Developmental Control and Plasticity of Fruit and Seed Dimorphism in Aethionema arabicum1[CC-BY]

Aethionema arabicum produces two different fruit and seed types with distinct anatomical, biomechanical, and physiological properties. Understanding how plants cope with changing habitats is a timely and important topic in plant research. Phenotypic plasticity describes the capability of a genotype to produce different phenotypes when exposed to different environmental conditions. In contrast, the constant production of a set of distinct phenotypes by one genotype mediates bet hedging, a strategy that reduces the temporal variance in fitness at the expense of a lowered arithmetic mean fitness. Both phenomena are thought to represent important adaptation strategies to unstable environments. However, little is known about the underlying mechanisms of these phenomena, partly due to the lack of suitable model systems. We used phylogenetic and comparative analyses of fruit and seed anatomy, biomechanics, physiology, and environmental responses to study fruit and seed heteromorphism, a typical morphological basis of a bet-hedging strategy of plants, in the annual Brassicaceae species Aethionema arabicum. Our results indicate that heteromorphism evolved twice within the Aethionemeae, including once for the monophyletic annual Aethionema clade. The dimorphism of Ae. arabicum is associated with several anatomic, biomechanical, gene expression, and physiological differences between the fruit and seed morphs. However, fruit ratios and numbers change in response to different environmental conditions. Therefore, the life-history strategy of Ae. arabicum appears to be a blend of bet hedging and plasticity. Together with the available genomic resources, our results pave the way to use this species in future studies intended to unravel the molecular control of heteromorphism and plasticity.

Fruits and seeds with very specific properties evolved as typical propagation and dispersal units to support the angiosperm life cycle in adaptation to the prevailing environment (Donohue et al., 2010;Linkies et al., 2010;Ferrandiz, 2011). Seeds provide the receptacle for the embryo and nutrients to aid germination and early seedling establishment. Innate morphological and physiological seed properties define the environmental conditions suitable for germination timing through various dormancy mechanisms (Finch-Savage and Leubner-Metzger, 2006) and thereby greatly influence seedling survival and plant fitness. Fruits are unique structures that enclose and protect angiosperm seeds (Scutt et al., 2006;Seymour et al., 2008). Like seeds, they have a large impact on the fate of a plant's offspring by providing various mechanisms for seed dispersal (Ferrandiz, 2011). Most plant species commit themselves to the propagation strategy of homomorphism, producing seeds and fruits of a single type that is optimally adapted to the respective habitat.
However, several angiosperm families independently evolved heteromorphism, characterized by the production of two or more distinct fruit and seed morphs on individual plants (Imbert, 2002). These differ in various properties, including fruit and seed size, shape, color, mechanisms of dispersal, dormancy, germination, and mucilage production upon imbibition (Takeno and Yamaguchi, 1991;Mandak and Pysek, 2001;Lu et al., 2010;Dubois and Cheptou, 2012;Baskin et al., 2014;Yang et al., 2015, and refs. therein). Heteromorphic plants thus can produce offspring with different fates, determined by the distinct properties of their fruits and seeds. Consequently, heteromorphism has been interpreted as a bet-hedging strategy in adaptation to unpredictable environments, where flexibility in terms of propagation is an important fitness advantage (Venable and Lawlor, 1980;Philippi and Seger, 1989;Evans and Dennehy, 2005;Abley et al., 2016). The facts that heteromorphic species occur primarily in stressful and frequently disturbed habitats (such as arid and semiarid environments) and that they mostly consist of annual species further support this conclusion (Venable et al., 1995;Imbert, 2002).
Heteromorphism represents a classical tradeoff. It may increase long-term reproductive success by reducing the risk of extinction, but it comes at the cost of decreasing the immediate fitness because only a fraction of propagules are optimally adapted to any given environment (Venable, 2007). Several studies suggest that at least some heteromorphic species diminish this problem by means of phenotypic plasticity, defined as the ability of a genotype to produce different phenotypes when exposed to different environmental conditions (Via et al., 1995;Sultan, 2000;Pigliucci et al., 2006;Abley et al., 2016). The fruit-morph ratio of heteromorphic species may vary in response to herbivory (Imbert and Ronce, 2001), nutrient availability and plant density (Mandák and Pyšek, 1999;Sadeh et al., 2009;Lu et al., 2013a), germination time (Yang et al., 2015), and soil moisture (Lu et al., 2013a), indicating that these plants can adjust their fruit development in response to certain environmental parameters. However, so far, very little is known about the molecular determinants of heteromorphism in general and the plastic developmental modulation of this phenotype in particular.
The monogeneric tribe Aethionemeae (genus Aethionema), the sister group to the rest of the Brassicaceae (core Brassicaceae), comprises 57 species with a distributional hotspot in the Middle East and Eastern Europe (Al-Shehbaz et al., 2006;Beilstein et al., 2010;Franzke et al., 2011). Heteromorphism in this lineage has been reported for six species (Solms-Laubach, 1901;Hedge, 1965), including Aethionema arabicum, a small diploid, annual, herbaceous plant whose genome sequence was published recently (Haudry et al., 2013). Analysis of the genome has shown that the genus Aethionema shares the ancient whole-genome duplication "At-alpha" with the crown-group Brassicaceae (Schranz et al., 2012); thus, it has been used for comparative molecular evolutionary analyses for several gene families (Hofberger et al., 2015;Mohammadin et al., 2015;Simon et al., 2015). The species has been described as forming two distinct fruit and seed morphs that may influence their propagation strategy (Solms-Laubach, 1901). We have now systematically analyzed the dimorphic phenotype of Ae. arabicum throughout the plant's life cycle. Our data characterize morphological and physiological features of the two distinct fruit and seed morphs and provide evidence for phenotypic plasticity. These findings, together with its phylogenetic position, available genome sequence, and life-history traits, make Ae. arabicum attractive for future research on both the proximate causes (molecular mechanisms) and the ultimate causes (selection regimes or genetic drift) of heteromorphism.

Heteromorphism Evolved at Least Twice within the Aethionemeae
Phylogenetic analysis (Fig. 1) of the genus Aethionema gives a tree topology with a well-supported backbone splitting in three clades: A, B, and C. Both Bayesian (Supplemental Fig. S1) and RaxML (Supplemental Fig.  S2) phylogenetic analyses with the chloroplast rbcL-a gene and the trnL-F intergenic spacer support this topology. Our results also confirmed that the four Noccaea spp. are not part of a monophyletic Aethionema (Khosravi et al., 2009;Al-Shehbaz, 2012). Five annual species form a monophyletic group within clade A, suggesting a single origin of the annual life history. The five annual species, as well as the perennial Aethionema saxatile, are heteromorphic. However, Ae. saxatile is not monophyletic with the annual species, suggesting an independent origin of its heteromorphism.

Ae. arabicum Produces Two Clearly Distinct Fruit Morphs
To test whether fruits of Ae. arabicum cluster into two distinct fruit morphs (Solms-Laubach, 1901;Hedge, 1965), we performed a morphometric analysis that incorporated data for fruit width and fruit length, the presence or absence of a septum, the number of seeds per fruit, and the production of seed mucilage. Plotting the width and length of ripe fruits against each other confirmed two clusters corresponding to the distinct fruit types (Fig. 2). This is further supported by hierarchical and two-step cluster analyses based on fruit width, fruit length, and the number of seeds per fruit (Supplemental Fig. S3). The clustering correlates to 100% with the presence of a septum and the production of seed mucilage in the large, two-to sixseed-containing fruit morph, as opposed to the absence of a septum and nonmucilaginous seeds in the small, singleseeded fruit morph. The two clearly distinct fruit morphs characterize Ae. arabicum as a dimorphic species.

Dehiscence, Abscission, and Movement of Dimorphic Fruits
Fruits of heteromorphic plant species have often been reported to differ in their dispersal mechanisms (Sorensen, 1986;Mandak and Pysek, 2001;Lu et al., 2010Lu et al., , 2013Dubois and Cheptou, 2012). To investigate whether this also holds true for the two Ae. arabicum fruit morphs, we studied their dehiscence and detachment behavior (Fig. 3). A random-impact test performed on ripe fruits revealed significant differences in dehiscence half-life between the two morphs (Fig. 3A). The many-seeded larger morph dehisced readily upon mechanical stimulation (mean halflife of 15 s) and is henceforth called the dehiscent fruit morph. The single-seeded smaller morph took approximately 10 times longer to open (mean half-life of 158 s) and is henceforth called the indehiscent fruit morph. In contrast, a fruit detachment force test (Fig. 3B) revealed that a slight touch (140 mN on average) was already sufficient for the abscission of ripe indehiscent fruits from the mother plant, while a significantly higher force (885 mN on average) was required to detach ripe dehiscent fruits. The moisture-induced movement of fruits or fruit parts often helps desert plant species to relate seed dispersal to rain events, so as to ensure optimal germination conditions (Gutterman, 1993). Therefore, we exposed dry Ae. arabicum infructescences to different degrees of humidity and measured changes in the angle between the pedicel and infructescence axis (Fig. 3C). The exposure Figure 1. Phylogenetic tree for the tribe Aethionemeae. Phylogenetic analysis of 38 Aethionema species identifies three wellsupported clades (A, B, and C). The corresponding RaxML and MrBayes analyses had the same topology except for the species marked with stars (for exact differences, see Supplemental Figs. S1 and S2). The topology suggests a single origin of the annual lifehistory strategy (gray bar) but indicates two independent origins of heteromorphism (highlighted in purple). Green brackets, posterior P . 0.8 and bootstrap percentage greater than 80; orange brackets, posterior P , 0.5 and bootstrap percentage less than 80; blue brackets, posterior P . 0.8 and bootstrap percentage less than 80.
to moisture resulted in a noticeable outward bending of pedicels, starting from an angle of approximately 30°in dry conditions (Fig. 3D) up to 90°after 30 min of exposure to water spraying ( Fig. 3E; Supplemental Video S1). This spatiotemporal movement pattern was similar for both fruit morphs (Fig. 3C), suggesting that rain may aid the dispersal of both morphs. Taken together, this shows that Ae. arabicum applies two alternative strategies, fruit dehiscence versus abscission, to disperse its seeds.

Comparative Morphology of the Dimorphic Fruits
In our search for anatomical features underlying the observed mechanical differences between the two fruit morphs, we compared their dehiscence and abscission zones (Fig. 4). Dehiscent fruits contain two to six seeds and a septum dividing their inside into two locules (Fig.  4, A and B), whereas indehiscent fruits contain a single seed tightly enclosed by the unilocular fruit without a septum (Fig. 4, A and C). Scanning electron microscopy (SEM) images of ripe fruits revealed the beginning of tissue separation, indicative for the presence of a dehiscence zone at the valve-replum transition in the dehiscent morph ( Fig. 4, D, E, and G), while valve and replum remain tightly connected in the indehiscent morph ( Fig. 4, F, H, and I). Cross sections of the valve-replum border of green silicles just prior to the onset of ripening-induced yellowing revealed the presence of two stripes of nonlignified cells separating the lignified cells of the replum and endocarp layer b (enb) in the dehiscent morph ( Fig.  4J), thus resembling the dehiscence zones of other Brassicaceae species with dehiscent fruits, including Arabidopsis (Arabidopsis thaliana; Ferrándiz et al., 2000;Østergaard et al., 2006;Arnaud et al., 2011;Mühlhausen et al., 2013). In contrast, in the indehiscent morph, the lignified cells of the enb and the replum are not separated, thus forming a continuous lignified band at the inside of the fruit valve completely enclosing the seed (Fig. 4K). Therefore, a dehiscence zone (dz and white arrow in Fig. 4, G and J) is present only in the dehiscent morph.
Programmed organ abscission often is mediated by the formation of an abscission zone, characterized by several layers of small, densely cytoplasmic cells at the base of the respective organ (Li et al., 2006;Estornell et al., 2013). An abscission zone is evident at the fruitpedicel junction of the indehiscent fruit morph, separating the lignified cells at the fruit base from those of the pedicel (az in Fig. 4M). In contrast, the dehiscent fruit morph is tightly connected with the pedicel through a continuous bridge of lignified cells (Fig. 4L). In order to characterize the molecular pathway underlying fruit morph development in Ae. arabicum, we performed homology searches to identify Ae. arabicum orthologs of the best characterized fruit regulatory genes in Arabidopsis, namely ALCATRAZ (ALC), APETALA2 (AP2), FILAMENTOUS FLOWER (FIL), FRUITFULL (FUL), INDEHISCENT (IND), REPLUMLESS (RPL), SHATTERPROOF1 (SHP1), and SHP2 (Dinneny et al., 2005;Ripoll et al., 2011). These also include genes that are known to show changes in expression pattern in other indehiscent Brassicaceae fruits (Avino et al., 2012;Mühlhausen et al., 2013). Phylogenetic analyses showed that, for each of these Arabidopsis fruit regulators, a single ortholog is present in the Ae. arabicum genome, henceforth called AearALC, AearAP2, AearFIL, AearFUL, AearIND, AearRPL, AearSHP1, and AearSHP2, respectively (Table I; Supplemental Fig. S4). Gene expression analyses via quantitative reverse transcription-PCR on outgrown, green fruits revealed that all genes were expressed substantially in the dehiscent morph. Comparing expression levels between the two fruit morphs showed that, of the investigated genes, only AearIND expression was significantly different and approximately 7-fold lower in indehiscent compared with dehiscent fruits (Fig. 5). This finding indicates that differential regulation of AearIND might be one of the key molecular mechanisms for the establishment of morph-specific differences during Ae. arabicum fruit development. Ae. arabicum fruits. Small fruits contain a single nonmucilaginous seed and no septum (dark gray circles), while large fruits contain two to six mucilaginous seeds and a septum (light gray circles).

Comparative Morphology of the Dimorphic Seeds
In addition to the strong dimorphism of Ae. arabicum fruits, the seeds developing within indehiscent or dehiscent fruits also exhibit remarkable morphophysiological differences (Fig. 6). The most obvious difference is the production of mucilage upon imbibition in mature seeds from dehiscent fruits (Fig. 6, C-H), henceforth referred to as the mucilaginous seed morph (M + ). This mucilage is mostly lacking in imbibed mature seeds from indehiscent fruits (Fig. 6, A and B), henceforth referred to as the nonmucilaginous seed morph (M 2 ).
The production of seed mucilage, known as myxospermy, occurs from the outer cell walls of the seed coat epidermal cells in response to seed hydration, forming a water-containing, gel-like pectinaceous layer surrounding the seed (Western, 2012). A number of species have cellulosic threads or thick fibers projecting from their mucilage cells. Our detailed analysis of the Ae. arabicum seed surface by light microscopy and SEM (Fig. 6) showed a smooth to slightly grooved surface structure of M 2 seeds (Fig. 6, A and B). In contrast, the surface of M + seeds was densely covered with dome-like structures and crinkles around their base, each corresponding to a mucilage-producing epidermal cell (Fig. 6F). Upon seed imbibition, these structures were irreversibly swelling, expanding, and forming conical mucilage papillae of up to 200 mm with a globe-like tip (Fig. 6, D-H). Upon drying, the papillae shrank in diameter and formed knob-shaped tips (Fig. 6,G and H). Each of the two seed morphs was unambiguously connected with the fruit morphs: M 2 seeds were found inside indehiscent fruits, while M + seeds were dispersed from dehiscent fruits.
Comparative morphological analysis showed that the surface differences were accompanied by different positions of the radicle in relation to the cotyledons and the internal cell layers (Fig. 7). The M + seeds were oblong and biconvex, having a notorhizal embryo (incumbent: radicle lying along the back of one cotyledon). M 2 seeds were ovate and planoconvex, and the embryo was either pseudonotorhizal (radicle situated near the margin of the cotyledons) or nearly pleurorhizal (accumbent: radicle applied to margins of both cotyledons; Fig. 7B). Light microscopic analysis highlighted the difference in the abundance of mucilage as well as embryo position between the two seed morphs (Fig. 7). The mature seed coat of both morphs was composed of multiple layers. The outermost epidermal layer formed large Astra Blue-stainable mucilage papillae in M + seeds (Fig. 7D), whereas M 2 seeds produced only a very thin film of mucilage from their epidermal layer (Fig. 7I). Directly adjacent to these epidermal layers, Figure 3. Comparison of fruit dehiscence, abscission, and movement between the two fruit morphs. A to C, Small and large fruit morphs of Ae. arabicum differ in dehiscence halflife (A) and detachment force (B) but have similar movement patterns in response to humidity (C). Values shown are means 6 SD for n = 19 (A), n = 47 (B; indehiscent), n = 44 (B; dehiscent), and n = 8 (C) replicate measurements. D and E, Fruit movement is further illustrated by exemplary images of a ripe dehiscent fruit under dry conditions (D; relative humidity less than 50%) and humid conditions (E; water spraying).
both seed morphs had an unstained single layer of small palisade cells (Fig. 7, D and I). Inward, this was followed by multiple layers of safranin-stainable crushed palisade cells (Fig. 7, F and K). All these outer tissue layers of the seed coat consisted of dead cells, as indicated by the absence of nuclei ( Fig. 8, A-F). Between the embryo and the seed coat, we identified a layer of living cells that completely surrounded the embryo in both seed morphs (Figs. 7, E and L, and 8, A-F). This layer appeared thin around most of the embryo but thicker around the radicle tip, where it could be multilayered (Fig. 7, G and M). In many Brassicaceae species, a living endosperm layer around the embryo plays an important role in the regulation of dormancy and germination of the seeds (Müller et al., 2006;Graeber et al., 2012). To determine if the identified cell layer in Ae. arabicum was part of the endosperm, we performed flow cytometric analysis of these layers in both seed morphs. C values of 2.97 6 0.05 (M + ) and 2.97 6 0.02 (M 2 ) confirmed triploidy and, therefore, endosperm origin (Fig. 8G). In conclusion, M + seeds from dehiscent fruits and M 2 seeds from indehiscent fruits differ in several anatomic features, affecting the position of the embryo and the seeds' outermost mucilage-producing epidermal layer.

Germination Physiology of Dimorphic Ae. arabicum
The dimorphic syndrome of Ae. arabicum has consequences for seed germination. In the case of the indehiscent fruits, in which the whole fruit represents the natural dispersal unit, the M 2 seeds need to germinate within the fruit (unless they are released from the fruit coat by external mechanical means). Interestingly, we found that 100% of mature M + seeds germinated readily within 3 d under laboratory conditions, whereas mature M 2 seeds within indehiscent fruits reached only 50% germination after about 3 weeks under the same conditions ( Fig. 9A). To test if these conditions are generally prohibitive for M 2 seed germination, we analyzed seeds manually dissected from indehiscent fruits. These isolated M 2 seeds were able to germinate faster and to a greater extent (70%) than those within indehiscent fruits, although the overall germination speed and maximum capacity remained lower compared with M + seeds (Fig. 9A).
Water uptake is vital for germination and can be controlled by various fruit or seed structures and by mucilage (Weitbrecht et al., 2011;Western, 2012). The distinct coat structures of the two seed morphs and their different germination kinetics prompted us to investigate the water-uptake patterns of mature M + seeds, isolated mature M 2 seeds, and intact mature indehiscent fruits during germination (Fig. 9B). M + seeds showed a classical water-uptake pattern: imbibition was characterized by a very rapid and steep increase in seed moisture content (phase I), followed by a plateau (phase II), which leads to another increase (phase III), coinciding with embryo elongation and radicle protrusion to complete germination. A three-phase water uptake also was evident for isolated M 2 seeds. However, compared with M + seeds, their overall moisture content during imbibition remained much lower, and phase II was prolonged, in agreement with the lack of mucilage and the later completion of M 2 seed germination ( Fig. 9). Interestingly, indehiscent fruits and M + seeds took up similar relative amounts of water during phase I, indicating that indehiscent fruit coats aid in water absorption, seemingly compensating for the lack of mucilage of M 2 seeds. Although the indehiscent fruits initially imbibed water similarly to M + seeds, they then remained in phase II, reflecting the delayed germination of the M 2 seeds within the fruits. To investigate if water taken up by the fruit coat was subsequently also taken up by the enclosed seed, we measured the water activity (a w ) during phase II of M 2 seeds imbibed within or without fruit coats. Values for a w related to the amount of freely available water, ranging from 0 (completely dry matter) to 1 (pure water). Dry M 2 seeds or indehiscent fruits (a w approximately 0.4) were imbibed for 18 h. The seeds subsequently dissected from fruit coats showed similar a w compared with seeds imbibed without fruit coat (a w approximately 0.8). Thus, the indehiscent fruit coat delayed germination, although the fruit coat was fully permeable for water.

Plasticity of Ae. arabicum Fruit Morph Production
Ae. arabicum plants originating from M + or M 2 seeds, respectively, did not differ significantly in their fruitmorph ratio or total number of fruits, indicating that plants developing from the two different seed morphs are indistinguishable upon maturity (Fig. 10A). Although the succession of dehiscent and indehiscent fruits along an individual infructescence can appear stochastic (Fig. 2B), the two fruit morphs were not distributed randomly on the whole plant: higher order side branches produced a larger fraction of indehiscent fruits than the main branch (Fig. 10B). Other heteromorphic Aethionema spp. have been reported to develop more dehiscent fruits after cutting off side branches (Zohary and Fahn, 1950). This prompted us to investigate the potential phenotypic plasticity in Ae. arabicum fruit morph production. The constant removal of all side branches significantly increased the production of dehiscent fruits on the main branch to approximately 95% (Fig. 10C). We further observed that sets of plants originating from the same seed batch but grown either in a phytochamber or a greenhouse had different fruit-morph ratios in their progeny. While indehiscent fruits clearly predominated (approximately 70%) in the phytochamber, the majority of fruits (approximately 80%) produced in the greenhouse belonged to the dehiscent morph (Fig. 10D). One of the factors that varied significantly between the nonconditioned greenhouse and the phytochamber was temperature. Thus, we grew plants under two different constant temperatures but otherwise identical conditions and again observed a strong influence on fruit morph production. At a growth temperature of 20°C, the fraction of indehiscent fruits was much higher compared with 25°C (Fig. 10E). This plastic response was not solely an indirect effect of a changed branching pattern, because the shift in fruit morph production could be detected throughout the whole plant (Supplemental Fig. S5). These results indicate that Ae. arabicum shows a plastic response of its fruit-morph ratio in response to certain environmental factors.   and survival of several plant species in environmentally unpredictable habitats (Imbert, 2002;Evans and Dennehy, 2005;Lu et al., 2010). Heteromorphic seeds and fruits provide distinct dispersal and germination strategies to aid the distribution of the (mostly annual) species in time and space. We found that Ae. arabicum, an annual Brassicaceae plant adapted to arid and semiarid environments, employs a dimorphic dispersal and germination strategy (Fig. 11). On the same infructescence, it produces dehiscent fruits with M + seeds and indehiscent fruits harboring M 2 seeds. Figure 11 depicts the Ae. arabicum dimorphic dispersal and germination strategy Figure 6. Morphological characterization of dimorphic Ae. arabicum seeds. Characteristic images show the two Ae. arabicum seed morphs taken with a binocular microscope (A, C, D, E, and H3) or a scanning electron microscope (B, F, G, H1, H2, and H4). Mature seeds from indehiscent fruits (M 2 seeds; A and B) have a smooth surface and do not produce mucilage upon imbibition. Mature seeds from dehiscent fruits (M + seeds; C-G) are densely covered with dome-like structures and crinkles. Upon imbibition, conical mucilage papillae with a globe-like tip emerge from these structures (D and H). When redried, papillae shrink in diameter and length and finally form dried, knob-shaped tips (E, G, and H). White arrowheads, sticky, ball-shaped tips; black arrowheads, dried, knob-shaped tips.
as part of the annual life cycle. Two clearly distinct lifehistory strategies are evident: quickly germinating M + seeds dispersing via fruit dehiscence and slowly germinating M 2 seeds separating from the mother plant via the abscission of indehiscent fruits (Fig. 11). The formation of morphs with low dispersal ability and delayed seed germination has been interpreted as a low-risk strategy because they stay in the approved habitat near the mother plant, and their fractionated germination increases the chance for at least some of them to encounter favorable environmental conditions (Venable and Lawlor, 1980;Venable and Levin, 1985;Venable et al., 1995;Lu et al., 2012Lu et al., , 2013Lu et al., , 2015. Morphs with high dispersal ability and quick germination, on the other hand, may represent a high-risk strategy that only pays off when the environmental conditions are beneficial. However, the exact opposite combination of the two traits (quickly germinating nondispersing morphs) also has been observed and interpreted as a way to overcome high sibling competition in local populations (Dubois and Cheptou, 2012;Rubio de Casas et al., 2015). The moisture-dependent pedicel movement (Fig. 3, C-E) may aid rain-operated seed dispersal (Gutterman, 1993;Parolin, 2006;Pufal and Garnock-Jones, 2010). Independent of their dispersal behavior (Fig. 11), we interpret the formation of M + seeds as a high-risk strategy, because quick and uniform germination in uncertain environmental conditions is always risky, especially for annuals. Thus, the delayed and fractionated germination of M 2 seeds would represent a low-risk strategy. However, more research investigating the connection between seed germination, dispersal, and survival under natural growth conditions is needed in order to refine our understanding of the dimorphic life-history strategy of Ae. arabicum. Various forms of seed and fruit heteromorphism have evolved independently, predominantly in the Asteraceae, Amaranthaceae, and Brassicaceae families (Imbert, 2002). Within the genus Crepis (Asteraceae), it has evolved independently several times (Imbert, 2002;Dubois and Cheptou, 2012). Within the Brassicaceae, heteroarthrocarpy has evolved in annual Cakile spp. . Light microscopic analysis of longitudinal (A and H) and transversal (C and J) whole seed sections highlight differences in the formation of blue-stained mucilage. The seeds' outermost epidermal layer (ep), which is directly adjacent to a single layer of palisade cells (p), forms large mucilage papillae (m) in the case of M + seeds (D) but only a very thin film of mucilage in M 2 seeds (I). Both seed morphs (F and K) possess multiple redstained crushed palisade cell layers (cp) in direct contact with the endosperm layer (en). The whole embryo (em) is surrounded by this continuous single-layered endosperm (black arrowheads; E and L), which appears multilayered and thick around the radicle tip (G and M). Cross sections were stained with safranin/Astra Blue. Astra Blue stains unlignified cell walls, cellulose, and mucopolysaccharides blue, and safranin stains lignified, suberized, or cutinized cell walls red. Bars = 100 mm. and other annual Brassiceae plants (Hall et al., 2006;Avino et al., 2012), as has a complex fruit and seed heteromorphism in the annual Chorisporeae plant Diptychocarpus strictus (Lu et al., 2010(Lu et al., , 2012(Lu et al., , 2015. We demonstrate here that heteromorphism has evolved twice within the genus Aethionema (38 species analyzed): in the perennial Ae. saxatile (Andersson et al., 1983) and independently for the five annual species including Ae. arabicum ( Fig. 1; Solms -Laubach, 1901;Zohary and Fahn, 1950;Hedge, 1965). In contrast to the Bayesian inference, a maximum likelihood analysis does not support an annual clade (Supplemental Fig. S2). Further phylogenetic research with higher species coverage is needed to address this issue better. We propose that the dimorphism of Ae. arabicum has evolved as an adaptation to unpredictable environments such as the arid and semiarid habitats of the Irano-Turanian region (southwest Asia; e.g. Turkey, Syria, Iran, and Iraq) to which this monophyletic group of annual plants is adapted.

Anatomy and Molecular Regulation of Fruit Dehiscence in Ae. arabicum
Brassicaceae fruits are typically dehiscent pods. Their opening mechanism depends mainly on the correct  . Ae. arabicum heteromorphic dispersal units differ in their germination and water uptake patterns. Mature M + seeds, M 2 seeds extracted from mature indehiscent fruits, and whole intact indehiscent fruits were incubated at 14˚C in continuous light. Germination over time (A) is shown in relation to water uptake kinetics (B) expressed as the percentage moisture content of fresh weight (FW). Three phases of water uptake (I-III) are indicated for M + and related to their germination kinetics. Note that three phases of water uptake also can be identified for M 2 . Phase III generally coincides with the completion of germination by radicle emergence, and, reflecting their delayed germination, seeds in indehiscent fruits remain in the plateau phase II during the investigated interval.
formation of a dehiscence zone at the valve-replum border Roberts, 1990a, 1990b;Spence et al., 1996). This zone typically consists of two directly adjacent and functionally complementary layers of cells: a separation layer and a lignified layer. Studies so far pointed toward a high conservation of the fruit-opening process within the Brassicaceae, and the presence of separation layer cells in the dehiscent fruits of Ae. arabicum indicates that fruit opening in this species also may function in a similar way ( Fig. 4; Hall et al., 2006;Østergaard et al., 2006;Mühlhausen et al., 2010Mühlhausen et al., , 2013Arnaud et al., 2011). The facts that single orthologs of all investigated fruit developmental genes are present in the genome of Ae. arabicum (Table I) and that they are all expressed in the dehiscent fruit morph further support the idea that not only the Brassicaceae-specific opening mechanism, but also its molecular regulation, might be at least partially conserved in the dehiscent Ae. arabicum fruits.
A special feature of Ae. arabicum dehiscent fruits is the lack of a typical lignified layer present in other Brassicaceae plants, where it forms a lignified bridge connecting the enb with the fruit exocarp (Hall et al., 2006;Østergaard et al., 2006;Mühlhausen et al., 2013). In Ae. arabicum, the separation layer is located between the lignified cells of the enb and the replum, a feature that may by characteristic for the Aethionemeae in general, as it is also observed in dehiscent fruits of Ae. saxatile (Mühlhausen et al., 2010). Another peculiar feature of the dehiscent fruits of Ae. arabicum is the presence of small cellulose-rich cells in the replum that are connected to the separation layer cells (Fig. 4J). They are anatomically similar to separation layer cells and only present in the dehiscent morph, implying that they may act as a functional extension of the separation layer. This could explain how fruit opening is achieved, although the separation layer itself does not extend toward the fruit exocarp in Ae. arabicum.
The anatomy of the valve-replum border of Ae. arabicum indehiscent fruits resembles that of indehiscent fruits in other Brassicaceae species (Hall et al., 2006;Mühlhausen et al., 2010Mühlhausen et al., , 2013. The dehiscence zone is absent and the lignified cells of the enb are connected directly to the lignified part of the replum, thus forming a continuous lignified band around the fruit and preventing fruit opening. In other species, these anatomical changes have been connected with altered expression patterns of genes orthologous to the fruit developmental genes SHP1, SHP2, IND, and ALC, which are known to control the formation of the dehiscence zone in Arabidopsis (Liljegren et al., , 2004Rajani and Sundaresan, 2001;Avino et al., 2012;Mühlhausen et al., 2013). Likewise, our expression analysis revealed a strong down-regulation of AearIND in indehiscent compared with dehiscent fruits in Ae. arabicum (Fig. 5). This downregulation alone could be sufficient to induce the anatomical changes underlying the indehiscence phenotype in the respective fruit morph (Fig. 4K), provided that gene functions and regulatory interactions are indeed similar to those in Arabidopsis. There, IND is crucial for the formation of the dehiscence zone by directly controlling the lignification of the lignified margin cells and by indirectly controlling the formation of the separation layer by mediating the release of ALC and SPATULA from DELLA repressor proteins (Liljegren Figure 10. Fruit morph production of Ae. arabicum shows phenotypic plasticity in response to environmental triggers. The ratios (top rows) and total numbers (bottom rows) of fruits belonging to the dehiscent (gray bars) or indehiscent (black bars) fruit morph are similar for plants originating from M + or M 2 seeds (A) but change significantly between main and side branches (B) in response to cutting off all side branches (causing a change in plant architecture; C), upon growth in a controlled phytochamber versus a non-temperature-controlled greenhouse (D), and at different defined growth temperatures (E). Data represent means with SD. et al., 2004;Arnaud et al., 2010;Groszmann et al., 2011). Thus, the differential expression of AearIND likely represents the first molecular key mechanism that has been identified to cause morph-specific differences during heteromorphic fruit development. It will be interesting to investigate which upstream regulatory mechanisms cause this differential expression and whether it is possible to identify a regulatory connection with other morphspecific differences such as fruit abscission or seed mucilage production.

Morphology and Ecophysiology of Dimorphic Seed Germination Differing in Myxospermy and Fruit Coat Constraint
Our comparative data about the seed morphology of Ae. arabicum (Figs. 6-8) allow us to address questions about the differential control of M + and M 2 seed development. The mature seed coat is of maternal origin, and its development has been characterized genetically in detail in Arabidopsis (Haughn and Chaudhury, 2005). The complex genetic regulation of mucilage secretory cell development from the outer integument in Arabidopsis seeds has been uncovered (Francoz et al., 2015). We show here that M + seeds develop mucilage papillae upon wetting, whereas M 2 seeds show reduced mucilage production because they lack mucilage secretory cells. In Arabidopsis, defects in the differentiation of the outer integument during seed development have been correlated with a lack of mucilage synthesis (Western, 2012). Several transcription factor mutants affecting outer seed coat differentiation, such as ap2 (Western et al., 2001) and the nac-regulated seed morphology1 (nars1) nars2 double mutant (Kunieda et al., 2008), do not produce any mucilage and show altered seed coat surface structure. Especially the shriveled appearance of the Arabidopsis nars1 nars2 seed coat looks strikingly similar to the Ae. arabicum M 2 seed surface. It is tempting, therefore, to speculate that Ae. arabicum is able to regulate its seed development differentially to produce seeds with or without mucilage by differentially employing conserved developmental regulators.
The adaptive value of seed mucilage also has prompted plant ecologists to propose a role in the long-distance dispersal as well as the local anchorage of seeds (Yang et al., 2012). The myxospermy of Ae. arabicum M + seeds (Figs. 6 and 11) could assist long-distance dispersal by adherence to animal vectors (Norton et al., 1997;Mummenhoff and Franzke, 2007). However, it also may allow M + seeds to adhere to soil particles, a common mechanism for seed retention in dry habitats (Huang et al., 2000;Lu et al., 2010;Gutterman, 2012). Beyond dispersal, seed mucilage may promote seed germination through the attraction and retention of water surrounding the seed (Yang et al., 2012), protect against osmotic stress (Yang et al., 2010), assist the repair of embryo DNA damage (Yang et al., 2011), and promote early seedling growth as an adaptation to harsh desert environments (Yang et al., 2012). Mucilage Figure 11. Scenario of different life-history strategies of Ae. arabicum seed morphs. After fruit maturation, indehiscent fruits with enclosed M 2 seeds abscise from the mother plant, while dehiscent fruits open to release M + seeds. These germinate quickly, while M 2 seeds remain in the soil seed bank for longer periods, mainly due to pericarp-mediated dormancy. Furthermore, both morphs may be subject to long-distance dispersal due to the presence of adhesive mucilage (M + ) or enclosure in winged indehiscent fruits (harboring M 2 seeds).
produced by the scattered Ae. arabicum M + seeds was associated with faster germination compared with the nonmucilaginous M 2 seeds encased by the indehiscent fruit coat (Fig. 9). A similar behavior has been described for other heteromorphic Brassicaceae plants (Zohary, 1962;Imbert, 2002;Lu et al., 2015). Interestingly, M 2 seeds removed from their surrounding fruit coat can germinate quickly, similar to M + seeds. Waterimpermeable cell layers in fruit coats can prevent imbibition and germination, which is referred to as physical dormancy (Finch-Savage and Leubner-Metzger, 2006). We show here that Ae. arabicum indehiscent fruit coats are water permeable. Therefore, the delay in germination is likely caused by a purely mechanical restraint of the fruit coat or by chemical inhibitors. The differential presence of chemical germination inhibitors in heteromorphic seeds and fruits has been described in many species (Matilla et al., 2005). Notably, the presence of larger amounts of the germination-inhibiting plant hormone abscisic acid in the fruit coats of one morph of the heteromorphic species Salsola komarovii caused delayed seed germination that could be overcome by removing the fruit coat (Takeno and Yamaguchi, 1991). Furthermore, mechanical restraint was proposed as the cause of the delayed germination of indehiscent siliques of heteromorphic D. strictus (Lu et al., 2015). Further biomechanical and biochemical studies of the indehiscent fruit coat of Ae. arabicum will shed light on its germinationinhibiting nature.

Phenotypic Plasticity of Fruit and Seed Heteromorphism
Phenotypic plasticity and bet hedging are two evolutionary modes of response to environmental variance (Simons, 2011;Abley et al., 2016). Plasticity describes a concerted change of a given trait over a range of environmental conditions and, thus, critically depends on the availability of cues that allow forecasting of the future state of the environment (Bradford and Roff, 1993;Simons, 2011). Bet hedging, on the other hand, describes a risk-spreading strategy as an adaptation to environmental unpredictability, producing only one fixed phenotypic condition that is suboptimally adapted to any given environment but maximizing the geometric mean fitness across generations (Philippi and Seger, 1989). Heteromorphism in plants has often been considered a mere bet-hedging strategy, emphasizing its independence from environmental conditions (Imbert, 2002). However, our data, in accordance with several other studies, demonstrate that, in some heteromorphic plant species, morph numbers and ratio show plasticity in response to certain environmental stimuli (Mandák and Pyšek, 1999;Imbert and Ronce, 2001;Sadeh et al., 2009;Lu et al., 2013b;Yang et al., 2015). Such a blend of bet hedging and plasticity should be expected to evolve when either the cue that predicts the future environment is weak or fitness is determined by predictable and unpredictable environmental factors alike (Bradford and Roff, 1993;Simons, 2011). Comparative analyses are a powerful approach for unraveling the evolutionary and genetic backgrounds of phenotypic plasticity in heteromorphic fruit and seed development.
The Potential of Ae. arabicum for Future Research on Heteromorphism and Plasticity A fascinating and underexplored aspect of heteromorphism is its genetic and molecular control. Since the differences between the morphs usually are multifaceted, the respective regulatory module(s) must be positioned upstream of several developmental pathways (such as the fruit dehiscence pathway discussed above) and regulate their action in a highly coordinated manner. Moreover, sensory elements are required when morph development is regulated in an environmentally dependent manner. Nevertheless, heteromorphism evolved many times independently (Fernández et al., 2001;Imbert, 2002;Cruz-Mazo et al., 2009), suggesting that its genetic basis is rather simple. A deep understanding of heteromorphism at the phenotypic level is crucial for subsequent molecular studies. Therefore, our data are important prerequisites to understand the developmental and molecular aspects of heteromorphism. Even more challenging will be to unravel the importance of heteromorphism for plant fitness under natural growth conditions. The life-history plasticity revealed by Ae. arabicum raises the question of whether it is adaptive. Unfortunately, to assess the adaptive value of plastic responses is not a trivial task (Sultan, 2000). Experimental approaches for that are available, but conclusive investigation of the complex framework of ecological evolutionary developmental biology depends very much on suitable model systems (Sultan, 2000;Gilbert et al., 2015).
Ae. arabicum is a good candidate to become such a model species because it is reliably dimorphic without any intermediate morphs and with striking differences in various anatomical and physiological features and evidence for developmental control and changes in gene expression patterns. In addition, it is easy to grow and has an advantageously short life cycle. With all these features, including a published genome sequence (Haudry et al., 2013), Ae. arabicum represents the current best organism in which to investigate and understand the molecular, evolutionary, and ecological aspects of heteromorphism.

Phylogenetic Inference
We sampled 38 Aethionema spp., four Noccaea spp. formerly included in Aethionema (Al-Shehbaz, 2012), and Tarenaya hassleriana as an outgroup. The geographic origins are listed in Supplemental Table S1. DNA was extracted from silica-dried and ground leaf material using the cetyl-trimethyl-ammonium bromide method (Bakker et al., 2016). DNA quality was controlled by agarose gel electrophoresis and Nanodrop 1000 spectrophotometry (Thermo Fisher Scientific). The chloroplast rbcL-a gene (forward primers from Huang and Shi [2002] and reverse primers from Fofana et al. [1997]) and the trnL-F intergenic spacer (primers from Dumolin-Lapegue et al. [1997]) were used as phylogenetic markers. Cycle sequencing was performed at Greenomics. Codoncode aligner was used to clean up and align the retrieved sequences. The genes were concatenated to a total alignment of 1,499 bp. For a Bayesian phylogenetic analysis (MrBayes version 3.2.2; nst = mixed, rates = gamma, ngen = 250,000,000, diagnfreq = 5,000, and temp = 0.05), we partitioned the data as follows: trnL-F, rbcL-a codon positions one and two, rbcL-a codon position three. We also ran a maximum likelihood analysis (RaxML version 8; 1,000 bootstraps and GAMMA model of rate heterogeneity) on an unpartitioned data set. The final alignment is available as a nexus file (Supplemental Data S1). Figures were made with FigTree version 1.4.2 and GIMP 2.8.10.

Plant Material and Growth Conditions
Experiments were conducted on Aethionema arabicum plants or seeds of accession 0000309 (obtained from Kew's Millennium Seed Bank) or accession ES1020 (obtained from Eric Schranz). Plants were grown on soil under long-day conditions (16 h of light/20°C and 8 h of dark/18°C) in a greenhouse or phytochamber (CambridgeHOK) or in a non-temperature-controlled greenhouse in summer.

Morphometric Analysis
Twenty ripe fruits were harvested from the most basal part of the main inflorescence of seven individual Ae. arabicum plants. Fruit length and width were determined using a Leica M205 FA stereomicroscope employing the Interactive Measurement module of the Leica Application Suite software. Subsequently, fruits were opened to assess the presence or absence of a septum and the number of seeds. For one seed per fruit, mucilage production was evaluated 5 min after incubating the seed in a drop of water. In order to identify clustering within our data set, hierarchic cluster analysis followed by two-step cluster analysis was performed using the SPSS 20.0 software package.

Quantification of Abscission, Dehiscence, and Hygrochastic Movement
To quantify fruit dehiscence, a random-impact test was performed on ripe fruits derived from 19 replicate plants. Principally, the test was performed as described previously (Lenser and Theißen, 2014) but using three 5-mm steel balls and an agitation force of 9 Hz.
To quantify fruit detachment force, ripe whole main-branch infructescences were fixed with a metal clamp on a rack with the tip of the infructescence pointing downward. Fruits in direct proximity to the fruit to be analyzed were removed. A hair was folded around the fruit-pedicel junction. Weights were attached to the hair so that a force specified by the attached weight was applied, pulling the fruit away from the infructescence. The force needed to detach the fruit from the stem was recorded.
Analyses to quantify the hygrochastic movement of fruit pedicels were carried out with ripe infructescences in an air-tight glass vessel. All fruits were removed from infructescences, and an equilibrium relative humidity of 53% and 89% at 23°C to 25°C was adjusted by saturated salt solutions as described by Greenspan (1977) and controlled by an electronic thermometer/hygrometer (PCE-313 A; Meschede). Furthermore, individual pedicels were sprayed with 300 mL of water every 10 min. Hygrochastic pedicel movement was documented photographically (Nikon D7100, Sigma 105-mm F2.8 EX DG MACRO OS; one photograph every 1-3 min) and quantified by an angle meter in Adobe Photoshop.

Microscopic Analysis of Fruits and Seeds
Fruits just prior to the onset of ripening-induced yellowing were fixed in 2% formaldehyde, 5% glacial acetic acid, 60% ethanol, and 0.1% Tween 20 at 4°C for 24 h, embedded in Paraplast (Carl Roth), and sectioned. Thin sections were dewaxed and stained for 2 min with safranin/Astra Blue (Sigma-Aldrich; Gerlach, 1984), followed by microscopic analysis using a Leica DM5500 B microscope.
Dry mature seeds were fixed as described (Lee et al., 2012) and embedded in Technovit 7100 (Heraeus Kulzer) according to the manufacturer's instructions. Prior to fixation, seeds were pierced three times with insect pins. Polymerization took place in truncated pyramid-shaped 8-mm-diameter polythene embedding capsules (BEEM). Cuts of 6 to 12 mm thickness were made with a Microm HM355S microtome (Thermo Scientific), using the specimen clamp, knife block N, knife holder C, and Histoblades (Heraeus Kulzer). Embedded samples were placed directly into the specimen clamp without the use of Histoblock. Dry cuts were stained for 5 min in a freshly made mixture (5:1) of Astra Blue (0.5% in 0.5% acetic acid) and safranin (1% in water). Samples were washed once with deionized water and subsequently differentiated with 1% HCl in 96% ethanol. For nuclei analysis, samples were mounted in 20 mL of Vectashield (Vector Laboratories) + 2 mg mL 21 DAPI. Microscopic analysis was done using an NiE Upright Microscope (Nikon) and the NIS-Elements Basic Research software.
For SEM analysis, specimens were dried over silica gel for 2 weeks, mounted on specimen stubs using a carbon adhesive disc (Plano), and coated with platinum-iridium with a sputter coater (K575X Turbo; Quorum Technologies). Surfaces were analyzed by SEM (Supra 55VP; Carl Zeiss).

Ortholog Identification
To identify orthologs of Arabidopsis (Arabidopsis thaliana) fruit developmental genes in Ae. arabicum, Arabidopsis query sequences were searched with BLASTP (Altschul et al., 1990) against a plant-specific, custom-made protein database that included genomes of the species listed in Supplemental Table S2. Results were filtered for having at least 80% query coverage and according to Rost (1999) to detect clearly homologous sequences only. Resulting sequences were aligned using MAFFT version 7.037b (Katoh and Standley, 2013) in automatic mode, and alignments were inspected manually and trimmed using Jalview version 2.8 (Clamp et al., 2004). Duplicated sequences were removed after inspection of initial trees. Final neighbor-joining phylogenies were constructed using Quicktree-SD (Howe et al., 2002;Frickenhaus and Beszteri, 2008) with 1,000 bootstrap samples and displayed and midpoint rooted with FigTree version 1.4.2 (http://tree.bio.ed.ac.uk/software/figtree/).

Quantitative Reverse Transcription-PCR
Fully outgrown green fruits of the dehiscent and indehiscent morph were collected separately from the same plants in three biological replicates. For each replicate, four dehiscent (four developing seeds in each fruit) or 16 indehiscent (one developing seed in each fruit) fruits were pooled into one sample, resulting in equal numbers of dehiscent and indehiscent seeds, respectively. Total RNA was extracted using the RNeasy Plant Mini Kit (Qiagen). Three micrograms of total RNA was treated with DNase I (Thermo Scientific) and precipitated with a one-tenth volume of 3 M sodium acetate (pH 5.2) and 2.5 volumes of ethanol. Two micrograms of DNase I-treated RNA was used for cDNA synthesis with random hexamer primers using the RevertAid H Minus First Strand cDNA Synthesis kit (Thermo Scientific). Quantitative reverse transcription-PCR using the primer pairs listed in Supplemental Table S3 was performed in duplicate using FastStart Essential DNA Green Master Mix (Roche) in the LightCycler 96 system (Roche) with the following parameters: 95°C for 10 min, 45 cycles at 95°C for 10 s and 60°C for 30 s, and one cycle at 95°C for 10 s, 60°C for 30 s, and 97°C for 1 s to obtain the melting curve for each reaction. Cycle threshold values were calculated using LightCycler 96 software (Roche). The geometric means of Ae. arabicum orthologs of ACTIN2 (AA26G00546), POLYUBIQUITIN10 (AA6G00219), and ANAPHASE-PROMOTING COMPLEX2 (AA61G00327) were used for data normalization. For each gene, the expression level in indehiscent fruits is presented as fold change relative to dehiscent fruits, for which the average expression level was set to 1.

Flow Cytometry
Two 7-d-old seedlings grown from seeds of each morph were used per sample. Endosperm tissue was prepared from 100 mature seeds per sample after imbibition for 12 h. Seedlings were chopped using a razor blade, either on their own or mixed with endosperm tissue, in 100 mL of Cystain UV Precise P extraction buffer (Partec). Samples were stained with 1 mL of Cystain UV Precise P DAPI fluorescent buffer (Partec) and filtered through a 30-mm filter. Three independent biological replicates consisting of 10,000 nuclei were analyzed using a Partec PAS flow cytometer. Histograms were analyzed using Flowing Software 2.5.1 (www.flowingsoftware. com), and the mean C values of endosperm nuclei with particular DNA contents were calculated relative to the mean 2C values of seedling nuclei.

Seed Germination, Moisture Content, and Water Activity Analysis
Dry mature seeds or fruits were placed in 3-cm petri dishes containing two layers of filter paper, 3 mL of distilled water, and 0.1% Plant Preservative Mixture (Plant Cell Technology). Plates were incubated in an MLR-350 Versatile Environmental Test Chamber (Sanyo-Panasonic) at 14°C and 100 mmol s 21 m 22 constant white light. Seed germination, scored as radicle emergence, of three biological replicates of 20 seeds or fruits each was analyzed. Dry seed moisture content was determined by drying four replicates of approximately 100 mg of air-dried fruits or seeds using the Mettler-Toledo moisture analyzer HB43-S for 4 h at 120°C. Water uptake during germination was analyzed by repeatedly weighing four replicates of approximately 100 seeds or 60 fruits imbibing under the same conditions as for the seed germination assay described above. Water activity was determined on four biological replicates of 20 M 2 seeds extracted from indehiscent fruits either before or after imbibition for 18 h under the conditions described above using a Labmaster-aw apparatus (Novasina).

Supplemental Data
The following supplemental materials are available.
Supplemental Figure S3. Fruits of Ae. arabicum fall into two discrete clusters.
Supplemental Figure S4. Phylogenies of Ae. arabicum orthologs of Arabidopsis fruit developmental genes.
Supplemental Figure S5. The temperature-induced shift in fruit-morph ratio is brought about by changes throughout the whole plant.
Supplemental Table S1. Geographic origins of Aethionemeae species used for the phylogeny.
Supplemental Table S2. List of species and respective sequences used for gene phylogeny reconstruction.
Supplemental Table S3. List of primers used for quantitative reverse transcription-PCR analysis.
Supplemental Data S1. Sequence alignment underlying species phylogeny.