Presence of a plastid-derived structure

We compared the microbial communities of A. lixula eggs, embryos, and larvae (that were never fed) using amplicon sequencing of the 16S rRNA gene. These communities were primarily composed of the bacterial phyla Bacteroidota (7.3%), Cyanobacteria (3.4%), Firmicutes (5.8%), and Proteobacteria (80.5%) (Fig 1A). The majority (78.4% ± 5.0%) of cyanobacterial sequences associated with the egg were, however, unidentified plastid sequences that could be reassigned to photosynthetic eukaryotes (Fig 1A and 1B). These plastid sequences were derived from the Archaeplastida (13.3%), Excavata (14.9%), Hacrobia (45.3%), and Stramenopiles (26.5%) (Figs 1B and S2; S1 Table). We re-analyzed the egg-associated microbiota for a dozen other sea urchin species [22] and found an identical pattern: cyanobacteria (which do not contain plastids) and photosynthetic eukaryotes were both present, but photosynthetic eukaryotes represented nearly all of these sequences (S3 Fig; S2 and S3 Tables).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Plastid DNA in sea urchin eggs.

(A) One of the main microbial phyla found to associate with the offspring of the sea urchin Arbacia lixula were Cyanobacteria (which do not contain plastids), but the majority of these sequences were unidentified plastids of photosynthetic eukaryotes. (B) All unidentified plastids could be traced to four major groups and less than a dozen ASVs of photosynthetic eukaryotes. (C) A substantial volume of A. lixula eggs would need to be allocated towards photosynthetic eukaryotes if their entire cell was maternally transmitted. A nuclear gene marker (18S rRNA gene) was disproportionately low in abundance in the metagenome of A. lixula embryos and larvae compared to a plastid gene marker (16S rRNA gene), implying that mothers may only provide her offspring with plastids. Corresponding raw data are presented in S1 and S8 Tables.

https://doi.org/10.1371/journal.pbio.3003705.g001

Plastid sequences were exclusive to and consistently present in the eggs of sea urchins that develop by planktotrohpy (i.e., were not detected in non-feeding lecithotrophs; S4 Fig). These plastid taxa [i.e., amplicon sequence variants (ASVs)] originated from eight eukaryotic phyla that fall within five kingdoms, with Ochrophyta being the most abundant and the Bacillariophyta (i.e., diatoms) representing ~90.0% of these sequences (Figs 1D and S5; S4 and S5 Tables). Approximately 91.6% of Ochrophyta ASVs—and ~74.6% of all ASVs—were derived from diatoms (Figs 1D and S5; S4 and S5 Tables). Sea urchin eggs have a species-specific collection of plastid ASVs [Analysis of Similarity (ANOSIM), p < 0.001], which can more clearly be defined by geography (ANOSIM, p < 0.001) and time (ANOSIM, p < 0.001) (S6 and S7 Figs; S6 and S7 Tables). Richness of these plastid ASVs directly relates to egg size (linear regression: F_1,109 = 7.80, p = 0.006, R² = 0.067), whereby sea urchin species with smaller eggs tend to specialize on a few dominant ASVs while those with larger eggs had a relatively even distribution of a dozen or so ASVs (linear regression: F_1,109 = 68.28, p < 0.0001, R² = 0.385; Figs 1E, S8, and S9).

Associating with diverse photosynthetic eukaryotes with cell sizes similar to sea urchin eggs would present a geometric problem for the provisioning of macronutrients that are essential for early development (S10 Fig) [13,25]. An alternative strategy would be for mothers to provide her eggs with only the plastids from these photosynthetic eukaryotes. We used shotgun metagenomics to quantitatively compare a plastid (16S rRNA) and nuclear (18S rRNA) gene marker for photosynthetic eukaryotes. We found that each A. lixula embryo had 25.2 (± 20.7) copies of the 16S rRNA gene and 0.05 (± 0.07) copies of the 18S rRNA gene from photosynthetic eukaryotes (Figs 1C and S11; S8 Table). Notably, zero copies of the 18S rRNA gene were detected for most eukaryotic phyla, while copies of the 16S rRNA gene were consistently present (S11 Fig and S8 Table). Abundance of the 16S rRNA gene decreased significantly in larvae (7.3 × ; unpaired t test, p = 0.006), while the abundance of the 18S rRNA gene remained consistent (unpaired t test, p = 0.207) (Figs 1C and S4; S3 Table).

We fixed A. lixula eggs for fluorescence and transmission electron microscopy to determine whether plastids are present in A. lixula eggs [26]. We observed cytoplasmic structures with autofluorescence in most eggs, which lacked chlorophyll but were most consistent for the presence of carotenoids (Figs 2A, 2B, and S12). The ultrastructure of these particles were elongated crystals (Figs 2C and S13). The presence of plastid DNA and autofluorescent particles implies that these structures are chromoplast-derived carotenoid crystals (Figs 1 and 2). We could not identify chromoplast-specific starch granules and plastoglobules from similar structures that are commonly found in sea urchin eggs. Chromoplast-specific carotenoids (e.g., xanthophylls) were, however, detected in the metabolome of A. lixula eggs (S9 and S10 Tables). The presence of multiple hallmark structures exclusive to chromoplasts [27,28] implies that a plastid-derived structure is present in these eggs.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Presence of chromoplast-derived carotenoid crystal.

(A, B) Micrographs of unfertilized eggs from the sea urchin Arbacia lixula without (top) and with (bottom) cytoplasmic autofluorescent particles. These particles were most consistent for the presence of carotenoids, as shown using (A) transmitted light and (B) fluorescence at an excitation of 561 nm. The light micrograph is a single optical slice, while the florescence micrograph is a maximum projection Z-stack. (C) These autofluorescent structures correspond with large crystals (CC) observed in transmission electron micrographs. Mitochondria (Mt) and yolk platelets (YP) are also noted. The combined fluorescence signature and cellular structure implies that these crystal structures are chromoplast-derived carotenoid crystals. Corresponding raw data are presented in S11 Table.

https://doi.org/10.1371/journal.pbio.3003705.g002

Benefits to offspring development

Chromoplast-derived carotenoid crystals influence plant reproductive fitness and are involved in light-dependent reactions that are independent of photosynthesis [29,30]. Sea urchin development, on the other hand, is not known to be directly impacted by light. If these chromoplast-derived carotenoid crystal structures are beneficial to A. lixula offspring, then a “loss of function” (i.e., darkness) should negatively impact development. We observed that A. lixula offspring undergo initial cell divisions more quickly in dark (i.e., at 4 hpf; paired t test, p = 0.025), but have no observable difference upon gastrulation (paired t test, p = 1.000; S14 and S15 Figs; S11 Table). By four days post-fertilization, 4.8× more offspring developed into 2-arm larvae in light than their siblings in dark (paired t test, p = 0.026; Figs 3A and S14; S11 Table). The feeding arms of larvae that were cultured in dark grew disproportionately relative to the larval body (unpaired t test, p = 0.012; Figs 3B, S16, and S17; S11 Table); thus, larvae in dark expressed the long-arm phenotype that corresponds with a nutritional restriction [31]. However, unlike typical morphological plasticity, the gut volume did not differ between A. lixula larvae that were cultured in light and dark (unpaired t test, p = 0.952; Fig 3B and S11 Table) and that this occurred without any response to exogenous resources (i.e., phytoplankton) [31].

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Benefits to offspring fitness.

(A) More offspring of the sea urchin Arbacia lixula developed into 2-arm larvae in light (gray) than their siblings in dark (black) by 4 days post-fertilization. (B) Feeding arms of larvae that were cultured in dark grew disproportionately to the larval body (left), but did not exhibit a change in gut volume (right); thus, larvae in dark expressed the long-arm phenotype that corresponds with a nutritional limitation. (C) Survival was higher for offspring cultured in light, of which was also affected by time for both treatments as well as the interaction between these factors. Differences in light-induced survival was observed on days two through five. (D) Larvae cultured in light and dark exhibit organism-wide differences in their metabolome (left). This was predominantly driven by a shift in fatty acid metabolism, from the catabolism of energy storage lipids in light to structural lipids in dark (Variable Importance in Projection, VIPs; center). Two of the 717 differentially abundant metabolites are apocarotenoids, which were more abundant for larvae cultured in light (right).

https://doi.org/10.1371/journal.pbio.3003705.g003

A reduced developmental rate and modification to nutrition are two key factors that directly impact developmental fitness [8], which can be assessed by quantifying survival. We observed that the survival of A. lixula offspring was significantly reduced when cultured in dark, as well as with time and the interaction between these factors (two-way ANOVA, time: F_5,54 = 143.6, p < 0.0001; treatment: F_1,54 = 109.9, p = 0.0007, interaction: F_5,54 = 10.31, p < 0.0001; Fig 3C and S11 Table). Differences in survivorship were observed two days post-hatching (i.e., four days post-fertilization) and in all subsequent days (p < 0.001 for each day), such that survivorship was 1.4× higher in light at five days post-hatching (Fig 3C and S11 Table).

We then sought to understand how these chromoplast-derived carotenoid crystal structures impact the developmental metabolism of A. lixula. Light induced an organism-wide divergence in the metabolome during early development (PERMANOVA, eggs: p = 0.181, p < 0.001; Figs 3D and S18; S12 Table). This was predominantly driven by a shift in fatty acid metabolism, from the catabolism of energy storage lipids (e.g., fatty acids, acyls, and esters) in light to structural lipids (e.g., phospholipids and sphingolipids) in dark (Figs 3D and S19; S9, S10, and S13 Tables). Two of the 717 differentially abundant metabolites were apocarotenoids (i.e., xanthophyll-derived phytohormones that promote growth and development, regulate lipid metabolism, and modulate environmental stress [32–34]). These apocarotenoids are a prenol lipid and an uncharacterized apocarotenoid, which were detected 2.7× more frequently in larvae in light and formed a molecular family with additional phytohormones and an energetic lipid (Figs 3D and S19; S9, S10, and S13 Tables). We, therefore, find that these chromoplast-derived carotenoid crystal structures and their derived metabolites have a broad and integral influence on the development of A. lixula.

Enhanced and trans-oceanic dispersal

Development and survival can directly impact dispersal [1,2,9,10]. We used our survival data to estimate the potential pelagic larval duration for A. lixula that do (i.e., light) and do not (i.e., dark) benefit from the light-dependent activity of chromoplast components (Fig 3C). We estimated that the instantaneous mortality rate for A. lixula is 0.093 day⁻¹ in light and 0.165 day⁻¹ in dark, which are within the expected range for the marine invertebrates that develop by planktotrohpy and are higher than expected because the larvae in these experiments were never fed [8,9,35]. By quantifying the fecundity of A. lixula (S10 Fig), we then estimate that offspring that benefit from the light-dependent activity of chromoplast components have a 1.8× longer dispersal duration than those without it (S20 Fig). We then used the VIKING20X circulation model of the Atlantic Ocean to simulate dispersal [36]. We estimate that offspring that benefit from the light-dependent activity of chromoplast components have the potential to disperse 1.6× further (paired t test, p < 0.0001) and across a 4.7× larger geographical area (paired t test, p < 0.0001) than those without that benefit (Figs 4A and S21). A longer planktonic larval duration is estimated to enable 2.9× more larvae additional dispersal paths for a successful settlement (paired t test, p < 0.001; S21 and S24 Figs; S15 Table).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Enhanced and trans-oceanic dispersal.

(A) Distribution of particles on the African shelf after 102 (i.e., without benefits from the light-dependent activity of chromoplast components; left) and 181 (i.e., with benefits from the light-dependent activity of chromoplast components; right) days of being released from Tenerife (black bordered island). Grey arrows show the 10-year mean velocity in the release depth from VIKING20X, with every 10th arrow is shown. Black contours around the African shelf mark the 500 m isobath. (B) Only particles that dispersed for 181 days from the southernmost point of Arbacia lixula’s range were predicted to disperse across the Northern Atlantic Ocean from the coast of Africa to Brazil. Fifty example trajectories are shown. Each plot is the accumulation of annual releases over 10 years. Grey arrows as in A, but only every 50th arrow is shown. Visualizations of larval dispersal used maps from cartopy [110], with the underlying vector map using NaturalEarth (naturalearthdata.com).

https://doi.org/10.1371/journal.pbio.3003705.g004

Offspring that benefit from the light-dependent activity of chromoplast components are also predicted to take fewer generations to spread across the longitudinal range of A. lixula (S25 Fig). At the southernmost point of this range, only larvae that benefit from the light-dependent activity of chromoplast components—an estimated 0.3% of the total—are predicted to disperse across the Atlantic Ocean (S26–S29 Figs). Nearly all of these larvae enter the Angola Current, merge with the South Equatorial Current, and then settle on the northern coast of Brazil (Figs 4B, S26, and S29). This east-to-west trans-Atlantic journey is successful each year despite the inter-annual variability in the strength of these currents (S28 Fig). Once on the northern coast of Brazil, offspring that benefit from the light-dependent activity of chromoplast components could disperse southward along the continental shelf, back to the African coast, and to Cape Verde (S30 Fig). It, however, would take 2.0× longer to reach the Azores and 4.3× longer to reach the Canary Islands than our estimated pelagic larval duration (S31 Fig and S15 Table).

Specimen collection

Adult Arbacia lixula were collected by snorkel from Radazul Beach (Tenerife, Canary Islands, Spain; 28.401487, −16.324368) from 2021 to 2024. Adult sea urchins were transferred dry to the University of La Laguna within one hour, where they were maintained in an aerated aquarium with seawater from Radazul Beach until spawning a couple hours later.

Offspring culturing

Gravid adults were spawned by an intracoelomic injection of 0.50 M KCl. Females were rinsed with 0.22 μm FSW to remove potential contamination by epibionts on the aboral surface—despite it being widely acknowledged that epibionts are unable to colonize sea urchin spines [53,54]—and then spawned into 5.0 μm filtered seawater (FSW) from Radazul Beach, while males were spawned dry and stored on ice. Eggs were subsequently rinsed three times with 0.22 μm FSW to remove potential contamination during spawning. Female and male gametes were collected to generate full-sibling replicates (n = 5) by following Strathmann [55].

Briefly, a few drops of diluted (1:1,000) sperm from the male pair was added to the total clutch of the female pair. Egg and sperm were gently mixed. Fertilization was assessed 30 min later using a compound microscope (Leica DM1000 with a Canon EOS 1200D camera). Cultures were diluted in 5.0 μm FSW (that allows developing embryos to be colonized by their natural microbiota) to 2 embryos/mL and each male-female pair was divided amongst separate light and dark beakers. Dark beakers were covered with multiple layers of tinfoil, equipped with a cardboard lid that was also covered with multiple layers of tinfoil and an overhang where the lid and beaker meet, and a loose-fit black garbage bag on the stirring paddle to exclude all possible light from entering. All beakers were stirred gently and ~95% of the FSW was replenished every other day with 5.0 μm FSW. Offspring in the dark beakers were minimally exposed to light during mandatory husbandry (i.e.,~5 min per water change). Eggs, embryos, and larvae were cultured in ambient FSW at 23 °C (i.e., ambient sea surface temperature during autumn in Canary Islands coastal waters [56]), provided a combination of ambient solar light and artificial laboratory light (photosynthetically active radiation of ~570.3 ± 100.0 μmol photons m⁻² s⁻¹ to reflect light incidence with alternate cloudy and sunny days in Canary Islands [57]), and never provided phytoplankton or another external dietary source.

Sample collection, extraction, and sequencing

We collected eggs (0 hours post-fertilization, hpf), embryos (blastula; 18 hpf), and early-stage larvae (4 days post-fertilization, dpf) from each “light” beaker (n = 5) for amplicon and metagenomic sequencing; both sequencing types were from the same sample. Approximately 200 offspring were collected from each male-female pair at each of the developmental stages. Offspring were concentrated using a microcentrifuge, the seawater was removed with a sterile pipette, and tissues were preserved in 250 µL of 70% ethanol for long-term storage at −20 °C. Samples were then transferred on blue ice to the GEOMAR Helmholtz Centre for Ocean Research (Kiel, Germany).

Total DNA was extracted from all samples and DNA kit blanks (n = 3) according to the manufacture’s protocol for the DNeasy Blood & Tissue Mini Kit (Qiagen). Total DNA was then quantified using the dsDNA HS Assay Kits for the Qubit Fluorometer (Thermo Fisher Scientific) following the manufacture’s protocol. Samples were diluted to 0.5 ng/µL for amplicon sequencing and were not diluted for metagenomic sequencing. Samples for amplicon (V3/V4 region of the 16S rRNA gene) and metagenomics were sent to the Competence Center for Genomic Analysis (Christian-Albrechts University of Kiel, Germany) for library preparation and sequencing by Illumina MiSeq (v3, 2 × 300 bp paired-end reads) and Illumina NovaSeq (S4, 2 × 150 bp paired-end reads), respectively.

Amplicon analysis of A. lixula

Raw amplicon reads with quality information were imported into QIIME 2 (v. 2022.11; [58]), where forward and reverse reads were paired using VSEARCH [59], filtered by quality score, and denoised using Deblur [60]. QIIME 2-generated “features” were analyzed as ASVs [61] and were assigned taxonomy using SILVA (v. 138; [62]). Sequences matching to Archaea, those that were disproportionately present in the DNA kit blanks, and samples with less than 3,000 reads were discarded. The filtered data table was then rarified to 3,482 sequences (S32 Fig and S16 Table). We summarized taxonomic groups of bacteria for each sample from the SILVA [62] assignment. We then filtered this data table and used PhytoREF [63]—a database for plastidial 16S rRNA gene sequences of photosynthetic eukaryotes—to re-assign taxonomy for all ASVs that were previously assigned to “o__Chloroplast.” It is worth noting that PhytoREF uses outdated taxonomic information for Eukaryotes [64,65], which may limit taxonomic accuracy at the finer levels. Taxonomic overlap in the plastid ASVs that were present in the eggs of each female A. lixula was determined using InteractiVenn [66].

Amplicon meta-analysis of sea urchins

We compiled all published 16S rRNA datasets for sea urchin eggs (S2 Table) [15,22,67,68], as well as single samples of Eucidaris thouarsii and Eu. tribuloides that was collected during, but not published with, Carrier and colleagues [15]. We generated taxonomic profiles for each sea urchin species and then tallied the relative abundance of “o__Chloroplast” to the non-plastid Cyanobacteria sequences, and then reassigned the taxonomy for “o__Chloroplast” using PhytoREF [63]. We transformed the plastid-only data table to relative frequency to avoid discarding invaluable sequences. Using this, we generated a phylogenetic tree for all plastid ASVs within QIIME 2 [58] using MAFFT [69] and FastTree [70]. This tree was then visualized using the Interactive Tree Of Life [71] and stylized using Adobe Illustrator (v. 24.0.1). Mean relative frequency of each plastid ASV was added as metadata to the phylogenetic tree. Next, we manually curated the prevalence of plastids in each sample and ASVs across species. We then calculated Jaccard distances for all samples, visualized these using principal coordinate analyses in Prism (v. 9.0.0), and stylized these using Adobe Illustrator. ANOSIM and PERMDISP, as well as their respective pairwise comparisons, were performed within QIIME 2 [58] to test whether community composition and dispersion varied between host species, geography, and years.

Jaccard distances were then used to construct a dendrogram within QIIME 2 [58]. The data table for this analysis was collapsed by host species and then was rarified to 447 sequences; this was the only rarified analysis. We then tested for phylosymbiosis by comparing topological congruence with a Cytochrome C Oxidase Subunit I (COI) gene tree using sequences from the National Center for Biotechnology Information (NCBI) (S2 Table). These sequences were imported into the Molecular Evolutionary Genetics Analysis software (v. 11.0.9; [72]), aligned with MUSCLE [73], and trimmed. This relationship was inferred using maximum likelihood with the optimized DNA substitution model (GTR + G + I, as determined by BIC criteria) and 1,000 bootstrap replicates. Phylosymbiosis was tested using the Robinson-Foulds metric in TreeCmp (v. 2.0; [74]) and matching cluster metrics with 10,000 random trees [75].

Four measures of alpha diversity (total ASVs, Faith’s phylogenetic distance, McIntosh evenness, and McIntosh dominance) were then calculated for all samples. These were compared across species using one-way ANOVAs and Tukey’s post-hoc tests for pairwise comparisons. Linear regression analysis was performed to compare total plastid ASVs for each sea urchin species to published average values of egg diameter [13,76–78], total energetic content [77,79–81], and pelagic larval duration [76,82–87] for these same sea urchin species in Prism. A linear regression analysis was also used to compare total plastid ASVs and McIntosh dominance in Prism.

Metagenomic analysis

Raw metagenomic reads were quality filtered, trimmed, and removed of host sequences within MetaWRAP [88]. The S. purpuratus genome (v. 5.0) from Echinobase [89,90] was used to minimize host reads because there was no published genome for A. lixula at the time of this analysis. Furthermore, we removed mitochondrial sequences using published mitochondrial genomes for A. lixula (NC_001770), Echinometra mathaei (NC_034767), Hemicentrotus pulcherrimus (NC_023771), Heterocentrotus mammillatus (NC_034768), and Lytechinus variegatus (NC_037785) from NCBI.

rRNA gene sequences were identified in the forward sequences using METAXA2 [91]. All rRNA gene sequences were imported into QIIME 2 [58] for taxonomic assignment using the default parameters of the VSEARCH [59] consensus classifier. Plastid sequences (i.e., 16S rRNA gene) were identified using PhytoREF, while nuclear sequences (i.e., 18S rRNA gene) were identified using the combined Protist Ribosomal Reference (PR²) [92]/EukRef [93] 18S rRNA databases. The quantity of 16S and 18S rRNA gene sequences were counted for all equivalent groups of photosynthetic eukaryotes (S8 Table). Sequences for 16S and 18S rRNA gene were then summed separately and compared between developmental stages using unpaired two-tailed t-tests within Prism.

Microscopy

Fluorescence microscopy was used to assess whether there were autofluorescent signatures consistent with plastid pigments in unfertilized eggs of A. lixula. Eggs (n = 10) were fixed at room temperature in 4% formaldehyde buffered in ambient seawater supplemented with 20 mM HEPES. This fixative was replaced after 1 hour with 1% formaldehyde buffered in ambient seawater with 20 mM HEPES buffer, before being stored long-term at 4 °C. Eggs were assessed for autofluorescence above 620 nm at excitation wavelengths of 405 nm, 488 nm, 561 nm, and 640 nm using an LSM 900 with Airyscan microscope (Zeiss) in the Multiplex mode. Maximum intensity projections of Z-stacks were generated using the ZEN Blue (v. 3.2) software (Zeiss). Transmitted light images at single Z-planes were acquired with the LSM mode. A Plan-Apochromat 20× (0.8 NA) objective was used for all applications.

Transmission electron microscopy was used to assess the ultrastructure of unfertilized eggs of A. lixula from the same females (n = 10) that were sampled for fluorescence microscopy. Eggs were fixed with 0.5% glutaraldehyde buffered in ambient seawater for 1 hour and then stored long-term at 4 °C [26]. For resin embedding, the eggs were washed with deionized water and glycine to remove the fixative, post-fixed with 1% osmium tetroxide in 1.5% potassium ferricyanide for 1 hour at room temperature, washed several times with deionized water, treated with 0.1% tannic acid in 50 mM HEPES (pH 7) for 3 min at room temperature, stained with 2% aqueous uranyl acetate for 1 hour, and then washed again with deionized water. Samples were dehydrated using a graded ethanol series, followed by 100% acetone, and then progressively infiltrated with Epon 812 substitute resin (Sigma-Aldrich) and heat polymerized. Ultrathin 80-nm sections, prepared using the UC7 ultramicrotome (Leica) and diamond knives (Diatome), were contrasted with saturated aqueous uranyl acetate and lead citrate. Grids were inspected in a Tecnai G2 Spirit BioTWIN transmission electron microscope using a side-entry CCD MegaView III G2 camera and iTEM software (v. 5.0) (Olympus Soft Imaging Solutions).

Development, morphology, and survival

We measured egg size, development, and morphology as part of the experiment outlined in Sample collection, extraction, and sequencing section. Egg diameter (n = 10 per replicate) was measured at 0 hpf for all “light” beakers, while the developmental stage (n = 25 per replicate per time point) at 0 hpf, 4 hpf, 18 hpf, 2 dpf, and 4 dpf was assessed for all beakers. Offspring were staged as either egg, 2-cell embryo, 4-cell embryo, 8-cell embryo, 16-cell embryo, blastula, gastrula, prism-stage larva, or 2-arm larva. We quantified average cell number per embryo at 4 hpf and frequency of 2-arm larva at 4 dpf to assess developmental rate, which were both compared statistically by a t test. Morphometrics (i.e., mid-body line, post-oral arms, stomach height, and stomach width) were performed on all 2-arm larva at 4 dpf using ImageJ (v. 1.9.2; [94]). Due to reduced development and survival in dark, 2-arm larvae from light and dark replicates (i.e., n = 49 in light and n = 10 in dark) were merged and then compared statistically within Prism using an unpaired two-tailed t test for each morphological measurement.

Survivorship was quantified in separate experiments using the exact same experimental design. Offspring from male-female pairs (n = 10) developed until hatching (i.e., 2 dpf), after which triplicate counts of each culture density was performed and were subsequently diluted to 2 embryos/mL in 1 L of FSW. Cultures were reversed-filtered to 100 mL each day, the number of embryos were counted in triplicate 1 mL aliquots using a dissecting microscope (Leica EZ4), and then returned to the beaker along with fresh FSW. This was performed for five days (i.e., until 7 dpf). Five of the biological replicates were performed in 0.22 μm FSW and the other five were performed in 5.0 μm FSW as a cross-validation for the agent being intracellular. Both survival datasets were merged because the agent was intracellular (Fig 2) and because the survival rates were statistically identical between both conditions (two-way ANOVA for light, time: F_5,40 = 102.8, p < 0.0001; treatment: F_{1 8} = 0.04, = , interaction: F_5,40 = 3.91, p < 0.0055; two-way ANOVA for dark, time: F_5,40 = 93.33, p < 0.0001; treatment: F_{1 8} = 0.38, p = 0.554, interaction: F_5,40 = 1.06, p = 0.399; S11 Table). A two-way ANOVA was performed within Prism to test whether survivorship differed between treatments and over time/development.

Metabolomics

Metabolomics was used in a separate experiment using the exact same experimental design as development and survival. Specifically, the clutch of spawned females (n = 5) was equally distributed between separate 1 L beakers for light and dark. Three technical replicates of ~1,000 eggs were collected from all beakers, which were immediately centrifuged to concentrate the tissues, the 0.22 μm FSW was removed, samples were flash frozen using liquid nitrogen, and then transferred to a −80 °C freezer for long-term storage. The remaining eggs were fertilized with dilute sperm to create full-sibling replicates. Excess sperm was removed 1 hour later by multiple 95% water changes and fertilized embryo were then provided two days to develop. Hatched blastulae were isolated, the cultures were diluted to 10 embryos/mL following multiple 95% water changes, and the cultures were then provided 3 additional days to develop (i.e., when there was a maximal differential in survival; Fig 3C). At 5 dpf, three technical replicates of ~1,000 offspring were collected from all beakers and processed as described above. Four and three biological replicates could be collected from the light and dark treatments of larvae, respectively. Blank tubes and 0.22 μm FSW seawater were also collected as controls (n = 5 each). All samples were then shipped on dry ice (with temperature being monitored) to the University of California, San Diego (CA, USA).

Samples were prepared for liquid chromatography tandem mass spectrometry (LC–MS/MS) according to an established protocol [95]. Untargeted LC–MS/MS was then performed using a Vanquish Flex Ultra-High-Performance Liquid Chromatography (UHPLC) system that was coupled with an Orbitrap Exploris 240 Mass Spectrometer (Thermo Fisher Scientific). Chromatographic separation was performed using a Kinetex 2.6 µm Polar C18 reversed phase UHPLC column with a constant flow rate of 500 μL/min (Phenomenex). LC–MS/MS acquisition followed an isocratic elution and a multistep linear gradient (S14 Table). Data-dependent acquisition mode was then used for acquisition of tandem MS (S14 Table). Analytical blanks, pool samples, and mixtures of sulfamethazine, sulfamethizole, sulfachloropyridazine, sulfadimethoxine, amitriptyline, and coumarin-314 at 10 μM were injected after every 10 samples as quality control for monitoring instrument performance.

Raw LC–MS/MS data files were processed in MZmine (v. 4.6.1; [96]) to generate a feature table, peak list, and batch file of all settings, with peak area of detected ions being used as metabolite abundance (S10 Table). Annotations and chemical classification of detected molecules were obtained with SIRIUS (v. 6.1.1.1; [97]) (S9 Table), with these annotations corresponding to level 3 of the Metabolomics Standards Initiative [98,99]. This was then used for a feature-based molecular network [100] in GNPS2 [101]. We generated a rarefaction curve to assess whether the metabolite diversity is fully represented (S33 Fig). We performed a principal component analysis and separate PERMANOVA in R to compare the metabolome across developmental stages as well as between light and dark for eggs and larvae. We then performed a Partial Least Squares Discriminant Analysis to identify differentially abundant metabolites (i.e., Variable Importance in Projection, or VIPs) of larvae in light and dark. We summarized the total contribution of these differentially abundant metabolites in light and dark at the pathway, superclass, and class levels. Lastly, we used normalized data to summarize the abundance of two VIPs and assessed which metabolites are within their respective molecular family from the feature-based molecular network [102].

Estimating and modeling dispersal

We calculated the duration required for the survival of the last larva from the clutch of A. lixula with and without the light-dependent activity of what we presume are chromoplast components. The abundance of A. lixula offspring (N_t) after a specific time interval (t; days) can be calculated using the instantaneous rate model: N_t = N₀ e^−Mt, where N₀ is the fecundity, e is the Naperian constant, and M is the instantaneous rate of mortality (day⁻¹) [8]. M was calculated using the survival data for A. lixula offspring in light and dark. Fecundity (N₀) was estimated by spawning individuals (n = 5) until eggs were no longer released, standardizing by total volume, performing two subsequent 1:100 dilutions for 1 mL fully concentrated eggs, and counting triplicate 0.5 mL aliquots of dilute eggs using a dissecting microscope (Leica EZ4). The specific time interval was then calculated for offspring that do and do not benefit from the light-dependent activity of chromoplast components. We compared our instantaneous rates of mortality to the literature to verify if larvae cultured in light and dark were both well within the expected range for the marine invertebrate larvae that develop by planktotrophy [8,9,35].

We integrated these specific time intervals into the VIKING20X [36], a full-scale, nested, and eddy-rich oceanographic model of the Atlantic Ocean. Specifically, we used the VIKING20X-OMIP [36] to perform our dispersal experiments because of its realistic ability to simulate atmospheric forcing [103], to accurately simulate the large-scale horizontal circulation, distribution of mesoscale, overflow, and convective processes, and the representation of regional current systems in the North and South Atlantic Ocean [104,105]. This combination of oceanographic features has previously enabled the VIKING20X to be used to simulate larval dispersal of coastal and deep-sea marine invertebrates [106,107]. We used the Lagrangian simulation software tool Parcels (v. 2.3.1; [108]) to simulate the 3D advection of virtual larvae based on the daily-mean velocities for ocean currents in the upper 12 vertical levels (~0–130 m) from 2007 to 2016. Virtual particles were released at depths between 0 and 30 m in the coastal waters around Tenerife (S21 Fig). Our simulation released 10,000 particles each week for 10 years (i.e., from 2007 to 2016) to account for seasonal and inter-annual variability. Virtual particles had a lifetime of either 102 (i.e., without benefits from the light-dependent activity of chromoplast components) or 181 days (i.e., with benefits from the light-dependent activity of chromoplast components). The abundance, geographical location, and distance of particles from each treatment that reached the continental shelf of Africa were tracked. Particles that reached the continental shelf of Africa in 25 days or less were not counted due to the minimal pelagic larval duration for A. lixula [78]. Each of these quantitative assessments were compared statistically using paired two-tailed t-tests within Prism.

Using this same oceanographic framework, we determined how many iterations (i.e., generations) it took for virtual larvae to spread along the African shelf and reach the southernmost range for A. lixula. Generation 0 was considered Tenerife and Generation 1 was the resulting dispersal. The location and abundance of particles from Generation 1 were then used to determine the origins and then spread in Generation 2. This was repeated until a particle reached the southernmost range for A. lixula. Next, we evenly distributed 2.0 million particles along the African shelf (from 28°N to 10°S; S25 Fig) to determine if any and which locations could be used as the departure area for trans-Atlantic dispersal. Only particles with a pelagic dispersal potential reflecting the benefits from the light-dependent activity of chromoplast components and initially located at the southernmost area could disperse across the Atlantic Ocean (Fig 4B), so we then tracked paths that these particles took and compared them to known oceanographic currents. Lastly, we evenly distributed 1.6 million particles along the northern coast of Brazil (from 5°S to 10°S; S30 Fig) to determine if larvae that benefit from the light-dependent activity of chromoplast components could disperse back to Macaronesia and towards Uruguay (i.e., the southernmost range for A. lixula in South America) [24,109]. This modeling experiment was performed for four years to determine when virtual particles reached the Azores and the Canary Islands. Visualizations of larval dispersal used maps from cartopy [110], with the underlying vector map using NaturalEarth (naturalearthdata.com).