Robustness of phylogenetic results and ancestral reconstructions
Our conclusions about the ancestry of T. pseudonana rest primarily on the assumption that our analyses have accurately resolved its position within the broader Thalassiosirales phylogeny. As suggested by the numerous taxonomic and nomenclatural uncertainties (Additional files 3 and 4), T. pseudonana has gross morphological similarities to species of Thalassiosira, Cyclotella, and Discostella. In fact, T. pseudonana is commonly confused with several species of Discostella and Thalassiosira (viz., the purely freshwater species D. pseudostelligera and the purely marine species T. guillardii and T. oceanica). We included each of these species in our molecular dataset, and they all resolved well away from T. pseudonana (Figure 1). Cyclotella and Thalassiosira are especially large and morphologically diverse genera, though years of monographic research have led to numerous informally recognized morphological groups within each of them [34–43]. Our molecular dataset includes at least one representative from each of the informally designated groups within both Thalassiosira and Cyclotella. By sampling much of the known range of morphological diversity within these two genera, we have included many of the lineages within the Thalassiosirales phylogeny where one might predict T. pseudonana to fall.
The molecular dataset ultimately resolved T. pseudonana as sister to a clade that includes the C. meneghiniana species complex and the nomenclatural type of the genus Cyclotella, C. tecta. Comparing phylogenies derived from morphological and molecular datasets is a powerful way to detect phylogenetic incongruence, or likewise, to strengthen an existing phylogenetic hypothesis based on just one of the two data types . Despite their different and largely non-overlapping taxon sampling schemes, the molecular and morphological datasets were congruent in their placement of T. pseudonana as sister to Cyclotella (Figures 1 and 2). Moreover, there was also congruence between morphology and molecules for the larger set of taxa common to both datasets, viz: (T. gessneri, (T. weissflogii, (T. pseudonana, C. meneghiniana))).
Our habitat codings were based on the habitat from which each culture originally was isolated. Although an individual culture strain might not capture the full range of genetic and phenotypic variability of its species, coding a culture strain by its natural habitat is conceptually similar to the common practice of representing a species in phylogenetic analyses with gene sequences from a single individual. Thalassiosira pseudonana can tolerate a wide range of salinities , so we could have coded each of the strains in our analyses as simply "euryhaline." For our purposes, however, individually coding each strain according to its natural habitat should have been more conservative, in that it allowed for a marine ancestral state reconstruction for both T. pseudonana and the nodes immediately surrounding it.
Although future data might show that some of the T. pseudonana strains considered here represent different biological species, it is not clear that those species would necessarily resolve along marine-freshwater lines. The finding by Guillard and Ryther  that three different T. pseudonana clones (3H, 5A and e.p.) maintained high growth rates from 0.5-37‰ salinity supports the hypothesis that T. pseudonana is, in fact, a single species that tolerates a wide range of salinities. Finally, the reconstruction of a freshwater ancestor for T. pseudonana (Figure 1) applies regardless of whether marine and freshwater strains are considered the same, or different, species.
To summarize, phylogenetic analyses place T. pseudonana as sister to a predominately freshwater clade of diatoms in the genus Cyclotella. The congruence between the molecular and morphological datasets provides especially strong support for this result . Ancestral state reconstructions showed that the common ancestor of this clade was either unequivocally freshwater (morphological data, parsimony analysis) or likely freshwater (molecular data, likelihood analysis). Like any hypothesis, the results are potentially subject to change as new data become available or different character codings are used, but with respect to T. pseudonana, phylogenetic results and ancestral state reconstructions appear to be at least moderately robust to taxon sampling, data type, and optimality criterion.
Is Thalassiosira pseudonana a good model for marine diatoms?
The tremendous importance of marine diatoms in global carbon fixation and marine food webs understandably compelled Armbrust et al.  to choose a marine diatom as the first for whole-genome sequencing . Thalassiosira pseudonana belongs to the large and predominantly marine genus Thalassiosira and so was reasonably considered representative of that genus and of marine diatoms more broadly . Subsequent phylogenetic analyses showed, however, that members of the genus Thalassiosira are spread across some 10 distinct evolutionary lineages (Figure 1). The extent of independent, unshared evolutionary history distinguishing these 10 lineages from one another is such that no single model species can possibly capture all of the diverse ecological, metabolic, and genomic properties of "Thalassiosira"—a prediction increasingly borne out by experimental and genomic data [45–47]. Of course, no one would expect uniformity across such a broad range of species diversity, but their shared classification in the genus Thalassiosira falsely suggests that the taxa in these 10 lineages constitute a more biologically coherent group than the phylogeny reveals them to be (Figure 1). The persistence of this, or any, misinformative classification represents a failure by systematists—ourselves included—to provide a phylogenetically based, and therefore biologically informative, alternative. Such an alternative would facilitate, for example, ecologically based selection of model species.
Salinity preference provides a glaring example of a trait that varies considerably across the 10 lineages of Thalassiosira, and Thalassiosirales as a whole (Figure 1). With a distribution that spans both marine and freshwater habitats, T. pseudonana embodies this variation. The evolutionary history of Thalassiosirales includes just a few major freshwater diversifications , and T. pseudonana marks an early divergence in one of them (Figure 1). In fact, phylogenetic analyses and formal mapping of habitat preference onto phylogenetic trees show that, 1) T. pseudonana is part of a large clade that is likely ancestrally freshwater (Figures 1 and 2), and 2) as a species, T. pseudonana is likely ancestrally freshwater as well (Figure 1). If so, these data suggest that marine strains of T. pseudonana represent recent recolonizations of higher salinity habitats by a derived freshwater ancestor (Figures 1 and 2).
Parker et al.  keenly raised the question as to how well our current selection of model microalgae represent their corresponding dominant forms in the marine environment. The data presented here suggest that the substantial freshwater component of T. pseudonana's phylogenetic history might confound its use of as a model for marine diatoms. For example, there is great interest in understanding how diatoms sequester, store, and use iron because it limits their growth over vast offshore regions of the world's oceans [14, 15, 46, 47]. By contrast, iron generally does not limit diatom growth in freshwater  and coastal ecosystems . Experimental data show that coastal marine strains of T. pseudonana and T. weissflogii require more iron to maintain a maximal growth rate than a marine strain of the pennate diatom, Phaeodactylum tricornutum, and the oceanic diatom, T. oceanica, respectively [14, 47]. As currently understood, these differences reflect either vastly different iron uptake  and storage  architectures between entirely different classes of diatoms (pennates vs. non-pennates), or slight modifications of shared iron uptake  or photosynthetic  architectures between closely related species. Some of these differences might, at least in part, reflect that T. pseudonana descended from a freshwater ancestor that experienced different (perhaps relaxed) constraints on its iron uptake and utilization machinery. Extending studies like that of Kustka et al.  to include strictly freshwater pennate and non-pennate diatoms along with strictly marine pennate and non-pennates might help tease apart the relative importance of shared history versus shared ecology in the evolution of iron metabolism in diatoms. In this context, "strictly" refers to diatoms that occur exclusively in either marine or freshwaters and whose immediate ancestor shared a similar distribution (e.g., Figure 1, "M").
Among the several hundred species of Thalassiosirales, which might provide a more representative model for marine diatoms? The phylogeny provides a powerful, predictive guide for selecting candidates. In this case, one might first consider the five strictly marine clades (Figure 1, "M") and then winnow down the candidate pool based on practical (e.g., genome size and amenability to cell culture) and ecological (e.g., oceanic vs. offshore) criteria.
What is Thalassiosira pseudonana?
The diatom originally described as Cyclotella nana from the River Wümme has also gone by the names T. pseudonana , C. pseudonana , and most recently, Discostella nana  (Additional file 3). Like grammar, common usage tends to prevail, and this diatom became universally known as T. pseudonana. Importantly, none of the name changes was based on direct observations of the type material, and none either had [27, 28] or took advantage of  the insights of a strongly supported phylogenetic hypothesis for the Thalassiosirales . A combination of evidence (Figure 3, Additional files 3 and 4) shows that the diatom originally described by Hustedt  as C. nana corresponds to the traditional concept of T. pseudonana [29–34]. Furthermore, phylogenetic analyses show that T. pseudonana is part of a strongly supported clade that includes C. tecta, which is the nomenclatural type of the genus Cyclotella (Figure 1). By contrast—short of creating a genus that includes nearly the whole of Thalassiosirales—a genus that would include both T. pseudonana and T. nordenskioeldii (the type species of Thalassiosira) cannot be reconciled with the phylogeny (Figure 1). We are therefore compelled to resurrect the original name of T. pseudonana, Cyclotella nana , and deprecate T. pseudonana , C. pseudonana , and D. nana  as synonyms.