Description of the Soil Diatom Sellaphora terrestris sp. nov. (Bacillariophyceae, Sellaphoraceae) from Vietnam, with Remarks on the Phylogeny and Taxonomy of Sellaphora and Systematic Position of Microcostatus

In material isolated from soils of Cát Tiên National Park, we isolated four strains that were assigned to the genus Sellaphora. Identification was carried out on the basis of morphological and molecular studies. We proposed a new species named Sellaphora terrestris sp. nov. An evolutionary distance matrix, based on the 18S rDNA gene including the V4 domain, showed the new species shared 94.1–97.2% similarities with other Sellaphora sequences. The new species is morphologically similar to species previously identified as representatives of the genus Microcostatus.

The monophyly of Sellaphora has been well supported in a number of studies [6,9,[13][14][15]. Currently, there are 550 sequences for 39 taxonomic units of Sellaphora deposited in GenBank (data on 1 August 2022), of which 17 have been identified to the species level, four taxa have been identified presumptively, and 18 have not been identified.
Sellaphora pupula is the type species of the genus [16]. The species complex of S. pupula (Kützing) Mereschkowsky (230 sequences in GenBank; data on 1 August 2022) has been the most studied taxa within the genus. It is a common and frequently reported species and has long been classified as a taxon of high morphological variability [17,18], resulting in many variations being described [19]. Study of the molecular structure of this species complex has been primarily related to questions of species boundaries, the biological significance of variations, the study of cryptic and pseudocryptic diversity, biogeography, relationships of phylogeny and morphometric characters, and problems of speciation [6,8,12,20,21]. This species complex was also used as a convenient model for testing the resolution of various genetic markers [9,22]. As a result, it has been shown that "S. pupula (Navicula pupula sensu Hustedt [17,23]; Krammer and Lange-Bertalot [24]) represents a paraphyletic group" [12], while morphological synapomorphies for species separation have not been clearly defined. Academy of Science, Moscow, Russia. Representative specimens. Strain VP304 (slide no. 07062a), VP272 (slide no. 07030), VP303 (slide no. 07061), VP299 (slide no. 07057); sample Kt33 (slide no. 07223), Kt59 (slide no. 06989), Kt60 (Slide no. 07232).
Etymology. The specific epithet "terrestris" refers to the terrestrial habitat from where the new species was extracted.
Distribution. So far only known only from the Cát Tiên National Park (agricultural and forest soils).
Сomment. We investigated the four strains of S. terrestris (VP304, VP272, VP299, VP303) and found that there were some morphological differences expressed across the strains. In particular, strain VP299 ( Figure 4) has smaller valves, almost elliptical, without distinct apices (length 5.2-6.9 μm and width 3.7-4.0 μm) in comparison with VP304, VP303, and VP272 (length 11-14.3 μm and width 4.5-5.1 μm). Apparently, the strain VP299 presented the smallest valves during the valve diminution series. Despite these  Etymology. The specific epithet "terrestris" refers to the terrestrial habitat from where the new species was extracted.
Distribution. So far only known only from the Cát Tiên National Park (agricultural and forest soils).
Comment. We investigated the four strains of S. terrestris (VP304, VP272, VP299, VP303) and found that there were some morphological differences expressed across the strains. In particular, strain VP299 ( Figure 4) has smaller valves, almost elliptical, without distinct apices (length 5.2-6.9 µm and width 3.7-4.0 µm) in comparison with VP304, VP303, and VP272 (length 11-14.3 µm and width 4.5-5.1 µm). Apparently, the strain VP299 presented the smallest valves during the valve diminution series. Despite these differences, all of these strains form one branch with maximum statistical support on the phylogenetic tree. , and 3M, black arrows). Raphe is straight, lying in raised, slightly asymmetrical conopeum ( Figures 1M,N, 2M,N and 3M, white arrowheads). Proximal raphe ends almost straight, drop-shaped ( Figures 1N and 2N, black arrows). Distal raphe ends bent to one side and extending to the valve mantle ( Figure 1M,O and 3M, white arrows). Areolae rectangular to rounded, 60-65 in 10 μm.
SEM, internal view (Figures 1P-R, 2P-R, 3N-P and 4M-O). The axial area is quite wide, moving into an even wider central area ( Figures 1P and 5O, black arrows). The striae in the central part are noticeably shortened, up to 3-5 areolae per striae ( Figures 1Q and  3N, white arrows). Areolae covered by hymenes ( Figure 2Q, white arrow). Raphe filiform, central raphe ends unilaterally slightly deflected ( Figure 1P; 6O, white arrowheads), distal   Figures 1P and 5O, black arrows). The striae in the central part are noticeably shortened, up to 3-5 areolae per striae ( Figures 1Q and 3N, white arrows). Areolae covered by hymenes ( Figure 2Q, white arrow). Raphe filiform, central raphe ends unilaterally slightly deflected ( Figures 1P and 6O, white arrowheads), distal raphe ends curved in opposite directions and terminate as small helictoglossae  Variability of the morphology of our species both in the natural material and in culture was noted (Table 1, Figure 5). This variability is expressed mainly in the shape of the valve ends. The cultivated strains are distinguished by having narrow valve apices. However, the ends of the valves range from beak-shaped (in large specimens) to broadly rounded (in small specimens). The shape of the valves themselves varies from linear-lanceolate and linear-elliptical to elliptical. At the same time, such features as the shape of the central and axial areas are the same. Quantitative characteristics in natural populations and images from cultures are quite similar. The length of valves in natural samples was 6.9-17.1 μm; in culture, it was 5.2-14.3 μm. The width of valves in natural samples was 4.2-6.1 μm; in culture, this parameter was 3.7-5.1 μm. Striae density in specimens from nature was 29-32 μm; in culture, the striae density was 28-35 μm (Table 1). Despite changes in the shape of the valves of the species introduced into culture, molecular genetic analysis allowed us to reliably establish the identity of the studied strains (Figures 6 and  7). Variability of the morphology of our species both in the natural material and in culture was noted (Table 1, Figure 5). This variability is expressed mainly in the shape of the valve ends. The cultivated strains are distinguished by having narrow valve apices. However, the ends of the valves range from beak-shaped (in large specimens) to broadly rounded (in small specimens). The shape of the valves themselves varies from linear-lanceolate and linear-elliptical to elliptical. At the same time, such features as the shape of the central and axial areas are the same. Quantitative characteristics in natural populations and images from cultures are quite similar. The length of valves in natural samples was 6.9-17.1 µm; in culture, it was 5.2-14.3 µm. The width of valves in natural samples was 4.2-6.1 µm; in culture, this parameter was 3.7-5.1 µm. Striae density in specimens from nature was 29-32 µm; in culture, the striae density was 28-35 µm (Table 1). Despite changes in the shape of the valves of the species introduced into culture, molecular genetic analysis allowed us to reliably establish the identity of the studied strains (Figures 6 and 7).

Molecular Analysis
Phylogenetic relationships within the genus Sellaphora were reconstructed using the maximum likelihood (ML, where LB is the bootstrap value) and Bayesian (Bayesian inference, where PP is the posterior probability) methods. The analysis used two approaches to determine the position of new Sellaphora soil strains: building trees based on the rbcL Mann remained outside these clades. In the rbcL tree, the studied soil strains of Sellaphora formed a single subclade with high statistical support (99 LB, 1.0 PP). This subclade is located in the basal part of the Sellaphora tree and was not included in the previously described clades 1-3. At the same time, S. seminulum strains, with the support of 78 LB, 1.0 PP act as a sister subclade. The addition of the 18S rDNA gene to the sequence analysis slightly changed the topology of the phylogenetic reconstruction, while the composition of clades 1-3 remained the same. It should be noted that the position of the S. minima strains was different between the two analyses. If in the rbcL tree they were between clades 2 and 3, then in the rbcL and 18S rDNA tree they are outside clades 1-3 ( Figure 7). Moreover, in the extended reconstruction, soil strains of Sellaphora moved away from strains of S. seminulum and formed a single clade with strains of S. laevissima (Figure 7).  [9], and Mann and Poulíčková [12] are in bold.   Strain numbers (if available) are indicated for all sequences. Clades indicated sensu Evans et al. [9], and Mann and Poulíčková [12] are at the right.
Partial ribosomal 18S rDNA gene fragments comprising the V4 domain of the new soil Sellaphora strains showed 100% similarity. The VP304 strain was selected for further comparison of the percentage similarity. The evolutionary distance matrix based on the partial 18S rDNA gene showed that the VP304 strain shared 94.1-97.2% similarities with the other Sellaphora strains (Table 2), and 96.7% and 97.2% similarities with the closest strains in the phylogenetic tree (S. laevissima THR1, S. laevissima THR4, Figure 7). Within clade 1B, similarity was noted at the level of 99.5-100%. Sellaphora strains within clade 2A are characterized by 98.2-99.2% similarity, with clade 2B having 99.0-99.7%. The greatest differences between Sellaphora strains were noted within clade 3, with up to 2.7% ( Table  2). In general, strains VP304, THR1, THR4, and B385 had a high similarity of the 18S rDNA gene region (97.2-99.5%), which allowed us to propose a new clade 4 within the genus Strain numbers (if available) are indicated for all sequences. Clades indicated sensu Evans et al. [9], and Mann and Poulíčková [12] are at the right.

Molecular Analysis
Phylogenetic relationships within the genus Sellaphora were reconstructed using the maximum likelihood (ML, where LB is the bootstrap value) and Bayesian (Bayesian inference, where PP is the posterior probability) methods. The analysis used two approaches to determine the position of new Sellaphora soil strains: building trees based on the rbcL gene, with the inclusion of all relevant nucleotide sequences available in GenBank (a total Plants 2022, 11, 2148 9 of 22 of 106 Sellaphora strains); and tree construction based on pooled rbcL and 18S rDNA gene alignments (35 Sellaphora strains in total). In both cases, as in previous studies on the phylogeny of Sellaphora strains, most of the sequences included in the analysis formed several well-supported clades (clades 1-3, Figure 6).
The strains of S. balashovae Andreeva, Kulikovskiy and Kociolek; S. laevissima (Kützing) D.G. Mann; S. minima (Grunow) D.G. Mann; and S. seminulum (Grunow) D.G. Mann remained outside these clades. In the rbcL tree, the studied soil strains of Sellaphora formed a single subclade with high statistical support (99 LB, 1.0 PP). This subclade is located in the basal part of the Sellaphora tree and was not included in the previously described clades 1-3. At the same time, S. seminulum strains, with the support of 78 LB, 1.0 PP act as a sister subclade. The addition of the 18S rDNA gene to the sequence analysis slightly changed the topology of the phylogenetic reconstruction, while the composition of clades 1-3 remained the same. It should be noted that the position of the S. minima strains was different between the two analyses. If in the rbcL tree they were between clades 2 and 3, then in the rbcL and 18S rDNA tree they are outside clades 1-3 ( Figure 7). Moreover, in the extended reconstruction, soil strains of Sellaphora moved away from strains of S. seminulum and formed a single clade with strains of S. laevissima (Figure 7).
Partial ribosomal 18S rDNA gene fragments comprising the V4 domain of the new soil Sellaphora strains showed 100% similarity. The VP304 strain was selected for further comparison of the percentage similarity. The evolutionary distance matrix based on the partial 18S rDNA gene showed that the VP304 strain shared 94.1-97.2% similarities with the other Sellaphora strains (Table 2), and 96.7% and 97.2% similarities with the closest strains in the phylogenetic tree (S. laevissima THR1, S. laevissima THR4, Figure 7). Within clade 1B, similarity was noted at the level of 99.5-100%. Sellaphora strains within clade 2A are characterized by 98.2-99.2% similarity, with clade 2B having 99.0-99.7%. The greatest differences between Sellaphora strains were noted within clade 3, with up to 2.7% (Table 2). In general, strains VP304, THR1, THR4, and B385 had a high similarity of the 18S rDNA gene region (97.2-99.5%), which allowed us to propose a new clade 4 within the genus Sellaphora based on these strains, and which has a separate position on the phylogenetic tree of the rbcL and 18S rDNA genes ( Figure 2). The only strains of Microcostatus sp. (MMDL 54701 and MMDL 54702) differed significantly, by 6.9%, according to the 18S rDNA gene sequence. At the same time, the differences with the available sequences of the 18S rDNA gene of Sellaphora strains reached 14.6%, which may indicate either a distant relationship with Sellaphora strains, or incorrect identification of strains MMDL 54701 and MMDL 54702, since there are no photomicrographs and other information for these strains.   S. lundii Wetzel, Barragán, and Ector [27] differs from S. terrestris sp. nov. by its elliptical shape of the valves (in S. terrestris sp. nov., the valves have a linear-elliptical shape) and wider ends of the valves (in S. lundii, the valves have a shape from rostrate to slightly capitate; in our species, the ends of the valves have a shape from rostrate to broadly rounded). The striae density in S. lundii is generally lower than in S. terrestris sp. nov. (23-28/10 µm versus 28-35/10 µm). The areolae of S. lundii are larger and rectangular, while the areole of S. terrestris sp. nov. are of a smaller diameter and only partially rectangular to mostly rounded. The main similarity between S. lundii and S. terrestris sp. nov. is the presence of distinct grooves in the central area of the valves on both sides of the proximal raphe ends.
When studying the morphological features of the new species, we noticed morphological similarity with some taxa from the genus Microcostatus. The genus Microcostatus was described in 1998 [44], primarily based on the presence of transversely oriented microcostae located in the axial area of the valve; the shape of the chloroplast, which is two plastids located along the cingulum of frustule [44]; and the lack of a conopeum but presence of an elevated sternum inside, with two longitudinal depressions alongside, sensu [44]. Initially, the type of genus (M. krasskei) showed open areolae and on this basis the authors assigned the genus to the family Naviculaceae [44]. Later, for representatives of the genus, a hymenes was shown that individually covered each of the areolae on the inner surface of the valves [41,43]. The diversity of morphological characters in the species currently assigned to the genus Microcostatus cannot but draw attention to itself. For example, the microcostae in the species can be expressed to a greater or lesser, the structure of the striae is different (they consist of one macroareola, for example, in the species Microcostatus elisabethianus Van de Vijver and Ector or several areolae in striae, as in most species of the genus), the conopeum/pseudoconopeum may be present or absent, and the hymenes closes the openings of the areolae in different species from the inside or even from the outside (M. elisabethianus). A common property for a number of species of Sellaphora, Okhapkinia alexanderii Glushchenko, Kulikovskiy, and Kociolek in Kulikovskiy et al. [5] and Microcostatus (e.g., M. dexteri, M. egregius, M. naumannii) is the presence of a round apical pit on each internal valve margin.
Sellaphora terrestris sp. nov. is morphologically similar to Microcostatus dexterii Stanek-Tarkowska, M. Rybak, and Czyż [43] and Microcostatus naumanii (Hustedt) Lange-Bertalot [41] ( Table 3). The greatest similarities between these species are observed using SEM. Microcostatus dexterii [43] and S. terrestris sp. nov. (Figures 3M-O, 4M,N and 5M) have a raised, slightly asymmetrical raphe sternum. The orientation of the outer central and distal raphe ends also coincides in the species. As in the case of S. terrestris sp. nov. (Figures 3M,N, 4M,N and 5M), M. dexterii has longitudinal depressions that widen into the valve center and form an almost longitudinal lanceolate central area [43]. The authors did not indicate any pronounced microcostae, which are important morphological feature of the genus Microcostatus. Both species have a structure typical for some Sellaphora, that is, round apical pits at the apex of the valve ends [43]. Microcostatus dexterii differs from S. terrestris sp. nov. by its valve shape (lanceolate to elliptical-lanceolate vs linear-elliptical to elliptical), lower striae density (20-24/10 µm vs 28-35/10 µm), and areolae (45-50/10 µm vs 60-65/10 µm). The shape of the areolae in M. dexterii is also almost square, in contrast to S. terrestris sp. nov., where the shape of the areolae can be either almost square or rounded. The ends of the valves of the species M. dexterii are widely rounded; in S. terrestris sp. nov., the shape of the ends is more varied, from rostrate to widely rounded.
Microcostatus naumanii is similar to S. terrestris sp. nov. in its valve width (4-5 µm in M. naumanii vs 3.7-6.1 in S. terrestris sp. nov.). The main similarity of the discussed species was observed in the SEM. The areolae of both species are round to rectangular in shape. In both species, the raphe is located on a raised and slightly asymmetrical sternum; the orientation of the outer central and proximal raphe ends of the raphe is the same [41] (pp. 146-420), (Figures 3M,P,Q, 4M,P and 5M-O). Both species, M. naumanii [41] and S. terrestris sp. nov. (Figures 3M,N, 4M,N and 5M), have prominent longitudinal depressions widening towards the central part of the valves; the central area in both species is longitudinally lanceolate [41], (Figures 3N and 4N). In addition, M. dexterii, some Sellaphora spp. (including S. terrestris sp. nov.), Okhapkinia alexanderi, and M. naumanii have one common structure: round apical pits at the apices of the valve ends [41] (Figures 3Q and 4P,R). Van de Vijver et al. [41] indicated the presence of very weakly expressed microcostae in the material of M. naumanii [41]; while in Zidarova et al. [42], microcostae were not present in the same species.

Molecular Discussion
Molecular genetic data on freshwater representatives of Microcostatus are not available. At present, the 18S rRNA gene sequences of two strains of Microcostatus (MMDL 54701, MMDL 54702) from the intertidal zone of China have been deposited in GenBank. Unfortunately, the work has not yet been published, and there is no way to assess the morphological features of these taxa. Furthermore, to date, Microcostatus has been known only as a freshwater taxon, raising questions as to the identity of the diatoms from which the Chinese sequence data have been obtained. Despite this, the results of calculating the genetic distance for the site at V4 (Table 2) showed a high degree of difference. Thus, the sequences of strains differ from each other by 6.9% and most likely represent different species. In relation to the species S. terrestris sp. nov., the differences were 4.6% and 8.2%, respectively, which clearly demonstrates their isolated position. However, to understand the relationship between Microcostatus and Sellaphora, a phylogenetic analysis is required involving an additional set of genes (for example, rbcL, cox1), since the 18S rRNA analysis did not provide the necessary resolution. For accurate identification, it is necessary to study the morphology of these strains. Both genera, as currently circumscribed, possess a tremendous amount of morphological variation. For example, Liu et al. [45] recognized four groups differing in their morphological attributes, and one of them (their group B) is mainly represented by taxa in molecular analyses. All this, in our opinion, indicates the need to revise both genera: Microcostatus and Sellaphora. Given that the monophyletic clade containing "S. seminulum" was distinct from other Sellaphora species in both analyses, it may be prudent to reinvestigate whether the genus Eolimna should be recognized as distinct.
Studies using molecular approaches indicate a high genetic diversity of Sellaphora. Using the example of S. pupula, a number of studies have shown groups of representatives with minor morphometric differences form separate well-maintained genetic lines [6,9,14]. The results of crossing some taxa confirmed the existence of reproductive barriers for morphologically similar populations [13,19]. In studies concerning the detailed study of the morphological and cryptic diversity of Sellaphora, many representatives with unclear differences in morphology and with pronounced genetic isolation were considered as phenodemes or demes [6,9,14]. A detailed revision of the morphological, molecular, and crossbreeding data was carried out using the example of four species of S. americana (Ehrenberg) D.G. Mann, S. bacillum, S. pupula, and S. laevissima sensu Krammer and Lange-Bertalot [24] from 40 sites in Britain and Northern Ireland by Mann et al. [8]. As a result, 54 demes were proposed, which can be considered the starting point for further research into the intraspecific and interpopulation relationships of Sellaphora.
Molecular data are most fully summarized in Mann and Poulíčková [12], where a phylogenetic analysis of 67 strains was carried out based on a concatenated five-gene dataset (partial nuclear 18S rDNA and 28S rDNA, plastid 23S rDNA and rbcL, and mitochondrial cox1). The authors noted that the tree topology supported the results obtained in previous works based on a two-gene dataset [9] (18S rDNA and rbcL); [20] (cox1 and rbcL)) except for S. cf. minima (clone BM42), whose position in early studies was inside the S. pupula-S. bacillum complex. In the latest phylogeny of S. minima (TCC 524) and S. cf. minima (BM42), these strains form a separate clade next to the clade including S. laevissima ( Figure 5) [12]. The results of the phylogenetic analysis by Mann and Poulíčková ( Figure 5) [12] indicated that most strains are grouped into three clades named 1, 2, and 3: clades 1 and 3 include the taxa S. pupula sensu lato, and clade 2 includes strains of S. bacillum and several demes of S. pupula, characterized by the presence of conopeum (continuous across the central area or interrupted accordingly), while representatives of S. laevissima, S. seminulum, and S. minima are grouped into separate external branches.
Our phylogenetic analysis for the 18S rDNA and rbcL genes supports the clades 1-3 identified in previous works [9,12]. However, the clade size is much smaller, due to the lack of 18S rDNA gene sequence data for many strains. The outer clades "minima", "seminulum", and "laevissima" are also clearly differentiated with sufficient support values (Figure 7). It should be noted that the identification of these external clades was also observed in previous works. Thus, the clade with S. minima (strains BM42, TCC524) had the highest support (1.0 PP) in Mann and Poulíčková [12], the clade with S. laevissima (indicated as clade 4 in Evans et al. [9]; Poulíčková et al. [21]), regardless of the genes used, had a stable isolated phylogenetic position: 1.0 PP for 18S rDNA and rbcL analyses [8], 85 LB for cox1 and rbcL analyses [20], 0.99 PP for five-gene dataset (18S rDNA, 28S rDNA, 23S rDNA, rbcL, cox1) [11]. In our analysis, this clade is supplemented by S. balashovae.
A high level of support for the clade with S. seminulum was first shown by Mann and Poulíčková [12], where three strains were included in the analysis. In previous works [9,21], only one strain (TM37) was included, presumably assigned to this species. Notably, in all analyses, this species always represented a separate line (branch) next to S. laevissima (clade 4).
We performed a more complete phylogenetic analysis for the rbcL region, since the number of sequences in GenBank for this gene for Sellaphora is much larger. Our phylogenetic tree includes 106 strains of Sellaphora, the taxonomic epithets listed in GenBank have been changed to the names of demes or species proposed in the relevant works (listed in Supplementary Data S1). The general topology of the tree is consistent with that of previous works [9,12,20,21]. To designate subclades, we have used the numbers proposed by Evans and co-authors [9] and Mann and Poulíčková [12]. Clade 1A includes six demes of S. auldreekie, clade 1B of S. bisexualis, and several demes of S. pupula. Clade 2 in our phylogeny is expanded, due to a separate subclade, which includes six demes of S. pupula. Clade 3 retains high statistical support, includes many S. pupula demes, and strains of S. blackfordensis D.G. Mann and S. Droop and S. capitata are grouped into subclades. The fourth clade is formed by strains of S. laevissima and S. balashovae. The new basal clade has maximum support, which includes two clades: S. seminulum and a clade consisting of strains of the species S. terrestris sp. nov. The clade with S. minima was supplemented by the strain Sellaphora sp. Ak1876, recently isolated at Lake Toro, Sibecya, Hokkaido, Japan [46]. In general, according to the molecular genetic data of Sellaphora terrestris sp. nov., it was most similar to the species S. seminulum, S. laevissima, and S. balashovae (Figures 6 and 7).
Thus, the morphological data and molecular genetic data indicate that our species belongs to the genus Sellaphora. The genus Microcostatus can currently be considered to require additional morphological and molecular genetic study of the species currently assigned to this genus.

Materials and Methods
Soil samples were collected from Cát Tiên National Park, Ðồng Nai Province, Vietnam, by E.S. Gusev and E.M. Kezlya in June 2019. Sampling occurred during an expedition of the Joint Russian-Vietnamese Tropical Research and Technological Centre (the "Ecolan 3.2" Project). The National Park is located 150 km northeast of Hồ Chí Minh City. The region belongs to the bioclimatic type of monsoon tropical climate with summer rains, relative humidity almost always exceeding 70%, and average annual temperature of about 26 • C. The main part of the territory is occupied by forests, which are of the monsoon, semi-deciduous type [47].

Sample Collection Procedure
The samples were collected as follows: the surface of the test site was examined in order to detect macrogrowth of algae, then a composite sample was taken from an area of 10-30 m 2 . The composite sample consisted of 5-10 individual samples. For an individual sample, the topsoil was removed from an area of 5 to 20 cm 2 with a metal scoop or shovel. After sampling, the instruments were cleaned and sterilized with ethanol. The samples were placed in labeled plastic zip bags and carried to the laboratory. The absolute humidity was measured immediately using the hot-drying method [48], and the samples were air-dried and packaged.

Soil Acidity Measurements
To measure pH, we mixed 30 g of soil with 150 mL of distilled water [49]. The suspension was poured into a clean glass beaker, and the measurements were performed with a Hanna Combo (HI 98129) device (Hanna Instruments, Inc., Woonsocket, RI, USA).

Culturing
Gathered materials were processed in the laboratory of molecular systematics of aquatic plants of the Institute of Plant Physiology of the Russian Academy of Sciences (IPP RAS). A sample of soil was thoroughly mixed and a small amount (15-20 g) was placed into a Petri dish (diameter 60 mm), then saturated with distilled water up to 60-80% of full moisture capacity and placed into an illuminated climate chamber. After a 10-day incubation period, the sample was diluted with a small amount of distilled water, mixed gently, and the suspension was transferred to a Petri dish for LM, using a Zeiss Axio Vert A1 inverted microscope. Algal cells were extracted with a micropipette, washed in 3-5 drops of sterile distilled water, and placed into a 300-µL well on a plate for enzymelinked immunoassay with Waris-H+Si [50]. Non-axenic unialgal cultures were maintained at 22-25 • C in a growth chamber with a 12:12 h light:dark photoperiod. After 20 days, strains were re-inoculated to a Petri dish with liquid Waris-H+Si medium. The strains investigated here were designated VP272, VP299, VP303, and VP304 (see Table 4). A list of all samples and strains examined in this study, with their GenBank accession numbers and geographic location of sampling sites, with measured ecological parameters is presented in Table 4. Terminology of the valve follows Kulikovskiy et al. [4], Mann et al. [8], Wetzel et al. [37], Van de Vijver et al. [41], and Stanek-Tarkowska et al. [43].

Preparation of Slides and Microscope Investigation
The monoclonal culture was boiled in 30% hydrogen peroxide at a temperature between 150-160 • C for 8 h, to dissolve organic matter. After decanting and refilling up to 50 mL with deionized water, the suspension was spread onto coverslips and left to dry at room temperature. Permanent diatom preparations were mounted in Naphrax ® (refraction index = 1.73). LM observations were performed with a Zeiss Axio Scope A1 microscope equipped with an oil immersion objective (×100, n.a. 1.4, differential interference contrast) and Axio Cam ERc 5s camera. Valve ultrastructure was examined in Papanin Institute for Biology of Inland Waters RAS, Borok, Russia, with a JSM-6510LV scanning electron microscope (JEOL Ltd., Tokyo, Japan), operated at 10 kV and an 11-mm distance. For scanning electron microscopy (SEM), parts of the suspensions were fixed on aluminum stubs after air-drying. The stubs were sputter coated with 50 nm of gold in an Eiko IB 3 (Eiko Engineering Co., Ltd, Ibaraki, Japan).

Molecular Investigation
The total DNA of the strains was obtained using ChelexTM 100 Chelating Resin, molecular biology grade (Bio-Rad Laboratories, Hercules, CA, USA), according to the manufacturer's protocol 2.2. Fragments of 18S rDNA gene (406-417 bp, including V4 domain), and partial rbcL plastid gene (951-957 bp) were amplified using the primers D512for and D978rev from Zimmerman et al. [51] for 18S rDNA fragments, and rbcL40+ from Ruck and Theriot [52] and rbcL1255 from Alverson et al. [53] for the rbcL gene. Amplifications of the 18S rDNA fragments and partial rbcL gene were carried out using the premade mix, ScreenMix (Evrogen, Moscow, Russia), for polymerase chain reaction (PCR).
The conditions of amplification for 18S rDNA fragments were as follows: an initial denaturation of 5 min at 95 • C, followed by 35 cycles at 94 • C for denaturation (30 s), 52 • C for annealing (30 s) and 72 • C for extension (50 s), and a final extension of 10 min at 72 • C. The conditions of amplification for partial rbcL were an initial denaturation of 5 min at 95 • C, followed by 45 cycles at 94 • C for denaturation (30 s), 59 • C for annealing (30 s) and 72 • C for extension (80 s), and a final extension of 10 min at 72 • C.
The resulting amplicons were visualized using horizontal agarose gel electrophoresis (1.5%), colored with SYBR Safe (Life Technologies, Carlsbad, CA, USA). Purification of DNA fragments was performed with an ExoSAP-IT kit (Affimetrix, Santa Clara, CA, USA), according to the manufacturer's protocol. 18S rDNA fragments and partial rbcL gene were decoded from two sides, using forward and reverse PCR primers and the Big Dye system (Applied Biosystems, Waltham, MA, USA), followed by electrophoresis using a Genetic Analyzer 3500 sequencer (Applied Biosystems, Waltham, MA, USA).
Editing and assembly of the consensus sequences were carried out by comparing the direct and reverse chromatograms using the Ridom TraceEdit program (ver. 1.1.0) and Mega7 [54].
With Mega7 software, the p-distance for 18S rDNA gene was determined and used to calculate the sequence similarity with the formula (1-p) ×100. Newly determined sequences and 18S rDNA gene and rbcL fragments from 39 other diatoms, which were downloaded from GenBank (taxa and Accession Numbers are given in the Supplementary Data S1), were used to generate two-gene trees (Supplementary Data S2). Since the 18S rDNA gene sequences are not available for the most Sellaphora strains, the rbcL reads were included in the alignments, along with corresponding sequences of 110 diatoms downloaded from GenBank (Supplementary Data S3). Three centric diatom species were chosen as the outgroups. The nucleotide sequences of the 18S rDNA and rbcL genes were aligned separately using Mafft v7 software and the E-INS-i model [55]. For the proteincoding sequences of the rbcL gene, we checked that the beginning of the aligned matrix corresponded to the first position of the codon (triplet). The resulting alignments had lengths of 404 (18S rDNA) and 957 (rbcL) characters. Two 18S rDNA gene sequences of the genus Microcostatus were used for the p-distance calculation.
The dataset was analyzed using the Bayesian interference (BI) method implemented in Beast ver.1.10.1 [56] for phylogeny reconstruction. For each of the alignment partitions, the most appropriate substitution model was estimated using the Bayesian information criterion (BIC), as implemented in jModelTest 2.1. 10 [57]. This BIC-based model selection procedure selected the following models, shape parameter α, and proportion of invariable sites (pinvar) for the concatenated tree: TIM3ef+I+G, α = 0.3800, and pinvar = 0.5940 for 18S rDNA gene; HKY+I, pinvar = 0.8430 for the first codon position of the rbcL gene; TVMef+I, pinvar = 0.9000 for the second codon position of the rbcL gene; and TPM3uf+G, α = 0.4070 for the third codon position of the rbcL gene. We used the HKY model of nucleotide substitution instead of TIM3ef and TPM3uf, and the GTR model instead of TVMef, since they were the best matching models available for BI. The following models, shape parameter α, and proportion of invariable sites were for the rbcL tree: HKY+I+G, α = 0.3370, and pinvar = 0.6700 for the first codon position of the rbcL gene; TrNef+I+G, α = 0.5340, and pinvar = 0.8280 for the second codon position of the rbcL gene; and TVM+I+G, α = 0.6170, and pinvar = 0.2080 for the third codon position of the rbcL gene. We used the HKY model of nucleotide substitution instead of TrNef, and the GTR model instead of TVM, for the individual rbcL tree.
A Yule process tree prior was used as a speciation model. The analysis ran for 5 million generations, with chain sampling every 100 generations. The parameters estimated convergence, effective sample size (ESS), and burn-in period were checked using the software Tracer ver. 1.7.1. [56]. The initial 25% of the trees were removed, and the rest were retqined to reconstruct a final phylogeny. The phylogenetic tree and posterior probabilities of its branching were obtained on the basis of the remaining trees, having stable estimates of the parameter models of nucleotide substitutions and likelihood. Maximum likelihood (ML) analysis was performed using the program RA×ML [58]. The nonparametric bootstrap analysis with 1000 replicates was used. The phylogenetic tree topologies are available online: Supplementary Data S4 for the concatenated rbcL and 18S rDNA genes tree and Supplementary Data S5 for the rbcL gene tree. The statistical support values were visualized in FigTree (ver. 1.4.2) and Adobe Photoshop CC (19.0).

Conclusions
We studied in detail the morphology of four soil strains of Sellaphora from Cát Tiên National Park and identified a number of differences from all currently known species. During the study of cultured and natural materials, variability of morphometric characteristics (length and width of valves, density of striae) was revealed, while molecular studies on rbcL and 18S rDNA genes including the V4 domain showed that the strains studied by us formed a separate branch, with high statistical support. Based on these data, we described the new for science species Sellaphora terrestris sp. nov. In addition, we noticed that some species assigned to Microcostatus J.R. Johansen and J.C. Sray have a very similar morphology to our new species. The results of the study of the morphological characteristics of a number of species of Sellaphora and Microcostatus indicated that many characteristics (robustness of microcostae, striae morphology, and presence of conopeum and hymenes) can be expressed to varying degrees. We cannot exclude the possibility that some species assigned to the genus Microcostatus do not belong there, and may be better placed within Sellaphora. Molecular genetic data for Microcostatus are limited and require further study. Funding: Isolation, cultivation and molecular analysis were obtained with financial support from the Russian Science Foundation (project number 19-14-00320Π). Light and scanning electronic microscopies, and finishing the manuscript were obtained within the state assignment of the Ministry of Science and Higher Education of the Russian Federation (theme no. 122042700045-3).