Mitochondrial Diversity of Early-Branching Metazoa Is Revealed by the Complete mt Genome of a Haplosclerid Demosponge

Abstract

The first mitochondrial (mt) genomes of demosponges have recently been sequenced and appear to be markedly different from published eumetazoan mt genomes. Here we show that the mt genome of the haplosclerid demosponge Amphimedon queenslandica has features that it shares with both demosponges and eumetazoans. Although the A. queenslandica mt genome has typical demosponge features, including size, long noncoding regions, and bacterialike rRNA genes, it lacks atp9, which is found in the other demosponges sequenced to date. We found strong evidence of a recent transposon-mediated transfer of atp9 to the nuclear genome. In addition, A. queenslandica bears an incomplete tRNA set, unusual amino acid deletion patterns, and a putative control region. Furthermore, the arrangement of mt rRNA genes differs from that of other demosponges. These genes evolve at significantly higher rates than observed in other demosponges, similar to previously observed nuclear rRNA gene rates in other haplosclerid demosponges.

Porifera, sponges, Demospongiae, mitochondrial genome, Metazoa, Amphimedon queenslandica

The acceptance of relative uniformity of metazoan mitochondrial (mt) genomes (Lang et al. 1999) has been weakened significantly by the recently published mt genomes of Porifera (Geodia neptuni, Tethya actinia, and Axinella corrugata, all class Demospongiae; Lavrov and Lang 2005; Lavrov et al. 2005) and Placozoa (Trichoplax adhaerens, Dellaporta et al. 2006). The mt genomes of these early-branching animals exceed the typical length of metazoan mt genomes, which is approximately 16 kb, and they possess long noncoding stretches of DNA, have no identifiable control region, and bear additional open reading frames (ORFs)—atp9 in demosponges and 5 putative ORFs in Trichoplax.

The 3 sponge mt genomes sequenced by Lavrov et al. (2005) are from different orders (table 1) and show relatively uniform features. All possess the metazoan standard set of 13 protein genes plus atp9, 2 rRNA genes, and 24–25 tRNA genes for a complete set of amino acids (see table 1 for differences). The arrangement of the protein and rRNA genes is identical (with the only exception of nad6 in T. actinia).

Table 1

Comparison of the Demosponge mt Genome Features

	Geodia neptuni (Astrophorida), AY320032	Tethya actinia (Hadromerida), AY320033	Axinella corrugata (Halichondrida), AY791693	Amphimedon queenslandica (Haplosclerida), DQ915601
Size	18,020 bp	19,564 bp	25,610 bp	19,960 bp
Gene overlaps	4	4	0	2
Noncoding	2.1%	7.6%	23.1%	12.1%
Longest sequence	59 bp	319 bp	931 bp	1,044 bp
Repeats	—	—	—	6x12 bp
Protein genes	14	14	14	13
tRNA genes	24	24	25	17
Doub./ident.a	I, L, R, S/-	I, L, R, S/M	I, L, R, S/-	R, S/M
Missingb	—	—	—	D, H, I, L, T, V
rRNA cluster	rns-trnG-trnV-rnl	rns-trnG-trnV-rnl	rns-trnG-trnV-rnl	rns-trnG-trnF-rnl
Positionc	nad5, cox2	nad5, cox2	nad5, (nad6)-cox2	nad3, nad4l
rns term. BLd	0.067	0.086	0.046	0.193
rnl term. BLd	0.069	0.079	0.062	0.168

a

Refers to double amino acid tRNA genes with/without identical anticodon.

b

Refers to missing amino acid tRNA genes (referring to the anticodon).

c

Indicates the protein genes flanking the rRNA cluster.

d

Refers to the terminal branch lengths (BL) as calculated by p-distances from a conservative alignment.

Despite this congruence, it remains unknown if these features are shared between all poriferans. Here, we show that mt genome evolution in basal metazoans is far more complex than previously appreciated and observed in eumetazoans, by presenting the complete sequence of the demosponge Amphimedon queenslandica (appears in databases and earlier reports under its working name “Reniera sp.”) (Hooper and van Soest 2006), which is the target species for the Sponge Genome Project (http://www.jgi.doe.gov/sequencing/why/CSP2005/reniera.html). Amphimedon queenslandica is a species of the order Haplosclerida, which is a pivotal order and probably the most successful demosponge taxon with the highest biodiversity in terms of species and habitat (van Soest and Hooper 2002). The A. queenslandica mt genome (fig. 1) is a circular molecule of 19,960 bp, which is a typical length for demosponges (table 1; see supplementary table A, Supplementary Material online for methods on A. queenslandica mt genome assembly, annotation, and analysis). There are only 2 gene pairs that overlap, and most genes are irregularly interspersed by noncoding regions, which comprise about 12% of the genome (fig. 1). The longest noncoding region is 1,044 bp and is the longest yet found in Porifera. It possesses the first mt repeat sequence detected in sponge mitochondria, which, in combination with its position within the mt rRNA cluster, resembles higher metazoan control region features. Therefore, it provides the first strong evidence for the presence of a mt control region in poriferan mitochondria. No other ORFs have been confirmed existing amongst the intragenic spacers, as has been observed in anthozoan cnidarians (e.g., mutS; Pont-Kingdon et al. 1995).

FIG. 1.—

Genomic map of the A. queenslandica mitochondrium. All protein genes have the same transcriptional orientation and no introns. Start codons are ATG with the exception of cox3 (GTG) and nad6 (TTG). All codons are present in the mt ORFs, although some are rare; for example, CGT (1x), CGC (2x), and CTC (3x). Above the rRNA genes, schematic drawings of the predicted secondary structures are given (Voigt, Erpenbeck, and Wörheide, unpublished data).

Amphimedon queenslandica possesses 32 mt genes, which is the smallest gene number observed in a demosponge mt genome to date. The A. queenslandica mt DNA codes for 13 proteins and lacks atp9, as observed in many eumetazoan mt genomes; atp9 is present in other demosponge mt genomes. Cob and trnS genes, which flank atp9 in other demosponges, overlap in A. queenslandica. We found no evidence for an atp9 pseudogene in the mt genome, suggesting there has been an excision event of atp9 in the A. queenslandica lineage. The nuclear atp9 encodes a protein that is 5 amino acids shorter than the mt counterparts of other sponges and is translatable into the same functional protein with both the universal and the poriferan mt code. Interestingly, the nuclear atp9 has flanking inverted terminal repeats (ITR) regions of 63 bp that are 87% identical (fig. 2). Their presence provides evidence for the transposon-mediated transport of atp9 from the mitochondrion to the nucleus. The high similarity of the ITR regions suggests a rather recent transposition event. The ITRs are located 72 bp and 2.5 kb from atp9 and enclose putative ORFs of unknown identity. The fixation status of the atp9 transposition in this genus can be estimated in future studies among closely related species (there are >40 Amphimedon species cf. World Porifera Database; http://www.vliz.be/vmdcdata/porifera/). The transposition of atp9 to a different locus prevents its usage for demosponge mt phylogenetics (see also Delsuc et al. 2005 on phylogenomics). Other peculiar protein gene features of A. queenslandica comprise deletions in cox1, nad1, 2, 5, 6, and atp6. Such deletions are not observed in other early-branching taxa (including the other demosponges).

FIG. 2.—

(A) Arrangement of atp9 in the nuclear genome. The ITRs are flanking a region comprising atp9 and 5 yet unidentified ORFs. (B) Alignment of the flanking ITR sequences.

Amphimedon queenslandica is the most early-branching Metazoa found without a complete set of tRNAs in its mt genome. Six amino acids are not coded by tRNAs (D, H, I, L, T, and V; table 1), a feature that is only known from Cnidaria but has not previously been detected in other diploblast mt genomes. Remarkable is the lack of trnL, which is usually present in 2 copies in metazoan mt genomes, and trnI_CAU, which is an ancestral feature lost in Eumetazoa (Lavrov and Lang 2005). An A. queenslandica tRNA^Ile is unlikely to be formed out of posttranscriptional edition of 1 of the 2 trnMs as both are clearly functionally distinguished into initiator (trnMf) and elongator (trnMe; Drabkin et al. 1998). The trnM tandem arrangement is a peculiar feature and, considering their high sequence difference (51%), unlikely to be the result of a single tandem duplication.

The A. queenslandica rRNA cluster is uniquely translocated among the demosponges within the genome (table 1) and diverges from the “demosponge + choanoflagellate + ‘many bilaterians’ − motif” rns-(trnG/trnV/trnG-trnV)-rnl (Lavrov et al. 2005). Amphimedon queenslandica rRNA secondary structures are bacterialike as in all demosponges but possess several unique structural features, including extra and missing helices (O. Voigt, D. Erpenbeck, and G. Wörheide, unpublished data). The “nuclear” rRNA of Haplosclerida has previously been shown to evolve in different patterns and significantly higher rates than in other demosponge orders (Erpenbeck et al. 2004). Interestingly, similar tests in our present study reveal higher evolutionary rates also for the “mt” rRNA genes and the mt protein–coding genes (P < 0.05, table 1 and fig. 3) of A. queenslandica opposed to the nonhaplosclerid sponges. These higher rates in both Haplosclerida genomes may cause their higher adaptation potential and biological success.

FIG. 3.—

Phylogenetic reconstruction from the mt protein sequences of representative taxa by Bayesian inference. Species names are followed by the GenBank accession numbers. Numbers above the branches are Bayesian inference posterior probabilities calculated under the Cprev + G + I + F model. Numbers below the branches are maximum likelihood bootstrap probabilities (100 repeats).

The availability of mt DNA in combination with nuclear genomic traces, which revealed transposon-mediated export of genetic material from the mitochondrium in this study, makes A. queenslandica an ideal candidate for mt evolution studies. The mt genome of A. queenslandica shares features with both known poriferan and eumetazoan (excluding T. adhaerens) genomes, consisting of ancestral and derived features. However, it is clear that this demosponge is not a “missing link” between both groups (see molecular analyses in fig. 3 and morphology [Hooper and van Soest 2006]). Instead, it appears that A. queenslandica displays convergent features of lower metazoan mt evolution, which may be a reflection of its higher rates of evolution that have been likewise observed in other nuclear genomes of this order (Erpenbeck et al. 2004). Consequently, elevated evolutionary rates in combination with given genetic precursors result in the evolution of the A. queenslandica mt genome convergent to its eumetazoan counterparts. It shows how gain and loss of genes, partial genes, and (control) regions occurred multiple times, presumably even throughout smaller lineages in basal Metazoa and that one can expect far more divergence at the root of the Metazoa before obtaining a realistic picture of metazoan mt evolution. In addition, the unique gene composition and arrangement in the A. queenslandica mt genome and its surprising similarities to eumetazoan mt loci suggests that using gene order to infer early-branching metazoan relationships may be unsound (see also Delsuc et al. 2005).

We thank S. M. Degnan and an anonymous reviewer for valuable comments and acknowledge financial support from the European Union (Marie-Curie outgoing fellowship MOIF-CT-2004) to D.E., the UQ Postdoctoral Fellowship Scheme to M.A., the German Research Foundation (DFG; Projects Wo896/3,5,6) and the European Union through the Marie-Curie project HOTSPOTS (MEST-CT-2005-020561) to G.W., and grants from the US Department of Energy Joint Genome Institute through the Community Sequencing Program and the Australian Research Council to B.M.D.

Author notes

Billie Swalla, Associate Editor