Lake Poso's shrimp fauna revisited: the description of five new species of the genus Caridina (Crustacea, Decapoda, Atyidae) more than doubles the number of endemic lacustrine species

Abstract Lake Poso, an ancient lake system on the Indonesian island Sulawesi, harbours an endemic species flock of six, four lacustrine and two riverine species of the freshwater shrimp genus Caridina. In this study, five new lacustrine species are described, bringing the total to eleven species altogether. The number of lacustrine species is more than doubled to nine species compared to the last taxonomic revision in 2009. One of them, Caridina mayamareenae Klotz, Wowor & von Rintelen, sp. nov., even represents the first case of an atyid shrimp associated with freshwater snails which is morphologically adapted to living in shells. An integrative approach was used by providing a combination of morphological, ecological, and molecular data. Based on standard morphological characters, distribution, substrate preferences, and colouration of living specimens in the field, five distinct undescribed species could be distinguished. To support our species-hypothesis based on the mitochondrial genes 16S and COI, a molecular phylogeny was used for all eleven species from Lake Poso. All species form a well-supported monophyletic group, but only four morphospecies consistently correspond to mtDNA clades – a possible reason could be introgressive hybridisation, incomplete lineage sorting, or not yet fixed species boundaries. These results are discussed further in the context of adaptive radiation, which turned out to be more diverse than previously described. Finally, yet importantly, subjecting all new species to similar threats and to the same IUCN category and criterion than the previously described species from the lake is recommended.


Introduction
Lake Poso (Fig. 1) is one of the two so-called ancient lakes systems on the Indonesian island of Sulawesi. This long-lived lake probably is more than 1 million years old (Vaillant et al. 2011) and regarded as a hotspot of biodiversity (von Rintelen et al. 2012). The lake is of tectonic origin and has an area of 323.2 km 2 , maximum depths of 450 m, is oligotrophic with a high transparency and low organic content (von Rintelen et al. 2012). It provides ideal conditions for the evolution of highly diverse endemic species flocks of freshwater organisms such as crustaceans, molluscs and fishes (see review in von Rintelen et al. 2012).
The endemic species flock of atyid freshwater shrimps of the genus Caridina in Lake Poso was first studied by Schenkel (1902) with the description of two new species from the lake itself and one riverine species from the lake's catchment. More than 100 years later, another lacustrine species was described by Cai and Wowor (2007), followed by von Rintelen and Cai (2009), who revised the entire species flock of four lacustrine and two riverine species in the lake system including the description of a new lacustrine and a new riverine species (Table 1).
Here, we use an integrative taxonomic approach to study newly collected material from Lake Poso to a) discover new, so far unknown species from the lake, b) provide a combination of morphological, ecological, and molecular data to describe the newly discovered species, c) provide two different identification keys (a regular key for preserved specimens and a key for pre-sorting living specimens in the field without having to use a microscope), and d) discuss the results in context of adaptive radiation and conservation status of the previously revised Lake Poso species flock.

Materials and methods
Specimens were caught by hand net and preserved in 75-95% ethanol during several fieldtrips to Lake Poso ( Fig. 1) between 2003 and 2019 (for sampling details, see Systematic accounts). Specimens were dissected and morphometrical data were taken using a BMS 143 Trino Zoom dissecting microscope with an ocular grid. Details on setae and mouthparts were observed using a Reichert Biovar compound microscope. Rostral characters were taken from all specimens examined. Drawings were made from microphotographs using Adobe Illustrator following Coleman (2003,2006). The se- tae terminology used mostly follows Short (2004). The two identification keys in this study were modified and updated from the previous keys in von Rintelen and Cai (2009). All new species were described by the first, third and last author.
All material examined is deposited in Museum Zoologicum Bogoriense, Cibinong, Indonesia (MZB), and the Museum für Naturkunde (Museum of Natural History), Berlin, Germany (ZMB). The following abbreviations are used in the text: cl., carapace length (measured from the postorbital margin to the posterior margin of the carapace); ov., ovigerous; E, east; S, south; N, north; W, west.
Maximum Parsimony (MP) analyses were performed using the heuristic search algorithm as implemented in PAUP* (Swofford 2002), with gaps treated as fifth base. Support for nodes was estimated by bootstrap analysis (1,000 fast stepwise-addition bootstrap replicates). Genetic distances were calculated using MEGA X (Kumar et al. 2018).

Results
We distinguished five morphologically distinct and undescribed species that could be separated clearly based on the examination of living specimens in the field (Figs 2E, 3A-C, E-H) and preserved in ethanol that turns specimens completely colourless. In Lake Poso, there are eleven species altogether, comprising a species flock of nine lacustrine and two riverine species (Table 1). This more than doubles the previously known lacustrine fauna of four species (von Rintelen and Cai 2009). Distribution data limited to Lake Poso and reproductive biology (few (5-36), large-sized eggs ca. 0.7-1.1 mm length of developed eggs with eyespots) indicative of direct larval development (Lai and Shy 2009) suggest endemism of all Lake Poso species (see Systematic accounts of this study; von Rintelen and Cai 2009).
Mouthparts and branchiae. Incisor process of mandible ( Fig. 5J) ending in irregular teeth, molar process truncated. Lower lacinia of maxillula ( Fig. 5K) broadly rounded, upper lacinia elongate, with numerous distinct teeth on inner margin, palp slender with few pappose setae and one conical spiniform seta near tip. Upper endites of maxilla ( Fig. 5L) subdivided, palp slender, scaphognathite tapering posteriorly, fringed with long, curved setae at posterior margin. Palp of first maxilliped (Fig. 5M, N) ending in a slender triangular extension. Podobranch on second maxilliped (Fig. 5O) reduced to a lamina. Third maxilliped ( Fig. 5P) with one well developed and one small arthrobranch, ultimate segment of maxilliped shorter than penultimate segment. First pereiopod with a small arthrobranch. Pleurobranchs present on all pereiopods. Epipod slightly reduced (without distal hook) on third maxilliped, absent from all pereiopods.

Pleopods. Appendix masculina
Colouration. Body dark reddish or brown with tiny light bluish dots, well-defined white transversal bands on the first, third, fifth, and sixth abdominal segments (Fig. 3H).
Etymology. The Latin word fuscus refers to the species' dark reddish or brown colouration (Fig. 3H).
Distribution. Caridina fusca sp. nov. is endemic to Lake Poso. Specimens were found at two localities within the lake, in a small bay south of the town of Tentena at the east shore and in a bay at the west shore.
Ecology. Caridina fusca sp. nov. is found under rocks in deep water (more than 5 m depth), while the morphologically similar species C. sarasinorum is usually found on various kinds of substrate like deposits of leaf litter, on wood or macrophytes (von Rintelen and Cai 2009).
Remarks. In life colouration, C. fusca sp. nov. might be confused with C. sarasinorum, also endemic to Lake Poso. In the latter, the transversal bands on the abdomen are less defined and scraggy compared to the sharply defined straight bands in C. fusca sp. nov. In preserved condition C. fusca sp. nov. can be differentiated from C. sarasinorum by the rostrum reaching to the end of the antennular peduncle, the dorsal and ventral margin armed throughout almost to the tip vs. reaching to the distal margin of the scaphocerite or beyond, unarmed in anterior one-third to half of the dorsal margin in C. sarasinorum. Epipods are reduced on the third maxilliped and absent on all pereiopods of C. fusca sp. nov. vs. well-developed on the third maxilliped and first pereiopod, absent on second to fifth pereiopods in C. sarasinorum. The chelae of the first pair of pereiopods are not inflated, 2.29-2.73 times as long as wide, 1.17-1.34 times as long as the carpus in C. fusca sp. nov. vs. slightly inflated, 1.74-2.10 times as long as wide, 1.35-1.48 times as long as the carpus in C. sarasinorum. The carpi of the first pair of pereiopods are more slender (2.33-4.00 times as long as wide) and hardly excavated distally vs. more stout (1.75-2.22 times as long as wide) and slightly excavated distally in C. sarasinorum. Description. Cephalothorax and cephalic appendages. Postorbital carapace length 1.7-3.1 mm (n = 33). Rostrum (Fig. 7A, B) very short, not overreaching distal margin of eyes, clearly convex on dorsal margin, abruptly tapering to a fine tip distally, 0.18-0.33 (median 0.26, n = 20) times as long as carapace, rostral formula 5-10 + 5-10 / 0. Antennal spine well separated from inferior orbital angle. Pterygostomial angle subrectangular. Eyes well developed with globular cornea. Antennular peduncle ( Fig. 7A, G), 0.70-0.79 (median 0.75, n = 6) times as long as carapace in females, 0.92 (n = 1) times as long as carapace in male, first segment 2.08-2.70 (median 2.42, n = 7) times as long as second segment, second segment 1.67-2.75 (median 2.50, n = 7) times longer than third segment. Stylocerite reaching to 0.68-0.89 (median 0.77, n = 6) of first segment of antennular peduncle. Scaphocerite (Fig. 7H) 3.43-4.62 (median 4.02) times as long as wide.
Etymology. Named after the second and last authors' first daughter who is very interested in field work and helped to observe and document this species while visiting the lake in 2019.
Distribution. Caridina lilianae sp. nov. is endemic to Lake Poso. Specimens were found at three localities within the lake, two within a bay south of the town of Tentena at the east shore and one in a bay at the west shore.
Ecology. Caridina lilianae sp. nov. lives on very fine sand or silt (soft substrate) in shallow water (1.5-2.5m). Remarks. With its small size and the less developed chelae with scarce setae at the tip of the fingers, C. lilianae sp. nov. is similar to C. mayamareenae sp. nov. but can easily be distinguished from this species by the very short, convex rostrum (vs. rostrum conspicuous high, reaching to end of second segment of antennular peduncle or slightly overreaching this segment) and the slender third pair of pereiopods bearing long stiff setae on merus and ischium but without any spiniform setae on flexor margin (vs. third pereiopod very robust, without long simple setae on merus and ischium and dactylus with five or six spiniform setae on flexor margin). These characters also distinguish C. lilianae sp. nov. from all other Caridina spp. known from the Lake Poso. Although C. lilianae sp. nov. and C. mayamareenae sp. nov. occur in sympatry in the lake, the microhabitats of these species are quite different. Caridina mayamareenae sp. nov. lives in empty shells of aquatic snails while C. lilianae sp. nov. on soft substrate. The long stiff simple setae attached to the posterior segments of the chelipeds and pereiopods could be interpreted as a morphological adaption to this kind of habitat by preventing them to subside into the soft substrate. This hypothesis would need to be tested, though. In the field, the whitish or creamcoloured body colouration is indiscernible on light-coloured sandy habitats (Fig. 3D). Description. Cephalothorax and cephalic appendages. Postorbital carapace length 1.44-3.07 mm (n = 19). Rostrum (Fig. 9A, B) long and slender, curved upwards or sigmoid, reaching slightly beyond end of scaphocerite or in small specimen, to end of the antennular peduncle, distal 0.16-0.40 (median 0.38, n = 11) of dorsal margin unarmed, ventral margin armed throughout, dorsal teeth more widely spaced distally, 0.95-1.54 (median 1.20, n = 14) times as long as carapace, rostral formula 3-6 (4-6) + 10-20 / 6-18. Antennal spine slightly separated from orbital margin. Ptery-gostomial angle broadly rounded. Eyes well developed with globular cornea. Antennular peduncle (Fig. 9A, B, G), 0.89-1.04 (median 0.96, n = 5) times as long as carapace, first segment 1.56-1.79 (median 1.71, n = 4) times as long as second segment, second segment 2.40-2.80 (median 2.58, n = 4) times longer than third segment. Tooth on distolateral margin of first segment of antennular peduncle acute. Stylocerite reaching to 0.78-0.83 (median 0.80, n = 4) of first segment of antennular peduncle. Scaphocerite (Fig. 9H) 4.10-4.25 (median 4.20, n = 3) times as long as wide.
Etymology. Named after the second and last authors' second daughter who is very interested in field work and helped to observe and document this species while visiting the lake in 2019.
Distribution. Caridina marlenae sp. nov. is endemic to Lake Poso. Specimens were found only at one locality in a bay south of the town of Tentena at the east shore of the lake.
Ecology. Caridina marlenae sp. nov. is found under rocks in deep water (more than 5 m).
Remarks. With its long rostrum, approximately anterior 0.4 unarmed, C. marlenae sp. nov. is similar to C. sarasinorum, C. schenkeli and C. longidigita, all endemic to Lake Poso. In the field, body colouration alone is sufficient to differ C. marlenae sp. nov. from C. sarasinorum or C. schenkeli but it might be confused with reddish specimens of C. longidigita. Caridina marlenae sp. nov. is showing large bright white dots on reddish colouration of the entire body. In C. sarasinorum, the body is coloured dark brown with faint light transversal bands on first, third, fifth and sixth abdominal segments (W. Klotz, pers. observation on the comparative material listed above). In C. schenkeli the colouration of the body is mostly transparent with some brownish or whitish blotches.
Mouthparts and branchiae. Incisor process of mandible (Fig. 12A) ending in irregular teeth, molar process truncated. Lower lacinia of maxillula (Fig. 12B) broadly rounded, upper lacinia elongate, with numerous distinct teeth on inner margin, palp slender with few simple setae and one conical spiniform seta near tip. Upper endites of maxilla (Fig.  11I) subdivided, palp slender, scaphognathite tapering posteriorly, fringed with long, curved setae at posterior margin. End of palp of first maxilliped (Fig. 11J, K) ending in blunt triangular shape. Podobranch on second maxilliped (Fig. 12C) reduced to a lamina. Third maxilliped (Fig. 12D) with one well developed and one arthrobranch reduced to a small worm-like structure. First pereiopod with an arthrobranch. Pleurobranchs present on all pereiopods. Epipod vestigial on third maxilliped, absent from all pereiopods.
Colouration. Body colouration of large females whitish, frequently with broad bright red stripes and blotches, eggs green (Fig. 3A, B), males mostly transparent with some white blotches (Fig. 2E).
Etymology. Named after the fourth author's, daughter for her strong interest in decapod crustaceans her father is working on.
Distribution. Caridina mayamareenae sp. nov. is endemic to Lake Poso. Specimens were found at five localities within the lake, three in the northern part and two at the eastern and western shores in the southern part of the lake.
Ecology. Caridina mayamareenae sp. nov. is hiding inside empty shells of the viviparid snail Celetaia persculpta (P. Sarasin and F. Sarasin, 1898) and Tylomelania spp. (Fig. 3D), and was not observed on any other substrate. On average, 1.4 shrimps were found per shell, but there is considerable variation (0.6-2.4 shrimps per shell) among the examined sites (Table 2). Up to four specimens were found in a single shell at the Dolidi Ndano locality. Caridina mayamareenae sp. nov. is also confined to deeper water; shells from depths of 7 m upwards did not contain any shrimps.
Remarks. Among all species of the genus Caridina known from Lake Poso, C. mayamareenae sp. nov. is unique by its short and conspicuous high rostrum, the less developed chelipeds with scarce setae at the tip of the fingers, and the strong third pair of pereiopods. A high and rather short rostrum is an infrequent character among lacustrine species of the genus Caridina from the Central Lakes of Sulawesi (compare revision in von Rintelen and Cai 2009). Many lacustrine species are showing slender, styliform rostrum shapes as seen in C. ensifera and C. caerulea, the most common species in Lake Poso. The conspicuous high rostrum and the strong third pair of pereiopods adapted for clinging on hard substrate are visible characters of a high grade of specialisation to the microhabitat of this species. Table 2. Abundance of Caridina mayamareenae sp. nov. in shells of aquatic snails in Lake Poso. The numbers in brackets in the "shells" column refer to numbers of shells of Celetaia persculpta / shells of Tylomelania spp.; in the "other taxa" column, the numbers are juvenile gecarcinucid crabs / n Cirolana spp. Description. Cephalothorax and cephalic appendages. Postorbital carapace length 2.6-3.8 mm (n = 36). Rostrum (Fig. 13A-C) very long and slender, curved upwards, reaching far beyond end of scaphocerite, distal 0.5 to 0.8 unarmed, ventral margin armed throughout, most proximal tooth placed below third tooth of dorsal margin in most specimens, 1.35-2.75 (median 2.01, n = 23) times as long as carapace, rostral formula 3-5(4) + 8-14 / 19-37. Orbital margin fused with an antennal spine. Pterygostomial angle broadly rounded. Eyes well developed with globular cornea. Antennular peduncle (Fig. 13A, C), 0.97-1.03 (median 1.01, n = 4) times as long as carapace in females, 1.07-1.19 (median 1.16, n = 4) times as long as carapace in males, first segment 1.48-1.78 (median 1.55, n = 5) times as long as second segment, second segment 2.25-2.88 (median 2.44, n = 5) times longer than third segment. Stylocerite reaching to 0.78-0.88 (median 0.83, n = 4) of first segment of antennular peduncle. Scaphocerite (Fig. 13H) 4.30-5.33 (median 4.75, n = 6) times as long as wide.
Colouration. Body and legs mottled with reddish and white dots arranged in rows, exopod of uropods with a black and white blotch, antennae dark red, chelae white with red fingers (Fig. 3E, F).
Etymology. The specific name is a noun in apposition after the type locality, Lake Poso.

Distribution.
Caridina poso sp. nov. is endemic to Lake Poso. Specimens were found at three localities within the lake, one at the east shore and two at the west shore.
Ecology. Caridina poso sp. nov. lives in packs of debris (small to medium-sized stones) close to the shore of the Lake Poso and thus could be considered a hard substrate dweller as defined in von Rintelen and Cai (2009). The species was never found on soft substrates such as dead leaves, wood or water plants.
Remarks. With its long and upturned rostrum, C. poso sp. nov. is similar to C. ensifera and C. caerulea, two endemic species to Lake Poso. In the field, colouration alone is sufficient to differenciate C. poso sp. nov. from these species. The much smaller species C. poso sp. nov. (carapace length 2.6-3.8 mm) is showing black and white blotches on the exopod of the uropods. In the larger species C. ensifera (cl. 3.5-5.3 mm), a dark red spot is seen on the exopod of the uropods. In C. caerulea (cl. 3.0-4.5 mm), the exopod of the uropods shows an elongate blue blotch (von Rintelen and Cai 2009).
In preserved condition, C. poso sp. nov. can be distinguished from C. ensifera by the absence of epipods on all pereiopods (a vestigial epipod is present on third maxilliped vs. epipods well developed, with distal hooks on the third maxilliped and first and second pereiopods in C. ensifera) and by the higher number of postorbital teeth on the rostrum (3-5 (mode 4) vs. 1-3 (mode 2) in C. ensifera). Further, C. poso sp. nov. differs by its slender chelipeds (chela of first pereiopod 3.1-3.4 times as long as high vs. 2.0-2.8 times in C. ensifera, dactylus of first cheliped 2.0-2.5 times as long as palm vs. 1.0-1.3 times in C. ensifera, dactylus of second cheliped 1.5-2.3 times as long as palm vs. 1.2-1.4 times in C. ensifera) and slender third pair of pereiopods (propodus 16.5-19.1 times as long as wide vs. 10-13 times in C. ensifera, carpus 8.7-10.4 times as long as wide vs. 4.7-6.1 times in C. ensifera, merus 11.3-13.9 times as long as wide vs. 9.2-11.4 times in C. ensifera). In contrast, the dactyli of the fifth pereiopods are shorter (dactylus 2.4-4.7 times as long as wide vs. 5.4-7.0 times in C. ensifera, propodus 4.4-5.3 times as long as dactylus vs. 2.5-3.0 times in C. ensifera). The dactyli of fifth pereiopods are armed with a smaller number of serrate setae on the flexor margin (30-34 vs. 51-57 in C. ensifera). Caridina poso sp. nov. differs from C. caerulea by the absence of epipods on all pereiopods (a vestigial epipod is present on the third maxilliped) vs. epipods well developed (with distal hooks) on the third maxilliped and first and second pereiopods and the higher number of postorbital teeth on the rostrum (3-5 (mode 4) vs. 2-4 (mode 2)) in C. caerulea; further by its shorter telson (telson 0.6-0.7 times as long as carapace vs. 0.8 times in C. caerulea), by the slender chelipeds (chela of first pereiopod 3.1-3.4 times as long as high vs. 1.9-2.7 times in C. caerulea, dactylus of first cheliped 2.0-2.5 times as long as palm vs. 1.1-1.4 times in C. caerulea, carpus of first cheliped 3.1-4.3 times as long as wide vs. 2.1-2.5 times in C. caerulea; chela of second cheliped 3.2-3.9 times as long as wide vs. 2.1-3.2 times in C. caerulea, dactylus of second cheliped 1.5-2.3 times as long as palm vs. 1.3-1.6 times in C. caerulea, carpus of second cheliped 6.5-8.4 times as long as wide vs. 4.1-5.4 times in C. caerulea). The third pair of pereiopods is more slender (propodus 16.5-19.1 times as long as wide vs. 12.9-16.3 times in C. caerulea, carpus 8.7-10.4 times as long as wide vs. 5.9-8.0 times in C. caerulea, merus 11.3-13.9 times as long as wide vs. 9.4-11.8 times in C. caerulea). Merus of fifth pereiopod slender, 11.5-12.9 times as long as wide vs. 8.5-11.3 times in C. caerulea.

Molecular phylogenetics
We used sequences of mitochondrial DNA to investigate the phylogenetic relationship among the species of Caridina from Lake Poso as described above. The resulting sequence alignments have a length of 781 bp (COI) and 540 bp (16S), respectively. In 16S, only two short and largely unambiguous indels (1-2 bp) were required to homologise positions in the alignment.
If support values are considered, the tree topologies reconstructed from 16S and COI are largely congruent (Suppl. material 2, Figs S4, S5). All well supported clades (BPP > 0.9) are found in both trees, while basal splits of both trees, particularly for 16S, are poorly supported.
The molecular phylogeny of Lake Poso species for this study and from the previous study with fewer species (von Rintelen et al. 2007a) revealed similar results: a) All 11 species from the lake as well as its catchment area (Table 1) form a well-supported monophyletic group ( Fig. 15; Suppl. material 2: Figs S1-S3); b) the nine species from the lake proper do not form a monophylum but cluster in separate groups as highlighted in Figure 15, partly clustering with the riverine species Caridina acutirostris and C. schenkeli; c) the match of morphospecies and genetic clades remains as heterogeneous as in von Rintelen et al. (2007a). Only four of the eleven morphospecies consistently correspond to mtDNA clades (C. acutirostris, C. caerulea, C. ensifera, and C. lilianae sp. nov.). Caridina mayamareenae sp. nov. also forms a single clade but contains one specimen of C. fusca sp. nov. Two more species (Caridina longidigita and C. sarasinorum) form distinct clades comprising the majority of sequenced specimens, but not all. A few specimens of these two species are also found within a clade comprising all sequences of C. marlenae sp. nov. and C. poso sp. nov. as well as several specimens of C. sarasinorum, C. schenkeli and the second known population of C. fusca sp. nov. The majority of C. schenkeli specimens form a paraphyletic group with respect to C. caerulea, and this group also includes one sequence of C. sarasinorum.

Lake Poso revisited -new insights from new material?
The molecular phylogeny and field observations (colour pattern, habitat, distribution, behaviour if applicable) were used to test and support the morphological studies of alcohol preserved specimens. The integrative taxonomic approach taken by von Rintelen et al. (2007a) and von Rintelen and Cai (2009) was again successfully applied in this study.
The match of morphospecies and genetic clades remains as heterogeneous as in von Rintelen et al. (2007a). Although all new species are morphologically distinguishable based on the characters described in this study, only four species form exclusively monophyletic groups (Fig. 15). The non-monophyly of the remaining species including three of the new species described here might be explained by introgressive hybridi-  sation or incomplete lineage sorting as discussed for the previously described species. It remains to be seen whether the hypothesis forwarded by von Rintelen et al. (2007a) and von Rintelen and Cai (2009) that the colour patterns of the hybridising species seem to be less obvious and stable than those of the monophyletic taxa, occasionally resulting in mating errors between lake species and between riverine and lake species, holds true. Among the newly described species, Caridina poso sp. nov. and C. marlenae sp. nov. show characteristic, stable, and in the latter also rather conspicuous colour patterns. This hypothesis, however, needs further testing. Another assumption explaining the insufficient resolution by molecular data in most of the new species, which possibly also led to their late discovery, is that their species boundaries are not as fixed yet and they are still in the process of becoming proper biological species. A similar case is known from four closely related Caridina species from Lake Towuti, Sulawesi (von Rintelen and Cai 2009). In a future study, this assumption could be tested and distinguished from introgressive hybridisation by applying a population genomics approach such as, e.g., RAD seq, which has been applied to much the same purpose in Lake Poso ricefishes (Sutra et al. 2019).

Adaptive radiation in Lake Poso
Von Rintelen et al. (2007a) positively tested the hypothesis of an adaptive radiation in the atyid shrimp species flock in Lake Poso, which met at least three of the four criteria defined by Schluter (2000): All eleven species, including those from the catchment area, showed common ancestry, indicative of one colonisation of the entire Poso lake system, and rapid radiation. Apart from the well supported monophyly of the entire species flock, the non-monophyly of the nine lake species are congruent with the results found by von Rintelen et al. (2007a). This time again, the data failed to provide conclusive evidence for an in situ radiation within the lake itself. The earlier study revealed a correlation of phenotype and environment (habitat preferences and divergence in trophic morphology) in Lake Poso species. By exhibiting species-specific and unusual colour patterns, the species seemed to have reached the third stage of adaptive radiation according to Streelman and Danley (2003), i.e., sexual selection suggested as a driving force of diversification (see review in von Rintelen et al. 2020).
The new species cluster within the Poso clade (Fig. 15) are thus part of the monophyletic species flock (common ancestry) that radiated into several specialised ancestors. All new lake species also show pronounced microhabitat preferences -the most extreme example is Caridina mayamareenae sp. nov. (Fig. 3A, B) -along with interspecific differences in cheliped morphology (Figs 6,8,10,12,14) and species-specific colour patterns (Figs 2E, 3A, B), although not always as clear as described in von Rintelen and Cai (2009). The adaptive radiation of shrimps in Lake Poso is thus more diverse than previously assumed, not just in terms of species number, but also with respect to habitat and trophic specialisation. A parallel case of adaptive and largely microhabitatdriven radiation in Caridina is know from the Malili lake system of Sulawesi (von Rintelen et al. 2010;Martin and Richards 2019).

Atyid shrimps in association with other organisms
The occurrence of a Lake Poso species in empty snail shells (i.e., Caridina mayamareenae sp. nov.; Fig. 3A, B, D) is rather unusual, as empty shells of aquatic snails were not reported as microhabitat of atyid shrimp up to now.
In Lake Tanganyika, East Africa, Roth-Woltereck (1958) described the small atyid species Limnocaridina iridinae based on two ovigerous females found in the gill chamber of the bivalve Iridina spekei. Later, no further specimen of this shrimp species was found. Only in November 2010, the first author received a single ovigerous specimen of L. iridinae (now deposited in the crustacean collection of the Oxford University Museum of Natural History, collection number OUMNH.ZC.2012-05-0012) for determination. This time, it was found in an empty shell of the viviparid snail Neothauma tanganyicense. According to the collector, approximately 100 specimens of Iridina spekei were checked for specimens of shrimp without any success (Heinz Büscher, pers. comm.). This might indicate that L. iridinae is not only associated with bivalves but likewise seeks shelter in empty snail shells.
So far, Limnocaridina iridinae and the parallel case of C. mayamareenae sp. nov. from Lake Poso are the only cases of freshwater shrimps in general, and particularly in ancient lake species flocks, associated with molluscs. As a possibly morphological adaptation to their habitat, both species share the less developed type of chelae bearing just some scarce setae at the tips of fingers in contrast to the brush-like dense tufts of setae found in other atyids. A similar association was only described for C. spongicola from the Malili lake system, Sulawesi. This species is associated with an endemic freshwater sponge from Lake Towuti and one of the most extreme specialisations found in the adaptive radiation of Caridina in the ancient lakes of Sulawesi (Zitzler and Cai 2006;von Rintelen et al. 2007b). All three cases might be seen as an example of ecological convergence between species of all three ancient lake systems.

Conservation status and sustainability
Following the IUCN categories, all previously described species from Lake Poso and catchment (Table 1) were assessed as Vulnerable under the D2 criterion (De Grave et al. 2013a-e, Wowor et al. 2013. This was justified based on the limited occurrence of endemic populations and the presence of an introduced fish species as a plausible threat. This originally African cichlid species was again observed in Lake Poso in 2019 (KvR, TvR pers. obs.). The five new Caridina species are thus subjected to similar threats and have a similarly limited occurrence in the lake system as the previously described species. We therefore suggest to include the new species under the same IUCN category and criterion. Currently, the third author and Indonesian colleagues are preparing measures to protect the habitats and fauna of the ancient lakes of Sulawesi, including the atyid species flocks from Lake Poso and the Malili lakes summarised in this study and in von Rintelen and Cai (2009).
The key to pre-sorting living specimens in the field can be used easily without having to use microscopic equipment. Shrimps can, for example, be observed while swimming or snorkelling or by putting them in small fish tanks, and releasing back into the water afterwards. The key can be used for pre-sorting in the field for scientific purpose but also for sustainable capacity building or citizen science projects without having to reduce the populations. This key, however, has not fully been tested in the field and would certainly be an ideal test case for a local citizen science or student project.

Conclusions
Even in relatively well studied areas like the ancient lakes of Sulawesi, the biodiversity of freshwater shrimps has largely been underestimated. An integrative taxonomic approach is the key to the discovery of new species and to a better understanding of the evolution of Lake Poso's fauna. This new knowledge can contribute to the prevention of biodiversity and habitat loss.  Figure S1. Phylogenetic relationships reconstructed by BI analyses of two mitochondrial gene fragments (topology based on concatenated 16S and COI datasets). The scale bar indicates the substitution rate. Original Bayesian posterior probabilities of Fig. 13. Figure S2. Phylogenetic relationships reconstructed by ML analyses of two mitochondrial gene fragments (topology based on concatenated 16S and COI datasets). The scale bar indicates the substitution rate. Figure S3. Phylogenetic relationships reconstructed by MP analyses of two mitochondrial gene fragments (topology based on concatenated 16S and COI datasets). Figure S4. Phylogenetic relationships reconstructed by ML analyses of one mitochondrial gene fragments (16S). The scale bar indicates the substitution rate. Figure S5. Phylogenetic relationships reconstructed by ML analyses of one mitochondrial gene fragments (COI). The scale bar indicates the substitution rate. Copyright notice: This dataset is made available under the Open Database License (http://opendatacommons.org/licenses/odbl/1.0/). The Open Database License (ODbL) is a license agreement intended to allow users to freely share, modify, and use this Dataset while maintaining this same freedom for others, provided that the original source and author(s) are credited. Link: https://doi.org/10.3897/zookeys.1009.54303.suppl2