Coming out of your shell or crawling back in: multiple interphylum host switching events within a clade of bivalve- and ascidian-associated shrimps (Caridea: Palaemonidae)

Taxon sampling

A total of 55 ingroup and five outgroup species were included in this study (supplementary appendix S1; table 1, supplementary table S1). Of those included in the ingroup, 39 species have their distribution in the Indo-West Pacific region (iwp), eight in the East Pacific region (ep), and another eight in either the West and/or East Atlantic Ocean (W/E Atl) (table 1). Five outgroup taxa were selected to span across the spectrum of hosts within the symbiotic Palaemonidae, representing echinoderm, sponge and cnidarian associates and one mostly free-living species (supplementary table S1). These are Actinimenes inornatus (Kemp, 1922), Actinimenes ornatus (Bruce, 1969), Cuapetes tenuipes (Borradaile, 1898), Periclimenes colemani Bruce, 1975 and Typton wasini Bruce, 1977. With the exception of C. tenuipes, all these outgroup species fall within Clade 5 of Horká et al. (2016), which is the sister clade of the present clade of focus. All outgroup species have an iwp distribution.

Fresh specimens were collected by scuba diving or other standard collection methods (supplementary table S1). Recently collected specimens were directly stored in 96% ethanol for dna barcoding and identification. Subsampling was done by removing the left second and third pleopod or left third and fourth pereiopod of each specimen. After subsampling, specimens were moved to 70% ethanol for morphological analysis and long-term storage in the Naturalis Biodiversity Center decapod collection (rmnh.crus.d.; Leiden, The Netherlands) or the Museum Zoologicum Bogoriense (mzb.; Bogor, Indonesia). The newly sequenced material consists of 24 species, spread over seven genera. In addition to the newly generated sequences, sequence data from GenBank (Sayers et al., 2020) was used (supplementary table S1).

dna extraction, amplification and sequencing

Total genomic dna was extracted from the subsampled tissue using the DNeasy Blood & Tissue Isolation Kit (qiagen), according to the manufacturer’s protocols for animal tissue (insects). Elution was performed twice using 150 μl Buffer ae instead of using 200 μl Buffer ae once, to increase overall dna yield and dna concentration. Partial segments of four genes were amplified by pcr: mitochondrial genes for cytochrome c oxidase subunit I (coi, 658 bp) and 16S rRNA (16S, ca. 501 bp); and nuclear genes for histone 3 (H₃, 310 bp) and 18S rRNA (18S, ca. 663 bp), partially based on the study of Horká et al. (2016). Polymerase chain reactions (pcr s) for the markers coi, 16S and 18S were performed in 25-μl volumes following Fransen & Reijnen (2012) and Brinkman & Fransen (2016) containing 1 μl of dna template, 1.00 μl (10 pMol/μl) forward and reverse primers, 0.5 μl (2,5 mM) dNTPs, 0.25 μl (5 units/μl) Qiagen Taq dna polymerase, 2.5 μl (10×) CoralLoad pcr Buffer, and 18.5 μl Ultrapure MilliQ (H₂O). pcr s performed for the marker H₃ differed slightly, with the addition of 5 μl Qiagen Q-solution (5×), accounting for the concentration with 13.8 μl Ultrapure MilliQ (H₂O). The mitochondrial genes for coi and 16S rRNA were amplified using the universal primer pairs lco 1490/hco 2189 (Folmer et al., 1994) and 16Sar/16Sbr (Palumbi et al., 1991) respectively. The nuclear genes for H₃ and 18S rRNA were amplified using the primer pairs H₃F/H₃R (Colgan et al., 2008) and 18Sa2.0/18S9r (Whiting, 2002). The pcr program for the amplification of the genes coi, 16S and 18S was as follows: 3 min at 96°C for initial denaturation, followed by 40 cycles of denaturing for 10 s at 96°C, annealing for 1 min at 50 °C, and extension for 1 min at 72°C, and a final extension step at 72°C for 5 min. The pcr program for the amplification of H₃ was as follows: 3 min at 96°C for initial denaturation, followed by 40 cycles of denaturing for 10 s at 96°C, annealing for 1 min at 48°C, and extension for 1 min at 72°C, and a final extension step at 72°C for 5 min. pcr products lengths were checked using Thermofisher sybr Safe dna stained 2% agarose gels. Sanger sequencing reactions were performed by BaseClear B.V., Leiden, The Netherlands.

Morphological character state analysis

All ingroup and outgroup species were scored morphologically, based on descriptions and illustrations in previous literature, as well as specimens present in the Naturalis collections (supplementary table S1). Specimens were studied with a dissecting stereomicroscope (Zeiss Discovery.V8) and a compound microscope (Olympus bx53). The morphological data matrix contains 85 characters (supplementary appendix S2, supplementary table S2), and was formatted based on previous analyses by Fransen (2002) and De Gier & Fransen (2018) on a subset of the species in the current analysis. Multistate characters 3, 4, 12, 14, 15, 18, 19, 23, 28, 29, 36, 47, 49, 50, 64, 66, 68, 70, 76–78 and 80 were treated as ordered (Fitch, 1971), while characters 1, 2, 5, 6, 11 and 16 were considered irreversible (Camin & Sokal, 1965) and thus accompanied by a step matrix (using character states A, B, C, and D, instead of numbers, following Fransen (2002)). The remaining characters were treated as unordered (Farris, 1970). Character states A and 0 were generally attributed to the outgroup taxon Cuapetes tenuipes, unless in the case of an assumed linear transformation series in characters 47, 68, 70 where the species does not exhibit one of the extremes, and when the character state in C. tenuipes was unknown (characters 23–26, 29) or inapplicable (characters 57, 58, 60, 84, 85). Unknown character states were treated as missing data (= ?) and inapplicable characters were coded with gaps (= -). The character states, step-matrices and resulting morphological dataset can be found in the appendices (supplementary appendix S2, supplementary table S2).

Phylogeny reconstructions

Multiple sequence alignments were obtained using the ClustalW (Thompson et al., 1994) algorithm (default settings), and their ends were trimmed manually in mega v10.2.5 (Kumar et al., 2018). Alignments of protein-coding genes coi and H₃ were subjected to Xia’s test of nucleotide substitution saturation (Xia et al., 2003) in dambe v7.3.0. (Xia, 2009). Only the third codon position of coi was found to be saturated and thus excluded, resulting in a total of 395 positions being included (table 2). After aligning, non-coding ribosomal 16S and 18S genes were examined for highly divergent blocks using Gblocks v0.91b (Talavera & Castresana, 2007) using default parameters but allowing gap positions, smaller finer blocks, and less strict flanking positions in both 18S and 16S. This resulted in the retention of 436 positions in 16S and 650 positions in 18S. Models for sequence evolution were calculated in mega v10.2.5 (Kumar et al., 2018) using default settings (table 2). Best models were chosen based on values for the Akaike Information Criterion (aic) and Bayesian Information Criterion (bic).

View Table

A concatenated dataset containing all molecular markers was constructed in mega v10.2.5 (Kumar et al., 2018). This alignment consisted of 105 sequences and 1791 positions (table 2). Phylogenetic analyses were performed under the maximum likelihood (ml) and Bayesian Inference (bi) criteria using RAxML via the on-line cipres (Miller et al., 2010) with the RAxML-ng BlackBox v1.0.0 tool (Kozlov et al., 2019) and MrBayes v3.2.7 (Ronquist & Huelsenbeck, 2003), respectively. For the single-marker ml analyses default settings were used and automatic bootstrapping with a cut-off of 0.03, under an unpartitioned model. The concatenated four-marker dataset used a partitioned model with scaled branch length linkage and automatic bootstrapping with a cut-off of 0.03. Rest of the settings were default. The bi analyses were conducted using a Markov Chain Monte Carlo (mcmc) method with two independent runs and four chains. Analyses were run with a minimum of 5.000.000 generations to ensure that the average standard deviation of split frequencies (asdsf) reached a value < 0.01, which would indicate the two runs have converged to a stationary distribution. The bi analysis on the concatenated molecular dataset did not reach this threshold after an additional increase of 15.000.000 generations (making a total of 20.000.000 generations) and was terminated with a asdsf value of 0.010234. Trees were sampled each 500 generations. The initial 25% of trees was discarded as burn-in and the remaining trees were used to generate consensus trees and to estimate Bayesian posterior probabilities (pp). The bi analysis was consequently examined using mcmc Convergence Diagnostics in Tracer v.1.7.1 (Rambaut et al., 2018), where the 0.25 burn-in fraction was taken into account. Closer inspection of the trace plots of the log-likelihood indicated that convergence was reached, with an effective sample size (ess) of 3477.3.

The morphological data matrix was analysed in paup v4.0a (Swofford, 2003) under the maximum parsimony (mp) criterion, using ordered characters and step matrices as mentioned above. A heuristic search was performed with 1000 repetitions. The strict consensus and the 50% majority rule consensus trees were obtained.

A total evidence (te) analysis was done, similar to the one used in Van Der Wal et al. (2019). For this approach, concatenation of the molecular and morphological data was performed manually using a standard text editor. This alignment consisted of 66 sequences and 1876 positions (see table 2). All species were represented by morphological data, and where possible, with a set of sequences of all four available markers. If a species was found to be represented in multiple lineages in the previous single-markers analyses, all lineages were represented with their own sequences, in addition to a copy of the morphological data for this species. The selected sequences used in the te analysis can be found in the appendices (supplementary table S1). Models for sequence evolution were as defined previously (table 2). Ordered and irreversible states were treated accordingly in MrBayes (Ronquist & Huelsenbeck, 2003), using the command ctype (table 2).

All trees were manually rooted by the outgroup taxon Cuapetes tenuipes and visualized in FigTree v1.4.4 (Rambaut, 2009). The resulting phylogenetic trees of the analyses based on single-gene markers (coi, H₃, 16S, and 18S) can be found in the appendices (supplementary fig. S1), while the others can be found in the results (see below). The trees in the results were edited in Adobe Illustrator cc (Adobe Systems, USA) to match the colour scheme of the phylogenetic trees in previous studies (Horká et al., 2016; Chow et al., 2021).

Biogeographical analysis

In order to reconstruct the ancestral biogeographic ranges of the studied ingroup, broad current geographical distributions were obtained from taxonomic and ecological literature (supplementary appendix S1). Distributions were classified as mentioned above (ep, iwp, E Atl, W Atl). The resulting ml tree from the te analysis was used as a backbone for figures, but an ultrametric version of the tree was used during the analyses. The ultrametric version of the phylogenetic tree was made using the force.ultrametric() function of the R-package phytools (Revell, 2012). Three different models of biogeographic evolution were assessed using the R-package BioGeoBEARS (Matzke, 2013), similar to other studies focussing on other marine organisms (e.g., Baraf et al., 2019). The following models were used for comparison: Dispersal-Extinction-Cladogenesis (dec) (Ree & Smith, 2008); Dispersal-Vicariance Analysis (diva) (Ronquist, 1997); and Bayesian inference of historical biogeography for discrete areas (bayarea) (Landis et al., 2013). aic scores were used to choose the best-fitting model, and can be found in the appendices (supplementary table S3). Given the recent concerns regarding the addition of the founder speciation event parameter “J” (Ree & Sanmartín, 2018), the current analyses resort to the most commonly used models for comparison without considering “jump” (e.g., Baraf et al., 2019). For all analyses, a restricted model was used to limit the movement of ranges to only adjacent areas. This was done using a matrix file, as described in the BioGeoBEARS manual (Matzke, 2013). The resulting pie-charts displaying the ml values were edited and added to the phylogeny in Adobe Illustrator cc (Adobe Systems, USA). ml-values for all probabilities can be found in the appendices (supplementary table S4).

Ancestral state reconstruction of host associations

To reconstruct the evolution of host-associations, an ancestral character state reconstruction was performed in R using the mp criterion and the ml tree from the te analysis as a backbone. The R-packages ape v. 5.5 (Paradis et al., 2019) and Geiger 2.0.7 (Penell et al., 2014) were used. Pie-charts displaying ml-values similar to the ones seen in comparable studies (e.g., Salis et al., 2018) were added in Abode Illustrator cc (Adobe Systems, USA). ml-values for all probabilities can be found in the appendices (supplementary table S4).

Molecular phylogeny reconstructions

Three major clades can be distinguished in the ingroup, with most deeper nodes bearing support values lower than 50 (fig. 2). Various intrageneric branches within Conchodytes, Ascidonia and Pontonia were recovered with both high bootstrap and Bayesian pp support values. All outgroups were separated from the ingroup, except for one Periclimenes colemani specimen of which only a 18S sequence was available, which nested within a clade of Dactylonia (see discussion). One of two branches of Platypontonia hyotis Hipeau-Jacquotte, 1971 was the most basal ingroup lineage, while the other branch seems to be a basal branch within the first clade (see below). The placement of P. hyotis is debated in the discussion.

Phylogeny based on the RAxML tree topology of the concatenated molecular dataset (coi, H₃, 16S, 18S). RAxML bootstrap support and Bayesian posterior probabilities expressed as percentages are indicated respectively. Dashes (--) indicate values <50; asterisk (*) indicates different topology of RAxML or MrBayes tree. Four branches are shortened for convenience. Three major clades can be recognized. Newly acquired barcodes are indicated with a collection accession number (rmnh.crus.d., mzb.), otherwise GenBank accession numbers are given. The selection of species can be found in the appendices (supplementary table S1). Colours indicate various host associations.

Citation: Contributions to Zoology 2022; 10.1163/18759866-bja10030

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Phylogeny based on the RAxML tree topology of the concatenated molecular dataset (coi, H₃, 16S, 18S). RAxML bootstrap support and Bayesian posterior probabilities expressed as percentages are indicated respectively. Dashes (--) indicate values <50; asterisk (*) indicates different topology of RAxML or MrBayes tree. Four branches are shortened for convenience. Three major clades can be recognized. Newly acquired barcodes are indicated with a collection accession number (rmnh.crus.d., mzb.), otherwise GenBank accession numbers are given. The selection of species can be found in the appendices (supplementary table S1). Colours indicate various host associations.

Citation: Contributions to Zoology 2022; 10.1163/18759866-bja10030

• Download Figure
• Download figure as PowerPoint slide

Phylogeny based on the RAxML tree topology of the concatenated molecular dataset (coi, H₃, 16S, 18S). RAxML bootstrap support and Bayesian posterior probabilities expressed as percentages are indicated respectively. Dashes (--) indicate values <50; asterisk (*) indicates different topology of RAxML or MrBayes tree. Four branches are shortened for convenience. Three major clades can be recognized. Newly acquired barcodes are indicated with a collection accession number (rmnh.crus.d., mzb.), otherwise GenBank accession numbers are given. The selection of species can be found in the appendices (supplementary table S1). Colours indicate various host associations.

Citation: Contributions to Zoology 2022; 10.1163/18759866-bja10030

• Download Figure
• Download figure as PowerPoint slide

The first clade contains the Atl/ep genera Ascidonia and Pontonia, the iwp genus Anchiopontonia, and one branch of Platypontonia hyotis. Although some of the intrageneric species relations of this clade are resolved with high bootstrap or Bayesian pp values, the topology of the deeper nodes is resolved with very low bootstrap values and different topologies in the Bayesian analysis. Basally, the second branch of Platypontonia hyotis and a branch with five sequences of Anchiopontonia hurii (Holthuis, 1981) are split off from the first clade respectively, after which the monophyletic genus Ascidonia (consisting of Ascidonia miserabilis (Holthuis, 1951), and four sequences of Ascidonia quasipusilla (Chace, 1972)) appears to be sister of the monophyletic genus Pontonia. Within Pontonia, various species seem to be phylogenetically paraphyletic: Pontonia manningi Fransen, 2000 can be found in a basal branch also containing one sequence of Pontonia margarita Smith, 1869, and in more derived branches, of which one is also containing an unidentified species of Pontonia; Pontonia mexicana Guérin-Méneville, 1855 can be found in two branches, one of which is grouped basally to the more derived species like Pontonia pinnophylax (Otto, 1821), and one sequence was grouped within the most derived branch, also containing sequences of Pontonia domestica Gibbes, 1850 and P. pinnophylax; and P. pinnophylax itself can be found in two groups, albeit both in the most derived branches of the genus. Whether these paraphyletic lineages can be attributed due to misidentifications, differences in distribution, or host choice is discussed below (see Discussion).

The second clade appears to be a sister to the third, much bigger clade. Species relations within these clades are not well resolved, with either low support values, different topologies in the Bayesian analysis, or both. The second clade contains all species of Dactylonia (except for one sequence of Dactylonia holthuisi Fransen, 2002) and (the monotypic genus) Cainonia, as well as a sequence of Odontonia plurellicola De Gier & Fransen, 2018 and of one specimen of the outgroup species Periclimenes colemani. Cainonia appears to be a sister clade to the rest of the species, although one sequence of Dactylonia holthuisi is grouped together with the only species of Cainonia. The other sequences of D. holthuisi within the second clade appear to be grouped together, with the inclusion of two sequences of Dactylonia okai (Kemp, 1922), the outgroup species Periclimenes colemani, and more basally, one sequence of Dactylonia ascidicola (Borradaile, 1898). The other branch of Dactylonia contains all other sequences of D. ascidicola, albeit in two to three different branches. One of these branches also contains two unidentified species of Dactylonia, the 18S sequence of Odontonia plurellicola, and one sequence of Dactylonia anachoreta (Kemp, 1922).

Lastly, all but one species of Odontonia, all of Conchodytes and one sequence of Dactylonia holthuisi form the largest third clade. Most basally, a branch containing eight sequences of Odontonia rufopuncata Fransen, 2002 splits off, after which two branches appear: one containing the remaining sequences of Odontonia (and one sequence of Dactylonia holthuisi (see Discussion)), and the other containing all sequences of Conchodytes. Odontonia seems to be split up further into two main branches, one of which is containing basally Odontonia seychellensis Fransen, 2002, with Odontonia katoi (Kubo, 1940) and O. kerangcaris splitting up as sister species. The other branch contains Odontonia sibogae (Bruce, 1973), with Odontonia bagginsi De Gier & Fransen, 2018 as its sister species. Conchodytes appears to be the most well-resolved, with various branches being resolved with high bootstrap support values. Conchodytes nipponensis (De Haan, 1844) and Conchodytes tridacnae Peters, 1852 both respectively split off first, after which the genus is subdivided into two major branches: one branch containing (from basally branching off to more derived) Conchodytes pteriae Fransen, 1994, Conchodytes chadi (Marin, 2011), and Conchodytes placunae (D. S. Johnson, 1967). The branch ends in a split with Conchodytes monodactylus Holthuis, 1952, and (four sequences of) Conchodytes biunguiculatus (Paulson, 1875) being recovered as sister species. One other sequence of C. biunguiculatus can be found nested in the other major branch, between the two branches of Conchodytes meleagrinae Peters, 1852.

Morphological phylogeny

The morphological analysis recovered all the genera as monophyletic (fig. 3). Cuapetes tenuipes, Actinimenes inornatus, A. ornatus and Periclimenes colemani were recovered as outgroup species, while Typton wasini clustered with the ingroup species Colemonia litodactylus Bruce, 2005 and Bruceonia ardeae (Bruce, 1981) in the 50% majority consensus tree (fig. 3). Five distinct, well-supported clades could be recognized, with the monotypic iwp genera Rostronia and Anchiopontonia branching off basally, accompanied by high (100) support values. The first clade contains the Atl/ep genera Pontonia and Ascidonia that together form a well-supported lineage (100). Branching off this first clade is the second well-supported (100) clade, containing the iwp genera Cainonia and Dactylonia. Sister of the first two clades are the following three clades, in succession: the third well supported (100) clade containing the outgroup species Typton wasini basally, and the two monotypic iwp genera Bruceonia and Colemonia; the fourth well supported (100) clade contains the two iwp genera Pseudopontonia and Opaepupu basally, and all species of Odontonia; the fifth and last clade contains Notopontonia most basally, after which Platypontonia and later Pinnotherotonia branched off, the latter forming a direct sister to Conchodytes.

The 50% majority rule consensus tree of the morphological analysis in paup. Five distinct clades can be recognized. Support values are given at every dichotomous or polytomous branching. Colours indicate various host associations. The host association of Pontonia longispina Holthuis, 1951 is unknown, indicated with a question mark (?), and the known host association of Pontonia chimaera Holthuis, 1951 is with gastropod molluscs, which is indicated with an outline of a shell.

Citation: Contributions to Zoology 2022; 10.1163/18759866-bja10030

• Download Figure
• Download figure as PowerPoint slide

View Full Size

The 50% majority rule consensus tree of the morphological analysis in paup. Five distinct clades can be recognized. Support values are given at every dichotomous or polytomous branching. Colours indicate various host associations. The host association of Pontonia longispina Holthuis, 1951 is unknown, indicated with a question mark (?), and the known host association of Pontonia chimaera Holthuis, 1951 is with gastropod molluscs, which is indicated with an outline of a shell.

Citation: Contributions to Zoology 2022; 10.1163/18759866-bja10030

• Download Figure
• Download figure as PowerPoint slide

The 50% majority rule consensus tree of the morphological analysis in paup. Five distinct clades can be recognized. Support values are given at every dichotomous or polytomous branching. Colours indicate various host associations. The host association of Pontonia longispina Holthuis, 1951 is unknown, indicated with a question mark (?), and the known host association of Pontonia chimaera Holthuis, 1951 is with gastropod molluscs, which is indicated with an outline of a shell.

Citation: Contributions to Zoology 2022; 10.1163/18759866-bja10030

• Download Figure
• Download figure as PowerPoint slide

Two somewhat supported polytomies can be found in the morphological phylogeny: one in the genus Conchodytes, branching off into six lineages (of which three contain two species); and one in the genus Pontonia, branching off to four lineages (of which one contains two species, and one three species).

Total Evidence analysis

Four clades can be distinguished based on the te approach (fig. 4). Unfortunately, the deeper nodes in the phylogeny had support values < 50 and inferences from these nodes should thus be made with caution. The in- and outgroup were confidently separated, illustrated by both high bootstrap and Bayesian pp support values. Colemonia litodactylus was the most basal ingroup lineage.

Phylogeny based on the RAxML tree topology of the te approach. RAxML bootstrap support and Bayesian posterior probabilities expressed as percentages are indicated respectively. Dashes (--) indicate values <50; asterisk (*) indicates different topology of RAxML or MrBayes tree. Four major clades can be recognized. Newly acquired barcodes are indicated with a collection accession number (rmnh.crus.d., mzb.), otherwise GenBank accession numbers are given. The selection of species can be found in the appendices (supplementary table S1), and species of which only morphological data was analysed are indicated with an asterisk (*) and no accession number. Colours indicate various host associations. The host association of Pontonia longispina Holthuis, 1951 is unknown, indicated with a question mark (?), and the known host association of Pontonia chimaera Holthuis, 1951 is with gastropod molluscs, which is indicated with an outline of a shell.

Citation: Contributions to Zoology 2022; 10.1163/18759866-bja10030

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Phylogeny based on the RAxML tree topology of the te approach. RAxML bootstrap support and Bayesian posterior probabilities expressed as percentages are indicated respectively. Dashes (--) indicate values <50; asterisk (*) indicates different topology of RAxML or MrBayes tree. Four major clades can be recognized. Newly acquired barcodes are indicated with a collection accession number (rmnh.crus.d., mzb.), otherwise GenBank accession numbers are given. The selection of species can be found in the appendices (supplementary table S1), and species of which only morphological data was analysed are indicated with an asterisk (*) and no accession number. Colours indicate various host associations. The host association of Pontonia longispina Holthuis, 1951 is unknown, indicated with a question mark (?), and the known host association of Pontonia chimaera Holthuis, 1951 is with gastropod molluscs, which is indicated with an outline of a shell.

Citation: Contributions to Zoology 2022; 10.1163/18759866-bja10030

• Download Figure
• Download figure as PowerPoint slide

Phylogeny based on the RAxML tree topology of the te approach. RAxML bootstrap support and Bayesian posterior probabilities expressed as percentages are indicated respectively. Dashes (--) indicate values <50; asterisk (*) indicates different topology of RAxML or MrBayes tree. Four major clades can be recognized. Newly acquired barcodes are indicated with a collection accession number (rmnh.crus.d., mzb.), otherwise GenBank accession numbers are given. The selection of species can be found in the appendices (supplementary table S1), and species of which only morphological data was analysed are indicated with an asterisk (*) and no accession number. Colours indicate various host associations. The host association of Pontonia longispina Holthuis, 1951 is unknown, indicated with a question mark (?), and the known host association of Pontonia chimaera Holthuis, 1951 is with gastropod molluscs, which is indicated with an outline of a shell.

Citation: Contributions to Zoology 2022; 10.1163/18759866-bja10030

• Download Figure
• Download figure as PowerPoint slide

The first clade contains the Atl/ep genera Ascidonia and Pontonia. Within Pontonia, Pontonia panamica was basal to the rest of the genus. While the multiple sequences of Ascidonia quasipusilla clustered together nicely, the two sequences of P. pinnophylax were divided by a branch containing P. mexicana.

The splitting of the second clade containing Dactylonia, Cainonia, and Rostronia was weakly supported by both the bootstrap values, as well as the Bayesian Inference pp support values. However, the internal branches were better supported. Rostronia basally branching off from this clade resulted in low support values in the bi analysis, while the bootstrap support value is 80. The two sequences of Dactylonia ascidicola did not cluster together, but were only separated by low support values and aberrant topologies of the MrBayes analysis. The third and fourth clades were recovered as sister clades with Opaepupu basal to these two clades, although these relationships were only weakly supported. The third clade contains a monophyly of Odontonia, Pseudopontonia, Bruceonia and Anchiopontonia. The placement of the monotypic genera directly basal to Odontonia had low support values. O. kerangcaris, was placed most basally to the genus together with the monotypic genus Anchiopontonia, with low support values. There were topological incongruencies between both analyses, but the sequences of Odontonia sibogae clustered together as expected.

The fourth clade contains a monophyly of all species of Conchodytes and a basal lineage of Platypontonia and Pinnotherotonia, with Notopontonia being basal to this entire clade. This topology between the four genera was supported relatively weakly in both analyses. Within Conchodytes, support values were generally low, with some high support values within and for the clade containing C. meleagrinae, Conchodytes kempoides Bruce, 2013, and one sample of C. biunguiculatus. In addition, C. nipponensis was recovered as the most basal species in Conchodytes, with somewhat high support values.

Historical biogeography analysis

An ancestral state reconstruction analysis was performed to identify ancestral biogeographic ranges of the studied ingroup taxa. Of the studied models, the dec model was retrieved with the lowest aic score (see supplementary table S3). The clade with the genera Pontonia and Ascidonia has a distribution outside Indo-West Pacific area (see table 1). The common ancestor of this clade has an ep/W Atl distribution (fig. 5: node A, P = 0.77). Pontonia has maintained an ep distribution until the common ancestor of P. manningi evolved with an ancestral range in the ep/W Atl (fig. 5: node B, P = 0.98).

Phylogeny based on the RAxML tree topology of the te approach (fig. 4), with ancestral biogeographic range reconstructions on the internal nodes (probabilities are shown as pie charts). Indo-West Pacific genera simplified as genus names, except for Odontonia kerangcaris Fransen, Groenhof & De Gier, 2021 due to its position outside of the genus. Colours indicate distribution ranges, both for the species as well as the ancestral distribution ranges. Species/genera of which only morphological data was analysed are indicated with an asterisk (*).

Citation: Contributions to Zoology 2022; 10.1163/18759866-bja10030

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Phylogeny based on the RAxML tree topology of the te approach (fig. 4), with ancestral biogeographic range reconstructions on the internal nodes (probabilities are shown as pie charts). Indo-West Pacific genera simplified as genus names, except for Odontonia kerangcaris Fransen, Groenhof & De Gier, 2021 due to its position outside of the genus. Colours indicate distribution ranges, both for the species as well as the ancestral distribution ranges. Species/genera of which only morphological data was analysed are indicated with an asterisk (*).

Citation: Contributions to Zoology 2022; 10.1163/18759866-bja10030

• Download Figure
• Download figure as PowerPoint slide

Phylogeny based on the RAxML tree topology of the te approach (fig. 4), with ancestral biogeographic range reconstructions on the internal nodes (probabilities are shown as pie charts). Indo-West Pacific genera simplified as genus names, except for Odontonia kerangcaris Fransen, Groenhof & De Gier, 2021 due to its position outside of the genus. Colours indicate distribution ranges, both for the species as well as the ancestral distribution ranges. Species/genera of which only morphological data was analysed are indicated with an asterisk (*).

Citation: Contributions to Zoology 2022; 10.1163/18759866-bja10030

• Download Figure
• Download figure as PowerPoint slide

The common ancestor of the genus Ascidonia was recovered to have an ep/W Atl distribution (fig. 5: node C, P = 0.91). This remains the ancestral range for the common ancestor of the clade containing Ascidonia pusilla (Holthuis, 1951) and A. quasipusilla. The ancestor of the other clade with sister species Ascidonia californiensis (Rathbun, 1902) and Ascidonia flavomaculata (Heller, 1864) was recovered as mainly W Atl (fig. 5: node D, P = 0.81).

Ancestral state reconstruction of host associations

An ancestral character state reconstruction analysis resulted in Maximum Likelihood probabilities for all 127 internal nodes, which translate to the probabilities an ancestral clade of shrimp would have been associated to a certain type of host. As expected, all internal nodes of the clade containing Conchodytes, Platypontonia, Pinnotherotonia are fully resolved to be associated to bivalve molluscs (P = 1.00), similar to most internal nodes of Pontonia (excluding P. panamica and P. longispina). Similarly, the clades containing Ascidonia, Dactylonia and most of Odontonia are fully resolved to be associated with ascidians.

Most interesting are the ancestral character states of the entire ingroup, and nodes where interphylum host-switches seem to occur. The character state of the ancestral node of the entire ingroup is recovered to be an ascidian-associated species, both with or without the inclusion of the basally branching Colemonia (fig. 6: node i and ii, P = 0.99 and P = 1.00). While the ancestral host association of the first major clade, containing Conchodytes, Odontonia and related genera, is recovered as a bivalve associate (fig. 6: node V, P = 0.62), the character state of the ancestor when including Opeapupu is still recovered as an ascidian-associate (fig. 6, node iii, P = 0.95). A host switch has occurred from an ascidian to a bivalve mollusc host at the base of this clade, but also separately in the branch containing Opeapupu. In addition, the ancestral character state of the clade including Conchodytes, Platypontonia, Pinnotherotonia, and Notopontonia, was recovered to be a bivalve-associate, even when taking the ascidian-associated Notopontonia into account (fig. 6: node vi, P = 0.64). This means that there has occurred a similar switch, albeit now from a bivalve mollusc to an ascidian host, within the branch of Notopontonia. Within the other part of the clade, comprised of Odontonia, Anchiopontonia, Pseudopontonia, and Bruceonia, two to three host switches, again from a bivalve mollusc to an ascidian host, can be observed. First one in the single branch of Pseudopontonia, and later in the branch containing O. rufopunctata, and in the other lineage of Odontonia (excluding O. kerangcaris).

Phylogeny based on the RAxML tree topology of the te approach (fig. 4), with ancestral character state reconstructions on the internal nodes (probabilities are shown as pie charts). Colours indicate various host associations, both for the species as well as the ancestral character states. The host association of Pontonia longispina Holthuis, 1951 is unknown, indicated with a question mark (?), and the known host association of Pontonia chimaera Holthuis, 1951 is with gastropod molluscs, which is indicated with an outline of a shell. Species of which only morphological data was analysed are indicated with an asterisk (*).

Citation: Contributions to Zoology 2022; 10.1163/18759866-bja10030

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Phylogeny based on the RAxML tree topology of the te approach (fig. 4), with ancestral character state reconstructions on the internal nodes (probabilities are shown as pie charts). Colours indicate various host associations, both for the species as well as the ancestral character states. The host association of Pontonia longispina Holthuis, 1951 is unknown, indicated with a question mark (?), and the known host association of Pontonia chimaera Holthuis, 1951 is with gastropod molluscs, which is indicated with an outline of a shell. Species of which only morphological data was analysed are indicated with an asterisk (*).

Citation: Contributions to Zoology 2022; 10.1163/18759866-bja10030

• Download Figure
• Download figure as PowerPoint slide

Phylogeny based on the RAxML tree topology of the te approach (fig. 4), with ancestral character state reconstructions on the internal nodes (probabilities are shown as pie charts). Colours indicate various host associations, both for the species as well as the ancestral character states. The host association of Pontonia longispina Holthuis, 1951 is unknown, indicated with a question mark (?), and the known host association of Pontonia chimaera Holthuis, 1951 is with gastropod molluscs, which is indicated with an outline of a shell. Species of which only morphological data was analysed are indicated with an asterisk (*).

Citation: Contributions to Zoology 2022; 10.1163/18759866-bja10030

• Download Figure
• Download figure as PowerPoint slide

The ancestral state of the clade containing Pontonia, Ascidonia, Dactylonia and related species is recovered as an ascidian-associated lineage (fig. 6: node iv, P = 1.00). Within this clade, two host switches from an ascidian host to a bivalve mollusc, and one within-phylum (from a bivalve mollusc to a gastropod mollusc) host switch can be found. The subclade containing Dactylonia, Cainonia and Rostronia can be observed to have an interphylum host switch at a basal branch of the clade (after Rostronia splits off, at the branching with Cainonia): the ancestral character state of this node is recovered as an ascidian associated species (fig. 6: node ix, P = 0.99). In contrast to the monotypic Cainonia, the various bivalve-associated Pontonia species form a separate clade after splitting off from their ascidian associated congener, P. panamica. The unknown host association of P. longispina causes the ancestral state reconstruction of their joint node to have a lower probability in favour of an ascidian-associated ancestor (fig. 5: node x, P = 0.67). The intergeneric host switch of P. chimaera from a bivalve to a gastropod mollusc, results in a fully resolved ancestral character reconstruction in favour of a bivalve mollusc host (fig. 5: node xi, P = 1.00).

Uncertain placement of ingroup species among different tree topologies

While most of the genera share a similar, monophyletic topology in all trees, there are some species with one or more questionable placements among the different tree topologies. In addition, various genera which were only represented by one species (monotypic) in the concatenated alignment, or in the te analysis differ in placement in the various tree topologies as well. The differences in placement among the various analysis-methods can be explained by missing data in the molecular alignment (e.g., Van Der Wal, 2019), uncertain identifications (Raupach & Radulovici, 2015), or bias in the analysis methods (e.g., due to convergent evolution of adaptive morphological features; Scotland et al., 2003; Lee & Palci, 2015; Caldas & Schrago, 2018). The current placements, possible explanations of these placements, and potential solutions will be discussed below for all notable species.

Anchiopontonia hurii can be observed to be most closely related to the recently described species Odontonia kerangcaris in the te analysis (fig. 4), while it was placed basally to Ascidonia and Pontonia in the concatenated and morphological alignment (figs. 2, 3). Bruce (1992) argued that A. hurii was closely related to some other species that have long dorsal telson spines, but this character can be found in various lineages of the studied ingroup. Previous phylogenetic studies have at least confirmed the basal placement of the species, but if the species is basal to some iwp genera (e.g., Fransen & Reijnen, 2013, Chow et al., 2021) or to the entire, or a part of the ingroup (e.g., Gan et al., 2015, Horká et al., 2016), remains elusive.

While the placement of Conchodytes as a genus is fairly well resolved in all phylogeny reconstructions, the placement of its species is far from similar in all analyses. In addition, previous literature based on morphological data does not seem to fully comply with the current results based on molecular data (e.g., Holthuis, 1952; Bruce, 1989; Fransen, 1994b; Bruce, 2013). In the current analyses, there seems to be a consensus about the basal placement of C. nipponensis and C. tridacnae (and the closely related Conchodytes maculatus Bruce, 1989, which is only represented by morphological data) to the rest of the genus, but the support values resulting from the te and concatenated alignment are rather low (figs. 2 and 4). The deviation in placements might be related to the morphological similarity of C. nipponensis to the other members of the genus (fig. 3) (e.g., Fransen, 1994b). Bruce (1989) mentioned that C. maculatus is morphologically quite distinct, suggesting a basal placement too in the genus. The remaining species of Conchodytes seem to cluster in two lineages in the te and concatenated approach (figs. 2 and 4), similar to the previous phylogenetic studies on the genus (Fransen & Reijnen, 2012, 2013): C. chadi, C. placunae, C. monodactylus and (one lineage of) C. biunguiculatus seem to cluster together in all analyses, complemented by one species which is only represented by morphological data (C. philippinensis Bruce, 1996). It is worth noting that the previous phylogenetic studies of this genus used only one molecular marker (coi). Interestingly, Bruce (1996) mentions C. philippinensis to be most related to C. nipponensis and C. maculatus, which are here recovered as one of the basal species within Conchodytes (fig. 4). While in both analyses C. pteriae is represented as a basal species related to this cluster, previous literature states C. pteriae to be a sister to C. meleagrinae, a species from the other group in the present analyses.

The second group in the present analyses is a cluster of two lineages of C. meleagrinae, and one lineage of C. biunguiculatus. In addition, this cluster also includes C. kempoides, which is only represented by morphological data from the limited published illustrations and description (Bruce, 1989, 2013). As mentioned above, C. meleagrinae and C. biunguiculatus are represented by two lineages each, the latter lineages even found in both Conchodytes groups. The possible explanations for these may be deduced from the localities and host of the species (supplementary table S1): the C. biunguiculatus specimen from Chow et al. (2020, 2021) which does not comply with the phylogenetic position found in previous studies (Fransen & Reijnen, 2012, 2013) comes from the Red Sea (Eilat, Israel), from a Pinctada margaritifera (Linnaeus, 1758). The other specimens are from more Eastern waters (Indian Ocean, iwp; supplementary table S1), presumably from different species of bivalves (presumably all Pinna spp.). The genetic distance between the two lineages of C. biunguiculatus can partially be explained by the geographical distance between both populations, although a placement as sister species would be expected (as can be seen in the genus Alpheus Fabricius, 1798 in Williams et al., 2001). Similarly, a difference in host species would possibly not affect the genetic distance between populations, as was proven in Z. soror by Antokhina & Sorokin (2010). This was however just proven in one species, and is worth examining in other palaemonid shrimp species with a smaller host-range. Physical examination of this specimen is needed to find out if this was correctly identified or whether it belonged to an undescribed species of Conchodytes. Holthuis (1952) already noted a resemblance of C. monodactylus to C. biunguiculatus, which is herein also confirmed based on molecular data of the “true” C. biunguiculatus (fig. 4). Similarly, there are two separate lineages of C. meleagrinae: one more basally placed to the other in both te and concatenated analyses. The possible explanation for this placement is difficult to identify: the localities and host choice of all analysed specimens are similar. Only one specimen (kf638630), one of the three specimens which were included by Fransen & Reijnen (2013), was found in an aberrant host species (Spondylus sp.). Fransen & Reijnen (2013) found one clade of C. meleagrinae, but a reasonable genetic distance between the three basal specimens (kf638630, jx185699, and jx185698) and the more derived specimen (kf638631). In addition, two specimens are indicated as “C. meleagrinae cf.” in our analyses. These specimens were found in a specimen of Pteria sp., and have some morphological character that are different from the other specimens. The specimens (rmnh.crus.d.57913 and rmnh.crus.d.57915) are smaller in body size (pocl of 3–5 mm), while other specimens are somewhat larger in size (pocl 5–7.5 mm). Consequently, various defining characters are underdeveloped in these specimens. They, however, share various characters on the third, fourth and fifth walking leg dactyli, and the second pair of chelae, that differ it from the original identification of C. pteriae (Fransen, pers. obs.). A more complete molecular alignment is needed to compare the specimens of this species, but also all other species within Conchodytes, correctly.

The two lineages recovered in the te analysis (fig. 4) within Dactylonia are hard to identify in the other molecular analysis (fig. 2). The first lineage seems to partially correspond to the previous study by Fransen (2002) and includes D. okai, D. holthuisi, and one species that is only represented by morphological data (Dactylonia franseni Bruce, 2003). The second lineage seems to be represented by D. anochoreta, Dactylonia borradalei Bruce, 2005, Dactylonia carinicula Bruce, 2006, Dactylonia monnioti (Bruce, 1990) (the latter three only represented by morphological data), and two lineages of D. ascidicola. Interestingly, Bruce (2003) already thought D. franseni to be related to D. okai, but also to D. ascidicola, which is in part congruent with the current analyses. In a similar situation, Bruce (2006) thought D. carinicula to be a close relative to D. ascidicola, D. anachoreta, or D. monnioti, which seems to correspond as well. At least two species with uncertain placements cause the phylogeny reconstruction to be far from resolved within the genus. As mentioned above, D. ascidicola is represented by two specimens in two lineages in the te analysis, and by eight specimens (as five separate lineages) in the concatenated alignment. Assuming the specimen identified as D. anachoreta is correctly sequenced, an indication of the true relations between the specimens identified as D. ascidicola can be seen. Due to both lineages being examined for their morphology, and the high genetic similarity, and therefore low intrageneric support values, the “true” lineage of D. ascidicola cannot be identified. However, there seems to be a consensus that D. ascidicola is closely related to D. anachoreta (and possibly to D. borradalei, as this species was described by Bruce (2005) on the basis of specimens previously attributed to D. ascidicola). The only known coi sequence of D. ascidicola is located in the other Dactylonia lineage (fig. 2: mh257317) and was later used in the te alignment (fig. 4), where it clustered with the above-mentioned clade. The other problematic species, D. holthuisi, can be observed to be represented in four lineages in the concatenated alignment: the presumably correct placement as two clustering with D. okai, one basally to Dactylonia clustering with Cainonia, and one within Odontonia. The first three lineages were correctly identified by physical examination, while the Taiwanese material from Gan et al. (2015) from the fourth lineage could not be examined. This specimen might have been mistaken for Odontonia katoi due to the shrimp’s small size and ascidian-inhabiting behaviour (the host of Gan et al.’s (2015) specimen is however unknown). The specimen related to Cainonia is represented by a single 16S sequence, potentially causing this placement (supplementary fig. S1C). Undersampling of genetic markers in the available specimens, as well as marker choice, can influence the resulting placement in a phylogenetic tree (e.g., da Silva et al., 2011). Similar to Conchodytes, more molecular data is needed in order to correctly place the species of Dactylonia in a phylogenetic framework.

Although the species within Odontonia are placed in a different order in all three tree topologies, there seems to be some kind of consensus on which species are more basal, and which species are more related to one another. The recovered topologies are largely similar to those of Fransen (2002). Odontonia seems to be split up into two main lineages in the analysis using only molecular markers, with O. rufopunctata being placed basal to the Odontonia and Conchodytes groups (fig. 2). Odontonia rufopunctata is placed in a basal position to the other ascidian-associated Odontonia species in the te analysis as well (fig. 4). Similarly, the bivalve-associated O. kerangcaris is placed basally to the other Odontonia species with Anchiopontonia hurii in the te analysis (fig. 4), while it is placed in Odontonia next to O. katoi in the tree resulting from the concatenated alignment (fig. 2). Odontonia kerangcaris is also placed basally in the morphological analysis, however, O. rufopunctata is placed in a derived branch of Odontonia in this tree topology. The basal placement of O. kerangcaris is as expected, since the species’ host and morphological features were found by Fransen et al. (2021) to resemble that of related, bivalve-associated genera such as Conchodytes. The placement of O. rufopunctata is less expected, but the placement away from the other species within the genus was also recovered in a previous study based on dna and morphological features (De Gier & Fransen, 2018). As mentioned above, the other species seem to be grouping in two separate lineages in the concatenated analysis: one lineage including O. katoi, O. seychellensis and O. kerangcaris (and one specimen of Dactylonia holthuisi); and one lineage including O. bagginsi and O. sibogae. The placement of the relatively basal O. katoi is interesting, since it here clusters with O. seychellensis, which was recovered to be a very derived species in the te analysis. The potential explanation for this can be found in the only successfully sequenced marker of O. katoi, 16S. Odontonia katoi can therefore only be compared to other 16S sequences, in this case being limited to only O. sibogae and O. rufopunctata (see De Gier & Fransen, 2018). Due to O. simplicipes and O. compacta being only represented by morphological data, and O. katoi, O. kerangcaris, O. plurellicola, O. bagginsi, and O. seychellensis only being represented by one molecular marker, the relation between these species and the other species in Odontonia is far from resolved.

The inclusion of the recently described genus Opaepupu is only based on similarities in morphological characters, and a similar host association (with a bivalve mollusc). Anker & De Grave (2021) even remarked that this monotypic genus might not necessarily be closely related to the other bivalve endosymbionts. Nonetheless, O. huna seems to bear various morphological characters that resulted in the current analyses in a basal placement to Conchodytes (fig. 3) or to a clade including Conchodytes, Odontonia, and other related genera (fig. 4). These morphological characters include: slender, seemingly unadapted dactyli on the ambulatory legs; a well-developed, acute antennal spine and dorsal and rostral carina; and a slender third maxilliped (Anker & De Grave, 2021; supplementary appendix S2, supplementary table S2). In a similar situation the sole member of Notopontonia, N. platycheles Bruce, 1991, was also included based on the morphological similarities of the described specimens in Bruce (1991) and Berggren (1999). Based on external characters, Bruce (1991) argued Notopontonia to be closely related to Pontonia. However, Bruce (1991) did not discover a host for the species, and suggested that the species was a bivalve associate due to its depressed body shape and the horizontal disposition of the second pereiopods (Berggren, 1999). Berggren (1999) found the species in the ascidian Herdmania momus (Savigny, 1816), but did not mention anything about its systematic placement. The ascidian-associated habit of N. platycheles, as well as its previously thought relatedness with Pontonia, lead us to believe the species is a member of the currently studied clade.

Platypontonia hyotis is represented by five individuals in the concatenated alignment, clustered in two groups of two and three barcodes respectively. The basal placement of the first group (fig. 2) raises questions about the similarities of these specimens, compared to the ones being basally placed next to Anchiopontonia, Ascidonia and Pontonia. The possible explanation for this aberrant placement can be found in the selection of the barcodes (supplementary table S1): the two basally placed barcodes are both coi-barcodes, while the other three clustered barcodes are a combination of 16S and H₃ barcodes. Physical examinations were done on four specimens, eliminating the chance of misidentifications. If the barcodes are combined and supplemented with morphological data (fig. 4), P. hyotis clusters together with Platypontonia brevirostris (Miers, 1884) (represented only with morphological data), as a sister to Pinnotherotonia (and together with Pinnotherotonia with Conchodytes). In contrast, previous analyses by Fransen & Reijnen (2012) and Chow et al. (2021) placed Platypontonia respectively basal to Conchodytes, and next to Anchiopontonia. Bruce (1968) even suggested a relatedness to Pontonia. Based on previous analyses and our data, it is likely that the genus is at least part of a clade containing Conchodytes, Odontonia and also Dactylonia.

Although the recovered topologies are largely similar to those of Fransen (2002), various species within Pontonia cause problems when studying the phylogeny reconstructions. Most problematic placements can be traced back to potential misidentifications and limited genetic information of some species, causing low intrageneric support values. Interestingly, the genetic distances between various species of Pontonia is very small in the concatenated alignment. These species groups are: P. domestica, one lineage of P. mexicana, and one lineage of P. pinnophylax; and P. margarita and one lineage of P. manningi (fig. 2). The placement of these specimens and more obvious problems are discussed below. First up, the placement of the Pontonia species with an aberrant host association (or potentially aberrant host association) (resp. P. panamica and P. longispina) are generally placed more basally to the bivalve associated species. This is the case in the morphological, as well as the te analysis. This is in congruence with previous remarks by Marin & Anker (2008). In the tree resulting from the concatenated alignment, a branch containing P. margarita and two lineages of P. manningi is placed basally to all other species of Pontonia. The possible explanation to this different topology can be found in the list of successfully sequenced molecular markers (supplementary table S1): only coi and H₃ have successfully been amplified for these specimens. Therefore, these specimens could only be compared to the other species of Pontonia using these markers, which do not give well-resolved topologies on their own (supplementary fig. S1A, B). In addition, P. manningi can be represented by two other, more derived, lineages: one related to an unidentified species of Pontonia, and one basal to this cluster and P. pinnophylax, P. domestica, and one lineage of P. mexicana. As mentioned above, the reason for this aberrant placement can be found in the successfully sequenced markers (supplementary table S1), as these two lineages of P. manningi are represented by the two ribosomal markers (16S and 18S). Both resulting lineages of P. manningi were physically examined, so misidentifications are ruled out. Similarly, P. mexicana is also represented by two lineages, one of which was checked for its morphology. The other specimen included in the concatenated alignment (fig. 2: gq227823; Baeza, 2010) could not be checked, and is represented by only one 16S sequence.

Inclusion of unidentified species

As can be observed in the phylogenetic tree resulting from the concatenated alignment, various unidentified species were included in the study (fig. 2). The inclusion of the species resulted in tentative identifications to the two unidentified species of Dactylonia and one of Pontonia. Both unidentified specimens of Dactylonia (from Kou et al., 2013 and Mitsuhashi et al., 2007) seem to group together with a large clade of D. ascidicola, which seems to be the best guess to the correct identification of the species. Both specimens have a distribution indicated as “Philippines, Panglao Island” (supplementary table S1), which corresponds with the known distribution of D. ascidicola (see Fransen, 2002). The host of these specimens are however unknown.

The unidentified specimen of Pontonia groups together with a lineage of P. manningi. Assuming these specimens were physically checked before their inclusion in the original study (Bracken et al., 2009), the possibility both specimens are the same species is very low. Both specimens are only represented by the two ribosomal markers (16S and 18S), making comparison with the other species difficult. Because physical examination was not performed, the true identity of this specimen remains elusive.

Outgroup species placement

Cuapetes tenuipes was assigned as outgroup, followed by the recovery of Actinimenes as being most related to this outgroup. The other two outgroup species were correctly recovered in the te analysis, but their placement varied in the molecular and morphological analyses. The four molecular sequences of Periclimenes colemani were split in two in the phylogeny reconstruction based on the concatenated alignment, due to the 16S sequence of this species being from a different sample (see supplementary table S1). The 16S sequence clusters together with members of Dactylonia (fig. 2). The 16S marker, compared to 18S, resolves the intrageneric relatedness quite well for the studied clade (supplementary fig. S1C), so the placement of this sequence of P. colemani is not expected. Since physical examination of the misplaced specimen used in Chow et al. (2021) was not performed, it is not possible to find an explanation for its placement in Dactylonia.

In both the analyses using the te and concatenated alignment, Typton wasini is recovered as the outgroup species most related to the ingroup. However, in the morphological dataset, the phylogenetic analysis recovered the species as a close relative to Colemonia and Bruceonia. The possible explanation for this placement can be found in the similar morphology of the sponge-inhabiting species due to convergent evolution causing homoplasy. T. wasini shares the following potentially ecomorphological characters with various species of the ingroup: 2 to 4 subdistal teeth on the rostrum; an absent supra-orbital tooth; a broadly rounded or straight orbital angle; a relatively large tooth on the scaphocerite; the teeth on the flexor margin of the corpus of the dactyli; and various camouflage characters (see supplementary appendix S2 and supplementary table S2). A morphological comparison between T. wasini and a typical ingroup species (e.g., Conchodytes meleagrinae) can be done by examining the published illustrations of the female habitus and walking leg dactyli by Bruce (1977) and Fransen & Reijnen (2013). whether these morphological characters have a similar (homologous) origin, can only be discovered by examining all related species of T. wasini in this context.

Historical biogeography

Combining the published and previously unpublished distribution data (supplementary appendix S1; table 1, supplementary table S1) with the presented phylogeny reconstruction from the te approach (fig. 4), gives an indication of potential evolutionary distribution patterns of the studied ingroup (fig. 6). Almost all studied genera can be found in tropical coral reefs in the iwp, ranging from the Red Sea to Vanuatu and Hawaii (supplementary appendix S1). The ep/Atl lineage, containing all members of Ascidonia and Pontonia, forms a monophyletic clade (fig. 4). This is in congruence with a previous study using morphological characters (Fransen, 2002). It reflects the combination of the Terminal Tethyan Event (tte, 12–18 Myr ago), forming a ‘hard’ barrier, and the East Pacific Barrier (epb, 5000 km open ocean, in effect throughout the past 65 Myr), forming a ‘soft’ barrier, separating iwp and ep populations. The internal architecture of the clade is mirrored in the subsequent closure of the Isthmus of Panama (iop, 3.1 Myr ago) separating the ep and Atl regions, as well as the ‘soft’ barrier caused by the opening of the Atlantic Ocean (oao, in effect throughout the past ca. 60 Myr) which can be observed in several marine groups (e.g., Fransen, 2002; Anker & Baeza, 2012; Cowman & Bellwood, 2013; Anker et al. 2017; Baraf et al., 2019). Examples of closely related paraphyletic or sister species pairs that possibly evolved as a consequence of the iop are: P. manningi/P. margarita and A. pusilla/A. quasipusilla. This also holds for the sister species A. flavomaculata and A. californiensis, although no representative of this clade has been recorded from the W Atl. This could indicate subsequent extinction in that region or it remains to be discovered there. The oao is best reflected in the species P. pinnophylax and P. mexicana. As indicated by Fransen (2002), morphological differences between these species are minute and not reliable. The genetic distance between the W Atl populations of P. mexicana and those of the E Atl P. pinnophylax could be seen as an indication of speciation in progress.

Dispersal events from the iwp to the E Atl seem to be missing in the current results, probably due to the natural dispersal of the species around South Africa being inhibited by cold water flows. In recent times, the Suez channel in Egypt, and the fishery business (oyster and mussel farmers) are giving the opportunity for Red Sea invertebrates to become introduced in Mediterranean waters (e.g., Galil et al., 2018).

The host switch from an ascidian host in P. panamica to a mollusc host in the remaining Pontonia species parallels the iop in the biogeographical analysis (fig. 5). The interactions between host switching and biogeographical events have been studied in detail before (e.g., Hoberg & Brooks, 2008). While this is the only apparent matching between the two analyses, it might be worth studying this in more detail.

Host switching and notes on ecomorphological characters

The ancestral character state with regard to the host choice of the entire monophyletic ingroup was recovered to be an ascidian-associated shrimp species (fig. 6). Interestingly, previous studies featuring ancestral state reconstruction methods resulted in an unresolved host association (Horká et al., 2016) or a bivalve-associated shrimp species (Chow et al., 2021) for the common ancestor. The explanation for this mismatch in ancestral character state recovery should be sought in the limited number of included species in the previous studies (Horká et al., 2016: n = 4; Chow et al., 2021: n =10), causing ambiguous results for the presently studied clade.

As mentioned before, there seem to be at least six to seven different instances of host switches between phyla (from an ascidian to a bivalve host, or from a bivalve to an ascidian host), and one instance of a host switch within a phylum (from a bivalve to a gastropod mollusc). It is worth noting that the number of host switches presented in this study is based on the tree topology resulting from the te-approach, which unfortunately is accompanied with low support values on various branches (figs. 4 and 6). Once more (complete) molecular data becomes available, the tree topologies and support values might change, revealing a different number of host switches. Thus, the number of host switches is tentatively identified as six to seven. Most of the host switches in the studied clades are characterized as a single occurrence: the host switch from an ascidian to a bivalve host in the genus Pontonia happened once after the branching of the ascidian-associated P. panamica, or after the branching of P. longispina, of which a host association remains uncertain (fig. 6). Similarly, within the clade containing Rostronia, Cainonia and Dactylonia, one host switch from an ascidian to a bivalve host occurred in the branch represented by the only species of Cainonia (fig. 6). In previous morphological analyses by Fransen (2002), Cainonia was considered to be part of Dactylonia, although the genus was later founded partly based on the aberrant host-association of the species (Bruce, 2005). Within the clade containing Conchodytes, Odontonia and all other smaller related genera, a switch from an ascidian-associated ancestor to a bivalve associated ancestor can be observed basally in the phylogeny reconstruction (fig. 6), after Opeapupu splits off. The ancestral character state of the clade including Opeapupu is still an ascidian association, meaning that in the branch of Opeapupu a host-switch has occurred from an ascidian to a bivalve host. After the initial switch, several unique host switching events occur in the clade, first in the branch containing Notopontonia, which seems to have switched back from a bivalve- to an ascidian-associated lifestyle. Similarly, in the clade containing Pseudopontonia, Anchiopontonia, and Odontonia, the bivalve-associated ancestors do not switch once, but twice back to an ascidian-associated species in Pseudopontonia, and later in Odontonia (fig. 6). Within Odontonia, the ancestor of the genus (excluding O. kerangcaris) is recovered as a bivalve-associate, suggesting that the ascidian-associated O. rufopunctata has switched to an ascidian-associated lifestyle independently from the rest of the genus.

It has been hypothesized that the switches from an ectosymbiotic to an endosymbiotic host association resulted in various species-specific associations with bivalve, ascidian and sponge hosts in multiple clades within the Palaemonidae (Horká et al., 2016). This can be observed in the high level of host-specificity in the studied endosymbiotic clade, exploiting 19 confirmed families (table 1; Chow et al., 2021). Chow et al. (2021), however, hypothesized that these clades of endosymbiotic shrimp do make smaller taxonomic jumps than ectosymbiotic shrimp, when colonizing new hosts. This is congruent with the host choices of the species in most of the studied genera (supplementary appendix S1), but this does not explain the seemingly difficult switch from one phylum to another. One explanation for these seemingly unexpected host switching events can be found in the internal structures of the host phyla: the studied shrimp species living inside bivalve molluscs are reported as inhabitants of the mantle cavity (e.g., Fransen, 1994), feeding from the mucus and pseudofeces which are built up near the gills (e.g., Kennedy et al., 2001; Ashelby et al., 2015; De Grave et al., 2021). This soft, mucus-producing internal cavity might be morphologically similar to the pharyngeal basket of phlebobranch and stolidobranch solitary ascidians (Horká et al., 2016, for comparison, see Monniot, 1991). A seemingly similar host switch between phyla has occurred in a lineage of the mainly sponge-inhabiting genus Periclimenaeus, where various species can be found in the sponge-like tunnelling systems of colonial (or compound) ascidian species (e.g., Fransen, 2006; Horká et al., 2016).

The ancestor of the gastropod-associated Pontonia chimaera and the closely related bivalve-associated congener was recovered to be a bivalve-associate. Although this intrageneric host switch can easily be distinguished in the currently presented phylogeny reconstruction, there are two other minor host switches present in the current study. De Gier and Fransen (2018) reported that Odontonia plurellicola was found in a species of Plurella Kott, 1973, an aggregating phlebobranch species. Similarly, Dactylonia holthuisi can also be found in the same and other Plurella species (Fransen, 2006; C. Fransen, pers. obs.). Members of Plurella, like all other members of the Plurellidae, are thought to have evolved from a solitary ancestor, resulting in an aggregating species characterized by a non-shared branchial sac. Contrary to the above-mentioned ecology of Periclimenaeus, this means that if individuals of O. plurellicola and D. holthuisi want to move to another ascidian in the aggregate, they have to enter it from the outside (De Gier & Fransen, 2018).

An ecomorphological question remains if the host switches, albeit within or between host phyla, result in novel morphological adaptations in the evolution of the endosymbiotic crustaceans (Fransen, 1994a; De Gier & Becker, 2020; Chow et al., 2021). Various studies have focussed on, or at least highlighted seemingly host-specific morphological features, like the reduction of spines (Fransen, 1994; 2002), adaptations of the eyes (Dobson et al., 2014, 2016), cryptic camouflage patterns (Horká et al., 2016), and the ornamentations on the walking leg dactyli (De Grave, 1999; Fransen, 2002). Hints to host-specific adaptations can also be found in previously unseen microstructures, studied with Scanning Electron Microscopy (sem) (e.g., De Grave, 1999; Dobson et al., 2014; Ashelby et al., 2015; De Gier & Fransen, 2018) and ct scanning studies (e.g., Bagge et al., 2017), also in other groups of symbiotic crustacea (e.g., De Gier & Becker, 2020).

As we detected multiple host switching events from one phylum to the other and vice versa in this lineage of symbiotic shrimp, we conclude that bivalves and solitary ascidians with their similar internal microenvironments function as hot spots for interphylum host switching. It is expected that interphylum host switches will be observed more often between phyla with matching internal microenvironments than between phyla without these matching internal microenvironments in various lineages of crustaceans and other symbionts.

Future perspectives

The present study features the most complete molecular and morphological dataset of the studied mollusc- and ascidian-associated shrimp clade. Although the studied dataset resolved some of the questions about monophyly, host-switching and generic relatedness, it demonstrates that more molecular data is needed in order to obtain better supported and resolved phylogeny reconstructions. New, slowly evolving, genetic markers might be needed in order to solve the deeper phylogeny of the studied clade (e.g., Bininda-Emonds, 2007). In addition, more field observations and studies addressing the feeding habits, host choice, and current distributions of palaemonid shrimps are needed in order to complement the datasets needed for analyses focussing on host associations, ancestral state reconstructions, and present and past biogeography. Fortunately, various recent studies have been addressing these subjects (e.g., Ďuriš et al., 2011; Wood et al., 2017; Levitt-Barmats & Shenkar, 2018; Chow et al., 2021; De Grave et al., 2021). As mentioned earlier, an ecomorphological approach might give other, new insights. New imaging methods, combined with ecological and molecular data might help us understand how the endosymbiotic species have adapted to their aberrant way of life.

To test the hypothesis that interphylum host switches will be observed more often between phyla with matching internal microenvironments than between phyla without these matching internal microenvironments, similar analyses could be performed with for instance pinnotherine crabs and amphipods that inhabit the same host phyla. It is expected that for instance the matching internal microhabitats of compound ascidians and sponges will be another hot spot for inter phylum host switching. A comparative study of inter phylum host switching throughout nature could reveal more of these hot spots.

Editor: R. Vonk