Cyanomargarita gen. nov. (Nostocales, Cyanobacteria): convergent evolution resulting in a cryptic genus

Two populations of Rivularia‐like cyanobacteria were isolated from ecologically distinct and biogeographically distant sites. One population was from an unpolluted stream in the Kola Peninsula of Russia, whereas the other was from a wet wall in the Grand Staircase‐Escalante National Monument, a desert park‐land in Utah. Though both were virtually indistinguishable from Rivularia in field and cultured material, they were both phylogenetically distant from Rivularia and the Rivulariaceae based on both 16S rRNA and rbcLX phylogenies. We here name the new cryptic genus Cyanomargarita gen. nov., with type species C. melechinii sp. nov., and additional species C. calcarea sp. nov. We also name a new family for these taxa, the Cyanomargaritaceae.

With the advent of molecular methods, many phycologists, including those who study cyanobacteria, began to recognize the existence of cryptic species , Casamatta et al. 2003, Erwin and Thacker 2008, Joyner et al. 2008, Premanandh et al. 2009, Reñ e et al. 2013, M€ uhlsteinov a et al. 2014a,b, Patzelt et al. 2014). However, while in these papers the existence of cryptic species was suggested, the species were not recognized taxonomically. Subsequently, cryptic species have been named in several algal groups, including euglenids (Marin et al. 2003, Kosmala et al. 2007, Kosmala et al. 2009, 2013, Linton et al. 2010, eustigmatophytes Fawley 2007, Fawley et al. 2015), chlorophytes (Fawley et al. 2005, Fu c ıkov a et al. 2014, and cyanobacteria (Osorio-Santos et al. 2014. Some cyanobacterial systematists have suggested the existence of cryptic genera as well (Kom arek et al. 2014, Dvo r ak et al. 2015a, although very few cryptic genera have actually been described. Pinocchia, which is morphologically identical to Pseudanabaena, has a phylogenetic position distant from that genus, and therefore was described as a cryptic genus (Dvo r ak et al. 2015b). Kovacikia, which is morphologically similar to Phormidesmis but molecularly distinct, would also fit the definition of a cryptic genus, although the authors did not label it as such (Miscoe et al. 2016). There are also a number of pseudocryptic genera (genera defined by morphological traits that are minor or phenotypically plastic, and therefore not always expressed in the population) as well, such as Nodosilinea, Oculatella, Limnolyngbya, Pantanalinema, and Alkalinema. These genera belong to the Synechococcales, an order containing taxa with few morphological characteristics (simple filamentous forms, variations in trichome width and sheath characteristics).
Outside of the Synechococcales, few cryptic genera have been recognized. In the Chroococcales, Chalicogloea is similar to Gloeocapsa and could be considered a cryptic genus (Rold an et al. 2013). There are likely more cryptic genera in this order, but identifying them is more problematic because members of the genus are difficult to grow in culture, and consequently, fewer sequences are available. In the Oscillatoriales, Ammassolinea is the one of the described cryptic genera (Ha sler et al. 2014), being morphologically inseparable from Phormidium, as it is presently defined, as well as Moorea, Okeania and Microseira (Engene et al. 2012, 2013, McGregor and Sendall 2015. Within the Nostocales, there is much greater morphological complexity than the nonheterocytous orders. Some pseudocryptic genera have been described, including Mojavia ( Reh akov a et al. 2007), Dapisostemon (Hentschke et al. 2016), and Pelatocladus (Miscoe et al. 2016).
We recently discovered a population of tapering, heterocyte-bearing trichomes embedded in a hemispherical to spherical mucilage investment in a small, spring-fed, unpolluted stream near the town of Apatity in the Kola Peninsula, Russia. It was completely consistent with the description of Rivularia C.Agardh ex Bornet & Flahault, the type genus of Rivulariaceae, which contains tapering, heterocytous taxa. This taxon fit no established species in Rivularia, and upon sequencing was determined to be phylogenetically distant from all members of that family. A second species belonging to the same clade as the Russian material was found, and sequenced several years earlier from a wet wall in the Grand Staircase-Escalante National Monument in Utah, USA. These two populations differ morphologically and ecologically, and are described herein as two new species based on a modified phylogenetic species concept (Mishler andTheriot 2000, Johansen andCasamatta 2005) in a newly proposed genus, Cyanomargarita gen. nov. This genus cannot be placed in any family-level grouping of taxa based on the phylogenetic analyses performedwe place these taxa in a new family to science, the Cyanomargaritaceae fam. nov.

MATERIALS AND METHODS
Isolation and strain characterization. Both strains of Cyanomargarita were isolated from natural populations into unialgal cultures using standard microbiological methods, including enrichment plates and direct isolation from the original samples, in Z8 medium (Kotai 1972, Carmichael 1986). Cultures were observed under a Zeiss Axioskop photomicroscope with both bright field and DIC optics. All morphological measurements were obtained using AxioVision 4.8 software provided by Zeiss (Oberkochen, Germany). Living cultures were deposited into the Cyanobacterial Culture Collection at John Carroll University, Cleveland, OH, USA. Natural populations of material from which the strain C. melechinii APA-RS9 was derived were dried and deposited as an isotype in the Herbarium of the Polar-Alpine Botanical Garden-Institute, Kola Science Centre, RAS, Kirovsk-6, Murmansk Region, Russia, and information about habitat, coordinates and locality can be found in the online database Cyanopro (Melechin et al. 2013). The dried holotype material of this strain was deposited in the Herbarium for Nonvascular Cryptogams in the Monte L. Bean Museum, Provo, UT, USA. Liquid materials of both species fixed in 4% formaldehyde, as well as dried materials of C. calcarea, were also deposited in the Herbarium for Nonvascular Cryptogams in the Monte L. Bean Museum, Provo, UT, USA.
Molecular methods. Genomic DNA was extracted following techniques described in Pietrasiak et al. (2014). PCR amplification of the 16S rRNA gene was accomplished following Osorio-Santos et al. (2014), with the exception that forward primer 8F was used instead of forward primer VRF2, for amplification of a longer sequence, starting near the beginning of the 16S rRNA gene (Perkerson et al. 2011). The 16S rRNA amplicons were cloned to avoid problems in sequencing of multiple operons (Siegesmund et al. 2008). PCR amplicons and sequences of rbcLX and rpoC1 genes were obtained exactly according to Rudi et al. (1998) and Seo and Yokota (2003), respectively. The nifD gene amplification was completed using a protocol described in Roeselers et al. (2007). All three protein-encoding genes (rbcLX, rpoC1, nifD) were sequenced after purification procedure using Wizard â SV Gel and PCR Clean-Up System (Promega, Madison, WI, USA); rather than cloned, because they are single-copy genes CYANOMARGARITA GEN. NOV. in the cyanobacterial genome. All sequences obtained in this study were deposited in the NCBI Nucleotide database, under accession numbersfor 16S rRNA gene: KY296602, KY296603, KY296604, KY296605, KY296606, KY296607, KY296608; rbcLX: KY296611, KY296612; nifD: KY296609, KY296610; rpoC1: KY296613, KY296614.
Phylogenetic analyses. All sequences chosen for alignment and phylogenetic analyses were obtained from our internal set of sequences and relevant sequences (chosen based on both BLAST searches and named taxon searches) from the NCBI Nucleotide database before September 1, 2016. Ribosomal (16S) sequences were aligned using MUSCLE in Mega v. 6.06 (Tamura et al. 2013), and checked manually in Microsoft Word (Microsoft Corp., Redmond, WA, USA) to ensure that alignments supported preservation of secondary structure (Luke sov a et al. 2009, Reh akov a et al. 2014. Protein coding genes were aligned as codons using MUSCLE in Mega v. 6.06 (Tamura et al. 2013).
The public software jModeltest2 (Darriba et al. 2012) was used to determine the optimal Maximum Likelihood (ML) model, which was GTR+I+G for 16S rRNA genes, and both ML and Bayesian Analysis (BA) were subsequently run using this type of model. The exact parameters of the substitution models were individually estimated from the data during analysis by MrBayes v. 3.2.6 (Ronquist et al. 2012) andRaxML v. 7.2.8 (Stamatakis et al. 2008) software. The jModeltest2, ML, and BA calculations were all run on CIPRES (Miller et al. 2012). In the BA, two runs of eight Markov chains were applied with 10 million generations, sampling every 100 generations, with 25% burn-in. The sump command in Mr. Bayes was used. The ML analysis was conducted in RAxML v.7.2.8 with 1,000 bootstrap replicates, using GTR+I+G substitution model. Parameters for the models, with rate matrix, base frequencies, and invariant and gamma settings are given in Table S1 in the Supporting Information. The Maximum-parsimony (MP) analysis was performed using PAUP v. 4.02b (Swofford 2002) with steepest descent, the tree bisection and reconnection branch swapping, and 1,000 bootstrap replicates.
Phylogenies utilizing 16S rRNA gene sequences can yield ambiguous or unsupported trees, and in such cases a multiple loci approach is recommended (da Silva Malone et al. 2015, Song et al. 2015. We treated the rbcLX alignment as codons in the BA calculation (Fawley et al. 2015), using the Ny98 equal rate ratio model available in MrBayes for analysis of proteincoding genes (Ronquist et al. 2012). In the ML analysis the GTR+I+G model chosen by jModelTest2 was applied. Parameters for the models are given in Table S1. We conducted BA calculation of the rbcLX alignment with two runs of eight Markov chains with 20 million generations, sampling every 100 generations, with 25% burn-in. Additionally, MP and ML analyses were performed for rbcLX gene using the same settings as in the 16S rRNA phylogeny. The tree topology from BA was chosen for the resulting visualization for both 16S rRNA gene and rbcLX gene phylogenies, as well as for the mapping of support values from MP and ML analyses.
For 16S rRNA gene the BA had an estimated sample size (ESS) exceeding 2,000 for most of the parameters (790-6,877, average = 2,621 from 13 parameters), well above the average of 200 typically accepted as sufficient by phylogeneticists (Drummond et al. 2006). The final average standard deviation of split frequencies was <0.016. The potential scale reduction factor (PSRF) value for all the estimated parameters in the BA was 1.000, indicating that convergence of the MCMC chains was statistically achieved (Gelman and Rubin 1992).
For rbcLX gene the BA had an ESS exceeding 2,000 for some of the parameters (79-2,193, average = 345 from 67 parameters), well above the average of 200 typically accepted as sufficient by phylogeneticists (Drummond et al. 2006).
The final average standard deviation of split frequencies was <0.010. The PSRF value for all the estimated parameters in the BA was 1.000, indicating that convergence of the MCMC chains was statistically achieved (Gelman and Rubin 1992). P-distance values for all sequences were calculated in PAUP v. 4.02b (Swofford 2002). Graphical representations of the ITS structures were created in Adobe Illustrator CS5.1 (Adobe Systems Inc., San Jose, CA, USA) based upon secondary structure configurations given by Mfold (Zuker 2003).
Line drawings. Drawings were made by the senior author using stippling technique, completed digitally with Wacom Cintiq 24HD Pen Display (Wacom Technology Corporation, Portland, OR, USA) utilizing the original photos as templates.

RESULTS
Phylogenetic analyses. The 16S rRNA gene phylogeny has posterior probability support on all nodes in the backbone of the BA, with the exception of four nodes at the base of the tree marked with small light gray circles ( Fig. 1). Overall topology of the tree is consistent with recent studies of Nostocales (Berrendero et al. 2011, Hauer et al. 2014 Cyanomargarita gen. nov. forms a cluster of two terminal OTUs (Operational Taxonomic Units) corresponding to two new species: C. melechinii and C. calcarea, with high support (Fig 1). Cyanomargarita is sister to a large clade, containing the well-defined monophyletic Gloeotrichiaceae, as well as the Fortieaceae, Aphanizomenonaceae, Nostocaceae, and Tolypothrichaceae. Additionally, Cyanomargarita is also related to the "Scytonema cf. crispum" group, which requires revision (i.e., it is not Scytonema C.Agardh ex E.Bornet & C.Flahault) and has an uncertain taxonomic position (incertae familiae), falling outside of the Scytonemataceae clade defined by the inclusion of the type species, Scytonema hofmannii C.Agardh ex Bornet & Flahault (the basal clade in Fig. 1). Cyanomargarita is found outside of the Rivulariaceae sensu stricto, despite the similar morphology between Cyanomargarita and Rivularia.
The BA in this study and the MP and ML analyses had highly similar topology, although the latter two analyses had poorer support. The MP analysis differed slightly in the placement of Roholtiella Bohunick a, Pietrasiak & J.R.Johansen and Gloeotrichia J.Agardh ex Bornet & Flahault, but these taxa were still associated with the Nostocaceae and Aphanizomenonaceae. In the parsimony analysis, Macrochaete Berrendero Gomez, J.R.Johansen & Ka stovsk y moved from a position proximate to Calothrix C.Agardh ex Bornet & Flahault to a position sister to Cyanomargarita, but Cyanomargarita remained sister to the same polyfamily clade as in the BA (Fig. 1). The ML analysis generally had the poorest node support, but had the same topology as the BA except for a slight change in the position of "S. cf. crispum." In all three analyses, Cyanomargarita was always a supported clade outside of the Rivulariaceae and outside of any other family-level clade. Another piece of evidence supporting the placement of Cyanomargarita outside the Rivulariaceae is that representatives of this new genus have ITS regions with only one tRNA gene (tRNA Ile ) across four different ribosomal operons (Table 1), which is different from both the Nostocaceae (with two or no tRNAs; see Boyer et al. 2001, Reh akov a et al. 2007) and the Rivulariaceae (with two tRNAs only). We conclude that, based on current phylogenetic evidence, Cyanomargarita requires its own family-level rank, and propose the family Cyanomargaritaceae fam. nov.
Cyanomargarita has low similarity in 16S rRNA gene sequence with most other Nostocalaean taxa ( Table 2). The highest similarity was with Gloeotrichia pisum Thuret ex Bornet & Flahault from an alkaline wetland in Ohio, USA (95.4%). However, our new taxon differs from members of Gloeotrichia based on the absence of paraheterocytic akinetes with well-developed exospore. Moreover, 16S rRNA gene similarity between our taxon and a Rivularia strain from Argentina is only~92.3%. Historically, less than 95% similarity among 16S rRNA gene sequences was considered good evidence for separation of prokaryotic genera (Stackebrandt and Goebel 1994), but within the heterocytous genera the cutoff is likely much higher , Patzelt et al. 2014, Berrendero G omez et al. 2016.
Cyanomargarita is also outside of Rivulariaceae sensu stricto, according to our rbcLX phylogenetic analysis (Fig. 2). According to this analysis, it is most closely related to the Tolypothrichaceae (containing the type species Tolypothrix distorta K€ utzing ex Bornet & Flahault) and diverse strains of Calothrix. In contrast, Rivularia forms a well-supported clade with Kyrtuthrix Ercegovic, distant from Cyanomargarita (Le on-Tejera et al. 2016). The rbcLX phylogeny, with posterior probability support in the BA showing separated clades of Rivularia and Cyanomargariata, is consistent with our conclusion based on the 16S rRNA gene phylogeny that Cyanomargarita is not congeneric with Rivularia, and, furthermore, is not even in the Rivulariaceae. The MP and ML phylogenies had much poorer support but similar topologies. Cyanomargarita was sister to the Tolypothrichaceae in MP and in a polytomy with Tolypothrichaceae and Calothrix/ Macrochaete as in the BA in the ML analysis (data not shown).
ITS analysis. The 16S-23S ITS sequences of Cyanomargarita calcarea are~50 nucleotides longer than the ITS sequences of C. melechinii, likely as a result of insertions flanking the tRNA Ile gene on the 3ʹ side of the gene (Table 1). In general, secondary structures of D1-D1ʹ, V3, and Box B helices show similar structures across both species with minor base substitutions in all three domains. Below, we compare the secondary structures of conserved ITS domains for homologous operons in the two species (e.g., operon 1, we recovered two additional  operons in Cyanomargarita. melechinii that were not obtained from C. calcarea). The configuration of D1-D1ʹ helices for both species share features seen in most members of the Nostocales: a small terminal loop; a sub-terminal bilateral bulge; and a basal unilateral bulge on the 3ʹ side of the helix, with a highly conserved basal clamp of five base pairs (GACCU-AGGUC). We detected four substitutions across the two species in the upper part of the D1-D1ʹ helix, with three of those located within the loop regions. The last substitution on D1-D1ʹ helix occurs within the basal 3ʹ unilateral bulge, a transition mutation from G to A (Fig. 3). The V3 helix was very similar in both species, but with some minor differences such as two substitutions in the apical loop and a compensatory change in a single base pair in the middle part of the stem (indicated by arrows). The V3 helix from operon 3 of C. melechinii has a short insertion (UAAU) within the terminus of the helix (Fig. 3). The Box B structure appears to be variable and informative, with a notably different terminal loop in operon 1 of each species. The Box B of operon 1 in C. calcarea is actually more similar to the Box B of operon 2 of C. melechinii. We do not know if this is a convergent mutation or gene conversion. In our particular case, differences of ITS structures across different operons inside one lineage can be more significant than differences detected between homologous operons of different species. The overall differences between the ITS sequences from homologous operons of the two species exceeds the differences used in the past to justify species separation (Osorio-Santos et al. 2014, Miscoe et al. 2016). Based on morphology, ecology, distribution, 16S rRNA gene phylogeny, p-distance analyses of 16S rRNA gene, rbcLX phylogeny, analysis of the secondary structure of the 16S-23S ITS region, and p-distance analysis of the 16S-23S ITS region, we conclude that the two strains of the Cyanomargarita clade appear to be evolutionarily independent lineages distant from representatives of Rivulariaceae, with the genus Cyanomargarita gen. nov. belonging to a monogeneric family, Cyanomargaritaceae. These taxa are described here.
Diagnosis: Morphologically similar to the members of the Rivulariaceae, but phylogenetically distinct from that family. Phylogenetically closely related to clade containing Gloeotrichiaceae, with which it bears morphological similarity, but separated from that family by phylogeny and the absence of FIG. 3. Secondary structures of the 16S-23S ITS region from both species. OP stands for different operons, with three operons recovered from Cyanomargarita melechinii and one operon (with two clones with identical sequence and structure) from Cyanomargarita calcarea. Arrows on C. calcarea structures indicate base changes from the homologous operon 1 for C. melechinii.

768
SERGEI SHALYGIN ET AL. paraheterocytic, elongated akinetes. Also related to the "Scytonema cf. crispum" clade, which is phylogenetically distant from Scytonema sensu stricto, but differing from that group by tapering, copious mucilage formation, and hemispherical to spherical colony formation.
Etymology: named for the single genus in the family, Cyanomargarita. Description: Macroscopic colonies in nature hemispherical to spherical to irregularly globular, with tapering trichomes embedded in the colonial mucilage but extending outside of the mucilage to impart a fuzzy appearance to the colony. Filaments with distinct lamellated sheath, which is often funnel-or collar-like at the distal ends. Trichomes typically largest at the base and tapering to a thin hair distally, arranged in parallel, singly, or doubly false branched, sometimes forming concentric layers in large colonies. Heterocytes basal or rarely intercalary. Akinetes absent, but large swollen arthrospores present in some species.
Etymology: named for the pearl-like appearance of blue-green colonies growing on mosses; cyaneus (L) = greenish-blue; margarita (L) = pearl.
Diagnosis: Akin to M. calcarea, but differing in possession of broad, colorless to slightly blackish sheaths and shorter hairs, with shorter spacer regions flanking the tRNA Ile region in the 16S-23S ITS, with percent identity between ITS sequences of both species >90.00%.
Description: Natural Populations (Figs. 4 and 5) -Macroscopic colonies slimy, spherical, or hemispherical, with appearance of small blue-green pearls attached to mosses, less commonly irregularly shaped, grayish blue-green to blue-green, attached to the substrate (in type locality on the submersed moss Fontinalis sp. and on stones), growing up to 5 mm in diameter. Filaments more or less radially arranged, sometimes arranged in concentric layers in the colony, attenuated towards the ends, densely arranged in parallel orientation, abundantly single false branched, with young, short filaments having geminate branching, 12.5-18 (21) lm wide near base, rarely with basal parts onion-like swollen. Sheaths thin to thick, 1-8 lm wide, often strongly lamellated with 3-5 distinct layers, colorless to slightly blackish in old filaments, funnel-like widened at the distal ends and near site of branching, rarely firm, compacted to give wavy or transverse striations. Trichomes usually gradually widened at the base, rarely onion-like swollen, sometimes narrowing towards the base, gradually tapering towards the distal ends, unconstricted, slightly constricted to distinctly constricted at the cross walls, typically constricted in the basal part, becoming unconstricted in the middle of long, mature trichomes, 7.5-12.5 lm wide near the base, distally elongated into long, thin hairs, as narrow as 1 lm. Cells usually granulated, rarely with large, spherical clear vesicular spaces devoid of thylakoids, bright blue-green to blue-green, when actively dividing as short as 2 lm long, near the base shorter than wide to isodiametric, usually longer than wide in the middle of long mature trichomes, up to 10 lm long, towards the ends less intensely pigmented or colorless, 8-20 (27) lm long. Heterocytes often solitary, rarely in pairs or up to three in a row, spherical, hemispherical, slightly conical, oval, or cylindrical, elongated, flattened, within, or outside of sheath, olive-brown in color, usually with an enlarged, single polar nodule, 10-15 (16) lm wide, 9-18 (20) lm long. Necridia and intercalary involution cells present.
Cultures (Fig. S1 in the Supporting Information) -Macroscopic colonies dark-green to blue-green, spreading far from the center, with several filaments upright from agar. Filaments very entangled, long, in liquid Z8 medium forming huge, abundant nodules (20-60 lm wide), on the solid medium, frequently having single-, and double-false branching as well as geminate loops prior to branch formation, when young forming stages similar to Tapinothrix clintonii Bohunick a et Johansen with one isopolar filament tapered at both ends fragmenting to produce two heteropolar filaments with widened base and tapered ends, rarely on nitrogen-free medium arranged in parallel like representatives of Coleodesmium Borzi ex Geitler, (8.1) 10-16 lm wide. Sheaths are always colorless, slightly lamellated, with 2-4 layers, usually straight, 1-6 lm wide. Trichomes in young stages taper, at the basal part always clearly constricted, rarely forming long unconstricted hairs, 1-2 lm wide, in mature stages also distinctly constricted, often slightly tapering or untapered but forming conical apical cells, usually long and entangled, releasing small tapered hormogonia, or with pairs of cells with zigzag arrangement at the middle of the trichomes, also forming abruptly conical apical cells on nitrogen-free medium, 3-10 lm wide. Cells often granulated, bright blue-green to olivegreen, when actively dividing short, 2 lm long, in middle of long trichomes, 5-10 lm long, in the hair 3-15 (17) lm long, in nitrogen-free medium dividing parallel to filament axis to form a pair of cells (proheterocytes?) at the basal end of the trichome. Heterocytes forming only in nitrogen-free medium, Single filaments with false branching, firm sheath, and constrictions at crosswalls. (F) Variably shaped heterocytes. Numerals indicate diagnostic characteristics used in species description: 1, Filament without constrictions; 2, sheath with wavy striations; 3, funnel-like widened sheaths; 4, two heterocytes in the row; 5, intercalary involution cells; 6, juvenile single trichome without individual sheath; 7, geminate branching on juvenile single trichome; 8, two necridia in a row; 9, different shaped heterocytes; 10, thin apical hairs. basal, slightly brownish or colorless, of different shapes, from oval or spherical to hemispherical, flattened or irregular, often solitary, rarely two in a row or two side by side, within or outside of sheath, 5-7 lm wide, 4-6 lm long. Necridia, intercalary involution cells, and dark-olive resting cells present.
Etymology: Named in honor of Alexey Melechin, the lichenologist who originally found Cyanomargarita in its type locality and informed the author of its existence.
Reference Strain: Cyanomargarita melechinii APA-RS9, deposited in the Cyanobacterial Culture Collection at John Carroll University.
Notes: According to morphology, most similar to the poorly known taxon, Rivularia compacta Collins, described from Northern America, from which it differs by larger size of the filaments and trichomes, as well as geminate branching and character of the sheath (Kom arek 2013).
Diagnosis: Akin to C. melechinii, but differing by possession of brownish sheaths closely attached to the trichomes, with longer hairs, with arthrospores, and with longer spacer regions flanking the tRNA Ile region in the 16S-23S ITS, with percent identity between ITS sequences of both species >90.00%.
Description: Cultures (Figs. 6, S2 in the Supporting Information) -Macroscopic colonies dark-green to olive-green when old, radiating far from the colony center, with several filaments erect from the agar, in liquid medium forming hemispherical colonies with parallel and radial arranged filaments. Filaments relatively long, entangled, sometimes irregularly coiled or screw-like coiled, frequently with single-, and double-false branching as well as with geminate loops prior to branch formation, gradually tapering from the base, 7-12 (16) lm wide, rarely with basal parts of filaments onion-like swollen. Sheath in the juvenile stages usually colorless, soft, thin, always attached to trichomes, maximally with two layers, 2 lm wide; in senescent cultures brown to slightly reddish, firm, covering only basal parts of trichomes, up to 5 lm wide, sometimes forming collars. Trichomes gradually attenuated, constricted at the cross walls when young, unconstricted when mature, 6-10 lm wide, tapering to a colorless hair many cells long, (2) 2.5-3 lm wide. Cells granulated, usually barrel-shaped or distinctly constricted, apical cells sometimes widened in comparison to the adjacent subterminal cells but abruptly narrowing to a conical end, blue-green, bright blue-green to dark olive-green, longer than wide, isodiametric, or shorter than wide, longer than wide towards the ends, 2-3.5 lm wide, 9-16 lm long. Heterocytes basal or intercalary, two or three in a row, flattened, quadratic, or elongated oval, with shape spherical, hemispherical, conical, or irregular, rarely with two heterocytes side by side, within or outside sheath, bright brown to olive in color, 6-12 lm wide, 9-12 lm long. Arthrospores variable in shape, spherical to barrel-shaped, also irregular and rhomboid, typically distinctly granulated, with thin walls, bluegreen, 7-10 lm wide, 7-12 (17) lm long. Necridia present.
Reference Strain: Cyanomargarita calcarea GSE-NOS12-04C, deposited in the Cyanobacterial Culture Collection at John Carroll University.
Notes: Natural material could not be obtained as populations were microscopic embedded in a mixture of other algae. DISCUSSION Originally, tapering cyanobacteria capable of producing heterocytes were placed either in the Rivulariaceae (Rivularia, Isactis Thuret ex Bornet et Flahault, Brachytrichia Bornet et Flahault and Gloeotrichia) Bornet et Flahault;Bornet andFlahault 1886-1888). In the early part of the 20th century, these taxa, as well as other tapering taxa, including non-heterocytous forms, such as Leptochaete Borzi ex Bornet et Flahault and Tapinothrix Sauvageau, were all placed in a single family, Rivulariaceae (Fr emy 1929, Geitler 1932. The nonheterocytous forms were removed from the family in the revision of the Nostocales completed by Kom arek and Anagnostidis (1989)this system continued in both Kom arek (2013)  (2014). Morphologically, these taxa are well-defined, although the colonial morphology and production of hairs is typically lost in culture. The type species for Calothrix, C. confervicola C.Agardh ex Bornet & Flahault, has not yet been sequenced, and is marine in origin. The accepted type species for Rivularia, R. dura Roth ex Bornet & Flahualt, has also not been sequenced, and is freshwater in origin.
Confusion regarding the diagnosis of Calothrix from Rivularia clearly exists in the modern literature. In Bergey's Manual of Systematic Bacteriology (Second Edition), the reference strains for Calothrix are all freshwater in origin (Rippka et al. 2001a), whereas the three reference strains for Rivularia are all from saline habitats (Rippka et al. 2001b). This ecological niche is the opposite of what one would expect based on the type ecology of the species. Subsequent to Rippka et al.'s (2001a,b) work, more sequences in the tapering group were found (Sihvonen et al. 2007), yielding a phylogeny with five groups: (i) Rivularia, mostly from marine habitats, including the Bergey's Manual reference strain Rivularia PCC 7716 (Rippka et al. 2001b), (ii) Calothrix marine clade I, (iii) Calothrix marine clade II, (iv) Calothrix freshwater clade, and (v) Gloeotrichia clade. Berrendero et al. (2008) confirmed this result (although Gloeotrichia was not in their phylogeny), but showed that all three marine clades had at least some strains assigned to Calothrix and some strains assigned to Rivularia. In subsequent papers (Berrendero G omez et al. 2016, Le on-Tejera et al. 2016), the five clades noted by Sihvonen et al. (2007) persisted in the phylogenetic analyses based on larger taxon sets. Our 16S rRNA phylogeny has the most taxa, and these five clades persist in our phylogeny as well (Figs. 1, S3 in the Supporting Information).
Although some confusion persists in the names assigned to strains in culture collections, the identity of these five clades is fairly stable. We suspect that the type for Calothrix, when it is isolated and sequenced, will fall within either marine Calothrix Clade I or Clade II; Rivularia dura, when sequenced, will fall in the Rivularia clade defined in Berrendero G omez et al. (2016) and Le on-Tejera et al. (2016). Gloeotrichia has already been moved to another family, the Gloeotrichiaceae (Kom arek et al. 2014). We anticipate that Calothrix-like taxa (Freshwater, Marine I, Marine II) likely will be revised and separated into three genera and placed in their own families, separate from the Rivulariaceae (Fig. 1). Based on either morphology or phylogeny, Cyanomargarita does not fall into any previously described families, and will be placed in the Cyanomargaritaceae.
Much of the confusion in cyanobacterial taxonomy today is the result of the assumptions by earlier authors that a number of morphological features evolved within the phylum only once, or, at best, only a few times. Tapering trichomes inhabiting soft mucilage to form adherent colonies, false branching, and true branching were all characteristics that were thought to be significant and sufficient to group taxa into relatively few higher level taxa. We now know that these derived characters have arisen multiple times through the process of convergent evolution. Tapering (Gugger and Hoffmann 2004, Wilde et al. 2014, Mare s et al. 2015. Indeed, in the Cyanomargaritaceae, cell division in two planes is present in both species, and this is a prerequisite character to true-branching, although at present we have only seen the phenomenon in the basal cells of the trichomes in culture material.
Polyphyly in cyanobacterial genera should not be a surprise. Given that relatively few characters were given inordinate weight by early taxonomists, thinking that these characters could arise independently did not seem parsimonious or likely. However, with a molecular understanding, we realize that many supposed synapomorphies in cyanobacteria are actually not homologous characters. It seems apparent that they are useful in the definition of genera, where they appear to be consistent across the entire group, but they fail in the definition of higher-level taxa. The exception appears to be the formation of heterocytes and akinetes, which are restricted to the Nostocales and therefore likely arose only once.
Given the convergence of morphological traits in evolutionarily distant lineages, the use of molecular sequence data to define family-and order-level taxa is likely going to increase. The morphological definition of families will likely be replaced by a phylogenetic definition (a monophyletic cluster of 774 genera). This is already happening in other algal groups, such as the Sphaeropleales (Fu c ıkov a et al. 2014). We anticipate that as more molecular sequence data become available for more genera, the difficulty in using existing family-level taxonomy will increase in many algal groups, including cyanobacteria, and more families will be described and recognized in order to maintain monophyly and to stabilize taxonomy. These families will, unfortunately, often be difficult to characterize morphologically, and so will lose their meaning and value to the taxonomic novice. However, a taxonomic system consistent with evolutionary history has long been the goal of taxonomists.
We thank John Carroll University for salary support for the first author, supplies and use of laboratory facilities, and supportive faculty and staff in the Biology Department. Additionally, we are grateful to all lab mates from the Johansen laboratory for useful discussion of the work. This work was completed with support from the projects 15-11912S of the Czech Science Foundation and with the long-term research development project of the Institute of Botany CAS, no. RVO67985939.

Supporting Information
Additional Supporting Information may be found in the online version of this article at the publisher's web site:   Uncollapsed phylogeny for Cyanomargarita spp. inferred by Bayesian analysis within Nostocales based on a maximum of 1,495 nucleotides from the 16S rRNA gene (240 OTUs). Branch support values are shown as BA/ ML/MP. Support values of 100 in ML and MP, or 1.00 (BA), is displayed as "*," respectively, nodes with no support are shown as "-." Table S1. Summary of likelihood substitution model parameters for the GTR+G+I nucleotide models and Ny98 codon model. CYANOMARGARITA GEN. NOV.