Expression of Aberrant Forms of AUXIN RESPONSE FACTOR8 Stimulates Parthenocarpy in Arabidopsis and Tomato

Fruit initiation in Arabidopsis ( Arabidopsis thaliana ) is generally repressed until fertilization occurs. However, mutations in AUXIN RESPONSE FACTOR8 ( ARF8 ) uncouple fruit initiation from fertilization, resulting in the formation of seedless, parthenocarpic fruit. Here we induced parthenocarpy in wild-type Arabidopsis by introducing either the mutant genomic (g) Atarf8 - 4 sequence or g AtARF8 : b -glucuronidase translational fusion constructs by plant transformation. Silencing of endogenous AtARF8 transcription was not observed, indicating that the introduced, aberrant ARF8 transcripts were compromising the function of endogenous ARF8 and/or associated factors involved in suppressing fruit initiation. To analyze the role of ARF8 in tomato ( Solanum lycopersicum ) we initially emasculated 23 tomato cultivars to test for background parthenocarpy. Surprisingly, all had a predisposition to initiate fertilization-independent fruit growth. Expression of g Atarf8 - 4 in transgenic tomato (‘Monalbo’) resulted in a signiﬁcant increase in the number and size of parthenocarpic fruit. Isolation of tomato ARF8 cDNA indicated signiﬁcant sequence conservation with AtARF8. SlARF8 may therefore control tomato fruit initiation in a similar manner as AtARF8 does in Arabidopsis. Two SlARF8 cDNAs differing in size by 5 bp were found, both arising from the same gene. The smaller cDNA is a splice variant and is also present in Arabidopsis. We propose that low endogenous levels of the splice variant products might interfere with efﬁcient formation/function of a complex repressing fruit initiation, thereby providing an explanation for the observed ovary expansion in tomato and also Arabidopsis after emasculation. Increasing the levels of aberrant Atarf8 - 4 transcripts may further destabilize formation/function of the complex in a dosage-dependent manner enhancing tomato parthenocarpic fruit initiation frequency and size and mimicking the parthenocarpic dehiscent silique phenotype found in homozygous Atarf8 - 4 mutants. Collectively these data suggest that similar mechanisms involving auxin signaling exist to inhibit parthenocarpic fruit set in tomato and Arabidopsis.

Signaling processes, initiated by pollination and fertilization, are normally required to initiate seed and fruit development (Raghavan, 2003). Subsequent fruit growth depends, in part, on the coordinated action of growth substances produced in the ovary and seed after pollination and/or fertilization (Gillaspy et al., 1993;Garcia-Martinez and Hedden, 1997). Various phytohormones, including auxin, cytokinin, and GAs, have been implicated in seed and fruit growth and development (Nitsch, 1952(Nitsch, , 1970Coombe, 1960;Garcia-Martinez and Hedden, 1997;Fos et al., 2000Fos et al., , 2001.
The effect of pollination and fertilization in stimulating fruit growth can be mimicked by hormone applications (Goodwin, 1978;Gillaspy et al., 1993;Vivian-Smith and Koltunow, 1999) or by expression of auxin biosynthesis genes in ovaries and ovules (Rotino et al., 1997;Ficcadenti et al., 1999;Carmi et al., 2003;Mezzetti et al., 2004) that lead to the formation of fertilizationindependent or parthenocarpic (seedless) fruit. Elevated or altered levels of phytohormones have also been observed during fruit growth in naturally occurring parthenocarpic plants (Mapelli et al., 1978;George et al., 1984;Talon et al., 1990Talon et al., , 1992. Recent studies in tomato (Solanum lycopersicum) and Arabidopsis (Arabidopsis thaliana) have revealed that the auxin signaling pathway is involved in controlling the early events of fruit initiation. Two components of the auxin signaling pathway, AUXIN RESPONSE FACTOR8 (ARF8) from Arabidopsis Goetz et al., 2006) and the Aux/IAA protein IAA9 from tomato (Wang et al., 2005), have been implicated in repressing fruit initiation in the absence of the fertilization cue. Antisense repression of IAA9 in tomato and recessive mutations in Arabidopsis ARF8 uncoupled fruit initiation from pollination and fertilization and gave rise to parthenocarpic fruit (Wang et al., 2005;Goetz et al., 2006). Aux/IAA proteins can bind to ARF proteins to activate or inhibit the transcription of auxin responsive genes (Ulmasov et al., 1999b;Hardtke et al., 2004;Tatematsu et al., 2004). It has been proposed that both Arabidopsis and tomato possess ARF8-and IAA9-like orthologs that interact and, together with potentially other as yet unknown proteins, form a protein complex that prevents fruit set prior to fertilization (Goetz et al., 2006;Swain and Koltunow, 2006).
Parthenocarpy in mutant and wild-type Arabidopsis has been examined by emasculation, which involves removing all of the floral organs surrounding the carpel before the anthers open and shed pollen onto the stigma (Vivian-Smith and Koltunow, 1999). Emasculation of wild-type Arabidopsis flowers in Landsberg erecta (Ler), Columbia (Col), and Wassiljewska (Ws) ecotypes results in a slight increase in carpel length and girth from cellular expansion but a dehiscence zone indicative of a differentiated fruit does not form (Fig. 1;Vivian-Smith and Koltunow, 1999). Recessive mutations arising from lesions in the coding region of AtARF8, such as those found in Atarf8-1 and Atarf8-4 mutant alleles, induce parthenocarpic silique development whereby carpel growth continues further than the initial growth that is observed in emasculated wildtype plants and differentiation of a seedless, dehiscent silique occurs. Atarf8-1 is a T-DNA insertion mutant and Atarf8-4 contains a mutation in the putative translation initiation codon. Both alleles produce transcripts that have the potential to encode truncated ARF8 proteins and it has been suggested that this might explain why full complementation of the mutant phenotypes is not achieved when wild-type genomic ARF8 sequences are introduced (Tian et al., 2004;Goetz et al., 2006). Analysis of the expression of a genomic wildtype AtARF8:GUS translational fusion in transgenic Ler 1 AtARF8:GUS plants showed that parthenocarpic phenotypes can be induced in some transgenic lines, suggesting that the altered ARF8 protein can have dominant-negative effects. Furthermore, the expression of the transgenically introduced AtARF8:GUS translational fusion was temporally and spatially altered in the arf8-4 mutant background relative to the expression pattern in the wild-type plants. This may relate to possible interactions between the gene products or to indirect effects on gene expression relating to the mutant arf8 background (Goetz et al., 2006). These results led us to propose that a putative complex repressing fruit initiation in Arabidopsis could be destabilized in at least two ways. Mutations in ARF8 may prevent the production of a functional protein so that the complex cannot form. Alternatively, aberrant ARF8 transcripts or proteins might be formed that disrupt complex formation or function via some form of dosage-dependent competitive interference and thus stimulate parthenocarpy (Goetz et al., 2006).
In this article, we directly examined whether the introduction of aberrant forms of ARF8 can induce parthenocarpic fruit initiation in Arabidopsis and tomato. These data together with the isolation and analysis of tomato ARF8 sequences suggest that ARF8 is also involved in the regulation of fruit initiation in tomato.

Induction of Parthenocarpy in Wild-Type Arabidopsis by gAtarf8-4
Preliminary results indicated that the Atarf8-4 gene can induce parthenocarpy in Arabidopsis (Vivian-Smith, 2001). To examine this further, the Atarf8-4 gene sequence was introduced into wild-type Ler plants by plant transformation. A total of 10 independent, homozygous lines containing one to three copies of the Atarf8-4 gene were recovered and analyzed. Carpel elongation after emasculation was observed in nine out of the 10 lines and formation of the dehiscence zone occurred in the three out of those 10 lines that also exhibited greatest elongation (Table I). These three lines also displayed other typical Atarf8-4 mutant phenotypes because the siliques were broad shouldered in appearance ( Fig. 1), reduced seed set was evident after fertilization in the basal part of the silique, and the carpels frequently protruded from unopened flower buds (data not shown). The carpels from the remaining seven lines elongated slightly after emasculation but otherwise did not show additional phenotypes related to Atarf8-4.
The point mutation in Atarf8-4 creates a cleaved amplified polymorphic sequence (CAPS) marker (Vivian-Smith, 2001) and expression of the introduced mutant gene was confirmed by using the CAPS marker in reverse transcription (RT)-PCR assays and also by quantitative real-time PCR. A mixture of mutant Atarf8-4 and wild-type AtARF8 transcripts was observed using the CAPS marker in all transformants (data not shown) and levels of total AtARF8 transcript were also found to be higher in transgenic plants compared to wild-type plants (Table I), indicating that cosuppression or RNA silencing of the endogenous AtARF8 gene were not the cause of the induced parthenocarpic phenotype. These data indicate that expression of arf8-4 mutant genes seem to induce variable silique elongation and dehiscence zone formation responses. Although it is not possible to determine the exact ratios of expression levels between the mutant and wild-type mRNAs from these experiments, the observed variation in the extent of parthenocarpy is probably due to differences in expression levels of the introduced Atarf8-4 copies. Variable responses, as observed in the transgenic lines here, are expected if competitive interference occurs between the introduced aberrant Atarf8-4 products and the endogenous wild-type ARF8.

Expression of a gAtarf8-4:GUS Translational Fusion Construct Induces Parthenocarpy
The genomic Atarf8-4 mutant sequence was translationally fused to GUS and introduced to Ler and arf8-4 Arabidopsis backgrounds by plant transformation to examine if a protein product is made. A total of eight independent, homozygous lines were generated in the Ler background and four independent, homozygous lines in the arf8-4 background and all the lines expressed GUS (Fig. 2). Since the GUS gene does not contain its own translation initiation codon, the detection of GUS expression indicates that translation of the fusion protein is initiated from an alternative start codon somewhere within the mutant Atarf8-4 gene. Consequently a protein fragment can be produced from the mutant allele despite the point mutation of the putative translation initiation codon of AtARF8 in Atarf8-4. The exact nature of this mutant protein is, however, unclear and remains to be determined in future experiments.
Interestingly, the spatial and temporal expression patterns of the translational gAtarf8-4:GUS construct in both the Ler and arf8-4 backgrounds during flower and seed initiation were identical to those previously published for gAtARF8:GUS in the Ler and arf8-4 backgrounds (compare Fig. 2 with Goetz et al., 2006). As for gAtARF8:GUS, expression of gAtarf8-4:GUS showed reduced expression levels in arf8-4 plants in some tissues (e.g. petals and anther filaments) and a temporal change in expression was observed, especially in the carpel walls, the septum, the funiculi, and the ovules, where GUS activity was detected earlier and persisted in the mutant background. A detailed description of the expression patterns is provided in Goetz et al. (2006). Thus, as for the gAtARF8:GUS translational construct, expression of the gAtarf8-4:GUS translational fusion is spatially and temporally altered in the arf8-4 background compared to the expression of the same constructs in the Ler background ( Fig. 2), indicating that the regulation of their expression is impaired in arf8-4. Therefore the arf8-4 mutant genetic background influences the spatial and temporal expression pattern of the introduced genes.
We previously observed that a dehiscence zone formed in some of the emasculated gAtARF8:GUS plants examined at that time (Goetz et al., 2006). In this study we emasculated a total of seven independent, homozygous lines transformed with the gAtARF8:GUS construct and five independent, homozygous lines transformed with gAtarf8-4:GUS construct to examine their efficiency in inducing parthenocarpy. All lines showed the formation of a dehiscence zone, an indicator for fruit differentiation, and silique elongation was observed in five out of seven Ler 1 gAtARF8:GUS lines and in four out of five Ler 1 gAtarf8-4:GUS lines with the greater silique elongation observed in the gAtARF8:GUS lines ( Figs. 1 and 3). Levels of endogenous AtARF8 mRNA did not detectably decrease in transgenic plants (data 3.6 mm 6 0.1 2 2 2.6 6 0.0 Ler 1 arf8-4 no. 4 4.6 mm 6 0.3 3 1 3.0 6 0.2 Ler 1 arf8-4 no. 5 3.8 mm 6 0.2 3 2 4.2 6 0.4 Ler 1 arf8-4 no. 6 5.9 mm 6 0.2 2 1 2.8 6 0.0 Ler 1 arf8-4 no. 7 3.6 mm 6 0.2 2 2 n.d. Ler 1 arf8-4 no. 8 3.5 mm 6 0.2 2 2 n.d. Ler 1 arf8-4 no. 9 3.8 mm 6 0.2 3 2 n.d. Ler 1 arf8-4 no. 10 6.1 mm 6 0.1 3 1 3.5 6 0.3 not shown) therefore RNA silencing is unlikely to be the cause of the phenotype. These data show that expression of the translational gAtARF8:GUS and gAtarf8-4:GUS fusion constructs induce dehiscence zone formation but variable silique elongation responses. These data together with those obtained from plants transformed with the Atarf8-4 mutant gene indicate that the introduced aberrant constructs are able to destabilize the function of endogenous ARF8 and/or interacting factors that restrict fertilizationdependent fruit initiation in some way, leading to the induction of parthenocarpic phenotypes.

Parthenocarpy Occurs in Many Commercial Tomato Cultivars
Arabidopsis produces a dry, dehiscent fruit. To examine the role of ARF8 during the initiation of fleshy fruit growth we selected tomato for further experiments. We emasculated a total of 23 different cultivars to examine their potential for parthenocarpic fruit initiation (see ''Materials and Methods''). All of the cultivars tested exhibited some ovary growth after emasculation and most were able to initiate seedless fruit for-mation to varying degrees where the parthenocarpic fruits colored up and ripened. Data concerning eight indicative cultivars is provided in Table II.
The extent of fruit set observed varied between the different tomato cultivars in terms of the percentage of emasculated buds that initiated fruit development, final fruit size, and in particular the time lag between emasculation and fruit initiation. Naturally smaller fruited cultivars were most efficient in setting parthenocarpic fruit ( Fig. 4A; Table II). In general, the size of ripe parthenocarpic fruit was smaller than fertilizationinduced fruit size (Table II), although we did not carry out an exhaustive analysis in this regard. In cultivars with poor fruit initiation, many emasculated flowers abscised at the abscission zone in the pedicel. If the emasculated flowers were retained some ovary expansion and fruit development was generally observed with very slow growth or early arrest in the varieties with poor fruit initiation (Fig. 4B). In some cultivars the parthenocarpic fruit had a different shape (Fig. 4C) and pulp formed better in the seedless fruit of some cultivars compared to others. Emasculated flowers of 'Monalbo' took between 63 and 143 d to form red, ripe parthenocarpic fruit with an average development GUS at anthesis (stage 13; A) and after fertilization (stage 15; B). C and D, GUS staining patterns in arf8-4 flowers and siliques transformed with gAtarf8-4:GUS at anthesis (stage 13; C) and after fertilization (stage 15; D). Goetz et al. (2006) contains a detailed description of the expression patterns during flower and fruit development in Ler and arf8-4 mutants containing the gAtARF8:GUS translational fusion for comparison with the figures presented here. The numbers at the bottom left indicate the stages of flower development (Smyth et al., 1990). a, Anther; gy, gynoecium; pd, pedicel; pe, petals; se, sepals; sp, septum; si, silique; st, stigma. Bars 5 1 mm. time of 90 d (618.4 d). When multiple flowers on a single truss were emasculated on the same day, different rates of fruit initiation and growth were detected on nine individual trusses observed. Figure 4D shows an example of such a truss from 'Sweet Cherry Gold' where a ripe fruit is senescing and different sized fruits have developed to varying degrees. Collectively, these data demonstrate that in most tomato cultivars examined after emasculation some limited ovary growth up to full fruit development occurs.
'Monalbo' was chosen for subsequent experiments. It has been selfed for many generations and is frequently used experimentally. It has moderate parthenocarpic ability and there was a substantial time difference between the initiation of fertilization-induced fruits at 8 d after pollination compared to parthenocarpic fruit development at around 35 d postemasculation. Fertilizationinduced fruits in 'Monalbo' were larger than the smaller seedless fruit obtained after emasculation (Fig. 5, A and B; Table II).

gAtarf8-4 Sequences Enhance Parthenocarpy in the Tomato 'Monalbo'
We transgenically introduced the gAtarf8-4 mutant allele as a genomic construct into the tomato 'Monalbo' to examine if this would influence parthenocarpic fruit initiation in tomato. We chose the Atarf8-4 allele because it stimulated fertilization-independent fruit initiation in Arabidopsis in the experiments described above via a non-gene-silencing mechanism. Phylogenetic studies suggested a close relationship between ARFs from different plant species, implying that they derived from a common ancestor (Wang et al., 2007). Therefore we hypothesized that using gAtarf8-4 should enable functional competition with tomato ARF8 sequences and also provide the opportunity to distinguish and detect transcripts formed from introduced and endogenous genes. Three independent primary transgenic 'Monalbo' lines containing one, three, and five copies of the Atarf8-4 gene were selected for further analysis. Expression analysis using quantitative real-time PCR showed that Atarf8-4 levels in the transgenic plants increased proportionately with the number of introduced copies of the Atarf8-4 gene in the tomato plants (Table III).
Emasculation of untransformed 'Monalbo' plants showed that the majority of parthenocarpic fruit developing on the control plants were small (Fig. 5A), while wild-type fruits that developed after fertilization were much larger (Fig. 5B). All three of the primary transgenic plants showed enhanced parthenocarpy  (2) and gives the total number of ARF8 and arf8-4 genes present in the lines. Bars show pistil lengths 6 SDs for the individual lines. with respect to an increase in the number of emasculated flowers that initiated fruit development, and in terms of final fruit size compared with the emasculated, untransformed control plants (Table III; Fig. 5C). The extent of this enhanced parthenocarpic capability correlated with an increase in the steady-state pool of introduced mutant Atarf8-4 transcripts (Table III).
Enhanced parthenocarpic fruit set and growth was maintained in segregating plants containing one to five copies of the Atarf8-4 gene obtained from self-pollinated primary transgenics. Retention of emasculated flowers and subsequent parthenocarpic fruit initiation increased up to 79% in the transgenic lines compared to 50% in wild-type plants (Table IV). Larger fruits were generally observed in plants containing more copies of the introduced Atarf8-4 gene. Many of the parthenocarpic fruit from transgenic lines were similar in size and weight to seeded fruit from untransformed control plants and a greater percentage of fruit grew much larger than the average small sized seedless fruit formed after the emasculation of untransformed 'Monalbo' flowers (Table IV; Fig. 5, A-C). The addition of the gAtarf8-4 mutant allele has resulted in an enhanced pathenocarpic phenotype in terms of increased parthenocarpic fruit set and also size, thus we conclude that fruit initiation in tomato is controlled in a similar manner as found previously in Arabidopsis and that a S. lycopersicum ARF8 (SlARF8) ortholog has a role in controlling fruit initiation in tomato.

Larger Parthenocarpic Fruit Contain Seed-Like Structures (Pseudoembryos)
Tomato fruit from the control and transgenic plants were also examined for seed, pulp, and flesh content. Emasculated flowers of wild-type 'Monalbo' predominantly formed small fruits with small locules and little pulp (Fig. 5D). After fertilization 'Monalbo' formed large fruits with large locules that were filled with pulp, in which the seeds were embedded (Fig. 5E). Fruits that developed after emasculation in the transgenic plants had pulp-filled locules, a thinner mesocarp, and enlarged central columella (Fig. 5F).
When assessed for seed content, the small parthenocarpic fruits from both emasculated control and transgenic flowers predominantly contained enlarged ovules ( Fig. 5G), compared to the fully developed, large, hairy seeds found in fertilization-induced fruit (Fig. 5H). In the larger parthenocarpic fruit from emasculated control 'Monalbo' and transgenic plants (generally 20 g and above) many of the enlarged ovules had further differentiated into well-developed seed-like structures often called pseudoembryos (Fig. 5J). The size and level of differentiation of these seed-like structures varied, but was generally more advanced the larger the parthenocarpic fruit were in which they formed. Fully differentiated pseudoembryos were about oneeighth of fertilization-induced seed size (compare Fig. 5, H and J). Seed coat differentiation had occurred producing soft, seed-like structures with thickened lignified radial walls and protruding hairs. Sections of the fertilization-induced tomato seeds contained embryo and endosperm (Fig. 5I), whereas the well-developed pseudoembryos from the large parthenocarpic transgenic fruit lacked both embryo and endosperm and the collapsed embryo sac was surrounded by deeply staining multilayered endothelium (Fig. 5K). A correlation between the frequency and extent of pseudoembryo development with fruit growth and final fruit size has been observed previously (Kataoka et al., 2003). In our study we noticed a similar association between the development of large pseudoembryos, final fruit size, and copy number of the introduced Atarf8-4 gene in 'Monalbo'. We also found pseudoembryos in parthenocarpic fruit following emasculation of other tomato cultivars. These data suggest that SlARF8 may also play a role in regulating seed initiation and/or differentiation in tomato.
Two Distinct ARF8 mRNAs Are Found in Tomato 'Monalbo' and Arabidopsis The tomato ortholog of ARF8 was isolated from 'Monalbo' to investigate potential links to parthenocarpy given that mutations in Arabidopsis ARF8 induce parthenocarpic fruit initiation and that expression of gAtarf8-4 was able to enhance parthenocarpy in transgenic tomatoes. Screening of tomato EST databases from flowers and fruits identified three overlapping unigene sequences with high homology to AtARF8. Based on these sequences, PCR primers were designed and RT-PCR and 5# and 3# RACE were used to amplify and clone the full-length mRNA clones of SlARF8.
Amplifications and cloning using these PCR primers led to the isolation of two nearly identical cDNA sequences. The longer ARF8 clone of 2,872 bp (SlARF8 [full]) had 74% sequence identity with AtARF8 at the DNA level (Supplemental Fig. S1A). The similarity between the predicted protein sequence from tomato and AtARF8 was 76%, while the predicted protein identity was 66% (Table V;   . Fruits and seeds of wild-type and transgenic tomato plants. A, Typical small, parthenocarpic fruit developing from 'Monalbo' wild-type plant after emasculation. B, Large fruit developing from 'Monalbo' wild-type plant after fertilization. C, Large, parthenocarpic fruit developing from 'Monalbo' 1 gAtarf8-4 number 7 plant after emasculation. D, Small, emasculated 'Monalbo' wild-type fruit cut in half. A little orange pulp is visible in the locule. E, Large, fertilized 'Monalbo' wild-type fruit cut in half. The locule is filled with pulp and the seed are clearly visible. F, Large, emasculated 'Monalbo' 1 gAtarf8-4 number 7 fruit cut in half. The central columella (c) is enlarged and the locule is filled with pulp. Some pseudoembryos can be seen buried deep in the pulp. G, Ovule traces from a typical small, parthenocarpic 'Monalbo' wild-type fruit. H, Seed from a fertilized 'Monalbo' wild-type fruit and cross section through the seed (I), showing the developing embryo inside. J, Pseudoembryo from a parthenocarpic fruit from 'Monalbo' 1 gAtarf8-4 number 7 and cross section through the pseudoembryo (K), showing a collapsed embryo sac without an embryo. Bars 5 1 cm in A to F and 1 mm in G to K. ces, Collapsed embryo sac; em, embryo; en, endosperm; sh, seed hair; t, testa.  6A). This 5 bp deletion is predicted to introduce an early stop codon immediately after the deletion within the DNA-binding domain (boxed in Fig. 6A) potentially resulting in a truncated ARF8 protein containing 185 amino acids. We considered that the shorter cDNA might arise from alterations in splicing at an exon/ exon boundary of the same gene from initial comparison of both SlARF8 cDNA sequences with AtARF8 mRNA and the AtARF8 genomic sequence. Alternatively, given the tomato genome is not yet sequenced, it could have originated from mutations or sequence deletions during recent gene duplication events. DNA gel-blot analysis using a 359 bp fragment comprising part of the Q-rich middle region and the beginning of the C-terminal protein interaction domain of SlARF8 (Fig. 6B) as a probe to 'Monalbo' genomic DNA showed the presence of multiple bands (data not shown). Given the high sequence similarity between ARFs, especially in the DNA-binding domain and the C-terminal protein interaction domain, a cross-reaction with other ARFs in the 'Monalbo' genome is to be expected.
To directly examine the origin of the shorter transcript, we used PCR to isolate an 1,856 bp genomic fragment of the SlARF8 gene spanning exons IV to VIII and including the region around the 5 bp deletion (Fig.  6B). A total of five cloned PCR products were examined and found to be identical in sequence. This tends to argue against multiple genes giving rise to the two SlARF8 mRNAs as intron sequences are rarely conserved to this extent even in recently duplicated genes and duplicate genes have been found to exhibit dramatically accelerated rates of protein evolution (Castillo-Davis et al., 2004).
Comparison of the two SlARF8 mRNA sequences with the SlARF8 genomic sequence, the intron/exon structure of AtARF8, and consensus splice donor and acceptor sites confirmed that the shorter mRNA is likely to be a product of alternative splicing. Intron/exon  Fig. S1C). A splice donor site (GC) marking the end of exon VI exists at 479 bp in the isolated SlARF8 genomic sequence (arrow in Fig. 6C). This indicates that there is an exon/exon boundary after 553 bp of the mRNA, immediately 5# to the 5 bp deletion found in SlARF8 (25 bp). The consensus sequence for splice acceptor sites is (T/C) AG (Simpson and Filipowicz, 1996) and analysis of the SlARF8 genomic fragment showed that there is one splice acceptor site that appears to be used to generate the longer (full length) SlARF8 mRNA ( Fig. 6C; TAG, splice acceptor site 1). However, an alternative splice acceptor site (CAG) is also present 5 bp downstream from that used to generate the large mRNA at position 1,564 to 1,566 ( Fig. 6C; splice acceptor site 2). We speculate that the presence of the shorter mRNA sequence results from an alternative splicing event in the same ARF8 gene when this downstream acceptor site is used.
To test this hypothesis we developed a dCAPS marker to examine the presence or absence of the two SlARF8 mRNA variants in 'Monalbo' and some of the other tomato cultivars known to form parthenocarpic fruit after emasculation. Using the dCAPS marker with control SlARF8 plasmid DNA resulted in a single band as expected. When mRNA from 'Monalbo' flowers collected around anthesis was used, the dCAPS marker showed two bands, indicating both SlARF8 (full) and SlARF8 (25 bp) mRNA variants were present (Fig.  6D). Twelve additional tomato cultivars were tested with the dCAPS marker. Both mRNA variants were found in flowers of all cultivars collected at anthesis as indicated by the presence of two bands (Fig. 6D).
We also tested the expression of SlARF8 in flowers of the primary transgenic lines transformed with the gAtarf8-4 construct. There was no detectable decrease in the levels of endogenous SlARF8 (full) and SlARF8 (25 bp) variant transcripts, indicating that cosuppression or silencing of the endogenous SlARF8 mRNAs did not occur in these plants (Fig. 6E). In addition, various 'Monalbo' tissues were sampled and tested and all samples contained the two different SlARF8 mRNA variants (Fig. 5E). The factors that give rise to the alternatively spliced mRNAs in tomato are thus not floral specific.
These observations in tomato prompted us to examine whether an alternative splice acceptor sequence is also present in the ARF8 gene sequence of Arabidopsis (Ler, Col, and Ws ecotypes). We found that the alternative splice acceptor site (CAG) is present at the same position in Arabidopsis (see position 553-555 of AtARF8 sequence in Fig. 6A) as previously identified in tomato. To determine if it is recognized by the splice machinery of Arabidopsis to produce an alternatively spliced AtARF8 mRNA variant, we performed the dCAPS marker test for the 5 bp deletion of the alternative splice variant with Arabidopsis sequence-specific primers using cDNAs from ecotypes Ler, Col, and Ws. The results show two bands were present for all three ecotypes tested (Fig. 6E), indicating the presence of the splice variant. As there is a single ARF8 gene in the Arabidopsis genome, both transcripts clearly arise from the same gene.
These data indicate that in both Arabidopsis and tomato low levels of endogenous, aberrant ARF8 transcripts exist in the form of the shorter splice variant mRNA that is predicted to give rise to a truncated protein. In most tomato cultivars we examined, we observed some limited ovary growth up to full fruit development following emasculation. Similarly, some carpel elongation occurs in emasculated Arabidopsis ecotypes (Vivian-Smith and Koltunow, 1999). This is in contrast to other plants in which unfertilized flowers usually abscise and indicates that the controls restricting fruit development in Arabidopsis and selected tomato cultivars in the absence of fertilization might not be as tightly regulated. It is tempting to speculate that the presence of the ARF8 splice variant in both tomato and Arabidopsis is possibly contributing to a capacity for this low level of initial growth. This needs further investigation. We have observed that the alternative splice acceptor sequence is also present in the gARF8 sequences from B. rapa and rice (Oryza sativa; data not shown) and further analysis of the gene sequences and expression patterns of ARF8 in these and other plants will be necessary to determine if an alternative splice variant of ARF8 exists in these species and what biological effects it might have on fruit initiation.

Transgenic Introduction of Aberrant ARF8 Transcripts Induces Parthenocarpy in Arabidopsis
Recent work in Arabidopsis has shown that ARF8 might be part of a regulatory complex that negatively regulates fruit initiation. Lesions in AtARF8 lead to parthenocarpic fruit formation (Goetz et al., 2006). Here we have shown that the introduction of aberrant ARF8 sequences to wild-type Arabidopsis plants in the form of a genomic Atarf8-4 sequence or mutant and wildtype genomic translational GUS fusions can stimulate parthenocarpy in Arabidopsis via a non-gene-silencing mechanism. Thus, we suggest that the introduction of aberrant molecules can destabilize the formation of a proposed inhibitory complex controlling the transition from flower to fruit development, allowing fruit initiation in the absence of fertilization. Mechanisms by which this might occur could employ dosage effects involving the competitive interference of aberrant proteins, inhibiting the N-terminal DNA-binding domain of ARF8 from binding to the correct promoter sequences in auxin responsive genes and/or preventing  mRNA to identify splice donor and acceptor sites. The splice donor site (gc) and the two potential splice acceptor sites (tag and cag) are underlined and the splice sites are indicated by arrows. The numbers are the running count of the bases in the SlARF8 gene fragment sequence. The alignment contains a large gap in the sequence of intron VI. D, Analysis of the tomato ARF8 dCAPS marker. DNA from a SlARF8 mRNA plasmid was used as control. cDNAs from flowers of 'Monalbo' and 12 additional tomato cultivars collected at anthesis were tested for the presence of the two SlARF8 mRNA variants. E, Analysis of cDNAs from various 'Monalbo' tissues collected at different developmental stages for the presence of the two SlARF8 mRNA variants. Expression of the C-terminal protein interaction domain to bind to Aux/IAA proteins or to dimerize with other proteins that may make up a repressive complex.
The extent of parthenocarpy induced in Arabidopsis varied depending on the construct used and individual lines analyzed, which is most likely due to differences in expression levels and effectiveness of the products of the various constructs. Variable responses like this are expected if competitive interference occurs between the introduced aberrant ARF8 protein and the regulatory complex. The stronger effects with the gAtARF8: GUS construct over the gAtarf8-4 and gAtarf8-4:GUS constructs seem to support this. Due to the previously identified point mutation in the translation initiation codon (Goetz et al., 2006), any translation from the Atarf8-4 mutant transcripts would have to start from an alternative start codon and is therefore likely to be less efficient than translation from the wild-type AtARF8 transcript. While equal or higher gene expression has been shown for all constructs via PCR methods and expression of the GUS fusion proteins was verified by analysis of the GUS staining patterns (Goetz et al., 2006), we were unable to confirm the presence and sizes of proteins formed in gAtarf8-4, gAtarf8-4:GUS, and gAtARF8:GUS plants by western-blot analysis due to the lack of effective and specific antibodies.
There were also differences between the translational GUS constructs and the gAtarf8-4 allele itself in terms of efficiency of parthenocarpic induction in Arabidopsis. Clear parthenocarpic phenotypes were more readily observed in lines containing the translational GUS constructs. Although the elongation response was variable with all constructs, dehiscence zone formation was only variable in lines transformed with the gAtarf8-4 construct, while it was present in all lines transformed with the translational GUS constructs. We suggest that one reason for this is that elongation and dehiscence zone formation are two separate pathways that might be controlled through the same complex, but activated in different ways.
Since the GUS protein is attached to the C-terminal protein-protein interaction domain (Fig. 6B) of Atarf8-4 and AtARF8, the greater effectiveness of the translational GUS constructs in inducing the dehiscence zone could be due to the presence of the GUS protein in that position. ARFs can homodimerize and also heterodimerize with other proteins including Aux/IAA proteins via the C-terminal protein-protein interaction domain (Liscum and Reed, 2002;Hardtke et al., 2004;Tatematsu et al., 2004) and these interactions may be important in regulating the formation of the dehiscence zone. In transgenic Arabidopsis plants containing the translational fusions, the presence of the GUS protein may interfere with the normal protein inter-action activity and/or binding of the DNA-binding domain (Fig. 6B) to promoters of early auxin responsive genes. These possibilities require further analysis at the protein level.
Collectively the data presented show that in addition to the genetic induction of parthenocarpy, as demonstrated in our previous work (Goetz et al., 2006), parthenocarpy can also be stimulated by expressing aberrant products that affect the auxin response pathway. Thus, interference with the proposed inhibitory complex in Arabidopsis either by genetic ablation of either protein or by introducing inhibitory or competitive effects through nonfunctional or inactive versions of transcripts or proteins, allows parthenocarpic fruit development by permitting fruit initiation to occur in the absence of fertilization.

Genetics, Expression, and Induction of Parthenocarpy in Tomato
Parthenocarpy is a desirable trait in horticultural crops, as it enables fruit set and growth to be independent of pollination, fertilization, and seed development. Therefore it may circumvent the environmental constraints on fruit production and ensure yield stability. Marketable parthenocarpy in tomato would provide the possibility to produce seedless fruits with high consumer appeal and could also be a valuable trait for industrial tomatoes because parthenocarpic fruit can have a higher percentage of soluble solids, improving yield and flavor of paste and reducing processing costs.
Fruit set in tomato is temperature dependent and commercial fruit quantities are obtained when night temperatures are maintained between 15°C to 21°C. Problems in pollen formation limit fruit set if night temperature is lower than 13°C or when the day temperature is higher than 38°C (Baksh et al., 1978;Rylski, 1979;Lin et al., 1983a;George et al., 1984;Ho and Hewitt, 1986;Vardy et al., 1989a;Lukyanenko, 1991;Adams et al., 2001). Breeding programs have been aimed at generating tomato cultivars that set seedy fruit under normal pollinating conditions and commercially marketable but seedless fruit under pollen formation-limiting conditions. At least four recessive parthenocarpic or pat loci Figure 6. (Continued.) SlARF8 in anthesis flower cDNAs from 'Monalbo' wild-type and 'Monalbo' 1 gAtarf8-4 lines was also determined with the help of the tomato ARF8 dCAPS marker. The Arabidopsis ecotypes Ler, Col, and Ws were also tested for the presence of the ARF8 (25 bp) variant using the dCAPS marker with Arabidopsis sequence-specific primers. have been identified in tomato by growing plants mainly at cool temperatures and scoring for seedless fruit formation. There are differences in the quantification of parthenocarpy in genetic studies but in addition to the pat loci other minor parthenocarpy loci that influence pat expression in tester lines used in the crosses have also been identified (Vardy et al., 1989a(Vardy et al., , 1989b. The molecular identity of all of these loci is unknown. Emasculation has rarely been used to score parthenocarpic fruit growth in tomato, which is understandable in the plant breeding context where the intention is to provide the grower with varieties that are facultative for parthenocarpy in that they set fruit under normal growth conditions and also temperatures that limit pollen formation. In our study plants were emasculated and subsequently monitored for long periods of time. Our analysis showed that many commercially available tomato cultivars and also experimentally used lines that have been selfed for many generations show some degree of natural parthenocarpy following emasculation ranging from ovary expansion to varying sized seedless fruits. Some of these small and misshapen fruits are likely to be similar to those referred to as pseudofruits, puffs, or nuts in the literature (Foster and Tatman, 1937;Vardy et al., 1989a;Pandolfini et al., 2002;Carmi et al., 2003). We have previously demonstrated that parthenocarpy in Arabidopsis arf8 mutants is not expressed if floral whorls surrounding the carpel are not removed, as signals from these tissues inhibit the expression of parthenocarpy . Our preliminary studies indicate that the pollen is the primary source of this inhibition (M. Goetz and A.M. Koltunow, unpublished data). It is not certain if similar floral whorl signals suppress parthenocarpy in tomato. But one explanation for the relatively high frequency of parthenocarpic growth we observed may be that by emasculating the tomato flowers we have removed floral whorl suppression signals allowing the expression of parthenocarpy.
Here we have established that ARF8 sequences are highly conserved between Arabidopsis and tomato at the mRNA and predicted protein level. When we introduced the mutant genomic Atarf8-4 sequence to 'Monalbo' a significant increase in parthenocarpic fruit initiation, fruit size, and weight was observed following emasculation above the background levels in control plants, supporting a role for ARF8 in regulating fruit initiation in tomato. Moreover, silencing of SlAux/ IAA9 has been demonstrated to induce parthenocarpic fruit initiation in tomato (Wang et al., 2005). Together, the involvement of ARF and Aux/IAA proteins implicates the auxin response pathway in the repression of tomato fruit initiation in the absence of fertilization. It is tempting to speculate that SlARF8 and SlAux/ IAA9 interact and the analysis of this together with the analysis of the expression of these two proteins in pat plants may provide informative clues concerning the regulation of parthenocarpy in tomato.

Aberrant Endogenous ARF8 Transcripts Together with Existing Data Provide a Model for Fruit Initiation in Tomato and Arabidopsis
When we isolated SlARF8 cDNA sequences from tomato, we found a second variant of the mRNA that is most likely formed as a result of alternative splicing. Analysis of ARF8 expression in wild-type Arabidopsis ecotypes Ler, Col, and Ws confirmed the presence of the same alternative splice form and both plants show ovary expansion and growth following floral emasculation. This is in contrast to other plants where unfertilized flowers abscise. The use of other splice sites resulting in the alternative splicing of plant genes has been reported previously and heat stress or temperaturedependent splicing has been described as the cause in several of these cases (Burr et al., 1996;Sablowski and Meyerowitz, 1998;Lazar and Goodman, 2000;McKibbin et al., 2002;Colot et al., 2005;Reddy, 2007). It is not clear, however, what cues lead to the generation of the ARF8 splice variant in Arabidopsis and tomato.
Our data suggests that ARF8 has a role in modulating fruit initiation in both Arabidopsis and tomato. Given the large amount of biochemical data available Figure 7. A model for the role of ARF8 and Aux/IAA9 proteins in the control of fruit initiation and growth. ARF8 and Aux/IAA9 proteins, together with potentially other as yet unknown proteins (5?), form a regulatory complex that can either directly block transcription of target (fruit initiation) genes, or act indirectly by preventing ARF8 from functioning as a transcriptional activator. After pollination and fertilization occur, auxin acts by binding to its receptor, TIR1, promoting degradation of Aux/IAA9 proteins via the SCF TIR1 ubiquitin ligase complex. In the absence of Aux/IAA9, ARF8 together with additional signals and activators (5A) stimulate expression of early auxin responsive genes, initiating fruit growth and development. Destabilization of the regulatory complex or reduction of its functionality by aberrant ARF8 transcripts and possible products can lead to a reduction or loss of the inhibition of transcription of the fruit initiation genes, resulting in parthenocarpic fruit growth.
about the auxin signal transduction pathway (Ulmasov et al., 1999a(Ulmasov et al., , 1999bGray et al., 2001;Rogg and Bartel, 2001;Liscum and Reed, 2002;Tiwari et al., 2003;Hardtke et al., 2004;Kepinski and Leyser, 2004;Tatematsu et al., 2004;Kepinski and Leyser, 2005), we have devised a model that proposes that ARF8 and Aux/IAA9 orthologs interact in these plants (potentially with other proteins) to form a regulatory complex (Fig. 7). This complex can bind to the promoters of a range of primary auxin responsive genes that play an essential role in repressing fruit initiation in the absence of fertilization. The presence of the ARF8 splice variant transcripts and possible products could act to partially destabilize this complex. In Arabidopsis (and possibly tomato) this destabilization may be revealed as some fruit growth following emasculation and coincident removal of floral whorl inhibitory signals. Assuming fertilization induces an auxin burst, which has been documented in orchids (O'Neill, 1997) and shown to target Aux/IAA proteins to the proteasome degradation pathway (Ulmasov et al., 1999b;Gray et al., 2001;Tiwari et al., 2001;Zensner et al., 2001), this would completely remove the transcriptional block allowing the expression of genes promoting fruit initiation. The involvement of auxin in fruit initiation is supported by studies showing that exogenous application of auxin (King, 1947;Asahira et al., 1967;Lin et al., 1983b) and the expression of auxin biosynthesis genes in transgenic tomato ovaries (Ficcadenti et al., 1999;Carmi et al., 2003) induce commercially acceptable parthenocarpic fruit and elevated auxin levels have been found in a range of parthenocarpic fruits (Mapelli et al., 1978;George et al., 1984;Talon et al., 1990Talon et al., , 1992. Experimental confirmation of the interacting partners of ARF8 and Aux/IAA9 and an examination of any potential sources of auxin during fertilization-dependent fruit set in both Arabidopsis and tomato are required to substantiate the model. The role of the splice variant in complex destabilization might be further tested by introducing genomic sequences that produce only the splice variant transcript into Arabidopsis and tomato.
Obviously additional signals and activators are required following removal of the block in fruit initiation to promote fruit development and growth. In fertilizationinduced fruits, seeds have long been known to produce growth-promoting hormones. Work in garden peas (Pisum sativum) suggests that seeds produce a modified auxin that stimulates GA production in the surrounding carpel tissues (Ozga et al., 2002;Ozga and Reinecke, 2003). Parthenocarpic fruits are likely to have other sources of growth hormones, but a link between auxin and GA in Arabidopsis is supported by the dependence of Atarf8 parthenocarpy on GA activity  and increased GA levels are associated with growth of parthenocarpic tomato fruit (Fos et al., 2000(Fos et al., , 2001. In tomato, the pat-2 and pat-3/pat-4 mutations, for example, have been found to independently accumulate high concentrations of GAs in unpollinated tomato ovaries via different pathways. It was therefore suggested that natural parthe-nocarpy induced by pat-2 and pat-3/pat-4 is due to increased synthesis of active GAs (e.g. GA1) in unpollinated ovaries as a result (Fos et al., 2000(Fos et al., , 2001. The pseudoembryos observed in parthenocarpic tomato may contribute to early growth of fruits as a substitute for developing seeds, either by producing phytohormones or by creating a sink for hormones to be transported into the growing fruits (Nitsch, 1952;Varga and Bruinsma, 1976;George et al., 1984;Kataoka et al., 2003). Well-developed pseudoembryos have been observed in parthenocarpic tomato lines controlled by the pat-2 allele (Kataoka et al., 2003), auxin-induced parthenocarpic fruits (King, 1947;Asahira et al., 1967;Lin et al., 1983b), transgenic tomatoes containing auxin biosynthesis genes (Carmi et al., 2003), and larger seedless fruits of the 'Monalbo' lines containing the Atarf8-4 gene in addition to parthenocarpic fruits formed following emasculation of a range of varieties in this study. They clearly correlate with the increased size of parthenocarpic fruit. Groot et al. (1987) suggested that GAs are important for seed growth and development in tomato. The application of uniconazole, an inhibitor of GA biosynthesis, to parthenocarpic tomato lines containing the pat gene and also auxin-induced parthenocarpic fruits 3 d after anthesis, abolished fruit growth and pseudoembryo development. GA treatment 2 d after uniconazole application restored fruit and pseudoembryo growth (Kataoka et al., 2003). This provides further support for the linkage of GA biosynthesis to fruit and pseudoembryo growth in parthenocarpic tomato once the block to fruit development is removed.
In our study, the development of ovules into pseudoembryos in parthenocarpic tomato fruit contrasted with the observed senescence of Arabidopsis ovules around 5 d after emasculation in both wild-type and arf8-4 plants. Arabidopsis ovules similarly senesced when parthenocarpy was induced by application of various growth hormones (Vivian-Smith, 2001). This suggests that there may be different dependencies on ovule and ovary communication events in tomato and Arabidopsis during fruit development and this requires further investigation.

Arabidopsis Growth and Scoring Parthenocarpy
Arabidopsis (Arabidopsis thaliana) seeds (Ler, Col, and Ws) were surface sterilized and grown as described previously (Vivian-Smith and Koltunow, 1999). Parthenocarpy was assessed by flower emasculation (Vivian-Smith and Koltunow, 1999;Vivian-Smith et al., 2001). Only siliques above flower position 20 were used, collected, and photographed and their lengths were determined using the Scion Image Beta 4.02 program (http://www.scioncorp.com/ frames/fr_scion_products.htm). Plants producing siliques that significantly and reliably elongated more than the corresponding wild-type plants and that formed a dehiscence zone were scored as parthenocarpic.

Cloning of arf8-4 and Translational GUS Constructs
The cloning of the arf8-4 mutant gene in the pGEM-T easy vector (Promega) was described earlier (Goetz et al., 2006). It was subcloned into the pBIN19 vector (Bevan, 1984) and transformed into Agrobacterium tumefaciens strain AGL1. Ler plants were transformed with the construct via the floral-dip method (Clough and Bent, 1998).
The cloning of the translational ARF8:GUS and arf8-4:GUS constructs and the plant transformations have been detailed earlier (Goetz et al., 2006).

RNA Preparation, RT-PCR, and Quantitative Real-Time PCR
Total RNA was extracted from Arabidopsis and tomato (Solanum lycopersicum) plant tissues using the RNeasy Plant Mini kit (Qiagen) according to manufacturer's instructions.
One microgram of total RNA treated with an on-column RNase-free DNase protocol (Qiagen) was used as template for cDNA synthesis with the ThermoScript RT-PCR system (Invitrogen).
Quantitative real-time PCR was performed as described before (Goetz et al., 2006).

Growth of Tomato and Cultivars Used
The tomato cultivars used were either obtained as young plants from nurseries and repotted or purchased as seed and grown as described below. 'Roma', 'LA1563', 'LA1714', 'LA3130', and the male sterile lines 'LA1222' and 'UC82b' were obtained from the C.M. Rick Tomato Genetics Resource Centre (University of California). 'Monalbo' and 'ChicoIII' were kindly donated by Andrea Mazzucato (University of Viterbo). 'Moneymaker', 'Beefsteak', 'Big-Boy', 'Roma', and 'Margold' were obtained from Eden Seeds. 'Mighty Red', 'Roma', 'Cocktail Supreme', 'Health Kick', 'Red Mamma', 'Sweet Cherry Gold', 'Grosse Lisse', 'Top Dog', and 'Patio Prize' were obtained from local nurseries and 'Cherry Ripe', 'Greek', and 'LM' were obtained from a local private garden. Flowers from each cultivar were emasculated, tagged, and checked after 3 to 4 weeks for signs of fruit growth. Most cultivars showed signs of fruit development after the indicated time.
Seeds of tomato ('Monalbo' ['LA2818' in Charles Rick collection, UC Davis]) were sown in a peat-based seed and modular compost and were germinated in a glasshouse compartment. Temperatures in the glasshouse varied between 10°C to 19°C minimum temperature at night and 24°C to 34°C maximum temperature during the day. After 3 weeks, seedlings were pricked out into 1-L pots containing peat-based potting compost. At first flowering, plants were transplanted into 10-L pots containing peat-based potting compost and plants were supported by canes.
The arf8-4 allele from Arabidopsis was used for transformation of 'Monalbo'. Tomato transformations were done essentially as described by Fillatti et al. (1987) and transgenic plants were selected by germinating sterilized seeds on selective medium (one-half Murashige and Skoog medium, 3% Suc, and 100 mg L 21 kanamycin). Flowers of transgenic lines were emasculated and tagged 1 to 2 d before anthesis, and developing parthenocarpic fruits were harvested, weighed, and analyzed when fruits were red and ripe.
Isolation of ARF8 mRNA from Solanum esculentum EST databases from Solanum esculentum were screened for sequences with high homology to AtARF8. The three UniGene sequences SGN-U228441, SGN-U238776, and SGN-U227556 were identified and analyzed further. Alignments of the three UniGenes showed that SGN-U227556 and SGN-U238776 overlap and that after a 70 bp gap SGN-U238776 and SGN-U228441 also overlap with high identity. Based on the consensus sequences, the primers SlARF8_SF1 to SlARF8_SF12 (Supplemental Table S1) were designed to amplify and clone this sequence and perform 5#-and 3#-RACE experiments to amplify and clone the full-length mRNA sequence of SlARF8.

dCAPS Marker for ARF8 Variants
The two SlARF8 variants can be distinguished with a dCAPS marker (Neff et al., 2002). Amplification with the primers SlARF8_SF11 and SlARF8_AvrII (Supplemental Table S1) from 1 mL of cDNA from tomato and subsequent digest with XmaJI (Fermentas) resulted in a band of 199 bp for the SlARF8 (full) variant and a band of 177 bp for the SlARF8 (25 bp) variant. An alternative dCAPS marker using the primers SlARF8_SF11 and SlARF8_SacII and digestion with Cfr42I (Fermentas) can also be used and resulted in the same bands as described above. For detection of the two AtARF8 variants in the Arabidopsis ecotypes, the primers AtARF8_SF11 and AtARF8_SacII (Supplemental Table  S1) and digestion of the PCR product with Cfr42I (Fermentas) were used. This resulted in the same-sized bands as described above.

Embedding and Sectioning of Seeds and Pseudoembryos
Seeds, ovule traces, and pseudoembryos were collected from tomato fruits and fixed in 3.1% glutaraldehyde in 0.1 M sodium cacodylate buffer pH 7.4. Fixed tissues were dehydrated in an acetone series and infiltrated with acetone: Spurrs resin (1/1) overnight at room temperature in a vacuum chamber. They were then transferred to 100% Spurrs resin overnight in a vacuum chamber, embedded in fresh Spurrs resin, and polymerized at 65°C overnight. Specimen were cut into 2 mm thin sections and then stained with 0.1% toluidine blue in 0.02% sodium carbonate to determine cell identity.

Microscopy and Photography of Fruits and Seeds
Photographs of whole and cut fruits were taken with a Nikon Coolpix 995 digital camera. Whole-mount seeds and seed sections were viewed with Stemi2000C or Axioskop microscopes (Carl Zeiss). Digital images were captured using a Spot II camera (Diagnostic Instruments Inc.). Image processing and reproduction were performed with Auto Montage Essentials (Syncroscopy) and Photoshop 7.0 (Adobe Systems).
Sequence data from this article can be found in the GenBank/EMBL data libraries under accession numbers AT5G37020 (AtARF8) and EF667342 (SlARF8).

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Table S1. Primer names and sequences used to amplify and clone SlARF8 and for amplification of the SlARF8 dCAPS markers.