LA1589 carries a single copy of fw3.2 and a wild-type (wt) allele of the M9 SNP. Seeds were sown directly in the soil in 72-count 6 pack trays and grown in a growth chamber under 16-h light/8-h dark photoperiod for 5–6 weeks. The 5- to 6-week-old seedlings were transplanted into 2.45 L pots and were cultivated in a greenhouse under 16-h light/8-h dark photoperiod in Athens, GA, United States. All the plants of each experimental evaluation were randomly arranged in the greenhouse.

The gRNA targeting the M9 SNP was designed using the CRISPR-P tool¹ (Lei et al., 2014). The CRISPR/Cas9 construct was assembled using the Golden Gate cloning method as previously described (Wu et al., 2018). Electroporation was used to introduce the final binary vector into Agrobacterium tumefaciens strain LBA4404, which was kindly provided by Dr. Joyce Van Eck, Cornell University. The LBA4404 harboring the binary vector was used for the transformations of LA1589. The genetic transformations of LA1589 were performed as described (Gupta and Van Eck, 2016) at Wayne Parrott’s Laboratory, University of Georgia.

For genotyping of T₀ transgenic plants, genomic DNA was extracted from leaves and was used for genotyping by PCR for the presence of the Cas9 using two specific primer pairs (14EP426/14EP427; 14EP438/14EP439) on the Cas9 coding region. The target region was amplified by PCR from the genomic DNA using specific primers flanking the gRNA target sites (17EP42/17EP47). The PCR products were purified and directly sequenced by Sanger sequencing. The mutations in T₀ generation were analyzed by decoding sequencing chromatograms (Ma et al., 2015c).

T₀ transgenic plants were backcrossed to LA1589, and the seeds of the F₁ generation were extracted and sown in 288-well plastic flats. Each progeny was genotyped by PCR using primer flanking of the target region (17EP218/EP2460), and PCR products were resolved on 3.5% (w/v) agarose gel. The progenies harboring small indels that cannot be easily identified by PCR assay were further genotyped by Derived Cleaved Amplified Polymorphic Sequences (dCAPS)-BsrBI, dCAPS-AciI, and dCAPS-HaeIII. For Cas9-free F₁ plants, they were self-pollinated for the generation of homozygous mutants in the F₂ generation. For the F₁ plants carrying the Cas9 transgene, they were backcrossed with LA1589, and the Cas9-free homozygotes were obtained in the F₃ generation. All the mutant alleles from the selected plants in F₁, F₂, and F₃ generations were confirmed by Sanger sequencing, and the presence of Cas9 was determined by two specific primers in its coding region (Supplementary Table 4).

The phenotyping was performed on homozygous mutants with three experimental evaluations during 2018 and 2019, each with at least three plants per genotype (Supplementary Table 3). Three stems were kept and trained on a bamboo stick, respectively. We kept six inflorescences per plant with eight fruits on each inflorescence. All the flowers were pollinated by hand. We numbered the fruit position from 1 to 8 according to the proximal to distal positions on each inflorescence, and the inflorescences were numbered on each plant from 1 to 6 according to the harvesting time. The fruit weight was measured individually using a precision balance (VWR 64B). For fruit weight distribution analysis, fruits were grouped into four categories based on the quartiles of the fruit weight from all genotypes in each experimental evaluation.

Since SlKLUH showed very low expression in developing fruits and very high expression in young flower buds in LA1589 (Chakrabarti et al., 2013), young flower buds at 9–13 days post initiation (dpi) (Supplementary Figure 7) were collected from at least three inflorescences per plant. Total RNA from meristems was then extracted using the TRIzol^® Reagent (Thermo Fisher, United States). Total RNA was used for cDNA synthesis with a High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher, United States). Quantitative real-time PCR (qRT-PCR) was performed with gene-specific primers using the SsoAdvanced™ Universal SYBR^® Green Supermix (Bio-Rad, United States) reaction system on the CFX96 Real-Time system (Bio-Rad, United States), following manufacturer’s instructions. Clathrin adaptor complexes medium subunit (CAC) gene was used as an internal control (González-Aguilera et al., 2016; Supplementary Table 4).

Statistical analysis was performed in R. The information of the statistical test is given in the respective figures.

The M9 SNP is located within the second repeat of four 30-bp tandem repeats (Supplementary Figure 1). Homology-based prediction using Multiple Em for Motif Elicitation (MEME) indicated that the four tandem repeats are conserved motifs in orthologous KLUH promoters (Figure 1A), suggesting that the motifs might be the important regulatory elements of the CYP78A genes. Since active gene regulatory elements are associated with open chromatin (Zhu et al., 2015; Lu et al., 2019; Yan et al., 2020), ATAC-seq was utilized in tomato meristem and leaf tissues (Hendelman et al., 2021). Significant peaks covering the repeats were detected in both tissues, indicating that this region might be an important regulatory region of SlKLUH (Figure 1B). Many genetic changes underlying traits of economic importance reside in cis-regulatory elements (CREs) and cis-engineering mediated by CRISPR/Cas for crop improvement can be utilized to expedite the modification of these CREs (Li Q. et al., 2020; Hendelman et al., 2021). Therefore, we hypothesized that engineering fruit weight variation could be implemented by editing the presumptive CRE of SlKLUH.

We tested the hypothesis by targeting the CRE using CRISPR/Cas9 with a single gRNA in the wild relative of the cultivated tomato LA1589 (Solanum pimpinellifolium) (Figure 1C). Seven first-generation transgenic (T₀) individuals were obtained (Supplementary Figure 2). The disrupted target sites detected by PCR and Sanger sequencing suggested that two plants were biallelic for m3_+1bp/m9_–5bp (T₀–1) or m3_+1bp/m6_–2bp (T₀–15), two plants were homozygous for m6_–2bp (T₀–9) or m16_–10bp (T₀–11), and other two plants were heterozygous for m3_+1bp (T₀–12) or chimeric for at least three alleles (T₀–10) (Supplementary Figure 2 and Supplementary Table 1).

To enrich for SlKLUH promoter mutant alleles covering a range of fruit weight variations, the CRISPR/Cas9-driven mutagenesis approach was utilized (Rodriguez-Leal et al., 2017). A sensitized population of 719 F₁ plants was generated by the backcrossing of T₀ lines with LA1589 and genotyped by PCR and Restriction Enzyme (PCR/RE) analysis (Supplementary Figure 3). F₁ progenies from T₀–9 crossed to LA1589 resulted in the identification of three new alleles, namely, m2_+4bp, m14_{–9 bp}, and m21_{–60 bp}, based on indels observed by gel electrophoresis mobility shift after PCR using primers flanking the target region (Figure 2; Supplementary Figure 3; Supplementary Table 2). To identify small indels that are different from m6_–2bp or wt, we developed a screen scheme that exploits dCAPS analysis according to the sequences of m6_–2bp and wt and the recognition sequences of the three REs (BsrBI, HaeIII, and AciI). The remaining progenies were genotyped by dCAPS-BsrBI, which can digest m6_–2bp and wt. The new mutations that do not carry the restriction site for the digestion with BsrBI were further analyzed by dCAPS-HaeIII and/or dCAPS-AciI, by which we identified one plant heterozygous for m8_–4bp (Figure 2; Supplementary Figure 4; Supplementary Table 2). With this approach, we identified another 13 new alleles from the progenies of T₀–10, T₀–12, and T₀–15 (Figure 2; Supplementary Figure 4; Supplementary Table 2).

Collectively, a total of 21 mutant alleles were created in the LA1589 background (Figure 2 and Supplementary Table 2), including three insertion alleles (1, 4, and 98 bp) and deletions of various sizes (from 1 to 60 bp) (Figure 2). Cas9 often cleaves double-strand DNA at a position of 3-bp upstream of the protospacer adjacent motif (PAM) sequence, and most mutations occurred at the 4th base from the PAM site (Jinek et al., 2012). However, we found that eight alleles (m5_–1bp, m9_–5bp, m11_–7bp, m12_–8bp, m15_–9bp, m17_–10bp, m19_–18bp, and m21_–60bp) did not occur right upstream of the predicted double-strand break (DSB) position. These results indicated the high efficiency of the CRISPR/Cas9-driven mutagenesis screen approach creating a collection of novel mutant alleles, including unexpected mutations, in tomatoes. F₁ plants carrying novel alleles that were Cas9 positive were backcrossed to LA1589 to segregate away the Cas9 transgene to avoid further edits and minimize off-target effects. The Cas9-free mutants homozygous for these mutant alleles were obtained in F₂ and F₃ generations and used for further analysis.

The plants homozygous for the 21 alleles were grown in December 2018 with at least three plants per genotype (Supplementary Table 3). They displayed a continuum of fruit weight variation between 0.92 and 1.14 g (Figure 3A and Supplementary Table 3). Compared to LA1589, plants homozygous for nine alleles (m2_+4bp, m3_+1bp, m5_–1bp, m18_–11bp, m13_–8bp, m14_–9bp, m21_–60bp, m1_+98bp, and m17_–10bp) showed more than 5% variation in fruit weight (Figure 3A). Notably, fruit weight of m2_+4bp, m3_+1bp, m5_–1bp, m18_–11bp, m13_–8bp, and m14_–9bp homozygotes was greater by 10.68, 8.74, 8.74, 8.74, 7.77, and 7.77% compared to LA1589, respectively. In contrast, fruit weight decreased by 8.74, 9.71, and 10.68% for m21_–60bp, m1_+98bp, and m17_–10bp, respectively (Figure 3A and Supplementary Table 3).

We then selected eight mutant alleles for further analysis, including five larger-fruited alleles (m2_+4bp, m3_+1bp, m5_–1bp, m13_–8bp, and m14_–9bp) and three smaller-fruited alleles with large insertion/deletions (m1_+98bp, m20_–46bp, and m21_–60bp). Notably, statistically significant increases in average fruit weight were observed in plants homozygous for m2_+4bp and m3_+1bp (Figure 3A and Supplementary Table 3). The increased fruit weight of the homozygotes m5_–1bp, m13_–8bp, and m14_–9bp was still observed and showed 5% or more variation in fruit weight compared to LA1589 in May 2019 and/or September 2019 (Figure 3A and Supplementary Table 3). Collectively, m2_+4bp, m3_+1bp, m5_–1bp, m13_–8bp, and m14_–9bp homozygous mutants had 14.07, 9.60, 5.09, 6.26, and 7.06% greater fruit weight compared to the LA1589 (Figure 3A and Supplementary Table 3). However, the smaller-fruited alleles with large insertions/deletions, m1_+98bp, m20_–46bp, and m21_–60bp, had inconsistent effects on fruit weight between December 2018 and September 2019 (Figure 3A and Supplementary Table 3), potentially due to environmental effects. In the following experiments, we focused only on the larger fruit mutant alleles (m2_+4bp, m3_+1bp, m5_–1bp, m13_–8bp, and m14_–9bp) as they showed consistent effects on fruit weight across all experimental evaluations.

To better describe the increased fruit weight in the mutants homozygous for m2_+4bp, m3_+1bp, m5_–1bp, m13_–8bp, and m14_–9bp, we performed a fruit weight distribution analysis. Fruits were grouped into four categories based on the quartiles of fruit weight in all genotypes in each experimental evaluation. The mutant alleles displayed a range of quantitative effects on the distribution of fruit weight in each experimental evaluation (Supplementary Figure 5). Importantly, the five large-fruited alleles (m2_+4bp, m3_+1bp, m5_–1bp, m13_–8bp, and m14_–9bp) showed a significant decrease in the proportion of the small fruits (gray bar) across all or the majority of the experimental evaluations (Figure 3B and Supplementary Figure 5). The results indicated that the decreased proportion of the small fruits is responsible for the increased fruit weight in the five mutants.

Previous studies in the domesticated tomato demonstrated that fruits in the same inflorescence generally differ in size from the larger ones at the proximal position (Beadle, 1937; Bangerth and Ho, 1984) and similar results were also observed in LA1589 (Supplementary Figure 6). Fruit weight showed a decreasing trend from the 1st (proximal) to 8th (distal) fruit in the inflorescence, and the decrease rate of fruit weight was 29.1% in December 2018, 35.0% in May 2019, and 13.8% in September 2019 (Supplementary Figure 6).

Given that the five mutant alleles can increase fruit weight and decrease the proportion of small fruits, we hypothesized that this could be achieved by reducing the decreasing trend along with the position of an individual inflorescence or increasing the weight of all fruits on each inflorescence. To explore this, two-way ANOVA was performed using the weight and position along the inflorescence. As expected, the fruit weight was significantly affected by genotype and fruit position (P < 0.05) (Table 1). However, the interaction of genotype and fruit position had no significant effect (P > 0.05) in all experimental evaluations (Table 1 and Figure 4), indicating that the five mutant alleles had no significant effect on the rate of fruit weight decrease along inflorescence. Therefore, the five mutant alleles can increase the weight of all fruits along inflorescence without affecting the decreasing rate.

The effects of promoter mutations on the phenotypic changes are often unpredictable and unexpected due to the complexity of transcriptional controls (Rodriguez-Leal et al., 2017). In this study, eight alleles (m1_+98bp, m2_+4bp, m3_+1bp, m4_–1bp, m5_–1bp, m6_–2bp, m7_–3bp, and m15_–9bp) had the unaltered wt M9 SNP (Figure 2), among which m2_+4bp, m3_+1bp, and m5_–1bp showed an increase in fruit weight, while other five showed no or inconsistent effects on fruit weight in this study (Figure 3 and Supplementary Table 3). These results suggested that the M9 SNP had no or minor effects on tomato fruit weight, which was supported by the finding of the SlKLUH gene duplication underlying fw3.2(Alonge et al., 2020). In addition, although many alleles shared overlapping deletions, they had different phenotypic effects. For example, m13_–8bp and m14_–9bp showed consistent effects on increasing fruit weight, while m20_–46bp and m21_–60bp had opposite effects on fruit weight between December 2018 and September 2019 (Figure 3A and Supplementary Table 3).

Given that the lower expression of SlKLUH results in smaller fruits (Chakrabarti et al., 2013; Alonge et al., 2020), we tested whether the expression levels of SlKLUH were upregulated in the mutants homozygous for the five larger fruit mutant alleles. We evaluated the SlKLUH expression in young flower buds at 9–13 dpi (Supplementary Figure 7). Although increased fruit weight was observed for the five large-fruited mutant alleles compared to LA1589, the SlKLUH expression levels from them are comparable to LA1589 (Figure 5A). Remarkably, there was a low correlation between the SlKLUH expression levels and fruit weights (Figure 5B). Similar findings were also observed for the tomato lc allele that is caused by two SNPs in a 15-bp repressor element downstream of tomato WUSCHEL (SlWUS) (van der Knaap et al., 2014; Rodriguez-Leal et al., 2017). Subtle differences in SlWUS expression level were not captured by RT-PCR, resulting in larger fruit size (Muños et al., 2011; van der Knaap et al., 2014; Rodriguez-Leal et al., 2017). These results support that allele types and transcriptional changes are poor predictors of phenotypic changes as previously noted (Rodriguez-Leal et al., 2017; Hendelman et al., 2021).