3D cellular architecture of choanoflagellates

Three randomly selected single cells and three randomly selected colonial cells from a single colony were chosen for the reconstruction of entire choanoflagellate cells and subcellular structures (Fig 1, S1–S3 Figs, S1–S6 Movies). Both single and colonial S. rosetta cells exhibited a prominent, central nucleus enveloped by a mitochondrial reticulum and basal food vacuoles—as well as intracellular glycogen reserves—consistent with the coarse choanoflagellate cellular architecture reported in previous studies [20,21] (reviewed in [13,22]) (Fig 1, S1–S3 Figs, S1–S6 Movies). However, with the increased resolution of electron microscopy, we detected three morphologically distinct populations of intracellular vesicles with distinct subcellular localisations (Fig 1G, S1I–S1L Fig): 1) large vesicles (extremely electron-lucent, 226 ± 53 nm in diameter) (S1J, S1J’, and S1J” Fig); 2) Golgi-associated vesicles (electron-dense inclusions, 50 ± 10 nm in diameter) (S1I, S1I’, and S1I” Fig); and 3) apical vesicles (electron-lucent, 103 ± 21 nm in diameter) (S1K, S1K’ and S1K” Fig). Extracellular vesicles were also observed to be associated with two of the single cells (electron-lucent, 173 ± 36 nm in diameter) and appeared to bud from the microvillar membrane (S1L, S1L’ and S1L” Fig). Choanoflagellate cells subjected to fluorescent labelling were congruent with 3D ssTEM reconstructions in terms of organelle localisation (Fig 1B and 1C), providing evidence that the 3D models presented herein are biologically representative.

Ultrastructural comparison between single and colonial choanoflagellate cells reveals surprising differences

Our 3D ssTEM reconstructions allowed for detailed volumetric and numerical comparisons among single and colonial S. rosetta cells (Fig 2, S2 Fig, S1 and S2 Tables). Overall, the general deposition of major organelles was unchanged in both cell types (Fig 1E–1L, Fig 2A and 2B, S2A–S2C Fig). In addition, single and colonial cells devote a similar proportion of cell volume to most of their major organelles (nucleus: single cells 12.92% ± 0.58% versus colonial cells 11.56% ± 0.27%; nucleolus: 1.85% ± 0.33% versus 2.2% ± 0.22%; mitochondria: 5.08% ± 1.14% versus 6.63% ± 0.42%; food vacuoles: 9.22% ± 2.75% versus 6.85% ± 0.87%; and glycogen storage: 8.71% ± 2.36% versus 7.50% ± 1.12%) (Fig 2, S2 Fig, S1 and S2 Tables).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. 3D ssTEM reconstructions allow for volumetric and numerical comparison of high-resolution single and colonial S. rosetta cells.

Shown are the mean volumetric breakdowns of three single (A) and three colonial (B) S. rosetta cells (left) and a generalised diagram of cell type ultrastructure (right). Colours are as in Fig 1. (C) Volumetric (%) (±SEM) (ER and endocytotic vacuoles) and numerical (μm⁻³) (±SEM) (endocytotic vacuoles, pseudopodia, Golgi-associated vesicles, and mitochondria) differences were found between single and colonial (n = 3) S. rosetta cells. *P < 0.05, **P < 0.01, ***P < 0.001. ER, endoplasmic reticulum; ssTEM, serial ultrathin transmission electron microscopy.

https://doi.org/10.1371/journal.pbio.3000226.g002

We did, however, uncover some interesting ultrastructural differences between single and colonial cells (Fig 2C). Colonial cells devoted a higher proportion of cell volume to endoplasmic reticulum (ER) (single: 3.27% ± 0.35% versus colonial: 6.86% ± 0.39%). This contrast was coupled to a differential ER morphology across cell types. The ER of colonial cells frequently displayed wide, flat sheets (Fig 3E), which were not observed in the reconstructed single cells. Single cells exhibited a higher number of Golgi-associated vesicles (single: 166.3 ± 32.7 versus colonial: 72.3 ± 26.5) and individual mitochondria than colonial cells (single: 25.3 ± 5.8 versus colonial: 4.3 ± 4.2) (Fig 2C, S2 Table) despite lacking volumetric differences between cell types.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Reconstructions of complete choanoflagellate RCs places colonial cells into context and unveils ultrastructural features involved in rosette formation and a novel cell type.

(A–D) 3D ssTEM reconstruction of a complete RC1. The plasma membrane was made transparent (B) to allow better visualisation of subcellular structures. Highlighted are contacting FP (C) and IBs (D). Cellular structures coloured as in Fig 1. Scale bar = approximately 1 μm. (E–L) 2D TEM and 3D ssTEM reconstructions of structures (*) differentially exhibited by colonial cells or involved in colony formation. Shown are the ER (E, F), IBs (G, H), EV (I, J), and FP (K, L). Scale bars = 200 nm. (M–P) Reconstruction of multiple S. rosetta colonies shows no strong pattern of volumetric distribution and bridge networks but reveals the presence of highly derived cell morphologies. (M) 3D ssTEM reconstructions of five complete rosettes (RC1–5) coloured by cell number (above), and 2D projections of bridge connections in 3D ssTEM reconstructions of RCs (below). Disconnected IBs marked by white arrowheads and lines. Asterisks mark the presence of highly derived cell morphologies in RC3 and RC4. Cells in RCs are numbered in order of their appearance along the z-axis. (N) Volumetric distribution of mean cell volumes (RC1–5) in RCs reveals no apparent pattern of cell distribution across the z-axis. (O, P). Two highly derived cell types, the ‘carrot cell’ (O) from RC3 and the ‘chili cell’ (P) from RC4, were identified in RCs. Colours as in Fig 1. Scalebar = approximately 1 μm. (Q–U) IBs in colonial S. rosetta exhibit a high diversity of morphologies, suggestive of disconnection. In addition to prior descriptions of IBs (arrowheads) and electron-dense septa (asterisks), bridges in colonial S. rosetta often display an asymmetrically distributed septum (Q), protracted and elongated morphology (R), disconnection from one of the contiguous cells (S), evidence of abscission (T), and putative inheritance of the septum (U). Scale bar = 200 nm. ER, endoplasmic reticulum; EV, endocytotic vacuoles; FP, filopodia; IB, intercellular bridge; RC, rosette colony; ssTEM, serial ultrathin transmission electron microscopy.

https://doi.org/10.1371/journal.pbio.3000226.g003

Finally, we found that colonial cells are characterised by a more amoeboid morphology than single cells (Fig 3A). Colonial cells exhibited a higher relative proportion of endocytotic vacuoles by volume (single: 0.07 ± 0.07 versus colonial: 0.32 ± 0.12)—a phenomenon coupled to a higher overall number of endocytotic vacuoles (single: 1 ± 1 versus colonial: 5 ± 2) and pseudopodial projections per cell (single: 1 ± 1 versus colonial 8 ± 2) (Fig 2C, S1 and S2 Tables). Many of the pseudopodial projections and endocytotic vacuoles bore the morphology of lamellipod ruffles and macropinosomes (Fig 3A), suggesting that colonial cells are typified by high macropinocytotic activity.

Reconstruction of multiple rosettes reveals colony-wide cell arrangement, different cell shapes, and complete cell–cell contact networks

While high-magnification 3D ssTEM enabled the high-resolution reconstruction of individual colonial cells, their context and interactions with neighbouring cells were lost. To address this, we reconstructed the subcellular structures of a seven-cell rosette colony (complete rosette, RC1) from 80-nm sections taken at lower magnification (Fig 3A–3D, S7 Movie) as well as the gross morphology of four larger rosettes (RC2–5) from 150-nm sections to provide a more representative survey (Fig 3E–3P). We found that individual cells in rosette colonies vary widely in volume (Fig 3M and 3N), although no pattern was detected in the volumetric cellular arrangement along the rosette z-axis (Fig 3M). In addition, mean cell size was comparable among different rosettes, including those that contained different numbers of cells (S4B Fig). However, we did find a positive correlation between cell number and the number of intercellular bridges per cell across rosette colonies (S4B Fig).

Importantly, we uncovered the presence of unusually shaped cells in two of the five S. rosetta rosette colonies (carrot-shaped cell 5 in RC3 and chili-shaped cell 5 in RC4, both labelled orange with an asterisk) (Fig 3M). These unusual cells were both found at the same location along the rosette z-axis, exhibited an elongated morphology distinct from other colonial cells (Fig 3O and 3P and S8 and S9 Movies), and were small in volume. Cells 5 from RC3 and RC4 were 9.87 μm³ and 13.35 μm³, respectively (Fig 3N)—the mean volume of the cells in RC3 and RC4 was 27.38 μm³ and 27.25 μm³, respectively (Fig 3N). While each of these unusual cells possessed a flagellum, a collar, connections to neighbouring cells via intercellular bridges, and had a similar proportion of cell volume dedicated to most of their major organelles as observed in other colonial cells, these cells devoted a larger volumetric percentage of the cell body to the nucleus (29.8% and 30.78%, respectively, versus the mean colonial proportion of 13.76% ± 0.49%).

Our 3D ssTEM reconstructions of rosette colonies also revealed the distribution of intercellular bridges and the connections formed between individual cells (Fig 3M). We found intercellular bridges in all analysed rosette colonies (RC1–5), totalling 36 bridges. There was no detectable pattern regarding bridge networking across rosette colonies. Bridges were distributed from the cell equator to either of the poles along the cellular z-axis, and the average bridge was 0.75 ± 0.38 μm in length (S4C Fig). Prior studies [12,17] of S. rosetta bridges suggested that bridges are typically short (0.15 μm), connecting two adjacent cells and containing parallel plates of electron-dense material. In contrast, the bridges detected in this study exhibited striking morphological diversity (Fig 3M, 3Q–3U), with lengths ranging from 0.21–1.72 μm. The majority of bridges consisted of a protracted cytoplasmic connection between two cells, and in many cases, the septum was localised asymmetrically along the bridge (S4C Fig). Most surprisingly, some bridges were not connected to any neighbouring cells at all, but rather the septum was situated on the end of a thin, elongated cellular protrusion (Fig 3S). In addition, we observed asymmetric bridge width and degraded electron-dense structures proximal to bridge remnants being incorporated into the cell body of a contiguous cell (Fig 3T and 3U). These data suggest that intercellular bridges could be disconnected from neighbouring cells and that the electron-dense septum may be inherited.

Comparison of 3D cellular architecture between choanoflagellates and sponge choanocytes

Both choanoflagellates and sponge collar cells influence local hydrodynamics by beating their single flagellum to draw in bacteria that are captured by the apical collar complex [23], however sponge choanocytes are part of an obligately multicellular organism (Fig 4A). Sponge choanocytes therefore provide an excellent representative of an early-branching animal collar cell against which to compare choanoflagellate cell architectures. Our 3D ssTEM reconstructions allowed for the reconstruction of five choanocytes and for the volumetric and numerical comparison of choanocyte and choanoflagellate subcellular structures (Fig 4B–4E, S5 and S6 Figs, S10 Movie). We detected little ultrastructural variability between the five choanocytes (S5 Fig, S3 and S4 Tables). All five cells exhibited a prominent basal nucleus, small and unreticulated mitochondria, food vacuoles scattered around the entire cell, and an apical Golgi apparatus (Fig 4B–4D, S5 and S6 Figs)—consistent with the coarse choanocyte cellular architecture reported in previous studies [23,24] (reviewed in [1,25]).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. 3D cellular architecture of sponge choanocytes.

(A) Choanocytes line interconnected chambers in members of the Porifera and serve as feeding cells. (B) Mean volumetric breakdown of five sponge choanocytes. Colours are as in Fig 1. (C–E) 3D ssTEM of a section of choanocyte chamber containing five complete cells (B). The plasma membrane was rendered transparent (D), and food vacuoles and ER were removed to allow better visualisation of subcellular structures (E). Colours are as in Fig 1. Scale bar = approximately 1 μm. (F–G) Reconstruction and comparison of the sponge choanocyte (F) and choanoflagellate (G) apical poles shows distinct differences between the two cell types. Shown in the choanocyte reconstruction are the basal foot (red, associated with basal body), food vacuole (light green), ER (yellow), flagellar basal body (light blue), flagellum (dark green), Golgi apparatus and Golgi-associated vesicles (purple), microtubules (grey), mitochondria (red), nonflagellar basal body (dark orange), Type 1 vesicles (light orange), and Type 2 vesicles. Shown in the choanoflagellate reconstruction are the apical vesicles (pink), food vacuole (light green), ER (yellow), flagellar basal body (light blue), flagellum (dark green), Golgi apparatus and Golgi-associated vesicles (purple), glycogen (white), large vesicles (brown), microtubules (grey), microtubular ring (red), and nonflagellar basal body (dark orange). Scale bars = 200 nm. Diagrams of the choanocyte fine kinetid (F) and choanoflagellate fine kinetid (G) structure highlight the distinct differences. ER, endoplasmic reticulum; ssTEM, serial ultrathin transmission electron microscopy.

https://doi.org/10.1371/journal.pbio.3000226.g004

Furthermore, our data showed many ultrastructural commonalities between sponge choanocytes and choanoflagellates. For example, the number of microvilli that surround the apical flagellum in single and colonial choanoflagellates is comparable to the number of microvilli in sponge choanocytes (single: 32 ± 2 versus colonial: 35.3 ± 4.9 versus choanocytes: 30.6 ± 4.1) (S6A Fig). We also found that the number of food vacuoles and the number and volumetric proportion of the Golgi apparatus are similar in all three cell types (S6A Fig). Although sponge choanocytes did not appear to exhibit the same macropinocytotic activity as colonial choanoflagellates throughout the cell (some micropinocytotic inclusions are present toward the cell apex [S6D and S6E Fig]), basal sections of choanocytes were heavily amoeboid (S6B and S6C Fig). These amoeboid protrusions may not only be for mechanical anchorage into the mesohyl but may play a role in phagocytosis, as we observed bacteria in the mesohyl to be engulfed by basal pseudopodia (S6F and S6G Fig). Thus, both choanocytes and colonial choanoflagellates are typified by high-amoeboid cell activity.

We also observed some ultrastructural differences between choanocytes and choanoflagellates. In contrast with cells from choanoflagellate rosettes, sponge choanocytes lack filopodia and intercellular bridges. Choanocytes also do not possess glycogen reserves and devote significantly less of their cell volume (9.25% ± 0.39%) than choanoflagellates (single: 12.92% ± 0.58% and colonial: 11.56% ± 0.27%) to the nucleus and less to mitochondria (2.5% ± 0.3% versus single: 5.08% ± 1.14% and colonial: 6.63% ± 0.42%) (S6A Fig). However, choanocytes devote significantly more of their volume to food vacuoles (20.7% ± 1.01%) than choanoflagellates (single: 9.22% ± 2.75% and colonial: 6.85% ± 0.87%) (Fig 4E). High-resolution reconstructions of the choanocyte and choanoflagellate apical pole (Fig 4F and 4G, S11 and S12 Movies) showed differences in terms of vesicle type and localisation, Golgi positioning, and collar arrangement (conical in choanoflagellates while cylindrical in choanocytes, as previously noted [23]). The flagellar basal body has previously been meticulously characterised in both choanocytes and choanoflagellates, and some differences have been reported between the two by other authors [26–31]. These findings are reiterated by our reconstructions and observations (Fig 4F and 4G).

Concluding remarks

While we recognise the limitations of our findings due to the morphological descriptive nature of this study and the small sample size, the comparative 3D reconstruction of collar cells from two different phyla, choanoflagellates and sponges, allowed for an unbiased view of their cellular architecture and for the reconstruction of key properties of the enigmatic ancestral collar cell. Our data reveal distinct ultrastructural features in single and colonial choanoflagellates and demonstrate that cells within rosette colonies vary significantly in their cell size and shape. The newly identified ‘carrot’ and ‘chili’ cells reveal that cells within choanoflagellate colonies do not simply consist of an assemblage of equivalent single cells, but some may represent a distinctly differentiated cell type displaying ultrastructural modifications. Likewise, our data suggest that sponge choanocytes are not simply an incremental variation on the choanoflagellate cell plan but are specialised feeding cells, as indicated by their high volumetric proportion of food vacuoles. Together, our data show a remarkable variety of collar cell architecture and suggest cell type differentiation may have been present in the stem lineage leading to the animals.

Cell culture

Colony-free S. rosetta cultures (ATCC 50818) were grown with coisolated prey bacteria in 0.22 μm filtered choanoflagellate growth medium [66] diluted at a ratio of 1:4 with autoclaved seawater. Cultures were maintained at 18°C and split 1.5:10 once a week. Colony-enriched S. rosetta cultures (PX1) were likewise maintained but monoxenically cultured with the prey bacterium A. machipongonensis [67] to induce rosette formation.

Fluorescent labelling of organelles

To support the annotation of organelles from ssTEM sections, the microanatomy of S. rosetta cells was chemically characterised by fluorescent vital staining. Cells were pelleted by gentle centrifugation (500x g for 10 min at 4°C) in a Heraeus Megafuge 40R (ThermoFisher Scientific) and resuspended in a small volume of culture medium. Concentrated cell suspension (500 μl) was applied to glass-bottom dishes, coated with poly-L-lysine solution (P8920, Sigma-Aldrich), and left for 10–30 min until cells were sufficiently adhered. PX1 cultures were concentrated into 100 μl of culture medium to promote the adherence of rosette colonies.

Adhered cells were incubated in 500 μl of fluorescent vital dye diluted in 0.22 μm filtered seawater. Cells were incubated with 4.9 μM Hoechst 33342 Dye for 30 min (to label nuclei), 1 μM LysoTracker Yellow HCK-123 for 1.5 h (to label food vacuoles), and 250 nM MitoTracker Red CM-H2Xros for 30 min (to label mitochondria). All vital dyes were from ThermoFisher Scientific (H3570, L12491, T35356, and M7513, respectively). Fluorescent-DIC microscopy was conducted under a 100x oil-immersion objective lens using a Leica DMi8 epifluorescent microscope (Leica, Germany). Vital dyes were viewed by excitation at 395 nm and emission at 435–485 nm (Hoechst 33342 Dye), 470 nm and emission at 500–550 nm (LysoTracker Yellow HCK-123 and FM 1–43 Dye), and 575 nm and 575–615 nm (MitoTracker Red CM-H2Xros). Micrographs were recorded with an ORCA-Flash4.0 digital camera (Hamamatsu Photonics, Japan). All cells were imaged live. No-dye controls using only the dye solvent dimethyl sulfoxide (DMSO) (D4540, Sigma-Aldrich) were run for each wavelength to identify and control for levels of background fluorescence. Chemical fixation during vital staining and TEM sectioning was avoided where possible in this study to reduce fixation artefacts.

To visualise cell bodies, flagella, filopodia, and collar-adherent cells were fixed for 5 min with 1 ml 6% acetone and for 15 min with 1 ml 4% formaldehyde. Acetone and formaldehyde were diluted in artificial seawater, pH 8.0. Cells were washed gently four times with 1 ml washing buffer (100 mM PIPES at pH 6.9, 1 mM EGTA, and 0.1 mM MgSO₄) and incubated for 30 min in 1 ml blocking buffer (washing buffer with 1% BSA and 0.3% Triton X-100). Cells were incubated with primary antibodies against tubulin (E7, 1:400; Developmental Studies Hybridoma Bank), diluted in 0.15 ml blocking buffer for 1 h, washed four times with 1 ml of blocking buffer, and incubated for 1 h in the dark with fluorescent secondary antibodies (1:100 in blocking buffer, Alexa Fluor 488 goat anti-mouse). Coverslips were washed three times with washing buffer, incubated with Alexa Fluor 568 Phalloidin for 15 min, and washed again three times with washing buffer. Coverslips were mounted onto slides with Fluorescent Mounting Media (4 ml; Prolong Gold Antifade with DAPI, Invitrogen). Images were taken with a 100x oil-immersion objective on a Leica DMI6000 B inverted compound microscope and Leica DFC350 FX camera. Images presented as z-stack maximum intensity projections.

Electron microscopy

3D reconstruction and analysis

ssTEM sections were imported as z-stacks into the Fiji [71] plugin TrakEM2 [72] and automatically aligned using default parameters, except for increasing steps per octave scale to 5 and reducing maximal alignment error to 50 px. Alignments were manually curated and adjusted if deemed unsatisfactory. Organelles and subcellular compartments were manually segmented and 3D reconstructed by automatically merging traced features along the z-axis. Meshes were then preliminarily smoothed in TrakEM2 and exported into the open-source 3D software Blender 2.77 [73]. Heavy smoothing of the cell body in TrakEM2 sacrifices fine structures associated with cellular projections or does not remove all distinct z-layers, which exist as reconstruction artefacts. Therefore, cell bodies were manually smoothed using the F Smooth Sculpt Tool in Blender of final distinct z-layers for presentation purposes only (S3 Fig). All organelles were subjected to the same smoothing parameters across individual cells. All analysis was conducted using unsmoothed, unprocessed meshes. Organelle volumes were automatically quantified by the TrakEM2 software and enumerated in Blender 2.77 by separating meshes in their total loose parts.

The microvillar collar and flagellum were excluded from volumetric analysis, as their total, representative length could not be imaged at this magnification. Cytosolic volume was calculated by subtracting total organelle volume from cell body volume and is inclusive of cytosol, ribosomes, and unresolved smaller structures excluded from 3D reconstruction. Endocytotic vacuoles were distinguished from food vacuoles by connection to the extracellular medium in ssTEMs or by localisation to a cell protrusion. Cells in rosette colonies are numbered in order of their appearance along the image stack z-axis. Rosette colony diameters were calculated by measuring the largest distance of the z-axis midsection. Bridge length was measured in one dimension along the bridge midsection. Mean vesicle diameters were calculated from 20 measurements (or as many as possible if the vesicle type was rare) from single cells.

Data analysis

Univariate differences in the volume and number of subcellular structures between the two cell types were evaluated using two-sample t tests. Shapiro–Wilk and Levene’s tests were used to assess normality and homogeneity of variance, respectively. Statistical comparisons were conducted using data scaled against total cell volume. Correlations between colony cell number, cell volume, and bridges per cell were assessed using Pearson correlation tests. All statistical analyses were conducted using R v 3.3.1 [74] implemented in RStudio v 0.99.903 [75].