Interkingdom gene fusions.

BACKGROUND: Genome comparisons have revealed major lateral gene transfer between the three primary kingdoms of life - Bacteria, Archaea, and Eukarya. Another important evolutionary phenomenon involves the evolutionary mobility of protein domains that form versatile multidomain architectures. We were interested in investigating the possibility of a combination of these phenomena, with an invading gene merging with a pre-existing gene in the recipient genome.
RESULTS: Complete genomes of fifteen bacteria, four archaea and one eukaryote were searched for interkingdom gene fusions (IKFs); that is, genes coding for proteins that apparently consist of domains originating from different primary kingdoms. Phylogenetic analysis supported 37 cases of IKF, each of which includes a 'native' domain and a horizontally acquired 'alien' domain. IKFs could have evolved via lateral transfer of a gene coding for the alien domain (or a larger protein containing this domain) followed by recombination with a native gene. For several IKFs, this scenario is supported by the presence of a gene coding for a second, stand-alone version of the alien domain in the recipient genome. Among the genomes investigated, the greatest number of IKFs has been detected in Mycobacterium tuberculosis, where they are almost always accompanied by a stand-alone alien domain. For most of the IKF cases detected in other genomes, the stand-alone counterpart is missing.
CONCLUSIONS: The results of comparative genome analysis show that IKF formation is a real, but relatively rare, evolutionary phenomenon. We hypothesize that IKFs are formed primarily via the proposed two-stage mechanism, but other than in the Actinomycetes, in which IKF generation seems to be an active, ongoing process, most of the stand-alone intermediates have been eliminated, perhaps because of functional redundancy.

Background

Comparative genome analysis has revealed major lateral gene transfer between the three primary kingdoms of life, Bacteria, Archaea, and Eukarya [1,2,3,4]. The best recognized form of lateral gene flux is the transfer of numerous genes from mitochondria and chloroplasts to eukaryotic nuclear genomes [5]. Far beyond that, however, the role of lateral gene exchange, along with lineage-specific gene loss, as one of the principal factors of evolution, at least among prokaryotes, is obvious from the fact that the great majority of conserved families of orthologous genes show a 'patchy' phyletic distribution [6,7]. In many cases, such families are shared by phylogenetically distant species (for example, bacteria and archaea), while they are missing in some of the more closely related species (for example, bacteria from the same lineage). Correlations have been noticed between the preferred routes of gene transfer and the lifestyles of the organisms involved. Thus, massive gene exchange seems to have occurred between archaeal and bacterial hyperthermophiles [8,9], whereas certain parasitic bacteria, for example, chlamydia and spirochetes, appear to have acquired significantly more eukaryotic genes than free-living bacteria [10,11,12].

Another evolutionary trend that is predominant in eukaryotes, but is important also in bacteria and archaea, involves the evolutionary mobility of protein domains that combine to form variable multidomain architectures [13,14,15,16,18]. Domain fusion is one of the foundations of most forms of regulation and signal transduction in the cell. Examples include prokaryotic transcriptional regulators, most of which consist of the DNA-binding helix-turn-helix domain fused to a variety of small-molecule-binding domains [19], the two-component signal transduction system that is based on fusions of histidine kinases with sensor domains and of receiver domains with DNA-binding domains [20], and the sugar phosphotransferase (PTS) systems that include complex fusions of several enzymes [21]. In the evolution of eukaryotes, domain fusion takes the form of domain accretion, whereby proteins from complex organisms (such as animals) that are involved in various forms of regulation and signal transduction tend to accrue multiple domains that facilitate the formation of complex networks of interactions [22].

We were interested in exploring the possibility of a meeting between these two major evolutionary phenomena - lateral gene exchange and gene fusion - which would result in the formation of multidomain proteins in which different domains display distinct evolutionary provenance. In particular, we sought to identify fusions between domains originating from different primary kingdoms - Bacteria, Archaea and Eukarya - which we term interkingdom gene (domain) fusions (IKFs), and obtain clues to the pathways of IKF origin through comparative genome analysis. We show that, although IKF in general is a rare phenomenon, one bacterial lineage, the Actinomycetes, displays a significantly increased frequency of such events; we also propose a probable mechanism for IKF formation.

Results and discussion

To identify IKFs, all protein sequences encoded in the analyzed genomes were compared to the non-redundant protein database, and those proteins in which distinct parts showed the greatest similarity to homologs from different primary kingdoms were identified (see the Materials and methods section). In most cases, the reported alignments were highly statistically significant, leaving no doubt that true homologs were detected (Table 1). On the few occasions when the database search statistics in themselves were not fully convincing (for example, the OB-fold nucleic acid-binding domain in the Bacillus subtilis protein YhcN and the methyltransferase domain in the YabN protein, also from B. subtilis), the homologous relationship was validated by detection of the salient sequence motifs known to be involved in the corresponding protein functions (data not shown). Such motif analysis was performed for all analyzed domains in order not only to validate homology, but also to distinguish between active and inactivated forms of enzymes. Figure 1 shows multiple alignments of two domains involved in an IKF, illustrating the conservation of the characteristic functional motifs and the specific similarity between each of the domains of the IKF protein (in this case from Aquifex aeolicus) and their archaeal and bacterial homologs, respectively.

Table 1

Interkingdom domain fusions and their probable origins

IKF gene	Best 'native' hit	Best 'alien' hit	Protein function	Stand-alone	Comment
(GI number and gene	(E-value, amino acid	(E-value, amino		paralog of the
name) and origin	residue range,	acid residue range,		alien domain
of domains	species)/domain	species)/domain
	function	function
Archaea
Aeropyrum pernix
5106104_	2621953_Mth	2633525_Bs	Hydroxymethyl-	None	Pyrococci encode proteins with
APE2400	5e-27;	4e-54;	pyrimidine		the same domain organization
Archaeal-bacterial	282-445;	16-272;	phosphate kinase		andclosest similarity to A. pernix;
	uncharacterized domain	hydroxymethyl-	involved in thiamine		M. jannaschii encodes a protein
	conserved among	pyrimidine phosphate	biosynthesis		with the same domain
	archaea (homolog	kinase	(additional function?)		organization but low similarity;
	of the amino-terminal				Mt encodes a HMP-kinase with
	domain of sialic acid				moderate similarity
	synthase)
Methanococcus jannaschii
1591138_	2128140_Mj;	7270033_At;	Unknown;	None	The amino-terminal domain is
MJ0434	1e-19;	0.003;	possible role		present in several stand-alone
Archaeal-	2-94;	120-222;	in stress response		copies in M. jannaschii, but
bacterial-eukaryotic	uncharacterized	AIG2-like			otherwise, is seen mostly in
	domain	stress-related			bacteria; the possibility of
		protein			acquisition of a bacterial gene
					by the Methanococcus lineage
					is conceivable
Methanobacterium thermoautotrophicum
2621249_	5103547_Ap;	1651798_Ssp;	Membrane-associated	None	In Ssp, the amino-terminal
MTH204	1e-34;	0.002;	5-formyl-		domain is fused to another
Archaeal-	137-326;	8-139;	tetrahydrofolate		uncharacterized domain. An
eukaryotic/	5-formyl-	uncharacterized	cyclo-ligase(?);		ortholog with conserved
bacterial	tetrahydrofolate	membrane-associated	exact function		domain organization is seen
	cyclo-ligase	domain	unknown		in Mycobacterium, but many
					other bacteria encode stand-
					alone versions of this domain,
					which could be the actual sources
					of horizontal gene transfer
2621673_	3256572_Ph;	2984130_Aa;	GTPase, possible	2621855
MTH594	3e-10;	6e-19;	role in signal
Archaeal-bacterial	5-137;	233-390;	transduction
	inactivated RecA	GTPase
	domain
2622642_	5105992_Ap;	2569943_Axy;	Glucose-1-phosphate	None
MTH1523	3e-36;	2e-05;	thymidylyl transferase/
Archaeal-bacterial	5-226;	226-334;	glucose-6-phosphate
	glucose-1-phosphate	mannose-6-	isomerase
	thymidylyl transferase	phosphate isomerase
Bacteria
Aquifex aeolicus
2983622_	2633696_Bs;	2650176_Af;	Signal	None
aq_1151	5e-65;	0.005;	transduction
Bacterial-archaeal	325-795;	116-279;	c-di-GMP
	c-di-GMP phospho-	PAS/PAC	phospho-diesterase
	diesterase	domain
2984285_	586875_Bs;	3915955_Mj;	Molybdenum	None
aq_2060	4e-63	3e-09;	cofactor
Bacterial-archaeal	1-252;	270-441;	bisynthesis enzyme(?)
	PHP superfamily	pyruvate
	hydrolase	formate-lyase
		activating enzyme
		(Fe-S cluster
		oxidoreductase)
Bacillus subtilis
2632283_yaaH,	4980914_Tm	399377_Rn	Chitinase	2635915	B. subtilis encodes two
1945087_ydhD	1e-06	2e-11			paralogous proteins with the
Bacterial-eukaryotic	2-92;	221-402;			same domain architecture
	LysM repeat domain	chitinase
2633242_yhcR	645819_Dr;	2622704_Mth;	Nuclease-nucleotidase	None
Bacterial-archaeal	1e-64;	0.008	(probable repair
	584-1068;	151-257;	enzyme)
	5'-nucleotidase;	nucleic acid-binding
	1175987_	domain (OB-fold)
	ECR100;
	2e-09;
	377-521;
	thermonuclease
2632325_yabN	4981449_Tm;	3873806_Ce;	Methyl-transferase/	None	Other than in chlamydiae,
Bacterial-eukaryotic	2e-62;	0.003;	pyro-phosphatase		the SWI domain is seen
	223-483;	7-125;	(metabolic enzyme		in eukaryotic chromatin-
	MazG (predicted pyro-	SAM-dependent	of an unknown		associated proteins, leading
	phosphatase)	methyl-transferase	pathway?)		to the suggestion that
					chlamydial topoisomerase
					is involved in chromosome
					condensation
Chlamydophyla pneumoniae
4377077_	730965_Bs;	3581917_Sp;	DNA topoisomerase I,	7189103	SWI is a typical eukaryotic
CPn0769	e-148;	3e-10;	possibly involved in		domain not found in
Bacterial-eukaryotic	1-727;	792-866;	chromatin		prokaryotes other than
	DNA topoisomerase I	SWI domain	condensation		chlamydia (the ortholog
					in Chlamydia trachomatis has the
					same domain architecture)
Deinococcus radiodurans
6459294_	7248325_Sco;	6754878_Mm;	DNase	None	The G9a domain is not
DR1533	0.001;	9e-28;			detectable in other prokaryotes.
Bacterial-eukaryotic	171-265;	4-148;			In eukaryotes, this domain so
	McrA family	G9a domain (DNA-			far has been found only as part
	endonuclease	binding?)			of multidomain nuclear proteins,
					including transcription factors
Escherichia coli
1787179_	94933_Ppu;	3747107_Rn;	Oxidoreductase	None	The eukaryotic domain is present
b0947	3e-10;	3e-32;			(as a partial sequence) also in the
Bacterial-eukaryotic	287-367;	4-261;			beta-proteobacterium Vogesella.
	ferredoxin	uncharacterized			This domain contains a conserved
		domain (thiol			pair of cysteines, which together
		oxidoreductase?)			with the ferredoxin fusion, may
					suggest a thiol oxidoreductase
					activity. Most of the eukaryotic
					proteins containing this domain
					appear to be mitochondrial,
					suggesting the possibility of an
					alternative evolutionary scenario
1787678_	487713_Sli;	5459012_Pab;	Methyl-transferase/	None
b1410	3e-05;	1e-17;	Lipase (exact function
Bacterial-archaeal	408-522;	33-274;	unclear)
	SAM-dependent	lyso-phospholipase
	methyl-transferase
1787679_ynbD	1591375_Mj;	7160233_Sp;	Membrane-associated	None	An unusual case of fusion
Archaeal-eukaryotic	4e-04;	1e-06;	bifunctional		between an apparently archaeal
	50-218;	346-415;	phosphatase		and a typical eukaryotic domain
	membrane-associated	tyrosine phosphatase			in a bacterium
	acid phosphatase
1788589_	5763950_Sco;	3860247_At;	Bifunctional enzyme;	None
b2255	4e-35;	1e-55;	exact function unclear
Bacterial-eukaryotic	1-259;	318-652;
	methionyl-tRNA	dTDP-glucose 4-6-
	formyl-transferase	dehydratase
1788938_yfiQ	929735_Nsp;	2649370_Af;	acetyl-CoA synthetase/	None
bacterial-Archaeal/	8e-32;	4e-85;	acetyl-transferase; exact
eukaryotic	637-874;	6-689;	function unclear
	acetyl-transferase	acetyl-CoA synthetase
Mycobacterium tuberculosis
2909507_	6469244_Sco;	4151109_Tbr;	Adenylate cyclase/	7476546,	M. tuberculosis encodes three
Rv2488c,	5e-64;	6e-04;	ATPase; probable	7476738	paralogous proteins that consist
2791528_Rv1358,	19-603;	6-167;	transcription regulator		of three domains, the eukaryotic-
1419061_	4726088_Rer;	adenylate cyclase			type adenylate cyclase, AP
Rv1358	2e-12;				(apoptotic) ATPase and DNA-
Bacterial-eukaryotic	818-1073				binding response regulator, and
					two stand-alone versions of
					adenylate cyclase, which show the
					closest similarity to the cyclase
					domain of the multidomain
					proteins
1314025_	120037_Tt;	178213_Hs;	Ferredoxin/	2076681	D. radiodurans also encodes the
Rv0886	1e-11;	4e-65;	ferredoxin reductase		eukaryotic-type ferredoxin
Bacterial-eukaryotic	2-79;	93-543;			reductase, but the ferredoxin
	ferredoxin	ferredoxin reductase			fusion is unique to mycobacteria
3261732_	2661695_Sco;	279520_Dd;	cAMP-dependent	4455714
Rv0998	3e-13;	7e-07;	acetyl-transferase(?)	(M. leprae)
Bacterial-eukaryotic	148-328;	30-105;
	acetyl-transferase	cAMP-binding domain
2326726_	421331_Cvi;	2645721_Mm;	Bifunctional enzyme of	1929080
Rv1683	1e-24;	6e-26;	poly (3-hydroxy-butyrate)
Bacterial-eukaryotic	23-359;	456-972;	synthesis
	poly (3-hydroxy-	very-long-chain
	butyrate) synthase	acyl-CoA synthetase
1403447_	6752338_Sco;	3892714_At;	Polyfunctional enzyme	2661651	In this protein, the domain of
Rv2006	2e-27;	8e-27;	of trehalose metabolism		apparent eukaryotic origin
Bacterial-eukaryotic	23-240;	264-521;			is flanked by bacterial domains
	phosphatase;	trehalose-6-phosphate			from both sides
	6448751_Sco;	phosphatase
	0.0;
	534-1320;
	trehalose hydrolase
2896788_	117648_Ec;	3073773_Mm;	Polyfunctional enzyme	2337823	The presence of the stand-alone
Rv2051c	1e-16;	4e-31;	of lipid metabolism	(M. leprae);	version of the eukaryotic
Bacterial-eukaryotic	94-514;	588-829;		6468712	domain in Streptomyces suggests
	apolipoprotein	dolichol-phosphate-		(Streptomyces	an ancient horizontal transfer
	N-acyltransferase	mannose synthase		coelicolor)
2791523_	6225563_Scy;	1098605_Cnu;	Multifunctional enzyme	None
Rv2483c	7e-16;	5e-22;	of phospholipid
Bacterial-eukaryotic	36-253;	289-492;	metabolism
	phosphoserine	1-acyl-sn-
	phosphatase	glycerol-3-phosphate
		acyltransferase
2894233_	2633801_Bs;	4538974_At;	Molybdopterin synthase	2076687	The same domain organization
Rv3323c	3e-19;	7e-06;			is seen in D. radiodurans, but in
Bacterial-eukaryotic	89-208;	2-82;			this case, both components
	molybdopterin	molybdopterin			appear to be of bacterial origin
	synthase large subunit	synthase small subunit
	(MoaE)	(MoaD)
2960152_	4753872_Sco;	466119_Ce;	cAMP-regulated	2501688	M. tuberculosis encodes two
Rv3728,	1e-35;	7e-20;	efflux pump(?)		strongly similar paralogs with
7477551_	56-428;	549-964;			the same domain architecture
Rv3239c	transmembrane	cAMP-binding domain-
Bacterial-eukaryotic	efflux protein	phosphoesterase
2960153_	4731342_Sl;	1591330_Mj;	Bifunctional enzyme	1806159	The amino-terminal domain
Rv3729	3e-14;	3e-58;	of molybdenum		stand-alone paralog is more
Bacterial-archaeal	510-776;	molybdenum	cofactor biosynthesis		similar to archaeal homologs
	C5-O-methyl-	cofactor biosynthesis			than to the stand-alone paralog,
	Transferase	protein MoaA			but nevertheless, the latter
	(mitomycin	(Fe-S oxidoreductase)			appears to be of archaeal origin
	biosynthesis)
3261806_	40487_Cg;	7304009_Dm;	Secreted protein	7649504	The stand-alone version of the
Rv3811	3e-12;	2e-12;		(S. coelicolor)	eukaryotic domain is present
Bacterial-eukaryotic	404-494;	198-384;			only in Streptomyces
	major secreted	peptidoglycan
	protein	recognition protein
Treponema pallidum
3322964_	7225946_Nm;	320868_Sc;	Uridine kinase	None	A co-linear ortholog is present
TP0667	9e-04;	2e-13;			in Thermotoga
Bacterial-eukaryotic	10-154;	290-488;
	threonyl-tRNA	uridine kinase
	synthetase (TGS and
	H3H domains)
Thermotoga maritima
4981276_	68516_Bs;	3218401_Sp;	Uridine kinase	None	A co-linear ortholog is present
TM0751	3e-07;	2e-11;			in Treponema
Bacterial-eukaryotic	11-200;	288-475;
	threonyl-tRNA	uridine kinase
	synthetase (TGS and
	H3H domains)
Eukaryotes
Saccharomyces cerevisiae
536367_	586134_Bt;	7450047_Aa;	Bifunctional signal-	5249	SurE homologs are not
Ybr094w	9e-10;	8e-09;	transduction protein	(Yarrowia	detectable in eukaryotes other
Eukaryotic/	tubulin-tyrosine ligase	acid phosphatase		lipolytica)	than yeasts
Bacterial-archaeal		(SurE)
1431219_	577625_Hs;	3328426_Ct
YDL141w	1e-39	5e-27;
Eukaryotic-	Biotin-[propionyl-	biotin protein ligase	Bifunctional biotin-	None	An ortholog with an identical
bacterial	CoA-carboxylase(ATP-		protein ligase		domain architecture is present
	hydrolysing)] ligase				in S. pombe
458922_	477096_Gg;	1653075_Ssp;	heat shock	NONE	An ortholog with an identical
YHR206W	8e-18;	7e-17;	transcription		domain architecture is present
Eukaryotic-bacterial	78-216	375-503;	factor		in S. pombe (3327019)
	heat shock	CheY domain
	transcription factor
	domain	2983676_Aa;	Siroheme synthase	2330809	S. pombe also encodes a co-linear
486539_	1146165_At;	1e-04;		(S. pombe)	ortholog (3581882); apparent
YKR069w	3e-34;	22-188;			displacement of the bacterial
Eukaryotic-bacterial	249-556;	precorrin-2 oxidase			precorrin-2 oxidase by a distinct
	urophorphyrin III				Rossmann fold domain
	methylase
1302305_	4938476_At;	3212189_Hi;	Multifunctional enzyme	None	Co-linear orthologs in S. pombe
YNL256w	5e-65;	5e-05;	of folate biosynthesis		(7490442) and Pneumocystis
Eukaryotic-bacterial	324-861	62-148;			carinii (283062)
	7,8-dihydro-6-	187-297;
	hydroxymethylpterin-	dihydro-neopterin
	pyro-phosphokinase+	aldolase
	Dihydro-pteroate
	synthase
1419887_	7297709_Dm;	5918510_Sco;	Bifunctional RNA	2213559	The known bacterial homologs
YOL066c	2e-72;	2e-10;	modification enzyme	(S. pombe)	have a two-domain organization;
Eukaryotic-bacterial	42-408;	436-574;			the evolutionary scenario could
	large ribosomal	pyrimidine deaminase			have included domain
	subunit pseudoU				rearrangements
	synthase
1419865_	2462827_At;	1075360_Hi;	Transcriptional regulator None		Yeast encodes three strongly
YOL055c,	1e-39;	6e-24;	of thiamine biosynthesis		similar paralogs with identical
2132251_	22-390;	342-549;	genes(?)		domain organization; co-linear
YPL258c,	phosphomethyl	transcriptional			orthologs are present in other
2132289_	pyrimidinekinase	activator			ascomycetes
YPR121w	(thiamine biosynthesis)
Eukaryotic-bacterial
1370444_ YPL214c	2746079_Bn;	2648451_Af;	Bifunctional thiamine	None	Except for the one from
Eukaryotic-archaeal/	1e-27;	9e-27;	biosynthesis enzyme		A. fulgidus, all highly conserved
Bacterial	9-233;	251-531;			homologs of the kinase domain
	thiamin-phosphate	hydroxyethyl-thiazole			of this protein are bacterial; it
	pyro-phosphorylase	kinase			appears likely that the A. fulgidus
					gene is the result of horizontal
					transfer

The following complete genomes were analyzed. Archaea: Aeropyrum pernix (Ap); Archaeoglobus fulgidus (Af); Methanococcus jannaschii (Mj); Methanobacterium thermoautotrophicum (Mth); Pyrococcus horikoshii (Ph); Bacteria: Aquifex aeolicus (Aa); Borrelia burgdorferi (Bb); Bacillus subtilis (Bs); Chlamydophila pneumoniae (Cp); Deinococcus radiodurans (Dr); Escherichia coli (Ec); Haemophilus influenzae (Hi); Helicobacter pylori (Hp); Mycobacterium tuberculosis (Mt); Mycoplasma pneumoniae (Mp); Rickettsia prowazekii (Rp); Synechocystis sp (Ssp); Thermotoga maritima (Tm); Treponema pallidum (Tp). No IKFs were detected in the genomes that are not shown in the table. Additional species name abbreviations: At, Arabidopsis thaliana; Axy, Acetobacter xylinus; Bn, Brassica napus; Ce, Caenorhabditis elegans; Cvi, Chromatium vinosum; Gg, Gallus gallus; Hs, Homo sapiens; Mm, Mus musculus; Rn, Rattus norvegicus; Sco, Streptomyces coelicolor; Sl, Streptomyces lavendulae.

Figure 1

Multiple alignments of two domains comprising an interkingdom domian fusion. Alignments of (a) the PHP-hydrolase domain [4] and (b) the pyruvate formate lyase activating enzyme domain of the IKF protein aq_2060 from A. aeolicus. The sequences of the aq_2060 domains are placed with the most similar sequences of the corresponding stand-alone enzymes, bacterial ones in the case of PHP-hydrolase and archaeal ones in the case of the pyruvate formate lyase activating enzyme. The phylogenetic trees produced form these alignments are shown in Figure 2c. The numbers in parentheses show the lengths of regions between the aligned blocks that are not shown. The consensus includes amino acid residues and residue classes that are conserved in 75% of the aligned sequences; the residue classes are as follows: h, hydrophobic; l, aliphatic; a, aromatic; s, small; u, tiny; p, polar; b, big; t, residues with high turn-forming propensity. Asterisks show the predicted active site residues; note the replacements in some of the sequences that are predicted to be inactivated versions of the respective enzymes (see text). The alignments were colored using the BOXSHADE program [30]; individual residues conserved in at least 50% of the aligned sequences are in red; residues similar to the conserved ones and groups of conserved similar residues are in blue.

In several cases, the chimeric origin of a gene was obvious at a qualitative level because no homolog of the 'alien' domain with comparable sequence similarity was detected in the recipient superkingdom (Table 1, Figure 2a,b). For the rest of the candidate IKFs, phylogenetic tree analysis was performed to corroborate the origin of the invading domain by horizontal transfer; statistically significant grouping of a candidate IKF domain with homologs from the donor superkingdom provides such evidence (Figure 2c,d). The overall number of confirmed IKFs is relatively small - 37 in 21 compared genomes (about 0.1% of the genes) - compared to the total number of likely interkingdom gene transfers. For completely sequenced bacterial genomes this has been conservatively estimated as 1-2% of the genes, with a greater fraction (2-10%) detected in archaea and hyperthermophilic bacteria ([23], and K.S. Makarova, L. Aravind and E.V.K., unpublished observations). Examination of the clusters of orthologous groups (COGs) of proteins from complete genomes [6], in which multidomain proteins are split into the constituent domains if the orthologs of the latter are present as stand-alone forms in some of the genomes, shows that IKFs constitute only a small fraction of all fusions of evolutionarily mobile domains (Figure 3). Generally, the small number of identified IKFs compared to the total number of inferred horizontal transfer events and the total number of domain fusions could be compatible with a random model of domain fusion subsequent to lateral gene transfer.

Figure 2

Examples of phylogenetic trees supporting the contribution of interkingdom horizontal gene transfer to the emergence of interkingdom domain fusions. The names of proteins from different primary kingdoms are color-coded: black, bacterial; pink, archaeal; green, eukaryotic; the domains involved in the apparent IKF are shown in red. Red circles show nodes with bootstrap support >70%, and yellow circles show nodes with 50-70% support. The bar unit corresponds to 0.1 substitutions per site (10 PAM). (a) IKF: Rv1683 (gi| 7476858) from M. tuberculosis. Fusion of a bacterial poly(3-hydroxy-butyrate) (PHB) synthase and eukaryotic very long chain acyl-CoA synthetase. Note the absence of eukaryotic homologs in the PHB synthase tree and of bacterial homologs other than the two from M. leprae in the acyl-CoA synthetase tree. (b) IKF: yeast YOL066c (gi|6324506). Fusion of a eukaryotic pesudouridylate synthetase with a bacterial pyrimidine deaminase. Note the absence of eukaryotic homologs, other than that from S. pombe, in the pyrimidine deaminase tree. (c) IKF: aq_2060 (gi|2984285) from Aquifex aeolicus. This protein is a fusion of a PHP superfamily hydrolase of apparent bacterial origin and a pyruvate formate-lyase activating enzyme of archaeal origin. (d) IKF: yeast YOL055c (gi|1419865), YPL258c (gi|2132251) and YPR121w (gi|2132289) from S. cerevisiae. Fusion of a eukaryotic phosphomethylpyrimidine kinase and a bacterial transcriptional activator. Species abbreviations: Bac.meg., Bacillus megaterium; Chr.vin., Chromatium vinosum; Thi.vi., Thiocystis violacea; Am.med., Amycolatopsis mediterranei; Coch.het., Cochliobolus heterostrophus; Dme, Drosophila melanogaster; Cel, Caenorhabditis elegans; Mus, Mus musculus; Spo, Schizosaccharomyces pombe; Ath, Arabidopsis thaliana; Strep.co., Streptomyces coelicolor; The.nea., Thermotoga neapolitana; Bac. am, Bacillus amyloliquefaciens; Shi.fl., Shigella flexneri; Hsa, Homo sapiens.

Figure 3

Overall numbers of domain fusions estimated using the COGs and interkingdom domain fusions encoded in completely sequenced genomes. The data for estimating the overall number of domain fusions were from the current COG release [6], which does not include several bacterial and archaeal species (for example, Aeropyrum pernix and Deinococcus radiodurans) that have been analyzed in the present work (Table 1). Accordingly, the data for these genomes are not shown in the figure. Species name abbreviations: Af, Archaeoglobus fulgidus; Mj, Methanococcus jannaschii; Mth, Methanobacterium thermoautotrophicum; Ph, Pyrococcus horikoshii; Sc, Saccharomyces cerevisiae; Aa, Aquifex aeolicus; Tm, Thermotoga maritima; Ssp, Synechocystis sp.; Ec, Escherichia coli; Bs, Bacillus subtilis; Mtu, Mycobacterium tuberculosis; Hi, Haemophilus influenzae; Hp, Helicobacter pylori; Mg; Mycoplasma genitalium; Mp, Mycoplasma pneumoniae; Bb, Borrelia burgdorferi; Tp, Treponema pallidum; Ct, Chlamydia trachomatis; Cp, Chlamydophila pneumoniae; Rp, Rickettsia prowazekii.

However, the distribution of IKFs among genomes is distinctly non-random, suggesting that such a simple model may be incorrect. Specifically, 12 IKFs were detected in Mycobacterium tuberculosis and 10 were found in the yeast Saccharomyces cerevisiae, but only a small number or none was identified in each of the other bacterial and archaeal genomes (Figure 2, Table 1). The excess of IKFs in Mycobacterium is particularly notable, given that the fraction of genes horizontally transferred from archaea and eukaryotes in the mycobacterial genome is only slightly greater than that in most of the other bacteria, and considerably lower than that in the hyperthermophilic bacteria Aquifex and Thermotoga (K.S. Makarova, L. Aravind and E.V.K., unpublished observations). Similarly, whereas the overall number of domain fusions in M. tuberculosis is greater than in most other bacteria, the difference is insufficient to account for the over-representation of IKFs; furthermore, the cyanobacterium Synechocystis sp. has an even greater overall number of fusions but does not have any detectable IKFs (Figure 3). At present, we cannot provide a defendable biological explanation for the comparatively high frequency of IKF in Mycobacterium. It is tempting to interpret this trend in terms of adaptation of this bacterium to its relatively recently occupied parasitic niche, but examination of the individual IKF cases does not offer immediate clues in mycobacterial biology. The yeast IKFs clearly represent relatively recent horizontal transfers distinct from the gene influx from the mitochondria following the establishment of endosymbiosis because, under the protocol of IKF detection used here, only those alien domains were identified that have no counterparts in other eukaryotes.

Most of the IKFs are unique, but B. subtilis, M. tuberculosis and yeast each also encode families of two to three paralogous IKFs, which apparently have evolved by duplication subsequent to the respective fusion events (Table 1). Strikingly, the same IKF, the three-domain uridine kinase, is shared by Treponema pallidum and Thermotoga maritima (Table 1). Given that these two bacteria are not specifically related and that Borrelia burgdorferi, the second spirochete whose genome has been sequenced, encodes a typical bacterial uridine kinase, the presence of a common IKF in Treponema and Thermotoga cannot be realistically attributed to vertical inheritance of this gene from a common ancestor. It thus probably reflects horizontal transfer of the gene encoding the three-domain protein subsequent to its emergence in either the spirochetes or the Thermotogales.

Two evolutionary issues pertaining to IKFs need to be addressed, namely the mechanism(s) of their origin and the selective forces responsible for their preservation. From general considerations, it seems likely that IKFs have evolved via a two-step process, which involves lateral transfer of the complete gene coding for the IKF's alien portion, followed by domain fusion. This scenario rests on the assumption that the acquired foreign gene is selectively advantageous, because otherwise it would have been inactivated by mutations before recombination could take place. Under this mechanism, the alien portion of an IKF is likely to be present in the recipient genome also as a stand-alone gene. A clear-cut case of such a duplication of a horizontally transferred domain has been noticed in Chlamydia, whose genomes encode the SWI domain, implicated in chromatin condensation, both as a stand-alone protein and as the carboxy-terminal portion of topoisomerase I [10]. Apart from this case, the IKFs fall into two readily discernible classes, namely those from Mycobacterium and all the rest. M. tuberculosis (the only complete genome of an actinomycete available) possesses considerably more IKFs than any other bacterial or archaeal species (see above), and typically, the alien portions of these proteins show high level of similarity to the homologs from the donor superkingdom (eukaryotes). Most significantly, there is also, with a single exception, a stand-alone counterpart in the mycobacterial genome; in some cases, such a counterpart is seen only in a closely related species, M. leprae, and in one case, it is found in Streptomyces, a distantly related actinomycete (Table 1). In the other genomes, the IKFs are generally less similar to the apparent donor and, with a few exceptions, stand-alone versions of the alien domains are missing (Table 1). The hypothesis that seems to be most compatible with these observations is that IKFs indeed evolve via a stand-alone, horizontally transferred intermediate, but in the case of ancient IKFs, these intermediates are typically eliminated during evolution, perhaps because their function becomes redundant with the formation of the IKF. The IKFs identified in actinomycetes appear to result from relatively recent gene fusion events so that the original, stand-alone transferred genes are still present in the genome.

The IKFs include a variety of protein functions. Only some of these are well understood such as, for example, those of the bifunctional nucleotide and coenzyme metabolism enzymes that are particularly abundant in yeast (Table 1). In other cases, the function of an IKF-encoded protein could be predicted only tentatively on the basis of the functions of its constituent domains (Table 1). The selective advantage of the formation of multidomain proteins, at least as far as enzymes are involved, lies in the possibility of effective coupling of the reactions catalyzed by the different domains [16]; this may be generalized also for functional coordination of non-enzymatic domains. Fusion may result in the addition of a regulatory function to an enzymatic one. For example, it appears most likely that the RNA-binding TGS domain [24] in the uridine kinases of Treponema pallidum and Thermotoga maritima is involved in autoregulation of translation. The unusual aspect of the IKFs appears to be the compatibility of evolutionarily distant domains.

Examination of the phyletic distribution of the multidomain architectures of IKFs may help in pinpointing the evolutionary stage at which the fusion (but not necessarily the preceding horizontal gene transfer) has occurred. For example, the fusion of the SWI domain with topoisomerase belongs after the radiation of Chlamydia from other bacterial lineages, but before the radiation of Chlamydia pneumoniae and Chlamydia trachomatis (Table 1). The majority of IKFs detected in the yeast S. cerevisiae are also present in Schizosaccharomyces pombe and/or other ascomycetes (Table 1, and data not shown), but not in any other eukaryotes, and accordingly, they should have evolved at a relatively early stage of fungal evolution, but not before the fungal clade diverged from the rest of the eukaryotic crown group.

Finally, it should be noted that formation of some of the IKFs might have required more complex rearrangements of the contributing proteins than simple domain fusion. Figure 4 shows the domain architectures of proteins that contribute domains to two IKFs. In each case, a simple fusion between genes encoding the respective individual domains is insufficient to explain the emergence of the IKF. For example, the uridine kinase example mentioned above (Figure 4a) should have involved isolation of the TGS-HxxxH domains of threonyl-tRNA synthetase before or concomitantly with their fusion with the uridine kinase. The specific molecular mechanism could have involved selective duplication of the upstream portion of the threonyl-tRNA synthetase gene. Similarly, the sialic acid synthase homologous domain, which is fused to hydroxymethylpyrimidine phosphate kinase in A. pernix and pyrococci, appears to have been derived from two-domain proteins that additionally contain a helix-turn-helix DNA-binding domain (Figure 4b). These hypotheses of a complex mechanism of gene fusion involved in the emergence of IKFs are based on a limited sample of sequenced genomes. An alternative possibility is that, before the postulated horizontal transfer event, the recipient domain(s) has been encoded by a stand-alone gene; such genes that do not contain the fused alien domain may yet be discovered in newly sequenced genomes. In fact, a stand-alone version of the sialic acid synthase homologous domain is seen in Methanobacterium, although it is considerably less similar to the IKF than the version fused to the HTH domain (Figure 4b).

Figure 4

Multidomain architectures of interkingdom fusion proteins and their homologs (examples). (a) The three-domain uridine kinase; (b) the sialic acid synthase homologous domain fused to hydroxymethylpyrimidine phosphate kinase. Domain name abbreviations: TTRS, threonyl-tRNA synthetase; UDK, uridine kinase; TGS and H3H, amino-terminal domains of TTRS; HMP-PK, hydroxymethylpyrimidine phosphate kinase; SISH, sialic acid synthase homologous domain; HTH, helix-turn-helix DNA-binding domain. Different shades represent distinct sequence families of each domain. Species name abbreviations: Tp, Treponema pallidum; Tm, Thermotoga maritima; Mth, Methanobacterium thermoautotrophicum; Mj, Methanococcus jannaschii; Ap, Aeropyrum pernix; Ph, Pyrococcus horikoshii; Pa, Pyrococcus abyssii.

The identification of IKFs underscores the complexity of the evolutionary process as revealed by comparison of multiple genomes. In and by itself, this phenomenon may not have a unique biological significance, but it reveals the overlap between two major evolutionary trends, horizontal gene transfer and protein domain rearrangement, and shows that domains, rather then entire proteins (genes), should be considered fundamental units of genetic material exchange.

Materials and methods

Protein sequences encoded in 21 complete genomes of archaea, bacteria and the yeast Saccharomyces cerevisiae were extracted from the Genome division of the Entrez retrieval system [25]. Each protein encoded in these genomes was used as the query in a comparison against the non-redundant protein sequence database (National Center for Biotechnology Information, NIH, Bethesda, USA) using the BLASTP program [26]. For each query, the set of local similarities detected by BLASTP was automatically (using a Perl script written for this purpose) screened for putative IKFs, that is situations in which the query did not have full-size homologs outside its immediate taxonomic group (for example, the Proteobacteria for Escherichia coli) and in which different regions of the query showed the greatest similarity to proteins from different primary kingdoms. The pseudocode for the script follows:

The script itself is available as an additional data file. The candidate IKF cases were further examined to detect situations where one or more distinct regions of the query could be classified as 'native' or 'alien' either on the basis of the lack of close homologs from the respective primary kingdom or using phylogenetic analysis. Multiple sequence alignments were generated using the ClustalW program [27], and when necessary, manually corrected to ensure the proper alignment of conserved motifs typical of the respective domains. Phylogenetic trees were constructed using the PROTDIST and FITCH programs of the PHYLIP package [28]. Trees were made separately for each domain of a putative IKF, and its mixed ancestry was considered confirmed if the affinities of the domains with different primary kingdoms were supported by bootstrap values of at least 50%. Additional iterative database searches were performed using the PSI-BLAST program [26,29] in order to predict functions of the individual domains of the identified IKFs in cases when these were not immediately clear.

Additional data

The following additional data are included with the online version of this paper: the Perl script used to screen local similarities for putative IKFs.

The Perl script used to screen local similarities for putative IKFs