This species is known from 32 type specimens housed in three different museum collections, the Laboratory of Ichthyology at the Zoological Institute of the Russian Academy of Sciences (ZIN), St. Petersburg, Russia (holotype: ZIN 49228; six paratypes: ZIN 49229–32), the Zoological Museum (ZMMU), Biological Faculty, M. V. Lomonosov Moscow State University, Moscow, Russia (21 paratypes ZMMU: P-17989–91; two paratypes P-18222) and the ichthyological collection of the Zoological Museum (ZMH) of the LIB in Hamburg, Germany [paratype ZMH 24994 (ex ISH 6-1989); paratype ZMH 24993 (ex ISH 5-1989)]. We sampled muscle tissue from the paratype specimen ZMH 24994 (Figure 2A) at the caudal peduncle using a biopsy needle for minimally invasive sampling. The tissue was preserved in the original preservation fluid of the storage container. The specimen was captured at 25°21′S and 85°8′W at a depth of 720 m on 24.04.1987. It is a juvenile male of 445 mm total length (Thiel et al., 2009; Straube et al., 2011a). Although not explicitly mentioned in the original description, or tested by us, the overall condition of the specimen indicates a fixation in formaldehyde: both body and eyes do not show bleaching of exclusively ethanol preserved samples. Furthermore, the common procedure during research cruises at the time of sampling was a fixation of specimens in 4% formaldehyde and long-term preservation in 70% ethanol.

Laboratory steps and analysis of test-sequencing data of this specimen is described in detail in Straube et al. (2021a). The sample was incubated in a GuSCN-based buffer (Rohland et al., 2004) applying the protocol by Dabney et al. (2013) for DNA purification. A single-stranded DNA library was then constructed, and test-sequencing was performed to check for the ratio of target DNA and contamination. After detection of endogenous DNA in the test-sequencing dataset, target capture for mitochondrial DNA was performed using home-made baits. These were generated from long-range PCR products amplified from the DNA of Etmopterus cf. molleri tissue housed in the tissue sample collection of the Bavarian State Collection of Zoology (registration number: Ich-P-CH-0264). For the long-range PCR protocol and primers see Straube et al. (2021a). Hybridisation capture was then performed following the protocol of González Fortes and Paijmans (2019), where the single-stranded library is mixed with the denatured bait library after addition of blocking oligos. Hybridisation of target DNA to baits was carried out for 24 h at 65°C. The captured library was then amplified, and the capture procedure and amplification repeated. The resulting double captured library was then sequenced using custom sequencing and index 2 read primers (Gansauge and Meyer, 2013; Paijmans et al., 2017) on an Illumina^® MiniSeq instrument. We used a mid-output kit in a pool of double indexed samples.

Paired-end raw reads were quality and adapter trimmed with Cutadapt v.1.16 (Martin, 2011) using default settings. The iterative mapping algorithm MitoBim v. 1.9.1 (Hahn et al., 2013) was then used to reconstruct the mitochondrial genome sequence, using default settings and Genbank entry KU892588 (Etmopterus pusillus; Chen et al., 2016) as reference for initial baiting. Annotation was performed by aligning the paratype consensus sequence to KU892588 in Geneious^® Prime 2021.1 (Biomatters Ltd. Auckland, New Zealand), and checked for internal stop codons. Protein coding genes could not be fully reconstructed. The tRNA-Phe and tRNA-Val transfer RNAs, and the 12S and 16s ribosomal RNAs could be completely reconstructed and were therefore extracted for phylogenetic analysis (2676 bp in total). Reads used in the last iteration of Mitobim were mapped back to the mitochondrial genome consensus sequence as well as to the tRNA and rRNA sequences using BWA aln v.0.7.17 (Li and Durbin, 2009), with default settings, to check if the reads could be unambiguously mapped. Further, BWA was used to align the trimmed and quality filtered reads excluding duplicates to the full mitochondrial genome sequence as well as the tRNA-Phe, the 12S ribosomal RNA, the tRNA-Val and the 16S ribosomal RNA of KU892588 to assess coverage. Obtained sequences were aligned to the sequences of specimens listed in Supplementary Table 1, covering nine of the eleven species of the E. spinax group (Straube et al., 2010; Ebert et al., 2021). Sequences used to determine the phylogenetic placement of E. litvinovi were obtained from the Chondrichthyan Tree of Life (2016) project¹ which are collected from vouchered and validated specimens, as described in White et al. (2018). A maximum likelihood tree was computed using RAxML v.8.2.4 (Stamatakis, 2014) under the general time reversible model. Heterogeneity of substitution rates among sites was modelled using a GAMMA distribution. To assess the statistical support for nodes, bootstrapping with 100 replicates was performed and plotted onto the maximum likelihood tree. A haplotype network was reconstructed with POPArt v. 1.7 (Leigh and Bryant, 2015) using the median joining network algorithm (Bandelt et al., 1999) under default settings. The RAxML tree served as a basis for calculating the p-distances between E. litvinovi and E. spinax clade species analysed herein using the Species Delimitation Plugin 1.4.5 (Masters et al., 2011) in Geneious^®.

This species is known from six specimens housed in three different museum collections, the ZIN (holotype: ZIN: 49226; two paratypes: ZIN: 49227); the ZMMU (paratype ZMMU: P-17992, paratype ZMMU P-17993) and the ZMH [paratype ZMH: 24995 (ex ISH 4-1989)]. We sampled tissue from the paratype specimen ZMH 24995 (Figure 2B) as described previously for the E. litvinovi paratype specimen. The specimen was captured at 25°56′ S and 88°33′ W at a depth of 580 m on 30.04.1987. It is an adult male of 426 mm total length (Thiel et al., 2009). As described for the E. litvinovi paratype, the overall condition and sampling date of the specimen suggests fixation with formaldehyde.

DNA extraction of the sample involved the same procedure as for E. litvinovi. Single stranded library preparation of E. pycnolepis DNA followed the protocol described in Gansauge et al. (2017). The E. pycnolepis sample underwent different laboratory procedures in comparison to the E. litvinovi sample, as the samples were processed with a considerable temporal gap, during which time the standard procedures in the historical laboratory at the University of Potsdam had been updated. Raw test-sequencing reads were analysed as in Straube et al. (2021a). FastQ Screen v0.14.0 (Wingett and Andrews, 2018) was used to check for unique hits to Etmopterus references and estimate contamination levels, before proceeding with target capture. After detection of target DNA in the test sequencing dataset, target capture was performed using an Arbor Bioscience myBaits^® RNA bait kit. The baits were part of a multi-locus, multi species museum specimen barcoding approach described in Agne et al. (2022). NADH2 bait sequences were derived from representatives of all four Etmopterus clades (Straube et al., 2010) deposited in Genbank: E. lucifer (JQ518963), E. gracilispinis (JQ518960), E. granulosus (KF861686) and E. bigelowi (JQ518959). The four sequences were initially published in Naylor et al. (2012) and Straube et al. (2015). The single stranded DNA library was captured twice following the protocol described in Huang et al. (2021) using a hybridisation temperature of 65°C for 24 h. Sequencing of the double-captured, indexed library was performed on an Illumina NextSeq 500 System at the University of Potsdam as described in Paijmans et al. (2017). After quality filtering and adapter trimming using Cutadapt v. 2.10 (Martin, 2011) under default settings, reads were processed as described for E. litvinovi to reconstruct the NADH2 sequence of the paratype, using the NADH2 sequence of E. lucifer (JQ518963; Naylor et al., 2012) as reference.

The NADH2 consensus sequence (1044 bp) of the paratype was subsequently aligned with NADH2 sequences of other Etmopterus species with focus on the E. lucifer clade (Supplementary Table 2). Comparative sequences were obtained from the Chondrichthyan Tree of Life (2016) project (see text footnote 1). For details of NADH2 amplification and sequencing see Naylor et al. (2005, 2012). Forward and reverse sequences were aligned based on chromatograms and edited using Geneious^® Pro v. 6.1.7 (Biomatters Ltd. Auckland, New Zealand). The consensus sequences were translated to amino acids and aligned with corresponding NADH2 sequences from representatives of closely related species using the MAFFT (Katoh et al., 2002, 2005) module in Geneious^®. The aligned amino acid sequences were translated back in frame to their original nucleotide sequences, to yield a nucleotide alignment 1044 base pairs in length. Analysed samples are listed in Supplementary Table 2 including 13 of the 15 E. lucifer clade species. Phylogenetic inference and species delimitation was performed as described for E. litvinovi.

A total of 3,734,481 trimmed and quality filtered reads were available after combining test-sequencing and target capture data, including duplicates. MitoBim ran for four iterations. 1,589,598 reads were used for baiting in the final iteration step. The consensus sequence shows 1.26% ambiguities scattered across the mitochondrial genome. Excluding duplicate sequences, 4985 reads map to the consensus sequence resulting from the Mitobim analysis providing an average coverage of 22 reads. The GC content is 40%. The mitochondrial tRNA and rRNA markers used for the phylogenetic analysis showed mapped read lengths mostly larger than 70 base pairs (Supplementary Figure 1A) and an average coverage of 58 reads, excluding duplicates (Supplementary Figure 2A). They did not show any ambiguous nucleotides.

The maximum likelihood phylogeny of the tRNA and rRNA sequences identifies lineages corresponding to species within the E. spinax clade. The relationships in the tree are mostly well-supported with many bootstrap values reaching 100% (Figure 3A). The E. litvinovi paratype sequence is sister to a sample identified as E. benchleyi. This clade also includes a specimen of E. benchleyi sampled in the Indian Ocean (GN4952). The clade as a whole is sister to the North Atlantic species E. princeps and E. spinax, which together form the sister clade to the Southern Hemisphere species E. viator (Figure 3A). The reconstructed haplotype network detected five haplotypes with 34 segregating sites and 16 parsimony-informative characters. Figure 3B shows that the haplotype sequence of the E. litvinovi paratype is identical to the E. benchleyi sample GN14570 and separated by a single mutational step from E. benchleyi sample GN4952. The species delimitation analysis shows that the interspecific K2P distance between two valid sister species within the E. spinax clade is on average 1.6% (Supplementary Table 3A), while the K2P distance between E. litvinovi and E. benchleyi is substantially smaller (K2P distance = 0.0518%; Supplementary Table 3A). Overall, our data does not support the validity of both species due to the phylogenetic placement of the E. litvinovi paratype sequence in the E. benchleyi clade and a very small K2P distance between both species.

The test-sequencing dataset of 398,752 trimmed reads detected the presence of Etmopterus DNA, as indicated by 5.62% unique hits to the E. spinax transcriptome used as reference in the FastQscreen analysis. 82.83% of reads were un-assigned to any of the provided references, and contamination with other samples processed simultaneously was not detected. 4,029,200 raw reads were produced by sequencing of the target captured library. Quality filtering and trimming reduced this to 2,680,159 reads. Of these, 139,219 reads mapped to the NADH2 gene of E. lucifer (JQ518963). The complete NADH2 sequence was reconstructed after three iterations in Mitobim, using 115,548 reads in the final iteration. The mapped read length distribution is shown in Supplementary Figure 1B. The modal fragment length is around 50 base pairs. Mapping those reads back to the reconstructed NADH2 consensus sequence, showed that 277 reads mapped with an average coverage of 13 reads, excluding duplicate sequences (Supplementary Figure 2B). The maximum likelihood NADH2 phylogeny shows well-supported clades within the E. lucifer group. Several clades do not correspond to species: E. brosei clusters with E. sculptus, E. cf. molleri clusters with E. cf. decacuspidatus and forms a distinct clade not including E. molleri. Etmopterus molleri and E. dislineatus do not form two distinct clades. The E. pycnolepis paratype specimen is sister to a clade containing the southern hemisphere samples of E. lucifer and E. sculptus (Figure 4). The species delimitation analysis shows that the interspecific K2P distance of the NADH2 gene between sequences of two valid species in the E. lucifer clade, excluding clades not corresponding to species, is on average 4% (Supplementary Table 3B). Comparing the K2P distance of E. pycnolepis to its closest sister taxon, E. lucifer, the K2P distance is 3.4%. Our data therefore supports the species status of E. pycnolepis.

The mitochondrial DNA we obtained from both paratype specimens, while fragmented (Supplementary Figure 1), was less degraded than other museum samples analysed in previous studies (Straube et al., 2021b). Fixation and preservation cause DNA damage (e.g., Stiller et al., 2016; Hahn et al., 2021; Straube et al., 2021a); however, the mitochondrial DNA of the two paratype samples analysed herein may be less affected due to the comparatively young age of 32 and 34 years, respectively, at the time of extraction. More comparative data is necessary to test if time is correlated with mitochondrial DNA fragmentation levels, and if fragmentation is ongoing under the current storage conditions.