Polycomb group genes control developmental timing of endosperm


  • Mathieu Ingouff,

    1. European Molecular Biology Organization (EMBO) YIP Team, Unité Mixte de Recherche 5667, Ecole Normale Supérieure de Lyon, Centre National de la Recherche Scientifique, Institut National de la Recherche Agronomique, Université de Lyon I, 46 Allée d'Italie, F-69364 Lyon Cedex 07, France, and
    Search for more papers by this author
    • Present address: Chromatin and Reproduction Group, Temasek Lifescience Laboratory, 1 Research Link, National University of Singapore, Singapore 117604.

  • Jim Haseloff,

    1. Department of Plant Sciences, University of Cambridge, Downing Street, Cambridge CB2 3EA, UK
    Search for more papers by this author
  • Frédéric Berger

    Corresponding author
    1. European Molecular Biology Organization (EMBO) YIP Team, Unité Mixte de Recherche 5667, Ecole Normale Supérieure de Lyon, Centre National de la Recherche Scientifique, Institut National de la Recherche Agronomique, Université de Lyon I, 46 Allée d'Italie, F-69364 Lyon Cedex 07, France, and
    Search for more papers by this author
    • Present address: Chromatin and Reproduction Group, Temasek Lifescience Laboratory, 1 Research Link, National University of Singapore, Singapore 117604.

(fax +65 687 270 07; e-mail fred@tll.gov.sg).


Polycomb (PcG) group proteins form modular complexes, which maintain repressed transcriptional states of target genes across cell divisions. As PcG complexes provide a memory of cell fate, such proteins might control temporal aspects of development. Loss-of-function of any of the FERTILIZATION INDEPENDENT SEED (FIS) PcG genes perturbs endosperm development. In this report we provide a detailed analysis of the phenotype of fis endosperm development using molecular and cellular markers. Wild type (WT) endosperm development undergoes a series of four major developmental phases timed by successive synchronous nuclei division. In fis endosperm the transition from phase 1, marked by a synchronous mode of nuclei divisions to phase 2, corresponding to the establishment of three mitotic domains, is absent. Accordingly, the expression of seven markers of phase 1 and phase 2 is temporally perturbed. In spite of such changes, specific sequences of developmental events still take place as in the WT. Overall, fis mutations are heterochronic mutations that cause a temporal deregulation in the ontogenic sequence of endosperm development.


Plants pass through determinate successive sequences of developmental phases. A vegetative phase is followed by an adult vegetative phase and by a reproductive phase including successive flowering, gametogenesis, fertilization and seed development. Heterochronic mutations alter the order of developmental steps by changing the relative timing of developmental events (Slack and Ruvkun, 1997). Most heterochronic mutations described in plants alter the sequence of morphological changes during the transition from a juvenile to an adult vegetative stage (Berardini et al., 2001; Dudley and Poethig, 1991; Evans et al., 1994; Itoh et al., 1998; Telfer and Poethig, 1998; Telfer et al., 1997). With the exception of the temporal control of flowering (Henderson et al., 2003; Mouradov et al., 2002; Putterill et al., 2004) reports of alterations of developmental timing of the reproductive programme have been scarce (Grimanelli et al., 2003; Raz et al., 2001).

In flowering plants, double fertilization results in the formation of two zygotic products within the seed. The embryo will give rise to the juvenile plant and the endosperm nurtures the developing embryo (Berger, 2003). Endosperm development is divided into a series of successive developmental steps timed by synchronous nuclei divisions (Boisnard-Lorig et al., 2001; Brown et al., 1999; Guitton et al., 2004; Mansfield and Briarty, 1990). Hence, the endosperm constitutes a good model to investigate mechanisms involved in temporal controls.

In Arabidopsis, the three FERTILIZATION INDEPENDENT SEED (FIS) genes MEDEA (MEA), FERTILIZATION INDEPENDENT SEED 2 (FIS2) and FERTILIZATION-INDEPENDENT ENDOSPERM (FIE) encode members of the Polycomb group family (PcG) and are homologous to Enhancer of Zeste (E[z]), Suppressor of Zeste 12 (Su(z)12) and Extra Sex Combs (ESC) respectively (Grossniklaus et al., 1998; Kiyosue et al., 1999; Luo et al., 1999; Ohad et al., 1999). PcG proteins assemble in chromatin remodelling complexes and repress transcriptional activity of target genes (Francis and Kingston, 2001). Plant genomes contain only homologues of the Polycomb repressive complex 2 (PRC2) defined by E[z]/Su(z)12/Esc (Reyes and Grossniklaus, 2003). Similar to Drosophila, the plant PRC2 complex contains MULTICOPY SUPPRESSOR OF IRA1 (MSI1), a WD40 protein homologous to the Drosophila Retinoblastoma (Rb) binding protein P55 (Köhler et al., 2003a; Mosquna et al., 2004). Mutations in MSI1 cause a pleiotropic phenotype and causes defects in endosperm similar to those reported in fis mutants (Guitton et al., 2004).

The fis mutants were originally reported for autonomous development of seed in the absence of fertilization (Chaudhury et al., 1997; Ohad et al., 1996; Peacock et al., 1995). If fertilization takes place, fis mutations perturb endosperm development. In contrast to wild type (WT) endosperm, fis endosperm does not undergo cellularization but remains syncytial and sustains a high rate of proliferation during late development (Chaudhury et al., 1997; Kiyosue et al., 1999). It was thus concluded that FIS genes negatively regulate endosperm growth and proliferation. Moreover, fis mutations cause ectopic development of multinucleate structures called nodules, which are located at the endosperm posterior pole in WT endosperm (Mansfield and Briarty, 1990; Scott et al., 1998). Accordingly, the posterior pole marker KS117 (Sørensen et al., 2001) and the gene PHERES1 (Köhler et al., 2003b) are ectopically expressed throughout fis endosperm. PHERES1 expression is directly controlled by the FIE-MEA complex (Köhler et al., 2003b). The enhancer trap line KS117 reports the expression of the actin nucleator FORMIN HOMOLOGY PROTEIN 5 (AtFH5) (Ingouff et al., 2005). We initially proposed that FIS genes repress posterior differentiation of the endosperm in the anterior domains (Guitton et al., 2004; Sørensen et al., 2001). However, the role of FIS as repressors of posterior differentiation is in apparent contradiction to overlapping WT expression patterns of FIS genes (Kinoshita et al., 1999; Luo et al., 2000; Vielle-Calzada et al., 1999) and of their target genes PHERES1 and AtFH5 (Ingouff et al., 2005; Köhler et al., 2003b). Alternatively the effect of fis mutations on endosperm development could result from temporal deregulation of posterior marker expression, leading to maintenance of their initial uniform expression pattern throughout endosperm development. In this report we provide support to this hypothesis and show that FIS PcG genes affect developmental timing of endosperm.


Impact of mutations in FIS genes on major features of endosperm development

Most fis mutations lead to collapse and death of the seed after the embryo heart stage (Chaudhury et al., 1997; Kiyosue et al., 1999). In order to detect morphological alterations in fis seeds prior to the mid-embryo heart stage we used the endosperm marker KS117 (Sørensen et al., 2001). As early as the embryo early globular stage, fis seeds with high uniform KS117 expression can be distinguished from WT seeds where KS117 expression is confined to the posterior pole (Figure 1a). We compared the developmental stages of the endosperm between the two classes of seeds in fis1/mea (n = 106), fis2 (n = 50) and fis3/fie (n = 123). Overall, we could not detect major morphological changes in endosperm development (Figure 1b,c) until the embryo mid-heart stage in WT seeds. During this early syncytial phase of development, growth and pace of nuclei divisions were similar in fis and in WT endosperm. After the embryo heart stage, the endosperm undergoes cellularization in WT seeds but not in fis seeds and endosperm proliferation and growth are more pronounced in fis seeds than in WT seeds (data not shown; Kiyosue et al., 1999). We conclude that fis mutations do not alter the basic cellular processes during endosperm development prior to cellularization in the WT.

Figure 1.

Determination of endosperm phenotype in fis seeds prior to the heart embryo stage.
(a) mGFP5 fluorescence in segregating developing seeds from a selfed fis/+; KS117/KS117 plant. A weak and localized mGFP5 activity identifies wild type (WT) seeds (arrows) and a high and uniform mGFP5 activity identifies fis seeds (arrowheads). After clearing the phenotype of the endosperm and the embryo in a WT seed (b) is identical to the phenotype of a mea/fis1 seed from the same silique (c). Bar = 250 μm for (a), 30 μm for (b) and (c).

Mutations in FIS genes perturb temporal expression of markers expressed in the endosperm

After the embryo heart stage, the cellularization-defective endosperm in fis mutants does not undergo arrest of proliferation and shows improper differentiation of the posterior pole (Guitton et al., 2004; Sørensen et al., 2001; Vinkenoog et al., 2000). In order to investigate whether other aspects of late endosperm development was impaired in the fis mutants, we used the two markers N9185 and G222 featuring an activity initiated at the time of cellularization around the embryo (Figure 2a and e respectively) and persisting throughout embryo maturation (Figure 2b and c, and f and g respectively). The expression patterns of N9185 and G222 are perturbed in fie/+ background. In fie seeds no reporter activity is detected in either marker line (Figure 2d,h). Similarly, expression of N9185 or of G222 is neither observed in mea nor in fis2 seeds (data not shown). The lack of expression was not the consequence of the absence of cellularization in fis mutants as G222 expression was not affected in spätzle endosperm defective for cellularization (Sørensen et al., 2002) (Figure S1). Thus, FIS genes are necessary for some aspects of endosperm molecular differentiation after cellularization.

Figure 2.

fis mutations prevent expression of late endosperm enhancer trap markers.
(a–c) mGFP5 expression pattern in the enhancer trap line N9185. mGFP5 activity is first detected in the endosperm surrounding the heart stage embryo (a). This specific expression remains in the endosperm around the embryo at torpedo (b) and bent cotyledon stage (c). In fie mutant endosperm with overgrown cyst (cy) and ectopic nodules (no), no expression of the N9185 marker is detected (d).
(e–g) GUS activity in line G222. A GUS staining is first evident in the endosperm region surrounding the heart stage embryo (e). The GUS activity expands in the peripheral endosperm at the torpedo embryo stage (f). At mature embryo stage, G222 expression persists in layers of endosperm cells (g).
(h) In the fie endosperm characterized by an overgrown cyst (cy) and ectopic nodules (no), GUS activity of the G222 marker line is not present. (a–d) Projections of z-series of confocal sections of GFP fluorescence and red autofluorescence. Bar = 60 μm. (e–h) Nomarski micrographs. Bar = 50 μm.

The absence of differentiation in the late fis endosperm is in agreement with a temporal extension of juvenile traits revealed by the pattern of expression of KS117 and PHERES1 (Köhler et al., 2003b; Sørensen et al., 2001). To determine whether juvenile aspects of endosperm development were affected in fis class mutants, we used enhancer trap markers with patterns specific to different phases of syncytial endosperm development. For each marker we assessed the impact of mea/fis1 and mea-6 alleles, fis2-1 and fis2-6 alleles, fie-10 and msi1-2. Syncytial endosperm developmental stages are defined according to the number of nuclei (Boisnard-Lorig et al., 2001). As mitotic divisions are nearly synchronous, each stage contains nearly twice as many nuclei as the previous stage. Syncytial endosperm development can be divided into three phases (Boisnard-Lorig et al., 2001; Guitton et al., 2004; Sørensen et al., 2002). Phase 1 corresponds to the initial three successive synchronous nuclei divisions (stages I–IV). Phase 2 corresponds to the three mitotic domains (stages V–VII) each identified by their own rate of nuclei division (Boisnard-Lorig et al., 2001). Phase 3 begins at stage VIII with nuclei migration to the posterior pole (Guitton et al., 2004). Phase 4 is initiated at stage IX by the onset of endosperm cellularization (Sørensen et al., 2002). In WT endosperm, KS22 expression marks phases 2 and 3 (Figure 3a–d). In fis/+ plants half of the seeds show a higher mGFP5 expression throughout endosperm development (Figure 3e). As these seeds show ectopic nodules and eventually collapse, we conclude that fis mutations are responsible for the temporal extension of KS22 expression. We observed similar persisting expression of mGFP5 in the endosperm of fis/+ plants in the enhancer trap lines M11 and N9319 with expression patterns mainly restricted to phases 2 and 3 and to phase 2 respectively (Figure 3f–j).

Figure 3.

Altered expression of early endosperm development markers in the fis mutants.
(a–t) Projections of z-series of confocal sections of GFP fluorescence and red autofluorescence. (a–d) mGFP5 expression pattern in the enhancer trap line KS22. mGFP5 activity is first present in all parts of the endosperm at stage V (a) and up to late stage VII (b). At stage VIII, mGFP5 fluorescence persists in the poles (c). A faint mGFP5 activity remains until stage X (d). In the fie endosperm, the uniform expression of KS22 is maintained beyond stage VIII (e, compare with d). (f–i) mGFP5 expression pattern in the enhancer trap line N9319. A mGFP5 activity is initially detected in all parts of the endosperm at stage V (f), stage VI (g) and until stage VII (h). A residual fluorescence remains in the poles at stage VIII (i). In the fie endosperm, N9319 remains uniformly expressed beyond stage VIII (j, compare with i). (k–n) mGFP5 expression pattern in the enhancer trap line KS117. The expression of mGFP5 is detected uniformly from stage III (k, l). At stage VII the fluorescence becomes confined to the posterior pole (m) and persists at least until the torpedo embryo stage (n). In the fie mutant endosperm, KS117 remains expressed uniformly from stage III beyond stage VIII (o, compare with n). (p–s) mGFP5 expression pattern in enhancer trap line N9307. The mGFP5 activity is restricted to the endosperm posterior pole (arrowhead) from stage VI (p) until stage IX (q, r). An additional mGFP5 expression is specifically detected in the cotyledons in the late embryo heart stage (s). (t) In the fie endosperm, mGFP5 activity persists at the endosperm posterior pole until the seed collapses. The mGFP5 activity of the N9307 marker is unchanged in fie embryo.
Bar = 20 μm for (a), 30 μm for (b) and (c), 40 μm for (d) and (e), 30 μm for (f) and (j), 20 μm for (k), 30 μm for (g)–(i), (l) and (m), 40 μm for (n), (o) and (s), 30 μm for (p) and (q), and 40 μm for (r) and (t).

Re-examination of the effect of fis on KS117 also revealed a temporal extension of expression. WT activity of mGFP5 in KS117 remains uniform until the end of phase 2 (Figure 3k,l) and becomes confined to the posterior pole (Figure 3m) where it persists at least until late torpedo embryo stage (Figure 3n). In fis endosperm KS117 does not undergo transition from the early expression pattern typical of phase 2 to the restricted pattern typical of phase 3 but instead becomes over-expressed uniformly until the end of endosperm development (Figure 3o).

Mutations in the FIS genes show a maternal gametophytic effect (Chaudhury et al., 1997; Grossniklaus et al., 1998; Kiyosue et al., 1999; Ohad et al., 1999; Sørensen et al., 2001). This means that fis mutations cause a phenotype only if they are inherited maternally and crosses of fis/+ ovules with WT pollen produce 50% mutant seeds. The expression of the early enhancer trap markers M11, N9319, KS117 and KS22 is perturbed in half of the seeds of fis/+ plants homozygous for the marker when pollinated with each corresponding marker line (Sørensen et al., 2001; data not shown). This observation confirms that fis mutations have a maternal gametophytic effect on the expression of the enhancer trap markers and affect their temporal pattern of expression.

We had originally interpreted the ectopic expression of KS117 in fis endosperm as an evidence for abnormal development of the posterior pole in fis mutants (Sørensen et al., 2001). In order to test whether features of the posterior endosperm properly differentiate in fis mutants we used the posterior pole marker line N9307. In this line mGFP5 fluorescence is initially detected at the posterior pole during phase 2 (Figure 3p). The expression of N9307 remains confined to the cyst (Figure 3q,r). Moreover, N9307 expression becomes detected in the embryo cotyledons as early as the embryo late heart stage (Figure 3s). In fis seeds, N9307 expression in the embryo is similar to the WT expression, although embryo development is arrested at the late torpedo stage (Figure 3t). In the fis endosperm, N9307 expression persists only in the posterior pole and is not present in ectopic nodules (Figure 3t). Hence fis mutations do not primarily prevent differentiation of a posterior identity. In the fis endosperm all syncytial phase markers remain expressed throughout development with a pattern similar to that observed during their earliest phase of expression. We hypothesize that fis mutations cause the temporal extension of several developmental features typical of phase 2.

The transition from a uniform to a posterior pattern of expression of two FIS reporter gene fusions is prevented in fis mutants

The expression of MEA and FIS2 is initially uniform in the syncytial endosperm and later becomes restricted to the posterior pole (Kinoshita et al., 1999; Luo et al., 2000). This transition takes place during the early syncytial stage prior to nuclei migration at the posterior pole (phase 3) and we examined whether it is affected by fis mutations (Table 1, Figure 4). In the WT, MEA:GUS expression is initially uniform in the endosperm (Figure 4a) but becomes restricted to the posterior pole during phase 2 (Figure 4b). After the end of phase 2, MEA:GUS expression is no longer detected in the endosperm (Figure 4c). In siliques of self-pollinated fie/+ plants, seeds no longer display the restriction of MEA:GUS expression to the posterior pole (Table 1, Figure 4d). Segregating seeds with fis phenotype in msi1 plants also showed temporal extension of uniform MEA:GUS expression beyond phase 2 (data not shown). Similar observations were made for the effect of fis mutations on the pattern of expression of FIS2:GUS. As previously reported (Luo et al., 2000), FIS2:GUS is expressed immediately after fertilization and up to the end of phase 1 in all parts of the endosperm (Figure 4e). During phase 2, GUS activity becomes restricted to the large nuclei of the posterior pole (Figure 4f). After endosperm cellularization, expression of FIS2:GUS remains localized to the cyst at least until torpedo embryo stage (Figure 4g). In contrast to WT endosperm, the transition from a uniform pattern to a posterior pattern does not occur in a fie/+ background and half of the seeds in a segregating population from selfed plants show uniform expression of FIS2:GUS after stage VIII (Table 1, Figure 4h). Pollination of fie/+; FIS2:GUS/FIS2:GUS plants with WT pollen produced half of seeds in a segregating population with uniform expression of FIS2:GUS which shows that the effect of fie mutation on FIS2:GUS expression is gametophytic maternal (Table 1). The fis2-1 mutation also perturbs the transition between the early to the late pattern of expression of FIS2:GUS but with lower penetrance than that observed in the fie/+ background (Table 1). In conclusion, fie and fis2 mutations cause temporal extension of the early uniform pattern of MEA and FIS2 expression.

Table 1.  Effect of the fie and fis2 mutations on the pattern of expression of MEA:GUS and FIS2:GUS
StageWT MEA:GUSfie/+; MEA:GUSWT FIS2:GUSfie/+; FIS2:GUSfie/+; FIS2:GUS X WT pollenfis2/+; FIS2:GUS
  1. fie affects the transition from the uniform expression of MEA:GUS to an expression restricted to the posterior pole. Similarly, fie and fis2 prevent the transition from a uniform pattern of expression of FIS2:GUS to a pattern restricted to the posterior pole. Percentage (%) of seeds with GUS activity detected uniformly or restricted to the posterior pole in the endosperm from self-pollinated fis/+; MEA:GUS/MEA:GUS or FIS2:GUS/FIS2:GUS plants or from fie/+; FIS2:GUS/FIS2:GUS plants pollinated with wild type (WT) pollen.

V100% Uniform (n = 97)100% Uniform (n = 97)100% Uniform (n = 225)100% Uniform (n = 97)100% Uniform (n = 97)100% Uniform (n = 97)
VI88% Restricted (n = 78)64% Restricted (n = 44)26% Restricted (n = 66)24% Restricted (n = 78)n.d.n.d.
VII38% Absent (n = 44)n.d.89% Restricted (n = 163)57% Restricted (n = 143)53% Restricted (n = 163)63% Restricted (n = 83)
VIII100% Absent (n = 121)62% Absent (n = 44)100% Restricted (n = 95)54% Restricted (n = 565)n.d.60% Restricted (n = 65)
Figure 4.

Altered activity of MEA:GUS and FIS2:GUS reporters in fie endosperm.
(a–c) MEA:GUS activity during endosperm development. (a) Uniform GUS activity seen at stage V. (b) Restricted GUS staining to the posterior pole at stage VI. (c) Absence of GUS activity after stage VIII. In the fie endosperm featuring an overgrown cyst (cy) and ectopic nodules (no) MEA:GUS activity remains uniform at stage VIII (d, compare with c).
(e–g) FIS2:GUS activity during endosperm development. (e) Uniform GUS activity at stage V (28 nuclei). (f) GUS staining is restricted to the posterior pole from stage VII and remains in the endosperm posterior cyst at the torpedo embryo stage (g). In the fie endosperm featuring an overgrown cyst (cy) and ectopic nodules (no), GUS staining is uniform at stage VIII (h, compare with g).
Bar = 25 μm for (a) and (e), 30 μm for (b) and (f), 40 μm for (c) and (g), and 50 μm for (d) and (h).

Impact of fis mutations on the establishment of mitotic domains

In the WT endosperm the transition from phase 1 to 2 is marked by the establishment of mitotic domains. As fis mutations perturb the expression of FIS genes during phase 2 we investigated whether fis mutations affect the establishment of mitotic domains. In the WT, after the fourth (Figure 5a) or the fifth cycle of synchronous division (Figure 5c) two or four nuclei at the posterior pole no longer divide while the remaining nuclei divide synchronously leading, at stage VI, to three types of endosperm that contain 26 (Figure 5b), 28 (Figure 4e) or 30 (Figure 5d) nuclei including 2, 4 or 2 large nuclei at the posterior pole respectively.

Figure 5.

Absence of mitotic domains in fis mutants.
(a–f) Micrographs of endosperm nuclei labelled by FIS2:GUS construct. (a–d) Successive stages of divisions in the wild type (WT) after the eight-nuclei stage. During the fourth round of divisions two nuclei at the posterior pole endoreduplicate while other nuclei divide leading to a 14-nuclei stage V (a) followed by a 26-nuclei stage VI (b). Alternatively, the fourth syncytial division is synchronous and generates 16-nuclei stage V endosperm (c). In the latter case two nuclei at the posterior pole endoreduplicate and the fifth round of division leads to the 30-nuclei stage (d). A clear difference in nuclei size is seen between nuclei at the posterior pole (arrowhead) and the other endosperm nuclei (b, c). In contrast, in the fis2 mutant, divisions remain strictly synchronous and generate stages with 16 nuclei (e) and with 32 nuclei (f). Bars represent 30 μm.
(g) The FIS2:GUS construct labelling the endosperm nuclei was used to determine the number of nuclei in syncytial endosperm at stages V and VI in seed populations of WT and selfed fis/+ plants. At WT stage V, the endosperm contains 14 nuclei (light blue) or 16 nuclei (light green). At the following mitosis (stage VI), the minor population of 14 nuclei endosperm, gives rise to a 26-nuclei endosperm population (dark green). Two subpopulations are derived from the major population of 16-nuclei endosperm, one with 28 nuclei (turquoise) (see Figure 4e) and the other with 30 nuclei (blue). All subpopulations typical of WT endosperm at stages V and VI are observed in fis mutants (mea, fis2 and fie). However, at stage VI, a unique population of seeds is detected with endosperm containing 32 nuclei (red). The number of analysed seeds (n) at stage V or stage VI of endosperm development is indicated at the top of each bar.

In contrast to WT, selfed fis/+ plants produce a quarter of seeds with 32 nuclei in the endosperm at stage VI (Figure 5f,g). The occurrence of endosperm with 32 nuclei of equal size indicates that the posterior mitotic domain is not established in the fis endosperm. In order to establish whether the perturbation of the mitotic domains was under maternal control, we fertilized fie/+; FIS2:GUS/FIS2:GUS plants with WT pollen. We observed 15% endosperm with 32 nuclei at stage VI (n = 70; 14% with 26 nuclei, 42% with 28 nuclei and 29% with 30 nuclei) (Figure S2). Hence, the fie/fis3 mutation has a maternal gametophytic effect on mitotic domains. The maternal effect can be explained by the paternal imprinting of FIE expression during phases 1 and 2 (Yadegari et al., 2000). We also observed the absence of posterior mitotic domain in the msi1-2 endosperm as msi1-2/+ plants pollinated with WT produce 41% of seeds with 32 nuclei in the endosperm at stage VI (n = 151). In conclusion, fis mutations prevent mitotic domain formation. The early uniform pattern of nuclei division typical of phase 1 is reiterated throughout endosperm development in parallel to the extension of the early pattern of expression of several markers of phases 1 and 2.


Mutations in the FIS PcG genes do not alter growth, cell proliferation and spatial patterning during early endosperm development

The major cellular defects reported so far for endosperm development in fis seeds were an absence of cellularization and a continued syncytial phase with an increased proliferation by the time WT endosperm cellularizes (Chaudhury et al., 1997; Kiyosue et al., 1999; Ohad et al., 1999). We further show that prior to endosperm cellularization, we could not detect changes in the overall rates of syncytial nuclei division and growth. Moreover, as in WT endosperm, the posterior pole of the fis endosperm differentiates a cyst marked by expression of the posterior marker N9307 and by production of Zn-phytate crystals (Otegui et al., 2002). These observations do not support our previous hypothesis of an overall perturbation of the antero-posterior pattern of endosperm development by fis mutations (Sørensen et al., 2001). In conclusion, during syncytial development, fis mutations do not appear to impair basic cellular processes such as growth and proliferation nor do they directly prevent antero-posterior patterning in the endosperm. Increased proliferation in the fis endosperm after the eight cycles of nuclei division likely results from the absence of cellularization as the pace of the cell cycle is higher in a syncytial state than in a cellular state as shown in Drosophila embryos (Edgar et al., 1994) and in Arabidopsis endosperm (Boisnard-Lorig et al., 2001). Alternatively, mutations in the FIS PcG genes may directly perturb cell proliferation as demonstrated for PcG genes in animals (Jacobs and van Lohuizen, 2002; Orlando, 2003) and suggested in Arabidopsis for the PcG gene CURLY LEAF (CLF) (Kim et al., 1998; Serrano-Cartagena et al., 2000). Moreover, FIE interacts with the Arabidopsis Rb homologue (Mosquna et al., 2004) and MSI1 is putatively able to interact with Rb in planta (Ach et al., 1997). As Rb controls the transition between the G1 and the S phase of the cell cycle (Ach et al., 1997), the FIS PcG complex could directly regulate the cell cycle. Such a role could also account for the absence of mitotic domains in the fis mutant endosperm.

The fis mutations affect temporal patterning in syncytial endosperm

Endosperm syncytial development is subdivided into three phases (Figure 6a). After phase 1 (consisting of three synchronous nuclei divisions), mitotic domains are defined by stage V (phase 2). Phase 3 starts with the onset of nuclei migration towards the posterior pole at stage VIII (Guitton et al., 2004). At stage IX cellularization marks the end of the syncytial phase (phase 4). Mutations in the FIS genes cause a general temporal extension of patterns of expression of seven markers typical of phase 2, including MEA and FIS2 reporters (Figure 6b). A similar change in pattern of expression has been reported for the MADS box gene PHERES1 in mea and fie endosperm (Köhler et al., 2003b). We thus conclude that fis mutations affect the transition from phase 2 to phase 3. Accordingly, nuclei migrations typical of phase 3 in the WT are not observed in the fis endosperm (Guitton et al., 2004). In addition, the molecular markers N9185 and G222 of phase 4 are never expressed in the fis endosperm consistent with a temporal prolongation of features of the syncytial endosperm. Similarly, the mitotic domains established during phase 2 in the WT are absent in the fis endosperm. This defect might be interpreted as the temporal extension of features typical of phase 1 when mitotic domains are still undefined. However, endosperm development in fis mutants is not arrested at the transition between phase 1 and phase 2. Such an arrest would block cell proliferation and growth at stage V. In such a scenario, markers KS22 and N9319 initially expressed in phase 2 in WT endosperm would not be detected in the fis endosperm, which is not the case. Similarly, endosperm development in fis mutants is not arrested at the transition between phase 2 and phase 3. This would result in a fis endosperm containing ca. 50 nuclei and no cyst. We rather observed the opposite phenotype with unaffected pace of nuclei division until the eighth mitosis and further over proliferation of the fis endosperm. Moreover, several aspects of antero-posterior patterning including N9307 expression take place in the fis seed with timing similar to WT seeds (Otegui et al., 2002). Similarly, some late markers of cellularized endosperm are expressed in the fis endosperm (Ohad et al., 1996). In summary, fis mutations can be defined as heterochronic as they alter the relative sequence of events during endosperm development.

Figure 6.

Altered developmental timing of endosperm development in fis mutants.
(a) Expression pattern of reporter genes in marker lines during endosperm in the wild type (WT).
Endosperm development can be divided into four phases and comprises 12 stages (roman numbers) defined by successive pseudo-synchronous mitoses (Boisnard-Lorig et al., 2001; Guitton et al., 2004; Sørensen et al., 2002; this work). The first three successive synchronous nuclei divisions (stages I–IV) correspond to phase 1. Phase 2 (stages V–VII) consists of the three mitotic domains (symbolized by black dots of different sizes in the endosperm). Phase 3 begins when migration of nuclei to the posterior nodules and cyst is initiated by stage VIII onwards (Guitton et al., 2004). Phase 4 marks the onset of endosperm cellularization at stage IX (Sørensen et al., 2002).
The expression pattern of each reporter gene is determined for marker lines during endosperm development in the WT (a) and the fis group mutants (b). Four classes of endosperm marker lines are defined based on the reporter gene activity in the endosperm: two late markers (brown) (N9185 and G222), three early markers (pink) (KS22, N9319 and M11), two markers of the endosperm posterior pole (green) (KS117 and N9307) and two promoter GUS fusion constructs that report the activity of MEA:GUS and FIS2:GUS (blue). A dotted pattern symbolizes a reporter gene expression restricted to the posterior pole.
(b) Expression pattern of reporter gene in marker lines during fis endosperm development.
All marker lines are introduced in the fis/+ background and the reporter gene activity is analysed in fis seeds and compared with the WT seeds from the same silique. When the marker line is analysed only in the fie mutant, it is indicated by a star (*). In the fis endosperm, late markers are never expressed (transparent rectangles with a cross-bar) whereas the uniform expression of the early marker lines (pink) and the KS117 line are perpetuated until the fis seed collapses. This temporal shift of expression pattern is observed for MEA:GUS and FIS2:GUS that remain uniformly expressed during late endosperm development with a pattern typical of the early syncytial phase in the WT. The absence of the posterior pole expression of MEA:GUS, FIS2:GUS and KS117 does not result from patterning defect as marker N9307 is still expressed in the posterior pole of the fis endosperm. Perpetuated juvenile expression pattern of endosperm markers in fis mutants is correlated with the persistence of cytological features such as syncytial proliferation typical of early endosperm. Moreover, mitotic domains are not defined at the transition between stages V and VI. This absence constitutes the earliest defect observed in the fis endosperm. This is parallel to the lack of restriction of MEA:GUS and FIS2:GUS expression to the posterior pole. Thus, we propose that fis mutations perpetuate several molecular and cellular features of the early syncytial phase to later phases of development. As a result, juvenile characters persist and certain late molecular and cytological markers are missing while others are expressed. fis mutations cause heterochronic endosperm development.

A conserved role for plant PcG proteins in the temporal control of developmental phases

The recent demonstration of conservation in Drosophila, mammals and plants of the PRC2 type of PcG complex containing homologues of E[z], Su(z)12, Esc and P55 suggests a conservation of its function in development (Chanvivattana et al., 2004; Guitton et al., 2004; Köhler et al., 2003a; Otte and Kwaks, 2003). This conservation is further supported by identification of homeotic genes as PcG target genes in animals (Francis and Kingston, 2001) and in plants (Goodrich et al., 1997; Katz et al., 2004; Kinoshita et al., 2001; Köhler et al., 2003b; Moon et al., 2003).

We report in this study that the Arabidopsis FIS complex controls developmental timing of endosperm. Together with our results, recent studies suggest that PcG complexes control timing of development in plants. During its life cycle a plant undergoes a series of transitions from the embryonic stage to the vegetative non-flowering stage and later to the flowering stage. Each transition is controlled by a distinct PcG complex (Hsieh et al., 2003; Wagner, 2003). Mutations in EMBRYONIC FLOWER 2 (EMF2), a PcG gene encoding a protein homologous to FIS2, lead to production of flowers by embryos (Yoshida et al., 2001). This phenotype likely results from a bypass of the vegetative growth phase and an immediate transition from the embryonic to the flowering phase. The interpretation of this phenotype is consistent with the definition of a heterochronic development. EMF2 acts in a PcG complex involving FIE (Chanvivattana et al., 2004). The transition to flowering is controlled by the PcG gene CURLY LEAF (CLF) belonging to the MEA family. Mutations in CLF cause early flowering (Goodrich et al., 1997). CLF and its homologue SWINGER interact with FIE and EMF2 (Chanvivattana et al., 2004) and maintain the repression of several homeobox genes during vegetative development (Katz et al., 2004). The similarities between phenotypes in clf plants and plants with reduced level of MSI1 transcripts (Hennig et al., 2003) suggest that the complex also contains MSI1 and controls the transition from the vegetative to the flowering phase. A similar PcG complex containing VERNALISATION 2 (VRN2), in place of EMF2, records exposure to cold (vernalization) (Chanvivattana et al., 2004). The duration of the initial vegetative development depends on the duration of vernalization. Increasing periods of vernalization favour precocious flowering in the WT but not in the mutant vrn2 (Gendall et al., 2001). The memory of vernalization is apparently mediated by methylation on K27 and K9 residues of histone H3 (Bastow et al., 2004). A similar histone methyltransferase activity has been originally ascribed to the PRC2 complex of Drosophila (Czermin et al., 2002; Müller et al., 2002). This suggests a potential conservation of the enzymatic properties of PcG complexes between animal and plants. In conclusion, we propose that PcG group proteins form modular transcriptional repressing complexes that regulate the timing of successive developmental phases required to fulfil the entire plant life cycle.

Experimental procedures

Plant strains

Lines KS117, KS22, M11, N9185, N9307 and N9319 (C24 accession) were identified after a screen in the Jim Haseloff's enhancer trap mGFP5 line collection (Haseloff, 1999; http://www.plantsci.cam.ac.uk). The enhancer trap GUS line G222 (Ler accession) was a generous gift from G. Jürgens and was isolated by G. Martin in a promoter trap line collection.

The fis alleles fis1/mea, fis2-3 and fis3/fie (Ler accession) and the transgenic lines (C24 accession) that contain the promoter MEA: GUS or promoter FIS2:GUS fusion constructs were kindly provided by A. Chaudhury (Canberra, ACT, Australia; Chaudhury et al., 1997; Luo et al., 2000). The mea-6, fis2-6 and the multicopy suppressor of ira 1-2 (msi1-2) alleles (C24 accession) used in this study originate from a screen reported by Guitton et al. (2004). The mutant spätzle (allele DRU 42, WS accession) originates from a screen of the Versailles collection (Sørensen et al., 2002). Plants were grown as reported previously (Garcia et al., 2003).

The following combinations of markers (homozygous) and fis/+ were obtained. fis1/mea, fis2-3 and fis3/fie were combined with all the markers used in this study. In order to ensure that the genetic background combination between Ler and C24 did not interfere, we crossed as well mea-6, fis2-6 with KS22, KS117 and FIS2:GUS. When compared with each other, alleles of mea and fis2 gave similar results (data not shown).

msi1-2 was combined with KS117, KS22, M11, N9319, MEA:GUS and FIS2:GUS. For each observation at least 100 seeds were observed for each developmental stage.

In order to test the maternal gametophytic effect of fis mutations on the expression of markers, lines homozygous for the markers and heterozygous for fis were emasculated prior to anthesis and pollinated after 1 day with a homozygous marker line in the FIS/FIS background. We tested at least two crosses for each combination.

Microscopic analysis of the phenotype of fis developing seeds

We used fis/+; KS117/KS117 plants. Developing seeds were isolated from individual siliques at stages of embryo development ranging from the early globular stage to the early heart stage. Each population of seeds was mounted in Hoyer's medium (Boisnard-Lorig et al., 2001) and fluorescence associated with the KS117 marker was readily observed with a Leica MZFLIII stereomicroscope coupled to a digital camera DC300F (Leica Microsystems, Heerbrug, Germany). Images were processed with the software FW4000 (Leica). After clearing, the phenotype was determined microscopically using differential interference contrast optics (Optiphot; Nikon, Tokyo, Japan) for each seed and linked to the associated genotype determined by the expression of KS117.

Analysis of reporter gene activity in the developing endosperm

GUS assay was performed as reported previously (Boisnard-Lorig et al., 2001). GFP fluorescence was imaged using laser scanning confocal microscopy (Zeiss LSM-510, Jena, Germany) with selective settings for GFP detection (excitation 488 nm and emission 510–550 nm) and non-specific settings for autofluorescence detection (excitation 543 nm and emission 560 nm) (Sørensen et al., 2001). Digital image processing was performed with Photoshop 5.5 and Illustrator 9.0 (Adobe Systems, San Jose, CA, USA).


We are grateful to Abed Chaudhury (Canberra, ACT) for the MEA:GUS, FIS2:GUS plants and the fis mutants, to Gerd Jürgens (University of Tübingen) for the G222 line, to Loïc Lepiniec (INRA, Versailles) for the DRU42 mutant and to Justin Goodrich (Institute of Cell of Molecular Biology, Edinburgh) for the N9185 marker line. We thank Jonathan FitzGerald, Anne-Elisabeth Guitton, Damien Garcia and Mathieu Gourgues for critical reading of the manuscript.

This work was supported by the French Génoplante II program (M.I.) and by the EMBO Young Investigator Program (F.B.).

Supplementary Material

The following material is available from http://www.blackwellpublishing.com/products/journals/suppmat/TPJ/TPJ2404/TPJ2404sm.htm

Figure S1. GUS activity of the marker line G222 restricted to the peripheral endosperm observed after cellularization in the WT is unchanged in the cellularization defective endosperm mutant spätzle. Bar = 50 μm.

Figure S2. Maternal effect of the fie mutation on the mitotic domain in the endosperm.
(a) Sixteen and (b) 32 nuclei endosperm in fie/+; FIS2:GUS/FIS2:GUS plants pollinated with WT plants. GUS staining in endosperm nuclei results from FIS2:GUS activity. Bar = 30 μm.