New Data on the Clupeoids of Lakes Mai-Ndombe and Tumba, Western Democratic Republic of the Congo, Using an Integrated Approach Combining Morphology and Barcoding

Morphological and molecular comparisons of clupeoid fishes in lakes Tumba and Mai-Ndombe in the Congo Basin revealed that a large-sized form of a common species in Lake Mai-Ndombe had not been distinguished from the small-sized endemic Mai-Ndombe dwarf sprat Nannothrissa stewarti Poll & Roberts, 1976 in ecological studies. The large-sized form differs from the Mai-Ndombe dwarf sprat in several morphological characters and resembles the smalltoothed pellonula Pellonula leonensis Boulenger, 1916, but in barcoding analysis appears more closely related to the bluntnosed sawtooth pellonuline Potamothrissa obtusirostris (Boulenger, 1909). This large-sized form is probably a species new to science but will remain undescribed for now, and is here referred to as Pellonulinae sp. BIG. A slender clupeoid, also from Lake Mai-Ndombe, has recently been referred to as the Lake Tumba dwarf sprat Nannothrissa parva (Regan, 1917), which would be the first record of this species from that lake. Specimens of a slender clupeoid collected from Lake Mai-Ndombe for this study were identified as the losera fangtooth pellonuline Odaxothrissa losera Boulenger, 1899. This is the first confirmed occurrence of this species in Lake Mai-Ndombe, whereas the presence of N. parva could not be confirmed. The phylogenetic analysis strongly supports the genus Nannothrissa as currently delimited, here barcoded for the first time, but also confirms the previously reported mismatch between morphological classification and molecular data in this group. We further note that the Mai-Ndombe dwarf sprat attains a larger size than the maximum size noted in the original description in 1976, and it is suggested that an intense beach-seine fishery that utilises mosquito nets might have led to a population decline followed by decreased intraspecific competition, enhancing individual growth. For the same reason, Pellonulinae sp. BIG might have been less common previously and so remained undetected until now.


Introduction
New data on the clupeoids of lakes Mai-Ndombe and Tumba, western Democratic Republic of the Congo, using an integrated approach combining morphology and barcoding and Stiassny 2006;Hardman and Stiassny 2008), as well as a list compiled of species exploited in the local fishery (Luhusu Kutshukina and Micha 2013).
The only and undisputed pellonulin species in Lake Mai-Ndombe is the endemic Mai-Ndombe dwarf sprat Nannothrissa stewarti Poll & Roberts, 1976, but there have been no data for that species or any other pellonulin species in the lake until recently.Micha et al. (2020) studied and discussed the local fishery for N. stewarti and the Lake Tumba dwarf sprat Nannothrissa parva (Regan, 1917), pointing out the prevalent but prohibited use of mosquito nets for beach seining as a threat to small-sized clupeoids.Given the fine mesh size of mosquito nets, fish larvae (even at yolk-sac stage) and the adults of small-sized species may be massively collected (Micha et al. 2020).Nannothrissa parva is a common species in Lake Tumba (Zanga et al. 2019) but has not previously been reported from Lake Mai-Ndombe.The genus Nannothrissa is characterised by vestigial or absent teeth, a deep lower jaw, a low number of pelvic-fin rays (i+6) and small body size, with maximum sizes of 22.5 and 30 mm SL for N. stewarti and N. parva, respectively (Whitehead 1985).However, specimens identified as N. stewarti and studied for stomach contents by Micha et al. (2020) were 29-72 mm SL (JE Svensson, University of Borås, pers.com.), far exceeding the maximum length reported in the original description by Poll and Roberts (1976).Two recent studies by Zanga et al. (2022Zanga et al. ( , 2023) also included relatively large specimens, up to 49.8 mm SL.Furthermore, Stiassny et al. (2021) reported N. stewarti of 40.7-49.5 mm SL from the Mfini and Tshe rivers in the central Congo Basin (Figure 1), suggesting this species could be larger in size and have a more extensive distribution outside the lake.This study combined morphological characters and genetic barcoding data on the clupeoids of lakes Mai-Ndombe and Tumba to confirm the occurrence of N. parva, and to study the large-sized form of N. stewarti in Lake Mai-Ndombe.
Several molecular studies focusing on clupeiform phylogeny and biogeography (Lavoué et al. 2007(Lavoué et al. , 2013;;Wilson et al. 2008) have included representatives of 2-6 pellonulin genera, but not Nannothrissa.For example, Lavoué et al. (2007Lavoué et al. ( , 2013) ) examined a total of seven mitogenomes covering 4 genera, and Wilson et al. (2008) covered 6 genera and 8 species sequenced for the mitochondrial genes 12S, 16S and cytochrome b.Furthermore, Sonet et al. (2019) published an extensive barcoding study of fishes from the region, but it does not include data from clupeids and limits all available pellonulin sequence data on GenBank for the mitochondrial barcoding gene cytochrome c oxidase subunit 1 (COI), including mitogenomes, to only nine species.
To avoid confusion, N. stewarti and N. parva from Lake Mai-Ndombe sensu Micha et al. (2020) are here referred to as Pellonulinae sp.BIG and Pellonulinae sp.SLENDER, based on their relative sizes and body shapes.

Materials and methods
This study followed three major steps: (i) roughly sorting formalin-fixed material to different hypothetical taxa; (ii) identifying the same taxa in ethanol-preserved material, as well as selecting and radiographing vouchers for tissue subsampling and barcoding; and (iii) selecting representative specimens in the formalin-fixed material for clearing and staining, allowing for the detailed study of teeth and jawbones as well as fin ray and ventral scute counts with high accuracy for a subset of the specimens.

Primary material
This collaboration between researchers from the Democratic Republic of the Congo and from Sweden (Micha et al. 2020) meant that some formalin-fixed material collected in 2016 from Lake Mai-Ndombe was deposited at the Swedish Museum of Natural History (NRM).Additional material collected in 2020 and 2021 consisted of both formalin-preserved and ethanol-preserved (for barcoding) specimens from lakes Mai-Ndombe and Tumba (Table 1).Collection with seine nets in the two lakes took place at or close to the localities named in Figure 1.

Morphological data
Methods for counts and measures largely followed Gourène and Teugels (1989).Measurements were restricted to standard length (SL) and body depth (BD), obtained either with a digital calliper from digital x-ray images or with a reticle eyepiece mounted on a stereomicroscope.Cleared and stained (C&S) specimens were prepared following the method of Taylor and Van Dyke (1985).Part of the formalin-fixed material subsequently preserved in 70% ethanol was stained with Alizarin red (Springer and Johnson 2000).Tissue DNA Kit (Thermo Fisher Scientific, Waltham, USA) following the manufacturer's recommended protocol.The CO1 fragment was amplified using the primer combination FishF2-FishR2 (Ward et al. 2005) with the PuReTaq Ready-To-Go PCR kit (Amersham Biosciences AB, Uppsala, Sweden), with 0.5 µL per primer (10 µM = 5 pmol), 2 µL of template, and water added for a total reaction volume of 25 µL.The PCR cycling program was: denaturation at 94 °C for 4 min, followed by 35 cycles of 94 °C for 30 sec, 52 °C for 30 sec, and 72 °C for 30 sec, followed by another 8-min elongation step at 72 °C.The PCR products were electrophoresed through a 2% agarose gel in 1x Tris-EDTA buffer, stained with GelGreen UV fluorescent dye and visualised with an UV-transilluminator.The PCR products were purified by adding 5 μL of a mix consisting of 20% Exonuclease I and 80% FastAP Thermosensitive Alkaline Phosphatase (Fermentas/Thermo Fischer Scientific, Gothenburg, Sweden) to each 25-μl PCR reaction, incubated at 37 °C for 30 min, and then heated to 80 °C for 15 min.Sequencing of both strands was carried out by Macrogen Europe (Amsterdam, The Netherlands).All sequences were proofread and assembled using the software Geneious Prime 2020.2.5 (https://www.geneious.com).All sequences were deposited in GenBank.The voucher specimens and sequences used in this study, along with the GenBank accession numbers, are listed in Table 2.

Molecular data
The CO1 sequences were screened for contamination by searching the GenBank 'nr' database.The sequences were then aligned using the Clustal Omega 1.2.2 (Sievers and Higgins 2014) plug-in for Geneious.The CO1 alignment comprises 100 sequences representing 17 nominal species and is 652 bp long.Alignment is available from the authors.The phylogenetic analysis was performed using the software MrBayes 3.2 (Huelsenbeck and Ronquist 2001;Ronquist et al. 2012).The CO1 data were partitioned according to codon position (first, second, third), and parameters were estimated separately for each partition.For phylogenetic inference, the GTR + Γ + I model was used as suggested by the program ModelTest (Posada and Crandall 1998).The analysis was run for 5 million generations, at which time average standard deviation of split frequencies reported by MrBayes was ≤0.01.Samples were taken every 1 000 generations and the first 25% of the samples were discarded as burn-in.End gaps were coded as missing data.The tree was rooted with Alosa fallax (Lacépède, 1803) and Sprattus sprattus (Linnaeus, 1758).See Supplementary Table S1 for a list of the sequences obtained from GenBank.

Results
Nannothrissa parva was tentatively identified from Lake Tumba and more detailed study of three C&S specimens confirmed the identification.The specimens (both formalin and ethanol) appeared morphologically homogenous.Ten specimens (17.4-31.4mm SL) were included in the barcoding analysis.
Formalin-fixed material from Lake Mai-Ndombe (collected in 2020) appeared morphologically homogenous and tentatively resembled N. stewarti by having no visible teeth and some meristic characters within the reported range of N. stewarti.However, in addition to their relatively large body size (up to 55 mm SL), the 2020 material had an overall higher gill-raker count of ~30, compared with 20 gill rakers in N. stewarti.Radiographs of 10 ethanol-preserved specimens of Pellonulinae sp.BIG chosen for barcoding revealed that eight of the specimens (12.9-44.7 mm SL) have their dorsal fin straight above the pelvic-fin origins, while two specimens (25.5 and 28.3 mm SL) have their dorsal fin aligned clearly posterior to the pelvic-fin origins (Figure 2).These two specimens had lower gill-raker counts and were determined to be N. stewarti.Additional specimens of both species were identified and included in the subsequent analyses (Table 2).
Phylogenetic analyses based on CO1 sequences (Figure 3) place the two Nannothrissa species together and Pellonulinae sp.BIG as the sister clade of the bluntnosed sawtooth pellonuline Potamothrissa obtusirostris (Boulenger, 1909).Querying the BOLD 'All records' database (Ratnasingham and Hebert 2007), N. stewarti, N. parva, and Pellonulinae sp.BIG had no exact counterparts in the database, differing from the most-similar sequences (identified as P. obtusirostris, BIN ID: BOLD:AAF7519) by 9.4%, 9.8% and 1.7%, respectively.Based on the somewhat distinct position of the dorsal fin in relation to the pelvic-fin origins, specimens of N. stewarti were identified in distinction from Pellonulinae sp.BIG in material from 2016 and 2021, which was confirmed by C&S preparation.Notably, 7 of the 8 C&S specimens have only i+5 pelvic-fin rays (Table 3).The largest specimen studied in detail measures 28.4 mm SL.In the abundant material collected in 2016, there were 144 Nannothrissa stewarti (20-30 mm SL) and 198 Pellonulinae sp.BIG (17-71 mm SL) identified.
Vestigial teeth in both species of Nannothrissa are present on the maxilla, appearing like a fine serration with ~20 minute teeth.In Pellonulinae sp.BIG the maxilla is completely smooth and toothless.The Pellonulinae sp.BIG C&S specimens (from 2020) reveal small or vestigial teeth on the premaxilla and dentary (Figure 4).Dentation on the dentary and premaxilla appears somewhat variable but is easily seen in C&S specimens; it was also detected in some specimens preserved in 95% ethanol because of shrinkage of soft tissue.In better-preserved (formalin-fixed) specimens the tiny teeth tend to remain undetected, as they are completely covered by soft, pigmented tissue.
Collecting in 2021 aimed to acquire material that would also including Pellonulinae sp.SLENDER, in addition to N. stewarti and Pellonulinae sp.BIG.This was successful and initial examination of formalin-fixed material suggested that they belong to the losera fangtooth pellonuline Odaxothrissa losera Boulenger, 1899.Our O. losera sequences were identical to sequences identified as O. losera (BIN ID: BOLD:ABW2669), making this the first record of this species in Lake Mai-Ndombe.Two specimens of O. losera were later identified and barcoded from the ethanol-preserved material from Lake Tumba.These two specimens are identical in their CO1 haplotype to specimens from Lake Mai-Ndombe.

Discussion
This study emphasises the importance of correctly identifying species in applied ecological studies, such as by Micha et al. (2020) and Zanga et al. (2022Zanga et al. ( , 2023)).Despite identification problems, the results of those authors remain important and relevant for the conservation and protection of the fish fauna in Lake Mai-Ndombe.The reported intense beach-seine fishery may also provide a clue, or at least give the basis for some ad hoc hypotheses regarding the emergence of Pellonulinae sp.BIG and the larger-sized N. stewarti.In the material we studied in detail, 10 of 17 specimens exceed the previously reported maximum length of 22.5 mm SL.
An obvious question is, why didn't Poll and Roberts (1976) report Pellonulinae sp.BIG from Lake Mai-Ndombe? Parker (1958) showed that thinning in populations of fish sometimes results in increased growth.Thus, an intense fishery that utilises mosquito nets, which can trap copious numbers of both larval fish and adults of small-sized species (Micha et al. 2020), may have led to a marked decrease in the abundance of clupeoids in the lake since the 1970s and allowed N. stewarti to grow larger owing to less competition for food.This could also be the same reason for the rarity and smaller body size of Pellonulinae sp.BIG.Poll and Roberts (1976) included 244 specimens in their description of N. stewarti.However, the paratype RMCA 75 75-50-P-2-3 [Royal Museum for Central Africa] (Figure 3-87 in Musschoot et al. 2021) suggests a juvenile of Pellonulinae sp.BIG more than an adult of N. stewarti, especially regarding the relative positions of the dorsal and pelvic fins.One possible explanation is that this material also included a few juvenile Pellonulinae sp.BIG specimens that were not studied in detail, and consequently the species remained undetected until now.Identification keys (Whitehead 1985;Musschoot et al. 2021), meristic characters (Table 3) and shape of the jaw bones (Figure 4) suggest Pellonulinae sp.BIG is very similar to Pellonula leonensis, except for a truly toothless maxilla and even smaller vestigial, or completely missing, teeth on the dentary and premaxilla (Figure 4), in addition to lower scale counts along the side of the body (Table 3).The relatively deep-bodied appearance, caudal peduncle depth (Figure 2) and shape of the jaw bones (Figure 4) suggest it is a member of the genus Microthrissa.But compared with species of Pellonula, Pellonulinae sp.BIG has several meristic characters, including anal-fin rays and gill-raker counts, that lie far outside the range of variation in species of Microthrissa (Gourène and Teugels 1989).
The slender Potamothrissa obtusirostris has very little in common with Pellonulinae sp.BIG regarding morphology (see also Table 3) with the notable exception of a truly toothless maxilla.However, in the phylogenetic analysis Pellonulinae sp.BIG is sister to P. obtusirostris, and the distance analysis was inconclusive: based on empirical evidence using the CO1 gene as the barcoding region, a pairwise distance of at least 2-3% indicates that the studied groups are distinct biological species (e.g. Ward 2009).Pellonulinae sp.BIG differed from sequences of P. obtusirostris by 1.7%.
Thus, the taxonomic identity of Pellonulinae sp.BIG remains an open question at present.Strong conflict in morphological and genetic data renders a straightforward assignment to a genus impossible.For future studies, properly identified vouchers for several taxa are needed.For now, Pellonulinae sp.BIG will be referred to as Pellonula sp.BIG in the NRM catalogue, but as Pellonulinae sp.BIG in GenBank.It is notable, though not surprising based on previous studies (Lavoué et al. 2007(Lavoué et al. , 2013;;Wilson et al. 2008), that the limited mtDNA dataset only supports Nannothrissa as a monophyletic genus (Figure 3), whereas Microthrissa and possibly Potamothrissa seem to be paraphyletic.
A phylogeny based only on CO1, including a high number of sequences in the ingroup, is not expected to provide strong support.Compared with the phylogeny of Lavoué et al. (2013) based on seven mitogenomes from the taxa in focus (Figure 3), the inclusion of additional genes in  Numbers at nodes are for posterior probabilities.All sequenced samples and comparative material are labelled with their GenBank accession numbers.Bold font denotes mitogenomes reported by Lavoué et al. (2013) the present dataset would most certainly support Pellonula.Microthissa is, however, paraphyletic in both Wilson et al. (2008) and Lavoué et al. (2013).Mitogenomes reported by Lavoué et al. (2013) strongly support Potamothrissa, but their dataset does not include Limnothrissa, Stolothrissa and Pellonulinae sp.BIG.Nannothrissa is included in the phylogeny by Gourène and Teugels (1994) and shows some similarity to the molecular data (i.e. a close affinity between Potamothrissa and Nannothrissa, with one species of Microthrissa as a sister-group).It is fully possible that nuclear DNA data give a different result than mtDNA and morphological data and that extensive introgression within this group of fishes has occurred.If so, that could explain the close affinity between Pellonulinae sp.BIG and Potamothrissa obtusirostris in mtDNA.The latter is widespread in the Congo Basin and close to Lake Mai-Ndombe, but so far has not been recorded from the lake itself (Musschoot et al. 2021).A proper description of Pellonulinae sp.BIG as new to science is justified, but additional comparative material, especially from Pellonula species from the Congo Basin, is needed, preferably also including material reported by Stiassny et al. (2021) from the Mfini and Tshe rivers.

Figure 1 :
Figure 1: Map showing where collecting took place (black circles) at lakes Tumba and Mai-Ndombe in the Congo River basin

Table 1 :
Occurrences and preservation of studied material from lakes Mai-Nombe and Tumba.C&S: cleared and stained; E = ethanol preserved for DNA study; F = formalin-fixed

Table 2 :
Clupeoid specimens barcoded for CO1 in the present study.Body depth (BD) is given as a percentage of standard length (%SL).NRM = Swedish Museum of Natural History, Stockholm