Simonsenia aveniformis sp. nov. (Bacillariophyceae), molecular phylogeny and systematics of the genus, and a new type of canal raphe system

The genus Simonsenia is reviewed and S. aveniformis described as new for science by light and electron microscopy. The new species originated from estuarine environments in southern Iberia (Atlantic coast) and was isolated into culture. In LM, Simonsenia resembles Nitzschia, with bridges (fibulae) beneath the raphe, which is marginal. It is only electron microscope (EM) examination that reveals the true structure of the raphe system, which consists of a raphe canal raised on a keel (wing), supported by rib like braces (fenestral bars) and tube-like portulae; between the portulae the keel is perforated by open windows (fenestrae). Based on the presence of portulae and a fenestrated keel, Simonsenia has been proposed to be intermediate between Bacillariaceae and Surirellaceae. However, an rbcL phylogeny revealed that Simonsenia belongs firmly in the Bacillariaceae, with which it shares a similar chloroplast arrangement, rather than in the Surirellaceae. Lack of homology between the surirelloid and simonsenioid keels is reflected in subtle differences in the morphology and ontogeny of the portulae and fenestrae. The diversity of Simonsenia has probably been underestimated, particularly in the marine environment.


Results
A new species of Simonsenia was found in natural population of diatoms collected along the southern coast of the Iberian Peninsula, in the estuaries of the Guadiana (3.85 km from the river mouth -the holotype habitat) and Arade rivers (6.48 and 9.85 km inland from the river mouth). In the Arade estuary its relative abundances ranged from 0.3 to 0.7 %.
Simonsenia aveniformis Witkowski, Gomes and Gusev, sp. nov.  Table 1. DESCRIPTION. Frustules very small, consistently nitzschioid (i.e. with the raphe systems of the daughter cells always formed on opposite sides of the parent following cell division), rectangular and with slightly bevelled corners in girdle view . Each cell with two simple chloroplasts, one located towards each valve end ( ETYMOLOGY: The specific name is derived from Latin name of oat = avena; "aveniformis" refers to the similarity of the new species to oat grains as observed under LM. Distribution and autecology. Simonsenia aveniformis is only known so far from the estuaries of the Guadiana and Arade Rivers in the Iberian Peninsula (Supplementary Dataset 1). At the sampling sites where S. aveniformis was observed, the salinity of the sediment interstitial water ranged between 19.94 (GS1) and 35.71 g/Kg (AS25) while pH varied from 6.77 (AS16) to 7.51 (GS1). The duration of the tidal inundation was between 15.4% (AS25) and 98.9% (GS1). The substratum grain-size varied between fine silt (AS14, AS16 and AS25) and very fine sand (GS1) and the organic carbon and nitrogen percentages ranged from 0.52% (GS1) to 4.25% (AS14) and from 0.06% (GS1) to 0.7% (AS14), respectively. Supplementary Dataset 2 presents in detail the physicochemical parameters measured at each sampling station where Simonsenia aveniformis was observed. In the surface waters of the river adjacent to the EI, PT and SF sampling transects, the salinity was similar to that of the sediment interstitial water, varying between 20.58 (SF, low tide) and 32.92 g/Kg (SF, high tide). Regarding pH, river water ranged from 7.8 (SF, low tide) to 8.31 (EI, high tide), the values being slightly higher than those measured in the sediment interstitial water. At sites where S. aveniformis was found, dissolved oxygen varied between 83.4% (EI, low tide) and 112.6% (EI, high tide). The temperature of the river water ranged between 17.50°C (EI, high tide) to 18.27°C (SF, high tide).
Raphe canal structure. Both cultivated and wild specimens were examined by SEM. A key aspect of S. aveniformis, linking it to the two Simonsenia described previously, is the structure of the raphe system. Along one side of each valve there is a marginally positioned keel, which is supported externally by solid rib-like braces (fenestral bars; Fig. 2). At its top the keel bears a tubular raphe canal (Figs 2B-D, 3C-D, 5A), which is elevated above the valve face to approximately the same extent over the whole of its length, except at the poles, where the height abruptly decreases and the canal joins rest of the valve. The outer walls of the raphe canal are solid and bear no ornamentation (Figs 2D, 3C,D). The raphe slit is simple and runs along the top of the raphe canal; it is not flanked by ridges. The fenestral bars are not visible internally; instead the course of the raphe canal is marked by a line of small circular portulae and wide fibulae, pierced by areolae like those of the valve face (Fig. 3). The tubular canal raphe is connected with the valve interior only through these portulae and at the apices. Correlated with the absence of central raphe endings, the central portula is no larger than any of the others (Fig. 3A). At the poles the raphe ends internally in a small helictoglossa (Fig. 3A,B), while externally there is a short, bent terminal fissure (Fig. 2B,C). The relationship between the fenestral bars and the fibulae and portulae is difficult to visualize, even from SEM images, and we have therefore provided 3-D reconstructions of the valve to aid interpretation (Fig. 4). The external braces (= fenestral bars) that support the raphe canal have almost exactly half the linear density (i.e. they are spaced twice as widely) as the striae, each brace lying opposite and between two of the transapical ribs separating the striae (Fig. 4). Near where each brace arises on the distal side of the raphe, the two transapical ribs opposite the brace converge and fuse (Fig. 4B, but see also 2A, 3A and 5A). Between some of the braces (roughly every second one), the inner wall of the raphe canal extends in towards the main part of the valve and fuses with it, creating the rounded portulae visible from the valve interior in SEM (Fig. 3A,B,E,F) or their openings into the raphe canal tube as seen in Focused Ion Beam (FIB) preparations (Fig. 6A) or in TEM (Fig. 3C,D). Internally, the valve face ribs and striae continue past where the transapical ribs fuse in pairs (i.e. the positions occupied externally by the bases of the fenestral bars) and merge to produce a somewhat irregularly porous fibula beneath the raphe (Figs 3A,B, 5G). Between adjacent portulae (i.e. directly above each fibula), there is a passage (fenestra) outside the cell, connecting valve face to mantle (Figs 5C, 6A). This means that, within the fenestra, the fibula is also the outer wall of the cell.
In SEM preparations of cultured material, we found some valves that had been in the process of formation when the material was fixed. In the earliest stages observed there was almost no sign of the portulae and fibulae: the valve consisted of (1) the raphe, with an incomplete U-shaped, open raphe canal beneath it, and (2) delicate transapical ribs and striae extending to the distal margin (Fig. 5A), though some modelling of the areolae was still ongoing. The only structures linking the raphe to the valve face or to the proximal margin at this stage were the incipient fenestral bars, each of which could be seen to be already linked to two of the transapical ribs on the distal side (i.e. the side where the bars join the valve face); on the proximal side each incipient bar terminated as a simple rib linked to the narrow valve margin (Fig. 6A). Close examination revealed that the incipient bars were differentiated internally into two types: (1) somewhat elevated bars terminating in "hammer-like" structures, and (2) almost flat (slightly concave) strips (Fig. 5D,E). The two types usually alternated, but occasionally hammer-type structures were formed on adjacent ribs. During the following stages of valve ontogeny, the hammer-like structures fused with each other and developed into the fibulae, whereas the flatter strips corresponded to where the portulae would form.
Internally, rows of knob-like projections developed on either side of each fenestral bar (Fig. 5D). Next, the U-shaped canal became modified into a tube, except for a hole -destined to become the portula -left opposite every second bar (arrows, Fig. 5E,F). Portula growth began with development of simple small, cylindrical structures on the inner side of the raphe canal (Fig. 5E), which grew inwards, elevating the raphe canal above the valve face. Through the further growth, branching and proliferation of the knobs, an inner porous wall was created beneath the raphe, forming the fibulae evident in the completed valve (Fig. 5G). This inner wall was continuous with the earlier-formed transapical ribs and striae of the valve face.  [38][39][40] , in this case the endosymbionts of Galeidinium rugatum, Durinskia baltica and Kryptoperidinium foliaceum. The newly described species Simonsenia aveniformis was placed in the Bacillariaceae clade and was sister to a clade containing Nitzschia inconspicua, N. frustulum, N. amphibia and Denticula kuetzingii with low support values (bs < 50%, pp = 0.89). This assemblage grouped with a clade composed of Pseudo-nitzschia species, Fragilariopsis cylindrus, N. frustulum and N. fonticola (bs = 63%, pp = 0.99). The trees indicated a wide separation between S. aveniformis and any members of the Surirellaceae (sensu 1 ), which instead were sister to Rhopalodiales (Figs 7, 8). However, in order to test further the evolutionary relationship between Simonsenia and the Surirellaceae, which share some characters with Simonsenia based on morphology, a constrained analysis was performed, grouping Simonsenia in the Surirellaceae. The result showed that the tree with S. aveniformis constrained to a clade with the Surirellaceae was significantly worse than the unconstrained tree by a SH test. Hence classification of S. aveniformis in or near the Surirellaceae can be rejected.

Discussion
Distribution and ecology of Simonsenia. It seems that Simonsenia is more widespread and species-rich than has previously been thought. Lange-Bertalot 16 noted that S. delognei (which was the only species known until 1983) was almost unrecorded for nearly a century after its original description, but now there are records of it from all continents apart from Antarctica (see Introduction). The discovery of S. aveniformis extends the known range of Simonsenia into marine habitats, adding another raphid genus to those already known to display evolutionary euryhalinity 41 . Due to the lack of diatom studies along the coastal zone of southern Iberia, S. aveniformis remained unknown until now. Here, though it was never abundant, it was present in small numbers in many samples 42 . Probably Simonsenia has been under-recorded because of its similarity in LM to small species of Nitzschia sect. Lanceolatae, which are notoriously difficult to separate and identify (e.g., 43 ). However, examination in SEM reveals a very different structure to that present in Nitzschia sect. Lanceolatae, where there are no fenestrae in the raphe keel and the fibulae are solid rib-or block like structures. Table 1 summarizes the morphological differences between S. aveniformis and the two previously described species of Simonsenia (we synonymize S. delognei subsp. rossii Lange-Bertalot & Krammer 29 with S. delicatula, following Witkowski 30 , since the diagnostic features of subsp. rossii given by Lange-Bertalot & Krammer are present in S. delicatula). Simonsenia aveniformis is clearly separated from the other two species in LM by the much finer structure of the striae. The major differences between the species, however, are best observed in SEM, and involve contrasting stria and fenestral bar structure. The differences appear to be fewer and less between S. delognei and S. delicatula than between either of these and S. aveniformis.
In terms of autecology, S. aveniformis is also distinct from S. delognei and S. delicatula. Simonsenia aveniformis was only observed in sites where the salinity of the sediment interstitial water varied between 19.94 and 35.71 g/Kg (Supplementary Dataset 2), suggesting that its occurrence may be at least influenced by salinity; thus the new species may be classified as a marine to brackish water species. In contrast, S. delognei and S. delicatula are reported in freshwater to brackish water environments (e.g., 25,28,30,42,44 ). It seems that the most favorable habitats for S. delognei and S. delicatula are riverine supralittoral areas, spring waters 28,30 or the upper reaches of the intertidal zones 42 , whereas S. aveniformis inhabits intertidal zones, regardless of the duration of tidal inundation and exposure. Despite its presence in both sandy and silty substrates (Supplementary Dataset 2) S. aveniformis, as well as S. delognei 30 , reveals a preference for silty sediments, thus it may be classified as a benthic epipelic diatom. According to the data presented by Witkowski et al. 30 and in Supplementary Dataset 2, both S. aveniformis and S. delognei occurred in environments where the pH of the sediment interstitial water was in the neutral range. Simonsenia aveniformis, like S. delognei 30 , also reveals high tolerance to pronounced variations in sediment organic carbon (0.52-4.25 %) and nitrogen content (0.06-0.7 %), which may indicate its resilience to organic pollution.
Systematic position of Simonsenia. In this paper we present for the first time a molecular phylogeny of canal raphe bearing diatoms that includes Simonsenia. Our results show that the Bacillariaceae and Surirellaceae are not closely related and their fibulate (canal-) raphe systems must have evolved independently. This is consistent with the phylogenetic analysis of Ruck and Theriot 10 , which was based on fewer taxa of canal raphid diatoms than ours but more genes (SSU and psbC in addition to rbcL). Our results clearly show that Simonsenia is unrelated to the Surirellaceae yet it has a similar fenestrated keel to some Surirella, Campylodiscus and Stenopterobia species, which must therefore reflect convergent evolution. It is interesting to examine whether this homoplasy is betrayed by subtle differences in keel structure, or in the way in which the fenestrae and fibulae are formed and this is considered further below (in "Structure and homology of the Simonsenia raphe system").
Classification of Simonsenia within the Bacillariaceae is consistent with chloroplast arrangement, since most of the family possess two chloroplasts, one towards each pole (reviewed by Mann 45 ), as in Simonsenia. Furthermore, in most other Bacillariaceae, like Simonsenia, the chloroplasts do not move around in the cell during the cell cycle, retaining their 'fore-and-aft' positions during cell division (Fig. 1A-C). It is also consistent with pore structure, since Simonsenia species have hymenate pore occlusions (Lange-Bertalot 16 , Lange-Bertalot & Krammer 29 , this paper) like other Bacillariaceae, but unlike the other major groups of canal raphe diatoms, the Rhopalodiales and Surirellales, where the areolae are closed by flaps of silica (e.g., 46 ).
The Bacillariaceae is shown to be monophyletic in our analyses, with good support, but within the Bacillariaceae, the ML and BI phylogenetic trees contain many nodes lacking statistical support, so that the relationships of Simonsenia cannot be fully determined. Nevertheless, both analyses indicate that Simonsenia aveniformis is located on a long branch and is closely related to a clade containing some species of Nitzschia sect. Lanceolatae and Denticula kuetzingii. Moreover, Bayesian analysis gives good support for a clade containing these taxa with some further members of Nitzschia sect. Lanceolatae (accessions of N. fonticola and N. frustulum CCMP558) and also Pseudo-nitzschia and Fragilariopsis. Hence the genus Nitzschia is para-or polyphyletic in our trees and this is consistent with other analyses of the Bacillariaceae that have used rbcL (e.g., 22,25 ) or other markers (e.g. 33,36 ) to reconstruct phylogeny.
Structure and homology of the Simonsenia raphe system. Our study shows that, although the molecular phylogeny indicates that the fibulate raphe systems of Simonsenia and other Bacillariaceae are essentially homologous, it is obvious that they are also very different, with Simonsenia having a much more complicated structure than other Bacillariaceae, with fenestrae in the keel and fenestral bars crossing them (cf. Lange-Bertalot 16,29,30,47 , and this paper). None of the other genera and clades of Bacillariaceae have such a raphe structure. Instead, they have a raphe canal that is integrated into the valve structure, rather than being elevated on a fenestrated wing as in the 'simonsenioid' raphe. Furthermore, the fibulae are solid rods, ribs or blocks (e.g., 1,45 ), rather than the perforated plates of Simonsenia, which function not only as bridges linking the two sides of the valve, but also as part of the external wall of the cell through which it communicates with its environment. The only other Bacillariaceae known to us with similarly porous fibulae to those of Simonsenia are some species of Tryblionella, e.g. T. debilis (Mann 45 ,  Fig. 730), in which the valve face striae continue without interruption across the interior of the fibula to the proximal mantle. The structure of the keel needs further study in these species, but it is already clear that fenestrations like those of Simonsenia are lacking.
Thus, we can distinguish a third type of canal raphe, the "simonsenioid" type, in addition to the "nitzschioid" and "surirelloid" ones recognized hitherto. In our Fig. 6 we provide a summary comparing the three types. This Figure illustrates the principal differences, based on FIB preparations of specimens from natural populations and supplemented with diagrammatic presentations modified from Ruck & Kociolek 22 in the case of Nitzschia and Surirella. The simonsenioid raphe system has converged towards the surirelloid type but the molecular phylogeny shows clearly that it developed from the nitzschioid type. In such cases of homoplasy, detailed inspection sometimes reveals differences that were previously overlooked and we therefore examined the Simonsenia raphe for hints of its separate evolutionary origin. At first sight, the canal raphe does seem extremely similar in Simonsenia and Surirellaceae, as can be seen by comparing the images of Surirellaceae in Round et al. 1 22 (particularly those of Stenopterobia delicatissima and S. densestriata) with those presented here for Simonsenia aveniformis (Fig. 6A-C) or by Witkowski et al. 30 for S. delognei. However, close examination suggests that there is a difference between the two in that the alar canals and fenestrae of Surirellaceae are formed by undulations (180° out of phase) of the valve face and mantle, which bring the two alternately close together (to fuse and create the fenestrae) and far apart (to create the alar canals). In Simonsenia, on the other hand, there are no undulations. This is consistent with the little that is known about ontogeny: in Surirellaceae the fenestrae and portulae are formed by changes in the orientations of the transapical ribs as the ribs extend out from the raphe sternum 48,49 , while in Simonsenia the raphe canal and portulae are formed after the valve face and mantle structure is essentially complete, if not yet fully thickened, essentially by extension and branching (rather than the orientation) of what become the fenestral ribs in the mature valve.

and Ruck & Kociolek
In contrast to the Surirellaceae, where the question is whether Simonsenia has a different raphe ontogeny despite a similar final morphology, in Bacillariaceae the point to examine is whether Simonsenia has a similar ontogeny to other genera, even though the mature valves have a very different structure. In Nitzschia and Denticula, the formation of the valve has been studied and begins with the deposition of the raphe canal, continues with the formation of the transapical ribs and poroids, and is completed by the addition of fibulae, which develop unilaterally, growing across from one side of the valve to fuse with the other side [50][51][52] . This pretty much resembles our observations in Simonsenia, except that in Simonsenia the fibulae seem to develop from both sides of the valve simultaneously (via the 'hammer structures').
Taxonomic conclusion. Molecular

Fieldwork. Fieldwork was carried out between 2010 and 2012 in the Guadiana and Arade River
Estuaries, yielding the material used here for microscope observations and species culture, and also the environmental measurements (cf. Supplementary Dataset 1 and Gomes et al. 53 ). Further information concerning the sampling approach and the environmental measurements is available in Gomes et al. 53 Microscopy. Light microscopy (LM) and electron microscopy (EM) observations involved the same processing procedure as published in Witkowski et al. 56 . LM observations of cleaned material were conducted with a Zeiss Axio Imager 2 (Carl Zeiss, Jena, Germany) with a 100x oil immersion objective (n.a. 1.46). For chloroplast imaging we used the same method as for LM examination of frustules, but in addition, to avoid squashing the cells we placed Tesa tape on the slide (to raise the cover glass and maintain space between the slide and cover glass). As the newly described species has a very small size, we supplemented our own LM observations with a few images kindly taken by ing. Wulff Herwig using his advanced light photomicrography system, for which a detailed description is presented at http://www. microscopy-uk.org.uk/mag/artmar11/Advanced_Light_Photomicrography.pdf.
Ultrastructural analysis was made with scanning and transmission electron microscopy (SEM and TEM, respectively). For SEM examination, a drop of the cleaned sample was filtered onto Whatman Nuclepore polycarbonate membranes (Fisher Scientific, Schwerte, Germany). Filters were air-dried overnight, mounted onto aluminum stubs, and coated with gold-palladium or osmium. SEM observations were made at the University of Frankfurt using a Hitachi S-4500 (Hitachi, Tokyo, Japan) and at the Warsaw University of Technology, Faculty of Materials Science and Engineering, using a Hitachi SEM/ STEM S-5500. The observations and deconstruction of Simonsenia aveniformis, Nitzschia sp. and Surirella sp. raphe were performed by means of a Hitachi NB5000 integrated system. The system consists of ultra-high performance Focused Ion Beam (FIB) (40 kV) and high resolution field emission (FE)-SEM (30 kV). This dual beam system enabled high throughput specimen preparation, high resolution imaging and analysis, and precision nanodeconstruction. It enables SEM imaging both during and after FIB deconstruction. The special low-damage deconstruction technique has been applied during diatom processing dedicated for materials sensitive to electron irradiation. The process of deconstruction was performed at accelerating voltage of 10 kV for FIB. The siliceous parts of Simonsenia aveniformis raphe were cut at nanoscale and imaged. The diatom specimens have been imaged prior and during particular steps of deconstruction. The deconstruction has been performed on specimens from natural samples. The raphe system was cut primarily along the transapical axis to image the canal supporting system and the connection of canal with the cell lumen via portulae; however, a few cross sections of the canal along the apical axis were made as well.
Modelling of the raphe system. Model images were generated using the computer aided drafting software AutoCAD. Images of Simonsenia were imported into the program in order to derive relative measurements and build a closely approximated basic model. Improvements were made using other images showing various other details and angles. The final model was completed after a number of iterations of the model building process DNA extraction, amplification and sequencing. Genomic DNA was isolated from 50 ml of a two-week-old algal cell culture using the Gene Elute ™ Plant Genomic DNA Miniprep Kit (Sigma-Aldrich, Hamburg, Germany) according to the manufacturer`s instructions. The chloroplast-encoded gene (rbcL) was amplified from genomic DNA using the proof reading polymerase Phusion (Finnzymes, Thermo Scientific, Schwerte, Germany) according to the manual. The primers used for amplification and the protocol of amplification are the same as in Witkowski et al. 56 specified for rbcL sequence.
Phylogenetic analysis. Maximum likelihood (ML) tree of the diatom phylogeny was constructed by means of the program RAxML version7.2.6 57 with 93 taxa using Ctenophora pulchella and Tabularia cf. tabulata as outgroups. The data, comprising the rbcL sequences of Simonsenia aveniformis and other rbcL sequences downloaded from GenBank (Supplementary Dataset 3) were partitioned by codon position and analysed with the GTR+ G+ I model. The analysis consisted of multiple runs (20), each with 1000 bootstrap replicates and the tree with the best log likelihood score was chosen as final maximum likelihood estimate. Using the same procedures as for the unconstrained analysis, a constrained ML analysis was performed, forcing Simonsenia aveniformis into the Surirellaceae family. The Shimodaira-Hasegawa test 58 was run using the best constrained tree and the best unconstrained tree from the previous analysis, also using RAxML v 7.2.6 (see Table 2).
For the Bayesian Inference tree, two Bayesian inference analyses each with 4 chains (one cold and three heated), were run with Mr Bayes v.3.2 using the same substitution model and partitioning method as the ML analyses. 10 8 generations were run per analysis with sampling every 1,000th iteration, generating in total of 10 5 samples. The final 10 3 trees were used to get a majority rule consensus tree and obtain posterior probabilities for nodes.  Table 2. Results of the Shimodaira-Hasegawa (SH) test for the comparison between the best phylogenetic tree and the constrained phylogenetic tree.