Massive Accumulation of Strontium and Barium in Diplonemid Protists

ABSTRACT Barium and strontium are often used as proxies of marine productivity in palaeoceanographic reconstructions of global climate. However, long-searched biological drivers for such correlations remain unknown. Here, we report that taxa within one of the most abundant groups of marine planktonic protists, diplonemids (Euglenozoa), are potent accumulators of intracellular barite (BaSO4), celestite (SrSO4), and strontiobarite (Ba,Sr)SO4. In culture, Namystinia karyoxenos accumulates Ba2+ and Sr2+ 42,000 and 10,000 times higher than the surrounding medium, forming barite and celestite representing 90% of the dry weight, the greatest concentration in biomass known to date. As heterotrophs, diplonemids are not restricted to the photic zone, and they are widespread in the oceans in astonishing abundance and diversity, as their distribution correlates with environmental particulate barite and celestite, prevailing in the mesopelagic zone. We found diplonemid predators, the filter-feeding zooplankton that produces fecal pellets containing the undigested celestite from diplonemids, facilitating its deposition on the seafloor. To the best of our knowledge, evidence for diplonemid biomineralization presents the strongest explanation for the occurrence of particulate barite and celestite in the marine environment. Both structures of the crystals and their variable chemical compositions found in diplonemids fit the properties of environmentally sampled particulate barite and celestite. Finally, we propose that diplonemids, which emerged during the Neoproterozoic era, qualify as impactful players in Ba2+/Sr2+ cycling in the ocean that has possibly contributed to sedimentary rock formation over long geological periods.

other organisms (10,19). In the world's oceans, diplonemids have only recently been recognized as omnipresent and one of the most diverse and abundant groups of microeukaryotes (comparable to microalgae), with a prevalence within the mesopelagic protist community (32)(33)(34). Although relatively rare, they are present in freshwater bodies as well (35). We analyzed their crystalline inclusions by a range of complementary approaches and discuss here their possible biological functions and role in biogeochemical cycles.

RESULTS
Light microscopy and analysis of crystals by Raman microscopy. To determine the chemical composition of biogenic crystals directly within intact cells, Raman microscopy, a vibrational spectroscopic method sensitive to molecular composition, was used. Out of 21 strains belonging to 15 diplonemid species, three members of the distantly related genera Lacrimia and Namystinia, represented by Lacrimia sp. YPF1808, Lacrima lanifica, and Namystinia karyoxenos, were shown to possess celestite crystals (Fig. 1A). Their Raman spectra were congruent with the spectra of mineral celestite and chemically prepared precipitates of SrSO 4 (Fig. 1F), matching also Raman spectra of celestite reported elsewhere (36). Due to the countercation sensitivity of the position of the most intense Raman band at around 1,000 cm 21 belonging to the symmetric 1 vibrational mode of the SO 4 2tetrahedron, biogenic celestite could be unambiguously identified as SrSO 4 and was easily distinguishable from barite, baritocelestite, gypsum (CaSO 4 ), or calcite (CaCO 3 ) (see Fig. S1 in the supplemental material). Small relative-intensity nuances of other Raman bands of biogenic celestite from various cells (Fig. S2) can be explained by differences in crystal structures (lattice defects), trace admixtures of Ba 21 , and/or orientations of the crystals, as these are present also in the spectra of mineral reference and chemical precipitates.
When N. karyoxenos was examined by light microscopy with differential interference contrast (DIC), the crystalline structures appeared as small birefringent particles moving fast by Brownian motion (Fig. 1B and C; Movie S1). They were most prominent within the enlarged lacunae, which are peripheral membrane-bounded compartments positioned directly beneath the subpellicular microtubular corset (Movie S1), following the addition of 0.1% (wt/vol) formaldehyde. Within a single culture, the size and quantity of crystals inside the cells ranged from a few small particles (Fig. 1B) up to multiple large, tightly packed, polygonal crystals reflecting the shape of orthorhombic prisms (Fig. 1C). The crystalline particles of Lacrimia sp. YPF1808 and L. lanifica were far less prominent under the light microscope than those of N. karyoxenos. However, large crystals were visible around the posterior vacuole with DIC ( Fig. 1D and E; Movie S1) and under polarized light (Movie S1).
Morphology, localization, and elemental analysis of intracellular crystals. Examination with light, Raman, transmission electron microscopy (TEM), and serial blockface scanning electron microscopy (SBF-SEM) showed that the crystalline inclusions in two clades of diplonemids differed in their localization and shapes. In semithin resin-embedded sections of N. karyoxenos, numerous orthorhombic prismatic and bipyramidal crystals were localized mostly inside the lacunae ( Fig. 2A, C, and D), with a preference towards the cell posterior. Occasionally, crystals were found inside the large posterior vacuole ( Fig. 2A) and in smaller vacuoles scattered throughout the cytoplasm ( Fig. 2A  and B). Only small crystals could be seen in semithin sections, while bigger crystals dropped out, leaving empty crystal-shaped holes. Due to frequent rupturing, it was not possible to visualize celestite crystals in semithin epoxy resin sections of Lacrimia species. Thus, we used the SBF-SEM approach, which showed that the celestite crystals of Lacrimia sp. YPF1808 appeared mostly in small membrane-bounded compartments with electron-transparent matrix ( Fig. 2H to L) adjacent to the large posterior vacuole (Fig. 2H, I, and K). Three-dimensional (3D) reconstruction revealed that each of these compartments contained one crystal of variable size (Movie S2). Less frequently, crystals were found inside the posterior vacuole ( Fig. 2I) or in compartments localized near the anterior flagellar pocket (Movie S2), while they were absent from the cytoplasm and other organelles. The crystals had a shape of rhombic prisms ( Fig. 2J and M) or asymmetric tabular prismatic structures with pyramidal and pedial terminations ( Fig. 2K and N). Although in L. lanifica the celestite crystals were mostly lost from the TEM sections, the positions of holes and ruptures within them and the analysis by Raman microscopy showed similar localizations and sizes of the crystals as those of Lacrimia sp. YPF1808 (Fig. 2D, E, H, and I). Likewise, the membrane-bounded compartments were positioned around the posterior vacuole (Fig. 2F), with small asymmetric flattened crystals preserved only occasionally in TEM sections (Fig. 2G).
The presence of celestite crystals ( Fig. 1 and 2) was further confirmed by elemental analysis using energy-dispersive X-ray (EDX) spectroscopy in the cryo-SEM-EDX mode of freeze-fractured Lacrimia sp. YPF1808 ( Fig. 3A and C; Fig. S3) and N. karyoxenos cells ( Fig. 3B and C; Fig. S4) and by TEM-EDX of whole air-dried cells of L. lanifica (Fig. 3E). Atomic percentages estimated by cryo-SEM-EDX analysis were 7.2% Sr and 7.2% sulfur (S), compared to 1.1% Sr and 1.8% S in Lacrimia sp. YPF1808 and N. karyoxenos, respectively. The dominance of C, N, and O atoms can be explained by the presence of ice and signals from other cellular contents obtained from deeper and/or surrounding areas.
The identities of celestite crystals in Lacrimia sp. YPF1808 (1 mm-thick sections from resin blocks used for SBF-SEM) and N. karyoxenos (250 nm-thick resin sections examined by TEM) were confirmed by SEM-EDX and TEM-EDX, respectively (Fig. 3D, F, and G; Fig. S5). Additionally, a significant amount of Ba 2+ was detected in the crystals from N. karyoxenos. Crystallographic analysis by electron diffraction showed that the diffraction of measured crystals corresponded to celestite structure (isostructural with BaSO 4 ) with space group Pnma and lattice parameters a = 8.3 Å, b = 5.3 Å, and c = 6.8 Å in L. lanifica. Larger lattice parameters (a = 8.7 Å, b = 5.5 Å, c = 7.1 Å) were observed in N. karyoxenos, which may be explained by the replacement of Sr 21 with larger Ba 21 in the structure of celestite.
Quantitative analysis by ICP-MS and SBF-SEM. SBF-SEM-based 3D reconstructions of Lacrimia sp. YPF1808 (Movie S2) showed the presence of celestite crystals in all 20 analyzed cells, ranging from 2 to 16 celestite particles per cell ( Fig. 4D; Table S2). In total, more than 100 crystals were analyzed, with a volume ranging from 0.017 to Accumulation of Strontium and Barium in Diplonemids mBio 7 mm 3 ( Fig. 4C; Table S2). The impacts of the measured celestite contents on the overall cell density ranged from 0.05% to 9%, with an average of 1.3 6 0.5% ( Fig. 4E; Table S2). The calculations were based on the measured volumes, known density of celestite (3.9 gÁcm 23 ), and common cellular densities of 0.985 to 1.156 gÁcm 23 reported elsewhere (37). The lack of celestite crystals in other analyzed species (Diplonema japonicum, Paradiplonema papillatum, and Rhynchopus sp. YZ270) was consistent with the minute 88 Sr content measured by ICP-MS. The high values of 88 Sr in N. karyoxenos, Lacrimia sp. YPF1808, and L. lanifica corresponded to the abundance of intracellular crystals Electron diffraction for h0l-oriented sections through the 3D ED data sets from the corresponding areas shown in panels E and F. The celestite unit cell is displayed as a yellow rectangle. Scale bar, 2 mm.
Accumulation of Strontium and Barium in Diplonemids mBio detected by Raman microscopy, TEM, and SBF-SEM. Since direct measurement of the dry mass was impossible due to the inevitable presence of salts from the medium, the elemental composition analysis by ICP-MS was calculated in atoms per cell or femtomoles per cell. To calculate Sr and Ba content per dry mass, the latter was subsequently estimated by quantitative phase imaging using holographic microscopy ( Table 1). The 88 Sr amounts ranged from 0.01 fmolÁcell 21 in P. papillatum to 5,500 6 570 fmolÁcell 21 (mean 6 standard deviation) in N. karyoxenos, corresponding to 340 6 38 mgÁg 21 . Lacrimia sp. YPF1808 and L. lanifica were also potent 88 Sr accumulators, with 370 6 58 fmolÁcell 21 (130 6 25 mgÁg 21 ) and 54 6 8 fmolÁcell 21 (64 6 13 mgÁg 21 ), respectively. Depending on the species, the intracellular concentration of 88 Sr was 1,200 to almost 10,000 times higher than in the surrounding medium (Tables 1 and S2).
Compared to the massive accumulation of 88 Sr in diplonemids, the naturally cooccurring Ba was present in much lower amounts (Table 1 and S2), slightly above the detection limit of TEM-EDX analysis (Fig. 3G), possibly reflecting a 12.5 times lower Ba 21 concentration in the seawater growth medium (Table S1). Nevertheless, the cells concentrated Ba 21 on average 1,000 to over 42,000 times above the level in the growth medium (Table S1), reaching 1,200 6 130 fmolÁcell 21 (120 6 14 mgÁg 21 ) in N. karyoxenos, with lower values in Lacrimia sp. YPF1808 (65 6 10 fmolÁcell 21 , or 36 6 7 mgÁg 21 ) and L. lanifica (2 6 1 fmolÁcell 21 , or 4 6 1 mgÁg 21 ). Altogether, (Ba,Sr)SO 4 accumulation reached 91%, 33%, and 14% of dry weight in N. karyoxenos, Lacrimia sp. YPF1808, and L. lanifica, respectively. All strains showed significantly different levels of accumulated Sr or Ba contents (Table 1). Although the concentration of Ba 21 in our experiment was 30 to 45 times higher than in nature, the stoichiometry per cell was 5 orders of magnitude lower than in the oceans (Table 1).

Accumulation of Strontium and Barium in Diplonemids mBio
Ba 2+ loading experiments and elimination of Sr 2+ and Ba 2+ from the medium. Under our cultivation conditions, Ba 21 was 20 times less abundant than Sr 21 (Table S1), which resulted in formation of celestite with Ba 21 admixtures in the examined species. In oceans, the proportion is even 30 times lower (11,12). Here, we tested whether increases in Ba 21 content would impact the biomineralization process by cultivation in artificial seawater loaded with equimolar amounts of Ba 21 and Sr 21 to 88 mM. To prevent spontaneous precipitation of barite, we supplemented sulfates with NaCl to maintain the osmolarity of the medium. This resulted in the formation of all mineral combinations: pure barite (Raman marker at 988 cm 21 ), celestite (1,000 cm 21 ), and mixed forms of (Ba,Sr)SO 4 (991 cm 21 ; strontiobarite or baritocelestite) (Fig. 5), revealing that diplonemids do not show a strong preference toward the accumulation of either element.
Both L. lafinica and Lacrimia sp. YPF1808 contained mostly mixed crystals of (Ba,Sr)SO 4 . In N. karyoxenos, pure celestite prevailed in the central part of crystals (999 cm 21 ) while barite dominated at the periphery (988 cm 21 ), and they gradually overlapped each other ( Fig. 5C to F). Additionally, after several passages in the artificial medium without Ba 21 and Sr 21 (Table S1), barite and celestite were no longer detectable by Raman microscopy. The lack of Ba 21 and Sr 21 did not result in altered morphology or growth impairment. We did not observe any Ca 21 -or Mg 21 -containing crystals despite high concentrations of both elements in the medium.
Feeding experiments. Fecal pellets are an important agent mediating sedimentation of biogenically accumulated minerals to the sea floor: large aggregates of crystals held together by undigested fecal organic matter enable their fast sinking, thus preventing dissolution of micrometer-sized crystals in the water column, which is undersaturated for barite and celestite (15,24). To experimentally address whether zooplankton feeds on diplonemids and whether their celestite crystals are carried into fecal pellets, we incubated N. karyoxenos and Lacrimia sp. YPF1808 with freshly captured filter-feeding marine copepods Centropages typicus, Temora longicornis, and Acartia sp., starved for 12 h prior to the experiment. After 5-day cocultivation, we determined by Raman microscopy that the fecal pellets contained copious amounts of celestite derived from diplonemids (Fig. 6B). In the control system of the same population of copepods fed with freshly collected marine plankton, the n/a n/a n/a 135 271 715 Concentration of Sr (folds)* b n/a n/a n/a 1,520 6 680 1,240 6 550 10,000 6 3,300 Ba (fmolÁcell -1 ) n/a n/a n/a 2.0 6 0.5 65 6 10 1,200 6 130 Ba (mgÁg -1 ) n/a n/a n/a 3.7 6 1.1 35.8 6 6.8 116 6 14 p n/a n/a n/a a b c BaSO 4 (mgÁg -1 ) n/a n/a n/a 6.3 61 198 Concentration of Ba (folds)* b n/a n/a n/a 1,130 6 660 4,300 6 1,900 42,000 6 14,000 Total Ba,Sr(SO 4 ) (mgÁg -1 ) n/a n/a n/a 141 332 913  Table S1); in N. karyoxenos, the amount of accumulated Ba exceeds total Ba available per volume of culture medium due to repeated passages of pelleted cells into fresh medium. n/a, not analyzed.

DISCUSSION
The most studied biominerals in protists are extracellular calcite scales of haptophytes and silicate frustules of diatoms, while studies on intracellular mineral crystals are far less common (38). After more than a century since the skeletons of marine acanthareans and freshwater streptophytes were found to contain celestite (16) and barite (39), respectively, we have identified potent accumulators of Ba 21 and Sr 21 in an unexpected group of eukaryotes, the diplonemids.
The heterotrophic diplonemids are widespread in the oceans and, as recently described, in astonishing abundance and diversity (31,33). Despite their abundance and extreme diversity, diplonemid flagellates remain a poorly known group of protists (34) that are abundant from the surface to the deep sea, with a wide peak in the The major source of trace elements is driven by river influx and less prominently by aerial deposition (9) and is mostly balanced by the same amount of total deposition in sediments, correlated with the total marine productivity and particulate organic matter deposition (24). A great proportion of these trace elements is being recycled by living organisms after release from dead cells. Acantharea (image adapted from referene 69) take up a substantial portion of Sr 21 and Ba 21 , which are further recycled in upper 400 m (25). Coccolithophorids build their scales from calcium carbonate with minor amounts of Ba 21 and Sr 21 that are proportional to the seawater contents (22), being partially recycled upon dissolution or transported to the marine sediments in fecal pellets (24). The Ba 21 accumulated by bacteria sediments in the aggregates of marine snow (26) or is recycled. We highlighted diplonemids as potential players in the marine cycle of both elements and drivers of biogenic formation of celestite and barite crystals found in suspended matter everywhere in the world oceans (24), and they can also feed on bacteria or particulate organic matter or scavenge the dead bodies of zooplankton as major heterotrophic protists in the mesopelagic zone. (B) Raman microscopy analysis of fecal pellets produced by copepods experimentally fed with diplonemids, which contained celestite crystals (cyan) and undigested lipids, including sterols (gray). In the control samples of copepods fed with microalgae, fecal pellets contained undigested carotenoids, chlorophyll (yellow-orange-red), and calcite particles (pink), with unexpected particles of polystyrene microplastics (measured as a single spectrum [not shown on the Raman map]). Scale bars, 20 mm. mesopelagic zone (32,33,40,41). The high capacity of intracellular Sr 21 and Ba 21 accumulation in some diplonemids outperforms that of any other reported organisms (10,13,21,(42)(43)(44). Indeed, while the intracellular concentration of Sr 21 in the most efficient accumulators known thus far (yeasts, desmids, and cyanobacteria) reaches a maximum of 220 mgÁg 21 per dry weight (1,10,43), N. karyoxenos contains as much as 340 mgÁg 21 Sr 21 together with 120 mgÁg 21 Ba 21 , which in the form of sulfate represents 90% of the cellular dry mass, pointing to the unique Sr 21 and Ba 21 accumulation capacity of this diplonemid, while both Lacrimia species are slightly less potent in this respect (Table 1). Interestingly, when both trace elements are provided in equimolar concentrations, diplonemids form pure celestite and barite and/or mixed forms of (Ba,Sr)SO 4 , apparently not discriminating one element over the other. Hence, we explain the higher content of Sr 21 over Ba 21 inside the crystals by the higher availability of the former element in seawater. Although the mechanisms behind intracellular accumulation of Sr 21 and Ba 21 are largely unknown, it has been suggested that mineral crystals typically occur in membrane-bounded compartments or vacuoles, in which they are formed from supersaturated solutions via precisely regulated nucleation (13). The Sr 21 uptake and transportation within eukaryotic cells have been shown to occur via commonly present transporters of divalent cations, i.e., the Ca 21 uniporter and H 1 /Ca 21 antiporter (45,46). The diplonemid nuclear genome is not yet available, but these transporters have been documented in the related kinetoplastid Trypanosoma brucei (47). Although the reported affinity to Ca 21 and Sr 21 is usually comparable (45,46), some organisms including diplonemids clearly favor Ba 21 and Sr 21 over Ca 21 (10). When such vacuoles contain sulfate solutions, they may function as a "sulfate trap" for those cations that precipitate easily in the presence of sulfates (2). At the same time, we did not observe CaSO 4 or any of its forms (gypsum, bassanite, anhydride, etc.), even though the concentration of Ca 21 in the cultivation medium or in the environment is several orders of magnitude higher than that of Sr 21 and Ba 21 .
Densities of celestite and barite of 3.9 gÁcm 23 and 4.5 gÁcm 23 , respectively, have been repeatedly reported as statoliths in ciliates or charophytic algae (13,17,18). In comparison to the seawater density of 1.03 gÁcm 23 and typical cell density range between 0.985 and 1.156 gÁcm 23 , the heavy crystals may help maintain appropriate buoyancy by counterbalancing light lipid droplets (0.86 gÁcm 23 ) (37,48). Indeed, the impact of celestite crystals is substantial, since they may increase the overall density of Lacrimia sp. YPF1808 and N. karyoxenos by up to 9% and 27%, respectively (Table S2). According to Stokes' law for small particles of low Reynolds numbers, the barite/celestite ballasting can significantly increase the sedimentary velocity for up to 50 to 200 m per month or 0.5 to 2 km per year (Table S2). Hence, while the function of biomineralization in diplonemids remains unknown, we speculate that they may benefit from gravitropic sensing, which would allow directed movement and/or enable passive sedimentation. Another intriguing impact of barite and celestite is associated with their propensity to strong absorption of UV and blue light (49). Hence, in surface waters, these minerals may contribute to UV protection. It is reasonable to assume that by forming celestite, protists adjust their inner osmolarity, the principle analogical to the formation of other cell inclusions, such as oxalate, calcite, or polyphosphate, that are either dissolved and osmotically active or crystallized or polymerized and osmotically inactive inside a vacuole (13,50).
Celestite-forming acanthareans are considered key players in the upper 400 m of the ocean, yet do not contribute to the sedimentary rock formation, as their skeletons dissolve upon decay of their cells (25). Coccolithophorids and bacteria produce carbonates (44) and/or phosphates (26) of Ba 21 /Sr 21 , which can also be converted to sulfates either on the bacterial extracellular polymeric substances or in the microenvironment of decaying matter of marine snow aggregates in the process of diagenesis (26). In the chemical continuum between pure barite and celestite, the latter represents 10 to 30% (24), gradually decreasing, depending on the depth (11,12). The majority of biogenic particulate barite and celestite is recycled by simple dissolution (25), microbial loop (26), or resuspension of sediments (24). However, the overall influx into the system is balanced by sedimentary deposition (9,24), which might have a biological driver. Seminal work of Dehairs et al. (24) scrutinized all potential sources of particulate barite and celestite, and they did not find experimental support for either Ba 21 incorporation in siliceous plankton or precipitations on decaying organic matter in sulfate-enriched microenvironments. Hence, they ultimately favored the biogenic origin of particulate barite/celestite being hypothetically formed by microorganisms inhabiting the highproductivity mesopelagic zone (24) only to remain unknown since then. These predictions nicely correlate with our measurements in diplonemids, indicating that micronsized celestite and sometimes barite crystals of variable Ba-Sr ratios (Fig. 2 to 5) are scattered throughout the water column of the world's oceans, with the highest prevalence in the mesopelagic zone (32). Moreover, particulate barite/celestite is often found in fecal pellets and aggregates of marine snow, and finally, in the sediments (24,27,32). By providing celestite-containing diplonemids to filter-feeding copepods, we found undigested celestite in their fecal pellets (Fig. 6), the main transport system of micrometric biominerals into the sediments, although the majority is recycled (24). Thus, diplonemids may be involved in Ba 21 /Sr 21 cycling and/or in sedimentary deposition of celestite or barite. Since these protists likely emerged during the Neoproterozoic era (590 to 900 million years ago [MYA]), overlapping with the Ediacaran period (51), their impact on biogenic marine sediments may cover several geological eras. The coccolithophores appeared around the same time as diplonemids, yet the onset of carbonate biomineralization has been timed to ;200 MYA (52).
As another ecological addition to the big picture of Ba 21 /Sr 21 cycling, diplonemids have been shown to ingest bacteria as one of their sources of nutrition (30); if bacteria were loaded with Ba 21 /Sr 21 in the form of (poly)phosphates, as reported elsewhere (26), diplonemids may further transform it into barite upon digestion. Additionally, diplonemids are likely to feed on the organic matter of marine snow providing preconcentrated Ba 21 , in which case they may accumulate more Ba 21 than Sr 21 . In principle, we experimentally supported such a scenario upon doping the cells with equimolar Ba 21 and Sr 21 concentrations (Fig. 5). Finally, we do not exclude that some species of diplonemids to be described in future would prefer Ba 21 over Sr 21 or that there are other as-yet-unknown microbial bioaccumulators of these trace elements.
Based on the ability of some diplonemids to store massive amounts of celestite and to lesser extent barite, we speculate that more as-yet-unknown diplonemid species may qualify as impactful players of Ba 21 /Sr 21 flow through the food web, eventually influencing the sedimentary records.

MATERIALS AND METHODS
Cell cultures, cultivation, and light microscopy. For all experiments, axenic cultures were grown in seawater-based Hemi medium (see Table S1 in the supplemental material) supplemented with 1% horse serum and 0.025 g/liter LB broth powder (53). An artificial seawater medium lacking Sr 21 , Ba 21 , and sulfates was prepared from 288 mM NaCl, 8 mM KCl, 718 mM KBr, 100 mM MgCl 2 , 12 mM CaCl 2 , 40 mM HBO 3 , and 60 mM NaF, supplemented with 1% (vol/vol) heat-inactivated horse serum (Sigma-Aldrich) and 25 mg LB broth powder (Amresco). The medium was used as rinsing solution for preparation of ICP-MS samples and cell microcrystal depletion to be measured via quantitative phase imaging (QPI) (see below). For Ba 21 loading experiments, BaCl 2 was added in equimolar amounts with respect to naturally occurring (88 mM) Sr 21 (12).

Accumulation of Strontium and Barium in Diplonemids mBio
Light microscopy images and videos were taken with an Olympus BX53 microscope equipped with a DP72 microscope digital camera using CellSens software v. 1.11 (Olympus) and processed with GIMP v. 2.10.14, Irfan View v. 4.54, and Image J v. 1.51 software. Polarized microscopy was performed using crossed polarizers installed to a Raman microscope (as specified below).
Environmental sampling. Zooplankton was collected in the Bay of Villefranche sur Mer, France (43°409N, 7°199E) with a 10-min haul from 10 m to the surface, using a 20-mm-mesh-size plankton net. Captured copepods were transferred into 0.5 liters of freshly filtered natural seawater and starved for 12 h. Centropages typicus, Temora longicornis, and Acartia sp. were then picked under a dissection microscope. All experiments were carried out at cultivation temperature of the prey species of diplonemids (13°C for Lacrimia sp. YPF1808, room temperature for N. karyoxenos). Ten copepods were kept in 20 mL of diplonemid culture (10 5 cells mL 21 ) for 5 days, after which their fecal pellets were collected under a dissection microscope and immediately analyzed by Raman microscopy (as specified below).
Raman microscopy. For the in situ determination of the chemical composition of intracellular structures, a confocal Raman microscope (alpha300 RSA; WITec, Germany) was used as previously described (56)(57)(58)(59)(60). To immobilize the fast-moving flagellates on the quartz slide, 5 mL of the cell pellet was mixed with 5 mL of 1% (wt/vol) solution of low-melting-point agarose (catalog number 6351.5; Carl Roth, Germany), immediately spread as a single-cell layer between a quartz slide and coverslip, and sealed with CoverGrip sealant (Biotium, USA). Two-dimensional Raman maps were obtained with laser excitation at 532 nm (20 mW power at the focal plane) and oil-immersion objective UPlanFLN 100Â, numerical aperture (NA) 1.30, or water-immersion objective UPlanSApo 60Â, NA 1.20 (Olympus, Japan). A scanning step size of 200 nm in both directions and an integration time of 100 ms per voxel were used. A minimum of 30 cells were measured for each strain. Raman chemical maps were constructed by multivariate decomposition of the baseline-corrected spectra into the spectra of pure chemical components by using Project Plus 5.1 software (WITec, Germany).
TEM, TEM-ED, and TEM-EDX. The protocol for the basic sample preparation of all kinds of electron microscopy approaches listed here is described in detail elsewhere (61). We used it with minor modifications, as stated below. Cell pellets were transferred to specimen carriers and immediately frozen in the presence of 20% (wt/vol) bovine serum albumin solution using a high-pressure freezer (Leica EM ICE, Leica Microsystems, Austria). Freeze substitution was performed in the presence of 2% osmium tetroxide diluted in 100% acetone at 290°C. After 96 h, specimens were warmed to 220°C at a step of 5°C/h. After another 24 h, the temperature was increased to 3°C (3°C/h). At room temperature, samples were washed in acetone and infiltrated with 25%, 50%, and 75% acetone/resin mixture for 1 h at each step. Finally, samples were infiltrated in 100% resin and polymerized at 60°C for 48 h. Semithin (250 nm) and ultrathin (70 nm) sections were cut using a diamond knife, placed on copper grids, and stained with uranyl acetate and lead citrate. TEM micrographs were taken with a Mega View III camera (SIS) using a JEOL 1010 TEM operating at an accelerating voltage of 80 kV.
For TEM-EDX, 10 mL of pelleted L. lanifica cells was spread over a holey carbon-coated copper grid, washed twice with 10 mL of distilled water in order to reduce the sea salts from the culture medium, and allowed to dry by evaporation at ambient temperature. Semithin sections of resin-infiltrated blocks of N. karyoxenos were prepared as stated above. For the identification of the crystalline phase, sections were studied by TEM on an FEI Tecnai 20 system (LaB6, 120 kV) equipped with an Olympus SIS chargecoupled-device camera Veleta (2,048 by 2,048 pixels) and an EDAX windowless EDX detector Apollo XLTW for elemental analysis. The diffraction data were collected by means of 3D electron diffraction (ED) (62). The data processing was carried out using PETS software (63). Structure solution and refinement were performed in the computing system Jana2006 (64).
Cryo-scanning electron microscopy with EDX. Cells pellets were high-pressure frozen as described above and transferred into a Leica ACE 600 preparation chamber (Leica Microsystems, Austria) precooled at 2135°C, fractured with a scalpel, freeze-etched at 2100°C for 1 min, and sputter-coated with 2.5 nm of gold-palladium at 2125°C. Specimens were transferred under vacuum using transfer system VCT100 (Leica Microsystems, Austria) and observed with a Magellan 400L SEM (FEI, Czech Republic and USA) precooled at 2125°C (cryo-SEM). Topographical images and EDX measurements were obtained using an EDT detector and EDAX detector (Octane Elect Super; EDAX, USA), respectively, either at 5 keV/0.1 NA or 10 keV/0.8 NA. The taken spectra were analyzed with EDAX TEAM software and quantified by the eZAF method.
Serial block-face SEM. The sample preparation of Lacrimia sp. YPF1808 by the high-pressure freezing technique followed the protocol for TEM sample preparation. After freeze-substitution, the samples were subsequently stained with 1% thiocarbohydrazide in 100% acetone for 1.5 h, 2% OsO 4 in 100% acetone for 2 h at room temperature, and 1% uranyl acetate in 100% acetone overnight at 4°C. After every staining step, the samples were washed 3 times with 100% acetone for 15 min. Samples were then infiltrated with 25%, 50%, or 75% acetone-resin mixture for 2 h at each step, and finally infiltrated in 100% Hard Resin Plus 812 (EMS) overnight and polymerized at 62°C for 48 h. Resin-embedded blocks were trimmed and imaged using an Apreo SEM equipped with a VolumeScope (Thermo Fisher Scientific, Germany). Serial images were acquired at 3.5 keV, 50 pA, 40 Pa with a resolution of 6 nm, 100-nm slice thickness, and dwell time per pixel of 4 ms. Image data were processed in Microscopy Image Browser v2.702 (65) and Amira v2020.2. The resin-embedded blocks were also collected in the form of 1-mm-thick sections on a silicon wafer and analyzed by SEM-EDX (Magellan 400L system, as described above).
Based on volumetric data, we calculated the percentage of increase in cell density based on measured volumes of crystals compared to the theoretical crystal-free cells of the same volume and reported average theoretical density of 1.07 gÁcm 23 (37).

ICP-MS.
For analysis of Ba and 88 Sr concentrations, cultures were grown in triplicates, counted, and washed three times with 1 M sorbitol solution (for P. papillatum, D. japonicum, and Rhynchopus YZ270 cl. 10) or Sr-and Ba-free artificial seawater rinsing solution (see above) (for L. lanifica JW1601, Lacrimia sp. 1808, and N. karyoxenos) to remove Ba 21 and Sr 21 present in the cultivation medium. Cultivated cells were harvested by centrifugation and rinsed twice with 50 mL and once with 2 mL of the rinsing solution, and the resulting pellets were freeze-dried. A 0.5-mL digestion acid mix (425 mL of 70% HClO 4 and 75 mL of 69% HNO 3 ) prepared as described elsewhere (66) was added directly to the dried biomass. The digestion was done using a Fuji PXG4 Thermoblock (AHF Analysentechnik AG, Germany). After evaporation of the acid mix, 0.5 mL of 5% HCl was added to each test tube to redissolve the salts. The glass tubes were heated to 90°C for 1 h to obtain clear solutions. The final volume of 1.5 mL was adjusted with double-distilled H 2 O. Appropriate dilutions were made with 0.2% HNO 3 . Indium was added as an internal standard at 1 ng/mL to each test solution. The ICP multielement standard solution VI (Merck, Germany) was used to prepare standard curves. Analyses were done using an inductively coupled plasma sector-field mass spectrometer (ICP sfMS) Element XR-2 with jet interface (Thermo Fisher Scientific, Germany) following a described protocol (67). Medium resolution of 4,000 was used in Ba and 88 Sr measurements in triplicate of each technical replicate, with the highest precision and lowest relative standard deviation. Additionally, the elemental composition of samples of standard growth medium and artificial seawater medium without sulfates, Ba 21 , or Sr 21 were analyzed.
Holographic microscopy and QPI. Samples for holographic microscopy were immobilized prior to measurement as described above for Raman microscopy. Imaging was performed at the Q-Phase microscope (Tescan Orsay Holding, Czech Republic). The holographic Q-Phase microscope is equipped with halogen lamp illumination through an interference filter (l[¼] 650 nm, 10 nm full-width, half-maximal) and microscope objective (Nikon Plan Fluor oil immersion, 60Â, numerical aperture 1.4, providing lateral resolution of 0.57 mm). The numerical reconstruction of acquired data was performed using Q-Phase software (Tescan Orsay Holding, Czech Republic). The technique enables automated cell segmentation and quantitative analysis of cellular mass based on the specific proportions of thickness and refractive indices of measured cells in comparison to the reference (68). Due to the high variability of cell contents and sizes, at least 150 cells were analyzed for each strain. Because crystalline inclusions caused artifacts during capturing due to the big difference in refractive indices, we analyzed crystal-free cells cultivated in the artificial seawater medium lacking Sr 21 , Ba 21 , and sulfates (as specified above). We calculated the total dry mass of the cells as the sum of crystal-free cells, measured by holographic microscopy, and SrSO 4 and BaSO 4 amounts measured via ICP-MS. The dry weight ratios of trace elements measured via ICP-MS were calculated based on the total dry weight of corresponding strains.
Statistical data analysis. Statistical analysis was conducted using SigmaPlot v. 12.5 and SPSS v. 23.0. Logarithmically normalized data were subjected to statistical tests (one-way analysis of variance [ANOVA] and Tukey's post hoc) on an alpha level of 0.05. Calculations of standard errors of the means based on independent methods (i.e., ICP-MS quantification and QPI dry mass quantification) with different levels of variability were done according to mathematical conversion using Taylor expansion.
Data availability. All data generated or analyzed during this study are included in the published article and its supplemental material files.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. MOVIE S1, MOV file, 12.6 MB. MOVIE S2, MOV file, 12.9 MB.