REPLY TO COMMENT BY THERIOT (2008) ON KACZMARSKA ET AL. (2006)¹

Theriot (2008) aimed to demonstrate nontrivial errors in the inferences made regarding evolution of a morphological character (fultoportula, syn. “strutted process”) in thalassiosiroid diatoms published by Kaczmarska et al. (2006). He stated that phylogenetic principles were not upheld in Kaczmarska et al. (2006) because, apparently, our hypotheses were nonparsimonious. He demonstrated this conclusion by applying Patterson’s (1988) tests for homology. Contrary to his impression, our three hypotheses of fultoportula emergence, as ascribed in the goals of our paper (Kaczmarska et al. 2006, p. 122), are sound and consistent with parsimony as a methodological principle. Parsimony is a part of scientific methodology in general and not restricted to phylogenetic systematics (Sober 1988, p. 39; Brooks and McLennan 1991, p. 45). For accuracy and clarity we will refer the reader to the specific contentious statements made by Kaczmarska et al. (2006) and Theriot (2008) in the following manner. The normal author-year citation will be supplemented by the page number, specific column of the text (either right column [RC] or left column [LC], when appropriate), and a number specifying the line of text when statements are short. Using the same principles and criteria that Theriot (2008) advocates, we show that his conclusion is incorrect in the case of our tubular hypothesis, while our annular hypothesis is more parsimonious than the one he proposes (we appear to agree on the areolar hypothesis). First of all, we wish to refer the reader to Kaczmarska et al. (2006, p. 135, LC 33), which gives clear indication that the presentation was made on an initial, original data set (no similar work was conducted prior to our initial submission in 2003). The analysis and hypotheses referred to taxa for which both molecular and morphological data were available (Kaczmarska et al. 2006, p. 127, RC 15–16) and within the framework of the sister relationship between thalassiosiroids and lithodesmioids (ibid., p. 122, RC 37). As much as possible, we refrained from assumptions where unsequenced species would fall onto a tree, however similar their valve morphology might be. Discoveries of unanticipated phylogenetic relationships inferred from molecular data and contradicting those inferred from morphology (including nonmonophyly of stephanodiscaceans first shown in our paper) warranted such caution. In this rebuttal, we focus on the major contentious issues to show the patterns of errors and misrepresentations in Theriot’s (2008) presentation of our work and in his conclusions. We start with providing schematic illustration of the three discussed hypotheses (1-3), as we proposed in our paper for comparison to Figure 4, illustrating the hypothesis that Theriot (2008, p. 831, RC 52, LC 1–20) claims is best supported. This analysis will follow with general comments relevant to the phylogenetic context in which we placed our work. In addition to major errors, Theriot (2008) contains many omissions and other lesser misrepresentations and mistakes describing our work. There are too many of these to address in point-by-point fashion without testing the patience of readers, but we will provide them on request. Schematic representation of hypothesized steps involved in tubular origin of fultoportula (steps 1 and 2) from Kaczmarska et al. (2006); steps involving only thalassiosiroids are in bold, steps in sister clade are in italics. Schematic representation of hypothesized steps involved in areolar origin of fultoportula (steps 1 and 2) from Kaczmarska et al. (2006); steps involving only thalassiosiroids are in bold, steps in sister clade are in italics. Schematic representation of hypothesized steps involved in annular origin of fultoportula (steps 1 and 2) from Kaczmarska et al. (2006, pp. 134, RC 40–44) with no consideration given to the evolution of “central structures” in extinct diatoms discussed in Gersonde and Harwood (1990) and Medlin et al. (2000). Schematic representation of hypothesized steps involved in annular origin of fultoportula as inferred by Theriot (2008, p. 831, LC 1–5) who states, “This possibility would make the most parsimonious interpretation of the multistrutted process (shown here as Step 2) to be intermediate stage between areola (shown as Step 0) and fultoportula (shown here as Step 4),” a total of four steps (1, 2, 3, 4). In order to avoid terminological suggestion of evolutionary relationship between fultoportula (formerly know as strutted process) and “multistrutted” process, we use the neutral term “annular process” instead. Comments regarding the phylogenetic trees. Two issues brought up by Theriot require rebuttal. First, the claim that our alignments were not available to him (Theriot 2008, p. 822, LC 39) is not true. In fact, he received the requested NEXUS files in January 2006 from one of the authors (L. M.), in accordance with Journal of Phycology standards and policies. These files, which contain the bases selected for analyses from the aligned sequences in a format ready for analysis by PAUP and MrBayes, are the only type of files required to reproduce or additionally analyze any of our original molecular data. The alignment alone is insufficient because one does not know from the alignment which bases were selected for analyses. Second, Theriot casts doubt on the reproducibility of our trees and on the soundness of the relationship between Thalassiosirales and Lithodesmiales. In response, we draw attention to the most recent publication by Alverson et al. (2007, fig. 3; work performed in Theriot’s lab and coauthored by Theriot) where the main clades of thalassiosiroids (with greater taxon sampling and an additional gene) are very similar to those presented in Kaczmarska et al. (2006, fig. 7), with our Clades B, C, and D species clearly identifiable in Alverson et al. (2007, fig. 3) in the same sequence of divergence and species association for those species in common between the two studies. Kaczmarska et al. (2006) Clade F is also clearly recognizable. The major difference between the two papers is a better resolution of “assemblage E” and its relationship to Clade F in Alverson et al. (2007). Kaczmarska et al. (2006) clearly addressed the issues of the topological differences in recovered trees (fig. 6, pp. 129–30) and referred to their work as initial (Kaczmarska et al. 2006, p. 135, LC 33), in need of follow-up work with greater taxon sampling (p. 135, RC 47), and collegially refrained from including genera that at that time were under investigation in Theriot’s lab (p. 122, RC 54). Furthermore, despite doubting the robustness (Theriot 2008, p. 831, RC 31, “Kaczmarska believes this to be robust…”) of the sister relationship between Lithodesmiales and Thalassiosirales shown in Kaczmarska et al. (2006), Theriot nonetheless writes (in Alverson et al. 2007, p. 195, LC 32) that “the four outgroup taxa were chosen because they consistently comprise the sister clade to Thalassiosirales.” All four of the outgroup taxa are lithodesmioids. The word “consistently” in that publication stands in stark contrast to his doubts expressed in Theriot (2008) regarding this very same relationship shown in our paper. Kaczmarska et al. (2006, p. 122, RC 39, pp. 132–3, and table 3) clearly recognized and addressed the issues of H1 being dependent on lithodesmioids being a sister clade to thalassiosiroids. Sequence data available up to that date (nuclear-encoded SSU rRNA, plastid-encoded SSU rRNA, and mitochondrial cytochrome oxidase subunit 1) generated by independent laboratories (Medlin et al. 1997, Ehara et al. 2000, Kooistra et al. 2003) demonstrated support for lithodesmioids and thalassiosiroids as sister clades, and we did address exceptions in Kaczmarska et al. (2006, p. 132, RC 23). The most current work in Theriot’s own lab (Alverson et al. 2007) adds to the significance of this relationship. Comments regarding our evolutionary inferences. Kaczmarska et al. (2006) discussed three hypotheses that could explain the origin of the fultoportula: tubular (H1), areolar (H2), and annular (H3) origin in the context of the sister relationship between lithodesmioid and thalassiosiroid diatoms recovered by molecular phylogeny, schematically illustrated in 1-3. Two hypotheses (H2 and H3, and their permutations) have been known for nearly two decades, and as such we had to address them in our paper. The areolar (H2) hypothesis is known in two variants, one proposed by Syvertsen and Hasle (1982) and another by Round and Crawford (1984); we supported the latter. The annular hypothesis (annular process and multistrutted process are two names for the same structure) involving some extinct diatoms grossly similar to frustule architecture seen among lithodesmioids was already developed and discussed extensively in earlier publications (Gersonde and Harwood 1990, Medlin et al. 2000). In accordance with Gersonde and Harwood (1990) and Medlin et al. (2000), the annular origin of the fultoportula (H3) was introduced in a broader evolutionary context than H1 and H2 and suggested a sequence of several stages, beginning with Lower Cretaceous and extinct species to modern diatoms to illustrate the hypothetical connection of extinct diatoms to extant lithodesmioids and thalassiosiroids, because no molecular data are available for these extinct species. Because of this fact, H3 involved additional scenarios, some published before (Gersonde and Harwood 1990, Medlin et al. 2000), others under investigation in (then unpublished) projects (Kühn et al. 2006, Sims et al. 2006), and thus were not taken beyond cursory presentation. In comparison of all three hypotheses (H1–H3), Kaczmarska et al. (2006, p. 134, RC 45) focused on the events immediately preceding emergence of the fultoportula (our H3) outside of the broader issue of evolution of central structures involving many extinct lineages (Gersonde and Harwood 1990, Medlin et al. 2000). The hypothesis of a tubular origin of fultoportulae (H1) was new, and thus needed and received the most detailed treatment. Because Theriot (2008) referred to our hypotheses using his own labeling system, for clarity we schematically illustrate our hypotheses in 1-3 and compare these to Theriot’s best supported hypothesis shown in Figure 4. The closest correspondence between our hypotheses (H1–H3) and labels used by Theriot (2008) is the following: tubular hypothesis (H1) carries the same label in both papers; areolar hypothesis in our paper (H2) is discussed in Theriot (2008) on p. 830, LC 55 to p. 832, LC 1–12 as one scenario of his hypothesis 5; our annular hypothesis (H3) is referred to in Theriot (2008) in separate hypotheses. The closest correspondence between his inferences and our H3 is Theriot’s hypothesis 3. Theriot’s hypotheses 2 and 4 discuss ideas advanced in previous publications (Gersonde and Harwood 1990, Medlin et al. 2000) and was work we were aware of being performed in another laboratory (Kühn et al. 2006, Sims et al. 2006). For this reason, it was not advanced beyond general outline in our paper. Tubular origin of fultoportula (H1 in Kaczmarska et al. 2006, fig. 1). This hypothesis is parsimonious, sound, and consistent with the data at hand, despite Theriot’s claim to the contrary. It is unclear why Theriot (2008) embarked on an argument regarding “ridges” in T. weissflogii (pp. 823–5). We can find no evidence that we suggested that this diatom has ridges. Consequently, those arguments are irrelevant to our H1. He is in error claiming that we suggest that the lithodesmioid ridge system is the evolutionary precursor to marginal fultoportulae (Theriot 2008, p. 823, RC 2); again, we did not claim this. See Kaczmarska et al. (2006, p. 127, LC 21–22 and p. 133, LC 15) for a clear indication of what we meant. We introduced “valve face protrusions” listing best-known formats (spines, tubular and hair-like protrusions, projections and siliceous flaps) when used for the first time. Later, we shortened the list of types of protrusions to “marginal ridges or fimbria” to avoid cumbersome repetition. Theriot (2008, p. 824, LC 9–15) is incorrect in writing that we did not hypothesize positional and functional homology of fultoportulae and marginal protrusions. We did in several locations in Kaczmarska et al. (2006, table 3, p. 132, RC 56 and p. 134, LC 34–38). The hypothesis that marginal protrusions (not ridges) are homologues of fultoportulae (H1) is supported by functional and positional similarity of the characters, passing the test for similarity (Fig. 5, and text below). We hypothesized that a “process” (meaning a sequence of steps) whereby a marginal protrusion transformed into external tubes of fultoportulae might have resembled structures similar to those that are formed during valve morphogenesis in T. weissflogii. The term “margin” is used correctly in our writing (Round et al. 1990, pp. 280 and 290). Only two steps are required for a fultoportula in a tube to evolve from marginal protrusions (Fig. 1). Phylogeny of Thalassiosirales inferred from Bayesian analysis of SSU rDNA sequence alignment as published in Kaczmarska et al. (2006), with figure modified to show widely present existence of linking by fultoportulae (by exudates and interlock), as expressed in that publication and reflecting the normal understanding of the function of the fultoportula (see text for details). Data regarding linkage taken from Cupp (1943), von Stosch (1977, 1986), Ramirez (1981), Syvertsen and Hasle (1982), Gleser et al. (1988), Fukuyo et al. (1990), Round et al. (1990), Bérard-Therriault et al. (1999), and Sarno et al. (2005). It should be noted that although the particular species of Cyclotella included in this tree is solitary, species are present in this genus where fultoportular threads facilitate formation of colonies (Gleser et al. 1988, pp. 38, 47). Theriot (2008) claims that the cell-linking function of fultoportulae is incongruent with our H1 (p. 825, LC 22–26). However, this paragraph of his paper is difficult to interpret because we both agree that in most of the taxa of interest, extrusion of chitin threads through the fultoportulae is used to join cells (compare Kaczmarska et al. 2006, p. 133, LC 40 and p. 135, LC 50 to Alverson et al. 2007, p. 194, RC 1–3). The most compelling claim of incongruence in Theriot (2008) makes reference to “14 independent losses” being required to resolve “Group G” taxa (Theriot 2008, p. 825, LC 56). However, because neither our original paper (Kaczmarska et al. 2006) nor Theriot (2008) contain a “Group G” in any tree, this critique is hard to rebut or even understand. In any case, when the cell-linking function of the fultoportulae (inclusive of their exudates, Hasle 1973, p. 68, LC 23) is correctly mapped onto the tree (Fig. 5), there is good internal consistency of this character distribution, thus passing the congruence test. The hypothesized tubular origin of the fultoportula fails the conjunction test, sensu Patterson (1988) if we accept that “some of the forms similar to hypothesized intermediate stages in the evolution of the fultoportula within the external tube might have resembled an occluded process” (Kaczmarska et al. 2006, p. 133, LC 54). If occluded processes were taken as representing one of the early serial homologues of fultoportulae, the co-occurrence of fultoportulae and occluded processes fails the test of conjunction because Patterson (1988) does not recognize serial homologues as homologues in the strict sense (Theriot 2008, p. 822, LC 27). However, as discussed below, all the hypotheses fail this test, including the one Theriot regards as best supported. In summary, in contrast to Theriot’s claims, we show that H1 passes the functional similarity test and congruence test (Kaczmarska et al. 2006, table 3, p. 132, RC 56, and p. 134, LC 34–38; and Fig. 5 in this presentation). Theriot (2008) overlooks the fact that the first diagnostic character of Lithodesmiales is its bilabiate process and not their marginal protrusions (ridges being only one of the types present), and a combination of characters to define taxa is often used in diatom taxonomy (Round et al. 1990). We concur that H1, as all hypotheses in science, is sensitive to taxon sampling, irrespective of the approach. Areolar origin of fultoportula (H2 in Kaczmarska et al. 2006, fig. 2). Kaczmarska et al. (2006, p. 133, RC 20) explicitly state that fultoportulae might have evolved from areolae more directly and illustrate a clear structural resemblance between the two (Kaczmarska et al. 2006, fig. 3d), as also suggested by Round and Crawford (1984) by modification of the basal cribra (two steps in Fig. 2). We did not hypothesize that areolar fultoportulae were derived via an occluded process; Syvertsen and Hasle (1982, text fig. 6) did. Consequently, Theriot’s arguments (2008, pp. 829–30) regarding serial homology of the occluded process and areolae is irrelevant to our hypothesis H2. Whenever appropriate, we gave credit to the work of Syvertsen and Hasle (1982), even though we did not support their hypothesis. The purpose of Theriot’s criticisms of Syvertsen and Hasle (1982) in our paper is unclear to us. We stated and illustrated (Kaczmarska et al. 2006, p. 133, RC 20, fig. 3d) which of the two previously published hypotheses relating fultoportulae to areolae we follow. We clearly stated that this hypothesis recognizes structural similarity (passes similarity test) between areolae and fultoportulae and requires only simple modifications. Theriot arrived at the same conclusion and could not reject the hypothesis that fultoportulae might have evolved from areolae via modification of the cribra, just as we did and others before us. H2 passes the congruence test (as Theriot 2008 agrees) but fails the conjunction test (as did H1; Theriot’s best supported hypothesis also fails). The areolar origin of the fultoportula is just as well supported as is H1 once errors and misrepresentations of Theriot (2008) are dispensed with. Annular origin of fultoportula (H3 in Kaczmarska et al. 2006, fig. 3). Disagreement here centers on the evolutionary significance of ancient structures called either multistrutted or annular processes. We felt compelled to include H3 in our discussion because of the published record of this hypothesis, even though we had no molecular data to substantiate the phylogenetic relationship between the extinct and extant diatoms considered in our paper. “Multistrutted” and “annular” processes were discovered independently by Hasle and Syvertsen (1985) and Nikolaev and Strelnikova (1987), respectively. It therefore appears that the “obvious” similarity (Theriot 2008, p. 826, LC 53) of this process to the fultoportula was obvious to Hasle and Syvertsen but not to Nikolaev and Strelnikova. The process is known in extinct diatoms from the genera Thalassiosiropsis and Gladiopsis. Similar to H1 and H2, the annular hypothesis passes similarity and congruence tests but fails the conjunction test, because areolae and annular processes co-occur on the same valves. Our H1 and H2 place the evolving fultoportula onto valve architecture consistent with known Thalassiosirales, sister to Lithodesmiales (as stated in our goals, p. 122, RC 40–44) and nearly all modern diatoms; they have rimoportulae. The lack of rimoportulae in diatoms carrying annular processes was the main concern we had in 2003 (when Theriot reviewed this manuscript for the first time) and still have, even though morphological transformations needed to derive fultoportulae from annular processes are simple (Fig. 3, two steps), as we said before. However, no currently known species carry both annular processes and rimoportulae. Theriot (2008) claims that the annular process as an intermediate between areolae and fultoportulae is the most parsimonious (Fig. 4). For Theriot’s hypothesis to have one of two have before thalassiosiroid diatoms either the an annular process rimoportulae they from the common with thalassiosiroids, before the Lower Cretaceous when no known diatom or the thalassiosiroids evolved from an and it Theriot (2008, p. 831, LC 1–5) to support the first illustrated in Figure 4. Kaczmarska et al. (2006, p. 134, RC considered the scenario (Fig. 3). it is that to derive the fultoportula from the annular process (Fig. 3) is as parsimonious as to derive this from marginal protrusions, via with (Fig. or directly from an areola with cribra (Fig. 2); all only two All are more parsimonious than the sequence of steps suggested by Theriot (2008) involving both the areola and annular process (Fig. which four Our hypotheses H1 and H2 place the fultoportula in the context of modern diatoms with the sister relationship between Lithodesmiales and Thalassiosirales and is supported by molecular data (Medlin et al. 1997, Ehara et al. 2000, Kooistra et al. Kaczmarska et al. 2006, Alverson et al. the by Theriot’s lab and with him as Theriot (2008) Sims as his of did not support a relationship between Thalassiosiropsis and Gladiopsis. that these diatoms in several and thus are not as as previous work suggested pp. and Theriot (2008) is of two extinct genera carrying and fultoportulae from in which on the understanding of the origin of the Kaczmarska and (2008) of an even fultoportula from an diatom same as the diatom in Kaczmarska et al. 2006) and simple fultoportular are into the of the valve consistent with emergence of the that fultoportulae are present in extinct species from investigation as Theriot should be taken with as a of these have been to genera (Gleser et al. 1988, In summary, it requires only two steps to derive the fultoportula from the annular process under our Only two steps are also required under either of our hypotheses H1 and H2. All are parsimonious, as we stated in our paper. four steps to derive fultoportulae from areolae via the annular process, Theriot’s hypothesis (Fig. is parsimonious to those discussed in our with or without the sister relationship between Thalassiosirales and Lithodesmiales. The relationship between the two sister clades is supported by molecular data (as stated in Kaczmarska et al. 2006) and has support in the more recent publication from Theriot’s own lab (Alverson et al. consideration of the origin of the fultoportula outside this relationship to scientific evidence independently in several Comments of general First of all, we to Theriot’s (2008) that the of parsimony and of similarity (with to a of can only be within the and methodological framework of is a of analysis known for and used by many of and it is not for phylogenetic analyses conducted in a also agree that as a methodological is a part of scientific methodology in general (Sober 1988, p. 39; Brooks and McLennan 1991, p. 45). The authors that of parsimony in phylogenetic systematics is no from its use in any other of or any other there is no to that the of parsimony was not in our work even if Patterson (1988) was is also to in the work of nor is or of and internal consistency of the data set at and are not for specific types of phylogenetic analysis performed with of These are in scientific methodology systematics being one of the using them with a very specific used by including Kaczmarska et al. (2006). These no for Furthermore, similarity of has been used to homology with from common Theriot (2008, p. 826, RC claims that homology even be proposed without reference to Patterson However, similarity be considered as a test of homology because it the very of homology It a that was derived from it in the first 1991, p. than a test, it appears to be a case of is what evolutionary to with or without Patterson Two other tests for conjunction and have been by many where they not and 2000, and and specific discussed The conjunction test is by where homologues co-occur on the same in and with for and or and in species of and in and in all of of for a of in and on of and and 2000, and and These the and scientific of the test, in the strict as Theriot (2008, p. 822, LC We that all of our as well as Theriot’s fail this test. a test as as the test of congruence consistency is a in any has been shown to in where molecular homology is well for and et al. et al. 2006). It is well that molecular for or even are not from the of the thus their in and species et al. et al. et al. 2006) but are available for and in 2000, and We therefore that our understanding of the the and of homologues is still too to accept that all character total of all independent of the of the The independent of the characters if our hypothesis H1 were to be (Theriot 2008, p. 825, LC 56 and p. 830, LC clearly of the We use the to in Theriot’s and only as a possibility that set of from those and a discussion of when parsimony should or could not be in phylogenetic are by and Theriot p. This is from recent of our work (Theriot We wish to that we scientific and from a of accuracy and We to our work being as because we homology no reference to Patterson We also show that of our molecular analyses are present only in Theriot’s (2008) published data (including Theriot’s own support the tree from Kaczmarska et al. (2006) and the sister relationship between Lithodesmiales and Thalassiosirales, providing a morphology independent framework to homology among species Theriot be right in both doubt on the of our analysis and at the same time his own work the same trees. We show that our hypotheses were sound within the goals we and data we had available in when work was that they up to in are more parsimonious than Theriot’s hypothesis of fultoportula emergence involving extinct diatoms with an annular

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.