Skimming genomes for systematics and DNA barcodes of corals

Abstract

Numerous genomic methods developed over the past two decades have enabled the discovery and extraction of orthologous loci to help resolve phylogenetic relationships across various taxa and scales. Genome skimming (or low‐coverage genome sequencing) is a promising method to not only extract high‐copy loci but also 100s to 1000s of phylogenetically informative nuclear loci (e.g., ultraconserved elements [UCEs] and exons) from contemporary and museum samples. The subphylum Anthozoa, including important ecosystem engineers (e.g., stony corals, black corals, anemones, and octocorals) in the marine environment, is in critical need of phylogenetic resolution and thus might benefit from a genome‐skimming approach. We conducted genome skimming on 242 anthozoan corals collected from 1886 to 2022. Using existing target‐capture baitsets, we bioinformatically obtained UCEs and exons from the genome‐skimming data and incorporated them with data from previously published target‐capture studies. The mean number of UCE and exon loci extracted from the genome skimming data was 1837 ± 662 SD for octocorals and 1379 ± 476 SD loci for hexacorals. Phylogenetic relationships were well resolved within each class. A mean of 1422 ± 720 loci was obtained from the historical specimens, with 1253 loci recovered from the oldest specimen collected in 1886. We also obtained partial to whole mitogenomes and nuclear rRNA genes from >95% of samples. Bioinformatically pulling UCEs, exons, mitochondrial genomes, and nuclear rRNA genes from genome skimming data is a viable and low‐cost option for phylogenetic studies. This approach can be used to review and support taxonomic revisions and reconstruct evolutionary histories, including historical museum and type specimens.

Bioinformatically pulling UCEs, exons, mitochondrial genomes, and nuclear rRNA genes from genome skimming is a viable and low‐cost option for phylogenetic studies. The mean number of UCE and exon loci extracted from the genome skimming data was 1837 ± 662 SD for octocorals and 1379 ± 476 loci for hexacorals; phylogenetic relationships were well resolved within each class. We also obtained loci from historical specimens, with 1253 loci recovered from the oldest specimen collected in 1886.

1. INTRODUCTION

The advent of novel genomic methods and analyses has revolutionized our ability to resolve phylogenetic relationships across the tree of life. Numerous genomic methods [e.g., whole‐genome sequencing (e.g., Laumer et al., 2019), transcriptomics (e.g., Whelan et al., 2017) restriction‐site associated sequencing (e.g., Herrera & Shank, 2016), target‐capture (e.g., McFadden et al., 2021)] developed over the past two decades have enabled the discovery and extraction of orthologous loci across multiple phyla. While high‐quality whole genomes or transcriptomes are ideal in many situations, obtaining this genetic information from most animal taxa is often not feasible due to, for example, high costs and the ability to obtain high quality and/or quantity of DNA/RNA from specimens. But over the past decade, the average cost of high‐throughput sequencing has rapidly decreased (Park & Kim, 2016). Now, we can multiplex many more taxa and obtain more genomic data (i.e., base pairs) per sample at a much lower cost than ever before. Therefore, genome skimming, or low‐coverage whole‐genome sequencing (i.e, low read depth resulting in highly fragmented and gapped assemblies), could be used to readily obtain enough orthologous loci, including conventional DNA barcodes, at a relatively low cost for phylogenomic studies (Liu et al., 2021; Trevisan et al., 2019).

Genome skimming has been used in prior studies to obtain whole mitochondrial genomes and nuclear DNA loci for phylogenetic studies (e.g., Malé et al., 2014; Liu et al., 2021; Golightly et al., 2022; Taite et al., 2023). In addition, genome skimming has increasingly been used to help build DNA barcode reference databases for applications such as environmental DNA (eDNA) sequencing (Hoban et al., 2022; Zeng et al., 2018; Zhang et al., 2023). This method's potential, however, for other applications remains unrealized, as typically more than 99% of the sequence data produced by skimming is not used (Bohmann et al., 2020). Low‐coverage genome skims could readily be used to bioinformatically pull out ultraconserved elements (UCEs), exons, and other genes of interest. And because this method does not necessarily need high‐quality DNA as other methods (i.e., RAD Sequencing), genome skimming might be useful for historical samples that are housed in natural history museums across the globe (see Bakker, 2017; Hoban et al., 2022; Liu et al., 2021; Tin et al., 2014; Yeates et al., 2016). Thus, this method should be more thoroughly explored for various applications across different qualities and quantities of genomic DNA.

Phylogenomic studies of marine invertebrates might benefit from a genome‐skimming approach. In particular, the subphylum Anthozoa (sensu McFadden et al., 2022; phylum Cnidaria) is in critical need of taxonomic revision and resolution across family, genus, and species levels. Taxonomic revision along with new species descriptions are essential to help classify taxa and improve estimates of species diversity and distribution. Anthozoans are a diverse group of marine invertebrates, including sea anemones and corals, which are essential in building marine ecosystems from polar to tropical regions and the coasts to the abyss. Anthozoans currently comprise ~7500 valid species (Daly et al., 2007) in two classes (Hexacorallia and Octocorallia, McFadden et al., 2022), but this number might be grossly underestimated (Bridge et al., 2023; Plaisance et al., 2011). Recently (i.e., in the past 5 years), the number of phylogenomic studies on anthozoans has grown rapidly. These studies have used a variety of methods, such as restriction‐site associated sequencing (RADSeq, Arrigoni et al., 2020; Herrera & Shank, 2016; Quattrini et al., 2019; Reitzel et al., 2013), transcriptomics (Zapata et al., 2015), and target‐capture genomics (e.g., Bridge et al., 2023; Glon et al., 2021; McFadden et al., 2021, 2022; Quattrini et al., 2020; Untiedt et al., 2021) to resolve questions at a range of scales. Target‐capture of UCEs and exons, in particular, has shown promise in resolving phylogenetic relationships of anthozoans across deep (i.e., orders, McFadden et al., 2021, 2022; Quattrini et al., 2020) to shallow (i.e., closely related species, Bridge et al., 2023; Erickson et al., 2021; Glon et al., 2023) time scales.

The original Anthozoa baitset targeting UCEs and exons was designed by Quattrini et al. (2018) and redesigned by Erickson et al. (2021) for Octocorallia and Cowman et al. (2020) for Hexacorallia. These baitsets target 1000s of loci, but do not include baits for mitochondrial genes or the nuclear ribosomal RNA (rRNA) genes. Although using mitochondrial genes and rRNA genes for phylogenomic studies of Anthozoa is cautioned (Figueroa & Baco, 2015; Herrera & Shank, 2016; Quattrini et al., 2023), the utility of these markers goes beyond phylogenomic analyses. For example, mitogenome evolution across Anthozoa is intriguing as they exhibit a range of properties unique among metazoans, including gene order rearrangements (Brockman & McFadden, 2012; Figueroa & Baco, 2015; Lin et al., 2014; Seiblitz et al., 2022), a mismatch repair enzyme in Octocorallia (mtMutS, Bilewitch & Degnan, 2011), gene introns in the Hexacorallia (e.g., a homing endonuclease, Barrett et al., 2020; Fukami et al., 2007), and bipartite mitogenomes (Hogan et al., 2019). In some cases, mitogenomes have been used as taxonomic characters, as certain mitochondrial gene orders appear to be restricted to certain families (see Seiblitz et al., 2022). Finally, with emerging efforts to monitor coral ecosystems with environmental DNA, there is a need to increase the number of taxa and loci in reference databases (McCartin et al., 2023). Because genome skimming enables the production of low‐coverage yet highly fragmented genomes, this method, followed by bioinformatic analyses, holds promise in obtaining whole mitogenomes, nuclear rRNA genes, UCEs and exons, and other genes of interest from a range of DNA sample types (i.e., contemporary to historical samples) for a relatively low cost.

Here, we tested the utility of using genome‐skimming data to bioinformatically obtain whole mitogenomes, nuclear rRNA genes, UCEs, and exons from hexacorals (mostly black corals) and octocorals (Figure 1). Although most of our efforts were focused on recently collected (<20 years) specimens preserved specifically for genetic purposes, we also tested the utility of this approach to obtain UCEs, exons, and mitogenomes from historical material collected more than 100 years ago.

FIGURE 1.

Images of anthozoan corals. (a) Eunicea flexuosa, Florida Keys, (b) Gorgonia ventalina, Florida Keys, (c) Swiftia exserta, Gulf of Mexico, (d) Callogorgia lucaya, Caribbean Sea, (e) Aphanipathes puertoricoensis, Caribbean Sea, and (f) Aphanipathes sp., Caribbean Sea.

2. MATERIALS AND METHODS

2.1. Collections

Octocorals (n = 177) and hexacorals (n = 32, including 30 antipatharians or black corals, one scleractinian [Javania], and one zoanthid [Umimayanthus]) were collected from the Gulf of Mexico, Caribbean Sea, and off the southeastern US coast from 2006 to 2019 on various expeditions. Specimens were collected with both Remotely Operated Vehicles (ROV) and SCUBA. Tissue samples were taken in the field, preserved in 95% ethanol and stored at −20°C, or flash frozen in liquid nitrogen and stored at −80°C. We also added historical, cataloged octocorals (n = 33) collected from 1886 to 2006 from locations worldwide. Most museum specimens were either stored dry or in 70‐95% EtOH. See Appendix S1 for further details.

2.2. Molecular lab work

DNA was extracted in various ways (Appendix S1). Contemporary samples were extracted with either a modified CTAB protocol (Appendix S2), a salting‐out protocol (Herrera, 2022), a GeneJet Genomic DNA Purification kit, or a Qiagen DNEasy extraction kit. Historical samples were all extracted with a Qiagen DNEasy kit. For some antipatharians and octocorals, DNA was cleaned with a Qiagen Power Clean Pro kit to remove PCR inhibitors (see Table S1). DNA was quantified with a fluorometer, either with a Quant‐iT or with a Qubit.

For most samples (204 of 242), library preparation was carried out in the Laboratories of Analytical Biology at the Smithsonian Institution. The quantity of genomic DNA input into a library preparation ranged from <0.65 to 93 ng total DNA; the average was 55 ± 15 (SD) ng DNA. Library preparation was carried out using the NEBNext Ultra II FS DNA Library Prep Kit for inputs ≤100 ng with the following modifications: the reaction volume was reduced by half, the fragmentation/end prep incubation was conducted for 10 min (contemporary samples) or 2.5 min (historical samples), 5 μL of iTru Y‐yoke adaptor (Glenn et al., 2019) was used instead of NEBNext Adaptor, adaptor ligation time was 30 min, bead cleanups were performed with KAPA Pure Beads, iTru i5 and i7 indices (Glenn et al., 2019) were used, and 10 cycles of PCR enrichment were conducted. A negative control was included on each plate during library preparation to test for any potential contamination. All DNA libraries were quantified and assessed with a Qubit fluorometer High Sensitivity Assay and a Tapestation, and final pools were created for sequencing on an Illumina NovaSeq (150 bp paired‐end (PE) reads, Appendix S1). Pool 1 contained 33 historical samples sequenced on one lane of a NovaSeq S4 with 347 other invertebrate samples for a target read number of 5 M PE reads per sample. Pool 2 contained 133 samples sequenced all together on one lane of a NovaSeq X for a target read number of 20 M PE reads per sample. Pool 3 contained 38 samples sequenced with 57 additional samples on one lane of a NovaSeq X Plus for a target read number of 10 M PE reads. To assess whether we could combine data from other DNA libraries, we included 38 DNA libraries (i.e., pool 4) that were prepared with an Illumina Nextera XT2 kit for NextSeq 500 sequencing at Biopolymers Facility at Harvard Medical School.

2.3. UCE and exon analyses

Demultiplexed reads were trimmed using Trimmomatic v 0.32 or v 0.39 (Bolger et al., 2014). Trimmed reads were assembled using Spades v. 3.1 or 3.13.0 (Bankevich et al., 2012). Spades assemblies were then passed to phyluce v 1.7 (Faircloth, 2016) to bioinformatically extract UCEs and exons using previously published bait sets for octocorals (octo‐v2, Erickson et al., 2021) and hexacorals (hexa‐v2, Cowman et al., 2020). The phyluce pipeline was used separately on octocorals and hexacorals as described in the online tutorials (https://phyluce.readthedocs.io/en/latest/tutorials/tutorial‐1.html) with some modifications following Quattrini et al. (2018, 2020). Before aligning with MAFFT v7.130b (Katoh & Standley, 2013), we combined the data from 208 octocoral samples and the zoanthid Umimayanthus with previously published target‐capture data obtained from 187 octocorals and 11 outgroups (Quattrini et al., 2018, 2020, Untiedt et al., 2021, Erickson et al., 2021, McFadden et al., 2022). We combined the data from 30 black coral samples and the stony coral Javania with previously published (Horowitz et al., 2022; Horowitz, Opresko et al., 2023; Horowitz, Quattrini et al., 2023; Quattrini et al., 2018, 2020) target‐capture data from 106 black corals and four outgroups. After alignment, phyluce was used to create a 60% taxon‐occupancy matrix of all loci to maximize locus number while keeping a majority of taxa present per locus. Loci were then concatenated separately for black coral (n = 141) and octocoral (n = 407) datasets. Phylogenomic analyses were conducted using maximum likelihood in IQTree v 2.1 (Minh et al., 2020) on the concatenated datasets with ultrafast bootstrapping (−bb 1000, Hoang et al., 2018) and the SH‐like approximate likelihood ratio test (−alrt 1000, SH‐aLRT Guindon et al., 2010). A partitioned model was used (−p). The best model of nucleotide substitution for each partition was found with ModelFinder (−m TESTMERGE, Kalyaanamoorthy et al., 2017) (Appendix S1). One octocoral sample, Tripalea clavaria, a dried museum specimen, was recovered as sister to all other octocorals. This sample was likely a contaminated sequence, which was pruned from the alignment. The alignment (n = 406 species) was then re‐run in IQTree using the abovementioned parameters.

2.4. Mitogenome analyses

For most samples (n = 204), trimmed reads were also passed to Mitofinder v. 1.4 (Allio et al., 2020) for mitogenome assembly and annotation using a reference database of either octocorals or hexacorals downloaded from GenBank. We used trimmed reads in the analyses with the –new‐genes parameter (to account for mtMutS and HEG) and the translation table (−o) 4 (coelenterate mitochondrial code). For the 38 samples from pool 4, mitogenomes were previously reported in Easton and Hicks (2019, 2020); thus, those results are not included in the present study.

2.5. Nuclear rRNA analyses

We also mapped, assembled, and extracted nuclear rRNA genes from the genome‐skimming data. To obtain a reference sequence for mapping and assembly of octocoral samples, an annotated nuclear rRNA operon sequence, including the nuclear rRNA genes as well as ITS1 and ITS2, was extracted from the NCBI‐annotated Xenia sp. genome (RefSeq assembly GCF_021976095.1, scaffold NW_025813507.1) at NCBI (https://www.ncbi.nlm.nih.gov/genome/annotation_euk/all/). As a reference for black corals, we used a 4721 bp sequence of Cladopathes cf. plumosa (GenBank: MT318868.1) from Barrett et al. (2020) that spans 18S, ITS1, 5.8S, ITS2, and the majority of 28S.

Trimmed read pairs were merged using BBMerge v 38.84 (Bushnell, 2017) with the normal merge rate and the default settings and then imported into Geneious Prime v. 2023.1.2 (https://www.geneious.com). Merged read pairs were mapped and assembled to the reference sequences using the “Map to Reference(s)” function in Geneious with the sensitivity set to “Medium‐Low Sensitivity/Fast” and with five mapping iterations. Consensus sequences were generated from the resulting assemblies with the following settings. At each position, the threshold was set to 90% identity across all mapped reads for base‐calling, a “?” was called if the coverage was less than 10 mapped reads, and the quality was assigned as the highest quality from any single base. Each consensus sequence was trimmed to its reference.

From the consensus sequences, we extracted and analyzed the rRNA genes 18S, 5.8S, and 28S. The consensus sequences were aligned using MAFFT v. 1.5.0 (algorithm E‐INS‐I, scoring matrix 100PAM/K = 2) as implemented in Geneious Prime 2023.2.1 (https://www.geneious.com). Two alignments were analyzed, one including ITS1 and ITS2 in addition to the rRNA genes and another with ITS1 and ITS2 removed (e.g., 18S, 5.8S, and 28S only). The alignments were trimmed at the 5′ end to the beginning of 18S using octocorals as a reference. While we were able to assemble the entirety of 28S for octocorals, we were only able to assemble about one‐half of the 28S gene in black corals, due to incompleteness of the black coral reference sequence used. Partitions were created for both alignments (with and without the ITS). Phylogenetic inference was then conducted with IQTree using the best model of evolution for each locus determined by Modelfinder (−m TEST, Kalyaanamoorthy et al., 2017) and 1000 ultrafast bootstrap replicates (−bb 1000).

In addition to analyzing these concatenated rRNA gene alignments, we also extracted a ~400 bp DNA barcode in the 28S region from the consensus sequences that is targeted by anthozoan‐specific meta‐barcoding primers (McCartin et al., 2023). This 28S DNA barcode was compared to sequences generated via conventional PCR/Sanger sequencing for seven black coral and 28 octocoral samples (McCartin et al., 2023). These barcoding sequences were aligned with MAFFT v. 7.49 (LINS‐I method) and phylogenetic inference was conducted in the same manner using IQTree as for the concatenated alignment of rRNA gene sequences. Best models of sequence evolution for the partitioned datasets were chosen by ModelTest as implemented by IQTree (−m TEST).

2.6. Statistical tests

For historical museum specimens sequenced in pool 1, we conducted analyses to determine whether collection year, library concentration, or DNA concentration impacted the number of reads or loci obtained. We first determined a significant correlation (r = .58, p = .001) between DNA and library concentration and thus removed DNA concentration from further analyses (Figure 2a). Then, we assessed both additive and multiplicative linear regression models on log‐transformed data to determine whether library concentration and collection year affected the dependent variables of number of reads and loci. The multiplicative models had a higher adjusted R‐squared value (.32, .69) than the additive models (.24, .65) for tests on loci and read recovery, respectively; thus, we report the results of the multiplicative model below. We also tested whether the number of loci recovered was influenced by the number of reads obtained per sample.

FIGURE 2.

Data for historical museum samples sequenced in pool 1. (a) Library concentration versus DNA concentration. (b) Number of loci by the number of paired‐end reads. (c–f) Number of reads and loci obtained by library concentration and collection year.

We also determined whether the number of reads obtained across pools 1–3 significantly affected the completion of mitogenome circularization when using MitoFinder. We used a one‐way analysis of variance on log‐transformed data to test whether mitogenome circularization was impacted by read number for both hexacorals and octocorals.

3. RESULTS

3.1. Assembly statistics

Of 242 samples, two failed sequencing with only 4926 and 89,916 PE reads obtained; thus, these samples were removed from subsequent analyses. The remaining 240 samples had between 854,547 and 55,565,170 PE reads, with an average of 17,382,298 ± 8,065,341 PE reads. Pool 1 had an average of 8,343,203 ± 2,922,102; Pool 2 had an average of 23,156,985 ± 3,323,082; Pool 3 had an average of 12,822,312 ± 5,512,007; and Pool 4 had an average of 9,884,551 ± 857,465 PE reads. Trimmed reads were assembled into a mean of 741,347 ± 484,057 SD contigs per sample with a mean length of 348 ± 139 bp (Appendix S1).

3.2. UCE and exon results

UCEs and exons were successfully recovered from the genome skimming data of octocorals and hexacorals. For octocorals, 7–2443 loci (mean 1837 ± 662 SD) of 3023 targeted loci were recovered from each individual. The mean locus size was 1266 ± 1048 bp with a trend of increasing numbers of loci obtained with increasing numbers of PE reads until ~10 M PE reads, where the recovery rate reached a plateau (Figure 3). Of 206 octocorals, <200 loci were recovered in only 3% of samples, which were from pools 1 and 4 with a range of collection ages from 1960 to 2017 and a 10‐fold range of obtained reads (973,960 to 9,534,512 PE reads).

FIGURE 3.

The number of loci recovered by the total number of paired‐end reads obtained per sample in Octocorallia and Hexacorallia.

We were able to recover 18 to 2361 loci (1422 ± 720 loci) from the historical museum specimens, with 1253 loci recovered from the oldest specimen collected in 1886 and 1336 loci recovered from the holotype of Sibogagorgia dennisgordoni, which was collected in 1997 (Figure 2). The mean locus size, however, was smaller (790 ± 578 bp) compared to the contemporary samples preserved specifically for genomics (1355 ± 1093 bp). In general, the number of loci recovered from the assemblies increased significantly (t = 3.663, p = .0009) with the number of reads obtained per specimen (Figure 2b). The number of reads increased significantly with an increase in library concentration (t = 2.31, p = .028), with an interaction effect of year of collection (t = −2.25, p = .032; Figure 2c,e). Likewise, the number of loci increased significantly with an increase in library concentration (t = 2.16, p = .039), with an interaction effect of year of collection (t = −2.14, p = .041; Figure 2d,f).

The phylogenetic tree that included all octocoral samples from genome skimming and prior target‐capture work (alignment: 1262 loci, 243,326 bp) was well supported (Figure 4, Appendix S3), and the genome‐skimmed samples were recovered in the phylogeny within their respective families except one dried museum specimen, Tripalea clavaria, which was recovered as sister to all other octocorals and was thus pruned from the phylogeny. We recovered the two reciprocally‐monophyletic orders, Scleralcyonacea and Malacalcyonacea, and added at least 55 species to the genomic‐scale phylogeny of octocorals. Of 405 nodes, 96% had SH‐aLRT values over 80%, and 89% had bootstrap support values over 95%; most of the low values were near the tips. The zoantharian used as an outgroup in the octocoral phylogeny was correctly recovered in its respective order.

FIGURE 4.

Maximum likelihood phylogeny of octocorals (purple = genome skim, black = target capture) based on UCEs and exons. Outgroups include hexacorals (Hexa). Node support values, represented by circles, include ultrafast bootstraps >95% (blue), 80%–95% (orange), and <80% (red). *Samples genome skimmed and target captured.

For hexacorals, 42 to 1783 loci (mean 1379 ± 476 SD) of 2476 targeted loci were recovered from each individual. The mean locus size was 2385 ± 1961 bp with a trend of increasing numbers of loci obtained with increasing numbers of PE reads until ~20 M PE reads, where the recovery rate slowed (Figure 3). Of 33 hexacorals, <200 loci were recovered in only 9% of samples, which were black corals collected in 2022 and sequenced in pool 2 with a range of obtained reads of 2,353,550 to 4,045,520 PE reads.

The phylogenetic tree that included all antipatharian samples (alignment: 467 loci, 110,353 bp) from genome skimming and prior target‐capture work was well supported, and the genome‐skimmed samples were recovered in the phylogeny within their respective families (Figure 5, Appendix S3). The newly incorporated genome‐skim data (representing all seven antipatharian families) reinforces the monophyletic relationships of Myriopathidae and the monogeneric family, Leiopathidae. All other families are polyphyletic; notably, the new genome skim data reveals that Aphanipathidae is polyphyletic, where Distichopathes hickersonae and Elatopathes abietina are divergent from the rest of Aphanipathidae. This new dataset added at least 10 species to the black coral genomic‐scale phylogeny. The scleractinian that was genome skimmed and included as an outgroup in the hexacoral phylogeny was also recovered in its correct order. Of 140 nodes, 70% had SH‐aLRT values over 80%, and 78% had bootstrap values over 95%. In all cases, the lower node support values were near the tips.

FIGURE 5.

Maximum likelihood phylogeny of black corals (purple = genome skim, black = target capture) based on UCEs and exons. Outgroups include Scleractinia (Scl) and Corallimorpharia (Co). Node support values, represented by circles, include ultrafast bootstraps >95% (blue), 80%–95% (orange), and <80% (red). Leio = Leiopathidae, Clado = Cladopathidae, Stylo = Stylopathidae, and *Species currently included within the polyphyletic family Aphanipathidae.

3.3. Mitogenome results

All mitochondrial protein‐coding genes (PCGs) and rRNA genes were successfully retrieved from 95% of the samples targeted for mitogenome recovery. Of the 170 octocorals, we recovered 14 PCGs and both rSUs in 168 individuals. Only 10 PCGs and mitochondrial rSUs were recovered in two octocorals; both were museum samples collected in 1993 (Callistephanus cf. koreni) and 2005 (Lateothela grandiflora). The mtMutS sequences obtained were successfully integrated with data produced from PCR/Sanger sequencing, resulting in an alignment of 1074 bp (Appendix S3). Placements of taxa in the mtMutS phylogeny were as expected and mtMutS sequence data were 100% identical to the same species that were Sanger‐sequenced. Most (70%) of octocoral mitogenomes were circularized with mitofinder. The majority of these were from Pool 2, which, on average, had the highest number of PE reads obtained across all pools (Figure 6). For octocorals, significantly more mitogenomes were circularized with a higher number of reads obtained (ANOVA, F = 96, p = .001). For the 32 hexacorals, only one individual failed mitogenome assembly, with only three PCGs obtained, yet this individual had over 3,406,440 PE reads. Only 40% of all hexacoral mitogenomes were circularized with Mitofinder, with the majority of these from Pool 2. For hexacorals, no significant differences were found between mitogenome circularization and number of reads obtained (ANOVA, F = 0.25, p > .05).

FIGURE 6.

Circularization of mitogenomes by the number of paired‐end reads from each sample for Octocorallia (n = 171, 114 circular, 57 non‐circular) and Hexacorallia (n = 32, 13 circular, 19 non‐circular). (*p = .001). Gray bars indicate the group mean.

3.4. Nuclear rRNA results

Nuclear rRNA genes were successfully obtained from all but one sample. Reads mapped to >95% of the reference sequence used, and the mean coverage across sites was 4317X. The length of the assembled consensus sequences ranged from 4142 to 6136 bp, and differences in length were mainly because the black coral reference did not include all of the 28S. Across the 478 bp alignment barcoding region of 28S, sequences generated from genome skimming were 100% (p‐distance) similar to their respective Sanger sequence except in the case of one specimen of Sibogagorgia cf. cauliflora (Appendix S3). In this species, the sequence assembled from genome skimming was 15% divergent from the Sanger sequence and had numerous ambiguous base calls across the 478 bp alignment. These ambiguities may reflect intragenomic variability in 28S across its multiple copies. Ambiguous base calls in the consensus sequences of other species in Scleralcyonacea, such as Plumarella pourtalesii, may similarly reflect intragenomic variation. The phylogenetic tree produced from the rRNA genes (6031 bp alignment) included taxa in positions as expected, based on nuclear 28S, UCEs/exons, and mitochondrial loci, except for Sibogagorgia cf. caulifora and Plumarella pourtalesii (Figure S5).

4. DISCUSSION

4.1. The utility of genome skimming

Genome skimming is an effective approach for obtaining a range of loci useful for systematics and reference DNA barcode libraries of anthozoans. In our study, we bioinformatically obtained >1300 UCE/exon loci on average from both hexacorals and octocorals. Our results, combined with studies in other taxonomic groups (e.g., Liu et al., 2021), indicate growing evidence that this approach is effective in obtaining loci, such as UCEs and exons, that have been generally captured through target enrichment.

To assess whether this method can yield results similar to those obtained through target‐capture enrichment, we compared our results with previously published, target‐capture data (Horowitz, Quattrini et al., 2023; McFadden et al., 2022). First, nine octocorals that were genome skimmed in this study were also target‐captured in prior work (McFadden et al., 2022). In all cases, the numbers of UCE loci obtained from the skimmed samples were slightly higher (by ~200 loci) than the same target‐captured samples, and five pairs of skimmed/target‐captured samples included in the phylogeny were recovered as sister taxa. Second, more UCE/exon loci were recovered in genome skimmed samples compared to target‐captured samples from prior studies (Horowitz, Quattrini et al., 2023; McFadden et al., 2022) and on average these loci obtained from skimming were longer (Table 1). The recovery of longer loci is perhaps due to the inclusion of more “off‐target” reads (i.e., reads not matched by baits) that are adjacent to or overlap with the target regions, resulting in longer assembled contigs. Third, the percentage of targeted UCE/exon loci recovered was higher in genome skimmed versus target‐captured samples for octocorals and black corals, however, the average read coverage of UCEs differed greatly between both approaches. Although the coverage is much lower for genome skimming, the results do not suggest that this difference presents an issue in resolving phylogenetic relationships with UCEs/exons. We caution, however, that these are not direct comparisons as different samples were used, different library preparations and protocols were used, the sequencing depth differed among studies, and the DNA quality varied across samples.

TABLE 1.

UCE/exon locus recovery in genome skimmed (GS) samples (this study) and target‐captured (TC) samples from prior studies ^a .

	Mean locus number (± SD)	Mean locus length (bp) (± SD)	Mean targeted loci recovered per sample	Mean coverage % (± SD)
Octocorals
GS	1271 ± 1048	1846 ± 648	62 ± 29%	12 ± 135
TC	1060 ± 393	1498 ± 509	50 ± 16%	262 ± 921
Hexacorals
GS	1380 ± 476	2385 ± 1961	56 ± 19%	20 ± 68
TC	920 ± 276	940 ± 238	37 ± 11%	571 ± 1450

^a Octocorals: McFadden et al., 2022, OCT Samples only, Hexacorals: Horowitz, Quattrini et al., 2023.

In addition to obtaining UCE/exon loci, we obtained most mitochondrial genes and nuclear ribosomal RNA gene sequences. Further, with minimal bioinformatic effort (i.e., just using one mitogenome assembly program), we were able to obtain complete, circularized mitogenomes for 60% of all the samples. Our results indicate that this approach can also be used on historical‐museum samples, as most target regions (i.e., mitochondrial genes, nuclear rRNA genes, UCES/exons) were successfully obtained. Other studies have indicated the utility of obtaining mt genes and nuclear RNA genes from skimming historical museum specimens, including herbaria specimens (Bakker, 2017; Liu et al., 2021) and fishes (Hoban et al., 2022), for example.

Historical specimens, many of which had highly‐degraded and low‐quantity DNA (Appendix S1), performed well with genome skimming. None of these historical specimens were preserved specifically for genetic applications. Yet, we recovered most, if not all, mt genes and nuclear rRNA genes and more than 1000 UCE/exon loci from 75% of the samples. In addition, this approach is useful for obtaining numerous loci from type specimens (i.e., Sibogagorgia dennisgordoni) and specimens collected over 100 years ago. Our results, however, suggest that DNA concentration is directly correlated with library concentration, and higher library concentrations yield more reads and, thus, more UCE/exon loci. In contrast to expectations, the collection year had minimal impacts on UCE/exon loci obtained from the skimming data. Museum specimens used in this study were preserved in various ways, including drying, 70% EtOH, and 95% EtOH. Some specimens were likely fixed in formalin, but this information is often not retained in museum records. Thus, preservation type could confound a direct relationship of collection year with the number of loci obtained. Therefore, we recommend that researchers try genome skimming on various museum samples, regardless of collection age or preservation type. We also urge the use of type material in genome skimming studies to help resolve taxonomic issues in both classes of hexacorals and octocorals. Because the first step of preparing genomic libraries is shearing DNA, one can skip or reduce the shearing time and use just the degraded DNA that is recovered from museum specimens in the DNA library preparation workflow. Our results here suggest that genome skimming is a simple genomic approach that can help unlock our historical museum collections, thus ultimately helping to resolve phylogenetic relationships across Metazoa.

There have been increasing efforts to use environmental (e)DNA sampling to characterize biodiversity and monitor health across ecosystems. It is clear, however, that the classification of eDNA sequences at a meaningful taxonomic resolution relies on the completeness of reference databases of DNA barcodes to which eDNA can be compared (Gold et al., 2021). But DNA barcodes remain missing for many metazoan taxa (e.g., Pappalardo et al., 2021; Ransome et al., 2017), and there are no standard barcodes that can be used to resolve species or even genera across diverse taxa, although both mitochondrial genes and nuclear rRNA genes are often used. Our results suggest that genome skimming is one way to improve reference sequence databases simply and rapidly for applications like eDNA metabarcoding. We provide evidence that 28S rRNA sequences recovered from the genome skimming data were largely congruent to sequences generated from conventional PCR amplification and Sanger sequencing. Furthermore, these data could be used to study intra‐genomic variability along the multi‐copy, nuclear RNA operon.

For the amount of data obtained, genome skimming is a relatively cost‐effective method compared to other genomic and genetic approaches, at least for taxa with genome sizes similar to anthozoans (i.e., ~500–750 MB). Library preparation, sequencing (10–20 M PE reads), and quantification cost ~$60–75 USD for this study. This same amount would facilitate sequencing ~6–7 loci (approximate costs, $6–8 for sequencing, $5 for PCR reaction) through Sanger sequencing. Although the average costs of genome skimming are relatively low compared to Sanger sequencing, the high costs and/or access to genomic sequencing facilities, high‐performance computing, and bioinformatics training might still be prohibitive for some researchers, particularly those in low‐income countries (see, e.g., Rana et al., 2020; Whiteford et al., 2023; Yek et al., 2022). However, samples from several, international research groups could be pooled for sequencing at one genomic sequencing facility, at least in situations where DNA exchange restrictions are not an issue.

4.2. New insights into octocoral phylogeny

At the genus level, the phylogeny of octocorals constructed here using a combination of data obtained from target‐enrichment and genome skimming was largely congruent with that published previously using data from target‐enrichment of UCEs and exons (McFadden et al., 2022). Relationships among families were also mostly in agreement with that previous analysis based on target capture, with the most notable exception being the recovery of the family Cladiellidae as the sister to the gorgonian families Euplexauridae and Paramuriceidae, as was also found by Quattrini et al. (2023). The subordinal‐level clades defined by McFadden et al. (2022) were not, however, as well supported by the analysis presented here (Figure 4). These discrepancies may be attributable to differences between analyses in taxon sampling or the numbers and identities of loci included (i.e., including loci with substitution saturation) and exemplify the challenges inherent in resolving the deepest nodes in a group of organisms that evolved in the pre‐Cambrian (McFadden et al., 2021).

Genomic data were obtained for the first time from representatives of 10 genera (Paracalyptrophora Kinoshita, 1908; Nicella Gray, 1870; Iciligorgia Duchassaing, 1870; Lateothela Moore et al., 2017; Hedera Conti‐Jerpe & Freshwater, 2017; Chromoplexaura Williams, 2013; Pseudoplexaura Wright & Studer, 1889; Placogorgia Wright & Studer, 1889; Aliena Breedy et al., 2023; and Thesea Duchassaing & Michelotti, 1860). Phylogenetic placement of each of these genera was congruent with expectations based on previous phylogenetic analyses of mitochondrial and nuclear rRNA gene trees (Breedy et al., 2023; Cairns & Wirshing, 2018; McFadden et al., 2022). The phylogenomic analysis recovered Thesea as polyphyletic, with some species grouping in the family Paramuriceidae and others in the Gorgoniidae, which is also congruent with previous phylogenetic analyses (Carpinelli et al., 2022). The paraphyletic relationships of Gorgonia to Antillogorgia and of Plexaura and Pseudoplexaura to Eunicea have also been recovered in previous studies (Grajales et al., 2007; Torres‐Suárez, 2014), as has the polyphyly exhibited by Leptogorgia (Poliseno et al., 2017).

Molecular data were obtained for the first time for four genera, allowing their familial relationships to be assessed. Acanthoprimnoa Cairns & Bayer, 2004, a genus whose membership in Primnoidae has never been questioned (Cairns & Bayer, 2004; Cairns & Wirshing, 2018), was instead found to be sister to Ifalukellidae. Tripalea Bayer, 1955, placed in Spongiodermidae based on morphology (Cairns & Wirshing, 2015), appears instead to belong to Incrustatidae in the order Malacalcyonacea. Finally, Caliacis Deichmann, 1936 and Pseudothelogorgia van Ofwegen, 1991, genera whose familial affinities were left incertae sedis by McFadden et al. (2022), each occupy unique positions within the clade of malacalcyonacean gorgonians, suggesting they each deserve family status. Before proposing those new families, however, it will be necessary to confirm the species‐level identification of the material we sequenced by comparison to original type material.

4.3. New insights into antipatharian phylogeny

The black coral phylogeny is mostly congruent with previous reconstructions (Horowitz, et al., 2022; Horowitz, Quattrini et al., 2023); however, this study includes three genera (Distichopathes Opresko, 2004, Plumapathes Opresko, 2001, and Tanacetipathes Opresko, 2001) that have been sequenced for the first time with high‐throughput genomic techniques, providing new insights into phylogenomic relationships within the order. Distichopathes Opresko, 2004 was recovered sister to Elatopathes Opresko, 2004. Along with Asteriopathes Opresko, 2004, these three genera are currently placed in Aphanipathidae Opresko, 2004, but they form a monophyletic clade divergent from the rest of Aphanipathidae. Instead, the three genera show affinity to Stylopathidae Opresko, 2006, a finding consistent with Opresko et al. (2020) based on three mitochondrial and three nuclear gene regions. The recovered monophyletic relationship of the myriopathid genera Plumapathes Opresko, 2001 and Tanacetipathes Opresko, 2001 is notable because they possess distinctly different branching characteristics (planar in Plumapathes vs. bottlebrush in Tanacetipathes). Horowitz, Opresko et al. (2023) (emphasized that smaller‐scale features, such as polyps and spines, are often more informative than branching characteristics, which are very common homoplasies among antipatharians. Indeed, most species within the Myriopathidae share similar spine and polyp characteristics. Thus, these genera within Myriopathidae, as well as other genera across Antipatharia, require further examination for a possible taxonomic revision.

Six out of the seven families in the order Antipatharia are polyphyletic based on this and previous phylogenetic reconstructions (Brugler et al., 2013; Horowitz et al., 2022; Horowitz, Quattrini et al., 2023). Notably, the family Aphanipathidae contains genera spread across the tree (identified by ‘*’ in Figure 5), highlighting the need for taxonomic revisions. However, a formal taxonomic review cannot be conducted yet because the type for Aphanipathidae by subsequent designation, Aphanipathes sarothamnoides Brook, 1889 has yet to be sequenced. Therefore it is not yet certain which clade represents the Aphanipathidae. This study demonstrates that genome skimming and target enrichment are suitable methods to yield high phylogenetic resolution of antipatharians. What is needed now are sequence data from holotype or topotype material representing each nominal and currently accepted genus to fill gaps and better support taxonomic revisions.

AUTHOR CONTRIBUTIONS

Andrea M. Quattrini: Conceptualization (equal); data curation (lead); formal analysis (lead); funding acquisition (equal); investigation (equal); methodology (equal); resources (equal); supervision (lead); visualization (lead); writing – original draft (lead). Luke J. McCartin: Data curation (supporting); formal analysis (supporting); methodology (equal); visualization (supporting); writing – review and editing (supporting). Erin E. Easton: Data curation (supporting); formal analysis (supporting); funding acquisition (equal); methodology (supporting); writing – review and editing (supporting). Jeremy Horowitz: Data curation (supporting); formal analysis (supporting); methodology (supporting); writing – original draft (supporting); writing – review and editing (equal). Herman H. Wirshing: Formal analysis (supporting); methodology (supporting); visualization (supporting); writing – review and editing (supporting). Hailey Bowers: Formal analysis (supporting); methodology (supporting); writing – review and editing (supporting). Kenneth Mitchell: Formal analysis (supporting); methodology (supporting); writing – review and editing (supporting). María del P. González‐García: Methodology (supporting); writing – review and editing (supporting). Makiri Sei: Methodology (supporting); writing – review and editing (supporting). Catherine S. McFadden: Conceptualization (equal); funding acquisition (equal); investigation (equal); methodology (supporting); resources (equal); supervision (supporting); writing – original draft (supporting); writing – review and editing (supporting). Santiago Herrera: Conceptualization (equal); funding acquisition (equal); investigation (equal); methodology (supporting); project administration (supporting); resources (equal); supervision (supporting); writing – review and editing (lead).

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicting interests.

Supporting information

Appendix S1.

Appendix S2.

Appendix S3.

Quattrini, A. M. , McCartin, L. J. , Easton, E. E. , Horowitz, J. , Wirshing, H. H. , Bowers, H. , Mitchell, K. , González‐García, M. d. P. , Sei, M. , McFadden, C. S. , & Herrera, S. (2024). Skimming genomes for systematics and DNA barcodes of corals. Ecology and Evolution, 14, e11254. 10.1002/ece3.11254

DATA AVAILABILITY STATEMENT

Sequence reads were submitted to GenBank under BioProject #PRJNA1043120. All code can be found on GitHub https://github.com/quattrinia/GenomeSkim_paper, and alignment and tree files can be found on Figshare https://doi.org/10.25573/data.24319078.