Expected results

Using the protocol as described here, we extracted DNA from 148 (144 fixed in formalin and 4 ethanol-only preserved) T. thynnus larvae ranging in length from 2.24–5.34 mm. All larvae were sourced from historical collections from Gulf of Mexico populations held at the National Oceanic and Atmospheric Administration Marine Southeast Fisheries Science Center in Miami, Florida, USA. The formalin-fixed larvae had been collected in 1996 and preserved with 10% seawater buffered formalin and transferred to 95% ethanol after 24 to 48 hours. The ethanol-only preserved larvae had been collected in 2018 and immediately preserved in 95% ethanol.

The total DNA yield from each formalin-fixed larva ranged from 2.88–115.2 ng (mean = 52.77; standard deviation = 24.62), which was a marked improvement over our results using a version of this protocol which utilised ethanol precipitation rather than SPRI beads to concentrate the purified DNA [7] and was similar to the yield from the ethanol-preserved larvae (mean = 52.58 ng). With the 16 formalin-fixed larvae yielding the highest DNA concentrations, as well as the four ethanol-preserved larvae, we prepared libraries for sequencing using the IDT xGen cfDNA and FFPE DNA kit from a uniform 15 ng of input DNA with 13 PCR cycles (Fig 1 and S1 Raw images). We sequenced the resulting libraries each with approximately 125 million paired-end 150 bp reads on the Illumina NovaSeq 6000 S4 platform.

Critically, our whole genome re-sequencing success is supported by using a trimming-free alignment approach as outlined in Hahn et al. [7], which aligns raw reads with the ngskit4b tool suite version 200218 (https://github.com/kit4b). Specifically here, we aligned to a hard-masked reference genome for T. albacares (yellowfin tuna, fThuAlb1.1) using the kalign function with options -c25 (--minchimeric = <int>; minimum chimeric length as a percentage of probe length) -l25 (--minacceptreadlen = <int>; after any end trimming only accept read for further processing if read is at least this length) -d50 (--pairminlen = <int>; accept paired end alignments with observed insert sizes of at least this) -U4 (--pemode = <int>; paired end processing mode: 4 –paired end no orphan recovery treating orphan ends as SE). From these alignments, we remove PCR and optical duplicates from the alignments using the MarkDuplicates function of PICARD enabling REMOVE_DUPLICATES = TRUE.

The mean insert size from the 16 formalin-preserved tuna larvae libraries ranged from 87.8–94.1 bp, which was slightly shorter than that of the ethanol-preserved specimens (mean = 122.5 bp). Pairing this extraction protocol with the IDT xGen cfDNA and FFPE kit, between 59.7% and 67.8% of the raw reads mapped to the reference genome compared to a maximum of 21% using alternative library preparation methods [7] and a mean of 51.0% for the ethanol-preserved larvae. Here, we achieved between 17.8X and 25.4X nuclear genome coverage from the formalin-fixed larvae (Table 1 and Fig 2), which again is a marked improvement over the maximum 3.1X nuclear genome coverage we achieved previously [7]. However, coverage is not directly comparable due to differences in genome size and sequencing effort between the two studies. In comparison, we achieved between 28.8X and 35.9X nuclear genome coverage from the ethanol-preserved larvae. Though the formalin-preserved larvae yielded lower coverage than the ethanol-preserved specimens, the coverage is sufficient for downstream applications. Focusing on the mitochondrial genome coverage, we achieved mean coverage of 1434X with the formalin-preserved larvae and 2215X with the ethanol-preserved larvae, demonstrating that the extraction method can be used to retrieve exceptionally high and even coverage of organellar genomes (Table 1 and Fig 2).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Alignment of Atlantic bluefin tuna reads mapped to the yellowfin tuna genome using kalign viewed in CLC Genomics Workbench.

Shown are (A) a section of the nuclear and (B) complete mitochondrial genome for sample 1 from the formalin-fixed larvae. After read deduplication, sample 1 yielded mean 19.8X coverage of the nuclear genome and 1784X coverage of the mitochondrial genome. A coverage graph is shown above in green with individual forward (green) and reverse (red) reads below. Fixed differences between the mapped reads and the reference can be observed in both A & B indicated in blue, yellow and red.

https://doi.org/10.1371/journal.pone.0296491.g002

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Exemplary DNA yield and mapping results.

https://doi.org/10.1371/journal.pone.0296491.t001

While our method enhances DNA yield and mapping success from formalin-preserved archival tissues, results remain contingent on the preservation quality of the specimen. Our protocol emphasises the importance of evaluating specimens to minimise unnecessary destructive sampling. Most critically, a thorough visual examination of internal organs is crucial to avoid sampling specimens displaying signs of decomposition before fixation. Due to variation in preservation practices, we cannot guarantee specimens identified as suitable will yield sequenceable DNA. Therefore, we recommend conducting a preliminary extraction on a few representative samples to assess DNA yield; more extensive sampling is recommended if the total yield exceeds 200 ng from 50 mg of tissue. We caution that tissue type can affect results and recommend conducting pilot extractions when exploring alternative tissues. Ideally, studies should report preservation quality metrics alongside sequencing results to enhance comprehension of the relationship between specimen quality and sequencing success. Such an understanding can help to guide further protocol refinement.

DNA extracted from formalin-preserved archival specimens tends to be highly fragmented, with a mean fragment length between 50 and 150 bp. Therefore, we recommend taking expected fragmentation into consideration when designing a sequencing strategy. For example, amplicon-based sequencing approaches may be appropriate only when designed to target fragments in this range. Alternatively, bait-capture sequencing methods can be applied with similar considerations made for the expected fragment sizes. We do not recommend de novo genome assembly using these data. Care must be taken in selecting an appropriate reference genome, as mapping success will be affected by both reference quality and the evolutionary distance. Additionally, for phylogenetic analysis, consideration must be given to the pangenomic differences between individuals as they may obscure detection of regions absent in the reference.

Sequencing data for the tuna specimens is being made available on the CSIRO Data Access Portal (DAP) and supporting publications will follow.