Gibberellin and miRNA156-targeted SlSBP genes synergistically regulate tomato floral meristem determinacy and ovary patterning Free

A wide variety of plant organ sizes and shapes exist in nature, and this variation can be explained in part by the control of the apical meristem activity. In maize, for instance, variation in meristem determinacy is responsible for differences in kernel row number between inbred lines, and high row number is a major driver of yield in cultivated hybrids (Doebley, 2004; Bommert et al., 2013). Regulation of meristem function occurs at several levels, including identity and determinacy; for example, floral meristem (FM) identity specifies that the meristem can initiate floral organs and, in most cases, the FM is determinate, i.e. it is consumed in the production of a limited number of organ primordia (Bartlett and Thompson, 2014). The determinacy and size of the FM set the number of cells available to initiate carpel primordia and, consequently, carpel number and ovary size (van der Knaap et al., 2014; Heidstra and Sabatini, 2014). Thus, variation in FM size and determinacy in response to hormonal and genetic pathways has had an important impact on ovary patterning and the control of seed dispersal, which have been pivotal to crop domestication and improvement (Purugganan and Fuller, 2009; Østergaard, 2009).

A classic genetic pathway involved in the regulation of meristem size and determinacy is based on a feedback circuit comprising the stem cell-promoting WUSCHEL (WUS) homeodomain transcription factor and CLAVATA (CLV) signalling peptides (Stahl and Simon, 2010). This circuit is conserved in diverse plants, including tomato (Solanum lycopersium). Larger meristems in tomato mutants with altered CLV-WUS activity produce fruits with a higher number of locules (the interior cavities of fruits). Importantly, natural mutations in CLV3 and WUS were essential for tomato domestication (Muños et al., 2011; Xu et al., 2015; Zsögön et al., 2018). Whereas most wild tomatoes and small-fruited cultivars have bilocular fruits, large-fruited cultivars can have eight or more locules (Tanksley, 2004). Most of this variation is due to two mutations (locule number or lc and fasciated or fas) with synergistic effects on the number of locules and thus fruit size (Lippman and Tanksley, 2001; Barrero and Tanksley, 2004). The lc mutation is localized downstream of tomato WUSCHEL [S. lycopersium (Sl)WUS], whereas a regulatory mutation repressing SlCLV3 (or FAS) expression underlies the fas mutant (Muños et al., 2011; Xu et al., 2015). However, although the core CLV-WUS circuitry is deeply conserved, recent work has shown that the pathway is modulated by diverse inputs (Rodriguez-Leal et al., 2019). Given the complexity of genetic and hormonal interactions that regulate meristem function (Lee et al., 2019), there is much potential for discovering novel molecular mechanisms that affect fruit development.

Meristem size is regulated by the microRNA156 (miR156) and its transcription factor targets, which are members of the SQUAMOSA PROMOTER-BINDING PROTEIN-LIKE (SPL or SBP) gene family. This evolutionarily conserved pathway was initially found to regulate age-dependent processes (Wu et al., 2009; Morea et al., 2016). More recently, Arabidopsis miR156-targeted SBP genes were shown to regulate the size of both root and shoot meristems (Fouracre and Poethig, 2019; Barrera-Rojas et al., 2020). Downregulation of SBP genes in miR156-overexpressing (156OE) plants also led to abnormal fruit development in both Arabidopsis and tomato (Silva et al., 2014; Xing et al., 2013), but with distinct phenotypic consequences: whereas miR156 overexpression only reduced the size of Arabidopsis gynoecia, tomato 156OE gynoecia displayed extra carpels and fruit-like ectopic structures (Silva et al., 2014; Xing et al., 2013). Interactions with other pathways also differ between Arabidopsis, in which miR156-targeted SBP genes control gynoecium patterning through interference with auxin homeostasis and signalling, and tomato, in which miR156-targeted SBP genes modulate the expression of genes involved in meristem maintenance (LeT6/TKn2) and organ boundary formation (GOBLET or GOB). Thus, similar miRNA-based pathways control initial steps of gynoecium patterning but with distinct functional consequences in dry and fleshy fruit-bearing plants (Correa et al., 2018).

Another molecular player in meristem and ovary patterning is the phytohormone gibberellin (GA). GA-regulated DELLA genes control shoot meristem function, modulating the size of inflorescence and FMs (Serrano-Mislata et al., 2017). Furthermore, the GA-deficient ga1-3 mutant displays delayed growth of all floral organs (Yu et al., 2004). GA controls ovary patterning through the interaction between DELLA and the basic helix-loop-helix (bHLH) proteins INDEHISCENT (IND) and ALCATRAZ (ALC), which specify the tissues required for fruit opening (Arnaud et al., 2010). In tomato, however, scarcely any information is available of how GA modulates FM determinacy and fruit development. The loss of PROCERA (PRO; DELLA in tomato) gene function in procera (pro) mutants or gibberellic acid (GA₃) application led to meristic changes in the flower (increased number of all floral organs) and occasional fruits with ectopic fruit-like structures (indeterminate ovaries), although the phenotypes are not as strong as in 156OE plants (Carrera et al., 2012; Silva et al., 2014).

In Arabidopsis, GA signalling and miR156-targeted SBP genes interact during the floral transition: the GA-regulated DELLA protein REPRESSOR OF GA1-3 (RGA) associates with LEAFY (LFY) and the miR156-targeted SBP proteins to promote APETALA1 (AP1) transcription (Yamaguchi et al., 2014). However, the interactions between the GA and miR156 pathways beyond the floral transition have not been reported. Given the similar phenotypic changes observed in flowers and fruits from tomato 156OE plants and pro mutant (Carrera et al., 2012; Silva et al., 2014), we hypothesized that miR156-targeted SlSBP genes and GA signalling also interact to orchestrate tomato FM size and determinacy, early gynoecium/ovary patterning and fruit development. Here, we show that the GA and miR156 pathways regulate FM size and that they partially interact at the molecular and genetic levels to control ovary determinacy and locule number, without primarily relying on modifications in the classical SlCLV3-SlWUS circuitry, thereby revealing a previously unreported mechanism in the control of tomato ovary patterning.

Because we showed that the miR156-silenced SlSBP genes interplayed with GA to control tomato floral transition (Silva et al., 2019), and that the rice mir156abcdfghikl mutant is hyposensitive to GA (Miao et al., 2019), we reasoned that the strong ovary indeterminacy observed in 156OE plants (Silva et al., 2014) could be associated with increased GA signalling. To initially test this conjecture, we quantified bioactive GA₁ and GA₄ levels in the tomato floral primordia. Whereas GA₁ levels were undetectable in both wild-type (WT) and 156OE primordia, GA₄ levels were significantly higher in 156OE primordia compared with those in WT primordia (Fig. S1A). This finding is consistent with the lower levels of several bioactive GAs (including GA₄) found in the mir156abcdfghikl mutant (Miao et al., 2019). Next, we treated WT and 156OE plants with commercial GA₃ and inspected the ovaries and fruits for the presence of at least one ectopic fruit-like structure growing from their stylar end, scored as ‘indeterminate’ fruits. ‘Malformed’ fruits represent an intermediate class comprising fruits that did not exhibit obvious fruit-like structures, but rather incomplete carpel fusion. Fruits were scored as ‘normal’ when the ovaries showed no presence of ectopic or malformed structures. Unlike Carrera et al. (2012), we did not observe indeterminate or malformed fruits in GA₃-treated WT plants (Fig. S1B). This divergence may be a result of the differences in the GA₃ treatment (Carrera et al., 2012; Silva et al., 2019). Nonetheless, externally applied GA₃ led to a 40% increase in indeterminate fruits in 156OE plants compared with those in mock-treated 156OE plants, suggesting that, in addition to higher GA levels, 156OE plants were more sensitive to GA compared with WT plants (Fig. S1B). Conversely, treatment of 156OE plants with paclobutrazol (PAC; a GA biosynthesis inhibitor; Jung et al., 2011) resulted in ∼2-fold more normal fruits than those in mock-treated 156OE plants, and almost no indeterminate and fewer malformed fruits (Fig. S1C).

To evaluate in more detail how GAs modulate fruit development in the 156OE plants, we inspected ovaries at anthesis and fruits from the hypomorphic pro mutant (pro harbours a mutation in the PRO gene that lessens protein activity; Carrera et al., 2012; Livne et al., 2015), GA20ox-overexpressing plants (GA20oxOE, which show high levels of bioactive GAs; García-Hurtado et al., 2012) and 156OE plants. Neither pro nor GA20oxOE plants showed obvious modifications in the ovaries, although 156OE ovaries displayed partially fused extra carpels, as previously described (Fig. 1A-D) (Silva et al., 2014). Around 3% of pro and 20% of 156OE fruits showed some degree of indeterminacy (likely initiated in the ovaries), as indicated by the development of one or more additional fruit-like structures at the style end of the fruit (Fig. 1I,J). The χ² statistical test indicated a significant (P<0.01) association between genotype and ovary/fruit indeterminacy, except for GA20oxOE and pro plants, which produced none or few indeterminate ovaries, respectively (Fig. 1I,J). Based on these observations, we hypothesized that high levels/responses of GA can generate strong ovary indeterminacy when SBP genes are expressed at low levels in reproductive primordia, as shown in GA₃-treated 156OE fruits (Fig. S1).

To genetically confirm the results obtained with GA₃ treatment, we evaluated ovary growth and patterning in 156OE;GA20oxOE and 156OE;pro plants. Strikingly, both 156OE;GA20oxOE and 156OE;pro plants showed 100% of ovaries with strong indeterminacy (Fig. 1E-J). 156OE;GA20oxOE and 156OE;pro flowers displayed gynoecia formed by supernumerary, partially fused carpels and ectopic pistil-like structures, which did not generate any noticeable locule-like structures (Fig. 1E-H). As a result, the 156OE;GA20oxOE and 156OE;pro amorphous fruits were seedless (Fig. 1I). Notably, all these gynoecia and fruits phenotypes were reminiscent of strong miR156-overexpressing tomato lines, which also produced amorphous, seedless fruits (Silva et al., 2014). Collectively, our results indicate that the indeterminate ovary and abnormal fruit phenotypes observed in 156OE;GA20oxOE and 156OE;pro plants are at least, in part, a result of the synergistic effects of GA and the miR156/SBP module converging on the regulation of meristem determinacy and floral organ identity.

Recent evidence indicates that exogenous GA treatment enhances locule number in tomato fruits (Li et al., 2019, 2020). Considering that increased locule number may result from extranumerary carpels due to a mild increase in FM indeterminacy, we checked how the interaction between GA and SlSBP genes that are silenced by miR156 affects locule number. In line with the link between indeterminacy and increased locule number, pro, GA20oxOE and 156OE fruits all displayed more locules than WT fruits. Whereas most WT fruits displayed two to three locules, the majority of pro and GA20oxOE fruits showed three to five locules (Fig. 2A,B). GA20oxOE and pro fruits displayed comparable number of locules, indicating that modifications in either GA levels or responses similarly affect locule development. Consistently, in PAC-treated WT plants, almost 60% of fruits had only two locules and 21% of fruits had three locules (Fig. 2C). The increase in locule number of pro and GA20oxOE was comparable, but not as severe as in 156OE plants, in which most fruits exhibited four to six locules (Fig. 2A,B) (Silva et al., 2014). We next inspected locule number in fruits from mock- and PAC-treated 156OE plants. PAC-treated 156OE plants showed a modest reduction in the number of fruits exhibiting five to seven locules (Fig. 2D), suggesting that miR156-targeted SlSBP genes and GA control tomato locule number through partially independent mechanisms.

Meristem determinacy and size are tightly correlated with the rate of organ initiation, as larger meristems produce more organs per unit time, including carpels (Xu et al., 2015; Je et al., 2016; Serrano-Mislata et al., 2017). Because pro and 156OE plants displayed variable degrees of ovary indeterminacy and locule number (Figs 1 and 2), we compared FM size in WT, pro and 156OE plants. To precisely measure FM size, we first established in tomato the modified pseudo-Schiff propidium iodide (mpSPI) and imaging methodology described for Arabidopsis inflorescence meristems (Bencivenga et al., 2016; Serrano-Mislata et al., 2017). To compare this parameter among different genotypes, we harvested tomato floral primordia with FMs displaying emergence of sepal primordia in a helical pattern (at 1-2 dpi; Fig. S2; Xiao et al., 2009). After segmenting and identifying FM cells in three dimensions (3D), we used the combined volume of cells in the L1 layer of the meristem dome as a proxy for meristem size (Fig. 3A). 156OE plants exhibited slightly larger meristems than pro plants, and both FMs were significantly larger than WT FMs (Fig. 3B). Importantly, these results correlate FM volume and the number of locules among these genotypes (Figs 2 and 3). To identify the cellular basis for the differences in meristem size, we analysed the cell number and cell volume in the L1 of tomato FMs. The mean number of cells in 156OE FMs was significantly greater than that in WT and pro FMs. The cell number of pro meristems was intermediate between that of WT and 156OE meristems (Fig. 3C). In addition, the enlargement in 156OE and pro FMs (Fig. 3B) was associated with increased cell volumes compared with those of WT FMs, although the cell volumes of 156OE and pro FMs were comparable (Fig. 3D). Our findings indicated that reduced PRO and SlSBP gene activities enlarged the tomato FM by increasing both cell size and number.

We have previously shown that some SlSBP genes and genes associated with boundary establishment (i.e. GOB) were mis-expressed in 156OE developing ovaries (Silva et al., 2014). To better understand the molecular mechanisms by which the miR156-targeted SlSBP genes interact with GA responses and how they modulate meristem activity (size and determinacy) and fruit patterning, we used RNA sequencing (RNA-seq) to monitor changes in gene expression in the 156OE floral primordia. RNA-seq experiments were conducted at 1-2 days post inflorescence (dpi) (when floral primordia comprise inflorescence meristems and FMs displaying emergence of sepal primordia; Fig. S3D; Xiao et al., 2009). At these developmental stages in WT plants, miR156 mature transcripts were weakly localized in the flanks of FMs and barely detected in the inflorescence meristem (Fig. S3C). In contrast, mature miR156 transcripts were readily detected in the dome of early and late vegetative meristems, as well as in leaf primordia (Fig. S3A,B), consistent with the role of miR156 as a master regulator of the vegetative phase (Hyun et al., 2016).

The RNA-seq analysis detected 240 differentially expressed genes (DEGs) between 156OE and WT floral primordia (Table S1). As expected, the DEGs included miR156-silenced SlSBP genes, which were downregulated in 156OE primordia. Two strongly repressed SlSBP genes (3- to 4-fold) were SlSBP3 (Solyc10g009080) and SlSBP15 (Solyc10g078700), which are representatives of the two main SBP clades (Table S1; Fig. 4A) (Morea et al., 2016). SBP3 and SBP15 are expressed during early floral development in Solanum pimpinellifolium but with distinct expression profiles (Fig. 4B; http://ted.bti.cornell.edu/cgi-bin/TFGD/digital/home.cgi; Chu et al., 2019). Based on the fruit tissue-specific expression atlas (https://tea.solgenomics.net/), SlSBP3 and SlSBP15 are preferentially expressed in the pericarp and septum of 0-days-post-anthesis (dpa) tomato fruits (Fig. 4C), which suggests that these SlSBP genes have roles in orchestrating the patterning of these tissues.

We identified genes directly associated with floral determinacy, such as the tomato CRABS CLAWa (SlCRCa; Solyc01g010240). SlCRCa and its paralog SlCRCb are positive, redundant regulators of tomato FM determinacy (Castañeda et al., 2022). It was surprising that SlCRCa was upregulated in FMs from 156OE plants, which have decreased ovary determinacy (Table S1; Fig. 4A,D). However, it has been reported that in the slcrcb null mutant, SlCRCa transcripts accumulated at higher levels than in WT plants, likely owing to a compensatory mechanism activated upon partial loss of ovary determinacy (Castañeda et al., 2022). We speculate that a similar mechanism activates SlCRCa in 156OE floral primordia, as 156OE ovaries also show a partial loss of determinacy (Fig. 1) (Silva et al., 2014).

Another interesting DEG was Solyc04g049800 (hereafter SlTOE-like), which was faintly expressed during early flower development and in fruit tissues (Fig. 4B,C) and was upregulated in 156OE primordia (Table S1; Fig. 4A,D). SlTOE-like is a target of miR172 (Karlova et al., 2013), which is repressed by the combined action of Arabidopsis thaliana (At)SPL9 and DELLA proteins to regulate the floral transition in Arabidopsis (Yu et al., 2012). Moreover, miR172 positively regulates tomato floral identity in a dose-dependent manner via AP2-like target genes, including SlTOE-like (Lin et al., 2021). This raised the possibility that the miR172/AP2 module also takes part in downstream responses to the miR156-targeted SlSBP genes in the tomato FM and, indeed, miR172 was downregulated in 156OE floral primordia (Fig. 4D).

In spite of the strong genetic interaction between 156OE and GA signalling, our RNA-seq data did not reveal changes in the expression of genes involved in GA biosynthesis and signalling associated with reduced expression of miR156-silenced SlSBP genes (Table S1). This result is consistent with the idea that most integration of miR156-targeted SlSBPs and GA-targeted DELLA proteins occurs by direct protein-protein interactions to regulate shared downstream targets (Yamaguchi et al., 2014). A candidate shared target was the MIKC^c-type MADS-box gene SlMBP18 (Solyc03g006830; Hileman et al., 2006), which was downregulated in 156OE primordia (Table S1; Fig. 4A,D). SlMBP18 is 49% identical to Arabidopsis AGAMOUS-LIKE 42 (AGL42), which promotes Arabidopsis flowering in a GA-dependent manner (Dorca-Fornell et al., 2011). In line with a role downstream of miR156-silenced SlSBP genes in floral primordia, SlMBP18 was expressed during early flower development in most tissues of 0-dpa fruits, with its expression partially overlapping with the SlSBP3 and SlSBP15 expression patterns (Fig. 4B,C).

Links to the CLV-WUS pathway were also not apparent in the RNA-seq data: SlWUS (Solyc02g083950) and SlCLV3 (Solyc11g071380) were expressed in our floral primordia samples but were not detected as differentially expressed (adjusted P-values=0.63 and 0.41, respectively). Other genes associated with the CLV-WUS pathway, such as WUSCHEL-RELATED HOMEOBOX 5 (Solyc06g076000; adjusted P-value=0.53; Rodriguez-Leal et al., 2019), and CLAVATA3 and ESR (collectively CLE) (Solyc05g009915 and Solyc11g066120, adjusted P-values=0.84 and 0.80, respectively; Rodriguez-Leal et al., 2019), were also not detected as differentially expressed. Considering the central role of WUS and CLV3 in meristem size (Stahl and Simon, 2010), we included SlWUS and SlCLV3 among the genes for which expression in 156OE and WT floral primordia was independently verified by quantitative real-time PCR (qRT-PCR) (Fig. 4D). SlCLV3 expression did not change significantly, but qRT-PCR did show downregulation of SlWUS in the 156OE floral primordia (Fig. 4D), perhaps because the qRT-PCR experiment had higher statistical power. Another possible explanation for this incongruity is that the SlWUS-SlCLV3 circuitry operates at later stages in tomato flower development (Chu et al., 2019; Rodriguez-Leal et al., 2019). However, the lack of changes in SlCLV3 expression and other genes associated with the CLV-WUS pathway makes it unlikely that SlSBP genes targeted by miR156 affect FM meristem size and determinacy primarily through the CLV-WUS pathway.

The results above suggested that the miR156-silenced SlSBP genes do not regulate FM function by modulating the expression of genes involved in GA signalling or in the CLV-WUS pathway. To test whether GA signalling or the CLV-WUS pathway modulate the expression of miR156-regulated SlSBP genes or their downstream targets, we compared the expression of selected genes in floral primordia of WT plants, the pro mutant and the fas mutant, which has reduced SlCLV3 expression (Barrero and Tanksley, 2004; Chu et al., 2019). miR172-targeted SlTOE-like was downregulated in fas floral primordia, but it was similarly expressed in pro and WT floral primordia. Conversely, SlCRCa was expressed at comparable levels in WT and fas primordia, but it was repressed in pro primordia (Fig. 4D). Neither pro nor fas affected SlSBP3 or SlSBP15 expression, but both mutants had lower expression of SlMBP18, as seen with 156OE plants (Fig. 4D; Fig. S4A). The levels of PRO transcripts did not change in fas and 156OE floral primordia compared with their levels in WT floral primordia, but PRO transcripts accumulated at higher levels in the pro mutant (Fig. S4B). Overall, our transcriptomic data suggest that the SlSBP genes targeted by miR156, PRO and SlCLV3 do not regulate each other's expression, but share overlapping sets of downstream targets during early floral development.

Overexpression of miR156 is expected to inhibit SBP gene expression, but the role of specific miR156-targeted SlSBP genes in tomato meristem size and fruit patterning is unclear. Given that SlSBP3 and SlSBP15 represent two distinct SlSBP clades (Morea et al., 2016) and were strongly downregulated in 156OE floral primordia (Table S1; Fig. 4), we next investigated how their de-repression would affect tomato FM activity and fruit patterning.

We initially generated tomato Micro-Tom (MT) lines overexpressing a miR156-resistant SlSBP3 allele (namely, rSBP3; Fig. S5A). We evaluated three rSBP3 lines, all of which showed much higher SlSBP3 transcript levels than WT, and no indeterminate fruits (Fig. 5A, Fig. S5B). rSBP3 lines displayed higher percentages (69 to 95%) of bilocular fruits when compared with those in WT (Fig. S5C). Given that rSBP3 lines #2, #4 and #6 displayed similar vegetative architecture and fruit phenotypes (Fig. S5C,D), we selected rSBP3#2 for further analyses. To study the effects of high levels of SlSBP15, we used a line with overexpression of a miR156-resistant SlSBP15 (rSBP15), which leads to axillary bud arrest and reduced lateral branching (Barrera-Rojas et al., 2023). Similar to rSBP3#2, rSBP15 plants produced no indeterminate fruits, and most of the fruits showed an apparent decrease in locule number (Fig. 5A). In summary, the effects of either SlSBP3 or SlSBP15 overexpression on fruit development were opposite to those seen in 156OE plants.

We next measured rSBP3#2 and rSBP15 FM size as described in Fig. 4. SlSBP3 overexpression only marginally reduced FM size, whereas rSBP15 meristems were significantly smaller than those of rSBP3#2 and WT plants (Fig. 5B; Fig. S6A). Cell numbers in rSBP15 FMs decreased compared with those in rSBP3#2 and WT FMs, which did not differ from one another (Fig. 5C). In contrast, both rSBP3#2 and rSBP15 FMs showed smaller cell volumes compared with those of WT FMs (Fig. 5D). These results suggest that the functions of different SlSBP genes targeted by miR156 do not fully overlap and that SlSBP15 has a more prominent role in FM size.

To evaluate how each of the two SlSPB genes could restore the functions inhibited by the miR156 overexpression, we crossed the rSBP3#2 and rSBP15 transgenes into the 156OE background. Only 34.8% of the 156OE;rSBP3#2 fruits had three locules, with the majority of the fruits having four or more locules (Fig. 5E,F). Moreover, most 156OE;rSBP3#2 ovaries at anthesis exhibited partially fused styles, similar to 156OE ovaries (arrows at Fig. 5E). These observations indicate that overexpressing SlSBP3 in 156OE is not sufficient to rescue the reproductive defects of 156OE plants. In contrast, 156OE;rSBP15 plants showed no indeterminate fruits and exhibited WT-like ovaries at anthesis. Importantly, over 95% of 156OE;rSBP15 fruits had two to three locules, comparable with WT fruits (∼82%) (Fig. 5E,F). Thus, high levels of SlSBP15 restored most of the developmental processes that are disrupted by miRNA156 overexpression. To test whether loss of SlSBP15 function is sufficient to explain the defects seen in 156OE plants, we examined CRISPR-Cas9 gene-edited SlSBP15 plants (sbp15^CRISPR; Barrera-Rojas et al., 2023). This loss-of-function mutant showed no fruit indeterminacy, and most fruits (over 80%) resembled WT fruits (Fig. S6B). Therefore, additional miR156-targeted SlSBP genes operating in the FM (Table S1) likely function redundantly with SlSBP15 to control FM activity and fruit patterning.

Because rSBP15 overexpression led to a stronger reduction in FM size and was sufficient to partially rescue WT-like ovary and fruit phenotypes (Fig. 5E,F), we monitored the SlSBP15 expression pattern in early flower development using in situ hybridization. At 1-2 dpi, SlSBP15 was mostly detected in flat FMs, which showed the emergence of sepal primordia at distinct developmental stages (Fig. 5G,H). By contrast, SlSBP15 transcripts were scarcely detected in sepal primordia or inflorescence meristems (Fig. 5H). At 4-5 dpi, SlSBP15 was expressed at low levels in floral buds (Fig. 5I). No signal was observed with the SlSBP15 sense probe (Fig. 5J). These findings reinforced the notion that SlSBP15 has an important role in controlling FM activity.

rSBP15 plants exhibited semi-dwarfism (Barrera-Rojas et al., 2023) (Fig. S7), a characteristic GA-deficient or GA-insensitive phenotype. Thus, we reasoned that high levels of SlSBP15 might attenuate GA responses or decrease GA sensitivity in tomato. To test this conjecture, we crossed the rSBP15 transgene into GA20oxOE plants, which have increased GA levels. In contrast to the typically tall GA20oxOE plants, rSBP15;GA20oxOE plants were semi-dwarf like rSBP15 plants (Fig. S7A). Importantly, rSBP15 also partially reverted the effects of GA20oxOE in the fruit; rSBP15;GA20oxOE plants produced a high percentage of WT-like fruits with two to three locules, in contrast with the three to four locules seen in most GA20oxOE fruits (Fig. 6A,B). Thus, SlSBP15 overexpression reduced the effects of high GA levels on tomato development. To test to what extent rSBP15 mimics the effects of reduced GA signalling, we also crossed the rSBP15 transgene into the pro background. The resulting rSBP15;pro plants showed similar vegetative architecture as that of pro plants (Fig. S7B) but had fewer indeterminate fruits than those of pro plants (Fig. 6C), even though the frequencies of locule numbers were similar (Fig. 6B). In conclusion, the genetic interactions suggest that rSBP15 opposes GA signalling during vegetative growth and in determining locule numbers but can also compensate for the effects of loss of GA signalling on fruit determinacy.

The data presented above suggested that the miR156-silenced SlSBP genes interact differently with GA signalling during vegetative development, ovary determinacy and fruit patterning. Furthermore, our RNA-seq data suggested that the effects of SlSBP genes in early FM development are not associated with clear changes in WUS/CLV3 expression. Similar to the fas mutant (Chu et al., 2019), pro and 156OE plants also showed an enlarged shoot apical meristem (SAM), whereas rSBP15 plants displayed a smaller SAM area compared with that of WT plants (Fig. S8A,B). We hypothesized that the SlSBP genes silenced by miR156 and PRO act along with the SlWUS-SlCLV circuitry at least partially via common pathways. To investigate how the CLV3-WUS pathway interacts with miR156-targeted SlSBP genes and GA signalling at the genetic level during fruit development, we next crossed 156OE, rSBP15 and pro with the fas mutant.

The combination of 156OE with fas showed dramatically enhanced defects in flower development and fruit patterning. Whereas 156OE and fas exhibited ∼20% of indeterminate ovaries, 156OE;fas plants had supernumerary, partially fused floral whorls such as sepals, petals, and carpels (Fig. S9A) and produced 100% of indeterminate ovaries (Figs 1J and 7A-D). Amorphous 156OE;fas fruits had ectopic fruit-like structures growing from their stylar end, with no apparent locular area (Fig. 7D; Fig. S9A), similar to fruits from 156OE;GA20oxOE and 156OE;pro plants (Fig. 2). Our data indicated that the combined loss of SlCLV3 and SlSBP gene function had synergistic effects on ovary determinacy and fruit development. Conversely, SlSBP15 overexpression in the fas mutant (rSBP15;fas plants) completely suppressed the ovary indeterminacy seen in the fas background (Fig. S8C). In contrast, fas mutants showed a large increase in locule numbers, which remained similar in rSBP15;fas plants (Fig. 7E,F). In summary, as seen above for the interaction with GA signalling, the miR156/SlSBP15 node and SlCLV3 interact differently in the control of locule number and ovary determinacy.

We next checked the interaction between fas and GA signalling. Arabidopsis DELLA proteins have been reported to restrict inflorescence meristem size independently of the canonical CLV-WUS circuitry (Serrano-Mislata et al., 2017). Similarly, application of the bioactive GA₄ or the GA inhibitor PAC modified locule number without relying on changes in the expression levels of SlWUS or SlCLV3 (Li et al., 2020), and mutation of the tomato DELLA (PRO) led to enlarged SAMs in the pro mutant (Fig. S8A,B). As seen for 156OE, combined loss of pro and fas had synergistic effects. The flowers from the double pro;fas mutant displayed excessive partially fused carpels containing extra carpels within, which were rarely observed in the pro or fas single mutants (Fig. S9B). Whereas ∼9% and 20% of indeterminate ovaries were observed in the pro and fas mutants, respectively, pro;fas plants produced 100% of indeterminate ovaries, all characterized by fruit-like structures growing from their stylar end, and no visible locular area (Fig. 7G; Fig. S9B). PAC treatment marginally reduced locule number in the fas mutant (Fig. S8D), suggesting that the increased locule number in fas is at least partially dependent on GA signalling.

Collectively, our genetic and molecular observations support the idea that miR156-silenced SlSBP genes, PRO, and SlCLV3 have overlapping functions in FM and fruit development, but their interactions vary between different aspects of fruit development, such as ovary determinacy and the regulation of locule number. The main genotypes described in this study and all the genetic interactions are summarized in Table 1.

The data presented so far suggested that changes in locule number and fruit determinacy must be under distinct regulation and do not simply unfold from early differences in FM size. One process that may change locule number without necessarily affecting determinacy is the formation of organ boundaries, in which genes of the CUP-SHAPED COTYLEDONS (CUC) family play a major role. Multiple lines of evidence implicate GOB (Solyc07g062840, a homologue of Arabidopsis CUC2) in the control of carpel and locule number in tomato. First, overexpression of miR164, which inhibits GOB, reduced locule numbers (Silva et al., 2014). Second, the loss of GOB function (gob-3 mutant) produced underdeveloped carpels (Berger et al., 2009) (Fig. S9B). Third, the semi-dominant GOB mutant (Gob-4d, which harbours a miR164-resistant version of GOB; Berger et al., 2009) exhibited gynoecia with ectopic, partially fused carpels, resulting in fruits with extra, malformed locules (Fig. S10A,C,D). Analysis of meristem size in Gob-4d/+ floral buds showed no significant differences from that of WT floral buds (Fig. S10F,G), further indicating that GOB is not associated with the control of FM size, but rather it affects fruit patterning at later developmental stages.

We have previously shown that GOB transcripts accumulate in 156OE developing ovaries (Silva et al., 2014) and recently reported that miR156-targeted SlSBP15 inhibited GOB activity in axillary buds (Barrera-Rojas et al., 2023). However, we did not find GOB transcripts in our transcriptomic data from 1- to 2-dpi floral primordia, suggesting that GOB activation only occurs at later stages of development. Indeed, GOB was upregulated in 6- to 8-dpi floral buds (when the carpel arises; Xiao et al., 2009) in both 156OE and pro plants, whereas it was downregulated in 6-to 8-dpi floral buds from rSBP15 plants (Fig. S10E). These observations suggested that GOB expression is regulated by PRO and the miR156/SlSBP15 node during the onset of carpel development. However, neither rSBP15 overexpression nor PAC treatment were sufficient to modify locule number in the Gob-4d/+ fruits (Fig. S10H,I), which indicates that additional factors (likely miR164) are necessary to fine-tune GOB expression in carpel development.

Some aspects of the complex interaction between the age-dependent miR156-targeted SBP genes and GA have been relatively well characterized in Arabidopsis and tomato, but how GA affects FM characteristics and fruit patterning in tomato has been unclear. We show that the repression or ectopic expression of miR156-silenced SlSBP genes (mainly SlSBP15) modulates tomato FM size, locule number and ovary determinacy. We also found that the fruit development roles of miR156-targeted SlSBP genes, GA signalling and the classic CLV-WUS pathway are different, but partially overlap. In particular, the reduced levels of SlSBP transcripts in 156OE plants had synergistic effects on fruit determinacy when combined with either high GA responses or with reduced activity of SlCLV3 (Table 1). Synergy generally arises when pathways that converge at a node or hub are disrupted (Pérez-Pérez et al., 2009). The miR156/SBP gene hub provides a notable example, given that SBP genes have been shown to connect many unrelated pathways (Wang and Wang, 2015).

Although the genetic interactions between the SlSBP genes targeted by miR156 and GA differed in tomato in terms of floral transition (Silva et al., 2019), their effect on meristem determinacy seems to be broadly conserved in other species. Meristem size is another conserved developmental aspect influenced by both GA and the miR156-targeted SBP genes. Tomato pro exhibited larger shoot and FMs, similar to plants overexpressing miR156 (Fig. 3; Fig. S7). The FM phenotype is largely explained by an increase in cell size and cell number in the L1 layer. Because we analysed only the morphology of cells in the L1 of FMs, variations observed in this cell layer may originate from additional morphological adjustments in the inner cell layers of the meristem. Although we did not analyse this conjecture in more detail, de-repression of the miR172-targeted AP2 in the central zone and organizing centre substantially increased Arabidopsis inflorescence meristem size (Sang et al., 2022). miR172 transcripts levels were reduced in the enlarged 156OE floral primordia, whereas miR172-targeted SlTOE-like was upregulated. On the other hand, SlTOE-like was similarly expressed in pro and WT primordia (Figs 4 and 7), suggesting that miR156-silenced SlSBP genes regulate the miR172/TOE-like node to modulate tomato FM size independently of GA.

In contrast, overexpressing rSBP15 in tomato attenuated GA responses in the vegetative development and in the establishment of locule number (Fig. 6; Fig. S6). Moreover, high levels of rSBP15 reduced the number of indeterminate ovaries in the pro mutant (Fig. 6), indicating that the miR156/SlSBP15 node and GA interact in distinct developmental contexts in tomato. Similar observations were reported for rice, in which the overexpression of Oryza sativa (Os)SPL14 blocks GA effects on seed germination, seedling growth and disease susceptibility (Liu et al., 2016; Miao et al., 2019). Loss of miR156 function in Arabidopsis leads to the production of a smaller vegetative SAM, whereas the quintuple mutant spl2;spl9;spl10;spl13;spl15, which disrupts miRNA156-targeted SBP genes, exhibits a larger SAM. In Arabidopsis, the miR156-targeted SBP genes appear to control SAM size by promoting WUS expression independently of CLV3 signalling (Fouracre and Poethig, 2019). However, our molecular and genetic data suggest that in the FM, SlSBP genes, GA and the SlWUS-SlCLV3 module converge on shared downstream targets. Future studies on the interactions with downstream targets such as SlCRCa and SlMBP18 may reveal how the different pathways are integrated at the molecular level, both in the SAM and FM.

In summary, our observations indicate that the miR156-silenced SlSBP genes function in tomato fruit development, but their interaction with GA signalling differs between vegetative and reproductive development. The hypothetical model showing the molecular link between GA and miR156-targeted SlSBP genes in fruit development is summarized in Fig. S11. Furthermore, SlSBP genes silenced by miR156 interact with both GA signalling and with the CLV-WUS pathway in ways that differ between developmental stages, from the regulation of FM size to setting locule numbers and the regulation of ovary determinacy. Thus, the interactions between these regulatory modules cannot be easily extrapolated between developmental contexts. Importantly, our work suggests that multiple genetic pathways are available to modify features of tomato fruit development that have so far been associated with the traditional CLV-WUS circuitry. For instance, although GOB may be another molecular link connecting GA and the miR156-targeted SlSBP genes (Fig. S11), the higher number of locules observed in the Gob-4d fruits is not a result of larger FMs, but rather modifications in the establishment of boundaries during carpel patterning (Fig. S10). Further studies on molecular mechanisms triggering floral determinacy and further gynoecium patterning will provide valuable information for tomato yield improvement, as the meristem size and final number of locules in the mature fruit are key factors for the establishment of the fruit size and shape. The accumulating knowledge of meristem regulatory pathways, and their relevance in regulating crop yield, might contribute to food security and sustainable agriculture in the next decades.

All genotypes of tomato (Solanum lycopersicum) described in this work were in the cultivar Micro-Tom (MT) background, which was used as wild-type (WT). The procera (pro) mutant, MT plants overexpressing miR156 (156OE), the miR156-resistant SlSBP15 allele (rSBP15) and the p35S::GA20ox construct (GA20oxOE), and sbp15^CRISPR plants were previously described (Garcia-Hurtado et al., 2012; Silva et al., 2014, 2017, 2019; Barrera-Rojas et al., 2023). The fasciated (fas), GOB 4-d (Gob4-d) and gob-3 alleles were introgressed into MT plants as described by Carvalho et al. (2011). Plants were grown as described by Silva et al. (2019). Floral primordia at 1-2 dpi and closed flower buds at 6-8 dpi were collected.

For crosses, flowers were emasculated and manually pollinated 2 days before anthesis to prevent self-pollination. Most crosses were evaluated in the F1 generation, except for 156OE;pro, rSBP15;pro, 156OE;fas, rSBP15;fas and pro;fas. These double mutants were evaluated in the F2 generation, in plants homozygous for the recessive allele pro and fas.

Fruits were inspected for the presence of at least one ectopic fruit-like structure growing from their stylar end, and scored as ‘indeterminate’ fruits. In addition, fruits were scored as ‘malformed’ when they exhibited an incomplete carpel fusion without any obvious fruit-like structure. Normal fruits (without the presence of ectopic or malformed structures) were cut in transverse sections to evaluate the number of locules. Each genotype was characterized by the percentage of the fruits that produced a specific number of locules. Over a hundred fruits were evaluated per genotype.

Total RNA was extracted from tomato leaves with TRIzol reagent (Thermo Fisher Scientific) and treated with Turbo DNAse (Thermo Fisher Scientific), and cDNA was synthesized using ImpromII Reverse Transcriptase (Promega). The SlSBP3 (Solyc10g009080) open reading frame was amplified and cloned into pENTR D-TOPO (Thermo Fisher Scientific). The miR156 recognition site-containing the 3′ untranslated region was removed from SlSBP3, therefore generating the miR156-resistant SlSBP3 allele (rSBP3). After sequencing, the cloned fragment was recombined into pk7WG2.0 (Gateway System) in front of the CaMV35S promoter using LR Clonase (Thermo Fisher Scientific), generating the p35S::rSBP3 construct. Tomato MT plants were transformed with the p35S::rSBP3 construct as described (Silva et al., 2014), generating rSBP3 plants. At least five transgenic events were obtained, and three were further analysed.

Total RNA was treated with DNAse and reverse transcribed to generate first-strand cDNA, as described above. PCR reactions were performed using GoTaq qPCR Master Mix (Promega) and analysed in a Step-OnePlus real-time PCR system (Applied Biosystems). Tomato TUBULIN (Solyc04g081490) was used as the internal control. Three technical replicates were analysed for three biological samples (each comprising at least ten floral primordia or closed buds), together with template-free reactions as negative controls. For miRNA quantification, cDNA synthesis and qPCR were performed as described (Varkonyi-Gasic et al., 2007). The threshold cycle (CT) was determined and fold changes were calculated using the equation 2^−ΔΔCT (Livak and Schmittgen, 2001). Oligonucleotide sequences are listed in Table S2.

Floral primordia at 1-2 dpi were collected from WT and 156OE plants and immediately frozen in liquid nitrogen. Two total RNA replicates from each genotype were sent to construct RNA-seq libraries and for high-throughput sequencing (Illumina NovaSeq platform) at Fasteris (https://www.fasteris.com/en-us; Switzerland). Raw sequencing reads were first cleaned by removing adaptor sequences and low-quality reads with Trimmomatic (Bolger et al., 2014) and BBDuk (https://sourceforge.net/projects/bbmap/), and library duplication was assessed with dupRadar (Sayols et al., 2016). The resulting high-quality reads were mapped to the tomato reference genome (Tomato Genome Consortium, 2012) (ITAG4.0) and transcripts were aligned, assembled and quantified using the Hisat2 and Salmon packages with default parameters (Kim et al., 2019; Patro et al., 2017). Differential expression analyses between WT and miR156-overexpressing plants were performed with the edgeR package (Robinson et al., 2010). DEGs in 156OE compared with WT (MT) with adjusted P≤0.05 and absolute fold change≥2.0 were filtered for further analysis.

DEGs were annotated based on the SOLGENOMICS database (version SL4.0 and Annotation ITAG4.0, https://solgenomics.net/) (Table S1). Enriched Gene Ontology terms for the DEG list were identified using the goseq package (https://bioconductor.org/packages/release/bioc/html/goseq.html) and the Gene Ontology Consortium database (Gene Ontology Consortium, 2021). The SOLGENOMICS database was used as a reference for the Gene Ontology analysis. Fisher's exact test with a false discovery rate correction with a cut-off of 0.05 was applied to determine enriched terms (Benjamini and Hochberg, 1995). Raw sequence data from this study have been deposited in Gene Expression Omnibus (GEO) of NCBI under the accession number GSE223674.

In situ hybridization was performed following the protocol described by Javelle et al. (2011). Primordia from 10-days-post-germination (dpg) WT seedlings and pre-anthesis WT ovaries were collected and fixed in 4% (w/v) paraformaldehyde. After alcoholic dehydration, plant material was infiltrated and embedded in Paraplast X-Tra (McCormick Scientific). The SlSBP15 (Solyc10g078700) probe was generated by linearizing (using BspHI; New England Biolabs) the pGEM-T Easy Vector (Promega Corporation) containing a 597-bp SlSBP15 fragment (nucleotides 523 to 1119 of the coding sequence). In vitro transcription was performed using the Digoxigenin (DIG) RNA Labeling Kit (SP6/T7, Roche). The sense SlSBP15 probe was used as a negative control. Locked nucleic acid probes with a sequence complementary to miR156 (5′-GTGCTCACTCTCTTCTGTCA-3′) and a negative control (scrambled miR, 5′-GTGTAACACGTCTATACGCCCA-3′) were synthesized by Exiqon (now QIAGEN), and DIG-labelled using a DIG oligonucleotide 3′ end-labelling kit (Roche Applied Science). Ten picomoles of each probe were used for each slide. All hybridization and washing steps were performed at 55°C as described by Javelle et al. (2011). Pictures were photographed using the Axio Imager.A2 (Carl Zeiss AG) light microscope. Oligonucleotide sequences are listed in Table S2.

Gibberellin (GA₃; Sigma-Aldrich; 10⁻⁵ M), paclobutrazol (PAC) (Sigma-Aldrich; 10⁻⁶ M) and mock solutions were applied to plants by watering as described (Silva et al., 2019).

At least five reproductive floral primordia at 1-2 dpi were dissected for each genotype, and the first FM was selected using a stereomicroscope for a standardized stage. Primordia bearing early FMs with sepal primordia were selected, stained by mpSPI, and imaged as previously described (Bencivenga et al., 2016; Serrano-Mislata et al., 2017).

For image analysis, Python scripts and Fiji macros were used to segment confocal image stacks, define the position of cells within the FM, and to delimit meristem cells and measure L1 volume. The function of each script and instructions on how to perform analysis with them were described in detail in Bencivenga et al. (2016) and Serrano-Mislata et al. (2017).

For SAM area measurements, shoot apices were collected at 4-6 dpg and photographed with the Leica S8AP0 stereomicroscope (Wetzlar, Germany) coupled to a Leica DFC295 camera (Wetzlar, Germany). Quantification of the SAM area was done in the ImageJ software (National Institutes of Health). We measured meristem width by drawing a horizontal line between the insertion points of the two youngest visible leaf primordia. From this line, we drew a vertical line to the top of the dome of the meristem to estimate its length (Xu et al., 2015; Rodriguez-Leal et al., 2019). At least four meristems were used for area measurements.

The data are presented as bar graphs and boxplots obtained with GraphPad Prism (https://www.graphpad.com). All statistical analysis was performed using GraphPad Prism. Two-tailed Student's t-test for unpaired data was used. For meristem analysis, two-way ANOVA followed by Tukey's pairwise multiple comparisons was performed. P-values greater than 0.05 were reported as not significant. For ovary/fruit indeterminacy analysis in Fig. 1J, we performed a Pearson χ² test (P<0.01) to compare whether the frequency of indeterminate fruits is independent of genotype in each of the pairwise comparisons with the WT. Degrees of freedom=1 (contingence table 2×2).

We thank the members of Dr Nogueira's laboratory for helpful discussions. We also thank the ESALQ/USP and John Innes Center (JIC) for providing Microscopy and Research facilities to develop this project.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.201961.reviewer-comments.pdf