Botryococcus braunii genome assemblies

High quality, high molecular weight genomic DNA [34, 35] was isolated from the A, B, and L races, and this DNA was used for sequencing on the Oxford Nanopore Technologies. This long-read genomic sequencing data was combined with previously obtained Illumina short-read sequencing for all three chemical races to combine the benefits of both sequencing technologies for downstream genome assembly. This produced 18-24x coverage (three samples multiplexed and sequenced on two flow cells) from the Nanopore data and 140-316x coverage from the Illumina data. Coverage was calculated based on total bases from sequencing divided by the estimated genome size for each race [16, 17]. For each race, de novo, reference-guided, and hybrid (combing Oxford Nanopore and Illumina sequences) assemblies were tested to determine which method produced an optimal draft genome assembly based on assembly metrics and BUSCO scores. For the L race, a de novo assembly approach using only the Oxford Nanopore sequence data produced the best assembly. For the A race, a hybrid assembly method, combining both Oxford Nanopore and Illumina sequence data, produced the most contiguous genome assembly. All attempts to produce an improved genome assembly for the B race resulted in sub-par final assemblies and were less contiguous compared to the current B race genome assembly [34].

The genome assembly statistics for three green microalgae with relatively complete assemblies, Chlamydomonas reinhardtii [36], Coccomyxa subellipsoidea [37], and Chlorella variabilis [38], are included in Table 1 as references to evaluate the contiguousness of the new B. braunii A and L race genome assemblies. Of the three B. braunii race genome assemblies, the L race was the most contiguous and whole with the largest N50 value and the smallest number of contigs/scaffolds, meaning it is less fragmented overall (Table 1). Even though the A race had a similar number of contigs/scaffolds as the B race genome assembly, its L50 was nearly half that of the B race assembly (Table 1), meaning the longest scaffolds assembled for the A race are longer and more contiguous than those of the B race. All three of the B. braunii race genome assemblies had a low percentage of fragmented BUSCOs while having varying amounts of single-copy, complete BUSCOs (Table 1). Overall, the new B. braunii A and L race genome assemblies are more contiguous than the current B race genome assembly. Interestingly the L race had the smallest assembly size of all three races at 135.6 Mb, and the A race had the largest at 188.3 Mb (Table 1). This contradicts earlier genome size estimates that showed the L race had the largest genome size at 211 Mb followed by the B race at 166 Mb and the A race at 160 Mb [16, 17]. Due to the difference in assembled genome sizes versus earlier estimations, all three race assemblies would benefit from additional sequencing including long-read and Hi-C sequencing in hopes of anchoring scaffolds to achieve chromosome resolution genome assemblies. This marks the first time that all three B. braunii races have had genomes assembled.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Assembly statistics for all three chemical races of B. braunii.

Assembly statistics for all three chemical races of B. braunii. A and L chemical races are novel in this study. The B. braunii B race, C. reinhardtii, C. variabilis, and C. subellipsoidea serve as a comparison to assess the contiguity of the A and L race genomes. Both A and L race assemblies are at scaffold resolution containing a mix of scaffolds and contigs. BUSCO scores were generated using the Chlorophyta ODBv10.

https://doi.org/10.1371/journal.pone.0304144.t001

Genome sizes vary among Trebouxiophyceae

Our early estimations of B. braunii genome sizes had placed the A, B, and L races among the largest known genomes for green microalgae at the time [16, 17]. Since these studies, more Chlorophyta species have been sequenced and genomes assembled, annotated, and deposited in public databases such as PhycoCosm [39]. This has allowed for a reexamination of B. braunii genome sizes and how they compare to other green microalgae.

For this comparison, we obtained genome size information from PhycoCosm for 14 species of green microalgae from the Trebouxiophyceae and Chlorophyceae clades as shown in S1 Table in S1 File. These species were chosen because of their close relation to B. braunii to serve as an in-clade comparison (Trebouxiophyceae) or as an outlier group for external clade comparison (Chlorophyceae). Additional criteria for comparison selection were datasets that included masked genomes, assembled transcriptomes, and annotated gene sets (gtf/gff3 formatted files). The comparison of the B. braunii genomes to these 14 species of green microalgae showed that the B. braunii genome assemblies are the largest among the Trebouxiophyceae species analyzed, but not the largest when including the Chlorophyceae species, with the Dunaliella salina genome [40] being the largest of those analyzed (S1 Table in S1 File). The closest known relative to B. braunii, C. subellipsoidea, had a substantially smaller (48.9Mb) genome assembly size [37] compared to any of the three B. braunii genome assemblies (Table 1 and S1 Table in S1 File). This large difference in genome assembly size would suggest that after divergence from a common ancestor either B. braunii experienced a genome size expansion event(s) or C. subellipsoidea experienced a genome size reduction event. The genome assembly sizes for the 9 species of Trebouxiophyceae (not including B. braunii) analyzed, are in the range of 14-60Mb (S1 Table in S1 File), suggesting B. braunii underwent a genome size expansion after delineation from a common ancestor shared with C. subellipsoidea.

It has been hypothesized that the cause of the large B. braunii genome sizes was a large number of genes encoded in the genomes of each race [16, 17]. This seemed to be confirmed when we published the first B. braunii genome from the B race [34], which had 20,765 genes. Now with the A and L race genomes assemblies, this finding can be reanalyzed by comparing genome size and gene count in several Trebouxiophyceae species and C. reinhardtii from Chlorophyceae. In general, an increase in genome size correlates with an increase in gene count for most Trebouxiophyceae species analyzed (Fig 1 and S1 Table in S1 File). It should be clarified that in this work the number of genes reported as encoded on the B race genome, 23,685, differs from the published value of 20,765 [34]. This is because the annotation software used to perform gene annotation in the A and L races was also applied to the B race genome (see Methods section for detailed description), due to this software’s increased capabilities over what was used when the B race genome was published. Genome size and gene count are not positively correlated among the three races of B. braunii. The B. braunii L race has a ~90Mb larger genome assembly size compared to C. variabilis and has ~6,000 more genes. (Fig 1 and S1 Table in S1 File). Similarly, B. braunii A race has a ~140Mb larger genome assembly size compared to C. variabilis and ~9,000 more genes. Within B. braunii, the A race, with the largest B. braunii genome assembly size, has ~5,200 fewer and ~2,600 more genes than the B and L races, respectively (Fig 1). Additionally, gene coding sequences in each race were analyzed for overlap with repeat content (see below for description of repeat content). This analysis revealed that only the A race contained gene/repeat overlaps with 3,189 genes having a significant overlap (>70%) to repeat elements (S1 File). Interestingly, Picochlorum renovo displays the smallest known Trebouxiophyceae genome size, yet has a remarkably large number of encoded genes given the small genome size (Fig 1) [41]. This compact genome may be a consequence of the fitness this species displays in thermo, salinity, and intense light tolerance [42–45]. Since gene count did not appear to be the sole factor accounting for the large B. braunii assembled genome sizes observed, a more detailed investigation into the genomes of B. braunii was carried out.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Comparison of gene number and genome size for select Chlorophyta species.

Grouped bar chart showing genome assembly size in blue and predicted gene count for each assembly in green. All algae except C. reinhardtii are Trebouxiophyceae. C. reinhardtii is shown as a reference for a similar size and gene count compared to the B. braunii races.

https://doi.org/10.1371/journal.pone.0304144.g001

Gene feature distribution of B. braunii

Having genome assemblies for the first time for each race of B. braunii allowed for direct comparison of gene feature distributions such as the length of 5’-UTRs, 3’-UTRs, and introns. These B. braunii gene features were compared to two Trebouxiophyceae species, C. variabilis and C. subellipsoidea, and the Chlorophyceae species C. reinhardtii. These species were chosen as a representative sample of different levels of relation to B. braunii given our phylogenomic analysis (see Fig 4): C. subellipsoidea as the closest relative, C. variabilis as a more distant relative, and C. reinhardtii as a distant relative form a different clade. The distribution of 5’- and 3’-UTRs was similar for each species analyzed with the largest number of UTR regions in each species being between ~100 and ~500 bp (Fig 2A and 2B). However, differences between species can be seen in the small number of large UTRs seen in each species. For example, among the three B. braunii races the B race has the longest 5’-UTRs and 3’-UTRs (Fig 2A and 2B). Interestingly, the closely related C. subellipsoidea had the smallest number of and shortest length of both types of UTRs (Fig 2A and 2B). For intron length distribution, the B race of B. braunii had a smaller number of and shorter length of introns compared to the A and L races (Fig 2C). This phenomenon could explain the large number of genes encoded in the B race genome assembly, but may also be an artifact of its current highly fragmented state.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Non-coding gene feature distribution analysis.

Box plots inside violin plots for (A) 5’-untranslated regions (UTRs), (B) 3’-untranslated regions (UTR)s, and (C) Intron lengths in the three races of B. braunii, C. subellipsoidea, C. variabilis, and C. reinhardtii. Box plots indicate length distribution with the median gene feature length indicated by a white dot inside the box plot. Violin plots show the frequency (width of the curve) of gene features at a given length (x-axis).

https://doi.org/10.1371/journal.pone.0304144.g002

DNA repeat content of B. braunii

The amount of DNA repeat content was examined to explain the relatively large genome sizes of the B. braunii races since a correlation between the number of putative genes and genome size was not seen (Fig 1). Transposable element (TE) and tandem repeat (TR; a.k.a. satellites and low-complexity repeats) content within the genomic space of all three B. braunii races were analyzed using RepeatModeler [46] and RepeatMasker [47] for TEs and the Tandem Repeats Finder program [48] for TRs. The B race exhibited the largest amount of total repeat content at 43% of the total genome size (Fig 3B), while the A and L races had less total repeat content at 39%, and 24%, respectively, of the total genome size (Fig 3A and 3C). The makeup of TEs in each race was also analyzed. All three races had more class I (retrotransposons) than class II (DNA transposons) TEs (Fig 3A–3C). Within the class I TEs, each race had a different major category with long interspersed nuclear elements (LINEs) dominating in the A race, long terminal repeats (LTRs) predominant in the B race, and LINEs and LTRs roughly equal in the L race (Fig 3A–3C). Short interspersed nuclear elements (SINEs) were only found in the B race as a minor component of class I TEs (Fig 3A–3C). Each race had different distributions and amounts of each subclass of class I and class II TEs (S1 File). The B race exhibited the most TR content at 22%, with the A and L chemical races at 16% and 11%, respectively (Fig 3A–3C).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. DNA repeat content of select Chlorophyta species.

Total repeat content, transposable elements (TE) plus tandem repeats (TR), and the breakdown of TE type was calculated for (A) B. braunii, race A, (B) B. braunii, race B, (C) B. braunii, race L, (D) C. incerta, (E) C. reinhardtii, (F) C. schloesseri, and (G) C. subellipsoidea. The pie chart represents the amount of the genome assembly for each microalga that is made up of repeat elements by showing the total class I TEs (magenta), class II TEs (salmon), unclassified TEs (blue), total TR (yellow), and non-repeat content (grey) for each genome. The bar graph shows the breakdown of class I TEs, SINEs (red), LINEs (green), and LTRs (light blue), and class II TEs, DNA transposons (orange), as a percent of the total TE content for each race. The distribution of all repeat elements, including class I and class II TE subclasses, for each species can be found in S1 File.

https://doi.org/10.1371/journal.pone.0304144.g003

To see how B. braunii total repeat content compared to other Chlorophyta species, the same DNA repeat analysis was performed on the closely related C. subellipsoidea and the three Chlamydomonas species. All of these species, except C. subellipsoidea, contained similar total repeat content compared to the three B. braunii races (Fig 3A–3G). While C. subellipsoidea contained little TE content, those TEs were made up of only class I LINEs and class II TEs (Fig 3G and S1 File). The three Chlamydomonas species contained class I LINEs as the main TE with varying amounts of LTRs and class II TEs (Fig 3D–3F). Like the observation that genome assembly size did not linearly correlate to the number genes annotated within each race of B. braunii (Fig 1), repeat content is also not linear with genome assembly size. For example, the A race has the largest genome assembly size of the three B. braunii races (Table 1), yet displayed 4% less total repeat content than the B race (Fig 3A and 3B). In contrast, C. schloesseri and C. Incerta have the largest genome size of the Chlamydomonas species (S1 Table in S1 File) and the most TE content (Fig 3D and 3E) analyzed here. Taken together, these data suggest TEs and TRs contribute to the observed large genome assembly sizes of the three races of B. braunii.

Comparative genomics analysis of the B. braunii races with other algal species

Phylogenomic comparison of multiple chlorophyta species.

Having genome assemblies of all three B. braunii races allowed for a comparative genomic analysis with other algal species. First, a phylogenetic comparison was carried out using a large dataset of genes conserved among 17 chosen algae species, including the three B. braunii races. By using a large dataset of conserved genes, more accurate phylogenetic relationships can be inferred through this type of phylogenomic analysis [49]. The species used for this analysis are listed in S1 Table in S1 File and were chosen based on having genomes, transcriptomes, and predicted gene datasets deposited and publicly available from PhycoCosm. Of the 17 species chosen, 12 were from Trebouxiophyceae (including the three B. braunii races) and five were from Chlorophyceae. To obtain the gene data set to be used in the phylogenomic analysis we utilized ChlorophytaODB v10 [50], which is a set of single-copy gene orthologs manually curated from 16 genomes of various Chlorophyta taxa included in the OrthoDB v10 database. At the time of this study, the ChlorophytaODB v10 had a total of 1,569 orthologous genes. A genome-wide evaluation of the 17 species revealed 1,189 of these genes were shared as single-copy orthologs while allowing any ortholog to be absent in up to three organisms.

Using these conserved, single-copy orthologs, a phylogenomic tree was constructed for the 17 species analyzed (Fig 4). This tree was constructed by first generating a maximum-likelihood (ML) tree for each of the 1,189 orthologous genes. Then all 1,189 ML trees were reconciled into a single phylogenomic tree that was rooted in the Chlorophyceae sub-clade containing Volvox carteri, the three Chlamydomonas species, and D. salina (Fig 4). Phylogenetic inference placed C. subellipsoidea, as the most common species branching from the same common ancestor as B. braunii (Fig 4). The B. braunii A race was the first to diverge from a common B. braunii ancestor with the B and L races diverging from each other at a later time point (Fig 4). Additionally, the A race of B. braunii exhibits more genetic divergence from the B and L races as compared to the genetic divergence among the three species of Chlamydomonas (Fig 4). This phylogenomic analysis led us to conclude at this point that, at minimum, the B. braunii A race is likely a separate species from the B. braunii B and L races.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Phylogenomics of chlorophyta.

Phylogenomic tree constructed using 1,189 conserved single-copy orthologs among 17 species from the Chlorophyta division. The Trebouxiophyceae class species including all three races of B. braunii are outlined in turquoise. The Chlorophyceae class species, used to root the tree, are outlined in olive green. Legend bar is in substitutions per site.

https://doi.org/10.1371/journal.pone.0304144.g004

Gene family analysis

The number of genes and gene families shared among the three races of B. braunii in comparison to separate but closely related species can be used to support defining the A race as a separate species and provide data to determine if the B and L races are separate species. Thus, gene relationships were inferred using OrthoFinder [51], a program package that reconstructs gene orthologs and paralogs based on sequence identity shared at both the inter- and intraspecies levels into groups that are called orthogroups (hereafter referred to as gene families). This analysis was carried out on the 17 species analyzed in the phylogenomic tree and found a combined total of 226,181 genes and 22,024 gene families (S2 Table in S1 File). Next, an analysis of gene family relationships between organisms of the same genus was performed. Since the phylogenomic analysis (Fig 4) suggests the A race is likely a separate species from the B and L races, we investigated the amount of A race unique gene families not shared with the B and L races, or if the B and L races shared more gene families. An UpSet plot [52] was generated to visualize gene family intersections between 9 species: 6 Trebouxiophyceae algae: C. subellipsoidea, C. variabilis, C. sorokinia, and the three B. braunii races, and three Chlorophyceae species: the three Chlamydomonas species C. reinhardtii, C. schloesseri, and C. incerta (Fig 5A and 5B). The top portion of the UpSet plot (Fig 5A) groups intersections of the reconstructed gene families represented by the accompanying connected dot matrix (Fig 5B). The three Chlamydomonas species together had the most genus-specific gene families at 4,641, and the next highest intersection of gene families was 3,029 families shared by all 9 algae (Fig 5A and 5B). While each race of B. braunii shared 847 gene families (red dots, Fig 5B), each race had a large number of gene families specific to each race (blue dots, Fig 5B). This was not seen for the three Chlamydomonas species. Interestingly, the B and L races shared 506 gene families (green dots, Fig 5B), while neither the B nor L shared a large (>200) number of gene families with the A race. This supports the relationships seen between the B. braunii races in the phylogenomic analysis (Fig 4) and suggests there are more genes, and as a consequence biosynthetic/metabolic pathways, that are similar or shared between the B and L races that the A race lacks.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Comparative gene family analysis.

UpSet plot showing the orthogroups or gene families identified by OrthoFinder grouped by species. (A) Bar chart showing the number of gene families that intersect with the given species in (B). (B) Shared gene families at different species intersections. Dots connect the species that contain the number of gene families shown in (A). Horizontal bar char in (B) corresponds to the total number of gene families with a given species. (C) and (D) Venn diagrams of shared gene families between the three Chlamydomonas species (C) and the three B. braunii races (D).

https://doi.org/10.1371/journal.pone.0304144.g005

It would be reasonable to expect a set of gene families shared by a common ancestor of all Trebouxiophyceae and have some portion of them retained in the divergent organisms. However, there was not a large number of gene families (>200) specific only to the Trebouxiophyceae algae (Fig 5A and 5B). Interestingly, the three races of B. braunii exhibited the largest amount of genus-specific gene families of the Trebouxiophyceae algae analyzed (red dots, Fig 5B).

The large difference in unique gene families identified in each B. braunii race led us to further investigate the gene families shared between the three races in comparison to the three closely related Chlamydomonas species using OrthoFinder [51] and weighted Venn diagrams (Fig 5C and 5D). There were a total of 12,903 gene families identified in the three Chlamydomonas species, jointly they shared 10,844, or ~83%, of these gene families, and the gene families shared between any two of the Chlamydomonas species ranged from 372 to 586 (Fig 5C). The three races of B. braunii contained 11,281 gene families, 6,151 of which, or ~55%, were shared among the three races, and the gene families shared between any two races ranging from 355 to 1,365 (Fig 5D). Additionally, there was a greater number of gene families specific to each race of B. braunii than specific to each Chlamydomonas species (Fig 5C and 5D). Taken together these data indicate there is a greater degree of diversity in the gene families between the three races of B. braunii than between the three Chlamydomonas species. These significant differences in the genic space between the three B. braunii races offer genome-wide evidence supporting each race of B. braunii is an individual species.

Gene duplications.

Commonly, some genes cannot be placed in an orthogroup by OrthoFinder when analyzing gene families. These genes are marked as unassigned and referred to as lineage-specific genes (LSGs) unique to a given species [53]. Over time an LSG may undergo gene duplication producing lineage-specific paralogs [54, 55], and are identified by OrthoFinder as a lineage-specific orthogroup (LSO). LSG and LSO analysis was carried out on the 17 algal species used in this study. In general, the three B. braunii races have more LSGs and LSOs than the other algae analyzed except Trebouxia sp. A1-2 and D. salina (Fig 6A and S2 Table in S1 File). Among the B. braunii races, the B race contained the largest number of LSGs with 5,160, while the A and L races had fewer at 2,160 and 1,660, respectively (Fig 6A). For the number of genes in LSOs, the A race had the most among the B. braunii races at 3,741 with the B and L races having 3,298 and 2,095, respectively (Fig 6A). On a percentage basis, each race of B. braunii had a high percentage of LSGs and LSOs compared to most of the other algae (Fig 6A). The genes comprising the LSOs and LSGs identified in each race of B. braunii were analyzed for the presence of annotated domains using the Pfam database. Only 17.4%, 15.3%, and 18.3% of the genes in the A, B, and L races, respectively, contained an annotated Pfam domain (data not shown).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Gene duplication event analysis in chlorophyta species.

(A) Distribution of the number and percentage of lineage-specific genes (LSGs) and lineage-specific orthogroups (LSOs) for each species listed. (B) Phylogenetic trees output for each single-copy ortholog from OrthoFinder were reconciled into a single tree with branch lengths mapped to gene gain/losses by Notung. At each evolutionary node break, the ratio of gene gain to loss from duplication events is labeled (e.g. 21.7:1). If the ratio indicates a net increase in gene gain it is colored in blue text. If the ratio indicates a net decrease in gene gain it is in red text. Ratios were calculated by dividing the total number of gains by losses for each branch. Tree branches represent the number of genes gained and loss from the total number of duplications calculated at each tree node. (C) Total gene duplication events per species inferred by OrthoFinder mapped as a horizontal bar chart.

https://doi.org/10.1371/journal.pone.0304144.g006

In comparison, the three closely related Chlamydomonas species had low and relatively equal LSGs and LSOs (Fig 6A), again indicating a large amount of genetic diversity between the three B. braunii races. Since no apparent whole genome duplication (WGD) events have been observed or inferred in B. braunii [56] it is likely the large number of LSGs and LSOs are the result of gene duplication events and subsequent functionalization [57–59]. This analysis also indicates that anywhere from 1 in 5 to 1 in 3 genes are unique to a given B. braunii race. In the Chlamydomonas only 1 in ~33 genes are unique to a given species between C. reinhardtii, C. incerta, and C. schloesseri. These findings further highlight the differences in the genomic space between the three B. braunii races that are not seen in the Chlamydomonas species.

Analysis of hydrocarbon biosynthesis gene distribution within the three races of B. braunii

Historically, it has been noted that the three races of B. braunii differ from each other mainly by the type of hydrocarbon produced by each race [7, 8, 62]. That is to say, the hydrocarbons of one race are not found in the other two races. It has been presumed this is due to each race having only the genes specific to the production of the hydrocarbons in that race. However, it is possible that hydrocarbon biosynthesis genes from one race exist in the other races but are not expressed. With the genome sequence for each race, the existence of these genes in each race of B. braunii can be analyzed. Thus, each B. braunii race was analyzed for the presence of known hydrocarbon biosynthesis genes from each race using HMMER searches with specific Pfam protein domains and BLAST searches for hydrocarbon genes from each race.

New Botryococcus species names

The B. braunii species was first named in 1849 by Friedrich Traugott Kützing [85] and was named after Alexander Karl Heinrich Braun, a 19^th century botanist [86]. The A and B races of B. braunii were classified in 1985 [62] and the L race in 1987 [8] based on the type of hydrocarbons produced in each race. Based on the data presented here we created new species names for the three races of B. braunii using the general rule for generating species names that the species name should match the gender of the genus name [87]. Botryococcus is considered as male [86]. Thus, we propose the following species names. The A race is renamed Botryococcus alkenealis. The general term for the A race hydrocarbons, alkene, is used with the Latin suffix -alis added, which is male or female and means pertaining to, i.e. pertaining to alkenes. The B race retains the name Botryococcus braunii to preserve the history of being the first to have its hydrocarbons identified, i.e. botryococcenes [88]. The L race is renamed Botryococcus lycopadienor. The root of the hydrocarbon name for the L race, lycopadiene, is used with the Latin suffix -or, which is a male suffix with no descriptive meaning. Hereafter, the three races of B. braunii will be referred to by these new species names and all researchers in the Botryococcus field are encouraged to adopt this new nomenclature.

Evidence in support of the chemical races as three different species

Culturing of B. braunii

Culture conditions for B. alkenealis, B. braunii, and B. lycopadienor were followed as described previously [18]. Briefly, B. alkenealis, (race A, Yamanaka strain [105]), B. braunii (race B, Showa strain [106]), and B. lycopadienor (race L, Songkla Nakarin strain [8]) were grown in modified Chu 13 medium under continuous aeration with filter sterilized air mixed at 2.5% CO₂. The composition of the modified Chu 13 medium was as follows: KNO₃ (0.4 g/L), MgSO₄ · 7H₂O (0.1 g/L), K₂HPO₄ (0.052 g/L), CaCl₂ · 2H₂O (0.054 g/L), FeNa EDTA (0.01 g/L), H₃BO₄ (2.86 mg/L), MnSO₄ · H₂O (1.54 mg/L), ZnSO₄ · 7H₂O (0.22 mg/L), CuSO₄ · 5H₂O (0.08 mg/L), NaMoO₄ · 2H₂O (0.06 mg/L), CoSO₄ · 7H₂O (0.09 mg/L). Cultures were grown under a light:dark cycle of 12:12h with a constant light intensity of 120 μE m^-2s^-1. Alga cells were subcultured by inoculation of 100 mL from a 6-week-old culture into 750mL of fresh medium.

DNA isolation and sequencing

Algal cells were harvested by vacuum filtration through a 5-μm or 10-μm nylon microsieve (BioDesign Inc. of New York) for B. lycopadienor and B. alkenealis/B. braunii, respectively. Cells were either used directly or snap-frozen in liquid nitrogen and stored at -80°C until further use. High molecular weight (HMW) genomic DNA (gDNA) free of RNA and polysaccharides, was isolated from harvested biomass using a protocol we developed specifically for Botryococcus species [35] that was adapted from previous studies [107, 108]. Briefly, harvested algae biomass was macerated by mortar and pestle in the presence of liquid nitrogen to a fine powder. Then 100-110mg of macerated biomass was resuspended in 1 mL of sorbitol wash buffer (100 mM Tris-HCl pH 8.0, 0.35 M sorbitol, 5 mM EDTA pH 8.0, 1% (w/v) polyvinylpyrrolidone (PVP-40)), sonicated at 30% power for 25 seconds (strong enough to homogenize the sample, but not to lyse the cells), centrifuged at 2,500 x g for five minutes at room temperature and the supernatant discarded. Due to the hydrocarbons in the sample, the pellet floats on top of the supernatant. Care was taken to remove the supernatant without disturbing the floating pellet. The washed pellet was then resuspended by vortexing for 5 seconds in 700 μL of pre-warmed (65°C), extraction buffer (100 mM Tris-HCl pH 8.0, 3 M NaCl, 3% CTAB-cetyl trimethylammonium bromide, 20 mM EDTA, 1% (w/v) Polyvinylpyrrolidone). Resuspended pellets were then incubated for 30 minutes in a 65°C water bath, mixed by inversion every 10 minutes, and allowed to cool to room temperature for 5 minutes. Then 700 μL of 24:1 (v:v) chloroform: isoamyl alcohol (CIA) was added, mixed by vortexing for 10 seconds, and then centrifuged at 2,500 x g for 10 minutes at room temperature. The aqueous phase was removed and treated with Rnase A (0.1 mg/mL) at 37°C for 15 minutes with mixing by inversion every 5 minutes, 500 μL of CIA added with mixing by inversion, centrifuged at 13,000 x g for 10 minutes at 4°C, and the aqueous phase was recovered and transferred to a fresh tube. HMW gDNA was then precipitated by the addition of 1/10 volume of 3M sodium acetate (pH 5.2), and 2/3 volumes of cold (-20°C) isopropanol, the sample was mixed by inverting 10 times, and incubated overnight at -20°C. The samples were then centrifuged at 13,000 x g for 10 minutes at room temperature, the supernatant discarded, the pellet air-dried, and the pellets washed with 1 mL of 70% ethanol followed by centrifugation at 13,000 x g for 10 minutes at 4°C. The HMW gDNA pellet was then dried under vacuum for 10 minutes, resuspended in 100 μL of nuclease-free water, and either stored at -80°C or prepped for sequencing.

Library preparation and sequencing for Illumina-based sequencing was performed on the Illumina HiSeq2000 platform by the Texas A&M AgriLife Genomics and Bioinformatics Service (TxGen). Oxford Nanopore-based sequencing was performed by the Sequencing Technologies and Analysis core facility at the Cold Spring Harbor Laboratory. For this sequencing, DNA concentration was determined via fluorometry and small/fragmented gDNA (<25kb) was removed using the Circulomics short read eliminator kit (SS-100-101-01). Samples were barcoded and adaptors ligated with the native barcode kit (Oxford Nanopore Technologies, EXP-NBD104). Sequencing was performed on the Oxford Nanopore PromethION platform, using R9.4.1 flowcell biochemistry and base calling was done with ONT’s proprietary Guppy (v4). In total, two flow cells were run due to complications with the quality of sequencing data on the first flow cell, the lower quality flow cell run will be referred to as flow cell 1 and the higher quality flow cell referred to as flow cell 2. Both flow cells’ data were used in the final genome assemblies for B. alkenealis and B. lycopadienor given recovery of read quality after read trimming. Both sets of reads from each flow cell were subjected to the same pre-assembly trimming process, however, flow cell 1 had sufficiently fewer reads post-trim than flow cell 2 reflecting the overall lower quality of the reads obtained.

De novo genome assembly for B. lycopadienor (L race)

The raw fastq files from the Nanopore sequences were trimmed with Porechop (v0.2.4) to remove any remaining non-gDNA sequences and poor-quality bases. The L race genome assembly was performed using the Flye long-read assembler [109] with the trimmed long-read sequences as input. The resulting scaffolds and contigs were then polished with the short-read Illumina DNA sequencing data that was trimmed for barcode, adapter, and poor quality sequence using Trimmomatic [110]. Scaffold polishing was done iteratively twice with Pilon [111], with the output of the first round of polishing used as the input for the next round of polishing. Scaffolds/contigs belonging to symbiotic or organelle DNA [112] were identified and removed using Blobtools (v1.1) [113]. Briefly, scaffolds/contigs that displayed GC% and k-mer frequency and were annotated taxonomically as non-green microalgae were individually reviewed for confirmation before removal. The nuclear draft genome was then masked using a combination of RepeatModeler (v1.0) [46] and RepeatMasker (v4.0) [47]. First, a de novo repeat library was assembled using RepeatModeler. The repeat library was then used to mask the genome with RepeatMasker using B. braunii B race CDS [34] (JGI: Bbraunii_502_v2.1.cds.fa) as trusted sequences to prevent over-masking by RepeatMasker. Finally, assembly metrics and quality were assessed using Quast [114] and BUSCO v5 [115], respectively. ChlorophytaODBv10 [50] was used as the ortholog database for BUSCO analysis.

Hybrid de novo genome assembly for B. alkenealis (A race)

The genome assembly pipeline used for B. lycopadienor was applied to B. alkenealis but did not result in a high-quality assembly. Thus, a modified approach was taken to produce a higher quality assembly for B. alkenealis to maximize the final N50, minimize the final L50, and reduce the number of “N”’s in the unmasked assembly. Trimming of both the Nanopore and Illumina sequencing data along with contig/scaffold assembly of the long read sequences by Flye remained the same as was done for B. lycopadienor. The hybrid assembler MaSuRCA (v4.0.1) [116] was used to generate a hybrid genome assembly from both the long and short read sequences as input. Then both the Flye assembly and MaSuRCA assembly were merged and de-replicated using Quickmerge [117]. The resulting merged assembly was then polished twice iteratively with the short-read sequence using Pilon. Removal of non-nuclear DNA, masking, and quality/metric assessment was done as described for the B. lycopadienor.

Attempts at genome assembly for the B. braunii (B race)

Attempts were made to improve the current B. braunii (B race) genome assembly [34]. The newly generated Nanopore sequencing data were used in identical workflows that were used for B. lycopadienor and B. alkenealis genome assemblies mentioned above. Unfortunately, neither resulted in an improved assembly. While N50 values were slightly higher than the current metrics, both contig number and L50 values were larger than their respective values in the current B. braunii (B race) genome assembly. BUSCO analysis also showed a higher amount of fragmented orthologs than the current assembly. An additional 10 assembly approaches were tested including a reference-guided genome assembly approach [118] using the current assembly as “trusted” contigs for inputs into MaSuRCA, Flye, and Canu assemblers. None of these advanced assembly tactics yielded an improved genome assembly for B. braunii. Likely, the Nanopore sequencing data did not contain enough sequencing depth to resolve the complexities of the B. braunii genome [34]. Future attempts at genome sequencing and assembly into B. braunii will need more long-read sequencing data and optical mapping to resolve complex, repeat-rich portions of the genome.

RNA isolation and sequencing

B. alkenealis and B. lycopadienor cultures were grown and biomass was harvested as described above. Total RNA was isolated and purified using the TRIzol Reagent (Thermofisher). The resulting RNA pellet was washed with 75% ethanol and then dried using a speed vac. To remove the polysaccharides from the samples, the RNA pellet was resuspended in 2M LiCl and centrifuged at 10,000 x g at room temperature for 10 minutes. The supernatant containing polysaccharides was discarded and the pellet retained. This process was repeated until the RNA pellet size remained the same after each wash step. The washed pellet was then resuspended in 1x TE buffer and extracted against and an equal volume of 1:1:1 phenol/chloroform/isoamyl alcohol. The suspension was then centrifuged at 10,000 x g at room temperature for 10 minutes with the pellet discarded and the supernatant saved. A final extraction against an equal volume of chloroform was done followed by centrifugation at 10,000 x g and the supernatant saved. Finally, the RNA was precipitated with 0.1 volumes of 3M sodium acetate and 2.5 volumes of 100% ethanol. The solution was centrifuged at 10,000 x g for 10 minutes at room temperature and the pellet was saved. The RNA pellet was resuspended in Rnase-free water and treated with Dnase (1 unit/μg RNA), incubated at 30°C for 30 min, treated with 1:1:1 phenol/chloroform/isoamyl alcohol, precipitated with sodium acetate and ethanol as described earlier, the DNA-free final RNA pellet was resuspended in nuclease-free water, and stored at -80°C until further use. RNA pellets were sent for sequencing through the National Alliance for Advanced Biofuels and Bioproducts (NAABB) to the Los Alamos National Laboratory for sequencing on the Illumina platform [119, 120].

Genome-guided transcript assembly for B. alkenealis and B. lycopadienor

RNA-seq data were trimmed of adapters and low-quality bases using Trimmomatic. Trimmed reads were then mapped to the corresponding genome assemblies using HISAT2 [121] with default settings. The resulting BAM file was indexed and fed as input for use in the Trinity suite [122] run in genome-guided mode. The Trinity raw assembled transcripts were then filtered for low-expression isoforms using Trinity’s built-in scripts, and the removal of redundant transcripts was done with CD-HIT-EST [123]. Final draft transcriptomes were assessed for quality and metrics by Quast and BUSCO, respectively, run in transcript mode. ChlorophytaODBv10 was used as the ortholog database for BUSCO analysis.

Annotation of genome features

Genes were predicted and annotated using the BRAKER2 pipeline [124–133]. Briefly, BRAKER2 was run in “evidence-based” mode where RNA-seq data was first mapped to a hard-masked version of the genome assemblies using HISAT2 [121]. The resulting BAM file was used as input into BRAKER2 with default parameters and the output was saved in GTF format for all three Botryococcus species. BRAKER2 was also used for the untranslated region (UTR) analysis using the GUSHR module. At the time of analysis, BRAKER2 did not support automatic intronic region annotation. So, intronic sequences were detected and annotated using a custom python script with the BRAKER2 GTF output files. Due to BRKAER2’s increased capability of fully automatic training of the gene prediction toolsets and leveraging RNA-seq and protein homology information into the final gene prediction sets, the decision was made to re-annotate the B. braunii (B race) v2 genome assembly, using this increased capability that was not available at the time of the original release of the v2 assembly. This re-annotation gave a different final number of genes when compared to the v2 genome assembly files. This new annotation predicted 23,685 genes while the v2 genome annotation predicted 20,765 genes as seen in Fig 1. Transposable element content was annotated using RepeatModeler [46], and RepeatMasker [47]. DNA tandem repeats (satellites) were annotated by Tandem Repeats Finder [48] using default parameters. In all three Botryococcus species, genes were scanned for coding sequence overlaps with TEs at an overlap threshold of ≥70%. This was achieved by converting the RepeatMasker output file and BRAKER2 GFF/GTF output files to the BED format. BEDTools [134] was then used to identify overlapping regions using the “intersect” program with the “-f 0.70” flag.

Phylogenomic analysis

The approach for a mass phylogenomic analysis using multiple microalgae species was modified slightly from a previous study [91]. Briefly, protein sequence files for each species analyzed in this study (B. alkenealis, B. braunii, B. lycapdienor, Asterchloris glomerata, Auxenochlorella protothecoides, Chlamydomonas incerta, Chlamydomonas reinhardtii, Chlamydomonas schloesseri, Chlorella sorokiniana, Chlorella variabilis, Coccomyxa subellipsoidea, Dunaliella salina, Micractinium conductrix, Picochlorum renovo, Picochlorum soloecismus, Trebouxia sp. A1-2, Volvox carteri) were screened by BUSCO in protein mode using the ChlorophytaODBv10 database [50]. Single-copy orthologous proteins that were conserved in at least 13 of the 17 species were extracted, counted, and analyzed. Multiple sequence alignments (MSAs) of each ortholog set were produced using MAFFT v7 [135]. Maximum likelihood trees were inferred for each ortholog set from the corresponding MSAs using IQTREE [136] run with the following parameters: `-m MFP -bb 1000 -T 5`. The final set of 1,189 ortholog trees was reconciled into a single alternative species tree using ASTRAL-III [137]. The final phylogenomic tree was rooted in the Chlorophyceae sub-clade.

Comparative genomic analysis

Protein sequence files for B. alkenealis, B. braunii, and B. lycopadienor were generated using BRAKER2 as described above. The remaining 14 species (A. glomerata, A. protothecoides, C. incerta, C. reinhardtii, C. schloesseri, C. sorokiniana, C. variabilis, C. subellipsoidea, D. salina, M. conductrix, P. renovo, P. soloecismus, Trebouxia sp. A1-2, V. carteri) used in this study were downloaded from JGI’s PhycoCosm [91, 138–143]. Comparative genomic analysis was initiated by first inferring gene family relationships within and between the different species using OrthoFinder [51].

To visualize the distribution of gene families among the 17 species, a custom script was written to parse the assigned orthogroups from OrthoFinder and visualize the number and intersection of orthogroups by generating an UpSet plot. Similarly, the same script also generated weighted Venn diagrams for both the three Botryococcus species and the three Chlamydomonas species.

Gene gain and gene loss analysis was performed as previously described [61] with slight modifications. Briefly, the gene trees for each orthogroup from the OrthoFinder analysis were reconciled using Notung [60] to determine gene loss and gene gain over an evolutionary period. A custom script was written to extract these losses and gains at each node of the reconciled tree and mapped to the gene duplication species tree from the OrthoFinder analysis. The total number of organism gene duplication events was also obtained from the gene duplication specie trees.

Identification of hydrocarbon genes within each Botryococcus species

Fatty acid elongase gene analysis was performed by by first obtaining FAE (Pfam: PF08392) and ELO (Pfam: PF01151) hidden-markov models (HMM) from Pfam [144, 145]. These HMM profiles were then used to search for homologous sequences using the HMMER suite [146] with an E-value cutoff of 10^-3. Sequences that had significant homology to the FAE and ELO profiles in each Botryococcus species were then aligned using MAFFT and phylogeny was inferred with IQTREE using the same parameters mentioned above. KCR and HCD genes do not currently have a Pfam domain entry. Instead, single-copy protein sequences from A. thaliana were obtained from NCBI (NP_564905.1, and NP_193180.1 respectively) and used as query sequences in a blastp analysis of the Botryococcus species’ proteomes. An e-value cutoff of 10^-3 was used. The Pfam ECR hmm domain (Pfam: PF12241) was used in a similar approach as the FAE and ELO searches. Due to the single-copy nature of KCR, HCD, and ECR genes, phylogenetic trees were not inferred. Squalene synthase (SS) and squalene synthase-like (SSL) gene analysis in the three Botryococcus species was performed in the same manner as the fatty acid elongases using the SQS-PYS HMM profile (Pfam: PF00494).

Identification of meiosis genes within each Botryococcus species

Potential meiosis genes were identified based on sequence homology to meiosis genes identified in C. subellipsoidea and C. variabilis [79]. Briefly, DMC1, HOP1, HOP2, MER3, MND1, MSH4, MSH5, REC8, and SPO11 protein targets were identified in the BRAKER2 protein annotation sets by protein-protein BLAST+ homology alignment. Targets identified by alignment over an E-value threshold of 0.001, were then searched for Pfam domains using InterPro [144] against the Pfam database. In both homology searches, only the top alignment (smallest E-value) is reported. C. subellipsoidea protein sequences used in the homology searches were HOP1 (XP_005651810), HOP2 (XP_005643862), MER3 (XP_005651102, XP_005649357, XP_005649238), MND1 (XP_005649122), MSH4 (XP_005646103), MSH5 (XP_005644118), REC8 (XP_005651890), and SPO11 (XP_005647003). The C. variabilis meiosis gene used in the homology searches was DMC1 (XP_005848077).