Gynoecium and fruit development in Arabidopsis

Multicellular life on Earth is inconceivable without plants. Our planet harbors close to 400,000 vascular plant species, of which approximately 94% are seed plants (Govaerts, 2001; Willis, 2017), which produce and reproduce themselves via seed. The vast majority of seed plants are angiosperms, which produce flowers, such as Arabidopsis thaliana of the Brassicaceae family (Fig. 1A). Flower and seed formation are adaptive advantages of plant sexual reproduction, which requires male and female gametes. In general, flowers have four different types of floral organs placed in four whorls, from outside to inside: sepals, petals, stamens and carpels (see Glossary, Box 1). The stamen, also called the androecium, is the ‘male’ part of the flower that produces the male gametophyte from which gametes differentiate. The ‘female’ part of the flower is formed by one or more carpels. In Arabidopsis, two congenitally fused carpels form the pistil, which is also called gynoecium (see Glossary, Box 1). Ovules are formed inside the gynoecium and the female gametophyte, which produces the female gametes, develops within each ovule (see Glossary, Box 1). Upon fertilization, seed development begins and, in most plant species, the gynoecium turns into a fruit (reviewed by Alvarez-Buylla et al., 2010; Dresselhaus et al., 2016).

The identification of AGAMOUS (AG) as a transcription factor controlling male and female reproductive organ development in plants (androecium and gynoecium, respectively), marked the beginning of a journey towards the understanding of the molecular aspects of flower formation (Bowman et al., 1989; Irish, 2017; Yanofsky et al., 1990). During three decades of research, many master regulators, such as SEPALLATAs (SEPs), CRABS CLAW (CRC), SPATULA (SPT), SHATTERPROOFs (SHPs), FRUITFULL (FUL) and SEEDSTICK (STK), among many others, have emerged as crucial regulators of reproductive development, especially of gynoecium development (reviewed by Alvarez-Buylla et al., 2010; Bowman et al., 1999; Ferrándiz et al., 2010; Reyes-Olalde et al., 2013; Roeder and Yanofsky, 2006; Simonini and Østergaard, 2019; Zúñiga-Mayo et al., 2019). Besides the identification of more transcription factors involved in gynoecium development, information on genes acting downstream of them have also been discovered (reviewed by Pajoro et al., 2014). Together, these discoveries have opened the road for charting gene regulatory networks (GRNs). In addition, several datasets from chromatin immunoprecipitation followed by sequencing (ChIP-seq) experiments, have allowed genome-wide analysis of binding events for many master regulators that participate in the transition to reproduction and flower development (Chen et al., 2018). Now, the fine-tuning aspects of flower development are starting to be revealed.

Over the years, many review articles on flower and fruit development have been published (Alvarez-Buylla et al., 2010; Ballester and Ferrándiz, 2017; Bowman et al., 1999; Ferrándiz et al., 2010; Marsch-Martínez and de Folter, 2016; Reyes-Olalde and de Folter, 2019; Reyes-Olalde et al., 2013; Roeder and Yanofsky, 2006; Sehra and Franks, 2015; Simonini and Østergaard, 2019; Smyth et al., 1990; Zúñiga-Mayo et al., 2019). However, there is currently no recent review that includes all the current GRNs known to date for gynoecium and fruit development. In 2015, we proposed approaches and tools to construct a comprehensive GRN for gynoecium development (Chávez Montes et al., 2015). Therefore, this Review focuses on recent discoveries and the integration of the GRNs that guide the development of Arabidopsis, starting from gynoecium initiation until fruit maturation.

Plants possess the ability to form new organs continuously (reviewed by Gaillochet and Lohmann, 2015; Sablowski, 2007). When environmental and endogenous genetic cues are met, the shoot apical meristem (SAM; see Glossary, Box 1) transitions to an inflorescence meristem (IM; see Glossary, Box 1) (Fig. 1B,E), marking the beginning of the reproductive phase of the plant (reviewed by Andrés and Coupland, 2012). At the flanks of the IM, auxin induces differentiation to give rise to flower primordia, each with a floral meristem (FM; see Glossary, Box 1). In Arabidopsis, the most recently formed flower primordium is referred to as ‘stage 1’ (Smyth et al., 1990). When the next flower primordium is formed, the previous one is called stage 2, and so on. At floral stage 3, the flower primordium increases in size and sepal primordia become visible (Fig. 1B,C,F). The FM gives rise to the different floral organs present in a mature flower (Fig. 1D,F) (reviewed by Denay et al., 2017). During the next stages, sepals continue to grow; during stage 5, petal and stamen primordia appear, whereas stage 6 is characterized by the complete coverage by the sepals of the FM and internal flower primordia, including the gynoecium primordium (see Glossary, Box 1), which becomes visible at stage 6 (Fig. 1F,G) (Alvarez-Buylla et al., 2010; Denay et al., 2017; Smyth et al., 1990).

In general, the genes that regulate meristem maintenance in the SAM and the FM are shared (reviewed by Chang et al., 2020; Gaillochet and Lohmann, 2015). The role of the WUSCHEL-CLAVATA3 (WUS-CLV3) circuit in plant stem cell niche maintenance is key (Brand et al., 2000; Schoof et al., 2000). The activity of WUS, together with LEAFY (LFY) and others, converges on the regulation of AG, which starts to be expressed at stage 3 in the FM (Fig. 2A). In turn, AG starts to repress WUS and activates several genes that negatively regulate WUS expression (Fig. 2).

The meristematic activity of the FM ends when the gynoecium primordium is formed at stage 6; however, the programs controlling FM termination begin at stage 3. The molecular mechanisms underlying FM termination have recently been reviewed (Chang et al., 2020; Lee et al., 2019; Shang et al., 2019; Xu et al., 2019; Zúñiga-Mayo et al., 2019). The GRNs that control FM termination (Fig. 2) and gynoecium initiation (Fig. 2B) contain at least 15 transcription factors and various other proteins. AG plays a leading role in both these processes, as illustrated in the ABC and quartet model for floral organ formation (reviewed by Coen and Meyerowitz, 1991; Theißen and Saedler, 2001). AG is therefore observed in the GRNs as the main hub (Fig. 2A). AG activates pathways related to auxin and cytokinin signaling (Ó’Maoiléidigh et al., 2018; Yamaguchi et al., 2017). Cytokinin signaling, in turn, also controls AG (Gómez-Felipe et al., 2021; Rong et al., 2018). The effects of AG in the control of FM termination and gynoecium development must be tightly controlled to perform those functions properly. Indeed, the mechanisms that control AG expression are diverse, involving transcription factors, microRNAs, histone deacetylase complexes and RNA-binding complexes (reviewed by Pelayo et al., 2021). Interactions with other proteins and the formation of protein complexes provide additional layers of control of AG activity (Box 2). For example, the interaction with SEP3, via the formation of dimers or tetramers, controls different biological processes (Hugouvieux et al., 2018; Lai et al., 2020).

Over the past few years, some new nodes have been included in the networks that underlie gynoecium initiation. AINTEGUMENTA (ANT), which has reported roles in the establishment of flower primordia and the development of floral organs, has now been identified to function at very early stages of flower formation (stages 3-6), and the roles of ANT are shared with AINTEGUMENTA-LIKE 6 (AIL6) (Krizek et al., 2020, 2021). At stage 3, both proteins directly regulate AG, as well as other genes, such as MONOPTEROS (MP) and REVOLUTA (REV), which are related to meristem activity regulation and new primordia formation (Krizek et al., 2021) (Fig. 2A). At stage 6, ANT positively regulates SPT, an important gene for tissue development in the medial domain, as described below (Fig. 2B) (Krizek et al., 2020).

Epigenetic silencing of AG is illustrated by the participation of POLYCOMB-GROUP (PcG) complexes with CURLY LEAF (CLF), EMBRYONIC FLOWERING 1 (EMF1) and EMF2. In a similar way, APETALA 2 (AP2) recruits TOPLESS (TPL) and HISTONE DEACETYLASE 19 (HDA19) to repress AG (Fig. 2A) (Pelayo et al., 2021). Now, some evidence points to the involvement of PcG in silencing of WUS (Sun et al., 2019) and the modification of chromatin in the control of auxin biosynthesis by AG and CRC (Fig. 2) (Yamaguchi et al., 2018).

KNUCKLES (KNU), a C2H2 zinc-finger transcription factor, also functions during FM termination in addition to its role in repressing WUS in the FM (Fig. 2B). A recent analysis of its expression pattern has revealed the presence of KNU in the CLV3 expression domain. The extended functions of KNU include suppression of the expression of several floral meristem regulators, such as CLV1 and CLV3, at stage 6 (Kwaśniewska et al., 2021; Shang et al., 2021).

The transcription factor ETTIN (ETT), also known as AUXIN RESPONSE FACTOR 3 (ARF3), is involved in gynoecium patterning (Heisler et al., 2001). An additional function of ETT is to inhibit cytokinin biosynthesis by repressing ISOPENTENYLTRANSFERASE (IPT) and LONELY GUY (LOG) genes, and the gene encoding the cytokinin receptor ARABIDOPSIS HISTIDINE KINASE 4 (AHK4) during stages 5 and 6. Through these mechanisms, ETT contributes to FM termination (Fig. 2B) (Zhang et al., 2018).

Conversely, cytokinin also activates transcription factors important for gynoecium initiation and patterning, which starts around stage 6 (Fig. 2B). For example, the master regulator AG is regulated by cytokinin via the Arabidopsis response regulators type-B (ARRs) (Gómez-Felipe et al., 2021; Rong et al., 2018). Furthermore, other transcription factors that act in parallel downstream of AG are CRC, SHP2 and SPT, which are important for gynoecium initiation and patterning, as well as being positively regulated by ARR type-B transcription factors (Fig. 2B) (Gómez-Felipe et al., 2021).

Many transcription factors involved in gynoecium development (Reyes-Olalde et al., 2013) are also expressed during gynoecium initiation. In total, over 60 genes are expressed at stage 4 and over 65 genes at stage 6 (Table S1) (Herrera-Ubaldo et al., 2018 preprint; Jiao and Meyerowitz, 2010). Current studies are revealing novel roles for well-known regulators; additionally, previously unreported regulators and interactions are continuously being discovered. Therefore, we expect that these GRNs will expand in the future. This expansion of GRNs also applies to the GRNs described for the subsequent stages of gynoecium and fruit development.

The transition from a gynoecium primordium into two, well-defined domains (lateral and medial; see Glossary, Box 1) marks the beginning of patterning and the formation of internal tissues (Fig. 1G; Fig. 3). During stages 7 and 8, the activation and maintenance of the carpel margin meristem (CMM; see Glossary, Box 1) is crucial for the subsequent development of specialized structures: in the lateral domains, valves (see Glossary, Box 1) develop to protect the ovules, and seeds are subsequently formed in the medial domain. Additional tissues, such as the septum and the transmitting tract (see Glossary, Box 1), also form in the medial domain to facilitate fertilization (Fig. 1G).

The process of gynoecium differentiation into medial and lateral domains requires several transcription factors and hormonal signals (Reyes-Olalde and de Folter, 2019; Reyes-Olalde et al., 2013). One important regulatory module involves the integration of auxin and cytokinin signaling with the activity of the transcription factors HECs and SPT in the medial domain around floral stages 7 and 8 (Müller et al., 2017; Reyes-Olalde et al., 2017; Schuster et al., 2015) (Fig. 3C). To summarize, auxin signaling is present in the lateral domains and cytokinin signaling in the medial domain in the CMM. In the medial domain, cytokinin signaling induces auxin biosynthesis and auxin transporters to create an auxin flux to transport auxin to the lateral domain. Auxin-induced proteins in the lateral domain repress cytokinin signaling genes to limit them to the medial domain, such as the cytokinin signaling inhibitor ARABIDOPSIS HISTIDINE PHOSPHOTRANSFER PROTEIN 6 (AHP6) and the auxin response factor ETT (Heisler et al., 2001; Reyes-Olalde et al., 2017). The created auxin flux is important for the apical-basal growth of the gynoecium. Cytokinin biosynthesis in the CMM is assumed to be regulated by SHOOT MERISTEMLESS (STM), as has been shown for other meristems (Jasinski et al., 2005; Yanai et al., 2005). STM is a KNOTTED (KNAT) 1-LIKE homeodomain transcription factor important for CMM formation and activity (Scofield et al., 2007). The STM gene is activated by the CUP-SHAPED COTYLEDON proteins (CUCs) (Kamiuchi et al., 2014). A positive regulatory loop between STM and CUC genes has been reported in the SAM and could be present in the CMM as well (Spinelli et al., 2011). The CUC transcription factors seem to also be important for SPT expression in the basal part of the gynoecium (Nahar et al., 2012).

The current GRN that controls activities in the CMM has been derived from functional studies of genes and reporter lines (reviewed by Reyes-Olalde and de Folter, 2019; Reyes-Olalde et al., 2013). The study and integration of additional regulators is currently in progress; there are more than 50 transcription factors related to gynoecium development with reported expression in this region (Table S1) (Herrera-Ubaldo et al., 2018 preprint). The coordination of biochemical and genetic processes underlying meristematic activity in the CMM, and its further differentiation, probably involves the participation of many proteins and pathways that are yet to be characterized (Kivivirta et al., 2021; Villarino et al., 2016; Wynn et al., 2011).

One of the most important functions of the gynoecium is to provide protection to the developing ovules. The determination of ovule identity occurs at the flanks of the CMM. Ovule primordia are formed at stage 9 by periclinal cell divisions within the epidermal tissue of the placenta (see Glossary, Box 1) (Fig. 1G; Fig. 4). Like previous examples, the GRNs involve the coordinated action of several transcription factors and hormonal signaling pathways that control the main aspects of ovule development: the establishment of boundaries and primordia (Fig. 4A), the control of ovule initiation and number, and ovule patterning (Fig. 4B,C) (Barro-Trastoy et al., 2020b; Cucinotta et al., 2014, 2020).

The intricate mechanisms that guide ovule initiation are far from fully understood, but advances have been made (Barro-Trastoy et al., 2020b; Cucinotta et al., 2020). Indeed, the identification of enzymes and proteins that act at different levels outside the GRN has shed light on fine aspects of the molecular mechanisms, and opens new questions and directions regarding our understanding of ovule development. For example, factors have been recently identified that control the spacing of ovules: two secreted peptides and their ERECTA (ER) family receptor kinases coordinate regular ovule primordia initiation coupled to fruit growth from the placenta and carpel walls (Kawamoto et al., 2020).

Processes that occur at the cell-wall level have also been discovered. A recent study has reported on the involvement of placenta-expressed genes encoding CELL WALL INVERTASE (CWIN) 2 and 4 as positive regulators of ovule initiation. CWINs hydrolyze sucrose into glucose and fructose, and have additional roles in sugar signaling. Specific suppression of the activity of these enzymes leads to a significant reduction in ovule number, as well as ovule and seed abortion. Based on transcriptome analysis in CWIN2- and CWIN4-silenced lines, it has been shown that the ovule-identity gene STK, as well as auxin signaling-related genes and genes encoding hexose transporters are downregulated (Liao et al., 2020).

Also at the membrane level, the localization of PIN auxin transporters in the placenta epidermis is crucial for ovule initiation (Fig. 2A). The distribution of PIN1 and auxin fluxes is very dynamic in the placenta and affects not only one ovule primordium, but also the neighboring ovule primordium. Analysis of marker lines and auxin response has revealed that ovules initiate asynchronously and follow this trend through later stages (Yu et al., 2020).

On average around 60 ovules are formed inside the gynoecium (reviewed by Barro-Trastoy et al., 2020b; Cucinotta et al., 2014, 2020; Yuan and Kessler, 2019). However, ovule numbers can range between 40 and 80, depending on the accession of Arabidopsis (Yuan and Kessler, 2019). Some modules of the GRN that control ovule number can also control gynoecium size (reviewed by Cucinotta et al., 2020). A recent study has uncovered a previously unidentified positive regulator of ovule number: NEW ENHANCER OF ROOT DWARFISM1 (NERD1) (Yuan and Kessler, 2019).

Transcription factors, such as MP, ANT and CUCs, are interconnected with hormone homeostasis mechanisms, mainly with auxins, cytokinins, brassinosteroids and gibberellins, to regulate ovule number (Fig. 4A) (reviewed by Barro-Trastoy et al., 2020b; Cucinotta et al., 2020; Qadir et al., 2021). As mentioned above, auxin positively regulates ovule primordia initiation and cytokinin signaling positively regulates ovule number (Bartrina et al., 2011; Cerbantez-Bueno et al., 2020; Cucinotta et al., 2018; Galbiati et al., 2013; Reyes-Olalde et al., 2017; Zúñiga-Mayo et al., 2018). It has now been shown that cytokinin metabolism in the epidermis of the placenta is also important for ovule number (Werner et al., 2021). Furthermore, brassinosteroids and gibberellins positively and negatively regulate ovule number, respectively (Barro-Trastoy et al., 2020a; Huang et al., 2013; Nole-Wilson et al., 2010; Gomez et al., 2018, 2019).

After the establishment of ovule identity and ovule number definition (Fig. 4A), ovule patterning takes place, which involves differentiation of the funiculus, chalaza and nucellus (see Glossary, Box 1) at stage 10 (Fig. 4B), and continues with the formation of the inner and outer integuments at stage 11 (Fig. 4C). Various recent reviews on these stages are available (Barro-Trastoy et al., 2020b; Pinto et al., 2019), and an in-depth discussion of this topic goes beyond the scope of this article. However, one important transcription factor, STK, is the master regulator of ovule identity (Favaro et al., 2003; Pinyopich et al., 2003), although this gene has many additional functions, as described below. Various other genes involved in ovule patterning have additional functions during gynoecium development.

During stages 11 and 12, gynoecium patterning continues to form all the tissues, and the gynoecium attains the final shape suitable for pollination (Fig. 1D). During this period, there is active patterning in the main axes: the abaxial-adaxial with funiculus, chalaza and nucellus (summarized in Fig. 5); the medial-lateral with the valves, valve margins, replum (see Glossary, Box 1), septum and transmitting tract (Fig. 5A); and the apical-basal with stigma (see Glossary, Box 1), style (see Glossary, Box 1), ovary and gynophore (Fig. 5B) (Chávez Montes et al., 2015; Deb et al., 2018; Marsch-Martínez and de Folter, 2016; Simonini and Østergaard, 2019; Zúñiga-Mayo et al., 2019). Most of the best-characterized regulators in these GRNs are transcription factors that affect tissue differentiation.

In a mature gynoecium, the medial-lateral axis is composed of the valves (the carpel walls), which protect the developing ovules. After pollination, the valves are attached to the replum (Fig. 1F,G; Fig. 5A). The formation of this axis (valves-replum-valves) requires the concerted action of medial factors, such as BP, RPL, WUSCHEL-RELATED HOMEOBOX 13 (WOX13) and/or NO TRASMITTING TRACT (NTT), which possess meristematic-related functions and repress the action of lateral factors, such as FILAMENTOUS FLOWER/YABBY3 (FIL/YAB3) and the ASYMMETRIC LEAVES 1/2 (AS1/2) (Alonso-Cantabrana et al., 2007; Dinneny et al., 2005; Marsch-Martínez et al., 2014; Romera-Branchat et al., 2013). Additionally, the so-called ‘boundary factors’ (CUCs and KNAT2/6) participate in between, marking the division of the meristematic and lateral organ functions (Fig. 5A) (reviewed by Ballester and Ferrándiz, 2017; Simonini and Østergaard, 2019).

Inside the ovary, transmitting tract formation is controlled by NTT (Crawford et al., 2007), the basic helix-loop-helix (bHLH) proteins CESTA/HALF FILLED (CES/HAF), BRASSINOSTEROID ENHANCED EXPRESSION 1 and 3 (BEE1, BEE3) (Crawford and Yanofsky, 2011), HECs (Gremski et al., 2007), SPT (Heisler et al., 2001) and STK (Di Marzo et al., 2020b; Herrera-Ubaldo et al., 2019) (Fig. 5A). In addition, several of these factors (NTT, SPT and STK) control the development of other tissues in the gynoecium that are related to transmitting tract formation; they control programmed cell death (PCD) and the deposition of extracellular matrix (ECM) (reviewed by Crawford and Yanofsky, 2008; Pereira et al., 2021).

Some processes occurring between stages 11 and 12 highlight the importance of cell wall modifications in medial-lateral gynoecium patterning. The leading role of ETT in gynoecium morphology is clear; ETT is known to participate, in coordination with auxin input, in the guidance of early patterning events (Fig. 2B; Fig. 3) (Sessions et al., 1997; Simonini et al., 2016). Now, an additional pathway of ETT participation has been identified, the regulation of pectin methylesterase (PME) activity in the valves (Andres-Robin et al., 2018). PME activity in valve cell walls increases the levels of demethylesterified pectins, allowing the reduction of cell wall stiffness and the elongation of the valves.

Inside the gynoecium, other modifications need to occur in the medial domain to allow the formation of the transmitting tract. Recently, NTT and STK have been shown to participate in the modulation of the cell wall composition by regulating a gene encoding a putative mannanase enzyme, as well as other genes encoding proteins involved in lipid biosynthesis and transport (Herrera-Ubaldo et al., 2019).

Major changes must be coordinated to achieve the distinct apical-basal features (stigma, style and ovary) observed in a mature gynoecium. The ovary is a bilateral structure – in contrast to the style, which exhibits radial symmetry (Fig. 1F,G). During floral stage 12, the style differentiates and the bilateral-to-radial transition is controlled by the modulation of auxin flux, as well as cytokinin sensitivity (Fig. 5B) (Carabelli et al., 2021; Moubayidin and Ostergaard, 2014). First, the transcription factors SPT and INDEHISCENT (IND) promote apolar localization of PIN proteins (auxin efflux transporters normally localized to cell poles to create polar auxin transport; Moubayidin and Ostergaard, 2014). SPT and HECs then control the expression of the adaxial-identity genes HOMEOBOX ARABIDOPSIS THALIANA 3 (HAT3) and ARABIDOPSIS THALIANA HOMEOBOX 4 (ATHB4), which orchestrate the final steps of the radicalization process (Fig. 5B) (Carabelli et al., 2021). In another study focusing on style development using genetics and protein-protein interaction experiments, it has been suggested that ETT, IND, BREVIPEDICELLUS (BP), REPLUMLESS (RPL) and SEUSS (SEU) work synergistically to regulate style morphology (Fig. 5B) (Simonini et al., 2018).

Sequential activation and function of transcription factors is required for stigma development: the NGATHA (NGA) and HECATE (HEC) protein families cooperatively regulate stigma development through the activation of IND, which later interacts with NGA (and probably HECs) to activate SPT. SPT is then probably integrated into the NGA-IND-HEC complex to activate target genes (Ballester et al., 2021). Recently, three angiosperm-specific regulators of the style and stigma have been identified called STIGMA AND STYLE STYLIST 1-3 (SSS1-3), which belong to a previously unreported family. They are expressed in the apical tissues of the gynoecium and act downstream of the NGA transcription factors to fine-tune the development of the stigma and style (Fig. 5B) (Li et al., 2020).

Pollination and fertilization involve a continuous and active communication between female and male tissues over several steps: pollen hydration, pollen germination, pollen tube growth, pollen tube attraction to the ovule, pollen tube reception, sperm cell delivery and gamete activation (reviewed by Cascallares et al., 2020).

As discussed earlier, the formation of the transmitting tract includes regulation of PCD and ECM deposition (reviewed by Pereira et al., 2021), as well as other changes in cell wall composition in the medial region (Herrera-Ubaldo and de Folter, 2018). The ECM provides nutrition, adhesion and guidance during pollen tube growth. During fertilization, the transmitting tract secrets chemical signals, such as arabinogalactan-proteins (AGPs), whereas other signals, such as LURE proteins, derive from the ovules to attract pollen tubes (reviewed by Johnson et al., 2019; Pereira et al., 2021).

Checkpoints during the pollination process control the acceptance or rejection of the pollen prior to fertilization (reviewed by Dresselhaus et al., 2016) and various small peptides and receptors are involved in these processes (reviewed by Zhang et al., 2021). Recently, some new discoveries have been made towards the understanding of communication and signaling cascades during pollination. One such discovery is the existence of an autocrine signaling pathway acting at the surface of stigmatic papillae, inducing reactive oxygen species (ROS) production; ROS levels are reduced upon pollination owing to an antagonistic peptide from the pollen coat, allowing pollen hydration and germination (Liu et al., 2021; Zhou et al., 2021). Another discovery also involves the receptor kinase FERONIA (FER) in the control of male-female interactions. FER, located at the entrance to the female gametophyte, mediates de-esterified pectins and the production of nitric oxide (NO), which is necessary for when the first pollen tube arrives at the ovule; NO accumulation stops the attraction of further pollen tubes (Duan et al., 2020). The process of pollination, and the subsequent events, can only occur in a specific time window, delimited by stigmatic receptivity. Two transcription factors, KIRA1 (KIR1) and ORESARA1 (ORE1), belonging to the NAC family promote PCD in the stigma cells, thereby controlling stigma longevity (Gao et al., 2018).

After fertilization, the gynoecium becomes a fruit, and its development in Arabidopsis continues over 12 days to finally release fully formed, fertile seeds. During the final stages of fruit development, the fruit grows, the seeds mature and developmental programs prepare for fruit dehiscence (see Glossary, Box 1) and seed dispersal (Alvarez-Buylla et al., 2010; Ballester and Ferrándiz, 2017; Roeder and Yanofsky, 2006). After fertilization (stage 13), the fruit grows mainly longitudinally to reach its maximum size by stage 17, followed by the start of senescence of the fruit (stage 18), and finally fruit dehiscence and seed release (stage 20) (Fig. 6A-C).

The involvement of cytokinin in fruit elongation has recently been reported: phenotypic alterations observed in fruit length of the stk mutant, which produces short fruits, resemble those of fruits of the ckx7 mutant (encoding for the enzyme CYTOKININ OXIDASE/DEHYDROGENASE 7), in which cytokinin degradation is affected (Di Marzo et al., 2020a). Therefore, cytokinin has a negative effect on fruit elongation. This additional function of STK also integrates pathways that control fruit size because STK positively regulates CKX7 expression in the fruit and indirectly controls FUL, which is involved in a pathway with miR172 and AP2 (Fig. 6E) (Di Marzo et al., 2020a; Ripoll et al., 2015). Additionally, STK regulates fruit growth by controlling the expression of α-XYLOSIDASE 1 (XYL1), which is involved in the modification of xyloglucans (XyGs), which form a component of cell walls. Indeed, XyG modifications contribute to the mechanical properties of the cell wall, affecting its extensibility and thus cell growth. The activity of XYL1 is also required in the growing seeds in order for them to achieve their final size (Di Marzo et al., 2021).

The involvement of hormones in fruit development and maturation goes beyond just auxin and cytokinin. Gibberellins induce fruit growth in Arabidopsis (Dorcey et al., 2009) and regulate the formation of the separation layer in the valve margins (Fig. 6D,F) (Arnaud et al., 2010). The role of hormones, such as abscisic acid, ethylene and brassinosteroids, are better described for fleshy fruit development (reviewed by Li et al., 2021; McAtee et al., 2013; Sotelo-Silveira et al., 2014).

Related to longitudinal growth of the silique (see Glossary, Box 1) or fruit, a recent study has reported a spatiotemporal map of fruit growth at the cellular level that captures quantitative data to measure cell division and cell expansion. Analysis of the ntt mutant affected in seed-set revealed that final fruit length correlates with the number of seeds (Ripoll et al., 2019).

Early studies on fruit development have described the initial models for fruit patterning related to fruit opening (dehiscence or pod shattering), highlighting the key roles of the transcription factors FUL, SHPs, RPL, ALC and IND (Ballester and Ferrándiz, 2017; Liljegren et al., 2004). The dehiscence zone, located on the valve margins, are specialized cell layers that allow valve detachment. SHP proteins control the expression of IND, which guides the formation of the lignification layer (Fig. 6D,F). Additionally, SHPs control ALCATRAZ (ALC), which guides the formation of the separation layer (reviewed by Ballester and Ferrándiz, 2017). Moreover, FUL controls the formation of a lignified layer in the valves (Liljegren et al., 2004).

Another role of STK is the control of lignification in the funiculus, which is necessary for seed abscission (Balanzà et al., 2016). In the funiculus, STK, together with the co-repressor SEU, represses the expression of HEC3, SPT and ALC (Fig. 6F). A rather similar process occurs in valve margin lignification, whereby SHPs regulate IND, SPT and ALC, but the regulation is inverted (i.e. SHPs activate gene expression) (Fig. 5A; Fig. 6F).

Important roles of hormones have been uncovered as well. IND regulates the expression of PINOID (PID) and WAG2 to control auxin transport outside the separation layer in the valve margins, creating a so-called ‘auxin minimum’ that is important for dehiscence zone formation (Sorefan et al., 2009). However, further studies regarding the participation of auxin in the control of dehiscence, using reporter lines to track auxin distribution have revealed that, at early fruit-development stages (stage 14-16), auxin is present at the valve margin for specifying the dehiscence zone (van Gelderen et al., 2016). Perhaps the ‘auxin minimum’ at later fruit stages (e.g. stage 17B) is still functionally relevant for cell separation (reviewed by Ballester and Ferrándiz, 2017; de Folter, 2016). The involvement of cytokinin signaling during fruit development has also been studied. The valve margins display cytokinin signaling, which is lost in mutants that are indehiscent (ind and shp1 shp2). When cytokinin is applied to these mutants, fruit dehiscence is restored (Marsch-Martínez et al., 2012). Furthermore, replum development is also dependent on the presence of cytokinin; with less cytokinin or cytokinin signaling, the width of the replum is reduced and vice versa (Marsch-Martínez et al., 2012; Reyes-Olalde et al., 2017). In a double Arabidopsis ntt rpl mutant that lacks the replum, replum formation is restored upon cytokinin application, suggesting that cytokinin signaling acts downstream of NTT and RPL (Zúñiga-Mayo et al., 2018). In addition, increased replum width has also been observed when NTT is overexpressed (Marsch-Martínez et al., 2014).

In addition to the hormonal and genetic control of dehiscence, a recent study adds the importance of environmental input. Temperature affects the expression of IND; higher temperature causes chromatin modifications at the IND locus, which results in increased IND expression that accelerates valve margin development (Fig. 6G) (Li et al., 2018).

The concerted action of GRNs during these stages will ultimately guide fruit growth and shape. Although a lot of information is available, further players are likely still to be discovered. A recent study has conducted transcriptome analysis to reveal gene expression dynamics during fruit growth and maturation, including silique samples, from 3, 6, 9 and 12 days after pollination (Mizzotti et al., 2018). This dataset is supported by another study, which has analyzed the pre-fertilization stages: carpel initiation (stage 5), elongation of carpel walls (stage 9), gynoecia during female meiosis (stage 11), and gynoecia before anthesis (stage 12) (Kivivirta et al., 2021). These two studies (and others, e.g. Klepikova et al., 2016; Villarino et al., 2016; Wynn et al., 2011) now provide a broad gene expression atlas for the complete developmental window from gynoecium-fruit development. In both works, samples included complete gynoecia and fruits, so additional work is required to resolve gene expression further or dissect transcriptomic data at the tissue- and single-cell level.

The current knowledge on gynoecium and fruit development is expanding. We have given an overview of the GRNs known to date. In the future, it is likely that these GRNs will expand because we know that many more transcription factors are expressed at the specific stages discussed (Table S1). Furthermore, many recent works have provided new datasets and information related to gynoecium formation, such as transcriptomics during development (Kivivirta et al., 2021; Mizzotti et al., 2018) or at specific times or scenarios (Krizek et al., 2021; Liao et al., 2020; Martínez-Fernández et al., 2020; Yu et al., 2020); variation by genome-wide association studies (Yuan and Kessler, 2019) or quantitative trait locus analysis (Kawamoto et al., 2020); and protein-protein interactions (Herrera-Ubaldo et al., 2018 preprint). The presented GRNs are based on data of Arabidopsis it is likely that these GRNs have variations and/or rewiring that may explain existing diversity in gynoecia and fruits of other species (Box 3).

Current technologies allow the study of morphogenesis in Arabidopsis in unprecedented detail. Live imaging, lineage tracking and geometric analysis, in combination with single-cell transcriptome profiling, are changing how we study plants and paving the way to study development in four dimensions. Some important advances have been made during early flower development, such as the construction of a 4D atlas that integrates cell growth quantification and gene expression data (Refahi et al., 2021); a 3D gene expression atlas of the FM based on the spatial reconstruction of single-nucleus RNA-sequencing data (Neumann et al., 2021 preprint); and a study of sepal (Zhu et al., 2020) and stamen growth (Silveira et al., 2021). In the case of ovule formation, the generation of a 3D reference atlas (Vijayan et al., 2021) or the analysis of the relationship between organ geometry and cell fate (Hernandez-Lagana et al., 2021) provide comprehensive views of ovule development.

These works are related to reproductive structures, but the field can also take advantage of broader studies performed at the whole-plant level. For example, the protein interactome for hormone-related proteins (Altmann et al., 2020) or the Arabidopsis proteome with samples from whole flowers and fruits (as well as carpels, seeds, valves and septum samples; Mergner et al., 2020), provide useful resources and valuable information to be integrated. Additionally, a recent dataset has facilitated the study of protein complexes at the pan-plant level (McWhite et al., 2020), and another allows the comparison of transcriptomic programs across land plants (Julca et al., 2021). Finally, the Plant Cell Atlas initiative will provide insight into the content and processes taking place in each cell type in the plant (Plant Cell Atlas Consortium et al., 2021; Rhee et al., 2019). The integration of all these data in the coming years will reveal many hidden aspects of morphogenesis during gynoecium and fruit formation and will achieve a detailed picture of the mechanisms involved, at a similar level to other systems, such as plant embryo patterning (Harnvanichvech et al., 2021).

We apologize to researchers in the field whose work has not been cited owing to space constraints. We thank Andrea Gómez-Felipe and Victor Zúñiga-Mayo for providing images for Fig. 1B,C and Fig. 1D, respectively. We also thank the two anonymous reviewers for their input.