We successfully isolated DNA from all 12 specimens including fixed soft tissues, bone and pelt (Table 1). DNA was typically fragmented with an average size of approximately 100 base pairs (bp). The hyper-variable region of the thylacine mtDNA contains a 72–75 bp repeat in triplicate. As the DNA samples were already considerably fragmented, amplification of the repeat using small overlapping amplicons would not clarify their relative position. Therefore, we amplified a small unique portion of the mtDNA control region directly after the repeat (Figure S1).

We were able to compare six individuals over 187 bp and 12 individuals over 117 bp of the mitochondrial hyper-variable region. There were only six nucleotide differences between all of the animals sequenced, and all of the polymorphisms were within the smaller 117 bp region. These polymorphisms were found in five individuals collected in 1852 (1 difference), 1871 (2 differences), 1887 (1 difference), 1888 (1 difference), and one of unknown collection date (1 difference). Genetic diversity among the specimens, or the chance of two randomly chosen sequences being different, was 60%±21.5% and 68±14% for the 187 and 117 bp fragments respectively. Nucleotide diversity between individual homologous nucleotides was 0.5%±0.47% and 0.8%±0.6% for the 187 and 117 bp fragments respectively, while the mean number of nucleotide differences between sequences was 1.0±0.77 and 1.0±0.72 for the 187 and 117 bp fragments respectively. To determine our expected error rate caused by extrinsic damage to the archived specimen's DNA we sequenced the haplotypes of two offspring derived from a single female specimen. The young were 100% identical to the mother over 187 bp of sequence, suggesting a very low extrinsic DNA damage rate (pouch young were not included in the population analysis).

Additional mitochondrial genome sequences were isolated from a 454 whole genome shotgun run using DNA derived from pelt (Specimen C-5748; Table 1). The sequences outside of the hyper-variable region included 16 fragments ranging in size from 142–458 bp (Figure 2). Eleven of these sixteen fragments were 99.91% similar to the published genome, which matches previous estimates of mtDNA variability in the thylacine [7]. However, the remaining five fragments spanning three locations were only 89.2% similar and contained numerous polymorphisms relative to the previously published thylacine mtDNA genomes [7].

Three of these polymorphic fragments formed one contiguous block of 301 nucleotides that mapped to a 79 bp portion of the ATP synthase subunit 8 gene (ATP8) and a 222 bp portion of the ATP synthase subunit 6 gene (ATP6; Figure 2). This block contained 37 polymorphisms that encoded 15 amino acid changes relative to the published genome [7]. The second variable region mapped to the cytochrome oxidase subunit 1 gene (COX1) and contained 73 polymorphisms over 458 bp. These polymorphisms encoded 11 amino acid substitutions (Figure S2).

We performed phylogenetic analysis of the nucleotide and amino acid (where appropriate) sequences encoded by the variable fragments with closely related marsupial sequences including those from the Tasmanian devil, Northern quoll (Dasyurus hallucatus) numbat (Myrmecobius fasciatus) and the distantly related South American gray short-tailed opossum (Monodelphis domestica). Analysis of the inferred amino acid sequences grouped the thylacine COX1 variable fragment separately from all marsupial orthologs (Figure 3). In contrast, the variable ATP8/ATP6 amino acid sequence grouped with the published thylacine homolog. Phylogenetic analysis of nucleotide sequences for both ATP8/ATP6 and COX1 also grouped the variable reads with the published thylacine homologs (Figure S3).

The third variable region mapped to the 16S ribosomal RNA gene and contained 15 polymorphisms over 389 bp. 16S rRNA is not translated, thus we cannot provide an amino acid comparison. Phylogenetic analysis also grouped the variable fragment with the published thylacine homolog similar to the ATP8/ATP6 and COX1 genes (Figure S3). All of the nucleotide phylogenies were supported by high bootstrap values (>91) except for the COX1 variable nucleotide sequence in which the bootstrap values, but not the groupings, differed between models (ranging from 56–78).

Contamination of our thylacine DNA with bacteria and nucleic acids from other sources was very low (Figure S4). Non-marsupial reads represented less than 10% of our sequences, caused largely by microbial contamination (8.1%). These contaminating reads were removed prior to our analyses.

Specimens were obtained with permission from the Berlin Museum für Naturkunde, Museum Victoria, University Museum of Zoology Cambridge, Museo Nacional de Ciencias Naturales, Staatliches Museum für Naturkunde Stuttgart, Hessisches Landesmuseum Darmstadt, Übersee Museum Bremen, Forschungsinstitut Senckenberg, University of Heidelberg Museum of Zoology and the Zoologisches Museum Greifswald.

A small sample of pelt tissue or nasal cartilage was cut discretely from individual specimens using a sterile razor blade and gloves. Specimens were kept in sterile cryotubes at room temperature and extracted at the earliest possible convenience. Samples of the alcohol preserved specimens were carefully dissected under sterile conditions. Liver tissue was collected into sterile cryotubes and held at 5°C until extraction.

DNA was extracted using the method of Pääbo [16] within a UV-sterilized reverse-flow containment hood, housed in a dedicated clean room separate from the area where PCR amplifications were performed. Bone powder (0.5 g) from one specimen was obtained by using a 1.5 mm drill bit. DNA was extracted by incubation of the bone powder for 48 hours at 42°C in 0.5 M EDTA. This preparation was centrifuged (5 minutes, 15,000 RPM) briefly and new EDTA and 20 µl of 10 mg/ml Proteinase K (Peqlab, Germany) added and incubated for a further 24 hours. The digested sample was centrifuged, 1 ml of supernatant was removed and combined with fresh phenol, mixed thoroughly and again centrifuged. The aqueous fraction was then removed and the process repeated again with fresh phenol, and then once more with chloroform/isoamyl alcohol (24∶1). The final supernatant was removed and 1 ml of ice-cold 100% ethanol added to the solution and incubated at −20°C overnight. This preparation was then centrifuged for 30 mins at 15,000 RPM, the ethanol removed, and the DNA resuspended in 20 µL of ultra clean water.

Mitochondrial sequences were amplified and isolated using a two-step PCR method. The first-round PCR was performed as a multiplex reaction in 50 µL total volume using high fidelity, proof reading polymerase master mix (25 µL; Bio-Rad, USA), three primer sets (1 µL each at 10 ng/µL), BSA at a concentration of 1 µg/µL, 1 µL of DNA and ultra sterile water to 50 µL. The conditions for the PCR were 98°C for 3 minutes, 25 cycles of 98°C for 10 seconds, 48°C for 10 seconds, 72°C for 30 seconds and a final extension of 72°C for 5 minutes. The whole PCR sample was then purified using a MinElute column (Qiagen, Germany) and resuspended in 10 µl of elution buffer ready for the second-round PCR. The second PCR was performed using specific primer sets in a 25 µL reaction with 12.5 µL Dream Taq (Thermo Scientific, Germany) or GoTaq (Promega, USA), 1 µL of primer (10 ng/µL), 2 mµ of purified PCR DNA, and water to 25 µL. Amplification conditions were as follows: initial denaturation 95°C for 4 minutes, 30 cycles of 95°C for 30 seconds, 48°C for 30 seconds, 72°C for 1 minute, and a final extension of 72°C for 5 minutes. Primer sets included (Forward primer 1: CTTTACCTAATATTTCTCAT; reverse primer 1: GGATTATGGTCGTAGATCAT; forward primer 2: ACTTACATAGACATTTCATC; reverse primer 2: AAATGACCGAGTTGCTGATC). All sequences were deposited in the Genbank database (accession numbers: JQ318672–JQ318683). No template controls were processed using the same conditions. Final PCR products were separated using electrophoresis on 1% polyacrylamide gels stained with GelRed (Biotrend, Germany). Positive bands were isolated under UV light and purified for sub-cloning.

Positive PCR fragments were sub-cloned into pBluescript or pGEM-T easy (Promega, USA) for sequencing. A minimum of 4 plasmids per PCR fragment were sequenced in both forward and reverse directions using T7 and SP6 primers. A consensus sequence was constructed for each individual based on multiple clones. Nucleotide positions that may indicate deamination sites relative to the sequences published by Miller et al [7] were excluded by performing a secondary PCR from the original DNA. The presence of nuclear mitochondrial sequences (NuMts) was excluded based on overall similarity of the sequences relative to published thylacine mtDNA genomes and the conservation of individual sequences generated from overlapping primer combinations.

Sequences were aligned using Mega version 5.0 [17] and standard molecular indices of diversity were generated using Arlequin version 3.1 [18]. Sequence analysis was performed on 187 bp of mtDNA sequence from 6 specimens (including those published by Miller et al., 2009; Table 1), and 117 bp of mtDNA sequence from 12 specimens. Sequences derived from pouch young were not included in this analysis because of their known relationship to the female specimen. Phylogenetic trees were constructed using Mega version 5. All trees were constructed using bootstrap tests (1000 replicates) and default settings for a range of phylogeny methods including maximum likelihood (Model: Jones-Taylor-Thornton for protein; Tamura-Nei for nucleotide), neighbour joining (Model: p-distance for protein; maximum composite likelihood for nucleotide) and minimum evolution (Model: p-distance for protein; maximum composite likelihood for nucleotide). All of these phylogeny methods produced identical groupings for each of our gene comparisons. Published sequences used for these comparisons included NCBI accession numbers; FJ515780, YP002519811, YP002519812, YP002519809 (Tasmanian tiger), FN666604, CBJ55445, CBJ55446, CBJ55443 (Tasmanian devil), NC007630, YP423966, YP423967, YP423964 (Northern quoll), FN666603, YP002519876, YP002519877, YP002519874 (Numbat) and NC006299, YP087183, YP087184, YP087181 (Opossum).

DNA was extracted from a thylacine pelt specimen (C5748) as previously described [6]. The total DNA sample (without size selection) was processed for library construction using the random shotgun approach and the GS FLX Titanium library preparation kit (454 Life Sciences) using 14 cycles for library amplification. Genomic library fragments were amplified with the GS FLX Titanium SV emPCR Kit (454 Life Sciences). The subsequent library was sequenced using the GS FLX system (454 Life Sciences).

Our run produced 4,976 reads with an average length of 301 bases. The reads were mapped to the previously published thylacine mitochondrial genomes FJ515780 and FJ515781 [7] using the program BWA 0.6.1 [19]. BWA was able to map 16 reads to each genome. Human contamination was computationally removed by mapping reads to the human genome, version GRCh37.64 from Ensembl, and any that matched were removed from the pool of reads. To evaluate the effectiveness of this, all 4,976 reads were mapped to NCBI's nr database using BLAST [20], [21]. The output was parsed using a custom bioperl script to identify the organism associated with the highest E-value. Reads with more than 50% that mapped to the lowest E-value hit were counted as that organism (Genbank accession numbers: JQ318684–JQ318693). We determined the high quality of our thylacine 454 library by assessing sequence similarity of 2699 uniquely mapping reads (not derived from repeat elements) to the Tasmanian devil genome. These reads showed an average of 98.5% sequence identity, consistent with the evolutionary divergence of these species and previous genomic comparisons.