Gomphonella olivacea (Bacillariophyceae) – a new phylogenetic position for a well-known taxon, its typification, new species and combinations

Background and aims – Within the project “German Barcode of Life – Diatoms” common diatoms of German waters were routinely isolated and cultivated. In order to understand the taxonomy and phylogeny of the genus Gomphonema , one of the most common taxa of Central Europe, known currently either under the name Gomphonema olivaceum in Europe or Gomphoneis olivacea in America, was studied. Methods – Twenty unialgal strains were established from five different water bodies in Germany and one from Lake Balaton, Hungary, which supplied molecular data (18S V4 and rbc L) besides morphometric and ultrastructural data. In addition, on eight populations from different water bodies including the type from Denmark, morphometric and micromorphological studies by light and scanning electron microscopy were performed. Key results – Molecular and micromorphological data show that the target taxon neither belongs to Gomphonema Ehrenb. nor to Gomphoneis Cleve. By reinstating the genus name Gomphonella Rabenh., the nomenclatural and taxonomic enigma of this taxon is solved, and with the presentation of the type by Hornemann the authorship of the epithet is clarified. Molecular data for the unialgal strains and several environmental clones show that there is more diversity in the Gomphonella olivacea clade than can be identified morphologically. In addition, the establishment of the new species Gomphonella coxiae and Gomphonella acsiae is supported. The molecular data classified Gomphonella species as belonging to the Cymbellales but not to the Gomphonemataceae. In addition, molecular data put Gomphoneis tegelensis R.Jahn & N.Abarca also into Gomphonella. In order to make the genera Gomphoneis and Gomphonema monophyletic, their astigmate members are transferred to Gomphonella . Conclusions – The results clarify that the gomphonemoid outline is not restricted to the family Gomphonemataceae but seem to be distributed across the entire order Cymbellales. This is shown in this paper for the revived genus Gomphonella , which contains the astigmate group of Gomphoneis and Gomphonema besides the longly disputed G. olivacea . Only a polyphasic approach, combining molecular and micromorphological data for taxonomy, nomenclatural evaluation, and observations from clonal cultures can reveal the full intricacies of evolutionary relations.


INTRODUCTION
The taxon currently known under the name Gomphonema olivaceum (Hornem.) Ehrenb. or Gomphoneis olivacea (Hornem.) P.A.Dawson ex R. Ross & P.A.Sims is a very common diatom in fresh waters of Central Europe. Its habit of living on stalks or being free-living and producing fair amounts of mucus had originally placed it into different genera since, at the beginning of diatom research, life forms where thought to be decisive for phylogeny. In 1810 Hornemann pictured an olive coloured mass of mucus from a Danish river and named it Ulva olivacea Hornem. (fig. 1). Lyngbye (1819) used his material to describe and draw G. olivacea-like cells and recombined the name as Echinella olivacea (Hornem.) Lyngb. Kützing (1833) named it Frustulia olivacea (Hornem.) Kütz. and Brébisson & Godey (1835) Cymbella olivacea (Hornem.) Bréb. & Godey. When it was recognized that those two growth habits -attached on stalks or freely moving -were just two different life forms of the same species, the species was subsumed under the genus Gomphonema Ehrenb. In 1838 Ulva olivacea was transferred twice: in July/August by Ehrenberg and in October by Brébisson. In 1853 Rabenhorst published the new genus name and combination Gomphonella olivacea (Hornem.) Rabenh., a name which was reduced to a section of Gomphonema by Brun (1880) and which, since then, has apparently been neglected by the diatom community; nevertheless, in the Index Nominum Genericorum (2018) Gomphonella Rabenh. is listed as a genus with an unassigned type of the name of the genus even though a type was given in Round et al. (1990).
In parallel to the above sketch of the nomenclatural history of the epithet olivacea, there are further names which seem to refer to the same taxon. Agardh (1824) gave it the superfluous name Meridion vernale C.Agardh. In 1830 Leiblein, using diatom material from waters near Würzburg, Germany, published a picture of an unnamed gomphonemoid diatom and discussed the taxonomic identity of Meridion vernale. Leiblein sent this material among others to C. Agardh in Lund who described Gomphonema leibleinii C.Agardh validly from this material (Agardh 1830). Kützing (1833) accepted Agardh's name whereas Ehrenberg (1838) put Gomphonema leibleinii into synonymy with his Gomphonema clavatum Ehrenb., which he had described validly in 1832 (in Ehrenberg 1830, it was a nomen nudum). Ehrenberg (1838) also used (and transferred) the taxon name Gomphonema olivaceum, which meant that for him Gomphonema olivaceum and Gomphonema clavatum (= Gomphonema leibleinii) were not conspecific. In 1844 Kützing put Gomphonema leibleinii into synonymy with Gomphonema olivaceum and also put Gomphonema clavatum into synonymy with Gomphonema subramosum C.Agardh (see also Reichardt 2015).
Despite its complex early history, this taxon has been known for 140 years as Gomphonema olivaceum. With the advent of the electron microscope it became obvious that its micromorphology was different from most Gomphonema. Dawson (1974) proposed that, because of its biseriate striation, it should be separated from Gomphonema and put into the genus Gomphoneis Cleve, which had been erected by Cleve (1894); this proposal was formally correctly executed by Ross & Sims in 1978. Since that time in the USA (Kociolek 2011) this taxon has been assigned to the genus Gomphoneis as Gomphoneis olivacea (Hornem.) P.Dawson ex R. Ross & P.A.Sims (1978) whereas in Central Europe it stayed within Gomphonema (Hofmann et al. 2013, Levkov et al. 2016 because biseriate striation was seen as common in this genus also and therefore not identified as a differentiating feature (Reichardt 2007). In addition, since it lacks an axial plate and mantle lamella it was seen as not fitting the genus Gomphoneis (Krammer & Lange-Bertalot 1985).
The logical approach for solving nomenclatural and taxonomic enigmas is to locate the type specimen or original material that was in the hands of the first describer. Locating this material was quite challenging, since 200 years ago Prof. Hornemann had worked in Copenhagen (Denmark), and Lyngbye, who had studied this material in more detail and published figures, had gotten his education in Copenhagen but was later based in Lund (Southern Sweden). Requests were made to both Herbaria, resulting in the finding of Agardh's Gomphonema leibleinii in the Lund Herbarium (plus comments made in 1982 by an earlier researcher). The type material of Ulva olivaceum had been loaned in 1965 to the Diatom Herbarium in Philadelphia, USA, by the Copenhagen Herbarium (C) but fortunately, an intensive search in Philadelphia resulted in the finding of this material after 52 years! Within the framework of the project "German Barcode of Life -Diatoms" we were finally successful in isolating and cultivating several strains of this taxon. In our studies to understand the taxonomy and phylogeny of the genus Gomphonema (Abarca et al., in prep.), we also questioned its phylogenetic position and taxonomic affiliation with morphological as well as molecular data. Since these data are very different from the core group of Gomphonema yet similar to Gomphoneis tegelensis R.Jahn & N.Abarca (Skibbe et al. 2018), we are publishing the results here separately.

MATERIAL AND METHODS
The original material from Denmark of 1810 was studied (Lectotype C-A9208; see Typification below). In addition, data from eight populations and 38 strains are included in the present study (see electronic appendices 1 & 2). Thirtytwo strains were established by the authors. The sequence data for the other six strains were downloaded from ENA/ Genbank (as part of the International Nucleotide Sequence Database Collaboration, INSDC). All sequences downloaded from INSDC were BLASTed (basic local alignment search tool) against the INSDC database to test for taxonomic consistency.

Field collection and cultivation
Freshwater samples were collected from Germany and Hungary between 2004 and 2017. Twenty unialgal strains of the target taxon were isolated from nine samples of six different waters in Germany and from Lake Balaton in Hungary (for details see electronic appendices 1 & 2). Clonal strains were established by micropipetting single cells using a stereo microscope (Olympus, Japan) and an inverted LM (Olympus, Japan). All strains were treated according to Romero & Jahn (2013). Non-axenic unialgal cultures were maintained at room temperature (19-25°C for cultures until 2016), 10°C (in 2016) and at 20°C (in 2017) in a growth chamber. A 12:12 h light/dark photoperiod from a daylight LED light source following Jahn et al. (2017) was applied. In addition, seven populations from the original samples from which clonal cultures were established were used for supporting documentation of the morphologies of the clones (for details see electronic appendix 2).

Documentation and vouchering
For all newly established strains the frustule preparation and morphological documentation were executed following Zimmermann et al. (2014). LM pictures of live cells ( fig. 2) and of permanent specimens on slides were taken with a Zeiss AxioImager.M2 (Zeiss, Germany). SEM images were taken with a Hitachi FE SEM 8010 (Hitachi, Japan) of unsputtered material. The vouchers for all new strains are deposited at B (Herbarium Berolinense), where long-term stable and semantic web compatible identifiers for specimens are used according to Güntsch et al. (2017). Molecular data for all isolates are deposited in INSDC (see electronic appendix 1). DNA samples are stored in the Berlin DNA Bank and are available via the Genome Biodiversity Network (GGBN, Droege et al. 2014); nomenclatural acts are registered (Turland et al. 2018, Art. 42) in PhycoBank (continuously updated). Data are available through AlgaTerra (Jahn & Kusber continuously updated).

Morphological criteria
Besides valve outline and morphometric measurements of each clone (length, width, number of striae in 10 μm (electronic appendix 2) valves were investigated under SEM to compare internal and external valve and girdle views (fig. 5A-C). Special attention was given to the presence or lack of

DNA extraction, sequencing and alignment
Cultured material was transferred to 1.5mL tubes. The DNA was isolated using the NucleoSpin® Plant II Mini Kit (Macherey and Nagel, Düren, Germany) or Qiagen® Dneasy Plant Mini Kit (Qiagen, Valencia, CA) following the respective product instructions. The DNA fragment size and concentrations were measured via gel electrophoresis (1.5% agarose gel) and Nanodrop® (PeqLab Biotechnology, Erlangen, Germany), respectively. The DNA samples were stored at −20°C for future use and finally deposited in the Berlin collection of the DNA bank network (Droege et al. 2014). The polymerase chain reaction (PCR) for rbcL was conducted following Abarca et al. (2014). The V4 section of the 18S SSU rRNA gene locus (18S V4) was amplified and PCR performed following Zimmermann et al. (2011). PCR products were visualized in a 1.5% agarose gel and cleaned with MSB SpinPCRapace® (Invitek LLC, Berlin, Germany) following standard procedures. DNA concentrations were measured using Nanodrop® (PeqLab Biotechnology) and samples were normalized to a total DNA content >100 ng μL −1 for sequencing.
Sanger sequencing was conducted by Starseq® (GENterprise, Mainz, Germany), rbcL gene according to Abarca et al. (2014) and 18S V4 according to Zimmermann et al. (2011). In both cases the same primers were used for amplification and sequencing. The editing, as well as the quality control of the pherograms for the new sequences, were done in Phyde® (Müller et al. 2010). The evaluated sequences were aligned using MUSCLE (Edgar 2010), as implemented in MEGA6 (Tamura et al. 2013) with subsequent manual adjustments in case of 18S V4. The lengths of the newly generated sequences were 432 bp for 18S V4 and 979 bp for rbcL. For comparison, the alignments for 18S V4 and rbcL also included other sequences of our own and some others from INSDC representing the Cymbellales D.G.Mann, as well as Achnanthidium saprophilum (H.Kobayasi & Mayama) Round & Bukht., which was added as the outgroup for the phylogenetic tree generation following Kermarrec et al. (2011). All added sequences were trimmed to fit to the newly generated sequences for 18S V4 as well as rbcL The accessions used are given in electronic appendix 1).

Phylogenetic analyses
Two different data sets (18S V4, rbcL) were used for the phylogenetic analyses. Each dataset was analysed using Maximum Likelihood (ML) as implemented in RAxML (Stamatakis 2006, 2014, Stamatakis et al. 2008) using the CIPRES platform (Miller et al. 2010) in both cases.
For the ML analysis of the molecular datasets, the optimal model of sequence evolution that best fits the sequence data was calculated under the hierarchical likelihood ratio test (hLRT) and the Akaike information criterion (AIC) using model test 3.7 (Posada & Crandall 1998). The best fitting model was GTR+G+I (Tavaré 1986). A ML analysis was conducted using RAxML 8.2.8 (Stamatakis 2006, 2014, Stamatakis et al. 2008, ML search option (GTR+G+I) and 1,000 bootstrap replicates (model GTRCAT as implemented in RAxML for the rapid bootstrap algorithm). Additionally, Bayesian phylogenetic inference was conducted for both data sets using MrBayes v. 3.1.2. (Huelsenbeck & Ronquist 2001, Ronquist & Huelsenbeck 2003) with the same model. The default settings were used, runs with four incrementally heated Metropolis-coupled Monte-Carlo Markow Chains and runs with 10 million generations were executed. The runs were sampled every 1000 generations, the first 25% generations being discarded as burn-in; the rest were used to calculate a 50% majority rule consensus tree. The best ML tree found by RAxML and the 50% majority rule tree of the BI analysis were compared for rbcL as well as 18S V4. In all cases the trees showed no different topologies and were therefore summarized in one tree for each marker, showing bootstrap statistics (> 75) for ML (LB) and (> 0.90) posterior probabilities (BI). Trees were drawn using FigTree v. 1.4.2 (Rambaut 2008) and Adobe Illustrator (Adobe Systems, San Jose, CA). Genetic distances for 18S V4 and rbcL were calculated using MEGA6 (Tamura et al. 2013) and the implemented pdistance option.
Clade 2 is the sister group to Clade 3 in the 18S V4 (1.00 posterior probabilities, fig. 3) as well as in the rbcL tree (83/1.00 bootstrap value/posterior probabilities, fig. 4). In the case of Clade 4 the two trees show slight differences in the topologies regarding the sister group relation. For 18SV4 ( fig. 3) Clade 4, supported by 99/1.00 bootstrap value/posterior probabilities, is the sister group to the branch with Clades 1a, b, 2 and 3. Clades 2 and 3 for their part are building the sister group to Clade 1a, b supported by 75/1.00 bootstrap value/posterior probabilities. In contrast, for the rbcL tree ( fig. 4) Clade 4 is the sister group to the branch with Clade 2 and 3 with a support of 87/0.98 bootstrap value/posterior probabilities and these three Clades (Clades 2, 3 and 4) are the sister group to Clade 1A, B with a support of 76/1.00 bootstrap value/posterior probabilities. Clades 1A and 1B -From a molecular point of view the 13 clones (for details see electronic appendix 1) from Tegeler See (samples D128 and D130), Müggelsee (D129) and river Main (D135_024) are the same (Clade 1A). They show well supported but small differences -p-distance = 0.7% for 18SV4 and 0.4% for rbcL -to the two clones of river Spree (D03_184), and Saale (D132_024) (Clade 1B). The molecular data from the environmental sample named Cymbellales from brook Westerhöver Bach (Brinkmann et al. 2015) sits in Clade 1A. Clade 2 -The two strains with data for 18S V4 and the four strains for which rbcL data are available, are all from Lake Balaton (D140; see electronic appendix 1). They show 0% differences between each other in 18SV4 and rbcL. There are well supported differences from the other three clades: For 18SV4 the p-distances are 1.6-2.3% to Clade 1, 1.6% to Clade 3 and 5.1% to Clade 4; for rbcL they are 3.5% to Clade 1, 2.9% to Clade 3 and 5.1% to Clade 4. Clade 3 -This clade is defined by the data of strain D201_007 for which 18S V4 and rbcL data are available. For 18S V4 the p-distances are 2.3-2.8% to Clade 1, 1.6% to Clade 3 and 5.3% to Clade 4; for rbcL they are 2.9% to Clade 1, 2.2% to Clade 2 and 4.6% to Clade 4. Clade 4 -This clade is defined by the data of the isolate D221_Gt, which has been separated into a new species, Gomphoneis tegelensis R.Jahn & N.Abarca (Skibbe et al. 2018). Concerning 18S V4 data, this strain has a p-distance of 4.9-5.3% to Clade 1, 5.1% to Clade 2 and 5.3% to Clade 3; concerning rbcL data, this strain has a p-distance of 5.0% to Clade 1, 5.1% to Clade 2 and 4.6% to Clade 3.

Morphology (figs 5-14)
The most conspicuous trait of all studied strains is their pronounced variability in outline: from the typical ovate-oblanceolate to symmetrically lanceolate, with apices that are They often grew fan-like (strains D129_043 and D132_024) or in lumps (strain D132_036) producing plenty of gelatinous material. Since the strains have such variable und often untypical outlines which might be due to long lasting cultivation, we also studied the populations of the samples from which the strains were isolated such as the populations from Tegeler See ( Concerning morphological synapomorphies, the most important and conspicuous feature of clades 1-4 visible in LM is that they do not have any stigmata ( fig. 5A & B); occasionally, there are a few isolated puncta visible in the central area (e.g. figs 7AR, AS & 8K) but they look like the areolae and seem to be continuations of the striae, just separated by a gap. In SEM, these puncta have no internal structure and are therefore not true stigmata (see also definition in www. diatoms.org). A further conspicuous trait visible only in SEM is their biseriate striation (in Clade 4 only, some striae are triseriate) with small round uniform areolae not occluded by siliceous flaps (see Skibbe et al. 2018: figs 17-21). The striae are not interrupted near the valve face/mantle junction and continue onto the valve mantle ( fig. 5C). These double rows of areolae can terminate either as single or as double rows along the axial area and mantle. But all striae taper into only a single row of areolae at the central area ( fig. 5A). The foot pole is composed of a bilobed field of porelli. The outside distal ends of the raphe extend through these porelli but not all the way to the end of the mantle. The raphe is either straight or very slightly undulated at both apices. Internally, the pseudosepta are wide, distinct and prominent at both apices. Also the helictoglossae are prominent at both poles and in some clones lie well away from the valve terminus.
The autapomorphies separating the clades (and species) are aspects of the striation, such as the number of striae in 10 µm, their parallel or radial direction throughout the valve face, and the form of the central area. Of special importance too are the porelli at the footpole, which are either of similar size and shape to the areolae of the striae or relatively larger and distinct; they are either arranged in double rows or with- out any order and they are located close to the striae or well separated. Clade 1 (figs 6-10 & 14A) -The valves are heteropolar, clavate with broadly rounded headpoles and acutely rounded footpoles ( fig. 6A-P). There is wide variation within the clone cultures: the valves can be heteropolar, clavate with broadly rounded headpoles and acutely rounded footpoles (see fig. 6Q), but they can also be heteropolar, lanceolate to linear lanceolate (see fig. 6R), or even be only slightly heteropolar valves with almost parallel to slightly convex margins and rounded headpole and footpole (see fig. 6S). The axial area is narrow, straight, expanded at the centre to form a rectangular, bow-tie-shaped to transversally elliptical central area bordered at the margins by 1-3 approximately equally- to clavate in smaller specimens, with narrowly rounded headpoles and acutely rounded footpoles. Wide variation of shape occurs within the clone cultures, since the valves can be slightly heteropolar, lanceolate and widest at the centre, or they can be heteropolar, clavate with broadly rounded headpoles and acutely rounded footpoles. The valves of the population of Clade 2 are on average longer and have a higher stria density than the club-shaped populations of Clade 1, but both have the same average width. Valves of Clade 2 are differentiated from valves of Clade 1 by a transapically widened central area, which is smaller and more rectangular  13J). The footpole has a large apical pore field with relatively large porelli, which differ in size from the areolae of the striae and are well separated from the striation. Valves of Clade 3 possess parallel transapical striae, which become radial at the centre as in Clade 2 ( fig. 13A-I), whereas in Clade 1 the striae are radial throughout the valve face. Clade 4 (illustrated in Skibbe et al. 2018) -The most prominent differences from Clade 1, 2 and 3 are the striation, which is bi-to triseriate, and the small axial plate and mantle lamella (for details see Skibbe et al. 2018). Otherwise, there are no stigmoids and the areolae are small, round and uniform, and are not occluded by siliceous flaps. The striae are not interrupted near the valve face/mantle junction and continue onto the valve mantle (Skibbe et al. 2018: figs 17, 18, 23).

Typification and nomenclature
Clades 2 and 3 cannot be identified with already known taxa and need to be described as new (see below). Specimens of Clade 4 were recently described as Gomphoneis tegelensis (Skibbe et al. 2018). Specimens of Clade 1 were described more than 200 years ago as Ulva olivacea ( fig. 1), which went through a number of name changes until today. We were able to study this material in LM and SEM for the first time (see figs 5A-C & 6A-H).
Since clade 1 obviously does not belong to the genus Gomphonema ( fig. 5D-F) and an earlier valid genus name exists, namely Gomphonella Rabenhorst (1853), we are here reinstating this name. Rabenhorst introduced this genus name and made the combination Gomphonella olivacea (1853: 61), describing it as "Eine gestielte Gomphonema in einer gestaltlosen Gallertmasse" [a stalked Gomphonema in an amorphous gelatinous mass].
Gomphonella Rabenh. (Rabenhorst 1853: 61, pl. IX) Original description -"bis 2/100 Mm lang, verkehrt-eiförmig-lanzettlich; Nebenseiten breit keilförmig, am Rande mit zarten Querstreifen. Durch ganz Europa." (Rabenhorst op. cit.) [up to 20 µm long, ovate oblanceolate; sides broadly wedge-shaped, with delicate bars at the margin. Throughout the whole of Europe]. Type species (lectotype) -Gomphonella olivacea (Hornem.) Rabenh. (Ulva olivacea Hornem.), designated in Round et al. (1990: 691). Registration -https://phycobank.org/100348 Emended description -The most important and conspicuous feature of this genus visible in LM, separating it from other gomphonemoid taxa, is that there are no stigmoids or stigmata present on the valve face. A further conspicuous trait separating it from Gomphonema s. str. but visible only in SEM is the bi-to triseriate striation with small round uniform areolae not occluded by siliceous flaps. The diameter of the areolae is about 100 nm. The striae sit in moderately deep alveolae between thicker vimines. The striae continue onto the valve mantle. The apical foot pole is composed of a bilobed field of porelli which are round and similar to the normal areolae in the striae (= undifferentiated AFPs). The raphe is filiform and either straight or very slightly undulate at both apices. Internally, both polar raphe endings end in prominent helictoglossae well away from the valve terminus and there is an intermissio in the centre with the raphe end-    fig. 6Q), or heteropolar and lanceolate to linear lanceolate (i.e. fig. 6R), or be slightly heteropolar with almost parallel to slightly convex margins and rounded headpole and footpole (e.g. fig. 6S). In both natural populations and clone cultures, an axial plate and mantle lamella is lacking. The axial area is narrow, straight, expanded at the centre to form a rectangular, bow-tie-shaped to transversely elliptical central area bordered at the margins by 1-3 approximately equally-shortened striae arched around the central area (figs 5A, B, 8B, E, H, K, 9B, E, H, K, 10B, E). Except for occasional isolated puncta, which seem to be simple areolae as continuations of the central striae (figs 7AR-AS &  10A). The footpole has a large apical pore field with porelli of the same size and structure as the areolae. The porelli appear to be arranged in double rows, located close to the striation and therefore undifferentiated structurally and spatially from them (figs 8C, I, & 9C, I). Both apices have distinct pseudosepta (figs 8D, F, J, 9D, F & 10D, F). Fig. 7AD-AN Fourteen strains show the same molecular data for 18SV4 and rbcL (see electronic appendix 1), but two strains (D03_184 Spree, fig. 7AF-AN; and D132_024 Saale; figs 7AD-AE & 10A-F) are slightly different and show p-distances from the others of 0.7% (18SV4) and 0.4% (rbcL). These two strains could represent a separate variety but since no morphological differences have been found, we are refraining here from naming and ranking this taxon, just using instead the neutral  term genodeme to mark molecular differences from genodeme 1 (i.e. the epitype of G. olivacea, see above).  fig. 11H; iso-: BP, slide HNHM-ALG-D002300). Description -The morphometric data of the type population (n = 37) are: length 19.2-44.5 μm, width 6.6-8.8 μm and 12-15 striae in 10 µm ( fig. 11A-F). Valves are slightly heteropolar, lanceolate in larger specimens to clavate in smaller specimens with narrowly rounded headpoles and acutely rounded footpoles. The morphometric data of the clone cultures (n = 58) are: length 13-39.7 μm, width 4.5-6.8 μm and 11-15 striae in 10 µm. The valves have a wide variation of shape within the clone cultures: they can be slightly heteropolar, lanceolate and widest at the centre, or heteropolar, clavate with broadly rounded headpoles and acutely rounded footpoles (figs 11G-S). In both, the natural population and clone cultures, an axial plate and mantle lamella is lacking. The axial area is narrow, straight, expanded at the centre to form a small, rectangular central area bordered at the margins by 1-3 irregularly-shortened striae slightly arched around the central area ( fig. 12B, E, H & K). Stigmoids in the central area are lacking. Raphe lateral, with external proximal ends slightly dilated ( fig. 12B & H), extending into the central area; external distal raphe endings extend straight ( fig. 12A) (in some valves slightly deflected, fig. 12C) onto the valve mantle at both poles. Internal proximal raphe endings curved in the same direction and located on a raised central nodule ( fig. 12E & K). Internal distal raphe ends -terminal nodules or helictoglossae -are distinct, in the foot pole positioned well before valve terminus ( fig. 12F & L). The striae are biseriate composed of small, round areolae not occluded by siliceous flaps that terminate as single rows around the central area ( fig. 12B, E, H & K). The internal structure of the single rows of areolae is more silicified in some valves, bordered by thickened vimines at the central area ( fig. 12E). At the headpole, the striae are composed of two alternating rows of areolae ( fig. 12A & G). Transapical striae are parallel, becoming radial at the centre 11-15 in 10 µm. The footpole has a large apical pore field with relatively large porelli, different in size from the areolae of the striae and well separated from the striation ( fig. 12C & I). Both apices have distinct pseudosepta (figs 12D, F & L).
In both the natural population and clone cultures the axial plate is lacking. The axial area is narrow, straight, expanded at the centre to form a broad, bow-tie-shaped to rectangular central area bordered at the margin by 2 or 3 approximately equally-shortened striae ( fig. 13L-M). Stigmoids are lacking. Raphe lateral, with external proximal ends dilated, extending into the central area ( fig. 13L); external distal raphe slightly bent before the apical point and extending onto the valve mantle at both poles ( fig. 13J & N). Internal proximal raphe endings curved in the same direction and located on a raised central nodule ( fig. 13M). Internal distal raphe endings -terminal nodules or helictoglossae -are distinct, positioned well before valve terminus ( fig. 13O). The striae are biseriate and composed of two small, round, and alternating but not occluded rows of areolae ( fig. 13J); they terminate as single rows but are not arched around the central area. Transapical striae parallel, becoming radial at the centre. The footpole has a large apical pore field with relatively large porelli, which differ in size from the areolae of the striae and are well separated from the striation ( fig. 13Na, b). Both apices have distinct pseudosepta ( fig. 13K & O). Registration -https://phycobank.org/100352 Etymology -We are dedicating this species to Dr. Eileen Cox for her outstanding contributions to diatom research as an author, editor and colleague. In addition, she organized and hosted the first German Speaking Diatom Meeting when she was doing research at the Limnological Station in Schlitz, Germany, in 1987.

New combinations
As explained further in the Discussion section, astigmate taxa of the genera Gomphoneis and Gomphonema from several localities around the world need nomenclatural transfer to the genus Gomphonella. We effect those transfers here. But we are refraining from recombining taxa where SEM data is missing for an unambiguous demonstration that there is no stigma, typical striation, areolae and porelli (see table 1). This means that we had to rely on recent descriptions or interpretations. And we are refraining from recombining many of the infraspecific taxa of G. olivacea because we think that they need to be studied also molecularly in order to find out if we are dealing here with species or only outline variations. Differentiating features are listed in table 1. (Kociolek &

Taxon name
Corresponding reference  Table 1 -Gomphonella taxa, their morphometrics and their ultra-structural differentiating features as documented in the corresponding reference.

Striae (in
Table 1 (continued) -Gomphonella taxa, their morphometrics and their ultra-structural differentiating features as documented in the corresponding reference.

Taxon name
Corresponding reference

Axial plate
Mantle lamella

Striae in central area
Porelli at footpole Pore field close to striae Basionym -Gomphonema perolivaceolacuum Levkov, Phytotaxa 30: 28, figs 185-198, 241-245. 2011(Levkov & Williams 2011 the jelly was the most conspicuous character ("Substantia gelatinosa lubrica supellucida"). Rabenhorst (1859) also used this feature to differentiate his new genus Gomphonella from Gomphonema: a stalked Gomphonema in an amorphous gelatinous mass ("Eine gestielte Gomphonema in einer gestaltlosen Gallertmasse"). When we took Gomphonella olivacea into culture we often noted that it grew in thick spherical lumps (O. Skibbe, unpubl. obs.). When we collected Gomphonella coxiae from reeds in Helenesee, it was surrounded by a 1 cm thick periphyton covering. When we sampled Lake Balaton, the stones at the shore of the lake were covered with a thick jelly which we expected to be Didymosphenia M.Schmidt (also known as 'rock snot' because of its jelly production). In the lab though, we did not find a single Didymosphenia but lots of Gomphonella specimens and some Cymbella. Gomphonella olivacea -identified as Cymbellales sp. because of missing taxonomic reference dataseems to be involved in the biogenic calcite precipitation and tufa formation of karstic streams (Brinkmann et al. 2015).

Distribution
Identified by its morphology alone (e.g. Kociolek 2011, Levkov & Williams 2011, Hofmann et al. 2013, Levkov et al. 2016, Gomphonella olivacea seems to be cosmopolitan, occurring with its many described varieties all over the world. For Central Europe, Hofmann et al. (2013) considered it to be the most frequent "Gomphonema". Many of its varieties have been raised to species rank but it needs to be studied if these represent real species or just outline variations. When searching recent literature for SEM pictures of Gomphoneis and Gomphonema species which unambiguously show relevant synapomorphies [astigmate valves, bi (to tri)-seriate striae with small round uniform areolae not occluded by siliceous flaps continuing unto the valve mantle, an apical foot pole composed of a bilobed field of porelli], we noticed that a diversity of apparently related species has been recently described from Asia such as China (You et al. 2013) and Lake Baikal  and to a lesser extent from Europe, such as from Lake Ohrid (Levkov & Williams 2011, Levkov et al. 2013) and Germany (this study). Howeverexcept for Gomphonella olivacea -none have been reported from the Americas, such as the USA (Spaulding & Edlund 2009) or Mexico (Vázquez et al. 2011. Whether this is a biogeographical pattern due to limited distribution or to missing investigations needs to be determined in future studies, when the autopomorphies of the genus Gomphonella in contrast to the genera Gomphoneis and Gomphonema have become better established. Nevertheless, these non-occurrences of Gomphonella in the Americas are paralleled by Cleve's (1894) original remark on the distribution of Gomphoneis taxa: that this genus is found in the freshwaters of North and Central America.

Classification
From a molecular point of view Gomphonella olivacea is not closely related to Gomphonema as defined by its type G. acuminatum; the gene trees show that it does not even belong to the Gomphonemataceae (figs 3 & 4; and Skibbe et al. 2018: fig. 27). This is also supported by the different morphologies, which are even discernible in LM since Gomphonella has no stigmoids, although occasionally there can be a few isolated puncta in the central area, set apart from the striae (figs 7AR, AS & 8K). The biseriate striation of the Gomphonella taxa previously included in Gomphonema, which has been explicitly assumed to be non-differentiating by Reichardt (2007) and Hofmann et al. (2013), is in fact constructed differently to the biseriate striation of Gomphonema s. str., which is clearly discernible if the valves are studied in the SEM (already seen by Dawson, 1974). To illustrate this fact, we here compare the micromorphology of Gomphonella olivacea ( fig. 5A-C) with the taxon Gomphonema minutum ( fig. 5D-F), since both have the same outline. In LM the only difference is the stigma in Gomphonema minutum. But in SEM, they look clearly different: Gomphonema minutum has biseriate areolae which are reniform and irregularly arranged and covered internally with flaps, in contrast to the round and regularly parallel arranged areolae with clear cut edges and no noticeable covering in Gomphonella. Also, the striae continue onto the valve mantle and end in a single well separated pore in Gomphonema minutum; in contrast to the abrupt stria endings in Gomphonella. Other differences include that the raphe is slightly bent at the foot pole and head pole in Gomphonema minutum; in contrast to the straight filiform raphe of Gomphonella; and the porelli at the foot pole are well set apart from the striae and the pores have a different pattern than the areolae of the striae (= differentiated APF) in Gomphonema minutum; in contrast to the undifferentiated apical porelli at the footpole of Gomphonella.
Molecularly, Gomphonella olivacea is also not closely related to Gomphoneis minuta (J.L.Stone) Kociolek & Stoermer (1988a) as published in Genbank. In contrast, it is closely related to G. tegelensis as studied by Skibbe et al. (2018). Most interesting though is that these two taxa (minuta and tegelensis), which had until now been put into the same genus Gomphoneis, are only distantly related (rbcL p-distance: 7.2-8.3%; 18SV4: 11.89-12.59%), with Gomphoneis minuta sitting closer to the Gomphonema acuminatum group (18SV4 and rbcL p-distances 3.3-3.5%) (Abarca et al. in prep.) within the Gomphonemataceae, whereas G. tegelensis is sister to other Gomphonella species. This shows that our current concept of the genus Gomphoneis is polyphyletic. Unfortunately, the type of the name of the genus Gomphoneis -Gomphonema elegans Grunow in Van Heurck 1880 as Gomphoneis elegans (Grunow) Cleve 1894 according to Boyer (1928) -has not been studied molecularly in order to guide us to understand where the real Gomphoneis belongs phylogenetically; we presume that the true Gomphoneis will also not belong to the Gomphonemataceae (Cox 2015). This means that currently we have to rely on morphological differences only to delimit Gomphonella from Gomphoneis.
Cleve ( formed this new genus for some species formerly considered as belonging to Gomphonema, but differing from it both in the structure and the presence of the longitudinal lines. In these characteristics they agree nearly with Scoliotropis, but differ in the straight median line, and the asymmetrical form of the valve. To Gomphoneis may perhaps also belong G. eriense Grun. The few known species of G. are all of freshwater habitat and are found in North and Central America. G. elegans Grunow 1880, G. herculeanum Ehrenb. 1845. plus var. robusta Grun. 1878." In contrast to Cleve's definition of Gomphoneis, Gomphonella olivacea has no stigma or stigmoid at all, no decussating lines of the double puncta and also no longitudinal lines. Nevertheless, Dawson (1974) transferred G. olivaceum and G. quadripunctatum to Gomphoneis (correctly done formally by Sims & Ross 1978) because of their double rows of simple round pores in contrast to the reniform poroidal structures of Gomphonema taxa.
In their paper on the phylogenetic relationship of Gomphoneis, Kociolek & Stoermer (1989) used morphologicalcladistic methods to introduce two subgeneric groups, namely the elegans-group which contains the type of the genus, and the herculeana-group. This grouping helps to define the true Gomphoneis. Gomphoneis minuta (≡ Gomphoneis herculeana var. minuta J.L.Stone) belongs to the herculeanagroup and therefore to the Gomphonemataceae and will not be discussed here any further (Abarca et al. in prep.). In a more recent paper on the Gomphoneis of Lake Baikal, Kociolek et al. (2013) defined the elegans-group further as having biseriate striae, undifferentiated porelli at the footpole, septa and pseudosepta, and a central nodule with internal proximal raphe ends recurved in the same direction. They proposed a further division into the typical elegans-group with four or more stigmoids (simple openings around the central area that may have siliceous ingrowths), and a group lacking stigmoids. We think that the taxa of the last group, those lacking stigmoids, should be recombined under the genus Gomphonella because the synapomorphies of Gomphonella olivacea are very similar to G. tegelensis: the same size and details of the areolae, the undifferentiated porelli at the footpole, the straight raphe and the missing stigmoids or stigmata (compare Skibbe et al. 2018). Conclusions drawn from these micro-morphologies are supported by the molecular data, which cluster Gomphonella olivacea, G. acsiae, G. coxiae and G. tegelensis into one group with very high bootstrap support (figs 3 & 4). The axial plate and mantle lamella, which are small but prominent features of G. tegelensis (see Skibbe et al. 2018) and, according to Cleve of the genus Gomphoneis in general, do not seem to play such an important differentiating role on the genus level as proposed by Kociolek & Stoermer (1988b, 1988c, 1989, 1993. Puzzling are the unclear differentiating features -autapomorphies -of the genus Gomphosinica, which was recently described by Kociolek et al. (2015), having been split off from Gomphoneis (for a morphological comparison see table 2). Although it looks like a Gomphonella concerning the areolae, undifferentiated porelli at the footpole and straight raphe endings, it has a stigmoid with internal stigmoid covering that seems to be typical for the Gomphoneis elegans group (containing the type species of Gomphoneis). This suggests to us that, also in diatom research, new genera should be described with molecular data supporting ultrastructural features. Otherwise taxonomical artefacts are produced which hide the true evolutionary scenario. Recently, You et al. (2013) reiterated that the genus Gomphoneis is monophyletic and that the small species without stigmoids -Gomphonella species in this study -are highly derived members of Gomphoneis, as Kociolek & Stoermer (1989, 1993 had postulated on morphological cladistical evidence alone. Even Nakov et al. (2014), who used molecular data, stated that Gomphoneis is monophyletic and Gomphonema polyphyletic even though molecular data for only one Gomphoneis species -several strains of Gomphoneis minuta -was available. However, the genus Gomphoneis was clearly polyphyletic until now, because unrelated groups of taxa -the herculeana-group and the two elegans-groups -were subsumed under this genus name. The polyphyly of Gomphoneis supported by molecular data was first shown by Skibbe et al. (2018); in this paper we are resolving at least part of this polyphyly by moving the astigmate taxa from Gomphoneis to the reinstated genus Gomphonella.

CONCLUSION
This study shows that the gomphonemoid outline is not restricted to the family Gomphonemataceae but seems to be distributed across the entire order Cymbellales. This had already been shown for Didymosphenia but now it is clear also for the revived genus Gomphonella, which contains the astigmate group of Gomphoneis and Gomphonema besides the long disputed G. olivacea. Although the Herculeana-group of the polyphyletic genus Gomphoneis seems to belong to the Gomphonemataceae, we presume that the true members of the genus Gomphoneis -the elegans group -will also not cluster within the Gomphonemataceae, although this cannot be tested since no molecular data of G. elegans are currently available. The outcome of this study shows that only a polyphasic approach, combining molecular and micromorphological data for taxonomy, nomenclatural evaluation, observations from clonal cultures, and ecology, will reveal the full intricacies of evolutionary relations within specific organism groups, laying the foundation for future evolutionary, taxonomical and ecological studies as well as the sound application in monitoring approaches using diatoms.

SUPPLEMENTARY DATA
Supplementaty Data are available in pdf at Plant Ecology and Evolution, Supplementary Data Site (https://www.ingentaconnect.com/content/botbel/plecevo/supp-data) and consists of: (1) list of material examined; and (2) morphometric data of the studied Gomphonella strains or populations.