Initially, members of the MYST group were classified as putative acetyltransferases based on a region in the MYST domain that is homologous to the canonical acetyl-CoA binding domain (motif A) found in GNAT superfamily acetyltransferases 29. The Arabidopsis genome encodes two closely related MYST family proteins HAM1 and HAM2 (87.9% identity, 92.5% similarity in amino-acid sequences), also known as respectively HAG4 and HAG5 6. Wolfe data http://wolfe.gen.tcd.ie/ postulated that HAM1 (At5g64610) and HAM2 (At5g09740) resulted from a duplication event, the α event according to 30, produced by a polyploidization in the Brassicaceae ancestor. The Arabidopsis thaliana HAM1 and HAM2 genes show a Ks value of 0.84. In their measure of divergence between duplicated genes, De Bodt et al. (2005) 31 conclude that the most recent polyploidization event corresponds to a modal Ks value between 0,7 and 0,9. This reinforce the previous observation that the HAM1 and HAM2 genes from Arabidopsis belong to duplicated segments produced by the most recent polyploidization event in the Brassicaceae ancestor.

In order to identify the closest sequences to HAM1 and HAM2, a phylogenetic analysis of the MYST proteins was performed by an extensive search in available databases. Amino acid alignements of the MYST domain were used as the basis for classifying MYST proteins. Fig. 1A shows an unrooted phylogenetic tree illustrating the relationship between 105 MYST proteins (listed in Additional file 1) selected on a total of 130.

The MYST family is divided into five unrelated classes, i.e. not related by significant bootstrap values. The class I comprises proteins from the green lineage including the Arabidopsis HAM1 and HAM2 sequences, two clades of sequences from Mammals, Teleostei, Insects and Cnidarian, and clades of sequences from Alveolata, Fungi, Nematods and Plathelminthes. The class II (bootstrap 79%) groups sequences of Fungi plus one sequence of insect. Classes III and IV (bootstrap 98% and 93%) are exclusively composed of sequences from Fungi. The class V (bootstrap 96%) enclosed sequences from Teleostei, Archausauria, Mammals and Insects. Three sequences (from Human, Insect and Ascidiaceae) form a small additionnal group (bootstrap 84%).

We also noticed that the MYST proteins from the green lineage (Eudicots and Monocots Angiosperms, Gymnosperms, Bryophytes, Chlorophyceae and Prasinophyceae) lie within a cluster supported by a significant bootstrap value (86%). This cluster is associated both with two sequences of Alveolata and 2 sequences of Dictyostelium discoideum, as is frequently observed for sequences from plants. A second cluster of MYST sequence from Prasinophyceae (Ostreococus) is observed. The green lineage phylogenetic tree appears robust with highly significant bootstrap values (Additional file 2). Members of the MYST family of acetyltransferases possess several protein domains. Structural analysis by using Pfam and SMART tools revealed that within each previously defined group, domain organizations of the MYST proteins are highly similar (Fig. 1B). Class I members possess a chromodomain (PF 00385) in the amino-terminus and a zinc finger (C2H2 type) contiguous to a MOZ-SAS acetyltransferase domain (PF 01853). The structure is reduced to the zinc finger and MOZ-SAS domains in classes II and III. Class IV possesses a PHD finger (plant homeodomain zinc finger, PF 00628) in the amino-terminus and a central MOZ-SAS domain. Finally, sequences from class V were much longer, with a linker histone H1 and H5 family domain (PF 00538), one or two PHD finger domain, both at the amino terminus, a C2H2 type zinc finger, contiguous to the MOZ-SAS domain and an extensin-like region (PF04554) at the carboxy terminus. The observed domain organization and protein sizes reinforced the idea of existence of several distinct MYST subfamilies.

In order to study the developmental function of plant MYST genes, a search in Arabidopsis T-DNA insertion mutant collections was performed. Three insertion lines in HAM1 gene were identified. HAM1 mutants, ham1-1, ham1-2 and ham1-3, disrupt the predicted coding region at 155, 1973 and 1959 bp in the genomic DNA and downstream of the initiation codon, respectively (Fig. 2A). A single insertion line, ham2, was identified in HAM2 gene. The T-DNA insert is located 809 bp downstream of the ATG (Fig. 2A). All these mutants are in Columbia-0 (Col-0) background, except for ham1-3 in the Wassilewskija (Ws) background.

Homozygous insertion plants were identified by PCR. RT-PCR experiments, with primers that span the insertion sites, were neither able to detect any HAM1 mRNA in the ham1 mutants nor HAM2 mRNA in ham2 mutant (Fig. 2B). There were no changes in the level of HAM2 mRNA in ham1 mutants lines compared with control. Similarly, HAM1 mRNA was unchanged in ham2 homozygous plants (Additional file 3). This indicates that the absence of one HAM transcript does not affect expression of the other.

We did not observed any abnormal phenotype in plants homozygous for either ham1 or ham2 mutations when grown under standard conditions. In addition, each of the mutant alleles displayed a normal Mendelian segregation ratio (data not shown). As HAM1 and HAM2 are closely related genes, functional redundancy might exist to prevent the appearance of a mutant phenotype in the homozygous mutant lines. Therefore, ham1-1, ham1-2 and ham1-3 homozygous plants were respectively crossed with ham2 homozygous plants to create double mutants. The resulting double-heterozygous F1 were allowed to be self-fertilized and individuals from the resulting progeny were genotyped using PCR. For each crosses, from more than 120 F2 individuals, no homozygous double mutants were detected. However, plants homozygous for insertion at one locus and heterozygous at the other were found. Such plants with either the genotype HAM1/ham1; ham2/ham2 or ham1/ham1; HAM2/ham2 were allowed to self-pollinate. For this F3 population, the expected frequency of ham1ham2 plants is 25%. From 166 seedlings from a HAM1/ham1-1; ham2/ham2 parent, no ham1-1ham2 plants were found (Additional file 4, line1). From 162 seedlings from a ham1-1/ham1-1; HAM2/ham2 parent, no ham1-1ham2 plants were found (Additional file 4, line 2). These data strongly indicate that ham1ham2 double mutant plants are not viable. Because F2 and F3 seeds appeared to be 100% viable, without seedling lethality after germination (data not shown), the loss of a double mutant plant must have occurred early during seed development or before fertilization and suggested that HAM1 and HAM2 have a redundant but important function for Arabidopsis embryo development and/or male/female gametophyte formation.

If only the ham1ham2 developing seeds were not viable, then the progeny of self-fertilize ham1/ham1; HAM2/ham2 and HAM1/ham1; ham2/ham2 plants allowed to self-fertilize would segregate 2/1 for heterozygous and homozygous wild type at the heterozygous locus of the parent. However, the observation was that the percentage of heterozygous seedlings was much less than 67%, with only 32.5% heterozygous at the HAM1 locus and 18.5% heterozygous at the HAM2 locus (Additional file 4, lines 1 and 2). This deviation from a standard inheritance pattern implies death of more than just the double null developing seed.

To confirm this hypothesis, the fertility of ham1/ham1; HAM2/ham2 and HAM1/ham1; ham2/ham2 mutants was determined after self-fertilization and compared to the wild-type. For different genotypes, the length of the siliques was reduced (Fig. 3A). For example, in the case of the HAM1/ham1-1; ham2/ham2 sesquimutant, the length of the siliques was 1.02 ± 0.10 cm compared to 1.44 ± 0.15 cm for the wild-type (n = 40). The number of seeds per silique was significantly reduced compared to wild-type siblings and was greater than the 25% loss expected by loss of the double mutant (Table 1). In addition, dissected siliques illustrated that unfertilized ovules were presents (Fig. 3B). This suggests that the loss of inheritance of the mutant allele occurred early, as a result of female gametophyte death.

To distinguish between arrested in embryo development and abnormalities in gametophytes, reciprocal crosses were performed to analyse inheritance via gametophytes. The null alleles, were successfully inherited from male gamete for both loci, although at a rate reduced from the expected 50% frequency (Additional file 4, lines 3 to 6). When pollen grains of the HAM1/ham1-1; ham2/ham2 or ham1-1/ham1-1; HAM2/ham2 mutants were used to pollinate the stigma of the wild-type female parent, the number of seeds was not significantly different from those observed in the wild-type (Table 1; lines 12 and 14). The frequency of inheritance of the ham allele from the male was 15.5% for ham1 and 9% for ham2 (Additional file 4, lines 4 and 6).

When HAM1/ham1-1; ham2/ham2 or ham1-1/ham1-1; HAM2/ham2 mutants were used as female parent, cleared siliques from these crosses showed that female gametophyte development was arrested in approximately one half of the ovules (Table 1; lines 13 and 15). This finding suggests that aborted ovules may correspond to a female gamete ham1ham2. Genotyping of the progenies confirmed that the inheritance of the ham alleles from the female was null for ham1 and ham2 (Additional file 4; lines 3 and 5).

To track the expression stage of the ham1ham2 mutations, we therefore examined the viability of pollens of the mutants. Pollens collected from wild-type and mutants bearing one copy of either HAM1 or HAM2 were stained with Alexander solution, which stained mature viable pollen grains as purple and dead or dying ones as dark green. The majority of examined pollens from wild-type were viable (Fig. 3C and 3E) with very few dead ones. In anthers of ham1-1/ham1-1; HAM2/ham2 mutants, however, only approximately 60% of pollens grains showed a staining pattern similar to that of the wild-type, the remaining were stained as dark green (n = 1400; Fig. 3D and 3F). These data are consistant with the reduce transmission of ham2 allele describe previously (Additional file 4). Morphologically, the dead pollens were misshapen and smaller (Fig. 3H), which could easily be distinguished from wild-type pollens (Fig. 3G).

We further analyzed male gametophyte development in the double mutant by fluorescence and microscopy. The pollen grain is the male gametophyte in Angiosperms. During microsporogenesis, meiosis of the microspore mother cell produces a tetrad of microspores. After release from the tetrad, during microgametogenesis, each microspore goes through an asymmetric cell division, pollen mitosis I (PMI), to produce a bicellular pollen grain containing a generative cell and a much larger vegetative cell. Only the smaller generative cell undergoes a second round of cell division, pollen mitosis II, to produce two sperm cells 3233. Pollen grains collected from open flowers of both wild-type and mutants bearing one copy of either HAM1 or HAM2, were examined by staining with DNA-specific dye 4',6-diamidino-2-phenylindole (DAPI). When wild-type flowers were open, pollen grains were already mature, and they had two brightly stained sperm nuclei and a faintly stained vegetative nucleus (Fig. 4C). In single mutant plants, microsporogenesis proceeds normally (Additional file 5). By contrast, a part of pollens derived from HAM1/ham1-1; ham2/ham2 and ham1-1/ham1-1; HAM2/ham2 sesquimutant flowers displayed one large DNA mass (~36%, n = 280; Fig. 4E and 4F). These observations suggest that the degeneration of ham1ham2 double mutant pollens occurred mainly at the uninucleated stage before the first pollen mitosis (PM I).

On the female side, during megagametogenesis, meiosis of the megaspore mother cell gives rise to four megaspores, but three degenerates and one survives. This cell undergoes three round of mitosis to form a seven-celled mature embryo sac (female gametophyte) at female gametophyte stage 6 (FG6; 34). The analysis of ovule development of ham1 and ham2 single mutants, using a chloral hydrate clearing protocol and Normarski optics, indicated that megasporogenesis occurred normally in single ham mutants (data not shown). In the ovules of sesquimutants bearing one copy of either HAM1 or HAM2, meiosis always resulted in a single surviving megaspore and as occurred in wild-type, only one of them survived. Initial abnormalities in megagametogenesis were observed only after the completion of meiosis. While half the ovules in both HAM1/ham1-1; ham2/ham2 and ham1-1/ham1-1; HAM2/ham2 siliques were mature showing a wild-type size and shape (Fig. 4G and 4I), the remaining ovules contained only one nucleus localized to the micropylar pole (corresponding to the stage FG1; Fig. 4H and 4J). They produced the 50% aborted ovules (Fig. 3B). These observations suggest that, as observed for the male gametophytic development, the degeneration of the ovule in ham1ham2 double mutants occurs after the uninucleated stage before the first mitosis.

The HAM1 and HAM2 mRNA levels in different organs and tissues were too low to be detected by Northern blots. RT-PCR experiments were used to analyse the expression pattern of HAM genes. Fig. 5 shows that HAM1 and HAM2 genes displayed a similar expression pattern in the different tested organs, with higher expression in flowers compared to leaves, stems and roots and in younger growing leaves compared to mature leaves.

Transgenic Arabidopsis lines were also generated to express the β-glucuronidase (GUS) reporter gene under the control of HAM1 or HAM2 promoter. About 1 kb DNA fragments encompassing the putative promoter regions of HAM genes were fused to the GUS coding sequence and these constructs were introduced into Col-0. A reproducible and overlapping expression pattern was found in three independent reporter lines for each promoter. Under standard growth conditions, promoter activity was detected in the shoot apex of the seedlings as well as in the cotyledons and leaves (Fig. 6A, B). In leaves, a patchy expression pattern was observed which corresponded to strong GUS staining at the basis of the trichomes (Fig. 6B). GUS activity was never detected in the hypocotyls and petiole. A faint GUS signal was occasionally detected in root hairs (Fig. 6A) but never in the internal root tissues. By contrast, GUS activity was strong in developing flowers, particularly in the anthers and gynoecia but not in mature flowers (Fig. 6C) in which a slight GUS activity was localized to the stigma. Transversal sections confirmed the GUS staining in developing gynoecia (Fig. 6D) and young pollens (Fig. 6E) and the absence of GUS expression in mature tricellular pollen grains (Fig. 6F).

ham1-1 and ham2 mutants carrying T-DNA insertions respectively within HAM1 (At5g64610; SALK_027726) and HAM2 (At5g09740; SALK_106046) were obtained from The Nottingham Arabidopsis Stock Center (NASC). ham1-2 was a GABI-KAT line (050B11) 45. acquired from the NASC. ham1-3 (EHQ293) was obtained from the Versailles collection. The following PCR primers were used to genotype plants carrying these T-DNA insertions. ham1-1 F: 5'-ATGGTGTGCGAATCTATGACC-3'; ham1-1 R: 5'-TCAAGGTCAAGCTGTTCAAGC-3'; ham1-2 F: 5'-TTACAGGTGGGCAAGAAG-3'; ham1-2 R: 5'-ACCATCCAGACAAAAGATTCC-3'; ham1-3 F: 5'-AAGGAAGGGCTATGGCAAAT-3'; ham1-3 R: 5'-CGTTTTACCATCCAGACAAAAG-3' ham2 F: 5'-GTCGAAGAAGAGGAAAATGGG-3'; ham2 R: 5'-CATATGCCTTTGAAGCTGCTC-3'; T-DNA left border: 5'-CGATTTCGGAACCACCATCAAACAGGA-3'. All of the T-DNA mutants and wild-type plants in this study were from the Columbia ecotype Col-0 excepted for ham1-3 mutant in the Ws background.

Arabidopsis plants were grown in a greenhouse under long-day conditions (16 h of light) at 19.5°C (day) and 17.5°C (night). For in vitro cultures, seeds were sown on 0.5 Murashige and Skoog medium, incubated at 4°C for 48 h, and then transferred to growth chambers.

Arabidopsis leaves were used for genomic DNA extraction. PCR were carried out using the Promega GoTaq polymerase. Total RNA was isolated with TRIzol reagents (Invitrogen). First strand cDNA was synthesized from 3 μg of total RNA using ImProm-II reverse transcriptase (Promega). Polymerase chain reaction primers specific to the predicted cDNA sequences of each gene were used. For HAM1: P1, 5'-ATGGGATCGTCTGCGGATACA-3'; P2, 5'-GAATTCGTGAGAGCGAGTATCGCA-3'; P3, 5'-AGTCATCTAAGGATATGCAGA-3'. For HAM2: P4, 5'-CCTTTAACTCCTGATC-AAGCTAT-3'; P5, 5'-CTACAGCGCACTCTACTGAATC-3'; P6, 5'-GACAGCCCGCTTTACTTACACA-3'. For Actin: ACT-FP, 5'-ACCCAAAGGCCAACAGAGAGA-3'; ACT-RP, 5'-TGCTTGGTGCAAGTGCTGTGA-3'.

To examine pollen viability, pollen grains were stained with Alexander solution 46. The pollen nuclei were stained with DAPI according to a method already describe by 47. For observations of ovules, siliques were fixed overnight using 3:1 acetic acid: ethanol and then washed with 70% ethanol before clearing in 8: 1: 2 chloral hydrate: glycerol: water. For scanning electron microscope (Hitachi S-3000) analysis, samples were slowly frozen at -18°C under a partial vacuum on the Peltier stage before observation under the environmental secondary electron detector mode.

The HAM1 and HAM2 promoters sequences used contain 1 kb upstream of the ATG and were amplified from Arabidopsis genomic DNA with the 5'-GCAGAATTCTCATTGTAGGTAAAAGAA-3' and 5'-GGATCCTTCTTTAGTCGGGTCGGA-3' primers for HAM1 and the 5'-GCGAATTCGTCTAACAGACTAAACGT-3' and 5'-CGGATCCTTCTCGGTCGGGTCGGAG-3' primers for HAM2. The corresponding PCR fragments were cloned into PUC19 vector and then transferred upstream the uidA gene in the plant transformation vector pPR97. The constructs were used to transform Col plants by the floral dip method. Transgenic plants were obtained on kanamycin containing medium and later transferred to soil for optimal seed production. For analysis of GUS activity, samples were prefixed in 90% acetone at room temperature for 20 min, rinsed in staining buffer without 5-bromo-4-chloro-3-indolyl-_-D-glucoronic acid (X-Gluc), infiltrated with staining solution (100 mM sodium phosphate buffer, pH 7, 5 mM potassium ferrocyanide, 5 mM potassium-ferricyanide, 1 mM X-Gluc) under a vacuum for 15 min, and incubated at 37°C for 14 h. After a progressive dehydration in a series ethanol concentrations up to 70%, samples were cleared in 8: 1: 2 chloral hydrate: glycerol: water. For sections, the stained samples were fixed in FAA at 4°C overnight, and then embedded in Leica historesin. Semithin sections (3 μm) were cut and analyzed under a microscope.

We searched for MYST sequences from protein databases at NCBI. 130 protein sequences were downloaded from numerous genomes. This allows to analyse sequences from Plants, Algae, Mycetozoans, Insects, Teleostei, Mammals (including Monotremata and Marsupial), Amphibia, Cnidarian, Echinozoa, Fungi, Plathelminthes, Nematods, Alveolata, Archausauria and Ascidiaceae. Bacteria do not have counterpart to the MYST proteins.

The amino-acid alignment was conducted using Clustal 4849 with default parameters. The generated alignment was adjusted manually. Amino acid alignments of the MYST domain were used as the basis for classifying MYST proteins. The unrooted tree was created using the PhYML algorithm and the maximum likelihood method 50. To assess support for each node, bootstrap analysis were performed using 500 bootstrap replicates. A bootstrap value of 70% is likely to be correct at the 95% level, and bootstrap values higher than 70% were taken as sufficient evidence for grouping.

To search for domain organization in the MYST proteins, we analyzed the sequences in Pfam http://pfam.sanger.ac.uk51, Prosite http://www.expasy.ch/prosite52 and SMART http://smart.embl-heidelberg.de/index2.cgi52 databases.

To assess the age of the divergence between sequences, we estimated the level of synonymous substitutions (Ks) between the Arabidopsis thaliana HAM1 and HAM2 genes. After removing gaps in the nucleotide alignment, per-site synonymous (Ks) and nonsynonymous (Ka) substitution rates were calculated using PAL2NAL http://coot.embl.de/pal2nal/54.