Analysis of histone acetyltransferase and histone deacetylase families of Arabidopsis thaliana suggests functional diversification of chromatin modification among multicellular eukaryotes

Abstract

Sequence similarity and profile searching tools were used to analyze the genome sequences of Arabidopsis thaliana, Saccharomyces cerevisiae, Schizosaccharomyces pombe, Caenorhabditis elegans and Drosophila melanogaster for genes encoding three families of histone deacetylase (HDAC) proteins and three families of histone acetyltransferase (HAT) proteins. Plants, animals and fungi were found to have a single member of each of three subfamilies of the GNAT family of HATs, suggesting conservation of these functions. However, major differences were found with respect to sizes of gene families and multi-domain protein structures within other families of HATs and HDACs, indicating substantial evolutionary diversification. Phylogenetic analysis identified a new class of HDACs within the RPD3/HDA1 family that is represented only in plants and animals. A similar analysis of the plant-specific HD2 family of HDACs suggests a duplication event early in dicot evolution, followed by further diversification in the lineage leading to Arabidopsis. Of three major classes of SIR2-type HDACs that are found in animals, fungi have representatives only in one class, whereas plants have representatives only in the other two. Plants possess five CREB-binding protein (CBP)-type HATs compared with one to two in animals and none in fungi. Domain and phylogenetic analyses of the CBP family proteins showed that this family has evolved three distinct types of CBPs in plants. The domain architecture of CBP and TAF_II250 families of HATs show significant differences between plants and animals, most notably with respect to bromodomain occurrence and their number. Bromodomain-containing proteins in Arabidopsis differ strikingly from animal bromodomain proteins with respect to the numbers of bromodomains and the other types of domains that are present. The substantial diversification of HATs and HDACs that has occurred since the divergence of plants, animals and fungi suggests a surprising degree of evolutionary plasticity and functional diversification in these core chromatin components.

INTRODUCTION

Gene expression in eukaryotes involves a complex interplay among transcription factors and chromatin proteins that pack chromosomal DNA into the confined space of the nucleus while poising genes for activation or repression (1). The basic unit of chromatin is the nucleosome core particle, a structure in which ∼146 bp of DNA is wrapped around a protein octamer made up of two subunits each of the core histones H2A, H2B, H3 and H4 (2). Core histones can exist in multiple alternative states of acetylation, methylation, phosphorylation, ubiquitination or ADP-ribosylation (3). The regulatory significance of these modifications for processes including gene repression, gene activation and replication is increasingly clear (4–6).

Lysines at the N-terminal ends of the core histones are the predominant sites of acetylation and methylation and a regulatory role for these modifications was proposed as early as 1964 (7). However, decades passed before it was demonstrated that active genes are preferentially associated with highly acetylated histones whereas inactive genes are associated with hypoacetylated histones (8). The N-termini of histones H3 and H4 were subsequently shown to be essential for repression of the silent mating type loci in Saccharomyces cerevisiae (9,10). Enhancer-dependent activation of other S.cerevisiae genes also required these N-terminal sequences (11–13). Collectively, these studies suggested that histones are integral to both gene activation and gene repression mechanisms. A breakthrough was the finding that a Tetrahymena thermophila protein with histone acetyltransferase (HAT) activity shared substantial similarity with S.cerevisiae Gcn5p (14), the catalytic subunit of several multi-protein complexes required to activate a diverse set of genes. A complementary breakthrough was the finding that a purified mammalian histone deacetylase (HDAC) was similar to Rpd3p (15), a protein which helps repress numerous genes in S.cerevisiae (16), also as part of a larger protein complex (17–19). Histone acetylation and deacetylation are thought to exert their regulatory effects on gene expression by altering the accessibility of nucleosomal DNA to DNA-binding transcriptional activators, other chromatin-modifying enzymes or multi-subunit chromatin remodeling complexes capable of displacing nucleosomes (20,21).

Sequence characterization reveals at least four distinct families of HATs and three families of HDACs (3,22,23). HATs include: (i) the GNAT (GCN5-related N-terminal acetyltransferases)-MYST family (24,25) whose members have sequence motifs shared with enzymes that acetylate non-histone proteins and small molecules; (ii) the p300/CREB-binding protein (CBP) co-activator family in animals implicated in regulating genes required for cell cycle control, differentiation and apoptosis (26,27); and (iii) the family related to mammalian TAF_II250, the largest of the TATA binding protein-associated factors (TAFs) within the transcription factor complex TFIID (28). These three families are widespread in eukaryotic genomes, and homologous proteins are also involved in non-HAT reactions in prokaryotes and Archaea. Mammals have a fourth HAT family that includes nuclear receptor coactivators such as steroid receptor coactivator (SRC-1) and ACTR, a thyroid hormone and retinoic acid coactivator that is not represented in plants, fungi or lower animals (22,29,30).

Major groups of HDACs include the RPD3/HDA1 superfamily, the Silent Information Regulator 2 (SIR2) family (31) and the HD2 family. RPD3/HDA1-like HDACs are found in all eukaryotic genomes. Interestingly, homologous proteins that have acetate utilization and acetylpolyamine aminohydrolase activities are also present in bacteria and Archaea, organisms that lack histones (17,32). The SIR2 family of HDACs is distinctive in that it has no structural similarity to other HDACs and requires NAD as a cofactor (33). In S.cerevisiae, SIR2 is known to play roles in repression of silent mating type loci (34), repression of rRNA gene recombination (35), and repression of protein-coding genes inserted near telomeres (34) or within rRNA gene arrays (36). Mutations in SIR2 also affect aging and longevity in S.cerevisiae (37,38). SIR2-related proteins form a large family with members present in all kingdoms of life, including bacteria (39). The third family, the HD2-type HDACs, were first identified in maize and appear to be present only in plants (40,41). HD2-type HDACs are homologous to a class of cis–trans prolyl isomerases present in other eukaryotes (42).

Limited information is available concerning the roles of most proteins in the four HAT homology groups and the three HDAC homology groups in control of gene expression in multicellular eukaryotes, especially in plants (43,44). Here, we present phylogenetic and domain analyses of HAT- and HDAC-related proteins identified in searches of the essentially complete Arabidopsis thaliana genome sequence. To test and correct open-reading frames (ORFs) predicted by exon-modeling algorithms, cDNA sequences were determined for most of these proteins. Alternative splicing was demonstrated for 3 of 16 genes encoding HDACs. Together, these data provide a foundation for the functional analysis of these important chromatin-modifying activities in Arabidopsis, as well as in other plants and model organisms.

MATERIALS AND METHODS

Database similarity searches of the Arabidopsis genome and other plant sequences

Known HDAC and HAT protein sequences available from a variety of eukaryotic organisms (Table S1) were used as queries to search the complete Arabidopsis genome sequence (45) using the TBLASTN and TFASTX programs (46,47). To assure that all homologous genes in these families had been identified, three additional searches were performed. First, all Arabidopsis protein sequences in GenBank including those predicted by genome annotation were searched with the query sequences using BLASTP, FASTA and SSEARCH. Secondly, these protein sequences were searched for protein family (Pfam) domains known to be present in previously characterized HDAC and HAT proteins using the program HMMER (http://hmmer.wustl.edu/). Thirdly, predicted Arabidopsis HDAC and HAT proteins were used as queries to search for additional paralogous genes in the Arabidopsis genome sequence using TBLASTN and TFASTX. Sequences having an E-value of 0.01 or less were investigated further. However, this third approach did not find any proteins in addition to those that had already been identified by the initial TBLASTN or TFASTX searches.

Gene nomenclature

The genes identified in this study are listed in Table 2. To designate newly identified genes, we used three-letter symbols that specify the homology group to which a gene belongs as follows: HAG for HATs of the GNAT/MYST superfamily, HAC for HATs of the CBP family, HAF for HATs of the TAF_II250 family, HDA for HDACs of the RPD3/HDA1 superfamily, SRT for HDACs of the SIR2 family (sirtuins), and HDT for HDACs of the HD2 family (‘HD-tuins’). To designate individual genes within a homology group, the three-letter symbol is followed by a numeral that does not imply orthology because in many cases it was not possible to determine orthology. To ensure that orthology is not inferred from numerals, a different series of numerals was assigned to different species: A.thaliana genes are indicated by the numerals 1–99, Zea mays by 101–199, S.cerevisiae by 201–299, Caenorhabditis elegans by 301–399, Drosophila melanogaster by 401–499 and Schizosaccharomyces pombe by 601–699 (for other organisms, see Table 2). Names of genes previously assigned in the literature or in GenBank were retained, except in Arabidopsis, for which we propose that the designations defined here should be used. To avoid possible confusion with HDA1 of S.cerevisiae, the Arabidopsis HDA series begins with HDA2.

Table 2. Sequence accession numbers for HAT and HDAC genes analyzed.

Genes (synonyms)	Organism	GenBank accession no.
		Protein	ESTs
HDACs
(1) RPD3/HDA1 family
Eukaryotes
Class I
atHDA6 (AtRPD3B)	Arabidopsis thaliana	BAB10553
atHDA7	Arabidopsis thaliana	BAB09994
atHDA9	Arabidopsis thaliana	CAB72470
atHDA19 (AtRPD3A)	Arabidopsis thaliana	AAB66486
zmRPD3	Zea mays	AAC50038
zmHd1b-II	Zea mays	AAD10139
scRPD3	Saccharomyces cerevisiae	P32561
scHOS2	Saccharomyces cerevisiae	P53096
scHOS1	Saccharomyces cerevisiae	Q12214
ceHDA301	Caenorhabditis elegans	CAB03984
ceHDA302	Caenorhabditis elegans	Q09440
ceHDA303	Caenorhabditis elegans	CAB03224
dmHDA401	Drosophila melanogaster	AAC61494
dmHDA402	Drosophila melanogaster	AAC83649
hsHDAC1	Homo sapiens	Q13547
hsHDAC2	Homo sapiens	NP_001518
hsHDAC3	Homo sapiens	NP_003874
hsHDAC8	Homo sapiens	NP_060956
spCLR6	Schizosaccharomyces pombe	CAA19053
spPHD1	Schizosaccharomyces pombe	O13298
osHDA702	Oryza sativa	AAK01712.1
gmHDA1201	Glycine max		BF066371, BF324960, AW396525, AW308961
			AI736699, AW460060, AW598663
mcHdeac1	Mesembryanthemum crystallinum	AAF82385
anRPD3A	Aspergillus nidulans	AAF80489
anHOS2A	Aspergillus nidulans	AAF80490
ttTHD1	Tetrahymena thermophila	AAG00980
pfHDA1	Plasmodium falciparum	AAD22407
lmHDA2	Leishmania major	CAC14522
Class II
atHDA5	Arabidopsis thaliana	NP_200914
atHDA15	Arabidopsis thaliana	BAB01118
atHDA18	Arabidopsis thaliana	NP_200915
zmHDA109	Zea mays		AW216192, AW256098, AW258010, BE238824
zmHDA110	Zea mays		AW231694, BE510782
scHDA1	Saccharomyces cerevisiae	P53973
ceHDA304	Caenorhabditis elegans	Q20296
ceHDA305	Caenorhabditis elegans	CAA90401
ceHDA306	Caenorhabditis elegans	AAB71243
ceHDA307	Caenorhabditis elegans	CAA21669
dmHDA404	Drosophila melanogaster	AAD21090
dmHDA405	Drosophila melanogaster	AAF48245
hsHDAC4	Homo sapiens	NP_006028
hsHDAC5	Homo sapiens	NP_005465
hsHDAC6	Homo sapiens	NP_006035
hsHDAC7	Homo sapiens	AAF63491
spCLR3	Schizosaccharomyces pombe	CAC01518
mmHDA2	Mus musculus	T13964
Class III
atHDA2	Arabidopsis thaliana	AAD40129
ceHDA308	Caenorhabditis elegans	CAA94910	Z71185
dmHDA403	Drosophila melanogaster	AAF56350
hsHDAC11	Homo sapiens		BE613615, HSM802049, AA346002
Unclassified
atHDA8	Arabidopsis thaliana	AAF22892
atHDA14	Arabidopsis thaliana	CAB38805
scHOS3	Saccharomyces cerevisiae	Q02959
Prokaryotes
aaAcu	Aquifex aeolicus	D70388
afAph	Archaeoglobus fulgidus	B69266
apAph	Aeropyrum pernix	BAA79001
bhAph	Bacillus halodurans	BAB06956
bsAcu	Bacillus subtilis	S39643
drAcu	Deinococcus radiodurans	AAF10411
haloAcu	Halobacterium sp. NRC-1	AAG18756
mjAph	Methanococcus jannaschii	AAB98526
mrAph	Mycoplana ramose	Q48935
mtAph	Methanobacterium thermoautotrophicum	C69026
nmAph	Neisseria meningitides	AAF41032
paAph	Pyrococcus abyssi	B75095
phAph	Pyrococcus horikoshii	H71071
psAph	Pseudomonas aeruginosa	D83174
stcoAcu	Streptomyces coelicolor	T36278
sxAcu	Staphylococcus xylosus	Q56195
synGln	Synechococcus PCC7002	CAA78367
sypAph	Synechocystis PCC6803	S74557
vcAcu	Vibrio cholerae	AAF95190
(2) HD2 familya
HDT1 (AtHD2A)	Arabidopsis thaliana (Arabidopsis)	AAB70032
HDT2 (AtHD2B)	Arabidopsis thaliana (Arabidopsis)	AAC02539
HDT3	Arabidopsis thaliana (Arabidopsis)	AAF70197
HDT4	Arabidopsis thaliana (Arabidopsis)	AAF70198
Hd2a	Zea mays (maize)	AAC61674
Hd2b	Zea mays (maize)	AAF68624
Hd2c	Zea mays (maize)	AAF68625
HDT101	Zea mays (maize)	AAK67143
HDT701	Oryza sativa (rice)		AU082425, AU68818, D22916, C72062, D15380, AU067990, D39074
HDT801	Triticum aestivum (wheat)		BE586047, BE404638, BE415745, BE400260
HDT802	Triticum aestivum (wheat)		BE518130, BE517665, BE402732, BE415484
HDT901	Sorghum bicolor (wheat)		AW565004, BE359652, BE356690, AW564222, BE594360, BE366313
HDT1001	Hordeum vulgare (barley)		BE194667
HDT1101	Lycopersicon esculentum (tomato)		AW218751, AI488715, AW625291, BE435034, AW092756, AW093039,AW623008, AW97980
HDT1102	Lycopersicon esculentum (tomato)		AW929159, AW429116, BE449829, AI776222, AW615938, AW039158
HDT1103	Lycopersicon esculentum (tomato)		AW037620, BE463291
HDT1202	Glycine max (soybean)		AW707012, BE022138
HDT1301	Medicago truncatula (barrel medic)		AA660660, AW688123, BE322886, AL375160,
HDT1302	Medicago truncatula (barrel medic)		BE318805, BE32160
HDT1401	Solanum tuberosum (potato)		BE343097, BE342419, BE340133, BE342346, BE922565
HDT1402	Solanum tuberosum (potato)		BE341470, BE343018, BE924296, BE921506, BE343072, BE473230,AW906030, AW906356
HDT1501	Gossypium arboretum (cotton)		AW730862
HDT1502	Gossypium arboretum (cotton)		AW729511
HDT1601	Mesembryanthemum crystallinum (ice plant)		BE035409
HDT1602	Mesembryanthemum crystallinum (ice plant)		BE034206
HDT1701	Lotus japonicus		AW720250, AV409362, AV422554
(3) SIR2 family
atSRT1	Arabidopsis thaliana	BAB09243
atSRT2	Arabidopsis thaliana	CAC05449
zmSRT101	Zea mays		AI734474, AI770366, AI861441, AW000351, AW000357, AW202474
scHST1	Saccharomyces cerevisiae	P53685
scHST2	Saccharomyces cerevisiae	P53686
scHST3	Saccharomyces cerevisiae	P53687
scHST4	Saccharomyces cerevisiae	P53688
scSIR2	Saccharomyces cerevisiae	NP_010242
ceSRT309	Caenorhabditis elegans	T24172
ceSRT310	Caenorhabditis elegans	T22325
ceSRT311	Caenorhabditis elegans	T22324
ceSRT312	Caenorhabditis elegans	T25520
dmSRT406	Drosophila melanogaster	AAC79684
dmSRT407	Drosophila melanogaster	AAF46055
dmSRT408	Drosophila melanogaster	AAF54513
dmSRT409	Drosophila melanogaster	AAF56851
hsSIRT1	Homo sapiens	NP_036370
hsSIRT2	Homo sapiens	AAD40850
hsSIRT3	Homo sapiens	AAD40851
hsSIRT4	Homo sapiens	NP_036372
hsSIRT5	Homo sapiens	NP_036373
hsSIRT6	Homo sapiens	NP_057623
hsSIRT7	Homo sapiens	NP_057622
spSIR2	Schizosaccharomyces pombe	T39571
spHST2	Schizosaccharomyces pombe	T40929
spHST4	Schizosaccharomyces pombe	AAD53752
osSRT701	Oryza sativa	AAD42226
taSRT803	Triticum aestivum		BE498844
leSRT1104	Lycopersicon esculentum		BG125699, BG134964, BE354229, BE450731
leSRT1105	Lycopersicon esculentum		BF050336, AW441312, BE354229
mtSRT1303	Medicago truncatula		AW686356, BM814514.1, BF647706.1
HATs
(1) GNAT-MYST family
GCN5
atHAG1 (atGCN5)	Arabidopsis thaliana	AAB92257
scGCN5	Saccharomyces cerevisiae	NP_011768.1
ceHAG303	Caenorhabditis elegans	AAF60658.1
dmHAG401	Drosophila melanogaster	AAC39102.1
hsPCAF	Homo sapiens	NP_003875.2
spHAG601	Schizosaccharomyces pombe	T37933
ELP3
atHAG3	Arabidopsis thaliana	BAB09451
scELP3	Saccharomyces cerevisiae	NP_015239.1
ceHAG304	Caenorhabditis elegans	CAB01454.1
dmHAG408	Drosophila melanogaster	AAF51012.1
hsHAG510	Homo sapiens	BAB14138.1
spHAG602	Schizosaccharomyces pombe	CAB10146
HAT1
atHAG2	Arabidopsis thaliana	BAB09892
scHAT1	Saccharomyces cerevisiae	U33335.1
ceHAG302	Caenorhabditis elegans	CAA88954.1
dmHAG409	Drosophila melanogaster	AAF51953.1
hsHAG509	Homo sapiens	XM_002242
spHAG603	Schizosaccharomyces pombe	CAB16872.1
MYST
atHAG4	Arabidopsis thaliana	BAB11428
atHAG5	Arabidopsis thaliana	CAB89356
scESA1	Saccharomyces cerevisiae	NP_014887.1
scSAS2	Saccharomyces cerevisiae	CAA88552.1
scSAS3	Saccharomyces cerevisiae	NP_009501.1
ceHAG305	Caenorhabditis elegans	T19693
ceHAG306	Caenorhabditis elegans	CAA96668.1
ceHAG307	Caenorhabditis elegans	CAB04552.2
ceHAG308	Caenorhabditis elegans	AAC78211.1
dmMOF	Drosophila melanogaster	AAC47507.1
dmHAG404	Drosophila melanogaster	AAF44628.1
dmHAG405	Drosophila melanogaster	CAA21829.1
dmHAG406	Drosophila melanogaster	AAF47164.1
dmHAG407	Drosophila melanogaster	AAF56792.1
hsTIP60	Homo sapiens	NP_006379
hsMOZ	Homo sapiens	NP_006757.1
hsMORF	Homo sapiens	AAF00095.1
hsHAG511	Homo sapiens	XP_050187
hsHBOA	Homo sapiens	NP_008998
hsHAG512	Homo sapiens	XP_018398
spHAG604	Schizosaccharomyces pombe	CAA93696.1
spHAG605	Schizosaccharomyces pombe	CAA22591.1
HPA2
scHPA2	Saccharomyces cerevisiae	NP_015519
scHPA3	Saccharomyces cerevisiae	NP_010848
spATS-1	Schizosaccharomyces pombe	NP_593494
(2) CBP family
atHAC1	Arabidopsis thaliana	AAC17068
atHAC2	Arabidopsis thaliana	AAB95246
atHAC4	Arabidopsis thaliana	AAF79331
atHAC5	Arabidopsis thaliana	AAF35947
atHAC12	Arabidopsis thaliana	AAG09087
ceCBP-1	Caenorhabditis elegans	P34545
dmCBP	Drosophila melanogaster	T13828
hsCBP	Homo sapiens	S39162
hsp300	Homo sapiens	A54277
mmCBP	Mus musculus	S39161
(3) TAFII250 family
atHAF1	Arabidopsis thaliana	AAF25977
atHAF2	Arabidopsis thaliana	BAB01700
scTAFII145	Saccharomyces cerevisiae	AAA79178
ceTAFII250	Caenorhabditis elegans	CAB04907
dmTAFII230	Drosophila melanogaster	P51123
hsTAFII250	Homo sapiens	NP_004597
spTAFII145	Schizosaccharomyces pombe	CAA91179

^aThe common names as presented in Figure 5 are included in parentheses next to the Latin name.

Gene annotation

High quality plant protein sequences for phylogenetic analysis were obtained in several steps. First, the gene prediction programs GeneMark (48), GenScan (49) and NetPlantGene (50) were used to produce gene models for these sequences. From these separate models, a single consensus model was derived. To verify the gene models, RNA gel blots were used to determine the length of the mRNA from each expressed gene. The positions of exons in the consensus model were then tested by analysis of available Arabidopsis EST sequences using the gene prediction tool, GeneSeqer (51). For genes that were not completely represented by EST sequences, EST clones were obtained from the ABRC and Kazusa stock centers and sequenced. Remaining gaps between known cDNA sequences were filled by sequencing RT–PCR amplification products obtained using total RNA as the template and using primers that annealed to predicted coding sequences. Although the actual start codon for each protein has not been identified with certainty, none of the predicted proteins lack known conserved N-terminal or C-terminal domains, suggesting that the modified gene models are reasonably accurate.

cDNA sequence was not determined for HDA10 and HDA17 because these genes are truncated in their HDAC domains, HAC2 because it could not be amplified by RT–PCR, and HAC12 and HAF2 because they are highly similar to HAC1 and HAF1, respectively. HAC12 and HAF2 were annotated according to the splicing models of HAC1 and HAF1. In the case of HAC4, the only transcript we detected carries a premature nonsense codon that would eliminate conserved regions of the protein, although the transcript extends beyond this and contains these conserved regions. Thus, for purposes of the phylogenetic and domain analyses presented here, we have used an algorithm-derived splicing model that predicts the conserved CBP-type HAT domain and cDNA sequence-derived splice junctions in the remainder of HAC4. Alternative splicing products were observed for three genes: HDA2, HDA15 and SRT2. For purposes of the phylogenetic analyses presented here, we used the predicted protein sequence that possessed intact conserved domains (HDA2alt1, HDA15alt1 and SRT2alt1).

The cDNA sequence data for the HAT and HDAC genes have been submitted to the GenBank data library under the following accession numbers: HAC4 (AF512559, AF512560, AH011643), HAC5 (AF512557, AF512558, AH011642), HAG2 (AF512724), HAF1 (AF510669), HDA2alt1 (AF510671), HDA2alt2 (AF510165), HDA7 (AF510166), HDA9 (AF512725), HDA15alt2 (AF510169), HDA15alt3 (AF510170), HDA18 (AF510670), SRT2alt2 (AF510171), SRT2alt3 (AF510172), SRT2alt4 (AF510173), SRT2alt5 (AF510174), SRT2alt6 (AF510175). For rest of the genes, cDNA sequences submitted by other groups were found in GenBank and were identical to the sequence data generated by Plant Chromatin Consortium. Their accession numbers are as follows: HAC1 (AF323954), HAG1 (AF338768), HAG3 (AY056323), HAG4 (AY099684), HAG5 (NM_121011), HDA5 (AY090936), HDA6 (AF195548), HDA8 (AY097371), HDA14 (AY052234), HDA15alt1 (NM_112737), HDA19 (AY093153), HDT1 (AF195545), HDT2 (AF044914), HDT3 (AF372889), HDT4 (AF255713), SRT1 (AF283757), SRT2alt1 (AY045873).

Similarity searches of non-plant genomes for HDAC and HAT genes

The genomes of a diverse group of organisms were searched with the query sequences (Table S1), as well as with any Arabidopsis HDAC and HAT sequences showing similarity to these query sequences. First, BLASTP searches of the individual proteomes of baker’s yeast (S.cerevisiae), nematodes (C.elegans), fruit flies (D.melanogaster), and several species of bacteria and Archaea were conducted. Secondly, genomic sequences of humans (Homo sapiens), fission yeast (S.pombe; http://www.sanger.ac.uk/Projects/S_pombe/), and leishmania (Leishmania major; http://www.sanger.ac.uk/Projects/L_major/) were individually searched for homologous sequences using TBLASTN. Thirdly, the public GenBank nr (non-repeating) databases were searched to identify homologs in additional species using BLAST and PSI-BLAST. Fourthly, a large number of plant EST collections (including Z.mays, Oryza sativa, Lycopersicon esculentum, Medicago truncatula, Glycine max, Triticum aestivum, Sorghum bicolor, Gossypium arboreum, Solanum tuberosum, Hordeum vulgare, Lotus japonicus and Mesembryanthemum crystallinum) were searched using TBLASTN. Plant ESTs were assembled into contigs using the FAKtory DNA sequence assembly system (http://bcf.arl.arizona.edu/faktory/) and the contigs were translated into amino acid sequences for further analysis.

Analysis of protein families

Phylogenetic analysis. Protein sequences and domains were aligned using Clustal W (52), edited with Genedoc (http://www.psc.edu/biomed/genedoc/), and an unrooted phylogenetic tree was constructed by the distance method using the neighbor-joining algorithm implemented in the program Neighbor in the PHYLIP (3.5) package (53). The Dayhoff PAM model of protein evolution was used to compute the distances between the sequences (54) using the PROTDIST program. This analysis allowed the identification of the most similar protein sequences in the same or different organisms based upon protein sequence similarity in the multiple sequence alignment. These alignments are available in Figures S1–S3. Identification of a paralogous family of sequences was revealed by the presence of a cluster of similar sequences from one organism or group of organisms that appeared to have arisen by gene duplication. Assignments of likely orthology were based upon the observation of a high level of sequence similarity among unique sets of sequences present in diverse organisms. In order to assess how well the multiple sequence alignment supported the branch patterns in the predicted phylogenetic tree of the sequences, a bootstrap analysis was performed using PHYLIP. This method resampled columns in the multiple sequence alignment to generate 500–1000 new alignments, each of which was used to produce a new tree. The number of alignments that support each branch pattern in the tree was then assessed and is reported in the appropriate figure. When a clear majority of bootstrap trees (>70%) were in agreement, support was considered to be good. In many cases, bootstrap support was excellent, in the 95–100% range.

For phylogenetic analysis of the HD2 family (Fig. 5), mRNA and EST sequences encoding the HD2-type HDAC domains were aligned by CLUSTALW. These alignments are available in Figure S4. Following some minor editing to match the codons to the protein sequence alignment, an unrooted tree was produced using the maximum likelihood method as implemented in the DNAML program using the default transition/transversion ratio of 2:1 in the PHYLIP suite.

Figure 5.

Maximum likelihood analysis of the plant HD2 family nucleic acid sequences. This analysis is based upon a codon-by-codon alignment of the first 273 positions of the maize HD2 cDNA sequence, corresponding to the HDAC domain, with other plant cDNA and EST sequences listed in Table 2. The common name for each species is listed in parentheses and where common names are not available, the Latin name is included. The gene names and their accession numbers are identified in Table 2. Confidence levels for the tree branches that are best supported by bootstrap analysis are shown as percentages.

Motif analysis of the RPD3/HDA1-related HDACs. To identify common motifs in the HDAC domain, a multiple sequence alignment of representative proteins in each of the three HDAC classes was generated by CLUSTALW. Each multiple sequence alignment was then searched for regions with strongly conserved patterns having high information content (55). Information content was determined by producing a sequence logo using WebLogo (http://www.bio.cam.ac.uk/cgi-bin/seqlogo/logo.cgi). A logo is a graph that displays the amount of information at each column in the alignment and is measured in bits (reduction in uncertainty above background amino acid frequencies). The logo also shows the contribution of each amino acid to this information.

Domain analysis of HAT proteins. HAT protein sequences identified in Arabidopsis and in the proteomes of other organisms were analyzed for the presence of any domain present in the Pfam (Protein Family) database. The collection of Pfam hidden Markov model (HMM) profiles for domain families (version 6.5) was downloaded from the Pfam web site. Sequence profile searches were performed using the software HMMER (http://hmmer.wustl.edu/). For certain domains such as the CBP-type HAT domain, for which a Pfam model is not available, a multiple sequence alignment generated using CLUSTALW was examined for the presence of the biochemically-defined HAT domain in human CBP protein (26) and a profile HMM was constructed using programs in the HMMER package. The Predict Protein resource based on neural networks (PHD at http://maple.bioc.columbia.edu/predictprotein/) and the Discrimination of protein secondary structure server (DSC at http://bioweb.pasteur.fr/seqanal/interfaces/dsc-simple.html) was used for predicting the secondary structure for proteins. The KIX domains in CBP-type HAT proteins were further searched against a database of position-specific-scoring-matrices representing conserved structural domains (3D-pssm at http://www.sbg.bio.ic.ac.uk/∼3dpssm/) to find similarity with the known KIX domain structure.

RESULTS

Identification of HDAC and HAT proteins and alternative transcripts encoded by the Arabidopsis genome

The Arabidopsis genome sequence was searched for homologs of known HDAC and HAT proteins as described in the Materials and Methods. A total of 16 Arabidopsis HDAC genes and 12 HAT genes were identified (Table 1). Of the 16 HDACs, 10 belong to the RPD3/HDA1 superfamily and were named with the symbol HDA, four belong to the HD2 family and were given the name HDT (‘HD-tuins’), and two belong to the SIR2 family and were named with the symbol SRT. Two additional members of the HDA family were found that have partial HDAC domains. Of the 12 HATs, five belong to the GNAT/MYST superfamily and were named with the symbol HAG, five belong to the CBP family and were named with the symbol HAC, and two belong to the TAF_II250 family and were named with the symbol HAF.

Table 1. Genes encoding HAT and HDAC homologs in Arabidopsis.

HDAC and HAT gene family	Arabidopsis gene name (synonym)	MIPS accession no.	BAC clone and genomic locus	Chromosome
RPD3/HDA1	HDA2	At5g26040	T1N24.9	V
	HDA5	At5g61060	MAF19.7	V
	HDA6 (AtRPD3B)	At5g63110	MDC12.7	V
	HDA7	At5g35600	K2K18.5	V
	HDA8	At1g08460	T27G7.7	I
	HDA9	At3g44680	T18B22.80	III
	HDA10a	At3g44660	T18B22.60	III
	HDA14	At4g33470	F17M5.230	IV
	HDA15	At3g18520	MYF24.23	III
	HDA17a	At3g44490	F14L2	III
	HDA18	At5g61070	MAF19.8	V
	HDA19 (AtRPD3A)	At4g38130	F20D10.250	IV
HD2	HDT1 (AtHD2A)	At3g44750	T32N15.8	III
	HDT2 (AtHD2B)	At5g22650	MDJ22.7	V
	HDT3	At5g03740	F17C15.160/MED24.1	V
	HDT4	At2g27840	F15K20.6	II
SIR2	SRT1	At5g55760	MDF20.20	V
	SRT2	At5g09230	T5E8.30	V
GNAT	HAG1 (atGCN5)	At3g54610	T14E10.180	III
	HAG2	At5g56740	MIK19.19	V
	HAG3	At5g50320	MXI22.3	V
MYST	HAG4	At5g64610	MUB3.13	V
	HAG5	At5g09740	F17I14.70/MTH16.20	V
CBP	HAC1	At1g79000	YUP8H12R.3	I
	HAC2	At1g67220	8 F1N21.4	I
	HAC4	At1g55970	F14J16.27/T6H22.23	I
	HAC5	At3g12980	MGH6.9	III
	HAC12	At1g16710	F17F16.21	I
TAFII250	HAF1	At1g32750	F6N18.20	I
	HAF2	At3g19040	K13E13.16	III

^aGenes with partial HDAC domains.

Consensus gene splicing models were first developed by comparison of several computationally determined models. Because computational methods do not predict all splice sites correctly, cDNA sequences were generated from EST clones and RT–PCR products (see Materials and Methods). In addition to revising splicing models, the cDNA sequence analysis detected multiple splicing products for three HDAC genes (HDA2, HDA15 and SRT2) (Fig. 1). Revised coding sequences, predicted proteins and alternative splicing products are available at the Plant Chromatin Database, ChromDB (http://www.chromdb.org).

Figure 1.

Alternative splicing of HDA2, HDA15 and SRT2. Sequence coordinates indicate the position of exons within the unspliced transcripts relative to the start of the ‘alt1’ RT–PCR product sequences. The approximate location of predicted protein domains and conserved amino acid motifs is marked by brackets, their Pfam accessions are listed here. HDAC, the histone deacetylase domain (PF00850); SIR2, the multidomain core [including the conserved GAG, NID and CYS motifs (89) (PF02146)]; zf-RanBP, Ran-binding protein zinc finger (PF00641); (NLS), sequence similar to the bipartite nuclear localization sequence. SRT2alt3 and alt6 completely lack exon 2, which contains the predicted translation initiation codon of SRT2alt1. The nearest downstream ATG codon for translation initiation from SRT2alt3 and alt6 is located at position 446 of the unprocessed transcript. A consequence of translation initiation at this position would be a protein lacking a putative nuclear localization signal. Alternative splicing of exon 2 in SRT2alt2 and alt5 removes 39 nt of the 5′-UTR. Alternative splicing of exon 5 in SRT2alt4, alt5, and alt6 introduces a premature nonsense codon within the conserved multi-domain SIR2 core; alternative splicing at this position is conserved between SRT2 and a putative ortholog in tomato represented by ESTs 12635152 and 12625887.

The phylogenetic and domain analyses presented here are based on alternative products designated ‘alt1’ (Fig. 1), each of which is predicted to encode intact, conserved HDAC domains. The HDAC domain is disrupted in alternative transcripts produced by HDA2 and HDA15. SRT2 produced six alternative transcripts via different combinations of the same splice sites, affecting a putative nuclear localization signal and the SIR2 domain. The alternative splice site in the SIR2 domain appears to be evolutionarily conserved because it also occurs in a putative ortholog in tomato. Alternative splicing in the 5′-untranslated region (5′-UTR) of SRT2alt2 and alt5 could affect translation efficiency or mRNA stability. Details are presented in Figure 1.

The RPD3/HDA1 superfamily of HDACs

A total of 10 representatives possessing the complete HDAC domain (Pfam designation PF00850) that defines the RPD3/HDA1 superfamily were identified in Arabidopsis (Table 1). Two additional predicted proteins, HDA10 and HDA17, were found that possess only the 30 and 40 C-terminal amino acids, respectively, of the HDAC domain.

Sequence similarity searches of a variety of eukaryotic and prokaryotic genomes, as well as other sequences available in public databases (including ESTs), led to the identification of a total of 72 RPD3/HDA1 superfamily protein sequences (including 10 in Arabidopsis) that possess an intact HDAC domain. For 80% of these sequences, the 300 amino acid HDAC domain constitutes more than half of the protein. For the remaining 20% of the sequences, additional sequences were present. Searching these larger proteins using Pfam (version 6.5) did not reveal any additional domains, although there is a possibility of the presence of additional domains that have not yet been identified.

Figure 2 shows an unrooted phylogenetic tree illustrating the relationships among the 72 RPD3/HDA1 superfamily proteins (listed in Table 2), produced by aligning their HDAC domains (for double-domain proteins each domain was analyzed separately). The analysis in Figure 2 is based on a mixture of both predicted and experimentally determined protein sequences. In order to confirm these results, the analysis was also performed using only experimentally derived sequences (i.e. those confirmed by cDNA sequences). The clustering patterns and the bootstrap support for these patterns were similar to those shown in Figure 2 (data not shown). The RPD3/HDA1 superfamily, represented by these 76 domain sequences, is divided into two major clades based on a strongly supportive bootstrap value (85%). These clades, shown in Figure 2 as two lightly shaded ovals, include three classes of eukaryotic proteins: Classes I and II, both of which have been reported previously based on a smaller number of sequences (56,57), and a new class of proteins, Class III. These Class III proteins include a recently cloned and characterized human HDAC11 (58).

Figure 2.

Phylogenetic analysis of the RPD3/HDA1 HDAC superfamily. Unrooted neighbor-joining tree of 76 RPD3/HDA1 superfamily sequences includes four double-domain sequences with each domain being analyzed separately. Confidence levels of the branching patterns are: filled circle, excellent support (>99% of bootstrap replicas); empty square, good or >70%; empty circle, majority support or >50%. Eukaryotic gene names and sequence accession numbers are listed in Table 2. The plant proteins are highlighted in bold and the three eukaryotic classes are represented in gray shaded ovals. Prokaryotic genes are represented by Acu (acetoin utilization proteins) or by Aph (acetylpolyamine aminohydrolase proteins). All the proteins have abbreviated species names as prefix. The proteins and their accession numbers are identified in Table 2. Abbreviations for species are: Aeropyrum pernix (ap), Arabidopsis thaliana (at), Archaeoglobus fulgidus (af), Aquifex aeolicus (aa), Aspergillus nidulans (an), Bacillus halodurans (bh), Bacillus subtilis (bs), Caenorhabditis elegans (ce), Deinococcus radiodurans (dr), Drosophila melanogaster (dm), Glycine max (gm), Halobacterium sp. NRC-1 (halo), Homo sapiens (hs), Leishmania major (lm), Mesembryanthemum crystallinum (mc), Methanobacterium thermoautotrophicum (mt), Methanococcus jannaschii (mj), Mus musculus (mm), Mycoplana ramose (mr), Neisseria meningitides (nm), Oryza sativa (os), Plasmodium falciparum (pf), Pseudomonas aeruginosa (ps), Pyrococcus abyssi (pa), Pyrococcus horikoshii (ph), Saccharomyces cerevisiae (sc), Schizosaccharomyces pombe (sp), Staphylococcus xylosus (sx), Streptomyces coelicolor (stco), Synechococcus PCC7002 (syp), Synechocystis PCC6803 (syn), Tetrahymena thermophila (tt), Vibrio cholerae (vc), Zea mays (zm).

The two major clades include proteins from both prokaryotes and eukaryotes. The rightmost clade includes acetylpolyamine aminohydrolase proteins from multiple species of Archaea and bacteria, suggesting that HDAC proteins in this clade could be derived from these prokaryotic proteins. The leftmost clade includes acetoin-utilizing proteins from bacteria (but no Archaea sequences), suggesting that the HDAC proteins in this clade could have originated from these bacterial proteins. Proteins from other lower eukaryotic organisms, including Plasmodium falciparum, T.thermophila and L.major, were present in only the leftmost clade of Figure 2. This evolutionary link between the prokaryotic proteins and the HDACs is also evident at the level of enzymatic activity. HDACs and acetylpolyamine aminohydrolases catalyze the removal of an acetyl group from acetylated aminoalkyls by cleaving an amide bond and reconstituting the positive charge on the substrate; acetoin utilization proteins catalyze deacetylation of acetoin (32).

Class I proteins. The total number of Class I proteins found in the Arabidopsis genome is similar to the numbers found in other sequenced genomes (Table 4). The four Arabidopsis proteins lie within a cluster comprised of S.cerevisiae RPD3p and several animal Class I proteins with good bootstrap support (70%) (Fig. 2). Three of the Arabidopsis proteins, along with other plant proteins, group into two branches forming clusters A and B, each with excellent bootstrap support (100%). The proteins in cluster A (including Arabidopsis HDA19) are 73–80% identical at the amino acid level and may comprise an orthologous group. The proteins in cluster B (which includes Arabidopsis HDA6 and HDA7) are somewhat more divergent than the proteins in A (58–74% identical at the amino acid level) cluster. The strongly supported separation of clusters A and B suggests the possibility of functional diversification. Because both clusters contain dicot and monocot proteins, they would seem to have originated by gene duplication predating divergence of the monocot and dicot lineages. Immunological data indicates zmRPD3 (cluster A) and zmHD1b-II (cluster B) to be associated with human Rbap46/48 like proteins (59) found in the NuRD and SIN3 HDAC complex (60).

Table 4. Summary of HDAC and HAT homologs found in plants, fungi and animals.

Homology groups	Plants	Fungi		Animals
	At	Sc	Sp	Dm	Ce
HDAC homology groups
RPD3/HDA1 family (HDA genes)
Class I	4	3	2	2	3
Class II	3	1	1	2	4
Class III	1	0	0	1	1
Unclassified	2	1	0	0	0
HD2 family (HDT genes)	4	0	0	0	0
SIR2 family (SRT genes)
Class I	0	5	3	1	1
Class II	1	0	0	1	2
Class IV	1	0	0	2	1
Total HDAC homologs	16	10	6	9	12
HAT homology groups
GNAT-MYST superfamily (HAG genes)
GNAT family
GCN5	1	1	1	1	1
ELP3	1	1	1	1	1
HAT1	1	1	1	1	1
HPA2	0	2	1	0	0
MYST family	2	3	2	5	4
CBP family (HAC genes)	5	0	0	1	1
TAFII250 family (HAF genes)	2	1	1	1	1
Total HAT homologs	12	9	7	10	9

The genes and their accession numbers are identified in Table 2. Organism abbreviations as follows: Arabidopsis thaliana (At), Saccharomyces cerevisiae (Sc), Schizosaccharomyces pombe (Sp), Caenorhabditis elegans (Ce), Drosophila melanogaster (Dm).

One of the Arabidopsis class I proteins, HDA9, is highly similar at the nucleotide level to HDA10 and HDA17, both of which possess an incomplete HDAC domain. HDA10 lies ∼11 kb from HDA9, and the interval between these two genes contains an ORF annotated in GenBank as encoding a ‘disease-resistance-like’ gene. HDA17 lies on a neighboring BAC clone, adjacent to a second copy of this ‘disease-resistance-like’ gene, suggesting that HDA10 and HDA17 were derived from HDA9 by sequence rearrangements that duplicated part of HDA9 and its flanking sequences. These events appear to be relatively recent in evolution, considering that the homologous regions of these three genes are 97% identical at the nucleotide level. Genetic and biochemical analyses will be required to determine whether HDA10 and HDA17 possess some function, perhaps related to that of HDA9, or are non-functional pseudogenes.

Class II proteins. The Arabidopsis genome possesses three Class II proteins (designated HDA5, HDA15 and HDA18), a total similar to that found in other sequenced genomes (Table 4). A subset of Class II proteins found in humans, mice, C.elegans and D.melanogaster are ‘double-domain’ proteins, i.e. they possess two tandem HDAC domains separated by a small, but variable, spacer region. In human and mouse proteins, each domain has been found to be an independently functional catalytic domain (57). Double-domain proteins have not been found in either S.cerevisiae or S.pombe, each of which has a single Class II protein with a single domain. Likewise, the Arabidopsis genome does not contain any double-domain Class II proteins. Recently, HDAC6, a human double-domain protein, has been shown to be a cytoplasmic tubulin deacetylase, not an HDAC (61).

Class II proteins are more divergent in sequence than are Class I proteins, resulting in longer, more poorly supported branches (Fig. 2), and making it impossible to definitively classify orthologous and paralogous groups. Two clusters of plant Class II proteins (indicated by brackets in Fig. 2) can be identified by phylogenetic analysis. HDA5 and HDA18 appear to be more closely related to the double-domain proteins from animals than to HDA15, and so may act on proteins other than histones. Sequence analysis revealed the presence of putative nuclear export signals in HDA5 and HDA18. Similar nuclear export signals in human and mouse class II proteins are known to be involved in shuttling these proteins between an active state in the nucleus and an inactive, phosphorylated state in the cytoplasm (62,63). Interestingly, HDA15 contains a RanBP zinc-finger domain. Such domains have been implicated in nucleocytoplasmic transport and nuclear envelope localization (64).

HDA5 and HDA18 occur immediately adjacent to each other on chromosome V, consistent with a gene duplication event. Their encoded proteins share 84% identity, mostly in the HDAC domain. The coding sequences of these genes share the same splice site positions throughout the HDAC domain which lies toward the 5′ end of the transcript, whereas their C-terminal regions are unrelated to each other. The C-terminal region of HDA5 does not possess any known protein domains, whereas that of HDA18 is predicted to encode a predominantly α-helical domain. This putative domain carries a leucine zipper motif and is similar to structural domains found in filamentous proteins, including coiled-coil dimers and two S.pombe proteins (cut3 and cut14) that are required for chromosome condensation and segregation (65).

A third gene (At5g61050), with partial homology to HDA5 and HDA18 outside the HDAC domain, was also found immediately downstream of HDA5 (Fig. 3). HDA5, HDA18 and At5g61050 are located within a 10 kb segment on chromosome V. The five exons of At5g61050 share similarity with some exons of HDA5 and HDA18, however, the region encoding the HDAC domain is missing in At5g61050, so it is not classified as an HDAC protein. The high degree of sequence identity in homologous regions of the three genes suggests two recent duplications of HDA5 to produce the progenitors of HDA18 and At5g61050. The duplication was apparently followed (or accompanied by) an internal deletion in one gene copy to form At5g61050 and acquisition of repeated sequences elements encoding an α-helical domain in the other gene copy to form HDA18. This gene duplication event is not shared by all the angiosperms, and appears to be unique to a lineage within the dicots including Arabidopsis. Whether this event resulted in diversification of function of Class II proteins remains to be determined.

Figure 3.

Schematic representation of the exon–intron and domain organization of the HDA18-HDA5-At5g61050 gene cluster on chromosome V. Coordinates indicate the position of the start and stop codons of the three genes in the P1 clone MAF19 (accession no. AB006696). The approximate location of predicted protein domains is marked by brackets. The dotted line indicates the missing HDAC domain in At5g61050. NES, nuclear export signal. Arrows indicate nucleic acid sequence repeats in HDA18.

Class III: a new class of proteins in the RPD3/HDA1 superfamily. A major finding of our analysis is a new class of HDAC proteins, which we designate Class III, represented in Arabidopsis by HDA2. Class III includes predicted proteins HDA403 from D.melanogaster, HDA308 from C.elegans and HDAC11, an EST contig from humans (Fig. 2) that has been recently identified (58). These proteins are conserved at the amino acid level, being 45% or more identical in pairwise sequence alignments. Additional members of this class were found in the EST database, but were not included in our analysis because their HDAC domains were incomplete. Class III proteins are a part of a cluster that includes three bacterial sequences encoding acetoin utilization proteins (vcAcu and drAcu) and a cyanobacteria glutamine synthetase protein (synGln) (Fig. 2), with good bootstrap support (99%). The presence of a well supported cluster of diverse proteins is consistent with a novel function for class III HDAC proteins in higher eukaryotes, possibly of bacterial origin. No class III proteins were detected in fungal genomes.

Multiple sequence alignments of Classes I, II and III proteins identified conserved motifs within the HDAC domain, with some amino acids common to all HDAC classes and others unique to a particular HDAC class (Fig. 4). A conserved but distinct pattern of amino acids for Class III proteins is evident, providing additional support for a novel biological function for these proteins.

Figure 4.

Class III proteins in the RPD3/HDA1 protein superfamily have distinct motifs in the HDAC domain. Alignment of the HDAC domain of Arabidopsis HDA2 protein with human HDAC11, D.melanogaster HDA403 and C.elegans HDA308. These proteins and their accession numbers are identified in Table 2. Shading was done based on degree of identity or conservation using the Genedoc program. Also shown below the multiple sequence alignment is a second alignment of consensus motifs found in the proteins in all the three classes of HDACs identified in Figure 2. These motifs represent the most highly conserved sequence positions in the HDAC domain. The consensus motif for each class was identified by generating a logo sequence. Each class of proteins is indicated by a consensus of the sequences in that class: black boxes, positions conserved across all three classes; underlined, positions highly conserved within a class; upper case letters, 98% conserved within a class; lower case letters, 60% conserved within a class; X, variable positions. The amino acid positions in each sequence class refer to the location of these motifs in Arabidopsis HDA19 (Class I), HDA5 (Class II) and HDA2 (Class III) proteins.

Unclassified proteins. The Arabidopsis genome encodes two additional HDAC proteins in the RPD3/HDA1 superfamily, HDA8 and HDA14. Although these proteins fall into the same major clade as Class II proteins, they do not cluster with them (Fig. 2). Instead, they are present in a poorly supported group of highly diverse proteins that includes acetylpolyamine aminohydrolases from the Archaea, as well as S.cerevisiae HDAC protein Hos3p. The low sequence similarity between S.cerevisiae Hos3p and Arabidopsis HDA8 and HDA14 and the poor bootstrap support for this grouping indicates that these proteins are not closely related. Searches of existing genome and EST databases, including plant sequences, using Hos3p, HDA8 and HDA14 as query sequences did not identify any additional proteins in this group.

To determine whether the sequences from Archaea and bacteria influence the classification of these eukaryotic proteins, the tree was regenerated without these sequences. In the resulting tree, S.cerevisiae Hos3p and Arabidopsis HDA14 protein moved into the class II cluster, but HDA8 did not. This test revealed that S.cerevisiae Hos3p and Arabidopsis HDA14 can not be assigned to any definitive cluster, but appear to be relatives of Class II proteins. Arabidopsis HDA8 seems to be more closely related to prokaryotic acetylpolyamine aminohydrolase proteins than to Class II; it is possible that this protein might have acetylpolyamine deacetylating activity or other deacetylating activity rather than histone deacetylation activity. In the motif analysis of all three HDAC classes shown in Figure 4, Hos3p, HDA8 and HDA14 share the conserved amino positions of Class II proteins, corresponding with their location in the same major clade as Class II proteins.

The HD2 family: unique to plants

Plants possess a family of HDAC proteins, the HD2 family, which is not found in animals or fungi (40) and is distantly related to cis–trans isomerases found in insects, S.cerevisiae and parasitic apicomplexans (42). Using maize HD2 as a query, four candidate proteins, HDT1, HDT2, HDT3 and HDT4, were identified in the Arabidopsis proteome (Table 1). The conserved N-terminus of these proteins contains the HD2-type HDAC domain of approximately 100 amino acids. The proteins are comprised of a conserved N-terminal domain, a central acidic domain and variant C-terminal domain. Two of these proteins, HDT1 and HDT2, have been analyzed in a recent paper showing that antisense silencing of HDT1 results in aborted seed development (41). A sequence comparison of Arabidopsis and maize HD2-type proteins has been made by Dangl et al. (66).

Plant EST sequence databases were searched to find HD2-type HDAC proteins in other plant species (listed in Table 2 and Fig. 5). Comparison of the HDAC domains of these proteins revealed a series of highly conserved motifs within the HDAC domain. A phylogenetic analysis of the nucleotide sequences encoding these conserved motifs in the HDAC domains was performed, producing the tree shown in Figure 5. A similar analysis using protein sequences produced a tree with similar topology and the same major features although with varying but somewhat lower bootstrap support than the DNA tree. This analysis permits two general observations to be made concerning the evolution of the HD2 gene family in plants. First, dicot and monocot sequences are separated into two distinct clades strongly supported by bootstrap analysis (98%), indicating that a single HD2 gene in the ancestor of monocots and dicots gave rise to all HD2 proteins in these groups. Secondly, the clustering pattern in dicots is consistent with a gene duplication event occurring before the diversification in dicot evolution that produced the families Solanaceae (tomato and potato), Malvaceae (cotton) and Aizoaceae (ice plant), although this conclusion is only weakly supported by bootstrap analysis (<50%). More recent duplications that are strongly supported by bootstrap analysis are also evident in several species [e.g. Arabidopsis HDT1 and HDT2 (100%), barrel medic HDT1301 and HDT1302 (90%), and maize HD2a, HD2b and HD2c (100%)]. It will be interesting to determine whether the considerable amount of genetic diversification of the HD2 family has been accompanied by functional diversification.

The SIR2 family of HDACs

Plants possess representatives of the SIR2 family of NAD-dependent HDAC proteins, known as sirtuins. Sirtuins occur across a wide range of organisms, including prokaryotes, fungi, plants and animals and are defined by a 175 amino acid domain (Pfam designation PF02146) comprised of a series of conserved motifs. Based on variation in this domain, the eukaryotic proteins fall into four main classes (31). A fifth class is present in some prokaryotes, but most prokaryotic sirtuins fall into Classes II and III (31). A search of the Arabidopsis genome identified two SIR2 family proteins, SRT1 and SRT2, fewer than are found in fungi and animals (Table 4).

In order to identify additional plant sequences for use in a phylogenetic analysis, Arabidopsis SRT1 and SRT2 proteins were used as queries of plant EST collections, revealing six related proteins (Table 2 and Fig. 6). Phylogenetic analysis of all plant SIR2 homologs and homologs from representative species in the Frye (31) classification of SIR2-like proteins is shown in Figure 6. Of the four classes of SIR2 proteins, plant proteins are only found within divergent plant lineages in Classes II or IV. Both classes contain plant and animal proteins but no fungal proteins (Table 4). Class IV includes two divergent animal lineages represented in flies and humans. All plant Class IV proteins cluster in a single, less divergent lineage associated with one of these animal lineages. Both plants and animals have a single lineage of Class II proteins. No plant proteins cluster with proteins of Class I, which includes all five S.cerevisiae sirtuins, as well as homologs in animals and S.pombe.

Figure 6.

Phylogenetic analysis of plant SIR2 proteins. Unrooted neighbor-joining tree of 31 SIR2-related proteins shows the four previously identified classes of SIR2 proteins. The two plant protein clusters are highlighted in bold. Confidence levels of the branching patterns are: filled circle, excellent support (>99% of bootstrap replicas); empty square, good or >70%; empty circle, majority support or >50%. The genes and their accession numbers are identified in Table 2. Abbreviations for species are: Arabidopsis thaliana (at), Caenorhabditis elegans (ce), Drosophila melanogaster (dm), Homo sapiens (hs), Lycopersicon esculentum (le), Medicago truncatula (mt), Oryza sativa (os), Saccharomyces cerevisiae (sc), Schizosaccharomyces pombe (sp), Triticum aestivum (ta), Zea mays (zm).

Representation of HATs in the GNAT/MYST superfamily in the Arabidopsis genome

In the GNAT/MYST superfamily of HAT proteins, GNAT proteins are defined by the presence of a HAT domain (Pfam designation PF00583) which is comprised of four motifs, A–D, whereas MYST proteins possess only the A motif of the HAT domain (22).

The GNAT family is generally considered to be comprised of four subfamilies designated GCN5, ELP3, HAT1 and HPA2. The HPA2 subfamily has in vitro histone acetylation activity (67), but it is not yet known whether these proteins play any role in the control of gene expression. In the Arabidopsis genome, we identified a single homolog of each of the GCN5, ELP3 and HAT1 subfamilies (HAG1, HAG3 and HAG2, respectively) and no homolog of the HPA2 subfamily. HAG1 (atGCN5) and its associated adaptor proteins [similar to yeast SAGA complex (22)] in Arabidopsis have been known for their involvement in cold regulated gene expression (68). Searches of the S.cerevisiae, S.pombe, D.melanogaster and C.elegans genomes, as well as the nearly complete human genome, also identified a single representative of the GCN5, ELP3 and HAT1 subfamilies in each; only fungi were found to possess the HPA2 subfamily (Table 4). Thus, Arabidopsis appears to have the same representation of GNAT family HATs as do animals, suggesting that the plant proteins may form complexes similar to those formed in yeast and animals (69).

The Arabidopsis genome was found to encode two MYST family proteins, HAG4 and HAG5. Fungal genomes were found to have two to three, and animal genomes four to six, MYST family proteins. Thus, the number of plant MYST family representatives is within the range found in other eukaryotic organisms, though at the lower end of this range, and below the numbers found in animals (Table 4).

The CREB-binding protein (CBP) family of HATs

The CBP family of HAT proteins is comprised of large, multi-domain proteins (Fig. 7A) which, until recently, had been reported only in animals. The histone acetylation domain of the CBP family is unrelated to that of the GNAT/MYST superfamily; we refer to this as the CBP-type HAT domain. The Arabidopsis genome encodes five CBP-type HAT domain proteins (HAC1, HAC2, HAC4, HAC5 and HAC12), whereas the number of CBP proteins predicted in animals is only one to two (Table 4). The absence of the CBP family in fungi suggests that this type of protein was lost during the evolution of fungi.

Figure 7.

Domain architecture of the CBP-type HAT family and phylogenetic analysis of their HAT domains. (A) Schematic representation of the domain organization of Arabidopsis and animal CBP proteins. Different domains are identified by different symbols and colors, and are shown at their approximate relative location in the protein sequence. The protein lengths are listed on the right. //, indicates position of extra sequence; /, indicates more sequence at the N- and C-terminus. The CBP-type HAT domain is conserved throughout its length between plants and animals, however, in plants a ZZ-type zinc finger domain is inserted near the C-terminus of the HAT domain. The Pfam accession number for the domain profiles is indicated in parentheses. (B) Unrooted neighbor-joining tree of 10 CBP-type HAT proteins based on the HAT domain. Distinct animals and Arabidopsis clusters are shown by two shaded ovals. Confidence levels of the branching patterns are: filled circle, excellent support (>99% of bootstrap replicas). The genes and their accession numbers are identified in Table 2. Abbreviations for species are as follows: Arabidopsis thaliana (at), Caenorhabditis elegans (ce), Drosophila melanogaster (dm), Homo sapiens (hs), Mus musculus (mm).

Phylogenetic analysis of the plant and animal CBP-type HAT domains indicates an early divergence of HAC2 from the lineage leading to the other four Arabidopsis HAC proteins (Fig. 7B). Consistent with this divergence, in vitro assays of HAC2 did not detect any HAT activity, whereas it was readily detected for HAC1 (70). Similarly, HAC4 has diverged significantly from HAC1, HAC12 and HAC5. Interestingly, the HAT domains of human and mouse CBP proteins are 96% identical, whereas the two closest Arabidopsis CBP paralogs (HAC1 and HAC12) are only 90% identical in the HAT domain.

The domain architecture of CBP-type HAT proteins differs between plants and animals (Fig. 7A) in four major respects. (i) Bromodomains. As was noted also by Bordoli et al. (70), plant CBP-type HATs lack a bromodomain. The role of the bromodomain in the animal proteins is to bind acetylated histones (71). The lack of a bromodomain in the plant proteins suggests that these proteins utilize a different domain to perform this function or that another bromodomain protein acts as a bridge between acetylated histones and CBP-type HATs. (ii) KIX domains. All animal CBP-type HAT proteins possess a KIX domain by which they bind the nuclear factor CREB (72). Bordoli et al. (70) reported that the Arabidopsis proteins lack KIX domains. However, we found a weakly defined KIX-like domain in four of the five Arabidopsis proteins (Fig. 7A). The KIX domain is known to be comprised of three α-helices joined by connecting loops (73). The plant KIX-like domains from HAC1, HAC5 and HAC12 have three α-helices with about the same spacing as in the animal KIX domain, whereas HAC4 has two α-helices. A search of all four plant KIX-like sequences against a database of position-specific-scoring-matrices representing conserved structural domains (3D-pssm) produced a match with the matrix representing the KIX domain. Interestingly, the location of the KIX domain relative to the TAZ-type zinc finger domain in the animal proteins differs from the location of the KIX-like domain relative to this domain in the plant proteins (Fig. 7A). (iii) Zinc finger domains. ZZ and TAZ types of zinc finger domains are found only in CBP-type proteins and are known to mediate protein–protein interactions with transcription factors (74). Animal CBP-type proteins have one ZZ-type zinc-finger domain located near the C-terminal end of the CBP-type HAT domain, whereas all the plant proteins have two such domains, one of which lies within the HAT domain. Both plant and animal proteins possess two TAZ-type zinc fingers, one on each side of the HAT domain. The N-terminal TAZ-type domain is located at a greater distance from the HAT domain in the animal proteins than in the plant proteins. (iv) Glutamine-rich regions. Animal CBP-type HATs possess an extensive glutamine-rich region near the C-terminus, which harbors the binding site for the unrelated mammal-specific HATs, SRC-1 and ACTR (75,76). Plant proteins lack such a C-terminus (70) (Fig. 7A), which is not particularly surprising given that plants lack this family of HATs (22), which we have confirmed by searching the Arabidopsis genome.

The TAF_II250 family of HAT proteins

The human TAF_II250 protein is a subunit of transcription factor IID (TFIID) (77) and has a HAT domain unrelated to the GNAT/MYST and CBP-type HAT domains. Using animal protein sequences as queries, two Arabidopsis TAF_II250 homologs were identified and designated HAF1 and HAF2 (Table 1). These long predicted proteins are 72% identical to each other at the amino acid level. A similar search against the complete C.elegans, D.melanogaster, S.pombe and S.cerevisiae genomes, and the nearly complete human genome, identified only one homolog in each organism. Hence, Arabidopsis is unusual in encoding two predicted TAF_II250 HAT proteins.

The human and D.melanogaster proteins have a 260 amino acid long TAF_II250-type HAT domain (28). A multiple sequence alignment revealed the presence of a domain in the Arabidopsis and C.elegans proteins that is similar in length to the human and D.melanogaster TAF_II250 HAT domains. This domain is 45–75% identical among this group of organisms. A similar type of HAT domain in S.cerevisiae is shorter in length, lacking amino acids at the C-terminus of the domain, but still has HAT activity (28). Thus, the plant proteins are more similar to the animal proteins in this respect than to the fungal proteins.

The overall domain architecture of TAF_II250-type proteins in plants, animals and fungi is presented in Figure 8 and shows three interesting features. (i) In addition to the TAF_II250-type HAT domain, the human and D.melanogaster proteins have two bromodomain copies on the C-terminal side of the HAT domain, whereas the Arabidopsis proteins possess only a single bromodomain in this region. (ii) A zinc-finger-type C₂HC domain is located at an approximately equal distance downstream of the HAT domain in each of the seven sequences, presumably with a role in DNA binding or protein–protein interactions. (iii) A conserved ubiquitin signature at the N-terminal side of the HAT domain was found in each Arabidopsis protein, but not in the animal or fungal proteins. No other Pfam ubiquitin-associated domain was found in the animal or fungal proteins. In D.melanogaster, the region of TAF_II230 responsible for ubiquitin-conjugating activity for histone H1 overlaps the TAF_II250-type HAT domain (78), and these regions are presumably present in the highly conserved TAF_II250-type HAT domains in the Arabidopsis proteins.

Figure 8.

Domain architecture of the TAF_II250 proteins. A schematic representation is shown of the domain organization of Arabidopsis and animal TAF_II250 proteins aligned by the N-terminus of the HAT domain. Different domains are identified by different symbols and colors, and are shown at their approximate relative locations in the protein sequences. The protein lengths are listed on the right. Pfam accession numbers for the domain profiles are indicated in parentheses underneath the alignment. The sequences and their accession numbers are identified in Table 2. Abbreviations for species are: Arabidopsis thaliana (at), Caenorhabditis elegans (ce), Drosophila melanogaster (dm), Homo sapiens (hs), Saccharomyces cerevisiae (sc), Schizosaccharomyces pombe (sp).

Arabidopsis bromodomain proteins

Because of the disparity in number and occurrence of bromodomain between plant and animal HAT proteins, we performed a preliminary search for all bromodomain- containing proteins in Arabidopsis using the bromodomain HMM profile from Pfam. Twenty-nine Arabidopsis bromodomain proteins were found (Table 3), all of which had only a single bromodomain. Although the majority of bromodomain proteins in fungi and animals also possess a single bromodomain, many have from two to five bromodomains (79). Thus, plants lack multi-bromodomain proteins.

Table 3. Arabidopsis bromodomain proteins and associated domains within these proteins.

Proteins	Number of bromodomains	Associated domains	GenBank protein accession no.	MIPS accession no.
BRD1	1	AT hook	AAF80635.1	At1g20670
BRD2	1	–	AAF16663.1	At1g76380
BRD3	1	Myb-DNA-binding domain	AAC16089.1	At2g44430
BRD4	1	Myb-DNA-binding domain	AAB71473.1	At1g52110
BRD5	1	–	AAG50696.1	At1g58025
BRD6	1	Myb-DNA-binding domain	CAB75926.1	At3g60110
BRD7	1	–	CAB67626.1	At3g57980
BRD8	1	AAAAtpase domain	AAF29398.1	At1g05910
BRD9	1	Myb-DNA-binding domain	AAB88641.1	At2g42150
BRD10	1	–	AAD03360.1	At2g15030
BRD11	1	4 WD40 repeats	BAB09913.1	At5g49430
BRD12	1	5 WD40 repeats	AAC62845.2	At2g47410
BRD13	1	–	BAB10578.1	At5g55040
CHR2	1	ATP binding and helicase domain	AAC62900	At2g46020
HAF1	1	TAFII250 HAT domain, C2HC zinc finger	AAF25977	At1g32750
HAF2	1	TAFII250 HAT domain, C2HC zinc finger	BAB01700	At3g19040
GTE1	1	ET domain	AAC12830	At2g34900
GTE2	1	ET domain	CAB89388	At5g10550
GTE3	1	ET domain	AAF18720	At1g73150
GTE4	1	ET domain	AAF80220	At1g06230
GTE5	1	ET domain	AAF97259	At1g17790
GTE6	1	ET domain	CAC07919.1	At3g52280
GTE7	1	ET domain	BAA98182.1	At5g65630
GTE8	1	ET domain	BAB02121.1	At3g27260
GTE9	1	ET domain	CAB87766.1	At5g14270
GTE10	1	ET domain	BAB10737.1	At5g63330
GTE11	1	ET domain	AAF01563.1	At3g01770
GTE12	1	ET domain	BAA97526.1	At5g46550
HAG1	1	HAT domain	AAB92257	At3g54610

Bromodomain proteins exist in diverse classes defined according to the presence of other domains in those proteins (80). We performed a domain analysis of the 29 Arabidopsis proteins for other Pfam domains. Unlike fungi or animal bromodomain proteins that commonly possess zinc fingers (81), none of the Arabidopsis bromodomain proteins possess any type of zinc finger, such as a PHD domain, with the exception of the C₂HC zinc knuckle observed in HAF1 and HAF2. As noted previously, bromodomains are often associated with certain other domain classes in other organisms, whereas the same associations are not observed in Arabidopsis proteins. In the case of CBP-type HATs, animal proteins contain both a bromodomain and multiple zinc fingers, whereas Arabidopsis CBP-type HATs contain only zinc fingers (Fig. 7A). Another interesting difference is that an animal homolog of fly Trithorax-related proteins (mouse protein AAK26242) has a bromodomain associated with a SET domain, whereas no bromodomain protein in Arabidopsis contains a SET domain. Thus, the utilization of bromodomains differs not only in HATs, but also in other types of chromatin protein, in plants as compared to animals and fungi.

DISCUSSION

The Arabidopsis genome is predicted to encode 16 HDAC and 12 HAT proteins, which is somewhat more than the number of such genes found in other sequenced eukaryotic genomes (Table 4). The distribution among different homology groups of HDACs and HATs in Arabidopsis differs from that in fungi and animals in several respects, as summarized in Table 4. Phylogenetic and domain analyses of these proteins predict that some have functionally diversified during plant evolution, whereas others appear to have conserved the functions of their ancestral homologs. In addition, the observed alternative mRNA splicing of three HDAC genes suggests the possibility of further functional diversification of these protein families and a complex relationship between gene number and the actual number of gene products encoded within plant genomes, as also appears to be the case for the human genome (82).

The most obvious indication of diversification of histone acetylation/deacetylation functions in plants as compared to animals and fungi is that plants possess a unique family of HDACs, the HD2 gene family (66). Because no homologs of HD2 are found in any animal or fungal genome, these proteins could serve a novel plant function or could provide a function similar to one carried out by a different type of HDAC in animals and fungi. Our phylogenetic analysis is consistent with a greater degree of functional diversification in the HD2 family in dicots than monocots. This analysis suggests that a gene duplication event may have occurred early in dicot evolution and that further diversification has occurred in the lineage leading to Arabidopsis, suggesting functional diversification of the HD2 subfamily.

We found that the SIR2 family is under-represented in plants as compared to fungi and animals. It is possible that the HD2 family has taken over some of the function(s) of sirtuins. Another possibility is that alternative splicing has provided added diversity of sirtuin functions. Plants possess two classes of sirtuins that are also represented in animals, but not in fungi. The SIR2 family has major biological significance including determining the life span of S.cerevisiae cells and aging in animals (37,38), but its function in plants remains unknown.

Phylogenetic analysis of the RPD3/HDA1 superfamily revealed another similarity between plants and animals, but not fungi, in that both possess representatives of Class III proteins, whereas fungi have none. It is possible that these unclassified proteins have an activity other than histone deacetylation.

The degree of evolutionary change differs significantly among HAT gene families (Table 4). At one extreme, gene number in three subfamilies of the GNAT family is completely conserved. The fourth GNAT subfamily (HPA2) is specific to fungi. At the other extreme, the CBP family has been amplified in plants to five genes as compared to a single representative in most animals, and none in fungi. There are two TAF_II250-type proteins in plants as compared to one in fungi and animals. The size of the MYST family ranges from two in Arabidopsis and S.pombe to five in D.melanogaster. Domain and phylogenetic analyses of the CBP-type proteins revealed three classes of these proteins in plants, as compared to a single class in animals, as well as major differences in domain architecture between plant and animal proteins. In addition, HAC2 appears to have diverged early in plant evolution. Its HAT domain appears to have evolved more rapidly than the lineage from which it diverged, and its N-terminal region lacks domains present in other plant CBPs, consistent with in vitro experiments that suggest it does not have HAT activity (70). HAC4 also appears to have evolved more rapidly than the lineage from which it diverged and has distinct features in its N-terminal region.

The Arabidopsis genome encodes proteins homologous to factors in yeast and mammals that associate with HAT complexes SAGA and ADA (GCN5, ADA2 homologs) and HDAC complexes NuRD and SIN3 (RPD3-like, Mi-2, MBD, RbAP46/48 homologs) (see http://www.chromdb.org), suggesting that the Arabidopsis GNAT family HATs and RPD3 family HDACs form complexes similar to those in other organisms (68). In contrast, an analysis of the domain structure of Arabidopsis CBP and TAF_II250 proteins suggests that these proteins may form complexes different from their animal relatives. Plant CBP proteins lack a bromodomain, whereas animal CBPs have one, and plant TAF_II250 proteins have a single bromodomain, compared to the two bromodomains found in their animal homologs. The plant proteins may utilize a different domain that serves the function of the second animal bromodomain in these proteins or may interact with a different bromodomain protein. A precedent for this possibility can be seen in TAF_II145 proteins in S.cerevisiae which do not have a bromodomain, but that interact with Bdf1p. Bdf1p contains two bromodomains and may substitute for the missing C-terminal sequences in the S.cerevisiae Taf_II145p protein (83). Although we identified a number of bromodomain-containing proteins in Arabidopsis, none of these have enough sequence similarity to Bdf1p to suggest a homologous function. However, the Arabidopsis genome encodes two proteins (SGA1 and SGA2; www.chromdb.org) that are similar to yeast Asf1p. Asf1p interacts with Bdf1p, and its counterpart in humans, CIA/ASF1, interacts with the two bromodomains of human TAF_II250 (84). Thus, the possibility exists that one of the many bromodomain proteins in Arabidopsis plays the role of Bdf1p and interacts with an Asf1p homolog. Interestingly, the Arabidopsis genome encodes two TAF_II250 proteins and two ASF1 homologs, whereas yeast and animals encode only one of each.

In addition, our analysis of the Arabidopsis genome sequence revealed that all Arabidopsis bromodomain- containing proteins have only a single bromodomain, in contrast to some animal and S.cerevisiae bromodomain proteins that have multiple copies, ranging from two to five bromodomains. Many bromodomain-containing transcription factors also possess a conserved PHD finger (85–87). Our finding of the absence of such a conserved feature in Arabidopsis bromodomain proteins suggests that the manner in which bromodomains are deployed and utilized differs between plants and animals.

Alternative splicing of two RPD3/HDA1 family genes and one SIR2 family gene could indicate alternative regulatory functions of the RNAs or the predicted protein products, different enzymatic or structural functions for the proteins, or no function at all. Alternative splicing that is conserved in Arabidopsis and tomato SIR2 homologs is suggestive evidence for function of an alternative splicing product, but it is also possible that this is a non-functional splicing product, merely an incidental consequence of a conserved RNA sequence.

These evolutionary differences in fundamental chromatin components among plants, animals and fungi suggest that there may be more evolutionary plasticity and more functional diversification in core chromatin components than might have been anticipated just a few years ago. This diversity is likely to reflect important differences in the manner in which chromatin controls gene expression in these three major kingdoms of eukaryotes, and supports the suggestion that plants have developed mechanisms of global gene regulation related to their unique developmental pathways and environmental responses (88).

SUPPLEMENTARY MATERIAL

Supplementary Material is available at NAR Online.

DDBJ/EMBL/GenBank accession nos⁺