In total we aligned sequences of 56 taxa (Table 1) for the large taxon set and 34 taxa for the reduced taxon set. Due to ambiguous homologisation of certain nucleotide positions in the alignments we excluded parts of high variability which were mainly due to inserts in certain taxa from the alignments. The lengths of the obtained alignments (after removal of ambiguous nucleotide positions) for the different datasets (complete taxon number versus reduced taxon number for combined analyses) are shown in Table 2.

Performance of the chi-square-test in PAUP yielded a homogeneous base composition in the 28S-alignments (P = 0.999). The 18S- (P = 0.001) and 16S- alignments (P = 0.007) showed heterogeneous base composition mainly due to the sequences of Nudipleura (Bathydoris clavigera, Dendronotus dalli, Pleurobranchus peroni, Tomthompsonia antarctica) (for 18S) and Orbitestellidae (for 16S). The COI-alignment showed heterogeneous base composition when using all three codon positions (P = 0.000). However, we identified substitution saturation in the third codon position (I_ss 0.830 >I_ss·c 0.771 for large taxon set; I_ss 0.753 >I_ss·c 0.737 for reduced taxon set) and subsequently removed third codon positions from the alignments. The P-value of the chi-square test then changed to 1.00 indicating homogeneous base composition for all taxa.

Additionally, we observed substitution saturation in the 16S-alignments (I_ss 1.071 >I_ss·c 0.784 for large taxon set; I_ss 1.02 >I_ss·c 0.750 for reduced taxon set).

The relative rate test revealed that evolutionary rates are different in the investigated taxa and genetic markers. This is especially true for the two ribosomal genes 18S rRNA and 28S rRNA, where the Nudipleura show extremely high Z-Scores (e. g. 18S: Dendronotus dalli vs Orbitestella sp. = 12.19; 28S: Dendronotus dalli vs Turbonilla sp. = 5.34).

Investigation of differences in incongruence length between the four different sequence partitions and two taxon sets revealed that combination of the partitions improves phylogenetic signal with a p value of 0.01 (10.000 replicates).

We performed different analyses which provided mostly congruent results regarding phylogenetic relationships of subgroups within Euthyneura.

Our analyses of concatenated sequences of 18S rRNA, 28S rRNA, 16S rRNA and COI of the large taxon set comprised 56 taxa representing all of the major taxa of Euthyneura (Fig. 1) and four non-euthyneuran taxa. The reduced dataset included 34 taxa with Trimusculoidea, Acochlidiacea, Thecosomata and Gymnosomata not represented (Figs. 2 and 3). Although 16S-sequences and 3^rd codon positions of the COI sequences exhibited varying degrees of substitution saturation we used all sequences in our combined analyses of the large taxon set (Fig. 1) and the reduced taxon set (Fig. 3) to avoid the loss of phylogenetic signal at lower taxonomic levels (e. g. relationships within superfamilies). In a previous study, Thollesson 18 demonstrated that 16S rRNA sequences provide useful information in recontruction of phylogenetic relationships within Euthyneura. However, to accommodate substitution saturation and to test its influence on the phylogeny inferred, we additionally performed combined analyses of the reduced dataset excluding 16S-sequences and 3^rd codon positions of COI sequences (Fig. 2).

The first striking result of this study is the inclusion of the taxon Turbonilla sp. (Pyramidelloidea) within Euthyneura as sister taxon to the Amphiboloidea (Figs. 1, 2 and 3). Within the Euthyneura Opisthobranchia are polyphyletic (Figs. 1 and 3) or paraphyletic (Fig. 2). The Opisthobranchia comprise several distinct clades. The most basal offshoot within Euthyneura is a clade composed of Acteonoidea and Nudipleura, both forming a sister group relationship. However, the internal branches leading towards Acteonoidea and Nudipleura are relatively long. Additionally, terminal branches of Bathydoris clavigera and especially Dendronotus dalli are extremely long indicating high substitution rates in these taxa. Therefore, this result should be considered with caution.

The second opisthobranch clade comprises the Cephalaspidea as sister to the Umbraculoidea plus Aplysiomorpha (Figs. 2 and 3). When Thecosomata and Gymnosomata are included (Fig. 1), they represent the immediate sister taxon to the Aplysiomorpha. The latter three taxa are sister to Umbraculoidea. The Sacoglossa are recovered closely related to the Pulmonata as sister to the Siphonariidae (Figs. 1 and 3). However, support for this sistergroup relationship is non existent in the analyses of the reduced dataset. Nevertheless, inclusion of Sacoglossa within Pulmonata is well supported in all analyses, rendering the Pulmonata paraphyletic. Inclusion of the enigmatic Acochlidiacea in the large taxon placed them within Pulmonata (Fig. 1).

Within Pulmonata we recovered monophyly for Eupulmonata. Whereas the Bayesian statistical support for a monophyletic Eupulmonata is rather low in the analysis of the large taxon set (Fig. 1) possibly due to missing data in these taxa (especially in Stylommatophora, see Table 1), the combined analyses of the reduced taxon set showed a high posterior probability (Fig. 2) and additionally high bootstrap support for a monophyletic Eupulmonata (Fig. 3). Within Eupulmonata, Onchidioidea are a well supported monophyletic clade whereas the Bayesian posterior probability for a monophyletic Stylommatophora is low in the analysis of the large taxon set probably due to the reasons mentioned above. Ellobioidea are poorly supported and their monophyly cannot be recovered unequivocally. Basommatophora are paraphyletic. Within Basommatophora, Siphonariidae and Hygrophila are well supported monophyletic taxa.

Two clades can be distinguished within Hygrophila. The first clade consists of Chilina spp. and Latia neritoides whereas the second clade is comprised of higher limnic Basommatophora. The relationships in this latter clade differ in the different analyses. Within higher limnic Basommatophora only a close relationship of Bulinus tropicus and Planorbis planorbis is consistent and well supported in all analyses.

Based on the phylogenetic hypothesis for Euthyneura deduced with a Bayesian analysis of the reduced data set we reconstructed the ancestral character states and character evolution relating to habitat types in euthyneuran subgroups.

We calculated posterior probabilities for each of the four habitat types for certain ancestral nodes within Pulmonata (Fig. 4). It can be seen from this analysis that colonisation of freshwater habitats occurred only once within Pulmonata by Hygrophila. The ancestor of Hygrophila probably already occurred in a freshwater habitat. Colonisation of terrestrial habitats in Eupulmonata has its seeds in marginal zones such as estuaries or mangroves. The ancestor of Eupulmonata, Amphiboloidea plus Pyramidelloidea and Hygrophila probably already lived in a marginal habitat as indicated by the high proportion of posterior probabilities for this habitat type at the respective node (Fig. 4). Otina ovata within Eupulmonata and Turbonilla sp. both inhabit marine environments.

With the current study we present a comprehensive molecular phylogenetic analysis of Euthyneura including representatives of all major subgroups. We used different approaches to infer a phylogenetic hypothesis for this taxon of Gastropoda. These approaches yielded for the most part constant results respective to the deeper nodes in the tree and indicated only minor differences at generic or familial level.

The unification of Opisthobranchia and Pulmonata in the group Euthyneura has been widely accepted since its original definition by Spengel 4. Consequently, the monophyly of Euthyneura has mostly been accepted in recent systematic and phylogenetic investigations 59111213152324. Although the defining character of Euthyneura, euthyneury, is considered to be a result of multiple convergence 1617, several autapomorphic characters supporting monophyly have been proposed for Euthyneura 111617.

Our results contradict a monophyletic Euthyneura, since Pyramidelloidea are included within euthyneurans as sister group to the Amphiboloidea. This is in congruence with Grande et al. 19 who found Pyramidelloidea deeply nested within Pulmonata in their molecular phylogenetic analysis. Although not having included Pyramidelloidea in his molecular phylogenetic analysis of Euthyneura, Thollesson 18 proposed a possible synapomorphy of Pyramidelloidea and Euthyneura in the presence of a gap in the helix G16 of the 16S rRNA gene, thus supporting inclusion of Pyramidelloidea within Euthyneura as proposed here. Studies considering additional pyramidelloid and other basal heterobranch taxa currently undertaken in the group of the senior author will hopefully shed more light on the position of the enigmatic Pyramidelloidea within Heterobranchia (Euthyneura plus Heterostropha sensu Ponder and Waren 10). Additionally, new character complexes, such as secondary structures of ribosomal genes, could provide further valuable phylogenetic information for the taxa in question, as has been demonstrated recently by Lydeard et al. 25.

Although the separation between Opisthobranchia and Pulmonata is rather distinct because of e. g. characters of the nervous system, it has been difficult to propose sound autapomorphies for a monophyletic Opisthobranchia 12. Although the subgroups of Opisthobranchia can be clearly distinguished and appear to be monophyletic 2126272829, monophyly of the Opisthobranchia altogether could not be shown by most recent phylogenetic analyses, regardless of whether they used morphological or molecular data 1819212230. The Opisthobranchia appear to be paraphyletic (cp. e. g. 718) rather than a natural grouping. Our data confirm this latter assumption by rendering the Opisthobranchia either polyphyletic or paraphyletic (depending on the analysis). Filtering our MCMC trees of the Bayesian analysis of the reduced dataset (including all genes) under the constraint of a monophyletic Opisthobranchia revealed that out of 100.000 trees not one single tree supports a monophyletic Opisthobranchia. Application of the Approximately Unbiased test 31 to both reduced datasets also rejected a monophyletic Opisthobranchia at the 5% significance level. In contrast, we can distinguish several clades within Opisthobranchia clearly distinct from each other gradually leading to the pulmonate level of organisation.

One such clade comprises the Nudipleura and Acteonoidea which in our trees appears at the base of the Euthyneura. Acteonoidea have traditionally been regarded as the most primitive opisthobranchs or have even been excluded from Opisthobranchia 222732 whereas Nudipleura are derived 2122. However, this unexpected grouping of both taxa has recently continuously been revealed by molecular systematic studies. 202133. Nevertheless, statistical support in our analyses is rather low and good synapomorphies for this sister group relationship are still warranted. Moreover, we observed deviant base composition and rate heterogeneity in Nudipleura which also could attribute to the basal position of this taxon and possibly artificially groups them with Acteonoidea.

A second well supported opisthobranch clade in the current analysis of the large taxon set comprises the Cephalaspidea, Umbraculoidea and Aplysiomorpha plus Thecosomata and Gymnosomata (Pteropoda). This close grouping of Cephalaspidea, Aplysiomorpha and Pteropoda has been suggested lately by several studies 1733. The position of Umbraculoidea within this clade, although well supported in our current analyses, should be considered with caution since only one umbraculoid taxon has been included. Moreover, no morphological synapomorphies are known to support a close relationship of Umbraculoidea and Aplysiomorpha/Pteropoda. More umbraculoid taxa must be considered in future studies to clarify the phylogenetic position of this taxon.

The phylogenetic position of Sacoglossa within Opisthobranchia/Euthyneura has been a matter of debate 202122. Although morphologically well defined as a monophylum 2627, different analyses assign Sacoglossa equivocally to different clades within the Euthyneura. Our data suggest a close affinity to primitive Pulmonata. However, Sacoglossa are definitely a taxon that needs to be paid more attention to in future phylogenetic studies.

The same holds true for the Acochlidiacea, an enigmatic taxon inhabiting interstitial marine habitats and even freshwater systems, a feature unique within Opisthobranchia. Due to their very small size (often less than 1mm in length) hardly anything is known about their morphology and life history. Recent morphological investigations utilising modern computer generated 3-D reconstructions 3435 have shed some light on the anatomy of these enigmatic gastropods, however, more studies are warranted. The phylogenetic position of the Acochlidiacea is still unclarified. Vonnemann et al. 21 found them as a basal taxon within Opisthobranchia, while our results imply a position within Pulmonata.

The monophyly of Pulmonata is widely accepted on the basis of morphological characters 67111316. In contrast, our data suggest paraphyly of Pulmonata due to the inclusion of Sacoglossa, Pyramidelloidea and possibly Acochlidiacea. If our phylogenetic hypothesis is correct, one must postulate that pulmonate autapomorphies (such as acquisition of a pneumostome and pulmonary vessels as well as the presence of a procerebrum and dorsal bodies) have secondarily been reduced in Sacoglossa, Acochlidiacea and Pyramidelloidea.

Paraphyly or even polyphyly of Pulmonata was recovered by molecular systematic studies 141936.

Within Pulmonata we can distinguish two monophyletic clades: Eupulmonata and Hygrophila. The unification of Siphonarioidea and Amphiboloidea in the Thalassophila based on morphological data 13 could not be confirmed by our data.

The Eupulmonata comprise a group of marine, semi-terrestrial and truly terrestrial gastropods showing a high diversity with regards to species number and life history characters. Monophyly of this taxon is strongly supported by Wade and Mordan 15 and the present results, however no morphological apomorphy is known to date 910. Relationships within Eupulmonata have been disputed in the past. Monophyly of Stylommatophora is undoubted 111314151737 and can be supported by several autapomorphies 1113. Our results also show a monophyletic Stylommatophora. Haszprunar and Huber 16 proposed a close relationship of Ellobiidae, Trimusculidae and Stylommatophora on the basis of apomorphic characters of the nervous system, whereas Dayrat and Tillier 1112 consider Stylommatophora and Systellommatophora (Onchidioidea and Veronicelloidea) to be closely related. Neither of these two hypotheses are confirmed by our findings since we find a close relationship between ellobioid taxa, Otinoidea, Onchidioidea and Trimusculoidea. The phylogenetic status of the Ellobioidea cannot be conclusively clarified from our data since they appear to be paraphyletic in the two analyses utilising all sequence data. Exclusion of 16S-sequences and 3^rd codon positions of COI-sequences, however, recovers them monophyletic.

The Basommatophora sensu Nordsieck 13 and Bouchet et al. 38 are paraphyletic in our analyses. No topology in our set of 100.000 MCMC trees of the Bayesian analysis of the reduced dataset supports a monophyletic Basommatophora. Moreover, testing an alternative tree topology with a constrained monophyletic Basommatophora with the Approximately Unbiased test 31 rejected monophyly of Basommatophora at the significance level of 5%. This is not surprising since the monophyly of Basommatophora has already been disputed by Hubendick 39. Within Basommatophora the Hygrophila are a well supported and defined group of pulmonates. Our results support earlier hypotheses of a common origin of these freshwater taxa 12131740. For the first time, however, crucial taxa like Latia are included in a molecular study. Within Hygrophila two clades can be distinguished; on the one hand the Chilinoidea (sensu Boss 3, including Chilinidae and Latiidae), and on the other the higher limnic Basommatophora. Chilina and Latia are sister-species, a relationship that has already been mentioned earlier 133941. Higher limnic Basommatophora can be divided into four well distinguishable families 39 while interfamilial relationships of these taxa vary in the current analyses thus rendering them beyond the scope of this study. These relationships are discussed in other publications 4243.

Former taxonomic classifications of Pulmonata united a group of primitive taxa (Ellobiidae, Otinidae, Amphibolidae, Siphonariidae, Chilinidae and Latiidae) in the Archaeopulmonata 44. Other authors followed this classification 6345. However, our results clearly indicate a paraphyletic nature of these taxa and therefore we propose the disuse of the name Archaeopulmonata.

One key step in the evolution of Euthyneura was the invasion of freshwater and terrestrial habitats by Pulmonata resulting in a multitude of taxa worldwide adapted to these habitats. It is undisputed that the first pulmonate gastropods were marine organisms 1246. Marine pulmonates can be found in truly marine submersed environments (such as Williamia) and, more frequently, in upper littoral zones (Siphonaria, Myosotella, Trimusculus). Certain taxa even occur in estuaries or mangroves (Onchidiidae, certain Ellobiidae, Amphibolidae). Terrestrial pulmonates inhabit a wide spectrum of habitats, ranging from caves to lowlands to mountains and from boreal to tropical ecological zones. The same holds true for freshwater taxa, which occur in different types of limnic environments worldwide.

Our reconstruction of character evolution relating to habitat type (Fig. 4) shows that within Pulmonata the freshwater habitat has only been conquered once by the Hygrophila. Calculation of posterior probabilities for the different habitat types at specific nodes indicates that the ancestor of Hygrophila probably already lived in a freshwater habitat. Acknowledging that there are extremely few onchidioidean species reported to live in brackish or freshwater it would not change the picture that the major radiation into freshwater happened only once leading to the Hygrophila. The same is true for the few acochlidiaceans that are known from freshwater habitats. Freshwater members of both taxa were not included in our data set, for the time being, we would also hypothesize a similar evolutionary pathway excluding a terrestrial step.

In non-pulmonates, the situation is different. 'Prosobranchs', such as the superfamily Cerithoidea, comprise predominantly marine taxa. Nevertheless, freshwater was invaded independently in several cerithoid lineages 47. Even the evolution of life history traits such as viviparity in the freshwater invaders was correlated with this evolutionary step 48.

The ancestor of Eupulmonata and Hygrophila, in contrast, appears to have lived in a marginal zone like supralittoral zones, estuaries or mangroves. Therefore we conclude, that colonisation of freshwater in Pulmonata occurred via an aquatic pathway directly from the marine realm and not via a terrestrial step.

The terrestrial habitats have most probably also been invaded via marginal zones, as can be seen from the Bayesian reconstruction of ancestral character states in Eupulmonata. Involvement of freshwater systems in colonisation of terrestrial habitats is less likely, since the freshwater taxa (Hygrophila) are clearly separated from Eupulmonata and the posterior probability for a freshwater habitat is basically non-existant at the base of the Eupulmonata.

Terrestriality on the contrary is thought to have evolved much more than 10 times independently (discussed in Barker 49). Barker 49 already emphasized that recent ellobiid terrestriality was clearly derived from marine littoral ancestors and not secondarily from terrestrial ancestors.

Invasion of marine habitats by Otina ovata within Eupulmonata and Turbonilla sp. obviously is the result of a secondary colonisation originating in marginal zones.

A total of 53 representative taxa of most major extant lineages of Euthyneura were included in the current phylogenetic study. Additionally, we included two basal heterobranch species (Turbonilla sp. and Orbitestella sp.) as well as the caenogastropod Littorina littorea. Specimens were collected from different locations worldwide (for details see Table 1). They were collected from the field by hand, snorkelling or scuba diving, preserved in 80–100% ethanol and stored cooled until further processing.

For some taxa (Stylommatophora, Amphibolidae, Ellobiidae, Opisthobranchia and Littorinidae) we obtained published sequences from Genbank). Accession numbers of all sequences used in the analyses are listed in Table 1.

Genomic DNA was extracted from muscle tissue or entire animals using the DNeasy Tissue Kit (Qiagen, Hilden, Germany) or the protocol given in 43. Voucher depositions are listed in Additional file 1. We amplified fragments of four genes, including 18S rDNA, 28S rDNA, 16S rDNA and COI. For details of primers used and PCR protocols see Additional file 2.

After the amplification products were purified from an agarose gel using the QIAquick Gel Extraction Kit from Qiagen (Hilden, Germany). The DNA was subjected to cycle sequencing using the ABI Prism Big Dye terminator kit (Perking-Elmer, Norwalk, CT, USA) or the CEQ DTCS Quick Start Kit (Krefeld, Germany). DNA sequences were obtained with an ABI 377 automated DNA sequencer at the Scientific Research Lab, Frankfurt/Main and by using a CEQ 2000 Beckmann Coulter sequencer. All fragments were sequenced in both directions using the same primer sequences as used for the PCR. For 18S and 28S several internal primers were used (Additional file 3).

Sequence alignments were performed with CLUSTAL W implemented in the software package BioEdit Version 7.0.5 50. Default parameters were used but subsequent manual correction was necessary and was performed with BioEdit. Table 2 provides an overview on the number of taxa and lengths of the sequence alignments used in the analyses. Several base positions have been excluded from the alignments of 18S rRNA, 28S rRNA and 16S rRNA (for details see Table 2) prior to reconstructing phylogeny due to inserts in certain taxa or due to high variability within the alignments (especially in the 28S- and 16S-fragment).

We investigated base compositions by means of the software PAUP 4.0b10 51. To test the degree of substitution saturation we used the test of Xia et al. 52, as implemented in the software package DAMBE 53. A relative rate test was performed with the software k2WuLi 54 in order to test for rate heterogeneity in the sequences. In order to examine whether there are significant differences in incongruence length between the four datasets, a homogeneity-partition test implemented in PAUP 4.0b10 51 was performed. Alternative tree topologies were tested for the reduced taxon sets using the Approximately Unbiased (AU) test 31. The likelihood at each nucleotide position was calculated for each alternative topology (constrained monophyletic Opisthobranchia, constrained monophyletic Basommatophora) as well as the topology under scrutiny using PAUP 4.0b10 51. Likelihoods were used to calculate p-values using CONSEL version 0.1 55.

Prior to phylogenetic analyses, we determined the optimal model of sequence evolution for each of the four partitions and each of the taxon sets using Modeltest 3.7 56 based on the Akaike information criterion (for details of the models determined refer to Table 2).

We applied several different analyses to our data. First, we used a concatenated alignment of all four partitions including all taxa for which sequences were available (large taxon sets). Missing sequences for certain taxa in certain partitions were coded as missing data in the respective alignments.

Secondly, we also followed a more conservative approach by analysing a combined dataset of all four partitions with only those taxa for which sequences of all partitions were available (reduced taxon set, 34 taxa). For all analyses Bayesian inference methods were applied with Mr Bayes 3_03b 57, using separate models of evolution for each of the four partions. The tree space was explored using four chains of a Markov Chain Monte Carlo algorithm (one cold, three heated) for 1,000000 generations. Likelihoods of the trees started at around -67470 and quickly converged upon a stable value of about -54600 after approximately 7,000 generations. Thus we computed a 50% majority rule consensus tree with the first 1,000 trees (= 10,000 generations) ignored as burn-in. Posterior probabilities were calculated for each node in the tree. We consider a BPP value of 0.95 and higher to give good support to a node.

Additional Maximum likelihood analyses were conducted for all data sets using RAxML-VI-HPC version 4.0.0 58). This program allows analyzing large combined data sets and uses a GTR model based approach with internal estimation of all free model parameters for all data partitions. Bootstrapping with 100 replicates was performed.

Trees were rooted with the caenogastropod Littorina littorea.

We reconstructed character evolution regarding habitat types with a Bayesian approach as implemented in the BayesTraits software package 59. The BayesMultiState option allows to automatically find posterior distribution of models of evolution using the reversible-jump MCMC method. Ancestral character states were reconstructed for selected nodes in the phylogeny resulting in respective probabilities of all character states for that node. Data on habitat types colonised by the investigated taxa were taken from the literature. Information on coding for individual taxa can be taken from Additional file 4. Rate deviation was set to 8.0. A hyperprior approach was used with exponential prior seeded from a uniform on the intervall 0 to 30. Thus, acceptance rates between 20 and 30 % were achieved, which are in the preferred range.