THE MOLECULAR PHYLOGENETIC AND GEOMETRIC MORPHOMETRIC EVALUATION OF MICRASTERIAS CRUX-MELITENSIS/M. RADIANS SPECIES COMPLEX¹

Origin and cultivation of strains, LM and SEM observations. The investigated strains were acquired from algal culture collections at the Hamburg (SVCK) and Prague (CAUP) universities (Table 1). Cultures were initiated with an inoculum of 10–15 cells and grown for 6 weeks in 100 mL Erlenmeyer flasks containing liquid oligotrophic medium developed by the Culture Collection of Algae of Charles University of Prague (CAUP) (http://botany.natur.cuni.cz/algo/caup.html). Strains were maintained at temperatures of 20°C and illuminated at 40 μmol photons · m⁻² · s⁻¹ from 18 W cool fluorescent tubes (Philips TLD 18W/33, Royal Philips Electronics, Amsterdam, the Netherlands), at a light:dark (L:D) regime of 12:12. Microphotographs were taken on an Olympus BX51 light microscope with Olympus Z5060 digital microphotographic equipment (Olympus Corporation, Tokyo, Japan). Projection of light microscopic images was made using the Deep Focus 3.0 module (Promicra s.r.o, Prague, Czech Republic) implemented in the QuickPHOTO CAMERA 2.3 software (Promicra s.r.o). For SEM, the acetone-washed glass coverslips (10 mm in diameter) were placed on a heating block and coated three times with a poly-L-lysine solution (1:10 in distilled water) to ensure better adhesion of cells. After cooling, a drop of the formaldehyde-fixed cell suspension was placed on the glass, and when almost dry, it was transferred into 30% acetone and dehydrated by an acetone series (10 min successively in 30, 50, 70, 90, 95, 99% and 2x in 100%). Finally, cells were dried to a critical point with liquid CO₂, subsequently sputter-coated with gold (Bal-Tec Sputter Coater SCD 050, Capovani Brothers Inc., Sconia, NY, USA), and examined using the JEOL 6380 LV scanning electron microscope (JEOL Ltd., Tokyo, Japan).

DNA isolation, amplification, and sequencing. DNA was extracted from the strains listed in Table 1. After centrifugation in an MPW 223 centrifuge (MPW Med. Instruments, Warsaw, Poland), algal cells were mechanically disrupted by shaking in the presence of glass beads (0.5 mm diameter, Sigma-Aldrich, St. Louis, MO, USA). Genomic DNA was extracted using the Invisorb Spin Plant Mini Kit (Invitek, Berlin, Germany) following the manufacturer’s instructions. The ITS rDNA regions were amplified by PCR using the algal-specific primer nr-SSU-1780-5′ (5′-CTGCGGAAGGATCATTGATTC-3′; Piercey-Normore and DePriest 2001) and a universal primer ITS4-3′ (5′-TCCTCCGCTTATTGATATGC-3′; White et al. 1990). Additionally, the trnG^uuc intron was amplified using newly designed primers trnG-uuc-F-5′ (5′-AGCGGGTATAGTTTAGTGGT-3′) and trnG-uuc-R-3′ (5′-GGTAGCGGGAATCGAACCCGC-3′). The new primers were designed based on a published chloroplast genome of Staurastrum punctulatum (GenBank accession no. AY958085; Turmel et al. 2005). All PCRs were performed in 20 μL reaction volumes (15.6 μL sterile Milli-Q Water [Millipore Corp., Bedford, MA, USA], 2 μL 10′ PCR buffer [Sigma-Aldrich], 0.4 μL dNTP [10 μM], 0.25 μL of primers [25 pmol · mL⁻¹], 0.5 μL Red Taq DNA Polymerase [Sigma-Aldrich] [1 U · mL⁻¹], and 1 μL of DNA [not quantified]). PCR reactions were performed in either an XP thermal cycler (Bioer, Tokyo, Japan) or a Touchgene gradient cycler (Techne, Cambridge, UK). PCR amplification of the ITS rDNA began with an initial denaturation at 95°C for 5 min and was followed by 35 cycles of denaturing at 95°C for 1 min, annealing at 54°C for 1 min and elongation at 72°C for 1 min, with a final extension at 72°C for 7 min. Identical conditions were used for the amplification of the trnG^uuc intron, except that an annealing temperature of 66°C was used. The PCR products were quantified on a 1% agarose gel stained with ethidium bromide and cleaned either with the JetQuick PCR Purification Kit (Genomed, Löhne, Germany) or with QIAquick Gel Extraction Kit (Qiagen Inc., Valencia, CA, USA) according to the manufacturer’s protocols. The purified amplification products were sequenced with the PCR primers with an Applied Biosystems (Seoul, Korea) automated sequencer (ABI 3730xl) at Macrogen Corp. in Seoul, Korea. Sequencing reads were assembled and edited using SeqAssem software (SequentiX, Klein Raden, Germany). After checking the diversity in ITS rDNA and trnG^uuc intron sequences, SSU rDNA regions were amplified for selected strains. PCR amplification was performed as described in Neustupa and Škaloud (2007). In total, nine ITS rDNA, nine trnG^uuc intron, and two SSU rDNA sequences were generated and submitted to GenBank (Table 1).

Sequence alignment. ITS2 rDNA and trnG^uuc intron sequences were manually aligned in MEGA 4 (Kumar et al. 2008). The trnG^uuc intron sequences were all of equal length (750 bases); their alignment was straightforward and unambiguous. However, ITS2 rDNA sequences showed considerable variability that made it difficult to create precise alignment. Thus, the common ITS2 rRNA secondary-structure transcript has been constructed using the mfold program (version 2.3; Walter et al. 1994, Zuker 2003) to align the sequences correctly. The final alignment (374 bases) was generated in 4SALE (Seibel et al. 2006, 2008), according to the secondary-structure information (Fig. S1 in the supplementary material). The final SSU rDNA alignment was generated as follows. First, we downloaded 280 aligned sequences of Desmidiales from the SILVA database, version 98 (Pruesse et al. 2007); then, the alignment was reduced to Desmidiaceae and analyzed by the neighbor-joining (NJ) method in PAUP*, version 4.0b10 (Nylander 2004). On the basis of an inferred phylogenetic tree and literary data (Gontcharov et al. 2003, Gontcharov and Melkonian 2008, Hall et al. 2008), the alignment was reduced to 63 sequences that were selected to encompass all Desmidiaceae lineages. In addition, the long-branched Cosmarium sequences AJ428113, AJ428114, AY964133, and AY964135 were removed to avoid the long-branch attraction (LBA) effect in phylogenetic analyses. Finally, two new SSU rDNA Micrasterias sequences, together with one closely related sequence revealed by BLAST searches, were added to the alignment and aligned with the help of the published SSU rRNA secondary-structure transcript of the genus Closterium (Denboh et al. 2001). The final alignment thus comprises 62 sequences (Table S1 in the supplementary material). Ambiguously aligned regions (loops at the end of stem 17 and between the stems 45 and 46) and positions with deletions in most sequences were removed from the alignment, resulting in an alignment comprising 1,725 base positions. All of the alignments were submitted to TreeBASE (ID: SN4729).

Phylogenetic analyses. The phylogenetic trees were inferred with Bayesian inference (BI) using MrBayes version 3.1 (Ronquist and Huelsenbeck 2003). The most appropriate substitution model was estimated for each data set using the Akaike information criterion (AIC) with PAUP/MrModeltest 1.0b (Nylander 2004). In the BI analysis, two parallel Markov chain Monte Carlo (MCMC) runs were carried out for three million generations, each with one cold and three heated chains. Bootstrap analyses were performed by maximum-likelihood (ML) and weighted parsimony (wMP) criteria using PAUP*, version 4.0b10. ML bootstrap analysis (100 replications) consisting of heuristic searches with 10 random sequence addition replicates, tree bisection-reconnection (TBR) swapping, and a rearrangement limit of 5,000 for each replicate. The wMP bootstrapping (1,000 replications) was performed using heuristic searches with 100 random sequence addition replicates, TBR swapping, random addition of sequences (the number limited to 10,000 for each replicate), and gap characters treated as a fifth character state. Congruence between ITS2 rDNA and trnG^uuc intron data sets was tested using the incongruence length difference (ILD) test (Farris et al. 1995), as implemented by the partition homogeneity test in PAUP* (heuristic search, simple addition, TBR branching swapping, 100,000 replicates).

Morphometric analyses. For each strain, 50 adult semicells were randomly chosen and photographed. On each semicell, 31 structurally corresponding landmarks and semilandmarks were depicted (Fig. 1). Whereas the landmarks were placed in fixed positions, the semilandmarks (nos. 10, 11, 15, 17, 21, and 22) were allowed to slide along the abscissa connecting adjacent landmarks (Zelditch et al. 2004). For most of the geometric morphometric analyses, the TPS-series software (available at http://life.bio.sunysb.edu/morph/) was used. Positions of landmarks and length and width of the cells were digitized in TpsDig, ver. 2.12. The landmark configurations of the entire set of 450 objects were superimposed by generalized Procrustes analysis (GPA) in TpsRelw, ver. 1.42. This widely used method standardizes the size of the object and optimizes the rotation and translation so that the distances between corresponding landmarks of investigated objects are minimized (Bookstein 1991, Zelditch et al. 2004). Correlation between Procrustes and the Kendall tangent space distances was assessed using TpsSmall, ver. 1.20, to ensure that the variation in shape was small enough to allow subsequent statistical analyses (Zelditch et al. 2004). The correlation of Procrustes and Kendall shape spaces was very high (r = 0.999), so we proceeded with further analyses. The Micrasterias semicells are bilaterally symmetrical, and their anterior and posterior sides do not differ (Neustupa and Škaloud 2007). As the asymmetry of semicells was not of interest in this study, cells were symmetrized prior to analysis following the standard formula of Klingenberg et al. (2002). This involved reflecting the landmark configurations (by multiplying the x-coordinates in all landmarks by −1). Then, the paired landmarks in the reflected copy were relabeled, and the original and mirrored configurations were averaged in the general Procrustes superimposition (Klingenberg et al. 2002). A principal component analysis (PCA) of geometric morphometric data was conducted on the entire set of 450 semicells. Scores of the objects on the all nonzero 29 principal component (PC) axes were used for canonical variate analysis (CVA) in PAST, ver. 1.89 (Hammer et al. 2001), to test for the differences in shape of individual strains. In addition, scores on PC axes were used for the two-group linear discrimination analyses between all pairs of investigated strains. Significance of the difference between group mean shape configurations was assessed by the Hotelling’s T² test. In parallel, the K-means clustering, a method based on the nonhierarchical clustering of multivariate data into a specified number of groups (Bishop 1995), was used to illustrate relative differences in shape of the individual strains. The K-means clustering into two groups was conducted in pairs of all investigated strains. Number of cells classified out of the original pattern of 50 versus 50 objects from two strains in each pair indicated their overall shape similarity. K-means clustering pattern identical with the underlying origin of cells from two strains indicated clear separation of groups, while a high number of cells placed into the wrong group illustrated the close relationship of two groups on the basis of their landmark-based shape data.

Phylogenetic analyses. The genetic diversity of nine Micrasterias strains was determined using the ITS2 of rDNA and the group II intron sequences of the plastid gene that encodes transfer RNA-Gly (trnG^uuc). Results of genotype and haplotype groupings among the strains are illustrated in Figure 2. These two markers differed considerably in the amount of their variation. The ITS2 rDNA showed a higher average genetic distance (Kimura 2-parameter model) between the lineages (0.168) than the trnG^uuc intron (0.032). The ITS2 rDNA sequences were so variable that the alignment could only be established with the aid of the common ITS2 rRNA secondary structure transcript (Fig. S1). On the other hand, the alignment of trnG^uuc intron sequences was straightforward and unambiguous, allowing us to analyze the phylogeny of nine investigated Micrasterias strains, together with the three related strains (M. crux-melitensis NIES 152, GenBank accession no. FN562171; M. truncata CAUP K606, FN562173; and Staurodesmus dickiei ASW 07056, FN562172). In general, genotype groups observed in the ITS2 rDNA sequences (Fig. 2A) corroborated the haplotype groupings observed in the trnG^uuc intron sequences (Fig. 2B). The congruence of both phylogenies was supported by the statistically significant partition homogeneity test (ILD test; P = 0.52). Generally, three lineages could be established from the phylogenetic analyses: (i) a large lineage comprising all the European and the single North American M. crux-melitensis isolates (CAUP K602, CAUP K607, SVCK 98, SVCK 128, SVCK 266, and SVCK 431, respectively); (ii) the single M. radians strain originating from Southeast Asia (SVCK 389); and (iii) two M. radians isolates of tropical African origin (SVCK 518 and SVCK 519). The single difference between the ITS2 and trnG^uuc phylogenies pertained to the genetic identity of the European strains. The ITS2 analysis showed the divergence of SVCK 431, whose sequence differed by two substitutions and one insertion from that of the other strains in the European lineage. By contrast, the trnG^uuc analysis revealed a single substitution in the SVCK 98 sequence that differentiated it from the rest of the strains. Moreover, the trnG^uuc data revealed the monophyly of M. crux-melitensis/radians complex (BI/ML/MP support 0.98/93/99). However, the strains pertaining to the same traditional morphospecies did not cluster together (Fig. 2B). According to the trnG^uuc sequences, European/North American M. crux-melitensis isolates formed a well-supported clade (1.00/98/100) with Asian M. radians SVCK 389. Similarly, the pair of African M. radians isolates significantly clustered with M. crux-melitensis NIES 152 (1.00/100/100).

Although both ITS2 rDNA and trnG^uuc markers were very useful to reveal genetic relationships among nine investigated Micrasterias strains, they were too variable for determining phylogenetic position of the three recognized lineages within the genus Micrasterias and Desmidiaceae as a whole. Therefore, SSU rDNA sequences of the selected representative strains (CAUP K602, SVCK 389, and SVCK 518) were compared with the SSU rDNA data set of 59 desmid species (Fig. 3). All three lineages formed a well-supported independent clade, together with M. crux-melitensis NIES 152, M. truncata CAUP K606, and S. dickiei SVCK 38. The strain of M. crux-melitensis CAUP K602 belonging to the European/North American cluster formed a clade (74/95/0.99) with the Asian M. radians SVCK 389, which did not include the African M. radians SVCK 518 (Fig. 3). Although the exact position within Desmidiaceae remained ambiguous, BI, ML, and MP analyses (latter two not shown) consistently showed all other Micrasterias sequences as the next closest relatives to this clade. However, neither the relationship with any of these other Micrasterias species nor the lineage encompassing the genus as a whole received significant internal branch support.

Morphology of cells. The cells of all strains were readily identifiable into M. crux-melitensis/M. radians species complex (Fig. 4). However, there were apparent differences in morphology of individual strains. Cell dimensions profoundly differed (Fig. 5, A and B). The strains SVCK 266, SVCK 431, and SVCK 98 consistently had the smallest cells, while cells of the SVCK 128, CAUP K602, CAUP K607, and SVCK 518 strains were ∼35% to 45% larger. Moreover, the degree of lobulation differed among strains. Most strains had well-developed second-order lobules present in all semicells (Fig. 4, A, B, E–I), with the notable exception of the SVCK98 strain, whose second-order lobules were only weakly developed (Fig. 4D). At the same time, this strain had the relatively smallest cells (Fig. 5, A and B), and it corresponded well to the M. crux-melitensis var. janeira, according to the traditional taxonomy (Krieger 1939). On the other hand, the SVCK128 strain of M. crux-melitensis var. superflua had third-order lobules present on most semicells (Fig. 4C) and relatively large cells (Fig. 5, A and B), so that it fit well into the description of this variety. The second strain, originally assigned as M. crux-melitensis var. supeflua, SVCK 266, was different from SVCK 128 as it had considerably smaller cells and most semicells did not develop the third-order lobules. The SVCK 431, CAUP K602, and K607 fit into descriptions of the type variety M. crux-melitensis var. crux-melitensis, even if their cell dimensions differed among strains. In contrast to other strains, the SVCK 518 and SVCK 519 had a pair of spines on each side of their polar lobes (Fig. 6, A–C). However, these polar lobe spines were not always symmetrically developed on all semicells, and, sometimes, there was just a single asymmetrical spine present (Fig. 6B). This characteristic presence of polar lobe spines, as well as the long furcate lobules of their semicells, led us to identify these strains as M. radians var. evoluta (W. B. Turner) Willi Krieg. according to traditional criteria (Krieger 1939). The SVCK 389 strain of M. radians var. bogoriensis (C. J. Bernard) Willi Krieg. fit well into the taxonomic description of this variety, mainly by the presence of long, slightly curved spines on tips of the polar lobes and lateral lobules (Fig. 6F). Furthermore, the third-order lobules were sometimes present, in accordance with the taxonomic delimitation of this variety. The cell walls of all the investigated strains were consistently provided with pore-linked mucilage extrusions (Fig. 6, D–F), and no differences among strains were detected in this genus-wide feature.

Geometric morphometrics. The PCA of the symmetrized Procrustes-superimposed landmark data yielded 29 nonzero PC axes. The first PC axis spanned 59.5% of the total variation. Semicells varied from being deeply lobed, with a narrow polar lobe in negative values, to compact shapes with shallow incisions and a wide polar lobe in positive marginal positions along the first PC axis (Fig. 7). Consequently, the strains SVCK 518, SVCK 519, and SVCK 389, which were characterized by deeply lobed semicells of the M. radians type were positioned close to the negative margin on PC1. Conversely, the SVCK 98 strain had the most positive PC1 scores corresponding to its compact semicells with shallow incisions (Fig. 8, A and B). The second PC axis, which spanned 10.44% of the total shape variation, was related to width of the upper lateral lobule. Semicells with negative scores on PC2 had narrow upper lateral lobules and open incisions, while semicells with positive PC2 scores were characterized by closed incisions and wide upper lateral lobules (7, 8). This axis primarily separated strains SVCK 518, SVCK 519, SVCK 389, and SVCK 98 positioned mostly in negative PC2 values from the rest of the strains.

The CVA illustrated the separation of strains on the basis of their shape characteristics (Fig. 8, C and D). The overall separation among strains was highly significant (Wilk’s λ = 1.46e10⁻⁵, P = 0). The SVCK 98 strain was the most dissimilar from all other strains. In addition, the three strains SVCK 518, SVCK 519, and SVCK 389, assigned originally as M. radians, were grouped together and more or less separated from the remaining strains (Fig. 8, C and D). The pair comparisons revealed that all strains were mutually statistically significantly different in their mean shape characteristics (Table 2). The P-value of the Hotelling’s T² tests was <10⁻⁵ in all pairs, and there were 100% correctly classified cells in most of the pair comparisons. The K-means clustering algorithm illustrated a higher number of objects classified across the original group assignment (Table 2). In this respect, the pairs of M. crux-melitensis strains from the Czech Republic (CAUP K602, K607) and the pair of African M. radians var. evoluta strains SVCK 518 and SVCK 519 had the most similarly shaped semicells.