A major driver of crop domestication was the modification of wild plant architectures into new forms that improved flower, fruit and seed production¹². In maize, rice and soybean, changes in branching¹³, uprightness^{14, 15} and compactness^{6, 7} enabled plants to be grown at higher density, simultaneously increasing yield and facilitating large-scale harvesting. In tomato, shoot architecture changed radically 85 years ago with the discovery of the self pruning (sp) mutant. This discovery transformed 'indeterminate' plants into a new 'determinate' form that spawned the processing tomato industry^{16, 17}. SP is the homolog of the Antirrhinum majus CENTRORADIALIS (CEN) and Arabidopsis thaliana TERMINAL FLOWER1 (TFL1) genes. SP encodes a repressor of flowering in the CETS protein family, which includes the flowering hormone florigen^{10, 18}. Wild tomatoes and classical cultivated varieties have an indeterminate vine architecture regulated by the opposing activities of SP and florigen^{10, 18}. In sympodial plants, such as tomato, pluripotent cells at shoot apical meristems terminate in flowers. Specialized lateral (sympodial) meristems continuously give rise to new sympodial shoots and inflorescences. In indeterminate tomato plants, each sympodial shoot develops three leaves before terminating in a multi-flowered inflorescence, resulting in endlessly growing vines (Fig. 1a)^{10, 18, 19}. Regularity and perpetuation of cycling are made possible by a balance between flower-promoting (florigen, encoded by SFT) and flower-repressing (SP) signals²⁰. In sp mutants, florigen is no longer inhibited, causing shoots to terminate progressively faster until cycling stops^{10, 20}. The result is a burst of inflorescence production and nearly synchronous fruit ripening on compact, bushy plants that can be grown at high density and harvested mechanically¹⁷.

Homologs of SP and SFT have been targeted for agricultural adaptations in many crops, including barley¹, beans³, beets⁸, grapes², rice⁴, roses⁵, soybean^{6, 7}, strawberries⁵ and sunflower⁹. Yet, despite the agricultural importance of these genes, the limited number of alleles available may not provide an optimal balance of flowering signals for maximum productivity. In our previous work, we found that mutations in SFT can increase yield in determinate plants²¹. In sft mutants, loss of florigen results in highly vegetative plants with few flowers and fruits. However, when determinate plants are heterozygous for sft mutations, a partial, dose-dependent reduction in florigen activity weakly suppresses sp, resulting in more sympodial shoots and inflorescences (Fig. 1a)²². These results suggested to us that available shoot architectures might be limiting tomato yields. We reasoned that new mutants that suppress the sp mutant phenotype could be exploited to modulate flowering, shoot architecture and yield in a quantitative manner.

We screened a mutagenesis population in the processing cultivar M82 and found three ethyl methanesulfonate (EMS)-derived suppressor of sp (ssp) mutants that restored sp mutants to indeterminate growth (Fig. 1b) (the sp designation in genotypes is hereafter omitted unless otherwise noted). The ssp-1906 mutant developed three-leaf sympodial shoots, as in wild-type plants. In contrast, the allelic ssp-2129 and ssp-610 mutants averaged two leaves per sympodial shoot, resulting in a higher density of inflorescences in comparison to wild-type plants (Fig. 1b–d and Supplementary Table 1). All three mutants flowered late, with the ssp-1906 mutant showing the strongest effect (Fig. 1e). In a functional SP background, each mutant flowered similarly late; however, sympodial shoot flowering was unaffected (ssp-2129) or only slightly delayed (ssp-1906) (Supplementary Fig. 1).

To identify the genes responsible, we mapped the locus for the ssp-1906 mutant to a 29-Mb interval on chromosome 3, which included SFT (Supplementary Fig. 2a). Sequencing the SFT coding region from the ssp-1906 mutant, we identified a missense mutation mapping to a conserved 14-residue external loop domain, which confers florigen activity in the Arabidopsis SFT ortholog, FLOWERING LOCUS T (FT) (Fig. 2a and Supplementary Fig. 2b)²³. The ssp-1906 mutant failed to complement the sft-4537 and sft-7187 mutants with stronger phenotypes (Supplementary Fig. 2c)¹⁸. The same amino acid change in FT causes weak late flowering in Arabidopsis²⁴. Thus, the ssp-1906 allele (renamed sft-1906) is a weak allele of sft.

We mapped the locus for the ssp-2129 mutant to a 2-Mb region on chromosome 2 and used RNA sequencing to identify the responsible mutation among 267 genes (Online Methods and Supplementary Fig. 3). Aligning Illumina-sequenced reads from an ssp-2129 cDNA library to the tomato reference genome, we identified four genes with missense mutations in the mapping interval (Online Methods and Supplementary Table 2). Among these was Solyc02g083520, encoding a basic region/leucine zipper (bZIP) transcription factor. The protein encoded by Solyc02g083520 was identified previously as an SP-interacting G-BOX (SPGB)²⁵ protein and a homolog of Arabidopsis FLOWERING LOCUS D (FD), variants in which delay flowering (Supplementary Fig. 4a)^{26, 27}. In rice, a homolog of FD (OsFD1) interacts with florigen via 14-3-3 scaffolding proteins to form a florigen activation complex (FAC), which activates the expression of flowering transition genes¹¹. OsFD1–14-3-3 interactions depend on a C-terminal SAP (S/TAP) motif found in other FD homologs, including Solyc02g083520 (Fig. 2b and Supplementary Fig. 4b). We found that Solyc02g083520 mutations in the ssp-2129 and ssp-610 mutants disrupted the SAP motif, converting the proline to a leucine and the threonine to an isoleucine, respectively (Fig. 2a,b and Supplementary Fig. 3). Yeast two-hybrid assays showed that both mutations abolished SSP interactions with 14-3-3 proteins, which also bind SP²⁵ and SFT (Fig. 2c and Supplementary Fig. 5). We validated these results in planta using bimolecular fluorescence complementation (BiFC) assays in tobacco leaf cells (Fig. 2d,e and Supplementary Fig. 5). Moreover, BiFC assays showed that predominantly cytoplasmic SFT–14-3-3 BiFC signal concentrated in the nucleus when SSP was coexpressed; however, nuclear localization was significantly (P < 0.01) reduced when ssp-2129 or ssp-610 was coexpressed (Fig. 2f,g and Supplementary Fig. 6). These results suggest that the suppression of determinacy by the ssp mutants is due to a disruption of FAC assembly.

The ssp phenotypes were restricted to flowering and meristem determinacy, and our tomato meristem maturation atlas¹⁹ showed that SSP is expressed in all meristems, with the highest expression in vegetative stages (Fig. 3a). In situ hybridization further showed expression in all meristem layers (as for FD^{27, 28}), with some expression in vasculature cells (Fig. 3b–e and Supplementary Fig. 7). These data suggest a scenario similar to that in Arabidopsis and rice^{11, 26, 27} in which SSP is continuously present in meristems, and FAC formation, flowering transition and sympodial cycling only begin when florigen (SFT) is transported from leaves to meristems. SSP mutations that severely disrupt FAC assembly should therefore resemble strong sft alleles; yet, the ssp mutant phenotypes were phenotypically and molecularly weak (Figs. 1 and 3f,g). One likely explanation is redundancy with an SSP paralog²⁸, whose gene product also interacts with 14-3-3 in a SAP motif–dependent manner and shows similar meristem expression as SSP (Supplementary Fig. 4).

Given our previous finding that heterozygosity for an sft mutation (for example, sft-4537/+) improves inflorescence production and yield through a dose-dependent suppression of the sp mutant phenotype^{21, 22} and that protein complexes are sensitive to dosage changes in their component parts²⁹, we tested whether dosage effects could be combined in hybrid plants (Supplementary Fig. 8). In this way, a set of genotypes providing a gradient of flowering signal levels could be established from which new architectural optima might arise. To explore this possibility, we created a series of single- and double-mutant heterozygous genotypes in the background of the homozygous sp mutation and compared the flowering and sympodial growth of the corresponding plants.

Plants heterozygous for either sft allele, alone or in combination with ssp allele heterozygosity, exhibited weak delays in the primary shoot flowering of only one leaf, showing similar dosage-dependent effects for the sft-1906 and sft-4537 alleles (Fig. 4a)^{18, 21}. In contrast, plants heterozygous for the ssp alleles transitioned to flowering normally (Fig. 4a). Flowering for all mutant genotypes was delayed within successive sympodial shoots. Double heterozygotes were more severely delayed than single heterozygotes, and the length of the delays increased progressively with the strength of each mutant allele (Fig. 4b). For example, delays in ssp-2129/+ sft-4537/+ plants were greater than in ssp-2129/+ sft-1906/+ plants (Supplementary Table 3). We observed similar progressively longer delays in the sympodial shoots derived from side shoots, indicating an effect on the whole plant (Supplementary Fig. 9). The range of dosage-dependent effects also progressively prolonged sympodial cycling, with ssp-2129/+ sft-1906/+ and ssp-610/+ sft-1906/+ double heterozygotes producing the largest number of sympodial shoots on still-determinate plants (Fig. 4c). ssp-2129/+ sft-4537/+ plants reached a threshold of sp mutant suppression that caused a switch to indeterminate growth in which the average sympodial shoot leaf number was less than in ssp-2129 homozygotes (Fig. 4d,e and Supplementary Fig. 10). We found that this phenotype was due to the frequent occurrence of one-leaf sympodial shoots in ssp-2129/+ sft-4537/+ plants in comparison to their rare occurrence in homozygous ssp-2129 mutants (Figs. 1b and 4d).

To determine whether our continuum of architectures translated to a continuum of yields, we grew plants with all genotypes under ideal field conditions (Online Methods). All plants with genotypes corresponding to determinate architecture were higher yielding than sp mutant controls, and yield increased gradually from single to double heterozygotes and from weaker to stronger alleles (Fig. 5a,b and Supplementary Fig. 11a). ssp-2129/+ sft-1906/+ plants produced the highest yield, nearly doubling that of sp mutants and exceeding the next highest yields, which corresponded to the sft-4537/+ and ssp-610/+ sft-1906/+ genotypes, by 15% and 18%, respectively (Fig. 5a,b and Supplementary Table 4). Processing tomato yield also incorporates fruit sugar content (Brix) into a multiplicative trait known as Brix yield. ssp-2129/+ sft-1906/+ plants produced the highest Brix yield, exceeding that of sp plants by 130% and the next closest yield, corresponding to the sft-4537/+ genotype, by 21% (Fig. 5b and Supplementary Table 4). Because yield and optimal genotypes often vary between environments, we performed a second yield trial in a more challenging location and observed the same trends (Fig. 5c and Supplementary Fig. 11e). Although yields were then more similar between ssp-2129/+ sft-1906/+, ssp-610/+ sft-1906/+ and sft-4537/+ plants, Brix yields for the double heterozygotes were 6% and 8% higher, respectively, than for sft-4537/+ plants (Fig. 5c and Supplementary Table 4). In both experiments, the plants with genotypes corresponding to indeterminate architecture yielded less than the sft-4537/+ heterozygotes, despite the ssp-2129/+ sft-4537/+ and ssp-2129 genotypes corresponding to the largest plants (measured by plant weight; Fig. 5 and Supplementary Fig. 11b,f). Likewise, homozygous ssp-2129 plants produced more leaves on each sympodial shoot yet yielded less than ssp-2129/+ sft-4537/+ plants (Figs. 4e and 5b–d), highlighting that, as the balance of growth shifts from reproductive to vegetative, yield drops. This trend of a peak in optimal yield was maintained in multiple genetic backgrounds and growing seasons; double heterozygosity with either the ssp-2129/+ sft-1906/+ or ssp-610/+ sft-1906/+ genotype gave the highest yields in hybrid crosses with three different elite inbred lines (Supplementary Fig. 12 and Supplementary Table 5). Moreover, the increased yields had no negative impact on fruit quality traits (Supplementary Fig. 11c,d,g,h). These results show that our ssp and sft mutations, through individual and combined heterozygous dosage effects, can fine-tune tomato yield. The new balance of flowering signals in ssp-2129/+ sft-1906/+ and ssp-610/+ sft-1906/+ hybrid plants results in a new optimum for productivity.

Here we identified a way to optimize the yield of field tomatoes by modifying the balance between the florigen hormone and its antagonist, SP^{20, 30}. Our results suggest a general model for increasing productivity in which optimum yields are achieved by balancing flowering signals to shift determinate growth to indeterminate growth (Fig. 5d). Our model suggests that a similar gradient of flowering signals and yields could be achieved through other means. For example, progressively reducing flower-repressing signals could be attained through weaker alleles of sp. As demonstrated in multiple genetic backgrounds, locations and growing seasons, alleles of the florigen pathway genes comprise a flexible toolkit to develop higher-yield hybrids (Supplementary Fig. 12). Moreover, introgression of the sp allele with the ssp-2129 or ssp-610 allele into 'cherry' and 'round' fresh market tomatoes results in plants with an indeterminate architecture and an improved balance of flowering to growth, indicating a potential to rapidly transform the entire tomato breeding industry (Supplementary Fig. 13).

Our findings point to a widespread potential to tap the universal florigen pathway in manipulating the yields of other crops. First, multiple florigen pathway genes can be targeted to achieve beneficial mutations, likely including functionally redundant paralogs of SSP (Supplementary Fig. 4). Second, available evidence suggests that dosage effects from heterozygous mutations in florigen pathway components will be conserved; Arabidopsis FT mutations show dosage effects in the heterozygous state²², and heterozygosity for mutations in a TFL1 homolog in the oil crop rapeseed (Brassica napus L.)³¹ and in an FT paralog in sunflower⁹ improve yield components. Hybrids and the associated benefits of hybrid vigor (heterosis) are the foundation for breeding in most major crops, making it possible to take advantage of combinations of mutations in florigen complex components in hybrid plants. Finally, our work highlights the power of tomato as a model for crops with similar growth habits. For example, soybean³² also undergoes multiple flowering events in a growing season and has both indeterminate and determinate cultivars. Notably, the determinate form is also based on a mutation in an SP homolog selected early in domestication⁷. With recent advancements in gene editing using the CRISPR/Cas system³³, the opportunity now exists to rapidly create targeted mutations to fine-tune and optimize yields in many crops.

Plant material and genotyping.

The ssp mutants were isolated from a large EMS and fast neutron mutation library in the background of the processing cultivar M82 (ref. 34). To map the ssp mutants, we introgressed the sp classic mutation into the background of the wild species Solanum pimpinellifolium by backcrossing five times, and F₂ mapping populations were established. From 216 F₂ plants generated from a cross between ssp-1906 sp and sp in the S. pimpinellifolium background, we identified 32 plants showing indeterminacy and then used bulk segregant mapping with known molecular markers¹⁸ and new markers³⁵ that we generated from the genomic sequence of S. pimpinellifolium to position ssp-1906 near the centromere of chromosome 3, in an interval that included SFT. The SFT coding region was sequenced by Sanger sequencing to identify the responsible mutation. We also used bulk segregant mapping to position ssp-2129 on the long arm of chromosome 2 and used mRNA sequencing to identify the responsible gene. Allelism tests were performed by crossing ssp-2129 sp and ssp-610 sp plants and by crossing sft-1906 sp plants and previously characterized sft mutants¹⁸. For phenotypic analyses, we backcrossed the ssp mutants to the sp parents in the M82 background at least two times to remove background mutations. To generate single-heterozygote genotypes, the sft mutants (sft-1906 sp and sft-4537 sp) and the ssp mutants (ssp-2129 sp and ssp-610 sp) were backcrossed with the sp mutant. To generate double heterozygotes, the sft-1906 sp or sft-4537 sp mutant was crossed with the ssp-2129 sp and ssp-610 sp mutants. To generate hybrids with different genetic backgrounds, ssp-2129 sp, ssp-610 sp, sft-1906 sp, sft-4537 sp, ssp-2129 sft-1906 sp or ssp-610 sft-1906 sp homozygotes in the original M82 background were crossed with three inbred determinate lines developed in the Zamir laboratory: SAR, REB and LEA. Control hybrids were generated by crossing M82 plants with each inbred line.

Identification of mutant variants using mRNA sequencing.

A cDNA library generated from ssp meristem RNA was sequenced on the lane of a HiSeq 2000 (Illumina), and 160 million 101-bp paired-end sequence reads were preprocessed on the basis of quality score before mapping to the tomato genome. Bases with quality scores lower than 36 were trimmed from the 3′ends of reads, and the trimmed reads were then mapped to predicted protein-coding sequences from tomato annotation version 2.3, allowing two mismatches per read, using the Burrows-Wheeler Aligner (BWA)^{35, 36}. Total variants were called using SAMtools, mpileup and bcftools with all default parameters³⁷. In total, 9,690 variants, including SNPs and small insertion-deletions (indels), were called from the mRNA sequencing library for the mutant, of which 2,324 variants were due to differences between M82 and the reference tomato genome, cultivar Heinz. Of the remaining 7,366 unique variants, only 8 were located within the interval on chromosome 2 harboring SSP. Five variants caused nonsynonymous mutations in four genes, including Solyc02g083520, which is the closest homolog of the Arabidopsis FD gene. The Solyc02g083520 mutation in ssp-610 was identified by Sanger sequencing.

In situ RNA hybridization.

We used standard protocols for mRNA in situ hybridization³⁸. Shoot apices were collected from seedlings at 10 DAG (days after germination), 13 DAG and 15–17 DAG to capture vegetative and reproductive meristem stages¹⁹. SSP transcripts were detected with full-length DIG (digoxigenin)-labeled SSP antisense probes.

Yeast two-hybrid analysis.

Protein interaction assays in yeast were performed according to the protocol for the Matchmaker Gold Yeast Two-Hybrid System (GAL4 based; Clontech). The coding sequences for bait proteins (SFT, SP and 14-3-3 proteins) were cloned into the pGBKT7 vector, and the resulting vectors were transformed into the Y2HGold yeast strain. The coding sequences for prey proteins (SSP, ssp-2129, ssp-610 and 14-3-3 proteins) were cloned into the pGADT7 AD vector, and the resulting vectors were transformed into the Y187 yeast strain. After mating the two yeast strains expressing bait and prey proteins, diploid yeast cells were selected and grown on dropout medium not containing leucine and tryptophan. Clones were then selected on triple-dropout medium not containing leucine, tryptophan or histidine for 3 d at 30 °C to assay protein-protein interactions. To block autoactivation caused by interaction of the 14-3-3 proteins, we added 5 mM and 10 mM 3-AT to the triple-dropout medium or removed adenine from the triple-dropout medium. The sequences for all primers used in cloning genes are provided in Supplementary Table 6.

Transcriptome profiling and the digital differentiation index.

We performed transcriptome profiling and DDI analyses as described previously¹⁹. More than 50 meristems from each meristem stage and each genotype were dissected for total RNA extraction. We used 1–2 μg of total RNA for library preparation according to the compatible protocols from the Illumina RNA sequencing library preparation kits (Epicentre, New England BioLabs and Bio Scientific). A pool of four or five libraries was sequenced (101-bp paired-end reads) using the Illumina HiSeq 2000 system. For read mapping, all reads were trimmed to 50 bp to remove low-quality bases and mapped using Bowtie³⁹ to the tomato reference CDS³⁵ with paired-end relationships maintained. Trimming the reads made libraries comparable to our previous mRNA sequencing libraries¹⁹ for combined DDI analyses. The resulting bam alignments were sorted and indexed by SAMtools³⁷, and the number of reads mapping to each coding sequence was determined to calculate the raw counts for all libraries. Raw counts for the meristem expression profiles were incorporated into a master meristem data set that included all raw counts from our previous meristem profiling experiments^{19, 22}. For the master data set, all raw counts were then summarized over replicates and normalized against the total number of mapped reads and the lengths of the coding sequences to calculate RPKM values for DDI analyses⁴⁰.

DDI used tissues with known or predetermined maturation states as calibration points and identified marker genes that showed maximum expression at each calibration point to characterize the calibration points molecularly. DDI checked marker gene expression in query ('unknown') samples and quantified maturation states relative to the calibration points. For each marker gene, DDI compared the expression levels of the unknown sample and each calibration point and calculated a 'maturation score'. Collectively, all marker genes were used to generate a distribution of maturation scores for the unknown sample⁴⁰. At the same time, a Student's t test of the average difference in maturation score between the calibration and unknown samples was conducted for each unknown meristem sample, yielding a P value for the significance of the difference in maturation state. For each prediction, this P value was obtained for comparison of the unknown sample and temporally successive calibration points to generate a 'gradient' of meristem similarity (plotted as heat maps in the form of scaled 1/(–log₁₀ P) values). For example, to predict the maturation state of ssp-2129 sp SYM using the first replicate of wild-type EVM, MVM, LVM, TM and FM samples as calibration points, P values were calculated for the maturation state comparisons SYM versus EVM, SYM versus MVM, SYM versus LVM, SYM versus TM and SYM versus FM. The P values were then transformed into 1/(−log₁₀ P) values and scaled across five values into a range of 0 to 1 (scaling was carried out independently for each prediction). Because smaller P values indicate larger differences in maturation score, the scaled 1/(−log₁₀ P) values quantify the relative similarity of the ssp-2129 sp SYM to each of the five calibration points.

With the master meristem data set, DDI analyses were conducted using five wild-type PSM stages as calibration points to predict the maturation stages for ssp-2129 sp TM and SYM, sft-1906 TM and SYM, and sp TM and SYM (control), and for sft-7187 sp TM. sft-7187 sp plants did not produce SYMs. As described previously^{19, 22}, one replicate of the calibration samples was used for marker gene identification, a second replicate of the calibration samples was treated as unknown and its mutation scores were predicted and plotted to set the boundaries of the maturation stages, and the averaged RPKM values for mutant meristems were used to generate and plot the predicted distribution of maturation scores. All parameters for DDI analyses were as previously described¹⁹. All analyses were carried out using modified R scripts as described previously¹⁹.

Bimolecular fluorescence complementation assays and subcellular localization.

The split-EYFP BiFC system was used to evaluate protein-protein interactions among components of the tomato FAC in tobacco leaf cells by an Agrobacterium-mediated transient gene expression system as described previously^{41, 42}. Coding sequences for the SFT, 14-3-3/74, SSP and ssp proteins were subcloned into the pENTR vector (Invitrogen) and recombined into the pBiFP-2 (nEYFP) or pBiFP-3 (cEYFP) Gateway vectors for protein interaction assays. For subcellular localization assays, the pMDC43 (GFP) and pH35CG (CFP) Gateway vectors were used. Agrobacterium cells carrying pUbi::4NLS-RFP were used to test transformation efficiency and as a nuclear marker. We used N. benthamiana leaves from 3-week-old plants for Agrobacterium infiltration. Through the first 4 d after infiltration, we observed fluorescence signals (EYFP, EGFP and RFP) in the injection area where the nuclear marker was expressed in over 70% of the cells using a Zeiss AXIO Imager Z1 microscope. Fluorescence and DIC (differential interference contrast) images were taken on the third day after infiltration using an LSM 710 laser-scanning confocal microscope (Zeiss). To assay the subcellular localization of SFT-14-3-3/74, the SSP, ssp-2129 and ssp-610 proteins were fused with CFP and coexpressed. Two negative controls were used (CFP expressed in whole cells and RFP expressed in nuclei), along with one positive control (intact SSP). EYFP fluorescence signals from confocal images were captured by z stack. To quantify the SFT-14-3-3/74, interaction signals in each cell, we also captured confocal images by z-stack scanning and quantified EYFP signals, visualizing SFT-14-3-3/74. Whole-cell and nuclear areas were independently drawn on the DIC image showing cell shapes and nuclei marked by fluorescence (Supplementary Fig. 6). Each signal was measured using the Volocity 6.3 application (PerkinElmer). The NLS-RFP and SSP-CFP proteins were used as nuclear markers. All experiments were repeated more than three times for each combination. The sequences of the primers used for cloning genes are listed in Supplementary Table 6.

Phenotyping.

Data on tomato shoot architecture and yield component traits were collected for plants with all genotypes grown in the fields and greenhouse at Cold Spring Harbor Laboratory and the Cornell Long Island Horticultural Experiment Station in Riverhead, New York. Seedlings were grown in 72-cell flats for 35 d and transplanted to the field at the end of May. Plants were grown under drip irrigation and standard fertilizer regimes (for example, 100 pounds of nitrogen per acre). In the greenhouse, the seeds for sp and ssp-2129 plants and single- and double-heterozygote lines were directly sown into pots. Two plants per pot were grown for 3 months under well-controlled water and nutrient conditions and with natural light supplemented with high-pressure sodium bulbs (50 μmol/m²/s; 16-h light/8-h dark cycle). Greenhouse phenotyping was used for a second evaluation of flowering time, shoot determinacy and sympodial shoot number in 3-month-old plants. Each genotype was represented by 16 biological replicates (plants) in the field and 10 replicates in the greenhouse. Damaged or diseased plants were not included in the analyses. Mean values were compared by statistical calculations, using the JMP 7.0.1 software package (SAS Institute), such as by Student's t test.

Field trials and statistical analyses.

Yield trials were performed as previously described²¹ at the Western Galilee Agricultural Experimental Station in Akko, Israel, and at the Cornell Long Island Horticultural Experiment Station in the years 2013 and 2014. Seedlings were grown in nurseries for 35–40 d and transplanted at the beginning of April in Israel and at the end of May in New York. Both experiments were conducted using a completely randomized design under wide spacing of 1 plant/m² for the yield trials in Akko, Israel, under spacing of 1 plant/0.75 m² for the Riverhead experiments using standard drip irrigation and fertilizer regimes. Each genotype was represented by a minimum of 15 biological replicates. Damaged or diseased plants were not included in the analyses. Harvests were performed when the majority of plants in a trial had 80% or more red fruit. Phenotypic measurements of fruit yield and plant weight were carried out after plants and fruits were manually removed from the soil and plant, respectively. Red fruits were collected as mature fruits, and total fruit yield was the sum of the red fruits and green fruits from each plant. Ten fruits were randomly selected to estimate average fruit weight and the total soluble sugar content (Brix) in fruit juice. Brix value (percent) was quantified using a digital Brix refractometer (ATAGO Palette). Brix yield was the total Brix content in the total yield per plant (BY, the multiplied output of Brix and total fruit yield measured in g/m²). Statistical calculations were performed using the JMP 7.0.1 software package (SAS Institute). Mean values for each measured yield parameter were analyzed using the Fit Y by X function and statistically compared using a Dunnett's 'compare with control' test, Tukey-Kramer multiple-comparison test or Student's t test (two-tailed), whenever appropriate.

Accession codes.

mRNA sequencing data have been deposited with the Solanaceae Genomics Network (SGN: http://solgenomics.net/) and are available from ftp://ftp.solgenomics.net/user_requests/LippmanZ/public_releases/by_experiment/Park_etal/.

1. Comadran, J. et al. Natural variation in a homolog of Antirrhinum CENTRORADIALIS contributed to spring growth habit and environmental adaptation in cultivated barley. Nat. Genet. 44, 1388–1392 (2012).
   • PubMed
   • Article
2. Fernandez, L., Torregrosa, L., Segura, V., Bouquet, A. & Martinez-Zapater, J.M. Transposon-induced gene activation as a mechanism generating cluster shape somatic variation in grapevine. Plant J. 61, 545–557 (2010).
3. Repinski, S.L., Kwak, M. & Gepts, P. The common bean growth habit gene PvTFL1y is a functional homolog of Arabidopsis TFL1. Theor. Appl. Genet. 124, 1539–1547 (2012).
   • CAS
   • PubMed
   • Article
4. Ogiso-Tanaka, E. et al. Natural variation of the RICE FLOWERING LOCUS T 1 contributes to flowering time divergence in rice. PLoS ONE 8, e75959 (2013).
5. Iwata, H. et al. The TFL1 homologue KSN is a regulator of continuous flowering in rose and strawberry. Plant J. 69, 116–125 (2012).
   • PubMed
   • Article
6. Liu, B. et al. The soybean stem growth habit gene Dt1 is an ortholog of Arabidopsis TERMINAL FLOWER1. Plant Physiol. 153, 198–210 (2010).
7. Tian, Z. et al. Artificial selection for determinate growth habit in soybean. Proc. Natl. Acad. Sci. USA 107, 8563–8568 (2010).
   • PubMed
   • Article
8. Pin, P.A. et al. An antagonistic pair of FT homologs mediates the control of flowering time in sugar beet. Science 330, 1397–1400 (2010).
   • CAS
   • ISI
   • PubMed
   • Article
9. Blackman, B.K., Strasburg, J.L., Raduski, A.R., Michaels, S.D. & Rieseberg, L.H. The role of recently derived FT paralogs in sunflower domestication. Curr. Biol. 20, 629–635 (2010).
   • CAS
   • PubMed
   • Article
10. Pnueli, L. et al. The SELF-PRUNING gene of tomato regulates vegetative to reproductive switching of sympodial meristems and is the ortholog of CEN and TFL1. Development 125, 1979–1989 (1998).
    • CAS
    • ISI
    • PubMed
11. Taoka, K. et al. 14-3-3 proteins act as intracellular receptors for rice Hd3a florigen. Nature 476, 332–335 (2011).
    • CAS
    • ISI
    • PubMed
    • Article
12. Meyer, R.S. & Purugganan, M.D. Evolution of crop species: genetics of domestication and diversification. Nat. Rev. Genet. 14, 840–852 (2013).
    • CAS
    • PubMed
    • Article
13. Doebley, J., Stec, A. & Hubbard, L. The evolution of apical dominance in maize. Nature 386, 485–488 (1997).
    • CAS
    • ISI
    • PubMed
    • Article
14. Jin, J. et al. Genetic control of rice plant architecture under domestication. Nat. Genet. 40, 1365–1369 (2008).
    • CAS
    • PubMed
    • Article
15. Tan, L. et al. Control of a key transition from prostrate to erect growth in rice domestication. Nat. Genet. 40, 1360–1364 (2008).
    • CAS
    • PubMed
    • Article
16. Yeager, A.F. Determinate growth in the tomato. J. Hered. 18, 263–265 (1927).
17. Rick, C.M. The tomato. Sci. Am. 239, 76–87 (1978).
    • Article
18. Lifschitz, E. et al. The tomato FT ortholog triggers systemic signals that regulate growth and flowering and substitute for diverse environmental stimuli. Proc. Natl. Acad. Sci. USA 103, 6398–6403 (2006).
    • CAS
    • PubMed
    • Article
19. Park, S.J., Jiang, K., Schatz, M.C. & Lippman, Z.B. Rate of meristem maturation determines inflorescence architecture in tomato. Proc. Natl. Acad. Sci. USA 109, 639–644 (2012).
    • PubMed
    • Article
20. Shalit, A. et al. The flowering hormone florigen functions as a general systemic regulator of growth and termination. Proc. Natl. Acad. Sci. USA 106, 8392–8397 (2009).
    • PubMed
    • Article
21. Krieger, U., Lippman, Z.B. & Zamir, D. The flowering gene SINGLE FLOWER TRUSS drives heterosis for yield in tomato. Nat. Genet. 42, 459–463 (2010).
    • CAS
    • PubMed
    • Article
22. Jiang, K., Liberatore, K.L., Park, S.J., Alvarez, J.P. & Lippman, Z.B. Tomato yield heterosis is triggered by a dosage sensitivity of the florigen pathway that fine-tunes shoot architecture. PLoS Genet. 9, e1004043 (2013).
23. Ahn, J.H. et al. A divergent external loop confers antagonistic activity on floral regulators FT and TFL1. EMBO J. 25, 605–614 (2006).
    • CAS
    • ISI
    • PubMed
    • Article
24. Ho, W.W. & Weigel, D. Structural features determining flower-promoting activity of Arabidopsis FLOWERING LOCUS T. Plant Cell 26, 552–564 (2014).
25. Pnueli, L. et al. Tomato SP-interacting proteins define a conserved signaling system that regulates shoot architecture and flowering. Plant Cell 13, 2687–2702 (2001).
    • CAS
    • ISI
    • PubMed
    • Article
26. Wigge, P.A. et al. Integration of spatial and temporal information during floral induction in Arabidopsis. Science 309, 1056–1059 (2005).
    • CAS
    • ISI
    • PubMed
    • Article
27. Abe, M. et al. FD, a bZIP protein mediating signals from the floral pathway integrator FT at the shoot apex. Science 309, 1052–1056 (2005).
    • CAS
    • ISI
    • PubMed
    • Article
28. Jaeger, K.E., Pullen, N., Lamzin, S., Morris, R.J. & Wigge, P.A. Interlocking feedback loops govern the dynamic behavior of the floral transition in Arabidopsis. Plant Cell 25, 820–833 (2013).
    • CAS
    • PubMed
    • Article
29. Veitia, R.A. A generalized model of gene dosage and dominant negative effects in macromolecular complexes. FASEB J. 24, 994–1002 (2010).
30. Lifschitz, E., Ayre, B.G. & Eshed, Y. Florigen and anti-florigen—a systemic mechanism for coordinating growth and termination in flowering plants. Front. Plant Sci. 5, 465 (2014).
31. Guo, Y., Hans, H., Christian, J. & Molina, C. Mutations in single FT- and TFL1-paralogs of rapeseed (Brassica napus L.) and their impact on flowering time and yield components. Front. Plant Sci. 5, 282 (2014).
32. Graham, P.H. & Vance, C.P. Legumes: importance and constraints to greater use. Plant Physiol. 131, 872–877 (2003).
    • CAS
    • PubMed
    • Article
33. Belhaj, K., Chaparro-Garcia, A., Kamoun, S. & Nekrasov, V. Plant genome editing made easy: targeted mutagenesis in model and crop plants using the CRISPR/Cas system. Plant Methods 9, 39 (2013).
34. Menda, N., Semel, Y., Peled, D., Eshed, Y. & Zamir, D. In silico screening of a saturated mutation library of tomato. Plant J. 38, 861–872 (2004).
    • CAS
    • ISI
    • PubMed
    • Article
35. Tomato Genome Consortium. The tomato genome sequence provides insights into fleshy fruit evolution. Nature 485, 635–641 (2012).
    • CAS
    • ISI
    • PubMed
    • Article
36. Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25, 1754–1760 (2009).
    • CAS
    • ISI
    • PubMed
    • Article
37. Li, H. et al. The Sequence Alignment/Map format and SAMtools. Bioinformatics 25, 2078–2079 (2009).
    • CAS
    • ISI
    • PubMed
    • Article
38. Jackson, D., Veit, B. & Hake, S. Expression of maize KNOTTED1 related homeobox genes in the shoot apical meristem predicts patterns of morphogenesis in the vegetative shoot. Development 120, 405–413 (1994).
    • CAS
    • ISI
39. Langmead, B. Aligning short sequencing reads with Bowtie. Curr. Protoc. Bioinformatics Chapter 11, Unit 11.7 (2010).
40. Efroni, I., Blum, E., Goldshmidt, A. & Eshed, Y. A protracted and dynamic maturation schedule underlies Arabidopsis leaf development. Plant Cell 20, 2293–2306 (2008).
    • CAS
    • ISI
    • PubMed
    • Article
41. Voinnet, O., Rivas, S., Mestre, P. & Baulcombe, D. An enhanced transient expression system in plants based on suppression of gene silencing by the p19 protein of tomato bushy stunt virus. Plant J. 33, 949–956 (2003).
    • CAS
    • ISI
    • PubMed
    • Article
42. Desprez, T. et al. Organization of cellulose synthase complexes involved in primary cell wall synthesis in Arabidopsis thaliana. Proc. Natl. Acad. Sci. USA 104, 15572–15577 (2007).
    • PubMed
    • Article

Download references

We thank members of the Lippman laboratory for discussions, C. Brooks for technical assistance, T. Mulligan and P. Hanlon for plant care, and staff from Cornell University's Long Island Horticultural Research and Extension Center in Riverhead, New York, and the Western Galilee Experimental Station in Akko, Israel, for assistance with the yield trials. We also thank M. Lodha and D. Jackson (Cold Spring Harbor Laboratory) for providing vectors, F. Yang and Y.K. Lee for guidance on in situ hybridization and Agrobacterium infiltration in tobacco, and S. Hearn for assistance with confocal microscopy. This research was supported by a European Research Council (ERC)-Advanced grant entitled YIELD to D.Z., by Israel Science Foundation (ISF; 1294-10) and Binational Agricultural Research and Development Fund (BARD; IS4536-12C) grants to Y.E. and by a grant from the National Science Foundation (NSF) Plant Genome Research Program (1237880) to Z.B.L.