Seventeen culture strains of the Ceramiales were obtained from the National Marine Biodiversity Institute of Korea (MABIK). The strains were maintained in the laboratory using L1 enriched seawater medium at 20°C with a 14 : 10 light / dark cycle. Culture tissue (~30 mg) was frozen in liquid nitrogen and ground using Automill TK-AM5 (Tokken Inc., Chiba, Japan). Total genomic DNA was extracted with a modified cetyltrimethylammonium bromide (CTAB) protocol (Doyle and Doyle 1987). The preheated (65°C) CTAB mixture (CTAB 750 μL, 1% 2-mercaptoethanol 8 μL, and 20 mg mL⁻¹ proteinase K 50 μL) was added to ground tissue and incubated for 10 min at 65°C, followed by a 5 min incubation at room temperature with RNase A (1% v/v). Phenol : chloroform : isoamyl alcohol (25 : 24 : 1) was added in volume equal to sample and centrifuged for 10 min at 14,000 rpm. The supernatant was transferred to a new tube, mixed with 600 μL of chloroform and centrifuged under the same condition. DNA precipitation was achieved by adding 500–600 μL of pre-cooled isopropanol and a volume of 3 M sodium acetate equal to 1/10 of the suspension, followed by incubation at –20°C for 30 min. The supernatant was decanted after centrifugation at 14,000 rpm for 20 min at 4°C. The pellet was washed twice with 1 mL of 70% ethanol, each time inverted and centrifuged at 14,000 rpm for 15 min at 4°C. The final pellet was air-dried and suspended in 50–100 μL of D.W. The quantity and quality of DNA were assessed using the NanoDrop (NABI micro digital, Seongnam, Korea) and Qubit 2.0 fluorometer (Thermo Scientific, Waltham, MA, USA). The concentration of DNA varied in the range of 8–180.8 ng μL⁻¹. The 260/280 absorbance ratios were within the optimal range of 1.9 to 2.1. However, the 260/230 absorbance ratios showed a wider range of 0.97–2.32.

Whole genome sequencing was conducted using Illumina NovaSeq 6000 (Illumina, San Diego, CA, USA) by DNA-link Inc. (Seoul, Korea). The pair-end sequencing library was prepared using the TruSeq Nano DNA Prep Kit (Illumina), which generated 10 Gb for each taxon. We obtained additional previously generated Illumina read data for Chondria armata (Kützing) Okamura (SRR15927350) and Digenea simplex (Wulfen) C.Agardh (SRR8325636), which were retrieved from the NCBI Sequence Read Archive (SRA). The adapter and low-quality reads were removed using Trimmomatic v0.39 (Bolger et al. 2014) with the following parameters: ILLUMINACLIP: TruSeq3-PE.fa: 2: 30: 10:2:True LEADING: 3 TRAILING: 3 MINLEN: 36.

The mitogenome was assembled using GetOrganelle v1.7.7.0 (Jin et al. 2020) and NOVOPlasty v.4.3.1 (Dierckxsens et al. 2017). Genome assemblies obtained with GetOrganelle were manually inspected using Bandage (Wick et al. 2015). The Illumina reads were mapped to the assembled mitogenomes using Bowtie2 (Langmead and Salzberg 2012) for validation and error correction. Primary annotation was performed automatically using MFannot (Beck and Lang 2010). Transfer RNA (tRNA) genes were verified using tRNAscan-SE v.2.0 (Lowe and Chan 2016). Other possible open reading frames in the intergenic region were searched by using BLASTx. All the predicted genes were confirmed manually by using Geneious Prime 2020.2.4 (https://www.geneious.com) and BLASTp (e-value, 1e-05). Inverted repeats were identified using Palindrome analyzer (Brázda et al. 2016). Palindrome analyzer describes the features of palindromic and inverted repeats through similarity, localization, and visualization. Long repeat sequences were analyzed using the Repeat Finder in Geneious Prime with minimum length of 20 bp and 5% of mismatch. The mitochondrial genome sequences have been deposited at GenBank under the accession numbers OR885323–OR885339, BK064963, and BK064964.

We collected a total of 42 mitogenomes, including 19 newly generated in this study and 23 from previously published genomes (Table 1). Rhodymenia pseudopalmata (J.V.Lamouroux) P.C.Silva (KC875852) and Schimmelmannia schousboei (J.Agardh) J.Agardh (KJ398162) were selected as outgroup. Nucleotide sequences of 23 protein-coding genes (PCGs) were aligned with MACSE v2 (Ranwez et al. 2018) and manually adjusted. The ambiguous region was trimmed with GBlocks v.0.19b (Castresana 2000) under a 50% gap threshold. Maximum likelihood analysis was performed using IQ-Tree2 v2.0.6 (Minh et al. 2020) with 1,000 ultrafast bootstrap replications under the GTR + F + I + R5 model. Phylogenetic trees were edited and visualized using FigTree v.1.4.2 (http://tree.bio.ed.ac.uk/software/figtree).

Nineteen newly assembled mitogenomes had an average coverage between 500× and 1,000×. We observed a general decrease in coverage, often to 300×, in palindromic regions. However, the lowest coverage in these regions was >100×, maintaining the integrity and reliability of our data (Supplementary Figs S1 & S2). Genome sizes and gene contents varied at the family and species level, ranging from 24,508 bp in Campylaephora sungminbooi (J.R.Hughey & G.H.Boo) Barros-Barreto & Maggs (KU145004) to 36,710 bp in Pseudoceramium tenerrimum (G.Martens) Barros-Barreto & Maggs (MABIK-LAS-310); the GC content was 25.0 ± 2.4% (Table 2). Overall, the mitogenomes contained 43–68 genes, consisting of 21–36 PCGs, 18–33 tRNAs, and 2–4 rRNA subunits (Table 2). A core set of 23 PCGs were shared in all families, however, up to 36 PCGs were found in the family Ceramiaceae (Table 2, Fig. 1). Most of the Ceramiales species contained two copies of the ribosomal RNA operon, while, depending on the number of copies of the 5S rRNA, three to four rRNAs were found in five species: Antithamnionella miharai (Tokida) Itono (4 rRNAs), Dasysiphonia japonica (Yendo) S.-H.Kim (3 rRNAs), Bostrychia moritziana (Sonder ex Kützing) J.Agardh (3 rRNAs), Pterosiphonia complanata (Clemente) Falkenberg (3 rRNAs), and Symphyocladiella dendroidea (Montagne) Bustamante, B.Y.Won, S.C.Lindstrom & T.O.Cho (3 rRNAs). All 19 newly analyzed mitogenomes had a group II intron between the nad5 and nad4 genes, and Pseudoceramium tenerrimum contain two introns (Table 2).

Phylogenomic analysis resolved the relationships of families in the Ceramiales using a concatenated dataset of 23 PCGs (16,263 bp in alignment), excluding duplicated PCGs (Fig. 2A). The Ceramiales was monophyletic, including four families, which were fully supported. The Rhodomelaceae and Delesseriaceae formed a monophyletic clade, which was a sister to the Wrangeliaceae. The Ceramiaceae diverged first within the Celamiales, which was far related from the remaining families.

We categorized four types of mitogenome architecture in the Ceramiales (Figs 2B & 3). Type A consisted of 23 PCGs, with 10 PCGs in the forward direction and 13 PCGs of the atp6-nad6 region in reverse direction between the cob and tatC genes. Type A was found throughout the four families and the outgroup species, and here we refer to this as the putative ancestral type.

Type B mitogenomes were characterized by having inverse transpositions (Fig. 3A), as found in Dasya binghamiae A.J.K.Millar and Dasysiphonia japonica. The inverse transpositions occurred in three sets of paired-genes: rps3-rpl16, sdh2-sdh3, tatC-rps12. A deletion of sdh3 gene was found in Dasya binghamiae only.

Type C mitogenomes had inverted duplications by repeats, where each pair of duplicated segments revealed a palindromic structure (Fig. 3B). The duplication was centered on either the 3′ end of the cob or nad6 genes, and the number of duplicated genes varied among species. Type C occurred in four species of the tribe Ceramieae: Campylaephora kondoi (Yendo) Barros-Barreto & Maggs, Campylaeophora sungminbooi, Ceramothamnion sp. MABIK-LAS-337, and Pseudoceramium tenerrimum. The duplication began with the cob gene in Ca. kondoi (cob-cox3) and Ca. sungminbooi (cob-rps3). Pseudoceramium tenerrimum and Ceramothamnion sp. MABIK-LAS-337 displayed inverted duplications that started from cob and nad6 genes, respectively. The nad6-nad3 region was duplicated in Ceramothamnion sp. A similar but longer duplication observed in Pseudoceramium tenerrimum, which spanned from nad6 to atp6.

Type D mitogenome had both an inversion and a duplication (Fig. 3B), and this type was found in the tribe Antithamnieae (Antithamnionella miharai). Specifically, there was an inversion of the genomic segment from nad6 to rps12 genes and duplications occurred for the rrn5-rps11 and nad1 genes.