|Electronic supplementary material
Additional details of constrained parsimony analysis
The scenario of sponge spicule evolution presented in Fig. 2a displays the most parsimonious tree to conform to the a priori defined homology statements. The paucity of available morphological characters combined with the experimental imposition of constraints means that formal computerized cladistic analysis is inappropriate, in this case, for testing between hypotheses. Therefore, justification is given here for the preference of the topology in Fig. 2a over alternative scenarios that have equal or closely comparable numbers of character transformations, but which differ in the positions of the fossil taxa (including in relation to the root), the permutations of character transformations, the interrelationships of the crown groups, or the role of the outgroup.
(i) Alternative position of Eiffelia as the sister taxon to (Calcarea plus Homoscleromorpha plus Eumetazoa).
In comparison to the topology of Fig. 2a, this scenario additionally requires a secondary loss of calcareous spicules, in (Homoscleromorpha plus Eumetazoa).
(ii) Alternative position of Eiffelia as the sister taxon to all remaining taxa, or to (Forteau spicules plus Hexactinellida plus Demospongiae).
These topologies conform to the a priori defined homology statements of single origins for spicule sheaths and orthogonal spicule symmetry, and a mineralogical transition between silicean and calcarean spicules. They furthermore conform in the polarity of mineralogical transition to the preferred scenario of Botting and Butterfield (2005). In terms of parsimony, they require a similar complexity of character evolution (including awkward transitions as well as losses) to that presented in Fig. 2a. However, they also require that spicules of calcareous biomineralogy and diverse symmetry (tri/hexaradiate, orthogonal) were present in the crown ancestor of Metazoa, making the absence of such spicules from the pre-Cambrian fossil record problematic. In contrast, the position of Eiffelia in Fig. 2a allows for a Cambrian appearance of calcareous sponge spicules, a scenario that is consistent with molecular clock estimates of divergence times (Sperling et al. 2010). Although the problem remains of a long missing fossil record for siliceous spicules (whether or not this character traces back to the crown ancestor of Metazoa), a specific preservational constraint has been proposed which may explain this (see Sperling et al. 2010).
(iii) Alternative character evolution in (Homoscleromorpha plus Eumetazoa).
An alternative to the loss of silica mineralization followed by the independent evolution of silica spicules in Homoscleromorpha (Fig. 2a) is to allow a transformation from sheathed spicules of orthogonal symmetry to (?)unsheathed spicules of homoscleromorph-type morphology, with silica mineralization lost in Eumetazoa. There seems to be no clear preference for one or other hypothesis on the basis of parsimony, particularly in the absence of an internal phylogeny of Homoscleromorpha, which includes aspiculate representatives.
(iv) Alternative position of the Forteau spicules as sister taxon to (crown-group) Hexactinellida.
In comparison to the topology of Fig. 2a, this position for the Forteau spicules requires an additional loss of spicule sheaths among siliceans (once in Hexactinellida and once in Demospongiae).
(v) Alternative constraint tree in which ‘Demospongiae’ is paraphyletic with respect to Hexactinellida.
This scenario, which is supported by some recent molecular phylogenetic analyses (discussed in Sperling et al. 2009) but not others (Philippe et al. 2009; Sperling et al. 2010), requires extra steps in character evolution (losses of sheaths and transformations in spicule symmetry), regardless of whether the Forteau taxon is positioned as in Fig. 2a, or as the sister taxon to Hexactinellida.
(vi) Alternative constraint tree in which Porifera is monophyletic.
A topology of extant sponge relationships in which Porifera is monophyletic and exhibits an internal arrangement of (Hexactinellida plus Demospongiae) (Calcarea plus Homoscleromorpha) has been recovered in the recent phylogenomic study by Philippe et al. (2009), although sponge monophyly has subsequently been considered an artefact of tree reconstruction methods by Sperling et al. (2009). Imposing this topology as a constraint tree, and retaining the relative positions of the Forteau spicules and Eiffelia from Fig. 2a, means that spicules can be primarily rather that secondarily absent from stem-eumetazoans, but otherwise does not alleviate the complexity of character transitions and losses required by the homology hypotheses arising from Botting and Butterfield (2005).
(vii) Outgroup analysis.
Choanoflagellates, as the outgroup to Metazoa (Sperling et al. 2009), potentially help to polarize biomineralization characters among sponges. However, it is not known whether the siliceous structures produced by some choanoflagellates are homologous to sponge spicules, and they are not considered in the present analysis.