Tomato facultative parthenocarpy results from SlAGAMOUS‐LIKE 6 loss of function

Summary The extreme sensitivity of the microsporogenesis process to moderately high or low temperatures is a major hindrance for tomato (Solanum lycopersicum) sexual reproduction and hence year‐round cropping. Consequently, breeding for parthenocarpy, namely, fertilization‐independent fruit set, is considered a valuable goal especially for maintaining sustainable agriculture in the face of global warming. A mutant capable of setting high‐quality seedless (parthenocarpic) fruit was found following a screen of EMS‐mutagenized tomato population for yielding under heat stress. Next‐generation sequencing followed by marker‐assisted mapping and CRISPR/Cas9 gene knockout confirmed that a mutation in SlAGAMOUS‐LIKE 6 (SlAGL6) was responsible for the parthenocarpic phenotype. The mutant is capable of fruit production under heat stress conditions that severely hamper fertilization‐dependent fruit set. Different from other tomato recessive monogenic mutants for parthenocarpy, Slagl6 mutations impose no homeotic changes, the seedless fruits are of normal weight and shape, pollen viability is unaffected, and sexual reproduction capacity is maintained, thus making Slagl6 an attractive gene for facultative parthenocarpy. The characteristics of the analysed mutant combined with the gene's mode of expression imply SlAGL6 as a key regulator of the transition between the state of ‘ovary arrest’ imposed towards anthesis and the fertilization‐triggered fruit set.


Introduction
Fruit development following fertilization is critical for the completion of the plant life cycle. In tomato, the ovary, which develops in concert with the rest of the flower organs (growth phase I, according to Gillaspy et al., 1993), ceases to undergo cell divisions shortly (1-2 days) before anthesis and enters an 'ovary arrest' state. Only if fertilization is successfully completed, a signal believed to be produced by the young embryo provokes the ovary to resume growth. This growth involves initially a phase of rapid cell division and expansion (designated phase II or 'fruit set') for 5-10 days (Bohner and Bangerth, 1988;Varga and Bruinsma, 1986), and subsequently (during phase III) growth is driven mainly by cell enlargement concomitant with nuclear polyploidization (Chevalier et al., 2014, and references therein). Once reaching full size, ripening processes initiate.
The default programme of fertilization-dependent fruit development ensures that resources are not wasted sustaining purposeless fruit development, whereas parthenocarpy, that is fertilization-independent seedless fruit development, is a counterproductive trait in all the sexually reproducing plant species. Sexual reproduction entails that the hypersensitivity of the microsporogenesis process, and the mature male gametes to moderately high or low temperatures, and to extreme humidity or light intensity (El Ahmadi and Stevens, 1979;Mesihovic et al., 2016;Picken, 1984;Sato et al., 2006), is a major hindrance for year-round fertilization-dependent tomato yielding. Consequently, breeding for parthenocarpy is considered a valuable goal especially in the context of maintaining sustainable agriculture in the face of global warming (Ariizumi et al., 2013;Gorguet et al., 2005;Ruan et al., 2012). Other advantages of parthenocarpy relate to consumers' preference of seedless over seeded fruits, improved fruit quality due to elevated content of total soluble solids (TSS) (Carmi et al., 2003;Casas Diaz et al., 1987;Falavigna et al., 1978;Ficcadenti et al., 1999) and saving of energy invested in separating the seeds from processed products.
Since tomato and other vegetables that could benefit from parthenocarpy are commonly propagated from seeds, hence only genetic sources for facultative parthenocarpy, where seeded fruits can develop following successful fertilization (Varoquaux et al., 2000), are of practical value. Presently, the most extensively characterized nontransgenic sources for facultative parthenocarpy in tomato are as follows: the three monogenic sources, pat (Beraldi et al., 2004;presumably a mutated Solyc03g120910, Selleri, 2011;Soressi and Salamini, 1975), procera (a mutated SlDELLA, Bassel et al., 2008) and entire (mutated SlAUX/IAA9, Mazzucato et al., 2015;Saito et al., 2011), all of which manifest undesired pleiotropic effects; and the three digenic sources, pat-2 (Hazra and Dutta, 2010;Vardy et al., 1989), IL5-1 and IVT-line 1 (Gorguet et al., 2008), all manifesting acceptable parthenocarpic phenotype; and the inferior oligogenic source pat-3/pat-4 (Nuez et al., 1986;Philouze and Maisonneuve, 1978). Despite the importance of this trait, exploitation of these mutants in breeding programmes is still rather limited. Some of them are associated with mild or severely undesirable pleiotropic effects (e.g. Ariizumi et al., 2013;Carrera et al., 2012;Lin et al., 1984;Mazzucato et al., 1998;Philouze, 1989). And introgression of a digenic source is much more laborious, especially since the identity of the genes underlying any of these three sources was not reported so far. Besides their applicative importance, parthenocarpic mutants are indispensable in the study of the mechanism underlying 'ovary arrest' at pre-anthesis and its fertilization-triggered release leading to fruit set.
In the present study, we demonstrate that mutated alleles of the MADS-box gene SlAGL6 enable tomato yielding under heat stress. The mutations confer facultative parthenocarpy manifested in the development of seedless fruits comparable in both weight and shape to wild-type (WT) seeded fruits and that without pleiotropic effects, thus making it a novel valuable source for parthenocarpy in tomato. The pivotal role of SlAGL6 in controlling the transition from the state of 'ovary arrest' to fertilization-triggered fruit set is also discussed.

Line 2012 is a new monogenic recessive mutant for parthenocarpy
A chemically, EMS-mutagenized M 2 population generated in the M82 cultivar (generated by J. Hirschenhoren and Y. Kapulnik, The Volcani Center, ARO) was screened for mutants yielding under extremely high temperatures ( Figure S3a), which prevented fertilization-dependent fruit set, as described in Data S1. Family No. 2012 included two plants that set high-quality parthenocarpic fruits with good jelly fill under these conditions, whereas the parental line set only tiny, hollow fruits, commonly dubbed 'nuts'.
One of these two plants served to pollinate emasculated flowers of M82 plant, which set seeded fruits. These BC 1 plants were not parthenocarpic. However, 7 out of 40 BC 1 F 2 progenies set seedless fruits under the extremely hot conditions prevailing in the late summer ( Figure S3b) when the parental line M82 managed to set tiny 'nuts' fruitlets only (Figure 1a vs. b). This indicated that the trait is governed by a single recessive mutation. Further, out of 30 plants from the same BC 1 F 2 population grown in the winter ( Figure S3c), in a nonheated glass house, nine set seedless fruits, whereas the rest of the siblings set no normal fruits, indicating that the mutation enables parthenocarpic fruit development also under temperatures too low to allow fertilization-dependent fruit set.

Mapping the 2012 mutation
To map the mutation, we chose to adopt the bulk segregation approach (Michelmore et al., 1991) and perform Illumina sequencing of two genomic libraries one coming from a bulk of 2012 BC 1 F 2 parthenocarpic plants and the other from their nonparthenocarpic siblings. For this purpose, BC 1 F 2 plants were grown again in the late summer when the day temperatures were high enough ( Figure S3d) to seriously damage microsporogenesis and hence prevented fertilization-dependent fruit development. Under these conditions, at the date of harvest, the parthenocarpic plants produced high-quality red parthenocarpic fruits with good jelly fill, whereas the nonparthenocarpic (NP) siblings produced only small distorted, green puffy 'nuts' fruits, which weight was significantly smaller than that of the red fruits harvested from their parthenocarpic siblings (Figure 1c).
The genomic library coming from a pool of 20 BC 1 F 2 plants characterized as 'clearly parthenocarpic' was designated '2012 library', and the other coming from a pool of 23 of their siblings characterized as 'clearly nonparthenocarpic' was designated 'NP (nonparthenocarpic) library'. Bioinformatics analysis of the sequenced libraries was performed as detailed in Data S1. This analysis pointed to a segment of 3.85 million nucleotides in chromosome 1, spanning between SL2.5ch01:85115654 and 88965277, as the likely location of the mutation (Table 1). The analysis revealed nine homozygous mutated single nucleotide polymorphisms (SNPs) in this region in the '2012 library' that were heterozygous in the 'NP library', that is only 19-37% of the reads in this library showed the mutated alleles, while the others included the expected wild-type (WT) allele.
To restrict the location of the mutation underlying the 2012 mutant, we tested, in two segregating populations, for cosegregation of SNPs dispersed along the chromosomal interval suggested as the mutation location (Table 1, column 2), with the parthenocarpic phenotype. Cosegregation with six SNPs was examined in a testcross (TC) population (designed to segregate in a 1 : 1 ratio of parthenocarpic and nonparthenocarpic plants) which was allowed to set fruit under heat stress ( Figure S3e). As shown in Table 2a, this analysis eliminated the candidacy of the mutations represented by SNP Nos. 1, 5 and 6 as they are not closely linked with parthenocarpy. To further zoom in on the location of the mutation, a 2012 BC 2 F 2 population was similarly analysed for cosegregation of the parthenocarpic phenotype with the mutated version of SNP Nos. 2 and 3. This analysis, which is summarized in  (Table 2b). Furthermore, the progenies (BC 2 F 3 ) of these three plants manifested clear parthenocarpy when allowed to set fruit in the winter, under suboptimal temperatures ( Figure S3g), suggesting that the nonparthenocarpic phenotype of their parents reflects the facultative nature of the mutation.
Together, these analyses strongly suggested the mutated SlAGL6 (represented by SNP No. 3, see Table 1), as the gene underlying the parthenocarpic mutation 2012. SlAGL6 encodes for a MADS-box protein belonging to the type II lineage MIKC C , subfamily AGL6, of the MADS-box transcription factor family (Smaczniak et al., 2012).

CRISPR/Cas9-mutated SlAGL6 confers parthenocarpy
To confirm that the mutated SlAGL6 is the gene underlying the 2012 parthenocarpy, CRISPR/Cas9 technology was exploited to knockout the SlAGL6 gene (Solyc01g093960). Synthetic gRNA was designed to target the second exon of SlAGL6 (Figure 2a). It was incorporated into a Cas9 expressing binary vector and transformed into tomato line MP-1.
Progenies of three R 0 plants designated sg1, sg4 and sg5 differing in the nature of their mutation (Data S1), but all leading to premature stop codon (Figure 2b), were chosen for analysis of their ability to set parthenocarpic fruits. R 1 plants genotyped as heterozygous (+/m) or homozygous (m/m) for mutated SlAGL6 were grown in a greenhouse, side by side with the parental line MP-1. Red fruits were picked, weighed and analysed for seeds bearing. As demonstrated in Figure 2c, heterozygous progenies produced seeded fruits only, whereas the progenies homozygous or bi-allelic for mutated versions of SlAGL6 set mostly parthenocarpic fruits, as well as a few underseeded fruits (bearing up to 10 seeds), and fruits containing more than 10 seeds (defined as seeded fruits). Thus similar to the 2012 mutant, they are manifesting facultative parthenocarpy. Besides being seedless, the parthenocarpic fruits are similar in shape and jelly fill to those of the parental line fruits or their heterozygous seeded siblings ( Figure S1).
The conclusive proof for allelism of the 2012 mutation and the CRISPR/Cas9-generated SlAGL6 mutants came from the parthenocarpic phenotype of most of the fruits produced on F 1 hybrid between sg1 plant homozygous for mutated SlAGL6 and 2012 plant homozygous for mutated SNP No. 3, whereas F 1 hybrid between MP-1 (+/+) and the same 2012 plant produced seeded fruits only (Figure 2d).
If the observed parthenocarpy is true ('vegetative'), namely requires no external trigger, it will set fruit under pollination restrictive conditions; if the mutant is 'stenospermocarpic', that is requires an external stimulus (provided, e.g., by damaged pollen or the young embryos before their abortion), it implies that this parthenocarpy still relies on pollination, even though an unsuccessful one (Varoquaux et al., 2000). To distinguish between the two, flowers were emasculated at pre-anthesis and tested for fruit set. As shown in Figure 3a, all the emasculated flowers of the R 1 heterozygous progenies and MP-1 aborted, whereas over 60% of the emasculated flowers of mutated homozygous progenies set fruits, testifying to the true vegetative nature of the Slagl6-induced parthenocarpy. In agreement, the enlarged ovules collected from fruits developed from emasculated flowers are similar in size and appearance to those collected from nonemasculated parthenocarpic fruits, both of which are substantially smaller than normal seeds (Figure 3b).

Mutated SlAGL6 maintains fruit weight
In many cases, parthenocarpy was claimed to reduce fruit size compared to that of seeded fruits. As a first step towards Fruit weight (g) a Figure 1 The 2012 mutation enables parthenocarpic tomato fruit development under heat stress. In BC 1 F 2 population segregating for the 2012 mutation, when yielding in the late summer: (a) parthenocarpic siblings set highquality seedless fruits, whereas (b) M82 plants set small distorted fruitlets. (c) Within the segregating BC 1 F 2 population, the weight of the fruit harvested from 23 parthenocarpic plants was significantly higher than that from 24 nonparthenocarpic siblings (t-test, P < 0.001).
assessing a possible penalty of the SlAGL6 mutation on yielding parameters, its effect on fruit weight was examined in three different populations/genetic backgrounds: (i) the segregating 2012 BC 2 F 2 population, in the background of the determinate cultivar M82, (ii) the segregating F 2 population derived from a cross between the big fruit semi-determinate cultivar Marmande and parthenocarpic 2012 plant, and (iii) R 1 progenies of the CRISPR/Cas9-derived line sg1 in the indeterminate MP-1 line background.
The effect of the mutation on fruit weight and yielding potential was assessed on the same 2012 BC 2 F 2 plants used to map the mutation (Table 2). This population grew under near-optimal temperatures for fertilization-dependent fruit setting (detailed in Data S1 and Figure S3f). Three and a half months after planting in the net house, all the red, breaker and mature green fruits were harvested from 11 to 14 plants carrying the following three alternative genotypes of SlAGL6: homozygous WT (+/+), heterozygous (+/m) or homozygous mutated (m/m) allele, as well as from eight M82 plants. As shown in Figure 4a, the yielding potential, expressed as the yield including mature green, breaker and red fruits, harvested from the plants homozygous for the mutated allele  does not differ from that of line M82, or the siblings either homozygous (+/+) or heterozygous (+/m) for the WT SlAGL6 allele. Further, the plants homozygous for the mutated allele were characterized by a profoundly earlier and more concentrated yielding, manifested in a significantly higher yield of red fruits at the date of harvest ( Figure 4b, and d vs. e). The average weight of the red fruits developed on the (m/ m) plants is significantly higher than that of the seeded red fruits harvested from the parental line M82, or its (+/+) and (+/m) siblings (Figure 4c).
In order to start estimating the potential of the 2012 mutation to support development of parthenocarpic fruits also when introduced into large fruit background, 2012 plant was crossed with the medium-large, multilocular fruit, semi-determinate open variety Marmande (www.rareseeds.com/marmande-tomato/), and the effect of the mutation on parthenocarpic fruit weight was estimated in the segregating F 2 population (described in Data S1). Fruit weight is governed by several QTLs (Tanksley, 2004), and in segregating F 2 population, the average fruit weight is always similar to that of the small fruit parent, with only small  portion of the progenies close in weight to that of the big fruit parent (Lippman and Tanksley, 2001;Perry, 1915). Thus, we examined whether, in this F 2 population, the mutation affects fruit weight and the tendency to generate parthenocarpic fruits close in size to the big fruit parent. Analysis performed on a small population including 48 F 2 plants grown in 5-L pots in a greenhouse indicated that among the 11 progenies homozygous for the mutated SlAGL6 (SNP No. 3), two produced seedless fruits of weight higher than that of the seeded fruits of the big fruit parent, and the average fruit weight of the mutated progenies was somewhat higher than in their nonmutated siblings (Figure S2a). Plants from the same F 2 population were also grown in the soil, in a net house, next to the BC 2 F 2 population described above. Fruits were picked and analysed from the 26 homozygous mutated plants. As shown in Figure S2b, c, two out the five plants with the highest average fruit weight showed a very strong parthenocarpic phenotype, as nearly all of their fruits were seedless. These analyses indicate that the mutation does not prevent the parthenocarpic F 2 progenies from reaching weight similar to that of the big fruit parent.
Surprisingly, 12 out of the 13 tested R 1 progenies of plant sg1 were found to be devoid of the transgenic cassette, while all the tested progenies of plants sg4 and sg5 contained it. Absence of the transgene enabled to assess in sg1 R 1 progenies the effect of the SlAGL6 mutation per se, on fruit weight. As shown in Figure 4f, the weight of seedless fruits harvested from sg1 R 1 progenies homozygous for the mutation was comparable to that of seeded fruits developed on MP-1 plants grown side by side under the same ambient conditions.

Slagl6 improves yielding under heat stress
Yielding under natural heat stress conditions was examined comparing MP-1 and sg1 line homozygous for Slagl6 and devoid of the Cas9 cassette (at R 2 ). Plants were planted in a net house on 20 April 2016 and the first harvest was performed 67 days later (as detailed in Data S1). During the months of May and June, the day temperatures were very high including a 3 days long spell (between 14 and 16 of May 2016) of extremely high temperatures (maximum day temperature 38°C and above, Figure S3h). These naturally occurring heat stress conditions fall under the definition of 'chronic mild heat stress' known to hamper microsporogenesis and hence fertilization-dependent fruit set (see Mesihovic et al., 2016; and references therein). As demonstrated in Figure 5a, under these climatic conditions, the red fruit yield of the parental line MP-1 was significantly lower than that of line sg1 (ca. 85% lower). This difference reflects mainly a dramatic difference in the number of fruits produced (Figure 5b, f, g), which was 83% lower in line MP-1, and also a significant lower fruit weight in the latter (Figure 5c Casas Diaz et al., 1987), the TSS content, expressed as Brix, of red ripe seedless fruits was significantly higher than that of seeded fruits of MP-1 (Figure 5d), while the acidity (pH) of the fruits remained similar (Figure 5e).

SlAGL6 mode of expression
To gain additional insight into the function of SlAGL6, its expression was queried in publically available data and complemented by quantitative RT-PCR of developing ovaries. According to the Expression Atlas of Tomato Tissues (http://tomatolab.cshl.edu/~lippmanlab2/allexp_query.html, Park et al., 2012;Tomato Genome Consortium 2012), SlAGL6 is highly expressed only in the flower meristem, flower bud and open flower, yet it sharply declines in the developing fruit ( Figure S4a). This explains why the Slagl6 mutation does not affect vegetative development, transition to reproductive stage, pollen viability, or fruit size and shape (Figure 6a-c, e-n). However, despite the reported expression in the flower meristem, the flower bud and the flower (Figure S4a), the only subtle difference noticed in the flowers is that the petals are paler and somewhat narrower and longer than in the WT (Figure 6d).
To examine SlAGL6 mode of expression during growth phase I, its expression was quantified in developing ovaries of line M82. As demonstrated in Figure  (a) Yield potential, defined as the yield (kg/plant) of all fruits reaching at least the mature green stage, is not affected by the mutation (one-way analysis of variance, P = 0.75); (b) compared to M82, only siblings homozygous for the mutation produce significantly (t-test, P < 0.001) higher yield of marketable red fruit; (c) the weight of their red fruits is significantly (t-test, P < 0.05) higher than that of M82 At the date of harvest; (d) in M82, most of the harvested fruits were still green, while (e) in (m/m), plant most of them were red ripe; and (f) parthenocarpic fruits of line sg1 plants homozygous for the mutation (at R 1 generation) are comparable in weight to seeded MP-1 fruits (t-test, P = 0.946). In panels a, b, c and f, columns accompanied by different lowercase letters differ significantly.
developing ovaries, and it peaks towards the stage of 'ovary arrest' (10-mm-long buds correspond to pre-anthesis) and remains high at anthesis. Yet 5DPA it sharply declines to the level found in ovaries of the young 4-mm-long flower buds, hence associating 'ovary arrest' with elevated SlAGL6 expression and fruit set with its decline. Genetic variation between M82 and Micro Tom or Solanum pimpinellifolium might explain the higher fold decline between anthesis and 4DPA reported before (Pattison et al., 2015;Tang et al., 2015;Wang et al., 2009), whereas smaller variation between M82 and Moneymaker could account for the observed similarity in fold decline (Figure 7 vs. Figure S4d, queried from Zhang et al., 2016). Evidence for preferential sublocalization of the SlAGL6 transcript within the ovules of the mature arrested ovary was provided by Pattison et al. (2015). Following transcriptomic analysis of lasercaptured tissues, they show that SlAGL6 level in the ovules is at least fourfold higher than in the other tissues comprising the ovary. However, at 4DPA, its expression in the embryo is already 15-fold lower than in the ovule (Fig. S4c). This finding was further corroborated by Zhang et al. (2016) (Fig. S4d).

Loss of SlAGL6 function results in facultative parthenocarpy that ensures fruit production under high temperatures
After isolating the EMS-induced 2012 parthenocarpic mutant (Figure 1a vs. b), next-generation sequencing (Table 1) followed by marker-assisted mapping (Table 2) and CRISPR/Cas9 gene knockout (Figure 2) confirmed that mutated SlAGAMOUS-LIKE 6 confers facultative parthenocarpy in tomato.
To fully determine whether mutated SlAGL6 imposes any unacceptable penalty on yielding potential or fruit characteristics, it waits to be introduced as a single mutation into elite tomato cultivars and tested when grown under ambient environmental conditions and established horticultural practices. Nonetheless, in three different genetic backgrounds, the weight of Slagl6 seedless fruits was comparable or even higher than that of the seeded WT ones (Figures 4c, f, 5c, Figures S1 and S2), and yielding potential was comparable to that of the parental cultivar M82 (Figure 4a).

MP-1
Yield red fruit (kg/plant) The exact conditions favouring seed setting in the mutated plants under fertilization permissive conditions remain to be elucidated. Vigorous inflorescence vibration resulted in many seeded fruits from plants that otherwise set mainly seedless ones (data not shown). In the absence of intentional vibration, enhanced tendency was clearly associated with two parameters: first, the small fruits developed on old plants, frequently bear seeds. This is an unusual phenomenon, as it was shown that seeded tomato fruits are larger than underseeded ones (Carmi et al., 2003;Imanshi and Hiura, 1975;Varga and Bruinsma, 1976). Second, fruits that set at temperatures mildly lower than optimal were frequently found to contain seeds. In both cases, seed production presumably reflects conditions slowing the rate of ovary expansion into fruit, thus allowing the MP-1 MP-1 sg1-8 sg1-8  The pericarp of parthenocarpic fruit of line sg1-8 (g-j) is similar to that of seeded fruit of MP-1 (k-n) of similar weight (ca. 23.6 g) and size (g, k scale = 2 cm). (h, l) The pericarp is of similar width and shape, scale = 1000 lm. The cells in the layers between the vascular bundle rim and the exodermis (the exocarp) (i, m) and those between the vascular bundle rim and the endodermis (endocarp) (j, n) are of similar appearance. In i, j, m, n, scale = 500 lm. Photographs (h-j) and (l-n) are of thin freehand transverse sections taken from the middle (equator) of the nearly mature green fruits presented in g and k, and photographed under light microscope. Line sg1-8 is described in legend to Figure 5. Ex, exodermis; En, endodermis; Vb, vascular bundle. pollen grains to complete germination, elongation and fertilization of ovules before the style is detached from the otherwise rapidly expanding ovary/fruit. The genetic background apparently affects the facultative manifestation as well: while in 50% (13/26) of the Slagl6/Slagl6 F 2 progenies of the 2012 9 Marmande hybrid, grown under ambient conditions, 8/8 fruits tested per plants were seeded ( Figure S2b), only 2.4% (3/126) of the Slagl6/Slagl6 plants of the 2012 BC 2 F 2 population grown side by side were completely seeded (see Table 2b). Although not tested during the relevant flowering period, genetic differences in anther dehiscence under the given humidity conditions, the rate of ovary enlargement pre-and postanthesis and/or the duration of stigma receptivity could be among the factors underlying the observed difference between these genetic backgrounds. The ability of the Slagl6-induced parthenocarpy to solve the problem of yielding under fertilization restrictive conditions such as imposed by chronic mild heat stress (Mesihovic et al., 2016) is its most important agronomic attribute. When challenged by continuous mild heat stress, which was worsen by a 3-day spell of acute stress (Figure S3f), the mutated line yielded over sixfold higher than the parental line ( Figure 5a) and that mainly because of the profoundly higher number of flowers that set (seedless) fruits under these fertilization restrictive conditions (Figure 5b, f,  g). The demonstrated capability of Slagl6 to yield under microsporogenesis restrictive conditions (Sato et al., 2006), together with the ripening uniformity along consecutive trusses exhibited in the M82 determinate background (Figure 4b, d-e), makes Slagl6 particularly suitable for breeding of processing tomato cultivars. This is because under fluctuating climatic conditions, it maximizes the marketable yield that can be obtained in a single mechanical harvest.

Slagl6 is an attractive gene for parthenocarpy
Three digenic sources for parthenocarpy which manifest only mild or no adverse pleiotropic effects and hence of practical value are as follows: pat-2 (Hazra and Dutta, 2010;Vardy et al., 1989), IL5-1 and IVT-line 1 (Gorguet et al., 2008) (as reviewed by Ariizumi et al., 2013). A reliable comparative assessment of the horticultural/parthenocarpic performance of these three sources with that of the Slagl6 mutation waits the identification of the mutated genes underlying these sources and the introducing of all of them to a common genetic background. Nonetheless, different from them, Slagl6 is a single recessive source for facultative parthenocarpy, which is not allelic to any of these three, since none of them was suggested to map to chromosome 1 (Gorguet et al., 2008;Nunome et al., 2014). Taken together, the simple mode of inheritance, the lack of pleiotropic effects ( Figure 6) or adverse effects on fruit weight or shape (Figures 4c, f, 5b, 6g-n, and Figures S1 and S2), the true vegetative nature of the induced parthenocarpy ( Figure 3) and its facultative manifestation make Slagl6 an attractive single recessive gene for parthenocarpy. The CRISPR/Cas9 technology now enables expeditious integration of Slagl6 into any elite cultivar of interest. That can be done either by direct CRISPR/ Cas9-mediated SlAGL6 knockout in the two parents of the elite hybrid cultivars, if they are amenable for transformation, or introgressed by backcrossing.

SlAGL6 functions before and after fertilization
Comprehensive transcriptomic studies of tomato fruit set found significant changes in the expression of hundreds of genes, including several down-regulated MADS-box genes (e.g. Pattison et al., 2015;Tang et al., 2015;Wang et al., 2009;Zhang et al., 2016). Among them, SlAG1 and SlAGL6 were suggested to play an important role in fruit set (Wang et al., 2009), and so was also the tomato MADS-box 29 (TM29) (Ruan et al., 2012). However, experimentally supported identification of any transcription factor actually involved in the regulation of 'ovary arrest' unless it is fertilized is still missing (Ruan et al., 2012). The severe homeotic malformations of the seedless fruit developed following silencing TM29 (Ampomah-Dwamena et al., 2002) or SlAG1 (Gimenez et al., 2016;Pan et al., 2010;Pnueli et al., 1994) indicate a role in stamen and carpel development rather than as activators of the normal fertilization-triggered fruit set, and experimental support for the SlAGL6 suggested function was not provided.
The studied SlAGL6 mutants together with relevant transcriptomic analyses identify SlAGL6 as a central player in the mechanism underpinning fertilization-dependent fruit development. The presented data strongly indicate that the principle, if not the sole, role of SlAGL6 is to serve as a key regulator of the transition between the 'ovary arrest' state reached at the end of growth phase I and the fertilization-triggered resumption of growth, that is fruit set. The information gathered in the current study led us to propose a model (Figure 8) according to which SlAGL6 accumulation at pre-anthesis, most likely in the maturing ovules, plays a pivotal role in the induction or at least retention of 'ovary arrest'. At anthesis, successful ovule fertilization signals for SlAGL6 down-regulation, and once its suppressive effect is alleviated, the ovary/fruit growth is resumed and continues to its full growth potential.
The reasoning behind the proposed model is as follows: first, although according to the Expression Atlas of Tomato Tissues (http://tomatolab.cshl.edu/~lippmanlab2/allexp_query.html; Park  Figure S4a); besides the subtle change in the petal hue and shape, the Slagl6 flowers are normal, and both male and female fertile (Figure 6d, e-f). This indicates that different from some other species where an E-function was attributed to the AGL6 clade (Dreni and Zhang, 2016), SlAGL6 is dispensable in determining the identity of any of the flower whorls according to the extended ABC(DE) and the quartet models (Coen and Meyerowitz, 1991;Theissen, 2001;Theissen and Saedler, 2001). Second, quantification of SlAGL6 transcripts in developing ovaries of line M82 demonstrates that it is not highly expressed throughout growth phase I (Figure 7), but rather gradually elevates and peaks at the stage when the ovary growth is arrested. However, 5DPA it already declines to the low level found in ovaries of the young 4-mm-long flower buds. Although our analysis was performed on whole ovaries, based on the demonstrated preferential expression of SlAGL6 in the ovules at anthesis ( Figure S4c, d, queried from Pattison et al., 2015;Zhang et al., 2016), it is reasonable to assume that the observed peak at pre-anthesis represents mainly increased expression in the ovules of ovaries which growth is arrested. The lower level of the transcript in the younger growing ovaries, together with the Slagl6 parthenocarpy, strongly suggests that SlAGL6 accumulation in the mature ovules acts as a key suppressor of ovary growth beyond anthesis unless it is fertilized, thus preventing accidental development of unfertilized ovary into purposeless fruit. Lastly, the normal size and shape of the seedless fruits indicate that development of the fruit to its full growth potential relies predominantly on the removal of the SlAGL6 suppressive signal (either following fertilization or by mutation), rather than on promoting signals emitted by the developing embryos independent of SlAGL6 down-regulation. Continuously emitted SlAGL6 suppressive signal from unfertilized ovules within the developing WT fruit (as suggested by Tang et al., 2015; see Figure S4b) could explain the often observed restricted development of underfertilized fruits (Carmi et al., 2003;Imanshi and Hiura, 1975;Varga and Bruinsma, 1976). This assumption is further supported by the significantly higher weight of the Slagl6 parthenocarpic fruits in the segregating 2012 BC 2 F 2 population compared to that of the seeded fruits of the parental line M82 (Figure 4c), as well as that of the seedless sg1 fruits compared to the MP-1 seeded ones when developed under heat stress (Figures 5c). In both cases, not all of the ovules are necessarily fertilized in the seeded WT fruits.

Neofunctionalization of AGL6 in tomato
It is noteworthy that in apple (Malus domestica), where the fleshy fruit is a pome derived from the floral tube fused to the carpels, parthenocarpy was found to be governed by mutated MdPI, the homolog of Arabidopsis PISTILLATA (Yao et al., 2001), which belongs to the DEF/GLO rather than the AGL6 clade of MADS-box genes (Smaczniak et al., 2012). Alternatively, parthenocarpy was not reported for any of the mutated SlAGL6 homologs characterized so far. In rice (Oryza sativa), the homeotic changes manifested by mutated OsMADS6 testify to its role in determining floral organ and meristem identities (e.g. Duan et al., 2012;Li et al., 2011;Ohmori et al., 2009;Zhang et al., 2010). Knocking down its Nigella damascene homolog affected structure of sepals and petals indicating an A-function (Wang et al., 2015). The Arabidopsis thaliana homolog (At2g45650) affects flowering time and axillary bud formation (Huang et al., 2012;Koo et al., 2010;Yoo et al., 2011). Interestingly, similar to Slagl6, the mutated Petunia hybrida homolog, Phagl6, caused only subtle effect on petals colour and indentation. Yet, although similar to tomato and Arabidopsis , PhAGL6 is highly expressed in the mature ovules, in this dry capsule fruit species, as in Arabidopsis, parthenocarpy was not reported (Rijpkema et al., 2009). Thus similar to other MADS-box genes that underwent neofunctionalization (Dreni and Kater, 2014;Scutt et al., 2006;Smaczniak et al., 2012;Zahn et al., 2006), in tomato, and presumably in other fleshy Solanaceae fruits, AGL6 acquired a new function, acting as the suppressor of ovary development beyond anthesis. Further identification of the signals and regulators involved in SlAGL6 downregulation and its immediate targets could point to additional candidate genes for parthenocarpy.

Experimental procedures
Populations for mapping the 2012 mutant Several populations were analysed, including 2012 BC 1 F 2 , a testcross (TC) population, 2012 BC 2 F 2 and F 2 population derived from 2012 9 Marmande cross. Generation of these populations, their growth conditions and parameters analysed in each of them are detailed in Data S1 and in Figure S3.

SNP genotyping of 2012 derived progenies
Genotyping was performed as a service by DYN R&D Ltd (Migdal Haemek, Israel), following the melting curve SNP method (Ye et al., 2002). Plants were genotyped in duplicate, on two separately sampled leaves.

Genomic DNA libraries, sequencing and bioinformatics analysis
From each of 20 plants derived from 2012 BC 1 F 2 population, defined as 'strong parthenocarpic', one young leaf (ca. 150 mg FW) was picked, and from the pooled leaves, DNA was extract. Similarly, DNA was extracted from a pool of leaves sampled at the same date from 23 'nonparthenocarpic' plants.

Construction of a CRISPR/Cas9 knockout plasmid and tomato transformation
The CRISPR/Cas9 construct was designed to create a deletion/ insertion after 212 bp of the Solyc01g093960 coding sequence (predicted exon 2, see Figure 2b). The 20-bp target sequence was chosen to be followed by protospacer adjacent motif (PAM), the requisite binding site for Cas9, TGG (depicted in Figure 2b). The selected sgRNA was amplified using the primers: SalI-gRNA-F: AGAgtcgacATAGCGATTGAGGATTAAGGCAACAACGTGTTTTAG AGCTAGAAATAGCAAG and HindIII-gRNA-R: TAAGCTaagcttC GATCTAAAAAAAGCACCGACT (the added restriction sites are presented in italics lowercase letters, and the specific target sequence (cRNA) is underlined in the SalI-gRNA-F primer sequence). The PCR product was restricted and cloned into the pRCS binary vector SalI-HindIII sites as previously described (Chandrasekaran et al., 2016). The binary vector was transformed to Agrobacterium tumefaciens strain EHA105, which served to transform the indeterminate tomato line MP-1 as previously described (Barg et al., 1997).

Detection of CRISPR/Cas9-induced mutations
The Solyc01g093960 gRNA target site was designed to include AclI restriction enzyme site overlapping three bp upstream from the PAM (see Figure 2a), the predicted cut site of the Cas9 nuclease, so that DNA double-strand break repair could disrupt the restriction site. R o plants were screened for the presence of chimeric section as detailed in Data S1, and the same procedure was used to genotype R 1 progenies. The presence of the Cas9/sgRNA cassette in the progenies was PCR-tested.

Quantitative RT-PCR analysis of SlAGL6
Analysis was performed on total RNA samples extracted from developing ovaries and young fruit collected from line M82 plants, as earlier described (Damodharan et al., 2016).

Pollen viability assay
Freshly harvested pollen grains were incubated in germination solution (Firon et al., 2012) for 18 h, at 24°C, and examined under light microscope. All primers used in this study are listed in Table S1.

Supporting information
Additional Supporting Information may be found online in the supporting information tab for this article: Figure S1. Parthenocarpic fruits of R 1 plants homozygous for three differently CRISPR/Cas9 mutated SlAGL6 alleles. Figure S2. Manifestation of the 2012 mutation in a larger fruit background. Figure S3. Daily maximum and minimum temperatures experienced by the plants in the various experiments. Figure S4. Reported modes of expression of SlAGL6. Table S1. Primers used in this study. Data S1. Methods.