Laboratory culture of the California Sea Firefly Vargula tsujii (Ostracoda: Cypridinidae): Developing a model system for the evolution of marine bioluminescence

Bioluminescence, or the production of light by living organisms via chemical reaction, is widespread across Metazoa. Laboratory culture of bioluminescent organisms from diverse taxonomic groups is important for determining the biosynthetic pathways of bioluminescent substrates, which may lead to new tools for biotechnology and biomedicine. Some bioluminescent groups may be cultured, including some cnidarians, ctenophores, and brittle stars, but those use luminescent substrates (luciferins) obtained from their diets, and therefore are not informative for determination of the biosynthetic pathways of the luciferins. Other groups, including terrestrial fireflies, do synthesize their own luciferin, but culturing them is difficult and the biosynthetic pathway for firefly luciferin remains unclear. An additional independent origin of endogenous bioluminescence is found within ostracods from the family Cypridinidae, which use their luminescence for defense and, in Caribbean species, for courtship displays. Here, we report the first complete life cycle of a luminous ostracod (Vargula tsujii Kornicker & Baker, 1977, the California Sea Firefly) in the laboratory. We also describe the late-stage embryogenesis of Vargula tsujii and discuss the size classes of instar development. We find embryogenesis in V. tsujii ranges from 25–38 days, and this species appears to have five instar stages, consistent with ontogeny in other cypridinid lineages. We estimate a complete life cycle at 3–4 months. We also present the first complete mitochondrial genome for Vargula tsujii. Bringing a luminous ostracod into laboratory culture sets the stage for many potential avenues of study, including learning the biosynthetic pathway of cypridinid luciferin and genomic manipulation of an autogenic bioluminescent system.


Species
Phylum Luciferin www.nature.com/scientificreports www.nature.com/scientificreports/ and invertebrates on the benthos since they can be collected via baited traps. In addition, some authors observed V. tsujii attacking the soft tissues of live fishes caged nearshore 45 . Vargula tsujii is an important prey item of the plainfin midshipman fish (Porichthys notatus), as it is the source of luciferin for the bioluminescence seen in the fish 46 .
In addition to evidence for the completion of the life cycle of Vargula tsujii in the laboratory, we provide a successful methodology for culturing populations of V. tsujii, descriptions and analyses of late stage embryogenesis and instar development, a reconstruction of its mitochondrial genome, and a brief description of the distribution of molecular diversity of V. tsujii across coastal southern California.  Table S1). We also provide locations in Fig. 1 and Table S1 where we set traps but did not collect V. tsujii. We collected the specimens used for culture and description of the life cycle of V. tsujii exclusively from Fisherman's Cove, Catalina Island, and we collected additional specimens from the Cabrillo Marina and Shelter Island public boat ramp to discern the genetic diversity of V. tsujii and test for the presence of any cryptic species. For specimen collection we used Morin pipe traps: cylindrical polyvinyl chloride (PVC) tubes with mesh funnels to prevent the ostracods from escaping once they entered 47 . We baited traps with dead fish or fresh chicken liver (roughly 1-2 cm 3 ). We set traps in 1-10 meters of water at sunset and collected them 1.5 to 2 hours after last light to allow time for the ostracods to accumulate. population genetic analyses. We obtained Vargula tsujii sequence data for 16S rRNA (16S) and cytochrome oxidase 1 (COI) from a previously published thesis 48 . We then aligned and visually inspected sequence chromatograms with Sequencher version 3.1 (Gene Codes Corp) (  50 . To assess whether sequences within populations cluster together, we analyzed each gene via Discriminant Analysis of Principal Components (DAPC) using the adegenet package in R 51 and compared models with one, two, and three clusters using the Bayesian Information Criterion (BIC) 52 . Description of aquarium system and feeding. We cultured Vargula tsujii at the University of California, Santa Barbara (UCSB) in aquaria designed for water to flow through sand and out the bottom. This allowed continuous water flow to keep water fresh, and not suck ostracods into a filter. Instead, ostracods remain in sand where they spend the daylight hours. We obtained the seawater for this system through intake structures at UCSB located 2,500 feet offshore of campus beach at a depth of 51 feet under local conditions (12-18 °C sea surface www.nature.com/scientificreports www.nature.com/scientificreports/ temperature). The water in our tanks tends to fall between 16-18 °C because it warms in the building before being distributed to tanks. We kept animals under a 12:12 light cycle with an indirect blue LED light.

Methods
We first modified a five gallon glass tank by drilling a hole in the bottom, which was then covered with the bottom of an undergravel filter and a 3 cm layer of commercial aragonite sand with grain sizes 1-2 mm for aquariums (CaribSea brand). To maintain continuous water flow, we used plastic tubing (diameter = 1 cm) connected directly to an overflowing reservoir tank, itself fed by UCSB's seawater system that pumps directly from the Pacific Ocean. The tube from the reservoir created a siphon system for moving water into the main tank (Fig. 2). The siphon provided continuous water flow when needed, and prevented overflow of the aquarium. In addition, to prevent a complete drain of the water if the siphon flow broke, which is common with larger siphons, we attached an outflow pipe to the bottom of the aquarium that drained halfway up the side of the tank (Fig. 2).
We also created a similar system for smaller experimental aquaria, but these appeared to be less effective long-term, as we saw high mortality in non-systematic studies. For the small aquaria ("condos"), we cut a large hole (~5 cm diameter) into the bottom of 500 mL Nalgene Square PETG Media Bottles with Septum Closures (Figs. 2F, S7). We then modified a 100 μm cell strainer to fit in the bore of the media bottle in such a way that it would remain there when the closure of the media bottle was screwed back on, allowing water but not sand to flow through. We inverted the bottle and added a ~3 cm layer of sand above the cell strainer, measured from the edge of the bottle cap. In plain terms, the condos are simply capped bottles, turned upside down, after creating a large hole in the bottom and a small hole in the cap. We placed a strainer in the cap so water can flow down through the cap and not get clogged by sand. For these smaller aquaria, we used 5 mm diameter tubing as siphons from the reservoir tank and for the outflow. Similar to the large tank, the outflow tubes (placed in the septum closure) drained roughly halfway up the side of the aquaria to prevent the water from draining completely.
Most recently, we created custom aquariums from acrylic (Fig. 2G), with similar dimensions to the 5-gallon glass aquarium. We installed a bulkhead in the bottom and attached PVC tubing as a drain, which we again raised www.nature.com/scientificreports www.nature.com/scientificreports/ to about half the height of the aquarium to prevent complete drainage if inflow of water stops accidentally. We added a valve to control the rate of outflow and instead of a siphon, we added a float switch to set inflow rate. In the latest iteration of the aquaria, we raise the outflow pipe to set the level of water in the aquarium and use continuous inflow directly to the main aquarium, which works well as long as the water is draining freely. We use a flow rate of approximately 0.4 L/min.
We initially offered the ostracods multiple diet choices, including: (1) imitation crab, which is primarily pollock, (2) carnivore fish pellets or TetraMin fish flakes (Tetra, Blacksburg, VA), (3) tilapia fillets, and (4) anchovies, which have each made for successful food sources in Caribbean ostracods (Todd Oakley, personal observation 32,47 ). Although the adults appeared moderately interested in these food sources in non-systematic trials, we found the juvenile ostracods did not eat them, but they did actively feed on raw chicken liver, as do Vargula hilgendorfii individuals from Japan 53 . In culture, we therefore provided pieces of chicken liver daily (~1 cm 3 per 200 animals). Due to the nocturnal life history of V. tsujii, we fed them in the late afternoon or early evening (around the beginning of the dark cycle), and removed the food from the aquarium in the morning (~3-5 hours into the light cycle).
Rearing of embryos. We tested two primary methods for rearing embryos. The first was less successful in terms of longevity of both the mothers and embryos, but provided usable data on embryogenesis. Regardless of method, adult female ostracods did not appear to eat while they were brooding.
For the first method, we kept females individually in circular wells in twelve-well dishes 32 , with approximately 1 mL of 17 °C seawater. We recorded visual observations of the females and their embryos and changed the water daily. The advantage of this method is ease of identification and observation of individual ostracods, allowing us to track the growth of specific embryos over a given period of time. However, all females which deposited embryos into the brooding chamber died before their embryos were released from the marsupium, although the precise cause of death is unknown. These embryos were then removed from the female marsupium by dissection, kept in wells, and observed until the animals hatched through the chorion (see Supplemental Video).
For the second method, we identified and removed brooding females from the aquarium and placed them in 50 mL Falcon tubes (2 females per tube). Roughly 5 mL of commercial aragonite sand for saltwater aquariums with most grain sizes 1.0-2.0 mm (Caribsea brand) was added to the bottom of each tube to allow females to burrow. Falcon tubes were then placed in a large seawater reservoir that ranged from 17-19 °C. We changed the water in each tube daily, and imaged brooding females every other day with a camera attached to a dissecting scope. We used key diagnostic features such as number of embryos and female size to distinguish individual females from one another. Of the thirteen individuals kept in Falcon tubes, only one died prior to embryo release from the marsupium.

Animal measurements and inferences of instar size classes.
We imaged all ostracods using a standard dissecting microscope and an Olympus brand camera. We then measured carapace length, carapace height, eye width, and keel width with the software Fiji (Sample size = 626 individuals; Table S3) 54 . These parameters are useful in the determination of the instar stages in other bioluminescent ostracods 32 . A subset of these individuals (9 males, 11 females) were dissected to confirm sex. We partitioned the data into clusters using the KMeans clustering algorithm implemented in scikit-learn 55 in Python 2.7 56 . KMeans fits the data into k clusters, specified in the arguments, sorting each data point into the cluster which has the minimum euclidean distance to the centroid of the cluster. We originally specified the value for k as eight based on the five known stages of instar development for related species of ostracods plus the embryonic stage and adult males and females. However, a k-value of eight clusters resulted in a split of the inferred adult female cluster into two clusters rather than the discrimination of adult males from inferred A-I instars. To avoid this apparent over-clustering, we ran our final analysis with a k-value of seven, which retained all inferred adult females in a single cluster but was unable to distinguish an adult male cluster from the inferred A-I instars. We assigned development stages to each cluster based on the relationship by comparison to the closely related Photeros annecohenae, which grew in 12-well dishes to allow more precise documentation of instars 32 . We further split the A-I instar/Adult male cluster using the KMeans algorithm and a k-value of 3 with more precise measurements of length, height, and eye width, as well as eye to keel distance (EKD) from an additional sample of individuals that fell within this cluster (Sample size = 43). We measured eye to keel distance by calculating the length between the farthest point of the eye from the front of the ostracod to midway on the keel of the animal. We then assigned three new clusters to development stages by comparing eye size and shape parameters with what we expect based on Photeros annecohenae 32 .
Whole genome sequencing. We froze four adult female Vargula tsujii, collected from the WMSC Dock on 2017-09-01 in liquid nitrogen and shipped the specimens on dry-ice to the Hudson Alpha Institute for Biotechnology Genomic Services Lab (HAIB-GSL; Huntsville, AL) for genomic DNA extraction and Chromium v2 Illumina sequencing library preparation. HAIB-GSL extracted the DNA using the Qiagen MagAttract Kit, yielding 141 ng, 99 ng, 153 ng, and 96 ng of extracted DNA respectively. This quantity of DNA was not sufficient for a size distribution analysis via agarose pulse field gel analysis, therefore we decided to proceed with Chromium v2 library prep with the 153 ng sample (sample # 5047/TCO4), without confirmation of a suitable high-molecular weight DNA size distribution. We sequenced this library on three lanes of a HiSeqX instrument using a 151 × 151 paired-end sequencing mode, resulting in a total of 910,926,987 paired reads (260.6 Gbp). We attempted whole genome assembly of the resulting sequencing data with the Supernova v2.0.0 genome assembly software 57 , using default parameters. Pseudo-haplotype (pseudohap) paths through the assembly graph were then extracted using the Supernova 'mkoutput' command. The 'raw' assembly graph with full headers was also output to FASTA format using supernova 'mkoutput' , and was then converted to the GFA assembly graph format using the super-nova2gfa utility of gfaview (https://github.com/lh3/gfa1). After sequencing and library construction, remaining www.nature.com/scientificreports www.nature.com/scientificreports/ DNA was shipped on dry-ice from HAIB-GSL to MIT. We then used the high-sensitivity Agilent Femto Pulse High Molecular Weight DNA capillary electrophoresis instrument at the MIT BioMicro Center (Cambridge, MA) to perform an analysis of the size distribution of the remaining extracted DNA, including the sample used for Chromium library preparation.
Mitochondrial genome assembly. To achieve a full length mitochondrial genome (mtDNA) assembly of V. tsujii, we partitioned and assembled sequences separately from the nuclear genome. We first mapped Illumina reads from our Chromium libraries (Table S4) to the known mtDNA of the closest available relative, Vargula hilgendorfii (NC_005306.1) 58 using bowtie2 (v2.3.3.1) 59 (parameters:-very-sensitive-local). We then extracted read pairs with at least one read mapped from the resulting BAM file with samtools (parameters: view -G 12), name-sorted, extracted in FASTQ format (parameters: samtools fastq), and input the reads into SPAdes (v3.11.1) 60 for assembly (parameters:-only-assembler -k55,127). We inspected the resulting assembly graph using Bandage (v.0.8.1) 61 and performed a heuristic manual deletion of low coverage nodes (deletion of nodes with coverage <200x, mean coverage ~250x), which resulted in two loops with sequence similarity to the V. hilgendorfii mtDNA that both circularized through a single path (Supplementary Figure S1). We then manually inspected this graph via a blastn (2.7.1+) 62 aligned against the V. hilgendorfii mtDNA through SequenceServer (v1.0.11) 63 and visualized with the Integrated Genomics Viewer (v2.4.5) 64 . We noted that the single circularizing path was homologous to the two duplicated control regions reported in the V. hilgendorfii mtDNA 58 . We surmised that this assembly graph structure indicated the V. tsujii mtDNA had the same global structure as the V. hilgendorfii mtDNA, and so used the Bandage "specify exact path" tool to traverse the SPAdes assembly graph and generate a FASTA file representing the circularized V. tsujii mtDNA with the putative duplicated control regions ( Figure S1). To avoid splitting sequence features across the mtDNA circular sequence break, we circularly rotated this full assembly with the seqkit (v0.7.2) 58,65 "restart" command, to set the first nucleotide of the FASTA file at the start codon of ND2. We then confirmed 100% of the nucleotides via bowtie2 re-mapping of reads and polishing with Pilon (v1.21) 66 (parameters:-fix all). We annotated the final assembly of the mitochondrial genome using the MITOS2 web server 67 with the invertebrate mitochondrial genetic code (Number 5). We manually removed low confidence and duplicate gene predictions from the MITOS2 annotation.
Results population genetics analyses. Our COI haplotype network revealed minimal structure among V. tsujii populations in southern California (Fig. 3A). Haplotypes from San Pedro and Catalina Island overlap or diverge by only one to three mutations. Haplotypes from San Diego individuals, although still very similar to haplotypes from Catalina Island and San Pedro, differ by at least eight mutations and show no overlap with haplotypes from other populations of V. tsujii. However, the haplotype network for 16S indicates that sequences from San Diego individuals are identical to some individuals from the other two populations (Fig. 3B). Furthermore, although the model with three clusters was favored in the DAPC analysis for each gene (Supplementary Figures S2 and S3), membership in five of the six clusters includes individuals from different populations ( Supplementary Figures S4  and S5). Therefore our analyses do not support three genetically distinct populations as defined by location. embryogenesis description. After fertilization and release into the marsupium, V. tsujii embryos resemble opaque, green, ovoidal embryos (Figs. 4 and 5). After five to eight days of growth, separation of the yolk begins, which appears as a light green mass condensed within the embryo. Around ten to eighteen days after deposition into the brooding chamber, two very small red lateral eyespots are visible on each embryo and the separation of the yolk continues. Daily observations over the next six days showed continual growth and darkening of the eyes. By day sixteen the yolk is almost completely separated and has begun to form the gut of the ostracod. By days fifteen to nineteen, the eyespots are mostly black and the light organ (a light brown/yellow mass) appears anteroventral to the eyes of the embryo. Around twenty days after deposition, developing limbs in the form of two rows of small brown masses are visible along the ventral side of the embryos. At this point the gut has finished forming within the carapace and is visible as a yellow-white mass on the posterior end of each embryo. Limb formation continues over the next few days, and the eyes of each embryo become entirely black. Twenty-five to thirty-eight www.nature.com/scientificreports www.nature.com/scientificreports/ days following embryo deposition, embryogenesis is complete, and first (=A-V, or adult minus 5) instars are released from the mother's marsupium.

instar development. We inferred five juvenile instar stages (A-I, A-II, A-III, A-IV, A-V) and an adult stage
for Vargula tsujii (Figs. 6, 7A; Table 2). Laboratory measurements of specimens of unknown sex were compared with the measurements of sexed individuals (9 males, 11 females) to confirm interpretation of the instar stage of size classes (Fig. 7C,D). Due to large amounts of overlap of A-I females with adult and A-I males, these three groups are currently indistinguishable in the large data set. However, when we attempted to distinguish these  www.nature.com/scientificreports www.nature.com/scientificreports/ stages using additional measurements (particularly eye-to-keel distance, EKD), we find that EKD differentiates A-I males and females, and that length separates the A-I instars from adult males (Figs. 6E-G, 8; Table 3). Sexual dimorphism in size becomes apparent in adulthood, with adult females (Length ± SD = 1.983 ± 0.0798) clearly larger than adult males (Length ± SD = 1.647 ± 0.028). However, compared to many other cypridinid ostracods, this dimorphism can be difficult to assess visually without a side-by-side comparison. evidence for complete life cycle in the lab. To assess whether a complete life cycle was obtained in the laboratory, we began with only brooding females in a small experimental aquarium ("condo"). Those brooding females, although not tracked individually, released a number of juveniles, which we measured intermittently over time (Fig. 7B). In this aquarium we found juveniles, and eventually adults, that fell across the full size range  www.nature.com/scientificreports www.nature.com/scientificreports/ of wild-caught ostracods. After all of the broods initially hatched, we removed all adult females from the experimental aquarium to assess whether the ostracods would mature and mate in the laboratory. To ensure that all adult females were removed, we disassembled the small experimental aquarium and examined each animal under a dissecting microscope. We then reassembled and added only the subadult animals back into the aquarium. Following removal of the original adult females, we found adult brooding females that grew up in experimental aquarium that themselves gave birth to the next generation in 3-4 months time. Overall, our final Vargula tsujii nuclear genome assembly is highly fragmented, with an indication of high levels of heterozygosity between haplotypes (~5%) and a high proportion of simple sequence repeats.
The final V. tsujii mtDNA (15,729 bp) (Fig. 9) aligned to the previously published V. hilgendorfii mtDNA with 98% coverage and 78.5% nucleotide identity, and is available on NCBI GenBank (MG767172). The V. tsujii mtDNA contains 22 tRNA genes, 13 protein coding genes, and 2 duplicated control regions in an identical positioning and orientation to the V. hilgendorfii mtDNA. The V. tsujii duplicate control regions (661 bp, 570 bp), are smaller than the homologous V. hilgendorfii control regions (855 bp, 778 bp). The alignment of the V. tsujii control regions to the V. hilgendorfii control regions indicate they possess reasonable nucleotide identity (~73%), but the presence of unalignable regions ~20% of the length of the regions suggests evolution constrained by size and/or GC% content rather than sequence identity alone. There is one polymorphic base in the final mitogenome assembly coding for a synonymous wobble mutation in the COII gene (COII:L76).  Table 2. Measurements for each cluster (and inferred instar) of Vargula tsujii in mm ± SD.  (Fig. 7A). These points were then clustered via KMeans analysis into three clusters, which we assigned to A-I Males (red), A-I Females (blue), and Adult Males (yellow).  Table 3. Additional measurements for further inferring A-I Male, A-I Female, and Adult Male instars, including length, height, eye width, and eye to keel distance (EKD) ± SD.

Discussion
Based on measurement data on the first and second generations of ostracods reared in the lab (Fig. 7), and the presence of brooding female ostracods in our experimental aquarium where adult females had previously been removed, we can infer that we were successful in rearing Vargula tsujii through its complete life cycle in the lab. This is a first for any bioluminescent ostracod, as rearing in the lab has remained a significant hurdle for laboratory studies in these animals 32,33 . However, despite recent improvements, challenges still remain. Recent improvements (as described in the methods) include adding a neoprene gasket to the edge of the false bottom (Supplementary Figure S7) to prevent sand and detritus from falling below and fouling the aquarium and periodically siphoning the surface of the sand in the aquarium to reduce build-up of detritus and fouling microorganisms. But previous challenges mean our current population of animals is relatively small (between 500 and 1000 individuals), and we do not yet know if it is possible to continue the culture indefinitely without new input of animals. Previously, we added new wild-caught animals about every 6 months to increase the population size, but at the time of re-submission we have a culture that has been thriving for 10 months without the addition of new individuals. The primary challenge currently is populations within small-volume aquariums (condos) are very prone to collapse, and animals do not survive long in 12-well dishes as they did in P. annecohenae 32 , limiting current experimental work because tracking individuals is not possible.
Our current population was cultured from animals collected near Wrigley Marine Science Center on Santa Catalina Island, which appears to contain much of the genetic diversity sampled across much of the known range of V. tsujii (Fig. 3). Our population genetic analyses support the hypothesis that populations of Vargula tsujii experience some gene flow across the localities sampled, including Catalina Island. These results are consistent with the hypothesis that V. tsujii represents a single, connected species. As such, we assume our culture of ostracods from Santa Catalina Island, and inferences made from it, may apply to populations of V. tsujii across southern California coastal systems. However, we cannot discount the possibility that there might be some variation in development times and instar sizes across the range of this species. Also of note is that this species is not easily collected throughout the reported range, and northern populations, if they exist, may be divergent from southern California populations.
Attempted whole genome reconstruction. The 1C genome size of Vargula tsujii was previously reported to be 4.46 ± 0.08 pg (Female, n = 3), and 4.05 pg (Male, n = 1) 68 , corresponding to a 1C genome size of 4.36 Gbp and 3.96 Gbp respectively and suggesting XO sex determination. Our nuclear genome assembly is congruent with these past reports and indicates a genome size of greater than 3.2 Gbp for an adult female V.  www.nature.com/scientificreports www.nature.com/scientificreports/ tsujii. Our reconstruction of the Vargula tsujii nuclear genome assembly is highly fragmented, however, with an indication of high levels of heterozygosity between haplotypes and a high proportion of simple sequence repeats. We speculate that our poor nuclear genome assembly results are likely due to the very small size distribution of the input DNA, as larger fragment sizes are especially critical for the reconstruction of large genomes such as V. tsujii. Further protocol development is needed to extract suitable High Molecular Weight DNA, which could allow long-read sequencing that would improve assembly, but we note that assembly of very large genomes of high heterozygosity is non-trivial and is an active area of genome assembler research. Our results may also be due to the poor handling of high-heterozygosity species by the 10x Genomics Chromium platform (anecdotally), which notably is now discontinued by 10x Genomics as a library preparation and sequencing product. Alternative approaches to improve our V. tsujii nuclear genome assembly are currently in progress.
The life cycle of Vargula tsujii. The embryogenesis of Vargula tsujii is similar to closely related species, and the 24-day average brooding duration falls within the range for other cypridinid ostracods (10-30 days) 69 . The primary stages of V. tsujii late-stage embryogenesis are yolk separation, formation of eye spots, development of the upper lip, development of limbs, and emergence from the chorion and marsupium (Fig. 5). In general, the late-stage embryonic development of Vargula tsujii is longer than that of other luminescent ostracods whose development has been described ( Table 4). Longer overall development in the temperate V. tsujii is consistent with work in other crustaceans, which indicates that temperature has an inverse relationship with development time 70 . However, earlier embryonic development in V. tsujii is faster than in the tropical P. annecohenae. In P. anncohenae, yolk separation begins on day 9 and eye development begins on days 14-15 32 , while those stages begin at 5-7 and 10-13 days, respectively, in V. tsujii. Additionally, the order of development in V. tsujii appears to differ from that of Vargula hilgendorfii, which starts to develop limbs on days 4-5, before eye and upper lip development 71 . By contrast, V. tsujii limbs do not appear until after eye-spot and upper lip formation. The reasons for this difference are unclear.
Juvenile development of Vargula tsujii is also quite similar overall to that of Photeros annecohenae, the only other bioluminescent cypridinid ostracod for which instar developmental data are available 32 . We inferred five distinct instar stages (A-V through A-I), and A-I instars develop into size dimorphic adult males and females (Tables 2 and 3). It is important to note that these inferences represent hypotheses that can be tested by rearing and measuring individuals through molts, which we have not yet been able to accomplish in our culture. However, average measurements of size, including carapace length and height, as well as eye width, were comparable between respective instar stages in V. tsujii (inferred) and P. annecohenae (observed). However, in V. tsujii average eye width remains within the range of 0.0600 mm to 0.0684 mm from the embryonic stage until the second instar, and makes a large jump by about two fold in the third instar (0.121 ± 0.0208 mm), as compared to a gradual increase in eye size in P. annecohenae. In our KMeans clustering analysis, the second-to-largest cluster of V. tsujii appears to contain adult males as well as A-I males and females. This finding differs from that in P. annecohenae, where adult males were found to cluster separately from A-I males and females and adult females based on size and shape, and A-I males and females could be distinguished by certain shape parameters, as well as eye size 32 . In V. tsujii, we can largely differentiate adult and A-I males from females by eye-to-keel distance (EKD), which was not measured in previous studies 32 , and adult males can be separated from A-I males by length ( Fig. 8; Table 3). We hypothesize that this apparent reduction of sexual dimorphism observed in the later stages of development, compared to P. annecohenae, may occur because V. tsujii may have lost the ability to use luminescence for courtship 26 , so selection acting on differences in male and female shape may have been reduced.
Benefits of completing the life cycle. Cypridinid ostracods represent an independent origin of endogenous autogenic bioluminescence. This group is a natural comparison to other, well-studied bioluminescent courtship systems, such as that of fireflies 72 . Other possible laboratory models for bioluminescence, such as brittle stars 73 and cnidarians 74 are useful for some research tasks, such as studies of luciferins obtained from the diet (e.g., coelenterazine). Others might be easier to rear in the laboratory, such as the self-fertilizing brittle star Amphiolus squamata 75 . However, ostracods produce their own luciferase and luciferin, unlike brittle stars and cnidarians. Ostracods also have relatively fast generation times (on the order of months), which can be modified somewhat by adjusting the temperature of the water, as is common among ostracods [31][32][33]76,77 . In comparison, fireflies are usually on a six-month to one-year life cycle that is difficult to alter [78][79][80] , and some brittle stars appear to spawn only seasonally 81,82 . Furthermore, culture of an organism (Vargula tsujii) from a clade with an independent origin of autogenic bioluminescence allows for comparative studies between luminescence in ostracods and the autogenic bioluminescence found in fireflies 17 or other organisms, which would not otherwise be possible.  Table 4. Comparison of embryogenesis timing between five different ostracod species post-deposition. Species with an asterisk (*) are bioluminescent cypridinids. A dash (−) indicates no data are available.
www.nature.com/scientificreports www.nature.com/scientificreports/ Ostracods will be incredibly useful for studying the physiology and biochemistry of bioluminescence 2,21 , as both diversification and behavioral changes in cypridinids related to luminescence have occurred using the same luciferin and homologous luciferases 30 . Furthermore, although Vargula tsujii itself does not possess bioluminescent signaling for courtship, we expect a similar culturing system could be useful for some signaling species as well, at least those that come to baited traps. Given that ostracod courtship signals may be simulated in a laboratory setting, this system could allow us to assess male and female behavior in response to changes in signal parameters, such as intensity, color, and pulse duration 83 . These tools will provide researchers with an excellent system for asking important questions related to selection on genes important for rapid diversification 84 , including those related to behavior 85 , sexual selection 86 , and the role of the biochemical properties of bioluminescence 29 .
Importantly, the biosynthetic pathway for luciferin production is currently unknown for any animal. One major obstacle to describing this pathway has been building laboratory cultures for animal species that produce luciferin themselves and also have a relatively fast generation time. Solving this bottleneck to accessing specimens for bioluminescent ostracods has set the stage for investigations into the biosynthetic pathway of cypridinid luciferin and its possible use in biomedical tools such as optogenetics 87 . Finally, this is an excellent system with which to consider developing genome editing techniques (e.g., CRISPR/Cas9) 88 to test candidate genes for luciferin biosynthesis and to assess how genetic variants in critical genes might affect production or perception of bioluminescent signals.

Data availability
Sequence data for COI and 16S are available in GenBank (Table S2) and our whole genome Chromium reads are available in the Sequence Read Archive (SRR9308458). The final V. tsujii mtDNA is available on GenBank (MG767172). Femto pulse results, the pseudohaplotype files, and converted assembly graph for our Supernova nuclear genome assembly are available in the Dryad repository (doi:10.6075/J0DZ06QQ). Scripts and data for the population genetics and k-means clustering analyses are provided on Github (https://github.com/goodgodric28/ vargula_tsujii_culture). An additional video which shows Vargula tsujii hatching from the chorion and outside the mother is provided on Youtube (https://youtu.be/8o5CqHTcTjI). Those interested in starting their own V. tsujii culture from our current laboratory culture should contact the corresponding author for details.