In the living cell, the nucleus of Oxyrrhis marina is located almost at the center of the cell, with a single prominent nucleolus positioned centrally within it ( Figure 1A ). Cells fixed with 1% glutaraldehyde were stained with DAPI and observed with confocal laser microscopy, demonstrating the presence of numerous elongated chromosomes in the nucleus except in the region of the nucleolus ( Figure 1B ).

Figure 1C is a transmission electron microscopy (TEM) image showing nuclear fine structures in O. marina by using conventional chemical fixation methods. As seen in the light microscopic observations ( Figures 1A, B ), a single nucleolus was present at the center of the nucleus. A number of chromosomes were observed as electron-dense granular or rod-like structures (red arrows in Figure 1C ). In the nucleoplasm, there were many small objects of high electron density (white arrows in the inset image of Figure 1C ), which were similar to the ones described by Dodge and Crawford (1971a) as fragmented chromosomes. Various organelles, such as mitochondria, trichocysts, and oil droplets, as well as crystal-like structures of very high electron density, were found within the cytoplasm, which were unevenly distributed around the nucleus and organelles. The remaining intracellular area could be observed as low electron density space. These morphological features are consistent with previous descriptions (Dodge and Crawford, 1971a, 1971b, 1974).

Conventional chemical fixation requires time for the fixative to diffuse through the cell membrane and form cross-linkages between molecules in the cell. Such relatively slow fixation can result in artifacts such as cytoplasmic gaps and organelle swelling. For example, Song et al. (2017) and Hoshina et al. (2021) reported that chemical fixation and subsequent dehydration led to vacuolar swelling, resulting in a gap between the perialgal vacuolar membrane and symbiotic Chlorella in ciliates. In chemically fixed O. marina, the cytoplasm was only distributed around the nucleus and organelles ( Figure 1C ), but this was not consistent with the images of live cells observed by differential interference contrast microscopy ( Figure 1A ), suggesting that conventional chemical fixation methods may produce structural artifacts in the cells. This prompted us to investigate the ultrastructure of O. marina cells using freeze-substitution fixation (which can fix cell structures much more rapidly) and compare it with conventional chemical fixation methods. In cells prepared by freeze-substitution fixation, the cytoplasm was uniformly distributed throughout the cell ( Figure 1D ; Supplementary Figures S1A, B ). In the subcortical region of the cell, the alveoli and mitochondrial cristae that were clearly observed in the chemically fixed cells were obscured. On the other hand, cell scales were well preserved in the freeze-substituted cells. ( Supplementary Figures S1C, D ). The interior of the nucleus was, in outline, similar to that observed with conventional chemical fixation, with one nucleolus and numerous chromosomes. However, small electron-dense objects, such as those described as fragmented chromosomes, were not found at all in the nucleus prepared by freeze-substitution ( Figure 1D ; Supplementary Figures S1E, F ). These findings suggest that freeze-substitution fixation, rather than conventional chemical fixation, is preferable for artifact-free ultrastructural observations of O. marina.

The three-dimensional (3D) structure of the nucleus of O. marina fixed by freeze-substitution was reconstructed using volume electron microscopy, and morphological features of the chromosomes, including the number of chromosomes, were examined ( Figures 2A–E ). Figures 2A–C ( Supplementary Movie S1 ) were reconstructed from 56 consecutive thin sections of 100 nm thickness taken by conventional TEM. Figures 2D, E were obtained from consecutive images sliced at 50 nm thickness by serial block-face scanning electron microscopy (SBF-SEM). The characteristics of the representative nuclei observed are summarized in Table 1 . The estimated number of chromosomes encapsulated in one nucleus based on conventional consecutive TEM images (specimen no. 1 in Table 1 and Figures 2A–C ) was 398, while that estimated from SBF-SEM images (specimen no. 2 in Table 1 , and Figures 2D, E ) was 405. Chromosomes were uniform in thickness (~0.1 µm) but varied in length from 0.1 to 2.5 µm ( Figure 2F ). The median length was 0.73 µm, and about 80% of all chromosomes were between 0.2 and 1.4 µm in length. To further confirm the number of chromosomes, two more nuclei (specimen no. 3 and 4; data not shown) were reconstructed using SBF-SEM. The results estimated 402 and 403 chromosomes, respectively, suggesting that the number of O. marina chromosomes is approximately 400.

In core dinoflagellates, the dinokaryon has multiple nucleoli (sites of ribosome biogenesis) to which many chromosomes adhere. In Prorocentrum micans, the ribosomal RNA gene locus is on chromosomes attached to the nucleolus (Géraud et al., 1991). From here, the region encoding the ribosomal RNA gene is released from the chromosome and extended to the low electron-dense central region, where active rRNA transcription occurs (Géraud et al., 1991). In O. marina, many chromosomes were observed adhering to the nucleolus, which is located in the center of the nucleus ( Figures 1C, D ). The structural relationship between the nucleolus and the chromosomes adhering to it was clarified by reconstructing it in 3D. Here, in addition to the four nucleoli analyzed in the nuclear fine structures in 3D (specimens no. 1 and 2 in Figure 2 and specimens no. 3 and 4), four more nucleoli (specimen no. 5–8: Figures 3 , 4 ) were reconstructed in 3D from the other serially sectioned TEM images (eight nucleoli in total). The total number of chromosomes remained constant (approximately 400) in the four newly observed nuclei, but the number of chromosomes adhering to the nucleolus varied. Only long chromosomes adhered to the nucleolus ( Figures 3 , 4 ). While some chromosomes surrounded the surface of the nucleolus (specimen no. 5 in Figures 3A, B ; Supplementary Movie S2A ), chromosomes penetrating the nucleolus were also observed in some nucleoli (specimen no. 6 in Figures 3C, D ; Supplementary Movie S2B ).

The nucleolus of core dinoflagellates can be divided into two regions according to electron density. The high-density peripheral regions are called granular pre-ribosome compartment (G), whereas the central low-density regions are called fibrillogranular compartment (FG), where active rRNA transcription takes place (de la Espina et al., 2005). The nucleolus of O. marina may also be classified into two similar regions based on differences in electron density ( Figures 4A, C ), where chromosomes penetrating the G regions and reaching the low-density FG regions were also observed (specimen no. 7 in Figures 4A, B ; Supplementary Movie S3A ). In addition, nucleofilaments branching off from chromosomes on the surface of the nucleolus were frequently observed. These nucleofilaments were approximately 6.5 nm thick and penetrated beyond the G compartment into the FG regions at the center of the nucleolus (specimen no.8 in Figures 4C, D ; Supplementary Movie S3B ). Figure 5 is a schematic diagram summarizing the structure of chromosomes and nucleofilaments penetrating the nucleolus.

The periodic arch-shaped structure is a common feature of dinokaryon chromosomes ( Figure 6A ). This feature has been considered to reflect the liquid cholesteric crystal structure of the dinoflagellate chromosome in which parallel nucleofilaments are repeatedly stacked in slightly different alignments (Bouligand et al., 1968). Such a periodic filament structure was not found in the O. marina chromosomes, but some disordered or randomly folded filaments appeared in the chromosome ( Figure 6B ). Since these nucleofilament morphologies were observed in the chemically fixed cells, an attempt was made to reevaluate the nucleofilament arrangements in O. marina chromosomes using freeze-substitution fixation.

The core dinoflagellate Heterocapsa circularisquama, which possesses a typical dinokaryon, was used for comparative analysis of chromosome structure. With freeze-substitution fixation, chromosomes in the 80-nm sections of both H. circularisquama and O. marina appeared as electron-dense images with no discernible internal structure (data not shown), suggesting the possibility of a loss of chromosomal constituents, particularly with conventional chemical fixation methods. To visualize the internal structures in the highly-dense chromosomes of the dinokaryon of H. circularisquama, we applied tomography analysis, which allowed us to find the arch-shaped structure within the chromosomes ( Figure 6C ; Supplementary Movie S4A ). The visualized nucleofilament arcs were similar to those by conventional chemical fixation methods but were more tightly aligned (compare Figures 6A, C ). In O. marina, however, no periodic filamentous structure appeared ( Figure 6D ; Supplementary Movie S4B ), suggesting that the nucleofilaments of O. marina are either tightly packed without periodic structure or so densely folded that individual filaments are indistinguishable.

Oxyrrhis marina was obtained from a publicly available algal culture stock center (the Culture Collection of Algae at Göttingen University; international acronym SAG), Strain No. 21.89. It was maintained in natural seawater sterilized by filtration and fed on the green alga Dunaliella tertiolecta (SAG Strain No. 13.86), which was grown in a modified f/2 medium under a 12:12 h light-dark (LD) cycle at 25°C (Fukuda and Endoh, 2006). The O. marina cells were also maintained under 12:12 LD at 25°C.

Heterocapsa circularisquama was sourced from the National Research Institute of Fisheries and Environment, Hiroshima, Japan, and cultured in a modified SWM3 medium under a 12:12 LD cycle (Kang et al., 2015).

To visualize nuclear morphology, O. marina cells were fixed with 1% glutaraldehyde for 30 min and washed several times with filtered natural seawater. Cells were then resuspended in filtered natural seawater containing 1 μg/mL DAPI (Nacalai-Tesque, Kyoto, Japan), and mounted on a glass slide under a coverslip. Microscopic observation was performed with a confocal laser scanning microscope (Olympus FV-1000; Evident Corporation, Tokyo, Japan), equipped with a x100 objective lens.

Cells were prefixed with 2% glutaraldehyde in filtered natural seawater. Cell suspension was mixed 1:1 with double-strength fixative solution at room temperature, and then the suspension was left at 4°C for 60 min. Fixed cells were washed three times with cooled filtered natural seawater and then transferred to 0.5% OsO₄ in filtered natural seawater for post-fixation at 4°C for 30 min. The fixed cells were then washed several times with filtered water, to eliminate fixative and salts, dehydrated through a graded ethanol series. The dehydrated sample was subjected to centrifugation, after which half of the supernatant was discarded, and an equal amount of Spurr resin (Polysciences Inc., Warrington, PA, USA) was added, followed by mixing. This process was repeated twice. Subsequently, the entire supernatant was replaced with Spurr resin and mixed again, repeating this process twice (30 minutes each). The sample embedded in resin was transferred to the Beem^® capsule, and the cells were concentrated at the bottom of the capsule through centrifugation. Finally, the capsulated sample was placed in a pre-heated polymerization oven and polymerized at 70°C for 8 hours. Ultrathin sections (approximately 80 nm thick) were prepared using an ultramicrotome (EM UC7; Leica Microsystems GmbH, Wetzlar, Germany), collected onto Formvar-coated grids (Cu, one slot), stained with 3% uranyl acetate and lead citrate, and examined using a H-7100 transmission electron microscope (Hitachi High-Tech, Tokyo, Japan) operating at 75 kV.

Cells were collected by low-speed centrifugation, and suspensions at high cell densities were cryofixed by slamming them onto a liquid nitrogen-cooled (-196°C) copper block using a metal-contact quick-freezing device (VFZ-101, Japan Vacuum Device Inc., Mito, Ibaraki, Japan). The frozen materials were transferred to cold (-80°C) acetone containing 1% OsO_4, and substitution was performed by incubation at -80°C for 72 h. The temperature was then manually elevated stepwise (-20°C for 1 h, 4–10°C for 0.5 h, and then room temperature for 0.5 h). The materials were washed with 100% acetone at room temperature and embedded in Spurr’s resin.

Two methods were used to obtain consecutive image series to create three-dimensional (3D) models. One was to collect consecutive sections and then take sequential images with a TEM; and the other was to automatically acquire sequential images using serial block face-scanning electron microscopy (SBF-SEM).

For the first method, consecutive ultrathin sections (approximately 100 nm thick) were prepared using an ultramicrotome and 56 consecutive sections were collected onto Formvar-coated grids (Cu, one slot) and stained with 3% uranyl acetate and lead citrate. They were then examined using TEM operating at 75 kV. The images taken at a nominal magnification of 17k were recorded in a 1k × 1k CCD camera (C4741-95; Hamamatsu Photonics, Hamamatsu, Japan) at a pixel size of 5.43 nm.

For SBF-SEM observation, the resin block containing the samples was trimmed and glued onto an aluminum rivet with conductive epoxy resin (SPI Conductive Silver Epoxy; Structure Probe Inc., West Chester, PA, USA) and coated with gold using an ion coater. A SBF-SEM system (Sigma VP, Carl Zeiss Microscopy, Jena, Germany) equipped with a back-scattered electron detector (3View; Gatan Inc., Pleasanton, CA, USA) was used to slice and image the specimen block surface. The SEM was operated at a low accelerating voltage of 1.0 kV to reduce charging. Consecutive image series were automatically acquired using Gatan Digital Micrograph software. All images were taken at an image size of 8192 × 8192 pixels (pixel size = 3 nm). Once an SEM image had been acquired, a 50-nm-thick layer was removed from the block face by the diamond knife, and the next freshly exposed surface was imaged by SEM in the same manner. This image acquisition cycle was repeated until images of a complete dinokaryon had been obtained. Two dinokaryon samples were obtained, one with 127 images and the other with 145 images. The consecutive images were aligned using the IMOD software package (Kremer et al., 1996). Segmentation in the 3D reconstruction was performed with Amira version 5.4.5 (FEI Visualization Science Group, Burlington, MA, USA).

Specimens of O. marina and H. circularisquama prepared by the freeze-substitution fixation method were also subjected to electron tomography. A 80 nm-thick section on a single-slot grid was stained with 3% uranyl acetate and lead citrate. A 200 kV electron microscope (JEM-2100F; JEOL, Tokyo, Japan) was used for data acquisition. Tilt series were recorded with a Gatan Tridium imaging filter with a 2k × 2k CCD camera at 2° increments in a tilt range from -60° to +60°. Automated data acquisition for electron tomography was performed using the Recorder module in the TEMography suite (System in Frontier Inc., Tokyo, Japan). After image alignment, the 3D reconstruction was performed by weighted back-projection using IMOD (Kremer et al., 1996).

Sample preparation methods and imaging conditions are listed in a Supplementary Table S1 .