Fruit weight regulation by a paralog of Cell Size Regulator (CSR) in tomato and other crops
Qian Feng, Lara Pereira, Manoj Sapkota, Krishna Sai Karnatam, Yanbing Wang, Padma Nimmakayala, Umesh K. Reddy, Esther van der Knaap
TL;DR
A gene called CSR-like1 influences tomato fruit weight by increasing cell size and is likely important in other crops too.
Contribution
CSR-like1 is identified as a new gene controlling fruit weight and is conserved across crops.
Findings
CSR-like1 increases fruit weight by enlarging pericarp cells in tomatoes.
The CSR-like1-D allele was selected during tomato domestication.
A pepper study found a similar gene linked to fruit weight.
Abstract
A paralog of Cell Size Regulator (CSR), CSR-like1, underlies the novel fw6.2 QTL in tomato. The gene and locus regulate fruit weight by increasing pericarp cell size and its function on fruit weight appears to be conserved in other crops. Fruit weight is a quantitative trait that was under strong selection during the domestication of fruit and vegetable crops such as tomato (Solanum lycopersicum). While numerous fruit weight QTLs have been identified, only three tomato fruit weight genes have been cloned. In this study, we utilized a genetically diverse tomato panel, the Varitome collection, to identify additional genetic loci that control fruit weight. We mapped and fine mapped two fruit weight QTLs on chromosome 6, fw6.1 and fw6.2, by using Genome Wide Association studies (GWAS) and linkage mapping in bi-parental populations. We identified a member of the Cell Size Regulator family,…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6- —http://dx.doi.org/10.13039/100000001National Science Foundation
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsPlant Molecular Biology Research · Genetic Mapping and Diversity in Plants and Animals · Postharvest Quality and Shelf Life Management
Introduction
Plant domestication has led to the transformation from wild species into high-yielding and nutritionally valuable crops. Genetic modifications through repeated selection and cultivation have reshaped the plant morphology, physiology, and development to suit human needs (Doebley et al. 2006). Increases in fruit or grain size is one of the key traits that differentiate the domesticated crops from their wild progenitors, collectively referred to as “domestication syndrome” (Doebley et al. 2006; Gross and Olsen 2010). For tomato (Solanum lycopersicum), huge variation in fruit size and shape is evident in the cherry-sized type S. lycopersicum var cerasiforme (SLC*)* and fully domesticated type S. lycopersicum var. lycopersicum (SLL) compared to its wild relative, S. pimpinellifolium (SP) (Tanksley 2004).
Several studies have revealed insights regarding the complex domestication history of tomato resulting in different models. One model illustrated a two-step process (Razifard et al. 2020): SP first originated in the Andean region of South America and gave rise to the semi-domesticated SLC. The more domesticated South American SLC groups then migrated northwards when some domestication traits, including fruit size, were lost. A second round of selection took place in Mesoamerica before the emergence of the fully domesticated SLL. A recent study suggested a more complex trajectory (Blanca et al. 2022): SP expanded much further northwards into Mesoamerica where it evolved into wild SLC progenitor species. Some SLCs subsequently migrated back to South America and admixed with SP in Ecuador and Peru. Ecuadorian and Peruvian SLC then underwent another northward migration to further evolve into fully domesticated SLL in Mexico.
Fruit weight is a character controlled by multiple quantitative trait loci (QTL) (Paran and van der Knaap 2007). Though over twenty QTL have been mapped in tomato (Causse et al. 2002; Grandillo et al. 1999; Illa-Berenguer et al. 2015; Pereira et al. 2021b), only three fruit weight genes have been cloned: fw2.2, fw3.2 and fw11.3. Fw2.2 underlies Cell Number Regulator (CNR) (Frary et al. 2000), a negative cell division regulator with a change in transcription dynamic that is associated with lower cell division rate (Cong et al. 2002). The putative casual mutation is thought to lie in the promoter region since the coding region is highly conserved in small-fruit allele and large-fruit allele (Frary et al. 2000). Fw3.2 (SlKLUH) encodes a P450 enzyme of the CYP78A subfamily that regulates cell number in the fruit (Chakrabarti et al. 2013). A tandem duplication of around 50 kb containing SlKLUH leads to higher expression of SlKLUH and an increase in fruit weight. (Alonge et al. 2020). Fw11.3 encodes Cell Size Regulator (CSR) and is the only fruit weight gene known to affect cell size (Mu et al. 2017). The derived allele (CSR-D) has a 3′ 1.4 kb deletion that affects the coding region as well as higher expression throughout fruit development compared to the wildtype allele (CSR-WT) (Mu et al. 2017). In addition to the fruit weight loci, two locule number loci, lc (Muños et al. 2011) and fas (Cong et al. 2008; Xu et al. 2015), can affect the fruit weight in tomato through regulating the number of locules in the fruit.
Here, we report the fine-mapping and identification of two linked fruit weight QTL in tomato, fw6.1 and fw6.2, and the cloning and characterization of a new fruit weight gene CSR-like1 at the second locus. Similar to its paralog CSR, CSR-like1 also controls the fruit mass through regulating the pericarp cell size. Yet different from CSR, CSR-like1 underwent selection in rather the early stages of domestication and was much higher expressed in fruit tissues. Transgenic plants that downregulated CSR-like1 expression showed decreased fruit weight that developed late during fruit growth, as well as effects on other traits such as ripening. We also present genetic evidence on the function of a pepper ortholog of CSR-like1 on fruit weight. The findings imply a conserved function of CSR-like1 in fruit size regulation in Solanaceae crops.
Materials and methods
Plant materials
The 166 Varitome collection accessions have been described previously (Mata-Nicolás et al. 2020; Razifard et al. 2020). The plants for the GWAS and mapping populations were grown at University of Georgia’s Horticulture Research Farm (Athens, GA, USA), Vidalia Onion and Vegetable Research Center (Lyons, GA, USA), Georgia Mountain Research and Education Center (Blairsville, GA, USA) and University of Florida’s North Florida Research and Education Center-Suwannee Valley (Live Oak, FL, USA). Fresh fruit weight of GWAS population was measured as described in Razifard et al. (2020). Three replicates were grown for each accession and approximately 20 fruits were weighed to calculate the average fruit weight for each replicate. Population development including crossing, generating F_1_ and the subsequent recombinant screening and progeny testing populations were conducted in the University of Georgia (Athens, GA, USA) greenhouses where plants were grown under standard conditions. Phenotypic characterizations including cellular measurement, days to ripening and plant height measurement were performed in greenhouse as well.
Variant calling and genome-wide association study (GWAS)
Short-read sequencing data for the Varitome tomato accessions were retrieved from NCBI (BioProject PRJNA454805) and aligned to the SL4.0 reference genome. The raw reads were aligned and processed for Variant calling for SNPs and INDELs (insertions and deletions) using the pipeline as previously described (Pereira et al. 2021a). Variants were filtered to retain only biallelic sites with missingness < 5%, minor allele frequency (MAF) ≥ 5%, and mapping quality ≥ 30 using bcftools (Danecek et al. 2021). After filtering, 659,002 INDELs were retained. For computational efficiency, SNPs were further thinned to one SNP per kilobase using VCFtools (Danecek et al. 2011), resulting in a final set of 609,034 SNPs. SNP genotypes were exported from VCF files using the R/vcfR package (Knaus and Grünwald 2017). INDELs were encoded as a presence/absence matrix and treated as pseudo-SNPs, with the reference allele representing absence and the alternate allele representing presence of the INDEL. Association mapping was performed using GWASpoly (https://jendelman.github.io/GWASpoly/GWASpoly.html) (Rosyara et al. 2016) under a diploid setting (ploidy = 2). The raw fruit weight data is shown in Supplementary Dataset 1 and the normalized fruit weight data was used for association mapping. The diplo-additive model was used as the primary framework, where genotypes are encoded as 0/1/2 copies with heterozygotes intermediate between homozygotes. This model was chosen for its simplicity, statistical power, and biological interpretability. Although other models were tested, results were broadly consistent; therefore, only the diploid additive model is presented. Population structure was controlled using the first three principal components, and kinship was modeled with a leave-one-chromosome-out (LOCO) approach (Cheng et al. 2013). Genome-wide significance thresholds were established using the Bonferroni correction at α = 0.05.
Analysis of orthologous CSR-like1 gene underlying fruit weight variation in the Capsicum Genus
A total of 594 accessions of Capsicum spp., comprising 82 C. annuum, 35 C. annuum var. glabriusculum, 243 C. baccatum, and 234 C. chinense, were obtained from the USDA–ARS Germplasm Resources Information Network (GRIN), Plant Genetic Resources Conservation Unit (Griffin, GA, USA). These accessions represented a wide range of geographical origins and fruit weight variation. Plants were cultivated for the three seasons in a randomized block design with three replications, and phenotypic data were recorded from five plants per accession in each replication. Genomic DNA was extracted from young leaf tissue, and genotyping-by-sequencing (GBS) was performed following the protocol described by Elshire et al. (2011) using the Illumina HiSeq 2500 platform. The resulting sequence reads were aligned to the Capsicum annuum cv. CM334 v1.55 reference genome (Kim et al. 2014) (http://peppergenome.snu.ac.kr/). SNP calling and filtering were performed according to Reddy et al. (2025), retaining 45,987 high-quality SNPs with a minor allele frequency (MAF) ≥ 0.05 and call rate ≥ 90%. Population structure was assessed through principal component analysis (PCA) using GAPIT 3.0 (Wang and Zhang 2021), and admixture proportions were inferred with ADMIXTURE v1.3 (Alexander et al. 2015) for K = 1–10. Genome-wide association studies (GWAS) for fruit weight were conducted in GAPIT 3.0 (Wang and Zhang 2021) using both the mixed linear model (MLM) and the Bayesian-information and linkage-disequilibrium iteratively nested keyway (BLINK) model to ensure robustness of detection.
Biparental mapping population development, finemapping and progeny testing
Five F_2_ biparental populations were generated to validate fw6.1 and fw6.2 (Supplementary Table S1). The populations were fixed for the other known fruit weight and shape genes. To fine map the fw6.1 locus, recombinant plants with one recombination event between marker 18EP962 and 19EP629 (Supplementary Table S2) were evaluated for fruit weight in an F_4_ population 20S156 (N = 202). Interval mapping results was plotted using R/qtl package (map.funciton = kosambi, model = normal) (Broman et al. 2003). Quantile normalization was performed for fruit weight data using qqnorm function in R. The logarithm of odds (LOD) threshold of α = 0.01 was calculated using permutation tests with 1000 permutations. To fine map fw6.1 and fw6.2 separately, we generated F_6_ and F_7_ recombinant population from a cross between large-fruited BGV06232 and small-fruited BGV008225. Population 22S33 (N = 137) segregated for fw6.1 and population 22S30 (N = 88) segregated for fw6.2. Family 22S33 was fixed at BGV006232 allele between marker 21EP44 and 18EP327 for fw6.2. Family 22S30 was fixed at BGV008225 allele between marker 18EP962 and 21EP41 for fw6.1. Nine recombinant plants from different generations were progeny tested. Pedigree information for fine mapping fw6.1 and fw6.2 is shown in Supplementary Fig. S1.
To genotype the plants, Kompetitive Allele Specific PCR (KASP) markers were designed using the Primer Express® Software version 3.0.1 (Applied Biosystems, Carlsbad, CA) as described in Topcu et al. (2021) and derived cleaved amplified polymorphic sequences (dCAPS) markers were designed using the tool indCAPS (Hodgens et al. 2017). Primer information is shown in Supplementary Table S2.
Knockout of CSR-like1
To utilize CRISPR-Cas9 editing to create knockout mutations of CSR-like1, a single guide RNA was designed using CRISPR-P (http://crispr.hzau.edu.cn/cgi-bin/CRISPR/CRISPR) (Supplementary Table S2) and cloned into p201N vector (Addgene plasmid #59,175). The construct was transformed into BGV007931 (SLC) from the Varitome collection and a breeding line Fla.8059 (SLL) (Scott et al. 2008) using standard transformation procedures (Van Eck et al. 2019). The transformation yielded two independent T_0_s in BGV007931, and three independent T_0_s in Fla. 8059 background (Supplementary Table S3). Primers for genotyping the presence of transgene and the mutation are listed in Supplementary Table S2.
Downregulation of CSR-like1
One artificial microRNA (amiRNA) targeting the coding region of CSR-like1 was designed using WMD3 Web MicroRNA Designer (http://wmd3.weigelworld.org/). Target search with maximum 5 mismatches of the selected amiRNA sequence (TCTTGAGTCGAGTTGCGTCAT) showed CSR-like1 as the only target. The amiRNA sequence and its pair amiRNA* (ATAACGCAACTCGTCTCAAGT) were cloned into the Arabidopsis miR319a precursor backbone. The amiRNA precursor was synthesized by Azenta Life Sciences (South Plainfield, NJ, USA) and cloned into pKYLX71 vector at the XhoI and SacI restriction enzyme sites. The construct was transformed into the large-fruited mapping parent BGV006232. Four T_0_s were obtained and verified for the presence of transgene using primers 20EP672/20EP673 (Supplementary Table S2). The BC_1_F_1_ families 24S156, 25S158, 25S159 and 25S160 were used for phenotypic evaluations. Pedigree information is shown in Supplementary Fig. S2.
Phenotypic evaluations
The mapped interval of fw6.2 spanned 374 kb between marker 21EP56 and 24EP416. We generated a fw6.2 NIL family 24S155 in which fw6.2-D plants carry the large-fruited derived allele from BGV006232 and fw6.2-WT plants carry the wildtype allele from BGV008225. For fw6.2 NIL family 24S155 and the amiRNA transgenic family 24S156, 6 to 10 ovaries per plant were collected and scanned using a flatbed scanner. Ovary size was measured using Fiji software (Schindelin et al. 2012). Flowers were hand-pollinated and tagged for additional fruit phenotyping. Per plant, five to 10 of the largest fruits at turning or red ripe stage were cut at the equatorial plane along the medio-lateral axis and scanned. The images were analyzed using Tomato Analyzer 4.0 (Rodríguez et al. 2010) to obtain the total cross-section area, pericarp area, pericarp area ratio, columella area and columella area ratio. Fruit weight at various developmental stages was also measured in grams: 3, 5, 10, 15, 20, 25 DPA with three to four replicates per plant; mature green and red ripe stage consisted of five to 10 replicates per plant. Pericarp thickness, cell layer and mesocarp cell size were measured following the previously established protocols (Mu et al. 2017). Mesocarp cell size was calculated as the mean of the five largest cells in a pericarp slice. The number of days between anthesis to red ripe was recorded for the amiRNA transgenic families 24S156, 25S158, 25S159 and 25S160. Plant height at 55 days after sowing was measured for transgenic families 25S158, 25S159 and 25S160.
Haplotype analysis of CSR-like1
The genetic diversity of CSR-like1 in the Varitome collection was analyzed following the previously described protocol (Pereira et al. 2021a). Briefly, SNPs and INDELs were extracted from within CSR-like1 as well as 3 kb upstream of the start site and 1 kb downstream of the termination site with VCFTools (Danecek et al. 2011). The locations and functions of the variants were annotated using SnpEff (Cingolani et al. 2012) with a reference SL4.0 tomato reference genome (https://solgenomics.net/). The haplotype heatmap was generated using the R package “pheatmap” (Kolde 2019), accompanied with the phylogeny of the accessions (Razifard et al. 2020). The fruit weight of each accession was classified into three groups: small (below 3 g), medium (3–20 g) and big (above 20 g). The p-values of pairwise comparison among haplotype clusters were conducted using the R package “emmeans” (Lenth 2022) and the Tukey test with a significance threshold at 0.05 was used. Single variant association with fruit weight and pericarp cell size was calculated by Kruskal–Wallis test and the p-values were adjusted by the Benjamini–Hochberg method to control the False Discovery Rate (FDR). The haplotype distribution of CSR-like1 was visualized in a geographic map as described in Sapkota et al. (2023) using R packages “rnaturalearth” (https://CRAN.R-project.org/package=rnaturalearth) and “ggspatial” (https://CRAN.R-project.org/package=ggspatial). The latitude and longitude of collection sites were retrieved from Razifard et al. (2020).
The haplotypes of CSR-like1 in pepper were identified using five GBS SNPs (S06_227195491, S06_227195529, S06_227195579, S06_227195589, and S06_227195619) and the pairwise comparison of fruit weight among haplotypes were performed using the same method as for tomato haplotype analysis.
Results
Identification and fine mapping of fw6.1 and fw6.2 in tomato
To identify novel genes regulating fruit weight in tomato, a GWAS analysis was performed using the Varitome collection which included 27 wild SP, 121 semi domesticated SLC and 18 ancestral landrace SLL accessions. A total of 15 fruit weight GWAS loci using SNPs and INDELs were identified in this study, which led to six unique fruit weight QTLs on chromosomes 2, 3, 6, 11, and 12, and another six QTLs that were scattered on chromosome 1 (Fig. 1a,b; Supplementary Table S4). The GWAS_fw2.2 locus is likely associated with the known fw2.2, whereas GWAS_fw11.1 is likely associated with the linked loci fw11.3 and fas. GWAS_fw3.2 is likely associated with fw3.2, however, the variant was 5 Mb from the known fruit weight gene. Among the novel GWAS loci, GWAS_fw6.1 has two linked INDEL markers associated with fruit weight (ISL4.0ch06_44805731 and ISL4.0ch06_45074906) and was further characterized in this study.Fig. 1. Identification of fw6.1 and fw6.2. a GWAS for fruit weight in the Varitome collection based on SNPs. b GWAS for fruit weight based on INDELs. The x-axis shows the chromosome number and y-axis shows the -log10(p-value) of the variants. The Bonferroni-corrected Genome-wide significance thresholds at α = 0.05 are indicated by the dotted lines. c Composite interval mapping of fw6.1 and fw6.2 in an F_4_ population 20S156. The markers used to genotype recombinant plants in 20S156 are shown below the QTL map. The markers flanking the two intervals are highlighted in red. The LOD threshold of α = 0.01 is indicated in a gray line
To validate the newly identified fw6.1 locus, we evaluated a total of five F_2_ biparental populations using flanking markers approximately 1–2 Mb upstream and downstream of the significant GWAS loci. The parental selection of the F_2_ populations was based on the predicted variants from the GWAS and the estimated effect of the QTL on fruit weight. For all F_2_ populations, the two parents of each population were fixed at the known tomato fruit weight and fruit shape genes (Supplementary Table S1). GWAS_fw6.1 was confirmed in the four F_2_ populations that segregated at the GWAS_fw6.1 locus, namely 18S33, 18S39,18S40 and 18S133 (Supplementary Table S5). The QTL showed a significant additive effect with a small, nonsignificant dominance deviation, indicating primarily additive gene action with partial dominance (Supplementary Table S5). The marker association was most significant in 18S39, and therefore this population was used to fine map the locus. Family 18S39 was derived from a cross between an Ecuadorian accession (BGV006232) and a Peruvian SLC accession (BGV008225), and the two parental accessions exhibited a fourfold difference in fruit weight (Supplementary Table S1).
To narrow the interval of fw6.1, we screened for recombinant plants using self-pollinated progeny from the 18S39 F_2_ population. Linkage mapping indicated the presence of two linked QTLs that were significantly associated with fruit weight on chromosome 6 (Fig. 1c). fw6.1 was flanked by marker 18EP962 (SL4.0ch06:39,930,114) and 20EP567 (SL4.0ch06:42,662,777*),* spanning a 2.73 Mb interval, whereas fw6.2 was flanked by marker 20EP1256 (SL4.0ch06:43,033,158) and 19EP629 (SL4.0ch06:44,319,744), spanning a 1.29 Mb interval*.* BGV006232 carried the large-fruited derived allele at both fw6.1 and fw6.2, as shown by the allelic effects of the highest associated markers at each QTL (Supplementary Fig. S3).
To map each locus further, we selected recombinant plants that were segregating at one locus while fixed at the other. By evaluating two recombinant populations (Fig. 2a,b) and nine progeny testing families (Fig. 2c; Supplementary Table S6), we narrowed fw6.1 to a 1.03 Mb region flanked by marker 22EP373 (SL4.0ch06:40,416,852) and 22EP37 (SL4.0ch06:41,451,757), and fw6.2 to a 374 kb region between marker 21EP56 (SL4.0ch06:43,262,334) and 24EP416 (SL4.0ch06:43,636,481).Fig. 2. Fine mapping of fw6.1 and fw6.2. a Interval mapping of fw6.1 in an F_6_ population 22S33. b Interval mapping of fw6.2 in an F_7_ population 22S30. The markers used to genotype recombinant plants in 22S30 and 22S33 are shown with grey bars and additional markers used to genotype the progeny testing families are shown with blue bars. For a and b, the LOD threshold of α = 0.01 is indicated in a gray line. c Progeny testing of fw6.1 and fw6.2. The colored boxes indicate the parental genotype of each progeny testing family. All plants were grown under field condition except greenhouse-grown 24S155. The p-values were calculated using Student's t-test, and p-values < 0.05 are highlighted in red. The markers flanking fw6.1 and fw6.2 are highlighted with red dotted lines
To carefully analyze fruit phenotypes that are controlled by fw6.2 and gain insights into its mechanism to regulate fruit weight, we used NILs from an F_7:8_ plant 22S310-30 in which fw6.1 was fixed for the derived BGV006232 allele. At fw6.2, fw.6.2-D NILs carried the derived BGV006232 allele and fw6.2-WT NILs carried the wildtype BGV008225 allele. Fruit weight was significantly different between the two genotypes (Fig. 3a; Table 1), whereas ovary size was not affected (Table 1). The difference in fruit weight started to manifest itself during the development of the fruit 25 days after pollination (Fig. 3a). We conducted additional morphological and cytological measurements (Table 2). Fw6.2-D NILs showed a larger total fruit and pericarp area than fw6.2-WT NILs. On the other hand, there was no significant segregation for pericarp area ratio, columella area, and columella area ratio. This suggests that the fruit weight increase in the fw6.2-D NILs was primarily driven by the enlargement of the pericarp. Moreover, the pericarp cell size was significantly larger in the fw6.2-D NILs, while the two genotypes showed a similar number of pericarp cell layers (Fig. 3b; Table 2). Therefore, fw6.2 is likely to affect fruit weight by regulating the cell size in the pericarp.Fig. 3. Fruit development in fw6.2 NILs and CSR-like1 transgenic family. a Fruit weight at various fruit development stages (3, 5 10, 15, 20 and 25 DPA, mature green (MG) and red ripe (RR)) for fw6.2 NILs. b Pericarp section of a representative mature green fruit of fw6.2-D and fw6.2-WT stained with toluidine blue. c Fruit weight at various fruit development stages for amiRNA transgenic family. d Pericarp section of a representative mature green fruit of CSR-like1-WT and CSR-like1-amiRNA. Student’s t-test is used to calculate the p-values in a and c. Asterisks denote levels of statistical significance: p < 0.05 (), p < 0.01 (), and p < 0.001 ()Table 1. Fruit weight and ovary size of fw6.2 NILs (24S155) and CSR-like1 amiRNA transgenic plants (24S156)Population24S155 (fw6.2 NILs)24S156 (CSR-like1 amiRNA BC_1_F_1_)Genotypefw6.2-D (n = 10)fw6.2-WT (n = 10)P-value^b^WT (n = 10)amiRNA (n = 10)P-valueAverage red ripe fruit weight (g)^a^7.59 ± 0.686.59 ± 0.500.0014511.62 ± 1.076.60 ± 0.571.13E-10Ovary size (mm^2^)1.26 ± 0.0961.23 ± 0.110.4981.50 ± 0.1041.35 ± 0.08262.74E-03^a^All measurements in the table are given as mean ± standard deviation (SD)^b^The p-values were calculated by Student’s t-testTable 2Weight attribute evaluations of fruits and cellular evaluations of fruit pericarp from the fw6.2 NILs (24S155) and CSR-like1 amiRNA transgenic plants (24S156)Population24S155 (fw6.2 NIL)24S156 (CSR-like1amiRNA BC_1_F_1_)Genotypefw6.2-D (n = 10)fw6.2-WT (n = 10)P-value^b^WT (n = 10)amiRNA (n = 10)P-valuePerimeter (cm)^a^8.25 ± 0.287.81 ± 0.184.29E-049.08 ± 0.218.41 ± 0.239.22E-06Total area (cm^2^)4.89 ± 0.334.37 ± 0.204.45E-045.85 ± 0.255.05 ± 0.278.88E-06Pericarp area (cm^2^)2.12 ± 0.151.90 ± 0.110.001092.35 ± 0.121.98 ± 0.129.31E-06Columella area (mm^2^)6.50 ± 0.375.88 ± 0.520.007136.24 ± 0.595.54 ± 0.310.00743Pericarp area ratio0.433 ± 0.0080.433 ± 0.0080.8920.401 ± 0.0120.393 ± 0.0110.121Columella area ratio0.013 ± 0.000700.013 ± 0.00110.7410.011 ± 0.000890.011 ± 0.000470.430Pericarp thickness (mm)2.11 ± 0.131.88 ± 0.130.001122.42 ± 0.162.00 ± 0.098.02E-06Pericarp cell size (mm^2^)0.0810 ± 0.00720.0671 ± 0.00777.74E-040.0974 ± 0.01170.0728 ± 0.00616.26E-05Pericarp cell layer12.54 ± 0.6512.38 ± 0.800.64211.53 ± 0.4911.20 ± 0.300.105^a^All measurements in the table are given as mean ± SD^b^The p-values were calculated by Student’s t-test
Candidate gene at fw6.2: CSR-like1
Since fw6.2 showed the highest PVE and smallest introgression, we evaluated the genes in the 374 kb interval in the tomato reference SL4.0 genome (Supplementary Table S7). Among the 51 genes, Solyc06g073940 stood out as the paralog of the known fruit weight gene CSR controlling pericarp cell size, CSR-like1 (Mu et al. 2017). The sequence analysis of CSR-like1 showed that the BGV006232 and BGV008225 alleles carried no nucleotide polymorphisms in the protein coding region, and instead three SNPs in the 5’ untranslated region (UTR), one SNP in the 3’UTR, and 16 SNPs in the putative 3 kb regulatory region upstream of CSR-like1.
To validate the function of CSR-like1 in fruit weight, we first sought to create knockout mutants using the CRISPR-Cas9 gene editing approach in two distinct accessions, a small-fruited SLC BGV007931 and a large-fruited SLL Fla. 8095 accession (Scott et al. 2008). We recovered two independent T_0_s in BGV007931 and three independent T_0_s in Fla. 8095 backgrounds, respectively, but all five T_0_s carried in frame deletions of 6 bp or 12 bp (Supplementary Table S3). Despite the CRISPR-edited alleles were in-frame and not nulls, we evaluated the fruit weight variation in these lines (Supplementary Fig. S4). As expected, the in-frame edited alleles did not show consistent fruit weight differences compared to wildtype. We concluded that CSR-like1 may play an essential role in plant development such that a complete knockout is likely to be lethal.
We next sought to downregulate the expression of CSR-like1 using the artificial microRNA (amiRNA) technology, and recovered four T_0_ in the large-fruited BGV006232 background. The fruit weight of CSR-like1-amiRNA plants was between 27 to 43% lower than controls under greenhouse conditions in the transformed lines, including a significant decrease in ovary size (Table 1; Supplementary Table S8). The pericarp area was significantly decreased in CSR-like1-amiRNA plants resulting from reduced cell size (Fig. 3d; Table 2). We also observed slower fruit development in CSR-like1-amiRNA plants, starting as early as 5DPA (Fig. 3c), and these plants required approximately 8–12 days more to reach the red ripe stage compared to CSR-like1-WT plants (Table 3). No vegetative effects were observed as the plants looked normal (Supplementary Table S8; Supplementary Fig. S5). Table 3. The number of days from anthesis to red ripe for CSR-like1-WT and CSR-like1-amiRNA fruits in four amiRNA transgenic familiesPopulationGenotypeAnthesis to red ripe (days)^a^24S156 (BC_1_F_1_)WT (n = 10)40.56 ± 1.76amiRNA (n = 10)52.28 ± 2.31P-value^b^1.88E-1025S158 (BC_1_F_1_)WT (n = 11)34.97 ± 1.40amiRNA (n = 9)44.82 ± 4.35P-value1.86E-0625S159 (BC_1_F_1_)WT (n = 10)36.19 ± 1.03amiRNA (n = 10)44.46 ± 2.60P-value2.46E-0825S160 (BC_1_F_1_)WT (n = 10)34.95 ± 1.40amiRNA (n = 10)42.92 ± 2.46P-value5.23E-08^a^All measurements in the table are given as mean ± SD^b^The p-values were calculated by Student’s t-test
Evolution of CSR-like1 in tomato domestication
Fruit weight was an important trait for selection during the domestication of most vegetable crops. To investigate the evolution of CSR-like1 during tomato domestication, we explored the genetic diversity at the locus by generating a heatmap containing six haplotype clusters in the Varitome collection (Fig. 4a). A total of 88 variants were identified at the locus, of which 12 were INDELs ranging in size from 1 to 9 bp in addition to 76 SNPs (Supplementary Table S9). Most of the variants were located in the regulatory region and the UTRs. Similar to the parents in the mapping population, the coding region of CSR-like1 was highly conserved in the Varitome collection with only 2 SNPs in distantly related accessions. SNP_ SL4.0ch06:43,344,537 resulted in a missense mutation from proline to leucine at amino acid position 355, while SNP_ SL4.0ch06:43,344,932 led to a synonymous mutation of glutamine at amino acid position 223. The clustering of haplotype groups was also associated with the phylogenetic groups of the accessions. Most of the SLC and all SLL accessions were found in Cluster I, II, and III. All of the wild tomato SP accessions were in Cluster IV, V and VI, carrying more mutations in the regulatory region and the UTRs. The missense mutation was only found in four accessions (1 SP and 3 SLC) in Cluster VI. We also surveyed an additional accession panel (SLL_CUL) including 73 modern cultivars, landraces and heirlooms (Tieman et al. 2017), and assigned their haplotypes using conserved SNPs across Cluster III to VI (Supplementary Table S10). These cultivated accessions also have either haplotype I or II, similar to the ancestral SLL in the Varitome collection (Supplementary Table S10). We will now refer haplotype I and II allele together as CSR-like1-D for derived which was most similar to the Heinz 1706 reference genotype, and haplotype III, IV, V, and VI together as CSR-like1-WT for wildtype.Fig. 4. Haplotype analysis of CSR-like1 in the Varitome collection. a Heatmap representing the genotypes of accessions (rows) at the CSR-like1 locus. Reference genotypes are represented in blue, alternate in red, heterozygous in yellow and missing data in white. The first left column indicates the fruit weight category (big, medium, or small) of each accession. The second left column shows the phylogenic group of each accession (SP, SLC, or SLL). The first top bar shows the location of the variant: Upstream (Upstr), 5’ or 3’ untranslated region (UTR), protein coding region (missense or synonymous (Synon) mutation), and downstream (Downstr). The second top bar indicates the type of variant (SNP or INDEL). The only two SNPs in the coding region are highlighted in the black box. Red asterisks indicate the four variants that are conserved in most of the accessions carrying CSR-like1-WT allele (Supplementary Table S11). b,c Violin plots of fruit weight and pericarp cell size of the Varitome accessions fixed with CSR-WT allele, grouped by haplotype clusters
For the two mapping parents of fw6.2, the large-fruited parent BGV006232 carries the CSR-like1-D allele (haplotype II) while the small-fruited parent BGV0008225 carries the CSR-like1-WT allele (haplotype IV). To further investigate the association between CSR-like1 haplotypes in the entire Varitome collection and fruit traits, we selected only accessions carrying the wildtype fw11.3/CSR allele since that gene has a large effect on fruit weight and pericarp cell size. The resulting 108 SLC and 27 SP accessions showed that fruit weight and pericarp cell size were larger in the accessions carrying CSR-like1-D alleles (Fig. 4b,c). Similarly, when fixing for CNR (fw2.2), KLUH (fw3.2) and all three known fruit weight genes, the CSR-like1 haplotype I and II were always associated with larger fruits and pericarp cells (Supplemental Figure S6), further supporting the notion that this gene is critical in the regulation of fruit weight via increased cell sizes. We also identified four SNPs in the upstream and downstream of CSR-like1 that are exclusively found in CSR-like1-WT alleles (red asterisks in Fig. 4a). These SNPs are the four most significant variants associated with fruit weight and pericarp cell size (Supplementary Table S11).
CSR-like1 associated with fruit weight in pepper
Orthologs of tomato fruit weight genes have been reported as likely candidate genes underlying fruit weight QTLs in other Solanaceae crops including pepper and eggplant (Rinaldi et al. 2016; Toppino et al. 2016; Zygier et al. 2005) as well as some Cucurbitaceae crops including melon, cucumber and watermelon (Monforte et al. 2014; Pan et al. 2020). To explore the potential conserved function of CSR-like1 in regulating fruit weight in pepper, we conducted GWAS using multi-year phenotypic data collected from four Capsicum species (C. annuum, C. annuum var*. glabriusculum*, C. baccatum, and C. chinense) (Supplementary Table S12). We identified three significant SNPs (S06_227195491, S06_227195529, and S06_227195619) in the mixed linear model (MLM), and one SNP (S06_227195491) in the BLINK model, all located within the coding region of CaCSR-like1 (CA06g22610) (Fig. 5a,b). The Capsicum ortholog CaCSR-like1 exhibited conserved motifs and domain structures characteristic of the CSR/FAF-like proteins (Mu et al. 2017), suggesting a conserved molecular mechanism in regulating cell size and fruit mass.Fig. 5GWAS for the fruit weight in pepper using BLINK (a) and MLM (b) models. The horizontal dotted line indicates the Bonferroni-corrected threshold (p < 0.05), and the dotted circle highlights significant SNPs linked to the CaCSR-like1 gene
To further characterize allelic diversity of CaCSR-like1, we identified a total of five SNPs (S06_227195491, S06_227195529, S06_227195579, S06_227195589, and S06_227195619) in the coding region of CaCSR-like1 among the Capsicum GWAS panel using the GBS markers (Fig. 6a). Six haplotypes (haplotype I to VI) were assigned using the five segregating sites, and only SNP S06_227195619, S06_227195579 and S06_22719549 led to non-synonymous changes at amino acid position 10, 24 and 53 (Fig. 6a). These non-synonymous mutations were outside of the FAF domain (Pfam accession: PF11250) which is located between amino acid position 230 and 278. The wild pepper C. annuum var*. glabriusculum* in this study only carried haplotype II or III, while the cultivated pepper C. annuum, C. baccatum, and C. chinense exhibited all six haplotypes across different accessions (Fig. 6b; Supplementary Table S13). There is also clear association of haplotype VI with higher fruit weight within each cultivated pepper species (Fig. 6b).Fig. 6. Haplotype analysis of CaCSR-like1. a Six haplotypes identified in the pepper GWAS panel. The gene structure of CaCSR-like1 is shown on the top in which the grey boxes represent the 5′ and 3′ UTR, the white box represents the coding sequence, and the straight line represents the intron. Colored triangles represent the five segregating SNPs. Part of the amino acid sequence (position 1 to 60) containing the SNPs is shown for each haplotype. b Haplotype association with fruit weight in cultivated species C. annuum, C. baccatum, and* C. chinense*
Discussion
Fruit weight is a critical component in modern agriculture as it is linked to yield and the price of produce. Hence a key focus in many vegetable breeding programs is to maintain or increase fruit weight. Understanding the genetic basis underlying this highly quantitative trait can provide us valuable insights into fruit development and help accelerate breeding programs. In this study, we successfully identified five unique fruit weight QTL from a GWAS with diverse SP, SLC and SLL accessions. Of these loci, three were likely to correspond to known fruit weight genes while two were novel. We subsequently fine mapped two QTLs on chromosome 6 using association mapping and progeny testing. Of these*, fw6.2* harbors the smaller interval of 374 kb. Based on changes in pericarp cell size in fw6.2 NILs, CSR-like1 was identified as the most likely candidate gene at the locus. CSR-like1 is a paralog of the only known fruit weight gene CSR that regulates cell enlargement.
An additional 50 genes are found at the fw6.2 locus. Therefore, we can’t exclude the possibility of other genes contributing to fruit weight at this locus. Even so, we successfully excluded SP (SELF-PRUNING; Solyc06g074350) that had recently been shown to control fruit weight (Ning et al. 2023). SP is located ~ 291 kb downstream of CSR-like1, and is primarily known to modulate plant growth habit by regulating the vegetative-reproductive switch in shoot growth (Pnueli et al. 1998). fw6.3 was mapped to a 62.6 kb interval that concluded SELF-PRUNING (SP) as the most likely candidate gene (Ning et al. 2023). We designed a marker 24EP416, locating in the third intron of SP, and were able to exclude SP from fw6.2 using progeny testing of families 23S94 and 24S155 (Fig. 2).
The haplotype analyses of CSR-like1 showed that smaller pericarp cell size was associated with accessions carrying CSR-like1-WT alleles in backgrounds where the other cloned fruit weight genes were fixed (Fig. 4c and Supplementary Figure S6). Moreover, among the F_2_ populations that were used to validate GWAS_fw6.1, four showed significant association with fruit weight at the locus (Supplementary Table S5). Of those that segregated at the GWAS locus, the large-fruited parent of each of the four populations carries CSR-like1-D allele while the other parent carries CSR-like1-WT allele except for 18S40 where both parents carry the same CSR-like1-D allele. The small effect on fruit weight at GWAS_fw6.1 may be the result of the other segregating fruit weight QTLs in the population and imply that other genes at the GWAS_fw6.1 locus affect the trait as well (Supplementary Tables S1 and S5). The fifth F_2_ population, 18S32, was neither segregating for polymorphisms at GWAS_fw6.1, nor for the CSR-like1 allele. Therefore, as expected, the locus was not associated with fruit weight in the 18S32 population. Together, genetic mapping and haplotype association analysis demonstrate that CSR-like1 is a likely candidate for fruit weight at the fw6.2 locus. In addition, the haplotype analyses revealed four most significant SNPs associated with fruit weight that could be utilized in future breeding programs for selecting desired fruit weight in tomato (Supplementary Table S11).
CSR-like1 was located 19 kb away from one of the selective sweep windows in the transition from SP to SLC (Razifard et al. 2020), suggesting that the gene may have been selected in the early stage of tomato domestication. This agrees with the distribution of CSR-like1 haplotypes in the genetically distinct groups classified in Razifard et al. (2020) (Supplementary Fig. S7). All SP accessions in the Varitome collection carried haplotype IV, V or VI, which were associated with smaller fruit weight. Based on the tomato domestication hypothesis proposed by Razifard et al. (2020), Ecuadorian SLC (SLC_ECU) accessions would be the most ancestral SLC group, and thus would be the first group where larger fruit-associated haplotype I and II allele of CSR-like1 appeared. The frequency of haplotype I and II allele would increase with further domestication. Haplotype III allele is associated with smaller fruit weight compared to haplotype I and II, and is mostly found in a group of SLC accessions from Mexico, Central America, and northern South America (together annotated as SLC_CA). The rise of haplotype III allele would coincide with the transition of fruit weight back to more wild-like tomatoes in SLC_CA, and we would expect to see a reselection of Haplotype I and II of CSR-like1 in the fully cultivated SLL from SLC_Mexico. However, such re-selection was not detected in the study of Razifard et al. (2020) at the CSR-like1 locus. In fact, the haplotype distribution of CSR-like1 could demonstrate another scenario in the evolution of tomato such as described in Blanca et al. (2022) (Supplementary Fig S8). SLC_CA was likely to be the intermediate group after the northward migration of SP towards Mesoamerica. Later SLC_CA migrated back to South America and admixed with Ecuadorian and Peruvian SP to give rise to Ecuadorian and Peruvian SLC in which haplotype I and II allele of CSR-like1 were further distributed. Some Peruvian SLC then migrated back to Mexico where it evolved fully to SLL.
We functionally validated the role of CSR-like1 in controlling fruit weight and pericarp cell size by downregulating its expression in transgenic plants. In the amiRNA lines, the effect of reduced expression of CSR-like1 on fruit development started to manifest itself in ovaries one day prior to anthesis. CSR-like1 is the only member of the CSR/FAF-like family in tomato that has relatively high expression in young flower buds and flowers at anthesis (Mu et al. 2017). This suggests that CSR-like1 might regulate cell differentiation and enlargement in floral development as well even though the NILs did not show a difference at that time. Similar differences between lines carrying natural alleles and lines carrying transgenes has been observed in fw3.2/KLUH (Chakrabarti et al. 2013). Compared to fw3.2 NILs, the downregulated transgenic lines showed additional phenotypic defects across the entire plant, including plant height, leaves and leaflets, seed number and side shoot number (Chakrabarti et al. 2013). Future experiments are needed to determine the developmental time frame for when and how CSR-like1 acts in floral development.
Tomato serves as an important model crop for studying fruit development, with many genes displaying conserved functions across plant species. Recently, two association mapping studies in C. annuum and C. chinense, respectively, reported SNPs in the orthologs of CSR-like1 to be significantly associated with fruit weight (Nimmakayala et al. 2016, 2021). Here we are adding another piece of evidence of a pepper CSR-like1 regulating fruit weight using a more expanded Capsicum collection with multi-model GWAS identifying significant SNPs within the coding region of CaCSR-like1. The presence of the distinct haplotypes associated with fruit weight further indicates that allelic diversification of CaCSR-like1 contributed to the phenotypic spectrum observed in cultivated Capsicum species. Haplotype VI of CaCSR-like1 was enriched among large-fruited accessions, likely represents a derived allele favored during domestication or selection for increased fruit mass.
There is evidence of CSR-like1 being relevant in other crops. In watermelon and cucumber, QTL mapping using recombinant inbred lines (RILs) has identified fruit weight/size QTLs that harbor the respective orthologs of CSR-like1 (ClCG02G022450 in watermelon; CsGy6G022740 in cucumber) (Guo et al. 2024; Weng et al. 2015). A recent GWAS utilizing the CucCAP cucumber core collection also identified a SNP which is ~ 250 kb upstream of CsGy6G022740 to be significantly associated with fruit size (Lin et al. 2024). In eggplant, though there have not been any reported fruit weight QTLs mapped close to CSR-like1, a fruit weight QTL of around 10 cM on chromosome 12 was mapped in an F_2_ population (S. melongena 305E40 x S. melongena 67/3), with SMEL4.1_12g014140.1, the ortholog of CSR, residing at the locus (Gaccione et al. 2023; Portis et al. 2014). Direct genetic modification in other crops could be applied to broaden our understanding on the potential conserved function of CSR-like1 and its related genes in regulating fruit size and other species-specific functions.
While the function on fruit size and cell size by CSR-like1 and CSR in tomato is apparent, its molecular function is less clear. CSR-like1 is predicted to encode a protein that contains a FANTASTIC FOUR (FAF) domain, and is an ortholog of FAF-like (AT5G22090) in Arabidopsis thaliana (Mu et al. 2017). FAF-like and its related family FAF were first reported in Arabidopsis, which carries four FAF members and one FAF-like (Wahl et al. 2010). Phylogenetic analysis in monocotyledonous and dicotyledonous plants suggest that the FAF genes have evolved from a FAF-like gene to become a dicotyledonous-specific gene family after gene duplication (Wahl et al. 2010). Many of the plant species in the Rosids and Asterids clades carry only one copy of FAF-like (Mu et al. 2017; Wahl et al. 2010). Interestingly, an expansion of FAF-like is observed in the Solanaceae family. Tomato, potato (S. tuberosum) and chili pepper (Capsicum annuum) carry four FAF-like genes whereas eggplant (S. melongena) carries three members (Mu et al. 2017). In Arabidopsis, FAF-like (AT5G22090) encodes a protein described as EAR1 (ENHANCER of ABA CO-RECEPTOR1), which can enhance the activity of clade A type 2C protein phosphatases (PP2Cs) by binding to their N termini, causing the inhibition of Snf1-related kinases2 (SnRK2s) (Wang et al. 2018). Meanwhile, the expression of many downstream targets of ABA signaling rely on phosphorylation by SnRK2s (Hasan et al. 2022). Therefore, EAR1 is a negative regulator of ABA signaling and shown to affect seed germination, primary root growth and drought tolerance (Wang et al. 2018). Interestingly, in pepper, CaCSR-like1 (CA06g22610; or CaFAF1) regulates the ABA signaling pathway but in a different manner than EAR1 (Lim et al. 2022; Wang et al. 2018). CaCSR-like1 does not interact with the known pepper PP2Ps in yeast two-hybrid assays, and unlike EAR1, CaCSR-like1 does not affect seed germination and primary root growth when overexpressed in Arabidopsis. Instead, CaCSR-like1 plays a positive role in drought stress and a negative role in salt stress. Therefore, CaCSR-like1’s function differs under certain abiotic stresses (Lim et al. 2022).
In this study, fruit ripening is delayed in the transgenically downregulated CSR-like1 lines. It is a complex process controlled by many regulators through phytohormone and environmental signals (Kou et al. 2021). ABA is known to be a key signaling hormone to regulate the ripening process in tomato through crosstalk with ethylene biosynthesis and metabolism (Kou et al. 2021; Mou et al. 2016; Zhang et al. 2009). As discussed earlier, EAR1 in Arabidopsis is a negative regulator of ABA signaling pathway (Wang et al. 2018). Silencing one of the tomato PP2Cs, SlPP2C3 (Solyc06g076400), was shown to accelerate the ripening process (Liang et al. 2021). Therefore, if CSR-like1 were to have a similar function as EAR1 and interact with PP2Cs like PP2C3, we would expect to see a shortened ripening process in our amiRNA transgenic plants; instead, we observed the opposite result. This suggests that CSR-like1 might have a divergent function from EAR1. As mentioned earlier, a similar case has been made in pepper where CaCSR-like1 was demonstrated to affect ABA signaling through a different mechanism under abiotic stresses (Lim et al. 2022). The expansion of CSR/FAF-like family especially in the Solanaceae crops also suggests CSR/FAF-like proteins might have evolved to have novel functions evolution.
In summary, our data support a critical role for the tomato FAF-like genes, CSR and CSR-like 1 on fruit weight in this species and potentially other crops, especially in the Solanaceae family. Further understanding its molecular function should lead to insights into fundamental plant processes while at the same time enabling breeders to implement the knowledge presented in their breeding programs.
Supplementary Information
Below is the link to the electronic supplementary material.Supplementary file1 (XLSX 22 KB)Supplementary file2 (XLSX 165 KB)Supplementary file3 (XLSX 108 KB)Supplementary file4 (PDF 4758 KB)