA). Gene expression was investigated by qRT-PCR. The primer pair sequences (designed using Primer Premier 3.0 [http://frodo.wi.mit.edu/primer3]) are listed in S15 Table. The cDNA synthesis and qRT-PCR steps were performed as previously described [60]. The Actin gene (Solyc11g008430) was used as an internal standard and qRT-PCR was performed with three repeats per gene (including ACTIN).

Subcellular localization

The coding sequences of SlbHLH59^SNP9Aand SlbHLH59^SNP9G without the stop codon was amplified from the cDNA of tomato (TS-186 for SlbHLH59^SNP9A and TS-265 for SlbHLH59^SNP9G) by PCR and then cloned into the expression vector p101YFP under the control of the CaMV35S promoter by homologous recombination. CaMV35S: SlbHLH59-YFP vector as well as cell nucleus marker CaMV35S: ERF-YFP was transformed into Agrobacterium tumefaciens strain GV3101 and co-infiltrated into leaves of N. benthamiana with the suspension as previously described [61]. After 48 h incubation at 25 °C, the tobacco leaves were used for YFP and RFP fluorescence signal observation using Leica Confocal software. CaMV35S:YFP acted as positive control.

Yeast one-hybrid assay

The yeast one-hybrid assay was performed as described in the Matchmaker One-Hybrid Library Construction and Screening Kit (Clontech). Briefly, the full-length of SlbHLH59 ORF sequence (amplified from TS-186 cDNA) and promoter sequences of SlPMM, SlPMI, SlGME1, SlGMP1, SlGMP2, SlGMP3 and SlGMP4 (amplified from TS-186 genomic DNA) were cloned into the pGADT7 and pAbAi vector (Clontech), respectively. The pAbAi bait vectors were introduced into the GOLD1 yeast and cultured on SD/–Ura. The pGADT7 prey vector was introduced into yeast strains containing pAbAi bait vectors and cultured on SD/–Leu. After 4 d incubation, the positive yeast strains were picked and diluted in double-distilled water to an OD600 of 0.1, and 2 μL of suspension was spotted on SD/–Leu, with or without ABA (0–20 ng/mL) (Sigma-Aldrich), followed by 3 to 7 d incubation at 30°C.

Transient expression in N. benthamiana leaves and Arabidopsis protoplasts

The full-length SlbHLH59 ORF was cloned into the effector vector pGreen II 62-SK under the control of CaMV 35S promoter. SlPMM, SlGMP2 and SlGMP3 promoter fragments were PCR amplified using specific primers and cloned into the reporter vector pGreen II 0800-LUC. Individual combinations of effector and reporter vectors were co-transformed into Agrobacterium GV3101 cells alongside the pSoup vector, and the transformed GV3101 cells were used to infiltrate young N. benthamiana leaves, in which transient expression was analyzed following a 2-d incubation. Firefly and Renilla luciferase signals were assayed with the dual luciferase assay reagents (Promega) using an Infinite M200 (Tecan).

The promoter activity analysis was carried out as described previously [62]. GUS activity and LUC activity were determined by Fluorescence FLx800 microplate fluorescence reader (BIO-TEK Instruments). Ratios of GUS to LUC activities were used to define relative promoter activity. Three biological replicates were performed for each construct. The cis-element analysis was conducted in PLACE (http://www.dna.affrc.go.jp/PLACE).

Light response and oxidative stress treatment

One-month-old seedlings ofSlbHLH59^SNP9A overexpression plants as well as TS-186 and TS-265 plants were grown in plastic pots in the greenhouse. For the light response characterization, from 8 am, the plants were exposed to continuous light at 25°C for 12 h followed by 12 h continuous dark under 25°C and recovered by12 h continuous light at 25°C. In a 36-h photoperiod, samples were taken every four hours (8:00, 12:00, 16:00 and 20:00 under light, 24:00 and 4:00 under light) to determine the expression of SlbHLH59 in tomato leaves.

To evaluate the performance of SlbHLH59^SNP9A overexpression plants and wild type (TS-186) plants under oxidative stress, the plants were sprayed with 75 μM MV (MV dissolved in water with 0.1% Tween-20) or water with 0.1% Tween-20 (control) once a day for 2 days. Phenotype was investigated and recorded one week after the end of the treatment. For the 3–3’-diaminobenzidine (DAB) staining, the leaves were cleaned and placed in 1 mg/mL DAB, pH 3.8, under light at 25°C for 8 h. The experiment was terminated by immersing the leaves in boiling 96% ethanol for 10 min. After cooling, the leaves were placed in fresh 96% ethanol for 4 h at room temperature and photographed. The deep brown polymerization product was produced via the reaction of DAB with H₂O₂. Also, leaves were collected and ground into fine powder in liquid nitrogen after the treatment. To assay chlorophyll levels, 1 ml of 80% (v/v) acetone was added to approximately 0.1 g of frozen powder in a 2-ml Eppendorf tube under low light intensity by the procedure described by Wellburn [63]. The MDA levels were measured as previously described [64].

Molecular diversity analysis

For the molecular diversity analysis, the π ratios and Tajima’s D [65], were used to identify the selective sweeps in SlbHLH59 associated with tomato domestication and improvement events. Briefly, π (π_PIM, π_CER and π_BIG) and Tajima’s D (Tajima’s D_PIM,Tajima’s D_CER and Tajima’s D_BIG) were calculated using DnaSP5.0 version 5.0 software [66], with a sliding window length of 100 bp and step size of 25 bp.