We reconstructed the plastome of 28 strains of Coelastrum. We isolated 11 new strains of Coelastrum from Gull Lake (3 strains), Swan Lake (4 strains), and Wintergreen Lake (2 strains) in Michigan, Coral Gables Canal in Florida (1 strain), and Lake Buchanan (1 strain) in Texas, all collected using a 20 μm mesh plankton net. The isolated strains were cultured and maintained in WC (Wright Chu) artificial freshwater media (Guillard, 1975). Another eight strains were obtained from the UTEX Culture Collection, and ten strains were acquired from the Experimental Phycology and Culture Collection of Algae (EPSAG). Additionally, one Coelastrum plastome and 18 Scenedesmaceae outgroup plastomes were downloaded from NCBI (http://www.ncbi.nlm.nih.gov/) for phylogenetic analysis: Coelastrum microporum (NC068582), Asterarcys sp. (MK995333), Coccoidesmus tetrasporum (OR350844), Coelastrella saipanensis (NC042181), Coronastrum ellipsoideum (PP979513), Crucigenia lauterbornii (PP979532), Crucigenia quadrata (PQ301446), Desmodesmus abundans (NC066651), Desmodesmus spinosus (PV295633), Pectinodesmus pectinatus (NC036668), Tetradesmus arenicola (NC086756), Tetradesmus bajacalifornicus (NC086755), Tetradesmus dimorphus (NC086754), Tetradesmus distendus (NC086753), Tetradesmus lancea (OR502671), Tetradesmus major f. lunatus (OR502665), Tetradesmus obliquus (NC008101), Tetradesmus obliquus var. spiraliformis (OR502672), and Tetradesmus reginae (NC086752).

All Coelastrum strains grown in our laboratory were observed under a light microscope for initial species identification based on morphological criteria, while cells were in exponential growth phase in WC media. Increases in cell abundance were determined by daily fluorescence in 25 mm diameter glass tubes in a Turner TD-700^® fluorometer. Scanning electron microscopy (SEM) was then utilized to evaluate ultrastructural variation in cell walls among species. Strains were fixed with formaldehyde or glutaraldehyde and dehydrated on a 25 mm diameter, 0.2 μm pore size membrane filter. Critical-point dried samples were mounted onto aluminum SEM stubs and coated with iridium using a sputter coater. Cell shape and ultrastructure were examined to confirm identity at the species levels based on authoritative references (Fenwick et al., 1966; Fenwick, 1968; Komárek and Fott, 1983; Hegewald et al., 2010).

Our identification of publicly available strains matched the names associated with those deposits and the literature. Our identification of our new strains was based upon agreement with the literature and with the morphology of the publicly available strains. We use the term “nominal species” to emphasize that any classification at any time is a hypothesis.

To obtain DNA, cultures grown in WC medium were harvested in exponential phase, and cells were pelleted by centrifugation at 4,500 rpm for 20 min. DNA was extracted from the collected pellets for next-generation sequencing (NGS) using a DNeasy^® Plant Mini Kit (Qiagen, Hilden, Germany) following the manufacturers protocol. DNA quantity was measured using the Qubit^® double stranded DNA High Sensitivity Assay Kit and the Qubit^® 2.0 Fluorometer, while DNA quality was assessed with the NanoDrop^® ND-1000 UV-Vis Spectrophotometer.

Total genomic DNA extracted from cultured Coelastrum strains was submitted to one of two facilities for high-throughput sequencing: the Genome Sequencing and Analysis Facility (GSAF) at the University of Texas at Austin or to Novogene (Beijing, China). For most strains, short-read sequencing libraries were prepared and sequenced on the Illumina HiSeq 4000 platform (Illumina, San Diego, CA), generating approximately 30 million paired-end reads (150 bp read length) per sample.

Short-read Illumina sequencing failed to yield complete plastome assemblies for three strains—SAG 2078, SAG 2248, and SAG 41.86, thus, long-read sequencing was performed on these three strains using the PacBio Sequel II platform with library preparation and sequencing carried out by Novogene.

Adapter sequences and low-quality bases were removed using BBDuk from the BBTools software suite (https://jgi.doe.gov/data-and-tools/bbtools/). Clean short reads were assembled de novo using NOVOPlasty v4.2.1 (Dierckxsens et al., 2017) on the Texas Advanced Computing Center (TACC) supercomputing platform, using an optimized k-mer size of 33 and an insert size of 300 bp. Long-read data were assembled with ptGAUL (plastid Genome Assembly Using Long reads) (Zhou et al., 2023), which is specifically designed for accurate reconstruction of plastid genomes from long-read datasets.

All resulting plastome assemblies were imported into Geneious Prime v2020.2.4 (Biomatters Ltd., Auckland, New Zealand). Assembly completeness, gene order, and IR boundaries for newly sequenced strains were examined, and read mapping was performed using BBMap (Bushnell, 2014) to verify coverage uniformity and support assembly accuracy. Taxonomic verification was conducted using BLAST searches against the NCBI nucleotide database (Altschul et al., 1990) to detect possible contaminants and confirm species identity. Plastome sequences were annotated in Geneious based on homologous genes from closely related Scenedesmaceae taxa and further validated with tRNAscan-SE v2.0 (Lowe and Chan, 2016) for tRNA identification and RNAmmer v1.2 (Lagesen et al., 2007) for rRNA gene prediction.

IR boundaries and annotations for all strains not sequenced by our laboratory followed annotations downloaded from NCBI.

Phylogenetic relationships were inferred using protein coding genes (CDSs) shared across all 29 Coelastrum plastomes and at least 16 of the 18 outgroup species (Supplementary Table S3). Coding sequences for these 62 genes were extracted from the annotated plastomes in Geneious. CDS sequences were aligned using MAFFT v. 7.450 (Katoh and Standley, 2013) with a default setting in Geneious, and the resulting alignment was used for phylogenetic inference. Missing genes were treated as missing data, and maximum likelihood phylogenies were constructed using IQ-TREE2 v1.6.12 (Minh et al., 2020) with 1,000 bootstrap replicates. The best-fit substitution model, GTR+F+R5, was selected by ModelFinder (Kalyaanamoorthy et al., 2017). The resulting phylogenetic tree was visualized in FigTree v1.4.3 (http://tree.bio.ed.ac.uk/software/figtree/).

The aligned data matrix is available from the authors upon request.

IR boundaries determined by methods below for newly obtained sequences, and from NCBI annotations for downloaded plastomes, were mapped onto the best ML tree under parsimony.

Intron presence and distribution were assessed across the newly assembled plastomes of Coelastrum strains and Coelastrum microporum (NC068582). Intron-containing genes were identified manually from the annotated plastomes using Geneious, and exon-intron boundaries were inferred by alignment with homologous plastid sequences from other Coelastrum species and from the related Scenedesmaceae genera Coelastrella, Pectinodesmus, Scenedesmus, and Tetradesmus (de Cambiaire et al., 2006; Wang et al., 2019; Zhao et al., 2022; Wang et al., 2024). To confirm intron identity and detect sequence conservation, all intron sequences were queried against the NCBI nucleotide database using BLAST. For each plastome, the total number of introns and the cumulative intron length (bp) were recorded. Intron variation was tabulated across strains and visualized alongside a maximum likelihood phylogeny to assess patterns of intron gain and loss across Coelastrum strains.

To evaluate the relationship between intron variation and overall plastome size, Pearson correlation tests were performed using R v4.3.0. Specifically, correlations were assessed between total intron number and size, and non-coding region size. This analysis aimed to identify which genomic features contribute most significantly to plastome size variation within Coelastrum.

Repeat content was analyzed across 29 Coelastrum plastomes to evaluate the contribution of repetitive DNA to genome size variation. Prior to analysis, one copy of the inverted repeat (IRA) was removed from each plastome to prevent redundancy. Tandem repeats were detected using Tandem Repeats Finder v.4.09 (Benson, 1999) via the web interface, with default parameters. The total number and cumulative length of tandem repeats were calculated for each plastome.

Dispersed repeats were detected by running a BLASTN search of each plastome against itself using BLAST v2.16.0+ (Altschul et al., 1990), with word size of 16 and a minimum identity threshold of 80%, following the methods of Lee et al. (2020). Dispersed repeats were identified by retaining BLAST hits whose aligned query and subject regions occurred at distinct, non-overlapping positions within each plastome. Self-hits were removed, and matches representing overlapping or adjacent tandem duplications were excluded to ensure that only dispersed repeats were retained. Total repeats and the proportion of repeat content were calculated for each plastome.

Our analysis of plastome data recovered monophyly of all strains of Coelastrum with 100% BS support (Supplementary Figure S1). Two nominal species were represented by only one strain each: C. cambricum (UTEX 2446), and C. indicum (SAG 2363). They were recovered along with C. microporum (UTEX 281) in a clade with 100% BS support. The latter species was represented by a total of seven strains. The other six were recovered in two separate clades, with strains UTEX 1354, CG1, CW5, and CS5 in one clade and strains CF1 and SAG 2292 in another. In short, plastome data suggested that the morphology associated with C. microporum may consist of several independent lineages. A clade comprising C. sphaericum (SAG 1.82 and SAG 32.81) and C. proboscideum (UTEX 184 and UTEX 282) had 100% BS support, but neither nominal species was resolved as monophyletic within this clade.

In contrast, all other nominal species were recovered as monophyletic with high BS support. All C. reticulatum strains and all C. morus strains were each resolved with 100% BS support and those clades were resolved as sister to one another with 100% BS support. All six C. astroideum strains (CW1, CG6, CS1, CS3, CS9, and SAG 33.88) and all four C. pseudomicroporum strains (CLB, SAG 2077, UTEX 1353, and UTEX 280) were each recovered as monophyletic with 100% BS support.

The newly assembled plastome sequences had a mean coverage ranging from 563.1 to 2,847.8X for 150 bp pair-end Illumina reads, and from 56.3 to 116.0X for long-read PacBio sequencing (Table 1). All plastomes were fully annotated and deposited in the NCBI GenBank database, with accession numbers provided in Table 1. Each plastome displayed a typical quadripartite structure, consisting of a large single copy (LSC) region and a small single copy (SSC) region with two inverted repeats (IRA and IRB).

Despite this conserved structural organization, plastome sizes showed substantial interspecific variation, ranging from 166,827 bp in Coelastrum pseudomicroporum (UTEX 1353) to 553,457 bp in Coelastrum morus (SAG 41.86) (Figure 1 and Table 1). The size of the single copy regions ranged from 137,407 bp to 423,190 bp (approximately a threefold range), whereas IR size ranged from 7,650bp to 65,144 bp, representing more than an eightfold range. These regions corresponded to 8.2% and 37.84% of total plastome size, respectively.

There were 4 types of inverted repeat boundaries (Figure 2). Inverted repeat expansion in certain lineages resulted in the translocation of four genes (psbC, atpF, atpH, and ftsH) across the LSC and the IR region. In Type 1, psbC was positioned in the LSC region near the IRB-LSC junction in the C. morus clade and in C. reticulatum CG10. It was also found in C. cambricum and all C. microporum and C. pseudomicroporum. The Type 2 boundary is similar, with the IR boundary bisecting the psbC gene, and was found in two C. reticulatum strains (SAG 8.81 and UTEX 1365) and in the C. sphaericum - C. proboscideum clade. Type 3 was restricted to C. indicum which retained psbC and atpF genes in the LSC but atpH and ftsH genes were found in the IRB region. Type 4 had a partial duplication of the ftsH gene within the IRB region of C. astroideum. Outgroup plastomes nearly all had the Type 1 boundary. Pectinodesmus pectinatus (NC036668) and Tetradesmus obliquus (NC008101) had Type 2 boundaries. Crucigenia lauterbornii (PP979532) and Coronastrum ellipsoideum (PP979513) had very different IR boundaries than any of the other plastomes. Regardless of where one might draw the root among outgroups, Type 1 still maps unambiguously as plesiomorphic to all of Coelastrum.

The total coding region length ranged from 89,638 bp to 98,170 bp (Table 1). Overall GC content across the genus was relatively consistent, ranging from 29.1% to 32.1%. However, one C. morus strain showed a lower GC content of 27.4%. In contrast, non-coding regions varied across strains, from 75,745 bp to 455,287 bp, comprising up to 82.3% of the total plastome length. This variation in non-coding content likely plays a central role in the plastome size differences observed among Coelastrum strains. To investigate the potential origins of the expanded intergenic regions in C. morus, we performed BLAST searches using representative intergenic sequences. These searches yielded no significant matches.

Comparative analyses of gene content across 29 Coelastrum plastomes revealed overall conservation in core gene composition, with the exception of minor variation in tRNA gene content (Supplementary Table S1). The most pronounced case of gene loss was observed in Coelastrum reticulatum (UTEX 1365), which lacked five contiguous genes, clpP, rpl2, rpl23, rps4, and rps19, that are typically co-localized within the small single copy (SSC) region in other Coelastrum strains (Figure 1). This does not appear to be an artefact as coverage was deep and uniform across the region of interest (Supplementary Figure S3). In addition, trnF-GAA and trnL-TAG were also absent in this strain, suggesting either a localized deletion event or significant genomic rearrangement in the SSC region. Independent tRNA gene losses were detected in two additional strains: trnR-UCU was absent in C. reticulatum (SAG 8.81), and trnS-GCU was missing from C. proboscideum (UTEX 282).

Across the 29 assembled Coelastrum plastomes, intron distribution showed considerable heterogeneity in both location and frequency, reflecting dynamic evolutionary events. A total of 13 genes were identified as intron-bearing, with psaA, psbA, psbC, rbcL, and rrn23 exhibiting the greatest variability in intron content and length (Figure 3). The psaA gene consistently displayed a trans-spliced structure with introns separating three exons. The cis-spliced configuration was observed in most Coelastrum strains.

Unique intron insertions were identified in several genes across the Coelastrum plastomes. Notably, an intron at position 66 in the psbE gene was detected in two of the three C. morus strains, but was absent in the third strain, indicating strain-specific intron retention or loss within the species. The psbZ intron (position 65) was observed only in C. cambricum, suggesting this to be a unique insertion event within Coelastrum. The atpA intron, located at site 489, was present in three strains of C. astroideum that did not form monophyletic group and one strain of C. microporum (UTEX 2446) in an unrelated clade. The canonical group I intron within trnL-UAA was conserved across all sampled plastomes, consistent with its widely reported vertical inheritance and evolutionary stability within Chlorophyta.

Four strains were outliers in terms of overall plastome size (Table 1). The two largest plastomes belonged to two C. morus strains (SAG 2248 and SAG 41.86). They were nearly identical in size to each other but were 85% larger than the third C. morus strain SAG 2078, which was 23% larger than that of C. reticulatum UTEX 1365. The latter was 16% larger than the next largest. After that, there was never more than a 4% difference in plastome size between any two strains. The reporting that follows refers to these four largest strains as the outliers, or as the four large strains.

There was relatively little variation in coding region size whether or not one considered outliers. For example, the coefficient of variation of the aggregate size of coding regions (from both single copy regions plus one IR copy) was about 2.2% with outliers and 1.5% without outliers (Table 1). The largest absolute contributor to plastome size was the non-coding region size which accounted for over 99% of the variation in total genome size whether or not outlier strains were included (r = 0.99, p < 0.001; Figures 4A, B).

We then analyzed contribution of various elements to aggregate non-coding region size. With outlier strains included, intron number and aggregate intron size were each positively correlated to non-coding size (Figures 4C, E) but removal of the outliers left no significant correlation (Figures 4D, F). However, regardless of whether or not outliers were included, total number and aggregate size of repeated elements were correlated with aggregate non-coding region size (Figures 4G-J).

That repeat elements were important contributors to size variation in all strains warranted more detailed investigation of these elements. Across the 29 Coelastrum plastomes, both dispersed and tandem repeats exhibited lineage-specific variation (Table 2). Repeat content was typically limited to 1–5 dispersed repeats per plastome, typically totaling <1,000 bp and contributing <0.5% of the genome. Tandem repeats in these compact plastomes were likewise limited, generally numbering fewer than 40 and spanning <2,000 bp, corresponding to <1% of total plastome size. In sharp contrast, repeat proliferation was dramatic in the C. morus lineage. The two giant plastomes, SAG 2248 and SAG 41.86, contained more than 26,000 dispersed repeats, representing ~200 kb of sequence and accounting for ~41% of their genome size. These same strains also harbored exceptionally high levels of tandem repeats, each exceeding 400 elements and ~53,000 bp in total length (~10% of the genome). Altogether, repetitive DNA surpassed 250 kb in these plastomes, contributing more than 50% of total genome size, whereas all other Coelastrum plastomes contained <5% repetitive DNA. Intermediate cases, such as C. reticulatum (UTEX 1365) and C. morus (SAG 2078), exhibited moderate repeat proliferation (100–120 dispersed repeats; ~7.5–8 kb), suggesting that repeat accumulation may have occurred in a stepwise and lineage-specific manner. These results again highlight repetitive DNA, both dispersed and tandem repeats, as a major driver of the genome size variation observed within Coelastrum.

The four outliers in overall plastome size also had the four largest IR sizes (Table 1). Size of the IR was positively correlated with total plastome size when outliers were included (Figure 4K), but not when outliers were excluded (Figure 4L). As for the entire plastome, non-coding DNA in the IR was positively correlated with IR size whether or not outliers were included (Figures 4M, N).

Our study reveals a nearly 3.3-fold variation in plastome size within Coelastrum, ranging from 166,827 bp in C. pseudomicroporum (UTEX 1353) to 553,457 bp in C. morus (SAG 41.86). A nearly two-fold difference in plastome size was observed between strains nominally identified as C. morus, representing an intraspecific size variation in green algae. The two strains of C. morus (SAG 2248 and SAG 41.86) stand out with an unprecedented genome size exceeding 550 kb, making them the largest plastomes reported within the order Sphaeropleales. Until now, plastomes of this magnitude have only been described in the green algal order Chlamydomonadales and Chaetopeltidales, such as Volvox carteri, Floydiella terrestris, Haematococcus lacustris, and H. pluvialis (Smith and Lee, 2009, 2010; Brouard et al., 2010; Bauman et al., 2018; Ren et al., 2021). These species often share several features including high GC content, repetitive DNA, and extensive intergenic regions. Plastome inflation has been proposed to be caused by error-prone DNA repair or mutational hazard in non-coding regions and repeat-rich regions (Smith and Lee, 2009; Brouard et al., 2010; Muñoz-Gómez et al., 2017; Smith, 2018, 2020; Ren et al., 2021).

In contrast, the plastomes of all four outliers, including the two largest C. morus strains, present a unique evolutionary case. Despite their extreme size, these four outliers do not exhibit unusually high GC content (range of GC = 27.4 - 30.5), nor do they show the remarkable intron density found in red algal lineages such as Bulboplastis apyrenoidosa and Corynoplastis japonica, which possess over 200 and 300 introns, respectively (Muñoz-Gómez et al., 2017). Instead, the two largest C. morus strains contain a modest 29 introns. Plastomes of the smaller outliers C. morus SAG 2078 and C. reticulatum UTEX 1365 contain only 23 and 19 introns. Taken together, these observations underscore that intron proliferation is not the main driver of non-coding region size and overall plastome enlargement in Coelastrum. This is particularly true when ignoring the outliers where we found lack of any significant correlation with intron number or size to total plastome size when outliers were removed from the analysis.

Our findings suggest that genome expansion in Coelastrum is primarily driven by the accumulation of non-coding sequence, particularly through the expansion of intergenic regions, rather than increased gene content or intron load. This mirrors patterns observed in Haematococcus lacustris, where plastome inflation is attributed to intergenic expansion (Smith, 2018), and in B. apyrenoidosa, where enlarged intergenic regions are linked to insertion sequences of bacterial origin (Muñoz-Gómez et al., 2017). Our BLAST search results of the expanded intergenic regions in C. morus yielded no significant matches, which may reflect limited genomic representation of closely related green algal taxa or potential bacterial donors in public databases.

In addition to intergenic expansion, repetitive DNA is a major driver of plastome enlargement. In the two largest C. morus plastomes, dispersed and tandem repeats exceed 250 kb, accounting for more than half of the total plastome length, whereas most Coelastrum plastomes contain only trace repeat content (<1%), with only a single dispersed repeat in approximately half of the samples. The magnitude of this proliferation parallels Floydiella terrestris, whose plastome is composed of nearly 50% short repeats (Brouard et al., 2010), and the repeat-rich architecture of Haematococcus lacustris, dominated by palindromic and dispersed repeats (Bauman et al., 2018; Smith, 2018). Within Coelastrum, repeat content scales with plastome size: compact genomes harbor little repetitive DNA, whereas enlarged lineages show progressively greater accumulations, culminating in the repeat-rich C. morus (Table 2). This pattern highlights repeat proliferation as a significant contributor to genome expansion. Although the exact molecular mechanisms remain elusive, inefficient DNA repair pathways, such as break-induced replication or the accumulation of palindromic repeats, may underlie intergenic expansion in C. morus plastomes, as proposed in other green and red algal systems (Muñoz-Gómez et al., 2017; Smith, 2020; Ren et al., 2021).

In summary, the pattern in Coelastrum aligns with previous findings in green algal lineages, where plastome enlargement has been attributed to the accumulation of non-coding DNA, repeats, IR, and introns (Brouard et al., 2010; Muñoz-Gómez et al., 2017; Turmel et al., 2017; Bauman et al., 2018; Cremen et al., 2018; Smith, 2018, 2020; Ren et al., 2021).

IR length ranged from 7,650 bp to 65,144 bp, contributing to differences in overall genome size. Multiple IR boundary shifts and gene duplications were observed, particularly involving psbC, atpF, atpH, and ftsH, indicating ongoing IR structural rearrangement. Patterns of IR boundary variation indicate that recently diverged lineages tend to exhibit more extensive IR expansions, reflected by the inclusion of additional genes into the IR (Figure 2). These findings align with the dynamic nature of IR expansion and contraction previously described in other members of green algae (Lemieux et al., 2014; Turmel et al., 2016).

Tetradesmus is the next most densely sampled Scenedesmaceae genus, with 13 strains (Cho and Lee, 2024). Total plastome size ranged from 148,816 to 196,309 bp. From this narrow perspective, the nearly tenfold range in total plastome size of Coelastrum is extraordinary, but we hope our findings will stimulate more detailed work on other Scenedesmaceae groups.

We focused our discussion on intron dynamics solely within Coelastrum because intron analysis has not been reported evenly across other Scenedesmaceae plastome studies. Nevertheless, Zhao et al. (2022) identified introns in seven single copy region genes, five of which also had introns in Coelastrum. The rrn23 gene is the only Pectinodesmus IR gene with an intron; all Coelastrum strains have at least one intron in the same gene. Coelastrella has seven genes with introns, and all seven are shared with Coelastrum (Wang et al., 2019). Coccoidesmus was reported by Wang et al. (2024) to have introns in the large RNA subunit gene and in the psbZ gene. Introns were found in all of our large RNA subunit genes, but only in one psbZ gene (in our single strain of C. cambricum UTEX 2446, see Supplementary Figure S2 and discussed below).

Intron diversity across the 29 Coelastrum plastomes revealed substantial lineage-specific variation in both presence and number, consistent with previous observations of dynamic intron evolution in Chlorophyta. Thirteen genes were identified as intron-bearing, with notable variability in psaA, psbA, psbC, rbcL, and rrn23. The distribution pattern largely mirrors those previously characterized in green algal plastomes, particularly in the order Sphaeropleales (Fučíková et al., 2016; McManus et al., 2018). The psaA gene consistently exhibited a trans-spliced architecture, with three exons interrupted by two introns, a synapomorphy of Sphaeropleales. This trans-splicing configuration is conserved across all Coelastrum strains sampled here. The cis-spliced configuration C1190 was absent in C. microporum (CG1) and the clade of C. reticulatum + C. morus. Within Coelastrum, this is most parsimoniously interpreted as parallel losses. The cis-spliced configuration is widely scattered across the Scenedesmaceae and certain Pediastrum species within the sister family Hydrodictyaceae (McManus et al., 2018), and we hesitate, without increased taxon sampling, to conclude whether this intron presence represents parallel gains across the Sphaeropleales, whether its absence represents parallel losses, or some more complex process.

With the caveat that taxon sampling and reporting of intron data are still uneven in the Scenedesmaceae, it does appear that several introns may be lineage-specific or represent recent insertion events. For instance, introns in psbE (position 66) and psbZ (position 65) were restricted to C. morus (SAG 2248 and SAG 41.86) and C. cambricum (UTEX 2446), respectively. As far as we are aware, there are no previous reports of introns in psbE in any green algae. Similarly, the atpA intron (position 489) was restricted to a clade within C. astroideum strains and to one C. microporum strain (UTEX 281). This intron is shared with the distantly related Neochloris aquatica, suggesting either deep homology or a horizontal transfer event (McManus et al., 2018). The group I intron in trnL-UAA was universally conserved across all Coelastrum plastomes. This intron has been reported as a vertically inherited feature across photosynthetic eukaryotes, from red algae to land plants (Besendahl et al., 2000), and its universal presence here further supports its evolutionary stability within the Chlorophyta.

A striking pattern of intron proliferation was observed in the psbA gene of C. morus. Strains SAG 2078, SAG 2248, and SAG 41.86 contained up to seven introns at unique insertion positions (81, 174, 273, 393, 486, 533, 741, 885), far exceeding the typical intron number (≤4) reported in Scenedesmus obliquus (de Cambiaire et al., 2006) and species within the sister family Hydrodictyaceae (McManus et al., 2018). This lineage-specific intron expansion suggests recent and rapid intron gain events.

Most Coelastrum plastomes retain a conserved gene complement consistent with other Scenedesmaceae. However, C. reticulatum (UTEX 1365) exhibits a unique gene loss cluster affecting five adjacent protein-coding genes (clpP, rpl2, rpl23, rps4, rps19) within the SSC region, along with tRNA genes trnF-GAA and trnL-TAG. Similar clustered gene loss events have been reported in other green algae and are thought to reflect localized rearrangements, deletions, or functional transfers to the nucleus (Brouard et al., 2010; Turmel and Lemieux, 2018). Additional tRNA gene losses in C. reticulatum (SAG 8.81), C. proboscideum (UTEX 282), and outgroups such as Pectinodesmus pectinatus suggest that tRNA genes are particularly prone to loss or replacement. The gene clpP encodes a subunit of the chloroplast Clp protease, while rpl2, rpl23, rps4, and rps19 encode ribosomal protein subunits, suggesting that the loss of these genes may affect plastid protease or ribosome function. Transcriptomic analysis/nuclear genome sequencing needs to be conducted to confirm loss or transfer of these genes to the nucleus.

Our recovery of Coelastrum as monophyletic highlights issues of both gene and taxon sampling in interpreting the phylogeny of the Scenedesmaceae. It is beyond the scope of this paper to resolve issues related to these problems, but we focus on two papers that clearly highlight them.

Hegewald et al. (2010) used a relatively short sequence (ITS2). Nevertheless, we consider this to be a fundamental study because of its breadth of taxon sampling, including 106 Scenedesmaceae strains from eleven nominal genera. In contrast to our results, they recovered a non-monophyletic Coelastrum. Two strains are of particular interest because we selected them for analysis because they occupied critical positions in the Hegewald et al. (2010) tree. Namely, our study and Hegewald et al. (2010) included SAG 2078 (C. morus) and SAG 8.81 (H. reticulata Dangeard = C. reticulatum (Dangeard) Senn). If one were to trim away strains unique to only one study or the other, C. reticulatum SAG 8.81 would be sister to C. morus SAG 2078 in both studies. However, differences in taxon sampling exist, and one cannot ignore that Hegewald et al. (2010) also recovered Coelastrella spec. SAG 217.5 (isolated from Finland), Dimorphococcus lunatus SAG 2241, and Asterarcys quadricellulare COMAS 1977–75 as close relatives to both SAG 8.81 and Coelastrum cambricum SAG 7.81. The latter strain was not available at the time we assembled cultures and data for analysis, so we used C. cambricum UTEX 2446. ITS2 data provided only weak support at the node linking Coelastrella spec. SAG 217.5 to other taxa, whereas plastome data using the already published Coelastrella saipanensis sequence (NC042181), placed that Coelastrella strain in a position distant from all three of our C. reticulatum strains, including SAG 8.81. It is possible that Coelastrum saipanensis (NC042181) and C. spec. SAG 217.5 belong to different lineages, and/or that other taxon sampling differences influenced the relationship of Coelastrella to Coelastrum. We also note that the deeper nodes of the Coelastrum sensu lato branch had very low BS support in Hegewald et al. (2010), indicating that differences between their tree and ours may also be due to gene sampling and low resolving power of the ITS2 at critical nodes.

A much more recent paper used multiple longer sequences in three independent analyses, and again illustrated that both gene and strain sampling affect tree topologies. Wang et al. (2024) analyzed nuclear ITS data (referred to only as the “ITS region” in that study), the nuclear SSU (18S), and the plastome gene tufA independently of one another, each with a similar but not identical taxon sampling design. Coelastrum was recovered as monophyletic with 18S data, but the only two species included were C. proboscideum (SAG 217-3) and C. sphaericum (SAG 217-2). Hariotina reticulata (SAG 8.81) was embedded in a monophyletic Hariotina and that genus was several nodes away from Coelastrum. However, neither genus was monophyletic when ITS data were analyzed, and different Coelastrum strains (C. astroideum var. rugosum RW10 and C. microporum FNY-1) were employed. Finally, with tufA data, and using a third pair of Coelastrum (C. sphaericum CCMA UFSCar 060, and C. sp. YN 15-2), a monophyletic Coelastrum was recovered as sister to a monophyletic Hariotina.

We are not criticizing Wang et al. (2024), as they used sequences available in NCBI, and they discussed several points of incongruence. We only use Hegewald et al. (2010) and Wang et al. (2024) to illustrate that molecular phylogenetic studies need to account for strain and taxon sampling as well as gene sampling when discussing classification. Finally, we note that the SSU and ITS data are from one compartment (nuclear genome) and the tufA data are from another (the plastome). The possibility of distinct gene trees and organismal trees must be considered as well when attempting to corroborate or reject monophyly for Coelastrum.

In summary, future studies of classification using phylogenomic data obviously need to include both more outgroup and ingroup taxa, and more strain representatives for each species, to conclusively support or reject Coelastrum as monophyletic.

Looking towards lower-level classifications, towards the tips of the plastome tree, there is general agreement between species names and plastome genetic diversity. Plastome data also recovered hierarchical structure within nominal species. Discovery of additional diversity within a historical morphological classification at the species level by the use of DNA sequence data often leads to the invocation of “cryptic species”, but morphology in algae is usually treated in some sort of non-canonical fashion, whereas molecular data are treated phylogenetically (Alverson, 2008).

The phylogenetic species concept demands only that a species be the smallest monophyletic group that a systematist cares to name. That is, grouping is objective, but naming is arbitrary (Frost and Kluge, 1994; Alverson, 2008; Mishler and Theriot, 2000). Multiple strains of each of C. morus, C. reticulatum, C. astroideum, and C. pseudomicroporum are each supported as monophyletic with BS support of 100%.

However, phylogenetic structure was recovered within each of these clades, with plastome size and sequence contributing to diversity in all four clades (Figure 1 and Table 1). The clade of C. morus strains recovered two (SAG 2248 and SAG 41.86) with plastomes of over 500 kb and a third (SAG 2078) whose genome was almost 300 kb (compared to a median of about 180 kb). The two largest plastomes could represent one phylogenetic species and SAG 2078 could possibly represent a second. Similarly, there was one strain of C. reticulatum (UTEX 1365) with a plastome nearly 40 kb larger than that of C. reticulatum CG10, and nearly 48 kb larger than that of C. reticulatum SAG 8.81. Coelastrum asteroideum is represented by at least two distinct clades each receiving 100% BS support (strains CG6 and CW1; strains CS1, CS3, CS9, and SAG 33.88). The former two strains had nearly identical sequences, and also the largest plastomes among nominal C. asteroideum differing in length by only 56 bases between the two. However, the Swan Lake strains (CS1, CS3, and CS9) differed by about 3 kb, and SAG 33.88 was another 2 kb larger than the largest Swan Lake strain (6 kb smaller than the CG6 and CW1 strains). Plastomes of C. pseudomicroporum were recovered in two subclades (CLB and UTEX 1353; SAG 2077 and UTEX 280). Sister strains UTEX 1353 and CLB had the smallest and largest plastomes of C. pseudomicroporum, being nearly 13 kb difference in size.

Plastome size differences within these Coelastrum clades ranged from 100 bp differences to several hundred thousand bp differences. The biological meaning is unclear and there are no clear guidelines to be taken from the general literature on Viridiplantae plastome size and species diagnosis. For example, in vascular plants, species of pears are known to interbreed when plastome sizes differ in terms of only a few hundred bases (Kim et al., 2024). On the other hand, differences in plastome size of as much as 20 kb are considered intraspecific in the widely cultivated Peucedanum japonicum (Joh et al., 2025). We are confident that increasing strain sampling in green microalgae will find other such perplexing and interesting observations.

The non-monophyly of C. microporum strains presents a different set of issues. There are three distinct lineages. One strain (UTEX 281) is embedded within a clade that includes C. indicum (SAG 2363) and C. cambricum (UTEX 2446), while CF1 and SAG 2292 form a clade of strains with nearly identical sequences, as do CS5, CW5 and UTEX 1354. Again, we observe variation among nominal strains of C. microporum of as much as 20 kb. These inconsistencies may reflect simple taxonomic misidentification, incomplete lineage sorting, and/or historical hybridization events.

Two species mentioned several times in various studies are particularly in need of revision and probably need to be considered synonymous. Coelastrum sphaericum and C. proboscideum have been difficult to resolve both with morphology and genetic data. Komárek and Fott (1983) claimed that these species differ in apical structure, with the former being formed by small bumps, and the latter representing a crown-like structure. Others have considered the two species synonymous (Hajdu et al., 1976). Our results are consistent with the findings of Hegewald et al. (2010), who suggested a persistent phylogenetic signal despite differences in molecular markers used (e.g., nuclear ITS). Such a signal, interpreted as shared apomorphic similarity in morphology and DNA sequence, strongly supports the classification of C. sphaericum and C. proboscideum as a single phylogenetic species. We note again, however, that the largest plastome is more than 4 kb larger than the smallest. Thus, such inferences will be more powerful once broader strain sampling is done across both data sets.

In summary, phylogenetic resolution may be constrained by limited taxon sampling (Hillis, 1998; Pollock et al., 2002; Zwickl and Hillis, 2002; Heath et al., 2008), especially in cases where only a single strain represents a species. In such instances, intraspecific variability remains unassessed, which may mask deeper population structure or contribute to apparent paraphyly. Finding strains that are very similar in coding region sequence, with shared morphological characteristics, but very different plastome sizes, in all of our clades underscores the need to expand strain sampling across multiple populations per species to better understand species classification and diagnosis in these green algae.