Arabidopsis MSI1 is required for epigenetic maintenance of reproductive development

Regulatory decisions that are made during development must be maintained correctly during the life of the organism. In multicellular eukaryotes,developmental programs are controlled largely at the level of transcriptional regulation and modulation of chromatin structure. Chromatin packages DNA into higher-order structures, but also maintains accessibility to DNA for regulatory DNA-binding proteins. Transcriptional activators also facilitate DNA accessibility by increasing chromatin fluidity. In contrast,transcriptional repressors establish and maintain chromatin structures that repress the activity of individual genetic loci or entire chromosome regions. During the last few years considerable insight has been gained into the biochemistry of chromatin metabolism and its role in development of plants and animals (for reviews see Habu et al.,2001; Muller and Leutz,2001; Verbsky and Richards,2001; Narlikar et al.,2002; Reyes et al.,2002).

Unlike in most animals, environmental factors strongly affect development of plants. Although the differences in developmental strategies are striking,it now emerges that plants utilise molecular mechanisms similar to those that control animal development (Habu et al.,2001; Verbsky and Richards,2001; Meyerowitz,2002). For example, chromatin-modifying proteins such as histone deacetylases, several polycomb-group proteins, like-heterochromatin protein 1,and chromatin assembly factor 1 (CAF-1) are also required for normal plant development (Goodrich et al.,1997; Chaudhury et al.,1997; Grossniklaus et al.,1998; Wu et al.,2000; Tian and Chen,2001; Gaudin et al.,2001; Kaya et al.,2001). Proteins similar to yeast MSI1 (multicopy suppressor of ira1) and to the mammalian retinoblastoma-associated proteins RbAp46/48 are WD40 repeat proteins encoded by small multigene families in most eukaryotes. These conserved proteins are components of complexes involved in chromatin metabolism, including CAF-1, pRb, histone acetyl transferases and histone deacetylases (Qian et al.,1993; Parthun et al.,1996; Verreault et al.,1996; Taunton et al.,1996). CAF-1, a trimeric complex that facilitates deposition of nucleosomes on newly synthesised DNA, has been identified in Arabidopsis, mammals, yeast, flies and Xenopus(Smith and Stillman, 1989; Bulger et al., 1995; Kaufman et al., 1997; Quivy et al., 2001; Kaya et al., 2001). Biochemical analysis has shown that, similar to their animal counterparts,plant MSI1-like proteins are found in CAF-1 and can bind to plant retinoblastoma-related proteins and histones(Ach et al., 1997; Kaya et al.,2001; Rossi et al., 2001). The biochemical data have not provided substantial insights into the biological function of MSI1-like proteins. Loss of MSI1 function in yeast does not produce apparent phenotypic alterations under optimal growth conditions. Detailed molecular analysis of msi1 mutants revealed, however, that silencing of telomeric regions and mating type loci is decreased and cells are more sensitive to UV light (Kaufman et al., 1997). The functional analysis of MSI1-like proteins has proved more difficult in multicellular eukaryotes. The mutation lin53in Caenorhabditis elegans revealed that LIN53 encodes a MSI1-like protein, which interacts with the retinoblastoma-like LIN35 protein and is required during vulva formation (Lu and Horvitz, 1998). RNAi-mediated interference with LIN53expression caused embryo lethality, suggesting that the gene product has an essential function during development. Additional mutants that affect MSI1-like proteins in other organisms could therefore provide important new insights into the function of this family of WD40 proteins in yeast and multicellular eukaryotes.

Four MSI1-like proteins have been identified in Arabidopsis(AtMSI1-4), but only AtMSI1 is most similar to RbAp48(Ach et al., 1997; Kenzior and Folk, 1998). Similar to MSI1-like proteins from mammals, Drosophila and yeast,AtMSI1 and its tomato homologue LeMSI1 are predominantly localised to the nucleus (Quian et al., 1993; Tyler et al.,1996; Ach et al.,1997; Zhu et al.,2000; Bouché et al.,2002). A direct biological role, however, has not been established for either mammalian or plant MSI1-like proteins. Here we show that reducing AtMSI1 expression in Arabidopsis disrupts several aspects of the developmental program. Our results suggest that AtMSI1 function is required during vegetative and reproductive growth and for maintenance of correct homeotic gene expression.

To construct plants in which AtMSI1 is ectopically expressed, the complete coding sequence was fused to the cauliflower mosaic virus (CaMV) 35S promoter into the binary vector pSLK7292. Tri-parental mating of Escherichia coli DH5a cells transformed with the binary plasmids, Agrobacterium tumefaciens (strain GV3101/mp90) and E. colihelper strain (pRK2013) (Koncz and Schell,1986) was performed to transfer the plasmid into Agrobacterium. Arabidopsis thaliana plants (accession Columbia) were transformed by floral dip (Clough and Bent,1998). T₁ seeds were plated on Murashige and Skoog (MS)medium containing 50 μg/ml kanamycin, and after approximately two weeks the transformed seedlings were transferred to soil. Subsequent generations were grown in the greenhouse under long days (16 hours of light) at 23°C(±1.5°C). Alternatively, Conviron growth chambers with mixed cold fluorescent and incandescent light (230 μmol/m²/second,23°C) were used to raise the plants. The AGAMOUS::GUS reporter line was kindly provided by Drs Nathaniel P. Hawker and John Bowman (UC Davis,CA). It is similar to the promoter-intron GUS construct described by Sieburth and Meyerowitz (Sieburth and Meyerowitz,1997) and contains all relevant intragenic sequences. The GUS expression in this line faithfully resembles spatial expression patterns of the endogenous AG gene (J. Bowman, personal communication). Seeds of Columbia, Landsberg erecta and Enkheim wild-type accession and of fas1-1 and fas2-1(Reinholz, 1966; Leyser and Furner, 1992)mutants were obtained from the Nottingham Arabidopsis Stock Centre(NASC On-Line Catalogue). fas1-1 fas2-1 double mutants were generated by crossing.

Protein extracts were prepared from Arabidopsis by grinding fresh or frozen tissue in extraction buffer (100 mM Tris, pH 7.5, 500 mM NaCl, 5 mM EDTA, 10 mM EGTA, 10% sucrose, 40 mM b-mercaptoethanol, 0.5 mg/ml Pefablock SC(Roche, Rotkreuz, Switzerland), 1 μg/ml pepstatin, 0.5 μg/ml leupeptin,40 μg/ml bestatin. Homogenates were centrifuged for 5 minutes at 14,000 g and 4°C. SDS-PAGE, protein blotting and detection were performed as described previously (Ach et al., 1997). Chemiluminescent detection was performed with ECL(Amersham, Uppsala, Sweden) according to manufacturer's instructions.

For coupled in vitro transcription and translation reactions, cDNAs of AtMSI1-5 were cloned in frame with the HA-tag in pGADT7 (Clonetech,Palo Alto, CA). Reactions were performed using a Promega TnT wheat germ extract system (Promega, Madison, WI) according to manufacturer's instructions.

RNA was extracted from leaves of 4-week-old plants using Trizol(Invitrogen, Carlsbad, CA) according to manufacturer's instructions. Fractionation of RNA on an agarose gel, transfer to nylon membranes,hybridisation with a random-primed ³²P-labelled probe and detection were performed as described previously (Ach et al., 1997). For RT-PCR analysis, 1 μg total RNA was treated with DNase I. Half of the DNA-free RNA (0.5 μg) was reverse-transcribed using an oligo(dT) primer and MMLV reverse transcriptase (Clonetech, Palo Alto, CA), while the remaining RNA was incubated without reverse transcriptase. Aliquots of the generated cDNA, which equalled 50 ng total RNA,were used as template for PCR with gene specific primers(Table 1).

Flowers or inflorescences were fixed in FAA (3.7% formalin, 5% acetic acid,50% ethanol) overnight at 4°C, and embedded in Technovit 7100 resin(Kulzer, Wahrheim, Germany). Sections of 5 μm were stained with Toluidine Blue and observed using an Axioplan 2 microscope (Zeiss, Jena, Germany). Tissues were cleared with chloralhydrate after fixation in ethanol:acetic acid(9:1) and observed under differential interference contrast (DIC) optics. Images were recorded with an Axiocam HRC CCD camera (Zeiss, Jena, Germany) and edited with ZeissVision software. For GUS expression analysis, flowers were fixed for 1 hour at 4°C in 90% acetone and washed three times with 50 mM phosphate buffer (pH 7.0) before overnight incubation at 37°C in reaction buffer (0.19 mM 5-bromo,4-chloro,3-indolyl-D-glucuronide, 10 mM EDTA, 0.1%Triton X-100, 0.5 mM potassium ferrocyanide, 0.5 mM potassium ferricyanide, 50 mM phosphate buffer, pH 7.2). Tissue was cleared in ethanol, mounted in 50%glycerol and observed with a Zeiss Axioplan microscope.

Samples were prepared according to the method of Ross et al.(Ross et al., 1996). Briefly,inflorescences were fixed in ethanol:acetic acid (3:1), washed in water and incubated for 2 hours at 37°C with 0.3% (w/v) each of cellulase,pectolyase and cytohelicase in 10 mM sodium citrate (pH 4.5) to remove cell walls. Nuclei from petals were spread on microscopic slides as described, and stained with 4′,6′-diamidino-2-phenylindole (DAPI). The fluorescence patterns were examined with a Zeiss Axioplan microscope, and images were recorded with an MagnaFire® CCD camera (Optronics, Goleta,CA). Digital images were quantified using ImageJ 1.27Z (W. Rasband, NIH, USA, http://rsb.info.nih.gov/ij/). Total fluorescence was determined for at least 40 representative nuclei. Regions corresponding to chromocentres were selected manually and their emitted fluorescence was quantified.

Sequences were aligned using CLUSTAL_X(Thompson et al., 1997) with default settings. Using the neighbour-joining algorithm implemented in CLUSTAL_X, an N-J tree was constructed that was corrected for multiple substitutions. The tree was bootstrapped and a graph of the tree was displayed using TreeView software (Page,1996).

Four MSI1-like genes, AtMSI1-4, were previously reported from Arabidopsis (Ach et al.,1997; Kenzior and Folk,1998). After analysis of the complete Arabidopsis genome(The Arabidopsis Genome Initiative,2000), we identified a fifth gene (At4g29730) encoding a protein similar to AtMSI1 that we termed AtMSI5. Results of RT-PCR analysis(see below) confirmed that AtMSI5 is expressed.

Phylogenetic analysis of selected MSI1-like proteins from plants, yeast and animals demonstrated that diversification of these proteins occurred independently several times in evolution(Fig. 1). We were unable to identify a clade containing MSI1-like proteins that could be used to predict a specific biochemical function. When proteins from chicken and Xenopuswere included in the analysis, we obtained similar results and found that the additional sequences grouped more closely with the human proteins RbAp46 and RbAp48, and Drosophila p55. In plants, divergence of MSI1-like proteins most likely occurred before the monocot and dicot split, because AtMSI1, AtMSI4 and AtMSI5 have close relatives in the monocot corn (Zea mays). AtMSI2/AtMSI3 and AtMSI4/AtMSI5 form pairs of very similar proteins (similarity of 90% and 83%, respectively), suggesting that the proteins are functionally redundant. Since AtMSI1 is sufficiently diverged from the other proteins, we expected that altering expression of AtMSI1 would likely provide new insights into the function of this WD40 protein.

We analysed the expression patterns of AtMSI1 to determine if the gene is differentially expressed. RNA blot analysis revealed that AtMSI1 transcripts are present in all analysed tissues, including etiolated and green seedlings, roots, mature vegetative leaves, cauline leaves, stems, flowers and siliques (Fig. 2A). To correlate the transcript profile with AtMSI1 protein levels, we used an affinity-purified anti-LeMSI1 antiserum(Ach et al., 1997). This antibody is specific because it cross-reacts only with the HA-tagged AtMSI1 fusion protein, but not with the HA-AtMSI2-5 fusion proteins that were synthesised in a coupled in vitro transcription/translation system(Fig. 2B). A control protein blot probed with the anti-HA antibody showed that all five AtMSI proteins were produced in nearly similar amounts (Fig. 2B, upper panel). Protein extracts from several Arabidopsis organs and tissues were subsequently separated by SDS-PAGE and analysed by immunoblotting with the anti-LeMSI1 antibody. AtMSI1 protein was clearly detected in seedlings, roots, young and mature leaves,stems, flowers and siliques and the highest accumulation was found in flowers(Fig. 2C).

RNAi-mediated interference with LIN53 expression in C. elegans caused embryo lethality (Lu and Horvitz, 1998), suggesting that the MSI1-like protein has an essential function during development. Ectopic expression of proteins also often provides new insights into their function. In plants, the activity of a transgene may also induce co-suppression, which results in the inactivation of both transgene and endogenous gene. We therefore transformed Arabidopsis plants with a construct containing the complete AtMSI1 coding sequence in the sense orientation under control of the constitutive 35S promoter. The transformation produced 70 kanamycin-resistant AtMSI1-OE plants, of which 11 T₁ plants accumulated up to 50-fold higher levels of AtMSI1 (data not shown). Two independently transformed T₁ lines, 1OEa3 and 1OEc2, were characterised in detail. All T₁ plants appeared normal and showed no discernible morphological changes, suggesting that increasing AtMSI1 levels has no adverse effect on Arabidopsis development.

Among the progeny of ten of the original 70 T₁AtMSI1-OE plants, including 1OEa3 and 1OEc2, several plants developed a severely stunted phenotype and were sterile. To determine the level of AtMSI1 protein accumulation in this class of AtMSI1-OE plants,protein extracts were prepared from rosette leaves of T3 siblings and subjected to gel blot analysis. Fig. 2D shows results from six 1OEa3 plants as an example. In plants that had stunted growth, AtMSI1 protein levels were reduced to less than 10%of control plants (Fig. 2D,lanes 3, 4 and 6). Siblings that appeared normal accumulated either wild-type(lane 2) or strongly increased AtMSI1 levels (lanes 7 and 8). Similar results were obtained for 1OEc2 T3 siblings (data not shown). The reduction of AtMSI1 levels in these plants was most likely the result of co-suppression of transgene and endogenous gene expression(Matzke and Matzke, 1995). Co-suppression of AtMSI1 was gene specific, as would be expected based on the limited amino acid sequence homology with AtMSI2-5. RT-PCR with cDNA-specific primers demonstrated that mRNA levels of AtMSI2-5 were not decreased in the AtMSI1-OE co-suppressed (subsequently termed AtMSI1-CS) plants (Fig. 2E). The strong AtMSI1 co-suppression phenotype was recovered in 10-30% of the progeny of at least 3 independently transformed plants, and AtMSI1-CS plants from lines 1OEc2 and 1OEa3 showed a similar phenotype for all aspects that were subsequently analysed.

The rosette leaves of AtMSI1-CS plants were irregular in shape,and the normal phylotactic pattern of leaf production from the shoot apical meristem was altered (Fig. 3A,B). After transition to flowering, the primary shoot of wild-type plants elongated and formed cauline leaves, lateral branches and flowers(Fig. 3C, left). In contrast,the primary shoot of AtMSI1-CS plants arrested early in development(Fig. 3C, right and Fig. 3D). After elongating for 5 to 15 mm, some shoots formed one or two aborted floral buds, but more frequently no organs developed except some bract-like structures.

Mature AtMSI1-CS plants were bushy and stunted, suggesting that they also had reduced apical dominance(Fig. 3E). Between 1 and 2 weeks after emergence of the aborted primary shoot, secondary shoots developed from axillary meristems. In contrast to the primary shoot, these shoots produced cauline leaves, lateral branches and flowers, although flower morphology was strongly altered. The number of unopened floral buds was increased in AtMSI1-CS plants relative to control plants. Unlike wild-type Arabidopsis flowers, which consist of sepals, white petals,pollen-producing stamens, and ovule-bearing carpels that are arranged in four concentric whorls (Fig. 4A,left), the flowers of AtMSI1-CS plants were deformed. The outer whorl of sepals appeared to be separated from the petals. The carpels were strongly elongated relative to the stamens and protruded from the enclosing petals and sepals (Fig. 4A, right). The morphological alterations that were already detectable in the first flowers on the axillary shots increased in severity in flowers that developed later. While overall morphology of older flowers still resembled that of wild-type flowers (Fig. 4A,B),younger flowers often lacked petals and anthers, the organs of whorls 2 and 3. Sepals developed carpel-like characteristics, including stigmatic surfaces(Fig. 4C). All flowers that developed on the axillary inflorescences were sterile. Reciprocal crosses showed that the sterility of AtMSI1-CS plants was caused by maternal defects.

To determine the basis of female sterility, ovules were analysed in AtMSI1-CS plants (Fig. 5). Longitudinal sections of early flowers showed that floral organs (sepals, petals, stamens and carpels) initiated normally, but both petals and anthers were shorter than in wild-type flowers (also shown in Fig. 5B). Initiation of ovules on the placenta was similar to wild-type Arabidopsis with no gaps(Fig. 5A). Mature wild-type ovules have a curved funiculus, integuments that enclose the nucellus, and a clearly visible embryo sac (Fig. 5B,C,H). Longitudinal sections of AtMSI1-CS flowers of a comparable developmental stage revealed that ovule development was severely delayed and the ovules lacked embryo sacs (Fig. 5D). Ovule development appeared to be arrested at megagametogenesis (Fig. 5G). Further examination, including optical sections using DIC optics, revealed that both integuments initiated normally, but outgrowth of integuments was strongly retarded (Fig. 5E,F,I). To determine whether ovule development was arrested with the observed morphological changes or could proceed further, using DIC optics we examined ovules from gynoecia of flowers that had just shed sepals, petals and anthers. Although the inner integument had continued to grow and enclosed the nucellus,the outer integument rarely showed the well-defined asymmetric growth resulting in the anatropy of mature wild-type ovules, and embryo sacs were missing (Fig. 5K,L).

The deformed and curled leaves, as well as abnormal flower morphology, in AtMSI1-CS plants are reminiscent of mutants in which control of floral homeotic gene expression is disrupted. For example, in curly leaf (clf) and methyltransferase 1 antisense(met1) mutants floral homeotic genes become ectopically expressed in leaves. Using RT-PCR, we tested for the presence of AGAMOUS(AG), SUPERMAN (SUP), APETALA1-3(AP1-3), WUSCHEL (WUS) and LEAFY(LFY) mRNAs in RNA isolated from leaves of wild-type and AtMSI1-CS plants (Fig. 6A). As expected, all genes were expressed in wild-type flowers,and only very low mRNA levels were detected in leaves of wild-type plants. In contrast, AG and AP2 mRNAs accumulated to significantly higher levels in AtMSI1-CS leaves than in wild-type leaves,suggesting that the control of their expression was deregulated when AtMSI1 levels are decreased below a certain threshold. No increase in mRNA levels was detected for WUS, AP1, AP3, SUP and LFY(Fig. 6A and data not shown). In order to test whether aberrant expression of floral homeotic genes is restricted to leaves or also occurs in flowers, we crossed AtMSI1-CS plants with an AG::GUS reporter line. Fig. 7A shows GUS-staining patterns for a segregating wild-type plant. As for the endogenous AGgene, expression of the reporter is restricted to the inner two whorls. The intermediate AtMSI1-CS flower shown in Fig. 7B is characterised by shorter stamens and petals and enlarged sepals. In contrast to wild-type sepals, which did not contain any GUS staining, the carpel-like stigmatic structures at the tip of AtMSI1-CS sepals clearly expressed the GUS reporter gene.

AtMSI1 was previously reported to function together with FAS1 and FAS2 in the Arabidopsis CAF-1 complex(Kaya et al., 2001), and fas1 and fas2 single mutants have been described(Reinholz, 1966; Leyser and Furner, 1992). To better understand the functions of the two FASCIATA genes relative to AtMSI1, we constructed fas1 fas2 double mutants. The mutant plants were viable and displayed the same set of developmental alterations as the parental single mutants or the fas1 fas2 double mutant(Kaya et al., 2001). In contrast, none of the characteristic traits of AtMSI1-CS plants (Figs 3, 4, 5) were observed (data not shown). Similarly, no AG or AP2 transcript could be detected in leaves of fas1 and fas2 single or fas1 fas2double mutants (Fig. 6B). Together, these results demonstrate that repression of homeotic genes in leaves requires AtMSI1 but is independent of CAF-1 function.

Arabidopsis chromocentres contain highly condensed heterochromatic DNA that consists of centromeric and pericentomeric repeats and rRNA genes(Maluszynska and Heslop-Harrison,1991; Fransz et al.,2002). Formation of chromocentres requires epigenetic imprints,such as DNA methylation and histone acetylation, because reduced amounts of heterochromatin and dispersion of pericentromeric sequences away from chromocentres were observed in decreased DNA methylation 1(ddm1) and met1 mutants(Soppe et al., 2002). Because AtMSI1 most likely participates in specific chromatin-modifying complexes, it could also be required for maintenance of functional heterochromatin. We therefore investigated the structure of nuclei in AtMSI-CS plants that had strongly reduced AtMSI1 levels using nuclear spreads and compared them to wild-type nuclei. The general appearance of AtMSI-CSinterphase nuclei, including nuclear size and number of chromocentres, was not significantly different (Fig. 8). We quantified nuclear fluorescence originating from heterochromatic chromocentres and from the remaining euchromatic regions of nuclei. In wild-type interphase nuclei, euchromatin fluorescence was 2.3 times higher than heterochromatin fluorescence. In contrast, this ratio increased to 4.0 in AtMSI1-CS plants (Fig. 8B, right). These observations indicate that recruitment of chromosomal DNA into heterochromatic chromocentres is strongly reduced in the absence of AtMSI1.

MSI1-like RbAp48-related proteins have been identified as components of complexes that regulate transcription activity and chromatin structure in mammals, yeast, flies and other eukaryotes, but their functional role in the control of gene expression is not well understood(Qian et al., 1993; Parthun et al., 1996; Verreault et al., 1996; Taunton et al., 1996). In organisms that encode multiple MSI1-like genes, it is not clear how much functional overlap exists between the different MSI1-like proteins. Drosophila has only one MSI1-like protein (p55), which is a subunit of CAF-1 and is found in complexes with histone deacetylases, histone acetyl transferases, the nucleosome remodelling factor NURF and histone methyltransferases (Tyler et al.,1996; Martinez-Balbas et al.,1998; Czermin et al.,2002). Human RbAp46 and RbAp48 appear to be more specialised. RbAp46 interacts with histone acetyltransferase HAT1, and only RbAp48 was reported as a subunit of human CAF-1(Verreault et al., 1996; Verreault et al., 1998). Similarly, yeast HAT2p is specific for HAT1p, and only CAC3/MSI1 is a component of yeast CAF-1 (Verreault et al., 1996; Kaufman et al.,1997). The MSI1-like proteins that participate in CAF-1 function in different organisms (human RbAp48, Drosophila p55, yeast MSI1,AtMSI1), however, are not more closely related to each other than they are related to other MSI1-like proteins. This suggests that functional divergence of these proteins has evolved independently in yeast, mammals and plants. Amino acid sequence comparison also suggests that plant MSI1-like proteins already diverged before the evolutionary split in monocot and dicot plants. It will be interesting to establish the functional significance of multiple MSI1-like proteins in yeast, mammals and plants, as compared to Drosophila, in which p55 is found to be associated with several chromatin-modifying complexes.

Reduced levels of AtMSI1 altered cotyledon and leaf shape, affected meristem function, and reduced fertility, indicating that AtMSI1 function is required during vegetative and reproductive development. This conclusion is consistent with the presence of AtMSI1 in all tissues and organs that were analysed. Moreover, AtMSI1 was expressed most strongly in flowers, which were also most severely affected by reduced AtMSI1 levels in AtMSI1-CSplants. RT-PCR with gene-specific primers confirmed that AtMSI2-5mRNA levels were not reduced in AtMSI1-CS plants. The morphological changes are therefore strictly correlated with the changes in AtMSI1 levels,suggesting that there is only limited functional redundancy between AtMSI1 and the other four MSI1-like proteins in Arabidopsis.

A CAF-1 complex active in chromatin assembly in vitro is also present in Arabidopsis and consists of FAS1, FAS2 and AtMSI1, which are similar to CAC1, CAC2 and CAC3 (Kaya et al.,2001). FAS1 or FAS2 are both encoded by single copy genes in Arabidopsis. Mutations in these genes cause fasciation as the result of an enlarged shoot apical meristem. Fasciated plants have an altered phyllotaxy and leaf shape, reduced root growth, broadened, flattened or bifurcated stems and increased numbers of flower organs(Reinholz, 1966; Leyser and Furner, 1992). Despite strongly reduced AtMSI1 protein levels AtMSI1-CS plants showed only mild symptoms of fasciation. It is possible that the AtMSI1 levels still present in these plants are sufficient to maintain minimal CAF-1 activity for functional chromatin assembly. Alternatively, one of the other MSI1-like proteins in Arabidopsis (AtMSI2-5) can substitute for AtMSI1 function in CAF-1 activity. AtMSI1-CS plants also have several phenotypic alterations not associated with the fasciation phenotype in fas1 and fas2, and repression of AG and AP2 expression in leaves is lost when AtMSI1 is strongly reduced, but not when FAS1 and FAS2 are absent. Because Arabidopsis accessions are known to differ in genetic potential and developmental programs(Alonso-Blanco and Koornneef,2000), both the fas1-1 and the fas2-1 allele were back-crossed into the Columbia accession. However, the different phenotypes of AtMSI1-CS and fasciata mutants remained independent of the genetic background (data not shown). Therefore, AtMSI1 functions not only in CAF-1 but also in other chromatin-modifying complexes that do not contain the two larger CAF-1 subunits. Studies on MSI1-like proteins in yeast and animals suggest potential biochemical functions involving histone acetylation, deacetylation or methylation and nucleosome remodelling (Tyler et al.,1996; Martinez-Balbas et al.,1998; Tie et al.,2001; Czermin et al.,2002). Protein binding studies revealed that AtMSI1 interacts with the Arabidopsis retinoblastoma-related protein (RBR) and histone deacetylase HDA1, and that AtMSI2 and AtMSI3 interact with FAS1 (Heidi Feiler,Lars Hennig, N. S., P. T. and W. G., unpublished data). Together, our data suggest AtMSI1-5 have different functional specificities. While some of the functions depend strictly on AtMSI1, other functions (e.g., CAF-1 mediated chromatin assembly) could also involve AtMSI2-5.

AtMSI1 deficiency affects the shoot apical meristem, and both floral meristems and primordia. The fate of the primary shoot meristem after transition to flowering was strongly dependent on AtMSI1 levels. While leaf development during the vegetative phase was affected but not abrogated,essentially no organs developed on the AtMSI1-CS primary shoot after transition to flowering. The appearance of bract-like structures demonstrated that primordia could be initiated by the inflorescence shoot apical meristem,but were unable to differentiate correctly. Secondary inflorescence shoots arising from axillary meristems of rosette leaves were significantly less affected and only occasionally showed a developmental arrest similar to the primary inflorescence shoot. Although the secondary inflorescence shoots usually gave rise to bracts, flowers and lateral shoots, flowers formed on these shoots displayed a progressive loss of floral morphology. As with leaf development during the vegetative phase, flower organs appeared to initiate normally, but their normal differentiation was disrupted. Since the severity of phenotypes increased with additional rounds of cell divisions, perhaps AtMSI1 is required to maintain the epigenetically controlled developmental pattern of gene expression during cell division.

In wild-type Arabidopsis plants, ovule development comprises primordia initiation, specification of identity, pattern formation,morphogenesis and cellular differentiation(Grossniklaus and Schneitz,1998; Schneitz,1999). Ovules arise as finger-like protrusions from the placental tissue of the carpel. After polarity has been established along a proximal-distal (PD) axis of symmetry, megasporogenesis begins and the two integuments initiate. Outgrowth of the outer integument shows strong polarity along the adaxial-abaxial (Ad-Ab) axis, which results in anatropy of mature ovules. Similar to leaf and flower development, initiation of ovule primordia and the integuments occurred normally. Loss of AtMSI1 did not affect polarity along the PD and Ad-Ab axis, but prevented further asymmetric growth and megagametogenesis. Several genes whose functions are required during ovule development have been identified already (for review, see Grossniklaus and Schneitz,1998; Schneitz,1999). In addition to ovule-specific genes, genes involved in other developmental processes are also required for proper ovule development. Among them SUP, AG, WUS and SPLAYED (SPY) are noteworthy, since mutants defective in these genes, or lines in which these genes are ectopically expressed, share several morphological alterations with AtMSI1-CS plants (Ray et al.,1994; Gaiser et al.,1995; Western and Haughn,1999; Groß-Hardt et al.,2002; Wagner and Meyerowitz,2002). In particular, ovule development arrests at similar stages in spy mutants and AtMSI1-CS. SPY is a SWI/SNF ATPase homolog, which is thought to modify activity of the LFY transcription factor by altering chromatin states, and AG is among the genes whose expression depends on LFY and SPY activity(Wagner and Meyerowitz, 2002). Given the phenotypic aspects shared among the mutants, it will be interesting to test possible genetic interactions between SPY and AtMSI1.

In AtMSI1-CS plants, reduction of AtMSI1 levels disrupts the spatial and temporal control of expression for several homeotic genes that regulate plant development and organ identity. The class C floral organ identity gene AG, whose expression is confined to the inner two whorls of flowers in wild-type Arabidopsis, was ectopically expressed in leaves of AtMSI1-CS plants. AP2, a class A floral organ identity gene that is also expressed weakly in leaves(Jofuku et al., 1994; Okamuro et al., 1997), was more strongly expressed in leaves of AtMSI1-CS plants than in wild-type plants. Expression of WUS, AP1, AP3 and SUP was not affected in AtMSI1-CS plants, suggesting that loss of AtMSI1 function affects only a selected class of regulatory genes. Aberrant expression of AG is not restricted to leaves, and activity of the AG::GUS reporter in AtMSI1-CS flowers is also consistent with the observed homeotic changes of organ identity. GUS activity was detected in the tips of enlarged sepals that acquired a carpel-like, abnormal identity characterised by stigmatic structures. GUS staining was much weaker in AtMSI1-CS leaves and was concentrated along the leaf veins (data not shown). Thus, transcriptional control of both the endogenous AGgene and also the reporter transgene depends on AtMSI1 function.

AP2 and AG are known to interact antagonistically and reciprocally inhibit their activation in floral whorls(Bowman et al., 1991). Because AP2 and AG expression domains overlap also in whorl 3 and whorl 4 of WT flowers, AP2 and AG gene products are not sufficient for transcriptional repression (Jofuku et al.,1994). Therefore, the simultanous expression of both AP2and AG in AtMSI1-CS leaves is quite conceivable. Current models suggest that homeotic genes, which control the developmental fate of meristems, are controlled by modulation of their chromatin structures and/or methylation status (Conner and Liu,2000; Jacobsen et al.,2000; Tian and Chen,2001; Yoshida et al.,2001; Wagner and Meyerowitz,2002). For example, AG, AP3 and SUP are ectopically expressed in mutants such as clf, met1 and others(Finnegan et al., 1996; Goodrich et al., 1997). Thus,it is possible that the higher-order chromatin structure or assembly of specific repressor complexes at promoters of AG and AP2depend, either directly or indirectly, on AtMSI1 function. Notably, this function is independent of chromatin assembly by CAF-1 because even the fas1 fas2 double mutant maintains the repressed state of AGand AP2 in leaves.

The functional analysis of AtMSI1 discussed above suggests that the protein may have a role in control of chromatin structure and dynamics. Our cytological analysis of AtMSI-CS nuclei subsequently revealed a significant loss of heterochromatin assembly into chromocentres. Recent reports have established a central role of nuclear chromocentres for the organisation of chromosomal DNA, and specific euchromatic loops were detected that extended from the condensed heterochromatin(Fransz et al., 2002; Soppe et al., 2002). In contrast to the DNA in such loops, DNA present in chromocentres was heavily methylated. Similar to our observations in AtMSI1-CS nuclei, ddm1 and met1, which both have reduced DNA methylation, also had smaller chromocentres. Methylation is a genomic imprint that is required for the maintenance of heterochromatin(Soppe et al., 2002). Strikingly, both met1 and AtMSI1-CS plants assemble less DNA into chromocentres, ectopically express AG and other floral homeotic genes, and share other phenotypic traits(Finnegan et al., 1996; Soppe et al., 2002; this study). We therefore tested if methylation of centromeric repeats, which is strongly decreased in met1 plants, was affected in AtMSI1-CS. Blots of genomic DNA digested with methylation-sensitive or -insensitive restriction endonucleases demonstrated, however, that centromeric methylation patterns were intact in AtMSI1-CS plants(data not shown). In summary, we propose that AtMSI1 is required downstream of DDM1- and MET1-dependent DNA methylation in order to facilitate formation of repressive chromatin structures. Alternatively, AtMSI1 might function in a pathway parallel to DNA methylation. Experiments are in progress to establish the role of AtMSI1 in heterochromatin condensation.

The extent of the phenotypic alterations that resulted from the reduction and loss of AtMSI1 function suggests that AtMSI1 has a fundamental role in development and cellular differentiation. Our view is consistent with the report that RNA-mediated interference (RNAi) of LIN53 expression,which encodes a protein similar to RbAp48 and MSI1, causes embryonic lethality in C. elegans (Lu and Horvitz,1998). It is important to note, however, that not all developmental processes are similarly affected in plants with reduced AtMSI1 levels. The role of AtMSI1 in specific developmental pathways will be understood better once the significance of its interactions with the retinoblastoma tumour suppressor protein, HDA-dependent transcriptional co-repressors, CAF-1, NURF and perhaps other chromatin-modifying complexes has been analysed in more detail. It will also be important to clarify the function of the other MSI1-like proteins in Arabidopsis, which may provide new insights into the biological functions of this class of WD40 proteins in other multicellular eukaryotes as well.

We thank Dr Celia Baroux for valuable technical advise and helpful discussions. We are grateful to Drs Nathaniel P. Hawker and John Bowman for the AG::GUS reporter line. We thank Vivien Exner for help with the cytological experiments. L.H. was supported by fellowships from EMBO and DFG. This research was supported by EU. Project QLG2-1999-00454 (ECCO) and funds from the Swiss Ministry for Science and Education (BBW-No. 00.0223) to W.G.