Mussel specimens, preliminarily identified as Cristaria cf. plicata, were collected from three localities of the Mekong Basin in Laos (Tables 1, 2). Tissue snips for DNA analyses were then preserved in 96% ethanol. The shell and tissue samples were deposited in the Russian Museum of Biodiversity Hotspots (RMBH thereafter), N. Laverov Federal Center for Integrated Arctic Research of the Ural Branch of the Russian Academy of Sciences (Arkhangelsk, Russia).

The type specimens, non-type museum lots, and images of Cristaria from Laos, Thailand, Cambodia, and East Asia were studied in the Museum of Comparative Zoology (MCZ), Harvard University, Cambridge, USA; Natural History Museum (NHMUK), London, Great Britain; Senckenberg Research Institute and Natural History Museum (SMF), Frankfurt, Germany; Muséum national d’histoire naturelle (MNHN), Paris, France; North Carolina Museum of Natural Sciences (NCSM), Raleigh, USA; Florida Museum of Natural History (UF), Gainesville, USA; Academy of Natural Sciences of Drexel University (ANSP), Philadelphia, USA; Illinois Natural History Survey (INHS), Champaign, USA; Field Museum of Natural History (FMNH), Chicago, USA; and University of Michigan Museum of Zoology (UMMZ), Ann Arbor, USA.

Occurrences of Cristaria spp. from the Mekong Basin were obtained based on museum data and published references (Fig. 1, Table 2; Suppl. material 1). The map was made through ESRI ArcGIS 10 software (https://www.esri.com/arcgis). The topographic base of the map was formed using free open sources such as Natural Earth Free Vector and Raster Map Data (https://www.naturalearthdata.com), Global Self-consistent Hierarchical High-resolution Geography, GSHHG v2.3.7 (https://www.soest.hawaii.edu/wessel/gshhg), and HydroSHEDS (https://www.hydrosheds.org).

Morphological analyses were conducted using parameters of shell shape, umbo position, structure of pseudocardinal and lateral teeth, as well as muscle attachment scars (Konopleva et al. 2017, 2019). All specimens were compared with the original descriptions and images of nominal taxa. We analyzed 59 shell contours of the Cristaria representatives, including 3 shells of the type specimens of Anodonta bellua, 15 shells of newly collected topotypes, 12 shells of C. plicata from East Asia, and 29 shells of C. plicata from Sundaland (Suppl. material 2). Images of the shells were processed using GIMP v2.10.3 (www.gimp.org). For morphometric analyses, we used Fourier coefficients calculated through SHAPE v1.3 (Iwata and Ukai 2002) as described in previous studies (Konopleva et al. 2017). The results of the Principal Component Analysis (PCA) of Fourier coefficients were visualized using PAST v4.06 (Hammer et al. 2001).

Molecular data of mitochondrial (COI and 16 rRNA) and nuclear (28S rRNA) markers were obtained for the eight newly collected Cristaria specimens (Table 1). Total genomic DNA was extracted from ethanol-preserved tissue samples using the NucleoSpin Tissue Kit (Macherey-Nagel GmbH & Co. KG, Germany), following the manufacturer protocol. Primers for amplification are shown in Table 3. For amplification we applied marker-specific PCR programs as follows: (i) COI: 95 °C (4 min), 27 repeats at 95 °C (50 s), 47 °C (50 s), 72 °C (50 s), and 72 °C (5 min); (ii) 28S rRNA: 95 °C (4 min), 35 repeats at 95 °C (50 s), 57 °C (50 s), 72 °C (50 s), and 72 °C (5 min); and (iii) 16S rRNA: 95 °C (4 min), 29 repeats at 95 °C (50 s), 46 °C (50 s), 72 °C (50 s), and 72 °C (5 min). Forward and reverse sequence reactions were executed on an ABI PRISM 3730 DNA analyzer (Thermo Fisher Scientific Inc., Waltham, MA, USA) with the ABI PRISM BigDye Terminator v. 3.1 reagents kit. The resulting sequences were checked visually using a sequence alignment editor BioEdit v. 7.2.5 (Hall 1999). The sequences were aligned through the MUSCLE algorithm in MEGA11 (Tamura et al. 2021). Uncorrected genetic p-distances between species within the genus Cristaria were calculated through MEGA11 (Tamura et al. 2021).

The mitochondrial phylogeny was based on the COI dataset using 130 sequences of Cristaria spp., including eight new samples of Cristaria bellua from the Mekong (Table 1; Suppl. material 1). Additional sequences of Cristaria were obtained from the NCBI’s GenBank database (Suppl. material 1). The COI dataset was collapsed to 96 unique haplotypes using an online FASTA sequence toolbox (FaBox v1.61; https://birc.au.dk/~palle/php/fabox/; Villesen 2007). The representatives of the subtribe Cristariina (Beringiana beringiana, Buldowskia suifunica, and Sinanodonta schrenkii) and the family Margaritiferidae (Margaritifera dahurica and Gibbosula laosensis) were used as outgroup. Maximum likelihood phylogenetic analysis was performed using the online server for IQ-TREE v1.6.12 (W-IQ-TREE) with automatic identification of the most appropriate evolutionary models (Chernomor et al. 2016) and ultrafast bootstrapping algorithm (UFBoot) with 5000 replicates (Hoang et al. 2017). Models of sequence evolution for each partition were calculated through Model Finder (Kalyaanamoorthy et al. 2017) based on Bayesian Information Criterion (BIC) as follows: 1^st codon of COI: F81 + I; 2^nd codon of COI: TN + G; and 3^rd codon of COI: TN + I. The Bayesian Inference (BI) phylogenetic analysis was performed in MrBayes v3.2.7 (Ronquist et al. 2012) at the San Diego Supercomputer Center through the CIPRES Science Gateway (Miller et al. 2010). The same evolutionary models were implemented in the COI dataset. We used the following parameters: two runs with four Markov chains (three heated and one cold, temperature = 0.2), 15,000,000 generations, and tree sampling every 1000^th generation, 15% of trees were discarded as burn-in and the majority rule consensus tree was calculated from the remaining trees. Convergence of the MCMC chains to a stationary distribution was checked visually based on the plotted posterior estimates using an MCMC trace analysis tool (Tracer v1.7; Rambaut et al. 2018).

For phylogeographic analysis, we trimmed the COI sequences of Cristaria spp. from the initial length of 659 bp to the final length of 615 bp. A median-joining network was constructed through Network v4.6.1.3 software with default settings (Bandelt et al. 1999). Locality for sequences AS13MT01 and AS13MT02 (Zhang et al. 2013), NC_012716 (Jiang et al. 2010), and KM233451 (Wang et al. 2014) was set as Central East China.

A time-calibrated multi-locus phylogeny (3 codons of СOI + 16S rRNA + 28S rRNA) was based on 71 haplotypes of the Unionidae (Suppl. material 1). The best-fit evolutionary models according to BIC were as follows: 1^st codon of COI: F81+I+G; 2^nd codon of COI: K3Pu+G; 3^rd codon of COI: HKY+I+G; 28S rRNA: TIM2+G; and 16S rRNA: TIM2+I+G. Instead of using the estimated best-fit models, we used the less complex HKY model with corresponding distributions for each partition to avoid overparameterization (Bolotov et al. 2017). Calculations were performed in BEAST v1.10.4 with a lognormal relaxed clock and Yule speciation process with continuous quantile parametrization as priors (Suchard et al. 2018). We used three fossil calibration points with Lamprotula, Cuneopsis, and Cristaria as MRCA (Bolotov et al. 2017) with an exponential distribution prior: mean (lambda) = 9.3, offset = 34. The MRCA of the Unionidae was set at 152 Ma (Lopes-Lima et al. 2021; Zieritz et al. 2021) with an exponential distribution prior: mean (lambda) = 2.7, offset = 152. Two independent runs of 25,000,000 generations were processed, with sampling every 1000 generations. The resulting tree sets were combined using LogCombiner v1.10.4 with 10% burn-in. The ESS values were checked using Tracer v1.7 (Rambaut et al. 2018) and each value was recorded as >400. A maximum clade credibility tree has been computed with TreeAnnotator v1.10.4.

Ancestral area reconstruction was based on three algorithms, i.e., Statistical Dispersal-Vicariance Analysis (S-DIVA), Dispersal-Extinction Cladogenesis (DEC), and Statistical Dispersal-Extinction Cladogenesis (S-DEC) implemented in RASP v3.2 (Yu et al. 2015) as described in Bolotov et al. (2017). We assigned two possible ancestral areas of the Cristaria species: (A) East Asia and (B) Sundaland. The three primary models were combined into an integrative model using the Combine Results option of RASP v3.2 (Yu et al. 2015).

The general outlines of newly collected shells of Cristaria from Laos are mainly similar to the type specimens of Anodonta bellua Morelet, 1866, described from the Mekong Basin (Fig. 2). Comparative morphological analyses did not reveal remarkable differences in the shell shape and teeth structure between representatives of Cristaria plicata, the type specimens of Anodonta bellua, and newly collected topotypes. Younger individuals usually differ by a well-developed and high post-dorsal wing, which can be smoothed with aging. Old specimens have large, solid, more ovate-elongated shells with more or less developed wings, sharp and long teeth, and well-marked muscle attachment scars.

PCA based on Fourier coefficients revealed six principal components (PCs) (Fig. 3), among which PC1 and PC2 explained the maximum of the total shell shape variance (69.8% and 12.7%, respectively). However, a Kruskal-Wallis test revealed only one significant component PC2 (P < 0.05; Suppl. material 3). This component reflects the position of the wing and the equilaterality of shells. Three 95% confidence ellipses corresponding to the topotypes of Anodonta bellua, the shells from Mekong Basin, and the shells from East Asia are almost completely overlapped. The lectotype and paralectotypes of Anodonta bellua also lie within the 95% confidence ellipses and mainly correspond to the coordinates of topotypes of Anodonta bellua and specimens from the Mekong Basin.

The phylogenetic analysis revealed five species-level clades of Cristaria, i.e. Cristaria plicata, C. clessini, C. truncata, C. bellua, and one undescribed lineage Cristaria sp. (Fig. 4). Cristaria bellua represents a separate phylogenetic lineage, which is sister to C. clessini with an uncorrected COI p-distance of 8.0 ± 1.0%. Genetic divergences (mean uncorrected COI p-distances, %) calculated between C. bellua and other related species of the genus are shown in Table 4. The topologies of the mitochondrial COI trees were identical for both the Bayesian and ML analyses. Cristaria bellua was recorded as a highly-supported monophyletic clade (1.00/99). Other clades were also well-supported, except for the node leading to Cristaria bellua and C. clessini, which was moderately supported in the ML analysis (BS = 60).

According to the median-joining network of the COI sequences, C. bellua also represents a divergent lineage and shares five unique haplotypes (Fig. 5). Two haplotypes from the Nam Ngum River are more divergent from others and differ by five nucleotide substitutions. The COI haplotypes of C. bellua are more closely related to C. clessini from Japan, but do not connect to other Cristaria species. The largest COI haplotype diversity is observed in Cristaria plicata, especially for individuals from China. Three species, Cristaria truncata, Cristaria sp. and C. clessini connect to Cristaria plicata.

Fossil-calibrated biogeographic modeling (Fig. 4; Suppl. material 4) suggests that the Cristaria MRCA originated within East Asia and Sundaland in the Eocene (mean age = 36.5 Ma, 95% HPD = 34.0–41.2 Ma, BEAST BPP = 1.00) with the probability for an East Asian origin of 76.7% and an East Asia + Sundaland origin of 23.3%. The speciation processes mainly occurred during the Miocene, starting from 20.4 Ma, when the ancestors of two clades (C. plicata + C. truncata and C. bellua + C. clessini) diverged.

The split between Cristaria bellua and C. clessini is placed in the mid-Miocene (mean age 15.8 Ma, 95% HPD = 6.5 – 29.1 Ma), but with a low support value (BEAST BPP = 0.51). This clade most likely originated in East Asia and Sundaland via a vicariance event (probability of 98.4%)

Family Unionidae Rafinesque, 1820

Subfamily Unioninae Rafinesque, 1820

Tribe Anodontini Rafinesque, 1820

Subtribe Cristariina Lopes-Lima et al. 2017

In the present study, Cristaria bellua was distinguished as a distinct species based on molecular and phylogenetic analyses. The COI haplotypes of Cristaria bellua are relatively close to its congener, C. clessini, endemic to Japan (Lopes-Lima et al. 2020). Populations of Cristaria bellua, collected from three localities of the Mekong Basin during the field surveys are rather distant, with each sample having its COI haplotype (Figs 4, 5), which may reflect their isolated existence in these water bodies. Nevertheless, comparative morphological and morphometric studies did not reveal any significant differences from its other congener, Cristaria plicata. Historically, Cristaria bellua has been synonymized with C. plicata by many authors (Haas 1969; Brandt 1974; He and Zhuang 2013; Do et al. 2018; Lopes-Lima et al. 2020), mainly, due to the high level of conchological similarity of both species. The two species are morphologically very similar and may be distinguished using molecular data and biogeographic patterns only.

For a long time, it was supposed that the range of Cristaria plicata covers the major basins of East and Southeast Asia, including the Mekong River catchment area (Brandt 1974; Lopes-Lima et al. 2020; Graf and Cummings 2021). Such a broad distribution was surprising because this species crosses the drainage divides between separate large freshwater systems such as the Mekong and the East Asian rivers. The crossing of separate drainages was described in previous studies of Southeast Asia, for example, the presence of disjunct mussel species populations in the Mekong and Chao Phraya (Konopleva et al. 2021; Pfeiffer et al. 2021). Earlier, Bolotov et al. (2020) noted that “records from the Mekong Basin may represent a historical human-mediated or natural dispersal event”. It is well-known that C. plicata has historically been used for food, pearl, nacre, and jewelry production (Landmann et al. 2001; Fiske and Shepherd 2007; Nagai 2013) and was dispersed by man (Schneider et al. 2013). Thereby Schneider et al. (2013) doubted the reliability of the genus Cristaria for paleogeographic reconstructions. In contrast, our molecular analyses showed very high COI genetic divergence between Cristaria bellua and all its congeners (uncorrected p-distance > 8.0%), and point to long-term isolation in the Mekong Basin with a subsequent speciation event.

Multiple fossils of Cristaria were found among the Paleogene Na Duong, Cao Bang, and Rhin Chua mollusk assemblages (Böhme et al. 2013; Schneider et al. 2013). These records were used in our fossil-calibrated phylogenetic modeling as one of the calibration points. According to our model, the diversification of Cristaria mainly occurred in the Miocene. The MRCA (Middle Miocene) of Cristaria bellua from the Mekong Basin and C. clessini from Japan suggests a probable ancient stream capture between the Mekong and the East Asian rivers and the start of the separate evolution of these lineages since this event. The moderate to high level of phylogenetic support for a node of the sister species Cristaria bellua and C. clessini as well as a high genetic distance between these lineages may indicate the presence of putative additional taxa, which have not yet been sampled, for example from remote and understudied areas of northern Laos, China, Vietnam or the Korean Peninsula.

Several geological studies support the hypothesis of an ancient stream capture between the paleo-Mekong and paleo-Red rivers. According to the geomorphological reconstruction of Clark et al. (2004), in the past, the Upper and Middle Yangtze, Upper Mekong, Upper Salween, and the Tsangpo (Upper Brahmaputra) rivers drained together to the South China Sea through the paleo-Red River. In this work, the Mekong River is considered a major tributary of the paleo-Red River system before the elevation of southeastern Tibet (Clark et al. 2004). Conversely, analyses by Hoang et al. (2009) showed that the Mekong headwaters (as well as the Ayeyarwady and the Salween) have not been connected to the Red River since the Late Miocene. Furthermore, the drainage of the paleo-Red River included the middle part of the present Yangtze and the headwaters of the modern Pearl River (Clift et al. 2008; Hoang et al. 2009).

There are examples of East Asian fish genera, which cross the Mekong drainage divide, with one or several species being endemic to the Upper Mekong. For instance, representatives of Pareuchiloglanis species inhabit the Upper and Middle Yangtze, Red, and Pearl River basins and at the same time the Mekong drainage (Li et al. 2020). The same distributional patterns are typical for another fish genus, Vanmanenia, which occurs in the Upper Yangtze, Pearl, and Red River drainages and the Upper Mekong (Lancangjiang) (Li et al. 2019).

Recently, the representative of the typical East Asian fish genus Carassius Nilsson, 1832 was described from the Upper Nam Ngum River in the Mekong Basin in central Laos, i.e., C. praecipuusKottelat 2017 (Kottelat 2017). As Kottelat (2017) claimed, the Nam Ngum River together with the Nam Ngiep and Nam Neun have headwaters on the Plain of Jars. One of these rivers, i.e. the Nam Neun, is adjacent to the Ma River Basin and flows through Vietnam into the South China Sea (the Gulf of Tonkin). The fact that the three rivers are separated only by very low divides may suggest a faunal exchange in the past, which has already been shown for a few species (Kottelat 2017). Consequently, Cristaria bellua may have been isolated in the Mekong Basin, for example, together with some East Asian fishes.

Further molecular studies and field sampling is necessary to better understand the evolutionary diversification and the biogeography of Cristaria.

Currently, Cristaria bellua is the sole representative of the genus Cristaria and the only native member of the subfamily Unioninae in the Mekong River drainage. Live specimens of Cristaria bellua were recorded in three water bodies of the Mekong basin in Laos, i.e., the Nam Ngum, Nam Don, and Nam Ngiep rivers. The majority of available museum lots of Cristaria from Laos, Cambodia, and Thailand were collected as late as the second part of the 20^th century or even earlier (see Table 2). Among the most recent museum records of Cristaria (January 2016), only shells from one locality of the Mun River, Thailand were registered (UF 507842). Additionally, one Cristaria shell was collected from the Choen River basin in Thailand in 2015 (Nahok et al. 2017). The fact that the present living specimens from Laos were collected from a considerable depth (>1.5–2.0) also revealed that mussels of Cristaria bellua are not easily accessible for sampling.

As mentioned in the taxonomic account, specimens of Cristaria bellua were infested by parasitic mites (Acari: Unionicolidae). However, this is not surprising, because there are multiple observations of mussel-associated mites and their eggs in the congeneric species Cristaria plicata from China (Wen et al. 2006; Wu et al. 2008; Zhang et al. 2018). Probably, Cristaria bellua may serve as a nutrition source and shelter for these parasites (Vidrine 1986; Wen et al. 2006).

The Mekong Basin harbors one of the richest fauna of the Unionidae globally with many endemic taxa (Pfeiffer et al. 2018, 2021; Zieritz et al. 2018; Jeratthitikul et al. 2021; Konopleva et al. 2021). The fauna in Laos is threatened by many anthropogenic pressures such as habitat degradation, water pollution, dam construction, and overharvesting of aquatic animals (Bolotov et al. 2014). Hence, Cristaria bellua requires conservation attention including accurate monitoring efforts, especially because viable populations in other parts of the Mekong, for instance in the type locality, have not been recorded during recent decades (Ng et al. 2020).