Introduction

The dominance of flowering plants on earth owes much to the evolution of maternal tissues such as seedcoat and fruit that protect and disseminate the seeds. However, how fertilization induces the development and differentiation of these maternal tissues is not well understood. Strawberry has a history of serving as a model to study how fertilization induces fruit development due to its unique flower and fruit structure. The fleshy fruit develops from the receptacle, an enlarged stem tip supporting hundreds of seed-containing ovaries (achenes)1,2. The achenes dotting the surface of the receptacle are easily accessible for dissection and manipulation, and their individual impact on the underlying receptacle fruit flesh development is visible in real time as they develop. When achenes were removed from the receptacle, no fruit flesh would form; however exogenous application of auxin promoted receptacle fruit enlargement in the absence of achenes3. Therefore, auxin was identified as an inductive signal produced by the seed to stimulate fruit development3,4,5. The essential role of auxin in fruit induction is supported by genetic studies in plants such as Arabidopsis and tomato, where mutants with defects in auxin signaling develop virgin (or parthenocarpic) fruits6,7,8,9,10,11, suggesting a conserved auxin-dependent mechanism in flowering plants. Genome-wide transcript profiling and reporter gene expression in diploid strawberry, Fragaria vesca, revealed that the expression of auxin biosynthesis genes, YUCs and TAAs, are primarily induced in the endosperm immediately post-fertilization5,12. However, the molecular mechanism of how fertilization induces auxin synthesis in the endosperm for fruit induction is presently unknown.

In seeds, the embryo and endosperm are the products of double fertilization of the egg cell and central cells, respectively. While the endosperm is largely thought of as a nutritive tissue for the embryo, increasing evidence suggests the endosperm serves as a fertilization ‘sensor’ that coordinates the development of the embryo and the surrounding maternal sporophytic tissues such as seedcoat and fruit13. In Arabidopsis, transgenic expression of auxin biosynthesis genes, DD65::TAA1 and DD65::YUC6, in the central cell initiates parthenocarpic endosperm proliferation, seedcoat differentiation, and fruit development14,15. These studies not only established auxin as the fertilization-induced signal but also pinpointed endosperm as the tissue that senses fertilization, produces auxin, and exports auxin to the seedcoat and fruit.

In most angiosperms, endosperm development initiates as a syncytium in which the fertilized central cell nucleus divides several times without cytokinesis. Sometime later, the timing varying between species, the endosperm nuclei become cellularized16,17. Importantly, the timing of endosperm cellularization determines the extent of endosperm nuclear proliferation and hence seed size and weight. In Arabidopsis, fertilization-induced auxin production is essential for endosperm nuclear proliferation14. Increased or prolonged auxin biosynthesis in the endosperm prevents endosperm cellularization and leads to seed arrest, but reduced auxin biosynthesis can restore cellularization, suggesting that the auxin level in the endosperm may determine the timing of endosperm cellularization18. Thus, understanding the regulation of auxin biosynthesis in endosperm has significant impact on grain and fruit yields.

In Arabidopsis, AtAGL62, a type I MADS-box gene, was identified as a regulator of endosperm cellularization. AtAGL62 is expressed exclusively in early endosperm development, during the nuclear proliferation phase, and its expression declines quickly before cellularization19,20. While atagl62 loss-of-function mutants showed precocious endosperm cellularization, ectopic AtAGL62 expression blocked endosperm cellularization19,21,22. As such, the effect of AtAGL62 mirrors that of auxin in promoting endosperm nuclear proliferation and inhibiting endosperm cellularization. However, the functional relationship between AtAGL62 and auxin biosynthesis remains unknown. A prior study suggested that AtAGL62 positively regulates the expression of an ABCB-type transporter that is required for auxin transport from endosperm to the seedcoat15,20. This observation does not, however, explain the similar effects of AtAGL62 and auxin in endosperm cellularization and development.

In this work, we investigated the mechanisms of how fertilization induces auxin biosynthesis in the endosperm to promote fleshy fruit initiation using diploid wild strawberry Fragaria vesca as our model. We showed that FveAGL62 promotes fruit development by promoting auxin biosynthesis in the endosperm. Further, we demonstrated that Arabidopsis AtAGL62 similarly regulates auxin synthesis in the endosperm. Thus, our investigation has uncovered an evolutionarily conserved mechanism underlying fertilization-induced auxin synthesis in the endosperm. Our finding is broadly relevant to flowering plants for producing seeds and fruits and provides a knowledge base for increasing grain and fruit size as well as creating virgin (parthenocarpic) fruits.

Results

FveAGL62 is essential for seed and fruit development in F. vesca

To identify key regulators that underlie fertilization-induced auxin biosynthesis in seeds, we mined prior F. vesca co-expression networks based on transcriptome data from distinct flower and fruit tissues at different stages5,23,24. A consensus co-expression cluster was identified with its eigengene highly and preferentially expressed in the stage 2 seeds (Supplementary Fig. 1a; Supplementary Data 1). Stage 2 seeds are formed at 2-4 days post anthesis (DPA)1 and contain transcripts induced by fertilization. Among the transcription factors in the cluster (Supplementary Fig. 1b), two type I MADS-box genes FvH4_2g03030 and FvH4_6g08460 exhibit highly specific and transient expression in the ghost (endosperm and seed coat) (Supplementary Fig. 1d, e) and are chosen for further characterization.

First, phylogenetic analysis of 45 F. vesca and 74 Arabidopsis type-I MADS-box genes (Supplementary Fig. 1c) indicates that FvH4_2g03030 clusters with AtAGL62 and FvH4_6g08460 clusters with AtAGL80 and are thereafter referred to as FveAGL62 and FveAGL80 respectively. Next, to verify their expression, transgenic F. vesca plants containing reporter genes ProFveAGL62::FveAGL62-GUS and ProFveAGL80::FveAGL80-GUS were generated. In each case, GUS expression was detected in the endosperm at stage 2 (2-4 DPA) but not in the pre-fertilization stage 1 ovule (Fig. 1a–c), confirming that FveAGL62 and FveAGL80 are specifically expressed in the endosperm soon after fertilization.

Fig. 1: Characterization of F. vesca mutants of FveAGL62 and FveAGL80.
figure 1

a ProFveAGL62::FveAGL62-GUS reporter expression in stage 1 ovule (pre-fertilization) and stage 2 seed (post-fertilization). b ProFveAGL80::FveAGL80-GUS reporter expression in stage 1 ovule and stage 2 seed. c Diagram of a stage 2 strawberry seed. d Mature F. vesca plants of WT, fveagl62 single, fveagl80 single, and fveagl62 fveagl80 double mutant. fveagl62 single and fveagl62 fveagl80 double mutants fail to produce fruits. e Comparison of receptacle fruit development at the three earliest stages of fruit development. f WT and fveagl62 stage 1 ovules and stage 2 seeds. g Fruits from self-cross or reciprocal crosses between WT and fveagl62 (female genotype is listed first). Only self-cross of fveagl62 leads to failed fruit growth. h GUS expression in stage 2 seeds derived from reciprocal crosses between ProFveAGL62::FveAGL62-GUS transgenic plants and WT. i fveagl62 mutant fruits and seeds after different hormone treatments. Scale bars: 0.12 mm (a, b, h), 5 cm (d), 0.5 cm (e), 1 mm (f), 1 cm (g), 0.5 cm (fruit) (i) and 1 mm (seed) (i).

CRISPR/Cas9 was used to knockout FveAGL62. Two sgRNA (sgRNA1 and sgRNA3) targeting different regions of the FveAGL62 ORF were independently cloned and transformed into F. vesca. Transgenic plants containing the Cas9-sgRNA1 led to three independent mutant lines, which carried either homozygous or biallelic frameshift mutations (Supplementary Fig. 2a). Transgenic plants containing the Cas9-sgRNA3 yielded three additional lines, all of which contained heterozygous frameshift mutations (Supplementary Fig. 2b). The three homozygous or biallelic fveagl62 mutant lines, fveagl62−1/−4, fveagl62−1/−1, and fveagl62−1/−5, showed similar fruit phenotypes. Specifically, they were morphologically similar to wild type during vegetative growth (Fig. 1d), but their fruits did not develop further beyond stage 2 (Fig. 1e; Supplementary Fig. 3a). While fveagl62 ovules at stage 1 showed no discernable difference from those of wild type, fveagl62 seeds at stage 2 were much smaller than wild type, appeared to dry up, and were not viable (Fig. 1f).

CRISPR/Cas9 was also used to generate a fveagl80−1/−1 mutant (Supplementary Fig. 2c), which exhibited normal vegetative growth and developed fruit similarly to wild type (Fig. 1d). Since FveAGL80 (FvH4_6g08460) clusters together with four additional FveAGL80-Like (FveAGL80L) genes in F. vesca (Supplementary Fig. 1c), functional redundancy may have prevented any mutant phenotype of fveagl80−1/−1. Finally, fveagl62−1/−4, fveagl80−1/−1 double mutants exhibited a similar phenotype as the fveagl62−1/−4 single mutant (Fig. 1d).

MADS-box proteins often form heterodimers or tetramers in executing their functions, so we investigated whether FveAGL62 and FveAGL80 act as heterodimers like their Arabidopsis homologs19,25. Both yeast two-hybrid (Y2H) and bimolecular fluorescence complementation (BiFC) assays showed that FveAGL62 and FveAGL80 interacted with each other but neither could homodimerize (Supplementary Fig. 2d, e). BiFC was also used to test any interaction between FveAGL62 and the other four FveAGL80Ls. Positive interactions were observed between FveAGL62 and all four FveAGL80Ls, but no interaction was found between FveAGL80 and FveAGL80Ls (Supplementary Fig. 4). The data suggest that FveAGL62 can form heterodimers with FveAGL80 as well as FveAGL80L1-4.

Characterization of fveagl62 mutants in strawberry

To determine whether FveAGL62 is subject to parental imprinting, reciprocal crosses between fveagl62 and wild type F. vesca were performed. The resulting F1 seeds and fruits were normal whether the fveagl62 mutant served as the female or male donor (Fig. 1g). Only when both parents carried the fveagl62 mutation, did the F1 progeny exhibit the mutant phenotype as shown in a self-cross of the fveagl62 mutant (Fig. 1g). Reciprocal crosses were performed between a ProFveAGL62::FveAGL62-GUS transgenic plant and a non-transgenic WT plant; the GUS reporter expression was detected in the endosperm nuclei of young F1 seeds whether the ProFveAGL62::FveAGL62-GUS transgenic plant served as the male or female parent (Fig. 1h). Therefore, FveAGL62 is not subjected to parental imprinting, which is similar to the Arabidopsis AtAGL6219.

We then asked whether the fveagl62 seed and fruit defects were caused by failures of auxin and GA biosynthesis. GA3 and auxin (NAA), alone or in combination, were applied to fertilized fveagl62 mutant flowers and shown to cause fruit and seed enlargement, with the combined GA3/NAA treatment having the strongest effect (Fig. 1i). However, hormone treated fveagl62 seeds eventually died, indicating that fveagl62 may cause additional defects other than hormone production. Nevertheless, this result suggested that a loss of FveAGL62 might result in reduced auxin and GA production and/or transport.

RNA-seq profiling of fveagl62 stage 2 seeds

To identify downstream genes and processes regulated by FveAGL62 during seed development, we used RNA-seq to profile stage 2 fveagl62-1/−4 seeds and wild type seeds, each with three biological replicates. Differential gene expression (DE) analysis showed that 1102 and 1572 genes are respectively down- and up- regulated in the fveagl62−1/−4 seeds (Supplementary Data 2). Among the down-regulated genes in fveagl62, enriched GO terms include cell wall modification and organization processes and seed development (Supplementary Fig. 5a; Supplementary Data 3). In contrast, enriched GO terms for up-regulated genes in fveagl62 include responses to biotic stimuli and response to abscisic acid (Supplementary Fig. 5b; Supplementary Data 3).

Among all 30 type I MADS-box genes expressed in the WT or fveagl62 stage 2 seeds (Supplementary Fig. 5c), FveAGL80, FveAGL80L1, and FveAGL80L2 were also significantly down-regulated in the fveagl62 seeds. Hence, FveAGL62 may positively regulate FveAGL80 and FveAGL80L expression. The existence of these similarly regulated FveAGL80/FveAGL80L genes also explains a lack of phenotype of fveagl80−1/−1 single mutants.

Most significantly, the RNA-seq data showed that the expression of several auxin biosynthesis genes, FveYUC1, FveYUC5, FveYUC10, FveTAA1, FveTAR1, and FveTAR2, are reduced in the fveagl62 stage 2 seeds (Fig. 2a). RT-qPCR confirmed reduced transcript levels of five auxin biosynthesis genes in the fveagl62 stage 2 seeds when compared with wild type (Fig. 2c). Similarly, the GA biosynthesis genes FveGA20OX1c, FveGA20OX1d, FveGA3OX1a, and FveGA3OX1b also showed reduced expression in fveagl62 mutant seeds (Fig. 2b). Together, a loss of FveAGL62 caused significant reduction in the transcript level of auxin and GA biosynthesis genes in stage 2 seeds.

Fig. 2: F. vesca fveagl62 mutant seeds show reduced auxin biosynthesis.
figure 2

a Hierarchical clustering of auxin biosynthesis gene expression in WT and fveagl62 stage 2 seeds. b Hierarchical clustering of GA biosynthesis gene expression in WT and fveagl62 stage 2 seeds. c RT-qPCR analysis of five auxin biosynthesis genes in WT and fveagl62 stage 2 seeds. Gene IDs are indicated in a. Y-axis indicates the relative expression level against the control gene FvePP2a (FvH4_4g27700). Significant difference by two-tailed Student’s t-test is indicated by ** (P < 0.01) or * (P < 0.05) between WT and fveagl62 mutant seed. Error bars indicate standard deviation. Similar results were obtained in three biologically independent experiments. d DR5ver2::GUS reporter expression in WT and fveagl62 stage 2 seeds. e DR5ver2::GUS expression in individual stage 2 seeds. f Confocal laser scanning microscopy images of WT and fveagl62 seeds at stage 2. Precocious cellularization of fveagl62 endosperm (arrows) is observed in stage 2 seeds. g Confocal laser scanning microscopic images of WT and fveagl62 stage 2 seeds mock- or NAA-treated. Scale bars: 1 mm (d), 120 μm (e), 100 μm (fg).

We subsequently examined a prior F. vesca RNA-seq dataset26 for the temporal expression pattern of these auxin and GA biosynthesis genes in seeds. We found that the transcripts of many auxin biosynthesis genes, FveYUC5 and FveYUC10, FveTAA1, and FveTAR1-3 (Supplementary Fig. 6a), and GA biosynthesis genes, FveGA3OX1a, 1b, 1c, 1d and FveGA20OX1d (Supplementary Fig. 6b), exhibited a gradual increase from stage 1 ovules to stage 3 seeds. RT-qPCR was performed for five auxin biosynthesis genes and four GA biosynthesis genes (Supplementary Fig. 6c), all of which showed the same expression trend from a low expression at stage 1 ovules to a high expression at stage 3 ghost (endosperm and seedcoat), consistent with their roles in fertilization-induced auxin and GA biosynthesis in seeds.

Characterization of fveagl62 endosperm phenotype

To determine whether the reduced transcript level of auxin biosynthesis genes leads to reduced auxin activity in the fveagl62 seeds, an auxin reporter DR5ver2::GUS12,27 was introduced into the fveagl62 mutant by genetic crosses. Compared with the wild type, DR5ver2::GUS expression is greatly reduced in the fveagl62 stage 2 seeds (Fig. 2d, e), although the expression pattern is similar with more intense staining at the chalazal end of the endosperm and seed coat, which was proposed as the auxin transport route to the receptacle12. The result established that the fveagl62 mutant seeds contain a lower auxin activity, which likely contributes to defects in the fveagl62 seed and fruit development.

We used confocal laser scanning microscopy (CLSM) to examine fveagl62 seeds in detail. Strawberry endosperm initially consists of a thin and elongated tube (Fig. 1c). The wild type stage 2 endosperm consists of several nuclei situated at the peripheral of the tube. In contrast, all three fveagl62 mutant lines showed premature endosperm cellularization when the seeds were examined at stage 2 (Fig. 2f; Supplementary Fig. 3b). Both the confocal image and the confocal stack video (Supplementary movies 1 and 2) clearly illustrate the tube-like endosperm filled with a single file of cells separated by cell walls. Hence, similar to Arabidopsis atagl62 mutants19, strawberry fveagl62 also exhibits premature endosperm cellularization.

To eliminate the possibility that the observed early cellularization phenotype was due to off-target mutations, we also examined fveagl62 mutant seeds derived from an independent CRISPR experiment, where a Cas9-sgRNA3 construct targeting a different sequence of FveAGL62 was created and transformed into F. vesca. Three transgenic lines respectively with a −1, −1, or −4 bp mutation were obtained (Supplementary Fig. 2b). Since they are all heterozygous, the self-fertilized seeds of line #1 with a −1 bp mutation were examined under the confocal microscope. 25.9% of the F1 seeds (n = 54) exhibited early cellularization of the endosperm (Supplementary Fig. 3c), supporting that the early endosperm cellularization was due to a defective FveAGL62.

We then tested if reduced auxin biosynthesis in fveagl62 mutant seed may cause early endosperm cellularization. We applied auxin (NAA) to fveagl62−1/−4 seeds at 1 DPA to test if the NAA application could reduce premature endosperm cellularization. 11.9% of fveagl62−1/−4 seeds (n = 42) were rescued by the NAA treatment, showing a lack of premature endosperm cellularization and enlarged nucellus 2 days after NAA treatment (Fig. 2g; Supplementary movie 4). Rescue was not observed in mock-treated fveagl62 seeds (n = 56) (Fig. 2g; Supplementary movie 3). Therefore, reduced auxin in fveagl62 seeds likely contributes to the premature endosperm cellularization, and auxin may act downstream of FveAGL62, consistent with a positive role of FveAGL62 for auxin biosynthesis gene expression (Fig. 2a, c).

Characterization of atagl62 mutants in Arabidopsis

In Arabidopsis, despite a similar requirement of auxin and AtAGL62 for endosperm nuclear proliferation14,19, functional relationships between AtAGL62 and auxin synthesis are unknown, although AtAGL62 has been shown to be required for the expression of P-GLYCOPROTEIN 10 (PGP10) that may transport auxin from the endosperm to the seedcoat15. Based on the similar endosperm phenotype observed in Arabidopsis atagl62 seeds and strawberry fveagl62 seeds, we hypothesized that AtAGL62 may play a similar positive regulatory role for auxin biosynthesis as the strawberry FveAGL62. To test this hypothesis, we crossed an established Arabidopsis line containing the R2D2 auxin reporter into an Arabidopsis atagl62-2 mutant plant. The R2D2 reporter construct contains DII-VENUS and mDII-Tdtomato reporter genes. While the DII-VENUS is subjected to auxin-mediated protein degradation, mDII-Tdtomato is not and serves as a control27. As shown in Fig. 3a, the DII-VENUS fluorescence signal is absent in the WT endosperm indicating auxin accumulation there. In contrast, strong nuclear DII-VENUS fluorescence is observed in atagl62-2 endosperm nuclei (Fig. 3a) suggesting an absence of auxin there. Furthermore, the ratio of DII-VENUS fluorescence to mDII-Tdtomato fluorescence is significantly higher in the atagl62-2 endosperm nuclei than the wild type (Fig. 3a), indicating an absence or severely reduced auxin accumulation in the atagl62-2 endosperm.

Fig. 3: Arabidopsis atagl62 endosperms have reduced auxin when compared with WT.
figure 3

a Confocal image of auxin reporter R2D2 in WT (Col-0) and atagl62-2 seeds at 2 DPA. Absence of DII-VENUS nuclear signal in the WT endosperm contrasts with strong DII-VENUS nuclear signals in the atagl62-2 endosperm. The scatter graph Y-axis indicates the ratio between DII-VENUS and mDII-Tdtomato signal per endosperm nucleus. A second independent experiment gave a silimar result. The elements of boxplots and source data are provided in the Source Data file. Significant difference indicated by ** (P = 7.55e-12 by two-tailed Student’s t-test) is found between WT and atagl62. b Confocal image of ProAtYUC10::3xnGFP signal in WT and atagl62 endosperms. Note the strong nuclear GFP signals in the WT endosperm and almost undetectable GFP signals in the atagl62 endosperm. c Confocal image of WT and atagl62 mutant seeds at 2 DPA either mock-treated or auxin (2,4-D)-treated. Different extent of endosperm cellularization is observed. Arrows indicate cell walls between endosperm cells. d Quantification of mock- or 2,4-D-treated seeds derived of atagl62-2 (−/+) parents. Y-axis is the percentage of endosperms with one of the three phenotypes: no cellularization, partial cellularization, and complete cellularization. Significant difference (two-tailed Fisher’s exact test) is indicated by ** (P = 0.0006) for complete cellularization or * (P = 0.0416) for no cellularization between mock and 2,4-D treatments of each phenotypic category. e Confocal image of Arabidopsis seeds at 2 DPA. Precocious endoserpm cellularization is absent in transgenic atagl62 plants containing the strawberry FveAGL62 gene. Scale bar in ae: 50 μm.

The decreased auxin level in the atagl62-2 endosperm may be caused by reduced auxin biosynthesis gene expression. A ProAtYUC10::3xnGFP transgene28 was introduced into atagl62-2 plants by genetic crosses to establish an F2 plant homozygous for the transgene and heterozygous for the atagl62-2 mutation. In F3, 95.4% (n = 121) of the wild type seeds show strong GFP fluorescence in the endosperm nuclei (Fig. 3b); in contrast, 92.86% (n = 56) of atagl62-2 (−/−) seeds showed undetectable or extremely weak GFP fluorescence in the endosperm nuclei (Fig. 3b). Together, the data revealed a significant reduction of auxin accumulation as well as reduced biosynthesis gene expression in the atagl62 mutant endosperm, supporting a similar function of AtAGL62 to FveAGL62 in promoting auxin biosynthesis.

If the premature cellularization of atagl62-2 endosperm is caused by reduced auxin, exogenous application of auxin may be able to rescue this endosperm defect. Fertilized Arabidopsis atagl62-2 (/+) siliques at 1 DPA were treated with a synthetic auxin (2,4-dichlorophenoxyacetic acid; 2,4-D). We chose 2,4-D due to its efficacy in Arabidopsis seeds14. Progeny seeds at 2 DPA were examined under the confocal microscope. Different endosperm cellularization phenotypes including no cellularization, partial cellularization, and complete cellularization were quantified (Fig. 3c, d). We observed 22.7% (138 out of 608) mock-treated seeds exhibiting complete cellularization, a proportion that closely matches the 25% atagl62-2 (−/−) seeds expected from the atagl62-2 (/+) parents. A small percentage (4.61%) of mock-treated seeds also showed partial endosperm cellularization (Fig. 3c, d). Upon 2,4-D treatment, only 15% (95 of 633) seeds exhibited complete cellularization (Fig. 3d). If one focuses on the homozygous atagl62-2 seeds, the proportion that exhibits complete cellularization would be reduced from 91 to 60% (22.7 and 15% divided by 25% respectively) by the auxin treatment. Hence, 2,4-D is able to partially rescue the endosperm defect of atagl62-2 seeds.

Since strawberry fveagl62 and Arabidopsis atagl62 mutants both exhibit early endosperm cellularization and reduced auxin biosynthesis and accumulation, the strawberry FveAGL62 and Arabidopsis AtAGL62 appear to possess a conserved function. In support of this, the strawberry FveAGL62 ORF flanked by the AtAGL62 promoter and terminator was transformed into atagl62-2. Three independent transgenic lines were characterized in depth; the FveAGL62 transgene is able to rescue all defects of atagl62-2 including early endosperm cellularization and seedcoat development (Fig. 3e; Supplementary Fig. 7a–c).

Seedcoat development in fveagl62 mutants

In Arabidopsis, atagl62 mutant seedcoat fails to differentiate indicated by an absence of proanthocyanidins in the endothelium layer (Supplementary Fig. 7c)15. Hence, we investigated whether FveAGL62 is also required for seedcoat development in strawberry. Stage 1 ovule and stage 2-3 seeds from fveagl62−1/−4 strawberry mutants were dissected out of the ovaries and stained with vanillin that stains proanthocyanidins29. At stage 1, both wild type and fveagl62 ovules showed no vanillin staining except mild staining at the two poles (Fig. 4a). At stage 2 however, strong vanillin staining was present in both wild type and fveagl62 seedcoat. When the fveagl62 mutant seeds arrested development at stage 3, they still stained dark red similarly to the wild type (Fig. 4a). CLSM optical sections revealed similar seed coat cell layers in wild type and fveagl62 with three integument cell layers exterior to the endothelium layer (Fig. 4b). Hence, unlike Arabidopsis, the strawberry FveAGL62 is not required for seedcoat development. Interestingly, FveAGL62 ORF driven by the cis-elements of AtAGL62 readily rescued the seedcoat defect of Arabidopsis atagl62 mutants (Supplementary Fig. 7c), suggesting that the distinct requirement of AtAGL62 for Arabidopsis seedcoat development is not due to different AGL62 protein function but rather differences afforded by different reproductive structures or processes.

Fig. 4: Characterization of F. vesca fveagl62 mutant seedcoat.
figure 4

a Vanillin-stained F. vesca WT and fveagl62 mutant seeds at the pre-fertilization stage 1 and post-fertilization stages 2–3. The proanthocyanidins in the endothelium layer stains red with the vanillin. b Confocal laser scanning microscopy images of WT and fveagl62 seeds at stage 2. The enlarged box highlights the three integument cell layers (red arrows) exterior to the endothelium. Scale bars, 500 μm (a), 100 μm (b).

FveATHBs may mediate the effect of FveAGL62 on auxin synthesis

The above results established FveAGL62 a positive regulator of auxin biosynthesis in the endosperm. Therefore, we examined if this positive effect of FveAGL62 on auxin biosynthesis is direct or indirect. A yeast one-hybrid (Y1H) screen using the promoter of an auxin biosynthesis gene FveYUC10 failed to identify FveAGL62 or FveAGL80 as binding factors. We then used Y1H assay to directly test if FveAGL62 and FveAGL80 could bind and activate auxin biosynthetic genes FveYUC10 and FveTAR1. No activation of FveYUC10 or FveTAR1 was observed even when both FveAGL62 and FveAGL80 were simultaneously introduced into the yeast (Supplementary Fig. 8a). Therefore, FveAGL62/FveAGL80 likely promotes the expression of auxin biosynthesis genes indirectly by regulating other transcription factors.

We re-examined the RNA-seq data from stage 2 fveagl62 and WT seeds described above and noticed a subfamily of ATHB genes that were significantly up-regulated in fveagl62 (Fig. 5a). This ATHB subfamily of transcription factors is composed of both zinc finger and homeobox domains, and one subfamily member was previously found to bind the FveYUC10 promoter in a Y1H screen (Supplementary Fig. 8b). FveATHB29a, FveATHB29b, and FveATHB30 not only exhibited an increased expression in the fveagl62 seeds (Fig. 5a) but also showed a decreasing expression trend from stage 1 ovules, stage 2 seeds, to stage3 ghosts (endosperm and seedcoat) in the wild type (Fig. 5b). This expression trend is consistent with the hypothesis that fertilization-induced FveAGL62/FveAGL80 represses FveATHB gene expression to relieve the inhibition of auxin biosynthesis gene expression during early seed development.

Fig. 5: FveAGL62/FveAGL80 regulates a family of FveATHB genes in strawberry.
figure 5

a Hierarchical clustering heatmap showing RNA-seq reads (TPM) in WT and fveagl62 mutant seeds at stage 2 for a subclass of FveATHB genes. Three FveATHB genes (marked by *) are more highly expressed in the fveagl62 mutant seeds than the WT. b Bar graphs showing the expression level (TPM) of three FveATHB genes at the stage 1 ovules, stage 2 seeds, and stage 3 ghosts in WT based on a prior RNA-seq dataset5. c, d Y1H assays showing binding and activation of promoters of FveATHB29b and FveATHB30 by the combined action of FveAGL62 and FveAGL80. e, f Transient repression of the LUC reporter driven by the FveATHB29b (e) or FveATHB30 (f) promoter. Effector proteins are FveAGL62 or FveAGL80 alone or in combination. Significant difference from the negative control (GFP) is marked by ** (P < 0.01 by two-tailed Student’s t-test). Error bars indicate standard deviation. Three biologically independent experiments gave similar results. g Y1H assay testing the binding of FveATHBs to ProFveYUC10::AbAi in the SD-Leu-Ura medium with or without 500 ng/ml AbA. Empty vector containing the AD serves as a negative control. h Transient repression of ProFveYUC10::LUC in tobacco leaves. Y-axis indicates the relative expression level of LUC compared to REN. Significant difference from the negative control (GFP) is marked by * (P < 0.05 by two-tailed Student’s t-test). Error bars indicate standard deviation. Three biologically independent experiments gave similar results.

To test if the FveAGL62/FveAGL80 directly represses the expression of these FveATHBs, we conducted Y1H assay and transient luciferase (LUC) assays in tobacco leaves. In yeast, either FveAGL62 or FveAGL80 failed to activate the promoters of FveATHB29b and FveATHB30; when expressed together however, FveAGL62 and FveAGL80 were able to activate both FveATHB genes (Fig. 5c, d). In tobacco leaves, the LUC reporter driven by either FveATHB29b or FveATHB30 promoter showed reduced expression when either FveAGL80 alone or FveAGL62 and FveAGL80 together were introduced into the tobacco leaves (Fig. 5e, f), supporting a direct repressive role of FveAGL62/FveAGL80 on the expression of FveATHB29b and FveATHB30.

As shown earlier, many auxin biosynthesis genes FveYUC5, FveYUC10, FveYUC11, FveTAA1, and FveTAR1 exhibited gradual increase in expression from stage 1 ovule to stage 3 ghost (Supplementary Fig. 6a, c), an expression trend opposite that of FveATHB genes (Fig. 5b). This opposite expression trend is consistent with the hypothesis that FveATHBs may repress the transcription of auxin biosynthesis gene pre-fertilization, and that this repression is gradually lifted from stage 1 to 3 through the action of FveAGL62/FveAGL80 upon fertilization. To test this hypothesis, Y1H and transient luciferase expression were used to test direct binding of FveATHBs to the FveYUC10 promoter. In Y1H, FveATHB29b and FveATHB30 each was able to activate ProFveYUC10::AbAi to allow yeast colony formation on media containing AbA500 (Fig. 5g). In the tobacco transient expression assay, FveATHB30 but not FveATHB29b repressed the expression of ProFveYUC10::LUC (Fig. 5h). Since FveATHB29b and FveATHB30 can heterodimerize (Supplementery Fig. 9) and the ATHB subfamily members were reported to function through homo- and heterodimerization30, the FveATHB29b/FveATHB30 heterodimers, as well as other in vivo heterodimer combinations, may act to directly repress FveYUC10 expression.

Characterization of FveATHB overexpression transgenic lines

To test the function of FveATHB29b and FveATHB30 in strawberry, a reasonable approach is to over-express (OE) FveATHB genes in F. vesca, which overcomes functional redundancy among FveATHB family members. Three independent transgenic lines of FveATHB29b-OE and two independent lines of FveATHB30-OE, all driven by the UBQ10 promoter, were generated and shown to increase FveATHB29b and FveATHB30 expression by 10–20 fold and 3-7 fold, respectively (Supplementary Fig. 10). All FveATHB29b-OE and FveATHB30-OE transgenic lines produced smaller fruits compared to the wild type (Fig. 6a–c). To determine if the small fruit phenotype of FveATHB-OE lines is due to reduced auxin, we applied NAA to the fertilized fruit and observed restoration of fruit size to wild type size (Fig. 6b). To test if the FveATHB-OE shows reduced expression of auxin biosynthesis genes, which may underlie the small fruit size, transcript levels of FveYUC10 and FveTAA1 in stage 2 seeds of FveATHB-OE lines were quantified by RT-qPCR. FveYUC10 and FveTAA1 transcript levels were reduced by 2–10 fold in the seeds of FveATHB-OE lines (Fig. 6d, e). Therefore, the small fruits of FveATHB-OE lines are likely caused by a reduction of auxin biosynthesis gene expression.

Fig. 6: Characterization of FveATHB-OE transgenic lines.
figure 6

a Plant and fruit phenotype of FveATHB-OE transgenic lines. Fruits of three FveATHB29b-OE lines and two FveATHB30-OE lines are shown. b Mock- and NAA-treated fruits of FveATHB29b-OE and FveATHB30-OE lines. c Fruit size measurement of WT, three FveATHB29b-OE lines, and two FveATHB30-OE lines. Significant difference from WT is marked by ** (n = 20, P < 0.01 by two-tailed Student’s t-test). The elements of boxplots and source data are provided in a Source Data file. d, e Relative transcript level (Y-axis) of auxin biosynthetic genes FveYUC10 and FveTAA1 in stage 2 seeds of WT and FveATHB29b-OE (d) and FveATHB30-OE (e) lines. Significant difference (two-tailed Student’s t-test) from WT is marked with ** (P < 0.01) and * (P < 0.05). Error bars indicate standard deviation. Three biologically independent experiments gave similar results. f Confocal laser scanning microscopy reveals extensive cell death (red arrowheads) and precocious cellularization (white arrows in enlarged inserts) in the endosperm of stage 2 seeds in FveATHB-OE lines. Scale bars, 1 cm (a, b), 100 μm (f).

If the reduced fruit size is due to reduced auxin biosynthesis in FveATHB-OE seeds, some of the seeds should exhibit a phenotype similar to fveagl62. Confocal microscopy showed that the majority of FveATHB-OE seeds remained small and contained a tube-like structure filled with dead cells (red arrowheads in Fig. 6f; Supplementary Fig. 11). A small percentage (8.5%) of the endosperms showed precocious cellularization (Supplementary Fig. 11). Sometimes, both cell death and premature endosperm cellularization were present in the same endosperm of FveATHB-OE seeds (see enlarged inserts in Fig. 6f). Since the expression of FveATHBs via the UBQ10 promoter occurs in maternal sporophytic as well as gametophytic tissues before and after fertilization, ectopic FveATHB expression is likely causing reduced auxin earlier and in more tissues than what is observed in fveagl62, which may lead to endosperm cell death. We showed that application with NAA could partially rescue the endosperm phenotypes in cell death and premature cellularization (Supplementary Fig. 11). Together, our data revealed a previously unknown regulatory module consisting of AGL62/AGL80 and ATHBs to regulate auxin biosynthesis in the endosperm of fertilized seeds.

Discussion

Auxin from the seeds has been known to induce strawberry fruit development for more than 70 years3. Recent studies in diploid wild strawberry F. vesca showed that the fertilization-induced auxin biosynthesis mainly occured in the endosperm5,12. However, the molecular mechanism underlying this fertilization-induced auxin biosynthesis in the endosperm is unknown. In this study, we show that the type I MADS-box gene FveAGL62 is required for auxin biosynthesis in the endosperm and FveAGL62 expression is induced by fertilization specifically in the endosperm, thus establishing FveAGL62 as a key gene that mediates fertilization-induced auxin production for strawberry seed and fruit development.

In addition, we show that AGL62 plays a conserved function in promoting auxin biosynthesis in strawberry as well as Arabidopsis. Both strawberry FveAGL62 and Arabidopsis AtAGL62 exhibit similar expression patterns: their transcripts increase immediately after fertilization in the sexual endosperm and then decline abruptly just before cellularization (Fig. 1a, Supplementary Fig. 1d)19. In addition, loss of AGL62 in strawberry or Arabidopsis causes a similar precocious endosperm cellularization phenotype, and the strawberry FveAGL62 gene rescues the Arabidopsis atagl62 mutant phenotypes. Both fveagl62 and atagl62 mutants exhibit significantly reduced expression of auxin biosynthesis genes and reduced auxin reporter signal in the endosperm, and application of exogenous auxin partially rescues each mutant’s endosperm cellularization defect. These data strongly support AGL62 as a conserved positive regulator of auxin biosynthesis in the sexual endosperm of multiple plant species.

Our finding is supported by previous genetic analysis of Arabidopsis auxin biosynthesis mutants and atagl62 mutants. Triple loss-of-function mutants in auxin biosynthesis genes had severe endosperm proliferation defects, and exogenous auxin application or central cell-specific activation of auxin biosynthesis genes induced autonomous endosperm development without fertilization14. In fis2 mutants with a defective FIS-PRC2 (Polycomb Repressive Complex 2), auxin is ectopically produced in the central cell, which correlates with central cell replication and autonomous endosperm formation without fertilization14. Interestingly, this autonomous endosperm development is suppressed by the atagl62 mutation14 and AtAGL62 was shown to be a direct target of FIS2-PRC221. These genetic analyses suggest that FIS-PRC2 may normally repress auxin and AtAGL62 to prevent endosperm development before fertilization. Nevertheless, despite their similar role in endosperm development, the relationship between AtAGL62 and auxin in endosperm development has not been resolved until now.

Based on the findings of this study, we propose a model shown in Fig. 7. Upon fertilization of the central cell and through an as yet undetermined mechanism, the expression of FveAGL62 and FveAGL80 is induced in the endosperm. FveAGL62/FveAGL80 heterodimers may directly bind and inhibit the transcription of FveATHB29b, FveATHB30, as well as other target genes. The decreased level of FveATHBs leads to de-repression of FveYUC10 and other auxin biosynthesis genes, enabling the synthesis and accumulation of active auxin, which is subsequently exported to the receptacle to initiate fruit development.

Fig. 7: A model illustrating the FveAGL62-FveATHB module.
figure 7

Upon fertilization, FveAGL62 and FveAGL80 expression is induced in the endosperm and the FveAGL62/FveAGL80 heterodimers act to switch off the transcription of FveATHB genes, lifting the transcriptional repression of auxin biosynthesis genes such as FveYUC10, leading to the synthesis and accumulation of auxin. Upon being transported to receptacle, auxin stimulates receptacle fruit development.

Other than auxin biosynthesis, AGL62/AGL80 almost certainly regulates other downstream processes including auxin transport and GA biosynthesis as shown by significantly reduced expression of GA biosynthesis genes (Fig. 2b). In both strawberry and Arabidopsis, agl62 mutant endosperm still undergoes a few rounds of nuclear division before premature cellularization. These initial nuclear divisions could be stimulated by a low dose of auxin brought in by the sperm cell or by an initial auxin synthesis triggered by the action of fertilization. Therefore, AGL62 may very well act in a positive-feedback loop to amplify auxin biosynthesis in the endosperm after an initial pulse of auxin that arises independently.

While our work focuses on fertilization-induced auxin synthesis in the endosperm, several prior studies investigated fertilization-induced seedcoat development in Arabidopsis, which revealed that the Arabidopsis seedcoat development relied on the AtAGL62-dependent transport of auxin produced from the endosperm15,20, and that atagl62 mutants not only failed to develop a seedcoat but also trap and elevate auxin in the endosperm15. While we similarly observed a lack of seedcoat development in the Arabidopsis atagl62 mutants (Supplementary Fig. 7c), we observed a significantly reduced auxin in the atagl62 mutant endosperm (Fig. 3a, b). We do not know reason behind this difference but note that the studies used different auxin reporters. The prior study used DR5::VENUS15, whereas our study utilizes R2D2 and ProAtYUC10::3xnGFP27,28.

Another distinction is the apparently normal seedcoat formation in strawberry fveagl62 mutants, suggesting that strawberry seedcoat development is independent of FveAGL62 (Fig. 4). We did not find an ABCB transporter among the down-regulated genes in fveagl62 stage 2 seeds as was observed in atagl62. However, an auxin efflux carrier, FvePIN2 (FvH4_4g06850), is down-regulated by tenfold in the strawberry fveagl62 seeds. As part of the essential role of FveAGL62 for fruit development, FvePIN2 could be activated by FveAGL62 to transport auxin to the receptacle fruit of strawberry. A possible explanation for the different seedcoat phenotype between atagl62 and fveagl62 is that the pulses of auxin that initiate seedcoat development could originate from different tissues in strawberry. For instance, the ovary wall of F. vesca was found to express auxin biosynthesis genes upon fertilization5 and could be the tissue that provides the initial auxin for initiating seedcoat development.

In summary, fertilization triggers seed and fruit development through the induction of AGL62 that leads to auxin synthesis in the endosperm. However, pollination/fertilization-induced auxin synthesis is not exclusive to the endosperm and may occur in other sporophytic tissues as is supported by reports of independent auxin synthesis in the seed integuments31,32. Subsequently, newly synthesized auxin not only initiates post-fertilization development in situ but also is transported to other seed and floral tissues to coordinate and sustain the post-fertilization development. Comparative functional analysis of AGL62 in other plant species will reveal both conserved and species-specific mechanisms for the induction of post-fertilization developmental programs.

Methods

Strains and growth conditions

The Fragaria vesca strain Yellow Wonder 5AF7 (YW5AF7)33 was used as wild type. CRISPR-knockout, overexpression, and GUS reporter studies are all conducted in YW5AF7. Strawberry plants were grown in a growth chamber with a 16-hour light at 25 °C followed by an 8-hour darkness at 22 °C with a relative humidity of 50%.

Arabidopsis WT (Col-0), a T-DNA insertion line in the Col background, SALK_022148 (atagl62-2)19, the R2D2 reporter (stock number: CS2105637) in the Col background27 and the ProAtYUC10::3xnGFP reporter (stock number: CS69899) in the Col background28 were all obtained from the Arabidopsis Biological Resource Center (ABRC) (https://abrc.osu.edu/). Seeds of DR5ver2::GUS in F. vesca (YW5AF7)12 were obtained from Dr. Chunying Kang.

Strawberry Gene IDs

All the genomic and CDS sequences of F. vesca genes mentioned in the manuscript can be found in GDR (rosaceae.org) using their gene IDs (genome version 2.0 annotation version 2.0; genome version 4.0 annotation version 2.0) as follows: FveAGL62 (gene01789; FvH4_2g03030), FveAGL80 (gene22916; FvH4_6g08460), FveAGL80L1 (gene18029; FvH4_6g21170), FveAGL80L2 (gene04949, FvH4_7g09730), FveAGL80L3 (gene22967; FvH4_6g08570), FveAGL80L4 (gene23924; FvH4_6g43410), FveATHB29b (gene09326; FvH4_5g17830), FveATHB30 (gene03815; FvH4_6g48610), FveATHB22 (gene01920; FvH4_1g20040), FveYUC5 (gene32686; FvH4_2g14550), FveYUC10 (gene27796; FvH4_2g24750), FveYUC11 (gene06886; FvH4_4g17980), FveTAR1 (gene31791/gene37056; FvH4_5g05900), FveTAA1 (gene03586; FvH4_4g25850), FveGA20OX1c (gene19436; FvH4_7g12610), FveGA20OX1d (gene13360; FvH4_7g28670), FveGA3OX1b (gene01060; FvH4_2g30010), FveGA3OX1c (gene01059; FvH4_2g30020), FvePIN2 (gene12312; FvH4_4g06850), FvePP2a (gene03773; FvH4_4g27700).

CRISPR and GUS constructs and transgenic strawberry generation

To generate the F. vesca agl62 mutants, gRNA1 (GGGTGCGCAAGGCGAGCCAGGGG) or gRNA3 (CTATCACAGACTCCACGCAGGGG) targeting the coding region of FveAGL62 were inserted into the JH4 entry vector and then incorporated into the binary vector JH19 via gateway cloning34. To generate the fveagl80 mutant, one gRNA (CGATGGACTCCCAACGGGGAAGG) targeting the coding region of FveAGL80 was inserted into the JH4 entry vector and then incorporated into the binary vector JH19. To confirm editing, genomic sequences spanning the target site of respective genes were amplified by PCR and sequenced. All primers used in this study are listed in Supplementary Table 1.

For ProFveAGL62::FveAGL62-GUS and ProFveAGL80::FveAGL80-GUS, ProFveAGL62::FveAGL62 (1512-bp upstream and the 660-bp coding region of FveAGL62) and ProFveAGL80::FveAGL80 (2461-bp upstream and 834-bp coding region of FveAGL80) were PCR amplified and cloned into pMDC162 binary vector35, in which the GUS reporter is fused at the C-terminus of the gene. The constructs were transformed into agrobacterium strain GV3101.

For F. vesca transformation, calli are induced from cut cotyledons of F. vesca (YW5AF7) on the 5++ medium (1 x MS, 2% sucrose, 3.4 mg/L benzyl adenine, 0.3 mg/L indole-3-butyric acid (IBA), 0.7% phytoagar, pH 5.8) for two-to-three weeks in the dark. The calli are then incubated with the agrobacterium GV3101 harboring the construct in the dark for 1 h in the co-cultivation buffer (1 x MS, 2% sucrose, 0.4 mg/ml acetosyringone) with gentle shaking. The infected calli were kept on the 5 ++  medium for three days in the dark, then washed with sterile water and put onto the MS medium with antibiotics (250 mg/L timentin + 250 mg/L carbenicillin), and grown in the dark for 2 weeks. Transformed calli were selected on the 5++ medium with 250 mg/L timentin, 250 mg/L carbenicillin and 4 mg/L hygromycin. The calli were moved to fresh medium (5++ medium with the same three antibiotics) every 2-to-3 week until shoots appear. When the shoots grow larger, they are transferred to the rooting medium (0.5 x MS, 0.01 mg/L IBA, 2% glucose, 0.7% phytoagar, pH 5.8) with 4 mg/L hygromycin. After about 1–2 months, the plants with roots are transferred to soil and genotyped5,36. Nine and 12 independent transgenic lines of ProFveAGL62::FveAGL62-GUS and ProFveAGL80::FveAGL80-GUS, respectively, were obtained and analyzed.

For overexpression, full-length CDS of FveATHB29b and FveATHB30 were PCR amplified from cDNA and assembled into vector JH23 at the digestion sites of KpnI and PacI through Gibson cloning37,38. Thus, FveATHB29b and FveATHB30 were driven by the Arabidopsis Ubiquitin 10 promoter. After transformation into F. vesca, 10 and 12 independent lines of FveATHB29b-OE and FveATHB30-OE respectively were confirmed by PCR genotyping.

Analyses of Arabidopsis atagl62 mutants

The full-length FveAGL62 ORF (663 bp) was PCR amplified and assembled with the upstream and downstream fragments of AtAGL62. The 2074 bp upstream fragment of AtAGL62 contains the promoter and 5’UTR, while the 1054 bp downstream fragment contains the 3’UTR of AtAGL62. This ProAtAGL62::FveAGL62:terAtAGL62 chimeric gene was cloned into pMDC9935 and transformed via floral dip into the Arabidopsis atagl62-2 (−/+) heterozygous plant. After germination on ½ MS medium with 30 µg/ml hygromycin, the hygromycin-resistant seedlings were further genotyped using Salk022148/atagl62-2 T-DNA genotyping primers (Supplementary Table 1). 12 independent agl62-2/ homozygote plants containing the ProAtAGL62::FveAGL62:terAtAGL62 transgene were obtained and analyzed.

To analyze the auxin reporters (R2D2 or ProAtYUC10::3xnGFP) in atagl62-2 seeds, atagl62-2 (−/+) plants were crossed with plants harboring R2D2 or ProAtYUC10::3xnGFP. The F1 plants containing both the atagl62-2 mutation and the reporter gene were confirmed by PCR and GFP fluorescence. F2 plants homozygous for the reporter gene and heterozygous for the atagl62-2 mutation were obtained by genotyping and segregation. Homozygous ProAtYUC10::3xnGFP is necessary to avoid impacts of parental imprinting. Fertilized seeds from F2 plants were analyzed under a Leica SP5X Confocal microscope (Leica Co. USA).

Seed and fruit characterization

To dissect and image seeds, ovules/seeds at stages 1 (opening flower), 2 (2–4 DPA), and 3 (6–7 DPA) were dissected out of achenes using tweezers and then imaged with a Zeiss Stemi SV 6 microscope equipped with an AxioCam digital camera.

For hormone treatment of strawberry, fveagl62 receptacles were treated with 500 μM NAA, 500 μM GA, or a combination of NAA and GA. A drop of Tween-20 was added to 10 ml solution. 10 ml water with a drop of Tween-20 was used as the mock control. These treatments were repeated every two days until day 10 when the images were taken. FveATHB-OE receptacles were treated with 500 μM NAA or mock control. These treatments were repeated every two days until day 30 when the images were taken.

To treat Arabidopsis seeds with 2,4-D, atagl62-2 (−/+) siliques at 1DPA were submerged in 200 μM 2,4-D solution for 30 sec15. For mock treatment, 0.05% Tween-20 solution was used instead. The seeds were collected the next day and processed for confocal imaging.

For confocal imaging, stage 2 seeds of F. vesca were dissected out from the achenes prior to fixation. The 2 DPA Arabidopsis seeds were dissected out of the siliques Seeds were fixed in 4% glutaraldehyde (in 12.5 mM cacodylate, pH 6.9), and kept under a vacuum for 2 h. After fixation, the seeds were dehydrated through an ethanol series [30, 50, 70, 90, 100% (v/v)] at 15 min per step. The dehydrated seeds were clarified in 2: 1 (v/v) benzyl benzoate: benzyl alcohol for at least 1 h before imaging39. A Leica SP5X Confocal microscope (Leica Co. USA) with 488 nm excitation was used to image autofluorescence of the seeds. For observing fluorescent reporters in Arabidopsis seeds, eGFP of ProAtYUC10::3xnGFP was excited at 488 nm and detected at 498-530 nm. For observing the R2D2 fluorescent reporter in Arabidopsis seeds, a Leica STELLARIS DLS Confocal microscope (Leica Co. USA) was used. DII-VENUS was excited at 515 nm and detected at 520–559 nm and mDII-Tdtomato was excited at 554 nm and detected at 562–726 nm. The pixels of VENUS and Tdtomato fluorescence in endosperm nuclei were measured by the LAS X software (Leica Co. USA) and used to calculate the DII/mDII ratio.

For seedcoat vanillin staining, seeds of Arabidopsis or strawberry were dissected and mounted in vanillin staining solution, 1% (w/v) vanillin (4-hydroxy-3-methoxybenzaldehyde) in 6 N HCl, for 30 min and then imaged with a Zeiss Stemi SV 6 microscope equipped with an AxioCam digital camera.

RNA-seq analysis

Strawberry stage 2 seeds were dissected out of achenes. About 50 seeds were collected for each biological replicate and three replicates were prepared for each genotype. Total RNA was extracted using the Monarch Total RNA Miniprep Kit (NEB) following the manufacturer’s instructions. A total of 0.5 to 2 μg RNA per sample was sent to the Novogene Corporation Inc for library preparation and sequencing with Illumina Platform. About 43 to 60 million PE150 raw reads per sample were obtained.

For read quantification, Salmon (v0.11.2)40 was applied to estimate the transcript abundance. Salmon index was generated using Fragaria_vesca_v4.0.a2.transcripts.fa as the reference transcriptome with parameters -k 31 and --keepDuplicates. The reference transcriptome was retrieved from Genome Database for Rosaceae (GDR, https://www.rosaceae.org/species/fragaria_vesca/genome_v4.0.a2)26,41. The paired-end RNA-Seq reads were quantified in Salmon mapping-based mode. The --seqBias flag was used to turn on sequence-specific bias correction. Eventually, the isoform expression was summarized into gene level by Tximport (v1.10.1)42.

For differential expression analyses, DESeq2 (v1.22.2)43 was utilized to identify differentially expressed genes (\({padj} \, < \, 0.05\) and \(\left|{{\log }}_{2}{FoldChange}\right| > 1\)) between WT and fveagl62. The longest strawberry protein variants from Fragaria vesca annotation v4.0.a2 were searched against Arabidopsis protein database via OmicsBox (v1.2.4) built-in BLAST44. The BLAST database was constructed using the longest protein sequences derived from TAIR10_pep_20101214 (https://www.arabidopsis.org/download/index-auto.jsp?dir=%2Fdownload_files%2FProteins%2FTAIR10_protein_lists) with E-value cutoff set to 0.001.

For Gene Ontology (GO) enrichment analysis, GO terms were assigned to the longest strawberry proteins (v4.0.a2) by OmicsBox (v1.2.4) based on BLAST and InterProScan results45,46. The protein sequences were searched against Swiss-Prot database (uniport_sprot.fasta, https://www.uniprot.org/downloads)47 locally using OmicsBox built-in BLAST. InterProScan was executed within OmicsBox to identify the protein domains with default settings. TopGO (v2.34.0)48 was employed to conduct GO enrichment analysis. The GO terms with P value ≤ 0.05 in Fisher’s exact test were considered to be significant.

For gene expression visualization, the heatmaps were made using the website Morpheus (https://software.broadinstitute.org/morpheus/) with the Hierarchical Clustering option.

Phylogenetic tree building

Full Pfam alignment files PF00319 and PF01486 (https://pfam.xfam.org/) were downloaded and analyzed by HMMER (version 3.3) (http://hmmer.org/)49 to identify the strawberry and Arabidopsis proteins with the MADS-box domain or K-box domain. The proteins that only have a MADS-box domain were considered as type I MADS-box proteins. Protein sequences of strawberry type I MADS-box genes were downloaded from GDR (www.rosaceae.org/) and those of Arabidopsis genes were downloaded from TAIR (https://www.arabidopsis.org/). The sequence alignment was made using Clustal Omega (http://www.ebi.ac.uk/Tools/msa/clustalo)50. The unrooted phylogenetic tree was constructed using the online tool PhyML 3.0 (http://www.atgc-montpellier.fr/phyml/) with the neighbor-joining statistical method.

RNA extraction, cDNA synthesis, and RT-qPCR

Total RNA was extracted from stage 1 (unfertilized ovules), stage 2 (fertilized seeds), and stage 3 (fertilized ghost) following the Cetyl Trimethyl Ammonium Bromide (CTAB) RNA extraction method51 and reverse transcribed to cDNA using SuperScript IV VILO Master Mix (Invitrogen, USA). RT-qPCR was performed using BioRad CFX96 Real-time system and SYBR Green PCR MasterMix (Applied Biosystems). The relative expression level was analyzed using a modified 2−ΔΔCT method52. For all RT-qPCRs, FvePP2a (FvH4_4g27700) was used as the internal control36. The RT-qPCR experiments reported in Figs. 2, 5, 6 and Supplementary Figs. 6, 11 are from one representative biological experiment (with three technical repeats). Biological replicates (2 to 3 times) gave similar results.

Testing protein-protein and protein-promoter interactions

For BiFC assays, full-length CDS of genes of interests were PCR amplified and ligated into either PXY105 (cYFP) at BamHI and XbaI sites or PXY106 (nYFP)53 at BamHI and XbaI sites through Gibson cloning37. BiFC constructs were transformed into Agrobacterium GV3101. Agrobacteria were harvested when OD600 reached 1.5, and resuspended in a solution (10 mM MgCl2, 10 mM MES (2-(N morpholine) ethanesulfonic acid) pH 5.6, and 200 μM acetosyringone) to a final OD600 of 0.854. The agrobacterial mixture was infiltrated into young Nicotiana benthamiana leaves. After 48 h, the YFP fluorescence signal was visualized and photographed using a Leica SP5X confocal microscope (Leica Co. USA).

For Y2H assays, full-length CDS of genes of interest were PCR amplified and ligated into pGADT7 (Clontech Inc.) at the EcoRI/BamHI digestion sites or pGBKT7 (Clontech Inc.) at the EcoRI/BamHI digestion sites by Gibson assembly37. Yeast strain PJ69-4A was transformed using the LiAc/PEG method55.

For Y1H assays, the Y1H constructs and procedures were performed based on the Matchmaker Gold Yeast One-Hybrid Library Screening System User Manual (PT4087-1). Full-length CDS of genes of interest were PCR amplified and ligated into pGADT7 (Clontech Inc.) at the EcoRI/BamHI digestion sites by Gibson assembly37. The target promoter sequence was amplified and ligated into pAbAi (Clontech Inc.) at the KpnI/SalI digestion sites by Gibson assembly37. Yeast strain Y1HGold was transformed using the LiAc/PEG method55.

For transient luciferase assays, the promoters of FveYUC10, FveATHB29b, and FveATHB30 were PCR amplified from genomic DNA, first Gibson assembled into a modified gateway entry vector LEI238, and then into the destination vector pLAH-LARm vector56 by LR recombination. The full-length CDS of FveAGL62 and FveAGL80 were PCR amplified from genomic DNA and assembled into the vector JH23 at the digestion sites of KpnI and PacI through Gibson cloning37,38. The constructs were transformed into agrobacterium GV3101. The agrobacterial solution was then infiltrated into young Nicotiana benthamiana leaves. At 48 h after agroinfiltration, the tobacco leaf tissue was collected and treated using the Dual-Luciferase Reporter Assay System (Promega). The luminescence activities of the firefly luciferase (LUC) and the Renilla luciferase (REN) were measured with a TD-20/20 Luminometer (Turner Designs). The Renilla luciferase gene (35S:REN) was used as a control to normalize the LUC readout. The assay was performed as previously described57. Each experiment was repeated two to three times with one representative result shown in Fig. 5e, f, h. All primers for the above assays are listed in Supplementary Table 1.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.