From historical expedition diaries to whole genome sequencing: A case study of the likely extinct Red Sea torpedo ray

Torpedo rays (Torpedinidae, Torpediniformes) are small to moderately large batoids that produce an electric discharge. They are distributed worldwide in temperate and tropical seas and are, as a result of their bottom‐dwelling behaviour, susceptible to trawl fishing and often end up as victims of bycatch. The distribution ranges of most recognized species seem to be restricted; however, their species‐level systematics is not adequately resolved. In the genus Torpedo, in which many species require revision, there are possibly several undescribed species, while numerous misidentifications add to the complexity of the issue. In the latest lists of living rays, 13 species are accepted in the genus Torpedo, including three of doubtful validity and several recently discovered undescribed species. Among the valid species is the critically endangered, possibly extinct, Torpedo suessii Steindachner, 1898, the Red Sea torpedo, of which only four specimens have been recorded in the literature until now, three of which still exist in the fish collection of the Natural History Museum of Vienna. Museum collections are the most important archive of biodiversity on Earth, and are increasingly being used for various studies, including phylogenetics, population genomics, and biogeography. Nevertheless, molecular analysis of old museum material remains challenging because the genetic material has degraded, is fragmented, and of low quantity. In molecular taxonomy, the necessity of including type specimens as name‐bearing specimens is increasingly recognized. Here, the extended specimen approach was applied to re‐describe the lectotype of T. suessii. The approach included research of historical information and whole genome sequencing, followed by genome assembly and phylogenetic analysis.


| INTRODUCTION
Torpedo rays (Torpedinidae, Torpediniformes) are small to moderately large batoids, up to 180 cm total length (TL), mostly <100 cm TL (de Carvalho, 2022), whose common feature is the production of a powerful electrical discharge of up to 45 volts, used for defence and to stun prey (Bigelow & Schroeder, 1953;Coates & Cox, 1942;Sheridan, 1965).These rays are distributed worldwide in temperate and tropical seas, on the continental shelf in shallow coastal regions or pelagic waters, often to a depth of 100 m.Some species may inhabit estuaries and also deeper waters of the continental slope though none enter freshwater (de Carvalho et al., 2016;McEachran & Aschliman, 2004).Due to their bottomdwelling behaviour, Torpedo rays are susceptible to trawl fishing and, while not of interest to commercial or industrial fisheries, they often end up as victims of bycatch (Jabado et al., 2017;Moazzam & Osmany, 2021;Stevens, 2000).As much as can be evaluated from the sporadically published records, the distribution ranges of most recognized species are limited (de Carvalho, 2022;Sujatha et al., 2014).Meanwhile, their systematics, in particular at the species level, is inadequately resolved (de Carvalho, 2022; see below for further details).
The genus Torpedo is taxonomically challenging (Pollom et al., 2019) as there are cases of both high levels of intraspecific variability (De Carvalho, 2022) and low levels of interspecific variability (Serena et al., 2020;Wallace, 1967).Therefore, many species require revision, especially since several undescribed species and coloration forms have recently been discovered.Furthermore, species ranges may be more restricted than previously thought, with the possibility that high levels of endemicity will be discovered during revisions or new species descriptions (Compagno et al., 2022;de Carvalho, 2022;de Carvalho et al., 2016;Ebert et al., 2021;Kyne et al., 2017;Serena et al., 2020).Meanwhile, numerous misidentifications also pose a problem (de Carvalho, 2022;de Carvalho et al., 2002;Serena, 2005;Serena et al., 2020).For example, records from off the coasts of the Indian Ocean Islands of Madagascar, Mauritius, and the Seychelles may represent misidentifications or species yet to be described (Compagno et al., 2022;de Carvalho, 2022;de Carvalho et al., 2016;Pollom et al., 2019).In the latest lists of living rays (de Carvalho et al., 2016;Weigmann, 2016) 13 species are accepted in the genus Torpedo, including three of doubtful validity, as well as several recently discovered undescribed species (Table S1; de Carvalho et al., 2016;Ebert et al., 2021;Weigmann, 2016): e.g., Torpedo sp.(Comoros), and Torpedo sp.(Mozambique) (de Carvalho, 2022: Plate 54).According to IUCN (http://www.iucnredlist.org), of the 10 indisputable species, one is listed as of least concern, two as vulnerable, four as endangered and one as critically endangered, while two are listed as data deficient (Table S1).
The critically endangered, possibly extinct, species is Torpedo suessii Steindachner, 1898, the Red Sea torpedo, of which only four specimens have been recorded in the literature until now.Three of these specimens were collected during the second Austrian Expedition to the Red Sea in 1897-1898 in the waters off Perim and Mokha [al-Mukhā] in Yemen (Steindachner, 1898a(Steindachner, , 1898b)).These individuals are deposited in the fish collection of the Natural History Museum of Vienna (NHMW) whereas one specimen is recorded only from a photograph and was not collected (de Carvalho et al., 2002;Michael, 1993).In the literature, T. suessii was considered a valid species or "closely allied to if not identical with" T. panthera Olfers, 1831 by Garman (1913: 308; in the genus Narcacion), or a synonym of T. sinuspersici Olfers, 1831 by Fraser-Brunner (1949: 947) and Randall (1995: 41).It was re-established as a distinct species (Compagno, 2005;de Carvalho et al., 2002de Carvalho et al., , 2016;;Weigmann, 2016), with de Carvalho et al. (2002) highlighting its very specific ornate color pattern.
In the present study, the Extended Specimen Network approach (Lendemer et al., 2020;Webster, 2020) was applied, following Bogutskaya et al. (2022), to re-describe the lectotype of T. suessii.This approach was widened to genomic data analysis by performing whole genome sequencing followed by genome assembly and phylogenetic analysis.The present study had two goals: (1) to provide genomic analysis of the type specimen of a rare (possibly extinct) torpediniform species, coupled with an in-depth research of the history of the specimen, while providing some insight into its phylogenetic position within Torpedo rays, and, based on this example, (2) to present an integrative approach toward re-description and digitalization of nomenclatural types, whereby the genomic data generated can be used beyond taxonomic classification and species identification.
The locality and other details of the three T. suessii specimens (lectotype and two paralectotypes) are discussed in the Results section below.
Measurements and vertebral counts, as well as the use of the term "subadult", follow de Carvalho et al. (2002).Analysis of the nomenclatural aspect of the collection specimens of the species was based upon a critical reexamination of relevant publications and unpublished sources deposited in NHMW.

| Extended specimen network approach applied for T. suessii
The approach applied follows the Extended Specimen Network (Lendemer et al., 2020;Webster, 2020) as described in Bogutskaya et al. (2022).A diagram of the procedure is presented in Figure 1.Briefly, all historical information, including cards and labels, were studied and scanned.The original and subsequent descriptions were re-checked, reviewed in detail, and scanned.All available literature in which the specimens are mentioned, as well as data on expedition routes, dates, and sampled localities, were reviewed with the specific goal of clarifying the type locality and the habitat.A map of the Second Austrian Red Sea Expedition is presented in Figure S1.Specimens relevant to the study were photographed and radiographed.

| Molecular analysis and genome assembly
Four pieces of tissue were dissected and weighed.DNA extraction was performed with four-fold replication from two pieces of kidney (Tsue1 and 2) and two of muscle (Tsue3 and 4), using the in-house phenol:chloroform:isoamyl alcohol (P:C:I) 25:24:1 extraction method, which yields the highest quantity of DNA (pers.obs.and, e.g., Hawkins et al., 2022;McDonough et al., 2018).The extraction buffer was freshly prepared to contain 150 μL Tris-HCL (20 mM), 150 μL NaCl (20 mM), 3 mg of SDS, 12 μL DTT (1 M) and 20 μL proteinase K (600 mAU/ml) per sample.After placing the dried sample tissue into the extraction buffer, the samples were incubated at 56°C and shaken at 450 rpm overnight (16 h) until the tissue had completely dissolved.The lysate was handled with care and only wide-bored filtered pipette tips were used.For DNA purification, the digested sample was transferred to a Quanta bio-Phase Lock Gel (Light) reaction tube to eliminate interphase-protein contamination during extraction.The DNA was washed three times with P:C:I (25:24:1) and once with C:I (24:1) by adding the same volume as the lysate-extraction buffer.The DNA was precipitated with a 1/10 volume of sodium acetate (NaAC; 3 M, pH = 5.2), and 2 volumes of ice-cold ethanol (70%), and left in the freezer (−20°C) for at least 1 h before final elution in 20 μL ultrapure water.
The quantity of double-stranded DNA was assessed by fluorometry (Qubit; ThermoFisher Scientific) using the double-strand DNA Broad-Range Assay Kit.The average DNA fragment length was measured on the TapeStation system (Agilent) using High Sensitivity DNA Screen Tape.
The extract with the longest average DNA fragment length was sent to Novogene for whole genome sequencing.For Illumina low-coverage genome sequencing (aiming at 20× coverage), a short-read library (insert size, 350 bp) was prepared using the NEBNext Ultra II Kit (Illumina) followed by 150 bp paired-end sequencing on an Illumina NovaSeq 6000 platform (San Diego, CA).
The complete mt genome was assembled independently in Geneious v. 10.2.6 (http://www.geneious.com) using a subset of 10 million reads, which allowed for sufficient coverage (>100) yet decreased the time needed for analysis.The reads were first blasted against a custom database based on the reference genome of T. marmorata, deposited in GenBank under MT274576 (Kousteni et al., 2021); the 'low complexity filter' button was unselected and max E-value set to 1e-10.The reads corresponding to the reference genome were designated to the 'Hits' folder, and aligned to the reference genome using the 'Align/Assemble' and 'Map to reference genome' functions, using High Sensitivity iterations up to five times and the 'do not trim' option.Finally, the complete mt genome was annotated with the 'Annotate & Predict' function, using the same reference -MT274576 -and a similarity of 85%.

| Phylogenetic reconstruction and pairwise distances
A detailed phylogenetic reconstruction of Torpediniformes was not the aim of the present study; rather phylogenetic analyses were performed at different hierarchical levels to infer the species status and phylogenetic position of the T. suessii lectotype.For this purpose, the barcoding region cytochrome oxidase I (COI) and the gene nicotinamide adenine dinucleotide dehydrogenase subunit 2 (NADH2) were used, including all available sequences of Torpediniformes of sufficient length and quality deposited in GenBank.As many of the sequences originate from unpublished data, it is hard to judge the reliability of the species identification of the specimens from which they were extracted.For more information, see Table S2.
For COI and NADH2, sequences were downloaded from GenBank (Table S2).Following various molecular studies (see Introduction) we chose a comprehensive outgroup that covers batoid genera Platyrhina, Platyrhinoidis, Raja, Rhinobates, Pristis, Potamotrygon, and Dasyatis.Sequences were aligned using ClustalW (Thompson et al., 1994) and overhanging ends were truncated to match the shorter sequences downloaded from GenBank.Subsequently, the sequences were collapsed to unique haplotypes with FaBox v1.5 (Villesen, 2007), and both trees were calculated with Bayesian inference implemented in Mr. Bayes v3.2 (Huelsenbeck & Ronquist, 2001;Ronquist & Huelsenbeck, 2003) using the GTR substitution model with gamma-distributed rate variation across sites, and a proportion of invariable sites.For COI, the program was run for 1 million generations, after which the standard deviation of split frequencies would fall below 0.01 and the scale reduction factor for all parameters was 1.0.Parameters were sampled every 500 generations.The first 25% of the trees were discarded as burn-in with the rest used to build a consensus tree based upon a 50% majority rule.For NADH2, the program was run for 10 million generations to ensure that standard deviations of split frequencies fell below 0.01.
The number of base substitutions per site between and within groups and their standard error estimate(s) were calculated on COI datasets with the Tamura 3-parameter model (Tamura, 1992) implemented in MEGA6 (Tamura et al., 2013).The dataset included only Torpedo species and, as an outgroup, Tetronarce nobiliana, and was calculated using 1000 bootstraps.The sequences were grouped according to species and geographical data.For the NADH2 dataset, the same method was used.As mostly only one sequence per species or locality was used, the pairwise distances were based on sequences and no groups were defined.
Using the same two aligned fasta files as for the pairwise distance analysis, Assemble Species by Automatic Partitioning software (ASAP, with Jukes-Cantor model; Puillandre et al. 2021) was used to estimate the number of species partitions in the datasets.
Beyond these two single genes, only limited data are available for comparison within Torpediniformes.Nevertheless, two additional hierarchical levels of phylogenetic analysis were conducted to show the potential of whole genome sequencing of type specimens.Thus, phylogenies based on complete mt genomes and nuclear BUSCO genes were constructed, which offer the possibility of resolving intra-and interfamilial phylogenetic relationships.For the phylogenetic analysis, one complete mt genome of T. marmorata (GenBank number NC_059941), as well as two of Narke and four of Narcine were downloaded, with outgroups from the genera Raja, Potamotrygon, Dasyatis, Rhinobatos, and Pristis (Table S2).The sequences were aligned in Geneious 10.2.6, using the alignment tool MUS-CLE and eight iterations.The control region was cut out of the alignment and the phylogenetic tree calculated with Mr. Bayes v3.2 (Huelsenbeck & Ronquist, 2001;Ronquist & Huelsenbeck, 2003) using the GTR substitution model with gamma-distributed rate variation across sites and a proportion of invariable sites.The program was run for one million generations, after which the standard deviation of split frequencies would fall below 0.0000001 and the scale reduction factor for all parameters was 1.0.Parameters were sampled every 500 generations.As with analysis for COI and NADH2, the first 25% of the trees were discarded as burn-in and the rest used to build a 50% majority rule consensus tree.
For phylogenetic reconstruction with BUSCO genes, 15 genomes of other Chondricthyes available in the Genomesync database (https://genom esync.org/)were downloaded, together with the genome of Danio rerio, which was used as an outgroup.Subsequently, amino acid sequences of the same orthologs, which were completely assembled in T. suessii with the described pipeline, were downloaded and aligned with MAFFT (Katoh & Standley, 2013), and each gene aligned separately.The aligned sequences were concatenated and the phylogenetic tree reconstructed with RAxML and the Prot-Gamma-GTR model to account for the nature of the raw data (amino acid sequences).Finally, 1000 bootstrap replicates were conducted, before plotting the phylogenies in R with the ggtree package (Yu et al., 2017).Original publication: Steindachner, 1898a: 199.The original description (Steindachner, 1898a: 199; Figure S2) is based on a 30-cm-long specimen and included also an unindicated number of syntypes.The fact that there was more than one described specimen is evident from the statement "Scheibe kreisrund, nur bei einem Exemplare von 30 cm Länge am Vorderande quer abgestutzt" (Disc circular, only in one specimen of 30 cm length transversely truncated at the front edge).

3
Steindachner (1898b: 784-786) subsequently published a more detailed description based on the 30-cm-long female and two other specimens: young males 13 and 24 cm in length, with the 13-cm-long male life-sized illustrated ("junges Männchen; in natürl.Grösse") (Plate II, Figure S5a), with small claspers not visible in the dorsal view.
The specimen had been long ago mislabeled and considered lost (De Carvalho et al., 2002), but was recently found again.
Clarification of the dates of publications of the journals issued by the Mathematics and Natural Sciences Class of the Imperial [Austrian] Academy of Sciences in Vienna [Matematisch-Naturwissenschaftliche Klasse, Kaiserliche Akademie der Wissenschaften in Wien] confirmed that Steindachner (1898a) predates Steindachner (1898b).The former was published in Anzeiger der Akademie der Wissenschaften in Wien (XXXV.Jahrgang.Nr.XIX; meeting of 14 July 1898) a few days after 14 July but prior to 13 October 1898.The latter was published in Sitzungsberichte, Akademie der Wissenschaften in Wien, Mathematisch-Naturwissenschaftliche Klasse, Bd. 107 after 15 December 1898 and before 19 January 1899 when it was announced as published (Anzeiger, 36: 13) at the meeting held that day.
The type status of the two specimens thought to be syntypes was apparently resurrected by Bernard Séret (at IRD-MNHN, Natural History Museum in Paris) and Barbara Herzig (curator of the fish collection, NHMW, at the time) in 1988 (judging by records on the cards).Designation of the lectotype (NMW 88240) was made by de Carvalho et al. ( 2002): 24-25, figure 14; the species name misspelled as "suessi" and figure 14a is vertically flipped so that the dorsal view is not natural but mirrored.
Lectotype: NMW 88240 (Figure 2a-c).Recent measurement: TL 296 mm (historical and recent measurements differ slightly because of some shrinkage of the body due to long-term deposition in alcohol).
Type locality, date, and collector: Neither the precise location nor collector are given in the original description (Steindachner, 1898a), apart from the fact that the specimen was, together with other species described therein, collected during Austrian expeditions to the Red Sea from October 1895 to May 1896 and September 1897 to March 1898 (von Pott, 1898(von Pott, , 1899)).Later (Steindachner, 1898b: 786), the locality is specified as "a side bay of the port of Perim and at Mocca" (Figure S1).
For the lectotype, the original jar label is completely damaged and unreadable.The subsequently attached label states the locality as "Perim oder Mocca" (Perim Island, Strait of Mandeb/Mokha (al-Mukhā), both Yemen, Red Sea) and the date December 1897; collected during the Second Red Sea Expedition.An old card in a catalogue of the fish collection by Steindachner (Steindachner's handwriting; found only recently, Figure 3a Published data on routes, dates, and sampling stations of the Second Austrian Red Sea Expedition (Luksch, 1898(Luksch, , 1901;;von Pott, 1899) indicate that studies in the area of Perim were undertaken during the period December 2-6, 1897.Luksch (1901)  Steindachner participated in the expeditions and curated the implementation of scientific research until December 8, 1897, it is likely that he was the collector of the lectotype.
Steindachner's card, under number 7137 (Figure 3b), reads "Torpedo Süssii Steind.Mocca, [acquisition number] 15/12 897-98 224b.II.R. M. Exp. 1 Exp.jun.male" and apparently refers to either the NMW 78016 paralectotype, a subadult male (Figures S3 and S4) or the NMW 87406 paralectotype, juvenile male (Figure S5).The date of collecting is given as 15 December 1897.Sampling in the area of Mocca (Mokha, al-Mukhā) was undertaken during the period 13-16 December 1897 (Luksch, 1901;von Pott, 1899).As Franz Steindachner had already left the Pola by this date, the collector would appear to be Paul Edler von Pott, who described (von Pott, 1899: 35) the collecting of several Torpedo rays on December 15 as "… since the weather remained clear throughout the day, our work could be carried out regularly and without problems and finished on the evening of the 15th.Two trawls with the trawl net, which were made in the morning of the 15th, despite the unfavorable bank conditions, when the wind had not yet freshened too much, provided us with a very welcome addition to our collections -among other things also several more specimens of one of the Red Sea new, blue-spotted Torpedos." According to Luksch (1901), no stations were implemented on 15 December 1897, though he provides data for proximate dates as follows: Station 347 on 14 December; the Mokka (Mokha, al-Mukhā) roadstead, at anchor (no coordinates given; present, ca.13°19′11″ N 43°14′07″ E), depth 0-8 m; water temperature 24.9-25.0°C,air temperature 25.Preservation: Primary fixative used for the discussed specimens is not known.However, though formaldehyde (and its aqueous solution under the commercial name formalin or formol) was introduced as a biological reagent (a fixative for biological objects) in 1893-1894 after the publications of Ferdinand Blum (e.g., Blum, 1894) (for details see Fox et al., 1985), it seems that alcohol was used rather than formalin during the Red Sea expeditions.This speculation is further confirmed based on the quality -average length -of the isolated DNA fragments (see 3.2), which are typically much shorter in the DNA isolated from formalinpreserved specimens (personal observation).
3.1.2| Morphological description of the lectotype and the paralectotypes Measurements and counts are reported in Table S3; for general appearance (dorsal and ventral views) and radiographs see Figure 2 and Figures S3-S6.In female lectotype, disc wider than long, broadly rounded in outline with truncate anterior margin, without any median protuberance anteriorly while it is longer than wide with slight median projection in male paralectotypes.It cannot be excluded that this difference is due to sexual dimorphism; in any case, similar variability was found in other Torpedo species (de Carvalho et al., 2002: 9, figures 1-10).Posterior pectoral faintly overlapping origins of pelvic fins.Preorbital length (6%-7% TL) equal to or slightly exceeding prenasal length.Spiracle relatively small, length of its opening (2% TL) smaller (markedly smaller in subadult male paralectotype) than orbit diameter (2.5%-4% TL).Spiracular papillae knob-like and weakly pronounced in female lectotype except for larger, pointed postero-central papilla, 8/9, or more pronounced (both male paralectotypes), triangular, with postero-central papilla and one lateral papilla markedly larger than others, 7/8 and 8/10.Eye and spiracle relatively closely set, with space between them roughly equal to orbital length (2.5% TL in lectotype), while in subadult male paralectotype eye larger (3.4% TL), and eye and spiracle relatively closer together, with space between them not as great as orbital length.Distance between eyes about equal to interspiracular distance (5%-6% TL).Distance between first gill slits equal to distance between last gill slits.Second and third gill slits of about same length and larger than others, and fifth gill slit smallest.Nasal curtain very short and wide, with straight outer margin, not reaching mouth cleft; nostril with prominent fold mostly surrounding its outer margin located lateral to level of mouth corner (Figure S6).
Pelvic fin originating just anterior to posterior disc margin.Pelvic fin long but not very wide, greatest width of pelvic fin less than one-half of disc width and slightly wider in male specimen.Anterior margin of pectoral fin straight, not projecting laterally in female, and rounded in male.Posterior margin of pelvic fin more convex, with slightly attenuated posterior apex.Tail relatively short and stout, 19% TL as measured from second dorsal fin origin, but tail length from cloaca different in female (38% of TL) and male (46% TL).First dorsal fin insertion anterior to level of posterior axil of pelvic fin and first dorsal fin base entirely over pelvic fin bases.First dorsal fin markedly larger than second in both depth and length of base.Distance between dorsal fins about equal (in lectotype female) or shorter (in male paralectotypes) than distance between second dorsal and caudal fins.Elongated claspers devoid of integumental flap in subadult male (Figures S3  and S4).In juvenile male (Figure S5c), claspers much shorter and slender, and not seen in dorsal aspect.No dermal denticles.
Teeth densely set in a quincunx arrangement, with overlapping bases, roughly similar in shape and in about 30 rows in either jaw.Flat-based crown with sharp and tall, slender cusp located about middle of base.In occlusal view, crown base rectangular (wider than long), with characteristic four-lobed appearance: lower part of labial face bifurcating into widely diverging, basally curbing lobes while median notch at lingual crown base very shallow.Tips of individual cusps lay over lingual bases of teeth positioned in next inner row (Figure S6a).Teeth of outermost rows with relatively wider bases and smaller cusps, cusps oriented toward lateral mouth corners on all teeth.Total vertebrae, 84: trunk vertebrae, 21; total tail vertebrae, 63; tail vertebrae to second dorsal fin, 26 (lectotype, Figure 2c); 87: 26, 61, and 25, respectively, in the NMW 78016 subadult male paralectotype (Figure S3c), and 86: 25, 61, and 25 in the NMW 87406 juvenile male paralectotype.
Coloration.The specimens (Figure 2; Figures S3, S5), now preserved in alcohol, are bright brown dorsally with eight large dark brown spots on the disc, arranged regularly in lateral rows of three spots in each and two in the midline, with lighter outlines (ocelli).There is also vague reticulation over the disc and disc edges, and smaller dark spots on the anterior disc region, pelvic fins (in the female lectotype, on the right side the anterior pelvic spot merges with the posterior spot from the disc), and dorsal fin bases.As Pott (1899: 35) apparently referred to just collected "new species of Torpedo" as "blue-spotted Torpedo", the dorsal spots were probably blueish (or with blueish tint) in live fish.

Torpedo species
The dorsal color pattern of T. suessii easily distinguishes it from congeners from the northwestern Indian Ocean, the Red Sea, and adjacent gulfs: from (1) T. panthera with small, irregular whitish spots; (2) T. sinuspersici characterized by a vermiculate or marbled whitish pattern over a dark brown background; (3) T. fuscomaculata, though variable but typically with numerous dark spots and blotches, each smaller in size than interorbital space; and (4) T. adenensis with a completely plain color pattern (de Carvalho et al., 2002(de Carvalho et al., , 2016;;Fraser-Brunner, 1949;Peters, 1855).
Torpedo suessii differs further from both T. panthera and T. adenensis by having no median integumental flap in the clasper glans region, and from T. adenensis and T. sinuspersici by smaller total number of vertebrae, 84-87 (vs. 96-102).It also differs from T. sinuspersici in having the base of the first dorsal fin not extending posterior to level of pelvic fin base, i.e., entirely over pelvic fin bases (vs.extending posterior to level of pelvic fin base, i.e., not entirely over pelvic fin bases).Furthermore, T. suessii can be distinguished from T. fuscomaculata by a shorter nasal curtain, with length about a third of internarial width (vs.half), and a longer distance between eye and spiracle, which is longer than the orbit diameter (vs.shorter).
For more details on morphological comparison, see Table S4.

| Molecular analysis, genome assembly
The measurements of DNA quantity and average fragment length are reported in Table 1.Accordingly, Tsue4 was chosen for further analysis.
The sequencing output was 120 GB (143,163,576 reads).The reads were deposited at Sequence Read Archive (SRA) at the National Center for Biotechnology (NCBI) under the project number PRJNA1008877.FastQC analyses of the raw sequencing quality showed that both the forward and reverse reads were of high quality, but characterized by an elevated amount of sequence duplication.The putative genome size was estimated to be approximate 1.1 GB.However, given the level of DNA fragmentation and the corresponding large number of scaffolds in the assembly, this estimate is highly unreliable.After the initial genome assembly undertaken with Spades, post processing was performed using two iterations of Racon polishing based on the trimmed reads.A comparison of BUSCO scores (see below) indicated that this did not further improve the assembly quality.
The descriptive statistics of the draft genome summarized with Quast revealed that the assembly consisted of 1,527,816 sequences of 1 bp or more, and 615,086 sequences of 1001 bp or more.The total length of the assembly was 1,672,532,237 bp; the N50 was 2037 bp and the largest contig had a length of 20,388 bp.The descriptive statistics were summarized further with a SnailPlot using BlobTools (Figure S7), while coverage, GC content, and the putative taxonomic assignment of each contig was visualized as a BlobPlot (Figure S8).The assembled genome was uploaded at NCBI under the project number PRJNA1008877.
The analysis indicated that several of the contigs are not of Chondrichthyes origin but rather from human (8013 contigs; 7.46%) or bacterial (9705 contigs; 9.04%) contamination.Based on these BLAST results, putatively contaminant contigs were removed using a custom Python script.Subsequent BUSCO analysis indicated that only one of the BSUCO genes was lost due to this filtering step, since it was located on a human-specific contig.

| Phylogenetic reconstruction and pairwise distances
A total of 175 sequences of Torpediniformes from Gen-Bank were aligned for the COI dataset along with the lectotype sequence, and seven outgroup sequences (Table S2; see also Figure 4).The trimmed alignment was 594 bp in length.Subsequently, the dataset was collapsed to 83 unique haplotypes.The resulting tree corroborated the monophyly of Torpedinoidea, where the genus Tetronarce was placed as a sister group to Torpedo, with a posterior probability of 1.However, in the Narcinoidea, the genus Narcine was not monophyletic.
The genus Torpedo was split into two well-supported groups (with posterior probability 0.95): (1) T. marmorata, and (2) T. fuscomaculata, T. torpedo, and T. sinuspersici, including the lectotype of T. suessii (Figure 4).Within this group, one sequence (KF489784), determined as T. sinuspersici collected at KwaZulu-Natal (Indian Ocean, South Africa), does not cluster to the same clade as other T. sinuspersici sequences.The sequence JQ350399, determined as T. fuscomaculata collected from north-eastern Madagascar (Indian Ocean), clustered as a sister group to other T. fuscomaculata sequences.The T. suessii lectotype represents a separate lineage forming a polytomy with T. torpedo, T. fuscomaculata, and possibly KF489784 T. sinuspersici (these splits are not supported statistically).
The pairwise-distance analysis based on the COI dataset included 56 Torpedo and four Tetronarce sequences, which were 594 bp long (Table S5).The shortest genetic distance was 2.6%, between the lectotype of T. suessii and T. fuscomaculata (JF494706) collected at Western Cape (Atlantic Ocean, South Africa).The distance between the lectotypes of T. suessii and T. sinuspersici collected from Pakistan was 3.5%, while that between T. suessii and T. torpedo was 4.2%.The shortest genetic distance among other taxa was between T. sinuspersici collected from Pakistan and T. sinuspersici collected from India.Distances within the group were longest in T. marmorata, at 0.003 (see Table S6).
According to the ASAP analysis, the best-fitting number of species partitions for the COI dataset was eight (see Table S7), and considered the lectotype of T. suessii as a separate species partition.The other species corresponded to the clades recovered by the phylogenetic reconstruction (Figure 4), whereby T. sinuspersici (KF489784) and T. fuscomaculata (JQ350399) were considered separate partitions.
In the NADH2 dataset, 41 sequences were used in addition to that of T. suessii, 34 sequences of Torpediniformes and seven sequences for the outgroups from GenBank (see  (Huelsenbeck & Ronquist, 2001;Ronquist & Huelsenbeck, 2003) using the GTR substitution model with gamma-distributed rate variation across sites and a proportion of invariable sites.Posterior probabilities above 0.95 are shown, and nodes with low statistical support are dimmed.
Table S2, Figure S9).These were collapsed to 34 unique haplotypes, with an alignment of 1044 bp.The phylogenetic tree confirmed the monophyly of Torpedinoidea, with two sister genera Tetronarce and Torpedo; but not the monophyly of the genus Narcine (Narcinoidea).T. suessii clustered together with Torpedo sp. from Sri Lanka, T. sinuspersici, T. fuscomaculata, and T. torpedo, which formed a sister group to T. mackayana.
The size of the assembled mitogenome was 17,681 bp.The gene order and content of all nine assembled mitogenomes was as expected for vertebrates and contained 13 protein coding genes, 2 ribosomal RNA genes, 22 transfer RNA genes, the control region (D-loop) and several small noncoding regions (Gen-Bank Accession Number OR506611).In the phylogenetic tree, T. suessii and T. marmorata form sister species (Figure S10).
Following BUSCO analysis, 184 complete genes were assembled by the bioinformatic pipeline described.The reconstructed phylogenetic tree of Chondrichthyes is shown in Figure 5.
According to the ASAP analysis, the best-fitting number of species partitions for the NAQDH2 dataset was nine (See Table S9), with the lectotype of T. suessii as a separate species partition.Other species corresponded to the clades recovered by the phylogenetic reconstruction (Figure S9), whereby Tetronarce nobiliana (JQ518933, JQ518931) and Tetronarce macneilli (JQ518927, JN184080) were recognized as one, and JQ518928 Torpedo marmorata as a separate species.

| DISCUSSION
The importance of undertaking detailed research into historic records to complement the analysis of genomic data of type specimens was highlighted by the present study (its first goal).On the one hand, rediscovery of the presumed lost paralectotype of T. suessii and details on the site of collecting were reported, as historical information and a careful review of the original description reveal the precise type locality of Perim (Strait of Mandeb, Yemen, Red Sea), as well as the depth and salinity at the collecting site.The advantages of undertaking whole genome sequencing of a name-bearing type was also highlighted, as exhaustive molecular analysis resulted in the construction of a phylogenetic tree of Torpediniformes based on complete F I G U R E 5 Phylogenetic reconstruction with Benchmarking Universal Single Copy Orthologs approach (BUSCO) genes, using 15 genomes of other Chondricthyes available in the Genomesync database (https:// genom esync.org/)and with Danio rerio as an outgroup.Amino acid sequences of the same orthologs, which were completely assembled in Torpedo suessii (184 genes), were aligned with MAFFT (Katoh & Standley, 2013), and each gene aligned separately.Furthermore, the aligned sequences were concatenated and the phylogenetic tree reconstructed with RAxML and the Prot-Gamma-GTR model to account for the nature of the raw data (amino acid sequences).Total of 1000 bootstrap replicates were conducted before plotting the phylogenies in R with the ggtree package (Yu et al., 2017).mitochondrial genomes, and of Chondrichthyes based on the amino acid sequences of 184 complete orthologs of highly conserved vertebrate nuclear genes.While the challenges within the genus -such as correct morphological identification (or its limits) -lie beyond the scope of the present paper, a comparison based on published morphological data (Bonfil & Abdallah, 2004;Cadenat et al., 1978;de Carvalho et al., 2002de Carvalho et al., , 2016;;Ravali et al., 2019;Roy et al., 2019;Sujatha et al., 2014;Wallace, 1967), and an examination of NHMW material (Table S4), was successful.Thus, consistent with the second goal of the study, the advantages of an integrative approach for analysing and digitalization of type material was clearly demonstrated.
Museum collections are the most important source of material for the genetic analysis of extinct (e.g., Feigin et al., 2017) and critically endangered species (e.g., Dussex et al., 2021).T. suessii is categorized as critically endangered, or possibly extinct, since only one other published record exists in addition to the three specimens described by Steindachner (1898aSteindachner ( , 1898b)); an image of an adult male Torpedo (under the name ocellated torpedo ray, Torpedo sp.(1) photographed by J.E. Randall was published by Michael (1993: 79), and later by De Carvalho et al. ( 2002): figure 16), Bogorodsky & Randall (2019: fig. 14.1a), and de Carvalho (2022: Plate 54).The specimen was recorded, but not collected, from the Red Sea near Sanganeb Atoll off Port Sudan.De Carvalho et al. ( 2002) and de Carvalho (2022) supposed that the specimen may constitute a further record of T. suessii and the differences in color pattern might represent ontogenetic or intraspecific variability.Nevertheless, as discussed below, the color patterns exhibit some stability as both males, markedly different in length, are completely similar in their color pattern.In congruence with the suggestion of intraspecific variation, the color pattern of the photographed specimen, with a pale brownish reticulate background and five well-defined round dark spots (without a whitish rim) of relatively medium size, is different from the two male paralectotypes.Furthermore, Michael (1993) analysed the difference from three other species in the Red Sea (not only in coloration but also in position of the spiracles and shape and size of the dorsal fins) and concluded that it represents a different (undescribed) species.Similarly, the photographic evidence was not accepted as T. suessii by Kyne et al. (2017), making the assignment of this record to T. suessii doubtful.Thus, besides the three specimens collected almost 130 years ago and housed in NHMW, T. suessii was never observed in nature.This makes the availability of the sequence data for the feature studies even more valuable.
Due to its benthic lifestyle, T. suessii is elusive to scientific research, a characteristic it shares with other torpedo rays (Jabado et al., 2017;Moazzam & Osmany, 2021), as well as many other marine fish species (Fraija-Fernández et al., 2020).Inaccessibility to traditional surveys may be partially overcome by the use of environmental DNA (eDNA), and there is increasing evidence showing the benefits of eDNA for monitoring of marine fish diversity (e.g., Rourke et al., 2022, and references therein).However, there is also a problem of missing genetic data, as only about one-third of marine fish species are represented in the eDNA databases, and reference databases first need to be built (Andruszkiewicz et al., 2017;Curd et al., 2018;Gold, Curd, et al., 2021;Gold, Sprague, et al., 2021;Gold et al., 2022).Museum collections are, therefore, an invaluable resource not only for rare (endangered or extinct) taxa but also for material from inaccessible geographic areas, and assortments of same species throughout their range (Raxworthy & Smith, 2021).In congruence, the number of studies based on museum specimens has been increasing rapidly in recent years, with new milestones constantly being reached due to evolving methods in the field of historical DNA research (Kjaer et al., 2022;Lalueza-Fox, 2022).The falling costs of sequencing have enabled biodiversity assessments made from museum collections to become a reality.Nevertheless, extracting and analysing historical DNA remains challenging, depending on the preservation condition of the sample (Hawkins et al., 2022;Zimmermann et al., 2008).In the present study, the samples collected from kidney tissue resulted in much higher DNA quantity than from skin tissue (Table 1), which traditionally is sampled from sharks and rays (e.g., Moore et al., 2011), confirming the idea that the choice of tissue is important (McDonough et al., 2018).In addition, contamination of samples with exogenous DNA was relatively slight (7%-9%; pers.obs.), with the average fragment length of the extracted DNA remarkably long (250-300 bp, vs. 150 bp or less; Zimmermann et al., 2008).Thus, after more than 125 years of ethanol storage under sub-optimal conditions, the extracted DNA enabled recovery of not just the complete mt genome but also an assembly of a low-coverage draft genome.Besides difficulties with DNA extraction and analysis, the lack of associated data for museum specimens and the limited trustworthiness are also problematic (see, e.g., false information on the label as described in Palandačić et al., 2020 and the references therein).A thorough evaluation of the historical information is thus essential for correct interpretation, as in the present study, where a historical card of Steindachner was found that helped to clarify the exact type locality (Figure 3).
A complete, in-depth morphological and molecular analysis of all type specimens in big (with thousands of specimens) collections is unrealistic.However, by providing the data in the form of an Extended Specimen Network, including photographs and radiographs, any interested scientist working on the group could easily incorporate them into their studies.Analysis of historical DNA, especially, needs expertise and a dedicated laboratory (Fulton & Shapiro, 2019); yet, the inclusion of genetic and genomic information as part of the digital package makes museum specimens accessible for a much wider spectrum of research.In this way, collections are now open to the broader scientific community, including scientists from disadvantaged or distant countries who cannot easily organize or afford to visit particular collections.For example, T. suessii was first described by Franz Steindachner, who was one of the most prolific curators of the NHMW Fish collection (Irmscher, 2021).He described more than 1000 fish species, of which 837 (5491 specimens) are still present in the collection.Due to his numerous expeditions, and productive career, which encompassed systematics widely, these specimens are of great importance for research groups around the world, particularly those from South America, and also the Middle East and Europe.Complete digitalization will enable access to his material for a wide scientific community working on his material.Meanwhile, digitalization of data means preservation of the type specimens in an electronic form and offers extra protection from catastrophic events (e.g., as in the case of the fire in Rio de Janeiro National Museum), from damage to single jars, their relocation or destruction, or from time, as DNA slowly degrades and the chances of successful analysis decline.Finally, there is an increasing demand on (destructive) sampling of type material, which is often questioned by the curators (see, e.g., "The curator's conundrum" in Raxworthy & Smith, 2021).When sampling irreplaceable material, it is of great importance to use the tissue to gain the maximum possible result i.e., genome-wide analysis.
The present study provides some insight into the phylogenetic position of T. suessii within Torpedo rays.However, it was hindered by the small amount of data that have been published so far.In contrast to the whole-genome approach in this study, molecular data for comparison are only available for five out of ten well-established Torpedo species (six if including the data from the present study) and are restricted to COI and, to a lesser extent, NADH2 sequences representing five and six species, respectively.In addition, for most of the species, only a few (2-3) individuals were sequenced.With this in mind, the gathered morphological and molecular data, in congruence with the opinion of de Carvalho and co-workers (de Carvalho, 2022;de Carvalho et al., 2002de Carvalho et al., , 2016)), indicate that T. suessii can be considered a valid species within Torpedo, at least to the point where a more complete dataset allows for a comprehensive revision.In addition to being visible as a separate clade on both phylogenetic trees, of COI (Figure 4) and NADH2 (Figure S9), T. suessii as a separate species partition is supported by the ASAP analysis of both datasets.The pairwise genetic distances within the clades are much shorter than between them (Tables S5  and S6).According to the COI dataset, the shortest such distance was found between the lectotype of T. suessii and T. fuscomaculata from Western Cape, South Africa (2.6%).Nevertheless, these taxa differ morphologically in the length of the nasal curtain, distance between the eye and spiracle, and length of the caudal fin, as well as coloration.According to these data, there seems to be no support for synonymizing T. suessii and T. fuscomaculata.The secondleast distant clade based on COI is T. sinuspersici, which according to Fraser-Brunner (1949) and Randall (1995) is considered a senior synonym of T. suessii based on morphological data.However, current knowledge suggests that the distribution ranges of T. sinuspersici and T. suessii overlap in the Red Sea (e.g., Carpenter et al., 1997;de Carvalho et al., 2002de Carvalho et al., , 2016;;Randall, 1995), and the sampling localities of the studied specimens -the Gulf of Oman and Strait of Mandeb -are proximate.Thus, the genetic distance between T. suessii and T. sinuspersici cannot be explained by the geographic distance between the sampling sites.Moreover, the morphological analysis performed in the present study, as well as the one reported in de Carvalho et al. (2002), suggests that T. suessii clearly differs from T. sinuspersici in the size of inter-spiracular width, number of total vertebrae, length of the first dorsal fin, and distances between the dorsal fins, again supporting the two as distinct species.
Besides T. sinuspersici, faunal revisions and compilations recently addressing the western Indian Ocean, Arabian Sea, Red Sea, and Persian Gulf recognize three additional valid species of Torpedo in these areas -T.adenensis and T. panthera (Carpenter et al., 1997;de Carvalho et al., 2002de Carvalho et al., , 2016;;Randall, 1995) -making them also sympatric with T. suessii.There are no molecular data available for these two species.Nevertheless, based on coloration (de Carvalho et al., 2002), T. suessii clearly differs from its congeners from the western Indian Ocean, the Red Sea, and adjacent gulfs, including from a proposed synonym, T. panthera, by small, irregular whitish spots.Regarding age and size variability, reliable published data indicate that coloration is not dependent on size and age in Torpedo.For example, in T. panthera and T. sinuspersici the color pattern of juveniles (and even fetus) is essentially the same as in adults (Bhagyalekshmi & Biju Kumar, 2017;de Carvalho et al., 2002).Finally, even when the color pattern is considered variable, as in T. torpedo (Capapé & Desoutter, 1981;El Kamel et al., 2009), T. fuscomaculata (de Carvalho, 2022;Moazzam &Osmany, 2021), andT. marmorata (Michael, 1993), the main characteristic features are nevertheless typical and diagnostic for the respective species (de Carvalho et al., 2002).Therefore, analysis of the combined (available) data points to the conclusion that T. suessii is a valid species.In some cases, the molecular data point to longer genetic distances within Torpedo species (e.g., between T. fuscomaculata from Western Cape, South Africa, and from Madagascar) than between species.Yet, as pointed out in many studies (e.g., Compagno et al., 2022;De Carvalho, 2022;de Carvalho et al., 2016;Ebert et al., 2021;Serena et al., 2020), there might yet be further non-identified or misidentified species in the genus Torpedo, as indicated in the phylogenetic trees of COI and NADH2, and corroborated by the ASAP analysis (e.g., Torpedo sp. from Sri Lanka; T. sinuspersici collected from South Africa).Here, only additional sampling would offer final support for the validity of the number of species in the genus.
Regarding phylogenetic relationships within Torpediniformes, the previously observed monophyly of Narcinidae (Claeson, 2014;Naylor et al., 2016) is likewise put into doubt by our results (Figure 4 and Figure S5).Moreover, the sister relationship of Torpedo and Tetronarce within Torpedinoidea is confirmed.According to several molecular (Aschliman, Claeson, & McEachran, 2012;Naylor et al., 2012Naylor et al., , 2016)), and morphological (Jambura et al., 2023;Villalobos-Segura et al., 2022;Villalobos-Segura & Underwood, 2020) studies, Platyrhinidae (with two genera: Platyrhinoidis and Platyrhina) were placed within Torpediniformes.However, this classification is unsupported by the molecular analysis of COI and NADH2 genes reported here, with both genera clustering with the outgroup (Figure 4 and Figure S5).Yet, as with species-level systematics within Torpedo, the lack of sufficient data on all known or accepted species hinders the construction of an all-inclusive phylogeny.More precisely, Platyrhinidae are missing from the complete mt genome and BUSCO gene analysis, confirming again the necessity to perform further genome-wide analyses of this group, ideally including their type specimens.
Measurements of DNA quantity and average fragment length of extracted DNA.

F
Phylogenetic reconstruction based on the cytochrome oxidase I sequence of the lectotype of Torpedo suessii (NMW 88240), 175 sequences of Torpediniformes downloaded from GenBank (respective numbers are included in the figure), and seven sequences for different outgroups.Reconstructed with Mr. Bayes v3.2