Cytokinin-dependent specification of the functional megaspore in the Arabidopsis female gametophyte


For correspondence (e-mail


The life cycle of higher plants alternates between the diploid sporophytic and the haploid gametophytic phases. In angiosperms, male and female gametophytes develop within the sporophyte. During female gametophyte (FG) development, a single archesporial cell enlarges and differentiates into a megaspore mother cell, which then undergoes meiosis to give rise to four megaspores. In most species of higher plants, including Arabidopsis thaliana, the megaspore closest to the chalaza develops into the functional megaspore (FM), and the remaining three megaspores degenerate. Here, we examined the role of cytokinin signaling in FG development. We characterized the FG phenotype in three triple mutants harboring non-overlapping T–DNA insertions in cytokinin AHK receptors. We demonstrate that even the strongest mutant is not a complete null for the cytokinin receptors. Only the strongest mutant displayed a near fully penetrant disruption of FG development, and the weakest triple ahk mutant had only a modest FG phenotype. This suggests that cytokinin signaling is essential for FG development, but that only a low threshold of signaling activity is required for this function. Furthermore, we demonstrate that there is elevated cytokinin signaling localized in the chalaza of the ovule, which is enhanced by the asymmetric localization of cytokinin biosynthetic machinery and receptors. We show that an FM-specific marker is absent in the multiple ahk ovules, suggesting that disruption of cytokinin signaling elements in Arabidopsis blocks the FM specification. Together, this study reveals a chalazal-localized sporophytic cytokinin signal that plays an important role in FM specification in FG development.


In angiosperms, both male and female gametophytes are embedded in the sporophytic tissues, and the haploid and diploid generations coexist in a single organ. The surrounding sporophyte supports meiosis and subsequent gametophyte development in anthers and ovules, the male and female organs. The female gametophyte (embryo sac), unlike pollen, remains physically connected to the maternal tissues before and after fertilization. Ovules arise from the placenta within the carpels in flowering plants and are the site of gametogenesis, which can be divided into megasporogenesis and megagametogenesis (Reiser and Fischer, 1993; Schneitz et al., 1995; Ma and Sundaresan, 2010; Drews and Koltunow, 2011). Ovules have three structural domains along the proximal–distal axis: the funiculus, the chalaza, and the nucellus (Figure 1a). During megasporogenesis, a hypodermal archesporial cell enlarges and differentiates into a megaspore mother cell (MMC) that undergoes meiosis and gives rise to a tetrad of four megaspores (Figure 1b). In higher plants with Polygonum-type embryo sac development, including Arabidopsis, Oryza sativa (rice) and Zea mays (maize), the megaspore closest to the chalaza develops into the functional megaspore (FM), whereas the other three degenerate (Figure 1c). During the subsequent megagametogenesis, the FM undergoes three rounds of mitosis, resulting in an eight-nuclei syncytium that partitions into four cell types after cellularization: two synergid cells that are important for pollen tube guidance; the egg and central cells, which receive sperm cells for double fertilization; and three antipodal cells that degenerate by the last developmental stage of female gametophyte (FG7) (Figure 1d; Ma and Sundaresan, 2010; Sprunck and Gross-Hardt, 2011).

Figure 1.

Two-component elements and female gametophyte development. (a–d) Schematic depiction of wild-type female gametophyte development in Arabidopsis. Abbreviations: MMC, megaspore mother cell; nu, nucellus; ch, chalaza; fu, funiculus; oi, outer integuments, ii, inner integuments; FM, functional megaspore; dm, degenerating megaspores; ccn, central cell nucleus; ecn, egg cell nucleus; syn, synergid cell nuclei. (e) Model of cytokinin response pathway in Arabidopsis. Cytokinin binds to the AHK receptors, which initiate a phosphorelay through AHPs and ultimately result in the phosphorylation of type–B ARRs. The activated type–B ARRs elevate the transcription of the type–A ARRs, which in turn act to negatively regulate the pathway. CKI1, which acts gametophytically to regulate FG development after FG4, also acts through the AHPs. The number of genes in each family in Arabidopsis is shown in parentheses below each element.

Given the proximal connection between the sporophytic tissues and gametophytic cells, it is anticipated that cellular communication is important during the developmental processes (Bencivenga et al., 2011; Grossniklaus, 2011). A sporophytic siRNA pathway involving ARGONAUTE 9 (AGO9) is crucial for specifying cell fate in the Arabidopsis ovule (Olmedo-Monfil et al., 2010). In ago9 plants, more than one sub-epidermal cell enlarges and contains a conspicuous nucleus in the ovule (Olmedo-Monfil et al., 2010). Recent studies have begun to reveal the interactive role of hormone signaling between two generations (Pagnussat et al., 2009; Bencivenga et al., 2011; Kinoshita-Tsujimura and Kakimoto, 2011; Pérez-España et al., 2011). The cell fate specification during syncytial development depends on a micropylar auxin gradient correlated with local auxin biosynthesis (Pagnussat et al., 2009). Mutations in the cytokinin receptors disrupt female gametogenesis through a sporophytic effect (Nishimura et al., 2004; Kinoshita-Tsujimura and Kakimoto, 2011), yet the underlying mechanisms are yet to be elucidated.

Cytokinin is important in a wide array of developmental processes. Cytokinin perception in plants is similar to bacterial two-component phosphorelay signal transduction systems (TCSs). In Arabidopsis, there are three cytokinin receptors [Arabidopsis histidine kinases 2, (AHK2), AHK3 and AHK4/CRE1/WOL] that autophosphorylate in response to the binding of cytokinin (Inoue et al., 2001; Higuchi et al., 2004). AHKs then relay this phosphoryl group to the Arabidopsis histidine phosphotransfer proteins (AHPs), which in turn transfer the phosphoryl group to the Arabidopsis response regulators (ARRs) (Figure 1e; Hutchison et al., 2006; Suzuki et al., 1998). CKI1, which encodes a histidine kinase that lacks a cytokinin-binding domain, can also feed into downstream TCS signaling (Kakimoto, 1996; Hwang and Sheen, 2001), and acts to regulate female gametophyte development (Pischke et al., 2002; Hejatko et al., 2003; Deng et al., 2010). The ARRs fall into four classes based on phylogenetic analysis and domain structure: type–A ARRs, type–B ARRs, type–C ARRs and the Arabidopsis pseudoresponse regulators (APRRs). The 11 type–B ARRs are positive elements in the primary cytokinin signal transduction network. The 10 type–A ARRs are rapidly transcriptionally upregulated in response to cytokinin via direct activation by the type–B ARRs, and act to negatively regulate cytokinin signaling (Brandstatter and Kieber, 1998; Imamura et al., 1998; D'Agostino et al., 2000; Hwang and Sheen, 2001; Sakai et al., 2001; To et al., 2004).

In Arabidopsis and rice, mutants defective in biosynthesis and perception of cytokinin displayed reduced female fertility, suggesting that cytokinin has a conserved role in regulating ovule development (Higuchi et al., 2004; Nishimura et al., 2004; Kinoshita-Tsujimura and Kakimoto, 2011; Yamaki et al., 2011; Bencivenga et al., 2012). However, a triple receptor mutant that showed cytokinin deficiency in multiple bioassays was able to form seeds (Riefler et al., 2006), raising the possibility that the female-sterile phenotype was conditional. Here, we examined the role of cytokinin in female gametogenesis using three different ahk triple mutants, high-order ahp and type–B arr mutants. We demonstrate that sporophytic cytokinin signaling is essential for female gametophyte development. In addition, we demonstrate that FM specification depends on cytokinin signaling in the surrounding sporophyte, contributed to, at least in part, by the chalaza-enriched expression of the cytokinin receptors and of IPT1, which encodes an enzyme involved in cytokinin biosynthesis. These results provide evidence for a cytokinin-dependent pathway involved in the communication between the sporophytic and gametophytic tissues during female gametophyte development.


Loss of cytokinin receptors results in female gametophytic lethality

Previous studies suggest that disruption of the three cytokinin histidine kinase receptors (AHKs) affected the viability of the female gametophyte, although the effect appeared to vary depending on which ahk alleles were examined (Higuchi et al., 2004; Nishimura et al., 2004; Riefler et al., 2006). The ahk2–5 ahk3–7 cre1–2 line was reported to form some, albeit few, seeds under favorable conditions (Riefler et al., 2006). In contrast, three other allele combinations (ahk2–1 ahk3–1 ahk4–1, ahk2–2tk ahk3–3 cre1–12, ahk2–7 ahk3–3 cre1–12) failed to produce any seeds (Higuchi et al., 2004; Nishimura et al., 2004; Kinoshita-Tsujimura and Kakimoto, 2011). To determine if the differences in female sterility reported for these mutants result from different growth conditions, or if they reflect the specific allele combinations used in each study, we examined the three triple mutants that have the least overlap in ahk alleles (Figure 2a). Grown simultaneously under identical conditions, the ahk2–5 ahk3–7 cre1–2 line was capable of producing a few seeds, whereas the other two allele combinations failed to generate any seeds. We examined the morphology of the female gametophyte at the last developmental stage (FG7; Figure 1d) in each of these allele combinations. There were substantially fewer ovules per gynoecium in the ahk2–5 ahk3–7 cre1–2 mutant (13.2 ± 1.1, n = 76), as compared with the wild type (58.3 ± 5.1, = 260). Similarly, the numbers of ovules in each gynoecium were also severely reduced in ahk2–1 ahk3–1 ahk4–1 (12.5 ± 3.7, = 50) and ahk2–7 ahk3–3 cre1–12 (14.2 ± 0.4, = 108), consistent with previous studies linking cytokinin to the placenta activity, and thus to the number of ovules per gynoecium (Bartrina et al., 2011).

Figure 2.

Phenotypic analysis of different ahk triple mutants. (a) Cartoon representation of the three AHK genes with the positions of the various T–DNA insertions present in three multiple mutants used in this study. Boxes represent exons and triangles represent the sites of integration of T–DNA in each line. The three ahk alleles that compose each triple mutant are coded with the same color. Black boxes, CHASE domains; gray boxes, histidine kinase domains; yellow boxes, receiver domains; red bars, transmembrane domains. (b) Percentage of arrested female gametophytes in the wild type and various ahk triple mutants (> 50 from at least three plants for each mutant). (c) Aerial phenotype of 4–week-old wild type and various ahk triple mutants. Scale bar: 1 cm.

Although there were fewer ovules in the ahk2–5 ahk3–7 cre1–2 mutant, most (89%, n = 76) were morphologically indistinguishable from wild-type ovules (Figure 2b). In contrast, many of the ovules in ahk2–1 ahk3–1 ahk4–1 (51%, n = 50) and ahk2–7 ahk3–3 cre1–12 (97.1%, = 108) mutants lacked a differentiated embryo sac (Figure 2b), consistent with previous observations of ahk2–2tk ahk3–3 cre1–12 ovules (Kinoshita-Tsujimura and Kakimoto, 2011). Significantly, we observed female gametophytic lethality in two independent triple ahk mutants, each composed of different ahk alleles, demonstrating that the mutant phenotype is the result of a loss of the cytokinin receptors.

We examined the phenotypes of these three ahk triple mutants at the vegetative stage to determine if the strengths of the phenotypes at this stage matched the relative strengths of the ovule phenotypes. Four-week-old ahk2–5 ahk3–7 cre1–2 mutant rosettes were substantially smaller than the wild type, yet relatively larger than the two other ahk triple mutants examined (Figure 2c), consistent with the relative strengths of the ovule phenotypes. These data suggest that the threshold of AHK function necessary to support gametogenesis is extremely low, as ahk2–5 ahk3–7 cre1–2 is strongly insensitive to cytokinin in a number of response assays (Riefler et al., 2006), yet has relatively normal female gametes. For further analysis of the role of cytokinin in female gametophyte development, we used the ahk2–7 ahk3–3 cre1–12 line, which consistently has the highest percentage of arrested gametophytes.

The ahk2–7 ahk3–3 cre1–12 line displayed the strongest effect on both rosette morphology and FG development, and thus is likely to represent the line closest to a cytokinin receptor null. However, even for this line, it has been suggested that there could be residual full-length transcript from the ahk3–3 allele (Higuchi et al., 2004). To explore this, we performed qRT-PCR using gene-specific primers spanning the T–DNA insertion sites for all three AHK genes in ahk2–7 ahk3–3 cre1–12. AHK4/CRE1 was not detected, even with saturating PCR cycles, whereas AHK2 and AHK3 were detected at severely reduced levels (approximately 0.02 and 0.8%, respectively, compared with the wild type; Figure S1b). Sequence analysis revealed that the splice junction spanning the intron containing the T–DNA insertion in ahk3–3 was correctly spliced in this residual transcript. Thus, none of the published AHK triple mutants represent a complete disruption of cytokinin receptors, leaving open the possibility that cytokinin may be essential for sporophytic plant development.

Impaired cytokinin signaling results in incomplete megasporogenesis

We analysed the female gametophyte lethality of the double and triple ahk mutants after allowing megagametogenesis to progress to FG7. The double mutants (ahk2–7 ahk3–3, ahk3–3 cre1–12 and ahk2–7 cre1–12) were indistinguishable from the wild type (> 207 ovules observed for each mutant; Figure 3a). Female gametophyte development did not differ from the wild type in the ahk2–7 ahk3–3/AHK3 cre1–12 line (one copy of AHK3 remaining; = 304), yet there was a slight increase in the frequency of arrested ovules in ahk2-7/AHK2 ahk3–3 cre1–12 (13.3%, = 352) and ahk2–7 ahk3–3 cre1–12/CRE1 (7.2%, = 270; Figure 3b). In contrast, in the ahk2–7 ahk3–3 cre1–12 triple mutant, an embryo sac was absent in nearly all of the ovules (97.1%, = 108 from seven different plants; Figure 3a). Together, these results suggest that the three cytokinin receptors have redundant functions in female gametophyte development.

Figure 3.

Characterization of the ovule phenotypes in the ahk2–7 ahk3–3 cre1–12 mutant. (a) Analysis of female gametophyte development at FG7 in multiple ahk mutants (n > 250 per line). (b) Analysis of female gametophyte development at FG7 in the double-homozygous single heterozygous ahk mutant. The frequency of arrested female gametophytes in the ahk2–7 ahk3–3/AHK3 cre1–12 (n = 304) line did not differ from that in the wild type; however, there is an increase in the ahk2–7/AHK2 ahk3–3 cre1–12 (n = 352) and ahk2–7 ahk3–3 cre1–12/CRE1–12 (n = 270) mutants. (c–j) Wild-type ovule development. (c) The ovule, shortly after initiation, consists of nucellus (nu), chalaza (ch) and a funiculus. A subepidermal cell enlarges and becomes the megaspore mother cell (MMC). (d) The MMC undergoes meiosis and gives rise to four megaspores. The cell closest to the chalaza becomes the functional megaspore (FM). The three non-functional megaspores (dm) degenerate shortly after meiosis. (e) The last stage of megagametogenesis, FG7. The FM undergoes three rounds of mitoses, resulting in an eight-nucleate syncytium that is later partitioned into seven cells consisting of four cell types. Antipodal cells degenerate by FG7 and are not shown. ccn, central cell nucleus; ecn, egg cell nucleus; syn, synergid cell nuclei. (f–j) Ovule development in the ahk2–7 ahk3–3 cre1–12 mutant. (f) The mutant ovules prior to meiosis with a single enlarged MMC. Note that there are fewer cells in the chalaza (ch) compared with the wild type in (a). (g) The mutant ovule after meiosis. Note that none of the tetrad of megaspores enlarges to form the functional megaspore as in the wild type (b). (h–j) Distinct phenotypes of the ahk2–7 ahk3–3 cre1–12 female gametophytes at the FG7 stage. (h) A cavity containing two nuclei, indicated by the arrowheads. (i) Absence of an embryo sac. (j) A developed embryo sac with a central cell-like nucleus. Note the altered morphology of this embryo sac compared with the wild type in (e). Scale bars: 10 μm (c–j).

We next examined the development in wild-type and ahk2–7 ahk3–3 cre1–12 ovules to determine the stage at which the morphology of the mutants first departs from that of the wild type. During megasporogenesis, one subepidermal archesporial cell differentiates into a megaspore mother cell (MMC) (Figure 3c). In ahk2–7 ahk3–3 cre1–12 ovules, the specification of MMC, meiotic division and tetrad formation did not differ from the wild type (Figure 3c,f). The earliest stage at which morphological defects were observed in the ahk triple mutant was at the specification of the functional megaspore (FM) after meiosis (Figure 3d,g). In the wild type, the megaspore closest to the chalaza enlarges and gives rise to the FM, and ultimately the embryo sac, whereas the other three megaspores degenerate shortly after meiosis (Figure 3d). In the ahk2–7 ahk3–3 cre1–12 mutant, ovule development was generally asynchronous, as compared with the wild type. The gametogenesis proceeded normally through tetrad formation. However, often the subsequent specification of the FM did not occur in the mutant ovules (37.1%, = 108; Figure 3g,i). In some cases (59.9%, = 108), an embryo sac-like cavity formed, but contained only one or two nuclei (Figure 3h). A few ovules (3%, = 108) contained a developing embryo sac, but the morphology differed from the wild type (Figure 3j). We observed similar phenotypes in another triple cytokinin receptor mutant: ahk2–1 ahk3–1 ahk4–1 (Figure S2).

CKI1, an Arabidopsis histidine protein kinase, is similar to the AHK cytokinin receptors, but it lacks the cytokinin-binding CHASE domain. CKI1 is essential for megagametogenesis: the earliest stage at which phenotypic abnormalities occurred in cki1 female gametophytes is at the four-nucleate stage (FG4), after the FM undergoes two rounds of mitosis. The cki1 phenotype becomes completely penetrant by the eight-nucleate stage (FG5/6) (Pischke et al., 2002; Hejatko et al., 2003). Previous studies have shown that the AHPs act downstream of CKI1 in megagametogenesis (Deng et al., 2010). Here, we explored whether the AHPs also act downstream of cytokinin receptor AHKs to regulate megasporogenesis.

The quintuple ahp1 ahp2–1 ahp3 ahp4 ahp5 mutant is disrupted for all five AHP genes in the Arabidopsis genome, although the ahp2–1 allele retains some AHP2 function (Hutchison et al., 2006). This ahp quintuple mutant displays strongly reduced sensitivity to cytokinin and reduced fertility (Hutchison et al., 2006). Most (81.7%, = 278) of the ovules in the ahp1 ahp2–1 ahp3 ahp4 ahp5 mutant were arrested at FG7 (Figure 4a). The majority of these (65.9%, = 278) displayed a cki1-like phenotype (increased number of nuclei in embryo sac or degenerating embryo sac; Figure 4c and S3), but a substantial number (15.8%, = 278) showed an ahk-like phenotype (lack of embryo sac; Figure  4b). This suggests that the AHPs act downstream of the AHKs in megasporogenesis, in addition to their function downstream of CKI1 in megagametogenesis.

Figure 4.

Both the ahk- and cki1-like phenotypes are present in multiple ahp and type–B arr mutants. (a) Percentage of arrested female gametophytes in the wild type and multiple ahp and arr mutants (n > 270 for each line). (b, c) The morphology of the arrested ovules in ahp1 ahp2–1 ahp3 ahp4 ahp5. (b) Lack of embryo sac formation in the ahp quintuple mutant, which is similar to the ovule phenotype of the ahk triple mutants (Figure 3i). (c) Degenerating embryo sac in the ahp quintuple ovule, which is similar to the ovule phenotype of cki1–5 (Figure S3a). (d) Percentage of ahk- and cki1-like ovule phenotypes present in multiple ahp and arr mutants. (e–f) The morphology of the arrested ovules in the arr1–3 arr2–2 arr10–2 arr12–1 mutant. (e) The ahk-like phenotype (lack of embryo sac). (f) The cki1-like phenotype (degenerating embryo sac). Scale bars: 10 μm.

To further decipher the downstream two-component elements in regulating megasporogenesis, we analysed ovules from a quadruple type–B arr mutant (arr1–3 arr2–2 arr10–2 arr12–1), which, like the multiple ahp and ahk mutants, displays insensitivity to high levels of exogenous cytokinin (Rashotte et al., 2006). We observed a substantial number of ovules arrested at FG7 in this quadruple type–B mutant (21.9%, = 276; Figure 4a), with 10.0% (= 276) showing an ahk-like phenotype (Figure 4e) and 11.9% (= 276) showing a cki1-like phenotype (Figure 4f). Mutations in ARR7 and ARR15, two of the type–A ARRs, were previously reported to affect female gametophyte development (Leibfried et al., 2005). More recent studies indicate that the lethality is the result of chromosome rearrangements present in the first alleles analyzed (Zhang et al., 2011). Most (98.9%, = 188) of the ovules in octuple arr3 arr4 arr5 arr6 arr7arr8 arr9 arr15 mutants were morphologically indistinguishable from the wild type, suggesting that these type–A ARRs do not play an essential role in female gametophyte development. Taken together, the dual ovule phenotype observed in the higher order ahp and type–B arr mutants (Figure 4d) suggest that these elements are essential in both megasporogenesis and megagametogenesis, acting downstream of both AHK and CKI1, respectively.

Cytokinin receptors and activity are enriched in the chalaza of the ovule

Mutations affecting embryo sac development may act through either the diploid sporophyte or the haploid gametophyte. The phenotypes of gametophytic mutations are solely determined by the haploid genotype of the embryo sac. In contrast, sporophytic mutations affect all gametophytes within the ovaries of a mutant plant: thus the progeny of a selfed heterozygous recessive mutation will segregate wild-type and mutant genotypes with a Mendelian ratio. Previous studies have suggested that female gametophyte (FG) development depends on the cytokinin receptors (AHKs) in the sporophyte (Nishimura et al., 2004; Kinoshita-Tsujimura and Kakimoto, 2011). Single heterozygous ahk2 showed a Mendelian segregation in a double homozygous receptor mutant background (ahk3 ahk4; Kinoshita-Tsujimura and Kakimoto, 2011; Nishimura et al., 2004). Consistently, we observed that 25% of progeny seedlings from selfed ahk2–7/AHK2 ahk3–3/ahk3–3 cre1–12/cre1–12 or ahk2–7/ahk2–7 ahk3–3/AHK3 cre1–12/cre1–12 mutants displayed the triple mutant phenotype (Table S1).

To better understand the functions of cytokinin during female gametophyte development, we analyzed transgenic plants carrying a GUS reporter driven by the native promoter of AHK2, AHK3 or CRE1. The intensity of the AHK2, AHK3 and CRE1 signal was high throughout megasporogenesis. At ovule stage 2III (the stage after that reported upon by Schneitz et al., 1995), when the outer integument is initiated and meiosis has not yet occurred, AHK2, AHK3 and AHK4 were expressed preferentially at the chalazal end of the ovule (Figure 5a–f). AHK2 and AHK4 were also expressed in the inner integument primordia (Figure 5a,b,e,f). When stained for extended times, AHK2 and AHK3 expression was observed diffusedly spread throughout the nucellus, but still with more intense expression at the chalaza (Figure 5b,d). AHK4 was expressed almost exclusively in the chalaza, with little or no expression in the nucellus during megasporogenesis (Figure 5e,f). Expression of these AHK genes in the chalaza continued through at least stage FG2. This pattern of expression is similar to that presented previously, though in that study, the authors did not note the enrichment of AHK expression in the chalaza (Bencivenga et al., 2012).

Figure 5.

Expression of cytokinin signaling elements in ovules. (a–f) Expression of cytokinin receptors as revealed by fusions of the respective regulatory element to a GUS reporter. Lines harboring an AHK2 promoter-protein:GUS fusion (a, b), an AHK3 promoter-protein:GUS fusion (c, d) or a CRE1 promoter::GUS fusion (e, f) were stained for GUS activity before meiosis, when the phenotype in ahk2–7 ahk3–3 cre1–12 was about to develop. Unsaturated (a, c, e) and saturated (b, d, f) staining are shown to reveal the preferential expression in each domain. (a–d) AHK2 and AHK3 are diffusedly detected in the ovule, but are enriched in the chalaza and the integument primordial. (e, f) AHK4 is specifically localized at the chalaza and inner integument primordia. (g) Visualization of the expression of the cytokinin-responsive ARR4::GUS transgene in pre-meiotic ovules. (h) Expression of IPT1 as revealed by the fusion of promoter and coding region to a GUS reporter. Abbreviations: ch, chalaza; ii, inner integument; MMC, megaspore mother cell; nu, nucellus; oi, outer integument. Scale bars: 10 μm.

To analyze the cytokinin responses in these tissues, we examined the expression of ARR4, one of the type–A ARRs that is a cytokinin primary response gene (Brandstatter and Kieber, 1998). The expression of ARR4 was localized in the chalaza at stage 2III (Figure 5g), overlapping with the expression pattern of the cytokinin receptors (Figure 5a–f). This is consistent with the pattern of expression of the TCS-GFP transgene (Bencivenga et al., 2012), which reports cytokinin signaling (Müller and Sheen, 2008). To further explore the origin of the cytokinin gradient, the expression of different isopentenyl transferases (IPTs) that catalyze the rate-limiting step of cytokinin biosynthesis in Arabidopsis were studied in the ovules. Plants carrying a fusion of the promoter and coding region of IPT1 to GUS were used to study cytokinin biosynthesis in the developing ovules. IPT1 expression was observed during megasporogenesis and megagametogenesis (Figures 5h and S4). In stage 2III, when the ovule phenotype in the ahk2–7 ahk3–3 cre1–12 mutant has not occurred, IPT1 was observed in the MMC, companion cells and the surrounding sporophytic tissues (Figure S4a). During tetrad formation, IPT1 expression was observed in the haploid megaspore and the diploid sporophytic cells, with higher intensity in the chalaza end of the ovule (Figure 5h and S4b). Together, these data indicate that cytokinin signaling and biosynthesis is asymmetrically distributed in the ovule, predominantly at the chalaza during megasporogenesis.

Chalazal cytokinin signaling is required for the selection of the functional megaspore

Megasporogenesis and megagametogenesis occur along the chalazal–micropylar axis of the ovule. In most species of flowering plants, including Arabidopsis, the megaspore located closest to the chalaza survives after meiosis, and differentiates into the FM. The lack of development of the FM in the triple cytokinin receptor mutant suggests that the elevated cytokinin signaling in the surrounding chalazal maternal tissues conveys positional information involved in the specification of the FM. To examine this hypothesis, we introduced an FM-specific marker (pFM2::GUS) into the ahk2–7 ahk3–3 cre1–12 background. The pFM2 promoter first drives expression specifically in the functional megaspore, but not in the degenerating megaspores, and its expression is sustained through FG7 (Olmedo-Monfil et al., 2010). In wild-type plants, pFM2 expression was detected in the chalazal-most megaspore (Figure 6a). In contrast, in an F3 population homozygous for pFM2::GUS in the ahk2–7 ahk3–3 cre1–12/CRE1 background, pFM2 was absent in the ovules that failed to form a morphologically recognizable FM (Figure 6b), whereas the sibling ovules in the same gynecium with a morphologically recognizable FM displayed GUS expression (Figure 6c). Because female gametogenesis is asynchronous in the ahk mutants and pFM2::GUS expression persists through FG7, we allowed megagametogenesis to progress to FG6/7 to avoid false-negative results. At FG6/7, pFM2 was present in 96.2% of the wild-type ovules (n = 373; Figure 6d,e). In pFM2::GUS; ahk2–7 ahk3–3 cre1–12/CRE1 plants, however, pFM2 was present in only 69.9% of the ovules (n = 302; Figure 6f–i). An embryo sac-like cavity (Figure 6g) was observed in some of the pFM2-negative ovules, whereas in others, the embryo sac was completely absent (Figure 6h). This frequency of FM2-negative ovules in the pFM2::GUS; ahk2–7 ahk3–3 cre1–12/CRE1 line (30.1%) is higher than the frequency of morphologically defective FG (8.9%, n > 200) in this line examined at the same time, suggesting that the AHKs are important in both the specification and maintenance of FM identity, or at least maintenance of FM2 expression.

Figure 6.

Cytokinin signaling is required for functional megaspore selection. (a) pFM2 expression in wild-type ovules after meiosis. Note the marker specifically stains the functional megaspore (FM) but not the degenerating megaspores (dm). (b, c) Expression of pFM2::GUS in ovules from the ahk2–7 ahk3–3 cre1–12/CRE1 mutants. (b) After meiosis, pFM2 was absent in the ovules that failed to form a morphologically identifiable FM. (c) pFM2 expression was detected in the sibling ovules from the same gynecium as shown in (b). (d–e) pFM2 expression in wild-type ovules persists through FG6–FG7. (f–i) FM specification is compromised in ovules from ahk2–7 ahk3–3 cre1-12/CRE1. (f) At FG6–FG7, pFM2 was absent in 30.1% of the ovules (n = 302) that had a cavity (g) or did not form an FM (h). (i) pFM2 was present in a subset of the sibling ovules of (g) that had a specified FM.


Two-component signaling in female gametophyte development

In this study, we demonstrate that all three elements of two-component signaling, the AHKs, AHPs and type–B ARRs, are necessary in the sporophyte for proper FG development (Figure 7). We grew three different ahk triple mutants simultaneously, and concluded that the differences in their fertility were not because of growth conditions, but rather because of the strength of the various alleles in these mutants. The cytokinin threshold required for FG development appears to be very low, as the partially fertile ahk2–5 ahk3–7 cre1–2 line is strongly insensitive to cytokinin in multiple response assays. We characterized the FG phenotype in the strongest mutant ahk2–7 ahk3–3 cre1–12 and found that the first deviation from the wild type occurred during functional megaspore (FM) specification, which resulted in the subsequent absence of embryo sac formation (ahk-phenotype). This is distinct from the phenotype described recently by Bencivenga et al., which focused on the lack of integument development in the ahk2–2tk ahk3–3 cre1–12 line (Bencivenga et al., 2012).

Figure 7.

Two-component signaling in female gametophyte development. In megasporogenesis, SPL/NZZ is required for the megaspore mother cell specification (Schiefthaler et al., 1999; Yang et al., 1999). The data presented here suggest that an AHK-AHP-type-B ARR signaling pathway acting in the sporophyte is responsible for FM specification. CKI1 is essential in the gametophyte during megagametogenesis (Pischke et al., 2002; Hejatko et al., 2003). The AHPs (Deng et al., 2010) and type–B ARRs also act in the CKI1-dependent pathway in the megagametogenesis.

A previous study has demonstrated an essential role for the AHPs in the megagameogenesis (Deng et al., 2010). Here, we demonstrate that disruption of the AHPs results in two distinct ovule phenotypes, suggesting that these elements act downstream of both the AHKs and CKI1, in the sporophyte and gametophyte, respectively. A sporophytic role for the AHPs is supported by the transmission efficiency of T–DNA insertions at the AHP3 or AHP5 locus in ahp1 ahp2–1 ahp3/+ahp4 ahp5 or ahp1 ahp2–1 ahp3 ahp4 ahp5/+, respectively, which were transmitted at the expected 1 : 1 ratio in reciprocal crosses (Deng et al., 2010). Together, these results indicate that the AHPs act in the sporophyte as well in the gametophyte.

Reproductive defects have not been reported in previous analyses of type–B arr mutants (Mason et al., 2004, 2005; Yokoyama et al., 2007; Argyros et al., 2008). The arr1–3 arr2–2 arr10–2 arr12–1 quadruple mutant examined in this study had two distinct (ahk- and cki1-like) phenotypes, with moderate penetrance. This quadruple mutant has mutations in four out of seven members of the subfamily of type–B ARRs linked to cytokinin signaling; therefore, other members may account for the remaining fertility, as these genes have been shown to be functionally redundant in other contexts (Mason et al., 2005; Yokoyama et al., 2007; Argyros et al., 2008; Ishida et al., 2008). Our data reveal a role for type–B ARRs in FG development, acting in the sporophyte and gametophyte through AHK- and CKI1-dependent pathways, respectively.

Cytokinin is essential for functional megaspore specification

Several lines of evidence support the notion that sporophytic cytokinin plays a role in the specification of the FM. Disruption of cytokinin signaling in the sporophytic tissue results in a lack of embryo sac formation. The earliest defects were observed in FM specification, as shown both morphologically and by using an FM-specific marker. Further, cytokinin signaling is enriched in the chalazal sporophytic tissue, as revealed by the localized expression of the cytokinin primary response gene ARR4 (Figure 5g), and by the expression of the cytokinin reporter TCS-GFP (Bencivenga et al., 2012). Together, these results suggest that chalazal-enriched cytokinin signaling confers positional information involved in specifying and maintaining FM development.

Whereas cytokinin signaling is clearly involved in FM specification, the FM-absent phenotype is not completely penetrant, even in the strongest cytokinin receptor mutant, ahk2–7 ahk3–3 cre1–12. In this mutant, however, residual full-length AHK3 transcript was detected from the ahk3–3 allele (Figure S1b), indicating that it is not a complete null for cytokinin receptors. Our analysis of three different ahk mutants suggests that the threshold of cytokinin function required for FG development is extremely low; thus the cytokinin signaling derived from the residual AHK3 in ahk2–7 ahk3–3 cre1–12 may be sufficient to drive some FM formation.

An alternative, not mutually exclusive, model for the residual FM formation in the triple cytokinin receptor mutants is that there are overlapping pathways mediating FM specification. One possibility for such a redundant factor is SPOROCYTELESS/NOZZLE (SPL/NZZ), the master transcription factor that links pattern formation and growth control during ovule development (Schiefthaler et al., 1999; Yang et al., 1999). Mutation in SPL/NZZ results in complete male and female sterility through a sporophytic effect (Schiefthaler et al., 1999; Yang et al., 1999). Recent transcriptome analysis of developing ovules revealed that multiple cytokinin-responsive type–A ARRs are elevated in the spl ovules (Johnston et al., 2007). In addition, AHK4 expression, but not AHK2 or AHK3, is abolished in spl/nzz ovules compared with wild type (Sánchez-León et al., 2012), and SPL/NZZ is induced in ovules upon cytokinin treatment (Bencivenga et al., 2012). Overall, these results suggest an intriguing, although complicated, interaction between SPL/NZZ and cytokinin signaling.

A recent study proposes that SPL/NZZ is required for the cytokinin-induced expression of auxin efflux protein PIN1 in the FG, and this regulation accounts for integument development in the ovules (Bencivenga et al., 2012). The lack of integuments (leading to finger-like structures), however, represents only approximately 10% of the phenotypes observed in the ahk2–2tk ahk3–3 cre1–12 line examined in that study, and is not observed in spl/nzz ovules (Bencivenga et al., 2012). In spl/nzz ovules, the subepidermal archesporial cell in the nucellus fails to differentiate into the megaspore mother cell (Schiefthaler et al., 1999; Yang et al., 1999). The phenotype of spl/nzz ovules occurs in a developmental window earlier than that which requires AHK function (FM specification post-meiotically; Figures 7 and S5).

Signaling between the sporophyte and the gametophyte involves cytokinin

Whereas cytokinin is important in FM specification, it is unlikely that cytokinin itself acts as the signal to determine FM fate in the gametophyte, as cytokinin receptor function is not required in the gametophytic cells (Table S1). Thus, the chalazal-enriched cytokinin signaling probably results in the generation of a distinct signal to direct FM fate, although further studies are needed to elucidate the nature of this signal. Although our studies suggest a cytokinin-dependent signal from the sporophyte to the gametophyte, the effect of the spl mutation on the expression of multiple cytokinin-responsive type–A ARRs (Johnston et al., 2007) suggests that gametophyte development also regulates cytokinin signaling in the surrounding sporophytic tissues. Thus, it appears that there is bidirectional communication between the sporophyte and the gametophyte that involves cytokinin.

The question arises as to when the putative sporophytic cytokinin-dependent signal acts during female gametogenesis. It has been hypothesized that the developmental cues mediating the selection of the FM may be present in the sporophytic tissues before meiosis (Willenmse, 1981; Huang and Russell, 1992; Reiser and Fischer, 1993), as prior to meiosis cellular organelles are asymmetrically distributed in the megaspore mother cell: more plastids are located in the chalazal region, whereas rough endoplasmic reticulum and vacuoles are more abundant at the micropylar region (Bajon et al., 1999). This asymmetry in the MMC may lead to an asymmetry in the resulting meiotic products, and may be set up by the localized cytokinin in the sporophytic tissue. Alternatively, the cytokinin may signal the single megaspore closest to the chalaza to adopt an FM fate after meiosis has occurred.

Origin of chalaza-localized cytokinin

Cytokinin signaling is enhanced in the chalaza of the sporophytic tissue of the ovule, as revealed by the localized expression of ARR4 (Figure 5g) and the TCS-GFP cytokinin reporter (Bencivenga et al., 2012). This probably results in part from localized expression of cytokinin signaling elements, but also from localized cytokinin biosynthesis, as revealed by localized IPT1 expression. However, in the atipt1 atipt3 atipt5 atipt7 quadruple mutant, in which four of the seven IPT genes are disrupted, no ovule nor seed set phenotypes were reported (Miyawaki et al., 2006). This suggests that residual local cytokinin biosynthesis coming from other lowly expressed IPTs in this tissue may be sufficient for ovule development, consistent with the notion that the threshold of cytokinin function required for FG development is very low. Alternatively, the mobile nature of cytokinin may allow it to serve systematically as a long distance signal. Other elements of cytokinin biosynthesis, such as the LOG genes (Kuroha et al., 2009), and cytokinin metabolism, such as the cytokinin oxidase genes (Bartrina et al., 2011), may also contribute to the generation of a localized cytokinin signal.


Plant hormones are involved in all aspects of plant growth and development. Perturbation in the signaling pathways of a number of phytohormones, including auxin, brassinosteroid (BR), cytokinin and ethylene results in defects in the development of ovules (Drews et al., 1998; Pagnussat et al., 2009; Pérez-España et al., 2011). Our data demonstrate that a chalazal-enriched cytokinin signal conveys positional information to specify FM. Future studies are needed to elucidate the nature of this cytokinin-dependent signal emanating from the diploid sporophytic tissues, and to determine how it regulates FM specification.

Experimental procedures

Plant materials and growth conditions

The ahk2–1 ahk3–1 ahk4–1 (Nishimura et al., 2004), ahk2–5 ahk3–7 cre1–2 (Riefler et al., 2006), ahk2–7 ahk3–3 cre1–12 (Figure S1) (Argyros et al., 2008), ahp1 ahp2–1 ahp3 ahp4 ahp5 (Hutchison and Kieber, 2007), arr1–3 arr2–2 arr10–2 arr12–1 (Rashotte et al., 2006), AHK2:GUS (Nishimura et al., 2004), AHK3:GUS, CRE1::GUS (Higuchi et al., 2004), ARR4::GUS (To et al., 2004) and pFM2::GUS (Olmedo-Monfil et al., 2010) mutants are in the Columbia (Col–0) ecotype. spl–1 (Yang et al., 1999) is in the Landsberg erecta (Ler) background. cki1–5 (Pischke et al., 2002) is in the Ws background.

Arabidopsis seeds were surface-sterilized, cold-treated at 4°C for 3 days and germinated on MS/1% sucrose vertical plates under constant light, as previously described (To et al., 2004). Seedlings were transferred to soil and kept in a growth chamber under long days (16–h light/8–h dark) at 22°C.

Histology and microscopy

For phenotypic analysis, we used the sixth flower or older, except for the ahk triple mutants, which had very few flowers. Ovules were dissected on a slide and fixed in Carnoy's fixative (3:1 ethanol:acetic acid) for 15 min, incubated in 70% ethanol for 15 min, transferred to water for 20 min, mounted in chlorohydrate:water (8:3) and finally photographed with a Nikon E800 photomicroscope equipped with a Nikon Plan Apo 100/1.40 oil immersion objective ( using differential interference contrast (DIC) optics.

For β–glucuronidase staining, dissected ovules were fixed in cold 90% acetone for 20 min, washed for 10 min in cold rinsing buffer (100 mm sodium phosphate, pH 7.0, 0.5% Triton X–100, 0.5 mm potassium ferro/ferricyanide) and vacuum infiltrated for 5 min in 1 mg ml−1 X–gluc (5–bromo-4-chloro-3-indolyl-β-d glucuronic acid). The samples were incubated at 37°C for 6–48 h, depending on lines. Ovules were mounted in chlorohydrate:water (8:3) and photographed under DIC optics, as described above.

Expression of AHK in ahk2–7 ahk3–3 cre1–12

RNA was isolated from the shoots of 4–week-old plants using an RNeasy kit (Qiagen, cDNA was prepared from the total RNA using SuperScript III reverse transcriptase (Invitrogen, following the manufacturer's instructions. Real-time PCR was performed sing SYBR Premix Ex Taq polymerase (TaKaRa, in a ViiATM 7 Real-Time PCR system (ABI, The following primers were used for gene expression: AHK2, 5′–ACGCGCCAGTTATATTTGCT–3′ and 5′–GCGGTAGGCTCGTGTCATAG–3′; AHK3, 5′–CAGCTCAAGAAAAAGGCTGAA–3′ and 5′–TGCGGTCCTAACATAATCCTG–3′; AHK4/CRE1, 5′–ACAATGGATAGAGGAGAGCCTTC–3′ and 5′–ATGGTGAGTTTCCAACAACCTAA–3′; TUB4, 5′–AGAGGTTGACGAGCAGATGA–3′ and 5′–ACCAATGAAAGTAGACGCCA–3′.


We would like to thank Tatsuo Kakimoto, Chiharu Ueguchi and Thomas Schmülling for providing the ahk mutants, Chiharu Ueguchi for providing the AHK2:GUS line, Tatsuo Kakimoto for providing the AHK3:GUS, CRE1::GUS, and AtIPT1:GUS lines, Venkatesan Sundaresan for providing the spl–1 mutant line and Jean-Philippe Vielle-Calzada for providing the pFM2::GUS marker line. We are grateful to Sara Ploense for technical guidance and helpful discussions. We thank Wenjing Zhang, Smadar Harpaz-Saad and Yu-Chang Tsai for stimulating discussions and their critical reviews of the article. This work was supported by National Science Foundation grant IOS 0618286 to J.J.K. and G.E.S. The authors declare no conflict of interest.