Seeds of S. lycopersicum cultivar “Micro-Tom”, both wild type (ID: TOMJPF00001) and an EMS parthenocarpic mutant iaa9-3 (ID: TOMJPE2811-1), were supplied by the National Bioresource Project archived at the TOMATOMA database¹. The plants were grown in rockwool blocks in a growth room set at 25°C under photosynthetically active light (75-110 mmol/m²/s) for 16 h and 20°C in the dark for 8 h.

Fresh anther cones were collected at the anthesis stage and then used to prepare three types of pollen, that is, dead pollen (DP), soft X-ray-treated pollen (XP), and intact/normal pollen (IP). To prepare DP, the anther cones were dried at 100°C for 2 h. IP were prepared by drying the anther cones at 35°C for 6 h. Finally, to make XP, fresh anther cones were first subjected to a soft X-ray irradiation (Model: OM-303M, OMIC Corporation, Japan) of 1000 Gy for 72.15 minutes followed by drying at 35°C for 6 h. The protocol for soft X-ray irradiation was derived from our work (unpublished data) at the University of Tsukuba.

Flowers were emasculated one day before anthesis (-1 DAA) to avoid self-pollination. The emasculated flowers were then pollinated on the next day (0 DAA) using the pollen types described in section 2.2. Styles were removed from IP-pollinated flowers 3 h, 6 h, 9 h, 12 h, and 24 h after pollination (denoted as IP3h, IP6h, IP9h, IP12h, and IP24h, respectively). For XP- and DP-pollinated flowers, styles were removed 24 h after pollination (denoted as XP24h and DP24h, respectively). In all cases (IP3h, IP6h, IP9h, IP12h, IP24h, XP24h and DP24h), ovaries that remained after style removal were left to stand on the plants (10 plants per each treatment) for further histological and RNA-Seq analyses as well as assessment of fruit set ratios. The ovaries were collected at 2 DAA (for RNA-Seq analysis) and from 2–10 DAA (for histological analysis). Ovaries from emasculated but unpollinated flowers (denoted as E) with style removal 24 h after anthesis were also collected at the same timepoints to represent control treatments.

Pistils were collected 3 h, 6 h, 9 h, 12 h, and 24 h after pollination with IP pollen, and 24 h after pollination with XP and DP. The pistils were then fixed in 3:1 ethanol (100%): acetic acid (100%) solution for 12 h, washed in 70% ethanol and finally softened in 5 N NaOH for 12 h. Softened pistils were washed five times in distilled water and stained overnight in the dark with 0.01% aniline blue solution in K₃PO₄ buffer. The stained pistils were then mounted in 100% glycerol on a slide and observed under a UV microscope (Olympus BX50, Olympus-Life Science Company, Japan). At least three pistils were observed for each of the seven treatments.

Fruit set ratios were evaluated using five different plants and a total of 21-30 flowers for each style removal (IP3h, IP6h, IP9h, IP12h, IP24h, XP24h and DP24h) treatment. The ratios were expressed as the percentage of ovaries which remained on the flowers and showed an increase in size at 10 DAA.

Histological analyses were carried out using Technovit 7100 (Kulzer Technik, Germany) according to the protocol of Yeung and Chan (2015) with slight modifications. Briefly, ovaries collected at 0, 2, 4, and 10 DAA were fixed as described in Section 2.4 and then dehydrated by passing them sequentially through graded ethanol (70%/2 h → 80%/2 h → 90%/2 h → 100%/overnight). This was followed by three infiltration steps in ethanol:Technovit solutions at different concentrations (2:1/2 h → 1:1/2 h → 1:2/2 h), and overnight immersion in pure Technovit. The ovaries were then embedded in a mixture of Technovit and Hardener II (15:1 v/v) and allowed to polymerize at room temperature for 12 h before sectioning at a thickness of 5 μm using a rotary microtome machine (Leica RM2235, Leica Biosystem Ltd., China). Finally, the sections were mounted in water on a glass slide, dried at 40°C, and stained with toluidine blue (Sigma-Aldrich T3260, Merck, USA). A drop of Entellan New (Sigma-Aldrich 107961, Merck, USA) was added to the slide before observation under a light microscope (Olympus BX50). At least three ovaries were observed for each style removal treatment. To estimate fruit growth, cell layer and cell volume measurements were conducted on the microscope images using ImageJ software.

Samples for RNA-Seq analysis included emasculated but unpollinated ovaries (E) as a control (absence of pollination, pollen tube growth and fertilization), and ovaries from IP6h, IP24h, DP24h, and XP24h style removal treatments. These samples were selected based on pollen tube growth observations; IP6h represented partial pollen tube growth (no penetration into the ovaries and hence no fertilization), IP24h represented full pollen penetration into the ovaries with fertilization, XP24h represented full pollen tube penetration without fertilization, while DP24h represented pollination only (no pollen tube growth and no fertilization). All samples were collected 48 h after pollination (2 DAA), and each sample contained 5 ovaries with three replications. Total RNA was extracted from the ovary samples using the RNeasy^® Plant Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. Treatment with DNase I (Nippon Gene, Tokyo, Japan) was carried out to remove genomic DNA contamination. mRNA was then purified from the total RNA using a poly(A) mRNA magnetic isolation module kit (New England BioLabs). The pure mRNA was used to construct paired-end libraries for Illumina using the Ultra™ II directional RNA library prep kit (New England BioLabs), and sequencing was performed on an Illumina NovaSeq 6000 platform (Illumina, Inc.).

RNA-Seq data were analysed on the Galaxy main server². The raw RNA sequence reads were qualified using the FastQC tool and then trimmed by the Trimmomatic tool (Bolger et al., 2014). The resultant high-quality sequences were then mapped to tomato reference genomes (v. SL4.0) using the Hisat2 tool with default parameters (Kim et al., 2019). Mapped reads were counted using the featureCounts tool (Liao et al., 2014). The read counts obtained were normalized to the gene expression level as transcripts per kilobase million (TPM) reads. Gene Ontology (GO) enrichment and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were conducted on the web application iDEP (v. 0.93)³ and ShinyGO (v. 0.77)⁴ (Ge et al., 2018).

Data obtained in this study are presented as average values ± SE. Mean comparisons were tested by Tukey’s HSD (honestly significant difference) test at P < 0.05. All statistical analyses were performed using Statistical Tool for Agricultural Research (STAR), version 2.0.1 (Gulles et al., 2014).

In vivo pollen tube growth assays, following pollination of tomato flowers with IP, XP and DP, were crucial in this study to determine the independent effects of pollination, pollen tubes, and fertilization on fruit setting. For IP-pollinated flowers, pollen tubes elongated to approximately one-third and three-quarters of the styles after 3 h and 6 h (Figure 1A), respectively and the earliest penetration of the pollen tubes into the ovaries, indicated by a limited signal below the style baseline, was observed 9 h after pollination. Full penetration of IP-generated pollen tubes into the ovaries occurred 12 h and 24 h after pollination, as the signal strengths inside the ovaries were comparable. Likewise, pollen tubes in XP-pollinated flowers elongated and fully penetrated the ovaries 24 h after pollination (Figure 1A). As expected, DP failed to germinate and hence pollen tubes could not be observed in the styles of DP-pollinated flowers even after 24 h. Therefore, at the specified times of style removal, pollen tubes were still in the styles for IP3h and IP6h, but there was partial penetration into the ovaries for IP9h. For IP12h, IP24h and XP24h, full penetration of the pollen tubes into the ovaries had already taken place.

Ovaries that were left on the plants after style removal treatments were then assessed at 10 DAA for their ability to develop into fruits. Results indicated that IP3h and IP6h ovaries failed to initiate fruit set, similar to DP24h (Figure 1B). On the other hand, IP9h ovaries displayed a considerable fruit set ratio (70.2%), although it was significantly lower than IP12h (84.4%) and IP24h (93.5%) ovaries. Interestingly, XP24h ovaries also showed a noticeably high fruit set ratio (78.3%), albeit significantly lower than that of IP24h ovaries. Together, these findings indicate that pollination alone (represented by DP24h) or partial pollen tube growth (represented by IP3h, IP6h and to some extent, IP9h) were not sufficient to initiate fruit setting. Contrarily, full penetration of pollen tubes into the ovaries even in the absence of fertilization (as shown in XP24h) can sufficiently trigger fruit set.

After fruit set assessments, we asked whether full penetration of pollen tubes into the ovaries minus fertilization can also lead to normal cell division and expansion. Histological analyses showed that both the number of cell layers and cell sizes of pericarp tissues of XP24h ovaries were higher at 2 DAA compared to the initial 0 DAA ovaries (Figures 2A–C). It is worth noting that both the number of cell layers and area of pericarp cells in XP24h ovaries at 2 DAA were not statistically different from those of IP12h and IP24h ovaries at the same timepoint (Figures 2B, C). By contrast, IP9h ovaries, which had only partial pollen tube penetration (Figure 1A), only showed larger cells at 2 DAA with no significant increase in cell layer numbers. Ovaries of IP6h, DP24h and the control E showed insignificant changes in both the cell area and number of cell layers, and the cell sizes were similar to those of ovaries at 0 DAA.

Pericarp cell measurements were also conducted on the young fruits (at 4 and 10 DAA) that developed from XP24h and IP24h ovaries. As shown in Figure 2D, young fruits of XP24h and IP24h showed a comparable number of cell layers both at 4 and 10 DAA. However, IP24h fruits showed a consistently larger pericarp cell area than XP24h fruits at the evaluated timepoints. It is worth noting that from 0 DAA to 2 DAA, there were no significant differences between IP24h and XP24h with regards to the increase in both cell layer numbers (26%) and cell area (30%) (Figures 2A–C). At 4 DAA, however, the increase in cell area relative to control E ovaries at 2 DAA was remarkably higher (210%) than that of the number of cell layers (46%). Furthermore, there was only a slight increase in cell layer numbers between 4 DAA and 10 DAA (11%), but a remarkably sharp increase in cell area (830%) was recorded. Altogether, these findings suggested that during early stages of fruit development (0–4 DAA), cell division (indicated by cell layer numbers) has a greater contribution than cell expansion (indicated by cell area). By contrast, cell expansion has a greater contribution than cell division during later fruit developmental stages (from 4 DAA).

To examine the impact of fertilization on late fruit developmental stages, fruit size and seed numbers in ripe fruits derived from style removal treatments that successfully set fruits (IP9h, IP12h, IP24h and XP24h) were also determined (Figure 3). The largest fruit weight was registered in ripe fruits obtained from IP12h and IP24h (3.69 g and 3.61 g, respectively) (Figure 3A). Fruits that developed from IP9h ovaries displayed the smallest average weight (1.77 g), but XP24h fruits had a slightly higher average weight (2.38 g), albeit significantly smaller than IP12h and IP24h fruits. Both the diameter and height of IP12h and IP24 h fruits were bigger than those of XP24h (Figure 3D), which correlated well with fruit weight data. These results suggested that the degree of fertilization might have a positive impact on the final size of fruits at maturity. Specifically, IP12h and IP24h ovaries underwent full penetration of pollen tubes (and hence presumably, complete fertilization) which likely led to the development of normal-sized fruits. However, IP9h ovaries underwent only partial pollen tube penetration, and hence partial fertilization, which might account for the small fruit sizes. Indeed, there was no significant difference in the number of seeds per fruit for IP12h and IP24h, but IP9h fruits had noticeably a smaller number of seeds (Figure 3B). Likewise, XP24h ovaries did not undergo fertilization even though full penetration of pollen tubes from the sterile X-ray-irradiated occurred, likely contributing to smaller fruit sizes than IP12h and IP24h. It is also worth noting that XP24h fruits did not produce regular seeds; instead, they produced many tiny seed-like structures (Figure 3D), which failed to germinate (data not shown). This finding therefore confirmed that the X-ray-treated pollen used in the present study indeed lost their fertilization function. Finally, combination of fruit weight and seed number data revealed a strong positive correlation (R² = 0.85) between these two phenotypes (Figure 3C).

After phenotypical analyses, RNA-Seq was performed to identify transcriptomic changes in ovaries at 2 DAA triggered by the different pollen types and style removal times. A total of 500 most variable genes were used to construct a heatmap showing the overall expression patterns based on the log₁₀ transformed TPM values. As shown in Figure 4A, the expression patterns in DP24h and IP6h ovaries were similar to those in the control (E) ovaries, while XP24h and IP24h ovaries showed highly similar patterns. Indeed, Pearson’s correlation coefficient analysis, based on 75% of the top variable genes, confirmed that there were highly positive correlations between XP24h and IP24h samples (r = 0.99), while there were high correlations (r=0.98~0.99) among E, DP24h, and IP6h samples (Figure 4B). In addition, principal component analysis (PCA) was conducted to explain 78% of the variation in the datasets (PC1 = 74%, and PC2 = 4%), in which XP24h and IP24h samples were also grouped together, but clearly separated from E, DP24h, and IP6h (Figure 4C).

The DEGs with a minimal fold change of 2 and FDR < 0.01 were considered significant and selected for enrichment analysis. As illustrated in Figures 5A, D, 65 DEGs were detected in IP6h ovaries, of which 11were upregulated while 54 were downregulated compared to the control (E). The numbers of DEGs in XP24h ovaries were 4383 (1401 up-regulated and 2982 down-regulated) (Figures 5C, D), while in IP24h ovaries, 5062 DEGs (1514 up-regulated and 3548 down-regulated) were detected (Figures 5B, D). By contrast, there were no DEGs between DP24h vs E samples or between XP24h vs IP24h (Figure 5D).

Venn diagrams were also constructed to visualize the commonly or uniquely up- or down-regulated DEGs under different pollen treatments. This analysis revealed that XP24h and IP24h samples shared the majority of up-regulated DEGs (1336 genes), while there were 11, 65, and 178 uniquely up-regulated genes in IP6h, XP24h, and IP24h samples, respectively (Figure 5E). Among the downregulated DEGs, 2800 genes were common between XP24h and IP24h samples (Figure 5F). We found that 53, 181, and 747 genes were uniquely downregulated in IP6h, XP24h and IP24h samples, but only 1 gene was commonly downregulated in all three samples (Figure 5F).

Term enrichment analysis of the identified DEGs was then performed in an attempt to zoom in on the molecular changes triggered in XP24h ovaries relative to E ovaries (Figure 6). We found that the 1,401 up-regulated genes were classified into 40 GO terms and 15 KEGG pathways (Table S1). The top biological processes were ‘regulation of cell cycle process’ (30 genes), ‘mitotic cell cycle’ (44 genes), ‘cell cycle process’ (52 genes), and ‘cell cycle’ (61 genes) (Figure 6A). Additionally, the tope enriched cellular components were ‘DNA packing complex’ (28 genes), ‘nucleosome’ (27 genes), and ‘protein−DNA complex’ (32 genes), while molecular functions that stood out were ‘cyclin-dependent protein serine/threonine kinase regulator activity’ (13 genes), ‘protein kinase regulator activity’ (14 genes), ‘kinase regulator activity’ (15 genes), and ‘protein kinase binding’ (15 genes). For the KEGG pathways, 15 terms were found among the upregulated genes and these were related to many critical pathways such as ‘photosynthesis’ (13 genes), ‘biosynthesis of cofactor’ (15 genes), ‘plant hormone signal transduction’ (13 genes), and ‘biosynthesis of secondary metabolites’ (57 genes) (Figure 6B; Table S1).

The 2,982 downregulated DEGs belonged to 32 GO terms and 10 KEGG pathways (Table S1). Notable GO terms included ‘response to hormone’ (63 genes), ‘response to endogenous stimulus’ (63 genes), ‘response to organic substance’ (72 genes), ‘response to chemical’ (86 genes), ‘SCF ubiquitin ligase complex’ (9 genes), ‘plasma membrane’ (86 genes), ‘cell periphery’ (99 genes), ‘ubiquitin conjugating enzyme activity’ (9 genes), ‘DNA-binding transcription factor activity’ (86 genes), and ‘transcription regulator activity’ (91 genes) (Figure 6A). Furthermore, the main enriched KEGG pathways were ‘plant hormone signal transduction’ (34 genes), ‘plant−pathogen interaction’ (14 genes), and ‘biosynthesis of secondary metabolites’ (72 genes) (Figure 6B; Table S1).

Interestingly, the term enrichment analysis results for IP24h ovaries were highly similar to those of the XP24h ovaries (Figure S1B; Table S1). In contrast, the GO terms and KEGG pathways for the 65 DEGs (11 up-regulated and 54 downregulated) identified in IP6h ovaries were different. Notable GO terms and KEGG pathways in IP6h ovaries were ‘deoxyribonucleotide biosynthesis process’ ‘nucleosome assembly’, ‘chromatin remodelling’, ‘nucleosome’, ‘DNA packing complex’, ‘asparagine synthase (glutamine-hydrolysing) activity’, ‘purine metabolism’ and ‘pyrimidine metabolism’ (Figure S1A; Table S1). These findings further supported the notion that partial pollen tube might have limited effects on the remodelling of genetic materials. The changes induced might be essential preparations for cell division, but they are likely not sufficient to initiate fruit set in tomato.

To explain the histological results that XP24h led to increased cell layer numbers and cell size at 2 DAA, we examined the expression patterns of some genes which are well-known for regulating cell division and expansion at the early stage of tomato fruit development. As illustrated in Figure 7A, thirteen members belonging to three subclasses (A, B, and D) of the cyclin family were significantly upregulated in XP24h and IP24h ovaries; highly expressed genes were cyclin B1_2 (Solyc10g080950) and cyclin B2_7 (Solyc03go32190). Likewise, five members of the expansin gene family and one gene encoding an expansin precursor (Soly02g088100) showed up to 3.6-fold higher expression in XP24h and IP24h ovaries (Figure 7B), than in IP6h and DP24h ovaries.

The vital role of phytohormones in regulating tomato fruit initiation and development was illustrated consistently through previous reports (Fenn and Giovannoni, 2021). In the present study, the KEGG enrichment analysis results also revealed that many DEGs were involved in the ‘plant hormone signal transduction’ pathway (Figure 6B). We therefore examined the expression pattern of genes involved in the synthesis, transportation, and signalling of various hormones (Figure 7C; Table S2). As indicated in Figure 7C, ethylene and auxin seemed to be the most active hormones in XP24h and IP24h ovaries. The expression of 8 genes related to ethylene (3 encoding ethylene receptors and 5 encoding ethylene biosynthesis enzymes) and 12 genes related to auxin (10 involved in auxin response and signalling, and 2 involved in auxin transport) were changed significantly. Specifically, there was significant downregulation (log₂FC = -2) of ARF7 and ARF5, which are key transcription factors that regulate fruit set and early fruit development. In addition, XP24h and IP24h ovaries displayed increased expression of two genes that positively regulate gibberellin synthesis (GA20ox1 and GA20ox2), and repression of GA2ox4, a negative controller of gibberellin catabolism (Figure 7C). Altogether, these results illustrated that complete penetration of pollen tubes (as in XP24h) might broadly affect hormonal responses to activate both cell division and cell expansion events by increasing the expression of cyclin and expansin genes.

In this study, pollination by X-ray-irradiated pollen resulted in a considerably high fruit set ratio (78.3%) of parthenocarpic fruits (Figure 1B), with an average weight approximately 65% that of seeded fruits on the same plant (Figure 3A). These effects are comparable to many previously reported parthenocarpy mutations in tomato (Sharif et al., 2022). We therefore examined the expression patterns of 23 well-known parthenocarpic genes in XP24h ovaries, compared to IP6h, IP24h, and DP24h. As a result, we found that 10 of these genes were differentially expressed (log₂FC ≥ 2), particularly in XP24h and IP24h ovaries (Figure 8A). The top variable genes were GA20ox1 (log₂FC = 3.09), NCED1 (log₂FC = -2.75), GA2ox2 (log₂FC = -2.26), AGL6 (pat-k) (log₂FC = -2.17), and ARF7 (log₂FC = -2.02).

Although many genes related to auxin signalling and transport were differentially expressed in XP24h ovaries, IAA9 which is an early auxin-responsive gene remained unchanged (Figure 8A). To further examine whether IAA9 is involved in parthenocarpic fruit development from XP24h ovaries, we pollinated the flowers of iaa9-3, a well-known parthenocarpic mutant (Saito et al., 2011), with X-ray-irradiated pollen. Compared to the control (E), X-ray-irradiated pollen slightly increased fruit set by approximately 14% (Figure 8B), but had little effect on cell layer formation (Figure 8C). In addition, the average weight of XP24h fruits (3.82 g) was similar to that of E fruits (Figure 8D), both of which were nearly 75% of the weight of IP24h fruits. These results suggested that full pollen tube penetration might regulate auxin signalling via altering the interaction between ARF7 and IAA9 proteins.