Redescriptions of Dussumieria elopsoides Bleeker, 1849 and related nominal species led to the revalidation of Dussumieria hasseltii Bleeker, 1851 and Dussumieria productissima Chabanaud, 1933 (Clupeiformes: Dussumieriidae)

Abstract Numerous rainbow sardine specimens, previously identified as Dussumieria elopsoides Bleeker, 1849, due to their elongate body, scales without posterior striae, and numerous gill rakers and branchiostegal rays, were examined and confirmed as representing four valid species: true D. elopsoides (restricted to Indonesia), Dussumieria hasseltii Bleeker, 1851 (previously regarded as a junior synonym of D. elopsoides; Arabian Sea to Philippines and Indonesia), Dussumieria modakandai Singh, Jayakumar, Kumar, Murali, Mishra, Singh and Lal, 2021 (Arabian Sea to Ryukyu Islands), and Dussumieria productissima Chabanaud, 1933 (previously regarded as a junior synonym of D. elopsoides; Gulf of Aden, Red Sea and eastern Mediterranean). Each species is redescribed in detail and a key based on morphological characters provided. The phylogenetic relationships among seven species of Dussumieria (including three of the species examined in the present study) were reconstructed from the cytochrome oxidase I (COI) gene, the genetic distinctiveness of each species being indicated by at least 4.1% mean p-distance divergence from the others.


Introduction
Dussumieria Valenciennes, 1847 is a genus of rainbow sardines characterized by 14-18 anal-fin rays, 12-18 branchiostegal rays, the pelvic-fin insertion below the dorsal-fin base, and the isthmus pointed anteriorly (Whitehead 1985;Munroe et al. 1999;. The genus was initially reviewed by Whitehead (1963), who considered it to be monotypic, including only Dussumieria acuta Valenciennes, 1847. Subsequently, Whitehead (1985 considered the genus to comprise two valid species, D. acuta and Dussumieria elopsoides Bleeker, 1849, the latter being distinguished by an elongate body (body depth <22% of standard length vs. >22% in D. acuta), higher counts of lower gill rakers on the first gill arch (21-32 vs. 19-26) and branchiostegal rays (13-17 vs. 12-15), and posterior striae absent on the body scales (vs. numerous longitudinal striae). Whitehead (1985) also regarded two nominal species (Dussumieria hasseltii Bleeker, 1851 and Dussumieria productissima Chabanaud, 1933) as junior synonyms of D. elopsoides, the latter being widely distributed in the Indo-West Pacific. However, recent taxonomic studies have revealed or suggested the existence of more than two valid species, including Dussumieria modakandai Singh, Jayakumar, Kumar, Murali, Mishra, Singh and Lal, 2021, morphologically resembling D. elopsoides, which was described from recently collected Indian specimens. During a revisionary study of the genus, re-examination of specimens previously identified as D. elopsoides following Whitehead's (1985) key numbers indicating the specific gill arch. Distances from the dorsal-fin origin to the pectoral-fin insertion, from the dorsal-fin origin to the pelvic-fin insertion, between the dorsal-fin and anal-fin origins, between the pectoral-fin and pelvic-fin insertions, and from the pelvic-fin insertion to the analfin origin are abbreviated as D-P1, D-P2, D-A, P1-P2, and P2-A, respectively. Osteological characters, including vertebral counts, were observed from radiographs of 7, 14, 33 and 16 specimens of D. elopsoides, D. hasseltii, D. modakandai and D. productissima, respectively. Institutional codes follow Sabaj (2020). All specimens listed in this paper, except for type specimens of D. modakandai, were examined in this study; data for the above type specimens are from Singh et al. (2021). Analysis of covariance (ANCOVA) was performed with EZR (Kanda 2012).
The genetic profiles of 102 specimens, comprising six described and one undescribed species of Dussumieria, were compared using the partial (648 bp) COI gene. This gene was newly sequenced for eight specimens of D. modakandai and 10 of D. hasseltii, and the resulting data were combined with COI sequences published by Lakra et al. (2011), Keskin and Atar (2013), Steinke et al. (2016), Lavoué et al. (2017),  and Singh et al. (2021) (Table I), from which the phylogenetic relationships among the species were inferred. The phylogenetic tree was rooted at its mid-point.
DNA was extracted from 95% ethanol-preserved tissue samples. Polymerase chain reaction (PCR) amplification and sequencing of the COI gene followed standard protocols (Ward et al. 2005), with annealing at 55°C. Amplification of the partial (about 670 bp) COI gene used the following primers: forward COI_FishF1 (5'-TCA ACC AAC CAC AAA GAC ATT GGC AC-3') and reverse COI_FishR1 (5'-CCG GTC TGA ACT CAG ATC ACG T-3'). PCR products were purified, and then sequenced in both directions with Sanger Sequencing technology using the same PCR primers. Sequences generated in this study have been deposited in the GenBank database (accession numbers given in Table I).
Alignments of the COI sequences were determined separately by eye, requiring neither insertions nor deletions. The final alignment combining the gene (for 102 specimens) comprised 648 nucleotide positions. Uncorrected pairwise genetic distances among and within species were calculated with MEGA X (Stecher et al. 2020). The maximum likelihood (ML) method of phylogenetic reconstruction was used to infer relationships within and between species. Weighting all positions equally, the ML phylogenetic tree was calculated under the general time-reversible model of nucleotide substitution with rate heterogeneity following a discrete gamma distribution and an estimated proportion of invariable sites, using the software MEGA X (Stecher et al. 2020). Bootstrap support (500 replicates) was calculated to assess the robustness of each relationship.

Molecular comparisons
Dussumieria hasseltii, D. productissima and D. modakandai diverged from each other (and from other species of Dussumieria) by at least 4.1% COI-based mean uncorrected genetic distance (minmax = 4.1-12.6%) (Table II). In contrast, each species was genetically uniform, with intra-group differentiation not exceeding 1% (except for D. modakandai [1.1%] and Dussumieria acuta [1.8%]). The ML phylogenetic tree using the COI gene ( Figure 1) was fully resolved, with each species forming a well-supported monophyletic group (bootstrap proportion >90%) in agreement with their genetic distinctiveness, thereby confirming their taxonomic status. The genetic results and morphological observations complement each other, supporting the validity of Dussumieria hasseltii, D. productissima and D. modakandai, among other species. COI sequences of two unexamined specimens collected off northern Australia (identified as Dussumieria sp. "Australia" in Figure 1) indicate that the species diversity of Dussumieria is still underestimated.  Turkey  KC500625 3  TR641EK  Turkey  KC500629 3  TR642EK  Turkey  KC500623 3  TR643EK  Turkey  KC500624 3  TR644EK  Turkey  KC500626 3  TR651EK  Turkey  KC500614 3  TR652EK  Turkey  KC500615 3 (Continued )

Morphological comparisons
Welch's T-test for comparison of meristic characters among the four species showed significant differences (p < 5) in at least four characters between each pair of species (Tables III-VI). In addition, ANCOVA of 22 morphometric characters, shown in Tables VII-X, showed significant differences (p < 5) in at least 10 characters between each species. Detailed comparisons with and identification of each species are included in "Comparisons" under each species account, and an identification key to the five species is provided at the end of the taxonomic account. As a result, the validity of D. elopsoides, D. modakandai and two nominal species (D. hasseltii and D. productissima) previously regarded as junior synonyms of D. elopsoides by Whitehead et al. (1988) is confirmed.
Description. Counts and measurements, expressed as percentages of SL, are given in Tables III and VII. Data for the lectotype are presented first, followed by other specimen data in parentheses (if different). Body cylindrical, deepest at dorsal-fin origin. Head and caudal peduncle laterally compressed. Dorsal profile of body elevated from snout tip to dorsal-fin origin, thereafter decreasing to uppermost point of caudal-fin base. Ventral profile of body lowering from lower-jaw tip to pelvic-fin insertion and subsequently rising to lowermost point of caudal-fin base. Abdomen soft, not covered with keeled scutes. Pelvic scute on pelvic girdle, W-shaped. Mouth terminal, small, posterior tip of maxilla not reaching vertical through anterior margin of orbit. Premaxilla rectangular. Single row of conical teeth on ventral margins of premaxilla and maxilla. The 1st and 2nd supramaxillae elongate. Vomer and parasphenoid without teeth. Tooth patches on anterior parts of palatine and pterygoid broad, with numerous rows ← Figure 1. Maximum-likelihood phylogenetic tree of Dussumieria, based on the cytochrome oxidase I (COI) gene (total: 648 base pairs) from 102 specimens, each species forming a monophyletic group. Each specimen is identified by COI sequence GenBank (GB) number followed by Museum Registration Number, and/or sequence origin (GB) and/or specimen code (see text and Table I for details) in parentheses, and geographical origin. Tree rooted at midpoint; branch lengths proportional to number of substitutions; bootstrap proportions (if >50%) indicated at nodes. Abbreviations: GB, GenBank; COI, cytochrome oxidase I.
Description. Counts and measurements, expressed as percentages of SL, are given in Tables IV and  VIII. Data for neotype are presented first, followed by other specimen data in parentheses (if different). Body cylindrical, deepest at dorsal-fin origin. Head and caudal peduncle laterally compressed. Dorsal profile of body elevated from snout tip to dorsal-fin origin, thereafter decreasing to uppermost point of caudal-fin base. Ventral profile of body lowering from lower-jaw tip to pelvic-fin insertion, subsequently rising to lowermost point of caudal-fin base. Abdomen soft, not covered with keeled scutes. Pelvic scute joined to pelvic girdle, W-shaped. Mouth terminal, small, posterior tip of maxilla not reaching vertical through anterior margin of orbit. Premaxilla rectangular. Single row of conical teeth on ventral margins of premaxilla and maxilla. The 1st and 2nd supramaxillae elongate. Vomer and parasphenoid without teeth. Tooth patches on anterior parts of palatine and pterygoid broad, with numerous rows of conical teeth (width of tooth patches on palatine and pterygoid varying with body size; see "Morphological comparisons"). Pterygoids densely covered with small conical teeth (Figure 8). Lower jaw with a single row of conical teeth. Basihyal with small dense conical teeth. Orbit, eye and iris round. Eye completely covered by well-developed adipose eyelid. Vertical slit opening on eyelid exposing central part of iris. Interorbital space flat. Nostrils close to each other, located on middle of snout. Posterior margins of opercle and preopercle smooth. Preopercular margin angular.

1148
H. Hata et al. Coloration of preserved specimens. Dorsum to upper part of lateral surface of body brownish to purplishblack, lateral surface thereafter uniformly pale or silver. All fins pale yellow. Posterior margin of caudal fin dusky. Melanophores scattered on 5th to 9th (rarely 12th) pectoral-fin rays from uppermost ray.
Distribution. Dussumieria hasseltii is widely distributed in the Indo-West Pacific, from the Persian Gulf to the Philippines and Indonesia ( Figure 5).
Morphological comparisons. Dussumieria hasseltii and D. elopsoides have broad tooth bands on the palatines and pterygoids (Figure 8), distinguishing them from all other species of the genus treated herein. Lower gill raker counts on each gill arch also characterize the former two species (Tables III-VI; Figure 9). Detailed comparisons between D. hasseltii and D. elopsoides are given in "Morphological comparisons" under the latter species. Dussumieria hasseltii is further characterized by a pre-dorsal-fin length >58% of SL (vs. <58% of SL in D. elopsoides, D. modakandai and D. productissima) (Figure 6(a)). The form of the palatine tooth patch in D. hasseltii changes with growth, individuals <80 mm SL having the patch restricted to a single or two rows of conical teeth. Additionally, pterygoid teeth are absent (Figure 8  (a)). In specimens 80-108 mm SL, the palatine tooth patch became broad, comprising several rows on the anterior part of the palatine and continuing onto the pterygoid (Figure 8(b)). Individuals >108 mm had broad tooth patches on the anterior parts of the palatine and pterygoid, comprising numerous rows of conical teeth (Figure 8(c)).
Remarks. Dussumieria hasseltii Bleeker, 1851 was described on the basis of 20 specimens (80-160 mm total length) collected from Batavia (currently Jakarta), Cheribon (Chirebon), Samarang (Semarang), and Surabaja, Java, Indonesia, the author noting that the species had distinct teeth on the tongue, palatine and pterygoids, and the dorsalfin origin located on the mid-point of the body. Dussumieria hasseltii defined herein closely matches Bleeker's (1851) D. hasseltii in having numerous conical teeth on the tongue, palatine and pterygoids, and a pre-dorsal-fin length 46.8-50.3% of total length (cf. 44.2-45.4% in D. elopsoides). Therefore, the species described by Bleeker (1851) is most likely D. hasseltii, as defined here. Because no syntypes of D. hasseltii could be found, Whitehead et al. (1966) selected BMNH 1867.11.28.21 (Figure 7 (c)), a specimen collected from Java, Indonesia, as the "putative neotype" for D. hasseltii. Although Whitehead et al. (1966) described that specimen in detail, having shown that neotype designation was necessary, a formal designation was not in fact given, therefore failing to meet the full mandatory requirements listed under Article 75.3.1 (ICZN 1999). To avoid taxonomic confusion among species of Dussumieria, we designate BMNH 1867.11.28.21 as the neotype of D. hasseltii, its Most specimens reported as D. elopsoides from the Indo-West Pacific region since Whitehead (1985) are probably D. hasseltii (see synonym list).  Description. Counts and measurements, expressed as percentages of SL, are given in Tables V and IX. The description is based only on specimens examined in this study (type specimens not included). Body cylindrical, deepest at dorsal-fin origin. Head and caudal peduncle laterally compressed. Dorsal profile of body elevated from snout tip to dorsal-fin origin, thereafter lowering to uppermost point of caudal-fin base. Ventral profile of body lowering from lower-jaw tip to pelvic-fin insertion, subsequently rising to lowermost point of caudal-fin base. Abdomen soft, not covered with keeled scutes. Pelvic scute joined to pelvic girdle W-shaped.

Dussumieria modakandai
Mouth terminal, small, posterior tip of maxilla not reaching vertical through anterior margin of orbit. Premaxilla rectangular. Single row of conical teeth on ventral margins of premaxilla and maxilla. First and second supramaxillae elongate. Vomer and parasphenoid without teeth. One or two rows of conical teeth on palatine. Pterygoids densely covered with small conical teeth (Figure 3(b)). Lower jaw with a single row of conical teeth. Basihyal with small dense conical teeth. Orbit, eye and iris round. Fresh coloration [based on color photographs of KAUM-I. 110283, 110284, 117471 and NMMB-P27124 ( Figure 10)]. Lateral surface of body uniformly whitish-silver. Dorsum dark blue. Dark blue and silver areas separated by light blue and yellowishgreen longitudinal bands. Lateral surface of anterior part of head yellowish-silver. Melanophores densely scattered on head from tips of both jaws to occiput. Iris yellowish-silver, pupil black. Preopercle reddish. Dorsal fin whitish-yellow. Pectoral, pelvic and anal fins semitransparent white. Caudal fin yellowishwhite, with dusky posterior margin. The fresh coloration of the holotype is shown in the original description by Singh et al. (2021, fig . 1a).
Coloration of preserved specimens. Lateral surface of body uniformly pale or silver. Dorsum to upper lateral body surface dark brown or purple. All fins pale yellow. Caudal fin with black margin. Melanophores scattered on 5th to 9th pectoral-fin rays from uppermost ray.
Distribution. Specimens of D. modakandai are known from the Indo-west Pacific from Chennai, Tamil Nadu State, eastern coast of India to Riau Archipelago and the Ryukyu Islands, Japan (Singh et al. 2021;this study). Moreover, molecular evidence indicates that the species is distributed westward to the Mediterranean (Singh et al. 2021; this study, Figure 1). Because BPBM 41429 (collected from Chennai, eastern coast of India) was caught with D. hasseltii (BPBM 41428, 152.0 mm SL) and D. albulina (BPBM 20500, 149.7 mm SL), the three species probably co-occur at least in the Bay of Bengal. Dussumieria modakandai is an abundant bycatch in midwater trawls targeting shrimps, and is mainly used for nutritional purposes in aquaculture in southern Taiwan (Hata 2019 as Dussumieria cf. elopsoides; this study).
Morphological comparisons. Dussumieria modakandai keys to D. elopsoides following Whitehead (1985) and Munroe et al. (1999), having body scales lacking numerous posterior striae, a body depth <22% SL, and branchiostegal rays and 1LGR numbering more than 12 and 23, respectively. However, D. modakandai clearly differs from true D. elopsoides, D. hasseltii and D. productissima in having body scales with 10 or fewer posterior longitudinal striae (Figure 4(c)) (vs. no distinct longitudinal striae in the latter three species; Figure 4(a-c)). Moreover, the narrow palatine tooth patch (1 or 2 rows) in D. modakandai separates that species from D. elopsoides and D. hasseltii (both having a broad palatine tooth patch, comprising numerous rows; Figures 3 and 8).
Distribution. Dussumieria productissima is distributed in the Gulf of Aden (Yemen), Red Sea (Egypt) and eastern Mediterranean ( Figure 5). The distribution of the species in the Mediterranean is considered to represent Lessepsian migration. In addition, molecular evidence indicates that the species is also distributed along the western coast of India (Figure 1), where it co-occurs with D. modakandai.
Although Whitehead (1985), who reviewed the genus, regarded this nominal species as a junior synonym of D. elopsoides, the validity of both nominal species was confirmed in the present study. Moreover, all nominal species included in the genus by Whitehead (1985), except for Clupea flosmaris Richardson, 1846, a name suppressed due to ambiguity in its original description by Opinion 901 (ICZN 1970), are regarded herein as valid species (Hata et al. , 2021Singh et al. 2021;this study).

Key to species identified as Dussumieria elopsoides by Whitehead (1985)
Whitehead's (1985) D. elopsoides was diagnosed by the lateral scales lacking numerous posterior longitudinal striae, a slender body (depth <22% of SL), 1LGR 21 or more, and branchiostegal rays 13 or more.