Cultivated tomato varieties accumulate less Cd in shoot than wild ancestors
Here, we investigated the differences in Cd accumulation capacity among 506 tomato accessions. These 506 accessions represent various geographical origins and improvement status, consisting of 53 wild accessions of Solanum pimpinellifolium (PIM), 179 domesticated accessions of Solanum lycopersicum var. cerasiforme (CER), and 274 improved accessions of Solanum lycopersicum (BIG) (Fig. 1A, fig. S1, and table S1). We found that the Cd content in the roots of the BIG group was significantly higher than that of the PIM and CER groups, while the Cd content in the shoots of the BIG group was significantly lower than that of the PIM and CER groups. Additionally, the Cd translocation factor of the BIG group was significantly lower than that of the PIM and CER groups. There were no significant differences observed in the Cd contents in the roots and shoots, as well as the translocation factor between the PIM and CER groups (Fig. 1, B to D, and table S2).
Taking the median value (3.0463 mg/kg) of Cd content in the shoots of 506 tomato accessions as the threshold, we found that the proportion of “low Cd” tomato (Cd content < 3.0463 mg/kg) increased gradually from the PIM group to the CER group and then to the BIG group, especially during the transition from CER to BIG, where it increased from 32.40–64.96% (Fig. 1E). This discovery suggests that unintentionally, as humans have domesticated and improved tomato yields, they have preserved the excellent trait of “low Cd” during the process. This is promising as it indicates that we can use existing cultivars for breeding to rapidly develop low Cd accumulating cultivars. Thus, the domestication and improvement of species for higher yields has not always resulted in the loss of desirable but invisible traits such as flavor(10) or stress resistance(11, 12), and may fortunately have led to the preservation of outstanding traits like “low Cd”.
Identification of a major locus for low Cd accumulation in tomato
The “Cd accumulation” trait has received relatively little attention during the process of tomato domestication and breeding. To understand why this trait underwent directional selection, we investigated the genetic mechanisms behind low Cd accumulation in tomato. Genome-wide association study (GWAS) is an effective strategy for uncovering the genetic basis of ion accumulation in plants(11, 13, 14). Here, 5,096,140 common SNPs with a minor allele frequency (MAF) > 0.05 and missing ratio < 10% were used to perform a GWAS. The P-values of 1.0 × 10− 7 were set as the significance threshold after Bonferroni-adjusted correction (Fig. 2A and fig. S2). One strong association signal (named LCT1 (Low Cadmium in Tomato 1)) on chromosome 3 was observed in the GWAS of Cd content in shoots, which includes 27 significant SNPs (Fig. 2A). We then analyzed the physical positions, genomic elements, and functional annotations surrounding the significant SNPs to select candidate genes underlying Cd content in shoots. Notably, three significant SNPs were found within the promoter region (SNP15-T), exon 2 (SNP16-T), and 3’UTR (SNP3-A) of the gene Solyc03g115220 (fig. S3 and table S3). The SNP15-T, located within the promoter region, is predicted not to reside within any known cis-regulatory elements. SNP16-T is a synonymous mutation. SNP3-A is predicted to be a binding site for miR8762, which is supposed to disrupt the miR8762 binding (fig. S4 and table S3). Haplotype analysis reveals that accessions with the 3’UTRAA genotype had lower Cd content compared to those with the 3’UTRTT genotype (Fig. 2B). The Solyc03g115220 encodes a flavonoid 3’-hydroxylase (SlF3’H), which belongs to the CYP450 family. Previous studies have reported that CYP450 is involved in regulating ion transport processes (15, 16). Hence, we focused on SlF3’H to understand its potential genetic mechanism underlying low Cd accumulation in tomato.
We generated two tomato SlF3’H overexpression (SlF3’H-OE) lines in background of “high Cd” accession TS-222 and confirmed their elevated SlF3'H RNA levels (fig. S5A). We measured the Cd accumulation abilities of SlF3’H-OE#1 and #2 lines together with TS-222 and found that overexpressing SlF3’H in tomato reduced the Cd content in the shoots while it had no significant effect on the Cd content in the roots (Fig. 2, C to F). Additionally, a functional experiment was performed by using CRISPR/Cas9 to knock out SlF3’H in TS-222 (fig. S5B). Interestingly, the successful knockout of SlF3’H also reduced the Cd content in the shoots (fig. S5, C to J). We speculate that this is due to the SlF3’H-OE and SlF3’H-CR lines affecting different pathways of Cd accumulation. The SlF3’H-OE lines likely function primarily through the cell wall and chelation pathways, whereas the SlF3’H-CR lines likely function mainly through the transporter pathway (fig. S6). Therefore, we proposed that SlF3’H was the causal gene for SlLCT1.
To validate the function of the flavonoid hydroxylase F3’H in other species, the rice orthologs, OsCYP75B2 and OsCYP75B4, were overexpressed in rice ZH11 (Oryza sativa L. ssp. japonica), and AtCYP75B1 was overexpressed in Arabidopsis, respectively (fig. S7). The results demonstrated that overexpressing OsCYP75B2 and OsCYP75B4 both reduced the Cd content in the leaves, stems, and grains of rice, while overexpressing AtCYP75B1 decreased the Cd content in the shoots of Arabidopsis (Fig. 2, G to L). Thus, these results suggested that the function of F3’H in reducing plant Cd accumulation is relatively conserved across different species. This could potentially provide a solution to the current issue of excessive Cd in rice production (17).
SlF3’H expression variance: a key to low Cd in cultivated tomato
In our analysis of SlF3’H expression levels in tomato, we observed that “low Cd” accessions exhibited higher expression levels relative to “high Cd” accessions, and that Cd treatment induced the expression of SlF3’H (Fig. 3, A and B). The 3’UTR of transcripts typically functions as a binding site for miRNA, negatively regulating gene expression (18, 19). Hence, employing the dual-luciferase reporter assay system, we fused the two 3’UTR sequences (3’UTRTT and 3’UTRAA) of SlF3’H with the luciferase gene sequence and co-expressed them separately with miR8762 in tobacco (Nicotiana tabacum) leaves (fig. S8). The results indicated that miR8762 exerted inhibitory effects on the luciferase fused with both types of 3’UTRs, but the luciferase construct fused with the 3’UTRAA showed a weaker inhibition (Fig. 3, C and D). Therefore, we speculate that the 3’UTR variation (SNP3-A) of SlF3’H buffered the inhibitory effect of miR8762 on SlF3’H expression, resulting in a reduced accumulation of Cd in tomato.
Promoter-GUS analysis revealed that SlF3’H is primarily expressed in the tomato hypocotyl (Fig. 3F) and subcellular localization indicated that SlF3’H is mainly located on the endoplasmic reticulum membrane (Fig. 3E). In hypocotyl cross sections, GUS activity was primarily detected in the vascular bundle sheath (Fig. 3, G and H). This implied that SlF3’H might be involved in the transport of Cd from the tomato root to the shoot.
Co-selection of LCT1 and fruit weight locus fw3.2
The findings above indicated that SlF3’H is the major gene controlling low Cd accumulation in tomato and plays a significant role in the low Cd accumulation in important crops like tomato and rice. Accordingly, we carried out an investigation to evaluate whether the trait of “Cd accumulation” in tomato, which has received little attention from breeders, has undergone directional selection. Since SNP3-A is supposed to be the casual variation underling the Cd content, the evolution and selection of 3'UTR of SlLCT1 was investigated.
We found that the majority of accessions in the PIM group belong to the SlLCT1TT genotype, while most of the accessions in the BIG group had the SlLCT1AA genotype (Fig. 4A). Throughout the domestication and improvement of tomato, there has been a gradual increase in the prevalence of the SlLCT1AA genotype, particularly during the transition from CER to BIG, where it rose from 20.67–85.77% (Fig. 4A). This is consistent with the trend for the change in proportion of tomato with “low Cd” phenotype in the three groups (Fig. 1E). We hypothesized that this phenomenon may be attributed to genetic hitchhiking, since there is no record of targeted selection for Cd accumulation traits during the tomato domestication and improvement process. Thus, our focus turned to fruit weight, a trait closely linked to yield, which is a major target for breeders. A significant negative correlation between fruit weight and Cd content in shoots was found among the 164 tomato accessions (Fig. 4B), providing preliminary validation for our hypothesis.
For genetic hitchhiking to occur, it is essential that two loci are located on the same chromosome and separated by a relatively short genetic distance (20, 21). In our study, we investigated QTLs associated with yield and identified a QTL, fw3.2, on chromosome 3 that controls fruit size, with a highly correlated SNP (ch03_64799226) in its promoter region (22, 23) (table S4). The linkage disequilibrium (LD) analysis revealed a tight linkage between the SNP (ch03_64799226) in the promoter of fw3.2 and the SNP (ch03_65001793) in the 3’UTR of SlF3’H (Fig. 4C). We examined the distribution of fw3.2CC and fw3.2TT alleles in the PIM, CER, and BIG accessions. The prevalence of the fw3.2TT allele, representing large fruit, gradually increased within the population, notably rising from 24.02–92.34% during the transition from CER to BIG (Fig. 4D). This trend closely paralleled changes in the prevalence of the LCT1 within the population (Fig. 4A). Accessions carrying the fw3.2CC allele mostly co-carried the LCT1TT allele, associated with “high Cd”, whereas accessions carrying the fw3.2TT allele predominantly co-carried the LCT1AA allele, linked to “low Cd” (Fig. 4E). Furthermore, analysis of near-isogenic lines (NILs) of LCT1 indicated that all accessions of the NIL (TT) genotype contain the fw3.2CC allele, while all accessions of the NIL (AA) genotype contain the fw3.2TT allele (Fig. 4F and fig. S9). We scanned the regions surrounding the LCT1 and fw3.2 genes on chromosome 3 and compared the ratios of nucleotide diversity between PIM and CER accessions, as well as between CER and BIG accessions. The results showed that both fw3.2 and LCT1 were found to be in the region associated with improvement sweep (Fig. 4, G and H). This suggests that fw3.2 and LCT1 experienced similar selection pressures during the improvement process from CER to BIG in tomato. Furthermore, the geographical distribution indicates that LCT1TT is mainly distributed in regions where tomato originate, such as Peru and Ecuador, whereas LCT1AA is mainly found in the United States, Russia, and along the Mediterranean coast (e.g., in Italy) (Fig. 5A). Thus, we proposed that the “low Cd” genotype LCT1AA has been retained in tomato during long-term domestication, and underwent strong selective pressure by genetic hitchhiking with fw3.2TT during the breeding process from CER to BIG. These events have led to the dominance of the “low Cd” genotype SlF3’HAA in cultivated tomato today (Fig. 5, A and B).