Four class A-AUXIN RESPONSE FACTORs promote tomato fruit growth despite suppressing its initiation

Abstract

In flowering plants, auxin produced in seeds after fertilization promotes fruit initiation. Application of auxin to unpollinated ovaries can also induce parthenocarpy (seedless fruit production). Previous studies showed that auxin signaling components SlIAA9 and SlARF7 (a class A-ARF) are key repressors of fruit initiation in tomato (Solanum lycopersicum). Similar repressive role of class A-ARFs in fruit set was also observed in other plant species. The paradox is the lack of evidence for any class A-ARF in promoting fruit development as predicted in the current auxin signaling model. Here, we generated higher order tomato mutants of four class A-SlARFs (SlARF5, SlARF7, SlARF8A and SlARF8B), and uncovered their precise combinatorial roles leading to suppressing and promoting fruit development. All four class A-SlARFs together with SlIAA9 inhibited fruit initiation, while remarkably promoted subsequent fruit growth. Transgenic tomato lines expressing truncated SlARF8A/8B lacking the IAA9-interacting PB1 domain displayed strong parthenocarpy, further confirming the promoting role of SlARF8A/8B in fruit growth. Altering doses of these four SlARFs led to biphasic fruit growth responses, showing their versatile dual roles as both negative and positive regulators. RNA-seq and ChIP-qPCR analyses further identified SlARF8A/8B target genes, including those encoding MADS-BOX transcription factors (AG1, MADS2 and AGL6) that are key repressors of fruit set. These results support that SlIAA9/SlARFs directly regulate transcription of these MADS-BOX genes to inhibit fruit set. Our study revealed the previously unknown dual function of four class A-SlARFs in tomato fruit development, and illuminated the complex combinatorial effects of multiple ARFs in controlling auxin-mediated fruit set and fruit growth.

Introduction

Fruit development is essential for the reproduction of flowering plants^1,2. Agronomically, fruits are a crucial food source. The transition from ovary to fruit, also called fruit initiation or fruit set, is arguably the most critical step in fruit development because the ovary is programmed to abort unless ovules are fertilized. The strict control that fruit initiation only occurs following successful fertilization provides the advantage of preventing resource wasted on producing seedless fruit (i.e., parthenocarpy), especially under adverse conditions that cause impaired anther/pollen development. On the other hand, parthenocarpy is a desirable trait for fruit crops because it is preferred by consumers and it ensures consistent fruit yield in variable environmental conditions. Studies on the model plant tomato (Solanum lycopersicum) and other species have shown that plant hormones play pivotal roles in fruit development, regulating every step from initiation to ripening^3,4. Auxin is one of the major hormones triggering fruit initiation. After pollination, auxin levels increase in developing seeds, and that is essential to trigger the transition from ovary to fruit development⁵. Auxin also promotes cell division and expansion during fruit growth^3,6. In addition, application of auxin to unfertilized ovary or genetic mutations that cause elevated auxin signaling can both induce seedless fruit formation^3,7. Thus, understanding the mechanism of auxin-induced parthenocarpy will have greater implication in developing climate-resilient crops.

The auxin signaling pathway is well conserved among land plants^8,9,10. The key components of early auxin signaling consist of three families of proteins: auxin coreceptors TIR1/AFB, AUX/IAA (IAA) transcription repressors and AUXIN RESPONSE FACTOR (ARF) transcription factors^11,12,13,14. When the auxin levels are low, IAA interacts with ARF at the target promoters and recruits co-repressor TOPLESS (TPL) to repress transcription by preventing ARF-Mediator complex formation¹⁵. When auxin levels increase, it binds to both TIR1/AFB and IAA to trigger IAA protein degradation, which releases ARFs to activate the auxin signaling pathway. This canonical AFB/IAA/ARF cascade is based mainly on class A-ARFs (also called activator ARFs)¹⁶. Recent studies have identified several auxin signaling components, all of which inhibit fruit initiation. Among the 25 members of SlIAAs¹⁷, SlIAA9 is the major repressor for fruit initiation because the tomato SlIAA9 null mutant entire^18,19 and antisense/CRISPR lines^20,21 all showed strong parthenocarpy. As for class A-ARFs, silencing of SlARF5 by an artificial miRNA or SlARF7 by RNAi led to strong parthenocarpy^22,23, although expression of several class A-ARFs, including SlARF5, SlARF7 and SlARF8B was reduced in the SlARF7 RNAi line¹⁹. Moreover, slarf5 and slarf7 single mutants do not display parthenocarpy¹⁹, suggesting that SlARF8A/8B may also inhibit tomato fruit initiation. In Arabidopsis, null alleles of AtARF8 showed parthenocarpy phenotype, indicating AtARF8 represses fruit set²⁴. The SmARF8 in eggplant (Solanum melongena) also inhibits fruit initiation as SmARF8 RNAi line presents strong parthenocarpy²⁵. In contrast, strawberry (Fragaria vesca) fvarf8 mutants produce larger fruits upon fertilization, but do not display parthenocarpy without pollination²⁶, suggesting FvARF8 acts as a major repressor for fruit growth, but not for fruit set. The different role of FvARF8 vs. other ARF8 orthologs may be due to variations in auxin-mediated fruit development in species with different fruit biology: receptacle-derived fruit in strawberry vs. ovary-derived fruit in tomato, Arabidopsis and eggplant. Nevertheless, these studies on the class A-ARFs all point to their inhibitory role in fruit initiation/growth. These findings are puzzling because auxin is known to play a key role in promoting fruit initiation^3,27 and because class A-ARFs are in general considered to be essential for activating auxin signaling¹¹. Therefore, a key question is which class A-ARF(s) are activators in mediating auxin-induced fruit initiation and/or in subsequent fruit development.

Here, we generated and characterized higher order mutant combinations of four class A-SlARFs (SlARF5, SlARF7, SlARF8A, SlARF8B) that are expressed at higher levels in tomato ovaries around anthesis. Surprisingly, we found that all four class A-SlARFs function as inhibitors (together with SlIAA9) in fruit initiation, but as activators in subsequent fruit growth. The parthenocarpic fruit sizes followed a biphasic bell-shaped curve in response to varying arf mutant combinations, revealing the fine-tuning capacity of fruit growth achieved by these four A-ARFs. The lack of placenta growth in slarf8a slarf8b double mutant further indicated that SlARF8A and SlARF8B are essential for placenta growth. Moreover, the four ARF proteins showed differential spatial localization in the tomato ovary, which is consistent with their mutant phenotypes. RNA-seq and ChIP-qPCR analyses identified three SlARF8A/8B target genes encoding MADS-BOX transcription factors that are key repressors of fruit set, suggesting IAA9/ARFs directly regulates transcription of these MADS-BOX genes. Together, our work demonstrated the four class A-ARFs function in a tissue-specific manner to modulate tomato fruit development by repressing fruit initiation and activating fruit growth.

Results

arf8a arf8b produced seedless fruits with reduced placenta

SlARF8A (Solyc03g031970) and SlARF8B (Solyc02g037530) in tomato are the closest orthologs to Arabidopsis AtARF8²⁸, and their encoded protein sequences share 81% identity. To examine the role of SlARF8A and SlARF8B in fruit development, we generated slarf8a, slarf8b knockout mutants in the Moneymaker (MM) cultivar using the CRISPR/Cas9 technology. In a single tomato transformation experiment, each ARF8 gene was targeted by a unique guide RNA (gRNA) at its coding sequence within the DNA-binding domain (Fig. 1a). Several frame-shift null slarf8a, slarf8b alleles were identified in T0 lines by DNA sequence analysis (Fig. 1b–1c), and each T0 line contains two different mutant alleles for SlARF8A and SlARF8B. To obtain the slarf8a and slarf8b single and double mutants without the Cas-9/gRNAs transgene, a T0 line that contained arf8a-1/8a-2 arf8b-1/8b-2 alleles was backcrossed to WT (wild type). We identified two sets of single and double arf8a arf8b homozygous mutants that contain two independent 8a and 8b alleles in the F2 generation: (1) arf8a-1, arf8b-1 and arf8a-1 arf8b-1; (2) arf8a-2, arf8b-2 and arf8a-2 arf8b-2. As shown below, these two sets of mutants showed consistent phenotypes. Therefore, arf8a-1(8a-1), arf8b-1(8b-1) and 8a-1 8b-1(8a 8b) lines were used for further studies.

Figure 1 |. arf8a arf8b crispr mutants displayed strong parthenocarpy.

a, Diagram of the SlARF8A/8B protein structure. DBD, DNA binding domain. MR, middle region. PB1, Phox and Bem1 domain. Arrow indicates the target position of CRISPR gRNAs. b-c, List of arf8a and arf8b crispr null alleles, respectively. Black numbers: distances from the ATG start codon of SlARF8 cDNA sequences. Letters in red: gRNA 20-mer sequences targeting SlARF8A or SlARF8B. PAM: protospacer adjacent motif. Blue numbers/letters: # of nucleotide deletions/insertions in each crispr allele, and the nucleotide insertion type. Dashes in blue or red: deletions or no insertion. d, Schematic of mature WT tomato fruit derived from self-pollinated flower. e-h, Parthenocarpic growth in arf8a-1, arf8b-1 single and double mutants. In e, pictures were taken five weeks after emasculation. Bar = 2 cm. In f, parthenocarpy frequency was calculated as % of parthenocparpic fruits developed from emasculated flowers. Means ± SE from two biological replicas. n = 29-39 for total number of emasculated flowers per line. i-k, entire failed to rescue 8a 8b placenta growth defect. In i, pictures were taken five weeks after emasculation. Bar = 2 cm. Means ± SE from two biological replicas. n =28-51 for total number of emasculated flowers per line. In f-h and j-k, different letters above bars represent significant differences, p < 0.05 (f, j, k) or p < 0.01 (g, h). The p values were made with one-sided (f, j) or two-sided (g, h, k) analysis. Exact n and p values for f-h and j-k were listed in Source Data Fig 1. In boxplots g, h and k, center lines and box edges are medians and the lower/upper quartiles, respectively. Whiskers extend to the lowest and highest data points within 1.5× interquartile range (IQR) below and above the lower and upper quartiles, respectively.

To examine fruit set without pollination, we emasculated flowers of these arf8 mutants to remove stamens and petals at 2 days before anthesis (−2 DAA), and recorded their parthenocarpic fruit growth after 5 weeks when the fruits had reached their maximum size at the mature green stage. WT did not display any parthenocarpy, while 8a-1 and 8b-1 single mutants produced seedless fruits from ~70% of unfertilized ovaries (Fig. 1e, 1f). The 8a 8b double mutant displayed even stronger parthenocarpy (~90%) than single mutants, and its parthenocarpic fruit size was also bigger than the single mutants (Fig. 1e, 1g), indicating that SlARF8A and SlARF8B have overlapping functions in repressing fruit initiation and development. However, 8a 8b almost completely eliminated placenta growth (Fig. 1e, 1h), which is the tissue for ovules/seeds attachment (Fig. 1d). The 8a and 8b single mutants also showed reduced placenta growth with intermediate phenotypes between WT and the double mutant. The strong parthenocarpy and the lack of placenta growth in 8a 8b mutant suggest dual function of SlARF8A/8B in repressing fruit initiation and promoting placenta growth, respectively.

We also examined phenotypes of mutant fruits after natural pollination, and found that the overall fruit sizes of 8a-1 and 8b-1 were smaller than WT (Supplementary Fig. 1a–1b). The placenta and locular tissue in these single mutants were also reduced in comparison to WT. The fruit phenotypes (after fertilization) in 8a-1 and 8b-1 point to the possible activator function of ARF8 in fruit growth. In contrast, the 8a 8b double mutant displayed obligatory parthenocarpy because it produced similar parthenocarpic fruits with or without emasculation (Fig. 1e and Supplementary Fig. 1a), indicating the double mutant is either male-and/or female-sterile. The 8a 8b fruit size was further reduced from the 8a-1 or 8b-1 single mutant (Supplementary Fig. 1b), indicating that both SlARF8A and SlARF8B promote fruit growth. We also found that almost every flower of 8a 8b was able to form fruit (95%, Supplementary Fig. 1c–1d) because of its high parthenocarpy frequency. In contrast, only 40% of WT flowers developed into fruits under the same growth condition. The total fruit yield (total weight) per cluster was also higher in 8a 8b, although average weight per fruit of 8a 8b was lower comparing to WT and arf8 single mutants (Supplementary Fig. 1e–1f). To confirm that the phenotypes in 8a-1 and 8b-1 were caused respectively by the slarf8a or slarf8b crispr allele but not off-target mutations, P_ARF8A:3xFLAG-2xHA-ARF8A (3F2H-ARF8A) and P_ARF8B:3xFLAG-2xHA-ARF8B (3F2H-ARF8B) transgenic lines were made and crossed to 8a-1 or 8b-1 respectively. 3F2H-ARF8A and 3F2H-ARF8B transgenes rescued the strong parthenocarpy phenotype in 8a-1 and 8b-1 mutants, respectively (Supplementary Fig. 2a–2d), verifying that 8a-1 and 8b-1 alleles lead to parthenocarpy.

As mentioned earlier, the placenta growth defect in 8a 8b is likely the consequence of missing the activator roles of SlARF8A/8B. But it could also result from the repressor functions of other class A-ARFs forming complexes with IAA9. To test the latter possibility, epistasis analysis was performed between 8a 8b and the sliaa9 null allele [entire (e)] to examine if removing SlIAA9 function would restore placenta growth in 8a 8b. As shown in Fig. 1i, the emasculated entire mutant flowers produced large seedless fruits with well-developed placenta and locular gel-like tissue, whereas the e 8a 8b triple mutant displayed similar placenta defect as in 8a 8b, suggesting that ARF8A/8B are the major ARFs promoting placenta growth in tomato. However, the e 8a and e 8b double mutants displayed more placenta growth than the 8a-1 and 8b-1 single mutants, indicating that removing SlIAA9 repression can enhance the activator role of the remaining functional SlARF8A or 8B in placenta. These results thus confirmed that the slarf8a slarf8b double mutations are epistatic to sliaa9 in regulating placenta growth.

In addition to fruit development, the arf8 mutants also showed additional defects in vegetative growth and in flower development. 8b-1 was shorter than WT and the 8a 8b double mutant was even shorter and produced smaller leaves than 8b-1 (Supplementary Fig. 3a–3e), indicating ARF8A and ARF8B additively promote stem growth and leaf expansion. The 8a-1 and 8b-1 mutations also additively reduced style length to 82% of WT, but increased anther cone length to 110% of WT (Supplementary Fig. 3f–3h). These subtle defects may reduce pollination efficiency, but the total infertility in 8a 8b is likely caused by additional factors as shown by the results of reciprocal crosses between WT and 8a 8b (Supplementary Fig. 3i–3k). We found that 8a 8b pollen was as efficient as the WT pollen to trigger fruit set when WT ovary was the recipient. However, the fruits from 8a 8b (♂) x WT only produced 50% of seeds than those from WT x WT, suggesting reduced pollen viability in 8a 8b. In contrast, WT (♂) x 8a 8b cross produced seedless fruits just like 8a 8b itself, indicating 8a 8b is female-sterile. The smaller leaf, shorter stem/style and female-sterile phenotypes in 8a 8b are consistent with the phenotype previously reported for the MIR167 transgenic tomato line with reduced ARF6/8 expression (targeted by overexpression of AtMIR167a), except that stamen was also shorter in that study²⁹. These results suggest that SlARF8A/8B mainly act as activators for promoting auxin-mediated flower development, except that they may inhibit anther cone growth. The parthenocarpy and whole plant phenotypes in arf8a-1, arf8b-1 mutants were reproducible in arf8a-2, arf8b-2 single and double mutants (Supplementary Fig. 4), confirming that the mutant phenotypes in 8a 8b are due to slarf8a and slarf8b mutations.

Class A-ARFs exhibited dual function in fruit development

The arf8a arf8b double mutant produced large parthenocarpic fruits, which are likely promoted by other class A-ARFs, such as SlARF5 and SlARF7 that are also expressed at elevated levels in the ovary around anthesis¹⁹. To test this idea, we generated and characterized higher order arf mutants. In this study, all single and higher order arf mutants are homozygous, except when specified to be heterozygous. The arf5-1 (arf5) and CR-arf7 (arf7) null alleles were originally generated in the M82 cultivar³⁰, whereas the arf8a arf8b mutants are in the MM cultivar. Therefore, we introgressed the arf5 and arf7 null alleles into MM cultivar by genetic crossing 5 times and then crossed these mutants with 8a 8b to make higher order arf mutant combinations. arf5 displayed severe flower development defects in all floral organs. The arf5 flower is missing petals and anthers, and only consists of tiny ovary, sometimes with sepals (Fig. 2a, Supplementary Fig. 5a). As a result, arf5 could produce neither normal fruit nor parthenocarpic fruit. The heterozygous arf5/+ mutant showed overall WT-like phenotypes. However, the heterozygous arf5/+ mutant produced small parthenocarpic fruits at a low frequency in comparison to those of 8a 8b (Fig. 2d–2f). The arf7 mutant showed WT-like phenotypes throughout vegetative and reproductive growth. Similar to arf5/+, arf7 produced small parthenocarpic fruits at a low frequency (Fig. 2a–2c). Both 5 8a 8b and 7 8a 8b triple homozygous mutants showed lower parthenocarpy frequency and produced smaller seedless fruits than those of 8a 8b (Fig. 2a–2c). The placenta growth defects in 5 8a 8b and 7 8a 8b were the same as in 8a 8b. These results indicated that SlARF5 and SlARF7 also function in repressing fruit initiation and promoting pericarp expansion during fruit growth, while they are not required for placenta growth.

Figure 2 |. ARF5, ARF7, ARF8A and ARF8B all contribute to fruit initiation and growth.

a-c, Parthenocarpy phenotypes of various higher order combinations of arf5, 7, 8a, 8b mutants. d-f, Parthenocarpy phenotypes of various combinations of arf 5/+, 7, 8a, 8b mutants. Parthenocarpy frequency (b, e) was shown as means ± SE from two biological replicas. In c and f, center lines and box edges are medians and the lower/upper quartiles, respectively. Whiskers extend within 1.5x IQR. n =26-49 (b, c) or 15-28 (e, f) for total number of emasculated flowers per line. Photos and measurements were taken five weeks after emasculation. Bar = 2 cm. Different letters above bars represent significant differences, p < 0.01 (in c) or p < 0.05 (in b, e, f). The p values were made with one-sided (b, e) or two-sided (c, f) analysis. Exact n and p values for b-c and e-f were listed in Source Data Fig 2.

To examine if SlARF8A and SlARF8B promote pericarp growth in addition to their strong roles in repressing fruit initiation, we analyzed the effect of adding 8a-1 and 8b-1 alleles into the arf5 arf7 background. Comparing to the arf5 single mutant, the arf5 arf7 double homozygous mutant showed even more severe floral defects as its flowers only formed a pin-like structure, which could not develop into fruits (Fig. 2a). Remarkably, both 5 7 8a and 5 7 8a 8b/+ mutants that contain two or one copies of WT 8B produced medium-size seedless fruits with 20-30% frequency, whereas the 5 7 8a 8b quadruple homozygous mutant showed no parthenocarpy at all (Fig. 2a–2c). Similarly, both 5 7 8b and 5 7 8a/+ 8b mutants that contain two or one copies of WT 8A produced medium-size seedless fruits with ~15% frequency (Supplementary Fig. 5b–5d). These results indicated that both SlARF8A and 8B also promote fruit growth. Our data revealed a biphasic response in parthenocarpic fruit expansion, which increased initially with removal of 1-2 SlARFs, then decreased when more SlARFs were knocked out, until no growth occurred when all four SlARFs were knocked out. We verified the roles of these SlARFs further using arf mutant combinations with arf5/+ for the parthenocarpy assay, because the homozygous arf5 mutant had severe ovary defect (during flower development) that may interfere with fruit development. The 5/+ 7 mutant displayed weak parthenocarpy as in arf5/+, again indicating their minor roles in repressing fruit initiation (Fig. 2d–2f). Similar to 7 8a 8b, 5/+ 8a 8b produced smaller parthenocarpic fruits than 8a 8b, demonstrating that removing one functional copy of SlARF5 in 8a 8b could reduce fruit growth. Furthermore, 5/+ 7 8a 8b seedless fruits were even smaller than those in 5/+ 8a 8b and 7 8a 8b (Fig. 2d–2f). Again, the placenta growth in these triple and quadruple mutants remained the same as in 8a 8b.

To verify further the role of SlARF8A and 8B in promoting fruit growth, we generated transgenic tomato lines P_ARF8A:3F2H-ARF8A-NT and P_ARF8B:3F2H-ARF8B-NT, which expressed either truncated SlARF8A or SlARF8B proteins lacking the C-terminal PB1 domain (1-716 a.a. for ARF8A-NT and 1-715 a.a. for ARF8B-NT). Because the PB1 domain is required for ARF8-IAA9 interaction¹⁹, expression of the SlARF8-NT proteins should lead to constitutive auxin signaling by uncoupling SlARF8A/8B activity from SlIAA9. The effect of IAA9 on ARF8A vs ARF8A-NT was first tested in a synthetic yeast system, which was modified from the auxin response circuit (ARC) that was previously proven functional for testing Arabidopsis auxin signaling components^31,32. In this yeast ARC system, we expressed SlARF8A or SlARF8A-NT in the presence or absence of SlTPL1^N100-SlIAA9 fusion (a truncated TPL-IAA9 fusion). The yeast strain also contains an auxin-responsive promoter P3_2x:Venus (an output reporter). Both SlARF8A and SlARF8A-NT induced P3_2x:Venus expression (Supplementary Fig. 6). However, co-expression of SlTPL1^N100-SlIAA9 dramatically down-regulated SlARF8A-induced P3_2x:Venus, but had no effect on SlARF8A-NT activity. Consistent with the yeast ARC results, both transgenic ARF8A-NT and ARF8B-NT tomato lines with high expression levels exhibited strong parthenocarpy (Fig. 3a–3d). Moreover, these ARF8-NT lines produced large seedless fruits with well-developed placenta and locular tissue, similar to the entire (sliaa9) seedless fruit (Fig. 1i) and pollinated WT fruit (Supplementary Fig. 1a). In contrast, most of the 31 P_ARF8A:3F2H-ARF8A transgenic lines produced WT-like seeded fruits (Supplementary Fig. 2a–2b), except one line with highest expression (referred to as ARF8A-OE) displayed obligate parthenocarpy (Fig. 3a–3d). This ARF8A-OE line produced parthenocarpic fruits that are smaller than those of 8A-NT and 8B-NT lines, although its ARF8A protein levels were higher than the ARF8-NT lines. These results support that deletion of the PB1 domain in ARF8A and ARF8B enabled them to promote parthenocarpic fruit growth more effectively than the full-length ARF8.

Figure 3 |.

Parthenocarpy phenotypes of 3F2H-ARF8A-OE, 3F2H-ARF8A-NT and 3F2H-ARF8B-NT transgenic lines. a-c, Photos and measurements were taken five weeks after emasculation. Bar = 2 cm. d, Detection of 3F2H-ARF8A (in both normal and OE lines), 3F2H-8A-NT, 3F2H-ARF8B and 3F2H-8B-NT proteins in −2 DAA ovaries by α-HA antibody. WT was included as a negative control, and α-tubulin (TUB) was included for immunoblotting to show similar loadings. Note that 3F2H-ARF8A-OE, 3F2H-ARF8-NT proteins were extracted from T0 hemizygous ovary while 3F2H-ARF8 from T1 homozygous ovary. b-c, Parthenocarpy frequency (b) and fruit diameters (c) of 8A-OE, 8A-NT and 8B-NT lines. b, Means ± SE from two biological replicas. n =30-32 for total number of emasculated flowers per line. In boxplot c, center lines and box edges are medians and the lower/upper quartiles, respectively. Whiskers extend within 1.5x IQR. Different letters above bars represent significant differences, p < 0.01 (in b, c). The p values were made with one-sided (b) or two-sided (c) analysis. Exact n and p values for b-c were listed in Source Data files.

Altogether, the results from the different arf combinations and the ARF8-NT transgenic lines strongly support that SlARF5, SlARF7, SlARF8A and SlARF8B all play dual function in fruit development: they act as inhibitors when associated with SlIAA9 during fruit initiation and as activators in fruit growth. SlARF8A and SlARF8B are stronger repressors than ARF5 and ARF7 in fruit initiation, while all four ARFs act as activators to promote pericarp growth. On the other hand, placenta growth is mostly promoted by SlARF8A/8B.

Spatiotemporal expression patterns of ARFs and IAA9 in ovary

To investigate the roles of these four class A-SlARFs and SlIAA9 in different ovary tissues (pericarp, ovule, placenta and septum) during fruit initiation and growth, absolute transcript levels of these genes were examined around anthesis by RT-qPCR analysis (Fig. 4a and Supplementary Fig. 7). In −2 DAA ovary, SlARF8A and SlARF8B transcript levels were similar in all four tissues, whereas SlARF5 and SlARF7 transcript levels were highest in ovule (Fig. 4a). SlIAA9 transcript levels were high in all four ovary tissues (Fig. 4a). In addition to −2 DAA ovary, we also monitored spatial and temporal expression patterns of these genes using 0, +2, +5 DAA ovaries (Supplementary Fig. 7). In general, the distribution patterns for each SlARF and SlIAA9 were similar as those in −2 DAA ovary. Most notably, expression of SlARF5 and SlARF7 in ovule/seeds peaked around 0 to +2 DAA, and dramatically decreased at +5 DAA, implying their roles during the ovule-to-seed transition period.

Figure 4 |. ARFs and IAA9 showed distinctive spatial localization in ovary.

a, Absolute transcript levels of SlARFs and SlIAA9 in four different tissues (pericarp, ovule, placenta, septum) of −2 DAA ovary. The transcript levels were calculated using standard curves¹⁹ (see Supplementary Fig. 7), and are shown as copies of ARFs/IAA9 cDNA per 10² copies of SlUBQ7. Means ± SE of 3 biological replicas are shown. b-h, ARFs/IAA9 protein localization in −2 DAA ovaries of different transgenic tomato lines as labeled by immunoblot analyses. Pericarp (Pc), Ovule (O), Placenta (Pl) and Septum (S). Transgenic tomato lines expressing epitope-tagged ARFs and IAA9 showed distinctive protein localization in different ovary tissues: (b) 3FLAG-2HA-ARF8A (3F2H-ARF8A); (c) 3F2H-ARF8B; (d) 3F2H-ARF8A-NT; (e) 3F2H-ARF8B-NT; (f) 3F-ARF5; (g) 3F-ARF7; (h) 3F2H-IAA9. Non-transgenic WT −2 DAA whole ovary was used as a negative control. i, Temporal distribution of 3F2H-IAA9 protein in ovaries before and after anthesis. WT −2 DAA ovary was used as a negative control. Arrow: epitope-tagged ARF or IAA9 proteins. *: non-specific proteins. Tagged-ARFs were detected by α-FLAG antibody while tagged-IAA9, ARF8A-NT or ARF8B-NT by α-HA antibody. α-Tubulin (TUB) was included for immunoblotting to show similar loadings. Source data for a-i are provided in Source Data files.

To monitor spatial distribution of SlARFs and SlIAA9 proteins, we generated transgenic tomato lines expressing epitope-tagged SlARF/IAA9 constructs under corresponding endogenous promoters (P_ARF:tag-ARF or IAA9). SlARF5 and SlARF7 constructs contained a 3xFLAG (3F) tag while SlARF8A, SlARF8B and SlIAA9 had a 3F2H tag. As described earlier, 3F2H-ARF8A and 3F2H-ARF8B are functional in planta to rescue the 8a-1 and 8b-1 mutant phenotypes, respectively (Supplementary Fig. 2a–2d). 3F-ARF5 and 3F2H-SlIAA9 also rescued all defects in the arf5 homozygous mutant and the sliaa9 (entire) mutant, respectively (Supplementary Fig. 2e–2i). ARF protein localizations were analyzed in four ovary tissues from −2 DAA flower: pericarp, ovule, placenta and septum. Immunoblot analysis indicated that SlARF8A and SlARF8B showed similar spatial expression patterns with high levels in pericarp, placenta and septum, and low levels in the ovule (Fig. 4b–4c). These protein expression patterns are mostly consistent with their transcript levels, except in the ovule, possibly due to downregulation of SlARF8A/8B translation by miR167 in ovule³³. Indeed, SlARF8A-NT/8B-NT that lack the miR167 target sequence within the PB1 domain showed similar levels in all four ovary tissues (Fig. 4d–4e). SlARF5 and SlARF7 showed opposite expression patterns to those of SlARF8A and SlARF8B in that they were mainly present in ovules (Fig. 4f–4g), although low levels of SlARF5 were also detected in the pericarp, septum and placenta (Fig. 4f). SlIAA9 was detected in all four tissues: high levels in ovules, medium levels in placenta and low in pericarp and septum (Fig. 4h), again showing discrepancy between transcript and protein distributions for this key auxin signaling component. We also examined SlIAA9 protein levels in the ovary daily from −2 DAA to +5DAA, and found that it was slightly reduced at +2 DAA and was significantly decreased at +5 DAA (Fig. 4i), consistent with the idea that auxin produced after anthesis results in degradation of IAA9 protein.

Taken together, the distinct spatial distributions of SlARFs and SlIAA9 proteins in ovary indicated specific roles of SlARFs and SlIAA9/ARF repressor modules in regulating development of different fruit tissues, and these spatial results were, in general, consistent with arf and entire mutant phenotypes.

SlARF8A/8B target genes during fruit initiation and growth

To identify putative SlARF8A/8B target genes that are involved in auxin-mediated fruit set and growth, transcriptome analysis was performed by RNA-Seq using the following samples: (1) −2 DAA ovaries of 8a 8b and WT; (2) −2 DAA WT ovaries treated with mock or 100 μM 2,4-D (a synthetic auxin) for 2, 6 or 24 h, respectively. These 2,4-D treatment time points were chosen based on our previous RT-qPCR results indicating clear changes in expression of most auxin-responsive genes at 6 h, although some genes only showed significant changes at 24 h. We also included 2h treatment in an attempt to identify early auxin response genes. Three biological repeats were included in each set of samples, except that the +2,4-D 2 h treatment used two biological repeats for RNA-seq analysis. The differentially expressed gene (DEG) lists for SlARF8A/8B-responsive genes (235) and for auxin-responsive genes (1731/3262/3212 in 2/6/24 h treatments) were identified with the following parameters: fold change > 1.5; p < 0.1 for 8a 8b vs. WT dataset; p < 0.05 for +2,4-D vs. mock treatment dataset (Supplementary Table 1). By comparing among these gene lists, we found that almost 60% of DEGs that were responsive to 8a 8b (139 DEGs out of 235 total) were co-regulated with at least one of the +2,4-D time points (Fig. 5a–5b, Supplementary Table 2). Furthermore, most of the co-regulated genes were found in the +2,4-D 6 h treatment and/or 24 h treatment dataset (108 DEGs out of 139 total, Fig. 5c).

Figure 5 |. Identification of 8a 8b- and auxin-responsive genes in ovary by RNA-seq analysis.

RNA-seq analysis was performed using −2 DAA ovaries of 8a 8b and WT, and −2 DAA WT ovaries ± 100 μM 2,4-D for 2, 6 or 24 h. The differentially expressed gene (DEG) lists for 8a 8b-responsive genes and for 2,4-D-responsive genes are in Supplementary Table 1. a-b, Venn diagrams of coregulated DEGs by 8a 8b and/or +2,4-D 2/6/24 h treatments. Down-regulated DEGs in a, Upregulated DEGs in b. c, Heat map of co-regulated DEGs among 8a 8b and +2,4-D 6/24 h. d, Selected gene groups from co-regulated genes between 8a 8b and +2,4-D 6 h/24 h. MADS-BOX genes: AG1 (AGAMOUS1), AGL6 and MADS2 (MADS-BOX PROTEIN 2. HD genes: HOX.1 (HOMEOBOX-LEUCINE ZIPPER PROTEIN 1), HD52 (Homeodomain protein H52) and ZF-HD1 (ZINC-FINGER HOMEODOMAIN 1). MYB genes: MYB21 (R2R3MYB21), MYB77 (R2R3MYB77). HSF (HEAT STRESS TRANSCRIPTION FACTOR), ACO4 (ACC OXIDASE 4), ABAH (ABA 8’-HYDROXYLASE), ERF4 (ETHYLENE RESPONSE FACTOR 4), GA20ox1 (GA20-OXIDASE 1) and CKX2 (CYTOKININ OXIDASE 2), XTH7 and XTH15 (for xyloglucan endotransglucosylase/hydrolase) and TPX1 (Peroxidase1), NR (Nitrate Reductase), SUS (for sucrose synthase), MIPS (for Myo-inositol-1-phosphate synthase).

For all co-regulated DEGs, Gene Ontology (GO) analysis was performed to identified 75 enriched GO terms that can be organized into 16 groups of different biological processes, cellular components and molecular functions (Supplementary Table 3). Although the co-regulated DEGs belong to a variety of biological pathways, certain groups were more representative, including transcription factors, hormone-related genes and growth-related genes (Fig. 5d). Eleven genes were chosen from these 3 groups to verify their expressions in 8a 8b by RT-qPCR (Supplementary Table 2). These include four genes encoding transcription factors [two MADS-BOX genes, AGAMOUS1 (AG1)³⁴ and MADS-BOX PROTEIN 2 (MADS2), HOMEOBOX-LEUCINE ZIPPER PROTEIN 1 (HOX.1) and HEAT STRESS TRANSCRIPTION FACTOR (HSF)³⁵]; four that function in hormone metabolism and signaling [ACC OXIDASE 4 (ACO4)³⁶, ABA 8’-HYDROXYLASE (ABAH)³⁷, ETHYLENE RESPONSE FACTOR 4 (ERF4)³⁸ and GA 20-OXIDASE 1 (GA20ox1)³⁹]; two genes related to cell wall modification [Xyloglucan endotransglucosylase/hydrolase 7 (XTH7)⁴⁰ and Peroxidase1 (TPX1)⁴¹]; and a Nitrate Reductase (NR)⁴² for converting inorganic nitrate to organic nitrite which is crucial for protein production in plants. Consistent with the RNA-Seq data, the RT-qPCR assays confirmed that expression of these genes was down-or up-regulated in 8a 8b (Fig. 6a, 6b) or by 2,4-D treatment (Fig. 6c, 6d). We further tested several 2,4-D-responsive growth-related genes by RT-qPCR, and found that XTH15 was also induced in 8a 8b ovary (Fig. 6b). The altered expression of hormone-related genes suggested that 8a 8b has reduced ethylene and ABA biosynthesis and signaling (by downregulating ethylene biosynthesis gene ACO4, upregulating ERF4 whose ortholog in Arabidopsis was shown to inhibit ethylene and ABA responses⁴³, and upregulating ABAH for ABA deactivation), and increased GA levels (by upregulating GA20ox1). Because fruit initiation is known to be repressed by ethylene and ABA^27,44, and promoted by GA³, the observed changes in hormone potentials would facilitate fruit initiation in the 8a 8b ovary. Likewise, XTH7, XTH15 and TPX1 function in promoting cell wall biosynthesis^40,41, their up-regulation in 8a 8b ovary points to increased potential for fruit growth.

Figure 6 |. Spatial expression patterns of auxin-responsive ARF8A/8B target genes.

a-d, RT-qPCR analysis of WT vs. 8a 8b (a-b) or WT after mock vs. 2,4-D 24h treatment (c-d) to verify selected down-regulated and up-regulated genes that were identified from RNA-seq data in Fig. 5. RNA from −2 DAA whole ovary of WT and 8a 8b (a, b) or WT after mock vs 2,4-D treatment for 24h (c, d) were used. e-i, RT-qPCR showing differential expression patterns of SlARF8A/8B target genes (e: AG1; f: MADS2; g: AGL6; h: XTH7; i: GA20ox1). RNA samples from four dissected tissues of −2 DAA ovaries of WT and 8a 8b were used for this analysis. For all RT-qPCR analyses, the housekeeping gene UBQ7 was used to normalize different samples. Means ± SE of 3 biological replicas are shown. a-i, expression level in WT whole ovary was set to 1. Asterisks above bars represent significant differences to their WT counterpart. * p < 0.05, ** p < 0.01. The p values were made with one-sided (b, e-i) or two-sided (a, c, d) analysis. Exact p values for a-i were listed in Source Data Fig 6.

Interestingly, two of the down-regulated transcription factors in 8a 8b are MADS-BOX genes AG1 and MADS2. Several MADS-BOX genes [e.g., AG1, AGL6 (AGAMOUS-LIKE 6), TM29] have been implicated as repressors of fruit initiation as their transcript levels were significantly down-regulated in ovary upon pollination or by auxin treatment^45,46. More importantly, the AGL6 null mutants display strong parthenocarpy phenotype⁴⁷, indicating that AGL6 is a major repressor in fruit set. RT-qPCR analysis confirmed that AGL6 transcript levels were repressed in 8a 8b ovary (Fig. 6a). Because SlARF8A and SlARF8B proteins in the −2DAA ovary are more abundant in most tissues except in the ovule, we further analyzed the MADS-BOX, XTH7 and GA20ox1 transcript levels to determine whether altered transcript distributions correlate with the spatial localization of SlARF8A/8B proteins. AG1, MADS2, AGL6 all showed reduced transcript levels in pericarp of 8a 8b, comparing to the corresponding tissues in WT (Fig. 6e–6g). In placenta and septum, these genes also showed reduced expressions in 8a 8b, although only AG1 had significant difference from WT. In contrast, only AG1 showed reduced expression in 8a 8b ovule in comparison to that in WT, while expression of MADS2 and AGL6 did not change significantly (Fig. 6e–6g), indicating that downregulation of these MADS-BOX genes in 8a 8b coincides with SlARF8A/8B protein localization in ovary. Similarly, expression of XTH7 and GA20ox1 was induced in pericarp/septum and pericarp/placenta tissues of 8a 8b respectively, while not in ovule (Fig. 6h–6i). These results suggest that the MADS-box genes, XTH7 and GA20ox1 are SlARF8A/8B targets. To test if SlARF8A/8B directly regulate these genes, ChIP-qPCR assay was performed using −2 DAA ovaries of the P_ARF8A:3F2H-ARF8A line. Cross-linked chromatin was immunoprecipitated using anti-FLAG beads, and qPCR was performed using 2-3 primer pairs that span the promoter region of each MADS-box gene, XTH7 and GA20ox1. On average, we found 1.7-to 2.3-fold enrichment in these promoters when normalized against UBQ7 promoter (Fig. 7a), supporting that SlARF8A directly binds to these target promoters to regulate their expression.

Figure 7 |. Confirmation of ARF8A/8B target genes by ChIP-qPCR and RT-qPCR.

a, ChIP-qPCR analysis showed SlARF8A binding to promoter regions of AG1, MADS2, AGL6, XTH7 and GA20ox1genes. −2 DAA ovaries of the P_ARF8A:3F2H-ARF8A line and anti-FLAG beads were used for the ChIP experiment. Numbers on x-axis represented nt distances (kb) from the start codon of each gene. The relative enrichment was calculated by normalizing against ChIP-qPCR of nontransgenic samples using UBQ7 coding region as control. Means ± SE of 3 biological replicas are shown. b, RT-qPCR analysis showing MADS-BOX genes were downregulated in −2 DAA ovaries of higher order arf mutants, entire and ARF8-NT transgenic lines in comparison to WT, except for MADS2 in 5 7 8a, 5 7 8a 8b/+ and 8A-NT lines. c, RT-qPCR analysis showing upregulation of XTH7 and GA20ox1 in −2 DAA ovaries of arf mutants, entire and ARF8-NT lines. The 8a 8b data in (b) and (c) were the same as in Fig. 6a. For all RT-qPCR analyses, the housekeeping gene UBQ7 was used to normalize different samples. Means ± SE of 3 biological replicas are shown. Expression level in WT whole ovary was set to 1. Asterisks above bars represent significant differences to their WT counterpart. * p < 0.05, ** p < 0.01. The p values were made with one-sided analysis. Source data including all p values are provided in Source Data Fig 7.

MADS-BOX expression in unpollinated ovary requires IAA9/ARFs

Our RNA-Seq and ChIP-qPCR results showed that the three MADS-BOX genes (AG1, MADS2, AGL6) that are downregulated in fruit initiation are direct targets of SlARF8A/8B. Repression of these MADS-BOX genes appeared to be useful biomarkers to assess whether phase transition from ovary to fruit set had occurred, especially when no further growth was observed (e.g., the 5 7 8a 8b mutant). RT-qPCR analysis indicated that −2 DAA ovaries of the quadruple 5 7 8a 8b mutant showed reduced expression of all three MADS-BOX genes, similar to those in the parthenocarpy mutants 8a 8b, 5/+ 7 8a 8b, 5 7 8a and 5 7 8a 8b/+ (Fig. 7b). The only exception was that MADS2 expression was not reduced in 5 7 8a and 5 7 8a 8b/+ comparing to WT, possibly due to the presence of SlARF8B. These results support that fruit initiation had occurred in these mutants, although the quadruple 5 7 8a 8b mutant ceased in subsequent fruit growth process. We further analyzed expression of growth-related genes in the higher-order arf mutants. Consistent with the mutant phenotypes, GA20ox1 showed increased expression in all parthenocarpy arf mutants, but was not elevated in 5 7 8a and 5 7 8a 8b mutants comparing to that in WT (Fig. 7c). In contrast, XTH7 expression was induced in all higher-order arf mutants, including 5 7 8a 8b (Fig. 7c), suggesting that induction of this gene is insufficient to trigger fruit growth. Likewise, in −2 DAA ovary of SlARF8A-NT and SlARF8B-NT lines, the MADS-BOX genes were all repressed comparing to WT, except for MADS2 in 8A-NT (Fig. 7b). Both XTH7 and GA20ox1 were induced in these ARF8-NT lines except for XTH7 in 8B-NT (Fig. 7c). In addition, the three MADS-BOX genes and GA20ox1 were also repressed (Fig. 7b) or induced (Fig. 7c), respectively, in −2 DAA ovary of the entire (iaa9) mutant comparing to WT. These results suggested that IAA9 also participates in the regulation of these genes, and that increased expression of these MADS-BOX genes in unpollinated WT ovary requires IAA9 and ARFs.

Current dual function model of class A-ARFs suggests that ARFs bind to their target promoters both in the presence and absence of IAAs. Our yeast ARC results shown earlier (Supplementary Fig. 6) is consistent with this model as both ARF8A and ARF8A-NT (without the PB1 domain) activated transcription of the synthetic auxin-responsive promoter P3_2x. To further test this in planta, RT-qPCR analysis was performed using +2 and +5 DAA WT fruits when the IAA9 protein levels were slightly or significantly reduced, respectively (Fig. 4i). In comparison to −2 DAA ovary, expression of five selected ARF8A target genes in +2 DAA fruits only displayed subtle changes (AG1, MADS2, and XTH7) or were unaltered (AGL6 and GA20ox1), consistent with the slight reduction of IAA9 protein amount at this time point. As expected, in +5 DAA fruits, all three MADS-BOX genes were repressed 2-5 fold, while XTH7 and GA20ox1 genes were induced 11-and 6-fold, respectively (Supplementary Fig. 8a–8b). More importantly, ChIP-qPCR analysis confirmed that ARF8A still bound to these target promoters in +5 DAA fruit where IAA9 was mostly degraded (Supplementary Fig. 8c). ChIP-qPCR using the ARF8A-NT line also showed binding of ARF8A-NT to the same ARF8 target promoters in −2 DAA ovary, further confirming that ARF8A can bind to these selected targets with or without forming complex with IAA9 (Supplementary Fig. 8d). Whether all ARF8A target genes remain the same in the presence or absence of IAA9, ChIP-Seq analysis would need to be performed.

Discussion

Our study illustrated the dual function of four class A-SlARFs (ARF5, ARF7, ARF8A, ARF8B) in inhibiting fruit initiation while promoting fruit growth in tomato (Fig. 8a). Importantly, altering doses of these four class-A SlARFs led to biphasic fruit growth responses as summarized in Fig. 8b, revealing the complex combinatorial effects of multiple ARFs in fine-tuning auxin-mediated fruit set and subsequent fruit growth. Generation of the higher order slarf mutant combinations and the identification of marker genes for fruit initiation are critical tools, which allowed us to dissect the separate roles of ARFs in fruit set and fruit growth. By doing so, we were able to unmask the activator function of these class A-SlARFs in fruit growth, which was impossible in previous studies because of their inhibitory role in fruit set. We found that SlARF8A and SlARF8B are central repressors (together with SlIAA9) in fruit set as 8a 8b led to maximum parthenocarpy frequency, although SlARF5 and SlARF7 also play minor repressive roles in fruit set as arf5/+ and arf7 mutants showed low rates of parthenocarpy. The parthenocarpy frequency, defined as percentage of fruits (diameter ≥ 1 cm) that developed from unpollinated ovaries, reflects the combined effects of both fruit set and subsequent growth. Hence, we used three MADS-BOX genes (AG1, AGL6 and MADS2) as molecular markers to gauge the onset of fruit initiation because these genes are direct targets of SlARF8A/8B (Fig. 5–7) and function as repressors of fruit set^45,47,48. Reduced expression of these MADS-BOX genes suggests fruit initiation has occurred (Fig. 7b), even in the absence of apparent fruit growth in 5 7 8a 8b quadruple mutant (shown as the dashed line in Fig. 8b).

Figure 8 |. Model of class A-ARFs and IAA9 in controlling fruit initiation and growth in tomato.

a, Model of the roles of SlIAA9 and class A-SlARFs (ARF8A, ARF8B, ARF5 and ARF7) in regulating auxin-induced early events in fruit development, divided into two phases. Phase 1: transition from ovary to fruit set. Phase 2: fruit set to fruit growth (including both cell division and cell expansion). The four A-SlARFs display a dual role. Before pollination, SlIAA9/ARF complexes repress fruit set by activating expression of MADS-BOX genes, although the mechanism of this transcription activation is unclear (?). Upon anthesis, auxin levels are elevated in the ovary, which lead to SlIAA9 degradation and release class A-SlARFs to activate fruit growth (by up-regulating growth-related genes). b, Biphasic fruit growth in response to combinatorial regulation of four class A-SlARFs. The graph summarizes parthenocarpy frequency (% of parthenocarpic fruits, green circles in graph) and fruit diameter (blue circles in graph) from emasculated flowers of WT and single, double, triple and quadruple arf mutants. All data were derived from Fig. 2 after normalization using 8a 8b data that are present in different datasets. The general trend of fruit initiation is shown based on parthenocarpy frequency (solid green line) or inferred from down-regulated expression of MADS-BOX genes (dashed green line). The general trend of fruit growth (blue curve) is plotted based on fruit diameters of different genotypes. b was created with BioRender.com free plan.

To dissect the individual and combined role of these SlARFs in fruit growth, we then focused on the effects of these arf mutants in altering fruit sizes. We found that parthenocarpic fruit diameter increased from arf5/+ or arf7 mutants to arf8 single mutants, then to 8a 8b double mutant (Fig. 2 and 8b). However, further adding arf5, arf5/+ or arf7 mutations to the 8a 8b background resulted in reduced fruit growth. Eventually, fruit growth was completely abolished in the arf5 7 8a 8b quadruple mutant, although fruit initiation is likely activated in this mutant judging by repressed expression of MADS-BOX genes (Fig. 7b, 8b). The strong parthenocarpic phenotype of the ARF8A/8B-NT transgenic lines (expressing truncated ARF8A/8B proteins lacking the PB1 domain) further illustrated the promoting role of ARF8A and ARF8B in fruit growth when uncoupled from IAA9 (Fig. 3a). These results strongly support that all 4 class A-ARFs play a dual role in fruit initiation and fruit growth, and they contribute to fruit growth in a dosage-dependent manner. Before anthesis (pollination), the repressive role of these class A-SlARFs in fruit set is dependent on SlIAA9 as the null sliaa9 mutant (entire) and the ARF8-NT lines all displayed strong parthenocarpy. Expression patterns of SlIAA9 and SlARF proteins before anthesis are consistent with their corresponding roles during fruit set. We found that SlIAA9 protein was present in the whole ovary. SlARF8A and SlARF8B proteins were mainly localized to pericarp, septum and placenta, so knocking out these SlARFs removed their repression in these tissues. In contrast, SlARF5 and SlARF7 proteins were mainly localized in ovules so that arf5/+ or arf7 mutant only displayed weak parthenocarpy because of the presence of the strong repressors SlARF8A/8B in other tissues. Previous studies showed that AGL6 and possibly additional MADS-BOX genes are key repressors of fruit set^45,47,48. Transcript levels of these genes are down-regulated by silencing or null mutation of IAA9⁴⁵, which was confirmed in our study (Fig. 7b). Importantly, we found that these MADS-BOX genes were also down-regulated in the 8a 8b mutant (Fig. 6a), and they are direct targets of SlARF8A by ChIP-qPCR (Fig. 7a). In addition, expression of these MADS-BOX genes was downregulated at +5 DAA WT ovary (when IAA9 protein levels were low) in comparison to that at −2DAA (when IAA9 levels were high) (Supplementary Fig. 8a and 8b, Fig. 4i). Taken together, these results strongly support that IAA9/ARF complexes directly promote transcription of MADS-BOX genes (AG1, MADS2 and AGL6) before anthesis (Fig. 8a). This is a surprising finding as IAAs are only known as transcriptional repressors^11,14. It remains to be elucidated how SlIAA9/ARF complexes directly activate transcription of these MADS-BOX genes. Based on our model, the reduction in the IAA9 protein levels after pollination would lead to reduced expression of the MADS-BOX genes to basal levels in the absence of the IAA9/ARF complexes. Alternatively, IAA9/ARFs may inhibit unidentified repressor(s) of MADS-BOX genes before anthesis, and ARFs may promote expression of repressor(s) of MADS-BOX genes after pollination when IAA9 levels are reduced.

Our ARF8A/8B and auxin co-regulated DEG list highly correlates with a recent study on tomato transcriptomic reprogramming after pollination or auxin treatment⁴⁹. Among our 139 DEGs, 96 and 91 genes overlap with the 4-day after pollination DEG list (4DPA) and the 4-day after IAA treatment DEGs (4IAA) in that study, respectively (Supplementary Table 4). For the 13 genes tested in Fig. 6a–6d, 10 of them showed correlation with 4DPA DEGs and 9 with 4IAA DEGs in the previous study, including AG1, MADS2, XTH7 and GA20ox1. Interestingly, the same study also identified a number of epigenetic regulators on their 4DPA/4IAA DEG list, nine of which were on our 24h 2,4-D DEG list (Supplementary Table 4). Their study further revealed strong correlation between changes in transcription levels and alterations in H3K9ac or H3K4me3 histone marks at target genes during pollination and IAA-induced fruit set⁴⁹. These results point to synergistic regulation of histone modification and gene expression by active auxin signaling at fruit set.

Interestingly, the lack of placenta development of 8a 8b double mutant is completely opposite to its whole fruit growth phenotype. The large parthenocarpic fruit of 8a 8b with very little placenta growth suggests that regulation of fruit growth could be divided into two compartments: the growth of fruit wall, i.e. pericarp and septum, and the growth of inner tissue, i.e. seeds and placenta. In the 8a 8b mutant, fruit growth is achieved solely by the growth of fruit wall. In contrast, the entire (iaa9) mutant produced large parthenocarpic fruits (from emasculated flowers), which resemble WT fruits (from self-pollinated flowers) with well-developed placenta and locular tissue, except seedless. However, entire could not rescue the placenta defects in 8a 8b. In addition, the ARF8A/8B-NT transgenic lines produced large parthenocarpic fruits with similarly well-developed placenta and locular tissue, confirming that placenta growth is exclusively promoted by SlARF8A and SlARF8B, which is different from the involvement of all four ARFs in promoting pericarp and septum growth.

In angiosperms, AFBs, IAAs and ARFs all belong to large-gene families, which make functional analysis difficult due to high degrees of functional redundancy^8,9. In contrast, the early emerging land plants, such as Physcomitrella patens (moss) and Marchantia polymorpha (liverwort), contain relatively simple auxin response systems with low or no redundancy. Recent studies using these simple model systems elegantly resolved the fundamental molecular mechanisms of these key components of early auxin signaling^10,50,51. The large family members encoding ARFs and Aux/IAA proteins in flowering plants, in contrast to the basal land plants, are thought to provide more complex auxin responses. Indeed, a recent transcriptome study⁸ identified more auxin-responsive genes, especially auxin-induced genes in angiosperms, than those in algae and bryophytes. However, the developmental outcomes affected by this increased complexity in target gene regulation are largely unknown. By analyzing high order mutants of four A-SlARFs, our study uncovered their activator role in mediating auxin-induced fruit growth despite their inhibitory role in fruit initiation. It remains to be determined whether a similar mechanism applies to A-ARFs in regulating fruit development in other plant species. Our work also demonstrated that growth of different tissues of tomato fruit is controlled by different combinations of SlARFs, which is potentially useful for agronomic applications in controlling growth of specific fruit tissues. In addition, the high parthenocarpy frequency of the 8a 8b mutant led to higher fruit yield than that of WT. Given the globally changing climate, the knowledge on the molecular mechanism of class A-SlARFs-controlled parthenocarpic fruit development could provide valuable tools to generate resilient crops with enhanced fruit yield to increase food security.

METHODS

Plant Materials, Growth Conditions and Statistical Analysis

Tomato (Solanum lycopersicum) cultivar Moneymaker (MM) was used as WT in this study. The entire (iaa9) mutant was backcrossed to MM five times from their original Alisa Craig background¹⁸. arf5-1 and CR-arf7 mutants were backcrossed to MM five times from their original M82 background¹⁹. CR-arf8a and CR-arf8b mutants were generated in this study. Genotyping primers for entire, arf5-1 and CR-arf7 were described previously¹⁹. Genotyping primers for the CR-arf8a, CR-arf8b mutants are listed in Supplementary Table 5. In this study, all single and higher order arf mutants are homozygous, except when specified to be heterozygous.

Tomato plants were grown in greenhouse with 16 h day/8 h night light cycle as described previously¹⁹. Parthenocarpy test was done as described previously¹⁹ with some modifications: −2 DAA flowers on mature tomato plants (2-6 weeks after first flowering) were emasculated and recorded fruit growth 5 weeks later; fruit diameter at least 1 cm was considered as parthenocarpy. For fruit yield analysis, # of flowers were counted for each young flower cluster of ~ two-month-old plants, and fruits developed on these clusters were harvested individually and weighed when they were at the red ripe stage. Whole plant phenotypes including plant height, internode numbers and length, and leaf morphology were recorded 5 weeks after sowing. For 2,4-D treatment, 10 μl of 100 μM 2,4-D, or mock solvent (5% MeOH, 0.1% Tween-20) was applied to −2 DAA ovaries of emasculated flowers.

All statistical analyses were performed using Student’s t-test.

Plasmid Construction

Primers for plasmid construction are listed in Supplementary Table 5. PCR-amplified DNA fragments in constructs were sequenced to ensure that no mutations were present. Detailed information on plasmid construction is described in the Supplementary Table 6.

Plant Transformation

The P_GENE:tag-GENE (including SlARF8A, SlARF8B, SlARF5, SlARF7, SlIAA9), P_GENE:tag-GENE-NT (including SlARF8A-NT and SlARF8B-NT) and CRISPR/Cas9-ARF8A/ARF8B constructs were introduced into tomato (MM) through agrobacterium-mediated plant transformation with strain GV3101 pMP90⁵². For P_GENE:tag-GENE transformation, T0 lines containing a single insertion were identified by their T1 seedlings showing a 3:1 ratio of kanamycin-resistant versus kanamycin-sensitive segregation patterns. For each construct, 3-5 independent homozygous lines with WT-like phenotype were tested by immunoblot analysis to identify representative spatial expression patterns in ovary, and a representative line (i.e. 3F-ARF5 # C7-3-3; 3F-ARF7 # C4-7-6; 3F2H-ARF8A # C5-24-4; 3F2H-ARF8B # C8-4-1; 3F2H-IAA9 # C6-4A-6) was chosen for final expression analysis and complementation test. Most of the 31 3F2H-ARF8A T0 lines had normal fruit and seed production, while one line #C5-5 was completely parthenocarpic. This line had the highest ARF8A protein level among all T0s and was named ARF8A-OE. For P_GENE:tag-GENE-NT transformation, 25 ARF8A-NT and 9 ARF8B-NT T0 lines were generated. These T0 lines were used directly for phenotyping, protein analysis and RT-qPCR analysis. The representative lines used in this study are 3F2H-ARF8A-NT # G1-2 and 3F2H-ARF8B-NT # G2-3. For CRISPR/Cas9-ARF8A/8B transformation, arf8a and arf8b mutant alleles were amplified by PCR using genotyping primers in T0 lines and sequenced to identify the molecular lesions. To remove possible off-target mutation(s) and the Cas9 transgene, a representative T0 mutant was backcrossed to WT once.

Synthetic Auxin Response System in Yeast

The published minimal ARC^sc (for ARC in yeast) contains a single AtTIR1/AFB, AtTPL^N100-AtIAA (truncated TPL-IAA fusion), AtARF and an auxin-responsive promoter:Venus (output reporter)^31,32. The Venus signal can be quantified by flow cytometry. We generated new ARC^sc with tomato components including 3F-SlTPL1^N100-SlIAA9 fusion, 3F-SlARF8A or 3F-SlARF8A-NT. All tomato genes are expressed using constitutive yeast promoters. Our system also contains a synthetic auxin-response promoter P3_2x fused to the fluorescent protein Venus coding sequence (P3_2x:Venus)³². pGP8A-3F-SlARF8A, pPG5G-3F-SlARF8A-NT constructs were separately transformed into the yeast reporter line in yeast strain MATa. Then pGP4G-3F-SlTPL1^N100-SlIAA9 was transformed to ARF8A or 8A-NT-containing lines or P3_2x:Venus line alone. All constructs were incorporated in yeast chromosome permanently. Venus fluorescence was recorded by flow cytometry with an FACSCanto Flow Cytometer (BD Bioscience). The median value of 10,000 cells was recorded as one reading and was included in Supplementary Data file.

Immunoblot Analysis

Whole ovary or dissected ovary tissues were extracted by grinding in 2x Laemmli buffer (Bio-Rad) and boiling for 10 min. After centrifugation, lysates were separated by SDS-PAGE, and detected by immunoblotting with either an HRP-conjugated anti-FLAG antibody (Sigma A8592 clone M2, for detection of 3F2H-ARF8A, 3F2H-ARF8B, 3F-ARF5, 3F-ARF7 with dilutions of 1:3,000, 1:2,000, 1:1,000 and 1:2,000, respectively) or a mouse anti-HA antibody (BioLegend #901503, for detection of 3F2H-IAA9, 3F2H-ARF8A-NT, 3F2H-ARF8B-NT with dilution of 1:1,000). As gel loading control, tubulin was detected with a mouse anti-tubulin antibody (Sigma T5168) at dilution of 1:500,000. HRP-conjugated donkey anti-mouse secondary antibody (Jackson ImmunoResearch) was used for anti-HA and anti-tubulin immunoblotting at dilution of 1:5,000 and 1:50,000, respectively.

RNA-Seq Analysis

Total RNAs were purified from −2 DAA ovaries of WT and 8a-1 8b-1, as well as −2 DAA WT ovaries treated with 100 μM 2,4-D for 2/6/24 h or mock treated (5% methanol, 0.1% Tween 20). RNA-Seq cDNA libraries (three biological repeats) were prepared with the QuantSeq 3’mRNA-Seq library prep kit FWD for Illumina (Lexogen). DNA sequencing was performed with Illumina Next-Seq500 High-Output 75bp SR. Sequence alignment and DE (differential expression) analysis were done online at lexogen.bluebee.com. Heatmap was created in RStudio with plotly package. Co-regulated genes among 8a 8b and +2,4-D 2/6/24 h gene lists were then identified (fold change > 1.5; p < 0.1 for 8a 8b vs. WT; p < 0.05 for +2,4-D vs. mock). Venn Diagrams were made using online tool at InteractiVenn.net⁵³. GO analysis was performed using agriGO v2 toolkit⁵⁴. Heatmap analysis was made in R language with the plotly package.

RT-qPCR Analysis

Total RNAs from whole ovary or dissected ovary tissues were isolated with Quick-RNA MiniPrep kit (Zymo Research). First-strand cDNA was then synthesized using a Transcriptor First Strand cDNA Synthesis kit (Roche). qPCR analyses were performed using FastStart Essential DNA Green Master mix (Roche) and LightCycler 96 instrument (Roche). The PCR program was performed as described before¹⁹. Three biological replicates from independent pools of tissues (2 technical repeats each) were included for each experiment. Primers for qPCR are either described before¹⁹ or listed in Supplementary Table 5.

For absolute qPCR analysis, the qPCR standard curves of UBQ7 and SlARF genes were made previously ¹⁹. For SlIAA9, linearized plasmid pENTR1A-IAA9 cDNA¹⁹ served as template for determining cDNA copy vs qPCR cycle number (Supplementary Fig. 6f). With these standard curves, the cDNA copy numbers of UBQ7, SlARFs and SlIAA9 were determined according to their respective cycle numbers.

ChIP-qPCR Analysis

Ovaries of the 3F2H-ARF8A transgenic line C5-24-4 (−2 and +5 DAA) and ARF8A-NT line (−2 DAA) were used for ChIP-qPCR. Experiment was performed as described¹⁹. Primer sequences for qPCR test are listed in Supplementary Table 5.

Data Availability

The RNA-seq data have been deposited in the NCBI Sequence Read Archive (SRA) under BioProject PRJNA929538. Source Data files are provided with this paper.